The α2A-Adrenergic Receptor Discriminates between Gi Heterotrimers of Different βγ Subunit Composition in Sf9 Insect Cell Membranes*

In view of the expanding roles of the βγ subunits of the G proteins in signaling, the possibility was raised that the rich diversity of βγ subunit combinations might contribute to the specificity of signaling at the level of the receptor. To test this possibility, Sf9 cell membranes expressing the recombinant α2A-adrenergic receptor were used to assess the contribution of the βγ subunit composition. Reconstituted coupling between the receptor and heterotrimeric Gi protein was assayed by high affinity, guanine nucleotide-sensitive binding of the α2-adrenergic agonist, [3H]UK-14,304. Supporting this hypothesis, the present study showed clear differences in the abilities of the various βγ dimers, including those containing the β3 subtype and the newly described γ4, γ10, and γ11 subtypes, to promote interaction of the same αi subunit with the α2A-adrenergic receptor.

Consistent with the steadily increasing number of G protein 1 ␤ and ␥ subtypes that has been revealed in recent years (1), in vivo studies have indicated a role for this structural diversity in the specificity of signaling. In this regard, antisense studies by Kleuss et al. (2,3) have demonstrated a specific requirement for the ␤ 1 and ␥ 3 subunits in the somatostatin receptor signaling pathway in rat pituitary GH 3 cells, with a similarly specific requirement for the ␤ 3 and ␥ 4 subunits in the muscarinic receptor signaling pathway. Also, a ribozyme study by Wang et al. (4) has shown a specific involvement of the ␥ 7 subunit in the ␤-adrenergic receptor signaling pathway in human kidney 293 cells. Taken together, these in vivo studies indicate that the composition of the ␤␥ dimer has important ramifications for the fidelity of signaling that is probably manifested at the level of the receptor.
A growing body of in vitro evidence supports a direct interaction between the receptor and the ␤␥ dimer (5). In particular, direct interaction of transducin ␤␥ with rhodopsin has been shown with a fluorescence energy transfer technique (6). This association was blocked by a synthetic peptide derived from the carboxyl-terminal tail of rhodopsin, suggesting a site of direct contact between ␤␥ and rhodopsin. Moreover, cross-linking studies have confirmed a receptor contact site on the ␤ subunit. A synthetic peptide derived from the carboxyl-terminal portion of the putative third cytoplasmic loop of the ␣ 2A -AR could be cross-linked to the carboxyl-terminal region of the ␤ subunit (7).
To date, in vitro studies examining the contribution of a limited number of ␤␥ dimers to the specificity of receptor coupling have not yielded the same high degree of discrimination shown in the in vivo studies cited above (8,9). The present study extended this analysis to the ␣ 2A -adrenergic receptor and to ␤␥ dimers that represent the most extensive degree of structural diversity examined to date. Since baculovirus expression has been shown to be an effective means for producing functional G protein subunits (10 -13) as well as G protein-coupled receptors (8,9,14,15), this system was used to measure the level of interaction between the recombinant ␣ 2A -adrenergic receptor expressed in Sf9 cell membranes and reconstituted in the presence or absence of purified G i proteins of varying ␤ or ␥ subtype composition. Among the two ␤ subtypes or eight ␥ subtypes tested, 30-fold differences were observed in their relative abilities to support coupling of the same ␣ subunit to the recombinant ␣ 2A -adrenergic receptor. These data demonstrate that the specificity of ␣ 2 -adrenergic receptor-G protein interactions is affected by the ␤␥ dimer composition.

EXPERIMENTAL PROCEDURES
Expression of ␣ 2A -Adrenergic Receptor-A pVL1392 transfer vector containing ␣ 2A -adrenergic receptor cDNA was generously provided by Dr. H. Kurose and R. Lefkowitz (Duke University, Durham, NC). Recombinant baculovirus encoding the ␣ 2A -adrenergic receptor was generated by co-transfection of ␣ 2A -pVL1392 with a linearized lethal deletion mutant of Autographa californica as directed by the supplier (BaculoGold, PharMingen Corp.). Expression by recombinant baculovirus was identified by specific binding of the ␣ 2 -adrenergic radioligand, [ 3 H]yohimbine (described below). A positive recombinant was isolated through four rounds of plaque purification. Receptors were expressed by inoculating Sf9 insect cells at an m.o.i. of 1 in IPL-41 medium, 1ϫ lipid concentrate, and 1% heat-inactivated fetal bovine serum (Life Technologies, Inc.) at a density of 2 ϫ 10 6 cells/ml. After 72 h, cell pellets were lysed by nitrogen cavitation (500 pounds/square inch for 30 min at 4°C) in 100 ml of ice-cold lysis buffer (25 mM Tris, pH 7.4, 1 mM EDTA, 10 mM MgCl 2 , 100 mM NaCl, 0.02 mg/ml phenylmethylsulfonyl fluoride, 0.03 mg/ml leupeptin, and 1 mM benzamidine) and centrifuged at 4°C for 10 min at 600 ϫ g. The supernatant was centrifuged at 40,000 ϫ g for 40 min at 4°C. The pellets were resuspended, washed once in lysis buffer (40,000 ϫ g, 40 min), and resuspended in 10 ml of lysis buffer. Protein concentration was determined by Coomassie assay (Pierce). Particulate fraction protein was snap-frozen with liquid N 2 in aliquots of 300 g each and stored at Ϫ80°C. Receptor expression was quantitated by saturation binding of [ 3 H]yohimbine, as described under "Experimental Procedures." A single 500-ml expression culture yielded adequate material to carry out all of the reconstitution experiments.
Expression and Purification of G Protein Subunits-Recombinant baculoviruses directing the expression of ␤ 1 , ␥ 1 , ␥ 2 , ␥ 3 , ␥ 5 , and ␥ 7 recombinant baculovirus were described previously (16,17). Isolation of human cDNAs encoding ␤ 3 (18) and ␥ 4 , ␥ 10 , and ␥ 11 (19) was described previously. In these cases, recombinant baculoviruses were obtained by co-transfection of Sf9 insect cells with pVL1393 transfer vectors containing ␤ 3 , ␥ 4 , ␥ 10 , or ␥ 11 and a linearized lethal deletion mutant of A. californica nuclear polyhedrosis virus as directed by the supplier (Bacu-loGold, PharMingen). Recombinant viruses were identified by immunoblotting Sf9 cell lysates for expression of the appropriate subunits. Subtype-specific antibodies were generated as described previously (20) using the following synthetic peptides: ␥ 4 , CKEGMSNNSTTSIS (amino acids 2-14); ␥ 10 , CKDALLVGVPAGSNPFREPR (amino acids 45-63); and ␥ 11 , CPALHIEDLPEK (amino acids 2-12). Other subtype-specific antibodies have been described previously (20,21). Recombinant virus encoding G␣ i1 containing a hexahistidine tag at amino acid position 121 was kindly provided by Dr. T. Kozasa (Southwestern Medical Center, Dallas, TX). One-liter cultures of Sf9 insect cells in IPL-41 medium, 1% heat-inactivated fetal bovine serum, and 1ϫ lipid mix (Life Technologies, Inc.) were inoculated at a density of 2 ϫ 10 6 cells/ml simultaneously with recombinant baculoviruses encoding ␣, ␤, and ␥ subunits as follows: 6his ␣ i1 at m.o.i. ϭ 2, ␤ 1 or ␤ 3 at m.o.i. ϭ 3, and each of the ␥ subtypes at m.o.i. ϭ 3. Under this condition, those ␥ subtypes predicted to contain a C-20 geranylgeranyl group are appropriately modified. However, those few ␥ subtypes predicted to contain a C-15 farnesyl group are variably modified at high levels of protein expression (10). Therefore, to optimize the addition of a C-15 farnesyl moiety, cultures of Sf9 cells expressing the ␥ 1 and ␥ 11 subtypes were also infected with recombinant baculovirus encoding both subunits of the mammalian farnesyltransferase at m.o.i. ϭ 0.2. This virus was kindly provided by Dr. Thomas Kost (Glaxo Corp.). Cultures of Sf9 cells infected with the farnesyltransferase virus displayed greater than 15-fold higher activity toward the Ha-Ras fusion protein substrate than cultures not so infected. Moreover, cultures of Sf9 cells infected with the farnesyltransferase virus resulted in the majority of the ␥ 1 and ␥ 11 subtypes being modified with the C-15 farnesyl moiety, as shown previously (10). Expression of G protein ␤ and ␥ subunits in particulate fractions was confirmed 72 h later by immunoblotting with subtype-specific antibodies (22).
Recombinant ␤␥ heterodimers were purified to apparent homogeneity using the procedure described by Kozasa and Gilman (11) for purification of ␤ 1 ␥ 2 . Following cholate extraction of particulate fractions, the cholate-soluble protein was diluted to 0.2% sodium cholate with 20 mM Hepes, pH 8.0, 100 mM NaCl, 1 mM MgCl 2 , 10 mM ␤-mercaptoethanol, 10 M GDP, and 0.5% polyoxyethylene 10-lauryl ether. The cholate-soluble extract was loaded onto a 4-ml Ni-NTA resin bed at 3-4-bed volumes/h (4°C) and washed with 100 ml of the same buffer containing 300 mM NaCl and 5 mM imidazole. ␤␥ dimers were eluted from the column by activation of 6his ␣␤␥ with AMF (30 M AlCl 3 , 50 mM MgCl 2 and 10 mM NaF)-containing buffer: 20 mM Hepes, pH 8.0, 50 mM NaCl, 10 mM ␤-mercaptoethanol, 10 M GDP, 1% sodium cholate, 5 mM imidazole, 50 mM MgCl 2 , 10 mM NaF, and 30 M AlCl 3 . The peak ␤␥-containing fractions were identified by immunoblotting for both ␤ and ␥ subunits and then pooled and diluted to less than 10 mM NaCl using 20 mM Hepes, pH 8.0, 1 mM EDTA, 3 mM DTT, 3 mM MgCl 2 , and 0.7% CHAPS. ␤␥ subunits were further purified by fast protein liquid chromatography on a Mono Q column (Amersham Pharmacia Biotech, HR 5/5) eluted with a linear NaCl gradient from 0 to 400 mM. Peak fractions were again confirmed by immunoblotting. The elution peaks were pooled and dialyzed overnight (3 buffer changes) against 20 mM Hepes, pH 8.0, 1 mM EDTA, 3 mM DTT, 3 mM MgCl 2 , 100 mM NaCl, and 0.7% CHAPS (Spectra/Por tubing, 6000 -8000 molecular weight cut-off, Spectrum Medical Industries, Houston, TX). A mixture of ␤␥ subunits purified from bovine brain by a previously described method (23) was also further purified on a Mono Q column by the same procedure. Following dialysis, purified ␤␥ subunits were concentrated to approximately 0.1 mg/ml in an Amicon ultrafiltration device (PM10 membrane), and the final protein concentrations were determined by staining with Amido Black. Purified ␤␥ subunits were snap-frozen in small aliquots with liquid N 2 and stored at Ϫ80°C.
The 6his ␣ i1 subunit was expressed alone (m.o.i. ϭ 3) in a 1-liter culture of Sf9 insect cells for subsequent purification. Protein extraction, loading, and washing of the Ni-NTA column were identical to that described for ␤␥ subunits. The 6his ␣ i1 was eluted from the Ni-NTA column with the following buffer, 20 mM HEPES, pH 8.0, 100 mM NaCl, 10 mM ␤-mercaptoethanol, 1 mM MgCl 2 , 0.5% polyoxyethylene 10-lauryl ether, 10 M GDP, and 150 mM imidazole, and was subsequently purified further on a Mono Q column (fast protein liquid chromatography) using the same procedure described for ␤␥ subunits with the exception that collection tubes contained an aliquot of GDP to yield a final concentration of 10 M GDP in each of the column fractions. Subsequent handling of 6his ␣ i1 was identical to ␤␥ subunits.
Reconstitution of Receptor-G Protein Coupling-Purified ␣ and ␤␥ subunits were combined in 20 mM Hepes, pH 8.0, 1 mM EDTA, 3 mM DTT, 10 M GDP, 0.02% sodium cholate to allow formation of G protein heterotrimers of defined subtype composition. The mixture was incubated in a total volume of 30 l on ice prior to reconstituting the G proteins into Sf9 cell plasma membranes. An aliquot of the Sf9 membrane preparation expressing ␣ 2A -adrenergic receptor was thawed and diluted to approximately 0.5 mg of protein/ml in 25 mM Tris, pH 7.6, 1 mM EDTA, 10 mM MgCl 2 , 1 mM benzamidine, 1 g/ml pepstatin A, and 1 g/ml aprotinin; then G i heterotrimers were added to the membranes at the desired ratio of G protein to receptor and incubated on ice for 30 min prior to receptor binding assays. The CHAPS concentration during this incubation was Յ0.04%.
ADP-ribosylation Assay-To assess the relative affinities of the various ␤␥ dimers for the ␣ subunit, 330 ng of the various purified ␤ 1 ␥ dimers were combined with 500 ng of purified 6his ␣ i1 at a final ratio of 50:75 (mol:mol) of ␤␥:␣. At this ratio, differences in the relative affinities of certain ␤␥ dimers for the receptor were detected. As described previously (24), incubation of the resulting G i heterotrimers was carried out in the presence of 200 ng of islet activating protein and [ 32 P]NAD at 30°C for 20 min and then terminated by precipitation with 30% trichloroacetic acid followed by rapid filtration over BA85 nitrocellulose filters. ADP-ribosylation of 6his ␣ i1 was measured by scintillation counting to detect [ 32 P] bound to filters.
Radioligand Binding Assays-[ 3 H]UK-14,304 and [ 3 H]yohimbine binding incubations were carried out in a total volume of 250 l at 24°C for 60 min in a reaction buffer consisting of 25 mM Tris, pH 7.6, 1 mM EDTA, 10 mM MgCl 2 (plus 100 mM NaCl in assays of [ 3 H]yohimbine binding). Radioligand binding was initiated by addition of 2-10 g of Sf9 membrane protein and terminated by addition of 3 ml of the same ice-cold reaction buffer followed by rapid filtration over Whatman GF/C glass fiber filters on a Millipore filtration apparatus. Filters were rinsed twice more with the same buffer. Bound radioligand was quantitated by liquid scintillation counting (Beckman LS6500). Nonspecific binding was determined in the presence of 100 M yohimbine. Statistical analysis of coupling (high affinity [ 3 H]UK-14,304 binding) supported by different ␥ subtypes was done by a two-tailed, Student's t test for comparison of the variation between two means, Յ 0.05 indicated statistical significance.
Materials-[imidazoylyl-4,5-3 H]UK-14,304, [methyl-3 H]yohimbine, [ 32 P]nicotinamide adenine dinucleotide, and G protein ␤-common antiserum (SW/1) were obtained from NEN Life Science Products; Ni-NTAagarose was obtained from Qiagen (Chatsworth, CA); oxymetazoline was obtained from Research Biochemicals International (Natick, MA); ECL reagent and horseradish-linked anti-rabbit IgG were obtained from Amersham Pharmacia Biotech (Buckinghamshire, UK); and other chemicals were obtained from Sigma.  Fig. 1A). This K D range of the recombinant receptor was consistent with that of the native ␣ 2A -adrenergic receptor (for review, see Ref. 25). Moreover, this moderate B max value was optimal for the purpose of this study since this level of recombinant receptor expression was readily measurable but required minimal amounts of purified G proteins for reconstitution. Finally, the properties of the recombinant ␣ 2A -adrenergic receptor were characteristic of those of the native receptor. The specific [ 3 H]yohimbine binding showed a NaCl sensitivity that is typical of the native receptor (25), i.e. the B max binding plateau was 45-55% lower in the presence of 100 mM NaCl (data not shown). Prazosin and oxymetazoline displaced the specific [ 3 H]yohimbine binding from the recombinant receptor with the same order of potency of the native receptor (K D values of 4 and 0.022 M, respectively). Taken together, these data confirmed the suitability of Sf9 cells for the expression of recombinant ␣ 2A -adrenergic receptor that is functionally similar to the native receptor (26).

Reconstitution of ␣ 2A -Adrenergic Receptor-G Protein Coupling in Sf9
Cell Plasma Membranes-Coupling of the ␣ 2A -adrenergic receptor was examined in the presence and absence of exogenous G i protein using the agonist [ 3 H]UK-14304. The direct agonist binding technique is generally considered to be more sensitive than agonist displacement studies for detecting receptor-G protein complexes (25). When Sf9 cell membranes expressing the recombinant ␣ 2Aadrenergic receptor were incubated in the absence of exogenous G i protein, a low level of specific [ 3 H]UK-14304 binding was detected, accounting for ϳ15% of the binding that was later observed in the presence of added G i protein (Fig. 2). By contrast, when Sf9 cell membranes expressing the recombinant ␣ 2A -adrenergic receptor were incubated in the presence of exogenous G i protein (at a molar ratio of 100:1 G i :receptor), the level of specific [ 3 H]UK-14304 binding was increased by more than 5-fold, representing coupling of the recombinant receptor to the added G i protein (Fig. 2). Moreover, the increased level of [ 3 H]UK-14304 binding was reversed by the addition of GTP␥S, reflecting uncoupling of the recombinant receptor from the added G i protein. Thus, reconstituted coupling was easily distinguishable from the background coupling in this experimental system, thereby confirming the suitability of this experimental system for measuring the coupling of the recombinant receptor to added G i proteins of varying ␤␥ composition. For optimal resolution between the reconstituted and background coupling, a 4 nM concentration of [ 3 H]UK 14,304 was used in subsequent experiments.
Receptor to G Protein Stoichiometry-The requirements of the [ 3 H]UK-14,304 binding assay for the G protein ␣ and ␤␥ subunits were examined further. Consistent with previous studies of the A 1 adenosine receptor (9), the combined interaction of both the G protein ␣ and ␤␥ subunits was required in order to detect the high affinity state of the recombinant ␣ 2Aadrenergic receptor with this binding assay. As shown in following reconstitution with purified G i heterotrimer at a ratio of 50:1, G i to receptor; E, binding in the presence of 100 M GTP␥S following reconstitution with G i ; f, specific binding without added G i ; Ⅺ, binding in the presence of GTP␥S with no added G i . A curve was fit to the data by non-weighted, nonlinear regression analysis using a one-site hyperbola fit. G i heterotrimer (9.25 pmol) was composed of purified 6his ␣ i1 and purified bovine brain ␤␥ (3:2, mol ␣/mol ␤␥). Nonspecific binding (presence of 100 M yohimbine) was virtually identical to low affinity binding (presence of 100 M GTP␥S); thus essentially all the radioligand binding within this concentration range represents high affinity, receptor-G protein complexes. recombinant receptor. Next, the recombinant receptor was reconstituted with a constant amount of 6his ␣ i1 subunit and increasing amounts of ␤␥ dimer. As shown in Fig. 3B, raising the amount of ␤␥ dimer increased the fraction of receptor in the high affinity state as measured by the higher level of [ 3 H]UK-14,304 binding. When the amounts of ␤␥ dimer and 6his ␣ i1 subunit approached a 1:1 ratio, the level of [ 3 H]UK-14,304 binding reached a plateau, accounting for ϳ60% of the total receptor population as determined by [ 3 H]yohimbine binding. A similar, maximal level of coupling was observed previously for the A 1 adenosine receptor (9), suggesting that not all of the recombinant receptors are accessible for reconstitution with added G proteins. Under these conditions, any observed differences in the magnitude of [ 3 H]UK-14,304 binding can be assumed to be attributable to selective interactions of ␤␥ dimers of varying composition with the receptor rather than to alterations in G protein ␣-␤␥ subunit interactions.
Purification of G Protein ␤ 1 ␥ Dimers-To produce recombinant ␤␥ dimers of varying ␥ composition, the ␤ 1 subtype was chosen since it has previously been shown to interact with all of the ␥ subtypes (17,27,28). The recombinant ␤␥ dimers were purified using a procedure originally described by Kozasa and Gilman (11). Sf9 cells expressing the 6his ␣ i1 , ␤ 1 , and one of the following ␥ 1 , ␥ 2 , ␥ 3 , ␥ 4 , ␥ 5 , ␥ 7 , ␥ 10 , or ␥ 11 subunits were prepared. Then, cholate-solubilized membrane extracts from these cells were bound to Ni-NTA agarose columns by virtue of the 6-histidine tag on the ␣ i1 subunit; the ␤␥ dimers were eluted from the columns by activating the bound heterotrimers with AMF; and the 6his ␣ i1 subunit was subsequently eluted from the columns with high imidazole. Further purification of the recombinant ␤␥ dimers and the 6his ␣ i1 subunit was achieved by applying their enriched fractions from the Ni-NTA columns to Mono Q columns.
The purity of the 6his ␣ i1 subunit was assessed by SDS-PAGE and silver staining (29). As shown in Fig. 4A, the purified 6his ␣ i1 preparation contained one major band of the size expected for the ␣ i1 subunit taking into account the added amino-terminal tag. The purity of the recombinant ␤␥ dimers was also compared by SDS-PAGE and silver staining (Coomassie was used in the case of ␤ 1 ␥ 1 ). As shown in Fig. 4A, each purified ␤␥ preparation was composed of two predominant bands by protein staining as follows: a 36-kDa band representing the ␤ 1 subunit, and a 5-8-kDa band representing one of the following ␥ 1 , ␥ 2 , ␥ 3 , ␥ 4 , ␥ 5 , ␥ 7 , ␥ 10 , or ␥ 11 subunits. The identity of each ␤␥ purified preparation was confirmed by immunoblotting with antibodies specific for each ␥ subtype. Antibodies specific for the ␥ 1 , ␥ 2 , ␥ 3 , ␥ 5 , and ␥ 7 subunits were used for this purpose previously (20). However, antibodies specific for the newly described ␥ 4 , ␥ 10 , and ␥ 11 subunits needed to be generated against synthetic peptides based on the unique amino acid sequences of these proteins. As shown in Fig. 4B, the identities of the purified ␤ 1 ␥ 4 , ␤ 1 ␥ 10 , and ␤ 1 ␥ 11 preparations were confirmed by immunoblotting with these antibodies.
Comparison  with increasing amounts of purified bovine brain ␤␥ to yield a final 6his ␣ i1 :receptor ratio of 75:1 (mol/mol) throughout the curve and the ␤␥:receptor ratio indicated on the ordinate. The curve was fit to the data by non-weighted, non-linear regression analysis using a one-site hyperbola (GraphPad Prism). The data are representative of three similar results. FIG. 4. Purified recombinant G protein subunits. A, 1st lane, 3 g of purified ␤ 1 ␥ 1 was trichloroacetic acid-precipitated and loaded onto a 15% polyacrylamide gel in 40 l of Laemmli sample buffer and stained with Coomassie Blue following electrophoresis; 2nd to 9th lanes, silverstained purified G protein subunits. 300 ng each (determined by Amido Black staining) of purified G protein subunits were trichloroacetic acidprecipitated and loaded onto 15% polyacrylamide gels in 40 l each of Laemmli sample buffer containing 120 mM DTT. Gels were silverstained following the method of Fawzi et al. (30). Lanes contain the following subunits: 2nd lane, following Western transfer, 300 ng each of purified ␤␥ subunits were probed with the following ␥-specific antibodies: ␥ 4 , E59 at 1/250; ␥ 10 , E57 at 1/250; and ␥ 11 , E60 at 1/300. The peak of ␤␥ elution occurred at Mono Q fractions 11 and 12, corresponding to 200 -240 mM NaCl, in each case. (19), the observation that they promote very different levels of coupling suggests that the relatively small number of amino acid differences between the two subtypes are important for recognition by the receptor. The ␤ 1 ␥ 2 , ␤ 1 ␥ 3 , ␤ 1 ␥ 4 , and ␤ 1 ␥ 7 dimers also produced high levels of coupling with the recombinant ␣ 2A -adrenergic receptor. These four ␥ subtypes are closely related, showing 66 -74% homology at the amino acid level, and therefore, the finding that they produce essentially identical levels of coupling is not unexpected. Finally, the ␤ 1 ␥ 5 and ␤ 1 ␥ 10 dimers yielded intermediate levels of coupling with the recombinant ␣ 2A -adrenergic receptor. These two subtypes are only distantly related, showing less than 53% homology to each other or to other ␥ subtypes. Taken together, these results showed measurable differences between ␤␥ dimers of varying ␥ composition to support coupling of the same ␣ subunit to the recombinant ␣ 2A -adrenergic receptor. Statistical analysis revealed the ␤ 1 ␥ 1 , ␤ 1 ␥ 5 , and ␤ 1 ␥ 10 dimers supported the lowest levels of coupling. Previously, several groups have reported that the ␤ 1 ␥ 1 dimer was less effective than other ␤␥ dimers in supporting coupling to numerous receptors (8,9,30) and that this problem could not be overcome by increasing its concentration. To extend further these observations, we showed that increasing the concentration of the ␤ 1 ␥ 5 dimer did not raise the level of receptor coupling (Fig. 5B), indicating the ␥ 5 subtype has a lower intrinsic ability to interact with the recombinant ␣ 2A -adrenergic receptor. Based on results for both the ␤ 1 ␥ 1 and ␤ 1 ␥ 5 dimers, a similar result could be predicted for the moderately effective ␤ 1 ␥ 10 dimer.
The lower activities of the ␤ 1 ␥ 1 , ␤ 1 ␥ 5 , and ␤ 1 ␥ 10 dimers could be due to differences in affinities for the ␣ subunit of the G protein rather than for the ␣ 2A -adrenergic receptor itself. To evaluate this possibility, the affinities of representative ␤␥ dimers for the 6his ␣ i1 subunit were measured by the pertussis toxin-dependent ADP-ribosylation assay. Under the reconstitution conditions used in this study, which employed relatively high concentrations of ␤␥ dimers, there were no real differences in the affinity of these ␤␥ dimers for the 6his ␣ i1 subunit (Fig. 6). Thus, the observed differences between ␤␥ dimers of varying ␥ composition reflect their intrinsic abilities to interact with the receptor, suggesting structural diversity among ␥ subtypes plays a role in agonist-stimulated receptor-G protein complex formation.
Purification of G Protein ␤ 3 ␥ Dimers-A variety of approaches has been used to examine the ability of the 6 ␤ and 11 ␥ subtypes to form ␤␥ dimers (17,19,27,28). Overall, these approaches have provided consistent results showing that all the known ␥ subtypes are able to interact with the ␤ 1 subtype and, to a lesser extent, the ␤ 2 subtype. However, these approaches have yielded conflicting results regarding the abilities of the known ␥ subtypes to interact with the ␤ 3 subtype. In this regard, the lack of a ␤␥ dimer containing the ␤ 3 subtype to serve as a positive control in the various assays has been a particular hindrance. To this end, the method of Kozasa and Gilman (11) was used to obtain a ␤␥ dimer containing the ␤ 3 subtype. Sf9 cells were infected with recombinant viruses encoding the 6his ␣ i1 , ␤ 3 , and ␥ 5 subunits either simultaneously or separately, and the ␤ 3 ␥ 5 dimer was then purified by the procedure described above. As shown in Fig. 7A, the co-expression of the 6his ␣ i1 , ␤ 3 , and ␥ 5 subunits resulted in the appearance of ␤ 3 and ␥ 5 subunits in the AMF activation fractions as detected by immunoblotting. This result was consistent with the release of  Fig. 5, 500 ng of purified 6his ␣ i1 were combined with 330 ng of the various purified ␤ 1 ␥ dimers at a final ratio of 75:50 (mol:mol) of ␣:␤␥. Incubation of the purified proteins in the presence of 200 ng of islet activating protein and [ 32 P]NAD was terminated after 20 min at 30°C by precipitation with 30% trichloroacetic acid followed by rapid filtration over BA85 nitrocellulose filters. ADP-ribosylation of 6his ␣ i1 was measured by scintillation counting to detect [ 32 P] bound to filters. Bars show the mean cpm Ϯ S.E. values from triplicate determinations with the specific ␤ 1 ␥ combination indicated. Background [ 32 P] bound to filters in the absence of 6his ␣ i1 has been subtracted from these values. ␤ 3 ␥ 5 dimer from the 6his ␣ i1 subunit in response to AMF activation. As shown in Fig. 7B, the identity of the AMF-released ␤ subunit as the ␤ 3 subunit was confirmed by immunoblotting with a ␤ 3 subtype-specific antibody (B-34). Taken together, these results supported the conclusion that the ␤ 3 and ␥ 5 subunits interact to form a functional ␤␥ dimer.
An example of the purity of the ␤ 3 ␥ 4 , ␤ 3 ␥ 5 , and ␤ 3 ␥ 11 subunits that can be obtained by this procedure is shown by silver staining. As shown in Fig. 8A, each purified ␤ 3 ␥ preparation was composed of two predominant bands by silver staining as follows: a 37-kDa band representing the ␤ 3 subunit and a 5-8-kDa band representing one of the ␥ subtypes. Confirmation that the ␤ 3 and ␥ 5 subunits interact to form a functional ␤␥ dimer is also shown using a previously developed tryptic digestion procedure (19,27). This method is based on the finding that ␤ monomers are cleaved at numerous sites by trypsin. By contrast, functional ␤␥ dimers are cleaved at a single site, resulting in the appearance of a 26-kDa fragment of the ␤ subunit that is resistant to further digestion by trypsin. Whereas the appearance of this stable 26-kDa fragment is a reliable marker for the formation of ␤␥ dimers containing the ␤ 1 and ␤ 2 subtypes, it is not clear whether formation of ␤␥ dimers containing the ␤ 3 subtype yields the appearance of a similar protected fragment. To date, such a protected fragment has not been detected for the ␤ 3 subtype, but these results are difficult to interpret due to the lack of availability of a positive control at that time (19). As shown in Fig. 8B, purified preparations of both the ␤ 1 ␥ 5 and ␤ 3 ␥ 5 subunits produced a 26-kDa protected fragment when digested under identical conditions with trypsin, as detected in each case by immunoblotting with a commercial ␤-antibody (DuPont SW/1, carboxyl terminus). Since equal amounts of ␤ 1 ␥ 5 and ␤ 3 ␥ 5 subunits were loaded, the differences in intensities of the ␤ 1 and ␤ 3 bands presumably reflect differences in affinities of the antibody used for immunoblotting. Taken together, these results confirmed that the ␤ 3 and ␥ 5 subunits are able to interact to form a functional ␤␥ dimer. Moreover, these results extended the utility of the trypsin digestion method as a reliable marker of ␤␥ dimer formation to the ␤ 3 subtype.
Comparison of ␤␥ Dimers of Varying ␤ Composition in Terms of Coupling to the ␣ 2A -Adrenergic Receptor-As shown in Fig. 9, all combinations of the 6his ␣ i1 subunit with the various ␤ 3 ␥ dimers were capable of inducing high affinity [ 3 H]UK-14,304 binding. Again, in all cases, the [ 3 H]UK-14,304 binding was completely abolished by addition of GTP␥S (data not shown). The ␤ 1 ␥ 4 and ␤ 3 ␥ 4 dimers showed similar abilities to reconstitute coupling with the recombinant ␣ 2A -adrenergic receptor. Likewise, the ␤ 1 ␥ 11 and ␤ 3 ␥ 11 dimers had essentially identical activities. On the other hand, the ␤ 3 ␥ 5 dimer showed a substantially higher level of coupling with the recombinant ␣ 2Aadrenergic receptor than the ␤ 1 ␥ 5 dimer. Increasing the concentration of the ␤ 1 ␥ 5 dimer did not raise the level of coupling (Fig. 5B), indicating the ␤ 1 subtype, when in association with the ␥ 5 subtype, has a lower intrinsic ability to interact with the recombinant ␣ 2A -adrenergic receptor. Taken together, these differences support the conclusion that the receptor recognition of the G protein is dependent on the particular combination of ␤ and ␥ subtypes. DISCUSSION The present study examined the potential of the ␣ 2A -adrenergic receptor to couple to G proteins differing in their ␤␥ subunit composition only. The selectivity of coupling was directly assessed by a high affinity agonist binding assay. Importantly, this assay was found to require the addition of the ␤␥ subunits in order to detect the interaction of the ␣ subunit with the receptor. From previous studies, this requirement for the ␤␥ subunits appears to reflect not only a general role of the ␤␥ subunits to stabilize the ␣ subunit (31) but also a specific role of the ␤␥ subunits to interact directly with the receptor (6, 7).  11 . B, protection of ␤ 3 from tryptic proteolysis by association with ␥ 5 . ␤␥ subunits were incubated for 40 min at 30°C with or without 1 g of trypsin, after which 6 g of trypsin inhibitor were added to each, and the samples were trichloroacetic acid-precipitated for 15% SDS-PAGE. The products are shown by Western blot using ␤-common antibody (SW/1, NEN Life Science Products). 1st and 2nd lanes, 400 ng each of ␤ 1 ␥ 5 ; 3rd and 4th lanes, 400 ng each of ␤ 3 ␥ 5 .
In view of these roles and the rich diversity of ␤␥ subunit combinations, the possibility was suggested that the nature of the ␤␥ subunits might contribute to the selectivity of the receptor interaction. Supporting such a possibility, the present study showed clear differences in the abilities of the various ␤␥ dimers, including those containing the ␤ 3 subtype and the newly described ␥ 4 , ␥ 10 , and ␥ 11 subtypes, to promote interaction of the same ␣ i subunit with the ␣ 2A -adrenergic receptor.
Influence of ␤␥ Composition on Receptor Coupling-Several lines of evidence support the validity of using Sf9 insect cell membranes expressing the recombinant ␣ 2A -adrenergic receptor as a suitable system for examining the specificity of coupling to purified, recombinant G proteins. First, the recombinant ␣ 2A -adrenergic receptor displayed the binding affinities and pharmacologic properties characteristic of the native receptor. Second, the recombinant ␣ 2A -adrenergic receptor showed a mostly uncoupled phenotype in the absence of added G proteins and a largely coupled phenotype in the presence of added G proteins of defined composition and stoichiometry. Using high affinity agonist binding as a quantitative measure of the coupled phenotype, this system was first used to examine the influence of the ␥ component on receptor coupling. G i proteins were produced from 6his ␣ i1 , ␤ 1 , and varying ␥ subtypes.
Among the eight ␤␥ dimers tested, 30-fold differences were observed in their abilities to support coupling of the 6his ␣ i1 subunit to the ␣ 2A -adrenergic receptor, with the ␤ 1 ␥ 2 , ␤ 1 ␥ 3 , ␤ 1 ␥ 4 , ␤ 1 ␥ 7 , and ␤ 1 ␥ 11 dimers displaying the most efficacy, the ␤ 1 ␥ 5 and ␤ 1 ␥ 10 dimers showing intermediate efficacies, and the ␤ 1 ␥ 1 dimer exhibiting the least efficacy. With the exception of the ␥ 11 subtype, the observed differences segregated with the structural diversity of the ␥ component along subclass lines. As defined previously, the human ␥ subunit family has been divided into three subclasses, with each subclass showing less than 50% amino acid homology to other subclasses (1). On this basis, subclass I contains the ␥ 1 , ␥ c , and ␥ 11 subtypes, which are modified by the less common C-15 farnesyl group; subclass II includes the ␥ 2 , ␥ 3 , ␥ 4 , ␥ 7 , and ␥ 12 subtypes, which are modified by the more common C-20 geranylgeranyl group; and subclass III contains the ␥ 5 and ␥ 10 subtypes, which again receive the more common C-20 geranylgeranyl group. The present study represents the most extensive functional analysis of the ␥ subunit family to date. Next, this system was used to examine the influence of the ␤ 1 versus the ␤ 3 subtype on receptor coupling. In vitro studies have revealed little or no functional differences due to the ␤ subunit (16,17). Since only the closely related ␤ 1 and ␤ 2 subtypes were examined, however, the present study extended this analysis to the more divergent ␤ 3 subtype. For this purpose, the method of Kozasa and Gilman (11) was used to produce and then purify functional ␤␥ dimers containing the ␤ 3 subtype. In addition to providing a source of material of defined composition, this approach also revealed new information on the selectivity of ␤-␥ interaction by confirming the ability of the ␤ 3 subtype to interact with the ␥ 4 , ␥ 5 , and ␥ 11 subtypes. Whereas interaction between the ␤ 3 and ␥ 4 subtypes had been predicted (2,3), the ability of the ␤ 3 subtype to interact with the ␥ 11 subtype was unexpected in view of the high homology between the ␥ 11 and ␥ 1 subtypes and the reported failure of the ␥ 1 subtype to interact with the ␤ 3 subtype (28). When the various ␤␥ dimers were tested for receptor coupling, only the ␤ 3 ␥ 5 and ␤ 1 ␥ 5 dimers showed substantive differences in their abilities to support coupling of the 6his ␣ i1 subunit to the ␣ 2A -adrenergic receptor. No such differences were observed with the ␤ 3 ␥ 4 and ␤ 1 ␥ 4 dimers nor the ␤ 3 ␥ 11 and ␤ 1 ␥ 11 dimers. These results suggested that it is the particular combination of ␤ and ␥ subtypes that ultimately determines receptor recognition. Interestingly, the ␤ 3 ␥ 5 dimer was shown previously to interact preferentially with a G protein-coupled receptor kinase, GRK3, indicating a role of the ␤ subtype in selective receptor desensitization (32).
Taken together, these data show that G i proteins containing different ␤␥ dimers produce different levels of coupling to the ␣ 2A -adrenergic receptor. This result could arise because the composition of the ␤␥ subunits alters the formation or stability of the G protein, the affinity of the ␤␥ dimer for the receptor, or some combination therefrom. Our data (17) and those from other laboratories (9,33) suggest that formation of the G protein is the least likely reason since the affinity of the ␣ subunit for the various ␤␥ dimers is similar. Instead, our data are most consistent with the composition of the ␤ and, particularly, the ␥ component affecting the affinity of the G protein for the ␣ 2A -adrenergic receptor. Studies of the A 1 -adenosine receptor support a similar conclusion (33).
Sites of Interaction of ␥ Component with Receptor-The observed differences in the abilities of various ␤␥ dimers to support coupling to the ␣ 2A -adrenergic receptor reside primarily in the ␥ component. Although cross-linking studies have yet to detect receptor-␥ contact sites (7), several studies point to the importance of the carboxyl-terminal amino acid region and the type of prenyl group on the ␥ subunit in determining the interaction of the ␤␥ dimer with receptor (34,35). These latter studies may explain the lower reconstitutive activity of the ␤ 1 ␥ 1 dimer in the present study. In this regard, the ␥ 1 subtype sequence is the most divergent of those determined to date, and its lipid modification is a C-15 farnesyl group rather than the C-20 geranylgeranyl group found on most other ␥ subtypes (1). With regard to the latter, a recent study comparing ␤␥ dimers with the two types of prenyl groups showed that the wild type, geranylgeranylated ␥ 2 subunit and the mutant, geranylgeranylated ␥ 1 subunit had similar abilities to interact with the A 1 adenosine receptor (34). By contrast, the wild type, farnesylated ␥ 1 subunit and the mutant, farnesylated ␥ 2 subunit were much less effective. Whereas these data indicate that type of prenyl group is critical, the primary structure of the ␥ subunit is of equal or greater importance. Underscoring this point, a synthetic peptide derived from the carboxyl-terminal sequence of the ␥ 1 subtype was able to stabilize the light-activated state of rhodopsin receptor to a much greater degree when the peptide was farnesylated than not (35). However, the effect was greatly attenuated when the amino acid sequence of the peptide changed by only two amino acid substitutions (F64T and High affinity agonist binding was calculated as the difference between total binding and binding in the presence of GTP␥S. Background (i.e. high affinity agonist binding without added G i ) was subtracted from each value to show only the reconstituted coupling. ␤ 1 ␥ 5 supported significantly less coupling than ␤ 3 ␥ 5 ( Յ 0.05, Student's t test). L67S) even though the farnesylated state was preserved. Thus, both the primary structure and the prenylation state of the ␥ 1 subtype are likely to contribute to its poor affinity for the ␣ 2A -adrenergic receptor in the present study and its contrastingly high affinity for the rhodopsin receptor in previous studies (35). When compared with the ␤ 1 ␥ 1 dimer, the higher reconstitutive activity of the ␤ 1 ␥ 11 dimer was unexpected since the newly described ␥ 11 subtype is farnesylated and has a carboxyl-terminal tail nearly identical to ␥ 1 subtype. This result suggests the importance of regions other than the carboxyl-terminal tail of the ␥ 11 subunit in promoting its strong interaction with the ␣ 2A -adrenergic receptor. By focusing on the few amino acid differences between the ␥ 11 and ␥ 1 subtypes and making the appropriate mutations, it should be possible to pinpoint other regions of ␥ protein structure that are selectively recognized by the ␣ 2A -adrenergic receptor. Given the wide tissue distribution of the ␥ 11 subtype (19) in contrast to the restricted expression of the ␥ 1 subtype, it is perhaps not so surprising that a receptor other than rhodopsin would exist which prefers the farnesylated ␥ 11 subtype in tissues other than the retina.
Finally, the high reconstitutive activities of the ␤ 1 ␥ 2 , ␤ 1 ␥ 3 , ␤ 1 ␥ 4 or ␤ 1 ␥ 7 dimers compared with the intermediate activities of the ␤ 1 ␥ 5 and ␤ 1 ␥ 10 dimers underscore the importance of primary structure in another way. Since all of the aforementioned ␥ subtypes are modified by the C-20 geranylgeranyl group (19), the observed differences between the two groups must relate to the differences in protein structure. In agreement with this, the ␥ 2 , ␥ 3 , ␥ 4 and ␥ 7 subtypes share a high degree of amino acid homology (66 -74%), and predictably, ␤␥ dimers containing these ␥ subtypes produce comparably strong levels of coupling of the 6his ␣ i1 subunit to the ␣ 2A -adrenergic receptor. By contrast, the ␥ 5 and ␥ 10 subtypes share only 35-50% identity with the aforementioned group of ␥ subunits (19), and accordingly, ␤␥ dimers containing these ␥ subtypes yielded significantly lower levels of coupling in comparison with the mean value calculated from the grouping of ␤␥ dimers consisting of the ␥ 2 , ␥ 3 , ␥ 4 , and ␥ 7 subtypes ( ϭ 0.025 or 0.05, respectively; Student's t test). Taken together, these data indicate that the primary structure, including regions other than the carboxyl-terminal tail, of the ␥ subunit is a most important factor in determining the selectivity of interaction with the ␣ 2A -adrenergic receptor. In this instance, the type of prenyl group is a less critical factor.
Sites of Interaction of ␤ Component with Receptor-Although the ␤ subunits are highly conserved in their predicted amino acid sequences, the observed difference between the ␤ 1 ␥ 5 and ␤ 3 ␥ 5 dimers indicates that selective receptor recognition does occur on the basis of the ␤ subtype. Cross-linking studies have revealed a site of contact between the 7th or possibly 6th WD-40 repeat of the ␤ subunit and a peptide derived from the third intracellular loop of the ␣ 2 -adrenergic receptor (7). The crystal structure of ␤␥ t indicates that residue 303, which lies at the center of this segment, resides on an exposed surface of the ␤␥ dimer (36). Thus, this site has the potential to participate in the preferential coupling of the ␣ 2A -adrenergic receptor to G i heterotrimers containing the ␤ 3 ␥ 5 dimer over those containing the ␤ 1 ␥ 5 dimer. Other possibilities are suggested: 1) regions other than the carboxyl-terminal tail of the ␤ subunit may interact with receptor; 2) a concerted interaction between the ␤ and ␥ subunits within the G protein binding pocket of the receptor; or 3) some combination therefrom.
Influence of ␤␥ Composition on Coupling to Other G Proteincoupled Receptors-Various ␤␥ dimers show a different order of potency depending on the type of receptor (8,9,37). It is speculated that protein-protein interactions between the ␤␥ subunits and receptors, as well as hydrophobic interactions due to the prenylation state of the ␥ subunit, will be important elements in modeling selective recognition between G protein and receptors. Distinct, yet to be revealed, structural features within the G protein binding regions of receptor subtypes must also be taken into account in such a model. For example, a previous study showed that the 5HT 1A receptor interacts similarly with G proteins containing ␤ 1 ␥ 2 , ␤ 1 ␥ 3 , or ␤ 1 ␥ 5 dimers (8), whereas the present study revealed that the ␣ 2A -adrenergic receptor prefers G proteins containing the ␤ 1 ␥ 2 or ␤ 1 ␥ 3 dimer over that containing the ␤ 1 ␥ 5 dimer. Thus, the G protein binding pockets of the ␣ 2A -adrenergic receptor and the 5HT 1A may possess subtle structural differences that result in either more or less receptor to G␣␤␥ contact depending on the identity of the ␥ subunit. Active-state receptors may possess discrete elements contacting ␣, ␤, and ␥ subunits within the same G protein that complement one another to some degree in order to activate the heterotrimer. The experimental approach used here is quite amenable to manipulating both the receptor and the G protein subunit composition in order to bring together selected elements of signaling proteins.
Summary-The data in the present study demonstrate the specificity of ␣ 2 -adrenergic receptor-G protein interactions is affected by the ␤␥ dimer composition, with protein-protein interactions forming the basis for the observed differences. These in vitro results complement a growing body of in vivo results demonstrating the ␤␥ subunit composition is an important determinant of the specificity of signaling pathways. Strikingly, antisense suppression of the ␤ 1 ␥ 3 or ␤ 3 ␥ 4 subtypes disrupts coupling between inhibition of a calcium channel and the somatostatin or muscarinic receptors, respectively, in GH3 pituitary cells (2,3). Similarly, ribozyme suppression of the ␥ 7 subtype attenuates coupling between stimulation of adenylylcyclase and the ␤-adrenergic receptor in HEK293 cells (4). Although these results could arise from an ordered arrangement of signaling proteins in the cell membrane, the in vitro results presented here implicate receptor-␤␥ "recognition" as an additional mechanism for determining the specificity of signaling. It is expected that a combination of in vitro and in vivo approaches will provide some of the answers needed for construction of a mechanistic model of specificity in G proteinmediated signaling pathways.