Allostery in Recombinant Soluble Guanylyl Cyclase from Manduca sexta*

Soluble guanylyl/guanylate cyclase (sGC), the primary biological receptor for nitric oxide, is required for proper development and health in all animals. We have expressed heterodimeric full-length and N-terminal fragments of Manduca sexta sGC in Escherichia coli, the first time this has been accomplished for any sGC, and have performed the first functional analyses of an insect sGC. Manduca sGC behaves much like its mammalian counterparts, displaying a 170-fold stimulation by NO and sensitivity to compound YC-1. YC-1 reduces the NO and CO off-rates for the ∼100-kDa N-terminal heterodimeric fragment and increases the CO affinity by ∼50-fold to 1.7 μm. Binding of NO leads to a transient six-coordinate intermediate, followed by release of the proximal histidine to yield a five-coordinate nitrosyl complex (k6-5 = 12.8 s-1). The conversion rate is insensitive to nucleotides, YC-1, and changes in NO concentration up to ∼30 μm. NO release is biphasic in the absence of YC-1 (koff1 = 0.10 s-1 and koff2 = 0.0015 s-1); binding of YC-1 eliminates the fast phase but has little effect on the slower phase. Our data are consistent with a model for allosteric activation in which sGC undergoes a simple switch between two conformations, with an open or a closed heme pocket, integrating the influence of numerous effectors to give the final catalytic rate. Importantly, YC-1 binding occurs in the N-terminal two-thirds of the protein. Homology modeling and mutagenesis experiments suggest the presence of an H-NOX domain in the α subunit with importance for heme binding.

The primary form of sGC is an ␣1/␤1 heterodimer composed of two evolutionarily related subunits that display several recognizable domains (Fig. 1A) (18). Heme is bound to the protein through proximal ␤1 His-105. NO binding to the heme leads to proximal histidine release and stimulation of cyclase activity, presumably through a change in protein conformation (reviewed in Ref. 19). The heme-binding domain has evolved from a widespread family of bacterial proteins called H-NOX (Heme-Nitric oxide/OXygen) domain proteins, of which the structures of three are known (20 -22). The central portion of sGC contains two PAS domains (18,23), and the C-terminal region contains a catalytic domain that is very similar to that of adenylyl cyclase (24,25).
In the 1990s, the anti-platelet activity of YC-1, a benzylindazole derivative (Fig. 1B), was found to derive from its ability to bind to and stimulate sGC (26,27), leading to a search for related compounds that might serve as sGC-targeted drugs for human health (28). The YC-1 mechanism of action remains unclear, as does the location of its binding site in the sGC protein. The nucleotide-like structure of YC-1, in conjunction with mutagenesis studies, has led to the suggestion that YC-1 binds to the cyclase domain (29), whereas cross-linking studies with YC-1-related compounds have indicated that they bind to the N-terminal domain of the ␣1 subunit (30).
Studies of insect sGC have lagged behind studies of the mammalian enzyme, but, as in mammals, insect sGC plays an important physiological role. In the Manduca sexta larva (tobacco hornworm), sGC is implicated in antennal lobe morphogenesis (31)(32)(33), and in the Manduca adult (hawkmoth), sGC is a key component for olfactory processing of odors (34 -37). The development of insect cell expression systems for mammalian sGC proteins has facilitated mechanistic studies (38 -40), but obtaining functional material in sufficient quantity remains a major difficulty in the field. Here we have developed Escherichia coli expression systems for both the full-length Manduca sGC heterodimer (msGC) and for two heterodimeric, N-terminal truncations (msGC-NT1 and msGC-NT2). Our analysis provides the first mechanistic results for an insect-derived sGC. We demonstrate that Manduca sGC behaves much like its mammalian counterparts and that binding of YC-1 occurs within the N-terminal portion of the protein.
Mutations in the H-NOX domain were introduced by PCR using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA), with the msGC-NT2 plasmid as the template. The primers used to generate the ␣1 L211A mutant were 5Ј-gaaccagtggcgtacgctttagtaggcagtctgaaag-3Ј and 5Ј-ctttcagactgcctactaaagcgtacgccactggttc-3Ј. For the ␣1 T223A mutant, primers were 5Ј-gccatagcgaaacgactggctgatacacagacagac-3Ј and 5Ј-gtctgtctgtgtatcagccagtcgtttcgctatggc-3Ј. Mutations were confirmed by complete sequencing of the genes.
Expression of soluble msGC-NT1 or -NT2 was carried out in the presence of 25 M ␦-aminolevulinate, a heme synthesis precursor. Cells were grown at 30°C, induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside, and harvested 6 h later.
Purification of Recombinant Manduca sGC from E. coli-All purification steps were performed at 4°C. Cell pellets were suspended in lysis buffer (50 mM NaPO 4 , pH 7.0, 300 mM NaCl, 25 mg/ml DNase I, 2 mM MgCl 2 , 1 mM phenylmethylsulfonyl flu-oride, and 1 mM benzamidine) and disrupted using a French press (1000 p.s.i.). Cell debris was pelleted by low speed centrifugation (GSA rotor, 12,000 rpm for 30 min) and ultracentrifugation (45Ti rotor, 40,000 rpm for 30 min). The clear supernatant was then loaded onto metal affinity resin (Clontech Talon cobalt or Qiagen Ni-NTA (Valencia, CA)) pre-equilibrated with phosphate-buffered saline (50 mM NaPO 4 , 300 mM NaCl, pH 7.0). The column was washed with 20 bed volumes of low concentration imidazole in phosphate-buffered saline or EDTA, and the protein was eluted with 150 mM imidazole in phosphate-buffered saline (cobalt) or 100 mM EDTA (Ni-NTA). Brown-colored fractions were pooled, concentrated, and loaded onto a Sephacryl S-200 size-exclusion column (GE Healthcare). Both msGC-NT1 and -NT2 proteins eluted as a major peak at ϳ115 ml and were concentrated and buffer-exchanged into protein buffer (50 mM KPO 4 , pH 7.4, 100 mM KCl, and 5% glycerol) using a Vivaspin 6 concentrator (Sartorius Corp., Edgewood, NY). Protein concentration was estimated by absorption of the Soret band (see below). The procedure yielded 1-2 mg of highly pure protein per liter of cell culture. The proteins were stored at Ϫ80°C.
Full-length msGC was expressed at a lower temperature (17°C) in Rosetta(DE3) pLysS cells and purified in a similar manner as msGC-NT1 and -NT2, except that only a single metal-affinity column purification step was used because of difficulty with stability. The procedure yielded 0.5-1 mg of partially pure protein per liter of cell culture.
UV-visible Spectroscopy-Spectra were recorded on a Cary Bio50 spectrophotometer at room temperature at a scan rate of up to 600 nm/min. In a typical NO binding experiment, a 1-ml sGC sample was deoxygenated in a septum-capped cuvette with an argon stream (30 min), while stirring with a stir bar, before DEA/NO was added using a gas-tight syringe. A DEA/NO stock solution was prepared fresh in 10 mM NaOH and quantified by its absorbance at 250 nm using the extinction coefficient ⑀ 250 ϭ 8000 M Ϫ1 cm Ϫ1 , a value adjusted to account for incomplete release of NO (41). Complete degradation of DEA/NO was assumed to release two molecules of NO after 10 min. For CO binding, the sGC sample was purged with CO gas for 15-20 min before the spectrum was recorded.
Heme Soret Extinction Coefficient-Molar extinction coefficients were measured using the pyridine hemochromogen assay as described previously (42). The protein solution (700 l) was mixed with 100% pyridine (300 l), 5 N KOH (20 l), and crystals of dithionite. The peak absorbance at 556 nm minus that at 700 nm was used to determine the hemin concentration, assuming an extinction coefficient of 32 mM Ϫ1 cm Ϫ1 . NO and CO complex Soret band and all Q-band extinction coefficients were estimated by their ratio to the unliganded Soret band.
CO Dissociation Constants-The msGC-NT samples were placed in a septum-capped cuvette with minimal head space at room temperature (22°C). Aliquots from CO-saturated protein buffer (50 mM KPO 4 , pH 7.4, 100 mM KCl, and 5% glycerol), assumed to be 1 mM in CO, were added to the cuvette and stirred for 10 min, and the spectrum was measured. When present, nucleotide (0.5-1 mM) or YC-1 compound (50 M) was added before addition of CO. CO binding was measured by the shift in Soret band after accounting for dilution because of the addition of CO. This shift was estimated as A 424 -A 437 times the dilution factor, except in the presence of YC-1, where A 422 was used rather than A 424 . Data were fitted to a single-site saturation ligand binding model using SigmaPlot (SPSS, Inc., Chicago). For the wild-type protein in the presence of YC-1, titration was also undertaken in a cuvette with a 10-cm path length, using an RSM-1000 spectrophotometer (Olis, Inc., Bogart, GA), which yielded a value indistinguishable from that measured with the 1-cm cuvette.
Kinetics for Proximal Histidine Release-The rates for release of ␤1 His-105 upon NO binding to msGC-NT were measured at 10°C by mixing msGC-NT and NO in an RSM-1000 stopped-flow spectrophotometer (OLIS, Inc.). Samples of msGC-NT (0.7-2 M) were prepared by first deoxygenating protein buffer through bubbling of argon gas for 10 min, followed by addition of protein and additional deoxygenation with an argon stream placed above the solution for ϳ30 min. The protein solution was then transferred to the instrument in a gas-tight syringe. NO solutions were prepared by addition of DEA/NO from a stock solution to argon-purged protein buffer in a gas-tight syringe and then connected to the stopped-flow device. Decomposition was allowed to proceed for 20 min at room temperature before transfer to the instrument, where the solution was allowed to equilibrate to the desired temperature (5 min). Absorbance changes (A 420 ) were fitted to single or double exponential equations using SigmaPlot; values reported are the average and standard deviation of 5-7 consecutive measurements. For experiments with nucleotides or YC-1, the compounds were pre-mixed with the protein sample.
Kinetics of NO Release-Rates for denitrosylation of msGC-NT-NO were estimated using a dithionite/CO trap, as described (7,43). A slight excess of NO (from DEA/NO) was added to an anaerobic msGC-NT sample prepared as described above, and a spectrum was measured to ensure saturation. Nucleotide or YC-1 was pre-mixed with the protein sample before NO addition. The trapping solution was prepared by bubbling CO gas (10 min) into a freshly prepared dithionite solution (Na 2 S 2 O 4 , 60 mM in protein buffer). Denitrosylation was initiated by mixing the trapping and protein solutions either in a stopped-flow device at 20°C, or in a cuvette at room temperature (22°C), and monitoring the change in absorbance (A 424 -A 413 or A 424 -A 412 ). Rate constants were obtained by fitting the relevant time interval to a single or double exponential equation, as appropriate. For the cuvette data, the first 100 s were discarded to remove the fast phase, which was better estimated in the stopped-flow experiment.
Kinetics of CO Binding and Release-CO binding rates were measured for a series of CO concentrations (0.05-0.5 mM) in a stopped-flow spectrophotometer by monitoring absorbance change (A 424 -A 412 ) and fitting k obs versus [CO], which displayed the expected linear dependence for a monophasic process. Protein-and CO-containing solutions were prepared as described above. The second-order rate constants reported are for the slope and error from the linear fit.
CO release rates from msGC-NT-CO were measured by replacing released CO with excess NO upon rapid mixing in the stopped-flow device. One syringe contained msGC-NT-CO, and the second contained protein buffer saturated with NO from DEA/NO (ϳ2 mM). All reactions were performed at 20°C, and YC-1, where included, was pre-mixed with the protein sample.
Guanylyl Cyclase Enzymatic Assay-The cGMP producing activity of msGC in cell lysates and metal-affinity column elution fractions was measured using an enzyme immunoassay kit (Cayman Chemical Co.), following the manufacturer's instructions. In a typical assay, 10 l of reaction buffer (0.5 M HEPES, pH 7.5, 30 mM GTP, 60 mM MgCl 2 , 20 mM dithiothreitol) was added to protein sample for a total reaction volume of 100 l. The mixture was incubated at room temperature for 10 min and then quenched with 200 l of 250 mM zinc acetate and 200 l of 250 mM sodium carbonate. For experiments measuring NO-activated enzyme activity, protein samples were pre-mixed with DEA/NO before initiating catalysis by addition of reaction buffer.
Molecular Modeling of N-terminal msGC-Sequences of the ␣1 and ␤1 subunits of msGC were submitted to the Robetta structure prediction server for domain analysis using Ginzu (44). Three domains were predicted for each subunit as follows: an N-terminal domain predicted to be an H-NOX domain (20 -22) (pdb-blast, confidence ϭ 30 and 36 for the ␣1 and ␤1 subunits, respectively), a largely helical middle domain, and a C-terminal cyclase domain with homology to the catalytic domain of adenylate cyclase (pdb-blast confidence 52 and 47). A 25-residue sequence between the second and third domains in each subunit was identified as a possible coiled-coil region. The 50 N-terminal residues of the ␣1 subunit, which have no counterpart in the ␤1 subunit, were predicted to be disordered. Ginzu did not predict the PAS domains.
After domain analysis, the sequence of the ␣1 N-terminal domain (minus the 50 N-terminal residues predicted to be disordered) was submitted to the 3D-Jury Meta Server for structure prediction and initial model building (45). A template structure file was not specified. C-␣ models and alignments of ␣1 H-NOX with known H-NOX structures (Protein Data Bank entries 1XBN (21) and 2O09 (20)) were returned from several severs. All had very significant 3D-Jury scores (100 -108). Fullatom models were built from these alignments using Modeler (46). The resulting models were nearly identical, varying only in the placement of a short insertion near the vacant heme cavity. The final model was minimized using NAMD (47).

Expression and Purification of Heterodimeric N-terminal
Fragments of Manduca sGC-Studies on NO signaling have been limited by difficulties in obtaining sufficient quantities of intact sGC. To overcome this shortcoming, we pursued functional domains of sGC from M. sexta. Sequence analyses and homology modeling (described below) suggested boundaries for the C-terminal cyclase domains and N-terminal H-NOX domains of both subunits. Two constructs were prepared, msGC-NT1 (␣ 1-471 /␤ 1-400 ), which lacks the cyclase domains and is expressed from a single plasmid, and msGC-NT2 (␣ 49 -471 /␤ 1-400 ), which is identical to msGC-NT1 except that it also lacks the putatively disordered N-terminal region of the ␣1 subunit (Fig. 1A). Both constructs produced soluble heterodimeric protein in an E. coli expression system, and both displayed identical functional behaviors, suggesting the ␣1 subunit N-terminal His tag did not affect activity; however, msGC-NT2 was the more stable of the two. Co-expression of both subunits and addition of heme precursor ␦-aminolevulinate was required to produce soluble recombinant protein; expression of individual subunits or expression without ␦-aminolevulinate led only to insoluble inclusion bodies.
Purification using the ␣1 subunit N-terminal His tag was complicated by a tendency of the protein to lose heme. This tendency was exacerbated by the imidazole in the elution buffer, because imidazole can coordinate to heme and facilitate its removal. Heme loss has long been known to occur in sGC during isolation. To circumvent this difficulty, we developed a purification procedure where sGC was released from the nickel-affinity column by the addition of EDTA, which chelates the nickel and releases the protein from the column. Rapid purification over nickel-affinity and gel filtration col-umns yielded ϳ1 mg of Ͼ90% pure protein/liter of cell culture ( Fig.  2A). The protein behaved as a heterodimer by size-exclusion chromatography and typically displayed an A 433 /A 280 ratio of 1.2. Best stability for the protein was achieved under anaerobic conditions and bound to CO, where little change was observed over several days. Soret band degradation is seen with time in the presence of the reductant dithiothreitol but not in the presence of tris(2-carboxyethyl)phosphine.
Truncated Manduca sGC Displays Typical Electronic Spectra-Truncated Manduca sGC heterodimers display Soret and Q-band absorption maxima that are typical for sGC proteins (Fig. 2B). The isolated protein has a broad Soret maximum centered at 433 nm, consistent with a high spin ferrous heme center. Binding of CO, which requires ferrous heme, shifts the Soret maximum to 425 nm and sharpens both the Soret and Q-bands, consistent with a low spin six-coordinate heme. The high purity of the protein allowed us to determine molar extinction coefficients for these bands, using the pyridine hemochromogen assay ( Table 1). The resulting values were in good agreement with those reported for globins, nitrophorins, and full-length and ␤1 truncated mammalian sGC proteins (42,48).
YC-1 Binding Increases the CO Affinity of msGC-NT-Small molecule effectors of sGC catalytic activity, including ATP, GTP, and the compound YC-1, have been described, but the binding site(s) for these molecules and their mechanisms of action remain unclear. The interface of the ␣1/␤1 catalytic domains in sGC contains both a catalytic site and a second pseudo-symmetric site that has been suggested to be a regulatory binding site much like the forskolin-binding site in the related protein, adenylyl cyclase (10,29,49,50). In contrast, cross-linking studies with the YC-1 related molecule BAY 41-2272 indicated binding was near ␣1 residues Cys-238 and Cys-243, well away from the cyclase domain (30).
To address this issue, we determined the effect of YC-1, ATP, and GTP on CO binding to msGC-NT1 and msGC-NT2, neither of which contains the sGC catalytic domain. CO binding to the full-length protein has been shown previously to increase dramatically in the presence of YC-1, leading to strong stimulation of sGC activity. CO binding to msGC-NT1 (not shown) and msGC-NT2 (Fig. 3) was monitored spectroscopically. Binding was similarly modest for both proteins in the absence of YC-1 and increased dramatically upon addition of YC-1 ( Fig.  3 and Table 2). Addition of ATP and GTP (Table 2) or cGMP (not shown) had little effect on CO binding. We conclude that the major YC-1-binding site lies in the N-terminal two-thirds of the protein, away from the catalytic domain.
Kinetics of NO-induced ␤1 His-105 Release from Heme-The hallmark of sGC activation by NO is the release of ␤1 His-105 from the heme, leading to allosteric stimulation of cyclase activity. Elucidation of the details of this step has been hampered by insufficient protein and remains unresolved and controversial in the literature (7, 8, 19, 51). Of particular importance is the

FIGURE 2. SDS-PAGE and UV-visible absorption spectra for msGC-NT2.
A, purified msGC-NT2 (4 g) visualized by SDS-PAGE (12%, stained with Coomassie Blue dye). B, electronic absorption spectra of purified msGC-NT2 (yellow), and its complexes with CO (blue) and NO (red). Inset shows the Q-band region of the spectra.  (20) a Wavelengths are given in nanometers, and extinction coefficients are shown in parentheses (mM Ϫ1 cm Ϫ1 ). The extinction coefficient for the unliganded Soret band was determined from direct measurement of heme content, with the error estimated from the standard deviation of four measurements. All other values were derived from the ratio of the absorption maximum to that of the unliganded Soret band. b ND indicates not distinct. sGC from Manduca sexta possibility that, beyond the NO-binding site on the heme, there is a second NO-binding site that affects both ␤1 His-105 release kinetics and allosteric stimulation of full-length sGC. We examined ␤1 His-105 release using stopped-flow spectroscopy ( Fig. 4 and Table 3). Mixing of NO and msGC-NT1 led to rapid formation of a six-coordinate intermediate with an apparent Soret maximum of 420 nm. Because this step largely occurred in the mixing dead time (ϳ3 ms), the associated rate could not be accurately measured in our device. For the rat protein, this rate has been estimated to be Ͼ1.4 ϫ 10 8 M Ϫ1 s Ϫ1 at 4°C (52). Decay of the six-coordinate intermediate was relatively slow for msGC-NT1, however, and was readily monitored. For moderate NO concentrations, ␤1 His-105 release was ϳ12 s Ϫ1 under all conditions. GTP, ATP, and YC-1 addition had no effect on this step (Table 3) nor did the kinetics of ␤1 His-105 release differ between msGC-NT1 and msGC-NT2 (not shown). Only under higher NO concentrations (Ͼ30 M) was a change observed, leading to a small ϳ2-fold increase in the release rate. Thus, the histidine release kinetics of msGC-NT exhibit neither the large NO concentration dependence described for the fulllength rat protein (k ϭ 2.4 ϫ 10 5 M Ϫ1 s Ϫ1 (52, 53)) nor the sensitivity to nucleotide binding described for the full-length bovine protein (8).
YC-1 Decreases the Rate of NO Release from msGC-NT-We examined NO release from msGC-NT1 to further investigate the role of small molecule effectors on the protein and to uncover which step(s) in binding and release are altered by YC-1 binding. YC-1, ATP, and GTP have all been reported to alter the NO release rates from sGC (7, 54). To measure NO release, we mixed excess CO and dithionite with msGC-NT1-NO, allowing CO to replace NO while preventing NO rebinding through reaction with dithionite. In similar experiments with full-length rat (55) and bovine (54) sGC, binding of YC-1 or GTP led to an increase in NO release rates. In the experiments with rat sGC, two release phases were detected. Using msGC-NT1, we also found multiphasic release rates, but we were unable to capture the fastest of these through simple mixing in a cuvette, in contrast to the previously reported studies. Therefore, we employed both stopped-flow and cuvette-based spectrophotometric measures of release.
The resulting NO release behavior and associated rate constants are shown in Fig. 5 and Table 4. By itself, msGC-NT1 displayed two prominent phases by stopped-flow analysis, a very fast phase with rate constant of 0.1 s Ϫ1 (56% of the total amplitude) and a slower phase with rate constant of 0.0066 s Ϫ1 (44%). Analysis in a cuvette, after allowing for decay of the fast phase, was best fit with a single exponential, yielding a more robust value for the second phase of 0.0015 s Ϫ1 . This value is ϳ5-fold smaller than that reported for the first phase of the rat protein (55) and ϳ3-fold larger than that of the single phase reported for the bovine protein (54). Addition of GTP or ATP had very little effect on msGC-NT1 rate measurements, whereas GTP or YC-1 greatly accelerated NO release for the full-length mammalian proteins (7,54). To our surprise, addition of YC-1 to msGC-NT1 had a profound effect on NO release; the fast phase was completely eliminated leaving only the second phase in place. This result is clearly shown in plots of the absorbance change associated with CO formation; in the stopped-flow experiment, msGC-NT1-CO formation was markedly delayed when YC-1 was present (Fig. 5A), whereas in the cuvette experiment, the amplitude for the absorbance change because of msGC-NT1-CO formation was much greater in the presence of YC-1, because the loss of NO was not diminished by the fast phase, which was not observable in the cuvette experiment.
YC-1 Decreases Rates for CO Release from msGC-NT-We also examined CO binding and release to further characterize changes in the protein. CO release from msGC-NT1-CO was measured in a stopped-flow device by replacement with NO, which binds more quickly and more tightly to the protein. CO release was faster than NO release; nonetheless, the release rate was decreased dramatically (by ϳ10-fold) by the addition of YC-1, much like with NO release (Table 5). In contrast to NO, however, CO had only one detectable release phase.
CO binding, unlike NO binding, was sufficiently slow for measurement in the stopped-flow device. The rate constants for CO binding were similar in the presence and absence of YC-1 ( Table 5). The k off /k on ratios were similar to the measured values for K d (71 versus 77 M in the absence of YC-1 and 8  versus 2 M in the presence of YC-1), indicating that our analysis strategy was valid. It should be noted that, for full-length sGC from bovine lung, Stone and Marletta (56) have reported that YC-1 binding has no effect on CO affinity or release rates, although Kharitinov et al. (57) have reported that YC-1 binding causes CO to bind 10-fold more tightly. Functional Expression of Full-length Manduca sGC in E. coli-We were also able to express full-length ␣1/␤1msGC in E. coli. As with the truncated protein, we expressed both subunits of the full-length protein from a single plasmid. Partial purification was achieved using an ␣1 N-terminal His tag, and yielded 0.5-1 mg of sGC protein per liter of cell culture, as estimated by the heme Soret absorption band. This value is likely an overestimation of total active protein, however, because the ␣1 subunit is extremely protease-sensitive and is partially degraded in our preparations.

TABLE 2 CO dissociation constants for msGC-NT2 measured by equilibrium titration
Cyclase activity for msGC was measured using an enzymelinked immunoassay for the detection of cGMP (Fig. 6). The partially purified protein displayed a basal activity of 6.3 nmol of cGMP mg Ϫ1 min Ϫ1 and a maximal activity of 1058 nmol of mg Ϫ1 min Ϫ1 , values that are 6 -12-fold smaller overall than those reported for mammalian sGC proteins (7,8,29). The lower value we obtained is most likely because of having a mixture of fulllength and ␣1-degraded material in the preparation, leading to overestimation of total intact protein from the Soret band absorption. However, the possibility that the intact Manduca protein is inherently less active then its mammalian counterparts cannot be ruled out.
NO stimulation of msGC was similar to that of the mammalian proteins. Typically, NO stimulates mammalian sGC by 100 -200-fold; in our study, NO stimulated the recombinant Manduca sGC by 135fold alone and by 175-fold in the presence of YC-1 (Fig. 6). Importantly, YC-1 alone was also a potent stimulator of catalytic activity (Fig.  6). For maximal activity, msGC required both NO and YC-1, as has been generally reported for the mammalian proteins (29, 58 -60), although one group has reported that only CO and YC-1, not NO and YC-1, are synergistic (56).
Mammalian sGC is inhibited by ATP, suggesting that the nucleotide is an allosteric regulator of the protein (7, 10, 61). When we examined inhibition of msGC, we found that 1 mM ATP, a physiologically relevant concentration (62), inhibited the enzyme by ϳ70% in the presence of stoichiometric NO concentrations (Fig. 6). Overall, msGC behaves much like its mammalian counterparts, indicating our results with msGC-NT1 and msGC-NT2 are generally applicable for the entire sGC family.
Modeling sGC Functional Domains-The foregoing data confirm that YC-1 binding occurs in the N-terminal two-thirds of msGC. We reasoned that the regulatory effect of YC-1 might occur through binding to the ␣1 subunit in the region homologous to the ␤1 heme-binding domain, which has recently been recognized to have evolved from an ancient prokaryotic hemoprotein called H-NOX (20 -22). We examined the domain structure of ␣1, searching for known folds using the Ginzu protocol as implemented in the Robetta server (see "Experimental Procedures" for details). This led to the prediction that ␣1 has an H-NOX fold between residues 51 and 247. A homology model of residues 61-234 based on Protein Data Bank entries 1XBN (21) and 2O09 (20) yielded a satisfactory fold with a dope score similar to that of the template structure 1XBN (Fig. 7).
Based on this result, we hypothesized that a small molecule binding pocket might reside in the site equivalent to the hemebinding site in the H-NOX domain and that conserved residues   sGC from Manduca sexta JULY 25, 2008 • VOLUME 283 • NUMBER 30 in this region, such as ␣1 Leu-211 and Tyr-223 (Fig. 7), might be of importance in YC-1 binding. Therefore, we produced the msGC-NT2 mutants ␣1 L211A and Y223A. Both proteins were expressed in soluble form; however, heme loading was reduced in both mutant proteins, suggesting that a direct contact between ␣1 and heme might have been lost. YC-1 stimulation of CO binding was unaltered for the heme-intact portion of both mutated proteins (Table 2), and the precise location of YC-1 binding therefore remains unknown.

DISCUSSION
We have expressed functional heterodimeric full-length and truncated sGC from the tobacco hornworm (M. sexta) in E. coli, the first time this has been accomplished for sGC from any species. Both express as soluble, heterodimeric proteins with intact ferrous heme. The partially purified full-length protein  3.1 Ϯ 0.3 27 Ϯ 6 1.9 Ϯ 0.01 17.9 Ϯ 0.04 a Ligand concentrations are as follows: ATP, 1 mM; GTP, 0.5 mM; YC-1, 50 M. b Data were measured by CO replacement/NO reduction with dithionite in a stopped-flow spectrophotometer. The data were fitted to a double exponential, k 1 and k 2 with amplitudes a 1 and a 2 . Errors are from the fitting algorithm. For experiments including YC-1, only the slower phase was observed. c Data were measured by CO replacement/NO reduction with dithionite in a cuvette. The data were fitted to a single exponential after 100 s to remove the faster phase.  The production of cGMP from GTP by partially purified msGC was measured using an enzyme immunoassay. The msGC concentration was estimated by Soret band absorbance using the extinction coefficient measured for msGC-NT2. Each point is the average of 3-4 measurements.

sGC from Manduca sexta
shows behavior similar to its mammalian counterparts, including stimulation by NO and compound YC-1, and inhibition by ATP (Fig. 6). The specific activity of this material is 6 -12-fold lower than that of its mammalian counterparts, most likely because of the presence of proteolytically degraded material in the preparation that retains heme but not catalytic activity. Truncated msGC, which contains the N-terminal two-thirds of both the ␣1 and ␤1 subunits but not the catalytic domains ( Fig.  1), binds YC-1, which decreases release rates for both NO (Table 4) and CO (Table 5) and increases the binding affinity for CO (Table 2) and presumably NO (the NO dissociation constant is too small to be readily measured). In contrast, ATP, GTP, and cGMP have little effect on NO and CO binding to msGC-NT. Taken together, these results indicate that a specific binding site for YC-1 and related compounds lies away from the catalytic domain and is distinct from the allosteric site proposed for GTP, ATP, cGMP, and pyrophosphate (PP i ). Below, we explore the implications of these results. YC-1 Binding Alters the Heme Pocket-In the absence of YC-1, NO release from msGC-NT is biphasic, whereas in the presence of YC-1, release is monophasic because of loss of the fast phase (Table 4). Similarly, CO release from msGC-NT is 10-fold faster in the absence of YC-1 than in its presence, but only a single phase is observed (Table 5). YC-1 binding also leads to an ϳ2-nm blue shift in the msGC-NT-CO heme Soret band (not shown), as reported previously for the full-length bovine protein (57), indicating a change in the heme pocket takes place. A Soret band shift is not observed in the absence of CO or in the presence of NO. Additionally, photolysis of CO from msGC-NT-CO in the presence of YC-1, but not in its absence, displays a large geminate recombination phase, where the photolyzed CO becomes trapped in the protein matrix and rapidly recombines with heme rather than escaping into bulk solvent. 3 Taken together, these data suggest that YC-1 binding closes the msGC-NT heme pocket such that access to solvent is reduced.
An alternative possibility is that YC-1 binding changes electrostatic stabilization of the sGC NO and CO complexes. Elegant studies on myoglobin by Olson and co-workers (63,64) indicate that dioxygen release is particularly sensitive to changes in distal pocket hydrogen bonding, and furthermore, discrimination between dioxygen and CO in myoglobin depends heavily on electrostatic stabilization of heme-bound dioxygen. However, their studies indicate that electrostatic effects have little influence on the release rates of NO and CO because of the apolar nature of the FeNO and FeCO moieties (63,64). Thus, our results fit better with a model invoking steric trapping of the molecules.
One of two mechanisms is likely to give rise to a blocked heme pocket. First, direct blockage of the distal pocket by YC-1 would be consistent with the kinetic data and has been suggested previously (65), but it is inconsistent with the limited effects on the heme absorption spectra. Alternatively, and more likely in our view, is an allosteric model wherein YC-1 binding leads to a change in protein conformation that closes the distal pocket. That such a mechanism is possible despite the propensity for NO and CO to diffuse away is illustrated by the nitrophorins from the kissing bug (Rhodnius prolixus), which are used for NO transport during blood feeding (66). Binding of NO to the Rhodinius nitrophorin heme leads to desolvation of the distal pocket, due in part to the hydrophobic nature of the NO molecule, and to a large change in loop conformation (67), which hinders escape of NO from the heme (68 -70). A similar mechanism can be envisioned for sGC.
Functional Interaction between ␣1 and ␤1 H-NOX Domains-The location of the YC-1-binding site in the N-terminal twothirds of sGC is still unknown. To aid in the discovery of the binding site, we undertook molecular modeling, which led us to conclude that the ␣1 N-terminal region contains an H-NOX domain that is much like that for ␤1, but lacking bound heme ( Figs. 1 and 7). Although the ␣1 domain is less conserved than the rest of the protein, a few conserved residues are present, including Phe-157, Leu-211, and Tyr-223. In the model, these residues lie in the region analogous to the ␤1 heme-binding pocket (Fig. 7) and are near ␣1 Cys-238 and Cys-243 in the human sGC sequence, which can be modified through photoaffinity labeling with a YC-1 analogue (30).
We reasoned that ligand binding in the ␣1 H-NOX pocket might have a regulatory effect much like that of heme binding in the ␤1 H-NOX domain, because of the symmetric domain structure of the protein, and therefore the ␣1 H-NOX pocket might represent the YC-1-binding site. We generated ␣1 mutant proteins L211A and Y223A to test this possibility. Neither mutation affected YC-1 binding ( Table 2), but to our surprise, both mutant proteins lost heme affinity and could not be isolated with a fully intact heme pocket, suggesting that the ␣1 and ␤1 H-NOX domains contact one another and possibly even share heme binding. Interestingly, the recombinant rat ␤1 H-NOX domain exists as a homodimer in solution (71), consistent with our conclusion that the ␣1/␤1 dimer shares an H-NOX interface. The binding site for YC-1, however, remains unknown.
Relation to Full-length Mammalian sGC-Marletta and coworkers (7,55) and Russwurm and Koesling (8) have demonstrated that NO stimulation of mammalian sGC is increased in the presence of GTP (Marletta and co-workers) or Mg 2ϩ / cGMP/PP i (Russwurm and Koesling), presumably through conformational linkage of the cyclase and H-NOX domains. Additionally, Marletta and co-workers (7) and Sharma and coworkers (54) have reported that the off rates for NO increase in the presence of GTP. For msGC-NT, we saw no such effect with GTP (Table 4) or cGMP (not shown), indicating functional binding for these ligands most likely occurs only in the cyclase domain.
Mammalian sGC has been shown to have both high and low output forms that display spectrally identical five-coordinate Fe-NO complexes (7,8,55). Russwurm and Koesling (8) have argued that the high and low output forms of sGC arise from two different Fe-NO binding geometries, one with NO bound to the distal side of the heme (high output) and one to the proximal side (low output), displacing ␤1 His-105, in a geometry reminiscent of that for nitrosylated cytochrome cЈ (72). Their data indicate that formation of the low output form only occurs in the absence of substrates or products. Marletta and co-workers (7,55) have argued that high and low output forms of sGC represent two different conformations of the protein, differing in both catalytic rate and NO release rates and arising from a second NO-binding site that influences the rate of ␤1 His-105 release (52). In our hands, decay of the msGC-NT sixcoordinate intermediate was quite fast (k 6-5 ϭ 12.8 s Ϫ1 , see Table 3) and independent of NO concentration at moderate NO levels (Fig. 4), consistent with formation of the high output form of full-length sGC without binding of a second NO molecule or influence from nucleotide binding. We conclude that binding of a regulatory nucleotide or a second NO molecule, if they occur for msGC, must therefore take place in the C-terminal catalytic domain.
It should also be noted that, curiously, the opposite effect of YC-1 binding on NO release was recently found by Marletta and co-workers (55) for the full-length rat protein. In that study, two phases were also described for NO release, but YC-1, like GTP, increased the proportion of the faster phase and, to a lesser extent, increased the rate constants themselves. The authors proposed a model quite similar to ours, but opposite in direction; they suggest that GTP and YC-1 bind to the equivalent site in the catalytic domain, leading to an open conformation with faster release kinetics. In contrast, Koesling and coworkers (49,58), by monitoring cGMP production in the presence of an NO scavenger, inferred decreased NO off rates for full-length bovine lung sGC when bound to YC-1, much as we see with msGC-NT, and Kharitinov et al. (57) reported that CO binds 10-fold more tightly to YC-1-ligated bovine lung sGC, again as we found for msGC-NT. Nonetheless, that a second YC-1-binding site exists in the cyclase domain is suggested by mutations to the catalytic domain that alter the response to the ligand (29,49). Clearly, additional studies are required to resolve this complicated issue.
A Two-state Model for sGC Activation-In conclusion, we have demonstrated that msGC behaves much like its mammalian counterparts, and that YC-1 binding occurs in the N-terminal two-thirds of the protein and leads to a change in heme pocket conformation that is consistent with distal pocket closure. The simplest model for activation that emerges from these data is one in which the protein moves between two predominant conformations, "high output" and "low output," such that multiple effector molecules, binding in multiple locations, can all shift the protein between the two states. In this respect, effector molecules for sGC act much like O 2 ; CO 2 and bisphosphoglycerate act on hemoglobin, where each molecule affects the equilibrium between R and T states but do so through binding, coordinating, or reacting at distinct sites on the protein. As with hemoglobin, the ever-expanding list of molecules suggested to influence sGC, including non-heme NO, CO, GTP, ATP, Mg 2ϩ /cGMP/PP i , YC-1, phosphorylation, nitrosylation, and protein partners, may all feed into a single conformational change pathway, with the sGC protein integrating all such signals to produce a final level of activity. In this model, release of ␤1 His-105 is unnecessary for sGC to achieve its high output conformation, but it is nonetheless favored when NO, with its strong trans effect, binds to the proximal heme site.