Gαi and Gβγ Jointly Regulate the Conformations of a Gβγ Effector, the Neuronal G Protein-activated K+ Channel (GIRK)*

Stable complexes among G proteins and effectors are an emerging concept in cell signaling. The prototypical Gβγ effector G protein-activated K+ channel (GIRK; Kir3) physically interacts with Gβγ but also with Gαi/o. Whether and how Gαi/o subunits regulate GIRK in vivo is unclear. We studied triple interactions among GIRK subunits 1 and 2, Gαi3 and Gβγ. We used in vitro protein interaction assays and in vivo intramolecular Förster resonance energy transfer (i-FRET) between fluorophores attached to N and C termini of either GIRK1 or GIRK2 subunit. We demonstrate, for the first time, that Gβγ and Gαi3 distinctly and interdependently alter the conformational states of the heterotetrameric GIRK1/2 channel. Biochemical experiments show that Gβγ greatly enhances the binding of GIRK1 subunit to Gαi3GDP and, unexpectedly, to Gαi3GTP. i-FRET showed that both Gαi3 and Gβγ induced distinct conformational changes in GIRK1 and GIRK2. Moreover, GIRK1 and GIRK2 subunits assumed unique, distinct conformations when coexpressed with a “constitutively active” Gαi3 mutant and Gβγ together. These conformations differ from those assumed by GIRK1 or GIRK2 after separate coexpression of either Gαi3 or Gβγ. Both biochemical and i-FRET data suggest that GIRK acts as the nucleator of the GIRK-Gα-Gβγ signaling complex and mediates allosteric interactions between GαiGTP and Gβγ. Our findings imply that Gαi/o and the Gαiβγ heterotrimer can regulate a Gβγ effector both before and after activation by neurotransmitters.

It is believed that signaling via G protein-coupled receptors (GPCRs) 5

occurs within multiprotein complexes that include
GPCRs, G proteins, and effectors (1)(2)(3). The G protein-activated K ϩ channel (GIRK, Kir3), an important mediator of neuronal inhibition (4), is activated by the binding of G␤␥ to the cytosolic N and C termini (NT and CT, respectively) of GIRK. G␤␥ associates with the channel before and after receptor activation (5,6). GIRK NT and CT segments also bind G␣ i/o (see Ref. 7), but it is not clear whether and how G␣ i/o regulates GIRK in vivo. G␣ clearly plays a role in determining specificity of signaling from GPCR to GIRK (8,9). In heterologous systems, expression of G␣ i reduces GIRK basal activity (I basal ) and increases the relative extent of activation by added or coexpressed G␤␥ (10 -12). Recently, we have demonstrated that this regulation is exerted by the GDP-bound form of G␣ i , G␣ i GDP (12); no role for G␣ i GTP in GIRK gating could be assigned so far. We proposed that regulation of GIRK by G␣ i relies upon the formation of the G␣ i GDP -G␤␥ heterotrimer, which forms a persistent, dynamic signaling complex with GIRK to ensure proper gating with low I basal and high signal-to-background ratio upon G␤␥ activation (11,12).
We hypothesized that within a multiprotein signaling complex, the partner proteins alter each other's conformation and activity at various stages of the signaling process. Thus, G␣ i and G␤␥ may regulate the conformation of the channel both before and after activation. We explored this hypothesis in vitro and in vivo using biochemical, functional, and intramolecular Förster resonance energy transfer (i-FRET) methods. Our data reveal interdependent triple interactions among G␣ i3 , G␤␥, and GIRK subunits, which correlate with distinct conformational and gating states of the GIRK channel.

EXPERIMENTAL PROCEDURES
Additional details on all methods are available in the supplemental methods.
cDNA Constructs and Electrophysiology-The cDNAs used in this study were obtained or prepared using standard PCRbased procedures. G1NC was constructed as described (12), with an 8-amino acid linker GSTASGST replacing the transmembrane segment (amino acids 85-184). Doubly labeled (DL)-GIRK1 and DL-GIRK2 were created by fusing CFP to the NT and YFP to the CT of the GIRK subunit, via linkers (see supplemental methods). Other cDNA constructs were as in Ref. 12. Xenopus oocytes were injected with RNAs, and whole-cell currents were measured using standard two-electrode voltage clamp procedures at 20 -22°C, in the ND96 (low K ϩ ) solution (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 solution (see Fig. 3, hK) (24 mM K ϩ , isotonically replacing NaCl in ND96) as described (12) (see supplemental methods). Differences in current amplitude, resulting from variations in GIRK expression, were corrected to channel expression, determined by the quantitation of YFP fluorescence as described (12). All currents were normalized to the control group.
Imaging and Spectral FRET Analysis-Fluorescent signals were collected with the Zeiss 510 META confocal microscope and analyzed as described (14), with modifications (see supplemental methods). Briefly, two spectra were collected from the animal hemisphere of each oocyte, with 405-nm (CFP excitation) and 514-nm (YFP excitation) laser lines. Net FRET signal of the CFP/YFP-labeled channels was calculated in the YFP emission range (with the 405 nm excitation) by consecutive subtraction of a scaled CFP-only spectrum (giving the A ratio parameter) and then of the ratio A 0 , which reports the direct excitation of YFP by the 405-nm laser. Therefore, Dynamic i-FRET-Ionic currents were recorded in ND96 solution at Ϫ80 mV (see supplemental methods). Oocytes were repetitively excited with the 405-nm laser every 40 s. An ϳ150 ϫ 100-m region of the membrane area was imaged. The CFP and YFP fluorescence was collected using 470 -500-and 505-550-nm band pass filters, respectively, and background signals were subtracted. YFP and CFP intensities at each point were normalized to an initial four measurements in each cell. This configuration allows leakage of CFP into the YFP recording window of ϳ0.30. Because this leakage is purely optical and constant irrespective of the FRET changes, we did not correct for it (15). The ratio of normalized intensities was denoted F YFP /F CFP (16). Some photobleaching occurred during repetitive excitations, but it had negligible effects on the F YFP / F CFP (see Fig. 4). Cells showing Ͼ15% bleaching of YFP or CFP were discarded.

GIRK-G␣-G␤␥ Triple Interactions in Vitro-
We constructed a cDNA encoding the complete cytosolic domain of GIRK1, G1NC, in which the NT and CT were fused whereas the transmembrane region was replaced by a short linker (Fig. 1A), to test its interaction with the G proteins. Similar constructs were previously shown to have a strong propensity to form stable tetramers in solution (17,18). The ivt 35 S-labeled G1NC gave a single protein band on SDS gel, as expected (Fig. 1B). Moreover, G1NC bound purified G␤ 1 ␥ 2 (12). We have recently reported (12) that G␤␥ enhances the interaction of G␣ i3 GDP with the proteins (normalized to that of G1NC to G␣ i3 GDP ) to GST-G␣ i3 , in the presence of either GDP or GTP␥S, with or without the addition of 3 g of purified G␤ 1 ␥ 2 . D, the interaction of G␣ i3 GA or G␣ i3 QL with purified G␤ 1 ␥ 2 , measured by Western blots as in E, in the presence of GDP or GTP␥S. E, the interaction of G1NC with G␣ i3 GA or G␣ i3 QL, in the presence of either GDP or GTP␥S, with or without G␤␥. The upper panel shows the summary of all experiments. Below the summary is a representative autoradiogram of pulled-down G1NC, and the bottommost panel shows Western blot of bound G␤. In D and E, in each experiment, binding of G␤␥ or G1NC was normalized to that measured with G␣ i3 QL GDP . Results are shown as mean Ϯ S.E. *, p Ͻ 0.05; **, p Ͻ 0.01, as compared with the control group.
G1NC. Here, we further characterized the interaction of GIRK1 with G␣ i3 GDP and also with G␣ i3 GTP . We similarly produced ivt CT of GIRK1 (G1C), the distal half of the CT (G1C 363-501 ) and YFP-fused NT of GIRK1, G1N YFP (Fig. 1A). A GST-fused G␣ i3 bound ivt G1NC as well as separate G1C and G1N YFP , whereas YFP or the distal CT (G1C 363-501 ) did not show detectable binding (Fig. 1, B and C). The latter observations confirm results obtained in a reciprocal configuration, with GST-fused channel parts and ivt G␣ i3 (9).
The addition of purified G␤ 1 ␥ 2 had no effect on the binding of GST-G␣ i3 to separate GIRK1 N and C termini, in the presence of either GDP or GTP␥S (Fig. 1B, summary in panel C). In contrast, the binding of G1NC to GST-G␣ i3 GDP was enhanced ϳ9-fold by the addition of G␤ 1 ␥ 2 , corroborating the previous report (12). Unexpectedly, G␤␥ also strongly enhanced the binding of G1NC to GST-G␣ i3 in the presence of GTP␥S (Fig. 1, B and C), a condition where GST-G␣ i3 poorly binds G␤ 1 ␥ 2 (supplemental Fig. S1). To alleviate the concern that a residual amount of G␣ i3 ␤␥ heterotrimers remaining in GTP␥S could bias the results, we constructed GST-fused versions of two well characterized G␣ i3 mutants that mimic G␣ i3 GTP and G␣ i3 GDP , respectively: constitutively active G␣ i3 Q204L (G␣ i3 QL) and constitutively inactive G␣ i3 G203A (G␣ i3 GA) (see Ref. 12). GST-G␣ i3 GA showed the expected strong interaction with G␤ 1 ␥ 2 in GDP and much less in GTP␥S, as reported (19). GST-G␣ i3 QL showed some interaction with G␤ 1 ␥ 2 in GDP (see also Ref. 20) but not in GTP␥S (Fig. 1D, example in panel E).
Both GST-G␣ i3 GA and GST-G␣ i3 QL bound G1NC in the absence of G␤␥ (Fig. 1E). G␤␥ enhanced the binding of both G␣ i3 GA and G␣ i3 QL to G1NC, in GDP as well as in GTP␥S, despite the great differences in their interaction with G␤␥ ( Fig.  1E). Thus, GIRK1 likely binds the active G␣ i3 (G␣ i3 GTP␥S and G␣ i3 QL) directly and not through G␤␥; by binding to GIRK1, G␤␥ enhances this interaction.
G Protein Subunits Induce Non-identical Conformational Rearrangements in GIRK1 and GIRK2-i-FRET was previously used to detect conformational changes in membrane proteins (16). We created a doubly labeled GIRK1 subunit (DL-G1) by fusing CFP (the donor) and YFP (the acceptor) to the extremes of NT and CT of GIRK1, respectively. DL-G1 was coexpressed with the neuronal GIRK2 subunit in Xenopus oocytes to give DL-G1/G2 channels, and visualized in the plasma membrane (PM) using a confocal microscope (Fig. 2). A linear relation of YFP to CFP fluorescence in DL-G1 confirmed the expected CFP/YFP molar ratio of 1:1 (supplemental Fig. S2C). We tested the functionality of all our tagged clones (Fig. 3) and found that DL-G1/G2 displayed similar function to wild-type G1/G2. The DL-G1/G2 channel had the characteristic basal activity (I basal ), was readily activated by acetylcholine (ACh) via a coexpressed muscarinic m2 receptor (m2R) and blocked by Ba 2ϩ (Fig. 3Aa). Moreover, DL-G1/G2 exhibited the typical inward rectification (Fig. 3Ab). Importantly, the regulation of the channel by coexpressed G␣ i3 GA, G␣ i3 QL and the G␤␥-scavenging protein m-phosducin, was also identical to that of the wild-type G1/G2 channel (Fig. 3B) (12). We found that G␣ i3 GA strongly reduced I basal and when coexpressed with G␤␥, G␣ i3 GA did not reduce the total G␤␥-dependent current, I ␤ ␥ (as also observed for the wild-type channel). Thus, G␣ i3 GA enhanced the relative acti-vation by coexpressed G␤␥ (R ␤ ␥) (supplemental Fig. S3A). G␣ i3 QL affected neither I basal , I ␤ ␥, nor R ␤ ␥. m-Phosducin reduced I basal but, in sharp contrast to G␣ i3 GA, also substantially diminished I ␤ ␥, acting as a typical G␤␥ scavenger ( Fig. 3B and supplemental Fig. S3A) (11,12).
We quantified the i-FRET signal using the spectral FRET technique (14) (Fig. 2C and supplemental Fig. S2E). Each oocyte was excited at two wavelengths, 405 and 514 nm, emission spectra were collected (Fig. 2, B and C, and supplemental Fig. S2E), and parameters characterizing bleed-through and energy transfer (ratios A 0 and A, respectively) were calculated. The extent of FRET is proportional to (A Ϫ A 0 ). Ratios A and A 0 were linear over a broad range of wavelengths, a testimony to the absence of optical and calculation artifacts (14) (Fig. 2D).

JOURNAL OF BIOLOGICAL CHEMISTRY 6181
Coexpression of G␤␥ with G␣ i3 -GA restored the active state with large I ␤ ␥ and with high i-FRET. In contrast, coexpression of G␤␥ did not reverse the m-phosducin-induced reduction in i-FRET and in I ␤ ␥ (Figs. 3B and 4A), supporting the differentiation of the actions of G␣ i3 GA from those of a simple G␤␥ scavenger (11). Expression of G␣ i3 QL did not significantly alter the basal i-FRET and currents, but further coexpression of G␤␥ led to a strong increase in i-FRET (ϳ30%), significantly greater than under all other conditions ( Fig. 4A;  Unlike GIRK1, G␤␥ does not enhance G␣ i3 -GIRK2 interaction, and G␣ i3 does not regulate I basal of homotetrameric GIRK2 (12). To assess whether G␣ i3 and G␤␥ confer conformational changes upon GIRK2 within the GIRK1/2 heterotetramer, we constructed a doubly labeled GIRK2 (DL-G2) as in GIRK1 (see "Experimental Procedures"). DL-G2 was expressed with wild-type GIRK1 to produce G1/DL-G2 heterotetramers, which showed adequate regulation by G␣ i3 and G␤␥ (supplemental Fig. S3B) and showed basal i-FRET of 0.25 Ϯ 0.02 (n ϭ 44). Similarly to DL-G1, i-FRET of DL-G2 was reduced by G␣ i3 GA (by ϳ20%), an effect reversed by G␤␥ (Fig. 4B). However, other parameters of i-FRET in the G1/DL-G2 heterotetramer were differently affected by the coexpressed G proteins. Coexpression of G␤␥ did not increase i-FRET above basal level either in the presence or in the absence of G␣ i3 . Most interestingly, concomitant expression of G␣ i3 QL with G␤␥ significantly reduced the basal FRET signal (by ϳ26%), opposite to its effect on DL-G1.
As a control, we engineered a doubly labeled IRK1 (Kir2.1) channel, DL-IRK1 ( Fig. 3C and supplemental Fig. S5). This inwardly rectifying K ϩ channel does not directly interact with, and is not regulated by, G proteins (21). DL-IRK1 exhibited PM localization, constitutively active gating, the typical strong inward rectification (Fig. 3, C and D) and strong basal i-FRET ((A Ϫ A 0 ) ϭ 0.35 Ϯ 0.01). Coexpression of G␤␥, G␣ i3 GA, G␣ i3 QL, or m-phosducin caused no significant changes in i-FRET (Fig. 4C) or in channel currents. These results confirm the specificity of G protein regulation of conformational states of GIRK1/2 by G protein subunits, revealed by i-FRET.
Dynamic Structural Rearrangements in GIRK1/2 upon GPCR Activation-The changes in i-FRET caused by the presence of G protein subunits probably reflect the conformations of the channel at different activation steps. To verify that these static measurements reflect the conformations adopted by the channel during its physiological activation process, we monitored dynamic changes in i-FRET caused by activation of a GPCR (m2R). In these experiments, the electrophysiological response (K ϩ current) and fluorescence signals were collected simultaneously from individual oocytes (Fig. 5A).
Oocytes expressing DL-GIRK1/2, m2R, and the wild-type G␣ i3 were voltage-clamped at Ϫ80 mV and continuously perfused with ND96 solution. Concomitantly, a region of  the oocyte membrane was excited with the excitation wavelength of the donor (405 nm) every 40 s. The emission signals of both donor and acceptor were recorded separately, and the F YFP /F CFP ratio was calculated. This ratiometric imaging method allows the monitoring of changes in the relative position of donor and acceptor fluorophores. For instance, reduction of the distance between donor and acceptor increases the energy transfer, causing an increase in YFP emission intensity concurrently with a decrease in CFP emission intensity and resulting in an increase in F YFP /F CFP ratio (16).
Under the recording conditions used, DL-G1/G2 showed considerable basal activity (11), corresponding to a mixture of "preactivated" states (see "Discussion"). The addition of 10 M ACh (Fig. 5B, gray area) caused additional activation of GIRK (Fig. 5B, top red trace) and a concomitant increase in F YFP /F CFP ratio (Fig. 5B). When ACh was washed from the system, the channel reassumed its basal, preactivated F YFP /F CFP ratio and basal activity. These dynamic changes match our static i-FRET data (see "Discussion"). No change in F YFP /F CFP was seen in control experiments, where oocytes were exposed to 30 M atropine, which completely inhibited the ACh-evoked K ϩ current (Fig. 5C). Importantly, pretreatment with pertussis toxin (PTX) completely inhibited both ACh-evoked currents and the ACh-induced change in i-FRET (Fig. 5D). The use of atropine rules out the involvement of the receptor in the induction of the change in conformation of the GIRK, and the latter is most probably caused by the direct interaction of the G proteins themselves. Finally, ACh did not induce any ionic currents or changes in F YFP /F CFP in DL-IRK1 channel (Fig. 5E).

G␤␥ Is Crucial for Strong Interaction between GIRK1 and G␣ i -
Our biochemical results strongly support the idea of a preformed signaling complex of heterotrimeric G␣ i ␤␥ with GIRK1, initially based upon the finding of a strong interaction of G␣ i1 GDP ␤␥ with the NT of GIRK1 in vitro (5). Indeed, G␤␥ greatly enhanced the interaction of GIRK1 with G␣ i3 GDP under conditions favoring the formation of G␣␤␥ heterotrimers ( Fig.  1) (12). Somewhat at odds with Huang et al. (5), our data suggest that both NT and CT are necessary for the formation of the strong GIRK1-G␣ i ␤␥ complex as the effect of G␤␥ was not present in separate N and C termini. Further quantitative binding studies may be needed to understand the reason for this discrepancy.
A novel and unexpected finding was that G␤␥ also enhanced the interaction of GIRK1 with G␣ i3 and the constitutively active G␣ i3 QL in GTP␥S, a condition when no G␣ i3 QL-G␤␥ interaction was detected by pulldown (supplemental Fig. S1). Therefore, we suggest that the conformational change in GIRK1 induced by the binding of G␤␥ allosterically improves the direct binding of G␣ i3 GTP to GIRK1. Alternatively, GIRK1 may still bind G␣ i3 GTP via G␤␥, utilizing a second G␤␥-binding site on G␣ i3 , distinct from the classical high affinity interaction site (22). In all, the biochemical data support the persistence of GIRK-G␣ i/o -G␤␥ signaling complexes, both before and after activation by agonist-bound GPCR. The continuous attachment of G␣ i to GIRK1/2 would ensure a high local concentration of G␣ and a diffusion-independent reassociation with G␤␥, which allows for fast termination of the physiological response upon removal of the agonist because of the fast kinetics of G␣ GDP -G␤␥ binding (23).
i-FRET and GIRK Conformation-The available partial crystal structures of GIRK tetramers indicate proximity between the NT of one subunit to the mid-CT of an adjacent subunit (17,18,24). In a tetrameric DL-G1/G2 channel, i-FRET could potentially arise from proximity of CFP-NT of one DL-G1 subunit to the CT-YFP of an adjacent DL-G1. This would require a 1-1-2-2 arrangement of subunits in the tetramer, which is probably viable, by analogy with 1-1-4-4 in GIRK1/GIRK4, although 1-4-1-4 is preferable (25,26). However, it is important to note that the proximal NTs (ϳ40 amino acids) and the distal CT (ϳ130 amino acids in GIRK1, amino acids 370 -501; ϳ45 amino acids in GIRK2) are omitted from the above crystal structures. The unique distal CT segment of GIRK1 is essential for the strong triple GIRK-G␣ i -G␤␥ interaction and for the differential regulation of GIRK1 and GIRK2 by G␣ i and G␤␥ (12). The flexible proximal NT and distal CT are probably long enough to allow for intrasubunit interactions. Such intrasubunit contacts were proposed to play an essential role in GIRK gating by G␤␥ (27). Therefore, we interpret an increase in i-FRET as nearing of the NT and CT of either the same or adjacent GIRK subunits. The exact structural rearrangements remain to be determined. However, it is certain that changes in i-FRET reflect G protein-induced changes in conformation of the channel (Figs. 3C and 5B).
G Protein-regulated Conformational States of the GIRK Channel-Measurements of single-channel or population currents in native and chimeric GIRK channels demonstrated that heterotetrameric GIRK channels assume a number of closed and open conformations with a variable number of G␤␥ molecules bound, from a closed, non-conducting state (interpreted as G␤␥-devoid), via intermediate states with one, two, or three G␤␥ molecules bound and correspondingly increasing open probability (P o ), to a fully G␤␥-activated state with four G␤␥ bound and the highest P o (27)(28)(29)(30). Analysis of i-FRET, when combined with biochemical and functional assays, extends our understanding of the rules and modes of G protein-GIRK interactions and of the regulation of channel conformation. A simplified schematic of some of the interactions and conformations is presented in Fig. 6.
Preactivated (resting) states of GIRK1/2 probably reflect a mixture of conformations of GIRK1/2 channels, associated with a variable number of G␣ i ␤␥ heterotrimers and/or G␤␥ that lack matching G␣ i GDP (Fig. 6, b and d), as suggested by functional data (11). Correspondingly, DL-G1/G2 and G1/DL-G2 expressed alone at high levels yield high I basal (10) and an intermediate i-FRET signal. A substantial basal GIRK activity has been reported in hippocampal and cortical neurons (31,32); therefore, the preactivated states are physiologically relevant.
Overexpression of high doses of the G protein subunits or G␤␥ scavengers shifts the channel population toward more homogenous states, some of which are detected by our i-FRET assay. 1) The first is G␤␥-free state, with low I basal (Fig. 6a). This state is promoted by m-phosducin, which removes G␤␥ away from the channel (6,11). It is characterized by low i-FRET and is depicted in Fig. 6 as having a large distance between NT and CT of GIRK1. 2) The second is G␤␥-activated state(s), with high i-FRET and GIRK currents, presumably with several (up to four in the full tetramer) G␤␥ dimers bound at activation sites (Fig. 6, d and e). 3) Joint expression of G␤␥ and G␣ i3 QL confers a conformation (Fig. 6c) characterized by the highest i-FRET in GIRK1 but the lowest i-FRET in GIRK2 (Fig. 3G). These opposite changes in i-FRET further point toward a difference in regulation of GIRK1 and GIRK2 subunits by G␣ i and G␤␥ (12). Neither G␣ i QL nor G␤␥ alone confer such changes in i-FRET, indicating concomitant binding of G␣ i3 GTP and G␤␥ to the heterotetrameric GIRK1/2. Cells overexpressing G␤␥, G␤␥ϩG␣ i GA, or G␤␥ϩG␣ i QL show high constitutive, non-desensitizing GIRK currents (11,12), clearly reporting the open states of the channel. Despite the different relative positions of GIRK N and C termini, these various open state(s) give similar whole-cell currents (but it remains to be seen whether they have distinct single-channel properties). The overall conformation of the channel in the presence of ACh should resemble that seen in static FRET experiments upon coexpression of G␤␥ or, even more closely, G␤␥ϩG␣ i3 QL. Accordingly, dynamic activation of GIRK by ACh via coexpressed m2R caused an increase in both GIRK current and i-FRET in DL-G1/G2. The increase in i-FRET most probably corresponds to the opening of the channel because GIRK channels expressed in Xenopus oocytes show little AChinduced desensitization over time periods of several minutes (33). Control experiments showed no changes in i-FRET upon exposure to the muscarinic antagonist atropine and after treatment with pertussis toxin, which prevents the activation of G␣ i by GPCRs. Finally, no G protein, or GPCR, changes in i-FRET were observed in the double-labeled G␤␥-insensitive IRK1 channel. These control experiments demonstrate the authenticity and specificity of i-FRET changes caused by G␣ i and G␤␥ in GIRK1/2.
The low i-FRET and small GIRK currents observed after the expression of G␣ i3 GA are indistinguishable from those seen in G␤␥-free channels in the presence of m-phosducin, as if G␣ i3 GA removed G␤␥ away from the channel. However, given the compelling evidence for a strong triple association between GIRK1, G␣ i GDP and G␤␥ ( Fig. 1) (5-7, 34), we hypothesize that the G␣ i3 GA-G␤␥ heterotrimers, formed after expression of G␣ i GA, remain mostly associated with GIRK1/2. In support, we observe specific differences in the effects of G␣ i3 GA (but not the QL mutant) on i-FRET and the total GIRK current when coexpressed with G␤␥. This corroborates the hypothesis that it is the GDP-bound G␣ i (or G␣ i ␤␥ heterotrimer) that regulates the basal activity of GIRK and "primes" the channel for activation by G␤␥ (11,12).
The new biochemical and i-FRET results, as well as previous functional observations, are compatible with a "two-site model" in which GIRK possesses an anchoring site for G␣ i/o (and/or G␣ i/o ␤␥) and a separate activation site, where the binding of G␤␥ leads to channel opening (12) (Fig. 6). The assumption of separate binding sites for G␣ i (or G␣␤␥) and G␤␥ is supported, although not proved, by the absence of competition among G␣ i and G␤␥ for binding to G1NC and G2NC under conditions that favor either the formation or the dissociation of G␣ i ␤␥ heterotrimers and by the more-than-additive character of i-FRET Distances between NT and CT are depicted schematically, where high i-FRET is interpreted to indicate proximity between the N and C termini (of the same or of adjacent subunits). Inversely, low i-FRET correlates with distancing of the termini. ϩphos., m-phosducin. changes induced by G␣ i QL and G␤␥. The two sites may lie in proximity (or even partly overlap). According to the two-site model, the high basal activity of overexpressed GIRK1/2 is due to an excess of G␤␥ associated with the channel, over G␣ i (11,12,35). Under these conditions, some channels have G␤␥ "uncompensated" by G␣ and thus bound to the activation sites (Fig. 6d). The coexpressed G␣ i3 GA associates with G␤␥ subunits, obstructing their interaction with the activation sites (hence the decrease in I basal and i-FRET). The newly formed G␣ i3 ␤␥ heterotrimers are docked at the anchoring site (Fig. 6b). When more G␤␥ is coexpressed, free G␤␥ subunits bind at the activation sites, hence the high i-FRET and currents (Fig. 6e). The two-site hypothesis provides an economical description of a normal physiological situation; upon activation of GPCR, G␤␥ detaches from G␣ i and interacts with the activation site, whereas G␣ i GTP stays at the anchoring site (Fig. 6c). (Alternatively, G␤␥ may stay at the same site, and it is G␣ i that shifts to another location. Also, G␣ i and G␤␥ do not have to fully disengage (16, 36); a rearrangement followed by an exposure of a few crucial residues on G␤␥ may be sufficient). Once GTP is hydrolyzed by G␣, G␤␥ rebinds G␣ and becomes docked again.
Conclusions-Coordinated biochemical, functional, and i-FRET data support the existence of a persistent, dynamic GIRK1/2-G␣ i/o -G␤␥ complex, both before and after activation by agonist-bound GPCRs. We propose that GIRK1 acts as the nucleator of the complex and that both active and inactive G␣ i3 remain bound to the channel, ensuring fast and specific activation and termination of the signal. New in vitro and in vivo data reveal intricate triple interactions among GIRK1, G␣ i3 , and G␤␥ and show that the presence of G␤␥ is crucial to ensure strong GIRK1-G␣ i interactions. At least some of the effects of G␤␥ on GIRK1-G␣ i interaction appear to be allosteric. G␣ i and G␤␥ induce mutually dependent conformational rearrangements in GIRK1/2, characterized by distinct changes in i-FRET in the GIRK subunits and often, but not always, by changes in channel activity.