Abstract
Calretinin (CR) is a major calcium binding protein widely expressed in the CNS. However, its synaptic function remains largely elusive. At the auditory synapse of the endbulb of Held, CR is selectively expressed in different subtypes. Combining electrophysiology with immunohistochemistry, we investigated the synaptic transmission at the endbulb of Held synapses with and without endogenous CR expression in mature CBA/CAJ mice of either sex. Two synapse subtypes showed similar basal synaptic transmission, except a larger quantal size in CR-expressing synapses. During high-rate stimulus trains, CR-expressing synapses showed improved synaptic efficacy with significantly less depression and lower asynchronous release, suggesting more efficient exocytosis than non-CR-expressing synapses. Conversely, CR-expressing synapses had a smaller readily releasable pool size, which was countered by higher release probability and faster synaptic recovery to support sustained release during high-rate activity. EGTA-AM treatment did not change the synaptic transmission of CR-expressing synapses, but reduced synaptic depression and decreased asynchronous release at non-CR-expressing synapses, suggesting that CR helps to minimize calcium accumulation during high-rate activity. Both synapses express parvalbumin, another calcium-binding protein with slower kinetics and higher affinity than CR, but not calbindin. Furthermore, CR-expressing synapses only express the fast isoform of vesicular glutamate transporter 1 (VGluT1), while most non-CR-expressing synapses express both VGluT1 and the slower VGluT2, which may underlie their lagged synaptic recovery. The findings suggest that, paired with associated synaptic machinery, differential CR expression regulates synaptic efficacy among different subtypes of auditory nerve synapses to accomplish distinctive physiological functions in transmitting auditory information at high rates.
SIGNIFICANCE STATEMENT CR is a major calcium-binding protein in the brain. It remains unclear how endogenous CR impacts synaptic transmission. We investigated the question at the large endbulb of Held synapses with selective CR expression and found that CR-expressing and non-CR-expressing synapses had similar release properties under basal synaptic transmission. During high-rate activity, however, CR-expressing synapses showed improved synaptic efficacy with less depression, lower asynchronous release, and faster recovery. Furthermore, CR-expressing synapses use exclusive VGluT1 to refill synaptic vesicles, while non-CR-expressing synapses use both VGluT1 and the slower isoform of VGluT2. Our findings suggest that CR may play significant roles in promoting synaptic efficacy during high-rate activity, and selective CR expression can differentially impact signal processing among different synapses.
Introduction
Calretinin (CR) is one of the major calcium-binding proteins widely expressed in the brain to regulate intracellular calcium, and thus calcium-mediated cellular function (Rogers et al., 1990; Schwaller, 2014, 2020). The synaptic functions of CR are of particular interest because of the pivotal role of calcium in synaptic transmission. Studies on the topic were limited, however, because of the difficulty in identifying and accessing distinct synapses with and without endogenous CR expression. Instead, CR function was investigated at synapses in transgenic CR-deficient mice (Schurmans et al., 1997; Gurden et al., 1998; Schiffmann et al., 1999; Schmidt et al., 2013; Brachtendorf et al., 2015), or synapses with artificial overexpression of CR (Bolshakov et al., 2019). Consistent with its general role in buffering calcium, CR was found in these studies to enhance short-term synaptic depression and indirectly impact the induction of long-term potentiation. Nonetheless, synapses investigated in these studies had disturbed CR expression, and were small with unreliable evoked responses to the tested (single or a few) limited stimulations. It remains unclear how endogenous CR affects synaptic transmission, especially during sustained activity at high rates when synaptic calcium accumulates to high levels on repetitive channel openings. CR binds calcium with positive cooperativity (Faas et al., 2007) and is expected to operate at greater buffering capacity under these high calcium conditions (Saftenku, 2012). The question is ideally investigated in large model synapses with easy identification, undisturbed endogenous CR expression, as well as large and reliable synaptic responses.
In cochlear nucleus (CN), the first neural station of the central auditory system, principal bushy neurons receive large axosomatic synapses from the auditory nerve (AN) named the endbulb of Held (Nayagam et al., 2011; Yu and Goodrich, 2014). For decades, these synapses have been studied as a model for the mechanisms of synaptic transmission as well as for their physiological function in processing temporal fine structure of sound (Joris et al., 1994a, b; Manis et al., 2011). Recently, we reported that endbulb of Held synapses can be classified into two groups based on their selective CR expression (Wang et al., 2021). CR-expressing endbulbs are presumably from spiral ganglion neurons (SGNs) with high spontaneous firing rate and low threshold, whereas non-CR-expressing endbulbs are from SGNs with medium/low spontaneous firing rate and medium/high threshold (Liberman, 1982; Petitpré et al., 2018; Sharma et al., 2018; Shrestha et al., 2018; Sun et al., 2018). These two types of endbulbs transmit different aspects of sound information (especially about intensity) and converge at various ratios onto CN bushy neurons with distinct physiological properties (Wang et al., 2021). In response to stimulus trains, bushy neurons that receive predominantly CR-expressing endbulbs fire more spikes in sustained pattern with better temporal precision, while those that receive predominantly non-CR-expressing endbulbs fire fewer spikes in transient or onset pattern with relatively compromised timing. It is unknown how endogenous CR impacts the synaptic transmission at these two types of endbulbs in contributing to the distinct responses of their postsynaptic bushy neurons.
Combining electrophysiology with immunohistochemistry, we studied the synaptic transmission at the endbulb of Held synapses with and without endogenous CR expression, using acute brain slices from normal hearing CBA/CaJ mice. We found that both types of synapses had similar release property under basal transmission, but CR-expressing endbulbs showed improved synaptic efficacy with less depression and reduced asynchronous release during sustained high-rate activity, which was commonly observed in AN under physiological condition (Taberner and Liberman, 2005; Wen et al., 2009). Our findings shed new light on activity-dependent CR function in synaptic transmission, especially during prolonged activity at high rates. We further showed that selective expression of CR at different endbulbs is accompanied by other key synaptic molecules including parvalbumin and different isoforms of vesicular glutamate transporters (VGluTs), which act collectively to accomplish distinctive functions at target synapses.
Materials and Methods
Research animals.
CBA/CaJ mice of either sex were obtained from The Jackson Laboratory, and housed and bred at the animal facility at The Ohio State University. Mice of either sex were used in this study at the age of 2–5 months. All experiments were conducted under the guidelines of the protocol approved by the Institutional Animal Care and Use Committee of The Ohio State University.
Brain slice preparation.
Acute brain slices were prepared as previously described (Xie and Manis, 2017; Wang et al., 2021). Briefly, mice were anesthetized via intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg) and decapitated, and the brainstem retrieved. Parasagittal slices containing the CN were cut at a thickness of 225–240 µm using a vibratome (either Vibratome 1000, Technical Products; or model VT1200S Microtome, Leica Biosystems). Slices were incubated for 30–45 min to recover in artificial CSF (ACSF) at 34°C. ACSF contained the following (in mm): 122 NaCl, 3 KCl, 1.25 NaH2PO4, 25 NaHCO3, 20 glucose, three myo-inositol, 2 sodium pyruvate, 0.4 ascorbic acid, 1.8 CaCl2, and 1.5 MgSO4, and was continuously gassed with 95% O2 and 5% CO2.
Electrophysiological recording.
After incubation, brain slices were transferred to a submersion recording chamber under a microscope (Axio Examiner, Carl Zeiss) and were bathed in the same ACSF continuously flowing at a rate of 2–3 ml/min. Whole-cell recording in voltage-clamp mode was performed from neurons in the middle- and high-frequency area of the anteroventral CN (Wang et al., 2021). Recording electrodes were made from borosilicate glass pipette (catalog #KG-33, King Precision Glass) using a micropipette puller (catalog #P-2000, Sutter Instrument), and filled with electrode solution containing the following (in mm): 130 CsMetSO3, 5 CsCl, 5 EGTA, 10 HEPES, 4 MgATP, 0.3 GTP, 10 Tris-phosphocreatine, and 3 QX 314, with pH adjusted to 7.20. The electrode solution also contained 0.01% Alexa Fluor 594 by weight to fill the recorded neurons for visualization. Electrophysiological data were acquired using Multiclamp 700B amplifier, Digidata 1550B Acquisition System, and pClamp 11 software (Molecular Devices). All recordings were made at 34°C. The study included only bushy neurons, which were identified morphologically by having short primary dendrites with heavy branches (Cant and Morest, 1979; Webster and Trune, 1982; Wu and Oertel, 1984; Manis et al., 2019) and were further confirmed by receiving large endbulb of Held synapses (see Immunohistochemistry). A junction potential of −7 mV was corrected in reported voltages. Cells were held at −67 mV, with series resistance compensated by ∼75% online. Two micromolar strychnine was added to the bath ACSF to block glycinergic inhibitory synaptic transmission. Spontaneous EPSC (sEPSC) events were recorded first, followed by evoked EPSCs (eEPSC) triggered with single- and paired-pulse stimulations at the AN to evaluate synaptic properties under quiescence. Each AN stimulation was a 0.1 ms voltage pulse, delivered through a 75-µm-diameter concentric stimulating electrode (Frederick Haer Company). The stimulus intensity was set at ∼30% above the intensity level that reliably triggered maximum eEPSCs in the target neuron. To investigate synaptic transmission during high-rate activity, 50-pulse stimulus trains were delivered at 400 Hz, followed by a single post-train stimulation at various delays between 10 and 1500 ms after the end of each train to evaluate synaptic recovery. For some neurons, experiments continued with bath application of 20 μm EGTA-AM for 15 min, and the recordings were repeated. After the completion of data acquisition, the recording electrode was slowly withdrawn from the neuron, which resealed and was filled with the Alexa Fluor 594 dye. The brain slice was then fixed in 4% paraformaldehyde in PBS for 15 min and washed in PBS for post hoc immunostaining to identify the subtypes of the endbulb of Held synapses on the recorded neurons (Wang et al., 2021).
Immunohistochemistry.
Immunostaining of the fixed brain slices with labeled neurons were performed as previously described (Káradóttir and Attwell, 2006; Lin and Xie, 2019; Wang et al., 2021). Slices with labeled neurons were first incubated in blocking solution (10% horse serum, 0.5% Triton X-100, and 0.05% NaN3 in PBS) at room temperature for 6 h, followed by overnight incubation with primary antibodies against CR (rabbit ant-CR; 1:500; catalog #214102, Synaptic Systems) and VGluT1 (guinea pig anti-VGluT1; 1:500; catalog #135304, Synaptic Systems) at 4°C. After rinsing, slices were then incubated with corresponding secondary antibodies (goat anti-rabbit IgG; Alexa Fluor 647 conjugated; 1:500; catalog #A21245, Thermo Fisher Scientific; and goat anti-guinea pig IgG; Alexa Fluor 488 conjugated; 1:500; catalog #A11073, Thermo Fisher Scientific) for 4 h at room temperature. The slices were rerinsed and mounted on slide using DAPI-Fluoromount-G mounting media (Southern Biotech). Confocal images of the target neurons and innervating AN synapses were acquired using a confocal microscope (model FV3000, Olympus).
Expression patterns of parvalbumin, calbindin, VGluT1, and VGluT2 in AN synapses were studied using cryo-sectioned CN slices (Lin and Xie, 2019). After cutting, acute brain slices containing CN were immediately fixed in 4% paraformaldehyde in PBS for 15 min, washed, and cryoprotected in 30% sucrose in PBS. Slices were then embedded in Cryo-Gel (Instrumedics) and resectioned in the same parasagittal orientation to 30 µm in thickness using a cryostat slicer (model CM3050 S, Leica Biosystems). Resectioned slices were triple stained using antibodies against CR (rabbit ant-CR; 1:500; catalog #214102, Synaptic Systems), VGluT1 (guinea pig anti-VGluT1; 1:500; catalog #135304, Synaptic Systems), and parvalbumin (chicken anti-parvalbumin; 1:500; catalog #195006, Synaptic Systems), calbindin (chicken anti-calbindin; 1:500; catalog # 214006, Synaptic Systems), or VGluT2 (chicken anti-VGluT2; 1:500; catalog #135416, Synaptic Systems). Corresponding secondary antibodies were used including goat anti-chicken IgY, DyLight 755 conjugated (1:500; catalog #SA5-10 075, Thermo Fisher Scientific).
Image processing.
We followed the same image analysis procedure as in our previous study (Wang et al., 2021) using Imaris software (version 9.5.0; Oxford Instruments). Three-dimensional structure of the dye-filled target bushy neuron, CR-labeled AN synaptic terminals, and VGluT1-labeled puncta surrounding the soma of the target neuron were analyzed using semiautomated Imaris tools. The volumes of the VGluT1-labeled puncta were measured, including puncta that located inside the CR-labeled synaptic terminals and those that did not. These puncta volumes were used to quantify the amount of CR-expressing and non-CR-expressing synaptic inputs onto the target neurons, respectively. Bushy neurons that received >70% CR-expressing synaptic inputs by volume were classified as the CR group, and those with <25% CR-expressing synaptic inputs were classified as the non-CR group.
Electrophysiology analysis.
sEPSC events were detected using an amplitude threshold of two times the SD of the background noise and measured in MATLAB (MathWorks). All other electrophysiological data were analyzed in Igor Pro (WaveMetrics) using custom-written algorisms, as described in previous studies (Xie and Manis, 2013, 2017) and specified in the Results section.
Experimental design and statistical analysis.
Experimental design of this study was two group comparison in different properties of synaptic transmission between CR-expressing and non-CR-expressing synapses. Statistical analyses were performed with GraphPad Prism (version 6.0 h). Group data were first analyzed using D'Agostino–Pearson omnibus test to determine whether they are normally distributed. Statistical comparisons were then made using either unpaired t test (data with normal distribution) or Mann–Whitney test (data that are not normally distributed). Two-way ANOVA was used to compare stimulus train-evoked responses between two groups, as specified in the Results section. All data are reported as the mean ± SD except in certain figure panels, which plotted mean ± SEM for purposes of clarity, as specified in the figure legends.
Results
Classification of neurons with CR- and non-CR-expressing synaptic inputs
Our previous study showed that CR-expressing and non-CR-expressing endbulb of Held synapses converge onto individual CN bushy neurons, and that the convergence ratio varies (Wang et al., 2021). Neurons that receive exclusively CR-expressing or non-CR-expressing synapses are rare. The present study used the same method as in our previous study (Wang et al., 2021) to quantify the volume of synaptic inputs based on the 3D reconstruction of the synapses, and classified postsynaptic bushy neurons based on the ratio of the types of synaptic inputs they receive. As shown in Figure 1, slices containing dye-filled neurons were immunostained with antibodies against VGluT1, which labels glutamate vesicles in all AN synapses, and CR, which labels CR-expressing synaptic terminals. VGluT1-labeled puncta surrounding the somata of the target bushy neurons were reconstructed in 3D from image stacks (Fig. 1, right panels), and their volumes were measured to quantify the ratio of CR-expressing synapses (Fig. 1, VGluT1 puncta inside CR-labeled synaptic terminals, yellow) and non-CR-expressing synapses (Fig. 1, VGluT1 puncta outside of CR-labeled terminals, green). We classified bushy neurons with >70% of synaptic inputs from CR-expressing synapses as CR dominant (Fig. 1A–C), and those with <25% as non-CR dominant (Fig. 1D–F). To investigate the synaptic transmission of synapses with and without CR, we performed voltage-clamp recording from these two groups of bushy neurons, including 35 neurons with CR-dominant synapses (referred to as the CR synapse group) and 31 neurons with non-CR-dominant synapses (referred to as the non-CR synapse group) recorded from a total of 57 CBA/CaJ mice. The average membrane resistance was 137 ± 104 MΩ in CR synapse group and 120 ± 79 MΩ in non-CR synapse group (Mann–Whitney test, p = 0.278). Correspondingly, the average membrane capacitance was 17.8 ± 4.9 and 19.9 ± 5.4 pF (Mann–Whitney test, p = 0.246), and the series resistance was 16.0 ± 4.8 and 14.7 ± 5.7 MΩ (Mann–Whitney test, p = 0.197). Because these groups contained heterogeneous populations of CR inputs (>70% vs <25%), it is likely that the differences observed in this study were an underestimate between synapses with and without CR expression.
Classification of neurons with CR- and non-CR-expressing synaptic inputs. A–C, Example bushy neurons that receive predominantly CR-expressing synapses. Panels from left to right: Alex Fluor 594-filled neurons, CR staining, VGluT1 staining, merged image, and 3D-reconstructed VGluT1-labeled puncta. Color representation in the right two panels: yellow, VGluT1-labeled puncta from CR-expressing synapses (overlap of red and green); green, VGluT1-labeled puncta from non-CR-expressing synapses. Numbers in the right panels mark the volume ratio of synaptic inputs from CR-expressing synapses. D–F, Example bushy neurons that receive predominantly non-CR-expressing synapses. Scale bar, 5 µm.
Synaptic transmission under quiescence is similar between CR and non-CR synapses
We first recorded sEPSCs from the target neurons to evaluate spontaneous vesicle release under quiescence (Fig. 2A). CR and non-CR synapses released sEPSCs with similar event frequency (Fig. 2B; CR synapses, 6.7 ± 5.5 Hz; non-CR synapses, 7.2 ± 5.4; Mann–Whitney test, p = 0.800) and average decay time constant (Fig. 2C; CR synapses, 0.35 ± 0.17 ms; non-CR synapses, 0.37 ± 0.18; Mann–Whitney test, p = 0.423). The average sEPSC amplitude, however, was significantly larger in CR synapses (Fig. 2D,E; CR synapses: –85.9 ± 19.6 pA, n = 34; non-CR synapses: –74 ± 28.6 pA, n = 31; Mann–Whitney test, p = 0.0057). As shown in Figure 2F, sEPSC amplitude distribution skewed toward large amplitudes at both synapse subtypes with an average skewness of 0.70 ± 0.29 (n = 34) in CR synapses and 1.08 ± 0.56 (n = 31) in non-CR synapses (Mann–Whitney test, p = 0.0004). The results indicate that both CR-expressing and non-CR-expressing synapses have similar spontaneous release property, except an average of 15% larger quantal size in CR-expressing synapses.
Synaptic transmission under quiescence is similar between CR and non-CR synapses. A, sEPSC events from two example bushy neurons with predominantly CR (red) and non-CR (blue) synapses. B, C, Summary of sEPSC event frequency (B) and average decay time constant (C) in CR and non-CR synapses. NS, Not significant. D, Average sEPSC traces from cells in A. E, Summary of sEPSC average amplitude. **p < 0.01. F, Amplitude distribution of sEPSC events in CR and non-CR synapses. Thin lines, Single neurons; thick lines, mean ± SEM.
Evoked release (eEPSCs) were then investigated in response to single (Fig. 3A) or paired-pulse stimulations (Fig. 3B,C). The average eEPSC peak amplitudes were similar between CR synapses (–3.6 ± 1.9 nA, n = 35) and non-CR synapses (–3.7 ± 1.5 nA, n = 31; Fig. 3D; Mann–Whitney test, p = 0.474). The decay phase of the average eEPSC from each cell was best fit with either a single or a double exponential function, from which the eEPSC decay time constant was derived (tau from single exponential fit or weighted tau from double exponential fit; Xie and Manis, 2013). On average, the eEPSC decay time constant was 0.50 ± 0.18 ms (n = 35) in CR synapses and 0.68 ± 0.40 ms (n = 31) in non-CR synapses (Fig. 3E; Mann–Whitney test, p = 0.155). Similarly, no significant difference was found in the paired-pulse ratio (PPR) at either 2.5 ms paired-pulse interval (Fig. 3F; CR synapse: 0.89 ± 0.10, n = 32; non-CR synapse: 0.92 ± 0.20, n = 31; Mann–Whitney test, p = 0.571), or 50 ms internal (Fig. 3G; CR synapse: 0.91 ± 0.06, n = 35; non-CR synapse: 0.91 ± 0.04, n = 31; Mann–Whitney test, p = 0.808). Therefore, CR and non-CR synapses show similar evoked release to single-pulse or paired-pulse stimulations, indicating that endogenous CR does not affect the basal synaptic transmission at the endbulb of Held synapses.
Basal evoked release is similar at CR and non-CR synapses. A, Averaged eEPSCs to single AN stimulation in two example bushy neurons that receive predominantly CR and non-CR synapses. Arrow, Stimulus onset. B, C, Paired-pulse stimulation evoked EPSCs from the same neurons in (A) at 2.5 ms (B) and 50 ms (C) interpulse intervals. D–G, Comparisons between CR and non-CR synapses in eEPSC amplitude (D), eEPSC decay time constant (E), and paired-pulse ratios at the interval of 2.5 ms (F), and 50 ms (G). NS, Not significant.
CR synapses show improved synaptic efficacy during high-rate activity
Under physiological condition, AN can be highly active and fire spikes at rates >400 Hz (Kiang et al., 1965; Taberner and Liberman, 2005; Wen et al., 2009). We therefore assessed the synaptic transmission at both CR and non-CR synapses in response to AN stimulation with 50-pulse trains at 400 Hz. Both synapses showed vigorous eEPSCs at the beginning of the stimulus train, followed by synaptic depression with reduced eEPSC peak amplitude until reaching a steady state toward the second half of the train (Fig. 4A). Synaptic depression, however, was less profound in CR synapses, in which the normalized eEPSCs were significantly larger in peak amplitude toward the later section of the 400 Hz train than those of non-CR synapses (Fig. 4B; two-way ANOVA; pulse number effect: F(49,2550) = 181, p < 0.0001; synapse type effect: F(1,2550) = 219, p < 0.0001; interaction: F(49,2550) = 0.976, p = 0.5209). On average, eEPSCs depressed to 26.6 ± 6.7% (n = 25) during the last 20 pulses of the train at CR synapses, compared with 18.6 ± 7.7% (n = 28) at non-CR synapses (unpaired t test: t(51) = 4.07, p = 0.0002). The results indicate that CR synapses have higher synaptic efficacy with significantly less depression, which is expected to promote the transmission of sustained activity at high rates.
CR synapses show improved synaptic efficacy during high-rate activity. A, Example eEPSCs to 400 Hz stimulus trains at CR and non-CR synapses. Ticks on top, Stimulus onset; dashed line, baseline. Notice that eEPSCs in non-CR synapse in blue showed more synaptic depression later in the train with smaller eEPSC peak amplitude and elevated summation current from the baseline. B, Comparison of eEPSC peak amplitude throughout 400 Hz trains (normalized to the first eEPSC) from CR and non-CR synapses. For clarity, error bars represent SEM. C, Cumulative plots of the eEPSC size from the same two example neurons in A. Lines, Linear regression of the steady-state release. The RRP size was estimated as the y-axis intercept of the fitted line. D, E, Summary in RRP size (D) and initial release probability (E) using the method in D at CR and non-CR synapses. ***p < 0.001; ****p < 0.0001. F, Plot of individual eEPSCs against the cumulative size of prior eEPSCs from the same two example neurons in A. Lines, Linear regression of the initial eEPSCs. The RRP size was estimated as the x-axis intercept of the fitted line. G, H, Summary in RRP size (G) and initial release probability (H) using the method in F at CR and non-CR synapses. *p < 0.05.
We calculated the readily releasable pool (RRP) size of the synapses using the 400 Hz eEPSC trains. As described by Schneggenburger et al. (1999), the quantal content of each eEPSC was calculated as the eEPSC peak amplitude divided by quantal size (average sEPSC amplitude), and the cumulative eEPSC size (in number of vesicles) was plotted throughout the stimulus train (Fig. 4C). The steady-state phase of the cumulative plot (from 21st to 50th pulse) was fit with a line, which was back-extrapolated to time 0 to obtain estimated RRP size. As shown in Figure 4D, CR synapses had significantly smaller RRP size (177 ± 82 vesicles, n = 25) than non-CR synapses (314 ± 143 vesicles, n = 28; Mann–Whitney test, p < 0.0001). The initial release probability (Pr), calculated as the first eEPSC size divided by the RRP size, was significantly higher in CR synapses than non-CR synapses (Fig. 4E; CR synapse: 0.244 ± 0.040, n = 25; non-CR synapse: 0.187 ± 0.051, n = 28; unpaired t test, t(51) = 4.44, p < 0.0001).
A high rate of synaptic transmission releases a large quantity of glutamate at the synaptic cleft that can desensitize and saturate postsynaptic AMPA receptors (AMPARs; Chanda and Xu-Friedman, 2010), especially at immature synapses (Taschenberger et al., 2002; Renden et al., 2005; Wang and Manis, 2008). This study used mature CBA/CaJ mice >2 months of age, and it is unclear whether AMPAR desensitization and saturation decrease the eEPSC responses during the steady-state phase of the stimulus train and cause an underestimation of RRP size in Figure 4, C and D. Therefore, we also calculated the RRP size using the method by Elmqvist and Quastel (1965), which relies on the initial eEPSCs of the stimulus train and is less prone to AMPAR desensitization and saturation issues. As shown in Figure 4F, the vesicle release of individual eEPSCs throughout the 400 Hz train was plotted against the cumulative release preceding the given stimulus, and the RRP size was estimated as the x-axis intercept of the line fitted to the first several eEPSCs of the train. Consistent with Figure 4, C and D, the calculated RRP size using the initial eEPSCs of the train was also significantly smaller in CR synapses (304 ± 135 vesicles, n = 25) than in non-CR synapses (450 ± 248 vesicles, n = 28; Fig. 4G; Mann–Whitney test, p = 0.017). The Pr was higher in CR synapses than non-CR synapses (Fig. 4H; CR synapse: 0.170 ± 0.046, n = 25; non-CR synapse: 0.138 ± 0.055, n = 28; Mann–Whitney test, p = 0.013). Overall, calculation using both the steady-state and the initial eEPSCs of the 400 Hz trains showed that CR synapses have smaller RRP size but larger Pr than non-CR synapses. The differences between these two methods are likely because of different model assumptions, as discussed by Neher (2015).
To further clarify the contribution of AMPAR desensitization to synaptic transmission at CR and non-CR synapses, we applied 50 μm cyclothiazide (CTZ) to block AMPAR desensitization and tested its effect in six CR synapses and seven non-CR synapses. On average, CTZ did not change the amplitude of single evoked EPSCs at both CR (control, –3.1 ± 1.2 nA; CTZ, –3.2 ± 1.3 nA; n = 6; paired t test: t(5) = 0.212, p = 0.841) and non-CR synapses (control, –3.6 ± 1.1 nA; CTZ, –4.6 ± 1.6 nA; n = 7; paired t test: t(6) = 2.42, p = 0.052). However, eEPSC kinetics became slower after CTZ application resulted in a significantly larger EPSC decay time constant at both synapses (CR control, 0.50 ± 0.17 ms; CR with CTZ, 2.65 ± 1.15 ms; n = 6; paired t test: t(5) = 5.33, p = 0.0031. Non-CR control, 0.84 ± 0.42 ms; non-CR with CTZ, 2.66 ± 1.13 ms; n = 7; Wilcoxon matched-pairs signed-rank test, p = 0.0156). In consequence, the baseline EPSC response during the 400 Hz train was elevated under CTZ at both synapses (Fig. 5A). Synaptic depression was only slightly relieved throughout the 400 Hz train under CTZ, but the change was significant at both synapses (Fig. 5B; CR synapses, two-way ANOVA: CTZ effect: F(1,500) = 18.0, p < 0.0001; pulse number effect: F(49.500) = 85.3, p < 0.0001; interaction, F(49,500) = 0.562, p > 0.9999; Fig. 5D; non-CR synapses, two-way ANOVA; CTZ effect: F(1,600) = 9.51, p = 0.0021; pulse number effect: F(49,600) = 64.4, p < 0.0001; interaction, F(49,600) = 0.189, p > 0.9999). The calculated RRP size using the steady-state eEPSCs of the trains was not significantly different between control and CTZ application at both CR (Fig. 5C; control, 168 ± 69 vesicles; CTZ: 178 ± 66 vesicles, n = 6; Wilcoxon test, p > 0.999) and non-CR synapses (Fig. 5E; control, 300 ± 181 vesicles; CTZ, 387 ± 241 vesicles, n = 7; Wilcoxon test, p = 0.375). The results suggest that AMPARs does desensitize during high-rate synaptic transmission at mature endbulb of Held synapses, but does not significantly contribute to the observed differences in RRP size between CR and non-CR synapses.
CTZ effects on synaptic transmission. A, Example eEPSCs to 400 Hz stimulus trains at CR and non-CR synapses. Ticks on top, Stimulus onset. Traces in black was under CTZ application. Trains were normalized to the first eEPSC peak for comparison. B, Comparison of eEPSC peak amplitude throughout 400 Hz trains (normalized to the first eEPSC) from CR synapses before and after CTZ application. For clarity, error bars represent SEM. C, Summary of estimated RRP size before and after CTZ application. NS, Not significant. D, Comparison of eEPSC peak amplitude throughout 400 Hz trains (normalized to the first eEPSC) from non-CR synapses before and after CTZ application. Error bars represent SEM. E, Summary of estimated RRP size before and after CTZ application. NS, Not significant.
Since eEPSC peak amplitude only reflects synchronous vesicle release, we additionally calculated the total vesicle release per stimulation throughout the train, which is the charge transfer under each eEPSC curve that includes both synchronous and asynchronous release. The total release was significantly more reduced during the stimulus train in CR synapses than non-CR synapses (Fig. 6A; two-way ANOVA; pulse number effect: F(49,2550) = 35.3, p < 0.0001; synapse type effect: F(1,2550) = 231, p < 0.0001; interaction: F(49,2550) = 0.127, p > 0.9999), reflecting more efficient exocytosis given their larger eEPSC peak amplitude (Fig. 4B). We defined the synchronous release as the eEPSC charge transfer within a 1 ms window surrounding the eEPSC peak (Fig. 6B, from 0.3 ms before to 0.7 ms after the peak, cyan area), and the rest as the asynchronous release (gray area). Throughout the stimulus train, CR synapses had significantly less asynchronous release than non-CR synapses (Fig. 6C; two-way ANOVA; pulse number effect: F(49,2550) = 16.9, p < 0.0001; synapse type effect: F(1,2550) = 725, p < 0.0001; interaction, F(49,2550) = 0.901, p = 0.6697), and the difference grew larger with increasing pulse number. During the last 20 pulses of the stimulus train, 35 ± 6% (n = 25) of the vesicle release were asynchronous in CR synapses, compared with 46 ± 8% (n = 28) in non-CR synapses (unpaired t test: t(51) = 5.01, p < 0.0001). The enhanced asynchronous release in non-CR synapses contributes to the elevated summation current from the baseline as exampled by Figure 4A.
CR synapses show less asynchronous release during high-rate activity. A, Comparison of eEPSC total charge throughout 400 Hz trains (normalized to the first eEPSC) from CR and non-CR synapses. For clarity, error bars represent SEM. B, Calculation of synchronous and asynchronous release within each stimulus cycle from the example neurons in Figure 4A. Only the last two eEPSCs of the trains were shown. Cyan, Synchronous release charge as quantified within a 1 ms window surrounding the eEPSC peak (from 0.3 ms before the peak to 0.7 ms after the peak); gray, asynchronous release outside of the 1 ms window within the stimulus cycle; arrows, stimulus onset. C, Relative asynchronous EPSC (aEPSC) charge within each stimulus cycle throughout the eEPSC trains at CR and non-CR synapses. Error bars represent SEM.
These results showed that while CR synapses were smaller in RRP size, they had higher Pr and their evoked vesicle release was more synchronous during high-rate activity than that for non-CR synapses. These features warrant effective exocytosis in CR synapses to improve high-rate synaptic transmission that supports reliable spikes in postsynaptic neurons. During the process, CR as a calcium-binding protein is expected to curtail the elevation of calcium within the synaptic terminal and reduce asynchronous vesicle release. The effect of CR calcium buffering would become more substantial during high-rate activity, as synaptic calcium can be significantly elevated on repetitive calcium influx over a prolonged time (Saftenku, 2012).
EGTA-AM improves synaptic transmission at non-CR synapses
To test whether the lower asynchronous vesicle release in CR synapses is due to more efficient calcium regulation within the synaptic terminal, we studied the effect of EGTA-AM treatment on synaptic transmission in both CR and non-CR synapses. EGTA-AM is a membrane-permeable form of the calcium-chelating agent EGTA that enters the synaptic terminal to generate intracellular EGTA, which reduces the level of free calcium ions and decreases asynchronous release (Kaeser and Regehr, 2014). As shown in Figure 7A, EGTA-AM treatment did not change the synaptic depression throughout the 400 Hz trains at CR synapses as measured in normalized eEPSC peak amplitude (n = 7 neurons; two-way ANOVA; pulse number effect: F(49,600) = 58.1, p < 0.0001; EGTA-AM effect: F(1,600) = 0.0849, p = 0.7709; interaction: F(49,600) = 0.0510, p > 0.9999). However, synaptic depression was significantly reduced at non-CR synapses (Fig. 7B; n = 5 neurons; two-way ANOVA; pulse number effect: F(49,400) = 47.3, p < 0.0001; EGTA-AM effect: F(1,400) = 63.1, p < 0.0001; interaction: F(49,400) = 0.360, p > 0.9999). Similarly, when measured in EPSC total charge, EGTA-AM did not change the transmission at CR synapses (Fig. 7C; two-way ANOVA; pulse number effect: F(49,600) = 34.1, p < 0.0001; EGTA-AM effect: F(1,600) = 2.69, p = 0.1014; interaction: F(49,600) = 0.0629, p > 0.9999), but showed a trend to relieve synaptic depression at non-CR synapses (Fig. 7D; two-way ANOVA; pulse number effect: F(49,400) = 2.92, p < 0.0001; EGTA-AM effect: F(1,400) = 2.93, p = 0.0876; interaction: F(49,400) = 0.0249, p > 0.9999). The relative asynchronous release was not changed throughout the EPSC train in CR synapses after EGTA-AM treatment (Fig. 7E; two-way ANOVA; pulse number effect: F(49,600) = 1.41, p = 0.0391; EGTA-AM effect: F(1,600) = 2.33, p = 0.1276; interaction: F(49,600) = 0.0409, p > 0.9999), but was significantly reduced in non-CR synapses (Fig. 7F; two-way ANOVA; pulse number effect: F(49,400) = 0.527, p = 0.9965; EGTA-AM effect: F(1,400) = 13.7, p = 0.0003; interaction: F(49,400) = 0.0337, p > 0.9999). It is worth noting that the effect of EGTA-AM treatment was more prominent toward the second half of the 400 Hz trains (Fig. 7B,D,F), where more calcium is expected to accumulate at the synaptic terminal after repetitive calcium influx. The results suggest that calcium accumulation is alleviated in CR synapses during high-rate activity, presumably because of the calcium buffering of CR, so that synaptic transmission is not affected by exogenous EGTA. In contrast, because of the lack of CR calcium buffering, non-CR synapses likely accumulate higher levels of calcium during high-rate activity, which is chelated by exogenous EGTA, resulting in reduced asynchronous release and improved synaptic efficacy.
EGTA-AM improves synaptic transmission at non-CR synapses. A, B, Effect of EGTA-AM treatment in eEPSC peak amplitude (normalized to the first eEPSC) throughout 400 Hz trains in CR (A) and non-CR (B) synapses. Ctrl, Control; EGTA, after EGTA-AM treatment. C, D, Effect of EGTA-AM treatment in eEPSC charge throughout 400 Hz trains (normalized to the first eEPSC) in CR (C) and non-CR (D) synapses. E, F, Effect of EGTA-AM treatment in the relative asynchronous EPSC charge in CR (E) and non-CR (F) synapses. For clarity, all error bars represent SEM.
Both CR and non-CR synapses express parvalbumin but not calbindin
CR, parvalbumin, and calbindin are the three most commonly expressed calcium-binding proteins in the nervous system (Schwaller, 2020). To test what other calcium-binding proteins are also used, we investigated the expression patterns of all three calcium-binding proteins at the endbulb of Held synapses using immunohistochemistry. As shown in Figure 8A, triple staining of VGluT1, CR, and parvalbumin revealed that both CR and non-CR synapses were labeled by parvalbumin. In contrast, neither synapse was labeled by calbindin, which only occasionally labeled some postsynaptic CN neurons (Fig. 8B). Therefore, CR synapses express both CR and parvalbumin as the main calcium-binding proteins, while non-CR synapses only express parvalbumin. Given the drastic differences in their calcium-binding kinetics and affinity (Schwaller, 2010, 2020), CR and parvalbumin may regulate synaptic calcium under different conditions and differentially impact the synaptic transmission at CR and non-CR synapses (for details, see Discussion).
Both CR and non-CR synapses express parvalbumin but not calbindin. A, Representative staining of endbulb of Held synapses against VGluT1 (green), CR (red), and parvalbumin (magenta). Note that all VGluT1-labeled synapses were also labeled by parvalbumin. Triangles, Example neurons with predominant CR synapses labeled by all three markers; stars, example neurons with predominant non-CR synapses labeled by VGluT1 and parvalbumin but not CR. B, Representative staining of endbulb of Held synapses against VGluT1 (green), CR (red), and calbindin (magenta). Note that all VGluT-labeled synapses were not labeled by calbindin. Triangles, Example neurons with predominant CR synapses labeled by VGluT1 and CR; stars, example neurons with predominant non-CR synapses labeled only by VGluT1.
CR synapses recover faster
Synaptic transmission is a dynamic process that involves vesicle depletion and recovery, which collectively determine the transmission efficiency, especially during high-rate activity. We studied the vesicle recovery in both CR and non-CR synapses by assessing the post-train recovery of eEPSCs, which were probed at 10–1500 ms after the end of 400 Hz stimulus train. As shown in Figure 9A, eEPSCs recovered faster in size at CR synapses than those at non-CR synapses. The difference in recovery speed was significant throughout the tested 1500 ms post-train window (Fig. 9B; two-way ANOVA; recovery time effect: F(11,450) = 52.4, p < 0.0001; synapse type effect: F(1,450) = 72.2, p < 0.0001; interaction: F(11,450) = 0.964, p = 0.479), and was most profound during the first several hundred milliseconds. To quantify the recovery speed, we best fit the eEPSC recovery of each cell with either a single or a double exponential function to calculate the eEPSC recovery time constant. A double exponential function was used only when the slow time constant was at least three times longer than the fast decay time constant, and the resulted fit had a χ2 value that was at least three times smaller than that of a single exponential fit of the same data. Overall, the recovery of eEPSC amplitude in eight CR synapses and nine non-CR synapses were best fit by a single exponential function with the average decay time constant of 64 ± 47 and 156 ± 165 ms, respectively. eEPSC recovery of 11 CR synapses were best fit by a double exponential function, with the average fast decay of 28 ± 13 ms (weight, 56 ± 27%) and slow decay of 152 ± 115 ms (weight, 44 ± 27%). Twelve non-CR synapses showed eEPSC recovery that were best fit by a double exponential function, with the average fast decay of 26 ± 13 ms (weight, 54 ± 13%) and slow decay of 834 ± 674 ms (weight, 46 ± 13%). We calculated the weighted decay time from synapses with double exponential fit and plotted all the synapses in Figure 9C. Overall, CR synapses recovered with an average time constant of 100 ± 176 ms (n = 19), compared with 276 ± 259 ms (n = 21) in non-CR synapses (Mann–Whitney test, p = 0.0011). Because of the large variation in eEPSC amplitudes, the recovery time constant estimated from individual synapses had a fairly wide range. When we averaged all synapses within each subtype (Fig. 9B), the eEPSC recovery in CR synapse was best fit by a single exponential function (red curve) with a decay time constant of 58 ms, and the recovery of non-CR synapse was best fit by a double exponential function (blue curve) with a fast decay of 31 ms (62% weight) and a slow decay of 791 ms (38% weight). These results suggest that synaptic vesicles recover significantly faster in CR synapses, which contributes to improved synaptic efficacy and better supports sustained activity at high rates than non-CR synapses.
CR synapses recover faster. A, Example EPSC recovery from two example neurons with CR (red) and non-CR (blue) synapses after 50-pulse stimulus train at 400 Hz. eEPSCs were normalized to the first eEPSC of the train (data not shown). Recovery was probed at 10 ms (first eEPSC), and 20, 40, 70, 120, 200, 300, 440, 600, 800, 1100, and 1500 ms (last eEPSC) after the end of stimulus train. B, Summary EPSC recovery in all CR and non-CR synapses. Error bars represent SEM. C, Comparison of EPSC recovery time constant. **p < 0.01. D, EPSC recovery time constant in CR synapses before (Ctrl) and after EGTA-AM treatment (EGTA). E, EPSC recovery time constant in non-CR synapses before (Ctrl) and after EGTA-AM treatment (EGTA). NS, Not significant.
Synaptic recovery was shown to be regulated by calcium, in which high calcium accelerates the speed of vesicle endocytosis and RRP replenishment (Dittman and Regehr, 1998; Neher and Sakaba, 2008), and buffering the residual calcium with EGTA at the synaptic terminal was reported to slow down EPSC recovery (Yang and Xu-Friedman, 2008; Cho et al., 2011; Ritzau-Jost et al., 2018). In this study, surprisingly, we did not observe any significant changes in EPSC recovery time constant after EGTA-AM treatment in either CR (Fig. 9D; n = 7; control, 74 ± 70 ms; EGTA, 123 ± 131 ms; Wilcoxon test, p = 0.9375) or non-CR synapses (Fig. 9E; n = 5; control, 159 ± 150 ms; EGTA, 98 ± 84 ms; Wilcoxon test, p = 0.3125). The results indicate that in addition to their distinctive expression of calcium-binding proteins, CR and non-CR synapses may differ in other synaptic machinery involving calcium-independent processes (Ritzau-Jost et al., 2018) that contribute to the difference in their synaptic recovery speed.
CR synapses do not express VGluT2
One rate-limiting step in synaptic recovery is vesicle recycling (Hori and Takahashi, 2012), including especially the process of refilling newly formed vesicles with neurotransmitter (Nakakubo et al., 2020). Both CR and non-CR synapses in this study are glutamatergic and require VGluTs to pack the vesicles with glutamate. The two major isoforms of VGluTs expressed in the CN are VGluT1 and VGluT2 (Gómez-Nieto and Rubio, 2011; Ito et al., 2011). While VGluT1 is expressed in both CR and non-CR synapses (Wang et al., 2021), it is unclear whether and how VGluT2 is expressed in these synapses that may underlie their differences in synaptic transmission. VGluT1 was shown to have faster kinetics than VGluT2 in loading glutamate into synaptic vesicles, which helps improve sustained synaptic transmission (Nakakubo et al., 2020).
We triple-stained CN against VGluT1, VGluT2, and CR to investigate the expression pattern of VGluT2 in CR and non-CR synapses. As shown in Figure 10, CR synapses were labeled by VGluT1 and CR (Fig. 10, merged panel, yellow), but not by VGluT2. In contrast, most non-CR synapses were labeled by both VGluT1 and VGluT2 (Fig. 10, merged panel, cyan), and not by CR. Thus, CR synapses only express the fast VGluT1 isoform that helps promote the speed of vesicle recycling, whereas most non-CR synapses coexpress both VGluT1 and the slower VGluT2 isoforms that presumably contribute to the lagged vesicle recycling and synaptic recovery. We also found some occasional non-CR synapses that were labeled by only VGluT1 and not by VGluT2 or CR (Fig. 10, merged panel, green), which suggests further diversification of non-CR synapses that may involve additional elements to serve different synaptic functions.
CR synapses do not express VGluT2. Representative staining of endbulb of Held synapses against VGluT1 (green), CR (red), and VGluT2 (blue). Note that all CR synapses colabeled by VGluT1 and CR (merged panel, yellow) were not labeled by VGluT2, while most non-CR synapses labeled by VGluT1 but not CR were labeled by VGluT2 (merged panel, cyan). Triangles, example neurons with predominant CR synapses labeled by VGluT1 and CR, but not VGluT2; stars, example neurons with predominant non-CR synapses labeled by VGluT1 and VGluT2, but not CR.
Discussion
Function of CR in synaptic transmission
CR is known to buffer calcium and regulate synaptic function. Because of its positive cooperativity in binding calcium (Faas et al., 2007; Saftenku, 2012), CR is expected to impact synaptic transmission in an activity-dependent manner. Specifically, CR binds calcium at a slow rate when calcium levels are low, but the rate drastically accelerates with increasing calcium as more calcium-binding sites are occupied. Under resting condition, the level of calcium is very low (<100 nm) and CR is mostly in calcium-free form (Schwaller et al., 2002; Schwaller, 2020). Action potential invasion triggers calcium influx and transient elevation of calcium, which quickly diminishes after calcium channel closure and activation of multiple calcium clearance pathways (Garaschuk et al., 1997; Billups and Forsythe, 2002; Kim et al., 2005). During sustained activity with repetitive calcium influx, calcium accumulates within synaptic terminal, resulting in enhanced residual calcium, which grows with increasing firing rate and duration (Saftenku, 2012). Therefore, during infrequent firing or the first several spikes of the stimulus train, CR likely binds calcium at a slow rate and functions with moderate impact on synaptic transmission. In this study, we found that CR-expressing and non-CR-expressing endbulbs had similar basal synaptic transmission and PPR (Figs. 2, 3), supporting the notion that CR buffering does not significantly impact synaptic transmission at low activity level. Our findings under high rate activity, during which residual calcium builds up to high levels (Zhuang et al., 2020) that engage greater CR buffering capacity, suggest that CR helps improve synaptic efficacy by alleviating the buildup of residual calcium and reducing asynchronous release. The impact of CR likely increases with accumulating calcium level as reflected by the gradually increased difference in asynchronous release throughout the stimulus trains between CR and non-CR synapses (Fig. 6C). A similar effect of CR in buffering calcium was simulated in cerebellar granule cells and was not mimicked by any concentrations of exogenous calcium chelators (Saftenku, 2012).
Function of parvalbumin at CR and non-CR synapses
We investigated the expression patterns of three major calcium-binding proteins at the endbulbs (Fig. 8) and found that both CR-expressing and non-CR-expressing synapses also express parvalbumin, a slower calcium buffer than CR (Schwaller, 2020). The calcium-binding sites of parvalbumin also bind magnesium, and are mostly occupied by magnesium under resting condition when the calcium level is low (<100 nm) and the magnesium level is at 0.3–0.6 mm (Li-Smerin et al., 2001; Schwaller et al., 2002; Schwaller, 2020). The slow dissociation of magnesium from the binding sites delays the rate of calcium binding to parvalbumin when intracellular calcium elevates. Furthermore, parvalbumin has significantly higher affinity to calcium (Kd, 50–250 nm; Lee et al., 2000; Schwaller, 2020) than CR (initial binding Kd, 28–36 μm; Faas et al., 2007), which results in differential calcium buffering under different activity levels. At resting condition with low calcium (<100 nm), a small fraction of parvalbumin binds calcium and helps to maintain a low calcium environment. During synaptic transmission, local synaptic calcium transiently rises to up to hundreds of micromoles (Augustine et al., 1991; Llinás et al., 1992a, b; Hsu et al., 1996; Wang and Augustine, 2014), causing more parvalbumin binding, which helps to curb calcium diffusion and regulate short-term synaptic plasticity (Caillard et al., 2000; Vreugdenhil et al., 2003; Collin et al., 2005). During repetitive transmission at high rates, however, the persistently elevated calcium is expected to saturate most parvalbumin and decrease its buffering capacity. Under such conditions, CR calcium binding accelerates and plays the dominant role. Therefore, the main function of parvalbumin is likely to buffer calcium at rest and during infrequent activity at both CR-expressing and non-CR-expressing synapses. In contrast, CR takes over the major role in buffering calcium at CR-expressing synapses to improve synaptic efficacy during sustained transmission at high rates. Lacking such a mechanism, non-CR-expressing synapses likely build up higher residual calcium within the terminals resulting in enhanced asynchronous release and more synaptic depression during high-rate activity. Interestingly, parvalbumin and calbindin were shown not to affect synaptic transmission at recurrent Purkinje neuron synapses during high-rate activity (Bornschein et al., 2013), suggesting that calcium-binding proteins may have distinctive functions at different central synapses.
Estimated RRP size at the endbulb of Held synapses
Multiple studies reported the RRP size of endbulb of Held synapses that spans from ∼100 vesicles (Oleskevich et al., 2004; Taruno et al., 2012), 200–600 vesicles (Butola et al., 2017; Xie and Manis, 2017), to >1000 (Lin et al., 2011). The large variation may be attributed to different animal models and ages, but especially the use of distinct methods and stimulation paradigms favoring different assumptions about vesicle pool models (Neher, 2015). Our estimated RRP sizes are within the range of previous findings using the same cumulative analyses (Butola et al., 2017; Xie and Manis, 2017). The large size of 1060 vesicles reported by Lin et al. (2011) is likely an overestimation, since their study depleted RRP using prolonged voltage jumps of up to 50 ms but neglected RRP replenishment during the process. As shown in Figure 9 and in other studies (Wang and Manis, 2008; Yang and Xu-Friedman, 2009; Mendoza Schulz et al., 2014), RRP recovers significantly within 50 ms and the replenished vesicles would contribute to the measured capacitance jump, resulting in exaggerated RRP size.
Synaptic recovery and VGluTs
Synaptic vesicles released during transmission are constantly replenished to maintain continuous activity. The faster synaptic recovery in CR-expressing synapses suggests that they have more efficient vesicle replenishment, as indicated by steeper overall slope of the cumulative EPSCs in Figure 4C. This recovery process is regulated by calcium and can be slowed down by calcium buffer EGTA (Cho et al., 2011; Ritzau-Jost et al., 2018). Surprisingly, our study showed that synaptic recovery was faster in CR-expressing synapses (Fig. 9A–C), which presumably have lower synaptic calcium because of CR buffering than that of non-CR-expressing synapses. It suggests that in addition to calcium buffering, other mechanisms may also contribute to the different recovery speed. Indeed, we found that CR-expressing endbulbs express only VGluT1, while non-CR-expressing endbulbs express both VGluT1 and VGluT2. In VGluT1 knock-out mice, Nakakubo et al. (2020) showed that the loss of VGluT1 at the calyx of Held significantly slowed down the speed of vesicle refilling because of the slow action of VGluT2, which did not change the basal synaptic transmission, but significantly increased the failure rate with impaired fidelity during sustained transmission. Therefore, it is likely that vesicle refilling is slower in non-CR-expressing endbulbs because of VGluT2, which contributes to their lagged synaptic recovery. It is also possible that the expression of other proteins like piccolo and bassoon is stronger in CR-expressing synapses allowing for faster vesicle reloading (Hallermann and Silver, 2013), which needs further investigation in future studies.
CR and auditory function
CR was used as a molecular marker to differentiate physiologically and anatomically distinctive SGN populations (Liu and Davis, 2014; Petitpré et al., 2018; Sharma et al., 2018; Shrestha et al., 2018; Sun et al., 2018) and is believed to play significant roles in auditory function. Specifically, CR is expressed in SGNs with high spontaneous rate and low threshold (Liberman, 1982), which constantly fire spikes at rates up to 120 spikes/s even in quiet environment (Liberman, 1978; Taberner and Liberman, 2005), and their intracellular calcium continuously accumulates because of repetitive channel openings. The buffering capacity of CR may be essential in these SGNs to keep the calcium level low and reduce calcium related cellular damages (Farber, 1990). In contrast, CR-deficient SGNs have a medium/low spontaneous rate and a medium/high threshold. These SGNs fire fewer spikes and respond to loud (and often transient) sound signals, during which intracellular calcium only temporarily rises. Excessive and prolonged calcium elevation may be detrimental to these neurons because of the lack of CR buffering capacity (Sharma et al., 2018), which happens under pathologic conditions like noise overexposure. Indeed, cochlear synaptopathy after noise insult and during aging preferentially damages low-spontaneous rate SGNs (Kujawa and Liberman, 2009; Sergeyenko et al., 2013; Kujawa and Liberman, 2015; Liberman et al., 2015; Liberman, 2017), whose central AN synapses are also more severely degenerated during age-related hearing loss (Wang et al., 2021). Therefore, CR may play a neural protective role in CR-expressing SGNs through enhanced calcium buffering.
Our previous study (Wang et al., 2021) showed that bushy neurons innervated by predominantly CR-expressing endbulbs fire sustained spikes to high-rate stimulation and are optimized to process constantly ongoing sound information. In contrast, bushy neurons with predominantly non-CR-expressing endbulbs fire only transient or onset spikes and are ideal to detect transient signals. Both types of responses are well supported by the distinct synaptic properties observed in this study, in which CR-expressing endbulbs show improved synaptic efficacy with less depression, lower asynchronous release and faster synaptic recovery promoting sustained transmission; whereas, the opposite features appear in non-CR-expressing endbulbs favoring the transmission of transient signals. Together, our results highlight a crucial role for CR in processing auditory information, in which differential calcium buffering and vesicle recycling at CR-expressing and non-CR-expressing endbulbs support unique synaptic properties required for proper auditory circuit function.
Footnotes
This work was supported by National Institutes of Health (NIH) Grants R01-DC-016037 and R56-DC-019093 to R.X. Images were generated at The Ohio State University Campus Microscopy & Imaging Facility, which is supported by NIH Grant P30-CA-016058. We thank Benjamin Seicol for comments on the manuscript.
The authors declare no competing financial interests.
- Correspondence should be addressed to Ruili Xie at ruili.xie{at}osumc.edu
