Structural insights into the mechanism of human soluble guanylate cyclase

Soluble guanylate cyclase (sGC) is the primary sensor of nitric oxide. It has a central role in nitric oxide signalling and has been implicated in many essential physiological processes and disease conditions. The binding of nitric oxide boosts the enzymatic activity of sGC. However, the mechanism by which nitric oxide activates the enzyme is unclear. Here we report the cryo-electron microscopy structures of the human sGCα1β1 heterodimer in different functional states. These structures revealed that the transducer module bridges the nitric oxide sensor module and the catalytic module. Binding of nitric oxide to the β1 haem-nitric oxide and oxygen binding (H-NOX) domain triggers the structural rearrangement of the sensor module and a conformational switch of the transducer module from bending to straightening. The resulting movement of the N termini of the catalytic domains drives structural changes within the catalytic module, which in turn boost the enzymatic activity of sGC. Cryo-electron microscopy structures of human soluble guanylate cyclase in inactive and activated states shed light on the activation mechanism of this enzyme by nitric oxide.

Nitric oxide (NO) is a gaseous signalling molecule that is involved in many important physiological processes, such as vasodilatation, neurotransmission, platelet aggregation, immunity, cell proliferation, and mitochondrial respiration 1,2 . The dysregulation of NO signalling has been linked to cardiovascular disease, sepsis, acute lung injury, and multiple organ failure 1,3,4 . NO signalling is initiated by the activation of NO synthase (NOS), which generates NO in response to physiological stimuli 4 . NO readily permeates target cell membranes, and after diffusing across the membrane, it binds and activates soluble guanylate cyclase (sGC), the primary NO acceptor 4 . sGC catalyses the cyclization reaction of guanosine triphosphate (GTP) to generate inorganic pyrophosphate and the secondary messenger cyclic guanosine monophosphate (cGMP) 5 . cGMP then acts on downstream effectors, including cGMP-regulated protein kinases, phosphodiesterases, and ion channels, to regulate physiological processes in the cell 5 . Genetic mutations of sGC in humans are associated with coronary artery disease 6 , moyamoya disease, achalasia, and hypertension 7,8 , and it is a validated drug target for the treatment of pulmonary hypertension and chronic heart failure 4 . The NO donor nitroglycerin has been widely used for centuries to alleviate angina pectoris, and the sGC stimulator riociguat has been approved for the treatment of pulmonary hypertension 4 . Drugs that activate or stimulate sGC also have therapeutic potential in fibrotic diseases, systemic sclerosis, chronic kidney diseases, neuroprotection, dementia, and sickle cell disease 9 .
sGC is a heterodimeric protein complex composed of one α-subunit and one β-subunit. In humans, the α1 and β1 subunits are widely expressed in many tissues, while the expression of α2 and β2 subunits is tissue-specific 5,10 . The αand β-subunits have some sequence homology and are similarly organized into modular domains, including an N-terminal H-NOX domain, a Per/Arnt/Sim (PAS) domain, a coiledcoil (CC) domain, and a C-terminal catalytic domain. The PAS and CC domains mediate protein-protein interactions, and the catalytic domain is responsible for enzymatic activity 5,11 . The H-NOX domain of the β-subunit contains a ferrous b-type haem prosthetic group that facilitates the high-affinity binding of NO 5,11 . Under pathological conditions or oxidative stress, the ferrous haem can be oxidized to ferric haem 12 , and haem-oxidized sGC has low activity even in the presence of NO 13 .
Several structures of isolated sGC domains have been solved by X-ray crystallography or NMR. These structures include the human β1 H-NOX domain (PDB ID: 5MNW), the Manduca sexta α-PAS domain 14 , the human β1 CC domain 15 , and the human α1β1 catalytic domain heterodimer 16,17 . Recent negative stain electron microscope studies 18 have revealed the general shape of the full-length mammalian sGC at a resolution of 25-40 Å, and hydrogen-deuterium exchange experiments mapped NO-induced structural changes onto the primary sequence of the full-length sGC 19 . Despite these pioneering structural efforts, the allosteric mechanism that underlies the activation of the distal catalytic domain in response to binding of NO to the β-subunit H-NOX domain remains unclear at the atomic level, mainly owing to the lack of high-resolution structural information on intact sGC in different functional states. Here, we have used cryo-electron microscopy (cryo-EM) to determine the structure of the human α1β1 sGC holoenzyme in both the inactive and NO-activated states at a resolution of 3.9 Å and 3.8 Å, respectively. We also obtained a 6.8 Å resolution cryo-EM map of the constitutively active β1(H105C) mutant. These structures uncover not only the detailed domain-domain interfaces, but also the activation mechanism of human sGC.
Article reSeArcH Videos 1,2). The qualities of composite maps obtained from multibody refinement were sufficient to trace the main chain of most residues with the aid of available high-resolution homologous structures (Extended Data Figs. 5, 6, Extended Data Table 1). We also obtained a 6.8 Å map for the β1(H105C) mutant (Extended Data Fig. 4g, h). At this resolution, the overall shape and domain organization of the β1(H105C) mutant were found to be similar to that of the NO-activated state, with a real space correlation of 0.96. However, the haem density is missing in the β1 H-NOX domain of the β1(H105C) mutant, as expected (Extended Data Fig. 4i). The atomic models of sGC in different states allowed us to characterize the domain-domain interfaces in detail (Extended Data Figs. 7,8).

Structure of sGC in the inactive state
Both the haem-unliganded and the haem-oxidized sGC were in a 'bent' conformation 18 ( Fig. 1c-e, Supplementary Video 3). In our cryo-EM reconstructions, we found that the overall structure of sGC in the haem-unliganded state (4 Å) is essentially the same as that in the haem-oxidized state (3.9 Å), with a root mean square deviation (r.m.s.d.) of only 0.28 Å (Extended Data Fig. 9a), in accordance with the functional data, which showed that the haem-unliganded and haem-oxidized states have low activity 13 (Fig. 1b). Therefore, both of the structures were considered as the inactive state, and the 3.9 Å haem-oxidized state is used in further discussion of the inactive state. The structure of the inactive sGC occupies a 3D space of 140 Å × 75 Å × 75 Å (Fig. 1c-e). The large N lobe is composed of α1 H-NOX, α1 PAS, β1 PAS, and β1 H-NOX domains. These domains are arranged in a pseudo-two-fold symmetric manner, with the scaffolding PAS domains at the centre and the H-NOX domains at the periphery (Fig. 1c). The H-NOX domains and PAS domains are essential for NO sensing and form the N-terminal sensor module of sGC (Fig. 1c). The CC domains of both subunits form the transducer module that bridges the N-terminal sensor module and the C-terminal catalytic module (Fig. 1d).
A haem molecule binds inside the β1 H-NOX domain, and its five-coordinated Fe ion is tightly bound to H105 of αF, as evidenced by the strong connecting density between them (Fig. 1f). By contrast, the α1 H-NOX domain does not bind haem owing to a sterical clash (Extended Data Fig. 9b). The structure of each PAS domain resembles that of the M. sexta sGC α-subunit 14 (PDB ID: 4GJ4; Extended Data The β1 H-NOX domain, especially the αE and αF helices, interacts with both the neighbouring PAS heterodimers and the transducer module (Extended Data Figs. 2g, 7a-e, Supplementary Note 1). These interactions further stabilize the transducer module in the bent conformation, in which both the α1 and β1 CC domains are broken into two short helices (αM and αN) connected by a near 90° turn (Extended Data Fig. 7f). The two αN helices pack in a 'leucine zippers' manner and interact extensively with the catalytic module (Extended Data Fig. 7f-h, Supplementary Note 2). In the catalytic module, the two subunits are organized in a pseudo-symmetric manner as well, but the angle between domains is different from that of the isolated catalytic module 17 (Extended Data Fig. 9e, f, Supplementary Note 3). Compared with the adenylate cyclase in the active state 22 (PDB ID: 1CJU), the structure of the catalytic module shows steric clashes between the substrate and the protein residues (Extended Data Fig. 9g). This suggests that the structure of sGC in the inactive state is incompatible with substrate binding, consistent with previous studies that showed that inactive sGC has a high Michaelis constant (K m ) 23 . The domain-domain interactions observed in the inactive state were further validated by cysteine cross-linking under oxidative conditions (Extended Data Fig. 7i-l, Supplementary Note 4).

Structure of sGC in NO-activated state
The NO-activated sGC has a dumbbell-shape extended structure 18 , in which the sensor module moves away from the catalytic module (Fig. 2, Supplementary Video 4). This is markedly different from the bent conformation of the inactive state. In addition, the overall structure of the constitutively active β1(H105C) mutant, in the absence of NO donors, shows a similar extended conformation (Extended Data Fig. 4g). This structural agreement suggests that this large overall conformational change is associated with enhanced enzymatic activity and the full activation of sGC, but probably does not result from the S-nitrosylation of sGC by NO, which is a covalent modification of cysteine residues that can lead to desensitization of sGC under certain conditions 24-26 (Extended Data Fig. 4j Article reSeArcH maintained (Figs. 1, 2). In the electron density map of the NO-activated state, the H105-Fe bond of β1 H-NOX is cleaved, as evidenced by the clear separation between each density (Fig. 2c). This suggests that the current conformation is likely to correspond to an NO-bound state, because excess NO donor DEA NONOate was added to the sample and NO binds sGC with picomolar-range high affinity 2 . However, we could not explicitly model the NO molecules or the haem deformation owing to the limited resolution. The binding of NO induces a conformational change in β1 H-NOX in which the C-terminal subdomain rotates relative to the N-terminal subdomain (Extended Data Fig. 9h). When αF (residues 96-107) of the β1 subunit was used as the reference to superimpose the structure of the NO-bound β1 H-NOX domain onto the structure of the inactive state, the Cα atom of N62 in the N-terminal subdomain was displaced by 4.6 Å (Extended Data Fig. 8a) and, more importantly, the NO-bound β1 H-NOX domain sterically clashed with the adjacent domains of the inactive state (Extended Data Fig. 8a). This indicates that the inactive state structure is incompatible with the NO-bound β1 H-NOX domain and, therefore, a structural rearrangement is required to accommodate the conformational change of the β1 H-NOX domain upon NO binding. Indeed, we observed structural changes within the sensor module in which α1 H-NOX underwent a small downward movement while β1 H-NOX underwent a large rotational and translational movement (Fig. 2d).
These conformational changes of the sensor module in the NO-activated state result in completely new interfaces between the NO-bound β1 H-NOX domain and its adjacent domains (Extended Data Figs. 4e, 8b, c, Supplementary Notes 6, 7). Many residues contribute to this new interface; among them, D106 on αF of the β1 H-NOX domain forms important polar interactions with other residues (Fig. 2e). We found that sGC with the β1(D106A) mutation had normal haem incorporation but impaired activation by NO (Fig. 2f, Extended Data Fig. 8d), suggesting that this interface is essential for sGC activity in the NO-activated state.
The markedly altered interactions between the CC domain and the sensor module lead to the conformational change of the transducer module (Fig. 3). Strikingly, the linkers between αM and αN fold into α-helical structures, which fuse αM with αN into the 71 Å-long αMN helices (Extended Data Fig. 8e). Specifically, R420-K426 of the α1 CC domain and L355-Q358 of the β1 CC domain fold into α-helical structures (Fig. 3a, Extended Data Figs. 5, 6). As a result, the transducer module switches from a highly bent conformation in the inactive state into a long, continuous coiled-coil structure in the NO-activated state (Fig. 3a). The folding of the αM-αN loops results in a decrease in the exchangeability of the main chain hydrogens as they form hydrogen bonds in α-helices. This is consistent with previous hydrogen-deuterium exchange mass spectrum results that showed that the αM-αN loops had a much slower exchange rate upon NO activation 19 . To determine the functional importance of this bending-straightening conformational change, we mutated residues in the αM-αN linker to either prolines or alanines. Prolines generate kinks in helical structures because they cannot form hydrogen bonds on the main chain. Therefore, proline mutations should destabilize the helical structures of αMNs in the NO-activated state, and these proline mutants may favour the inactive conformation. Indeed, proline mutations of D423 in the α1 CC domain or G356 in the β1 CC domain rendered sGC unresponsive to NO activation, although these mutants could incorporate haem normally ( Fig. 3a, b, Extended Data Fig. 8d). By contrast, mutations of the same set of residues into alanines had no such effect ( Fig. 3a, b), indicating that the continuous helical structures of the αMNs are essential for activation of sGC by NO.
In the NO-activated conformation, the interface between the α1 and β1 CC domains is markedly different from that observed in the inactive state (Fig. 3a, Extended Data Fig. 7f, Supplementary Note 8). Besides the overall bending-straightening movements of each CC domain, the αN helix of the α1 subunit rotates approximately 70° around the αN helix of the β1 subunit (Fig. 3c). The separation of the C termini of the transducer modules also decreases. The distance between the Cα atoms of P459 of the α1 subunit and P399 of the β1 subunit shrinks from 26 Å to 20 Å (Fig. 3d). This drives the structural reorganization of the connecting catalytic module, in which the catalytic domain of the α1 subunit rotates 17° relative to the β1 catalytic domain (Fig. 3e, Extended Data Figs. 8f, 9i, j). These movements increase the volume of the central pocket from 1,375 Å 3 to 1,549 Å 3 and reorganize the catalytic centre (Extended Data Fig. 9k). This not only permits the binding of the substrate GTP and the cofactor Mg 2+ ions but also alters the local chemical environment of the pocket to make it possible for small stimulators to plug in and activate the enzyme (Supplementary Note 9). In the map of the NO-activated state, we observe a strong density corresponding to the substrate analogue GMPCPP (Fig. 3f), which was added during cryo-EM sample preparation. By comparing the current structure with the active adenylate cyclase structure 22 (PDB ID: 1CJU) (Extended Data Fig. 9l), we found that the residues responsible for substrate binding   and catalysis are in similar positions, indicating that the current sGC structure represents a catalytically competent conformation.

Structural mechanism of sGC activation
By analysing the structures of individual sGC domains in both inactive and NO-activated states, we found that the structures of the scaffolding PAS dimer remain relatively unchanged among the different states, with a r.m.s.d. of 0.91 Å (Extended Data Fig. 9m). Therefore, we used the structures of the PAS dimer as a reference point to align and compare the two full-length structures (Fig. 4a). During activation, the α1 H-NOX domain makes a small concomitant downward movement, while the interfaces between α1 H-NOX and its adjacent domains are largely maintained (Fig. 2d). This suggests that the α1 H-NOX domain may play a role that is mainly structural instead of being involved in NO signal transduction (Supplementary Video 5). This is in agreement with the finding from an activity assay that the H-NOX domain of the α1 subunit is dispensable for NO activation 27 (Fig. 4b). By contrast, the local conformational change of the β1 H-NOX domain upon NO binding drives the structural rearrangement of the sensor module ( Fig. 2d), which, along with previous functional data, suggests that the H-NOX domain of the β1 subunit has an essential role in NO sensing 5 . Indeed, complete removal of the β1 H-NOX domain rendered the sGC enzyme trapped in a relatively low activity state and unresponsive to NO activation (Fig. 4b). This suggests that the β1 H-NOX domain in the NO-bound state is necessary to stabilize the sGC enzyme in an active conformation. Further supporting this conclusion, disruption of the interactions between β1 H-NOX and adjacent domains by mutation also diminished activation by NO (Fig. 2f). The structural changes in the sensor module upon binding of NO trigger the bending-straightening conformational switch of the transducer module. As a result, the distal catalytic module rotates 86° in a swing-like manner and the centre of mass of the catalytic module is spatially displaced by 101 Å (Fig. 4a). Inhibition of the straightening by proline mutations abolishes activation by NO (Fig. 3b), which suggests that the conformational    Article reSeArcH change in the transducer module is essential for activation of the catalytic module. Furthermore, we found that the catalytic module changes from a conformation that cannot bind substrate to a catalytically competent conformation (Fig. 3e, f, Extended Data Figs. 8f, 9g, k), which explained how the binding of NO decreases the K m (GTP) and increases the catalytic constant (k cat ) of sGC 23 (Fig. 4c). It has been previously proposed that the activation of sGC by NO involves two steps 2 , and our structure observations are compatible with this two-step hypothetic model (Supplementary Note 10). Notably, information flow in the reverse direction, from the catalytic module to the sensor module, has been suggested by published functional data 28-30 . Therefore, the transducer module acts as an allosteric structural coupler between the sensor module and the catalytic module to allow the bi-directional flow of information within the sGC molecule (Supplementary Note 11).

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41586-019-1584-6. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Methods
Cell culture. HEK293F (Thermo Fisher Scientific) suspension cells were cultured in Freestyle 293 medium (Thermo Fisher Scientific) or SMM 293-TI medium (Sino Biological) supplemented with 1% FBS at 37 °C with 6% CO 2 and 70% humidity. It is reported that HEK293F is a female cell line. Sf9 insect cells (Thermo Fisher Scientific) were cultured in SIM SF (Sino Biological) at 27 °C. The cell lines were routinely checked to be negative for mycoplasma contamination but have not been authenticated. Protein expression and purification. cDNA of Drosophila melanogaster 31 , mouse, and human sGC were cloned into a modified BacMam expression vector 32,33 and transfected into HEK293F cells for screening by fluorescence-detection size-exclusion chromatography (FSEC) 34 on a Superose 6 increase 5/150 GL. The combination of C-terminal GFP-tagged human α1 and non-tagged β1 subunits yielded a stable heterodimer. sGC protein composed of an α1 and a β1 subunit is the most predominant isoform, and it has been widely used as a model protein to elucidate the biochemical, biophysical, and structural properties of mammalian sGC 5 .The coding sequences of human α1 and β1 subunits were transformed into the pFastbac dual vector and expression was driven by p10 or polyhedrin promoters. The corresponding baculovirus was generated using the Bac-to-Bac system. Sf9 insect cells at a density of 4 × 10 6 /ml in SIM SF medium were infected with the baculovirus and cultured at 27 °C in a shaker for 72 h before harvesting and storage at −80 °C. Cells corresponding to 500 ml culture were thawed and resuspended with 20 ml lysis buffer (50 mM Tris pH 8.0 at 4 °C, 150 mM NaCl) containing 1 μg/ml aprotinin, 1 μg/ml pepstatin, 1 μg/ml leupeptin, 1 mM phenylmethanesulfonyl fluoride (PMSF), 2 mM dithiothreitol (DTT), and 1 mM ethylenediaminetetraacetic acid (EDTA). Cells were broken by sonication in 5 s intervals followed by a 5 s pause at 50% output for 20 min. Unbroken cells, cell debris, and membranes were removed by ultracentrifugation at 40,000 r.p.m. for 1 h at 4 °C using a Ti70 rotor (Beckman). An excess amount of purified glutathione S-transferase-tagged GFP-nanobody 35 was added to the supernatant and incubated at 4 °C for 10 min with rotation. Samples were then loaded onto 4-ml Glutathione Sepharose 4B columns (GE Healthcare) and washed with TBS buffer (20 mM Tris, pH 8.0, 150 mM NaCl) containing 1 mM DTT at 4 °C. Protein was eluted with elution buffer (50mM Tris, pH 8.5, 10 mM reduced glutathione, 1 mM DTT) at 4 °C. The eluate was diluted with buffer A (20 mM Tris, pH 8.0 at 4 °C) to a conductivity lower than 5 mS/cm and loaded onto a 1-ml HiTrap Q HP (GE Healthcare). The protein was eluted with buffer B (20 mM Tris, pH 8.0, 500 mM NaCl) at 4 °C in a linear gradient using the AKTA pure system (GE Healthcare). The peak fractions containing sGC were pooled and incubated with prescission protease overnight to cleave the tag from the protein. The digested protein was further purified by Superdex 200 increase (GE Healthcare) running in buffer containing 20 mM HEPES (pH 7.4), 50 mM NaCl and 2 mM tris (2-carboxyethyl) phosphine (TCEP). The peak fractions containing the sGC protein were pooled and concentrated. UV-vis spectrum was measured using a spectrometer (Pultton) in the cuvette mode. Activity assay. The protein used for cryo-EM sample preparation was diluted with 20 mM triethanolamine (TEA, pH 7.6), 300 mM NaCl, 1 mM DTT and subjected to activity assay as described below. For the haem-oxidized sample, the protein was diluted with 20 mM TEA (pH 7.6), 300 mM NaCl and preincubated with 20 μM NS2028 at 25 °C for 30 min and then added DTT to the final concentration of 1 mM for activity assay. To generate the sGC mutant protein for activity assay, the coding sequences of the α1 subunit with a C-terminal GFP-strep tag and the β1 subunit were cloned into pFastBac1 expression vectors, respectively. To generate α1(ΔNOX) and β1(ΔNOX) constructs, the N-terminal 273 amino acids of the α1 subunit and 200 amino acids of the β1 subunit were removed, respectively. Constructs carrying the desired point mutations were generated by Quick Change and corresponding baculoviruses were generated using the Bac-to-Bac system. Sf9 insect cells at a density of 4 × 10 6 /ml in SIM SF medium were infected with baculovirus and cultured in a shaker at 27 °C for an additional 72 h before harvest. Cells were resuspended with buffer containing 50 mM Tris (pH 8.0 at 4 °C), 150 mM NaCl, 1 μg/ml aprotinin, 1 μg/ml pepstatin, 1 μg/ml leupeptin, 1 mM PMSF and 1mM DTT, and broken by passing through a syringe needle with a 0.45-μm inner diameter six times. Cell debris and membrane were removed by ultracentrifugation at 40,000 r.p.m. for 30 min at 4 °C using a TLA55 rotor (Beckman). The supernatants were loaded onto Streptactin Beads 4 FF (Smart-Lifesciences) and washed with buffer containing 20 mM Tris (pH 8.0 at 4 °C), 150 mM NaCl and 1 mM DTT. The protein was eluted with buffer containing 50 mM Tris (pH 8.5 at 4 °C), 50 mM NaCl, 10 mM d-desthiobiotin and 1 mM DTT. The eluates were diluted with an equal volume of 20 mM TEA (pH 7.6), loaded onto Q Beads 6 FF (Smart-Lifesciences) by gravity and washed with buffer containing 20 mM TEA (pH 7.6), 150 mM NaCl and 1 mM DTT. The protein was eluted by 20 mM TEA (pH 7.6), 300 mM NaCl and 1 mM DTT. The protein concentrations of various GFP-tagged sGC mutants were determined by comparing their GFP fluorescence signal to that of a purified GFP-tagged sGC standard on FSEC 34 . The activity assay mixture contained 10 nM sGC, 60 mM TEA (pH 7.6), 150 mM NaCl, 0.5 mM DTT, 5 mM MgCl 2 , 200 μM GTP with or without 200 μM DEA NONOate (Cayman Chemical) in a final volume of 20 μl. The assay mixture was incubated at 25 °C for 10 min and stopped by adding 80 μl 125 mM Zn(OAc) 2 and 100 μl 125 mM Na 2 CO 3 . The GTP-ZnCO 3 precipitation was removed by centrifugation at 17,000g for 5 min and the supernatant was used for cGMP quantification with the Cyclic GMP ELISA Kit (Cayman Chemical) according to the instructions. Each assay was independently repeated at least three times. For the measurement of UV-vis spectra of sGC mutants, proteins eluted from Streptactin Beads 4FF (Smart-Lifesciences) were digested with prescission protease overnight and further purified by Superdex 200 increase column (GE Healthcare) with buffer containing 20 mM HEPES (pH 7.4), 50 mM NaCl and 1 mM TCEP. The peak fractions were pooled and concentrated. UV-vis spectrums of sGC mutants with or without 400 μM DEA NONOate were measured using a spectrometer (Pultton). EM sample preparation. We prepared a haem-unliganded sGC sample, in which no exogenous ligand was supplemented, and then supplemented different small molecules to stabilize the purified sGC protein into functionally distinct states. The compound NS2028 has been reported to efficiently oxidize the Fe(ii) in sGC to Fe(iii) 36 . Indeed, we found that incubating sGC with NS2028 almost completely shifted the Soret peak from 431 to 392 nm (Extended Data Fig. 1c). Therefore, we incubated purified sGC with NS2028, Mg 2+ ions, and substrate GTPγS 37 to obtain the haem-oxidized state. To achieve the NO-activated state, we supplemented the purified protein with excess NO donor DEA NONOate 38 , Mg 2+ ions, and noncyclizable substrate analogue GMPCPP 29 . In detail, the purified sGC was concentrated to A 280 = 6 with an estimated concentration of 55.9 μM. For the haem-unliganded state sample, 5 mM MgCl 2 , 0.5 mM fluorinated octyl-maltoside (FOM, Anatrace) were added; for the haem-oxidized state sample, 5 mM MgCl 2 , 0.5 mM FOM, 100 μM NS2028 (Cayman Chemical), and 1 mM GTPγS (Sigma) were added; for the NO-activated state, 5 mM MgCl 2 , 0.5 mM FOM, 1 mM noncyclizable substrate analogue GMPCPP (Biorbyt), and 1 mM DEA NONOate (Cayman Chemical) were added. For the β1(H105C) mutant sample, we added 5 mM Mg 2+ ions, 1 mM GMPCPP and 0.5 mM FOM to the protein. Protein samples were loaded onto glow-discharged Quantifoil 0.6/1 holey carbon gold grids and plunged into liquid ethane by Vitrobot Mark IV (Thermo Fisher Scientific). Disulfide bond cross-linking. To generate the less-Cys sGC construct (sGC LC ), the cys-rich N-terminal 63 amino acids of α1 subunit were removed. Additional mutations of C176A, C239A, C669S, C455Y, and C460G were created in the α1 subunit and C292N in the β1 subunit. The coding sequences of α1 LC with a C-terminal GFP-strep tag and β1 LC without tags were cloned into modified BacMam expression vectors 32,33 . Then specific amino acids were mutated into cysteines using the Quick Change method. Cysteine mutants were transfected into HEK293F cells with polyethylenimine (PEI) (Polysciences) at a density of 2.0 × 10 6 /ml. Cells were harvested 72 h after transfection and broken by passing through a syringe needle with 0.45 μm inner diameter ten times. Unbroken cells and large debris were removed by centrifugation at 14,800 r.p.m. for 10 min at 4 °C. sGC proteins were purified from supernatants using Streptactin Beads 4FF resin (Smart-Lifesciences, China). Protein samples were cross-linked on ice for 30 min by adding Cu(ii) (1,10-phenanthroline) 3 to a final concentration of 30 μM to promote disulfide bond formation. Protein samples were subjected to 4-15% gradient SDS-PAGE (Beyotime Biotechnology, China) for separation either in non-reducing condition or reducing condition (in the presence of 100 mM DTT). The fluorescence was detected using a ChemiDoc MP (Bio-Rad) fluorescence imaging system. Cryo-EM data acquisition. Cryo-grids were screened on a Talos Arctica electron microscope (Thermo Fisher Scientific) operating at 200 kV using a Ceta 16M camera (Thermo Fisher Scientific). The screened grids were transferred to a Titan Krios electron microscope (Thermo Fisher Scientific) operating at 300 kV with an energy filter set to a slit width of 20 eV. Images were recorded using a K2 Summit direct electron camera (Thermo Fisher Scientific) in super-resolution mode at a nominal magnification of 130,000×, corresponding to a calibrated super-resolution pixel size of 0.5225 Å. The defocus range was set from −1.5 μm to −2 μm. Each image was acquired as a 7.68-s movie stack (32 frames) with a dose rate of 6.25 e − Å −2 s −1 , resulting in a total dose of about 48 e − Å −2 . All data acquisition was done using SerialEM. Cryo-EM data processing. The data processing workflows are illustrated in Extended Data Figs. 1-4 and Extended Data Table 1. Super-resolution movie stacks were motion-corrected, mag-distortion corrected, dose-weighted, and binned to a pixel size of 1.045 Å by MotionCor2 1.1.0 using 5 × 5 patches 39 . Contrast transfer function (CTF) parameters were estimated from non-dose-weighted micrographs using Gctf v1.06 40 . Micrographs with ice or ethane contamination, empty carbon, and poor CTF fit (>5 Å) were manually removed. All classification and reconstruction was performed with Relion 3.0 41 unless otherwise stated. Particles were picked using Gautomatch (developed by Kai Zhang) and subjected to reference-free 2D classification to remove bad particles. Initial models were generated by cryoSPARC 42 using the selected particles from 2D classification. The particles Article reSeArcH were further subjected to 3D classification to remove bad particles using the initial model, which was low-pass filtered to 30 Å as the reference. The particles selected from good 3D classes were re-centred and re-extracted, and their local CTF parameters were individually determined using Gctf v1.06 40 . These particles were imported into cisTEM 43 and subjected to 3D classification with auto-masking. The particles from the best 3D classes calculated by cisTEM were exported into Relion 3.0 and subjected to 3D auto-refinement to generate the consensus map. However, the two large lobes of sGC in the consensus maps showed blurry features, which were indicative of continuous conformational heterogeneities. Therefore, we divided the whole molecule into two bodies-the larger N lobe and the smaller C lobe-for further multibody refinement (Extended Data Figs. 1-4) in Relion 3.0, and the subsequent local map qualities were greatly improved (Extended Data Figs. 2a, b, 3e, f). In detail, two soft masks that cover the N lobe and C lobe were generated from the consensus map, which was edited manually in UCSF Chimera using the volume eraser tool 44 . 3D multi-body refinements 21 were performed using the two soft masks of the lobes and the parameters determined from previous 3D auto-refinement. The motions of the bodies were analysed by relion_flex_analyse in Relion 3.0. The two half-maps of each lobe generated by 3D multi-body refinement were subjected to post-processing in Relion 3.0. The masked and sharpened maps of each lobe were aligned to the consensus map using UCSF Chimera and summed to generate the composite map for visualization and interpretation. All of the resolution estimations were based on a Fourier shell correlation (FSC) of 0.143 cutoff after correction of the masking effect 45 14 and human α1-β1 catalytic domain heterodimer (PDB: 3UVJ) 16 . The models were placed into the corresponding composite maps using UCSF chimera 44 and manually rebuilt in Coot 47 . The composite maps were then converted into mtz files and the models were further refined by Phenix in reciprocal space 48 and Coot in real space. During model building, we found that the structures of the catalytic module in the haem-oxidized state and the haem-unliganded state were essentially the same, but we observed a positive difference density around the βK-αP loop of the α1 catalytic domain in the haem-oxidized state sample (Extended Data Fig. 2i). During the preparation of the haem-oxidized sample, we supplemented oxidizing reagent NS2028, substrate GTPγS and cofactor Mg 2+ ions into the sGC protein. Therefore, based on the local chemical environment, this positive density might represent Mg 2+ ions together with highly negatively charged phosphate groups that possibly came from the decomposition of the GTPγS molecule. However, these putative phosphate groups were not modelled. Volumes of the catalytic pocket were calculated using Caver 49 with the large probe radius 5 Å and the small probe radius 2.4 Å. Quantification and statistical analysis. Global resolution estimations of cryo-EM density maps are based on the 0.143 FSC criterion 50 . The local resolution was estimated using Relion 3.0 41 . The number of technical replicates (n) and the relevant statistical parameters for each experiment (such as mean or standard deviation) are described in the figure legends. No statistical methods were used to pre-determine sample sizes. Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this paper.
Article reSeArcH extended data table 1 | Cryo-eM data collection, refinement and validation statistics *The numbers outside the brackets are from the consensus refinement. Numbers inside brackets are from the multibody refinement (N-lobe/C-lobe).

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted

Software and code
Policy information about availability of computer code Data collection SerialEM 3.6.11 Data analysis MotionCor2, GCTF, Gautomatch, RELION 3.0, cisTEM, cryoSPARC, PHENIX, Coot, UCSF Chimera, Pymol, GraphPad Prism 6, Microsoft Excel, caver For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A list of figures that have associated raw data -A description of any restrictions on data availability Cryo-EM maps and atomic coordinates of the heme-unliganded, heme-oxidized, NO-activated and β1 H105C mutant sGC have been deposited in the EMDB and PDB under the ID codes EMDB: EMD-9883, EMD-9884, EMD-9885, EMD-9886 and PDB: 6JT0, 6JT1, 6JT2, respectively.

nature research | reporting summary
October 2018 Field-specific reporting Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
No predetermination of sample size was performed. Sufficient cryo-EM data were collected to achieve adequate map resolutions for model building. The enzymatic activity assay experiments were performed with three biological replicates. The sample size was based on previous studies in the field and clearly indicated in the legends.
Data exclusions Cryo-EM micrographs with ice or ethane contamination, empty carbon, and poor CTF fit (> 5 Å) were excluded manually. Particles belonging to bad classes were discarded and the data processing flowchart were summarized in Extended Data Fig. 1f, 3c. These criteria were preestablished and the procedure is a common practise in cryo-EM image analysis. No data was excluded in functional studies.

Replication
All attempts at replication were successful according to the detailed protocol described in the methods section. The numbers of replication were described in figure legends.
Randomization For cryo-EM 3D refinement, all particles were randomly split into two groups.

Blinding
The investigators were blinded to group allocation during cryo-EM data collection and analysis.
Reporting for specific materials, systems and methods We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.

Authentication
None of the cell line used was authenticated.

Mycoplasma contamination
All cell lines were tested negative for mycoplasma contamination.
Commonly misidentified lines (See ICLAC register) No commonly misidentified cell lines were used.