Kinetics of Ternary Complex Formation with Fusion Proteins Composed of the A1-Adenosine Receptor and G Protein α-Subunits*

High affinity agonist binding to G protein-coupled receptors depends on the formation of a ternary complex between agonist, receptor, and G protein. This process is too slow to be accounted for by a simple diffusion-controlled mechanism. We have tested if the interaction between activated receptor and G protein is rate-limiting by fusing the coding sequence of the human A1-adenosine receptor to that of Gαi-1(A1/Gαi-1) and of Gαo(A1/Gαo). Fusion proteins of the expected molecular mass were detected following transfection of HEK293 cells. Ternary complex formation was monitored by determining the kinetics for binding of the high affinity agonist (−)-N 6-3[125I](iodo-4-hydroxyphenylisopropyl)adenosine; these were similar in the wild-type receptor and the fusion proteins over the temperature range of 10 to 30 °C. Agonist dissociation may be limited by the stability of the ternary complex. This assumption was tested by creating fusion proteins in which the Cys351 of Gαi-1 was replaced with glycine (A1/Gαi-1C351G) or isoleucine (A1/Gαi-1C351I) to lower the affinity of the receptor for the G protein. In these mutated fusion proteins, the dissociation rate of the ternary complex was accelerated; in contrast, the rate of the forward reaction was not affected. We therefore conclude that (i) receptor activation per se rather than its interaction with the G protein is rate-limiting in ternary complex formation; (ii) the stability of the ternary complex is determined by the dissociation rate of the G protein. These features provide for a kinetic proofreading mechanism that sustains the fidelity of receptor-G protein coupling.

Signaling by G protein-coupled receptors is initiated by the formation of a ternary complex which consists of agonist (H), 1 receptor (R), and G protein (G) (1,2); upon activation of the receptor by an agonist, the activated receptor (R*) associates with its cognate G protein(s). In the resulting ternary HR*G complex, the agonist is bound with high-affinity. It is not clear whether the conformational transition to R* is rate-limiting or whether ternary complex formation is limited by the association of receptor and G protein (2,3). Based on kinetic arguments, it has been suggested that a significant portion of the receptors were precoupled, i.e. there is a fraction of preformed R⅐G complexes in intact cells and membranes because receptors can bind G proteins in the absence of agonists (for review, see Refs. 4 -6). However, earlier experiments that were designed to investigate the kinetics of adenylyl cyclase regulation by the ␣ 2 -adrenergic receptor failed to detect an initial burst of receptordependent inhibition although this would be predicted for precoupled receptors (7). Similarly, the mechanism of signal transduction in the visual system is inconsistent with precoupling of rhodopsin and transducin (8).
In contrast to photoreceptors (or other specialized sensory cells) where a receptor is only confronted with a single type of G protein, in most cells a given receptor selects its cognate G protein(s) from a multitude of available G protein oligomers; the specificity in this interaction that can be observed is remarkable in many instances (reviewed in Ref. 9). The mechanism by which this fidelity is achieved is not fully understood. In the present work, we used the human A 1 -adenosine receptor because its interaction with G proteins has been extensively characterized in experiments with purified and defined components (10 -15); we have generated precoupled R/G tandems by fusing the coding sequence of the human A 1 -adenosine receptor to that of G␣ i-1 (or G␣ o ) to test if the formation of the ternary complex was limited by the association of receptor and G protein. In addition, we have altered the affinity of the G protein for the receptor by introducing mutations at the carboxyl terminus of the G␣ subunit, a site that is critical for R/G interaction (16,17). Using this approach, we show that the association of receptor and G protein is not rate-limiting; in contrast, the stability of the ternary complex is limited by the dissociation rate of the G protein. This suggests that the fidelity of receptor-G protein coupling is achieved by a kinetic proofreading mechanism.

EXPERIMENTAL PROCEDURES
Materials-Adenosine deaminase and Complete TM protease inhibitor tablets were from Roche Molecular Biochemicals (Germany). Hepes and CHAPS were from Biomol (Munich, Federal Republic of Germany); suramin was obtained from Research Biochemicals (Natick, MA). The materials required for SDS-polyacrylamide gel electrophoresis were from Bio-Rad. Fetal calf serum was from PAA Laboratories (Linz, Austria), Dulbecco's modified Eagle's medium, non-essential amino acids, ␤-mercaptoethanol, L-glutamine, penicillin G, streptomycin, and G418 (geneticin) were obtained from Life Technologies, Inc. (Grand Island, NY). cAMP, CPA, forskolin, and pertussis toxin were purchased from Sigma. Buffers and salts were from Merck (Darmstadt, Germany); [ 3 H]adenine and [ 125 I] were from NEN Life Science Products Inc. (Boston, MA). [ 125 I]HPIA was synthesized according to Linden (18). Centrifuge tubes and tissue culture plates were from Greiner (Vienna, Austria) and Corning Costar (Acton, MA). Plasmid preparation kits were from Qiagen (Hilden, Germany). The cDNA coding for the human A 1 -adenosine receptor in the plasmid vector pBC-A1R (19) was kindly provided by M. J. Lohse (University of Wü rzburg). The vector pEGFP-C1 was obtained from CLONTECH (Palo Alto, CA).
A 1 Adenosine Receptor/G␣ Fusion Proteins-The construction of A 1adenosine receptor/G␣ i-1 (and G␣ o ) fusion proteins is described in Ref. 20. In short, the NH 2 terminus of pertussis toxin-resistant forms (Cys 351 of G␣ i-1 replaced by Gly or by Ile) of G␣ i-1 or of wild-type G␣ i-1 were fused to the COOH terminus of the human A 1 -adenosine receptor. This converted the Asp residue at the carboxyl-terminal of the receptor to Ala and preserved the methionine, which normally functions as the initiator in G␣ i-1 (and G␣ o ).
Cell Culture and Transfections-The conditions for cell culture and transfection are described in Ref. 21. In brief, HEK293 cells were maintained in Dulbecco's modified Eagle's medium at 5% CO 2 and 37°C. Culture media were supplemented with 10% fetal calf serum, 2 mM L-glutamine, ␤-mercaptoethanol, non-essential amino acids, 100 units/ml penicillin G, and 100 g/ml streptomycin. Media for the culture of stably transfected cells were supplemented with 0.2 mg/ml geneticin (G418) in order to maintain the selection pressure. Co-transfection of pEGFP-C1, a vector carrying a red-shifted variant of wildtype green fluorescent protein cDNA from the jellyfish Aequoria victoria served as a control to monitor transfection efficiency by fluorescence microscopy.
Membrane Preparation and Protein Purification-Cells were harvested and membranes were prepared as described previously (21). The final membrane pellet was resuspended in buffer (25 mM Hepes-NaOH, pH 7.5, 1 mM EDTA, 2 mM MgCl 2 ) at a protein concentration of 8 -10 mg/ml and stored in aliquots at Ϫ80°C. In some experiments cells were treated with 100 ng/ml pertussis toxin for 1-24 h; at the end of the incubation period the cells were lysed by immersing the dish in liquid nitrogen followed by rapid thawing. The cells were scraped off and membranes were prepared. Membranes obtained from pertussis toxintreated cells expressing the human A 1 -adenosine receptor were reconstituted with purified oligomeric G proteins as described (21). In order to deplete membranes from ␤␥-dimers, membranes prepared from HEK293 cells expressing the A 1 -adenosine receptor/G␣ i-1 fusion protein were incubated in buffer (25 mM Hepes-NaOH, pH 7.5, 1 mM EDTA, 2 mM MgCl 2 ) containing 16 mM CHAPS (the ratio of detergent to membrane protein was 2:1); after 1 h on ice, the solubilized material and the extracted membrane pellet were recovered by centrifugation. The supernatant was concentrated to a protein concentration of 7 g/l and the pellet was resuspended in CHAPS-free buffer (composition as given above). About one-third of the receptors was retained in the pellet as assessed by [ 3 H]DPCPX binding; immunoblots revealed that Ͼ90% of the ␤␥-dimers had been extracted from the membrane. Recombinant G␣ i-1 was expressed in Escherichia coli JM109 harboring a plasmid encoding yeast myristoyl-CoA transferase and purified from bacterial lysates (22). Oligomeric G proteins were purified from porcine brain membranes (23) and free ␤␥-dimers were chromatographically resolved from the ␣-subunits (24) with minor modifications (23).
Immunodetection of the A 1 -Adenosine Receptor/G␣ Fusion Proteins Expressed in HEK293 Cells-Membranes (ϳ50 g of protein) prepared from HEK293 cells stably expressing the A 1 /G␣ i-1 or the A 1 /G␣ o tandem were separated on a 10% polyacrylamide SDS gel and transferred to a nitrocellulose membrane. Blots were probed with a G␣ i-1 specific antiserum (I1C, raised against the residues 160 -169 of G␣ i-1 , see Ref. 25) and with G␣ o -specific antisera (ON1, OC2 raised against the aminoterminal 16 and the carboxyl-terminal 10 amino acids of G␣ o , respectively, see Ref. 26). The immunoreactive bands on nitrocellulose blots were detected by chemiluminescence using SuperSignal chemiluminescence substrate (Pierce, Rockford, IL) and a horseradish peroxidaseconjugated anti-rabbit immunoglobulin antibody (Amersham Life Science; Buckinghamshire, United Kingdom).
Binding Experiments-Binding of the agonist radioligand [ 125 I]HPIA and of the antagonist radioligand [ 3 H]DPCPX to membranes expressing the human A 1 -adenosine receptor and the human A 1 -adenosine receptor-G protein fusion tandems were carried out as follows: membranes (150 -300 g) were resuspended in 600 l containing 25 mM Hepes-NaOH, pH 7.5, 2 mM MgCl 2 , 1 mM EDTA, 8 g/ml adenosine deaminase. The reaction was initiated by adding 900 l of prewarmed buffer containing the indicated concentrations of [ 125 I]HPIA or [ 3 H]DPCPX; an aliquot of this total binding reaction was immediately withdrawn and 1 M CPA was added to determine nonspecific binding which was typically less than 10% of total binding at 1 nM of either radioligand. Aliquots (50 l) were withdrawn from the binding reaction at the time points indicated and immediately filtered over glass fiber filters. At 30 -60 min (as indicated), the incubation was split into 2 aliquots that I]H-PIA association and dissociation were analyzed by nonlinear leastsquares curve fitting using either single exponential or double exponential equations. An F-test based on the extra sum of squares principle was carried out to determine whether the more complex (double exponential) model produced a better fit than the simpler (single exponential) model. If not indicated otherwise, the experiments were carried out at 25°C. In order to assess the energy requirements of ternary complex formation, the association and dissociation of [ 125 I]HPIA binding was determined at different temperatures (10 -30°C) and the calculated rate constants were used to generate Arrhenius plots. The resulting plots were fitted by linear regression to calculate the slopes. The activation energies were calculated from the linearized version of the Arrhenius equation: ln k ϭ A ϩ E a /RT, where ln k denotes the natural logarithm of the rate constant, R the gas constant ϭ 8.314 kJ/mol, A the pre-exponential term, and E a the activation energy. Time course experiments at the different temperatures were carried out for the following incubation periods: 3 h (10°C), 2.5 h (15°C), 2 h (20°C), 1.5 h (25°C), and 1 h (30°C). Experiments were carried out with membranes prepared from stably and transiently transfected cells; when directly compared, there was no appreciable difference between the rate constants observed.
Inhibition of Cellular cAMP Formation-Cells were grown to confluence in 6-well plates. The adenine nucleotide pool was metabolically labeled by incubating confluent cells in 6-well plates for 18 h with [ 3 H]adenine (2 Ci/well). Subsequently, the medium was replaced and the cells were preincubated for 1 h with 100 M of the phosphodiesterase inhibitor RO201724 and adenosine deaminase (1 unit/ml). The accumulation of cAMP was stimulated by the addition of 25 M forskolin and receptor-dependent inhibition was elicited in the presence of CPA (0.1-10 nM). After 3 min, the incubation was terminated by the addition of 2.5% perchloric acid; [ 3 H]cAMP was isolated by sequential chromatography on Dowex AG 50W-X4 and neutral alumina (27). Each experiment reported was carried out at least three times.

High Affinity Agonist Binding to the A 1 -Adenosine Receptor in the Presence of Varying Amounts of rG␣ i-1 Proteins-High
affinity agonist binding to G protein-coupled receptors depends on the formation of a ternary complex HRG between agonist (H), receptor (R), and G protein (G) (1). Ternary complex formation can therefore be followed by measuring the apparent association rate of a high affinity agonist radioligand. Rate constants that are observed are in the order of Յ10 6 M Ϫ1 s Ϫ1 (see also Table I) and the process is therefore too slow to be accounted for by a simple diffusion controlled mechanism. One possible rate-limiting step in HRG formation is the interaction of the activated receptor and its cognate G protein(s). We have therefore tested if ternary complex formation can be accelerated by raising the concentration of G proteins in the mem- to the A 1 -adenosine receptor and fusion proteins comprising the receptor and the indicated G protein ␣-subunits The apparent (pseudo-first order) association rate (k app ) and the dissociation rate (k off ) were derived from experiments carried out as outlined in the legends to Figs. 4 and 8; the incubation temperature was 25°C. The experimental data points were fitted to monoexponential equations or to equations describing the sum of two exponential processes; the latter models did not improve the fit significantly (p Ͼ 0.05 in all instances, F-test). The association rate constant (k on ) was calculated from the relation: k on ϭ (k app Ϫ k off )/L, where L denotes the radioligand concentration (0.92-1.1 nM). Data are means Ϯ S.E. of three to six experiments carried out in duplicate. brane. The A 1 -adenosine receptor is coupled to the pertussis toxin-sensitive G proteins of the G i /G o group (10 -15). HEK293 cells stably transfected with the human A 1 -adenosine receptor were treated with pertussis toxin (100 ng/ml for 24 h) to fully inactivate the endogenous G proteins (see Fig. 6A). Membranes prepared from these cells were incubated with 10 nM (q in Fig.  1) or 50 nM (E in Fig. 1) rG␣ i-1 and a 4-fold molar excess of purified brain ␤␥-dimers to reconstitute high affinity agonist binding. It is evident from Fig. 1A that the increase in the concentration of the heterotrimer did not result in any appreciable change in the association rate. In contrast, an increase in the concentration of the radioligand augments the apparent association rate (Fig. 1B). The interaction between receptor and G protein is essentially confined to a two-dimensional plane, i.e. the inner leaflet of the lipid bilayer, that is limited by the size of the vesicle. Because G proteins cannot be inserted into an individual vesicle to arbitrarily high levels, the variation in the concentration of G proteins that can be achieved in the vicinity of the receptor is presumably modest. As an alternative, the diffusion step, in which the activated receptor collides with the appropriate G protein, can be eliminated by fusing the receptor directly to the G protein ␣-subunit.
Expression of the A 1 -adenosine Receptor/G␣ Fusion Proteins-We have verified that fusion proteins consisting of the A 1 -adenosine receptor and of the G␣ subunits were expressed by immunodetecting the A 1 /G␣ i-1 and the A 1 /G␣ o construct with appropriate antisera ( Fig. 2A). Diffuse bands in the 70 -80-kDa range were observed in the lanes which contained membrane proteins of cells transfected with the plasmids encoding the fusion proteins. These were absent in untransfected cells or in cells expressing the A 1 -adenosine receptor. The apparent molecular mass of the purified A 1 -adenosine receptor is ϳ35 kDa (28) Membranes (ϳ10 g of protein) prepared from pertussis toxin-treated cells expressing the human A 1 -adenosine receptor were preincubated with 10 nM (q) or 50 nM (E) rG␣ i-1 which had been combined with a 4-fold molar excess of purified brain ␤␥-dimer in the presence of 10 mM CHAPS. The binding reaction was started by adding buffer prewarmed to 25°C and containing [ 125 I]HPIA (0.5 nM final concentration) yielding a final CHAPS concentration of 2.5 mM. Aliquots were withdrawn at the time points indicated and immediately filtered over glass fiber filters. Data are expressed as percent of the respective B eq values which were 68 Ϯ 9 fmol/mg for 10 nM and 146 Ϯ 23 fmol/mg for 50 nM rG␣ i-1 . Nonspecific binding determined in the presence of 1 M CPA did not change over the time course of the assay and amounted to less than 10% of the total binding seen with 10 nM rG␣ i-1 . k app was estimated to be 0.0243 Ϯ 0.004 min Ϫ1 for 10 nM and 0.0271 Ϯ 0.003 min Ϫ1 for 50 nM G␣ i-1 , respectively. Panel B, concentrationdependent increase in the apparent association rate constant k app of [ 125 I]HPIA. The apparent association rate constant was determined at the indicated concentrations of [ 125 I]HPIA; the binding reaction was carried out with membranes from HEK293 cells expressing the A 1adenosine receptor. Assay conditions were as described for panel A. The pseudo-first order rates were calculated by nonlinear least squares curve fitting and the linear regression of the secondary plot was calculated. The y intercept yielding an estimate of k off was 0.02 min Ϫ1 ; k on estimated from the slope of the regression line was 4. were resolved on a 10% SDS-polyacrylamide gel and transferred to nitrocellulose. The blot was probed sequentially with antiserum I1C and ON1, which are specific for G␣ i-1 and G␣ o , respectively. Lane 5, purified rG␣ i-1 (50 ng) was applied as a standard. Panel B, two aliquots (50 g/lane) of HEK293 cell membranes containing the A 1 -adenosine receptor (lanes 2 and 4) or the A 1 /G␣ o fusion protein (lanes 1 and 3) were applied onto a 10% SDS-polyacrylamide gel. After electrophoretic transfer, the nitrocellulose membrane was cut into halves; these were probed with either antiserum ON1 or OC2 which are directed against the amino and carboxyl terminus of G␣ o, respectively. The migration of molecular mass markers are indicated. respectively. Thus, the size of the immunoreactive material is consistent with the expected molecular mass of the fusion proteins; the broad staining pattern presumably arises from microheterogeneity due to glycosylation of the receptor moiety in the fusion protein rather than from partial proteolysis. We have employed the antisera ON1 and OC2 which are directed against the amino and carboxyl terminus of G␣ o , respectively (26). Immunostaining with these two antisera visualized a predominant band in the range of 70 -80 kDa; in addition, antiserum ON1 detected a band at about 50 kDa (lane 1, Fig.  2B). This band was not seen in cells lacking the fusion protein (lane 2, Fig. 2B) and in the immunoblot with antiserum OC2 (lane 3 in Fig. 2B). Because this band retains the aminoterminal epitope, it may result from proteolysis within the G␣ o moiety of the fusion protein. It is, however, evident that this cleavage product represents only a very minor fraction of the total immunoreactivity. We therefore conclude that the bulk of the fusion proteins comprise an intact G protein moiety. Saturation binding experiments with the antagonist radioligand [ 3 H]DPCPX revealed that the fusion proteins A 1 /G␣ i-1 and A 1 / G␣ o were stably expressed up to levels of ϳ5 and 10 pmol/mg, respectively. Because HEK293 cells do not express detectable levels of G␣ o , we subsequently present data on the comparison of the fusion protein A 1 /G␣ i-1 with the wild-type receptor (which interacts with the G␣ i complement endogenous to HEK 293 cells). We stress, however, that the ligand binding kinetics of the fusion protein A 1 /G␣ o were essentially identical to that of A 1 /G␣ i-1 (data not shown; see also Table I).
The association (Fig. 3A) and dissociation kinetics (Fig. 3B) of the antagonist [ 3 H]DPCPX were found to be virtually identical in binding assays that were carried out in parallel with membranes carrying the unfused A 1 -adenosine receptor (E in Fig. 3, A and B) and those containing the fusion protein A 1 / G␣ i-1 (q in Fig. 3, A and B). For the A 1 -adenosine receptor, the apparent (pseudo-first order) association and the dissociation rates were calculated as 0.23 Ϯ 0.05 min Ϫ1 and 0.14 Ϯ 0.06 min Ϫ1 , respectively; the corresponding values for the fusion protein A 1 /G␣ i-1 were k app ϭ 0.37 Ϯ 0.08 min Ϫ1 and k off ϭ 0.16 Ϯ 0.05 min Ϫ1 . The kinetically derived estimates for the dissociation constant K D were 1.5 and 0.76 nM for the A 1adenosine receptor and for A 1 /G␣ i-1 , respectively. The capacity of the A 1 /G␣ i-1 tandem to impinge on an effector was investigated by assessing the inhibitory regulation of adenylyl cyclase. It is evident from the data shown in Fig. 3C that inhibition of forskolin-induced [ 3 H]cAMP formation was observed over a reasonably comparable concentration range of the A 1receptor agonist CPA in cells expressing the A 1 -adenosine receptor (E in Fig. 3C) and in cells expressing the fusion protein A 1 /G␣ i-1 (q in Fig. 3C). Half-maximum inhibition was observed at 0.3 Ϯ 0.1 and 0.6 Ϯ 0.1 nM CPA for the A 1 -adenosine receptor and A 1 /G␣ i-1 , respectively. Taken together, these data indicate that the A 1 /G␣ i-1 fusion protein is expressed as a functional receptor capable of regulating its typical second messenger pathway.
High Affinity Agonist Binding to the A 1 /G␣ i-1 Fusion Protein-Given the close proximity of receptor and G protein in the A 1 /G␣ i-1 fusion protein, a more rapid association rate of [ 125 I]H-PIA was to be expected, if the association of receptor and G protein were the rate-limiting step in the formation of the ternary complex. The experiments summarized in Fig. 4 and Table I show that this was not the case; the apparent (pseudofirst order) association (k app ) and dissociation (k off ) rates of the agonist [ 125 I]HPIA on membranes prepared from HEK293 cells stably expressing the human A 1 -adenosine receptor (E in Fig.  4) or the A 1 /G␣ i-1 fusion protein (q in Fig. 4) were comparable; accordingly the calculated association rate constants k on were similar within the experimental error (Table I). We have, furthermore, ruled out that the energy requirements for formation (or break-up) of the ternary complex differed between the na- tive A 1 -adenosine receptor and the fusion protein A 1 /G␣ i-1 by determining the kinetics of agonist binding over a temperature range from 10°C to 30°C. The apparent (pseudo-first order) association rates k app (open symbols in Fig. 5, A and B) and the dissociation rates (closed symbols in Fig. 5, A and B) obtained in these experiments were used to generate Arrhenius plots for [ 125 I]HPIA binding to the human A 1 -adenosine receptor (Fig.  5A) and the A 1 /G␣ i-1 fusion protein (Fig. 5B). The difference between k app and k off corrected for the ligand concentration yields the Arrhenius plot for the rate constant k on of the forward reaction (Fig. 5C). It is evident that linear Arrhenius plots were obtained in all cases; more importantly, a comparison of the slopes of the lines calculated for the native receptor and the fusion protein A 1 /G␣ i-1 shows that they are reasonably similar. Thus, the two proteins do not differ with respect to their temperature dependence of ternary complex formation.
Interaction of the A 1 /G␣ i-1 Fusion Protein with ␤␥-Dimers-In the absence of ␤␥-dimers, the A 1 -adenosine receptor interacts only poorly with G␣ i and G␣ o subunits and the affinity of the receptor is greatly enhanced by the presence of ␤␥-dimers (10). As mentioned above the fusion protein A 1 /G␣ i-1 accumulated to fairly high levels (stable cell lines expressing up to 5 pmol/mg); these amounts still represent only a minor fraction of the total membrane level of G protein ␤␥-dimers (the lower limit being 100 pmol/mg as detected by an antiserum recognizing ␤ 1 and ␤ 2 ). Nevertheless, the G␣ moiety in the fusion protein may not have access to the total membrane pool of ␤␥-dimers because these may not exchange freely between the G␣-subunits endogenous to the membrane and the G␣ moiety of the fusion protein. In order to rule out that ␤␥-dimers were limiting, we transiently co-expressed ␤ 1 ␥ 2 -subunits together with the fusion protein A 1 /G␣ i-1 . A comparison between membranes from control cells and those from cells co-transfected with plasmids encoding ␤ 1 and ␥ 2 showed that the overexpression of ␤␥-dimers did not affect the kinetics of [ 125 I]HPIA binding to the A 1 /G␣ i-1 fusion protein (data not shown). Hence, the levels of ␤␥-dimers in the membrane may suffice to support the interaction of the receptor and ␣-subunit moiety in the fusion protein. Alternatively, it is conceivable that ␤␥-dimers are not necessary to support the interaction when receptor and ␣-subunits are fused into a single molecule. The amino-terminal ␣-helix is required for binding the ␤␥-dimer (17). Because this part of the protein is directly tethered to the receptor moiety in the fusion protein, the interaction with ␤␥-dimers may be sterically hindered. Pertussis toxin-catalyzed ADPribosylation of G␣ i or G␣ o is supported by ␤␥-dimers and the rate of the reaction depends on the interaction of ␣-subunits and ␤␥-dimers (24). We have exploited this property of the toxin to test if, in the intact cells, the fusion protein combines with ␤␥-dimers. As expected, incubation of cells expressing the human A 1 -adenosine receptor with pertussis toxin led to a time-dependent loss of high affinity [ 125 I]HPIA binding (E in Fig. 6A); this was also seen in cells expressing the A 1 /G␣ i-1 fusion protein (q in Fig. 6A). The differences between the two cell lines were modest. If the data were fitted to an equation describing a monoexponential decay, rate constants of 0.47 Ϯ 0.1 and 0.68 Ϯ 0.16 h Ϫ1 were calculated for the A 1 /G␣ i-1 fusion protein and the human A 1 -adenosine receptor, respectively. It is also evident from Fig. 6 that there was a delay before the action of pertussis toxin was detectable. This hysteresis presumably resulted from the transmembrane permeation and activation of the toxin. Finally, a fraction of the A 1 /G␣ i-1 fusion protein (about 15%) was resistant to the action of pertussis toxin. Access of the toxin to the carboxyl terminus of the G protein ␣-subunit may be sterically hindered by the presence of the receptor moiety. Regardless of the underlying reasons for these minor differences between fusion protein and wild-type receptor, it is safe to conclude that the bulk of the fusion protein interacted with ␤␥-dimers in the intact cells. Furthermore, the modest difference in the rate constants suggested that the affinity of the fused ␣-subunit moiety for ␤␥-dimers was somewhat reduced, but not dramatically compromised.
While these observations indicate that an interaction of the fusion protein with ␤␥-dimers is possible, they do not address the question if ␤␥-dimers are required to support ternary complex formation upon binding of an agonist to the fusion protein. This requirement can be seen for the native human A 1 -adenosine receptor in Fig. 6B; if membranes were prepared from pertussis toxin-treated cells (in which high affinity [ 125 I]HPIA binding was abolished), addition of rG␣ i-1 (at a limiting concentration of 12.5 nM) did not reconstitute [ 125 I]HPIA binding per se but required the addition of ␤␥-dimers to restore high affinity agonist binding (E in Fig. 6A). Because this approach cannot be employed with the A 1 /G␣ i-1 fusion protein, we have gener-ated membranes that were depleted of ␤␥-dimers (Ͼ90% as estimated by immunoblotting) by detergent extraction. Under these conditions about one-third of the fusion protein is retained in the particulate fraction (as assessed by [ 3 H]DPCPX binding). Addition of ␤␥-dimers promoted high affinity agonist binding (q in Fig. 6B) and the half-maximum effect was seen at 22 Ϯ 5 nM ␤␥-dimers; this affinity estimate is somewhat lower than that determined in parallel for the unfused A 1 -adenosine receptor (EC 50 ϭ 11 Ϯ 3 nM). More importantly, these observations indicate that ␤␥-dimers are required for ternary complex formation by the A 1 /G␣ i-1 fusion protein.
Alterations in the Affinity of the Receptor for the G Protein Affects the Dissociation of the Ternary Complex-In order to test the hypothesis that the dissociation rate rather than the rate of the forward reaction is crucial for ternary complex formation, we have replaced cysteine 351 of G␣ i-1 by glycine or isoleucine. This substitution not only renders the G protein resistant to ADP-ribosylation by pertussis toxin but also yields G proteins with an altered affinity for their cognate receptors (29). These fusion proteins A 1 /G␣ i-1 C351G and A 1 /G␣ i-1 C351I were expressed in HEK293 cells. We stress that cells expressing these fusion proteins were always pretreated with pertussis toxin to prevent the receptor in the fusion protein from interacting with the G i proteins endogenous to the membrane (30). Two approaches were used to estimate the affinity of the receptor for the mutated G protein moiety in the tandem. First, we have determined the ability of GDP to suppress the formation of the ternary complex. Because the activated receptor reduces the affinity of the G protein for GDP (by promoting GDP release), an excess of GDP conversely lowers the affinity of the G protein for the receptor (1). Fig. 7A shows the inhibition of high affinity [ 125 I]HPIA binding to membranes prepared from HEK293 cells expressing the wild-type fusion protein A 1 /G␣ i-1 (q) as well as the mutated versions A 1 /G␣ i-1 C351I (E) and A 1 -G␣ i-1 C351G (⌬) by GDP. Half-maximum inhibition by GDP was observed at 5.01 Ϯ 1.55 M for A 1 /G␣ i-1 , 1.44 Ϯ 0.35 M for A 1 -G␣ i-1 C351I, and 0.39 Ϯ 0.16 M for A 1 -G␣ i-1 C351G. The second approach relied on the use of suramin. This compound binds directly to G protein ␣-subunits (31) and competes with the activated receptor for binding to the G protein; suramin can therefore be employed to estimate the affinity of a receptor for a G protein (21,32). High affinity agonist binding to A 1 /G␣ i-1 fusion proteins mutated at Cys 351 of the G␣ i-1 moiety was more readily suppressed by suramin (Fig. 7B). The IC 50 of suramin was 8.55 Ϯ 2.2 M for A 1 /G␣ i-1 (q in Fig. 7B), 2.99 Ϯ 0.67 M for A 1 /G␣ i-1 C351I (E in Fig. 7B), and 1.02 Ϯ 0.06 M for A 1 /G␣ i-1 C351G (⌬).
Based on the data summarized in Fig. 7, we concluded that G␣ i-1 C351G exhibited the lowest affinity for the A 1 -adenosine receptor; therefore the kinetics of ternary complex formation of the fusion protein A 1 /G␣ i-1 C351G were investigated in detail. The mutation did not affect the forward reaction (⌬ in Fig. 8A), the calculated k on being comparable within experimental error to that seen in the fusion protein containing the wild type version of G␣ i-1 (Table I); in contrast, the ternary complex formed by A 1 /G␣ i-1 C351G dissociated more readily (⌬ in Fig.  8B). The k off of A 1 /G␣ i-1 C351G was about five times faster than that determined for the fusion protein A 1 /G␣ i-1 (Table I) and thus in reasonable agreement with the estimated difference in affinity obtained by the approaches in Fig. 7. The reduced stability of the ternary complex was seen over the entire temperature range investigated for A 1 /G␣ i-1 C351G (⌬ in Fig. 8C) and was also seen in the fusion protein A 1 /G␣ i-1 C351I (E in Fig. 8C). prepared from pertussis-toxin treated cells expressing the A 1 -receptor (E) or prepared from cells expressing the A 1 /G␣ i-1 fusion proteins (q) and subsequently depleted of ␤␥-subunits by detergent extraction (see "Experimental Procedures") were preincubated with (E) or without (q) 12.5 nM rG␣ i-1 and increasing amounts bovine ␤␥-subunits and [ 125 I]H-PIA binding (0.5 nM) was carried out as described in the legend to Fig.  1. Data are means of duplicate determinations; two additional experiments gave comparable results.

DISCUSSION
A fusion construct of the ␤ 2 -adrenergic receptor and G␣ s was originally shown to yield a functional protein capable of efficiently activating its prototypic effector adenylyl cyclase (33,34); this approach has more recently been extended to other receptor/G protein tandems and several aspects of receptor-G protein coupling have been investigated with these constructs (29, 35; for review, see Ref. 36). In the present work, we have used fused tandems of the A 1 -adenosine receptor and G protein ␣-subunits to test if the association of receptor and G protein was rate-limiting for ternary complex formation. It is evident from our experiments that the direct fusion of the receptor to the G protein ␣-subunit does not accelerate the rate of ternary complex formation, although the local reactant concentration must, by definition, be very high. We rule out that the slow association rate is due to a lack of ␤␥-dimers because overexpression of ␤␥ does not affect the rate of the complex formation. Similarly, in reconstitution experiments, no appreciable acceleration in the rate of ternary complex formation was observed upon variation in the concentration of G protein oligomers. Our findings rather support the interpretation that the rate-limiting step(s) are those that govern the transition from the inactive to the active conformation of the receptor (3). In fact, in the absence of a G protein, the rate of agonist binding to the purified ␤ 2 -adrenergic receptor has been inferred from the alterations in fluorescence of appropriately labeled receptors; the rate of fluorescence change was also too slow to be accounted for by a simple diffusion-controlled reaction and was therefore proposed to reflect the rate-limiting change in conformation to the active species HR* (37). Finally, fusing the receptor and G protein moiety did not lower the activation energy required for complex formation; this is also consistent with the interpretation that the interaction of receptor and G protein is not ratelimiting. For both, the A 1 -adenosine receptor and the fusion protein A 1 /G␣ i-1 , the activation energies E a were estimated in the range of 45 kJ mol Ϫ1 . While this value is higher than that reported for the ␣ 2 -adrenergic receptor in human platelet membranes (38), it is comparable to estimates that can be calculated from kinetic data that have been reported for high affinity agonist binding to the A 1 -adenosine receptor in brain membranes (39).
Our interpretation relies on two simplifications: (i) we assume the rate of [ 125 I]HPIA dissociation is limited by the stability of the interaction between R* and G; this assumption is justified for a high affinity agonist because the affinity of R* for G␣ i . ␤␥ is lower than that for [ 125 I]HPIA. The apparent affinity of the human A 1 -adenosine receptor in HEK 293 membranes for G␣ i . ␤␥ has previously been estimated by two independent methods and is in the range of 10 nM (21); this value is also consistent with the affinity estimate obtained in reconstitution experiments (12). In contrast the K D for high affinity binding of [ 125 I]HPIA is about 1 nM. (ii) Ternary complex formation is treated as a bimolecular reaction; because the rate of ternary complex formation is independent of G protein concentration (i.e. not limited by the association of the receptor with the G protein), we assume the transition of the inactive conformation R to R* as the rate-limiting step. The agonist may stabilize R* by shifting the spontaneous equilibrium between R and R* or induce the formation of R* (3); regardless of which of these two models is applicable, ternary complex formation can be treated as the product of a bimolecular reaction of R* and G.
It has long been appreciated that the carboxyl terminus of G␣-subunits is important for binding of receptors; accordingly, mutations that substitute the cysteine residue at the carboxylterminal position Ϫ4 are expected to affect the affinity of the G protein for the receptor (16,17). This has recently been systematically investigated by examining the interaction of the ␣ 2A -adrenergic receptor with G␣ i-1 , in which this residue was replaced by the other 19 naturally occurring amino acids (29). While substitution with glycine reduces the affinity of both, the A 1 -adenosine and the ␣ 2A -adrenergic receptor, it is worth noting that the replacement by isoleucine has different effects; it increases the affinity of the mutated G protein moiety in the ␣ 2A -adrenergic receptor/G␣ i-1 fusion protein (29) but reduces the affinity in the A 1 /G␣ i-1 tandem. Analogous discrepancies have been noted in related studies. Some, but not all, G qcoupled receptors can couple to mutated forms of G␣ s in which the last 5 amino acids were replaced with the corresponding residues of G␣ q ; the same is true for G s -coupled receptors that are confronted with a carboxyl-terminal altered G␣ q (40). Similarly, peptides derived from the carboxyl terminus of the cognate G protein ␣-subunits are capable of stabilizing the receptor in the conformation that binds agonists with high affinity; this can be seen with some, but not all, receptors (41,42). Taken together, these findings highlight the different modes by which receptors engage the same G protein ␣-subunit and support the concept that the contact site is different enough in individual receptor-G protein complexes to allow this site to be considered as a potential target for inhibitors (32,43).
The ability of a given receptor to engage a G protein may be limited by its ability to associate with the appropriate G protein; alternatively, the activated receptor may form complexes with various types of G proteins but only those complexes that dissociate slowly are stabilized to support efficient signaling. In order to differentiate between these two possibilities, we have created fusion proteins in which the affinity of the receptor for the G protein was lowered by substituting the critical cysteine residue with glycine or isoleucine. Our data clearly show that the rate of ternary complex formation is not affected by lowering the affinity of the G protein for the receptor; in contrast, the ternary complex dissociates more rapidly. These findings provide an explanation for the fidelity of signaling that is usually observed; they also account for the observation that upon overexpression of receptors the fidelity is lost; i.e. overexpressed receptors have the propensity to interact with G proteins and activate down-stream signaling pathways that are not subject to their physiological regulation (reviewed in Ref. 6). Under these conditions, low concentrations of agonists typically suffice to promote the interaction of the receptor with the cognate G protein; however, high concentrations of agonists, i.e. in excess of those necessary to saturate the receptor, typically result in activation of one or more additional G proteins and their downstream signaling pathways. Our observations indicate that this high agonist concentration is required to increase the lifetime of the ternary complex resulting from the interaction of the receptor with the non-physiological G protein(s). In contrast, at low agonist concentrations, the "strength of signal" (6) does not suffice, because the lifetime of these ternary complexes is too short to support signal transduction. This model also predicts that a partial agonist is converted to an antagonist, if the affinity of the G protein for the receptor is lowered. This has indeed been recently observed; the intrinsic activity of the partial agonist clonidine is absent in a fusion protein composed of the ␣ 2A -adrenergic receptor and G␣ i-1 G351C and the compound acts as an antagonist (44). Thus, the different dissociation rates of the ternary complex allow for a kinetic proofreading mechanism; the activated receptor can associate with various G proteins but only the cognate G protein(s) are retained in stable ternary complexes. Obviously, in intact cells additional factors contribute to the specific interaction of receptors and G proteins because the signaling molecules are compartmentalized. Receptors, for instance, are not uniformly distributed over the membrane in polarized cells (45) and in neurons (46) and components of the cytoskeleton and additional proteins are involved in the organization of G proteins and effectors (47,48). Nevertheless, we propose that kinetic proofreading is important for a receptor to faithfully select its cognate partner(s) from the total G protein pool present in its vicinity.