Mechanism and regulation of ferrous heme-nitric oxide (NO) oxidation in NO synthases

Nitric oxide (NO) synthases (NOSs) catalyze the formation of NO from l-arginine. We have shown previously that the NOS enzyme catalytic cycle involves a large number of reactions but can be characterized by a global model with three main rate-limiting steps. These are the rate of heme reduction by the flavin domain (kr), of dissociation of NO from the ferric heme-NO complex (kd), and of oxidation of the ferrous heme-NO complex (kox). The reaction of oxygen with the ferrous heme-NO species is part of a futile cycle that does not directly contribute to NO synthesis but allows a population of inactive enzyme molecules to return to the catalytic cycle, and thus, enables a steady-state NO synthesis rate. Previously, we have reported that this reaction does involve the reaction of oxygen with the NO-bound ferrous heme complex, but the mechanistic details of the reaction, that could proceed via either an inner-sphere or an outer-sphere mechanism, remained unclear. Here, we present additional experiments with neuronal NOS (nNOS) and inducible NOS (iNOS) variants (nNOS W409F and iNOS K82A and V346I) and computational methods to study how changes in heme access and electronics affect the reaction. Our results support an inner-sphere mechanism and indicate that the particular heme-thiolate environment of the NOS enzymes can stabilize an N-bound FeIII-N(O)OO− intermediate species and thereby catalyze this reaction, which otherwise is not observed or favorable in proteins like globins that contain a histidine-coordinated heme.

Nitric oxide (NO) synthases (NOSs) catalyze the formation of NO from L-arginine. We have shown previously that the NOS enzyme catalytic cycle involves a large number of reactions but can be characterized by a global model with three main ratelimiting steps. These are the rate of heme reduction by the flavin domain (k r ), of dissociation of NO from the ferric heme-NO complex (k d ), and of oxidation of the ferrous heme-NO complex (k ox ). The reaction of oxygen with the ferrous heme-NO species is part of a futile cycle that does not directly contribute to NO synthesis but allows a population of inactive enzyme molecules to return to the catalytic cycle, and thus, enables a steady-state NO synthesis rate. Previously, we have reported that this reaction does involve the reaction of oxygen with the NO-bound ferrous heme complex, but the mechanistic details of the reaction, that could proceed via either an inner-sphere or an outersphere mechanism, remained unclear. Here, we present additional experiments with neuronal NOS (nNOS) and inducible NOS (iNOS) variants (nNOS W409F and iNOS K82A and V346I) and computational methods to study how changes in heme access and electronics affect the reaction. Our results support an inner-sphere mechanism and indicate that the particular heme-thiolate environment of the NOS enzymes can stabilize an N-bound Fe III -N(O)OO ؊ intermediate species and thereby catalyze this reaction, which otherwise is not observed or favorable in proteins like globins that contain a histidinecoordinated heme.
NO synthases (NOSs; 4 EC 1.14.13.39) are homodimeric enzymes that catalyze formation of the biological messenger molecule nitric oxide (NO) from L-arginine (L-Arg) (1)(2)(3). Three main isoforms of NOS have been described in mammals: neuronal (nNOS), endothelial (eNOS), and inducible (iNOS). All isoforms share significant sequence homology and consist of two main modules: an N-terminal oxygenase domain (NOSoxy) that binds heme, tetrahydrobiopterin (H 4 B), and the substrate L-Arg and a C-terminal reductase domain that binds FMN, FAD, and NADPH. A calmodulin-binding region is located between NOSoxy and the reductase domain. NOS reductase domains are closely related to dual-flavin NADPH oxidoreductase enzymes like cytochrome 450 reductase, whereas proteins similar to the NOSoxy have been found in bacteria, plants, fungi, and basal metazoans (4,5).
NO synthesis is a complex multistep process (6). The kinetics of NO synthesis are complicated by the fact that not all of the NO synthesized by NOS is effectively released out of the enzyme, but instead can remain bound to the heme and then be oxidized to nitrate. To describe this process, we have proposed a global mechanism that includes the NO biosynthetic reactions, the NO release step, and a futile cycling of NOS through its ferrous heme-NO complex without net NO release (Fig. 1A) (6 -8). It is important to note that the futile cycle becomes an inherent property of the NOS catalytic cycle; as the electron transfer from the reductase domain to the heme is involved in the initial cycle step (k r ) but also in the inactivation of the enzyme via the reduction of the ferric-NO complex (k r ٞ), it follows that an optimal value for k r that maximizes NO output exists for each NOS isoform (6). The model can explain a number of paradoxical catalytic behaviors observed in different NOS isoforms and mutants (6,7,9).
The oxidation of the ferrous heme-NO complex of NOS (designated by k ox in Fig. 1A) is actually an NO dioxygenation reaction that destroys the NO made by the NOS enzyme. It has received somewhat less attention than the other kinetic parameters of NOS catalysis, particularly the rate of heme reduction by the reductase domain (k r in Fig. 1A), or the rate of NO dis-sociation from the ferric heme-NO product complex (k d in Fig.  1A) (6 -8). Nevertheless, the k ox rate setting is important for NOS enzymes, because it helps to determine how much NOS enzyme accumulates as the ferrous heme-NO complex during steady state NO synthesis (Fig. 1, B and C), which in turn markedly impacts the apparent K m O 2 and magnitude of the steadystate NO synthesis activity of the given NOS enzyme (Fig. 1, B and C) (6 -8). Indeed, computer simulations of the global model in Fig. 1A have indicated how different k ox settings can impact the steady-state rates of NO synthesis and apparent K m O 2 of a given NOS. Such simulations have been helpful for explaining the catalytic phenotypes of various NOS mutants with altered k ox settings (10), as well as understanding the three WT NOS isozymes, which possess different k ox settings (6,7). For example, the relatively fast oxidation rate of the ferrous heme-NO complex of iNOS (i.e. it has the highest k ox setting) helps to explain why it displays an NO synthesis activity that is equivalent to or exceeding that of nNOS, even though the heme reduction rate (k r ) in iNOS, which determines its inherent rate of NO biosynthesis, is 4 -6 times slower than in nNOS (7).
Previous measures of k ox were typically made at single O 2 concentrations (i.e. using air-or O 2 -saturated buffers) (7,8). We investigated the NO dioxygenase (NOD) reaction of nNOS and iNOS as a function of O 2 concentration, using the bacterially expressed oxygenase domains (nNOSoxy and iNOSoxy) (11), which are useful surrogates for the full-length NOS enzymes in such studies (7,8). We restricted our analysis to the catalytically relevant H 4 B-plus L-Arg-bound NOSoxy forms and found the following. (i) The nNOSoxy and iNOSoxy have a 100-fold difference in their respective k ox settings (k 1 ϭ 230 M Ϫ1 s Ϫ1 for nNOSoxy and k 1 ϭ 26,500 M Ϫ1 s Ϫ1 for iNOSoxy, where k obs ϭ k 1 [O 2 ] ϩ k Ϫ1 ). (ii) Their kinetic behaviors suggest an NOD reaction mechanism whereby the ferrous heme-NO complex of both proteins reacts directly with O 2 to yield the ferric enzyme plus nitrate. This is novel and contrasts with the NOD reactions of globin ferrous heme-NO complexes (i.e. hemoglobin, myoglobin, bacterial NO dioxygenases), which are not directly reactive toward oxygen and instead have their NOD reactions rate-limited by an initial NO dissociation from the ferrous heme, which allows subsequent O 2 binding to form a ferrous heme-O 2 complex that reacts with NO in the distal pocket (12,13). (iii) Because we observed no buildup of reaction intermediates in the NOSoxy reactions even at the highest possible O 2 concentration, it appears that O 2 access and/or its reactivity with the ferrous heme-NO complex is rate-limiting for the NOD reactions of the two H 4 B-and L-Arg-bound NOSoxy proteins.
Here, we expanded our study by investigating how substrate and cofactor occupancy, protein structural changes, and heme electronic features impact k ox in iNOSoxy and nNOSoxy. In particular, we focus on factors that are known to modify the stability of the Fe II -O 2 and/or Fe II -NO complexes of NOS, such The rates of heme reduction (k r , k r Ј, k r Љ, and k r ٞ), Fe III NO dissociation (k d ), and Fe II NO oxidation (k ox ) are the key rate-limiting steps, as oxygen binding and the catalytic steps (k cat1 and k cat2 ) are comparatively faster. B, effect of k ox rates on instantaneous NO synthesis rates. The rate of NO synthesis versus time was simulated according to the global model using rates for nNOS at 25°C (63) and varying k ox values (0.01-1.0 s Ϫ1 ; values for each trace are indicated). The k ox rate does not change the initial rates of NO synthesis but becomes more important in setting steady-state NO synthesis rates. C, effect of k ox rates on steady-state NO synthesis rates (steady NO synthesis rate calculated at 10 s from simulations conducted as those in B).

Regulation of Fe II -NO oxidation in NOS
as heme thiolate environment (10,14,15), NOS dimerization (16,17), high-spin/low-spin heme transitions (18,19), and H 4 B cofactor binding (20 -22). Additionally, we examine the k ox reaction using computational methods to predict the thermodynamics and the feasibility of the direct reaction of the ferrous NO adduct of NOS with O 2 , using a NOS active site model. These calculations were then repeated using a globin active site model to elucidate the observed differences in NOD reactivity between NOS and globins. The results increase our understanding of how k ox is controlled in NOS enzymes and provide further insight into their NOD reaction mechanism.

Effects of bound H 4 B and L-Arg
Reactions were run at 10°C under different conditions of substrate or cofactor occupancy and across a range of O 2 concentrations, using iNOSoxy and nNOSoxy proteins that we purified in the absence of L-Arg or H 4 B. Adding L-Arg or H 4 B alone or together to iNOSoxy stabilized its ferric high-spin form (23) and enabled the generation of a stable six-coordinate Fe II -NO complex. In contrast, the L-Arg-and H 4 B-free iNOSoxy and nNOSoxy proteins did not form stable six-coordinate Fe II -NO complexes and converted rapidly to five-coordinate Fe II -NO complexes that reacted very slowly with O 2 , as reported previously (11). These were not studied further.
The oxidation reactions of the L-Arg-and H 4 B-bound iNOSoxy Fe II -NO complexes were monophasic, with no buildup of detectable spectral intermediates, and formed ferric high-spin iNOSoxy as the product ( Fig. 2A), identical to what we observed previously (11). Reaction of the iNOSoxy Fe II -NO complex that only contained H 4 B proceeded at similar rates across the range of O 2 concentrations (Fig. 3A) and gave essentially an identical k ox rate constant as the H 4 B-and L-Argbound protein ( Table 1). The reactions of the iNOSoxy Fe II -NO complex that only contained L-Arg were again very similar regarding the spectral evolution (data not shown) but in this circumstance occurred at substantially increased rates relative to the two other conditions (Fig. 3A), giving a 2.7 times higher k ox rate constant ( Table 1). The rate versus oxygen concentration relationship was linear in all three conditions of substrate and cofactor occupancy (Fig. 3A). Our results suggest that bound H 4 B retards k ox in iNOSoxy, whereas bound L-Arg does not, and that O 2 access or reactivity remains rate-limiting under all three conditions. We performed additional experiments with an iNOSoxy containing dihydrobiopterin (H 2 B) instead of the native cofactor (H 4 B). H 2 B can occupy the same binding site in NOS enzymes, but it is redox-inactive during the NO synthesis process (22,24). The oxidation reaction of the iNOSoxy Fe II -NO complex in the presence of H 2 B and L-Arg yielded similar kinetic results to those obtained for H 4 B (Table 1). Thus, the biopterin reduction state does not influence its capacity to diminish the k ox setting in iNOSoxy.
We similarly studied the k ox reactions of the nNOSoxy Fe II -NO complex under the three conditions as noted above. Overall, the same pattern was observed (Figs. 2B and 3B), although the magnitudes of k ox obtained across the range of O 2 concentrations were much less than for iNOSoxy reactions, as we previously observed (11). The nNOSoxy Fe II -NO complex containing H 4 B alone, or along with L-Arg, had similar k ox rates, and in the absence of H 4 B, the L-Arg-bound complex again showed increased k ox rates (Fig. 3B). However, the increase seen for the H 4 B-free nNOSoxy was very modest compared with that seen for iNOSoxy and resulted in only a 30% increase in the k ox rate constant (Table 1). Nevertheless, the data suggest that the k ox reactions of nNOSoxy and iNOSoxy are similarly influenced by L-Arg-and/or H 4 B-binding site occupancy.

Effect of restricting access to the distal heme pocket
In iNOS, the magnitude of several kinetic parameters is sensitive to relatively small changes in the protein structure near the distal heme pocket (25). Replacing a conserved Val at the entrance to the distal heme pocket with the bulkier Ile altered the observed rates of O 2 and NO binding to or dissociation from the iNOS heme, which could be explained by changes in ligand access to or egress from the distal heme pocket (25). These effects on heme-NO binding kinetics were also observed for a bacterial NOSoxy protein when similar substitutions were made (26). We therefore used the V346I mutation to study the

Regulation of Fe II -NO oxidation in NOS
effect of distal heme pocket changes on k ox rates in iNOSoxy. The protein formed a stable six-coordinate Fe II -NO complex in the presence of either H 4 B or L-Arg, and as we observed for the WT iNOSoxy, a five-coordinate Fe II -NO complex appeared in the absence of H 4 B and L-Arg. The spectral evolution observed during reaction of a V346I iNOSoxy Fe II -NO complex in the presence of L-Arg and H 4 B was very similar to that seen for WT iNOSoxy (Fig. 2C). Nevertheless, the k ox reaction rates across the range of O 2 concentrations were about 40% that of WT iNOS (Fig. 3C). The observed rates remained linearly dependent on O 2 concentration, and the k off term was too small to be accurately determined. The calculated value for the k on rate constant was ϳ10,200 M Ϫ1 s Ϫ1 , which is 38% of the magnitude observed for WT under identical conditions ( Table 1). The effects of individual H 4 B or L-Arg site occupancy were remarkably similar in the mutant compared with WT iNOSoxy (Fig. 3, A and C), with little difference in the k ox rates observed when H 4 B was bound with or without L-Arg and a significant (2.5fold) rate increase when only L-Arg was bound (Table 1). Our finding that the k ox reaction is sensitive to the protein structure determining the opening to the distal heme pocket is consistent with O 2 access being a rate-limiting feature of the k ox reaction.

Effect of NOSoxy dimeric structure
NOSoxy subunit dimerization is required in the enzyme's post-translational assembly and is associated with specific protein structural changes that decrease the solvent exposure of the bound heme and help to create the L-Arg-and H 4 Bbinding sites (20,21). We therefore studied the k ox reaction of the K82A iNOSoxy mutant, which is monomeric as purified but dimerizes in the presence of H 4 B (17). After incubating K82A iNOSoxy at high H 4 B and L-Arg concentrations (0.5 mM H 4 B, 20 mM L-Arg) to fully induce dimerization, the enzyme formed a stable six-coordinate Fe II -NO complex, and the kinetics of Fe II -NO oxidation reaction were comparable with WT iNOS, giving a linear dependence of k ox versus O 2 concentration (Fig.  3D) and a k on of ϳ19,000 M Ϫ1 s Ϫ1 (Table 1). To study the reaction of the monomeric enzyme, we incubated K82A iNOSoxy with L-Arg alone, which is unable to promote dimerization in the mutant (17). In this case, the enzyme formed a Fe II -NO complex that was a blend of five-and six-coordinate, with Soret bands at 393 nm and at 438 nm for the five-and six-coordinate Fe II -NO complex, respectively (Fig. 2D). This mix of complexes was found to be stable over the experimental time frame. The reaction of the monomeric form of the K82A iNOSoxy Fe II -NO complex with O 2 proceeded with two observable phases (Fig.  2D): a fast phase associated with the loss of 438-nm absorbance and a concomitant increase at 398 nm and a second, much slower phase in which the Soret band shifted toward higher wavelengths. We assigned these changes to two reactions. The fast oxidation is that of the six-coordinate Fe II -NO complex, with an absorbance decrease near 438 nm and buildup of high spin Fe III (396 nm), and a subsequent slow oxidation represents

Regulation of Fe II -NO oxidation in NOS
the conversion of the five-coordinate Fe II -NO complex (386 nm) to low-spin Fe III (419 nm). Absorbance changes at 440 and 398 nm (Fig. 2D, inset) were fitted to a bi-exponential equation, and the fast rates were plotted versus O 2 concentration (Fig.  3D). The observed rates for the fast phase (representing oxidation of the six-coordinate complex) were similar at either wavelength, giving k on of ϳ73,300 M Ϫ1 s Ϫ1 . This was significantly faster than for the H 4 B-bound dimeric form of K82A iNOSoxy and represented a 4-fold increase (Table 1). This k ox rate constant was similar to what we observed for the WT iNOSoxy enzyme without H 4 B (see Table 1). Thus, the Fe II -NO complexes in the iNOSoxy monomer and in the H 4 B-free iNOSoxy dimer were kinetically similar regarding their reactivity toward O 2 , despite any possible differences between their heme environments or ligand access. These results suggest a more open heme pocket in the absence of H 4 B.

Effect of modifying the heme midpoint reduction potential
NOS isoforms contain a conserved tryptophan (Trp-409 in nNOS) that is located on the proximal side of the heme that engages in a -stacking interaction with the heme porphyrin and forms a hydrogen bond with the axial cysteine thiolate heme ligand (15). Studies using Trp-409 mutants suggested that the loss of the Trp-Cys hydrogen bond should increase the negative charge density on the cysteine thiolate and thus lower the heme midpoint reduction potential (10,15). In fact, a number of studies on the homologous Trp residue of iNOS (27), eNOS (28), and bacterial NOS proteins (28 -31) indicate a correlation between the strength of the hydrogen bonding to the thiolate, the heme midpoint potential, and the stability/reactivity of the Fe II -NO/Fe II -O 2 complexes. To test this hypothesis, we utilized a spectroelectrochemical method (27) to measure and compare the heme midpoint potentials of WT and W409F nNOSoxy in the presence of both L-Arg and H 4 B. The fittings of the experimental values to the Nernst equation are shown in Fig. 4. The calculated E m values were Ϫ275 Ϯ 4 mV for nNOSoxy WT, and Ϫ335 Ϯ 4 mV for W409F nNOSoxy. The midpoint value we obtained for WT is close to previously reported values (23,32,33). Thus, the W409F substitution decreased the ferric/ferrous heme midpoint potential by Ϫ60 mV. This confirmed the predicted effect of the Trp-Cys hydrogen bond and made the W409F nNOSoxy mutant useful for our investigation.
The W409F nNOSoxy mutant is known to form a six-coordinate Fe II -NO complex that shifts rapidly to a mixture of a weaker six-coordinate complex and some five-coordinate complex (10,18), similar to the behavior of K82A iNOSoxy described above. We measured k ox rates as a function of O 2 concentration for the Fe II -NO complex of the W409F nNOSoxy mutant in the presence of L-Arg and H 4 B (Fig. 2E), focusing on the reciprocal absorbance changes associated with oxidation of the six-coordinate complex at 438 and 396 nm. The reaction was monophasic and proceeded with the typical spectroscopic transitions, but the observed rates were much higher than for the WT nNOSoxy enzyme (Fig. 3E). The k ox rates increased linearly with oxygen concentration across the entire range and gave a value for k on of ϳ16,400 M Ϫ1 s Ϫ1 (Table  1). This represents a 70-fold increase in k ox value relative to WT (Table 1) and thus demonstrates that decreasing the ferric/ferrous heme midpoint potential in nNOSoxy has a large impact on the k ox .

Effect of pH
Increasing the pH can impact two aspects of the k ox reaction. First, it can strengthen the heme thiolate bond in W409F nNOSoxy and so stabilize the six-coordinate Fe II -NO complex (18). Second, it may conceivably increase the stability of a putative heme-bound peroxynitrite intermediate that could form as an immediate product of the k ox reaction (11). We therefore studied the k ox reaction at pH 9.5 for several proteins: iNOSoxy, V346I iNOSoxy, nNOSoxy, and W409F nNOSoxy.
Increased pH did not alter the spectroscopic properties of the protein Fe II -NO complexes except in W409F nNOSoxy, where it completely stabilized the six-coordinate complex having a Soret absorbance at 438 nm (Fig. 2F), as previously reported for the nNOSoxy W409Y mutant (18). Running the reactions at pH 9.5 did not alter the O 2 dependence of the k ox reaction for W409F nNOSoxy or for any of the other proteins, as their k ox reactions remained monophasic with no detectable buildup of a reaction intermediate, and their observed k ox rates still increased linearly with oxygen concentration across the entire range (data not shown). We also observed only small changes (within 20%) in the k ox rate constants for the reactions run at pH 9.5 versus 7.4 (Table 1). Thus, increasing the pH had minimal effects on the O 2 reactivity of the Fe II -NO complexes and did not alter the reaction trajectory regarding stabilization of possible heme-bound intermediates.

Computational investigation of the k ox reaction
To further understand the reaction of NOS Fe II -NO complexes with O 2 , we have performed density functional theory (DFT) calculations on a computational model of the NOS active site. The model used for these calculations was constructed by taking the coordinates from the X-ray crystal structure reported of the Fe II -NO complex of nNOS (34) and keeping

Regulation of Fe II -NO oxidation in NOS
only the Fe II -NO protoporphyrin IX complex, the axial bound cysteinate (thiolate) ligand (Cys-415), the conserved axial tryptophan (Trp-409), and a glycine residue (Gly-417), which also forms a hydrogen bond with the axial thiolate ligand. Further details of the NOS active site model used for all calculations are provided under "Experimental procedures." The geometry of this NOS Fe II -NO truncated active site model was fully optimized using BP86/TZVP with water as the polarized solvent continuum (via a self-consistent reaction field approach). Fig.  S1 shows the structure obtained from these calculations. A triplet O 2 ( 3 O 2 ) molecule was then introduced into the system, with the unpaired elections of the 3 O 2 antiparallel to the unpaired electron of the S ϭ 1 ⁄ 2 Fe II -NO unit, to allow for potential bond formation between these two radical species. Initially, the O 2 molecule was constrained at a distance of 3.26 Å from the nitrogen atom of the Fe II -NO unit (see Fig. 5 for the optimized structure). Geometric parameters of this optimized structure are compared with the nNOS Fe II -NO crystal structure in Table S1. This comparison shows that our simplified active site model replicates the bond distances and angles observed in the NOS crystal structure reasonably well, without including any distal secondary sphere amino acids. Using this optimized structure as the starting point (ON-OO distance ϭ 3.26 Å), a potential energy surface (PES) scan was performed where the 3 O 2 was moved closer to the nitrogen atom of the Fe II -NO unit until an ON-OO distance of 1.42 Å was reached (Fig. S2). From these calculations, initial movement of the 3 O 2 toward the Fe II -NO unit is predicted to result in formation of a favorable van der Waals complex (ON-OO distance ϭ 2.11 Å), which is Ϫ4. For the NOS k ox reaction to proceed through a peroxynitrite intermediate, further stabilization of this species is needed (to prevent the back reaction to O 2 and the Fe II -NO complex). One possibility in this regard would be protonation of the newly formed peroxynitrite ligand. Free peroxynitrite has a pK a of 6.8 (35), meaning it could be converted to peroxynitrous acid under biological conditions. Protonation of our Fe III -N(O)OO Ϫ complex in the NOS model in silico and optimization of the structure resulted in a stable ferric-peroxynitrous acid complex (Fe III -N(O)OOH) with a ON-OO bond length of 1.483 Å, without the need to constrain the ON-OO bond for the optimization. However, the fact that experimentally, only a small pH dependence of the NOD reaction in NOS is observed (Table 1) argues against this possibility.
Alternatively, we hypothesized that instead of protonation of the peroxynitrite ligand, a simple hydrogen bond to this species could be sufficient for stabilization, allowing the NOD reaction to proceed. To test this possibility, a water molecule was placed in close proximity to the peroxynitrite complex, and the structure was then re-optimized without constraining the peroxynitrite ON-OO bond. This resulted in a stable structure with only a slightly elongated ON-OO bond of 1.508 Å (see Fig. 6, inset). In this structure, the H 2 O molecule is predicted to make a reasonably strong hydrogen bond to the anionic ONOO Ϫ ligand with an OH-OONO distance of 1.69 Å and O-H-O(ONO) angle of 176.7°. This structure is calculated to be 8.1 kcal/mol lower in energy relative to the complex without the H 2 O hydrogen bond (ON-OO ϭ 1.50 Å). A PES scan was then performed with the inclusion of the H 2 O molecule as shown in Fig. 6 (black points). Results from this PES scan predict that the ferric peroxynitrite intermediate is now only 1.3 kcal/mol higher in energy relative to the Fe II -NO/O 2 van der Waals complex (observed here at ON-OO ϭ 2.108 Å), and that the activation energy barrier for formation of the peroxynitirite intermediate is reduced to 2.4 kcal/mol relative to the van der Waals complex. Together, these results indicate that formation and stabilization of a heme-thiolate Fe III -ONOO Ϫ intermediate would be quite feasible in NOS when assisted by a hydrogen bond from a water molecule. This is certainly a reasonable possibility, considering that water molecules have been identified in the NOS active site by X-ray crystallography, as shown in Fig. 5B (34). To compare the NOS model to globins, we built a model including the distal histidine conserved in globins (see details in the supporting information and Fig. S3). A PES scan of the globin model (Fig. 6, red points)

Regulation of Fe II -NO oxidation in NOS
indicates that the distal imidazole does not stabilize the N-bound peroxynitrite complex, in agreement with the experimental observation that reaction of 3 O 2 with the NO adduct of a ferrous globin requires initial dissociation of NO (and binding of O 2 to the heme) for the reaction to proceed.

Nature of the peroxynitrite intermediate
In the NOD reaction of globin proteins, where NO reacts with oxyglobin to generate nitrate, the reaction likely proceeds through formation of a terminal O-bound peroxynitrite intermediate (Fe III -OONO Ϫ ). Other computational studies have investigated the formation and conversion of this intermediate to nitrate and have concluded that a sequential isomerization mechanism is the most thermodynamically feasible pathway (over a concerted mechanism) (36,37). In this mechanism, the ferric globin peroxynitrite complex (Fe III -OONO Ϫ ) undergoes O-O(NO) bond homolysis, resulting in formation of an Fe IV ϭO complex (Compound II in the deprotonated state) and the nitrogen dioxide radical ( ⅐ NO 2 ), which then quickly recombine to generate a ferric-nitrate complex (Fe III -NO 3 Ϫ ). For a similar sequential isomerization mechanism to occur for the proposed NOS N-bound peroxynitrite intermediate, a switch in the peroxynitrite coordination mode would first need to occur. To help determine the feasibility of this possibility, the geometries of the three other binding modes of peroxynitrite to our heme-thiolate NOS model were optimized to allow for a comparison of their relative energies (see Table 2). Full optimization of the three additional peroxynitrite coordination modes was possible without constraining the ON-OO bond. Geometry optimization of the structures in the presence of the H 2 O hydrogen bond donor shows that the terminal O-bound complex (Fe III -OONO Ϫ ) is the most favorable binding mode, which is 5.8 kcal/mol lower in energy than the stable N-bound structure. The (ON)O(O)-bound peroxynitrite complex is 4.8 kcal/mol higher in energy than the N-bound species (and 10.6 kcal/mol higher in energy than the terminal OONObound complex), and the O(NOO)-bound peroxynitrite complex is 5.2 kcal/mol higher in energy than the N-bound species (ϩ11.0 kcal/mol compared to the OONO-bound complex). Note that the relative energies of the different peroxynitrite complexes in the absence of a hydrogen bond are also listed in Table 2.
This analysis suggests a mechanism whereby the initially generated N-bound peroxynitrite molecule would dissociate from the ferric heme and rebind through its terminal oxygen to form the most energetically favorable species (Fig. 7). Indeed, optimization of the N-bound Fe III -N(O)OO Ϫ complex in the high-spin state resulted in dissociation of the ONOO Ϫ /H 2 O ligand. As dissociation of the N-bound peroxynitrite ligand would result in a conversion of the heme Fe from low-spin to high-spin, it is plausible that this transition would allow adequate time for rotation of the unbound peroxynitrite in the NOS active site, which would then allow rebinding of peroxynitrite in its most favorable coordination mode (i.e. terminal O-bound Fe III -OONO Ϫ ). The subsequent conversion of bound peroxynitrite to nitrate could occur by the same mechanism as proposed for the NOD reaction in globins (36,37). The end product of this reaction, a low-spin ferric nitrate complex (Fe III -NO 3 Ϫ ), was calculated to be 33.4 kcal/mol lower in energy than the terminal O-bound Fe III -OONO Ϫ complex. The relative energetics of the steps and species proposed in this mechanism are shown in Fig. 8. Alternatively

Regulation of Fe II -NO oxidation in NOS
trans O-bound peroxynitrite complex (Fe III -OONO Ϫ ). The trans Fe III -OONO Ϫ complex is calculated to be Ϫ5.1 kcal/mol lower in energy relative to the Fe III -(O)O(NO) Ϫ starting complex. Following the mechanism of the NOD reaction for globins, this would then lead to O-O(NO) bond homolysis and formation of an Fe IV ϭO complex and ⅐ NO 2 . In this case, the nitrogen atom of the released ⅐ NO 2 would be in an ideal position to undergo radical rebound with the Fe IV ϭO intermediate, to give nitrate (Fig. 7). Details and results of the calculations performed to investigate this alternative pathway for formation of a trans Fe III -OONO Ϫ complex are provided in the supporting information. The breaking of the Fe III -(O)ONO Ϫ O-N bond and conversion to the trans Fe III -OONO Ϫ complex is calculated to have an activation barrier of 12.9 kcal/mol and is therefore energetically feasible. Fig. 8 shows the energetics of this alternative pathway compared with the mechanism that was discussed previously.

Discussion
NOS catalysis involves two different pathways that either lead to the production of NO via reaction of the enzyme hemeoxy species and release of NO from the Fe III -NO species (productive cycle) or lead to the production of other nitrogen oxides via formation of an enzyme Fe II -NO complex (futile cycle, Fig.  1A) (6). The rate of oxidation of the Fe II -NO complex (k ox ) was shown to be a key factor in determining the distribution and cycling of enzyme species during catalysis (6,7). Our results indicate that the reactivity of the Fe II -NO complex can be finely tuned by the NOS heme environment.

Oxygen accessibility and k ox
Oxygen binding to the NOS ferrous heme is extremely fast (higher than 1 ϫ 10 5 M Ϫ1 s Ϫ1 (38,39)). Nevertheless, the k ox rates obtained in this work are 1-4 orders of magnitude smaller. This difference might be due to an electrostatic, repulsive effect of the heme-bound NO, but it could also indicate that oxygen entry in the heme pocket is not the rate-limiting step. Thus, for WT nNOS, oxygen access is not a rate-limiting factor as the nNOS W409F mutant is able to catalyze a much faster reaction without altering the heme distal pocket. This suggests that in WT nNOS, the [Fe II -NO⅐⅐⅐O 2 ] complex is not very reactive, and oxygen can diffuse out of the heme pocket without

Regulation of Fe II -NO oxidation in NOS
reacting. On the other hand, for the iNOS enzyme, there is a marked dependence of the rates on O 2 concentration, which suggests that oxygen entry and formation of the [Fe II -NO⅐⅐⅐O 2 ] complex is almost rate-limiting. The V346I mutant shows that a small reduction in the size of the heme pocket can cause a 2-fold decrease in the rates compared with WT iNOS. It must be noted that this effect may be due to changes not only in oxygen accessibility but also in the stability of the [Fe II -NO⅐⅐⅐O 2 ] complex.
Further evidence for the influence of dioxygen accessibility on iNOS k ox rates is the increase observed in the absence of H 4 B. Although H 4 B has a significant effect on the electronic properties of the heme, it can also modify heme accessibility. These H 4 B effects on heme ligand access are likely related to structural changes that occur upon H 4 B binding within the NOSoxy domain, which also help create the dimer interface and thus promote dimerization of the eukaryotic NOS enzymes. Such structural changes caused by H 4 B are indicated by a comparison of the H 4 B-bound NOSoxy dimer structure with the structures of the iNOSoxy monomer (40) and the H 4 B-free, dimeric form of Bacillus subtilis NOS (16). In our present study, it is apparent that the different k ox rates that we observed for the K82A iNOSoxy mutant in its monomeric versus H 4 B-bound dimeric form can be attributed to the absence or presence of H 4 B. Although we were unable to study the oligomerization effect separately from the presence of H 4 B, the bulk of evidence suggests that the oligomerization state of the NOS protein plays a smaller role in determining its k ox than do structural changes within the dimeric NOS that are caused by H 4 B binding itself.

The proximal cysteine environment and k ox rates
A number of reports have proven the relevance of the Trp-409 residue in nNOS (10,14,15,18). Our results show that replacement of Trp-409 with phenylalanine causes a decrease in the redox potential and significant changes in the oxidation of the Fe II -NO complex. The nNOS W409F mutant behaves similarly to WT iNOS, with a high k ox rate and a marked oxygen dependence of the rates. The higher reactivity of the mutant can be attributed to the more reducing heme iron in the mutant, which accelerates the formation of the peroxynitrite intermediate. Increasing the pH stabilizes the six-coordinate Fe II -NO complex but still allows high k ox rates. A likely explanation of this effect is that the more alkaline pH can prevent the protonation of the cysteine, which is linked in computational studies to the cleavage of the Fe-S bond (41) as proposed previously (18). It remains unclear why the stabilization was previously found only for the W409Y mutant, but the proposed formation of a hydrogen bond between W409Y and the cysteine may not be the main factor, as (i) we were able to detect the same effect in the W409F mutant, and (ii) the crystal structure indicates that unless an important rearrangement occurs, the -OH group of the tyrosine will not be in an adequate geometry for a hydrogen bond with the cysteine, although the OH group will be in close vicinity of the sulfur and can have an electrostatic effect. Interestingly, the effects are in agreement with studies of the homologous tryptophan in bacterial NOS proteins (28 -31) (Table S2). In both B. subtilis NOS and Staphylococcus aureus NOS, the heme redox potential decreases in the sequence His Ͼ Trp(WT) Ͼ Phe (29,31). The histidine mutants show features consistent with stronger Fe II -NO/ Fe II -O 2 complexes (stronger bonding between the Fe and the diatomic ligand) and slower k ox rates. Alternatively, the elimination of the hydrogen bond to the cysteine thiolate in the Phe mutant leads to weaker -bonding and increased -backbonding in the interaction with distal ligands (29). Thus, the Phe mutants show increased reactivity of their Fe II -NO and Fe II -O 2 complexes. In summary, from the nNOSoxy results, it appears that an increased basicity of the proximal ligand can induce a greater push effect and a higher Fe II -NO reactivity, consistent with other observations in different NOS proteins (27)(28)(29)(30)(31). We observe a loss of the thiolate bond upon NO binding associated with mutants with very low heme midpoint potentials. Consequently, Fe II -NO oxidation can be modulated, but only within certain constraints of heme-thiolate bond viability.

Mechanistic implications
We have reported previously that NOS enzymes are able to catalyze the fast oxidation of Fe II -NO complexes, whereas most heme enzymes known to catalyze a similar process either catalyze the reaction at a much slower rate (neuroglobin, hemopexin) or proceed via a mechanism involving the reaction of Fe II -O 2 with NO (hemoglobin, myoglobin, flavohemoglobin) (11).
The reaction of Fe II -NO with O 2 may include formation of a transient Fe III -N(O)OO Ϫ intermediate (Fig. 7) (12,42,43). This species has been proposed for the symmetrical reaction of oxyhemoglobin with nitric oxide at alkaline pH, but not under neutral or acidic conditions (43). This has been related to the pH-dependent stability of cis-peroxynitrite. Depending on the geometry, the dissociation of this intermediate can hypothetically produce either cis-or trans-peroxynitrite. Trans-peroxynitrite would rearrange to nitrate fast, whereas cis-peroxynitrite is more stable. The stability of cis-peroxynitrite is highly dependent on pH, as protonation to peroxynitrous acid (ONOOH) lowers the barrier for isomerization (44).
Overall, our results indicate that the effects of increased pH, if any, are negligible. Thus, according to the proposed reaction scheme ( Fig. 7; see also Figs. S4 -S6 and Table S3), our data suggest that any step involving peroxynitrite formation is not rate-limiting.
Our previous mechanistic investigation of nNOS-and iNOSheme-nitrosyl oxidation points rather to a fast and direct reaction with dioxygen (11). A proposed alternative mechanism suggested the possible involvement of an outer-sphere electron transfer step between incoming molecular oxygen and the Fe II -NO heme complex of iNOSoxy or nNOSoxy (11). A comparison of outer-sphere electron transfer rates with observed k ox rates informs on this argument. In this regard, it is possible to estimate the expected rate of a limiting case of outer-sphere electron transfer (k outer-sphere ) between molecular oxygen and the heme-nitrosyl complex in nNOS using the simplified Marcus cross-relationship: k outer-sphere ϭ (k 11 ⅐k 22 ⅐K eq ⅐f 12 ) 1 ⁄ 2 (45). In this case, k 11 is the self-exchange rate for the O 2 /O 2 ⅐Ϫ couple (ϳ450 M Ϫ1 s Ϫ1 ) (46); k 22 is the self-exchange for the nNOSheme-ferrous/ferric nitrosyl couple. K eq (ϳ2.6 ϫ 10 Ϫ6 at 25°C) is the thermodynamic equilibrium constant of redox exchange

Regulation of Fe II -NO oxidation in NOS
between the O 2 /O 2 ⅐Ϫ couple (E 0 Ј ϭ Ϫ0.330 V) and the nNOS ferrous/ferric heme-nitrosyl couple (E 0 Ј Ϸ 0.0 V) (32). The factor f 12 is a function of the two self-exchange rates k 11 and k 22 and is given by the equation, ln(f 12 ) ϭ (ln(K eq )) 2 /4(ln(k 11 ⅐k 22 /Z) (where Z is the collision frequency ϳ10 11 ). Although the selfexchange rate of the nNOS ferrous/ferric heme-nitrosyl is not known, it is possible to use estimates based on measured rates in other heme proteins. Measured self-exchange rates for various cytochromes show a range extending over 5 orders of magnitude (from 10 2 to ϳ10 7 M Ϫ1 s Ϫ1 ) (47). Much lower rates have also been reported for other systems (e.g. cytochrome c peroxidase k 22 ϭ 1.9 ϫ 10 Ϫ9 M Ϫ1 s Ϫ1 ) (48). Although there are no data on self-exchange rates in cytochromes P450, these are expected to be small. In general, metalloproteins in which redox changes are associated with larger reorganization energies exhibit slow electron self-exchange rates. Resonance Raman studies point to changes in heme deformation when going from heme-Fe III -NO to heme-Fe II -NO in NOSs (19,49) as well as a significant change in the Fe-N-O angle (50). It is thus expected that the self-exchange rate for the nNOS-Fe-hemenitrosyl would not exceed the moderately fast hemeprotein cases with less reorganization (e.g. k 22 Ϸ 3 ϫ 10 3 M Ϫ1 s Ϫ1 ). Based on this estimate and the values for the other parameters mentioned above, one can estimate an expected outer-sphere electron transfer rate of k 12 (calc) Ϸ 3 ϫ 10 Ϫ4 s Ϫ1 using an [O 2 ] ϭ 270 M. This rate is 400 times slower than the observed rate for ferrous heme nitrosyl oxidation (k ox ϭ 0.115 s Ϫ1 at 25°C) (the ratio is even higher than 400 times at higher temperatures). Fig. S7 shows that this relationship holds even if we assume self-exchange rates for the nNOS-heme ferrous/ferricnitrosyl couple as high as 10 6 M Ϫ1 s Ϫ1 or even 10 8 M Ϫ1 s Ϫ1 . An outer-sphere electron transfer step limiting k ox is therefore unlikely. Instead, the analysis suggests that, either a direct attack of O 2 on the bound heme-nitrosyl or, alternatively, an inner-sphere electron transfer complex, leading to a structured transition state, is more likely to be the mechanism that governs k ox in nNOS.

Discussion of DFT results
Taken together, the results from our DFT calculations are in agreement with the conclusion that Fe II -NO oxidation in NOS (but not globins) occurs via direct attack of 3 O 2 on the Fe II -NO complex. Our DFT results show that the reaction of the NOS heme-thiolate Fe II -NO complex with 3 O 2 generates a stable, N-bound Fe III -N(O)OO Ϫ intermediate. In contrast, a direct reaction of globin Fe II -NO complexes with 3 O 2 is not supported by our DFT analysis. This is in agreement with the similar rates observed for globin Fe II -NO oxidation and NO dissociation and further supports the idea that globin Fe II -NO oxidation occurs through a rate-limiting NO/O 2 ligand exchange process, followed by the typical NOD reaction between the oxyglobin complex and NO. In NOS, such a ligand exchange is not necessary.
Our results show that stabilization of the N-bound Fe III -N(O)OO Ϫ intermediate can actually be accomplished by a simple hydrogen bond, which is in good agreement with our kinetic data. Based on this result, we re-examined the crystal structure of the nNOS Fe II -NO complex, which in fact shows a H 2 O molecule directly above the Fe II -NO unit in the active site (Fig. 5B) (34). In this structure, the H 2 O appears to be in hydrogenbonding distance of the oxygen atom of NO and one of the nitrogen atoms of L-Arg, which was co-crystallized in the active site. H 2 O molecules are also observed directly above the heme in the crystal structure of ferric iNOS ox (PDB code 1NOD) (51), again indicating that H 2 O molecules are commonly found in the active site of NOS, close to the heme, and in the needed position to interact with a peroxynitrite ligand. Based on these observations, it seems highly likely that a H 2 O molecule would be available to form the proposed hydrogen bond with the newly formed ONOO Ϫ ligand under standard catalytic conditions. We thus conclude that this is the most likely scenario for the stabilization of an N-bound Fe III -N(O)OO Ϫ intermediate, which is unique to NOS enzymes.
Finally, we investigated two pathways in which the peroxynitrite coordination mode could change from the initial, N-coordinated form to the energetically most favored, terminal oxygen atom-bound Fe III -OONO Ϫ form. We propose that this would then allow NOS-assisted sequential isomerization to nitrate through a similar mechanism as the NOD reaction proposed for globin proteins (Figs. 7 and 8; see also Figs. S4 -S6 and Table  S3). Our DFT calculations predict that formation of a hemethiolate Fe III -NO 3 Ϫ complex would be 33.4 kcal/mol lower in energy than the terminal cis O-bound Fe III -OONO Ϫ complex and 38.8 kcal/mol lower in energy than the trans Fe III -OONO Ϫ complex, which is in agreement with the fast second order rate constants determined for the iNOS k ox reaction, which indicates that there is a large driving force for nitrate formation, once peroxynitrite has been generated. These results are also in good agreement with gas-phase calculations at the UB3LYP/6-311G(d,p) level of theory, reported for the NOD reaction in globins, where the heme-imidazole Fe III -NO 3 Ϫ complex was predicted to be 28.2 kcal/mol lower in energy than the cis Fe III -OONO Ϫ peroxynitrite complex (37).

Experimental procedures
Reagents H 4 B and H 2 B were purchased from Schircks Laboratories (Jona, Switzerland). Oxygen and nitrogen gas (Medipure grade) were from Praxair (Danbury, CT), and NO gas was from Linde LLC (Murray Hill, NJ). Other reagents were purchased from sources reported previously (11,15).

Protein expression and purification
The oxygenase domains of rat nNOS and mouse iNOS were overexpressed in Escherichia coli BL21 (DE3) cells carrying the corresponding gene in the pCWori vector (38,52). WT and mutant proteins were purified as described previously (15,17,25,38,52). Protein concentration was determined from the absorbance at 445 nm of the ferrous heme-CO complex, using an extinction coefficient of 74 mM Ϫ1 cm Ϫ1 (53).

Ferrous heme-NO oxidation
The reaction was studied as described previously (11). Briefly, ferrous heme-NO complexes were prepared by titrating anaerobic solutions of ferric protein with dithionite and then with NO from a NO-saturated solution. The titrations were

Regulation of Fe II -NO oxidation in NOS
monitored using either a Cary100 or a Shimadzu UV-2401 PC spectrophotometer. The resulting protein Fe II -NO complexes were rapidly mixed with O 2 -containing buffer with the same composition as the protein buffer (including H 4 B and/or L-Arg if present in the protein buffer). Reactions were carried out in a Hi-Tech SF-61 stopped-flow instrument (Hi-Tech Scientific, Salisbury, UK) equipped with a diode array detector. Experiments were performed at 10°C. Sample solutions contained 10 M NOS in 40 mM EPPS buffer, pH 7.6 (or 40 mM CHES buffer, pH 9.5), 10% glycerol, 150 mM NaCl, and 0.5 mM EDTA. Depending on the experiment, different combinations of H 4 B or L-Arg were added. The concentrations used were 20 M H 4 B and/or 2.5 mM L-Arg unless stated otherwise. To study the oxygen dependence of the rates, different concentrations of oxygen were produced by mixing oxygen-and nitrogen-saturated buffers.

Kinetic analysis
Reaction rates were calculated from the absorbance changes at the wavelength of the Soret peak of Fe III -heme (ϳ396 nm, ferric heme recovery rate) and heme Fe II -NO (ϳ436 nm, ferrous heme-NO decay). Data were fitted to a single-exponential equation. The oxygen dependence of the observed rates versus O 2 concentration was in all cases best fit to a linear equation in the form, k obs ϭ k 1 [O 2 ] ϩ k Ϫ1 , where k 1 and k Ϫ1 are the apparent rates of O 2 association/dissociation to/from the Fe II -NO complex. For iNOSoxy proteins and nNOS W409F mutant, the value of k Ϫ1 is much smaller than k 1 and cannot be determined accurately. For these proteins, a linear fit through zero in the form k obs ϭ k 1 [O 2 ] was preferred.

Calculation of potential shift induced by NO binding
The dissociation constants for a ligand are related to the shift in redox potential, because the Nernst equation can be rewritten as follows (54), where E°c is the redox potential for the heme-NO complex, E°is the potential of the heme in the absence of NO, and K d III and K d II are the dissociation constants for NO from the Fe III and Fe II states, respectively.

Redox potentiometry
The redox potentials for the NOS heme Fe III /Fe II couples were determined as previously described in detail (27,55,56). The experiments were carried out in a glove box (Belle Technology, Dorset, UK) under nitrogen atmosphere with oxygen levels below 5 ppm. Absorption spectra were recorded in a Cary 50 spectrophotometer using a dip probe detector, and the potentials were monitored using an Accumet AB15 potentiometer (Fisher Scientific) coupled to an Ag/AgCl electrode saturated with 4 M KCl. Protein samples were ϳ10 M in potassium phosphate buffer, 100 mM, pH 7.0, with 125 mM NaCl, 25 M H 4 B, and 2.5 mM L-Arg. 1-5 M anthraquinone 2-sulfonate (E m ϭ Ϫ225 mV), phenosafranine (E m ϭ Ϫ252 mV), and benzyl viologen (E m ϭ Ϫ358 mV) were used as redox mediators. Meas-urements were done at 15 Ϯ 1°C. The fraction of protein oxidized was plotted versus the redox potential and fitted to the Nernst equation using Origin version 7.5 software (OriginLab, Northampton, MA). Reported values for the redox potential measurements are mean values Ϯ S.D. of three determinations. Actual errors may be larger, as other factors such as electrode accuracy or equilibration errors are not factored in the calculations.

Computational methods
All DFT calculations were performed with the Linux version of the program Gaussian 09 (57). All geometry optimizations of the DFT models were performed at the UBP86/TZVP level, using the unrestricted formalism of Becke's 1988 exchange functional (58) and the gradient corrections of Perdew, along with his 1981 local correlation functional P86 (BP86) (59). TZVP is a triple polarized basis set by Ahlrichs and co-workers (60,61). All calculations were performed using the polarizable continuum model as implemented in Gaussian 09 with water selected as the solvent.
The NOS active site model used for all DFT calculations was generated by starting with the coordinates from the X-ray crystal structure of the Fe II -NO complex of nNOS (PDB code 2G6K) (34) and keeping only the coordinates of the Fe-NO protoporphyrin IX complex, Cys-415, Trp-409, and Gly-417. The Cys-415 and Trp-409 residues where then truncated to their side chains, an ethanethiolate and a 3-methylindole, respectively. Gly-417 was converted to an N-ethylacetamide to better mimic its structure in the NOS active site (i.e. as an amide group, with the peptide bond present to its neighboring amino acid Val-416). As indicated in Fig. S1, the coordinates of the terminal carbon atoms of the ethanethiolate, the N-ethylacetamide, the protoporphyrin IX propionic acid groups, and the methyl group of 3-methylindole were frozen to conserve their native positions found in the NOS active site. All other atoms were left unconstrained, and the geometry of the structure was fully optimized, with a total charge ϭ Ϫ1 and a total multiplicity ϭ 2. After this initial geometry optimization, the coordinates of the protoporphyrin IX propionic acid oxygens and of the protons of the frozen methyl groups were frozen as well for all following calculations. This was necessary to prevent random rotations of the methyl and carboxylic acid groups that were observed in early calculations, which increase computational time without having any relevance for the NOD reaction studied here. Triplet oxygen ( 3 O 2 ) was then introduced into the optimized geometry of the NOS Fe II -NO complex, with the triplet spin of the O 2 molecule antiparallel to the doublet spin of the Fe II -NO complex to allow for bond formation between these two radicals. To achieve this, the wave function of the overall complex was constructed with a total multiplicity ϭ 2 and a total charge ϭ Ϫ1, and four fragments were generated for the different components of the NOS model. An "initial guess" was generated and then used to start geometry optimizations or PES scan calculations. The charge and multiplicity of the four fragments were assigned as follows: (i) Fe II -NO and 3 O 2 (charge ϭ 2, multiplicity ϭ 2); (ii) the ethanethiolate ligand (Ϫ1, 1); (iii) protoporphyrin IX (Ϫ2, 1); (iv) the 3-methylindole and N-ethylacetamide (0, 1). For successful generation of the

Regulation of Fe II -NO oxidation in NOS
correct initial wave function, with the S ϭ 1 ⁄ 2 heme Fe II -NO complex and 3 O 2 with antiparallel spins, an ON-OO distance of at least 3.26 Å was found to be necessary. This distance was thus used as the starting point in the following PES scan calculations.
The globin active site model used for the DFT calculations was generated from the crystal structure of a WT Mb Fe II -NO complex (PDB code 1NPF) (62). Here, we used the coordinates of the Fe II -NO protoporphyrin IX complex, the axial His-93 and the distal His-64 residues. Both His-93 and His-64 were truncated to the relevant side chains and modeled as 5-ethylimidazole groups (see Fig. S3). The coordinates of the terminal carbon atoms of the 5-ethylimidazole groups and of the protoporphyrin IX propionic acids were frozen, and the rest of the coordinates were fully optimized. In the next step, the carboxylic acid groups were converted to methyl groups to prevent undesired interactions of the carboxylic acids with the 5-ethylimidazole groups based on their native positioning, which does not naturally occur in Mb due to the formation of other interactions (i.e. salt bridges) between the carboxylate propionate groups and neighboring amino acids not included in our DFT model. The model was then re-optimized, and this model was then used for all following calculations. This is the same overall approach as that for the NOS model. A procedure analogous to that for the NOS model was also used to construct the intended wave function of the Fe II -NO globin model with added 3 O 2 (see above). Here, a total multiplicity ϭ 2 and a total charge ϭ 0 were used for the calculation. For the successful generation of the wave function with a doublet Fe II -NO complex and a triplet O 2 with antiparallel electron spins, it was necessary to first use an ON-OO distance of ϳ3 Å. From this starting structure, the PES scan was then conducted.