The Stoichiometry of G bg Binding to G-protein-regulated Inwardly Rectifying K 1 Channels (GIRKs)*

G-protein-coupled inwardly rectifying K 1 (GIRK; Kir3.x) channels are the primary effectors of numerous G-protein-coupled receptors. GIRK channels decrease cellular excitability by hyperpolarizing the membrane potential in cardiac cells, neurons, and secretory cells. Although direct regulation of GIRKs by the heterotrim-eric G-protein subunit G bg has been extensively studied, little is known about the number of G bg binding sites per channel. Here we demonstrate that purified GIRK (Kir 3.x) tetramers can be chemically cross-linked to exogenously purified G bg subunits. The observed laddering pattern of G bg attachment to GIRK4 homotetramers was consistent with the binding of one, two, three, or four G bg molecules per channel tetramer. The fraction of channels chemically cross-linked to four G bg molecules increased with increasing G bg concentrations and approached saturation. These results suggest that GIRK tetrameric channels have four G bg binding sites. Thus, GIRK (Kir 3.x) channels, like the distantly related cyclic nucleotide-gated channels, are tetramers and exhibit a 1:1 subunit/ligand binding stoichiometry. Roughly 2% of the human genome encodes G-protein-coupled receptors (1). Agonist binding to these G-protein-coupled receptors catalyzes the activation of G a and G bg subunits of hetero-trimeric G-proteins. The free G a and G bg subunits can then interact independently or in concert with numerous effectors. immunoblotting experiments. Only one anti-G bg antibody (anti-KTREGNVRVSREL, Chemicon International, Inc. Temecula, CA) reacted with G bg after DTSSP treatment. Typically, DTSSP treatment reduced total antigenic signal by . 90%, . 60%, and . 95% for anti-GIRK4, anti-GIRK1, and anti-G bg antibodies, respec-tively. Transfer times for immunoblot analysis were extended to . 2 h at 15 V to improve transfer of the high molecular weight complexes. A GS-700 imaging densitometer (Bio-Rad Laboratories, Hercules, Califor- nia) was used to analyze the protein gels and immunoblots. Molecular masses were calculated using densitometry profiles from a combination of prestained high molecular mass markers (Bio-Rad) and low and high molecular mass markers (Amersham Pharmacia Biotech). In a portion of the gels, thyroglobulin (Amersham Pharmacia Biotech) was added to ensure linearity up to 330 kDa.

Homotetrameric and heterotetrameric combinations of the four known mammalian GIRK subunits are activated by neurotransmitters in the nervous system, pancreas, and heart. Muscarinic (m2, m4), ␥-aminobutyric acid (GABA B ), D 2 -dopamine, ␣ 2 -adrenergic, opiate, somatostatin, and adenosine all employ the G␣ i -G␤␥ signal transduction system to activate GIRK channels via direct G␤␥ binding to the tetrameric channel. GIRK4-knockout mice have irregularities in heart rate variability (5) and difficulties with spatial learning (24). GIRK2-knockout mice are prone to seizures (25). Weaver mice have a mutation in the pore domain of the GIRK2 subunit (26) that renders the channel nonselective (27) and results in the degeneration of cerebellar granule cells (28) and the dopaminergic neurons of the substantia nigra (29,30). The native atrial I KACh channel is composed of two GIRK1 subunits and two GIRK4 subunits (31)(32)(33) that comprise a channel that mediates neuronal regulation of heart rate. Biochemical studies indicate that G␤␥ binds the native I KACh complex with a K d of 55 nM (9). G␤␥ binds both recombinant GIRK1 (K d ϭ 125 nM) and GIRK4 (K d ϭ 50 nM) (9). GIRK1 subunits are unable to form functional homomultimers (34), whereas GIRK4 homomultimers have been biochemically isolated from bovine atria (35). GIRK2/3 and GIRK1/2 heteromultimers have also been isolated from brain (1,36). The C-terminal tail of GIRK1 and GIRK4 subunits bind G␤␥ (9, 32, and 37-46), but the detailed steps of how this binding leads to channel gating is not known. Furthermore, there is limited data about the areas of G␤␥ that bind GIRK channel subunits (43,47) and about how many G␤␥ subunits can bind the tetrameric channel complex.
We have used a biochemical approach to determine how many G␤␥ subunits bind GIRK tetramers. By extending our previous chemical cross-linking studies (31), which indicated that GIRKs form tetramers, we demonstrate that GIRK4 homotetramers bind four G␤␥ subunits in their natural membrane environment.
Expression and Isolation of GIRKs from COS7 and CHO Cells-Plasma membrane proteins containing GIRK1-AU5 and GIRK4-AU1 were isolated from COS7 cells and solubilized as described (49).
G␤␥ Purification-G-proteins were isolated from bovine brain and separated into G␣ and G␤␥ subunits (50) and were further purified by affinity chromatography using immobilized G␣ (51).
G␤␥ Binding in Membranes-Isolated COS7 cells or native atrial membranes were treated for 1 h with 100 mM dithiothreitol and then dialyzed against 20 -50 mM HEPES, 100 mM NaCl, pH 7.4 -7.5 (G␤␥ binding buffer). Individual membrane aliquots were preincubated with purified bovine brain G␤␥ and rotated for 20 min at room temperature prior to cross-linking. The G␤␥ stock solution was in G␤␥ binding buffer containing 0.1% CHAPS (0.1% G␤␥ binding buffer). The final CHAPS concentration was less than or equal to 0.1%.
G␤␥ Binding to Solubilized Protein-Solubilized COS7 membrane proteins were treated for 1 h with 100 mM dithiothreitol and then dialyzed against 0.1% G␤␥ binding buffer. Individual aliquots were preincubated with purified brain G␤␥ (supplied in 0.1% G␤␥ binding buffer) and rotated for 20 min at room temperature prior to cross-linking.
Chemical Cross-linking-5 mM dithiobis[sulfosuccinimidylpropionate] (DTSSP, Pierce Chemical, Rockford, IL) was prepared as an 11ϫ stock solution immediately prior to use in 100 mM HEPES-containing buffer, pH 7.5. Iodine was added only to solutions containing purified I KACh . Cross-linking reactions were allowed to proceed for 30 min at room temperature and then quenched with 50 mM Tris. Typically 5-10 g of membrane proteins were used per reaction in a final volume of 15 l.
SDS-PAGE and Immunoblotting-Atrial membrane proteins or recombinant GIRK proteins were resuspended in Laemmli sample buffer containing 100 mM dithiothreitol (or 30 mM iodoacetamide when a cross-linking agent was used) for 15 min at 50°C, 30 min at room temperature, and 15 min at 50°C. 3-10% separating, 3% stacking, and pre-cast 2-15% (ISS) gels were utilized. Samples were analyzed by immunoblotting with anti-GIRK4 antibodies (generated against amino acids 19 -32, Ref. 31) and/or anti-GIRK1 antibodies (generated against the last 156 amino acids of GIRK1, Ref. 31). Several antibodies were tested for use in the anti-G␤␥ immunoblotting experiments. Only one anti-G␤␥ antibody (anti-KTREGNVRVSREL, Chemicon International, Inc. Temecula, CA) reacted with G␤␥ after DTSSP treatment. Typically, DTSSP treatment reduced total antigenic signal by Ͼ90%, Ͼ60%, and Ͼ95% for anti-GIRK4, anti-GIRK1, and anti-G␤␥ antibodies, respectively. Transfer times for immunoblot analysis were extended to Ͼ2 h at 15 V to improve transfer of the high molecular weight complexes. A GS-700 imaging densitometer (Bio-Rad Laboratories, Hercules, California) was used to analyze the protein gels and immunoblots. Molecular masses were calculated using densitometry profiles from a combination of prestained high molecular mass markers (Bio-Rad) and low and high molecular mass markers (Amersham Pharmacia Biotech). In a portion of the gels, thyroglobulin (Amersham Pharmacia Biotech) was added to ensure linearity up to 330 kDa.

RESULTS
Previous chemical cross-linking studies demonstrated that GIRK subunits form tetrameric channels and that the native atrial channel I KACh is composed of 2 GIRK1 and 2 GIRK4 subunits (31). Complete cross-linking of purified atrial I KACh formed a single adduct with a total molecular mass that was most consistent with a tetramer. In addition, partial crosslinking of purified I KACh produced subsets of molecular weight adducts consistent with monomers, dimers, trimers, and tetramers. In this study, we extended our previous experiments to determine how many G␤␥ molecules can be cross-linked to GIRK tetramers.
G␤␥ Cross-linking to Purified Native I KACh -To test whether GIRK1/GIRK4 heteromultimers could be directly and specifically cross-linked to G␤␥, we used isolated native atrial GIRK1 and GIRK4 subunits (31) and bovine brain G␤␥ (9). Isolated GIRK1 and GIRK4 heterotetramers were preincubated with isolated G␤␥, followed by cross-linking with DTSSP ( Fig. 1). Although the predicted molecular masses of GIRK1 and GIRK4 subunits are 56 and 45 kDa, respectively, the glycosylated GIRK1 migrates in a broad band between 67-72 kDa (9). In the absence of G␤␥, a band formed at 230 kDa, consistent with the total molecular mass of two GIRK1 (56, 67-72 kDa) and two GIRK4 (45 kDa) subunits. In the presence of G␤␥, a band corresponding to a molecular mass of 390 kDa was detected. Because G␤␥, GIRK1 and GIRK4 were the predominant proteins present, we interpreted the 160-kDa shift as the result of direct cross-linking of G␤␥ to GIRK channels. The molecular mass of G␤␥ is 42 kDa, suggesting that the 160-kDa shift was because of cross-linking of several G␤␥ molecules to the GIRK1/ GIRK4 heterotetramers. Because GIRK4 can form homotetramers, we repeated the previous cross-linking experiment using recombinant GIRK4 subunits. In the absence of G␤␥, crosslinking of recombinant GIRK4 resulted in a band at 170 kDa, corresponding to GIRK4 homotetramers. When G␤␥ was added to recombinant GIRK4, cross-linking yielded a band at ϳ320 kDa (not shown). This banding pattern is most consistent with four specific and saturable G␤␥ binding sites per GIRK4 homotetramer.
Cross-linking of Membrane-confined GIRK4 Homotetramers-It is important to study GIRK-G␤␥ binding in its membrane environment because phosphatidylinositol bisphosphate (PIP 2 ) (52, 53) is involved in the G␤␥-mediated activation of GIRK channels. Our previous GIRK cross-linking studies employed isolated solubilized channels (31). In this study, we tested whether GIRK subunits could be cross-linked into tetramers in membranes. After DTSSP cross-linking of membranes from COS7 cells expressing either recombinant GIRK4, GIRK1, or GIRK1 and GIRK4, SDS-PAGE yielded 180 -220-kDa bands (Fig. 2B, lanes 1-3). These bands are similar in molecular mass to those produced when solubilized GIRKs are cross-linked into tetramers (31). Of the GIRK tetramers, the chemically cross-linked GIRK4 homotetramers produced the narrowest band, around 190 kDa (Fig. 2B, lane 1). In addition, the GIRK4 band cross-linked directly in membranes (Fig. 2B, lane 1) was narrower than that of GIRK4 that had been solubilized before cross-linking (31).
Partial G␤␥ Cross-linking to Membrane-confined GIRK Tetramers-GIRK4 homotetramers were used in our membraneconfined GIRK-G␤␥ binding experiments because cross-linking of GIRK4 homotetramers in membranes yielded the narrowest bands. We altered our cross-linking conditions to verify that there were indeed four G␤␥ binding sites in the GIRK tetramer. DTSSP and G␤␥ concentrations were adjusted so that variable numbers of G␤␥ molecules were cross-linked to the GIRK4 homotetramers. COS7 cells were transiently transfected with GIRK4, and their membranes were divided into separate aliquots. Each aliquot was treated with variable amounts of G␤␥ and DTSSP and then analyzed by SDS-PAGE and immunoblotting. Untreated GIRK4, migrated as a 47-kDa trichloroacetic acid-disruptable monomer (Fig. 3B, lane 1). GIRK4 crosslinked with DTSSP migrated as a 170-kDa tetramer (Fig. 3B,  lane 2). GIRK4, preincubated with G␤␥ and then cross-linked with DTSSP, resulted in a laddering pattern of four main adducts (in addition to the GIRK4 homotetramer adduct) with consistent 40 -45-kDa increments between bands (Fig. 3B,  lanes 3 and 4). The proportion of high molecular weight adducts increased with increasing G␤␥ concentrations. Unlike our previous experiments that used solubilized GIRK protein, a population of the membrane-confined GIRK4 homotetramers (166 kDa) remained resistant to any G␤␥ binding. One possible explanation for this observation is that a subpopulation of GIRK4 homotetramers may not have been accessible to the exogenously applied G␤␥. In five independent trials, four GIRK-G␤␥ adducts consistently appeared. In some trials, high molecular mass, lower intensity smears formed, but these bands were not consistently reproducible. A laddering pattern was not formed when G␤␥ was boiled prior to its addition to membranes (data not shown). We hypothesize that the five adducts formed by treatment of G␤␥ and GIRK4-containing solutions with DTSSP represent the binding of zero, one, two, three, and four G␤␥ molecules to GIRK4 homotetramers.
Multiple lines of evidence suggest that G␤␥ is directly crosslinked to GIRK channels in our experiments. G␤␥ has been coimmunoprecipitated with GIRK subunits under the conditions used in our experiments (9). The ϳ45-kDa increments between cross-linked GIRK-G␤␥ adducts are consistent with the stepwise addition of 42-kDa G␤␥ subunits to the channel. Finally, similar results were obtained even when I KACh and G␤␥ were purified to Ͼ95% homogeneity prior to cross-linking. Because, I KAch and G␤␥ are the predominant proteins in solution, the molecular mass shift with G␤␥ addition strongly suggests that G␤␥ is being directly cross-linked to the channel. As a final precaution, we tested whether the putative GIRK-G␤␥ adducts are recognized by anti-G␤␥ antibodies. COS7 cells were transiently transfected with GIRK4 and their membranes were isolated. The membranes were treated with G␤␥ and DTSSP, followed by SDS-PAGE analysis. Immunoblots were probed with anti-GIRK4 antibodies then stripped and reprobed with anti-G␤␥ antibodies (Fig. 4, lanes 1 and 2, respectively). The anti-G␤␥ antibodies recognized bands at molecular masses that correspond to the putative GIRK-G␤␥ adducts.

DISCUSSION
The present study of G␤␥ binding to channel proteins has several advantages over other approaches. First, we ensured that we were using intact tetramers throughout our G␤␥ binding experiments. In addition, we purposely studied G␤␥ binding in membranes to approximate physiological conditions. This is especially important because PIP 2 , a component of the cell membrane, plays a role in G␤␥-mediated activation of GIRKs (52,53). Nonprenylated G␤␥ mutants do not activate GIRK channels (54,55), indicating that G␤␥ association with cell membranes may be a prerequisite for G␤␥ binding. We have paid careful attention to detergent concentrations, because low detergent conditions can potentially expose hydrophobic patches on GIRKs, producing nonspecific binding. Indeed, we found it difficult to prevent GIRK and G␤␥ aggregation in low detergent concentrations.
The stoichiometry of the I KACh -G␤␥ interaction has been repeatedly estimated by using the Hill equation to fit the G␤␥-I KACh dose-response curve. Estimates of the Hill coefficient for I KACh activation varied from 1.5 (9) to 3 (56,57) whereas it was ϳ1 in the study of the direct binding of purified I KACh and G␤␥ proteins (9). Although often used to infer binding stoichiometry of G␤␥ with GIRK subunits, the Hill coefficient is a measure of cooperativity, not the number of binding sites. For the Hill coefficient to equal the G␤␥ binding stoichiometry, two criteria need to be met or approximated. The G␤␥ molecules must bind the channel simultaneously and G␤␥ must bind with infinite cooperativity (58). In addition, the Hill equation does not take into account the increasing open probability of the channel with each ligand molecule bound. Thus, the stoichiometry of G␤␥ binding to I KACh is not adequately determined by fits of the Hill equation to the G␤␥ dose-response relations. Even the more complicated Monod, Wyman, and Changeux (MWC) formula does not properly describe the subunit gating of the cyclic nucleotide-gated channel (59). Nevertheless, the cooperativity in I KACh activation (9,56,57) and the G␤␥-dependent shifts in its gating modes (60,61) suggest that GIRK channels have multiple G␤␥ binding sites.
Given the inadequacy of available models, a direct biochemical approach was used to determine GIRK-G␤␥ binding stoichiometry. Solutions containing purified GIRK1 and GIRK4 were treated with the cross-linking reagent DTSSP in the presence or absence of G␤␥. A 230-kDa band was observed in the absence of G␤␥, compared with a 390-kDa band when G␤␥ was present. We concluded that the 160-kDa shift was the result of covalent linkage of multiple 42-kDa G␤␥ molecules to the channel. Next, solubilized recombinant GIRK4 homotetramers were treated with DTSSP in the presence and absence of G␤␥. A 170-kDa band formed without G␤␥ in contrast to the 320-kDa band in the G␤␥-containing experiment. The 150-kDa shift in the presence of G␤␥ is most consistent with the chemical cross-linking of four 42-kDa G␤␥ molecules to the GIRK4 homotetramers with complete G␤␥ binding site saturation and cross-linking. Finally, a variety of membrane-associated GIRK channels were treated with DTSSP and analyzed by SDS-PAGE. In each case, completely cross-linked GIRK tetramers resulted. To verify that there were four G␤␥ binding sites in GIRK tetramers, we altered our cross-linking conditions. DTSSP and G␤␥ concentrations were adjusted so that variable numbers of G␤␥ molecules were cross-linked to the GIRK4 homotetramers. Five adducts, representing zero, one, two, three, and four G␤␥ molecules cross-linked to the channel, were detected. We were unable to cross-link more than four G␤␥ molecules to the channel, even with G␤␥ concentrations two orders of magnitude higher than the K d for G␤␥ binding to GIRK subunits. We conclude that four G␤␥ subunits can bind to a GIRK tetramer.
Currently, the G␤␥ binding site on GIRK subunits is thought to reside primarily on the cytoplasmic C-terminal region shortly after the second transmembrane domain. But a detailed description of the GIRK/G␤␥ binding site will undoubt-edly require direct structural determination. For example, it is not possible to determine with our experiments whether the G␤␥ binding pockets were formed within subunits or between subunits. Short of direct structural determination, in future experiments it may be possible to cross-link G␤␥ to GIRKs during patch clamp recording. Such a technique has proven valuable in evaluating cyclic nucleotide binding to cyclic nucleotide-gated channels (59). FIG. 4. Anti-G␤␥ antibodies recognize bands at molecular masses that correspond to the putative GIRK-G␤␥ adducts. 3-10% SDS-PAGE analysis and immunoblotting of membranes from COS7 cells that were transfected with GIRK4, pretreated with G␤␥, and then treated with DTSSP. Lane 1, GIRK4-containing membranes pretreated with 6.3 M G␤␥ and cross-linked with 4 mM DTSSP probed with anti-GIRK4 antibodies. Lane 2, (lane 1) stripped and reprobed with anti-G␤␥ antibodies.