Sites for Gα Binding on the G Protein β Subunit Overlap with Sites for Regulation of Phospholipase Cβ and Adenylyl Cyclase*

Heterotrimeric G proteins, composed of α and βγ subunits, forward signals from transmembrane receptors to intracellular effector enzymes and ion channels. Free βγ activates downstream targets, but its action is terminated by association with GDP-liganded α subunits. Because α can inhibit activation of many effectors by βγ, it is likely that the α subunit binding surfaces on βγ overlap the surfaces necessary for effector activation. To test this hypothesis, we mutated residues on β shown to contact α in the recently published crystal structures of the αβγ heterotrimer (Wall, M. A., Coleman, D. E., Lee, E., Iniguez-Lluhi, J. A., Posner, B. A., Gilman, A. G., and Sprang, S. R. (1995) Cell 83, 1047–1058; Lambright, D. G., Sondek, J., Bohm, A., Skiba, N. P., Hamm, H. E., and Sigler, P. B. (1996) Nature 379, 311–319.). The α subunit binds to the flat, top surface of the toroidal β subunit and also extends a helix along the side of the β subunit at blade 1. We mutated four residues on the top surface of β (Hβ1[L117A], Hβ1[D228R], Hβ1[D246S], and Hβ1[W332A]) and two residues on the side of β that contacts α (Hβ1[N88A/K89A]). Each of the mutant proteins was able to form βγ dimers, but they differed in their ability to bind α and to activate phospholipase C β2 (PLCβ2), PLCβ3, and adenylyl cyclase II. Mutation of residues along the side of the torus at blade 1 diminish affinity for α but do not prevent activation of any of the effectors. Mutations on the α binding surface differentially affected PLCβ2, PLCβ3, and adenylyl cyclase II. Residues that affect PLCβ and adenylyl cyclase II activity are found on opposite sides of the central tunnel, suggesting that PLC and adenylyl cyclase, like the α subunit, make many contacts on the top surface. None of the mutations affected the ability of βγ to inhibit adenylyl cyclase I. We conclude that α, PLCβ2, PLCβ3, and adenylyl cyclase II share an interaction on the top surface of β. The importance of individual residues is different for α binding and for effector activation and differs even between closely related isoforms of the same effector.

Heterotrimeric G proteins composed of ␣ and ␤␥ subunits forward signals from transmembrane receptors to intracellular effector enzymes and ion channels. Activation of the G protein through the receptor causes dissociation of ␣ from ␤␥. Each of the subunits is then able to regulate downstream targets. All known effectors are regulated only by the dissociated ␣ or ␤␥ subunits and not by the ␣␤␥ heterotrimer. When GTP bound to ␣ is cleaved to GDP, the subunits reassociate. Reassociation with ␤␥ is not obligatory for deactivation of ␣ because the conformational change in ␣ that accompanies hydrolysis of GTP to GDP contributes to the termination of ␣ signaling (for a recent review, see Ref. 1). There are at least two potential explanations for the universal finding that formation of the ␣␤␥ heterotrimer turns off any signal transmitted through ␤␥. First, there may be a major conformational change in ␤␥ induced by binding to ␣. Second, all effectors may share a part of the ␣ binding site on ␤␥, so that the inhibition of effector activation by ␣ would be primarily steric. The recent publication of the crystal structure of the ␣␤␥ heterotrimer (2, 3) and of the isolated ␤␥ subunit (4) suggests that the latter explanation is correct. The structure seems to be quite rigid, and no major conformational differences were seen between the ␤␥ subunit in a ␣␤␥ heterotrimer versus the free ␤ subunit. These observations suggest that a place to look for effector contact sites is on a surface of ␤␥ that interacts with ␣.
The ␤ subunit consists of a symmetrical seven-bladed propeller structure with four kinds of surfaces: the flat surface at the narrow end that we call the "top," the flat surface at the wide end (the "bottom"), the outer surface of the torus, and the surface that lines the tunnel through the middle of the molecule (Fig. 1). The ␣ subunit contacts ␤ in two of these regions. The first ␣/␤ interface is between the amino-terminal helix of ␣ and the first blade of the ␤ propeller. The amino terminus of ␣ has long been known to be important for the formation of heterotrimers (5)(6)(7), and the crystal structure beautifully reveals why this is so. The major points of contact along this interface include residue Lys 89 that contacts residues Leu 15 and Leu 19 on ␣ (2, 3). The second interface between ␣ and ␤␥ is made up of residues on the top surface of the ␤ torus. These residues that contact ␣ are located in turns between the short ␤ strands that make up the blades of the propeller.
In order to test the hypothesis that residues in the ␤ subunit that are important for interaction with ␣ may overlap with residues important for activation of effectors, we have analyzed the consequences of mutating residues on the surface of ␤ that interacts with the switch II region of ␣ and residues on the sides of blade 1 of ␤ that contact the amino terminus of the ␣ subunit. The positions of the mutated residues are shown in Fig. 1. We compared the ability of the mutants to activate two isoforms of phospholipase C␤ and to regulate two isoforms of adenylyl cyclase. The function of ␤␥ subunits containing these mutations were analyzed in three different expression systems: in vitro translation, transient expression in COS-7 cells, and in vitro reconstitution with proteins purified from baculovirusinfected Sf9 cells. None of the mutations interfered with the ability of ␤ subunits to form ␤␥ dimers. As expected, some, but not all, mutations affected the ability of mutant ␤␥ dimers to interact with ␣. Most importantly, the results show that the ␣ contact surface on the flat, narrow end of the propeller is important for effector activation. Moreover, the mutations did not have equal consequences for the effectors tested, nor even between closely related subtypes of the same effector.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-COS-7 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Transfection was done with LipofectAMINE according to the manufacturer's protocol using 6 -7 g of LipofectAMINE/ml of culture medium (Opti-MEM, Life Technologies, Inc.) for 5 h or overnight. The medium containing LipofectAMINE was removed, and cells were washed twice with Dulbecco's modified Eagle's medium/5% fetal bovine serum.
Mutagenesis and Plasmid Construction-Site-directed mutagenesis was done in the pAlter vector (Promega) according to the manufacturer's instructions. The cDNA of the ␤ 1 subunit was cloned into pAlter vector as described (8,9). The single-stranded ␤ 1 DNA was produced and used as a template for mutagenesis. To facilitate the transfer of mutants among different vectors, a silent mutation corresponding to amino acids 144 and 145 was introduced into ␤ 1 to create a unique KpnI site. To construct a hexahistidine-tagged ␤ 1 subunit (H␤ 1 ), 1 the initial methionine was mutated into glutamine, and at the same time, a HindIII site and a PstI site were introduced. An annealed doublestranded DNA encoding the first methionine and six histidines was synthesized and ligated between this new HindIII site and the EcoRI site from the pAlter vector. The amino acid sequence added to the amino-terminally His-tagged ␤ 1 subunit is MSHHHHHHGSLLQ. For expression in COS-7 cells, the EcoRV-KpnI fragment of H␤ 1 was transferred to the pCDNA3 (Invitrogen) ␤ 1 vector using the blunted HindIII and the KpnI sites. The H6␤ 1 in pAlter was used as a template for the creation of the mutated subunit N88A/K89A (AACAAG/GCCGCG); L117A (CTG/GCG); D228R (GAG/AGA); D246S (GAC/AGC); and W322A (TGG/GCG). The mutation of H6␤ 1 was confirmed by doublestranded sequencing. The mutated part of the ␤ 1 subunit was transferred into pCDNA3 containing H6␤ 1 , using the HindIII and KpnI sites, or KpnI and BamHI as appropriate. PLC␤ 3 (obtained from S. G. Rhee) was transferred to the pCDNA3 vector between the EcoRV and XhoI sites. The EcoRV site was abolished. PLC␤2 was used in the pMT 2 vector as obtained from M. Simon. The construction of HA␥ 2 (␥ 2 tagged at the amino terminus with hemagglutinin epitope) was described in Ref. 10.
Transient Expression in COS-7 Cells and PLC Assay in COS-7 Cells-COS-7 cells (1.5-3 ϫ 10 5 /well) in six-well plates were transfected with 1.8 g of DNA and 10 g of LipofectAMINE in Opti-MEM (Life Technologies, Inc.) per well. Forty to 48 h after transfection, cells were starved for 1 h in methionine-and cysteine-deficient RPMI 1640 medium containing 5% dialyzed fetal bovine serum for 30 min and then labeled with 100 Ci/ml Trans[ 35 S]-Label (ICN) for 3 h. Labeled cells were lysed at 4°C in 0.7 ml of HMSDET buffer. All further steps were at 4°C. The lysates were incubated with 10 -15 l of packed protein A-Sepharose for 20 -30 min and centrifuged at 14,000 rpm in a microcentrifuge for 10 min. The supernatants were incubated 1 h or overnight with 0.5-2 l of 12CA5 monoclonal antibody (Babco) against the hemagglutinin epitope on the ␥ 2 subunit and then incubated with 40 l of protein A-Sepharose (50% v/v) for 30 min and centrifuged for 30 s. The pellets were washed twice with HMSDET and once with phosphate-buffered saline and then buffer for ADP-ribosylation (100 mM Tris-HCl, pH 7.6, 2 mM MgCl 2 , 1 mM EDTA, and 10 mM DTT). ADP-ribosylation was carried out in a 40-l volume containing 1 mM ATP, 1 mM NADP, 10 M NAD, 100 M GTP, 7.5 mM thymidine, 60 mM Tris, pH 7.6, 1.2 mM MgCl 2 , 0.6 mM EDTA, 6 mM DTT, 0.5 Ci [ 32 P]NAD, and 0.25 g of activated pertussis toxin. The reaction proceeded at 37°C for 30 min. The final products were analyzed by 11% SDS-PAGE. For exposure of the 32 P signal without a contribution from 35 S, a piece of black film was placed between the gel and the film to be exposed.
[ 3 H]Inositol phosphate formation was measured by modifications of the methods described in Refs. 11 and 12. COS-7 cells (0.5-1.5 ϫ 10 5 /well) in 12-well plates were transfected with PLC␤ 2 (pMT 2 vector) or PLC␤ 3 (pcDNA 3 ) and wild-type ␤ 1 , histidine-tagged ␤ 1 (H␤ 1 ), or histidine-tagged ␤ 1 mutants. A mixture of 0.8 -0.9 g of DNA and 3.5 g of LipofectAMINE in 0.5 ml of Opti-MEM (Life Technologies, Inc.) was added to the cells. The day after transfection, cells were incubated with 2 Ci/ml myo- [2-3 H]inositol in inositol-deficient Dulbecco's modified Eagle's medium with 4% fetal bovine serum. Fifteen min later, LiCl 2 (final concentration, 10 M) was added to each well, and cells were incubated overnight at 37°C. The cells were extracted twice with 0.5 ml of 20 mM formic acid. The extracts were combined, neutralized to pH 7.5 with 30 mM ammonium hydroxide, and loaded on 0.5 ml AG1-X8 anion exchange columns. Prior to use, the columns were washed with 2 ml of 1 M NaOH and 2 ml of 1 M formic acid and equilibrated with water to neutrality. The columns were washed with 10 bed volumes of water and 10 bed volumes of 5 mM borax and 60 mM sodium formate. The inositol phosphates were eluted with 10 bed volumes of 1 M ammonium formate and 0.1 M formic acid. A 2-ml aliquot of the eluates was counted in a scintillation counter.
In Vitro Translation, Immunoprecipitation, and Cross-linking-All subunits were transcribed and translated using the TNT-coupled reticulocyte lysate system (Promega). Typically, 1 g of plasmid DNA and 20 Ci of [ 35 S]methionine were used in a 50-l reaction. In all cases, transcription was directed by the T7 promoter. Synthesis of the desired product was routinely verified by running 5 l of the translation mixture in a small 11 or 13% polyacrylamide gel (13), followed by autoradiography with overnight exposure. Independently translated ␤ and ␥ subunits were mixed together and incubated at 37°C for 90 min to allow dimer formation. Because ␥ translation was usually more efficient, 10 -15 l of ␥ translation mixture was typically added to 50 l of ␤ translation mixture. Fifty l of the ␤␥ mixture was passed over an 8 ml AcA 34 column (Sepracor) equilibrated with HMSE plus 0.05% Lubrol PX at 4°C in order to remove DTT and to separate the ␤␥ dimers from undimerized ␤. The fractions containing ␤␥ were concentrated 5-10-fold using a Centricon-30 concentrator (Amicon).
For cross-linking, 30 l of this sample was mixed with 10 l of ␣ o (2-5 g) purified from bovine brain (14) in HMSE or 10 l of HMSE buffer alone, and the reaction was initiated by the addition of 1.6 l of freshly prepared 50 mM bismaleimide hexane (BMH) (Pierce) in Me 2 SO (8,9). In control un-cross-linked samples, 20 mM DTT was added prior to BMH. After 40 min at 4°C, DTT (20 mM) and/or Laemmli sample buffer containing 15% ␤-mercaptoethanol was added, and the samples were boiled and resolved by SDS-PAGE on 9% polyacrylamide gels (13). Dried gels were soaked in Enhance and then used for autoradiography. The radioactive bands could be visualized after 2-7 days of exposure at Ϫ70°C.
Recombinant ␣ q and wild-type ␤ 1 ␥ 2 were purified from Sf9 cells as described (17). Myristoylated ␣ i1 was purified from Escherichia coli as described (19). PLC␤ 2 was purified from Sf9 cells and kindly provided by Dr. Paul C. Sternweis (University of Texas Southwestern Medical Center). Protein was measured as described in Ref. 20.
To measure adenylyl cyclase activity, purified ␤␥ mutants were reconstituted with 10 g of membranes from Sf9 cells expressing type I or type II adenylyl cyclase for 3 min at 30°C in a final volume of 20 l. Assays were then performed as described (22) for 7 min at 30°C in a total volume of 50 l containing 4 mM MgCl 2 and 0.2% octyl ␤-Dglucopyranoside. The presence of the hexahistidine tag at the amino terminus of either ␤ 1 or ␥ 2 did not affect any of the enzymatic assays (data not shown). 1 and Their Association with ␣-To evaluate the relationship between surfaces of ␤ that bind ␣ and that activate PLC␤, we mutated four residues on the flat, top surface of ␤ and two residues on the side of blade 1 (see Fig. 1). The mutations were introduced into the background of rat ␤ 1 tagged at the amino terminus with six additional histidine residues. Addition of the hexahistidine tag was extremely important for assays in COS-7 cells because the size difference between H␤ 1 and wild-type ␤ 1 allowed us to discriminate transfected, mutated ␤ subunits from endogenous ␤ subunits. We could detect no differences in these assays between the hexahistidine ␤ 1 (H␤ 1 ) and wild-type ␤ 1 (see below). Before we could assess the ability of mutated ␤ subunits to activate PLC, it was essential to establish that they could form dimers. We expected that some, but not all, mutations would also affect the ability of ␤␥ to form heterotrimers with ␣. To evaluate these two issues, transfected COS-7 cells were labeled with Trans[ 35 S]-Label and H␤ 1 or the mutant proteins were immunoprecipitated through cotransfected HA␥ 2 (10). Because we immunoprecipitated through one subunit (HA␥ 2 ) but measure the other, this assay measured only the amount of ␤␥ dimers that accumulate, and not the total synthesis of ␤. When HA␥ 2 is transfected, it dimerizes with both wild-type endogenous ␤ and transfected H␤ 1 subunits. Therefore, both types of ␤ subunits were immunoprecipitated with the anti-HA␥ 2 antibody, but the two can be readily distinguished by their mobility in SDS-PAGE ( Fig. 2A).

Formation of ␤␥ Dimers by Mutant H␤
We also used immunoprecipitation to measure the ability of mutant H␤ 1 ␥ 2 dimers to associate with transfected ␣ i2 . Cotransfection of ␣ i2 had no reproducible effect on the amount of ␤␥ dimer formed. Because antibody to HA␥ 2 precipitates both endogenous and transfected ␤␥, it was essential to be able to subtract the amount of ␣ i2 coprecipitating with a dimer containing endogenous ␤ and HA␥ 2 from the amount coprecipitating with a dimer of transfected H␤ 1 mutants and HA␥ 2 . The amount of ␣ coprecipitated with endogenous ␤ was determined from the lysates of cells transfected with HA␥ 2 , ␣ i2 , and vector but no additional ␤. From the relative density of the ␣ and the ␤ bands (taking into account the number of methionine and cysteine residues), we calculated that 0.6 -0.8 mol of ␣ were precipitated per mol of endogenous ␤. 2 To determine whether the ␣ i2 immunoprecipitating with H␤ 1 HA␥ 2 dimers was associating correctly, we measured the ability of coimmunoprecipitated ␣ to be [ 32 P]ADP-ribosylated by pertussis toxin. Although only the ␣ subunit is ADP-ribosylated, the substrate for the toxin is the ␣␤␥ heterotrimer (14). Alternate lanes in the top panel of Fig. 2 show 35 S label only or 35 S ϩ 32 P. The bottom panel shows 32 P only. These experiments are quantitated and summarized in Fig. 2B and Table I. The results show that mutation H␤ 1 [W332A] has little effect on the ability of ␤␥ to interact with ␣, but each of the other mutations diminishes coimmunoprecipitation of ␣ through ␤␥. For each mutant, the results were the same whether we measured the amount of ␣ by 35 S or by ADP-ribosylation. This correlation suggests that there are no dramatic differences in the ability of mutant ␤␥ to support ADP-ribosylation of ␣ once a complex has formed.
Another way to assess the interaction of mutant ␤␥ dimers with ␣ is chemically to cross-link mutant ␤␥ dimers to ␣ using the cysteine-specific reagent, BMH. We have previously shown that this reagent specifically cross-links cysteine 215 of ␣ o FIG. 1. Structure of ␤␥. Space-filling model of ␤␥ seen from the surface of ␤ that interacts the ␣. The ␤ subunit is shown in yellow, and the ␥ subunit is shown in red. The residues mutated in these studies are indicated in blue, and the residues cross-linked to ␣ by BMH are shown in white. The figure was drawn from coordinates kindly provided by Dr.
Stephen Sprang, University of Texas Southwestern Medical Center (Dallas, TX). either to cysteine 204 or cysteine 271 of ␤, giving two crosslinked products (8,9). In the wild-type ␤␥, the two cross-linked products are formed approximately equally (Fig. 3). From the crystal structure, we know that the distance between the residues on ␣ o and on ␤ is very close to that of the fully extended cross-linking reagent. The ability of the reagent to reach to one or the other of the cysteines on ␤ depends on a correct orientation of ␣ with respect to ␤. The ␤␥ formed from one of the mutants on the top surface of ␤ (H␤ 1 [D246S]) gave the same two cross-linked products as wild-type in the same ratio. In ␤␥ containing each of three other mutants, the ratio of crosslinked products was different. The upper band produced by cross-linking ␣ Cys 215 to ␤ Cys 204 was decreased in ␤␥ dimers containing H␤ 1 [W332A] and H␤ 1 [L117A]. In H␤ 1 [D228R], the lower band (produced by cross-linking ␣ Cys 215 to ␤ Cys 271 ) was missing. Mutating the two residues that contact the aminoterminal ␣ helix of ␣ (H␤ 1 [N88A/K89A]) greatly diminished the affinity of ␤␥ for ␣. Indeed, ␤␥ containing this mutation produced barely detectable cross-linked products, although both bands were faintly visible. The cross-linking reaction is an irreversible reaction and is therefore able to reveal even low affinity interactions between the subunits. We explain the altered cross-linking pattern of H␤ 1 [W332A], H␤ 1 [L117A] and H␤ 1 [D228R] by suggesting that the ␣ subunit is still able to interact with the mutated ␤, but that it is tilted on its binding site. It is unlikely the changes in the cross-linking pattern are due to local effects of the mutations themselves, because no mutated residue is adjacent to the cysteine whose cross-linking it affects.
Activation of Phospholipase C␤ Isoforms by ␤ 1 Mutants-PLC␤ 2 and PLC␤ 3 are two isoforms of PLC that are activated by the ␤␥ dimer (11, 23, 24). We used two methods to compare the ability of mutant ␤␥ proteins to interact with ␣ and to activate the two isoforms of PLC. First, we cotransfected COS-7 FIG. 2. Coimmunoprecipitation and ADP-ribosylation of ␣ i2 with dimers of HA␥ 2 and H␤ 1 mutants. A, COS-7 cells in 6-well plates were transfected with a combination of 0.6 g of ␣ i2 , 0.6 g of HA␥ 2 , and either 0.6 g of H␤ 1 , H␤ 1 mutants or 0.6 g of HA␥ 2 , 0.6 g of ␣ i2 , and 0.6 g of vector DNA (for the left two lanes). 35 S-Labeled cell lysates were immunoprecipitated with the monoclonal antibody directed against the HA epitope on ␥ 2 . One sample of each duplicate pair was [ 32 P]ADP-ribosylated by pertussis toxin after immunoprecipitation as described under "Experimental Procedures." The top panel shows duplicate samples, the first labeled with 35 S, the next with 35 S and 32 P (in ␣ only). ADP-ribosylation slightly slows the mobility of ␣, so the 32 P band is slightly above the 35 S band. The bottom panel shows the radioautogram of 32 P only. The film was shielded from 35 S during exposure. B, the amount of ␣ i2 in the experiments of the type shown in Fig. 3 was quantitated by densitometry. H␤ 1 was defined as 1.0, and each H␤1 mutant was expressed as a fraction of H␤ 1 Ϯ S.E. (n ϭ 3). In each case, the amount of ␣ accounted for by coprecipitation with endogenous ␤␥ has been subtracted (see text). Open bars show data from experiments with 35 S only, and hatched bars show data from experiments in which 35 S was shielded and only 32 P exposed the film. b The amount of ␣ i2 immunoprecipitated was quantitated by densitometry of the 35 S autoradiogram for three separate experiments. The amount of immunoprecipitated ␣ i2 accounted for by the endogenous ␤HA␥ 2 was subtracted from the total value. c The relative phospholipase C activation was calculated as follows.
FIG. 3. Cross-linking of in vitro synthesized ␤ 1 mutants to purified brain ␣ o . In vitro translated wild-type ␤ 1 or H␤ 1 mutants were dimerized with ␥ 2 -HA and cross-linked in the presence of ␣ o as described under "Experimental Procedures." Both treated (ϩBMH) and untreated (ϪBMH (20 mM DTT added before BMH)) samples were analyzed by 9% SDS-PAGE followed by autoradiography. Cross-linked products were visualized after a 2-day exposure. The positions of the molecular mass markers are indicated at the left. Shown is a representative of three experiments for each mutant. cells with wild-type and mutant H␤ 1 subunits, HA␥ 2 and PLC␤ 2 , or PLC␤ 3 and measured the increase in inositol phosphate production. Second, we synthesized the proteins in Sf9 cells, purified them, and measured activation of PLC␤ 2 in vitro. Transfection of ␤␥ into COS-7 cells did not significantly affect basal PLC activity (probably because the ␤␥ level is elevated only in the fraction of the cells that took up the cDNA, whereas all cells contribute to the basal activity) (data not shown). Transfection of PLC␤, together with ␤ and ␥, caused a 3-fold increase in inositol phosphate production compared with transfection of PLC␤ alone (Fig. 4A). Neither ␤ nor ␥ alone increased the activity of transfected PLC␤ (data not shown). Addition of the hexahistidine tag had no effect on the activity of ␤, and HA␥ 2 was as effective as ␥ 2 . As was previously shown by Katz et al. (11), we found that ␤␥ dimers that contain a mutant ␥ that cannot be prenylated at the carboxyl terminus do not activate PLC␤ in the COS-7 cells (data not shown). Finally, activation of PLC␤ by ␤␥ was blocked by cotransfection of ␣ i2 (see below). Taken together, these controls, together with published in vivo and in vitro studies (11,(25)(26)(27)(28)(29), support the interpretation that the elevation of inositol phosphates that we measured reflects activation of PLC␤ by ␤␥. This interpretation is further strengthened by agreement of the data obtained in transfected cells with those obtained with purified proteins.
As shown in Table I 3 almost as well as wildtype, but was blunted in its ability to activate PLC␤ 2 . In contrast, H␤ 1 [L117A] was fully active with respect to PLC␤ 2 but inactive with respect to PLC␤ 3 . These results are consistent with a model in which the interaction interfaces of ␤␥ with ␣ or ␤␥ with different effectors overlap, but the importance of specific residues for each function is different.
Cotransfection of ␣ i2 blocks PLC␤ activation by ␤␥ (Fig. 4A), even when the ␤␥ has a diminished affinity for ␣ in solution (for example, H␤ 1 [L117A] or H␤ 1 [D246S]). Of the mutations we made, changes in residues on the side of the ␤ torus (H␤ 1 [N88A/K89A]) had the most profound effect on the affinity for ␣ i2 in solution, as measured by immunoprecipitation. Nevertheless, in cells, expression of ␣ i2 blocked activation of PLC␤ 2 by H␤ 1 [N88A/K89A] with a dose-response curve similar to its inhibition of wild-type ␤␥ (Fig. 4B). Analysis of ␣ i2 expression by Western blot at each cDNA concentration showed that the ␣ i2 levels rose approximately equally in cotransfections with PLC␤ 2 and wild-type or mutant ␤ (data not shown).
To confirm the results in COS-7 cells, recombinant ␤␥ mutant proteins were synthesized in Sf9 cells and purified; SDS-PAGE analysis of these samples is shown in Fig. 5. All five ␤ 1 mutants were purified as complexes with the ␥ 2 subunit. In these studies, we used a single ␤ 1 [K89A] rather than a double H␤ 1 [N88A/K89A] mutant at the side of the ␤ 1 torus. All of the complexes supported ADP-ribosylation of ␣ i1 by pertussis toxin, although the potency of H␤ 1 [K89A]␥ 2 was about half that of the wild-type protein, presumably reflecting the lower affinity of this mutant for ␣ i , consistent with the properties of the double mutant (data not shown).
Activation of PLC␤ 2 by purified wild-type and mutant ␤␥ complexes is shown in Fig. 6. Consistent with results in COS-7 cells, H␤ 1 [K89A]␥ 2 and H␤ 1 [L117A]␥ 2 were approximately equal to wild-type ␤␥ in activating PLC␤ 2 , but the other three mutations on the top surface of ␤ 1 were severely blunted in their ability to activate PLC␤ 2 . Although there are quantitative differences in the degree of impairment of D228R, D246S, and W332A in the two experimental systems, the conclusion that mutating each of the three residues diminishes PLC␤ 2 activation is consistent in both. In analyzing a large number of mutations at various sites in ␤, we have sometimes observed differences in the ability of mutant proteins to fold correctly, FIG. 4. Cotransfection of ␣ i2 inhibits PLC␤ activation by wildtype and mutant ␤␥. A, inhibition of PLC␤ 3 by coexpression of ␣ i2 . COS-7 cells in 12-well plates were transfected with the indicated cDNAs in the following amounts: 0.2 g of H␤ 1 or mutant H␤ 1 , 0.2 g of ␥ 2 , and 0.2 g of ␣ i2 . In all cases, vector DNA was added to give a final DNA concentration of 0.8 g/well. PLC␤3 activation was measured as described under "Experimental Procedures." The data shown are representative of two experiments. The error bars indicate the range of duplicate assays. Where there are no error bars, the range was too small to display. Filled columns, without ␣ i2 ; open columns, with ␣ i2 . B, cotransfection of ␣ i2 inhibits PLC␤ 2 activation by H␤ 1 [N88A/K89A]. COS-7 cells were transfected and assayed as described under "Experimental Procedures." The x axis represents the logarithm of ng of ␣ i2 cDNA transfected, and the y axis shows activation of PLC␤ by H␤ 1 or H␤ 1 mutants. Shown is a representative of three experiments. Activation of PLC␤ 2 by H␤ 1 ␥ 2 or H␤ 1 [N88A/K89A]␥ 2 was taken as the 100% value for each. OE, H␤ 1 [N88A/K89A]␥ 2 ; f, H␤ 1 ␥ 2 .
depending on the expression system, with the most native state achieved when the protein is made in mammalian cells (28). It is possible that the final conformation of the mutant ␤ subunit is slightly different when they are made in mammalian COS-7 cells as opposed to insect cells.
Activation and Inhibition of Adenylyl Cyclase by ␤ 1 Mutants-The effects of mutant ␤ 1 ␥ 2 complexes on adenylyl cyclase activities are shown in Fig. 7. ␤␥ activates type II adenylyl cyclase in the presence of ␣ s , but it inhibits type I adenylyl cyclase (29). The apparent affinities of H␤ 1 [D246S]␥ 2 and ␤ 1 [W332A]␥ 2 for type II adenylyl cyclase are clearly diminished; we were unable to assess unequivocally their maximal capacities to activate the enzyme because of our inability to achieve higher concentrations of these proteins in the assay (Fig. 7B). Within a similar range of concentrations, H␤ 1 [D228R]␥ 2 did not activate type II adenylyl cyclase. The ␤ 1 [K89A]␥ 2 and ␤ 1 [L117A]␥ 2 mutants were indistinguishable from wild-type complex. In contrast, all five mutant ␤␥ com-plexes inhibited type I adenylyl cyclase (Fig. 7A). These inhibitory activities were lost after inactivation of the proteins at 95°C for 5 min (data not shown). The observation that H␤ 1 [D228R]␥ 2 is able to inhibit type I adenylyl cyclase (albeit with the lowest apparent potency of the group tested), whereas it is inactive in PLC␤ 2 , PLC␤ 3 , and type II adenylyl cyclase assays, confirms the conclusion, based on coimmunoprecipitation and cross-linking studies, that the protein is not grossly misfolded (Figs. 2 and 3).
The inability of any of the mutations on the top surface of ␤ to interfere with the inhibition of type I adenylyl cyclase raises the possibility that ␤␥ inhibition of type I adenylyl cyclase would not require the ␣ binding surface and would be an exception to the rule that association with ␣ blocks interaction of ␤␥ with all effectors. However, incubation of wild-type ␤␥ with GDP-␣ q interfered with both activation of type II adenylyl cyclase and inhibition of type I adenylyl cyclase ␤␥ (Fig. 8), suggesting overlap of ␣ with these interacting surfaces. The interface between ␤␥ and the two adenylyl cyclases must require different parts of the ␤ top surface. DISCUSSION The interpretation of the functional consequences of mutation introduced into a protein structure depends on demonstrating, as far as possible, that the mutation produces only a local change and not a global one. We have mutated some of the residues in ␤ known to contact ␣ (2, 3) in order to test the hypothesis that ␣ and effectors share a common surface. We have analyzed the properties of the mutant ␤ subunits in three kinds of expression systems, which allows us to evaluate different aspects of their function. None of the mutant proteins reported in this paper appeared to have global effects on ␤ structure or its ability to assemble with ␤␥. Indeed, in one assay (inhibition of type I adenylyl cyclase), all were fully active. When the same property was evaluated in more than one system (e.g. activation of PLC␤ 2 in COS-7 cells or reconstituted with pure proteins in vitro), we found general agreement in the results.
Our results show that some residues that are important for binding of ␣ by ␤␥ are also important for activation of PLC␤ 2 , PLC␤ 3 , and adenylyl cyclase II. Mutating the residues on the side of blade 1 (Asn 88 and Lys 89 ) has little effect on activation of either PLC isoform or regulation of adenylyl cyclase I and II. These results suggest that the side of the torus at blade 1 is not involved in activation of these effectors, although it is very important for ␣ binding. In contrast, mutations on the flat top surface of ␤ affect activation of both PLC isoforms and adenylyl cyclase II. The ␣ subunit, PLC␤ 2 , PLC␤ 3 , and adenylyl cyclase II appear to share an overlapping region on the top surface of ␤␥, but the relative importance of particular residues is different among them. Changing Trp 332 to Ala severely inhibits activation of both PLC enzymes, but the W332A mutation only modestly affects the ability of ␤␥ to activate adenylyl cyclase II. Mutating Leu 117 to Ala diminishes the ability to activate PLC␤ 3 but has no effect on PLC␤ 2 or adenylyl cyclase II. Mutating Asp 246 to Ser has the opposite phenotype: the ability to activate PLC␤ 2 is substantially more affected than the ability to activate PLC␤ 3 . PLC␤ 2 and PLC␤ 3 are very similar to each other overall, but they are quite different in regions that have been found to be important for activation by ␤␥ (25,30). Therefore, it makes sense that mutation of residues on ␤␥ would have different consequences for the two enzymes. Mutating Asp 228 to Arg produces a molecule that binds ␣ and activates PLC␤ 2 , PLC␤ 3 , and adenylyl cyclase II poorly. Nevertheless, ␤␥ containing the purified H␤ 1 [D228R] protein is able fully to inhibit adenylyl cyclase I, confirming our studies in vitro and in COS-7 cells that this mutation does not prevent folding of the mutant ␤ into ␤␥.
The observation that ␤␥ dimers containing each of the mutated ␤ subunits were able to inhibit type I adenylyl cyclase was surprising. Therefore, we wanted to be sure that ␤␥ inhibition of type I adenylyl cyclase could, indeed, be inhibited by ␣. As shown in Fig. 8, activation of type II adenylyl cyclase and inhibition of type I adenylyl cyclase follow the rule that ␣ blocks all effector regulation by ␤␥. We propose that type I adenylyl cyclase interacts with a different part of the top surface from PLC␤ 2 , PLC␤ 3 , and type II adenylyl cyclase. The footprint of ␣ on the top surface of ␤ is quite large. Steric hindrance could prevent interactions over a much greater portion of the surface than that immediately forming contact sites for ␣ on ␤.
Residues that affect PLC␤ and adenylyl cyclase II activity are found on opposite sides of the central tunnel (see Fig. 1), suggesting that PLC␤ and adenylyl cyclase II, like ␣, makes many contacts with the top surface. Overlap of the footprint of ␣, PLC or adenylyl cyclase II on ␤ explains why binding of ␣ to ␤␥ prevents activation of these effectors by ␤␥. Analysis of conserved residues on ␤ subunits from several species also suggested that the ␣ binding surface is important for effector activation (31).
The amount of ␣ coimmunoprecipitated with wild-type or mutant ␤␥ gives information about the relative importance of some residues in ␤ for binding to ␣ in solution. However, for molecules to coimmunoprecipitate requires that they have a fairly high affinity for each other. Another assay for the ability of mutant ␤␥ subunits to interact with ␣ is chemical crosslinking with BMH, an irreversible reaction. In addition to affecting the apparent affinity, some mutations of the ␣ binding surface H␤ 1 [L117A], H␤ 1 [D228R], and H␤ 1 [W332A] appear to cause the proteins to interact with ␣ abnormally. They give cross-linking patterns that differ from wild-type, suggesting that ␣ is not docking properly on the ␤␥ surface. The irreversible cross-linking reaction traps even transient associations, so that even some proteins that are poorly immunoprecipitated can still be cross-linked (H␤ 1 [D228R] and H␤ 1 [L117A]). They can also interact with ␣ in cells as shown by the ability of cotransfected ␣ i2 to block PLC␤ activation by all the mutants (Fig. 4A). Even the mutant ␤␥ most severely affected with respect to ␣ binding (H␤ 1 [N88A/K89A]) can be inhibited by cotransfected ␣ i2 with a cDNA dose-response curve similar to FIG. 8. GDP-␣ q interferes with the ability of ␤␥ to inhibit type I adenylyl cyclase or to stimulate type II adenylyl cyclase. The indicated concentration of ␤␥ was incubated with (f) or without (q) 100 nM GDP-␣ q on ice for 10 min. These samples were then reconstituted with 10 g of Sf9 cell membranes from cells expressing type I adenylyl cyclase (A) or type II adenylyl cyclase (B) in the presence of 50 nM GTP␥S-␣ s . Adenylyl cyclase activity was measured as described under "Experimental Procedures." Data shown are the average of duplicate determinations from a single experiment that is representative of two such experiments. reversal of PLC␤ 2 activation by wild-type ␤␥ (Fig. 4B). We suggest that when ␣ and the mutant ␤␥ proteins are correctly positioned in the membrane, the local concentration is high enough to allow productive interactions.
Besides ␣, two other molecules that virtually universally block the action of ␤␥ have been described. The first is the carboxyl-terminal region of the ␤-adrenergic receptor kinase (32). This region has been shown to bind ␤␥. When expressed in COS-7 cells, the ␤ARK fragment is able to inhibit activation of adenylyl cyclase and PLC by ␤␥. The ␤ARK fragment does not bind to the ␣␤␥ heterotrimer. We presume, therefore, that like PLC and type II adenylyl cyclase, it binds to the G␤ on the surface that faces ␣. The second is a 12-amino acid peptide derived from the putative ␤␥ binding site on adenylyl cyclase that blocks ␤␥ activation of adenylyl cyclase, the K ϩ channel, and PLC and may bind to the top surface of ␤␥ (33,34).
The region of ␤␥ that we have identified as important for interaction with PLC␤ and type II adenylyl cyclase is unlikely to be the only surface of ␤␥ that interacts with these enzymes, although it is probably the one crucial for regulation by ␣. Yan et al. (35) recently described interaction in the yeast two hybrid system of the first 100 residues of G␤ (including the aminoterminal coiled coil plus blade 1) with the amino terminus of the muscarinic K ϩ channel and a portion of adenylyl cyclase II. The interactions have not yet been verified by any other means, but the data are consistent with the genetic findings in S. cerevisiae that the amino-terminal region may contain binding sites for some effectors. In addition to the crystal structure of the ␣␤␥ dimer (2, 3), the structure of ␤␥ in a complex with phosducin, a potential regulator of ␤␥ function, has recently been solved (36). Like ␣, phosducin binds on the top surface of ␤ as well as to the side of the torus. It is likely that PLC and adenylyl cyclases also have important contacts along the side of the G␤ torus. Future studies will determine whether such contacts exist and where they are.