Purification and characterization of a soluble form of mammalian adenylyl cyclase.

An engineered, soluble form of mammalian adenylyl cyclase has been expressed in Escherichia coli and purified by three chromatographic steps. The enzyme utilizes one molecule of ATP to synthesize one molecule of cyclic AMP and pyrophosphate at a maximal specific activity of 12.8 micromol/min/mg, corresponding to a turnover number of 720 min-1. Although devoid of membrane spans, the enzyme displays all of the regulatory properties that are common to mammalian adenylyl cyclases. It is activated synergistically by Gsalpha and forskolin and is inhibited by adenosine (P-site) analogs with kinetic patterns that are identical to those displayed by the native enzymes. The purified enzyme is also inhibited directly by the G protein betagamma subunit complex. After adenovirus-mediated expression in adenylyl cyclase-deficient HC-1 cells, the enzyme can be stimulated synergistically by Gs-coupled receptors and forskolin.

Ten isoforms of mammalian adenylyl cyclase have been identified during the past 7 years, and these enzymes display a complex array of both shared and distinct regulatory properties (1,2). All members of the family are activated by the ␣ subunit of the heterotrimeric G protein G s , and receptor-mediated liberation of the GTP-bound form of G s␣ is the dominant mechanism for stimulation of cyclic AMP synthesis. Furthermore, all of the characterized mammalian adenylyl cyclases are activated by the diterpene forskolin (3); stimulation of catalysis by forskolin and G s␣ is highly synergistic in many cases (type II and IV-VI adenylyl cyclases). Finally, all family members are also inhibited by adenosine analogs known as P-site inhibitors (4). The physiological significance of P-site inhibition is not clear, but at least some relevant compounds appear to be present at sufficient concentrations in vivo to affect adenylyl cyclase activity (5).
Each mammalian adenylyl cyclase is thought to contain a short and variable amino terminus, followed by two repeats of a unit composed of six transmembrane spans and a large (ϳ40 kDa) cytosolic domain (6). Although there is little sequence homology among the various enzymes within the transmembrane spans, the overall identity of the different adenylyl cyclases within the cytosolic regions (denoted C 1 and C 2 ) ranges from 50 to 90%. The most highly conserved sequences are within the amino-terminal half of each cytosolic domain (C 1a and C 2a ). Furthermore, these domains are ϳ50% similar to each other (within a single adenylyl cyclase) and to the catalytic domains of the guanylyl cyclases.
Attempts to understand the mechanisms of regulation of adenylyl cyclases and to relate their common and distinct features to structure have been frustrated both by low levels of expression and the requirement for detergent for solubility. To overcome these hurdles, Tang and Gilman (7) constructed a soluble adenylyl cyclase by linkage of the C 1a domain of the type I enzyme with the C 2 domain of the type II enzyme. Expression of this molecule in Escherichia coli complemented the catabolic defect in an adenylyl cyclase-deficient strain of the bacterium. The crude lysate of such bacteria displayed adenylyl cyclase activity that was activated synergistically by G s␣ and forskolin and that was inhibited by 2Ј-d3Ј-AMP, 1 a potent P-site analog. To characterize this molecule further, we have developed methods to improve its expression and to purify it in a hexahistidine-tagged form. The basic biochemical properties of this protein are described herein.

MATERIALS AND METHODS
Plasmid Construction-The expression vector pTrcH6 was created by ligation of phosphorylated linkers (5Ј-CATGCATCACCATCACCAT-CACGCCGCCATGGA and 5Ј-GATCTCCATGGCGGCGTGATGGTG-ATGGTGATG) with NcoI-and BamHI-digested pTrcHisA (Invitrogen, San Diego, CA). To produce a construct in frame with the hexahistidine tag, the 1.64-kilobase BspHI-HindIII fragment of pTrcIC 1 IIC 2 L 3 , which encodes the soluble adenylyl cyclase (7), was ligated with pTrcH6 that had been digested with NcoI and HindIII. The BspHI-HindIII fragment of pTrcIC 1 IIC 2 L 3 was also cloned into the NcoI-and HindIII-digested H6pQE60 vector (8). The adenovirus shuttle vector pH 6 (271)I 1 II 2 L 3 -ACCMV was created by ligation of the 1.6-kilobase EcoRI-KpnI fragment from H 6 pQE(271)I 1 II 2 L 3 with the vector ACCMV after digestion with the same enzymes.
Antibodies-Two peptides were synthesized corresponding to amino acid sequences in the C 1 domain of type I adenylyl cyclase and the C 2 domain of type II adenylyl cyclase: IC 1 -452 (CGDYEVEPGHGH-ERNSF) and IIC 2 -870 (CRSLKNEELYHQSYD). The peptides were coupled to the purified protein derivative of tuberculin (Stantens Seruminstitut, Copenhagen) (9). Antibodies were produced in New Zealand White rabbits and affinity-purified as described (10).
The frozen cells were resuspended with a Polytron in 1.5 liters of lysis buffer (50 mM Tris-HCl (pH 8.0), 10 mM ␤-mercaptoethanol, 50 mM NaCl, 0.5 mg/liter aprotinin, and mixed protease inhibitors), followed by the addition of 0.2 mg/ml lysozyme while stirring. After 30 min at 4°C, 32 mg of DNase and 5 mM MgCl 2 were added, and the suspension was incubated for 30 min. The membranes and cell debris were pelleted by centrifugation for 30 min at 100,000 ϫ g. The supernatant was supplemented with NaCl (250 mM final concentration) and loaded onto a 10-ml Ni 2ϩ -NTA column equilibrated with lysis buffer. The column was washed with 25 column volumes of buffer A (50 mM Tris-HCl (pH 8.0), 10 mM ␤-mercaptoethanol, 2 mM MgCl 2 , 400 mM NaCl, 5 mM imidazole, and mixed protease inhibitors), followed by 15 column volumes of buffer A containing 15 mM imidazole and 10 column volumes of buffer B (buffer A, except with 10 mM NaCl and 15 mM imidazole). The hexahistidine-tagged protein was eluted with 50 ml of buffer B containing 150 mM imidazole. The Ni 2ϩ -NTA eluate was filtered and loaded onto a Mono Q HR 10/10 fast protein liquid chromatography column. The column was washed with 40 ml of buffer C (20 mM NaHepes (pH 8.0), 2 mM MgCl 2 , 1 mM EDTA, and 2 mM dithiothreitol) and then eluted with a gradient of 0 -400 mM NaCl. Fractions containing adenylyl cyclase activity were pooled, and (NH 4 ) 2 SO 4 was added to a final concentration of 0.4 M. This solution was loaded onto two phenyl-Superose HR 5/5 columns run in tandem, and protein was eluted with two sequential gradients (using buffer C): 0.4 to 0 M (NH 4 ) 2 SO 4 over 30 ml, followed by 0 -15 mM CHAPS over 30 ml. Fractions were pooled based on adenylyl cyclase activity and purity (silver-stained SDS-polyacrylamide gels). The protein was stored in small aliquots at Ϫ70°C in buffer C containing 11 mM CHAPS.
Adenylyl Cyclase Assays-Adenylyl cyclase activity was measured as described by Smigel (13). All assays were performed for 15-20 min at 30°C in a final volume of 100 l unless stated otherwise. The final concentration of MgCl 2 was 10 mM. Adenylyl cyclase and G proteins were first incubated on ice in a total volume of 40 l. Experiments involving the kinetics of P-site inhibition and determinations of K m were initiated by the addition of enzyme. Assays that contained G protein ␤␥ subunits were performed in the presence of 1.2 mM CHAPS.
To quantitate both the reactants and products of the adenylyl cyclase reaction, incubations were performed in the absence of an ATP-regenerating system with either ␣-32 P-or ␥-32 P-labeled ATP. [␥-32 P]ATP was used to measure production of pyrophosphate using a modification of the charcoal method (14) in which charcoal was equilibrated with 50 mM phosphate and 50 mM pyrophosphate; this permitted the complete recovery of both molecules. [␣-32 P]ATP was used to quantitate loss of ATP and production of cyclic AMP after their separation by Dowex and alumina chromatography (15). A portion of each reaction mixture and the separated products were analyzed by thin layer chromatography in Tb buffer (16) to assess their purity and to quantitate contaminants in the labeled ATP. Thin layer chromatography demonstrated no production of phosphate during the course of the reaction. Recovery of cyclic AMP during chromatography was determined by quantitation of [ 3 H]cAMP, which was added at the end of the reaction.
Production of Recombinant Adenovirus-The HC-1 and 293 cell lines were maintained in Dulbecco's minimal essential medium (Life Technologies, Inc.) supplemented with 10% heat-inactivated calf serum. To generate recombinant adenovirus, 1 g of the shuttle vector pH 6 (271)I 1 II 2 L 3 ACCMV and 4 g of the plasmid pJM17 (17) were cotransfected into 293 cells using the modified bovine serum transfection kit (Stratagene) according to the manufacturer's protocol. After transfection, the culture medium was changed every 2-3 days until cytopathic effects were observed. Primary virus was identified by polymerase chain reaction (18) using primers within the cytomegalovirus early gene promoter and the coding region of the adenylyl cyclase gene. The virus was plaque-purified (19), and the appropriate viral plaques were amplified and stored in cell culture medium. Viral titers were estimated by plaque assay and ranged from 0.5 to 1.5 ϫ 10 5 plaqueforming units/ml. The appropriate ratios for infection of HC-1 cells were determined by immunoblotting and by assay of adenylyl cyclase or ␤-galactosidase activity in cytosolic fractions of infected cells.
Cyclic AMP Accumulation in HC-1 Cells-Changes in intracellular concentrations of cyclic AMP were measured by determining the ratio of radioactivity in cyclic AMP to that in the total ATP and ADP pools in [ 3 H]adenine-labeled cells (20). HC-1 cells growing in 12-well plates (0.4 ϫ 10 6 cells/well) were infected with 100 -150 l of adenovirus encoding ␤-galactosidase or the soluble adenylyl cyclase. Medium was removed after 30 h and replaced with 1.5 ml of medium containing 2 Ci/ml [ 3 H]adenine. After an additional 18 h, the cells were washed once with Dulbecco's minimal essential medium containing 20 mM NaHepes (pH 7.4) and then treated at 37°C with 1.0 mM 1-methyl-3isobutylxanthine (a cyclic-nucleotide phosphodiesterase inhibitor) and various effectors as indicated. Incubations were terminated by aspiration of medium and addition of 0.9 ml of ice-cold 5% trichloroacetic acid containing 0.1 mM cyclic AMP. The cells remained on ice for 45 min, and [ 32 P]cAMP (5000 cpm) was added to each well to quantify recoveries from columns. Acid-soluble nucleotides were separated on Dowex and alumina columns according to the method of Salomon et al. (15). Assays were performed in duplicate, and the values shown are representative of three experiments.

RESULTS AND DISCUSSION
Expression and Purification of the Soluble Adenylyl Cyclase-Expression of the hexahistidine-tagged soluble adenylyl cyclase (H 6 (271)I 1 II 2 L 3 ) in E. coli results in the accumulation of ϳ2-5 nmol of adenylyl cyclase activity/min/mg in the cytosol. However, Ͼ90% of the adenylyl cyclase protein is found in inclusion bodies and cannot be solubilized in an active form. Growth of the bacteria at room temperature permitted the maximal accumulation of soluble and active protein. Concurrent expression of thioredoxin or the chaperonins GroEL and GroES increased the amount of adenylyl cyclase protein in the cytosol (detected by immunoblotting), but failed to increase adenylyl cyclase activity in these cells (data not shown). A proteolysis product representing the IIC 2 domain (detected by antibody IIC 2 -870) also accumulated (Fig. 1A), but fragments containing the IC 1 domain were not detected with antibody IC 1 -454 (data not shown).
The soluble adenylyl cyclase activity in the lysate can be enriched 50-fold using Ni 2ϩ -NTA chromatography ( Table I).
The yield is only ϳ30%, and much of the lost activity is not found in the flow-through or washes of the column. The protein is susceptible to proteases in crude extracts, and rapid manipulation at this stage improves recovery. Chromatography over the Mono Q column results in only a 3-fold purification, but several major contaminants are removed. The low enrichment is due to a broad elution profile on this and many other columns (Fig. 1C). The protein is eluted from the final (phenyl-Superose) column with CHAPS in the absence of salt. Although the detergent can be removed with retention of enzymatic activity, the stability of the protein is greatly enhanced by CHAPS during freezing and thawing. The soluble adenylyl cyclase migrated on sodium dodecyl sulfate-containing polyacrylamide gels with an apparent M r of 55,000, as judged by silver staining and immunoblotting, and appeared to be Ͼ90% pure (Fig. 1A); the predicted M r of the protein is 56,900.
Our functional criterion for adequate purification was the absence of interfering enzymatic activities (ATPases, phosphodiesterases, pyrophosphatases, etc.). For each molecule of ATP consumed by the preparation, one molecule of cyclic AMP and PP i were produced (Fig. 2). There is no detectable additional consumption of ATP even in the absence of any stimulator of adenylyl cyclase (data not shown). The enzymatic reaction approaches the equilibrium for the interconversion of ATP, cyclic AMP, and PP i as measured by Hayaishi et al. (21) with the adenylyl cyclase from Brevibacterium liquefaciens (K eq ϭ 0.065 M). Under conditions where ATP is not limiting, enzymatic activity is linear for Ͼ90 min (data not shown). Since mammalian adenylyl cyclases have two similar domains that could be involved in nucleotide binding, it is significant that only one molecule of ATP is utilized for each molecule of cyclic AMP that is synthesized. Any conformational changes that occur upon activation of the enzyme are not mediated by hydrolysis of ATP at a regulatory site.
Activation of the Soluble Adenylyl Cyclase-Although the soluble adenylyl cyclase is a chimera and lacks its integral membrane domains, it displays most of the regulatory features that are characteristic of its native counterparts. The enzyme has very low basal activity (20 -50 nmol/min/mg), but can be stimulated Ͼ100-fold by GTP␥S-G s␣ or forskolin. The K m for Mn 2ϩ -ATP is ϳ50 M and is little affected by the presence of enzyme activators. However, the K m for Mg 2ϩ -ATP remains relatively low when the enzyme is stimulated by GTP␥S-G s␣ (65 M), but increases to 620 M in the presence of forskolin (Fig. 3). When both GTP␥S-G s␣ and forskolin are present, the V max is 12.8 mol/min/mg, and the K m remains elevated (300 M). Similar changes have been observed with adenylyl cyclase from bovine brain (22) or with type I adenylyl cyclase expressed in Sf9 cells (23); the latter enzyme has a K m of 38 M with Mn 2ϩ and forskolin and a K m of 440 M with Mg 2ϩ and forskolin. Hill plots of the data in Fig. 3 were linear and gave a value of n ϭ 1.0 with each activator (data not shown).
The turnover number for the soluble chimeric adenylyl cyclase is similar to the values for the native forms of the enzyme.  The turnover number is 360 min Ϫ1 in the presence of forskolin and increases to 720 min Ϫ1 with both forskolin and GTP␥S-G s␣ . The turnover numbers for the purified type I and II adenylyl cyclases are 890 and 260 min Ϫ1 , respectively, in the presence of forskolin and Mn 2ϩ (24). The low V max for the GTP␥S-G s␣stimulated soluble adenylyl cyclase (Fig. 3) is due only to the amount of GTP␥S-G s␣ used in the assay compared with the EC 50 for activation by G s␣ (Fig. 4A). With maximal amounts of GTP␥S-G s␣ , the soluble adenylyl cyclase displays a V max approaching that observed in the presence of both GTP␥S-G s␣ and forskolin (Fig. 4A).
The affinity of the soluble adenylyl cyclase for activated E. coli-derived G s␣ is ϳ100-fold lower than that of type I adenylyl cyclase (Fig. 4A) (25). However, forskolin potentiates the effect of low concentrations of GTP␥S-G s␣ , shifting the apparent affinity of the soluble adenylyl cyclase for GTP␥S-G s␣ by 2 orders of magnitude. For example, adenylyl cyclase activity was 9.2 mol/min/mg in the presence of 225 nM GTP␥S-G s␣ and 2 M forskolin, while the activities in the presence of these activators alone were 1.2 and 0.8 mol/min/mg, respectively. This striking synergy is typical of type II adenylyl cyclase, but is not characteristic of the type I enzyme (25,26). Half-maximal activation of the soluble adenylyl cyclase was observed with 3 M forskolin (Fig. 4B), the same concentration required for type I and II adenylyl cyclases (22). Low concentrations of GTP␥S-G s␣ shift the apparent affinity of forskolin by 2 orders of magnitude. Unlike the situation with GTP␥S-G s␣ , maximal activity in the presence of forskolin is significantly less than that observed with GTP␥S-G s␣ or GTP␥S-G s␣ plus forskolin (Fig. 4B).
Ca 2ϩ -calmodulin had no stimulatory effect on the soluble adenylyl cyclase (data not shown); calmodulin activates type I adenylyl cyclase. However, the C 1b region of the type I enzyme that has been implicated in calmodulin binding is not present in the soluble protein studied here (27)(28)(29).
Inhibitors of the Soluble Adenylyl Cyclase-Type I and II adenylyl cyclases differ dramatically in their responses to G i␣ proteins and the G protein ␤␥ subunit complex. G i␣ , G o␣ , and ␤␥ all inhibit type I adenylyl cyclase (1,30). The type II enzyme is not affected by G i␣ and is greatly stimulated by ␤␥ in the presence of G s␣ (1,30). The chimeric type IC 1 /type IIC 2 soluble enzyme is unresponsive to G i␣ , even at 10 M concentrations (data not shown), but it can be inhibited almost completely by ␤␥ (Fig. 5). This is surprising since the extent of inhibition of purified type I adenylyl cyclase by ␤␥ is modest (30%) compared with the more dramatic inhibition of the membrane-bound protein (24). Inhibition of the soluble adenylyl cyclase is not due to interaction of ␤␥ with G s␣ since ␤␥ inhibits forskolinand G s␣ -activated enzymatic activity equally. The apparent affinity of the soluble adenylyl cyclase for ␤␥ is substantially lower (10-fold or more) than that of type I or II adenylyl cyclase. However, inhibition is still dependent on prenylation of the ␥ subunit since the nonprenylated ␤ 1 ␥ 2 Cys 68 3 Ser mutant was inactive at the highest concentrations tested (Fig. 5).
Chimeras have been used successfully to identify regions necessary for activation of type I and II adenylyl cyclases by calmodulin and protein kinase C, respectively (31). By coexpression of combinations of membrane-bound halves of type I and II adenylyl cyclases (noncovalent chimeras), Tang and Gilman (30) tentatively assigned the site for synergistic activation of adenylyl cyclase by ␤␥ to the carboxyl-terminal half of the molecule. Chen et al. (32) made a peptide corresponding to the sequence of a region in the C 2 domain of type II adenylyl cyclase that blocks the interactions of ␤␥ with several effectors, consistent with the assignment made by Tang and Gilman. It is thus perhaps surprising that the soluble adenylyl cyclase described herein is inhibited by ␤␥ rather than activated. However, we are hesitant to draw further conclusions about this phenomenon until comparisons can be made with nonchimeric soluble forms of the type I and II enzymes and until binding interactions between ␤␥ and these domains can be examined definitively.
It is not clear why G s␣ and ␤␥ have reduced affinities for the soluble adenylyl cyclase relative to its membrane-bound counterparts or what role the membrane spans play in signal transduction. ␤␥ is associated with the plasma membrane, at least in part because of prenylation of ␥. Its apparent higher affinity for membrane-bound adenylyl cyclase may be due partially to the concentrating effect of membrane localization. It is also possible that ␤␥ interacts directly with the membrane-spanning regions of adenylyl cyclase or normally interacts with both the C 1 and C 2 domains and that these interactions are constrained in the chimeric molecule. Despite these uncertainties, the fact that nonprenylated ␤␥ failed to inhibit the soluble adenylyl cyclase points to specific interactions that are dependent on the lipid modification, rather than to simple concentration at the membrane. Similar arguments may apply to G s␣ . However, the recombinant (E. coli-derived) G s␣ used here does not contain lipid modifications and already has a lower affinity for membrane-bound adenylyl cyclases than does tissue-derived G s␣ (33). All mammalian adenylyl cyclases are activated by G s␣ in vitro with few quantitative differences. It thus seems unlikely that the chimeric nature of the soluble enzyme would contribute to the loss of affinity. Again, it seems most likely that domains of adenylyl cyclase removed in the engineered protein (e.g. loops between membrane spans, C 1b , etc.) or conformational constraints imposed by linkage of C 1 and C 2 are responsible for the loss of affinity.
Natural, membrane-bound mammalian adenylyl cyclases are inhibited directly by adenosine analogs known as P-site inhibitors, so designated in reference to the requirement for an intact purine ring (4). The most potent inhibitors are 2Ј-or 5Ј-deoxy and 3Ј-phosphoryl compounds (34). P-site inhibition displays several unique features. Stimulated forms of the enzyme are substantially more sensitive to inhibition than are nonactivated forms. Inhibition is dependent on metal and is characteristically noncompetitive or mixed noncompetitive with respect to metal-ATP (35-37). However, inhibition of the G s␣ -activated enzyme is noncompetitive with respect to Mg 2ϩ -ATP (38). All of these features are preserved in the soluble adenylyl cyclase. This enzyme is significantly more sensitive to 2Ј-d3Ј-AMP when activated by forskolin, GTP␥S-G s␣ , or the combination of the two regulators than when assayed with Mn 2ϩ alone (Fig. 6). Thus, the GTP␥S-G s␣ -and forskolin-stimulated enzyme, which is 100-fold more active than the Mn 2ϩstimulated enzyme, is Ͼ300 times more sensitive to P-site inhibition (IC 50 ϭ 3 M versus ϳ1 mM). The IC 50 of bovine brain adenylyl cyclase for 2Ј-d3Ј-AMP is 2-3 M in the presence of forskolin or GTP␥S-G s␣ (35). P-site inhibition of the GTP␥S-G s␣ -stimulated soluble adenylyl cyclase is noncompetitive with respect to Mg 2ϩ -ATP. This is also true of detergent-solubilized adenylyl cyclase activity in rat brain (38). Noncompetitive inhibition is rare in unireactant systems and is characterized by a lower V max , a decreased apparent K m , and an unchanged K m /V max . The interpretation is that the inhibitor binds only to the enzyme-substrate complex, yielding an inactive enzyme. All other activated forms of the soluble adenylyl cyclase show a mixed noncompetitive pattern of inhibition with respect to metal-ATP, where both 1/V max and K m /V max increase with increasing inhibitor concentrations (Fig.  7, B-D). True noncompetitive inhibition implies that the inhibitor binds with equal affinity to the free enzyme and the enzyme-substrate complex; inhibitor binding is thus independent of substrate concentration, and the K m remains constant. V max , K m , and V max /K m all decrease with increasing concentrations of inhibitor when the soluble adenylyl cyclase is stimulated with forskolin (Mg 2ϩ or Mn 2ϩ ) or GTP␥S-G s␣ in the presence of Mn 2ϩ .
The x intercepts of replots of 1/V max or K m /V max versus the concentration of 2Ј-d3Ј-AMP represent the values of ␣K i and K i , respectively, where K i is the dissociation constant of 2Ј-d3Ј-AMP from the enzyme-inhibitor complex and ␣K i is the dissociation constant of 2Ј-d3Ј-AMP from the enzyme-substrate-inhibitor complex (Table II) (39). The calculated ␣K i values correspond to the observed IC 50 values for P-site inhibition and correlate well with those shown in Fig. 6. The factor ␣ is equal to 1 in the case of pure noncompetitive inhibition, Ͼ1 when the inhibitor favors binding to the free enzyme, and Ͻ1 when the inhibitor favors binding to the enzyme-substrate complex. The values of ␣ determined here are Ͻ1 in all cases, showing a clear preference of the inhibitor for the adenylyl cyclase-ATP complex compared with adenylyl cyclase alone. Despite these observations, it has not been possible to make a simple assignment of regulatory or catalytic function to the C 1 or C 2 domain by a mutagenic approach (23).
The role of metals in influencing K m , V max , and the inhibition patterns is complex. Mn 2ϩ (compared with Mg 2ϩ ) decreases dramatically the K m for ATP in the presence of forskolin, but has no effect on the K i for a P-site inhibitor with this activator. When GTP␥S-G s␣ is the activator, the type of metal determines if 2Ј-d3Ј-AMP binds solely to the adenylyl cyclase-ATP complex or to both this complex and free adenylyl cyclase. Mn 2ϩ can also influence the V max , usually increasing activity by ϳ2-fold. Whether the metal causes each of these changes by working at the active site, an alternative metal-binding site, or both is still unclear.
Adenovirus-mediated Synthesis of the Soluble Adenylyl Cyclase in HC-1 Cells-The rat hepatoma HC-1 cell has no detectable adenylyl cyclase activity, but contains G s␣ and binding activity characteristic of the ␤-adrenergic receptor (40). Although these cells constitute an ideal null background in which to express adenylyl cyclase, they are difficult to transfect. We thus constructed a recombinant adenovirus in which expression of the soluble adenylyl cyclase is driven by the cytomegalovirus early gene promoter. After infection of cells at virus/cell ratios of 10:1 to 40:1, adenylyl cyclase activity (0.7 nmol/min/ mg) and immunoreactivity were detectable. We have used this system to study receptor-mediated cyclic AMP synthesis in vivo. An adenovirus encoding ␤-galactosidase was utilized as a control.
There was no detectable accumulation of cyclic AMP in noninfected HC-1 cells or in those expressing ␤-galactosidase after exposure to isoproterenol and/or forskolin (Fig. 8). Similarly, cyclic AMP did not accumulate in cells expressing the soluble adenylyl cyclase after treatment with isoproterenol alone. However, there was substantial accumulation of cyclic AMP in such cells after incubation with forskolin, and, for data averaged over five time points, isoproterenol further enhanced cyclic AMP accumulation by 50%. Despite the absence of all transmembrane-spanning sequences, the soluble adenylyl cyclase can be activated by a G protein-coupled receptor when its apparent affinity for G s␣ is increased by forskolin.
The failure to detect stimulation of cyclic AMP synthesis in the presence of isoproterenol alone was not unexpected. The apparent affinities of the soluble adenylyl cyclase for the stimulator G s␣ and the inhibitor ␤␥ are similar. However, in the presence of forskolin, the apparent affinity of the enzyme for G s␣ is increased dramatically, while that for ␤␥ is not altered. Furthermore, the soluble adenylyl cyclase studied here is not unique in its failure to respond to G s␣ in vivo in the absence of forskolin. When expressed by transfection, several membranebound adenylyl cyclases have responded only to synergistic combinations of forskolin and appropriate receptor agonists (31,(41)(42)(43).
Since the soluble adenylyl cyclase studied here displays most of the regulatory features that are characteristic of the membrane-bound enzymes (and cytosolic domains necessary for other phenomena, such as activation by calmodulin and inhibition by G i␣ , may be missing), one wonders if the membrane spans of the native enzymes play only quantitative roles and are not important qualitatively for regulation of cyclic AMP synthesis. If so, what is the explanation for the complex topology of these enzymes, which is conserved in all of its known isoforms? The best current hypothesis is that of Reddy et al. (44), who have proposed that the activity state of at least certain mammalian adenylyl cyclases may be responsive to transmembrane potential.
In summary, a soluble form of mammalian adenylyl cyclase that displays all of the regulatory features that are common to the various isoforms of the enzyme can be synthesized in E. coli and purified. This approach offers many advantages over study of the native enzymes, which can be obtained only in microgram quantities from native sources or after expression in heterologous systems. We hope that the availability of this FIG. 8. Synergistic stimulation of soluble adenylyl cyclase activity by forskolin and isoproterenol in HC-1 cells. HC-1 cells infected with ␤-galactosidase adenovirus (␤-gal; E) or H 6 (271)I 1 II 2 L 3 adenovirus (H 6 -AC; q, å, f) were exposed to 10 M isoproterenol (INE; f), 50 M forskolin (Fsk; å), or 10 M isoproterenol ϩ 50 M forskolin (E, q) as described under "Experimental Procedures." Relative accumulation of cyclic AMP was determined as described under "Experimental Procedures." Each time point was performed in duplicate, and the experiment shown is representative of three experiments.

TABLE II Effects of activators and P-site inhibitors on K m , V max , and K i values for the soluble adenylyl cyclase
The K m for ATP and V max were determined by Eadie-Hofstee analysis. Reactions were performed as described in the legend to Fig. 7 The ␣K i for activation by Mn 2ϩ is an approximation based on the IC 50 from Fig. 6 and Eadie-Hofstee analysis (data not shown). b Fsk, forskolin.
protein and related constructs will permit detailed appreciation of mechanisms of regulation of cyclic AMP synthesis and their structural counterparts.