Structure and Reactivity of a Thermostable Prokaryotic Nitric-oxide Synthase That Forms a Long-lived Oxy-Heme Complex*

In an effort to generate more stable reaction intermediates involved in substrate oxidation by nitric-oxide synthases (NOSs), we have cloned, expressed, and characterized a thermostable NOS homolog from the thermophilic bacterium Geobacillus stearothermophilus (gsNOS). As expected, gsNOS forms nitric oxide (NO) from l-arginine via the stable intermediate N-hydroxy l-arginine (NOHA). The addition of oxygen to ferrous gsNOS results in long-lived heme-oxy complexes in the presence (Soret peak 427 nm) and absence (Soret peak 413 nm) of substrates l-arginine and NOHA. The substrate-induced red shift correlates with hydrogen bonding between substrate and heme-bound oxygen resulting in conversion to a ferric heme-superoxy species. In single turnover experiments with NOHA, NO forms only in the presence of H4B. The crystal structure of gsNOS at 3.2 AÅ of resolution reveals great similarity to other known bacterial NOS structures, with the exception of differences in the distal heme pocket, close to the oxygen binding site. In particular, a Lys-356 (Bacillus subtilis NOS) to Arg-365 (gsNOS) substitution alters the conformation of a conserved Asp carboxylate, resulting in movement of an Ile residue toward the heme. Thus, a more constrained heme pocket may slow ligand dissociation and increase the lifetime of heme-bound oxygen to seconds at 4 °C. Similarly, the ferric-heme NO complex is also stabilized in gsNOS. The slow kinetics of gsNOS offer promise for studying downstream intermediates involved in substrate oxidation.

Nitric-oxide synthases (NOSs) 2 are highly regulated proteins that catalyze the two-step oxidation of L-arginine to nitric oxide (NO) and citrulline via the stable intermediate N -hydroxy L-arginine (1)(2)(3). NO functions in mammals as a potent signaling molecule and a cytotoxic agent to protect against pathogens. Mammalian NOSs consist of a reductase domain that has binding sites for FAD, FMN, and NADPH and an oxygenase domain that binds iron protoporphyrin IX (heme), substrate L-arginine, and the cofactor (6R)-5,6,7,8-tetrahydro-L-biopterin (H 4 B) (1)(2)(3). Proteins similar to mammalian NOSs have been found in a variety of lower eukaryotes including insects, fungi (4 -7), and bacteria (8,9). Genes responsible for cofactor (H 4 B) biosynthesis are present in bacteria such as Bacillus subtilis and Geobacillus kaustophilus but not in other NOS-containing bacteria such as Deinococcus radiodurans (9 -11). Bacterial NOS-like proteins are similar to the oxygenase domain of mammalian NOSs but contain no associated reductase module. Reductase partners for bacterial NOSs are yet to be identified. NOSs from D. radiodurans, B. subtilis, Staphylococcus aureus, and Bacillus anthracis are well characterized and have been shown to produce nitrogen oxides (NO x ) in vitro (11)(12)(13)(14). Conservation of nearly all the key residues involved in substrate and cofactor binding among mammalian and bacterial NOSs (11,13,15) suggests a similar mechanism of NO formation in the two classes of proteins. Interestingly, D. radiodurans NOS can support L-Arg-based NO x formation with cofactors other than H 4 B, such as the ubiquitous cofactor tetrahydrofolate and even tryptophan (11,16). The ability to react with L-tryptophan may be significant as NOSs from certain Streptomyces strains participate in biosynthetic tryptophan nitration (17).
The NOS reaction sequence is well understood (Fig. 1), although the nature of the heme-oxygen complexes directly involved in substrate oxidation remains largely unknown (18). L-arginine is first hydroxylated at the guanidino nitrogen, and then the resultant N -hydroxy-L-arginine (NOHA), an enzyme-bound intermediate (19), is further oxidized to NO and citrulline. In both the L-arginine and NOHA reactions, reduction of the Fe(III) heme enables oxygen binding and formation of a heme-dioxygen complex, which is best described as a ferric superoxy species (Fe(III)-O 2 . ) (1, 20 -23). This intermediate does not react with L-arginine but may (24) or may not (25) react with NOHA. The addition of oxygen to reduced eNOS forms two distinct heme-oxy species, which have been interpreted as the ferrous-dioxygen complex and the ferricsuperoxy complex (26). H 4 B acts as an electron donor to the ferric-superoxy species in both steps of NO synthesis (Fig. 1). In the first step the reductase domain reduces the H 4 B . ϩ radical (18,(27)(28)(29)(30)(31)(32)(33). In the second step a downstream reaction intermediate (25), possibly a ferrous-heme NO complex (34), reduces the H 4 B . ϩ radical. Reduction of the Fe(III)-O 2 . species at cryogenic temperatures results in a ferric heme-peroxo species that rapidly reacts at higher temperatures with either L-arginine or NOHA to form products (35). Unlike heme-oxygenases such as cytochrome P-450 or heme oxygenase (36,37), a ferric heme-hydroperoxo species has not been observed in cryo annealing experiments. Bacterial NOSs retain NO in their heme pockets for longer times compared with their mammalian counterparts (38). In B. subtilis NOS (bsNOS), the release of NO is 20-fold slower than that in mammalian NOSs due to a bacterially conserved Val to Ile switch, which offers more steric hindrance for the heme-bound NO to diffuse away from the heme. An Ile to Val mutation in bsNOS increases the rate of NO release 3.6 times, and a Val to Ile mutation in mouse iNOSoxy decreases the rate of NO release by 3 times (38). Thus, the NOS heme pocket can tune the reactivity of heme ligands.
Otherwise unstable reaction intermediates of cytochromes P-450, another class of well studied heme-containing monooxygenases, have been observed at cryogenic temperatures after radiolytic reduction of the heme (39 -41). Additionally, in cysteine-ligated heme proteins, such as P450cam and chloroperoxidase, rapid formation of Compound I ([Fe(IV)AO] . ϩ heme/thiolate radical) and related species can be achieved on reaction with peracids (42)(43)(44)(45). In these cases Compound I (absorption peak ϳ367 nm) forms in ϳ10 ms after rapidly mixing 3-chloroperbenzoic acid with the ferric enzyme. Within 40 ms, this species converts to an inactive product with a peak ϳ406 nm (42). In NOS, the heme-oxy species that react with L-arginine or NOHA have not been observed, although it is widely thought that the Compound I species is involved, in analogy to cytochrome P-450-type reactions (43,44,46). Thermophilic prokaryotes can be exploited as a source of thermostable enzymes that have slow reaction profiles at temperatures below 25°C (44). With this motivation, we have cloned, expressed, and characterized a thermophilic nitric-oxide synthase like protein from Geobacillus stearothermophilus (gsNOS). Herein, we characterize biophysical and biochemical properties of gsNOS and demonstrate its enhanced thermostability and slower reaction kinetics compared with other bacterial NOSs. We also show that the N-terminal extension contributes to this thermal stability. The most striking property of gsNOS is that it forms a stable heme-oxygen complex that persists on the timescale of seconds at 4°C. The crystal structure of gsNOS at 3.2 Å of resolution reveals very high similarity to the structure of bsNOS and provides insight into the slower reactivity of gsNOS.

EXPERIMENTAL PROCEDURES
Materials and Methods-Dioxane and sodium chloride were obtained from Mallinckrodt, (Ϯ)2-methyl-2,4-pentane diol from Hampton Research and Tris (hydroxymethyl) aminomethane from Fisher. All other chemicals were obtained from Sigma-Aldrich unless otherwise noted. All UV-visible spectra and kinetic data were recorded using an Agilent 8453 UV-visible spectroscopy system. Single-value decomposition (SVD) analysis was done using the program SPECFIT (47)(48)(49)(50).
Molecular Biology-The NOS gene of G. stearothermophilus (ATCC strain number 12980) was amplified by PCR from genomic DNA. The 5Ј primer generated an NdeI site before the start codon, and the 3Ј primer generated an XhoI site after the stop codon. The amplified fragment was cloned into the pET28 expression vector (Novagen) and transformed into Escherichia coli BL21(DE3) cells.
Protein Expression and Purification-The full-length NOS (gsNOS) and a shorter construct with the first 13 residues removed from the N terminus (gsNOSϩ13) were overexpressed in E. coli BL21(DE3) cells with a His 6 tag. The proteins were purified using nickel-chelate chromatography and then size-exclusion chromatography after removal of the His 6 tag with thrombin. Both the constructs could be concentrated to ϳ100 mg/ml, as estimated by the Bradford assay.
Structure Determination-Diffraction data were collected at 100 K with synchrotron radiation ( ϭ 1.002 Å) on beamline X-25 of the National Synchrotron Light Source at Brookhaven National Laboratory. The data sets were reduced and scaled using HKL2000 (51). Initial phases were determined by molecular replacement (AmoRe) (52) with the structure of NOS from B. subtilis as the probe (PDB entry 1M7V). The model was then refined in CNS (53) using standard positional and thermal factor refinement, and the structure was adjusted with XFIT to F obs Ϫ F calc and 2F obs Ϫ F calc maps. The addition of L-arginine, heme, and water molecules amidst cycles of refinement produced the final model. Nitrite Formation with Peroxide-30 -50 M enzyme was incubated at different temperatures with 1 mM L-arginine and 20 mM H 2 O 2 , and the reaction was stopped at different times by adding Griess reagents R1 (sulfanilamide) and R2 (N-(1-naphthyl)ethylenediamine). The product (pink dye) formation was monitored by measuring the absorbance at 540 nm. The amount of nitrite generated in the solution was calculated using the Griess assay kit from Cayman Chemicals. Each activity reported represents an average from at least three experiments.
Single Turnover Experiments-Concentrated full-length gsNOS was cycled through a degassing chamber and left in an anaerobic glove box for ϳ30 min. All buffer solutions were extensively degassed and placed under argon. Degassed buffer solution was used to dilute the protein solution suitably for UV-visible spectroscopy. To observe the spectral changes the enzyme undergoes during reaction with oxygen, gsNOS (13 M) was reduced by titration with sodium dithionite (30 -50 M). The cuvette containing reduced gsNOS (13 M) was sealed using a rubber septum inside the glove box and then transferred to a UV-visible spectrophotometer. The temperature of the cell with the sample in the UVvisible spectrophotometer was lowered to 4°C. Nitrogen gas was blown around the cuvette to prevent water condensation due to lowered temperature. Ice-cold air saturated buffer was injected into the cuvette through the rubber septum to start the reaction, at which point the sample contained ϳ8 M gsNOS, ϳ20 -30 M dithionite, and ϳ160 M oxygen (54). The solutions were mixed rapidly using a magnetic stir bar in the cuvette.

General Properties
gsNOS was PCR cloned and expressed with a His 6 affinity tag in E. coli. The protein has 65% sequence identity when compared with the NOS protein from the related mesophile B. subtilis (Supplemental Fig.  1). After nickel nitrilotriacetic acid affinity purification and proteolytic cleavage of the His tag, gsNOS (ϳ43 kDa/subunit) elutes on a gel filtration column as a dimer (apparent molecular mass ϳ86 kDa). The UVvisible spectra of free gsNOS (absorption maxima at 403 and 519 nm), imidazole-bound enzyme (427 and 553 nm), L-arginine-bound enzyme (399 and 517 nm), and reduced enzyme with L-arginine bound (415 and 552 nm) (Fig. 2) are very similar to those of mammalian and other bacterial NOS oxygenase domains (1, 11, 12, 20 -22). Shifts in the Soret peak indicate that the heme coordinates imidazole (427 nm) and can be displaced by L-arginine (Fig. 2).

Thermal Stability
gsNOS shows high thermal stability. Molar ellipticity measurements (⍜ 222 ) of NOS dimers indicate irreversible loss of secondary structure with increasing temperature (Fig. 3). We define the melting temperature as that at which half of the ellipticity (⍜ 222 ) is lost as the temperature is raised. Whereas bsNOS melts at 60°C, full-length gsNOS melts at 80°C. gsNOSϩ13, in which an N-terminal amino acid extension has been removed, melts at an intermediate temperature of 66°C (Fig. 3). In all cases, loss of secondary structure, as evidenced by CD, is irreversible.

Activity of gsNOS and bsNOS
Activity of gsNOS and bsNOS were compared at different temperatures by evaluating the rate of nitrite produced from substrate L-arginine in the presence of hydrogen peroxide ( Table 1). The amount of nitrite formed was quantitated by the Griess reaction (11) at 25°C after incubation of the L-arginine-saturated enzyme with peroxide at various temperatures and times. Product formation was linear with time. We followed the conversion of L-arginine rather than NOHA to NO x  because NOHA breakdown at elevated temperatures resulted in significant background product formation. At increased temperatures, gsNOS gives a lower rate of product formation than bsNOS. At 50°C, the bsNOS activity deviates greatly from that of gsNOS, which increases gradually with temperature. At temperatures above 50°C, bsNOS denatures and precipitates. gsNOS reactions with peroxide and NOHA instead of L-arginine give ϳ2 times more nitrite formation at room temperature.

Single Turnover Experiments
Substrate-free gsNOS-In the absence of any substrate, the spectrum recorded ϳ5 s after the introduction of oxygen showed a red-shifted Soret peak at 413 nm (Fig. 4), and the line shape was considerably different than substrate-free reduced gsNOS, which has a Soret peak at 411 nm and a broad shoulder from 440 to 490 nm (Fig. 4). Furthermore, the extinction coefficient for the Soret band of the intermediate is less than that of both the ferric-and ferrous-free enzyme. The 413 species was stable for ϳ1 min at 4°C and slowly decayed back to the ferric form (403 nm) with a change in line shape. Conversion to the ferric form of the enzyme was not complete even after 3 min at 4°C and required increasing the temperature of the cell to 25°C. Because the new 413 species results from mixing air-saturated buffer with the reduced protein, it likely represents the Fe(II)-O 2 complex; although given the limited time resolution, the initial spectrum observed on the addition of oxygen could include some contribution from free ferric enzyme. Nevertheless, combination of spectra from the substrate-free ferric and ferrous forms cannot explain the optical features of the intermediate observed in the presence of oxygen. When the same reaction was carried out in the presence of 40 M H 4 B, which is known to accelerate the decay of the ferrous-oxy species in mammalian and bacterial NOSs (21, 32), the conversion of the 413 species to the ferric form was complete in ϳ40 s at 4°C. With 100 M H 4 B, the intermediate decayed to the Fe(III) form within 5 s, and it could not be observed by these methods. In the presence of H 2 B, an oxidized form of H 4 B that binds NOS and is redox-  inactive, the reaction mixture behaved very similarly to the reaction in the absence of pterin. Complete conversion to ferric form was again possible only after heating the cell to 25°C. L-Arginine Bound gsNOS-The aforesaid reactions were repeated in the presence of 50 mM L-arginine. The ferric form of gsNOS with L-arginine bound has a Soret peak at 399 nm, whereas dithionite-reduced gsNOS had a Soret peak at 415 nm (Fig. 1). After oxygen injection, the first spectrum taken at t ϭ 5 s showed a Soret peak at 427 nm, which decayed completely to 399 nm in ϳ1 min (Fig. 5). Four isosbestic points in the spectra indicate a smooth transition from the 427-nm species to the ferric heme species. Rate constants derived from loss of the 427-nm peak or gain of the 399-nm peak were identical within error. This behavior resembles the ferrous-oxy complex of mammalian NOSs at Ϫ30°C (21,22,26) or at ambient temperatures using stop-flow techniques (20). A 428-nm band is also characteristic of a ferrous-heme-oxy complex in chloroperoxidase (55).
H 4 B accelerates the decay of the ferrous-oxy species and appearance of the ferric heme species (Fig. 6). In the presence of 15 M H 4 B, the decay rate of the 427-nm band and the appearance rate of the 399-nm band were both biphasic, each with a dominant phase that was much faster than that observed in the absence of H 4 B ( Table 2). Increasing the concentration of H 4 B to 40 M did not noticeably affect the relative proportions the slow and the fast phase. Furthermore, rate constants for the slow phases are not the same as those observed in the absence of pterin ( Table 2). Because of these considerations and a binding constant of H 4 B to B. subtilis NOS (100 nM) (12) that is much lower than the concentrations being used here, the slow phase in the presence of H 4 B is unlikely to derive from gsNOS free of H 4 B.
With H 4 B, the fast phases were more difficult to parameterize from the 399-nm peak appearance rate than from the 427-nm peak disappearance rate due to apparent smaller amplitudes ( Table 2). This may indicate that the kinetics of the optical changes at these two wavelengths are not the same when H 4 B is present. Such behavior could result from the presence of another intermediate between the ferrous-oxy and Fe(III) states of gsNOS with H 4 B. Better time resolution is necessary to address whether this is truly the case.
In contrast, 15 M H 2 B had a more minor, but opposite effect on the rates for Fe(III) formation or Fe(II)-O 2 complex decay ( Table 2). Rates derived from 427 and 399 nm were again equivalent and monophasic but slightly less than with free enzyme, suggesting a modest stabilization of the ferrous-oxy species by H 2 B.
N -Hydroxy-L-arginine Bound gsNOS-The same reactions were repeated under identical conditions with NOHA (1 mM) as the sub-   Table 3). This rate also matches the SVD-derived rate constant for conversion of the "mixed" intermediate to product Fe(III) heme (0.04 s Ϫ1 ). Simulation of the kinetics for the sequential conversion of heme-oxy to Fe(III)-NO to Fe(III) indicates that the heme-oxy and Fe(III)-NO will appear to decay with the same observed rate constant during the period of observation if the rate constant for the Fe(III)-NO decay is ϳ2 times that for the conversion of heme-oxy to Fe(III)-NO. Similar to the Fe(II)-O 2 complex, the Fe(III)-NO in gsNOS appears to be much more stable (5-50 times) than in other NOSs (Table 3).

Crystal Structure
At 3.2 Å of resolution (Table 4), the overall structure of gsNOS (Fig.  8A) appears strikingly similar to the structure of bsNOS (15) except for a few significant changes near the heme that result in a more compact active site. gsNOS maintains the overall fold of mammalian NOSoxy enzymes, with a conserved ␤-winged core surrounded by ␣-helices (57). As with other bacterial NOS proteins, the N-terminal hook, the pterin binding segment, and the zinc-binding site are absent. The dimer interface in gsNOS is identical to the bsNOS dimer interface, with the exception of a change from Ala-334 to Thr-343 (gsNOS) in the helical lariat region (15). The side chain of Thr-343 hydrogen bonds with the peptide carbonyl of a conserved Ser-340 (gsNOS) of the same monomer but does not alter the backbone conformation. A long loop in bsNOS consisting of 10 residues joining two ␤-strands (Pro-110 to Val-119) is replaced in gsNOS by a shorter loop consisting of only six residues (Arg-124 to Val-129). The overall structure of gsNOS appears some-

TABLE 2 Kinetic parameters for single turnover experiments
Rates of decay of the heme-oxygen complex (as monitored at 427 nm) and the appearance of ferric heme enzyme (399 nm) in the presence of 2.5 mM L-arginine compared to the rates of other NOSs for which similar rates are known. Reduced pterin, H 4 B (15 M) accelerates both the decay of the heme-oxygen complex and the ferric heme appearance. Each kinetic parameter represents an average from at least three separate experiments. The relative amplitude of each phase ( f ) is given in parentheses for bi-exponential kinetics. deiNOS, D. radiodurans NOS. what more compact than the structure of bsNOS (Fig. 8B). The lower part of the dimer contracts, and the upper part slightly expands. The dimer interface, however, overlaps nearly exactly. Perhaps the most significant difference in the structure of gsNOS is in the active site (Fig. 9). A Lys-356 (bsNOS) to Arg-365 (gsNOS) substitution results in a hydrogen bonding interaction between this residue and Asp-225 in gsNOS (Asp-216 in bsNOS) that alters the orientation of the Asp carboxylate relative to bsNOS. Reorientation of Asp-225 displaces Ser-224 and pushes Ile-223 about ϳ0.6 Å closer toward the heme iron in gsNOS (Fig. 9). Although the limited resolution of the gsNOS structure (3.2 Å) prevents a precise evaluation of the Ile-223 side-chain position, difference Fourier electron density maps clearly show that Ile-223 and its ␤-strand reside closer to the heme than in bsNOS. Surprisingly, there is no apparent electron density for the N-terminal extension that confers enhanced thermal stability to gsNOS. SDS-PAGE analysis of crystals showed that the N terminus was not proteolytically removed from the crystallized protein.

DISCUSSION
We have cloned and characterized a NOS-like protein from G. stearothermophilus that is very similar to all other well characterized NOSs in terms of its sequence (Supplemental Fig. 1), structure, substrate specificity, activities, and spectral properties. We demonstrate that gsNOS produces nitrite from L-arginine and NOHA in the peroxide shunt reaction and an optical signal characteristic of bona fide NO in single turnover experiments when only NOHA and H 4 B are present. The structure of bsNOS reveals the absence of the zinc site and N-terminal hook of mammalian NOSoxy proteins, and thus, this protein further demonstrates that these regions are not required for catalysis provided the dimer remains intact (11,12,58). gsNOS has a small N-terminal extension (ϳ10 residues) compared with NOSs from other bacteria like D. radiodurans, B. subtilis, and S. aureus. For all thermophiles for which genome sequences are available, only one other, G. kaustophilus, also has a NOS gene. G. kaustophilus NOS shares 94% sequence identity with gsNOS but has a much longer N-terminal extension (ϳ50 residues). G. kaustophilus has genes coding for enzymes that synthesize H 4 B, and the G. stearothermophilus genome (currently incomplete) contains homologs with ϳ90% sequence identity.
Unlike other NOSs, gsNOS has high thermal stability, with a melting temperature of 80°C, 20°C higher than the melting temperature of bsNOS. Interestingly, removal of 13 residues that extend the gsNOS protein beyond bsNOS reduces the melting temperature to only 6°C above that of bsNOS. Surprisingly, these residues are completely disordered in the crystal structure. The gsNOS rate of catalysis is much slower compared with other bacterial NOSs and mammalian NOSoxy domains at ambient temperatures, consistent with the enzyme operating above 60°C in vivo. This sluggish reactivity has allowed us to carry out single turnover experiments that reveal a slow rate of conversion to products and uncharacteristically stable heme-oxy intermediates.
Oxygen complexes of reduced NOSs, cytochrome P450s, and chloroperoxidases are well characterized (1, 20 -23, 36, 39, 41, 59 -64). In    (1, 11, 12, 20 -22, 26, 41, 59) (Table  5). These spectral differences have been correlated with changes in hydrogen bonding to the proximal heme thiolate, which blue-shifts the Soret peak, and hydrogen bonding to the proximal oxy-species, which red-shifts the Soret peak (46,70,71). The studies of eNOS, mentioned above, correlate well with this general trend. In nNOS, substrate L-arginine or NOHA binds closely to the site of O 2 heme coordination and participates in an H 2 O guanidinium hydrogen-bonding network with the coordinated O 2 (35,57,58,72). Thus, interactions of the substrate are expected to increase hydrogen bonding to the coordinated oxygen and red-shift the Soret. In accord with increased hydrogen bonding, the Soret spectra of the heme-oxy complex in eNOS red shifts from 420 to 432 nm in the presence of L-arginine and 428 nm in the presence of NOHA. In nNOS, the heme-oxy complex has a Soret peak at 427 nm in the presence of L-arginine (20,73), a 416-nm peak in the absence of L-arginine (21), and a 419-nm peak in the presence of inhibitor N Gmethyl L-arginine, which cannot form a hydrogen bond with hemebound oxygen (Table 5). Similar trends are seen in gsNOS at 4°C, except that the substrate-free form (413 nm) and the substrate-bound form (427 nm) have slightly blue-shifted Soret bands relative to those of eNOS and similar to those of nNOS. In nNOS, the 430-nm heme-oxy species has been definitively assigned as a ferric-superoxy species by resonance Raman spectroscopy (1, 74). Thus, increased hydrogen bond- FIGURE 9. Comparison of the active sites of bsNOS (orange) and gsNOS (yellow). In bsNOS, Lys-356 does not interact with Asp-216. A Lys to Arg substitution in gsNOS allows Arg-365 to hydrogen bond with Asp-225 (3.2 Å), altering its side chain position. This change in structure appears to be correlated with movement of Ser-224 that in turn pushes Ile-223 into the active site, reducing the distance between the ␦-carbon of Ile-223 and the heme iron atom from 6.7 Å (bsNOS) to 6.1 Å (gsNOS).  (20 -22,26). Anaerobic stopped-flow methods will be better able to define the rate constants and progression of these intermediates, but these data do provide strong evidence for formation of an Fe(III)-NO after reaction of heme-bound oxygen with NOHA in the presence of H 4 B. Because no Fe(III)-NO complex forms with H 2 B, these data also corroborate H 4 B acting as an electron donor in the second step of the NOS mechanism (18,25,75). Rapid-freeze EPR experiments on the mouse iNOS oxygenase domain show that a tetrahydrobiopterin radical forms and then becomes reduced during NOHA oxidation, resulting in Fe(III)-NO formation (25). In contrast to the mammalian NOSs and other bacterial NOSs, where the ferric-superoxy appears on millisecond timescales at 10°C or at extremely low temperatures (Ϫ30°C) (20 -22, 26), the rate of decay of the ferric-superoxy species in gsNOS in ϳ10 -100 times slower. Similarly, the Fe(III)-NO species in gsNOS in 5 times more stable than in bsNOS and 50 times more stable than in iNOS ( Table 3).
The crystal structure of gsNOS provides some insight into why the heme-ligand complexes survive 1-2 orders of magnitude longer than counterparts do in other bacterial NOSs and mammalian NOSoxy ( Table 2). Given the different behaviors of the mesostable bsNOS and the thermostable gsNOS, there are very few structural differences between the two enzymes. The stability of the heme-oxy complex in gsNOS could derive from a number of factors that include a decreased heme redox potential (76), increased hydrogen bonding to the heme ligand, or affects on steric and dynamic properties of the heme pocket. Changes to heme redox potential or heme-ligand hydrogen bonding are unlikely because there are no differences in charged residues or hydrogen bond donors in the heme pockets of gsNOS compared with the bsNOS. However, there is a conformational difference in gsNOS at a key residue (Ile-233) known to generally affect NOS heme-ligand dissociation rates.
A conserved valine residue in mammalian NOS that resides above the heme pocket switches to a conserved isoleucine in bacterial NOSs (Supplemental Fig. 1). This Ile has been shown to reduce the release rates of heme-bound NO (38). A comparison of the respective crystal structures shows that the same Ile-223 (present in both bsNOS and gsNOS) is ϳ0.6 Å closer to the heme in gsNOS than in bsNOS. This appears to be linked to the substitution of Arg-365 (gsNOS) for Lys-356 (bsNOS) (Fig. 9). A more constrained heme pocket in gsNOS may disfavor ligand dissociation. Consistent with proximity of the residue at position 223 to the heme iron decreasing ligand dissociation rates, the Fe(III)-NO species that forms after reaction with NOHA resides on the heme even longer in gsNOS than in bsNOS (Table 3). Sequence alignment with all other known bacterial NOSs and mammalian NOSoxy domains (data not shown) shows that only gsNOS and the other thermophilic G. kaustophilus NOS have arginine at this key position.
Recent structures of cytochromes P-450 bound to oxygen reveal that the water molecule structure in the active center changes when oxygen ligates the heme (61, 63, 64). Similarly, water molecules not present in the ligand-free structures of gsNOS may affect the stability of the hemeoxy species in bacterial NOS. Even small motions in Ile-233 could in turn influence hydrogen bonding patterns between solvent and heme ligand and thereby contribute to differences in ferrous oxy stability.
There is no obvious single contributing factor to the overall thermal stability of gsNOS, although the structure does appear more compact than that of bsNOS (Fig. 8B). Surprisingly, removal of N-terminal extension (ϳ13 residues) decreases the T m almost to that of bsNOS; yet, this region is completely disordered in the crystal structure. We have previously observed that the stability and solubility of bacterial NOSs depend highly on the N terminus (15). Perhaps removal of the extension is sufficiently destabilizing to overcome small structural features that collectively generate increased stability in gsNOS.

CONCLUSIONS
We have characterized a thermostable nitric-oxide synthase with a slower reaction profile. As a result, the heme-bound oxygen intermediates can be observed at ambient temperatures on the timescale of seconds. Dramatic shifts in the heme-oxy spectra correlate with substrate hydrogen bonding to heme-coordinated O 2 . This interaction likely converts a heme-oxy complex from an electronic state best described as ferrous-oxy (413 nm) to one better represented as ferric-superoxy (427 nm). Structural comparisons suggest that subtle rearrangements in the distal heme pocket contribute to increased stability of heme-oxy complexes in gsNOS. The slow catalytic profile of the enzyme may be an advantage for identifying the nature of the reactive heme-oxygen intermediates directly involved in the NO synthase mechanism.