Galphai1 and Galphai3 differentially interact with, and regulate, the G protein-activated K+ channel.

G protein-activated K(+) channels (GIRKs; Kir3) are activated by direct binding of Gbetagamma subunits released from heterotrimeric G proteins. In native tissues, only pertussis toxin-sensitive G proteins of the G(i/o) family, preferably Galpha(i3) and Galpha(i2), are donors of Gbetagamma for GIRK. How this specificity is achieved is not known. Here, using a pull-down method, we confirmed the presence of Galpha(i3-GDP) binding site in the N terminus of GIRK1 and identified novel binding sites in the N terminus of GIRK2 and in the C termini of GIRK1 and GIRK2. The non-hydrolyzable GTP analog, guanosine 5'-3-O-(thio)triphosphate, reduced the binding of Galpha(i3) by a factor of 2-4. Galpha(i1-GDP) bound to GIRK1 and GIRK2 much weaker than Galpha(i3-GDP). Titrated expression of components of signaling pathway in Xenopus oocytes and their activation by m2 muscarinic receptors revealed that G(i3) activates GIRK more efficiently than G(i1), as indicated by larger and faster agonist-evoked currents. Activation of GIRK by purified Gbetagamma in excised membrane patches was strongly augmented by coexpression of Galpha(i3) and less by Galpha(i1). Differences in physical interactions of GIRK with GDP-bound Galpha subunits, or Galphabetagamma heterotrimers, may dictate different extents of Galphabetagamma anchoring, influence the efficiency of GIRK activation by Gbetagamma, and play a role in determining signaling specificity.

The importance of G protein-activated K ϩ channels (GIRK; Kir3) 1 is 2-fold: as abundant mediators of inhibitory neuronal signaling and as the first known and one of the best studied direct effectors of G␤␥ (1,2). GIRK channels convey inhibitory signals of many neurotransmitters acting via G protein-coupled receptors (GPCRs). They mediate much of the negative inotropic effect of acetylcholine (ACh) acting via the m2 muscarinic receptor (m2R) in the heart (3,4) and many of the inhibitory actions of opioids, serotonin, ␥-aminobutyric acid, dopamine, and other transmitters in the brain (2,5). They also play important roles in neuronal regulations of behavior and in mediation of GPCR-and alcohol-induced analgesia (6,7). GIRKs are inwardly rectifying, tetrameric K ϩ channels. GIRK1/2 heterotetramers, abundant in the brain, and GIRK1/4, prevailing in the heart, display a GIRK1 2 GIRKx 2 stoichiometry. Each subunit consists of a core transmembrane domain with two membrane-spanning ␣-helixes M1 and M2, flanked by cytoplasmic N and C termini, and a re-entrant helix-P-loop between M1 and M2. Recently resolved crystal structures of the cytoplasmic domain of GIRK1 and of an entire bacterial inward rectifier KirBac1.1 (8,9) show that the cytosolic domains of Kir channels tetramerize to form a cytoplasmic water-filled channel that presents continuation of the transmembrane pore (see Fig. 1A).
It is now unequivocally established that G␤␥, but not G␣, activates GIRK (10). G␤␥ binds to sites located in the N and C termini of each GIRK subunit and triggers channel opening by an unknown mechanism (11). The role of G␣ subunits in GIRK signaling has been controversial for many years (reviewed in Ref. 10), but it is now becoming clear that G␣ subunits are not mere donors of G␤␥ for GIRK activation. They physically interact with GIRK, and it has been proposed that they contribute to channel function in several ways: determining the specificity of activation by neurotransmitters, anchoring the G␤␥ subunits to GIRK to ensure fast and specific activation (12,13), and regulating GIRK basal activity and G␤␥-dependent gating (14). The latter seems to involve G␣ GDP rather than G␣ GTP . In Xenopus oocytes, expression of G␣ i3 reduces an excessively high basal activity of overexpressed GIRK channels and enhances agonist-induced activation. In excised patches, coexpressed G␣ i3 enhanced GIRK activation caused by added G␤␥ protein. This suggested that G␣ i3 (probably in its GDP-bound form) not only sequesters free G␤␥ and donates it upon agonist activation, but also predisposes (primes) the channels to G␤␥ activation (14). It is not known whether the priming effect is also produced by G␣ subunits other than G␣ i3 .
The role of G␣ in the specificity of signaling from GPCR to GIRK is widely accepted (15), but the mechanisms are unclear. GIRK is similarly activated by all G␤␥ tested (16,17) except G␤ 5 which, in combination with G␥ 2 , inhibits GIRK (17). However, in native tissues, G␤ 5 is associated mainly with certain RGS (regulators of G protein signaling) proteins rather than G␥ (18,19). Therefore, it is unlikely that selectivity of coupling is determined by G␤␥; G␣ subunits must play a role. In cardiac and neuronal cells only pertussis toxin-sensitive G proteins, G␣ i/o , activate GIRK (10). Selectivity is observed even within the G i/o family: in atrial and certain neuroendocrine and neuronal cells, upon activation of a variety of GPCRs, G␣ i2 and G␣ i3 rather than G␣ i1 and G␣ o appear to be the preferred donors of G␤␥ for GIRK (see "Discussion"). The molecular mechanisms determining the specific activation of GIRK only by certain G proteins remain unknown.
In contrast to the normal physiological situation, in heterologous expression systems (Xenopus oocytes, Chinese hamster ovary and human embryonic kidney cells) GIRK can be activated by G␤␥ released from a variety of G protein heterotrimers. Any G␣ i/o (14,20,21), including the pertussis toxin-insensitive G␣ z , activates GIRK (22,23). Even G␣ s can donate G␤␥ to activate GIRK in heterologous expression systems or in native cardiomyocytes if various components of the signaling pathway (␤-adrenergic receptors, G␣ s , or G␤␥) are overexpressed (24 -26). Of all G␣ tested, only G␣ q is unable to donate G␤␥ in expression systems; activation of G q by the relevant GPCRs inhibits GIRK, apparently via indirect pathways that involve second messengers (27,28).
How, then, is the selective activation of GIRK via G i/o in native cardiac and neuronal cells ensured? Evidently, selective coupling between GPCRs and G␣␤␥ heterotrimers contributes to the specificity of signaling to GIRK by neurotransmitters (20,29), but this cannot explain the selectivity at the G protein-GIRK interface (e.g. why G q does not activate GIRK). Thus, factors such as colocalization or scaffolding may be involved in assigning the relevant G␣ to GIRK. It has been proposed that pre-formed complexes of GIRK with certain G␣␤␥ heterotrimers may exist in native cells, assuring fast and selective activation of GIRK by G␤␥ derived from the GIRK-anchored heterotrimer (12,13,30). Supportive evidence is that G␣ i1 and G␣ i3 bind to the N terminus and mutations in N terminus reduce the speed of activation of GIRK by agonists of GIRK1 (12,14,31). However, this attractive hypothesis has not been conclusively confirmed.
Fundamental questions regarding the proposed complexes of GIRK with regulatory proteins remain open: whether GIRK subunits besides GIRK1 bind G␣ i ; whether additional G␣-binding segments besides N terminus exist in GIRK; whether other G␣ subunits besides G␣ i1 and G␣ i3 bind to GIRK; and whether differences in binding of various G␣ subunits to GIRK contribute to signaling specificity and gating. To start to address these problems, we compared G␣ i3 and G␣ i1 subunits, which showed striking differences in their binding to GIRK in a preliminary comparative assay, for their ability to directly interact with the channel molecule and for their ability to regulate the physiological parameters of channel function. We identified and mapped G␣ i3 binding sites in N-and C-terminal parts of GIRK1 and GIRK2 subunits and then performed a comparative study of binding and physiological effects of G␣ i3 and G␣ i1 . We found that G␣ i3 was a much better binding partner for GIRK1 and GIRK2 (at all binding sites) and a better physiological companion for G␤␥ in GIRK activation. The correlation between binding and function of these two G␣ subunits supports a role for physical interaction between G␣ GDP and GIRK in determining selective regulation of GIRK by different G␣ subunits.
The cDNAs, mRNAs, GST fusion proteins, and total oocyte membrane fraction were prepared as described in our previous publications (14,32), with minor modifications. Two buffers were used in pull-down experiments: phosphate-buffered saline or a 150 mM-K ϩ buffer (in mM: 50 Tris, 5 MgCl 2 , 1 EDTA, 150 KCl, pH 7.5). Identical binding patterns were seen in both buffers. [ 35 S]Methionine-labeled G␣ proteins were synthesized in reticulocyte lysate for 1.5-2 h, then 2-5 l of lysate were incubated with 100 M of either GDP or GTP␥S for 30 min at 30°C in 50 or 95 l of binding buffer with the detergent of choice, 0.05% Tween 20 or 0.5% CHAPS. To measure G␣ binding to GST fusion proteins, 5-10 g of the GST fusion protein were added, the total reaction volume brought to 300 or 500 l, and the incubation continued in the presence of 100 M GTP␥S or GDP for 1-2 h at room temperature, and binding to glutathione-Sepharose beads and elution with 15 mM glutathione were done as described previously (14). The eluted proteins were run on 10% or 12% polyacrylamide-SDS gels (SDS-PAGE). The radioactive signals from protein bands of the gels were imaged and quantitated using PhosphorImager and the software ImageQuaNT (Molecular Dynamics). The readings were corrected for the differences in the amounts of GST fusion proteins eluted from the beads as measured from Coomassie Blue-stained gels (in most cases such differences did not exceed 50%). The full-length N terminus of GIRK1, G1 1-84 , bound G␣ i3 as well as its distal part, G1 34 -86 (data not shown), and the results obtained with these two proteins were pooled. For trypsin digestion, 1 l of 0.25% stock of trypsin was diluted ϫ100 and incubated for 20 min at 30°C with 30 l of 150 mM-K ϩ buffer with 0.5% CHAPS containing 2 l of G␣ lysate, treated with GTP␥S or GDP as described above. Western blots were done using standard procedures, using a common G␣ i antibody (Calbiochem) and ECL reagents from Pierce Inc. Intensity of Coomassie Blue labeling was quantified using TINA software (Raytest, Straubenhardt, Germany).
Results are shown as mean Ϯ S.E.; numbers of cells, patches, or separate biochemical determinations are shown above the bars in the figures. Multiple group comparison was done using one-way analysis of variance followed by Student-Newman-Keuls or Tukey's tests. One or two asterisks indicate p Ͻ 0.05 and p Ͻ 0.01, respectively, compared with control group. Values of t 90% parameters in G␣ i1 -and G␣ i3 -expressing groups, which report measurements independent of the control group, were compared by t test.

RESULTS
We explored the binding of G␣ i3 to GST-fused fragments of cytoplasmic N and C termini of GIRK1, using a standard pulldown method (Fig. 1B). The GST fusion proteins were immobilized of glutathione-agarose affinity beads. [ 35 S]Methioninelabeled G␣ i3 was synthesized in vitro in rabbit reticulocyte lysate, as it was done previously to study the interaction of G␣ i3 with the N terminus of GIRK1 (14). In vitro synthesized G␤␥ has also been successfully utilized to identify G␤␥ binding sites of voltage-gated Ca 2ϩ and GIRK channels (32,33). The binding experiments were routinely conducted in buffers containing 0.5% CHAPS, which minimized nonspecific binding of G␣ i3 to GST. All initial screening and mapping experiments reported in Figs. 1 and 2 were performed with GDP-bound G␣ i3 (see "Experimental Procedures"). In each experiment the amount of bound G␣ protein was measured from autoradiograms using a PhosphorImager and normalized to the amount of the bound G1-C3 segment, which was present in all experiments (Fig.  1C). We stress that this methodology gives a highly sensitive and accurate quantitation of bound protein, because the radioactive signal from labeled G␣ is measured from dried gels by using the PhosphorImager directly, in an almost unlimited linear range, and with high precision. In comparison, in Western blotting, several additional intermediate steps are involved, which reduce the accuracy of measurement.
In initial experiments, we found that not only the N terminus of GIRK1, but also the full-length C terminus of GIRK1 (G1 183-501) and of GIRK2 bound G␣ i3 ( Fig. 1C and data not shown). We then examined the binding of G␣ i3-GDP to the whole C terminus and to six non-overlapping GST-fused fragments of C terminus, G1-C1 through G1-C6 (Fig. 1B), previously used to map the G␤␥ binding sites (32). Fig. 1C shows that three fragments covering all of the proximal part of C terminus of GIRK1, G1-C1 through G1-C3, bound G␣ i3-GDP ; G1-C3 showed the strongest binding (summarized in Fig. 1D). Interestingly, the whole C terminus (G1 183-501 ) bound G␣ i3 far weaker than G1-C3. It is possible that the complete C terminus is folded in such a way that the distal part beyond aa 380 (which was proposed to play a role of a "switch" that regulates the activation of the channel in protein kinase A-dependent manner (34 -36)) impedes the binding of G␣ i to the main C-terminal binding segments.
We also examined G␣ i3 binding to GIRK2 using GST-fused GIRK2 segments shown in Fig. 2A (sequence alignments of N termini and parts of the C termini of GIRK1 and GIRK2 are shown in Fig. 2B). The strongest binding of G␣ i3-GDP was observed in the large segment G2-C2, aa 310 -414, and in its smaller subdivision G2-C1-2 (aa 310 -380) ( Fig. 2C; summarized in Fig. 2D). This segment mostly overlaps G1-C3 of GIRK1 except the first 20 aa (Fig. 2B, part b). The first half of the N terminus did not bind G␣ i3 , whereas the second half bound G␣ i3-GDP , although apparently somewhat weaker than G2-C2-1. The absence of G␣ i3 binding to G2-C1-2, which lacks 20 C-terminal aa compared with the homologous G1-C2 that binds G␣ i3 well, hints that this 20-aa stretch contains an important binding determinant for G␣ i3 .
Incubation of in vitro synthesized G␣ i3 with GTP␥S protected G␣ i3 from trypsin digestion, as expected for a GTP␥S-bound G␣ (37) (Fig. 3A). This result confirms GTP␥S binding and the consequent change in G␣ conformation and attests to functional integrity of in vitro synthesized G␣ i3 protein. Under mild non-ionic detergent conditions (0.05% Tween 20), the extents of binding of G␣ i3-GDP and G␣ i3-GTP␥S to the N terminus of GIRK1 were comparable (Ref. 14 and data not shown). However, under the standard conditions used in most experiments, with 0.5% CHAPS, the binding of G␣ i3-GTP␥S was weaker than that of G␣ i3-GDP (Fig. 3C). The same was observed with the C-terminal G1-C3 segment (Fig. 3, B and C). The summary in Fig. 3C shows that, in 0.5% CHAPS, the binding of G␣ i3-GTP␥S to all G␣-binding fusion segments of GIRK1 and GIRK2 was only about a third of that of G␣ i3-GDP (Fig. 3C).
The G␣ i3 binding segments identified by pull-down also bind G␤␥ (32). To rule out the possibility that G␣ GDP binds to GIRK indirectly, via G␤␥ present in the reticulocyte lysate, we performed two types of control experiments. First, we verified that binding of G␣ i3 to different segments was not increased by the addition to the reaction mixture of various amounts of untreated reticulocyte lysate, suggesting that the binding of G␣ i3-GDP to GIRK was not via an endogenous protein present in the lysate (data not shown). Second, addition of lysate containing in vitro synthesized G␤ 1␥2 did not increase the binding of G␣ i3-GDP to G1 1-84 and G1-C1 (data not shown). Two additional facts argue against involvement of G␤␥ in mediating G␣ i3 binding to GIRK: 1) the binding preference for G␤␥ in the C terminus of GIRK1 (G1-C1 Ͼ G1-C2 Ͼ G1-C3; see Ref. 32) is opposite to that of G␣ i3 and 2) G␣ i1-GDP does not bind to G1-C3 (see below), despite the very high affinity of binding of G␣ i1-GDP to G␤␥ (38). We conclude that, under our experimental conditions, G␣ i3 binds directly to GST-fused segments of GIRK1.
Preliminary screening of several G␣ subunits indicated weak binding of G␣ i1-GDP to fusion proteins of GIRK1. We thus focused on the comparison of G␣ i3 and G␣ i1 . Functional integrity of G␣ i1 was verified by the trypsin digestion test (data not shown). For each GST fusion protein, the binding reaction was performed with equal volumes of reticulocyte lysate containing  (60). The dashed lines represent the approximate boundaries of the lipid phase of the membrane. Cytosolic domains of two out of four subunits are shown according to Nishida and MacKinnon (8). The folding of the C terminus of GIRK1 between residues 372 and 501 is not known and is shown arbitrarily. Linkers between cytoplasmic and transmembrane parts have not been resolved in the crystal structure of GIRK. Parts of the C terminus of GIRK1 tinted black, gray, and light gray correspond to G1-C1, G1-C2, and G1-C3, respectively. B, linear presentation of GIRK is shown at the top, and the GST-fused segments, with numbers of first and last amino acids, are shown below. NT, N terminus; CT, C terminus; TM, transmembrane domain. C, a representative pull-down experiment showing the binding of G␣ i3-GDP to GST fusion proteins of GIRK1. The upper panel shows Coomassie Blue-stained proteins eluted from the beads and subjected to SDS-PAGE, and the lower panel shows a PhosphorImager autoradiogram of the bound [ 35 S]methionine-labeled G␣ i3 . Note that the G␣-binding GST fusion proteins in this experiment (G1-C1 through G1-C3, and G1 183-501 ) have been added in lower amounts than GST fusion proteins of the second half of C terminus (G1-C4 through G1-C6); despite this, the latter did not bind G␣ i3 . D, G␣ i3 binding to GST fusion proteins of GIRK1, normalized to G1-C3. the labeled G␣ subunits, and the amount of bound G␣ was normalized to the amount of G␣ added to the reaction in the beginning ("input"). In all segments of GIRK1 tested, the binding of G␣ i1-GDP was consistently weaker than that of G␣ i3-GDP (Fig. 4, A and C), and the binding of G␣ i1-GTP␥S was too weak to be reliably quantitated (data not shown). Some G␣ i1 binding was observed in the N terminus, confirming a previous finding (12), but it amounted only to about 20% of G␣ i3-GDP . G1-C3 did not bind G␣ i1-GDP at all. In GIRK2, both G␣ i3 -binding segments (G2-N2 and G2-C2-1) gave very weak signal with G␣ i1-GDP (Fig. 4, B and C).
We utilized Xenopus oocytes to examine how coexpressed G␣ i3 and G␣ i1 affect the function of GIRK and to see whether their effect on various parameters of channel function correlate with the observed differences in their binding. To avoid promiscuous interactions caused by excessive expression of signaling proteins, which may obscure specificity (39), it is important to avoid overexpression and to titrate protein levels (which do not necessarily correlate with the amount of injected RNA). Titration of protein levels was especially important when comparing the effects of two homologous G␣ proteins, G␣ i3 and G␣ i1 , to avoid a situation in which the weakly interacting G␣ i1 , present at high levels, "mimics" the effects of the strongly interacting but poorly expressed G␣ i3 .
The protein levels of G␣ i3 and G␣ i1 were measured upon injection of 1 ng of RNA/oocyte, an amount widely used in functional experiments (see below). The amounts of G␣ were quantitated using Western blot methodology with a common G␣ i antibody. Because it is possible that the antibody does not bind with the same affinity to G␣ i1 and G␣ i3 , we performed a calibration procedure. [ 35 S]Met-labeled G␣ i1 and G␣ i3 were synthesized in reticulocyte lysate, run on gel, and transferred to a blot membrane. The radioactive signal from the membrane (Fig. 5A, "PhosphorImager") was compared for the two proteins and corrected for the number of methionines in each molecule, giving the relative molar amounts of each protein on the blot membrane (normalized to G␣ i3 ). The chemiluminescence signal obtained from Western blot of the same blot membrane (Fig.  5A) could then be calibrated to give a correction factor (equal to 1 if the antibody recognized the two proteins equally well), which can be used to correct the quantitative results from Western blots of G␣ in any cell. G␣ i1 and G␣ i3 (1 ng of RNA) were expressed in oocytes, and signals from Western blots of membrane fractions (Fig. 5B) were quantitated and corrected. The results indicated similar levels of expression of the two G␣ proteins, with a slight preference toward G␣ i1 (ϳ40%; Fig. 5C).
In most cases the choice of amounts of RNAs injected for functional experiments was based on titrations made in our previous works and designed to avoid excessive protein levels. Heterotetrameric GIRK1/2 channels were expressed either at a low level (density), 0.025-0.05 ng of RNA/oocyte of each GIRK subunit, or at a high density (1 ng of RNA/oocyte). The amounts of RNAs of G␣ subunits were varied depending on the density of expressed channel (14). To activate GIRK, human m2 receptor (m2R) was expressed at 0.5 ng of RNA per oocyte, sufficient to activate all expressed GIRK channels (23). No coexpression of G␤␥ is needed, because the oocytes usually have high levels of endogenous G␤␥ sufficient for activation of all GIRK channels even when those are at high density (14).
Whole cell GIRK currents were studied in intact oocytes under two-electrode voltage clamp. Exchanging the external solution from a low K ϩ (2 mM) ND96 solution to a high K ϩ (24 mM) solution revealed the basal GIRK current, I basal . Addition of 10 M acetylcholine elicited the agonist-evoked current, I ACh (Fig. 6A). When GIRK was expressed at low density (15-50 pg of RNA of each channel subunit per oocyte), coexpression of an excess of G␣ i3 or G␣ i1 (0.5 ng of RNA) reduced I basal by about half (Fig. 6B, part a). I ACh was not altered in G␣ i1 -expressing cells but augmented in G␣ i3 -expressing cells (p Ͻ 0.01, Fig. 6B, part c). The total GIRK current (I total ϭ I basal ϩ I ACh ) was not affected by the expression of either G␣ i1 or G␣ i3 (Fig. 6B, part e). The relatively modest effects of G␣ i3 at low densities of GIRK can be attributed to the presence of sufficient amounts of endogenous G␣ i/o that are able to sustain agonist-induced currents that are above basal agonist-independent activity, even without the coexpressed G␣ i (14). When GIRK was expressed at high densities (1 ng of RNA/ oocyte), expression of 0.5 ng of RNA of G␣ i1 or G␣ i3 did not affect I basal and slightly enhanced I ACh (data not shown). Expression of excess G␣ i1 or G␣ i3 (5 ng of RNA) greatly reduced I basal and significantly enhanced I ACh (Fig. 6B, parts b and d), as reported (14). However, I total was reduced, especially by G␣ i1 (Fig. 6B, part f). G␣ i3 appeared to cause a stronger enhancement than G␣ i1 , but the difference was not statistically significant. The strong reduction in I basal and a decrease in I total suggest a substantial scavenging of free G␤␥, and much of the enhancement of I ACh with both G␣ proteins may reflect a greater available pool of G␤␥ donors. Also, it is possible that excess of the "poor" interacting partner (G␣ i1 ) saturates the expressed channels, mimicking the preferred partner (G␣ i3 ) and masking the differences in efficiency of coupling with the channel. Therefore, we injected moderate amounts of RNA of G␣ i1 or G␣ i3 , 1 ng/oocyte. This caused only a slight ϳ20% reduction of I basal (Fig. 5B, part b), whereas I ACh was substantially increased: by 89 Ϯ 14.6% by G␣ i1 , and by 178 Ϯ 18% by G␣ i3 (Fig. 5B, part d; see examples in Fig. 5A). The enhancement by G␣ i3 was significantly greater than by G␣ i1 (p Ͻ 0.01). I total was not changed by G␣ i1 and slightly but significantly increased by G␣ i3 . If the G i3 heterotrimer is superior to G i1 in forming a complex with GIRK, and if such complexes exist prior to activation of the GPCR by agonist, a faster kinetics of activation of I ACh with G␣ i3 may be expected (13,40). Fig. 6C shows that this was indeed the case: coexpression of G␣ i3 significantly (p Ͻ 0.01) accelerated the kinetics of activation of I ACh , measured as time to 90% of peak I ACh (t 90% ). G␣ i1 also reduced the activation time (p Ͻ 0.05), but less than G␣ i3 . Although finite speed of perfusion (about 1 s) imposed a lower limit on measurements of activation time in these experiments, the remarkable acceleration of activation by G␣ i3 (from 8.8 Ϯ 2.1 s in control to 1.8 Ϯ 0.3 s in the presence of G␣ i3 ) cannot be an artifact of insufficiently fast wash in of the solution.
In intact cells, the increase in I ACh may result from a larger pool of G␣␤␥ available as donors of free G␤␥, but also from an effect of G␣ i on GIRK gating (14). The relative contribution of these factors to G␣ i -induced increase in I ACh cannot be explicitly estimated from whole cell recordings. In contrast, when GIRK is activated by purified G␤␥ in excised membrane patches, activation is receptor-independent and bypasses the donor stage. Activation of GIRK by G␤␥ in excised patches, relative to basal activity, is augmented by G␣ i3 (a priming effect) (14). To inquire whether G␣ i1 can prime GIRK for activation by G␤␥ in excised patches, we injected the oocytes with amounts of RNAs of G␣ i1 or G␣ i3 , which were slightly greater than of GIRK (1 ng of each GIRK subunit and 2 ng of G␣, or 2 ng of GIRK and 3 ng of G␣). In most experiments m2R was not expressed. Typical records made 3-4 days after RNA injection FIG. 4. G␣ i3 binds to GIRK subunits better than G␣ i1 . A, comparison of binding of G␣ i1-GDP and G␣ i3-GDP to GST-fused fragments of GIRK1. Results of several representative experiments are shown. The terms input and binding have the same meaning as in Fig. 2C. B, comparison of binding of G␣ i1 and G␣ i3 to GST-fused fragments of GIRK2. A representative experiment is shown. C, summary of experiments like those in A and B for G␣ i1-GDP and G␣ i3-GDP . For each GST fusion protein, the binding of G␣ i1-GDP was presented as percent of binding of G␣ i3-GDP , after correction for differences in inputs of G␣ i1 and G␣ i3 .
are shown in Fig. 7A. On the average, levels of basal activity recorded in cell-attached (c.a.) configuration were reduced by each G␣ compared with the control group where the channel was expressed alone (Fig. 7B, part a). The basal activity in these experiments corresponded to medium to high channel density (see Ref. 14).
After 1-2 min of c.a. recording, the patch was excised into a Na ϩ -and GTP-free, high K ϩ bath solution that contained 2 mM MgATP (the latter is needed to sustain viable levels of phosphatidylinositol bisphosphate, necessary for proper GIRK function (41)). After an initial reduction, which was similar in all oocyte groups (not shown), the basal activity stabilized. 20 nM purified recombinant G␤␥ was added 3 min after excision, and activation relative to basal activity in the same patch was measured. The results are presented as -fold activation relative to either basal c.a. activity (Fig. 7B, part b) or to basal activity FIG. 5. Estimating the levels of expression of G␣ i1 and G␣ i3 subunits after calibration of Western blot signals. A, comparison of radioactive (PhosphorImager) and chemiluminescence (Western blot) signals from in vitro synthesized, [ 35 S]methionine-labeled G␣ subunits. B, Western blot of total oocyte membrane fraction with the same antibodies is shown. All RNAs were 1 ng/oocyte except m2R, which was 0.5 ng/oocyte. C, relative amounts of expressed G␣ measured from blots like that in B (n ϭ 3; right panel), after correction using correction factors (left panel) for antibody signals measured from three experiments like that shown in A.
FIG. 6. Effects of coexpression of G␣ i3 and G␣ i1 on whole-cell GIRK1/2 currents in Xenopus oocytes. The code for bar shading for panels B and C is shown in the inset in the middle of the figure. A, representative current records in oocytes expressing the channel alone (left), with G␣ i3 (center), or with G␣ i1 (right). Each oocyte was injected with 1 ng of each RNA except m2R, which was 0.5 ng. B, summary of effects of G␣ i1 and G␣ i3 on amplitudes of basal, ACh-evoked, and total GIRK currents. I basal , I ACh , and I total in each oocyte were normalized to the average current recorded in control group (no G␣) of the same oocyte batch, and averaged across all experiments (see Ref. 14). C, comparison of kinetics of activation of I ACh in oocytes expressing different G␣ proteins. The upper panel shows the first 14 s of I ACh in the same oocytes shown in A. Currents were scaled to allow a direct comparison of speed of activation. The lower panel shows a summary of measurement of t 90% from 18 -20 oocytes. *, p Ͻ 0.05; **, p Ͻ 0.01 compared with control (no G␣). Filled circle with stars indicates significant difference between G␣ i1 and G␣ i3 groups, p Ͻ 0.01. in excised configuration (Fig. 7B, part c). Both types of analysis showed that both G␣ i3 and G␣ i1 enhanced G␤␥-induced activation, but G␣ i3 was more efficient, causing ϳ4-fold priming. G␣ i1 caused 2-to 3-fold priming. The absolute values of G␤␥-activated currents (which correspond to I total in whole cell recordings) did not differ significantly in the three groups, although a certain tendency to decrease was observed in the presence of G␣ i1 (Fig. 7B, part d). DISCUSSION We have identified G␣ i3-GDP binding sites in N-and C-terminal parts of the ubiquitous GIRK1 and the neuronal GIRK2 subunits and investigated the correlation between G␣ i binding and the effect on channel function. Coexpression of G␣ i3 improved the speed and extent of agonist-and G␤␥-induced activation of GIRK. A homologous member of the G␣ i/o family, G␣ i1 , showed substantially weaker binding to GIRK than G␣ i3 , and correspondingly served as a poorer donor of G␤␥ and a weaker enhancer of G␤␥-dependent activation. Our results support the postulate (13,14) that direct interactions of GDP-bound G␣ i subunits (or G␣␤␥ heterotrimers) with GIRK regulate the biological effect of G␤␥ and take part in determining signaling specificity. Active modulation of a G␤␥ effector by G␣ GDP (or G␣ GDP ␤␥) is a novel aspect of G protein signaling and warrants attention as a phenomenon that may exist in other G proteincoupled signaling pathways.
G␣ i3-GDP Binds to Strategically Important Parts of GIRK-In addition to the previously identified N-terminal binding site in GIRK1 (12,14), we found G␣ i3 binding sites in the N terminus of GIRK2 and in the C termini of both GIRK1 and GIRK2. The strongest binding was to the G1-C3 segment in GIRK1 (amino acids 320 -370), and a homologous segment G2-C2-1 in GIRK2 (aa 310 -380, correspond to aa 299 -369 of GIRK1; see Fig. 3B). G␣ i1 binds much weaker than G␣ i3 to both GIRK subunits. The existence of multiple N-and C-terminal G␣ i -binding segments implies large G␣-interacting surfaces, which probably contribute to three-dimensional G␣ i binding sites formed by one or more GIRK subunits.
According to the crystal structure of the cytosolic domain of GIRK1 (8), the G␣-binding segments lie at the surface of this domain and are exposed to cytosol, which makes them available for protein-protein interactions (see Fig. 1A). These seg-ments harbor (or lie close to) aa residues crucial in the regulation of GIRK gating, for instance glutamate 304 in GIRK1 (315 in GIRK2) and some adjacent residues (42,43). The sequence VDY (340 -342 in GIRK1, 346 -348 in GIRK2), which is homologous to IDY (aa 298 -300) in KirBac1.1, and adjacent segments in other Kir channels, are sites of interaction with the N terminus, which play an important role in gating processes in Kir channels (9,44). The N-terminal G␣ binding sites in GIRK subunits contain, at the boundary with the first transmembrane ␣-helix, a stretch (aa 64 -77 in GIRK1) highly conserved among all Kir channels (9). This segment in KirBac1.1 forms a "slide" helix linking the transmembrane, pore-forming part of the channel with the cytosolic domain; it has been proposed to play a key role in Kir gating by linking movements of the cytosolic domain and of transmembrane helixes M1 and M2 (9). Hence, the N-and C-terminal segments of GIRK that interact with G␣ overlap, or are located close to, parts of channel molecule that are crucially involved in channel gating.
The binding sites for G␣ i3 are also spatially close to, and may partially overlap, the G␤␥ binding sites found in the N and C termini (12,32,45,46). This implies a complex pattern of interaction between G protein subunits and the channel, which may involve elements of competition and/or synergistic interactions. However, because the continued presence of coexpressed G␣ i3-GDP does not impair (actually enhances) activation by added G␤␥ in excised patches, competition between G␣ GDP and G␤␥ for binding to GIRK under physiological conditions appears unlikely.
G␣ i as an Anchor and Donor of G␤␥-Anchoring of G␤␥ (within a G␣␤␥ heterotrimer attached to the channel) and donating it upon addition of agonist is the most acknowledged potential function of G␣. Our working hypothesis, based on suggestions of Lily Jan and collaborators (12,13) and later our own (14), is that physical interaction between G␣ i and GIRK allows fast coupling between the GPCR and GIRK. It enables the formation of a signaling complex that includes G␣, G␤␥, and the effector and enhances signaling specificity. Here, we utilized the differences in G␣ i1 and G␣ i3 binding to GIRK1 and GIRK2 to further explore this hypothesis in Xenopus oocytes. These cells are uniquely suited for quantitative titration of protein levels by exactly varying the amounts of injected RNA FIG. 7. Coexpression of G␣ i3 , but not G␣ i1 , enhances G␤␥-evoked GIRK response in excised patches. A, representative records from patches of three different oocytes expressing GIRK1/2 alone (a), with G␣ i1 (b), or G␣ i3 (c). B, parameters of GIRK activity recorded in patches: a, average basal activity in cellattached patches, in pA; b, average extent of activation by 20 nM G␤␥ relative to basal activity in the same patch in cellattached configuration; c, extent of activation by 20 nM G␤␥ relative to basal activity in the same patch measured after excision, during the last minute before the addition of G␤␥; d, total G␤␥-induced currents (in pA). *, p Ͻ 0.05; **, p Ͻ 0.01 compared with control (no G␣). Filled circles with stars indicates significant difference between G␣ i1 and G␣ i3 groups, p Ͻ 0.01. over a wide range. Here and previously (14), levels of expressed proteins have been also assessed by direct immunochemical measurements. Uncertainties introduced by possible differences in GPCR-G␣ coupling were avoided by using m2R, which couples equally well to all members of G␣ i/o family (47).
Mass action law dictates that, for similar expression levels of two G␣ i species, more GIRK-G␣␤␥ complexes will be formed with G␣ that binds GIRK with higher affinity, as long as the expression level of G␣ is non-saturating. The existence of such a complex will ensure faster activation by agonist than in the case when G␤␥ diffuses from a non-anchored heterotrimer (30). Therefore, kinetics of activation provide reliable information on the efficiency of G␣-GIRK coupling (23,48). Since I ACh shows desensitization (reviewed in Ref. 5), faster activation should also result in a greater agonist response amplitude (49). However, the kinetic parameter is more informative, because the difference in amplitudes of I ACh may also stem from the less well understood phenomenon of priming (see below) or from an increase in surface expression of GIRK caused by coexpression of G␣ i3 (14). We found both faster kinetics of activation and larger I ACh with coexpressed G␣ i3 than with G␣ i1 , supporting the hypothesis described above. The functional inferiority of G␣ i1 was seen at all doses of G␣ tested. However, it was most prominent at moderate expression levels, and became less pronounced at high levels of G␣ i1 . Such "promiscuous" coupling between poorly interacting proteins is common in heterologous expression systems upon overexpression of one of the components (39). This is in agreement with mass action law: when the concentration of G␣ i1 becomes saturating (presumably in great excess over GIRK), it should be able to cause the same maximal effect as G␣ i3 if the only difference between them is the affinity of interaction with GIRK. However, when expressed in excess, both G␣ i1 and G␣ i3 also act as general scavengers of G␤␥, forming G␣␤␥ heterotrimers not complexed with GIRK and reducing I total (14). If less G␣ i1 is complexed with GIRK compared with G␣ i3 , more G␣ i1 is available for general G␤␥ scavenging. Our results support this notion: although G␣ i1 accelerates the activation and increases I ACh at moderate and high expression levels, it never reaches the efficiency attained by G␣ i3 and causes a greater reduction in I total than G␣ i3 at 5 ng of RNA/oocyte (Fig. 6B). In all, our results suggest that G␣ i1 is less efficient than G␣ i3 as a G␤␥ donor for GIRK under most conditions.
The strength of G␣-GIRK interaction is not the only factor that determines the identity of G␣ i/o that couples a GPCR to GIRK in native cells. Additional factors include coupling specificity at the GPCR-G␣ interface (21) and G␣ availability. Theoretically, a cell with excess G␣ i1 may preferentially use it for agonist signaling to GIRK. Yet, published data imply a preferential use of G␣ i3 (or G␣ i2 ) over G␣ i1 and G␣ o to activate GIRK. An antibody-and antisense-based study suggested that somatostatin-induced GIRK current in AtT-20 cells critically depends only on G␣ i3 , whereas in locus coeruleus neurons it is mediated by G␣ i2 and not by G␣ i1 , G␣ i3 , or G␣ o (50). Notably, all G␣ i/o subunits are present in the brain (51). In the atrium, despite the abundance of G␣ o (52,53), only G␣ i2 and G␣ i3 , but not G␣ o mediate the m2R-mediated GIRK current I K(ACh) , as shown by targeted inactivation and re-expression of G␣ subunits (54 -56). Importantly, m2R does not prefer G␣ i2 or G␣ i3 over G␣ o , because it specifically uses G␣ o to inhibit ventricular Ca 2ϩ channels (55). Unfortunately, the role of G␣ i1 has not been specifically examined in these works, and its abundance in the heart is controversial (51,53). A correlation between relative abundance of G␣ i2 or G␣ i3 and their use as the main donors of G␤␥ cannot be ruled out at present.
Our results show that, in the zwitterionic detergent CHAPS, binding of G␣ i3-GTP␥S to GIRK is significantly weaker than of G␣ i3-GDP . Stronger binding of G␣ i1-GDP than G␣ i1-GTP ␥S to the N terminus of GIRK1 has also been reported (12). However, under mild non-ionic detergent conditions (0.05% Tween 20) the binding of G␣ i3 to N terminus of GIRK1 was not reduced by GTP␥S (14). Even if measurements of binding in CHAPS accurately reflect the changes in G␣ binding that take place in vivo, our results indicate that G␣ i-GTP remains attached to the channel following the GPCR-catalyzed exchange of GDP to GTP. When GTP is hydrolyzed, fast recapture of G␤␥ by closely positioned G␣ would ensure fast deactivation upon washout of agonist. Indeed, for fast-dissociating agonist-receptor complexes, macroscopic deactivation is limited only by the speed of hydrolysis of GTP but not G␤␥ dissociation form GIRK (57). The role of GPCR and of RGS proteins and the mode of their interaction with the G protein within such a multiprotein complex, remain to be elucidated. G␣ i as a Regulator of GIRK Gating and Function-Experiments in excised patches provide the opportunity to examine whether G␣ affects the function of the channel in separation from the donor function. Added G␤␥ (not G␤␥ released from an endogenous G␣␤␥ "donor" pool as in whole cell experiments) activates GIRK channels in the absence of coexpressed G␣. Coexpression of G␣ i3 not only reduces I basal but also enhances activation of GIRK by G␤␥, thus G␣ i3 serves as a primer for G␤␥ effect (14). We proposed that this priming effect and part of the reduction in I basal are the consequences of a direct interaction of G␣ i3 with the channel; we also envisaged a link between the two effects (14). The patch clamp experiments presented here showed that the extent of priming correlated with binding of G␣ i1 and G␣ i3 . G␣ i3 enhanced GIRK activation caused by G␤␥ in the absence of agonist or GPCR, and under the conditions of little general scavenging of G␤␥ (as witnessed by the preservation of the amplitude of total G␤␥-activated current). The weaker priming by G␣ i1 supports the view that this phenomenon depends on direct interaction between G␣ and GIRK. Because the patches were excised into a GTP-free solution, it is most probable that the priming effect is caused by G␣ GDP or G␣␤␥ rather than G␣ GTP . On the other hand, G␣ i1 and G␣ i3 caused a similar reduction in I basal (both in patches and in whole cells). This weakens the possibility of a direct inhibitory effect of G␣ i on basal activity. However, because the latter appears to be mostly G␤␥-dependent (14,58), a direct scavenging of G␤␥ may play a major role in the reduction of I basal by G␣, masking a minor contribution of a direct effect of G␣. At present we do not understand the mechanism of the priming effect of G␣ GDP and its linkage to the reduction in the basal activity. A possible scenario is one in which the channel possesses two types of G␤␥ binding sites, one underlying the basal activity (high affinity site) and another, low affinity site responsible for agonist-evoked activity (46,59). G␣, by forming a triple complex with G␤␥ and the channel, prevents channel activation via the high affinity site, reducing I basal , but changes the conformation of GIRK to enhance activation via the low affinity site once free G␤␥ becomes available. Variants of this model may explain the known facts related to basal and evoked activities of GIRK, but a more specific formulation will require further experimentation.