Characterization of heme-deficient neuronal nitric-oxide synthase reveals a role for heme in subunit dimerization and binding of the amino acid substrate and tetrahydrobiopterin.

Neuronal nitric-oxide (NO) synthase contains FAD, FMN, heme, and tetrahydrobiopterin as prosthetic groups and represents a multifunctional oxidoreductase catalyzing oxidation of L-arginine to L-citrulline and NO, reduction of molecular oxygen to superoxide, and electron transfer to cytochromes. To investigate how binding of the prosthetic heme moiety is related to enzyme activities, cofactor, and L-arginine binding, as well as to secondary and quaternary protein structure, we have purified and characterized heme-deficient neuronal NO synthase. The heme-deficient enzyme, which had preserved its cytochrome c reductase activity, contained FAD and FMN, but virtually no tetrahydrobiopterin, and exhibited only marginal NO synthase activity. By means of gel filtration and static light scattering, we demonstrate that the heme-deficient enzyme is a monomer and provide evidence that heme is the sole prosthetic group controlling the quaternary structure of neuronal NO synthase. CD spectroscopy showed that most of the structural elements found in the dimeric holoenzyme were conserved in heme-deficient monomeric NO synthase. However, in spite of being properly folded, the heme-deficient enzyme did bind neither tetrahydrobiopterin nor the substrate analog N(G)-nitro-L-arginine. Our results demonstrate that the prosthetic heme group of neuronal NO synthase is requisite for dimerization of enzyme subunits and for the binding of amino acid substrate and tetrahydrobiopterin.

Nitric oxide, an important effector and signaling molecule in the nervous, immune, and cardiovascular systems (1)(2)(3)(4), is enzymatically generated from L-arginine and molecular oxygen by different nitric-oxide synthases (NOS, 1 EC 1.14.13.39). To date, two constitutively expressed, Ca 2ϩ -dependent NOS isoforms as well as a cytokine-inducible, Ca 2ϩ -independent protein have been characterized (5). The isozyme purified from brain (5-7) termed neuronal NOS (nNOS), was identified as a 320-kDa homodimer, containing close to stoichiometric amounts of heme, FAD, and FMN as well as variable amounts of tetrahydrobiopterin (H 4 biopterin) (8 -14). nNOS exhibits sequence similarities to cytochrome P450 reductase (15) and to the heme-binding sequence in P450 proteins (16), suggesting that the enzyme represents a fusion protein of a cytochrome P450-like protein and a P450 reductase (17). In support of this hypothesis, nNOS was shown to catalyze the reduction of cytochrome c (18) and to possess a cysteine-ligated heme-iron (19,20) with spectral properties characteristic of a P450 protein (9,12,21). Further studies have confirmed the bidomain structure of nNOS (20,22,23), with the reductase domain shuttling NADPH-derived electrons in a calmodulin-triggered fashion from the flavins to the heme moiety (24,25), which is located in close proximity to the amino acid substrate and pteridine binding sites (26) and catalyzes a two-step mono-oxygenation of L-arginine to L-citrulline and NO with N G -hydroxy-L-arginine as intermediate (27,28).
Recent work has focused on the allosteric regulation of NOS by L-arginine, H 4 biopterin, and heme. The cytokine-inducible NOS dimer from macrophages was shown to form catalytically inactive monomers in the absence of H 4 biopterin or L-arginine and, notably, to lose its prosthetic heme group in the absence of these ligands. Reassociation of the monomers to catalytically active dimers required the coincident presence of heme, Larginine, and pteridine, pointing to a role of these compounds in the post-translational processing of inducible NOS (29). However, in contrast to the inducible isozyme, native nNOS maintained its catalytically active, dimeric conformation even in the absence of pteridine and L-arginine (14). Thus, it is currently unclear how binding of heme to nNOS is related to subunit assembly, catalytic activity, and pteridine binding. To address this issue, we have purified heme-deficient nNOS (nNOS(hϪ)) from a baculovirus overexpression system and analyzed the virtually heme-free protein for catalytic activity, cofactor content, amino acid substrate, and H 4 biopterin binding and determined the secondary and quaternary structural features of the enzyme.
Preparation of nNOS-Large scale purification of heme-saturated nNOS (nNOS(hϩ); Ն0.9 eq of heme per monomer from ϳ4.5 ϫ 10 9 Sf9 cells infected in the presence of hemin chloride (4 mg/liter) with rat brain nNOS-recombinant baculovirus (32) was performed by sequential affinity chromatography on 2Ј,5Ј-ADP-Sepharose and calmodulin-Sepharose as described in detail recently (33). Partially heme-deficient nNOS (ϳ0.3 eq of heme per monomer) was obtained from Sf9 cells, which had been infected in the absence of hemin chloride. For the isolation of nNOS(hϪ), the partially heme-deficient protein was concentrated to ϳ0.10 mM by means of Vivapore concentrators (Vivascience Ltd., Stonehouse, UK), and 0.20-ml aliquots were subjected to gel filtration chromatography as described below. Fractions (0.3 ml) containing Ͼ0.4 nmol of nNOS and Ͻ0.1 eq of heme per monomer were pooled and stored at Ϫ70°C. Purified nNOS(hϪ) had a concentration of ϳ4.5 M and contained Ͻ0.04 eq of heme per monomer.
Gel Filtration Chromatography-Aliquots of 0.10 -0.20 ml containing 0.5-20 nmol of nNOS were injected into an HPLC system (LiChro-Graph L-6200, Merck) equipped with a low pressure gradient controller and a gel filtration column, providing a separation range from approximately 5-5,000 kDa (Superose 6 HR 10/30, Pharmacia Biotech, Vienna, Austria). nNOS was eluted at room temperature at a flow rate of 0.5 ml/min with a 50 mM triethanolamine/HCl buffer, pH 7.4, containing 0.50 M NaCl and detected by its absorbance at 280 nm (UV/VISdetector L-4250, Merck). Fractions of 0.30 ml were collected for protein and cofactor determinations (see below). The column was calibrated with the gel filtration calibration kit from Pharmacia Biotech (Vienna, Austria), including blue dextran 2,000 (for the determination of the column void volume), thyroglobulin (669 kDa, Stokes radius ϭ 8.50 nm), ferritin (440 kDa, 6.10 nm), catalase (232 kDa, 5.22 nm), aldolase (158 kDa, 4.81 nm), and bovine serum albumin (67 kDa, 3.55 nm). Calibration curves were obtained by plotting (Ϫlog K av ) 0.5 against the Stokes radii of the above standard proteins (K av with V e , V o , and V t denoting the elution volume of the protein, the column void volume, and the total bed volume of the column, respectively. Reconstitution of nNOS(hϪ) with Flavins and Heme-Unless otherwise indicated, nNOS was preincubated at room temperature in a 50 mM triethanolamine/HCl buffer, pH 7.4, containing 12 mM 2-mercaptoethanol, with a 2-fold molar excess of FAD and FMN or hemin chloride for 5 min (flavins) or 30 min (hemin chloride). Unbound flavins and hemin were removed by gel filtration chromatography as described above.
Determination of Protein, Heme, FAD, FMN, and H 4 biopterin-Protein was determined with the Bradford method (34) using bovine serum albumin as standard. The amount of bound flavins and H 4 biopterin was determined by reversed phase HPLC and fluorescence detection with authentic FAD, FMN, and H 4 biopterin as standards (35). Enzymebound heme was quantified by reversed phase HPLC and UV/VISdetection using myoglobin as standard (33). Calculations of molar cofactor/NOS ratios were based on a subunit molecular mass of 160 kDa as calculated from the deduced amino acid sequence of the enzyme (15).
Determination of Enzyme Activity-NOS activity was determined as rates of L-[2,3,4,5-3 H]citrulline formation from L-[2,3,4,5-3 H]arginine (36). Incubations were performed for 10 min at 37°C in 0.1 ml of 50 mM triethanolamine/HCl buffer, pH 7.0, containing 0.6 -2.4 pmol of nNOS, 0.1 mM L-[2,3,4,5-3 H]arginine (ϳ80,000 cpm), 0.5 mM CaCl 2 , 10 g/ml calmodulin, 0.2 mM NADPH, 10 M H 4 biopterin, 5 M FMN, and 5 M FAD. The detergent CHAPS was included in the incubation mixture at a final concentration of 0.2 mM, since purified rat brain NOS was found to exhibit only ϳ20% of control activity if assayed in the absence of an added detergent (not shown). Calmodulin-dependent NADPH:oxygen and NADPH:cytochrome c oxidoreductase activities of nNOS were determined in the absence of L-arginine as described previously (8,18). Unless otherwise indicated, purified NOS (0.3-1.2 pmol) was incubated in final volumes of 0.2 ml of 50 mM triethanolamine/HCl buffer, pH 7.0, in the presence of 0.5 mM CaCl 2 , 10 g/ml calmodulin, 0.1 mM NADPH, and 0.2 mM CHAPS at 37°C. For the determination of cytochrome c reductase activity, 0.2 mM cytochrome c was additionally present. The changes in absorbance at 340 nm (NADPH) or 550 nm (cytochrome c) were monitored continuously against calmodulin-deficient blank samples, which exhibited Յ2 and 12% of enzyme activity determined in the presence of calmodulin, respectively. Rates of NADPH oxidation and cytochrome c reduction were calculated using extinction coefficients of 6.3 and 21 mM Ϫ1 ϫ cm Ϫ1 , respectively. For the determination of enzyme kinetic parameters, substrate concentrations were 1-50 M (L-arginine) and 5-100 M (cytochrome c), respectively. V max and K m values were obtained from weighted Lineweaver-Burk plots (37). Turnover numbers (k cat ) were calculated from the respective V max values and represent the maximum number of converted L-arginine molecules per min and per number of active sites, i.e. the number of heme-containing NOS molecules.
Radioligand Binding Studies-Binding experiments were performed as described previously (38,39). Briefly, nNOS (18 -40 pmol) was incubated for 10 min at 37°C with 12 nM L-[ 3 H]NNA (ϳ70 nCi) or [ 3 H]H 4 biopterin (ϳ17 nCi) and increasing concentrations of the respective unlabeled ligand (10 nM-10 M) in 0.1 ml of a 50 mM triethanolamine/HCl buffer, pH 7.0. Reactions were stopped by polyethylene glycol precipitation followed by vacuum filtration. The amount of the bound radioligand retained on the filters was determined by liquid scintillation counting. Data were corrected for nonspecific binding determined in the presence of 1 mM unlabeled ligand. K D and B max values were calculated using the GIPMAX nonlinear least squares regression curve fitting program (40). Given the low overall recovery (Ͻ25%) of this binding assay (36,38,39), the calculated B max values are only semiquantitative estimates and do not allow the determination of the number of pteridine or L-arginine binding sites.
Gel Electrophoresis-nNOS was subjected to low temperature SDS-PAGE essentially as described recently (14). Briefly, nNOS (ϳ150 pmol) was incubated for 5 min at 37°C in 50 l of 50 mM triethanolamine/HCl buffer (pH 7.0) in the absence or presence of H 4 biopterin (0.1 mM) and L-arginine (1 mM). Incubations were terminated by the addition of 50 l of chilled Laemmli buffer (41), containing 0.125 M Tris-HCl (pH 6.8), 4% (w/v) SDS, 20% (w/v) glycerol, and 0.02% (w/v) bromphenol blue. 2-Mercaptoethanol, which interfered with the heme-staining method described below, was omitted from the Laemmli buffer without having any effect on the migration pattern nNOS. Samples, containing ϳ30 pmol of nNOS, were subjected to SDS-PAGE for 60 min at a constant current of 30 mA on discontinuous 6% SDS-gels (70 ϫ 80 ϫ 1 mm). Gels and buffers, which had been prepared according to Laemmli (41), were equilibrated at 4°C prior to electrophoresis, and the buffer tank was cooled by means of an ice bath during electrophoresis. Gels were either stained for protein with Coomassie Blue R-250 or for heme with 3,3Јdimethoxybenzidine/H 2 O 2 following a published method (42) with slight modifications. Gels were washed for 10 min in methanol/sodium acetate (0.25 M, pH 5.0) ϭ 3:7 (v/v) and subsequently incubated in the dark for 20 min in a freshly prepared solution, containing 7 parts of 0.25 M sodium acetate, pH 5.0, and 3 parts of 6 mM 3,3Ј-dimethoxybenzidine dihydrochloride in methanol. Gels were developed for 60 min by adding H 2 O 2 to a final concentration of 60 mM, washed for 30 min in H 2 O/ methanol/acetic acid ϭ 8:1:1 (v/v/v), dried, and photographed. Relative amounts of protein or heme were estimated by densitometric analysis using the vds 800 video system and H1D-software of Hirschmann (Analysentechnik Hirschmann, Taufkirchen, Germany). The molecular mass of nNOS was estimated using a calibration kit from Bio-Rad, including thyroglobulin (subunit molecular mass ϭ 330 kDa), ferritin (subunit molecular mass ϭ 220 kDa), phosphorylase b (94 kDa), bovine serum albumin (67 kDa), and catalase (subunit molecular mass ϭ 60 kDa).
Static Light Scattering-Static light scattering can be used for the determination of the molecular mass of macromolecules in solution by extrapolation of the scattered intensity to zero angle (Zimm plot) (43,44). Accurate determinations of the specific refractive index increments (␦n/␦c) of the solute particles are required for molecular mass determinations by this method. Due to limitations in the amount of the available nNOS(hϪ) protein, reliable ␦n/␦c measurements were not possible. However, the ␦n/␦c values for related proteins are identical, and there is virtually no angular dependence of the scattering intensity (I 90 ϳ I 0 ). Thus, the scattering intensity of different nNOS species measured at a fixed scattering angle depends exclusively on (molecular mass ϫ protein concentration), allowing the determination of the molecular mass ratio of nNOS(hϪ) and nNOS(hϩ). Light scattering measurements were performed on an arbitrary scale with equally concentrated solutions of nNOS(hϪ) and nNOS(hϩ) (5.4 M) in rectangular microcuvettes (30 l) at a scattering angle of 90 degrees at 8°C, using an Ar ϩ laser (Spectra-Physics, model 2060 -65) at a wavelength of 514.5 nm and an output power of 400 milliwatts. Measuring times were 20 s, with each sample being scanned 20 times.
CD Spectroscopy-For CD measurements, solutions of nNOS (ϳ15 M) were diluted 1:10 in 50 mM sodium phosphate buffer, pH 7.4 and reconcentrated, using Centricon-30 microconcentrators (Amicon, purchased from W. D. Haider, Kottingbrunn, Austria). This procedure was repeated four times to minimize the amount of 2-mercaptoethanol, EGTA, triethanolamine, and NaCl present in the enzyme preparations. CD spectra were recorded on a Jasco J-500A spectrophotometer, using a 0.1-mm path length cell (Helma) thermostated at 20°C. Spectra and buffer base lines (filtrate from the last concentration step on Centricon-30 microconcentrators) were averaged from 10 scans, each recorded at 0.1 nm intervals, using a scanning rate of 1 nm/min and a time constant of 2 s. The instrument was calibrated as described (45), and base line corrections were made by means of the software provided by Jasco. Calculations of ⌬⑀ values were based on amino acid analysis, and the secondary structure of nNOS was computed by means of the variable selection method using a set of 33 reference proteins (46).

RESULTS
To investigate the role of heme in nNOS structure and function, we attempted to purify recombinant heme-deficient nNOS from a baculovirus overexpression system. From 4.5 ϫ 10 9 cells, which had been infected with the rat brain NOS-recombinant baculovirus for 48 h in the absence of hemin chloride, we obtained ϳ80 mg of nNOS with a heme content of 0.29 Ϯ 0.03 eq per monomer (n ϭ 3). The amount of heme bound to the recombinant protein was not significantly reduced by preincubation (24 -48 h) and infection (48 h) of Sf9 cells in the presence of 0.25 and 0.50 mM succinyl acetone, an inhibitor of heme biosynthesis (47), although cytosolic heme levels were reduced down to 10% of untreated control cells under these conditions (not shown). Partially heme-deficient nNOS was further analyzed for H 4 biopterin, FAD, and FMN, revealing that substoichiometric amounts of these cofactors were incorporated into the protein (0.12 Ϯ 0.01, 0.17 Ϯ 0.01, and 0.11 Ϯ 0.01 eq per monomer, respectively; n ϭ 3). In the presence of saturating concentrations of exogenous H 4 biopterin, FAD, and FMN, the partially heme-deficient enzyme exhibited only low specific activity (0.30 Ϯ 0.03 mol of L-citrulline ϫ min Ϫ1 ϫ mg Ϫ1 , n ϭ 3) as compared with the holoenzyme (1.08 Ϯ 0.14 mol ϫ min Ϫ1 ϫ mg Ϫ1 , n ϭ 3).
To investigate whether partially heme-deficient nNOS can be reconstituted with flavins, we have preincubated the protein with a 2-fold molar excess of FAD and FMN for 5 min at room temperature and subsequently removed free flavins by gel filtration chromatography. Under these conditions, the amount of protein-bound FAD and FMN was increased ϳ6-fold to 0.98 Ϯ 0.06 and 0.65 Ϯ 0.04 eq per monomer, respectively (n ϭ 3). Increasing the incubation time to 60 min did not further enhance FMN binding. To confirm that the flavins were incorporated into the nNOS protein in a catalytically active form, we have determined cytochrome c reductase activities of the reconstituted protein in the absence of exogenous flavins. Enzyme kinetic analysis revealed that reconstitution of nNOS with flavins had no effect on the affinity of cytochrome c for nNOS but was accompanied by a ϳ5-fold increase of cytochrome c reductase activity as compared to controls (K m , 14 Ϯ 3 versus 15 Ϯ 1 M; V max , 2.6 Ϯ 0.6 versus 15 Ϯ 2 mol ϫ min Ϫ1 ϫ mg Ϫ1 ; n ϭ 3). Maximal rates of L-citrulline formation as well as L-arginine-and H 4 biopterin-independent NADPH oxidation were not affected (not shown), demonstrating that the amount of enzyme-bound heme was limiting for NOS activity of the protein.
Gel filtration chromatography of the partially heme-deficient enzyme preparations, which had been reconstituted with FAD and FMN, revealed that NOS activity eluted in a single and well defined peak centered at an elution volume of 11.8 ml (Fig. 1, upper panel, filled circles). nNOS eluting in this peak fraction contained 0.56 Ϯ 0.06 eq of heme per monomer (n ϭ 3) and had a specific activity of 0.88 Ϯ 0.09 mol ϫ min Ϫ1 ϫ mg Ϫ1 (n ϭ 3) when assayed in the presence of saturating concentrations of exogenous H 4 biopterin and flavins. Cytochrome c reductase activity eluted in a broad peak with two maxima at 11.8 and 13.0 ml, and ϳ40% of the total reductase activity were found in fractions which did not exhibit detectable citrulline formation (Fig. 1, upper panel, open circles), suggesting the separation of two protein species. As shown in the lower panel of Fig. 1, comparable elution profiles were found when the eluate was assayed for protein (filled circles), FAD (open circles), and FMN (filled squares), demonstrating that both protein species contained virtually stoichiometric amounts of FAD and ϳ0.6 eq of FMN per monomer. Substoichiometric binding of FMN as well as a rightward shift of the FMN elution profile points to dissociation of the flavin during gel filtration. In contrast to FAD and FMN, which co-eluted with the reductase activity of nNOS, heme and H 4 biopterin strictly co-eluted with L-citrulline forming activity, demonstrating that heme-deficient nNOS was separated from the active heme-containing enzyme.
Virtually pure heme-deficient nNOS (nNOS(hϪ), Յ0.04 eq of heme per monomer) was obtained by pooling gel filtration column fractions which contained Ͼ0.4 nmol nNOS and Ͻ0.1 mol eq of heme per monomer. nNOS(hϪ) was found to contain only marginal amounts of heme and H 4 biopterin (0.030 Ϯ 0.006 and 0.007 Ϯ 0.001 eq per monomer, respectively; n ϭ 3). The amount of flavins bound to nNOS(hϪ) was almost identical with that of the starting material used for the isolation of the heme-deficient protein (0.87 Ϯ 0.14 versus 0.98 Ϯ 0.06 and 0.61 Ϯ 0.08 versus 0.65 Ϯ 0.04 eq per monomer, respectively; n ϭ 3), indicating that heme deficiency does not affect the affinity of nNOS for FAD and FMN.
We have recently shown that nNOS purified from porcine brain is converted to an SDS-resistant dimer upon binding of H 4 biopterin and L-arginine (14). Heme saturation of the porcine enzyme (13) precluded, however, investigation of the contribution of the prosthetic heme group to this tight interaction of nNOS subunits. To address this issue, we have subjected nNOS(hϪ) and nNOS(hϩ), which had been preincubated with 2% SDS in the absence and presence of H 4 biopterin (0.1 mM) and L-arginine (1 mM), to low temperature SDS-PAGE. In the absence of the added ligands (Fig. 2, upper panel, lane A), the ratio of SDS-resistant nNOS(hϩ) dimers (ϳ300 kDa) to monomers (ϳ150 kDa) was ϳ15:85. Preincubation of nNOS(hϩ) with exogenous H 4 biopterin (0.1 mM) and L-arginine (1 mM) increased the relative amount of dimeric nNOS to ϳ50%. With nNOS(hϪ), we did not detect any SDS-resistant protein dimers both in the absence (lane C, upper panel) and presence (lane D, upper panel) of saturating amino acid substrate and cofactor concentrations. Heme staining of the gels by means of the dimethoxybenzidine/H 2 O 2 method (Fig. 2, lower panel) shows that the ratio of heme-containing nNOS(hϩ) dimers to monomers was ϳ90:10 in the absence of H 4 biopterin and L-arginine (lane A, lower panel) and, thus, markedly higher than that determined by means of protein staining (see lane A, upper panel). Preincubation in the presence of exogenous pteridine and L-arginine led to virtually complete dimerization of the heme-containing protein (lane B, lower panel), demonstrating that ϳ50% of nNOS(hϩ) had lost their prosthetic heme group during electrophoresis. As expected, nNOS(hϪ) monomers did not stain for heme ( lanes C and D, Fig. 2, lower panel).
To investigate the involvement of the prosthetic heme group in the dimerization of nNOS under native conditions, we have analyzed nNOS(hϩ) and nNOS(hϪ) by means of gel filtration chromatography and static light scattering. In the course of gel permeation chromatography on Superose 6, heme-saturated (Ͼ0.90 eq of heme per monomer) and -deficient (Ͻ0.05 eq of heme per monomer) nNOS eluted at 11.6 and 13.1 ml, respectively (Fig. 3A). From these data, we calculated Stokes radii of 6.3 Ϯ 0.3 and 8.1 Ϯ 0.1 nm (n ϭ 3) for nNOS(hϪ) and nNOS(hϩ), respectively (Fig. 3B), showing that the hydrodynamic volume of nNOS(hϪ) (1.05 Ϯ 0.05 ϫ 10 Ϫ24 m 3 ) is 2.1 Ϯ 0.1-fold smaller than that of the holoenzyme (2.23 Ϯ 0.03 ϫ 10 Ϫ24 m 3 ). Together with the observation that the native, heme-saturated holoenzyme forms a 320-kDa homodimer (11,14), our results suggest that nNOS(hϪ) is a 160-kDa monomer. Alternatively, the smaller Stokes radius of nNOS(hϪ) may result from a more compact, globular conformation compared to the elongated holoenzyme, which was shown to exhibit an axial ratio of ϳ20:1 (14). This was clarified by means of static light scattering, a technique which allows us to determine the molecular mass ratio of related macromolecules in solution. The  (Fig. 3C), demonstrating that the molecular masses of nNOS(hϪ) and nNOS(hϩ) differed by a factor of 1.9 Ϯ 0.3.
To determine whether reconstitution of nNOS(hϪ) monomers with heme results in the reassociation of enzyme subunits into a homodimer, we have incubated heme-deficient monomers (4 M) with a 2-fold molar excess of hemin chloride for 30 min at ambient temperature prior to removal of unbound heme by gel filtration chromatography (Fig. 4). As estimated from the peak areas, the relative amount of dimeric nNOS was increased ϳ10-fold from ϳ5% to ϳ55% upon incubation with hemin. The inset to Fig. 4 shows that dimerization of nNOS subunits was accompanied by an ϳ10-fold increase in proteinbound heme from 20 Ϯ 9 pmol to 195 Ϯ 35 pmol (n ϭ 3) which corresponded to a heme content of 0.5 eq of per monomer. Increasing the incubation time to 60 min, varying the protein: heme ratio from 1:1 to 1:5, or co-incubation with L-arginine and H 4 biopterin (8 M each) gave essentially the same results (data not shown). Reconstitution with heme in the presence or absence of H 4 biopterin/L-arginine did not restore enzyme activity as determined by L-arginine-independent NADPH oxidation and formation of L-citrulline from L-arginine (not shown).
Interestingly, the amount of H 4 biopterin bound to nNOS(hϪ) (0.007 Ϯ 0.001 eq per monomer) was markedly reduced as compared with the heme-saturated enzyme (0.45 Ϯ 0.03 eq per monomer). To find out whether enzyme-bound heme is requisite for pteridine binding, we have performed binding studies with nNOS(hϩ) and nNOS(hϪ) using [ 3 H]H 4 biopterin as radioligand (Fig. 5A). In accordance with previous studies performed with nNOS from porcine brain (36,38), nNOS(hϩ) bound It should be pointed out that the binding assay is not quantitative and does not allow, therefore, to determine the absolute amount of bound H 4 biopterin (see "Experimental Procedures").
The apparent loss of amino acid substrate and pteridine binding sites in nNOS(hϪ) suggests that heme deficiency may be accompanied by unfolding of the protein. To address this issue, we have analyzed nNOS(hϩ) and nNOS(hϪ) by CD spectroscopy. As shown in Fig. 6, nNOS(hϩ) displayed a well defined far-UV CD spectrum with minima at 208 and 220 nm, a maximum at 192 nm, and base line crossovers at 200 nm and ϳ180 nm (solid line). Similar spectra were obtained for the heme-deficient enzyme (dashed line), and secondary structure analysis revealed that both nNOS species were virtually identical with regard to their content of parallel ␤-sheet, turns and other structures (Table I). However, nNOS(hϩ) and nNOS(hϪ) differed slightly in their content of ␣-helical structures (0.27 versus 0.34) and antiparallel ␤-sheet (0.15 versus 0.11), pointing to subtle secondary structure changes occurring upon heme binding.

DISCUSSION
It was the objective of the present study to isolate and characterize heme-free nNOS in order to find out how binding of the prosthetic heme group affects the catalytic and structural features of the enzyme. Infection of Sf9 cells with rat nNOSrecombinant baculovirus in the absence of added hemin yielded a partially heme-deficient nNOS which was purified and subjected to gel filtration chromatography for separation of the heme-free protein from the holoenzyme. Consistent with the essential role of heme in NOS catalysis, nNOS(hϪ) exhibited only marginal NOS activity, which was apparently due to contamination with holoenzyme (ϳ7% of total protein). Heme deficiency neither affected the kinetic parameters for cytochrome c reduction nor the ability of the enzyme to bind FAD and FMN, showing that nNOS(hϪ) contains a fully intact reductase domain. These data are in agreement with previous reports suggesting a bidomain structure of NOS (20,22,23,50).
Results from gel filtration chromatography and static light scattering revealed that the hydrodynamic volume and accordingly the molecular mass of nNOS(hϪ) was about half of the respective values calculated for nNOS(hϩ). Based on the identification of latter species as 320-kDa homodimer (7,14), these data demonstrate that nNOS(hϪ) is monomeric. Reconstitution of nNOS(hϪ) with hemin resulted in pronounced incorporation of heme and consequent enzyme dimerization. However, the reconstituted heme-containing dimers did not exhibit detectable NOS activity, indicating that co-translational heme binding is essential for expression of catalytically active NOS. In striking contrast with inducible NOS dimerization, which was reported to require the coincident presence of heme, H 4 biopterin, and L-arginine (29), the heme-induced dimerization of nNOS(hϪ) occurred in a pteridine-and L-arginine-independent fashion. Thus, in accordance with a recent report showing that porcine nNOS homodimers are stable in the absence of H 4 biopterin (14), our data suggest that heme is the sole cofactor controlling the assembly of nNOS subunits.
As previously shown for the enzyme purified from porcine brain (14), recombinant rat nNOS(hϩ) homodimers dissociated in the course of low temperature SDS-PAGE, unless H 4 biopterin was present to induce formation of superstable SDS-resistant dimers. However, while this effect was virtually complete with the pig enzyme, we observed only ϳ50% of rat nNOS(hϩ) dimers under comparable experimental conditions, i.e. upon preincubation with saturating concentrations of both H 4 biopterin and L-arginine. As revealed by the heme-iron staining experiments, this incomplete dimerization was apparently due to loss of heme during SDS-PAGE of nNOS(hϩ). Thus, the enzyme obtained from porcine brain seems to bind the prosthetic heme group more tightly than the recombinant rat brain NOS. We are currently investigating whether this reflects species differences in the heme environment of NOS or results from a specific feature of the baculovirus overexpression system.
There is experimental evidence that the prosthetic heme group is part of the amino acid substrate site and interacts with the pteridine binding domain. Thus, L-arginine and H 4 biopterin induce perturbations of the heme spectrum (20,23,48,51), sequences located near the axial cysteine ligand of the heme are apparently involved in L-arginine and pteridine binding (26), and the heme site inhibitor 7-nitroindazole was found to antagonize L-arginine and H 4 biopterin binding to NOS (36). In the present study, we demonstrate that nNOS(hϩ) binds H 4 biopterin and the amino acid substrate analog L-NNA with high affinity, whereas binding of these ligands to the hemedeficient protein was negligible. The minor alterations in the CD spectrum point to an effect of heme deficiency on the secondary structure of the protein. Based on the CD spectra, it cannot be ruled out that the lack of binding activity of nNOS(hϪ) is due to a loss of secondary structure, but this appears to be unlikely because most of the structural elements found in dimeric nNOS(hϩ) were conserved in the heme-deficient monomer. As compared with the holoenzyme, nNOS(hϪ) contained slightly more ␣-helical structures accompanied by some loss of antiparallel ␤-strands, indicating that heme binding and concomitant subunit dimerization involve a subtle rearrangement of secondary structure necessary for formation of appropriately folded amino acid substrate and cofactor binding sites.
In conclusion, our data support a model for dimeric nNOS assembly as shown in Fig. 7. Heme-free NOS is monomeric, contains the flavins FAD and FMN, and exhibits only cytochrome c reductase activity. Depending on the intracellular availibility of free heme, nNOS is expressed as a loosely associated homodimeric hemeprotein, which readily dissociates in the presence of SDS. In this conformation, nNOS acts as FIG. 6. CD spectra of nNOS(h؊) and nNOS(h؉). CD spectra of nNOS(hϪ) (solid line) and nNOS(hϩ) (dashed line) were recorded at 20°C with 12 M protein in 50 mM sodium phosphate buffer, pH 7.4, as described under "Experimental Procedures." The spectra shown are representative of two.

TABLE I Secondary structure of nNOS(hϩ) and nNOS(hϪ)
The CD spectra shown in Fig. 6 were analyzed for secondary structure by means of the variable selection method, using a basis set of 33 reference proteins (46). The secondary structures of nNOS(hϩ) and nNOS(hϪ) were computed from a final set of 29 combinations constructed from 29 protein spectra each and a set of 12 combinations including 24 spectra, respectively. Data represent the average structural content calculated from these selected combinations, in which the total of secondary structures was 1.00 and which gave a root-meansquare error of less than 0.20.  7. Role of heme in nNOS structure and function. nNOS(hϪ) represents a monomeric flavoprotein, catalyzing the reduction of cytochrome c and cytochrome P450 (18). The heme-deficient protein does not exhibit binding sites for L-arginine and H 4 biopterin. Upon heme binding, the H 4 biopterin-deficient enzyme forms a loosely associated homodimer, which functions as an NADPH oxidase, catalyzing the formation of superoxide and H 2 O 2 (52,53). Finally, this protein species is converted by H 4 biopterin (H 4 B) to a tight dimer exhibiting full NOS activity.
NADPH oxidase and catalyzes the formation of superoxide and H 2 O 2 (52,53). Binding of H 4 biopterin to nNOS(hϩ) further modifies the enzyme structure in that the subunits adopt a much tighter conformation resulting in the formation of superstable SDS-resistant dimers (14). This pteridine-induced conformational change, which may occur either co-or post-translationally, is requisite for the coupling of reductive oxygen activation to L-arginine oxidation and, thus, for the conversion of the enzyme from an NADPH oxidase to a fully active NOS. Accordingly, the role of the prosthetic heme group is not confined to NOS catalysis but extends to the modulation of protein structure, controlling the interaction of nNOS subunits as well as the formation of pteridine and amino acid substrate binding sites. It remains to be seen whether the expression of catalytically active nNOS dimers is limited by heme availability in certain (patho)physiological situations. If so, co-translational processing of nNOS may represent a novel mechanism for the regulation of neuronal NO biosynthesis in vivo.