Animal breeding and experiments were approved by the regional councils of Rhineland-Palatinate according to the German Animal Welfare Act (TSchG §4/3) and followed the guidelines for the protection and welfare of laboratory animals. C57BL/6 J mice were bred in the animal facilities of the University of Kaiserslautern-Landau, and both sexes were analyzed at postnatal days (P) 11 ± 1.

Coronal brainstem slices containing the MNTB and the LSO were prepared in accordance with the methodology previously described by Hirtz et al. (2012). Slices were stored at room temperature (RT) in carbonated artificial cerebrospinal fluid (ACSF), which consisted of the following components (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 2 sodium pyruvate, 3 myo–inositol, 0.44 L (+) ascorbic acid, 25 NaHCO₃, 10 glucose, 1 MgCl₂, 2 CaCl₂. The solution had an osmolality of 290 ± 10 mOsm/L and was bubbled with carbogen to achieve a pH of 7.4. Individual slices were transferred into the recording chamber on an upright microscope (Nikon Eclipse E60FN, Minato, Japan) equipped with infrared-differential interference contrast optics (Nikon 4x CFI Achromat, 0.1 ∞; 60x CFI Fluor W, 1.00 W ∞; Zeiss Fluar CFI 5x/0.25 ∞/0.17; Olympus LUMPlanFL N 60x/1.00 W ∞/0/FN26.5) and an Orca-05G camera (Hamamatsu, Hamamatsu City, Japan). Whole-cell recordings were performed at 35 ± 1°C (Slice Mini Chamber I equipped with temperature controller TC07, Luigs & Neumann, Ratingen, Germany). LSO principal neurons were identified by their fusiform somata, and this identification was subsequently confirmed through an examination of their electrophysiological properties (Sterenborg et al., 2010). Patch pipettes and stimulation pipettes (GB150F-8P with filament, Science Products, Hofheim am Taunus, Germany) were constructed with a P-1000 horizontal puller (Sutter Instruments, Novato, CA, United States). The input resistance of the patch pipettes ranged from 2.5 to 4.5 MΩ when filled with an internal solution containing the following components (in mM): 140 K-gluconate, 10 HEPES, 5 EGTA, 1 MgCl, 2 CaCl₂, 2 Na₂ATP, 0.3 Na₂GTP. The pH was 7.3 and the osmolality was 280 ± 10 mOsm/L. The recordings were conducted using an EPC10 amplifier and controlled with PatchMaster software (HEKA, Lambrecht/Pfalz, Germany). Recordings were sampled at 20 kHz and low-pass filtered at 2.9 kHz. The liquid junction potential (15.4 mV) was corrected online. Series resistance amounted up to 20 MΩ and compensated between 40 and 60%. If the input resistance changed by >30%, the recording was discarded. Stimulation pipettes (tip diameter ~10 μm), filled with ACSF and connected to a stimulus isolator (STG4002; MultiChannel Systems; Reutlingen; Germany), were placed either at the lateral edge of the MNTB (orthodromic stimulation) or at the medial edge of the LSO (antidromic stimulation). In both cases, 100-μs long electrical current pulses were applied to evoke inhibitory postsynaptic currents (eIPSCs) or APs, respectively. At a holding potential of −70 mV, eIPSCs were outward currents. Stimulus amplitudes were adjusted to values which yielded stable maximal eIPSCs during low-frequency stimulation (0.5 Hz to avoid synaptic depression). For APs, once the rheobase was reached, the stimulus amplitude was doubled. We employed three different stimulation protocols (Figure 1): Protocol 1 started with a 1-Hz|60-s stimulation period to establish a baseline value (BL, average eIPSC peak amplitude = 100%). Thereafter, MNTB-LSO synapses were electrically stimulated for 60-s at various frequencies in ascending order (Challenge period: Chal 1-4; 10, 50, 100, 200-Hz). Each challenge period was immediately followed by a 60-s period at 1-Hz (Recovery period: Recov1-4). A total of 21,900 stimuli were applied. A 6-s pause was introduced for data storing after BL and between the last stimulus of the recovery period and the 1st stimulus of the subsequent challenge period. Protocol 2 started with a 1-Hz|60-s BL, followed by a 4-min | 100-Hz stimulation (Chal 1-4). The challenge period was followed by a 180-s period at 1-Hz (Recov1-3). A total of 24,240 stimuli were applied. A 6-s pause was introduced for data storing after BL and between the last stimulus of the recovery period and the 1st stimulus of the subsequent challenge period. Protocol 3 consisted of a 1-Hz|60-s period to assess the BL, followed by 60-s challenging periods at 50-Hz and 100-Hz stimulation (Chal1-2). Each challenge period was followed by a 60-s period at 1-Hz to assess recovery from depression (Recov1-2). Thus, a total of 9,180 stimuli were applied. A 6-s pause was introduced for data storing after the BL period and between the last stimulus pulse of Recov1 and the 1st stimulus pulse of Chal2.

Lethally anesthetized mice (10% ketamine, 220 μg/g body weight plus 2% xylazine, 24 μg/g body weight, intraperitoneal) were transcardially perfused with 100 mM phosphate-buffered saline (PBS, pH 7.4) at RT, followed by ice-cold 4% paraformaldehyde (PFA) for 20 min (Ecoline VC-360 pump, IsmaTec, Chicago, United States). Brains were removed from the skull, postfixed for 1 h in 4% PFA (4°C) and stored in 30% sucrose-PBS at 6°C overnight. 30-μm-thick coronal brainstem slices were cut with a sliding microtome (HM 400 R, MICROM) and transferred to 15% sucrose-PBS, followed by three rinses in PBS (30 min, RT). For NHE6, a heat-induced epitope retrieval protocol was used prior to antibody treatment. For this, slices were immersed in 10 mM sodium citrate buffer (pH 6.0) plus 0.05% Tween 20 (Roth, Karlsruhe, Germany) and heated at 95°C for 20–40 min. They were then returned to RT, followed by three rinses in PBS (5 min, RT). Slides were applied free-floating at 4°C for 1 in blocking solution (0.3% Triton-X-100, 10% goat serum, 2% bovine serum albumin (BSA) in PBS). Antibodies against GlyT2 (1:200, host: mouse, Synaptic Systems, Göttingen, Germany) and NHE6 (1:200, host: rabbit, Sigma Aldrich, St. Louis, United States) were applied free-floating at 4°C overnight in carrier solution (0.3% Triton-X-100, 1% goat serum, 1% BSA in PBS), followed by three rinses in PBS (30 min, RT). Slices were then incubated in the dark for 2 h in carrier solution and secondary antibody (1,1,000; goat-anti-mouse, Alexa Fluor 488; goat-anti-rabbit, Alexa Fluor 647, Thermo Fisher Scientific, Waltham, United States), followed by three rinses in PBS (30 min, RT). Slices were mounted on gelatin-coated glass slides and covered with mounting medium containing 2.5% DABCO (1,4-diazabicyclo[2.2.2]octane; Sigma-Aldrich, St. Louis, MO, United States). Images were acquired with a TCS SP5 X confocal microscope equipped with an HCX PL APO 63 × oil objective (Leica Microsystems, Wetzlar, Germany). For quantitative analysis, intensity profiles at inhibitory synapses were analyzed using Fiji ImageJ 1.54 as described previously (Fischer et al., 2019). Line scans were drawn from the center of the soma across membrane-associated GlyT2 signals. Eight intensity profiles were obtained per cell. In total, two P11 animals, 16 cells and 128 intensity profiles were determined. The position of maximum GlyT2 signal was set to zero, with negative and positive scan positions representing intracellular and extracellular locations, respectively. A rolling ball background subtraction (radius 20) was applied to images in Fiji ImageJ 1.54 (Schindelin et al., 2012), followed by brightness adjustment. To assess colocalization of the immunosignals, Fiji’s Coloc2 plugin was used to determine the Pearson correlation coefficient and the Costes’ test (P_Costes), a measure of reliability (Costes et al., 2004). P_Costes = 1 indicates >95% certainty that colocalization exists.

Foot-to-peak eIPSC amplitudes were analyzed with a custom-written plugin (Dr. Alexander Fischer, University of Kaiserslautern-Landau) implemented in IGOR Pro 6.37 or 9.01 (WaveMetrics Inc., Lake Oswego, OR, United States). Negative eIPSC amplitudes were set to zero. Data were then normalized to BL (mean eIPSC peak amplitude = 100%; Galarreta and Hestrin, 1998).

Recovery behavior was calculated by three means: (a) RecovA: last 10 eIPSCs recovery vs. BL (= 100%); (b) RecovB: last 10 eIPSC recovery vs. last 10 eIPSC of previous challenge; (c) Fractional recovery (FR):

sIPSCs were analyzed during 1-Hz|60-s stimulus periods (BL and recovery periods) and manually detected using Clampfit 10.2.0.14 (Molecular Devices, CA, United States). Histograms of the sIPSCs peak amplitudes were generated and fitted with Gaussian curves. The quantal size (q) was determined by a Gaussian fit from 0 pA to the bin after the first local maximum (see Krächan et al., 2017; bin width 5 pA, in some cases 3 pA). The cumulative current evoked through complete release of the RRP (I_RRP) was assessed by forward extrapolation of a linear fit through eIPSC_1-5 obtained at 100 Hz (Elmqvist and Quastel, 1965). The number of SVs in the RRP (N_RRP) was determined as the ratio I_RRP/q. The quantal content (m) was obtained as eIPSC₁/q_Recov2. The release probability (Pv) was calculated as I_RRP/eIPSC₁. The transfer rate (TR) was evaluated as described (Müller et al.). Briefly, TR was assessed from the slope of a line fit through the last 10 s (s_50-60) of the cumulative eIPSC curve in a challenge period.

Statistical analysis was performed with GraphPad Prism 9 (GraphPad Software). Data are presented as mean ± SEM. For distribution analysis, the Kolmógorov–Smirnov test was used. Normally distributed data were compared using paired or unpaired two-tailed t-tests. If the distribution was not normal, the Wilcoxon rank-sum test or the Mann–Whitney U test was used. Details on the specific tests used are provided in the Supplementary Tables S1–S12. The post hoc tests were corrected using Bonferroni’s adjustments where applicable (α_adjusted = α/k, where α is the significance level and k the number of comparisons). Since k was maximally 2, the adjusted significance levels are as follows:

To corroborate the anecdotal evidence derived from the two example neurons shown in Figure 2, we proceeded to examine the cohort behavior (n = 11, 11, 9; Figure 3). The BL values were found to be similar between the Ctrl and Bafi10 cohort (mean: 1,023 vs. 914 pA or 100:89; Figures 3A1,A2,A4,A5). In contrast, Bafi30 values were significantly reduced, reaching only 35% of the Ctrl (354 pA; Figure 3A3). Furthermore, the time courses of depression and subsequent recovery to the four challenge frequencies were in accordance with those observed for the example neurons (Figures 3A1–A3; compare with Figure 2).

Subsequently, statistical analyses were conducted across the four 60-s challenge periods (eIPSC₁, s_0.9-1, s_9-10, s_50-60; Figures 3B2–E5; Supplementary Table S1). Only one of the 17 Ctrl_vs_Bafi10 comparisons yielded a statistically distinguishable result (Figures 3A4,B2–B4,C2–C4,D2–D4,E2–E4). In contrast, Bafi30 amplitudes were observed to be significantly lower than those of the Ctrl in 15 out of the 17 comparisons. For example, at s_50-60 of the 50-Hz challenge, Ctrl amplitudes were ~3-fold higher than Bafi30 amplitudes (267 vs. 88 pA, Figures 3C1,C5). With the harshest 200-Hz challenge, peak amplitudes declined markedly in each of the three cohorts (Figures 3E1–E5). By s_50-60, they were down to 29, 18 and 10 pA (equivalent to 2.8, 1.8 and 1.0%; Figure 3E5), indicating a near-complete depletion of transmitter release, particularly in the presence of the drug. This may explain the lack of statistical significance shown in panels E4 and E5. To further assess the near-complete collapse of synaptic transmission during the 200-Hz challenge in the two Bafi cohorts, the latest time at which the response amplitude reached ≥ 100 pA/pulse was determined. The values were 4.45, 3.58 and 0.05 s for Ctrl, Bafi10 and Bafi30, respectively (Figure 3E1). This further highlights the drastic difference observed in the Ctrl_vs_Bafi30 comparison, but not for Ctrl_vs_Bafi10 (numbers refer to 890, 716 and 10 stimulus events, respectively). To evaluate the resilience of the synapses, it is noteworthy that each cohort could maintain synaptic transmission for a considerably longer duration during the 100-Hz challenge. The last response exceeding 100 pA was observed at 59.27, 23.78 and 5.75 s for the Ctrl, Bafi10 and Bafi30 cohort, respectively (Figures 3D1–D5).

The sample results confirm that treating the brainstem slices with 2 μM Bafi for 10 min is insufficient to substantially reduce transmission at MNTB-LSO synapses. The postulated drug effect becomes evident only after a longer application time of ≥ 30 min. As a caveat it is also necessary to consider the possibility of a more severe stimulation history and/or rundown effects after this time. The issue is addressed below (see Supplementary Figures S1, S2A1–A3). In any case, the results obtained with Protocol 1 demonstrate that synaptic transmission persists for thousands of stimulus pulses, even in the Bafi30 scenario. It can thus be argued that the pharmacological blockade of V-ATPase, as conducted here, results in an incomplete cessation of SV (re)filling, thus leading to only partial synaptic fatigue. As demonstrated in the section “Recovery from depression is reduced but not abolished by Bafi treatment,” the resilience of the MNTB-LSO synapses is also evidenced by their ability to recover from depression, even in the presence of Bafi.

The above analysis included only some aspects of the synaptic performance, namely windows with a total duration of 11.1 s. To assess the transmission capacity during the entire 60 s for each challenge period, we determined the cumulative eIPSC amplitude (Müller et al., 2022). Each cohort showed the highest value at 50 Hz, indicating that this is the “best transmission frequency” (Figures 4A1,B1,C1,D1,E1,E3). Ctrl synapses generated a cumulative amplitude of 1,077 nA at this frequency, whereas Bafi10 and Bafi30 values were 899 and 343 nA, respectively (83 and 32% of Ctrl; Figures 4B1,B2). No statistically significant difference occurred in any of the four Ctrl_vs_Bafi10 comparisons, whereas significance was evident in each Ctrl_vs_Bafi30 comparison (Figures 4A2,B2,C2,D2; Supplementary Table S2). Notably, synapses treated with Bafi30 performed at ~1/3 of the Ctrl level at each challenge frequency.

To quantify neurotransmission during the steady-state phase of the challenge periods, and thus the continuous replenishment rate, we analyzed the transmission rate during the last 10 s of the challenge period (TR_50-60, [nA/s]). The results confirmed those obtained for the entire 60-s periods, namely a statistically significant difference for Ctrl_vs_Bafi30 at each frequency, but no difference for Ctrl_vs_Bafi10 (Figures 4A3,B3,C3,D3). For example, the Ctrl had a mean TR_50-60 of 17.3 nA/s at 50-Hz challenge, compared to Bafi10 and Bafi30 values of 14.6 and 5.7 nA/s, respectively (ratios: 100:84:33; Figure 4B3). Very similar ratios were obtained at the three other challenging frequencies (10 Hz: 100:82:32; 100 Hz: 100:88:34; 200 Hz: 100:60:32).

We also analyzed the overall transmission by determining the cumulative eIPSC amplitude over all four challenge periods, thereby assessing the response strength to 21,600 pulses. On average, the Ctrl cohort was able to elicit 2,842 nA (on average 132 pA/event), whereas the Bafi10 and Bafi30 values were reduced to 79 and 33%, respectively (2,256 and 927 nA; Figure 4E2). Again, the difference was statistically significant for Ctrl_vs_Bafi30, but not for Ctrl_vs_Bafi10.

Taken together, the results obtained from the analysis of overall transmission further support our conclusion that synaptic transmission at MNTB-LSO synapses is substantially reduced only after prolonged Bafi application (≥ 30 min).

Since the Bafi30 BL was ~3-fold lower than the Ctrl BL (354 vs. 1,023 pA; Figures 3A4,A5), we wondered whether this initial difference alone, present before the first challenge period, could explain the transmission differences described so far. To address this question, we set the BL of each neuron to 100% (BL = 1), allowing us to assess its relative performance. For the 10-Hz challenge, the mean final cumulative value was 381 for the Ctrl, very close to the Bafi10 and Bafi30 values of 348 and 344, respectively (Figure 4E3). Thus, the relative performance was remarkably similar between cohorts (100: 91: 90). Values for the 50-Hz challenge were 1,079, 1,009 and 854 BL (100: 94: 79), and confirming results were obtained at 100 Hz (847, 783 and 659 BL; 100: 92: 78) and 200 Hz (575, 419 and 449 BL; 100: 73: 78). None of the eight comparisons reached statistical significance (Figure 4E3). Finally, the normalized current integrals across 4-min challenging also showed statistically indistinguishable performances (Figure 4E4). Together, these results imply a weaker transmission strength with prolonged Bafi treatment (Bafi30), but similar transmission kinetics.

Effective SV (re-)filling is of great importance, not only for the robustness of neurotransmission during ongoing activation, but also for the recovery from synaptic depression. Accordingly, we examined the recovery behavior in the context of V-ATPase inhibition and compared it to the Ctrl situation (Figure 5; Protocol 1). After normalization of the eIPSC peak amplitudes of each neuron to its own baseline (BL = 100%), considerable overlap of the challenge-related traces was observed across cohorts, whereas the recovery traces showed greater divergence (Figure 5A1). In particular, the Bafi30 cohort exhibited lower values (Figures 5B1,C1,D1,E1).

Recovery was quantified in three ways (Figures 5A2,B2,C2,D2,E2; cf. Brill et al., 2019; Brill et al., 2021). The RecovA metric assesses the ability of the system to accurately return to the BL level. The Ctrls were able to do so after each challenge period (Figures 5B3,C3,D3,E3; end values 99, 103, 89 and 81%). Bafi10 was successful in three of four instances (Figures 5B3,C3,D3) and thus only slightly inferior. In contrast, Bafi30 was unsuccessful in three instances (Figures 5C3,D3,E3), implying replenishment impairment.

The RecovB metric focuses on the fold increase from the steady-state depression level (the last 10 challenge events), and thus does not consider the BL. For each cohort, the RecovB behavior was statistically significant (Figures 5B4,C4,D4,E4). The results demonstrate the ability of MNTB-LSO synapses to recover from depression, even after a ≥ 30-min treatment with the V-ATPase blocker.

We also determined the fractional recovery (FR). The Bafi30 cohort showed significantly lower FR values than the Ctrl at two frequencies (Figures 5C5,D5; Supplementary Table S3). In contrast, no statistically significant difference was observed between Bafi10 and Ctrl at any frequency (Figures 5B5–E5). Collectively, these findings provide further evidence that the effects of 2 μM Bafi were relatively subtle and became effective only after prolonged treatment (≥ 30 min) and challenge frequencies ≥ 50 Hz.

As a final step, we compared the time courses of recovery between cohorts. For this purpose, we calculated the weighted time constant τ_w after fitting the curves with a biexponential function. During recovery after the 10-Hz challenge, τ_w values were low, implying fast replenishment (Ctrl: 2.1, Bafi10: 2.1, Bafi30: 2.1 s; ratios 100:100:100). They were substantially higher for the recovery after 100-Hz (Ctrl: 23.7, Bafi10: 23.8, Bafi30: 34.4 s; ratios 100:100:145) and 200-Hz challenge (Ctrl: 35.2, Bafi10: 34.9, Bafi30: 39.5 s; ratios 100:99:112). The observation that the ratios differed less than 1.5-fold in each comparison suggests that blocking the V-ATPase activity did not affect the recovery kinetics.

Impaired (re-)filling of SVs upon V-ATPase blockade should result in a smaller quantal size (q). A complete absence of (re-)filling should even abolish the transmission completely. To test these hypotheses, we determined the peak amplitude of spontaneous IPSCs (sIPSCs) during the baseline and the four recovery periods after challenge (BL and Recov1-Recov4). Example current traces for the three cohorts during the BL and the Recov4 periods are shown in Figures 6A1–A6. For each neuron, q was determined from the Gaussian distribution of its sIPSC peak amplitudes (Figures 6B1–B6; see Krächan et al., 2017). In the Ctrl_vs_Bafi10 comparison, we found no significant change in q (Figure 6C3; Supplementary Table S4). In contrast, a significantly lower q value was observed in the Ctrl_vs_Bafi30 comparison (Figure 6C3). The q values averaged 26 pA in Ctrl, 27 pA in Bafi10 and 20 pA in Bafi30, corresponding to ratios of 100:104:77 (Figure 6C3). We consider the 23% reduction in Bafi30 to be moderate. The frequency of sIPSCs was determined across the BL and Recov1-4 periods. Values were 5.7, 5.4 and 3.4 events per second for the Ctrl, Bafi10 and Bafi30 cohorts, respectively (not shown). They correspond to ratios of 100:95:63 and provide further evidence that Bafi treatment did not abolish the (re-)filling process.

After determining the q values, we reanalyzed the TR_50-60 results (cf. Figures 4A3,B3,C3,D3) in order to obtain a vesicle-oriented value of the replenishment rate (SV fusion and SV replenishment are balanced in steady state). At the end of the 50-Hz challenge period, 642 SVs per second were replenished in the Ctrl. The corresponding values for Bafi10 and Bafi30 were 542 and 302 SVs/s, respectively. These results refer to ratios of 100:84:47, and similar ratios were obtained at the other challenge frequencies (10 Hz: 100:82:42; 100 Hz: 100:91:51; 200 Hz: 100:64:44).

Taken together, these results show a delayed and gradual appearance of incompletely filled SVs upon V-ATPase blockade. The moderate reductions in q and replenishment rate demonstrate that MNTB axon terminals still maintain a pool of readily releasable SVs whose glycine content is high enough to evoke a postsynaptic current of ~20 pA per SV, i.e., about 70% of the Ctrl level. Thus, refilling of SVs is maintained despite inhibition of V-ATPase activity.

To analyze synaptic parameters in more detail, we examined RRP, m and P_v. The forward extrapolation method was used to determine the RRP (Elmqvist and Quastel, 1965) and the 100-Hz challenge period was analyzed (Figure 6D1). Mean values for I_RRP were 11.3, 8.5 and 2.4 nA for Ctrl, Bafi10, and Bafi30, respectively (Figure 6D2). With the q values shown above, these numbers correspond to an N_RRP of 434, 389, and 125 SVs, respectively (100:90:29), thus demonstrating a 3.5-fold and statistically significant reduction in Bafi30.

We also determined the quantal content m for eIPSC₁ of the 100-Hz challenge period (m₁ = eIPSC₁/q; q was determined from the preceding Recov2 period). Again, the Ctrl_vs_Bafi30 comparison showed a statistically significant difference (47 vs. 21 SVs, i.e., ratio of 100:45; Figure 6D3). Finally, we observed a statistically significant difference for the release probability P_v in the Ctrl_vs_Bafi30 comparison (P_v = m₁/I_RRP). While P_v was 11% in the Ctrl, it was significantly increased to 16% in Bafi30 (~1.5-fold; Figure 6D4).

The 3.5-fold reduction in N_RRP was confirmed by means of a fluctuation analysis assuming a binomial release of SVs (not shown). We have recently used this analysis to evaluate statistical fluctuations of m at MNTB-LSO synapses (Vahdat et al., 2025) and found a P_v value of 93%, which is considerably higher than the value found in the present study.

Collectively, the reduction in q and m values (down to 70 and 45%, respectively; in combination down to 32%) underlies the >3-fold variation in eIPSC₁ values in the Ctrl_vs_Bafi30 comparison at the onset of the 100-Hz challenge (cf. Figure 3D2).

The above results demonstrated diminished yet persistent activity at MNTB-LSO synapses even after prolonged treatment with 2 μM Bafi (≥30 min; Figures 2–5). We thus sought to determine whether a higher Bafi concentration (5 μM) and more intense stimulation conditions would result in a greater reduction in neurotransmission, or even its complete abolition. The underlying procedure, designated “Protocol 2” or “half-marathon” (Figure 1C) comprised a Ctrl episode (4-min|100-Hz challenge followed by 3-min|1-Hz recovery, totaling 24,180 stimulus pulses) which was followed by a Bafi10 episode (paired recordings) and a final Bafi30 episode (from a different neuron). The stimulus conditions for each Bafi episode were identical to those employed for the Ctrl episode. The results are summarized in Figure 7, in which panel A1 illustrates the time course of the eIPSC amplitudes for the three cohorts. From the third challenge period onwards, that is not before 12,000 – 18,000 stimulus pulses, a significant reduction in steady-state eIPSC amplitudes was observed in the Bafi10 cohort (Figure 7A2, s_170-180). This finding differs from the results obtained with Protocol 1 (see Figure 3). By the conclusion of the 4-min challenge period (s_230-240), the steady-state amplitudes had decreased to 51, 25, and 17 pA (Figures 7A1,A2). The ~2-fold lower performance in Bafi10 and the 3-fold lower performance in Bafi30, in both cases statistically proven, unambiguously demonstrate the presence of a drug effect.

A more detailed examination of the time course of eIPSCs revealed that the largest amplitude differences occurred during the initial 20 s between the Ctrl and Bafi10 cohorts on one hand and the Bafi30 cohort on the other (left inset in Figure 7A1). This suggests that Bafi30 may experience significantly accelerated synaptic fatigue, whereas Ctrl and Bafi10 can draw from an intact RRP. However, the kinetics of depression remain largely indistinguishable (right inset in Figure 7A1).

Analysis of the cumulative eIPSC amplitudes corroborated the observation that a higher drug concentration and a more severe stimulation condition yielded had a greater effect on synaptic transmission (Figure 7B1). The Ctrl cohort was observed to elicit 1,567 nA, whereas the Bafi10 and the Bafi30 cohorts achieved only 66 and 33%, respectively (1,027 and 524 nA). For the Ctrl_vs_Bafi30 comparison, a statistically significant difference was observed (Figure 7B2; Supplementary Table S5). The results indicate that the inhibition of V-ATPase considerably impacts SV (re)filling when three conditions are met: first, a highly concentrated solution of Bafi is administered; second, the application time is considerable; third, the stimulation conditions are harsh. The results for the TRs_230-240 corroborate this conclusion (Figure 7B3; Ctrl: 6; Bafi10: 3; Bafi30: 2 nA/s; ratio 100:50:33).

Our subsequent investigation focused on the recovery behavior following the half-marathon challenge, with particular attention paid to s_230-240 (Figures 7C1–C4). Regarding RecovA, the Ctrl cohort exhibited complete recovery, reaching 107% of BL. In contrast, the Bafi10 and Bafi30 cohorts showed incomplete recovery, reaching only 63 and 68% of the BL level, respectively, and these values were significantly different from the BL (Figure 7C2). The ability of each cohort to reach higher eIPSC amplitudes after exhibiting very low levels at the end of the 4-min|100-Hz challenge was evaluated by assessing RecovB. The Ctrl and Bafi10 cohorts exhibited significantly higher amplitudes (24- and 23-fold, respectively; Figure 7C3, see also panel A2, s_230-240). In contrast, RecovB did not reach statistical significance for Bafi30 (16-fold). Finally, there was a significant difference in FR between the Ctrl and Bafi30 cohorts, but not between the Ctrl and the Bafi10 cohorts (Figure 7C4; Ctrl: 107%; Bafi10: 61%; Bafi30: 46%).

During the harsh 4-min|100-Hz challenge, we made an intriguing observation following the 6-s pauses required for data storage at minutes 1, 2, and 3, namely short-term recovery (Figure 7A1; NB: a similar observation was made by Qiu et al. (2015) at Calyx-of Held – MNTB synapses). As the stimulation was terminated during these three pauses, we were able to assess and quantify this short-term recovery without interference. To achieve this, we applied a variant of the FR formula (eIPSC1of Chal2−mean amplitudes50−60of Chal1BL−mean amplitudes50−60of Chal1 etc.) and found lower values in the Bafi cohorts compared to Ctrl, but still ongoing recovery (Figure 7A1). The Ctrl: Bafi10: Bafi30 ratios were: min₁: 100:95:89; min₂: 100:71:64; min₃: 100:59:48. This implies that the presence of Bafi did not abolish SV replenishment, even when a brief time domain of just a few seconds and the harsh half-marathon conditions are considered. The results confirm those obtained for the recovery periods lasting 60 s.

Taken together, the results from the half-marathon experiments show that a 10-min application time of Bafi at an extraordinarily high concentration (5 μM) is sufficient to elicit blocking effects on synaptic transmission. However, the effects do not become evident before at least 12,000 stimulus pulses are given within ~2 min. Moreover, they still enable eIPSC amplitudes of sufficient height (~50% of Ctrl) during sustained activation, resulting in statistically indistinguishable cumulative eIPSC amplitudes. Recovery from depression is impaired but still reaches about two third of the BL level, implying ongoing SV (re)filling. As expected, the Bafi30 effects are generally more severe.

One potential pitfall of the experiments discussed above is that the observed decrease in eIPSC amplitudes cannot be unambiguously attributed to a drug effect. Rather, the decrease may be time-dependent. To address this concern, we performed sham control experiments essentially identical to Protocol 1 (see Figure 3), except that only the solvent for Bafi was washed in (0.1% EtOH). We observed an almost perfect superposition of the current traces (Supplementary Figures S1A1–A3,B1,C1,D1,E1; Supplementary Table S6). In essence, these results impressively demonstrate that there was no time dependence. In both EtOH10 and EtOH30, the synaptic performance was statistically indistinguishable from Ctrl (in ACSF) for BL analysis (Supplementary Figures S1A4,A5), for challenge periods (Supplementary Figures S1B1,B2,C1,C2,D1,D2,E1,E2), and for robust recovery (Supplementary Figures S1F1–F3,G1–G3,H1–H3,I1–I3).

Taken together, the data show no effect of EtOH on synaptic transmission, suggesting that the impairment is solely due to the drug (Bafi) and not due to the solvent or the time.

We also compared EtOH effects with Bafi effects (Supplementary Figure S2; Supplementary Table S7; see also Figure 3) and found significantly impaired transmission in Bafi30_vs_EtOH30 (Supplementary Figures S2C1–C3) but neither in Ctrl_Bafi_vs_Ctrl_EtOH (Supplementary Figures S2A1–A3) nor in Bafi10_vs_EtOH10 (Supplementary Figures S2B1–B3). Impaired transmission was evident during the last 10 s at each challenge frequency (Supplementary Figure S2C2) and during each recovery period (Supplementary Figure S2C3). Collectively, the results from the sham control experiments confirm that the reduced synaptic strength observed in the presence of bafilomycin is caused by V-ATPase inhibition. Time-dependent or solvent-dependent effects can be excluded.

In a complementary set of experiments, we assessed the effects of inhibited V-ATPase activity by repeating the experiments with Protocol 1, except that Bafi was replaced by folimycin (Foli, 1 μM). The results are shown in Supplementary Figures S3, S4. Foli essentially recapitulated the effects of Bafi in direction and magnitude (Supplementary Tables S8, S9). For 86% (67/78) of the parameters that we quantified in both the Bafi and Foli cohorts, the statistics were congruent, thus demonstrating that both drugs affected V-ATPase activity in a very similar manner (Supplementary Table S9). For example, the BL was significantly reduced in Foli30 compared to the Ctrl (49%; 633 vs. 1,321 pA), but not in the Foli10 (87%; 1,151 pA; Supplementary Figures S3A4,A5). A consistent observation was made for the Bafi cohort, namely 38 and 89%, respectively (Supplementary Table S9; cf. Figures 3A4,A5). Furthermore, the Foli30 cohort showed increased synaptic depression compared to the Ctrl at each challenge frequency and in each time window (16 parameters), whereas the Foli10 cohort was statistically indistinguishable from the Ctrl for each of those 16 parameters (Supplementary Figures S3B1–B5,C1–C5,D1–D5,E1–E5). A similar scenario had been found for Bafi (Supplementary Table S9; cf. Figure 3). Finally, application of Foli did not result in significantly lower q values in Foli10, but it did so in Foli30 (Supplementary Figure S4E3), consistent with the previous findings on Bafi.

Treatment with Foli30 also resulted in impaired recovery after each challenge period (Supplementary Figures S4A1,A2,B1,B2,C1,C2,D1,D2; Supplementary Table S10). For Foli10, however, such an impairment was only detectable after a 200-Hz exposure (Supplementary Figures S4A2,B2,C2,D2). RecovB reached statistical significance in each case (Supplementary Figures S4A3,B3,C3,D3), suggesting that refilling mechanisms are still functional, albeit significantly reduced. FR values showed a significant difference in each comparison of Ctrl_vs_Foli30 (Supplementary Figures S4A4,B4,C4,D4).

The I_RRP of Foli10 was statistically indistinguishable from the Ctrl, but it was drastically reduced in Ctrl_vs_Foli30 (~4-fold; Supplementary Figures S4E1,E2). The N_RRP in the Ctrl, Foli10, and Foli30 contained 832, 620 and 374 SVs, respectively (Supplementary Figure S4E2). Thus, the N_RRP differed only 2.2-fold between Ctrl and Foli30. The lower N_RRP ratio compared to I_RRP is explained by the finding that q values were almost 2-fold lower in Foli30 than in Ctrl (13 vs. 23 pA; Supplementary Figure S4E3). Notably, this reduction was greater than that shown for Bafi30 in Figure 6C3. The Foli30 cohort also showed a statistically significant reduction in m, similar to the Bafi30 results (64 SVs/38 SVs = 1.7-fold; Supplementary Figure S4E4). In contrast, P_v values were not altered by Foli (10, 11, 9%; Supplementary Figure S4E5; different from Bafi30).

The similar results obtained with Bafi or Foli, when considered together, underscore the need for prolonged drug treatment to impair synaptic transmission. Statistically significant differences occurred in 39 out of 46 Ctrl_vs_Bafi30 comparisons and in 39 out of 39 Ctrl_vs_Foli30 comparisons, representing 85 and 97%, respectively (Supplementary Table S9). In contrast, with shorter drug treatment, only 3 out of 46 comparisons in Ctrl_vs_Bafi10 and 0 out of 39 in Ctrl_vs_Foli10 showed statistical significance (7 and 0%, respectively). These results impressively highlight the delayed onset of impaired transmission in the presence of V-ATPase blockers. In conclusion, the effects of Foli essentially recapitulated those of Bafi in direction and magnitude. They provide consistent evidence that V-ATPase activity is necessary to generate glycine-filled SVs and ensure robust synaptic transmission at MNTB-LSO synapses. However, SV (re)filling does not appear to depend on V-ATPase activity alone.

Because our Bafi and Foli analyses had revealed an only incomplete blockade of transmission, despite severe treatment conditions, we wondered whether there might be another proton pump besides the V-ATPase that could acidify the SVs, thereby creating a favorable proton gradient for VIAAT-mediated glycine import. One candidate is the Na⁺/H⁺ exchanger (NHE; see Figure 1C). Thirteen isoforms have been identified (Pedersen and Counillon, 2019), of which NHE6-NHE9 are located in the membranes of intracellular organelles (Alexander et al., 2017).

In order to assess the distribution of NHE proteins in the LSO and other superior olivary complex (SOC) nuclei, we used immunohistochemistry, thereby focusing on the NHE6 isoform. Our approach was motivated by mRNA sequencing results obtained in our group from laser-dissected tissues (Maraslioglu-Sperber, 2021). These results demonstrated high expression of Slc9A6, the gene encoding NHE6, in the LSO and MNTB of juvenile mice, among nine NHE isoforms analyzed. Slc9A6 expression was the most pronounced, with a level of 11 transcripts per million in the LSO and 12 in the MNTB. Interestingly, NHE6 belongs to the subset of Na⁺/H⁺ exchangers that are found in recycling endosomes (Brett et al., 2002; Masereel et al., 2003; Orlowski and Grinstein, 2004; Nakamura et al., 2005) and secretory organelles (Miyazaki et al., 2001; Kondapalli et al., 2014). Consistent with this, NHE6 has been identified in SVs (Preobraschenski et al., 2014; Taoufiq et al., 2020). Finally, NHE6 has been implicated in luminal pH regulation (Ohgaki et al., 2010; Ohgaki et al., 2011). Taken together, these findings identified NHE6 as an interesting candidate for our MNTB-LSO study.

To assign NHE6 signals to glycinergic neurons, our immunohistochemical analysis included a double-labeling strategy for NHE6 and GlyT2, which is a marker protein for glycinergic neurons (Friauf et al., 1999; Eulenburg et al., 2005). We found strong NHE6 immunofluorescence in the MNTB, the SPN and the LSO (Figure 8A1). At the cellular level, the cytoplasm of virtually every glycinergic MNTB soma was NHE6 positive (Figures 8A2,A3). This pattern stood in contrast to that observed in the LSO, where neuronal somata exhibited almost complete absence of NHE immunoreactivity, yet were decorated with double-labeled puncta in a perisomatic fashion (Figures 8A4–A6). We attribute these puncta to axon terminals of MNTB neurons. Quantification of NHE6 and GlyT2 co-localization confirmed the close proximity of the immunosignals (Figures 8B1–B3). Taken together, these results suggest that terminals of MNTB axons that contact LSO somata contain substantial amounts of NHE6.

No selective antagonist for NHE6 has been identified to our knowledge. Nevertheless, NHE-mediated Na⁺/H⁺ exchange in general is reportedly suppressed by the “broadband” antagonist ethyl-isopropyl-amiloride (EIPA; Vigne et al., 1983; Masereel et al., 2003; Hori and Takamori, 2021). We therefore characterized transmission at MNTB-LSO synapses in the presence of EIPA (100 μM). The procedure is illustrated in Figure 1C and outlined in Protocol 3. It consists of a Ctrl episode performed in ACSF and comprised a 1-min|50-Hz challenge, a 1-min|100-Hz challenge and two 1-min|1-Hz recovery periods, corresponding to 9,120 stimulus pulses in total. Subsequently, an EIPA10 episode was initiated that was identical in structure to the preceding Ctrl episode (paired recordings).

In the Ctrl, the synaptic depression during each challenge period, along with the subsequent recovery, occurred in a manner consistent with our expectations (Figure 9A1). However, we made an unanticipated observation during the 10-min wash-in period of EIPA, during which MNTB axons were stimulated at a low frequency of 0.5 Hz. The eIPSC amplitudes decreased from 1,619 pA at the end Ctrl-Recov2 to 799 pA at the beginning of EIPA-BL (Figures 9A1,A2). This resulted in a markedly reduced BL of 55% (Figures 9A3,A4), suggesting an initial inhibitory effect of EIPA on transmission that cannot be attributed to challenge-induced SV depletion. During each challenge period, the transmission in EIPA10 exhibited a ~ 4-fold reduction compared to the Ctrl (Figures 9B1,B2,D1,D2; Supplementary Table S11). RecovA showed no statistically significant difference from the preceding BL in each cohort following 50-Hz challenge (Figures 9C1,C2), indicating that the efficacy of recovery remained unchanged. However, after 100-Hz challenge, RecovA was insufficient (Figures 9E1,E2). Furthermore, the FR was significantly lower than that for the Ctrl after each challenge period (Figures 9C3,E3; 50 Hz: 119 vs. 68%, 100 Hz: 106 vs. 56%), suggesting that EIPA may indeed inhibit SV (re)filling. The observation that the synaptic strength exceeded 500 pA during each recovery period, however, indicates that SV (re)filling is not completely abolished by the drug (Figures 9C1,E1).

As before, we determined the readily releasable pool (I_RRP) at the begin of the 100-Hz challenge period (Figure 9F1). We found a severe reduction in the presence of EIPA10 (~4-fold, from 30.9 to 7.3 nA; Figure 9F2). These I_RRP values relate to N_RRP values of 930 and 254 SVs, respectively. The quantal size q remained unaltered (Figure 9F3), while the quantal content m was almost 3-fold lower (Figure 9F4), in line with the diminished N_RRP. Finally, the release probability Pv was found to be unaltered (Figure 9F5). In summary, the EIPA10 experiments yielded some surprising results, including the extraordinary time course of eIPSC amplitudes during the drug wash-in period (Figure 9A2), which ultimately resulted in an almost 2-fold lower BL compared to the Ctrl (Figure 9A3). Therefore, a word of caution is warranted regarding the interpretation of the EIPA10 results. This issue will be further discussed in the following section.

While performing the EIPA experiments, we noticed that the drug affected the AP behavior of LSO neurons. In the absence of EIPA, Ctrl neurons fired an onset AP in response to a 200-ms depolarizing current pulse, which is typical for LSO principal neurons (Sterenborg et al., 2010; Friauf et al., 2019). This onset AP was not elicited in the presence of EIPA (Figure 10A1), where each of the eight neurons analyzed exhibited a complete cessation in firing (Figure 10A2). This unanticipated outcome prompted us to be concerned about the activation efficiency of the MNTB neurons in the synaptic stimulation experiments. To address this concern, we performed antidromic stimulation experiments on MNTB neurons as previously described (Müller et al., 2022). The stimulation regime was according to Protocol 3, and APs with peak amplitudes <1 mV were considered as failures. The Ctrl MNTB neurons showed mild failure behavior at the end of the 50-Hz and the 100-Hz challenge period, but the amplitudes of their APs remained largely unaltered (example traces in Figure 10B1 and the 60-s time courses for the cohort in Figure 10C1). In contrast, EIPA10 MNTB neurons exhibited a high failure rate and a drastic reduction in AP peak amplitude, resulting in the emergence of mini spikes (spikelets) whose amplitude was only ~3 mV (Figure 10B2; time course in panels C2 and D). The AP fidelity of the Ctrl during s_0-10 was 99 and 89% at 50-Hz and 100-Hz challenge, respectively (Figure 10C1). The corresponding values for EIPA10 were only 45 and 20% (Figure 10C2). By s_50-60, the fidelity values were 86 and 74% for the Ctrls and 24 and 11% for EIPA10 (Figure 10E). The Ctrls demonstrated the ability to regain perfect fidelity values of 100% during each recovery period. In contrast, the corresponding values for EIPA10 were only 60 and 49% (Figure 10C2). As mentioned above, the AP fidelity was only 45% during s_0-10 of the 50-Hz challenge (Figures 10B2,C2). It is noteworthy that a similarly low level (56%) was already present during the BL period, and it emerged during the 2nd half of the 10-min wash-in period of EIPA at 0.5-Hz stimulation (Figure 10C2). In sum, the results illustrate considerable and deleterious effects of EIPA10 on AP generation and/or AP conduction.

In a next step, we conducted antidromic stimulation experiments in the presence of the V-ATPase inhibitor Bafi (2 μM) to ascertain whether the drug might also exert any deleterious effects on AP generation, as we had demonstrated for EIPA. Protocol 1 was employed to emulate the synaptic stimulation scenario previously utilized for Bafi (cf. Figures 1, 2). Detailed results are presented in Figures 10F1–F3,G1–G4 and in Supplementary Table S12. For the Ctrl, the s_50-60 fidelity values were 100, 84, 70 and 33%, respectively, at challenge frequencies of 10, 50, 100 and 200 Hz. The corresponding values for Bafi10 were 89, 82, 61 and 35% (Figures 10G1–G4). None of these values was statistically different from the Ctrl. Finally, the Bafi30 values were 88, 66, 44 and 19%, respectively, and again, no value was statistically different from the Ctrl. Furthermore, the recovery from depression was robust in each case, ranging from 86 to 100% (Figure 10F3). Collectively, these findings imply that Bafi elicits no considerable aversive side effects on AP behavior, in clear contrast to EIPA.

To consider the influence of AP failures on the overall transmission efficiency of MNTB-LSO synapses, we performed a mathematical analysis. This analysis aimed to separate effects on AP generation and/or AP conduction from those caused by drug-induced synaptic fatigue. For this purpose, mean eIPSC values from the challenge periods were corrected by assuming 100% AP fidelity. This allows to conservatively attribute the changes in eIPSC amplitude to changes in the presynaptic release machinery. The following example illustrates the procedure: If the AP fidelity rate was 40% at a given time point and the corresponding eIPSC amplitude was 210 pA, the latter value was multiplied by 2.5, thereby obtaining a “failure-corrected” value of 525 pA. The procedure was employed to correct the EIPA10, Bafi10 and Bafi30 means presented in Figures 3, 9, respectively. The results from 50- and 100-Hz challenge are shown in Figure 11 and Supplementary Table S13 provides a detailed quantification of the findings for the s_50-60 periods. Two arbitrary thresholds were set (50 and 200%), and they revealed no effect on synaptic transmission in the Bafi10 experiments upon failure corrections, thus confirming the aforementioned results (Figures 11A1,A2,C1,C2). In contrast, synaptic transmission was affected in Bafi30 upon correction, as described before. However, there was no complete abolishment, because eIPSC amplitudes were only moderately decreased, as documented by drugCtrl ratios in the range of 42-46%. For the EIPA experiments, our failure-correction revealed that this drug appears to leave SV (re-)filling unaffected (Figures 11B1,B2,D1,D2). An unchanged q value upon EIPA treatment (Figure 9F3) provides further evidence that EIPA does not appear to affect SV (re)filling.

The main findings of the present study are as follows: (1) In the absence of any V-ATPase blocker, synaptic transmission at inhibitory MNTB-LSO synapses is robust and remarkably resistant to fatigue, even under severe high-frequency stimulation conditions in the minute range and beyond (the half-marathon experiments included 24,000 stimulus pulses in 4 min; Figure 7). Furthermore, recovery from synaptic depression is rapid and complete, demonstrating great resilience. (2) Interfering with SV refilling by pharmacologically blocking V-ATPase activity with Bafi or Foli reduces steady-state eIPSCs only after prolonged drug application (≥30 min; Figures 3, 4; Supplementary Figures S3, S4). (3) However, the amplitude reduction is only moderate, so that transmission remains at a level of ~30%. (4) In line with the amplitude reduction, the quantal size q is reduced, but only to ~70% and only after prolonged blocker use (Figure 6; Supplementary Figure S4E3). (5) The ability to recover from synaptic depression is maintained in the presence of any V-ATPase blocker (Figure 5; Supplementary Figure S4). (6) Even when the drug concentration is increased and the stimulus conditions are most severe (half-marathon), Bafi10 effects are only evident after > 12,000 stimulus pulses (Figure 7). This further demonstrates the resistance of MNTB-LSO synapses to pharmacological treatment. (7) Application of EIPA to block NHE activity greatly reduces AP fidelity and thus affects neurotransmission (Figure 9). However, after correction of the AP failures, the hypothesis that EIPA affects SV refilling cannot be maintained (Figures 10, 11). Collectively, these major results suggest that effective SV refilling at MNTB-LSO synapses, especially under harsh stimulus conditions, does not depend solely on V-ATPase activity. Rather, another mechanism likely coexists that provides an H⁺ source for VIAAT and thus helps to generate SVs with a sufficiently high glycine content.

We argue that an incomplete pharmacological block of V-ATPase activity does not satisfactorily explain the persistence of transmission at MNTB-LSO synapses. Persistence was evident from substantial steady-state amplitudes, preserved quantal size, and recovery from depression. Notably, Bafi had to be applied for >30 min before its effect reached statistical significance (Figures 3–6; Protocol 1). Furthermore, relatively high concentrations of Bafi were required to elicit an effect (2 μM in Protocol 1; 5 μM in Protocol 2). Similar results were obtained with Foli (1 μM in Protocol 1; Supplementary Figure S3). The available literature supports this view. In vitro studies on dissociated cells, including primary neuron cultures, have used Bafi concentrations ranging from 0.25–4 μM (Granseth and Lagnado, 2008; Li et al., 2011; Peng et al., 2012; Ratnayaka et al., 2012; Santos et al., 2013; Siksou et al., 2013; Vargas et al., 2014; Di Giovanni and Sheng, 2015; Tang et al., 2015; Tagliatti et al., 2020). Comparable concentrations were used in organotypic culture studies (0.2–2 μM; Rose et al., 2013; Allain et al., 2016), purified synaptosomes (1 μM; Geerlings et al., 2001), acute brain slices (1–6 μM; Zhou et al., 2000; Harrison and Jahr, 2003; Satake et al., 2012; Muller et al., 2013; Song et al., 2013), an ex vivo optic nerve preparation (4 μM; Micu et al., 2016), and whole-larvae Zebra fish neuroblasts (1 μM; Einhorn et al., 2012). These concentrations are ~1,000 times higher than the K_i values for Bafi, which lie in the (sub-)nanomolar range (reviewed in Dröse and Altendorf, 1997). In most of the 20 cited studies, pretreatment with Bafi lasted only 30 s or 1–10 min (when specified). The five acute brain slice studies reported longer incubation times of 1 to ≥2.5 h. Thus, our drug application protocols appear reasonably consistent with established experimental parameters. Further support for complete V-ATPase block by Bafi comes from preliminary results demonstrating a gradual decline of eIPSC amplitudes when MNTB-LSO synapses are continuously stimulated at 1 Hz for ~40 min. These results predict an eventual abolishment of transmission and complete exhaustion of the releasable SV pool after ~51 min (see Supplementary Information).

A comparison of our drug-related results with those obtained in other synaptic systems reveals both similarities and striking differences (Table 1). Regarding the effects of V-ATPase inhibition on spontaneously released SVs, only one study reported a complete loss of events (Song et al., 2013), whereas most found moderately reduced q values. This suggests that across synaptic systems, transmitter-filled SVs remain available when V-ATPase activity is inhibited. Notably, drug-treated MNTB-LSO synapses displayed much smaller effects on sPSC frequency than other synapse types.

Far fewer studies have investigated evoked transmission during sustained high-frequency stimulation. At calyx of Held-MNTB synapses, V-ATPase inhibition did not completely abolish responses but reduced them to 10% of control during intense stimulation at 20 Hz for 400 s (Qiu et al., 2015). Thus, it appears that these auditory synapses also exhibit partial resistance to V-ATPase blockade, though less so than MNTB-LSO synapses, where steady-state transmission remained at ~25%. Taken together, we conclude that SV replenishment relies on both common and system-specific mechanisms. The persistence of transmission at both auditory synapse types, even under strong stimulation, points to an additional mechanism that ensures reliable SV (re-)filling at these high-fidelity synapses.

During Protocol 1, we delivered 21,600 stimulus pulses. In the absence of Bafi, MNTB-LSO synapses reached a cumulative eIPSC amplitude of 2,842 nA (Figure 4E2). Combined with a mean q value of 26 pA (Figure 6C3), this corresponds to ~109,000 released SVs. Given that N_RRP comprised 434 SVs (Figure 6D2), the RRP was turned over ~250 times. Analysis of the Foli experiments yielded similar results, with ~190,000 released SVs and ~230 RRP turnovers. Sustaining neurotransmission at such levels requires highly efficient replenishment mechanisms, with candidate resources being the recycling pool and the reserve pool. The recycling pool is generally thought to contain ~5–20% of all SVs, whereas the reserve pool forms the largest reservoir (~80–90%; Rizzoli and Betz, 2005). The RRP typically accounts for only ~1–2%. For MNTB-LSO synapses, specific pool sizes have not been determined. However, estimates from calyx of Held-MNTB synapses suggest ~180,000 SVs in the reserve pool, ~7,000 in the recycling pool, and ~3,000 in the RRP (95, 3.5, and 1.5% of the total pool, respectively; Rizzoli and Betz, 2005). If a similar distribution applies to MNTB-LSO synapses, the total SV pool would contain only ~29,000 SVs. Consequently, the release of 109,000 or more SVs cannot be explained by consumption of a pre-existing reserve. Instead, replenishment must exceed the size of the total pool nearly 4-fold during the four minutes. In this context, it is noteworthy that the entire recycling pool of hippocampal synapses is exocytosed within ~20 s during 10-Hz stimulation (200 stimulus pulses; Fernández-Alfonso and Ryan, 2004), and ~40% of the total SV pool participates in recycling (Fernández-Alfonso and Ryan, 2006).

In our Bafi and Foli experiments, we aimed to block V-ATPase to prevent the reacidification of SVs in the recycling pool. In this scenario, endocytosed SVs remain trapped in an alkaline state, thereby preventing neurotransmitter reloading and rendering them unable to replenishment the RRP (Sankaranarayanan and Ryan, 2001). Results from the Bafi30 experiments showed a cumulative eIPSC amplitude of 927 nA, 33% of the Ctrl value (Figure 4E2). With a q value of 19 pA and an N_RRP of 125 SVs (Figures 6C1–C5,D2), this corresponds to ~49,000 released SVs and a 392-fold RRP turnover (compared to a 250-fold turnover in the Ctrl). Again, we conclude that such robust replenishment cannot be explained by recruitment from the reserve pool alone, nor by de novo synthesis. Transport of SVs from the soma to the axon terminal would occur on a time scale of hours, far too slow to sustain transmission under acute synaptic stress. In the following section, we will discuss an alternative scenario.

We propose that SVs are reacidified by additional H⁺ transporters besides the V-ATPases, and we suggest that NHEs are strong candidates (Figure 12). Slc9a6, the gene encoding NHE6, exhibited the highest expression level, similarly high as that of Slc32a1 (VIAAT). Prominent immunoreactivity in MNTB axon terminals targeting LSO somata provides further support (Figure 8). Since NHE2–NHE5 expression was not detected in either region, NHE6 appears to be the most likely candidate. However, pharmacological attempts to block NHE activity with the antagonist EIPA were confounded by adverse effects on AP generation, preventing specific assessment. We acknowledge that functional evidence for NHE6’s role in SV acidification is lacking, and our immunohistochemical and gene expression data alone are insufficient to confirm its involvement.

We therefore propose the following hypothetical scenario, which aligns with existing literature across various synaptic systems (Kononenko and Haucke, 2015). After AP-triggered SV fusion and rapid transmitter release, SV membranes are retrieved and SVs via either classical clathrin-mediated endocytosis or clathrin-independent ultrafast endocytosis (Brockmann and Rosenmund, 2016; Eddings and Watanabe, 2025). The former occurs with a time constant of ~10 s, while the latter can proceed within hundreds of milliseconds (Dittman and Ryan, 2009; Soykan et al., 2017; Watanabe and Boucrot, 2017) or even faster (<100 ms; Watanabe et al., 2013; Delvendahl and Hallermann, 2016). In addition, activity-dependent bulk endocytosis, similar to micropinocytosis, can occur (Camblor-Perujo and Kononenko, 2021). Subsequently, reacidification takes place locally with time constants of 4–40 s (Atluri and Ryan, 2006; Egashira et al., 2015). NHEs could mediate rapid reacidification, particularly during heavy synaptic activity. In hippocampal synapses, a full exo-endocytosis cycle ~30 s (Fernández-Alfonso and Ryan, 2006), but can be as short as 15 s (Ryan and Smith, 1995; Hori and Takahashi, 2012). In the auditory brainstem, it may be faster. A recent article by Gallimore et al. (2025) provides further insight into the complete SV cycle.

As noted above, SVs refilling takes tens of seconds. One might argue that the rapid recovery observed after high-frequency stimulation in Bafi-treated synapses (within a few seconds; Figure 5) is inconsistent with this time course. However, steady-state amplitudes during challenge were determined in the s_50-60 window. Therefore, SVs endocytosed early (e.g., at s_5-25) could have recycled within 25–55 s, consistent with published data.

In our proposed dual-transporter model, vacuolar NHEs operate under highly favorable conditions for transporting H⁺ into the SV lumen. Immediately after endocytosis, the lumen contains extracellular fluid. Hence, luminal [Na⁺] (which was 153 mM in our ACSF) is ~30-fold higher than in the cytosol, driving Na⁺ efflux from the SVs. Because NHEs are antiporters, this drives H⁺ influx into the lumen, promoting reacidification. In addition, the luminal pH immediately after endocytosis is 7.4, compared to a cytosolic pH of 7.2, corresponding to a ~ 1.6-fold lower luminal [H⁺] (40 vs. 63 nM). This gradient also favors H⁺ entry into the lumen, an effect even more pronounced when V-ATPase activity is absent. Interestingly and perhaps surprisingly, a luminal pH of 7.4 corresponds to only 0.000275 free H⁺ ions per SV. A single free H⁺ ion in the SV lumen would result in pH 4 (Preobraschenski et al., 2014). However, the SV lumen has a buffer capacity of ~57 mM/pH, provided by vesicle proteins, ATP, and transmitter molecules (Edwards, 2007), which allows storage of ~1,200 H⁺ ions per SV (Egashira et al., 2015).

The equilibrium situation illustrated in Figure 1 is also relevant for NHE-mediated ion transport at rest. At a luminal pH of 6.4 versus a cytosolic pH of 7.2, the H⁺ gradient favors efflux from the SV. NHE activity under these conditions would dissipate the H⁺ gradient, counteract V-ATPase activity, and regulate alkalinization (Ouyang et al., 2013; Takamori, 2016). Further studies are needed to clarify the role of additional H⁺ transporters beyond V-ATPase in supplying protons to VIAAT, thereby ensuring sufficient SV filling, particularly during prolonged high-frequency activity. This issue applies to both inhibitory and excitatory synapses (Goh et al., 2011; Huang and Trussell, 2014; Lee et al., 2021).

Finally, in addition to NHE6, neurotransmitter transporter 4 (NTT4, also known as XT1 or B0AT3), might also contribute to SV refilling with glycine when the proton-motive force generated by the V-ATPase is diminished or absent. NTT4, encoded by Slc6A17, is exclusively expressed in the nervous system and associated with neuronal SVs (El Mestikawy et al., 1997; Fischer et al., 1999; Masson et al., 1999; Zaia and Reimer, 2009). Immunoreactivity has been demonstrated in both excitatory and inhibitory neurons (Hägglund et al., 2013). NTT4 catalyzes the transport of neutral amino acids, including glycine (Parra et al., 2008; Nicoli et al., 2025). Importantly, NTT4 is a sodium-dependent symporter that is inhibited by low pH (Zaia and Reimer, 2009), making it ideally suited for the initial stages of SV refilling (see Figure 12).

Several important questions remain open, leaving ample room for future experimentation. First, does sustained high-frequency stimulation after pharmacological inactivation of V-ATPase completely abolish neurotransmission in systems other than the MNTB-LSO synapses described here? Is the resilience in SV refilling unique to the auditory system, or does it extend to other robust, high-frequency synapses such as cerebellar mossy fiber-granule cell synapses (Delvendahl and Hallermann, 2016)? Could it even represent a more general property of synapses? Second, does NHE(6) directly contribute to refilling? This question requires functional evidence from assays employing specific NHE(6) antagonists or gene silencing. Third, are there additional vesicular transporters that maintain transmitter supply during high-frequency synaptic activity? Finally, can the robustness and resilience of neurotransmission at MNTB-LSO synapses be substantiated at the ultrastructural level using functional electron microscopy? Answering this question will require technically demanding “flash and freeze” or “zap and freeze” methods (Vandael and Jonas, 2024; Eddings and Watanabe, 2025).