Active site topologies and cofactor-mediated conformational changes of nitric-oxide synthases.

The active site topologies of neuronal (nNOS), endothelial (eNOS), and inducible (iNOS) nitric-oxide synthases heterologously expressed in Escherichia coli have been examined using three aryldiazene (Ar-N=NH) probes. The topological information derives from (a) the rate and extent of aryl-iron complex formation in the presence and absence of tetrahydrobiopterin (H4B), Ca2+-dependent calmodulin (CaM), and L-arginine, and (b) the N-phenylprotoporphyrin IX regioisomer ratios obtained upon migration of the phenyl of the phenyl-iron complex to the heme nitrogen atoms. The N-phenylprotoporphyrin ratios indicate that the three NOS isoforms have related active site topologies with unencumbered space above all four pyrrole rings but particularly above pyrrole ring D. H4B binds directly above the heme pyrrole ring D or causes a conformational change that constricts that region, because H4B markedly decreases phenyl migration to pyrrole ring D. Small CaM-dependent changes in the nNOS N-phenylporphyrin isomer pattern are consistent with a conformational link between the CaM and heme sites in this protein. The ceiling height directly above the heme iron atom differs among the isoforms and is lower than in the P450 enzymes because only nNOS and iNOS react with 2-naphthyldiazene, and none of the isoforms reacts with p-biphenyldiazene. L-Arg blocks access to the heme iron atom in all three NOS isoforms and nearly suppresses the phenyldiazene reaction. The data indicate that topological differences, including differences in the size of the active site, are superimposed on the structural similarities among the NOS active sites.

require heme, FMN, FAD, calmodulin (CaM), and tetrahydrobiopterin (H 4 B) as cofactors and utilize NADPH and O 2 as co-substrates. The NOS are modular hemoproteins in which a heme domain is linked via a CaM-binding sequence to a twoflavin domain (6 -8). During catalytic turnover, the flavin domain transfers NADPH-derived reducing equivalents to the prosthetic heme group. H 4 B is essential for NO synthesis, but its role in the function of the protein remains uncertain (9).
One salient difference among the NOS isoforms is their dependence on CaM. nNOS and eNOS bind CaM reversibly in a Ca 2ϩ -dependent manner, and their activity is regulated by the local Ca 2ϩ concentration (10,11). In contrast, iNOS binds CaM essentially irreversibly, and its activity is regulated not by changes in the Ca 2ϩ concentration but by changes in the rate at which the protein is synthesized (12). nNOS and eNOS, the two Ca 2ϩ -regulated isoforms, differ in that nNOS has an additional N-terminal domain that may be involved in subcellular targeting (13), and eNOS is membrane-bound due to the presence of myristoylation and palmitoylation sites (14,15).
Little information is available on the active sites topologies of the NOS isoforms and on the structural changes that may accompany the binding of cofactors. Studies with diverse hemoproteins have demonstrated that those with open active site architectures react with aryldiazenes (Ar-NϭNH) to give -bonded aryl-iron complexes (16). The nature of the phenyliron complex is clearly established by crystal structures of the myoglobin and P450 cam complexes (17,18). The aryl-iron complexes are usually stable, but if the trans-iron ligand is a thiolate, ferricyanide oxidation of the aryl-iron complex results in migration of the aryl group from the iron to one of the heme nitrogen atoms. The resulting N-arylporphyrins can be separated, and the four possible regioisomers can be quantitated by HPLC (18). As the aryl shift occurs within the intact active site, the isomer ratio provides direct information on the active site topology because steric obstruction in one region and not in another results in preferential shift of the phenyl toward the least obstructed nitrogen atom(s). In a preliminary study, we established that a phenyl-iron complex is formed in the reaction of nNOS with phenyldiazene and showed that the active site is least sterically encumbered in that isoform over pyrrole ring D (19). We report here extension of this analysis to the active sites of eNOS and iNOS, determination of the rates and extents of reaction of the three NOS isoforms with aryldiazenes, and, most importantly, definition of some of the active site topological changes caused by CaM, H 4 B, and L-Arg.

EXPERIMENTAL PROCEDURES
Materials and General Methods-L-Arg was obtained from Aldrich. Methyl phenyldiazene-carboxylate azo ester was purchased from Research Organics, Inc. (Cleveland, OH). Naphthyl-and biphenyldiazene were synthesized by the procedure of Huang and Kosower (20). H 4 B was from Alexis Biochemicals (San Diego). H 4 B solutions were prepared as a 10 mM stock in 100 mM DTT. Negative controls for the effects of H 4 B in a given assay included the appropriate amount of DTT. Spectra were * This work was supported by National Institutes of Health Grant GM25515, with facility support provided by Liver Center Core Grant 5 P30 DK26743. 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 U.S.C. Section 1734 solely to indicate this fact.
Enzymes-Rat brain nNOS and bovine eNOS were heterologously expressed in Escherichia coli and purified as previously reported (19,21). Murine iNOS was co-expressed with human CaM in E. coli and was purified by ADP-Sepharose chromatography. 2 Related systems for the co-expression of CaM and iNOS have recently been reported (22,23). The three NOS proteins were purified in the absence of H 4 B and L-Arg. Protein concentrations were determined by the Bradford assay (Bio-Rad) using bovine serum albumin as the standard. Concentrations of heme-containing NOS were determined using the extinction coefficients ⑀(400 nm) ϭ 100 mM Ϫ1 cm Ϫ1 for the ferric enzyme and ⌬⑀(444 -490 nm) ϭ 76 mM Ϫ1 cm Ϫ1 for CO-bound NOS (24).
Aryl Shift and Quantitation of N-Arylporphyrins-In a microcuvette, one or more of the following were added to 2 nmol of NOS in 600 l of 0.1 M Hepes (pH 7.5): 15 M H 4 B, 150 M DTT (if no H 4 B added), 6 nmol of CaM, or 800 M L-Arg. The reaction was initiated by addition of 1 l of stock aryldiazene solution, and spectra were periodically recorded. The stock diazene solutions were made by placing 1 l of a 125 mM stock solution of the methyl carboxylate azo ester of the diazene in 100 l of 1 N NaOH. The reaction of the protein with the diazene was allowed to proceed for 30 min before addition of three 5-l aliquots of 20 mg/ml aqueous potassium ferricyanide. Samples were removed 30 and 60 min after the first ferricyanide addition to monitor the formation of the N-aryl isomers. The samples were worked up and chromatographed as previously reported (19).
Statistical Analysis-A paired Student's t test was carried out to determine whether the difference between any pair of N-phenylprotoporphyrin IX regioisomer values was significant. The following equation for t was used: where x 1 and x 2 are the means of each of the two samples being compared, N 1 and N 2 are the number of measurements carried out on samples 1 and 2, respectively, and s p is the pooled standard deviation: N is the total number of measurements and is equal to (N 1 ϩ N 2 ϩ . . . ϩ N k ), x i1 , x i2 , . . . ,x ik are the individual values in each set of measurements, and k is the number of sets of analysis (in all cases samples were compared pairwise, therefore k ϭ 2). The value obtained for t was compared to the tabular value of t with the appropriate degrees of freedom at a 95% confidence level. If the value obtained for the two sets of data was larger than the tabulated value at a 95% confidence level, we conclude that there is a significant difference between the two values. nm) ϭ 30 mM Ϫ1 cm Ϫ1 for H 4 B-free NOS and 25 mM Ϫ1 cm Ϫ1 for H 4 Bbound NOS. To determine the percent of NOS that formed an aryl-iron complex, the concentration of aryl-iron complex was divided by the starting concentration of heme-containing NOS.

RESULTS
The NOS active site topology was examined by a three-step process involving: (a) reaction of the enzyme with aryldiazenes to give aryl-iron complexes, (b) ferricyanide-mediated shift of the aryl group to the heme nitrogens, and (c) HPLC determination of the resulting N-arylprotoporphyrin IX isomer ratios. The nNOS, eNOS, and iNOS employed in the investigation were heterologously expressed in E. coli and were purified by affinity chromatography in the absence of H 4 B and L-Arg. The recombinant proteins were therefore obtained in an H 4 B-free state suitable for investigation of the influence of H 4 B on the active site topology (21). In the case of iNOS, it was necessary to co-express CaM with the enzyme in order to obtain active iNOS. It was therefore not possible to investigate the effect of CaM on the active site topology of iNOS.
Reaction with Phenyldiazene-The three NOS isoforms react with phenyldiazene to form a phenyl-iron complex with an absorption maximum at ϳ470 nm ( Fig. 1), but the rates and extents of the reactions differ. This absorbance maximum is slightly lower than the value of ϳ480 nm for most P450 aryliron complexes but resembles that of the P450 1A1 complex at 474 nm (25). The absorption peak of the NOS phenyl-iron complex is relatively broad.
Effect of Cofactors on the Phenyldiazene Reaction-The extent of the phenyldiazene reaction depends on the presence or absence of CaM, H 4 B, and L-Arg (Fig. 2, Table I). A very small amount of the phenyl-iron complex is formed in the reaction of nNOS with phenyldiazene if L-Arg is present (Fig. 1C). In general, the binding of L-Arg greatly decreases the extent of phenyl-iron complex formation: almost no complex is formed with eNOS and only minor amounts with nNOS and iNOS. The anomalous finding that L-Arg increases iNOS complex formation in the absence of H 4 B is not reliable because the spectrum of iNOS in the absence of H 4 B indicates that the structure of the protein is strongly perturbed. In the absence of L-Arg, the extent of the reaction is highest for CaM-and H 4 B-bound nNOS (80%) and iNOS (59%), but is higher for eNOS (88%) when only CaM is bound. CaM generally increases the extent of complex formation, but the effect of H 4 B is less consistent. It decreases the degree of complex formation with eNOS and iNOS, but with nNOS sometimes increases and sometimes decreases complex formation.
The first order rate constant for phenyl-iron complex formation, like the extent of the reaction, depends on the NOS isoform and on the presence or absence of CaM, H 4 B, and L-Arg. In general, L-Arg decreases the rate of complex formation (Table  I). However, because so little complex is formed in the presence of L-Arg, little weight can be placed on the rates of complex formation in its presence. For nNOS, the binding of CaM increases, and the binding of H 4 B decreases, the rate 2-fold ( Fig.  2, Table I). When CaM is bound at the same time as H 4 B, the effect of the latter agent dominates and the rate is less than that of the protein without CaM (Fig. 2, Table I). A similar pattern is observed for eNOS, except that the increase caused by CaM is smaller and the decrease caused by H 4 B is larger. Again, the effect of H 4 B dominates over the effect of CaM. CaM-free iNOS is not available, and the effect of CaM on its reaction with phenyldiazene could not be evaluated. However, H 4 B decreases the rate of phenyl-iron complex formation, as it does with the other two isoforms. Under any given set of reaction conditions, nNOS reacts most rapidly with phenyldiazene, iNOS at an intermediate rate, and eNOS most slowly.
Reaction with 2-Naphthyl-and p-Biphenyldiazenes-To examine the NOS active site at greater distances above the heme iron than are sampled by the phenyl probe, the reactions of the NOS isoforms with 2-naphthyldiazene and p-biphenyldiazene in the absence of H 4 B and L-Arg were investigated. 2-Naphthyldiazene does not form an aryl-iron complex with eNOS, but 60% of the nNOS and 30% of the iNOS readily form the complex with first order rate constants of 0.08 and 0.20 min Ϫ1 , respectively (Table I). These extents of complex formation are comparable to those obtained with phenyldiazene even though the naphthyldiazene reaction rates are slower. p-Biphenyldiazene does not detectably form a p-biphenyl-iron complex with any of the NOS isoforms. The differences in rates for the larger aryldiazenes could be due to differences in the steric and/or hydrophobicity requirements for the reaction, but the fact that the extent of naphthyl-iron complex formation is eventually the same as phenyl-iron complex formation is more consistent with steric than hydrophobicity constraints in determining whether the aryl group is bound to the iron.
Iron-Nitrogen Shift of the Aryl Groups-Experiments with P450 enzymes have established that ferricyanide-mediated oxidation of the aryl-iron complexes of thiolate-ligated hemoproteins causes the aryl group to migrate to the porphyrin nitrogens (16). Correlation of the resulting N-arylprotoporphyrin IX regioisomer ratios with the crystal structures of three P450 enzymes indicates that the migration is controlled by the steric effects of active site residues (16,26). The aryl groups invariably migrate toward open regions in the active site rather than to more lipophilic regions, as expected if steric rather than entropic effects dominate the rearrangement process. The Narylporphyrin ratios thus provide direct information on the active site topology.
The reactions of the NOS isoforms with the aryldiazenes were monitored at 470 nm until maximum complex formation was observed. Ferricyanide was then added, and the N-arylprotoporphyrin IX adducts were isolated and quantitated by HPLC (Fig. 3). The isomers elute from the HPLC in the order N B :N A :N C :N D , where the subscript denotes the pyrrole ring bearing the aryl group. The regioisomer patterns were obtained for the two constitutive NOS isoforms in the presence and absence of H 4 B and CaM and for iNOS in the presence and absence of H 4 B (Fig. 3, Table II). In those cases where the extent of phenyl-iron complex formation is low, the yields of N-phenylprotoporphyrin IX isomers are also low, and the integrated ratios are less reliable. The phenyl-iron shift was therefore not investigated in the presence of L-Arg due to the very small extent of phenyl-iron complex formation.
The binding of CaM to nNOS in the absence of H 4 B causes a small (9%) but reproducible decrease in the proportion of the N D regioisomer and a compensatory increase (12%) in the proportion of the N A isomer (Table II). Both of these changes are statistically significant at the 95% confidence level by a paired Student's t test (Table III). The proportion of the N D isomer also appears to decrease when CaM binds to eNOS, but the difference is not statistically significant. The effect of CaM on the iNOS active site topology could not be evaluated because the CaM-free protein is not available. The CaM-mediated nNOS regioisomer changes provide the first experimental data suggesting that CaM binding alters the NOS active site topology.
The binding of H 4 B to NOS in the absence of CaM alters the N-phenylprotoporphyrin IX ratios more than the binding of CaM in the absence of H 4 B (Table II). The proportion of the N D isomer decreases in nNOS from 74 to 41% and in eNOS from 81 to 53% when the aryl shift is carried out in the presence versus absence of H 4 B. The decrease in the N D isomer is compensated for by increases in the other regioisomers. All of the following are statistically significant by a paired Student's t test (Table  III): the decreases in the N D isomer in nNOS and eNOS, the increases in N A and N C in nNOS, and the increases in N B and N A in eNOS. These patterns suggest that a decrease in the space above pyrrole ring D due to the binding of H 4 B increases the shift of the phenyl toward the other pyrrole nitrogens.
The changes in the regioisomer ratio when both CaM and H 4 B are bound are similar to those obtained with H 4 B alone (Table II). The binding of CaM if H 4 B is already present does not significantly alter the regioisomer ratio, but the ratio is altered when H 4 B is added to the CaM-bound enzyme. The changes observed in N B , N C , and N D in nNOS, and all the changes observed in eNOS, are statistically significant by a paired Student's t test (Table III). This also appears to be true for iNOS, although the change due to CaM alone cannot be determined. In the case of nNOS, the binding of H 4 B brings  about a structural change in the same region that is perturbed by the binding of CaM. This change involves a diminution of the available space above pyrrole ring D, a finding consistent with the decrease in both the rate and the extent of phenyl-iron complex formation due to the binding of H 4 B to nNOS and eNOS. In the presence of CaM and H 4 B, all three NOS active sites appear to have similar active site topologies (Table II). DISCUSSION The active sites topologies of the three NOS isoforms differ, but the differences are superimposed on what appear to be global topological similarities. The similarities are most evident in the N-phenylprotoporphyrin IX isomer ratios for the CaM-and H 4 (Table  II). The small differences among these ratios indicate that the active sites of the three NOS isoforms are similarly structured with respect to the prosthetic heme group. In all three active sites, the region above pyrrole ring D is the most open, but none of the pyrrole rings is completely masked. However, significant differences in the dimensions of the active site cavity are revealed by the reaction with larger aryldiazenes. Differences are also found in the topological alterations caused by the binding of CaM, H 4 B, and L-Arg.
The reactions of the three NOS isoforms with 2-naphthyldiazene and p-biphenyldiazene indicate that the head space directly above the heme iron atom is limited. In contrast to the P450 enzymes, no p-biphenyl-iron complexes are formed in the NOS reactions with p-biphenyldiazene. A difference in ceiling height among the NOS isoforms is evident from the reaction with 2-naphthyldiazene, which yields a 2-naphthyl-iron complex with nNOS and iNOS but not with eNOS. The formation of the 2-naphthyl-iron complex is ϳ6 times slower with nNOS, and 3.5 times slower with iNOS, than the formation of the phenyl-iron complex (Table I). These rate differences reflect the differences in the lipophilicities of the phenyl (log p ϭ 2.1) and 2-naphthyl (log p ϭ 3.3) groups as well as differences in their size. Models of the 2-naphthyl-and p-biphenyl-iron complexes based on the crystal structure of the P450 cam phenyl-iron complex indicate that the 2-naphthyl ligand extends to a height of ϳ7.1 Å above the iron and the p-biphenyl to a height of ϳ9.9 Å (27). Formation of the nNOS and iNOS 2-naphthyl but not p-biphenyl complexes thus suggests that a protein residue closes the active site directly above the iron atom at a distance of 8 -10 Å. The capping residue in eNOS, which forms the phenyl but not 2-naphthyl complex, is presumably lower than that in the other two isoforms. The NOS ceiling directly above the iron is lower than that in myoglobin or P450 because myoglobin and most P450 enzymes form both the 2-naphthyland p-biphenyl-iron complexes (16,28).
Clear differences in the NOS active sites are evident in the differential effects of CaM, H 4 B, and L-Arg on the phenyldiazene reaction. L-Arg decreases the extent (from 80 to 17%) but modestly increases the rate (from 0.30 to 0.55 min Ϫ1 ) of the reaction of phenyldiazene with CaM-and H 4 B-saturated nNOS (Table I). This result provides quantitative support for the earlier qualitative observation that L-Arg inhibits the reaction of CaM-and H 4 B-saturated nNOS with phenyldiazene (19). Indeed, L-Arg decreases the rate, and particularly the extent, of   (Table I). The binding of L-Arg appears to close the active site of the enzyme, leaving only a small population of the protein with an active site that is still able to react with phenyldiazene. The nature of this small population is unclear. Evidence that the binding of L-Arg to NOS causes a low-to high-spin shift suggests that L-Arg binds close to the heme (21, 29 -31), as it must if it is oxidized by a heme-bound species. The finding that the reaction of phenyldiazene with the enzyme is largely suppressed by L-Arg is consistent with binding of the natural substrate near the heme.
The binding of CaM to nNOS and eNOS modestly increases the rate and extent of the phenyldiazene reaction, with the exception of the eNOS rates in the absence of either H 4 B or L-Arg (Table I). The changes in the nNOS N-phenylprotoporphyrin IX isomer patterns obtained after migration of the phenyl group from the iron to the porphyrin nitrogens indicate that CaM causes a small but statistically significant (Table III) decrease in the proportion of the N D isomer in the absence, but not presence, of H 4 B. A similar small, but in this case statistically insignificant, change is observed in the proportion of the N D isomer when CaM binds to eNOS. These CaM-mediated changes in the nNOS active site topology suggest that CaM binding has at least a small direct effect on the active site conformation. It is not known if this small conformational change has any relation to the role of CaM in facilitating electron transfer from the flavins to the heme iron atom.
The effects of H 4 B on the rate and extent of the phenyldiazene reaction vary. H 4 B decreases the rate of the nNOS reaction except when L-Arg is present, and decreases the extent except when CaM is also bound ( Table I). H 4 B decreases both the rate and extent of the phenyldiazene reaction with CaMbound eNOS and (rate only) iNOS. These rate decreases are consistent with constriction of the active site in a manner that interferes with the phenyldiazene reaction. More direct evidence for topological alteration of the NOS active site is provided by the phenyl shift data (Table II). The binding of H 4 B causes the largest change in the N-phenylprotoporphyrin IX patterns. With all three NOS isoforms, H 4 B binding results in a major statistically significant decrease (Table III) in migration of the phenyl group to the nitrogen of pyrrole ring D with compensatory increases in its shift to the other pyrrole ring nitrogens, particularly pyrrole ring B. These changes in isomer ratio indicate that H 4 B binding increases the steric constraints above pyrrole ring D. The simplest interpretation of this finding is that H 4 B binds above pyrrole ring D, an interpretation supported by independent evidence that H 4 B binds near the heme group (32,33). However, the possibility cannot be excluded that the decrease in space above pyrrole ring D is the result of a conformational alteration associated with the binding of H 4 B at a distal site. In either case, the results clearly demonstrate that binding of H 4 B specifically decreases the available space above pyrrole ring D.
In summary, the active site topologies of nNOS, eNOS, and iNOS are similar, although the ceiling height directly above the iron atom differs. The differences in these ceiling heights, in the rates of the phenyldiazene reactions with the three isoforms, and the ability of the isoforms to react with naphthyldiazene, suggest that the size of the active site decreases in the order nNOS Ͼ iNOS Ͼ eNOS. In all three active sites, the binding of H 4 B markedly decreases the space above pyrrole ring D, suggesting that the cofactor binds, or causes a conformational constriction, in that region. The available space in the active site is further curtailed by the binding of L-Arg, which severely impairs the phenyldiazene reaction by physically blocking access to the heme. Despite an overall similarity, the NOS active sites exhibit topological differences that can possibly be exploited for the development of isoform-specific inhibitors.