Two Distinct Aspects of Coupling between Gαi Protein and G Protein-activated K+ Channel (GIRK) Revealed by Fluorescently Labeled Gαi3 Protein Subunits*

G protein-activated K+ channels (Kir3 or GIRK) are activated by direct interaction with Gβγ. Gα is essential for specific signaling and regulates basal activity of GIRK (Ibasal) and kinetics of the response elicited by activation by G protein-coupled receptors (Ievoked). These regulations are believed to occur within a GIRK-Gα-Gβγ signaling complex. Fluorescent energy resonance transfer (FRET) studies showed strong GIRK-Gβγ interactions but yielded controversial results regarding the GIRK-Gαi/o interaction. We investigated the mechanisms of regulation of GIRK by Gαi/o using wild-type Gαi3 (Gαi3WT) and Gαi3 labeled at three different positions with fluorescent proteins, CFP or YFP (xFP). Gαi3xFP proteins bound the cytosolic domain of GIRK1 and interacted with Gβγ in a guanine nucleotide-dependent manner. However, only an N-terminally labeled, myristoylated Gαi3xFP (Gαi3NT) closely mimicked all aspects of Gαi3WT regulation except for a weaker regulation of Ibasal. Gαi3 labeled with YFP within the Gα helical domain preserved regulation of Ibasal but failed to restore fast Ievoked. Titrated expression of Gαi3NT and Gαi3WT confirmed that regulation of Ibasal and of the kinetics of Ievoked of GIRK1/2 are independent functions of Gαi. FRET and direct biochemical measurements indicated much stronger interaction between GIRK1 and Gβγ than between GIRK1 and Gαi3. Thus, Gαi/oβγ heterotrimer may be attached to GIRK primarily via Gβγ within the signaling complex. Our findings support the notion that Gαi/o actively regulates GIRK. Although regulation of Ibasal is a function of GαiGDP, our new findings indicate that regulation of kinetics of Ievoked is mediated by GαiGTP.

G protein-coupled receptor (GPCR) 3 signaling is one the most important cellular signaling cascades as witnessed by the hundreds (ϳ800) of genes in the human genome (1), with some cells expressing as many as a hundred different receptor types (2). An agonist-bound GPCR initiates GDP-GTP exchange at the G␣ subunit, which serves as the "on-off" switch of the GPCR-initiated process and as the trigger for the release of the G␤␥ dimer. Both the active GTP-bound G␣ subunit (G␣ GTP ) and G␤␥ directly interact and regulate various effector molecules such as phospholipase C␤, adenylyl cyclase, and other enzymes, as well as a number of ion channels (1,(3)(4)(5)(6). In particular, the G protein-activated inwardly rectifying K ϩ (GIRK) channel is the archetypal G␤␥ effector activated by direct binding of G␤␥ (7). GIRK is an important mediator of inhibitory actions of G i/o -coupled GPCRs, notably the vagal inhibition of cardiac pacing and of the inhibitory actions of a large number of neurotransmitters in the brain (8 -10).
It is becoming clear that GIRK, the prototypical G␤␥ effector (11), is also an effector of the G␣ subunit, which appears to play multiple roles in modulation of GIRK activity. G␣ i controls the gating of GIRK by acting as a servo-type regulator, operating both as a damper of persistent inhibitory signaling by reducing the basal activity of GIRK (I basal ) and, when activated by GPCR, as the generator of the main inhibitory signal, the evoked current of GIRK (I evoked ) (12,13). Several lines of evidence point to an active modulation of GIRK both by the nonactivated GDPbound form of G␣ i3 (G␣ i3 GDP ) (12,13) and by the activated G␣ i GTP (14 -16), but the details and mechanism of G␣ action are poorly understood. In line with the notion that G␣ i might affect channel behavior are recent findings suggesting that the GIRK channel serves as a multiprotein scaffold in signaling complexes with G␤␥, G␣ i/o , regulators of G protein signaling, and possibly certain GPCRs (12, 16 -21). This multiprotein complex is believed to enable specific and fast interactions of many protein partners, such as G␣ i and GIRK, thus enabling them to modulate one another. However, the composition of such complexes and the roles of individual protein partners are yet to be determined (9,22).
To assess modes of interactions between G␣ i3 and GIRK within the hypothetical signaling complex, spectroscopic approaches appear particularly attractive. Tagging with fluorescent proteins (xFPs; supplemental Fig. S1A) allows both detection of the tagged protein and assessment of its interac-tion in vivo with other proteins using resonance energy transfer methods, mainly Förster RET (FRET) (23) or bioluminescence RET (24). However, fusion of xFP to the G␣ subunit has proved challenging. Only a handful of positions have been reported to tolerate insertion of xFPs to produce functional G␣ i/o proteins (25)(26)(27)(28)(29). Even when normal function in one or several aspects is reported, it is often debated whether other actions are preserved after xFP tagging (25,28). Another notable controversy, relevant to the question of GIRK-related signaling complex, pertains to the direct G␣ i/o -GIRK interaction in vivo. A measurable bioluminescence resonance energy transfer between GIRK1-GFP and G␣ i1 -RLuc (fused with luciferase) was interpreted to suggest a stable GIRK-G␣ i1 signaling complex, probably within the endoplasmic reticulum (21). However, no measurable FRET between GIRK2 and G␣ o could be detected (20). These discrepancies could arise from inadequate functioning of some xFP-fused GIRKs or G␣ i and stress the need for their rigorous testing.
Here, we investigated the functional coupling between G␣ i and GIRK using three distinct G␣ i3 constructs (G␣ i3 xFPs) labeled with cyan or yellow fluorescent proteins (CFP or YFP, respectively). We have found that xFP labeling of G␣ i , even at positions widely regarded as producing functional G␣ proteins, may impede G␣-effector interactions and regulations. Furthermore, none of the G␣ i3 xFPs, even the one that best mimicked the effects of G␣ i3 WT, produced a measurable FRET signal with GIRK1, contrasting the strong FRET between GIRK1 and G␤␥. Taken together with biochemical data, these results suggested that G␣ i3 associates with GIRK primarily via G␤␥. Importantly, the use of G␣ i3 xFPs revealed differential regulation of basal and evoked GIRK activities by G␣, which was confirmed by titrated expression of G␣ i3 WT. We conclude that control of the basal activity of GIRK and the fast provision of "free" G␤␥ for GIRK activation (following GPCR activation) are distinct functions of the G␣ i subunit in this signaling cascade.

EXPERIMENTAL PROCEDURES
cDNA Constructs-The cDNAs used in this study were obtained or prepared using standard PCR-based procedures. All cDNA constructs were inserted in two vector types as follows: pGEX (for production of GST fusion proteins in Escherichia coli) or into high expression oocyte vectors containing 5Ј-and 3Ј-untranslated sequences of Xenopus ␤-globin: pGEMHE, or its derivative pGEMHJ, or pBS-MXT. Rat GIRK1 (U01071), mouse GIRK2-1 (U11859), and human G␣ i3 (J03198) were used for oocyte expression and as the basis for modifications. The cDNA for GST-G1NC was kindly provided by Craig A. Doupnik, University of South Florida. GST-G1NC was constructed with a His 6 linker replacing the transmembrane segment (amino acids 85-184). GST was fused to its N terminus via a 6-amino acid linker (Leu, Val, Pro, Arg, Gly, and Ser). GST-G␣ i3 was prepared as described previously (31). G1NC for in vitro translation was prepared as described previously (16); its transmembrane segment (amino acids 85-184) was deleted and replaced by an 8-amino acid linker GSTASGST. GIRK1 CFP was created by replacing its stop codon with an XbaI restriction site. The coding sequence of CFP was inserted between the XbaI site and a HindIII site immediately following the stop codon of the CFP. YFP-labeled G␣ i3 constructs were obtained or constructed as follows. G␣ i3 117, where YFP was inserted in the ␣b-␣c loop, with the first Met of YFP placed at position 117 of G␣ i3 , was kindly provided by Alfred G. Gilman (28). Myr-YFP-G␣ i3 was designed similarly to Ref. 27, but instead of a 20-amino acid GAP43 palmitoylation signal, it started with a 17-amino acid myristoylation signal derived from the Src protein, as described previously (32), followed by the first Met of CFP or YFP and then by G␣ i3 (the stop codon of xFP was removed). G␣ i3 -CT-YFP was constructed like GIRK1 CFP , where the G␣ stop codon was removed and XbaI restriction site was added instead. YFP was inserted immediately after the G␣ i3 insert and terminated with a stop codon and a HindIII restriction site. PTX-insensitive G␣ i3 s were created using PCR, by mutating the cysteine at position 351 (4th position from the C terminus) to isoleucine, C351I.
Experimental System and Electrophysiology-Experiments were approved by the Tel Aviv University Institutional Animal Care and Use Committee (permit no. 11-05-064). Xenopus oocytes were prepared and injected with RNA, as described previously (12), and incubated for 3-5 days at 20 -22°C in ND96 solution (96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 5 mM HEPES, pH 7.5), supplemented with 2.5 mM sodium pyruvate, 100 g/ml streptomycin, and 62.75 g/ml penicillin. Whole-cell currents were measured using standard two-electrode voltage clamp procedures at 20 -22°C, in the ND96 (low K ϩ ) solution or in a high K ϩ solutions (24 mM K ϩ , isotonically replacing NaCl in ND96) as described previously (16). When designed, protomer A of pertussis toxin (PTX) was injected into oocytes 2-20 h before current recordings (33). Currents were recorded at Ϫ80 mV, filtered at 500 Hz, and sampled at 5 or 10 kHz. Data acquisition and analysis were done using pCLAMP (Molecular Devices). Calculation of activation time constant ( act ) was done using a standard mono-exponential fit using the Clampfit program of the pCLAMP suite.
Pulldown Assays-GST-fused proteins were purified using standard protocols (31). [ 35 S]Methionine-labeled proteins were synthesized in rabbit reticulocyte lysate. For pulldown experiments, the following procedure was used: G␣ i3 (when present) was incubated at 30°C for 20 min in a total volume of 50 ml of high-K ϩ binding buffer containing 33 M of either GDP or GTP␥S and 0.01% Lubrol (31). Next, G␤ 1 ␥ 2 was added for another 20 min. The full cytosolic domain of GIRK1 (G1NC) was then added, and the total reaction volume was brought to 300 l (with binding buffer containing 30 M GDP or GTP␥S), and the incubation was continued for 1 h at room temperature. Then 5 l were removed and used to measure the loaded protein ("input"). Binding to glutathione-Sepharose beads and elution with 15 mM glutathione were done as described previously (31). The eluted proteins were separated on 12% SDS-polyacrylamide gels. Using this procedure, different pulldown experiments were done as follows: 3 g of purified histidinetagged G␤ 1 ␥ 2 was used to pull down 5 l of reticulocyte lysate containing 35 S-labeled G␣ i3 xFP proteins or 5 l of YFP alone ( Fig. 2; no 35 S-labeled G1NC was added in this experiment). 3 g of purified GST-fused G1NC was used to pull down 5 l of reticulocyte lysate containing 35 S-labeled G␣ i3 WT and 5 l of G␤ 1 ␥ 2 (Fig. 3). Autoradiograms were obtained by imaging and quantitating the dried gels using PhosphorImager and the software ImageQuaNT (GE Healthcare). Western blots were performed using G␤ antibody (Santa Cruz Biotechnology) and ECL reagents from Pierce.
Imaging Analysis and FRET-Fluorescent signals were collected with the Zeiss 510 META confocal microscope. Immunolabeling of GIRK1 in giant excised PM patches was done using GIRK1 antibody as described previously (12,13,34). In brief, the vitelline membrane was peeled off, and the oocytes were placed on glass coverslips. After sticking to the coverslip, the oocyte was removed; pieces of membranes attached to the coverslip, with their cytosolic leaflet surface exposed to the external solution, were washed and fixated, and nonspecific sites were blocked with donkey immunoglobulin G (IgG, whole molecule, 1:400, Jackson ImmunoResearch), and coverslips were incubated with GIRK1 antibody (Alomone Labs, Jerusalem), followed by incubation with secondary antibody (CY3 donkey anti-rabbit IgG, 1:400, Jackson ImmunoResearch). The CY3 fluorescence was imaged by exciting the dye at 514 nm; the emitted light was collected between 540 and 615 nm using the spectral mode of the Zeiss 510 Meta (beam splitter HFT 405/514/633).
Intact oocytes were imaged in ND96 solution in a 0.7-mm glass-bottom dish, as described previously (16). Fluorescent signals were collected from circular regions of interest of ϳ75 pixels in the membrane area, close to the midline of the animal hemisphere, and from three background regions of interest outside the oocyte using a 20ϫ air objective. CFP was excited using a 405 nm laser line, and emission spectrum was collected using the META detector. Peak emission (481-492-nm band) was chosen for the comparison of expression levels of CFPlabeled proteins. YFP was excited using a 514 nm, and peak emission (524 -535 nm) was used for comparison of expression levels.
FRET experiments were performed as described previously (16,35). 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 report the direct excitation of YFP by the 405 nm laser, as in Equations 1 and 2, Because of the use of different experimental settings in independent experiments, the fluorescence intensities cannot be used to construct FRET titration curves. We imaged a doublylabeled protein expressing both CFP and YFP at a 1:1 stoichiometry (YFP-IRK1-CFP, DL-IRK1(16)) in each experiment to convert the fluorescence of CFP and YFP into their molar ratio (see supplemental Fig. S6). This was achieved by calculating the slope of the CFP to YFP fluorescence of DL-IRK1. Then this slope was used as a reference to convert CFP and YFP fluorescence, in the other groups of the same experiment, into a molar ratio, which can be collected and used from many experiments and implemented into one titration curve. Calculation of E app and distances between fluorophores was done as described previously (36,37).
Statistical Analysis-Results are shown as mean Ϯ S.E. Multiple group comparison was done using one-way analysis of variance (ANOVA) followed by Tukey all-pairwise analysis. Two group comparisons were done using two-tailed t test or paired t test when applicable. Correlations between two parameters were examined using Spearman test. Asterisks indicate statistically significant differences as follows: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.

RESULTS
To address GIRK-G␣ i interactions in vivo using optical methods, we constructed xFP-labeled GIRK1 and G␣ i3 and compared with their wild-type counterparts. Functional tests were done in Xenopus oocytes, where the regulation of GIRK by GPCRs, G␤␥, and G␣ has been extensively characterized. GIRK currents were recorded under voltage clamp at Ϫ80 mV. The neuronal GIRK1/2 (heterotetramer composed of GIRK1 and GIRK2 subunits) shows a sizeable basal activity (I basal ), especially at high levels of channel expression, presumably because of excess of associated G␤␥ and/or lack of sufficient channelassociated G␣ i/o (32). I basal is revealed by exchanging a physiological, low-K ϩ (ND96; 2 mM) solution to a high-K ϩ ("hK"; 24 mM) solution (Fig. 1A). When coexpressed with the muscarinic receptor 2 (m2R), GIRK is readily responsive to activation by agonist (acetylcholine; ACh). In the absence of coexpressed G␣, the ACh-evoked current (I evoked or I ACh ) reports the activation of endogenous G␣ i/o of the oocyte (Fig. 1A, left panel) (38).
I basal is largely G␤␥-dependent and is strongly sensitive to coexpressed G␣ i/o , which reduces I basal in a dose-dependent manner and concomitantly increases I evoked , thereby increasing the ratio of activation, R a (R a ϭ I total /I basal , where I total ϭ I basal ϩ I evoked ) ( Fig. 1, A-C). The reduction in I basal is a function of G␣ i GDP and reflects the formation of G␣␤␥ heterotrimers, which predisposes ("primes") the channel for subsequent activation by the GPCR (12,39). Among several G␣ i/o subunits tested, we have chosen G␣ i3 as the preferred donor of G␤␥ and regulator of I basal . Coexpressed G␣ i3 provides for fast kinetics of agonist response (Figs. 1A, inset, and 4), accelerating its time constant ( act ) compared with endogenous G␣ i/o (40). Notably, G␣ i3 WT does not reduce the total GIRK current (Fig. 1B), as opposed to G␤␥-"scavenging" proteins (39). In fact, I total may actually increase following moderate expression of G␣ i3 WT (40). I basal , I evoked , I total , R a , and act have been chosen as the essential set of parameters to assess the functional coupling of xFP-labeled G␣ i3 and GIRK1/2.
Fluorescent Labeling of GIRK1 Does Not Perturb Regulation by G␣ i3 -xFP-labeled GIRK channels have been found functional with respect to trafficking to the PM and regulation by GPCRs and G␤␥ (13,16,17,20,21,(41)(42)(43), but regulation by G␣ i/o has not been tested. We examined the regulation by G␣ i3 of heterotetrameric GIRK1/2 formed by GIRK1 labeled with CFP or YFP at the C terminus and the wild-type GIRK2 (GIRK2WT). The resulting GIRK1 xFP /2 showed fast activation by ACh and inward rectification like the GIRK1/2WT (supplemental Fig. S2). The GIRK1 xFP /2 channel also exhibited proper PM targeting and expression, comparable with the WT GIRK1/2 channel, as detected by antibody labeling of GIRK1 in giant excised plasma membrane patches (Fig. 1D, top panel) (34). Expression of GIRK1 YFP was also monitored in whole intact oocytes (Fig. 1D, bottom panel). Like the GIRK1/2WT, the GIRK1 xFP /2 showed dose-dependent regulation by G␣ i3 WT; I basal was reduced, and I evoked was concomitantly increased, and I total slightly increased or remained unchanged ( Fig. 1, A and B). As described for GIRK1/2WT (12), positive correlation between the increase in the dose of injected G␣ i3 RNA and subsequent increase in R a was observed (Fig. 1C). Overexpression of G␣ i3 WT and/or G␤␥ did not significantly affect the PM levels of GIRK1 xFP /2 ( Fig. 1, E and F). In all, GIRK1 xFP /2 is properly regulated by G␣ i3 WT.
YFP-labeled G␣ i3 Proteins Bind GIRK1 and G␤␥ but Appear Deficient in a Triple GIRK1-G␣-G␤␥ Interaction Test-We next tested three different xFP-G␣ i3 constructs as follows: Nand C-terminally labeled (G␣ i3 NT and G␣ i3 CT, respectively) and an internally labeled one (G␣ i3 117). G␣ i3 NT has been constructed to start with a 15-amino acid sequence containing a myristoylation signal, followed by CFP or YFP and then the full-length G␣ i3 . N-terminally lipid-modified and xFP-labeled G␣ i subunits have been reported to successfully couple GPCRs to GIRK, reducing I basal and eliciting I evoked (27). G␣ i3 117, in which YFP was inserted in the ␣b-␣c loop of G␣ i3 (with its methionine starting at amino acid 117) of G␣ i3 via a two-amino acid linker, showed proper uncatalyzed GDP-GTP exchange in vitro and an ␣ 2 -adrenergic receptor-induced dissociation (rear-rangement) from G␤␥ in mammalian cells (28). G␣ i3 CT is a novel construct, in which YFP is fused to the end of the C terminus of G␣ i3 via a two-amino acid linker.
We first examined the interaction of G␣ i3 xFP proteins with G␤␥. 35 S-Labeled G␣ i3 YFP proteins, synthesized in vitro in reticulocyte lysate, were pulled down by purified histidinetagged G␤ 1 ␥ 2 (His 6 G␤␥) in the presence of GDP or GTP␥S. In the presence of GDP, all three G␣ i3 YFP-bound His 6 G␤␥ at least as well as G␣ i3 WT, whereas 35 S-labeled YFP did not bind His 6 G␤␥ (Fig. 2, A and B). As expected, G␣ i3 -G␤␥ binding was significantly reduced in the presence of GTP␥S (Fig. 2, A and C) (44). The residual binding could be attributed to the presence of a second nucleotide-independent G␤␥-binding site in G␣ i (45), but this issue was not pursued. In all, although semiquantitative, this assay showed the expected guanine nucleotide-dependent interaction of all three G␣ i3 YFP constructs with G␤␥.
GIRK subunits bind both G␤␥ and G␣ i (18, 19, 40, 46 -49). Furthermore, G␤␥ enhances the interaction of G␣ i3 with the full cytosolic domain of GIRK1 (13,16). Although the exact functional correlate of this effect is not known, we hypothesized that integrity of triple G␣-G␤␥-GIRK interactions within the presumptive signaling complex may be an important factor in proper GIRK regulation. To examine whether the G␣ i3 YFP proteins interact with GIRK1 and how this regulation is affected by G␤␥, we used a purified GST-fused full cytosolic domain of GIRK1 containing both the N-and C-terminal cytoplasmic domains but lacking its transmembrane region, GST-G1NC. GST-G1NC properly bound 35 S-labeled, in vitro trans-  (1 ng), and m2R (0.5 ng). Membrane potential was Ϫ80 mV. Switching from a physiological low K ϩ solution (2 mM K ϩ ; ND96) to a high K ϩ solution (24 mM K ϩ ; hK) reveals the basal activity of the channel (I basal ). Addition of the agonist ACh (10 M) shows the evoked current (I ACh ). Ba 2ϩ (5 mM) was then added to inhibit GIRK currents, allowing us to calculate the net I basal and I total . Right trace, coexpression of G␣ i3 WT (5 ng) reduces I basal and enhances I ACh . Inset, zoom on the activation phase of I ACh . Monoexponential fit (black line) used to calculate the activation time constant ( act ) is shown superimposed on the actual trace (gray line). B, increasing amounts of G␣ i3 WT significantly reduces I basal of both GIRK1WT/2 and GIRK1 CFP /2, and enhances I ACh without reducing I total , hence causing a significant increase in the relative extent of activation, R a (C). D, GIRK1WT/2 and GIRK1 YFP /2 channels express at comparable levels in the plasma membrane. PM levels of GIRK1 and GIRK1 YFP were assessed using the giant excised membrane patches method (top images), and GIRK1 YFP was also visualized with YFP fluorescence in intact oocytes (bottom images). No GIRK1 could be detected by any of these methods in native oocytes. E, summary of the expression level of both channels. Coexpression of G␤␥ does not cause any significant change in surface expression levels of GIRK1WT/2 and GIRK1 xFP /2. F, coexpression of G␤␥ or G␣ i3 WT does not cause any significant changes in GIRK1 YFP PM expression level. In all figures, number of oocytes is shown within or above the bars. *, p Ͻ 0.05; **, p Ͻ 0.01 compared with the control group; a.u., arbitrary units. lated (ivt) 35 S-G␣ i3 WT and G␤␥, in the presence of either GDP or GTP␥S (Fig. 3A). Notably, the binding of GST-GIRK1 to G␤␥ was 4 -5-fold stronger than to G␣ i3 WT, as indicated by the smaller percentage of bound versus added protein (Fig. 3B). Furthermore, the binding of 35 S-G␣ i3 WT to GST-G1NC was enhanced in the presence of purified G␤␥, ϳ2.5-fold in GDP and ϳ2-fold in GTP␥S (Fig. 3, C and E). These data corroborate our previous report that used a reciprocal protocol, with purified GST-G␣ i3 and in vitro synthesized 35 S-G1NC (16).
Next, we tested the interaction of GST-G1NC with G␣ i3 YFP proteins and its modulation by G␤␥. GST-G1NC bound all three 35 S-G␣ i3 YFP constructs at least as well as G␣ i3 WT, and it did not bind ivt 35 S-YFP (Fig. 3, D and E, supplemental Fig. S3).
In the presence of GDP, all of the G␣ i3 YFPs showed the expected G␤␥-dependent enhancement of interaction with GIRK1 (Fig. 3D, top, summary in E) but failed to do so in GTP␥S (Fig. 3, D, bottom, F). These results suggest a possible flaw in function of the YFP-labeled G␣ i3 subunits within the presumptive GIRK-G␣-G␤␥ signaling complex.
In contrast to the effect of G␤␥ on GIRK1-G␣ i3 binding, addition of purified GST-G␣ i3 did not affect the interaction between His 6 G␤␥ and 35 S-G1NC (Fig. 3F). These observations indicate that, within the signaling complex, G␣ i3 WT is probably bound to the channel via G␤␥ and not vice versa (see below). In view of the absence of any effect, we did not pursue the matter with the YFP-labeled G␣ i3 subunits.
YFP Labeling and PM Level Titration of G␣ i3 Reveal Duality of G␣ i3 Regulation of GIRK-We next examined functional regulation of GIRK1 CFP /2 by YFP-labeled G␣ i3 constructs (Fig. 4).
In each experiment, we monitored the expression of the fluorescent subunits at the PM (Fig. 4A) along with the regulation of the GIRK1 CFP /GIRK channel (Fig. 4, B-D). We first examined the effects of G␣ i3 NT and compared them with those of G␣ i3 WT measured in the same experiment. Increasing the amount of mRNA of G␣ i3 NT resulted in a concomitant increase in its PM expression (Fig. 4B), without markedly affecting the PM level of GIRK1 CFP (supplemental Fig. S4A). In most experiments (n ϭ 4 out of 6), G␣ i3 NT was unable to reduce I basal and did not substantially increase I evoked in a wide range of G␣:channel mRNA stoichiometries. Accordingly, R a was unchanged, suggesting that G␣ i3 NT cannot proficiently control the basal activity of the channel (Fig. 4B). Nevertheless, coexpression of G␣ i3 NT accelerated the activation kinetics of ACh-induced I evoked , reducing act to below 1 s, similarly to G␣ i3 WT (Fig. 4B), suggesting that G␣ i3 NT efficiently transduces activation from GPCR to GIRK. It thus appears that reduction in I basal and acceleration of agonist response may be separate functions of G␣ i .
Next, we examined G␣ i3 117 and G␣ i3 CT (Fig. 4, C and D). Both proteins expressed well at the PM and, unlike G␣ i3 NT, reduced I basal . G␣ i3 117 did not reduce I total , therefore increasing R a , although not as well as G␣ i3 WT. However, G␣ i3 117 persistently slowed down act , to 6 -8 s, substantially slower than act seen with endogenous G␣ (Fig. 4C). Strikingly, G␣ i3 CT decreased I basal , I ACh , and I total at all expression levels (Fig. 4D), highly reminiscent of a G␤␥ scavenger (13), and markedly slowed act . In all, G␣ i3 CT was the least functional of all three YFP-labeled constructs. These results indicate the possibility of imperfect coupling of G␣ i3 117 and G␣ i3 CT to m2R or to GIRK.
We have previously reported that to substantially reduce I basal of GIRK1/2, 2.5-5-fold excess of G␣ i3 WT RNA over the RNA of the channel was necessary, suggesting that G␣ i "titrates" the expressed channel (12). We noticed that G␣ i3 NT expresses less well than G␣ i3 117 or G␣ i3 CT (e.g. Fig. 4A), raising the possibility that this is the reason for the inability of G␣ i3 NT to reduce I basal . Therefore, we titrated the amount of RNA of G␣ i3 NT and G␣ i3 117 to reach similar PM expression levels, keeping the RNA dose of the channel constant (1 ng of each subunit) (Fig. 5, A and B; see complete results of two experiments in supplemental Fig. S4, B and C). When high doses of the RNA of G␣ i3 NT were injected and expressed at the PM equally with G␣ i3 117 (Fig. 5, A and B), it was able to reduce I basal by ϳ60% and increase the R a , although G␣ i3 117 reduced I basal even more, by ϳ80% (Fig. 5C). Importantly, when G␣ i3 NT and G␣ i3 117 showed equal expressions, G␣ i3 NT significantly accelerated act (as G␣ i3 WT), whereas G␣ i3 117 significantly decelerated the activation (Fig. 5D). experiments; N is shown above the bars. C, statistical analysis of the effect of GTP␥S on the interaction between His 6 G␤␥ and ivt synthesized G␣ i3 xFP proteins. Percent binding in GTP␥S relative to GDP was calculated in each experiment, and mean % binding calculated from all experiments is shown for each G␣ i3 construct. For each construct, the effect of GTP␥S was analyzed using paired t test; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. The differences between the four proteins were analyzed using one-way ANOVA and found nonsignificant (n.s.; p Ͼ 0.05).
In the experiment of Fig. 5, A-D, G␣ i3 WT was expressed at a relatively low dose that did not decrease I basal , but acceleration of act was clearly observed (Fig. 5D). However, correction for expression level changes indicated that a slight increase in PM expression of the channel caused by G␣ i3 WT could compensate for the decrease in I basal (Fig. 5C, open bars). Therefore, we performed additional experiments with low doses of G␣ i3 WT RNA (1-2.5 ng). One of them is shown in Fig. 5, E-H. These experiments confirmed that acceleration of act by G␣ i3 WT (Fig. 5H) takes place under conditions when I basal is unchanged (Fig. 5F), and the expression of the channel is not altered (Fig.   5E). Thus, reduction in I basal and acceleration of agonist response appear to be separable functions of G␣ i .
Suppression of Endogenous G␣ i/o Accentuates Differences among xFP-labeled G␣ i3 -In studying the properties of heterologously expressed G␣ i/o , it is customary to suppress the activation of endogenous G␣ i/o with PTX. However, since PTX prevents the dissociation of G␣ i/o from G␤␥ and "freezes" the heterotrimeric form (50), there is a chance that it may also lock the hypothetical complex of GIRK with the inactivated G␣ i/o ␤␥ heterotrimer and interfere with actions of the expressed G␣. With this reservation, we used PTX to alleviate the interference FIGURE 3. YFP-labeled G␣ i3 proteins bind GIRK1 but appear deficient in a triple GIRK1-G␣-G␤␥ interaction test. A, pulldown of in vitro translated, 35 S-labeled G␣ i3 WT and G␤ 1 ␥ 2 by purified GST-fused full cytosolic domain of GIRK1, GST-G1NC. Upper and middle panels show representative autoradiograms of bound and added 35 S-labeled proteins, respectively (binding and input as in Fig. 2). The lower panel shows Coomassie staining of the added proteins.
GST-G1NC appears as two bands, with upper band corresponding to the correct calculated molecular size (ϳ72 kDa). GST-G1NC, but not GST, binds G␣ i3 WT or G␤ 1 ␥ 2 , no matter the nucleotide added (GDP or GTP␥S). B, summary of binding of G␣ i3 WT and G␤ 1 ␥ 2 by GST-G1NC, expressed in % of input. C, purified G␤ 1 ␥ 2 enhances the binding of 35 S-labeled G␣ i3 WT to GST-G1NC in the presence of GDP. Pulldown was done using glutathione-Sepharose, with GST alone as negative control. Similar result was observed in the presence of GTP␥S (see summary in E). D, representative experiment showing the binding of 35 S-labeled ivt G␣ i3 xFP proteins to GST-G1NC, in the absence or presence of purified G␤ 1 ␥ 2 . E, summary of the effect of G␤ 1 ␥ 2 on the interaction between G1NC and 35 S-labeled ivt G␣ i3 xFP from four to nine independent experiments. In GDP, G␤ 1 ␥ 2 significantly enhances the interaction between G1NC and all 35 S-labeled ivt G␣ i3 xFP (top), but it fails to do so in GTP␥S (bottom). n.s., nonsignificant. F, G␣ i3 does not enhance interaction between His 6 G␤ 1 ␥ 2 and 35 S-labeled ivt G1NC (representative experiment, out of two). His 6 G␤ 1 ␥ 2 (3 g) pulls down ivt G1NC in the absence or presence of two different amounts of purified GST-G␣ i3 WT. Upper and middle panels show binding and 1/60 input of ivt G1NC, respectively. The lower panel shows Western blot of His 6 G␤ 1 ␥ 2 .
from the endogenous G␣ i/o . We constructed PTX-insensitive G␣ i3 *WT and G␣ i3 *YFP constructs (asterisk denotes the C351I mutation that prevents ADP-ribosylation by PTX (51)). The oocytes were injected with the catalytic subunit of PTX 2-20 h prior to current measurement. PTX eliminated I evoked in the absence of coexpressed G␣ i (Fig. 6A and supplemental Fig.  S5B). PTX treatment also tended to reduce I basal and the PM levels of coexpressed G␣ i3 proteins and of GIRK1 xFP (supplemental Fig. S5). However, I evoked was faithfully restored after coexpression of G␣ i3 *WT, although in these two experiments the reduction of I basal by G␣ i3 WT did not reach statistical significance (Fig. 6, A and B). act was 2.2 Ϯ 0.19 s (n ϭ 12), slower than with PTX-sensitive G␣ i3 WT, probably reflecting an interference of the C351I mutation with G␣ activation by the GPCR (52). In contrast, G␣ i3 *CT was completely unable to activate GIRK after PTX treatment (Fig. 6, B and C), suggesting that the small agonist-induced activation seen after overexpression of G␣ i3 CT in the absence of PTX (Fig. 4D) was due to endogenous G␣ i/o ; the slowing of act probably resulted from competition between the two G␣ i pools (53).
Both G␣ i3 *NT and G␣ i3 *117 showed robust RNA dose-dependent expression in the PM and restored I evoked (Fig. 6, B-D). Although G␣ i3 *NT did not reduce I basal , it properly restored the fast activation of I evoked , with act identical to that obtained with G␣i3*WT (Fig. 6, B and D). It is probable that the lack of reduction in I basal stems from its relatively low expression pattern along with the tendency of PTX to further reduce its expression (supplemental Fig. S5A). In contrast, both G␣ i3 *CT and G␣ i3 *117 reduced I basal ; at high expression levels G␣ i3 *117 also robustly reduced I evoked and I total . Notably, I evoked with G␣ i3 *117 activated very slowly with a act of ϳ10 s, as observed with nonmutated G␣ i3 proteins (Fig. 5E). In all, of the three G␣ i3 *YFP constructs, G␣ i3 *NT best conveys the activation from GPCR to GIRK. G␣ i3 *117 is able to regulate I basal but is deficient in providing the proper fast activation. The striking differences in actions of G␣ i3 NT and G␣ i3 117 further support the notion that regulation of GIRKs basal activity and fast donation of G␤␥ for GPCR-induced GIRK activation are separate functions of G␣ i/o .
FRET Indicates an Intimate Contact of GIRK1 with G␤␥ but Not with G␣ i3 xFP-Although none of xFP-labeled G␣s fully reproduced all GIRK regulations and interactions seen with G␣ i3 WT, G␣ i3 NT and G␣ i3 117 were functional enough to attempt the assessment of G␣ i3 xFP-GIRK interaction by FRET. GIRK1 CFP /2 and G␣ i3 -YFP were expressed in oocytes, and spectral FRET analysis was performed using Zeiss 510 meta confocal microscope (Fig. 7, supplemental Fig. S6) (16,35). CFP-labeled and G protein-insensitive inwardly rectifying K ϩ channel IRK1 (Kir2.1) was used as the negative control. We titrated the expression of each of the proteins to construct FRET saturation curves (36,54). To overcome variability inherent to fluorescence measurements, we utilized a previously reported doubly xFP-labeled YFP-IRK1-CFP expressing CFP and YFP at a 1:1 molar ratio, thus enabling us to calibrate the readings of CFP and YFP and to assess their actual molar ratio in each experiment (supplemental Fig. S6B) (16). The obtained molar ratio is independent of the experimental settings and allows combining the results of multiple experiments into one saturation curve. The "saturating" value of FRET reflects the distance between the centers of the fluorophores, CFP and YFP (36), assuming an absence of anisotropic effects (17). We estimated that the maximal density of the expressed fluorescent proteins was usually below 100 molecules/m 2 , 4 which is mild compared with many hundreds of molecules/m 2 usually exploited in FRET experiments in mammalian cells (55). This strategy helps to avoid "crowding" effects that may lead to FRET among weakly interacting molecules due to random collisions (54).
We detected very small, if any, FRET signals between GIRK1 CFP /2 and G␣ i3 117-YFP or G␣ i3 NT-YFP, even at very high donor/acceptor molar ratios. There were no significant differences between the FRET signals obtained by either GIRK1 CFP or IRK1 CFP (our negative control) and the G␣ i3 -YFPs (Fig. 7, A, C and D). We also extended our FRET assay to an N-terminally labeled GIRK1 (CFP-GIRK1) and similarly found no evidence for a specific FRET signal (Fig. 7E). We also tested the interaction between GIRK1 CFP /2 and G␤␥ YFP , in which G␥ 2 was labeled (see under "Experimental Procedures"). Confirm-4 S. Berlin and N. Dascal, unpublished observations.  Fig. S4, B and C, for a full summary of this and an additional similar experiment). The amount of expressed GIRK1 CFP /2 was also monitored. A, representative images of oocytes expressing G␣ i3 NT (20 ng of RNA) and G␣ i3 117 (2 ng of RNA) at comparable expression levels. The summary of PM expression levels of all constructs is shown in B. Note that coexpression of G␣ i3 caused only slight, statistically insignificant, changes in channel expression. C and D, effect of G␣ i3 WT and of equally expressed G␣ i3 NT and G␣ i3 117 on channel activity (I basal , I ACh , I total ; R a is shown in the inset, and act in D). Correction of I basal for minor changes in GIRK expression (see Refs. 13,16) in different test groups did not yield major changes except for accentuating the decrease in I basal caused by G␣ i3 WT. E-H, mild expression of G␣ i3 WT (2.5 ng of RNA) does not affect GIRK expression (E) and does not reduce I basal (F), but it still significantly increases R a (G) and accelerates activation (H). A representative experiment, out of three, in which the RNA dose of G␣ i3 WT was chosen or titrated to be just below the level that significantly reduces I basal , is shown. Number of oocytes is shown within bars. *, p Ͻ 0.05; **, p Ͻ 0.01. n.s., nonsignificant (compared with control group (where no G␣ subunit was coexpressed)).
ing previous studies by Reuveny and co-workers (17) and Hebert and co-workers (21), GIRK1 CFP and G␤␥ YFP gave a saturable and specific FRET signal in Xenopus oocytes, much higher that the nonspecific signals obtained for GIRK and G␣ or IRK and G␤␥ (Fig. 7, B and F). The apparent maximal FRET efficiency obtained by fitting a Michaelis-Menten-type equation to the data, E app(max) , was 14.5%. This corresponds to a distance of ϳ65 Å assuming random fluorophore orientation (36), similar to a previous estimate of ϳ60 Å obtained with GIRK1/4 (17). Our findings strengthen the notion that G␤␥ is in close proximity to GIRK even prior to activation of the receptor. If G␣ i3 is present in the signaling complex, it should lie at a greater distance (Ͼ10 nm).

DISCUSSION
We investigated the role and mechanisms of regulation of the neuronal GIRK1/2 channel by G␣ i using wild-type and three xFP-fused G␣ i3 constructs. An extensive array of functional assays showed that none of the G␣ i3 xFPs can fully reproduce all regulatory effects of the wild-type G␣ i3 . Of the three G␣ i3 xFP, only the N-terminally labeled, myristoylated G␣ i3 NT mimicked all aspects of regulation, but it was less efficient in reducing the basal activity (I basal ). The distinct functional properties of the different G␣ i3 xFPs and dose-dependent differences in effects of G␣ i3 WT revealed that regulations of basal and agonist-evoked activity of GIRK1/2 are separate and independent FIGURE 6. Suppression of endogenous G␣ i/o with PTX accentuates the differences between the xFP-labeled G␣ i3 . A, representative currents showing the effect of PTX treatment. PTX almost completely abolished I ACh (black trace), whereas coexpression of PTX-insensitive G␣ i3 WT (denoted G␣ i3 WT*) completely restores I ACh (red trace). B, representative current traces and summary of act showing the activation of GIRK using PTX-insensitive YFP-labeled G␣ i3 subunits. G␣ i3 CT* does not mediate channel activation via m2R. The lower set of traces in B shows normalized ACh-evoked currents, demonstrating differences in kinetics of activation (icons show the color code). C-E, summary of PM expression of the PTX-insensitive YFP-labeled G␣ i3 subunits: G␣ i3 NT*, G␣ i3 117*, and G␣ i3 CT*, respectively, and their functional effects on GIRK currents. *, p Ͻ 0.05; **, p Ͻ 0.01. n.s., nonsignificant (compared with control group (where no G␣subunit was coexpressed)). a.u., arbitrary units.
functions of G␣ i . These functions are probably carried out by the two different guanine nucleotide-associated forms of G␣, although G␣ i GDP regulates I basal , G␣ i GTP accelerates the G␤␥induced channel opening following activation of the GPCR. A combination of direct biochemical binding and FRET measurements indicated that, within the GIRK-G protein signaling complex, G␣ is attached to GIRK via G␤␥. The findings reported here provide new structural and functional insight into the mechanisms of GPCR-G protein-GIRK signaling pathway and support the idea that GIRK is an effector of both G␤␥ and G␣ i .
xFP Tagging of G␣ May Disrupt Regulation of Effectors-Proteins tagged with derivatives of GFP (xFPs) are generally reported to retain native function, despite their large size (ϳ27 kDa; see supplemental Fig. S1) (56). We find that C-or N-terminal xFP-tagged GIRK1 produces functional channels properly gated not only by GPCRs and G␤␥, as shown previously (17,41), but also by G␣ i WT. However, this was not the case with xFP-tagged G␣ i3 . Our results reveal striking incompetence of two of the three xFP-labeled constructs used here in adequately restoring GPCR-GIRK coupling. Only the N-terminally myristoylated xFP-fused G␣ i3 NT produced activation as fast as G␣ i3 WT. GPCR-induced activation with G␣ i3 117 was substantially slower, and it was completely absent in the C-terminally xFP-labeled G␣ i3 CT. The latter is probably due to defective activation by the GPCR, which crucially depends on interaction of GPCR with an intact C terminus of G␣ (1). G␣ i3 CT probably competes with the endogenous G␣ for m2R and G␤␥, thus reducing both free [G␤␥] (and thus I basal and I total ) and the amount of GPCR available for interaction with G i/o .
Of particular interest is the flaw in coupling to GIRK exhibited by G␣ i3 117, as evidenced in kinetic slowing of agonistinduced activation (compared with the activation via the endogenous G␣ i/o of the oocyte), which stands in contrast to the acceleration caused by G␣ i3 WT and G␣ i3 NT. Incidentally, the kinetics of activation of GIRK by G␣ i3 117 and similar con- . CFP-labeled GIRK shows FRET with YFP-labeled G␤␥ but not with YFP-labeled G␣ i3 subunits. A, emission spectra of groups expressing GIRK1 CFP /2 (black trace) or GIRK1 CFP /2 and G␣ i3 NT (light gray trace) when the donor CFP was excited with the 405 nm laser. Note that the spectra almost overlap. In contrast, B shows the emission spectra of GIRK1 CFP /2 or GIRK1 CFP /2 and G␤␥ YFP . Note the large increase in emission in the YFP emission range (520 -580 nm), revealing the energy transfer between the donor and acceptor molecules. C-F, apparent FRET efficiency (E app ) of resonance transfer between GIRK1 CFP and G␣ i3 117 (C), GIRK1 CFP and G␣ i3 NT (D), CFP-GIRK1 and G␣ i3 117 (E), and GIRK1 CFP and G␤␥ YFP (F). IRK1 CFP was used for the assessment of nonspecific FRET signal (negative control). Measurements from multiple experiments were incorporated in a single curve, by using donor/acceptor molar ratio (see "Experimental Procedures"). Curves were fitted to a hyperbolic function.
structs have not been tested in the past, whereas other aspects of their function have been described as normal. Gibson and Gilman (28) have demonstrated that G␣ i3 117 and analogous constructs of G␣ i1 and G␣ i2 have adequate uncatalyzed GDP-GTP exchange in vitro and robust coupling to several isoforms of ␣2 adrenergic receptors, as monitored by rearrangements of G␣ and G␤␥ reported by FRET. Fast FRET or bioluminescence resonance energy transfer measurements indicated rapid (Ͻ1 s) GPCR-induced rearrangements of G␣ and G␤␥, using constructs of G␣ o and G␣ i1 labeled with xFPs at positions analogous to 117 in G␣ i3 , as well as G␣ i/o labeled in the ␣B-␣D loop (position 91 in G␣ i1 ) (25,29). Thus, the coupling of G␣ i3 117 to GPCR and to G␤␥ appears intact, and the slowing of GIRK activation probably reflects a problem in the G␣ i3 117-GIRK dialog, in line with the idea that G␣ i actively regulates GIRK gating (12,39). Indeed, interaction with GIRK and specific activation of GIRK by G␣ i/o -coupled GPCRs involve the helical domain of G␣ i/o (57). We propose that xFP insertions in the G␣ helical domain obstruct G␣ i/o -GIRK interaction via the interface between the helical domain of G␣ i/o and the putative G␣-interaction site in GIRK1, interfering with normal regulation of GIRK by G␣ i/o .
Regulation of Basal and Evoked Activities of GIRK Are Separate Functions of G␣ i -Heterologously expressed GIRK1/2 (or GIRK1/4) channels have substantial G␤␥-dependent basal activity, and G␣ is necessary to keep the basal activity low and to ensure robust activation by agonists (12,58). The reduction of I basal is carried out by G␣ i/o GDP , by forming G␣␤␥ heterotrimers, which presumably remain attached to the channel (18,19,39). The formation of this minimal signaling complex, by itself, is expected to accelerate activation of the channel by agonists (59), with no need to assume that the activated G␣ i/o GTP contributes to activation. The reasoning is proximity (59) and stoichiometry; a GIRK channel has four G␤␥-binding sites (60,61), and extent of activation is a graded function of number of bound G␤␥ molecules (30,62). If some of the heterologously expressed GIRK channels are associated with less than four G␣ i/o ␤␥ heterotrimers, addition of lacking G␣ i/o could enhance and accelerate activation. Accordingly, in our previous work, we attributed the acceleration of act by coexpressed G␣ i3 to formation of G␣ i3 ␤␥-GIRK complexes (40). If this is the case, it is possible that G␣ i3 117 does not form such complexes (slow dissociation from G␤␥ can be excluded; see above).
However, the minimal hypothesis implicating G␣ i GDP alone in the accelerating effect of coexpressed G␣ i appears insufficient to account for our new observations. First, we find that the acceleration of act by G␣ i3 WT and G␣ i3 NT takes place in the absence of a reduction in I basal , which is a hallmark of formation of G␣ i/o ␤␥ heterotrimers functionally associated with the channel (13). In fact, the two constructs that best reduce I basal , G␣ i3 117 and G␣ i3 CT, are deficient in GPCR-driven coupling to GIRK, which further emphasizes the separation of the two functions of G␣ i . Second, in protein interaction assays, G␣ i3 xFP GDP constructs bound GIRK1 at least as well as the wild-type G␣ i3 , both in the absence and presence of G␤␥ (Figs. 2 and 3), suggesting that the formation of G␣ i ␤␥-GIRK complex by these proteins is retained.
In view of these considerations, we hypothesize that acceleration of act is carried out by G␣ i GTP , which acts synergistically with G␤␥ in GIRK1/2 activation. So far no direct positive regulation of mammalian GIRKs by "active" G␣ GTP has been demonstrated. Rather, an inhibitory regulation of GIRK1/4 by G␣ i1 GTP␥S (but not G␣ i3 GTP␥S ) has been observed at very low expression levels of the channel, suggesting a mechanism involving interference with endogenous G␣ i/o (12,14). However, a synergistic, mutually obligatory activation by G␣ i GTP and G␤␥ in artificial membranes has been recently found in a chimeric channel composed of a transmembrane domain of a bacterial K ϩ channel and a truncated cytosolic domain of mammalian GIRK1 (15). Furthermore, coexpression of a constitutively active (GTP-bound) mutant of G␣ i3 significantly modified G␤␥-induced conformational changes in the cytosolic domains of GIRK1 and GIRK2 subunits within the GIRK1/2 heterotetrameric channel, also suggesting more than additive effects of G␣ i3 GTP and G␤␥ (16). In accelerating act , G␣ i/o GTP could regulate GIRK by a direct interaction, as supported by biochemical evidence (e.g. Fig. 3); however, an indirect effect cannot be ruled out. Thus, regulation of GIRK channels by G␣ i/o GTP appears plausible but requires further study.

G␣ i/o ␤␥ Heterotrimers Contact GIRK1 Primarily via G␤␥-
The idea that GIRK channel forms a signaling complex with the G protein has become widely accepted (for review, see Refs. 9, 22), but structural and molecular details are poorly understood. The interaction of GIRK1 subunit with G␣ i is enhanced by G␤␥, supporting the notion that heterotrimeric G␣ i ␤␥ is anchored to GIRK1 (13,16,19). It has been proposed that, at least in GIRK2 subunit, anchoring occurs via G␣ i (18), suggesting a tight contact between GIRK and G␣ i/o preceding GPCR activation. However, studies that sought support for this hypothesis using resonance energy transfer yielded inconsistent results (see "Introduction") (20,21). Because both studies (20,21) used G␣ constructs with luciferase or xFP insertions within the helical domain of G␣ i , we initially hypothesized that the defect in functional interaction of these proteins with GIRK could account for the inconsistent results. We therefore used the N-terminally labeled G␣ i3 , which carried the closest resemblance to G␣ i3 WT in its regulation of GIRK1/2. Yet, neither G␣ i3 NT nor G␣ i3 117 produced a significant FRET signal in our experiments. Under the same conditions, a strong FRET between G␤␥ and GIRK1 was observed, confirming previous publications (17,21).
What could be the reason for the lack of FRET between G␣ i3 and GIRK1? Below we are listing four types of possibilities, starting from the least plausible. 1) When GIRK and G␣ i3 are bound to each other, the distance between the xFP labels exceeds the limit of FRET resolution (ϳ100 Å), or the angle between the donor and acceptor dipoles is unfavorable for FRET. We deem this as unlikely, given the moderate sizes of these proteins (supplemental Fig. S1) and the fact that placing xFP labels on either N or C terminus of GIRK1, or N terminus of G␣ i3 , or in the helical domain did not make a difference. At present, we cannot rule out other obstructions to FRET such as quenching of some of the fluorescence by the protein environment of GIRK-bound G␣ i3 xFP. 2) Subtle differences between G␣ i3 NT and G␣ i3 WT bring about such a dramatic change in GIRK1-G␣ i3 interaction that the close interaction between the channel and G␣ i ␤␥, via G␣ i , is lost. This is also unlikely; our functional and biochemical data do not lend support for this notion.
3) The hypothetical G␣ i ␤␥-GIRK complex is not stable but dynamic, with relatively low affinity of interaction that is not captured in our FRET assay. This possibility cannot be ruled out, despite the strong functional and biochemical evidence in support of a stable complex (22).
We favor a fourth alternative explanation; within the GIRK1/ 2-G␣ i/o ␤␥ signaling complex, G␣ i/o is anchored to the channel primarily via G␤␥. The molecular sizes of G␣ and G␤␥ (supplemental Fig. S1), and the estimated 60 -65 Å distance between the centers of xFPs associated with GIRK1 and G␥, suggest that the distance between GIRK1-and G␣ i -attached fluorophores may be big enough to thwart FRET if G␣ i is associated with GIRK1 through its binding with G␤␥. This possibility is further supported by several lines of new and published data. 1) Our previous functional and immunocytochemical studies suggested that heterologously expressed GIRK1/2 is preferentially associated with G␤␥, whereas, under certain conditions, G␣ i/o appears to be missing from the signaling complex (32,39). These findings are in line with the above assumption. 2) The binding of G␣ i3 to GIRK1 is several times weaker than that of G␤␥ (Fig. 3). Lack of strong binding of G␣ GTP , compared with the strong binding of G␤␥, to immunoprecipitated subunits of cardiac GIRK channel has been reported previously (47). It is important to emphasize that here we used the full cytosolic domain of GIRK1, rather than purified separate G␤␥or G␣-binding segments, because the actual G␤␥-binding surface of GIRK1 (and probably other GIRK subunits) is composed of several N-and C-terminal cytosolic segments (48). Importantly, the enhancing effect of G␤␥ on GIRK1-G␣ i3 binding occurs only in the context of a full cytosolic domain of GIRK1, but not with separate N-or C-terminal domains (16). This is the first time that a direct comparison of G␣ i3 and G␤␥ binding to full cytosolic domain of GIRK1 is done. The pulldown of G␣ i3 and G␤␥, although semiquantitative, was performed under identical conditions and in the same experiment, enhancing our confidence in the result. 3) G␤␥ enhances the interaction of GIRK1 with G␣ i3 , but G␣ i3 does not affect the interaction of GIRK1 with G␤␥ (Fig. 3); the opposite could be expected if G␤␥ were anchored via G␣ i/o . In all, our findings support a model in which the GIRK is the scaffold of the signaling complex that anchors G␤␥, whereas G␣ i/o is associated with the complex via its binding to G␤␥.