Genetic selection of mammalian adenylyl cyclases insensitive to stimulation by Gsalpha.

We describe the development of a genetic system allowing for the isolation of mutant mammalian adenylyl cyclases defective in their responses to G protein subunits, thus allowing for the identification of structural elements within the cyclase that are responsible for the recognition of these regulators. Expression of mammalian type V adenylyl cyclase in a cyclase-deleted yeast strain can conditionally complement the lethal phenotype of this strain. Type V adenylyl cyclase-expressing yeast grow only when the cyclase is activated by coexpression of Gsalpha or addition of forskolin to the medium; however, growth arrest is observed in the presence of both activators or under basal conditions. Utilizing this genetic system, we have isolated 25 adenylyl cyclase mutants defective in their response to Gsalpha. Sequence analysis and biochemical characterization of these mutants have identified residues in both cytoplasmic domains of the cyclase that are involved in the specific binding of and regulation by Gsalpha.

Regulation of intracellular cyclic AMP concentrations is principally controlled at the level of its synthesis, through the hormonal regulation of adenylyl cyclase, the enzyme responsible for the conversion of ATP into cyclic AMP. The adenylyl cyclase system comprises three components: seven transmembrane-spanning receptors for a variety of hormones and neurotransmitters, heterotrimeric G proteins, 1 and the catalytic entity itself. Currently, nine isoforms of membrane-bound adenylyl cyclases have been identified by molecular genetic approaches, and studies of these enzymes reveal both common and unique regulatory features (1,2). All isoforms tested to date are activated by the GTP-bound form of G s␣ and by forskolin; for some of the isoforms, such as the type V, these stimulators synergistically activate the enzyme. All isoforms of adenylyl cyclase are further regulated by additional inputs in an isoform-specific pattern. For example, G i␣ inhibits the types I, V, and VI isoforms (3)(4)(5); ␤␥ subunits can activate (types II, IV, and VII) or inhibit (type I) adenylyl cyclase activity (6 -10). Increases in intracellular calcium concentrations will inhibit the types V and VI isoforms (11)(12)(13) while indirectly activating (via a calmodulin-dependent process) types I and VIII (14,15) and inhibiting (by calmodulin kinase) the type III isoform (16).
Structural motifs responsible for the recognition of these regulatory molecules by the adenylyl cyclases are starting to be uncovered. A region of type II adenylyl cyclase (residues 956 -982 from the C 2 domain) containing a QXXER motif has been shown to be important for the regulation of this enzyme by G protein ␤␥ subunits (17). This site has been proposed to interact with the ␤␥ subunit of the G protein, within the aminoterminal 100 residues of ␤ (18,19). Synthetic peptide and mutational approaches have identified sequences located within the first cytoplasmic (C 1a ) region of type I adenylyl cyclase (residues 495-522) important for calmodulin activation (20,21); this sequence has a hydrophobic/basic composition and an aromatic amino acid in its NH 2 -terminal portion, typical of most calmodulin binding domains, and is therefore likely to function as the calmodulin binding domain in the cyclase. The recently solved crystal structure of a homodimer of the type II adenylyl cyclase second cytoplasmic domain (C 2 ) bound with forskolin has revealed the binding site(s) of this regulator (22). Mutational analysis of G s␣ has revealed three regions of the protein critical for the ability of G s␣ to activate adenylyl cyclase (23,24); however, at the time that this study was initiated, the G s␣ -binding site on the adenylyl cyclase molecule had not been determined.
The yeast adenylyl cyclase, encoded by the CYR1 gene, is structurally distinct from and exhibits much different regulation than the mammalian adenylyl cyclases (25). Yeast strains containing a disruption of the CYR1 gene are not viable but can be propagated in medium containing cAMP, demonstrating the importance of cAMP for growth of the organism (26). Based on this requirement of cAMP for growth, we have expressed mammalian adenylyl cyclases in a CYR1-deleted yeast strain to take advantage of the readily manipulatable genetic properties of the yeast for the isolation of mutant mammalian adenylyl cyclases defective in their regulatory responses. In this report, we describe a genetic system utilizing the expression of mammalian type V adenylyl cyclase in yeast and demonstrate its utility in isolating mutant adenylyl cyclases defective in their regulation by G s␣ . Genetic and biochemical analyses of these mutants reveal structural information about the G s␣ binding site on the cyclase, as well as regions of the molecule important for responses to activating regulators.
Construction of Plasmid Libraries of Mutant Type V Adenylyl Cyclase-Randomly mutated libraries of plasmid pADHprACVLeu were generated by passage in the mutator bacterial strain XL1 Red (Stratgene). Libraries with different extents of mutagenesis were prepared by isolating plasmids after 24, 48, and 72 h of growth in this strain. Libraries containing random mutations in the C 1 or C 2 domains were constructed by error-generated PCR mutagenic techniques. For the C 1 domain, sense (nucleotides 1125-1144) and antisense (nucleotides 1807-1826) primers were used and the PCR products were digested with NcoI and NdeI and ligated into these same sites present in the pADHprACVLeu yeast expression plasmid. C 2 domain mutant libraries were constructed by replacement of the carboxyl-terminal SphI fragment with the SphI-digested PCR fragment generated with the C 2 domain primers (sense ϭ nucleotides 2918 -2937, antisense ϭ nucleotides 3754 -3773). PCR was performed using standard conditions for Taq polymerase except MgCl 2 and deoxynucleotide triphosphate (dNTP) concentrations were as follows (for low, medium, and high level of mutagenesis, respectively): 1.5 mM MgCl 2 , 0.2 mM dNTP; 7 mM MgCl 2 , 1 mM dNTP; 7 mM MgCl 2 , 1 mM dCTP, 1 mM dGTP, 0.2 mM dATP, and 0.2 mM dTTP. PCR was performed for 30 cycles (1 min at 94°C, 1 min at 45°C, 1 min at 72°C).
Selection of Mammalian Type V Adenylyl Cyclase Mutants in Yeast-Plasmid libraries were introduced into yeast strain 12229 by standard yeast transformation protocols using lithium acetate. Transformants were plated on yeast plates containing CM/ϪLeu medium supplemented with 100 M CuSO 4 , 200 M forskolin, and 2% dimethyl sulfoxide and incubated at 30°C. Plasmids were recovered from yeast using Wizard Clean-up columns (Promega) and introduced into bacteria by electroporation as per manufacturer's specifications (Bio-Rad). Recovered plasmids were reintroduced into 12229 and TC41-1 yeast strains by transformation and replica-plated.
DNA Sequencing-Sequencing of the adenylyl cyclase mutants was performed using Thermal Sequenase (Amersham Pharmacia Biotech) as per procedures supplied by the manufacturer. In some cases, automated sequencing was performed using an Applied Biosystems sequencer (Applied Biosystems, division of Perkin-Elmer).
Retesting Mutations in Yeast-Plasmids encoding mutant adenylyl cyclases localized (by sequencing) to the C 1 region were produced by ligating the NcoI-NdeI fragment from isolated mutant plasmids into NcoI-NdeI-digested parental plasmid pADHprACVLeu. Plasmids encoding mutant adenylyl cyclases mapping to the C 2 region were produced by replacing the SphI-HindIII fragment (encoding the carboxyl terminus of type V adenylyl cyclase) in wild type plasmid pADHprACV-Leu with the SphI-HindIII fragment of the isolated mutant plasmid. In all cases, the replacement fragment contained only a single point mutation. These plasmids were reintroduced into 12229 and TC41-1 yeast strains by transformation and replica-plated to verify that the mutation correlated with the growth phenotype of the originally isolated mutant plasmid.
Generation of Recombinant Baculovirus-Generation of recombinant baculovirus was performed using the fastbac system (Life Technologies, Inc.) as per manufacturer's specifications. For the production of recombinant baculovirus encoding the wild type type V adenylyl cyclase, the BamHI fragment containing the entire coding region of dog type V was ligated to the BamHI-digested pfastbac transfer vector to generate pfastbacACV. Viruses encoding mutant adenylyl cyclases mapping to the C 1 region were produced by ligating the NcoI-NdeI fragment from isolated mutant pADHprACVLeu yeast expression plasmids into NcoI-NdeI-digested pfastbacACV. Viruses encoding mutant adenylyl cyclases mapping to the C 2 region were produced by replacing the SphI-HindIII fragment (encoding the carboxyl terminus of type V adenylyl cyclase) from pfastbacACV with the SphI-HindIII fragment of the isolated mutant pADHprACVLeu yeast expression plasmids.
Sf9 Cell Culture and Preparation of Cell Membranes-Procedures for the culture of Sf9 cells and the amplification of recombinant baculovirus have been outlined by Summers and Smith (28). Sf9 membranes containing individual adenylyl cyclase isoforms were prepared as described elsewhere (29).
Purification of G Protein Subunits-Recombinant G s␣ and G i␣ were synthesized in bacteria and purified as described by Lee et al. (30). Protein concentrations were estimated by staining with Amido Black (31). The ␣ subunits were activated by incubation with 50 mM Na-HEPES (pH 8.0), 5 mM MgSO 4 , 1 mM EDTA, 1 mM dithiothreitol, and 400 M GTP␥S at 30°C for 30 (G s␣ ) or 120 (G i␣ ) min (32); unbound GTP␥S was removed by gel filtration.
cAMP Assay-Yeast cells were grown to mid log phase and incubated with appropriate activators/inducers (forskolin/copper) for 3 h, harvested by centrifugation, and cAMP content was measured as described by Uno et al. (33).
Adenylyl Cyclase Assay-Adenylyl cyclase activity was measured using the procedure described by Smigel (34). All assays were performed for 10 min at 30°C in a final volume of 100 l containing 20 g of membrane protein and a final concentration of 10 mM MgCl 2 .
Adenylyl Cyclase Purification and Immunoblotting-Wild type and mutant type V adenylyl cyclase proteins were purified from Sf9 membranes using published methods (9). Samples were resolved by SDSpolyacrylamide gel electrophoresis, transferred to nitrocellulose, and immunoblotted as described (29), using a primary rabbit antibody specific for type V/VI adenylyl cyclase (AC V/VI ) (35) and a secondary 125 Iconjugated anti rabbit antibody (NEN Life Science Products). Proteins were visualized on a PhosphorImager (Molecular Dynamics).
G s␣ Binding Assay-The labeled G s␣ for binding experiments was prepared by incubating ϳ50 pmol of G s␣ with 1.2 M [ 35 S]GTP␥S (1250 Ci/mmol) for 1 h at 30°C in 20 mM Na-HEPES (pH 8.0), 5 mM MgSO 4 , 1 mM dithiothreitol, and 1 mg/ml bovine serum albumin in a final volume of 80 l. The reaction mixture was then gel-filtered through a Sephadex G-50 column to remove free [ 35 S]GTP␥S. The G s␣ binding assay was performed by mixing 20 g of Sf9 cell membranes expressing wild type or mutant adenylyl cyclase, 80 fmol of [ 35 S]GTP␥S-G s␣ (final concentration 4 nM) and unlabeled GTP␥S-activated G s␣ as described previously (5,36).
Data Analysis-Data were analyzed using the GraphPad Prizm program. Minimum values for the EC 50 for G s␣ stimulation were estimated from the dose-response curves; for curves that did not approach saturation, values represent concentrations of G s␣ that stimulate enzymatic activity half as much as the highest concentration of G s␣ tested (3 M).

FIG. 1. Expression of mammalian adenylyl cyclase in yeast.
Mammalian type V adenylyl cyclase was expressed in CYR1-disrupted strains of S. cerevisiae. The growth characteristics of these strains were evaluated under the following conditions: addition of 2 mM cAMP to the medium (cAMP), no regulator (Basal), addition of 100 M forskolin to the medium (Fsk), coexpression of G s␣ (G s␣ ), coexpression of G s␣ and addition of 100 M forskolin to the medium (Fsk ϩ G s␣ ), or coexpression of G s␣ and G i␣ (G s␣ ϩ G i␣ ).

FIG. 2. Growth phenotypes of adenylyl cyclase mutants.
Growth characteristics of CYR1-disrupted yeast expressing wild type or mutant type V adenylyl cyclase constructs were evaluated in the absence of cyclase activators (Basal), addition of 2 mM cAMP (cAMP), addition of 100 M forskolin (Fsk), coexpression of G s␣ (G s␣ ), or coexpression of G s␣ and addition of 100 M forskolin (Fsk ϩ G s␣ ). Intracellular G s␣ concentrations (under the control of the copper-inducible CUP1 promoter) were modulated by the omission of copper (G s␣ low), addition of 10 M CuSO 4 (G s␣ med), or addition of 100 M CuSO 4 (G s␣ high) to the growth medium.

RESULTS
Yeast (Saccharomyces cerevisiae) adenylyl cyclase (encoded by the CYR1 locus) is an essential gene product; cyr1 strains lacking a functional cyclase are non-viable but can be propagated in the presence of cAMP in the growth medium. As shown in Fig. 1, expression of the mammalian type V (ACV) adenylyl cyclase (under the control of the yeast ADH promoter) on a multicopy episomal plasmid can conditionally rescue the lethal phenotype of the CYR1-deleted strain TC41-1. The expression conditions employed have allowed us to discern the different regulatory states of the type V adenylyl cyclase based on growth characteristics of the yeast. In the absence of cyclase activators, TC41-1 cells expressing ACV fail to grow, presumably due to low cAMP levels produced by this isoform under basal conditions (0.02 pmol of cAMP/mg of protein). However, under activating conditions (in the presence of 100 M forskolin in the medium or coexpression of G s␣ ), these cells now grow. Growth was accompanied by an increase in intracellular cAMP levels (0.25 pmol of cAMP/mg of protein for either condition). In vitro, ACV is synergistically activated by G s␣ and forskolin in combination, and under these conditions we were surprised to observe a no growth phenotype rather than robust growth of this yeast strain despite measuring enhancement of intracellular cAMP levels (1.2 pmol of cAMP/mg of protein).
The observed failure of the yeast strain expressing ACV to grow under concomitant activation of the cyclase by forskolin and G s␣ suggested a strategy to isolate adenylyl cyclase mutants defective in their response to G s␣ . We reasoned that G s␣ -insensitive mutants should represent a subset of those that would allow for growth in the presence of G s␣ and forskolin. Toward this end, we have constructed libraries encoding randomly generated mutant type V adenylyl cyclases in a yeast expression vector, introduced these libraries into yeast strain 12229 (a derivative of TC41-1 expressing rat G s␣ under the control of the CUP1 copper-inducible promoter), and selected for transformants that grow in the presence of forskolin and copper. Several hundred colonies were isolated from Ͼ5 ϫ 10 6 transformants derived from nine libraries; three libraries differing in the extent of mutagenesis were generated for each target region of ACV (C 1a , C 2 , and the entire coding region). to carboxyl terminus (COOH), is as follows: M 1 , first set of membrane-spanning regions; C 1 , the first large cytoplasmic domain comprising conserved C 1a and variable C 1b subdomains; M 2 , the second set of membrane-spanning regions; C 2 , the second large cytoplasmic domain comprising conserved C 2a and variable C 2b subdomains. Alignment of G s␣ -regulated adenylyl cyclase sequences (types I-IX and Drosophila rutabaga) in the regions containing isolated G s␣ -insensitive mutants are shown in panels B (C 1a ) and C (C 2a ). Top line is the sequence of type V adenylyl cyclase with numbering corresponding to positions within this isoform; mutations are indicated by asterisks.
Positive colonies selected in these genetic screens were replica-plated on medium lacking forskolin to assess growth in the presence of differing levels of expressed G s␣ (controlled by the concentration of copper in the growth medium) alone. For mutants that exhibited reduced sensitivity to G s␣ stimulation (observable by lack of growth in the presence of G s␣ ), plasmids encoding the mutant cyclases were recovered and introduced into TC41-1 cells to assess their sensitivity to stimulation by forskolin and back into the parental yeast strain, 12229, to verify that the growth phenotype segregated with the mutant adenylyl cyclase plasmid. Fig. 2 depicts the growth characteristics of 13 independently isolated mutants defective in responses toward G s␣ . All of these mutants (isolated for their ability to grow in the presence of G s␣ and forskolin stimulation) fail to grow in the presence of G s␣ levels sufficient to obtain growth for the wild type ACV-expressing cells; maximal activation of the CUP1 promoter by high concentration of copper in the medium, however, will elevate G s␣ levels enough to permit growth of yeast expressing some of these mutant cyclases. All but one mutant (D424N) impart a growth phenotype in the presence of forskolin, consistent with a defect limited to G s␣ regulation. Growth of yeast (expressing these mutants or wild type) is slower at lower concentrations of forskolin and not detected at concentrations below 10 M (data not shown).
Twenty-five independent mutants 2 were subjected to DNA sequence analysis to localize the sites of mutations responsible for the observed phenotypes. Eleven positions (the C 1a and C 2 domains of the cyclase containing 2 and 9 sites, respectively) were identified, as shown in Fig. 3. Several point mutants were independently recovered multiple times from different mutant libraries (F379L and F1093S, four times; L967P and G1016S, three times; N1013D and N1094D, two times), and two positions were found mutated to two different amino acids (Leu-967 to Pro or Arg, and Asn-1094 to Asp or Tyr). As depicted in Fig.  3, all mutated positions are highly conserved among all mammalian adenylyl cyclase family members.
We next examined the biochemical properties of these mutant type V cyclases to determine whether defective regulatory properties of these mutants could account for the altered growth phenotypes observed when they are expressed in yeast. Unfortunately, characterization of the regulatory properties of mutant or wild type adenylyl cyclase in yeast homogenates or membranes was not possible due to extremely low enzymatic activities of these preparations. Therefore, the mutant adenylyl cyclase constructs were introduced into recombinant baculoviruses and overexpressed in Sf9 cells; membranes from these cells were then used as a source of cyclase for biochemical studies. To address potential differences in relative levels of active cyclase protein, the wild type and mutant adenylyl cyclase proteins were purified; equivalent amounts of these cyclases (based on activity measurements) were analyzed by Western blotting with an anti-AC5/6 antibody (data not shown). For comparative purposes, levels of active cyclase protein were normalized with respect to that of the wild type enzyme and are summarized in Table I; the values range from 43% to 385% of wild type and may reflect differences in expression levels or stabilities of the mutant cyclases. C 1a Mutants-As shown in Fig. 4  adenylyl cyclases in Sf9 membranes Expression levels of active type V adenylyl cyclase mutants in Sf9 membrane preparations were determined by Western blotting and are reported as the percent of wild type (WT) expression level. First, purified wild type or mutant adenylyl cyclase protein (equivalent activities) were quantitated by Western blot analysis to calculate the specific activities of the mutants relative to the wild type enzyme. Relative specific activity ϭ (catalytic activity loaded on Western blot)/(optical density of cyclase band as % of wild type). Next, expression levels were calculated by dividing the activity of Sf9 membrane preparations by the specific activity of the wild type and mutant enzymes. Relative expression level ϭ (catalytic activity of Sf9 membranes)/(relative specific activity of mutant cyclases) . Data represent the average of five independent experiments.  4. Regulation of C 1 adenylyl cyclase mutants by G s␣ . Membranes (20 g) from Sf9 cells expressing either the wild type recombinant type V adenylyl cyclase (q), the F379L mutant (f), or the D424N mutant (OE) were assayed for adenylyl cyclase activity in the presence of indicated amounts of GTP␥S-bound G s␣ . Activities are expressed as nmol⅐min Ϫ1 ⅐mg Ϫ1 , and have been adjusted to reflect the differences in mutant expression levels reported in Table I. Assays were performed in duplicate (bars, S.D.), and the results are representative of at least three experiments. Ͼ700, and Ͼ600 nM, respectively). We could not determine whether the maximal stimulation by G s␣ was also reduced by these mutations, because we were unable to reach saturating G s␣ concentrations for the mutant type V adenylyl cyclases.
We next examined the responses of these mutants to forskolin to determine whether these mutants had alterations in their response to this activator and to examine the ability of these mutants to integrate concomitant forskolin and G s␣ activation. Examination of Fig. 5 reveals that the sensitivity of the F379L mutant to forskolin (EC 50 ϭ 37 M) is comparable to wild type (EC 50 ϭ 44 M). In the presence of G s␣ , where synergistic activation of the wild type ACV is apparent as an increase in sensitivity to forskolin (EC 50 lowered to 6 M) and an increase in the magnitude of the forskolin response (Fig.  5A), the F379L mutant shows neither of these effects (Fig. 5B), further demonstrating its defective response to G s␣ . In marked contrast is the behavior of the D424N mutant shown in Fig. 5C; this mutant has a marked decreased sensitivity to forskolin (the dose-response curve does not appear to saturate at the highest concentration tested). Surprisingly, the synergistic activation of this mutant by forskolin plus G s␣ (EC 50 ϭ 5 M) resembles that of the wild type despite the observed loss of responsiveness to either activator alone. These observations suggest that the defect associated with the D424N mutant is a more global defect affecting regulation by multiple activators, whereas the F379L mutant is specifically defective in its response to G s␣ , likely due to a reduced ability to bind G s␣ .
To test this hypothesis, we examined the ability of these mutants to directly bind 35 S-labeled GTP␥S-bound G s␣ . As shown in Fig. 6 (inset), wild type ACV binds labeled G s␣ in the presence of forskolin, and this binding is competed by the addition of excess unlabeled G s␣ . No binding of labeled G s␣ could be detected for the F379L mutant under these conditions, consistent with the reduced sensitivity of this mutant toward G s␣ stimulation, and inability to respond to G s␣ in the presence of forskolin. The D424N mutant, by contrast, did bind to G s␣ , although the initial level of binding appears reduced, the affinity of this mutant for G s␣ is comparable to that of the wild type enzyme (data not shown). Taken together, these data suggest a role for Phe-379 in G s␣ binding and a more structural role for Asp-424 (see "Discussion").
C 2 Mutants-The majority of the G s␣ -insensitive mutants that were isolated in yeast were localized to the C 2 domain of ACV. Biochemical analysis of these cyclase mutants, after expression in Sf9 cells, reveals that all are defective in responding to stimulation by G s␣ . As shown in Fig. 7, all mutants have reduced responses to stimulation by G s␣ compared with the wild type enzyme. EC 50 values of these mutants for G s␣ are shown in Table II. FIG. 5. Regulation of C 1 adenylyl cyclase mutants by forskolin. Membranes (20 g) prepared from Sf9 cells expressing either the wild type adenylyl cyclase (A), the F379L mutant (B), or the D424N mutant (C) were assayed for adenylyl cyclase activity in the presence of indicated amounts of forskolin. The effect of forskolin on basal (empty symbols), or G s␣ -stimulated (filled symbols, 10 nM G s␣ ) adenylyl cyclase activity was assessed. Activities are expressed as nmol⅐min Ϫ1 ⅐mg Ϫ1 , and have been adjusted to reflect the differences in mutant expression levels reported in Table I 6. Binding of labeled G s␣ to C 1 adenylyl cyclase mutants. Sf9 membranes (20 g) containing the wild type or mutant type V adenylyl cyclase constructs were incubated with 4 nM [ 35 S]GTP␥Sbound G s␣ for 10 min at 30°C. Binding was performed in the presence of 10 M forskolin and assessed by filtering the reaction mixture through Duropore filters. Radioactivity associated with filters was quantified by scintillation counting; specific binding to wild type, mutant, and control membranes was determined by calculating the binding that could be competed with excess (500 nM) G s␣ . Graphs represent data averaged from three independent experiments (bars, S.D.); each data point within an experiment was performed in duplicate. Inset, the binding of labeled G s␣ to Sf9 membranes (20 g) from cells overexpressing the wild type adenylyl cyclase or control (Lac Z) was assessed in the presence of indicated amounts of unlabeled G s␣ . G s␣ binding was adjusted to reflect the differences in mutant expression levels reported in Table I. Assays were performed in duplicate (bars, S.D.), and the results are representative of at least three experiments.
All mutants tested respond to stimulation by forskolin; forskolin-stimulated activities (measured in the presence or absence of manganese) ranged from 30 to 230% of those of the wild type enzyme, and EC 50 values are similar to that of wild type (Table II).
As G s␣ potentiates the magnitude of the forskolin-stimulated activity for the wild type enzyme, the sensitivity of these mutants to G s␣ was further examined by assessing this synergism. These data are presented in Table II. Based on this analysis, the C 2 mutants fall into two distinct groups. Both cyclases containing mutations at position Leu-967 display normal synergistic activation by forskolin and G s␣ despite reduced ability to be stimulated by G s␣ alone. The ability of these mutants to bind G s␣ was assessed and is shown in Fig. 8. Consistent with the activity measurements, both mutants can bind G s␣ in the presence of forskolin albeit to a lesser extent than wild type ACV; some binding of G s␣ to L967R can also be detected in the absence of forskolin (data not shown). Taken together, these data suggest that this residue is not involved in direct interactions with G s␣ but rather, is important in more global integration of regulatory inputs. Consistent with this interpretation is the observation that these mutants also display impaired inhibition by G i␣ (data not shown).
The second group of C 2 mutants (F1006L, N1013D, G1016S,  Table  I. Assays were performed in duplicate (bars, S.D.), and results are representative of at least three experiments.

TABLE II
Forskolin and G s␣ -stimulated activities of C 2 adenylyl cyclase mutants Sf9 membranes (20 g) expressing wild type (WT) ACV or the indicated C 2 mutants were assayed for G s␣ -and forskolin-stimulated activities for 10 min at 30°C as described in the methods section. All assays were performed in duplicate and are representative of at least three experiments. a Cyclase activity was assessed in the presence of 100 M forskolin and 5 mM Mn 2ϩ . b Forskolin-stimulated activities in the presence of Mn 2ϩ were adjusted to reflect differences in adenylyl cyclase expression levels (see Table I). c Values for the EC 50 for forskolin stimulation were calculated from forskolin dose-response curves using GraphPad Prizm software. The plots comprise 8 points with forskolin concentrations ranging from 0.3 to 300 M; plots were fitted with sigmoidal dose-response functions.
d Minimum values for the EC 50 of G s␣ stimulation were estimated from the dose-response curves depicted in Fig. 7; values represent concentrations of G s␣ that stimulate enzymatic activity half as much as the highest concentration of G s␣ tested (3 M).
e Synergistic activation of adenylyl cyclase by forskolin and G s␣ was evaluated by comparing forskolin-stimulated catalytic activities (100 M forskolin) in the presence and absence of 10 nM G s␣ . Activities are expressed as the percent enhancement of forskolin-stimulated activity upon addition of G s␣ . V1017D, R1021Q, V1022A, F1093S, N1094D, and N1094Y) exhibits a reduced ability to respond to G s␣ as assessed by G s␣ activation (Fig. 7) and the ability to synergistically activate the enzyme in the presence of forskolin (Table II), but displays normal regulation by G i␣ (data not shown) and forskolin. All of these mutants except N1013D also have dramatically reduced ability to bind G s␣ (Fig. 8). These mutants (F1006L, G1016S, V1017D, R1021Q, V1022A, F1093S, N1094D, and N1094Y) therefore define residues critical for high affinity interaction of the adenylyl cyclase with G s␣ either by participating in G s␣ binding or maintaining the structural integrity of the G s␣ binding site. The N1013D mutant possesses properties intermediate to this group and the Leu-967 mutants discussed above; it displays a reduced ability to be activated by G s␣ , exhibits modest synergism between G s␣ and forskolin activation (62%), and consistent with this activity profile, is only able to bind G s␣ in the presence of forskolin.

DISCUSSION
The 25 G s␣ -defective adenylyl cyclase mutants isolated represent point mutations at 11 positions that are highly conserved among all the identified adenylyl cyclase isoforms; the mutations are localized to the cytoplasmic regions (C 1a and C 2 ) of the molecule. These mutations can be placed into two general groups based on their biochemical properties. The D424N, L967P and L967R are defective in some of their responses to G s␣ but also display other regulatory abnormalities. The defects of the second group of mutants are limited to G s␣ re-sponses (in all aspects tested), including G s␣ binding, activation, and synergistic interaction with forskolin activation (and, for those mutants examined, stimulation of forskolin binding); the biochemical properties of these mutants are consistent with the involvement of these residues in the G s␣ binding site.
Based on the homologies among the cytoplasmic domains of mammalian adenylyl cyclase isoforms and between the C 1 and C 2 regions within a given cyclase isoform, all but the Leu-967 mutants identified in this study can be mapped to positions on the recently solved crystal structure of the ACII C 2 homodimer (22). Asp-424 is analogous to the conserved Asp-923 of the B chain of the ACII homodimer located on helix 2. Closer inspection of this aspartate residue reveals that its carboxyl side chain intimately associates with Asn-1124 and Arg-1125 of the C 2 domain (ACII residues 1012 and 1013 of the A chain). Mutation of this aspartate would disrupt its ability to hydrogen-bond to the backbone nitrogen of these residues, thereby decreasing the interaction of the two cytoplasmic domains. In light of the finding that both G s␣ and forskolin activate adenylyl cyclase by increasing the affinity of C 1 for C 2 (37,38), the predicted consequence of mutating Asp-424 would be precisely what was observed for the D424N mutant: reduced sensitivity to stimulation by either forskolin or G s␣ alone, and a retention of the ability to synergistically integrate these stimulatory inputs.
The second group of mutants identified serves to identify three regions of the adenylyl cyclase molecule critical for its FIG. 8. Binding of labeled G s␣ to C 2 adenylyl cyclase mutants. Sf9 membranes (20 g) containing the wild type adenylyl cyclase construct or the indicated C 2 mutants were incubated with labeled G s␣ in the presence of forskolin. Specific binding to the wild type, mutant, and control membranes was assessed by calculating the radioactivity competed with 1 M unlabeled G s␣ . G s␣ binding was adjusted to reflect the differences in mutant expression levels reported in Table I interaction with G s␣ . Phe-1093 and Asn-1094 reside at the junction between ␣ helix 3 and ␤ strand 4; Phe-1006, Asn-1013, Gly-1016, Val-1017, Arg-1021, and Val-1022 lie at the base of (and loop preceding) ␣ helix 2. The residues in these two regions are highlighted (Fig. 9). The third region is in C 1a (Phe-379) and is represented in dark green in Fig. 9 (the actual residue highlighted is the amino-terminal most residue of the C 2 homodimeric structure, Glu-875, which is 2 residues carboxyl-terminal to the homologous residue of Phe-379). All of these residue (except Gly-1016) are located on the same surface of the adenylyl cyclase protein displaying a localized binding site for G s␣ docking. Residue Gly-1016 is buried, and it is likely that G1016S was isolated as a G s␣ -insensitive mutant in our genetic selection because this substitution would undoubtedly disrupt the geometry of the exposed surface of ␣ helix 2.
All mutations isolated in this study are located in positions highly homologous among all G s␣ -regulated adenylyl cyclases. It is interesting to note that the one exception, residue Asn-1013, is absolutely conserved among types I-VIII, but differs in type IX (Tyr) and Drosophila rutabaga (Ser) cyclases. This may explain the relatively less drastic biochemical defect observed for the isolated N1013D mutant; it retained much of its G s␣ binding and activation (in the presence of forskolin) despite showing decreased binding to and stimulation by G s␣ alone.
The fact that the G s␣ binding site that we have determined by the mutational analysis comprises residues located in both C 1 and C 2 domains is consistent with the observation that G s␣ increases the affinity of the cytoplasmic domains for each other (37,38). That most of the G s␣ -insensitive mutants isolated in this study map to residues in the C 2 domain is consistent with the observation that recombinant soluble C 2 domain constructs can bind to G s␣ in vitro, whereas the C 1 domain cannot but can be covalently attached to G s␣ in the presence of chemical crosslinkers (39). A recent study utilizing alanine substitutions of cyclase residues has defined a number of mutations that disrupt G s␣ activation of soluble C 1 (type I) and C 2 (type II) cyclase constructs analogous to several isolated in our random mutagenic approach (40). Our findings confirm the importance of C 1 residue Phe-379 (ACV numbering), and C 2 residues (Arg-1021, Val-1022, Phe-1093, and Asn-1094) in G s␣ activation, and further demonstrate the importance of these residues for G s␣ binding and their function in the context of the full-length membrane-bound enzyme.
It is unclear how G s␣ precisely binds to the cyclase because the three-dimensional structure of G s␣ is not yet solved; however, molecular modeling using the available crystallographic structure of G i␣ (41) suggests a possible orientation of this interaction. In the G i␣1 , a hydrophobic pocket is formed by Ile-212, Phe-215, and Trp-258 when GTP is bound. The homologous amino acids in G s␣ could accommodate Phe-1093 in type V adenylyl cyclase, thereby orienting residues (shown to be important for the ability of G s␣ to activate adenylyl cyclase; see Refs. 23 and 24) toward ␣ helix 2 and Phe-379 of the cyclase. We are currently using our yeast genetic system to identify second-site suppressors, i.e. mutations in G s␣ that have gained the ability to stimulate the G s␣ -insensitive adenylyl cyclase mutants, in an effort to identify the precise interactions between G s␣ and its binding site on adenylyl cyclase.
Expression of mammalian type V adenylyl cyclase in a cyclase-deficient mutant of S. cerevisiae has allowed us to take a genetic approach toward elucidating the structural basis underlying the regulation of adenylyl cyclase isoforms. The utility of this genetic system is evidenced by the isolation of mutant adenylyl cyclases defective in their regulation by G s␣ . The genetic system should be readily applicable to the selection of additional adenylyl cyclase mutants defective in their responses to other regulators such as G i␣ , G protein ␤␥ subunits, or protein kinases, thus leading to the identification of the binding sites for these molecules.