Mice with cre-mediated excision (Lakso et al., 1996) of exon 14 of the Pclo gene and insertion of a neomycin resistance cassette in the adjacent 3′ intron (Mukherjee et al., 2010; PicMut), and their wildtype littermates (PicWT), of either sex, were studied from postnatal day 14–23. The mouse line was derived by heterozygous breeding with C57Bl/6J genetic background. Animals were genotyped, and re-genotyped post experiments, using PCR. PicBsn animals, with only one intact allele of the Bsn gene in addition to Piccolo mutation, were used. These were derived by heterozygous breeding of PicMut with Bsn^ΔEx4/5 mice (exons 4 and 5 of Bassoon gene deleted (Altrock et al., 2003). All experiments were performed in compliance with the guidelines of the German animal welfare act and were approved by the board for animal welfare of the University Medical Center Göttingen and the animal welfare office of the state of Lower Saxony.

Animals were transcardially perfused with 2% freshly prepared ice-cold paraformaldehyde with pH adjusted to 7.4. The fixed brain was then removed and brainstem was dissected with a coronal cut few millimetres nasal to the junction between occipital cortex and cerebellum. The brain block was washed overnight in 30% sucrose solution in PBS. For sectioning, the brain block was embedded in Tissue Tek Cryomatrix (Thermo Fisher Scientific, Waltham, MA, USA) and then fixed on the stage of the cryostat (Figocut E cryotome, Reichert-Jung, Depew, NY, USA) such that the caudal aspect was facing upwards and the dorsal side was towards the blade. Advancing from caudal to nasal, 30 μm coronal sections were cut (chamber temperature: −20°C, object temperature: −22°C) and discarded until the appearance of the 7th cranial nerve. Subsequent sections containing aVCN were collected onto electrostatically charged microscope slides (SuperFrost Plus, ThermoFisher Scientific, MA, USA). For parallel processing, one slice of each genotype was collected per slide. Thereafter, the slices were washed for 10 min in PBS and incubated 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) for 1 h, followed by incubation in primary antibodies diluted in GSDB, for 3 h, in a wet chamber at room temperature. After washing 2 × 10 min with wash buffer (450 mM NaCl, 0.3% Triton X-100, 20 mM phosphate buffer) and 2 × 10 min with PBS, the slices were incubated with secondary antibodies diluted in GSDB for 1 h in a light-protected wet chamber at room temperature. The slices were then washed 2 × 10 min with wash buffer, 2 × 10 min with PBS and 1 × 10 min in 5 mM phosphate buffer, and finally mounted with a drop of fluorescence mounting medium based on Mowiol 4-88 (Carl Roth, Karlsruhe, Germany) and covered with a thin glass coverslip.

For inner hair cell immunohistochemistry, apical turns of freshly dissected organ of Corti were fixed in 4% formaldehyde for 10 min and processed as described for fixed brain slices.

Primary antibodies used were: rabbit anti-Piccolo (Antibody #1; 1:200), guinea pig anti-Piccolo (Antibody #2; 1:200), mouse anti-Otoferlin (1:1000; Abcam Biochemicals, Cambridge, UK), guinea pig anti-VGLUT1 (1:500), rabbit anti-VGLUT1 (1:1000), mouse anti-Gephyrin (1:500), mouse anti-Sap7f407 to Bassoon (1:500; Abcam, Cambridge, UK), guinea pig anti-Bassoon (1:500), mouse anti-Calretinin (1:300; Swant, Marly, Switzerland), guinea pig anti-VGAT (1:600), rabbit anti-Munc13-1 (1:200), rabbit anti-RIM1 (1:200), rabbit anti-RIM2 (1:200). Unless stated otherwise, primary antibodies were purchased from Synaptic Systems, Göttingen, Germany. Secondary antibodies used were: AlexaFluor488-, AlexaFluor568- and AlexaFluor647-labeled antibodies (1:200, Invitrogen).

Confocal images were acquired using a laser-scanning confocal microscope (Leica TCS SP5; Leica Microsystems) equipped with 488 nm (Ar) and 561/633 nm (He-Ne) lasers, and 63×/1.4 NA oil-immersion objective. Samples of both genotype: PicWT and PicMut were processed and imaged in parallel, using same laser power, gain and microscope settings.

Animals were transcardially perfused using Karlsson-Schultz buffer (Karlsson and Schultz, 1965; 4% paraformaldehyde, 2.5% glutaraldehyde, 0.5% NaCl, with pH adjusted to 7.4), and the isolated brain was fixed for an additional hour in Karlsson-Schultz buffer and subsequently 150 μm thick parasagittal slices from cochlear nuclei were obtained as described for in vitro physiology, except that the sections were cut at a blade feed rate of 1 mm/s and an amplitude of 1 mm, and a separate vibratome (Leica VT 1200S) was used for sectioning fixed tissue. Slices were fixed overnight on ice using 2% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.2 and washed three times in sodium cacodylate buffer. Post-fixation was performed for 1 h with 1% osmium tetroxide (in 0.1 M sodium cacodylate buffer), followed by a 1 h washing step in sodium cacodylate buffer and three brief washing steps in distilled water. En bloc staining using 1% uranyl acetate in distilled water for 1 h on ice was followed by a brief washing step with distilled water. Finally, slices were dehydrated at room temperature with increasing ethanol concentrations and infiltrated in Epoxy resin (Agar 100 kit, Plano, Wetzlar, Germany; 100% EtOH/Epoxy 1:1, 30 and 90 min; 100% Epoxy, overnight). Slices were further incubated in fresh 100% Epoxy resin and placed on embedding molds. After polymerization for 48 h at 70°C, 70–75 nm sections were cut and placed on formvar-coated copper slot grids (Athene, Plano, Wetzlar, Germany). Slices were post-fixed and stained with uranyl acetate replacement solution/lead citrate following standard protocols. Micrographs were taken with a JEOL electron microscope (JEM 1011, JEOL, Freising, Germany) equipped with a GatanOrius 1200A camera (Gatan, Munich, Germany) using the Gatan Digital Micrograph software package at 12,000×. Endbulb AZs of the aVCN were chosen by the appearance of the vesicles (round) and the synaptic site (asymmetric, denoted as PSD). Only excitatory synapses were taken into account. For analysis, only those AZs were used that showed a clear pre- and postsynaptic membrane.

Electrophysiology data were analyzed using Igor Pro (Wavemetrics, Lake Oswego, OR, USA), Mini Analysis (Synaptosoft Inc., Fort Lee, NJ, USA) and GraphPad Prism software (La Jolla, CA, USA). Confocal images were analyzed using ImageJ software, Imaris (Bitplane AG, Zurich, Switzerland) and Matlab (Mathworks). Endbulb terminals were tracked and counted visually using ImageJ from Calretinin-stained confocal image stacks. Figures were assembled for display using Adobe Illustrator (Adobe Systems, Munich, Germany). Statistical significance between groups was determined by either unpaired Student’s t-test (in case of normally distributed data with comparable variances between the groups) or Wilcoxon rank sum test (when data distribution did not satisfy the criteria). Normality of distribution was tested with Jarque-Bera test and variances were compared with F-test. Data were presented as mean ± SEM when compared using Student’s t-test. In case of Wilcoxon’s rank sum test, data were presented as box and whisker plots showing grand median (of the means of all recordings), lower/upper quartiles, 10–90th percentiles). *, **, *** indicate p < 0.05, 0.01 and 0.001 respectively.

This study employed constitutive Piccolo mutants with a targeted deletion of exon 14 of Pclo gene and insertion of neomycin resistance cassette in the adjacent 3′ intron, described previously (hereafter dubbed “PicMut”), which lowers the protein levels of Piccolo to approximately 5% in the brain (Mukherjee et al., 2010). A shorter, ~330 kD, C-terminal truncated Pclo splice variant, Piccolino (Figure 1A), is the predominant Pclo isoform at the ribbon synapse of IHC (Regus-Leidig et al., 2013), directly preceding the endbulb of Held. Labeling with antibody #2 (directed against a central epitope, Figure 1A), we found that IHC ribbon synapses in PicMut still express Piccolino (Figure 1C), while the full-length variant, Piccolo, was absent (identified by antibody #1 directed against a C-terminal epitope, Figures 1A,C’). Recordings of ABR (ABR; see “Materials and Methods” section) showed normal sound thresholds and unaltered amplitudes of the spiral ganglion compound action potential, reflected in wave I of the ABRs, indicating normal cochlear sound encoding (Figure 2). In contrast to IHCs, both the excitatory AZs of the endbulb of Held synapse and inhibitory AZs, in the aVCN showed a near complete loss of Piccolo staining (10% of control levels; p-value < 0.001, Wilcoxon rank sum test) indicating a major reduction of Piccolo expression at these central auditory synapses (Figures 1D–E). Our data also indicated a lower fluorescent intensity for Piccolo at PicWT inhibitory AZs as compared to PicWT excitatory endbulb AZs (p-value < 0.01, Wilcoxon rank sum test), suggesting that Piccolo may be differentially expressed in the two types of AZs. However, in this study we focus exclusively on the excitatory endbulb synapses and the exploration of Piccolo’s role at inhibitory synapses would require an independent study. The excitatory AZs at the BCs were identified by co-localization of Piccolo with vesicular glutamate transport 1 (Vglut1; maximal center of mass distance 0.4 μm in xy and 1.2 μm in z), while inhibitory AZs were distinguished by co-localization with Gephyrin (maximal center of mass distance 0.3 μm in xy and 0.9 μm in z). Immunofluorescence intensities of Vglut1 (Figure 1F; p-value ≥ 0.05, Wilcoxon signed rank test) and Gephyrin remained unchanged (Figure 1F’; p-value ≥ 0.05, paired Student’s t-test). We note that this did not strictly differentiate AZs of endbulbs from those of glutamatergic bouton endings, which, however, are much fewer in number (Nicol and Walmsley, 2002; Gómez-Nieto and Rubio, 2009). Therefore, and since Vglut1 is primarily associated with terminals of SGNs (Heeringa et al., 2016), our estimates were strongly dominated by endbulb AZs. Functionally, wave II of the ABRs, thought to arise from activity in the cochlear nucleus (Melcher and Kiang, 1996), demonstrated reduced amplitude suggesting a functional impairment of synaptic transmission from SGNs to the aVCN neurons (Figure 2). Interestingly, later ABR waves seemed unaltered suggesting compensation of Piccolo deficiency.

We performed further semi-quantitative immunohistochemistry to analyze the effect of Piccolo disruption on the number of the endbulb of Held synapses and their AZs, as well as on the molecular composition of the AZs. We quantified the number of endbulbs converging on to BCs by visually tracing and counting Calretinin-stained endbulbs (Figure 3A; Lohmann and Friauf, 1996; Mendoza Schulz et al., 2014), not differentiating between globular and spherical BCs. BCs of both genotypes received 3-4 endbulbs on average. 3.56 ± 0.24 for PicMut and 3.67 ± 0.26 for PicWT, Figure 3B; p-value ≥ 0.05, Student’s t-test, which agrees with the number reported in the literature (Cao and Oertel, 2010; Mendoza Schulz et al., 2014) for mice after onset of hearing (p15–p21).

Next, we quantified the number of excitatory AZs per endbulb. In stacks of confocal sections of BCs (Figure 3C), we counted puncta immunofluorescent for the AZ markers (such as Bassoon, RIM1, RIM2 or Munc13-1), which gave us the total AZ count. Subtracting the number of immunofluorescent puncta juxtaposed with Gephyrin immunofluorescence (inhibitory AZ number, see above) from the total count of AZs yielded the number of excitatory AZs, which was unaltered in the PicMut synapses. 428.40 ± 17.93; (median 403) for PicMut and 455.16 ± 18.66; (median 445) for PicWT, Figure 3D, p-value ≥ 0.05, Wilcoxon’s rank sum test. Dividing this number by the average count of endbulbs gave us the number of AZs per endbulb which was also comparable between the genotypes (120.34 for PicMut and 124.14 for PicWT) and agreed with previous reports (Nicol and Walmsley, 2002; Mendoza Schulz et al., 2014). Hence there was no discernible change in the convergence of endbulbs to BCs or the number of AZs therein.

To study the changes in molecular composition, we quantified the immunofluorescence intensities of the CAZ proteins Bassoon, RIM1, RIM2 and Munc13-1. All CAZ proteins exhibited a spot-like fluorescence pattern around the BCs depicting AZs. Bassoon, a close homolog of Piccolo (Fenster et al., 2000), demonstrated significantly increased immunofluorescence intensity at both excitatory endbulbs and inhibitory synapses (Figures 4A,B) in PicMut mice. Immunofluorescence intensities of RIM1 and RIM2 seemed overall weaker in PicMut mice, whereby RIM1 fluorescent intensity was significantly reduced at endbulb AZs and RIM2 at AZs of inhibitory synapses (Figures 4C–F). Munc13-1 immunofluorescence was slightly higher at AZs of inhibitory synapses in PicMut mice, while the intensity was not altered at endbulb AZs (Figures 4G,H). Since the discussion on the changes observed at inhibitory AZs is beyond the scope of this study, we focused mainly on the changes observed at excitatory endbulb AZs. We conclude that Piccolo disruption leads to a reduction in the abundance of RIM1 at AZs of endbulb synapses while Bassoon is upregulated, potentially as a compensatory mechanism.

Using electron microscopy, we investigated ultrastructural changes due to Piccolo deficiency at the AZs of endbulb synapses in ultrathin sections of aVCN (Figures 5A,B). While the length of the postsynaptic density (PSD) was unaltered in the mutant (Figure 5C, p-value ≥ 0.05, Wilcoxon’s rank sum test), the number of SVs per 100 nm of the PSD was significantly reduced at Piccolo-deficient endbulb AZs (Figure 5D, p-value < 0.01, Student’s t-test). For a more comprehensive analysis we compared the SV distribution within 100 nm of the presynaptic AZ membrane (perpendicular to the presynaptic membrane into the cytosol of the presynaptic terminal) in ten 10 nm bins. For each bin, there was a general trend towards fewer SVs per 100 nm of the PSD length in the mutant endbulb AZs (Figure 5E). To assess the number of membrane proximal SVs, which likely contribute to the functionally-defined RRP (Imig, 2013), we focused on the number of SVs per 100 nm of the PSD length within 40 nm of the presynaptic AZ membrane (perpendicular to the presynaptic membrane into the cytosol of the presynaptic terminal). We observed a significant reduction in the membrane proximal SVs in the mutant AZs (Figure 5F, p-value < 0.001 Wilcoxon rank sum test), suggesting a role of Piccolo in the organization and maintenance of the RRP at the AZ.

To determine the functional consequences of Piccolo disruption at the endbulb of Held synapse, we studied synaptic transmission in acute sagittal slices of the brainstem of PicMut and PicWT mice by recording EPSCs from BCs at postnatal days 15–21. BCs were distinguished from stellate cells (SC; another major cell type in the aVCN) by the faster kinetics of their postsynaptic currents (Isaacson and Walmsley, 1995) and their characteristic short-term plasticity (Chanda and Xu-Friedman, 2010). In addition to such functional identification, each recorded cell was filled with fluorescent dye Alexa 488 via the patch pipette for morphological distinction. BCs are spherical in appearance with one primary dendrite terminating in a dense bush-like dendritic tree (Wu and Oertel, 1984), distinct from SC, which are asymmetrical in shape and have multiple dendrites branching off in various directions giving them a star-like appearance (Figures 6A,B).

We first studied miniature EPSCs (mEPSC), in the presence of 1 mM Kynurenic acid (Elmslie and Yoshikami, 1985) and 100 μM CTZ (Yamada and Tang, 1993), to check if quantal size or kinetics of single vesicle release were altered at Piccolo-deficient endbulbs. As previously reported (Oleskevich and Walmsley, 2002), application of TTX does not alter the spontaneous EPSCs (sEPSCs) at the mouse BCs. This has been corroborated by various other studies (Lu et al., 2007; Mendoza Schulz et al., 2014). Therefore, mEPSCs were recorded as spontaneous events in whole-cell recordings of BC in Pic WT (Figure 6C) and PicMut (Figure 6C’) that were voltage-clamped at −70 mV (Figure 6). We did not observe differences in the mEPSC amplitude (Figure 6D), kinetics (Figures 6E–G) and frequency (Figure 6H; p-value ≥0.05 for all three quantities, Figures 6D–F,H Student’s t-test, Figure 6G Wilcoxon rank sum test). To eliminate the possibility of CTZ obscuring the changes in AMPA receptor composition, we compared the decay kinetics of mEPSC events in the absence of CTZ. The mEPSC decay times for PicWT (157.10 ± 2.86 μs; median 158 μs) and PicMut (161.72 ± 3.27 μs; median 155 μs) were comparable (p-value ≥ 0.05, Wilcoxon’s rank sum test).

Next, we investigated evoked synaptic transmission in the presence of 1 mM Kynurenic acid and 100 μM CTZ to avoid saturation and desensitization of AMPA receptors (Chanda and Xu-Friedman, 2010), respectively. Evoked EPSCs (eEPSC) were elicited by minimal electrical stimulation of the auditory nerve fibers by a monopolar electrode placed in the proximity of the recorded BC, whereby each stimulus is aimed to elicit one action potential in one endbulb (Yang and Xu-Friedman, 2008). Analysis of eEPSC evoked by single stimulations (Figure 7) revealed reduced eEPSC amplitude in BCs of PicMut mice (3.15 ± 0.25 nA; median 2.83 nA) as compared to PicWT (4.43 ± 0.37 nA; median 3.52 nA), Figures 7A,B, p-value < 0.01, Wilcoxon rank sum test. The kinetics of rise and decay (Figures 7C–E), and the synaptic delay (Figure 7F) remained unaltered; p-value ≥ 0.05, Figures 7C,D,F Wilcoxon’s rank sum test and Figure 7E Student’s t-test. Synaptic delay was calculated as the time between the start of stimulus (voltage output of the amplifier as dictated by the experiment protocol) and the time when the respective EPSC response reaches 10% of its peak amplitude.

Quantal size (mEPSC amplitude) being unaltered, we next investigated the RRP size, release probability and pool dynamics to investigate the cause of the reduced eEPSC amplitude. We studied responses to high frequency train stimulation: 50 consecutive stimuli delivered at 100 and 200 Hz (in the presence of 1 mM Kynurenic acid and 100 μM CTZ). At these frequencies, both PicMut and PicWT mice exhibited comparable short-term depression (Figures 8A,A’). Both genotypes demonstrated similar kinetics of short-term depression during train stimulation (Tau (τ), Table 1; Figures 8B,B’). The paired pulse ratio (PPR) was not significantly altered at inter-stimulus intervals of 5 ms and 10 ms (Table 1). The extent of depression assessed as the amplitude in the steady-state reached after 30 train pulses was comparable (EPSC_30–50/EPSC₁, Table 1; Figures 8B,B’).

The size of the RRP, the release probability (P_r) and the SV replenishment rate were estimated by applying two variants of the cumulative analysis to the EPSC trains: Schneggenburger-Meyer-Neher (SMN) method (Schneggenburger et al., 1999) and Elmqvist and Quastel (EQ) method (Elmqvist and Quastel, 1965; both recently reviewed in Neher, 2015). In the SMN method (Figures 8C,C’), EPSC amplitudes from trains are plotted cumulatively against the stimulus number. A line fit to the steady-state points (last 10 of the 50 points) is back-extrapolated to the y-axis, and the y-intercept divided by the mEPSC amplitude estimates the RRP size, while P_r is estimated by the ratio of vesicle content of EPSC₁ to that of the RRP. The slope of the linear fit itself, approximates the rate of vesicle replenishment during the train. The SMN analysis (Figures 8C,C’, Table 1) only revealed a non-significant trend towards a reduced RRP size in the mutant. Another finding was the reduced rate of vesicle replenishment in the mutant that reached significance at 100 Hz stimulation (p-value = 0.012, Student’s t-test). The estimate of P_r was not significantly changed upon Piccolo disruption at any stimulation frequency. Asynchronous release following the train stimulation was unchanged: the rate of mEPSCs was not different in first 100 ms after the end of 100 Hz trains between BCs of both genotypes (Figure 8A” inset).

In the EQ method (Figures 8D,D’ used previously at central auditory synapse in Taschenberger et al., 2002), absolute EPSC amplitude in response to a given stimulus, n (EPSC_n), in the train is plotted against the cumulative EPSC amplitudes of the stimuli prior to stimulus n (Cumulative amplitudes of EPSC₀ to EPSC_n−1). A line fit to the first 3–5 points is forward extrapolated to the x-axis and the x-intercept divided by the mEPSC amplitude estimates the RRP size, while P_r is determined as the slope of the linear fit. In contrast to the SMN method, the EQ method more robustly reported the reduced RRP size in Piccolo mutant endbulb synapses (Figures 8D,D’, Table 1), demonstrated by a significant reduction in RRP at both 100 and 200 Hz stimulation frequencies. As with the SMN method, the estimates of P_r were comparable between the two genotypes.

To further verify our finding of slower vesicle replenishment in PicMut endbulbs at 100 Hz suggested by SMN analysis, we studied the recovery from short-term depression, by measuring eEPSC amplitudes elicited by single stimuli presented at varying time intervals after a conditioning 100 Hz train of 50 pulses (Figure 9A). Recovery is displayed as the eEPSC amplitudes normalized to the amplitude of the first eEPSC of the conditioning train (Figure 9B). The time course of recovery was fitted with a double exponential function: f(x)=y0−A1e−Δtτ1−A2e−Δtτ2

A double exponential time course of recovery at physiological temperatures has been previously reported for calyceal synapses: endbulb of Held (Yang and Xu-Friedman, 2008) and calyx of Held (Kushmerick et al., 2006), with time constants comparable to our wildtype estimates. The amplitude and tau (τ) values for the fits were as follows: for PicMut A₁ = 0.24 ± 0.04, τ₁ = 101.9 ms ± 35.4; A₂ = 0.43 ± 0.03, τ₂ = 2.98 s ± 0.57 and PicWT A₁ = 0.07 ± 0.01, τ₁ = 30.2 ms ± 14.8; A₂ = 0.44 ± 0.01, τ₂ = 3.05 s ± 0.23. PicMut endbulbs showed a significantly slower recovery during the initial phase (Figure 9B) with longer time constant for the fast component of the double exponential fits. The kinetics of the slow components were comparable.

In addition to the mutation of Piccolo, we also studied the effect of disrupting (deleting exons 4 and 5, Bsn^ΔEx4/5, Altrock et al., 2003) one allele of Bassoon on vesicle replenishment rate. The mutant is henceforth referred to as PicBsn. Using the recovery paradigm described above, we estimated the kinetics of SV recovery after short-term depression by means of double exponential fits (Figure 10). The amplitude and tau (τ) values of the fit at PicBsn synapses were as follows: A₁ = 0.30 ± 0.04, τ₁ = 141.2 ms ± 50.7; A₂ = 0.45 ± 0.04, τ₂ = 4.03 s ± 1.05. In previous studies (Mendoza Schulz et al., 2014), heterozygosity for Bsn^ΔEx4/5 did not have any phenotype on synaptic transmission. However, in the absence of Piccolo, even partial loss of Bassoon seems to exacerbate the slowing of vesicle replenishment (Figure 10B).

The aggravation of slowed recovery from vesicle depletion upon Bassoon disruption in addition to Piccolo mutation, provided the impetus to investigate synaptic transmission in PicBsn mice. For a comprehensive analysis all three modes of release (Kaeser and Regehr, 2014): spontaneous, asynchronous and evoked, were studied, just like for PicMut mice. All recordings were made in the presence of 1 mM Kynurenic acid and 100 μM CTZ to eliminate postsynaptic factors influencing the inference.

mEPSC recorded as spontaneous events from postsynaptic BCs were comparable between PicWT (Figure 11A) and PicBsn (Figure 11A’). The analysis did not reveal any differences in the mEPSC amplitude, kinetics or frequency of events (p-value ≥ 0.05 for all three quantities, Figures 11B–E Student’s t-test, Figure 11F Wilcoxon rank sum test).

Interestingly, the reduced eEPSC amplitude phenotype observed in Piccolo-deficient synapses showed a “pseudo-rescue” upon additional deletion of a single Bassoon allele. eEPSC amplitudes were largely unaltered in PicBsn endbulbs (Figures 12A,B; 4.10 ± 0.25 nA for PicBsn vs. 4.43 ± 0.37 nA for PicWT, p-value ≥ 0.05, Student’s t-test). The kinetics of rise and decay (Figures 12C–E), and the synaptic delay (Figure 12F) remained unaltered; p-value ≥ 0.05, Figures 12C,D,F Wilcoxon’s rank sum test and Figure 12E Student’s t-test. Again, in contrast to PicMut, endbulbs of PicBsn mice demonstrated a lower PPR in response to pairs of stimuli presented at inter-stimulus intervals of 5 and 10 ms (Table 2).

A lower PPR is indicative of a higher release probability which was not seen in PicMut mice. To further analyze the effects of Bassoon disruption in addition to the absence of Piccolo, responses to high frequency train stimulations were studied (Figures 13A,A’) to assess the RRP size, release probability and pool dynamics. The evoked response was quantified using cumulative analysis as described above (Figures 13C–D’; Table 2). Undetected in the SMN analysis, the EQ analysis revealed a smaller RRP and an increased P_r in the PicBsn endbulbs as compared to PicWT (Table 2). Short-term depression was also stronger at endbulbs of PicBsn mice than in PicWT as determined by a smaller τ (time constant of decay of the response to train stimulation) and a reduced steady-state amplitude (EPSC_30–50/EPSC₁, significant only at 100 Hz; Figures 13B,B’, Table 2). This contrasts with the phenotype observed with Piccolo mutation alone, where no discernible change in short-term depression was noted. Endbulb of Held synapses in the PicBsn mice tended to have more asynchronous activity as compared to PicWT synapses, but this did not reach statistical significance (Figure 13A” inset).

Deletion of exon 14 of Pclo along with insertion of a neomycin resistance cassette drastically reduced Piccolo expression as observed by semi-quantitative analysis by immunofluorescence. The reduction of Piccolo immunofluorescence on average amounted approximately to 90%, which is close to what was found by Western blotting of whole brain homogenate (95%; Mukherjee et al., 2010). Therefore, manipulation of Piccolo employed at the endbulb synapse was incomplete and, hence, the analysis likely underestimated the role of the protein. This calls for a refined genetic deletion that targets all splice variants. However, this drawback of the currently used Pclo manipulation also offered an advantage: the C-terminus, affected by the mutation, is absent from its shorter variant, Piccolino (Regus-Leidig et al., 2013, 2014), the predominant Piccolo isoform at the ribbon synapses of the IHCs. The data illustrate that the Piccolino expression remains intact at the IHCs of PicMut. ABR recordings corroborated the notion of intact IHC synapses and indicated overall normal cochlear function. This provided a unique opportunity to selectively study the effects of Piccolo disruption at the endbulb without any bias from the preceding synapse that was confounding the previous analysis of Bassoon function at the endbulb synapse (Mendoza Schulz et al., 2014).

The gross morphology of the endbulb synapse was not affected by Piccolo disruption: the number of the auditory nerve fibers converging onto BCs and the number of AZs per endbulb remained unaltered. Since Piccolo interacts directly or indirectly with most other CAZ proteins (Wang et al., 2009), we also studied how the molecular composition of AZ is affected upon its disruption. Semi-quantitative expression analysis was employed by integrating immunofluorescence of excitatory and inhibitory AZs onto BCs. Since excitatory AZs were required to co-localize with Vglut1 that is thought to distinguish endbulb terminals from Vglut2-positive bouton endings (Heeringa et al., 2016), it is reasonable to consider the analysis of excitatory AZs to mostly report properties of endbulb AZs. Out of the four CAZ proteins (Bassoon, RIM1, RIM2 and Munc13-1) investigated, expression levels of Bassoon and RIM1 were significantly altered at endbulb AZs. It is interesting to note that the two proteins that were altered share regions of homology with Piccolo (Wang et al., 1999). RIM and Piccolo both exhibit Zn finger domains, a PDZ domain and two C-terminal C₂ domains and Bassoon shares ten highly conserved PBH domains with Piccolo (tom Dieck et al., 1998; Wang et al., 1999; Fenster et al., 2000). RIM1 immunofluorescence intensity was reduced while that of Bassoon was increased. It can be speculated that the reduction in RIM1 could be a consequence of it not being integrated properly at the synapse in the absence of Piccolo, as Piccolo is thought to play an integral role in AZ assembly and scaffolding (Gundelfinger et al., 2016). Bassoon, on the other hand, thought to have partially overlapping functions with Piccolo (Mukherjee et al., 2010; Waites et al., 2013), could be upregulated in PicMut as a compensatory mechanism. This is in line with a previous report (Mendoza Schulz et al., 2014), where Piccolo expression was upregulated in Bassoon deficient synapses, likely to compensate for the loss of some of Bassoon’s function.

So far, a unifying picture of the role of Piccolo in synaptic transmission has been missing. While one study implicated Piccolo as a negative regulator of SV exocytosis (Leal-Ortiz et al., 2008), others stated Piccolo plays no major role in SV release (Mukherjee et al., 2010) or, together with Bassoon, contributes to maintenance of the structural integrity of the AZ (Waites et al., 2013). In the present study of the endbulb of Held synapse, an amplitude reduction of eEPSCs evoked by single action potentials was observed. BCs are responsible for high fidelity transmission of timing information (Oertel, 1999). Hence, even this mild reduction (~20%) in eEPSC amplitudes at the level of individual synapses potentially seems to impair the synchronized activity in the cochlear nucleus as demonstrated by the reduction of ABR wave II amplitude.

Kinetics of release, on the other hand, were not changed either for eEPSCs or for mEPSCs. The unaltered mEPSCs amplitude (quantal size) and kinetics rule out post-synaptic changes in abundance or properties of AMPA receptors. This is supported by the unchanged spatial extent of the PSD in PicMut endbulb AZs, reported by the electron micrograph analysis of the synaptic ultrastructure. In contrast, global disruption of Bassoon increased the quantal size and significantly enlarged the PSD length (Mendoza Schulz et al., 2014). This was suggested to reflect homeostatic upscaling of excitatory synaptic contacts in response to reduced SGN input to BCs due to impaired synaptic sound encoding in the cochlea. Excluding postsynaptic alterations and changes in vesicular glutamate content based on the unaltered mEPSCs, leaves two possibilities, a reduction in either the release probability (P_r) or the RRP size to explain the reduced eEPSC amplitude. The EQ analysis of eEPSC trains (Elmqvist and Quastel, 1965) indicated a reduced RRP size as the primary reason for the reduced eEPSC amplitude. While, the SMN (Schneggenburger et al., 1999) method failed to disclose a significant reduction in RRP size, our notion was corroborated by a reduced synaptic vesicle pool at the AZ observed on an ultrastructural level using electron microscopy. P_r, on the other hand, seemed largely unaltered as estimated by both the analysis methods.

After cessation of the stimulus following depression, the recovery protocol revealed an impairment in SV recovery, where the fast phase of recovery was slowed by a factor of three. Piccolo’s contribution to vesicle replenishment could be attributed to its interaction with Actin regulatory protein like Daam1, required for F-actin assembly (Wagh et al., 2015), that may regulate conversion of reluctant SVs to the fast releasing RRP (Lee et al., 2012). Moreover, the reduction in RIM1 might contribute to the observed reduction in RRP, as RIMs play a crucial role in making vesicles fusion competent (Wang et al., 1997; Gracheva et al., 2008; Deng et al., 2011; Han et al., 2011; Fernández-Busnadiego et al., 2013; Kintscher et al., 2013; Jung et al., 2015) and hence in determining the RRP size (Han et al., 2011, 2015). With its extended reach into the cytosol, tens of nanometers beyond the dense projections of the CAZ (Limbach et al., 2011), Piccolo might escort SVs from further away in the presynaptic cytosol and deliver them to RIM to make them release ready. The reduction in RIM1 expression reported in this study could be in line with this synaptic interplay. Indeed, our electron microscopy analysis demonstrating a reduced pool of SVs at the Piccolo-deficient AZs lends support to this hypothesis.

Interestingly, a trend towards further slowing of the fast phase of recovery from short-term depression was observed upon disrupting one allele of Bassoon, in addition to the mutation of Piccolo. Nevertheless, finding reduced vesicle replenishment upon disruption of either Bassoon (Hallermann et al., 2010; Mendoza Schulz et al., 2014) or Piccolo suggests that neither protein alone is sufficient to maintain wildtype performance of SV replenishment, even despite a likely compensatory, up-regulation of the respective other gene.

This implies partially additive function, probably even at the same AZs as the majority of them contain both proteins (Dondzillo et al., 2010; Mendoza Schulz et al., 2014), although Piccolo and Bassoon may promote replenishment through different molecular pathways. As discussed above, Piccolo might employ an interplay of F-actin assembly, interaction with other CAZ proteins like RIM and its Ca²⁺ sensing domain. Bassoon on the other hand, might tether SVs via interactions with the SV associated protein Mover (Ahmed et al., 2013; Körber et al., 2015) and could also assist in positioning of Ca²⁺ channels (Davydova et al., 2014) and SVs close to each other (Hallermann et al., 2010).

Future work will be required to dissect the precise molecular mechanisms by which Piccolo acts in vesicle replenishment. This should also target roles of Piccolo in SV endocytosis and low-affinity Ca²⁺ sensing. Piccolo binds Actin associated proteins like Abp1 (Fenster et al., 2003), and GIT1 (Kim et al., 2003), both of which have been linked to SV endocytosis. As stated earlier, Piccolo plays a role in dynamic assembly of F-actin, which is needed for retrieval of SVs both by bulk (Nguyen et al., 2012) and ultrafast (Watanabe et al., 2013) endocytosis. The finding that only the fast component of recovery was affected would imply a rapid mechanism, such as ultrafast endocytosis or release-site clearance (Haucke et al., 2011), to be involved. The C₂A domain of Piccolo acts as a low- affinity Ca²⁺ sensor (Gerber et al., 2001; Garcia et al., 2004), which is ideal to sense Ca²⁺ build-up during high-frequency activity at the synapse. At the endbulb of Held, Ca²⁺ drives the fast component of recovery (Wang and Manis, 2008; Yang and Xu-Friedman, 2008) and Piccolo is an interesting candidate as the sensor driving this Ca²⁺ dependent rapid recovery.

To elucidate the synergistic function of Bassoon and Piccolo, physiology of Piccolo mutant mice with additional heterozygosity for a partial deletion of the Bassoon gene (PicBsn mice, Bsn: deletion of exons 4 and 5; Altrock et al., 2003), was further probed. The phenotype observed was intermediate between the changes in neurotransmission seen in Piccolo and Bassoon-deficient endbulbs (Mendoza Schulz et al., 2014). Bassoon heterozygosity did not exert any synaptic deficits by itself (Mendoza Schulz et al., 2014). However, at the endbulb of Held, in the absence of Piccolo, Bassoon heterozygotes demonstrate a reduced PPR indicative of an increased release probability, faster and greater extent of depression to train stimulation, and a tendency for higher asynchronous release upon cessation of high frequency stimulation.

Interestingly, the eESPC amplitude, which was reduced in Piccolo-deficient endbulbs, seemed comparable to the PicWT values in PicBsn mice demonstrating a “pseudo-rescue” of PicMut phenotype. In Bassoon KO, the eEPSC amplitude was comparable to that of the wildtype likely because of homeostatic upscaling of excitatory synaptic contacts in response to reduced SGN input to BCs due to impaired synaptic sound encoding in the cochlea (Mendoza Schulz et al., 2014). However, this is not valid for PicBsn mice, as the cochlea does not suffer from Piccolo deficiency and one allele of Bassoon seems sufficient to sustain neurotransmission at the cochlear ribbon synapses (Frank et al., 2010). This is reflected in the unaltered mEPSC amplitude, as opposed to an increase observed at endbulbs of Bassoon KO (Mendoza Schulz et al., 2014).

EQ analysis, but not SMN analysis, estimated a reduction in RRP. This discrepancy in the estimations of the two methods arises from the assumptions on which they are based. The SMN analysis estimates the pool size by the decrement in the vesicle pool during stimulation and is thus influenced by the extent of depression during the train (Neher, 2015). Hence, an increased depression (lower steady-state amplitude) at PicBsn endbulbs could overestimate RRP size giving values, comparable to that of the PicWT. EQ analysis on the other hand, estimates the RRP by forward-extrapolation, as it would have been before the onset of stimulation and is hence independent of the stimulation frequency used and the synaptic depression resulting therein (Neher, 2015). Considering the reduced RRP at the PicBsn synapses, we speculate that the “pseudo-rescue” of eEPSC amplitude is due to an increase in release probability as indicated by a lower PPR and higher P_r by the EQ cumulative analysis.

Taken together, partial lack of Bassoon in Piccolo-deficient synapses seems to add changes in the release probability to the phenotype observed in PicMut mice. It is interesting to speculate that Piccolo and Bassoon maintain synaptic vesicle clustering at the synapse (Mukherjee et al., 2010), but while they both have a collaborative role in vesicle replenishment, Bassoon might have a role in regulation of Ca²⁺-influx, SV release and/or their coupling as well. This hypothesis opens questions and a need for further investigations to probe the changes in interactions and compositions of the CAZ proteins and the active zone ultrastructure in the absence of two of its largest proteins.