The Ferrous Dioxygen Complex of the Oxygenase Domain of Neuronal Nitric-oxide Synthase*

The mechanisms by which nitric-oxide synthases (NOSs) bind and activate oxygen at their P450-type heme active site in order to synthesize nitric oxide from the substrate L -arginine are mostly unknown. To obtain information concerning the structure and properties of the first oxygenated intermediate of the enzymatic cycle, we have used a rapid continuous flow mixer and resonance Raman spectroscopy to generate and identify the ferrous dioxygen complex of the oxygenase domain of nNOS (Fe 2 1 O 2 nNOSoxy). We detect a line at 1135 cm 2 1 in the resonance Raman spectrum of the intermediate formed from 0.6 to 3.0 ms after the rapid mixing of the ferrous enzyme with oxygen that is shifted to 1068 cm 2 1 with 18 O 2 . This line is assigned as the O-O stretching mode ( n O-O ) of the oxygenated complex of nNOSoxy. Rapid mixing experiments performed with nNOSoxy saturated with L -arginine or N v -hydroxy- L -arginine, in the presence or absence of (6 R )-5,6,7,8-tetrahydro- L - biopterin, reveal that the n O-O line is insensitive to the presence of the substrate and the pterin. The optical spectrum of this ferrous dioxygen species, with a Soret band wavelength maximum at 430 nm, confirms the identification of the previously reported oxygenated complexes generated by stopped flow techniques.

Nitric oxide (NO) 1 is an important messenger and an effector molecule involved in numerous physiological functions in the cardiovascular, nervous, and immune systems of mammals (1)(2)(3). It is generated by two constitutive isoforms of nitricoxide synthase (NOS) that were first isolated from rat brain (nNOS, NOSI) and vascular endothelium (eNOS, NOSIII) and a cytokine-inducible form that was first isolated from macrophages (iNOS, NOSII) (4). NOSs are homodimers composed of 130 -160-kDa subunits, each comprising an N-terminal oxygenase domain that contains binding sites for heme, L-arginine, and (6R)-5,6,7,8-tetrahydro-L-biopterin (H 4 B) and a C-terminal reductase domain that contains binding sites for NADPH, FAD, and FMN (5)(6)(7)(8)(9)(10). As in cytochrome P-450 and chloroperoxidase, the heme of the oxygenase domain is axially coordinated to the thiol group of an endogenous cysteine residue. The electron flow from the reductase domain to the heme domain is controlled by a Ca 2ϩ /calmodulin binding site located in the central portion of each of the NOS isoforms (11).
Several lines of evidence indicate that the heme-iron is involved in NO synthesis. NOS activity is inhibited by carbon monoxide (12)(13)(14) and by newly synthesized NO molecules, which form a ferrous nitrosyl complex under turnover conditions (15). Moreover, the recently determined structures of the oxygenase domain of iNOS (16,17) and eNOS (17,18) show that the binding site of the substrate lies above the heme on the distal side. The H 4 B cofactor also binds in the vicinity of the heme, close to the heme propionates in a perpendicular orientation. Both the substrate and H 4 B molecules participate in an extensive hydrogen bond network that includes a heme propionate, water molecules, and several amino acids. The proximity of the L-arginine binding site to the heme is consistent with the interactions of substrates with heme-bound ligands such as CO, NO, and imidazole, which were revealed by optical (19), EPR (20), and resonance Raman spectroscopies (21)(22)(23). In addition, the substrates and H 4 B, when present, cause a shift in the spin state of the ferric enzyme (24). It is not clear if H 4 B is also involved in catalysis, although recent results suggest that H 4 B may provide an electron required for oxygen activation (25).
The formation of nitric oxide by NOSs involves the NADPHdependent oxidation of L-arginine to citrulline using oxygen as a co-substrate. The complete enzymatic cycle requires two successive hydroxylation reactions proceeding via N -hydroxy-Larginine (NOHA) as an intermediate (26). The first steps in the catalytic cycle of NOS are thought to involve reduction of the heme and binding of an oxygen molecule. Subsequently, the hydroxylation of L-arginine to NOHA is thought to occur through a P-450 type mechanism involving an oxyferryl porphyrin cation complex, Fe 4ϩ ϭ O, while the second step, the hydroxylation of NOHA to citrulline and NO, has been suggested to involve a hydroperoxo species, Fe 3ϩ (OOH), formed by abstraction of a hydrogen atom from the substrate NOHA (27). Recent results with the oxygenase domain of nNOS (nNOSoxy) (28,29) and full-length nNOS (25,30) showed the formation of intermediates characterized by optical spectra that are consistent with the formation of the dioxygen complex, although different optical spectra were reported depending on the specific technique that was used. To obtain detailed structural information about intermediates of the NOS catalytic cycle, we have used resonance Raman spectroscopy to characterize oxygenated intermediates of nNOSoxy generated by the rapid mixing of the ferrous enzyme with oxygen. We show the rapid formation of the ferrous dioxygen species following mixing, and we unambiguously identify the O-O stretching mode of the heme-* This work was supported by National Institutes of Health Grants GM54806 and GM54812 (to D. L. R.) and GM51491 (to D. J. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed: bound dioxygen molecule. We also show that the O-O line is insensitive to the presence of substrates and H 4 B. The data from the oxy complexes are in contrast to results obtained with carbon monoxide that revealed steric or polar interactions between the CO and the L-arginine molecules and thereby define structural differences between the ferrous dioxygen and carbon monoxide complexes.

EXPERIMENTAL PROCEDURES
Materials-H 4 B was purchased from Alexis Biochemicals (San Diego, CA). NOHA and L-arginine were from Sigma. N 2 and 16 O 2 gas were from Tech Air, and 18 O 2 gas was from Icon (Mt. Marion, NY).
Enzyme Preparation-The oxygenase domain of nNOS was expressed in Escherichia coli from the cloned cDNA and was purified as described previously (28) in the absence of L-arginine and H 4 B. The hemoprotein concentration was determined from the absorbance at 444 nm of the CO complex using an extinction coefficient of 76 mM Ϫ1 cm Ϫ1 (14). When required, L-arginine, H 4 B, and NOHA were added to the purified enzyme in 100-, 3-5-, and 10-fold excess concentration over heme, respectively. The incubation of the enzyme with substrates and cofactor was performed at room temperature for ϳ12 h. The binding of these compounds was followed by optical spectroscopy from the change in the spin state of the ferric enzyme (31). The enzyme preparations were then used immediately without freezing.
Resonance Raman and Optical Spectra of nNOSoxy Oxygenated Intermediates-The buffer used for kinetic measurements was 40 mM EPPS, pH 7.6, containing 1 mM DL-dithiothreitol. To prepare the ferrous form of nNOSoxy, the ferric enzyme (80 -125 M) was equilibrated with prepurified nitrogen gas for 30 min at room temperature, and the heme was then reduced with sodium dithionite. The amount of sodium dithionite used was sufficient to reduce the enzyme and leave an excess of 50 M over heme. The concentration of the dithionite solution used to reduce nNOSoxy was determined from the reduction of ferric cytochrome c. Complete reduction of nNOSoxy was verified by optical spectroscopy. Oxygen-containing buffer solutions were prepared by equilibrating deoxygenated buffer with 16 O 2 or 18 O 2 at 1 atmosphere pressure.
The rapid mixer used here was described previously (32). Prior to use, oxygen was removed from the mixer, the observation chamber, and the buffer lines with a solution of 10 mM sodium dithionite, which was left in the apparatus for 1 h. The apparatus was then washed with anaerobic buffer containing 50 M sodium dithionite prior to connecting the syringes containing the ferrous protein and oxygen-saturated buffer.
The output at 413.1 nm from a krypton ion laser (Spectra Physics), ϳ50 milliwatts, was focused to a ϳ30-m spot on the continuously flowing sample in the flow cell. The position of the laser focusing point was moved along the flow direction to obtain the desired time point. The flow rate used varied between 1.2 and 1.5 s/m. The data were measured at time delays after mixing ranging from 0.6 to 3 ms. All measurements were recorded at room temperature (ϳ25°C). The resonance Raman spectra were calibrated with the lines of indene. The resonance Raman spectrum of reduced cytochrome c was recorded prior to each mixing experiment and was used to adjust small differences in the calibration of spectra from different mixing experiments.
To obtain the optical spectrum of the ferric, reduced, and ferrous dioxygen forms of nNOSoxy, the light emitted by a UV-visible lamp was focused on the flow cell, and the transmitted light was collected and focused on the entrance slit of the spectrometer. The grating used was 30 lines/mm instead of the 1200 lines/mm grating used to record resonance Raman spectra. The optical spectra were calibrated with the lines in the absorption spectrum of holmium oxide.

RESULTS AND DISCUSSION
Resonance Raman Spectrum of Fe II (O 2 ) nNOSoxy-To obtain the resonance Raman spectrum of the oxygenated form of the nNOSoxy domain, we first measured the spectrum of the ferric protein as shown in Fig. 1a. The high frequency of the strong line at 1371.5 cm Ϫ1 , 4 , the oxidation state marker line, confirms the ferric oxidation state of the heme. Upon reduction by minimal dithionite the ferrous spectrum, shown in Fig. 1b, is obtained, in which 4 shifts to lower frequency, forming a doublet with maxima at 1349 and 1359 cm Ϫ1 . When the reduced enzyme is reacted with oxygen, using a submillisecond continuous flow mixer (32), the spectrum shown in Fig. 1c is obtained within 0.6 ms of the mixing. As in the oxy forms of other heme proteins, the oxidation state marker line as well as the other lines in the spectrum shift to frequencies similar to those seen in the oxidized forms of the heme. However, the line in the 1132-cm Ϫ1 region is noticeably stronger in the spectrum of the oxygen-reacted species than in either of the other forms of the nNOSoxy domain.
To determine if any oxygenated products contribute to the spectrum, the reaction was carried out both with 16 (Fig. 2b). We attribute the low intensity line, which remained at 1132-cm Ϫ1 in the Fe 2ϩ ( 18 O 2 ) spectrum, to an overlapping porphyrin mode based on similar observations in ferrous dioxygen complexes of myoglobin (33) and cytochrome P450 (34). This porphyrin mode is canceled in the Fe 2ϩ ( 16 O 2 ) minus Fe 2ϩ ( 18 O 2 ) difference spectrum and the oxygen-sensitive line is then very well defined with a strong positive peak at 1135 cm Ϫ1 and a strong negative peak at 1068 cm Ϫ1 (Fig. 2c). We assign the 1135-cm other oxygenated complexes of heme proteins (Table I). With cytochrome P450 model compounds, it has been shown that the O-O line of the ferrous dioxygen complex is sensitive to modifications of the axial-thiolate ligand (34) ( Table I). Crystallographic data indicate that in iNOSoxy and chloroperoxidase, the proximal cysteine is involved in additional hydrogen bonds with surrounding residues as compared with cytochrome P450 (35,36). This observation is consistent with resonance Raman studies of the CO complexes of nNOS and chloroperoxidase, which indicate that the cysteine-iron linkage is weaker in these proteins than in cytochrome P450 (21). Thus the slightly lower frequency of the O-O line in nNOSoxy may result from a weaker iron-cysteine bond as compared with that of cytochrome P450.
Optical Spectrum of Fe II (O 2 ) nNOSoxy-By resonance Raman scattering, we have unambiguously identified the ferrous dioxygen intermediate of nNOSoxy produced by the rapid mixing of the reduced enzyme with oxygen. To compare this intermediate with those reported by others, we have recorded the optical spectrum of this species in the same continuous flow mixing apparatus that was used to obtain the resonance Raman spectrum. In the optical spectrum, recorded 0.6 ms after mixing (Fig. 3), the wavelength maximum of the Soret band is located at 430 nm, a value very close to that of the oxygenated intermediates of nNOSoxy (28) and full-length nNOS (30) gen-erated in stopped flow apparatus by the rapid mixing of the ferrous enzymes with oxygen (427 nm) ( Table II). The Soret band of the dioxygen complex of nNOS and nNOSoxy is located at a higher value than that of the dioxygen complex of cytochrome P450, 418 nm, but at a similar position to that of chloroperoxidase (ϳ430 nm) ( Table II). The longer wavelength for the Soret transition in chloroperoxidase has been suggested to result from a more polar heme pocket as compared with that in cytochrome P450 (37). It is noteworthy that the wavelength maximum of the dioxygen complex of a heme-mercaptide P-450 model compound exhibits a strong dependence on the polarity of the solvent, ranging from 422 nm in hydrophobic solvent to 430 nm in more polar solvent (38). In NOS, although the heme pocket is mostly hydrophobic, several polar amino acids are close to the substrate binding site, including tyrosine 367, glutamate 371, and the carboxyl group of tryptophan 366, and may increase the polarity of the environment around the oxygen binding site (16 -18). On the other hand, the longer wavelength for the Soret transition of Fe 2ϩ (O 2 ) nNOS may be caused by proximal effects resulting from the additional hydrogen bonds to the heme cysteine ligand (35,36).
The optical spectrum of the dioxygen complex of nNOS and nNOSoxy, obtained by the rapid mixing of the ferrous enzyme with oxygen, differs from that reported for the dioxygen complexes of nNOS (25) and nNOSoxy (29) obtained at low temperature after the addition of oxygen to the ferrous enzymes. In the latter studies, performed at Ϫ30°C, the wavelength maximum of the Soret band was observed at 416 and 419 nm, respectively. For cytochrome P450 and chloroperoxidase, the optical spectrum of the dioxygen complex does not appear to be affected by temperature (Table II). Experiments are currently under way to record low temperature resonance Raman spectra of Fe 2ϩ O 2 nNOSoxy to determine if the intermediate obtained at low temperature by other groups is the dioxygen complex nNOS and, if so, to determine what structural changes can explain the different optical spectra. The intermediates obtained by the two methods differ in another way. At low temperature, the oxygenated complex of the oxygenase domain could be observed only in the presence of substrate and biopterin (29), presumably because in their absence, the complex is not stable and the initial ferrous form of the enzyme decayed to the ferric form without forming a detectable intermediate. Here, the ferrous dioxygen complex of nNOSoxy is formed in the absence or presence of substrate and H 4 B (see below) as was reported previously (28). with the bound oxygen. However, we cannot exclude the possibility that at longer reaction times there may be differences.

Substrate and Cofactor Sensitivity of the O-O Line of Fe
It is somewhat surprising that the binding of substrates and H 4 B have no detectable effect on the position of the O-O line. In CO-nNOS, it was shown that the heme pocket undergoes a conformational change upon binding of L-arginine that results in a shift of the Fe-CO line from 487/501 to 503 cm Ϫ1 , a shift of the ␦ Fe-CO line from 562 to 566 cm Ϫ1 , and a shift of the C-O line from 1930/1949 to 1929 cm Ϫ1 (21,23). In the L-arginine-bound conformation, it was concluded that the heme pocket adopts a closed structure, where the substrate exerts a strong polar and/or steric effect on the heme-bound CO molecule. Because of the apparent absence of interaction between the substrate and the heme-bound dioxygen molecule in the ferrous dioxygen complex, we have examined the interaction of substrates with CO in CO-nNOSoxy. In the absence of substrate and H 4 B, a broad Fe-CO line was identified at 491 cm Ϫ1 , and the ␦ Fe-CO line was identified at 562 cm Ϫ1 (Fig. 4, c-e). With L-arginine-and L-arginine/H 4 B-bound enzyme, the Fe-CO and the ␦ Fe-CO line were detected at 502 cm Ϫ1 and 565 cm Ϫ1 , respectively (Fig. 4,     a and b). These results indicate that the heme pocket structure of the oxygenase domain is similar to that of the full-length enzyme. Therefore, the observations of interactions between the substrates and the heme-bound CO are in contrast to the case of the oxygenated heme of nNOSoxy, in which the substrate does not display any strong steric/polar interactions with the dioxygen molecule.
Several hypotheses could explain these observations. First, the vibrational modes of the oxygen molecule may not be very responsive to steric/polar interactions involving nearby residues. In myoglobin, it was shown that the resonance Raman lines associated with the Fe-O 2 linkage are not as sensitive as those of the Fe-CO linkage to the mutation of amino acid residues of the heme pocket. This was suggested to occur because the dioxygen molecule is highly polar and exists in only one resonance structure, while the CO molecule is apolar and exists in two resonance structures that involve net changes in bond order and alteration in the charge of the oxygen atom (42). A second explanation could be that the dioxygen molecule of Fe 2ϩ (O 2 ) nNOSoxy adopts a different orientation in the heme pocket as compared with that of CO. Typically, the heme-bound CO molecule adopts a nearly linear conformation (39 -41), while in oxy complexes, the dioxygen molecule is bent with an angle of ϳ120 o (42). If this is the case in nNOSoxy, then the bending of the dioxygen molecule away from L-arginine could explain the absence of interaction of the dioxygen molecule with substrate. Third, the structure of the heme pocket of the Fe 2ϩ (O 2 ) nNOSoxy complex could differ in such a way that the L-arginine molecule is located at a greater distance from the heme-iron as compared with its position in the CO complex, thus decreasing polar/steric interactions between the substrate and the dioxygen molecule. Last, we cannot exclude the possibility that in substrate-free Fe 2ϩ (O 2 ) nNOSoxy, the dioxygen molecule is involved in steric/polar interactions with a heme pocket residue. This proposition could account for the fact that in substrate-free CO nNOS (21) and CO nNOSoxy (Fig. 4c), approximately half of the enzyme adopts the closed conformation with the Fe-CO line at 502 cm Ϫ1 . The cofactor is not responsible for the fraction of enzyme that adopts the closed conformation, since this splitting is observed in the absence of biopterin (Fig. 4c). Therefore, in the closed conformations of substrate-free CO nNOSoxy, it is likely that an amino acid residue or water molecule(s) interact with the CO. A similar situation could occur with oxygen. If this is the case, then the involvement of an amino acid residue or water molecule with the dioxygen molecule in substrate-free nNOSoxy could mask the effect of L-arginine in the Fe 2ϩ (O 2 ) complex of substrate-bound enzyme.
Conclusions-The measurements that we have reported here unequivocally demonstrate the formation of the ferrous dioxygen complex of the oxygenase domain of nNOS shortly after rapid mixing oxygen with the ferrous enzyme. The optical spectrum of the ferrous dioxygen species of nNOSoxy shows that the Soret band is located at 430 nm, in good agreement with previous experiments performed with a stopped flow apparatus at 10°C, but different from the values of the Soret band reported for oxygenated complexes obtained at low temperature. Further experiments are needed to characterize the structure of the intermediate detected at Ϫ30°C in which the Soret transition is located at 416 -419 nm. Finally, the observation of the oxygenoxygen stretching mode in the oxy complex lays the foundation for identifying other oxygen modes such as those associated with the activated oxygen intermediates and thereby unraveling the catalytic mechanism of this fascinating enzyme.