Stargazin Modulates Neuronal Voltage-dependent Ca2+ Channel Cav2.2 by a Gβγ-dependent Mechanism*

Loss of neuronal protein stargazin (γ2) is associated with recurrent epileptic seizures and ataxia in mice. Initially, due to homology to the skeletal muscle calcium channel γ1 subunit, stargazin and other family members (γ3–8) were classified as γ subunits of neuronal voltage-gated calcium channels (such as CaV2.1-CaV2.3). Here, we report that stargazin interferes with G protein modulation of CaV2.2 (N-type) channels expressed in Xenopus oocytes. Stargazin counteracted the Gβγ-induced inhibition of CaV2.2 channel currents, caused either by coexpression of the Gβγ dimer or by activation of a G protein-coupled receptor. Expression of high doses of Gβγ overcame the effects of stargazin. High affinity Gβγ scavenger proteins m-cβARK and m-phosducin produced effects similar to stargazin. The effects of stargazin and m-cβARK were not additive, suggesting a common mechanism of action, and generally independent of the presence of the CaVβ3 subunit. However, in some cases, coexpression of CaVβ3 blunted the modulation by stargazin. Finally, the Gβγ-opposing action of stargazin was not unique to CaV2.2, as stargazin also inhibited the Gβγ-mediated activation of the G protein-activated K+ channel. Purified cytosolic C-terminal part of stargazin bound Gβγ in vitro. Our results suggest that the regulation by stargazin of biophysical properties of CaV2.2 are not exerted by direct modulation of the channel but via a Gβγ-dependent mechanism.

Stargazin (␥ 2 ) is a neuronal protein belonging to a family of transmembrane AMPA 3 receptor regulatory proteins, whose mutations are associated with recurrent epileptic seizures and ataxic gait in mice (1). Its most generally accepted role is in regulating early intracellular transport, synaptic targeting, anchoring, pharmacology, and function of the AMPA receptor in cerebellar neurons (2,3). However, when first identified, stargazin and its other six family members (termed ␥ 3-8 ) were classified as ␥ subunits of neuronal calcium channels for several reasons. First, the neurological deficits observed in stargazer mice are associated with mutations in the Ca 2ϩ channel subunits Ca V 2.1 or ␣ 1A (tottering), Ca v ␤ 4 (lethargic), and ␣ 2 ␦ (ducky) (4). Second, stargazin shares structural similarity with the ␥ 1 subunit of skeletal muscle Ca 2ϩ channel (25% amino acid identity, exon-intron organization, and predicted secondary structure of four transmembrane domains with intracellular N and C termini) (1). Finally, functional studies revealed similar inhibitory effects of ␥ 1 and stargazin on neuronal calcium channels (5).
Neuronal voltage-dependent calcium channels (Ca V 2.1-2.3) are crucial regulators of neurotransmitter release from presynaptic nerve terminals (6). They control neuronal excitability and communication and are tightly modulated by various regulatory mechanisms. One important modulatory mechanism is their inhibition by the G protein ␤␥ subunit dimer, G␤␥, derived from the activation of G protein-coupled receptors (GPCRs) acting via G i/o (7,8). Binding of G␤␥ to the ␣ 1 subunit of Ca V 2.2 produces an array of biophysical effects, including a decrease in current amplitude, which can be relieved by strong depolarization (termed voltage dependent facilitation, VDF); slowing of the kinetics of activation and inactivation; and a depolarizing shift of voltage dependences of activation and inactivation (for review, see Ref. 8). Ca 2ϩ channels are multisubunit proteins comprised of a main pore-forming ␣ 1 subunit and three ancillary subunits: ␣ 2 ␦, ␤ and, in skeletal muscle, also ␥ 1 (9). The identification of the putative neuronal ␥ subunits of Ca 2ϩ channels have prompted a multitude of studies, both in heterologous systems and native tissue, mainly with Ca V 2.1 and Ca V 2.2. Stargazin was found in association with these channels using sucrose gradient fractionation and immunoprecipitation methods (5,10). At the cellular level, coexpression of Ca V 2.2 channels with stargazin or other ␥ proteins in heterologous expression systems caused down-regulation of channel expression and decreased current amplitudes (5,(11)(12)(13). Transfection of ␥ 7 in cultured sympathetic neurons and of stargazin in cultured dorsal root ganglion neurons, which natively express Ca V 2 channels, produced variable effects, with no changes in channel expression and trafficking in the former (12) but with a marked decrease in Ca 2ϩ current density in the latter (11). In all, however, negative regulation of Ca V 2 channel expression by ␥ proteins is well established.
However, whether or not stargazin regulates the function of Ca V 2 channels remains controversial. Coexpression of stargazin with Ca V 2.1 or Ca V 2.2 in mammalian cells affected the volt-age dependence of channel activation (11) and inactivation (1,14), whereas coexpression of stargazin with Ca V 2.2 in Xenopus oocytes caused a deceleration of activation kinetics (5). In contrast to these reports, others (15,16) maintained that coexpression of ␥ 2 , ␥ 3 , or ␥ 4 with Ca V 2.1 in Xenopus oocytes had no major effects on current amplitude or the biophysical properties of the channel.
Here, we report a novel mechanism of Ca V 2.2 channel modulation by stargazin. When coexpressed in Xenopus oocytes, it disrupts the G␤␥-mediated modulation of Ca V 2.2, mimicking the effects of well characterized G␤␥ scavenger proteins. Interestingly, this effect is not specific to Ca V 2.2, because stargazin also opposes the effect of G␤␥ on the G protein-activated inwardly rectifying K ϩ channel (GIRK), which is activated by direct binding of G␤␥. Accordingly, we find that stargazin binds G␤␥ in vitro. On the other hand, stargazin reduced Ca V 2.2 expression but apparently by a G␤␥-independent mechanism. We propose that stargazin is not a genuine subunit of Ca 2ϩ channels; it regulates Ca V 2 and probably other G␤␥ effectors by opposing G␤␥-mediated effects. The controversial reports regarding the effects of stargazin on Ca V 2 gating may reflect variable levels of endogenous free G␤␥ in cells used for heterologous expression.
Xenopus Oocyte Preparation and Electrophysiology-Experiments were approved by the Tel Aviv University Institutional Animal Care and Use Committee (permit no. 11-05-064) as described (21). Briefly, portions of ovary were removed through a small incision on the abdomen from female frogs anesthetized in a 0.15% solution of procaine methanesulphonate (MS222). Oocytes were defolliculated using collagenase type IA (Sigma), isolated, injected with cRNA, and incubated for 3 days in ND-96 solution containing: 96 mM NaCl, 2 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 5 mM Hepes/NaOH; pH 7.6) and supplemented with 2.5 mM sodium pyruvate and 50 g/ml gentamycine. All experiments were performed at 20 -22°C.
Whole cell Ca 2ϩ channel currents were measured with the Gene Clamp 500 amplifier (Axon Instruments) using the twoelectrode voltage clamp method, with Ba 2ϩ as the charge carrier in a solution containing: 5 or 20 mM Ba(OH) 2 , 85 mM NaOH, 2 mM KOH, 5 mM HEPES; pH was adjusted to 7.5 with methanesulphonic acid (Sigma)). Oocytes were injected with Ca 2ϩ chelators BAPTA or EGTA (25 nl of 50 mM stock) prior to current measurement to inhibit endogenous calcium-dependent chloride currents. Ba 2ϩ currents (I Ba ) from oocytes expressing ␣ 1B /␣ 2 ␦ channels were recorded in high (20 mM) Ba 2ϩ solution, while currents from oocytes expressing ␣ 1B / ␣ 2 ␦/␤ 3 channels were recorded in low (5 mM) Ba 2ϩ as well as high (20 mM) Ba 2ϩ solutions to achieve currents ranging from 200 -5000 nA. At the end of experiment, recording protocols were repeated in the presence of 200 M Cd 2ϩ to block I Ba , and net I Ba was obtained by subtracting the residual Cd 2ϩ -resistant leak and endogenous currents. Low Ba 2ϩ (5 mM) and coexpression of Ca V ␤ 3 cause maximal activation at lower voltages than in 20 mM Ba 2ϩ or without Ca V ␤ 3 , respectively. Thus, oocytes expressing ␣ 1B /␣ 2 ␦ channels exhibited peak currents at ϩ20 mV, whereas oocytes expressing ␣ 1B /␣ 2 ␦/␤ 3 channels exhibited peak currents at 0 or ϩ10 mV. Whole-cell GIRK currents were measured using standard procedures in the ND96 (low K ϩ ) solution containing: 96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 5 mM HEPES, pH 7.5) or in a high K ϩ solutions (24 mM K ϩ , isotonically replacing NaCl in ND96) as described (21). Acetylcholine (ACh) was used at 1 nM (low dose) and 10 M (high dose), diluted in either Ba 2ϩ or high K ϩ solution depending on the channel studied. Agar-cushion electrodes (24) filled with 3 M KCl with resistances of 0.1-0.4 megohms were used in all experiments. Data acquisition and analysis were done using the pCLAMP 9 software (Axon Instruments). I-V curves were fitted with a standard Boltzmann equation in the form I Ba ϭ G max (V m Ϫ V rev )/(1 ϩ exp(Ϫ(V m Ϫ V 0.5 act /K a ))), where K a is the slope factor, V 0.5 act is the voltage that causes half-maximal activation, G max is the maximal conductance, V m is membrane voltage, I Ba is the current measured at the same voltage, and V rev is the reversal potential of I Ba . The conductance-voltage (G-V) curves were plotted with the values of V a and K a obtained from the fit of the I-V curves, using the following form of the Boltzmann equation: G/G max ϭ 1/(1 ϩ exp((Ϫ(V m Ϫ V 0.5 act )/ K a ))). Steady state inactivation curves were fitted with a standard Boltzmann equation in the form I Ba ϭ f ϩ ((1 Ϫ f)/(1 ϩ exp((V m Ϫ V 0.5 inact )/K i ))), where K i is the slope factor, V 0.5 inact is the voltage that causes half-maximal inactivation, and f is the fraction of channels that remained activated at a given voltage.
Protein Purification and Pulldown Assay-GST-G␣ i1 and GST-stg(204 -323) were purified as described (25). His 6 tagged G␣ i1 and G␤␥ were purified as described previously (26,27). [ 35 S]Methionine-labeled G␤␥, G␣ i1 , and PSD-95 were synthesized in vitro in rabbit reticulocyte lysate (Promega Corp., Madison, WI). Interaction among GST-fused GST-stg(204 -323) or GST-G␣ i1 with the in vitro translated (ivt) [ 35 S]methioninelabeled proteins was studied by pulldown on glutathione affinity beads in high K ϩ buffer containing: 150 mM KCl, 50 mM Tris, 5 mM MgCl 2 , 1 mM EDTA, pH 7.0 with an addition of 0.1% Lubrol and 30 M GDP) in a reaction volume of 300 l, as described (25). Amount of the GST-fused proteins in the reactions was: 4 g GST-G␣ i3 ; 10 g GST-stg(204 -323); 10 g GST. The following volumes of reticulocyte lysate containing the in vitro translated 35 S-labeled proteins were taken for pulldown reactions: 5 l G␣ i3 ; 5 l G␤␥; 15 l PSD-95. The eluted proteins were separated on 12% SDS polyacrylamide gels. The radioactive signals from protein bands of the gels were imaged and quantitated using PhosphorImager and ImageQuant software (Molecular Dynamics). The interaction between the GSTstg(204 -323) or the GST-G␣ i1 and purified His 6 -tagged G␤␥ was performed using pulldown on glutathione affinity beads. Amounts of GST and His-tagged proteins for Western blot experiments were: 4 g GST-G␣ i1 ; 10 g GST-stg(204 -323); 10 g GST; 2 or 10 g His-G␤␥; 8 g His-G␣ i1 . The eluted proteins were separated on 12% SDS-PAGE gels and transferred to nitrocellulose membranes for Western blotting. Membranes were blocked with 5% milk in TTBS solution (containing 10 mM Tris-HCl, 100 mM NaCl, and 0.1% Tween 20, ph 7.4) and then incubated with appropriate antibodies for 2 h at room temperature. Secondary horseradish peroxidase conjugated anti-rabbit antibody was then applied to visualize the bands using standard commercial kit. The proteins on the membrane were visualized using Ponceau (Sigma) for input and using antibodies for binding. Rabbit anti-G␤ 1 and anti-G␣i 3/1 primary antibodies (Santa Cruz Biotechnology), as well as goat anti-rabbit horseradish peroxidase-conjugated secondary antibody (Jackson ImmunoResearch Laboratories) were used in Western blotting experiments. Bound proteins were visualized on film with ECL reagents from Pierce.
Confocal Imaging-Imaging of proteins in the plasma membrane was performed in whole oocytes. Oocytes were placed in a chamber with a transparent bottom, and fluorescence imaging from YFP-or CFP-tagged expressed proteins was performed with a Zeiss 510 Meta confocal scanning microscope (ϫ20 objective, zoom ϫ2, pinhole 3 Airy units). CFP and YFP fluorescence was excited by 405 and 514 laser lines, respectively, and collected using 470 -500 nm and 505-550 band pass filters, respectively, and background signals were subtracted. All images were obtained from optical slices from the animal hemisphere close to the equator of the oocyte. Quantification of the intensity of fluorescence in the plasma membrane was measured by averaging the signal obtained from three standard regions of interest as described (20). Net fluorescence intensity per unit area was obtained by subtracting the background signal measured in uninjected oocytes. In all confocal imaging procedures, care was taken to completely avoid saturation of the sig-nal. In each experiment, all oocytes from the different groups were studied using constant LSM settings.
Statistical Analysis-Multiple group comparisons were done using one-way analysis of variance followed by all-pair Student Newman-Keuls test. Pairwise comparisons were done by nonpaired one-tailed t tests.

RESULTS
In initial experiments, we have noticed that stargazin appeared to oppose the G␤␥-mediated inhibition of Ca V 2.2. To elucidate whether stargazin acts on Ca V 2.2 directly or by opposing the effects of G␤␥, we investigated the mutual influence of stargazin and G␤␥ on the gating of the channel.
Stargazin Attenuates G␤␥-mediated Voltage-dependent Facilitation-G␤␥ inhibition of Ca V 2.2 is voltage-dependent, and it is relieved by strong depolarization (for review, see Refs. 8,28,29). This relief, called VDF, is a hallmark of G␤␥ modulation of Ca V 2 channels. To monitor VDF, we measured Ba 2ϩ currents (I Ba ) via the Ca V 2.2 channels under voltage clamp conditions, using a standard protocol (Fig. 1A, a), as described (30 -33). A reference Ba 2ϩ current, termed I Ba (Ϫpp), was elicited by a test pulse from Ϫ80 to 0 or ϩ20 mV. After 8 s, the test pulse was repeated, but this time, it was preceded by a short (60 ms) prepulse to ϩ100 mV. The prepulse removes much of the voltagedependent G␤␥ inhibition (if present), and the I Ba elicited by this test pulse, I Ba (ϩpp), is greater than I Ba (Ϫpp) (Fig. 1A, b). The extent of VDF is quantified by facilitation ratio, [I Ba (ϩpp)]/ [I Ba (Ϫpp)] (Fig. 1A, b). VDF is absent when the facilitation ratio is 1.
VDF in Ca V 2.2 expressed in Xenopus oocytes shows a complex dependence on G␤␥ dosage. As shown previously (30), expression of increasing doses of G␤␥ generated a biphasic effect on VDF (Fig. 1C, empty bars). In the ␣ 1B /␣ 2 ␦/␤ 3 channel, the highest facilitation ratio was obtained when low and intermediate doses of G␤␥ (0.2/0.04 ng and 2.5/0.5 ng of RNA/oocyte) were coexpressed, while a high dose (5/1 ng of RNA/oocyte) decreased VDF. This bell-shaped dose dependence highlights the proposed molecular mechanism of VDF: a depolarization-induced decrease in affinity of G␤␥ binding to the channel (29). As the dose of G␤␥ rises, the inhibition of I Ba grows, and so does the relief of inhibition caused by the depolarizing prepulse. However, at very high doses of G␤␥, the decrease in affinity caused by the depolarization is overridden by excess G␤␥.
Coexpression of stargazin (0.2-5 ng of RNA/oocyte) caused a significant decrease of basal VDF and of that caused by the low dose of coexpressed G␤␥ (Fig. 1C) and attenuated VDF caused by an intermediate dose of G␤␥ (2.5/0.5 ng of RNA). Significant suppression of VDF was observed in some experiments already at low doses of stargazin (0.2-1 ng of RNA). However, stargazin had no effect on VDF induced by the high dose of G␤␥ (Fig. 1C). Qualitatively similar results were obtained in the ␣ 1B /␣ 2 ␦ channels (supplemental Fig. S1A).
For comparison, we expressed the myristoylated C-terminal part of the ␤-adrenergic receptor kinase (c␤ARK; 5-10 ng of RNA/oocyte) or myristoylated phosducin (10 ng of RNA/oocyte). c␤ARK and phosducin are "G␤␥ scavengers" that bind G␤␥ with high affinity (ϳ30 -50 nM in solution) (39,40) and are widely used to counteract G␤␥-induced modulation of G protein-regulated Ca 2ϩ and K ϩ channels (18,21,30,41,42). Both proteins elicited virtually identical effects, eliminating G␤␥mediated modulation of Ca V 2.2. Coexpression of c␤ARK or phosducin significantly decreased basal and G␤␥-induced VDF in ␣ 1B /␣ 2 ␦ channels at all G␤␥ doses (supplemental Fig. S1A). In ␣ 1B /␣ 2 ␦/␤ 3 channels, c␤ARK eliminated the basal VDF but failed to eliminate VDF at a high dose of G␤␥ (Fig. 1C, gray  bars). A slight increase in VDF in the latter case was not significant and could reflect an incomplete G␤␥ scavenging, possibly leaving an intermediate level of free G␤␥ and thus a higher VDF.
Importantly, when stargazin and c␤ARK were expressed together, their joint effect on facilitation was not significantly different than that exerted by either of the proteins expressed alone (supplemental Fig. S1A, four middle bars). In all, the results of this series of experiments showed similarity in actions of stargazin and G␤␥ scavenger proteins on Ca V 2.2. Notably, like in mammalian cells (5,11), stargazin greatly reduced surface expression of Ca V 2.2 in Xenopus oocytes, but this effect was not reversed by coexpression of G␤␥ (supplemental Fig. S5A).
Stargazin Opposes G␤␥-induced Changes in Activation and Inactivation Kinetics and Voltage Dependence of Ca V 2.2-A prominent voltage-dependent effect of G␤␥ is the slowing of activation kinetics of Ca V 2 channels (8). Fig. 2A shows exemplary traces from oocytes expressing ␣ 1B /␣ 2 ␦/␤ 3 channels with or without G␤␥ and stargazin, and Fig. 2B presents a summary of data. Kinetics of activation was assessed by calculating the time from the beginning of the depolarizing pulse, without prepulse, to 90% of peak amplitude (t 90% ). G␤␥ caused a dose-dependent slowing of activation kinetics, from 5-6 ms to 13-16 ms, with higher doses eliciting greater effect ( Fig. 2B and supplemental Fig S1B, empty bars). Coexpression of stargazin counteracted the kinetic slowing induced by G␤␥ (Fig. 2B), completely eliminating the effect of G␤␥ at low and intermediate G␤␥ doses and attenuating it at the highest G␤␥ dose. Elimination of G␤␥-induced kinetic slowing by stargazin was also observed in the absence of Ca V ␤ 3 (supplemental Fig. S1B; note that the ␣ 1B /␣ 2 ␦ channel showed less pronounced kinetic slowing than ␣ 1B /␣ 2 ␦/␤ 3 ). Similarly to stargazin, coexpression of c␤ARK or phosducin counteracted the kinetic slowing even at the highest dose of G␤␥ ( Fig. 2B and supplemental Fig. S1B). Stargazin and c␤ARK had no effect on the basal level of activation kinetics (i.e. in the absence of coexpressed G␤␥), presumably because endogenous concentrations of G␤␥ were too low to cause measurable changes in this parameter.
G␤␥ also interferes with the voltage-dependent inactivation (VDI) of Ca V 2 channels (43,44). To examine whether stargazin opposes this effect of G␤␥, we measured the time course of VDI by applying 2500-ms pulses from the holding potential of Ϫ80 mV to voltages ranging from Ϫ10 mV to ϩ40 mV. The extent of VDI was quantitated as the fraction of the current remaining 850 ms after the beginning of the depolarizing pulse (R 850 ). In ␣ 1B /␣ 2 ␦ channels, G␤␥ caused a marked dose-dependent slow- ing of inactivation kinetics (Fig. 3, A and B). This decelerating effect was voltage-dependent, as it was absent at stronger depolarizations. Coexpression of either stargazin or c␤ARK eliminated the effect of G␤␥ and even accelerated the inactivation kinetics compared with control (channel alone) when a low dose of G␤␥ was expressed (Fig. 3A). Coexpression of both stargazin and c␤ARK (Fig. 3A) did not produce an effect significantly different from that caused by either of the proteins expressed alone. When high dose (5/1 ng) of G␤␥ was expressed, stargazin significantly accelerated inactivation kinetics though it did not restore the initial speed of VDI, whereas phosducin completely abolished the effect of G␤␥ (Fig. 3B).
Coexpression of Ca V ␤ 3 significantly accelerated the inactivation kinetics at all voltages tested (34) and rendered VDI with a biphasic time course and complex U-shaped voltage dependence, as observed previously (5,45). Coexpression of G␤␥ accelerated inactivation kinetics in a dose-and voltage-dependent manner (supplemental Fig. S2). Stargazin decelerated VDI at the low dose of G␤␥ but had no effect at 5/1 ng of G␤␥. c␤ARK decelerated VDI kinetics even at the high dose of G␤␥ (supplemental Fig. S2).
Another important manifestation of G␤␥ modulation of Ca V 2.2 is rendering the channel less available for activation by voltage. This is seen as a depolarizing (positive) shift of voltage dependence of activation (assessed from current-voltage (I-V) and corresponding conductance-voltage (G-V) relations) and inactivation (steady state inactivation relation) (8). We examined the effect of stargazin on these parameters. In ␣ 1B /␣ 2 ␦ channels, coexpression of the high dose of G␤␥ shifted the I-V and G-V curves to positive voltages; half-activation voltage (V 0.5 act ) changed by ϳϩ10 mV (Fig. 4A and B; see Table 1 for details of Boltzmann fits). Coexpression of stargazin counteracted the effect of G␤␥, shifting the V 0.5 act by ϳϪ5 mV, whereas phosducin restored V 0.5 act to control values (Fig. 4B, Table 1). Steady state inactivation (Fig. 4C) was studied with 2500-ms prepulses to different voltages, followed by a test pulse to ϩ20 mV. G␤␥ (5/1 ng) shifted half-maximal inactivation voltage by ϳ ϩ10 mV (Fig. 4C, Table 2), whereas coexpression of stargazin or phosducin reversed this effect (Fig. 4C, Table 2). A low dose of G␤␥ had little or no effect on voltage dependence of activation and inactivation (supplemental Fig. S3 and supplemental Tables S1-S4). In the presence of Ca V ␤ 3 , stargazin efficiently countered the effects of a low dose of G␤␥ but not of a high dose (supplemental Fig. S3B and Tables S3 and S4).
In summary, all of the inhibitory actions of G␤␥ on Ca V 2.2 gating were countered by stargazin, both in the presence and absence of coexpressed Ca V ␤ 3 . The effects of stargazin resembled, and were not additive with, those of G␤␥-scavenging proteins m-phosducin and c␤ARK. The effects of stargazin could be overcome by increasing the dose of G␤␥.
Stargazin Reduces Inhibition of Ca V 2.2 caused by GPCR Activation-In native cells, Ca V 2.2 channels are inhibited by the G␤␥ released following activation of G i/o -coupled GPCRs. However, no effects of stargazin of this classical modulation have been reported, possibly because the stargazin-G protein connection has not been suspected. We tested the effect of stargazin on regulation of Ca V 2.2 by coexpressed muscarinic type 2 receptor (m 2 R). To simulate low and high doses of G␤␥, we used two concentrations of the m 2 R agonist acetylcholine (ACh), a low one (10 Ϫ9 M) and a high one (10 Ϫ5 M), as described previously (30). The protocol of Fig. 1A was used to measure the effect of agonist application on current amplitude and VDF. Fig.  5A shows typical traces from representative oocytes expressing ␣ 1B /␣ 2 ␦/␤ 3 channels with m 2 R (upper traces) and either stargazin (middle traces) or c␤ARK (bottom traces). Application of ACh caused a dose-dependent inhibition of current amplitude and an increase in VDF. The effect of ACh was stronger in ␣ 1B /␣ 2 ␦ than in ␣ 1B /␣ 2 ␦/␤ 3 channel (compare Fig. 5, B and C). Stargazin substantially attenuated the decrease in current amplitude caused by 1 nM ACh but did not alter the effect of the high ACh dose, 10 M (Fig. 5B, a and C, a). Similarly, stargazin almost fully reversed the increase in VDF caused by 1 nM ACh but only partially attenuated it (in ␣ 1B /␣ 2 ␦ channels) or was ineffective (in the ␣ 1B /␣ 2 ␦/␤ 3 channels) when the high dose of ACh was applied (Fig. 5B, b and C, b, middle bars). Finally, coexpression of c␤ARK reversed ACh-dependent modulation under all conditions tested (Fig. 5, B and C, right set of bars). In all, stargazin counteracted G␤␥-mediated modulation caused through activation of a GPCR. The inhibitory effect of stargazin was weakened by increasing the dose of ACh and by the presence of Ca V ␤ 3 .

FIGURE 2. Effect of stargazin on kinetics of activation of I Ba (t 90% ).
A, net I Ba currents in representative oocytes expressing ␣ 1B /␣ 2 ␦/␤ 3 channels with or without 5/1 ng G␤␥ and either stargazin (stg) c␤ARK. The currents were obtained by steps from Ϫ80 mV to 0 mV. B, summary of data on kinetics of activation in oocytes expressing ␣ 1B /␣ 2 ␦/␤ 3 channels. Empty bars represent oocytes expressing the channel (control) with various doses of coexpressed G␤␥ as indicated below the graph. All doses of G␤␥ caused significant kinetic slowing compared with control (p Ͻ 0.05, compare empty bars). Number of oocytes in each group (n) is shown above the respective bar. Statistical significance is shown in each treatment relative to the group expressing the channel and the respective dose of G␤␥. **, p Ͻ 0.01; ***, p Ͻ 0.001.

Is the Functional Effect of Stargazin Exclusive to Ca V 2.2 Channel?-
The above results suggest that stargazin modulates the Ca V 2.2 channel by counteracting G␤␥-mediated effects. The similarity of effects of stargazin and the G␤␥ scavengers suggests similarity of mechanisms. If this is the case, its effect should not be limited to the Ca V 2.2 channel; rather, it may modulate other G␤␥ effectors. To explore this hypothesis, we tested the effect of stargazin on another well characterized effector of G␤␥ -the G proteinactivated inward rectifying K ϩ channel (GIRK), which is activated upon direct binding of G␤␥ (46). When expressed at high levels in Xenopus oocytes, the neuronal GIRK1/2 channel, composed of GIRK1 and GIRK2 subunits, shows high basal current, I basal , which is almost fully determined by ambient G␤␥ and allows an easy monitoring of the effects of G␤␥ coexpression or G␤␥ scavengers (19,47). GIRK1/2 was expressed together with m 2 R, with or without G␤␥ and/or stargazin or m-phosducin. To control for the surface expression of the channel, we expressed YFP-or CFP-tagged GIRK1 subunit with the wild-type GIRK2 and monitored the surface expression of the channel using confocal microscopy (see Ref. 21). Surface expression of GIRK1/2 was not affected by G␤␥ but was reduced by stargazin and phosducin (supplemental Figs. S5B and S6). Therefore, to calculate the net effect of expressed proteins, currents were corrected for the reduction in channel expression (21). Fig. 6A shows representative traces recorded at Ϫ80 mV. I basal was revealed by a change from a low K ϩ external solution (ND96) to a    high K ϩ solution (24 mM K ϩ ). Additional current was evoked by addition of ACh (Fig. 6A) or by coexpression of G␤␥ (Fig. 6, B and C, striped bars). 5 mM Ba 2ϩ was added at the end of the record to fully block GIRK, allowing calculation of the net GIRK current. Fig. 6, B and C, presents the summary of GIRK1/2 currents (corrected for changes in GIRK surface expression). Stargazin significantly decreased I basal of GIRK1/2 (Fig. 6, B and C, empty  bars) as well as the extra current evoked by a low dose of G␤␥, 0.5/0.1 ng of RNA (Fig. 6B, striped bars). The effect of stargazin was milder but still significant also when a high dose (5/1 ng) of G␤␥ was expressed (Fig. 6C, striped bars). Coexpression of m-phosducin caused significant inhibition of GIRK currents under all conditions. These results further support the notion that stargazin opposes the effects of G␤␥ and demonstrate that this is not unique to the Ca V 2.2 calcium channel but may be a general mechanism involving additional G␤␥ effectors.
Is Stargazin a G␤␥-binding Protein?-One possible mechanism by which stargazin may counteract G␤␥-mediated effects is by binding G␤␥ and competing with the G␤␥ effector. We thus tested whether stargazin can bind G␤␥ in vitro (Fig. 7). Stargazin is a transmembrane protein with cytosolic N and C termini. Its C terminus is its largest cytosolic part containing a PDZ-binding motif (2). We have constructed and purified a GST-fused full C terminus of stargazin, amino acids 204 -323 (GSTstg(204 -323)) (48). We first examined whether this protein can bind in vitro translated (ivt) G␤␥ by a pulldown method (Fig. 7A). As controls we used ivt PSD-95, which was previously shown to bind the C terminus of stargazin (49 -51) and the ivt G protein ␣ i3 subunit, G␣ i3 . As expected, GST-stg(204 -323)bound ivt G␤␥ but not the ivt G␣ i3 . GST alone did not interact with either of the tested proteins (Fig.  7A). The binding of G␤␥, however, was weaker (relative to input) than that of ivt PSD-95 and also weaker than the high affinity binding of G␣ i3 to ivt G␤␥. The latter was measured by pulldown of ivt G␤␥ with a GST-fused G␣ i3 in the presence of GDP (Fig. 7A, left lane) (19).

Stargazin Modulates Ca V 2.2 by Opposing the Actions of G␤␥-
Our results suggest that stargazin acts on Ca V 2.2 indirectly, by modulating its regulation by G␤␥. This stargazin-G␤␥-Ca V 2 connection has not been recognized previously. The main lines of evidence in favor of an indirect, G␤␥-mediated effect are as follows. 1) Stargazin suppresses, fully or partially, all of the known effects of G␤␥ on Ca V 2.2: the G␤␥-induced changes in FIGURE 5. Effect of stargazin on m 2 R-mediated inhibition of Ca V 2.2. A, exemplary traces from oocytes expressing ␣ 1B /␣ 2 ␦/␤ 3 channels obtained before and after application of 10 Ϫ9 and then 10 Ϫ5 M ACh. The currents were measured using the VDF protocol shown in Fig. 1A. In each record, the two traces represent I Ba obtained by test pulse without the prepulse (usually the current with the smaller amplitude) and I Ba after the prepulse. B, summary of the effects of m 2 R activation on the current amplitude before prepulse (B, a) and on facilitation ratio (B, b) of ␣ 1B /␣ 2 ␦ channels. C, summary of the effects of m 2 R activation on current amplitude before prepulse (C, a) and facilitation ratio (C, b) of ␣ 1B /␣ 2 ␦/␤ 3 channels. In B, a, and C, a, current amplitude in each oocyte was normalized to the amplitude measured before the application of ACh, and the normalized data were averaged. Number of oocytes in each group is shown in the respective bars. Statistical analysis included the comparison of each test group (ϩstg or ϩc␤ARK) to same treatment (either a low or high dose of ACh) in the control group. n.s., non significant; *, p Ͻ 0.05; **, p Ͻ 0.01. kinetics and voltage dependence of activation and inactivation, voltage-dependent facilitation, and current amplitude (in the case of GPCR-induced modulation).
2) The effects of stargazin could be replicated by G␤␥ scavenging proteins m-c␤ARK or m-phosducin, coexpression of which also inhibited the G␤␥ modulation with even greater efficiency.
3) The effects of stargazin could be overcome by coexpression of higher doses of G␤␥. 4) Effects of stargazin were occluded by m-c␤ARK, suggesting a common mechanism. If stargazin were to directly affect Ca V 2.2 gating, as expected from a genuine channel subunit, it should still do so after complete G␤␥ scavenging, which is not the case. Two crucial lines of evidence suggest that stargazin acts as a G␤␥-scavenger. First, stargazin inhibited the G␤␥-dependent basal and evoked activity of another G␤␥ effector, the GIRK channel. Second, purified GST-fused C-terminal part of stargazin bound G␤␥ but not G␣ i . This interaction was weaker than the previously described high-affinity interactions between stargazin and PSD-95 (50), or between GST-G␣ i and G␤␥ (e.g. Ref. 25). Together, these results strongly support the idea that stargazin is a G␤␥-binding protein and that its interaction with G␤␥ is of relatively low affinity, in line with the observations obtained in the functional assays.
The proposed mechanism of modulation of Ca V 2.2 by G␤␥ sequestration may explain discrepancies among previous studies regarding the effects of stargazin and other ␥ proteins in heterologous expression systems (1, 5, 11, 14 -16). Some studies reported no effects, whereas others described altered channel gating. In our study, all previously reported effects of star-gazin (acceleration of inactivation and hyperpolarizing shift in activation and inactivation voltage) were seen consistently, time after time, if a moderate amount of G␤␥ was also coexpressed. Similar to some of the previous reports (15,16), we also observed that expression of stargazin without G␤␥ had little or no effects on most parameters of the channels except basal VDF. VDF appears to be the most sensitive reporter of endogenous G␤␥, but effects of stargazin on VDF have not been studied previously. It appears, therefore, that the endogenous levels of G␤␥ are insufficient to affect most G␤␥-modulated parameters of channel gating. Moreover, the extent of G␤␥ or starga-  7-9). Statistical analysis shown in the graphs represents the following: for empty bars, each test group was compared with control group expressing channel alone. Statistical significance is represented by #. For striped bars, each test group was compared with the group with channel, and the respective dose of G␤␥ was coexpressed. Statistical significance is represented by an asterisk. #, *, p Ͻ 0.05; **, p Ͻ 0.01 by one-way analysis of variance, post hoc Student Newman-Keuls. zin modulation of gating was sometimes modified in the presence of Ca V ␤. Therefore, we propose that the modulation of Ca V 2 channels by stargazin reported in some studies was the result of sequestration of endogenous, Ca V 2-associated G␤␥ by the coexpressed stargazin. Divergent channel subunit compositions, together with varying endogenous G␤␥ levels in the diverse expression systems, can account for the discrepancies regarding the presence of stargazin-induced changes in channel gating.
Calcium Channel ␤ 3 Subunit and Stargazin: Role in Modulation by G␤␥-Ca V ␤ subunits are important players in G protein-mediated inhibition of Ca V 2 channels; however, whether they enhance or weaken this modulation remains debatable (8,28,33,36,37,52,53). Expression of Ca V ␤ 3 often enhanced the efficiency of low doses of G␤␥ but sometimes reduced the maximal effect of high G␤␥ doses. These results support the notion that of Ca V ␤ is essential for G␤␥ modulation of Ca V 2.2 (33,52) but also corroborate previous reports (36,37) of reduction of the maximal effect of G␤␥. Ca V ␤ 3 also often reduced the potency of the action of stargazin on Ca V 2.2 gating, but, in general, the effects of stargazin were qualitatively similar in the absence or presence of Ca V ␤ 3 . Although it is unclear how exactly Ca V ␤ alters the actions of stargazin, it may stem from the G␤␥-dependence of the effect of stargazin, as G␤␥ modulation of ␣ 1B is altered by Ca V ␤.
Is Stargazin a Subunit of Neuronal Ca 2ϩ Channels?-A major debate in recent literature regarding stargazin is whether it is a ␥ subunit of neuronal calcium channels as was proposed initially (1). Lack of strong supporting data and the crucial role stargazin (and other ␥ proteins) play in AMPA receptor function (2) had weakened the notion that stargazin is a subunit of calcium channels but have not ruled it out. We find that the modulatory effects of stargazin on Ca V 2.2 gating are indirect and not exclusive to this channel. On the other hand, the well characterized reduction by stargazin of the protein level of Ca V 2.2, which involves the unfolded protein response (11), was still observed in the presence of excess G␤␥. G␤␥ did not affect the surface expression of Ca V 2.2 or GIRK (supplemental Fig. S5). Thus, the decrease in surface expression caused by stargazin does not appear to depend on G␤␥. However, stargazin also reduced the surface expression of GIRK and even of the IRK1 channel, which is not functionally modulated by G␤␥ (54). It seems that neither the effect of stargazin on protein expression, nor its effect on channel function, are unique to the Ca V 2.2.
In conclusion, our results do not support the hypothesis that stargazin is a neuronal calcium channel subunit. Taken together with previous studies, our results suggest that stargazin (besides being an important subunit of AMPA receptors) may have additional roles in cellular mechanisms, including regulation of protein synthesis and surface expression as well as functional modulation of G␤␥ effectors. For instance, in certain neurons Ca V 2 channels are subject to tonic G protein inhibition by ambient free G␤␥ (e.g. Ref. 32). Stargazin may act to inhibit this tonic modulation by G␤␥, making calcium channels more available to activation (and GIRK channels, if present, less basally active), thus contributing to the delicate balance of neuronal excitability.