Differential Gating and Recruitment of P/Q-, N-, and R-Type Ca²⁺ Channels in Hippocampal Mossy Fiber Boutons

Abstract

Voltage-gated Ca²⁺ channels in presynaptic terminals initiate the Ca²⁺ inflow necessary for transmitter release. At a variety of synapses, multiple Ca²⁺ channel subtypes are involved in synaptic transmission and plasticity. However, it is unknown whether presynaptic Ca²⁺ channels differ in gating properties and whether they are differentially activated by action potentials or subthreshold voltage signals. We examined Ca²⁺ channels in hippocampal mossy fiber boutons (MFBs) by presynaptic recording, using the selective blockers ω-agatoxin IVa, ω-conotoxin GVIa, and SNX-482 to separate P/Q-, N-, and R-type components. Nonstationary fluctuation analysis combined with blocker application revealed a single MFB contained on average ∼2000 channels, with 66% P/Q-, 26% N-, and 8% R-type channels. Whereas both P/Q-type and N-type Ca²⁺ channels showed high activation threshold and rapid activation and deactivation, R-type Ca²⁺ channels had a lower activation threshold and slower gating kinetics. To determine the efficacy of activation of different Ca²⁺ channel subtypes by physiologically relevant voltage waveforms, a six-state gating model reproducing the experimental observations was developed. Action potentials activated P/Q-type Ca²⁺ channels with high efficacy, whereas N- and R-type channels were activated less efficiently. Action potential broadening selectively recruited N- and R-type channels, leading to an equalization of the efficacy of channel activation. In contrast, subthreshold presynaptic events activated R-type channels more efficiently than P/Q- or N-type channels. In conclusion, single MFBs coexpress multiple types of Ca²⁺ channels, which are activated differentially by subthreshold and suprathreshold presynaptic voltage signals.

Introduction

The mossy fiber–CA3 pyramidal neuron synapse is a key synapse in the trisynaptic circuitry of the hippocampus (Brown and Johnston, 1983; Henze et al., 2002; Nicoll and Schmitz, 2005; Bischofberger et al., 2006a). Although this synapse is widely used as a model to study presynaptic plasticity (Salin et al., 1996; Toth et al., 2000), the basic mechanisms of presynaptic Ca²⁺ signaling are incompletely understood.

Similar to other glutamatergic synapses, multiple types of Ca²⁺ channels contribute to synaptic transmission at hippocampal mossy fiber synapses (Castillo et al., 1994). The selective P/Q-type Ca²⁺ channel blocker ω-agatoxin IVa substantially reduces the amplitude of evoked EPSCs at both mossy fiber–pyramidal neuron and mossy fiber–interneuron synapses (Castillo et al., 1994; Pelkey et al., 2006). However, the selective N-type Ca²⁺ channel antagonist ω-conotoxin GVIa also inhibits synaptic transmission at mossy fiber synapses (Castillo et al., 1994; Pelkey et al., 2006). Finally, blockers of R-type Ca²⁺ channels inhibit mossy fiber long-term potentiation (LTP), although it has remained controversial whether these channels contribute to basal synaptic transmission (Gasparini et al., 2001; Breustedt et al., 2003; Dietrich et al., 2003). Recent imaging experiments suggested a large but highly variable (0–70%) contribution of R-type Ca²⁺ channels to action potential-evoked presynaptic Sr²⁺ transients in hippocampal mossy fiber boutons (MFBs) (Tokunaga et al., 2004; Miyazaki et al., 2005). Thus, multiple types of Ca²⁺ channels contribute to mossy fiber transmission, but their proportions at the level of single boutons and the conditions that lead to their activation have remained unclear.

Several recent studies suggest that electrical signaling in mossy fiber axons and boutons is more complex than previously thought (for review, see Bischofberger et al., 2006a) (see also Debanne, 2004). First, the duration of the presynaptic action potential is not constant, but shows broadening during repetitive stimulation (Geiger and Jonas, 2000). Second, activation of presynaptic GABA_A receptors may exert an excitatory or inhibitory influence on MFBs (Ruiz et al., 2003; Price and Trussell, 2006). Finally, recent experiments revealed that subthreshold EPSPs propagate passively from the somatodendritic domain into the axon of granule cells (Alle and Geiger, 2006). Previous studies suggested that presynaptic Ca²⁺ channels in MFBs are activated efficiently by action potentials (Bischofberger et al., 2002). However, how the different subtypes of Ca²⁺ channels are activated by various voltage waveforms is unknown. This information is essential to understand the coupling between electrical signals and exocytosis or plasticity at hippocampal mossy fiber synapses.

To rigorously address these questions, we performed direct recordings from hippocampal MFBs (Bischofberger et al., 2006b). This approach allows us to accurately measure the presynaptic Ca²⁺ current under voltage-clamp conditions, as previously achieved at the squid giant synapse (Llinás et al., 1981; Augustine et al., 1985). Using presynaptic recordings, we addressed the following questions. First, do P/Q-, N-, and R-type Ca²⁺ channels coexist in a single presynaptic terminal? Second, does the gating differ between Ca²⁺ channel subtypes? Third, how efficiently are the different types of Ca²⁺ channels activated by presynaptic action potentials or subthreshold voltage signals?

Materials and Methods

Presynaptic recording.

Transverse 300-μm-thick slices were cut from the hippocampus of 21- to 22-d-old Wistar rats, using a commercial (Dosaka DTK-1000, Kyoto, Japan) or a custom-built vibratome (Bischofberger et al., 2006b). The animals were killed by decapitation, in accordance with national and institutional guidelines. Slices were kept at 35°C for 30 min after slicing and then at room temperature. For the dissection and storage of slices, a solution containing 87 mm NaCl, 25 mm NaHCO₃, 10 mm glucose, 75 mm sucrose, 2.5 mm KCl, 1.25 mm NaH₂PO₄, 0.5 mm CaCl₂, and 7 mm MgCl₂ was used (equilibrated with 95% O₂/5% CO₂ gas mixture). For experiments, the slices were superfused with physiological saline containing 125 mm NaCl, 25 mm NaHCO₃, 25 mm glucose, 2.5 mm KCl, 1.25 mm NaH₂PO₄, 2 mm CaCl₂, and 1 mm MgCl₂ (95% O₂/5% CO₂).

Patch pipettes were pulled from borosilicate glass tubing (2.0 mm outer diameter, 0.6 mm inner diameter) and filled with a solution containing 145 mm CsCl, 2 mm MgCl₂, 2 mm Na₂ATP, 0.3 mm Na₂GTP, 5 mm Na₂-phosphocreatine, 10 mm EGTA, and 10 mm HEPES (pH adjusted to 7.3 with CsOH); pipette resistance was 7–9 MΩ. Mossy fiber boutons in stratum lucidum of the hippocampal CA3 region at a depth of ∼30 μm from the slice surface were identified by their small diameter (3–5 μm), small capacitance (0.8–2 pF), and high input resistance (>1 GΩ) (Bischofberger et al., 2006b). Patch pipettes were positioned using a Kleindiek micromanipulator (Kleindiek Nanotechnik, Reutlingen, Germany). Currents were recorded with an Axopatch 700B amplifier (Molecular Devices, Sunnyvale, CA). Series resistance (≤50 MΩ) was compensated (60–85% correction, bandwidth 10 kHz). Data were filtered at 6 kHz (internal four-pole low-pass Bessel filter), and sampled at 50 kHz using a 1401 power interface (Cambridge Electronic Design, Cambridge, UK). For display purpose and nonstationary fluctuation analysis, data were passed through an additional 2.5 or 5 kHz low-pass digital filter. Pulse sequences were generated using FPulse (U. Fröbe, Physiologisches Institut der Universität Freiburg, Freiburg, Germany) running under Igor (version 5.01; WaveMetrics, Lake Oswego, OR) on a personal computer. Test pulses were applied at a frequency of 0.2 Hz. To isolate Ca²⁺ currents pharmacologically, 1 μm tetrodotoxin, 20 mm tetraethylammonium chloride (TEA), and 5 mm 4-aminopyridine (4-AP) were added to the bath solution. Capacitive and leakage currents were subtracted using the pipette capacitance compensation circuit of the amplifier and online digital P over −4 correction. Boutons were held at −80 mV throughout the experiment. Recordings were made at 23 ± 2°C.

ω-Conotoxin GVIa and ω-agatoxin IVa were from Bachem (Bubendorf, Switzerland), SNX-482 was from Peptides International (Louisville, KY), and nifedipine was from Sigma (St. Louis, MO). Other chemicals were obtained from Sigma, Merck (Darmstadt, Germany), Riedel-de Haën (Seelze, Germany), or Gerbu (Gaiberg, Germany). Peptide toxins were applied using a recirculation system with a peristaltic pump (Ismatec, Wertheim-Mondfeld, Germany). The total volume of the system was ∼5 ml, and the solution was equilibrated with 95% O₂/5% CO₂. Bovine serum albumin (Sigma) was added at a concentration of 1 mg/ml to prevent adsorption of peptides to the surfaces of the system. In eight experiments, ω-conotoxin GVIa was applied before ω-agatoxin IVa, and in six other experiments, the order of application was reverse. Because the proportions of toxin-sensitive components of the Ca²⁺ current were similar, data were pooled. Because saturating concentrations of toxins and low-frequency stimulation were used, complications resulting from voltage-dependent toxin unblock were highly unlikely (Mintz et al., 1992; Forsythe et al., 1998).

Data analysis.

The proportion of P/Q- and N- components in the presynaptic Ca²⁺ current was quantified by subtracting Ca²⁺ currents in the presence of ω-agatoxin IVa or ω-conotoxin GVIa from those before toxin application and measuring the amplitudes of the difference currents. Putative R-type channels were analyzed in the presence of ω-agatoxin IVa + ω-conotoxin GVIa. Throughout the paper, we refer to the residual current as R-type component, because (1) 500 nm SNX-482 blocked >50% of the residual current (Tottene et al., 2000; Metz et al., 2005), (2) 20 μm nifedipine had no effect on the residual current, excluding L-type channels, and (3) hyperpolarizing prepulses had no effect on the compound Ca²⁺ current, making T-type channels unlikely (see Results). For summary purpose, each proportion was normalized to the sum of all fractions (Fig. 1C). The time course of Ca²⁺ channel activation (0–5 ms after pulse onset) was fitted with an exponential function with delayed onset:

where I₀ is the amplitude, τ_a is the activation time constant, and δ is the delay; the contribution of low-pass filtering (55 μs at 6 kHz) was subtracted. The time course of deactivation (0–3 ms after pulse end) was fitted with a single exponential. Voltage-clamp conditions were assessed from the 20–80% rise time of the tail current after a test pulse to 0 mV; the average value was 37.3 ± 2.0 μs in control conditions and 39.8 ± 2.6 μs in the presence of ω-conotoxin GVIa + ω-agatoxin IVa (n = 19; p > 0.1), indicating adequate voltage clamp for P/Q-, N-, and R-type components throughout the experiment. I–V relations were fitted with a product of a modified Goldman-Hodgkin-Katz equation and a Boltzmann term, giving the following function:

where P is an amplitude factor, C and D determine current rectification and reversal potential (Sala, 1991; Bischofberger et al., 2002), V_mid is the midpoint potential of the steady-state activation curve, and k is the corresponding slope factor.

Figure 1.

Pharmacological dissection reveals three components of the presynaptic Ca²⁺ current in hippocampal MFBs. A, Presynaptic Ca²⁺ current evoked by a 20 ms pulse to 0 mV under control conditions, in the presence of 500 nm ω-agatoxin IVa (ω-aga), and after subsequent addition of 1 μm ω-conotoxin GVIa (ω-cono). The residual component was blocked by 200 μm Cd²⁺, confirming that it was mediated by Ca²⁺ channels. B, Plot of Ca²⁺ current amplitude against time during application of the toxins. The time of application of the blockers is indicated by the horizontal bars. Same bouton as shown in A. C, Summary bar graph illustrating the proportions of ω-agatoxin IVa-sensitive, ω-conotoxin GVIa-sensitive, and toxin-resistant component of the Ca²⁺ current. Symbols (triangles, inverted triangles, and circles) represent individual experiments, and bars represent mean ± SEM. Data are from 12, 9, and 15 boutons. D, Plot of the proportions (Prop.) of the three components against the total Ca²⁺ current amplitude. Note the lack of correlation (Pearson's r, 0.11, −0.01, and 0.28; p > 0.2 in all cases).

For nonstationary fluctuation analysis (Sigworth, 1980), 15–138 Ca²⁺ current traces were analyzed. Ensemble mean versus time was determined by averaging all traces, whereas ensemble variance versus time was measured as the average of variances from overlapping pairs of traces, to minimize errors caused by run-down (Heinemann and Conti, 1992). Variance was plotted against mean for each time point. If the relation showed a parabolic shape, data were fitted with the following function:

where σ² is variance, I is the mean current, i is the single-channel current, N is the number of channels in the patch, and σ₀² is the variance of the baseline noise (Sigworth, 1980). Alternatively, variance-mean data were fitted with the function σ²(I) = iI + σ₀². For the estimation of maximal open probability and channel number, only experiments that could be fitted with the parabolic function were used (8/11 boutons under control conditions, 4/6 boutons in the presence of ω-conotoxin GVIa + ω-agatoxin IVa). This procedure may lead to a slight overestimation of maximal open probability and a corresponding underestimation of channel number.

Data analysis was performed using C-Stimfit (C. Schmidt-Hieber, Physiologisches Institut der Universität Freiburg, Freiburg, Germany), Graphpad Prism 4.0 (Graphpad, San Diego, CA), and Mathematica 4.1.2 or 5.0 (Wolfram Research, Champaign, IL). The values given in the text and the symbols and error bars in the figures indicate mean ± SEM. Traces shown in figures represent averages of 3–10 single sweeps, unless specified differently. Membrane potentials are given without correction for liquid junction potentials. Differences were tested for statistical significance using a two-sided Wilcoxon signed-rank or a two-sided Mann–Whitney test at the significance level (p) indicated.

Modeling of Ca²⁺ channel gating.

For each pharmacologically isolated Ca²⁺ current component, the total dataset comprised of steady-state activation curve, activation time constant and delay (−20 mV or −10 mV to +50 mV), deactivation time constant (−80 mV to −20 mV), and open probability at 0 mV was fitted with a multistate gating model that depicts the macroscopic current properties (Bischofberger et al., 2002; Engel and Jonas, 2005). The model consisted of five closed states and a single open state; all transitions except the last one preceding the open state were assumed to be voltage dependent. Because Ca²⁺ channels in MFBs showed only minimal inactivation during 20 ms test pulses to 0 mV (see Fig. 6D) or after 200–500 ms prepulses to potentials between −120 and −40 mV (not illustrated), inactivated states were not included in the model. Voltage-dependent forward and backward rates were calculated as follows:

where α_i,0 and β_i,0 are the forward and backward rates at 0 mV, k_i is the slope factor, and i = 1 … 4. The sum of squares of differences between experimental observations and model predictions was minimized using a random walk algorithm and subsequently the FindMinimum function of Mathematica. Weight factors were set to the inverse of the maximum value in the dataset for kinetic data and to three times the inverse of this value for activation curve and open probability. The model is based on recordings at 23°C and thus depicts Ca²⁺ channel gating at this temperature.

Figure 6.

Summary of gating properties of presynaptic P/Q-, N-, and R-type channels. A, Semilogarithmic plot of steady-state activation curve of P/Q- (triangles), N- (inverted triangles), and R- (circles) type components. The P_open–V relations were obtained from the I–V relations by dividing the current by the Goldman-Hodgkin-Katz factor in Equation 2, with values of P, C, and D taken from the fit of I–Vs. Note that the activation curves of P/Q and N-type component are very similar, whereas the activation curve of the R-type component is less steep. B, Plot of activation time constant against voltage. C, Plot of deactivation time constant against voltage. Note that both activation and deactivation time constants are similar for P/Q- and N-type components, but slower for the R-type component. Data points in A were fitted with logarithmic Boltzmann functions, and data points in B and C were fitted by exponential functions (for voltages ≥0 mV in B; continuous, dashed, and dotted curves). D, Summary bar graph of ratio of Ca²⁺ current at the end (15–20 ms) and the beginning (5–10 ms) of 20 ms pulses to 0 mV for P/Q-, N-, and R-type components. In all cases, the ratio is close to 1, indicating the lack of inactivation.

To determine the efficacy of Ca²⁺ channel activation by physiological voltage waveforms, an action potential evoked by remote stimulation of the mossy fiber axon was used as a template. The action potential used had a duration at half-maximal amplitude of 0.54 ms [corresponding to the shortest in a sample of six propagated action potentials in our previous study (Bischofberger et al., 2002)]. To represent both variability of half-duration across MFBs and the effects of activity-dependent action potential broadening (Geiger and Jonas, 2000), longer action potentials were then generated from this template by scaling the repolarization phase in time. Subthreshold voltage signals were simulated as exponentially rising and decaying waveforms to approximate the shape of excitatory presynaptic potentials (EPreSPs) passively propagated to MFBs via the mossy fiber axon (Alle and Geiger, 2006).

Results

We characterized the Ca²⁺ channel subtypes expressed in hippocampal MFBs from 21- to 22-d-old rats. To isolate Ca²⁺ currents pharmacologically, 1 μm tetrodotoxin, 20 mm TEA, and 5 mm 4-AP were added to the bath solution. Figure 1A shows representative recordings of compound Ca²⁺ currents evoked by 20 ms test pulses to 0 mV obtained under these conditions. Presynaptic Ca²⁺ currents in MFBs showed activation properties similar to those reported previously (Bischofberger et al., 2002). In addition, Ca²⁺ currents showed only minimal steady-state inactivation tested with 100–500 ms prepulses (−120 mV to −40 mV; five boutons).

Coexistence of multiple Ca²⁺ channel subtypes in single mossy fiber boutons

To determine the contribution of different Ca²⁺ channel subtypes to the compound Ca²⁺ current, we applied the selective P/Q-type Ca²⁺ channel blocker ω-agatoxin IVa and the specific N-type channel blocker ω-conotoxin GVIa at saturating concentrations (Randall and Tsien, 1995; McDonough et al., 2002) (Fig. 1). 500 nm ω-agatoxin IVa blocked 61 ± 3% of the presynaptic Ca²⁺ current in hippocampal MFBs (12 boutons) (Fig. 1B,C). ω-Conotoxin GVIa (1 μm) blocked 24 ± 4% of the presynaptic Ca²⁺ current (nine boutons) (Fig. 1B,C). The remaining component, which had a relative amplitude of 15 ± 2%, was blocked by 200 μm Cd²⁺, indicating that it was mediated by toxin-resistant Ca²⁺ channels. The three components were present in all boutons examined. Whereas the proportion of the P/Q-type component was relatively constant (with a coefficient of variation of 0.19), the proportion of both N- and R-type channels was more variable (with coefficients of variation of 0.55 and 0.39). Regression analysis further revealed that the variability in the proportion of N- and R-type channels was independent of the Ca²⁺ current amplitude (p > 0.2 in all cases) (Fig. 1D).

The ω-agatoxin IVa-resistant, ω-conotoxin GVIa-resistant Ca²⁺ current component could be mediated by R-type, L-type, or (less likely) T-type Ca²⁺ channels (Catterall et al., 2005). We therefore tested the effects of the selective R-type channel blocker SNX-482 (Newcomb et al., 1998) (Fig. 2). After the application of a combination of 500 nm ω-agatoxin IVa and 1 μm ω-conotoxin GVIa, 500 nm SNX-482 was applied. On average, SNX-482 blocked 56 ± 3% of the ω-agatoxin IVa-resistant, ω-conotoxin GVIa-resistant Ca²⁺ current, corresponding to 10 ± 2% of the total Ca²⁺ current (five boutons). Because SNX-482 was previously shown to block only a subpopulation of R-type channels (Tottene et al., 2000; Metz et al., 2005), these results are consistent with the hypothesis that the toxin-resistant component is primarily mediated by R-type Ca²⁺ channels. To further test for the presence of L-type channels, we examined the effects of 20 μm nifedipine after coapplication of 500 nm ω-agatoxin IVa and 1 μm ω-conotoxin GVIa. However, application of 20 μm nifedipine had no detectable effect on the Ca²⁺ current amplitude (three boutons). To further test for the presence of T-type channels, we examined the effects of 100 ms prepulses to −120 mV on the Ca²⁺ current amplitude in the absence of toxins. However, hyperpolarizing prepulses increased the amplitude of the Ca²⁺ current by only 2.4 ± 1.9% of the total current (five boutons). Based on the effects of 500 nm SNX-482, which blocked >50% of the ω-agatoxin IVa-resistant, ω-conotoxin GVIa-resistant Ca²⁺ current, and the lack of effect of both nifedipine and prepulses, we conclude that the residual current is likely to be mediated by R-type channels.

Figure 2.

The agatoxin/conotoxin-resistant Ca²⁺ current component is primarily mediated by R-type channels. A, Presynaptic Ca²⁺ current evoked by a 20 ms pulse to 0 mV under control conditions, in the presence of 500 nm ω-agatoxin IVa (ω-aga) + 1 μm ω-conotoxin GVIa (ω-cono), and after subsequent addition of 500 nm SNX-482. The residual component was completely blocked by 200 μm Cd²⁺, confirming that it was mediated by Ca²⁺ channels. B, Plot of Ca²⁺ current amplitude against time during application of the toxins. The time of application of the blockers is indicated by the horizontal bars. C, Summary bar graph illustrating the proportions of ω-agatoxin IVa + ω-conotoxin GVIa-sensitive, SNX-482-sensitive, and toxin-resistant component of the Ca²⁺ current. Symbols (diamonds, circles, and squares) represent individual experiments, and bars represent mean ± SEM. Data are from seven, five, and five boutons.

Differential gating of presynaptic P/Q-, N-, and R-type channels

To address the possible functional significance of the expression of P/Q-, N-, and R-type Ca²⁺ channels in MFBs, we characterized the voltage dependence of activation of the three types of channels (Fig. 3, Table 1). Test pulses (20 ms) to voltages between −70 and +60 mV were applied, and the corresponding Ca²⁺ currents were measured at the end of each pulse. P/Q- and N-type components were isolated by subtraction of current traces in the absence and presence of either ω-agatoxin IVa or ω-conotoxin GVIa, whereas R-type channels were characterized in the presence of a combination of the two blockers. Analysis of current–voltage relations of the three components revealed that the gating properties of the three components were significantly different. The activation threshold, quantified as the potential in which activation reached 10% of the maximal value, was −18.5 ± 1.0 mV for the P/Q-type component, −14.8 ± 3.4 mV for the N-type component, and −30.7 ± 1.8 mV for the R-type component (p < 0.005 for R-type versus P/Q- or N-type). Furthermore, the midpoint potential of the steady-state activation curve was −4.3 ± 0.7 mV for P/Q-type, −2.6 ± 2.0 mV for N-type, and −7.7 ± 2.1 mV for R-type channels, slightly, but not significantly, different between channels (p > 0.05). Finally, the slope factors of the activation curve were 4.7 ± 0.2 mV for P/Q-type, 4.2 ± 0.5 mV for N-type, and 6.7 ± 0.3 mV for R-type channels (p < 0.01 for R-type versus P/Q- or N-type). Thus, presynaptic R-type channels showed a lower activation threshold and a less steep voltage dependence of activation than P/Q- or N-type channels. P/Q-, N-, and R-type components showed only minimal inactivation during 20 ms pulses; the ratio of Ca²⁺ current at the end of a pulse to 0 mV (15–20 ms after onset) to that at the beginning (5–10 ms) was 1.05 ± 0.03, 1.06 ± 0.03, and 1.15 ± 0.07, respectively.

Figure 3.

Voltage dependence of activation of P/Q-, N-, and R-type components of the presynaptic Ca²⁺ current. A–C, Single traces of P/Q-type (A), N-type (B), and R-type (C) components of the presynaptic Ca²⁺ current obtained by pharmacological isolation and digital subtraction. Pulse protocol: holding potential −80 mV, test pulses to between −70 mV and +40 mV (10 mV increment), and step back to −80 mV. Traces in A–C are from three different boutons. D–F, Average current–voltage relations of the three components. Current amplitudes were measured at the end of the 20 ms pulses and normalized (norm.) to the amplitude at 10 mV (D, E) or 0 mV (F). Data points were fitted with Equation 2 (continuous, dashed, and dotted curves). Data are from six, five, and six boutons.

Table 1.

Gating and conductance properties of P/Q-, N-, and R-type Ca²⁺ channels in hippocampal MFBs

	Vmid (mV)	k (mV)	τact at 0 mV (ms)	τdeact at −80 mV (ms)	i at 0 mV (pA)	Proportion of numbers of channels (%)
P/Q	−4.3 ± 0.7	4.7 ± 0.2	0.79 ± 0.05	0.09 ± 0.01	0.13 ± 0.02a	66
N	−2.6 ± 2.0	4.2 ± 0.5	0.93 ± 0.14	0.06 ± 0.01	0.13 ± 0.02a	26
R	−7.7 ± 2.1	6.7 ± 0.3	1.79 ± 0.28	0.55 ± 0.08	0.21 ± 0.05	8

^aValues were obtained from analysis of the total Ca²⁺ current.

Next, we characterized the kinetics of activation and deactivation (Figs. 4, 5, Table 1). The activation time constant was measured from the onset of the Ca²⁺ current during the test pulses (Fig. 4A), and the deactivation time constant was determined using 20 ms test pulses to 0 mV followed by steps to voltages between −20 and −80 mV (Fig. 5A). Both the activation and deactivation time constants of P/Q- and N-type components were in the submillisecond range and were not significantly different from each other (p > 0.1). In contrast, the activation time constants of R-type channels were significantly slower between −10 and +10 mV (p < 0.05), and the deactivation time constants were substantially slower over the entire voltage range tested (p < 0.05). At −80 mV, the deactivation time constant for the R-type component was 0.55 ms, approximately six times slower than the deactivation of the P/Q-type component (0.09 ms). Thus, whereas the activation and deactivation kinetics of P/Q- and N-type channels were similar, activation and deactivation kinetics of R-type channels were markedly slower.

Figure 4.

Activation kinetics of presynaptic P/Q-, N-, and R-type Ca²⁺ channels. A, Traces of P/Q- (top trace), N- (center trace), and R- (bottom trace) type components of the presynaptic Ca²⁺ current obtained by pharmacological isolation and digital subtraction. Pulse protocol: holding potential −80 mV, test pulse to 0 mV. Traces were fitted with Equation 1 (continuous lines). B–D, Activation time constant (inverted triangles, triangles, and circles) and delay (squares) for P/Q- (B), N- (C), and R- (D) type components. Time constant values for voltages ≥0 mV were fitted with exponential functions, and delay values were fitted by linear regression. Data are from five, five, and seven boutons.

Figure 5.

Deactivation kinetics of presynaptic P/Q-, N-, and R-type Ca²⁺ channels. A, Traces of P/Q- (top trace), N- (center trace), and R- (bottom trace) type components of the presynaptic Ca²⁺ current obtained by pharmacological isolation and digital subtraction. Pulse protocol: holding potential −80 mV, 20 ms test pulse to 0 mV, and step back to −80 mV. Traces were fitted with an exponential function (continuous line). B–D, Deactivation time constant for P/Q- (B), N- (C), and R- (D) type components. Data points were fitted with exponential functions. Data are from five, four, and seven boutons.

In summary, as depicted in Figure 6 and Table 1, P/Q-, N-, and R-type Ca²⁺ channels in MFBs differ in gating properties. Presynaptic R-type channels show a more negative activation threshold, less steep voltage dependence of activation, slower activation, and slower deactivation than P/Q- or N-type channels. Thus, fast, high voltage-activated and slow, low voltage-activated Ca²⁺ channels coexist in single hippocampal MFBs.

Single-channel conductance and channel numbers

We next determined the single-channel conductance (i), the maximal open probability, and the number of Ca²⁺ channels (N) present in a single MFB by nonstationary fluctuation analysis (Sigworth, 1980) (Fig. 7). Ca²⁺ currents were recorded using 20 ms test pulses to 0 mV. The ensemble variance was calculated from the fluctuation of currents, plotted against the ensemble mean, and fitted with a parabolic function (Eq. 3) (Fig. 7B). Under control conditions, the estimated single-channel current was 0.13 ± 0.02 pA, the maximal open probability of the channels at 0 mV was 0.74 ± 0.04, and the estimated number of channels was 2007 ± 455 (11, 8, and 8 boutons). In the presence of ω-agatoxin IVa + ω-conotoxin GVIa, which pharmacologically isolates R-type channels, the estimated single-channel current was 0.21 ± 0.05 pA, slightly, but not significantly, larger than in control conditions (p > 0.1; six boutons). Furthermore, the maximal open probability for the R-type component at 0 mV was 0.84 ± 0.04 (p > 0.1) and the estimated number of channels was 161 ± 41 (four boutons). Assuming that P/Q- and N-type channels have the same single-channel conductance, these results indicate that a large hippocampal MFB contains on average ∼1300 P/Q-, ∼500 N-, and ∼160 R-type channels.

Figure 7.

Single-channel conductance and number of Ca²⁺ channels in hippocampal MFBs. A, C, 10 consecutive traces of Ca²⁺ currents evoked by steps from −80 to 0 mV (black traces) and mean variance calculated from 15 adjacent points (red trace) under control conditions (A) and in the presence of 500 nm ω-agatoxin IVa + 1 μm ω-conotoxin GVIa (C). B, D, Plot of variance against mean under control conditions (B) and in the presence of 500 nm ω-agatoxin IVa + 1 μm ω-conotoxin GVIa (D). Data were fitted by Equation 3 (red curve). Baseline noise was subtracted. The single-channel current and the total number of channels were 0.24 pA and 664 under control conditions (B) and 0.31 pA and 148 in the presence of ω-agatoxin IVa + ω-conotoxin GVIa (D). Data in A and B were obtained from a different bouton from those in C and D. E, Summary bar graph of single-channel current (i) in control conditions and in the presence of ω-agatoxin IVa + ω-conotoxin GVIa. F, Summary bar graph of number of channels per MFB in control conditions and in the presence of ω-agatoxin IVa + ω-conotoxin GVIa.

Differential activation of Ca²⁺ channel subtypes by presynaptic voltage waveforms

To determine the efficacy of Ca²⁺ channel activation by subthreshold and suprathreshold presynaptic voltage signals, we developed kinetic models of the gating of P/Q-, N-, and R-type Ca²⁺ channels (Hille, 2001) (Fig. 8, Table 2). For each pharmacologically isolated Ca²⁺ channel subtype, we fitted the steady-state activation curve (Fig. 8A,C,E), activation time constant and delay, deactivation time constant (Fig. 8B,D,F), and open probability at 0 mV with a gating model that consisted of five closed states and a single open state; all transitions except the last one preceding the open state were assumed to be voltage dependent. The four voltage-dependent transitions were necessary to account for the delay in activation. The final, voltage-independent step was required to describe the maximal open probability of Ca²⁺ channels and the relatively slow and voltage-independent deactivation of the R-type channels (Fig. 8F). As shown in Figure 8, the gating of all three types of channels is closely reproduced by the model. However, for R-type channels, the forward rates from C4 to O and the backward rates from O to C4 were smaller than those of P/Q- and N-type channels. This was necessary to account for the relatively slow and voltage-independent deactivation kinetics of R-type channels.

Figure 8.

Gating models for presynaptic P/Q-, N-, and R-type channels. A, B, Modeling of voltage dependence of steady-state activation (A, open circles), activation time constant (B, open circles), delay (B, open squares), and deactivation time constant (B, filled circles) for presynaptic P/Q-type channels. C–F, Similar analysis for N-type (C, D) and R-type channels (E, F). The scheme on top indicates the structure of the kinetic model. The predictions of the models are indicated by red (activation curve, activation time constant, deactivation time constant) and green curves (delays). For model parameters, see Table 2.

Table 2.

Gating models of P/Q-, N-, and R-type Ca²⁺ channels in hippocampal MFBs

αi,0 (ms−1)	βi,0 (ms−1)	ki (mV)
P/Q-type channel model
α1,0 = 5.89	β1,0 = 14.99	k1 = 62.61
α2,0 = 9.21	β2,0 = 6.63	k2 = 33.92
α3,0 = 5.20	β3,0 = 132.80	k3 = 135.08
α4,0 = 1823.18	β4,0 = 248.58	k4 = 20.86
α = 247.71	β = 8.28
N-type channel model
α1,0 = 4.29	β1,0 = 5.23	k1 = 68.75
α2,0 = 5.24	β2,0 = 6.63	k2 = 39.53
α3,0 = 4.98	β3,0 = 73.89	k3 = 281.62
α4,0 = 772.63	β4,0 = 692.18	k4 = 18.46
α = 615.01	β = 7.68
R-type channel model
α1,0 = 9911.36a	β1,0 = 0.62a	k1 = 67.75
α2,0 = 4.88	β2,0 = 21.91	k2 = 50.94
α3,0 = 4.00	β3,0 = 51.30	k3 = 173.29
α4,0 = 256.41	β4,0 = 116.97	k4 = 16.92
α = 228.83	β = 1.78

^aFor this step, α_1,0is large, and β_1,0 is small. Thus, omitting C0 and the associated transitions would only make minimal differences.

Using the experimentally constrained models of P/Q-, N-, and R-type channel gating, we examined the activation of the three types of channels by physiological voltage waveforms (Fig. 9). We first tested whether the models accurately predicted the Ca²⁺ current evoked by a short propagated action potential (duration at half-maximal amplitude 0.54 ms) in MFBs. Figure 9A shows a Ca²⁺ current activated by a short action potential in an MFB under control conditions and in the presence of ω-agatoxin IVa + ω-conotoxin GVIa, isolating the R-type component. Figure 9B shows the corresponding predictions of the models, which accurately reproduced the experimentally recorded action potential-evoked Ca²⁺ currents. For all three components, the peak of the Ca²⁺ current occurred during the repolarization phase of the action potential, as reported previously for the compound Ca²⁺ current in MFBs at both 23°C (Bischofberger et al., 2002) and 34°C (Geiger and Jonas, 2000) (but see Sabatini and Regehr, 1999).

Figure 9.

Efficacy of Ca²⁺ channel activation by natural presynaptic voltage waveforms. A, Contribution of P/Q-, N-, and R-type channels to the Ca²⁺ current evoked by a short action potential (AP) in a whole-cell recorded MFB. Top traces, action potentials used as voltage-clamp commands. Bottom traces, Ca²⁺ currents evoked by the stimulus waveform under control conditions and in the presence of 500 nm ω-agatoxin IVa + 1 μm ω-conotoxin GVIa. B, Simulated Ca²⁺ currents carried by P/Q-type channels (red curve), N-type channels (green curve), and R-type channels (blue curve) during a short action potential. C, D, Open probability of P/Q-type channels (red curves), N-type channels (green curves), and R-type channels (blue curves), during short (C) and broadened (D) action potential waveforms (black curves). Broadened action potentials were generated by scaling of the repolarization phase in time by a factor of three. E, Plot of peak efficacy of Ca²⁺ channel activation against the scaling factor, normalized (norm.) to the peak efficacy obtained with the unscaled action potential. For P/Q-type Ca²⁺ channels (red circles), the efficacy increased only moderately, whereas for N- (green circles) and R-type Ca²⁺ channels (blue circles) the efficacy increased substantially as the action potential was prolonged. F, Open probability of P/Q-type channels (red curves), N-type channels (green curves), and R-type channels (blue curves), during a subthreshold depolarization with slow exponential rise and decay (rise time constant 20 ms, decay time constant 100 ms), approximating the shape of previously recorded EPreSPs in MFBs (Alle and Geiger, 2006).

We then used the models to determine the efficacy of Ca²⁺ channel activation for the three types of presynaptic Ca²⁺ channels. The short action potential activated P/Q-, N-, and R-type channels with peak efficacies of 0.65, 0.50, and 0.49, respectively (Fig. 9C). Thus, short action potentials activated P/Q-type channels more efficiently than N- or R-type channels. In contrast, a broadened action potential in which the repolarization phase was scaled by a factor of three (Geiger and Jonas, 2000) activated P/Q-, N-, and R-type channels with efficacies of 0.90, 0.82, and 0.76 (Fig. 9D). Thus, broader action potentials activated the three types of channels with more comparable efficacy. When the duration of the action potential repolarization phase was systematically increased by scaling the repolarization phase of the short action potential, the efficacy of P/Q-type channel activation increased only moderately, whereas the efficacy of N- and R-type channel activation increased substantially (Fig. 9E). Finally, EPreSPs to −50 mV activate R-type channels with an efficacy of 1 × 10⁻³, approximately five times more efficiently than P/Q- and N-type channels (2 × 10⁻⁴) (Fig. 9F). Furthermore, a series of five EPreSPs to −50 mV will generate a total Ca²⁺ inflow via R-type channels comparable with that induced by a single action potential (71%). Thus, despite the low number of R-type channels, their activation by subthreshold stimuli is expected to have a large impact on slow Ca²⁺-dependent processes in hippocampal MFBs.

Discussion

We characterized the abundance and kinetics of Ca²⁺ channel subtypes at a hippocampal presynaptic terminal by whole-cell voltage-clamp recording. Three major findings are reported. First, we find that P/Q-, N-, and R-type Ca²⁺ channels coexpressed in single MFBs differ in gating properties. Whereas presynaptic P/Q- and N-type Ca²⁺ channels show high activation threshold and fast gating, presynaptic R-type channels show a lower activation threshold and slower gating. Second, our results show that R-type channels are preferentially activated by broadened action potentials and subthreshold presynaptic voltage signals, which readily explains their selective involvement in mossy fiber plasticity (Breustedt et al., 2003; Dietrich et al., 2003). Finally, using nonstationary fluctuation analysis, we obtained the first quantitative estimates of single-channel conductance, absolute open probability, and numbers of Ca²⁺ channels per bouton. Our results are critically important for understanding how action potentials and subthreshold presynaptic voltage signals trigger and modulate transmitter release from hippocampal MFBs.

Channel types, conductance properties, and channelnumbers

Previous imaging data with Sr²⁺ as a charge carrier suggested a large contribution of R-type Ca²⁺ channels to action potential-induced divalent inflow in MFBs (Tokunaga et al., 2004; Miyazaki et al., 2005). Whereas ω-agatoxin IVa and ω-conotoxin GVIa reduced the amplitude of the Sr²⁺ transients by 33 and 20%, the remaining fraction (47%) was resistant to these blockers but was partially inhibited by SNX-482 (Tokunaga et al., 2004; Miyazaki et al., 2005). In contrast, our direct recording from MFBs revealed that the P/Q-type Ca²⁺ current component (61%) dominates over both N- and R-type components (24 and 15% of the total Ca²⁺ current). These findings appear to be more consistent with the observation that transmitter release at mossy fiber synapses is primarily mediated by P/Q-type Ca²⁺ channels (Castillo et al., 1994; Pelkey et al., 2006).

We further obtained quantitative estimates of both single-channel conductances and channel numbers for the three types of Ca²⁺ channels by nonstationary fluctuation analysis. With Ca²⁺ as a charge carrier at physiological concentrations, the single-channel current at 0 mV was 0.13 pA for the total channel population and 0.21 pA for R-type channels, corresponding to chord conductances of 2.2 and 3.5 pS (assuming a reversal potential of +60 mV). Fluctuation analysis also allowed us to determine the total (i.e., open + closed) number of Ca²⁺ channels in a hippocampal MFB. On average, a single large bouton contained ∼1300 P/Q-, ∼500 N-, and ∼160 R-type channels. In relative terms, 66% of Ca²⁺ channels are P/Q-type, 26% N-type, and 8% R-type. As MFBs, on average, have 29.7 active zones at an age comparable with that used in the present study (four fully reconstructed MFBs at P28) (Rollenhagen et al., 2007) (see also Chicurel and Harris, 1992; Acsády et al., 1998), these total channel numbers correspond to 44 P/Q-type channels, 17 N-type channels, and 5 R-type channels per active zone. It will be interesting to compare these numbers with the number of presynaptic particles in MFB freeze-fracture material, which are thought to be primarily Ca²⁺ channels (Pumplin et al., 1981).

Differential gating of presynaptic Ca²⁺ channel subtypes

Understanding the coupling between presynaptic action potentials and exocytosis at cortical glutamatergic synapses requires quantitative analysis of the gating properties of presynaptic Ca²⁺ channels. Previously we analyzed the kinetic properties of the compound Ca²⁺ current in MFBs (Bischofberger et al., 2002). However, as the new data show, the presynaptic Ca²⁺ current is carried by a functionally heterogeneous population of Ca²⁺ channel subtypes. The properties of these subtypes are difficult to infer from the analysis of recombinant channels, because the subunit composition of the presynaptic Ca²⁺ channels including accessory subunits and other Ca²⁺ channel interaction partners is not precisely known. These factors may substantially affect Ca²⁺ channel gating (for review, see Arikkath and Campbell, 2003). For example, association of Ca²⁺ channel α subunits with α2/δ subunits shifts the activation curve in the hyperpolarizing direction (Felix et al., 1997), whereas association with syntaxin was reported to enhance inactivation (Bezprozvanny et al., 1995; Hurley et al., 2004). Thus, direct analysis of gating of native Ca²⁺ channels in presynaptic terminals is required.

The electrotonic structure of the mossy fiber terminals and the presynaptic location of the Ca²⁺ channels allowed us to study their gating properties under almost ideal voltage-clamp conditions. The rise time of the Ca²⁺ tail currents, which are almost instantaneous, indicates that the Ca²⁺ channels are located in mossy fiber terminals rather than adjacent axons. For the MFB, this means that the speed of voltage clamp is limited by the kinetics of the charging of the bouton, which occurs with a time constant τ = R_s × C_MFB ≈ 15 MΩ × 2 pF = 30 μs, where R_s is the uncompensated series resistance and C_MFB is the bouton capacitance (Geiger and Jonas, 2000; Hallermann et al., 2003) (see Materials and Methods). Thus, direct recording from hippocampal MFBs allows us to study the gating of identified Ca²⁺ channel subtypes with unique temporal resolution.

Under these near-optimal voltage-clamp conditions, we found subtle differences between the gating of P/Q- and N-type channels, but markedly different functional properties of R-type channels. Presynaptic N-type channels showed a slightly, but not significantly, more positive midpoint potential of activation curve, but almost identical activation and deactivation kinetics than P/Q-type channels. These findings are consistent with the slight differences between native and recombinant P/Q- and N-type channels reported previously (Wheeler et al., 1996; Colecraft et al., 2000). In contrast, presynaptic R-type channels in MFBs showed a markedly lower activation threshold and a shallower steady-state activation curve than P/Q- or N-type channels, consistent with previous data for recombinant channels (Bourinet et al., 1996). Furthermore, presynaptic R-type channels showed slower activation and deactivation kinetics than P/Q- or N-type channels. Thus, presynaptic Ca²⁺ channel types differ in gating properties, suggesting differential involvement in synaptic transmission.

The present study also suggests several differences between the gating of Ca²⁺ channels in the hippocampal MFB and the calyx of Held. First, the midpoint potential of activation of all Ca²⁺ current components in MFBs is more positive than that of the total Ca²⁺ current in the calyx (−23 mV) (Borst and Sakmann, 1998). Second, in MFBs the activation threshold of R-type channels is lower than that of P/Q- and N-type channels, whereas in the calyx the voltage-dependencies of the Ca²⁺ channel components are more similar (Wu et al., 1999). Finally, in MFBs the coexistence of P/Q-, N-, and R-type channels is maintained at relatively mature developmental stages. In contrast, in the calyx of Held, N- and R-type channels coexist with P/Q-type channels until postnatal day 10 and are downregulated later in development (Iwasaki et al., 2000). These findings suggest that the Ca²⁺ channel properties are highly specialized in the MFB, but more homogeneous in the calyx, especially in mature synapses.

Differential recruitment of Ca²⁺ channels by presynaptic voltagewaveforms

Several recent studies suggest that signaling in mossy fiber axons and MFBs is highly complex (for review, see Bischofberger et al., 2006a) (see also Debanne, 2004). First, the duration of the presynaptic action potential is not constant, but shows broadening during repetitive stimulation (Geiger and Jonas, 2000). Action potential broadening has been also reported in neocortical axons (Shu et al., 2006), suggesting a general feature of cortical axons and presynaptic terminals. Second, recent experiments revealed that subthreshold synaptic events propagate passively from the somatodendritic domain of granule cells to MFBs (Alle and Geiger, 2006). This may suggest that information in the mossy fiber system is coded by both action potential (“digital”) and subthreshold (“analog”) signals. Both spike duration coding and analog coding may enrich the computational repertoire of mossy fiber axons (Alle and Geiger, 2006; Bischofberger et al., 2006a).

How do these electrical events activate the different types of presynaptic Ca²⁺ channels in MFBs (Sabatini and Regehr, 1999)? Whereas short action potentials activate P/Q-type channels with higher efficacy than N-type or R-type channels, broadened action potentials activate the different channel types with more similar efficacy. Hence, action potential broadening will lead to selective recruitment of N- and R-type Ca²⁺ channels. Furthermore, subthreshold voltage signals will selectively activate R-type channels. For a subthreshold EPreSP reaching a peak amplitude of −50 mV, the open probability of R-type channels is approximately five times larger than that of P/Q- and N-type channels. In the calyx of Held, a minimal Ca²⁺ inflow through P/Q-type Ca²⁺ channels activated by subthreshold stimuli induces a significant increase in synaptic strength (Awatramani et al., 2005). In the MFB, the larger Ca²⁺ inflow through R-type Ca²⁺ channels is expected to increase synaptic strength even more profoundly, thus contributing to posttetanic potentiation or LTP at mossy fiber synapses (Salin et al., 1996; Breustedt et al., 2003; Dietrich et al., 2003). Because granule cells in the behaving animal fire with low frequency (Jung and McNaughton, 1993), but receive rhythmic excitatory synaptic input during theta oscillations (Ylinen et al., 1995), Ca²⁺ inflow through R-type channels activated by subthreshold events may be relevant for mossy fiber plasticity in vivo.