Mutation of the Five Conserved Histidines in the Endothelial Nitric-oxide Synthase Hemoprotein Domain

Five conserved histidine residues are found in the human endothelial nitric-oxide synthase (NOS) heme domain: His-420, His-421, and His-461 are close to the heme, whereas His-146 and His-214 are some distance away. To investigate whether the histidines form a non-heme iron-binding site, we have expressed the H146A, H214A, H420A, H421A, and H461A mutants. The H420A mutant could not be isolated, and the H146A and H421A mutants were inactive. The H214A mutant resembled the wild-type enzyme in all respects. The H461A mutant had a low-spin heme, but high concentrations of l-Arg and tetrahydrobiopterin led to partial recovery of activity. Laser atomic emission showed that the only significant metal in NOS other than calcium and iron is zinc. The activities of the NOS isoforms were not increased by incubation with Fe2+, but were inhibited by high Fe2+ or Zn2+ concentrations. The histidine mutations altered the ability of the protein to dimerize and to bind heme. However, the protein metal content, the inability of exogenous Fe2+ to increase catalytic activity, and the absence of evidence that the conserved histidines form a metal site provide no support for a catalytic role for a non-heme redox-active metal.

The nitric-oxide synthase (NOS) 1 isoforms are self-sufficient monooxygenases that utilize O 2 , NADPH, and L-Arg as substrates to synthesize NO, NADP ϩ , and citrulline (1)(2)(3)(4)(5)(6). The bi-domain structure of NOS contains a C-terminal reductase domain that binds NADPH, FAD, and FMN and an N-terminal domain that binds heme, H 4 B, and L-Arg. The two domains are connected by a CaM consensus binding sequence. Binding of CaM to this interdomain hinge triggers the electron transfer from the FMN to the heme required to oxidize L-Arg to NO and citrulline. Three distinct mammalian NOS isoforms have been cloned and characterized: nNOS (NOS-I), iNOS (NOS-II), and eNOS (NOS-III). The activities of nNOS and eNOS are regulated by the cellular Ca 2ϩ levels. However, CaM binds essentially co-translationally to iNOS, and its activity is primarily regulated at the transcriptional level rather than by the Ca 2ϩ concentration (1)(2)(3)(4)(5)(6).
A high degree of sequence similarity exists among the NOS isoforms, in accord with their similar cofactor requirements and enzymatic properties (1)(2)(3)(4)(5)(6). In this context, one of the unresolved mechanistic questions concerns the role of H 4 B in the NOS catalytic cycle. Two lines of evidence argue for a redox role for the pterin in NOS catalysis: (a) redox-inactive H 4 B analogs emulate the structural effects of H 4 B (dimer stabilization, increased substrate affinity, heme low-to high-spin shift), but do not sustain catalysis (7,8), and (b) low-temperature experiments suggest that H 4 B provides the electron necessary to reduce the oxyferroheme intermediate to the species that oxidizes arginine to N-hydroxyarginine (9).
A precedent for a redox role of H 4 B is provided by the aromatic amino acid hydroxylases, in which H 4 B reacts with O 2 to form a 4␣-hydroperoxy complex. Reaction of this peroxy intermediate with a non-heme iron atom generates the Fe 4ϩ ϭO species actually involved in amino acid hydroxylation (10). The crystal structure of tyrosine hydroxylase shows that the iron is coordinated by a "2-His-1-carboxylate facial triad" (11), a motif also present in other non-heme iron proteins (12). In tyrosine hydroxylase, these metal-coordinating residues are His-331, His-336, and Glu-376, with His-331 as the axial ligand and two water molecules at equatorial positions completing a square pyramidal geometry.
Perry and Marletta (13) have reported that iNOS and nNOS are purified with roughly equivalent amounts of bound zinc and copper, respectively, but after desalting can be reconstituted with ferrous chloride to give proteins with one non-heme iron atom per monomer. This iron reconstitution reportedly increases the NO-synthesizing activity of both nNOS and iNOS (13). These authors argued that the DHH motif found in all the NOS enzymes (Asp-419, His-420, and His-421 in human eNOS) resembles the HHDAST motif in enzymes such as lysyl hydroxylase that catalyze iron-dependent hydroxylations. In lysyl hydroxylase, the aspartate and the second histidine coordinate to the non-heme iron atom (14). Perry and Marletta proposed that the aspartate and the second histidine in NOS similarly coordinate Fe 2ϩ . The increased catalytic activity upon incubation with Fe 2ϩ was ascribed to binding of the metal close to H 4 B. They furthermore found the nNOS H652A mutant (equivalent to the human eNOS H421A mutant) to be inactive and argued that the loss of activity was due to the role of this histidine in coordinating to the divalent metal. By analogy with lysyl hydroxylase, another of the proposed metal ligands in nNOS would be His-692, which corresponds to His-461 in human eNOS.
The three-dimensional structure of the bovine eNOS heme domain indicates that the NOS enzymes have a metal-binding site at the monomer-monomer interface consisting of a Zn 2ϩ coordinated to four cysteine residues, two from each monomer (15,16). Crystallization of the iNOS heme domain can result in loss of the zinc and the formation of a disulfide bridge involving Cys-109 of the two monomers (16,17). No evidence was found in the crystal structures for the additional histidine-dependent metal-binding site invoked by Perry and Marletta (13). Moreover, incubation of eNOS with 10 mM FeSO 4 failed to identify new metal-binding sites (15). We report here the site-directed mutagenesis of the five conserved histidines in the human eNOS heme domain, determination of the metal content of NOS expressed in Escherichia coli and purified by affinity chromatography, and analysis of the effect of exogenous Fe 2ϩ and Zn 2ϩ on the activity of human eNOS and nNOS.

EXPERIMENTAL PROCEDURES
Materials and General Methods-nNOS was coexpressed with human CaM and purified as reported (18,19). Expression of the eNOS heme domain, consisting of residues 1-521 plus a six-His tag, in pCWori will be reported separately. 2 L-Arg was obtained from Aldrich; 2Ј,5Ј-ADP-Sepharose and 2Ј-AMP were from Sigma, and H 4 B was from Alexis Biochemicals (San Diego, CA). The Superdex HR200 gel filtration column was from Amersham Pharmacia Biotech; Vent polymerase and restriction enzymes were from New England Biolabs Inc. (Beverly, MA); and the pGEM-T vector and isopropyl-␤-D-thiogalactopyranoside were from Promega (Madison, WI). DNA purification kits and Ni 2ϩ -NTA-agarose were purchased from QIAGEN Inc. (Chatsworth, CA). Bacto-yeast extract, Bacto-Tryptone, and E. coli DH5␣ cells were from Life Technologies, Inc., and BL21 DE3-competent cells were from Novagen (Madison, WI).
Construction of the His-to-Ala Mutants-The five His-to-Ala mutants were constructed by overlap polymerase chain reaction, with two complementary mutagenic primers that introduced a new restriction site (silent mutations) to facilitate screening of the clones. The reverse mutagenic primer was used together with one primer that annealed at the polyhistidine N-terminal tag of pCWori (18) in order to perform the first partial reaction. The forward mutagenic primer was used together with a primer that annealed after the unique KpnI site for the second partial reaction. Once the two partial reactions were completed, the two fragments were copurified and gel-extracted with the purpose of using them as templates for the long polymerase chain reaction in which only the primers that annealed at both ends were used. The sequences of the forward primers that were used in the mutagenesis are given below (the reverse primers used are the exact complementary strand), with the residues in boldface indicating the new restriction site and the underlined residues the mutated site: H146A, 5Ј-AGCGGCTCCCAGGC-CGCGGAACAGCGGCTTCAA-3Ј, which introduces a new SacII site; H214A, 5Ј-GAAATGTTCACCTACATATGCAACGCCATCAAGTATG-CCACC-3Ј, which introduces a new NdeI site; H420A, 5Ј-GTCACCAT-CGTCGACGCCCACGCCGCCACG-3Ј, which introduces a new SalI site; H421A, 5Ј-ACCATCGTGGACCACGCCGCGGCCACGGCC-3Ј, which introduces a new SacII site; and H461A, 5Ј-CCCATCTCGGGA-AGCTTAACTCCTGTTTTCGCTCAGGAGGAGATGGTC-3Ј, which introduces a new HindIII site.
The polymerase chain reaction products were double-digested with NdeI plus KpnI and ligated into the corresponding site of human eNOSpoly-His-pCWori (19) from which the wild-type sequence had been removed. The positive colonies were screened for the new restriction sites introduced with the mutation. The positives were sequenced at the University of California, San Francisco, sequencing facility to confirm the presence of the mutation.
Expression and Purification of the Mutant Proteins-The mutant DNA was used to transform BL21 DE3-competent cells containing a human CaM coexpression system (19). A fresh colony was picked and used to inoculate 2 ml of LB medium that was grown overnight at 37°C in the presence of 0.1 mg/ml ampicillin plus 0.034 mg/ml chloramphenicol. This culture was used to initiate a larger culture of 1.5 liters of 2ϫ YT (16 g of bacto-tryptone, 10 g of bacto-yeast, 5 g of NaCl in 1 liter of deionized water, pH 7.0) that was grown in 2-liter flasks at 200 rpm at 37°C. When an absorbance of 0.8 was reached, 1 mM isopropyl-␤-Dthiogalactopyranoside was added, and the cultures were allowed to grow at 23°C for 16 h (18,19). The cells were then harvested by centrifugation in a GS3 rotor and frozen as a film in a plastic bag at Ϫ70°C until purification of the protein was undertaken. Six liters of medium were used in a typical preparation. All subcloning and cloning was performed in E. coli DH5␣ cells, whereas protein expression was always performed in the protease-deficient E. coli BL21 DE3 strain.
The cell paste was resuspended in buffer A (50 mM Hepes, pH 7.5, 100 mM NaCl, 0.1 mM phenylmethylsulfonyl fluoride, 1 mM leupeptin, 1 mM pepstatin, 1 mM antipain, and 10% glycerol) plus 0.5 mg/ml lysozyme. When needed, 1 mM L-Arg plus 10 M H 4 B were also added to the lysis buffer and kept throughout the entire purification process. The cells were then disrupted by pulse sonication (18,19). Purification of the mutants was performed by binding the proteins to the Ni 2ϩ -NTAagarose after removal of the cell debris by ultracentrifugation. The column was first washed with 5 column volumes of buffer A (with 500 rather than 100 mM NaCl) plus 100 M CaCl 2 , and the proteins were then eluted from the column with 150 mM imidazole in buffer A.
The protein eluted from the Ni 2ϩ -NTA-agarose column was diluted 2-fold with distilled water to lower the ionic strength and was loaded onto the 2Ј,5Ј-ADP-agarose column pre-equilibrated with buffer B (50 mM Hepes, pH 7.5, 0.1 mM phenylmethylsulfonyl fluoride, 1 mM leupeptin, 1 mM pepstatin, 1 mM antipain, and 10% glycerol) in the presence, when desired, of 1 mM L-Arg plus 10 M H 4 B. The column was washed with 3 volumes of buffer B containing 100 M CaCl 2 , and the protein was eluted with 10 mM 2Ј-AMP in buffer C (50 mM Hepes, pH 7.5, 1 M NaCl, protease inhibitors, and 10% glycerol). The purified proteins were stored at Ϫ70°C, and freeze-thawing was avoided.
The only exception to the above procedure was the purification of EDTA-treated iNOS. In this case, the procedure was as described previously (20), except that imidazole was never used, and the protein was eluted from the Ni 2ϩ -NTA column by stripping the Ni 2ϩ with 50 mM EDTA in 50 mM Tris, pH 7.5, 500 mM NaCl, 10% glycerol, and protease inhibitors (as above). The protein was then bound to ADP-agarose and washed with 4 column volumes of 50 mM EDTA in buffer D (50 mM Hepes, pH 7.5, 300 mM NaCl, and 10% glycerol) to remove proteinbound Ni 2ϩ . Another wash was performed with 8 column volumes of buffer D. Purification was then continued as before in the absence of EDTA. Since the H421A and H146A mutants displayed a high tendency to lose the heme prosthetic group, spectral analyses were also performed for these proteins immediately after the protein was eluted from the Ni 2ϩ -NTA-agarose column.
Restoration of the Heme Environment in the H461A Mutant-To study the time-dependent binding of H 4 B to the protein, the base-line spectrum of the protein was recorded in the presence of 10 mM L-Arg. Addition of 100 M H 4 B induced a progressive increase in the high-spin component and a progressive decrease in the low-spin component of the spectrum. The first difference spectrum was recorded 1 min after H 4 B addition, and successive spectra were recorded every 5 min at 25°C. The spectra were plotted, and the A 390 nm Ϫ A 430 nm values were calculated. The spin change (A 390 nm Ϫ A 430 nm ) was plotted versus time, and the plot was fitted to the following equation: where A is the maximum absorbance value and k is a kinetic refolding constant.
FPLC Assays-Aliquots of the mutant proteins (200 l, ϳ0.5 mg/ml) were analyzed at 25°C on an LCC 500-Plus FPLC system equipped with two P-500 pumps, a Superdex HR200 column, and a monitor set at 280 nm. The flow rate was 0.5 ml/min, and the buffer consisted of 50 mM Hepes, pH 7.0, and 100 mM NaCl. When desired to promote dimer formation, the buffer included 1 mM L-Arg and 0.5 M H 4 B. The column void volume was determined with dextran blue, and the total volume with potassium ferricyanide. The column was calibrated with the following proteins (Sigma): thyroglobulin (669 kDa), apoferritin (443 kDa), ␤-amylase (200 kDa), alcohol dehydrogenase (150 kDa), bovine serum albumin (66 kDa), carbonic anhydrase (29 kDa), and horse myoglobin (17 kDa). The calibration curve was obtained by plotting the (v e Ϫ v 0 )/(v t Ϫ v 0 ) ratio against the log of the molecular mass, where v e is the elution volume of the protein, v 0 is the void volume of the column, and v t is the total bed volume of the column. The elution positions of the dimer (ϳ10.5 ml) and monomer (ϳ11.5 ml) were used as a reference (19).
Anaerobic Formation of CO Complexes-Absorption spectra were obtained by placing the proteins (2-5 M) in a septum-sealed cuvette under vacuum and adding NADPH (100 -300 M). No CaM was added since the proteins were coexpressed with human CaM (19). The Ca 2ϩ concentration was kept constant at 100 M. After recording the spectrum, CO was purged into the cuvette for 3 min, and the cuvette was resealed to preserve the CO atmosphere. Spectra were recorded imme-diately on a Varian Cary 1E spectrophotometer connected to a Lauda circulating water bath set at 30°C.
NO Assay and Incubation with Divalent Cations-The activities of the eNOS mutant proteins were determined at 37°C by measuring the conversion of oxyhemoglobin to methemoglobin (18,19). The initial rates of NO production were determined from ⌬A 401-411 nm (⌬⑀ 401-411 nm ϭ 60 mM Ϫ1 cm Ϫ1 ) on a Cary 1E spectrophotometer. The final assay sample volume was 0.5 ml. The protein concentration was calculated using ⑀ 400 nm ϭ 100 mM Ϫ1 cm Ϫ1 for the ferric enzyme under conditions that maximized the high-spin state (i.e. high concentrations of L-Arg and H 4 B).
When the NO-releasing activities of the NOS enzymes were assayed in the presence of divalent cations, freshly prepared stock solutions of 4 mM ferrous sulfate or zinc sulfate in 100 mM Tris buffer, pH 7.0, were utilized. These solutions were used to prepare 0.4 mM stock solutions of both salts, again in 100 mM Tris buffer, pH 7.0. Addition of 5 l of these fresh solutions to a final assay volume of 0.5 ml (see above) resulted in a 100-fold dilution (final concentrations of 40 and 4 M). The Fe 2ϩ or Zn 2ϩ solutions were added to the assay mixture, and the rates were immediately measured.
Metal Determination-The metal content in NOS samples was determined by inductively coupled plasma atomic emission spectroscopy by Ernest Appelhans of Garratt-Callahan Co. (Millbrae, California).
Cytochrome c Reduction Assay-The rate of reduction of cytochrome c was measured at 37°C using ⑀ 550 nm ϭ 21 mM Ϫ1 cm Ϫ1 . Reaction mixtures containing 50 mM Hepes, pH 7.5, 100 g/ml bovine serum albumin, 100 M CaCl 2 , 500 M NADPH, 50 M cytochrome c, 50 units of catalase, and 2-5 g of eNOS were initiated by adding cytochrome c and NADPH. Both H 4 B and dithiothreitol were avoided when cytochrome c reduction was assayed to minimize the buffer-mediated reduction.

RESULTS
Characterization of the H146A Mutant-His-146 is located far from the active site in the ␣2 helix of eNOS ( Fig. 1) (15). Expression of the H146A mutant and elution from Ni 2ϩ -NTAagarose yielded a protein with a Soret absorption maximum at 408 nm despite the presence of the 150 mM imidazole used to elute the protein from the column (Fig. 2, trace 1). Addition of up to 30 mM L-Arg plus 30 M H 4 B did not induce a low-to high-spin shift in the Soret maximum. The absence of a distinct 427 nm absorption peak in the H146A mutant in the presence of imidazole indicates that either the heme distal site is occupied by another ligand or the proximal thiolate ligand has been lost. The absence of a Soret band shift in the presence of high concentrations of L-Arg and H 4 B confirms that major alterations have occurred in the heme environment. When the H146A mutant was reduced with NADPH under a CO atmo-sphere, the reduced CO difference spectrum exhibited only a cytochrome P420 peak (Fig. 2, inset).
The H146A mutant eluted from Ni 2ϩ -NTA-agarose was further purified by chromatography on a 2Ј,5Ј-ADP-agarose column. The absorption spectrum of the resulting protein (Fig. 2, trace 2) shows that the heme was lost during the column washes, leaving only the residual electronic absorbance of the flavin domain. The H146A mutation, despite its distance from the heme group, perturbed the active site sufficiently that the heme was incorrectly coordinated and dissociated from the protein during normal chromatographic procedures.
Characterization of the H214A Mutant-His-214 is located at the C-terminal end of the ␣4 helix at some distance from the heme group (Fig. 1) (15). The recombinant protein was obtained in high purity by the procedure employed previously to purify wild-type human and bovine eNOS. Spectroscopically, this mu- tant is indistinguishable from wild-type eNOS (Fig. 3). The Soret maximum at 395 nm (Fig. 3, trace 1) obtained in the presence of L-Arg and H 4 B indicates that the heme is in the high-spin state in the presence of these factors. Addition of 30 mM imidazole to this protein induced a high-to low-spin shift, as indicated by the final Soret maximum at 427 nm with an ␣/␤-band at 550 nm (Fig. 3, trace 2). When this mutant was analyzed by gel filtration on a Superdex HR200 column, the protein dimer was found to predominate. An asymmetric peak was seen at ϳ10.5 ml, with a broad shoulder at longer elution volumes that masked the position of the monomer at ϳ11.5 ml (Fig. 3). This elution profile is identical to that for the wild-type protein analyzed under similar conditions (18,19).
The NO-synthesizing activity of the H214A mutant (230 nmol min Ϫ1 mg Ϫ1 ) is slightly higher than that of wild-type human eNOS (200 nmol min Ϫ1 mg Ϫ1 ). The cytochrome c-reducing activity of the H214A mutant in the presence of CaM (2500 nmol min Ϫ1 mg Ϫ1 ) is similar to that of the wild-type enzyme (3000 nmol min Ϫ1 mg Ϫ1 ).
Characterization of the H420A and H421A Mutants-His-420 and His-421 are located at the beginning of the ␣9 helix ( Fig. 1) (15) in a region close to the heme site proposed to undergo a large rearrangement upon dimerization (17). The imidazole side chains of His-420 and His-421 are hydrogenbonded to Asp-397 in the neighboring subunit and thus participate in interactions that stabilize the dimer structure.
We were unable to obtain the full-length H420A mutant, apparently because it is proteolyzed in the E. coli expression system. Heme insertion in the isolated protein was imperfect, as the protein eluted from the Ni 2ϩ -NTA-agarose column did not exhibit the usual low-spin spectrum due to the imidazole complex. Its absorption maximum was at 406 nm and did not shift to a high-spin spectrum in the presence of large concentrations of L-Arg and H 4 B (data not shown).
The full-length H421A mutant could be expressed and purified, although its heme group was mostly in the low-spin state (Fig. 4, left panel). Addition of NADPH under a CO atmosphere to the protein eluted from the Ni 2ϩ -NTA column resulted in reduction of the flavins and the progressive formation of an Fe 2ϩ ⅐CO complex with a maximum at 450 nm (Fig. 4, left panel  and inset). Although formation of the cytochrome P450 complex did not go to completion, the extent to which it was formed correlated with the fraction of full-length dimers in the sample as judged by FPLC. Since electron transfer in the NOS dimer appears to be a trans-process (i.e. from the flavins of one monomer to the heme of the other) (21), only full-length dimers can form the reduced CO complex with NADPH as the reductant. These results contrast with those for the nNOS H652A mutant, equivalent to the H421A mutant studied here (13), for which the 450 nm Fe 2ϩ ⅐CO complex could be obtained only when the iron was reduced with dithionite. This difference is not surprising given the higher stability of the eNOS than nNOS dimers and, consequently, the likelihood that the nNOS H652A mutant is primarily in the monomeric state.
The dimer/monomer ratio for the H421A mutant varied from one preparation to another, depending on the amount of heme removed during the column washes. Fig. 4 (right panel) shows FPLC traces of three different preparations of the H421A mutant after elution from Ni 2ϩ -NTA-agarose. Peaks due to contaminating fragments that eluted after 13 ml have been omitted for clarity. H421A dimers and monomers were consistently observed, but in a widely varying ratio.
When the H421A mutant was further purified on an ADPagarose column, the heme Soret maximum of the protein was at 415 nm, and even large concentrations of L-Arg and H 4 B did not induce a low-to high-spin shift (data not shown). The purified H421A mutant had no NO-synthesizing activity, but had wild-type cytochrome c-reducing activity (2700 nmol min Ϫ1 mg Ϫ1 ).
Characterization of the H461A Mutant-His-461 in human eNOS (His-463 in bovine eNOS) is located at the end of the ␣11b helix. Although it does not interact directly with H 4 B, it contributes to the electrostatic interactions that stabilize the pterin moiety ( Fig. 1) (15). One of the His-461 imidazole nitrogens forms a hydrogen bond with a water molecule that, in turn, hydrogen bonds to a hydroxyl of the pterin dihydroxypropyl moiety from a different monomer (15). In iNOS (17), the backbone carbonyl group of Phe-460 is also directly hydrogenbonded to the dihydroxypropyl group. Replacement of His-461 by an alanine gives a protein that is defective in the binding of H 4 B (and consequently of L-Arg). This permits the low-to high-spin shift associated with the binding of H 4 B to be monitored (18,19,22).
The H461A mutant purified in the absence of H 4 B and L-Arg was almost completely in the low-spin state (Fig. 5, trace 1). This contrasts with the wild-type protein, which, even in the absence of the cofactor and substrate, is predominantly in the high-spin state (18,19,22). Incubation of the H461A mutant with 1 mM L-Arg in the absence of H 4 B engendered a moderate low-to high-spin shift (Fig. 5, trace 2), but even prolonged incubation with up to 10 mM L-Arg did not produce a complete shift to the high-spin state. Binding of H 4 B thus appears to be required for a complete shift in the spin state. The spin state shift in the mutant appears to involve a slow refolding process keyed to the residues that bind H 4 B and L-Arg because only after incubation of the mutant with 10 mM L-Arg plus 100 M H 4 B for 30 min at 30°C did we obtain a high-spin absorption spectrum identical to that of the wild-type protein (Fig. 5, trace  3). However, even under these optimal conditions, the H461A mutant was less active than the wild-type enzyme (see below).
H 4 B-and L-Arg-dependent alterations in the monomer-dimer equilibrium were also evident when the H461A mutant was examined by FPLC. The relative ratio of dimers to monomers was preparation-dependent, ranging from a maximum of 70:30 to a minimum of 40:60 (Fig. 5, right panel). Incubation of this mutant with 10 mM L-Arg plus 100 M H 4 B elevated the dimer population at the expense of the monomer population regardless of the initial dimer/monomer ratio (data not shown). However, in no instance did we observe the 90:10 dimer/monomer ratio typical of the wild-type protein under identical conditions (19). NO synthesis was not detected when the H461A mutant was assayed immediately after the protein was purified and 1 mM L-Arg and 1 M H 4 B were added. However, after incubation of the H461A mutant with 10 mM L-Arg plus 100 M H 4 B for 30 min at 30°C, a value of ϳ50 nmol min Ϫ1 mg Ϫ1 was reproducibly obtained using the hemoglobin method. This corresponds to ϳ25% of the wild-type eNOS activity.
To characterize the refolding of the H461A mutant in the presence of high concentrations of L-Arg and H 4 B, we recorded the low-to high-spin Soret band changes over a 30-min interval. The spectra were recorded at 25°C in the presence of 10 mM L-Arg after addition of 100 M H 4 B (Fig. 6A). Plotting the A 390 nm Ϫ A 430 nm absorbance difference versus time yielded a kinetic constant for the binding of H 4 B (Fig. 6B). The average value of the constant from four independent measurements is k ϭ 0.2 Ϯ 0.1 min Ϫ1 . This slow binding of the pterin to the H461A mutant is to be compared with the binding of H 4 B to the wild-type protein, for which both the low-to high-spin Soret transition and the NO-releasing activity of the H 4 B-deficient enzyme are recovered immediately upon addition of H 4 B (18). All of these results argue that the binding of H 4 B is associated with a slow restructuring of the protein that restores the integrity of the heme site and the protein-protein interactions required for dimer formation and catalysis. The integrity of the reductase domain of the H461A mutant was confirmed by the cytochrome c-reducing activity in the presence of CaM (3500 nmol min Ϫ1 mg Ϫ1 ), which compares favorably with the activity of the wild-type protein (3000 nmol min Ϫ1 mg Ϫ1 ).
Metal Content of the NOS Isoforms-The metal content of the NOS isoforms was determined by inductively coupled plasma atomic emission spectroscopy. As the iron content of the protein was difficult to quantitate precisely, the results are expressed relative to the iron content (Fig. 7). The metals that were analyzed for were calcium, cobalt, copper, iron, magnesium, molybdenum, nickel, and zinc. The calcium level in all the samples was 3-6-fold higher than the iron level (data not shown). The eNOS heme domain contained, in addition to calcium and iron, significant amounts of only zinc. Full-length nNOS contained, in addition to calcium and iron, a substoichiometric amount of zinc (relative to iron) and traces of copper and nickel. iNOS purified by the normal procedure contained a variety of metals, but when purified with EDTA-treated buffer washes, contained only calcium, iron, and small amounts of zinc and magnesium. No other metal apart from iron and calcium appeared to be present in significant amounts. In particular, copper was not found as a significant constituent, in contrast to an earlier report that copper is a major constituent of nNOS when it is isolated (13).
Incubation of eNOS and nNOS with Divalent Cations-Fe 2ϩ reportedly increases NOS activity (13). Zn 2ϩ , which is present in the eNOS structure (15,16), has been shown at high concentrations to inhibit the enzyme (23,24). We compared the effect of these two divalent cations on NOS activity (Fig. 8). In our hands, 4 M Fe 2ϩ had no effect on eNOS activity and only slightly enhanced the nNOS and inhibited the iNOS activities. Addition of 40 M Fe 2ϩ decreased the activities of eNOS and iNOS and caused little change in that of nNOS. Larger concentrations of Fe 2ϩ (Ͼ80 M) decreased all three NOS activities. Zn 2ϩ is a potent NOS inhibitor, and 40 M Zn 2ϩ almost suppressed the eNOS, nNOS, and iNOS activities. This inhibition of NOS by Zn 2ϩ was observed previously by other investigators (23,24) and was tentatively attributed to defective reduction of the heme iron in the presence of the divalent cation (23). DISCUSSION The roles of the five histidines conserved among all the known mammalian NOS sequences have been examined by site-directed mutagenesis. These studies have a direct bearing on the report that the activity of NOS is increased by divalent cations that coordinate to one or more of these histidine residues (13). This proposed histidine-dependent metal-binding site is distinct from the zinc site at the dimer interface (15,16), in which the metal is coordinated to four cysteine residues.
Three of the five conserved histidines (His-146, His-420, and His-421) are part of secondary structural elements that are required for the enzyme either to fold properly or to assemble into the catalytically active eNOS dimer. His-146 is distant from the active site ( Fig. 1), but its replacement with an alanine produces an inactive protein with an improperly incorporated heme group (Fig. 2). These long-distance effects presumably reflect a critical role of His-146 in the folding of the ␣2 and ␣1 helices.
The DHH motif containing His-420 and His-421 lies at the dimer interface and is involved in multiple hydrogen bonds that implicate it as a critical region in the monomer-monomer interface ( Fig. 1) (15). Mutation of either of the two histidine residues in the DHH motif destabilizes the protein. The H420A mutant is apparently misfolded and is sufficiently sensitive to proteolytic digestion that the intact protein could not be isolated. The purified H421A mutant has a low-spin heme Soret maximum and is both unstable and inactive. However, the dimeric form of the protein retains the heme prosthetic group and forms an Fe 2ϩ ⅐CO complex when reduced with NADPH (Fig. 4). The CO complex is formed only by the lower (versus wild-type) fraction of the protein in the dimeric state, as electron transfer to the iron reportedly occurs from the reductase domain of one monomer to the heme domain of the other (21). This behavior contrasts with that observed when His-652, the equivalent residue in nNOS, was mutated (13). The inability of NADPH to reduce the His-652 mutant of nNOS led to the proposal that loss of a divalent metal coordinated to the histidine interrupted the electron transfer pathway. The present results with the eNOS H421A mutant make this explanation untenable, at least as a general one. Our data indicate that the inactivity of the eNOS H421A mutant, and presumably its nNOS H652A counterpart, is due to the formation of a defective heme environment and dimer interface, as evidenced by an increase in the proportion of the protein in the monomeric state. Although the His-420/His-421 pair from each of the eNOS monomers faces the corresponding His-420/His-421 pair of the other monomer, the arrangement of these four imidazole side chains does not seem to be appropriate for metal coordination since no metal was bound in this region when the eNOS heme domain was incubated with 10 mM FeSO 4 (15). 3 His-214 does not fulfill an apparent major catalytic or structural role because the H214A mutant displays a wild-type phenotype (Fig. 3).
Mutation of His-461, which binds to H 4 B, alters the mono-3 T. Poulos, personal communication.

FIG. 7.
Metal content of NOS isoforms determined by inductively coupled plasma atomic emission spectroscopy. The iNOS, eNOS heme domain, and nNOS samples were purified by normal procedures. The iNOS (EDTA-treated) sample was obtained by the usual procedures, except that the protein was exposed to 50 mM EDTA when eluted from the Ni 2ϩ -NTA-agarose column and while bound to the ADP-agarose column (for details, see "Experimental Procedures"). Only the results for copper, iron, and zinc are shown. None of the other metals tested (cobalt, magnesium, molybdenum, and nickel) was found in amounts much above the detection limit except in the EDTA-untreated iNOS sample, which contained appreciable amounts of magnesium, molybdenum, and nickel. These metals were not present after the EDTA washes.
FIG. 8. Effect of divalent cations on NOS activity. Fe 2ϩ or Zn 2ϩ at concentrations of 40 and 4 M were included in the NOS assay mixture, and NO formation was determined at 37°C with the hemoglobin assay. The data are plotted using the eNOS, nNOS, or iNOS activity in the absence of any added metal as the reference 100% value. The activities of recombinant human eNOS (gray bars), rat nNOS (cross-hatched bars), and mouse iNOS (black bars) were examined. Similar results were obtained in three independent measurements. mer-monomer interaction, the heme chromophore (Fig. 5), and the NO-releasing activity of eNOS. Indeed, one of the electrostatic interactions necessary for proper binding of H 4 B involves a hydrogen bond through a water molecule to one of the imidazole nitrogens of His-461. This stabilizing interaction implicates the histidine residue of one subunit in binding the pterin molecule of the other subunit (15). In contrast to wild-type eNOS, which binds H 4 B almost instantaneously (18), the H461A mutant binds H 4 B slowly even in the presence of 10 mM L-Arg. This suggests that H 4 B binding requires gradual refolding of the biopterin site and dimer interface (Fig. 5). This reorganization of the biopterin-binding site, which is accompanied by a low-to high-spin shift of the iron, occurs over a period of ϳ30 min at 25°C (Fig. 6). When reconstituted with H 4 B, the H461A mutant has ϳ25% of the catalytic activity of the wildtype enzyme. The lower activity of fully reconstituted H461A eNOS, despite its full conversion to the high-spin state, stems, at least in part, from a subtle disruption of the monomermonomer interface. Direct evidence for this is provided by FPLC analysis, which shows that the dimer/monomer ratio for the H461A mutant is lower than that for the wild-type protein (Fig. 5, right panel). A comparable alteration in the binding of H 4 B has been reported for the nNOS C331A mutant, which requires prolonged incubation with L-Arg and H 4 B to bind the biopterin and to reconstitute the catalytic activity (25). Sequence comparisons suggest that Cys-331 is one of the cysteines in nNOS that coordinate to the Zn 2ϩ cation. The structural alteration due to defective binding of Zn 2ϩ is presumably overcome by the slow binding of L-Arg and H 4 B (25).
Metal analysis of the eNOS, nNOS, and iNOS samples indicates that the only significant metal associated with the proteins, apart from calcium and iron, is zinc. Zinc is present in the eNOS heme domain at a ratio of ϳ0.6 eq relative to the total iron; in nNOS at a ratio of ϳ0.3 eq; and in iNOS at ratios of ϳ0.7 and 0.2 eq in the enzyme isolated with normal and EDTAtreated buffers, respectively (Fig. 7). The zinc content roughly correlates with the stability of the homodimer to dissociation because the eNOS homodimer is the most stable and the iNOS homodimer the least stable among the isoforms. It is likely that the monomeric enzyme does not retain the zinc because the zinc-binding site, as shown by crystal structures of eNOS and iNOS (15,16), is at the dimer interface and is composed of two thiolate groups from each of the monomers in the dimer. Thus, a completely dimeric enzyme would be expected to have 0.5 eq of zinc relative to heme iron, close to the value found for the eNOS heme domain and iNOS purified without exposure to EDTA. Monomerization of the protein should decrease the proportion of bound Zn 2ϩ , as found for nNOS and EDTA-treated iNOS. The metal content studies provide no support for the presence of a catalytic metal other than the heme iron, although it has not been possible to quantitate the iron content sufficiently well to determine whether any non-heme iron is present.
The metal independence of the activities of eNOS and nNOS found here differs from the report that the activities of nNOS and iNOS increase upon incubation with Fe 2ϩ (13). In our hands, the maximum increase in NO production by nNOS upon incubation with iron, as judged by the hemoglobin assay, is only 18% (Fig. 7). No increase in the eNOS and iNOS activities was observed upon incubation with 4 M iron, and a reproducible decrease in the these activities was observed with 40 M Fe 2ϩ . The reason for the discrepancy in our findings and those of Perry and Marletta (13) is unclear. One difference in the two experiments is that Perry and Marletta measured activity by measuring the conversion of [ 14 C]arginine to [ 14 C]citrulline, whereas we determined the production of NO by the hemoglo-bin assay. The oxidation of guanidoximes such as N-hydroxyarginine is catalyzed by a variety of oxidizing systems, including cytochrome P450 (26), horseradish peroxidase/H 2 O 2 (27), hemoglobin (II), H 2 O 2 alone (28), and superoxide (29,30). This oxidation primarily produces nitrites or nitrates rather than NO. It is possible that exogenous iron accelerates the NOS reaction by oxidizing N-hydroxyarginine via a non-physiological mechanism that yields citrulline, but does not produce NO. A second difference in the two sets of experiments is that Perry and Marletta intentionally used metal-depleted buffer in their assays, a treatment that might result in loss of the Zn 2ϩ bound to the Zn 2ϩ -binding site at the dimer interface. The different stabilities of the three isoforms toward monomer formation suggest that this might have the most severe consequences for iNOS, for which the dimer is the least stable, and the least consequences for eNOS, for which it is the most stable (31). A recovery of activity could reflect a divalent metal-induced increase in dimer formation due to reformation of catalytically active dimer. This explanation, however, is inconsistent with the stoichiometry reported by Perry and Marletta, who found that both nNOS and iNOS can be reconstituted with one nonheme iron per monomer. Even if one of these iron atoms occupies the Zn 2ϩ -binding site at the monomer-monomer interface, a second site would have to be found for the additional iron atom.
In summary, one of the conserved histidines distant from the active site (His-214) is not critical for protein folding or catalytic activity, whereas the other (His-146) fulfills an important role in assembly of the protein and its active site. The three conserved histidines near the heme and biopterin are all critical for intersubunit interactions and catalytic activity: mutations of His-420 and His-421 irreparably alter the active site, but mutations of His-461 can be partially overcome by high concentrations of L-Arg and H 4 B. Malfunction of the protein due to the histidine mutations appears to reflect alterations in the protein structure and its ability to bind heme and to form dimers rather than loss of a metal coordination site. The metal content of the proteins and their lack of activation by exogenous Fe 2ϩ , in conjunction with the findings on the conserved histidines, provide no support for the proposed involvement of a metal other than the heme iron (and the structural Zn 2ϩ ) in the catalytic process (13).