The FAD-shielding Residue Phe1395 Regulates Neuronal Nitric-oxide Synthase Catalysis by Controlling NADP+ Affinity and a Conformational Equilibrium within the Flavoprotein Domain*

Phe1395 stacks parallel to the FAD isoalloxazine ring in neuronal nitric-oxide synthase (nNOS) and is representative of conserved aromatic amino acids found in structurally related flavoproteins. This laboratory previously showed that Phe1395 was required to obtain the electron transfer properties and calmodulin (CaM) response normally observed in wild-type nNOS. Here we characterized the F1395S mutant of the nNOS flavoprotein domain (nNOSr) regarding its physical properties, NADP+ binding characteristics, flavin reduction kinetics, steady-state and pre-steady-state cytochrome c reduction kinetics, and ability to shield its FMN cofactor in response to CaM or NADP(H) binding. F1395S nNOSr bound NADP+ with 65% more of the nicotinamide ring in a productive conformation with FAD for hydride transfer and had an 8-fold slower rate of NADP+ dissociation. CaM stimulated the rates of NADPH-dependent flavin reduction in wild-type nNOSr but not in the F1395S mutant, which had flavin reduction kinetics similar to those of CaM-free wild-type nNOSr. CaM-free F1395S nNOSr lacked repression of cytochrome c reductase activity that is typically observed in nNOSr. The combined results from pre-steady-state and EPR experiments revealed that this was associated with a lesser degree of FMN shielding in the NADP+-bound state as compared with wild type. We conclude that Phe1395 regulates nNOSr catalysis in two ways. It facilitates NADP+ release to prevent this step from being rate-limiting, and it enables NADP(H) to properly regulate a conformational equilibrium involving the FMN subdomain that controls reactivity of the FMN cofactor in electron transfer.

a C-terminal reductase domain by a calmodulin (CaM)-binding sequence. The NOS oxygenase domain contains binding sites for iron protoporphyrin IX (heme), (6R)-5, 6, 7, 8-tetrahydro-Lbiopterin (H 4 B), and L-Arg and is the site where oxidative catalysis takes place. The NOS reductase domain (NOSr) contains binding sites for FMN, FAD, and NADPH and functions to transfer reducing equivalents from NADPH to the oxygenase domain.
NOSr belongs to a small family of structurally related dualflavin reductases that also includes cytochrome P450 reductase (CYPR) (7,8), methionine synthase reductase (9), and novel reductase-1 (10). These proteins are comprised of separate FMN and FAD/NADPH modules attached by a flexible hinge region (11,12). It is believed that these reductases are the product of gene fusion because their FMN and FAD/NADPH subdomains show a high similarity to flavodoxins (13) and ferredoxin NADP ϩ reductases (FNR) (14), respectively. In NOSr and related flavoproteins, the FAD receives electrons from NADPH via hydride transfer and then sequentially passes the electrons to the FMN cofactor. Ultimately, it is the 2-electron reduced FMN hydroquinone that transfers an electron to the ferric heme in the NOS oxygenase domain (15). This electron transfer allows the heme to bind dioxygen and initiate its reductive activation in conjunction with H 4 B as is required for NO synthesis (16). The FMN hydroquinone in NOSr can also reduce external electron acceptors such as cytochrome c (17), and this activity is widely used as a tool to study electron transfer by NOSr. The binding of CaM to NOS elicits responses in the enzyme that include activation of the ferric heme reduction (18,19) and increasing the rate of cytochrome c reduction. The rate of ferric heme reduction by NOSr is very important because it and the rates of ferric heme-NO dissociation and ferrous heme-NO oxidation are the three key kinetic parameters identified by this laboratory that define the characteristics of NO release by any given NOS (20,21).
The mechanism through which structural elements in NOSr regulate its electron transfer is a topic of current interest. The related flavoprotein CYPR has been extensively characterized with respect to its electron transfer mechanisms (22)(23)(24)(25)(26). Although these studies serve as a guide, it is important to appreciate that in NOSr there are additional complexities that do not exist in related flavoproteins. For example, rates of NOSr electron transfer reactions are repressed relative to most other related flavoproteins, but this repression is relieved upon CaM binding (27). Studies of NOSr and its unique regulatory elements typically utilize purified NOSr domains constructed to include their adjacent CaM-binding motifs so that the native electron transfer and CaM-response profiles of the proteins are retained (28 -33). Besides CaM binding, other components involved in controlling NOSr electron transfer include an auto-inhibitory insert within the FMN module (34 -37), a C-terminal extension (38 -40), multiple phosphorylation sites (41)(42)(43)(44)(45)(46), a loop within the connecting domain (47), and NADP ϩ binding (30). Although their mechanisms of action remain speculative, these elements are thought to act in concert to positively and negatively regulate electron transfer by NOSr.
This laboratory recently reported that Phe 1395 in rat neuronal NOS (nNOS) is also involved in regulation (48). Phe 1395 is representative of a conserved group of aromatic amino acids that shield the re face of the FAD isoalloxazine ring in NOS, related flavoproteins, and simpler FNR proteins (12,49,50). The aromatic side chains of these shielding residues must move away from the FAD isoalloxazine ring to allow a productive binding interaction with the nicotinamide ring of NADPH that is essential for hydride transfer to FAD (12). In addition to an influence on NADPH binding specificity, we found that the aromatic side chain of Phe 1395 was needed to repress electron transfer in CaM-free nNOS and to fully relieve the repression upon CaM binding (48). The unexpected significance of Phe 1395 prompted us to study our most interesting mutation, F1395S, in the isolated nNOSr in order to understand better how Phe 1395 helps regulate electron transfer and catalysis by nNOS.

EXPERIMENTAL PROCEDURES
General Methods and Materials-UV-visible spectra and steadystate kinetic data were obtained by using a Hitachi U-2000 or Cary 100 Bio-spectrophotometer. Single wavelength stopped-flow kinetic experiments were performed by using a Hi-Tech (Salisbury, UK) SF-51MX instrument equipped for anaerobic work and photomultiplier detection. Full spectra stopped-flow experiments were performed by using a Hi-Tech SF-61 instrument equipped for anaerobic work and rapid-scanning diode array detection. Data from multiple identical stopped-flow experiments were averaged to improve the signal-to-noise ratio and fit to exponential functions using software (IS-2) provided by the instrument manufacturer. The buffer used for all experiments and protein purifications (Buffer A), unless noted otherwise, contained 40 mM EPPS, pH 7.6, 10% glycerol, and 150 mM NaCl. Samples were made anaerobic when necessary by repeated cycling of vacuum followed by a positive pressure of catalyst-deoxygenated nitrogen in an airtight anaerobic cuvette or by the addition of glucose oxidase (10 units/ml), catalase (134 units/ml), and glucose (2-3 mM) to the sample. Larger volumes of buffer were made anaerobic by extended bubbling of catalyst-deoxygenated nitrogen through the liquid. Wild-type nNOSr was prepared for use by oxidizing the purified air-stable semiquinone form with potassium ferricyanide followed by passing the mixture through a PD-10 desalting column. F1395S nNOSr was prepared for use by briefly incubating with 1 mM 2Ј-AMP and passing the mixture through a PD-10 column, unless noted otherwise. The concentration of nNOSr samples was determined by using an extinction coefficient of 22,900 M Ϫ1 cm Ϫ1 at 457 nm for the fully oxidized form (29). Graphing of all data and curve fitting of data not obtained in stopped-flow experiments was done by using Origin software. Human CaM was expressed with the pACYC plasmid in Escherichia coli and purified with standard procedures utilizing phenyl-Sepharose chromatography. 5-Deazariboflavin, prepared using the method of Yoneda and co-workers (51,52), was obtained as a gift from Dr. John K. Hurley (University of Arizona). PD-10 desalting columns, 2Ј,5Ј-ADP-Sepharose, and CaM-Sepharose were obtained from Amersham Biosciences. NADPH was obtained from Alexis Biochemicals. All other reagents were obtained from Sigma and used without further purification.
Generation of Wild-type and F1395S nNOSr Proteins-The wild-type rat nNOSr construct (Met 695 to Ser 1429 ) was created by using a PCRbased method with primer 1 (AGATTCCATATGCTCAACTATAGACT-CACC) that inserted an NdeI site adjacent to the translation start site at Met 695 and primer 2 (CCCCCACGCAGAACACATCACAG). The PCR product fragment so generated and pCWori vector containing fulllength rat nNOS DNA were both digested with NdeI and SphI, and the resulting 256-bp PCR fragment was cloned into the double-digested pCWori/nNOS vector at these restriction sites.
For the F1395S mutant, rat nNOSr cDNA was inserted into the pCWori vector by using its 5Ј NdeI and 3Ј XbaI restriction sites. The F1395S substitution in the nNOSr cDNA was engineered by subcloning a PCR-generated fragment from pCWori/nNOSr with a 5Ј-oligonucleo-tide containing the desired mutation. This mutated nNOS cDNA fragment, coding from the KpnI unique restriction site at position 4170 to the XbaI restriction site at position 4441, was amplified using the following primers: F1395S forward primer, AACGGTACCACGAGGA-CATCTCTGGAGTCACCCTCAGAACG; F1395S reverse primer, AAA-TCTAGAAGGACCAGGACACAGCAACAGGACAAG. The restriction sites are indicated by underlines, and the mutation site is denoted with boldface. The PCR product and wild-type pCWori/nNOSr were digested by both KpnI and XbaI restriction enzymes. The double-digested fragment of wild-type pCWori/nNOSr was replaced with the double-digested PCR fragment to generate the recombinant plasmid. After DNA sequencing to confirm the desired results, E. coli BL21 (DE3) cells were transformed with the nNOSr plasmids and selected with ampicillin for protein expression. When applicable, the cells were also transformed with a pACYC plasmid containing human CaM and selected with chloramphenicol.
Protein Expression and Purification-The nNOSr proteins and CaM (when present) were expressed in E. coli. The cells were lysed and centrifuged, and the protein was precipitated from the supernatant with ammonium sulfate in the presence of CaCl 2 using a procedure described previously (53). The resulting protein pellets were resuspended in Buffer A containing phenylmethylsulfonyl fluoride (1 mM) and applied to a 2Ј,5Ј-ADP-Sepharose column pre-equilibrated with Buffer A plus ␤-mercaptoethanol (5 mM), FAD (2 M), FMN (2 M), and EDTA (0.5 mM) when necessary to remove the co-expressed CaM. The bound protein was washed extensively with the equilibration buffer and then eluted with the same buffer containing 10 mM NADPH. The eluted protein sample was immediately brought to a final concentration of 2 mM CaCl 2 and applied to a column of CaM-Sepharose pre-equilibrated with Buffer A plus 2 mM CaCl 2 . The bound protein was washed extensively with the equilibration buffer and then eluted with Buffer A containing 3 mM EGTA. The resulting protein sample was concentrated and stored frozen in aliquots at Ϫ80°C. The flavin content of the nNOSr proteins was obtained by boiling a known amount of nNOSr protein for 3 min followed by centrifugation and determination of the flavin concentration in the supernatant by using an extinction coefficient of 12.2 mM Ϫ1 cm Ϫ1 at 447 nm.
Steady-state Cytochrome c Reduction Assays-Cytochrome c reductase activities of the nNOSr proteins were determined by using an assay procedure as described previously (53) without the addition of superoxide dismutase or catalase.
Interaction between the nNOSr Proteins and NADP ϩ -UV-visible spectral changes resulting from the association of NADP ϩ with 20 M nNOSr proteins were observed by calculating a difference spectrum between the proteins in solution and the same protein sample after treatment with a 3-fold excess of NADP ϩ . Determination of NADP ϩ apparent dissociation constant (K d ) values was done by titrating the nNOSr proteins (20 M) with aliquots of a known concentration of NADP ϩ and recording the visible spectrum after a short (3-4 min) equilibration time at 20°C. Mathematical determination of the K d values was attempted by plotting ⌬(A 510 nm Ϫ A 700 nm ) versus NADP ϩ concentration for each titration point and fitting the data by using nonlinear regression analysis to the protein-ligand complex model developed by Wang et al. (54). Rate constant (k off ) values for the dissociation of NADP ϩ from the oxidized nNOSr proteins were determined by rapidly mixing a solution containing nNOSr (30 -40 M) and NADP ϩ (100 M) with a solution of 2Ј,5Ј-ADP (4 mM) in the stopped-flow instrument at 10 or 23°C and fitting the absorbance decreases at 510 nm to single exponential functions.
Anaerobic Equilibrium Photoreduction-The nNOSr protein samples with and without a saturating amount of NADP ϩ (based on Fig. 2) were diluted in Buffer A containing a final concentration of 3 mM EDTA and a catalytic amount of 5-deazariboflavin with care taken to shield the sample from ambient light. Each titration point was obtained by illuminating the sample with a commercial slide projector bulb for a fixed length of time (3-10 s) and then acquiring a visible spectrum after a 2-3-min equilibration. The titration was complete when further irradiation of the sample produced no additional absorbance changes. Reduction of F1395S nNOSr in the presence of NADP ϩ yielded a stable reduced nNOSr-NADP ϩ charge transfer complex. The extinction coefficient of this complex (1200 M Ϫ1 cm Ϫ1 at 700 nm) was estimated from a 25 M sample of F1395S nNOSr protein that gave ⌬A 700 ϭ 0.030 at the end of the titration (1-cm cuvette).
Anaerobic Stopped-flow Flavin Reduction Kinetics-The absorbance changes associated with nNOSr flavin reduction by NADPH were recorded by rapidly mixing a solution of oxidized nNOSr (6 -10 M) containing either EDTA (1 mM) or CaCl 2 (2 mM) ϩ CaM (18 -24 M) with a solution of 60 -100 M NADPH (excess NADPH) or a single molar equivalent of NADPH at 10°C in the stopped-flow instrument. The maximum absorbance value for a given protein sample at 457 nm during single wavelength experiments was obtained by replacing the NADPH solution in one of the stopped-flow syringes with buffer only and recording additional mixing events. The individual rate constants associated with absorbance changes at 457 nm were first estimated by analysis of experiments of varying lengths. The final reported values were obtained by fitting an experiment on a time scale capturing all four rate constants to a quadruple exponential function such that the residuals were minimized and contained little or no systematic deviation between the fit curve and the actual data. Percent absorbance changes were calculated based on the percent absorbance change in the instrument dead time and the relative proportions of the ⌬ A values for each kinetic phase obtained from the fitting program. Analysis of absorbance changes observed at 600 nm and correlation of these data with rapid-scanning stopped-flow diode array spectra were done as described (see "Results").
Anaerobic Pre-steady-state Cytochrome c Reduction-A solution of nNOSr (16 M), glycine (3 mM), 5-deazariboflavin (catalytic), and either EDTA (1 mM) or CaCl 2 (2 mM) ϩ CaM (30 M) was completely photoreduced in an anaerobic cuvette by using a commercial slide projector bulb until no changes in the UV-visible spectrum of the sample were observed upon further irradiation of the sample. The pre-reduced protein sample was rapidly mixed with a solution of cytochrome c (4 M) at 10°C, and the absorbance changes at 550 nm were recorded. In some cases 1 mM NADPH was added to the pre-reduced protein sample, and the mixture was incubated at 10°C for at least 15 min prior to mixing. Absorbance data were fit to a single exponential function.
Oxidation of Reduced F1395S nNOSr-A solution of F1395S nNOSr protein (8 M) containing either EDTA (1 mM) or CaCl 2 (2 mM) ϩ CaM (20 M) in air-saturated buffer was reduced by adding NADPH (160 M) and then allowed to autoxidize at room temperature in an open cuvette while following the process at 457 nm and recording visible spectra at the indicated time points during the experiment (see "Results").
Preparation of EPR Samples-All EPR experiments were carried out in 20 mM HEPES buffer, pH 7.4, with 25% glycerol (v/v). The nNOS protein concentrations used were 40 -60 M. The dysprosium(III) complex with HEDTA (Dy(III)-HEDTA) was prepared by chelating For all samples containing CaM, the molar ratio of CaM to nNOS was 2:1, and the Ca 2ϩ concentration was 1-2 mM. Protein samples were treated with NADPH followed by air oxidation to generate the flavin semiquinone radical and then transferred to sealed EPR tubes. Sperm whale myoglobin (MbNO) was prepared as described previously (55) and used as a reference.
EPR Spectroscopy-EPR spectra were recorded at 35 K using a Varian E-9 EPR spectrometer equipped with an Oxford Instruments ESR-9 flowing helium cryostat. The microwave frequency was 9.28 GHz, modulation amplitude was 4 G at 100 kHz, and the field center was set as 3300 G. Microwave power saturation data were plotted and fit to the following empirical equation as described by Rupp et al. (56): where S is the signal height; K is a proportionality factor; P is the microwave power; P1 ⁄2 is the power required for half-saturation; and b is the inhomogeneity parameter which was set at 1 in this case (55).

RESULTS
Protein Expression and Purification-Given recent data contradicting the effects of CaM on flavin reduction kinetics (31), special consideration was given to the protein expression and purification methods used in this work. Proteolytic degradation of the CaM-binding site in nNOSr is often observed during its expression and purification (28,31). One strategy to protect the CaM-binding site is co-expression of the nNOSr with CaM (30,31) and subsequent purification in the presence of Ca ϩ2 . A final purification step involving CaM affinity chromatography can also increase homogeneity of the product and ensure that the purified protein can reversibly bind CaM. The effectiveness of these principles is illustrated in Table I. WT nNOSr was expressed in the absence of CaM and purified using only 2Ј,5Ј-ADP-Sepharose chromatography. Its steady-state cytochrome c reductase activity was then measured both before and after an additional purification step with a CaM-Sepharose affinity column. The 2Ј,5Ј-ADP purified sample exhibited only a modest rate enhancement upon CaM binding, whereas the CaM affinity column purified fraction showed a CaM effect nearly double that of the original sample. Not surprisingly, the nNOSr protein fraction that did not bind to the CaM-Sepharose resin showed little or no CaM effect in this assay. Finally, qualitative analysis of nNOSr protein samples by SDS-PAGE showed that those prepared using the two sequential affinity chromatography steps had the greatest level of purity (data not shown).
To answer the question of whether or not co-expression with CaM has an effect on the properties of the final nNOSr produced, multiple batches of proteins were prepared that either had or had not been co-expressed with CaM. After their purification via sequential 2Ј,5Ј-ADP-Sepharose and CaM-Sepharose affinity chromatography, no differences were found between the two types of proteins significant enough to account for the discrepancies suggested in the literature regarding the effect of CaM on stimulation of flavin reduction kinetics (31). Because the nNOSr proteins co-expressed with CaM were obtained with higher yields and had a slightly higher CaM-dependent rate enhancement in the cytochrome c reductase assay, only nNOSr proteins that were co-expressed with CaM and purified according to the dual affinity chromatography method were used for the remainder of this work. Yields of purified proteins were at least 10 mg/liter culture, and flavin analysis showed that they contained the proper 2:1 flavin-to-protein stoichiometric ratio.
Steady-state Cytochrome c Reductase Activities-Cytochrome c reductase activities of WT and F1395S nNOSr proteins prepared according to the final procedure described above are listed in Table II. The WT protein exhibited a 9 -10-fold rate enhancement upon CaM binding. In contrast, the activity of the CaM-free F1395S mutant was nearly 4-fold greater than that of WT, indicating a poor repression of this catalytic activity in the mutant. Additionally, there was little or no rate enhancement upon binding CaM to F1395S nNOSr. Similar behavior was observed previously for the full-length F1395S nNOS (48).
Interaction of NADP ϩ with nNOSr Proteins-Full-length F1395S nNOS has altered nicotinamide cofactor binding properties as reflected by decreased apparent K m values for NADPH and NADH (48). In oxidized FNR (49,57) and CYPR (50) enzymes, occupancy of the nicotinamide binding adjacent to the FAD isoalloxazine in a productive conformation for hydride transfer causes a shift in the flavin absorbance spectrum. Difference spectra between the NADP ϩ -bound and -free forms exhibit a characteristic peak centered near 510 nm. We characterized the binding of NADP ϩ to oxidized WT and F1395S nNOSr and then utilized the observed spectral changes to compare their NADP ϩ affinities, amounts of productive binding, and binding kinetics. Difference spectra obtained from the binding of NADP ϩ to the nNOSr proteins are shown in Fig. 1. Panel A represents a control experiment that contained FAD in solution and no protein. As shown, a spectral perturbation occurred in the 400 -500-nm range upon addition of NADP ϩ . This provides a measure of the intrinsic interaction between the two compounds in the absence of protein. Thus, any additional absorbance changes occurring upon addition of NADP ϩ to nNOSr can be attributed to its binding with the enzyme. In addition, some mutant FNR enzymes that have an increased affinity for NADP ϩ co-purify with an equivalent of NADP ϩ and thus require treatment with excess 2Ј-AMP and gel filtration to remove the bound cofactor (49). Therefore, as shown in Fig. 1, spectral properties of the nNOSr proteins were examined with and without preincubation with 2Ј-AMP to determine whether either of them also contained bound NADP ϩ after purification. Fig. 1, panel B, shows that a spectral shift centered near 510 nm occurred in WT nNOSr along with some modulation of the spectrum between 400 and 500 nm when excess NADP ϩ was added regardless of whether nor not the sample was treated with 2Ј-AMP. A similar difference spectrum was obtained for the F1395S mutant in Fig. 1, panel C, but only after it had been incubated with 2Ј-AMP and passed through a PD-10 desalting column. Thus, both WT and F1395S nNOSr proteins exhibited similar spectral shifts that are characteristic of NADP ϩ binding in the productive conformation to FNR-like proteins. However, the F1395S mutant also showed an increased affinity for NADP ϩ as evidenced by its being purified fully in the NADP ϩbound form.
The WT and F1395S nNOSr proteins at equal concentration (20 M) were then titrated with NADP ϩ , and the absorbance increase at 510 nm was recorded. Fig. 2 (panel A) shows that the F1395S mutant achieved an ϳ3-fold greater total absorb-ance change than WT upon saturation, indicating that the mutant bound NADP ϩ with a greater percentage of the nicotinamide ring held in the productive conformation with the FAD isoalloxazine ring. CaM binding did not affect the fraction of WT nNOSr having NADP ϩ bound in the productive conformation ( Fig. 2, panel B). An attempt was made to fit the spectral titration curves to a model for protein-ligand binding (54) that does not place constraints on relative protein and ligand concentrations. The data from the WT nNOSr fit well to this model giving an apparent K d of 5.9 Ϯ 1.1 M for NADP ϩ (Fig. 2, panel A), consistent with values previously reported for nNOSr and plant FNR (33,57). Note that this apparent K d value is based only on the fraction of NADP ϩ bound in the productive conformation and thus reflects an equilibrium between productively bound NADP ϩ and all other forms (nonproductively bound, free in solution, etc.) of NADP ϩ . The affinity of F1395S nNOS toward binding NADP ϩ in the productive conformation was so high that its titration curve was essentially linear, and the protein was saturated with the addition of 1 eq of NADP ϩ (Fig. 2, panel A). These results were also observed in similar mutants of plant FNR (57), and under such conditions an apparent K d value using the ligand binding model cannot be calculated. Regardless, it is clear that F1395S nNOSr has a higher affinity toward binding NADP ϩ in the productive conformation and does so to a significantly greater extent than WT nNOSr.
Stopped-flow kinetic measurements of NADP ϩ binding were attempted to complement the equilibrium data. Unfortunately, when we tried to measure the association rate (k on ) of NADP ϩ with the nNOSr proteins at 10°C, much of the spectral gain at 510 nm associated with NADP ϩ binding took place in the dead time of the instrument. Measurement of the NADP ϩ release rate (k off ), however, was successful. As shown in Fig. 3, mixing the NADP ϩ -bound proteins with an excess of 2Ј,5Ј-ADP resulted in time-dependent absorbance decreases at 510 nm. The amount of 2Ј,5Ј-ADP needed for this experiment was deter- , and NADP ϩ and 20 M F1395S nNOSr (panel C) as described under "Experimental Procedures." Panels B and C also demonstrate the effect of treating the proteins with 2Ј-AMP followed by rapid gel filtration prior to carrying out these experiments. The major peak observed in panels B and C has a maximum at 510 nm.

FIG. 2. Titration of WT and F1395S nNOSr with NADP ؉ . Panel
A, each flavoprotein (20 M) was titrated with a solution of NADP ϩ , and the absorbance changes at 510 nm were recorded after equilibration as described under "Experimental Procedures." For WT nNOSr, its titration curve was fit to a protein-ligand binding model (calculated best fit curve displayed) yielding a K d value of 5.9 Ϯ 1.1 M. The model was not applicable with F1395S due to its high affinity for NADP ϩ . Panel B, difference spectra between WT nNOSr and the same sample after saturation with NADP ϩ in the CaM-free and CaM-bound states. mined by titrating 26 M NADP ϩ -saturated F1395S nNOSr, a concentration similar to that used in k off measurements, with 2Ј,5Ј-ADP and monitoring NADP ϩ dissociation at 510 nm (Fig.  3, inset). Based on these results, the concentration of 2Ј, 5Ј-ADP used was over 10 times that required to fully displace bound NADP ϩ from the proteins. The magnitude of absorbance changes during the kinetic measurements was ϳ5 times greater in the F1395S mutant, consistent with the mutant displaying a greater occupancy of NADP ϩ bound in the productive conformation and the WT protein having a faster off rate (i.e. more dissociation took place in the instrument dead time). The absorbance changes were fit to a single exponential function giving a k off at 10°C for WT nNOSr (64.1 Ϯ 0.6 s Ϫ1 ) that was approximately eight times faster than the k off for F1395S nNOSr (8.2 Ϯ 0.01 s Ϫ1 ). The k off for NADP ϩ in F1395S nNOSr was also determined at 23°C and was found to be 31.4 Ϯ 0.2 s Ϫ1 . CaM binding did not alter the rate of NADP ϩ dissociation in either protein (data not shown). Together, these results provide a quantitative measure of the higher NADP ϩ affinity of F1395S nNOSr and reveal that CaM does not significantly alter the NADP ϩ binding properties of WT and F1395S nNOSr.
Equilibrium Photoreduction Titrations-The bright yellow color of purified F1395S nNOSr indicated that both of its flavins were completely oxidized during the purification process as observed previously with full-length F1395S nNOS (48) and in contrast to WT nNOSr which is purified with an air-stable neutral FMN semiquinone. Fig. 4 contains data from anaerobic equilibrium photoreduction titrations of oxidized F1395S and WT nNOSr performed in the presence and absence of a known saturating concentration of NADP ϩ . The initial visible spectrum of F1395S nNOSr generally resembled that of WT nNOSr and other related flavoproteins. Only subtle absorbance differences between WT and F1395S nNOSr in the 450 -510-nm region, consistent with slightly different flavin environments, were observed. A typical complete titration is shown in Fig. 4 (panel A) for F1395S nNOSr without NADP ϩ . A graded decline in flavin absorbance at 457 nm was observed throughout the reduction along with the buildup and subsequent decay of a broad absorbance band centered near 600 nm. This latter feature indicated that thermodynamically stable equilibrium mixtures of flavin semiquinone species were formed during the titrations. In Fig. 4, panels B-E, each shows three selected visible spectra obtained during the progress of four more photoreductions: 1) initial spectrum; 2) point of maximum absorbance buildup at 600 nm; and 3) final spectrum of the titration. Two features are of particular note. First, the absorbance maxima at 600 nm were calculated as a percentage of each corresponding initial (oxidized) absorbance value at 457 nm. These percentages (Fig. 4) were nearly identical in each case. This provides evidence that the equilibrium midpoint potentials of the flavin redox couples in F1395S nNOSr are not significantly altered from those in WT nNOSr (58) and, furthermore, that these potentials are relatively unaffected by NADP ϩ binding. Second, the final spectrum for the reduction of F1395S nNOSr saturated with NADP ϩ is characterized by an absorbance increase at longer wavelengths (Fig. 4, panel E). This is identical to the behavior of the corresponding Y308S mutant of pea FNR when it was photoreduced in the presence of NADP ϩ (49). This type of broad, structureless absorbance extending into the long wavelength region was demonstrated by Massey and Palmer (59) to be characteristic of a charge transfer complex between reduced flavoprotein and oxidized pyridine nucleotide. In our case the complex was most likely between FAD hydroquinone (FADH 2 ) and bound NADP ϩ , consistent with the experimental conditions and the fact that the charge transfer signature did not appear until the end of the titration when FADH 2 accumulated. Based on the enzyme concentration used in Fig. 4, panel E (25 M), and the initial and final spectra obtained, an extinction coefficient for the reduced nNOSr-NADP ϩ charge transfer complex of 1200 M Ϫ1 cm Ϫ1 at 700 nm was calculated. The absence of such a charge transfer absorbance in the corresponding spectrum of WT nNOSr (Fig. 4, panel D) under these equilibrium conditions is consistent with the lower affinity of NADP ϩ for the WT enzyme and smaller percentage of NADP ϩ bound in the productive conformation when Phe 1395 is present. Finally, the apparent inability to completely reduce the nNOSr proteins with ease at the end of two titrations (Fig. 4, panels D and E) can be explained through partial oxidation of the flavins by the excess NADP ϩ present in these two experiments.
Kinetics of Flavin Reduction by NADPH-Flavin reduction data from this laboratory using full-length nNOS (see Refs. 18 and 60 for examples) and independent results obtained by others (29,30,33) using the isolated nNOSr established a role for CaM in providing a significant activating effect in this reductive reaction of nNOSr. Recently, Knight and Scrutton (31) reported a new detailed examination of flavin reduction and electron transfer in nNOSr. Interestingly, they did not observe a significant effect of CaM binding on flavin reduction kinetics. To examine the issue further in this laboratory, flavin reduction of oxidized WT and F1395S nNOSr proteins by NADPH was examined. Single wavelength stopped-flow traces obtained at 457 nm for the reduction of WT and F1395S nNOSr in the CaM-free and CaM-bound states are shown in Fig. 5. For each experiment, the initial absorbance value representing no flavin reduction was also obtained. In all cases a significant amount of the total absorbance change (typically near 30%) took place in the dead time of the instrument, consistent with results published previously (30). The magnitude of absorbance decrease at 457 nm was eventually the same in the CaM-bound and CaM-free forms of both enzymes, as judged from experiments that monitored flavin reduction over a longer period (data not shown). The recorded absorbance changes at 457 nm (Fig. 5) fit well to a quadruple exponential function (31) using the process described under "Experimental Procedures," and thus four rate constants were obtained in each experiment. Representative values for these rate constants (k 1 to k 4 ) are given in Table III. Flavin reductions were repeated with multiple independent batches of protein and although, not surprisingly, some minor variations were observed between batches, the overall results and relative trends reported here (Fig. 5 and Table III) were consistently reproduced.
We observed a significant effect of CaM binding on the flavin reduction kinetics of WT nNOSr. This is apparent through visual inspection of the traces in Fig. 5 and is quantified by ϳ4 -6-fold increases in the catalytically relevant rate constants k 1 , k 2 , and k 3 (Table III). Thus, our current data reaffirm the stimulatory effect of CaM on the NADPH-dependent flavin reduction kinetics in nNOSr. In contrast, CaM had no significant effect on the kinetics of flavin reduction in F1395S nNOSr. In both the CaM-bound and CaM-free states, the rate constants associated with flavin reduction in F1395S nNOSr most closely resembled those obtained with WT nNOSr in its CaM-free state. Thus, Phe 1395 appears to be required for CaM binding to stimulate the rate of NADPH-dependent flavin reduction in nNOSr.
Absorbance change at 600 nm during the reduction of NOSr and related flavoproteins corresponds to the formation and/or decay of flavin semiquinone species as well as charge transfer complexes. We utilized a combination of single wavelength and rapid-scanning diode array stopped-flow experiments to characterize the absorbance changes observed at 600 nm during the reduction of WT and F1395S nNOSr by NADPH. Reduction of both proteins by a single molar equivalent of NADPH was examined first, and the results are given in Fig. 6. There were no significant differences caused by the F1395S mutation on the overall qualitative behavior of the proteins under these conditions. The absorbance increases observed at 600 nm can be described as an initial fast phase taking place within the first 200 ms (Fig. 6, panels A and B) followed by a much slower phase that required nearly a minute for completion (Fig. 6, panels C and D). The amount of absorbance increase taking place in the slow phase was approximately double the amount of increase in the fast phase for each protein. The absorbance traces were fit to single exponential functions, and the calculated rates are given in Table IV. CaM binding increased the rates of absorbance changes at 600 nm for WT nNOSr, although these increases were relatively small, whereas CaM had no measurable effect on the rates obtained with F1395S nNOSr (data not shown). Visible spectra corresponding to the beginning and end of each phase obtained by rapid-scanning diode array detection are given in Fig. 6, inset of each panel. In Fig. 6, panels A and B, the decrease in absorbance at 457 nm and the similarity of the broad absorbance increases at longer wavelengths to those observed previously (Fig. 4, panel E) indicate that the fast phase of absorbance increase at 600 nm corresponds mainly to hydride transfer and formation of an FADH 2 -NADP ϩ charge transfer complex in both proteins. The ϳ3-fold faster formation of this complex in the F1395S enzyme when compared with WT nNOSr (Table IV) is consistent with the mutant having greater affinity and a higher percentage of NADPH productive binding. Based on the extinction coefficient for the reduced nNOSr-NADP ϩ charge transfer complex, the absorbance increases observed in Fig. 6, panels A and B, are consistent with the protein in each sample accumulating as a charge transfer complex in the initial fast phase. The subsequent slow phase corresponds to a relaxation of the system into a thermodynamically stable state that clearly involves the formation of a significant population of flavin semiquinone species in both proteins. For F1395S nNOSr, the absorbance at 600 nm  Table III. The calculated best fit curves are plotted as dotted lines.
in the final spectra obtained (Fig. 6, panel D) is 26% of the absorbance at 457 nm in the first spectrum obtained prior to the fast phase (Fig. 6, panel B). Considering the fact that some flavin reduction likely took place in the dead time of the stopped-flow instrument prior to collecting the first spectrum in Fig. 6, panel B, the agreement of this value with the percentage obtained during equilibrium reductive titration of F1395S nNOSr (Fig. 4) supports the assignment of thermodynamic equilibration to this slow phase. Semiquinone buildup in the slow phase for WT nNOSr was accompanied by some absorbance increase at 457 nm, implying that a net oxidation of the protein occurred.
We next examined the kinetics of absorbance change at 600 nm in experiments corresponding to Fig. 5 where WT or F1395S nNOSr proteins were mixed with a 10-fold excess of NADPH. For WT nNOSr, as shown in Fig. 7, there was an initial rapid absorbance increase at 600 nm followed by a slower partial decrease (Fig. 7, panel A). Each phase was fit to a single exponential function, and the results are reported in Table IV. Rapid-scanning diode array spectra corresponding to the three time points indicated in Fig. 7, panel A, are given in panels B and C. The spectra show that the initial absorbance increase at 600 nm is part of a broader absorbance increase extending past 700 nm into the long wavelength region and a decrease in absorbance in the 450 -500-nm range. This indicates that the spectral intermediate observed to build up at 600 nm under these conditions contains a charge transfer complex, most likely between FADH 2 and NADP ϩ . Analysis of the observed absorbance increases (Fig. 7) based on the extinction coefficient of the reduced nNOSr-NADP ϩ charge transfer complex indicates that at least two-thirds of the WT protein samples exist in charge transfer complexes at the point of maximum 600 nm absorbance buildup. The shapes of the spectra corresponding to time point 2 (Fig. 7, panels B and C) also imply the existence of some flavin semiquinone species, most noticeable in the CaM-free sample of this particular data set. However, the magnitude of these signals is small when compared with that observed in equilibrium experiments (Fig. 4). The spectra also imply that decay of the 600-nm signals corresponds to further reduction of the flavins beyond the 2-electron level, consistent with our kinetic results obtained at 457 nm (Fig. 5). For F1395S nNOSr, the absorbance trace at 600 nm recorded during reduction by excess NADPH (Fig. 8, panel A) showed essentially no change in amplitude in the first 0.5 s followed by a slow gain in absorbance (Fig. 8, panel A, inset) that is also known to accompany the reduction of WT nNOSr by excess NADPH on a similar time scale (31). CaM binding did not alter these results (data not shown). Spectra collected at the three time points indicated are given in Fig. 8, panel B. They show that little or no flavin semiquinone species were present at either 0.07 or 0.5 s after mixing and that nearcomplete flavin reduction was achieved within 0.5 s. The presence of some flavin semiquinone species was indicated at the end of the 10-s slow phase, perhaps due to thermodynamic equilibration within the system.
Pre-steady-state Cytochrome c Reduction-Recently, Craig et al. (30) described a stopped-flow experiment to investigate the kinetics of cytochrome c reduction by an excess of pre-reduced nNOSr. Use of this method enabled us to evaluate how CaM binding, occupancy of the NADPH-binding site, and the F1395S mutation influence the kinetics of electron transfer between the FMN hydroquinone and cytochrome c. The rate of electron transfer between the photoreduced nNOSr proteins and cytochrome c was determined by monitoring the absorbance gain at 550 nm under pseudo-first order conditions (excess nNOSr proteins), and the effects of CaM and/or NADPH binding on these rates were evaluated. Fig. 9 contains representative stopped-flow traces demonstrating the effect of binding NADPH in the absence of CaM on the kinetics of cytochrome c reduction by photoreduced WT and F1395S nNOSr. All absorb-

FIG. 6. Kinetics of anaerobic flavin reduction in WT and F1395S nNOSr by a single molar equivalent of NADPH (600 nm).
Stopped-flow traces were collected at 600 nm after rapidly mixing oxidized WT nNOSr (panels A and C) and F1395S nNOSr (panels B and D) with a single molar equivalent of NADPH at 10°C as described under "Experimental Procedures." Single wavelength data shown in panels A and B were obtained by using 8 M nNOSr proteins prior to mixing. All other data shown were obtained using 10 M nNOSr proteins prior to mixing. The absorbance changes were separated into initial fast phases (panels A and B) and subsequent slow phases (panels C and D) with representative traces given. Inset plots obtained from rapid-scanning diode array experiments illustrate visible spectral changes associated with the corresponding 600-nm single wavelength traces in each panel as follows: initial spectrum of the kinetic phase (dotted line); spectrum after plateau of absorbance at 600 nm (solid line). Base-line reference levels are indicated with arrows (panels A and B). Rate constants obtained by fitting these data to single exponential functions are reported in Table IV. The calculated best fit curves are plotted as solid lines.

TABLE III
Rates of anaerobic flavin reduction of nNOSr proteins by excess NADPH at 457 nm Reductions were carried out in the stopped-flow instrument at 10°C with a 10-fold excess of NADPH while monitoring the absorbance changes at 457 nm. Data were fit to a quadruple exponential function as described under ''Experimental Procedures.'' ance curves were fit to a single exponential function. A summary of the data obtained under all experimental conditions is given in Fig. 10. In the absence of both CaM and NADPH, the WT and F1395S proteins existed in an intermediate state with respect to their rates of electron transfer to cytochrome c (13.2 Ϯ 0.2 and 24.5 Ϯ 0.4 s Ϫ1 , respectively). To facilitate comparison of the effects of ligand binding on the two proteins, these absolute rate values were each set as 100%, and the data in Fig. 10 are given as relative percentages of these original values. Both nNOSr proteins were kinetically activated toward electron transfer to cytochrome c upon binding CaM, and the binding of NADPH and CaM simultaneously provided an additional activating effect. However, as shown in Figs. 9 and 10, there was a striking difference in how NADPH affected the CaM-free WT and F1395S nNOSr proteins. Adding NADPH to photoreduced WT nNOSr slowed its rate of electron transfer to cytochrome c, reproducing the observation by Craig et al. (30) that NADPH binding stabilizes a conformation of WT nNOSr that impedes electron transfer from the FMN hydroquinone. In contrast, adding NADPH to photoreduced F1395S nNOSr actually increased its rate of cytochrome c reduction to an extent even greater than that achieved with CaM binding alone. This indicates that Phe 1395 has an essential function in the mechanism by which NADPH binding regulates nNOSr electron transfer to cytochrome c and suggests that when Phe 1395 is absent NADPH binding stabilizes a conformation of nNOSr that enables electron transfer from the FMN hydroquinone. Solvent Accessibility of the FMN Semiquinone-Measuring the effect of the soluble electron-spin relaxing agent Dy(III)-HEDTA on the microwave power saturation of a protein-bound radical species is an established EPR method for determining the average solvent accessibility of the radical (55, 61, 62). We  7. Kinetics of anaerobic flavin reduction in WT nNOSr by excess NADPH (600 nm). Panel A, stopped-flow traces were collected at 600 nm after rapidly mixing 8 M oxidized WT nNOSr with a 10-fold excess of NADPH at 10°C as described under "Experimental Procedures." Rate constants obtained from fitting the absorbance increases and decreases to single exponential functions are reported in Table IV. The calculated best fit curves are plotted as solid lines. Spectra obtained from equivalent rapid-scanning diode array experiments using 10 M nNOSr protein samples prior to mixing corresponding to points 1, 2 (point of maximum absorbance buildup), and 3 during the first 0.5 s of the reduction are given for the CaM-free and CaM-bound protein in panels B and C, respectively. Base-line reference levels are indicated with arrows. Inset spectra (panel C) were collected on a shorter time scale (100 ms) and more clearly capture the initial long wavelength absorbance increases associated with reduction of CaM-bound WT nNOSr. applied this technique to examine further the characteristics of how NADPH, CaM, and Phe 1395 may influence the shielding of the FMN cofactor in nNOSr. Because the EPR method requires that the FMN semiquinone radical be present, it was important to establish its formation and lifetime in F1395S nNOSr as shown in Fig. 11. We collected visible spectra at key time points (Fig. 11, inset) both before and after mixing the enzyme with excess NADPH in air-saturated buffer while continuously monitoring the absorbance at 457 nm. The time points are as follows: 1) prior to NADPH addition; 2) during steady-state NADPH oxidation; and 3) after all NADPH had been oxidized.
Upon addition of NADPH most of the absorbance in the 400 -500 nm range was lost, indicating that a significant proportion of each flavin was in its 1-and 2-electron reduced forms during steady-state NADPH oxidation. The broad absorbance centered near 600 nm indicates that F1395S nNOSr contained some flavin semiquinone species during this time. After approximately 2 min, there was a re-oxidation of enzyme flavins leading to the relatively stable species 3. Based on its absorbance levels at 457 nm, the broad absorbance at 600 nm, and analogy to WT nNOSr, species 3 is assigned as containing a large proportion of the FMN semiquinone. The slight shift in absorbance toward longer wavelengths most obvious near 510 nm in spectrum 3 is consistent with the enzyme containing bound NADP ϩ at this point. Although species 3 was stable within the time frame of our study, it underwent further oxidation over the next 60 min regenerating fully oxidized F1395S nNOSr (data not shown). CaM binding had minimal effects on the flavin oxidation kinetics. These results show that the F1395S mutation increased the O 2 reactivity of the reduced flavins relative to WT nNOSr and establish that the transient FMN semiquinone radical in F1395S nNOSr is sufficiently air-stable to perform the desired EPR experiments.
Full-length nNOS, WT nNOSr, and F1395S nNOSr were each treated with a slight excess of NADPH and then allowed to air-oxidize until they reached their 1-electron reduced FMN semiquinone forms. The samples were frozen, and the effects of Dy(III)-HEDTA on the microwave power saturation properties of the FMN semiquinone radicals were evaluated. Representative power saturation curves for full-length nNOS in the presence of four different concentrations of Dy(III)-HEDTA are shown in Fig. 12. In each case, the microwave power saturation of the FMN semiquinone radical was relieved by the addition of Dy(III)-HEDTA in a concentration-dependent manner. A set of experiments was performed for each of the three proteins to determine how CaM binding to each NADP ϩ -bound protein would affect the ability of Dy(III)-HEDTA to alter their power saturation curves. The curves were fit as described under "Experimental Procedures" and characterized by the parameter P1 ⁄2 .
The values of P1 ⁄2 obtained in the absence of Dy(III)-HEDTA FIG. 9. Stopped-flow traces illustrating the NADPH-dependent activation or inhibition of electron transfer to cytochrome c in pre-reduced CaM-free WT and F1395S nNOSr. Photoreduced samples of WT (top) and F1395S (bottom) nNOSr with and without NADPH were prepared and rapidly mixed with a substoichiometric amount of cytochrome c in the stopped-flow instrument at 10°C as described under "Experimental Procedures" while recording the absorbance increases at 550 nm. Rate constants obtained by fitting these data to single exponential functions were used in the preparation of Fig. 10. The calculated best fit curves to the traces shown here are plotted as dotted lines.
FIG. 10. Summary of pre-steady-state cytochrome c reduction rates by pre-reduced WT and F1395S nNOSr. The photoreduced proteins were rapidly mixed under various conditions with a substoichiometric amount of cytochrome c in the stopped-flow instrument at 10°C as described under "Experimental Procedures," and the absorbance increases at 550 nm were fit to single exponential functions. The relative percent reduction rates were calculated and are plotted for each protein to demonstrate the effects of CaM and NADPH binding on the electron transfer to cytochrome c.

FIG. 11. Autoxidation of NADPH-reduced F1395S nNOSr.
F1395S nNOSr was treated with an excess of NADPH in air-saturated buffer at the point indicated and allowed to re-oxidize at room temperature in the spectrophotometer. The reduction and re-oxidation were followed at 457 nm with and without bound CaM. Inset visible spectra (1)(2)(3) were recorded at the points indicated during the course of the reaction (see text).
were subtracted from each P1 ⁄2 value obtained with added Dy(III)-HEDTA to obtain ⌬P1 ⁄2 , a parameter that reflects only the influence of Dy(III)-HEDTA on the FMN semiquinone radical power saturation curves. The ⌬P1 ⁄2 values determined in the presence and absence of CaM for each of the three NADP ϩbound nNOS proteins are plotted versus the concentration of Dy(III)-HEDTA in Fig. 13. In each case, there was a linear relationship between ⌬P1 ⁄2 and Dy(III)-HEDTA concentration. Linear regression analysis of the data gave the slopes of the lines, m (mW/mM), which are reported in Table V. The values of m provide a measure of the physical accessibility of the Dy(III)-HEDTA molecules in solution to the protein-bound FMN semiquinone radical. A larger value for the slope indicates greater dependence of ⌬P1 ⁄2 on added Dy(III)-HEDTA and thus greater accessibility of the FMN semiquinone radical to the solvent. Both full-length nNOS and WT nNOSr exhibited a significant increase of m upon binding CaM. In contrast, F1395S nNOSr in the CaM-free state had an m value that was comparable with both of the WT proteins in their CaM-bound forms and showed only a slight increase in m with CaM binding. These data are consistent with NADP ϩ stabilizing a conformation of WT nNOSr characterized by a less accessible FMN in the absence of CaM, but being incapable of doing so and actually having the opposite effect in CaM-free F1395S nNOSr. The data also show that CaM binding reverses the effect of NADP ϩ and thus promotes greater FMN accessibility.

DISCUSSION
Phe 1395 is of interest because of its participation in the unique and complex regulatory mechanism of nNOS (48). The impact of the F1395S mutation on different aspects of nNOSr function and control revealed in this work are discussed individually and then collectively to explain the role of Phe 1395 in determining the steady-state catalytic activities of nNOSr.
Phe 1395 and the Interaction of nNOSr with NADP(H)-Our data indicate that F1395S nNOSr has an increased affinity for NADP ϩ and, by extension of FNR structural data (49), binds NADP ϩ with a significantly higher percentage of its nicotinamide ring occupying a productive conformation relative to WT enzyme. If one accepts that the absorbance change obtained at 510 nm upon saturating the proteins with NADP ϩ is directly proportional to the fraction of productively bound cofactor and that the F1395S mutation, as assumed with plant FNR mutants (57), enables 100% occupancy of the productive binding mode in nNOSr, then ϳ35% of the NADP ϩ bound to WT nNOSr is in the productive conformation. A similar fraction for NADP ϩ productive binding was observed in plant FNR (57) and CYPR (50). These relatively low percentages are not surprising given that the nicotinamide ring is bound in a nonproductive conformation away from the FAD in the crystal structures of both these enzymes (8,14) and in the FNR fragment of WT nNOSr (12). In fact, the productive NADP ϩ -binding mode has only been observed in crystal structures of FNR and CYPR mutants whose flavin-shielding residues corresponding to Phe 1395 were mutated to remove the aromatic side chain (49,50). Our findings in nNOSr are consistent with the multiple-interaction bipartite binding model for NADP (H) (57, 63, 64), where dis-  placement of Phe 1395 from the FAD isoalloxazine to achieve nicotinamide ring binding in the productive conformation carries with it an energetic cost. When that cost is lowered by mutation of Phe 1395 , a greater fraction of bound NADP(H) can achieve the productive conformation, and the resulting increase in noncovalent binding interactions between the nicotinamide and FAD isoalloxazine rings creates an overall higher affinity toward NADP(H). Most important, this mutational effect can slow catalysis by the flavoprotein if it causes NADP ϩ dissociation to become rate-limiting in steady-state catalysis. Such a phenomenon was observed in the form of a dramatic decrease in cytochrome c reduction activity by the structurally related W676H and W676A mutants of CYPR (63) and also becomes relevant with F1395S nNOSr as discussed below, although the two proteins are fundamentally different in the sense that the F1395S mutation in CaM-free nNOSr causes a net increase in its cytochrome c reduction activity. Although a productive nicotinamide-FAD interaction is prerequisite for flavin reduction by NADPH, the degree of productive NADP ϩ binding in nNOSr did not correlate with the rates of flavin reduction. For example, CaM binding increased the rate of flavin reduction in nNOSr without changing its percentage of productive NADP ϩ binding. Also, the F1395S mutant had either equivalent (ϪCaM) or slower (ϩCaM) rates of NADPH-dependent flavin reduction relative to WT nNOSr despite its exhibiting a 3-fold greater percentage of productive NADP ϩ binding. This may suggest that productive and nonproductive NADP(H) conformers are in rapid equilibrium within these flavoproteins. Thus, the specific ability of Phe 1395 to control the distribution of NADP ϩ -binding modes does not appear to be a mechanism by which this residue influences flavin reduction kinetics or its regulation by CaM in WT nNOSr.
Flavin Reduction-In general, reduction of nNOSr and related flavoproteins begins with hydride transfer from NADPH to FAD followed by an interflavin FAD to FMN electron transfer yielding a 2-electron reduced di-semiquinone form of the protein. Subsequently, and possibly corresponding, events include NADP ϩ dissociation, binding of a second molecule of NADPH, additional interflavin electron transfer, and a second hydride transfer to FAD. The stopped-flow traces we collected at 457 nm during reduction of fully oxidized nNOSr by excess NADPH span the entire process described above and fit well to a multiple exponential function yielding four distinct rate constants. Although our kinetic analysis of the stopped-flow data was purposely identical to that recently published for nNOSr by Knight and Scrutton (31), CaM binding caused significant increases in each of the catalytically relevant rate constants (k 1 , k 2 , and k 3 , Table III) only in the present work. It is noted that our protein construct was 38 amino acids longer at the N-terminal end than the one used by Knight and Scrutton (31). Interestingly, however, the rat nNOSr protein used by Iyanagi and co-workers (29) was also truncated relative to ours at the N-terminal end by over 20 amino acids, but they observed a stimulation of flavin reduction kinetics by CaM as well. Besides the effect on flavin reduction kinetics, we found that CaM did not significantly impact the amount of absorbance change occurring in the instrument dead time, the percentage of absorbance change associated with each of the individual rate constants, or the total absorbance change achieved during the reduction. This suggests that CaM accelerates multiple steps in the flavin reduction process.
CaM binding did not significantly change the flavin reduction kinetics of F1395S nNOSr at 457 nm, which resembled those of WT nNOSr in its kinetically repressed, CaM-free state. We can thus conclude that a change in flavin reduction kinetics is not responsible for the increased steady-state cytochrome c reductase activity of CaM-free F1395S nNOSr, and furthermore that repressed flavin reduction rates likely do not limit steady-state cytochrome c reduction rates in our CaM-free WT nNOSr. The inability of CaM to increase the flavin reduction rates of F1395S nNOSr, particularly with respect to the phases corresponding to k 2 and k 3 , could be due to the slower NADP ϩ off rate of the mutant if electron transfer within nNOSr is gated by NADP ϩ release (see below). It could also be due to another regulatory effect of Phe 1395 . At this point, we can say that displacement of the phenyl side chain from the FAD isoalloxazine ring upon productive binding of NADP(H), a process absent in the mutant, may be an important molecular motion tied to the relief of kinetic inhibition of flavin reduction by CaM.
The buildup and/or decay of flavin semiquinone species as well as charge transfer complexes during flavin reduction in proteins like nNOSr is indicated by absorbance changes at 600 nm. Current literature (29,31) on nNOSr is equivocal as to whether or not anaerobic reduction of oxidized WT rat nNOSr by excess NADPH is accompanied by a significant buildup of flavin semiquinone species. Our combined single wavelength stopped-flow traces, rapid-scanning diode array spectra, and equilibrium reductive titrations indicate that with our protein samples the initial absorbance gain at 600 nm during reduction with excess NADPH corresponds to buildup of charge transfer complexes mixed with only a minor amount of flavin semiquinone species. When compared with literature data on this topic, our spectra and interpretation appear most similar to those of Knight and Scrutton (31). The presence or lack of flavin semiquinone buildup during reduction by excess NADPH is only a question of relative rates since these species must always form at least transiently due to a thermodynamically favored FADH 2 to FMN single electron transfer known to occur at some point following the first hydride transfer from NADPH. Buildup of semiquinone species is expected only if interflavin FADH 2 3 FMN electron transfer is fast relative to events such as NADP ϩ release and hydride transfer from a second NADPH molecule. The lack of a large semiquinone buildup demonstrated here and the slow equilibration to semiquinone species in the absence of excess NADPH observed in Fig. 6 both support the previously proposed concept (30, 31) that our rat nNOSr contains a kinetic limitation on interflavin electron transfer which is relieved, at least in part, by NADP ϩ release.
The lack of absorbance change at 600 nm during the first 0.5 s of F1395S nNOSr flavin reduction by excess NADPH is interesting given that we observed rapid charge transfer complex formation when the mutant protein was mixed with a single equivalent of NADPH. The increased NADPH affinity of F1395S nNOSr may have made charge transfer complex formation so fast that it occurred in the instrument dead time when the protein was mixed with an excess of NADPH. This could explain the lack of rapid absorbance gain at 600 nm. The lack of subsequent absorbance decay at 600 nm may indicate that charge transfer complexes are maintained throughout the reduction process by a first and second molecule of NADPH. A more conclusive explanation for this phenomenon must await future studies.
These reduction experiments also illustrate a difference between F1395S nNOSr and related CYPR mutants (W676H and W676A). When the CYPR mutants are mixed with excess NADPH, the product is a stable 2-electron reduced protein due to NADP ϩ remaining tightly bound and preventing further reduction (65). In contrast, absorbance changes obtained during reduction of F1395S nNOSr by excess NADPH indicate that it is reduced beyond the 2-electron level, revealing that these aromatic residues, despite their structural conservation, impact NADP ϩ affinity to different degrees within the two flavoproteins.
Regulating nNOSr Conformational States Relevant to Electron Transfer-On the basis of steady-state cytochrome c reductase activities alone, it appears that the CaM-free F1395S mutant lacks repression and its CaM-bound form lacks the ability to become fully activated, or de-repressed, when compared with WT nNOSr. Previous work with three related flavoproteins (50,66,67) led to the idea that some type of FMN subdomain reorientation relative to the rest of the flavoprotein must occur during catalytic reduction of an external electron acceptor. Such a model is illustrated in Fig. 14. A crystal structure of CYPR depicts a conformation where the isoalloxazine rings of FMN and FAD are positioned next to each other (8) such that electron transfer between the two flavins is enabled (Fig. 14, species A), similar to what is also observed in a recent crystal structure of a complete nNOSr. 2 Once the FMN hydroquinone is formed in this conformation, however, it is thought to be incapable of delivering electrons to an acceptor such as the NOS heme or cytochrome c without first adopting a more open conformation (Fig. 14, species B) (50,66,67). 2 Conformations A and B (Fig. 14) likely represent limits of a dynamic continuum that are both essential for catalytic turnover by nNOSr and related flavoproteins. Conformation A is required to generate FMN hydroquinone, whereas conformation B is necessary to transfer electrons from the FMN hydroquinone to external acceptors. Thus, the steady-state activity of NOSr is determined at least in part by the position and dynamics of the equilibrium between conformations A and B. Efficient catalysis results when the energy barrier for interconversion between the two states of the FMN module is low enough to allow facile electron transfer through the nNOSr via the FMN cofactor. Conversely, a NOSr with either form significantly stabilized relative to the other will appear to be catalytically repressed with respect to reduction of external electron acceptors such as cytochrome c. To the extent that external ligands (CaM and NADPH) and the NOSr regulatory elements influence the relative energies, and thus the equilibrium, between these two forms of the FMN subdomain, they can control the overall catalytic activity of the NOSr. This model assumes that electron transfer between the FMN hydroquinone and the acceptor is fast, as appears to be the case when excess cytochrome c is the acceptor in the presence or absence of CaM (30).
This conceptual model of NOSr regulation via control of a conformational equilibrium (Fig. 14) mandates that surface or solvent accessibility of the FMN cofactor should directly correlate with the observed rate of electron transfer to external acceptors, and any exceptions to this rule require further explanation. Measuring the kinetics of pre-steady-state cytochrome c reduction by an excess of pre-reduced nNOSr is one way to determine the relative proportion of FMN hydroquinone accessible to an electron acceptor under various conditions (30). In complementary experiments, the average solvent accessibility of the FMN semiquinone radical under each condition can be determined by using EPR spectroscopy. The data we obtained for WT and F1395S nNOSr ( Fig. 10 and Table V) are consistent with the model and reveal a key difference between these two proteins with regard to ligand control of their FMN subdomain equilibria. Upon binding NADPH, a greater proportion of the FMN subdomain in CaM-free WT nNOSr becomes shielded as judged by the pre-steady-state cytochrome c reduction rate decreasing by ϳ75%. Remarkably, the very same conditions (ϪCaM, ϩNADPH) shifted the conformational equilibrium of F1395S nNOSr toward the FMN de-shielded form, as judged by its greater than 2-fold increase in cytochrome c reduction rate over the original value. Binding of CaM alone to either nNOSr protein favors their FMN de-shielded forms, and in the case of WT nNOSr this overrides the shielding effect of NADPH binding. These effects are fully supported and verified by the EPR experiments. Based on these data, the following can be concluded. (a) Structurally, the F1395S mutant has a normal CaM response since CaM binding appears to induce conformational changes similar in nature to those observed with WT nNOSr. Therefore, any catalytic properties of F1395S nNOSr that fail to reflect a normal CaM effect cannot simply be explained by a defective CaM response. (b) Phe 1395 is critically involved in the NADPH-dependent regulatory mechanism controlling the equilibrium, and thus relative energies, between FMN-shielded and -de-shielded forms of nNOSr. Interestingly, the F1395S mutation did not simply eliminate NADPH-dependent regulation, but it actually reversed the response to NADPH such that it became an activating ligand for electron transfer from the CaM-free mutant protein. (c) The excellent correlation between the EPR data and pre-steady-state cytochrome c reduction results validates the use of the latter for evaluating FMN accessibility.
Implications for Steady-state Catalysis-Craig et al. (30) described the nNOSr as being poised on a conformational "seesaw" with regard to repression of its electron transfer with NADP(H) acting on one side and CaM on the other. 3 Our data support their description and suggest that the "see-saw" actually corresponds to the relative positioning of a dynamic equilibrium between shielded and de-shielded states of the FMN subdomain (Fig. 14). As the relative energies of the two conformations change due to CaM binding, the equilibrium shifts toward the less shielded conformation, and the FMN module becomes more catalytically active. Of course, if the equilibrium shifts too far toward the de-shielded conformation, the protein is predicted to lose catalytic activity due to an inability to reform the FMN hydroquinone at a sufficient rate.
Craig et al. (30) proposed that the repressed steady-state cytochrome c reduction of CaM-free nNOSr is due to NADP(H)dependent conformational effects because a substantial proportion of nNOSr is NADP(H)-bound during turnover. This concept is consistent with our current data. A similar NADP(H) occupancy must occur in CaM-free F1395S nNOSr during its catalytic turnover, but in this case NADP(H) binding favors FMN de-shielding and results in a higher steady-state cytochrome c reductase activity relative to CaM-free WT. Thus, the steady-state catalytic difference between the two CaM-free nNOSr proteins is due to NADP(H) having an opposite effect on a conformational equilibrium rather than any intrinsic energetic difference between them. The divergent NADPH response in the mutant may also explain our observation of some NO synthesis by CaM-free full-length F1395S nNOS (48). Thus, our results extend the existing hypothesis (48) that movement of the Phe 1395 side chain upon binding NADP(H) helps trigger conformational changes involving the adjacent C-terminal control element of nNOSr, which together with the autoinhibitory insert may control the equilibrium between FMN shielded and de-shielded forms (12,30).
If CaM binding de-shields the FMN subdomain of F1395S nNOSr to a similar level as in WT enzyme ( Fig. 10 and Table  V), then why is the steady-state cytochrome c reductase activity of CaM-bound F1395S nNOSr only 50% that of CaM-bound WT enzyme? The answer is the slow dissociation of NADP ϩ from F1395S nNOSr. The F1395S k off for NADP ϩ at 23°C is ϳ31 s Ϫ1 (1860 min Ϫ1 ), in excellent correlation with the steady-state cytochrome c turnover rate of CaM-bound F1395S nNOSr at 23°C (2044 Ϯ 300 min Ϫ1 ). Although our measured k off rates are from the fully oxidized protein and not the partially reduced forms existing during steady-state catalysis, we believe that collective consideration of all the data presented gives strong support to this conclusion. 4 Thus, the increased NADP ϩ affinity in the mutant results in slower NADP ϩ release, and this step becomes rate-limiting for steady-state catalysis as evidenced by the relatively slow cytochrome c reduction activity relative to CaM-bound WT nNOSr. In addition, since NADPH binding to CaM-free F1395S nNOSr de-shields its FMN module nearly as much as binding CaM, the steady-state cytochrome c reductase activity of CaM-free F1395S nNOSr appears to be limited by NADP ϩ release as well. In other words, the steadystate cytochrome c reduction activities of CaM-free and CaMbound F1395S would be significantly higher if not for this kinetic limitation caused by removal of Phe 1395 . These conclusions are supported by analysis of NADH-versus NADPH-dependent cytochrome c reduction by CaM-bound F1395S nNOS (48). Because NADH lacks the 2Ј-phosphate group, its affinity for the protein is lower, and its dissociation rate (k off ) is ex-pected to be faster than that of NADPH. Thus, the V max cytochrome c reductase activity of F1395S nNOS should be greater for NADH than for NADPH, which was indeed observed (48).
Our data may help explain how Phe 1395 impacts NOS heme reduction and NO synthesis. Although CaM binding de-shields the FMN subdomain in F1395S nNOSr to an extent comparable with WT nNOS, its rates of heme reduction and NO synthesis are much lower (48). Thus, without Phe 1395 side chain movement, the nNOSr regulatory elements may not fully relieve the block on NOS heme access in response to CaM, which appears to be a separate function (19) occurring along with the de-shielding of the FMN subdomain. This possibility can now be explored.
Conclusions-Our data suggest a regulatory model for nNOSr in which its observed catalytic activity is influenced by an equilibrium between relatively shielded and de-shielded conformations of the FMN subdomain. Phe 1395 is required to stabilize the shielded conformation when NADP(H) binds in the absence of CaM, and this is associated with catalytic repression. CaM binding shifts the equilibrium toward a deshielded conformation giving the FMN hydroquinone greater access to cytochrome c, but Phe 1395 is not required for this effect of CaM. Once the protein has been conformationally activated by CaM binding, however, Phe 1395 is required to prevent NADP ϩ release from limiting the rate of cytochrome c reduction in the steady state. Phe 1395 appears to have distinct roles in regulating NADP(H) affinity, transmitting the effect of NADP(H) binding on the conformational equilibrium of CaMfree nNOSr and facilitating reduction of flavin and heme centers in CaM-bound nNOS. Its phenyl side chain controls the extent of productive NADP(H) binding, and this explains its effect on NADP(H) affinity and related catalytic impact. However, we speculate that the other roles of Phe 1395 are manifested through phenyl side chain movement away from FAD upon productive NADP(H) binding, as it is this movement that is expected to influence the behavior of an adjacent regulatory element in nNOSr.