Voltage-dependent G-protein inhibition of presynaptic Ca2+ channels is a key mechanism for regulating synaptic efficacy. G-protein βγ subunits produce such inhibition by binding to and shifting channel opening patterns from high to low open probability regimes, known respectively as “willing” and “reluctant” modes of gating. Recent macroscopic electrophysiological data hint that only N-type, but not P/Q-type channels can open in the reluctant mode, a distinction that could enrich the dimensions of synaptic modulation arising from channel inhibition. Here, using high-resolution single-channel recording of recombinant channels, we directly distinguished this core contrast in the prevalence of reluctant openings. Single, inhibited N-type channels manifested relatively infrequent openings of submillisecond duration (reluctant openings), which differed sharply from the high-frequency, millisecond gating events characteristic of uninhibited channels. By contrast, inhibited P/Q-type channels were electrically silent at the single-channel level. The functional impact of the differing inhibitory mechanisms was revealed in macroscopic Ca2+ currents evoked with neuronal action potential waveforms (APWs). Fitting with a change in the manner of opening, inhibition of such N-type currents produced both decreased current amplitude and temporally advanced waveform, effects that would not only reduce synaptic efficacy, but also influence the timing of synaptic transmission. On the other hand, inhibition of P/Q-type currents evoked by APWs showed diminished amplitude without shape alteration, as expected from a simple reduction in the number of functional channels. Variable expression of N- and P/Q-type channels at spatially distinct synapses therefore offers the potential for custom regulation of both synaptic efficacy and synchrony, by G-protein inhibition.
Voltage-dependent G-protein inhibition of N- and P/Q-type channels is a ubiquitous and potent mechanism for presynaptic inhibition of neurotransmitter release (Lipscombe et al., 1989; Wu and Saggau, 1994; Takahashi et al., 1996;Wu and Saggau, 1997). Such inhibition is believed to result from direct binding of G-protein βγ subunits (Gβ γ) to N- and P/Q-type channels (Herlitze et al., 1996; Ikeda, 1996; Zhang et al., 1996; De Waard et al., 1997; Zamponi et al., 1997), which then shifts gating from willing (normal) to reluctant (inhibited) modes of opening (Bean, 1989; Boland and Bean, 1993; Patil et al., 1996).
Whether Gβ γ-bound channels remain electrically silent or can open and close with distinctive kinetics (reluctant activity) during action potentials would lead to divergent effects on the waveform of Ca2+ entry (Brody et al., 1997). The waveform would be reduced in amplitude but maintain the same shape in the electrically silent case. By contrast, the waveform could exhibit both diminished amplitude and altered shape in the latter case. These two scenarios could impact the timing and efficacy of synaptic transmission in fundamentally different ways (Bertram et al., 1996;Borst and Sakmann, 1998; Sabatini and Regehr, 1999).
We have recently suggested, based on indirect whole-cell recordings, that N-type, but not P/Q-type channels can open in the reluctant mode (Colecraft et al., 2000). The strength of this suggestion is critically dependent on direct, single-channel resolution of this core contrast in the prevalence of reluctant openings. Presently, however, no single-channel records of P/Q-type channels during G-protein inhibition are available. Moreover, despite several single-channel studies of inhibited N-type channels (Lipscombe et al., 1989; Delcour and Tsien, 1993; Carabelli et al., 1996; Patil et al., 1996; Lee and Elmslie, 2000), there is no clear consensus about the prevalence of reluctant openings. Particularly challenging issues for such single-channel experiments include: (1) the requirement for unambiguous isolation of unitary N-type or P/Q-type channel activity (Lipscombe et al., 1989;Delcour and Tsien, 1993; Elmslie, 1997); (2) additional, voltage-independent inhibitory mechanisms that could produce activity mimicking reluctant gating (Hille, 1992; Luebke and Dunlap, 1994;Carabelli et al., 1996); and (3) the presumed voltage dependence of reluctant openings (Colecraft et al., 2000), which could render the usual voltage range attainable in single-channel studies (Carabelli et al., 1996; Patil et al., 1996) inadequate for resolution of reluctant gating (but see, Lee and Elmslie, 2000).
Here, we report the single-channel mechanisms underlying voltage-dependent G-protein inhibition of recombinant N- and P/Q-type channels, as reconstituted in human embryonic kidney (HEK 293) cells. This system permits unambiguous identification of channel type and reconstitution of a purely voltage-dependent G-protein inhibitory mechanism (Patil et al., 1996; Colecraft et al., 2000). Using quartz pipette technology to enhance resolution of single-channel currents at higher voltages, we provide direct evidence for the occurrence of reluctant openings in N-type channels and their absence in P/Q-type channels. In addition, reluctant openings of N-type channels activated fast enough to alter the shape of Ca2+currents evoked by neuronal action potential waveforms, whereas corresponding P/Q-type currents exhibited no such behavior.
MATERIALS AND METHODS
Transfection of HEK 293 cells. HEK 293 cells were maintained as previously described (Brody et al., 1997) and were transiently transfected using the calcium phosphate precipitation method with 8 μg each of calcium channel subunit cDNAs encoding rat brain α1A (Starr et al., 1991) or human brain α1B (Williams et al., 1992), rat brain β2a (Perez-Reyes et al., 1992), and rat α2δ (Tomlinson et al., 1993), together with 2 μg of m2 muscarinic receptor cDNA (Peralta et al., 1987). The use of the β2a subunit minimized voltage-dependent inactivation (Patil et al., 1998; Colecraft et al., 2000), allowing clear resolution of the slow phase of current activation and prepulse facilitation, during G-protein inhibition.
Electrophysiological recordings. Both whole-cell and single-channel currents were recorded at room temperature, 2–3 d after transfection, using an Axopatch 200A patch-clamp amplifier (Axon Instruments, Foster City, CA). For whole-cell recordings, cells were continuously perfused with a bath solution containing in mm: 150 tetraethylammonium methanesulfonate (TEA-MeSO3), 2 CaCl2, 1 MgCl2, and 10 HEPES, adjusted to pH 7.4 with TEA-OH. Where noted, 50 μm carbachol was added to the bathing medium to activate G-proteins via the cotransfected m2 receptor. The internal solution contained (in mm): 135 cesium methanesulfonate (Cs-MeSO3), 5 CsCl, 10 EGTA, 1 MgCl2, 10 HEPES, and freshly added 4 MgATP, pH 7.3, adjusted with CsOH. Electrodes were fashioned from borosilicate glass capillaries (WPI MTW 150-F4) and had pipette resistances ranging from 1.2 to 2.5 MΩ, before series resistance compensation of 75%. Voltage pulses were applied at 30 sec intervals. Currents were filtered at 5 kHz for action potential (AP) trains, or at 10 kHz for tail current measurements (four-pole Bessel), and digitized at 25 or 50 kHz, respectively. Voltages were corrected for a liquid junction potential of −11 mV before recording, and leak and capacitance transients subtracted by a P/8 protocol. AP voltage templates were based on recordings from the Calyx of Held (Borst et al., 1995), as previously described (Patil et al., 1998).
Single-channel and macropatch currents were obtained in the cell-attached patch-clamp configuration. To zero the membrane potential, the bath contained a solution consisting of (in mm): 132 K-glutamate, 5 KCl, 5 NaCl, 3 MgCl2, 2 EGTA, 10 glucose, and 10 HEPES, pH 7.4, adjusted with KOH. The pipette solution contained 90 mm BaCl2, 20 mm TEA-MeSO3, 50 μmcarbachol (where indicated), and 10 mm HEPES, pH 7.4, adjusted with TEA-OH. Patch pipettes were fashioned from fused quartz capillaries (Garner Glass Company, Claremont, CA) using a laser puller (Sutter Instrument, Novato, CA) and coated with Sylgard. Pipettes typically had resistances ranging from 5 to 20 MΩ when filled with pipette solution. Voltage pulses were applied at a repetition interval of 6 sec; data were sampled at 40 μsec intervals and filtered at 2 kHz (single-channel) or 5 kHz (single-channel and macropatch) in some instances (−3 dB, four-pole bessel). Reported voltages have been corrected for a liquid junction potential of −17 mV. Smooth functions fitted to leak and capacitative transients (P/8 protocol) were subtracted from single-channel records in a semiautomatic manner using custom software written in Matlab (MathWorks, Natick, MA).
Single-channel data analysis. All patches analyzed in this study contained a single active channel. Custom analysis software, written in Matlab, was used throughout. Records were additionally filtered to a 1.8 kHz bandwidth using a digital Gaussian filter and idealized using half-height criteria. Idealized records were used for generation of histograms, ensemble averages, andP oo distributions. Ensemble average currents, first-latency distribution (FL), and conditional open probability (P oo) were calculated as previously described (Imredy and Yue, 1994). Open and shut time histograms were fitted to the indicated number of exponentials using the method of maximum likelihood (Colquhoun and Sigworth, 1995), with the bin size maintained at the sampling interval (40 μsec) throughout. The maximum number of significant exponentials required to best fit each dwell time histogram, was determined by the likelihood ratio test (McManus and Magleby, 1988). The first two or three bins of each open or shut dwell time histogram, respectively, was excluded from the analysis to prevent distortion by events near the dead time.
Statistics. Throughout, pooled data are presented as mean ± SEM. All p values were calculated from two-tailed t-test; p < 0.05 was considered as significant.
Large surface-potential shifts complicate physiological interpretation of single-channel data
To facilitate resolution of single, Ca2+-channel gating events with adequate frequency bandwidth, pipette solutions typically contain ≥90 mm Ba2+ to increase unitary current amplitude. Such high concentrations of divalent cations screen negative surface charges much more effectively than would physiological concentrations of Ca2+ (Frankenhaeuser and Hodgkin, 1957; Krafte and Kass, 1988; Green and Andersen, 1991), resulting in potentially large, but as-yet-undetermined depolarizing voltage shifts in the gating properties observed at the single-channel level. The unspecified magnitude of this voltage shift constituted a particularly critical ambiguity in this study, because earlier macroscopic current experiments (using physiological Ca2+) indirectly hinted that the prevalence of reluctant openings may increase sharply with depolarization (Colecraft et al., 2000). Hence, in testing for reluctant openings at the single-channel level, it was essential to establish the corresponding voltages over which unitary reluctant openings would be expected to occur.
High-level expression of recombinant N-type (α1B β2aα2δ) channels in HEK 293 cells permitted robust determination of the sought after voltage shift, by enabling direct comparison of tail-activation curves obtained in whole cells using 2 mm Ca2+ as charge carrier, with those obtained in cell-attached macropatches using 90 mm Ba2+. Measurements of well resolved tail currents for the two configurations (Fig.1 A,B) facilitated construction of corresponding tail-activation curves (Fig. 1 C, symbols). Fits of Boltzmann functions with identical slope but differing voltage midpoints (smooth curves) established a large ∼50 mV depolarizing shift of the single-channel (filled circle) compared with whole-cell (open circle) activation curves. A ∼55 mV depolarizing shift of P/Q-type (α1A β2aα2δ) channel-gating properties was established by analogous methods (Fig. 1 D). Thus, the physiological voltages corresponding to single-channel experiments could be obtained by subtracting ∼50 mV from the values reported in the figures and in the text.
Single-channel gating properties of uninhibited N-type channels
The surprisingly large surface-potential shift of voltage-dependent gating under single-channel recording conditions challenged us to undertake considerably stronger single-channel depolarizations than previously attempted, to establish gating properties with greater relevance for physiological behavior, and to test appropriately for unitary reluctant openings. Such stronger depolarization would incur substantial drops in electrochemical driving force with resulting diminution of unitary current amplitudes. Nonetheless, stronger depolarization was an absolute requirement here, given that our traditional single-channel experiments (Patil et al., 1996), nominally performed with +20 mV steps, would correspond to physiological depolarizations of only −30 mV, well below the potentials required for either the bulk of activation (Fig. 1 C, open circles) or the anticipated occurrence of reluctant openings (Colecraft et al., 2000).
As a prelude to G-protein modulation experiments, Figure2 illustrates good resolution of the baseline characteristics of single, unmodified N-type channels, despite test-pulse depolarization to +45 mV. Quartz pipettes proved invaluable, here and throughout, in reducing noise sufficiently to distinguish elementary events during strong depolarization (Levis and Rae, 1998). Under these control conditions, N-type channels activated quickly after the start of the voltage step and subsequently displayed a dense bursting pattern (Fig. 2 A), with millisecond mean open times (Fig. 2 E). Ultimately, these features gave rise to fast-activating and maintained ensemble average currents (Fig.2 B). As expected for unmodified channels, a large depolarizing prepulse had no appreciable effect on either the gating pattern of single-channel traces or on the amplitude and kinetics of the ensemble average current (Fig. 2 A,B). These visual impressions were entirely confirmed by in-depth kinetic analysis of channel gating statistics averaged from multiple patches (Fig.2 C–F). The first-latency distribution (FL) (Fig. 2 C), which represents activation properties by plotting the probability of first opening before timet in the test pulse, was identical in the presence (black line) or absence (gray line) of a prepulse. Also unaffected by prepulse depolarization was the conditional open probability (P oo) (Fig.2 D), which reflects aggregate gating behavior after first opening. Here and throughout,P oo is the channel open probability determined with a delay t after first opening. Together,FL and P oo specify the total gating characteristics of the channel (Imredy and Yue, 1994;Patil et al., 1996), and their invariance therefore demonstrates the lack of effect of a prepulse on the baseline gating of N-type channels. Similarly, traditional open and closed time distributions (Fig.2 E,F) were essentially unchanged by a depolarizing prepulse.
Beyond demonstrating the feasibility of acquiring single-channel data at strongly depolarized test potentials, these results also provided immediate dividends by permitting the first calibration of macroscopic tail-activation curves (Fig. 1) in terms of absolute open probabilitiesP o, a long-sought-after feat with important implications for the physiological performance of presynaptic Ca2+ channels (Borst and Helmchen, 1998). The steady-state P o under single-channel conditions at +45 mV could be determined directly by dividing the unitary current amplitude into the plateau level of ensemble average currents (Figs. 2 B, 5 B), yielding a P o ∼0.3 or 0.1 for N- and P/Q-type channels, respectively. Plotting these values at the appropriate positions on the tail-activation curves in Figure 1,C and D (open squares), enabled overall calibration of curves in terms of open probability, shown on the right axes.
Single-channel detection of reluctant N-type channel openings during G-protein inhibition
We next turned to single-channel investigation of N-type channels during G-protein inhibition, using strong depolarizations as might be required to elicit reluctant openings (Fig.3). The protocols and display format were identical to those used in the controls (Fig. 2), except that G-protein activation was produced by the inclusion of saturating 50 μm carbachol (CCh) in the patch pipette to activate coexpressed m2 muscarinic receptors (Patil et al., 1996).
Strong G-protein modulation was immediately evident in ensemble average currents (Fig. 3 B), as the waveform observed without a prepulse was initially small and exhibited a slow phase of activation (“kinetic slowing”; Luebke and Dunlap, 1994). Furthermore, a prepulse transiently removed inhibition, yielding a large initial amplitude and no slowly activating component (“prepulse facilitation”; Elmslie et al., 1990; Boland and Bean, 1993; Patil et al., 1996; Colecraft et al., 2000). Inspection of single-channel records (Fig. 3 A) pointed to the underlying basis for these macroscopic hallmarks of G-protein inhibition. Traces corresponding to inhibited channels (Fig. 3 A, left column, without prepulse) frequently displayed a considerable delay to first opening after the start of the voltage step (traces 1, 2, 3, and 5), and subsequent gating was usually dominated by brief, submillisecond openings (Fig.3 G) interspersed by long closed periods. Within a sweep, this sparse gating mode could persist throughout the test pulse (traces 2 and 5), or convert to a dense-bursting pattern (trace 3) that appeared identical to that seen in unmodulated channels (Fig. 2). This long delay to the onset of dense bursting rationalized the small initial amplitude and subsequent slow-activating phase of ensemble average currents. Single-channel traces after a prepulse (Fig.3 A, right column, with prepulse) differed dramatically, usually showing rapid activation to a dense-bursting pattern (traces 1, 2, 4, and 5) that appeared identical to that in unmodulated channels (Fig. 2). The prepulse-induced conversion of single-channel gating behavior toward its unmodulated form provides an explanation for the prepulse facilitation evident in ensemble average currents.
Averages from multiple patches confirmed the features suggested above by exemplar single-channel sweeps. First, the existence of a large fraction of slow-activating sweeps is supported by the slow component of the FL without prepulse (Fig. 3 C, gray curve). Such sweeps are virtually absent after a prepulse, as evident from the fast and uniform rise of the correspondingFL curve (black curve), which is closely similar to those in control (Fig. 2 C). Second, the persistence of a sparse gating mode in sweeps without prepulse is evident from the maintained deficit of the corresponding conditional open probability (P oo) (Fig. 3 D, gray trace), compared with the P oocomputed for sweeps with prepulse (black trace). The virtual absence of sparse gating after prepulse is established by the close similarity of this P oo for sweeps with prepulse (Fig. 3 D, black trace) and those for control sweeps (Fig. 3 D, fitted curve). Third, the indistinguishability of the dense-burst gating in sweeps with and without prepulse during G-protein modulation, as well as in control patches, is corroborated by the near identity of open and closed time distributions for the various conditions (Figs. 2, 3, panels E and F). Although computed from all events, these distributions are largely identical to those that would be obtained from pure dense-burst gating, because the sparse gating pattern contributes a comparatively negligible number of open and closed events (Fig. 3 A, left column). Many of these single-channel characteristics have been previously observed and shown to account for the characteristics of macroscopic current waveforms (Patil et al., 1996), albeit at far less depolarized potentials.
The important new finding came with the sparse gating pattern observed above (Fig. 3 A, left column, sweeps 2, 3, 4), which featured openings with properties expected of the reluctant events whose existence we sought to determine. First, sparse gating was only appreciable where G-protein inhibition was expected to favor channel occupancy of the reluctant mode: present in modulated patches when observed without a prepulse, and essentially absent after a prepulse, or in control patches (where little basal modulation is present;Colecraft et al., 2000). Second, fitting with the anticipated voltage-dependent prevalence of reluctant openings (Colecraft et al., 2000), sparse gating was present in G-protein-modulated patches with strong, single-channel depolarization to +45 mV, but absent with weaker depolarization to +20 mV (Patil et al., 1996). Third, when present, sparse gating appeared as the first type of activity in a sweep, before conversion to dense-burst gating. The converse was seldom observed, as would be predicted from the inferred steady-state predominance of willing versus reluctant mode occupancy for these potentials (Colecraft et al., 2000). Finally, the brief openings and long closures of sparse gating were clearly consistent with the anticipated lowerP o of reluctant mode gating (Colecraft et al., 2000). All these reasons argue that the sparse gating activity observed in single-channel records represents direct resolution of gating in the reluctant mode.
Beyond supporting the existence of reluctant gating, the single-channel data could also provide in-depth kinetic information about such gating in isolation, after visual selection for sweeps (without prepulse) that started out in the sparse gating pattern. Lifetime distributions for the first openings of selected sweeps, which should mainly reflect reluctant gating, indicate brief, clearly submillisecond open times (τ = 0.22 msec) (Fig. 3 G, reluctant). Likewise, the P oo for selected sweeps, which should mostly reflect reluctant gating for at least tens of milliseconds (Fig. 3 D), decays rapidly to a plateau value of ∼0.02 (Fig. 3 H, reluctant). This value enables direct calculation of the steady-state open probability in the reluctant mode (Fig. 1 C, legend, triangle), providing the first explicit comparison of the steady-state open probabilities of gating in the reluctant versus willing mode. Reassuringly, kinetic analysis for the remainder of sweeps (without prepulse) recapitulated behavior expected of willing mode gating, as represented by the fits in Figure 3,G and H (willing). Such concordance bolstered confidence that our visual sorting of gating patterns was unbiased. Overall, the briefer open times and precipitousP oo decay for reluctant gating could produce altered waveforms of Ca2+ entry during action potential-evoked activity, if reluctant openings could be evoked by such brief stimuli.
Impact of reluctant openings of N-type channels at physiological potentials
Indeed, the remaining challenge for proposals of a physiological contribution of reluctant openings was that they might not activate fast enough to contribute during the brief 1–4 msec duration of neuronal action potentials; the slow phase of the FLdistribution that tallies the first occurrence of such openings (Fig.3 C, gray) fits a time constant of 116 msec. However, the +45 mV voltage step used in Figure 3 corresponds to −5 mV in 2 mm Ca2+ (Fig. 1), which is well short of the +30–35 mV amplitude of neuronal action potentials. Given the presumed voltage dependence of the activation of reluctant openings (Boland and Bean, 1993; Jones and Elmslie, 1997;Colecraft et al., 2000), we investigated whether they could activate fast enough at higher voltages, using the strongest single-channel depolarizations that permitted resolution of unitary activity.
Figure 4 summarizes the results of single-channel experiments on inhibited N-type channels, using voltage steps to +60 mV (corresponding to +10 mV in 2 mmCa2+). The protocols and format are otherwise identical to those in Figure 3. Despite the decrease in unitary current amplitudes (≤0.5 pA) caused by the stronger depolarization, single-channel activity was still clearly resolved, including the sparse gating pattern believed to correspond to reluctant gating (Fig. 4 A, left column, sweeps 1, 2, 4, and 5). As observed previously with depolarizations to +45 mV (Fig. 3), all of the data and analysis here (Fig. 4) supported the identification of sparse gating as reluctant mode activity, as well as the same elementary basis for G-protein modulation. Two important quantitative differences concerned the increased likelihood of evoking reluctant gating at physiological potentials. First, the steady-state probability of reluctant opening was now increased to ∼0.07 (Fig. 4 H, reluctant), consolidating projections favoring an appreciable reluctant open probability near the peak of action potentials (Fig. 1 C, triangles). Second, the activation of presumed reluctant openings was markedly accelerated, as evident from exemplar traces (Fig.4 A, left) and the near coalescence of slow and rapid phases of the FL distribution (Fig. 4 C, gray). At even greater depolarizations approaching action potential peaks, the activation of reluctant openings would be accelerated still further. Whereas single-channel experiments could not be performed explicitly at these higher physiological potentials, because of reduced signal amplitude, extrapolation of the results in Figures 3 and 4 strongly established the likelihood that reluctant N-type channel openings would contribute during neuronal action potentials.
Before considering subsequent experiments that tested directly for such a physiological contribution, we next turned to single-channel experiments that established fundamental differences in the G-protein modulation of P/Q-type channels. This order permitted subsequent, side-by-side assessment of the physiological impact of differences in the behavior of these two channel types.
Absence of reluctant openings in G-protein-inhibited P/Q-type channels
On the basis of previous whole-cell experiments (Colecraft et al., 2000), we have suggested that G-protein inhibition of P/Q-type and N-type channels may differ in two key regards. First, N-type channels exhibit a greater degree of inhibition than P/Q-type channels in a variety of preparations (Bourinet et al., 1996; Zhang et al., 1996;Currie and Fox, 1997; Roche and Treistman, 1998; Colecraft et al., 2000). The second postulated difference, based on indirect inferences from whole-cell data (Colecraft et al., 2000), concerns a more fundamental qualitative difference in that only N-type channels seem to open in the reluctant mode.
To directly investigate these potential differences, we conducted experiments on single P/Q-type channels during G-protein inhibition (Fig. 5), using the identical protocols and format as with N-type channels (Fig. 3). The ensemble average current of P/Q-type channels displayed clear-cut kinetic slowing and prepulse facilitation (Fig. 5 B), indicating strong G-protein modulation. Nevertheless, close inspection of ensemble currents indicated that the degree of inhibition of P/Q-type channels was clearly less than that obtained for N-type channels (Fig.3 B), demonstrating that the weaker inhibition of macroscopic currents has its basis at the level of individual channels. Single-channel records (Fig. 5 A) contrasted sharply from those of N-type channels, in that there was no hint of the sparse gating pattern attributed to reluctant openings (Fig. 5 D), despite clear delays in the onset of first opening in sweeps without prepulse (Fig. 5 A, traces 1, 2, 4, and 5; also Fig. 5 C). Gating subsequent to first opening always followed a dense-burst pattern that was unchanged by prepulse and identical to that in unmodulated controls, as established explicitly by the overlay of P oo for the various conditions (Fig. 5 D) and the similarity of duration histograms (Fig.5 E,F).
Convolution analysis (Aldrich et al., 1983; Yue et al., 1990) provided a further powerful test that slowing of first openings was the sole mechanism of G-protein inhibition of P/Q-type channels and that a sparse gating mode was absent. If these features held true, then the convolution of the following two entities should predict the ensemble average current waveform without prepulse (Fig. 5 B, left): (1) The derivative of the FL without prepulse (dFL/dt -pre), which incorporates the slowing of first latency in sweeps without prepulse. (2) The conditional open probability with prepulse (P oo,+pre), which would faithfully represent gating after first opening in sweeps lacking sparse gating. This prediction was met precisely, as shown by the overlay of the convolution (Fig. 5 B, smooth curve) and the data. By contrast, if sparse gating were present in sweeps without prepulse, then the same convolution would overpredict the ensemble average data, because P oo,+pre would overestimate the propensity for being open after the first event, given the greater prevalence of sparse gating. In fact, such a convolution exceeded the ensemble average data for N-type channels (Figs. 3 B,4 B, left, smooth curve), where sparse gating was in fact prevalent. The convolutions in Figures 3 Band 4 B (−prepulse) also provide important information regarding the rate of activation of reluctant openings, because they would begin to deviate from the ensemble average current at the onset of such reluctant gating activity. With +45 mV depolarization, the convolution overlaid the early part of the ensemble average current, but began to deviate later on in the pulse (Fig.3 B, left) suggestive of slowly activating reluctant gating activity. With +60 mV depolarization, however, the convolution deviated from the ensemble average current very early in the test pulse (Fig.4 B, left), consolidating our argument that reluctant openings occur with much faster activation kinetics at this elevated voltage. Reassuringly, in all cases where gating would be almost exclusively of the dense-bursting (willing) pattern, the convolution of dFL/dt +preand P oo,+pre precisely predicted ensemble average currents (Figs. 3 B, 4 B,5 B, with prepulse). These results clearly demonstrated the absence of reluctant openings in P/Q-type channels, at least with +45 mV depolarizations.
Because of the faster gating kinetics of P/Q-type channels, it was not possible to routinely obtain well resolved, single-channel records at +60 mV, as was possible for N-type channel (Fig. 4). However, the following macroscopic current experiments confirmed the relative paucity of reluctant openings of P/Q-type channels at voltages greater than +45 mV.
Physiological impact of the differential occurrence of reluctant openings
To explore the physiological implications of the contrasting prevalence of reluctant openings in N-type and P/Q-type channels, we investigated the effects of G-protein inhibition of whole-cell Ca2+ currents evoked by action potential waveforms. If reluctant openings of N-type channels activate fast enough to occur during neuronal action potentials, their briefer open times and lower open probability suggests that G-protein inhibition would serve to both decrease the amplitude and alter the shape of Ca2+ entry waveforms during physiological stimulation. By contrast, the absence of reluctant openings in P/Q-type channels predicts that G-protein modulation would simply reduce the number of functional channels, resulting in decreased amplitude of the Ca2+ entry waveform without shape alteration.
Direct tests of these predictions are summarized in Figure6. Whole-cell Ca2+ currents were elicited by action potential waveforms modeled after those recorded in the calyx of Held (Borst et al., 1995), which in this case spanned a variety of widths at half-maximal amplitude (Fig. 6 A). Corresponding N-type and P/Q-type currents are shown below (Fig.6 B,C, respectively). Under control conditions (−CCh, black traces), N-type channels had broader waveforms of Ca2+ entry than did P/Q-type channels, consistent with the generally slower kinetics observed above in the single-channel experiments. Activation of G-proteins with 50 μm CCh (+CCh, dotted traces) markedly decreased the amplitude of Ca2+current carried by either channel type, although the anticipated stronger inhibition of N-type channels was clearly apparent.
To test for effects on the shape of Ca2+entry waveforms, we scaled up +CCh traces (gray traces) to match peak amplitudes with those of −CCh currents. For N-type channels, the amplified +CCh traces were clearly advanced in time from control traces, as confirmed by nonzero population averages for Δt 90 shown in Figure6 B (bottom). Given that Ca2+ entry elicited here by action potential waveforms is largely akin to a tail current, such temporal advance of +CCh waveforms would be predicted from the faster deactivation kinetics of reluctant versus willing channels (Elmslie et al., 1990; Boland and Bean, 1993; Colecraft et al., 2000). Indeed, as more reluctant channels activate with prolongation of action potential waveforms (Fig. 6 A), the extent of temporal advance would be expected to increase as the fraction of tail current carried by reluctant channels accrues, an effect that we observed experimentally (Fig. 6 B). This same mechanism explains the overall temporal advance of Ca2+-entry waveforms for P/Q-type versus N-type channels (Fig. 6, compare B, C), as deactivation kinetics of P/Q-type channels are considerably faster than those for N-type channels (Colecraft et al., 2000). For P/Q-type channels, analogous amplification of +CCh traces (Fig. 6 C) indicated precise conservation of waveform shape with inhibition. This result is consistent with the absence of reluctant gating in P/Q-type channels, at least during the 2–5 msec time span of the applied action potential waveforms. These distinctions in regard to the temporal advance or invariance of the shape of Ca2+-entry waveforms could enrich the dimensions of synaptic modulation during G-protein inhibition, as will be discussed.
This study provides the first single-channel comparison of N- and P/Q-type calcium channels during G-protein inhibition. Single, inhibited N-type channels exhibited a sparse pattern of opening, which likely represents direct resolution of gating in the reluctant mode with Gβ γ subunit(s) bound to the channel. By contrast, inhibited P/Q-type channels were electrically silent, presumably reflecting an inability of these channels to open while complexed with Gβ γ. These striking contrasts in single-channel properties were shown to produce fundamental differences in the effect of G-protein inhibition on the waveform of Ca2+ entry elicited by physiological action potential waveforms. Our findings help to clarify divergent viewpoints of previous studies of reluctant gating and raise new possibilities regarding both the molecular basis of G-protein inhibition, as well as the dimensions of synaptic modulation that ensue from such channel inhibition.
Previous studies of reluctant gating
Several whole-cell studies have provided indirect evidence that N-type channels can open in the reluctant mode during G-protein inhibition (Elmslie et al., 1990; Boland and Bean, 1993; Jones and Elmslie, 1997; Colecraft et al., 2000). In contrast to the consensus at the whole-cell level, direct single-channel resolution of reluctant gating of N-type channels has been elusive. Early studies may have been challenged by improper isolation of single N-type channels (Delcour and Tsien, 1993; Elmslie, 1997). A subsequent study of inhibited N-type channels in IMR32 cells (Carabelli et al., 1996) reported a sparser form of gating after first opening, suggestive of reluctant gating. However, the presence of both voltage-dependent and voltage-independent forms of G-protein inhibition in this preparation (Carabelli et al., 1996) made it difficult to unambiguously associate such sparse gating with reluctant openings, because the voltage-independent mechanism might itself produce such effects. Our contemporaneous single-channel study (Patil et al., 1996), using recombinant N-type channels expressed in HEK 293 cells, offered the advantage that only voltage-dependent G-protein inhibition is present in this system (Patil et al., 1996; Colecraft et al., 2000). In this setting in which a sparser pattern of gating could be directly attributed to reluctant gating, inhibited N-type channels were electrically silent. However, the challenge of resolving small unitary events at strong depolarizations limited this study to voltages that might be too hyperpolarized to elicit reluctant gating, according to inferences drawn from whole-cell analyses (Jones and Elmslie, 1997; Colecraft et al., 2000) (see also Fig. 1 B). Subsequent results confirm this scenario. In frog sympathetic neurons, Lee and Elmslie (2000) observed reluctant gating of single N-type channels, with quantitative resolution of unitary events at depolarizations as high as +40 mV (approximately −10 mV at physiological Ca2+). However, the slow activation of such reluctant openings made it difficult to argue that they would contribute to physiological responses. Here, using quartz pipettes to enhance resolution of unitary currents at even higher voltages with near doubling of bandwidth, we find not only that single N-type channels support reluctant openings (Fig. 3), but that such openings have rapid activation at voltages approaching the peak of action potentials (Fig. 4). Our experiments with action potential waveforms (Fig. 6) explicitly confirm the impact of reluctant gating during physiological stimuli.
Moreover, our current study provides the first report of single P/Q-type channel gating during G-protein inhibition. Our results directly establish the absence of reluctant gating in P/Q-type channels, irrespective of voltages spanning the entire physiological range. An important caveat here is that we might not be able to detect reluctant P/Q-type channel openings should they occur much too slowly to impact currents evoked by action potential waveforms. Such a scenario, however, would be inconsistent with our previous whole-cell data (Colecraft et al., 2000).
Features of channel structure underlying differing prevalence of reluctant gating
Much research has sought to determine the structural determinants of G-protein inhibition of neuronal calcium channels, using channel mutations and chimeras to identify critical regions on the corresponding pore-forming α1 subunits. To some extent, the overall mechanism and structural basis for G-protein inhibition has been considered to be largely conserved across the neuronal channels that are subject to G-protein modulation (N-type (α1B), P/Q-type (α1A), and R-type (α1E)). In this view, structure–function studies focused on different members of this group of channels, have collectively identified at least three regions of the α1 backbone that may be essential for G-protein inhibition, namely: (1) the N terminus (Zhang et al., 1996; Page et al., 1998; Simen and Miller, 1998; Canti et al., 1999); (2) the cytoplasmic linker between domains I and II (De Waard et al., 1997;Herlitze et al., 1997; Zamponi et al., 1997; Furukawa et al., 1998a,b); and (3) the C terminus (Zhang et al., 1996; Qin et al., 1997; Furukawa et al., 1998a,b).
Whether these regions all contribute similarly to G-protein inhibition across neuronal channel types seems less certain, given our finding that N-type, but not P/Q-type channels can open in the reluctant mode. Within the context of the willing-reluctant model of channel inhibition (Elmslie et al., 1990; Boland and Bean, 1993; Patil et al., 1996), this finding implies that Gβ γ-bound N-type channels can open, whereas P/Q-type channels remain entirely closed while complexed with Gβ γ. This striking contrast in molecular properties gives reason to wonder whether channel-specific differences in one, or more of the three regions can account for the very different allosteric coupling between Gβ γ binding and channel opening.
Physiological implications of reluctant openings
The differential occurrence of reluctant openings in N-type versus P/Q-type channels raises rich possibilities for the dimensions of synaptic modulation produced by G-protein inhibition, especially because the involvement of these two channel types differs at spatially distinct synapses (Luebke et al., 1993; Takahashi and Momiyama, 1993;Wheeler et al., 1994; Takahashi et al., 1996; Poncer et al., 1997; Reid et al., 1998).
Our results suggest that at synapses where P/Q-type channels trigger release, presynaptic channel inhibition by G-proteins is attributable solely to delayed activation kinetics of inhibited channels (Fig. 5), effectively silencing them for the brief duration of neuronal action potentials. Hence, G-protein inhibition would reduce the amplitude of the waveform of Ca2+entry during action potentials but preserve waveform shape (Fig.6 C). In this case, synaptic efficacy would be reduced, but the timing and precision of neurotransmission (Sabatini and Regehr, 1999) would be preserved.
At synapses dominated by N-type channels, the effects of G-protein inhibition on synaptic communication could be considerably different. Here, the slower activation of inhibited channels would also cause many of them to remain closed throughout a given action potential, but a significant fraction of inhibited channels would contribute reluctant openings during a neuronal spike. Accordingly, whereas the silencing of inhibited channels would diminish the overall amplitude of Ca2+ entry, the faster deactivation kinetics pertaining to reluctant openings (Colecraft et al., 2000) would result in temporal advancement of the Ca2+-entry waveform during an action potential (Fig. 6 B). Such temporal acceleration of Ca2+ entry holds intriguing ramifications for synaptic modulation, because the changes in presynaptic Ca2+ entry could be tightly linked to analogous alterations in the waveform of postsynaptic responses (Sabatini and Regehr, 1997; Borst and Sakmann, 1999). A close linkage might be expected, given that the presynaptic release probability (RP) can be the rate-limiting function for specifying the EPSC waveform at cerebellar synapses (Chen and Regehr, 1999; Sabatini and Regehr, 1999), and the time course of RP may be largely shaped by the waveform of presynaptic Ca2+ entry (Sabatini and Regehr, 1996; Bollman et al., 2000) (W. Regehr, personal communication) (J. G. G. Borst, personal communication). In the case of such linkage, G-protein inhibition at synapses dominated by N-type channels would not only decrease synaptic efficacy, but also shorten delays between presynaptic and postsynaptic responses. The latter effect would impact the timing and coincidence detection properties of such synapses (Sabatini and Regehr, 1999).
Another potential effect of reluctant openings is to modulate the coupling efficiency of open N-type channels to release sites. The effect can be understood by considering that individual channels support a “calcium microdomain” surrounding the cytoplasmic mouth of the channel pore. At many synapses, the overlap of several domains with a single release site enables multiple channels to trigger the cognate release site (Mintz et al., 1995; Borst and Sakmann, 1998); at others, only one domain overlaps with a given release site (Augustine et al., 1991; Stanley, 1993; Bertram et al., 1996). Because the size of such microdomains is determined by both the unitary current flux and the duration of typical openings (Bertram et al., 1996; Neher, 1998), N-type channels supporting briefer reluctant openings should have smaller microdomains than those exhibiting willing openings. It is quite plausible, then, that an N-type channel manifesting willing openings would be coupled to a given release site, although the same channel would become uncoupled when opening reluctantly. Hence, G-protein inhibition could shift synapses toward regimes where a smaller proportion of active channels are coupled to release sites. This effect would deepen the presynaptic inhibition produced by a specified decrease in N-type Ca2+ current, over that which would be produced if N-type channels were incapable of reluctant gating. The deeper inhibition would occur because Ca2+ entering through formerly coupled channels would be “wasted” in terms of transmitter release. By contrast, the proportion of open P/Q-type channels coupled to release sites should remain invariant with G-protein inhibition.
Overall, these distinctions in G-protein inhibition of N- and P/Q-type channels contribute to the growing awareness that differences in the operational features of seemingly similar neuronal calcium channels, and their splice variants, may provide an important basis for custom regulation of synaptic function.
This work was supported by grants from the National Institutes of Health (D.T.Y.) and a National Institutes of Health Medical Scientist Training Program Award (D.L.B.). We thank SIBIA Neurosciences for the human α1B-1 clone, T. P. Snutch for the α1A-a and α2δ clones, E. Perez-Reyes for the β2a clone, Ernst Peralta for the m2 clone, and Devi Rathod and Rebecca Alvania for technical assistance.
Correspondence should be addressed to David T. Yue, Program in Molecular and Cellular Systems Physiology, Departments of Biomedical Engineering and Neuroscience, Johns Hopkins University School of Medicine, Ross Building, Room 713, 720 Rutland Avenue, Baltimore, MD 21205. E-mail:.