Mechanism of Superoxide Generation by Neuronal Nitric-oxide Synthase*

Neuronal nitric-oxide synthase (NOS I) in the absence of l-arginine has previously been shown to generate superoxide (O·̄2) (Pou, S., Pou, W. S., Bredt, D. S., Snyder, S. H., and Rosen, G. M. (1992) J. Biol. Chem. 267, 24173–24176). In the presence ofl-arginine, NOS I produces nitric oxide (NO⋅). Yet the competition between O2 and l-arginine for electrons, and by implication formation of O·̄2, has until recently remained undefined. Herein, we investigated this relationship, observing O·̄2 generation even at saturating levels ofl-arginine. Of interest was the finding that the frequently used NOS inhibitor N G-monomethyll-arginine enhanced O·̄2 production in the presence of l-arginine because this antagonist attenuated NO⋅formation. Whereas diphenyliodonium chloride inhibited O·̄2, blockers of heme such as NaCN, 1-phenylimidazole, and imidazole likewise prevented the formation of O·̄2 at concentrations that inhibited NO⋅ formation from l-arginine. Taken together these data demonstrate that NOS I generates O·̄2 and the formation of this free radical occurs at the heme domain.


of this free radical occurs at the heme domain.
At a time before the physiologic properties of endotheliumderived relaxation factor were associated with nitric oxide (NO ⅐ ) 1 (1)(2)(3), it was found that this free radical activated soluble guanylate cyclase from crude homogenates of brain tissue (4). The significance of this observation would, surprisingly, remain dormant for nearly a decade, even though it was known that L-arginine was the endogenous activator of this enzyme (5). With the purification and characterization of a unique monooxygenase, NOS I, capable of oxidizing L-arginine to Lcitrulline, and NO ⅐ (6, 7), a new class of small molecules (8) acting as transient second messengers in the brain was discovered (9). The versatility of this free radical in controlling a myriad of brain functions will undoubtedly result in new and provocative findings. Of special interest will be research that can distinguish NO ⅐ from NOS I-secreting neurons versus NO ⅐ from NOS III-containing endothelial cells and NO ⅐ from NOS II-stimulated microglial cells. Until the recent development of specific antagonists for each of the NOS isozymes (10 -12), it has been difficult to address this question, which is of particular significance when one considers, as described in this article, that NOS I, NOS II, and NOS III may produce NO ⅐ and O 2 . under differing cell conditions.
Nitric-oxide synthase is known to catalyze the production of NO ⅐ from L-arginine (13,14). In the absence of substrate, we have previously demonstrated that purified NOS I can use O 2 as the terminal electron acceptor, generating O 2 . (15). These findings have been confirmed because O 2 . has been spintrapped in L-arginine-depleted NOS I-transfected human kidney cells (16). During the course of our earlier studies (15), we noted to our surprise that purified NOS I appeared to produce O 2 . even in the presence of L-arginine. The current study, therefore, explores this phenomenon in depth. Herein, we demonstrate that NOS I, like NOS II (17) and NOS III (18 -20), can generate O 2 . and NO ⅐ despite saturating levels of substrate.
However, unlike NOS II, the heme of NOS I is the locus for the production of both free radicals. Finally, we discuss the implications of our findings; particularly relevant is the ability of L-NMMA to enhance NOS I-derived O 2 . even in the presence of saturating concentrations of L-arginine.
NOS I Purification-NOS I-transfected kidney 293 cells were cultured in Dulbecco's modified Eagle's medium:nutrient mixture F-12 containing 10% fetal calf serum, penicillin G (100 units/ml), and streptomycin (100 g/ml). NOS I was purified from these cells by the method of Bredt and Snyder (6). Briefly, cells were removed from the culture flasks and washed three times with phosphate-buffered saline via centrifugation. The pellet was resuspended in buffer containing phenylmethylsulfonyl fluoride (4 mg/ml) and homogenized with a Polytron (Brinkmann Instruments, model PCU-2 at setting 2 for 10 s). The remaining mixture was centrifuged at 15,000 rpm for 20 min to separate unbroken cells, and the supernatant was applied to a 2Ј,5Ј-ADP-Sepharose affinity column. After washing the column three times with 0.45 M NaCl and standard buffer, NOS was eluted with standard buffer containing 10 mM NADPH. Excess NADPH was removed by washing * This work was supported in part by Grants AG-14829 and CA-69538 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  Estimation of NOS I Activity-NOS I activity was assayed by measuring the formation of L-[ 14 C]citrulline from L-[ 14 C]arginine as described previously (23). Briefly, purified NOS I was added into a reaction mixture containing L-[ 14 C]arginine (0.6 Ci/ml), L-arginine (1 mM), NADPH (1 mM), calmodulin (100 units/ml; 23.6 g/ml), free calcium ions (CaCl 2 , 1 M; calculated as described in Ref. 24), and standard phosphate buffer at pH 7.4 (final volume ϭ 0.150 ml). After incubating at room temperature for exactly 10 min, the reaction was terminated by adding HEPES (20 mM, 2 ml) containing EDTA (2 mM, pH 5.5). L-[ 14 C]Citrulline was separated by passage through an 0.5-ml column of Dowex-50-X8 cation exchange, and radioactivity was counted using a Beckman ␤ counter.
Spin Trapping and EPR Spectroscopy-Spin trapping experiments with purified NOS I were conducted by mixing all the components described in the figure legends to a final volume of 0.25 ml. The experiment was initiated by adding freshly purified NOS I. Reaction mixtures were then transferred to a flat quartz cell and fitted into the cavity of the EPR spectrometer (Varian Associates, model E-109), and spectra were recorded at room temperature 1 min after addition of the enzyme. Instrumentation settings are presented in the figure legends.
Superoxide Detection-The assessment of the effect of L-NAME on the ability of xanthine/xanthine oxidase to produce O 2 . was evaluated as described previously (25). Briefly, xanthine oxidase was added to a solution containing xanthine (400 M) and ferricytochrome c (80 M) such that the rate of O 2 . formation, measured as the SOD-inhibitable reduction of ferricytochrome c at 550 nm (26) was 1 M/min. The effect of various concentrations of L-NAME on the rate of O 2 . generation was estimated. NADPH Oxidation-The rate of NADPH oxidation by NOS I was determined spectrophotometrically at 340 nm (⑀ ϭ 6.2 ϫ 10 3 M Ϫ1 cm Ϫ1 ).

RESULTS AND DISCUSSION
Before determining the ability of NOS I to generate O 2 . , our initial series of studies were devoted to determining the K m and the V max for NOS I, because these constants will help dictate future experimental designs. Based on these studies, the K m and the V max were found to be 3.47 M and 216 nmol min Ϫ1 mg Ϫ1 , respectively, well within the ranges (K m of 2-4.3 M and V max of 74 -3400 nmol min Ϫ1 mg Ϫ1 ) reported by others (27).
In the absence of L-arginine, NOS I has previously been shown to generate O 2 . (15,16), whereas at high concentrations of L-arginine (1 mM), complete inhibition of O 2 . production had been noted (15). We therefore examined the effect that intermediate concentrations of L-arginine would have on the ability of NOS I to generate O 2 . . As shown in Fig. 1 production (18,19). The initial step in the generation of NO ⅐ by NOS is the transport of electrons from NADPH to the oxidized flavin, FAD, resulting in FADH 2 , after abstraction of a proton from the surrounding milieu. Disproportionation between the flavins leads to FADH ⅐ /FMNH ⅐ . The electron donation from FMNH ⅐ to . (15,(17)(18)(19). When L-arginine is present, however, there is a binding of the guanidino nitrogen in an ordered position near the heme (30), which allows the oxidation of L-arginine to proceed. Based on a K m of 3.47 M for the NOS I oxidation of L-arginine and on data presented in Fig. 1, NO ⅐ and O 2 . are both generated. The rate of each free radical produced, however, cannot accurately be estimated by spin trapping. Fig. 1 indicates that NOS I at L-arginine concentration around the K m is capable of producing O 2 . at about 60% of the rate of that generated in the absence of L-arginine. For NOS II, electron transport from FMNH ⅐ to Fe 3ϩ appears not to be so tightly coupled, because some leakage from the flavin domain results in the formation of O 2 . , even in the presence of 1 mM L-arginine (17).
For NOS I, the site of O 2 . generation has not yet been defined.
We suggest, however, that during the oxidation of L-arginine to L-citrulline and NO ⅐ , direct competition with O 2 results in O 2 .
formation. At a fixed concentration of O 2 , the ratio of NO ⅐ and O 2 . is dependent, therefore, on the concentration of L-arginine.
To explore the mechanism of O 2 . generation by NOS I, it is important to determine how electrons are transferred through the enzyme. Thus, we undertook a series of inhibition experiments exploring the effects of two well known inhibitors of NOS, L-NAME and L-NMMA, on O 2 . production. Before conducting these experiments, we estimated the capacity of these NOS inhibitors to block L-citrulline formation and, by implication,  Fig. 2. L-NMMA was found to be slightly more potent (EC 50 ϭ 5 M) than L-NAME (EC 50 ϭ 10 M). For NOS III, L-NAME has been reported to be a more effective inhibitor of this isozyme than L-NMMA (31). With these data in mind, we then defined the ability of L-NAME and L-NMMA to prevent O 2 . production by NOS I. Fig. 3 shows the effect of increased concentrations of L-NAME on the ability of NOS I to generate O 2 . . Similar to L-arginine, L-NAME antagonized the spin trapping of this free radical by DMPO in a dose-dependent manner, with an EC 50 ϭ 40 M. In contrast, L-NMMA did not appreciably depress the formation of O 2 . by NOS I even at concentrations as high as 10 mM (Fig. 3). Consistent with these findings, the rate of NADPH oxidation by NOS I was inhibited by Ͼ70% as the concentration of L-NAME reached 100 M, whereas the rate of NADPH oxidation remained constant at about 50% of control with increasing concentration of L-NMMA up to 10 mM (Fig. 4). These results suggested that L-NAME, but not L-NMMA, inhibited the formation of O 2 . by impeding the electron transport to O 2 .
To confirm this hypothesis, the effect of L-NAME (1 mM) on the xanthine/xanthine oxidase production of O 2 . , as measured by the SOD-inhibitable reduction of cytochrome c, was assessed. Within experimental error, this rate was unchanged. These data indicate that L-NAME did not scavenge O 2 . but rather that L-NAME acted specifically on NOS I, inhibiting generation of this free radical. Because L-NMMA is a potent antagonist of NOS-generated NO ⅐ , we explored whether L-NMMA could block the ability of L-arginine to inhibit NOS I production of O 2 . .
These findings are presented in Fig. 5. As expected, L-arginine (100 M) almost completely inhibited the NOS I formation of O 2 . (Fig. 5B), which confirmed earlier studies (15,32). Surprisingly, L-NMMA, in a dose-dependent manner, reversed the inhibitory properties of L-arginine (Fig. 5, C-E), almost reaching control values, in the absence of L-arginine, at 1 mM (Fig. 5, E and F). Our findings support the theory that L-NAME antagonizes the transfer of electrons to either L-arginine or O 2 , whereas L-NMMA prevents the oxidation of L-arginine by competing for the same binding site on the enzyme (33). Although NO ⅐ production is inhibited by the presence of L-NMMA, NOS I still has the capacity to transfer electrons from NADPH to O 2 ( Fig. 4 and 5).  7B). In contrast, DPI (10 M) did not inhibit the reaction of DMPO with O 2 . generated from FMN/NADPH (Fig. 8D). Even though these data indicate that the electron flow is through the flavin domain, these experiments cannot establish whether O 2 . is produced solely at this site or at the heme domain. To further address this query, we investigated the effect of NaCN and two imidazoles (known to inhibit cytochrome P-450 and NOS (37,38) by blocking the heme site) on the generation of this free radical. First, however, we had to demonstrate that NaCN, imidazole, and 1-phenylimidazole inhibited the metabolism of L-arginine to L-citrulline by NOS I. generation were revealed, then the heme was most likely the site of this free radical formation. At 1 mM NaCN, there was a minimal decrease in the peak height of DMPO-OOH (Fig. 7C), whereas at 10 mM NaCN inhibition of O 2 . was nearly complete (Fig. 7D). In an independent series of experiments, NaCN at 10 mM was found to inhibit (by 50%) the spin trapping of O 2 . from the model O 2 . generating system consisting of FMN/NADPH (Fig. 8E). Because NaCN is known to react with DMPO (21), thereby decreasing the effective concentration of DMPO available to react with O 2 . , the findings in Fig. 8E can only suggest that the heme domain in NOS I was the site of O 2 . production.
Alternative antagonists for NOS were sought. Recently, imidazoles have been reported to block the heme site of NOS (38). Thus, we investigated the ability of imidazole and 1-phenylimidazole to attenuate O 2 . formation by NOS I. As shown in Fig. 7, 1-phenylimidazole (10 mM) and imidazole (10 mM) inhibited the formation of DMPO-OOH by almost 50 (Fig. 7E) and 80% (Fig.  7F), respectively, whereas, unlike NaCN, they had no effect on the spin trapping of this free radical from the O 2 .
-generating system of FMN/NADPH (Fig. 8, B and C). Data gathered from inhibitory experiments with the antagonists 1-phenylimidazole, imidazole, and NaCN support the heme domain as the site of O 2 . production by NOS I, similar to the findings for cytochrome P-450 (35) (Fig. 9). For NOS II, it appears that the reductase domain is not as tightly coupled to the heme domain as seen in NOS I and cytochrome P-450. This weak coupling allows some electron leakage to O 2 , generating O 2 . , even though sufficient electron flow to the heme permits a Finally, we investigated the role that BH 4 might play in regulating production of O 2 . by NOS I in the absence of Larginine. Tetrahydrobiopterin, at 10 and 100 M, inhibited O 2 . secretion by NOS I by Ͼ90%. As shown in Fig. 10, B and C, the EPR spectra are characteristic of DMPO-OH and to a lesser extent of DMPO-OOH. The source of DMPO-OH appears to be BH 4 . At 1 mM BH 4 (Fig. 10D), we were still able to observe some DMPO-OOH, even though DMPO-OH dominated the EPR scan. At 10 mM BH 4 , DMPO-OH (Fig. 10E) was the only spin-

FIG. 7. Effect of NOS inhibitors on O 2 . generation by purified NOS I. Scan
A was obtained in the absence of inhibitors. Scans B-F were obtained in the presence of DPI (10 M), NaCN (1 mM), NaCN (10 mM), 1-phenylimidazole (10 mM), and imidazole (10 mM), respectively. Microwave power was 20 megawatts, modulation frequency was 100 kHz with an amplitude of 1 G, sweep time was 12.5 G/min, response time was 1 s, and receiver gain was 6.3 ϫ 10 4 .
trapped adduct recorded. These data suggest that at low concentrations (1- Although the implications of our findings have yet to be fully realized, two recent publications may shed some light on the importance of our observations. First, during anoxia/reoxygenation of cardiomyocytes, generation of O 2 . was found to be markedly enhanced when L-NMMA was included (54), which supports the data presented in Fig. 5. Second, elevated levels of MnSOD were an essential element of viable NOS I containing neurons exposed to N-methyl-D-aspartate (NMDA) (55). Even though the source of O 2 . in these studies (54,55) was not identified, NOS must certainly be considered a possible contributor to the origin of this free radical. Scans B-E were obtained in the presence of BH 4 with increasing concentrations from 10 M, 100 M, 1 mM, and 10 mM, respectively. Microwave power was 20 megawatts, modulation frequency was 100 kHz with an amplitude of 1 G, sweep time was 12.5 G/min, response time was 1 s, and receiver gain was 8 ϫ 10 4 for scans A-D and 2 ϫ 10 4 for scan E.