Substrate-specific Interactions with the Heme-bound Oxygen Molecule of Nitric-oxide Synthase*

We report the characterization by resonance Raman spectroscopy of the oxygenated complex (FeIIO2) of nitric-oxide synthases of Staphylococcus aureus (saNOS) and Bacillus subtilis (bsNOS) saturated with Nω-hydroxy-l-arginine. The frequencies of the νFe–O and νO–O modes were 530 and 1135 cm–, respectively, in both the presence and absence of tetrahydrobiopterin. On the basis of a comparison of these frequencies with those of saNOS and bsNOS saturated with l-arginine (νFe–O at 517 cm–1 and νO–O at 1123 cm–1) and those of substrate-free saNOS (νFe–O at 517 and νO–O at 1135 cm–1) (Chartier, F. J. M., Blais, S. P., and Couture, M. (2006) J. Biol. Chem. 281, 9953–9962), we propose two models that account for the frequency shift of νFe–O (but not νO–O) upon Nω-hydroxy-l-arginine binding as well as the frequency shift of νO–O (but not νFe–O) upon l-arginine binding. The implications of these substrate-specific interactions with respect to catalysis by NOSs are discussed.

Nitric-oxide synthases (NOSs) 2 are a family of heme proteins that synthesize NO by catalyzing the two-step oxidation of L-arginine using an O 2 -dependent mechanism (1)(2)(3)(4)(5). NOSs first hydroxylate L-arginine to N -hydroxy-L-arginine (NOHA). This intermediate then becomes a substrate for the second reaction, which produces L-citrulline and NO.
To carry out these two reactions, the heme iron at the active site of NOSs must first be reduced to be able to bind molecular oxygen (O 2 ) and to form an oxygenated complex (Fe II O 2 ). The electron is supplied by NADPH via the reductase domain in mammalian NOSs and from an independent reductase for the bacterial NOSs (6). The Fe II O 2 complex must then be activated to form the oxidizing intermediate that oxidizes the substrates (L-arginine and NOHA). The hydroxylation of L-arginine requires an additional external electron initially supplied by tetrahydrobiopterin (H 4 B) to allow oxygen activation. Oxygen activation is thought to occur via a mechanism similar to that proposed for cytochrome P450 and to involve a Compound I intermediate defined as an oxyferryl (Fe IV ϭO) heme with an associated radical (2,5,(7)(8)(9). Strong evidence suggests that an electron and possibly a proton are supplied from the H 4 B cofactor to form a peroxy (Fe III O 2 . ) or hydroperoxy (Fe III OO Ϫ H) complex (10). A second proton is then recruited, leading to the heterolytic cleavage of the O-O bond and the formation of a water molecule and a Compound I-type intermediate.
Unlike the hydroxylation of L-arginine, the conversion of NOHA to NO and citrulline requires only a transiently supplied electron to activate the heme-bound O 2 . Many mechanisms for the hydroxylation of NOHA have been put forward (8,(11)(12)(13). In general, it has been suggested that an activated form of the oxygenated complex (peroxy or hydroperoxy), produced after the transfer of an electron from H 4 B, can perform an electrophilic attack on the guanidino carbon of NOHA to form a tetrahedral intermediate that undergoes conversion to L-citrulline and NO with a concomitant electron donation back to H 4 B (14).
NOSs likely evolved two different mechanisms of oxygen activation to oxidize their two substrates in the first and second catalytic cycles, respectively. Interactions between the oxygen atoms of the Fe II O 2 complex and the substrates, together with the electron/proton donor properties of the cofactor, likely play critical roles in determining how NOSs catalyze these reactions. Although H 4 B has been more extensively characterized, including its involvement as an electron donor in the first and second reactions (10, 14 -18) as well as its likely role as a proton donor (19,20), specific details on the interactions between the hemebound O 2 and substrates and the involvement of H 4 B in modulating these interactions are scarce (21)(22)(23).
We used stopped-flow spectrophotometry and continuousflow resonance Raman spectroscopy to probe the Fe II O 2 complex of Staphylococcus aureus NOS (saNOS) with the substrate L-arginine (23). Our results showed that L-arginine binding causes a downshift in the frequency of the O-O mode but has no effect on the frequency of the Fe-O mode (23). A similar downshift of the O-O frequency has also been reported for the oxygenated complex of mammalian neuronal NOS (nNOS) (24). Analysis of the kinetics of formation and decay of the Fe II O 2 complex revealed that L-arginine stabilizes the complex against autoxidation to the ferric form in saNOS. These findings indicate that hydrogen-bonding interactions involving the heme-bound O 2 and L-arginine stabilize the Fe II O 2 complex against autoxidation. Hydrogen-bonding interactions also decrease the frequency of the O-O mode by modulating the amount of -backbonding from the heme iron (23).
To complement our first study of the Fe II O 2 complex of saNOS in the presence of L-arginine (23), we now report an investigation of the interactions of the Fe II O 2 complex in the presence of NOHA for saNOS and Bacillus subtilis NOS (bsNOS) by resonance Raman spectroscopy. Specifically, we report the frequencies of the O-O and Fe-O modes of NOHAbound saNOS and bsNOS as well as of L-arginine-bound bsNOS. Differences in the frequencies of the Fe-O and O-O modes were observed in the presence of NOHA with respect to those measured with L-arginine, indicating that the hemebound O 2 is involved in substrate-specific interactions in both NOSs. The implications of these results for the mechanisms of oxygen activation by NOSs are discussed. Enzyme Preparation-saNOS and bsNOS were expressed in Escherichia coli from the cloned genes and purified as described previously (25,26). Samples were maintained in 40 mM HEPES (pH 7.6), 150 mM NaCl, and 1 mM DL-dithiothreitol. Where indicated, NOHA (500 M) was added to the purified enzymes. For samples containing H 4 B, a concentration of 500 M was used. The association of H 4 B and NOHA with saNOS and bsNOS and that of L-arginine with bsNOS were monitored by measuring the displacement of DL-dithiothreitol bound to the ferric enzymes by optical spectroscopy (25).

EXPERIMENTAL PROCEDURES
Stopped-flow Spectroscopy-Stopped-flow experiments were carried out as described previously (23). Briefly, anaerobic protein samples (5 M) were reduced with the minimum amount of sodium dithionite required to reduce the heme. Complete reduction of the sample was verified by optical spectroscopy. Rapid mixing experiments with reduced saNOS and bsNOS and molecular oxygen were carried out at 21°C. O 2 gas (100%) was used to saturate anaerobic buffer with O 2 . This solution was then used to prepare buffers of the specified oxygen concentrations.
The kinetics of formation and decay of the Fe II O 2 complex were followed at individual wavelengths in kinetic scanning mode. Kinetic traces were recorded at 5-nm intervals from 380 to 460 nm and from 515 to 610 nm, which generated optical spectra versus time data sets. The kinetic data were analyzed using SPECFIT global analysis software (Spectrum Software Associates, Chapel Hill, NC) with a kinetic model involving four states for the data obtained from saNOS and bsNOS saturated with NOHA: state A corresponded to the initial reduced protein; state B corresponded to the transient Fe II O 2 complex; state C corresponded to an intermediate; and state D corresponded to the resting ferric form. For the data obtained from bsNOS saturated with L-arginine, a three-state model was used: state A corresponded to the initial reduced protein; state B corresponded to the transient Fe II O 2 complex; and state C corresponded to the resting ferric form. The deconvoluted optical spectra, the fits at all wavelengths, and the time course of the appearance and decay of the four kinetic states were obtained from these analyses.
Resonance Raman Spectroscopy-Resonance Raman spectra of the oxygenated intermediates were acquired with a custommade continuous-flow T-mixer as described previously (23). Prior to mixing, the reduced forms of saNOS and bsNOS (80 M) were prepared by equilibrating the enzymes with pure argon gas for at least 30 min at room temperature and by adding the minimum amount of sodium dithionite required to reduce the heme completely. Optical spectra in the visible region (from 450 to 700 nm), recorded directly from the 10-ml syringe containing the reduced proteins, were recorded to assess the complete reduction of the protein. Pure 16 O 2 and 18 O 2 gases were used to prepare oxygenated buffers of known concentrations.
The rapid mixer was made anaerobic using a 10 mM sodium dithionite solution, followed by washing with anaerobic buffer to remove the dithionite. Oxygenated buffers ( 16 O 2 and 18 O 2 in two separate mixing experiments) and the reduced proteins were mixed at a 1:1 ratio from 10-ml syringes. The output at 441.6 nm from a helium-cadmium laser (Liconix laser, Melles Griot, Ottawa, Ontario) at ϳ10 milliwatts was focused on the sample inside the channel of the quartz flow cell. The resonance Raman spectra of saNOS/NOHA/H 4 B and saNOS/NOHA samples were acquired 4.5 and 10 ms after mixing, respectively, using previously described equipment (25). Those of bsNOS/L-Arg and bsNOS/NOHA were acquired 22 and 32 ms after mixing, respectively. Several 30-s spectra (12)(13)(14)(15) were acquired and averaged. All resonance Raman spectra were obtained at room temperature (25°C) and were calibrated with the lines of indene. The spectrum of reduced myoglobin was recorded before each experiment to check for small calibration differences on different days. The spectrum of the buffer was also recorded for each experiment to subtract the quartz diffraction signal originating from the detection tube of the continuousflow mixer.

RESULTS
Stopped-flow Spectroscopy-Stopped-flow optical spectroscopy was used to obtain the rates of formation and decay of the Fe II O 2 complex of saNOS in the presence of NOHA (without H 4 B) (Fig. 1). The kinetics of formation and decay of the Fe II O 2 complex of saNOS in the presence of NOHA and H 4 B have already been reported (Table 1) (23). Reduced saNOS saturated with NOHA was mixed with O 2 -saturated buffer. The data were acquired by kinetic scanning in the 380 -460 nm (Soret region) (Fig. 1A) and 515-610 nm (supplemental Fig. 1) regions of the absorption spectrum. These kinetic scanning data were first studied by global analysis using a three-state sequential kinetic model (A 3 B 3 C). The three-state model did not fit the data properly (supplemental Fig. 2), and inconsistent rates were calculated from the two optical region data sets.
A four-state model (A 3 B 3 C 3 D) best fit the data and produced consistent results in both regions of the absorption spectrum ( Fig. 1 and supplemental Fig. 1). The kinetic traces recorded at 430 and 450 nm and the corresponding fits to the four-state model are shown in Fig. 1B. These wavelengths show that four states are indeed needed to fit all the transitions observed. The residuals of the fits at all wavelengths were small in amplitude and randomly distributed (Fig. 1C), indicating that a good fit to the four-state model was obtained. State A, with a Soret optical transition centered at 410 nm, corresponded to reduced saNOS (Fig. 1E (23). The high frequency region of the resonance Raman spectra was first obtained to determine the oxidation and spin states of the Fe II O 2 complex. The spectra were recorded at 4.5 and 10 ms after mixing reduced saNOS/NOHA with the O 2 -saturated buffer in the presence and absence of H 4 B, respectively. At these times, the Fe II O 2 complexes had reached maximum concentrations of 81 and 72% of the heme available for saNOS/ NOHA with (23) and without H 4 B (Fig. 1D), respectively. The spectra showed that the oxidation state marker band 4 was at 1374 cm Ϫ1 and that the coordination and spin state marker band 3 was at 1501 cm Ϫ1 in samples with H 4 B (Fig. 2, traces A and B) and without H 4 B (traces C and D). The Fe II O 2 complexes of saNOS/NOHA and saNOS/NOHA/H 4 B were thus ferric and low spin like the oxygenated complexes of other heme proteins (24,30).
The shoulder on the 4 mode at 1349 cm Ϫ1 and the small intensity 3 mode at 1468 cm Ϫ1 indicated that a small amount of reduced five-coordinate saNOS was present, likely the remains of starting material predicted to correspond to 12 and 11% of the reaction mixtures at 4.5 and 10 ms following the initiation of the reactions based on the kinetic data with NOHA/H 4 B (23) and NOHA (Fig. 1D), respectively. These assignments are supported by the resonance Raman spectra of saNOS/NOHA obtained later after the initiation of the reactions (Fig. 3). At 56 ms (Fig. 3, trace B) and 81 ms (trace A) after mixing, the 3 and 4 lines of reduced saNOS were no longer observed, unlike the 10 ms spectrum (trace C). These results are consistent with stopped-flow data indicating that no reduced form remained at those times (Fig. 1D). Also consistent with stopped-flow data, the spectrum obtained at 81 ms displayed a 3 line at 1488 cm Ϫ1 , which corresponded to ferric five-coordinate saNOS and which represented 8% of the heme at that time. Also, at 81 ms, state C had reached its maximum concentration. Our results show that, with a 4 line at 1372 cm Ϫ1 and a 3 line at 1501 cm Ϫ1 , this intermediate was mostly six-coordinate and low spin.
Isotopic substitution was used to identify an oxygen-sensitive mode in the high frequency region. The 16 (Table 2). However, the line at 1135 cm Ϫ1 was significantly sharper in the spectrum of the Fe II O 2 complex with NOHA (width at a halfheight of 20 cm Ϫ1 ) than in that of substrate-free saNOS (width at a half-height of 22 cm Ϫ1 ). In the 16 O 2 -minus-18 O 2 difference spectrum of saNOS/NOHA recorded 81 ms after mixing (Fig. 2, trace G), the intensity of the O-O line was much lower than at 10 ms (trace F), indicating that only a small amount of the Fe II O 2 complex remained after 81 ms, which was expected based on the stopped-flow data (Fig. 1D).
Resonance Raman Spectroscopy in the Low Frequency Region-The resonance Raman spectra were obtained in the low fre-    (Fig. 4, trace I).
Analysis of the resonance Raman spectra revealed that substrate and H 4 B binding caused changes in the intensity of some lines. The intensity of the 7 out-of-plane mode at 677 cm Ϫ1 of the spectrum of substrate-free saNOS was much higher than that of the ␥ 15 out-of-plane mode at 693 cm Ϫ1 (23). As observed previously with L-arginine (23), upon binding of NOHA, the intensity of 7 decreased and that of ␥ 15 increased. These lines were therefore of nearly equal intensity in the presence of substrate (Fig. 4, traces A-D). Although H 4 B did not affect the frequencies of the Fe-O and O-O stretching modes, H 4 B binding decreased the intensity of the heme mode at 397 cm Ϫ1 (Fig.  4, traces A-D), which is in the region were the bending modes of the heme propionates are usually observed (31). A mode at 548 cm Ϫ1 was observed in the Fe II O 2 spectra obtained with NOHA and H 4 B (Fig. 4, traces A and B) that was absent in the spectra obtained with NOHA alone (traces C and D). The line at 548 cm Ϫ1 likely corresponded to the ␦ Fe-NO mode of the small population of the Fe III NO complex present 4.5 ms after mixing (23).

Substrate-specific Interactions with O 2 in NOS
plemental Fig. 4). The rate of formation of the Fe II O 2 complex (201 s Ϫ1 ) and the rate of decay (1.6 s Ϫ1 ) measured for bsNOS saturated with L-arginine were very similar to those obtained for saNOS/L-Arg (Table 1) and are in agreement with previously published data for bsNOS/L-Arg (28). The results from the complete analyses of the stopped-flow kinetic data with bsNOS/L-Arg with a three-state model are presented in supplemental Fig. 3. With NOHA-saturated bsNOS, the rate of formation of the Fe II O 2 complex was 201 s Ϫ1 . There was no evidence that state C, observed in the course of reaction of saNOS/ NOHA with O 2 (Fig. 1), was formed with bsNOS/NOHA. However, the decay of the Fe II O 2 complex could not be modeled by a single exponential, so a four-state model was used from which we obtained rates of decay to the ferric state of 4.9 and 1.2 s Ϫ1 , respectively. The optical spectrum calculated for the Fe II O 2 complex of bsNOS/NOHA (supplemental Fig. 4, E and F) was very similar to that of saNOS/NOHA (Fig. 1, E and  F), with a Soret transition blue-shifted with respect to that of saNOS and bsNOS saturated with L-arginine ( Table 1). The results from the complete analyses of the stopped-flow kinetic data with bsNOS/NOHA are presented in supplemental Fig. 4. The resonance Raman spectra of the Fe II O 2 complexes of bsNOS/L-Arg and bsNOS/NOHA were acquired at 22 and 32 ms, respectively, after mixing the reduced proteins with buffer saturated with 40 and 100% oxygenated buffers ( 16 O 2 and 18 O 2 ), respectively. We used a spectral window spanning the 400 -1200 cm Ϫ1 region that allowed the simultaneous identification of the Fe-O and O-O modes (Fig. 5). The 16 O 2 -minus-18 O 2 difference spectra of L-arginine-saturated bsNOS (Fig. 5, trace E) and NOHA-saturated bsNOS (trace F) revealed two isotopesensitive lines. Similar to saNOS/L-Arg (23), the lines corre-sponding to the O-O modes of bsNOS/L-Arg are large (Fig. 5, trace E). Peak fitting in this region allowed the identification of two O-O modes at 1124 and 1136 cm Ϫ1 that shifted to 1064 and 1073 cm Ϫ1 with 18 O 2 , respectively (supplemental Fig. 5B). The Fe-O mode was identified at ϳ517 cm Ϫ1 with 16 O 2 and at ϳ487 cm Ϫ1 with 18 O 2 (Fig. 5, trace E). These frequencies are identical to those of saNOS/L-arginine (Table 2). For the oxygenated complex of bsNOS/NOHA, a single O-O mode was identified at 1136 cm Ϫ1 that shifted to 1071 cm Ϫ1 with 18 O 2 (Fig. 5, trace F; and supplemental Fig. 5A). The Fe-O mode was observed at ϳ530 cm Ϫ1 with 16 O 2 and at ϳ497 cm Ϫ1 with 18 O 2 (Fig. 5, trace F). These frequencies are nearly identical to those of saNOS/NOHA ( Table 2). The resonance Raman spectrum of the oxygenated complex of substrate-and pterin-free bsNOS was not obtained, as the stopped-flow kinetic experiments did not show the accumulation of an oxygenated complex with a Soret band in the 425-430 nm range in the course of the reaction of ferrous bsNOS with oxygen (data not shown).

Correlation between the Fe-O and O-O Frequencies-
There is a well established negative correlation between the Fe-XO and X-O frequencies of the Fe III NO (six-and five-coordinate), Fe II CO (six-and five-coordinate), Fe II NO (five-coordinate), and Fe II O 2 complexes (five-coordinate). This correlation arises from -backbonding from the Fe d orbital to the * orbital of the heme-bound ligand, which decreases the X-O frequency and increases the Fe-X frequency. In Fe II CO complexes, the inverse correlation line for the Fe-CO and C-O frequencies of six-coordinate heme proteins is displaced below the curve of five-coordinate heme complexes as the ligation of the heme sixth ligand, competes with CO for -bond formation with the heme iron and lowers the frequencies of Fe-CO . Although the negative correlation between Fe-O and O-O was observed for five-coordinate heme complexes, it does not seem to be conserved for the six-coordinate Fe II O 2 complexes (Fig. 6). Indeed, six-coordinate Fe II O 2 complexes of heme proteins and heme models do not, as a whole, display a linear relationship (Fig. 6). Spiro and co-workers (32,33) have pointed out that, in contrast to Fe II CO complexes, the Fe-O frequencies of six-coordinate heme proteins are higher than those of five-coordinate compounds, which is expected based solely on -backbonding. This led them to speculate that the O 2 -orbital is better matched to the Fe dz2 orbital than that of CO so that the axial sixth ligand of six-coordinate Fe II O 2 complexes is less effective in -bonding competition. Nevertheless, the Fe-O frequencies for thiolatecoordinated hemeproteins are lower (517-541 cm Ϫ1 ) than those of histidine-coordinated heme proteins (542-570 cm Ϫ1 ) (Fig. 6), which fall on the extension of the correlation of fivecoordinate complexes, indicating that the proximal cysteine of thiolate-coordinated heme proteins is able to compete with O 2 for -bonding with the heme iron. In fact, the competition for -bonding by the proximal cysteine was proposed to explain the general trend toward a positive correlation for the data points of thiolate-coordinated heme proteins (Fig. 6, black squares (saNOS) and gray circles (P450 cam )) and thiolate-coordinated heme models (black triangles) (30). The positive correlation was proposed to arise from a combined effect of increased -backdonation from the heme iron to the * orbital of O 2 upon binding of the proximal cysteine (which lowers the Resonance Raman Spectra of State C of saNOS-Previous equilibrium kinetic studies on H 4 B-deficient NOS/NOHA using NADP(H) 2 or peroxide as the reductant revealed that nitroxyl is synthesized instead of NO because an Fe II NO complex was observed rather than an Fe III NO complex (29,34). Our conditions were different because reduced saNOS/NOHA reacted in single turnover condition without an excess source of electron. We used resonance Raman spectroscopy to obtain information about state C, which had a broad optical Soret transition (Fig. 1E). The resonance Raman spectrum of this intermediate acquired 81 ms after mixing confirmed that an Fe III NO complex was not formed, as no ␦ Fe-NO mode at 548 cm Ϫ1 was observed in the low frequency region (Fig. 4, traces E and F) (23,26). An Fe II NO complex, which would be in the six-coordinate and low spin state if NOHA was still bound to saNOS (26), was not formed either because the strong Fe-NO / ␦ Fe-N-O modes expected near 555 cm Ϫ1 were not detected. Spectra obtained in the high frequency region revealed that state C was six-coordinate and low spin with 4 and 3 lines at 1372 and 1501 cm Ϫ1 , respectively (Fig. 3, trace A). The frequency of the 4 line did not correspond to the 4 frequencies established for the Fe III NO and Fe II NO complexes of saNOS (26). In addition, a 3 line near 1508 cm Ϫ1 , which would be expected for a five-coordinate Fe II NO complex if no NOHA was bound to saNOS (26), was not detected. We thus conclude that nitroxyl is not synthesized by H 4 B-free saNOS/NOHA. State C could be an intermediate formed by the reduction and/or protonation of the oxygenated complex, but no O-O and Fe-O stretching modes associated with state C could be detected. The identity of state C was not investigated further.

DISCUSSION
Interactions of the Heme-bound O 2 with L-Arginine and NOHA-We reported previously that the presence of L-arginine causes a 12 cm Ϫ1 downshift in the O-O frequency of the Fe II O 2 complex of saNOS, but does not modify the frequency of the Fe-O mode (23). This effect of L-arginine correlates with the increased stability of the Fe II O 2 complex, which decays at rates of 5.6 and 1 s Ϫ1 compared with 39.6 s Ϫ1 without L-arginine (23). These results suggest that hydrogen-bonding interactions with the heme-bound O 2 stabilize the complex against autoxidation to the ferric form.
In the Fe II O 2 complex of saNOS/NOHA presented here, the frequency of the O-O mode was 1135 cm Ϫ1 , which is the same as that for substrate-free saNOS (23). The lack of sensitivity to NOHA binding of the O-O frequency has already been observed with the nNOS oxygenase domain (22,24) and very recently with the mammalian inducible nitric-oxide synthase (iNOS) oxygenase domain (35). However, unlike L-arginine, NOHA caused an upshift of the Fe-O frequency from 517 to 530 cm Ϫ1 in saNOS and bsNOS, assuming that the substratefree protein has a Fe-O mode at 517 cm Ϫ1 like substrate-free saNOS. These results indicate that NOHA binding modifies the frequency of the Fe-O mode, but not that of the O-O mode, whereas L-arginine binding modifies the frequency of the O-O mode, but not that of the Fe-O mode.
With regard to H 4 B, our results indicate that a pterin is required for the synthesis of NO from NOHA in saNOS, as an Fe III NO complex was observed in single turnover experiments performed in the presence of H 4 B (23), but not in its absence (Fig. 1). This means that H 4 B is able to support NO synthesis with saNOS as with other bacterial NOSs (28,36), but it does not mean that H 4 B is the natural cofactor of saNOS. The Fe II O 2 intermediate formed with NOHA-saturated saNOS was the same irrespective of whether H 4 B was present or not, as indicated by the similarity of the optical spectra (Table 1) (Table 2). These results suggest that H 4 B is involved in reactions that occur after the formation of the oxygenated complex, as is the case with L-arginine (23), and are consistent with the results of Wei et al. (10,14), who showed that an H 4 B radical is formed only after the The polarity effect, including hydrogen-bonding interactions of substrates with heme-bound ligand, is very likely given that hydrogen bond formation was demonstrated for nNOS and iNOS from the deuterium effect on heme-bound CO (38 -40) and suggested from the crystal structures of heme-CO and heme-NO complexes of nNOS, mammalian endothelial NOS (eNOS), and bsNOS (41,42). Particularly interesting with respect to the results presented in this study are the crystal structures the Fe II NO complexes of bsNOS/L-arginine and bsNOS/NOHA revealing that the heme-bound NO is likely involved in hydrogen-bonding interactions with the guanidinium group of L-arginine and water for the L-arginine-bound enzyme. With NOHA, only hydrogen bonds to the NH of NOHA would be formed as the water molecule is displaced away from the heme (Fig. 7) (42).
Several studies from the last 10 years indicate that that the Fe-O mode is sensitive to hydrogen-bonding interactions involving the proximal oxygen atom. Indeed, the truncated hemoglobins and HemAT sensor protein display smaller frequencies of the Fe-O mode (554 -564 cm Ϫ1 ) than myoglobin (570 cm Ϫ1 ) in response to hydrogen bonding with distal residues (43,44). It has been established that the hydrogen bond with the proximal oxygen (the oxygen directly ligated to the heme) causes the downshift of the Fe-O frequency, as hydrogen-bonding interactions with the terminal oxygen, as in myoglobin, do not shift the Fe-O frequency (43)(44)(45). The O-O mode is also sensitive to hydrogen-bonding interactions, but unlike the Fe-O mode, it is sensitive to hydrogen bonding involving the terminal oxygen (46).
By analogy to the truncated hemoglobins and the HemAT sensor, the upshift of the Fe-O frequency from 517 to 530 cm Ϫ1 upon NOHA binding could be attributed to the disruption of a hydrogen bond to the proximal oxygen, already present in substrate-free saNOS and L-arginine-bound saNOS (Fig. 7, Scheme  A). This hydrogen bond would rather likely involve a water molecule(s), which is consistently observed in the heme pocket of crystal structures of NOSs (41,42) because no heme pocket group appears to be close enough to the heme-bound ligands to play that role. In Scheme A, additional hydrogen bonding to the terminal oxygen upon L-arginine binding would cause a downshift of O-O due to modulation of -backbonding. Because the frequency of the O-O mode does not change upon NOHA binding with respect to substrate-free saNOS, Scheme A implies that no additional hydrogen-bonding interactions are made with the terminal oxygen. The kinetics of formation and decay of the oxygenated complex support this model, as the decay rate remains high with NOHA-bound saNOS (Table 1). This model may also explain the low frequency of the Fe-O mode in saNOS and bsNOS with respect to cytochrome P450 cam , which was unexpected because the Fe-Cys bond is actually weaker in NOSs (Fe-Cys stretching mode at 337 cm Ϫ1  (42). Also shown are models of the oxygenated complexes for saNOS and bsNOS (substrate-free (C) and with L-arginine (D) and NOHA (E)) based on our resonance Raman results (water molecules are included based on A and B). In Scheme A, hydrogen bonds are formed in substrate-free NOS from water to both oxygen atoms of the heme-bound O 2 . With L-arginine, additional hydrogen-bonding interactions with the terminal oxygen increase -backbonding from the heme iron and thus lower the O-O frequency (ϳ70% of the enzymes display the 1123 cm Ϫ1 mode) (23). With NOHA, the hydrogen bond to the proximal oxygen is disrupted, thus increasing the frequency of Fe-O . In Scheme B, there is no hydrogen bond involving the heme-bound O 2 in the absence of substrate. With L-arginine, 70% of the enzymes form strong hydrogen bonds with the heme-bound terminal O 2 , and ϳ30% remain hydrogen bond-free (not shown) (23). With NOHA, all saNOS molecules form hydrogen bonds with the heme-bound O 2 (strong with the proximal oxygen). The hydrogen bond is drawn from NH instead of NOH, as the crystal structures in A and of mammalian NOSs (61,62) indicate that the hydroxyl is pointing away from the heme iron. for iNOS oxygenase domain (47), 338 cm Ϫ1 for eNOS (48), and 342 cm Ϫ1 with bsNOS (47)) than in P450 cam (Fe-Cys at 351 cm Ϫ1 ) (49,50). Scheme A is, however, difficult to reconcile with the hydrogen-bonding pattern inferred for both atoms of NO from the crystal structures of the Fe II NO complex of bsNOS/ NOHA described above, analysis of the oxygenated complex of a NOS model by density functional theory methods indicating that hydrogen-bonding interactions involving both oxygen atoms of the heme-bound O 2 are likely (13), and the discussion of Poulos and co-workers (41) describing the orbitals of NOHA that are better matched for hydrogen-bonding interactions with the heme-bound O 2 than those of L-arginine. Also from Scheme A, one has to assume that a water molecule can be stabilized in the heme pocket in the absence of substrate and make strong hydrogen bonds to the proximal and distal oxygen atoms of the heme-bound O 2 .
A second scenario may be envisioned (Fig. 7, Scheme B). As discussed earlier, the proximal cysteine of thiolate-coordinated heme proteins is able to compete for the Fe dz2 orbitals and thus weaken the -bond between O 2 and iron (reduced Fe-O frequency). In both mammalian and bacterial NOSs, the proximal cysteine shares hydrogen bonds with the nitrogen atom of the side chain of a Trp and amide protons (1,51,52), so the electron-donating properties of the proximal cysteine may be modulated by strengthening or weakening these interactions (53,54). Interestingly, it was recently shown that the strength of the Fe-Cys bond is modulated in response to different substrate binding in P450 (55). With these ideas in mind, we propose that the interactions between the heme-bound O 2 with NOHA or an NOHA/water network involve the formation of a strong hydrogen bond to the proximal oxygen. This hydrogen bond should decrease the Fe-O frequency. However, if the new hydrogen-bonding interactions impose a better alignment of the O 2 orbitals with the Fe zz2 orbitals, the proximal cysteine may become a poorer competitor for -bond formation with the heme iron so that the net effect could be a stronger -bond between the iron and O 2 (increased Fe-O frequency). This hydrogen bond network would also stabilize Fe-O-O in a more rigid configuration that is indeed suggested by the increased sharpness of the O-O line with NOHA present (20 cm Ϫ1 in saNOS and 15.4 cm Ϫ1 in bsNOS) than without NOHA (22 cm Ϫ1 in saNOS). This scenario is consistent with studies by density functional theory methods indicating that hydrogen bonds to both oxygen atoms are formed in the oxygenated complex of NOS/NOHA (12). It is also consistent with the crystal structure of the Fe II NO complex of bsNOS discussed previously (42). With L-arginine, a strong hydrogen bond to the terminal oxygen would cause a downshift of O-O to 1123 cm Ϫ1 due to the modulation of -backbonding (Fig. 7, Scheme B). Implications for Catalysis-As presented in the Introduction, the catalytic cycles of NOSs start with the reduction of the heme and the binding of O 2 . In mammalian NOSs, this electron is transferred from NADPH via the reductase domain, whereas in bacterial NOSs, because they lack the reductase domain, this electron would be transferred from a separate protein (6). It is proposed that the Fe II O 2 complex is then activated with an electron to form a peroxy complex. In this context, Scheme B, which proposes that hydrogen-bonding interactions between the heme-bound O 2 and L-arginine involve primarily the distal oxygen, would be consistent with EPR/electron nuclear double resonance investigations of the cryoreduced peroxy complexes of eNOS/L-arginine, which indicated that hydrogen-bonding interactions involving the distal oxygen atom of the peroxy complex and a water/guanidinium network occur (21). Interactions of the water/guanidinium network with the terminal oxygen of the heme peroxy are expected to favor double protonation of the distal oxygen and to ultimately lead to O-O bond scission and Compound I formation. In contrast, quantum mechanical/molecular mechanical study of the electron transfer from H 4 B to the oxygenated complex in an L-argininebound NOS model suggests that protonation of both oxygen atoms of the oxygenated complex occurs before electron transfer to produce Compound I by a novel mechanism (9). If this is indeed how Compound I is formed in NOSs, Scheme A, which involves hydrogen bonds to both oxygen atoms of the hemebound O 2 , would be attractive, as the protonation of both oxygen atoms could be favored. All our results were obtained from the characterization of oxygenated complexes (Fe III O 2 . ), thus offering no clue as to whether electron or protons are first transferred to activate the oxygen. With regard to the second catalytic cycle, the formation of a hydrogen bond to the proximal oxygen of the Fe II O 2 complex with NOHA or an NOHA/water network proposed in Scheme B could remain in the peroxy complex, thus favoring protonation of the proximal oxygen atom, which would inhibit cleavage of the O-O bond (13,56,57). It must be pointed out that similar but not identical EPR signals have been obtained upon cryoreduction of the Fe II O 2 complex of the eNOS oxygenase domain saturated with L-arginine and NOHA, suggesting that substrate-specific interactions within the peroxy intermediate do occur (21). An alternative scenario, i.e. that the peroxy intermediate itself hydroxylates NOHA, may also be envisaged (21). In this case, the preferred interaction of the proximal oxygen with NOHA proposed in Scheme B may be essential to stabilize the peroxy intermediate and to prevent protonation of the hemebound O 2 . Finally, in the context of Scheme A, it is the displacement of the water molecule away from the heme that could be critical for the proper oxidation of NOHA. Instead of protecting the heme-bound O 2 from protonation by hydrogen-bonding interactions as in Scheme B, the displacement of the water molecule might protect the O-O bond from being cleaved by withdrawing a source of proton(s) and by being not in direct hydrogen-bonding interactions with the proximal and distal oxygen atoms of O 2 . Clearly, more studies are needed to fully understand the interactions between O 2 , substrates, and water at the active site of NOSs and to obtain evidence as to whether Scheme A or B better describe those interactions.