RIM-BP2 knockout mice (RIM-BP2^−/−) and their wild-type littermates RIM-BP2 (RIM-BP2^+/+) (Grauel et al., 2016) of either sex were used for experiments. For immunofluorescence microscopy, electron tomography, patch-clamp, and confocal Ca²⁺-imaging all mice were examined at the age of 2–3 weeks (i.e., after hearing-onset). Systems physiology and extracellular recordings from single SGNs were carried out at the age of 8–10 weeks. All experiments complied with national animal care guidelines and were approved by the University of Göttingen board for animal welfare and the animal welfare office of the state of Lower Saxony.

Freshly dissected apical turns of the organs of Corti of 3 week old RIM-BP2^+/+ and RIM-BP2^−/− mice were either fixed with 4% formaldehyde (FA) in phosphate buffered saline (PBS) on ice for 10 min or with methanol at −20°C for 2 min. Thereafter, the tissue was washed 3 × 10 min in PBS and incubated for 1 h in goat serum dilution buffer (GSDB: 16% normal goat serum, 450 mM NaCl, 0.3% Triton X-100, 20 mM phosphate buffer, pH 7.4) in a wet chamber at room temperature. Primary antibodies were dissolved in GSDB and applied overnight at 4°C in a wet chamber. After washing 3 × 10 min (wash buffer: 450 mM NaCl, 20 mM phosphate buffer, 0.3% Triton X-100) the tissue was incubated with secondary antibodies in GSDB in a wet light-protected chamber for 1 h at room temperature. Then, the samples were washed 3 × 10 min in wash buffer and 1 × 10 min in 5 mM phosphate buffer, placed onto the glass microscopy slides with a drop of fluorescence mounting medium (Mowiol) and covered with thin glass coverslips. The following antibodies were used: mouse-IgG1-anti-CtBP2 (also recognizing the ribbon protein RIBEYE, BD Biosciences, 1:200), guinea pig-anti-Synapsin1/2 (Synaptic Systems, 1:500), rabbit-anti-Ca_V1.3 (Alomone Labs, 1:50), guinea pig-anti-bassoon (Synaptic Systems, 1:500), rabbit-anti-GluA2/3 (Chemicon, 1:200), rabbit-anti-RIM-BP2 (Synaptic Systems, 1:200) as well as secondary AlexaFluor568- and AlexaFluor647-labeled goat-anti-mouse and goat-anti-rabbit antibodies (Invitrogen, 1:200) and STAR580-, and Star635P-labeled goat-anti-mouse, goat-anti-rabbit and goat-anti-guinea pig antibodies (Abberior, 1:50). Confocal images were acquired using a laser scanning confocal microscope (Leica TCS SP5, Leica Microsystems GmbH, Mannheim, Germany) equipped with 561 and 633 nm lasers for excitation and a 63x oil immersion objective (1.4 NA, Leica). STED images were acquired using a 2-color STED microscope (Abberior Instruments, Göttingen, Germany) equipped with 561 and 640 nm excitation lasers, a 775 nm laser for STED (1.2 W) and a 100x oil immersion objective (1.4 NA, Olympus). Samples were treated in parallel and images were acquired in parallel, using same laser power, gain, and microscope settings. Images were analyzed using ImageJ, Igor Pro, OriginPro and Imaris software, and assembled for display in Adobe Illustrator software.

Embedding of apical turns of the organ of Corti of 3-week-old RIM-BP^+/+ and RIM-BP^−/− mice was performed as described previously (Wong et al., 2014). Following conventional embedding and trimming of the blocks, 250 nm sections were obtained for electron tomography with an Ultracut E microtome (Reichert-Jung). Sections were poststained for 40 min using uranyl acetate-replacement solution and subsequently 30 s with lead citrate. Electron tomography, tomogram generation and rendering were essentially performed as described previously (Strenzke et al., 2016) and references therein. Tomograms were generated using the IMOD software program etomo. Rendering of 3-dimensional (3D) models and quantitative image analysis was performed blinded using 3dmod. The size of synaptic ribbons and presynaptic densities (PD) was determined from the surface area of the reconstructed 3D objects. For “ribbon-associated” vesicles (RA-SVs) the first row of vesicles around the synaptic ribbon within a distance of 80 nm was counted. For “membrane-proximal” vesicles (MP-SVs) the vesicles with a maximum lateral distance (membrane to membrane) of 100 nm from the PD and ≤ 50 nm from the presynaptic plasma membrane were counted (Jung et al., 2015).

IHCs from apical coils of freshly dissected organs of Corti of 2-week-old RIM-BP2^+/+ and RIM-BP2^−/− mice were patch-clamped at room temperature (22–24°C) as described (Moser and Beutner, 2000). For perforated-patch-clamp recordings of whole-cell Ca²⁺-current and exocytosis the pipette solution contained (in mM): 130 Cs-gluconate, 10 TEA-Cl, 10 4-AP, 10 HEPES, 1 MgCl₂, amphotericin B (300 μg/ml), pH 7.2. For confocal Ca²⁺-imaging we performed ruptured-patch-clamp recordings with the solution described above excluding amphotericin but including 4 mM Mg-ATP, 0.3 mM Na-GTP, 1 mM EGTA, 0.4 mM Fluo-4FF (Invitrogen), and 0.01 mM carboxytetramethylrhodamine(TAMRA)-conjugated dimeric RIBEYE-binding peptide (Francis et al., 2011). The extracellular solution contained (in mM): 110 NaCl, 35 TEA-Cl, 2.8 KCl, 2 CaCl₂, 1 MgCl₂, 10 NaOH-HEPES, 11.3 D-glucose, pH 7.3. EPC-10 amplifiers (HEKA Elektronik, Lambrecht, Germany) controlled by “Patchmaster” software (HEKA) were used for measurements. All voltages were corrected for liquid junction potentials (14 mV). Currents were leak-corrected using a p/10 protocol. Current-voltage relationships (IVs) were calculated from the last 8 ms of currents evoked by step depolarizations to various potentials. From these, fractional activation curves were calculated by first calculating the conductance of the calcium current as G(V)=ICa(V-Vrev), where V is the command potential and V_rev is the reversal potential of the current obtained by extrapolating a line fit from 6 to 26 mV. After normalizing these traces to the maximum conductance in the range of −20 to +10 mV, they were fit with a Boltzmann equation Gn(V)=Gn, max1+eVhalf-Vk with G_n,max, as the maximum conductance, V as the command potential, V_half as the voltage of half-maximal activation, and k as the slope factor. I_Ca inactivation kinetics were calculated from fitting a single exponential function to the last 195 ms of I_Ca traces during 200 ms depolarizations to −14 mV (corrected for liquid junction potential). For membrane capacitance (C_m) measurements, IHCs were stimulated by depolarizations to −14 mV with intervals of 60–120 s to allow for recovery of IHC exocytosis. Two pulse protocols were used: (i) single depolarizations of different durations and (ii) paired depolarizations with variable inter-pulse interval. We measured C_m changes (ΔC_m) using the Lindau-Neher technique (Lindau and Neher, 1988) as previously described (Moser and Beutner, 2000). Briefly, the exocytic ΔC_m was quantified as the difference of the averaged C_m 400 ms before and after the depolarization. To avoid impact of C_m-transients related to conductance or gating of ion channels on ΔC_m estimation (Moser and Beutner, 2000; Neef et al., 2007) we skipped the first 100 ms (single depolarizations) or 25 ms (paired depolarizations) of post-depolarization C_m for estimating the average. Mean ΔC_m and Ca²⁺-current estimates present grand averages calculated from the mean estimates of individual IHCs, where each depolarization was repeated 2–3 times. This avoided dominance of IHCs contributing more sweeps.

The apparent Ca²⁺-cooperativity m of exocytosis was addressed via gradual reduction of the number of open Ca²⁺-channels by slow wash-in (0.5 ml/min) of the dihydropyridine L-type Ca²⁺-channel antagonist isradipine (10 μM) in the extracellular solution (here using 5 mM CaCl₂ and 104 mM NaCl). The exocytic ΔC_m and corresponding I_Ca evoked by 20 ms depolarizations (to −14 mV, corrected for liquid junction potential) were recorded while slowly reducing the number of unblocked Ca²⁺-channels and ΔC_m to the Ca²⁺-current integral (Q_Ca) for each individual IHC and fit by a power function ΔC_m = A (Q_Ca)^m, where the power m describes the apparent Ca²⁺-cooperativity of synaptic vesicle exocytosis.

For confocal Ca²⁺-imaging we used the low affinity Ca²⁺-indicator Fluo-4FF (K_D = 10 μM, Invitrogen) and also loaded the IHCs with TAMRA-labeled RIBEYE-binding peptide for visualization of the ribbons (Zenisek, 2008; Frank et al., 2009). Line scans (0.7 kHz) across the center of fluorescently labeled AZs were performed during IHC depolarization (20 ms to −7 mV, three repeats per AZ) to measure Fluo-4FF fluorescence changes as a consequence of synaptic Ca²⁺-influx (Frank et al., 2009). For quantification of line-scans, the Fluo-4FF ΔF-signal was calculated as the average evoked change in intensity from three peak-centered pixels that showed the maximal fluorescence intensity (F_max) over the last 10 ms of depolarization. ΔF was normalized to F₀, for which we used the fluorescence intensity of the same three peak-centered pixels before IHC depolarization.

ABR, DPOAE, and extracellular recordings from single SGNs were performed as described before (Jing et al., 2013; Strenzke et al., 2016) on 8–10 week-old mice. For extracellular recordings from individual SGNs, mice were anesthetized by i.p. injection of urethane (1.32 mg/kg), xylazine (5 mg/kg), and buprenorphine (0.1 mg/kg), a tracheostomy was performed, and the mice were then placed in a stereotactic system. After partial removal of the occipital bone and cerebellum to expose the anteroventral cochlear nucleus (AVCN), a glass microelectrode was advanced through the posterior AVCN portion to reach the auditory nerve where it enters the AVCN. Acoustic stimulation was provided by an open field Avisoft ScanSpeak Ultrasonic Speaker (Avisoft Bioacoustics), and “putative” auditory nerve fibers (formed by the central SGN axons) were identified and distinguished from cochlear nucleus neurons based on their stereotactic position (>1.1 mm from the surface of the cochlear nucleus), spontaneous and noise-burst induced firing, peristimulus time histogram (PSTH), regularity of firing, and first spike latency. Recordings were performed using TDT system III hardware and an ELC-03XS amplifier (NPI electronics).

Data analysis and statistical tests were done in Igor Pro for electrophysiology, Ca²⁺-imaging, and immunohistochemistry. Normality of distribution was tested with Jarque-Bera test and variances were compared with F-test. Unpaired, two-tailed Wilcoxon rank test (Mann-Whitney test) was used to compare non-normal or data with unequal variances. Else, unpaired, two-tailed Student's t-test were used. For comparing the variance of the ΔF/F₀ estimates for synaptic Ca²⁺-signals we used the Brown-Forsythe-modified Levene test. The distance of MP-SVs from the plasma membrane was compared between genotypes in the ranges of 0–25 and 26–50 nm by the Kolmogorov–Smirnov test. Data analysis for systems physiology was done using waveform-based spike detection using custom-written MATLAB software. Further statistical analysis was done in Igor Pro as described above for all parameters except spontaneous rate distributions, which were compared using Mann Whitney U-test in GraphPad Prism.

First, we studied the localization of RIM-BP2 at IHC AZs using immunolabeling and confocal microscopy of mouse IHCs. RIM-BP2 immunofluorescence at CtBP2/RIBEYE-marked ribbon-occupied AZs of RIM-BP2^+/+ IHCs and was absent from the basal pole of IHCs of RIM-BP2^−/− IHCs confirmed specificity of the immunolabeling (Figure 1A). In addition, some RIM-BP2 immunofluorescence spots were found near IHCs but away from the ribbons, most likely reflecting their presence in conventional presynaptic terminals formed by efferent lateral olivocochlear neurons onto the postsynaptic boutons of the afferent SGNs near the base of the IHCs (connectivity illustrated in Figure 1B). This notion was confirmed by triple antibody staining for RIM-BP2, CtBP2/RIBEYE and Synapsin1/2, marking the presynaptic terminals of efferent lateral olivocochlear neurons (Safieddine and Wenthold, 1999) (Figure 1B). RIM-BP2 immunofluorescence co-localized with either CtBP2/RIBEYE or Synapsin1/2, but was not visible outside the marked presynaptic regions, suggesting specific localization of RIM-BP2 to AZs of IHCs and presynaptic terminals of efferent lateral olivocochlear neurons. Using 2D 2-color STED microscopy we observed that RIM-BP2 immunofluorescence at the ribbon-type AZs was mostly arranged in single or double stripe-like shapes (Figure 1C). This pattern was similar for other CAZ proteins of the IHC AZ: bassoon and RIM2 both predominantly form stripe-like clusters at the base of the synaptic ribbon (Frank et al., 2010; Jung et al., 2015). Likewise, Ca_V1.3 Ca²⁺-channel clusters at IHC AZs primarily assume single or double-stripe like shapes (Frank et al., 2010; Wong et al., 2014). Thus, the spatial organization of RIM-BP2 seems to be very similar to that of the other CAZ proteins at the IHC AZ, where these proteins form the presynaptic density and co-align with the presynaptic Ca_V1.3 Ca²⁺-channels cluster. Unfortunately, co-immunostaining of RIM-BPs together with Ca_V1.3 Ca²⁺-channels was not possible, since the antibodies available to us were from the same species (see section Materials and Methods).

Using constitutive RIM-BP2 knockout mice (RIM-BP2^−/−), we first addressed the afferent connectivity of IHCs using confocal microscopy and 3D analysis of immunolabeled synapses (Khimich et al., 2005). RIBEYE/CtBP2 (synaptic ribbon) served as a presynaptic synapse marker and GluA2/3 (postsynaptic AMPA receptors) as a postsynaptic synapse marker (Figure 2A). We found the number of ribbons (RIBEYE/CtBP2 spots) (p = 0.9) and ribbon-occupied synapses (number of juxtaposed RIBEYE/CtBP2 and GluA2/3 spots) (p = 0.6) to be comparable in RIM-BP2^+/+ and RIM-BP2^−/− IHCs (Figure 2B), indicating that the afferent hair cell connectivity was unaltered.

Since RIM-BPs are known to interact with the pore-forming Ca_V1.3α subunit (Coppola et al., 2001; Hibino et al., 2002), we next studied the effect of RIM-BP2 disruption on the presynaptic Ca²⁺-influx using perforated patch-clamp recordings of voltage-gated Ca²⁺-influx in RIM-BP2^−/− IHCs. We observed a significant reduction of the maximal whole-cell I_Ca amplitude in RIM-BP2^−/− IHCs (~20%) (Figure 3A, Table 1). We then analyzed the voltage-dependence of I_Ca activation (voltage of half-maximal activation, V_half and slope factor k; Figure 3B) as well as the kinetics and extent of I_Ca inactivation (Figure 3C), which were unaltered in RIM-BP2^−/− IHCs (Table 1). While our data seem not to support a regulation of Ca_V1.3 gating by RIM-BP2 in IHCs, we cannot exclude a reduced open probability as cause of the diminished Ca²⁺-influx. However, we favor the simple interpretation of the reduction of Ca²⁺-influx in RIM-BP2-deficient IHCs that RIM-BP2 positively regulates the number of Ca²⁺-channels in IHCs.

Since the whole-cell recordings sum I_Ca of all AZs and the extrasynaptic membrane, we sought to further analyze Ca²⁺-influx at the single AZ level using laser scanning confocal Ca²⁺-imaging (Frank et al., 2009; Wong et al., 2014). Using ruptured patch-clamp, we loaded IHCs with a fluorescently labeled RIBEYE-binding peptide to visualize the location of synaptic ribbons (Zenisek et al., 2004; Ohn et al., 2016) and a low affinity Ca²⁺-indicator Fluo-4FF. Under the chosen conditions ([Ca²⁺]_e: 2 mM, Ca²⁺ buffering: 1 mM EGTA, K_D of Fluo-4FF for Ca²⁺: 10 μM) we expect the Fluo-4FF fluorescence changes (ΔF/F₀) to be a good proxy for the synaptic Ca²⁺-influx (Frank et al., 2009). Line scans across the center of fluorescently labeled AZs were performed during IHC depolarization (20 ms to −7 mV) to measure Fluo-4FF ΔF/F₀ as a consequence of synaptic Ca²⁺-influx (Figure 4A). We found a significant reduction (~41%, p = 0.01) of the Fluo-4FF maximal fluorescence change (ΔF_max/F₀) at AZs of RIM-BP2^−/− IHCs, which exceeded the reduction of their whole-cell Ca²⁺-current (~20%, p = 0.01) (Figure 4B). This suggests that RIM-BP2 disruption caused a preferential loss of synaptic Ca²⁺-channels. As previously described for wild-type IHCs, the maximal ΔF/F₀ estimates varied greatly among AZs within IHCs. This variance is likely explained by the heterogeneity of synapses regarding their AZ size and Ca²⁺-channel number (Frank et al., 2009; Wong et al., 2013; Ohn et al., 2016). We assessed the synapse heterogeneity by calculating the coefficient of variation (CV) of the ΔF_max/F₀ signals, which was comparable between RIM-BP2^−/− (CV = 0.86) and RIM-BP2^+/+ AZs (CV = 0.88). Hence, despite the reduced synaptic Ca²⁺-influx, presynaptic heterogeneity was unchanged at RIM-BP2^−/− AZs (Figure 4B). Furthermore, we estimated the spatial extent of the synaptic Ca²⁺ signals by fitting a 1D Gaussian function to the ΔF amplitudes of the line scans, 15 ms after depolarization onset. The full width at half maximum (FWHM) at RIM-BP2^−/− AZs were not significantly different to that of RIM-BP2^+/+ AZs (RIM-BP2^+/+ FWHM = 1.43 ± 0.03 μm, RIM-BP2^−/− FWHM = 1.34 ± 0.03 μm, p = 0.5, Figure 4B). As the confocal Ca²⁺-imaging does not provide insight into the nanoscale organization of Ca²⁺-channels at the AZ we turned to STED super-resolution microscopy of Ca_V1.3 immunofluorescence.

Ca_V1.3 Ca²⁺-channels remained clustered at ribbon-occupied AZs in RIM-BP2^−/− IHCs (Figure 5A). We subjectively classified the Ca²⁺-channel clusters according to their appearance, whereby the majority of them (66%) assumed a “stripe-like” or “double stripe-like” arrangement of Ca²⁺-channels. Only 12% of Ca²⁺-channel clusters formed “round” and 22% “complex” cluster shapes. We note that we focused on synapses seemingly lying en face in the imaging plane, and refer to apparent shapes as we did not assess the clusters with 3D STED resolution. We found the classes of apparent shapes and their representation to be very similar in RIM-BP2^−/− IHCs with a slight trend toward less stripe-shaped (62%) and more round (16%) appearance. We approximated the dimensions of “stripe-shaped” Ca²⁺-channel clusters by fitting a 2-dimensional (2D) Gaussian function to STED images of Ca_V1.3 immunofluorescence (Figures 5A,C, quantifications Figure 5D, Table 2). The FWHM of the long axis was significantly shorter (p = 0.009) and the cluster area (22% reduction, p = 0.01) as well as the integral of the 2D Gaussian fit function (~49% of WT, p < 0.0001) were significantly smaller in RIM-BP2^−/− AZs (Figure 5D), whereas the short axis was unaltered (p = 0.1). Consequently, there was a trend toward a smaller FWHM ratio of long over short axis, which, however, did not reach statistical significance (p = 0.1). We interpret the calculated area and integral of the 2D Gaussian fit as a measure of Ca²⁺-channel cluster size and the integrated Ca_V1.3 immunofluorescence as a proxy for the number of Ca²⁺-channels. However, we note that Ca_V1.3 immunofluorescence might not relate linearly to the number of Ca²⁺-channels. In summary, our morphological STED imaging indicates a “rounder” shape, a smaller size and a halved integrated Ca_V1.3 immunofluorescence of the Ca²⁺-channel clusters of RIM-BP2-deficient AZs, which is consistent with the 41% reduction of the ΔF_max/F₀ found by Ca²⁺-imaging. The same analysis was performed for bassoon clusters (Figure 5B, quantifications Figure 5E, Table 2). Here, we found no difference in the FWHM of the long axis, short axis, long/short axis ratio and cluster area between AZs of the two genotypes. However, a significant increase in bassoon cluster integral (21%) was found in RIM-BP2^−/− AZs potentially indicating a compensatory upregulation of bassoon.

RIM-BP2 has been indicated to serve as molecular link between Ca²⁺-channels and release ready SVs via Rab3/RIMs interaction (Acuna et al., 2015; Grauel et al., 2016). Hence, next we investigated the role of RIM-BP2 in regulating synaptic transmission either via its promotion of Ca²⁺-channel abundance, coupling Ca²⁺-channels and vesicular release sites or by directly affecting SV exocytosis. We addressed this possibility by perforated patch-clamp recordings of exocytic membrane capacitance changes (ΔC_m) in response to voltage-gated Ca²⁺-influx in RIM-BP2^+/+ and RIM-BP2^−/− IHCs. We studied RRP and sustained exocytosis by probing ΔC_m in response to depolarizations of various durations to the potential eliciting the maximum Ca²⁺-current (−14 mV, Figure 6A). RRP exocytosis, approximated as ΔC_m evoked by 20 ms depolarizations (Moser and Beutner, 2000), tended to be smaller in RIM-BP2^−/− IHCs (reduced by 21%, but not reaching statistical significance: p = 0.1). The sustained phase of SV exocytosis, probed by responses to stimuli longer than 20 ms, is thought to primarily reflect vesicle resupply to the RRP and subsequent fusion (Schnee et al., 2005; Goutman and Glowatzki, 2007; Meyer et al., 2009; Neef et al., 2009). The reduction in sustained exocytosis was more pronounced (e.g., 38% reduction of ΔC_m for 50 ms, p = 0.0007). In addition, a line was fit to the sustained phase of exocytosis (from 50 to 200 ms) to estimate the kinetics of SV exocytosis during ongoing depolarization (Figure 6A, dashed lines in top left panel). In RIM-BP2^−/− IHCs, SV exocytosis rates were significantly slower (193 fF/s) compared to RIM-BP2^+/+ IHCs (279 fF/s, p = 0.001, Figure 6A, inset).

Next, we tested, whether relating the exocytic ΔC_m to the corresponding integrated Ca²⁺-current Q_Ca (Figure 6B) reduced the discrepancy of the exocytic responses between the RIM-BP2^+/+ and RIM-BP2^−/− IHCs. Indeed, the ΔC_m/Q_Ca ratio was indistinguishable between both genotypes for 20 ms depolarizations (p = 0.9). However, we still observed a significant difference between the two genotypes regarding the sustained phase of exocytosis (ΔC_m/Q_Ca for 50 ms: p = 0.003, for 100 ms: p = 0.04, for 200 ms: p = 0.01). This indicated that the reduction of presynaptic Ca²⁺-influx does not fully account for the impaired sustained exocytosis, suggesting a role of RIM-BP2 in SV-replenishment. When raising the extracellular Ca²⁺ concentration ([Ca²⁺]_e) from 2 to 5 mM, thereby boosting single channel Ca²⁺-current, we observed a rescue of the exocytic response of RIM-BP2^−/− IHCs that then was indistinguishable from RIM-BP2^+/+ at 2 mM [Ca²⁺]_e (50 ms p = 0.9, 100 ms: p = 0.9, 200 ms: p = 0.4) (Figure 6B). This suggests that boosting single Ca²⁺-channel current can overcome the alteration imposed by RIM-BP2 disruption, potentially because Ca²⁺-domains around open Ca²⁺-channels spread further out reaching more remotely positioned fusion-competent vesicles.

Next, we studied Ca²⁺-influx–exocytosis coupling by probing the apparent Ca²⁺-cooperativity m of RRP exocytosis (Augustine et al., 1991). We had previously found m to be close to unity for IHCs after the onset of hearing, indicating a Ca²⁺-nanodomain-like control of exocytosis in IHCs (Brandt et al., 2005; Wong et al., 2014; Pangršič et al., 2015). If RIM-BP2 was a molecular linker between the Ca_V1.3 Ca²⁺-channel complex and the vesicular release site, essential for establishing this Ca²⁺-nanodomain-like control, we would expect an increase in m. We repetitively applied 20 ms depolarizations separated by at least 60 s for SV-replenishment to secure complete RRP recovery and successively reduced the number of open Ca²⁺-channels by slow perfusion of the dihydropyridine channel L-type Ca²⁺-channel antagonist isradipine (10 μM, Figure 7). We obtained m by fitting the relationship of ΔC_{m,20 ms} and corresponding Q_Ca for each cell with a power function ΔC_m = A(Q_Ca)^m restricted to the range that did not show obvious saturation of ΔC_{m,20 ms} (Figure 7). The estimates of m were statistically indistinguishable between RIM-BP2^−/− IHCs and RIM-BP2^+/+ IHCs (1.55 ± 0.04 vs. 1.64 ± 0.17, p = 0.5) arguing for intact Ca²⁺-influx–exocytosis coupling for RRP SVs, at least after full RRP recovery. This, however, does not rule out a function of RIMB-BP2 in promoting the engagement of newcoming vesicles with Ca²⁺-channels. In fact, the reduction of sustained exocytosis as well as its increased Ca²⁺-dependence would be in line with such a scenario.

Our next step was to test for a role of RIM-BP2 in replenishing SVs during RRP recovery from depletion using a paired-pulse paradigm. We used a 20 ms depolarization to −14 mV to deplete the RRP and then probed its recovery by a consecutive 20 ms pulse applied within variable intervals (Δt) to the first pulse (Figure 8A). RRP recovery was slowed in RIM-BP2^−/− IHCs as evident from significantly smaller responses at an interval of 50 ms (p = 0.02, Figure 8B). The time course of RRP recovery was estimated by fitting the paired-pulse data with a single exponential function (Figure 8B, dashed lines) where the RRP refilling time constant amounted to 51 ms in RIM-BP2^−/− IHCs compared to 27 ms in RIM-BP2^+/+ IHCs. These two data sets indicate that in IHC ribbon synapses, RIM-BP2 is required for fast recruitment of new incoming SVs to the release sites after RRP-depletion. We speculate that RIM-BP2 could help to quickly guide SVs into close proximity of a nearby Ca²⁺-channel. Longer RRP recovery times (>50 ms) were sufficient for normal RRP replenishment in RIM-BP2-deficient IHCs (Figure 8B). Likely, longer recovery times increase the probability of proper positioning of replenished SVs to their release site by alternative protein-protein interactions.

To examine the effect of RIM-BP2 disruption on an ultrastructural level, we studied IHC ribbon synapses in RIM-BP2^−/− and RIM-BP2^+/+ mice using electron tomography (Figure 9). The overall synapse ultrastructure appeared normal: the size of synaptic ribbons was unaltered (Figure 9B) and synaptic ribbons were anchored to the presynaptic membrane via the presynaptic density (PD) (Figure 9A), which was normal in size as well (Figure 9B). Further, the number of SVs directly facing the AZ membrane (membrane-proximal SVs, MP-SVs, see section Materials and Methods) and of SVs directly adjacent to the ribbon (ribbon-associated SVs, RA-SVs) was unaltered (Figure 9B).

Next, we analyzed the distribution of MP-SVs regarding their distance to the plasma membrane and PD (Figures 9C,D). On average, no difference in the MP-SV distance to the membrane or PD was observed (Figure 9C). The cumulative distribution function of membrane-to-membrane distance however, shows that a large fraction (70%) of MP-SVs was located within ~25 nm from the plasma membrane and a smaller fraction (30%) of all MP-SVs between ~25 and 50 nm at AZs of both genotypes (Figure 9D top). Interestingly, the distributions differed for both fractions of MP-SVs between RIM-BP2^+/+ and RIM-BP2^−/− AZs (small fraction: p = 0.02 and large fraction: p = 0.004). There was a tendency of more MP-SVs in RIM-BP2^−/− AZs to populate the smaller distance bins (Figure 9D), while we observed less SVs with intermediate distance (~20–40 nm), but a second accumulation of MP-SVs further away from the membrane (~40–50 nm) that seemed absent at RIM-BP2^+/+ AZs. The lateral distribution of MP-SVs with respect to the PD was unaltered in RIM-BP2^−/− AZs (Figure 9D bottom).

Presence and unaltered size of ribbon and PD are consistent with our above immunofluorescence analysis (Figures 1, 2, 4). The normal count of MP-SVs suggests that exocytosis is not limited by the availability of SVs. We speculate that RIM-BP2 might instead promote fast replenishment of readily releasable vesicles by rapidly engaging them with Ca_V1.3 channels, i.e., by positional priming (Neher and Sakaba, 2008), that we might not have resolved by our static tomographic analysis.

In order to probe the impact of RIM-BP2 on auditory systems function we first recorded auditory brainstem responses (ABR) and distortion product otoacoustic emissions (DPOAE) in RIM-BP2^−/−mice. While DPOAE test the function of the cochlear amplification mediated by outer hair cells, ABR reflect the synchronized neural activation of the early stages of the auditory pathway. ABR thresholds were mildly but significantly elevated in RIM-BP2^−/− mice (Figure 10A; p_4kHz = 0.005, p_8kHz = 0.02, p_16kHz = 0.002). In addition, ABR wave I amplitude, reporting the compound action potential of the SGNs, was significantly reduced (Figure 10B; p = 0.03). As the presence of otoacoustic emissions with normal amplitudes (for all F2 intensities: p > 0.05) (Figure 10C) indicates normal mechanoelectrical transduction and cochlear amplification, this suggests a mild synaptopathic hearing impairment (Moser and Starr, 2016). Potentially, the hearing phenotype might still be attenuated by the presence of RIM-BP1 or −3 in IHCs, which should be tested in future studies.

Using extracellular recordings we studied sound encoding by single postsynaptic SGNs, which offers in vivo analysis of single AZ function as each SGN is thought to receive input from just one AZ of one IHC (Liberman, 1978). We stereotactically targeted microelectrodes to where the auditory nerve fibers (central axons of SGNs) enter the AVCN and recorded spontaneous and sound evoked firing in neurons that most likely represent SGNs (“putative” SGNs, hereafter dubbed SGNs for simplicity, see section Materials and Methods). We observed a larger proportion of SGNs exhibiting low spontaneous firing in RIM-BP2^−/− mice (Figure 11A; p = 0.04), which might result from the reduction in the Ca²⁺-channel abundance at the IHC ribbon synapses (Figures 4, 5). We found normal frequency tuning (Figure 11B) and sound thresholds (Figure 11B′) of RIM-BP2^−/− SGNs, which supports our above notion of intact cochlear amplification (Figure 10C). We then studied the neural response to 50 ms tone bursts at the characteristic frequency at a sound pressure level of 30 dB above threshold (Figure 12A), which is considered to reflect the presynaptic RRP dynamics convolved with postsynaptic refractoriness (Wittig and Parsons, 2008; Buran et al., 2010; Frank et al., 2010). The peak firing rate reports the maximal initial rate of SV release from the SV-occupied release sites of the standing RRP, (Pangršič et al., 2012). The partial depletion of the RRP governs the rapid spike rate adaptation and the adapted firing rate reflects the balance of SV fusion and replenishment. We found a reduced peak firing rate in RIM-BP2^−/− SGNs (p = 0.02), while the adapted firing rates where indistinguishable between SGNs of both genotypes (Figure 12A). The first spike latency was significantly delayed in RIM-BP2^−/− SGNs (p = 0.004), but the trend toward an increased variance of first spike latency did not reach statistical significance (Figure 12B). Together, these findings indicate a mild impairment of sound onset coding, which likely results in less synchronized firing of the SGN population thereby explaining the ABR phenotype. Sound coding during continued stimulation, therefore also SV-replenishment, on the other hand, seemed unaltered, which contrasts our results on IHC physiology (Figure 8). Therefore, we tested SV-replenishment as recovery on sound onset coding at variable times after cessation of stimulation (forward masking paradigm (Harris and Dallos, 1979), which provides an estimate of the kinetics of the RRP recovery after partial depletion (Figure 12C). Probe responses at short intervals to a 100 ms masker tone tended to be smaller for RIM-BP2^−/− SGNs for 4 ms (p = 0.07) and significantly so for 16 ms (p = 0.04), but from the interval of 64 ms onwards, the recovered response was comparable to that of RIM-BP2^+/+ SGNs, and the time constant for recovery from masking was unaltered. In summary, recordings from single SGNs indicated a mild impairment of sound onset coding and suggested that fast vesicle replenishment is slowed in the absence of RIM-BP2.

Sensory hair cells provide great access for biophysical analysis of Ca²⁺-influx and exocytosis enabling presynaptic studies of Ca²⁺-channels and their regulation of transmitter release. Ribbon-type AZs of IHCs, on average, are thought to cluster ~80–100 Ca²⁺-channels (Brandt et al., 2005; Zampini et al., 2013), which is very similar to frog saccular AZs (90 Ca²⁺-channels) (Roberts et al., 1990). So far, a positive regulation of the number and nanoscale organization of Ca_V1.3 Ca²⁺-channels was shown for the scaffold proteins bassoon and RIM. RIMs, intensively studied interacting partners of Ca_V channels (Coppola et al., 2001; Kiyonaka et al., 2007; Gebhart et al., 2010; Kaeser et al., 2011; Picher et al., 2017), promote clustering of Ca_V1.3 Ca²⁺-channels at the AZs of IHCs, but stripe-like Ca²⁺-channels clusters are maintained despite RIM2α disruption (Jung et al., 2015). Disruption of bassoon, on the other hand, not only led to a major reduction in the number of synaptic Ca²⁺-channels, but in addition caused a disintegration of the Ca²⁺-channel clusters at the IHC AZ (Frank et al., 2010). However, the interpretation of bassoon disruption is complicated as it also causes the loss of synaptic ribbons (Dick et al., 2003; Khimich et al., 2005; Jing et al., 2013). Our present study, using functional and morphological imaging in RIM-BP2-deficient IHCs indicated a ~40% reduction of the number of synaptic Ca²⁺-channels that remained clustered beneath the presynaptic density of the ribbon-occupied AZs. The decline of synaptic Ca²⁺-influx exceeded the reduction of the whole-cell Ca²⁺-influx (~20%), suggesting that AZs fail to cluster the available Ca²⁺-channels. Consistent with impaired synaptic clustering of Ca²⁺-channels we found a reduced length of the stripe-like clusters, which also showed lower integrated Ca_V1.3 immunofluorescence intensity. Importantly, the length of the presynaptic density, measured by STED microscopy of bassoon immunofluorescence, was normal for the AZs of RIM-BP2-deficient IHCs. Together, this suggests that the reduction of Ca²⁺-channel number occurs despite the presence of the ribbon and of a normal bassoon-containing presynaptic density, placing RIM-BP2's function in channel clustering downstream of bassoon. Both, the width and area of Ca_V1.3 clusters revealed by STED microscopy and the spread of the Ca²⁺ signal at the AZ appeared normal arguing against a mis-localization of Ca²⁺-channels to outside the area marked by the presynaptic density. Moreover, the biophysical properties of IHC Ca²⁺-channels, voltage-dependence of activation as well as extent and kinetics of inactivation, were unaltered. We cannot fully exclude changes in Ca²⁺-channel open probability due to delayed synapse maturation upon RIM-BP2 disruption. However, several findings argue against a developmental deficit in RIM-BP^−/− IHC AZs: (i) Ca²⁺-current inactivation was unaltered and mature (Grant and Fuchs, 2008), (ii) even though Ca²⁺-channel clusters were rounder, they were spatially confined and co-localize with immunolabeled presynaptic ribbons, which were present in mature numbers (Sendin et al., 2007; Zampini et al., 2010; Wong et al., 2014), (iii) finally the apparent Ca²⁺-cooperativity of exocytosis of 1.5 in RIM-BP2^−/− IHCs indicates tight Ca²⁺-nanodomain-like control of exocytosis (Wong et al., 2014). This study is the first to report a role of RIM-BPs in regulating the number of Ca²⁺-channels at the AZ of a mammalian synapse, thereby contrasting findings at hippocampal synapses and the calyx of Held synapse (Acuna et al., 2015; Grauel et al., 2016). However, a reduction in presynaptic Ca²⁺-influx was found at the neuromuscular junction of Drosophila larvae, which also displays an elaborate dense projection (T-bar) and requires d-RIM-BP for sufficient Ca²⁺-influx and transmitter release (Liu et al., 2011; Müller et al., 2015).

How does RIM-BP2 contribute to the regulation of synaptic transmission at IHC synapses that drive sound encoding in SGNs? We approached this question by analysis of IHC exocytosis, extracellular recordings from SGNs and measurements of ABR that rely on synchronous activation of SGNs. We tested two hypotheses: (i) reduction of exocytosis in proportion with the loss of presynaptic Ca²⁺-channels due to preserved Ca²⁺-nanodomain-like control of exocytosis and vesicle replenishment, (ii) reduction of exocytosis beyond that of the Ca²⁺-influx due to additional disruption of Ca²⁺-nanodomain-like control of exocytosis and/or impaired vesicle replenishment. Emerging evidence shows for neuromuscular junction of drosophila larvae (Liu et al., 2011; Müller et al., 2015), hippocampal synapses (Grauel et al., 2016) and the calyx of Held (Acuna et al., 2015) that genetic inactivation of RIM-BP disrupts the tight coupling between Ca²⁺-channels and release site.

Here, we tested the apparent Ca²⁺-cooperativity m of RRP exocytosis for changes in Ca²⁺-influx brought about by successive dihydropyridine block of Ca_V1.3 Ca²⁺-channels. We had established by various experiments and modeling that the near linear m (typically around 1.5) for channel block, which contrasts an m of 3–4 during changes of single channel current, is an indicator for Ca²⁺-nanodomain-like control of exocytosis at the mature IHC AZ (Brandt et al., 2005; Wong et al., 2014; Pangršič et al., 2015). When testing m of RRP exocytosis during dihydropyridine block with inter-pulse intervals of 60 s and more, in order to secure full RRP replenishment, we found no indication for a looser Ca²⁺-influx—exocytosis coupling in RIM-BP2-deficient IHCs. The mild effect of genetic deletion RIM-BP2 on RRP exocytosis and the intact Ca²⁺-nanodomain coupling are surprising given the findings of previous studies on RIM-BP function and might suggest that IHC ribbon-type AZs can use additional molecular mechanisms to establish tight coupling of Ca²⁺-channels and release sites in the absence of RIM-BP2. In this respect it is worth mentioning, that disruption of the RIM-BP interaction partners bassoon or of RIM2α did not lead to an obvious switch to Ca²⁺-microdomain control of RRP exocytosis at IHC synapses either (Frank et al., 2010; Jung et al., 2015). It is possible that in IHCs, only combinatory deletion of several proteins like for example RIM and RIM-BP2 would elicit any measurable effect on the tight coupling of Ca²⁺-channels and release sites. Further, the observation of tight Ca²⁺-nanodomain-like control of exocytosis for fusion-competent SVs of the RRP in the absence of RIM-BP2 was made under the artificial conditions of the biophysical experiment, where the synapse was inactivated due to voltage-clamp to hyperpolarized potential for full RRP recovery. In physiology, on the other hand, IHC AZs mediate massive vesicle turnover, which requires rapid replenishment at rates of hundreds of vesicles per second (Pangršič et al., 2012). For precise sound coding under conditions of high-frequency transmission, the engagement of newcoming vesicles with one or few nearby Ca²⁺-channel complexes needs to proceed at a similar pace (Cho et al., 2011). This might require multiple mechanisms of molecular interaction posing a higher requirement for each of the candidate molecular linkers. Interestingly, we found an impairment of RRP replenishment in RIM-BP2-deficient IHCs, which seems consistent with such an increased requirement for RIM-BP2 during turnover. One conceivable interpretation would be that RIM-BP2 acts as an indirect molecular linker between SVs and Ca²⁺-channels e.g., through Rab3—RIM—RIM-BP2—Ca²⁺-channel interaction (Hibino et al., 2002; Kaeser et al., 2011). Thereby, RIM-BP2 might engage molecularly primed SVs with a nearby Ca²⁺-channel after RRP depletion. This step was found to be rate-limiting as it converts slow releasing SVs into fast releasing SVs and was called “positional priming” (Neher and Sakaba, 2008). “Positional priming” was found to be actin dependent (Lee et al., 2013), for which we did not test here. However, a recent study reported that actin secures SV replenishment in IHCs at high rates and prevents SV exhaustion during the sustained phase of exocytosis (Guillet et al., 2016). Upon Actin network disruption in IHCs, SVs, which would only be available for release after RRP depletion during longer stimulations (≥50 ms) in control condition, fused with the plasma membrane ahead of time, indicating a role of actin in preventing premature exocytosis (Guillet et al., 2016). Further, the observed shift in the MP-SV distribution with respect to the plasma membrane in RIM-BP2^−/− EM tomograms suggests a change in SV position with respect to release sites. The observed change was small, but likely, the recruitment process of SVs is so fast that it is not fully reflected in our EM data. Both paired-pulse experiments and forward masking point out a significant delay in RRP refilling for short inter-stimulus intervals, and thereby support the hypothesis that RIM-BP2 plays a role in fast RRP refilling. Additionally, the reduced number of synaptic Ca²⁺-channels in RIM-BP2 deficient AZs might enhance the delay during “positional priming,” since fewer channels are available for SVs to connect. Hence, SV turnover speed would be limited by a lower probability of SV—release site matching due to a reduced number of Ca²⁺-channels and possibly also through the slowed engagement of SVs with the remaining Ca²⁺-channels. Given the possibility of RIM-BP2 interaction with the ribbon anchor bassoon (Davydova et al., 2014) another conceivable interpretation for the slow RRP replenishment upon RIM-BP2 disruption could be a slightly increased distance between Ca²⁺-channels and the presynaptic density and thus, ribbon-tethered SV. Both, STED microscopy and Ca²⁺-imaging did not reveal changes in the width and area of Ca_V1.3 clusters and spread of the Ca²⁺ signal, arguing against major changes in Ca²⁺-channel localization. In addition, the distance of MP SVs to the PD was unaltered in RIM-BP2 deficient synapses. However, we cannot rule out that already a slight increase of the distance of Ca²⁺-channels from the PD could slow down the engagement of new incoming ribbon-tethered SVs with a Ca²⁺-channel during RRP replenishment. Interestingly, the reduced efficiency of Ca²⁺ influx to trigger exocytosis could be rescued by elevating extracellular [Ca²⁺]_e in RIM-BP2^−/− IHCs. Elevated extracellular [Ca²⁺]_e could lead to accelerated Ca²⁺-dependent SV trafficking and replenishment which was reported for hair cells (Moser and Beutner, 2000; Spassova et al., 2004; Cho et al., 2011; Goutman and Glowatzki, 2011; Schnee et al., 2011) and other synapses (e.g., calyx of Held synapses Wang and Kaczmarek, 1998; Hosoi et al., 2007). Lower presynaptic Ca²⁺ signaling due to the reduced number of synaptic Ca²⁺-channels in RIM-BP2 deficient AZs might therefore explain their slower SV replenishment. In addition, or alternatively, increasing Ca²⁺-influx through elevation of extracellular [Ca²⁺]_e might recruit more distant release sites so that the exact spatial distribution and matching of Ca²⁺-channels and SVs is less critical (Fuchs et al., 2003; Pangršič et al., 2015) thereby bypassing “positional priming” as a potential rate-limiting step of exocytosis in RIM-BP2 deficient IHC synapses. A role for RIM-BP in positional priming was previously proposed based on observing slower resupply of high release probability SVs at DRBP-deficient Drosophila NMJ (Müller et al., 2015). Therefore, loss of RIM-BP2 likely weakens the SV replenishment at IHC ribbon synapse, leading to a reduction of the SVs available for release (“standing RRP”) explaining the impaired synchronous activation of SGNs reflected by the reduced wave I in the auditory brainstem recording (ABR). Since RIM-BP2 is also knocked out in efferent olivocochlear neurons that provide inhibitory feedback onto the afferent terminals of type I SGNs (Ruel et al., 2001), we cannot rule out effects caused through alterations in efferent modulation. Nonetheless, the sound encoding phenotype upon RIM-BP2 disruption was rather mild. Future studies should test the presence and potential function of other RIM-BP isoforms (RIM-BP1 and RIM-BP3) that might partially compensate the function of RIM-BP2.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.