Dynamic Integration of α-Adrenergic and Cholinergic Signals in the Atria

Numerous heptahelical receptors use activation of heterotrimeric G proteins to convey a multitude of extracellular signals to appropriate effector molecules in the cell. Both high specificity and correct integration of these signals are required for reliable cell function. Yet the molecular machineries that allow each cell to merge information flowing across different receptors are not well understood. Here we demonstrate that G protein-regulated inwardly rectifying K+ (GIRK) channels can operate as dynamic integrators of α-adrenergic and cholinergic signals in atrial myocytes. Acting at the last step of the cholinergic signaling cascade, these channels are activated by direct interactions with βγ subunits of the inhibitory G proteins (Gβγ), and efficiently translate M2 muscarinic acetylcholine receptor (M2R) activation into membrane hyperpolarization. The parallel activation of α-adrenergic receptors imposed a distinctive “signature” on the function of M2R-activated GIRK1/4 channels, affecting both the probability of Gβγ binding to the channel and its desensitization. This modulation of channel function was correlated with a parallel depletion of Gβ and protein phosphatase 1 from the oligomeric GIRK1 complexes. Such plasticity of the immediate GIRK signaling environment suggests that multireceptor integration involves large protein networks undergoing dynamic changes upon receptor activation.

Numerous heptahelical receptors use activation of heterotrimeric G proteins to convey a multitude of extracellular signals to appropriate effector molecules in the cell. Both high specificity and correct integration of these signals are required for reliable cell function. Yet the molecular machineries that allow each cell to merge information flowing across different receptors are not well understood. Here we demonstrate that G protein-regulated inwardly rectifying K ؉ (GIRK) channels can operate as dynamic integrators of ␣-adrenergic and cholinergic signals in atrial myocytes. Acting at the last step of the cholinergic signaling cascade, these channels are activated by direct interactions with ␤␥ subunits of the inhibitory G proteins (G␤␥), and efficiently translate M 2 muscarinic acetylcholine receptor (M2R) activation into membrane hyperpolarization. The parallel activation of ␣-adrenergic receptors imposed a distinctive "signature" on the function of M2R-activated GIRK1/4 channels, affecting both the probability of G␤␥ binding to the channel and its desensitization. This modulation of channel function was correlated with a parallel depletion of G␤ and protein phosphatase 1 from the oligomeric GIRK1 complexes. Such plasticity of the immediate GIRK signaling environment suggests that multireceptor integration involves large protein networks undergoing dynamic changes upon receptor activation.
The heart functions in a continuously changing environment responding to excitatory and inhibitory signals from the sympathetic and parasympathetic branches of autonomic nervous system. Both branches use members of the large family of seven transmembrane G protein-coupled receptors (GPCRs) 2 as membrane gateways for their opposing signals. At rest, parasympathetic signals reduce the heart rate through activation of G i/o -coupled M2Rs. During exercise and stress, sympathetically derived catecholamines target coexisting ARs and increase the cardiac output by accelerating the heart rate and enhancing contractility. The ␤ 1 -, ␤ 2 -, and ␣ 1 -ARs are the most predominant AR subtypes in cardiac myocytes (1). These receptors activate different signaling cascades with ␣ 1 -ARs coupling to G q , ␤ 1 -ARs coupling to G s , and ␤ 2 -ARs coupling to both G s and G i pathways. Although the cellular processes triggered upon activation of individual adrenergic and M 2 receptors have been well delineated (1), how cardiac myocytes integrate the inhibitory and excitatory signals transmitted by these receptors remains less well understood.
Signal integration might arise from heterodimerization of members of distinct subfamilies of GPCRs. Such receptor heterodimerization expands the operational features of the receptors involved, modulating either receptor trafficking or function (2). Importantly, heterodimerization of ␤ 1 -and ␤ 2 -ARs in intact cardiac myocytes was recently reported to give rise to a new population of ␤-ARs with enhanced signaling properties (3).
Assembly of receptor heterodimers, however, is perhaps only one of the diverse mechanisms used by different GPCRs to achieve exquisite and dynamic control of membrane excitability in cardiac myocytes. Integration in vivo might also rely on highly organized signaling networks in which relevant proteins are physically maintained together. Recent studies indicate that atrial GIRK1/4 channels might play a role in a similar network responsible for the coordination of adrenergic and cholinergic signals in the heart (4). These channels are heterotetrameric membrane proteins consisting of two pore-forming subunits, GIRK1 (5,6) and GIRK4 (7). Acting downstream from M2Rs, GIRK channels are responsible for membrane hyperpolarization and ensuing reduction in heart rate following cholinergic stimulation. Channel activation is conferred by direct G␤␥ binding to multiple domains on the intracellular surface of the channel (reviewed in Refs. 8 and 9). Upon G␤␥ activation, atrial GIRK1/4 channels adopt four functional modes, characterized by different open probabilities and rendered active by putative binding of an increasing number of G␤␥ subunits to the channel (10,11). As the G␤␥ concentration increases, the equilibrium among the four functional modes shifts gradually from low to high efficient gating. This operational feature enables GIRK1/4 channels to control membrane excitability with great precision, which is a prerequisite for efficient integration of opposing signals. Furthermore, a fraction of atrial GIRK1/4 channels has recently been shown to form large protein complexes (12). Immunopurification of these macromolecular complexes identified GIRK1 association with G␤␥, G proteincoupled receptor kinase, cyclic adenosine monophosphate-dependent protein kinase, two protein phosphatases, PP1 and PP2A, receptor for activated C kinase 1, and actin, implicating these proteins as potential modulators of channel function (4). Oligomeric complexes containing GIRK channels, adenylyl cyclase, and GPCRs have also been identified (13,14), suggesting the existence of large G protein-coupled signaling networks in the membrane.
To serve as integrators in similar networks, the GIRK channels must conform to two requirements. First, activation of different GPCRs should leave a unique mark on channel function at the single molecule level. Second, the signaling framework of the GIRK1 complexes should "memorize" previous events of GPCR activation. Here we examined whether the GIRK channels met these minimal requirements following concurrent stimulation of M 2 and ␣-adrenergic receptors in atrial myocytes. First reported by Braun et al. (15), the inhibition of GIRK channel currents by G q PCRs is a well established phenomenon (reviewed in Ref. 16), which provides an optimal venue for addressing these questions. Remarkably, the G q -coupled receptors recruit a large repertoire of parallel cellular processes to accomplish the inhibition of GIRK channel currents. Depletion of phosphatidylinositol 4,5-bisphosphate (PIP 2 ), activation of protein kinase C (PKC), and direct binding of G␣ q subunits to the channel have been all shown to contribute to this phenomenon (4, 16 -25). How these processes modulate GIRK channel function at single molecule level remains less well understood. We therefore investigated GIRK1/4 channel function and the mechanisms underlying ␣-AR-mediated inhibition in intact atrial myocytes, using the single-channel experiments. These experiments record the stochastic time trace of repetitive transitions of an individual molecule undergoing conformational fluctuations triggered by receptor activation and allow measurements of molecular properties unattainable from ensemble data. Initially, we established the equivalence between the M2R and G␤␥ activation of GIRK1/4 channels. Accordingly, the M2R-activated channels adopt four functional modes conferred by the apparent binding of an increasing number of G␤␥ subunits to the channel protein. This finding validates the concept of G␤␥-driven control of four functional states, evolving from the same molecule, as a robust framework for understanding GIRK1/4 function and regulation. In this framework, the ␣-AR stimulation perturbs GIRK channel-G␤␥ interactions, leading to a 2-fold reduction in the probability of G␤␥ binding to the channel. In addition, the impaired channel function is associated with concerted changes in GIRK1/4 signaling environment, manifested by differences in the composition of the oligomeric GIRK1 complexes. Intriguingly, cholinergic and ␣-adrenergic signals differentially affect this environment, supporting the notion that the oligomeric GIRK1 assemblies provide a dynamic protein framework involved in the integration of these signals.

EXPERIMENTAL PROCEDURES
Materials-Mouse monoclonal antibodies against G␤, PP1, phosphoinositide-specific phospholipase C (PLC) ␤1, phosphoinositide 3-kinase (PI 3-kinase), ␤-arrestin, G␣ i , G␣ s , and G␣ q were purchased from BD Transduction Laboratories. Mouse monoclonal anti-M2R, rabbit polyclonal anti-K ϩ channel GIRK1 antibodies, and normal mouse IgG were from Oncogene Research Products (Boston, MA). Other antibodies used in this study were as follows: the rabbit polyclonal anti-M2R and rabbit polyclonal anti-Ca v 1.2a (Alomone Labs, Jerusalem, Israel); mouse monoclonal 4A9 anti-phosphoserine (Calbiochem); rabbit polyclonal anti-actin, rabbit polyclonal anti-PLC ␤2 (Q-15), and rabbit polyclonal anti-adenylyl cyclase (R-32) (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Horseradish peroxidase-linked mouse and rabbit whole antibodies, ECL Plus reagents, and Rainbow Molecular Weight Markers were purchased from Amersham Biosciences. Dynabeads M-280 sheep anti-mouse IgG, M-280 sheep anti-rabbit IgG, and M-450 rat anti-mouse IgM were from Dynal Biotech Inc. (Lake Success, NY). Immobilon-P SQ transfer membrane was purchased from Millipore (Bedford, MA). Protease inhibitor mixture set III (PI) and phosphatases inhibitor mixture set I (PhI) were purchased from Calbiochem. N-Glycosidase F was purchased from Prozyme (San Leandro, CA). Electrophoretic and Western blotting reagents were provided by Bio-Rad. All other chemicals used in this study, including deoxycholic acid sodium salt and IGEPAL CA-630, were of molecular biology grade and were purchased from Sigma.
Electrophysiological Recordings-Viable atrial myocytes were obtained from 1-to 2-day-old Sprague-Dawley rats by a trypsin/chymotrypsin/elastase dissociation procedure and cultured as described previously (10). Unitary currents through K ACh channels were recorded from cell-attached and inside-out patches using standard high resolution patch clamp methods (26) as described previously (10,11). The composition of the bath solution was as follows (in mM): 150 KCl, 5 EGTA, 5 glucose, 1.6 MgCl 2 , 5 HEPES, pH 7.4; the composition of the pipette solution was as follows (in mM): 150 KCl, 1 CaCl 2 , 1.6 MgCl 2 , 5 HEPES, pH 7.4. Only recordings from a single GIRK1/4 channel or a pair of GIRK1/4 and scGIRK channels (11), residing in the same patch, were included in the analysis of GIRK1/4 channel gating behavior. In the latter case, the two types of GIRK channels were investigated as individual entities, and the gating equilibrium of GIRK1/4 channel was analyzed independently of the presence of a scGIRK channel in the same patch.
The modal behavior of M2R-activated GIRK1/4 channels was analyzed in the context of a model that assumes four functional modes of a single GIRK channel arising from different occupancy of four equivalent and independent G␤␥ sensors in the channel (10,11). Briefly, single channel recordings were divided into 400-ms consecutive segments, and the frequency of apparent openings, f, was calculated for each data segment. Initially, the data recordings were examined for homogeneity, by averaging the mean frequency, F, over 30-s data recordings (from 75 consecutive data segments) and plotting the F values versus time, t. The resulting F Ϫ t plots accurately identified the apparently homogeneous sections of each recording (11). A data recording was judged as homogeneous if the mean F values calculated from the first and the second halves of this recording were statistically indistinguishable. The f values calculated from the individual 400-ms data segments in a homogeneous section of the recording were further used to generate a frequency histogram. The f histograms were subsequently fitted by the sum of several gamma distributions, ⌫ k (f) as shown in Equation 1 and as described previously (11), where ⌫ k (f) is the distribution expected for the sum of two variables, when each variable has an exponential distribution, (27). Previously, we have shown that f histograms generated from G␤␥-activated GIRK channels were fitted rather well by a sum of gamma terms over a wide range of G␤␥ concentrations (11). The f histograms were further used to determine the number of gating modes accessible to the GIRK channels, the characteristic f k values, and the relative occupancy of each mode, F k (F k ϭ N k /⌺N k ). For each f histogram, the probability of G␤␥ binding, P, was computed from the relative occupancy of mode 1, F 1 , according to Equation 2, The relative occupancy of the remaining modes, F k , was then examined as a function of the parameter P to verify that experimental F k estimates were consistent with the theoretically derived probability of observing mode k as shown in Equation 3, where P is the probability of G␤␥ binding to one of four putative G protein sensors in GIRK1/4 channel structure, and N ϭ 4.
For each experimental condition, the fraction of blank data segments caused by GIRK1/4 channel desensitization, F D , was calculated as the difference, (F 0E Ϫ F 0T ), where F 0E is the experimentally determined fraction of blank data segments, and F 0T is the predicted fraction of time the channel spends with no G␤␥ bound to it. The F 0T fraction, F 0T ϭ (1 Ϫ P) 4 , was calculated from the probability of G␤␥ binding, P, according to Equation 3.
Immunoprecipitation and Immunoblotting-The composition of GIRK1 signaling complexes was determined as described previously (4), with some modifications. Briefly, rat atrial tissue was homogenized in HME buffer containing 20 mM HEPES-KOH, pH 8.0, 2 mM MgCl 2 , 1 mM EDTA, supplemented with 0.32 M sucrose, PI and PhI. The homogenate was centrifuged at 500 ϫ g for 15 min at 4°C, and the supernatant was centrifuged further at 100,000 ϫ g for 45 min. The resulting crude membrane pellet was washed, resuspended in HME-PI-PhI buffer containing 0.32 M sucrose, quickly frozen in liquid nitrogen, and stored in aliquots at Ϫ80°C until used. The crude atrial membranes contained both the core p54 and glycosylated p65 forms of GIRK1 protein (7), migrating at 54 and 65 kDa, respectively. For the experiments illustrated in Figs. 8 and 9, the membrane fraction was further purified as follows. The atrial tissue homogenate was first centrifuged at 10,000 ϫ g for 15 min at 4°C to remove the nuclei and mitochondria, and plasma membranes were then obtained by centrifugation (100,000 ϫ g for 65 min) of the resulting supernatant on a cushion of HME-PI-PhI buffer with 1.3 M sucrose (28). The membranes at the interface were collected, washed twice, resuspended in HME-PI-PhI buffer containing 0.32 M sucrose and frozen as above. This modification of membrane preparation yielded plasma membranes containing the glycosylated p65 GIRK1 form only.
For the experiments involving stimulation of M2Rs, excised atria were equilibrated in oxygenated solution containing (in mM) 140 NaCl, 5.4 KCl, 1 CaCl 2 , 1.6 MgCl 2 , 5 HEPES, 5 glucose, pH 7.4, stimulated with 1 M acetylcholine (ACh) for 5 min at room temperature and frozen in liquid nitrogen. To determine the effect of concurrent stimulation of M 2 and ␣-adrenergic receptors on the composition of the GIRK1 complexes, atria were preincubated with ␤-AR antagonist propranolol (1 M) for 5 min at room temperature, then treated with phenylephrine at a concentration of 1 M for 5 min, followed by stimulation with 1 M ACh for an additional 5 min in the presence of 1 M phenylephrine. After a brief rinse with ice-cold physiological solution, the tissue was frozen in liquid nitrogen. Unstimulated tissue processed in parallel was used as a control.
The atrial membranes were solubilized in MEB buffer (20 mM HEPES, pH 8.0, 2 mM MgCl 2 , 1 mM EDTA, 100 mM NaCl, 10 mM ␤-mercaptoethanol, 1% sodium deoxycholate, and 0.45% IGEPAL CA-630) with PI and PhI as described previously (4,10). The detergent-insoluble material was removed by centrifugation at 100,000 ϫ g for 1 h, and the supernatant, containing the solubilized membrane-associated proteins, was immediately used for immunoprecipitation. Protein concentration was determined by Bio-Rad Dc protein assay against a standard curve of bovine serum albumin. Equal amounts of membrane proteins were used in each reaction.
Immunoprecipitations were performed with Dynal magnetic beads according to the manufacturer's protocols using the MPC magnetic device (Dynal, Lake Success, NY) as described previously (4). The magnetic immunopurification eliminates all centrifugation steps from the classical immunoprecipitation process. As a result, any possible contamination of the immunoprecipitation product by large protein complexes that might be trapped within antibody-protein A/G beads latticework during centrifugation is also eliminated. Briefly, equal amounts of beads were incubated with the anti-phosphoserine (4A9 clone) or anti-G␤ antibodies (3 g per 1.2 ϫ 10 7 M-450 rat anti-mouse IgM beads) or anti-M2R antibody (Oncogene) (3 g per 1.8 ϫ 10 7 M-280 sheep anti-mouse IgG beads) for 2 h with rotation at 4°C. In each experiment, equal amounts of beads coated with normal mouse IgG (3 g per 1.8 ϫ 10 7 M-280 sheep anti-mouse IgG beads) were used as a negative control. In addition, the specificity of the antibodies used for immunoprecipitation was tested in each experiment to ensure that these antibodies recognized only the protein of interest in the atrial membranes. Coated beads were washed and incubated with the solubilized membrane proteins for 2 h at 4°C with rotation. The complexed beads were collected by magnetic precipitation, washed, resuspended in equal volumes of Laemmli sample buffer, and heated for 5 min at 95°C (70°C for M2R detection). Equal sample volumes were loaded on an 11% gel; the proteins were sep-arated by Tris-glycine SDS-PAGE and electrophoretically transferred to Immobilon-P SQ membranes. The proteins of interest were detected by Western blotting, and the immunoreactive complexes were visualized by enhanced chemiluminescence detection system. The immunoreactive protein bands were quantified using VersaDoc Imaging System and Quantity One (version 4.5.2) software (Bio-Rad). The changes in the composition of GIRK1 complexes after receptor stimulation were assayed as follows. The pixel intensities (arbitrary units) of the immunoreactive bands for each protein were quantified and normalized to the intensity of the corresponding protein band in the control sample. The ratio between the normalized amount of a particular protein and the normalized amount of GIRK1 protein in the same complex was then determined and the mean Ϯ S.E. of four independent experiments reported. Finally, to verify that the protein bands of interest were properly quantified, in each experiment, a range of known amounts of atrial membrane proteins was also loaded on the same gel, and the relative protein levels were examined to confirm that pixel intensity of the immunoreactive bands grows linearly with increasing protein amount.

Effects of ␣-AR Stimulation on GIRK1/4 Channel Function-
Numerous studies, involving different cell types, have indicated that activation of ␣ 1 -adrenergic or other G q -coupled receptors leads to inhibition of whole-cell GIRK channel currents (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25). To investigate the molecular mechanisms underlying such inhibition, we focused on GIRK1/4 channel function at the single molecule level and examined this function in cell-attached configuration following concurrent stimulation of M 2 and ␣-adrenergic receptors in atrial myocytes. M2Rs activate two populations of GIRK channels in atrial myocytes, canonical 35-picosiemen GIRK1/4 channels and small conductance (sc) GIRK channels with a conductance of ϳ16 picosiemens (11). This report focuses on ␣-AR regulation of the classical GIRK1/4 channels, because they are the major contributors to the inhibitory synaptic transmission in the heart.
Unitary currents, recorded from two different atrial myocytes in the presence of 1 M acetylcholine, added to the pipette solution, are shown in Fig. 1. The M2R-activated channels exhibited inward rectification and direct dependence on G␤␥ for their activation. Patch excision in GTP-free solution abruptly and completely eliminated channel activity, whereas subsequent perfusion of the cytoplasmic side of the patch with nanomolar concentration of purified recombinant G␤ 1 ␥ 7 brought the activity back (data not shown). For each cell-attached experiment, data acquisition began 30 s after formation of a gigaseal. This delay was necessary for adjustment of holding potential and the appropriate gain setting of the patch clamp amplifier. As illustrated in Fig. 1, the sensitivity of individual GIRK1/4 channels to M2R stimulation varied from one channel to another. Regardless of such variability, channel gating remained stable over the course of the cell-attached recordings (usually 7-10 min).
In contrast, activation of atrial ␣-ARs changed the profile of GIRK channel response to M2R stimulation. In whole-cell assays, inactivation of GIRK channels by G q -coupled receptors was quantified as the ratio of steady-state current, after application of specified G q -agonist, to the control steady-state current evoked in the same cell (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25). In our experiments, however, attempts to record M2R-activated single channels for several minutes and then perfuse the experimental chamber with an agonist frequently led to patch excision, because of the mechanical perturbations associated with the perfusion. For this reason we reversed the order of M 2 and ␣-adrenergic receptor activation. The atrial myocytes were incubated with ␤-AR antagonist, propranolol (1 M), for 5 min and then treated with ␣-AR agonist phenylephrine (1 M) for additional 5 min. At the end of the 5th min, the electrophysiological experiments were conducted under conditions identical to the conditions in the control experiments (1 M ACh was added to the patch pipette). Propranolol and phenylephrine were present in the experimental chamber for the duration of the experiment. Fig. 2 illustrates the activity of a GIRK1/4 channel recorded in cell-attached configuration, when ␣-ARs and M2Rs are concurrently activated. The initial response of GIRK1/4 channels to M2R stimulation in the phenylephrine-treated myocytes was similar to the response in the control group. Combined activation of ␣-ARs and M2Rs, however, led to extensive inactivation of GIRK1/4 channels within less than 5 min.
Gating Mechanism of M2R-activated GIRK1/4 Channels-Understanding the principles of integration of opposing stimuli by GIRK1/4 channels required a quantitative analysis of chan- nel gating. We therefore analyzed the kinetic behavior of GIRK1/4 channels in control and phenylephrine-treated myocytes. We have previously shown that upon G␤␥ activation, atrial GIRK1/4 channels undergo fast transitions between four functional modes rendered active by the binding of an increasing number of G␤␥ subunits to the channel (10,11). This G␤␥dependent GIRK channel behavior is further complicated by slow modal transitions, which are independent of G␤␥ concentration (11). The slow modal transitions, however, occur on a time scale ϳ1000-fold slower than the transitions between the four functional modes of GIRK1/4 channel. Such heterogeneity of channel behavior significantly hinders the interpretation of the single-channel recordings. We therefore developed quantitative approaches for detecting homogeneous sections of recordings and classification of heterogeneous channel gating into functionally distinct modes (10,11).
Initially, we examined the recordings from M2R-activated GIRK1/4 channels for the presence of slow modal transitions. Each recording was divided into 400-ms consecutive segments, and the frequency of openings, f, was calculated for individual data segments. The mean frequency, F, was then calculated for 30-s intervals of data recordings (by averaging the f values from 75 consecutive data segments) and plotted against time. Analysis of GIRK1/4 channel activity recorded from seven control and seven phenylephrine-treated myocytes is presented in Fig.  3. The mean frequency of channel gating measured in cell-attached patches from control myocytes during the first 60 s of recording was 19.9 Ϯ 2.5 Hz (n ϭ 7). This value is statistically indistinguishable from the frequency estimate obtained for the GIRK1/4 channels in phenylephrine-treated myocytes for the first 60 s of recording (18.2 Ϯ 2.3 Hz, n ϭ 7). However, inactivation of M2R-activated GIRK1/4 channels in the control myocytes was virtually absent for the duration of our experiments; the mean frequency of channel gating measured 4 min after the onset of the cell-attached recording was 18.9 Ϯ 2.5 Hz (n ϭ 7). In contrast, M2R-activated GIRK1/4 channels in phenylephrine-treated myocytes showed a prominent inactivation under identical recording conditions; the mean frequency was only 4.8 Ϯ 1.6 Hz (n ϭ 7) 4 min after the beginning of the recording. Thus, stimulation of ␣-ARs in atrial myocytes controlled differentially the amplitude of the initial M2R-activated response and GIRK1/4 channel inactivation. The initial amplitude of the response was unaffected (Fig. 3), whereas the channel inactivation was enhanced from 8.4% of the initial response in control myocytes to 74.4% in phenylephrine-treated myocytes.
Having generated the F Ϫ t plots for the individual singlechannel recordings, we compiled the f values within the apparently homogeneous section of each recording to construct a frequency histogram. The f histograms were then fitted with gamma functions to quantify the equilibrium among the gating modes of M2R-activated GIRK1/4 channels. Fig. 4 illustrates these two stages in the analysis of the M2R-activated GIRK1/4 channels shown in Fig. 1. The F Ϫ t plots (Fig. 4, A and B, top panels) verified the homogeneity of the channel activity upon M2Rs activation. More importantly, our analysis revealed that the f histograms (Fig. 4, A and B, bottom panels) were amenable to fitting with gamma functions, suggesting that M2R-activated GIRK1/4 channels have gating properties similar to those of G␤␥-activated channels. Fig. 5 illustrates the profile of GIRK1/4 channel activity in a phenylephrine-treated myocyte (the unprocessed data shown in Fig. 2), as revealed by analysis of mean frequency of gating F (top panel). The F Ϫ t plot accurately detected the transition from high to low efficient GIRK1/4 channel gating, triggered by the concurrent activation of ␣-adrenergic and M 2 receptors in the membrane. To obtain quantitative estimates of channel gating, we generated separate f histograms (Fig. 5, middle and bottom panels) from the initial and late phase (denoted by the horizontal lines in the F Ϫ t plot) of the recording. A comparison of these histograms revealed that the inactivation of GIRK1/4 channel was associated with a shift in the modal equilibrium toward the less efficient gating modes. Previously we have shown that similar shifts in modal equilibrium of G␤␥activated channels are predictive of reduction in the probability of G␤␥ binding to the channel (11). Receptor-mediated activation of the GIRK channels, however, relies on conformational changes in heterotrimeric G proteins, triggered upon agonist binding to the M2Rs. Accordingly, the interactions between M2Rs and endogenous G proteins theoretically can modulate the gating equilibrium of GIRK channels. We therefore examined whether the gating of M2R-activated channels differs from that of G␤␥-activated channels. To address this question, we analyzed the gating equilibrium of GIRK1/4 channels activated at 1 M ACh. The probability of G␤␥ binding to the channel, P, was computed from the relative occupancy of gating mode 1, assuming this mode ensues from the binding of a single G␤␥ to the channel (see "Experimental Procedures"). The relative occupancy of the remaining modes was then examined as a function of P and compared with the theoretically derived distributions for the probability of binding of increasing numbers of G␤␥ subunits to the channel. Fig. 6 illustrates this analysis for the two sets of experimental data for M2R-activated GIRK channels from control (A) and phenylephrine-treated myocytes (B). In the latter case, the probability of G␤␥ binding to the channel, P, was derived from the f histograms generated from the low efficient phase of GIRK1/4 channel gating. The curves labeled "mode 2" and "modes 3 and 4" show the probability of binding of 2, and 3 and 4 G␤␥ subunits to the channel, respectively, as calculated from the binomial distribution (see Equation 3). The theoretical curves superimposed on the experimental data for the occupancy of mode 2 (Fig. 6, diamonds) and modes 3 and 4 (triangles) illustrate the compatibility between Although the single-channel activity was recorded from different atrial myocytes, the data recorded under identical experimental conditions were combined on the same grid to illustrate the global effect of ␣-ARs on the activity of GIRK1/4 channels and membrane hyperpolarization. The initial response of the GIRK1/4 channels in the phenylephrine-treated group was undistinguishable from the response in the control group. However, six of seven GIRK channels in the phenylephrine-treated group exhibited a prominent inactivation within 5 min of concurrent ␣-AR and M2R stimulation. In contrast, the gating of the GIRK channels in the control group remained stable over the course of the recording. The frequency of channel gating for GIRK channels 1 and 2 in the control group (left panel) was derived from the initial 7 min of GIRK1/4 channel recordings illustrated in Fig.  1, A and B, respectively. The F Ϫ t plot for GIRK channel 2 in the phenylephrine-treated group (right panel) was derived from the initial 7 min of GIRK1/4 channel recording illustrated in Fig. 2. the model and the experimental observations. Altogether, the analysis of modal equilibrium confirmed that the M2R-activated GIRK1/4 channels follow the predictions of the model in which incremental binding of G␤␥ to four equivalent and independent sensors in the protein complex confers the four functional modes of the channel.
In the framework of this model, we next examined the probability of G␤␥ binding to GIRK1/4 channels after M2R activation and the effect of ␣-ARs on G␤␥ binding. At 1 M ACh, the probability of G␤␥ binding for M2R-activated GIRK channels from control myocytes ranged from 0.36 to 0.63. In contrast, the P values for channels from phenylephrine-treated myocytes were clustered in the range of 0.09 -0.27. The mean probability of G␤␥ binding to the channel (at 1 M ACh) was 0.55 Ϯ 0.03 for the control (n ϭ 7) and only 0.23 Ϯ 0.07 for the phenylephrine-treated myocytes (n ϭ 7). These results identified the reduction in the probability of G␤␥ binding to GIRK1/4 channels as a major mechanism underlying the ␣-AR inhibition of these channels.
Although the analysis of both G␤␥-activated and M2R-activated GIRK1/4 channels overwhelmingly supported the model of channel gating in four conformations rendered active by the binding of an increasing number of G␤␥ subunits to the channel, this model failed to predict accurately the probability of observing the nonconducting conformations of the channel. Thus, upon G␤␥ activation, the proportion of blank data segments in GIRK1/4 channel recordings significantly exceeded the probability of observing gating mode 0 (no G␤␥ bound to the channel) calculated from the binomial distribution (11). This discrepancy strongly suggests that an additional G␤␥-dependent desensitization process might contribute to the efficiency of GIRK1/4 channel gating. Because this process could prove vital in channel modulation, we examined GIRK1/4 channel desensitization in control and phenylephrine-treated myocytes. The fraction of desensitized channels, F D , was defined as the difference (F 0E Ϫ F 0T ) between the experimentally determined fraction of blank data segments, F 0E , and the predicted fraction of time the channel being with no G␤␥ bound, F 0T , calculated from the probability of G␤␥ binding, P, according to Equation 3. At 1 M ACh, F D was 9.82 Ϯ 2.88% for the control (n ϭ 7) and 19.07 Ϯ 3.82% for the phenylephrinetreated myocytes (n ϭ 7). Altogether, these results establish that parallel activation of ␣-ARs in atrial myocytes imposes a unique modulatory "signature" on M2R-activated GIRK1/4 channels, affecting both the probability of G␤␥ binding to the channel and its desensitization.
Effect of Receptor Stimulation on the Composition of GIRK1-G␤ Signaling Complexes-Having established the effect of ␣-AR stimulation on GIRK1/4 channel function in vivo, we asked whether channel modulation could be correlated with concurrent changes in channel signaling environment in vitro.
To address this question we determined the composition of immunopurified GIRK1-G␤ signaling complexes after coactivation of M 2 and ␣-adrenergic receptors. Fresh atrial tissue was first preincubated with propranolol for 5 min, followed by stimulation with 1 M phenylephrine for 5 min, and then stimulated with 1 M ACh for an additional 5 min. This protocol accurately replicates the receptor stimulation protocol from our electro- physiological experiments. Two additional samples, one incubated with vehicle for the ligand only (control) and another one stimulated with 1 M ACh for 5 min, were processed in parallel. Following receptor stimulation, the GIRK1-G␤ signaling complexes were initially immunopurified using the phosphoserinespecific 4A9 antibody. Previously, we have shown that the 4A9 antibody selectively detected the core p54 form of GIRK1 protein in rat atrial membranes, and this antibody could be used for immunopurification of the oligomeric GIRK1-G␤ complexes (4). The resulting 4A9 protein complexes were analyzed by immunoblotting for the presence of GIRK1, G␤, actin, and PP1. Fig. 7 (top left) illustrates the agonist-mediated changes in the composition of the 4A9 complexes. To quantify these changes, the amount of each protein was normalized to the amount of the same protein in the control 4A9 complexes. Thus, the relative amount of each component was set at a value of unity in the absence of receptor stimulation. Because both GIRK1 isoforms, the core p54 and the glycosylated p65 form, were present in the 4A9 complexes (Fig. 7, top left panel), we used the sum of the pixel intensities of the 54-and 65-kDa bands for quantification of the amount of GIRK1 protein in these complexes. To confirm that the sum of the intensities of these two bands accurately reflects the amount of total GIRK1 protein, we allowed N-glycosidase F to completely deglycosylate the p65 form of GIRK1, and then we quantified the intensity of a single GIRK1/ p54 band. We observed that the total amount of GIRK1 protein in the 4A9 complexes was a rather stable parameter in all three experimental conditions (data not shown).
To examine the stoichiometry of the 4A9 complexes after receptor stimulation, the relative amount of each protein was normalized to the amount of total GIRK1 protein in the same complexes. The results from this analysis for G␤, actin, and PP1 are illustrated in Fig. 7 (dark bars). A comparison of G␤ immunoreactive bands in the 4A9 complexes yielded data consistent with the results from the analysis of GIRK1/4 channel gating (Fig. 6C). The level of G␤ increased by 2.40 Ϯ 0.30-fold (n ϭ 4) following the M2R stimulation but only by 1.54 Ϯ 0.15-fold after concurrent stimulation of M 2 and ␣-adrenergic receptors. Furthermore, the levels of actin and PP1 increased (1.77 Ϯ 0.22fold and 1.60 Ϯ 0.09-fold, respectively, n ϭ 4) after M2R stimulation; however, these protein levels were not affected by the simultaneous activation of M 2 and ␣-adrenergic receptors. These results suggest the intriguing possibility that the signaling environment of GIRK channels undergoes dynamic transformations upon receptor stimulation.
To confirm that the identified changes in the 4A9 complexes were a consequence of receptor stimulations, we immunopuri- fied the oligomeric GIRK1 complexes using G␤-specific antibodies as an alternative "bait" and examined their composition. The selection of a G␤ antibody for these experiments was based on information about the immunogen used to generate the antibody and localization of this immunogen in the three-dimensional G␤␥ structure (29,30). According to the supplier (BD Transduction Laboratories), the monoclonal G␤ antibody was generated against G␤ residues 130 -145. This immunogen includes a surface-localized residue, Thr-143, essential for the interaction between G␤␥ and G␣-GDP that is not accessible in heterotrimeric G proteins (29,30). We therefore hypothesized that the G␤ antibody should be able to recognize specifically the native G␤␥ subunits but not the heterotrimeric G␣␤␥ proteins. Furthermore, the same G␤ residue (Thr-143) is required for G␤␥ interaction with PLC ␤2 and G protein-coupled receptor kinase but not for G␤␥ activation of GIRK1/4 channels and adenylyl cyclase (31). Accordingly, the G␤ antibodies should be able to bind G␤␥-GIRK1/4 channel complexes but not G␤␥-PLC ␤2 and G␤␥-G protein-coupled receptor kinase complexes. We have previously established that the G␤-complexes, immunopurified from atrial membranes, indeed consist of the same set of proteins found in the 4A9 complexes (4). Additionally, comparison of the stoichiometry of G␤ and 4A9 complexes in the present study (Fig. 7, light and dark bars, respectively) revealed a remarkable consistency in the composition of these complexes.
Finally, to rule out the possibility that other G␤␥-binding proteins were associated with the G␤ and 4A9 complexes, we analyzed these complexes by immunoblotting for the presence of G␣ i , G␣ s , adenylyl cyclase, PLC ␤1, PLC ␤2, PI 3-kinase, and ␤-arrestin. We have not been able to detect association of these proteins with both the G␤ and 4A9 complexes isolated from atrial membranes (data not shown). Notably, selective binding of G␣ i/o (32) and adenylyl cyclase (13,14) to neuronal GIRK channels has been well documented in heterologous expression systems and brain. The absence of G␣ i/o and adenylyl cyclase in the atrial GIRK1 oligomeric complexes most probably reflects the ability of different cell types to assemble divergent signaling networks.
Taken together, the analysis of the G␤ and 4A9 complexes indicates that both antibodies selectively recognize the native GIRK1-G␤ complexes in atrial membranes. Accordingly, we can accurately identify receptor-induced changes in the GIRK1 protein environment using either G␤or 4A9protein complexes.
The presence of two GIRK1 isoforms in the complexes isolated from crude membrane preparations, however, introduces ambiguity in the interpretation of our data, because the two forms may localize to different membrane compartments and associate with different protein partners. To verify that the oligomeric GIRK1 complexes are relevant to the physiological GIRK1/4 channel function, we further purified the atrial membranes to yield a fraction enriched in plasma membranes. This fraction contained the p65 form of GIRK1 only (Fig. 8, top left  panel). We directly confirmed the identity of the p65 GIRK1 form by deglycosylation of the membrane proteins with N-glycosidase F, prior to electrophoresis. This treatment converted the p65 form of GIRK1 (Fig. 8, Load) into the core p54 form (Fig. 8, Load*). Neither M2R stimulation nor coactivation of M 2 and ␣-adrenergic receptors triggered changes in the membrane density of the GIRK1/p65 protein (Fig. 8, top left panel, Load,  lanes 1-3).
Next, we sought to purify the GIRK1/p65-G␤ oligomeric complexes and to determine the effect of receptor stimulation on their composition. The 4A9 antibodies do not recognize the glycosylated form of GIRK1 on Western blots (4). Furthermore, these antibodies exhibit very low affinity for the native GIRK1/ p65-G␤ complexes and immunoprecipitate less than 2% of the GIRK1/p65 protein from the plasma membranes (data not shown). In contrast, the products of parallel immunoprecipitations using G␤ antibodies contained ϳ25% of the GIRK1/p65 protein (Fig. 8, Load, lanes 1-3 versus lanes 7-9; IP, anti-G␤). The specificity of G␤ association with GIRK1/p65 protein was further confirmed by the absence of other G␤␥ binding proteins, G␣ i , G␣ s , adenylyl cyclase, PLC ␤1, PLC ␤2, PI 3-kinase, and ␤-arrestin, in the G␤-protein complexes. We therefore used the G␤ antibodies for immunopurification of the oligomeric GIRK1/p65 complexes from the . Crude membrane lysates were incubated with Dynal beads coated with antibodies against phosphoserine (clone 4A9), which specifically recognizes the core p54 form of GIRK1 protein in atrial membranes. The immunoprecipitation (IP) products were collected, and the components were separated by 11% SDS-PAGE. Proteins were transferred to a polyvinylidene difluoride membrane, and sections were probed with the antibodies indicated to the left (Blot). Parallel immunoprecipitation reactions using equal amounts of Dynal beads coated with G␤-specific antibodies yielded oligomeric complexes with composition similar to that of the 4A9 immunocomplexes (not shown). Bar graphs, stoichiometry of 4A9 and G␤ immunocomplexes. The protein levels of G␤, actin and PP1 were quantified and normalized to the level of total GIRK1 protein in the oligomeric complexes (see "Experimental Procedures"). The relative changes in the protein levels are averaged from four independent experiments, and the stoichiometry of 4A9 (dark bars) and G␤ immunocomplexes (light gray bars) is shown. In each experiment, equal amounts of Dynal beads coated with normal mouse IgG were incubated with the membrane lysates as a negative control. None of the protein bands observed in the 4A9 and G␤ immunoprecipitates were detected in the products from the parallel immunoprecipitations using normal mouse IgG (data not shown).
atrial plasma membranes. Once again, analysis by immunoblotting revealed receptor-dependent changes in the composition and stoichiometry of multimeric GIRK1 complexes (Fig.  8, bar graphs). We found a significant increase in G␤ association with GIRK1/p65 protein 5 min after M2R stimulation (1.67 Ϯ 0.20-fold, n ϭ 4, p Ͻ 0.05; Fig. 8, bottom left). This increase, however, could be blocked 5 min after concurrent stimulation of M 2 and ␣-adrenergic receptors (0.89 Ϯ 0.14 of the control, n ϭ 4; Fig. 8, bottom left). Additionally, we found a significant increase in actin association with the GIRK1/ p65-G␤ complexes after M2R stimulation (1.50 Ϯ 0.12-fold, n ϭ 4, p Ͻ 0.05), and this increase was not eliminated by the simultaneous activation of M 2 and ␣-adrenergic receptors (1.48 Ϯ 0.10-fold, n ϭ 4, p Ͻ 0.05; Fig. 8, top right). Finally, the profile of PP1 association with the multimeric complexes (Fig. 8, bottom right) followed the pattern observed for the G␤ protein. The level of PP1 was increased by 1.94 Ϯ 0.27fold (n ϭ 4, p Ͻ 0.05) after M2R activation; however, it was unaffected (1.01 Ϯ 0.14 of the control, n ϭ 4) after concurrent stimulation of M 2 and ␣-adrenergic receptors.
Because direct interactions between G␣ q and the neuronal GIRK channels, as well as between G␣ q and the heterologously expressed GIRK1 and GIRK4 subunits, have been recently shown (33,34), we investigated whether G␣ q associates with the GIRK1-G␤ complex following ␣-AR stimulation. G␣ q was detected as a single band of 42 kDa in the atrial membranes. Immunoblotting of the GIRK1-G␤ complexes for G␣ q did not detect G␣ q association with the oligomeric complexes (Fig. 8, top left). The existence of G␣ q -GIRK complexes in the atrial membrane, however, cannot be entirely ruled out, because some protein interactions could have been disrupted by the detergent solubilization of the membrane proteins.

In Vitro Kinetic Assay for Monitoring Changes in GIRK1 Oligomeric
Complexes-Altogether, our data indicate that receptor stimulation triggers changes in the signaling environment of GIRK channels, and these changes can be accurately monitored by analysis of GIRK1-G␤ complexes. This analysis, however, provided only a static picture of the events following GPCR stimulation. In addition, electrophysiological recordings of the GIRK channel activity have shown that the channel open probability approaches zero in the absence of ACh in the pipette (data not shown). On the contrary, our biochemical experiments revealed a significant association between G␤ and GIRK1, even when the oligomeric complexes were immunopurified from control atrial tissue in the absence of receptor stimulation. This apparent discrepancy between the control electrophysiological and biochemical experiments can be attributed to a detergent-induced release of G␤␥ from the heterotrimeric G proteins along the steps of membrane solubilization and immunoprecipitation, even in the absence of a receptor agonist. We therefore sought to develop a biochemical assay, which would allow us to monitor dynamic changes in the composition of GIRK1-G␤ complexes regardless of receptor-independent generation of G␤␥. We hypothesized that a population of GIRK1-G␤ complexes, formed under equilibrium conditions, can be captured on magnetic beads and then modified by brief application of receptor agonists or antagonists for different GPCRs present in the reac- Equal amounts of plasma membrane proteins were incubated with equal amounts of Dynal beads coated with either G␤-specific antibodies or normal mouse IgG as a negative control. The immunoprecipitation products were collected, and the components were separated by 11% SDS-PAGE. Plasma membrane proteins were run on the same gel (lanes 1-3, Load). The products from the immunoprecipitations (IP) using normal mouse IgG were loaded in lanes 4 -6; and the G␤-immunocomplexes were loaded in lanes 7-9. The segments probed for G␣ q , actin, PP1, G␤, and IgG contain 5% of each membrane lysate used for immunoprecipitation, whereas the segments probed for GIRK1 contain 20% of the lysates before (Load) and after (Load*) treatment with N-glycosidase F. Proteins were transferred to a polyvinylidene difluoride membrane, and sections were probed with the antibodies indicated to the left (Blot). Bar graphs, stoichiometry of GIRK1/p65 complexes. The G␤ immunoprecipitation products of four independent experiments were analyzed; the protein levels of GIRK1, G␤, actin, and PP1 were quantified and normalized to the level of GIRK1 (p65) protein in the oligomeric complexes. tion system. If an external stimulus affects the composition of the GIRK1-G␤ complexes, we could examine this process with high temporal resolution, because the beads together with the bound protein complexes can be separated instantly from the remaining components of the reaction mixture. To test our premise, we immunopurified the GIRK1/p65-G␤ complexes in three identical and parallel reactions and examined their composition after three different treatments (Fig. 9A). Beads coated with G␤-specific antibodies were incubated with solubilized plasma membrane proteins from control atria for 2 h at 4°C. At the end of the incubation, the solubilized atrial membranes were supplemented either with a vehicle for the agonists (Fig. 9A, 1, Control) or a mixture of 40 M Mg-ATP, 1 M GTP, and 1 M propranolol (Fig. 9A, 2, Mg-ATP ϩ GTP) for 20 min at 4°C. In the third reaction, the solubilized membranes were stimulated with a mixture of 1 M ACh and 1 M phenylephrine for 10 min in the presence of Mg-ATP, GTP, and propranolol (Fig. 9A, 3, ACh ϩ Phe). At the end of each reaction, the GIRK1-G␤ complexes were quickly separated from the rest of the membrane proteins and washed, and their composition was determined by immunoblotting. The association of GIRK1 with G␤, actin, and PP1 was not affected by the presence of Mg-ATP, GTP, and propranolol in the reaction mixture (n ϭ 3; Fig. 9A, Blot, lanes 7 and 8). At the same time, the simultaneous activation of the M 2 and ␣-adrenergic receptors by ACh and phenylephrine resulted in an ϳ50% reduction in the association of G␤ and PP1 with GIRK1 (n ϭ 3, Fig. 9A, Blot, lane 9). Equally important is the fact that the receptor activation did not affect the amount of GIRK1 protein associated with the magnetic beads. Because the GIRK1-G␤ complexes are bound through G␤-specific antibodies to the beads, our biochemical data are consistent with the scheme illustrated in Fig.  9A. In this scheme, each GIRK channel molecule binds more than one G␤␥; one of the G␤␥ subunits is sufficient to tether the GIRK channel to the bead, whereas the activation of different receptor types can regulate the number of G␤␥ subunits bound to the remaining G␤␥ sensors in the channel. Altogether, our experiments confirmed the premise that the receptor activation modulates the composition of the preformed GIRK1-G␤ complexes. Furthermore, we revealed features of this modulation that are independent of the integrity of the cellular structures and can be studied in a preparation of isolated membrane proteins.
Having established the feasibility of the new assay, we used the same approach to monitor the temporal pattern of G␤ and PP1 dissociation from GIRK1 after concurrent stimulation of M 2 and ␣-adrenergic receptors (Fig. 9B). Once again, the  3). The segment probed for actin, PP1, and G␤ contains 5% of the membrane lysates used for immunoprecipitation, whereas the segment probed for GIRK1 contains 20% of these lysates. The products from three parallel immunoprecipitations using Dynal beads coated with normal mouse IgG were treated in the same manner and loaded in lanes 4 -6. B, GIRK1-G␤ complexes were bound under equilibrium conditions to Dynal beads coated with G␤-specific antibodies in five identical and parallel reactions, and the solubilized membranes were then treated with a mixture of 1 M ACh and 1 M phenylephrine for different time intervals (1, 2, 5, 10, and 20 min, respectively) in the presence of Mg-ATP, GTP, and propranolol. At the end of each reaction, the GIRK1-G␤ complexes were separated by magnetic immunoprecipitation from the remaining membrane proteins. The resulting oligomeric complexes were analyzed by immunoblotting for the presence of GIRK1, G␤, actin, and PP1 (top panel). Lanes 1-5 correspond to 1, 2, 5, 10, and 20 min of ␣AR and M2R stimulation. The intensities of the protein bands were quantified as before; however, the relative amount of each protein was normalized to the level of the same protein in the GIRK1-G␤ complexes obtained after 1 min of concurrent stimulation of M 2 and ␣-adrenergic receptors (instead in the control complexes) for consistency with the electrophysiological experiments illustrated in Fig. 3. After a 10-min stimulation of M 2 and ␣-adrenergic receptors, the relative amount of actin associated with GIRK1-G␤ complexes was 85.5 Ϯ 4.5% of the amount associated after a 1-min stimulation (middle panel, n ϭ 4), whereas the relative amount of PP1 was only 44.7 Ϯ 3.6% (bottom panel, circles, n ϭ 4), and the relative amount of G␤ was 50.0 Ϯ 5.3% (bottom panel, gray bars, n ϭ 4).
GIRK1-G␤ complexes were bound to the Dynal beads under equilibrium conditions in five identical and parallel reactions, and the solubilized membranes were then treated with a mixture of 1 M ACh and 1 M phenylephrine for different time intervals (1, 2, 5, 10, and 20 min, respectively) in the presence of Mg-ATP, GTP, and propranolol. At the end of each reaction, the GIRK1-G␤ complexes were segregated by magnetic immunoprecipitation from the remaining membrane proteins. The resulting oligomeric complexes were then analyzed by immunoblotting for the presence of GIRK1, G␤, actin, and PP1 (Fig.  9B, top). The results from the analysis of four independent experiments are illustrated in the middle and bottom panels of Fig. 9B. In these series of experiments, the intensities of the protein bands were quantified as before; however, the relative amount of each protein was normalized to the level of the same protein in the GIRK1-G␤ complexes obtained after 1 min of concurrent stimulation of the M 2 and ␣-adrenergic receptors (instead in the control complexes) for consistency with the electrophysiological experiments illustrated in Fig. 3. In this manner, we were able to explore the correlation between the GIRK1/4 channel inactivation and the composition of GIRK1-G␤ complexes. The experimental measurements of the G␤ protein levels as a function of the length of receptor stimulation demonstrated a G␤ depletion profile with a striking resemblance to GIRK channel inactivation (Fig. 3), but with a slightly slower time course. Such slowing is probably due to the fact that all biochemical experiments were carried out at 4°C using solubilized membrane proteins, whereas the biophysical experiments were performed in intact atrial myocytes at room temperature. Furthermore, the magnitude of G␤ depletion is in excellent agreement with the results from the biophysical analysis of the probability of G␤␥ binding to the channel, P (Fig. 6C). Remarkably, the G␤ dissociation from GIRK1-G␤ complexes was coupled with a similar depletion of PP1, whereas the relative amount of actin was only slightly decreased. Overall, our results revealed a dynamic modulation of the GIRK1-G␤ assemblies by the concurrent stimulation of the M 2 and ␣-adrenergic receptors. Such modulation might be a reflection of several different scenarios. One possibility is that receptor stimulation generates a homogeneous population of oligomeric GIRK1 complexes. Alternatively, GPCR activation might control equilibrium among several heterogeneous populations of signaling complexes similar to the equilibrium between the four functional modes of GIRK1/4 channels. Both possibilities are compatible with our biochemical data. Regardless of the exact scenario, the dynamic depletion of PP1 and G␤ from the GIRK1-G␤ signaling complexes upon M 2 and ␣-AR activation is an indicator of the plasticity of the GIRK1/4 channel environment.
Finally, we examined the potential role of M2Rs in such plasticity. To test for changes in M2R levels, we quantified the amount of M2Rs in the atrial membranes following stimulation with either 1 M ACh or with a combination of 1 M ACh and 1 M phenylephrine. Again, the receptor stimulation protocol strictly followed the protocol used in our electrophysiological experiments. Endogenous M2Rs were immunoprecipitated from the atrial membrane lysates, and the resulting immunocomplexes were probed with M2R-specific antibodies (Fig. 10).
The relative amount of M2Rs in atria treated for 5 min with 1 M ACh alone was 81.5 Ϯ 1.7% of the control (n ϭ 3). In contrast, the amount of M2Rs in atria incubated for 5 min with 1 M phenylephrine and then treated for 5 additional min with a combination of 1 M ACh and 1 M phenylephrine was only 32.6 Ϯ 7.2% of the control (n ϭ 3). The immediate implication of these findings is that altered density of M2Rs in the atrial membrane could also contribute to the inhibition of GIRK channel activity following ␣-ARs stimulation. The extent of these contributions, however, is currently unknown. Previous studies in CHO-K1 cells indicated that the homologous desensitization of GIRK channels caused by stimulation of M2Rs might be independent of the agonist-induced phosphorylation and internalization of these receptors (35). In addition, upon stimulation, a single GPCR molecule undergoes multiple catalytic cycles, activating a large number of heterotrimeric G proteins. Accordingly, the abundance of free G␤␥ in the membrane is not expected to decrease linearly with the reduction in the amount of M2Rs. Furthermore, the reduction in the probability of G␤␥ binding to GIRK1/4 channels, after ␣-ARs stimulation, and concurrent reduction in the amount of G␤ protein associated with GIRK1 complexes might be attributed to a diminished channel affinity for G␤␥ binding even at a constant G␤␥ concentration. Both PIP 2 depletion and PKC activation decrease G␤␥ affinity, rendering GIRK channels less sensitive to G␤␥ activation (4,16,22,36). Along these lines, the exact roles of M2Rs and different components of GIRK1 complexes in the coordination of ␣-adrenergic and cholinergic signaling pathways remain to be established. Nevertheless, the data in the present study demonstrate that such coordination involves extensive networks of membrane proteins.

DISCUSSION
We have established the ability of GIRK1/4 channels to coordinate the processing of ␣-adrenergic and cholinergic signals in the membrane of atrial myocytes. Functionally, this coordination is manifested as inhibition of GIRK1/4 channel activity upon concurrent stimulation of ␣-adrenergic and M 2 receptors in the myocytes. Channel inhibition, however, is not static. Although the initial GIRK response is comparable with the response evoked by activation of M2Rs alone, channel activity quickly declines when ␣-ARs and M2Rs are activated together. At a mechanistic level, channel inhibition is brought about by a combination of diminished probability of G␤␥ binding to the channel and increased channel desensitization. Furthermore, the dynamic modulation of channel function is accompanied by changes in the composition of GIRK1 oligomeric complexes. Such plasticity of the channel environment together with PIP 2 depletion and PKC activation might be paramount in the coordination of ␣-adrenergic and cholinergic signals in the membrane. These different aspects of our findings will be discussed below.
Initially, we examined whether the gating properties of the M2R-activated GIRK1/4 channels differ from those of G␤␥activated channels. The heterogeneity of GIRK channel gating requires new methods for interpretation of single-channel data. We therefore developed a quantitative approach for classification of channel behavior in functionally distinct modes (10,11). We first asked, "How long a GIRK channel resides in a particular functional mode before switching to another?" To answer this question, we analyzed the bursting behavior of GIRK channels. Our analysis indicated that the mean burst duration is ϳ400 ms, and this duration is independent of the gating pattern (10). Accordingly, the recordings were divided into consecutive, 400-ms intervals, and the channel frequency of openings, f, and the channel open probability, p o , were calculated within these intervals. The f and p o values were then compiled to generate f and p o histograms. Because each gating mode contributes a single kinetic component to these histograms, the histogram analysis reveals the number of functional modes accessible to the channel. Additionally, the f and p o histograms provide the means to quantify the equilibrium among different modes (11). Upon G␤␥ activation, our analysis identified the presence of four functional modes accessible to the GIRK1/4 channels. These four modes differ in their open probability and characteristic frequency of gating. More importantly, the concentration dependence of the modal equilibrium is consistent with a simple model of G␤␥ binding to the channel. In this model, the incremental occupancy of four equivalent and independent G␤␥ sensors in the tetrameric channel structure confers the four functional modes, corresponding to singly, doubly, triply and quadruply G␤␥-occupied GIRK1/4 channels (11). Accordingly, the probability of observing each gating mode would be a function only of the probability of G␤␥ binding to one of the four G␤␥ sensors, P, and this probability can be determined from analysis of modal equilibrium in each experiment. Here, we used the same approach to analyze the gating of M2R-activated GIRK1/4 channels and to probe the effects of ␣-ARs on the channel function. Our analysis revealed that the basic properties of G␤␥ activation of GIRK1/4 channels are conserved after M2R stimulation. Thus, receptor activation partitions the channels among four functional modes with characteristic frequencies comparable with those of G␤ 1 ␥ 7 -activated channels. Furthermore, the equilibrium probability (or relative occupancy) of different modes is binomially distributed as predicted by the model entertained previously for G␤␥ activation of the channel. The equivalence between the M2R and G␤␥ activation of GIRK channels yields an important insight into GIRK channel regulation; it indicates that M2R activation neither affects the equivalence of the four putative G␤␥ sensors in the channel nor obstructs G␤␥ binding to these sensors.
The molecular mechanisms underlying the heterogeneous GIRK behavior might be much more convoluted. Yet the manifestation of these mechanisms in the form of equilibrium between four functional modes of the channel creates a powerful framework for quantification and interpretation of heterogeneous single-channel behavior. Moreover, the analysis of the gating equilibrium allowed us to determine the probability of G␤␥ binding to the channel, P, in the intact myocytes following M2R activation and to evaluate the impact of ␣-ARs on this probability. In phenylephrine-treated myocytes, the fraction of GIRK1/4 channels residing in the high efficient modes 3 and 4 was substantially lower than that in control myocytes. Thus, the mean probability of G␤␥ binding to the channel declined ϳ2-fold in the phenylephrine-treated myocytes (from P ϳ 0.55 to P ϳ 0.23). Concurrently, the fraction of channels residing in nonconducting states increased ϳ2-fold beyond the predicted fraction of channels with no bound G␤␥. Such dual modulation of GIRK1/4 channels can be attributed to several parallel cellular processes triggered by stimulation of ␣-ARs in atrial myocytes. These receptors activate phospholipase C, which uses PIP 2 as a substrate to generate diacylglycerol and inositol 1,4,5trisphosphate, thus depleting membrane PIP 2 . In parallel, accumulation of diacylglycerol promotes activation of PKC. The G q -mediated inhibition of GIRK channels arises essentially from the PIP 2 depletion (16 -20, 24, 25) and PKC activation (4,(21)(22)(23), although G␣ q subunits might also contribute directly to this phenomenon (16,33,34). Furthermore, Keselman et al. (36) have recently shown that PIP 2 depletion and PKC activation positively reinforced one another in their inhibition of GIRK channels. Numerous studies from different groups, however, have reported conflicting findings on the relative contributions of PIP 2 depletion and PKC activation to channel inhibition (reviewed in Refs. 23 and 36). Such diverse findings might be attributed to differences in the experimental conditions employed in the different studies. Alternatively, the mode of operation in G q PCR-mediated inhibition of GIRK channels might depend on the highly organized GIRK1 complexes, which maintain cell-specific order among the diverse signaling components. To do so the GIRK1 complexes should have a dynamic nature and change their composition upon receptor stimulation. We therefore probed the composition of GIRK1 complexes in a series of biochemical experiments, which closely replicated the electrophysiological experiments. These experiments produced several interesting observations. First, the propensity of G␤ to associate with GIRK1 complexes (Fig. 9B) correlates remarkably well with the probability of G␤␥ binding, derived from the analysis of channel gating (Fig. 6C). This observation independently confirms that modal behavior of GIRK channels reflects the interactions between the channels and G␤␥. Hence, the analysis of gating equilibrium can be reliably used for quantification of these interactions.
More importantly, our biochemical experiments revealed receptor-and time-dependent differences in the association of actin and PP1 with GIRK1-G␤ complexes. The increased association of actin with these complexes upon M2R activation is particularly interesting in the light of a recent study by Cho et al. (25), which implicates the integrity of actin cytoskeleton in the control of the membrane mobility of PIP 2 . Therefore, future studies will investigate whether changes in the protein levels of actin in GIRK1 complexes would affect PIP 2 sensitivity of the GIRK channels. Even more importantly, our studies revealed that the temporal pattern of PP1 association with GIRK1-G␤ complexes closely resembles the association profile of G␤ with GIRK1/4 channels. Although the precise role of such protein choreography in the regulation of the channel function is not known, the studies presented here demonstrate for the first time a dynamic link between channel signaling environment and function. In addition, our studies launch several new areas of inquiry. What is the temporal sequence of association and dissociation of different protein kinases with the GIRK1 complexes? Which events orchestrate the dynamic modulation of GIRK1/4 channel function by PIP 2 and which events favor PKCdependent modulation? Finally, is the GIRK1/4 signaling environment also governing the reduction of M2R amount in the membrane? The answers to these questions will begin to define the role of protein context in the integration of adrenergic and cholinergic signals in the heart. Beyond providing mechanistic insight, these inquiries will also have significant ramifications for development of new therapies for management of clinical conditions like congestive heart failure in which persistent AR stimulation is a hallmark. Based on the functional data presented here (Fig. 3), such persistent activation of ␣-ARs would inhibit ϳ74% of the GIRK1/4 channel response to parasympathetic stimulation, leading to distortion of the delicate balance between sympathetic and vagal signals in the atria.