THE JOURNAL OF BIOLOGICAL CHEMISTRY © 1996 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 271, No. 51, Issue of December 20, pp. 32563–32571, 1996 Printed in U.S.A.

Characterization of Endothelial Nitric-oxide Synthase and Its Reaction with Ligand by Electron Paramagnetic Resonance Spectroscopy* (Received for publication, July 17, 1996, and in revised form, September 27, 1996)

Ah-Lim Tsai‡§, Vladimir Berka‡, Pei-Feng Chen‡, and Graham Palmer¶ From the ‡Division of Hematology, Department of Internal Medicine, University of Texas Medical School, Houston, Texas 77030 and ¶Department of Biochemistry and Cell Biology, Rice University, Houston, Texas 77251

Electron paramagnetic resonance was used to characterize the heme structure of resting endothelial nitricoxide synthase (eNOS), eNOS devoid of its myristoylation site (G2A mutant), and their heme complexes formed with 16 different ligands. Resting eNOS and the G2A mutant have a mixture of low spin and high spin P450-heme with widely different relaxation behavior and a stable flavin semiquinone radical identified by EPR as a neutral radical. This flavin radical showed efficient electron spin relaxation as a consequence of dipolar interaction with the heme center; P1⁄2 is independent of Ca21-calmodulin and tetrahydrobiopterin. Seven of the 16 ligands led to the formation of low spin heme complexes. In order of increasing rhombicity they are pyrimidine, pyridine, thiazole, L-lysine, cyanide, imidazole, and 4-methylimidazole. These seven low spin eNOS complexes fell in a region between the P and O zones on the “truth diagram” originally derived by Blumberg and Peisach (Blumberg, W. E., and Peisach, J. (1971) in Probes and Structure and Function of Macromolecules and Membranes (Chance, B., Yonetani, T., and Mildvan, A. S., eds) Vol. 2, pp. 215–229, Academic Press, New York) and had significant overlap with complexes of chloroperoxidase. A re-definition of the P and O zones is proposed. As eNOS and chloroperoxidase lie closer than do eNOS and P450cam on the truth diagram, it implies that the distal heme environment in eNOS resembles chloroperoxidase more than P450cam. In contrast, 4-ethylpyridine, 4-methylpyrimidine, acetylguanidine, ethylguanidine, 2-aminothiazole, 2amino-4,5-dimethylthiazole, L-histidine, and 7-nitroindazole resulted in high spin heme complexes of eNOS, similar to that observed with L-arginine. This contrasting EPR behavior caused by families of ligands such as imidazole/L-histidine or thiazole/2-aminothiazole confirms the conclusion derived from parallel optical and kinetic studies. The ligands resulting in the low spin complexes bind directly to the heme iron, while their cognate ligands induce the formation of high spin complexes by indirectly perturbing the heme structure and excluding the original axial heme ligand in the resting eNOS (V. Berka, P.-F. Chen, and A.-L. Tsai (1997) J. Biol. Chem. 272, in press). The difference in EPR spectra of

these high spin eNOS complexes, although subtle, are different for different homologs.

The free radical nitric-oxide serves as an important messenger in a wide variety of physiological and pathophysiological processes. Biosynthesis of this simple gaseous molecule is catalyzed by three different isoforms of nitric-oxide synthase (NOS).1 These isoforms share ;60% overall homology in their protein sequences, and their reaction mechanisms are also very similar (1, 2). All three isozymes contain heme, FAD, and FMN and require the participation of Ca21-calmodulin and tetrahydrobiopterin for enzyme catalysis (1– 4). Characterization of NOS by different biophysical methods including UV-Vis (5–7), electron paramagnetic resonance (EPR) (7–9), resonance Raman (10 –12), and magnetic circular dichroism spectroscopy (13) in the last few years has provided insights into the structure-function relationship and the oxidation-reduction behavior of each catalytic component. EPR is superior to UV-Vis electronic spectroscopy in providing information about heme symmetry, identifying heme ligands, and understanding the electronic structure of the heme iron. The much higher sensitivity of EPR to changes in heme symmetry than other spectroscopic methods proves to be a useful tool in delineating the structural perturbation caused by different axial ligands (14). Stuehr and Ikeda-Saito (7) pioneered the EPR study of neuronal NOS (nNOS) and firmly established the cytochrome P450 nature of the heme moiety and its magnetic interaction with the flavin semiquinone radical. Salerno and his co-workers (8) then carefully examined the effect of L-arginine and selected analogs on the EPR of nNOS. Enhancement of the spin relaxation of flavin semiquinone radical by the heme group was further evaluated by power saturation and pulsed EPR measurements on nNOS apo- and holoenzyme (9). The EPR properties of the other two subtypes of NOS have not yet been described. In this study, we report the EPR characterization of both the wild-type endothelial NOS (eNOS) and an eNOS lacking the myristoylation site prepared by specific replacement of the second residue in the primary sequence, a glycine, to an alanine in its primary sequence (G2A mutant).2 An evaluation of the effect of several heme ligands both exogenous and those arising from amino acid side chains has also been conducted. EXPERIMENTAL PROCEDURES

* This work was supported by United Public Health Service Grants GM44911 and GM21337 and Welch Foundation Grant C636. 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. § To whom correspondence should be addressed: Division of Hematology, Dept. of Internal Medicine, University of Texas Medical School at Houston, P.O. Box 20708, Houston, TX 77225. Tel.: 713-792-5450; Fax: 713-794-4230; E-mail: [email protected]. This paper is available on line at http://www-jbc.stanford.edu/jbc/

Imidazole, 2-methylimidazole, 4-methylimidazole, pyridine, 4-ethylpyridine, pyrimidine and 4-methylpyrimidine, thiazole, 2-aminothia1 The abbreviations used are: NOS, nitric-oxide synthase; eNOS, endothelial nitric-oxide synthase; nNOS, neuronal nitric-oxide synthase; CHAPS, 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate. G2A mutant, mutant with the glycine replaced by an alanine at residue number 2; CaM, calmodulin. 2 P-F. Chen, A.-L Tsai, V. Berka, and K. K. Wu, manuscript in review.

32563

32564

EPR of Endothelial Nitric-oxide Synthase

zole, 2-amino-4,5-dimethylthiazole, ethylguanidine, and acetylguanidine are from Aldrich. Potassium cyanide, L-arginine, L-lysine and L-histidine, NADPH, 7-nitroindazole, adenosine 29,39-monophosphate, CHAPS, and calmodulin were purchased from Sigma. L-Citrulline was obtained from ALEXIS Biochemicals, San Diego, and L-[2,3,4,5-3H]arginine was obtained from Amersham Corp. (specific activity, 77 Ci/mmol). Recombinant human endothelial nitric-oxide synthase (eNOS) was prepared using a baculovirus expression system as described previously with slight modification (15, 16). The cell pellets were suspended in buffer A (20 mM Tris-HCl, pH 7.5, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol, 20 mM CHAPS, 10% glycerol, 1 mM antipain, 1 mM leupeptin, 1 mM pepstatin, 1 mM phenylmethylsulfonyl fluoride) and sonicated three times for 10 s each. The 100,000 3 g supernatant was directly applied to a 29,59-ADP-Sepharose column (1.5 3 2 cm) preequilibrated with buffer A. The column was washed with 25 ml of buffer A plus 0.5 M NaCl, followed with 10 ml of buffer A and then eluted with buffer B (20 mM Tris-HCl, pH 7.5, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol, 5 mM CHAPS, 10% glycerol) plus 30 mM adenosine 29,39monophosphate. The eluate was concentrated by Centriprep-100 (Amicon) at 500 3 g and then applied to a 10-DG column (Bio-Rad). The yield was about 10 –20 mg of purified eNOS from 3 3 109 Spodoptera frugiperda cells. Soluble endothelial nitric-oxide synthase was prepared by specific replacement of the second amino acid, glycine, to alanine through oligonucleotide-directed mutagenesis described elsewhere.2 Expression and purification of this mutant eNOS were very similar to the wild-type enzyme but without detergent solubilization, and the enzyme was purified directly from the soluble fraction after the step of sonication. Details of the characterization of both wild-type and G2A mutant eNOS were presented elsewhere.2 Activity of eNOS was measured as the conversion of tritium-labeled L-arginine to L-citrulline as described previously (15, 16). A 3-min incubation of the reaction mixture at 37 °C before quenching was applied to maintain a linear production of L-citrulline. EPR was recorded mainly at liquid helium temperature and occasionally at liquid nitrogen temperature on a Varian E-6 spectrometer with an Air Product liquid helium transfer line (17). Calibration of magnetic field was done by using either a,a9-diphenyl-b-picrylhydrazyl standard (g 5 2.0037) or the stable flavin semiquinone for the high field region, and sperm whale myoglobin (g' 5 5.905) for the low field signal (18). A Hewlett-Packard HP5342 microwave frequency counter was used to record the frequency. Data obtained from progressive power saturation were fitted to log~S/P1/2! 5 2b/2 log~P1/2 1 P! 1 b/2 log~P1/2! 1 log~K!

(Eq. 1)

where P1⁄2 is the power to achieve half-saturation of the signal; b is set to 1 for inhomogeneous broadening as found for most hemeprotein samples, and K is a proportionality factor (19). In the case of low spin heme, the gmin (gx) was either measured directly, or calculated from (20, 21) gx 1 gy 1 gz 2 gxgy 2 gxgz 1 gygz 1 4gx 2 4gy 2 4gz 5 0 2

2

2

(Eq. 2)

(rather than gx2 1 gy2 1 gz2 5 16 because low spin P450 always exhibits a gmax value smaller than 3). The parameters, V and D, the rhombic and axial ligand field terms, were then calculated in units of l (the spinorbit coupling constant) using the equations derived by Taylor (22): V/l 5 gx/~gz 1 gy! 1 gy/~gz 2 gx!

(Eq. 3)

D/l 5 gx/~gz 1 gy! 1 gz/~gy 2 gx! 2 V/2l

(Eq. 4)

using an axis systems defined by Blumberg and Peisach (23). For the high spin heme, gx and gy were measured directly, and E and D were calculated using Equation 5. gz was measured directly when observable, otherwise it was calculated using Equation 6 and the previously determined values of E and D. The ratio of rhombic (E) and axial (D) ligand field components is ,0.1 for NOS (9) and most other P450s (14, 24). gx, y 5 3ge 6 24 E/D 2 168E2/~9D2!

(Eq. 5)

gz 5 ge 2 304E /~9D !

(Eq. 6)

2

2

where ge is 2.0.

2 P.-F. Chen, A.-L. Tsai, V. Berka, and K. K. Wu, manuscript in review.

FIG. 1. EPR spectra of wild-type eNOS and G2A mutant. 10 mM wild-type eNOS (Myr1) and 13 mM G2A mutant (Myr2) in the absence (Resting) or presence (1 L-Arg) of 200 mM L-arginine were used for recording these spectra. EPR conditions were: 10 mW microwave power, 10 G modulation, and 11 K. The g values of both the high spin and low spin heme are labeled by vertical lines. The high field signal of both types of heme have been confirmed using other instrument settings to minimize the mutual overlap and potential heterogeneity. RESULTS

EPR and Heme Spin State of the Wild-type eNOS and G2A Mutant—Wild-type eNOS isolated from insect cells appears to contain both high and low spin heme as revealed by its EPR spectrum (Fig. 1, top panel) and as was also reflected in the location of its Soret maximum of 398 – 400 nm2 (25). The high spin heme exhibits gmax and gmid values of 7.67 and 4.34, and the low spin heme had a gmax and gmid of 2.45 and 2.30, respectively. The gmin of the high spin heme was easily determined at high microwave power to be 1.84, while gmin for the low spin heme, around 1.87, was observed at low microwave power and high field modulation. There are in fact two low spin heme complexes resolved in the EPR measured at 0.2 mW (Fig. 3, spectrum a), each with distinctive g values of 2.54, 2.29, 1.84 and 2.42, 2.29, 1.90, respectively (Table I). Binding of L-arginine to eNOS caused a shift of the Soret maximum from 400 to 396 nm with a corresponding change of the EPR to that of a pure high spin P450 heme (Fig. 1, top panel). There was very little change of the g values for the heme, i.e. 7.65, 4.33, and 1.84, compared with resting enzyme. The amplitude of the low field resonance increased by a factor of 5 with little obvious change in line shape and no low spin signal is apparent after this treatment. On the assumption that only high and low spin forms of heme are present in the enzyme, this suggests that more than 80% of the heme present in wild-type eNOS is low

EPR of Endothelial Nitric-oxide Synthase TABLE I EPR parameters of eNOS and its complex with different ligands

1

Values in square brackets are calculated from Equation 1. Values in square brackets are calculated from Equation 6. Values for the high spin heme was calculated as 6.25 3 (gx 2 gy), and values for the low spin heme was calculated as 100 3 V/D. 4 Calculated according to Equation 3 in units of l, the spin-orbital coupling constant. 5 Calculated according to Equation 4 in units of l also. 2 3

32565

32566

EPR of Endothelial Nitric-oxide Synthase

FIG. 2. Progressive power saturation of the flavin semiquinone radical and the low- and high spin heme of eNOS. Wild-type eNOS (B), L-arginine-bound eNOS (C), and imidazole eNOS (A) complex were used for this study. The half-saturation power (P1⁄2) for the low spin heme (circles), high spin heme (triangles) and flavin radicals (squares) were obtained by a least square fit to Equation 1 and summarized in Table II. Solid lines are the fit to Equation 1. Similar results were obtained for samples at the presence (solid symbols) and absence (open symbols) of 10 mM calmodulin and 2 mM calcium. The G2A mutant exhibited the same power dependence as the wild-type eNOS for the high spin heme (data not shown). Inset of A, EPR of the stable flavin semiquinone radical. Both the wild-type and G2A mutant sample exhibited identical EPR for the flavin radical. Only the EPR of 13 mM G2A mutant sample is shown. Essentially the same spectrum was obtained for the imidazole eNOS and L-arginine-bound eNOS. EPR conditions were: 0.05 mW, 2 G modulation, and 16 K. A similar power dependence was also obtained for the resting eNOS in the presence of 320 mM tetrahydrobiopterin. TABLE II Half-saturation power (P1/2) of eNOS samples with different heme spin states Heme species Samples

eNOS eNOS eNOS eNOS eNOS eNOS

1 1 1 1 1

CaMb Imc Im 1 CaM d L-Arg L-Arg 1 CaM

Low spin

High spin

240 490 520 620

NSa NS NS NS

Flavin semiquinone

44 50 50 54 49 49

a

Not saturable in the range of our microwave power, 0 –200 mW. Concentrations of calmodulin and calcium were 10 mM and 2 mM, respectively. c Imidazole (Im) was present as 6 mM. d L-Arginine was present as 0.2 mM. b

spin. Low spin heme signals have a much smaller amplitude than high spin signals, a consequence of the derivative representation of the absorption envelopes of the two heme species (9, 14). The eNOS sample isolated from the G2A mutant cells exhibits an EPR with high spin heme being the dominant species. Approximately 60% of the total heme estimated from the amplitude of either gx or gy signal was present as high spin. The gz component is easily identified at 1.84, indistinguishable from that of the L-arginine-treated sample (Fig. 1, bottom panel). The EPR results corroborate the optical data that show the Soret peak is more blue-shifted for the resting G2A mutant than that of the wild-type eNOS.2 The relaxation behavior of the high spin and low spin heme EPR are quite different. The low spin heme showed inhomogeneous saturation with an average P1⁄2 of 0.3 mW (Fig. 2B and Table II). In contrast, the high spin heme EPR did not show any saturation even at 200 mW, the maximal output of our klystron with power leveler enabled. This difference in relaxation rate enabled a clean discrimination between the gmin originating from the high spin or low spin heme centers by selective saturation of one heme species. For example, the spectrum recorded at 0.2 mW gave mainly the low spin heme signals (Fig. 3, spectrum a) with only a small contribution from the high spin heme.

EPR of the Stable Flavin Radical and Its Interaction with Heme Centers—A prominent radical signal was observed in all isolated eNOS samples (Fig. 1) and was attributed to a flavin semiquinone radical (7–9). This EPR signal is isotropic and centered at g 5 2.004 with an overall line width of 20 G, indicative of a neutral flavin semiquinone radical (Fig. 2A, inset) (26). Double integration of this radical revealed that it represented about 15–20% of the total FMN, the flavin species proposed to generate the stable radical. A power saturation study gave a P1⁄2 of 50 mW. To evaluate the effect of heme spin states on the relaxation behavior of this flavin radical, eNOS was also prepared as the imidazole derivative containing pure low spin heme and also as the L-arginine-bound form, a pure high spin hemoprotein, for this study. These reagents did not change the EPR spectral characteristics of the flavin radical. As shown in Fig. 2 and Table II, the P1⁄2 of the flavin radical in the imidazole eNOS (Fig. 2A) and L-arginine bound eNOS (Fig. 2C) were the same as that of the resting eNOS (Fig. 2B). Addition of calcium plus calmodulin to either of these three samples did not change the P1⁄2 of the flavin radical and produced only a marginal change of the high spin or low spin heme centers (Fig. 2 and Table II). The presence and absence of tetrahydrobiopterin (data not shown) did not have any effect on P1⁄2 either, indicating the relaxation rate of the flavin radical is not influenced by the substrate binding or binding of either Ca21-calmodulin or tetrahydrobiopterin. Structural Perturbation of Heme by Different Ligands—As an extension of our recent study on the effect of various ligands on the heme optical spectral change of eNOS (43), we have evaluated the influence of selected ligands on the EPR behavior of this hemeprotein. Only wild-type eNOS was used in this study as the G2A mutant showed essentially the same EPR behavior as the wild-type enzyme, and the latter is available in larger quantities from our expression system. We have reacted each ligand with eNOS at a concentration an order of magnitude higher than its dissociation constant as determined previously by optical titration (43). In order of increasing heme rhombicity the ligands that resulted in low spin P450 complexes are pyrimidine, pyridine, thiazole, L-lysine, cyanide, imidazole, and 4-methylimidazole

EPR of Endothelial Nitric-oxide Synthase

FIG. 3. EPR of different low spin heme complex of eNOS. Spectra were recorded for the wild-type eNOS at 9 or 13 mM heme (a) and as complexes with the following individual ligands: 50 mM pyrimidine (b), 50 mM pyridine (c), 50 mM thiazole (d), 25 mM L-lysine (e), 150 mM cyanide (f), 25 mM imidazole (g), and 35 mM 4-methylimidazole (h). Spectra of the resting enzyme was recorded at 0.2 mW, and the rest were recorded at 4 mW power, 20 G modulation, and 11 K. The radical region of each spectrum was cropped for clarity. Numbers next to each arrow are the g values.

(Fig. 3 and Table I). The EPR spectra of these low spin heme complexes are more anisotropic than is the less anisotropic component of the resting enzyme. The gmax values vary from 2.48 to 2.70 compared with 2.42 found in the resting eNOS, and the gmin values decrease from 1.86 to 1.75 compared with 1.90 found for the resting enzyme. With the exception of the two imidazole ligands, each low spin heme complex exhibited a unique set of g values with the cyanide and thiazole complexes displaying much broader gmax and gmin signals. A minimum of two low spin heme species are present in the imidazole complex, as characterized by two discrete sets of g values, 2.70, 2.30, 1.75 and 2.57, 2.30, 1.83 (Fig. 6). The splitting of g 5 2.57 signal is not reproducible from batch to batch and is best treated as one single species but raises the possibility of the existence of more than two low spin heme complexes in this case. The cognate 4-methylimidazole complex showed very broad signals for gmax and gmin making it difficult to assign accurate g values for either of them. Axial (D) and rhombic (V) ligand field parameters were calculated from the g values according to Equations 3 and 4 (Table I). The resting enzyme, which has two low spin components, gave two axial components of 4.8 and 5.7 and two rhombic component of 1.8 and 2.9. The other seven low spin complexes have values of D ranging from 4.0 to 5.1 and values of V parameters ranging from 1.8 to 2.5. The rhombicity, calculated

32567

FIG. 4. EPR spectra of high spin heme complexes of eNOS. ENOS with 10 or 13 mM heme were used for these measurements. Spectra were recorded after reaction with 50 mM 4-ethylpyridine (a), 35 mM 4-methylpyrimidine (b), 0.1 mM acetylguanidine (c), 1 mM ethylguanidine (d), 4 mM 2-aminothiazole (e), 0.1 mM 2-amino-4,5-dimethylthiazole (f), 20 mM L-histidine (g), and 0.2 mM 7-nitroindazole (h). The radical signal is cropped for clarity. All spectra were recorded at 4 mW microwave power, 10 G modulation, and 11 K.

as the 100 3 V/D, of these seven low spin complexes fell in the range between 40 and 65%, relatively higher than those of the wild-type low spin heme species, 32 and 49%. The ligands that led to the formation of high spin heme complexes were 4-ethylpyridine, 4-methylpyrimidine, acetylguanidine, ethylguanidine, 2-aminothiazole, 2-amino-4,5-dimethylthiazole, L-histidine, and 7-nitroindazole. The important features of the spectra are shown in Figs. 4 and 5 and are summarized in Table I. Compared with the low spin heme complexes, the changes appear to be smaller in extent and manifested primarily as a subtle shift in all three g values following reaction with different ligands. The gmax varied between 7.70 and 7.83, gmid varied between 4.06 and 4.20, and gmin ranged from 1.81 to 1.84. However, each high spin complex was significantly different to the resting enzyme in two aspects. The first is a major change in heme spin state composition, with these heme complexes essentially present as pure high spin in contrast to a mixture of high and low spin heme found for the resting eNOS. The second difference is that the rhombicity of all these high spin complexes is larger than that of the resting enzyme as recognized by the larger separation between gmax and gmid (Fig. 4 and Table I). The degree of perturbation appears to be related to the

32568

EPR of Endothelial Nitric-oxide Synthase

FIG. 5. Shift of the EPR signal of high spin eNOS derivatives from that of resting eNOS. The low field signal shown in Fig. 4 was enlarged to show the change in line width and magnetic field position. The spectra follow the same sequence as shown in Fig. 4, with additional two spectra i and j as the eNOS treated with 200 mM L-arginine and resting eNOS, respectively.

chemical structure of each ligand. For example, the homologs acetylguanidine and ethylguanidine caused similar shifts in g value from the wild-type spectrum; the homologs 4-ethylpyridine and 4-methylpyrimidine or the two homologs of 2-aminothiazole resulted in a shift in g value of similar extent. L-Arginine, although converting the resting enzyme to a pure high spin hemeprotein, as evidenced by the sizable increase of the EPR amplitude, caused only marginal changes in the g values (Fig. 5, i and j). This insignificant shift in EPR was also reported by Galli et al. (9) and is in contrast to the results of Salerno and his co-workers (8) who observed distinct shift of the nNOS EPR signal upon arginine binding. The high spin EPR signal generated by acetylguanidine and 7-nitroindazole appears to have broader peaks for their low field signal with the 7-nitroindazole complex clearly showing heterogeneity in its EPR, indicating that more than one paramagnetic species is present in the EPR sample. This spectral heterogeneity is not a result of chemical impurity in the 7-nitroindazole for NMR, and TLC data indicate a purity greater than 99% (technical information from Sigma). DISCUSSION

The heme spin states of NOS appear to be dependent on the source of the enzyme. The proportion of low spin heme in the resting NOS varies within each subtype and between prepara-

tion to preparation of the same subtype (7–9). In our study, the EPR of wild-type eNOS is dominated by the low spin heme, whereas the soluble G2A mutant that has lost the myristoylation site is mainly high spin. There is no correlation between the heme spin state and enzyme activity.2 The substrate, Larginine, converts both forms of eNOS to pure high spin with identical EPR characteristics. The kinetics of binding with many different ligands is very similar in the two forms of eNOS2 (15, 43). We thus believe that the membrane-associated wild-type eNOS has almost identical structures in the heme and substrate binding domains as does the soluble G2A mutant eNOS. The subtle difference in their heme spin states simply reflects slight differences in the solvent accessibility of the distal ligand site. It is likely that coordination of a water ligand at the distal heme site of the wild-type enzyme was induced by detergent solubilization during enzyme preparation. The observation of two different low spin hemes in the resting enzyme has been noted for other P450 hemeproteins (27) and is attributed to different distal ligands contributed by different amino acid side chains or to the same ligand present in two different forms, such as H2O and OH2 or imidazole and imidazolate. The large difference in the microwave power dependence between the high spin and low spin hemes has been observed for many cytochrome P450 enzymes (24, 27, 28) including eNOS. The marked difference in relaxation behavior of the two species reflects the different mechanisms that dominate in each case. The low spin center relaxes via Raman relaxation and exhibits an approximate T 7 temperature dependence arising via the low frequency modulation of the orbital magnetism (41). The relaxation of high spin heme is dominated by coupling to nearby Kramer’s doublets, an Orbach process (42). For P450type high spin hemes these lie at ;7 cm21 (42). The fast relaxation of the high spin heme presumably accounts for the substantial flavin semiquinone radical signal at liquid helium temperature even at incident microwave powers as high as 10 –20 mW (Fig. 1 and Refs. 7–9). As inferred from a comparison of the line widths of flavin radicals in 16 different flavoproteins (26), the peak-to-trough separation of 20 G of the semiquinone EPR signal indicates a neutral rather than an anionic radical. Neutral flavin radicals typically have a 19 G EPR line width, in contrast to the 15 G for the anionic flavin radical; this extra width arises from the additional H at N5 in neutral flavin and can be almost eliminated by increasing the pH or using D2O buffer (26). The optical spectrum of both the wild-type and G2A mutant eNOS shows clear peaks at 500 – 650-nm region consistent with a blue neutral flavin chromophore rather than a red anion flavin radical (25). The intensity of the flavin radical is ;20% of the concentration of FMN. This amount of EPR-detectable radical exhibited no correlation with the concentration of the high spin heme (data not shown). On the other hand, EPR line shape and relaxation behavior of the flavin radical remained the same for the pure low spin imidazole eNOS complex as well as the pure high spin L-arginine-bound eNOS (Fig. 2 and Table II). Although the high spin heme, with five unpaired electrons, is more magnetic than the low spin heme, the latter might exhibit a more potent intrinsic spin-lattice relaxation rate. This might bring about a dipolar interaction between the low spin heme with the flavin radical comparable with that of the high spin heme center. Incidentally, the same P1⁄2 value was obtained by fitting the data of power saturation to Equation 1 using either one or two components in the relaxation process (data not shown). All these analyses led to the conclusion that it is not the heme spin state but most likely the heme redox potential that dictates the amount of radical formed. The dipolar interaction between the flavin semiquinone and

EPR of Endothelial Nitric-oxide Synthase

FIG. 6. Effect of pH on the low spin EPR of imidazole derivatives. EPR spectra of imidazole (a and b) and 4-methylimidazole (c and d) complexes were recorded for samples prepared at either pH 8.0 (a and c) or pH 6.9 (b and d). The samples for pH 8.0 were the same as those shown in Fig. 3, whereas the sample of pH 6.9 was prepared using 9 mM eNOS and 20 mM imidazole or 8 mM eNOS and 25 mM 4-methylimidazole.

the heme center results in a significantly enhanced spin-lattice relaxation rate. As P1⁄2 is inversely proportional to the relaxation time, P1/2 5 1/g2T1T2

(Eq. 7)

and the increase in relaxation rate by dipolar interaction with the heme center is proportional to the inverse 6th power of the distance between these two paramagnetic centers (19), ~1/T1!obs 2 ~1/T1!intrinsic 5 ~1/T1!dipolar}g26

(Eq. 8)

we would expect that a small change in physical distance between the flavin center in the reductase domain and the heme center in the oxygenase domain would exert significant effect on the relaxation behavior of the flavin radical. Thus, the lack of any convincing effect of Ca21-CaM on the saturation behavior of the flavin radical does not support the idea that Ca21-CaM binding causes a significant change in the separation between these two paramagnetic centers (29). The possibility of reorientation of the heme and flavin centers relative to each other by Ca21-CaM or shifts in the redox potential of either the heme or the flavin by calmodulin binding should be considered as alternative mechanisms for the functional role of calmodulin. Our parallel study of spectral perturbations on the optical spectra of eNOS by various heme and amino acid ligands has yielded useful information regarding the physical dimension of the distal heme ligand site and the spatial relationship between the L-arginine binding site and the heme (43). In that study, we found contrasting behavior within several cognate ligand groups. For example, imidazole and L-histidine generated type II and type I optical changes, respectively. The EPR data gave similar results with imidazole producing the low spin state while L-histidine converted eNOS to a pure high spin complex. Likewise, pyridine and pyrimidine caused type II spectral perturbations, whereas 4-ethylpyridine and 4-methylpyrimidine led to type I optical changes. Similar contrasting

32569

results were obtained for thiazole and 2-aminothiazole or Larginine versus L-lysine (Figs. 3–5). The optical data and the EPR results are in good agreement and thus confirm the contrasting spectral behavior exhibited by a family of ligands with common functional moiety. We conclude that those ligands that cause type II spectral changes and yield low spin EPR spectra are directly ligated to the heme iron. On the other hand, those ligands that result in type I spectral changes and high spin EPR are not directly coordinated to heme but rather cause a heme structural perturbation by one or more indirect mechanisms such as puckering or doming of the heme porphyrin, change of bond length between heme iron and the sulfur atom, or change of the dihedral angle between Fe-S and the N1-N3 pyrrole axis. Presumably these ligands, either because they are too bulky or because of unfavored steric factors, are unable to bind to the iron. It also appears that these indirect perturbations lead to a displacement of the original 6th heme ligand of eNOS. Dawson, Sono, and their colleagues (31, 32) have made an extensive study on the effect of ligands on the heme spin state of P450cam and found more than 20 different oxygen, nitrogen, and sulfur donors that led to the formation of low spin heme complexes. The proximal thiolate has an abnormally strong crystal field, and other than a few halide anions most other ligands produce a low spin 6-coordinate complex (31, 33). This also explains the smaller dispersion of g values caused by different ligands in P450 heme complexes (31) than in those hemeproteins that have histidine as the proximal ligand. Precautions must thus be taken in the assignment of the unknown 6th heme ligand of P450 complexes based solely on the similarities in EPR g values. One example of this difficulty in assignment is that the axial ligand at the distal position of both low spin P450cam and chloroperoxidase, each with a proximal thiolate ligand, is water (34, 35), rather than the nitrogen or carboxylate ligands proposed earlier (30, 36). With the exception of cyanide, the seven ligands that induce the formation of low spin heme complex are all nitrogen donors. Thiazole is potentially ambiguous, but “nitrogen on” rather than “sulfur on” is supported by the observation that no hyperporphyrin spectrum is observed for this complex (data not shown). The heterogeneous EPR of both imidazole ligands might be the result of two different chemical forms of the same ligand, i.e. a neutral imidazole versus imidazolate anion. The latter should give an EPR signal with less anisotropy, due to its stronger ligand field. The same situation occurs with the methylimidazole complex of eNOS. The very broad low field and high field EPR signals prohibited the accurate location of gmax and gmin (Fig. 3). Table I also reveals a big difference in gmin between that observed and that calculated using gmid, gmax, and Equation 2. The proposal that EPR heterogeneity is a consequence of the distal ligand present as two different protonated forms is not supported by the EPR changes induced by pH. With both imidazole and 4-methylimidazole, a decrease of sample pH from 8.0 to 6.9 led to an increase of the high field gmax at the expense of the low field component (Fig. 6), a change opposite to that expected from simple arguments involving the ligand field strength of the distal ligand. One possible explanation is that the two signals represent two orientations of the imidazole ring relative to the heme plane and that lowering the pH eliminates a H bond that stabilizes one of the two orientations. The Truth Diagram originated by Blumberg and Peisach (23, 24) correlates the electronic effect of the axial ligands, represented by D, with the heme rhombicity, expressed as the ratio of V/D, for various heme complexes. This diagram provides some empirical guidance for the assignment of axial ligands

32570

EPR of Endothelial Nitric-oxide Synthase

FIG. 7. Correlation between the tetragonal field and rhombicity of the low spin eNOS complexes and comparison with complexes of chloroperoxidase, P450cam, and other low spin hemeproteins with different proximal ligands. The BlumbergPeisach convention for the principal axes (23) was used. Data included in Table I on eNOS were used in this analysis and are represented by open circles. There are two entries each for the resting enzyme and the imidazole complexes. The complexes of chloroperoxidase (solid circles) included in this diagram are native, imidazole, pyridine, 4-phenylimidazole, cyanate, thiocyanate, cyanide, and formate (30); the P450cam complexes (open squares) include native, imidazole, pyridine, N-methylimidazole, 2-methylimidazole, 1-octylamine, metyrapone, indole, cyanide, and formate (30, 32); the complexes classified in the O zone (solid triangles) are leghemoglobin (alkaline form) (40), myoglobin pH 10.1, and myoglobin pH 12.8 (23), cytochrome a3 (23), catalase imidazole, and leghemoglobin phenolate with catalase imidazole as the only outlier (17); the complexes falling in the H zone are cytochrome b5, pH 12.1, and b2, pH 12 (23), hemoglobin azide, cytochrome a3 azide (23), and horse erythrocyte catalase azide, beef liver catalase azide (23); the complexes included in the C zone (open triangles) are beef heart cytochrome c, cytochrome c2, cytochrome c, cytochrome cL (M.m.), cytochrome cL (M.e.) cytochrome c551, cytochrome c6 (38); the complexes included in the B zone (reverse solid triangles) are cytochrome b5 (pH 6 –10), cytochrome b2 (pH 4.9), cytochrome a-bisimidazole, cytochrome b-bisimidazole, prostaglandin H synthase, prostaglandin H synthase imidazole, hemoglobin imidazole, and myoglobin imidazole (17) and several leghemoglobin complexes; solid squares: native, nicotinate, 5-fluoronicotinate, pyridine, methylamine, and heptylamine with the pyridine derivative having exceptionally small tetragonal parameter (40); the cyanide derivatives with proximal ligands other than thiolate (solid diamonds) include myoglobin cyanide and leghemoglobin cyanide (40), cytochrome c cyanide and hemoglobin cyanide (23). The P and O zones defined by the dashed lines are the original definition of Blumberg and Peisach (23).

and has been widely used (8, 14, 23, 33, 37). Although a number of pitfalls have been found in the use of this correlation diagram, and it should not be used indiscriminately as the only method for ligand assignment (38), many successful examples warrant its continued application. The correlation diagram constructed from the rhombic and tetragonal field strength according to Blumberg and Peisach (23) is shown in Fig. 7. The four domains that are about the same in their heme rhombicity are zone C (cytochrome c, with methionine and histidine ligands), zone B (cytochrome b, with two histidine or imidazole ligands), zone H (hemoglobin histidine, with histidine/imidazolate or histidine/azide ligands) and zone O (hemoglobin hydroxide, with histidine and hydroxide ligands). Complexes in these four fields all have histidine as the proximal ligands, thus their heme geometry is relatively similar, defined by the four pyrrole nitrogen ligands and the imidazole ligand of the histidine. Ligation of the sixth ligand modulates the tetragonal field effect with the more electronegative ligand showing the stronger influence (14). Compounds falling in the P zone having a proximal thiolate ligand and appear to exhibit a different geometry from compounds in the

other four domains. For simplicity, we selected P450 and chloroperoxidase and their derivatives containing mainly a nitrogen donor at the sixth position for direct comparison with our low spin eNOS complexes. To provide a contrast with these three thiolateligated hemeproteins, we also included a number of hemeproteins containing non-thiolate proximal ligands but with a nitrogen ligand at the sixth position. The seven eNOS low spin heme complexes (two from the imidazole complexes) in this study clustered in a distinct region between the P and O zones of low spin heme centers as originally defined by Blumberg and Peisach (23) (dashed line in Fig. 7). There is substantial overlap between the heme derivative of chloroperoxidase and eNOS and also, although to a lesser extent, between eNOS and P450cam. The region defined by these three enzymes is clearly separated from the other hemeproteins in this presentation (Fig. 7). Comparison of Fig. 7 with the original description of Blumberg and Peisach (23) reveals two major differences. There is now a significant overlap between the C and B zones and an expansion of zone P with a corresponding shrinking of zone O. This latter modification was due to substantial overlap among the three thiolate-ligated hemeproteins, P450, chloroperoxidase, and NOS in Fig. 7, and is substantiated by the extensive comparisons between P450 and chloroperoxidase of Sono and Dawson (30 –32). As pointed out previously, thiolate is a sufficiently dominating ligand that attenuates the effect of the distal ligand. One example, shown in Fig. 7, is that the cyanide derivative of all three thiolate-liganded hemeproteins falls nicely in the P zone. In contrast, cyanide significantly changes the location of b- and c-type hemeproteins in the truth diagram, because cyanide dominates when there are N or O donors at the proximal position (Fig. 7, CN2 zone). A similar logic can be applied to the significant overlap between the C and B zones. In this case, the proximal imidazole becomes the decisive ligand and significantly outweighs the influence of methionine, thus attenuating differences in electronic effects exerted by His/Met and His/His ligand pairs for some of the c- and b-type cytochromes (Ref. 38 and Fig. 7). The x-ray crystallographic data of P450cam (39) and choloroperoxidase (35) reveal the differences of the heme environment at the distal ligand pocket. P450cam and other P450 proteins have a very hydrophobic pocket for the distal heme ligand (39); by contrast chloroperoxidase resembles other peroxidases with much more polar surroundings in the distal pocket (35). The similarity in tetragonal field between eNOS and chloroperoxidase contrasted with P450cam suggests that the eNOS distal ligand has a more hydrophilic environment than regular P450. This is not unexpected considering that it must interact with the relatively polar substrates, L-arginine or N-hydroxylarginine during catalysis. In summary, these EPR studies are in strong support of our earlier optical characterization in defining changes of heme spin states and heme geometry produced by different heme and amino acid ligands. Some ligands result in low spin heme complexes and change the heme structure by directly coordinating with the heme iron. Other ligands yield high spin complexes presumably by perturbing the heme geometry by indirect mechanisms and by excluding the original distal heme ligand of the resting enzyme. The structural perturbation induced by the high spin ligands, although apparently subtle, does seem to depend on the chemical structure of the ligands. The flavin radical of eNOS is a neutral semiquinone radical. The high stability of the flavin semiquinone detected by EPR is the consequence of efficient dipolar interaction with a heme center, either the low spin or high spin heme. Calcium plus

EPR of Endothelial Nitric-oxide Synthase calmodulin did not change the saturation of the flavin radical, indicating that calmodulin binding might not bring about a significant change of the distance between the reductase and the oxygenase domains. The distal heme ligand of eNOS has a more polar environment than P450 and more closely resembles that of chloroperoxidase; this is undoubtedly dictated by its catalytic function. REFERENCES 1. Kerwin, J. F., Jr., Lancaster, J. R., Jr., and Feldman, P. L. (1995) J. Med. Chem. 38, 4343– 4358 2. Stuehr, D. J., Kwon, N. S., Nathan, C. F., Grifith, O. W. Feldman, P. L., and Wiseman, J. (1991) J. Biol. Chem. 266, 6259 – 6263 3. Feldman, P. L., Griffith, O. W., and Stuehr, D. J. (1993) Chem. Eng. News 71, 26 –38 4. Marletta, M. A. (1993) Adv. Exp. Med. Biol. 338, 281–284 5. McMillan, K., and Masters, B. S. S. (1993) Biochemistry 32, 9875–9880 6. Matsuoka, A., Stuehr, D. J., Olson, J. S., Clark, P., and Ikeda-Saito, M. (1994) J. Biol. Chem. 269, 20335–20339 7. Stuehr, D. J., and Ikeda-Saito, M. (1992) J. Biol. Chem. 267, 20547–20550 8. Salerno, J. C., Frey, C., McMillan, K., Williams, R. F., Masters, B. S. S., and Griffith, O. W. (1995) J. Biol. Chem. 270, 27423–27428 9. Galli, C., MacArthur, R., Abu-Soud, H. M., Clark, P., Stuehr, D. J., and Brudvig, G. W. (1996) Biochemistry 35, 2804 –2810 10. Wang, J., Stuehr, D. J., Ikeda-Saito, M., and Rousseau, D. L. (1993) J. Biol. Chem. 268, 22255–22258 11. Wang, J., Rousseau, D. L., Abu-Soud, H. M., and Stuehr, D. J. (1994) Proc. Natl. Acad. Sci. U. S. A. 91, 10512–10516 12. Wang, J., Stuehr, D. J., and Rousseau, D. L. (1995) Biochemistry 34, 7080 –7087 13. Sono, M., Stuehr, D. J., Ikeda-Saito, M., and Dawson, J. H. (1995) J. Biol. Chem. 270, 19943–19948 14. Palmer, G., (1979) in The Porphyrins (Dolphin, D., ed) pp. 313–350, Academic Press, New York 15. Chen, P.-F., Tsai, A.-L., and Wu, K. K. (1995) Biochem. Biophys. Res. Commun. 215, 1119 –1129 16. Chen, P.-F., Tsai, A.-L., and Wu, K. K. (1994) J. Biol. Chem. 269, 25062–25066 17. Tsai, A.-L., Kulmacz, R. J., Wang, J.-S., Wang, Y., Van Wart, H. E., and Palmer, G. (1993) J. Biol. Chem. 268, 8554 – 8563 18. Aasa, R., and Va¨nngård, T. (1975) J. Magn. Reson. 19, 308 –315

32571

19. Hales, B. J. (1993) Methods Enzymol. 227, 384 –395 20. Bohan, T. H. (1977) J. Magn. Reson. 26, 109 –118 21. Walker, F. A., Huynh, B. H., Scheidt, W. R., and Osvath, S. R. (1986) J. Am. Chem. Soc. 108, 5288 –5297 22. Tayler, C. B. S. (1977) Biochim. Biophys. Acta 491, 137–149 23. Blumberg, W. E., and Peisach, J. (1971) in Probes and Structure and Function of Macromolecules and Membranes (Chance, B., Yonetani, T., and Mildvan, A. S., eds) Vol. 2, pp. 215–229, Academic Press, New York 24. Peisach, J., and Blumberg, W. E. (1970) Proc. Natl. Acad. Sci. U. S. A. 67, 172–179 25. Chen, P.-F., Tsai, A.-L. Berka, V., and Wu, K. K. (1996) J. Biol. Chem. 271, 14631–14635 26. Palmer, G., Mu¨ller, F., and Massey, V. (1971) in Flavins and Flavoproteins, 3rd International Symposium (Karmin, H., ed), pp. 123–140, University Park Press, Baltimore 27. Tsai, R., Yu, C. A., Gunsalus, I. C., Peisach, J., Blumberg, W., Orme-Johnson, W. H., and Beinert, H. (1970) Proc. Natl. Acad. Sci. U. S. A. 66, 1157–1163 28. Sato, M., Kon, H., Kumaki, K., and Nebert, D. W. (1977) Biochim. Biophys. Acta 498, 403– 421 29. Griffith, O. W., and Stuehr, D. J. (1995) Annu. Rev. Physiol. 57, 707–736 30. Sono, M., Hager, L. P., and Dawson, J. H. (1991) Biochim. Biophys. Acta 1078, 351–359 31. Sono, M., and Dawson, J. H. (1982) J. Biol. Chem. 257, 5496 –5502 32. Dawson, J. H., Andersson, L. A., and Sono, M. (1982) J. Biol. Chem. 257, 3606 –3617 33. Hollenberg, P. F., Hager, L. P., Blumberg, W. E., and Peisach, J. (1980) J. Biol. Chem. 255, 4801– 4807 34. Goldfarb, D., Bernardo, M., Thomann, H., Kroneck, P. M. H., and Ullrich, V. (1996) J. Am. Chem. Soc. 118, 2686 –2693 35. Sundaramoorthy, M., Terner, J., and Poulos, T. L. (1995) Structure 3, 1367–1377 36. Sono, M., Dawson, J. H., Hall, K., and Hager, L. P. (1986) Biochemistry 25, 347–356 37. Chevion, M., Peisach, J., and Blumberg, W. E. (1977) J. Biol. Chem. 252, 3637–3645 38. Teixeira, M., Campos, A. P., Aguiar, A. P., Costa, H. S., Santos, H., Turner, D. L., and Xavier, A. V. (1993) FEBS Lett. 317, 233–236 39. Poulos, T. L., Finzel, B. C., and Howard, A. J. (1987) J. Mol. Biol. 195, 687–700 40. Appleby, C. A., Blumberg, W. E., Peisach, J., Wittenberg, B. A., and Wittenberg, J. B. (1976) J. Biol. Chem. 251, 6090 – 6096 41. Herrick, R. C., and Stapleton, H. J. (1976) J. Chem. Phys. 65, 4778 – 4785 42. Herrick, R. C., and Stapleton, H. J. (1976) J. Chem. Phys. 65, 4786 – 4790 43. Berka, V., Chen, P.-F., and Tsai, A.-L. J. Biol. Chem. 272, in press