Structural Basis for the Inhibition of Mammalian Membrane Adenylyl Cyclase by 2 ′(3′)-O-(N-Methylanthraniloyl)-guanosine 5 ′-Triphosphate*

Membrane-bound mammalian adenylyl cyclase (mAC) catalyzes the synthesis of intracellular cyclic AMP from ATP and is activated by stimulatory G protein α subunits (Gαs) and by forskolin (FSK). mACs are inhibited with high potency by 2 ′(3′)-O-(N-methylanthraniloyl) (MANT)-substituted nucleotides. In this study, the crystal structures of the complex between Gαs·GTPγS and the catalytic C1 and C2 domains from type V and type II mAC (VC1·IIC2), bound to FSK and either MANT-GTP·Mg2+ or MANT-GTP·Mn2+ have been determined. MANT-GTP coordinates two metal ions and occupies the same position in the catalytic site as P-site inhibitors and substrate analogs. However, the orientation of the guanine ring is reversed relative to that of the adenine ring. The MANT fluorophore resides in a hydrophobic pocket at the interface between the VC1 and IIC2 domains and prevents mAC from undergoing the “open” to “closed” domain rearrangement. The Ki of MANT-GTP for inhibition of VC1·IIC2 is lower in the presence of mAC activators and lower in the presence of Mn2+ compared with Mg2+, indicating that the inhibitor binds more tightly to the catalytically most active form of the enzyme. Fluorescence resonance energy transfer-stimulated emission from the MANT fluorophore upon excitation of Trp-1020 in the MANT-binding pocket of IIC2 is also stronger in the presence of FSK. Mutational analysis of two non-conserved amino acids in the MANT-binding pocket suggests that residues outside of the binding site influence isoform selectivity toward MANT-GTP.

The nine membrane-bound isoforms of mammalian adenylyl cyclase (mAC), 1 which convert ATP to the ubiquitous second messenger cAMP, respond differently to a variety of regulatory molecules (1)(2)(3). All of the mAC isoforms are activated by the stimulatory G protein ␣ subunit (G␣ s ) and (except for type IX) the diterpene forskolin (FSK) and its soluble derivatives (1, 4 -7). ACs require Mg 2ϩ or Mn 2ϩ for catalytic activity, although, in general, Mn 2ϩ has greater affinity for mAC and is a more effective activator than the physiological ligand Mg 2ϩ (7)(8)(9).
Adenylyl cyclases are inhibited by adenosine and certain adenosine derivatives such as 2Ј-5Ј-dideoxyadenosine and 2Јdeoxyadenosine 3Ј-monophosphate that possess intact adenine rings and are known as "P-site" inhibitors (9 -13). These compounds bind preferentially to the FSK-and G␣ s ⅐GTP␥S-activated state of mAC (14). P-site inhibitors are dead-end inhibitors that bind to the catalytic site in the presence of pyrophosphate (15)(16)(17) and exhibit un-or noncompetitive inhibition in the presence of Mg 2ϩ or Mn 2ϩ , respectively, in the direction of cAMP synthesis (18). Certain substrate analogs, such as the R p stereoisomer of ATP␣S (19) and particularly ␤-L-2Ј,3Ј-dd-5Ј-ATP (K i ϭ 24 nM with native mAC from rat brain) have been identified as potent competitive inhibitors (20). More recently, it has been demonstrated that nucleotide triphosphates derivatized at the ribose 2Ј-or 3Ј-exocyclic ribose ring oxygen atoms by the fluorescent MANT moiety are also highly potent inhibitors of mAC (21,22). The unexpected inhibitory activity of MANT-GTP suggests novel routes for the design of inhibitors for mAC, which, as a significant target of G protein-coupled receptor stimulation, may be considered an appropriate target for drug development.
Crystal structures have been determined of complexes between the homologous cytosolic catalytic domains of mAC bound to FSK and GTP␥S-activated G␣ s (16). In these studies, the C1 and C2 domains, which constitute the N-and C-terminal halves of the catalytic core, were derived from type V and type II isoforms of AC and are thus designated VC1 and IIC2, respectively. The crystallographic studies revealed that both P-site inhibitors and substrate analogs bind, together with the metal ion co-factors, to the catalytic site located at the interface between C1 and C2. FSK occupies a site related by 2-fold pseudosymmetry to the catalytic site. Crystal structures have also been determined of the G␣ s -activated mAC catalytic core complexed with various P-site inhibitors and substrate analogs (17,23).
Upon binding to potent P-site inhibitors and substrate analogs such as 2Ј-deoxyadenosine 3Ј-monophosphate (K i ϭ 1.2 M) (24) and ␤-L-2Ј,3Ј-dd-5Ј-ATP, VC1⅐IIC2 undergoes a transition from an "open" to a "closed" conformation in which the ␣1-␣2, ␤2-␤3, and ␣3-␤4 loops of C1 and the ␤7Ј-␤8Ј loop of C2 move toward each other, thereby closing the gap between the purinebinding pocket in the C2 domain and the triphosphate binding loop (P-loop) located primarily in the C1 domain (16). This conformational change also places the N terminus of the C1 domain ␣1 helix (the P-loop) in hydrogen bond contact with the ␤and ␥-phosphates of ATP. Therefore, the architecture of the closed conformation constitutes the functional catalytic structure that facilitates substrate binding. In contrast, weak inhibitors comparable in affinity with the substrate ATP (K m ϭ 0.4 mM), such as the nonhydrolyzable ␣-␤ methylene derivative of ATP, bind to the open conformation of the enzyme (8).
In the C1 domain, two divalent metal ions are coordinated by the invariant aspartic acid residues Asp-396 and Asp-440 (type V AC numbering). These metal cations, which bind to sites designated "A" and "B" are, in the closed conformation of the enzyme, in position to bind the ␤and ␥-phosphates of ATP. The metal ion in site A is proposed to function as a Lewis acid by coordinating the 3Ј-ribosyl hydroxyl group (17,25). mAC activity is reduced ϳ2000-fold with Mg 2ϩ and 200-fold with Mn 2ϩ when either of the two aspartic acid residues are substituted with alanine or asparagine residues (23). These data, together with biochemical and modeling studies (25) led to the understanding that the cyclization of ATP is catalyzed by a "two-metal ion" mechanism similar to that used by DNA polymerases (17,23,26). Also apparent from these studies is the plasticity of the mAC-binding site, which allows variation in the binding mode by which phosphate groups (derived from the nucleotide or from pyrophosphate) and metal ions are accommodated (16,17,23).
Nucleotides derivatized at the ribose moiety by the fluorescent MANT group, several of which have been used as probes of nucleotide binding and hydrolysis in G proteins and other nucleotidases, were found to be highly potent competitive inhibitors of mAC (21). A systematic analysis of MANT nucleotide inhibition of several mAC isoforms showed MANT-GTP to have the greatest affinity and potency toward VC1⅐IIC2 (K i ϭ 4.2 nM) and to show selectivity among mAC isoforms (22). Additionally, replacement of Mg 2ϩ by Mn 2ϩ increases the inhibitory potency of MANT-GTP␥S by 9-fold. In the present work, we have determined the crystal structures of the FSK and G␣ s ⅐GTP␥S-activated VC1⅐IIC2 with MANT-GTP bound to the catalytic site in the presence of Mg 2ϩ or Mn 2ϩ and have further probed the structure of the complexes by fluorescence spectroscopy. Our results show that the accommodation of the MANT group by the mAC catalytic site results in a novel mechanism of inhibition.

EXPERIMENTAL PROCEDURES
Preparation of Proteins and Materials-Plasmids encoding the wild type and mutant AC C1a domain from canine type V (VC1), C2a domain from rat type II (IIC2), and bovine G␣ s , consisting of residues of 364 -580, 874 -1081, and 1-396, respectively, were expressed in Escherichia coli BL21 (DE3) cells, and the proteins were purified and stored as described previously (27). Mutants VC1-A409P (substitution of Ala-409 with Pro) and IIC2-I1006V (substitution of Ile-1006 with Val) were generated by site-directed mutagenesis (QuikChange; Stratagene). Except for IIC2, both VC1 and G␣ s constructs contained a hexahistidine tag at their N termini. G␣ s was activated by incubation with 10 M GTP␥S and 2 mM MgCl 2 at 30°C for 1 h, and the resulting G␣ s ⅐GTP␥S complex was further digested by trypsin to a smaller fragment comprising residues 39 -387 (27).
Adenylyl Cyclase Complex Formation and Crystallization with MANT-GTP-Purified recombinant VC1, IIC2, and trypsin-treated G␣ s ⅐GTP␥S were mixed in a 1.5:1:1 molar ratio to form a heterotrimeric complex in the presence of excess MP-FSK and GTP␥S. The protein mixture was purified by gel filtration using tandemly arranged Superdex 75 and 200 columns (Amersham Biosciences), and fractions containing the heterotrimeric complex were collected and concentrated to 8 mg/ml in a buffer of 20 mM Na ϩ HEPES, pH 8.0, 2 mM EDTA, 2 mM MgCl 2 , 2 mM dithiothreitol, 100 mM NaCl, 25 M MP-FSK, and 10 M GTP␥S for crystallization.
Structure Determination and Model Refinement-Diffraction data sets were collected by the oscillation method (1°/frame, 45 s/frame) at the Advanced Photon Synchrotron SBC-CAT BM19 Beamline with an incident beam wavelength of 1.0393 Å. The images were processed with the HKL2000 package (28) ( Table I). Because of anisotropy, data with l index Ͼ21 for crystals of the MANT-GTP⅐Mg 2ϩ complex and l index Ͼ18 for crystals of the MANT-GTP⅐Mn 2ϩ complex were excluded from the dataset. Crystals of the VC1⅐IIC2⅐FSK⅐MANT-GTP⅐G␣ s ⅐GTP␥S complex were approximately isomorphous with those of the VC1⅐IIC2⅐FSK⅐G␣ s ⅐GTP␥S complex (16), and the refined coordinates of the latter (Protein Data Bank code 1AZS) were used as the initial phasing model. Atomic positions and thermal parameters were refined by successive rounds of rigid body, simulated annealing, Powell minimization, and grouped B factor refinement using the CNS 1.1 program suite (29). The MANT-GTP and metals in the structure were located in the SIGMA-A weighted ͉F o ͉ Ϫ ͉F c ͉ omit map computed with phases from the refined model. The model was iteratively improved by manual refitting into weighted 2͉F o ͉ Ϫ ͉F c ͉ maps using the computer graphics program O (30) and subsequent cycles of refinement with CNS. Final refinement statistics are listed in Table I. Coordinates for the MANT-GTP⅐Mn 2ϩ and MANT-GTP⅐Mg 2ϩ structures have been deposited in the Protein Data Bank with the codes 1TL7 and 1U0H, respectively.
Fluorescence Spectroscopy and Data Analysis-All of the experiments were conducted using a Cary Eclipse fluorescence spectrophotometer equipped with a Peltier thermostatted multicell holder at 25°C (Varian, Walnut Creek, CA). The measurements were performed in a quartz fluorescence microcuvette (Hellma, Plainview, NY). The final assay volume was 150 l. The reaction mixtures contained a buffer consisting of 100 mM KCl, 10 mM MnCl 2 , or 10 mM MgCl 2 and 25 mM Na ϩ HEPES, pH 7.4. Steady-state emission spectra were recorded at low speed with ex ϭ 350 nm ( em ϭ 370 -500 nm) and ex ϭ 280 nm Data Analysis and Data Reproducibility-Fluorescence recordings were analyzed with the spectrum package of the Cary Eclipse software (Varian). Base-line fluorescence (buffer alone) was subtracted. Fig. 4 shows superimposed original fluorescence recordings. Similar results were obtained in four independent experiments using two different batches of VC1 and IIC2. Substrate saturation experiments and competition experiments with MANT-GTP were analyzed by nonlinear regression using the Prism 3.02 software (GraphPad).

Crystal Structures of MANT-GTP⅐Mn 2ϩ and MANT-GTP⅐Mg 2ϩ
Complexes-Structures of the G␣ s ⅐GTP␥S⅐VC1⅐ IIC2⅐FSK protein complex containing MANT-GTP⅐Mn 2ϩ or MANT-GTP⅐Mg 2ϩ were determined to resolutions of 2.8 and 2.9 Å, respectively ( Fig. 1). The ͉F o ͉ Ϫ ͉F c ͉ omit electron density for the MANT-GTP in the catalytic site is well defined and of sufficient quality to allow the position and orientation of the purine, ribose, MANT, and 5Ј-triphosphate groups to be unambiguously determined (Fig. 2). In particular, the electron density of the planar purine ring and the ribose moieties delimits the positions of the hydrogen bonding groups of the former in the binding cavity, even though they are not clearly resolved in the electron density (see Fig. S1 published online as supplementary material). Electron density for the B-site metal ion is well defined in both structures (see Fig. S2 published as supplementary material on line), but density for the A-site is weaker than that for the B-site in the Mg 2ϩ -bound complex. MANT-GTP⅐Mn 2ϩ appears better ordered than MANT-GTP⅐Mg 2ϩ . With the assumption of 100% occupancy, the average temperature factors of bound FSK and GTP␥S are 32 and 38 Å 2 in the MANT-GTP⅐Mn 2ϩ and MANT-GTP⅐Mg 2ϩ structures, respectively. However, the average temperature factors for MANT-GTP⅐Mn 2ϩ and MANT-GTP⅐Mg 2ϩ are 55 and 73 Å 2 , respectively, indicating greater disorder or lower occupancy of MANT-GTP⅐Mg 2ϩ in complex with VC1⅐IIC2.
The MANT fluorophore is bound in a cavity at the interface between the ␤1-␣1-␣2 loop of VC1 and the ␣4Ј-␤5Ј and ␤7Ј-␤8Ј loops of IIC2 (Fig. 1, A and B). To accommodate the MANT group within the cavity, the MANT-GTP adopts an orientation in the catalytic site that is related to that of other substrate or P-site analogs by a ϳ180°rotation about the purine-phosphate axis. The consequences of this unexpected binding mode on the conformation of C1⅐C2, recognition of the nucleotide moiety and metal ions by the enzyme and the mechanism of inhibition by MANT-GTP, are described below.
The conformation of the MANT-GTP-bound C1⅐C2 complex is midway between the open and closed states. As shown in Fig. 1B, in which the MANT-GTP-bound, ␤-L-2Ј,3Ј-dd-5Ј-ATP-bound, and unbound structures are superimposed, the ␤1-␣1-␣2 and ␣3-␤4 loops of VC1 and ␤7Ј-␤8Ј of IIC2 shift toward the domain interface but fail to close fully. Superposition of the MANT-GTPbound complex with unliganded (open) and ␤-L-2Ј,3Ј-dd-5Ј-ATPbound (closed) complexes yields an overall rmsd of 0.6 and 0.8 Å, respectively, for all equivalent C␣ positions of the VC1⅐IIC2 complex. However, rmsd values for the C␣ atoms in the latter interfacial structural elements are considerably larger: 1.9 Å with respect to the closed complex and 1.0 Å with respect to the open complex. For comparison, the rmsd for the same C␣ atoms after superposition of open and closed complexes is 1.9 Å. The MANT fluorophore that is bound between ␤1-␣1-␣2 of VC1 and ␤7Ј-␤8Јof IIC2 thus acts as a wedge to prevent VC1⅐IIC2 from adopting a fully closed conformation.
Other than as noted above, global and local structural differences between the Mn 2ϩ and Mg 2ϩ complexes are small, as indicated by the rmsd between equivalent C␣ positions of 0.4 Å. The largest differences are apparent in the ␣1 helix, ␤1-␣1 loop, and ␣3-␤4 loop of VC1 in the MANT-GTP⅐Mg 2ϩ complex, which adopt a slightly more open conformation than the MANT-GTP⅐Mn 2ϩ structure. This structural difference may reflect the higher affinity of VC1⅐IIC2 for MANT-GTP⅐Mn 2ϩ (21,22).
The MANT-GTP used in this study is presumably a racemic mixture of ribosyl 2Ј-O-and 3Ј-O-MANT derivatives. Our previous molecular modeling studies suggested that MANT nucleotides bind to mAC preferentially in the 2Ј-O-MANT conformation (21). However, it is clear that crystals of mAC contain only the 3Ј-O-enantiomer. A stereochemically reasonable model of the 2Ј-O-enantiomer cannot be accommodated in the experimental electron density. The possibility that the latter is a substrate for VC1⅐IIC2 and is thus consumed by the enzyme is unlikely because the putative reaction product MANT-cGMP shows no fluorescence increase with VC1⅐IIC2, and there was no time-dependent decrease in MANT-GTP fluorescence upon binding to VC1⅐IIC2 (data not shown; see below for discussion of MANT-GTP fluorescence). The backbone nitrogen of Asn-1025 from the IIC2 domain appears to form a hydrogen bond with the 2Ј-ribosyl hydroxyl, whereas in the AC complex with ␤-L-2Ј, 3Ј-dd-5Ј-ATP, the carboximido NH 2 group of Asn-1025 is the mean intensity after rejections.
c Due to anisotropy, data with an l index greater than 18 (Mn 2ϩ complex) or 21 (Mg 2ϩ complex) were omitted from refinement.
and F c (h) are the observed and computed structure factors; no I cut-off was used during refinement. e 5% of the complete data set was excluded from refinement to calculate R free .
contacts the endocyclic ring oxygen of the ribose moiety. The conformation of Asn-1025 also differs from that in complexes with MANT-GTP and thereby avoids steric conflict with the 3Ј-MANT-ribosyl moiety. The polar substituents of the MANT moiety are oriented toward the exterior of the binding cleft between ␤1-␣1-␣2 of VC1 and ␣4Ј-␤5Ј of IIC2 (Figs. 1B and 3A) and form no hydrogen bonds with the protein. The nonpolar surface of the N-anthraniloyl ring is in van der Waals contact with residues in the hydrophobic pocket formed by Phe-400, Ala-404, Ala-409, Leu-412, and Leu-413 from VC1 and Val-1006, Trp-1020, Gly-1021, and Asn-1022 from IIC2 (Fig. 3A). A structure-based sequence alignment of selected mammalian AC family members (Fig. 3B) shows that the residues in the MANT-binding pocket are well conserved among type I-VIII membrane-bound mAC isoforms with the exception of type II at residues 307 (residue 409 for type V) and 1006. Several aromatic residues are part of, or close to, the MANT-binding pocket; in particular, the indole ring of Trp-1020, which is located less than 5 Å from the N-methylanthraniloyl ring of MANT-GTP.
The positions occupied by the purine ring and 5Ј-triphosphate substituents of MANT-GTP and the two metal ions are roughly similar to those of the corresponding moieties in P-site inhibitor⅐PP i or ␤-L-2Ј,3Ј-dd-5Ј-ATP complexes with VC1⅐IIC2 but differ in significant details (Fig. 1C).  (16,23), were superimposed and depicted as transparent, rose-and gray-colored elements, respectively. C, interactions are shown among protein residues, MANT-GTP, and two metal ions in the VC1⅐IIC2 substrate-binding site. MANT-GTP and protein residues are shown as stick models. Stick models of ␤-L-2Ј,3Ј-dd-5Ј-ATP and selected side chains in Asn-1025 in 1CJU (see A) are shown in yellow. The gray dashed lines depict the hydrogen bonds between MANT-GTP and protein residues and coordination of the metal ions at sites A and B (see text). Because of the high thermal parameters of the nucleotide (see text), the hydrogen bond distances are approximate. The coordination between metal ions and MANT-GTP is similar to that in the complex with ␤-L-2Ј,3Ј-dd-5Ј-ATP; metal A has four ligands, whereas metal B has five ligands. D, detailed view of the purine-binding pocket in the mAC substrate-binding site (see text) The superimposed adenosine group of ␤-L-2Ј,3Ј-dd-5Ј-ATP is shown in yellow. The distances of hydrogen bond between guanine ring and surrounding protein residues are as indicated.
ion at site B is in octahedral coordination with oxygen atoms from each of the ␤and ␥-phosphates, the ␤-␥ phosphate bridging oxygen and three VC1 residues (Asp-396, Asp-440, and Ile-397). Metal A is also ligated by Asp-396 and Asp-440 and oxygen atoms from the ␣and ␤-phosphates. Although interactions between the two metal ions and the nucleotide phosphates of MANT-GTP differ from those observed with the phosphate groups of P-site inhibitors and substrate analogs, the coordination of the two metal ions by their protein ligands is similar (Fig. 1C). As noted above, the occupancy of Mg 2ϩ at the A site appears to be considerably lower than that of Mn 2ϩ . We cannot exclude the possibility that an ordered water molecule occupies the A site in the MANT-GTP⅐Mg 2ϩ -mAC complex.
The guanine ring of MANT-GTP is located at the same position occupied by the adenine rings of substrate analogs but differs from these analogs in its orientation with respect to protein hydrogen bonding partners (Fig. 1D). The guanine ring is stacked upon the peptide planes of Leu-438 and Gly-439 in the ␤2-␤3 loop of VC1 and could form hydrogen bonds from the guanine N-1 atom to Asp-1018, the guanine N2 atom to the backbone carbonyl of Ile-1019, and from the guanine O-6 atom to Lys-938. The guanine ring adopts an anti conformation ( ϭ 234°) to the ribose ring, in contrast to the glycosidic angle of 107°observed in ␤-L-2Ј,3Ј-dd-5Ј-ATP and of 157°for ATP␣S-R p . Thus, the orientation of the guanine ring differs by ϳ180°(with respect to the purine N-9 -N-1 axis) from that of adenine in complexes with ATP analogs. However, the guanine ring in this reversed orientation presents a similar pattern of hydrogen bond donors and acceptors to the enzyme, compared with adenine. To accommodate the orientation of the guanine ring and the 3Ј-O-ribosyl-MANT group, the ribose of MANT-GTP adopts a C2Ј-endo conformation (Fig. 1, C and D).
Kinetic analysis of VC1⅐IIC2-In previous studies we showed that the potencies of MANT nucleotides for inhibiting various mACs are higher in the presence of Mn 2ϩ than with Mg 2ϩ (21,22). Those data suggested that formation of a catalytically competent conformation by mAC might facilitate binding of MANT nucleotides to the substrate-binding site. To further substantiate this notion, we examined the activity of the VC1⅐IIC2 protein complex in the presence of Mg 2ϩ or Mn 2ϩ in the absence or presence of FSK and G␣ s ⅐GTP␥S or both activators (Table II; (22). Consequently, values for kinetic constants determined here differ in some cases by 2-3-fold from those reported previously, but relative values for the same inhibitors examined under various activation conditions are similar. In the presence of Mn 2ϩ , mAC activities were generally higher than under the corresponding conditions in the presence of Mg 2ϩ . As reported in other studies (32,33), both FSK and G␣ s ⅐GTP␥S increased apparent V max in the presence of both cations, and FSK and G␣ s ⅐GTP␥S exhibited synergistic effects on enzyme activity. In the presence of Mg 2ϩ or Mn 2ϩ , the potency of MANT-GTP for inhibition of mAC correlated with V max . For each experimental condition (basal, ϩG␣ s ⅐GTP␥S, ϩFSK, ϩG␣ s ⅐GTP␥S and FSK), the potencies of MANT-GTP were higher for Mn 2ϩ than for Mg 2ϩ . Thus, although MANT nucleotides are competitive mAC inhibitors (21,22), they have in common with the noncompetitive (P-site) inhibitors the characteristic that inhibitor potency increases with enzyme activity (18). Accordingly, from a pharmacological perspective, receptor agonist-stimulated mAC activities should be more sensitive to inhibition by MANT nucleotides than basal mAC activities.
Except for two positions (residue 409 in VC1 and residue 1006 in IIC2), the amino acids that constitute the MANTbinding site are conserved among mAC isoforms (Fig. 3B). To assess their contribution to the inhibitory potency of MANT-GTP, we constructed cyclase domains in which the two nonconserved residues are mutated and assayed the affinities of the mutant VC1⅐IIC2 enzymes for MANT-substituted nucleotides. Substitutions of Ala-409 of VC1 with proline and Ile-1006 of IIC2 with valine were engineered to mimic the MANT-binding site composition of type II or type V mAC, respectively (Fig.  3B). Neither mutation substantially affected K m for ATP in the presence of Mn 2ϩ , G␣ s ⅐GTP␥S, and FSK (Table III). However, VC1-A409P (in combination with IIC2) exhibits ϳ3-fold higher V max for cAMP synthesis than VC1⅐IIC2, whereas that for IIC2-I1006V (in conjunction with VC1) is 70% of wild type. The inhibitory potencies of MANT nucleotides toward mutant mAC proteins were 2-10-fold lower than toward wild type VC1⅐IIC2 depending on the combinations of mutant mAC protein subunits studied (Table III). These data indicate that the nonconserved residues in the MANT-binding pocket influence the affinity of the mAC catalytic core for MANT nucleotides. Both the VC1⅐IIC2-I1006V complex, which mimics the putative hydrophobic binding pocket of type V mAC, and the VC1-A409P⅐IIC2 complex, which mimics that of type II mAC have less affinity for MANT nucleotides than the wild type mAC  (Table III).
We also found that the potency of 2Ј-deoxy-3Ј-O-(N-methylanthraniloyl)-adenosine 5Ј-triphosphate is between 3-and 13fold less than that of 2Ј(3Ј)-MANT-GTP in the presence of both Mn 2ϩ and Mg 2ϩ , respectively. These findings are consistent with the structural data showing possible hydrogen bond formation between the 2Ј-ribosyl-OH of MANT-GTP and the backbone nitrogen of Asn-1025 from IIC2 (Fig. 1C). Additional studies on MANT nucleotides with different purine substitutions show that 2Ј(3Ј)-O-(N-methylanthraniloyl)-xanthosine 5Јtriphosphate is a less potent inhibitor of activated mAC in the presence of either Mn 2ϩ or Mg 2ϩ than MANT-GTP, whereas MANT-ATP is a very potent inhibitor of activated mAC, and its inhibitory potency increases with the substitution of Mn 2ϩ for Mg 2ϩ .
Effects of C1/C2 and MP-FSK on Emission Spectra of MANT-GTP-Fluorescence emission from MANT-GTP was measured in the absence and presence of VC1⅐IIC2 and MP-FSK either by FRET from excitation of tryptophan at 280 nm (35) (Fig. 4, A and C) or by direct excitation of MANT fluorescence at 350 nm (34) (Fig. 4, B and D). The experiments were conducted in the presence of either Mn 2ϩ (Fig. 4, A and B) or Mg 2ϩ (Fig. 4, C  and D).
The addition of VC1⅐IIC2 in the presence of Mn 2ϩ increased MANT fluorescence by 70% upon excitation at 350 nm and shifted the MANT emission maximum from 450 to 420 nm (Fig.  4B). The change in fluorescence emission intensity is consistent with the transfer of the MANT group to a hydrophobic environment (34) as is the blue shift of the emission maximum toward shorter wavelengths (36,37). The addition of MP-FSK further increased fluorescence emission of MANT-GTP to about 3-fold above the value for the unbound compound. Also in the presence of VC1⅐IIC2 and Mn 2ϩ , excitation at 280 nm produces a broad emission peak at 350 nm (Fig. 4A) that is not observed in the absence of protein, and a shoulder of FRET-stimulated emission at 420 nm. Upon the addition of MP-FSK, there was a loss in fluorescence intensity at 350 nm with a concomitant increase in emission at 420 nm. These data show that binding of MP-FSK to VC1⅐IIC2 allows for efficient FRET between a tryptophan residue, presumably Trp-1020, in the MANT-binding site and the MANT fluorophore.
Mutation of conserved residues in mAC hydrophobic environment results in modest changes in the fluorescence emission spectrum because of binding of the MANT fluorophore. In the absence of FSK, the increase in the intensity of fluorescence emission upon direct excitation of MANT ( ex ϭ 350 nm) caused by binding is unchanged for VC1-A409P⅐IIC2, reduced by ϳ50% for VC1⅐IIC2-I1006V, and reduced by ϳ70% in the double mutant (see Fig. S4 in supplementary material published on line). Similarly, FRET-stimulated emission ( ex ϭ 280 nm) is markedly reduced for VC1⅐IIC2-I1006V and the double mutant. However, in the presence of MP-FSK, the intensity of both direct and FRET-stimulated emission from MANT is similar to that from VC1⅐IIC2.
In contrast to the results obtained in the presence of Mn 2ϩ , the addition of VC1⅐IIC2 to MANT-GTP in the presence of Mg 2ϩ resulted only in a small fluorescence increase upon excitation at 350 nm (Fig. 4D). The increase in MANT-GTP fluorescence in the presence of both Mg 2ϩ and MP-FSK was similar to that observed in the presence of Mn 2ϩ alone (Fig. 4B). The blue shift in the fluorescence emission spectrum of MANT was also much less pronounced and was only observed upon the addition of MP-FSK (Fig. 4D). Finally, in the presence of Mg 2ϩ , MP-FSK was much less efficient at promoting FRET than in the presence of Mn 2ϩ (Fig. 4, C versus A). Collectively, these data are indicative of a much stronger interaction between VC1⅐IIC2 and MP-FSK in the presence of Mn 2ϩ than in the presence of Mg 2ϩ and corroborate the enzymatic and crystallographic data. DISCUSSION MANT-GTP is the most potent member of a newly described family of competitive inhibitors for mAC (21,22). The crystallographic results described here show that MANT-GTP acts by a novel inhibitory mechanism in which the MANT substituent occupies a hydrophobic pocket at the interface between the homologous cytosolic domains that form the catalytic core of the enzyme. This pocket is not occupied by substrate and could FIG. 3. Structure-based sequence alignment and molecular surface of amino acid residues that form the binding pocket for the MANT fluorophore of MANT-GTP in VC1⅐IIC2. A, atoms of residues in the hydrophobic pocket are drawn as spheres with van der Waals' radii, and the nonconserved residues in the MANT-binding site are Ala-409 of VC1 (green) and Ile-1006 of IIC2 (pink). MANT-GTP is drawn as a stick model. B, sequence alignment of selected mammalian mAC isoforms and the soluble guanylyl cyclase ␣ subunit (sGC␣), for residues that form the hydrophobic binding pocket for the MANT group of MANT-GTP. The matching secondary structure assignments for the C1 and C2 domains are drawn above the alignment. Coils represent ␣ helices (pink), arrows represent ␤ strands (blue), and solid lines represent turns or irregular secondary structure. Red amino acid symbols correspond to residues that form the hydrophobic pocket surrounding the MANT fluorophore. Nonconserved amino acid residues in the MANT-binding site are colored green for VC1-Ala409 and pink for IIC2-Ile1006. be exploited in the design of specific mAC inhibitors. At the same time, MANT-GTP also occupies the substrate nucleoside and triphosphate-binding sites, although by a different mode than that used by substrate analogs and P-site inhibitors.
On binding to VC1⅐IIC2, the MANT fluorophore of MANT-GTP stabilizes a conformation that is intermediate between the open and closed forms of the enzyme by blocking the movement of ␤1-␣1-␣2 and ␣3-␤4 of VC1 toward ␤7Ј-␤8Ј of IIC2. The partial collapse of these structural elements creates a hydrophobic binding site for the MANT substituent. The increase in blue-shifted MANT fluorescence that is observed upon binding of MANT-GTP to the enzyme is consistent with this mode of interaction. The binding energy caused by the MANT group is substantial, as indicated by the 60-fold difference in affinity for VC1⅐IIC2 between inosine 5Ј-[␥-thio]triphosphate (IC 50 ϭ 1200 nM) and MANT-inosine 5Ј-[␥-thio]triphosphate (IC 50 ϭ 19 nM) (22). The interaction between the MANT group and VC1⅐IIC2 is highly complementary: substitution of a larger bulky hydrophobic group, such as a BODIPY moiety at the 2Ј(3Ј)-O-ribosyl positions, reduces potency up to 40-fold (IC 50 ϭ ϳ960 -15,000 nM), depending on the mAC isoform (21). The inhibitory potency of MANT nucleotides is reduced by mutations of residues that form the MANT-binding pocket. Crystals of FSK-and G␣ s -GTP␥S-activated VC1⅐IIC2 appear to preferentially bind the 3Ј-O-ribosyl derivative of MANT-GTP. The 2Ј-d-3Ј-MANT derivative, in which the fluorophore is fixed at the 3Ј-O-ribosyl position, is less potent than MANT-GTP, possibly because of the loss of a hydrogen bond between the 2Ј-ribosyl-hydroxyl and the backbone nitrogen of Asn-1025 in the IIC2 domain (Table II).
The ␥-phosphate moiety of MANT-GTP contributes substantially to the binding energy of the complex with mAC. The K i of MANT-GDP is 70-fold greater than that for MANT-GTP and replacement of ␥-phosphate of MANT-GTP with ␥-thiophosphate reduced its potency by 6-fold (22). These substitutions are predicted to reduce or eliminate coordination with the metal ion at the B site and neighboring protein side chains. Inhibitory potency is substantially reduced (IC 50 ϭ ϳ79,000 nM) if a bulky BODIPY group is attached to the ␥-thiophosphate of GTP (21).
Mg 2ϩ and Mn 2ϩ both serve as co-factors for mAC catalytic activity (Table II), but the inhibitory potency of MANT-GTP and other MANT nucleotide inhibitors (with the exception of 2Ј-deoxy-3Ј-O-(N-methylanthraniloyl)-adenosine 5Ј-triphosphate) is 3-10-fold greater in the presence of Mn 2ϩ compared with Mg 2ϩ . This increased inhibitory potency is paralleled by the magnitude of both intrinsic and FRET-stimulated emission from the MANT. No significant structural differences are apparent between the MANT-GTP⅐Mn 2ϩ and MANT-GTP⅐Mg 2ϩ complexes. However, MANT-GTP⅐Mg 2ϩ has a relatively high temperature factor and overall weak electron density compared with the MANT-GTP⅐Mn 2ϩ . Although the resolution of the data is insufficient to allow accurate measurement of metal coordination bond lengths, it is possible that Mg 2ϩ , which possesses a smaller ionic radius than Mn 2ϩ , is more loosely tethered to coordinating groups in the enzyme.
To accommodate the fluorophore of MANT-GTP in the active site of VC1⅐IIC2, the nucleotide must bind in the reverse orientation to that exhibited by adenine-containing compounds. In this orientation, the guanine N-2 atom occupies the same position as the adenine N-6, and both serve as hydrogen bond donors to the carbonyl oxygen of Ile-1019. The protonated guanine N-1 atom donates a hydrogen bond to the carboxylate of Asp-1018, whereas the unprotonated adenine N-1 is a hydrogen bond acceptor for the amine of Lys-938. The keto oxygen substituent of guanine C-6, which has no counterpart in adenine, accepts a hydrogen bond from Lys-938. The 25-90-fold reduction in binding energy for activated VC1⅐IIC2 incurred by substitution of the guanine for xanthine is not unexpected, because substitution of the guanine N-2 atom with oxygen abolishes a hydrogen bond and may lead to an electrostatically unfavorable interaction with the carbonyl oxygen of Ile-1019  ( Fig. 1D). MANT-substituted nucleotides are also substantially less potent inhibitors toward soluble guanylyl cyclase, which, although homologous to mAC, has different specificity-determining residues in its purine-binding site (22) In view of the "reverse" purine orientation observed for MANT-GTP in the active site of VC1⅐IIC2, it is remarkable that MANT-ATP is also a potent inhibitor of the fully activated mAC (K i ϭ 13 nM). That MANT-ATP binding is accompanied by an increase in intrinsic MANT fluorescence is evidence that the MANT group is accommodated in a hydrophobic binding site in the enzyme. However, the absence of FRET-stimulated emission from MANT upon excitation of tryptophan fluorescence 2 is evidence that the mode of MANT-ATP binding differs from that of MANT-GTP. Crystallographic studies will be necessary to better define the binding mode for MANT-ATP.
The several isoforms of mammalian mAC are all activated through G protein-coupled receptors by G␣ s but otherwise have multiple and distinct regulatory properties. Many mAC isoforms are differentially expressed in tissues. In particular, type V mAC is the predominant isoform in the heart (38). Hence, pharmacologically active compounds specific for that isoform of mAC may have some value in the treatment of heart failure (13, 39 -41). Several good inhibitors of mAC have been identified, including the more potent P-site inhibitors with 3Јpolyphosphate modifications. Essentially substrate analogs, Psite inhibitors bind to conserved catalytic residues of mAC (16,17,23). Although substantial isoform specificity is not expected, modest selectivities in the 3-30-fold range have been reported for P-site analogs (42).
MANT-substituted nucleotides are highly effective inhibitors of mAC and have potential for development as isoform-specific inhibitors. The potency of MANT-GTP (IC 50 ϭ ϳ20 nM) is comparable with that of ␤-L-2Ј,3Ј-dd-5Ј-ATP (IC 50 ϭ ϳ24 nM), which is among the strongest inhibitors so far identified (13,20,22). A comprehensive analysis of MANT-substituted nucleotides revealed that types V and VI mAC are ϳ10-fold more selective for MANT-GTP than type II mAC (22). The difference in selectivity among these isoforms might reasonably be attributed to two residues, Ala-409 and Val-1108 in type V mAC, that are substituted by Pro and Ile, respectively, in type II mAC. These are the only residues at the MANT-binding site that are not conserved among mAC isoforms. Mutation of these residues affects the potency of MANT-GTP inhibition, consistent with their location in the inhibitor-binding pocket. As expected, MANT-GTP has less potency toward VC1-A409P⅐IIC2, which possesses a MANT-binding pocket similar to that of type II mAC, than toward VC1⅐IIC2. The reduction in the potency of the inhibitor is not reflective of a loss in the catalytic activity of the enzyme because of the mutation. Rather, the A409P mutation increases catalytic activity (Table III), consistent with partial restoration of type II-specific contacts across the domain interface. However, MANT-GTP is also less active toward VC1⅐IIC2-I1006V, which is expected to have a MANT-binding pocket similar to that of type V mAC. Enzyme activity is also compromised by this mutation. Thus, we are not able to elicit a "gain-of-function" response with respect to MANT-GTP inhibition by substitution of a single residue that partially restores a type V C1-C2 interface near the MANT-binding pocket. It is well to note that isoform specificity toward MANT-GTP was demonstrated for mAC holoenzymes but was not explored for soluble domains (22). It is probable that the determinants of isoform selectivity for MANT-GTP include residues distant from the inhibitor-binding site, as well as residues outside of the core catalytic domains (for example in C1b) that might influence interdomain contacts.
The structure of VC1⅐IIC2 bound to MANT-GTP has revealed a ligand-binding site that is not utilized by substrates and perhaps is therefore not highly conserved. It may be possible to design analogs of MANT-GTP to increase the strength and selectivity of binding at the MANT-binding site of mAC. Because a substantial amount of binding energy is derived from interactions at this hydrophobic site, modification or replacement of other nucleotide substituents might increase potency or specificity without compromising bioactivity. Aside from their potential as mAC inhibitors, MANT-GTP and its relatives are also convenient spectroscopic probes for mAC activation because they bind preferentially to the G␣ s -and FSK-bound conformational states of mAC.