Calmodulin-binding Sites on Adenylyl Cyclase Type VIII*

Ca2+ stimulation of adenylyl cyclase type VIII (ACVIII) occurs through loosely bound calmodulin. However, where calmodulin binds in ACVIII and how the binding activates this cyclase have not yet been investigated. We have located two putative calmodulin-binding sites in ACVIII. One site is located at the N terminus as revealed by overlay assays; the other is located at the C terminus, as indicated by mutagenesis studies. Both of these calmodulin-binding sites were confirmed by synthetic peptide studies. The N-terminal site has the typical motif of a Ca2+-dependent calmodulin-binding domain, which is defined by a characteristic pattern of hydrophobic amino acids, basic and aromatic amino acids, and a tendency to form amphipathic α-helix structures. Functional, mutagenesis studies suggest that this binding makes a minor contribution to the Ca2+ stimulation of ACVIII activity, although it might be involved in calmodulin trapping by ACVIII. The primary structure of the C-terminal site resembles another calmodulin-binding motif, the so-called IQ motif, which is commonly Ca2+-independent. Mutagenesis and functional assays indicate that this latter site is a calcium-dependent calmodulin-binding site, which is largely responsible for the Ca2+ stimulation of ACVIII. Removal of this latter calmodulin-binding region from ACVIII results in a hyperactivated enzyme state and a loss of Ca2+ sensitivity. Thus, Ca2+/calmodulin regulation of ACVIII may be through a disinhibitory mechanism, as is the case for a number of other targets of Ca2+/calmodulin.

Mammalian adenylyl cyclases are a diverse group of variously regulated signaling molecules. Details are emerging on some of the molecular features conferring catalytic and regulatory properties on these enzymes. All of the nine cloned adenylyl cyclases are large (1080 -1248 amino acids) polypeptides that are proposed to comprise two cassettes of six transmembrane-spanning domains, each cassette being followed by a large cytoplasmic domain (1,2). The transmembrane domains are not highly conserved among adenylyl cyclases. However, parts of two of the cytoplasmic domains (termed C1a and C2a) are highly conserved, and, when expressed separately, they can combine to display basic catalytic activity (3)(4)(5)(6)(7)(8). These molecules have been crystallized, and the combination of C1a and C2a, each consisting of a three layer ␣/␤ sandwich, forms an active catalytic core to generate cAMP from ATP (9,10). The other major cytoplasmic domains of mammalian adenylyl cy-clases, the N terminus, the C1b region, and the C2b region, are not conserved at all and are speculated to reflect regulatory features of specific adenylyl cyclases (1,(11)(12)(13).
Ca 2ϩ elicits a prominent stimulation of ACI 1 and ACVIII, which is mediated by loosely bound calmodulin (3,11,14). Although the likely calmodulin-binding domain on ACI has been localized to the C1b region (15,16), the corresponding regulatory domain has not been identified on ACVIII. Indeed, ACVIII does not possess analogous calmodulin-binding sites in the C1b region. In the case of ACI, peptides corresponding to putative calmodulin-binding domains were used to identify a site in the C1b region as the likely site of calmodulin binding (16). Mutagenesis studies strongly supported this assignment (15,17). However, no information is yet available on the possible domains that mediate the Ca 2ϩ /calmodulin responsiveness of ACVIII. Given that ACI and ACVIII do not share similar C1b domains (they are only ϳ40% similar at the amino acid level, compared with 80% similarity in the C1a region) and also that their regulation by Ca 2ϩ /calmodulin shows distinct properties, it might not be unexpected if different motifs and/or locations were involved.
Identifying calmodulin-binding sites on proteins still mainly depends on experimentation, although some predictive criteria are available to guide experiments. For instance, many known Ca 2ϩ -dependent calmodulin-binding proteins possess a region that is often characterized by an amphipathic helix consisting of approximately 20 amino acid residues (18). In these regions, basic amino acids are interspersed among hydrophobic residues, and aromatic amino acids normally appear near either end (16,18,19). However, sequence analysis based on these criteria does not always identify calmodulin-binding regions, and indeed, regions of proteins that bind to calmodulin sometimes do not fit these criteria. Another calmodulin-binding motif is the so-called "IQ motif," consensus sequence IQXXXRGXXXR, which often (18) but not always (22,23) binds calmodulin in a Ca 2ϩ -independent manner.
The present studies used a combination of calmodulin overlay assays, mutagenesis, and peptide inhibition studies to locate the calmodulin regulatory domains on ACVIII. Surprisingly, a primary site was located in the C2b region, while an ancillary site that appeared to play a minor autoinhibitory role was located in the N terminus. Production of His-tagged Protein Fragments from ACVIII-Eight His-tagged fusion proteins were generated for calmodulin overlay experiments and are referred to as follows, with the appropriate ACVIII residues in parentheses: Nt (Met 1 -Glu 179 ); Nn (Met 1 -Ser 110 ); Nc (Glu 108 -Glu 179 ); C1 (Ala 346 -Asn 712 ); C1a (Ala 346 -Ser 593 ); C2 (Gly 913 -Pro 1248 ); C2a (Gly 913 -Pro 1184 ); C2b (Leu 1137 -Pro 1248 ). Nt, Nn, C2, C2a, and C2b were constructed by amplifying nucleotides 777-1314, 777-1136, 3513-4520, 3513-4329, and 4185-4520, respectively, by polymerase chain reaction (PCR) and subcloning the PCR products between EcoRI (5Ј-end) and HindIII (3Ј-end, blunted) sites of pRSETb (Invitrogen, Carlsbad, CA). C1 and C1a were constructed by amplifying nucleotides 1811-2912 and 1811-2546 by PCR and subcloning the PCR products between NcoI (5Ј-end) and HindIII (3Ј-end, blunted) sites of pRSETb. Nc was created from the Nt construct by cutting with EcoRI and BspI, blunting the two ends and ligating them back. Constructs were confirmed by sequencing. These cDNA constructs were transformed into BL21 (DE3) pLys S cells. The cells were grown in Luria's broth containing 100 g/ml ampicillin and 34 g/ml chloramphenicol at 37°C until they reached an A 600 of 0.8. Expression of these proteins was induced by the addition of 0.5 mM isopropyl ␤-D-thiogalactoside and incubation of the cells overnight at 37°C. The cells were harvested by centrifugation at 6000 ϫ g for 10 min and resuspended in ice-cold extraction buffer (20 mM Tris-HCl, pH 8.0, 50 mM NaCl, 1% Triton X-100, 1 mM ␤-mercaptoethanol, and protease inhibitor mixture). The cell lysate was sonicated (4 ϫ 10 s) and separated by centrifugation at 10,000 ϫ g for 10 min. The pellets were resuspended in equal volumes of 2ϫ sample buffer, boiled, and spun at 13,000 ϫ g for 5 min to eliminate insoluble material. The sample buffer containing bacterial proteins was loaded onto SDS-polyacrylamide gel electrophoresis gel. The expressed fusion proteins could be visualized readily from the gel stained by Coomassie Blue. These protein bands were also confirmed by Western blot probed with RGS-His antibody (Qiagen, Valencia, CA).
Calmodulin Overlay Assay-The calmodulin overlay assay was performed by fractionating His-tagged fusion proteins by SDS-polyacrylamide gel electrophoresis, transferring to nitrocellulose membranes, and probing with biotinylated calmodulin, as described by the manufacturer (Calbiochem). The blot membranes were stained with India ink.
Measurement of cAMP Accumulation-HEK 293 cells were maintained and transfected as described previously (25). cAMP accumulation in intact cells expressing different ACVIII mutant constructs was measured according to the method of Evans et al. (1984) as described previously (26,27) with some modifications. HEK 293 cells on 12-well plates were incubated in minimal essential medium (60 min, 37°C) with [2-3 H]adenine (1.5 Ci/well) to label the ATP pool. The cells were then washed twice and incubated with a nominally Ca 2ϩ -free Krebs buffer (900 l/well) containing 120 mM NaCl, 4.75 mM KCl, 1.44 mM MgSO 4 , 11 mM glucose, 25 mM HEPES, and 0.1% bovine serum albumin (fraction V) adjusted to pH 7.4 with 2 M Tris base. The use of Ca 2ϩ -free Krebs buffer in experiments denotes the addition of 0.1 mM EGTA to the nominally Ca 2ϩ -free Krebs buffer. All experiments were carried out at 37°C in the presence of the phosphodiesterase inhibitors, 3-isobutyl-1methylxanthine (500 M) and Ro 20 -1724 (100 M), which were preincubated with the cells for 10 min prior to a 1-min assay. Unless stated otherwise, cells were preincubated for 10 min with the Ca 2ϩ -ATPase inhibitor thapsigargin at a final concentration of 100 nM. This has the effect of passively emptying the Ca 2ϩ stores, establishing a low basal [Ca 2ϩ ] i , and priming the cells for capacitative Ca 2ϩ entry (28). Assays were terminated by the addition of 5% (w/v final concentration) trichloroacetic acid. Unlabeled cAMP (100 l, 10 mM), ATP (10 l, 65 mM), and [␣-32 P]ATP(ϳ7000 cpm) were added to monitor recovery of cAMP and ATP. After pelleting, the [ 3 H]ATP and [ 3 H]cAMP content of the supernatant were quantified according to the standard Dowex/alumina methodology (29). The accumulation of cAMP is expressed as the percentage of conversion of [ 3 H]ATP into [ 3 H]cAMP; means Ϯ S.D. of triplicate determinations are indicated. The basal activity of some constructs was high, with the result that significant amounts of cAMP had accumulated before the beginning of the 1-min incubation period. In such cases, the Krebs buffer bathing the cells was exchanged with buffer containing the same concentrations of 3-isobutyl-1-methylxanthine/Ro20-1724, EGTA, and thapsigargin, at the beginning of the 1-min assay, as indicated in the figure legends.
Western Blotting Experiments-Two antibodies were used in Western blot experiments. Ab VIII-A 1229 -1248 was raised against amino acids 1229 -1248 of ACVIII at the very end of the C terminus; the other, Ab VIII-A 666 -682, was raised against amino acids 666 -682 in the C1b region of ACVIII (33). The experimental procedure was described previously (33). 10-g membrane proteins were loaded on each lane of a 10% polyacrylamide gel. After electrophoresis, the contents of the polyacrylamide gel were transferred to polyvinylidene difluoride membranes (Micron Separations Inc.), which were incubated with TBST (20 mM Tris-HCl, pH 7.4, 137 mM NaCl, 0.1% Tween 20) containing 5% nonfat dry milk, for 1 h at room temperature. The primary antibody (1:20,000 dilution) was then added, and the membrane was incubated for another 1 h. After washing in TBST, the membrane was incubated in TBST containing 5% nonfat dry milk with goat anti-rabbit IgG horseradish peroxidase conjugate (1:1000 dilution; Bio-Rad) at room temperature for 1 h. The immune complex was detected by enhanced chemiluminescence (manufacturer's protocol; Amersham Pharmacia Biotech).
Adenylyl Cyclase Activity Measurements-Determination of adenylyl cyclase activity in vitro was performed as described previously (34) (34). The reaction mixture (final volume, 100 l) was incubated at 30°C for 20 min. Reactions were terminated with sodium lauryl sulfate (0.5%); [ 3 H]cAMP was added as a recovery marker, and the [ 32 P]cAMP formed was quantitated as described previously (29). Data points are presented as mean activities Ϯ S.D. of triplicate determinations.
Protein Secondary Structure Prediction and Synthetic Peptide Experiments-The protein sequence was analyzed with DNASIS 2.5 (Hitachi Software Engineering Co. Ltd., San Bruno, CA). The secondary structures of the N-terminal fragment and the C2b region were predicted using the Chou and Fasman program. Two peptides were synthesized (Genemed Synthesis Inc., South San Francisco, CA). One was 21 residues, called 8Ncam (RPQRLLWQTAVRHITEQRFIH), corresponding to amino acids 32-52 of ACVIII; the other one was 25 residues, called 8Ccam (YSLAAVVLGLVQSLNRQRQKQLLNE), corresponding to amino acids 1186 -1211 of ACVIII. The N termini of the two peptides were protected by acetylation, and the C termini were protected by amidation, to prevent self-circulation and to mimic the in vivo conformation. Both peptides were purified above 95%. The molecular weight for 8Ncam and 8Ccam is 2686 and 2885, respectively. The purity and molecular weight of the peptides were confirmed by high pressure liquid chromatography and mass spectroscopy, respectively (Genemed Synthesis). The peptide (CamkII, LKKFQARRKLKGAILTTMLA) of the CAM kinase II calmodulin-binding domain was purchased from Calbiochem. The peptide 8CT (TPSGPEPGAQAEGTDKSDLP) has 20 residues, corresponding to amino acids 1229 -1248 of ACVIII, which was originally synthesized for raising antibody Ab VIII-A 1229 -1248 (33). Stock solutions (2 mM) were made in water for all four peptides, which were used in in vitro adenylyl cyclase assays as described above, to generate competition and inhibition curves. The inhibition curves were fitted by the program Inplot 4.04 (GraphPad Software Inc.) with the following equation: Y ϭ A ϩ (B Ϫ A)/(1 ϩ (X/C)), where Y represents the adenylyl cyclase activity relative to the activity in the absence of peptides and X is the peptide concentration. The maximal value (A), the minimal value (B), and IC 50 (C) were obtained by curve fitting. The IC 50 for each peptide was calculated from three independent experiments.

RESULTS
Calmodulin Overlay Assays-Overlay assays were an initial strategy adopted to identify putative regions for calmodulin regulation of ACVIII. It seemed reasonable to assume that a sizable molecule (ϳ19 kDa) such as calmodulin, which is dissociable from ACVIII by washing in EGTA-containing buffers (14), should bind somewhere in the three large intracellular loops. Consequently, eight fusion proteins including the regions comprising these three loops were generated, designated Nt, Nn, Nc, C1, C1a, C2, C2a, and C2b ( Fig. 1A) as described under "Experimental Procedures." Western blotting with an anti-RGS-His antibody indicated that most of the expressed fusion proteins were in the pellets from the Escherichia coli cell lysates (data not shown). Therefore, cell lysate pellets solubilized with sample buffer were used in calmodulin overlay assays. Following transfer from SDS-polyacrylamide gel electrophoresis, nitrocellulose membranes with fractionated E. coli proteins were incubated with biotinylated calmodulin, and the bands that bound calmodulin were detected by horseradish peroxidase-streptavidin incubation and enhanced chemiluminescence (see "Experimental Procedures"). The N-terminal frag-ments, Nt and Nn, yielded strong signals for calmodulin binding (Fig. 1, C and E). This signal was Ca 2ϩ -dependent and was removable by washing the membrane with EGTA-containing buffer (data not shown). No significant signal was detected for any other domains (Fig. 1, C and E). Two other fusion proteins including the C1b region were also generated, neither of which showed positive results in calmodulin overlay assays (data not shown). Thus, these data indicate that the N-terminal exclusively binds calmodulin. However, since proteins on nitrocellulose membranes are denatured, it is quite possible that a site binding to calmodulin in the native state would not bind calmodulin in an overlay assay. Therefore, mutagenesis studies and functional experiments were considered critical to evaluate the significance of any apparent interactions derived from overlay assays.
Ca 2ϩ Sensitivity of ACVIII Mutants in Vivo-The C1a and C2a regions, which are the catalytic domains (4, 6, 35) are highly conserved across the mammalian adenylyl cyclase family. These domains have already been crystallized and well characterized (5,9,10,36,37). A putative calmodulin-binding site was not revealed in these highly conserved regions by either sequence analysis or earlier experimental attempts (15,16). On the other hand, the N terminus, C1b, and C2b regions are poorly conserved and are the regions where type-specific regulatory domains would be expected to reside. Therefore, a group of deletions was made, concentrating on these three areas (Table I) A, diagram of protein fragments from AC-VIII cytoplasmic domains. Five amino acids and their positions are marked in the one-dimensional structure diagram of ACVIII from left (N terminus) to right (C terminus). Vertical black boxes represent the predicted transmembrane domains; two thick lines represent the highly conserved C1a and the C2a region. The positions of eight cytoplasmic fragments (horizontal black boxes) are shown below, which were generated into His-tagged fusion proteins as described under "Experimental Procedures." B-E, calmodulin overlay assays were performed as described under "Experimental Procedures" in the presence of 0.5 mM CaCl 2 . The fusion proteins loaded on the gel are indicated at the top of the panels. The molecular weight is marked at the left. B and D, transferred membrane blots stained with India ink. C and E, calmodulin binding shown in the x-ray film from the membranes of B and D, respectively. The positions of the fusion proteins are indicated (arrowheads), which were confirmed by Western blot probed with anti-RGS His antibody. Nonspecific binding to horseradish peroxidase-streptavidin is also indicated (asterisk). When 0.5 mM EGTA was substituted for CaCl 2 in the overlay assay, the calmodulin-binding bands in C and E were no longer evident (not shown).
with 100 nM thapsigargin to deplete the intracellular Ca 2ϩ stores and prime the cells for capacitative Ca 2ϩ entry, which selectively regulates the activities of Ca 2ϩ -sensitive adenylyl cyclases (25,38,39). Thus, the Ca 2ϩ sensitivity of adenylyl cyclases can be determined by comparing the cAMP accumulation in cells pretreated with thapsigargin in the presence or absence of Ca 2ϩ during the assay. Surprisingly, the N terminus deletion of ACVIII, N ⌬1-106 (mutant 1 in Fig. 2A) remained sensitive to Ca 2ϩ , although this deleted region bound calmodulin in the overlay assay (Fig. 1). The -fold stimulation by Ca 2ϩ for N ⌬1-106 and two other N terminus deletions, N ⌬1-49 ( Fig.  2A, mutant 2) and N ⌬1-106, ⌬158 -169 (mutant 3) was only about half of that for wild type ACVIII; nevertheless, clear Ca 2ϩ regulation was retained. These data strongly suggested that the N-terminal region was not the only functionally significant calmodulin-binding region within ACVIII.
Deletions in the C1b region were next evaluated. Deleting the entire C1b region (Pro 587 -Ser 701 ) fully inactivated ACVIII (data not shown). The mutant C1 ⌬635-700 ( Fig. 2A, mutant 4), which is a naturally occurring splice variant of ACVIII (the "C form" (33), which lacks 66 amino acids in the C1b region), was stimulated by Ca 2ϩ . A deletion C1 ⌬588 -644 ( Fig. 2A, mutant 7), missing the remaining part of the C1b region, appeared to lose its sensitivity to Ca 2ϩ , but its activity was very low. Two related deletions, C1 ⌬588 -619 ( Fig. 2A, mutant 5) and C1 ⌬620 -644 (mutant 6) , were similar to the wild type ACVIII in terms of their -fold stimulation by Ca 2ϩ , although the forskolinand Ca 2ϩ -stimulated activity of C1 ⌬620 -644 was much higher than that of C1 ⌬588 -619 . These results suggest that the C1b region is important for ACVIII activity, and, consequently, deletions in the C1b region could modify the Ca 2ϩ regulation and the catalytic activity of ACVIII. However, it seems unlikely that this region is involved in the direct regulation by calmodulin. A double deletion, N ⌬1-106, ⌬588 -619 ( Fig. 2A, mutant 10), lacking approximately two-thirds of the N terminus and a major part of the C1b region, was also Ca 2ϩ -stimulable. This mutant further confirmed that the N terminus and the C1b region are unlikely to be the calmodulin-binding region responsible for stimulation by Ca 2ϩ .
In the assays described above, the basal activity of adenylyl cyclase is the cAMP that accumulates both inside and outside the cells before (10 min) and during the 1-min assay, in the presence of PDE inhibitors to prevent the breakdown of cAMP. Basal accumulation of cAMP is normally trivial during the preincubation period; however, where basal activities of some of these constructs are high (particularly C2b deletions), a considerable accumulation of cAMP can occur prior to the onset of the 1-min assay. Therefore, to clearly examine the activity of NC2 ⌬1-106, ⌬184 -1248 within the 1-min assay period, we modified our in vivo assay protocol. The medium was exchanged with new buffer just before the 1-min assay to minimize the influence of accumulated extracellular cAMP. In addition, the intracellular cAMP that had accumulated was measured by exchanging the medium and lysing one group of cells with trichloroacetic acid at the beginning of the 1-min assay. When HEK 293 cells transfected with pcDNA3, ACVIII, C2 ⌬1184 -1248 , or NC2 ⌬1-106, ⌬1184 -1248 were assayed using this modified protocol, the cAMP accumulation of C2 ⌬1184 -1248 and NC2 ⌬1-106, ⌬1184 -1248 in all of the conditions was much reduced. However, stimulation by Ca 2ϩ of wild type ACVIII was evident even in the basal state and was strikingly evident when activity was also stimulated by forskolin (Fig. 2B). A clear stimulation of NC2 ⌬1-106, ⌬1184 -1248 by forskolin was also revealed (Fig. 2B), while in the original in vivo assay no significant forskolin stimulation was detectable, probably due to masking by the high extracellular cAMP accumulation ( Fig. 2A). Most significantly, again, in keeping with the original in vivo assay results ( Fig. 2A), both C2 ⌬1184 -1248 and NC2 ⌬1-106, ⌬1184 -1248 were quite insensitive to Ca 2ϩ (Fig. 2B). Overall, these data strongly indicate that the C2b region of ACVIII is critical to the Ca 2ϩ stimulation. The activity of these deletions was also assessed in in vitro assays to confirm or disprove the conclusions from the intact cell assays.
Although the activity of this latter construct was lower than those of the other mutants, it was still significantly higher than the control (130 versus 20 pmol/mg/min; Fig. 3). Thus, the Ca 2ϩ sensitivity, from both the in vivo and in vitro assays, of all of the mutants agree well and confirm that the C2b region is critical for the Ca 2ϩ sensitivity of ACVIII. However, it is notable that the activities of C2 ⌬1184 -1248 and NC2 ⌬1-106, ⌬1184 -1248 are lower than that of the wild type in vitro (Fig. 3), while in the intact cells in the presence of forskolin, their activity is much higher than the wild type activity (Fig. 2B). This discrepancy could result from (among other things) misdirection of the adenylyl cyclase constructs to inappropriate membrane loca-tions. Therefore, to confirm the expression of all of the mutants and to estimate the amount of protein being expressed in the plasma membrane, Western blot experiments were performed on the membranes that were used in in vitro assays.
Characterization of the Three Deletions, N ⌬1-106 , C2 ⌬1184 -1248 , and NC2 ⌬1-106, ⌬1184 -1248 -From the foregoing data, three of the apparently more informative mutants, N ⌬1-106 , C2 ⌬1184 -1248 , and NC2 ⌬1-106, ⌬1184 -1248 , were chosen for more detailed analysis. Ca 2ϩ and calmodulin concentration-response curves were generated for these three mutants and compared with the wild type ACVIII. For wild type ACVIII, activity began to increase at 0.4 M Ca 2ϩ , reached a plateau between 1 and 10 M, and declined when the free Ca 2ϩ concentration exceeded 10 M (Fig. 5A). The decrease of ACVIII activity at high [Ca 2ϩ ] free is considered to reflect Ca 2ϩ competing with Mg 2ϩ at an allosteric regulatory site (1,40,41). This inhibition by high [Ca 2ϩ ] is a property of all mammalian adenylyl cyclases, apart from AC3, regardless of their response to Ca 2ϩ in the submicromolar range (25,41,42). The Ca 2ϩ concentration-response curve of N ⌬1-106 is similar to that of the wild type, except that the Ca 2ϩ -stimulated activity is somewhat higher (Fig. 5A). On the other hand, the C2 ⌬1184 -1248 construct was unaffected by low [Ca 2ϩ ] free , although its activity started to decline when [Ca 2ϩ ] free exceeded 10 M (Fig. 5A). Ca 2ϩ also did not stimu-late, but rather inhibited, the activity of NC2 ⌬1-106, ⌬1184 -1248 when concentrations surpassed 1 M (Fig. 5A).
Ca 2ϩ concentration-response experiments were also performed in the absence of calmodulin. The wild type ACVIII was stimulated by approximately 50% with supramicromolar Ca 2ϩ (Fig. 5B). This stimulation apparently resulted from residual calmodulin in the plasma membrane, although the preparation was washed twice with assay buffer containing 800 M EGTA. Such persistence of calmodulin with adenylyl cyclase has been long encountered (41). It had been considered to reflect the persistent association of calmodulin with plasma membranes or the presence of calmodulin in other assay components, e.g. albumin or creatine phosphokinase (43). However, it is conceivable that there is a tight binding site for calmodulin on adenylyl cyclases. The latter possibility is supported by two observations: (i) the Ca 2ϩ stimulation of wild type ACVIII observed without the addition of exogenous calmodulin can be abolished by the peptide 8Ncam, which binds to calmodulin (data not shown), and (ii) without exogenous calmodulin, Ca 2ϩ has little effect on N ⌬1-106 (Fig. 5B). Thus, the deletion of amino acids 1-106 appears to result in no residual calmodulin. As expected, no stimulation was seen with either C2 ⌬1184 -1248 or NC2 ⌬1-106, ⌬1184 -1248 , and activities started to decline at 1 M free Ca 2ϩ concentration (Fig. 5B).
Calmodulin concentration-response experiments were performed on the mutants in the presence of 22.4 M free Ca 2ϩ . (A high [Ca 2ϩ ] was chosen to saturate high concentra-

FIG. 3. In vitro measurement of the Ca 2؉ sensitivity of various mutants.
Adenylyl cyclase activity was determined in plasma membranes (5 g/reaction) prepared from cells transfected with either vector alone (pcDNA3.0, as control), wild type ACVIII (AC8), or the mutants indicated as described under "Experimental Procedures." Activity was measured in the presence of calmodulin (1 M) and forskolin (20 M), with (white bars) or without (black bars) 2 M free Ca 2ϩ . Mutant numbers from Table I are used. Values shown are from an experiment that was repeated at least three times with similar results.

FIG. 4. Western blots of various constructs.
The same membrane proteins assayed in Fig. 3 were loaded on SDSpolyacrylamide gel (10 g/lane) and probed as described under "Experimental Procedures." Mutant numbers (Table I) are indicated at the top of the corresponding lanes. The molecular masses are marked in kDa on the left. A, Ab VIII 1229 -1248 antibody was used to probe the mutants with intact C termini. B, Ab VIII-A 666 -682 was used to probe the mutants with intact C1b regions but with a deleted C terminus. This result is representative of two experiments yielding similar results. tions of calmodulin.) N ⌬1-106 elicited a similar stimulatory profile as wild type ACVIII, although the -fold stimulation (approximately 6-fold) was higher than that of the wild type (approximately 4-fold; Fig. 5C). Deletions C2 ⌬1184 -1248 and NC2 ⌬1-106, ⌬1184 -1248 were relatively insensitive to calmodulin when the concentration was lower than 1 M (Fig. 5C). However, when the calmodulin concentration was increased to 10 M, the activities of these two mutants were about 2-fold higher than the activity in the absence of calmodulin (Fig. 5C). This increase in the activities of C2 ⌬1184 -1248 and NC2 ⌬1-106, ⌬1184 -1248 , in the presence of a high concentration of calmodulin, most likely reflected a reversal of the inhibitory effects of high (22.4 M) Ca 2ϩ on these mutants (see Fig. 5B) due to calmodulin chelation of Ca 2ϩ .
Peptide Competition Studies-The calmodulin overlay assay data and the mutagenesis screen identified two putative calmodulin-binding sites on ACVIII; one is in the N terminus, amino acids Met 1 to Ser 110 , and one is in the C terminus, amino acids Pro 1184 to Pro 1248 . Using the Chou and Fasman secondary structure prediction program, we located only one putative site in the N terminus that could form a helical structure -from Gln 34 to His 52 . This site resembles a classic Ca 2ϩ -dependent calmodulin-binding site ( Fig. 6A; see Ref. 18), in that it has basic amino acids (net charge, ϩ4) and two aromatic amino acids, Trp 38 and Phe 50 . The hydrophobic amino acid distribution pattern is also similar to some known Ca 2ϩ -dependent calmodulin-binding sites, such as the death-associated protein kinase (a serine/threonine kinase associated with mediation of interferon-induced cell death (19)), the human erythrocyte Ca 2ϩ pump (20), a Ca 2ϩ -regulated nitric-oxide synthase (16), rabbit skeletal muscle myosin light chain kinase (21), the inositol 1,4,5-trisphosphate type I receptor, and Bacillus anthracis adenylyl cyclase ( Fig. 6A; Ref. 18). At the C terminus of ACVIII, between amino acids Pro 1184 and Pro 1248 , the best candidate for a calmodulin-binding site would be from amino acid Tyr 1187 to Glu 1211 . This site is similar to the IQ motif (18), although it lacks one glycine residue (Fig. 6B). It also tends to form a helical structure as predicted by the Chou and Fasman analysis and shares homology with other calmodulin-binding sites, such as the ones in inositol 1,4,5-trisphosphate receptor (18), the ⑀-subunit of protein kinase C (18), and the ␤-subunit of cyclic nucleotide-gated channel (Ref. 23; Fig. 6B). This site would be an atypical calmodulin-binding site if it indeed bound calmodulin, since the IQ motif normally binds calmodulin independently of Ca 2ϩ , although some studies have shown that the IQ motif can also bind to calmodulin in a Ca 2ϩ -dependent manner (22,23). Two peptides from ACVIII were synthesized for competition studies, one corresponding to amino acids Gln 34 to His 52 , termed 8Ncam, the other corresponding to amino acids Tyr 1187 -Glu 1211 , termed 8Ccam. The N termini of the two peptides were protected by acetylation, and the C termini were protected by amidation to prevent self-circulation of the peptide and to mimic the in vivo conformation.
These peptides were used in in vitro assays in competition studies. A calmodulin concentration-response curve was generated for ACVIII, in the presence or absence of added peptides (at 4 M). CamkII, the peptide corresponding to the calmodulinbinding site of CAM kinase II, was used as a positive control. 8CT, the peptide corresponding to the C terminus of ACVIII, from Thr 1229 to Pro 1248 , was used as a negative control. In the presence of 1 M calmodulin and 7.74 M free Ca 2ϩ , 8CT had no effect on the calmodulin concentration-response curve, while CamkII suppressed the ACVIII activity and shifted the curve to the right (Fig. 6C). Both the 8Ncam and 8Ccam peptides also suppressed the calmodulin stimulated activity of ACVIII, although not quite as efficaciously as the CamkII peptide (Fig. 6C). Peptide inhibition experiments were also performed for these four peptides in the presence of a fixed calmodulin concentration (0.3 M; and 7.74 M free Ca 2ϩ ). Again, 8CT had no effect on ACVIII activity even at its highest concentration (Fig. 6D). CamkII began to inhibit ACVIII activity when its concentration exceeded 0.1 M (Fig. 6D; IC 50, 0.15 Ϯ 0.1 M; n ϭ 3). 8Ncam began to inhibit ACVIII activity at 0.3 M, (IC 50, 1.15 Ϯ 0.25 M; n ϭ 3; Fig. 6D). 8Ccam was a little more potent than 8Ncam (IC 50, 0.65 Ϯ 0.17 M; n ϭ 3; Fig. 6D). These experiments clearly demonstrate that both peptide sequences derived from the N and C termini of ACVIII are bona fide calmodulinbinding sequences and that the most potent peptide is also the one that is of most functional significance. DISCUSSION This study has explored the calmodulin-binding sites on AC-VIII. The related Ca 2ϩ -stimulable ACI binds calmodulin in the C1b region of the molecule. However, ACVIII and ACI are very dissimilar (only 40% homologous) in this region. Therefore, it might not have been unexpected that different sites would mediate the Ca 2ϩ stimulation of ACVIII. Calmodulin overlay assays revealed one putative Ca 2ϩ -dependent calmodulin-binding site in the N terminus of ACVIII (Fig. 1). However, without this region the enzyme was still sensitive to Ca 2ϩ (Figs. 2 and  3). On the other hand, using mutagenesis and functional assays, only those mutants lacking the C2b region, such as FIG. 6. Two calmodulin-binding sites located on ACVIII by structure prediction and synthetic peptide studies. A, prediction of a putative calmodulin-binding site in the N-terminal fragment of ACVIII, which had bound calmodulin in overlay assays (Fig. 1). Amino acid numbers are placed above the sequence. The predicted secondary structure was obtained using the Chou and Fasman program. The putative calmodulin-binding site, called 8Ncam, is indicated in boldface type. The sequence of 8Ncam is aligned with those of calmodulin-binding sites from death-associated protein (DAP) kinase (a serine/threonine kinase associated with mediation of interferon-induced cell death), the human erythrocyte Ca 2ϩ pump, Ca 2ϩ -regulated nitric-oxide synthase (NO synthase), rabbit skeletal muscle myosin light chain kinase (MLCK), and B. anthracis adenylyl cyclase. Their hydrophobic amino acid patterns and net charges are compared. B, prediction of a putative calmodulin-binding site in the C terminus of ACVIII whose binding to calmodulin was suggested by mutagenesis studies. The secondary structure of amino acids 1182-1248 of ACVIII was predicted using the Chou and Fasman program. The putative calmodulin-binding site is indicated with boldface type. The sequences of IQ motif, 8Ccam, IP3 receptor, protein kinase C (⑀-subunit), and CNG channel (␤-subunit) are aligned, and important amino acids are compared. C2 ⌬1184 -1248 and NC2 ⌬1-106, ⌬1184 -1248 , could not be stimulated by Ca 2ϩ either in vivo or in vitro (Figs. 2 and 3). This suggests that the C-terminal region is responsible for the Ca 2ϩ /calmodulin stimulation of ACVIII. Moreover, the high basal activities of C2 ⌬1184 -1248 and NC2 ⌬1-106, ⌬1184 -1248 suggests the removal of autoinhibitory domains (the C2b region playing the major role), which suppress the activity of wild type ACVIII. The binding of Ca 2ϩ /calmodulin to the autoinhibitory domain apparently relieves the inhibitory binding and activates the enzyme. Such a disinhibitory mechanism is employed in a number of Ca 2ϩ /calmodulin-activated enzymes, such as Ca 2ϩregulated nitric-oxide synthase (44), Ca 2ϩ /calmodulindependent protein kinases (45)(46)(47), a Na ϩ /H ϩ exchanger (48), and calcineurin (49). It appears as though most of the C2b region might participate as the inhibitory binding domain, since the two synthetic peptides 8Ccam (25 residues) and 8CT (20 residues) from the C2b region did not inhibit ACVIII activity in the absence of Ca 2ϩ and calmodulin (Fig. 6C). This disinhibition mechanism of Ca 2ϩ /calmodulin stimulation is different from that of forskolin, which is thought to stabilize the C1/C2 heterodimer to activate the adenylyl cyclase activity (9,10). The latter suggestion is supported by the fact that forskolin and Ca 2ϩ /calmodulin synergistically stimulate ACVIII.
The putative calmodulin-binding site in the C terminus has the signature sequence of an IQ motif, which generally reflects Ca 2ϩ -independent calmodulin binding. It is possible that the binding of calmodulin is Ca 2ϩ -independent and that Ca 2ϩ binding changes the conformation of bound calmodulin to activate the enzyme. Alternatively, the binding of calmodulin may be Ca 2 -dependent as is the case with the ␤-subunit of rod photoreceptor cyclic nucleotide-gated channel (23). It is of some interest that the peptide synthesized from this region (8Ccam) is only 5 times less effective than 8CamkII, which is a conventional calmodulin-binding peptide. This observation underscores how much we still need to learn about the molecular characteristics of calmodulin-binding sequences.
The putative calmodulin-binding site for the N terminus of ACVIII is a conventional Ca 2ϩ -dependent calmodulin-binding site, which is reinforced by the results of overlay assays. The fact that the double deletion NC2 ⌬1-106, ⌬1184 -1248 has higher activity in vivo (Fig. 2) and can be inhibited by a lower concentration of Ca 2ϩ (Fig. 5A) than C2 ⌬1184 -1248 suggests that this site contributes to the Ca 2ϩ stimulation of ACVIII, although this contribution must be minor.
The fact that there is apparently more residual calmodulin in ACVIII wild type membrane preparations than in those of N ⌬1-106 (Fig. 5B) might suggest a role of the N terminus of ACVIII as a Ca 2ϩ -independent calmodulin trap, notwithstanding the apparently conflicting evidence of the Ca 2ϩ -dependent manner of the N-terminal calmodulin-binding site from overlay assays.
Unlike the two regulatory domains (the N terminus and the C2b region) discussed above, the C1b region does not have a free end, which suggests that the disruptions on this region could more easily change the activity of adenylyl cyclases. However, continuous deletions in the C1b region of ACVIII could not eliminate the Ca 2ϩ stimulation of ACVIII ( Fig. 2A), while, by contrast, a point mutation in this region of ACI abolished its Ca 2ϩ sensitivity (17). The different calmodulin-binding sites on ACVIII and ACI are underlined by some differences in their regulation by Ca 2ϩ /calmodulin; for instance, ACI is more sensitive to lower concentrations of Ca 2ϩ than is ACVIII, and ACVIII is more stimulable by Ca 2ϩ /calmodulin than ACI (25). The calmodulinbinding site in ACI is rich in basic amino acids (net charge is ϩ7), and the binding is Ca 2ϩ -dependent (15,16). Since no hyperactivity was observed by mutating the C1b region on ACI (15,17) and the movement of the C1b region is likely to be more restrained than those of the N terminus and C terminus, the mechanism of Ca 2ϩ /calmodulin regulation of ACI might be to stabilize the C1/C2 heterodimer, as has been proposed for forskolin and G s␣a (9,10), unlike the disinhibitory mechanism we have proposed for ACVIII.
In conclusion, two calmodulin-binding sites exist on ACVIII, one (at the C terminus) is of profound regulatory significance, whereas the other (at the N terminus) plays a more minor role. Whether these two domains of ACVIII physically interact to share the same molecule of calmodulin remains to be determined in future studies. Given that the C1a and C2a regions clearly interact for catalytic activity (4,5,7,9,10,35), the tantalizing possibility that adenylyl cyclase could adopt a transporter-like structure (11,50) would be greatly strengthened by interactions between the N and C termini.