Calmodulin Activates Electron Transfer through Neuronal Nitric-oxide Synthase Reductase Domain by Releasing an NADPH-dependent Conformational Lock*

Neuronal nitric-oxide synthase (nNOS) is activated by the Ca 2 (cid:1) -dependent binding of calmodulin (CaM) to a characteristic polypeptide linker connecting the oxygenase and reductase domains. Calmodulin binding also activates the reductase domain of the enzyme, increas-ing the rate of reduction of external electron accep-tors such as cytochrome c . Several unusual structural features appear to control this activation mechanism, including an autoinhibitory loop, a C-terminal extension, and kinase-dependent phosphorylation sites. Pre-steady state reduction and oxidation time courses for the nNOS reductase domain indicate that CaM binding triggers NADP (cid:1) release, which may exert control over steady-state turnover. In addition, the second order rate constant for cytochrome c reduction in the absence of CaM was found to be highly dependent on the presence of NADPH. It appears that NADPH induces a conformational change in the nNOS reductase domain, restricting access to the FMN by external electron acceptors. CaM binding reverses this effect, causing a 30-fold increase in the second order rate constant. The results show a star-tling interplay between the two ligands, which both exert control over the conformation of the domain to influence its electron transfer properties. In the full-length enzyme, NADPH binding will probably close the conformational lock in vivo , preventing electron transfer to the oxygenase domain and the resultant stimula-tion of nitric oxide synthesis. m M m M B), 1 of 50 m M 1 m M , 0.5 M NaCl, and 1 liter of buffer B. The was from the in 50 m M Tris-HCl, pH 7.5, 1 m M CaM-containing were identified by and before being dialyzed in B to remove The was dried and stored at (cid:3) 80 °C. Analysis by the Bio-Rad protein assay showed that the freeze-dried residue was 70% protein by weight, whereas SDS-PAGE indicated that the protein was (cid:4) 90% CaM. CaM prepared in this manner was indistin- guishable from the CaM supplied by Sigma when used in the CaM activation assay described above. Anaerobic Stopped-flow Spectrophotometry— Stopped-flow experiments were performed on an SX.18MV stopped-flow spectrophotometer (Applied Photophysics) at a temperature of 25 (cid:5) 1 °C contained within an anaerobic glove box (Belle Technology) to prevent the reaction of reduced nNOSrd with molecular oxygen. Oxygen levels were main-tained below 5 ppm in a nitrogen atmosphere at all times. were using Origin 6.1 (Microcal). Averaged traces four or more measurements were used for all stopped-flow experi- the was 50 m M Tris-HCl, pH m M supple-mented with 2 m M made 2 h before , , ,5 dithionite in

structures are available for several NOS oxygenase domain dimers (5,6) and for the FAD binding subdomain of neuronal NOS (nNOS) (7). The reductase domain closely resembles mammalian cytochrome P450 reductase (7)(8)(9)(10)(11) and similarly catalyzes NADPH dehydrogenation at the FAD site and electron transfer to the FMN. The oxygenase domain of one subunit accepts electrons from the reductase domain of the other subunit (12)(13)(14) and generates NO from L-arginine via a unique two-step monooxygenation reaction (1)(2)(3)(4)15). The two domains are linked by a functional peptide of 20 -25 amino acids which binds calmodulin (CaM) reversibly at elevated Ca 2ϩ concentrations in the nNOS and endothelial NOS (eNOS) isoforms but irreversibly in the inducible isoform (iNOS). CaM binding activates nNOS and eNOS, providing them with a rapid response mechanism during their participation in signaling cascades. The inducible isoform, on the other hand, is regulated at the transcriptional level. CaM binding has been shown to control NO synthesis by activating electron transfer through the enzyme (16,17). This effect is manifested in both the reductase domain and the enzyme as a whole by increases in the rates of steady-state cytochrome c reduction and NOS heme reduction. The isolated nNOS reductase domain (nNOSrd) retains its CaM-dependent cytochrome c reductase activity in the absence of the heme domain (18,19), whereas the isolated iNOSrd is not CaM-dependent (19,20). The three NOS isoforms show close sequence homology but can be readily identified by the presence, or absence, of several unusual isoform-specific inserts and extensions, which are responsible for defining their particular functional characteristics. Many of these regulatory control elements occur in the reductase domain (21) and include an autoinhibitory loop in the FMN binding subdomain of 40 -50 amino acids (20 -27), a C-terminal extension to the FAD binding subdomain of 20 -40 amino-acids (28,29), and kinase-dependent phosphorylation sites (30 -36). All influence the CaMdependent activation mechanism by either altering the CaM binding affinity or by affecting enzyme activity directly. Mutation of these control elements (e.g. by deletion) activates the reductase domains of eNOS and nNOS in the absence of CaM (Table I) and in some cases triggers NO synthesis (25,29). The reduction potentials of the FAD and FMN cofactors of the nNOSrd are unperturbed by CaM binding (37), indicating that the activation mechanism relies on a large scale structural rearrangement. In this paper, we probe the effect of CaM binding on the isolated nNOSrd by following reduction by NADPH and oxidation by cytochrome c under pre-steady-state conditions. The results indicate that the CaM-free enzyme forms a conformationally locked complex with NADPH, which has poor electron transfer capabilities. The importance of the NADPH and NADP ϩ dissociation rates to the CaM-dependent activation mechanism is discussed.

EXPERIMENTAL PROCEDURES
Preparation of Recombinant nNOS Reductase Domain-Recombinant nNOSrd was expressed in Escherichia coli strain JM109 (DE3) using plasmid pCRNNR as described previously (19). This plasmid coexpresses the rat nNOSrd residues 695-1429 (including the CaM binding site) and synthetic bovine brain calmodulin. The enzyme was purified essentially as described previously, in 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 5% glycerol (buffer A) plus 2 mM CaCl 2 on 2Ј,5Ј-ADP-agarose (Sigma). Bound CaM was removed by washing with buffer A containing 10 mM EGTA. The enzyme was eluted in buffer A plus 10 mM NADP ϩ and 2 mM CaCl 2 . The nNOSrd fractions were purified further on CaMagarose (Sigma) and eluted in buffer A plus 10 mM EGTA. The enzyme was exchanged into 50 mM Tris-HCl, pH 7.5, 2 mM CaCl 2 , 100 mM NaCl, and 5% glycerol and stored at Ϫ80°C. Enzyme concentrations were calculated from the absorbance at 457 nm of nNOSrd fully oxidized by potassium ferricyanide, using an extinction coefficient of 22,900 M Ϫ1 cm Ϫ1 (38). Enzyme activity and CaM sensitivity were checked using the cytochrome c turnover assay (without FAD, FMN, superoxide dismutase, and catalase) described by Gachhui et al. (18). Absorbance and kinetic measurements were made on a Shimadzu UV-1601 spectrophotometer.
Purification of CaM-The CaM-rich 2Ј,5Ј-ADP-agarose wash fraction was made 10 mM with CaCl 2 and then loaded onto a 200-ml phenyl-Sepharose column and washed with 800 ml of 50 mM Tris-HCl, pH 7.5, 1 mM CaCl 2 (buffer B), 1 liter of 50 mM Tris-HCl, pH 7.5, 1 mM CaCl 2 , 0.5 M NaCl, and 1 liter of buffer B. The CaM was eluted from the column in 50 mM Tris-HCl, pH 7.5, 1 mM EDTA. CaM-containing fractions were identified by SDS-PAGE and pooled before being dialyzed in buffer B to remove EDTA. The CaM was freeze dried and stored at Ϫ80°C. Analysis by the Bio-Rad protein assay showed that the freeze-dried residue was 70% protein by weight, whereas SDS-PAGE indicated that the protein was Ͼ90% CaM. CaM prepared in this manner was indistinguishable from the CaM supplied by Sigma when used in the CaM activation assay described above.
Anaerobic Stopped-flow Spectrophotometry-Stopped-flow experiments were performed on an SX.18MV stopped-flow spectrophotometer (Applied Photophysics) at a temperature of 25 Ϯ 1°C contained within an anaerobic glove box (Belle Technology) to prevent the reaction of reduced nNOSrd with molecular oxygen. Oxygen levels were maintained below 5 ppm in a nitrogen atmosphere at all times. Data were analyzed using Origin 6.1 (Microcal). Averaged traces from four or more measurements were used for analysis. For all stopped-flow experiments, the buffer was 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, supplemented with 2 mM CaCl 2 when CaM was present, unless otherwise stated. Buffer was made anaerobic by bubbling with nitrogen for 2 h before being left to equilibrate in the anaerobic box overnight. NADPH, NADP ϩ , NADH, NAD ϩ , 2Ј,5Ј-ADP, cytochrome c (horse heart, type I; Sigma), and dithionite were brought into the anaerobic box in powder form and dissolved in anaerobic buffer. Concentrations were determined by absorption spectroscopy on a Cary 50 Biospectrophotometer within the anaerobic box using the extinction coefficients 29,500 M Ϫ1 cm Ϫ1 at 550 nm (dithionite-reduced sample), 6,200 M Ϫ1 cm Ϫ1 at 340 nm, 18,000 M Ϫ1 cm Ϫ1 at 260 nm, 6,220 M Ϫ1 cm Ϫ1 at 340 nm, 18,000 M Ϫ1 cm Ϫ1 at 260 nm, and 15,400 M Ϫ1 cm Ϫ1 at 260 nm. Reduced nNOSrd was generated by titration with a concentrated solution of NADPH or dithionite until no further spectral change occurred. To produce oneelectron-reduced enzyme, nNOSrd was incubated in the anaerobic box in a 1 mM solution of dithiothreitol overnight. Enzyme (reduced or oxidized) was made anaerobic and free of excess reductant by passage down a 1.5 ϫ 20-ml Sephadex G-25 (Sigma) size separation column immediately prior to use. CaM was brought into the box in a 3 mM solution containing 30 mM CaCl 2 and added to the nNOSrd to achieve a 2:1 concentration ratio of CaM to nNOSrd. Concentrations of components of stopped-flow reaction mixtures are given as final concentrations after mixing.
Determination of Second Order Rate Constants for Cytochrome c Reduction-2 M cytochrome c was mixed with nNOSrd (NADPH-or dithionite-reduced) at concentrations of 5-40 M with and without bound CaM and with and without a 10-fold excess of NADPH (incubated with the enzyme 10 min prior to use). Cytochrome c reduction was monitored at 550 nm, and pseudo-first order rate constants were calculated by fitting the resultant traces to single exponential functions. Second order rate constants were determined by plotting the pseudofirst order rate constants against enzyme concentration and performing linear regression analysis constrained to pass through the origin.
Inhibition of Cytochrome c Reduction-Inhibition of cytochrome c reduction by substrates and substrate analogs was studied by mixing 2 M cytochrome c with 10 M dithionite-reduced nNOSrd, plus or minus bound CaM, in the presence of 5-1,000 M NADPH, NADP ϩ , NADH, NAD ϩ , or 2Ј,5Ј ADP. Rate constants for the reduction of cytochrome c were calculated from the change in absorbance at 550 nm as described above.
Flavin Oxidation by Excess Cytochrome c-Oxidation of the nNOSrd flavins by an excess of cytochrome c was conducted by mixing 2 M enzyme (with and without bound CaM and reduced by either NADPH or dithionite) with 10 -130 M cytochrome c. Cytochrome c reduction was monitored at 550 nm. At low concentrations of cytochrome c oxidation was slow, allowing the entire reaction course to be observed. These traces were used to calculate the number of electron equivalents removed from the nNOSrd sample, based on ⌬⑀ 550 ϭ 22,640 for cytochrome c reduction. Faster reactions using higher cytochrome c concentrations, in which much of the absorbance change occurred in the stopped-flow dead time, were rescaled accordingly.
Flavin Reduction-Reduction of the flavins of nNOSrd by NADPH was achieved by mixing 5 M one-electron-reduced enzyme (prepared with bound CaM; with and without 5 mM EGTA) with 200 M NADPH. The change in absorbance at 458 nm was followed and the resultant traces fitted to double or triple exponential functions.
Single Turnover Cytochrome c Reduction-To study the rate of electron transfer from NADPH through to cytochrome c, 10 M one-electron-reduced nNOSrd (prepared with bound CaM) with 200 M cytochrome c, with and without 5 mM EGTA, was mixed with 2 M NADPH. The change in absorbance was monitored at 550 nm. The traces were fitted to a two-step consecutive reaction model in which both phases were fixed to have the same absorbance change.

RESULTS
Cytochrome c Reduction-Electron transfer from nNOS or nNOSrd to cytochrome c occurs from the FMN hydroquinone (two-electron-reduced FMN) and is essentially unidirectional (25,39). A kinetic barrier prevents oxidation of the blue FMN semiquinone by cytochrome c despite the reaction being thermodynamically favorable (29,37,44). Therefore, during catalytic turnover with NADPH and cytochrome c, the FMN alternates between the hydroquinone and semiquinone redox states, receiving electrons from FAD and passing them to cytochrome c, as in Scheme 1. The rate of reduction of cytochrome SCHEME 1. Cytochrome c reduction. c by an excess of reduced nNOSrd is dependent only on the rate of formation of the binary complex (and its dissociation constant) and the rate of the actual electron transfer from the FMN hydroquinone. Pseudo-first order rate constants for cytochrome c reduction were measured by mixing NADPH-or dithionite-reduced enzyme with a substoichiometric concentration of cytochrome c in a stopped-flow spectrophotometer. The rate constants, derived from fitting to single exponential functions, were linearly dependent on enzyme concentration ( Fig.  1). Second order rate constants for cytochrome c reduction were calculated in the presence and absence of NADPH and bound CaM by linear regression analysis (Table II). The rate constants are greatest for CaM-bound nNOSrd, indicating that the FMN is most accessible for reaction with cytochrome c in the CaM-bound enzyme. These values are independent of whether the enzyme was reduced initially by sodium dithionite (which was removed) or by NADPH (present in 10-fold excess). In the absence of CaM the second order rate constant for dithionitereduced nNOSrd more than halves, suggesting that the FMN is less accessible. However, in the presence of NADPH this decrease is much more dramatic, more than 30-fold, indicating a large change in FMN accessibility. A similar effect was seen when the enzyme was initially reduced by dithionite, with the NADPH being added subsequently, which rules out the possibility that a combination of partial reduction and a flavin redox potential shift might be responsible. It appears, therefore, that NADPH binding strongly inhibits cytochrome c reduction by CaM-free nNOSrd, whereas no inhibition occurs for the CaMbound enzyme, even at much higher NADPH concentrations. Fig. 2 the concentrations of NADPH and NADP ϩ are plotted against the pseudofirst order rate constant for cytochrome c reduction by 10 M CaM-free nNOSrd (prereduced by dithionite). Data were fitted to single binding site noncompetitive inhibition models to determine values for percentage inhibition and K I (inhibition constants). These data, and parameters for NAD ϩ , NADH, and 2Ј,5Ј-ADP, are shown in Table III. NADPH inhibited the reaction to the greatest extent (87%) and had a low K I (2.7 M), indicating tight binding. NADP ϩ inhibited the reaction in two phases, the first resulting in 46% inhibition caused by tight binding (2 M), the second requiring much more NADP ϩ and possibly resulting from partial oxidation of the flavins by excess NADP ϩ . This effect complicates the data, adding a degree of uncertainty to the values. NAD ϩ had a similar effect. NADH inhibited cytochrome c reduction much less than NADPH, and 2Ј,5Ј-ADP had no significant effect, consequently, no accurate K I values could be obtained for these. It appears, therefore, that the inhibition effect relies on the nicotinamide substituent and occurs whether this is reduced or oxidized. Surprisingly, the physiological substrate of the enzyme, NADPH, appears to be its most potent inhibitor. No inhibition effect was observed

Inhibition of Cytochrome c Reduction-In
Determination of second order rate constants for presteady-state cytochrome c reduction by excess nNOSrd. Shown are pseudo-first order rate constants for the reduction of 2 M cytochrome c on mixing with excess nNOSrd prereduced by excess dithionite (E), in the presence of a 10-fold excess of NADPH (q), prereduced by excess dithionite and in the presence of a 10-fold excess of NADPH (ࡗ), prereduced by excess dithionite with bound CaM and in the presence of 1 mM Ca 2ϩ (‚), or with bound CaM in the presence of 1 mM Ca 2ϩ and a 10-fold excess of NADPH (OE). Rate constants were determined from the absorbance changes at 550 nm by fitting the resultant traces to single exponential functions. Data are shown fitted to straight lines through the origin, giving the second order rate constants (Table II).
In the presence of a 10-fold excess of NADPH.

FIG. 2.
Dependence of the pre-steady-state rate constant for reduction of cytochrome c by nNOSrd on the concentration of NADPH and NADP ؉ . Pseudo-first order rate constants for the reduction of 2 M cytochrome c by 10 M nNOSrd prereduced by excess dithionite and preincubated with varying concentrations of NADPH (q) and NADP ϩ (E) were determined as described in the legend to Fig. 1. Data are shown fitted to a noncompetitive inhibition model as described under "Experimental Procedures" to determine K I values (Table III). with CaM-bound nNOSrd with any of the ligands.
Oxidation of nNOSrd by Excess Cytochrome c-The rates of cytochrome c reduction determined above are linearly dependent on the concentration of nNOSrd, suggesting that the actual rate of FMN to cytochrome c electron transfer is very rapid. In the case of CaM-bound nNOSrd, rates are measured above 600 s Ϫ1 and must occur beyond the resolution of the stopped-flow instrument. Mixing 10 M reduced nNOSrd with 130 M excess cytochrome c will therefore lead to rapid, successive electron transfers until the nNOSrd has been oxidized completely to its stable one-electron-reduced form. For example, the bimolecular reaction of 130 M cytochrome c with nNOSrd (the first electron transfer) will occur at 4,700 s Ϫ1 , based on the second order rate constant determined above. The rates of subsequent electron transfer events will depend on the rate at which reduced cytochrome c dissociates and the rate at which the second electron arrives at the FMN. This series of reactions represents part of the catalytic cycle for cytochrome c reduction as illustrated in Scheme 1. In the scheme, Steps 1-5 represent the nNOSrd catalytic cycle in terms of individual electron transfer events. The flavin oxidation experiment begins with Step 1a (the first electron transfer to cytochrome c) and ends with Step 5. As can be seen, a total of three electron equivalents are transferred from nNOSrd to cytochrome c. Fig. 3 shows stopped-flow time courses for complete nNOSrd oxidation by excess cytochrome c using CaM-bound and CaM-free enzyme reduced by either sodium dithionite or NADPH. In Fig. 3 the absorbance change observed at 550 nm caused by cytochrome c reduction has been converted to molar reducing equivalents by determining the start point from identical experiments with lower cytochrome c concentrations (in which all of the trace is captured) and by using molar absorption coefficients. The traces, therefore, indicate how quickly each of the nNOSrd electrons are transferred to cytochrome c. Traces i, ii, and iii all capture the transfer of between two and three electron equivalents to cytochrome c. This is consistent with Scheme 1. For CaM-bound nNOSrd, the first two electron equivalents are transferred within 10 ms, whether the enzyme was reduced by dithionite or NADPH. The same was observed for dithionite-reduced CaMfree nNOSrd. The third electron equivalent was transferred more slowly in a multiphasic process to complete the oxidation. It is difficult to calculate rate constants for these steps with any certainty, partly because of the sheer number of steps involved and probable heterogeneity in the enzyme sample, which may be caused by less than total CaM binding or less than full reduction. Nevertheless, exponential fitting indicates that the second electron is transferred to cytochrome c at a rate of more than 150 s Ϫ1 in all three cases and the third electron at more than 30 s Ϫ1 . It can be concluded, therefore, that CaM-bound nNOSrd is not oxidized at a significantly faster rate than CaM-free nNOS and that the interflavin electron transfer event ( Step 2) is neither rate determining nor significantly CaM-dependent. Trace iv in Fig. 3 shows successive electron equivalents being transferred from NADPH-reduced nNOSrd (CaM-free) to cytochrome c under identical conditions. This trace is clearly very different from the others. First, the absorbance change observed is much larger, corresponding to between four and five electron equivalents. The obvious cause of this is that NADPH is bound at the active site of the enzyme prior to oxidation. Apparently, NADPH binds to the reduced enzyme so tightly that a stoichiometric amount remains after gel filtration. This means that the oxidation process will require two cycles of Scheme 1 to reach completion, the first occurring with bound NADPH and the second beginning after hydride transfer from the NADPH (i.e. Step 1). Second, the rate of nNOSrd oxidation is far slower for the CaM-free NADPH-bound enzyme form; the trace is shown fitted to a single exponential function with rate constant 4.5 s Ϫ1 . The fact that the trace fits so well to a single exponential function suggests that a single kinetic event may limit the oxidation process. There is a fast phase at the start of the trace, but this is nowhere near large enough to account for a single electron equivalent. From the second order rate constant for cytochrome c reduction by NADPH-reduced nNOSrd (CaM-free) we would expect the initial electron transfer (i.e. Step 1a) to occur at 140 s Ϫ1 . This is not the case for the bulk of the enzyme sample. It appears that a small portion of the sample reacts quickly, whereas the remainder (Ͼ90%) is oxidized at 4.5 s Ϫ1 . This heterogeneity may result from incomplete NADPH binding, leading to a fraction of the enzyme behaving as in trace ii or from the slow interconversion of two very different protein conformations. Overall, it is clear that NADPH binding dramatically inhibits the rate of nNOSrd oxidation by cytochrome c and that the effect is exclusive to CaM-free enzyme.
Flavin Reduction-The nNOSrd used in these experiments was the one-electron-reduced form generated by incubating the enzyme overnight with 1 mM dithiothreitol. The high potential blue FMN semiquinone is known to be stable in the presence of excess dithiothreitol, even under aerobic conditions (e.g. during enzyme preparation). This is the native form of the reductase domain present in vivo and corresponds to species A in Scheme 1, i.e. the starting point for the catalytic cycle. Stopped-flow traces showing the reduction of nNOSrd (with and without bound CaM) by excess NADPH, monitored at 458 nm, are shown in Fig. 4. The total absorbance change (marked on Fig.  4) was determined from the visible spectrum of the enzyme before and after NADPH addition. A split time base has been used to demonstrate the relative amplitudes of the fast and slow phases observed. To aid interpretation, a model representing the kinetic events occurring during pre-steady-state reduction is shown in Scheme 2. This is similar to models proposed for the reduction of the related enzyme, mammalian cytochrome P450 reductase (40,41). The scheme contains too many different intermediates to be fitted directly to the data with any certainty, but these form two distinct isoelectronic groupings (Species 1 and 2). The formation of Species 1 occurs after the reaction of nNOSrd with an equimolar amount of NADPH. A complex equilibrium is then established involving hydride transfer from NADPH to FAD and electron transfer from FAD to FMN. Both steps are fully reversible such that, in Species 1, the flavins are only partially reduced. It is apparent from Fig.  4 that the CaM-free enzyme has two distinct phases, each comprising an approximately equal percentage of the absorbance change. For the CaM-bound enzyme, ϳ80% of the absorbance change occurs within the first 50 ms. In both cases a significant amount of the reaction occurs in the stopped-flow dead time, i.e. before observation begins. The first and most obvious interpretation is that the slow phase obtained with CaM-free nNOSrd is accelerated on CaM binding. According to Scheme 2 and consistent with the cytochrome P450 reductase models, the slow phase is dependent on NADP ϩ dissociation from Species 1, which is immediately followed by another molecule of NADPH binding. The amount of flavin reduction observed in Species 1 is dependent on the reduction potentials of the individual cofactors at the time of reaction (i.e. within this particular complex). Reduction of the flavins is only complete when the excess of NADPH present is able to force over the equilibrium established in Species 1. Thus, the slow phase represents the rate of NADP ϩ dissociation. Table IV displays rate constants determined by fitting the reduction traces to multiple exponential functions. For CaM-free nNOSrd it was found that the fast phase is itself biphasic, probably as a result of a slow internal electron transfer event (k 2 ); therefore, three rate constants were obtained. These are assigned to the three steps: k 1 , k 2 , and k 3 in Scheme 2. The first step involves rapid NADPH binding and reversible hydride transfer to FAD (k 1 ). The second step involves electron transfer from FAD to FMN at 57 s Ϫ1 (k 2 ). Finally, NADP ϩ dissociation allows full flavin reduction to occur at 5.2 s Ϫ1 (k 3 ). The CaM-bound nNOSrd trace was fitted to a double exponential function, representing rapid formation of Species 1 (k 1 and k 2 ) and subsequent NADP ϩ dissociation at 86 s Ϫ1 (k 3 ). A further slow phase was also observed comprising less than 15% of the overall absorbance change. This may result from partial reduction of the enzyme to the four-electron-reduced species or from the presence of some fully oxidized enzyme in the initial reaction mixture. Incubation of the enzyme with NADP ϩ prior to reaction with NADPH did not induce a conformationally locked enzyme form. The rate of hydride transfer was decreased by an amount consistent with normal competitive inhibition.
Single Turnover Cytochrome c Reduction-The complex nature of the above reactions causes some ambiguity in both the fitting and the assignment of the rate constants to kinetic steps. Therefore, to test the model presented in Scheme 2 further and to confirm the above results, the rate at which an NADPH molecule could pass both of its electrons to cytochrome c via nNOSrd was determined (i.e. single turnover kinetics). Reaction of 10 M nNOSrd with 2 M NADPH was performed in the presence of 100 M cytochrome c so that the rate of transfer of the two electrons generated by NADPH dehydrogenation through the enzyme could be monitored at 550 nm by their rate of arrival on cytochrome c. Stopped-flow traces for cytochrome c reduction by CaM-free and CaM-bound nNOSrd under these conditions are illustrated in Fig. 5. The trace for CaM-bound nNOSrd displays only half the expected absorbance change, indicating that the first electron is transferred from NADPH to cytochrome c in the dead time of the instrument. This confirms that both k 1 and k 2 (Scheme 2) are extremely fast. The bimolecular reaction of reduced CaM-bound nNOSrd with 100 M cytochrome c is expected to occur at more than 3,000 s Ϫ1 , according to the second order rate constant presented in Table  II. Fitting the trace to a model for two consecutive kinetic steps with equal absorption intensity indicated that the second electron was transferred at 86 s Ϫ1 (Table IV). This step involves electron transfer from the FAD semiquinone to FMN, which may be either intrinsically slower or dependent on the rate of NADP ϩ release. Both mechanisms may also contribute to the overall rate constant. For CaM-free nNOSrd, the rate of transfer of both electron equivalents to cytochrome c could be resolved. The resultant trace, shown in Fig. 5, was fitted to the same model. The first electron was found to be transferred to cytochrome c at a rate of 57 s Ϫ1 , consistent with the results of the nNOSrd reduction experiments for k 2 in Scheme 2 (Table  IV). The second electron was transferred at 19 s Ϫ1 , which is again dependent on the rate of electron transfer from the FAD semiquinone to FMN, the rate of NADP ϩ release, or both. Note that NADP ϩ release from this species is related to k 3 in Scheme 2, but not identical to it, because the enzyme is only two-electron-reduced at this stage rather than threeelectron-reduced as in Species 1. Overall, the rate constants determined are in agreement with those from the reduction experiments (Table IV) and indicate that CaM binding accelerates both electron transfer between the flavins of nNOSrd and NADP ϩ dissociation from the reduced enzyme, i.e. k 2 and k 3 in Scheme 2.  Table IV. The horizontal line indicates the predicted start point for the absorbance change. SCHEME 2. Flavin reduction.

DISCUSSION
nNOS is regulated by CaM, which binds reversibly to the interdomain linker region of the enzyme at elevated Ca 2ϩ concentrations. CaM binding is thought to induce a large structural rearrangement, reorienting the oxygenase and reductase domains with respect to each other to facilitate interdomain electron transfer. This probably involves the FMN binding subdomain interacting with the oxygenase domain near Lys-423 of rat nNOS, where the heme edge is almost exposed (5,42,43). CaM binding simultaneously activates the reductase domain, which becomes a more effective cytochrome c reductase. The isolated reductase domain also exhibits this effect, when purified with the CaM binding site intact (nNOSrd) (18,19). The rate of interdomain (FMN to heme) electron transfer in full-length nNOS is much slower than the rate of steady-state cytochrome c reduction, indicating that this step determines the overall rate of NO synthesis (44), although the electron transfer itself may be limited by conformational gating (43). The relationship between the rate of interdomain electron transfer and the rate of NO synthesis is also complicated by the tendency of the enzyme to form a kinetically trapped heme⅐NO complex at high rates of electron flux (45)(46)(47). Several unusual structural features found within the nNOS and eNOSrds appear to control the rate of electron flux in the presence and absence of CaM both within the reductase domain itself and through the holoenzyme (21). Most of these appear to be inhibitory control elements. Mutation of kinase-dependent phosphorylation sites (35,36) in rat nNOS (S1412D) and bovine eNOS (S1179D) and deletion of the autoinhibitory loops (20, 24 -27) and C-terminal extensions (29) from the same enzymes caused the cytochrome c reductase activities of the CaM-free enzymes to increase ( Table I). Deletion of the autoinhibitory loop of nNOS (25) and the C-terminal extensions of nNOS and eNOS (29) have also been shown to stimulate NO synthesis in the absence of CaM. In previous stopped-flow experiments, CaM binding was reported to increase the rate of NADPH-dependent flavin reduction in nNOS (29,36,39), eNOS (29), and nNOSrd (18). In the absence of CaM, flavin reduction was also found to be faster in the C-terminal truncation mutants of nNOS and eNOS (29) and in the S1412D and FMN-free mutants of nNOS (36,39). The apparent correlation between the steady-state and pre-steady-state data led to the conclusion that the first step, hydride transfer from NADPH to FAD, determines the overall rate of reaction. However, our experiments indicate that although the rate of flavin reduction is increased on CaM binding, this is mainly dependent on the rate of NADP ϩ dissociation from the FAD site (k 3 in Scheme 2). This interpretation of the pre-steady-state reduction kinetics ( Fig. 4 and Table II) is based on similar studies conducted on the related enzyme, cytochrome P450 reductase (40,41). The model for flavin reduction proposed by Gutierrez et al. (41) was used as the basis for Scheme 2. In this model, Species 1 is assigned to a "charge transfer" complex; however, we believe that hydride transfer has occurred prior to formation of this complex and that Species 1 is in fact an equilibrium, as shown in Scheme 2. The hydride is effectively shared between the NADP ϩ and the FAD at this point. Confirmation of this can be found in the single turnover kinetics experiments ( Fig. 5 and Table IV). These indicate that with CaM-bound nNOSrd, the first electron equivalent passes from NADPH to cytochrome c in the dead time of the stopped-flow instrument, i.e. at a rate of Ͼ600 s Ϫ1 and, in the absence of CaM, at 60 s Ϫ1 . Both of these rate constants are vastly in excess of the respective catalytic turnover rates for cytochrome c reduction, which are 60 s Ϫ1 (CaMbound) and 4.5 s Ϫ1 (CaM-free) per electron. Clearly it is impossible for catalytic turnover to be limited by either hydride transfer or electron transfer from the FAD hydroquinone to FMN, i.e. Steps 1 and 2 of Scheme 1. This interpretation is also consistent with steady-state kinetic isotope effect data reported for turnover with NADPD. Wolthers and Schimerlik (49) showed that the kinetic isotope effects on V max /K m for cytochrome c reduction are 1.19 and 1.31 in the absence and presence of CaM, respectively. These values are too close to unity for hydride transfer to be rate-determining in either case, leading the authors to speculate that binding or release of NADPH or protein conformational changes limit turnover.
The remainder of the catalytic cycle for cytochrome c reduction (Scheme 1) involves oxidation of the nNOSrd flavins. This process is illustrated by the cytochrome c reduction time courses displayed in Fig. 3. Interestingly, flavin oxidation is rapid even in the absence of CaM but is severely retarded in the CaM-free enzyme by NADPH binding. Flavin oxidation under  (Fig. 4).  Table IV. these circumstances is monophasic and fits to a single exponential of rate constant 4.5 s Ϫ1 . This is similar to the rate of steady-state cytochrome c reduction (4.5 s Ϫ1 per electron), although it includes the transfer of four electron equivalents. It seems likely, therefore, that steady-state cytochrome c reduction is limited by a flavin oxidation step rather than by reduction. Interestingly, the slow phase of CaM-free nNOSrd reduction (Fig. 3) also occurs at a similar rate (5.2 s Ϫ1 ), which has been interpreted as being the rate of NADP ϩ dissociation from the reduced enzyme. Consequently, there appears to be a link between the rate of NADP ϩ /NADPH dissociation from the CaM-free enzyme and its rate of steady-state cytochrome c reduction. It is possible that both events are limited by a similar, slow conformational change.
The pre-steady-state second order rate constants for cytochrome c reduction ( Fig. 1 and Table II) provide an indication of the accessibility of the FMN cofactor. There is unlikely to be a specific binding site for cytochrome c on the nNOSrd, but we do know that electron transfer only occurs from the FMN, which must be in its fully reduced hydroquinone redox state. The FAD is buried within the protein and is far less accessible; consequently FMN-free nNOS is unable to catalyze cytochrome c reduction (7,25,39). Changes in the second order rate constant for cytochrome c reduction are therefore likely to be caused by structural changes in the vicinity of the FMN. The x-ray crystal structure of the related enzyme, cytochrome P450 reductase, shows the FMN to be buried in the interface between the FMN and FAD domains (8). It was speculated that large scale rearrangement of the protein would be necessary to enable electron transfer from cytochrome P450 reductase to its redox partner (8,48). In CaM-free nNOSrd, NADPH binding appears to lock the FMN into a similarly inaccessible position, although no structural details are available in this case. The second order rate constant for CaM-free NADPH-bound nNOSrd is only 1.1 M Ϫ1 s Ϫ1 (Table II), 30-fold less than for the CaM-bound enzyme (35.5 M Ϫ1 s Ϫ1 ). In the absence of NADPH and CaM the FMN is still reasonably accessible (k 2nd ϭ 15.4 M Ϫ1 s Ϫ1 ), suggesting that NADPH binding is required to lock nNOSrd into the inactive conformation. In the steady state, the second order rate constants for cytochrome c reduction by nNOS (k cat /K m ) were determined to be 4.3 M Ϫ1 s Ϫ1 (CaM-free) and 100 M Ϫ1 s Ϫ1 (CaM-bound), i.e. a 23-fold increase on CaM binding (50). These experiments are conducted in the presence of NADPH, but because the enzyme is turning over it is difficult to predict what is actually bound to the enzyme during the electron transfer events. The correlation with the pre-steadystate results suggests that a substantial proportion of the CaMfree enzyme is NADPH-bound. NADP ϩ also causes inhibition of cytochrome c reduction, albeit to a lesser extent (Table III) and is even more likely to be present in the active site of the enzyme if its dissociation rate is slow (see Scheme 1). The inhibitory effect of NADP ϩ on pre-steady-state cytochrome c reduction is complicated by its ability to complex with the reduced enzyme as Species 1 (Scheme 2) and to oxidize the FAD by dissociating as NADPH. Steady-state studies indicate that NADP ϩ is a simple competitive inhibitor of NADPH during cytochrome c reduction assays, with K I values of 1.2 and 2.1 M in the absence and presence of CaM, respectively, whereas the K m values for NADPH are 0.02 and 0.25 M (50). These values suggest that NADPH binds more tightly to the oxidized CaMfree enzyme than to the CaM-bound enzyme, whereas the effect of CaM on the affinity of the oxidized enzyme for NADP ϩ appears to be less. Stopped-flow reduction experiments were also conducted with nNOSrd incubated with NADP ϩ . If the one-electron-reduced CaM-free enzyme had formed a locked, NADP ϩ -bound complex under these conditions, the rate of reaction with NADPH may have been limited by slow NADP ϩ dissociation. However, the time courses (not shown) indicate that NADP ϩ binds in a rapidly reversible manner competitively inhibiting the reaction with NADPH, consistent with the steady-state inhibition assays. The rates of dissociation of NADP ϩ and NADPH from reduced CaM-free nNOSrd appear to be much slower, such that NADP ϩ dissociation limits the reduction kinetics ( Fig. 3 and Table IV), and a stoichiometric amount of NADPH remains bound to CaM-free nNOSrd after gel filtration. Surprisingly, both NADPH and NADP ϩ are strong inhibitors of cytochrome c reduction, despite being substrate and product, respectively.
Considering the catalytic cycle for cytochrome c reduction in Scheme 1, we have established that Steps 1 and 2 are rapid for both the CaM-free and -bound enzyme forms. CaM binding accelerates the electron transfers to cytochrome c (Steps 3 and 5) and the dissociation of NADP ϩ , which will occur during either Step 3 or 4. NADPH is likely to bind to the enzyme as soon as NADP ϩ dissociates. This will ensure that either substrate or product is bound during Steps 3 and 5. From the single turnover kinetic rates, it seems likely that NADP ϩ is bound during Step 3 and NADPH during Step 5. The latter step will therefore be especially slow and is probably the rate-determining event in the oxidation of CaM-free enzyme, i.e. Fig. 3, trace iv.
The discovery that NADP/H binding controls the conformation of the reductase domain in the absence of CaM helps to explain how the various autoinhibitory modules function. Although unresolved in the recent x-ray crystal structure, the C-terminal extension of nNOS and the phosphorylation site contained within it appear to be ideally placed to interfere with both NADP/H binding and the interaction between FAD and FMN (7,8). The autoinhibitory insert contained within the FMN binding subdomain is also thought to lie in this vicinity. NADPH binding may therefore trigger the displacement of the C-terminal extension, causing the FMN binding subdomain to lock into place. Interactions formed between the inserts and the two subdomains may then stabilize this complex, shielding the FMN from reaction with cytochrome c, as suggested by Zhang et al. (7), and slowing down the rate of NADP/H dissociation. Consequently, mutagenesis of the inhibitory inserts leads to faster rates of cytochrome c reduction (Table I) and NADP ϩ dissociation in CaM-free nNOS. The latter effect is also likely to be responsible for the apparent acceleration of pre-steadystate flavin reduction in the C-terminal truncation mutants of nNOS and eNOS (29) and the nNOS phosphorylation site mutant (36). The CaM-dependent release mechanism essentially involves the destabilization of the locked conformation, probably via the direct interaction of bound CaM with the autoinhibitory inserts. Both the C-terminal extension (29) and autoinhibitory loop (20,25,27) of nNOS destabilize CaM binding during catalytic turnover, suggesting that this is the case. The reductase domain is therefore poised on a conformational seesaw with CaM acting on one side and NADPH on the other. In full-length nNOS, which has cytochrome c reductase activity and CaM dependence similar to those of nNOSrd (Table I), the same mechanism must also apply. Inaccessibility of FMN to cytochrome c implies inaccessibility to the oxygenase domain; consequently, interdomain electron transfer (i.e. from FMN to heme) is the key CaM-dependent enzyme activation step. If the CaM-free enzyme binds NADPH in vivo, as seems likely, the reductase domain will be conformationally locked and unable to transfer electrons efficiently to either the oxygenase domain or other electron acceptors. CaM binding unlocks this complex at elevated Ca 2ϩ concentrations and reorients the two domains, triggering interdomain electron transfer and NO syn-thesis. Structural relaxation of the reductase domain appears to be an important part of the CaM-dependent activation mechanism. This enables the FMN binding subdomain to interact with two different redox sites (FAD and heme) in rapid succession during the transfer of electrons. The CaM-bound enzyme is therefore unlikely to be a rigid assembly of its component domains but is reliant on dynamic flexibility.