Expression and Characterization of Calmodulin-activated (Type I) Adenylylcyclase"

A complementary DNA that encodes a bovine brain, calmodulin-sensitive (type I) adenylylcyclase has been inserted into the baculovirus genome under the control of the strong polyhedron promoter. Expression of the recombinant adenylylcyclase in Sf9 cells using recom- binant baculovirus increases adenylylcyclase activity in cell membranes to 10-20 nmol*min"-mg" (approx-imately 0.1% of membrane protein). The catalytic ac- tivity of the recombinant adenylylcyclase can be stimulated by G,, calmodulin, or forskolin, and it can be inhibited by adenosine analogs and by G protein B-y subunits. The specific activity of the purified recom- binant protein approximates 5 pmol*min"*mg-'. This is similar to that of the enzyme purified from bovine brain. Type I adenylylcyclase has a quasiduplicated structure. There are two membrane-spanning domains, each with six putative transmembrane helices, and there are two presumed nucleotide-binding domains that are about 55% similar to each other. No catalytic activity is detectable when each half of the adenylylcyclase molecule is expressed by itself. However, coexpression of the two halves results in considerable enzymatic activity. Interaction between the two halves of adenylylcyclase may be necessary for catalysis.

calmodulin. Although there are distinct regions of the four proteins that differ markedly in their amino acid sequence, hydropathy analysis suggests that they all share the same general topology with regard to the membrane. Each protein is hypothesized to contain two domains that are associated with the lipid bilayer, each of these consisting of six transmembrane helices. There appear to be two large cytoplasmic domains; the first of these lies between the two sets of membrane-spanning helices, and the second is at the carboxyl terminus (see Fig. 8).
Expression of individual forms of adenylylcyclase and alterations of these proteins offer obvious paths to their more detailed characterization. We describe herein the expression of type I adenylylcyclase in Sf9 (fall army worm ovarian) cells using the recombinant baculovirus expression system (21-23). We document the utility of this approach for expression of this complex protein and present the results of its initial characterization.

MATERIALS AND METHODS
Plasmid Construction-For convenience of cloning, a 4.1-kb EcoRI fragment of DNA containing the entire coding sequence of type I adenylylcyclase was transferred from pMKAC (18) to the EcoRI site of pBluescript SK-. The resulting plasmid was linearized by treatment with Hind111 and was then digested with exonuclease 111 and nuclease S1 to delete approximately 100 bp of untranslated sequence from the 5' end of the cDNA (24). After treatment with EcoRI, a mixture of 4 kb of DNA was isolated and cloned into pBluescript SK-(digested with SmaI and EcoRI). Clones were screened by doublestranded DNA sequencing, and one clone (designated pSKACA58) was selected in which all of the putative 5"noncoding region was removed and a new NcoI site was generated at the initiating methionine residue. To create a better site for initiation of translation (AXXATGG) (25), an EcoRI-bluntedNcoI fragment was excised from pSKACA58 and ligated into pBluescript SK-that had been cut with EcoRV and EcoRI; this plasmid is designated pSKACA58-13. For expression of adenylylcyclase in human 293 cells and simian COS-m6 cells, a 4.0-kb fragment of DNA was excised from pSKACA58-13 with Hind111 and XbaI and was ligated into pCMV-5 that had been digested with the same enzymes; this plasmid is designated pCMVACA58. pCMVAC was described previously as pCMVAC-N (18). For construction of recombinant baculovirus, the NcoI-EcoRI fragment from pSKACA58 was ligated into baculovirus vector pAcC4.
For construction of mutant adenylylcyclases, convenient restriction enzyme sites were chosen to release DNA fragments, and oligonucleotides were ligated with these fragments to create initiation or stop sites. For mutant AN, a SacII-EcoRI fragment was isolated from pSKACA58. For mutants NMICI, fragments were isolated from pCMVACA58 by digestion with EcoRI and partial digestion with RsaI. For mutant M,C, a HincII-EcoRI fragment was isolated from pSKACA58. These fragments were ligated with phosphorylated adaptors (sequences listed below for each mutant) and pBluescript SKthat had been digested with XbaI and EcoRI. The mutant AN was cloned into the XbaI and EcoRI sites of the baculovirus vector pVL1393. The EcoRI-blunted XbaI fragments of mutants NMICl and MzCz were cloned into the SmaI and EcoRI sites of baculovirus vectors pVL1392 and pVL1393, respectively.

5' C T A G A A A C A T G G G C 3' T T T G T A C C
Adapter NMIC,
Production of Recombinant Baculouirus-To generate recombinant baculovirus, Sf9 cells (2 X IO6) were transfected with 1 pg of wildtype viral DNA and 18 pg of baculovirus vector containing DNA encoding wild-type or mutant adenylylcyclase; a calcium phosphate precipitation method was used as described (27). The virus was harvested after 6 days and was screened by limiting dilution and dot hybridization (28). Positive clones were then screened with plaque assays (27). One or two rounds of plaque purification were usually sufficient to obtain pure virus. Viral titers were estimated by end point dilution (27). Antibodies-Two peptides were synthesized corresponding to amino acid sequences in the C, and C, domains of type I adenylylcyclase: CI-251 (CIEDRLRLEDENEK) and CI-1115 (CGLAPGPPG-QHLPPGASGKEA); the arabic number for each peptide designates the position of the initial residue in the adenylylcyclase sequence. The peptides were coupled to the purified protein derivative of tuberculin (with 0.042% glutaraldehyde in phosphate-buffered saline) (29). Purified protein derivative of tuberculin was obtained from the Statens Seruminstitut, Copenhagen, Denmark.
New Zealand White rabbits were injected subcutaneously (four sites on the back) with a total of 200 pg of the peptide-carrier protein complex in complete Freund's adjuvant. A boost of 100 pg of antigen in incomplete Freund's adjuvant was administered 3 weeks later. Rabbits were bled every 2 weeks thereafter until the titer to peptide decreased noticeably as judged by enzyme-linked immunosorbent assay. A boost of 100 pg of antigen in incomplete Freund's adjuvant was then administered. Antisera (Ab CI-251 and Ab CI-1115) were affinity-purified as described (30), except that the initial binding of the diluted serum to the resin was performed by rocking the mixture for 1 h at room temperature. Bovine brain adenylylcyclase was partially purified by forskolin-Sepharose chromatography (18). Reactivity of the antisera was ini-tially determined by probing nitrocellulose to which varying amounts of the forskolin-Sepharose eluate had been transferred. Membranes from infected Sf9 cells and bovine brain were heated to 80 "C in the presence of 2% SDS and 0.2 mM dithiothreitol for 5 min; they were then treated with 50 mM N-ethylmaleimide for 10 min prior to electrophoresis, transfer to nitrocellulose, and immunoblotting (31).
This treatment significantly reduced the smearing of type I adenylylcyclase during SDS-PAGE.
Transient Expression-Human 293 cells (lo6 per 60-mm culture dish) were transfected with 2 pg of test DNA (pCMV constructs) plus 12 pg of carrier DNA (pUC-18) by calcium phosphate precipitation (32); chloroquine (100 p~) was added during transfection to reduce lysosomal degradation of DNA. Transfection of COS-m6 cells was performed as described (18), and membranes were harvested 48 h later. Protein concentrations were determined by dye binding (33), using bovine serum albumin as a standard. Adenylylcyclase assays were performed as described for 20 min, using approximately 10 pg of membrane protein (5).
Expression of Adenylylcyclase in Sf9 Celk"sf9 cells were usually infected with 1 plaque-forming unit per cell of baculovirus and were harvested 54-60 h after infection. Cells were lysed and membranes were washed and resuspended in HME buffer (20 mM NaHepes (pH 8.0), 2 mM MgCl,, 1 mM EDTA) plus protease inhibitors B, as buffer (20 mM Hepes (pH 8.01, 1 mM EDTA, 1 mM EGTA, 2 mM described (18). On some occasions membranes were washed in HEED DTT) containing protease inhibitors B prior to suspension in HME.
Adenylylcyclase activity was assayed for 10 min in the presence of 10 mM MgCl, using 5 pg of membrane protein or for 5 min using appropriate aliquots of the purified enzyme (5). rG., was activated for 30 min at 30 "C in 50 mM NaHepes (pH 8.0), 1 mM EDTA, 1 mM DTT, 5 mM MgSO,, and 100 p M GTPyS (34). Unbound GTPyS was removed by gel filtration; GTPrS does not dissociate from G,, over the course of several hours if the concentration of free Mg2' is greater than 1 p~. The stoichiometry of GTPrS binding was usually between 50 and 70%. Calmodulin and/or activated G,, was incubated with membranes for 10 min or with the purified enzyme for 2 min at 30 "C prior to assay; forskolin, P-site inhibitors, and/or G protein Pr subunits were added immediately prior to assay. Assays that contained G protein & subunits were performed in the presence of 0.1% sodium cholate. Assays that contained calmodulin also contained 50 or 100 p~ CaC1,.
Purification of Recombinant Adenylylcycluse-Sf9 cells (1 X lo9 in 1 liter) were infected with recombinant baculovirus containing the adenylylcyclase cDNA. Cells were harvested after 54 h, suspended at a density of 5 X 106/ml in 20 mM NaHepes (pH 8.0), 2 mM DTT, 1 mM EGTA, 5 mM EDTA, 150 mM NaC1, and protease inhibitors C, and lysed by nitrogen cavitation at 550 p.s.i. for 30 min. Nuclei were removed by centrifugation for 10 min at 500 X g. Membranes were then collected by centrifugation at 70,000 X g for 30 min. The membrane pellet was suspended by homogenization in 20 mM Na-Hepes (pH 8.0), 1 mM EDTA, 2 mM DTT, 200 mM sucrose, and protease inhibitors C. Lubrol PX was added from a deionized 10% solution to a final concentration of 0.6%. The mixture was homogenized with a motor-driven Teflon homogenizer (three cycles of 20 strokes per cycle) and was rocked gently for 15 min between each cycle. The supernatant was collected after centrifugation for 30 rnin at 50,000 X g. The insoluble material was extracted again with the pooled. NaCl was added to a concentration of 0.5 M, and the detergent same detergent-containing buffer, and the two supernatants were extract was then incubated with 3 ml of forskolin-Sepharose for 9 h at 4 "C with gentle rocking. The resin was poured into a column and washed sequentially with 100 ml of buffer A (10 mM Hepes, pH 8.0, 1 mM EDTA, 1 mM MgCl,, 2 mM dithiothreitol, 0.05% Lubrol Px) containing 0.5 M NaCl and protease inhibitors C; 100 ml of buffer A with 2 M NaCl plus protease inhibitors A and 10 ml of buffer A with 0.5 M NaCl, 2% dimethyl sulfoxide, and protease inhibitors A. Adenylylcyclase was finally eluted by incubation of the column for 15 min with 4 ml of 200 p~ forskolin in buffer A containing 0.5 M NaCl and protease inhibitors A. This elution procedure was repeated four times. Fractions containing purified adenylylcyclase were gel-filtered (Sephadex G-50) to remove forskolin prior to assay. To check the purity of the protein, a portion of the eluate was precipitated with 15% trichloroacetic acid. The precipitate was collected by centrifugation, washed with acetone for 1 h, and pelleted again. Samples were subjected to SDS-PAGE followed by silver staining or immunoblotting. The protein concentration was determined by staining with Amido Black (35).
Biosynthetic Labeling of Recombinant Proteins-Metabolic labeling of Sf9 cells was performed as described (27) with few modifications. Cells were labeled 66 h after infection with 30 pCi of [:"SS]methionine/ 0.5 ml in methionine-free Grace's medium for 4-6 h. Cells were incubated in methionine-free medium for 30-60 min prior to labeling, and medium was changed by sedimentation and resuspension of the cells. Cells were then harvested and lysed in HME buffer containing protease inhibitors B. Supernatant and pellet fractions were separated by centrifugation and were analyzed by SDS-PAGE. For analysis of N-linked glycosylation, tunicamycin (10 pg/ml) was added to the medium 4 h prior to methionine starvation. modestly from 40 to 130 pmol. min" mg" and forskolinactivated enzymatic activity from 380 to 840 pmol.min". mg" (18). The construct used (pCMVAC) has an ATG sequence 8 base pairs upstream from the putative initiation codon, and the 102-base pair 5"noncoding sequence is extremely rich in guanine and cytidine residues (91%). In an attempt to facilitate the expression of adenylylcyclase we removed all of the 5"noncoding sequence and constructed a better site for initiation of translation (ATCATGG). Unfortunately, this construct (pCMVACA58) was only marginally better. Mn2+-activated adenylylcyclase activity was 160 pmol . min" . mg" and forskolin-stimulated activity was 1100 pmol e min" . mg" after transfection of COS-m6 cells.

Transient Expression of
We also attempted to express type I adenylylcyclase transiently in human 293 cells. These cells are readily transfected, and they efficiently express genes driven by the cytomegalovirus early promoter. They also have relatively low basal adenylylcyclase activity: 10-20 pmol. min"mg". Cells transfected with pCMVAC showed a 10-15-fold increase in Mn2+activated adenylylcyclase activity but only a 2-3-fold increase in forskolin-stimulated activity (compared to cells transfected with pCMV5). The effect of expression of type I adenylylcyclase on forskolin-stimulated enzymatic activity is less dramatic because the endogenous adenylylcyclase of 293 cells is markedly activated by forskolin (30-fold). Again, results were only slightly better when pCMVACA58 was used.
The rat HC-1 cell is a hepatoma (HTC) cell variant with little or no detectable adenylylcyclase activity. Attempts to express type I adenylylcyclase in these cells (transiently or permanently) have failed.
Expression of Adenylylcyclase in Sf9 Cells-Recombinant baculovirus (B-rAC) was constructed to contain the cDNA for adenylylcyclase under the control of the strong polyhedron promoter, and adenylylcyclase activity was then assessed in Sf9 cell membranes a t various times after infection with wildtype and recombinant virus (Fig. lA). Uninfected cells displayed specific activities of 25-30 pmol . min" . mg" (Mn2+) or 250 pmol.min". mg" (Mn2+ and forskolin). When cells were infected with wild-type baculovirus, adenylylcyclase activity fell to about 10 pmol. min" . mg-I (Mn") or 100 pmol . min" . mg" (Mn2+ and forskolin) within 2 to 3 days. For cells infected with B-rAC, however, adenylylcyclase activity increased dramatically by 30 h after infection, reaching a max-
imum a t 2-3 days ( 5 nmol. min" . mg" with Mn2+ and 14 nmol .min" .mg" with Mn2+ and forskolin). Based on activity, 99% of the adenylylcyclase in cells infected with B-rAC is the recombinant type I enzyme. We estimate that approximately 0.1% of the protein in crude membranes from infected Sf9 cells is the recombinant protein (based on the specific activity of the purified protein of 8 pmol . min" . mg" ( 5 ) ) .
Six peptides that correspond to various hydrophilic regions of type I adenylylcyclase have been synthesized and used to immunize rabbits. To date, we have obtained two antisera that are useful to immunoblot native and recombinant aden-ylylcyclase in membrane preparations. One of these (Ab CI-251) was raised against a peptide that represents the aminoterminal end of the C1 domain, while the other (Ab CI-1115) was obtained using a peptide that corresponds to the 20 carboxyl-terminal amino acid residues of the protein. These affinity-purified antisera reacted with a protein with an apparent molecular weight of approximately 110,000 in Sf9 cells infected with B-rAC, and the intensity of the signal correlated reasonably well with the adenylylcyclase activity of these membranes, particularly during the first few days after infec-  (0) is also shown. The inset shows fractional adenylylcyclase activity at various concentrations of Pr. tion (Fig. 1B). The subsequent decrement in the intensity of the signals seen with immunoblotting compared to the adenylylcyclase activity in these membranes suggests that there is proteolysis, but that this does not eliminate catalytic activity. Fragments (36 and 43 kDa) of potential interest were visualized with Ab CI-1115; fragments were not detected with Ab Antibodies were also utilized to detect type I adenylylcyclase in bovine brain membranes (Fig. 2). A protein that migrated at 110 kDa was detected with both antisera. Type I adenylylcyclase expressed in Sf9 cells migrated slightly more rapidly than did the bovine brain protein, presumably because of differences in posttranslational modification (e.g. glycosylation). There appeared to be 20-30-fold more type I adenylylcyclase in Sf9 cell membranes than in bovine brain membranes (based on immunoblotting). This is roughly consistent with the measured enzymatic activities. Thus, adenylylcyclase activity in membranes from Sf9 cells infected with B-rAC is 4-fold higher than that observed in bovine brain membranes, and we estimate that only 20-30% of the brain enzyme is type I adenylylcyclase (based on the elution profile from forskolin-Sepharose chromatography) (18).
Activators of Recombinant Type I Adenylylcyclme-Since at least 99% of the adenylylcyclase activity in Sf9 cell membranes is contributed by the recombinant enzyme after infection of cells with B-rAC, these membranes offer a unique opportunity to study the properties of this form of adenylylcyclase in a membrane-bound and reasonably native environment. Most of the enzyme appears to be in the plasma membrane, based on indirect immunofluorescence using permeabilized cells and Ab CI-1115 (data not shown). As anticipated, enzymatic activity is enhanced by the addition of calmodulin, activated G,,, forskolin, or Mn2+ when compared with assays performed in the presence of Mg2+ alone (basal activity) (Fig. 3).
The basal activity of adenylylcyclase in Sf9 cell membranes containing type I adenylylcyclase decreased by 50-60% after washing with EGTA (a procedure designed to remove endogenous calmodulin (36)). Addition of Ca2+ and calmodulin together then increased adenylylcyclase activity about 8-fold to a value of 4 nmol-min".mg" (Fig. 3A). Half-maximal values were achieved at a calmodulin concentration of 20 nM. Similar maximal activities were observed when Ca2+ and calmodulin were added to membranes that had not been washed with EGTA (not shown). There was no effect of calmodulin if Ca2+ was replaced by EGTA during the assay. GTPyS.rG,, activated type I adenylylcyclase about 5-fold (Fig. 3B). The half-maximally effective concentration for activated rG,, was about 8 nM; again, the maximal activity achieved was nearly 4 nmol . min" . mg".
The diterpene forskolin interacts with adenylylcyclase directly and activates the enzyme. When forskolin was added to EGTA-washed Sf9 cell membranes containing type I ad-CI-251. enylylcyclase a maximal specific activity of approximately 4-6 nmol.min-l -mg" was observed (although it is notoriously difficult to achieve a stable maximum in the presence of high concentrations of forskolin). Thus, calmodulin, G,,, and forskolin activate type I adenylylcyclase to a similar degree when tested alone.

-
A. Although adenylylcyclase activity assaved in the presence of Mn'+ also appeared to approximate 4 nmol.min". mg-l ( Fig. lA), this effect was reduced substantially when membranes were washed with EGTA (Fig. 311). Addition of calmodulin restored the effects of Mn2+. Thus, much of the effect of Mn'+ appears to be due to a stimulatory effect of Mn2+calmodulin on enzymatic activity.

6-6-
The reduced effect of forskolin in EGTA-washed membranes (compare Figs. 1A and 3 C ) suggests that there might be synergistic effects between activators of t-ype I adenvlvlcyclase; such interaction between forskolin and G.,. and between G,,, and calmodulin have been described previouslv (37, 38). Greater than additive activation of type I adenvlvlcyclase is seen when either GTPyS.G.,, or forskolin is tested in the presence of calmodulin (Fig. 4). Interestingly, there is little effect of activated Ga,, in the presence of maximally effective concentrations of forskolin, despite the fact that this interaction is prominent with at least certain other types of adenylylcyclase (37, 39) (and has been demonstrated with the type I1 and type IV proteins in Sf9 cell membranes).'," Forskolin does not activate adenylylcyclase in the presence of maximally effective concentrations of calmodulin plus GTPyS. G., (not shown).

Inhibitors of Recombinant Type I Adenyl.vlc?/cla.~e-Rovine
brain adenylylcyclase is inhibited directly by adenosine and certain of its analogs (see Ref. 40 for review). These compounds have been termed P-site inhibitors, since an intact purine ring is an important structural requirement: 2'-or 5 'deoxy and 3"phosphoryl compounds are the most potent. Inhibition is not competitive with respect to metal-ATP, and stimulated forms of adenylylcyclase are more sensitive to inhibition than is basal activity. P-site inhibitors were tested for their effects on type I adenylylcyclase activity in Sf9 cell membranes (Fig. 5). As expected, forskolin-stimulated adenylylcyclase activity was more sensitive to inhibition than was basal activity (not shown). The order of potency was 2'deoxy-3"AMP > 3'-AMP > 2"deoxyadenosine > adenosine, as described previously for the bovine brain enzyme (41). Changes of both slopes and intercepts of Lineweaver-Rurk plots were a linear function of inhibitor concentration, consistent with prior observations that inhibition of enzvmatic activity was noncompetitive with respect to M$+ .ATP.
The G protein @y subunit. complex can inhibit adenvlvlcvclase activity; multiple mechanisms have been proposed (1). Inhibition that is apparent,ly not caused bv association of Py with GB,, has been ascribed to interaction between 87 and calmodulin (42). Such inhibition was noted to be most prominent with the calmodulin-stimulated form of the enzvme and required high concentrations of &; concentrations approximately equal to those of calmodulin. Multiple forms of adenylylcyclase may have been present in the partially purified preparation that was used.
Inhibition of recombinant type I adenylylcyclase by By was prominent when calmodulin-stimulated activity was tested (Fig. 6). Activity stimulated by GTP-yS G,, or forskolin was less subject to this effect. However, the concentration of P-y required to inhibit adenylylcyclase activity half-maximally was about 5 nM, and this value was independent of the activator (Fig. 6A) and of the concentration of calmodulin (Fig. 6B). Thus, inhibition by Py is neither competitive with calmodulin, nor can it be explained by interaction between By and calmodulin, since concentrations of calmodulin were, at times, in great excess of those of By. Inactivated P-y did not inhibit adenylylcyclase activity.
Purification of Recombinant Type I Adenylylcyclase-Procedures for the purification of recombinant type I adenylylcyclase were based on those described previously for the bovine brain enzyme (18), with minor modifications. The efficiency of solubilization of recombinant adenylylcyclase with Lubrol PX is about 50-60% (Table I), which is significantly less than that for the bovine brain enzyme. About 80% of this activity bound to forskolin-Sepharose after 4-6 h at 4 "C. Extensive washing of the column was critical for removal of contaminating proteins. Recovery of adenylylcyclase activity was only about 10% of the total. The highest specific activity of the protein recovered in the eluates was 4.3 pmol . min" . mg". This is slightly lower than (but comparable to) that observed with the purified brain protein (8 pmol. min" . mg") (5). As expected, the purified protein migrated as a species with a molecular weight of approximately 110,000, as judged both by silver staining and immunoblotting (Fig. 7).
Purified recombinant type I adenylylcyclase is not activated by GTPyS, indicating that the preparation is free of Gs,. The enzyme can be activated by G,, . GTPyS, Ca2+ + calmodulin, and forskolin (Table 11). Although the solubilized enzyme was inhibited by /3y prior to purification, we were unable to observe an effect of B-y on the purified preparation of adenylylcyclase. We are currently attempting to elucidate the mechanism of this effect of P-y.
Characteristics of Truncated Forms of Adenylylcyclase-As mentioned above, the four types of mammalian adenylylcyclase that have been cloned to date appear to be similar structurally (see Fig. 8, top). For convenience of discussion, we have divided the structure into seven regions. The aminoterminal domain is designated N, while the two membranespanning domains are designated M1 and M2. These regions differ markedly from one type of adenylylcyclase to another. The cytoplasmic domains are designated Cla, Clb, c2a, and C2b. C , and CPa are about 55% similar to each other, and they are also similar to the corresponding domains of other mammalian adenylylcyclases and to the catalytic domains of membrane-bound guanylyl cyclases (43, 44). The Clb regions are similar between type I1 and type IV adenylylcyclases; however, the Clb regions of types I and I11 differ from each other and from types 11 and Iv. Cab is largely absent in types 11 and I v and differs in the type I and I11 enzymes.
Three types of truncated adenylylcyclase constructs have been made. In the first, ACAN, the first 52 amino acid residues of the N region have been removed. Membranes from Sf9 cells that were infected with baculovirus containing this cDNA (B-rACAN) displayed an elevated level of adenylylcyclase activity that was stimulated by G,,, calmodulin, and forskolin (Fig. 8). However, the specific activity of adenylylcyclase was only 10% of that observed with B-rAC-infected cells. When the level of synthesis of this mutant protein was evaluated by either biosynthetic labeling (Fig. 9A) or immunoblotting (not shown), expression was found to be about 20fold greater than that observed with wild-type adenylylcyclase. Thus, it is possible that the ACAN mutant has an extremely low specific activity; alternatively, improper folding and/or orientation of the mutant protein in the membrane may account for at least some of its low activity.
The two other types of truncations that have been explored to date represent expression, roughly, of the amino-terminal and carboxyl-terminal halves of the protein. NMICI-l (long) and NMIC1-s (short) consist of the N, M1, and C1. regions and part of Clb; the long and short versions of this construct terminate 10 and 43 amino acid residues prior to the beginning of MP, respectively. The MzCz construct begins in region Clb, 46 amino acid residues prior to M2, and goes to the carboxyl terminus. Expression of these truncated proteins in Sf9 cells resulted in little or no change in adenylylcyclase activity. However, coinfection of cells with recombinant baculovirus encoding NMIC, and M2C2 resulted in the appearance of a substantial level of adenylylcyclase activity that is stimulatable by activators (particularly GTPyS . G,, and forskolin) and is inhibitable by 2'-deoxy-3'-AMP and G protein subunits (data not shown). Similar results were obtained if the C2b region of MPC2 was removed (data not shown): Biosynthetic labeling studies indicate that the level of expression of NMICl and M2C2 proteins is high (roughly 10-fold greater than that for wild-type adenylylcyclase when NMICl and M2C2 are expressed simultaneously, Fig. 9A).
The major products of ACAN and M2Cz migrate as doublets during SDS-PAGE (99 and 104 kDa for ACAN and 52 and 55 kDa for MzCz (Fig. 9, A and B ) ) . The apparent molecular weights of NMICI-l and NMIC1-s are 62,000 and 55,000, respectively (Fig. 9A). All of these species migrate as if their molecular weights are 5,000-10,000 units smaller than predicted; the reason for this deviation is unknown. Products that are presumed to result from proteolysis of these truncated forms of adenylylcyclase are also evident in Fig. 9A and by immunoblotting (not shown).
Sf9 cells expressing the truncated forms of adenylylcyclase were treated with tunicamycin to determine if the doublet seen with M2C2 was a result of differences in N-linked glycosylation. Indeed, only the lower band of the M2C2 doublet was apparent when protein was synthesized in the presence of tunicamycin (Fig. 9B). The sizes of the NMIC1-l and NMIC1s products were not changed by tunicamycin. There are four consensus sites for N-linked glycosylation in the sequence of type I adenylylcyclase. Our model of the orientation of the protein in the bilayer suggests that only one of these (in the Mz region) would be utilized. Results obtained with tunicamycin are consistent with this model.

DISCUSSION
Adenylylcyclase is a complex enzyme. It exists normally at very low concentrations (0.01-0.001% of membrane protein) and is labile and difficult to manipulate in detergent-containing solutions. The existence of multiple forms of the protein confounds interpretation of experiments performed with native membranes, particularly with preparations from sources as complex as brain. These considerations speak to the desirability of expression of individual forms of adenylylcyclase, which can then be studied in situ or after purification to homogeneity. However, expression of complex, intrinsic membrane proteins in quantities sufficient for biochemical characterization has often been difficult. The use of recombinant baculovirus permits expression of quantities of adenylylcyclase that are adequate for several needs. The activity of the recombinant protein exceeds that of the endogenous enzyme by loo-fold, permitting an uncluttered view of the adenylylcyclase of interest. Large numbers of Sf9 cells can be grown, and it may be possible to scale the system for production of milligram quantities of protein. However, it would certainly be desirable to increase the level of expression of adenylylcyclase that has been obtained to date.
Toward this end we note that there has been a negative T. Nguyen, W.-J. Tang, and A. G. Gilman, unpublished observations. correlation between the amount of adenylylcyclase protein synthesized during our attempts at expression and the intrinsic activity of the specific protein under consideration. Thus, ACAN, which retains most of the structural features of the native protein, has poor activity but accumulates to levels of more than 1% of membrane protein. NM,Cl and M2C2 are also inactive and accumulate to high concentrations; however, when they are expressed together (with resultant activity), the level of expression decreases substantially (see Fig. 9). Perhaps these considerations also explain our failure to achieve a more efficient transient expression system by modification of the 5' end of the adenylylcyclase cDNA in the pCMV constructs (pCMVAC versus pCMVACA58). Excessive production of cyclic AMP may inhibit transcription and/or translation. Inclusion of cyclic AMP analogs in the Sf9 cell culture medium did inhibit expression of adenylylcyclase, although millimolar concentrations of chlorophenylthio cyclic AMP were required to observe this effect. We attempted to control concentrations of cyclic AMP in Sf9 cells by coinfection of these cells with separate viruses encoding both adenylylcyclase and a low K , cyclic nucleotide phosphodiesterase (cDNA kindly provided by Dr. Michael Wigler, Cold Spring Harbor Laboratories, Cold Spring Harbor, NY). Although cyclic nucleotide phosphodiesterase activity increased markedly (to 100 nmol. min" . mg"), adenylylcyclase activity increased by only 2-fold at best. Type I adenylylcyclase expressed in Sf9 cells displays all of the properties known to be characteristic of calmodulin-sensitive adenylylcyclase purified from brain, and certain of these properties are clearly distinct from those of other adenylylcyclases. We note in particular that the enzyme can be stimulated equally by calmodulin and G,,, that calmodulin acts synergistically with other activators (Ga, or forskolin), and that there is no synergistic interaction between G,, and forskolin. Most interesting, perhaps, is the effect of G protein Py subunits. Inhibition of calmodulin-stimulated enzymatic activity was most notable, as previously described (42). However, we cannot ascribe this effect to interaction between & and calmodulin, as mentioned above.
Discussion of interactions of G protein & subunits with effectors such as adenylylcyclase, ion channels, and phospholipases has been controversial (45-47). The inhibitory effect of @y on a purified preparation of calmodulin-sensitive, bovine brain adenylylcyclase was noted previously (5). However, the effect was lost after further manipulation of the protein with alterations of detergent, and it was tentatively ascribed to contamination of adenylylcyclase with Gs,. We, too, have failed to detect inhibitory effects of by on purified preparations of type I adenylylcyclase. Experiments in progress are designed to determine if an additional protein is necessary to observe this effect. The concentration of Py required to inhibit adenylylcyclase is high relative to the concentrations of G protein a subunits required to activate the enzyme or to affect channel function.
(Although By and G,, were effective in the same concentration range in the experiments described above, recombinant (Escherichia coli-derived) G,, was utilized. This protein has an affinity for adenylylcyclase that is approximately 20-fold lower than that of native G,, (34).) Nevertheless, Py is a relatively abundant protein complex, particularly in brain, and its total concentration exceeds that of G,, by a substantial margin (>lOO-fold in brain). For G, to serve as the physiological activator of adenylylcyclase, the stimulatory effect of G,, must, of course, dominate the inhibitory effect of By. It will be interesting to determine if individual forms of differ in their ability to inhibit adenylylcyclase and if individual forms of adenylylcyclase are equally susceptible to this effect. Preferential effects of /3y on the calmodulin-stimulated activity of the type I enzyme suggest that different adenylylcyclases may respond to /3y in unique ways.
The quasisymmetrical structure of adenylylcyclase with its "duplicated C1, and CZe domains is curious. Although these regions do not have sequences that are characteristic of classical nucleotide-binding domains, we assume that at least one of them must have this function. There is a weak homology between C1, and Cza and the catalytic domain of yeast adenylylcyclase, and there is strong similarity between these regions and the catalytic domains of several membrane-bound guanylyl cyclases. Thus, we separated these regions by expression of NMICl and M2C2 in an attempt to locate the catalytic domain of adenylylcyclase. The results suggest that interactions between these regions may be necessary for catalysis, but they clearly do not rule out phenomena such as interactions necessary for proper folding of the molecules. Nevertheless, other observations are consistent with the possibility that interactions between C1, and Cza play a role in catalysis. Notably, kinetic analysis suggests that the inhibitory P-site and the catalytic site are separate but interacting domains (48); it may be possible to assign C1, and CZa to these respective functions. The adenylylcyclases of Bacillus anthrasis and Bordetella pertussis have duplicated A-type nucleotide-binding consensus sequences, and mutational analysis indicates the importance of these sites for catalysis (49-51). Soluble guanylyl cyclase is a heterodimer. Each of its subunits has a region that is homologous to C1, and CZa, and the subunits must be expressed concurrently in order to observe catalysis (19). Perhaps the membrane-bound guanylyl cyclases, with but one such domain, function as homodimers.