Isolation and Characterization of Constitutively Active Mutants of Mammalian Adenylyl Cyclase*

A genetic screen in Saccharomyces cerevisiae identified mutations in mammalian adenylyl cyclase that ac-tivate the enzyme in the absence of G s a . Thirteen of these mutant proteins were characterized biochemi-cally in an assay system that depends on a mixture of the two cytosolic domains (C 1 and C 2 ) of mammalian adeny- lyl cyclases. Three mutations, I1010M, K1014N, and P1015Q located in the b 4- b 5 loop of the C 2 domain of type II adenylyl cyclase, increase enzymatic activity in the absence of activators. The K1014N mutation displays both increased maximal activity and apparent affinity for the C 1 domain of type V adenylyl cyclase in the absence of activators of the enzyme. The increased affinity of the mutant C 2 domain of adenylyl cyclase for the wild type C 1 domain was exploited to isolate a com- plex containing VC 1 , IIC 2 , and G s a -guanosine 5 * -3- O -(thio)triphosphate (GTP g S) in the absence of forskolin and a complex of VC 1 , IIC 2 , forskolin, and P-site inhibi- tor in the absence of G s a -GTP g S. The isolation of these complexes should facilitate solution of crystal structures of low activity states of adenylyl cyclase and thus determination of the mechanism of activation of

A genetic screen in Saccharomyces cerevisiae identified mutations in mammalian adenylyl cyclase that activate the enzyme in the absence of G s ␣. Thirteen of these mutant proteins were characterized biochemically in an assay system that depends on a mixture of the two cytosolic domains (C 1 and C 2 ) of mammalian adenylyl cyclases. Three mutations, I1010M, K1014N, and P1015Q located in the ␤4-␤5 loop of the C 2 domain of type II adenylyl cyclase, increase enzymatic activity in the absence of activators. The K1014N mutation displays both increased maximal activity and apparent affinity for the C 1 domain of type V adenylyl cyclase in the absence of activators of the enzyme. The increased affinity of the mutant C 2 domain of adenylyl cyclase for the wild type C 1 domain was exploited to isolate a complex containing VC 1 , IIC 2 , and G s ␣-guanosine 5-3-O-(thio)triphosphate (GTP␥S) in the absence of forskolin and a complex of VC 1 , IIC 2 , forskolin, and P-site inhibitor in the absence of G s ␣-GTP␥S. The isolation of these complexes should facilitate solution of crystal structures of low activity states of adenylyl cyclase and thus determination of the mechanism of activation of the enzyme by forskolin and G s ␣.
Mammalian adenylyl cyclases are membrane-bound enzymes that catalyze the synthesis of the intracellular second messenger cyclic AMP from ATP. Nine isoforms of the enzyme have been detected, and they display characteristic regulatory properties and patterns of cellular distribution (1,2). Cellular rates of cyclic AMP synthesis are controlled by a variety of extracellular ligands that interact with heptahelical receptors in the plasma membrane. Relevant receptors can either stimulate cyclic AMP synthesis, usually via the intermediacy of a G protein 1 (G s ) that activates adenylyl cyclase, or inhibit cyclic AMP synthesis, often by interaction with an inhibitory G protein, G i . Mammalian adenylyl cyclases can also be activated by the diterpene forskolin (3) and inhibited by certain adenosine analogs and adenine nucleotides called P-site inhibitors (4). Certain adenylyl cyclases are also regulated by Ca 2ϩ , Ca 2ϩcalmodulin, and phosphorylation (5).
Mammalian adenylyl cyclases are integral membrane proteins that appear to contain two sets of six membrane-spanning helices that are separated by a large (ϳ40 kDa) cytoplasmic loop and followed by a similarly sized carboxyl-terminal cytosolic domain (6). The cytosolic domains, termed C 1 and C 2 , have been extensively studied; they are responsible for catalytic activity and most of the regulatory properties of the enzymes (7). The first 200 -250 amino acids of each cytosolic domain, designated C 1a and C 2a , are the most highly conserved regions among adenylyl cyclases. Strikingly, the C 1a and C 2a domains are approximately 50% similar and 25% identical to each other within a single isoform of adenylyl cyclase, and they are 20 -25% similar to the catalytic domains of membrane-bound and cytosolic guanylyl cyclases.
The C 1a and C 2a domains of adenylyl cyclases can be expressed separately and purified as recombinant proteins. When mixed together, they display the characteristics of membranebound adenylyl cyclase with respect to regulation by G s ␣, G i ␣, forskolin, and P-site inhibitors (8 -10). The crystal structures of the soluble catalytic core of adenylyl cyclase bound to G s ␣ and forskolin (11) and of this complex bound with the competitive substrates ␤-L-2Ј,3Ј-ddATP and ATP␣S (12) have provided detailed insights into mechanisms of catalysis of cyclic AMP synthesis and regulation of the activity of the enzyme.
Less information is available about the biochemical and structural properties of low activity states of adenylyl cyclase. Crystallization of the C 1 and C 2 domains in the absence of G s ␣ and/or forskolin would permit structural comparison of high and low activity states of the enzyme. We have approached this goal with a combination of genetic and biochemical techniques. Full-length mammalian adenylyl cyclase was introduced into a mutant strain of Saccharomyces cerevisiae that does not express the endogenous enzyme. Co-expression of the mammalian protein together with G s ␣ relieves dependence of cyclic AMP for growth at non-permissive temperatures. We have utilized a genetic screen to isolate constitutively active mutants of adenylyl cyclase within this strain. We describe the isolation of several partially active mutants, as well as the biochemical consequences of these mutations in the context of the soluble, recombinant adenylyl cyclase system.

EXPERIMENTAL PROCEDURES
Genetic Screen-To achieve high level constitutive expression of type II adenylyl cyclase (ACII) in yeast, a 0.94-kb fragment containing the S. cerevisiae PGK1 promoter and a 3.3-kb fragment encoding rat ACII (PCR-amplified from a template plasmid provided by Randall Reed, The John Hopkins University) were subcloned into pRS425 (13). The resulting expression plasmid, Cp1512 (genotype LEU2 PGK1p-ACII 2-ori REP3 bla), was used as the template for error-prone PCR amplification of two separate fragments as follows: nucleotides 5514 -6384, which encode the C 1a domain, and nucleotides 7033-8264, which encode the C 2a domain (primer sequence available upon request). The mutagenized fragments encoding the C 1a domain were then used to transform yeast strain TC41F2-1 (genotype MATa cyr1::ura3 cam1 cam2 cam3 leu2-3 leu2-112 his3-532 his4 ura3; provided by Warren Heideman, University of Wisconsin) along with a "gapped" vector prepared as follows. Cp1512 was digested with ApaI (which cuts at nucleotide 5682 of ACII) and DraIII (which cuts at nucleotide 6254), and the 10.1-kb fragment, which lacks the C 1a coding sequence, was isolated. Similarly, TC41F2-1 was co-transformed with mutagenized fragments encompassing the C 2a domain and a gapped vector prepared as follows. Cp1512 was digested with PvuII (which cuts at nucleotide 7111) and BamHI (which cuts at nucleotide 8168), and the 9.7-kb fragment, which lacks the C 2a coding region, was isolated. For both transformations, cells were plated on SC-Leu (14) without supplemental cyclic AMP and incubated at room temperature. Colonies from these plates were replica-plated to SC-Leu plates and incubated at 34°C. Colonies that grew at 34°C were expected to harbor stably replicating episomes that result from in situ recombination between the gapped vector and a PCR-amplified fragment and that, as a result of this recombination, express an adenylyl cyclase that has constitutive (i.e. G s ␣-independent) activity.
In all cases, the plasmid dependence of growth under selective conditions (on SC-Leu at 34°C) was further tested by allowing spontaneous loss of plasmid under non-selective conditions (on YEPD with supplemental cyclic AMP at 30°C), followed by re-testing under selective conditions. Plasmids were rescued from those colonies whose growth under selective conditions was dependent on plasmid; the plasmids were then reintroduced into naive host strain TC41F2-1, and the transformants were tested for growth on SC-Leu at 34 and at 37°C. All plasmids reported here confer plasmid-dependent growth at 34°C when introduced into TC41F2-1.
Mutagenesis of Adenylyl Cyclase Domains-The mutants that appeared to confer the highest level of constitutive activity in yeast were analyzed using the soluble mammalian adenylyl cyclase system. The vectors pQE60-H6-VC 1 (364 -591)FLAG (10), pQE60-IIC 2 -H6 (8), and pQE60-ArgC-IIC 2 (11) served as templates for site-directed mutagenesis (QuikChange TM , Stratagene). The mutations in the C 1a domain were made in the conserved residues in the C 1a domain from canine type V adenylyl cyclase corresponding to the mutations in the rat type IIC 1 domain obtained from the yeast screen. The mutations in the C 2 domain from the yeast screen were made in the C 2 domain of rat type II adenylyl cyclase using the vector pQE60-IIC 2 -H6. The vector pQE60-ArgC-IIC 2 was used to generate a non-tagged IIC 2 -K1014N. Sequences of synthetic mutagenic sense and antisense primers are available upon request.
Expression and Purification of Proteins-Wild type and mutant H6-VC 1 (591)FLAG and IIC 2 -H6 were expressed in Escherichia coli and purified as described previously (8,10). Non-tagged ArgC-IIC 2 wild type and mutant K1014N were expressed and purified as described (11). Bovine G s ␣ (short form) was purified and activated by GTP␥S as described (15).
Adenylyl Cyclase Assays-Adenylyl cyclase activity was measured as described by Smigel (16). All assays were performed in a volume of 100 l for 10 min at 30°C. The final concentration of MgCl 2 was 10 mM. The concentration of ATP was 1 mM unless otherwise stated. Kinetic constants were determined by varying MgATP from 20 M to 5 mM with a fixed excess of Mg 2ϩ . Unless otherwise stated, assays contained a limiting concentration of VC 1 and an excess of IIC 2 for both wild type and mutant assays. All specific activities reported are with respect to the concentration of the limiting domain. Each experiment was repeated two to three times.
Gel Filtration-Proteins (G s ␣ and adenylyl cyclase domains) were separated by fast protein liquid chromatography with tandemly arranged Superdex 75 and 200 (HR10/30) columns. Proteins were eluted in a buffer containing 20 mM NaHepes, pH 8.0, 2 mM MgCl 2 , 1 mM EDTA, 2 mM dithiothreitol, and 50 mM NaCl. Flow rates were 0.2 ml/min, and 300-l fractions were collected. An aliquot of each fraction

RESULTS
Genetic Screen-We isolated 24 plasmids that conferred cyclic AMP-independent growth at 34°C when introduced into strain TC41F2-1; this phenotype is presumed to reflect the expression of G s ␣-independent adenylyl cyclase activity. The mutagenized regions of these plasmids were sequenced, and their mutations are listed in Table I. Cp4465, which encodes a Growth was determined by visual inspection of streaked colonies. The rate of growth was given a relative score so that a colony with a growth of 2 grew to a given size 4 times faster than a colony with a growth of 0.5. TC41F2-1 transformed with CP1512, which encodes wild type ACII, is 0.25 at 34°C and 0 at 37°C. ACII I259V/Y402C, was used to generate the two plasmids ACII I259V and ACII Y402C; similarly, Cp4522 was used to generate two plasmids encoding ACII M253V and ACII C305R. In this way it was found that ACII C305R and ACII Y402C are constitutively active mutants. In several cases mutations that were observed in genetically selected plasmids were engineered by site-directed mutagenesis into wild type ACII to test their effects independently of coincident mutations. Table II lists the mutations and the growth characteristics of all constitutively active ACII single mutants.
It is possible that plasmid-dependent variations in copy number could account for the phenotypes conferred by the ACII mutants. We addressed that possibility by subcloning six of the genetically selected missense mutants into centromeric vectors whose copy number is stably maintained at 1-3 per cell (17). In each case the phenotype conferred by the centromeric plasmid was the same as that imparted by its high copy equivalent (data not shown). We thus suggest that all 14 mutants listed in Table II have elevated specific activities.
Adenylyl Cyclase Assay-The mutations identified in the yeast screen were generated in the soluble mammalian adenylyl cyclase system to facilitate biochemical characterization. The mutations were made in the C 1a domain of canine type V adenylyl cyclase at the residues that correspond to those mutated in the C 1a domain of rat type II adenylyl cyclase. The mutations F298Y, C305R, N315S, K334R, and V377I from rat ACIIC 1a were made to the following residues in canine ACVC 1a : F400Y, C407R, N417S, K436R, and V479I. A similar screen in yeast using rat type IV adenylyl cyclase 2 generated mutations Y265H, V388I, G968S, and K998N; these mutations were made in VC 1 and IIC 2 , respectively, as Y383H, V506I, G970S, and K1014N. No conversion was necessary for the remainder of the mutations since they were already in rat ACIIC 2 . All mutant proteins, with the exception of C407R, K436R, and V506I, were expressed and purified to degrees comparable to levels described previously for the wild type proteins. The mutants were assayed for activity at limiting concentrations of the VC 1 domain and increasing concentrations of the IIC 2 domain in the absence of activators or in the presence of forskolin. The apparent affinity between the two soluble domains of adenylyl cyclase was expressed as the EC 50 , and the apparent V max was extrapolated from the asymptote of the curve. The apparent EC 50 and V max values for the mutant proteins are summarized in Table III.
Surprisingly, most mutant proteins displayed only modest differences in the apparent affinity or V max when compared with the corresponding wild type domain. Mutants C407R and K436R yielded very little protein with no measurable adenylyl cyclase activity. The V506I mutation in VC 1 caused an increase in basal enzymatic activity and a 3-fold decrease in EC 50 in the presence of forskolin. However, the low level of expression of this protein precluded detailed characterization. The mutations I1010M, K1014N, and P1015Q, all located in the ␤4-␤5 loop of IIC 2 , caused an increase in V max in both the basal and the forskolin-stimulated conditions.
The K1014N mutation appeared to cause the largest degree of constitutive activation; IIC 2 -K1014N was thus purified to homogeneity by anion-exchange chromatography and characterized biochemically. The apparent affinity of IIC 2 -K1014N for wild type VC 1 was determined in various activation states. The activities shown in Fig. 1 were determined with a limiting concentration of VC 1 and either variable concentrations (Fig. 1, A-C) or a saturating concentration (Fig. 1D) of IIC 2 . The apparent affinity for VC 1 and the maximal catalytic activity were both 3-fold greater for the K1014N mutant compared with wild type IIC 2 in the absence of activators (Fig. 1A). Similarly, the apparent affinity of IIC 2 -K1014N for VC 1 was 3-fold higher in the presence of forskolin (Fig. 1B) and 10-fold higher in the presence of G s ␣-GTP␥S (Fig. 1C) compared with its wild type counterpart. No significant changes were observed in V max under these conditions. There was no difference in either the EC 50 or the V max in the presence of both forskolin and Gs␣-GTP␥S, demonstrating that the mutation does not create a hyperactive enzyme. Similar results were observed using limiting concentrations of IIC 2 and varying concentrations of VC 1 (data not shown). The apparent affinity of G s ␣-GTP␥S for adenylyl cyclase is shown in Fig. 1D. The EC 50 for G s ␣-GTP␥S was 0.05 M for K1014N compared with 0.4 M for the wild type IIC 2 . These values were 6 and 25 nM, respectively, in the presence of forskolin.
Determination of Kinetic Constants-The kinetic constants for substrate were also determined under various conditions ( Fig. 2 and Table IV). Reconstituted adenylyl cyclase containing IIC 2 -K1014N exhibited a K m value for ATP that was 6-fold less than that observed with the wild type protein under basal conditions. As noted above, V max is increased under this condition. Assays performed with activated G s ␣ revealed a 6-fold decrease in the K m value for ATP when the mutation was present. No changes were obvious in the presence of forskolin, with or without G s ␣-GTP␥S.
Gel Filtration of the VC 1 ⅐IIC 2 -K1014N⅐G s ␣ Complex-Purified H6-VC 1 and wild type or IIC 2 -K1014N were combined with G s ␣-GTP␥S and gel-filtered using tandem Superdex 75 and 200 columns. Fractions were analyzed by SDS-PAGE (Fig. 3). In the absence of forskolin, there is no evidence for formation of a complex between VC 1 , wild type IIC 2 , and G s ␣ (Fig. 3B). The largest apparent species (78 kDa) is likely a heterodimer consisting of IIC 2 and G s ␣; similar results have been reported previously (10). In contrast, protein in the mixture of VC 1 , IIC 2 -K1014N, and G s ␣ eluted as two major peaks with the largest species representing a 100-kDa complex. Analysis by SDS-PAGE indicates a complex of VC 1 , IIC 2 -K1014N, and G s ␣ with an apparent stoichiometry of 1:1:1 (Fig. 3A). Similar results were observed with wild type IIC 2 only when forskolin was present (10).
Isolation of a Complex of VC 1 ⅐IIC 2 ⅐Forskolin⅐2Јd,3Ј-AMP⅐ PP i -Purified H6-VC 1 and non-tagged wild type or IIC 2 -K1014N were combined and applied to a metal chelate chro- a Growth was determined by visual inspection of streaked colonies. The rate of growth was given a relative score so that a colony with a growth of 2 grew to a given size 4 times faster than a colony with a growth of 0.5. TC41F2-1 transformed with CP1512, which encodes wild type ACII, is 0.25 at 34°C and 0 at 37°C. b These mutants are in ACIIC 2 .
matographic column (Talon TM , CLONTECH). Samples were eluted with imidazole and analyzed by SDS-PAGE (Fig. 4). No complex was detected by SDS-PAGE when H6-VC 1 and IIC 2 or IIC 2 -K1014N were incubated with forskolin. Some IIC 2 -K1014N was retained on the column in the presence of the P-site inhibitor 2Јd,3Ј-AMP⅐PP i ; wild type IIC 2 was not. When both forskolin and 2Јd,3Ј-AMP⅐PP i were present, a complex of H6-VC 1 and either IIC 2 or K1014N was isolated. These complexes had apparent stoichiometries of 1:1 (determined by scanning densitometry). Similar results were obtained when Mn 2ϩ replaced Mg 2ϩ (data not shown).

DISCUSSION
The crystal structure of the cytosolic portions of adenylyl cyclase demonstrates that the C 1 and C 2 domains are arranged as a pseudo-2-fold symmetrical dimer (see inset in Fig. 5) (11). The contributions of several residues within each domain to substrate and Mg 2ϩ binding, as well as catalysis, have been investigated in previous studies. Adenylyl cyclases, and presumably guanylyl cyclases, contain palm domains. These domains were defined previously in DNA polymerases, enzymes that catalyze very similar reactions (18). Crystal structures of the cytosolic domains of adenylyl cyclase have revealed significant conformational changes upon substrate binding (11). The ␣1 and ␣2 helices and the ␣3 and ␤4 helix/strand of C 1 and the ␤7-␤8 loop of C 2 collapse around the nucleotide and align the nucleotide and two metal ions for catalysis. Located within the active site is the ␤2-␤3 loop of C 1 , containing aspartate residues 396 and 440 that coordinate two Mg 2ϩ ions. These divalent cations participate in deprotonation of the 3Ј hydroxyl of the ribose moiety (a critical step in the synthesis of cyclic AMP) and stabilize the pentavalent transition state. The conserved aspartate residues are also found among DNA polymerases and  guanylyl cyclases (11, 12, 19 -22).
The ␤4-␤5 loop of C 2 supports the ␤2-␤3 loop of C 1 (Fig. 5). Perturbations in either the contact regions or the loop-fold could have dramatic effects on C 1 :C 2 structure and hence activity. Several residues in both the ␤2-␤3 loop and the ␣2 helix of C 1 and the ␤4-␤5 loop of C 2 have been investigated by site-directed mutagenesis and have various effects on adenylyl cyclase activity. Structural evidence strongly suggests that Asp-424 in the ␣2 helix and Arg-434 in the ␤2 sheet of C 1 engage in extensive hydrogen bonding with the ␤4-␤5 loop of C 2 (Fig. 5). Asp-424 forms a salt bridge with Arg-434 and a hydrogen bond with the backbone nitrogens of Ala-1012 and Gln-1013. The side chain of Arg-434 forms a hydrogen bond with the side chain of Gln-1016; the backbone carbonyl of Arg-434 forms a hydrogen bond with the side chain of Gln-1013. Mutations of either Asp-424 or Arg-434 have previously been shown to have detrimental effects on adenylyl cyclase activity. Mutations of these residues have broad effects on cyclase activity without affecting the affinity of the enzyme for G s ␣, as follows: R434A increases the IC 50 value for P-site inhibitors (23); R434S increases the K m value for MgATP, the K i value for ATP␣S, and the EC 50 value for Mg 2ϩ (24); D424A and D424N decrease forskolin-and G s ␣-stimulated enzymatic activity (23,25).
Mutations of residues in the ␤4-␤5 loop of C 2 have also been shown to affect adenylyl cyclase activity. The mutations Y1017A and D1018A (Y999A and D1000A in type I adenylyl cyclase) obliterate activity without eliminating G s ␣ binding (23). Asp-1018 coordinates substrate binding through the pu-   rine ring and is responsible for dictating nucleotide specificity (11,26). Alteration in neighboring side chains that perturb the conformation of the amino acid chain backbone would likely perturb activity. Residues Ile-1010, Lys-1014, and Pro-1015, investigated in this study, are all located in the ␤4-␤5 loop of C 2 and therefore are intimately involved in the arrangement of the ␤2-␤3 loop of C 1 . The mutations I1010M, K1014N, and P1015Q are all within bonding distance of the ␤2-␤3 loop of C 1 . The P1015Q mutation, which displays slightly elevated affinity between C 1 and C 2 , was previously identified as a second site suppresser of a catalytically inactive mutant (N1025S) but unfortunately was not characterized alone (27). Prediction of the positions of the side chains of the ␤4-␤5 loop of C 2 is difficult because of its flexibility; however, some explanations of the activating mutations can be extracted from the crystal structure. We expect K1014 upon mutation to asparagine to pack between adjacent glutamines in the ␤4-␤5 loop of C 2 and form a stabilizing hydrogen bond with Arg-434 in the ␤2-␤3 loop of C 1 . P1015Q could rearrange the ␤4-␤5 loop, causing a more active conformation. I1010M fills space in the hydrophobic pocket with a larger hydrophobic residue. The introduction of new side chain interactions contributed by mutations I1010M or K1014N or the removal of main chain constraints with mutation P1015Q may alter the C 1 ␤2-␤3 loop and enhance activity. Taken together, these mutations suggest that proper formation of a competent active site is inhibited by decreased and promoted by increased interactions between the ␤4-␤5 loop of C 2 and the ␤2-␤3 loop of C 1 .
The ␤4-␤5 loop of the C 1 domain has a congruous interaction with the ␤2-␤3 loop of C 2 because of the pseudosymmetrical structure of adenylyl cyclase. Of the clones obtained from the genetic screen, only V506I displayed significant enhancement of activity when tested in the soluble adenylyl cyclase system. V506I adds a methyl group that may form a primary contact with forskolin and increase the hydrophobicity of the forskolinbinding pocket. Another possible explanation is that substitution of the isoleucine may enhance C 1 -C 2 interactions by altering van der Waals contacts with neighboring residues in the ␤4-␤5 loop of C 1 . The structural effect of this minor change at the base of the loop may be amplified along the length of the loop and thus alter interactions with the ␤2-␤3 loop of C 2 .
It is difficult to determine why other mutations displayed strong phenotypes in yeast but failed to produce substantial changes in the soluble adenylyl cyclase assays. As demonstrated in this study, mutations that increase favorable interactions between the ␤2-␤3 loop of C 1 and the ␤4-␤5 loop of C 2 likely account for the increased activity that was observed. Alternatively, increased activity may be related to the method of protein expression, since the soluble C 1 and C 2 domains remain as homodimers when expressed and purified individually. Mutations in the interface region may alter homodimerization. Mutations that impair homodimerization may favor heterodimerization and hence increase adenylyl cyclase activity. More likely, the lack of increased basal activity in the in vitro assays of many of the mutants could be explained by the inherent sensitivity to small changes in cyclic AMP concentrations of the yeast screen. The lack of change in activity in vitro could also be explained by the ablation of the membrane domains and the putative regulatory C 1b domain in the soluble constructs. For example, the F400Y mutation has been shown to increase both basal activity and sensitivity to the activators G s ␣ and forskolin and to abrogate inhibition by G i ␣ (28). However, when assayed in the soluble system, this mutation caused no increase in basal activity or sensitivity to forskolin compared with the wild type enzyme. The contribution of the C 1b domain to activity are not known, nor is there any structural information on this domain.
As mentioned previously, all complexes of adenylyl cyclase whose structures have been determined to date contain both G s ␣ and forskolin. Endogenous forskolin-like substances have yet to be discovered, begging the questions of the physiological significance of the C 1 ⅐C 2 ⅐forskolin⅐G s ␣-GTP␥S structure and the degree to which it resembles the structure of C 1 :C 2 :G s ␣-GTP␥S. Or rather, what is the mechanism of activation of adenylyl cyclase by forskolin? This is a particularly interesting question, since some forms of adenylyl cyclase (types II, IV, and V-VII) are activated synergistically by forskolin and G s ␣-GTP␥S, whereas others (2) are activated only additively (types I, III, and VIII). The K1014N mutation facilitated the isolation of a C 1 ⅐C 2 ⅐G s ␣-GTP␥S complex in the absence of forskolin. Determination of the structure of this complex would further our understanding of activation of adenylyl cyclase by the diterpene.
The crystal structure of a homodimer of the C 2 domain bound with two molecules of forskolin has also been determined (29). FIG. 5. Catalytic core of adenylyl cyclase. Inset, the heterodimeric complex formed by VC 1 (khaki) and IIC 2 (mauve) viewed along its pseudo-2-fold axis toward the hypothesized cytoplasmic face. Forskolin (Fsk) and ATP bind between VC 1 and IIC 2 and are shown as stick models. The ␤4-␤5 loop of IIC 2 containing I1010M, K1014N, and P1015Q is highlighted in green. Bottom right, interactions of ␤4-␤5 loop of IIC 2 . The side chains of the ␤4-␤5 loop (green) of IIC 2 (mauve) and their interactions with VC 1 (khaki) are shown. Dashed gray lines show side chain-side chain and side chain-main chain hydrogen bonds. Carbon atoms are gray, nitrogen atoms blue, and oxygen atoms red.
The structure also contains a 2-fold symmetrical arrangement of the domains and has been used as a model for the basal, nonactivated form of adenylyl cyclase. However, the presence of two forskolin molecules in a complex without an active site inherently precludes this structure as a precise model of nonactivated adenylyl cyclase. It is our hope that the structure of the constitutively active mutants described herein, particularly that of VC 1 associated with IIC 2 -K1014N, may represent a closer approximation of the low activity basal state.