Interaction of Gsα with the Cytosolic Domains of Mammalian Adenylyl Cyclase*

Forskolin- and Gsα-stimulated adenylyl cyclase activity is observed after mixture of two independently-synthesized ∼25-kDa cytosolic fragments derived from mammalian adenylyl cyclases (native M r ∼ 120,000). The C1a domain from type V adenylyl cyclase (VC1) and the C2 domain from type II adenylyl cyclase (IIC2) can both be expressed in large quantities and purified to homogeneity. When mixed, their maximally stimulated specific activity, 150 μmol/min/mg protein, substantially exceeds values observed previously with the intact enzyme. A soluble, high-affinity complex containing one molecule each of VC1, IIC2, and guanosine 5′-O-(3-thiotriphosphate) (GTPγS)-Gsα is responsible for the observed enzymatic activity and can be isolated. In addition, GTPγS-Gsαinteracts with homodimers of IIC2 to form a heterodimeric complex (one molecule each of Gsα and IIC2) but not detectably with homodimers of VC1. Nevertheless, Gsα can be cross-linked to VC1 in the activated heterotrimeric complex of VC1, IIC2, and Gsα, indicating its proximity to both components of the enzyme that are required for efficient catalysis. These results and those in the accompanying report (Dessauer, C. W., Scully, T. T., and Gilman, A. G. (1997) J. Biol. Chem. 272, 22272–22277) suggest that activators of adenylyl cyclase facilitate formation of a single, high-activity catalytic site at the interface between C1 and C2.

proteins by virtue of their two large hydrophobic domains, each of which is hypothesized to contain six membrane-spanning helices. The first of these hydrophobic regions follows a short amino-terminal sequence and precedes a roughly 40-kDa cytoplasmic domain (C 1 ). The second hydrophobic region separates C 1 from a second cytosolic domain (C 2 ) of comparable size. Each of the two cytosolic domains includes a sequence of 200 -250 amino acid residues that is typically 50% similar to its consort, 50 -90% similar to the corresponding domains of other adenylyl cyclase isoforms, and 20 -25% similar to the catalytic domains of membrane-bound and cytosolic guanylyl cyclases.
Detailed biochemical characterization of adenylyl cyclase is impaired by the insolubility, instability, and sparsity of the native enzyme, as well as our incapacity to express necessary amounts of the protein in heterologous systems. To overcome these hurdles we have synthesized (in Escherichia coli) portions of the two cytosolic domains of adenylyl cyclase in the absence of the remainder of the protein, first as a 55-kDa chimeric fusion protein containing the C 1a domain of type I adenylyl cyclase linked to the C 2 domain of the type II enzyme (3,4). The specific activity of this engineered enzyme is remarkably high, it is soluble in the absence of detergent, and it is adequately stable. Importantly, it is activated synergistically by G s␣ and forskolin and inhibited by so-called P-site inhibitors and the G protein ␤␥ subunit complex, providing ample justification to pursue investigation of this and similar artificial entities. To overcome the remaining hurdle of a relatively low level of accumulation of the chimera in bacterial cytosol, we and others prepared the two cytosolic domains of adenylyl cyclase as distinct entities and found that enzymatic activity with similar regulatory properties can be reconstituted by simple mixture of the two roughly 25-kDa proteins (5,6). Large amounts (50 -100 mg or more) of the C 2 domain of type II adenylyl cyclase can be prepared readily, but similar results are difficult to achieve with the C 1a domain of the type I enzyme. We have now extended this approach by utilizing a fragment of the C 1 domain of type V adenylyl cyclase, which can be expressed in reasonable (but not exuberant) quantities, and we have characterized the interactions of the two cyclase fragments with each other and with G s␣ .

MATERIALS AND METHODS
Plasmid Construction-DNA containing nucleotides 961-2010 (amino acid residues 321-670) of canine type V adenylyl cyclase (7) was generated with the polymerase chain reaction using oligonucleotides A: 5Ј-CCATGGCTGAGGTCTCCCAG-3Ј; and B: 5Ј-TTGCGGCCGCGGA-TCCGGTCAGGCTCCCTGAAGG-3Ј, as primers. Note that oligonucleotide A is 5Ј to a unique NcoI site in the cDNA for canine type V adenylyl cyclase and includes the codon for Met 364 . Further truncations of this fragment were generated by taking advantage of unique restriction sites for AvrII (at nucleotide 1815), BstBI (at nucleotide 1858), and Bsu36AI (at nucleotide 1900) to create VC 1 (364 -606), VC 1 (364 -620), and VC 1 (364 -635), respectively. VC 1 (364 -567), VC 1 (364 -591), and VC 1 (364 -591)Flag were generated with the polymerase chain reaction using the following pairs of oligonucleotide primers: A and 5Ј-TTCTC-CGGATCCAAGCTTGCAGCGCAGGA-3Ј, A and 5Ј-GATTCGAAGCTT-GTGCCCAATGGAG-3Ј, and A and 5Ј-GCTAATTAAGCTTGTCATCGT-CGTCCTTGTAGTCGTGCCCAATGGAGTTG-3Ј. The last construct contains a carboxyl-terminal Flag epitope (Kodak), DYKDDDDK, as well as a cleavage site for enterokinase. All of the VC 1 constructs were ligated into pQE60-H6 (8) with NcoI and HindIII. Thus, each of the encoded proteins contains a hexa-histidine tag at its amino terminus, followed by the residue corresponding to Met 364 of canine type V adenylyl cyclase, and terminates at the stop codon contained in pQE60-H6. We will subsequently designate these proteins by reference to their carboxyl-terminal residue.
Expression and Purification of Proteins-E. coli strain BL21 was co-transformed with pREP4 and pQE60-H6-VC 1 (various) and grown overnight at 30°C in 200 ml of LB medium containing ampicillin (50 g/ml) and kanamycin (50 g/ml). This culture was used to inoculate 10 liters of T7 medium (8), which was incubated at 30°C until OD 600 reached 1.3. Synthesis of VC 1 was then induced with 30 M isopropyl thiogalactoside, and incubation was continued for 4 h at room temperature. Cells were harvested by centrifugation, frozen in liquid N 2 , and stored at Ϫ70°C. Frozen cells were pulverized and suspended in 750 ml of buffer A (50 mM Tris-HCl, pH 8, 120 mM NaCl, 1 mM ␤-mercaptoethanol, and a mixture of protease inhibitors (4)) prior to incubation with 0.25 g/ml lysozyme in buffer A for 30 min at 4°C and subsequently 8 g/ml DNase and 1 mM MgCl 2 in buffer A for an additional 30 min. Cellular debris was removed by centrifugation at 100,000 ϫ g for 40 min at 4°C.
All purification steps were performed at 4°C. The supernatant (750 ml) was applied to a 5-ml metal chelate column (Talon, CLONTECH) equilibrated in buffer A, and the column was washed with 20 volumes of buffer A containing 500 mM NaCl and 10 volumes of buffer B (20 mM Tris-HCl, pH 8.0, 20 mM NaCl, 1 mM ␤-mercaptoethanol, and protease inhibitors). H6-VC 1 was eluted from the column in 2-ml fractions with buffer B containing 100 mM imidazole. Peak fractions were pooled, diluted with 2 volumes of buffer C (20 mM NaHepes, pH 8.0, 20 mM NaCl, 2 mM dithiothreitol, and protease inhibitors) and applied to a Mono Q 5/5 FPLC column (Pharmacia). This column was washed with 5 ml of buffer C and protein was eluted with a 25-ml linear gradient of NaCl (20 -300 mM in buffer C). Peak fractions (about 150 mM NaCl) were pooled and concentrated to 3 mg/ml unless noted otherwise.
The C 2 domain of type II adenylyl cyclase (IIC 2 ) was expressed in E. coli and purified as described previously (5), as was bovine G s␣ (short form) (8). G s␣ was activated with GTP␥S by incubation with 50 mM NaHepes, pH 8.0, 10 mM MgSO 4 , 1 mM EDTA, 2 mM dithiothreitol, and 800 M GTP␥S for 30 min at 30°C.
Sedimentation Equilibrium-Sedimentation equilibrium analysis was performed by centrifugation of samples (100 l) for 60 h at 4°C in a Beckman TL100 centrifuge as described previously (9). The buffer contained 50 mM NaHepes, pH 8.0, 10 mM MgCl 2 , 1 mM EDTA, 20 mM ␤-mercaptoethanol, 50 mM NaCl, 50 mM 6-O-(3Ј-(piperidino)propionyl)forskolin, and 5 mg/ml dextran T40 to provide density stabilization. Immediately after centrifugation, 7-l fractions were collected with a Brandel microfractionator. The concentration of [ 35 S]GTP␥S-G s␣ in each fraction was determined by liquid scintillation spectrometry. The concentration of adenylyl cyclase in each aliquot was determined by quantification of catalytic activity under linear conditions with addition of either 1 M VC 1 (591) or IIC 2 in the presence of 50 M forskolin and 0.2 M GTP␥S-G s␣ . The addition of exogenous GTP␥S-G s␣ and either the C 1 or C 2 domain of adenylyl cyclase was necessary to assure line-arity of activity throughout the range of concentrations found in the gradient. Molecular weights were calculated from the slope of a plot of ln C/C o versus r 2 , where C is the protein concentration in a given fraction, C o is the initial concentration, and r is the radial distance in the centrifuge (cm). Similar results were obtained for all conditions at two different rotational velocities.
Chemical Cross-linking-Cross-linking studies were performed utilizing the bifunctional amine-coupling reagent disuccinimidyl suberate (Pierce). Purified proteins were diluted into a buffer containing 20 mM NaHepes (pH 8.0), 1 mM EDTA, and 2 mM MgCl 2 . Cross-linking was initiated by addition of freshly prepared disuccinimidyl suberate (100 M final concentration) and allowed to progress for 15 min at room temperature. Reactions were terminated with SDS-PAGE sample buffer containing 10 mM glycine and 20 mM dithiothreitol, and proteins were resolved by SDS-PAGE. Proteins were visualized by immunoblotting using antibodies specific for H6-VC 1 (364 -591)Flag (anti-Flag), IIC 2 -H6 (10), or G s␣ (11).
Adenylyl Cyclase Assays-Adenylyl cyclase activity was measured as described by Smigel (12). All assays were performed for 10 min at 30°C in a volume of 100 l. In reconstitutive assays the final concentration of IIC 2 was at least 1-5 M to maintain linearity with variable concentrations of VC 1 . The final concentration of ATP was 1 mM unless stated otherwise, and an ATP regenerating system was used only when crude preparations of enzyme were assayed.

Expression and Purification of Proteins Containing the VC 1a
Domain-The C 1 domain of type V adenylyl cyclase was truncated at its carboxyl terminus in attempts to obtain a pure, stable protein with high specific enzymatic activity when reconstituted with IIC 2 . Bacterial lysates containing constructs as large as VC 1 (670), for example, supported high forskolinstimulated adenylyl cyclase activity (Ͼ50 nmol/min/mg in the lysate), but immunoblot analysis of SDS-PAGE gels indicated contamination with several proteolytic products (data not shown). Extracts containing truncations VC 1 (635), VC 1 (621), or VC 1 (606) displayed progressively less such contamination, but susceptibility to degradation was again high with VC 1 (567); the latter protein is analogous to the C 1 domain of type I adenylyl cyclase described previously (5). However, extracts of bacteria expressing VC 1 (591) or its carboxyl-terminally Flag-tagged counterpart, VC 1 (591)Flag, contained only a single significant proteolytic product (molecular mass ϳ 25 kDa; mass of VC 1 (591) ϳ 28 kDa), and its accumulation was minimized by shortening the length of time for expression to 4 h. The specific activities of such lysates for reconstitution of adenylyl cyclase activity with IIC 2 were typically 50 -100 nmol/ min/mg extract protein. These values are roughly 1% of those observed with purified native adenylyl cyclases (12)(13)(14).
Metal chelate column chromatography (Talon resin) of extracts containing VC 1 (591)Flag resulted in 100-fold or more purification of the protein with loss of 60% of the total activity (Table I). The remaining activity was not found in the flowthrough or column washes. Subsequent Mono Q column chromatography resulted in 2-3-fold purification and removal of the remaining major contaminants. This final product appeared essentially homogeneous when analyzed electrophoretically ( Fig. 1) and had a reconstitutive specific activity of 30 mol/min/mg when assayed with maximally effective concen- trations of purified IIC 2 and 50 M forskolin as the sole activator. This purified protein will subsequently be designated VC 1 . Adenylyl Cyclase Activity Reconstituted from VC 1 and IIC 2 -The adenylyl cyclase activity observed in mixtures of VC 1 and IIC 2 resembles that seen previously with mixtures of IC 1 and IIC 2 (5, 6); many of the regulatory features that characterize native membrane-bound adenylyl cyclases are retained. Activities shown in Fig. 2, A, B, and D, were measured with low concentrations of IIC 2 and either near saturating (2 M, Fig. 2, A and B) or variable (Fig. 2D) concentrations of VC 1 . Activities shown in Fig. 2C were measured with low concentrations of VC 1 and near saturating (2 M) concentrations of IIC 2 . The reconstituted enzyme is activated by either forskolin (EC 50 Ͼ 10 M) or G s␣ (EC 50 ϳ 400 nM) (Fig. 2, A and B). Unlike the mixture of IC 1 and IIC 2 , GTP␥S-G s␣ could not activate a mixture of VC 1 and IIC 2 maximally. The effects of forskolin and GTP␥S-G s␣ are synergistic; the presence of one activator increasing the apparent affinity (EC 50 ) of the enzyme for the other by 40-fold or more. In the absence of an activator, the rate of cyclic AMP synthesis is low and the two domains of the enzyme have an apparent affinity for each other of greater than 5 M (Fig. 2D, inset). This value is lowered in the presence of activators; the apparent affinity of VC 1 and IIC 2 is 0.66 M with GTP␥S-G s␣ , 1.2 M with forskolin, and 150 nM when both activators are present.
Activation of adenylyl cyclase activity by G s␣ is dependent on the nature of the bound nucleotide, but not to the extent usually assumed. GDP-G s␣ is only 10-fold less potent than GTP␥S-G s␣ and is equally efficacious (in the presence or absence of forskolin (Fig. 2B)). Identical results were obtained with limiting concentrations of either VC 1 or IIC 2 (data not shown).
Gel Filtration and Sedimentation Equilibrium Centrifugation-We have examined the interactions of VC 1 and IIC 2 with each other and with G s␣ by gel filtration, sedimentation equilibrium centrifugation, and (see below) chemical cross-linking. Gel filtration of VC 1 or IIC 2 on tandem columns of Superdex 75 and 200 suggests that each of these proteins exists in solution as roughly 50-kDa homodimers (Fig. 3, A and B); each individual protein has a molecular mass of approximately 28,000. Similarly, sedimentation equilibrium analysis of IIC 2 revealed a molecular weight of 46,000 (Fig. 6D). Gel filtration of a mixture of VC 1 and IIC 2 in the presence (Fig. 3B) or absence (not shown) of forskolin revealed a single peak of protein with an apparent molecular weight of 50,000, while sedimentation equilibrium analysis of the mixture indicates a molecular weight of 52,000 (Fig. 6B). Because of the similarity of the molecular masses of VC 1 and IIC 2 , we do not know if these values represent homodimers, heterodimers, or a mixture of the two.
Mixture of VC 1 , IIC 2 , and GTP␥S-G s␣ (in the presence of forskolin) results in formation of a high-affinity complex. This is detected by gel filtration with the appearance of a new peak of optical density, migrating with an apparent molecular weight of roughly 100,000 (Fig. 3B). SDS-PAGE analysis of fractions obtained by gel filtration of such a mixture containing an excess of IIC 2 indicates that this high molecular weight peak contains all three proteins, with an apparent stoichiometry of 1:1:1 (by scanning densitometry) (Fig. 4). (Note that essentially all of the VC 1 is found in this high molecular weight peak, presumably indicating that most of the protein is active.) Similarly, equilibrium sedimentation of such mixtures reveals a 104 -107 kDa complex when either G s␣ or adenylyl cyclase activity is monitored (Fig. 6, A and B). The anticipated molecular weight of a 1:1:1 complex of G s␣ , VC 1 , and IIC 2 is 102,000.
Interaction between G s␣ and IIC 2 alone can also be detected by gel filtration and equilibrium sedimentation. Incubation of 50 M [ 35 S]GTP␥S-G s␣ with 5, 35, or 250 M IIC 2 causes elution of radioactivity at a position consistent with progressively higher molecular weights (Fig. 5C). At the highest concentration of IIC 2 , the complex is consistent with molecular mass ϳ70 kDa. No such interactions were detected between VC 1 and GTP␥S-G s␣ (Fig. 5A) or IIC 2 and GDP-G s␣ (Fig. 5B). With sedimentation equilibrium, the apparent molecular weight of G s␣ was increased by 24,000 following mixture with IIC 2 (but not with VC 1 ) (Fig. 6C) and analysis of adenylyl cyclase activity revealed a complex between GTP␥S-G s␣ and IIC 2 with mass ϭ 64 kDa. (Note: observation of enzymatic activity in this experiment required addition of VC 1 ; the complex of GTP␥S-G s␣ with IIC 2 does not have detectable adenylyl cyclase activity.) All of these observations indicate formation of a relatively low affinity 1:1 complex between IIC 2 and GTP␥S-G s␣ , and, thus, that interaction of the adenylyl cyclase domains with the activated G protein ␣ subunit disrupts the homodimeric interactions characteristic of VC 1 and IIC 2 .
Chemical Cross-linking-Although direct interactions between VC1 and G s␣ were not detected by gel filtration or sedimentation equilibrium, chemical cross-linking studies do suggest that the molecules are at most 11 Å away from each other when tightly complexed with IIC 2 . Fig. 7A illustrates the forskolin (and IIC 2 -, not shown)-dependent appearance of a 70-kDa species representing covalent coupling of G s␣ and VC 1 by the 11-Å cross-linker, disuccinimidyl suberate. Analysis of the same fractions with an anti-G s␣ antibody also reveals a 70-kDa species that appears in a forskolin-dependent manner. It is unclear in this panel if this species represents G s␣ -VC 1 , G s␣ -IIC 2 , or both, since the cyclase domains have similar molecular weights. This was also complicated by the inability of available IIC 2 antibodies to detect IIC 2 -G s␣ complexes by immunoblotting. Cross-linking studies performed at higher protein concentrations (2 M VC 1 , IIC 2 , and G s␣ ) and analyzed with an anti-G s␣ antibody reveal a IIC 2 -G s␣ cross-linked species in the absence of VC 1 (Fig. 7E, lane 4). Moreover, a large species (M r ϳ 95,000) appears in the presence of all three proteins and 100 M forskolin; we presume this to be the VC 1 -IIC 2 -G s␣ heterotrimer (Fig. 7E, lane 7). Observation of both the cross-linked heterotrimer and the IIC 2 -G s␣ heterodimer is dependent on the presence of GTP␥S-activated G s␣ (Fig. 7E, lane 3 versus 4 and  lane 6 versus 7).
Chemical cross-linking also revealed forskolin-dependent formation of heterodimers between VC 1 and IIC 2 (Fig. 7, C and  D). Although each domain exists as a homodimer under nonactivating conditions, formation of heterodimers occurs in a forskolin (and G s␣ -)-dependent manner. DISCUSSION We have expressed and purified a fragment of the C 1 domain of type V adenylyl cyclase, consisting of amino acid residues 364 -591 and including hexa-histidine and Flag tags at the amino and carboxyl termini, respectively. Although this protein itself has no adenylyl cyclase activity, catalysis of cyclic AMP synthesis is restored by simple mixture of VC 1 with an appropriate fragment from the second cytosolic domain of the enzyme, such as IIC 2 . This interaction is similar to that described previously between IC 1 and IIC 2 (5, 6); the major ad-vantages are the yield of VC 1 , which exceeds that of IC 1 by 20-fold, and the apparent homogeneity of the product. The adenylyl cyclase activity of the VC 1 -IIC 2 mixture is stimulated markedly by either G s␣ or forskolin, and these two regulators activate the enzyme synergistically when present simultaneously. These are characteristics of native type II and type V adenylyl cyclases. A notable difference between this reconstituted adenylyl cyclase and the native enzymes is the maximal stimulated activity, which typically exceeds 100 mol/min/mg in the case of the mixture of VC 1 and IIC 2 ; values of 10 mol/min/mg typify purified preparations of native enzymes (12)(13)(14). Although the source of this discrepancy is not known, we suspect that the values observed with the VC 1 /IIC 2 mixture may indeed approximate a true V max for mammalian adenylyl cyclase. Several factors may cause underestimation of maximal activity when dealing with a native adenylyl cyclase. Overexpression of these enzymes in Sf9 cells is plagued by production of nonfunctional protein; detergents are necessary to maintain solubility of the native proteins but may alter estimates of specific activity; lengthy purification schemes may cause denaturation of these labile entities. Alternatively, it is possible that inhibitory domains have been removed from the soluble constructs described above or that the membrane spanning domains of the enzymes do not permit optimal orientation of the interacting cytosolic segments. Another notable difference between native adenylyl cyclases and the engineered soluble enzymes (whether or not the cytosolic domains are linked covalently) is the relatively low (reduced 20 -50-fold) apparent affinities of the soluble enzymes for G s␣ . It is possible that the transmembrane spans, the loops that connect them, and/or residues immediately surrounding the remnants of the C 1 and C 2 domains in the constructs utilized may contribute to the binding site for G s␣ . Nevertheless, the essential features of activation of adenylyl cyclase by the G protein ␣ subunit are retained.
We demonstrate herein that the C 1 and C 2 domains of adenylyl cyclase interact to form a catalytically active adenylyl cyclase and that the apparent affinity of C 1 for C 2 is enhanced in the presence of G s␣ and/or forskolin. Gel filtration, equilibrium sedimentation, and cross-linking analyses all demonstrate that G s␣ interacts with the C 2 domain of adenylyl cyclase in a GTP-enhanced manner and that this interaction is further stabilized by C 1 . Direct interactions between VC 1 and G s␣ were not detected by gel filtration or sedimentation equilibrium. However, these two proteins could be cross-linked by disuccinimidyl suberate in a IIC 2 -dependent manner. We further demonstrate that G s␣ , C 1 , and C 2 form a relatively high-affinity complex with a 1:1:1 stoichiometry. Although this is not surprising, the similarities in the primary sequence of the C 1 and C 2 domains of adenylyl cyclase raised the possibility of binding sites for two molecules of G s␣ . Several adenylyl cyclase isoforms are inhibited by G i␣ . Although high concentrations of G i␣ can apparently compete for the G s␣ -binding site, inhibition of adenylyl cyclase activity by G i␣ is not dependent on G s␣ and is not competitive with the stimulatory ␣ subunit (15). We presume that there is a distinct binding site for G i␣ on certain adenylyl cyclases; G i␣ may well be found to interact predominantly with C 1 .
Homodimerization of C 1 and C 2 is difficult to interpret. The phenomenon may clearly be an artifact of protein engineering and have no significance with regard to the membrane-bound native enzyme. However, we have previously detected oligomers of near native adenylyl cyclases in detergent solution, suggesting the possibility of relevance of homodimerization of the cytosolic domains (16).
The relative capacities of GTP-and GDP-bound G ␣ proteins to interact with their effectors has been long debated and may be dependent on the system in question. However, no such interaction has probably been examined previously in a system containing such highly purified proteins. The G s␣ , VC 1 , and IIC 2 proteins utilized herein are all expressed abundantly in bacteria, greatly facilitating their purification to a high degree. Significant levels of contaminating nucleotide kinases are extremely unlikely, obviating such concerns as conversion of GDP to GTP in the presence of ATP. Nevertheless, the apparent affinity of GDP-G s␣ for adenylyl cyclase is only about 10-fold less than that of GTP␥S-G s␣ . The GTPase activity of a G pro- tein ␣ subunit thus facilitates deactivation of the system by reducing the affinity of G ␣ for its effector. However, the affinity of the G protein ␤␥ subunit complex for G ␣ is more highly dependent on the nature of the bound nucleotide. GTP hydrolysis thus greatly favors association of ␣ with ␤␥, and it is this interaction that prevents access of either ␣ or ␤␥ to effectors.
More critical and microscopic analyses of the interactions of the two cytosolic domains of adenylyl cyclase with each other and with their regulators are necessary. Although this need has been recognized for a long time, progress has been greatly limited by the availability and purity of reagents. Milligram quantities of G protein-regulated adenylyl cyclase domains may now be prepared readily, as can a high-affinity complex of G s␣ with the crucial components of the cyclase. We hope that all relevant tools can now be utilized to understand the complex regulatory features of these interesting enzymes.