Modulation of L-type Ca 2 1 Channels by G bg and Calmodulin via Interactions with N and C Termini of a 1C *

Neuronal voltage-dependent Ca 2 1 channels of the N ( a 1B ) and P/Q ( a 1A ) type are inhibited by neurotransmit- ters that activate G i/o G proteins; a major part of the inhibition is voltage-dependent, relieved by depolarization, and results from a direct binding of G bg subunit of G proteins to the channel. Since cardiac and neuronal L-type ( a 1C ) voltage-dependent Ca 2 1 channels are not modulated in this way, they are presumed to lack interaction with G bg . However, here we demonstrate that both G bg and calmodulin directly bind to cytosolic N and C termini of the a 1C subunit. Coexpression of G bg reduces the current via the L-type channels. The inhibition depends on the presence of calmodulin, occurs at basal cellular levels of Ca 2 1 , and is eliminated by EGTA. The N and C termini of a 1C appear to serve as partially independent but interacting inhibitory gates. Deletion of the N terminus or of the distal half of the C terminus eliminates the inhibitory effect of G bg . Deletion of the N terminus profoundly impairs the Ca 2 1 /calmodulin-de-pendent inactivation. We propose that G bg and calmodulin regulate the L-type Ca 2 1 channel in a concerted manner via a molecular inhibitory scaffold formed by N and C termini of a 1C . Voltage-dependent Ca 2 1 channels (VDCCs) 1 are crucial for

Neuronal voltage-dependent Ca 2؉ channels of the N (␣ 1B ) and P/Q (␣ 1A ) type are inhibited by neurotransmitters that activate G i/o G proteins; a major part of the inhibition is voltage-dependent, relieved by depolarization, and results from a direct binding of G␤␥ subunit of G proteins to the channel. Since cardiac and neuronal L-type (␣ 1C ) voltage-dependent Ca 2؉ channels are not modulated in this way, they are presumed to lack interaction with G␤␥. However, here we demonstrate that both G␤␥ and calmodulin directly bind to cytosolic N and C termini of the ␣ 1C subunit. Coexpression of G␤␥ reduces the current via the L-type channels. The inhibition depends on the presence of calmodulin, occurs at basal cellular levels of Ca 2؉ , and is eliminated by EGTA. The N and C termini of ␣ 1C appear to serve as partially independent but interacting inhibitory gates. Deletion of the N terminus or of the distal half of the C terminus eliminates the inhibitory effect of G␤␥. Deletion of the N terminus profoundly impairs the Ca 2؉ /calmodulin-dependent inactivation. We propose that G␤␥ and calmodulin regulate the L-type Ca 2؉ channel in a concerted manner via a molecular inhibitory scaffold formed by N and C termini of ␣ 1C .
Voltage-dependent Ca 2ϩ channels (VDCCs) 1 are crucial for neuronal and muscular excitability (1). Mammalian VDCCs fall into several families distinguished by pharmacological and biophysical properties (L, N, P/Q, T, and R type) and the molecular identity of the main, poreϪforming subunit, ␣ 1 (2-4). The neuronal N-and P/Q-type channels, based on ␣ 1B and ␣ 1A , respectively, are crucial for neurotransmitter release (3). Ltype Ca 2ϩ channels containing the "cardiac-type" ␣ 1C subunit regulate contraction of cardiac and smooth muscle, and excitability and gene expression in the brain (2, 5, 6). The ␣ 1 subunits contain four homologous membrane domains numbered IϪIV and 5 large intracellular segments: N terminus (NT), C terminus (CT), and linkers between the domains (often called loops L 1 , L 2 , and L 3 ). There is also a large number of short intracellular linkers between transmembrane segments within each domain.
Activation in all voltage-dependent channels is initiated by a voltage-driven shift in charged transmembrane elements (7). Nevertheless, the parts of the channel and the auxiliary subunits which are not exposed to the membrane electrical field may substantially modulate the gating (for reviews related to Ca 2ϩ channels, see Ref. 3). In particular, VDCCs are strongly and specifically modulated by neurotransmitters acting via heterotrimeric G proteins, via actions on the cytosolic parts of the channel. Some of the modulations are mediated by G protein-triggered second messenger cascades, often via protein kinases A and C (PKA and PKC, respectively), others by a direct interaction with G protein subunits (1, 8 -14). Both PKC and PKA alter VDCC gating parameters acting via cytosolic parts of ␣ 1 or via the ancillary ␤ subunit (15)(16)(17)(18)(19).
Cytosolic parts of ␣ 1 subunits are also involved in inactivation gating. A major part of inactivation of L-type VDCCs is triggered by the entry of Ca 2ϩ (33). Calmodulin (CaM) has been recently identified as the Ca 2ϩ sensor indispensable for the Ca 2ϩ -dependent inactivation in L (␣ 1C )-and P/Q (␣ 1A )-type channels (34 -37). A CaM-binding site has been identified in the CT of ␣ 1C and ␣ 1A . Ca 2ϩ -dependent CaM interaction with this domain has been found crucial not only for the Ca 2ϩ -dependent inactivation, but also for an opposite effect of Ca 2ϩ , called Ca 2ϩ -dependent facilitation (34 -37).
Despite the outstanding role of cytosolic segments in channel modulation, our ideas of how these parts affect the gating of the VDCCs are vague. The L-type (␣ 1C ) channel is the best studied in this respect; but even here, only the roles of N and C termini have been examined. Removal of the distal half of the CT increases L-type channel currents and open probability by improving the coupling between gating charge (voltage sensor) movement and pore opening (38). It has been proposed that the CT acts as an inhibitory gate that conformationally restrains the opening of the channel (38,39). Similarly to the CT, removal of the NT also enhances the open probability and the macroscopic currents in L-type channels, and a similar role (of an inhibitory gate) for the NT has been proposed (16,17).
The L-type channel is not inhibited by neurotransmitters in a voltage-dependent manner (32, 40 -42), and the L 1 loop of ␣ 1C does not bind G␤␥ (15,26,27). Voltage-dependent facilitation has been demonstrated in L-type VDCC, but it was PKArather than G protein-dependent (Refs. 40 and 43; however, see Ref. 44). Therefore, ␣ 1C has been assumed to lack any interaction with G␤␥. ␣ 1C even served in several studies as a donor of presumably G␤␥-indifferent parts, to create chimeras with other ␣ 1 types, in the search for parts of ␣ 1 that determine the sensitivity to G␤␥. However, here we demonstrate that G␤␥ binds to NT and CT of ␣ 1C and inhibits the L-type VDCC in a voltage-independent but calmodulin-dependent manner. We identify a novel CaM-binding site in the NT, which, like the previously identified CT-binding site, is an important determinant of the Ca 2ϩ /CaM-dependent inactivation. We propose a model in which NT and CT of the L-type channel form a scaffold that plays a role of an inhibitory gate which integrates the regulatory effects of G␤␥ and CaM.

EXPERIMENTAL PROCEDURES
cDNA Constructs and mRNA-The cDNAs of the G protein subunits (bovine G␤ 1 , bovine G␤ 2, human G␤ 3, mouse G␤ 4 and bovine G␥ 2 ; provided by M. Simon, Caltech) were either amplified by polymerase chain reaction to create EcoRI sites at the 5Ј and 3Ј ends (G␤ 1 and G␥ 2 ) or excised with EcoRI from the original vectors (all the others) and then subcloned into the EcoRI site of the pGEMHE (G␤ 1 , G␤ 2, and G␥ 2 ) and pGEMHJ (G␤ 3 and G␤ 4 ) vectors (45,46). The cDNAs of CaM and CaM 1234 (47) were provided by J. P. Adelman. cDNAs and RNAs of rabbit Ca 2ϩ channels subunits ␤ 2 A and ␣ 2 /␦ were as described (48). The rabbit heart ␣ 1C cDNA (49) and all its mutants used here were subcloned into SalI/HindIII sites of the pGEM-SB vector (16). The Nterminal deletion mutant of ␣ 1C , ⌬N 2-139 , was prepared as described (16). To create the ␣ 1C C-terminal truncation mutant ⌬C 1700-end , polymerase chain reaction amplification of a C-terminal part of ␣ 1C was performed to create a stop codon after nucleotide 5274 (numbering by Ref. 49) followed by a HindIII site. The truncated cDNA was subcloned back into pGEM-SB. The ⌬N⌬C mutant was constructed by cutting and ligating the appropriate parts of ⌬N 2-139 and ⌬C 1700-end mutants. The RNAs were prepared using a standard procedure (50).
Oocytes and Electrophysiology-Xenopus laevis frogs were maintained and dissected as described (50). Oocytes were injected with equal amounts (by weight) of the mRNAs of ␣ 1C or its mutants with of ␣ 2 /␦, with or without ␤ 2 A, and incubated for 3-5 days at 20 -22°C in NDE96 solution (96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 2.5 mM Na pyruvate, 50 g/ml gentamycine, 5 mM HEPES, pH 7.5). Whole cell currents were recorded using the Gene Clamp 500 amplifier (Axon Instruments, Foster City, CA) using the two-electrode voltage clamp technique in a solution containing 40 mM Ba(OH) 2 or 40 mM Ca(OH) 2 , 50 mM NaOH, 2 mM KOH, and 5 mM HEPES, titrated to pH 7.5 with methanesulfonic acid. In some cases, a solution with 2 mM Ba 2ϩ was used (2 mM Ba(OH) 2 , 96 mM NaOH, 2 mM KOH, 5 mM Hepes, pH titrated to 7.5 with methanesulfonic acid). Stimulation, data acquisition, and analysis were performed using pCLAMP software (Axon Instruments). Ba 2ϩ currents were measured by a 200 or 400 ms step to 20 mV from a holding potential of Ϫ80 mV.
Interaction between GST Fusion Proteins and in Vitro Synthesized G ␤␥ and Calmodulin-The procedures were essentially as described (16). In brief, [ 35 S]Met/Cys-labeled G␤ 1 , G␥ 2 , CaM, and CaM 1234 were translated on the template of in vitro synthesized RNAs using a rabbit reticulocyte translation kit (Promega). The fusion proteins were synthesized and extracted from Escherichia coli using the Amersham Pharmacia Biotech kit. Purified GST fusion proteins (5-10 g) or purified GST (10 g) were incubated with 5 l of the lysate, containing the 35 S-labeled proteins in 500 l of phosphate-buffered saline with 0.05% Tween 20, for 2 h at room temperature, with gentle rocking. In some experiments the incubation was done in the presence of 1 mM CaCl 2 or 5 mM EGTA. In the experiments shown in Fig. 5C, the incubation was done in the same buffer but with varying concentrations of free Ca 2ϩ , in the presence of 2 mM EGTA; free Ca 2ϩ concentration was calculated using the MAXC program. Then 30 l of glutathione-Sepharose beads (Amersham Pharmacia Biotech) were added, and the mixture was incubated for 30 min at 4°C and washed four times in 1 ml of the same buffer. Following washing, GST fusion proteins were eluted with 30 l of 20 mM reduced glutathione in elution buffer (120 mM NaCl, 100 mM Tris-HCl, pH 8). CaM was analyzed on 15%, G␤, on 12% SDS-polyacrylamide gels. G␥ was analyzed on Tricine ready-made gels (Bio-Rad). The labeled products were identified and quantified by autoradiography using a PhosphorImager (Molecular Dynamics) as described (51). Phosphorylation of the N-terminal GST fusion protein was performed as described (17).
Presentation and Analysis of the Experimental Results-The data are presented as mean Ϯ S.E. To overcome the problem of batch-to-batch variability in current amplitudes, the results were normalized as follows (54): in each oocyte, I Ba was normalized to the mean amplitude of I Ba in the control group of oocytes of the same donor. These normalized values were averaged across all oocyte batches tested. Comparisons between two groups (e.g. control and G␤␥-expressing groups) were tested for statistically significant differences (p Ͻ 0.05 or better) using two-tailed unpaired t tests. Comparison between several groups was done using one-way analysis of variance (ANOVA) followed by Dunnet's or Tukey's tests, using the SigmaStat software (SPSS Corp.).

RESULTS
G␤␥ Binds to Intracellular Segments of ␣ 1C -Direct interaction between intracellular segments of ␣ 1C with G␤␥ was studied in vitro using fragments of ␣ 1C fused to GST, covering all the large intracellular segments of ␣ 1C (Fig. 1A). They included the whole NT (N 1-154 ), the full-length CT (C), three subdivisions of the C terminus (C 0 , C 1 , and C 2 , as shown in Fig. 1A, inset), and three interdomain linker loops (L 1 , L 2 , and L 3 ). The binding of G␤␥ to NT, CT, or its parts, and L 3 has not been examined in the past. The GST fusion proteins were immobilized on glutathione-agarose beads and assayed for interaction with in vitro translated, 35 S-labeled G␤ 1 ␥ 2 subunits. Unexpectedly, NT and CT bound G␤ 1 ␥ 2 , whereas GST alone and loops L 1 , L 2 , and L 3 did not bind G␤␥ (Fig. 1B, upper panel). The results of all experiments were quantitated by normalizing the 35 S-G␤␥ signal obtained from each GST fusion protein to that of the NT obtained in the same experiment (Fig. 1B, lower panel). The CT exhibited the strongest interaction with G␤␥, mainly localized to its proximal half (C 0 ; aa residues 1505-1846). The distal half of the CT, C 2 , showed a weak but reproducible G␤␥ binding. Since there is no binding in the middle of CT (C 1 ), it is plausible that G␤␥ binds to two separate sites in the CT, roughly in its first quarter (from the beginning of C 0 to the beginning of C 1, between aa 1505 and 1664) and the last half. NT showed intermediate G␤␥ binding. Fig. 1C shows that the binding of G␤␥ to NT and CT was Ca 2ϩ -independent: it was identical in the presence of 1 mM Ca 2ϩ or 5 mM EGTA, or with no additions (control). The binding of G␤␥ to NT (which is a PKC substrate: see Ref. 17) was not affected by phosphorylation by PKC or by the presence of the Ca 2ϩ channel ␤ 2 A subunit (data not shown).
Of the 5 known G␤ proteins (55), the highly homologous G␤ 1 through G␤ 4 often show considerable selectivity in modulating the voltage-dependent Ca 2ϩ channels (Refs. 56 -58; but see Ref. 59). However, in ␣ 1C , all four ␤ subunits (presented, in all cases, with G␥ 2 ) bound well to NT and CT (Fig. 1D). G␤ could bind to NT of ␣ 1C without the G␥, while G␥ 2 did not bind (Fig.  1E), as also has been shown for L 1 of ␣ 1B (58).
Coexpression of G␤␥ Attenuates the L-type Ca 2ϩ Channel Currents-Although L-type Ca 2ϩ currents in neurons are down-regulated by activation of G i/o -coupled neurotransmitter receptors (see "Discussion"), such modulations could not be reproduced in two expression systems, oocytes and HEK cells (31, 32, 40 -42). It seemed the G␤␥, when released from G proteins after activation of the relevant neurotransmitters by agonists, did not directly regulate the L-type channel. This leaves open the question of the functional consequences of the interaction between ␣ 1C and G␤␥ revealed by experiments of Fig. 1. To address this problem, we used coexpression methodology. L-type channels were expressed in Xenopus oocytes in full subunit composition (␣ 1C , ␣ 2 ␦ and ␤ 2 A), or without the ␤ subunit (␣ 1 ␣ 2 ␦ combination). Currents were measured using the two-electrode voltage clamp technique. At standard levels of expression of the channel used here (1 ng of RNA/oocyte for ␣ 1 ␣ 2 ␦␤, 2.5 ng/oocyte for ␣ 1 ␣ 2 ␦), the average I Ba was 1064 Ϯ 47 nA (n ϭ 58 oocytes, N ϭ 8 batches) and 324 Ϯ 15 nA (n ϭ 149, N ϭ 19), respectively.
In accord with previous reports, activation by acetylcholine of a coexpressed muscarinic m2 receptor (which couples to G i/o proteins) did not cause any consistent modulation of I Ba (data not shown). However, upon additional coexpression of G␣ o , which is indispensable for muscarinic inhibition of L-type Ca 2ϩ channels in pituitary neurons (60), acetylcholine caused a substantial inhibition of I Ba in 2 out of 5 oocyte batches tested (data not shown). The inconsistency of modulation suggests that additional unidentified proteins (lacking in some oocyte batches) are important for the effect of acetylcholine.
In contrast, coexpression of G␤ 1 ␥ 2 reproducibly decreased I Ba by ϳ40% in all oocyte batches tested (Fig. 2, A and C). In each batch, currents were normalized to the average I Ba in control group; this allowed the comparison and quantitation of effects of G␤␥ coexpression in many batches. At standard levels of channel expression, the current decrease caused by the coexpression of G␤␥ in the full subunit combination was 39.4 Ϯ 2.7% (n ϭ 58 oocytes from 8 batches), and in ␣ 1 ␣ 2 ␦ combination the decrease was 38.6 Ϯ 2% (n ϭ 149 oocytes from 19 batches; Fig. 2C). The inhibitory effect of coexpressed G␤␥ on I Ba was suppressed by coexpression of G␤␥ scavengers: free G␣ i1 and the myristoylated C-terminal fragment of GIRK1 K ϩ channel, GIRK 183-501 src (61) (data not shown). The inhibition by G␤␥ was clearly dose-dependent; a maximal inhibition was observed already at 0.5-1 ng/oocyte G␤ RNA (0.1-0.2 ng/oocyte G␥ RNA), and half-maximal effective doses were 0.17 and 0.07 ng/oocyte G␤ RNA for the full and ␣ 1 ␣ 2 ␦ subunit combinations, respectively (Fig. 2B). In all following experiments, G␤ and G␥ RNAs were always injected at supramaximal doses, 5 and 1 ng/oocyte, respectively.
We noticed that the inhibitory effect of G␤␥ coexpression became smaller when more channels were expressed. This was investigated with ␣ 1 ␣ 2 ␦ channels expressed in increasing amounts by injecting 1.25, 2.5, or 5 ng/oocyte of each subunits RNA (Fig. 2C). At the highest concentration of channel RNA, the inhibition was lost and even replaced by a small but statistically significant 17 Ϯ 5% enhancement (Fig. 2C, right  panel). The same trend was observed in the channels of full subunit combination (Fig. 2C, left panel). G␤␥ containing any of the four G␤ subunits tested in Fig. 1, G␤ 1 through G␤ 4 , inhibited I Ba approximately to the same extent as G␤ 1 with the standard RNA concentrations (Fig. 2D).
To exclude the possibility of a G␤␥ effect on the level of Ca 2ϩ channel expression, in two separate oocyte batches the RNA of G␤ 1 ␥ 2 was injected 4 days after the injection of RNAs of ␣ 1 and ␣ 2 ␦, and I Ba was measured 3 days later. (I Ba reaches a steady level 3-4 days after RNA expression; data not shown.) In these experiments, the decrease in I Ba was 35.5 Ϯ 3.7% (n ϭ 16, N ϭ 2). experiments in which the binding of G␤ 1 ␥ 2 to the various channel fragments has been compared. In each experiment, the intensity of the signal given by each fragment was normalized to that produced by G␤␥ bound to the N terminus (which was present in all experiments). The lane showing the signal obtained from the labeled protein synthesized in the reticulocyte lysate ("lysate") was loaded with 1 l of lysate, whereas 5 l were added to each reaction with GST fusion proteins. C, the binding of G␤␥ to N and C termini of ␣ 1C is Ca 2ϩ -independent. There is no difference in the level of G␤␥ binding in control conditions (when no additions to the reaction mixture have been made) and in the presence of 1 mM Ca 2ϩ or 5 mM EGTA. D, binding of G␤ 1 ␥ 2 , G␤ 2 ␥ 2 , G␤ 3 ␥ 2 , and G␤ 4 ␥ 2 to the NT and CT of ␣ 1C . E, the presence of the ␥ subunit is not necessary for the binding of G␤ 1 to the N terminus. G␤ 1 and G␥ 2 were synthesized in vitro separately and their interaction with the GST-fused N terminus was tested, also separately, by the standard procedure. The eluates of the two reactions were run on the same gel. G␤ 1 , but not G␥ 2 , was found to bind to the N terminus.
The effect of coexpression of G␤␥ on the total level of expression of ␣ 1C protein was further studied by quantitative immunoprecipitation of 35 S-labeled ␣ 1C from in vitro metabolically labeled oocytes. ␣ 1C was immunoprecipitated from manually separated plasma membranes and from the rest of the oocytes ("internal fraction") as described (51). Previously, we used 5 ng of RNA/oocyte to visualize ␣ 1C in plasma membranes (16). This time, to assure that these biochemical measurements are compatible with the electrophysiological ones described above, we used the standard RNA doses (2.5 ng/oocyte; ␣ 1 ␣ 2 ␦ subunit combination). Despite the low level of ␣ 1 C protein expression in the plasma membrane obtained under these conditions, it was clear that the amount of ␣ 1C in the plasma membrane was rather increased by coexpression of G␤␥ (Fig. 2E, left panel). Similarly, a slight increase of the amount of immunoprecipitated ␣ 1C was observed in the internal fraction containing the cytosol and the intracellular membranes (Fig. 2E, right panel). Similar results were obtained in two additional experiments with 2.5 ng of RNA/oocyte, and with higher levels of subunit expression (5 ng of RNA/oocyte; data not shown). Thus, G␤␥induced inhibition does not correlate with changes in the amount of ␣ 1C protein in plasma membrane; it is due to an effect on channels gating. Since the inhibition is lost at high levels of channel expression, we hypothesized that this effect of G␤␥ depends on the presence of a factor which is in short supply in oocytes.

Roles of N and C Termini in Gating and in
Mediating the G␤␥ Effects-To examine the functional impact of interaction with G␤␥ with NT and CT, we expressed three truncated constructs of ␣ 1C : ⌬C 1700-end ("⌬C") which lacks the distal part of the CT; ⌬N 2-139 ("⌬N") missing most of NT; and the double deletion mutant ⌬N⌬C, missing both pieces. Since it has been shown previously that neither NT nor CT deletions significantly alter the amount of ␣ 1C in the plasma membrane (16,38), we injected the standard amounts of channel RNA to obtain similar levels of channel expression. Wild-type (WT) and truncated ␣ 1C were coexpressed with ␣ 2 ␦, with or without G␤␥. Peak I Ba was increased 10-, 22-, and 80-fold by the deletion of the NT, CT, and both termini, respectively (Fig. 3A). The current-voltage (I-V) relationship was not significantly affected by each individual deletion, but I-V of ⌬N⌬C was shifted to the left (Fig. 3B). Fig. 3C shows that all deletions resulted in abolition of the G␤␥-induced decrease in I Ba (which was 26.1 Ϯ 3.7% in these three experiments). In all cases, the inhibition was lost and, in ⌬C 1700-end and ⌬N⌬C, there was even a small enhancement of the current. Thus, both N and C termini are indispensable for the G␤␥-induced inhibition.
The Inhibitory Effect of G␤␥ is Ca 2ϩ -and Calmodulindependent-Although Ba 2ϩ was used as the charge carrier, we were surprised to find that, in oocytes injected shortly before the experiment with a high-affinity Ca 2ϩ chelator, EGTA, the inhibition by coexpressed G␤␥ was lost (Fig. 4A), suggesting that the G␤␥-induced inhibition requires the presence of a certain basal level of Ca 2ϩ . Since G␤␥ binding to CT and NT was Ca 2ϩ -independent (Fig. 1), it was plausible that a Ca 2ϩ -binding protein such as CaM was involved in the G␤␥induced inhibition. To test this idea, we coexpressed the L-type channel with either the wild-type calmodulin, CaM WT , or its dominant negative mutant, Ca 2ϩ -insensitive mutant CaM 1234 (47). Interestingly, expression of CaM itself increased I Ba . At the standard level of channel expression (2.5 ng of RNA/oocyte; ␣ 1 ␣ 2 /␦ channels), the increase was not significant (22.6 Ϯ 12.7%, n ϭ 20, N ϭ 3; p ϭ 0.12). However, at the high level of channel expression (5 ng/oocyte) there was a large, highly significant increase in I Ba (63.6 Ϯ 15%; n ϭ 20, N ϭ 3, p Ͻ 0.001). This is in line with the notion that, at lower channel levels, there is enough endogenous CaM; therefore, the effect of expressed CaM becomes evident only at high levels of expressed channels. Similar results were obtained when CaM RNA was injected 2-3 days after the channels RNA (data not shown). In contrast, CaM 1234 reduced I Ba by 31 Ϯ 10% (n ϭ 13, N ϭ 2, p Ͻ 0.05) at the standard channel concentration.
As shown in Fig. 4B, at standard levels of channel expression, CaM WT did not alter the effect of G␤␥, while CaM 1234 significantly attenuated the G␤␥-induced inhibition. These results support the assumption that CaM may participate in G␤␥-dependent inhibition. It was therefore likely that CaM may be the endogenous factor "missing" for the inhibitory effect of G␤␥ when high doses of channel are expressed. To examine this assumption, we coinjected either standard (2.5 ng/oocyte) or high (5 ng/oocyte) doses of RNA of ␣ 1 and ␣ 2 ␦ subunits, with RNAs of G␤␥ and CaM WT . As in the previous experiments, we observed the inhibitory effect of G␤␥ on I Ba with the standard RNA concentration (Fig. 4C, left panel), and the opposite effect of G␤␥ when the channel was expressed at a higher level (Fig.  4C, right panel). Coexpression of CaM WT together with G␤␥ in the conditions of high level of expression of the channel resulted in a partial restoration of the inhibitory effect of G␤␥ (Fig. 4C, right panel). None of the effects of CaM coexpression could be attributed to a change in the level of ␣ 1C in the oocytes,

FIG. 3. Role of N and C termini in the inhibition of ␣ 1C by G␤␥.
A, comparison of the effects of deletions of N and C termini on the amplitude of I Ba . All ␣ 1C RNA constructs were injected at the same concentration (2.5 ng/oocyte), with 2.5 ng/oocyte ␣ 2 ␦ RNA. n ϭ 6, N ϭ 1. The inset shows Ba 2ϩ currents in representative oocytes. I Ba in each oocyte was normalized to the average current in the control group (WT ␣ 1C ). B, current-voltage (I-V) curves of Ba 2ϩ currents via channels containing WT and mutant ␣ 1C in oocytes of one batch (6 oocytes in each group). In each oocyte, I Ba at all voltages was normalized to the peak current. C, wild type ␣ 1C (2.5 ng of RNA/oocyte) or its truncated mutants ⌬N 2-139 (0.8 ng of RNA/oocyte), ⌬C 1700-end (0.8 ng of RNA/oocyte), and ⌬N⌬C (0.2 or 0.4 ng of RNA/oocyte) were coexpressed with ␣ 2 ␦ (same RNA amounts as ␣ 1C ), with or without G␤␥. In each oocyte, I Ba was normalized to the mean amplitude of I Ba in the control group of oocytes of the same donor, expressing the same ␣ 1C variant. The normalized values were averaged across all oocyte batches tested. *, statistically significant differences (p Ͻ 0.05 or better). Each bar represents results from 22 to 29 cells from three batches.

FIG. 4. The inhibitory effect of G␤␥ is Ca 2؉ -and calmodulindependent.
A, injection of EGTA (50 mM solution, 20 nl/oocyte) abolishes the inhibitory effect of G␤␥ and uncovers the G␤␥-induced enhancement of I Ba . This effect is observed with L-type channels of the full subunit composition (left pair of bars) or of the ␣ 1 ␣ 2 ␦ composition (right pair of bars). Each bar shows data from 14 -26 oocytes (3 batches) normalized as described under "Experimental Procedures." *, statistically significant differences between control and G␤␥-expressing groups (p Ͻ 0.05 or better). B, coexpression of the dominant negative mutant of calmodulin, CaM 1234 (10 ng of RNA/oocyte), but not the WT calmodulin, attenuates the inhibition caused by G␤␥. Summary of two experiments; 13 oocytes in each group. *, statistically significant differences between control and G␤␥-expressing groups (p Ͻ 0.01 or better). C, coexpression of the WT calmodulin restores the inhibitory effect of G␤␥ in oocytes expressing large amounts of L-type channels. Summary of experiments in 3 oocyte batches; each bar summarizes results from 20 oocytes. The left half of the chart shows the attenuation of I Ba in these 3 batches of oocytes under standard conditions (channel subunits RNAs at 2.5 ng/ oocyte; ␣ 1 ␣ 2 ␦ composition), with or without calmodulin. The right half of the chart shows the same but at high levels of channel expression, with 5 ng of RNA of each subunit. *, statistically significant differences between control (no G␤␥) and G␤␥-expressing groups (p Ͻ 0.05 or better). **, p Ͻ 0.001. D, coexpression of CaM does not alter the amount of ␣ 1C protein. CaM WT (10 ng of RNA/oocyte) was coexpressed with the channel (full subunit composition; 1 ng of RNA/oocyte of each subunit). After 4 days of incubation, ␣ 1C protein was immunoprecipitated from 5 whole oocytes. since the latter remained unchanged, as determined by immunoprecipitation of 35 S-labeled ␣ 1C from whole oocytes (Fig. 4D).
Interaction of Calmodulin with Cytosolic Parts of ␣ 1C and the Role of the N Terminus- Fig. 5A shows that CaM WT binds in vitro to GST fusion proteins of full-length CT and to C 0 , the fragment which includes the CaM-binding IQ motif, mapped to the proximal CT (34). We have also identified a novel CaM WTbinding site in the NT (Fig. 5A). These interactions were Ca 2ϩdependent; at 0.1 M free [Ca 2ϩ ], no binding of CaM either to NT or CT was observed, and a maximum binding was achieved at about 3 M (Fig. 5B). These results are similar to those previously reported for the CT (34,36). The CaM 1234 mutant did not bind to any of the GST fusion proteins neither in the presence nor in the absence of Ca 2ϩ (data not shown).
To study the role of the NT and the distal part of the CT in Ca 2ϩ -dependent inactivation of I Ba , we used the deletion mutants described above: ⌬N 2-139 , ⌬C 1700-end , and ⌬N⌬C. The decay of Ba 2ϩ versus Ca 2ϩ currents in oocytes injected with BAPTA was compared using the protocol of Peterson et al. (37) and Zuhlke et al. (36), shown in Fig. 5C. The changes in the extent of inactivation were quantitated by measuring the residual current fraction remaining after 400 ms of the depolarizing pulse, r 400 (36). As expected, with WT ␣ 1C , much stronger inactivation was observed in the high-Ca 2ϩ than in high-Ba 2ϩ solution (Fig. 5C). The extent of inactivation with either Ba 2ϩ or Ca 2ϩ as the charge carrier was not changed in the ⌬C 1700-end mutant, whereas deletion of the NT had profound effects on inactivation. With Ba 2ϩ , in the ⌬N 2-139 mutant the inactivation was significantly faster than in WT, whereas the inactivation in the Ca 2ϩ -containing solution was weaker than in WT (Fig. 5C). The difference between the values of r 400 in Ba 2ϩ and Ca 2ϩ was still significant (p Ͻ 0.05) but greatly reduced as compared with WT. The double deletion mutant ⌬N⌬C behaved like ⌬N 2-139 . Thus, the acceleration of L-type channel inactivation by Ca 2ϩ was severely compromised by the deletion of the NT but not of the distal portion of CT.
To study whether G␤␥ affects the Ca 2ϩ /CaM-dependent inactivation, we utilized the CaM 1234 mutant. When expressed in the oocytes, CaM 1234 suppresses the Ca 2ϩ -dependent L-type channel inactivation, presumably by replacing the endogenous CaM WT (36, 37) (see Fig. 6). However, the Ca 2ϩ -dependent inactivation remained unaffected by coexpression of G␤␥, either in the absence or presence of CaM 1234 . The inactivation of Ba 2ϩ currents was not affected by either G␤␥ or CaM 1234 (Fig.  6). Thus, coexpressed G␤␥ did not affect the inactivation process.

DISCUSSION
This work urges to reconsider the view that L-type Ca 2ϩ channels lack interaction with G␤␥ and also provides new insights into the mechanisms of L-type channel gating and its modulation by calmodulin. We demonstrate that NT and CT of ␣ 1C are partially independent, interacting inhibitory gates. G␤␥ directly binds to NT and CT. Coexpression of G␤␥ reduces the L-type Ca 2ϩ channel currents; this inhibition requires the presence of CaM. CaM binds to CT and to a novel site in NT, which plays an important role in Ca 2ϩ /CaM-dependent inactivation.
Interaction of the ␣ 1C Subunit of the L-type Channel with G␤␥ Affects the Channels Function-Using GST fusion proteins of cytosolic segments of ␣ 1C , we demonstrate that G␤␥ binds to the NT and to two separate sites in the CT of ␣ 1C . The strongest interaction is in the first half of the CT and is probably confined to the first quarter of the CT. The NT shows weaker but also substantial G␤␥ binding. In agreement with previous reports (15,26,27), the L 1 loop of ␣ 1C does not bind G␤␥, and no substantial binding is seen in other intersegment linker loops. Some isoforms of neuronal ␣ 1C have a shorter N terminus than the cardiac-type ␣ 1C used here (62). However, ␣ 1C with a fulllength N terminus is also abundant in the brain, at least in rat FIG. 5

. CaM binds to NT and CT in a Ca 2؉ -dependent manner.
A, binding of 35 S-labeled CaM to different fragments of ␣ 1C . Immobilized GST fusion proteins were incubated with calmodulin in the presence of 1 mM Ca 2ϩ . After washing, calmodulin bound to fusion proteins was eluted and resolved by SDS-PAGE. Identical data were obtained in two more experiments. B, CaM binding to N (lower panel) and C 0 (upper panel) GST fusion proteins at different free Ca 2ϩ concentrations (numbers are shown between the panels in M). C, Ca 2ϩ -dependent inactivation is attenuated by the removal of the NT. G␤␥ and/or calmodulin were coexpressed with the channel subunits (␣ 1C ϩ ␣ 2 ␦ ϩ ␤ 2 A; 1 ng/oocyte). Ca 2ϩ -dependent inactivation was studied in oocytes injected with 20 nl of 50 mM solution of BAPTA, 10 -60 min before measuring the current. Current were monitored in extracellular solutions containing 40 mM Ba 2ϩ (WT) or 2 mM Ba 2ϩ (other groups), or 40 mM Ca 2ϩ (all groups). Upper panel shows the summary from two experiments (6 -8 oocytes in each group), lower panel shows examples of I Ba or I Ca in individual oocytes (currents were scaled to peak amplitudes of I Ba or I Ca ). Single asterisks (*) show statistically significant differences in r 400 for I Ba (p Ͻ 0.05) from the WT group; double asterisks (**) show the same but for I Ca (one-way ANOVA was performed separately for I Ba and I Ca ).
FIG. 6. Depletion of G␤␥ attenuates the Ca 2؉ -dependent inactivation of L-type of Ca 2؉ channel. Ca 2ϩ -dependent inactivation is attenuated by CaM 1234 but not by the coexpression of G␤␥. G␤␥ and/or calmodulin were coexpressed with the channel subunits (␣ 1C ϩ ␣ 2 ␦ ϩ ␤ 2 A; 1 ng/oocyte) as in the previous experiments. The lower panel shows representative currents, normalized to peak, for each experimental condition used in this experiment. The upper panel shows the mean Ϯ S.E. of all cells tested (n ϭ 4 -5, N ϭ 1). (17). Therefore, the present results are relevant both to cardiac and neuronal L-type Ca 2ϩ channels.
The electrophysiological experiments in Xenopus oocytes showed that coexpression of G␤␥ caused a reduction of the whole cell the Ca 2ϩ channel current. This phenomenon was in the focus of this study. This inhibition was not due to a decrease in the amount of channel protein. This was shown by direct immunochemical measurements and by demonstrating that expression of G␤␥ still reduced I Ba at a time when the number of the functional channels had reached a steady level. Thus, the inhibition caused by coexpression of G␤␥ reflected a change in the function of the L-type channels. Both NT and CT were indispensable for this inhibition, which was lost after deletion of either NT or CT. The disappearance of G␤␥-induced inhibition could not be due to an increase in the level of ␣ 1C protein by one of the deletions, since removal of either NT or CT 1700-end does not increase the amount of ␣ 1C in the oocytes plasma membrane (16,17,38).
Under conditions that weakened the inhibitory effect of G␤␥ (EGTA, high levels of channel expression, NT and CT deletions), we often observed a small enhancement of I Ba . This phenomenon was much less pronounced and has been less well characterized. It could be due to a small increase in the amount of ␣ 1C observed upon coexpression of G␤␥ (Fig. 2E).
Presumably, overexpression of G␤␥ may cause general changes in the metabolic state of the cell and activate various signal transduction pathways, causing permanent changes in Ca 2ϩ channels state. At present we cannot exclude the possibility that these phenomena might contribute to the observed effects of G␤␥. However, the findings that the inhibition was removed acutely after EGTA injection, and by the deletion of G␤␥-binding elements, especially the NT, argue in favor of a more direct effect of G␤␥.
Possible Role of G␤␥ Modulation of the L-Type Ca 2ϩ Channel-Although our results do not explicitly assign a physiological role for the observed effects of G␤␥ discovered here, several possibilities can be proposed. The most exciting one is that the direct interaction of G␤␥ with ␣ 1C plays a role in mediating the inhibitory effects of G protein-coupled neurotransmitters on L-type channels. Neuronal L-type Ca 2ϩ channels are inhibited by neurotransmitters acting via G proteins of the G q class through an unidentified second messenger (see Ref. 63), and also via pertussis toxin-sensitive G i/o proteins (e.g. Refs. 60, 64, and 65). All these effects are voltage-independent. Although specific G␤ subunits have been shown to be indispensable for G i/o -mediated inhibitions (56), the details of the pathway(s) that involve this class of G proteins remain unknown. Even in cardiac cells, where the inhibitory effect of acetylcholine of L-type Ca 2ϩ channel has been extensively studied (8,9), the mechanism of inhibition of basal I Ca by ACh (in the absence of stimulation of adenylyl cyclase) is not fully understood (e.g. Ref. 66). Our and others' (32,40,41,67) inability to reconstitute neurotransmitter-induced inhibition of L-type I Ca in heterologous expression systems precluded the study of the mechanism of such inhibition. We stress that failure to obtain such reconstitution does not argue against either membrane delimited or second messenger-mediated mechanism. It simply suggests that an additional unknown component (e.g. a protein), present in neurons and heart, is missing in the oocytes and HEK cells (as is partially supported by our preliminary data concerning coexpression of G␣ o ). In a "normal" cell, this missing element may be needed, for example, to anchor G␤␥ near ␣ 1C ; overexpression of G␤␥ as done in our experiments is expected to increase the local concentration of G␤␥ near the channel, partially alleviating the need for the anchor.
An alternative explanation for the discrepancy between the inhibition of L-type Ca 2ϩ channel by overexpression of G␤␥ versus the absence of effects of neurotransmitters is that, indeed, there is no "fast" modulation by G␤␥ released from G␣␤␥ heterotrimers. If this is the case, the direct G␤␥-␣1C interaction still may be an important regulator on a long time scale, whereby the function of the L-type channel may be altered by long-term changes in G␤␥ concentration or its subcellular localization.
The Role of Ca 2ϩ and CaM in Regulation of L-type Channel by G ␤␥ -The inhibition by G␤␥ appeared to be Ca 2ϩ -dependent, since it was eliminated by intracellular Ca 2ϩ buffering with EGTA, although not by BAPTA (data not shown). On the other hand, G␤␥ inhibition did not require Ca 2ϩ entry (Ba 2ϩ currents were monitored), suggesting that a basal cellular level of Ca 2ϩ is sufficient. BAPTA, a faster chelator than EGTA but of lower affinity for Ca 2ϩ (68), may not be as efficient as EGTA in reducing the basal Ca 2ϩ levels under the conditions used here (about 1 mM EGTA or BAPTA in the cytosol). An alternative possibility of a Ca 2ϩ -independent "pharmacological" effect of EGTA cannot be discounted at present.
Several lines of evidence suggested that an oocyte's endogenous calmodulin is needed for the inhibitory effect of G␤␥: (i) the G␤␥-induced inhibition disappeared at high levels of channel expression, suggesting that an endogenous protein that normally regulates the channel is in short supply; (ii) coexpression of CaM restored the inhibitory effect of G␤␥ at high levels of channel expression; and (iii) the inhibitory effect of G␤␥ was strongly attenuated by the negative dominant CaM mutant, CaM 1234 , which misses all four Ca 2ϩ -binding sites (47).
CaM is an extremely important regulator of Ca 2ϩ and other ion channels (69), and our results provide new insights into the mechanisms of its action. We find that, in addition to the previously identified C-terminal CaM-binding site in ␣ 1C (containing the IQ domain, aa 1655-1665 (34,37)), CaM also binds to the NT. Deletion of NT greatly attenuates the Ca 2ϩ -dependent inactivation, supporting an important role for this CaMbinding site.
G␤␥ appears to affect CaM action on the L-type Ca 2ϩ channel. Since expression of G␤␥ does not affect the Ca 2ϩ /CaM-dependent inactivation (see Fig. 6), it is unlikely that G␤␥ interferes with this process. The facilitating effect of CaM is less obvious than inactivation in the oocyte system, but we propose that facilitation is reflected in the increase in I Ba caused by coexpression of CaM. This effect was observed under "resting" conditions (no Ca 2ϩ influx, no coexpressed G␤␥). In support of this hypothesis, the anti-CaM treatment by CaM 1234 decreased I Ba . Furthermore, coexpression of CaM increased I Ba at high levels of channel expression, when the endogenous CaM might be in short supply, but only marginally affected the channel at low expression levels (when the endogenous CaM is sufficient). These data suggest that, under resting conditions, CaM constitutively facilitates the function of the Ca 2ϩ channel. The inhibition by G␤␥ would then oppose this facilitating effect of CaM, explaining the Ca 2ϩ /CaM and expression level dependence of G␤␥ modulation. There are precedents for opposing interactions between CaM and G␤␥: G␤␥ binding to CaM prevents the formation of G␣␤␥ heterotrimers (70); G␤␥ and CaM oppositely regulate certain types of adenylyl cyclases (71); CaM promotes dissociation of bound G␤␥ from a glutamatergic metabotropic receptor (72).
Several questions remain open. First, the affinity of free CaM for Ca 2ϩ (K d of about 10 M; see Ref. 73) is too low to support the G␤␥ inhibition, which occurs at basal cellular levels of Ca 2ϩ (probably close to 0.1 M). However, the presence of CaM effectors (␣ 1C and/or G␤␥ in our case) is known to increase the affinity of Ca 2ϩ /CaM interaction (73). Although CaM does not interact with GST fusion proteins of either NT or CT at basal Ca 2ϩ levels (Fig. 5), higher affinity may be envisioned within a hypothetical binding site formed jointly by NT and CT (see below). Thus, theoretically, basal Ca 2ϩ levels may suffice to support CaM-␣ 1C interaction and thus G␤␥ inhibition, but this point requires further clarification. Second, CaM is an abundant cellular protein. It seems surprising that there is not enough for the modulation of Ca 2ϩ channels when they are expressed at relatively high levels. However, most of the cellular CaM may not be free but pre-assembled with its other effectors.
Role of N and C Termini in L-type Ca 2ϩ Channel Gating: an NT/CT Scaffold Model-In L-type Ca 2ϩ channels, NT and CT appear to act as inhibitory gates that attenuate the channels activation by depolarization; their removal increases the ionic currents without an increase in expression levels (16,38,39). Joint removal of both NT and CT 1700-end (in the ⌬N⌬C mutant), performed here for the first time, revealed an even greater increase in the macroscopic I Ba than the removal of each terminus separately. Peak I Ba was increased ϳ10-fold by the deletion of the NT, ϳ22-fold by the deletion of the CT 1700-end , and ϳ80-fold by the deletion of both (Fig. 3). Thus, each inhibitory gate (NT and CT) has an inhibitory effect of its own, and neither one is permissive for the other, otherwise the removal of either one would already have caused a maximal effect.
On the other hand, the maximal effect of the joint removal of two independent gates would be a multiple of the individual effects, i.e. a 220-fold increase in I Ba (the open probability of a single ␣ 1 ␣ 2 ␦ channel is below 0.0005 (52), and such a change is theoretically possible). Furthermore, the current-voltage curve was not significantly affected by either CT or NT deletion (16,38,39), but was shifted to the left by deletion of both termini (Fig. 3), suggesting their synergistic effect on activation gating. Finally, both NT and CT were indispensable for the G␤␥induced inhibition: removal of either terminus eliminated the inhibition. Therefore, we conclude that CT and NT are inter-acting gates. Importantly, the major G␤␥-binding site of the CT is located before aa 1700. The fact that the deletion of the CT beyond aa 1700 fully eliminates the G␤␥ effect implies that it is the removal of the gating element, rather than of the G␤␥binding site, which causes this effect. (Alternatively, it may suggest a great functional importance for the weak G␤␥-binding site found in the second half of CT.) Based on these considerations, we propose that C and N termini act as partially independent but interacting gates that form a molecular scaffold at the cytoplasmic side of the channel, which deters channel opening, probably by an allosteric mechanism. Its conformation changes as a function of the presence of Ca 2ϩ and of the extent of interaction between G␤␥ and CaM; other cytosolic parts of ␣ 1C and the Ca 2ϩ channel ␤ subunit may contribute to scaffolds formation. We propose that factors that "tighten" the interaction between NT and CT strengthen the inhibitory control exerted by the scaffold. Fig. 7 presents a simplified scheme of this model and of the way G␤␥ and CaM regulate gating. The scaffold harbors at least two binding sites for each CaM and G␤␥, in each of the termini. G␤␥ binds to the scaffold in the absence of elevated Ca 2ϩ and reduces the current by strengthening the interaction between CT and NT and/or by opposing a "basal" facilitating effect of CaM. Influx of Ca 2ϩ via the open channel is followed by Ca 2ϩ binding to CaM which, in turn, improves the interaction of CaM with its binding sites and further tightens the scaffold, causing more inhibition (which is apparent as Ca 2ϩ -dependent inactivation). The latter effect is G␤␥-independent.
This model actually presents a further elaboration of that proposed by Peterson et al. (37) to explain Ca 2ϩ -dependent inactivation. It has been envisaged that CaM, which in the absence of Ca 2ϩ must be tethered to a site distinct from the C-terminal IQ domain, binds to the latter following Ca 2ϩ influx and binding. Together with the yet undefined tethering element, the CT has been proposed to form a scaffold which, after binding of Ca 2ϩ /CaM, induces inactivation. The necessity for a CaM tethering element that should bind CaM in the absence of Ca 2ϩ was also stressed by Qin et al. (34) and Zuhlke et al. (36).
However, although we have tested the binding of CaM to all large cytosolic parts of ␣ 1C , no Ca 2ϩ -independent binding sites could be found. Furthermore, the CaM 1234 mutant, which is supposed to act by replacing the endogenous WT CaM at the tethering site, did not bind to any of our GST fusion proteins either in the absence or presence of Ca 2ϩ . Possibly, the threedimensional structure of the N/C-terminal scaffold is such that the NT and CT CaM-binding sites actually form one, highaffinity site which may anchor CaM (or CaM 1234 ) in the absence of elevated Ca 2ϩ (alternatively, the small linkers within the homology domains may contribute to the anchoring site). A high-affinity site for G␤␥ may also be jointly formed by the three sites (NT, proximal CT, distal CT) identified here, eventually providing for a 1:1 stoichiometry for G␤␥ binding (as appears to be the case with ␣ 1B ; see Ref. 25).

FIG. 7. Schematic representation of the NT/CT scaffold model.
A cartoon of channel states is shown at the top; corresponding Ba 2ϩ or Ca 2ϩ currents are drawn at the bottom. In the resting state, at basal cellular Ca 2ϩ levels, NT and CT are loosely folded; CaM may be anchored with only a fraction of its Ca 2ϩ -binding sites occupied by Ca 2ϩ ; depolarization causes large, poorly inactivating currents in a high-Ba 2ϩ , Ca 2ϩ -free solution (left). Binding of G␤␥ causes tightening of the scaffold and a decrease in the current (middle). A switch to a Ca 2ϩcontaining solution and the entry of Ca 2ϩ causes a forceful tightening of the scaffold and inactivation, due to strong binding of CaM to its NTand CT-binding sites; this process is largely unaffected by G␤␥ .