PAS-mediated Dimerization of Soluble Guanylyl Cyclase Revealed by Signal Transduction Histidine Kinase Domain Crystal Structure*

Signal transduction histidine kinases (STHK) are key for sensing environmental stresses, crucial for cell survival, and attain their sensing ability using small molecule binding domains. The N-terminal domain in an STHK from Nostoc punctiforme is of unknown function yet is homologous to the central region in soluble guanylyl cyclase (sGC), the main receptor for nitric oxide (NO). This domain is termed H-NOXA (or H-NOBA) because it is often associated with the heme-nitric oxide/oxygen binding (H-NOX) domain. A structure-function approach was taken to investigate the role of H-NOXA in STHK and sGC. We report the 2.1Å resolution crystal structure of the dimerized H-NOXA domain of STHK, which reveals a Per-Arnt-Sim (PAS) fold. The H-NOXA monomers dimerize in a parallel arrangement juxtaposing their N-terminal helices and preceding residues. Such PAS dimerization is similar to that previously observed for EcDOS, AvNifL, and RmFixL. Deletion of 7 N-terminal residues affected dimer organization. Alanine scanning mutagenesis in sGC indicates that the H-NOXA domains of sGC could adopt a similar dimer organization. Although most putative interface mutations did decrease sGCβ1 H-NOXA homodimerization, heterodimerization of full-length heterodimeric sGC was mostly unaffected, likely due to the additional dimerization contacts of sGC in the coiled-coil and catalytic domains. Exceptions are mutations sGCα1 F285A and sGCβ1 F217A, which each caused a drastic drop in NO stimulated activity, and mutations sGCα1 Q368A and sGCβ1 Q309A, which resulted in both a complete lack of activity and heterodimerization. Our structural and mutational results provide new insights into sGC and STHK dimerization and overall architecture.

The ability to sense small molecules is key for every life form and provides information about the extracellular milieu, monitors intracellular physiological status, or establishes cell-cell communication. Sensory signaling proteins are often modular in nature with distinct domains for ligand sensing and for output signals. A number of these domains are conserved in bacteria and animals. A striking example is the heme-nitric oxide/ oxygen-binding (H-NOX) 2 (previously termed H-NOB) domain that can be a stand-alone protein in Nostoc cyanobacteria, or can be part of a multidomain protein such as in the mammalian soluble guanylyl cyclase (sGC) (1, 2) (Fig. 1A). An additional evolutionary relationship was detected between sGC and 2 other cyanobacterial signaling proteins (1,2): the H-NOX associated H-NOXA (or H-NOBA) domain in sGC is also present at the N terminus of a cyanobacterial signal transduction histidine kinase (STHK) and 2-component hybrid sensor and regulator (2-CHSR) ( Fig. 1) postulated to have a PAS-like fold (1). Both genes of these cyanobacterial proteins are adjacent to genes coding for stand-alone H-NOX domains (Fig. 1A) suggesting they might work in concert. This H-NOXA evolutionary link adds to the already complex and poorly understood regulation of sGC stimulating the need to study this ancient domain.
Upon NO activation, sGC increases its production of the second messenger cGMP (3)(4)(5) leading to vasodilation, platelet aggregation, and induction of host defense mechanisms (6,7). sGC is therefore a potential drug target for treating hypertension and erectile dysfunction (8). sGC is comprised of an ␣ and ␤ subunit with the ␣1/␤1 isoform being the most ubiquitous (3). The H-NOXA domain is a central subdomain in both subunits: in sGC␤1, it is flanked by the N-terminal H-NOX domain and C-terminal predicted coiled-coil (CC) (9,10) and catalytic guanylyl cyclase (GC) domain (Fig. 1A). sGC␣1 has a similar subunit arrangement except that its N-terminal domain does not bind heme. As in sGC␣1 and sGC␤1, the H-NOXA domains in STHK and 2-CHSR are followed by a CC domain further strengthening their evolutionary connection (1) (Fig. 1A).
Structural information on sGC is limited and comes mostly from the structures of the homologous adenylyl cyclase catalytic domain (11)(12)(13) and bacterial H-NOX domains (14 -16). Even less is structurally known about the H-NOXA/CC region except that it was found to contain two distinct regions key for dimerization: one comprises H-NOXA ␤1 residues 204 -244, and the second contains CC residues 379 -408 with the intervening sequences postulated to contribute to a functional binding region (17). Analogous regions in sGC␣1 have also been shown to be important for sGC activity via deletion studies (9). These regions in sGC␣1 and sGC␤1 have recently been further narrowed down (18). The ␣1 and ␤1 H-NOXA domains are 38% sequence identical and also share ϳ35% sequence identity with the H-NOXA domains in the STHK from Nostoc punctiforme PCC 73102 (NpSTHK) and the 2-CHSR from Nostoc sp. PCC 7120 (Fig. 1B). Unlike the sGC subunits, the H-NOXA domain of the STHK was amendable to crystallographic analysis. We describe here the 2.1-Å crystal structure of the dimerized H-NOXA domain of NpSTHK revealing a PAS fold, a fold commonly found in sensory signaling proteins. We carried out additional structure-function studies in sGC and find that the H-NOXA subdomains of sGC could likely form a similarly arranged PAS-type heterodimer with an important role for functional sGC heterodimerization by the H-NOXA domains.
Expression and Purification of NpSTHK H-NOXA Domains-Both the 1-121 and truncated ⌬7 H-NOXA domains were expressed as N-terminal His-tagged protein in Escherichia coli BL21(DE3) Star cells (Invitrogen) using isopropyl 1-thio-␤-Dgalactopyranoside for induction. The cells were pelleted and sonicated for 5 min on ice in the following buffer: 20 mM Tris, pH 7.5, 100 mM NaCl, 5 mM ␤-mercaptoethanol. The cell lysate was centrifuged at 16,000 ϫ g for 10 min at 4°C and the supernatant was incubated with nickel-nitrilotriacetic acid (Qiagen) beads. The beads were washed with the washing buffer: 20 mM Tris, pH 7.5, 100 mM NaCl, 5 mM ␤-mercaptoethanol, 20 mM imidazole. The protein was released from the beads by thrombin (Enzyme Research labs) digestion. Further purification was performed by gel filtration in a Sephadex-75 (GE Healthcare) column equilibrated with 5 mM Tris, pH 7.5, 100 mM NaCl, and 1 mM ␤-mercaptoethanol.
Subcloning, Expression, and Purification of the H-NOXA Domain of sGC␤1-Residues 202-344 of the H-NOXA domain of rat sGC␤1 were subcloned into pET22b using the forward primer, 5Ј-ggaaattccatatgggtacccaggactcccgtatc-3Ј, and reverse primer, 5Ј-gcgaattctcagtggtggtggtggtggtgtccagggatgtcactcaggtacag-3Ј. The C-terminal His-tagged sGC␤1 H-NOXA protein was expressed and purified similarly to the homologous NpSTHK counterpart except that the protein was eluted from the nickel-nitrilotriacetic acid beads by an imidazole gradient instead of thrombin digestion. The Ala-scanning mutants of the sGC␤1 H-NOXA domain were introduced into pET22b plasmid using the QuikChange site-directed mutagenesis kit (QuikChange, Stratagene) to confirm the sequence.
Mutagenesis of Rat sGC and Transfection in COS-7 Cells-Templates were cDNAs encoding the ␣1 and ␤1 subunits of rat sGC cloned into the mammalian expression vector pCMV5 (19). Mutations described in the text were introduced by PCR (QuikChange, Stratagene) and sequenced. 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 for 48 h with Superfect reagent using the protocol of the supplier (Qiagen).
Cytosol Preparation and Western Blot Analysis-COS-7 cells were washed twice with ice-cold phosphate-buffered saline and then scraped off the plate in cold lysis buffer: phosphate-buffered saline containing protease inhibitors, 50 mM HEPES, pH 8.0, 1 mM EDTA, and 150 mM NaCl. Cells were broken by sonication (3 pulses of 3 s) and centrifuged at 16,000 ϫ g for 10 min at 4°C to collect the soluble fraction (referred to as cytosol in the text). To determine the efficiency of transfection for wt sGC or mutants in COS-7 cells, 15 g of cytosol was resolved on 10% SDS-PAGE and analyzed by immunoblotting with anti-sGC (anti-␣1 subunit and anti-␤1 subunit; Cayman Chemicals).
sGC Activity Assay-GC activity was determined by formation of [␣-32 P]cGMP from [␣-32 P]GTP, as previously described (20). 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 COS-7 cytosol transfected with either wt or mutants were used in each assay reaction. All assays were done in duplicate and each experiment repeated twice. Enzymatic activity was stimulated with the NO-donor SNAP (Calbiochem) at 100 M. sGC activity is expressed in pmol min Ϫ1 mg Ϫ1 and mean Ϯ S.E.
NpSTHK Crystallization-Crystals of the NpSTHK-(8 -121) protein construct were grown at 4°C by sitting-drop vapor diffusion. Protein was concentrated to 20 mg/ml in 5 mM Tris, pH 7.5, 100 mM NaCl, 1 mM ␤-mercaptoethanol and was mixed with an equal volume of the reservoir solution: 1.7-1.9 M ammonium sulfate, 100 mM Tris-HCl, pH 7.7-9.0. Crystals were prepared for data collection by fast transfer into cryoprotectant solution containing 2.0 M ammonium sulfate, 100 mM Tris-HCl with 15% glycerol. The 1-121 NpSTHK protein was concentrated to 15 mg/ml after the gel-filtration step and mixed with an equal volume of 0.1 M HEPES, pH 7.5, and 1.5 M lithium sulfate monohydrate. Crystals appeared after 3 days at 20°C. For data collection, the crystals were soaked in cryoprotectant containing 25% glycerol in addition to the crystallization solution and dunked into liquid nitrogen prior to data collection. Crystals of selenomethionine substituted 8 -121 protein were used for single wavelength anomalous dispersion (SAD) phasing.
NpSTHK Structure Determination and Refinement-Due to great difficulty in obtaining diffraction quality crystals for the 1-121 NpSTHK construct, the initial structure determination was carried out using crystals of the shorter 8 -121 construct. A native 2.0-Å resolution dataset for the 8 -121 NpSTHK protein was collected at ALS (Beamline 4.2.2) and processed with D*trek (21). The crystals belong to space group P6 1 22, with cell dimensions a ϭ b ϭ 72.3 Å, c ϭ 169.1 Å and two molecules in the asymmetric unit. To obtain crystallographic phase information, a SeMet crystal of the 8 -121 NpSTHK construct was used for SAD phasing. A 2.6-Å SAD data set was collected at the selenium peak wavelength at ALS. SOLVE/RESOLVE (22)

RESULTS
NpSTHK Structure Reveals a PAS Fold-The structure of the H-NOXA domain of NpSTHK was solved via SAD of first a smaller ⌬7 N-terminal-truncated fragment, followed by molecular replacement to solve the full-length NpSTHK H-NOXA domain structure (Table 1 and Fig. 1C). Two independent molecules in the asymmetric unit form a dimer. The final refined model of the NpSTHK dimer contains residues 1-107 of molecule A and 1-105 of molecule B.
The NpSTHK monomer structure is comprised of a 7-stranded anti-parallel ␤-barrel flanked by several ␣-helices (Fig. 1C). The structure of the NpSTHK monomer is similar to that of the PAS fold: its most similar structural neighbors are two heme containing sensors bjFixL and EcDOS (Protein Data Bank codes 1DRM and 1V9Z), followed by dPER (1WA9), HIF-2␣ (1P97), and photoactive yellow protein (3PYP) with DALI Z-scores (29) ranging from 8.3 to 7.6, respectively. A superposition of the NpSTHK and EcDOS PAS domain (30) (Fig. 1D) details their close structural similarity (r.m.s. deviation of 2.2 Å for 70 C␣ atoms). Their structure-based sequence alignment (Fig. 1B) reveals only 10% sequence identity. The canonical PAS fold contains 5 ␤-strands (31) but both NpSTHK and EcDOS contain an additional short ␤C strand; NpSTHK contains even an additional 7th (␤D) strand (Fig. 1D). The region between the two N-terminal strands (␤A/␤B) and three C-terminal strands (␤E/␤F/␤G) usually harbors 4 helices in PAS domains (31). Three of those helices are present in NpSTHK (Fig. 1B) although helix ␣D is in a somewhat different orientation (Fig. 1D). This multihelix containing region between ␤B and ␤E is structurally the most variable part of PAS domains, and can have as few as 2 helices such as CitA (32), possibly dictated by whether or not the PAS domain binds a ligand and if so, the type of ligand (31,33) (Fig. 1D). The similarity between NpSTHK and EcDOS is thus remarkable despite little sequence identity and extends beyond the PAS fold because they both contain an additional N-terminal helix, ␣A (Fig. 1D).
A structure-based sequence analysis of the homologous sGC subunits using the NpSTHK H-NOXA domain structure was used to probe sequence conservation and sites of insertions and deletions (Fig. 1B). Both sGC subunits have a large insertion between the ␤E and ␤F strands in their H-NOXA domain: a 9-residue insertion for ␣1 and 19 for ␤1. In addition, there are some smaller 1 residue insertions and deletions around the ␤C strand (Fig. 1B). To test whether the sGC subunit sequences are compatible with the PAS-fold of NpSTHK, MODELLER (34) was used to build homology models for the ␣1 and ␤1 H-NOXA domains using the alignment in Fig. 1B. The resulting models yielded no negative VERIFY3D structure validation (35) scores (or even scores with values lower than 0.1) indicating that both subunits can adopt the PAS fold of NpSTHK (the 19-residue insertion in ␤1 was omitted from the modeling due to its size and lack of template).
NpSTHK Dimer Organization-The two dimerized NpSTHK molecules make extensive interactions burying 2,395  with many water-mediated interactions, and is likely nonphysiological as indicative of the lower complexation significance score and PITA dimer evaluation scores of 0.44 and 47.1, respectively, the latter being below the threshold value of 67 (38). The two NpSTHK domains comprising the fulllength dimer are related by an approximate 2-fold non-crys-tallographic axis (178.5°) and are very similar in conformation (r.m.s. deviation of 0.45 Å for 105 C␣ atoms). Dimerization involves juxtapositioning of the N-terminal helices and preceding residues and the face of the ␤-sheet (Fig. 2, A-D). The dimer interface contains a central cluster of hydrophobic residues (including Leu 8 , Leu 13 , Phe 17 , Phe 100 , and Leu 102 ) and hydrogen bonds (involving side chains of residues Thr 7 , Ser 9 , Gln 31 , Gln 89 , and Gln 93 and main chain atoms of several other residues) (Fig. 2, B and C).
The resemblance of NpSTHK with EcDOS extends beyond the monomer fold because their dimer organizations are also similar involving N-terminal helix juxtapositioning (Fig. 3). The dimeric NpSTHK and EcDOS structures can be superimposed with r.m.s. deviations of 3.5 Å (for 2 ϫ 70 C␣ atoms), which is higher than the r.m.s. deviations of 2.2 Å for the monomer superposition yet low enough to indicate a similar dimer arrangement. Their similar dimer organization is attained because the hydrophobic nature of several residues at the core of the dimer interface is conserved (Leu 13 , Phe 100 , and Leu 102 in NpSTHK correspond to Leu 26 , Leu 127 , and Leu 129 , respectively, in EcDOS; see Fig. 1B). The same fold and dimer organization are also present in another heme containing oxygen sensor, RmFixL (39), also an STHK protein (Fig. 3). Furthermore, such PAS dimerization was also recently observed in AvNifL (40) (Fig. 3), which led to the suggestion that this is a conserved dimerization motif for a subset of PAS domains (40).
Evidence for a NpSTHK-type Dimer for the H-NOXA Domains of sGC-The N-terminal region of the H-NOXA domain of NpSTHK provides most of the dimer contacts, which is in agreement with the corresponding sGC␤1 residues 204 -244 found to be important for sGC dimerization (Fig. 1B) (17). Furthermore, the dimer interface region is largely conserved in sGC␣1 and sGC␤1 (Fig. 4A), suggestive of a similar dimerization function in sGC. The hydrophobic nature of all 5 residues comprising the central cluster of the dimerization interface (Leu 8 , Leu 13 , Phe 17 , Phe 100 , and Leu 102 ) is conserved as well suggesting that sGC␣1 and sGC␤1 H-NOXA domains could similarly heterodimerize, or homodimerize as observed in solution for the sGC␤1 H-NOXA domain (Fig. 4B). To test whether a heterodimeric sGC indeed could adopt a similar NpSTHK dimer arrangement, a homology model was constructed for the ␣1␤1 sGC H-NOXA heterodimer using MOD-ELER, which resulted in similar (non-negative) VERIFY3D scores as was observed for the individual monomers. The modeled sGC heterodimer structure was additionally validated by carrying out dimer analysis calculations using PISA resulting in a ⌬G for dimer formation of Ϫ16.9 kcal/mol and 2,156 Å 2 of buried surface (␣2␤1 yielded similar results). These values are very similar to that calculated for the experimentally determined NpSTHK dimer itself, whereas a MODELLER generated ␣1␤1 heterodimer homology model based on the likely nonphysiological ⌬7 NpSTHK dimer yielded only a ⌬G of Ϫ7.9 kcal/mol (1,742 Å 2 of buried surface). These automated modeling and analysis results suggest that the sGC H-NOXA domain sequences are indeed compatible with formation of a NpSTHK-type heterodimer.
To further validate the presence of a NpSTHK-type dimer interface in sGC we mutated 8 putative H-NOXA interface residues to alanines (see Figs. 1B, 2, B and C, and 4 for the NpSTHK interface residue positions that were targeted for mutagenesis of the corresponding residue in sGC). We tested the effect of these mutations on dimerization of the isolated sGC␤1 H-NOXA domain and the dimerization and activity of full-length sGC. The 8 Ala mutations were made in sGC␤1; two of the mutations were also generated at the corresponding positions in sGC␣1. All mutations had a negative effect, with varying degrees, on dimerization of the ␤1 H-NOXA domain (Fig. 4B). The ␤1 mutations that caused the largest effects on H-NOXA dimerization are Q231A, Q309A, and L322A, as more than half of the mutant protein became monomeric at 1 mg/ml (Fig. 4B). Even the mutations that were the least disruptive, S206A and F217A, still had a small but measurable negative effect on dimerization indicative of a role in dimerization (Fig. 4B). The latter F217A mutation also caused a widening of the dimer peak, perhaps suggesting an (additional) larger structural destabilization caused by this mutation (Fig. 4B).
The 8 sGC␤1 and 2 sGC␣1 Ala-scanning mutations had in most cases less of an effect on dimerization of the full-length sGC as probed by immunoprecipitation and Western blotting (Fig. 4, C and D). Exceptions are the sGC␤1 mutations Q309A (and equivalent Q368A mutation in sGC␣1), L322A, and to a lesser degree Q231A and I208A (Fig. 4, C and D). The Q309A (and sGC␣1 equivalent) caused a complete loss of sGC heterodimerization as partnering subunits could not . ORF, open reading frame. B, structure-based sequence alignment of the H-NOXA domains of STHK, 2-CHSR, sGC␣1, sGC␤1, and sGC␤2 subunits, and EcDOS (PDB code 1V9Z) and RmFixL (PDB code 1D06). Residues identical to sGC␤1 are in bold. Semi-conserved hydrophobic residues, excluding heme PAS domains because they have a heme as their hydrophobic core, are highlighted in yellow; fully conserved residues in green. Residues at the NpSTHK dimer interface are labeled with a "#"; the putative sGC interface residues probed by mutagenesis are circled. Residues of EcDOS and RmFixL that are in structurally equivalent positions with the H-NOXA structure are in uppercase. The EcDOS and RmFixL sequences contains a few insertions at positions labeled "ˆ" and their insertion sizes are 3, 6, and 2, respectively, for both EcDOS and RmFixL. sGC␤1 residues 204 -244 have previously been shown to be key for dimerization (17) and are boxed. The expressed construct end (Lys 121 ) and start of the CC of NpSTHK are shown (hooked arrow). The red underlined stretches of residues highlight deletions that resulted (18) in lack of sGC heterodimerization and concomitant loss of activity; deletion of the orange underlined stretches of residues resulted in sGC heterodimerization but severely compromised cyclase activity; deletion of the blue underlined stretches of residues resulted in increased cyclase activity and decreased EC 50 values for BAY41-2272 in combination with nitric oxide; deletion of the dotted blue underlined stretch of residues had a mixed effect in that this deletion caused a decrease in cyclase activity yet a decrease in EC 50 for BAY41-2272 in combination with nitric oxide (18). be pulled down with antibodies directed against either the ␣ or ␤ subunit (Fig. 4D). It is interesting to note that when this Gln mutation occurs in the sGC␣1 subunit, the wt sGC␤1 subunit can be pulled down using a ␤-specific antibody at comparable levels, whereas when the mutation is introduced in the sGC␤1 subunit, wt sGC␣1 itself cannot be pulled down with an ␣-specific antibody despite being present in the cell lysate probably as a result of aggregation. This differential sGC subunit behavior is similar at physiological receptor concentrations in mice with individual sGC subunits knocked-out because only the sGC␤1 subunit could be detected in sGC␣1 knock-outs (41) but not vice versa (42). A number of the mutations had a comparable effect on both sGC␤1 H-NOXA homodimerization and full-length sGC heterodimerization. sGC␤1 mutations Q309A, Q231A, and L322A caused a significant decrease in dimerization in both experiments, whereas F217A and S206A both had only a minor effect on H-NOXA homodimerization and little if any effect on full-length sGC dimerization (Fig. 4).
The effect of the Ala sGC mutations on guanylyl cyclase activity was also measured in mutant-transfected COS-7 cells. The effect on activity was most pronounced for the Q309A sGC␤1 mutant and the equivalent sGC␣1 Q368A mutant, which both resulted in a complete loss of NO-stimulated and basal activity, in agreement with their inability to even form a heterodimer (Fig. 4D), the minimum needed for sGC basal activity (43). The ␤1 F217A and equivalent ␣1 F285A mutations caused a 3.6-and 6.1-fold drop in NO-stimulated activity, respectively (Fig. 4D). This result is somewhat unexpected because both Phe 3 Ala mutants had only limited effects on H-NOXA homo-and sGC heterodimerization (Fig. 4, B and D). Nevertheless, the substantial decrease in NO-stimulated activity by these Phe 3 Ala mutations could perhaps be explained by either a more global destabilization of the H-NOXA domains as mentioned above or perhaps an altering of the relative orientation of the H-NOXA dimer within the mutant sGC (described in more detail below). Except for the mutants shown in Fig. 4D, the remaining mutants had only a modest negative effect on guanylyl cyclase (Fig. 4C) yielding a maximal decrease of NO-stimulated activity of just under 3-fold. However, the basal activity of the cell lysate is also lower for these mutants such that the fold activity enhancement upon NO stimulation above for these latter mutants is similar to wt being around 15. The one exception is the L322A sGC␤1 mutant, which harbors a 29-fold increase in stimulated activity over basal activity (Fig. 4C).
The Ala-scanning mutagenesis revealed that some of the residues at the putative H-NOXA dimerization interface in sGC are critical for dimerization and/or activity. The most critical residues are Phe 285 and Gln 368 in sGC␣1, and the equivalent Phe 217 and Gln 309 in sGC␤1, and correspond to Phe 17 and Gln 89 in NpSTHK (Figs. 1B and 2, B and C). Interestingly, Phe 17 and Gln 89 cluster and are both located at the bottom of the NpSTHK interface (Fig. 2C) yet still partially solvent exposed in proximity of the C termini of this domain. This proximity sug-gests that this region could have a key role in perhaps additionally interacting with the CC domain (or GC domain) or perhaps be important for the precise orientation of the H-NOXA domains and its C-terminal located domains. The Phe 3 Ala and Gln 3 Ala mutations will be discussed in more structural detail below within the context of the NpSTHK structure.
Phe 17 is partly solvent exposed and provides hydrophobic interactions via mostly its tip with the symmetry related pair Phe 17 Ј and also with Leu 102 Ј (Fig. 2B) and buries only 40 Å 2 of surface upon dimerization. Because this residue is near the 2-fold axis, when it is mutated in the isolated ␤1 H-NOXA homodimer, the amount of buried surface lost is much less than double because this residue is mostly interacting with its symmetry Phe 17 Ј mate. The limited dimer interface contribution of Phe 17 might provide an explanation for the only minor decrease in dimerization seen in the corresponding sGC mutants. However, its potentially critical position along the 2-fold dimer axis situated toward the C-terminal CC ϩ GC domains might be crucial for H-NOXA domain orientation and when mutated might affect CC/GC interactions (or orientation) impeding signal transduction and thus provide an explanation for the strong effects on NO-stimulated activity.
Residue Gln 89 of molecule A of NpSTHK forms a hydrogen bond and a water-mediated hydrogen bond with the backbone oxygens of Ala 16 Ј and Lys 15 Ј, respectively, in addition to packing against residue Met 91 being a key hydrophobic interface residue (Fig. 2, B-D). Note that due to the 1.5°deviation from the perfect 2-fold axis of the crystallized NpSTHK dimer, residue Gln 89 Ј of the other subunit is located 1 Å more distant from the dimer interface although still involved in packing against Met 91 , a likely key interface residue. Minor deviations from a perfect 2-fold axis have been observed to be as large as 7°in other dimer protein crystal structures (44) and we anticipate a symmetrical dimer when in solution.
Of all the mutants generated in this study, mutations at the equivalent position of STHK:Gln 89 in sGC (Q368A and Q309A in sGC␣1 and sGC␤1, respectively) resulted in drastic loss of sGC␤1 H-NOXA homodimerization, loss of sGC heterodimerization, and loss of basal activity. This glutamine residue position is thus clearly crucial and, like Phe 17 , is positioned at the edge of the dimer interface toward the C-terminal CC ϩ GC subunits. We speculate that its mutation in the sGC␤1 H-NOXA homodimer affects its homodimerization via loss of 2 direct hydrogen bonds and 2 water-mediated hydrogen bonds across the interface, and reorientation of its neighboring inter- face residues Met 91 and Met 91 Ј.
That the corresponding mutations in the full-length sGC receptor also lead to complete loss of heterodimerization is somewhat surprising because there are additional dimerization interactions in fulllength sGC (CC domain and GC domain). However, the protein concentrations of the full-length sGC heterodimerization experiments in cell lysates are much lower compared with those in the purified H-NOXA homodimerization experiments that would render the sGC dimerization state more sensitive to mutations even if it were present in only one of the domains. Nevertheless, it is still a remarkable effect because only one of the Gln residues is mutated per mutant due to the heterodimeric nature of sGC. We hypothesize that the reason that even the single ␣1 Q368A and ␤1 Q309A mutants causes loss of sGC heterodimerization is the cumulative effects of (a) loss of a direct hydrogen bond and a water-mediated hydrogen bond across the interface, (b) possible reorientation of its neighboring interface residue Met 91 equivalent (␣Ile 370 and ␤Ile 311 ), and (c) possible reorientation of the H-NOXA domains and/or loss of interactions with C-terminal domains due to the location of the Gln 89 interface residue toward these CC ϩ GC domains (Fig. 2C).
In summary, Ala scanning mutagenesis of the putative dimerization interfaces of the H-NOXA domains of sGC␤1 and sGC␣1 can lead to loss of dimerization and/or guanylyl cyclase activity pointing to the role of H-NOXA in sGC dimerization and activity. The two mutations that were carried out in both sGC␣1 and sGC␤1 subunits resulted in similar behavior (Fig.  4D) suggesting that they have roughly equally important roles. Together, these results are compatible with the postulate that the H-NOXA domains of sGC arrange themselves in a NpSTHK-like fashion. Such heterodimeric arrangement in sGC could be either rigid, , and Leu 322 (Leu 102 ). A mixture of molecular weight protein standards was also applied to the column for reference (gray). The theoretic molecular mass for the monomer of wt sGC␤1 H-NOXA domain is 17 kDa, which includes the His tag. For comparison, the crystallized NpSTHK, with a theoretical molecular mass of 14 kDa, elutes at 10.6 ml (not shown), which is similar to that of the wt sGC␤1 H-NOXA domain confirming the dimeric state for both proteins. The injecting volume for each run is 100 l and protein concentration is 1 mg/ml. C, guanylyl cyclase activity and Western blot analysis of co-immunoprecipitated wt and 6 sGC␤1 mutants of sGC. Basal and NO-stimulated guanylyl cyclase activity of cell lysates with transfected sGC subunits are shown (upper panel). The Western blot (lower panel) includes co-immunoprecipitated sGC subunits pulled down with anti-sGC␤1 antibody (top gel) as well as the contents of the cell lysates (bottom). The Western blot was probed simultaneously with anti-sGC␣1 and anti-sGC␤1 antibodies. D, guanylyl cyclase activity measurements and Western blot co-immunoprecipitation analysis of additional mutants of sGC generated in both ␣1 and ␤1 subunits. These mutations include the sGC␤1 F217A and Q309A mutations and are also generated at the equivalent positions in sGC␣1 (␣1 mutations F285A and Q368A, respectively). Data presentation are similar as in C but includes additional co-immunoprecipitation (IP) using an anti-sGC␣1 antibody (top gel). This Western blot was also probed similarly as in C. The difference in specific activity observed between experiments C and D (as reflected in the wt activity) is probably due the fact that two different batches of COS-7 cells were used in C and D. For each set of experiments, COS-7 cells were transfected three times and activity assays repeated three times with each measurement done in duplicate. required to pivot/reorient, or be more transient during one of the stages of sGC activation. The implications of such a possible arrangement are discussed below.

DISCUSSION
The PAS fold and postulated NpSTHK-type dimer organization for the H-NOXA domains of sGC (Figs. 2-4) has interesting consequences regarding preferential sGC heterodimerization, possible allostery, and the overall domain architecture of sGC.
Possible Implications for Preferential Heterodimerization of sGC-Our structural results regarding the 1-121 and ⌬7 truncated 8 -121 NpSTHK constructs point to a critical role for the N terminus in dimer formation. Deleting the first 7 residues of NpSTHK resulted in a flipped dimer suggesting that these residues are critical for providing specificity for correctly orienting the monomers within the dimer. Within these first 7 residues, residues Pro 4 , Leu 6 , and Thr 7 are involved in dimer interface interactions via either hydrogen bonds or van der Waals interactions (Fig. 2, B and C) in the 1-121 NpSTHK dimer structure. Residue Leu 6 has a major role in that both its side chains makes interactions, with a relatively conserved hydrophobic region at the interface involving Ala 22 and Val 30 , and its main chain N and O atoms make interface hydrogen bonds with the conserved Gln 31 residue. N-terminal H-NOXA interactions thus likely provide the specificity for correct monomer:monomer juxtapositioning in NpSTHK yet could also have a role in preferential sGC heterodimerization as discussed next.
In addition to functional ␣1␤1 and ␣2␤1 heterodimerization, sGC homodimers are also observed for ␤1 (43), ␤2 (45), and ␣1 (46), the latter being less stable. Homodimers of sGC are inactive (43), except for the ␤3 subunit from Manduca sexta (47) and possibly the ␤2 sGC subunit although its weak activation needs non-physiological manganese, thus providing no conclusive evidence that its H-NOXA domains are dimerized under physiological conditions (45). A physiological equilibrium is thought to be present between homo-and heterodimeric sGC (43) although homodimeric ␤1 (41) and in particular homodimeric ␣1 are found to be very unstable in vivo (42). This balance is likely influenced by each of the three known interfaces involving the GC, H-NOXA, and CC domains (9,17,48). Such preferential heterodimerization tendency within these sGC domains is likely a consequence of complementary interface differences between ␣1 (or ␣2) and ␤1 subunits. Although speculative, sequence comparisons of interface residues and modeling of homodimeric and heterodimeric H-NOXA dimers suggests that sGC residues corresponding to NpSTHK residues Leu 6 and Ala 22 Ј and Val 30 Ј could be, in part, responsible for preferential heterodimerization of sGC. These residues cluster (Fig. 2, B and C) and size variations across the interface suggest that a heterodimer might be more sterically compatible. In sGC␣1 (and sGC␣2), these residues are Leu 274 , Met 290 Ј, and Leu 298 Ј, respectively, yet are smaller in sGC␤1 being Ser 206 , Ile 222 Ј, and Thr 230 Ј, respectively, all within the stretch of residues 204 -244 shown to be key for binding to ␣1 (17). Largersized interface residues at all these positions, such as in an ␣1 (or ␣2) homodimer, would likely cause steric hindrance between Met 290 Ј and Leu 274 such that the latter cannot position itself to form key main chain hydrogen bonds (equivalent to NpSTHK residue Leu 6 ). This disruption would disfavor such dimerization perhaps analogous to the disruptive effects of the ⌬7 deletion in NpSTHK. sGC␤1 has smaller residues at these 3 positions likely permitting homodimer formation yet a heterodimeric ␣1␤1 could possibly be favored via improved van der Waals packing by complementary combinations of small and larger residues at these 3 interface positions. Although it is interesting that the NpSTHK dimer structure could suggest a structural basis for a possible H-NOXA-contributing role in sGC heterodimerization, it remains speculative. Such a H-NOXA contributing role for sGC heterodimerization awaits future experimental validation especially in light of that residues preceding the N-terminal helix could possibly adopt different conformations such as a ␣-helical configuration in RmFixL (39).
Possible Role for PAS Dimer in Signaling or Allosteric Regulation in sGC and NpSTHK-The possibility of PAS-based allostery in sGC and NpSTHK is intriguing because the PAS domain is an important ancient sensory domain that can, for example, sense redox potential, small ligands such as oxygen, and light, in addition to maintaining protein-protein interactions (49,50). PAS sensory domains are often linked with output domains such as a histidine kinase, phosphodiesterase, or adenylyl cyclase domains, and are remarkably abundant in cyanobacteria (51). Our results reveal a PAS dimer organization for NpSTHK indicating that it has the same subdomain architecture as RmFixL STHK: a parallel dimerized (oxygen-sensing) PAS domain followed by a CC domain and histidine kinase domain. As noted earlier, such parallel oriented PAS domain dimers are also observed in EcDOS and AvNifL (Fig. 3), which lead to its recognition as a conserved PAS domain dimerization motif (40). RmFixL, EcDos, and AvNifL either sense redox potential or oxygen each involving postulated signaling mechanisms that lead to changes at the PAS-domain interface (30,40). The presence of a PAS domain in NpSTHK and sGC and the noted dimerization similarities raises the possibility that these PAS domains in NpSTHK and sGC are also used for signal transduction purposes or perhaps even sensory purposes. Obviously, the major sensory domain in sGC is the NO-sensing H-NOX domain yet it is tempting to speculate that the PAS domain could harbor a second allosteric regulatory module in sGC. This module could be present as an evolutionary remnant, for perhaps heme binding, or as a site for an unknown regulator. Intriguing candidates could be the allosteric sGC regulators YC-1, as well as ATP and GTP, whose binding site, or sites (52), to sGC has not been unambiguously mapped (6,(53)(54)(55)(56). Alternative to a small molecule, the pocket of a PAS domain is also capable of harboring a tryptophan side chain of a different subunit as observed in PERIOD (57). If such a PAS-mediated signaling or sensory regulation were indeed to occur in sGC or NpSTHK, some opening up of the PAS domain would be necessary for binding. Other possible PAS regulatory mechanisms have been observed as well and include N-terminal PAS helix unfolding (58), a speculated helix-swap event (59), or PAS pocket mediated inter-subunit interactions with a C-terminal helix (57) indicating that PAS-mediated contacts seems to play a key role in a number of these signal transduction events. A regulatory/mechanistic role for the centrally positioned H-NOXA domains is a sincere possibility as discussed below.
NpSTHK-type H-NOXA Dimer in sGC and the Implications for sGC Domain Organization and Activation-The H-NOXA domain is located in the center of both the ␣1 and ␤1 subunits of sGC and the structure of the dimerized homologous NpSTHK H-NOXA domain therefore offers new insights into the arrangement of the flanking H-NOX and CC-GC domains. The NpSTHK dimer positions the two C termini at ϳ18 Å distance (for residues Ile 107 and Pro 105 Ј) at the bottom face of the dimer (Fig. 2, A and C). The N termini protrude in opposite directions and are located on the side of the NpSTHK dimer, each positioned about ϳ30 Å from the C terminus of the other monomer. The relative close proximity of the C termini of the NpSTHK domain dimer suggests that the succeeding CC domain, present in both NpSTHK and sGC, is in a parallel arrangement in accord with a previous sequence analysis study that had named this CC region a signaling helix or S-helix (10). The locations of the N and C termini has interesting consequences for the overall architecture of the heterodimeric sGC when combined with additional constraints in that the GC domains need to form a catalytically active heterodimer and that there is a direct interaction between ␤1 H-NOX and the GC domain (48). The structure of sGC H-NOX and GC domains are not known but we have taken their homologous NsSTHK (16) and adenylyl cyclase crystal structures (11,12). The CC region is depicted as parallel CC segments as previously predicted (10), although we do not rule out other CC arrangements such as a 4-helix bundle. Despite some limitations, we generated a model for the entire sGC heterodimer (Fig. 5). In this modeled composite structure, part of the H-NOX domain is able to reach the GC domain (Fig. 5) with a stretch of 15 residues between the heme domain and the H-NOXA domain allowing such conformational flexibility. Although it is not known where and whether a protein:protein regulatory site is present on the sGC GC domains, such a site is present in the homologous adenylyl cyclase domains. The adenylyl cyclase activity is regulated by G␣ s binding to the AC2 domain, which is homologous to the ␤1 cyclase domain (11). Assuming that sGC uses the same site on its catalytic domain for regulation, the H-NOX domain is oriented such that its region near loop L1 (containing Asp 44 -Asp 45 , speculated to be near the activation switch (14,60) and shift up activation (16)), is closest to, and can be sensed by, the ␤1 cyclase domain site that corresponds to the surface where G␣ s binds adenylyl cyclase. The above sGC model is speculative, in particular as it is in large part based on the NpSTHK-type PAS dimer organization and PAS domains can be capable of undergoing (mechanistic) reorganization such as in PilT (61). Nevertheless, this is a new global model of the sGC domain arrangement that provides a framework for future structure-guided experiments and can aid in the interpretation of a recent deletion mutagenesis study (18) as discussed below.
The above noted abundance of inter-subdomain interactions within sGC␣1␤1 provides the possibility that some of these interfaces are more involved in the sGC activation mechanism than needed for sGC heterodimerization. A recent study detailing a systematic deletion mutagenesis analysis (18) in light of our structural results hints at such a possibility. This study first observed a roughly equal importance of the ␣1 and ␤1 CC residues for overall sGC heterodimerization (18); the results from this study are also included in Fig. 1B. This study also points to a critical dimerization role for ␣1:363-372 and ␤1:304 -313, which are both located in H-NOXA (18). This is in agreement with our structure-function results as these homologous stretches of sGC residues include the critical ␣1:Q368 and ␤1:Q309 as well as ␤1:I311, which also has a role in dimeriza- FIGURE 5. Possible subunit arrangement of sGC. Composite figure depicting possible subunit arrangement of heterodimeric sGC. The structure of the helical putative CC region is not known (labeled "?") as well as the structure of the ␣1 N-terminal domain. The H-NOX domain is oriented such that helix ␣F and loop L1 in H-NOX are in proximity to the site that corresponds to where G␣ s binds and regulates the homologous adenylyl cyclase (marked "X"). In H-NOX, both ␣F and the N-terminal helical subdomain (␣A-␣C), which includes the loop L1 containing the potential switch residues Asp 44 -Asp 45 (14,60), are postulated to shift upon activation (16). To illustrate the N-terminal subdomain shift, we have depicted the superimposed NsH-NOX (red) (16) and TtH-NOX (blue) (14) structures that are postulated to represent the basal and activated state, respectively, of an H-NOX domain. Note that we cannot rule out direct interactions between the H-NOXA domain and the GC catalytic domain, because we do not know the conformation of the intervening sequences and the structure and position of the CC region. tion (Fig. 4B). In addition to the H-NOXA interface, we had noted above that some of these residues might also provide key interactions with the downstream CC region. In further concord with our results, the deletion study found additional stretches of H-NOXA residues that either affected sGC dimerization and/or sGC activity (18) (Fig. 1B); all deletions contain residues that would correspond to the H-NOXA dimer interface or are between the H-NOXA and CC domains (deletions depicted in Fig. 1B). However, equivalent deletions in ␣1 and ␤1 have a different effect (except for the noted ␣1:363-372 and corresponding ␤1:304 -313). Deletions in ␤1-H-NOXA caused loss of dimerization, and thus all activity, whereas ␣1-H-NOXA deletions ⌬283-292 and ⌬373-382 lead to an increase in cyclase activity and decreased EC 50 for BAY41-2272 in the presence of 1 nM DEA/NO indicating that they are more readily and more potently activated (18). These differences from this deletion study are intriguing but need to be taken with the caveat that their system actually introduced an additional artificial dimerization interface as the ␣1 and ␤1 sGC subunits were each fused with a different half of YFP that only fluoresces when the halves come together (18). Nevertheless, these unexpected observed differences when equivalent stretches of ␣1 or ␤1 H-NOXA residues are deleted likely provides new insights into the role of the H-NOXA domains for the activation mechanism of sGC. We propose that the CC domains provide the main, although not sole, driving force for sGC dimerization, whereas the H-NOXA subdomain heterodimer possibly serves as a regulatory interface needed to bring along and position the adjacent NO-regulated H-NOX domain such that it can interact and regulate the catalytic GC subunits (48). This could explain why deletions within the ␤1-H-NOXA domain are deleterious for heterodimerization as they likely disrupt not only the H-NOXA:H-NOXA interface but also indirectly disrupt the H-NOX:GC interface because H-NOX can likely no longer be properly positioned by ␤1-H-NOXA. However, deletions of most stretches of residues within the ␣1-H-NOXA domain did not cause loss of dimerization but, unexpectedly, made sGC more active even at lower concentrations of BAY41-2272 ϩ 1 nM DEA/NO (18). We therefore speculate that ␣1-H-NOXA is a likely negative regulatory subdomain that holds ␤1-H-NOXA, and thus indirectly also the adjacent H-NOX domain, in an inhibitory position/conformation prior to NO activation. Disruption of this interaction by deletions in ␣1-H-NOXA could possibly relax the ␤1-H-NOXA/H-NOX subdomains such that it is now more susceptible to BAY41-2272/NO activation yielding also higher cyclase activity. Further studies are needed to determine whether the H-NOXA domains in sGC, or PAS domain in STHK, are indeed used for such regulatory, or even sensory, mechanisms. Nevertheless, this structural interpretation of the Rothkegel et al. (18) deletion data using our structural results suggests that targeting the H-NOXA dimer interface or putative PAS-ligand binding pocket could lead to new therapeutic sGC stimulators and activators. It is worth noting that this approach of targeting a PAS domain with no known ligand was successful for the PAS kinase domain, whose pocket is mainly closed as well, using the SARby-NMR approach (62).