Gβγ-dependent and Gβγ-independent Basal Activity of G Protein-activated K+ Channels*

Cardiac and neuronal G protein-activated K+ channels (GIRK; Kir3) open following the binding of Gβγ subunits, released from Gi/o proteins activated by neurotransmitters. GIRKs also possess basal activity contributing to the resting potential in neurons. It appears to depend largely on free Gβγ, but a Gβγ-independent component has also been envisaged. We investigated Gβγ dependence of the basal GIRK activity (AGIRK,basal) quantitatively, by titrated expression of Gβγ scavengers, in Xenopus oocytes expressing GIRK1/2 channels and muscarinic m2 receptors. The widely used Gβγ scavenger, myristoylated C terminus of β-adrenergic kinase (m-cβARK), reduced AGIRK,basal by 70–80% and eliminated the acetylcholine-evoked current (IACh). However, we found that m-cβARK directly binds to GIRK, complicating the interpretation of physiological data. Among several newly constructed Gβγ scavengers, phosducin with an added myristoylation signal (m-phosducin) was most efficient in reducing GIRK currents. m-phosducin relocated to the membrane fraction and did not bind GIRK. Titrated expression of m-phosducin caused a reduction of AGIRK,basal by up to 90%. Expression of GIRK was accompanied by an increase in the level of Gβγ and Gα in the plasma membrane, supporting the existence of preformed complexes of GIRK with G protein subunits. Increased expression of Gβγ and its constitutive association with GIRK may underlie the excessively high AGIRK,basal observed at high expression levels of GIRK. Only 10–15% of AGIRK,basal persisted upon expression of both m-phosducin and cβARK. These results demonstrate that a major part of Ibasal is Gβγ-dependent at all levels of channel expression, and only a small fraction (<10%) may be Gβγ-independent.

G protein-activated, inwardly rectifying K ϩ channels (GIRK, Kir3) 1 mediate postsynaptic inhibitory effects of various neu-rotransmitters in the brain and atrium via seven-helix, G protein-coupled receptors (GPCRs) linked to pertussis toxin-sensitive G proteins of the G i/o family. Opening of the channels is the result of a direct binding of G␤␥ subunits released from the G␣ i/o ␤␥ heterotrimers (1)(2)(3)(4). The channel can also be activated by cytosolic Na ϩ and membranal phosphatidylinositol 4,5bisphosphate (PIP 2 ); the latter is essential for proper GIRK gating by both Na ϩ and G␤␥ (3,5,6).
Whereas the physiological role of neurotransmitter-induced GIRK activity is well established, the basal activity of these channels (A GIRK,basal ) is often regarded as negligible and physiologically unimportant. This feature distinguishes GIRK from many other K ϩ channels of the Kir family, such as Kir1 and Kir2, which show high intrinsic activity under physiological conditions and are often referred to as "constitutively active." Low A GIRK,basal is supposed to ensure high signal-to-noise ratio for GIRK-related neurotransmitter signaling and to minimize participation of GIRK in resting membrane K ϩ conductance (see Ref. 2). However, some classical and many recent studies challenge this concept. In sinoatrial node cells, GIRK channels may contribute a major part of basal K ϩ conductance (7). Recent studies in intact neurons indicate that basal activity of GIRK may substantially contribute to resting K ϩ conductance, shunting the excitatory postsynaptic potentials and reducing the responses to glutamate (8). Substantial A GIRK,basal has been reported in hippocampal pyramidal cells (9,10) and in rat locus coeruleus slices (11). In murine locus coeruleus, where GIRKs mediate most of the inhibitory effects of opioids, knock-out of GIRK subunit genes depolarizes the cell's resting potential by as much as 20 mV (8). In many cases, it was not clear whether the GIRK activity observed in resting cells under a variety of experimental conditions was a truly agonist-independent one or depended on the presence of "ambient" neurotransmitters or local hormones. For instance, in hippocampal pyramidal cells, one report assigned to ambient adenosine a major role in sustaining most of A GIRK,basal (9), whereas another one contended that A GIRK,basal was independent of adenosine or other neurotransmitters (10). In any case, it appears that GIRK plays a substantial role in determining the resting membrane potential at least in some neurons and in sinoatrial pacemaker cells. It is therefore conceivable that not only activation of GIRK by G i/o coupled neurotransmitters, but also its inhibition by neurotransmitters that activate G␣ q (12)(13)(14)(15)(16) may serve as an important mechanism of regulation of neuronal excitability. These considerations urge for a better understanding of the mechanisms of regulation of A GIRK,basal .
An important contributor to A GIRK,basal is ambient free G␤␥. This has been shown in heterologous expression systems; co-expression of G␣ subunits and a myristoylated C-terminal segment of ␤-adrenergic receptor kinase (m-c␤ARK), used in the past as standard G␤␥ scavengers in GIRK studies (17)(18)(19)(20)(21)(22), has been reported to reduce A GIRK,basal to different extents, between 50 and 90%. It is not clear whether the remaining activity is G␤␥-independent or simply reflects insufficient G␤␥ sequestration, leaving open the question of whether a G␤␥independent A GIRK,basal exists. Such G␤␥-independent basal activity could be due to a direct activation by cytosolic free Na ϩ (23)(24)(25) and/or reflect an intrinsic G␤␥-and Na ϩ -independent channel activity. The variability of the reported results regarding the extent of inhibition of A GIRK,basal by G␤␥ scavengers calls for a better quantitative assessment of the G␤␥-dependent component. Furthermore, the use of G␣ as a G␤␥ scavenger is problematic. Coexpressed G␣ i subunits may affect the level of channel expression (20); they interact with GIRK itself and have been hypothesized to affect the gating directly (20, 26 -28). This greatly complicates the interpretation of G␤␥-scavenging effects of G␣ subunits. It is not known whether m-c␤ARK, widely used to study modulation of ion channels by G␤␥ (19, 20, 29 -31), directly interacts with GIRKs or affects their expression. Therefore, further study is needed to reliably estimate the G␤␥-dependent part of A GIRK,basal .
An intriguing phenomenon related to A GIRK,basal has been discovered in Xenopus oocytes (20). At low expression levels (densities) of GIRK, A GIRK,basal is low, and activation by agonist and by purified G␤␥ (in whole cells and in excised patches, respectively) is strong, as expected. However, when more GIRK channels are expressed, their basal activity becomes excessively large at the expense of agonist-or G␤␥-evoked activity. Coexpression of a G␣ i subunit restores the normal gating pattern, with low A GIRK,basal and high agonist-or G␤␥-evoked activity. We proposed that A GIRK,basal is controlled by a number of factors, most prominently G␤␥ and G␣, and that the balance between these factors collapses when the channel is expressed at high levels in Xenopus oocytes, partly because of a lack of G␣ (20,28). The excessively high A GIRK,basal at high levels of GIRK expression remained incompletely understood. Theoretically, it seemed possible that at high densities GIRK behaves as a constitutively active channel.
To better understand the fundamental mechanisms of A GIRK,basal , we have initiated a systematic study of G␤␥ dependence of A GIRK,basal in Xenopus oocytes at different levels of GIRK expression. This heterologous expression system is superior to others for quantitative studies, since it allows both a controlled expression of proteins in very wide ranges and an accurate measurement of expressed proteins in the whole cell and in the plasma membrane (PM). Titrated expression of m-c␤ARK and a novel membrane-targeted G␤␥ scavenger, myristoylated phosducin (m-phosducin), revealed that at least 90% of A GIRK,basal is G␤␥-dependent at all levels of expression of GIRK. G␤␥-independent activity, if any, constitutes only a small fraction of A GIRK,basal . The excessively high basal activity observed at high levels of channel expression may be due to a constitutive association of GIRK with G␤␥, in the absence of sufficient G␣ needed to preserve low A GIRK,basal .

EXPERIMENTAL PROCEDURES
cDNA Constructs and RNA-The coding sequences of all cDNAs used in this study for preparation of RNA were inserted into high expression oocyte vectors containing 5Ј-and 3Ј-untranslated sequences of Xenopus ␤-globin: pGEMHE or its derivative pGEMHJ or pBS-MXT, as described in our previous publications (12,32), unless stated otherwise. All PCR-derived cDNAs were sequenced in full at the Tel Aviv University sequencing facility. The original cDNAs of human m2R, human 5HT-1A receptor, mouse GIRK2, rat m-c␤ARK (aa 452-689 end ) (12, 32), human ␤ARK (GB X61157), and bovine phosducin (GB M33529) were kindly provided by E. Peralta, P. Hartig, P. Kofuji, E. Reuveny, R. J. Lefkowitz, and M. Lohse, respectively.
PCR-derived full-length coding sequence of wild type phosducin was inserted into the EcoRI restriction site of pGEMHJ, creating the template for RNA synthesis and expression of wild-type, nonmyristoylated phosducin (n-phosducin). To create a membrane-attached derivative of phosducin, we first constructed a new vector, pGEMHE-myr, on the basis of the pGEMHE vector (33) modified by E. Reuveny, using standard PCR techniques. A DNA sequence, atggggagtagcaagagcaagcctaaggaccccagccagcgccgg, was inserted between SmaI and EcoRI restriction sites of the pGEM-HE vector. On the protein level, this adds the first 15 N-terminal aa of Src, MGSSKSKPKDPSQRR (34), followed by Glu-Phe (encoded by the nucleotide sequence of the EcoRI site), to any proteins whose coding sequence is inserted between EcoRI and any one of the following restriction sites of pGEMHE polylinker. PCR product corresponding to the entire coding sequence of bovine phosducin, flanked by EcoRI and HindIII, was inserted into the respective restriction sites of pGEMHE-myr.
cDNA for nonmyristoylated C terminus of human ␤ARK1, n-c␤ARK, was created by inserting the cDNA sequence encoding aa 460 -689 end of ␤ARK, preceded by a methionine, flanked by EcoRI and HindIII restriction sites, into the pGEMHJ vector. The comparison of physiological effects of m-c␤ARK (rat) and n-c␤ARK (human) is justified by high homology between these proteins (97% identity). To create the fusion proteins m2R-c␤ARK, 5HT1A-c␤ARK, and 5HT1A-1TMD-c␤ARK, PCR-derived coding sequences of m2R, 5HT1A receptor (5HT1AR), and aa 1-67 of 5HT1AR, respectively, flanked by EcoRI restriction sites, were inserted in frame just prior to the n-c␤ARK sequence made as described above. Point mutations in m-c␤ARK were done using the QuikChange site-directed mutagenesis kit (Stratagene).
RNA was synthesized in vitro using a protocol that ensures incorporation of the GTP cap mainly into the 5Ј portion of the RNA (35) rather than along the whole length as with commercially available kits. Accordingly, the resulting RNAs are usually more efficient than those obtained by other standard protocols. 2 Unless indicated otherwise, RNAs were injected into the oocytes in the following amounts: m2R, 0.5 ng/oocyte; n-c␤ARK, m2R-c␤ARK, 5HTA1-c␤ARK, 5HTA1-1TMD-c␤ARK, and m-c␤ARK R587Q , 5 ng/oocyte; n-phosducin, 10 ng/oocyte.
Interaction between GST fusion proteins and phosducin and c␤ARK was studied as described (21). Briefly, GST-CT-GIRK1 fusion protein was purified, and its interaction with [ 35 S]methionine-labeled proteins (the various forms of c␤ARK, phosducin, and G␤ 1 ␥ 2 ) synthesized in rabbit reticulocyte lysate (Promega) was monitored using a standard pull-down procedure. GST-GIRK1-CT or GST-GIRK1-NT (5-10 g) was incubated for 1 h with 5 l of reticulocyte lysate containing the designated [ 35 S]methionine-labeled proteins in 300 l of a high K ϩ buffer (150 mM KCl, 50 mM Tris, 5 mM MgCl 2 , 1 mM EDTA, pH 7.0) with the addition of 0.5% CHAPS. Then 30 l of glutathione-Sepharose beads (Amersham Biosciences) were added, and the mixture was incubated for 30 min at 4°C and washed in 1 ml of the same buffer (once with 0.5% and twice with 0.1% CHAPS). Bound proteins were eluted with 30 l of 15 mM reduced glutathione and analyzed on 12% SDS-polyacrylamide gels. The labeled products were imaged and quantified by autoradiography using a PhosphorImager (Amersham Biosciences).
Western Blot Analysis-The oocytes were homogenized on ice in homogenization buffer (20 mM Tris, pH 7.4, 5 mM EGTA, 5 mM EDTA, and 100 mM NaCl) containing the Roche Applied Science protease inhibitor mixture. Debris was removed by centrifugation twice at 1000 ϫ g for 15 min at 4°C. Total cellular membrane fraction from 5-17 oocytes was obtained by 1 h of ultracentrifugation at 100,000 ϫ g. For Western blots, protein samples were separated on SDS-12% polyacrylamide gels. Antibodies against c␤ARK, phosducin, and G␤ common (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) were used. Visualization of protein bands was performed using ECL reagents obtained from Pierce. The intensity of labeling was quantified using TINA software (Raytest, Straubenhardt, Germany).
Confocal Imaging of GIRK, G␣, and G␤ in Plasma Membrane-The levels of expressed proteins in PM were measured as described (20,37). Briefly, large oocyte membrane patches were attached to coverslips with their cytoplasmic leaflet facing the external solution; fixed; stained by 0.75 ng/l of a specific GIRK1 antibody (Alomone Laboratories, Jerusalem), 0.5 ng/l of a common G␤ antibody (Santa Cruz Biotechnology), or 2.5 ng/l of a common G␣ antibody (Calbiochem); and visualized using a Cy3-conjugated rabbit IgG (Jackson Immunoresearch Laboratories) using a Zeiss LSM 410 or a Leica TCS SP2 confocal microscope. Intensity of labeling (OD units) was measured with TINA software. In control experiments (Fig. 8, A and B), to estimate the specificity of labeling by the G␣ and G␤ antibodies, purified G␣ i1 -GDP or purified G␤ 1 ␥ 2 , respectively, was added in 20-fold excess (w/v) over the antibodies. The purified G␣ i1 and G␤ 1 ␥ 2 were a generous gift from C. W. Dessauer (University of Texas, Houston).
Data Presentation and Statistics-Data are presented as mean Ϯ S.E. Comparison between two groups of treatment was done using the two-tailed Student's t test. Comparisons among several groups of data have been performed using one-way analysis of variance followed by Dunnet's test. The level of statistical significance is indicated in the figures as follows: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.

RESULTS
m-c␤ARK Reduces I basal , but It Also Binds to the C Terminus of the GIRK1 Subunit-To separate the G␤␥-dependent component of A GIRK,basal , we have initially utilized the widely used G␤␥ scavenger, m-c␤ARK. GIRK activation by G␤␥ is membrane-delimited (1), so that only PM-attached G␤␥ can activate the channel. Targeting of m-c␤ARK to the membranes is provided by the myristoylation signal (the first 15 aa of Src added to the N terminus of m-c␤ARK) (29,34). For comparison, we have also constructed and tested a nonmyristoylated c␤ARK (n-c␤ARK), which should not be targeted to the membrane and thus is supposed to be less efficient than m-c␤ARK in inhibiting GIRK. As shown in Fig. 1, both m-c␤ARK and n-c␤ARK were well expressed in reticulocyte lysate (Fig. 1A) and in oocytes (Fig. 1B). m-c␤ARK showed two characteristic bands on SDS gels (Fig. 1, A and B). The lower band probably corresponded to the presence of a fraction of protein that remained nonmyristoylated, since only the upper band was present in the total membrane fraction (Fig. 1B, bottom). Only a small proportion of n-c␤ARK was found in the membrane fraction. Myristoyla-tion of m-c␤ARK appeared to be more efficient in reticulocyte lysate than in the oocytes (Fig. 1, compare A and B). m-c␤ARK ran on the gel shown in this figure below the 35-kDa molecular size marker, but its actual molecular size was closer to 35 kDa, as deduced from its comparison with G␤ 1 (ϳ37 kDa) and phosducin (ϳ33 kDa; see Figs. 3 and 4). In summary, in Xenopus oocytes, m-c␤ARK is synthesized, myristoylated, and preferentially targeted to the membranes, as expected.
Next, we examined the effects of m-c␤ARK and additional G␤␥ scavengers based on c-␤ARK on GIRK channels composed of subunits GIRK1 and GIRK2 (GIRK1/2). Oocytes were injected with either 0.2 ng/oocyte of RNA of each channel subunit (low/intermediate density of GIRK) or 1 ng/oocyte (high density of GIRK) (20) and 0.5 ng of RNA of the muscarinic m2 receptor, m2R. Fig. 2A shows typical recordings of K ϩ currents without the coexpression of m-␤ARK ("control") or with m-c␤ARK (5 ng of RNA/oocyte). In these whole-cell records, A GIRK,basal is mirrored in the basal GIRK current, I basal (i.e. the inward current that develops when the high Na ϩ extracellular solution ND96 (96 mM Na ϩ , 2 mM K ϩ ) is exchanged to a high K ϩ solution (hK) with either 96 mM K ϩ ( Fig. 2A) or 24 mM K ϩ (not shown). The agonist (acetylcholine; ACh)-induced current, I ACh , was evoked by 10 M ACh. The total GIRK current, I total , is I basal ϩ I ACh .
Oocytes that expressed c␤ARK showed greatly reduced I basal and I ACh . The inhibitory effect of m-c␤ARK was qualitatively and quantitatively similar in both 96 and 24 mM K ϩ solutions (data not shown).
We examined the dose dependence of the effect of m-c␤ARK, titrating its expression level by injecting various amounts of RNA (RNA of GIRK subunits was 0.2 ng/oocyte in these experiments). As shown in Fig. 2B, the effect of m-c␤ARK on I basal was dose-dependent and did not fully saturate at the highest dose of RNA used, 10 ng/oocyte. In contrast, I ACh was completely inhibited at 5 ng of RNA/oocyte of m-c␤ARK. The data were fitted to a Michaelis-Menten-like equation, assuming that, at saturating doses, m-c␤ARK would fully block the GIRK currents. The fit yielded values of ED 50 (effective doses of RNA that cause 50% inhibition) of 3.04 and 0.1 ng of RNA/oocyte for I basal and I ACh , respectively. The estimate of ED 50 for I ACh was intrinsically less accurate than for I basal , due to very strong inhibition of I ACh observed already at 2 ng of RNA/oocyte. Nevertheless, it reflects a much higher sensitivity of the evoked current to m-c␤ARK than of the basal one. Unfortunately, at 10 ng/oocyte, the expressed m-c␤ARK appeared to be toxic to oocytes, which often did not survive the 3-4-day incubation period. Therefore, in the following experiments, 5 ng/oocyte of m-c␤ARK RNA were routinely injected into the oocytes. We note that, in our hands, 5 or 10 ng of RNA/oocyte usually yielded very high levels of expressed proteins, because RNAs used in this study were designed to give maximal protein yield upon expression in the oocytes (see "Experimental Procedures").
To make sure that the decrease in GIRK currents by m-c␤ARK is not linked to changes in channel expression, we monitored the relative amounts of expressed GIRK channels in the PM using a confocal microscope-assisted immunoimaging method (Fig. 2C). The channels were visualized in very large plasma membrane patches (usually several thousand m 2 ), adhered to a coverslip, with their cytoplasmic surface facing the external medium (37). The patches were extensively washed and nominally free from endoplasmic reticulum. 3 As summarized in Fig. 2C, b, coexpression of m-c␤ARK did not affect the amount of channel protein in the PM.
Myristoylated m-c␤ARK seems to be suitable for the purpose of studying the G␤␥ dependence of A GIRK,basal , since it 3 N. Dascal and E. Artzy, unpublished observations. does not block the direct Na ϩ -induced activation (19). However, "side effects" are possible. One potential pitfall is that the G␤␥-binding segment is within a plekstrin homology domain, which also binds PIP 2 (38); thus, heavily overexpressed c␤ARK may in principle reduce the amount of PIP 2 in the PM. Second, it is not known whether the myristoyl moiety affects the function of GIRK. To control for these possibilities, we constructed several membrane-targeted G␤␥ scavengers based on c-␤ARK. First, we made a point mutation, R587Q, in m-c␤ARK, which greatly impairs the ability of c␤ARK to bind G␤␥, but not PIP 2 (39). Expression of this construct in the oocytes did not have any significant effect on GIRK currents (Fig. 2D), suggesting that PIP 2 chelation is not involved in the effect of m-c␤ARK. To check for possible "side effects" of the myristoyl moiety, we expressed the nonmyristoylated n-c␤ARK and found that it did not significantly affect GIRK currents at 5 ng of RNA/oocyte (Fig. 2D). Although n-c␤ARK (aa 460 -689 of c␤ARK) is 8 aa shorter than m-c␤ARK (aa 452-689), both constructs contain the G␤␥-binding domain (aa 643-670) as well as the full plekstrin homology domain (aa 561-655) (39). Therefore, the difference in length probably does not determine the difference in the effect on GIRK currents. Next, we constructed fusion proteins in which c-␤ARK was connected to C-terminal ends of several membrane proteins, which were expected to anchor these tandems to the PM, instead of the myristoyl moiety. The proteins chosen were m2R, the serotonin receptor 5HT1AR, and the first transmembrane domain of 5HT1AR, 5HT-1TMD. The first two constructs did not inhibit GIRK currents (Fig. 2D) after 3 days of incubation. Mild inhibitory effects of m2R-c␤ARK and 5HT1A-c␤ARK were seen after long periods of oocyte incubation, 4 -5 days (data not shown), suggesting that these very large fusion proteins may be poorly expressed in the oocytes. However, 5HT-1TMD-c␤ARK significantly reduced I total though less than m-c␤ARK (by ϳ55%, versus ϳ72% inhibition by c␤ARK; Fig. 2D). This result suggests that the presence of the myristoyl moiety is not necessary for c-␤ARK to inhibit GIRK. Here and in the following figures, the statistical significance is denoted as follows: **, p Ͻ 0.01; ***, p Յ 0.001. The brackets below the asterisks show groups subjected to pairwise comparisons. The asterisks above the bars not denoted by brackets indicate statistically significant differences compared with the control group only.
Next, we compared the effect of m-c␤ARK on GIRK channels expressed at two different densities. Coexpression of m-c␤ARK (5 ng RNA/oocyte) was less effective in reducing I basal and I ACh when more channels were expressed (Fig. 2E). At high channel density (1 ng RNA/oocyte), I basal was decreased by ϳ70%, and I ACh was decreased by ϳ75%, in comparison with ϳ80 and ϳ95% decrease, respectively, at lower channel density, 0.2 ng of RNA/oocyte. The same held true for I total . These differences are not unexpected for a pure G␤␥ scavenger whose efficiency depends on the ratio of its own concentration to that of the G␤␥-binding effector (GIRK) under study, as well as on the cellular level of G␤␥ (see Fig. 8). Another possibility is that the scavenger interacts with the effector (GIRK in our case) and is itself sequestered by high doses of the channel, so that the efficiency of G␤␥ sequestration is reduced. Indeed, GIRK and ␤ARK co-immunoprecipitate, along with a number of other components of a multiprotein signaling complex, from atrial cells (40).
The possibility of a direct interaction between c␤ARK and GIRK was probed using the standard pull-down methodology. We used a purified GST fusion protein containing the fulllength cytosolic C terminus of GIRK1, GST-GIRK1-CT (aa 183-501), which was previously shown to bind G␤␥ (21,27,41). Unexpectedly, we found that GST-GIRK1-CT bound [ 35 S]methionine-labeled m-c␤ARK and m-c␤ARK R587Q synthesized in vitro in reticulocyte lysate almost as well as it bound G␤␥ (Figs. 3 and 4B). n-c␤ARK also bound to GST-GIRK1-CT, but distinctly more weakly than m-c␤ARK. At present, we do not know whether this difference was due to the absence of myristoylation or of the first 8 aa in n-c␤ARK. Because m-c␤ARK R587Q , which binds GIRK1 but not G␤␥, does not affect GIRK currents (Fig. 2), it is plausible that most of the inhibitory effect of m-c␤ARK is independent of its binding to GIRK and represents a genuine G␤␥ scavenging effect. However, the direct interaction between m-c␤ARK and GIRK1 is bound to complicate the interpretation of the results of this and previous studies.
A Novel G␤␥ Scavenger, m-phosducin, Reveals That Most A GIRK,basal Is G␤␥-dependent at All Channel Densities-In a search for a G␤␥ scavenger that does not interact with GIRK, we constructed a myristoylated first intracellular loop of the N-type voltage-dependent Ca 2ϩ channel, which is known to bind G␤␥ (42). However, expression of this construct in the oocytes led to a rather weak inhibition of GIRK currents (data not shown). Next, we turned to phosducin, a soluble cytosolic G␤␥-binding protein originally identified in retinal photoreceptor cells, which has already been used as a G␤␥ scavenger in heterologous expression systems (43). Phosducin also binds G␣, but this interaction is of much lower affinity than that with G␤␥ (44). In this paper, we designate the original cytosolic phosducin as n-phosducin to distinguish it from the myristoylated construct made by us.
Coexpression of n-phosducin (10 ng of RNA/oocyte) reduced I basal by ϳ70% and I ACh by only ϳ42% (see Fig. 5B). To increase the likelihood that phosducin will access G␤␥ in the PM, we constructed a DNA for myristoylated m-phosducin. The resulting protein has an added 15-aa myristoylation signal, the same as in m-c␤ARK, at the beginning of its N terminus. n-Phosducin expressed in the oocytes was preferentially associated with the cytosolic fraction, whereas m-phosducin was enriched in the membrane fraction (Fig. 4A). Pull-down experiments showed that neither n-phosducin nor m-phosducin bound GST-GIRK1-CT (Fig. 4B). Since the N terminus of GIRK1 also binds G␤␥ (27), we examined whether the GST-fused N terminus of GIRK1 (GST-GIRK1-NT) binds c␤ARK and phosducin. Fig. 4D shows that GST-GIRK1-NT bound G␤␥ and, to a lesser extent, m-c␤ARK, but neither n-phosducin nor m-phosducin interacted with the N terminus of GIRK1. Coexpression of m-phosducin did not affect the amount of expressed GIRK channels in the PM of Xenopus oocytes (Fig. 4D). Thus, m-phosducin has the desired properties of a PM-enriched G␤␥ scavenger that does not interact with GIRK.
Coexpression of m-phosducin with GIRK at 10 ng RNA/ oocyte efficiently decreased GIRK currents (Fig. 5A). The effect of phosducin was dose-dependent (see below) and did not saturate at the maximal RNA dose used, 10 ng/oocyte (this dose was not toxic and was routinely used in all experiments, unless stated otherwise). We compared the effects of m-phosducin (10 ng of RNA/oocyte) on GIRK currents at low/intermediate (0.2 ng/oocyte) and high (1 ng of RNA/oocyte) channel densities (Fig.  5, B and C). m-Phosducin reduced I basal to about the same extent, by 85-90%, at both channel densities. However, it was less effective than m-c␤ARK in reducing peak I ACh in comparison with m-c␤ARK. The inhibition of I ACh by m-phosducin was stronger at lower channel densities, suggesting competition among G␣, GIRK, and m-phosducin for available G␤␥. n-Phosducin was significantly less effective than m-phosducin in reducing GIRK currents. To examine whether the 10 -15% I basal remaining in the presence of m-phosducin are G␤␥-dependent, we coexpressed, in addition, m-c␤ARK (5 ng/oocyte). This treatment did not further reduce I basal , supporting the possibility of the presence of measurable G␤␥-independent basal activity in GIRK channels (Fig. 5B). However, the main outcome of these experiments is that up to 90% of A GIRK,basal is G␤␥-dependent.
I ACh of GIRK channels expressed in Xenopus oocytes slowly decays in the continuous presence of ACh, reaching a steady state after 3-5 min (45,46). We noticed that, in the presence of m-phosducin, the decay was strongly accelerated. The steady state level of I ACh was greatly diminished compared with control (see Fig. 5A). The kinetics of decay of I ACh were specifically studied in a separate experiment (n ϭ 6 oocytes). Fig. 6A shows representative records from two oocytes, with and without coexpressed m-phosducin, after scaling the peak I ACh currents to facilitate the comparison of decay kinetics. Fig. 6B summarizes the effects of m-phosducin on the amplitudes of I basal and I ACh measured at the peak and 1.5 min after the addition of ACh. On the average, within 1.5 min, I ACh decayed to a mere 2.7 Ϯ 1. presence of m-phosducin; thus, in the latter case, by 1.5 min a steady state was reached (Fig. 6C). No change in speed of activation could be detected, but the speed limit of the perfusion system used (ϳ1 s) did not enable a conclusive measurement. The effect of phosducin is readily understood in general kinetic terms and can be quantitatively described by a kinetic model (not shown), assuming a simple competition among GIRK, G␣ GDP , and m-phosducin for available G␤␥, as follows. The extent of inhibition of I basal is determined by the concentrations of the proteins involved and by the actual K D values, since the system has sufficient time to reach equilibrium. At equilibrium, a certain percentage of G␤␥ is bound to G␣ within heterotrimers and is liberated from G␣ upon activation by agonist. This leads to an initial increase in free G␤␥ concentration, followed by re-equilibration and capturing of much of the G␤␥ by the scavenger. The initial large peak of I ACh can be explained kinetically by assuming that G␤␥ released upon transmitter activation first gets access to the channel and only later can be captured by the scavenger. This supports the view that a G␣␤␥ complex is closely associated with GIRK (see "Discussion"), although a similar property can also be conferred by a kinetic characteristic (a "slow" scavenger with a low on-rate).
Measurements of the kinetics of I ACh in the presence of coexpressed m-c␤ARK have been attempted but proved less reliable, since m-c␤ARK reduced peak I ACh more potently than m-phosducin. However, in those oocytes where I ACh could be reliably measured in the presence of m-c␤ARK, I ACh did not show any fast decay (Fig. 6D; representative of at least 20 cells).
The dependence of the effect of m-phosducin on level of its expression was studied in two separate experiments. The results are shown in Fig. 7, where they are compared with the dose dependence of m-c␤ARK from Fig. 2 (gray lines). m-phosducin reduced I basal similarly to m-c␤ARK, with an ED 50 of 2.1 ng of RNA/oocyte. Remarkably, the effect of m-phosducin on I ACh was strikingly different from that of c␤ARK. The dose dependence of inhibition of peak I ACh (open circles) was very similar to that for I basal (closed circles), practically falling on the same Michaelis-Menten curve. Because ED 50 is conceptually similar to dissociation constant, K D , which is a measure of the equilibrium state, we have also measured the dose dependence of inhibition of steady state I ACh (Fig. 7, triangles). Again, it was very similar to that of I basal and peak I ACh .
Expression of GIRK Is Accompanied by an Increase in the Amount of G␤␥ in the PM-We have previously shown that the amount of endogenous G␤␥ measured by Western blot methodology in total cellular membrane fraction of Xenopus oocytes does not change appreciably when GIRK is overexpressed (20). How can A GIRK,basal remain 90% G␤␥-dependent when more and more channels are expressed if free G␤␥ concentration stays constant and there is less and less G␤␥ per channel? One possibility is that the expression of GIRK leads to an increase in the amount of G␤␥ in the PM. To explore this possibility, we monitored changes in GIRK1/2, G␤␥, and G␣ in large patches of PM, using the imaging method described above (see Fig. 2C). The proteins were visualized by a GIRK1 antibody, a common G␤ antibody that recognizes G␤ subunits 1-4 but not G␤ 5 , and a common G␣ antibody that recognizes all G␣ subunits. Fig. 8 demonstrates that the increase in the amount of GIRK in the membrane, brought about by injection of RNA, was accompanied by an increase in the amounts of endogenous G␣ and G␤ in the PM. Specificity of labeling was controlled for by incubating the PM patches with the antibodies in the presence of an excess of the respective antigens (purified G␣ i1 or G␤ 1 ␥ 2 ). With both G␣ and G␤␥, the background labeling remaining in the presence of the antigen was the same in native and GIRK-expressing oocytes, suggesting a complete block of the specific labeling by both antigens. In uninjected (native) oocytes, the endogenous G␣ label was clearly above the background (Fig.  8A), whereas the endogenous G␤␥ could not be clearly detected in this experiment (Fig. 8B). In the presence of coexpressed GIRK (2.5 ng of RNA/oocyte), both G␤ and G␣ rose significantly above the control level. After subtraction of the background fluorescence, net increase in G␤ label was from an undetectable level to ϳ25 OD units (p Ͻ 0.001), whereas the level of G␣ increased from ϳ25 to ϳ38 OD units (p Ͻ 0.05) (i.e. less than 2-fold). Although this rough comparison does not provide true quantification of the increase in G␣ versus G␤ levels, it appears to indicate that the relative increase in the level of G␤, caused by the expression of GIRK, was greater than that of G␣. Fig. 8C shows that the expression of the GIRK protein in the PM increased with an increase in the dose of injected RNA. The background labeling with the GIRK1 antibody in uninjected oocytes was very low, suggesting high specificity of the antibody and the expected absence of GIRK1 protein in native oocytes. Fig. 8D summarizes a separate series of experiments in which the levels of G␣ and G␤ in the PM were measured upon the expression of GIRK at two different levels. The levels of both G protein subunits were increased already when a low dose of GIRK RNA, 0.1 ng/oocyte, was injected, and a further increase was observed upon injection of a high dose of GIRK RNA, 2.5 ng.

Most of the Basal Activity of GIRK Is G␤␥-dependent at All
Levels of GIRK Expression-Basal activity of GIRK channels, A GIRK,basal , is observed in the absence of agonist and, in heterologous expression systems, of any coexpressed receptor (2). To understand the mechanisms underlying A GIRK,basal , we utilized Xenopus oocytes, unsurpassed for titrated expression of proteins for rigorous quantitative studies (which we term "expression pharmacology"). Of the several G␤␥ scavenger constructs used here, myristoylated derivatives of c␤ARK and phosducin proved most efficient. The two scavengers have similar high affinities of binding to G␤␥ in vitro: K D of 42 nM in a detergent solution for phosducin (47) and 32 nM in phospholipid vesicles for ␤ARK (48). Myristoylation ensured preferential incorporation of these proteins into membrane fractions and, in parallel, stronger inhibition of GIRK activity than by nonmyristoylated congeners. The strong reduction of GIRK currents by c␤ARK fused to the first transmembrane domain of the 5HT1A recep- tor suggests that the myristoyl moiety itself is not essential; rather, membrane anchoring appears to be the important factor that ensures efficient competition between a G␤␥ scavenger and GIRK for membrane-attached G␤␥. Unfortunately, we find that, in addition to sequestering G␤␥, m-c␤ARK directly binds to GIRK1. This interaction does not seem to interfere with the activation of the channel by G␤␥, because m-c␤ARK R587Q , which does not bind G␤␥ but still binds GIRK1, has no effect on GIRK currents. Despite this consideration, the binding of m-c␤ARK to GIRK1 is a serious disadvantage that perplexes the interpretation of the experimental results.
Both m-phosducin and m-c␤ARK reduced I basal in a dose-dependent manner. RNA titration experiments showed a slightly lower ED 50 for m-phosducin, possibly due to better protein expression or PM accumulation of m-phosducin. Titrated expression of these two potent G␤␥ scavengers, performed in a rigorous quantitative manner for the first time here, led us to the main empirical conclusion of this work; at least 90% of A GIRK,basal in intact oocytes is G␤␥-dependent at all densities of GIRK channels.
The source of G␤␥ that determines A GIRK,basal in the oocytes and other cells is not clear. It is widely believed that the G␤␥-dependent component of A GIRK,basal reflects the concentration of free G␤␥ in the membrane (2). The most obvious and best studied sources of free G␤␥ in the PM are the G␣␤␥ heterotrimers. The latter dissociate and supply free G␤␥ by several mechanisms: basal dissociation of G␣ GDP from G␤␥ in the absence of GDP-GTP exchange (49); basal GDP-GTP exchange that results in great reduction of the affinity of G␣ to G␤␥ (50); and GPCR-catalyzed GDP-GTP exchange that occurs as a result of a constitutive activity of the GPCR in the absence of an agonist (51) or is induced by an ambient agonist present in the extracellular medium (the latter seems unlikely in isolated, follicle-free, constantly perfused oocyte preparations). Estimation of the relative contribution of these processes to A GIRK,basal in oocytes and other cells will require further study.
GIRK May Be Constitutively Associated with G␤␥-The idea that many end effectors of GPCRs operate within multiprotein complexes that include the GPCRs, heterotrimeric G proteins, and effectors themselves has gained wide acceptance (52). It has been proposed that, under normal physiological conditions, GIRK is associated with a G␣ i/o ␤␥ complex, which guarantees both low A GIRK,basal and rapid and efficient channel activation by G␤␥ released upon an encounter with an agonist-bound GPCR (20, 26 -28, 40, 53). The observation that the levels of endogenous G␤␥ and G␣ in PM increase upon expression of GIRK (Fig. 8) provides new support to this hypothesis and further extends it, indicating that GIRK is preassociated with G␤␥ and G␣ in the endoplasmic reticulum or the Golgi complex before being relocated to the PM. Joint trafficking of

FIG. 8. Expression of GIRK is accompanied by an increase in the amount of endogenous G␣ and G␤ in the plasma membrane. In
A and B, the levels of G␣ (A) and G␤ (B) were measured by immunocytochemistry in large PM patches. The left panels show representative confocal microscope images of the membrane patches labeled with the respective antibodies. The right panels summarize the label intensities measured in patches from oocytes of the same donor frog (numbers of patches imaged are shown above the bars). The labeling remaining in the presence of excess antigens (black bars; G␣ i1 -GDP in A and G␤ 1 ␥ 2 in B) represents nonspecific background fluorescence. The levels of nonspecific signal were compared with total signal observed in the absence of the antigen (asterisks above black bars) separately for native and for GIRK-expressing oocytes. C, the level of expressed GIRK1 in PM patches increases as a function of the amount of injected RNA. The right panel summarizes data from three oocyte batches. Nonspecific labeling in the absence of expressed channel was very low; therefore, the level of GIRK1 was normalized to that observed in oocytes injected with 0.1 ng of RNA. D, summary of data from two or three oocyte batches showing the increase in the level of endogenous G␣ and G␤ upon coexpression of GIRK1/2 at two RNA doses, 0.1 and 2.5 ng/oocyte. Data were normalized to control levels of G␤ or G␣ detected in uninjected (Uninj.) oocytes. In these experiments, nonspecific labeling was not estimated, and the measured values represent total fluorescence signal (background ϩ specific). ion channels and associated proteins is a well documented phenomenon (e.g. see Ref. 54). The change in G␤␥ levels was not previously detected in total membrane fraction (20), probably because of the very low surface/volume ratio in the oocyte, where the PM constitutes at best a few percent of total cellular membranes.
The protein complex comprising GIRK and G protein subunits may also contain other components, such as protein kinase A, phosphatases, and ␤ARK (40). It is probable that the complex is a dynamic one and that the interactions within it are reversible, as indicated by the ability of exogenously expressed G␣ such as G␣ s and G␣ z to replace the endogenous G␣ i/o (32,55) and by the finite affinities of protein-protein interactions in vitro. On the other hand, there are indications that the complex may be rather stable: the components are co-precipitated by a single antibody (40); pull-down experiments show strong binding between GIRK and G␣, G␤␥, and ␤ARK (see Refs. 21, 27, and 41 and Fig. 3); and known affinities of interactions among some of the protein components are in the nanomolar range (G␤␥-GIRK, Ͻ10 nM; G␣-G␤␥, 0.2-2 nM) (discussed in Ref. 56).
A stable association of GIRK with G␤␥, together with the increase in G␤␥ levels in PM, provides the basis for an understanding of the excessively high A GIRK,basal at high GIRK densities. Although the level of endogenous G␣ also rises upon coexpression of GIRK, it appears insufficient to prevent the disproportional increase in A GIRK,basal , possibly because the increase in the level of G␤␥ in the PM is greater than that of G␣. Comparison of the relative increases in G␣ and G␤ caused by the expression of GIRK (Fig. 8, A and B) seems to support this possibility. However, at present this interpretation remains presumptive, because the immunoimaging method used here, although sensitive, does not provide quantitative information on the absolute levels of proteins measured with different antibodies. It is also possible that the observed rise in G␣ levels includes that of "irrelevant" G␣ (not G␣ i/o ) available in the oocytes, which cannot efficiently reduce A GIRK,basal and/or donate G␤␥ upon activation of m2R.
In summary, our observations so far conform to the proposal (20, 28) that A GIRK,basal is controlled by both G␤␥ and G␣. Expression of GIRK is accompanied by a rise in ambient G␤␥ available for activation of GIRK, which is not accompanied by a comparably strong rise in the level of G␣ i/o subunits, thus causing an excessively high A GIRK,basal . Taking together the previously accumulated evidence regarding the existence of GIRK-G␣␤␥ complexes and the new data presented here, we propose that GIRK is persistently, although reversibly, associated with G␣ GDP ␤␥ throughout its biosynthesis; G␣ is a nonobligatory part of the complex, whereas G␤␥ is bound persistently.
Using m-phosducin and m-c␤ARK to Assess the Existence of Different G␤␥-binding Sites in GIRK-It has been proposed, on the basis of mutational analysis of GIRK1 and GIRK4, that agonist-evoked activity is mediated by binding of G␤␥ to a low affinity site(s), and A GIRK,basal depends on the interaction of G␤␥ with another, high affinity, site(s) in the GIRK protein (57). Later works suggested that an intact putative site underlying A GIRK,basal is also crucial for agonist-evoked activity (58) and that a residue implicated in low affinity G␤␥ binding (57) may be involved in channel gating rather than G␤␥ binding (21,59). Thus, the issue of two types of G␤␥-binding site remains unresolved.
G␤␥ scavengers can be used to probe the existence in GIRK of G␤␥-binding sites with different affinities to G␤␥. The concentration of G␤␥ required to induce A GIRK,basal via the high affinity site of GIRK would be lower than that required to induce I ACh via the low affinity site. Consequently, less G␤␥ scavenger would be needed to suppress activation of GIRK by G␤␥ released following GPCR activation than by ambient G␤␥ that determines A GIRK,basal . Alas, the results with the two potent G␤␥ scavengers used here were not unequivocal. mphosducin and m-c␤ARK showed striking differences in the apparent potency of inhibition of peak I ACh and in ED 50 . The data obtained with c␤ARK appear to support the two-site model of He et al. (57); I ACh was inhibited by much lower doses of m-c␤ARK than I basal . However, m-phosducin, which showed an even greater apparent affinity in inhibiting A GIRK,basal than m-c␤ARK (Fig. 7), exhibited an identical apparent affinity for I ACh .
A clue for the understanding of these differences may be provided by invoking the fact that m-c␤ARK binds GIRK. Phosducin reduced I basal stronger than I ACh but accelerated the decay of I ACh . These effects of m-phosducin are compatible with those of a simple G␤␥ sink (see "Results"). In contrast, m-c␤ARK inhibited I ACh with a high apparent affinity but did not accelerate its decay. It can be shown that such properties can be conferred by very fast G␤␥ scavenging by c␤ARK, such that it captures G␤␥ released from the heterotrimeric G␣␤␥ before G␤␥ can reach the channel. Anchoring of c-␤ARK to the channel would enable such extra fast scavenging. More complicated schemes, which assume the formation of a high affinity triple c␤ARK-G␤␥-GIRK complex, or a competition between m-c␤ARK and G␤␥ for binding to GIRK may also explain the effects of m-c␤ARK described here. Unfortunately, all interpretations are complicated by the fact that GIRK directly binds c␤ARK, and its use as a "straightforward" G␤␥ scavenger in GIRK studies should be avoided.
To summarize, we contend that, as a G␤␥ scavenger, m-c␤ARK is inferior to m-phosducin in GIRK studies. The results obtained with m-phosducin are consistent with the existence of a single type of G␤␥ binding site underlying both basal and agonist-evoked activation of GIRK or the existence of separate sites with very similar affinities for G␤␥.
Possible Mechanism of G␤␥-independent Basal Activity-Inhibition of the whole-cell basal GIRK current, I basal , by both scavengers used here was dose-dependent and did not fully saturate even at the highest dose of scavengers used. Therefore, we cannot state with certainty that the remaining ϳ10 -15% of A GIRK,basal is G␤␥-independent. The fact that the expression of c-␤ARK at 5 ng of RNA/oocyte (a dose that normally inhibits 70 -80% of I basal ) on top of phosducin did not further reduce I basal supports the existence of a G␤␥-independent component. This putative G␤␥-independent component of A GIRK,basal , although small, is not negligible and may be physiologically relevant. It is also interesting from the biophysical point of view. What determines the open-closed equilibrium of GIRK in the absence of the main physiological gating factor, G␤␥? "Spontaneous" transitions from closed to open state are observed in many ion channels in their "resting" state, but it is seldom clear to what extent such transitions are controlled by an extrinsic modulatory factor. In GIRK channels, one such factor may be the ubiquitously present PIP 2 , which is an obligatory gating factor in these channels (60). Also, it has been proposed that intracellular Na ϩ is an important regulatory factor that may substantially regulate A GIRK,basal (23,46). Na ϩ may activate GIRK in a G␤␥-dependent manner (61) (this component should be absent in oocytes expressing the G␤␥ scavengers) and also via a direct binding (23). The EC 50 of the direct Na ϩ effect is 20 -30 mM (62), and the extent of direct activation is probably minor in resting cells, at physiological Na ϩ concentrations of 5-10 mM. Further studies will be necessary to clarify whether a G␤␥-independent component of A GIRK,basal exists and what is the underlying mechanism.