The G Protein β Subunit Is a Determinant in the Coupling of Gsto the β1-Adrenergic and A2a Adenosine Receptors

The signaling specificity of five purified G protein βγ dimers, β1γ2, β2γ2, β3γ2, β4γ2, and β5γ2, was explored by reconstituting them with Gs α and receptors or effectors in the adenylyl cyclase cascade. The ability of the five βγ dimers to support receptor-α-βγ interactions was examined using membranes expressing the β1-adrenergic or A2a adenosine receptors. These receptors discriminated among the defined heterotrimers based solely on the β isoform. The β4γ2 dimer demonstrated the highest coupling efficiency to either receptor. The β5γ2 dimer coupled poorly to each receptor, with EC50 values 40–200-fold higher than those observed with β4γ2. Strikingly, whereas the EC50 of the β1γ2 dimer at the β1-adrenergic receptor was similar to β4γ2, its EC50 was 20-fold higher at the A2a adenosine receptor. Inhibition of adenylyl cyclase type I (AC1) and stimulation of type II (AC2) by the βγ dimers were measured. βγ dimers containing Gβ1–4 were able to stimulate AC2 similarly, and β5γ2 was much less potent. β1γ2, β2γ2, and β4γ2inhibited AC1 equally; β3γ2 was 10-fold less effective, and β5γ2 had no effect. These data argue that the β isoform in the βγ dimer can determine the specificity of signaling at both receptors and effectors.

Signal transduction involving heterotrimeric G proteins 1 is a universal mechanism for the integration of extracellular stimuli such as hormones, neurotransmitters, odorants, and light (1,2). The components involved in this signaling cascade are diverse, including a large number of receptors, G protein ␣ and ␤␥ subunits and effectors. Even though the diversity of the proteins in this system could potentially account for the known specificity of signaling in differentiated cells, the mechanisms for determining specificity are not completely defined. The ␤-adrenergic receptor is one of the most well characterized seven transmembrane spanning receptors, and provides an excellent example of selective coupling to a particular ␣ subunit, G s . When activated, G s can stimulate all nine adenylyl cyclase isoforms (3,4). The G protein ␤␥ dimer, when released after receptor activation, is also able to regulate adenylyl cyclase (5). However, the regulation of the various isoforms of adenylyl cyclase by the ␤␥ dimer is much more selective; apparently, only AC2, AC4 (6,7), and AC7 (8) are stimulated by ␤␥, whereas the neuronal-specific AC1 (4) and possibly AC5 and AC6 are inhibited by the dimer (9). Moreover, there is evidence that AC2 does not respond well to dimers composed of certain ␤ and ␥ subunits (10) or to dimers containing the phosphorylated ␥ 12 subunit (11). Thus, to understand fully the regulation of adenylyl cyclase by a G s -coupled receptor, one needs to know which ␤␥ dimers are most likely to support receptor G protein coupling and the effects of ␤␥ dimers on the various isoforms of adenylyl cyclase.
The number of functionally distinct ␤␥ dimers is potentially very large, with seven G protein ␤ isoforms (including two splice variants) and 12 ␥ isoforms characterized to date (12)(13)(14). Most in vitro studies involving coupling of receptors to G s ␣ or regulation of adenylyl cyclases by distinct ␤␥ dimers have used dimers containing the ␤ 1 , ␤ 2 , or ␤ 5 subunits (15,16). The ubiquitous cellular and tissue distribution of G s ␣ provides the potential for interaction with all five ␤ isoforms and underscores the importance of understanding the role of the different ␤ isoforms on signaling pathways involving G s ␣. For example, the antisense studies of Kleuss et al. (17)(18)(19) suggest that specific isoforms of the heterotrimer couple to different receptors, and a number of in vitro studies imply that defined ␤␥ dimers may be released upon receptor activation (16,20,21). In addition, isolation of G protein heterotrimers from a variety of tissues using chromatography or immunoprecipitation has shown that certain ␤ and ␥ subunits preferentially associate with one another as well as with distinct ␣ O isoforms (22,23). These data suggest that specific combinations of G protein subunits do exist in vivo and may have specialized roles in various signaling cascades.
To examine the roles of the various ␤ subunits in receptor-G s coupling, and in regulating adenylyl cyclase, recombinant G s ␣ and ␤␥ dimers containing ␤ 1-5 complexed with ␥ 2 were expressed in baculovirus-infected Sf9 insect cells and purified. Proteins were then reconstituted into partially purified Sf9 cell membranes overexpressing either the ␤ 1 -adrenergic receptor, the A2a adenosine receptor, AC1 or AC2. The effects of the ␤ 1-5 ␥ 2 combinations were measured in four assays as follows: 1) the ability to couple the G s ␣ subunit to the ␤ 1 -adrenergic receptor; 2) the ability to couple the G s ␣ subunit to the A2a adenosine receptor; 3) the ability to stimulate AC2; and 4) the ability to inhibit AC1. Clear differences were observed among the five ␤␥ dimers in both receptor coupling and effector regulation, suggesting that the diversity of the ␤ subunit contributes extensively to signaling specificity.

EXPERIMENTAL PROCEDURES
Construction of Recombinant Baculoviruses-Construction of baculoviruses encoding the ␤ 1 , ␤ 2 , ␤ 5 , ␥ 2 , ␥ 2FH , the G s ␣ and, G i1 ␣ subunits has been described (11, 24 -26). The viruses encoding AC1 and AC2 were the kind gift of R. Iyengar (27,28). Baculoviruses encoding the rat ␤ 1 -adrenergic receptor and the human A2a adenosine receptor were gifts from E. Ross (University of Texas, Southwestern Medical Center) and J. Linden (University of Virginia), respectively (29). The human ␤ 3 cDNA (30), a gift from S. R. Ikeda (Guthrie Institute), was excised from pCI with EcoRI and NotI; the mouse ␤ 4 cDNA (31), a gift from W. F. Simonds (National Institutes of Health) was excised from pcDNA3 with BamHI and XbaI. The human ␤ 3 s cDNA, a gift from D. Rosskopf (Institute for Pharmacology, Essen, Germany), is a truncated variant of the full-length ␤ 3 cDNA, in which the ␤ 3 s protein product has a deletion of amino acids 168 -208 (32); excision of the ␤ 3 s cDNA from pGEMT was accomplished with BamHI and PstI. The existing restriction sites were used to ligate digestion products into the multiple cloning site immediately downstream of the polyhedron promoter in the baculovirus transfer vector, pVL1393. All clones were sequenced to confirm the fidelity of the cDNA in pVL1393. Recombinant baculoviruses for ␤ 3 , ␤ 3 s, and ␤ 4 were prepared by co-transfection of linear wild type BaculoGold ® viral DNA (PharMingen) with pVL1393 transfer vectors containing the specific ␤ sequences into Sf9 cells as described (26) and purified with one round of plaque purification (33).
Expression and Purification of Recombinant G Protein ␣ and ␤␥ Dimers-Sf9 cells were infected with recombinant baculoviruses encoding the desired ␣ and/or ␤␥ dimer combinations at a multiplicity of infection of three and harvested 48 -60 h after infection. ␤␥ dimer combinations containing ␤ 1-4 and ␥ 2 were purified by G i1 ␣ affinity chromatography as described (34). The dimer containing ␤ 5 was expressed with a ␥ 2 subunit engineered to have a hexahistidine and FLAG tag (26) at the N terminus (␥ 2HF ) and purified from isolated Sf9 cell membranes by FLAG affinity and Ni 2ϩ affinity chromatography, followed by anion exchange chromatography (16).
Mass spectrometry was used to examine post-translational processing of the ␥ 2 subunit. Purified ␤␥ dimers were analyzed by matrixassisted laser desorption ionization-mass spectrometry to obtain masses of the ␥ subunits as described in Lindorfer et al. (35). For ␤␥ dimers with a protein concentration of less than 150 ng/l, acetone precipitation was used to concentrate the protein before mass analysis (36). Post-translational processing of the ␥ 2 isoform includes cleavage of the N-terminal methionine, acetylation of the resulting N-terminal alanine, geranylgeranylation at the cysteine four residues from the C terminus, cleavage of three C-terminal residues, and carboxymethylation of the resulting C-terminal geranylgeranylated cysteine. These post-translational modifications have been observed in ␥ subunits isolated from bovine brain (37) and in Sf9 cells (35). The predicted mass of the properly processed ␥ 2 isoform is 7750 Da; insertion of a His-Flag (HF) tag at the N terminus increases the predicted mass to 10,013 Da. Mass spectra of the purified ␤␥ isoforms containing either ␥ 2 or ␥ 2HF demonstrated that the major mass in each spectrum was compatible with these predicted masses within the accuracy of the instrument (35). For example, in one set of purified ␤␥ dimers, the experimental masses of the ␥ 2 subunits were as follows: ␤ 1 ␥ 2 , 7758 Da; ␤ 2 ␥ 2 , 7764 Da; ␤ 3 ␥ 2 , 7760 Da; ␤ 4 ␥ 2 , 7759 Da; and ␤ 5 ␥ 2FH , 10,020 Da.
Attempts were made to purify ␤ 3 s combined with various ␥ subunits. A protein with the appropriate molecular weight was expressed in Sf9 cells as judged by immunoblotting with a ␤-common antibody (PerkinElmer Life Sciences 808); however, the major barrier to purification was that it was not possible to solubilize the protein from the Sf9 cell pellet. For example, soluble extracts of whole cell pellets prepared using 1% (v/v) Genapol, 1% (w/v) CHAPS, or 1% (w/v) Cholate contained little ␤ 3 s protein in supernatant fractions that could be detected with the ␤-common antibody. Expression of various ␣ and or the ␥ 2FH , ␥ 5 , and ␥ 7 subunits with ␤ 3 s in Sf9 cells did not affect the solubility of the protein (data not shown), and thus characterization of this protein was not pursued. G s ␣ was overexpressed with a ␤ 1 subunit engineered to have a hexahistidine and FLAG tag (11) at the N terminus (␤ 1HF ), along with ␥ 2HF , and purified using a modification of the method described by Kosaza and Gilman (38). Briefly, harvested cells were resuspended in half the infection volume with cell lysis buffer (20 mM Tris, pH 8.0, 10 M GDP, 17 g/ml PMSF, and 2 g/ml pepstatin, leupeptin, and aprotinin). After resuspension, cells were lysed by nitrogen cavitation (25), and membranes were collected by centrifugation at 28,000 ϫ g for 20 min at 4°C. A Potter-Elvehjem homogenizer was used to resuspend the pellets in a quarter of the original resuspension volume (ϳ63 ml) of cell lysis buffer containing 10 g/ml DNase. After a 15-min incubation on ice, membranes were collected again by centrifugation at 28,000 ϫ g for 20 min at 4°C and resuspended with a Potter-Elvehjem homogenizer in a volume of extraction buffer (20 mM Tris, pH 8.0, 150 mM NaCl, 1 mM MgCl 2 , 0.5% (v/v) Genapol, 1 mM ␤-mercaptoethanol, 50 M GDP, 17 g/ml PMSF, and 2 g/ml of leupeptin, aprotinin, and pepstatin) equivalent to 10 times the weight of the original cell pellet. Membranes containing expressed G protein were resuspended and stirred with extraction buffer for 1 h at 4°C, followed by centrifugation at 142,000 ϫ g for 1 h at 4°C; the solubilized G s ␣/␤ 1FH ␥ 2FH supernatant extracts (typically about 100 ml) were flash-frozen in liquid nitrogen and stored at Ϫ80°C.
To begin the G s ␣ purification, the extract was diluted with an equal volume of Ni 2ϩ column buffer (20 mM Tris, pH 8.0, 150 mM NaCl, 1 mM MgCl 2 , 0.2% (v/v) Genapol, 1 mM ␤-mercaptoethanol, 10 M GDP, 5 mM imidazole, 17 g/ml PMSF, and 2 g/ml pepstatin, leupeptin, and aprotinin) and loaded onto a Ni 2ϩ -NTA Superflow column at 2 ml/min. Unless otherwise noted, all steps were performed at 4°C. The volume of the column bed was ϳ5% of the volume of the Genapol extract. The column was washed with 6 column volumes of Ni 2ϩ column buffer, 6 column volumes of Ni 2ϩ column buffer containing 300 mM NaCl, and 3 more column volumes of Ni 2ϩ column buffer. At this point, the column and buffers were warmed to room temperature for 10 -20 min, and G s ␣ was activated and eluted with 4 column volumes of activation buffer (Ni 2ϩ column buffer containing 50 mM MgCl 2 , 10 mM NaF, and 30 M AlCl 3 ) also warmed to room temperature. Although the increased temperature facilitates activation of the ␣ subunit, this step should be completed as quickly as possible, as functional activity of ␣ decreases with prolonged elevation of temperature. Pilot experiments using SDS-PAGE to identify the G s ␣ subunit indicated that the first 8 ml of eluate after the void volume contained the protein. Thereafter, these fractions were collected on ice and pooled. The fractions containing G s ␣ were diluted 5-fold with 15Q buffer (20 mM Tris, pH 8.0, 1 mM EDTA, 1 mM MgCl 2 , 0.1% (w/v) CHAPS, 1 mM DTT, 10 M GDP) and loaded onto a 200-l 15Q anion exchange column. This dilution facilitates adsorption of the protein to the column by reducing of the Cl Ϫ concentration to ϳ50 mM. In addition to concentrating the protein, the 15Q step is necessary to remove AlF 4 Ϫ , imidazole, and Genapol, which would itself be concentrated along with the protein in the next concentration step. After the protein was loaded, the column was washed with 15Q buffer containing 10 mM NaCl (15Q buffer A) for 20 min at 1 ml/min. Protein was then eluted with 15Q buffer containing 600 mM NaCl (15Q buffer B) in a linear gradient of 0 -50% 15Q buffer B over 15 min. One-ml fractions were collected, and 12% SDS-PAGE followed by either immunoblotting or silver staining with purified Gs ␣ as a standard was used to determine which fractions contained G s ␣.
Fractions from the 15Q column containing G s ␣ were pooled and concentrated with a Centricon 30 that had been passivated with a 1% BSA solution as described (16). The concentrated protein was diluted 10-fold with 15Q buffer containing 100 mM NaCl to reduce the high NaCl concentration that resulted from the elution from the 15Q column, and then concentrated once more to a volume of 100 -200 l, aliquoted, and stored at Ϫ80°C. The yield of purified G s ␣ from 10 g of Sf9 cell pellet (wet weight) was typically 10 -20 g. All protein estimates were determined using scanning densitometry of silver-stained gels as described previously (26), with standard curves generated from ovalbumin standards. G i1 ␣ was purified by a similar method with the following exceptions. The Ni 2ϩ column buffer contained 20 mM Tris, pH 8.0, 150 mM NaCl, 0.2% (w/v) CHAPS, 1 mM ␤-mercaptoethanol, 10 M GDP, 5 mM imidazole, 17 g/ml PMSF, and 2 g/ml pepstatin, leupeptin, and aprotinin. Protease inhibitors and imidazole were removed from the elution step, and the G i1 ␣ was taken directly to a Centricon 30 where it was concentrated and diluted 10-fold with Ni 2ϩ column buffer supplemented with 2 mM MgCl 2 . This step was repeated, and 100 -200 l of G i1 ␣ at 100 -200 ng/l were stored in aliquots at Ϫ80°C. As one criterion for the viability of the G s and G i1 ␣ subunits, the ability of the proteins to bind GTP␥S in solution was measured. The stoichiometry of nucleotide binding of two preparations of G s ␣ averaged 0.9 mol/mol. The G i1 ␣ subunit bound GTP␥S at a stoichiometry of ϳ0.3 mol/mol and also coupled effectively to the A1 adenosine receptor in assays similar to the one shown in Fig. 2A (data not shown).
Preparation of Membranes Containing Recombinant ␤ 1 -Adrenergic Receptors, A2a Adenosine Receptors, or Adenylyl Cyclases-Sf9 cells were infected with recombinant baculoviruses encoding either the rat ␤ 1 -adrenergic receptor, the human A2a adenosine receptor, and either type I or type II adenylyl cyclase (27,28). In the case of the ␤ 1adrenergic receptor, harvested cells were resuspended in membrane homogenization buffer (20 mM HEPES, pH 7.5, 2 mM MgCl 2 , 1 mM EDTA, 17 g/ml PMSF, and 2 g/ml leupeptin and aprotinin), and cells were lysed by nitrogen cavitation. The cell lysate was centrifuged at 750 ϫ g to pellet unbroken cells and nuclei. Membranes were prepared from the supernatant of the low speed spin by centrifugation at 28,000 ϫ g for 20 min at 4°C. Endogenous G proteins were then stripped from the membranes or inactivated by incubation with urea. Membranes containing the ␤ 1 -adrenergic receptor were homogenized with resuspension buffer (50 mM HEPES, pH 7.5, 3 mM MgSO 4 , 1 mM EDTA, 17 g/ml PMSF, and 2 g/ml leupeptin and aprotinin) containing 7 M urea and allowed to incubate for 30 min at 4°C. Resuspension buffer was used to dilute the membranes to 4 M urea prior to centrifugation at 142,000 ϫ g for 30 min at 4°C. Membranes were washed twice with resuspension buffer, and homogenization buffer was used to resuspend the membranes, which were stored in aliquots at Ϫ80°C. Membranes containing the A2a adenosine receptor were prepared using essentially the same method, except that homogenization buffer consisting of 25 mM HEPES, pH 7.5, 100 mM NaCl, 1% (w/v) glycerol, 17 g/ml PMSF, and 2 g/ml leupeptin, and aprotinin was used throughout the preparation. Radioligand binding experiments with [ 125 I]iodocyanopindolol and 125 I-ZM-241385 were used to determine receptor number and affinity for the ␤ 1 -adrenergic and A2a adenosine receptors, respectively. Stripping membranes with urea did not greatly affect the pharmacological properties of these two receptors (data not shown). The GTP␥S binding experiments presented under "Results" were obtained using a single preparation of membranes expressing either the ␤ 1 -adrenergic or the A2a adenosine receptor. Membranes expressing AC1 or AC2 were prepared as described previously (28). Total membrane protein concentration was determined by BCA assay using bovine serum albumin as a standard, and aliquots of membranes were stored at Ϫ80°C.
Measurement of Agonist-stimulated GTP␥S Binding to G s ␣ after Reconstitution with ␤␥ into Membranes Expressing either the ␤ 1 -Adrenergic Receptor or the A2a Adenosine Receptor-Kinetic parameters of agonist-stimulated binding of [ 35 S]GTP␥S to G s ␣ in the presence of different concentrations of ␤ 1 ␥ 2 were established with time course experiments. Aliquots of Sf9 cell membranes containing the ␤ 1 -adrenergic receptor were pelleted by centrifugation and resuspended in 100 -400 l of GTP␥S binding buffer (25 mM HEPES, pH 7.4, 100 mM NaCl, 1 mM EDTA, 5 mM MgCl 2 , 1 mM DTT, 0.1% BSA, 0.5 M GDP, and 1 M AMP-PNP) with a 28-gauge needle. The membrane suspension was reconstituted with 5 nM G s ␣ subunit such that the G s ␣:receptor ratio was 26:1; different concentrations of ␤ 1 ␥ 2 were then added and allowed to incubate for 30 min at 4°C. The incubation temperature was increased to 25°C for 10 min to equilibrate the reconstituted system to the reaction temperature; additions of [ 35 S]GTP␥S (final concentration 10 nM) and isoproterenol (final concentration 1 mM) initiated the time course. The binding of [ 35 S]GTP␥S to receptor-activated G s ␣ was measured at 1-min intervals by vacuum filtration. Increasing concentrations of ␤␥ increased the rate of receptor-catalyzed exchange of GDP for GTP␥S on G s ␣. The observed rates were relatively linear (see Fig. 2A), thus the effect of ␤␥ was quantified by the amount of [ 35 S]GTP␥S binding measured at a reaction time of 7 min. Seven minutes were chosen as a compromise that allowed a measurable signal before the reaction rate slowed.
By using this protocol, 10 different ␤␥ concentrations, ranging from 0.08 to 20 nM ␤ x ␥ 2 , were examined to determine the efficiency of coupling G s ␣ to receptor. Membranes were reconstituted with 5 nM G s ␣ and the indicated ␤␥ concentrations in a total volume of 30 l per tube. The concentrations of ␤␥ were prepared by serial dilution with GTP␥S binding buffer containing 0.1% CHAPS; 2 l of each ␤␥ concentration was diluted to the final incubation volume of 40 l, giving a CHAPS concentration of 0.005% for all but the highest concentrations of ␤␥. After the incubation protocols described above, 8 l of buffer containing [ 35 S]GTP␥S (ϳ1,000,000 dpm) and isoproterenol were added to each tube to start the 7-min reaction. The reaction was terminated as described above, and efficiency of coupling was determined by plotting [ 35 S]GTP␥S binding as a function of ␤␥ concentration (Fig. 2B). Receptor specificity for G s ␣ was demonstrated using the same protocol by reconstitution of G i1 ␣ with the ␤ 1 -adrenergic receptor (Fig. 2B).
A slight modification of this protocol was used to obtain dose response experiments with ␤␥ and the A2a adenosine receptor. The G s ␣:receptor ratio was 1.3:1, and in order to break down endogenous adenosine that is continuously generated in membrane preparations, adenosine deaminase was added to the membrane suspension before the 30-min incubation at a concentration of 14 units of activity/ml. The A2a adenosine receptor was activated with 100 nM 5Ј-N-ethylcarboximide adenosine.
Measurement of Adenylyl Cyclase Activity-In addition to activated G s ␣, the diterpene forskolin can stimulate all isoforms of adenylyl cyclase, whereas Ca 2ϩ /calmodulin can stimulate AC1, AC3, and AC8 (3). G s ␣ and forskolin were natural choices for this study, as they both activated AC1 and AC2, whereas Ca 2ϩ /calmodulin will not activate AC2. Forskolin was used to activate successfully AC1, but inhibition of the activity by ␤␥ was not as robust as in the case of the G s ␣-activated AC1 (data not shown); G s ␣ was therefore chosen as the activator for both AC1 and AC2. GTP␥S-activated G s ␣ was prepared by incubation of protein with gel filtration buffer (50 mM HEPES, pH 8.0, 150 mM NaCl, 5 mM MgCl 2 , 1 mM EDTA, 1 mM DTT, and 0.1% CHAPS) with the addition of 5 mM MgCl 2 (10 mM total) and 10 M GTP␥S for 30 min at 30°C. Unbound GTP␥S was removed by centrifugation with a 2-ml P6 desalting column equilibrated with gel filtration buffer using the method described in Yasuda et al. (39). Reconstitutions of G s ␣ and ␤␥ were performed as described by Lindorfer et al. (16). Maximal activity of AC1 and AC2 was confirmed by stimulation with increasing concentrations of GTP␥S-activated G s ␣. Cyclic AMP production was measured using a radioimmunoassay (40). Enzymatic activity reached a maximal rate of ϳ5 nmol of cAMP/min/mg protein for both AC1 and AC2, which is consistent with previous preparations from this laboratory (16), and with published data (41). For experiments involving the effect of different ␤␥ isoforms on adenylyl cyclase, activated G s ␣ is required at concentrations below what is necessary for maximal stimulation of ACII; therefore, 10 nM activated G s ␣ was chosen as the concentration suitable for co-activation of AC2. Complete activation of AC1 is required to observe inhibition by ␤␥; for this reason, 50 nM G s ␣ was used for activation of AC1. Adenylyl cyclase experiments utilized a single preparation of each membrane type.
Calculation and Expression of Results-In experiments using the ␤ 1 -adrenergic and A2a adenosine receptors, data from at least three experiments and two different sets of ␤␥ dimers were normalized as a percentage of maximal GTP␥S binding as determined by the one-site binding curves generated by GraphPad Prism. After normalization, the data were averaged for each ␤␥ isoform, and GraphPad Prism was used to obtain estimates of the EC 50 values and statistical analysis of the binding curves. These data are presented in Table I.
GraphPad Prism was used to estimate EC 50 and V max values for the potentiation of G s ␣-stimulated AC2 activity by ␤␥. At least three experiments using data from two different sets of ␤␥ were analyzed, and average values were reported in Table I. For AC1, GraphPad Prism was used to generate inhibition curves for each of the experiments with the different ␤␥ dimers; the data were then normalized as percent inhibition of the estimated rate of cAMP production with 50 nM GTP␥Sactivated G s ␣ in the absence of ␤␥. Normalized data from at least three experiments and two different sets of ␤␥ dimers were averaged and analyzed by GraphPad Prism to obtain IC 50 estimates (Table I). Statistical significance for differences among binding curves for both receptors and AC1 and AC2 was determined using the F-statistic; this technique is able to discern small but significant differences between two binding curves (42).
Materials-Reagents for Sf9 cell culture and purification of ␤␥ dimers has been described previously (16,25,26,34). 125 I-ZM-241385 was a kind gift from J. Linden, University of Virginia; baculovirus transfer vector was from Invitrogen; the BaculoGold kit was from PharMingen; DNase, GDP, imidazole, isoproterenol, and HEPES were from Sigma; adenosine deaminase, CHAPS, and GTP␥S were from Roche Molecular Biochemicals; P-6 desalting gel was from Bio-Rad; 10% Genapol C-100 was from Calbiochem; Ni 2ϩ -NTA Superflow resin was from Qiagen; [ 35 S]GTP␥S and [ 125 I]iodocyanopindolol from PerkinElmer Life Sciences; Source 15Q anion exchange resin was from Amersham Pharmacia Biotech; type HA 0.45-m nitrocellulose filters and Centricon 30 concentrators were from Millipore. All other materials were of the highest available purity. Fig. 1A presents a silver-stained gel showing the purity of the five ␤␥ dimers used in this study. Fig. 1B presents a similar gel showing the purity of the G s ␣ used. Significantly, both the ␤␥ dimers and the G s ␣ subunit were purified using biological affinity columns. The dimers containing ␤ 1-4 were purified with a G i1 ␣-agarose column ensuring that the proteins bound to ␣ subunits with high affinity and that the C terminus of the ␥ subunit was properly modified (see "Experimental Procedures"). Even so, a ␤ doublet was occasionally observed in the SDS-PAGE analysis of ␤ 3 ␥ 2 . The reasons for this behavior are not understood. The ␤ 5 ␥ 2HF dimer was purified from the membrane fraction of Sf9 cells using Ni 2ϩ /FLAG chromatography. Mass spectrometry was used to demonstrate that the ␥ subunit in this dimer was modified to the same extent as dimers purified on the G i1 ␣-agarose column, and the biological activity of ␤ 5 ␥ 2HF was vetted by determining its ability to activate PLC-␤ (10,16) to the same extent as ␤ 1 ␥ 2 (data not shown). G s ␣ was purified by AlF 4 Ϫ elution from a Gs ␣:␤ 1HF ␥ 2HF heterotrimer bound to a Ni 2ϩ -NTA-agarose column. The activity of the G s ␣ preparation was verified by its ability to activate AC1 and AC2. Fig. 1, C and D, shows that the ␣ subunit activated these cyclases 4 -8-fold, with EC 50 values of 1.9 nM for AC2 and 8.5 nM for AC1. These results are similar to other values reported in the literature for G s ␣ expressed in Sf9 cells (43), validating the quality of both the purified G s ␣ and the membrane preparations of AC1 and AC2.

Preparation of G Protein ␣ and ␤␥ Subunits-
The Ability of Different ␤␥ Subunits to Support Coupling of Receptors to G s ␣-A major goal of this study was to examine the possibility that different ␤ subunits interact selectively with certain G protein-coupled receptors. Since exchange of GDP for GTP is the first step of G protein signaling subsequent to receptor activation, an agonist-dependent GTP␥S-binding assay was used. Fig. 2A presents an experiment performed with membranes expressing the ␤ 1 -adrenergic receptor reconstituted with G s ␣ and two concentrations of the ␤␥ dimer. Note that the rate of isoproterenol-stimulated GTP␥S binding is nearly linear and highly dependent on the concentration of ␤␥ dimer. The triangles in Fig. 2A represent a basal rate of GTP␥S binding to membranes reconstituted with G s ␣; this rate was observed in the absence of ␤␥ (as illustrated in the figure) or with a fully reconstituted system in the absence of isoproterenol. Coupling of receptor to G protein is a composite of many biochemical interactions, the most important being the interactions of ␣-␤␥ and receptor-␣-␤␥. Receptor-␤␥ interactions were probed with a variation of the protocol designed to be poised on the concentration of the ␤␥ dimer. To define precisely the ability of ␤ 1 ␥ 2 to support coupling of G s ␣ to the ␤ 1 -adre- FIG. 1. Purity of G protein ␣ and ␤ subunits. A, the five isoforms of the ␤ subunit were overexpressed in Sf9 insect cells with the ␥ 2 or ␥ 2FH subunit and purified by G i1 ␣ affinity chromatography (␤ 1-4 ␥ 2 ) or Ni 2ϩ -NTA affinity chromatography (␤ 5 ␥ 2FH ). B, G s ␣ was overexpressed in Sf9 insect cells with a ␤ 1 ␥ 2 dimer containing a hexahistidine tag. The heterotrimer was adsorbed to a Ni 2ϩ -NTA column, and G s ␣ was eluted specifically with AlF 4 Ϫ . Purity of ␤␥ dimers (250 ng of each isoform) and Gs ␣ (150 ng) was visualized by silver staining after separation by 12% SDS-PAGE; positions of molecular weight markers are indicated at the right. C, Sf9 cell membranes expressing AC2 were incubated with increasing concentrations of G s ␣ activated with GTP␥S, and cAMP levels were determined using a radioimmunoassay; the calculated EC 50 for the experiment shown is 1.9 nM. D, Sf9 cell membranes expressing AC1 were characterized as in C; the calculated EC 50 for the experiment shown is 8.5 nM.
FIG. 2. Concentration dependence of ␤ 1 ␥ 2 on coupling G s ␣ to the ␤ 1 -adrenergic receptor. A, time course of [ 35 S]GTP␥S binding to Sf9 cell membranes containing the ␤ 1 -adrenergic receptor and reconstituted with 5 nM G s ␣ and 0, 0.5, or 10 nM ␤ 1 ␥ 2 . Ten nM [ 35 S]GTP␥S and 1 mM isoproterenol (Iso) were added at time 0, and bound nucleotide was determined at 1-min intervals by vacuum filtration. B, increasing concentrations of ␤ 1 ␥ 2 were reconstituted with G s ␣ or G i1 ␣ into membranes containing the ␤ 1 -adrenergic receptor, and the reaction was initiated as described in C; the amount of [ 35 S]GTP␥S bound was determined at the 7-min time point. The EC 50 for ␤␥ supported coupling of G s ␣ to receptor was 0.7 nM as determined by fitting the data to a one-site model. nergic receptor, a reaction time of 7 min was chosen to generate concentration-response curves with the different ␤␥ isoforms. A representative concentration-response curve performed with ␤ 1 ␥ 2 is presented in Fig. 2B; the apparent EC 50 estimated from this curve is 0.7 nM. The fidelity of receptor-␣ interactions was confirmed by the reconstitution of the ␤ 1 -adrenergic receptor with G i1 ␣, which demonstrates that even high concentrations of ␤ 1 ␥ 2 (20 nM) do not support isoproterenol-dependent GTP␥S binding (Fig. 2B) to the G i1 ␣ subunit. In contrast, when the same G i1 ␣ was reconstituted with ␤ 1 ␥ 2 into membranes containing the A1 adenosine receptor, robust agonist-dependent GTP␥S binding was observed (data not shown, but see Ref. 39).
The experiment performed with G s ␣ in Fig. 2B was repeated using each of the five ␤␥ isoforms, and the normalized data are shown in Fig. 3A. Highest in coupling efficiency was ␤ 4 ␥ 2 , with an EC 50 of 0.5 nM; ␤ 2 ␥ 2 was considerably less efficient with an EC 50 of 2.7 nM. There were slight but statistically significant differences in most of the ␤␥ isoforms, with only ␤ 1 ␥ 2 and ␤ 3 ␥ 2 showing no differences and an EC 50 of 1.0 nM. Poorest of all at coupling G s ␣ to the ␤ 1 -adrenergic receptor was ␤ 5 ␥ 2 with an EC 50 of 17.1 nM (see Table I).
To determine if the rank order of affinities determined with the ␤ 1 -adrenergic receptor was the same with another G slinked receptor, we examined the ability of the panel of ␤␥ dimers to support coupling of G s ␣ to the A2a adenosine receptor. Normalized data for each of the ␤␥ isoforms and the A2a adenosine receptor are shown in Fig. 3B; the coupling efficiencies and statistical analysis are presented in Table I. Highest in affinity was ␤ 4 ␥ 2 with an EC 50 of 1.3 nM; ␤ 2 ␥ 2 and ␤ 3 ␥ 2 were lower in affinity, both with EC 50 values around 6 nM. Strikingly, ␤ 1 ␥ 2 was much less efficient at coupling to the A2a adenosine receptor, with an EC 50 of 15.7 nM. Lowest in coupling efficiency was ␤ 5 ␥ 2 , with an EC 50 Ͼ100 nM. Fig. 3, C-G, illustrates that there are striking differences in the ability of the two G s -linked receptors to couple to the five different ␤␥ isoforms. Perhaps the most dramatic differences occurred with the ␤ 1 ␥ 2 isoform, which coupled 15-fold more efficiently to the ␤ 1 -adrenergic receptor than to the A2a adenosine receptor (Fig. 3C). Similarly, ␤ 3 ␥ 2 demonstrated a 7-fold difference between the two receptors (Fig. 3E). Note that ␤ 4 ␥ 2 was the most effective at coupling G s ␣ to either receptor (Fig.  3F), and that ␤ 5 ␥ 2 coupled poorly (Fig. 3G). In contrast, there are minimal differences in the ability of ␤ 2 ␥ 2 to couple to either receptor (Fig. 3D). It is important to stress that the only differences in these five sets of experiments are the types of recombinant receptor expressed in the Sf9 membranes. The G protein ␣ and ␤␥ subunits reconstituted into the membranes were identical in each case. Thus, the data clearly demonstrate that specific G protein ␤ subunits exhibit distinct preferences for different receptors. Importantly, these preferences are a result of interactions of receptor with the type of ␤ subunit in the dimer, since the G s ␣-␤␥ interactions are presumably identical for both receptors.
Activation of AC2 by ␤␥ Isoforms-The dramatic differences in the ability of the panel of ␤␥ dimers to couple to G s -linked receptors imply that different ␤␥ dimers might be released by receptor activation to act on downstream effectors. Since ␤␥ is a known potentiator of G s ␣-stimulated AC2 activity, and differences have been observed in the ability of dimers containing the ␤ 1 or ␤ 5 subunits to stimulate AC2 (9,16), the ability of all five ␤ subunits to activate AC2 was compared. The role of ␤␥ in the activation of AC2 is particularly interesting in that ␤␥ can increase the rate of cAMP production approximately 5-fold over the maximal effect of G s ␣ (Fig. 4), suggesting ␤␥ can regulate cAMP levels in vivo. Ten nM GTP␥S-activated G s ␣ and increasing concentrations of the five purified ␤␥ isoforms were reconstituted with Sf9 membranes expressing AC2, and cAMP production was measured. A representative experiment is presented in Fig. 4. Dimers containing ␤ 1-4 were similar in their ability to potentiate AC2 activity in the presence of acti- FIG. 3. Comparison of ability of different ␤␥ isoforms to couple G s ␣ to the ␤ 1 -adrenergic and A2a adenosine receptors. A, five ␤ x ␥ 2 isoforms were reconstituted with 5 nM G s ␣ and membranes containing the ␤ 1 -adrenergic receptor and the efficiency of coupling measured as described under "Experimental Procedures." Data from three experiments were normalized as a percent of maximal binding of [ 35 S]GTP␥S, and the averaged data plotted; error bars, most of which were within Ϯ10%, were omitted for clarity. B, five ␤ x ␥ 2 isoforms were tested as in A, but with the A2a adenosine receptor. Data from at least three experiments were normalized and plotted as in A. EC 50 estimates of the data sets for each ␤␥ dimer can be found in Table I. C-G, data from A and B were replotted to highlight differences in each particular ␤ x ␥ 2 isoform between the ␤ 1adrenergic receptor (␤) and the A2a adenosine receptor (A2a). Dotted lines indicate ␤␥ concentrations of 1 and 10 nM on the x axis. C, vated G s ␣, with all dimers increasing cAMP production 4 -6fold (V max values ranging from 30 to 40 nmol of cAMP/min/mg of protein). Consistent with previous reports (16), ␤ 5 ␥ 2 was significantly less active. Data from at least three similar experiments were normalized and averaged to determine the EC 50 values for each ␤␥ dimer, and the results are summarized in Table I. The EC 50 values for ␤ 1-4 range from 3.5 to 13 nM, and the value for ␤ 5 ␥ 2 is significantly higher at 76 nM. Careful analysis of the data indicates that the ␤ 2 ␥ 2 dimer is 3-fold less potent than the ␤ 1 ␥ 2 and ␤ 4 ␥ 2 dimers. Also of interest is the observation that dimers containing ␤ 3 and ␤ 4 , which had not been previously tested on adenylyl cyclase, were as effective at stimulation of AC2 as dimers containing ␤ 1 and ␤ 2 . These results suggest that all four ␤ isoforms can effectively participate in G s ␣ signaling pathways affecting the regulation of cAMP via AC2.
Inhibition of AC1 by ␤␥ Isoforms-AC1 is in very high concentration in neuronal tissue and is notable among the adenylyl cyclase isoforms in that it can be markedly inhibited by the ␤␥ dimer. The activity of AC1 expressed in membranes was demonstrated with purified G s ␣, which stimulated cAMP production with an EC 50 of 8.5 nM (Fig. 1D). This activation of AC1 by G s ␣ provided the opportunity to compare the inhibitory properties of all five ␤␥ dimers. Increasing concentrations of purified ␤␥ isoforms were reconstituted with 50 nM GTP␥Sactivated G s ␣ into Sf9 membranes expressing AC1. A representative experiment illustrating the robust inhibition of AC1 elicited by ␤ 1 ␥ 2 is presented in Fig. 5A, where the dimer reduced cAMP production by over 50%. Similar experiments were performed with the full panel of ␤␥ dimers. The data were normalized and are presented in Fig. 5B. Dimers containing the ␤ 1 , ␤ 2 , and ␤ 4 subunits were not different in their ability to inhibit AC1, with IC 50 values ranging from 10 to 17 nM. In contrast, the ␤ 3 ␥ 2 dimer had an almost 10-fold higher IC 50 of 110 nM, and surprisingly, no inhibition was observed with ␤ 5 ␥ 2 at the concentrations tested ( Fig. 5B and Table I). These data suggest that ␤␥ dimers composed of different ␤ isoforms, especially those containing ␤ 3 and ␤ 5 , may differentially regulate the type I and type II adenylyl cyclase isoforms. DISCUSSION Although considerable research has focused on the role of the ␣ subunit in receptor signaling via G proteins, it is also clear that the ␤␥ dimer is required for coupling the ␣ subunit to receptors (16,39,44,45), and both the ␤ and ␥ subunits appear to contribute to the interaction. Current evidence indicates that the C-terminal 10 amino acids of the ␥ subunit and the nature of the prenyl group (farnesyl versus geranylgeranyl) are very important determinants of the coupling of G proteins to receptors (20,21,39). Compared with ␣ and ␥, less is known about the important domains in the ␤ subunit, but cross-linking experiments indicate that the C terminus of the ␤ subunit is able to interact directly with the receptor (46). However, the regions close to the C terminus of the ␤ subunit are likely to be quite discrete because mutations in amino acids His 311 , Arg 314 , and Trp 332 have no effect on the ability of the dimers to support receptor coupling but cause a major disruption in the ability of the ␤␥ dimer to activate PLC-␤ or AC2 (47).
Whereas the diversity of the ␤ and ␥ subunits offers attractive possibilities for determining the specificity of cellular signaling, the functional significance of this heterogeneity has not been completely elucidated. Some insight has come from an elegant set of experiments in which antisense mRNAs to various G protein ␣ and ␤␥ subunits were injected into the nuclei of GH3 cells, leading to the observation that specific receptors couple to distinct isoforms of the G protein heterotrimer. These experiments indicated that the M 4 muscarinic receptor preferred a heterotrimer composed of ␣ O1 :␤ 3 ␥ 4 , whereas the soma-  Table I. tostatin receptor preferred a heterotrimer composed of ␣ O2 :␤ 1 ␥ 3 (17)(18)(19). These results have been extended to other systems using the antisense approach (48), where it has been demonstrated that selectivity of receptor coupling to ␣ subunits of the G i (49), G q (50), and G 12/13 (51) families depends upon the composition of the ␤␥ dimer. However, few experiments have examined the composition of ␤␥ dimers that couple receptors to G s ␣. Moreover, reproduction of this signaling specificity using in vitro systems has proven difficult.
A major contribution of the experiments presented in this report is that a panel of highly active recombinant ␤␥ dimers was used in a sensitive assay to compare the ability of five ␤ subunits to couple to two different G s -linked receptors. The ability of a defined heterotrimer to support coupling of an ␣ subunit to a particular receptor depends primarily on the affinity of the receptor for the heterotrimer (48), and presumably on the affinity of the interaction between the ␣ and ␤␥ subunits (16,26). Since the order of coupling efficiencies in the present experiments was quite different between the ␤ 1 -adrenergic and the A2a adenosine receptors, the possibility that the differences observed here were due to differences in the ␣:␤␥ affinity seems unlikely. Thus, the order of EC 50 values for each receptor is most likely a reflection of its preference for the particular ␤ subunit examined. The ␤ 5 ␥ 2 dimer, which couples M 1 mus-carinic receptors efficiently to the G q ␣ subunit (16), was particularly ineffective at coupling the G s ␣ subunit to either receptor. In contrast, the ␤ 4 ␥ 2 dimer was consistently highest in coupling efficiency for either receptor. Even more surprising was the finding that the ␤ 1 ␥ 2 dimer, which in most assays of ␤␥ function is potent and efficacious, was very poor at coupling G s ␣ to the A2a adenosine receptor. These results lead to the following conclusions: 1) receptors can discriminate among the G protein heterotrimers based on the ␤ isoform alone; 2) all ␤ 1-4 isoforms can function in signaling cascades involving G s ␣; and 3) dimers containing ␤ 5 are not likely to be released from G s -coupled receptors. Finally, since different ␤␥ dimers may be released upon receptor activation, the downstream effects of the distinct ␤ isoforms may have different signaling properties.
One of the immediate downstream targets for ␤␥ is the family of receptor kinases that phosphorylate G proteincoupled receptors upon recruitment to the membrane by the dimer (52). Experiments have examined the ability of defined ␤␥ dimers to interact with the receptor kinases. For example, dimers containing ␤ 1 and ␤ 2 interact with GRK2, the kinase responsible for down-regulation of the ␤-adrenergic and A2a adenosine receptors (53) far better than dimers containing ␤ 3 (54). Another study examined the ability of a variety of defined ␤␥ dimers to promote the phosphorylation of both the ␤ 2adrenergic receptor and rhodopsin by GRK2. These results indicate a significant difference in the ability of the various ␤␥ dimers to promote phosphorylation of the ␤ 2 -adrenergic receptor or rhodopsin and suggest that the type of ␤ subunit could determine selectivity between the two receptors (55). Thus, even though ␤ 1-4 are over 80% identical, the accumulating evidence suggests the type of ␤ subunit in the dimer may have a larger role in determining signaling specificity than previously appreciated.
Another immediate downstream target for the ␤␥ subunit is the effector adenylyl cyclase (3). Results presented here demonstrate that dimers containing ␤ 1 , ␤ 2 , or ␤ 4 were able to regulate either type of adenylyl cyclase effectively. In contrast, ␤ 5 ␥ 2 was not particularly effective at inhibiting AC1 or in stimulating AC2 (16). Intriguingly, ␤ 3 ␥ 2 was almost 10-fold weaker at inhibiting AC1 as compared with the ␤ 1-4 isoforms (Table I), whereas it was equally effective on AC2. These results suggest that upon stimulation of certain G s -linked receptors, co-activation of AC2 by ␤␥ is relatively nonspecific, whereas inhibition of AC1 by ␤␥ is more selective and may be receptor-dependent. This specificity of interaction between AC1 and the different ␤ isoforms suggests that dimers containing the ␤ 3 subunit have signaling roles distinct from those containing ␤ 1 , ␤ 2 , or ␤ 4 .
The regions of the ␤␥ dimer that are thought to interact with AC1 and AC2 have been examined using competition experiments with synthetic peptides and alanine mutagenesis. Peptides identical to residues 86 -105 and 115-135 of the ␤ 1 subunit were able to inhibit stimulation of AC2 by the ␤␥ dimer, implicating these residues of the ␤ subunit (56), as well as others (57), as sites on the molecule that interact with AC2. Moreover, the QEHA peptide, which represents a sequence from AC2 thought to interact with the ␤␥ dimer, was able to bind directly to the ␤ subunit. Molecular modeling of the QHEA peptide-␤␥ interaction also identified the region of ␤ defined by residues 75-165 as a potentially important effector-binding domain (58). Mutagenesis experiments have suggested that three residues in the ␤ subunit involved in ␣:␤ interface, Asp 228 , Asp 246 , and Trp 332 , are important for the activation of AC2 but have no effect on the ability of the dimer to inhibit AC1 (59). Studies of the outer surface of the ␤ torus show that Asn 132 in blade two of the protein is important for inhibition of  Table I. AC1, but seems to have little effect on activation of AC2 (60). The observation that ADP-ribosylation of Arg 129 of ␤ 1 , a residue also present in ␤ 2-4 , prevents the inhibition of AC1 by the ␤␥ dimer supports the argument that this is an important domain in the interaction of the dimer with AC1 (61).
Examination of the regions identified by the experiments discussed above in the ␤ 1-4 subunits shows minimal differences in the amino acid sequence. This is consistent with the observation that these four ␤␥ dimers activated AC2 equally. A similar conclusion applies to the ability of three of the dimers to inhibit AC1; the intriguing exception was ␤ 3 ␥ 2 , which was far less effective. Unfortunately, there are no obvious differences in the amino acid sequence of the ␤ 3 subunit in these regions to explain the differences in activity, suggesting that some other region in the molecule is also involved in the interaction with AC1. There are, however, sequence variations that could explain the lack of activity of ␤ 5 ␥ 2 on either isoform of cyclase. In contrast to the near identity of the ␤ 1-4 subunits in the regions outlined above that are thought to interact with AC2 and AC1, the ␤ 5 subunit has 13 amino acid differences in these regions when compared with the ␤ 1 subunit. Moreover, there is a two-amino acid insert in the region between residues 130 -132, a site identified by ADP-ribosylation experiments as being important for the inhibition of AC1 (61). Although not definitive, these differences provide a reasonable starting point for future experiments.
Despite the homology in amino acid sequence among these ␤ isoforms, differences in localization have been observed. For example, in contrast to the other ␤ isoforms that were localized to the membrane, ␤ 3 was observed predominantly in cytoplasmic fractions in both heart (62) and retina (63). This property of ␤ 3 does not seem likely to be responsible for the observed differences in inhibition of AC1, since ␤ 3 ␥ 2 was just as effective as the other ␤␥ dimers at activating the membrane-bound protein AC2 and supporting coupling to the ␤ 1 -adrenergic and A2a adenosine receptors. A more reasonable explanation is that certain residues unique to ␤ 3 impart specificity either through altering the contacts with an effector molecule such as AC1, or those unique residues slightly influence the conformation of the ␤ subunit, thereby altering interactions with other proteins. Whatever the reason, one important conclusion is that differences in AC1 inhibition may result from the release of different ␤␥ dimers by receptor activation.
This concept appears to apply especially well to dimers containing the ␤ 5 subunit. Even though the current data suggest that the ␤ 5 ␥ 2 dimer is unlikely to be released from G s -linked receptors, it clearly can be released by activation of G q -linked receptors (16). However, accumulating evidence shows that the ␤ 5 ␥ 2 dimer does not regulate a variety of effectors, including AC1, AC2, phosphatidylinositol 3-kinase, PLC-␤3 and the mitogen-activated protein kinase pathway (this report and Refs. 64 and 65). Even though the ␤ 5 ␥ 2 dimer did not regulate adenylyl cyclase in our experiments, transfection of the dimer into COS-7 cells caused an inhibition of both AC1 and AC2 (66). These conflicting data may be explained by other potential partners for ␤ 5 , such as RGS 6,7,9,and 11 (67,68), which may impinge upon the adenylyl cyclase cascade in vivo. This evidence of multiple partners for the ␤ 5 subunit suggests the ␤ 5 protein may have functions not normally associated with ␤ subunits and makes the physiological role of ␤ 5 on effectors unclear. Especially interesting is the possibility that the ␤ 5 subunit may exist as a monomer and exchange between ␥ subunits and RGS proteins (69). Although the role of ␤ 5 in signaling is clearly complex, one conclusion from this information is that receptors that couple to and release dimers containing the ␤ 5 subunit are less likely to generate cross-talk between signaling systems because of the limited effect of ␤ 5 containing dimers on downstream effectors.
Brain is one tissue where all of the signaling components studied in this report are expressed at high levels (70 -73). Thus, the differential effects of the ␤ isoforms demonstrated in this report could have major effects on signaling cascades in the brain. Some experimental support for this concept comes from experiments showing that small amounts of ␤␥ derived from G s activation can inhibit the neuronal-specific AC1 in vivo (74). Further information needed to corroborate these proposed differences in ␤␥ signaling includes cellular and subcellular localization of these molecular components. Once the subcellular architecture in these tissues is better understood with respect to G proteins, signaling models based on specific G protein ␤ subunits can be refined with precision.