Soluble guanylyl cyclase (sGC) plays an important role in cardiovascular function and catalyzes formation of cGMP. sGC is activated by nitric oxide and allosteric stimulators and activators. However, despite its therapeutic relevance, the regulatory mechanisms of sGC are still incompletely understood. A major reason for this situation is that no crystal structures of active sGC have been resolved so far. An important step toward this goal is the identification of high-affinity ligands that stabilize an sGC conformation resembling the active, “fully closed” state. Therefore, we examined inhibition of rat sGCα1β1 by 38 purine- and pyrimidine-nucleotides with 2,4,6,-trinitrophenyl and (N-methyl)anthraniloyl substitutions at the ribosyl moiety and compared the data with that for the structurally related membranous adenylyl cyclases (mACs) 1, 2, 5 and the purified mAC catalytic subunits VC1:IIC2. TNP-GTP [2′,3′-O-(2,4,6-trinitrophenyl)-GTP] was the most potent sGCα1β1 inhibitor (Ki, 10.7 nM), followed by 2′-MANT-3′-dATP [2′-O-(N-methylanthraniloyl)-3′-deoxy-ATP] (Ki, 16.7 nM). Docking studies on an sGCαcat/sGCβcat model derived from the inactive heterodimeric crystal structure of the catalytic domains point to similar interactions of (M)ANT- and TNP-nucleotides with sGCα1β1 and mAC VC1:IIC2. Reasonable binding modes of 2′-MANT-3′-dATP and bis-(M)ANT-nucleotides at sGC α1β1 require a 3′-endo ribosyl conformation (versus 3′-exo in 3′-MANT-2′-dATP). Overall, inhibitory potencies of nucleotides at sGCα1β1 versus mACs 1, 2, 5 correlated poorly. Collectively, we identified highly potent sGCα1β1 inhibitors that may be useful for future crystallographic and fluorescence spectroscopy studies. Moreover, it may become possible to develop mAC inhibitors with selectivity relative to sGC.
Soluble guanylyl cyclase (sGC) is activated by nitric oxide (NO) and plays an important role in the regulation of cardiovascular and neuronal functions (Friebe and Koesling, 2009; Kots et al., 2011; Derbyshire and Marletta, 2012). Binding of NO to the heme moiety of the enzyme induces a conformational change so that sGC then effectively catalyzes the conversion of GTP into the second messenger cGMP (Friebe and Koesling, 2009; Kots et al., 2011; Derbyshire and Marletta, 2012). NO-containing drugs are used clinically to improve symptoms in coronary heart disease (Daiber et al., 2010). However, a major problem in the clinical use of NO-containing drugs is desensitization, rendering long-term therapy difficult (Daiber et al., 2010). The more recently developed allosteric sGC activators and sGC stimulators are devoid of this serious problem and have substantial potential for the treatment of various cardiovascular diseases, including pulmonary hypertension and heart failure (Gheorghiade et al., 2013; Stasch and Evgenov, 2013; Follmann et al., 2013). Despite its profound therapeutic relevance, the molecular regulatory mechanisms of sGC function are still incompletely understood (Baskaran et al., 2011; Kumar et al., 2013; Underbakke et al., 2013).
The active sGC holoenzyme is a heterodimer formed by the two subunits sGCα and sGCβ. The catalytic sGC domains, sGCαcat and sGCβcat, are closely related to the catalytic domains of membranous ACs (mACs), C1 and C2, respectively (Sunahara et al., 1998; Seifert et al., 2012). Recently, crystal structures of an sGCβcat homodimer and an sGCαcat/sGCβcat heterodimer have been resolved (Allerston et al., 2013); however, they represent catalytically inactive, “open” conformations. In contrast, the “fully closed” active state of mACs is well known from several structures of the heterodimeric hybrid VC1:IIC2 (Tesmer et al., 1997, 2000). A major technical advance in the resolution of crystal structures of VC1:IIC2 was achieved by the use of high-affinity competitive mAC inhibitors, i.e., TNP [2′,3′-O-(2,4,6-trinitrophenyl)]– and (M)ANT [2′,3′-O-(N-methylanthraniloyl)]–substituted NTPs as stabilizing ligands (Mou et al., 2005, 2006; Hübner et al., 2011). These nucleotides bind to the catalytic site of VC1:IIC2, and the TNP or (M)ANT group projects into a pocket adjacent to the catalytic site, conferring high affinity of the ligands for mACs and stabilizing inactive mAC conformations between the open and closed states (Mou et al., 2005, 2006; Hübner et al., 2011; Seifert et al., 2012). Hydrophobic interactions of the TNP- or (M)ANT-nucleotides were also explored to monitor conformational changes in mACs by fluorescence spectroscopy, providing major new insights into the dynamics of enzyme regulation (Mou et al., 2005, 2006; Pinto et al., 2009, 2011; Suryanarayana et al., 2009; Hübner et al., 2011).
In previous studies we showed that several TNP- and (M)ANT-nucleotides are potent inhibitors of sGC conventionally purified from bovine lung (Gille et al., 2004; Suryanarayana et al., 2009). The most potent inhibitor of bovine sGC is TNP-ATP (Ki, 7.3 nM) (Suryanarayana et al., 2009). However, for mechanistic and biophysical studies, the recombinant enzyme rat sGCα1β1 is much better suited (Hoenicka et al., 1999; Stasch et al., 2001; Beste et al., 2012). In addition, since we completed our previous inhibitor studies on bovine sGC (Gille et al., 2004; Suryanarayana et al., 2009), various new (M)ANT-nucleotides were developed and characterized at ACs 1, 2, 5 (Geduhn et al., 2011). All these data prompted us to systematically characterize the interaction of recombinant rat sGCα1β1 with a series of 38 TNP- and (M)ANT-nucleotides with the goal to identify high-affinity inhibitors. Figure 1 shows the structures of the compounds studied. Moreover, based on the availability of crystal structures of sGCαcat/sGCβcat and mAC VC1:IIC2 (Tesmer et al., 1997; Mou et al., 2005; Allerston et al., 2013), we performed docking studies on a rat sGCα1β1 model.
Materials and Methods
GTP (≥95%), triethylamine, phosphocreatine, creatine kinase, dithiothreitol, EGTA, sodium nitroprusside, sodium acetate, and bovine serum albumin were purchased from Sigma-Aldrich (Seelze, Germany). 3′-MANT-2′-dATP [3′-(N-methyl-anthraniloyl)-2′-deoxy-ATP], 2′-MANT-3′-dATP [2′-O-(N-methyl-anthraniloyl)-3′-deoxy-ATP], TNP-ATP (2′,3′-O-trinitrophenyl-ATP), TNP-GTP (2′,3′-O-trinitrophenyl-GTP), TNP-CTP (2′,3′-O-trinitrophenyl-CTP), TNP-UTP (2′,3′-O-trinitrophenyl-UTP), MANT-AppNHp [2′,3′-O-(N-methyl-anthraniloyl)-adenosine-5′-[(α,β)-imido]triphosphate], MANT-GppNHp [2′,3′-O-(N-methyl-anthraniloyl)-guanosine-5′-[(α,β)-imido]triphosphate], MANT-ATPγS [2′,3′-O-(N-methyl-anthraniloyl)-adenosine-5′-(thio)-triphosphate], MANT-GTPγS [2′,3′-O-(N-methyl-anthraniloyl)-guanosine-5′-(thio)-triphosphate], and MANT-ITPγS [2′,3′-O-(N-methyl-anthraniloyl)-inosine-5′-(thio)-triphosphate] were obtained from Jena Bioscience (Jena, Germany). MANT-ITP [2′,3′-(N-methyl-anthraniloyl)-ITP], MANT-CTP [2′,3′-(N-methyl-anthraniloyl)-CTP], MANT-UTP [2′,3′-(N-methyl-anthraniloyl)-UTP], and other (M)ANT-nucleotides were synthesized as described (Geduhn et al., 2011). cGMP (cGMP·Na) was supplied by Biolog (Bremen, Germany). Manganese chloride tetrahydrate, hydrochloric acid, and ammonium acetate were purchased from Fluka (Buchs, Germany). Acetonitrile, methanol, and water were supplied by Baker (Deventer, The Netherlands) and acetic acid was purchased from Riedel-de Haën (Seelze, Germany). Tenofovir was obtained through the National Institutes of Health AIDS Research and Reference Program, Division of AIDS (cat. no. 10199) (Bethesda, MD). Rat sGCα1β1 was purified as described (Beste et al., 2012).
The sGC assay was performed as described recently in detail (Beste et al., 2012). Briefly, assays contained 1 ng of sGCα1β1 per tube, 50 μM GTP, 3 mM MnCl2, 9 mM phosphocreatine, 2 mg/ml creatine kinase, 50 mM triethylamine, pH 7.5, 100 μM EGTA, 1 mM dithiothreitol, 1 mg/ml bovine serum albumin, 100 μM sodium nitroprusside, and 10–30,000 nM of inhibitors 1–38 and were incubated at 37°C. After 20 minutes, assays were stopped with heating at 95°C for 10 minutes. After cooling, mixtures were diluted with 97:3 (v/v) water/methanol solution containing 50 mM ammonium acetate, 0.1% (v/v) acetic acid, and 100 ng/ml tenofovir. Denatured protein was precipitated by centrifugation for 10 minutes at 20,000g. Supernatants were used for quantitation of generated cGMP using high-performance liquid chromatography–tandem mass spectrometry as described (Beste et al., 2012).
The recently published (Allerston et al., 2013) crystal structure of an sGCαcat/sGCβcat heterodimer [Protein Data Bank (PDB) ID 3UVJ] represents a catalytically inactive, “open” conformation and could therefore not be used directly for docking studies. However, several structures of mAC VC1:IIC2 in complex with (M)ANT- and TNP-nucleotides (PDB IDs 1TL7, 1U0H, 2GVZ, 3G82, and 2GVD) (Mou et al., 2005, 2006; Hübner et al., 2011) may serve as templates for the generation of a “partially closed” sGC conformation that binds this class of inhibitors. mAC VC1:IIC2 in complex with 3′-MANT-GTP (PDB ID 1TL7) was selected because of the highest resolution (2.8 Å). The subunits sGCαcat and sGCβcat were provided with missing side chains, with hydrogens and separately fitted to mAC VC1 and IIC2, respectively, by alignment of the backbones of corresponding amino acids being within 3.5 Å around 3′-MANT-GTP in the mAC structure (10 residues in each subunit). This initial sGCαcat/sGCβcat model was still not able to bind (M)ANT- and TNP-nucleotides because of steric hindrance by the sGCαcat β2-β3 loop. Therefore, the segment 524-KVETIGDAY-532 was individually adapted to the corresponding mAC VC1 segment 434-RIKILGDCY-442. The (M)ANT-nucleotide binding sites of the resulting sGCcat model and the mAC template fit well (root-mean-square distance of the backbones of the 20 site residues, 0.85 Å). 2′-MANT-3′-dATP, TNP-GTP and bis-MANT-ITP were manually docked into the sGCcat model based on the binding modes of 3′-MANT-GTP and TNP-GTP in the mAC crystal structures 1TL7 and 2GVD, respectively, and on a previous model of mAC in complex with bis-Br-ANT-ITP (Geduhn et al., 2011). The positions of two Mn2+ ions were also adopted from these templates. For 2′-MANT-3′-dATP and bis-Br-ANT-ITP, appropriate docking poses were only obtained with a 3′-endo conformation of the ribosyl moiety. Charges were assigned to the models (proteins and water molecules—AMBER_FF99, ligands—Gasteiger-Hückel). Mn2+ ions received formal charges of 2. Each complex was refined in a stepwise approach. First, 25 minimization cycles with fixed ligand using the AMBER_FF99 force field (Cornell et al., 1995) (steepest descent method); second, 100 minimization cycles of the ligand and the surrounding (distance up to 6 Å) protein residues (Tripos force field) (Clark et al., 1989); third, minimization with fixed ligand (AMBER_FF99 force field, Powell conjugate gradient); and fourth, minimization with fixed protein (Tripos force field) until a root-mean-square force of 0.01 kcal/mol × Å–1 was approached in both latter cases. To avoid overestimation of electrostatic interactions, a distance-dependent dielectric constant of 4 was applied. Modeling, docking, and preparation of figures was performed with the molecular modeling package SYBYL 8.0 (Tripos, Inc., St. Louis, MO) on an Octane workstation (Silicon Graphics International, Fremont, CA). Molecular surfaces and lipophilic potentials (protein variant with the new Crippen parameter table; Heiden et al., 1993; Ghose et al., 1998) were calculated and visualized by the program MOLCAD (MOLCAD, Darmstadt, Germany) contained within SYBYL.
Data are based on four to nine independent experiments and presented as Ki ± S.E. and pKi (–log Ki). GraphPad Prism software version 5.01 (San Diego, CA) was applied for nonlinear regression and calculation of Ki, pKi, and S.E. using the GTP concentration of 50 μM and an Km value of GTP of 14.2 μM as constants.
Inhibition of sGC by (M)ANT- and TNP-Nucleotides.
As expected from previous studies with bovine sGC (Gille et al., 2004; Suryanarayana et al., 2009), (M)ANT-nucleotides were potent inhibitors of recombinant rat sGCα1β1 (Table 1). Bovine sGC was studied with a much more limited inhibitor set than rat sGCα1β1 (Table 1). However, the pKi values for this set of inhibitors studied at both sGC isoforms correlated well (Fig. 2A). In contrast, despite the structural similarities between sGCα1β1 and mAC VC1:IIC2, inhibitor potencies at the two enzymes were only poorly correlated (Fig. 2B). In general, inhibitors were more potent at mAC VC1:IIC2 than at sGCα1β1, MANT-xanthosine 5′-triphosphate (XTP) (4), 2′-MANT-3′-dATP (7), TNP-ATP (22) and TNP-GTP (23) being notable exceptions. The correlations of pKi values for sGCα1β1 versus mAC2 (Fig. 2C), mAC5 (Fig. 2D) and mAC1 inhibition, respectively, were not strong. Moderate sGCα1β1 selectivity was observed in the case of MANT-XTP (4), 2′-MANT-3′-dATP (7), Cl-ANT-ATP (13, only mAC2), AcNH-ANT-ATP (18), TNP-GTP (23) and bis-Pr-ANT-ATP (35). By contrast, all ITP derivatives showed high mAC selectivity due to low inhibitory activity in the sGCα1β1 assay.
At sGCα1β1, adenine and xanthine were the preferred bases when (M)ANT-nucleotides were considered (1, 5 → 2-4, 6; 9 → 10, 11; 13 → 14). However, among bis-(M)ANT-nucleotides (26–38) this preference for adenine was not evident anymore. The triphosphate chain was important for high inhibitor potency (27 → 29 → 30). sGCα1β1 did not show striking preference for purines relative to pyrimidines (1–6; 22–25; 26–28). Removal of the 3′-OH group from the ribosyl moiety, fixing the (M)ANT group in the 2′-position, increased the potency of MANT-ATP about 10-fold (1 → 7), whereas removal of the 2′-OH group, fixing the (M)ANT group in the 3′-position, decreased the potency more than 4-fold. Presumably, both effects are not simply due to the isomerization of the (M)ANT substituent in MANT-ATP (eutomer-distomer ratio ≤2 in this case), but should be based on a direct unfavorable or favorable influence of the 2′- and the 3′-OH group, respectively. Substitution of the triphosphate chain by a bulkier γ-thiophosphate chain (1 → 9, 2 → 10, 3 → 11) had no major impact on inhibitor potency. Exchange of triphosphate against β,γ-imidiodiphosphate did moderately decrease potency in case of guanine (3 → 21) but not in case of adenine (1 → 20). Substitution of (M)ANT by the smaller ANT group (1 → 12) had little effect. Introduction of 5-Cl into ANT increased the affinity in case of adenine about 4-fold (1 → 13), whereas larger substitutions in 5-position (propyl, 16) had no effect or even decreased potency (AcNH2, 18). In case of hypoxanthine, 5-substituted ANT groups (Cl, 14; Br, 15; Pr, 17; or AcNH2, 19) did not yield inhibitors with higher affinity. Compared with the corresponding (M)ANT-nucleotides, TNP-nucleotides were considerably more potent inhibitors of sGCα1β1 (1 → 22; 3 → 23; 5 → 24; 6 → 25). In case of (M)ANT-nucleotides, adenine was preferred relative to guanine (1 → 3), whereas the opposite was true for TNP-nucleotides (22 → 23). Introduction of a second (substituted) (M)ANT group had no major impact on inhibitor affinity or slightly decreased inhibitor affinity in several cases (1 → 26; 2 → 27; 5 → 28; 14 → 32; 15 → 34; 17 → 36; 19 → 38). In some cases with adenine (16 → 35; 18 →37), the second substituted (M)ANT group led to a more pronounced decrease of inhibitor affinity.
Docking Studies of Nucleotide Inhibitors to an sGCα1β1 Model.
Based on the recently published (Allerston et al., 2013) crystal structure of an sGCαcat/sGCβcat heterodimer in a catalytically inactive, “open” conformation (PDB ID 3UVJ), a model of a (M)ANT-nucleotide binding state was generated using the complex of mAC VC1:IIC5 with 3′-MANT-GTP (PDB ID 1TL7) as template (Mou et al., 2005). When bound to mAC VC1:IIC5, the (M)ANT group acts as a wedge between β1-α1-α2 of VC1 and β7′-β8′ of IIC2, preventing the formation of the “fully closed”, catalytically active conformation (Mou et al., 2005). It is very reasonable to suggest similar binding modes of (M)ANT nucleotides to mAC and sGC by the following reasons: 1) Without the hypoxanthine derivatives with explicit causes of mAC selectivity [hydrogen bonds of the base with Lys938 and Asp1018 (Hübner et al., 2011), replaced by Glu473 and Cys541, respectively, in sGCβ], the correlations of sGC and mAC pKi values (Fig. 2) substantially increase (sGC versus mAC1: r, 0.62, versus mAC2: r, 0.68, mAC5: r, 0.67). 2) The pKi scales of the mAC and sGC assays are similar even in the case of more bulky bis-(M)ANT derivatives; there is no “selectively inactive” compound. 3) Among the 20 amino acids within 3.5 Å around 3′-MANT-GTP in the mAC VC1:IIC5 structure, only eight are different in sGC (mostly conservative substitutions, VC1 Ala404 versus sGCα Cys494, Ala409 versus Pro499, Val413 versus Ile503, Leu438 versus Ile528, IIC2 Lys938 versus sGCβ Glu473, Asp1018 versus Ser541, Ile1019 versus Leu542, and Trp1020 versus Phe543).
Based on the fit of these 20 amino acids, an sGCαcat/sGCβcat model resulted that differs from the crystal structure of the inactive state mainly by a 27-degree rigid body rotation and a translocation of the α subunit (Fig. 3A). The close analogy with mAC VC1:IIC5 and in particular of the corresponding (M)ANT-nucleotide binding sites become obvious (Fig. 3B). A model of the fully closed sGC state generated by Allerston et al. (2013) using mAC VC1:IIC5 in complex with ddATP as template (Tesmer et al., 1997) is similar to our model. In this case, the best fit with catalytically active mAC was obtained by a translocation and 26-degree rotation of sGCα.
Figure 3 shows the pseudosymmetric quaternary structure of (M)ANT-nucleotide–bound sGCαcat/sGCβcat. The active site is occupied by 2′-MANT-3′-dATP. A mirror-image site, binding forskolin in various mAC crystal structures (Fig. 3B), is potentially suited to interact with allosteric modulators that may be nucleotide derivatives, too (Beste et al., 2012). Both sites together form a deep and spacious cavity between the α- and the β-subunit. The (M)ANT-nucleotide binding site consists of amino acids of the β1-α1-α2 motif and the β2-β3 loop of sGCα as well as of the β1 strand, the β2-β3 loop, the β5 strand, the α4 helix, and the β6-β7 loop of sGCβ. In detail, the (M)ANT substituent interacts with sGCα α1-α2 and sGCβ α4, the ribosyltriphosphate moiety with sGCα β1-α1, sGCα β2-β3, sGCβ α4, and sGCβ β6-β7, and the nucleotide base with sGCα β2-β3, sGCβ β1, sGCβ β2-β3, sGCβ β5 and sGCβ α4. Two Mn2+ ions were transferred from the mAC template. They coordinate the triphosphate chain and two aspartates in sGCα β1 and β2-β3, respectively.
Figure 4 represents the putative binding modes of three potent sGC inhibitors of different types from Table 1, 2′-MANT-3′-dATP (7, Fig. 4A), TNP-GTP (23, Fig. 4B), and bis-MANT-ITP (27, Fig. 4C). Since the docking poses of the triphosphate groups were derived from the mAC templates (Mou et al., 2005, 2006; Hübner et al., 2011), the given conformations are only exemplary. The great flexibility of the triphosphate chain, of the surrounding residues, and of the Mn2+ positions enable different fits with, however, conserved interactions. Each Mn2+ ion coordinates sGCα Asp486 and Asp530 as well as two of the phosphate groups (α- plus β-, and β- plus γ-phosphate, respectively). One Mn2+ is additionally coordinated with the backbone oxygen of sGCα Ile487 (not shown). The phosphate groups form salt bridges with sGCβ Arg552 and Lys593. Apart from putative hydrogen bonds of the ring oxygen with the backbone-NH of Asp530 (not shown), the ribosyl moieties of the three ligands are not involved in specific interactions with sGC but in hydrophobic contacts with sGCα Phe490.
The low base selectivity of sGC (see Table 1, 2–6 and 22–25) may be explained by the spacious, rather unspecific binding pocket between sGCα and sGCβ, which is mainly hydrophobic (Fig. 4D) due to residues sGCβ Phe424, Leu542, Val547, and Phe543. Singular value decomposition analysis of Ki values of (M)ANT- and TNP-nucleotides indicated that the base contributed less to mAC binding compared with the triphosphate and the (M)ANT or TNP groups (Mou et al., 2006). ATP selectivity of mAC VC1:IIC2 obviously results from hydrogen bonds of the adenine base with Lys938 and Asp1018 (Tesmer et al., 1999). The same residues account for the hypoxanthine selectivity in the case of (M)ANT-nucleotides (Hübner et al., 2011). Replacement by sGCβ Glu473 and Cys541 is suggested to determine the GTP selectivity of sGC by hydrogen bonds with two nitrogens (N1, N6) and the oxygen, respectively, of guanine (Sunahara et al., 1998; Allerston et al., 2013). However, both residues do not form specific interactions with the bases in our models, indicating that the binding modes of nucleotides are more or less different from those of their (M)ANT and TNP derivatives. Such differences were also observed for mAC binding of ATP and 3′-MANT-ATP, which does not interact with Lys938 and Asp1018 and holds the adenine ring in a syn orientation with respect to the ribosyl moiety (Mou et al., 2005). In our models, only the adenine base of 2′-MANT-3′-dATP may form a specific interaction, namely a hydrogen bond of the 6-NH2 group with the side chain oxygen of sGCβ Thr474. All other interactions of the bases with sGC are hydrophobic and/or of the van der Waals type.
Derived from the binding modes of 3′-MANT-ATP and 3′-MANT-GTP to mAC VC1:IIC2 (Mou et al., 2005, 2006), the 2′-MANT groups of 2′-MANT-3′-dATP and bis-MANT-ITP, as well as the TNP group of TNP-GTP are suggested to act as wedges between α1-α2 of sGCα on one and α4 as well as β6-β7 of sGCβ on the other side (Fig. 3). With a 3′-MANT substituent, the ribosyl moiety exists in a 2′-endo envelope conformation (Mou et al., 2005), leading to an equatorial orientation of the 2-hydroxyl incompatible with binding of a 2′-MANT substituent (strong sterical clashes with mAC VC1). However, as in the case of mAC, docking of 2′-MANT-NTP to sGC is possible if the ribosyl moiety adopts a 3′-endo envelope conformation enabling an axial 2′- and an equatorial 3′-substituent. Thereby the 2′-MANT groups of 2′-MANT-3′-dATP and bis-MANT-ITP are aligned with the 3′-MANT moiety of mAC-bound 3′-MANT-ATP and -GTP. Moreover, equatorial orientation of the 3′-MANT group in bis-MANT-ITP facilitates interaction with an additional binding pocket between α1 of sGCα and α4 as well as β6-β7 of sGCβ (Fig. 4C).
The 2-MANT substituents and the corresponding TNP group interact with sGCα Phe490, Thr491, Cys494, Pro499, Val502, Ile503, as well as with sGCβ Phe543, Asn545, and Asn548. Both asparagine side chains are potential donors in hydrogen bonds either with the 2′-MANT nitrogen (Fig. 4A) or with one of the ortho nitro substituents in TNP derivatives (Fig. 4B). These interactions may be a reason for the substantially higher potency of 2′-MANT-3′-dATP compared with its positional isomer 3′-MANT-2′-dATP (Table 1), because corresponding hydrogen bonds with Asn1025 and Asn1028 are not possible in the structure of mAC VC1:IIC2 complexed with 3′-MANT-ATP (Mou et al., 2005). The docking mode of 2′-MANT-3′-dATP requires a 3′-desoxy position since a 3′-hydroxyl would clash with the side chain of Asn548; i.e., 2′-MANT-ATP, one of the isomers of 1, is less potent than 2′-MANT-3′-dATP. By contrast, the effect of a 2′-hydroxyl seems to be favorable in both sGC and mACs (compare 1 and 8) although the structure of mAC VC1:IIC2 does not show direct 2′-OH protein interactions. Definite reasons for the common sGC versus mAC VC1:IIC2 selectivity of TNP derivatives cannot be derived from the putative docking mode of TNP-GTP (Fig. 4B) since the suggested NO2-Asn hydrogen bonds are generally possible in the case of mAC, too, but not present in the corresponding crystal structure (Mou et al., 2005). In all of the mAC VC1:IIC2 structures in complex with (M)ANT- and TNP-nucleotides, internal water molecules apparently do not contribute to binding of the substituents.
Phe490 of sGCα, corresponding to Phe400 of mAC VC1, seems to play a key role in the binding modes of the three ligands in Fig. 4. Its phenyl ring may interact face to edge with the 2′-MANT group of 2′-MANT-3′-dATP and bis-MANT-ITP and align with the TNP moiety of TNP-GTP. Contacts with the ribosyl functions of all ligands and with the 3′-MANT group of bis-MANT-ITP are present, too. Phe490 forms something like a hydrophobic edge at the entrance of the 2′-MANT binding pocket (Fig. 4D). The 3′-MANT group of bis-MANT-ITP is in additional hydrophobic and/or van der Waals contact with sGCα Thr491 and Cys494 as well as with sGCβ Asn548 and Lys593. The side chain of Arg552 (sGCβ) may form a hydrogen bond with the aminomethyl substituent.
In this study, we have identified a number of potent inhibitors of sGCα1β1, 2′-MANT-3′-dATP (7) and TNP-GTP (23) being the most notable compounds (Table 1). These compounds may be useful for future crystallographic and fluorescence spectroscopy studies with sGCα1β1. By analogy to previous studies with mAC VC1:IIC2 (Mou et al., 2005, 2006; Pinto et al., 2009, 2011; Suryanarayana et al., 2009), such biophysical studies with sGCα1β1 are expected to provide major insights into the molecular mechanisms underlying sGC regulation by NO, sGC activators, and sGC stimulators (Daiber et al., 2010; Derbyshire and Marletta, 2012; Follmann et al., 2013; Stasch and Evegenov, 2013). In fact, Busker et al. (2014) has recently reported on the application of (M)ANT-nucleotides for monitoring conformational changes in sGC upon activation. While most compounds examined here exhibited lower potency at sGCα1β1 than at mAC VC1:IIC2 and mACs 1, 2, 5 (Table 1), the striking preference of sGCα1β1 for 2′-MANT-3′-dATP points to some unique structural features in the latter enzyme, indicating that development of potent and selective sGC inhibitors is feasible. In the case of Bacillus anthracis edema factor AC toxin, defined 2′- and 3′-isomers of (M)ANT-nucleotides turned out to be exceptionally valuable tools for probing conformational states (Seifert and Dove, 2013). Due to the base preference of sGCα1β1 among (M)ANT-nucleotides (Table 1), 2′-MANT-3′-dXTP is predicted to constitute a particularly potent sGCα1β1 inhibitor.
Our present sGCα1β1 inhibitor data should also be viewed from the perspective of mAC inhibitors. Specifically, it has been suggested that selective AC5 inhibitors may be valuable drugs for the treatment of heart failure (Vatner et al., 2013). However, it is difficult to develop highly selective AC5 inhibitors (Braeunig et al., 2013; Brand et al., 2013; Seifert, 2014), and in the context of heart failure, additional inhibition of sGC would be detrimental for cardiovascular function (Schermuly et al., 2011; Gheorghiade et al., 2013). Rather, sGC activation would be desirable (Schermuly et al., 2011; Gheorghiade et al., 2013). In principle, the development of potent AC5 inhibitors with selectivity relative to sGC is feasible, MANT-ITP (2, >400-fold selectivity) and Cl-ANT-ITP (14, >600-fold selectivity) providing good starting points (Table 1). This is a very promising result for future research in this field, since in the present study, we investigated only a small set of compounds (<40) and, nonetheless, achieved selectivity for AC5 relative to sGCα1β1. Similarly, selectivity for ACs 1 and 2 relative to sGCα1β1 is possible (Fig. 2; Table 1). Selectivity of inhibitors among various mAC isoforms is probably more difficult to achieve (Seifert et al., 2012; Brand et al., 2013; Braeunig et al., 2013; Seifert, 2014).
The fact that not only purine- but also pyrimidine-nucleotides inhibited sGCα1β1 was not surprising in view of the finding that UTP and CTP are also substrates for this enzyme both at the level of the purified enzyme and at the intact cell level (Beste et al., 2012; Bähre et al., 2014). Purine- and pyrimidine-nucleotides are also quite potent mAC inhibitors (Gille et al., 2004; Mou et al., 2005, 2006; Suryanarayana et al., 2009; Hübner et al., 2011; Pinto et al., 2011), but currently it is not known whether mACs also accept UTP and CTP as substrates. This will be an important research topic for future studies, considering the fact that mammalian cells endogenously contain substantial concentrations of 3′,5′-cyclic CMP and 3′,5′-cyclic UMP and that sGC produces these cyclic pyrimidine-nucleotides in intact cells (Beste et al., 2013; Bähre et al., 2013).
In a recent study, we have identified bis-Cl-ANT-ATP as a highly potent inhibitor of Bordetella pertussis CyaA AC toxin with >100-fold selectivity relative to mACs 1, 2, 5 (Geduhn et al., 2011). Bis-Cl-ANT-ATP is a valuable starting point for the development of CyaA inhibitors that are of potential value in the treatment of whooping cough (Seifert and Dove, 2012). Selectivity of bis-Cl-ANT-ATP for CyaA relative for sGCα1β1 is also quite good (>30-fold), and it is possible that introduction of bases other than adenine or hypoxanthine further improves CyaA-selectivity. In contrast to the situation with CyaA, we have not observed a situation for sGCα1β1 in which introduction of a second (M)ANT substitutent into the inhibitor increased affinity, indicating a more constrained catalytic site in sGCα1β1 than in CyaA.
TNP-nucleotides are similarly potent sGC and mAC inhibitors (Table 1). Thus, in terms of nucleotidyl cyclase isoform-selectivity, the TNP group is not favorable. However, TNP-nucleotides monitor different conformational changes in mACs than (M)ANT-nucleotides (Suryanarayana et al., 2009). Considering the high affinity of sGCα1β1 for TNP-CTP and TNP-UTP, these nucleotides may become particularly valuable tools for answering the question why UTP and CTP are good sGC substrates in the presence of Mn2+ but not in the presence of Mg2+ (Beste et al., 2012).
Although, in general, both sGC and mACs exhibit broad-base promiscuity (Table 1) (Gille et al., 2004; Mou et al., 2005, 2006; Suryanarayana et al., 2009; Hübner et al., 2011; Pinto et al., 2011), there are some striking differences, specifically with respect to purine bases. In the case of mACs, hypoxanthine is the preferred base, and accordingly MANT-ITP is the most potent mAC inhibitor known so far (Hübner et al., 2011). In contrast, MANT-XTP exhibits only low affinity for mAC (Mou et al., 2005; Hübner et al., 2011), whereas in case of sGC, xanthine is preferred relative to hypoxanthine. This differential inhibition of nucleotidyl cyclases by nucleotides possessing different purine bases will be important for future studies on isoform selectivity of inhibitors.
The major sGC isoform in lung is sGCα1β1 (Koesling et al., 1990; Harteneck et al., 1991). Conventionally purified bovine lung sGCα1β1 (Humbert et al., 1990) and recombinant rat sGCα1β1 (Hoenicka et al., 1999) exhibit a similar pharmacological inhibition profile (Table 1). Therefore, it appears that in the case of sGC, species selectivity of inhibitors is not a major issue. In contrast, other signal transduction proteins, bovine and rat, specifically histamine receptors, show striking species selectivity (Strasser et al., 2013). To this end, species specificity of mAC inhibitors has not yet been studied, but it should be noted that the most commonly studied mACs, i.e., mACs 1, 2, 5, were cloned from different species (Gille et al., 2004).
In contrast to sGC, the atrial natriuretic peptide–stimulated particulate guanylyl cyclase (pGC)-A does not accept UTP or CTP as substrates (Beste et al., 2013). Based on this differential substrate specificity it may also become possible to develop selective pGC inhibitors. To the best of our knowledge (M)ANT-nucleotides have not yet been examined at pGC. However, such studies are worthwhile, and we expect a different inhibitor profile for pGC compared with sGC and mACs. First steps toward the development of selective allosteric pGC modulators have already been taken (Robinson and Potter, 2012; Seifert and Beste, 2012). Identification of high-potency pGC-A inhibitors may support biophysical studies with this enzyme.
In conclusion, here, we have presented a comprehensive structure/activity relationship study for sGC inhibitors. We have obtained several potent sGC inhibitors that could be used as experimental tools for fluorescence spectroscopy and crystallographic studies. Docking of representative inhibitors into an sGCαcat/sGCβcat model has provided binding modes accounting for the high affinity of derivatives from different structural classes and suggesting some possible reasons for their sGC-versus-mAC selectivity. Based on our data, the rational development of even more potent sGC inhibitors is possible. With respect to treatment of cardiovascular diseases, inhibition of mAC5 may be desirable, whereas sGC inhibition is not. Our present study also provides guidelines for the development of potent mAC5 inhibitors with low affinity for sGC. Furthermore, CyaA inhibitors may be valuable for the treatment of whooping cough, and again development of CyaA inhibitors with high selectivity relative to sGC is achievable. The crystallization of sGC with one of the inhibitors described in this study will further aid the development of nucleotidyl cyclase inhibitors that do not target sGC because low sGC activity is unfavorable. In other words, the catalytic site of sGC is an antitarget, and the present study characterized this antitarget at the molecular level.
The authors thank Prof. Dr. Burkhard König and Dr. Jens Geduhn for providing several (M)ANT-nucleotides and Annette Garbe for expert technical assistance.
Participated in research design: Dove, Danker, Seifert.
Contributed new reagents or analytic tools: Stasch, Kaever.
Performed data analysis: Dove, Danker, Seifert.
Wrote or contributed to the writing of the manuscript: Dove, Danker, Stasch, Kaever, Seifert.
- Received November 26, 2013.
- Accepted January 27, 2014.
This work was supported by the Deutsche Forschungsgemeinschaft (Research Training Group 1910; S.D. and R.S.).
S.D. and K.Y.D. contributed equally to this paper.
Institute of Cardiovascular Research, Bayer HealthCare, Wuppertal, Germany, Primary laboratory of origin: Institute of Pharmacology, Hannover Medical School.
- membranous adenylyl cyclase
- nitric oxide
- Protein Data Bank
- particulate guanylyl cyclase
- soluble guanylyl cyclase
- xanthosine 5′-triphosphate
- Copyright © 2014 by The American Society for Pharmacology and Experimental Therapeutics