The G Protein β5 Subunit Interacts Selectively with the Gq α Subunit*

The diversity in the heterotrimeric G protein α, β, and γ subunits may allow selective protein-protein interactions and provide specificity for signaling pathways. We examined the ability of five α subunits (αi1, αi2, αo, αs, and αq) to associate with three β subunits (β1, β2, and β5) dimerized to a γ2 subunit containing an amino-terminal hexahistidine-FLAG affinity tag (γ2HF). Sf9 insect cells were used to overexpress the recombinant proteins. The hexahistidine-FLAG sequence does not hinder the function of the β1γ2HF dimer as it can be specifically eluted from an αi1-agarose column with GDP and AlF4 −, and purified β1γ2HF dimer stimulates type II adenylyl cyclase. The β1γ2HF and β2γ2HF dimers immobilized on an anti-FLAG affinity column bound all five α subunits tested, whereas the β5γ2HF dimer bound only αq. The ability of other α subunits to compete with the αqsubunit for binding to the β5γ2HF dimer was tested. Addition of increasing amounts of purified, recombinant αi1 to the αq in a Sf9 cell extract did not decrease the amount of αq bound to the β5γ2HF column. When G proteins in an extract of brain membranes were activated with GDP and AlF4 − and deactivated in the presence of equal amounts of the β1γ2HF or β5γ2HF dimers, only αq bound to the β5γ2HF dimer. The αq-β5γ2HF interaction on the column was functional as GDP, and AlF4 −specifically eluted αq from the column. These results indicate that although the β1 and β2 subunits interact with α subunits from the αi, αs, and αq families, the structurally divergent β5 subunit only interacts with αq.

All cells possess multiple signaling pathways that transmit signals from the hormones, autacoids, neurotransmitters, and growth factors in their environment. Complex biochemical mechanisms exist to discriminate, integrate, and modulate a cell's response to this constantly changing set of stimuli. One of the best characterized signal transduction systems is the pathway used by receptors coupled to heterotrimeric G proteins 1 (1,2). Our current understanding of this signaling path-way shows it to be surprisingly complex with large families of proteins comprising the receptors, G proteins, and effectors (1,3,4) and important roles for both the ␣ and ␤␥ subunits of the heterotrimer in activating effectors (2,4,5). Moreover, some ligands activate multiple G proteins (1,2,6), and certain receptors activate the MAP kinase pathway (6,7) and/or other tyrosine kinase signaling pathways (8). Thus, an important unsolved question in cell signaling is how a cell selects a response from the multiple possibilities available.
Current evidence holds that specificity is determined at many levels. In addition to the tissue-specific expression of receptors, G proteins, or effectors (3), there are important protein-protein interactions involving the ␣ and ␤␥ subunits of the G protein heterotrimer that determine specificity. For example, the ␣ t subunit couples selectively to rhodopsin and the ␣ s subunit to the ␤-adrenergic receptor (1,2,6). Furthermore, it is clear that the ␤␥ dimer is required for efficient coupling of the ␣ subunit to receptors (9,10), and there is growing evidence supporting specific interactions of receptors with the ␤␥ dimer (11)(12)(13). Both the ␣ and ␤␥ subunits of transducin appear to contact rhodopsin (14,15), and the presence of the ␤␥ dimer significantly increases the affinity of the G t ␣ subunit for rhodopsin (14). In this regard, the carboxyl terminus of the ␥ subunit and its prenyl modification have emerged as important determinants of the interaction of G proteins with receptors (12,13). Experiments using antisense RNA to selectively remove G protein subunits in GH 3 cells also support a role for the diversity of the G protein ␣␤␥ subunits in determining signaling specificity. In these cells, the G␣ o1 ␤ 3 ␥ 4 heterotrimer couples preferentially to the muscarinic receptor, G␣ o1 ␤ 2 ␥ 2 to the galanin receptor, and the G␣ o2 ␤ 1 ␥ 3 combination to the somatostatin receptor (16,17). Similar experiments using rat basophilic leukemia cells suggest that the m1 muscarinic receptor couples selectively to ␣ q , ␣ 11 , ␤ 1 , ␤ 4 , and ␥ 4 (18). Thus, the existence of multiple isoforms of the ␣, ␤, and ␥ subunits and the participation of both ␣ and ␤␥ subunits in receptor coupling implies that the diversity of the subunits in the G protein heterotrimer could play an important role in signal transduction.
Although a large number of studies have focused on specific interactions of the ␣ and/or ␤␥ subunits with effectors (4,19), there are few investigations of the role of the ␣-␤␥ interaction in cell signaling. Perhaps it has been assumed that because of the similarity of the known ␤ subunits, all ␣ subunits would associate with all ␤ subunits. Recently, two more divergent members of the ␤ subunit family, ␤ 5 and ␤ 5L , have been described (20,21). Whereas the amino acid sequences of ␤ 1 , ␤ 2 , ␤ 3 , and ␤ 4 are 80 -90% identical, ␤ 5 is only 52% identical and 64% similar to ␤ 1 (20). In addition, ␤ 5 has an eight amino acid extension near the amino terminus and three short amino acid insertions within the WD repeat regions of the molecule. The ␤ 5 subunit is expressed predominantly in the brain, with only trace amounts detected by Northern analysis in the kidney. The ␤ 5L subunit appears to be expressed only in retina. Both the ␤ 5 and ␤ 5L subunits can stimulate PLC-␤ 2 activity when transiently transfected into COS-7 cells with the ␥ 2 subunit (20,21). However, the ␤ 5 ␥ 2 dimer fails to activate the MAP kinase pathway when transfected into these cells (22). This observation suggests that dimers containing the ␤ 5 subunit may have different functions from those containing other ␤ subunits. In the experiments reported here, we have tested the ability of several ␣ subunits to interact with ␤␥ dimers containing the ␤ 1 , ␤ 2 , or ␤ 5 subunit to determine if the variations in amino acid sequence observed for the ␤ subunits are manifested as differences in affinity for ␣ subunits. Sf9 cells were co-infected with an affinity tagged ␥ 2 subunit and various ␤ subunits. The resulting ␤␥ 2HF dimers were immobilized via the affinity tag and allowed to interact with a variety of recombinant ␣ subunits expressed in Sf9 cells. The results show that the ␤ 1 ␥ 2HF and ␤ 2 ␥ 2HF dimers interact with five different ␣ subunits from three families, whereas the ␤ 5 ␥ 2HF subunit only interacts with the ␣ q subunit.

EXPERIMENTAL PROCEDURES
Construction of Recombinant Baculoviruses for the ␥ 2HF and ␤ 5 Subunits-The polymerase chain reaction was used to modify the cDNA encoding the ␥ 2 subunit (23) by adding XbaI and BamHI restriction sites to the 5Ј and 3Ј ends of the ␥ 2 coding region, respectively. The primers used were Sense primer: 5Ј-AACTCTAGAATGGCCAGCAA-CAACACCGC-3Ј XbaI; and Antisense primer: 5Ј-CCTGGATCCTTA-AAGGATAGCACAGAAAAACTTC-3Ј BamHI. The products of the polymerase chain reaction were digested with XbaI and BamHI and ligated into the pDoubleTrouble (pDT) vector (24) to add the nucleotide sequences for the hexahistidine and FLAG affinity tags to the 5Ј end of the ␥ 2 coding region. To construct useful restriction sites for subcloning into the baculovirus transfer vector pVL1393, the ␥ 2HF coding sequence was excised from pDT with KpnI and BamHI and subcloned into the pCNTR shuttle vector using the Prime Efficiency Blunt-End DNA Ligation Kit (5 Prime 3 3 Prime). The ␥ 2HF coding region was excised from pCNTR with BamHI and ligated into the BamHI site of pVL1393 to place the ATG of the hexahistidine sequence 75 bases downstream of the polyhedron promoter. The mouse ␤ 5 cDNA in a Bluescript SKII vector was kindly provided by Dr. Melvin I. Simon of the California Institute of Technology. The 1803-base pair BamHI-XbaI fragment of the ␤ 5 cDNA was subcloned into the BamHI-XbaI sites of pVL1393. To ensure fidelity, both completed transfer vectors were sequenced in the forward and reverse directions using dye terminator sequencing on an automated sequencer (Applied Biosystems, model 377). Recombinant baculoviruses were isolated following co-transfection of the transfer vector and linearized BaculoGold viral DNA into Sf9 cells using the PharMingen BaculoGold ® kit. Briefly, 2 ϫ 10 6 Sf9 cells were co-transfected with 1 g of linear BaculoGold DNA and 3-5 g of recombinant baculovirus transfer vector DNA using calcium phosphate/DNA precipitation. Following a 4-h incubation at 27°C, the co-transfection medium was removed, and the monolayer was rinsed with fresh TNM-FH medium (25) supplemented with 10% fetal bovine serum, 50 g/ml gentamicin sulfate, and 2.5 g/ml amphotericin B. The plates were incubated at 27°C with 5 ml of fresh medium for 4 -6 days. Recombinants were detected by observing swollen, extremely large cells associated with a low cell density and a large amount of cell debris. Recombinant baculoviruses were purified by one round of plaque purification using standard techniques (25). The construction of the recombinant baculoviruses coding for the ␣ i1 , ␣ i2 , ␣ o , ␣ s , and ␣ q subunits and the ␤ 1 and ␤ 2 subunits has been described (26 -28). The baculovirus encoding the avian ␣11 protein (29) was the kind gift of Dr. T. K. Harden.
Expression and Purification of Recombinant G Protein ␣ and ␤␥ Subunits-Recombinant G protein subunits were overexpressed in suspension cultures of Sf9 insect cells as described (26,27,30). In most experiments, the recombinant ␣ and ␤␥ subunits were extracted from cell pellets using the detergent Genapol C-100 at a concentration of 0.1% (w/v). All steps were performed at 4°C. Frozen pellets were thawed in 15 ϫ their wet weight in lysis buffer containing 20 mM Hepes, pH 7.5, 150 mM NaCl, 3 mM MgCl 2 , 1 mM EDTA, 17 g/ml phenylmethylsulfonyl fluoride, 20 g/ml benzamidine, and 2 g/ml each of aproti-nin, leupeptin, and pepstatin, and burst by nitrogen cavitation (600 p.s.i. for 20 min) at 0°C. The crude lysate was mixed with an equal volume of lysis buffer supplemented with 0.2% (w/v) Genapol C-100 and stirred for 1 h. The Genapol extract was centrifuged at 100,000 ϫ g for 60 min and the supernatant decanted, and aliquots were frozen in liquid N 2 . The ␣ s subunit used in the adenylyl cyclase assays was prepared as above except 0.1% (w/v) CHAPS was substituted for 0.1% (w/v) Genapol C-100. Partially purified ␣ i1 subunits were used in some experiments. Sf9 cell pellets overexpressing the ␣ i1 subunit were extracted without detergent; a 100,000 ϫ g supernatant was prepared and the protein purified on a DEAE column exactly as described (26). This preparation of ␣ i1 subunit is approximately 95% pure, as determined by quantitation of a silver-stained gel.
Preparation of the ␤␥-Anti-FLAG Affinity Column-All column steps were carried out at 4°C. Typically, 1 ml of the Genapol C-100 extract of Sf9 cells overexpressing the desired ␤␥ 2HF dimer was applied to a 0.5-ml anti-FLAG M2 affinity gel column equilibrated with column buffer (lysis buffer containing 0.1% Genapol C-100 and 1 mM ␤-mercaptoethanol) at a flow rate of 0.2 ml/min. The resulting ␤␥ 2HF -anti-FLAG affinity gel column (␤␥ 2HF affinity column) was washed three times with 3 ml of column buffer. This procedure resulted in a highly pure preparation of ␤␥ dimers immobilized on the column (see Fig. 1). The amount of ␤␥ dimer immobilized on the column was about 6 g/0.5 ml of resin as judged by silver staining of the ␤␥ dimer eluted from the column with 0.1 M glycine, pH 3.5. This represents about 5% of the nominally available FLAG binding sites. Usually the interaction of an ␣ subunit with a particular ␤ subunit was measured by applying 2 ml of a Genapol C-100 extract of Sf9 cells expressing the desired ␣ subunit to a ␤␥ 2HF affinity column at a flow rate of 0.2 ml/min. The resulting ␣␤␥ 2HF affinity column was washed 4 times with 4 ml of column buffer. In some experiments, the ␣␤␥ 2HF heterotrimer was eluted with 0.1 M glycine, pH 3.5. In the experiment presented in Fig. 5A, the procedure was modified such that 2 ml of ␣ i2 extract and 2 ml of ␣ q extract were mixed and applied to the ␤␥ 2HF affinity column. In the experiment presented in Fig. 6, a range of 17.5-525 g of partially purified ␣ i1 was mixed with 2 ml of ␣ q extract and applied to the ␤␥ 2HF affinity column.
Specific Elution of ␣ Subunits from ␤␥ 2HF -Anti-FLAG Affinity Gel-To demonstrate a functional interaction between ␣ subunits and the immobilized ␤␥ subunits, a ␤␥ 2HF affinity column (0.5 ml) was prepared at 4°C as described above, washed three times with 3 ml of column buffer, and then equilibrated in ␣ subunit binding buffer (20 mM Hepes, pH 8.0, 100 mM NaCl, 1 mM MgCl 2 , 0.3% (w/v) C 12 E 10 , 10 mM ␤-mercaptoethanol, and 5 M GDP). Then, 17.5 g of purified ␣ i1 subunit diluted in 1 ml of ␣ subunit binding buffer was applied to the ␤␥ 2HF affinity column at a flow rate of 0.2 ml/min. The ␤␥ 2HF affinity column was washed 4 times with 4 ml of ␣ subunit binding buffer and twice with 2 ml of ␣ subunit binding buffer containing 300 mM NaCl. The column was incubated at room temperature for 15 min. The ␣ i1 subunit was eluted with 4 ϫ 0.5 ml of 20 mM Hepes, pH 8.0, 50 mM NaCl, 1 mM EDTA, 1.0% cholate, 10 mM ␤-mercaptoethanol, 10 mM MgCl 2 , and 100 M GTP␥S, also at room temperature. The column was washed twice with 2 ml of ␣ subunit binding buffer containing 300 mM NaCl, before final elution with 0.1 M glycine, pH 3.5.
Slight modifications of the above procedure were used to examine the interaction of the ␣ q subunit with the ␤ 5 ␥ 2HF dimer. The 0.5-ml ␤ 5 ␥ 2HF affinity column was prepared and washed as described in the previous section, except that the GDP concentration was increased to 50 M in the ␣ subunit binding buffer and the column buffer. Two ml of a Genapol C-100 extract of Sf9 cells expressing the ␣ q subunit was applied at a flow rate of 0.2 ml/min. The ␣ q ␤ 5 ␥ 2HF affinity column was washed twice with 1 ml of column buffer, four times with 1 ml of ␣ subunit binding buffer with 0.2% (w/v) C 12 E 10 , and finally twice with 2 ml of ␣ subunit binding buffer containing 300 mM NaCl and 0.2% (w/v) C 12 E 10 . The column was brought to room temperature for 15 min, and ␣ q was specifically eluted with 20 mM Hepes, pH 8.0, 50 mM NaCl, 1 mM EDTA, 0.05% cholate, 10 mM ␤-mercaptoethanol, 10 mM MgCl 2 , 10 mM NaF, 30 M AlCl 3 , and 50 M GDP. The cholate concentration was reduced to 0.05% in this buffer because higher cholate concentrations dissociated ␤ 5 from ␥ 2HF .
Extraction of G Proteins from Bovine Brain Membranes and Activation with GDP-AMF-Frozen bovine brains were obtained from Pel-Freeze and membranes prepared according to the method of Sternweis and Robishaw (31), with the addition of 0.2 g/ml aprotinin to all buffers. The membrane preparations were stored at Ϫ80°C. Membranes were thawed, washed once with ice-cold 20 mM Tris, pH 8.0, 1 mM EDTA, 100 mM NaCl, and pelleted at 40,000 ϫ g for 30 min at 4°C. The washed membrane pellet (1 g of protein) was extracted for 1 h at 4°C with 200 ml of 0.5% (w/v) C 12 E 10 in 50 mM Tris, pH 8.0. The extract was clarified by centrifugation at 143,000 ϫ g for 60 min and stored at Ϫ80°C. To examine the interaction of the G proteins in the membrane extracts with the various ␤ subunits, the extracts were thawed and mixed with the different ␤␥ 2HF Genapol extracts from Sf9 cells and activated by addition of concentrated stocks to give final concentrations of 3 mM MgCl 2 , 5 mM NaF, 15 M AlCl 3 , and 2.5 M GDP. The mixture was incubated at 30°C for 45 min (32). The ␣ subunits were deactivated by addition of 500 mM EDTA to a final concentration of 20 mM EDTA and incubated for an additional 30 min at 30°C. Typically, 4 ml of the deactivated mixture was applied to 0.5 ml of anti-FLAG affinity gel equilibrated with 20 mM Hepes, 0.5% (w/v) C 12 E 10 , 5 M GDP, pH 8.0. The loaded affinity column was washed with 5 ml of equilibration buffer containing 400 mM NaCl and eluted with 1.0 ml of 0.1 M glycine, pH 3.5, as described above.
Silver Staining, Immunoblotting, and Quantitation-Samples were prepared for electrophoresis, loaded on 0.75 mm, 12% acrylamide gels, subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and the gels stained with silver according to the method of Bloom et al. (33). Purified bovine brain G i/o heterotrimer (31) was used as a standard. The protein concentrations in the gel were compared with ovalbumin concentration standards and quantitated following silver staining using a BioImage scanning densitometer and the Whole Band ® software (BioImage, Ann Arbor, MI). For Western blots, gels were transferred to nitrocellulose and immunoblotted using the following primary antibodies: an anti-G protein ␤ subunit antibody (NEN Life Science Products, catalog number 808), 1:1000 dilution; an anti-G protein ␣ subunit antibody (Calbiochem, catalog number 371737), 1:1000 dilution; and an anti-␣ q /␣ 11 antibody (Santa Cruz, catalog number sc-392), 1:100 dilution. The primary antibodies were detected using goat antirabbit IgG(Fc) alkaline phosphatase conjugate (Promega) or donkey anti-rabbit IgG F(abЈ) 2 horseradish peroxidase conjugate (Amersham Corp.). The density of the bands on autoradiographs obtained following ECL detection was also estimated using the Whole Band ® software.
Adenylyl Cyclase Assays-Recombinant baculovirus encoding a FLAG epitope-tagged rat type II adenylyl cyclase was kindly provided by Dr. Ravi Iyengar, Mount Sinai School of Medicine. Sf9 cells were infected with the cyclase baculovirus and harvested 72 h later, when viability was approximately 80%. The cell pellet was washed three times with 6.8 mM CaCl 2 , 55 mM KCl, 7.3 mM NaH 2 PO 4 , 47 mM NaCl, pH 6.2, and membranes prepared according to the procedure of Taussig et al. (34). The washed membrane pellet was resuspended in 20 mM Hepes, pH 8, 200 mM sucrose, 1 mM EDTA, 2 mM DTT, 17 g/ml phenylmethylsulfonyl fluoride, 16 g/ml N-␣-p-tosyl-L-lysine chloromethyl ketone, 16 g/ml N-tosylphenylalanyl chloromethyl ketone, 2 g/ml leupeptin, and 3 g/ml lima bean trypsin inhibitor at a final total protein concentration of 1.5 mg/ml, as determined by the method of Bradford (35). Aliquots were frozen in liquid N 2 and stored at Ϫ80°C. A 0.1% (w/v) CHAPS extract of Sf9 cells overexpressing ␣ s (see above) was activated with 100 M GTP␥S in 5 mM MgSO 4 , 1 mM DTT, and 1 mM EDTA, pH 8, for 30 min at 30°C (34). Excess GTP␥S was removed by centrifugation through P6 resin (Bio-Rad) equilibrated in 50 mM Hepes, pH 8, 150 mM NaCl, 5 mM MgSO 4 , 1 mM DTT, 1 mM EDTA, 0.1% CHAPS, as described previously (13). The first elution fraction, containing activated ␣ s , was held on ice. Reaction tubes containing a total of 25 l of type II cyclase membranes (12 g of protein/assay tube), activated ␣ s , ␤␥, and/or buffer were prepared at room temperature. The reaction was begun by addition of 75 l of reaction mix pre-equilibrated at 30°C. The standard reaction mixture contained 25 mM Hepes, pH 8, 10 mM phosphocreatine, 10 units/ml creatine phosphokinase, 0.4 mM 3-isobutyl-1-methylxanthine, 10 mM MgSO 4 , 0.5 mM ATP, and 0.1 mg/ml bovine serum albumin. Reactions were carried out for 7 min at 30°C. Cyclic AMP production was stopped by the addition of 1.0 ml of 0.11 N HCl and cyclic AMP quantified by radioimmunoassay (36).
Expression of Results-Experiments presented under "Results" are representative of three or more similar experiments.
Materials-All reagents used in the culture of Sf9 cells and for the expression and purification of G protein ␣ and ␤␥ subunits have been described in detail (26,30). The baculovirus transfer vector, pVL1393, was purchased from Invitrogen; the BaculoGold ® kit was from PharMingen; 10% Genapol C-100 and the anti-␣ common subunit antibody were from Calbiochem; Prime Efficiency blunt-end DNA Ligation Kit was from 5 Prime 3 3 Prime; anti-FLAG ® M2 affinity gel was from Eastman Kodak; polyoxyethylene 10 lauryl ether (C 12 E 10 ) was from Sigma; the anti-␣ q /␣ 11 antibody was from Santa Cruz; the NEN-808 anti-␤ subunit antibody was from NEN Life Science Products, and nitrocellulose was from Schleicher and Schuell. All other reagents were of the highest purity available.

RESULTS
The objective of this study was to determine if selectivity in ␣-␤␥ interactions could be observed in vitro with recombinant G protein subunits isolated from baculovirus-infected Sf9 cells. We first constructed a recombinant baculovirus encoding sequential affinity tags on the amino terminus of the ␥ 2 subunit, a hexahistidine tag followed by a FLAG epitope tag (24). When used in conjunction with an anti-FLAG antibody covalently linked to agarose beads, the FLAG epitope tag provides a convenient method for separating ␣ subunits bound to ␤␥ 2HF from ␣ subunits free in solution. Previous work has shown that addition of a hexahistidine or FLAG affinity tag to the amino terminus of the ␥ subunit does not prohibit association with the ␤ subunit (37)(38)(39) or the subsequent association of the ␤␥ dimer with ␣ subunits (37). The heterotrimeric G protein crystal structure also suggests that an extension of the amino terminus of the ␥ subunit would be unlikely to interfere with ␣-␤␥ interactions (40,41).
The silver-stained SDS-polyacrylamide gel in Fig. 1 illustrates the steps involved in the preparation of ␤␥ and ␣-␤␥ affinity columns. Sf9 cells were co-infected with recombinant baculovirus encoding for the ␤ 1 and ␥ 2HF subunits, and the recombinant ␤ 1 ␥ 2HF protein was extracted from membranes of cells harvested 48 h post-infection. Crude detergent extracts were applied to anti-FLAG M2 affinity gel columns and washed with 5-10 column volumes. The resulting product of this onestep purification is shown in lane 3 of Fig. 1. The FLAG epitope tag on the ␥ 2HF subunit is available for binding to the anti-FLAG antibody and produces a dramatic increase in purity in a single step. The presence of the hexahistidine tag and FLAG epitope results in reduced electrophoretic mobility of the ␥ 2HF subunit relative to ␥ 2 (lane 1 versus 3). Approximately 12 g of ␤ 1 ␥ 2HF were captured per ml of anti-FLAG M2 affinity gel suspension, as determined by quantitation of the eluted ␤ subunit on a silver-stained gel. In subsequent experiments FIG. 1. Establishment of function ␣-␤␥ 2HF interactions on the FLAG affinity column. Three experiments were performed using the anti-FLAG M2 affinity gel: the ␤ 1 ␥ 2HF dimer was purified from an extract of Sf9 cells; the ␣ i1 subunit was bound stoichiometrically to a ␤ 1 ␥ 2HF affinity column; and the ␣ i1 subunit was specifically eluted from a ␤ 1 ␥ 2HF affinity column. Elution fractions from each experiment were resolved on a 12% SDS-polyacrylamide gel and stained with silver. The migration position of the bovine brain ␣, ␤, and ␥ subunits are indicated on the left. Lane 1, bovine brain standard; lane 2, pass-through (PT) after the application of a ␤ 1 ␥ 2HF Genapol C-100 extract onto the anti-FLAG M2 affinity gel; lane 3, the ␤ 1 ␥ 2HF eluted with glycine from anti-FLAG M2 affinity gel containing only immobilized ␤ 1 ␥ 2HF ; lane 4, partially purified ␣ i1 subunit prior to application onto the designed to monitor ␣-␤␥ interaction, the ␤ 1 ␥ 2HF was captured as before and the resulting ␤␥ 2HF affinity column used to specifically bind partially purified ␣ i1 subunits. To determine first the amount of ␣ i1 necessary for stoichiometric binding to immobilized ␤ 1 ␥ 2HF , replicate ␤ 1 ␥ 2HF affinity columns were prepared and then varying amounts of ␣ i1 subunit applied. Stoichiometric binding was achieved at a 3-7-fold excess (w/w) of ␣ i1 over immobilized ␤ 1 ␥ 2HF (data not shown). Lane 4 shows the ␣ i1 preparation used for these experiments. Lane 5 shows the resulting ␣ i1 ␤ 1 ␥ 2HF eluted with 0.1 M glycine after an excess of ␣ i1 was applied to the ␤ 1 ␥ 2HF affinity column.
To demonstrate that the immobilized ␤ 1 ␥ 2HF was properly folded and functional, a 0.5-ml ␣ i1 ␤ 1 ␥ 2HF affinity column, prepared identically to that shown in lane 5, was treated with 100 M GTP␥S. The ␣ i1 eluted specifically, in a volume of 0.5 ml, as shown in lane 6. Lane 7 shows the elution of the remaining ␤ 1 ␥ 2HF dimer with glycine. Thus, the ␣ i1 subunit can be dissociated from the ␤ 1 ␥ 2HF subunit with GTP␥S treatment, analogous to activation of the native heterotrimer in solution (42). Heterotrimeric G proteins can also be activated with GDP-AMF resulting in dissociation of the ␤␥ dimer (43). When an ␣ i1 ␤ 1 ␥ 2HF affinity column similar to that described above was activated with GDP-AMF, the ␣ i1 subunit was specifically eluted (data not shown).
To test the functionality of the ␣ i1 ␤ 1 ␥ 2HF interaction in another way, we subjected a detergent extract of ␤ 1 ␥ 2HF to our normal ␤␥ purification strategy, DEAE ion exchange chromatography followed by ␣ i1 -agarose affinity chromatography (30). The ␣-agarose affinity chromatography exploits the ability of GDP-AMF to dissociate the ␤␥ subunit from the ␣ subunit. Fig.   2A shows a Western blot, developed with an anti-␤ antibody, of the DEAE pool (PL) applied to the ␣ i1 column and the ␣ i1 column pass-through (PT). Comparison of lanes 1 and 2 shows a typical result for ␤ 1 ␥ 2 . A very high proportion of the ␤ 1 ␥ 2 present in the DEAE pool binds to the ␣ i1 -agarose. Lanes 3 and 4 show a very similar result obtained when a DEAE pool containing ␤ 1 ␥ 2HF was applied to the ␣ i1 -agarose. This observation is consistent with the result obtained with immobilized ␤ 1 ␥ 2HF and ␣ i1 free in solution, as described above. To obtain further evidence of functional ␣ i1 ␤ 1 ␥ 2HF interaction, we treated the ␤ 1 ␥ 2HF -loaded ␣ i1 -agarose column with GDP-AMF. Fig. 2B shows a silver-stained SDS-polyacrylamide gel of the ␣ i1 -agarose column pass-through (lane 2), wash fractions (lanes 3-5), and subsequent elution of ␤ 1 ␥ 2HF by treatment with GDP-AMF (lanes 6 -8). Thus, ␤ 1 ␥ 2HF binds tightly to immobilized ␣ i1 and elutes upon activation of ␣ i1 with GDP-AMF.
We next tested the ability of ␤ 1 ␥ 2HF to stimulate one effector, type II adenylyl cyclase. It is known that ␤ 1 ␥ 2 is a potent activator of type II cyclase in the presence of an ␣ s subunit activated with GTP␥S (44). The data in Table I compare the stimulation of type II cyclase by ␤ 1 ␥ 2 and ␤ 1 ␥ 2HF in the concentration range 0 -100 nM. At 100 nM, the ␤ 1 ␥ 2HF dimer is capable of a 15-fold stimulation of cyclase over the effect of GTP␥S-␣ s alone. However, at each concentration tested, the ␤ 1 ␥ 2 dimer activates adenylyl cyclase to a significantly greater extent than does the ␤ 1 ␥ 2HF dimer. This reduced stimulation could be due to a decreased effective concentration of ␤ 1 ␥ 2HF relative to ␤ 1 ␥ 2 at the adenylyl cyclase-containing membrane surface, or to a specific interference between the affinity tags on the ␥ subunit's amino terminus and type II cyclase. This matter is under further investigation. We conclude that the presence of the hexahistidine and FLAG epitopes on the amino terminus of the ␥ 2 subunit does not abrogate interaction of ␤ 1 ␥ 2HF with at least one effector, type II adenylyl cyclase.
Previous work with the adenosine A1 receptor showed little difference between the ability of ␣ i1 , ␣ i2 , and ␣ i3 to support high affinity binding of agonist in the presence of ␤ 1 ␥ 2 (13,45). Since this observation implies similar affinity of the three ␣ i subunits for ␤ 1 ␥ 2 , we tested the ability of another ␣ i isoform, the ␣ i2 subunit, to bind to a ␤ 1 ␥ 2HF affinity column. Two ml of a crude detergent extract of Sf9 cells infected with recombinant baculovirus encoding the ␣ i2 subunit was applied to a 0.5-ml ␤ 1 ␥ 2HF affinity column as described above, washed extensively, and the bound ␣ i2 and ␤ 1 ␥ 2HF eluted with glycine. A silver-stained polyacrylamide gel of the product is shown in Fig. 3A, lane 2. Thus, the immobilized ␤ 1 ␥ 2HF was also able to bind ␣ i2 , and a 2-ml volume of crude extract containing ␣ i2 subunit was FIG. 2. The ␤ 1 ␥ 2HF dimer associates with an ␣ i1 -agarose affinity column. A, extracts of Sf9 cells overexpressing ␤ 1 ␥ 2 and ␤ 1 ␥ 2HF dimers were partially purified on a DEAE ion exchange column, and the ␤ 1 ␥ 2 and ␤ 1 ␥ 2HF dimers were applied to ␣ i1 -agarose affinity columns. Aliquots of the pooled DEAE fractions (PL) and ␣-column pass-throughs (PT) were resolved on a 12% SDS-acrylamide gel and transferred to nitrocellulose. Lanes 1-4 were probed with an anti-␤-common primary antibody detected using a horseradish peroxidase-conjugated secondary antibody. The migration position of the ␤ 1 subunit is indicated on the left; the ␤␥ combinations are indicated above the appropriate lanes. B, purification of the ␤ 1 ␥ 2HF dimer on an ␣ i1 -agarose column. The ␤ 1 ␥ 2HF was specifically eluted from the ␣ i1 -agarose support with GDP-AMF. Proteins in the pass-through (PT), washes (W1-W3), and eluates (E1-E3) were resolved on a 12% SDS-acrylamide gel and stained with silver. Lane 1, bovine brain standard; lane 2, pass-through (PT) after application of the ␤ 1 ␥ 2HF DEAE pool onto the ␣ i1 -agarose column; lanes 3-5, wash fractions before application of GDP-AMF; lanes 6 -8, the ␤ 1 ␥ 2HF dimer eluted from ␣ i1 -agarose column by treatment with GDP-AMF. The migration positions of the bovine brain ␣, ␤, and ␥ subunits are indicated on the left. Migration positions of the ␤ 1 and ␥ 2HF subunits are indicated on the right.

TABLE I
Stimulation of type II adenylyl cyclase by native and affinity tagged ␤␥ subunits Sf9 cells were infected with a recombinant baculovirus encoding for type II adenylyl cyclase, membranes prepared, the membranes stimulated with GTP␥S-␣ s , and the indicated concentration of ␤␥ subunit for 7 min, and the cyclic AMP produced measured using a radioimmunoassay. The ␤␥ subunits were purified on a DEAE and ␣ i1 affinity column. The basal rate of cAMP production without GTP␥S-␣ s was 1.0 pmol/ml/min. See "Experimental Procedures" for details. The data are averages of 2-3 duplicate determinations. Having demonstrated functional activity for ␤ 1 ␥ 2HF and its ability to bind both purified ␣ i1 and a crude detergent extract of the ␣ i2 subunit, we tested the ability of other ␤␥ 2HF dimers to bind ␣ i2 . The ␤ 1 -, ␤ 2 -, and ␤ 5 ␥ 2HF columns were first constructed by application of appropriate crude cell extracts to anti-FLAG M2 affinity gel. Pilot experiments were performed to ensure that equivalent amounts of ␤ 1 -, ␤ 2 -, and ␤ 5 ␥ 2HF were bound to anti-FLAG columns by applying a sufficient excess of each ␤␥ 2HF detergent extract to saturate the available FLAG binding sites. Equal volumes of a crude detergent extract of Sf9 cells infected with recombinant baculovirus encoding the ␣ i2 subunit were then applied to each ␤␥ 2HF affinity column. The columns were washed extensively to remove any nonspecifically bound ␣ i2 . Finally, the ␣ i2 ␤␥ 2HF was eluted with glycine. The elution products were analyzed by gel electrophoresis followed by silver staining and Western blotting with an anti-␣ common antibody (Fig. 3, A and B). Under these conditions, the ␤ 1 ␥ 2HF and ␤ 2 ␥ 2HF columns captured equivalent, and roughly stoichiometric, amounts of ␣ i2 (Fig. 3A, lanes 2 and 3). Since the ␤ 5 subunit co-migrates with the ␣ i2 subunit under the electrophoresis conditions employed, it is not possible to determine from the silver-stained gel whether ␤ 5 ␥ 2HF captured ␣ i2 (Fig.  3A, lanes 4 and 5). However, the Western blot in Fig. 3B demonstrates clearly that ␤ 5 ␥ 2HF bound little, if any, ␣ i2 under conditions where ␤ 1 ␥ 2HF and ␤ 2 ␥ 2HF bound amounts of ␣ i2 easily detectable with the same primary antibody (compare lanes 2 and 3 with 4).
To determine if ␤ 5 ␥ 2HF was unable to bind ␣ i2 due to steric constraints of immobilization, we tested the ability of ␤ 5 ␥ 2HF free in solution to bind to an ␣ i1 -agarose column, an experiment analogous to the ␤ 1 ␥ 2HF -␣ i1 -agarose system illustrated in Fig.   2. The concentration of ␤ 5 ␥ 2HF in the DEAE pool applied to the ␣ i1 -agarose column was compared with the concentration of ␤ 5 ␥ 2HF in the column pass-through. Both the DEAE pool and the ␣ i1 column pass-through gave similar intensity when developed with an anti-␤ subunit antibody (data not shown), indicating little or no binding. Furthermore, no ␤ 5 ␥ 2HF product was detected on silver-stained polyacrylamide gels after treatment of the ␣ i1 -agarose with GDP-AMF. Therefore, the low affinity of ␤ 5 ␥ 2HF for ␣ i1 is not due to immobilization of the ␤␥ 2HF .
We then selected representatives of three ␣ subfamilies, ␣ i/o , ␣ s , and ␣ q (46), to investigate the ability of ␤ 5 to interact with other ␣ subunits. Crude detergent extracts of Sf9 cells infected with recombinant baculovirus encoding the appropriate ␣ subunit isoform were applied to three ␤␥ 2HF affinity columns constructed as before. After extensive washing, the specifically bound ␣ subunits were eluted, along with their ␤␥ 2HF counterparts, by treatment with 0.1 M glycine. The resulting products were analyzed by Western blotting with either an anti-␣ common antibody in the case of ␣ i2 , ␣ o , and ␣ s or with an anti-␣ q specific antibody for ␣ q . The results are shown in Fig. 4. As observed earlier using ␣ i2 (Fig. 3), ␤ 1 ␥ 2HF and ␤ 2 ␥ 2HF bind easily detectable amounts of ␣ i2 under conditions where ␤ 5 ␥ 2HF does not (Fig. 4, lanes 1-3). This pattern is repeated with respect to binding ␣ o and ␣ s (lanes 4 -9). However, when detergent extracts containing recombinant ␣ q were applied to each ␤␥ 2HF column, ␤ 1 ␥ 2HF , ␤ 2 ␥ 2HF , and ␤ 5 ␥ 2HF all bound ␣ q equally (lanes 10 -12). Thus, ␤ 5 ␥ 2HF appears to bind the ␣ q subunit selectively. To determine if the ␤ 5 ␥ 2HF dimer was able to interact with other members of the G q family, we have performed pilot experiments with a recombinant, avian ␣ 11 subunit (29). This protein is 96% identical in amino acid composition to the mouse ␣ 11 subunit (47) and 100% identical in the 20 amino acids shown to contact the ␤ subunit in the x-ray structure of the heterotrimer (41). Preliminary results indicate that crude detergent extracts containing ␣ 11 bind equally well to all three ␤␥ 2HF dimers (data not shown). Thus, at least one other member of the G q family binds to the ␤ 5 ␥ 2HF dimer. The interaction of the other members of the family (the G 14 -16 ␣ subunits) with the ␤ 5 subunit is currently under investigation.
The observation of selective ␣ q -␤ 5 ␥ 2HF interaction raises the FIG. 3. The ␣ i2 subunit does not associate with ␤ 5 ␥ 2HF . A, a detergent extract of the ␣ i2 subunit overexpressed in Sf9 cells was applied to ␤ 1 -, ␤ 2 -, and ␤ 5 ␥ 2HF affinity columns. After extensive washing, the ␣␤␥ 2HF heterotrimers were eluted from the anti-FLAG M2 affinity gel with 0.1 M glycine, pH 3.5, and immediately neutralized with 1 M Tris, pH 8. A ␤ 5 ␥ 2HF affinity column to which no ␣ i2 extract was applied was also eluted with 0.1 M glycine (lane 5). Proteins in each eluate were resolved on a 12% SDS-acrylamide gel and stained with silver. Lane 1, bovine brain standard; lane 2, eluate from the ␤ 1 ␥ 2HF affinity column; lane 3, eluate from the ␤ 2 ␥ 2HF affinity column; lane 4, eluate from the ␤ 5 ␥ 2HF affinity column; lane 5, ␤ 5 ␥ 2HF standard. B, immunoblot of the samples in A probed with an anti-␣-common primary antibody. The primary antibody was detected using an alkaline phosphatase-conjugated secondary antibody. The migration position of the ␣ i2 subunit is indicated on the left .   FIG. 4. Interaction of four ␣ subunits with three different ␤␥ 2HF affinity columns. Detergent extracts of Sf9 cells overexpressing the ␣ i2 , ␣ o , ␣ s , and ␣ q subunits were applied to ␤ 1 -, ␤ 2 -, and ␤ 5 ␥ 2HF affinity columns. After extensive washing, specifically bound ␣␤␥ 2HF heterotrimers were eluted from the anti-FLAG M2 affinity gel with 0.1 M glycine. The proteins in each eluate were resolved on a 12% SDSacrylamide gel and transferred to nitrocellulose. The ␣ subunits applied are indicated above each panel; the identity of the ␤ subunit in each ␤␥ 2HF affinity column used is indicated above each lane. Lanes 1-9 were probed with an anti-␣-common primary antibody, and lanes 10 -12 were probed with an anti-␣ q/11 primary antibody. Both primary antibodies were detected using an alkaline phosphatase-conjugated secondary antibody. The appropriate molecular weights for each ␣ subunit are indicated.
issue of relative affinities. One approach to this question is to determine the ability of other ␣ subunits to compete with ␣ q for binding to ␤ 5 ␥ 2HF . Since the ␣ q preparation was from a crude cell extract, we first selected a similar preparation of the ␣ i2 subunit for competition experiments. The ␤ 1 ␥ 2HF and ␤ 5 ␥ 2HF columns were constructed as before, and then a mixture of equal volumes of ␣ i2 and ␣ q detergent extracts were applied to each. After extensive washing, the ␣␤␥ 2HF complexes were eluted with 0.1 M glycine. The ␣ i2 /␣ q mixture applied to the columns (L) was compared with the glycine elution fractions (E) by Western blot using anti-␣ common or anti-␣ q antibodies. Lanes 1-4 of Fig. 5A show the result obtained with ␤ 1 ␥ 2HF . As expected from the individual ␣ subunit experiments, the ␤ 1 ␥ 2HF affinity column captured both ␣ i2 and ␣ q subunits (lane 2 versus 4). The ␤ 5 ␥ 2HF affinity column also bound ␣ q (lane 8) at a level roughly comparable to that bound by ␤ 1 ␥ 2HF (lane 4 versus 8). However, ␤ 5 ␥ 2HF bound no detectable ␣ i2 subunit (lane 2 versus 6).
Since all the above experiments were performed with recombinant proteins, we tested the ␣ subunit selectivity of the ␤␥ affinity columns with native ␣ subunits. As bovine brain membranes are known to contain a complex mixture of ␣ subunits, including ␣ i , ␣ o , and ␣ q (3, 48), we used an extract of bovine brain membranes as a starting material for these experiments. Crude detergent extracts of Sf9 cells infected with recombinant baculovirus encoding for the ␤ 1 ␥ 2HF or ␤ 5 ␥ 2HF dimers were mixed with brain membrane extract and the ␣ subunits activated by treatment with GDP-AMF as described under "Experimental Procedures." The mixtures were deactivated by addition of excess EDTA, resulting in the association of a fraction of the brain ␣ subunits with the recombinant ␤␥ 2HF dimers. The deactivated mixture was applied to an anti-FLAG affinity column and the ␣␤␥ 2HF complexes eluted. The proteins in the mixtures applied to the affinity columns and the glycine elution fractions were resolved on acrylamide gels, transferred to nitrocellulose, and probed with anti-␣ subunit antibodies. The ␤ 1 ␥ 2HF dimer bound ␣ subunits which gave positive signals with anti-␣ common antibodies and anti-␣ q/11 antibodies (Fig.  5B, lanes 2 and 4). However, the ␤ 5 ␥ 2HF dimer only associated with ␣ subunits detected by the anti-␣ q/11 antibody (lane 6 versus 8), in agreement with the selectivity observed with recombinant ␣ subunits (Fig. 5A). Interestingly, the ␤ 5 ␥ 2HF dimer binds a clearly resolved doublet from the brain extract (Fig. 5B,  lane 8). The anti-␣ q/11 antibody used in these experiments does not cross-react with ␣ i/o subunits, and therefore this doublet is most likely ␣ q and/or ␣ 11 . Thus, when the ␤ 5 ␥ 2HF dimer was presented with a complex mixture of native heterotrimeric G proteins, it selectively bound the ␣ q/11 subunits. Moreover, it did not appear to interact with the ␣ i or ␣ o subunits which are present at high concentrations in brain membranes.
Because the ␣ q , ␣ i2 , and bovine brain preparations are all crude detergent extracts, it is not possible to estimate the molar ratio of competing ␣ subunit to ␣ q subunit applied to the immobilized ␤␥ 2HF . To address this issue in part, we employed the purified ␣ i1 preparation described previously. The ␣ i1 subunit in this preparation represents approximately 95% of the intensity on a silver-stained gel. Increasing amounts of this ␣ i1 stock were diluted into a fixed, larger volume of ␣ q crude extract. Based on the amount of ␤ 5 subunit immobilized on the column as estimated from silver-stained gels, a 3-100-fold excess of ␣ i1 was added to the ␣ q extract. These mixtures were applied to immobilized ␤ 5 ␥ 2HF , washed, and eluted with 0.1 M glycine. The loading mixture, the last wash, and the elution fractions were examined by Western blot using an anti-␣ q/11 antibody (Fig. 6A) and an anti-␣ common antibody (Fig. 6B). Even at the largest excess of ␣ i1 over the immobilized ␤ 5 ␥ 2HF , there was no detectable competition by ␣ i1 for ␣ q binding to ␤ 5 ␥ 2HF . The ECL signal representing bound ␣ q in Fig. 6A was quantitated on a scanning densitometer. Fig. 6C shows a plot of this integrated intensity versus excess ␣ i1 present. Note that there is no apparent diminution of ␣ q binding at ratios of ␣ i1 to ␤ 5 ␥ 2HF far in excess of the ratio required for stoichiometric binding of ␣ i1 by ␤ 1 ␥ 2HF (about 3:1).
To demonstrate that the ␣ q ␤ 5 ␥ 2HF interaction was functional, we constructed an ␣ q ␤ 5 ␥ 2HF affinity column as described in Fig. 4 and treated the immobilized heterotrimer with GDP-AMF to activate and thereby dissociate ␣ q . Fig. 7A presents a silver-stained gel of the GDP-AMF elution product (E1-E3-AMF, lanes 4 -6). Because ␣ q and ␤ 5 co-migrate under these electrophoresis conditions, we verified the identity of the GDP-AMF and glycine elution products by Western blot with an anti-␣ q/11 antibody (Fig. 7B). Comparison of lanes 4 and 7 in Fig. 7B shows that the majority of the ␣ q bound to ␤ 5 ␥ 2HF eluted specifically with GDP-AMF. Thus, the ␣ q subunit is associating with the immobilized ␤ 5 ␥ 2HF in a manner that permits the ␣ q subunit to be activated and to dissociate from the ␤ 5 ␥ 2HF .
FIG. 5. The ␤ 5 ␥ 2HF dimer selectively associates with the ␣ q subunit in the presence of recombinant and native G proteins. A, a mixture of ␣ q and ␣ i2 subunits overexpressed in Sf9 cells was prepared as described under "Experimental Procedures" and applied to ␤ 1 ␥ 2HF and ␤ 5 ␥ 2HF affinity columns. After extensive washing, specifically bound ␣␤␥ 2HF heterotrimers were eluted from the anti-FLAG M2 resin with 0.1 M glycine. The proteins in the load (L) and eluate (E) were resolved on a 12% SDS-acrylamide gel and transferred to nitrocellulose. B, the G proteins in an extract of bovine brain membranes were activated with GDP-AMF, mixed with equal aliquots of Sf9 cell extracts expressing the ␤ 1 ␥ 2HF or the ␤ 5 ␥ 2HF dimers, incubated for 45 min, and quenched with EDTA as described under "Experimental Procedures." Each mixture was applied to a separate anti-FLAG M2 affinity gel column. After extensive washing, specifically bound ␣␤␥ 2HF heterotrimers were eluted with 0.1 M glycine, pH 3.5. The proteins in the load (L) and eluate (E) were resolved on a 12% SDS-acrylamide gel and transferred to nitrocellulose. In both A and B, lanes 1, 2, 5, and 6 were probed with an anti-␣-common primary antibody. Lanes 3, 4, 7, and 8 were probed with an anti-␣ q/11 primary antibody. Both primary antibodies were detected using an alkaline phosphatase-conjugated secondary antibody. The migration positions of the ␣ i and ␣ q subunits are indicated on the left.

DISCUSSION
The data presented in this report provide clear evidence that the diversity of the subunits in the G protein heterotrimer can have important functional consequences for the interaction of certain ␣ and ␤ subunits. Although all the ␣ subunits examined interact with the ␤ 1 or ␤ 2 subunit, the structurally different ␤ 5 subunit interacts selectively with the ␣ q subunit and the nearly identical ␣ 11 subunit. We inspected two heterotrimeric crystal structures for sites of intersubunit contact which might be responsible for the observed selectivity (40,41). These structures show that nine locations involving 16 amino acids on the ␤ 1 subunit are primarily responsible for interacting with the Switch I, Switch II, and the amino-terminal regions of the ␣ subunit. Of these 16 amino acids, only 3 are different in the ␤ 5 subunit (Leu 55 3 Gly, Tyr 59 3 Leu, and Ser 98 3 Thr, based on the ␤ 1 sequence). Although the essential residues necessary for a WD repeat (20,49) are conserved in the ␤ 5 subunit, the overall amino acid sequence of the protein is only 52% identical and 62% similar to that of the ␤ 1 subunit. Thus, there are amino acid differences in the sequences surrounding the direct ␣ subunit contact sites and other regions of dissimilarity distributed throughout the entire ␤ 5 sequence. Similarly, examination of the sequences of the ␣ subunits shows multiple differences in the amino acids contacting the ␤ subunits in the ␣ i , ␣ o , ␣ s , and ␣ q subunits, but there is only one site where the ␣ q/11 subunit is unique (41). The ␣ i , ␣ o , and ␣ s subunits have a Phe at position 195 in the beginning of the Switch II region, and the ␣ q/11 subunit share a Val at this position (41). Since there are multiple differences in sequence in both the ␣ and ␤ isoforms under consideration relative to the isoforms that have been crystallized, it is not possible to suggest a molecular basis for the selective interaction of the ␤ 5 ␥ 2HF dimer with the ␣ q subunit. However, the net effect of the various differences in ␣-␤ contacts must be substantial, as we have found that a large excess of the ␣ i1 subunit does not measurably compete with the ␣ q subunit for binding to the ␤ 5 subunit (see Fig. 6).
In evaluating the selectivity of the ␤ 5 ␥ 2HF dimer for ␣ subunits in the G q family, it is important to consider the fidelity with which Sf9 cells modify recombinant proteins. The ␣ subunits of most G proteins are modified with myristoyl and/or palmitoyl groups at their amino terminus, and the ␥ subunits are modified with a prenyl group at their carboxyl terminus (50). These modifications markedly affect the affinity of the ␣ subunits for the ␤␥ dimers (51). The available evidence suggests that the proteins used in this work are properly modified. Recombinant G i and G o ␣ subunits have been shown to be myristoylated (26), and the G q and G 11 ␣ subunits are able to activate phospholipase C-␤ equally with native proteins (52). The G s ␣ subunit produced in Sf9 cells fully activates adenylyl cyclase and is 50-fold more potent than the protein expressed in Escherichia coli but is not as potent as ␣ s purified from liver (53). The carboxyl terminus of the ␥ 2 subunit expressed in Sf9 cells appears to be properly and fully processed (54). Thus the available experimental evidence supports the hypothesis that recombinant proteins isolated from Sf9 cells are properly modified, and therefore the interactions reported here with recombinant proteins mimic those in intact cells. Most importantly, the major result of the study is considerably strengthened by the data shown in Fig. 5B demonstrating that the ␤ 5 ␥ 2HF dimer also selectively associates with the ␣ q/11 subunits in a mixture of native G proteins extracted from brain membranes.
Little is known about the biological role of the six different ␤ subunits in determining the specificity of cellular signaling.
FIG. 6. The ␤ 5 ␥ 2HF dimer selectively associates with the ␣ q subunit in the presence of an excess of partially purified ␣ i1 subunit. A, 2 ml of an extract of Sf9 cells overexpressing the ␣ q subunit was combined with increasing amounts of partially purified ␣ i1 (17-525 g, a 3-100-fold excess of ␣ i1 over ␤ 5 (w/w)) and applied to separate ␤ 5 ␥ 2HF affinity columns. After extensive washing, specifically bound ␣:␤␥ 2HF heterotrimers were eluted from the anti-FLAG M2 affinity gel with 0.1 M glycine. Proteins in the load (L), wash (W), and eluate (E1, E2, and E3) were resolved on a 12% SDS-acrylamide gel and transferred to nitrocellulose. Samples were probed with an anti-␣ q/11 primary antibody and detected using a horseradish peroxidase-conjugated secondary antibody. The migration position of the ␣ q subunit is indicated on the left. B, immunoblot of the same column samples as in A but probed with an anti-␣-common antibody. The migration position of the ␣ i1 subunit is indicated on the left. Lanes 1-5 contained no ␣ i1 ; lanes 6 -10, 3 ϫ ␣ i1 ; lanes 11-15, 10 ϫ ␣ i1 ; lanes 16 -20, 30 ϫ ␣ i1 ; lanes 21-25, 100 ϫ ␣ i1 . C, a plot of the integrated intensity of the ␣ q subunit signal from fractions E1-E3 shown in A versus 3-100-fold excess of the ␣ i1 subunit over the immobilized ␤ 5 (w/w). 7. Functionality of the G q ␣ subunit-␤ 5 interaction. The ␣ q ␤ 5 ␥ 2HF affinity column was prepared as described under "Experimental Procedures." The ␣ q subunit was specifically eluted from the immobilized ␤ 5 ␥ 2HF using GDP-AMF, and the ␤ 5 ␥ 2HF was eluted from the anti-FLAG M2 affinity gel using 0.1 M glycine. Proteins in each eluate were resolved on a 12% SDS-polyacrylamide gel and stained with silver (A) or transferred to nitrocellulose and probed with anti-␣ q/11 antibody (B). A, lane 1, bovine brain standard; lane 2, pass-through (PT) after the application of the ␣ q subunit Genapol C-100 extract onto a ␤ 5 ␥ 2HF affinity column; lane 3, final wash fraction (W) before application of GDP-AMF; lanes 4 -6, the ␣ q subunit eluted from the ␣ q ␤ 5 ␥ 2HF affinity column by treatment with GDP-AMF; lane 7, subsequent elution with 0.1 M glycine of residual ␣ q subunit and ␤ 5 ␥ 2HF dimer. The migration position of the bovine brain ␣, ␤, and ␥ subunits are indicated on both the left and the right. B, corresponding Western blot. The migration position of the ␣ q subunit is indicated on the left.

FIG.
The ␤ 1 -␤ 4 subunits are widely expressed, each contain 340 amino acids and are 80 -90% identical in sequence (3). In contrast, Northern analysis of various murine tissues shows the ␤ 5 subunit to be expressed predominantly in the brain (20), but more recently ␤ 5 subunit expression has been detected in rat portal vein (55). Expression of the similar ␤ 5L subunit which has a 42-amino acid amino-terminal extension appears restricted to certain areas of the retina (21). These two ␤ subunits do appear to be localized to the membrane (21) and are thus presumed to be involved in G protein-mediated signaling in sensory and nervous tissue. The data in this report suggest that the ␤ 5 subunit (and possibly the ␤ 5L subunit) participates in signaling via ␣ q -linked receptors. Interestingly, treatment of rod outer segment membranes with GTP␥S failed to release the ␤ 5L subunit (21). Because members of the ␣ q family are slow to exchange guanine nucleotides and are more readily activated by AMF (52), this observation is consistent with our finding of a specific interaction between the ␤ 5 and ␣ q subunits.
The biological implications of the restricted tissue distribution and the divergent sequences of the two ␤ 5 subunits are not fully understood. The ␤␥ dimer has multiple roles in G proteinmediated signaling. In addition to being required for the ␣ subunit to couple to receptors (9,10,14), the dimer can regulate the activity of multiple effectors including certain isoforms of PLC-␤, K ϩ , and Ca 2ϩ channels, phosphatidylinositol 3-kinase, adenylyl cyclase, the MAP kinase pathway and can help translocate receptor kinases to the plasma membrane (4). The functional roles of the two ␤ 5 subunits have not been fully explored, but they have been demonstrated to form functional dimers with the ␥ 2 , ␥ 3 , ␥ 4 , ␥ 5 , and ␥ 7 subunits (20,21). Analysis of the interaction of the ␤ and ␥ subunits using the yeast two-hybrid technique also shows an interaction between the ␤ 5 subunit and multiple ␥ subunits (56). Moreover, the ␤ 5 ␥ 2 and ␤ 5L ␥ 2 dimers markedly increase inositol phosphate breakdown in COS-7 cells transfected with the cDNAs for either ␤ 5 subunit, the ␥ 2 subunit, and PLC-␤ 2 (20 -22). Although the ␤ 5 ␥ 2 dimer can activate PLC-␤ 2 in transfected COS cells, it does not activate the MAP kinase or JNK kinase pathways in these cells (22). In contrast, transfection of the ␤ 1 ␥ 2 dimer is able to activate both PLC-␤ and the kinases (22,57,58). Our preliminary experiments show that the ␤ 5 ␥ 2HF dimer is not able to activate type II adenylyl cyclase. Thus, the ␤ 5 subunit (and possibly the ␤ 5L subunit) may not interact with certain important effectors.
The data described above combined with the data in this report suggest a number of possibilities for the biological role of the ␤ 5 subunits in signaling. First, heterotrimers containing the ␤ 5 subunit are most likely to couple to the ␣ q subunit, and thus only ␣ q -linked receptors may generate a ␤ 5 ␥ dimer to regulate effectors. The ability of other members of the G q family to couple to the ␤ 5 subunit needs to be explored. Second, ␤␥ dimers containing the ␤ 5 subunit may only be capable of interacting with a subset of the effectors regulated by other ␤␥ dimers. In the retina, the ␣ q -linked pathways have been assumed to play a minor role in visual signal transduction (59), but recent studies of mouse retina using immunological techniques have demonstrated the presence of the ␣ 11 subunit and PLC-␤ 4 (60). Thus, a function for this signaling pathway may emerge. A wide variety of ␣ q -linked receptors exist in neural tissue (61). One interesting pathway regulated by m1 or bradykinin receptors via the ␣ q subunit involves inhibition of M-type potassium currents (62,63). The known ability of the ␤␥ dimer to regulate K ϩ and Ca 2ϩ channels via multiple mechanisms (4, 61, 64) suggests interesting potential roles for dimers containing the ␤ 5 subunit in the regulation of ion channel activity. As multiple G protein-mediated signals are often in-tegrated by a single neuron (61,64), selective inputs by different ␤␥ dimers may allow distinct cellular responses. The observation that the ␤ 5 ␥ 2 dimer does not appear to activate the MAP kinase pathway (22) reinforces this possibility and indicates that dimers containing the ␤ 5 subunit may regulate a limited range of effectors. Thus, there may be an advantage to a more restricted ␤␥ signal in retina and neurons. Since recombinant ␤␥ dimers of defined composition have not been tested against all the known effectors regulated in this manner, it will be important to determine which effectors are regulated by dimers containing the ␤ 5 subunit. This information may help explain the restricted tissue distribution of these proteins.
In summary, the data in this report provide partial understanding for the large diversity of the proteins comprising the G protein heterotrimer. The finding that the ␤ 5 subunit interacts selectively with the ␣ q subunit suggests that it will be important to examine this issue in a number of signaling systems using recombinant proteins.