Catalytic Intermediates of Inducible Nitric-oxide Synthase Stabilized by the W188H Mutation*

Background: Trp-188 plays a role in regulating the activity of nitric-oxide synthase (NOS). Results: W188H mutation stabilizes a 420-nm intermediate by distorting the heme macrocycle. Conclusion: The 420-nm intermediate is a hydroxide-bound ferric heme species with a tetrahydrobiopterin radical center. Significance: The data provide the first evidence for a critical intermediate in NOS. Nitric-oxide synthase (NOS) catalyzes nitric oxide (NO) synthesis via a two-step process: l-arginine (l-Arg) → N-hydroxy-l-arginine → citrulline + NO. In the active site the heme is coordinated by a thiolate ligand, which accepts a H-bond from a nearby tryptophan residue, Trp-188. Mutation of Trp-188 to histidine in murine inducible NOS was shown to retard NO synthesis and allow for transient accumulation of a new intermediate with a Soret maximum at 420 nm during the l-Arg hydroxylation reaction (Tejero, J., Biswas, A., Wang, Z. Q., Page, R. C., Haque, M. M., Hemann, C., Zweier, J. L., Misra, S., and Stuehr, D. J. (2008) J. Biol. Chem. 283, 33498–33507). However, crystallographic data showed that the mutation did not perturb the overall structure of the enzyme. To understand how the proximal mutation affects the oxygen chemistry, we carried out biophysical studies of the W188H mutant. Our stopped-flow data showed that the 420-nm intermediate was not only populated during the l-Arg reaction but also during the N-hydroxy-l-arginine reaction. Spectroscopic data and structural analysis demonstrated that the 420-nm intermediate is a hydroxide-bound ferric heme species that is stabilized by an out-of-plane distortion of the heme macrocycle and a cation radical centered on the tetrahydrobiopterin cofactor. The current data add important new insights into the previously proposed catalytic mechanism of NOS (Li, D., Kabir, M., Stuehr, D. J., Rousseau, D. L., and Yeh, S. R. (2007) J. Am. Chem. Soc. 129, 6943–6951).

Nitric-oxide synthase (NOS) is a heme-containing flavoenzyme that synthesizes nitric oxide (NO) from L-arginine (L-Arg) in a two-step process (Scheme 1). In the first step of the reaction, one molecule of O 2 and two electrons from NADPH are consumed for the conversion of L-Arg to N-hydroxy-L-arginine (NOHA). 2 In the second step of the reaction, another molecule of O 2 and an additional electron from NADPH are used to convert NOHA to L-citrulline and NO. Previous studies suggest that the two steps of the reaction follow distinct mechanisms meditated by a compound I (Cmpd I) type of ferryl intermediate and a peroxyl intermediate, respectively (1)(2)(3)(4)(5)(6)(7). These mechanisms, however, remain elusive, as none of the putative intermediates have been experimentally observed under solution conditions, although (hydro)peroxo intermediates have been identified at cryogenic temperatures by radiolytic reduction methods (8,9); in addition, a Cmpd I intermediate has been observed after peroxyacid treatment (10).
Three isoforms of NOS have been identified in mammals: neuronal NOS, endothelial NOS, and inducible NOS (iNOS). Similar to the P450 class of enzymes, the heme prosthetic group in all three isoforms of NOS is coordinated by a thiolate sidechain group of an intrinsic cysteine residue in the proximal heme pocket. In P450s, the thiolate ligand forms a H-bond with a peptide NH group (11), whereas in NOSs the analogous thiolate ligand accepts a H-bond from the side chain of a conserved tryptophan residue (Trp-188 in iNOS). It is believed that the H-bonding interaction with the tryptophan residue reduces the electron donating capability of the thiolate ligand in NOSs, thereby modulating the oxygen chemistry occurring in the distal heme pocket of the enzymes (1,(12)(13)(14)(15). The mutation of the conserved tryptophan (Trp-409) in neuronal NOS to Phe or Tyr was shown to increase the rate of NO synthesis during multiple turnover conditions by decreasing the heme reduction rate and the degree of NO autoinhibition (15,16). Comparable mutants of iNOS, W188F, and W188Y, could not be overexpressed as stable recombinant forms (17); however, the W188H mutant was successfully expressed, purified, and studied (18).
It was shown that the W188H mutation slowed down the L-Arg hydroxylation reaction by stabilizing a new intermediate with a Soret maximum at 420 nm, which had never been observed during the wild type reaction, and that the formation of the 420-nm intermediate coincides with the disappearance of the ternary complex of the enzyme and the formation of a H 4 B radical, whereas its decay was concurrent with the recovery of the resting ferric enzyme. Tejero et al. (18) postulated that the 420-nm species is a catalytically competent oxygencontaining intermediate, such as a Cmpd I type of ferryl species. Regardless of the identity of the intermediate, the data demonstrated that the mutation modulates the structural properties and biochemical reactivity of the enzyme. However, the crystallographic data of the W188H mutant of the oxygenase domain of iNOS (iNOS oxy ) revealed that its active site structure is strikingly similar to that of the wild type enzyme (18). In particular, the side chain of His-188, like that of Trp-188 in the wild type enzyme, formed a H-bond with the thiolate ligand of the heme.
To determine how the W188H mutation modulates the oxygen chemistry of iNOS oxy without significantly perturbing the active site structure of the enzyme, we carried out a series of studies of the W188H mutant with optical absorption, resonance Raman, and EPR spectroscopic methods under steadystate and single turnover conditions. We discovered that the mutation introduced a unique out-of-plane distortion to the heme macrocycle that stabilizes the 420-nm intermediate populated during both the L-Arg and NOHA reactions and at the same time destabilizes the NO bound to the ferric heme during the NOHA reaction. The results are summarized and discussed in the context of the previously postulated NOS mechanism (1).

EXPERIMENTAL PROCEDURES
NOHA was purchased from Cayman Chemical (Ann Arbor, MI). H 4 B was purchased from Schircks Laboratories (Jona, Switzerland). Naturally abundant gases (N 2 and O 2 ) and liquid N 2 /He were purchased from Tech Air (White Plains, NY). Isotopically labeled gases ( 18 O 2 and 13 C 18 O) were purchased from Icon (Summit, NJ). EPPS, arginine, citrulline, dithiothreitol (DTT), sodium dithionite, and all other chemicals were purchased from Sigma. All chemicals were used without further purification and prepared with deionized water (Millipore, Billerica, MA).
All measurements were carried out with the oxygenase domain (residues 65-498) of murine iNOS (iNOS oxy ), which was prepared as described earlier (19). Briefly, the iNOS oxy gene with a C terminus six-histidine tag was cloned into a pCWori vector and overexpressed in a BL-21 strain of Escherichia coli. The cells were lysed using a microfluidizer (Microfluidics Corp., Newton, MA) and purified in the absence of cofactor and substrate by a Qiagen nickel-nitrilotriacetic acid affinity column (Germantown, MD). The purified enzyme was stored at 77 K in 40 mM EPPS buffer (pH 7.6) in the presence of 1 mM DTT. All purification steps were carried out at 4°C, whereas the experimental measurements were done at 20 Ϯ 1°C unless otherwise indicated.
Optical Absorption Measurements-The optical absorption measurements were carried out on a UV2100 instrument from Shimadzu Scientific Instruments, Inc. (Columbia, MD) with a spectral slit width of 1 nm. The path length of the optical cuvette was 2 mm. The protein concentrations were ϳ5-10 M. The substrate (L-Arg or NOHA) and cofactor (H 4 B) concentrations were at least 500-fold larger than the protein concentration unless otherwise indicated.
Stopped-flow Measurements-The stopped-flow measurements were carried out with a PiStar-180 system equipped with a photodiode array detector from Applied Photophysics Inc. (Leatherhead, UK). The enzyme, substrate, and cofactor were premixed and purged with N 2 for ϳ4 h in an airtight cuvette and subsequently titrated with a minimum amount of deoxygenated sodium dithionite. The complete binding of substrate/ cofactor and heme reduction was confirmed by optical absorption measurements. The reactions were initiated by 1:1 mixing of the ferrous enzyme (final concentration ϳ5 M) with airsaturated buffer. The data were analyzed with ProK software from Applied Photophysics Inc.
Resonance Raman Measurements-The equilibrium resonance Raman measurements were acquired with the system described previously (20). Briefly, the 413.1-nm output from a krypton ion laser (Spectra Physics, Mountain View, CA) was focused to an ϳ30-m spot on a quartz cuvette rotating at ϳ1000 rpm. The scattered light was collected at a right angle to the incident beam and focused on a 100-m entrance slit of a 1.25-m Spex spectrometer equipped with a 1200 grooves/mm grating (Horiba Jobin Yvon, Edison, NJ), where it was dispersed and detected by a liquid nitrogen-cooled charge-coupled device (CCD) detector (Princeton Instruments, Trenton, NJ). The incident laser light was removed by a holographic notch filter (Kaiser, Ann Arbor, MI). The Raman shifts were calibrated by using indene. Cosmic ray artifacts were removed using the Winspec software from Roper Scientific (Princeton, NJ). The laser power was Ͻ5 milliwatts at the sample for all measurements. The total integration time for each spectrum was 30 min. For the measurements of the CO related modes, the 441.6-nm laser output from a helium-cadmium laser (Kimmon, Centennial, CO) was used. The acquisition time for each spectrum was 30 min and 2 h for the measurements of the Fe-CO stretching mode ( Fe-CO ) and C-O stretching mode ( C-O ), respectively. For the measurements of the iron-cysteine stretching mode ( Fe-Cys ), the 363.8 nm output from an argon laser (Spectra Physics, Mountain View, CA) was used. The acquisition time for these measurements was 1 h.
The continuous-flow resonance Raman measurements were carried out with a homemade continuous-flow apparatus as described elsewhere (21). The enzyme samples (80 M) with L-Arg and/or H 4 B were purged with N 2 gas for several hours before reduction with ϳ2-fold excess of dithionite. The reduced enzyme and oxygen-saturated buffer (1.3 mM) were transferred to the homemade continuous-flow apparatus with gas-tight syringes (Hamilton, Reno, NV). The two solutions were mixed and flowed through a quartz cuvette (250-m path length) at a rate of 3.3 ml/min. The progression of the reaction SCHEME 1

Catalytic Intermediates in the W188H Mutant of iNOS
was probed by a laser beam at a desired time point. The scattered Raman light was collected perpendicular to the probe beam and measured as described above for the equilibrium measurements.
Rapid Freeze-Quench EPR Measurements-The EPR samples were prepared with a homemade rapid freeze-quench apparatus as described elsewhere (22). The dithionite-reduced enzyme (150 M) with 10 mM L-Arg and 2 mM H 4 B was rapidly mixed with oxygen-saturated buffer at room temperature. The mixed solution jet was directed onto a liquid N 2 -cooled copper wheel, where it was instantly frozen and ground into a fine powder. The freeze-quenched samples were measured in a liquid N 2 -cooled finger Dewar with X-band EPR on a Varian E-line spectrometer at 77 K. The modulation amplitude, microwave power, and receiver gain was 1.6 G, 1 milliwatt, and 1.25 ϫ 10 4 , respectively. The microwave frequency was 9.105 GHz.
Variable temperature X-band (ϳ9 GHz) EPR spectra were measured with a Bruker ESP 300 spectrometer equipped with an ESR 10 continuous-flow cryostat (Oxford Instruments). The temperature was regulated by a 3120 temperature controller (Oxford Instruments). The temperature readout of the controller was calibrated using a Lakeshore Cernox Resistor immersed in glycerol in a 4-mm EPR tube that was placed in the EPR cavity at the position of the sample. Temperature uncertainty was estimated to be Ϯ 1 K. D-band (130 GHz) EPR (two pulse echodetected) measurements were carried out on a spectrometer described elsewhere (23,24) at 7 K with 50-ns 90°pulses and 150 ns between pulses at a repetition rate of 30 Hz (30 averages per point). For both the X-and D-band measurements, the field was calibrated by Mn 2ϩ -doped MgO. Simulation of the X-and D-band spectra was performed using scripts written with MATLAB.

Optical Absorption and Raman Spectra of the Ferric
Enzyme-To investigate how the W188H mutation affects the structural properties of the enzyme, we first characterize the ferric derivatives of the W188H mutant in the absence or presence of substrate (L-Arg or NOHA) and/or cofactor (H 4 B) with respect to the wild type (WT) enzyme with optical absorption spectroscopy. It is noted that a small amount of DTT (100 M) was added to the H 4 B-containing samples to maintain H 4 B in the reduced state; the same amount of DTT was added to all other samples for consistency.
As reported previously (25), the wild type enzyme exhibits a Soret maximum at ϳ420 nm, indicating a six-coordinate low spin (6CLS) heme with a water bound to the ferric heme iron (Fig. 1a). The binding of substrate (L-Arg or NOHA) and/or cofactor (H 4 B) to the enzyme introduced structural changes to it that destabilized the water ligand, thereby leading to a fivecoordinate high spin (5CHS) heme, as indicated by the shift of the Soret maximum to ϳ395-400 nm. The broad Soret band and weak shoulder at 460 nm indicate that a small amount of the enzyme has DTT bound to the heme iron as the distal ligand. The presence of a shoulder at ϳ420 nm in the H 4 B-only sample indicates a minor contribution of the residual 6CLS water-bound ferric heme.
As in the wild type enzyme, the distal side of the ferric heme in the substrate and cofactor-free (SF) mutant enzyme was occupied by a water with a 6CLS configuration, as indicated by the Soret maximum at 417 nm (Fig. 1b); the binding of both substrate and cofactor (H 4 B ϩ L-Arg or H 4 B ϩ NOHA) or L-Arg alone destabilized the water ligand, leading to a 5CHS heme, as indicated by the shift of the Soret maximum to 395 nm. In contrast, the binding of either NOHA or H 4 B alone only slightly shifted the Soret maximum to 421 and 416 nm, respectively, indicating that the majority of the enzyme remained in the 6CLS state. Taken together the data revealed that the W188H mutation preferentially stabilizes the 6CLS configuration of the heme in the NOHA-alone or H 4 B alone state without significantly perturbing the configuration of the enzyme in other substrate-and/or cofactor-bound states. Fig. 2b shows the resonance Raman spectra of the mutant in the 1100 -1700 cm Ϫ1 window. The spectrum contains vibrational modes sensitive to the oxidation state ( 4 ) as well as the coordination and spin state ( 3 and 2 ) of the heme iron. Consistent with the optical absorption data, the SF mutant enzyme exhibited a pure 6CLS configuration, as indicated by the 2 / 3 modes at 1574/1502 cm Ϫ1 , whereas the addition of both substrate and cofactor (H 4 B ϩ L-Arg or H 4 B ϩ NOHA) led to a pure 5CHS configuration, as indicated by the 2 / 3 modes at 1561/1488 cm Ϫ1 . On the other hand, the addition of either NOHA or H 4 B alone led to a mixed 6CLS/5CHS configuration, with the 6CLS component dominating the spectrum, whereas the L-Arg alone enzyme exhibited a mixed 6CLS/5CHS configuration but with the 5CHS component dominating the spectrum. The presence of the minor contribution of the 6CLS component that is invisible in the optical absorption spectrum (Fig. 1b) was possibly a result of the fact that, with 413.1-nm laser excitation, the Raman lines associated with the 6CLS species were more resonantly enhanced with respect to those of the 5CHS species. Nonetheless, it is evident that the addition of H 4 B to the L-Arg or NOHA-alone enzyme enhanced the 5CHS component. Fig. 2a shows the resonance Raman spectra of the W188H mutant in the 200 -800 cm Ϫ1 window. In contrast to the high frequency spectra, the low frequency spectra are sensitive to the structure of the porphyrin macrocycle. Previous studies reported by Li et al. (26) demonstrate that the heme prosthetic group in the wild type enzyme exhibits an unusually high degree of out-of-plane deformation that is sensitive to substrate and cofactor binding. Likewise, the current studies revealed that substrate and/or cofactor binding to the mutant significantly perturbed the out-of-plane distortion of the heme and the peripheral groups attached to it (including the propionate and vinyl groups). It also led to the enhancement of several modes, including the out-of-plane modes, ␥ 15 (692 cm Ϫ1 ) and ␥ 11 (713 cm Ϫ1 ), the pyrrole breathing mode, 15 (752 cm Ϫ1 ) (27,28), the 8 mode (346 cm Ϫ1 ), the propionate bending mode (376 cm Ϫ1 ), and the vinyl bending mode (407 cm Ϫ1 ) as well as the perturbation to the line shapes and positions of the ␥ 12 (ϳ497 cm Ϫ1 ), 5 (1128 cm Ϫ1 ), and the methylene twist modes (1226 cm Ϫ1 ).
The Fe-S Bond Strength-In cytochrome P450s, the thiolate ligand of the heme accepts a H-bond from a peptide amide, whereas the analogous thiolate ligand in NOSs accepts a H-bond from the side-chain group of a Trp residue (Trp-188 in iNOS). The unique H-bond between the thiolate and Trp in NOSs leads to a weaker Fe-S bond that is believed to be critical for tuning the chemical reactivity of the enzymes, such that NO synthesis can be efficiently catalyzed by the various isoforms for diverse physiological demands.
To determine how the W188H mutation affects the H-bond thereby modulating the Fe-S bond strength, we measured the frequency of the Fe-S stretching mode ( Fe-S ) of the mutant with respect to the wild type enzyme. As the Fe-S mode is only observable in the 5CHS ferric state (29), we examined the L-Arg Ϯ H 4 B-and NOHA ϩ H 4 B-bound mutant versus the wild type enzyme with excitation at 363.8 nm. As shown in Fig. 3, a and b, in the L-Arg ϩ H 4 B-bound wild type enzyme, the Fe-S mode was detected at 337.0 cm Ϫ1 , consistent with that reported previously (30). The same Fe-S frequency was observed with L-Arg alone (data not shown). The Fe-S frequency of NOHA ϩ H 4 B-bound enzyme was slightly up-shifted to 339.2 cm Ϫ1 . The W188H mutation led to ϳ1-cm Ϫ1 higher frequencies for the analogous L-Arg ϩ H 4 B and NOHA ϩ H 4 B derivatives (338.1 and 339.8 cm Ϫ1 , respectively), indicating a weaker H-bond between the residue 188 and the thiolate ligand.
The observed Fe-S frequencies of iNOS oxy are significantly lower than those of P450s, which are ϳ351 cm Ϫ1 (31), in good agreement with the presence of a strong H-bond between residue 188 and the thiolate ligand of the heme. In the crystal structure of the W188H mutant (PDB code 3DWJ), the side chain of His-188, similar to Trp-188 in the wild type enzyme, forms a H-bond with the thiolate ligand. Although the H-bond distance is similar to that in the wild type enzyme, Tejero et al. (18) proposed that the H-bond between His-188 and the thiolate ligand in the mutant is stronger, as the redox potential of the mutant is 88 mV higher than that of the wild type enzyme. In contrast to this hypothesis, our data show that the Fe-S frequencies of the W188H mutant were ϳ1 cm Ϫ1 higher for the mutant as compared with the wild type enzyme, indicating only a slightly weaker H-bond in the mutant. The absence of a large frequency difference in the Fe-S modes of the mutant versus wild type enzymes is consistent with the preservation of the H-bond between residue 188 and the thiolate ligand of the heme with a strength similar to that in the WT enzymes.
CO-related Vibrational Modes-In addition to direct measurement of the Fe-S frequency, the proximal Fe-S bond strength could be estimated by the offset of the Fe-CO -CO inverse correlation line shown in Fig. 3c. CO has been demonstrated to be an informative structural probe for the distal ligand environment in hemeproteins, as the electron density distribution on the Fe-C-O moiety is sensitive to the electro- static potential of the protein environment surrounding the heme iron bound CO due to the back-bonding effect (32). When the electrostatic potential is positive, the Fe-CO -CO data points fall on the left side of the correlation line; in contrast, when the positive electrostatic potential is reduced, the Fe-CO -CO data points fall on the right side of the correlation line (see Refs. 20 and 32 for a complete discussion). The electron density distribution on the Fe-C-O moiety is also sensitive to the electronic properties of the proximal heme ligand. Accordingly, the C-O -Fe-CO correlation line of hemeproteins with histidine as the proximal ligand (such as globins, see the L ϭ His line in Fig. 3c) lies higher than that of hemeproteins with cysteine as the proximal ligand (such as P450s) (32).
It has been shown that the mammalian NOSs (mNOS) correlation line is up-shifted from the P450 line due to its weaker Fe-S bond, as indicated by their lower Fe-S frequencies (ϳ340 versus 350 cm Ϫ1 ). Intriguingly, the data of some bacterial NOSs, such as those from Bacillus subtilis (30,33), lie on a separate line (designated as the bsNOS line in Fig. 3c) in between the mNOS and P450 lines, suggesting that the Fe-S bond strength is stronger in these bacterial NOSs as compared with mNOSs.
The CO-associated modes of the W188H mutant were determined by resonance Raman spectroscopy (data not shown) and are listed in Table 1. The assignments of all the modes were confirmed by CO isotopic substitution experiments. As shown in Fig. 3c, the C-O / Fe-CO data points of both the wild type and W188H mutant of iNOS oxy (indicated by black squares and red circles, respectively) are sensitive to the substrate and/or cofactor bound to the active site. In the SF wild type enzyme, two Fe-CO conformers were identified, with the major and minor population labeled as point #1 and 1Ј, respectively. H 4 B binding to the enzyme led to the merging of the two data points into one lying between them (point #2). L-Arg binding, on the other hand, shifted the data points along the correlation line to the upper left corner (Fig. 3c, points 5  It is noteworthy that similar studies of the W56H mutant of a bacterial NOS from Staphylococcus aureus, analogous to the W188H mutant of iNOS oxy , were reported recently (34). However, it was found that the W56H mutation only slightly perturbed the Fe-CO and C-O frequencies in all substrateand/or cofactor-bound states examined, indicating the structural properties of S. aureus NOS are distinct from those of mNOS.
The 420-nm Intermediate-To understand how the W188H mutation modulates the oxygen chemistry catalyzed by iNOS oxy , the L-Arg and NOHA reactions of the W188H mutant were carried out in the presence of H 4 B under single turnover conditions. Comparable reactions of the wild type enzyme were examined as a reference. As shown in the inset in Fig. 4a, the L-Arg reaction of the wild type enzyme resulted in rapid formation of the O 2 complex, with a Soret maximum at 428 nm. The O 2 complex subsequently converted to the ferric 5CHS species with a Soret maximum at 395 nm at a rate of 13 s Ϫ1 without populating any intermediates (Equation 1), as indicated by the presence of a clear isosbestic point.
The L-Arg reaction carried out by the W188H mutant also led to the rapid production of the O 2 complex, with a Soret maximum at 430 nm, which subsequently converted to the ferric 5CHS species with a Soret maximum at 394 nm (see the inset in Fig. 4c). However, no isosbestic point was observed during the conversion, indicating the presence of an additional intermediate, as described in Equation 2. Global analysis revealed that the The I 420 nm intermediate was maximally populated at ϳ500 ms and subsequently decayed to the ferric species at a rate of 0.33 s Ϫ1 (Fig. 4e), similar to the previously reported data (18). The same experiment carried out in the absence of DTT (the reductant used to prevent H 4 B oxidation) showed similar results, indicating that the presence of small amount of DTT did not interfere with the reaction kinetics.
In the NOHA reaction of the wild type enzyme, a similar O 2 binding and oxidation process was observed (Fig. 4b). Global analysis revealed transient population of the NO-bound ferric species with a Soret maximum at 440 nm due to the binding of the product NO to the ferric heme. The NO-bound ferric species ultimately converted to the 5CHS ferric species with a Soret maximum at 396 nm by releasing NO, as described in Equation 3.
In the NOHA reaction of the W188H mutant, the O 2 complex with a Soret maximum at 428 nm was observed within the dead time of the instrument, which ultimately converted to the ferric 5CHS species with a Soret maximum at 396 nm (Fig. 4d). Surprisingly, global analysis revealed that the NO-bound ferric intermediate observed in the wild type reaction was not populated in the mutant reaction. Instead, a transient intermediate with a Soret maximum at 420 nm, similar to that observed in the L-Arg reaction, was maximally populated at ϳ185 ms and decayed to the ferric species at a rate of 0.17 s Ϫ1 (Fig. 4f) The Identity of the 420-nm Intermediate-To examine if I 420 nm is an oxygen containing intermediate, we carried out continuous-flow resonance Raman measurements of the intermediate populated during the L-Arg reaction. On the basis of the kinetic profile shown in Fig. 4e, 3 time points, 14, 100, and 500 ms (corresponding to a population of 0, 50, and 100% intermediate, respectively), were examined. To identify oxygen-containing vibrational modes, the 16 O 2 -18 O 2 isotope difference spectra were calculated. As shown in Fig. 5 On the basis of the previously proposed mechanism (1), other than the primary O 2 -complex, two potential oxygen-containing intermediates, a ferric peroxo species (Fe 3ϩ -O 2 2Ϫ ) and a Cmpd I type of ferryl species (Fe 4ϩ ϭ O 2Ϫ ) might be populated during the L-Arg and NOHA reactions. Our data showed that at all three time points, no signal was detected in the ϳ700 -950cm Ϫ1 region of the 16 O 2 -18 O 2 difference spectra, where the characteristic oxygen-related modes of typical peroxo and fer-  ryl species are located. It is important to note that although our data show no evidence supporting the assignment of I 420 nm to either the peroxo or ferryl species, we could not rule out the possibility that these modes were too weak to be detected with the 413.1-nm excitation. The Soret maximum of the peroxo derivatives of Geobacillus stearothermophilus NOS (9) and P450 (36) have been identified at ϳ440 nm by cryoreduction methods. It is inconsistent with the Soret maximum of I 420 nm , ruling out the assignment of I 420 nm to a peroxo species. To examine if I 420 nm is associated with any minute peroxo species, which was too weak to be detected by absorption measurements, we repeated the continuous flow resonance Raman measurements with 441.6-nm excitation (to resonantly enhance the modes associated with the potential peroxo species). Again, we observed no evidence for peroxo species, indicating that the peroxo species is not populated during the mutant reaction.
The absorption spectrum of the Cmpd I ferryl derivative of NOS has not been reported. We sought to generate the Cmpd I derivative of iNOS oxy by reacting the ferric enzyme with m-chloroperbenzoic acid, which had been widely used to generate ferryl derivatives of hemeproteins (37,38). We found that immediately after the initiation of the reaction, a species with a Soret maximum at 421 nm, which we assign to the m-chloroperbenzoic acid-bound ferric species, was observed (data not shown). The spectral properties of the Cmpd I derivative of iNOS oxy are consistent with that of the Cmpd I derivative of P450 identified in the m-chloroperbenzoic acid reaction with the ferric enzyme (37) and in the photo-oxidation reaction of a Cmpd II intermediate (39). It is also similar to the Cmpd I derivative of chloroperoxidase derived from the H 2 O 2 reaction (40). These data support the scenario that I 420 nm is not a Cmpd I type of ferryl species.
To further confirm that I 420 nm is not a ferryl species, we examined its lifetime in the presence and absence of ascorbate, which is a good reductant for Cmpd I or Cmpd II types of ferryl species (41), by carrying out sequential mixing experiments in a stopped-flow system. In these experiments the substrate-and cofactor-bound ferrous mutant was first mixed with O 2 -containing buffer in the first mixer. The reaction mixture was allowed to age until I 420 nm was maximally populated before it was mixed with a buffer in the presence or absence of ascorbate in the second mixer. The population of I 420 nm was then followed as a function of time by monitoring the absorption increase at 395 nm. The data show that the presence of ascorbate did not affect the decay rate of I 420 nm (Fig. 6). Comparable experiments carried out with ferrocyanide (instead of ascorbate), which is also a potential reductant for ferryl species, showed similar results (data not shown), confirming that I 420 nm is not a ferryl species.
The 420-nm Intermediate Is Associated with an H 4 B .
ϩ Radical-It has been proposed that during the NOS reaction, the H 4 B cofactor acts as an electron donor to the heme, yielding a one electron oxidized cation radical (H 4 B . ϩ ) that is subsequently re-reduced to the neutral species (42)(43)(44)(45)(46). The EPR spectra of the H 4 B . ϩ derived from the wild type iNOS oxy has been reported by Stoll et al. (47). To examine if I 420 nm was associated with any radical species, such as H 4 B . ϩ , we used a home-built rapid freeze-quenching apparatus (22) to trap I 420 nm derived from the L-Arg reaction and characterized it with both conventional X-band (9.4 GHz) and high-field D-band (130 GHz) EPR spectroscopy.
Our data (Fig. 7, right panels) confirmed that I 420 nm populated during the L-Arg reaction is indeed associated with a radical as reported by Tejero et al. (18). The X-band spectrum exhibited hyperfine structure that is consistent with the well characterized biopterin radical (42,44,(47)(48)(49). The D-band spectrum could be simulated with g 1 ϭ 2.0043, g 2 ϭ 2.0031, and g 3 ϭ 2.0021 by using the H 4 B . ϩ parameters determined in the wild type reaction (47) as a starting point. The X-band spectrum could be nicely fitted with these g values determined by the D-band data, confirming the assignment of the radical to It is important to note that the g values we determined differ slightly from those reported by Stoll et al. (47) for the wild type enzyme, in which the g 2 value was 2.00353. It is unclear if the small difference in the g value reflects the structural difference in the H 4 B . ϩ radical itself or in the environment surrounding the radical. Nevertheless, our data demonstrated that the EPR spectra of I 420 nm could be accounted for by a single set of g values associated with H 4 B . ϩ , indicating that only a single radical is associated with the intermediate. In accord with the conclusions drawn from our optical absorption and resonance Raman data, the EPR data exclude the possibility that I 420 nm is a Cmpd I ferryl species, as the EPR spectra of the -cation radical associated with a Cmpd I ferryl are typically highly anisotropic and broad with rhombic symmetry (50,51). Timedependent EPR measurements along with comparable stopped-flow experiments demonstrate that the H 4 B . ϩ radical is kinetically coupled to I 420 nm during both steps of the mutant reaction (Fig. 7, a and b), confirming that the radical is specifically associated with the intermediate.
It has been shown that during the L-Arg reaction of the wild type iNOS oxy domain, the re-reduction of the H 4 B .
ϩ radical to the neutral species occurs slowly via random processes because of the lack of the reductase domain; hence, the radical can be significantly populated and is readily detectable (42)(43)(44)(45)(46). In contrast, the H 4 B . ϩ radical associated with the NOHA reaction cannot be populated to an appreciable level, because it can be rapidly re-reduced by the immediate product of the reaction, HNO or NO Ϫ (leading to the generation of the final product NO ⅐ ); consequently, the radical decays at least 10 times faster than that observed during the L-Arg reaction (42,44,46). The surprisingly high level of the H 4 B . ϩ radical observed during the NOHA reaction of the W188H mutant (Fig. 7b) suggests that either the structural perturbation introduced by the mutation retarded electron transfer from HNO/NO Ϫ to H 4 B . ϩ or the enzyme is incapable of producing HNO/NO Ϫ . The latter scenario is excluded as the enzyme has been shown to be active in producing NO during multiple-turnover conditions (18).
The 420-nm Intermediate Is Associated with a Hydroxidebound Heme-The aforementioned spectroscopic data demonstrated that I 420 nm is not a peroxo or ferryl species, and it has a H 4 B . ϩ radical associated with it. To further investigate its identity, we surveyed the absorption spectra of all potential derivatives of the enzyme and found that the NOHA-bound 6CLS ferric species has an optical absorption spectrum similar to it (inset in Fig. 8). In addition, the resonance Raman spectrum of the NOHA-bound 6CLS ferric species strikingly resembled that of I 420 nm (Fig. 8). Accordingly, we propose that I 420 nm is a 6CLS species in which the heme iron is coordinated by a hydroxide ligand with an H 4 B . ϩ radical occupying the cofactor-binding site. A hydroxide ligand instead of a water ligand is proposed on the following basis. (a) Our earlier studies (1) suggested the same intermediate for the NOHA reaction, although the intermediate has never been experimentally observed (b) Structural data suggest that an unique out-of-plane distortion of the heme was introduced by the W188H mutation, which specifically stabilized the hydroxide ligand (see below). (c) The negatively charged hydroxide ligand plausibly can be stabilized by the H 4 B .
ϩ cation radical associated with the intermediate. Intriguingly, we observed that high concentrations of H 4 B significantly shortened the lifetime of I 420 nm (Fig. 6)   L-Arg reaction under four different conditions, with L-Arg and H 4 B systematically varied. As shown in Fig. 9, a versus c, the increase of H 4 B from 5ϫ to 500ϫ of the enzyme concentration led to the disappearance of I 420 nm . Similar H 4 B dependence was observed independent of L-Arg (Fig. 9, b versus d), although when L-Arg was low, global analysis revealed a new ferric intermediate (ferricЈ) with a Soret maximum at 395 nm (Fig. 9d), as described in Equation 5. It is noted that the weak shoulder at ϳ420 nm in the spectrum of the ferricЈ intermediate was attributed to a minor population of I 420 nm unresolved by the deconvolution process.
Comparable H 4 B-dependent studies of the NOHA reaction showed that, like the L-Arg reaction, I 420 nm was detected only when the H 4 B concentration was low regardless of the NOHA concentration (data not shown). However, when the NOHA concentration was low, the ferricЈ intermediate was not detected; instead, the reaction followed a two-step mechanism, Systematic H 4 B-dependent studies of the L-Arg reaction (Fig.  9e) showed that the formation rate of the intermediate from the oxygen complex remained constant, ϳ5 s Ϫ1 , whereas its decay rate to the resting ferric state increased linearly with increasing H 4 B concentration. A linear fit of the data yielded k on ϭ 0.78 mM Ϫ1 s Ϫ1 and k off ϭ 0.33 s Ϫ1 , based on the slope and intercept of the best fitted line. The data confirm that the excess of H 4 B reduces or replaces the H 4 B . ϩ radical associated with I 420 nm , thereby facilitating its conversion to the resting ferric state.
The Structural Change Introduced by the W188H Mutation-How does the proximal W188H mutation stabilize the distal hydroxide ligand and the H 4 B .
ϩ associated with it during the NOS reaction? The structure of the W188H mutant has been resolved in the H 4 B bound state (without any substrate), whereas that of the wild type enzyme has been determined in various substrate and cofactor-bound states. In the H 4 B-bound wild type enzyme, Trp-188 donates a H-bond to the thiolate ligand of the heme via its side chain; in addition, its indole ring forms a -stacking interaction with the porphyrin macrocycle (Fig. 10b). In the H 4 B-bound mutant enzyme, the side chain of His-188 also forms a H-bond with the thiolate ligand; however, its imidazole ring is rotated out of the parallel position with respect to the porphyrin (Fig. 10a), which creates a steric clash with a pyrrole ring of the porphyrin, thereby introducing a unique out-of-plane twist to it.
The nonplanar deformation of the heme in the wild type iNOS oxy as well as its critical role in controlling the catalytic and NO autoinhibition properties of the enzyme, have been reported previously by Li et al. (1). To quantify how the W188H mutation affects the out-of-plane distortion of the heme, we employed the Normal Coordinate Structural Decomposition (NSD) analysis method (52) to evaluate the structure of the heme in the W188H mutant versus the wild type enzyme (PDB code 3DWJ and 2NOD, respectively). The data show that the mutation significantly enhances the out-of-plane distortion of the heme, with a composite out-of-plane displacement of 1.18 and 0.77 Å for the mutant and the wild type enzyme, respectively (Table 2). We found that in the mutant the largest enhancement was along the B 1u (ruffling) symmetry (0.88 Å),  whereas that in the wild type enzyme was along the B 2u (saddling) symmetry (0.58 Å).
It has been shown that in H-NOX (a heme nitric oxide/oxygenbinding protein), heme distortion significantly reduces the electron density at the heme iron, which leads to an increase in the redox potential of the heme and a decrease in the pK a of the distal water bound to it (53). We hypothesize that the enhanced heme distortion in the W188H mutant also reduces the electron density at the heme iron (evident by the 88-mV higher redox potential (18)), thereby stabilizing the hydroxide ligand in I 420 nm . The unique heme distortion introduced by the mutation might further stabilize I 420 nm by perturbing the interactions between the heme, substrate (L-Arg or NOHA), and cofactor (H 4 B), which are linked together via an extended H-bonding network mediated by a propionate group of the heme.

DISCUSSION
Our previous studies suggest that the NOHA reaction catalyzed by NOS goes through a hydroxide-bound ferric heme intermediate, with H 4 B . ϩ associated with it (1), but it is too short-lived to be experimentally observed. In this work we demonstrated that this intermediate is stabilized by the W188H mutation and that the intermediate exhibits a Soret maximum at 420 nm. We showed that the intermediate is populated not only during the NOHA reaction but also during the L-Arg reaction. On the basis of this new finding, we postulate a modified catalytic mechanism as illustrated in Fig. 11.
In the first step of the reaction, the ferric 5CHS heme is reduced to the ferrous form by accepting an electron from the reductase domain. It is followed by O 2 binding to generate the O 2 -complex (I), with a ferric superoxide electronic configuration. The ferric superoxide species is reduced to a ferric peroxy species (II) by accepting an electron from H 4 B, leaving a cation radical on H 4 B. The ferric peroxo species, with the terminal oxygen forming H-bonds with L-Arg and a water molecule next to it (1,54), accepts a proton, which triggers the heterolytic O-O bond cleavage, leading to a Cmpd I type of ferryl species, with a -cation radical on the porphyrin ring (III). At the same time, a water molecule is released from the active site. The ferryl oxygen of the Cmpd-I species is then inserted into L-Arg to generate a postulated O-bound product heme complex (IV). The O-bound product accepts a proton from a water molecule, which triggers its dissociation from the heme iron. Concurrently, the hydroxide from the water molecule binds to the ferric heme to generate the I 420 nm species. The subsequent oneelectron reduction of H 4 B . ϩ back to the neutral species (by the reductase domain in the full-length enzyme or by random pro-cesses in the oxygenase domain) promotes the uptake of a proton to the active site to maintain its charge neutrality (55) and converts the hydroxide ligand to water. The water subsequently dissociates from the heme iron, thereby regenerating the resting 5CHS species (V), with NOHA bound to it. It should be noted that the current data could not exclude the possibility that H 4 B . ϩ decays via oxidation to H 2 B ϩ 2H ϩ (44). In the wild type reaction, H 4 B . ϩ was transiently populated after the decay of the O 2 complex (I), but no heme intermediate was detected as the reaction was rate-limited by the electron transfer from H 4 B to the O 2 complex (1).
In the second step of the reaction the ferric 5CHS heme is reduced to its ferrous state by accepting an electron from the reductase domain. The ferrous enzyme then binds O 2 to form the ferric superoxide species (VI). The ferric superoxide species accepts an electron from H 4 B to produce the ferric peroxo species (VII) with H 4 B .
ϩ associated with it. The proximal oxygen atom of the heme iron-bound peroxide forms a H-bond with the substrate, NOHA, which induces the addition of the terminal oxygen to the substrate, leading to the ferric iron-bound alkyl peroxo species (VIII). The subsequent bond rearrangement produces HNO and citrulline as well as the hydroxidebound ferric heme intermediate, I 420 nm . HNO is oxidized to NO by transferring an electron to H 4 B . ϩ to regenerate its neutral form; at the same time the proton converts the hydroxide ligand to water, which subsequently dissociates from the iron to regenerate the resting 5CHS ferric heme species (IX).
In the wild type reaction, the product, NO, binds to the ferric heme iron, generating a transient NO-bound ferric intermediate (shown in brackets) before it diffuses out of the active site

Normal-coordinate structural decomposition analysis (52) of the heme in the subunit A of the H 4 B-bound W188H mutant and wild type of iNOS oxy (PDB codes 3DWJ and 2NOD, respectively)
The distortion of the heme was decomposed to six symmetry types. Dip and dip columns show the composite in-plane distortions and the mean deviations, respectively; Doop and doop columns show the composite out-of-plane distortions and mean deviations, respectively.  FIGURE 11. The two-step mechanism of NOS. The mechanism was modified from that reported by Li et al. (1). The NO-bound inhibitory complex observed in the second step of the wild type enzyme, but unresolved in the W188H mutant, is shown in the bracket in b.

W188H
into free solution. This NO-bound ferric intermediate, surprisingly, was not observed in the mutant reaction. On the contrary, I 420 nm observed in the mutant reaction was not observed in the wild type reaction. These data indicate that the structural changes to the enzyme introduced by the W188H mutation stabilize the hydroxide ligand bound to the ferric heme iron and the H 4 B . ϩ associated with it; at the same time they prevent transient binding of NO to the heme iron before it escapes out of the enzyme. The fact that the increase in H 4 B facilitates the decay of I 420 nm to the resting 5CHS ferric species (Fig. 9e) suggests that H 4 B . ϩ can be reduced not only by HNO but also by the excess H 4 B in free solution. The observation that I 420 nm is kinetically coupled to the formation and decay of H 4 B . ϩ demonstrates that the protonation of hydroxide to water and its subsequent dissociation was triggered by the reduction of H 4 B . ϩ bound in the vicinity of the heme; in other words the negatively charged hydroxide ligand in I 420 nm is stabilized by the cation radical located on H 4 B.