The Dynamics of Formation and Action of the Ternary Complex Revealed in Living Cells Using a G-protein-gated K (cid:1) Channel as a Biosensor*

Traditionally the consequences of activation of G-pro-tein-coupled receptors (GPCRs) by an agonist are studied using biochemical assays. In this study we use live cells and take advantage of a G-protein-gated inwardly rectifying potassium channel (Kir3.1 (cid:1) 3.2A) that is activated by the direct binding of G (cid:2)(cid:3) subunit to the channel complex to report, in real-time, using the patch clamp technique the activity of the “ternary complex” of agonist/receptor/G-protein. This analysis is further fa-cilitated by the use of pertussis toxin-resistant fluorescent and non-fluorescent G (cid:4) i/o subunits and a series of HEK293 cell lines stably expressing both channel and receptors (including the adenosine A 1 receptor, the ad- renergic (cid:4) 2A receptor, the dopamine D 2S receptor, the M4 muscarinic receptor, and the dimeric GABA-B 1b/2 receptor). We systematically analyzed the contribution of the various inputs to the observed kinetic response of channel activation. Our studies indicate that the combination of agonist, GPCR, and G-protein isoform uniquely specify the behavior of these channels and thus support the importance of the whole

The activation of G-protein-coupled receptors (GPCRs) 1 by extracellular ligands is an important mechanism involved in a multitude of physiological responses and is of central importance in drug development and therapeutics (1). The activated receptor couples to G-proteins of various subtypes that then activates effector pathways either directly or indirectly. This combination of agonist, receptor, and G-protein is referred to as the "ternary complex" and is thought to be the key essential determinant of the magnitude of the downstream response (2,3). The most recent formulations propose a cubic ternary complex model with a large number of equilibrium constants between various states governing efficacy (2,4,5). The important species is the activated receptor/agonist/G-protein complex. It is proposed that for any combination of these three elements the particular active conformation (or conformational space) is unique and can thus have distinctive signaling consequences (2, 4 -6). An agonist binds more favorably to the active receptor species and thus at equilibrium favors its' formation. Recently this model has been extended to also incorporate the kinetics of G-protein activation and deactivation and indicate that a kinetic model, as opposed to an equilibrium model, may potentially have quite different properties (7).
Generally these phenomena have been studied by the use of biochemical assays, using cell homogenates or fractions, or by measuring the behavior of a physiological response many steps downstream from the G-protein cycle. It is apparent that there is a gap in our understanding about how these signaling pathways behave dynamically in intact cells. This is important, because, in reality, the release of hormones and neurotransmitters varies over the second time scale. Agonist binding to receptor is generally agreed to be diffusion-limited and much faster than the activation of downstream signaling events. However, there are a number of more controversial issues regarding models of receptor activation of G-proteins. Is there kinetic evidence for the unique conformation of the ternary complex? What is the role of both the isoform and concentration of G-protein in dictating the dynamic behavior? Is the encounter of receptor with the G-protein rate-limiting, and do receptor and G-protein exist in a pre-coupled complex? To address these questions we have used members of the Kir3.0 family of inwardly rectifying K ϩ channels that are gated by G-proteins as a reporter. G-protein-gated inwardly rectifying K ϩ channels were first identified in atrial myocytes where they are activated by acetylcholine at muscarinic M 2 receptors (8 -10). It was subsequently shown that this activation was membranedelimited (11), mimicked by non-hydrolysable GTP analogues (12), and sensitive to pertussis toxin (PTx), implicating the inhibitory family of G-proteins (G i/o ) (13). It is now apparent that G-protein-gated inwardly rectifying K ϩ currents are also present in many neuronal cell types (14 -16). Cloning efforts have revealed the molecular counterparts of these currents, and the channel is a heteromultimer of members of the Kir3.0 family of K ϩ channels (16 -23). It is now accepted that activation of native and cloned Kir3.0 channels involves a direct interaction with the G␤␥ dimer not the G␣ subunit (16,24,25).
In our previous studies we have developed a series of molecular tools to study these issues including stable cell lines expressing the channel complex along with GPCRs and both fluorescent and non-fluorescent G␣ i/o subunits engineered to be resistant to the action of pertussis toxin (26 -28). In this study we combine the use of these tools with whole-cell patch clamping to record Kir3.1ϩ3.2A currents in response to GPCR stim-ulation by agonists applied using a rapid and localized drug application system to assay the G-protein cycle on the hundreds of millisecond time scale. We focus on how quickly the channel activates and use this parameter to examine how the ternary complex dictates the final channel response.

EXPERIMENTAL PROCEDURES
Molecular Biology, Cell Culture, and Transfection-PTx-resistant G i/o mutant ␣ subunits and CFP-tagged PTx-resistant G␣ i/o were generated and used as previously described (27,28). GABA-B 1b and GABA-B 2 were expressed in the dual promoter vector pBudCE4.1 (Invitrogen). Standard molecular cloning techniques were used to excise the relevant clones from the previous vector (GABA-B 1b was excised from pcDNA3.1/neo/(ϩ) with PmeI/XhoI and GABA-B 2 from pcDNA3.1/neo/(ϩ) with KpnI/XhoI), and they were introduced into the two polylinkers in ScaI/SalI sites for GABA-B 1b and KpnI/XhoI sites for GABA-B 2 . Inducible expression of G␣ i3 -CFP was achieved using the TRex system (Invitrogen). G␣ i3 -CFP was removed from the previously described construct (28) and subcloned into pcDNA5/TO using a KpnI/ NotI digest. The A 1 -Gi1␣(C351G) fused construct was as previously described (29), and cDNA was excised from pcDNA3 and subcloned in to pcDNA3.1/Zeo/(ϩ).
Transiently transfected cells suitable for patch clamping were identified by epifluorescence from co-transfection of 100 ng of the enhanced variant of the green fluorescent protein (pEGFP-N1, Clontech). Data were obtained from at least two independent transfections.
Radioligand Binding-Radioligand binding was performed on crude membrane preparations isolated from the relevant stable lines (HKIR3.1/3.2/A1, HKIR3.1/3.2/␣2, and HKIR3.1/3.2/D2). Cells were harvested into binding buffer (50 mM Tris-HCl, pH 7.4) and stored at Ϫ80°C. Cells were hypotonically shocked (10 mM Tris-HCl and 10 mM EDTA) on ice and then homogenized using a glass-on-glass Dounce homogenizer. The homogenate was spun at 600 ϫ g (4°C) for 15 min to sediment nuclei and large cell debris. The membrane fraction was then obtained by spinning the supernatant at 100,000 ϫ g av in an ultracentrifuge (Beckman, Optima LE-80K). The pellet was resuspended in binding buffer and incubated with radioligand at room temperature for 1 h. Specific binding was assessed using saturating concentrations of radiolabeled receptor antagonists: 8 nM (D 2 ). Binding was performed in triplicate and repeated at least four times. Data were corrected for total protein content in each sample and are expressed as fmol/g of protein (mean Ϯ S.E.).
Electrophysiology-Whole-cell membrane currents were recorded using an Axopatch 200B amplifier (Axon Instruments). Patch pipettes were pulled from filamented borosilicate glass (Clark Electromedical) and had a resistance of 1.5-2.5 M⍀ when filled with pipette solution (see below). Prior to filling, tips of patch pipettes were coated with a Parafilm/mineral oil suspension. Data were acquired and analyzed using a Digidata 1200B interface (Axon Instruments) and pClamp software (version 6.0; Axon Instruments). Cell capacitance was ϳ15 picofarads, and series resistance (Ͻ10 M⍀) was at least 75% compensated using the amplifier circuitry. Recordings of membrane current were carried out after an equilibration period of ϳ5 min. Immediately following patch rupture, a current-voltage relationship was performed to establish that currents were inwardly rectifying. Thereafter cells were voltage-clamped at Ϫ60 mV, and agonist-induced currents were measured at this potential. For current-voltage relationships, records were filtered at 1 kHz and digitized at 5 kHz. For continual data acquisition where cells were voltage-clamped at Ϫ60 mV, records were digitized at 100 Hz.
Rapid Drug Application and Barium Calibration-Drugs were applied using a "sewer pipe" system (Rapid Solution Changer RSC-160, Bio-Logic) whereby an array of perfusion capillaries was placed in the bath ϳ40 m from the recorded cell. This system allowed rapid solution switching between capillary tubes and localized application of drugs due to the laminar flow over the studied cell from the tubes as previously described (31). A number of parameters were determined using this system (Fig. 1A, part ii). Upon agonist application current activated with an initial delay (lag) followed by a rapid rise to peak amplitude (time-to-peak (ttp)). Current subsequently became desensitized during continued agonist application. In this study agonist was applied for 20 s. Upon agonist removal currents deactivated back to baseline levels.
For each cell we assessed whether there were any flow artifacts resulting from the pressure of drug application. We did this by applying bath solution from one of the sewer pipes and recording any flowinduced currents. If such a current was observed, then the position of the perfusion head was moved to minimize it. Furthermore, to control for variations in positioning of the sewer pipe system relative to the cell, we calibrated this system using the kinetics of channel block by barium. The cell was positioned in the center of the field using crosshairs in the microscope eyepieces. Barium (1 mM) was applied to the cell in the presence of agonist when the agonist-induced current had reached a plateau phase. Block of the current occurred with an initial delay before reaching equilibrium. It was assumed that this lag reflected the intrinsic delivery time to the cell. A barium calibration was performed prior to the start of experiments to ascertain correct positioning of the sewer pipe and was repeated on several cells during each recording session. In general the results were highly reproducible (the lag time for barium block was 237.3 Ϯ 11.7 ms (n ϭ 73)).
Confocal Microscopy and Western Blotting-Confocal microscopy and acquisition of images were as previously described (28). In the current study we used a 40ϫ oil objective, 40% laser power, gain was set to 50%, and iris aperture was opened to 1.5 nm (optimum aperture, 1.1 nm). Western blotting of CFP-tagged G␣ i/o subunits was performed using a polyclonal rabbit GFP antibody as previously described (28).
Data Analysis-Membrane currents were measured at Ϫ60 mV, and all data are presented as mean Ϯ S.E., where n indicates the number of cells recorded from which data were recorded. Time measurements (lag plus ttp) were reciprocated prior to statistical analysis, because the reciprocal of time is normally distributed. Data are shown untransformed. Data were analyzed for statistical significance using either Student's t test or one-way repeated measures analysis of variance tests with Bonferroni correction as appropriate (*, p Յ 0.05; **, p Յ 0.01; and ***, p Յ 0.001).

RESULTS
In our previous studies (26,27), we generated an HEK293 stable cell line expressing the Kir3.1ϩ3.2A channel complex.
On this background of channel expression we subsequently generated dual receptor-plus-channel stable lines in which we have investigated the kinetic properties of receptor-mediated currents (see "Experimental Procedures"). Fig. 1A shows the profile of current activation following stimulation of the A 1 adenosine receptor by a concentration of agonist (NECA, 1 M) that would lead to full receptor occupancy. We observed an initial lag followed by a rapid rise to a peak amplitude of current. With prolonged agonist application current amplitude wanes as the response desensitizes, and upon removal of agonist it deactivates back to baseline levels. In this study we focus upon the initial activation phase, which we measured as the sum of the "lag" plus the "time-to-peak" (lag plus ttp) and investigate the effects of the ternary complex on this response. The experiments just described were performed using maximal concentrations of agonist likely to result in full receptor occupancy. We next examined the effects of agonist concentration, and thus receptor occupancy, on channel activation. We used the HKIR3.1/3.2/GGB ( Fig. 2A) and the HKIR3.1/3.2/A1 cell lines (Fig. 2B) and used the agonists at a high, saturating concentration and at a lower concentration, which was approximately the EC 50 value. We found that, with the lower concentration of agonist, channel activation was significantly slowed (Fig. 2, A and B).
We have shown previously that by using engineered PTxresistant G␣ subunits, it is possible to look exclusively at coupling between a receptor and the channel via specific G␣ i/o isoforms (27). Furthermore, we have recently made a series of cyan fluorescent protein (CFP)-tagged PTx-resistant G␣ i/o isoforms and have shown that they are both membrane-targeted and functional (coupling to both ␤␥ subunits and the adenylate cyclase pathway), and we have established conditions where these constructs are expressed at equivalent levels (28). In our previous work we have shown that the A 1 receptor appears to couple to the channel with equal efficacy and potency via all the G␣ i/o isoforms tested (27). We also now demonstrate that all CFP-tagged G␣ i/o subunits are able to participate in A 1 -mediated channel activation with similar magnitudes of response and similar kinetic profiles except via G␣ oA -CFP, where we observed that channel activation via this G-protein exhibited slower activation kinetics (Fig. 3) (28). However, we know that other receptors, for example the M4 muscarinic and the GABA B heterodimeric receptors, show more selective patterns of coupling to G␣ i/o subunits (27), and so we examined the dynamics of coupling via G␣ i2 and G␣ oA with the M4 and GABA-B 1b/2 receptors (using the HKIR3.1/3.2/M4 and HKIR3.1/3.2/GGB lines, respectively). With the HKIR3.1/3.2/M4 cell line we used the CFP-tagged G␣ subunits (G␣ i2 -CFP and G␣ o -CFP, at equivalent concentrations) and observed that, although channel activation via this receptor was intrinsically slower than through the A 1 receptor, there was no significant difference in channel activation via G␣ i2 -CFP and G␣ o -CFP (Fig. 4). However with the HKIR3.1/3.2/GGB line we were unable to rescue coupling between the GABA-B 1b/2 receptor and the channel in PTxtreated cells using the CFP-tagged G-proteins. This is the only receptor to date where we have observed this, and such lack of coupling may be related to their unique heterodimeric receptor formation. Instead we used the non-CFP-tagged G␣ i2 C352G and G␣ oA C351G to study channel activation through the GABA-B 1b/2 receptor. In the HKIR3.1/3.2/A1 cell line, expression of these constructs yielded comparable activation kinetics to the CFP-tagged variants (not shown). In contrast to both the A 1 and M4 receptors, activation via this receptor was much faster through G o than G i2 (Fig. 5). For comparison the magnitude of current potentiation and kinetics of activation via native G-proteins are included (see Figs. [3][4][5]. Importantly, we see different kinetic profiles of channel activation through different receptor and G-protein combinations despite robust coupling apparent from the magnitude of the response. G-protein ␣ isoforms are present at varying levels in different cells, and thus it is important to know what role the G-protein concentration has in determining these kinetic responses. To do this we established an inducible system whereby we could regulate the levels of expression of G␣ i3 -CFP in a cell line stably expressing the channel complex and the A 1 receptor (referred to as HKIR3.1/3.2/A1/G␣ i3 -T). This was done using a commercially available system (see "Experimental Procedures") whereby gene expression is conditional on the addition of the antibiotic tetracycline (Tet). We titrated the concentration of Tet (0.01-100 g/ml) to determine a high, medium, and low level of G␣ i3 -CFP expression. Fig. 6A shows the induction of G␣ i3 -CFP expression at the membrane in the HKIR3.1/3.2/ A1/G␣ i3 -T cell line with increasing concentration of Tet. Using Western blotting we showed the induction of graded expression of G␣ i3 -CFP (Fig. 6B). In addition, in PTx-treated cells, increasing concentrations of Tet progressively enhanced the amplitude of NECA-induced currents (Fig. 6C). However, importantly the kinetics of activation of the channel response was not altered (Fig. 6D).
Finally, we examined the potential role that diffusion of the receptor to G-protein might have on activation kinetics. We took advantage of an approach in which the receptor is physically tethered to an engineered PTx-resistant G-protein, specifically the A 1 receptor and the G␣ i1 C351G subunit. We both transiently and stably expressed this construct in the HKIR3.1/ 3.2 cell line. In addition, we characterized receptor expression density using radioligand binding with [ 3 H]DPCPX (8 nM) in a clonal isolate and found that the fused A 1 receptor was expressed at more that 2-fold higher levels (72.7 Ϯ 13 fmol/g of protein, n ϭ 9) than the non-fused A 1 receptor (28.6 Ϯ 7.2 fmol/g of protein, n ϭ 6) (Fig. 1C). We compared the activation kinetics of A 1 via endogenous G-proteins to that of the fused A 1 -G␣ i1 construct after both transient and stable expression, with and without PTx treatment (Fig. 7). Under both sets of conditions activation was significantly slower via the fused construct both before and after PTx treatment. However, if G␣ i1 C351G was transfected into the HKIR3.1/3.2/A1 line, signaling was also slowed in an analogous manner. Thus, it is the nature of the mutant G-protein subunit rather than the tethering per se that determines the change in activation kinetics. We also observed that the deactivation rate was increased and this is discussed below.

DISCUSSION
In this report we have taken advantage of the fact that Kir3.0 channels are gated directly by G␤␥ subunits, and thus they act as biosensors for G␤␥ concentration at the plasma membrane. One of the major advantages of this approach is that the release of G␤␥ is directly measured and the system output is not dependent upon downstream events. In combination with electrophysiological recordings (and rapid agonist application) this results in high temporal resolution. We have analyzed the kinetic contribution of the ternary complex to channel activation. We have a number of major findings, namely that increased occupancy of the receptor accelerates activation kinetics, that the activation kinetics via a G i/o isoform are determined by the particular receptor/G-protein combination, and that receptor diffusion to the G-protein and the concentration of the G-protein have little influence on the activation kinetics. The G-protein amount simply determines the amplitude of the response. Our data all support the hypothesis that the assembly of the ternary complex of agonist, receptor, and G-protein is not rate-limiting. It is, however, the unique conformation of the active ternary complex that specifies the kinetic behavior of the channel response.
Channel activation kinetics through a number of different receptors with saturating agonist concentration occurs rapidly via the mixed pool of G-proteins endogenously expressed in HEK293 cells. However, there are significant differences with activation, with that through M4 being the slowest and ␣ 2A being the most rapid. Indeed, the nature and numerical details of channel kinetics are very similar to those occurring with the channel expressed in neurons (15,32). Receptor-mediated currents elicited using high agonist concentration have a typical profile comprising an initial lag followed by a subsequent sharp rise to a peak amplitude after drug application. This pattern  bar. B, the bar chart shows the magnitude of current responses before, on, and after agonist application in conditions as indicated. C, the bar chart summarizes the data obtained from a number of these experiments. Mean data shows that there is no significant difference between G␣ i2 -CFP and G␣ oA -CFP in channel activation.
likely reflects the occurrence of a number of sequential steps. In a classic "collision" coupling view these steps might consist of agonist binding to receptor followed by diffusion of the agonist/ receptor complex to the G-protein, activation and dissociation of the G-protein heterotrimer, diffusion of G␤␥ to the channel, and finally, activation and opening of the channel. Binding of G␤␥ and channel activation are assumed to be fast and not rate-limiting. Indeed there is also evidence that channel activation is intrinsically cooperative, because the Hill coefficient for G␤␥-mediated channel stimulation is between 1.5 and 3 and more than one G␤␥ subunit needs to occupy one of the four equivalent binding sites to initiate channel opening (33,34). Such considerations would account for why there is a small discrepancy between the point at which induced G-protein is detected and the point at which significant coupling begins to occur in Fig. 6.
As might be expected from the principles of mass action, agonist concentration clearly influences the onset kinetics. Agonist binding to receptors is diffusion-limited, agonists bind preferentially to active receptor conformations, and thus within the timescale of signaling events the establishment of this equilibrium between active and inactive receptor conformations will not be rate-limiting. It is the proportion of active species that will subsequently determine the kinetics of the downstream response. But, what about other steps in the Gprotein signaling pathway? To address the role of receptor diffusion to the G-protein we fused the A 1 receptor directly to G␣ i1 . It is possible that a receptor might activate sequentially multiple G-proteins resulting in signal amplification. In this case, a slower response would ensue following stimulation of a fused receptor G-protein construct. Alternatively, diffusion might be a rate-limiting step in which case fusion should accelerate signaling. Experimentally, we observed that channel activation kinetics was slower (despite more than 2-fold higher levels of receptor expression in stable lines expressing the A 1 -G␣ i1 fused complex) compared with non-fused receptor signaling via endogenous G-proteins. However if channel activation via the A 1 -G␣ i1 construct is compared with activation when the A 1 receptor is simply constrained to (but not tethered to) G␣ i1 C351G, then there is no significant difference between these two conditions. Thus it is the G␣ i1 itself rather than the lack of amplification that influences kinetics. It has been previously noted that the nature of the cysteine mutation does result in lowered affinity for the receptor (35,36), and our finding appears to be a kinetic reflection of this observation. Although this may be an issue with this approach, it is essentially unavoidable. A second caveat is that the levels of receptor expression are higher in our system than might reasonably be expected to occur with natively expressed receptors: under conditions of lower expression amplification may be important. Thus, it seems that diffusion of the receptor to the G-protein is not kinetically important; it is the subsequent intrinsic activation process that is rate-limiting. Our results with the A 1 -G␣ i1 fusion are consistent with biochemical studies using this construct (36), but a study on a receptor-G z fusion found diffusion to be key, although this may be related to the unusual properties of that particular G-protein (37). Furthermore, we show that deactivation rates are increased and this too is consistent with data of Waldhoer et al. (36). They find that the A 1 -G␣ i1 C351G fusion releases bound radioligand agonist more readily than the A 1 receptor and argue that the ternary complex is less stable (36). It is difficult to be quite so categorical with our approach, but our data are potentially consistent with this hypothesis.
Receptor occupancy is not the only factor that has important consequences for channel kinetics. We demonstrate that acti- vation occurs faster via G i1 , G i2 , and G i3 than G oA for the A 1 receptor. This pattern was reversed when channels were activated via the GABA-B 1b/2 receptor, and a similar non-significant trend was observed with the M4 receptor, thus illustrating the particular receptor/G-protein combination dictates the response. These observations can be accounted for by greater "kinetic efficacy," i.e. some agonist-receptor-G-protein ternary complexes promote the faster release of GDP from the G-protein ␣ subunit. This proposal is not unreasonable given the accumulating data supporting the idea that different GPCRs have differing affinities for the various G␣ isoforms and that even different agonists at the same receptor may couple with varying degrees to different G-protein isoforms (2,6,27,38,39). It has been argued on theoretical grounds that there may be significant differences in the predicted behavior of signaling cascades when considered in kinetic models compared with that in equilibrium models (3,7). Our data argue that the ternary complex uniquely determines the kinetic as well as the steady-state properties.
Intriguingly, the G-protein amount simply modifies the amplitude of the response and does not influence the activation kinetics. This is consistent with the potential existence of a complex between the G-protein ␣ subunit and channel (40) or a high degree of precoupling between receptor and G-protein being important for channel activation (26,41), and it is con-FIG. 6. Inducible G␣ i3 -CFP expression system in the HKIR3.1/ 3.2/A1/G␣ i3 -T stable cell line. A, upper panel shows laser scanning confocal images of live cells treated with tetracycline for 24 h to induce the expression of G␣ i3 -CFP. Concentrations of tetracycline (Tet) are indicated (g/ml). The lower panels show a brightfield image of the same field of cells. The scale bar represents 10 m. Cells were imaged under identical conditions in each instance (see "Experimental Procedures"). B, we performed a Western blot using an antibody directed against GFP as previously described (28) to show graded expression of G␣ i3 -CFP (band indicated by the asterisk) in response to various concentrations of Tet as indicated. Markers are for 50-and 80-kDa proteins. WT indicates lysate of HEK293 cells, and C indicates zero Tet in the HKIR3.1/3.2/A1/G␣ i3 -T line. The symbol "ϳ" indicates a background band that we observe using this antibody (as previously reported (28)) but that has a different mobility to GFP. C, summary of electrophysiological data showing mean current densities measured at Ϫ60mV from HKIR3.1/3.2/A1/G␣ i3 -T cells treated with combinations of Tet (concentrations as indicated) and PTx (100 ng/ml) for 24 -30 h prior to recording. A representative sample of cells from each condition was recorded from. Not all Tet-induced cells produced agonist-stimulated currents: 25% of patched cells, which had been treated with Tet and PTx, were unresponsive to agonist and so were not included in the analysis. NECA-induced currents were significantly larger for the 10 g/ml Tettreated group than the 0.5 g/ml group (p Ͻ 0.05). D, measured activation kinetic parameters show no significant differences between treatment groups.  , 20 s). B, bar chart showing a comparison of channel activation kinetics between the fusion protein A 1 -G␣ i1 ("Fused") and the A 1 receptor ("Unfused") transiently transfected into the HKIR3.1/3.2/cell line. Channels were significantly slower (p Ͻ 0.001) to activate when the A 1 receptor was fused to G␣ i1 C351G. C, we compared channel activation kinetics via the A 1 receptor either alone ("Unfused"), when co-expressed with G␣ i1 C351G ("ϩG␣ i1 "), or when fused to G␣ i1 C351G ("Fused/-PTx") and treated with PTx ("Fused/ ϩPTx"). Here channel activation kinetics were not significantly different. sistent with theoretical studies (3). Our studies here have mostly focused on the A 1 receptor, and it is possible that variations in precoupling, for example promoted by RGS proteins, might influence activation kinetics via other receptors. It is intriguing that a recent report has shown dopamine receptors and Kir3.0 channels potentially existing in complexes suggesting that all three components may be associated in a macromolecular complex (42). It is also interesting that the kinetics of K ϩ and Ca 2ϩ channel modulation in G␣ o knockout mice are slowed (43): our results suggest that this may be related to differences in the efficacy of GPCR coupling to the remaining G-proteins rather than changes in amount of the total G-protein pool.
Our results and conclusions depend on the use of a number of tools, in particular that of PTx-resistant G-proteins with a point mutation at the cysteine four amino acids from the end of the protein. Despite the advantages of this approach, there are some essentially unavoidable caveats that should be appreciated. It is apparent from biochemical studies that the mutation of the cysteine and the particular amino acid substituted affects the efficacy of the response (35,44). Of these mutations the C-I replacement seems to best preserve coupling and is the one we have used in our fluorescent protein G-protein alpha subunit chimaeras (28,35,44). However we have previously shown little difference in efficacy of coupling between C-I and C-G mutations of G␣ i1 (27). However, using these reagents (and the G␣ i2 C352G and G␣ oA C351G mutants) in our system, it is clear that the magnitude and the time to peak activation varies little between coupling via native G-proteins in the absence of PTx and coupling via a selected G-protein mutant in PTxtreated cells. For example, if one compares coupling to endogenous G␣ via G␣ o -CFP with the M4 receptor (Fig. 4, B and C) and via G␣ oA C351G with GABA-B (Fig. 5, B and C), it can be seen that there is little difference in magnitude and activation kinetics. The only place where such issues may be pertinent is with the studies shown in Fig. 7, but it is difficult to see an alternative strategy that would ensure strict coupling between the receptor and fused G-protein. A second consideration is that different G␤␥ combinations in the heterotrimer may also play a role (45), however, in our hands all four G-protein ␣ isoforms can robustly couple with endogenous G␤␥ dimers in HEK293 cells (27,28).
In summary our data support the concept that the formation of the ternary complex is not rate-limiting for signaling. It is the release of GDP from the G-protein heterotrimer that is important, and the rate at which that happens is determined by the conformation of the ternary complex.