The three nitric-oxide synthases differ in their kinetics of tetrahydrobiopterin radical formation, heme-dioxy reduction, and arginine hydroxylation.

The nitric-oxide synthases (NOSs) make nitric oxide and citrulline from l-arginine. How the bound cofactor (6R)-tetrahydrobiopterin (H4B) participates in Arg hydroxylation is a topic of interest. We demonstrated previously that H4B radical formation in the inducible NOS oxygenase domain (iNOSoxy) is kinetically coupled to the disappearance of a heme-dioxy intermediate and to Arg hydroxylation. Here we report single turnover studies that determine and compare the kinetics of these transitions in Arg hydroxylation reactions catalyzed by the oxygenase domains of endothelial and neuronal NOSs (eNOSoxy and nNOSoxy). There was a buildup of a heme-dioxy intermediate in eNOSoxy and nNOSoxy followed by a monophasic transition to ferric enzyme during the reaction. The rate of heme-dioxy decay matched the rates of H4B radical formation and Arg hydroxylation in both enzymes. The rates of H4B radical formation differed such that nNOSoxy (18 s(-1)) > iNOSoxy (11 s(-1)) > eNOSoxy (6 s(-1)), whereas the lifetimes of the resulting H4B radical followed an opposite rank order. 5MeH4B supported a three-fold faster radical formation and greater radical stability relative to H4B in both eNOSoxy and nNOSoxy. Our results indicate the following: (i) the three NOSs share a common mechanism, whereby H4B transfers an electron to the heme-dioxy intermediate. This step enables Arg hydroxylation and is rate-limiting for all subsequent steps in the hydroxylation reaction. (ii) A direct correlation exists between pterin radical stability and the speed of its formation in the three NOSs. (iii) Uncoupled NO synthesis often seen for eNOS at low H4B concentrations may be caused by the slow formation and poor stability of its H4B radical.

Nitric-oxide synthases (NOSs) 1 (EC 1.14.13.39) are flavoheme enzymes that catalyze the oxidation of L-arginine to nitric oxide (NO) and L-citrulline (1,2). The overall reaction involves two separate steps with the first being the NADPHand O 2 -dependent hydroxylation of Arg to form N -hydroxyl-Larginine (NOHA) as an enzyme-bound intermediate (Scheme 1). Bound NOHA is then oxidized to citrulline and NO in a second NADPH-and O 2 -dependent reaction. Both reactions occur within a heme-containing oxygenase domain dimer of NOS (NOSoxy) that receives electrons from two attached flavoprotein domains (3,4). NOSoxy proteins can be expressed independently and can catalyze the Arg or NOHA oxidation reactions under single turnover conditions (5,6).
Three concepts arose from our kinetic studies of the iNOSoxy Arg hydroxylation reaction: (i) formation of the pterin radical is kinetically coupled to reduction of the Fe III O 2 Ϫ intermediate and to the rate of Arg hydroxylation; (ii) the rate of pterin electron transfer is directly proportional to the stability of the resultant pterin radical; and (iii) a sufficiently fast rate of pterin electron transfer must be maintained for iNOSoxy to couple its Fe III O 2 Ϫ formation to Arg hydroxylation (18). Although it is established that an H 4 B radical also forms in neuronal and endothelial NOS (nNOS and eNOS) during their Arg hydroxylation reactions (19,20), the kinetic relationships between H 4 B radical formation, processing of the Fe III O 2 Ϫ intermediate, and product formation in these NOSs are largely unresolved. Because this knowledge is needed to understand eNOS and nNOS catalysis, we measured and compared the kinetics of heme transitions, pterin radical formation, and Arg hydroxylation in single turnover reactions catalyzed by human eNOSoxy and rat nNOSoxy enzymes that contained H 4 B or 5MeH 4 B. Protein Expression and Purification-The H 4 B-free rat nNOSoxy was prepared as described previously (16). The plasmid for human eNOSoxy was a kind gift of Anders Aberg at Astra Zeneca. The expression and purification of human eNOSoxy in the presence and absence of H 4 B was performed as described elsewhere (46). The enzyme concentration was determined from the 444 nm absorbance of the ferrous CO complex by using an extinction coefficient of 76 mM Ϫ1 cm Ϫ1 . The concentrated protein was stored in a buffer containing 50 mM EPPS, pH 7.5, 2 mM ␤-mercaptoethanol, 10% glycerol, and 0.25 M NaCl. Fulllength eNOS and nNOS were purified as described elsewhere (16,21,22). Their NO production was determined using the spectrophotometric oxyhemoglobin assay (16).

Chemicals-[U-
UV-visible Spectrometry-Conventional spectra were obtained using a Cary 100 BIO instrument (Varian, Inc) equipped with temperature control and automatic stirring.
Kinetic Experiments-Ferrous NOSoxy proteins were prepared as described previously (23). Solutions containing concentrated NOSoxy, H 4 B (or 5MeH 4 B), and Arg in a buffer (40 mM EPPS, pH 7.5, 0.125 M NaCl) were preincubated for at least 1 h at 4°C prior to making the sample anaerobic and performing the dithionite titration. To initiate the reaction, the anaerobic ferrous hemeprotein solution was mixed with an O 2 -containing buffer at 10°C. The reactions were then monitored using rapid scanning stopped-flow spectroscopy, rapid quench with HPLC product analysis, and rapid freeze with EPR measurements as described previously (12). For rapid quench experiments, the reaction was initiated by mixing an anaerobic solution of 2-5 M ferrous hemeprotein containing 10 M [C 14 ]Arg and H 4 B (5MeH 4 B) with an equal volume of air-saturated buffer (40 mM EPPS, pH 7.5, 0.125 M NaCl) that contained 1 mM Arg to ensure a single turnover condition with respect to the radiolabeled substrate (24). The mixture was then aged in the instrument for various times followed by rapid quenching with 1 N HCl solution containing 1 mM citrulline and 1 mM NOHA. Each reacted sample was collected into a vial from the sample loop, and 5 l of isopropane was added. The resulting solution was vigorously vortexed for 30 min to release the substrate. The work up and amino acid determination was the same as reported previously (12). For the EPR measurements, an anaerobic buffer solution containing substrate, pterin, and ϳ400 M ferrous NOS protein was rapidly mixed at 10°C with an equal amount of O 2 -saturated buffer (O 2 concentration ϳ2.2 mM at 0°C (10)). The mixture was aged for various times, and then ejected into a rapid freezing solution (12) housed in a bath apparatus from Update Instruments, Inc. (Madison, WI). The samples were then kept at liquid nitrogen temperature until measurement. EPR spectra of the frozen samples were recorded in a Bruker ER300 spectrometer equipped with an ER 035 NMR gauss meter and a Hewlett-Packard 5352B microwave power controller. Temperature was controlled using Oxford Instruments ESR 900 continuous flow liquid helium cryostat and ITC4 temperature controller. All spectra were recorded at 150 K using a microwave power of 2 milliwatts, a frequency of 9.5 GHz, a modulation amplitude of 10 G, and a modulation frequency of 100 kHz.
Determining the Redox State of H 4 B after Arg Hydroxylation-This step was performed as described previously (13). Briefly, 100 M NOSoxy was incubated with 1 mM H 4 B (or 5MeH 4 B) and 10 mM Arg followed by rapid filtration to remove unbound H 4 B. Additional Arg was then added to the filtrate to compensate for its loss. The NOSoxy sample was then made anaerobic, reduced with dithionite, and mixed with O 2 -saturated buffer to initiate the reaction. The mixture was allowed to fully react for 1-2 min, aliquots were transferred to the analytical alkaline and acidic iodine/iodide oxidation solutions, and the content of H 4 B and its oxidation products were determined by HPLC with fluorescence detection.
Data Analysis-At least two different batches of each NOSoxy protein were used for the rapid kinetic measurements. Light absorbance spectra of heme species obtained from stopped-flow spectrometry were analyzed using the global analysis software SPECFIT/32 version 3.0 (Spectrum Software Associates, Marlborough, MA). In all cases, the fitting of absorbance cross-sections at specific wavelengths was also performed to check the accuracy. Radical formation and decay rates were determined by fitting using OriginPro 7.5 software (OriginLab Cooperation, Northampton, MA) as described previously (12).

RESULTS
We first used UV-visible spectroscopy to confirm 5MeH 4 B binding in our nNOSoxy and eNOSoxy enzymes. In the absence of 5MeH 4 B and Arg, both proteins had a heme Soret absorbance peak near 420 nm, consistent with previous results (14,16,25). Adding 5MeH 4 B at concentrations between 200 M and 1 mM caused the Soret peak to shift to 400 nm, consistent with 5MeH 4 B binding (17). The addition of 0.2 mM Arg to these samples caused a further shift in the Soret peak to 390 nm (data not shown). This demonstrated that 5MeH 4 B and/or Arg were able to stabilize the ferric heme in a high spin state, which is essential for catalytic activity. Using full-length NOS enzymes, we also found that 5MeH 4 B supported NO synthesis from Arg by eNOS and nNOS as determined by the spectrophotometric oxyhemoglobin assay (activities were 16.7 Ϯ 1.6 and 55 Ϯ 3.1 NO min Ϫ1 , respectively). These specific activities were equivalent to the activities supported by H 4 B.
We next ran single catalytic turnover reactions using eNOSoxy and nNOSoxy samples that contained Arg plus H 4 B (or 5MeH 4 B). The enzymes were reduced to their ferrous forms with dithionite under anaerobic conditions and then were rapidly mixed at 10°C with an O 2 -containing buffer to initiate the reaction. We monitored the reactions by rapid scanning UVvisible spectroscopy to follow the kinetics of heme transitions, by rapid freezing EPR spectroscopy to follow the kinetics of biopterin radical buildup and decay, and by rapid quenching and HPLC to follow the kinetics of Arg hydroxylation.
Our optical data for the eNOSoxy/H 4 B single turnover reaction are shown in Fig. 1. Global analysis indicated a best fit to an A to B to C kinetic model, such as we had observed previously in single turnover reactions of iNOSoxy and nNOSoxy (12,26). The spectra of the three species as determined by the global analysis are shown in Fig. 1A. The calculated spectra matched well with spectral traces that were collected at specific points during the eNOSoxy single turnover reaction (Fig. 1B). The Soret peak positions are listed in Table I (12), nNOSoxy (26), and full-length eNOS (32) under similar reaction conditions. Fig.  1C shows the change in concentration for each of the three species during the single turnover reaction. The rates of Fe III O 2 Ϫ formation and decay as determined by the global analysis were 55.6 and 6.7 s Ϫ1 , respectively (Table II). These rates match well with the rates estimated by fitting the absorbance change at 429 nm to the A 3 B 3 C reaction model (k formation ϭ 56.7 s Ϫ1 , k decay ϭ 6.5 s Ϫ1 ) (Fig. 1D).
We next ran single turnover reactions using eNOSoxy that contained 5MeH 4 B in place of H 4 B. The optical spectra of the three heme species discerned from global analysis are shown in Fig. 2A. In general, the spectra match those obtained using H 4 B, except for some modest shifting in the Soret positions ( Table I). The calculated spectra match the actual absorbance traces that were collected at specific time points during the reaction, except for the Fe III O 2 Ϫ intermediate (Fig. 2B), whose Soret position in the collected trace (427 nm) is shifted 3 nm relative to the Soret peak in the global analysis spectrum (430 nm). However, it was estimated that the spectrum collected at 38.8 ms (Fig. 2B) represents a mixture of 53% Fe III O 2 Ϫ , 26% ferrous, and 21% ferric species. When deconvolution was done to take this into account, it eliminated the apparent blue shift in the collected spectrum such that the Soret peak of the Fe III O 2 Ϫ species now matched the peak observed in the global analysis spectrum (data not shown). The Soret peak positions for the three eNOSoxy species in the 5MeH 4 B reaction ( Ϫ conversion to ferric enzyme was ϳ3-fold faster in the 5MeH 4 B-bound eNOSoxy reaction (Table II). The absorbance at 430 nm versus time of the reaction is shown in Fig.  2D. A fit of this data to the A 3 B 3 C kinetic model is shown in Fig. 2D and confirms that the rate of Fe III O 2 Ϫ disappearance was increased in the 5MeH 4 B-containing eNOSoxy reaction (k ϭ 26 s Ϫ1 ).
We then performed identical experiments using H 4 B-or 5MeH4B-bound nNOSoxy. The optical spectra and appearance/ disappearance kinetics of the three heme species obtained in our present study with the H 4 B-bound enzyme were very similar to data we reported previously (26) and are summarized in Tables I and II (Table II).
We next studied the kinetics of biopterin radical formation and decay in the Arg single turnover reactions catalyzed by eNOSoxy and nNOSoxy. Biopterin radical signals that built up during the reactions catalyzed by H 4 B-or 5MeH 4 B-bound eNOSoxy are shown in Fig. 3. Similar biopterin radical signals built up in the reactions catalyzed by nNOSoxy (data not shown) and in iNOSoxy (17). The biopterin radicals were centered at g ϭ 2.0 and had a peak to trough width of ϳ40 G. There was an additional hyperfine structure present in the radical signals from the 5MeH 4 B-bound eNOSoxy (Fig. 3) and nNOSoxy (data not shown) reactions. These EPR spectral characteristics are highly similar to the H 4 B and 5MeH 4 B radicals that were found to form in iNOSoxy during Arg single turnover  reactions run under identical conditions (17,28). Lowering the EPR measurement temperature to 4 K did not alter the line shape of the biopterin radical signals, and doubling the 5MeH 4 B concentration in the reactions did not increase the intensity of radical signal accumulation (data not shown). We detected no EPR signals in control reactions that contained biopterin-free NOSoxy enzymes run under identical conditions. Together, these data confirm that bound H 4 B and 5MeH 4 B are oxidized to their radicals in eNOSoxy and nNOSoxy during the Arg single turnover reaction, which is consistent with previous results (17,19,20,28). Fig. 4 compares the buildup and decay kinetics of the H 4 B and 5MeH 4 B radical signals during Arg single turnover reactions catalyzed by eNOSoxy (upper panel) or nNOSoxy (lower panel). The radical intensities that we observed at each time point were fitted to an A 3 B 3 C reaction model to determine their formation and decay rates, incorporating an initial lag time to account for O 2 binding to the ferrous NOS heme. The fitting gave rates of H 4 B radical formation and decay in the eNOSoxy reaction of 6.5 and 3.8 s Ϫ1 , respectively, and indicate that ϳ70% of the bound H 4 B participated in radical formation (per heme) during the reaction (Table II). Similarly, the fitting gave rates of 5MeH 4 B radical formation and decay in the eNOSoxy reaction of 19 and 0.7 s Ϫ1 , respectively. Thus, the 5MeH 4 B radical formed ϳ3-fold faster than the H 4 B radical during Arg hydroxylation by eNOSoxy, and it decayed 5-fold more slowly than the H 4 B radical (Table II). For the Arg reaction of nNOSoxy, the fitting gave rates of H 4 B radical forma-tion and decay of 20 and 0.6 s Ϫ1 , respectively. Again, 5MeH 4 B supported a faster radical formation (51 s Ϫ1 ) and a slower radical decay (0.2 s Ϫ1 ) in nNOSoxy when compared with H 4 B (Table II).
We then used rapid quench and HPLC methods to determine the rate of Arg hydroxylation in companion single turnover reactions catalyzed by eNOSoxy and nNOSoxy. These experiments were designed such that a substoichiometric amount (with respect to heme) of [ 14 C]Arg is converted to [ 14 C]NOHA during the single turnover reaction to precisely measure the rate of Arg hydroxylation (12). In all cases, [ 14 C]citrulline formation was minimal (Ͻ5%) compared with the content of [ 14 C]NOHA. The kinetic data are summarized in Table II. The rate of Arg hydroxylation in H 4 B-bound eNOSoxy was slower compared with H 4 B-bound nNOSoxy (5.8 versus 18.2 s Ϫ1 , respectively). Substituting eNOSoxy with 5MeH 4 B increased its rate of Arg hydroxylation by 4-fold (Fig. 5). Similarly, 5MeH 4 B substitution increased the rate of Arg hydroxylation 3-fold in nNOSoxy (Table II).
Finally, we determined the redox state of H 4 B after the radical signal decayed in the single turnover reactions of eNOSoxy and nNOSoxy. The experiments were performed in the same fashion as in our recent report on iNOSoxy using alkaline and acidic iodine/iodide oxidation (13). We found that only 32.3 Ϯ 7.2% and 25.6 Ϯ 6.1% of the biopterin was fully reduced (H 4 B) following Arg hydroxylation by eNOSoxy and nNOSoxy, respectively (data not shown). These values are consistent with our previous results and demonstrate that the H 4 B  2. Light absorbance spectra and kinetics of the heme species observed in the Arg hydroxylation reaction catalyzed by 5MeH 4 B-bound eNOSoxy. The reaction was identical to that described in Fig. 1  radical signal decay following Arg hydroxylation is because of an oxidative decay of the radical. 3

DISCUSSION
NOSs are the only enzymes known to support single-electron transitions of a bound H 4 B cofactor (29,30). We now know the kinetics of H 4 B redox transitions and the times when transitions occur relative to other key catalytic events during Arg hydroxylation in all three NOSs. Combined, the evidence supports an identical kinetic mechanism for Arg hydroxylation in the three NOSs (Scheme 2). This common mechanism incorporates the following tenets: (i) The H 4 B radical forms only after O 2 binds to the ferrous heme in each NOS. As discussed previously (12,23,27,28), there must be some property of the NOS Fe III O 2 Ϫ species that promotes the one electron oxidation of enzyme-bound H 4 B, which otherwise does not oxidize at an appreciable rate when the NOS heme is in the ferric state or present as a ferric heme-NO complex. ( (10,11,20,29). Thus, reaction of such a heme-oxy species with Arg and all of the subsequent chemical transformations that generate products must be as fast or faster than H 4 B reduction of the Ϫ intermediate in all three NOSs. We found that 5MeH 4 B transfers an electron to the Fe III O 2 Ϫ intermediate faster than H 4 B in eNOSoxy and nNOSoxy, and this was associated with a faster Arg hydroxylation in their single turnover reactions. We had previously observed identical behavior for 5MeH 4 B in iNOSoxy in single turnover reactions (17). In that case, the crystal structure data show that there is very little change in the protein or pterin struc- tures when 5MeH 4 B is bound in place of H 4 B (17). Thus, the driving force for faster radical formation seems to rely on the distinct electronic properties of 5MeH 4 B itself, namely its ability to better stabilize the radical through its 5-Me group (29). Apparently, a similar situation exists in eNOSoxy and nNOSoxy, suggesting that the three NOSs have common structural and electronic features that govern their pterin radical formation and stability. This is consistent with their protein sequences displaying good conservation of the residues that surround the H 4 B ring (31).
Interestingly, the rates of H 4 B radical formation and decay differed among the three NOSoxy enzymes during their Arg hydroxylation reactions. The H 4 B radical formation rates displayed a rank order of nNOSoxy (20 s Ϫ1 ) Ͼ iNOSoxy (11 s Ϫ1 ) Ͼ eNOSoxy (6.5 s Ϫ1 ) in the single turnover reactions run at 10°C, whereas the rank order of their radical decay rates was essentially reversed: eNOSoxy (3.8 s Ϫ1 ) Ͼ iNOSoxy (0.71 s Ϫ1 ) ϭ nNOSoxy (0.6 s Ϫ1 ). The rank order of these kinetic relationships held among different batches of NOSoxy enzymes, suggesting that it is an intrinsic difference between the three NOS isoenzymes. The physical basis for these kinetic differences is not readily apparent from the crystal structures or biochemical data and so will require further investigation. In any case, the measured rates and their differences have some important implications. For example, they show that there is a good correlation between the speed of H 4 B radical formation and the stability of the resulting H 4 B radical among all three NOSs (Fig. 7). We had previously proposed this relationship based on data obtained with iN-OSoxy and some mutants (17). The seemingly broad nature of the relationship implies that stabilization of the H 4 B radical is a common, primary strategy by which NOSs regulate the speed of their H 4 B electron transfer reactions.
The speed of H 4 B radical formation also helps to determine whether the Fe III O 2 Ϫ intermediate will receive an electron from H 4 B or from the NOS flavoprotein domain. Our kinetic measurements indicate that the rates of H 4 B radical formation in the three NOSs are 5-60-fold faster than the rates of flavoprotein-catalyzed electron transfer, as judged by the measured rates of NOS ferric heme reduction at the same temperature (10°C; 4 s Ϫ1 , 1 s Ϫ1 , and 0.1 s Ϫ1 in nNOS, iNOS, and eNOS, respectively) (22,(32)(33)(34) (36). In any case, the sensitivity of eNOS and nNOS coupling to the rate of H 4 B electron transfer can now be examined in detail.
The decay rate of the H 4 B radical may also be important. In our single turnover Arg hydroxylation reactions, the H 4 B radical decay simply reflects its stability within the enzyme after the reaction has ended. Our previous work had shown that SCHEME 2 FIG. 6. Relationship between the rates of biopterin radical formation and Arg hydroxylation in the three NOSoxy enzymes. Data are from reactions run at 10°C using eNOSoxy (squares), nNOSoxy (triangles), and iNOSoxy (circles). The iNOSoxy data are from Refs. 17 and 18. A least squares line of best fit is shown with a slope of 0.89 Ϯ 0.06 (r ϭ 0.989).

FIG. 7.
Relationship between the rate of biopterin radical formation and the half-life of the resulting biopterin radical in Arg hydroxylation reactions catalyzed by the three NOSoxy enzymes. Data are from Arg hydroxylation reactions run at 10°C using eNOSoxy (squares), nNOSoxy (triangles), and iNOSoxy (circles). The iNOSoxy data are from Refs. 17 and 18. A least squares line of best fit is shown with a slope of 13.5 Ϯ 2.8 (r ϭ 0.881).