The Catalytic Mechanism of Mammalian Adenylyl Cyclase

The mechanism of P-site inhibition of adenylyl cyclase has been probed by equilibrium binding measurements using 2′-[3H]deoxyadenosine, a P-site inhibitor, and by kinetic analysis of both the forward and reverse reactions (i.e. cyclic AMP and ATP synthesis, respectively). There is one binding site for 2′-deoxyadenosine per C1/C2 heterodimer; the K d is 40 ± 3 μm. Binding is observed only in the presence of one of the products of the adenylyl cyclase reaction, pyrophosphate (PPi). A substrate analog, Ap(CH2)pp (α,β-methylene adenosine 5′-triphosphate), and cyclic AMP compete for the P-site in the presence of PPi, but P-site analogs do not compete for substrate binding (in the absence of PPi). Kinetic analysis indicates that release of products from the enzyme is random. These facts permit formulation of a model for the adenylyl cyclase reaction, for which we provide substantial kinetic support. We propose that P-site analogs act as dead-end inhibitors of product release, stabilizing an enzyme-product (E-PPi) complex by binding at the active site. Although product release is random, cyclic AMP dissociates from the enzyme preferentially. Release of PPiis slow and partially rate-limiting.

Adenosine and various analogs of the nucleoside have both stimulatory and inhibitory effects on adenylyl cyclase activity (reviewed in Ref. 1). Londos and Wolff (2) categorized these effects mechanistically, based on their structure-activity relationships. Two types of adenosine-reactive sites were identified: those with strict requirements for the ribose moiety, designated R sites, and those with strict structural constraints for interaction with the purine ring, designated P sites. R sites are the ligand-binding sites of adenosine-specific G protein 1 -coupled receptors, which can either stimulate or inhibit adenylyl cyclase activity indirectly, while P sites, whose occupancy inhibits cyclic AMP synthesis, are structural features of adenylyl cyclases themselves (2)(3)(4)(5)(6)(7). The physiological significance of Psite inhibition is unclear, but concentrations of 3Ј-AMP found in vivo appear sufficient to inhibit adenylyl cyclase activity (8).
Using engineered, soluble forms of mammalian adenylyl cyclase, we and others have shown that the conserved cytosolic domains of the enzymes contain the structural components necessary for G s␣ -and forskolin-stimulated adenylyl cyclase activity, as well as the characteristic features of P-site inhibition (14 -18). We have also utilized the competitive substrate analog Ap(CH 2 )pp to identify a single substrate binding site on the enzyme (19). Binding of Ap(CH 2 )pp to adenylyl cyclase is completely unaffected by the addition of a potent P-site inhibitor. Neither of the two cytosolic domains of adenylyl cyclase contains a classical nucleotide binding motif, although these domains share approximately 200 amino acid residues of similar sequence. These observations add credence to the possibility that the substrate binding and P sites on the enzyme are structurally distinct.
The experiments described herein further define the nature of the P site and the mechanism of P-site inhibition. P-site inhibition is also used as a tool to facilitate understanding of the reaction catalyzed by adenylyl cyclases.

EXPERIMENTAL PROCEDURES
Materials-2Ј-[ 3 H]Deoxyadenosine (7 Ci/mmol) was purchased from ICN and lyophilized regularly to remove [ 3 H]H 2 O. Ap(CH 2 )pp was also purchased from ICN. Hexokinase and type XI glucose-6-phosphate dehydrogenase were purchased from Boehringer Mannheim and Sigma, respectively; these preparations were centrifuged, and the pelleted enzymes were resuspended in 20 mM NaHepes (pH 8.0) prior to use.
Protein Purification-Recombinant G s␣ and the two cytosolic domains of adenylyl cyclase, VC 1 (591)Flag and IIC 2 , were purified after expression in Escherichia coli as described (16,18,20). G s␣ was activated (whenever used) by incubation with 50 mM NaHepes (pH 8.0), 20 mM MgSO 4 , 1 mM EDTA, 2 mM dithiothreitol, and 400 M GTP␥S at 30°C for 30 min; free GTP␥S was then removed by gel filtration.
Adenylyl Cyclase Assays-Synthesis of cyclic AMP was measured as described (21) for 10 -15 min at 30°C in a final volume of 100 l. GTP␥S-G s␣ was present (400 nM) unless otherwise indicated. Activities are expressed per mg of the limiting adenylyl cyclase domain in the assay (VC 1 ). The other cytosolic domain (IIC 2 ) was present in excess (1 M) to drive the interaction between the two protein fragments. To determine kinetic constants, the concentration of MgATP was varied from 10 M to 2.56 mM with a fixed excess of 10 mM Mg 2ϩ . Initial velocities were linear with time, and less than 10% of the ATP was consumed at the lowest substrate concentrations. All points were measured in duplicate and experiments shown were repeated two to four times. Values are reported Ϯ S.E. of the mean.
Synthesis of ATP from cyclic AMP and PP i , the reverse reaction, was measured spectrophotometrically in the presence of glucose, hexokinase, NADP, and glucose-6-phosphate dehydrogenase. Reaction velocities were calculated from the linear increase in A 340 resulting from the reduction of NADP. Reactions contained 20 mM NaHepes (pH 8.0), 50 mM glucose, 0.8 mM NADP, 3 mM free MgCl 2 , 1.7 units of hexokinase, and 0.3 units of glucose-6-phosphate dehydrogenase in a volume of 500 l. The concentrations of substrates varied: 1.25-20 mM for cyclic AMP and 0.125-4 mM for MgPP i . PP i was always added last to avoid precipitation. Reactions were typically started by addition of 0.4 M VC 1 , 2 M IIC 2 , and 1 M GTP␥S-G s␣ (final concentrations) to the other reaction components (at 30°C), and absorbance changes were measured for 10 -15 min in a Beckman DU650 spectrophotometer with a temperaturecontrolled cuvette holder. The increase in absorbance in the absence of adenylyl cyclase was subtracted as background (Ͻ0.006 OD units/min). Optical densities of greater than 1.5 were excluded from analysis.
Equilibrium Dialysis-Equilibrium dialysis was performed essentially as described (19). To quantify binding of 2Ј-deoxyadenosine, each chamber contained 20 mM NaHepes (pH 8.0), 2.5 mM MgCl 2 , 2 mM dithiothreitol, 75 mM NaCl, 2Ј-[ 3 H]deoxyadenosine (12.5-250 M), and other additions as indicated. One chamber contained 18 M VC 1 , 18 M IIC 2 , and 25 M GTP␥S-G s␣ ; both of the cytosolic domains of adenylyl cyclase were necessary to observe binding. The opposite chamber contained buffer in lieu of the proteins. Samples were removed after dialysis for 24 h at 4°C with rotation. Duplicate 15-l aliquots from each chamber were analyzed by liquid scintillation spectrometry. Binding data have been normalized to protein concentrations based on the amount of active protein in the preparation, which was determined by titration with GTP␥S-G s␣ (19). Preparations of VC 1 and IIC 2 were more than 60% active by this criterion.
Steady-state Kinetic Model-We derived the rate equation for the reaction scheme presented in Fig. 5 using steady-state assumptions for the catalytic steps and rapid equilibrium for inhibitor binding. The critical features of the model include random release of products and no binding of P-site inhibitor to the free enzyme. The rates of both the forward and reverse reactions are complicated functions, described as where v is the velocity in M/s and [E t ] is the total molar concentration of enzyme. The equations that describe this function are presented in the "Appendix." Because of their complexity, they have not been rearranged to conform to standard Michaelis-Menten format. Curve fitting of individual experiments and modeling of kinetic data based on the steady-state rate equations were performed using Sigma Plot (Jandel Scientific).

P-site Inhibition of Adenylyl
Cyclase-Inhibition of the G s␣stimulated catalytic activity of our preparation of adenylyl cyclase by the P-site inhibitor 2Ј-deoxyadenosine is uncompetitive with respect to MgATP (Fig. 1). These reactions (and those described below) were performed with a limiting concentration of the C 1 domain of type V adenylyl cyclase and an excess of the C 2 domain of type II adenylyl cyclase to drive the proteinprotein interaction that is necessary for catalytic activity. Similar uncompetitive kinetics has been observed previously with a more potent P-site inhibitor, 2Ј-deoxy-3Ј-AMP, using G s␣ -stimulated bovine brain adenylyl cyclase (22) (presumably a mixture of isoforms) or a soluble system in which the two cytosolic domains of the enzyme were linked covalently (15). Although 2Ј-deoxyadenosine has an unfortunately high K i (240 Ϯ 60 M) compared with 2Ј-deoxy-3Ј-AMP (5 M, data not shown), we have utilized 2Ј-deoxyadenosine for this work because of the availability of the compound in radiolabeled form. Adenylyl cyclases (native enzymes or the soluble system utilized here) display noncompetitive or mixed noncompetitive inhibition by P-site analogs with respect to MnATP (3,4,10,11,15,23,24). We have focused our analysis on G s␣ -activated adenylyl cyclase activity with MgATP as substrate (in the absence of Mn 2ϩ ) to avoid the complications of mixed patterns of inhibition and the use of a second divalent cation (particularly Mn 2ϩ in the presence of PP i , which has limited solubility).
Equilibrium Dialysis with 2Ј-[ 3 H]Deoxyadenosine-Uncompetitive inhibition implies that 2Ј-deoxyadenosine and ATP do not combine with the same form of the enzyme; noncompetitive inhibition is consistent with such a mechanism (but does not demand it). Florio (25) suggested that P-site ligands were deadend inhibitors that formed a complex with the PP i -bound form of the enzyme (see Fig. 5). Alternatively, Johnson and Shoshani (22) suggested that the P site is distinct from the active site and that both inhibitor and substrate could be bound simultaneously. To distinguish between these possibilities, we utilized equilibrium dialysis to examine directly the requirements for binding of 2Ј-deoxyadenosine to adenylyl cyclase. This was not possible previously because of limiting quantities of protein.
We were unable to detect binding of 80 M 2Ј-deoxyadenosine to 18 M VC 1 and IIC 2 in the presence of G s␣ and Mg 2ϩ or Mn 2ϩ ( Fig. 2A); similar results were obtained after addition of the substrate analog Ap(CH 2 )pp. However, a modest level of binding was detected upon addition of 5 mM ATP under conditions (24-h incubation and very high enzyme concentration) where ATP and the products of the adenylyl cyclase reaction, cyclic AMP and PP i , should be in equilibrium. (The equilibrium constant, 0.065 M (26), implies final concentrations of roughly 0.33 mM ATP and 4.67 mM cyclic AMP and PP i .) We thus tested the capacity of cyclic AMP and PP i to support binding of 2Ј-deoxyadenosine. Binding of the P-site inhibitor was readily observed in the presence of PP i ; this was not true in the case of cyclic AMP. Binding of 2Ј-deoxyadenosine required both VC 1 and IIC 2 and was not observed with the individual proteins in the presence or absence of G s␣ and/or pyrophosphate ( Fig. 2A and data not shown).
Analysis of 2Ј-[ 3 H]deoxyadenosine binding to adenylyl cyclase in the presence of activated G s␣ , Mg 2ϩ , and 2.5 mM PP i revealed a single binding site per C 1 /C 2 heterodimer with a K d of 40 Ϯ 3 M (Fig. 3). Both cyclic AMP and Ap(CH 2 )pp inhibited the binding of 2Ј-deoxyadenosine, implying that these molecules compete with P-site inhibitors for a single binding site (Fig. 2B). Inhibition by the substrate analog, Ap(CH 2 )pp, renders highly unlikely the possibility of 2Ј-deoxyadenosine binding to a site that is distinct from the catalytic site but that still requires PP i to be manifest. This is the first direct evidence in support of Florio's hypothesis (25) of dead-end inhibition of adenylyl cyclase by P-site agents.
Product Inhibition-Binding of a P-site inhibitor that is observable only in the presence of a reaction product suggests that release of product from the enzyme is at least partially rate-limiting. Patterns of inhibition of enzymatic activity by product are useful in determining if release of product is an ordered or random event. Adenylyl cyclase activity is inhibited by both reaction products, and previous studies with the enzyme from Brevibacterium liquefaciens demonstrated their random release (27). Our enzyme system also displays random release of products, since the kinetics of inhibition of enzymatic activity by both cyclic AMP and PP is mixed; this is the pattern predicted by steady-state models. The intersection points for double-reciprocal plots (1/activity versus 1/substrate concentration) at increasing concentrations of either product, PP i ( Fig.  4A) or cyclic AMP (Fig. 4B), are above the abscissa and to the left of the ordinate. A purely competitive pattern of inhibition is expected for both products only if rapid equilibrium kinetics applies (28). The slopes of Fig. 4B are not a linear function of cyclic AMP concentration and tend to curve upwards at high concentrations of the cyclic nucleotide. This is also indicative of a steady-state system where the assumption of rapid equilibrium does not apply (28). Finally, both cyclic AMP and PP i can compete with Ap(CH 2 )pp for binding to adenylyl cyclase (equilibrium dialysis data not shown); this again indicates that release of product is random.
The equilibrium binding studies and patterns of product inhibition described to this point permit formulation of a model for inhibition of G s␣ -stimulated adenylyl cyclase activity by P-site inhibitors (Fig. 5). We observe binding of 2Ј-deoxyadenosine only in the presence of PP i , and therefore the inhibitor is hypothesized to bind to only a single intermediate along the reaction coordinate. This is consistent with the uncompeti- tive kinetic pattern between P-site inhibitor and substrate. If P-site inhibitors bind to the enzyme in the absence of PP i , the interaction is apparently quite weak compared with that with the enzyme-PP i complex.
Interaction between P-site Inhibitors and Products-Analysis of inhibition of adenylyl cyclase activity by both a P-site inhibitor and PP i or cyclic AMP provides additional kinetic information about the mechanism of inhibition and catalytically important steps. A diagnostic test for the interaction between two inhibitors is provided by Dixon plots of 1/velocity versus the concentration of one inhibitor at a constant concentration of substrate and different fixed concentrations of the other inhibitor (28) (Fig. 6). The family of curves obtained at different 2Ј-deoxyadenosine concentrations with respect to cyclic AMP are parallel, indicating that the actions of these two inhibitors are mutually exclusive (Fig. 6B). The IC 50 for inhibition by cyclic AMP increases as the concentration of 2Ј-deoxyadenosine increases in a manner analogous to the interaction between a competitive inhibitor and substrate. This result is consistent with equilibrium binding data, which suggested a competitive interaction between cyclic AMP and 2Ј-deoxyadenosine.
A Dixon plot of 1/velocity versus the concentration of PP i at different fixed concentrations of 2Ј-deoxyadenosine yields a family of intersecting lines, indicating that the two inhibitors do not bind to adenylyl cyclase in a mutually exclusive fashion (Fig. 6A). The slope of the line for one inhibitor now depends on the concentration of the second. The intersection of these curves above the [PP i ] axis indicates slight synergy between PP i and 2Ј-deoxyadenosine, and such synergy is anticipated from the binding data. However, the degree of synergy is dependent on a number of factors. The complexity of this result will be discussed below.
The Reverse Reaction-The reaction catalyzed by adenylyl cyclase is readily reversible (26,29,30). The equilibrium constant, measured with the enzyme from B. liquefaciens (26), is 0.065 M (pH 7.3, 25°C) and thus actually favors ATP synthesis under standard conditions of 1 M concentrations of reactants and products. The synthesis of ATP from cyclic AMP and PP i is activated by both G s␣ and forskolin, as anticipated (Table I). The velocity of the reverse reaction is approximately 6% of that of the forward reaction assuming infinite substrate concentrations (Table II); a similar value was obtained with the pyruvate-stimulated bacterial enzyme (30).
We have measured the rate of ATP synthesis at varying concentrations of the two substrates, cyclic AMP and PP i . The  7B) (28). The K d for binding of cyclic AMP to adenylyl cyclase, K cAMP , is 2.3 Ϯ 0.8 mM, while that for PP i , K PP i , is 0.12 Ϯ 0.04 mM (see Table II). Secondary plots of the apparent 1/V max versus 1/[cyclic AMP] (from Fig. 7A) or 1/[PP i ] (from Fig. 7B) provide the V max of the system at infinite substrate concentrations (3.9 Ϯ 0.3 mol/min-mg [1.8 Ϯ 0.1 s Ϫ1 ]) and the apparent dissociation constants for binding of the second substrate (␣K cAMP ϭ 12 Ϯ 3 mM, ␣K PP i ϭ 0.7 Ϯ 0.3 mM). If steady-state kinetics apply, ␣K cAMP and ␣K PP i cannot be assumed to represent binding constants but instead represent the K m for cyclic AMP and PP i , respectively (28).
P-site Inhibition of the Reverse Reaction-Lineweaver-Burk analysis of the effect of three fixed concentrations of 2Ј-deoxyadenosine on the velocity of the reverse reaction, measured at variable concentrations of PP i , reveal parallel lines indicative of uncompetitive inhibition. Thus, 2Ј-deoxyadenosine and PP i combine with different forms of the enzyme: E-PP i versus E-cAMP or E alone. Saturation with PP i cannot overcome P-site inhibition; rather, it enhances it.
Florio's (25) dead-end model of P-site inhibition clearly predicts a competitive interaction between 2Ј-deoxyadenosine and cyclic AMP with regard to the kinetics of the reverse reaction. This is shown clearly in Fig. 8B. High, saturating concentra- FIG. 5. Model for the mechanism of P-site inhibition. Odd-and even-numbered rate constants describe the forward and reverse reactions, respectively. Estimates of the values of these rate constants are provided in Table II. The species I represents the P-site inhibitor, 2Ј-deoxyadenosine. K i represents the dissociation constant for binding of 2Ј-deoxyadenosine to the E-PP i complex. tions of PP i drive the reaction toward the formation of E-PP i , for which cyclic AMP and P-site inhibitors compete. The apparent K i for 2-deoxyadenosine is 180 Ϯ 30 M, compared with a K d of 40 M measured by equilibrium dialysis for binding of the inhibitor to adenylyl cyclase in the presence of saturating concentrations of PP i . To obtain a true K i for 2Ј-deoxyadenosine, assays must be performed at different concentrations of PP i . Experiments of this sort indicate an approximate K i of 160 M (data not shown). The difference between the K d and the K i may be due to the difference in the temperature required for these measurements (binding at 4°C; kinetics at 30°C).
Calculation of the Steady-state Equation for P-site Inhibition-Since assumption of rapid equilibrium appeared to be invalid, at least in terms of analysis of product inhibition, we derived the rate equation for the model shown in Fig. 5 under steady-state conditions (see "Appendix"). We have made few assumptions in the derivation of the rate equation except for the exclusion of Mg 2ϩ as a participant in the reaction. This equation and the rate constants shown in the legend of Table  II, estimated by simultaneous fit of all of the data shown in Figs. 1, 4, and 6 -8, permit simulation of all of the kinetic patterns described herein. A comparison of the K i values, K m values, and velocities obtained experimentally versus those predicted by the given set of rate constants is shown in Table II (and see Fig. 10). We do not pretend that the rate constants listed in Table II represent exact values, but they do describe the kinetic patterns observed and, we believe, emphasize the validity of the model. The only points at which the simulated lines diverge from the data are in the degree of synergy displayed in Dixon plots of PP i versus 2Ј-deoxyadenosine and the degree of inhibition caused by cyclic AMP (Table II and see Fig. 10). Notable is the estimate that k 5 and k 9 are small compared with k 7 and k 11 . Thus, in the forward reaction the release of cyclic AMP occurs preferentially, prior to the release of PP i from the enzyme; in the reverse reaction the enzyme prefers to bind PP i before cyclic AMP. In addition, the slowest steps of the forward reaction appear to be k 3 and k 11 , the rate of cyclization of ATP and the release of PP i , respectively. These rate constants dictate the degree of P-site inhibition because of their control of the steady-state concentration of the enzyme-PP i complex.
We have also solved the rate equations for a number of different models for comparison. None predicts the various inhibition patterns to the same extent as does the model shown in Fig. 5 as determined by graphical evaluation of fits to the data and analysis of statistical parameters. In particular, an ordered product release mechanism, in which the release of cyclic AMP must precede the release of PP i , fails to predict inhibition by cyclic AMP in the forward direction (Figs. 4B and 5B). Models that permit binding of P-site analogs to both the enzyme-PP i complex and the enzyme alone do not display uncompetitive inhibition by 2Ј-deoxyadenosine (Fig. 1), even when the relative affinity of the P-site inhibitor for the enzyme-PP i complex exceeds that for the free enzyme by 20-fold.
Activation of G s␣ -stimulated Adenylyl Cyclase by Forskolin-Inclusion of forskolin in reaction mixtures increases the V max of both the forward and reverse reactions and lowers the apparent K i for 2Ј-deoxyadenosine (Table III). However, the patterns of inhibition by the P-site analog remain unchanged (not shown). The K d for binding of 2Ј-deoxyadenosine to adenylyl cyclase in the presence of PP i is also lowered compared with the value observed in the presence of G s␣ alone (21 Ϯ 2 M, Fig. 9). Increased activation of adenylyl cyclase by the combination of forskolin and G s␣ thus further stabilizes the enzyme-2Ј-deoxyadenosine-PP i complex. DISCUSSION Adenylyl cyclase contains a single ATP binding site that may lie at the interface between the C 1 and C 2 domains of the protein (19,31). Potent P-site inhibitors have no effect on the binding of the substrate analog Ap(CH 2 )pp to this site (19). This result is reflected in the uncompetitive nature of P-site inhibition with respect to MgATP. A classical uncompetitive inhibitor does not bind to the free enzyme; rather, the inhibitor binds to the enzyme only after substrate is bound.
Several mechanisms for P-site inhibition have been suggested. The first involves formation of a dead-end complex, initially proposed by Wolin (27) to describe inhibition of adenylyl cyclase from B. liquefaciens by adenosine. It was suggested that uncompetitive inhibition by adenosine reflected the formation of an enzyme-ATP-adenosine complex at the catalytic site  Figs. 1, 4, 6, 7, and 8 to the steady-state rate equation shown in the "Appendix" to obtain estimates of K i and the rate constants k 1 through k 12 . These constants were used to plot simulations of the data, which are shown in Fig. 10, and these lines were used to calculate the values shown in the columns labeled "Simulated fit." The rate constants used to simulate experimental data are: k 1 , 2.62 ϫ 10 5 M/s; k 2 , 89.5 s Ϫ1 ; k 3 , 59 s Ϫ1 ; k 4 , 2.6 s Ϫ1 ; k 5 (Fig. 8A) a 0.39 Ϯ 0.09 mM 0.37 mM V max (Figs. 1 and 4) 30 (Fig. 1) 240 Ϯ 60 M 256 M K m(cAMP) (Fig. 8B) a 8.7 Ϯ 0.4 mM 8.8 mM K i(PPi ) (Fig. 4A) 0.31 Ϯ 0.02 mM 0.16 mM V max (Fig. 8B) (Fig. 7) 0.12 Ϯ 0.04 mM 0.16 mM K cAMP (Fig. 7 that was not in rapid equilibrium with active forms of the enzyme. Florio (25) and Florio and Ross (7) suggested a related mechanism for mammalian adenylyl cyclase, a dead-end complex containing enzyme, PP i , and P-site inhibitor. An alternative mechanism was later proposed by Johnson and Shoshani (22), in which P-site inhibitor and substrate bind simultaneously to distinct sites on the enzyme. This type of inhibition, an allosteric mechanism (7), was particularly appealing because of the homologous nature of the two cytoplasmic domains, each possibly containing a nucleotide binding site.
To discern the reaction intermediate(s) involved in P-site interactions, we have used equilibrium dialysis to detect requirements for binding of 2Ј-deoxyadenosine. Interactions were not detected with enzyme alone, nor in the presence of Ap(CH 2 )pp or cyclic AMP. By contrast, binding was readily detected in the presence of the product PP i ; one molecule of 2Ј-deoxyadenosine bound per C 1 /C 2 heterodimer with a K d of 40 Ϯ 3 M. Cyclic AMP or Ap(CH 2 )pp reduced or abolished binding. These observations provide clear evidence to support the hypothesis that P-site analogs act as dead-end inhibitors of product release.
Analysis by us and others of both mammalian and bacterial adenylyl cyclases indicates that the release of product is random (22,27). This fact and the binding data permit formulation of the reaction scheme shown in Fig. 5. Our remaining kinetic data support this mechanism and provide additional information about product release. Dixon plots describing the inhibition of the forward reaction by both 2Ј-deoxyadenosine and products show parallel or intersecting lines for cyclic AMP or PP i , respectively. Parallel lines indicate that cyclic AMP and P-site inhibitors act in a mutually exclusive fashion, each inhibitor reducing the effectiveness of the other. This is certainly consistent with the notion that both molecules bind to the same site.
The Dixon plot that describes the interaction between 2Јdeoxyadenosine and PP i (lines intersecting above the abscissa; Fig. 6A) indicates that both molecules are bound to the enzyme simultaneously and display modest synergy as inhibitors. This result was demonstrated previously by Florio with forskolinactivated adenylyl cyclase activity in cyc Ϫ S49 cell membranes and was the basis for his dead-end inhibitor hypothesis (25). (We have also observed a synergistic interaction between 2Јdeoxyadenosine and PP i with our forskolin-activated enzyme; data not shown.) However, Johnson and Shoshani failed to observe any synergy between these inhibitors using a proteolyzed rat brain adenylyl cyclase preparation activated with GTP␥S-G s␣ (22). This result shaped their model of an allosteric mechanism. Modelling experiments using the steady-state rate equations for dead-end inhibition (see "Appendix") indicate that vastly different degrees of synergy can be observed, depending on the individual rate constants. The dissociation and association rate constants for ATP, PP i , and cyclic AMP all contribute to the degree of synergy observed with the combination of 2Ј-deoxyadenosine and PP i acting as inhibitors of the forward reaction. Thus, small differences in rates caused by variation in adenylyl cyclase isoform, species, or mode of activation may lead to different degrees of apparent synergy between P-site inhibitors and PP i . Measurement of ATP synthesis by adenylyl cyclase provides a substantial amount of information about the binding of cyclic AMP, PP i , and 2Ј-deoxyadenosine. Inhibition of the reverse reaction by 2Ј-deoxyadenosine is uncompetitive with respect to PP i , indicating that binding of inhibitor occurs after interaction of PP i with the enzyme. Inhibition by 2Ј-deoxyadenosine is competitive with respect to cyclic AMP in the presence of saturating concentrations of PP i , which drives the formation of E-PP i ; both cyclic AMP and 2Ј-deoxyadenosine compete for this complex. Thus, the binding and kinetic data for the reverse reaction indicate that a single inhibitor molecule binds after the addition of PP i , and the inhibitor is competitive with cyclic AMP at the catalytic site of the enzyme.
The model shown in Fig. 5 is consistent not only with the kinetic data shown above but also with classical features of P-site inhibition. These inhibitors display a strict requirement for an adenine ring (2,5,12), consistent with binding at the catalytic site. Crystallographic data and evidence from mutagenesis suggest that the active site lies at the interface of the C 1 and C 2 domains (19,31). A mutation that dramatically lowers the IC 50 for P-site inhibition (K923A in type I adenylyl cyclase) is also located at this interface (31,32). P-site inhibitors containing a 3Ј-ribose phosphate (e.g. 2Ј-deoxy-3Ј-AMP) have a greatly increased potency (12), and this suggests that the enzyme-PP i -inhibitor complex might resemble a structure close to a product-like transition state, accommodating both PP i and a nucleotide containing a 3Ј-phosphoryl substituent. Furthermore, the catalytic site must be capable of accommodating several phosphates and large substitutions at the 3Јribose position, since 3Ј-polyphosphates are more potent than other P-site inhibitors, and other modifications at this position are easily tolerated (12,13). Isoform-dependent sensitivity of adenylyl cyclases to various adenosine analogs may reflect differences in the capacities of their active sites to accommodate the 3Ј-moieties of these analogs (33).
Activated forms of adenylyl cyclase are more sensitive to P-site inhibition than are nonstimulated forms of the enzyme (3-5, 7, 11). This result is explained by a dead-end inhibitor model. To observe P-site inhibition of the forward reaction, release of PP i from the enzyme must be at least partially rate-limiting, otherwise the enzyme-PP i complex would never be present at sufficient concentrations to bind inhibitor. Note that k 3 and k 11 in our model are approximately equal (Table II). Any alteration of the enzyme that slows reaction steps prior to the release of product, such that product release is no longer rate-limiting, should decrease the potency of P-site inhibitors. This is the equivalent of saying that a higher concentration of a P-site inhibitor will be required to trap the E-PP i complex if it is present in lower steady-state concentrations. For example, if, under basal conditions, the rate of synthesis of cyclic AMP and PP i is slow relative to the rate of release of these products, enzyme-PP i will not accumulate and P-site inhibitors will be impotent. Activation of adenylyl cyclase by forskolin or G s␣ may increase the rate of synthesis of cyclic AMP and PP i (k 3 ), leading to higher steady-state concentrations of enzyme-PP i and thus increased potency of P-site inhibitors.
Alternatively, if the enzyme-PP i -P-site inhibitor complex mimics a transition state and activators stabilize the transition state of the enzyme, they will increase the affinity of adenylyl cyclase for P-site inhibitors. Binding and kinetic data for G s␣and forskolin-activated adenylyl cyclase (compared with the G s␣ -activated enzyme) suggest that both mechanisms serve to increase the potency of P-site inhibitors. Thus, addition of forskolin to G s␣ -activated adenylyl cyclase causes a 3-4-fold increase in V max of the forward and reverse reactions. The apparent K i for 2Ј-deoxyadenosine is decreased (4-fold) as expected. In addition, the K d for binding of 2Ј-deoxyadenosine is also decreased (2 fold) in the presence of both activators (Fig.  Fig. 7B). b V max and the apparent K i for 2Ј-deoxyadenosine measured with respect to MgPP i (see Fig. 7A). 9), suggesting that activators stabilize the enzyme-PP i -P-site inhibitor complex. If this complex represents a product-like transition state, the binding data imply that G s␣ and forskolin increase the catalytic activity of adenylyl cyclase by stabilizing the transition state of the enzyme.
A related feature of P-site inhibition is the decreased potency of these compounds when analogs of ATP are used as substrates (22). These analogs have either a decreased affinity for the enzyme or they impair the rate of cyclization. Thus, the release of product is no longer rate-limiting compared with the other steps of the reaction, and the potency of P-site inhibitors decreases. The large number of P-site phenotypes obtained as a result of mutagenesis (32) can also be explained by any decrease in k cat (or increase in K d for ATP) such that the ratelimiting step occurs prior to release of product.