Identification of a Giα Binding Site on Type V Adenylyl Cyclase*

The stimulatory G protein α subunit Gsα binds within a cleft in adenylyl cyclase formed by the α1-α2 and α3-β4 loops of the C2 domain. The pseudosymmetry of the C1 and C2 domains of adenylyl cyclase suggests that the homologous inhibitory α subunit Giα could bind to the analogous cleft within C1. We demonstrate that myristoylated guanosine 5′-3-O-(thio)triphosphate-Giα1 forms a stable complex with the C1 (but not the C2) domain of type V adenylyl cyclase. Mutagenesis of the membrane-bound enzyme identified residues whose alteration either increased or substantially decreased the IC50 for inhibition by Giα1. These mutations suggest binding of Giα within the cleft formed by the α2 and α3 helices of C1, analogous to the Gsα binding site in C2. Adenylyl cyclase activity reconstituted by mixture of the C1 and C2 domains of type V adenylyl cyclase was also inhibited by Giα. The C1b domain of the type V enzyme contributed to affinity for Giα, but the source of C2 had little effect. Mutations in this soluble system faithfully reflected the phenotypes observed with the membrane-bound enzyme. The pseudosymmetrical structure of adenylyl cyclase permits bidirectional regulation of activity by homologous G protein α subunits.

The hormone-sensitive adenylyl cyclase system is a well studied paradigm of G protein-mediated signal transduction. All isoforms of mammalian adenylyl cyclase are stimulated by the heterotrimeric G protein G s . Appropriate agonist-bound, heptahelical receptors activate G s by catalyzing the exchange of GDP for GTP. The GTP-bound ␣ subunit of G s in turn activates adenylyl cyclase, increasing the rate of synthesis of cyclic AMP from ATP (1,2). Many other regulatory influences are also brought to bear on various adenylyl cyclases (nine isoforms are known to date), and these enzymes thus serve critical roles as integrators of these diverse inputs. Other physiological regulators of adenylyl cyclases include G i␣ , G z␣ , and G o␣ , which are inhibitors; G protein ␤␥ subunits, which can activate or inhibit; free Ca 2ϩ and Ca 2ϩ bound to calmodulin or calcineurin; and various kinases (3)(4)(5). The sites and mechanisms of interaction of these regulators with adenylyl cyclases are being defined by a variety of techniques, particularly including x-ray crystallography and mutagenesis (6 -15).
Mammalian adenylyl cyclases are pseudosymmetrical, membrane-bound enzymes. The bulk of the protein consists of two repeats of a unit that includes a hydrophobic membrane-asso-ciated domain (modeled to contain six transmembrane spans) and a cytoplasmic domain. The pseudosymmetry results from the fact that the two cytoplasmic domains (C 1 and C 2 ) each contain a region of approximately 230 amino acid residues that are roughly 40% identical (C 1a and C 2a ). The most striking interaction between G s␣ and adenylyl cyclase is the insertion of the G s␣ switch II helix into the groove formed by the ␣2Ј helix and the ␣3Ј-␤4Ј loop of the C 2a domain of adenylyl cyclase (15). A similar groove is formed by the corresponding structural elements of C 1a .
Only adenylyl cyclases types I, V, and VI are inhibited by G i␣ and G z␣ (16 -19); G o␣ inhibits the type I enzyme alone (17). These G protein ␣ subunits are both myristoylated and palmitoylated. Myristoylation of G i␣ and G o␣ increases their affinity for ␤␥ (20) and is required for the interaction of G i␣ with adenylyl cyclase (16) and its association with the plasma membrane (21,22).
The assignment of regulatory properties to individual isoforms of adenylyl cyclase resulted primarily from expression of the proteins in Sf9 cells and exposure of membranes from these cells to regulators of interest (1)(2)(3)(4). More recently, the synthesis of large amounts of the two cytoplasmic domains of adenylyl cyclase as soluble proteins in Escherichia coli and their functional reconstitution in vitro have allowed exploration of more detailed mechanistic questions (23)(24)(25) and solution of the crystal structure of the complex containing G s␣ , forskolin (a diterpene activator of adenylyl cyclases), the C 1 domain from type V adenylyl cyclase, and the C 2 domain from the type II enzyme (15). We and others (26 -30) have shown that these domains contain all of the necessary structural components for synergistic activation of adenylyl cyclase by G s␣ and forskolin as well as for inhibition by adenine nucleotide analogs called P-site analogs.
Although the chimeric constructs used in the past were invaluable for investigation of properties shared by all isoforms of adenylyl cyclase, they are not as useful to study regulatory properties that are unique to one or a subset of isoforms. We have now developed a soluble enzyme system composed of the C 1 and C 2 domains from type V adenylyl cyclase and have used these proteins, in conjunction with the full-length type V enzyme expressed in Sf9 cells, to identify sites of interaction between adenylyl cyclase and G i␣ .

MATERIALS AND METHODS
Plasmid Construction-To create VC 1 (670) with a C-terminal hexahistidine tag, VC 1 (364 -670) (23) was digested with NcoI and BamHI and ligated with the expression vector pQE60 that had been digested with NcoI and BglII. DNA encoding amino acid residues 954 -1184 of canine type V adenylyl cyclase (31) was generated with the polymerase chain reaction using oligonucleotides A (5ЈAACACCATGGTCCAGGC-CTACAACCGGCGG 3Ј) and B (5Ј ATTCAAGCTTTACTACTGAGCGG-GGGCCCA 3Ј) as primers. The fragment was cloned into H 6 pQE60 using the restriction enzymes NcoI and HindIII to generate VC 2 with an N-terminal hexa-histidine tag. Mutations within type V adenylyl cyclase were generated using the pAlter mutagenesis kit (Promega) or the QuikChange mutagenesis method (Stratagene). The desired clones were identified by sequencing. Type V adenylyl cyclase was subcloned from the pVL1392 vector (16) into the pAlter-1 mutagenesis vector (Promega) using the restriction sites for BamHI and HindIII. Singlestranded DNA was generated using R408 Helper phage. Regions surrounding the mutations were subcloned back into type V adenylyl cyclase in pVL1392 or VC 1 (670)H 6 for expression in Sf9 cells or E. coli, respectively.
Sf9 Cell Culture and Recombinant Baculoviruses-Procedures for the culture of Sf9 cells and the production, cloning, and amplification of recombinant baculoviruses have been described by Summers and Smith (32). The plasmids containing mutations in type V adenylyl cyclase were cotransfected into Sf9 cells with baculoviral DNA (BacVector-3000, Novagen), and the recombinant viruses were screened by assay of adenylyl cyclase activity in Sf9 cell membranes.
G Protein Subunits-All G protein ␣ subunits were synthesized in E. coli as described by Lee et al. (34). G i␣1 was coexpressed with yeast protein N-myristoyltransferase (20) for synthesis of myristoylated protein. Purification of recombinant ␣ subunits was achieved by modifications of the methods of Linder et al. (35), as described by Lee et al. (34). Purified ␣ subunits were activated by incubation with 50 mM NaHepes (pH 8.0), 5 mM MgSO 4 , 1 mM EDTA, 2 mM dithiothreitol, and 400 mM [ 35 S]GTP␥S 1 at 30°C for 30 min for G s␣ or for 2 h for G i␣ . Free GTP␥S was removed by gel filtration. All G proteins were so activated unless stated otherwise. G i␣1 bound with GDP␤S was prepared by incubation of the protein with 400 M GDP␤S instead of GTP␥S for 2 h on ice, followed by gel filtration.
Expression and Purification of Adenylyl Cyclase in E. coli-The C 2 domain of type II adenylyl cyclase (IIC 2 ) and the C 1a domain of type V adenylyl cyclase (VC 1 (591)) were expressed in E. coli and purified as described previously (23,29). VC 2 and VC 1 (670) were expressed and purified as described for VC 1 (591). Briefly, these plasmids were cotransformed with pREP4 into the E. coli strain BL21(DE3). Cultures were grown in LB medium at room temperature and induced with 30 M isopropyl-1-thio-b-D-galactopyranoside at an A 600 of 1.2. Cells were harvested after 4 h and frozen in liquid N 2 . After incubation with lysozyme to facilitate lysis of cells, cellular debris was removed by centrifugation, and the supernatant was applied to a metal chelate column (Talon, CLONTECH). The column was washed with 500 mM NaCl and then eluted with 100 mM imidazole. The protein was then applied to a Mono Q 5/5 fast protein liquid chromatography column (Amersham Pharmacia Biotech) and eluted with a 30-ml linear gradient of NaCl (0 -300 mM). Peak fractions were pooled and concentrated. VC 1 (670) mutant proteins and a wild-type control were purified only by metal chelate chromatography. These proteins were then dialyzed overnight against 20 mM NaHepes (pH 8.0), 1 mM EDTA, 20 mM NaCl, 2 mM dithiothreitol, and 5% glycerol. The proteins were then concentrated to approximately 1 mg/ml.
Adenylyl Cyclase Assays-Adenylyl cyclase activity was measured as described by Smigel (36). All assays were performed for 7-10 min at 30°C in a final volume of 50 -100 l. The concentration of MgCl 2 was 5 mM. When used, forskolin was added at the start of the assay. To assess inhibition of adenylyl cyclase activity by G i␣ , membranes and G protein subunits were incubated for 15 min on ice in a total volume of 25 l prior to initiation of the assay (50 l total volume). In assays containing C 1 and C 2 domain proteins (reconstitution assays), limiting concentrations of the C 1 domain protein (diluted in HMED buffer containing 0.5 mg/ml BSA) were first incubated with G i␣ for 15 min on ice, followed by addition of G s␣ and IIC 2 or VC 2 prior to initiation of the assay. The final concentration of either IIC 2 or VC 2 was at least 0.5 M to promote interaction between the C 1 and C 2 proteins. The final concentration of ATP was 0.2 mM for membrane assays and 0.5-1 mM for reconstitution assays.
Sedimentation Equilibrium-Sedimentation equilibrium analysis was performed by centrifugation of 100-l samples for 60 h at 4°C in a Beckman TL100 centrifuge as described previously (23). The concentration of [ 35 S]GTP␥S-G i␣1 in each fraction was determined by liquid scintillation spectrometry. Molecular masses were calculated from the slope of a plot of ln C/Co versus r 2 , where C/Co is the relative protein concentration in a given fraction and r is the radial distance in the centrifuge (cm).

Inhibition of Type V Adenylyl
Cyclase by G i␣1 -The crystal structure of the catalytically active cytoplasmic domains of adenylyl cyclase associated with G s␣ demonstrated that the switch II region of the G protein ␣ subunits inserts into a groove formed by the ␣2Ј helix and the ␣3Ј-␤4Ј loop of the C 2a domain of adenylyl cyclase; the G protein also makes an additional contact with the extreme N terminus of C 1 . The pseudosymmetrical structure formed by C 1 and C 2 contains a comparable cleft in the C 1a domain. Because G i␣ and G s␣ are not competitive regulators of adenylyl cyclase (17), this cleft is an attractive candidate for interaction with G i␣ proteins (15). However, examination of the structure of adenylyl cyclase associated with both forskolin and G s␣ indicated that the cleft in C 1 was not large enough to accommodate the binding of an ␣ subunit in a manner comparable with that observed with the C 2 domain. We were surprised to note that we (and others) have characterized inhibition of G s␣ -or forskolin-stimulated adenylyl cyclase activity by G i␣ proteins in the past but have not examined the inhibitory effect of G i␣ in the presence of both G s␣ and forskolin (16 -18, 37-39). We now note that G i␣ is a poor inhibitor of G s␣ -and forskolin-stimulated type V adenylyl cyclase activity compared with effects seen in the presence of G s␣ or forskolin alone (Fig. 1). These results suggest that the crystal structure of the fully activated enzyme may not be appropriate to model sites for interaction with G i␣ .
Gel Filtration and Sedimentation Equilibrium Centrifugation-We have expressed in E. coli the C 2 domain of type V adenylyl cyclase and two forms of the C 1 domain, VC 1 (591) (23) FIG. 1. Inhibition of G s␣ -and G s␣ /forskolin-stimulated adenylyl cyclase activity by myristoylated G i␣1 . Membranes (20 g) from Sf9 cells infected with a baculovirus encoding type V adenylyl cyclase were assayed for 8 min at 30°C with the indicated concentrations of GTP␥S-G i␣1 in the presence of 100 nM G s␣ (q) or 100 nM G s␣ plus 50 M forskolin (f). All determinations were performed in duplicate and are representative of three experiments. and VC 1 (670), which contain C 1a and the entire C 1 domain, respectively. These proteins were purified to homogeneity (Fig.  2). We have examined the interactions of the C 1 and C 2 domains with activated G i␣1 by gel filtration and sedimentation equilibrium. We have also used both myristoylated G i␣1 and the nonmyristoylated protein as a control, as inhibition of adenylyl cyclase activity by G i␣1 requires this covalent lipid modification. Under the conditions used, the myristoylated protein migrates more slowly during gel filtration than do both nonmyristoylated G i␣1 and a 40-kDa standard, presumably due to the interaction of myristate with the gel filtration matrix (Fig.  3). This gel filtration profile is unaltered in the presence of VC 2 . However, when VC 1 (591) and myristoylated G i␣1 are mixed, the migration of both proteins is altered, consistent with formation of a 1:1 complex (compare the lower panels showing migration of VC 1 in the presence of myristoylated and nonmyristoylated G i␣1 and the chromatogram of myristoylated G i␣1 in the presence and absence of VC 1 ). Nonmyristoylated G i␣1 does not interact with either VC 1 (591) or VC 2 (Fig. 3B).
Sedimentation equilibrium demonstrates the same pattern of interactions as shown by gel filtration (Fig. 4). Myristoylated G i␣1 displays a molecular mass of 37 kDa (actual, 37.7 kDa) in the presence or absence of VC 2 . Addition of VC 1 (591) shifts the molecular mass to 72 kDa but has no effect on the nonmyristoylated protein. The predicted mass for a 1:1 complex of myristoylated G i␣1 and VC 1 (591) is 66 kDa. The slightly curved plot of ln[G i␣1 /total] versus r 2 for the complex may indicate an equilibrium of several different multimeric species.
Mutagenesis of Full-length Type V Adenylyl Cyclase-Based on the clear interactions between VC 1 and myristoylated G i␣1 , we created a series of mutations focused on the ␣2 helix and the ␣3-␤4 loop within the C 1 domain of type V adenylyl cyclase in an attempt to identify residues necessary for G i␣ -mediated inhibition (Fig. 5A). We also mutated additional residues that are conserved between type V and VI adenylyl cyclase but differ in other isoforms of the enzyme that are not inhibited by G i␣ . Baculoviruses encoding native type V adenylyl cyclase containing these point mutations were constructed, the proteins were expressed in Sf9 cells, and membrane preparations FIG. 2. Purification of C 1 and C 2 domains of type V adenylyl cyclase. Proteins were expressed and purified as described under "Materials and Methods." Samples were resolved electrophoretically on an 11% polyacrylamide gel containing sodium dodecyl sulfate, and proteins were visualized by staining with Coomassie Blue. Lanes 1-4 contain 1.5 g of VC 1 (591), VC 1 (670), VC 2 , and IIC 2 , respectively. were assayed for adenylyl cyclase activity. All of the mutant proteins were active, and they displayed varying sensitivities to myristoylated G i␣ (Fig. 6, Table I).
The largest difference between the C 2 domain, to which G s␣ binds, and the C 1 domain is the length of helix 1 and the loop between the ␣1Ј and ␣2Ј helices, which includes an extra 4 amino acid residues not found in the C 1 domain (Fig. 5A). This is one feature that makes the cleft in the C 1 domain significantly smaller than its counterpart in C 2 . Mutation of two residues within the ␣1 helix had no significant effect on the IC 50 for inhibition of adenylyl cyclase by G i␣ (Table I, Fig. 6A). The lack of a substantial effect of the Q406E or Q406A mutations was surprising, as this residue is exposed to solvent and extends into the cleft formed by the ␣2 and ␣3 helices. In addition, this residue is conserved in only those isoforms of adenylyl cyclase that are inhibited by G i␣ (Fig. 5A).
Mutations within the ␣2 helix did cause large changes in the response of adenylyl cyclase to G i␣ (Table I, Fig. 6A). The double mutation of Met-414 and Thr-415 to Ala and the point mutant E411A increased the IC 50 for inhibition by G i␣1 by 25and 75-fold, respectively. In contrast, mutation of Glu-418 to Ala enhanced inhibition, decreasing the IC 50 for G i␣1 by 6-fold. These residues all lie on the same face of helix 2 and extend their side chains into the cleft formed by helices 2 and 3. Glutamates 411 and 418 are conserved among nearly all isoforms of adenylyl cyclase; thus, these residues may not be involved in specificity, but they have a clear effect on inhibition by G i␣ .
Several mutations in the region surrounding ␣3-␤4 also increased the IC 50 for inhibition by G i␣ (Fig. 6B). The mutations L472A and M476A weakened the apparent affinity of G i␣ for adenylyl cyclase by 33-and 60-fold, respectively. The amino acid residue at position 476 is reported to be Val (31) (as is the case for type VI adenylyl cyclase). However, our type V clone clearly contains a Met at this position. Mutation of this residue to Val actually decreased the IC 50 for G i␣ by 3-fold. Although conserved in type V and VI adenylyl cyclase, mutation of Asn-480 to Asp caused little change. Surprisingly, the relatively small change of Val-479 to an Ile decreased the IC 50 for G i␣ by 4-fold. Except for E398V, none of the residues that were mutated had any effect on the apparent affinity of adenylyl cyclase for G s␣ , the synergy between G s␣ and forskolin, or the K m for ATP (Table I). E398V may have a more global effect on adenylyl cyclase, as its K m for ATP is 2-fold higher than the wild-type value and the EC 50 for activation by G s␣ is increased by 3-5-fold.
Alteration of other residues that are conserved throughout C 1 and the N-terminal region of C 2 had no effect on inhibition by G i␣ (Table I). Sequence-alignment strategies have not provided many candidates for possible interactions between G i␣ and the N terminus of C 2 (analogous to the interaction between C 1 and G s␣ ). In fact, many of the residues that have the most dramatic effect on inhibition by G i␣ are conserved in a number of isoforms that do not respond to G i␣ . Therefore, we cannot rule out additional contacts with other regions of adenylyl cyclase, in particular the N terminus of C 2 . However, it is clear that the majority of the binding is contributed by the C 1 domain and is clustered around the cleft formed by the ␣2 and ␣3 helices, as originally predicted. The structural location of residues whose mutation caused more than a 4-fold effect on inhibition by G i␣ is shown in Fig. 5B.
Inhibition of Reconstituted Adenylyl Cyclase Activity by G i␣1 -Adenylyl cyclase activity can be reconstituted by mixture of VC 1 (591) or VC 1 (670) with C 2 domains from either type V or type II adenylyl cyclase. The properties of these mixed proteins resemble those described previously with VC 1 (591) and IIC 2 (Ref. 23 and data not shown). G i␣1 inhibits adenylyl cyclase activity obtained by mixture of VC 1 (670) with VC 2 (Fig. 7A). Myristoylated GTP␥S-G i␣1 has an IC 50 of 0.3 Ϯ 0.1 M. The IC 50 for the GDP␤S-liganded protein is increased 5-fold, whereas nonmyristoylated GTP␥S-G i␣ is more than 30-fold less potent than the myristoylated protein. The GTP␥S-and GDPbound forms of both G i␣1 and G s␣ (23) display relatively small differences in their apparent affinity for adenylyl cyclase. As previously discussed (23), the primary effect of GTP (compared with GDP) is to cause conformational changes that promote dissociation of GTP-␣ from ␤␥ rather than major changes in the affinity of ␣ for its effectors.
The source of the C 2 domain makes little difference; inhibition is nearly equally effective with either IIC 2 or VC 2 , again consistent with primary interaction of G i␣ with the C 1 domain (Fig. 7B). Although the C 2 domain may contact G i␣1 , the residues involved do not seem to impart to individual isoforms of adenylyl cyclase the capacity to be inhibited by the G protein ␣ subunit. In the case of G s␣ , its interaction with the N terminus of C 1 is critical for function, and the residues are conserved as expected because activation by G s␣ is conserved among all isoforms of adenylyl cyclase.
The C 1b domain does contribute to the potency of inhibition by G i␣ . The IC 50 levels for inhibition of the adenylyl cyclase activity of mixtures of C 1 and C 2 domain proteins containing VC 1 (670) (both the C 1a and C 1b domains) are six times lower than those containing VC 1 (591) protein, which includes only a few residues of C 1b . The C 1b domains of the various isoforms of adenylyl cyclase differ significantly. It is possible that much of the selectivity for inhibition of adenylyl cyclase activity by G i␣ proteins is contributed by this region of adenylyl cyclase. The elevated IC 50 and changes in the assay conditions explain why inhibition was not detected previously with VC 1 (591) and IIC 2 (see "Materials and Methods").
Effect of Mutations in VC 1 (670) on Inhibition by G i␣1 -To determine whether the reconstituted system reflects the inhib- itory effects of G i␣ observed with full-length type V adenylyl cyclase, we made the mutations described above in VC 1 (670), expressed the proteins in E. coli, and purified them by Talon metal chelate chromatography. Except for E411A, all of the mutant proteins were synthesized at levels comparable with the wild-type protein (data not shown) and had similar specific activities. Adenylyl cyclase activity was measured by reconstitution of these mutant proteins with VC 2 in the presence of activated G s␣ and myristoylated G i␣1 (Fig. 8). As demonstrated with the full-length protein, mutants E411A, MT414A, L472A, and M476A all showed a clear reduction in G i␣ -mediated inhibition. Although the enhancement of inhibition of mutant M476V is not as pronounced as that shown in the full-length protein, there is a very marked increase in the potency of inhibition of mutant E418A by G i␣1 . DISCUSSION The inability of myristoylated G i␣ to inhibit effectively G s␣and forskolin-stimulated type V adenylyl cyclase activity suggests that the binding site for G i␣ is highly sensitive to the conformational changes imposed on the enzyme by stimulators of catalysis. Hints of this phenomenon were revealed previously by Taussig et al. (17). Thus, the capacity of myristoylated G i␣ to inhibit type V adenylyl cyclase decreased as the concentration of G s␣ used to activate the enzyme increased (although the IC 50 seemed largely unchanged). The simplest explanation of these results is that G s␣ and G i␣ can bind simultaneously to adenylyl cyclase. However, given sufficient concentrations of G s␣ or addition of another activator such as forskolin, the enzyme can largely overcome the inhibitory effect of G i␣ , despite apparent binding of the inhibitory G protein ␣ subunit. These studies also indicate that the recently solved threedimensional structure of the forskolin-and G s␣ -bound VC 1 ⅐IIC 2 complex may not be the optimal state of adenylyl cyclase to model its inhibitory interactions with G i␣ .
Previous work by Taussig et al. (17) suggested separate binding sites for G i␣ and G s␣ . Considering the pseudosymmetry of adenylyl cyclase, Sunahara et al. (23) and Tesmer et al. (15) suggested that G i␣ may bind to the C 1 domain of the enzyme in FIG. 5. Alignment of adenylyl cyclase isoforms and the position of mutations within type V adenylyl cyclase. A, structure-based alignment of amino acids that correspond to the proposed G i␣ binding site within C 1 . The analogous G s␣ -binding site within IIC 2 is shown for comparison. The amino acids that directly interact with G s␣ are shown in blue. Amino acids mutated within VC 1 in this study are shown in red. Type VI adenylyl cyclase is nearly identical to the type V enzyme in these regions. The species and references of the isoforms shown are as follows: V, canine (31); II, rat (46); III, rat (47); VII, mouse (48); VIII, rat (49); IX, mouse (5). B, the complex between G s␣ (gray and red), VC 1 (gold), IIC 2 (purple), and forskolin is shown. The side chains of residues whose mutation had a greater than 4-fold effect on the IC 50 for inhibition by G i␣ are highlighted. Switch II of G s␣ is shown in red; the ␣3-␤5 loop and surrounding structures are in gray. The figure was drawn with the program BOBSCRIPT and rendered with RASTER3D. a location analogous to that for interaction of G s␣ with the C 2 domain. We have expressed and purified fragments of the C 1 and C 2 domains of type V adenylyl cyclase and demonstrate herein the interaction of the C 1 domain of type V adenylyl cyclase with G i␣ by two independent methods. Both gel filtration and sedimentation equilibrium indicate formation of a 1:1 complex between VC 1 and G i␣1 , as was observed previously with G s␣ and IIC 2 (23). We were unable to detect interactions between VC 2 and myristoylated G i␣1 . We have to date been unable to detect formation of the tetrameric complex containing VC 1 , VC 2 , G s␣ , and G i␣1 . Nevertheless, we believe that G s␣ and G i␣ can bind simultaneously to adenylyl cyclase, as stated above. Binding of G i␣ to a complex of G s␣ , VC 1 , and VC 2 may weaken interactions between the cytosolic domains of the enzyme. Detection of simultaneous binding of G s␣ and G i␣ to adenylyl cyclase may thus require covalent linkage of C 1 and C 2.
A series of mutations was created in the C 1 domain of type V adenylyl cyclase that specifically targeted the ␣1-␣2 and ␣3-␤4 cleft, the site analogous to that for binding of G s␣ to IIC 2 . Several mutations within the ␣2 and ␣3 helices led to either increased or decreased IC 50 levels for inhibition of adenylyl cyclase activity by G i␣ . Mutations in other regions of the enzyme had no effect on this parameter. The point mutations E411A, L472A, and M476A and the double mutation M414A,T415A increased the IC 50 for G i␣ -mediated inhibition of these enzymes by 25-75-fold, whereas the mutations E418A, M476V, and V479I decreased the IC 50 for G i␣ inhibition by 4 -6-fold. All of the side chains at these positions point into the cavity created by the flanking helices, ␣2 and ␣3, except for Leu-472 (Fig. 5B). This side chain might participate in contacts with the ␣3-␤5 loop of G i␣ , as do residues in the C 2 domain with G s␣ or even with the C 1b domain of adenylyl cyclase.
Residues within the G s␣ binding pocket are conserved among the different isoforms of adenylyl cyclase as expected, as all isoforms respond to this activator. However, only type I, V, and VI adenylyl cyclases are inhibited by G i␣ , and we expected to find the greatest effect of mutation of residues that are conserved among only these isoforms. However, residues Glu-411 and Glu-418 are conserved in nearly all isoforms of adenylyl cyclase, and Val-479 is conserved in about half of these, including several that are not inhibited by G i␣ . Residues Leu-472, Met-476, Met-414, and Thr-415 are not conserved among other isoforms of adenylyl cyclase, but they also differ significantly in the type I enzyme. A residue such as Gln-406, located at the entrance of the ␣2/␣3 cleft and conserved between type V, VI, and I adenylyl cyclases, had no significant effect on inhibition by G i␣ when mutated to either alanine or glutamate. It is thus unclear how specificity for interaction with G i␣ is achieved.
The main structural difference between the G s␣ binding pocket and that proposed for G i␣ is the size of the cleft, which is largely dictated by the angle between the ␣2 and ␣3 helices and the insertion of four residues in C 2 in the loop between ␣1Ј and ␣2Ј. This provides a wider entrance to this cleft for contact with switch II of G s␣ . There are significant conformational differences in the G s␣ binding pocket when one compares the IIC 2 homodimeric structure (14), proposed to mimic an inactive ground state, with that of the G s␣ ⅐Fsk⅐VC 1 ⅐IIC 2 complex (15). It is unknown what changes may occur in the C 1 domain and the G i␣ binding pocket as the enzyme is converted from a low to a high activity state. However, this cleft undergoes conformational changes upon P-site inhibitor and pyrophosphate binding.
Inhibition of adenylyl cyclase activity by G i␣ was also detected using the purified C 1 and C 2 domains of the type V enzyme. The IC 50 for inhibition was only 2-fold greater than that observed with full-length, membrane-bound type V adenylyl cyclase (0.3 versus 0.15 M). This is a small difference FIG. 6. Inhibition of type V adenylyl cyclase mutants by myristoylated G i␣1 . Membranes (30 g) from Sf9 cells expressing either wild-type V adenylyl cyclase or mutant proteins were assayed with 100 nM activated G s␣ in the presence of 7 nM--7 M myristoylated GTP␥S-G i␣1 . A, assay of adenylyl cyclase proteins with mutations located in the ␣1-␣2 helix region. B, assay of adenylyl cyclase proteins with mutations in the ␣3 helix. The endogenous adenylyl cyclase activity of membranes from Sf9 cells expressing ␤-galactosidase are shown as a control. All activities are expressed as percentages of control values measured in the absence of G i␣ (see Table I for specific activities of each). Assays were performed in duplicate, and the results are representative of two experiments. a The location of each mutant residue is based on the secondary structure of the G s␣ ⅐ Fsk ⅐ VC 1 ⅐ IIC 2 complex. b The variation in the V max for each mutant is due in large part to differences in the age of Sf9 cells used for expression rather than the stability of individual proteins. Endogenous adenylyl cyclase activity of Sf9 membranes infected with ␤-galactosidase is 0.03 nmol/min/mg. c The EC 50 for G s␣ stimulation was determined from dose-response curves in the absence and presence of 50 M forskolin with G s␣ concentrations ranging from 3 to 300 nM. compared with the 10-fold or more loss of apparent affinity for G s␣ when the purified soluble VC 1 /IIC 2 system was compared with the holoprotein. Full inhibition by G i␣ required both activation of the protein with GTP␥S and myristoylation. Myristoylation was also required for binding of G i␣ to VC 1 as measured by gel filtration and sedimentation equilibrium. Modeling of G i␣ in the binding pocket identified by mutagenesis (Fig. 9) places its N terminus and attached myristate near the presumed location of the plasma membrane. The requirement for myristoylation of G i␣ for inhibition of membrane-bound adenylyl cyclases is not understood. However, insertion of myristate in the plasma membrane would increase the local concentration and apparent affinity of G i␣ for adenylyl cyclase, and there are also several examples of intermolecular and intramolecular protein-lipid interactions. Recoverin and Arf contain myristate groups that are sensitive to the conformational states of these proteins (40 -42). Lipid-mediated protein-protein interactions are prevalent in binding of regulatory molecules to Ras-related G proteins (43). The requirement for myristate in a completely soluble system suggests that myristate does not serve simply to concentrate G i␣ at the membrane but rather interacts directly with either the protein moiety of G i␣ , adenylyl cyclase, or both to increase the affinity of their interaction.
The inhibitory effect of G i␣ on reconstituted adenylyl cyclase activity was largely unchanged when the C 2 domain of the type II enzyme replaced VC 2 , suggesting that C 2 does not play a significant role in binding G i␣ and thereby dictating isoformspecific recognition of the regulatory protein. However, inhibition by G i␣ was enhanced when the full C 1 domain (VC 1 (670)) was included compared with C 1a alone (VC 1 (591)). Isoform specificity may be partially dictated by C 1b . This region is highly variable among the different isoforms of adenylyl cyclase. The difference in inhibition by G i␣ between VC 1a and VC 1 (670) was not observed when the C 1 and C 2 domains were linked covalently (30). It is not clear why this should make a difference. However, mutations in VC 1 (670) do faithfully mimic the phenotypes of those made in the full-length protein.
Tesmer et al. (15) and Sunahara et al. (44) suggested that the specificity for recognition of G s␣ at its binding site was dictated primarily by the backbone conformation of the ␣ subunit and not its primary sequence. There are only two conservative changes between G s␣ and G i␣ in the switch region that interacts with adenylyl cyclase. However, the ␣3-␤5 and ␣4-␤6 loops of G s␣ and G i␣ , proposed by mutagenesis studies to interact with adenylyl cyclase, do have differences in their primary amino acid sequences and structural conformations. Crystallographic data suggest that the ␣4-␤6 loop of G s␣ does not contact the conserved cytoplasmic domains of adenylyl cyclase. Based on the location of this loop in G i␣ relative to that of the C terminus of C 1a , it is possible that the C 1b domain of adenylyl FIG. 7. Inhibition of the soluble form of type V adenylyl cyclase by G i␣1 . A, purified VC 1 (670) (50 nM) was reconstituted with 0.5 M VC 2 and assayed with 400 nM activated G s␣ in the presence of 10 nM--10 M myristoylated GTP␥S-G i␣1 (q), myristoylated GDP␤S-G i␣1 (f), or nonmyristoylated GTP␥S-G i␣1 (OE). B, VC 1 (670) (q, f) or VC 1 (591) (OE, ƒ) was reconstituted with 0.5 M VC 2 (q,OE) or IIC 2 (f,ƒ) and assayed with 400 nM G s␣ in the presence of the indicated concentrations of myristoylated GTP␥S-G i␣1 . Activities are expressed as percentages of control values: 3.3, 5.6, 8.7, and 8.5 nmol/ min/mg for VC 1 (670)/VC 2 , VC 1 (670)/IIC 2 , VC 1 (591)/VC 2 , and VC 1 (591)/IIC 2 , respectively. All values shown are the averages of duplicates, and the results are representative of three experiments. cyclase could interact with this region of G i␣ , adding additional stability to this complex.
The mechanism of activation of adenylyl cyclase by G s␣ is not completely understood, and discussion of the mechanism of inhibition by G i␣ can thus be only speculative. However, the close proximity of the G i␣ binding site to the active site of the enzyme certainly poises G i␣ to affect catalysis by a number of mechanisms. A simple model of binding of G i␣ to adenylyl cyclase based on the enzyme's interactions with G s␣ (Fig. 9) does not explain all of the effects of the mutations in adenylyl cyclase described above or those in G i␣ reported previously (45). However, this model is consistent with the general location of the binding site for G i␣ that is indicated by the mutations. Therefore, much in the same way that binding of G s␣ is proposed to widen the binding cleft in C 2 , inducing a domain rotation that reorients residues within the active site (15), G i␣ may widen its binding pocket in C 1 and generate a movement of the ␣1-␣2 loop toward the catalytic site to disrupt or block the active site. The addition of forskolin to the G s␣ -bound enzyme increases the affinity of the two domains for one another and may serve to lock the enzyme in a highly activated state that does not favor G i␣ binding to C 1 . However, we suggest that G s␣ and G i␣ do bind simultaneously to adenylyl cyclase, forming a large complex (200 kDa with the native enzyme) whose pseudosymmetry extends from the duplicated structure of adenylyl cyclase to these two regulatory proteins. The pseudosymmetry of adenylyl cyclase is a clear structural correlate of bidirectional regulation of the enzyme by homologous G protein ␣ subunits. FIG. 9. Two views of a theoretical complex between the catalytic domains of adenylyl cyclase and the stimulatory and inhibitory G protein ␣ subunits G s␣ and G i␣ . Evidence for such a complex is from kinetic studies of full-length adenylyl cyclase that revealed noncompetitive inhibition of G s␣ -stimulated activity by G i␣ (17). As yet, there is no evidence that such a tetrameric complex exists for the soluble system whose components are portrayed here. Thus, these models are only meant to show the pseudosymmetrical relationship of the G s␣ binding site to that of the proposed G i␣ binding site and the relationship of both ␣ subunits to the active site of adenylyl cyclase. A, side view of the complex. The N and C termini of all the subunits are found along the top of the model, where the lipid bilayer of the plasma membrane is proposed to exist. The positions of G s␣ , C 1 , and C 2 are derived from the crystal structure of G s␣ ⅐C 1a ⅐C 2 bound to forskolin, 2Ј-deoxy-3Ј-AMP, and pyrophosphate (15). The position of G i␣ was generated by first superimposing the structural elements of the C 2 domain that bind G s␣ on their equivalents in C 1 , then subjecting the model of G s␣ to the same transformation, and then superimposing G i␣ on top of the new position of G s␣ . The switch II helix of G s␣ is red, whereas that of G i␣ is green. The switch II helix of G i␣ is thought to interact directly with adenylyl cyclase (45), and as modeled, many of the residues shown herein to be important for inhibition of adenylyl cyclase by G i␣ are close to this helix. The molecule of GTP bound to each ␣ subunit is drawn as a ball-and-stick model. B, cytoplasmic face of the complex. Forskolin and 2Ј-deoxy-3Ј-AMP/PP i , which bind in a long cleft between the C 1 and C 2 domains of adenylyl cyclase, are shown as ball-and-stick models. 2Ј-deoxy-3Ј-AMP/PP i is bound in the active site of the enzyme. G i␣ may function by disrupting the active site, decreasing the affinity of C 1 for C 2 , or both. Both panels were generated with the program BOBSCRIPT and rendered using RASTER3D.