|Home | About | Journals | Submit | Contact Us | Français|
Type V and VI mammalian adenylyl cyclases (AC5, AC6) are inhibited by Ca2+ at both sub- and supra-micromolar concentration. This inhibition may provide feedback in situations where cAMP promotes opening of Ca2+ channels, allowing fine control of cardiac contraction and rhythmicity in cardiac tissue where AC5 and AC6 predominate. Ca2+ inhibits the soluble AC core composed of the C1 domain of AC5 (VC1) and the C2 domain of AC2 (IIC2). As observed for holo-AC5, inhibition is biphasic, showing “high affinity” (Ki = ~ 0.4 μM) and “low-affinity” (Ki = ~100 μM) modes of inhibition. At micromolar concentration, Ca2+ inhibition is non-exclusive with respect to pyrophosphate (PPi), a non-competitive inhibitor with respect to ATP, but at > 100 μM Ca2+, inhibition appears to be exclusive with respect to PPi. The 3.0Å-resolution structure of Gαs•GTPγS/forskolin-activated VC1:IIC2 crystals soaked in the presence of ATPαS and 8 μM free Ca2+ contains a single, loosely coordinated metal ion. ATP soaked into VC1:IIC2 crystals in the presence of 1.5 mM Ca2+ is not cyclized, and two calcium ions are observed in the 2.9Å resolution structure of the complex. In both of the latter complexes VC1:IIC2 adopts the “open”, catalytically inactive conformation characteristic of the apoenzyme, in contrast to the “closed”, active conformation seen in the presence of ATP analogs and Mg2+ or Mn2+. Structures of the pyrophosphate (PPi) complex with 10 mM Mg2+ (2.8Å) or 2 mM Ca2+ (2.7Å) also adopt the open conformation, indicating that the closed to open transition occurs after cAMP release. In the latter complexes, Ca2+ and Mg2+ bind only to the high-affinity “B” metal site associated with substrate/product stabilization. Ca2+ thus stabilizes the inactive conformation in both ATP- and PPi-bound states.
The nine membrane-bound mammalian adenylyl cyclase (AC)1 isoforms and a soluble AC (sAC) catalyze the conversion of ATP to the intracellular second-messenger cAMP and pyrophosphate (PPi) (1, 2). Biochemical and crystallographic studies show that this enzymatic process requires Mg2+ or Mn2+ as co-factors (3-5). Membrane-bound ACs are activated by the stimulatory G protein α subunit (Gαs) and other regulatory molecules (1, 6, 7). The catalytic activity of several ACs is profoundly influenced by submicromolar Ca2+ (8, 9). At resting physiological concentration of ~100 nM (10), calcium ions stimulate Type I and Type VIII isoforms of AC but inhibit Types V and VI. These isoforms, which are differentially expressed in tissues, integrate the effects of calcium- and cAMP-mediated signaling to control diverse cellular functions (11-13). Calmodulin mediates the calcium stimulation of both Type I and Type VIII AC and could contribute to mechanisms of learning and memory (14-17). Type III AC is regulated by calmodulin and CAM Kinase II (18, 19), and Type IV is subject to calcium inhibition through the action of calcineurin (20) The two calcium-inhibited ACs (Type V and VI) are highly expressed in cardiovascular tissue and have been proposed to be key negative feedback regulators of cardiac rhythmicity (21-23). At high (>10 μM) concentration Ca2+inhibits all mammalian AC isoforms (9, 11, 24, 25).
The catalytic site of AC, as for all Class III cyclases, resides at the interface of two homologous Cyclase Homology Domains (CHD) (26). The model for the prototypical AC catalytic core is derived from crystal structures of the complex formed by the C1 CHD domain from AC5 and the C2 CHD domain from AC2 (VC1:IIC2) bound to two activators, forskolin (FSK) and GTPγS-activated Gαs (27). Structural studies of activated VC1:IIC2 complexes bound to substrate analogs show that the active site of AC contains two Mg2+ (or Mn2+) sites, designated “A” and “B”, that must be occupied for catalytic activity (5). Upon binding to potent substrate analogs, AC undergoes a transition from an “open” to a “closed” conformation by moving several structural elements in both CHDs toward the catalytic site. In the catalytically active conformation, both divalent metal ions coordinate with the substrate and with two conserved aspartic acid residues, Asp-396 and Asp-440 (residue numbers refer to AC5) in the C1 domain. The metal ion in site “A” is thought to activate the 3′-ribosyl hydroxyl group for nucleophilic attack on the oxygen of the α-phosphate whereas the metal ion in site “B” coordinates with β- and γ-phosphates of ATP (4, 5). The ATP phosphates are themselves recognized by a phosphate binding loop at the junction of the C1 domain β1-α1 element. Kinetic analysis shows that cAMP is released from the enzyme more rapidly than the second product, pyrophosphate (PPi) (28). The enzyme adopts the open state upon product release, but it is not known whether the transition to this state occurs upon cAMP formation, or release of cAMP or PPi from the enzyme.
Previous studies have shown that Ca2+ inhibits AC5 and AC6, by a mechanism that is largely non-competitive with respect to ATP (9). Calcium ion antagonizes the activation of these enzymes by Mg2+, and this inhibition is biphasic with respect to Ca2+ concentration (9, 25, 29). Kinetic analysis of AC5 shows that, at submicromolar concentration, Ca2+ is a non-competitive inhibitor of Mg2+ activation, and, at supramicromolar concentration, Ca2+ inhibition is directly competitive with Mg2+ (25). The interdependence of Mg2+ activation and Ca2+ inhibition in site-directed mutants of AC5 suggests that inhibition of AC by Ca2+ involves the Mg2+-binding loci at the AC catalytic site. Indeed, mutations distal to the Mg2+ coordination site that affected Mg2+ binding also displayed impaired or diminished Ca2+-inhibition in direct proportion to diminution of activation by Mg2+ (25). That high-affinity Ca2+ binding is non-competitive with respect to Mg2+ is suggestive of an allosteric mode of action, wherein Ca2+ modulates the conformational equilibrium of the enzyme. However, the mechanisms by which Ca2+ regulates AC remain to be understood at the molecular level.
Here, we present structural data that provides insight into the molecular mechanism by which Ca2+ inhibits AC5. Ideally, such an investigation would focus on the AC5 holoenzyme, or its catalytic domain, which is subject to the same mechanism of Ca2+ inhibition (25). However, biochemical studies have demonstrated that the catalytic and many of the regulatory properties of holo-AC are faithfully recapitulated by the Gαs-activated soluble VC1:IIC2 catalytic core of AC, which is amenable to crystallization (30, 31). Experiments conducted with chimeric AC molecules demonstrate that isoform-specific high-affinity inhibition by Ca2+ is conferred by the C1 domain, but not the C2 domain of AC. Chimeric AC molecules composed of the Transmembrane 1 (TM1) and C1 domains of AC5 and the TM2 and C2 domains of AC2 retain susceptibility to inhibition submicromolar Ca2+. In contrast, the corresponding constructs in which TM1/C1 is derived from AC2 and TM2/C2 from AC5, are not inhibited. High-affinity Ca2+ inhibition was also observed in experiments in which half molecules composed of the TM and C1a (CHD only) domain of AC5 and the TM and C2 domain of AC2 were co-expressed (25). These experiments, together with evidence that Ca2+ acts at the Mg2+ cofactor binding sites, have prompted us to use Gαs•GTPγS:VC1:IIC2 as a model system to investigate the structural mechanism for high-affinity inhibition of AC5 by Ca2+.
Here, we present kinetic evidence to demonstrate that, in the presence of Gαs•GTPγS and FSK, Ca2+ shows the same pattern of high- and low-affinity inhibition towards VC1:IIC2 that has been observed for holo AC5. We have determined a series of crystal structures of Gαs•GTPγS and FSK-activated VC1:IIC2 in the presence of Ca2+ that provide direct insight into the mechanism of Ca2+ inhibition. These structures were determined, 1) in the presence of non-reactive ATP analog and free Ca2+ at a concentration below the EC50 for high affinity inhibition; 2) in the presence of ATP and free Ca2+ in the millimolar range sufficient to saturate both high and low affinity sites and 3) in the presence of pyrophosphate and saturating Ca2+. In conjunction with the latter, we have also determined the structure of the Gαs•GTPγS and FSK-activated AC catalytic core bound to the cofactor Mg2+ and pyrophosphate, the binary product complex. Together, these structures elucidate the mode of Ca2+ binding in the high- and low-concentration regimes and show that Ca2+ stabilizes the inactive (open) state of the enzyme in both ATP and PPi-bound states. Further, the structure of the PPi-bound AC catalytic core provides insight into the nature of conformational changes that accompany the product-release phase of the AC catalytic cycle.
Recombinant mammalian adenylyl cyclase cytosolic VC1 and IIC2 domains and bovine Gαs proteins were expressed and purified as described (32). Before complex formation, the Gαs protein was activated by GTPγS and then trypsin-digested. A mixture of individually purified recombinant VC1, IIC2 and trypsin—treated Gαs•GTPγS was passed through sizing columns in the presence of excess MP-FSK and GTPγS. Fractions containing the heterotrimeric complex were identified by gel electrophoresis on a 4-20% Tris-HCl polyacrylamide gel (BIO-RAD, Hercules, CA). Fractions containing the complex were collected and concentrated to ~8 mg/ml in a buffer containing 20 mM Na+HEPES (pH 8.0), 2 mM EDTA, 2 mM MgCl2, 2 mM DTT, 100 mM NaCl, 25 μM MP-FSK, and 10 μM GTPγS for crystallization.
Crystals of the protein complex were grown, harvested and cryoprotected as described previously (32, 33). Before cryoprotection, crystals were soaked in one of following reservoir solutions: 2 mM ATPαS and 50 μM CaCl2 (for which the free [Ca2+] was found to be 8 μM using the dye calibration method of Linse (34)); 5 mM ATP and 1.5 mM CaCl2; 3 mM PPi and 10 mM MgCl2; or 3 mM PPi and 2 mM CaCl2 for 1-2 hr at room temperature and then harvested in cryoprotectant solution containing the same concentrations of the respective ligands and metal ions. For crystallization of mAC at low [Ca2+], crystals were soaked in reservoir solution containing 2mM ATPαS and 50 μM CaCl2 ,which were likewise present at the same concentrations in cryoprotectant solutions. Diffraction data sets were collected by the oscillation method at the Advanced Photon Synchrotron SBC-CAT ID-19 beamline or the Stanford Synchrotron Radiation Laboratory 9-1 beamline with an incident beam wavelength of 1.0454Å or 1.0231Å, respectively. Image processing and data reduction utilized the HKL2000 package (35). Due to anisotropy, data with l index > 20 for crystals of the Gαs•GTPγS and FSK-activated VC1:IIC2 complexes with PPi•Mg2+, PPi•Ca2+, ATPαS•Ca2+, or ATP•2Ca2+ were excluded from the dataset used for refinement. The atomic coordinates from the isomorphous crystals of the MP-FSK,Gαs•GTPγS:VC1:IIC2 complex (PDB code: 1AZS), were used to compute initial phases for all four complexes (27). Atomic positions and thermal parameters were refined by successive rounds of rigid body refinement, simulated annealing, Powell minimization, and B factor refinement using the CNS 1.1 program suite (36) or REFMAC5 (37) as implemented in the CCP4 program Suite (38). Ligands and metal ion(s) were located in SIGMA—A weighted (39) difference omit maps computed with phases from refined models. Atomic models were iteratively improved by manual refitting into weighted 2|Fo|—|Fc| maps using the computer graphics model building programs O and Coot (40, 41), and subsequent cycles of refinement using CNS or REFMAC5. Diffraction and refinement statistics are given in Table 1. Figures were generated using PyMOL (DeLano Scientific LLC, San Carlos, CA. http://www.pymol.org). Coordinates for the AC•PPi•Ca2+, AC•PPi•Mg2+, AC•ATP•2Ca2+ and AC•ATPαS•Ca2+ complexes have been deposited in the Protein Data Bank with the codes 3C14, 3C15, 3C16, and 3E8A, respectively.
Adenylyl cyclase activity was determined as described previously (42-44) with some modifications. Briefly, for meaurement of Ca2+ inhibition of Mg2+, FSK, Gαs•GTPγS activated VC1:IIC2, adenylyl cyclase activity of purified VC1 (50 nM) and IIC2 (250 nM) was measured in the presence of the following components: 12 mM phosphocreatine, 2.5 units of creatine phosphokinase, 0.1 mM cAMP, 0.1 mM ATP, 0.04 mM GTP, 0.5 mM phosphodiesterase inhibitor, isobutylmethylxanthine, 1.25 μCi of [α-32P]ATP, 0.75 mM MgCl2, 10 μM FSK or 500nM Gαs•GTPγS, as indicated. Free Ca2+ concentrations were established using an EGTA-buffering system as described previously (45) and the BAD4 computer program (24). The reaction mixture (final volume, 100 μl) was incubated at 30 °C for 20 min. Reactions were terminated with sodium lauryl sulfate (0.5%); [8-3H]cAMP (~6000 cpm) was added as a recovery marker, and the [32P]cAMP formed was quantitated as described previously (46).
For PPi or Ca2+/PPi inhibition experiments, the activity of purified VC1 (100 nM) and IIC2 (500 nM) was measured in a buffer of 25mM Na+Hepes (pH 8.0) containing 20mM phosphor(enol)pyruvate, 0.1 mM GTP, 3 units pyruvate kinase, 0.1mM ATP, 0.1mM cAMP, 0.3 mM MgCl2, 1 μCi of [α-32P]ATP, 1 μM Gαs•GTPγS, and 100 μM FSK, unless stated otherwise. Free Ca2+ concentrations were established and calibrated using an EGTA-buffering system as described previously (34). Following a 2 min pre-incubation at 37°C, reactions were conducted for 15 min at 37°C and were terminated by the addition of 20 μL of 2.2 N HCl containing [3H]cAMP. Denatured protein was additionally heated to 95°C for 4 min, cooled on ice, and finally sedimented by a 1 min centrifugation at 15,000 × g. Reaction mixtures were applied onto disposable columns filled with 1.3 g neutral alumina. cAMP was separated from [α-32P]ATP by elution of cAMP with 0.1 M ammonium acetate, pH 7.0. [32P]cAMP and [3H]cAMP was measured by liquid scintillation spectrometry where [3H]cAMP was calculated for the efficiency of cAMP recovery of each tube. Nonlinear regression curves and kinetic parameters were obtained using SigmaPlot software (SYSTAT Software Inc, San Jose, CA). Adenylyl cyclase activity is expressed per weight of VC1 and data points are presented as mean activities ± S.D. of triplicate determinations.
The measurement of ATP synthesis from cAMP and PPi was performed as described previously (28) with some modifications. The reverse reaction of adenylyl cyclase was measured spectrophotometrically in the presence of the following components: 50 mM glucose, 20 mM Hepes, 1.2 units of hexokinase, 0.8 mM NADP, 2 mM MgCl2, 2 mM PPi, 0.5 units of glucose-6-phosphate dehydrogenase, and the various concentrations of cAMP and CaCl2. The reaction was started by the addition of 0.4 μM VC1, 2 μM IIC2, and 1 μM Gαs•GTPγS (final volume 400 μl, pH 7.4), incubated at 30 °C for 40 min. ATP synthesized was determined by the rate of NADP reduction as measured by the change in absorbance at 340 nm using an a Ultrospec 2100 Pro UV/Visible spectrophotometer. A standard curve was obtained by measuring the rate of NADP reduction at a range of ATP concentration in the absence of protein, cAMP, and PPi. ATP synthesis activities were expressed per weight of VC1 and fitted with Lineweaver-Burk analysis using GraphPad Prism Version 4. Data points were presented as mean activities ± S.D. of triplicate determinations.
GTPγS was obtained from Roche (Indianapolis, IN). FSK and MP-FSK were obtained from Merck (Nottingham, UK) or Calbiochem (La Jolla, CA). [α—32P]ATP (3,000 Ci/mmol) was from Perkin—Elmer Life Sciences (Boston, MA) or GE Healthcare (Little Chalfont, UK). [8-3H]cAMP were obtained from GE Healthcare (Little Chalfont, UK) and all other reagents were purchased from Sigma (Poole, UK).
As noted in the introduction, Ca2+ was shown to exert biphasic inhibition of AC5 and AC6 in cell lysates (9, 25). In the present work, we extended these studies to investigate Ca2+ inhibition of a purified soluble recombinant complex consisting of VC1 and IIC2. This complex possesses high catalytic activity upon activation by both FSK and Gsα•GTPγS protein (31), and thus experiments were conducted in the presence of 10 μM FSK and 500 nM Gαs•GTPγS. As is the case for holoenzyme preparations, inhibition of FSK and Gsα•GTPγS activated VC1:IIC2 by Ca2+ is biphasic (Figure 1A), with EC50 values of 0.37 ± 0.03 μM and 100 ± 30 μM, respectively. The Ki of the high affinity component is not significantly changed, whereas that of the low affinity inhibition increased two-fold over the range of [Mg2+] from 0.3 mM to 3 mM (data not shown). These results are comparable to data taken from cell lysates expressing AC5 where the range for high and low affinity of Ca2+ inhibition was 0.28 ± 0.21 μM and 32 – 163 μM, respectively (25) indicative of non-competitive and competitive inhibition of Mg2+ activation of AC through the high- and low-affinity binding modes of Ca2+.
AC activity in the direction of cAMP synthesis is inhibited by both reaction products, cAMP and PPi (28). Product inhibition by PPi has been shown to be non-competitive with respect to ATP with an observed Ki of ~ 300 μM. Over a range of PPi concentrations, Ca2+ exhibits biphasic inhibition of VC1:IIC2 in the presence of Gαs•GTPγS and FSK (Figure 1A). A Dixon plot of 1/velocity versus the concentration of Ca2+ in the high affinity range (200 nM – 1.4 μM ) intersect at a common point, consistent with non-exclusive inhibition by Ca2+ and PPi of adenylyl cyclase (47) (Figure 1B, upper panel). The Ki of PPi remains constant over this range of Ca2+, indicating that the two inhibitors do not bind cooperatively. Values of reciprocal velocity at varying [PPi] at [Ca2+] in the low affinity range (30 – 300 μM) give rise to a set of parallel lines, indicating that PPi and Ca2+, acting at a low-affinity site, are exclusive inhibitors (Figure 1B, lower panel). These kinetic data can be rationalized by the crystal structures of Ca2+ complexes with VC1:IIC2 described below. The origin of the sharp rise in AC activity in the 2.5 - 5 μM Ca2+ range, observed in the presence of PPi is not apparent. It is possible that, in this narrow concentration range, Ca2+ potentiation of ATP binding offsets the combined inhibitory affects of Ca2+ and PPi on catalytic activity.
To determine whether Ca2+ is an inhibitor of AC in the direction of ATP synthesis, we measured the rate of ATP production from cAMP and PPi in the presence of Mg2+ at various Ca2+ concentrations. The reverse reaction, like the forward reaction, is also activated by Gsα•GTP and FSK, but with less potency (28). Double reciprocal plot analysis shows that Ca2+ causes a slight increase in the Km for cAMP, and no significant change in Vmax. However, the inhibitory potency of Ca2+ reaches a plateau in the 2 μM to 5 μM range. Thus, Ca2+ appears to be a competitive partial inhibitor with respect to cAMP in the direction of ATP synthesis (Figure 2) with an inhibition constant similar to that for “high affinity”inhibition of cAMP synthesis.
We determined the structure of the Gαs•GTPγS/FSK-bound VC1:IIC2•ATPαS-Rp complex from crystals soaked in reservoir solution containing 8 μM free Ca2+, well below the Ki for low-affinity Ca2+ inhibition. The isosteric and non-reactive ATP analog ATPαS-Rp was used in this instance because VC1:IIC2 retains appreciable catalytic activity in the presence of micromolar Ca2+. The structure of the VC1:IIC2•ATP complex was also determined from crystals containing 1.5 mM Ca2+, sufficient to achieve maximal inhibition. The structures of the two complexes were determined at 3.0Å (ATPαS-Rp) and 2.9Å (ATP) resolution (Table 1), and are isomorphous with previously reported Gαs•GTPγS/FSK-bound VC1:IIC2 structures (27). Both the Ca2+-bound ATP and ATPαS-Rp complexes adopt the “open” conformation characteristic of apo-VC1:IIC2 (Figure 3A), in contrast to the “closed” state seen in the complex of ATPαS-Rp with Mg2+ and Mn2+ (5). In the open state, the enzyme is not able to form the full complement of protein-nucleotide interactions that are possible for the closed state. Although VC1:IIC2 adopts an open conformation in both complexes, the mode of nucleotide-Ca2+ binding in the ”low-Ca2+”, ATPαS-Rp complex differs significantly from that in the “high-Ca2+” complex with ATP. The enzyme active site of the latter adopts a more open conformation due to a rotation of the C1 domain α1 helix away from the subunit interface (Figure 3B).
The (|Fo| — | Fc|) difference electron density map, for the ATPαS-Rp complex reveals continuous density for ATPαS-Rp and a single diffuse peak which we attribute to a Ca2+ ion (Figure 4A). This assignment is confirmed by the presence of positive difference density observed upon modeling a water molecule at that site (Figure 4B). Magnesium ion, which could be expected to bind in similar fashion, was not present in the soaking solution. We refer to this structure as the ATPαS•Ca2+ complex.
The nucleotide in the ATPαS•Ca2+ complex adopts a compact, arched conformation. The electron density, together with stereochemical constraints imposed during refinement dictate a gauche conformation for the C(4′)-C(5′) ribosyl bond, and a glycosyl bond angle in the mid-anti range, which together cause the α thiophosphate to project away from the metal center. This mode of ATPαS-Rp binding contrasts with that in the ”closed:” complex with Mg2+ and Mn2+ bound to the metal A and B sites, respectively (5) (Figure 5A,B). The purine ring of ATPαS-Rp in the Ca2+ complex is within hydrogen bonding distance of the carbonyl oxygen atom of Ile-1019, affording interactions that provide specificity for adenine in preference to other nucleotide bases (26, 48, 49). However, as a consequence of the compact configuration of the nucleotide, and the open conformation of VC1:IIC2, the β and γ phosphate moieties cannot form hydrogen bonds with main-chain amides in the β1-α1 loop of VC1 (Figure 5B). The diffuse electron density for these phosphate moieties is indicative of a flexible binding mode.
The single Ca2+ ion observed in the ATPαS•Ca2+ complex occupies a position intermediate between the Mg2+ A and B sites (Figure 5B), although it is coordinated by the β and γ phosphates and both Asp-396 and Asp-440 as is typical for B-site metal interactions (5). As noted above, the α thiophosphate is not within coordination distance of the calcium ion. Electron density extending from the ribosyl group to the Ca2+can be modeled as a water molecule bridging the ribose O(3′) and the calcium ion (Figure 4A). Even at the modest resolution of the structure, it is apparent that Ca2+ - oxygen ligand distances are longer than typically observed in protein-Ca2+ complexes, indicating that the metal is not rigidly coordinated (50).
The crystal structure of Gαs•GTPγS/FSK-activated VC1:IIC2 bound to ATP in the presence of millimolar Ca2+ exhibits two Ca2+ ions at the metal-binding site (Figure 4C). An |Fo| — | Fc| electron density map computed with refined coordinates of the protein and bound nucleotide, show difference density at both Ca2+ sites, as well as weaker density that we attribute to ordered water molecules (Figure 4D). Accordingly, we refer to this structure as the ATP•2Ca2+ complex. The metal ions are separated by only 3Å, well under the separation distance expected for adjacent calcium ions. It is possible that the observed distance is an artifact of static disorder, as discussed below. The nucleotide triphosphate exhibits an extended conformation similar to that of ATPαS-Rp in the closed complex with Mg2+ and Mn2+, but is translated about 1Å towards the β1-α1 relative to the latter (Figure 6A,C). This relative translation may be due to the expansion of the metal coordination sphere required to accommodate two calcium ions, which have larger radii (1.14Å) than Mg2+ (0.86 Å) and Mn2+ (0.89Å). The expansion of the coordination sphere forces the metal ligands in the β1-α1 loop (carboxylate of Asp 396 and carbonyl oxygen of Ile 397) away from the catalytic site, with the collateral loss of interactions between the ATP β and γ phosphates and the main chain amides of Gly-399 and Phe-400 observed in the closed conformation of VC1:IIC2. Consequently, the CHD domains of ATP•2Ca2+ adopt a somewhat more open conformation than that of ATPαS•Ca2+, such that α1 helix is rotated about 10° further away from the C1:C2 interface (Figure 3B). Likewise, the adenosine moiety is not close enough to Asp-1018 and Ile-1019 to form hydrogen bonds or van der Waals contacts with these side chains in the base recognition pocket (Figure 6A,C). Further, the more open conformation at the C1:C2 domain interface does not allow for an ion pair interaction between Arg-1029 of the C2 domain with the α-phosphate oxygen atoms of ATP as observed in the ATPαS•Mn2+•Mg2+ complex. This interaction has a key role in transition state stabilization (27, 51, 52).
Water-mediated interactions contribute to ATP binding in the open complex with Ca2+. In the closed state, Lys-1065 of the C2 domain β7′–β8′ hairpin loop forms ionic contacts with α- and γ-phosphates of the nucleotide. In the ATP•2Ca2+ complex, this ionic interaction is mediated by a water molecule (Figure 4C,D and and5C).5C). A network of water molecules tethers ATP to Ca2+ and protein. The 3′-hydroxyl of the ribose forms water-mediated contacts with the A site Ca2+. The triphosphate of ATP is also stabilized by a water molecule that coordinates with two oxygen atoms of the β- and γ-phosphates and two backbone amine groups of Gly-399 and Phe-400 (Figure (Figure4D,4D, ,5C)5C) at the N-terminus of the C1 domain α1 helix. These hydrogen bonds match those interactions between the β1-α1 loop and triphosphates of ATPαS in the closed state. Thus, ATP•2Ca2+ is accommodated in the catalytic site despite the loss of direct interactions with C2 domain while retaining contacts with the C1 domain. These observations are consistent with the evidence that the C1 domain mediates the effect of Ca2+ on the catalytic activity of AC (25).
The triphosphate moiety of ATP is fully engaged in coordination of the two Ca2+ ions in the ATP•2Ca complex. The resolution of the structure does not permit accurate determination of Ca2+-ligand coordination distances, yet the probable protein ligands can be identified. The equatorial plane of Ca2+ ligands comprises an α phosphate oxygen, one or both carboxylate oxygen atoms of Asp-396 and a water molecule. On one face of the equatorial plane, an axial ligand is provided by a carboxylate oxygen of Asp-440, while the opposite face is empty, possibly occupied by a disordered water molecule. The A-site Ca2+ is shifted about 2Å closer to the Asp-396 carboxylate than the corresponding Mg2+ of the closed ATPαS complex, and so is able to form more contacts with protein ligands than the latter. The B-site Ca2+ is still more tightly bound with nearly hexadentate geometry. The equatorial plan is populated by one oxygen atom of Asp-440, a β phosphate oxygen, the backbone carbonyl oxygen of Ile-397, and the ATP α-β phosphate bridging oxygen. A carboxylate oxygen of Asp-396 and the ATP γ phosphate contribute axial oxygen ligands. This mode of Ca2+ coordination is nearly identical to that observed for Mn2+ in the closed complex with ATPαS.
Crystals of VC1:IIC2 bound to PPi in conjunction with either Ca2+ or Mg2+ diffracted to 2.7Å and 2.8Å resolution, respectively (Figure 6). These represent, to our knowledge, the first structures to be determined of a AC terminal product complex with pyrophosphate. Comparison with the unliganded enzyme shows that the β1—α1—α2 and α3—β4 loops of VC1 and β7′—β8′ of IIC2 in the PPi•Ca2+- and PPi•Mg2+-bound complexes conform to the open conformation of AC (Figure 3A). Global and local structural differences among apo VC1:IIC2 and the PPi•Ca2+, and PPi•Mg2+ complexes are minimal with root mean square deviations among equivalent Cα positions of 0.41 – 0.55 Å.
The electron density omit map for both the PPi•Mg2+ and PPi•Ca2+ complexes of VC1:IIC2 can be unambiguously fit to a model consisting of pyrophosphate bound to a magnesium ion with a water molecule, or a single calcium cation, respectively (Figure 6A and 6B). Superposition of PPi•Ca2+-bound VC1:IIC2 complex with that of the enzyme bound to Mg2+/Mn2+ and ATPαS, confirms that the calcium ion binds to the metal B site in the pyrophosphate complex. Difference Fourier analysis of the diffraction data obtained from crystals soaked in PPi•Ca2+ and PPi•Mg2+ revealed strong |FCaobs - FMgobs| difference electron density at the metal “B” site (Figure 6D) confirming that the B site contains Ca2+ in the former complex. Neither the PPi complex with Mg2+ nor that with Ca2+ show electron density at the metal A site.
The orientation of pyrophosphate in the active site of the open state AC complexes described here differs from the β-γ diphosphate moiety of ATPαS in the closed state complex with Mg2+ and Mn2+ (5) (Figure 4B) and from pyrophosphate in the closed complex with the P-site analog 2′-d-3′-AMP (53). Essentially, the Pβ-Pγ axis of the pyrophosphate rotates in conjunction with the change in the orientation of the C1 domain α1 helix in the transition from the closed to the open states of the enzyme. Consequently, hydrogen bond contact between the backbone amide groups of the α1 helix and PPi is maintained in both open and closed conformations. In the open state complexes with Mg2+ or Ca2+, the PPi phosphate distal to the C1 domain α1 helix (Pγ) is located in approximately the same position as the γ phosphate of the substrate analog in the ATPαS complex (Figure 4B) and its counterpart in the 2′-d-3′-AMP•PPi complex (5) (Figure 7). An oxygen atom from Pγ forms ion pair bonds with the positively charged guanidinium amides of Arg-484 from VC1 as seen in other structures of AC bound to substrate analogs, “P-site” and fluorophore-substituted inhibitors (Figure 6C) (5, 33, 53, 54). The proximal phosphate (Pβ) is oriented toward the β1—α1 loop of VC1 and forms hydrogen bonds with the backbone amides of Gly-399 and Phe-400. This is in contrast to the closed complex with 2′-d-3′-AMP•PPi, where the PPi Pγ phosphate forms hydrogen bonds with the amide of Gly-399, and the proximal phosphate is hydrogen bonded to the amide groups of Phe-400 and Thr-401 (53). Thus, the Pβ-Pγ axis of the pyrophosphate, which is aligned with the peptide bonds of residues 399-401 in the closed P-site complex, rotates ~35° in the PPi•Ca2+(Mg2+) complexes. This rotation allows the Pγ moiety to maintain hydrogen bond contact with the α1 helix as it swings into the open state (Figure 7).
The amino acid side chains that interact with the metal ion and PPi maintain similar conformations in the Ca2+ and Mg2+ complexes with two significant exceptions. First, Asp-396 adopts a different rotomeric state in the complex with Ca2+, wherein both carboxylate oxygen atoms contact the metal ion, in comparison to the Mg2+ complex, in which only one of the carboxylate oxygen atoms is a metal ligand (Figure 6C-D). Secondly, Lys-1065 forms an ion pair with the bridging oxygen of PPi in the PPi•Mg2+ complex but not in the complex with Ca2+. It appears that the loss of this ionic interaction is compensated by an additional coordinating contact between the Pβ phosphate and the metal ion. Interactions between Lys-1065 and phosphate groups of substrate analogs are generally observed in complexes with ATP analogs or fluorophore-substituted nucleotides, in which the enzyme adopts a closed or semi-closed conformation (5, 54)
B site Ca2+ coordination in the PPi complex is heptadentate, with some differences in the atoms that form the coordination sphere. A carboxylate oxygen of Asp-396 and an oxygen atom of Pγ form axial ligands, and five equatorial ligands contributed by Asp-396, Asp-440, the carbonyl of Ile-397 and Pβ form a rough pentagonal array. B-site Mg2+coordination in the PPi•Mg2+ complex is, in contrast, octahedral, involving oxygen ligands from Pβ, three VC1 residues (Asp-396, Asp-440, and Ile-397), and a water molecule (Figure 6C) but no interaction with Pγ. The mode of PPi — metal coordination observed here differs from that in the closed complex with 2′-d-3′-AMP•PPi•Mg2+ a mimic of the cAMP•PPi complex, where the 3′ nucleotide phosphate serves as the axial ligand of the B-site metal in place of a water molecule (53). In the closed complex with ATPαS, the α-thiophosphate acts in a similar role (Figure 5A). In both of the latter complexes the Pγ of PPi, or the corresponding nucleotide γ phosphate, participates in metal coordination.
The data presented here provide strong evidence that Ca2+ inhibits AC5 by displacing the Mg2+ co-factors at the catalytic site. Accordingly, mutations of residues adjacent to (Cys-441 or Tyr-442) or remote from (Phe-423 and Arg-434) the Mg2+ binding site that diminish sensitivity of AC5 to inhibition by Ca2+ also impair enzyme activation by Mg2+ (25). Further, Ca2+ stabilizes an open, catalytically inactive conformation of the C1 and C2 domains of AC. This is the case for both the substrate complex with ATP and pyrophosphate product complex. Because AC assumes an open conformation in the PPi-bound state, Ca2+ inhibits AC by stabilizing the enzyme:product complex.
The biphasic inhibition of AC5 by Ca2+ is recapitulated in the hybrid catalytic core composed of VC1 and IIC2. Inhibition by Ca2+ at micromolar concentration, a hallmark of AC5 and AC6, arises from binding at a location close to the A site. However, coordination of the metal is B site-like, such that Ca2+ is loosely bound in an electronegative cage formed by the β and γ phosphates of ATP and by Asp-396 and 440. In this complex the substrate analog ATPαS adopts an arched conformation that does not permit interaction of its α phosphate with either Ca2+ or with residues, such as Arg-1029 that provide transition state stabilization (Figure 5B) (27). The compact conformation of the nucleotide is unlikely to be due to the presence of the Rp thiolate at the α phosphate since ATPαS adopts an extended conformation in the closed-state complex with Mg2+ and Mn2+ (Figure 5A) (5).
The non-competitive relationship reported earlier between Mg2+ activation and high-affinity Ca2+ inhibition (9) arises in part from stabilization, by Ca2+, of an open conformation of the ATPαS•Ca complex. Further, because the nucleotide β and γ phosphates are engaged in binding Ca2+ near the A site, they are unavailable for Mg2+ binding at site B. A conformational transition, similar to that which occurs upon binding a second calcium ion (see below), could accommodate Mg2+ at the B site, leaving Ca2+ bound to site A. However, unlike Mg2+, a calcium ion would not activate the O-3′ hydroxyl of ATP for nucleophilic attack upon the α phosphate. Thus, A-site Ca2+ binding is inhibitory.
At ~100 μM concentration, Ca2+ binds to both A and B sites of open-state VC1:IIC2. Because, at that concentration a calcium ion is already bound to AC with B site-like coordination, it is likely that A site coordination is responsible for the low-affinity limb of the biphasic Ca2+ inhibition pattern. Binding of the second Ca2+ could be accompanied by translation of the first-bound Ca2+ to the B-site locus. This is likely to occur in concert with the conformational transition of ATP to an extended conformation (Figure 6C) (supplementary movie). However, because the expanded Ca2+ coordination sphere does not permit direct interactions between the β and γ ATP phosphate with the β1-α1 loop, AC remains in the open conformation. Rather, these phosphate moieties form water-mediated interactions with the amide groups in the open form of that loop. Bound to the open conformation of AC, ATP is miss-aligned with the catalytic residues. Interaction of Arg-1029 with the α phosphate of ATP, which provides transition-state stabilization, and hydrogen bonding of Asn-1025 to the ribose endocyclic oxygen, which may be important for substrate orientation (5), do not occur in the Ca2+ Complex (Figure 4D). Non-productive binding, and the failure of Ca2+ to provide catalytic activation at the A site, ensures that ATP does not turn over in crystals of Ca2+-bound VC1:IIC2.
The distance between the two Ca2+ ions bound to the A and B sites of the ATP•2Ca complex is considerably less than that observed between adjacent divalent ions (Mg2+ or Mn2+) that share coordinating ligands in other mAC complexes (5) or in RNA polymerase (55). The ionic radius of Ca2+ is larger than that of Mg2+ and Mn2+, consistent with separations between adjacent calcium ions ranging from 3.8 - 4.3Å in the 2.5Å resolution structure of the C2 domain of protein kinase Cβ (56). It is possible that the more weakly bound site A Ca2+ is partially occupied, such that the B site Ca2+ comes to occupy an intermediate position in asymmetric units in which Site A is unfilled. The Ca2+ positions reported here only represent average values for the two calcium ions in the refined structures. In fact, the distances between the two calcium ions range from 2.8 to 4.5 Å depending on the refinement constraints, consistent with dynamic interactions among two calcium ions, ATP, and neighboring protein residues and the modest resolution to which the crystals of the ATP•2Ca complex diffract.
Product release from AC does not follow an obligate ordered mechanism, but cAMP dissociates more rapidly than pyrophosphate from the catalytic site (28). The structural data reported here are consistent with the hypothesis that AC reverts to an open conformation upon release of cAMP, while pyrophosphate and Mg2+ are still bound to the enzyme. Mg2+ occupies the B site at this stage of the catalytic cycle, and is tethered in an octahedral coordination sphere to oxygen atoms of the Pβ pyrophosphate phosphate, and to the two conserved, and catalytically essential aspartate residues at positions 396 and 440. Hence, in the absence of stabilizing interactions with the purine moiety of the adenine nucleotide, and of metal-ion mediated interactions that link the nucleotide monophosphate and pyrophosphate groups, AC reverts to the relaxed, open conformation. The structure of the PPi•Mg2+ complex shows that PPi is a non-competitive inhibitor of AC because it stabilizes an open conformation of the enzyme to which ATP cannot bind productively.
High-affinity Ca2+ inhibition is non-exclusive with respect to pyrophosphate (Figure 1B), but appears to be exclusive at the low-affinity site. Even at millimolar concentration, only a single Ca2+ is bound to the VCI:IIC2•PPi complex, where it occupies the Mg2+ B site, and forms similar interactions with PPi and the conserved aspartate residues. The exclusivity between inhibition by PPi and by Ca2+ acting at the low-affinity site is probably due to the absence of the low-affinity A site in the open complex to which PPi binds. The A site is formed in part by the α phosphate of ATP, which has no analog in the PPi complex.
Ca2+ is a micromolar inhibitor of ATP synthesis, but is competitive with cAMP. However, since increase in Ca2+ from 2 to 5 μM is not accompanied by an appreciable decrease in AC activity, it is possible that Ca supports catalytic activity, possibly through interaction at the B site. The mechanism of this interaction requires further study.
Most vertebrate adenylyl cyclases, AC2, AC4 and AC7, for example, are not inhibited by micromolar Ca2+. Molecular Modeling of the Type II AC CHD domain onto the open conformation of VC1:IIC2 structure (1AZS) reveals several non-conserved amino residues around the catalytic site. The insensitivity of Type II AC to sub-micromolar Ca2+ might be attributed to local features of the binding site (see supplemental figure S1 and S2). For example, substitution of Ala-409 of VC1 with Pro-307 of IIC1 might restrict the movement of α1-α2 helixes upon binding of Ca2+ to the B site.
In contrast to AC5, bicarbonate-activated soluble adenylyl cyclase (sAC) is stimulated by Ca2+, which, with an EC50 = ~750 μM, decreases the KM of the enzyme for ATP from 10 mM to 1mM (57, 58). Crystallographic studies of the homologous sAC from Cyanobacteria have revealed open and closed conformational states that approximate those observed in VC1:IIC2 (59). Structures of the Cyanobacterial sAC catalytic core in complex with ATPαS or AMP(CH)2PP and Ca2+ (or Ca2+ mimetics Sr+2 and Eu+3) reveal exclusive binding of Ca2+ at the B metal site that leaves the Mg2+ site unoccupied . The Cyanobacterial sAC adopts an open conformation, particularly with respect to the β1-α1 element, that is similar to that of the Ca2+- bound VC1:IIC2 complexes. However, ATP is tightly bound — with contacts to both nucleoside and phosphate moieties - in the open, Ca2+-bound state of sAC. Thus, by stabilizing the nucleotide β— and γ— phosphates at the active site, the hepta-coordinate Ca2+ increases affinity for nucleotide, without fully destabilizing the metal A site for Mg2+ binding. We propose that global features of the catalytic site in the basal states of Class III ACs dictate their susceptibility to either activation or inhibition by Ca2+. It is probable that these structural features, which determine the volume of the active site and disposition of purine, metal ion and phosphate binding residues, are determined by many residues at and surrounding the interface between the CHD domains.
The mechanisms by which AC activity is dynamically regulated by Ca2+ are complex, involving differential responses by various AC isoforms, contingent on patterns of tissue-specific expression. The data presented here, in the context of previous studies, reveal the molecular mechanism by which fluctuations in the physiological concentration of Ca2+ might directly attenuate cAMP production by AC5 and AC6.
We thank the staffs at the Advanced Proton Synchrotron SBC-CAT ID-19 beamline and the Stanford Synchrotron Radiation Laboratory for their assistance with data collection.
†This work was supported by NIH Grant R01-DK46371 (S.R.S.) and Wellcome Trust grant RG31760 (DMFC). DMFC is a Royal Society Wolfson Research Fellow.