Modal Gating of Human Ca_V2.1 (P/Q-type) Calcium Channels : I. The Slow and the Fast Gating Modes and their Modulation by β Subunits

Voltage-gated P/Q-type calcium channels (Ca_V2.1) are located in presynaptic terminals and somatodendritic membranes throughout the brain (Volsen et al., 1995; Westenbroek et al., 1995), and have a prominent role in controlling neurotransmitter release (Dunlap et al., 1995). The somatodendritic localization of Ca_V2.1 channels points to additional postsynaptic roles in, for example, neural excitability (Bayliss et al., 1997; Pineda et al., 1998; Mori et al., 2000), synaptic integration and plasticity (Eilers et al., 1996; Magee et al., 1998), and gene expression (Sutton et al., 1999). The importance of Ca_V2.1 channels in brain function is stressed by the evidence that mutations in CACNA1A, the gene encoding Ca_V2.1α₁ subunits, cause a group of dominantly inherited human neurological disorders, including familial hemiplegic migraine, episodic ataxia type-2, and spinocerebellar ataxia type 6 (Ophoff et al., 1996; Zhuchenko et al., 1997). Mutations at the mouse orthologue cause a group of recessive neurological disorders, including the tottering, leaner, and rocker phenotypes with ataxia and absence epilepsy, and the rolling Nagoya phenotype with ataxia without seizures (Fletcher et al., 1996; Pietrobon, 2002). In the lethargic mouse, a mutation in the gene encoding the auxiliary calcium channel β₄ subunit causes a clinical phenotype very similar to that of tottering (Burgess et al., 1997). Knockout mice with a null mutation in the Ca_V2.1α₁ gene show severe cerebellar ataxia and dystonia and selective progressive cerebellar degeneration (Jun et al., 1999; Fletcher et al., 2001).

The single channel gating properties of Ca_V2.1 channels are critical determinants of the time course and magnitude of the various Ca²⁺-dependent processes they control. Both experiments and simulations have shown that even small changes in the kinetics of channel opening and/or closing and in channel open probability can strongly affect the time course and magnitude of Ca²⁺ influx during an action potential (McCobb and Beam, 1991; Borst and Sakmann, 1998; Sabatini and Regehr, 1999; Colecraft et al., 2001; Bischofberger et al., 2002; Meinrenken et al., 2002, 2003). Given the steep dependence of neurotransmitter release on Ca²⁺ influx (Dodge and Rahamimoff, 1967; Bollmann et al., 2000; Schneggenburger and Neher, 2000), the detailed gating properties of single Ca_V2.1 channels will have a strong impact especially on the time course and magnitude of neurotransmitter release at central synapses, where P/Q channels appear to be more effectively coupled to release than other Ca²⁺ channel types (Mintz et al., 1995; Wu et al., 1999; Qian and Noebels, 2001). Despite the critical role of Ca_V2.1 channel gating in determining neurotransmission efficacy at central synapses, surprisingly few data are available on the single channel gating properties of native P/Q-type Ca²⁺ channels (Usowicz et al., 1992; Forti et al., 1994; Tottene et al., 1996) or recombinant Ca_V2.1 channels (Yatani et al., 1994; Bourinet et al., 1999; Hans et al., 1999; Colecraft et al., 2001; Tottene et al., 2002). The discovery that mutations causing familial hemiplegic migraine increase the open probability and the single channel Ca²⁺ influx through human Ca_V2.1 channels (Hans et al., 1999; Tottene et al., 2002) and, as a consequence, facilitate the induction and the propagation of cortical spreading depression (van den Maagdenberg et al., 2004) fosters the interest in extending our knowledge of the single channel properties of human Ca_V2.1 channels. Therefore, one of our aims here was to obtain a detailed characterization of the gating properties of human Ca_V2.1 channels at the single channel level.

In heterologous expression systems, β subunits are powerful regulators of both channel activity and number of channels expressed in the membrane (Walker et al., 1998; Dolphin, 2003a). In particular, for the Ca_V2.1 channel, there is evidence for both a chaperone-like effect of β subunits on channel trafficking to the plasma membrane (Brice et al., 1997; Bichet et al., 2000) and modulation of the voltage range of activation and inactivation as well as the kinetics of inactivation of the whole-cell current (Stea et al., 1994; De Waard and Campbell, 1995; De Waard et al., 1995; Cens et al., 1996). Four different genes encode β subunits (β_1,2,3,4) that are differentially expressed in different neurons and during development (Tanaka et al., 1995; Witcher et al., 1995; Ludwig et al., 1997; Volsen et al., 1997; Vance et al., 1998; Burgess et al., 1999). Regional differences in brain expression pattern can also occur for splice variants of the same β subunit (Helton et al., 2002). Native P/Q-type channels can contain each of the four different β subunits (Liu et al., 1996), and the fractional contribution of a particular β subunit for channel formation varies among different brain regions (Pichler et al., 1997). Different combinations of a Ca_V2.1α₁ subunit with auxiliary β subunits most likely contribute to generate the large functional diversity of native P/Q-type calcium channels (Mintz et al., 1992; Usowicz et al., 1992; Randall and Tsien, 1995; Tottene et al., 1996; Forsythe et al., 1998; Mermelstein et al., 1999). Whole-cell current recordings in heterologous expression systems revealed quite different kinetics of inactivation and slightly different voltage ranges of activation depending on the β subtype combined with the Ca_V2.1α₁ subunit (Sather et al., 1993; Stea et al., 1994; De Waard and Campbell, 1995; De Waard et al., 1995; Cens et al., 1996; Moreno et al., 1997; Krovetz et al., 2000; Restituito et al., 2000; Sokolov et al., 2000; Sandoz et al., 2001; Helton and Horne, 2002; Helton et al., 2002; Tsunemi et al., 2002). However, the effect of different β subunits on the single channel gating properties of Ca_V2.1 channels (and also of the other Ca_V channels, with the exception of Ca_V1.2) remains unknown. Interestingly, distinct β subunits confer unique single channel gating properties to L-type channels extending well beyond differences in inactivation (Colecraft et al., 2002). The regulation of Ca_V2.1 channel properties by variations in β subunit composition appears as a potential mechanism for tuning channel behavior to support a given physiological role, and might contribute to create the great diversity of release efficacy and short-term synaptic plasticity at different synapses (Atwood and Karunanithi, 2002). Therefore a second aim here is to study how the presence of different auxiliary β subunits in the human Ca_V2.1 channel alters gating at the single channel level.

Single channel patch-clamp recordings on human embryonic kidney HEK293 cells expressing human Ca_V2.1 channels revealed a complex modal gating of these channels, which is described in this and the accompanying paper (Fellin et al., 2004). Here, we report the characterization of two modes of gating of human Ca_V2.1 channels, the slow mode and the fast mode, that differ in mean closed times and latency to first opening, in voltage dependence of the open probability, in kinetics of inactivation, and voltage dependence of steady-state inactivation. The study of Ca_V2.1 channels containing four different β subunits shows that both the relative occurrence of the slow and fast gating modes and the inactivation properties within each mode are modulated by the type of β subunit. We also show that the prevailing mode of gating of Ca_V2.1 channels lacking a β subunit is a low-p_o mode different from both the fast and the slow gating modes.

HEK293 cells (American Type Culture Collection, CRL-1533) were stably transfected with cDNA constructs encoding the human Ca_V2.1α₁ (α_1A-2), α_2bδ-1, and β_1b (A68-90 cell line) or β_2e (E1H2 cell line) or β_3a (PB1-14 cell line) or β_4a (10–13 cell line) subunits using a standard calcium phosphate–mediated procedure (Hans et al., 1999). Antibiotic-resistant colonies were selected in medium consisting of Dulbecco's modified Eagle's medium (DMEM; Life Technologies) supplemented with 6% bovine calf serum, penicillin, streptomycin, and G418 or a combination of G418 and Zeocin depending on the constructs used. Cells were maintained on standard tissue culture plates at 37°C and 5% CO₂. Parental lines and subsequent subclones were selected based on the following criteria: (a) functional responses in a fluo3 dye–based assay where increases in intracellular Ca²⁺ are measured in response to KCl depolarization; (b) Northern and Western analysis, and (c) electrophysiological characterization. All cell lines were incubated at 28°C for 12–24 h before electrophysiological measurements (and in the case of the PB1-14 cell line also before the fluo3 assay) (Hans et al., 1999).

In the case of transient transfection of HEK293 cells with cDNA constructs encoding the human Ca_V2.1α₁ (α_1A-2) and α_2bδ-1 subunits (or of PB1-14 cells with cDNA constructs encoding the β_3a subunit), CD4 expression plasmids were included to permit the identification of transfected cells as in Hans et al. (1999).

In untransfected HEK293 cells, neither Ca_V2.1α₁ nor β transcripts were detected by Northern blots and neither Ca_V2.1α₁ nor β protein was detected on Western blots (not depicted). In contrast, both Northern and Western blot assays revealed the presence of an endogenous α_2bδ-1 subunit in the same cells (not depicted).

Single channel patch-clamp recordings were performed following standard techniques, as in Hans et al. (1999). Currents were recorded at room temperature with a DAGAN 3900 patch-clamp amplifier, low-pass filtered at 1 kHz (−3 dB, 8-pole Bessel filter), sampled at 5 kHz, and stored for later analysis on a PDP-11/73 computer. All single channel recordings were obtained in the cell-attached configuration. The pipette solution contained (in mM) 90 BaCl₂, 10 TEA-Cl, 15 CsCl, 10 HEPES (pH 7.4 with TEA-OH). The bath solution contained (in mM) 140 K-gluconate, 5 EGTA, 35 l-glucose, 10 HEPES (pH 7.4 with KOH). The high-potassium bath solution was used to zero the membrane potential outside the patch. Liquid junction potential at the pipette tip was +12 mV, and this value should be subtracted from all voltages to obtain correct values of membrane potentials.

Linear leak and capacitative currents were digitally subtracted from all records used for analysis. Open-channel current amplitudes were measured by manually fitting cursors to well-resolved channel openings. Values at each voltage are averages of many measurements. Open probability, p_o, was computed by measuring the average current, 〈I〉, in a given single channel current record and dividing it by the unitary single channel current, i. To obtain activation curves, p_o values were calculated in patches containing only one channel by averaging the open probabilities measured in each sweep with activity at a given voltage (excluding the last shut time). Single channel activation curves were best fitted with the Boltzmann equation p_o = p_omax × {1 + exp[−(V − V_1/2)/k]}⁻¹, where k = RT/zF. For open and closed time histograms, a channel opening or closure was detected when more than one sampling point crossed a discriminator line at 50% of the elementary current. Histograms of open and closed times were fitted with sums of decaying exponentials. Log binning and fitting of the binned distributions were done as described by McManus et al. (1987) and Sigworth and Sine (1987). The first closed time was used to generate the cumulative first latency histogram. The best fit was determined by the maximum likelihood maximization (Colquhoun and Sigworth, 1983), and the best minimum number of exponential components was determined by the maximum likelihood ratio test (Rao, 1973). All values are given as mean ± SEM.

Cell-attached single channel recordings on HEK293 cells stably coexpressing human Ca_V2.1α₁ (α_1A-2), β_1b, and α_2bδ-1 subunits revealed that, in many patches containing only one channel, the same channel showed two different modes of gating. Fig. 1 A shows consecutive traces from a single channel patch displaying activity of the same Ca_V2.1 channel in two different periods at the same voltage (+30 mV). Analysis of the open and closed time histograms in the two periods shows that the two modes of gating differ mainly in the intervals of time spent by the channel in closed states (Fig. 1 B and Fig. 2 A). We have called these two modes of gating “slow” and “fast” for reasons that will become clear below, when the kinetics of inactivation and the latencies to first opening of a Ca_V2.1 channel in the two modes will be considered.

The transitions between the two modes were infrequent. The channel shown in Fig. 1 was in the slow gating mode during the first 8 min of the recording, and then switched to the fast mode and remained in this gating mode until the end of the recording (for almost 40 min). In different single channel patches, the intervals of time spent by the channel in each gating mode varied from tens of seconds to tens of minutes. Average values of the different gating parameters for the two gating modes in isolation were obtained from single channel patches in which, by visual inspection, the channel showed only (or mainly) one gating mode or two clearly separated periods in the two modes (Fig. 2 A).

In both gating modes, open and closed time histograms were best fit by the sum of two and three exponential components, respectively. Both the time constants and the relative areas of the two exponential components best fitting the open time histograms were similar for the two gating modes, whereas two of the closed time constants (τ_c2 and τ_c3) were significantly larger in the slow mode than in the fast mode (Fig. 1 B and Fig. 2 A). Moreover, their relative areas were different in the two gating modes, showing a larger contribution of the longest closed time intervals in the slow with respect to the fast gating mode. As a result, at +30 mV, the open probability of the channel in the fast gating mode was about twice that in the slow mode (Fig. 1 C and Fig. 2 B). Measurements of the open probability, p_o, as a function of voltage revealed that the Ca_V2.1 channel activates at more negative voltages in the fast gating mode than in the slow mode (Fig. 1 C and Fig. 2 B). Fit with a Boltzmann function of the average activation curve of Ca_V2.1 channels containing the β_1b subunit gave V_1/2 values of 25 and 32 mV (with k values of 5.3 and 4.2 mV) for the fast and slow gating mode, respectively. The single channel currents and conductance were identical in the two gating modes (not depicted, c.f. Table I).

The fast and slow modes of gating were observed in cells coexpressing human α_1A-2, α_2bδ-1, and any of the four different β subunits. Fig. 2 B shows that the activation curves of single channels containing either β_1b, or β_2e, or β_4a subunits in the fast gating mode were similarly shifted to more negative voltages with respect to the corresponding channels in the slow mode. Fits with Boltzmann distributions gave V_1/2 values of 28 and 27 mV for channels containing either β_2e or β_4a subunits in the fast gating mode (with k values of 6.6 and 5 mV), and V_1/2 values of 36 and 34 mV for the corresponding channels in the slow gating mode (with k values of 5.9 and 4 mV). The average activation curve of channels containing the β_3a subunit in the fast gating mode was similar to that of channels containing the other β subunits (V_1/2 = 23 mV, k = 3.7, n = 6, not depicted; the only significant difference was with respect to the V_1/2 value of channels containing the β_2e subunit). An average activation curve in the slow gating mode is not available, given the relatively low occurrence of that gating mode in channels containing the β_3a subunit (see below).

Single channels containing the different β subunits in either the fast or the slow gating mode had similar open and closed time histograms at +30 mV, as shown by the average values of the open and closed time constants in Table I. Only the channels with the β_2e subunit displayed some significant difference with respect to those with the β_1b subunit, and, precisely, both τ_c2 and τ_c3 in the slow gating mode were larger (P < 0.05). As a consequence, the open probability at +30 mV in the slow gating mode was smaller with the β_2e than with the β_1b subunit (P < 0.05).

Overall, from the data in Table I and Fig. 2, we can conclude that channels containing different β subunits in either the fast or the slow gating mode show only minor differences in the rate constants of the transitions between closed and open states within each gating mode.

In addition to different properties of activation, the human Ca_V2.1 channel showed different properties of inactivation in the two modes of gating (Figs. 3 and 4). During depolarizations to +30 mV, lasting 720 ms, single channels containing the β_4a subunit showed both inactivating and noninactivating activity in the fast gating mode, but almost exclusively noninactivating activity in the slow mode (Fig. 3 A). Thus, as shown by the average single channel ensemble currents, inactivation was more rapid in the fast than in the slow gating mode. The kinetics of inactivation were more rapid in the fast than in the slow gating mode independently of the type of β subunit combined with the Ca_V2.1α₁ subunit (Fig. 3 B). However, the kinetics of inactivation in each gating mode were different depending on the type of β subunit. Judging from the fraction of peak current remaining at the end of the test pulse, the most rapidly and slowly inactivating channels in the fast gating mode were those containing the β_4a and the β_2e subunit, respectively (with intermediate similar values with β_3a and β_1b; c.f. also average fraction of inactivating traces in Table II). Judging from the time constant of inactivation, τ_i, obtained from fitting the ensemble averages of the inactivating sweeps (Table II), Ca_V2.1 channels containing the β_4a subunit inactivated more slowly than those containing the β_3a or the β_1b subunit. The discrepancy may arise in part from the existence and different contribution of an additional noninactivating gating mode of Ca_V2.1 channels, the b mode, that will be described in the companion paper (Fellin et al., 2004). The different kinetics of inactivation of channels with different β subunits in the slow gating mode could be best appreciated at voltages >30 mV (not depicted).

Also the voltage dependence of steady-state inactivation was different in the two gating modes. Fig. 4 A shows the open probability in successive depolarizations at the same voltage for two representative patches containing a single Ca_V2.1 (α_1A-2–β_4a–α_2bδ-1) channel during recordings in which the holding potential was changed as indicated above the horizontal bars. When the holding potential was changed from −80 to −40 mV, the channel in the slow gating mode did not inactivate, in contrast with the clear inactivation shown by the channel in the fast mode. Steady-state inactivation occurred at more negative voltages in the fast than in the slow gating mode independently of the type of β subunit, as shown in Fig. 4 B for Ca_V2.1 channels containing either β_1b or β_4a subunits. However, the voltage dependence of steady-state inactivation in each gating mode depended on the type of β subunit, as shown for the fast gating mode in Fig. 4 C. The average open probabilities at different holding potentials in Fig. 4 (B and C) were obtained by averaging the open probability of all consecutive sweeps at a given holding potential in several experiments like those in Fig. 4 A. In the slow gating mode, like in the fast gating mode, the channels containing the β_1b and β_4a subunits inactivated at much more negative voltages than those containing the β_2e subunit. With the β_2e subunit, the open probability of a channel in the slow gating mode could be obtained at several holding potentials only in a single channel patch, and in this case, the channel did not show any steady-state inactivation up to a holding potential of −10 mV (at which voltage, channels containing β_1b and β_4a subunits in the slow gating mode were largely inactivated). In single channel patches, the fraction of traces without activity at a holding potential of −80 mV (nulls) was significantly larger for channels gating in the fast than in the slow mode, and, independently of the gating mode, was significantly larger for channels containing a β_1b or β_4a subunit than for those containing a β_2e subunit (Table II).

The relative occurrence of the slow and fast modes of gating of Ca_V2.1 channels appeared to depend on the type of β subunit. The fraction of time spent by single recombinant Ca_V2.1 channels in each gating mode was evaluated in patches containing only one channel, whose activity could be recorded for at least 12 min. In each patch, only the first 12 to 16 min of single channel activity were considered, to avoid possible distortions due to different durations of the recording in different patches. Moreover, due to the difficulty of separating with a rigorous criterion individual sweeps with either fast or slow mode activity, only single channel patches in which the channel showed (by visual inspection) only (or mainly) one gating mode or clearly separated long periods in the two gating modes were considered. The classification of the activity in a certain patch or period as either slow or fast gating mode was then based on the open/closed time histograms and/or on the voltage dependence of the open probability. In 168 min of recording (from 11 single channel patches), channels containing the β_1b subunit spent 46% of the time in the slow gating mode. A larger fraction of time (58%, in 183 min of recordings from 12 single channel patches) was spent in the slow gating mode by channels containing the β_2e subunit. Channels containing the β_4a subunit spent most of the time in the slow gating mode (68%, in 117 min of recordings from nine single channel patches). The opposite was true for channels containing the β_3a subunit (34%, in 87 min of recordings from six single channel patches). According to Fisher exact test, the differences in occurrence of the fast and slow gating modes among channels with the different β subunits are all statistically significant (P < 0.05 or 0.001). Transitions from the slow gating mode (present at the beginning of the recording) to the fast mode were observed in 2, 3, 0, and 2 single channel patches with the β_1b, β_2e, β_3a, and β_4a subunit, respectively. The opposite transition from the fast gating mode (present at the beginning) to the slow mode was observed only in one single channel patch with the β_4a subunit. In the remaining single channel patches, the channel remained mainly or only in either the fast (5, 3, 4, and 1 patches) or slow gating mode (4, 6, 2, and 5 patches).

In contrast with the gating properties, the permeation properties of Ca_V2.1 channels were not affected by the type of β subunit. Ca_V2.1 channels containing the four different β subunits had identical single channel currents and conductances (Fig. 5).

Given the infrequent transitions between the slow and fast modes of gating, and the well established fact that Ca_V2.1 channels containing a β subunit activate and inactivate at more negative voltages than channels lacking a β subunit (Stea et al., 1994; De Waard and Campbell, 1995), one might wonder whether the slow gating mode (showing activation and inactivation at more positive voltages than the fast mode) corresponds to channels that have lost the β subunit. This hypothesis is made unlikely by the observation that most of the observed transitions between gating modes were from the slow to the fast mode, and is ruled out by the different properties of inactivation of channels containing different β subunits in the slow gating mode. An additional conclusive evidence against the same hypothesis is the fact that, in HEK293 cells transfected with only Ca_V2.1α₁ and α_2bδ-1 subunits, single Ca_V2.1 channels did not gate in the slow gating mode. Interestingly, channels containing only Ca_V2.1α₁ and α_2bδ-1 subunits showed a low-p_o mode of gating, different from both the fast and the slow gating modes (Figs. 6 and 7).

Fig. 6 shows consecutive traces of activity at +30 mV from a patch containing a single α_1A2–α_2bδ-1 channel and the open–closed time histograms and activation curve of the same channel. Fig. 7 compares the gating parameters of the low-p_o mode typical of α_1A2–α_2bδ-1 channels with those of the slow and fast gating modes of Ca_V2.1 channels containing a β subunit. In comparison with the slow and fast gating modes, the low-p_o mode of Ca_V2.1 channels lacking the β subunit was characterized by shorter open times and longer closed times (Fig. 6, A and B, and Fig. 7 A), by longer first latencies to channel opening (Fig. 7 C), by activation at more positive voltages, lower open probability, and shallower dependence of the open probability on voltage (Fig. 6 C and Fig. 7 B), and by a much larger number of nulls (64 ± 5% out of 578 traces at +30 mV in five single channel patches). The time constants of the two exponential components best fitting the open time histograms were both smaller in the low-p_o gating mode and the contribution of the shortest open times was larger; on the other hand, the three closed time constants were all significantly larger and the contribution of the shortest closed times was smaller in the low-p_o gating mode than in both the fast and slow gating modes (Fig. 7 A). As a result, the open probability of the channel in the low-p_o gating mode (at +30 mV) was 0.41 and 0.23 times that in the slow and fast modes, respectively (Fig. 7 B). Fits with a Boltzmann function of the average activation curves of single channels in the three gating modes gave V_1/2 (and k) values of 44 (k = 8.9, n = 6) mV, 33 (k = 4.2, n = 13) mV, and 26 (k = 5.2, n = 17) mV for the low-p_o, slow, and fast gating modes, respectively (for the latter gating modes, the values from channels containing the different β subunits were averaged together). The cumulative first latency histograms of single Ca_V2.1 channels in both fast and slow gating modes were best fit by the sum of two exponential components (τ₁ = 2.0 ms, 94%, τ₂ = 8.7 ms for the fast mode, and τ₁ = 1.8 ms, 55%, τ₂ = 12 ms for the slow mode), whereas three exponentials were necessary for the low-p_o mode of α_1A2–α_2bδ-1 channels (τ₁ = 2.3 ms, 31%, τ₂ = 18 ms, 59%, and τ₃ = 78 ms, 10%). The time constant of the fastest component was similar in the three gating modes (∼2 ms) but its contribution was very different (from almost 100% for the fast mode, to only 55% and 31% for the slow and low-p_o modes, respectively). Thus, Ca_V2.1 channels in the fast gating mode open more rapidly than those in the slow mode (and the latter more rapidly than those without β subunit in the low-p_o mode).

In addition to the prevailing low-p_o mode of gating, Ca_V2.1 channels containing only α_1A2 and α_2bδ-1 subunits showed a high-p_o mode of gating similar to the fast gating mode of the α_1A2–α_2bδ1-β channels. In three single channel patches, the average open probability at +30 mV of the channel in this gating mode was 0.267 ± 0.04; the open time constants were 0.58 ± 0.22 and 1.4 ± 0.2 ms with relative contribution of 56 ± 4 and 44% and the closed time constants were 0.45 ± 0.05, 2.4 ± 0.1, and 8.8 ± 1.6 ms with relative contributions of 54 ± 4, 30 ± 2, and 16 ± 2%. The half voltage of activation was 27 ± 2 mV (with k = 4.9 ± 0.2). Much less frequently, α_1A2–α_2bδ-1 channels also showed a mode of gating similar to the slow mode. One may wonder whether the fast and slow gating modes recorded in cells transfected with only α_1A2 and α_2bδ-1 subunits may correspond to channels containing endogenous β subunits (that might be expressed at such low level to be undetectable by Western and Northern blots; Meir et al., 2000). This interpretation appears unlikely in light of the following observation. A low-p_o mode of gating very similar to that just described was the prevailing gating mode present (together with the fast and the rarer slow gating mode) in our initial single channel recordings from the cell line PB1-14, which was supposed to stably express α_1A2, α_2bδ, and β_3a subunits. The low-p_o mode of gating was however totally absent after transfection of PB1-14 cells with β_3a cDNA, suggesting that for some reasons these cells were not expressing sufficient amounts of β_3a subunits, and therefore most of the recorded single channels lacked the β_3a subunit. Indeed, the single channel open probability (0.07 ± 0.01, n = 3), the open and closed time constants (τ_o1 = 0.26 ± 0.04 ms, τ_o2 = 0.97 ± 0.26 ms; τ_c1 = 0.72 ± 0.10 ms, τ_c2 = 8.8 ± 0.3 ms, τ_c3 = 34 ± 5 ms, n = 3), and the fraction of nulls (64 ± 6%, n = 4) obtained for the low-p_o gating pattern in PB1-14 cells were very similar to those obtained for the low-p_o activity in HEK293 cells transfected with only α_1A2 and α_2bδ-1 subunits. Interestingly, in two single channel patches from PB1-14 cells, transitions of the same channel from the slow to the low-p_o gating mode, and in one of them, from the low-p_o to the fast gating mode and back to the low-p_o mode were observed. This observation is consistent with the hypothesis that the three gating patterns reflect different modes of gating of a Ca_V2.1 channel lacking a β subunit rather than different combinations of Ca_V2.1α₁ subunits and auxiliary subunits. The low-p_o gating mode shown in Figs. 6 and 7 is the prevailing mode of gating of Ca_V2.1 channels without a β subunit, but is completely absent in channels containing a β subunit. In fact, in hundreds of cell-attached patches on HEK293 cells transiently transfected with human Ca_V2.1α₁, α_2bδ-1, and β_2e subunits, the low-p_o gating mode was never observed (Hans et al., 1999; Tottene et al., 2002). Likewise, it was never observed in cells transiently transfected with only α_1A2 and β_2e subunits (n = 8).

The main findings reported in this paper can be summarized as follows. (a) Single recombinant human Ca_V2.1 channels show two different modes of gating (slow and fast), characterized by different mean closed times and latency to first opening (both longer when the channel is in the slow mode), different voltage dependence of the open probability (larger depolarizations are necessary to activate the channel in the slow mode), different kinetics of inactivation (slower in the slow mode), and different voltage dependence of steady-state inactivation (occurring at less negative voltages when the channel is in the slow mode). (b) Ca_V2.1 channels containing any of the β subtypes (β_1b, β_2e, β_3a, or β_4a) can gate in either the slow or the fast mode, with only minor differences in the rate constants of the transitions between closed and open states within each mode. With all β subunits, inactivation is more rapid and steady-state inactivation occurs at more negative voltages in the fast than in the slow gating mode. However, in both gating modes, Ca_V2.1 channels display different rates of inactivation and different steady-state inactivation depending on the β subtype, with β_2e giving a considerably slower rate of inactivation and less negative voltage range of steady-state inactivation than the other β subunits. (c) The relative occurrence of the slow and fast gating modes of Ca_V2.1 channels is modulated by the type of auxiliary β subunit; β_3a promotes the fast mode, whereas β_4a promotes the slow mode of recombinant Ca_V2.1 channels expressed in HEK293 cells. (d) The prevailing mode of gating of Ca_V2.1 channels lacking a β subunit is a low-p_o mode different from both the fast and the slow gating modes. A channel in the low-p_o mode shows shorter mean open times, longer mean closed times, longer first latency, a much larger fraction of nulls, and activates at more positive voltages, displaying a shallower voltage dependence, than in either the fast or the slow gating mode.

Modal gating appears to be a highly conserved feature among different ion channel types. Discrete modes of single channel activity, presumably corresponding to different sets of protein conformations, have been described for voltage-dependent Ca²⁺ (Hess and Tsien, 1984; Yue et al., 1990; Plummer and Hess, 1991; Delcour et al., 1993; Forti and Pietrobon, 1993; Rittenhouse and Hess, 1994; Mantegazza et al., 1995), Na⁺ (Patlak and Ortiz, 1986; Nilius, 1988; Moorman et al., 1990; Zhou et al., 1991; Alzheimer et al., 1993; Bohle et al., 1998), K⁺ (Marrion, 1996; Singer-Lahat et al., 1999), and Cl⁻ (Blatz and Magleby, 1986) channels, and also for Ca²⁺-activated and G protein–activated K⁺ channels (McManus and Magleby, 1988; Smith and Ashford, 1998; Yakubovich et al., 2000) and for ligand-activated channels (for example see Naranjo and Brehm, 1993; Popescu and Auerbach, 2003). In most cases, it remains unclear whether the transitions between modes of gating (occurring in time frames ranging from hundreds of milliseconds to several minutes) reflect slow conformational changes intrinsic to the channels or conformational changes driven by chemical reactions or interaction with other proteins. However, there are several clear instances in which phosphorylation–dephosphorylation reactions either modulate the rate constants of interconversion between different gating modes of the channel or drive the channel into different gating modes corresponding to different states of phosphorylation (c.f. for L-type Ca²⁺ channels, Ochi and Kawashima, 1990; Yue et al., 1990; Herzig et al., 1993; Ono and Fozzard, 1993; Dzhura et al., 2000; for different K⁺ channels, Marrion, 1996; Smith and Ashford, 1998; Singer-Lahat et al., 1999; for a cation channel in Aplysia neurons, Wilson and Kaczmarek, 1993). In other cases, it has been shown that the interconversion between different gating modes is either modulated or driven by the interaction of the channel with other proteins, e.g. auxiliary subunits (Naranjo and Brehm, 1993; Chang et al., 1996; Singer-Lahat et al., 1999; Wakamori et al., 1999; Meir et al., 2000) and/or G protein subunits (Delcour et al., 1993; Lee and Elmslie, 2000; Colecraft et al., 2001), or calmodulin (Imredy and Yue, 1994; Peterson et al., 1999; Zuhlke et al., 1999). In addition, voltage dependence of the rate constants of interconversion between different gating modes has been clearly shown for L-type Ca²⁺ channels (Pietrobon and Hess, 1990; Forti and Pietrobon, 1993; Cloues et al., 1997; Hivert et al., 1999).

This (and the following, Fellin et al., 2004) is the first report of modal gating of Ca_V2.1 channels. The fast and slow modes of gating, described in this study, are quite different from the gating modes previously reported for L-type and N-type Ca²⁺ channels (Hess and Tsien, 1984; Yue et al., 1990; Plummer and Hess, 1991; Delcour et al., 1993; Rittenhouse and Hess, 1994; Mantegazza et al., 1995). The distinguishing features are (a) the long mean lifetimes (time frame of several minutes), pointing to quite slow rate constants for the transitions between the two gating modes, and (b) the different kinetics and voltage dependence of inactivation of the channel in the two gating modes, together with different closed times (but similar open times) and different voltage dependence of activation. The gating modes of L-type Ca²⁺ channels are characterized by large differences in open and closed times of the channel in the different modes, leading to very different open probabilities, without reported clear differences in inactivation properties (Hess and Tsien, 1984; Yue et al., 1990; Mantegazza et al., 1995). Gating modes with different kinetics and voltage dependence of inactivation have been reported for N-type Ca²⁺ channels (Plummer and Hess, 1991), but the much shorter mean lifetime (few seconds) of the inactivating and noninactivating gating modes of N-type channels and the similar open and closed times of the channel in the two gating modes clearly distinguish the modal behavior of N-type (Ca_V2.2) channels from that of Ca_V2.1 channels described here (but see Fellin et al., 2004). On the other hand, inactivating and noninactivating gating modes with mean lifetimes in the time frame of several minutes, which share several properties with the fast and slow gating modes of Ca_V2.1 channels, have been described for Na⁺ and K_V1.1 channels (Zhou et al., 1991; Singer-Lahat et al., 1999; Tabarean et al., 1999). For both Na⁺ and K⁺ channels, there are indirect evidences, from macroscopic current recordings, that interaction of the channel with the cytoskeleton and with Gβγ subunits either modulate or drive switching between the inactivating and noninactivating gating modes (Levin et al., 1996; Ma et al., 1997; Jing et al., 1999; Tabarean et al., 1999). It will be interesting to investigate with future work whether similar protein interactions modulate or drive switching between the fast and slow gating modes of Ca_V2.1 channels. For the time being, we have established that interaction with syntaxin1A is not involved in this mode switching, since incubation with botulinum toxin C1, to cleave syntaxin, did not alter modal gating of human Ca_V2.1 channels expressed in HEK293 cells (unpublished data; c.f. Sutton et al., 1999).

Our data suggest that the equilibrium between the slow and fast gating modes of Ca_V2.1 channels is modulated by the type of auxiliary β subunit. Thus, β subunits exert a dual regulation of inactivation of Ca_V2.1 channels; on one hand, the β subtype affects the inactivation properties of the channel in each gating mode, and, on the other hand, it affects the fraction of time spent by the channel in the slowly inactivating and more rapidly inactivating gating modes. The order of efficiency with which the different β subunits shift the equilibrium between gating modes toward the inactivating fast mode (β_3a > β_1b> β_2e > β_4a) is quite different from that with which they increase the rate of inactivation of the channel (β_4a > β_1b ≅ β_3a >> β_2e, judging from the fraction of current inactivating during 720 ms, or β_3a ≥ β_1b > β_4a >> β_2e, judging from the time constant of inactivation of the ensemble current of the inactivating traces) or shift steady-state inactivation toward negative voltages (β_4a ≅ β_1b > β_3a >> β_2e). These observations suggest that the β subunit modulation of inactivation properties and of modal gating occurs by different mechanisms and molecular interactions. It might be interesting to note that the two β subunits that seem to favor the slow gating mode are those that specifically interact with the COOH and NH₂ terminals of the Ca_V2.1α₁ subunit (Walker et al., 1998, 1999), with β_4a showing both a higher affinity for these binding sites and a larger occurrence of the slow gating mode.

The dual independent regulation of channel inactivation by β subunits might contribute to explain the disagreement in the literature regarding the order of efficiencies with which different β subunits modulate the inactivation properties of Ca_V2.1 channels (Stea et al., 1994; De Waard and Campbell, 1995; Restituito et al., 2000). In fact, different splice variants of Ca_V2.1α₁ might well display a different equilibrium between the slow and fast gating modes or a different regulation of modal gating by β subunits. Moreover, as found for many channels, mode switching might be modulated by other factors (e.g., interactions with other proteins or metabolic processes) that may be cell type specific.

A large variability of the inactivation properties (both kinetics and voltage range of inactivation) of the whole cell Ca²⁺ current in both HEK293 cells and oocytes expressing rat Ca_V2.1 channels has been reported (Moreno et al., 1997; Restituito et al., 2001; Rousset et al., 2001). A similar variability has been noted also in tsA201cells expressing Ca_V2.2 (N-type) channels (Hurley et al., 2000). Restituito et al. (2001) have excluded both different levels of palmitoylation and different levels of β expression as the cause of the variable inactivation properties, and have proposed that the variability may reflect a variable fraction of channels that have lost the β subunit after being targeted to the membrane. Our single channel recordings do not support the presence of a significant fraction of Ca_V2.1 channels lacking a β subunit in transfected HEK293 cells. In fact, in hundreds of cell-attached patches on cells transiently transfected with human Ca_V2.1α₁, α₂δ, and β₂ subunits, the low-p_o gating mode, typical of Ca_V2.1 channels lacking a β subunit, was never observed. On the other hand, a variability in the equilibrium between the gating modes described in this and the companion paper (Fellin et al., 2004) may well explain the variable inactivation properties of recombinant Ca_V2.1 channels in different cells (c.f. variable fraction of time spent by a single Ca_V2.1 channel in the different gating modes).

In many central synapses, Ca_V2.1 channels are preferentially located at the release sites and are more effectively coupled to neurotransmitter release than other Ca²⁺ channel types (Mintz et al., 1995; Wu et al., 1999; Qian and Noebels, 2001). At these synapses, the action potential–evoked Ca²⁺ influx and the local Ca²⁺ increase that triggers neurotransmitter release are mainly determined by the kinetics of opening and closing, the open probability, and the unitary conductance of Ca_V2.1 channels (Borst and Sakmann, 1998; Sabatini and Regehr, 1999; Meinrenken et al., 2002, 2003). Given the steep dependence of neurotransmitter release on Ca²⁺ influx (Dodge and Rahamimoff, 1967; Bollmann et al., 2000; Schneggenburger and Neher, 2000), even small changes in these properties have a large effect on the number of vesicles released by an action potential (Sabatini and Regehr, 1999; Meinrenken et al., 2002). The impact of a relatively small change in the kinetics of activation of presynaptic Ca²⁺ channels on the Ca²⁺ current and neurotransmitter release evoked by an action potential at a central synapse has been shown by Borst and Sakmann (1998). The presynaptic Ca²⁺ channels (mainly P/Q) of the Calyx of Held activated more rapidly after a brief prepulse; the faster activation resulted in an increased action potential-evoked presynaptic Ca²⁺ current, which could be simulated with a small increase in the rate constant for gate opening (using a Hodgkin and Huxley–type kinetic scheme to model Ca²⁺ channel gating). The importance of the kinetics of activation of presynaptic Ca²⁺ channels in determining the fraction of channels that open during a short action potential and the ensuing Ca²⁺ current waveform and neurotransmitter release has been shown also by Sabatini and Regehr (1999) with similar simulations.

Human Ca_V2.1 channels gating in the slow or the fast gating mode have quite different latencies to first opening, activation curves, and open probabilities. The magnitude and timing of Ca²⁺ influx in response to an action potential are then expected to be quite different for Ca_V2.1 channels in the fast or the slow gating mode. One can approximately infer the impact that the different latencies to first opening of channels in the fast and slow mode may have on the Ca²⁺ current evoked by an action potential by considering the impact of G protein modulation of Ca_V2.1 channels on the action potential–evoked Ca²⁺ current as measured by Colecraft et al. (2001). The relative difference in latencies to first opening between G protein–inhibited single rat Ca_V2.1 channels and uninhibited channels in Colecraft et al. (2001) appears in fact comparable to that between human Ca_V2.1 channels in the slow and fast mode (at nearly equivalent voltages, taking into account that both the po–V and i–V curves measured by Colecraft et al. appear to be shifted by ∼15 mV toward positive voltages with respect to those measured here). The longer latency to first opening of G protein–inhibited channels produced a marked reduction of the whole cell Ca²⁺ current elicited by an action potential (Colecraft et al., 2001). One can predict that the slow mode of Ca_V2.1 channels would produce a comparable reduction of the Ca²⁺ current elicited by an action potential. Note that the lower open probability of a channel in the slow mode, compared with the fast mode, would also contribute to reduce Ca²⁺ influx in response to an action potential.

Synaptic transmission and other physiological responses that are dependent on action potential–evoked Ca²⁺ influx through Ca_V2.1 channels would then be potently modulated by any factor that regulates the equilibrium between the slow and fast modes of gating of Ca_V2.1 channels. The auxiliary β subunit appears to be one of these factors since the relative occurrence of the fast and slow gating modes was different in single human Ca_V2.1 channels containing different β subunits. The β subunit composition of Ca_V2.1 channels could thus contribute to fine tune presynaptic Ca²⁺ influx to local physiological requirements and contribute to create the great diversity of release efficacy at different synapses (Atwood and Karunanithi, 2002). Given the different inactivation properties of the fast and the slow mode and the dual regulation of inactivation of human Ca_V2.1 channels by β subtypes (affecting both the equilibrium between modes and inactivation within modes), Ca_V2.1 channels containing different β subunits may also lead to different amounts and timing of Ca²⁺ influx in response to presynaptic trains of action potentials (Forsythe et al., 1998) or to postsynaptic complex voltage waveforms (containing both excitatory postsynaptic potentials and action potentials) as those generated in response to high-frequency presynaptic stimulation (Liu et al., 2003). Liu et al. (2003) have shown that different kinetics of inactivation and especially voltage- dependence of steady-state inactivation of Ca²⁺ channels have a striking impact on the amount and temporal pattern of Ca²⁺ influx in response to the repetitive firing waveforms that simulate the postsynaptic response of a pyramidal neuron to high-frequency and theta rhythm presynaptic stimulation. The β subunit composition of Ca_V2.1 channels may therefore also contribute to fine tune postsynaptic Ca²⁺ influx and Ca²⁺-dependent processes to specific stimuli and local physiological requirements. It may also contribute to create the great diversity of short-term synaptic plasticity at different synapses (Forsythe et al., 1998; Atwood and Karunanithi, 2002).

As discussed above, modal gating is widespread among channels and regulation of the equilibrium between gating modes appears as a widespread mechanism for neuromodulation of channel function. Ca_V2.1 channels are known to be regulated by many different transmitters and various signaling pathways (Catterall, 2000; Dolphin, 2003b; Elmslie, 2003). Except for G proteins (Colecraft et al., 2001), modulation of Ca_V2.1 channels has not been studied at the single channel level. We can exclude that the slow gating mode corresponds to the reluctant gating mode of Ca_V2.1 channels bound to Gβγ (Colecraft et al., 2001). However, a shift in the equilibrium between slow and fast gating modes may be possibly involved in the regulation of Ca_V2.1 channels by, e.g., phosphatidyl-inositol-4,5-bisphosphate (Wu et al., 2002), or oxidation (Li et al., 1998; Chen et al., 2002) or hippocampal A_2b adenosine receptors (Mogul et al., 1993). In each case, at the whole-cell level, the modulation produced changes in the voltage dependence of Ca²⁺ current activation. While it remains to be seen whether these changes were due to shifts between the gating modes of Ca_V2.1 channels, certainly the regulation of the complex modal gating of these channels (described here and in Fellin et al., 2004) provides a potent and versatile mechanism to fine tune Ca²⁺ influx and Ca²⁺-dependent processes to specific stimuli in a changing physiological environment. Given the critical role of Ca_V2.1 channels in controlling neurotransmitter release at central synapses, regulation of their modal gating could have profound consequences on synaptic transmission and plasticity.

The financial support of Telethon-Italy and the Italian Ministry of Education University Research (PRIN, FIRB, ST-L.449/97-CNR-MIUR, FISR-L.16/10/2000-CNR-MIUR) to Daniela Pietrobon is gratefully acknowledged.

Olaf S. Andersen served as editor.