Structure of Cinaciguat (BAY 58–2667) Bound to Nostoc H-NOX Domain Reveals Insights into Heme-mimetic Activation of the Soluble Guanylyl Cyclase*

Heme is a vital molecule for all life forms with heme being capable of assisting in catalysis, binding ligands, and undergoing redox changes. Heme-related dysfunction can lead to cardiovascular diseases with the oxidation of the heme of soluble guanylyl cyclase (sGC) critically implicated in some of these cardiovascular diseases. sGC, the main nitric oxide (NO) receptor, stimulates second messenger cGMP production, whereas reactive oxygen species are known to scavenge NO and oxidize/inactivate the heme leading to sGC degradation. This vulnerability of NO-heme signaling to oxidative stress led to the discovery of an NO-independent activator of sGC, cinaciguat (BAY 58–2667), which is a candidate drug in clinical trials to treat acute decompensated heart failure. Here, we present crystallographic and mutagenesis data that reveal the mode of action of BAY 58–2667. The 2.3-Å resolution structure of BAY 58–2667 bound to a heme NO and oxygen binding domain (H-NOX) from Nostoc homologous to that of sGC reveals that the trifurcated BAY 58–2667 molecule has displaced the heme and acts as a heme mimetic. Carboxylate groups of BAY 58–2667 make interactions similar to the heme-propionate groups, whereas its hydrophobic phenyl ring linker folds up within the heme cavity in a planar-like fashion. BAY 58–2667 binding causes a rotation of the αF helix away from the heme pocket, as this helix is normally held in place via the inhibitory His105–heme covalent bond. The structure provides insights into how BAY 58–2667 binds and activates sGC to rescue heme-NO dysfunction in cardiovascular diseases.

Heme is a key evolutionarily conserved cofactor involved in many processes ranging from signal transduction, gas transport, circadian rhythm, microRNA processing, and drug metabolism (1)(2)(3). Heme, as part of NO signaling through sGC, 2 is critical for regulation of cardiovascular processes via cGMP-dependent pathways, making sGC a prime drug target in cardiovascular diseases (4). sGC is a heterodimeric protein, mostly in the ␣1␤1 isoform, that stimulates the production of cGMP upon NO activation (5). Under normal physiological conditions, NO-activated sGC stimulates vasodilation and is required for platelet disaggregation as well as neurotransmission (6). During the pathophysiological conditions of oxidative stress common to vascular disease, the NO-sGC signaling pathway is disrupted due to sGC desensitization and degradation of sGC upon heme loss, thus limiting the extent to which traditional treatment methods such as nitroglycerin can increase sGC activity (7). BAY 58 -2667 can overcome these defects by activating sGC and protecting heme-oxidized sGC from proteasomal degradation (4,8). These beneficial characteristics contributed to the pharmaceutical potential of BAY 58 -2667, and this compound is in clinical trials to treat acute decompensated heart failure (9 -11). The initial steps of BAY 58 -2667 activation of sGC are fundamentally distinct from activation by NO (Fig. 1). NO binds a 5-coordinated His 105 -liganded heme, forming a brief 6-coordinated intermediate and subsequently causing breakage of the Fe-His 105 bond to yield a 5-coordinated NO-bound heme in the activated state (Fig. 1B). In contrast, BAY 58 -2667 activation of sGC is heme-independent yet takes place in the NO sensory H-NOX domain of sGC when it is heme-depleted ( Fig. 1C) (12). This crucial H-NOX domain is the N-terminal domain of the sGC␤1 subunit, which is adjacent to a Per-Armt-Sim (PAS-like) H-NOXA domain, followed by a coiled-coil domain, and the C-terminal catalytic guanylyl cyclase domain. The sGC␣1 subunit has a similar domain arrangement, except that its N-terminal domain does not bind heme. sGC has not been amendable to crystallization, and structural studies of sGC have therefore been limited to individual domains of homologous proteins (13)(14)(15)(16). The H-NOX domain used in these studies that is most homologous to the H-NOX domain of sGC is Nostoc sp. H-NOX (Ns H-NOX), which shares 35% sequence identity with sGC, having similar properties including its ability to bind CO and NO (supplemental Fig. 1) (13). We present here the 2.3-Å crystal structure of the Ns H-NOX⅐BAY 58 -2667 complex and mutational results revealing insights into the molecular mechanisms of sGC activation by a heme mimetic.

EXPERIMENTAL PROCEDURES
Preparation of the BAY 58 -2667⅐H-NOX Complex-We expressed and purified a C-terminally truncated version of Nostoc sp. H-NOX comprising residues 1-183 similarly as described previously for the full-length 1-187 protein (13). The heme was replaced by BAY 58 -2667 by adding a 10-fold molar excess of the heme oxidizer NS-2028 (Alexis Biochemicals) and 5-fold molar excess of BAY 58 -2667 (obtained from Dr J. P. Stasch, Bayer Schering Pharma AG) at 37°C prior to Superdex 75 chromatography to remove the displaced heme and unbound BAY 58 -2667.
Crystallization and Structure Determination-Colorless crystals of the Ns HNOX domain bound to BAY 58 -2667 were obtained using sitting drop crystallization at room temperature with a protein concentration of ϳ10 mg/ml and a well solution of 1.8 M sodium malonate at pH 7.3. Crystals were cryoprotected in 3.0 M sodium malonate, pH 7.3, prior to dunking the crystal in liquid nitrogen. Data were collected at the Stanford Synchrotron Radiation Lightsource beamline 11-1 to 2.3-Å resolution and processed using HKL2000 (17). Crystals of BAY 58 -2667 bound HNOX were in the same space group as for the heme-bound protein with two molecules in the asymmetric unit. Twinning analysis revealed a twinning fraction of close to 0.5 (18), which was refined in REFMAC (19) using the amplitude-based twin refinement with the H-NOX coordinates, without the heme, as the starting model (Protein Data Bank code 2O09; 13). The structure was subsequently refined using alternating cycles of fitting using COOT (20) and REFMAC. Heme density was absent, yet strong density for the two copies of BAY 58 -2667 was present after the initial refinement. Subsequently, two BAY 58 -2667 molecules were added in refinement using a stereochemistry library file that was generated with PRODRG (21). The structure was refined to a final R-factor of 0.150 with an R free of 0.193 and validated using PROCHECK (22). Data collection and refinement statistics are listed in Table 1. Figures are generated using PyMOL.
Expression in COS-7 Cells of WT and Mutant sGC-COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum, penicillin, and streptomycin (100 units/ml). Cells were transfected with SuperFect TM reagent using the protocol of the supplier (Qiagen). Cells (100-mm dish) were transfected with 10 g of plasmid encoding each wild-type or mutated subunit. After 40 h, cells were washed twice with 5 ml of phosphate-buffered saline. Cytosolic Preparation-After washes, cells are scraped off the plate in cold lysis buffer: phosphate-buffered saline buffer contained protease inhibitors, 50 mM HEPES (pH 8.0), 1 mM EDTA, and 150 mM NaCl. Cells were broken by sonication (three pulses of 3 s). The resulting lysate was centrifuged at 16,000 ϫ g for 10 min at 4°C to collect the cytosol.
sGC Activity Assay-sGC activity was determined by formation of [␣-32 P]cGMP from [␣-32 P]GTP, as described previously (23). Reactions were performed for 5 min at 33°C in a final volume of 100 l, in a 50 mM HEPES, pH 8.0, reaction buffer containing 500 M GTP, 1 mM dithiothreitol, and 5 mM MgCl 2 . Typically, 40 g of cell cytosol was used in each assay reaction. Enzymatic activity was stimulated with the NO donor S-nitroso-N-acetyl-penicillamine (SNAP) (Calbiochem) at 100 M. sGC activity is expressed in pmol/min mg.

RESULTS
To obtain the BAY 58 -2667⅐Ns H-NOX complex, the heme of Ns H-NOX was first oxidized with the sGC-specific heme oxidizer NS-2028 (4H-8-bromo-1,2,4-oxadiazolo(3,4-d) benz(b)(1,4)oxazin-1-one); the heme was replaced by adding excess of BAY 58 -2667 (cinaciguat; 4-((4-carboxybutyl)(2-((4-phenethylbenzol)oxy) phenethyl) amino)methyl(benzoic) acid). Subsequent crystallization and structure determination yielded the 2.3-Å resolution crystal structure of Ns H-NOX bound to BAY 58 -2667 with two copies in the asymmetric unit (Table 1 and Fig. 2). Electron difference density in the heme-binding pockets of each copy clearly revealed the trifurcated BAY 58 -2667 ligand density (Fig. 2B). The two carboxylate groups of BAY 58 -2667 make extensive interactions: the carboxy-butyl moiety interacts with the side chains of Tyr 134 , Ser 136 , and Arg 138 , whereas the carboxylate moiety of the benzoic acid interacts with Arg 138 and the main chain nitrogen of Tyr 2 (Fig. 2C). Both copies of BAY 58 -2667 in the asymmetric unit have a similar binding mode, although one of the carboxylates is slightly rotated (supplemental Fig. 2). Electron density for the third moiety of BAY 58 -2667 extending from the tertiary amine, the extended tri-benzyl ring containing moiety, is well defined except for the terminal benzyl ring (Fig. 2B); the orientations for this terminal benzyl ring also vary slightly between the two crystallographic copies (supplemental Fig. 2). This hydrophobic third branch of BAY 58 -2667 makes extensive hydro-   (Fig. 2C). Finally, the trifurcated tertiary amine is 4.1 Å from Trp 71 , apparently forming a cation/pi interaction of the positively charged amine and the pi clouds of the tryptophan. Despite having a shape dissimilar to heme, the trifurcated BAY 58 -2667 folds up and occupies the heme cavity with its carboxylate groups superimposing on those of the heme propionate groups in the apo-Ns H-NOX structure (Fig. 3) published previously (13). BAY 58 -2667, therefore, mimics heme in that it contains two hydrophilic charged carboxylates that interact with the YxSxR motif found to be crucial for heme and BAY 58 -2667 binding (12,24), yet, on the other hand, it is mostly hydrophobic and can "fold" in a somewhat planar hemelike fashion (Fig. 3C). A key functional difference between the two molecules is that BAY 58 -2667 is a sGC activator, whereas heme is an inhibitor as the latter forms the covalent inhibitory bond with His 105 (that normally is broken via NO binding). The Ns H-NOX structure with BAY 58 -2667 in the activated state reveals a 3°rotation of the ␣F helix, normally tethered by the heme via His 105 , as analyzed by the program ESCET ( Fig. 4 and supplemental movie) (25). This rotation culminates in a ϳ1-Å shift of the carboxyl terminus of ␣F comprising residues 110 -111 (see supplemental movie). NO activation of sGC would likely induce a similar conformational change of ␣F as NOheme binding causes the breakage of the ␣F-His 105 -Fe bond (26) and would allow this helix to shift to alleviate the steric clashes between the heme and His 105 due to van der Waals repulsion interactions. This conformational change distorts the surface region comprising residues 110 -116 and nearby flanking residues 40 -46 (Fig. 4, A and B) and highlights a region that is likely to be used by H-NOX to transmit its activation signal to the rest of sGC upon NO or BAY 58 -2667 activation. We therefore probed the importance of this region for sGC activation via alanine-scanning mutagenesis of surface-exposed H-NOX residues Arg 40 , Ile 111 , and Arg 116 . WT and mutant versions of sGC were expressed in COS-7 cells, and their guanylyl cyclase activity was assayed (Fig. 5B), revealing that these residues affect NO-stimulated activity. Mutation of residues Ile 111 and Arg 116 have an ϳ4and 3-fold, respectively, effect on max- imal sGC activation, whereas mutation of residue Arg 40 has the largest effect on NO-stimulated and basal sGC activity (Fig. 5B). These results suggests that sGC residues Arg 40 , Ile 111 , and Arg 116 play a role in communicating H-NOX conformational changes to the rest of sGC and that there are likely some similarities between BAY 58 -2667 and NO activation of sGC.

DISCUSSION
Our results regarding the 2.3-Å resolution structure of BAY 58 -2667 bound to Ns H-NOX reveals the molecular details of the remarkable ability of BAY 58 -2667 to activate heme-depleted sGC. By being able to mimic the heme yet avoid forming a covalent inhibitory bond with His 105 , BAY 58 -2667 is able to bind to the heme pocket and yet allow the conformational activation change involving helix ␣F and its His 105 residue. The largest shifts of the BAY 58 -2667-bound Ns H-NOX structure compared with the heme-bound structure are in the C-terminal part of helix ␣F (Fig. 4). We postulate that this region, near residue 111, and the flanking region comprising residues 40 -46 are therefore likely involved in communicating the activation conformational changes of the H-NOX domain to the rest of sGC. The importance of this region for sGC activation was probed by mutagenesis and indicated that the ␤1 H-NOX mutations R40A, I111A, and R116A negatively affect NO-stimulated cyclase activity (Fig. 5). Because these residues are mostly solvent exposed and not involved in direct heme binding or domain folding, these residues are likely involved in direct H-NOX/GC interactions (such a direct H-NOX/GC interdomain interaction had been previously observed (27)). In earlier work, residue Asp 45 also was found to be important in sGC activation (28); this residue also falls within this surface region providing additional support for the proposed activation relevance of this region near residues 110 -116 and 40 -46. Future structural studies are needed to indeed confirm such a direct interaction between these sGC domain and also what the differences and similarities of the NO-bound 5-coordinated sGC structure is compared with BAY 58 -2667.
Structure-Activity Relationship of BAY 58 -2667-BAY 58 -2667 is the result of substantial structure-activity relationship optimization (9), and its critical features agree well with its H-NOX binding mode and induced conformational activation  The observed activation shift of helix ␣F (blue), including residues Ile 111 (red), would cause subsequent shifts in this region hypothesized to be sensed by the rest of sGC. Surface residues targeted for mutagenesis are shown in red, and the heme is shown in magenta sticks. B, basal (gray) and NO-stimulated (white) WT and sGC mutants expressed in COS-7 cells. Basal activities were similar between WT and mutants but not for R40A (for which basal activity was close to background values). NO-stimulated activity of cytosols showed that mutants have a reduced response to NO stimulation compared with WT. The NO donor SNAP was used at 100 M. Experiments were repeated three times, from two independent transfections, with each measurement done in duplicate and expressed in pmol/min⅐mg Ϯ S.E. changes (Fig. 6). A large number of BAY 58 -2668 derivatives have been generated and their likely structural consequence within the H-NOX binding pocket, and their effects on rabbit saphenous arteries, will be correlated next. First, the pivot point of the BAY 58 -2667 induced helix rotation of ␣F is near residue Leu 101 , a well conserved residue in sGC␤1. Leu 101 interacts with the phenyl ring of the phenethylamino moiety of BAY 58 -2667 (ring labeled 1 in Fig. 6), a group found to be key for BAY 58 -2267 activity (9). Second, this phenyl moiety is in a hydrophobic internal H-NOX region, which explains why ring nitrogens introduced to convert the phenyl (labeled 2 in Fig. 6) to a heterocycle cannot be accommodated without negatively impacting the IC 50 (9). Third, there is considerable variability possible with respect to the linker length of the terminal hydrophobic group being 4-phenethylbenzol in BAY 58 -2667 (red oval labeled 3 in Fig. 6). Having just one phenyl ring at this position yields a potent sGC activator, although filling up the H-NOX cavity via longer linkers and having a second linearly connected phenyl ring (labeled 3b in Fig. 6) improve the IC 50 values. In contrast, widening this linker at the nonterminal phenyl ring (labeled 3a in Fig. 6) or adding polar atoms to this linker moiety have negative effects likely due to the only narrow and hydrophobic space available to this moiety, which is flanked by Tyr 83 and its own benzoic acid moiety (labeled 5b in Fig. 6) (9). These observations are in accord with our structure, which shows flexibility in the terminal BAY 58 -2667 phenyl moiety and partial displacement of Phe 112 , suggesting that variations there can be accommodated with even longer variants such as BAY 60 -2770 (9). Fourth, being released from the covalent bond with the heme, the His 105 residue is pushed down ϳ0.6 -0.7 Å by the ethylamino moiety (labeled 4 in Fig. 6; Fig. 4). This latter group contains the trifurcated chiral branch atom of BAY 58 -2667, the correct chirality at this position, and therefore likely also how far His 105 is pushed away, has a major impact on its IC 50 (9). Finally, the BAY 58 -2667 structure also explains why the butyl-carboxylate (Fig. 6, 5a) is more important compared with the methyl-benzoic acid (Fig. 6, 5b) because replacing the latter with a hydrogen leads to only a 63-fold IC 50 increase, whereas replacing the former leads to a drastic 4800fold increase (9). The BAY 58 -2667 H-NOX structure shows that the butyl-carboxylate makes key interactions with all three of the pivotal YxSxR motif residues known to be crucial for BAY 58 -2667 activation (24,28) holding this H-NOX region in place, whereas the benzoic acid carboxylate only interacts with Arg 138 and the backbone nitrogen of Tyr 2 (Fig. 2C). The dominance and importance of one of the carboxylates explains the potency of recently developed sGC activators with only one carboxylate, as described in a recent U.S. patent application (29). In summary, this analysis shows that two of the most critical features of BAY 58 -2667 are its butyl-carboxylate moiety, which interacts with the YxSxR motif, and a long hydrophobic tail that can bend and occupy the mostly hydrophobic heme pocket. This analysis suggests that our BAY 58 -2667 complex structure is a great tool for analyzing current analogs and likely also future analogs of BAY 58 -2667.
In summary, the Ns H-NOX protein continues to be a great tool for understanding sGC. In addition to being the closest bacterial homolog to the H-NOX domain of sGC in terms of sequence and ligand binding (13), we show here that Ns H-NOX is, like sGC, capable of having its heme displaced by the sGC activator BAY 58 -2667 and that the observed binding mode of BAY 58 -2667 strongly correlates with the published sGC structure activity relationships (SAR) data on BAY 58 -2667 analogs (Fig. 6). In addition, the BAY 58 -2667 binding mode is in agreement and structurally explains previous mutagenesis results that pointed to the strong role for the YxSxR motif present in both sGC and Ns H-NOX (12,24). And finally, the BAY 58 -2667 induced shift of the His 105 -containing ␣F helix and adjacent loop point to the importance of the surface patch comprised of regions 110 -116 (and flanking residues 40 -46), which we subsequently probed with sGC mutagenesis experiments (Fig. 5). These combined results point to the importance of this surface region for sGC activation by BAY 58 -2667, which possibly shares some mechanistic similarities with NO activation of sGC. These cohesive results therefore aid in our understanding of sGC activation by a hememimetic and how BAY 58 -2667 protects sGC from degradation by stabilizing the heme cavity and offer new structurebased design strategies to pharmaceutically further target the sGC protein to treat certain cardiovascular diseases.