Action potentials (APs) play a crucial role in evoking Ca2+ currents (ICa) through voltage-gated calcium channels (VGCCs) and transmitter release. During development and neuromodulation, both depolarization and repolarization phases of APs change, but how such changes affect the characteristics of ICa and its efficacy at central synapses is not clear. By paired voltage-clamp recordings of ICa and excitatory postsynaptic currents (IEPSC) with pseudo-APs and real APs, we examined these issues in the developing calyx of Held synapse of postnatal mice. We found that speeding the AP depolarization rate primarily reduces the number of activated VGCCs, whereas shortening the AP repolarization phase decreases the number of activated VGCCs and accelerates their kinetics. The ICa–IESPC relationships are well predicted by the integral but not the amplitude of ICa, and exhibit development- and temperature-dependent shifts toward left, indicating an enhancement in downstream Ca2+ coupling efficacy. Cross-correlation analyses of ICa and IEPSC evoked by real APs and pseudo-APs demonstrated that AP shortening in the half-width from 0.4 ms at postnatal day 8 (P8)–P12 to 0.27 ms at P16–P18 decreases ICa integral by 36%, but increases IEPSC by 72% as a result of developmental upregulation in coupling efficacy. These counteracting actions maintain the release fraction evoked by an AP at ∼10% of the maximal quantal output. We suggest that AP narrowing is a critical adaptation for the calyx of Held synapse to control the quantal output per AP and is likely important for the efficient use of the readily releasable pool of synaptic vesicles during high-frequency neurotransmission.
- calyx of Held synapse
- action potential
- Ca2+ current
- transmitter release
- developmental plasticity
- temporal fidelity
Pioneer work in several invertebrate synapses has demonstrated that the waveform of action potentials (APs) is a crucial element in regulating Ca2+ influx into nerve terminals and transmitter release (Klein and Kandel, 1980; Llinas et al., 1981, 1982; Augustine, 1990; Delaney et al., 1991). However, typical central synapses are too small to access, having largely precluded biophysical analysis of voltage-gated Ca2+ channels (VGCCs) and their couplings to transmitter release. Recent establishment of optical techniques (Sabatini and Regehr, 1996) and model preparations with large terminals (Forsythe, 1994; Borst et al., 1995; Geiger and Jonas, 2000) has led to the first quantitative description of the interplay between the AP waveform, Ca2+ influx, and quantal output in central synapses, but a series of issues arise from such studies. First, contrary to the findings in invertebrate synapses, in which only a small fraction of VGCCs are activated by an AP (Llinas et al., 1982; Augustine, 1990), these studies demonstrate that an AP can effectively open the majority of VGCCs (Sabatini and Regehr, 1997; Borst and Sakmann, 1998, 1999; Geiger and Jonas, 2000; Bischofberger et al., 2002). Second, at squid giant synapse, AP broadening increases the number of activated VGCCs and transmitter release in a linear manner, implying that release of single vesicles can be gated by as few as one channel (Augustine, 1990). In contrast, the fraction of VGCCs activated by an AP at central terminals approaches the submaximum and additional broadening mainly alters the kinetics of VGCCs (Sabatini and Regehr, 1997; Geiger and Jonas, 2000; Bischofberger et al., 2002). Cooperative actions of multiple channels are required for the release of single synaptic vesicles (Wu and Saggau, 1994; Mintz et al., 1995; Borst and Sakmann, 1996; Wu et al., 1999; Fedchyshyn and Wang, 2005). These studies raise the question whether changes in the AP waveform affect the number of activated VGCCs and/or their kinetics selectively or simultaneously. This issue is important not only for understanding different operating modalities for transmitter release among various synapses but also for differentiating subcomponents of Ca2+ currents (ICa) that may be targeted by neuromodulators or pharmacological reagents. Finally, previous studies have been done mostly in immature synapses at room temperature. It is unclear how development and temperature influence ICa evoked by real APs and consequently the quantal output under physiological conditions.
In light of these, we have used paired voltage-clamp recordings to investigate the effects of changing AP waveform on ICa and postsynaptic excitatory currents (IEPSC) at the mouse calyx of Held synapse. We report that the number of VGCCs or their kinetics can each account for approximately one-half of the total Ca2+ influx and quantal output. Developmental AP shortening reduces the number of recruited VGCCs and accelerates the kinetics of ICa, leading to a marked decrease in Ca2+ charge transfer at both 22 and 35°C. Quantal output evoked by smaller ICa is, however, upregulated by a developmental increase in the coupling efficacy.
Materials and Methods
Mice were housed in the facility certified by the Canadian Council of Animal Care and used for this study according to a protocol approved by the Hospital for Sick Children Animal Care Committee. Brainstem slices were prepared from postnatal day 8 (P8) to P18 mice (CD1×C57 hybrid), similar to that previously described for rats (Forsythe and Barnes-Davies, 1993). After decapitation with a small guillotine, brains were immediately immersed into semifrozen artificial CSF (ACSF) containing the following (in mm): 125 NaCl, 2.5 KCl, 10 glucose, 1.25 NaH2PO4, 2 Na-pyruvate, 3 myo-inositol, 0.5 ascorbic acid, 26 NaHCO3, 1 MgCl2, and 2 CaCl2 at a pH of 7.3 when oxygenated (95% O2 and 5% CO2) followed by rapid dissection. Transverse slices of the auditory brainstem containing the medial nucleus of the trapezoid body (MNTB) were cut at a thickness of 150–250 μm using a microtome (VT1000S; Leica Microsystems, Ernst-Leitz-Strasse, Wetzler, Germany) followed by incubation at 37°C for 1 h before experimentation. For current-clamp recordings of real APs, thicker slices (250 μm) were used to preserve afferent axons. For paired voltage-clamp recordings, thinner slices (150–200 μm) were prepared to minimize presynaptic axon length and reduce voltage-clamp errors. Most experiments were performed at room temperature (∼22°C), except for a subset of experiments in Figure 7, which were performed at 35°C using an inline heater with feedback thermistor (TC-324B; Warner Instruments, Hamden, CT).
All recordings were made with a dual-channel amplifier (MultiClamp 700A; Molecular Devices, Union City, CA). ACSF was supplemented with bicuculline (10 μm) and strychnine (1 μm) to block inhibitory inputs during recording. For recording of presynaptic ICa, tetrodotoxin (0.5 μm), tetraethylammonium (TEA) (10 mm) and 4-aminopyridine (4-AP) (0.3 mm) were added to block Na+ and K+ channels respectively, whereas extracellular Ca2+ concentration ([Ca2+]e) was set at 1 mm. Patch electrodes were fabricated from thin-wall glass with filaments (World Precision Instruments, Sarasota, FL) and coated with dental wax under a dissection microscope. Resistances of these pipettes were 4–6 MΩ for presynaptic terminals and 2.5–3 MΩ for postsynaptic neurons. For paired voltage-clamp recordings, presynaptic and postsynaptic series resistances were 6–10 MΩ and 4–8 MΩ, respectively, and compensated to 90%. Cells showing higher resistances were omitted from the analysis. Intracellular recording solution for presynaptic ICa contained the following (in mm): 10 CsCl, 40 HEPES, 0.5 EGTA, 1 MgCl2, 2 ATP, 0.5 GTP, 12 phosphocreatine, 20 TEA, and 3 K-glutamate, and pH adjusted to 7.3 with CsOH. Intracellular solution for postsynaptic recordings contained the following (in mm): 97.5 K-gluconate, 32.5 CsCl, 5 EGTA, 10 HEPES, 1 MgCl2, 30 TEA, and 3 lidocaine N-ethyl bromide, pH 7.2. The holding potential was −70 or −80 mV for presynaptic terminals and −60 mV for postsynaptic neurons, respectively. Presynaptic ICa was evoked by various voltage-command paradigms, indicated in the text, and subtraction of capacitive current and leak was done with the on-line P/4 protocol. Real APs were recorded from calyces at a sampling rate of 50 kHz and a filtering frequency of 4 kHz in the current-clamp configuration by stimulating afferent axon fibers with a bipolar platinum electrode. The intracellular solution contained the following (in mm): 97.5 K-gluconate, 32.5 KCl, 0.5 EGTA, 40 HEPES, and 1 MgCl2, pH 7.3 (no supplement of glutamate). After manually removing stimulation artifacts preceding the APs, the digitized values were used as voltage-command templates and fed back into the amplifier as stimulation files (Axon Text File) through pCLAMP 8 software at the same frequency as their acquisition (50 kHz). These real APs were then used as presynaptic voltage-command waveforms for experiments in Figures 1, 6, and 7. Details for pseudo-APs and real APs as voltage-clamp paradigms are given in the text and figure legends. Reagents were from Sigma (St. Louis, MO), Tocris Bioscience (Ellisville, MO), and Alomone Labs (Jerusalem, Israel).
Voltage-clamp quality and stability of ICa and IEPSC evoked by APs.
A series of pseudo-APs and real APs were used as voltage-clamp templates to evoke presynaptic ICa. The validity of such an approach relies on the assumption that space clamp of the entire terminal is adequate, so that voltage-clamp command templates delivered through a patch electrode do not degrade via the equivalent circuit with the access resistance (Rs) and membrane capacitance (Cm) in the whole-terminal configuration. We performed three sets of experiments to directly evaluate the quality of voltage clamp and stability of ICa in our study. In the first set of experiments, two electrodes were patched onto the same calyx and the whole-terminal configuration was sequentially established for each pipette. We delivered a real AP template through one electrode in voltage-clamp mode while the other electrode was used to record changes in membrane potential in current-clamp mode. Figure 1A shows such a recording in which patch electrodes were positioned on the opposite side of a synapse with a half-circle distance of ∼30 μm between two pipettes (diameter of postsynaptic neuron ∼20 μm; left panel). We found that the shape and size of recorded APs were very similar to those of the command AP template except for a short delay time (Fig. 1A, right panel), as predicted by the equivalent circuit (i.e., τD = RsCm). In three such recordings, we estimated the mean τD to be 0.18 ± 0.01 ms, whereas the mean amplitude of the recorded AP was decreased by 5.90 ± 2.22 mV or 5.4 ± 2.0% compared with the command template (n = 3). These results demonstrated that the fidelity of the AP waveform was largely preserved under our experimental conditions. Although the drop in the AP amplitude may be smaller than 6 mV for membrane area <30 μm away from the patch electrode, it should be noted that even a small decrease in the AP amplitude may have a significant impact on IEPSC (Wu et al., 2004), because IEPSC is a power function of ICa. To reduce such space-clamp errors and mimic the manner by which axonal APs propagate into the calyx, we positioned the presynaptic patch pipette onto the calyx membrane immediately next to the axon heminode for a majority of experiments included in this study. We found such a technical maneuver was critical for us to maintain low access resistance throughout the experiments (<10 MΩ) and obtain large and fast ICa. This is particularly important for recordings from P16–P18 calyces that have undergone morphological transformation into thin finger digits, from which delivery of voltage-clamp paradigms to the entire calyx with fidelity is difficult to achieve.
In the second set of experiments, we examined the kinetic changes of ICa evoked by a real AP in response to elevated extracellular Ca2+ concentration. If the quality of voltage clamp was poor, one could predict that significant slowing of activation and deactivation kinetics occur as ICa grows in size with an increased driving force. Figure 1B (left panel) shows the recordings of ICa in extracellular solutions containing low [Ca2+]e/high [Mg2+]e (0.5/4.5 mm) and a high [Ca2+]e/low [Mg2+]e (4/1 mm) ACSF. When Rs was kept <10 MΩ, we found that at low and high Ca2+ concentrations, the rise times of ICa were 0.19 ± 0.01 and 0.23 ± 0.03 ms, and decay times were 0.37 ± 0.02 and 0.35 ± 0.03 ms, respectively (Fig. 1B, right top panels), which are comparable with those obtained with double-electrode voltage-clamp experiments (Borst and Sakmann, 1998). The kinetics of ICa was not significantly changed (p > 0.05; n = 4) despite a nearly fourfold increase (from 0.71 ± 0.17 to 2.61 ± 0.68 nA) in the amplitude of ICa (Fig. 1B, right bottom panel), indicating that voltage-clamp errors are minor. To minimize such errors, all of the subsequent experiments in this study were done in 1 mm [Ca2+]e, which typically allowed generation of ICa under 2 nA. Low [Ca2+]e also helped reduce saturation and desensitization of postsynaptic AMPARs.
In the third set of experiments, we examined the stability of ICa and IEPSC evoked by voltage-clamp AP template after the breakthrough of the presynaptic and postsynaptic patches. We found that under the condition of low [Ca2+]e (1 mm) and low Rs (<10 MΩ), both currents were reasonably stable (Fig. 1C, left panels). Compared with the first responses, ICa and IEPSC after 10 min of recordings were 93.41 ± 4.29% and 96.08 ± 6.26% (n = 4), respectively (Fig. 1C, right panels), running down by ∼5%. Given this observation, we collected the majority of data within the first 10 min while the sequence of stimulation paradigms for different synapses was randomized so as to reduce the errors attributable to the rundown.
Data were acquired on-line, filtered at 4 kHz, digitized at 50 kHz and analyzed off-line using the pCLAMP 8 software package (Molecular Devices) and Excel 2000 (Microsoft, Redmond, WA). For analyses of the input (ICa)–output (IEPSC) relationship, the ICa integral was measured using the analysis window that began at the onset of ICa and ended at the point where the decay phase of ICa intercepted with baseline; the IEPSC integral was measured from the onset of EPSC to the point at which EPSC decayed back to 10% of the EPSC peak amplitude. Absolute or normalized current integrals were plotted against each other and fitted with the Hill function in the following form: IEPSC = IEPSC-Max/[1 + (ICa/IEC50)nH], where nH and IEC50 denote the Hill coefficient and the ICa integral at which the half-maximal IEPSC (IEPSC-Max) was evoked. In some cases, the integral of IEPSC was plotted against the amplitude of ICa to illustrate their relationship under different experimental conditions (e.g., Fig. 6). All raw traces shown in this study were single sweeps except for ICa and IESPC evoked by real APs, which were averages of three sweeps (see Figs. 6, 7). For other correlation analyses, least-squares linear regression was performed using Excel 2000 (Microsoft). Statistical tests of significance were two-tailed, unpaired Student’s t tests with a p value cutoff of <0.05. Data were expressed as the mean ± SE from a population of synapses (n).
Distinct effects of AP depolarization and repolarization rate on kinetics of presynaptic ICa
The calyx of Held synapse is an axosomatic synapse in the auditory brainstem, being capable of high-fidelity neurotransmission at extraordinarily high rates (up to 600 Hz) for sound localization (Trussell, 1999; von Gersdorff and Borst, 2002). During the early postnatal developmental stage (P7–P18) at this synapse, the AP width shortens dramatically because of accelerated depolarization and repolarization rates, whereas the AP amplitude remains relatively unchanged (Taschenberger and von Gersdorff, 2000; Fedchyshyn and Wang, 2005). These changes are presumably attributed to the concurrent upregulation of Na+ and K+ conductances (Dodson et al., 2003; Elezgarai et al., 2003; Ishikawa et al., 2003; Leao et al., 2005). However, it is not clear how the depolarization or repolarization rate affects the size and kinetics of presynaptic ICa. To address this question, we made patch-clamp recordings of ICa, from visually identified calyces, in response to three sets of AP-like voltage-clamp paradigms that were designed to distinguish the effects of changing each AP component (i.e., depolarization or repolarization) on ICa (Fig. 2A–C, top panels). In the first pseudo-AP-like voltage-clamp paradigm (i.e., AP-DEP) (Fig. 2A), the depolarization time was increased from 0.1 to 1 ms whereas the amplitude (130 mV, from −70 mV to +60 mV) and repolarization time (0.4 ms) remained constant. We found that presynaptic ICa evoked by this voltage paradigm gradually increased in amplitude and eventually saturated (the maximal amplitude: 1.79 ± 0.12 nA, n = 5) as the depolarization time was prolonged (Fig. 2A, middle panel). When the rise and decay time of ICa were quantified and plotted against the depolarization time, we found that the rise time was ∼0.16 ms and unaffected by the changes in the depolarization duration, but the decay time was increased slightly (from 0.23 ± 0.03 to 0.29 ± 0.04 ms, p > 0.05; n = 5) (Fig. 2A, bottom panel). Application of the second voltage paradigm, in which only the repolarization time of pseudo-APs was prolonged (from 0.2 to 1.6 ms) without changing the amplitude and depolarization time (0.2 ms) (i.e., AP-REP) (Fig. 2B), initially produced a gradual increase in the ICa amplitude, which subsequently saturated after the first few pseudo-APs at a level similar (1.49 ± 0.20 nA) to that evoked by AP-DEP paradigm (Fig. 2B, middle panel). However, the rise and decay time of ICa continued to increase (rise time: from 0.10 ± 0.01 to 0.49 ± 0.02 ms, p < 0.05; decay time: from 0.21 ± 0.04 to 0.49 ± 0.08 ms, p < 0.05; n = 5), even after the ICa amplitude reached its maximum (Fig. 2B, bottom panel). Because the driving force for Ca2+ in AP-DEP and AP-REP is identical, the similar maximal amplitudes of ICa generated by the two paradigms likely reflect activity of the same maximal number of recruited VGCCs. However, different kinetic profiles of ICa in response to AP-DEP and AP-REP raise the possibility that the repolarization phase is not only involved in determining the number of activated VGCCs but also simultaneously affects their kinetics.
To clearly separate the effects of the depolarization and repolarization phase on the number of recruited VGCCs and their kinetics, we designed a third set of pseudo-APs (Fig. 2C). In this paradigm, incremental prolongation of plateau duration (0–0.9 ms) was used to recruit an increasing number of VGCCs, while the depolarization and repolarization time course (0.2 and 0.2 ms, respectively) of command voltage ramps were unchanged (i.e., AP-STEP). Because voltage steps to +60 mV in the AP-STEP protocol presumably maximizes the open probability of VGCCs, and the driving force for Ca2+ remains constant for all open channels (Borst and Sakmann, 1998, 1999; Gentile and Stanley, 2004; Fedchyshyn and Wang, 2005), the only variable at any given plateau duration would be the number of recruited VGCCs. In response to AP-STEP, we found that ICa sharply increased in amplitude as the plateau duration was extended (Fig. 2C, middle panel) and reached the maximum amplitude of 1.82 ± 0.37 nA within 0.5 ms of plateau duration, but again the rise time was independent of the plateau duration, whereas the decay kinetics showed an increase (from 0.21 ± 0.02 to 0.31 ± 0.03 ms, p < 0.05; n = 5) (Fig. 2C, bottom panel). Under this protocol, the profile of ICa in its amplitude and kinetics resembled that evoked by AP-DEP (Fig. 2A,C), except that AP-STEP appeared to be more effective than AP-DEP in evoking ICa. The increase in the decay time constant of ICa evoked by AP-DEP and AP-STEP may reflect additional recruitment of distal axonal VGCCs (Borst and Sakmann, 1998) or a different gating mode of calyceal VGCCs in response to prolonged depolarization (e.g., from reluctant to willing mode) (Colecraft et al., 2000).
Together, these three sets of experiments with pseudo-APs suggest that the depolarization and repolarization phases have distinct effects on presynaptic ICa. The depolarization phase appears to selectively determine the number of recruited VGCCs, whereas the repolarization phase also regulates their kinetics. Hence, subtle changes in AP waveform may potentially provide the nerve terminal with different ways of shaping Ca2+ transients. We suggest that the convergent action of shortening both AP depolarization and repolarization phases at the calyx of Held synapse may reduce the amplitude of ICa and ensure rapid activation and deactivation kinetics of VGCCs and hence fast and brief Ca2+ transients. Such changes may be particularly important for controlling the amount of Ca2+ influx and transmitter release at this synapse.
Dependence of presynaptic ICa integral on AP depolarization and repolarization rate
To investigate how depolarization and repolarization rates of an AP influence the amount of Ca2+ influx, we quantified the total integral of ICa evoked by AP-DEP, AP-REP, and AP-STEP. When the first and last sweeps of recorded ICa in each paradigm were contrasted (Fig. 3A–C), it was apparent that the slow kinetics of ICa evoked by AP-REP led to a significantly greater total integral of ICa than that generated by AP-DEP or AP-STEP. The maximal ICa integrals produced by AP-DEP and AP-STEP are comparable (0.66 ± 0.12 nA·ms vs 0.52 ± 0.11 nA·ms, p > 0.05) but significantly lower than those evoked by AP-REP (1.18 ± 0.15 nA·ms, p < 0.05) (Fig. 3D). These results reinforce our interpretation that AP depolarization rate selectively determines the number of activated VGCCs with little effect on their kinetics, whereas AP repolarization rate affects the number and kinetics of recruited channels, both of which contribute to the total Ca2+ charge integral during an AP. In Figure 3E, ICa integrals from each of three voltage-clamp paradigms were normalized to their own maximum and plotted against the time of the depolarization, repolarization, or plateau duration. We noted that time-dependent increases in the ICa integral were well described with a single-exponential function, giving time constants of 0.18, 0.31, and 1.46 ms for AP-STEP, AP-DEP, and AP-REP, respectively. The greater ICa integral and slower time constants associated with AP-REP imply that the kinetic changes of ICa, as a result of altering AP repolarization rate, may provide the nerve terminal with a large dynamic range within which the amount of Ca2+ inflow during an AP may be fine tuned.
Fractional contribution of the number and kinetics of presynaptic ICa to quantal output
Knowing that both the number and kinetics of recruited VGCCs by an AP contribute to the total integral of ICa, we examined the relative contribution of these two components to the maximal quantal output. Using the same synapses, we performed paired recordings in which presynaptic ICa was evoked with the three sets of voltage-clamp paradigms, whereas IEPSC was simultaneously recorded as readout of transmitter release (Fig. 4). We found that in response to either AP-DEP or AP-STEP, IEPSC grew in size as ICa increased but remained relatively stable once the amplitude of ICa reached its maximum (Fig. 4A,C). However, when the AP-REP paradigm was applied, IEPSC continued to increase even after the amplitude of ICa had saturated (Fig. 4B), implying that the kinetic component of ICa contributed to the additional Ca2+ influx and transmitter release. To quantify these results, we measured the integrals of ICa and IEPSC and constructed input–output relationships for three pseudo-AP paradigms as shown in Figure 4D. We found that the maximal quantal output (IEPSC integral) produced by AP-DEP and AP-STEP was similar (16.40 ± 3.22 nA·ms vs 14.41 ± 2.42 nA·ms, p > 0.05, n = 5), although these two sets of input–output curves did not overlap. This nonoverlapping profile likely resulted from a difference in the effectiveness of the two protocols in recruiting VGCCs, as illustrated in Figure 3E. In contrast, AP-REP activated an increasing number of VGCCs and also prolonged their kinetics, doubling quantal output (28.07 ± 4.26 nA·ms, p < 0.05, n = 5) compared with AP-DEP and AP-STEP (Fig. 4D,E). These results suggest that both the number of activated VGCCs and their kinetics are two integral parts of the Ca2+ transient evoked by an AP, each of which could contribute up to ∼50% of the maximal transmitter release.
By plotting the integral of IEPSC against that of ICa on a log-log scale, we further examined input–output relationships for three paradigms and found that the slope factors (i.e., calcium cooperativity m) derived from linear fits to the data sets generated by AP-DEP and AP-STEP paradigms were very similar (∼4) (Fig. 4F). However, the input–output relationship for the AP-REP paradigm had two linear components with m values being 4.2 and 1.9, respectively. The m value for the steep component, which mainly covered data points before the ICa amplitude became saturated, was similar to those obtained with AP-DEP and AP-STEP paradigms. This is in agreement with our interpretation that the changes in the ICa amplitude resulting from AP-REP paradigm reflect different numbers of recruited VGCCs, as is also the case for AP-DEP and AP-STEP paradigms. The low m value component is likely associated with kinetic changes of ICa, which become the main contributor to the total ICa integral after the ICa amplitude reaches the maximum. Therefore, although the number of recruited channels and their kinetics can each contribute equally to the maximal quantal output, changing the former has much more robust effects on transmitter release than changing the latter. This implies that by changing the depolarization or repolarization time course, nerve terminals may modulate different components underlying the characteristics of the Ca2+ transient and cooperative nature of transmitter release.
Developmental characteristics of native APs, quantal output, and the fraction of VGCCs recruited by real APs
By systematically changing individual components of pseudo-APs, we have thus far separated the effects of changing the depolarization or repolarization rate on the amplitude and kinetics of ICa and on quantal output. However, how developmental changes in AP waveform affect synaptic strength remains to be established. To this end, we first characterized the APs evoked by axonal stimulation (axon-APs) from P8–P12 and P16–P18 calyces in both cell-attached voltage-clamp and whole-terminal current-clamp configurations while IEPSC was simultaneously recorded from the postsynaptic neurons (Fig. 5A,B). In cell-attached configuration, axon-APs were registered as an inward current and an outward current (Fig. 5A,B, top left panels), the peak of which usually corresponds to the maximal rate of AP depolarization and repolarization, respectively. The peak-to-peak time between these two currents, which approximates the half-width of an AP (Sabatini and Regehr, 1997), was estimated to be 0.41 ± 0.04 ms (n = 6) and 0.26 ± 0.01 ms (n = 8, p < 0.05) for P8–P12 and P16–P18 synapses, matching well the half-width of APs subsequently recorded from the same calyces (P8–P12: 0.41 ± 0.01 ms; P16–P18: 0.27 ± 0.02; p < 0.05) (Fig. 5C). In contrast, the amplitude of APs between two age groups was not significantly different, being 109.9 ± 7.3 mV (n = 6) and 103.0 ± 4.4 mV (n = 8, p > 0.05) for P8–P12 and P16–P18 calyces. These observations clearly demonstrated that axon-APs we recorded in current-clamp mode recapitulate developmental AP shortening, attested by the decrease in the peak-to-peak time with noninvasive cell-attached recording configuration that minimally disturbs intraterminal homeostasis. Despite a significant AP shortening, which presumably decreased ICa, IEPSC recorded in 1 mm [Ca2+]e before the rupture of presynaptic terminals actually increased in size from 3.36 ± 0.63 nA (n = 7) at P8–P12 to 4.72 ± 0.56 (n = 10) at P16–P18 (Fig. 5D), although not significantly (p > 0.05). Such an age-dependent increase in IEPSC is qualitatively consistent with the developmental trend in IEPSC as reported for the mouse calyx of Held synapse in 2 mm [Ca2+]e (Futai et al., 2001; Joshi and Wang, 2002).
From the above experiments, we selected two axon-APs and used their waveforms as voltage-clamp command templates to measure the fraction of VGCCs recruited by these real APs (Fig. 5E, right panel) (API for P8–P12 synapses: amplitude 110 mV with a half-width of 0.4 ms; APM for P16–P18 synapses: amplitude 110 mV with a half-width of 0.27 ms). We normalized their amplitude to 110 mV to specifically compare the effects of AP shortening in width on ICa. In the same calyces, we applied API, APM,and a voltage step (10 ms) to evoked ICa. Because a prolonged step presumably recruits all VGCCs on the calyx, the peak of tail ICa may be used as the maximal number of channels to estimate the relative fraction of VGCCs activated by real APs. By calculating the amplitude ratio of ICa evoked by real APs and that by the voltage-step, we found that API and APM activated ∼50 and 35% of VGCCs (49.7 ± 4.2% vs 34.9 ± 3.8; p < 0.05, n = 4). Because our previous work showed that the maximal size of ICa and its voltage dependence are not different between P8–P12 and P16–P18 calyces, but the physical coupling between VGCCs and synaptic vesicles tightens (Fedchyshyn and Wang, 2005), we postulate that developmental AP narrowing is an effective and necessary adaptation for the calyx of Held synapse to reduce the number of VGCCs engaged in triggering transmitter release during an AP.
Development and temperature dependence of quantal output evoked by real APs
To directly examine physiological consequence of AP shortening, we next compare the effects of AP narrowing on ICa and IEPSC. Because both depolarization and the repolarization phases speed up concurrently at the developing calyx of Held synapse (Fig. 5) (also see Taschenberger and von Gersdorff, 2000), we designed a series of pseudo-APs (Fig. 6A,B, top left panels), which had the same amplitude (i.e., −80 to +30 mV, 110 mV) as real APs, but different depolarization and repolarization rates. Using the same synapses, we performed voltage-clamp recordings of presynaptic ICa evoked by both pseudo-APs and real APs, paired with simultaneous recordings of IEPSC from either the P8–P12 (Fig. 6A) or P16–P18 age group (Fig. 6B). Independent of the maturity of the synapses, we found that the amplitude of ICa increased in response to the first few APs and then slightly decreased, as both depolarization and repolarization phases of pseudo-APs were prolonged. The rise and decay time continually increased, leading to a monotonic growth in the total ICa integral (Fig. 6C). In parallel, these kinetic changes in ICa also led to slower IEPSC rising phase in both age groups (Fig. 6D). When the total integrals of ICa and IEPSC were quantified and plotted against each other (Fig. 6F,G), both sets of input–output curves were well described by a Hill function. In contrast, the integral or amplitude of IEPSC could not be properly correlated with the amplitude of ICa, because the ICa amplitude reached the maximum and then declined, whereas IEPSC continued to increase in response to additional AP broadening (Fig. 6E). By fitting input–output curves based on the absolute values of ICa and IEPSC integrals from P8–P12 (n = 6) and P16–P18 synapses (n = 5) with the Hill function, we obtained an ICa integral of 0.70 ± 0.01 nA·ms and 0.48 ± 0.01 nA·ms (IEC50) for the half-maximal quantal output, and the Hill coefficient (nH) of 3.59 ± 0.14 and 2.85 ± 0.12, respectively (Fig. 6F,G). This change in nH parallels the developmental decrease in Ca2+“channel/domain cooperativity” as we previously reported (Fedchyshyn and Wang, 2005), in which we have shown that tighter coupling between VGCCs and synaptic vesicles in old synapses increases the release efficacy. Indeed, the maximal output (IEPSC-Max) in P16–P18 synapses (38.09 ± 0.64 nA·ms) was significantly higher than that in P8–P12 synapses (25.78 ± 0.44 nA·ms). The increase in the integral of IEPSC-Max could be somewhat underestimated if translated into the total number of quanta, because developmental subunit switching in AMPARs from slow-gating GluR1-dominant to fast-gating GluR3/4-dominant phenotype (Joshi et al., 2004; Koike-Tani et al., 2005) actually leads to IEPSC with faster kinetics and reduced the integral of unitary quantal event in older synapses.
When real AP templates (i.e., API and APM) were used to evoke ICa, we found that in P8–P12 calyces (n = 6), ICa produced by API and APM had a half-width of 0.30 ± 0.02 ms and 0.23 ± 0.01 ms and an average amplitude of 1.21 ± 0.15 nA and 0.79 ± 0.11 nA. In P16–P18 calyces (n = 5), ICa evoked by API and APM showed comparable results (half-width: 0.29 ± 0.02 ms and 0.22 ± 0.01 ms; amplitude: 1.38 ± 0.12 nA and 1.14 ± 0.13 nA). Hence, regardless of the maturity of synapses, APM generated much faster and smaller ICa than API, decreasing ICa integral activated during an AP. The absolute amplitude of IEPSC generated by API was 1.56 ± 0.90 nA for P8–P12 synapses and 4.54 ± 1.40 nA by APM for P16–P18 synapses. Compared with that of IEPSC evoked by axonal stimulation for the same age groups (P8–P12: 3.36 ± 0.63 nA; P16–P18: 4.75 ± 0.56 nA) (Fig. 5D), the former is substantially lower, and the latter correlates well. The difference in IEPSC generated by axon-APs and by voltage clamp with real APs may be attributed to the difficulty in our selection of real APs (i.e., API) as the voltage-clamp template. Immature calyces exhibit variable AP half-width ranging from 0.3 to 0.6 ms, and small variations in the AP width may lead to big changes in quantal output as a result of high Ca2+ cooperativity (m) (Fedchyshyn and Wang, 2005). In contrast, both the AP width and m have stabilized in mature synapses (P16–P18), resulting in a good correlation in IEPSC assayed with two different protocols.
Cross-correlation analyses of ICa and IEPSC integrals evoked by real APs with those generated by pseudo-APs showed that APM evokes 35.9% less ICa integral than API (0.25 ± 0.01 nA·ms vs 0.39 ± 0.04 nA·ms) but yields higher IEPSC integrals [3.94 ± 1.04 nA·ms (Fig. 6G, blue filled circle)] in P16–P18 synapses than API [2.29 ± 0.32 nA·ms (Fig. 6F, red open circle)] in P8–P12 synapses, representing an increase of 72.1% in quantal output. When the two sets of input–output curves are normalized and superimposed (Fig. 6H), we noted that API generates ∼30% (27.8 ± 0.7%, red open circle) of the total ICa integral, yielding ∼10% of the IEPSC-Max (9.7 ± 1.4%) for P8–P12 synapses, whereas APM produces ∼20% (18.0 ± 0.7%, blue filled circle) of the total ICa integral, resulting in slightly more than 10% of the IEPSC-Max (11.4 ± 2.9%) for P16–P18 synapses (Fig. 6H). Had AP shortening not occurred in P16–P18 synapses, API would generate less than onefold increase in ICa integral (0.44 ± 0.02 nA·ms) but more than sixfold increase in IEPSC integral (15.98 ± 2.73 nA·ms) (Fig. 6G, filled red circle) and fourfold increase in the release fraction per AP (46.5 ± 7.6%) (Fig. 6H, filled red circle). These results imply that AP shortening is a necessary adaptation for the calyx to efficiently control quantal output. Despite APM generating smaller total ICa integral than API, a prominent increase in the coupling efficacy counteracts the effects of decreased Ca2+ influx produced by narrow AP in mature calyces and effectively maintains the level of quantal output per AP.
It is well known that temperature has a profound impact on AP waveform and transmitter release (Charlton and Atwood, 1979; Sabatini and Regehr; 1996; Kushmerick et al., 2006). To investigate the temperature effect, we recorded APs from P8–P12 calyces (n = 4) at the near physiological temperature of 35°C and noted that APs were shortened in width (0.27 ± 0.04 ms) with little change in amplitude (107.3 ± 5.1 mV). We again normalized API at two temperatures to the same amplitude (110 mV) (Fig. 7A,B, top right panels). Based on these real APs, voltage-command templates were generated for the cross-correlation analyses as in Figure 6. Using immature calyces, we found that both ICa and IEPSC recorded at 35°C displayed accelerated kinetics, compared with those recorded at 22°C (Fig. 7A,B). When API was used to evoke ICa, raising temperature from 22 to 35°C reduced the half-width of ICa from 0.30 ± 0.02 to 0.22 ± 0.01 ms (Fig. 7E) (p < 0.05) with little effect on its amplitude (1.21 ± 0.15 and 1.50 ± 0.19 nA). With respect to IEPSC at high temperature, we found that IEPSC saturated at a lower ICa integral than that at 22°C with a similar maximal quantal output (IEPSC-Max: 25.78 ± 0.44 nA·ms for 22°C vs 26.10 ± 0.45 nA·ms for 35°C) (Fig. 7C). Comparison of these two sets of curves indicated that raising temperature led to an upregulation in release efficacy (IEC50: 0.70 ± 0.01 nA·ms for 22°C vs 0.46 ± 0.01 nA·ms for 35°C). API produced similar ICa integral at 35°C to that at 22°C (0.35 ± 0.03 nA·ms vs 0.39 ± 0.04 nA·ms) but significantly increased IEPSC by 2.9-fold (6.74 ± 2.11 nA·ms vs 2.29 ± 0.32 nA·ms). When the input–output relationship at 35°C was normalized to the maximum and compared with that at 22°C, we again observed a leftward shift in the curve (IEC50: 49.7 ± 0.6% for 22°C vs 41.3 ± 0.5% for 35°C). Although at 35°C API produced a subtle increase in normalized Ca2+ integrals (from 27.8 ± 0.7% to 32.0 ± 0.2%), a robust increase in the fraction of IEPSC relative to IEPSC-Max was observed at a high temperature (API: 9.7 ± 1.4% at 22°C and 27.9 ± 10.6% at 35°C) (Fig. 7D). Interestingly, the rise time of IEPSC was also reduced from 0.28 ± 0.01 to 0.23 ± 0.01 ms (p < 0.05) when temperature was increased from 22 to 35°C (Fig. 7F), implying that fast ICa transients may help synchronize release events. These results suggest that additional narrowing of APs at 35°C may lead to very brief Ca2+ transient but temperature-dependent upregulation in release efficacy downstream from Ca2+ influx can boost quantal output by each AP without compromising the synchrony of vesicular release.
Integral components of presynaptic ICa evoked by an AP
In this study, we have taken a reductionist approach to elucidate the effects of different components of an AP on the amplitude and kinetics of presynaptic ICa and its coupling efficacy for triggering vesicular release. We found that the number of recruited VGCCs, but not their kinetics, is highly sensitive to changes in the AP depolarization phase (Fig. 2A). Because the depolarization rate is mainly determined by voltage-gated Na+ currents, this surprising finding implies that small changes in Na+ channels, e.g., caused by neuromodulation or neural plasticity (Levitan, 1994; von Gersdorff and Borst, 2002; Engel and Jonas, 2005; Leao et al., 2005), may affect depolarization rate and indirectly modulate the number of active VGCCs during an AP and hence the amount of downstream Ca2+ inflow. At the calyx of Held synapse, Na+ channels (Nav1.6) have been shown to cluster in the axon heminode and upregulate with development, likely underlying the acceleration in the AP depolarization rate or subtle changes in the AP amplitude (Taschenberger and von Gersdorff, 2000; Leao et al., 2005). At this synapse, the density of high-threshold K+ channels (Kv 3) also shows an age-dependent increase, which may contribute to the rapid repolarization phase (Dodson et al., 2003; Elezgarai et al., 2003; Ishikawa et al., 2003). Our experiments with AP-REP paradigm demonstrated that the repolarization phase is also involved in determining the number of activated VGCCs, particularly when the AP width is narrow, as manifested by the increase in the ICa amplitude with AP broadening. However, the kinetic component of activated VGCCs becomes the dominant contributor to the total Ca2+ integral once the ICa amplitude saturates (Fig. 2B,D).
Our observations may help reconcile the opposing views on the effect of spike broadening on presynaptic ICa among different synapses. At the squid giant synapse, classical work showed that an AP opens only a small fraction of presynaptic VGCCs (∼10%) (Pumplin et al., 1981; Llinas et al., 1982) and that AP broadening enhances calcium entry primarily by increasing the number of calcium channels that open, leading to large increases in the amplitude of ICa (Augustine, 1990). In contrast, theoretical computer simulations and experimental evidence in central synapses, including the calyx of Held synapse, demonstrate that an AP effectively opens a majority of VGCCs (>70%), and AP broadening increases calcium influx mainly, if not exclusively, by prolonging the kinetics of calcium currents with little effect on the amplitude of ICa (Sabatini and Regehr, 1997; Borst and Sakmann, 1998, 1999; Geiger and Jonas, 2000; Bischofberger et al., 2002). In fact, much of the past work done in central synapses was based on biophysical properties of VGCCs in immature terminals, in which propagating APs usually have wide waveform [e.g., cerebellar granule–Purkinje cell synapse with AP half-width ∼1 ms from P10–P18 rats (Sabatini and Regehr, 1997); hippocampal mossy fiber–CA3 synapse with AP half-width ∼0.85 ms from P20–28 rats (Geiger and Jonas, 2000; Bischofberger et al., 2002); calyx of Held synapse with AP half-width ∼0.5 ms from P8–P10 rats (Borst and Sakmann 1998, 1999)]. As a result, AP broadening will only reveal changes in the ICa kinetics but not its amplitude. It is, however, conceivable that the differences in the AP amplitude among various synapses may also contribute to the apparent discrepancy. In addition, experiments in invertebrate synapses were usually performed at lower temperatures (e.g., 18°C for squid giant synapse) than those in central synapses, which may significantly slow the activation and deactivation kinetics of VGCCs and reduce the fraction of VGCCs recruited by an AP.
In this study, we have recapitulated distinct aspects of the AP broadening effects described in different synapses. In the experiments with real APs (Fig. 5), API (half-width, 0.4 ms) and APM (half-width, 0.27 ms) elicited ∼50 and 35% of the maximal ICa. The former value is lower than that (∼70%) reported for the rat immature calyx in Borst and Sakmann (1998), but this may be accounted for by the difference in the AP width (i.e., 0.4 vs 0.5 ms) and other subtle difference in experimental protocols. Regardless of the developmental stage of this synapse, APM always activates a smaller fraction of VGCCs than API, leaving the remaining channels for additional recruitment by AP broadening before the kinetic changes of ICa become the predominant contributor to the total integral (Figs. 6, 7). Therefore, the initial AP width defined by depolarization and repolarization time course may dictate whether AP broadening induces changes in the amplitude or the kinetics of ICa. We suggest that by accelerating AP depolarization and repolarization rates, developmental AP shortening at the calyx of Held synapse primarily reduces the number of VGCCs recruited by an AP and ensures the brevity of Ca2+ transients.
Efficacy of presynaptic ICa in triggering transmitter release
We have used paired recordings to characterize the coupling efficacy of presynaptic ICa evoked by various pseudo- and real APs with transmitter release. Probing with AP-DEP, AP-REP, and AP-STEP, we found that both the number of VGCCs and their kinetics contributed to the total Ca2+ integral (Fig. 3), with each of the two components contributing to ∼50% of the maximal quantal output (Fig. 4). Experiments with other pseudo-AP paradigms (Figs. 6, 7), mimicking concurrent changes in both depolarization and repolarization phases during development, demonstrate that there is a strong dependence of quantal output on the total integral of ICa, with all input–output relationships being well described by the Hill function independent of synaptic maturity and temperature (correlation coefficient R > 0.95 in all cases). By manipulating both the number and kinetic components of activated VGCCs, synapses may effectively expand their dynamic operating range and potentially present disparate targets for neuromodulation.
Our observations are in line with the notion that the integral of the Ca2+ transient rather than its amplitude determines the efficacy of transmitter release (Bollmann and Sakmann, 2005). AP narrowing from APM to API decreases absolute ICa integral by ∼36% while IEPSC increases by 72%, indicating enhanced synaptic strength. Cross-correlation analyses of normalized input–output curves between pseudo-APs and real APs revealed that API in P8–P12 synapses generates ∼28% of the maximal ICa integral and sets IEPSC <10% of the IEPSC-Max, whereas APM in P16–P18 synapses significantly decreased the total ICa integral to ∼18% and yet maintained IEPSC at 11% of the IEPSC-Max (Fig. 6). The mechanism underlying such a stability in the release fraction evoked by a reduced Ca2+ integral may reside in the opposing action of enhanced coupling efficacy, as evidenced by an age-dependent leftward shift in the input–output curve and an increase in IEPSC-Max. We have demonstrated previously that the spatial coupling between VGCCs and synaptic vesicles tightens as maturation progresses, transforming the release modality from loosely coupled “microdomain” to tightly coupled “nanodomain” (Fedchyshyn and Wang, 2005). Such a transformation in the active zones evidently enhances the coupling efficacy of Ca2+ influx to vesicular fusion. Similarly, raising temperature in immature synapses can also increase the coupling efficacy by reducing the minimal number of overlapping domains required for triggering release of single vesicles (Fedchyshyn and Wang, 2005). Hence, the effect of AP shortening on ICa can be counteracted by the increase in coupling efficacy. An additional extrapolation from our results is that the calyx of Held synapse at maturity has built up the mechanisms for controlling quantal output by regulating the number of recruited VGCCs and their couplings to synaptic vesicles, releasing transmitters in a manner reminiscent of that at the squid giant synapse (Llinas et al., 1981, 1982; Pumplin et al., 1981; Augustine, 1990).
Physiological implications of developmental AP shortening in high-fidelity and high-frequency synaptic transmission
At any central synapse, the relationship between the amount of Ca2+ influx and quantal output can be approximated by a power function in the form of IEPSC α [ICa] m, where m denotes Ca2+ cooperativity (Dodge and Rahamimoff, 1967). Experiments presented in this study demonstrate that AP narrowing is a highly effective adaptation for the calyx to reduce ICa integral. We have demonstrated that both developmental maturation and high temperature reduce the m value (Fedchyshyn and Wang, 2005). In the context of these lines of evidence, we suggest that the developmental reductions in ICa and m converge to constrain the number of synaptic vesicles released, whereas parallel upregulation of both coupling efficacy downstream from Ca2+ entry and the number of active zones at the mature calyx (∼600) (Taschenberger et al., 2002) may counteract such effects. Ultimately, synaptic strength is enhanced while the release fraction per AP is maintained. These adaptations help prevent the terminal from depleting the readily releasable pool of synaptic vesicles during high-frequency synaptic activity (Wang and Kaczmarek, 1998; Schneggenburger et al., 1999; Iwasaki and Takahashi, 2001; Taschenberger et al., 2002). Furthermore, small and brief ICa generated by narrow APs promotes synchronized release events (Fig. 7F) and hence the fidelity of synaptic transmission. Such biophysical plasticity along with concurrent morphological reorganization in active zones (von Gersdorff and Borst, 2002) is likely important for the development of high-fidelity neurotransmission required for sound localization at this synapse.
This work was supported by an individual operating grant, “The Synapse” group grant from the Canadian Institutes of Health Research, and by the Burroughs Wellcome Fund (L.-Y.W.). L.-Y.W. holds a Tier II Canada Research Chair. We thank Dr. Milton Charlton for valuable suggestions and other members of the Wang Laboratory for assistance and discussion.
- Correspondence should be addressed to Dr. Lu-Yang Wang, Division of Neurology, The Hospital for Sick Children, 555 University Avenue, Toronto, Ontario, Canada M5G 1X8. Email: