Interaction of nitric oxide with human heme oxygenase-1.

NO and CO may complement each other as signaling molecules in some physiological situations. We have examined the binding of NO to human heme oxygenase-1 (hHO-1), an enzyme that oxidizes heme to biliverdin, CO, and free iron, to determine whether inhibition of hHO-1 by NO can contribute to the signaling interplay of NO and CO. An Fe(3+)-NO hHO-1-heme complex is formed with NO or the NO donors NOC9 or 2-(N,N-diethylamino)-diazenolate-2-oxide.sodium salt. Resonance Raman spectroscopy shows that ferric hHO-1-heme forms a 6-coordinated, low spin complex with NO. The nu(N-O) vibration of this complex detected by Fourier transform IR is only 4 cm(-1) lower than that of the corresponding metmyoglobin (met-Mb) complex but is broader, suggesting a greater degree of ligand conformational freedom. The Fe(3+)-NO complex of hHO-1 is much more stable than that of met-Mb. Stopped-flow studies indicate that k(on) for formation of the hHO-1-heme Fe(3+)-NO complex is approximately 50-times faster, and k(off) 10 times slower, than for met-Mb, resulting in K(d) = 1.4 microm for NO. NO thus binds 500-fold more tightly to ferric hHO-1-heme than to met-Mb. The hHO-1 mutations E29A, G139A, D140A, S142A, G143A, G143F, and K179A/R183A do not significantly diminish the tight binding of NO, indicating that NO binding is not highly sensitive to mutations of residues that normally stabilize the distal water ligand. As expected from the K(d) value, the enzyme is reversibly inhibited upon exposure to pathologically, and possibly physiologically, relevant concentrations of NO. Inhibition of hHO-1 by NO may contribute to the pleiotropic responses to NO and CO.

Nitric oxide (NO) 1 functions as a signaling molecule in a diversity of physiological responses, including vasodilation and regulation of normal vascular tone, neuronal signal transmission, cytotoxicity against pathogens and tumors, and regulation of cellular respiration (1)(2)(3). Most of these responses result from interaction of NO with the heme group of the receptor guanylyl cyclase (1). A role akin to that of NO in signaling pathways has also been postulated for CO (4). CO is produced in mammals from heme by two heme oxygenases, HO-1 and HO-2 (5)(6)(7)(8). The involvement of CO has been invoked as a factor in atherosclerosis (9), psoriasis (10), vascular constriction (11), chronic renal inflammation (12), cellular protection (13), hyperoxia-induced lung injury (14), and other physiological situations. The role of CO as an NO-like signaling molecule has received strong support from studies of heme oxygenase and nitric-oxide synthase knockouts (15), but much of the evidence, particularly that which depends heavily on inhibition of heme oxygenase by metalloporphyrins such as tin protoporphyrin IX, is tainted by ambiguities concerning the specificity of the inhibitors (16). Nevertheless, the collective evidence makes a persuasive case for at least a limited role for CO in mammalian signaling systems.
Evidence has accumulated that interactions of CO and NO may influence the physiological responses to each of these agents through interactions at the level of the biosynthetic enzymes. Thus, NO has been shown to elevate the levels of heme oxygenase-1 mRNA and protein (17)(18)(19)(20)(21)(22)(23)(24)(25), and this response appears to be mediated by a guanylate cyclase-independent mechanism that may subserve a more generalized antioxidant response (18). Conversely, CO has been reported to elevate the steady state level of NO (26), but increased levels of heme oxygenase have been shown to decrease NO concentrations, possibly by consuming the heme required for assembly of the nitric-oxide synthases (27)(28)(29).
As both NO and CO are small molecules that can coordinate to the iron in heme proteins, it is possible that at physiological concentrations NO may directly inhibit the heme oxygenases, and conversely, CO may inhibit the nitric-oxide synthases. In studies of the identity of the proximal ligand in HO-1, NO has been shown by resonance Raman, and EPR to bind to the ferrous heme iron atom (30,31). In HO-2, a heme regulatory motif binds a secondary non-catalytic heme that binds NO and, through an undetermined mechanism, inactivates the protein (32). Indirect evidence also exists for the inhibition of heme oxygenase in tissue homogenates by endogenously formed NO (33), but no focused study has been carried out of the inhibition of HO-1 by NO.
Enzymes-Catalase and glucose oxidase were from Sigma. Rat biliverdin reductase and rat cytochrome P450 reductase were purified by published procedures (34 -36). The hHO-1 construct used encoded hu-* This work was supported by National Institutes of Health Grants DK30297 (to P. R. O. M.) and GM18865 (to P. M. L.). 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.
NO Treatments-The concentrated NO donor solution stocks were prepared fresh in 0.01 N NaOH. NOC9 and Dea/NO are stable under alkaline conditions, but they decompose spontaneously when small amounts of the stock solutions are added to the standard pH 7.4 buffer. Concentrated stock solutions of the NO donors were used to minimize the changes in pH. NOC9 treatment was carried out at room temperature (ϳ25°C) and that with Dea/NO at 37°C. These temperatures were chosen so that the two NO donors had similar rates of NO release. All treatments with NO donors were performed aerobically. The NO gas solutions used in stopped-flow experiments were made anaerobic by bubbling NO into argon-equilibrated standard buffer.
The concentration of the NO donors in the enzyme binding assays was varied at a fixed enzyme concentration of 4 M. NO binding was monitored at wavelengths between 250 and 700 nm. Formation of the hHO-1-heme Fe 3ϩ -NO complex was determined from the ratio of the 416 to 404 nm absorbance. Formation of the corresponding met-Mb Fe 3ϩ -NO complex was estimated from the ratio of the absorbance at 417 to 408 nm.
Spectroscopic Characterization-The UV-visible spectra of the hHO-1 proteins were recorded in standard buffer on a Cary Varian model 1E spectrophotometer.
The enzyme concentration for RR experiments was ϳ125 M in 0.1 M potassium phosphate buffer at pH 7.4. 14 NO and 15 NO gas purchased from Aldrich was bubbled through a 0.1 M KOH solution to remove higher nitrogen oxides. Formation of the NO adduct was achieved by addition of an NO-saturated buffer solution to an argon-purged hHO-1 solution in the Raman capillary cell to reach a final concentration of ϳ1 mM NO. Before freezing the samples, the completion of the reaction was confirmed by UV-visible spectroscopy in the same Raman capillary cell using a Cary 50 spectrophotometer. Once the RR experiments were completed, the sample was thawed to obtain its UV-visible spectrum and confirm the stability of the complex during the laser illumination at 90 K.
RR spectra were obtained on a custom McPherson 2061/207 spectrograph (set at 0.67 m with variable gratings) equipped with a Princeton Instruments liquid N 2 -cooled CCD detector (LN-1100PB). Kaiser Optical supernotch filters were used to attenuate Rayleigh scattering. Excitation sources consisted of an Innova 302 krypton laser (413 nm). Spectra were collected on frozen samples kept at ϳ90 K with N 2 cold finger in a backscattering geometry (41). Frequencies were calibrated relative to indene and CCl 4 standards and are accurate to Ϯ1 cm Ϫ1 . CCl 4 was also used to check the polarization conditions.
In the FTIR experiments, the enzyme concentration was increased to ϳ1 mM using a Microcon 10 ultrafiltration device (Amicon). The concentrated hHO-1 solution was made anaerobic in a vial before exchanging the head space with pure NO gas to reach a final concentration of ϳ2 mM NO in solution. The protein sample was then injected into an IR cell consisting of CaF 2 windows separated by a 50-m Teflon spacer. The formation of the NO adduct was confirmed by UV-visible spectroscopy in the IR cell using a Cary 50 spectrophotometer.
FTIR spectra were obtained at room temperature on a PerkinElmer Life Sciences system 2000 equipped with a liquid N 2 -cooled MCT detector. Sets of 20-min accumulations were acquired at a 2-cm Ϫ1 resolution on the samples and the identical cell filled with buffer for background subtraction.
Stopped-flow Kinetic Analyses-Kinetic studies were carried out with an SF-61 DX2 double mixing stopped-flow system (Hi-Tech Scientific). The stopped-flow instrument was made anaerobic by first rinsing with argon equilibrated standard buffer, followed by an overnight incubation with anaerobically prepared catalase/glucose/glucose oxidase solution. Kinetic traces were taken at room temperature at wavelengths between 380 and 700 nm. The data were analyzed with KinetAsyst2 software and were fit to a first-exponential expression. In this experiment, 3 M ferric hHO-1-heme and 1.5 M horse met-Mb were used. The solubility of NO gas under 1 atmosphere at 20°C is ϳ2 mM. Dilution of the NO-saturated solution was made with argon-equilibrated standard buffer in gas-tight syringes (Hamilton).
Bilirubin Activity Assay-The reaction mixture contained ferric hHO-1-heme (1 M), hemin (30 M), bilirubin reductase (4 M), and cytochrome P450 reductase (0.4 M) in the standard buffer. The reaction was initiated by the addition of NADPH (400 M). The production of bilirubin at room temperature was monitored at 468 nm for 100 s. The initial rate of the reaction was calculated using the value ⑀ 468 ϭ 43.5 mM Ϫ1 cm Ϫ1 for the bilirubin product. To study the effects of NO on hHO-1 activity, two sets of experiments were carried out. In one set, biliverdin reductase, cytochrome P450 reductase, hemin, and hHO-1 were incubated with various concentrations of NOC9 for different times before NADPH was added to measure the bilirubin forming activity. In the other set, hHO-1 was incubated with 1 mM NO donor and then the rest of the assay mixture was added for the activity assay. Separately, hemin (30 M) and bilirubin reductase (4 M)-cytochrome P450 reductase (0.4 M) were separately incubated with the NO donors before the other components of the reaction system were added, and bilirubin formation was quantitated.

RESULTS
Absorption Spectrum of the hHO-1-Heme Fe 3ϩ -NO Complex-Ferric hHO-1-heme has a Soret maximum at 405 Ϯ 1 nm in 0.1 M phosphate buffer at pH 7.4. Upon addition of a 1 mM concentration of the NO donor NOC9 or Dea/NO, the ferric hHO-1-heme Soret maximum immediately shifted to 416 nm (Fig. 1). This spectroscopic shift was reversible, suggesting that the NO species generated by the NO donors interacted with the heme in hHO-1. Ferric heme alone has a very broad Soret absorption at about 380 nm, and its incubation with the NO donors caused a sharpening of the Soret band with a decrease in its absorption intensity (not shown). In contrast to the reaction with ferric hHO-1-heme, the spectral change observed with free hemin was not reversible.
RR and FTIR Spectroscopy-The ferric hHO-1-heme NO complex prepared by bubbling NO through a solution of the ferric hHO-1-heme complex was observed to be very photolabile in the resonance Raman experiments, but at 90 K the efficiency of this process was sufficiently diminished to allow the experiments to be carried out. As shown previously (31), the spectrum of ferric hHO-1 reveals a hexacoordinated high spin/ hexacoordinated low spin equilibrium with 3 , 2 , and 10 modes at 1482/1508, 1566/1585, and 1608/1640 cm Ϫ1 , respectively ( Fig. 2A). After exposure to NO, the porphyrin skeletal modes are observed at higher frequencies with 4 , 3 , 2 , and 10 at 1378, 1511, 1588, and 1645 cm Ϫ1 , respectively (Fig. 2B). These frequencies identify the NO adduct as a hexacoordinated low spin complex, although a minor hexacoordinated high spin population revealed by the weak 3 at 1483 cm Ϫ1 is assigned to some photodissociation. In heme ferric nitrosyl complexes, the (N-O) is not resonance-enhanced with Soret excitation, but it is easily observed in the FTIR spectrum. This vibration is detected at 1918 cm Ϫ1 in ferric hHO-1-heme, only 4 cm Ϫ1 lower than that in met-Mb (Fig. 3). Such (N-O) frequencies are characteristic of linear six-coordinated {Fe(NO)} 6 complexes (42). A significant difference between these two signals resides in the ϳ20-cm Ϫ1 halfwidth of this stretching mode in hHO-1 compared with the 9-cm Ϫ1 half-width observed in met-Mb. In met-Mb the configuration of the nitrosyl group is clearly defined by the presence of the imidazole ring from the distal histidine, but the absence of distal polar side chains above the heme iron of hHO-1 and a greater solvent accessibility may permit greater fluctuation of the NO ligand, resulting in substantial inhomogeneous broadening of the (N-O).
In the low frequency region of the RR spectra (Fig. 4), the identification of the Fe-N-O vibrational modes was facilitated by the use of isotopic labeling and the close similarity of these frequencies with those observed in other hemoproteins (43). In the met-Mb ferric-nitrosyl complex, the (Fe-NO) and ␦(Fe-N-O) are observed at 595 (Ϫ6) and 573 (Ϫ11) cm Ϫ1 , respectively. In the ferric nitrosyl complex of hHO-1, two bands at 596 and 588 cm Ϫ1 that shift to 590 and 573 cm Ϫ1 with 15  hHO-1 Fe 3ϩ -NO complex was found to be ϳ50-fold faster, and k off 10-fold slower, than for horse met-Mb, resulting in an ϳ500-fold tighter binding of NO. The resulting calculated K d value for the binding of NO to ferric hHO-1-heme is 1.4 M.
NO Binding to Mutant Ferric hHO-1-Heme Complexes-As shown in Fig. 5, the binding of NO to ferric hHO-1-heme can also be seen with the NO donors NOC9 and Dea/NO. The concentrations of NOC9 and Dea/NO that cause half-maximal binding of NO to ferric hHO-1-heme are ϳ80 and 100 M, respectively, although the actual concentration of NO in these experiments is much lower. Under the same conditions the binding of NO to horse met-Mb did not reach a plateau even at much higher concentrations of the NO donors, as expected from the higher K d value for this protein (Table I and Fig. 5).
The high affinity of the ferric hHO-1-heme complex for NO is unusual for a ferric hemoprotein, although the binding of NO to ferric catalase reportedly occurs with a comparably low K d of ϳ0.5 M (45, 46). In our efforts to identify specific residues that contribute to this high binding affinity, we investigated the binding of NO to hHO-1 in which individual active site amino acids had been mutated. hHO-1 residues that could stabilize distal iron ligands by hydrogen bonding or other interactions include Glu-29, Gly-139, Asp-140, Ser-142, Gly-143, Lys-179, and Arg-183. Glu-29 is close enough to form a hydrogen bond with His-25, the proximal iron ligand (47). Gly-139, Asp-140, and Gly-143 are part of a hydrogen bonding network on the distal side that interacts with distal ligands, and mutation of any one of them to an alanine results in dissociation of the distal water ligand (36,39). Mutation of Ser-142 shift the pK a value of the distal water ligand toward more basic values. 2 Lys-179 and Arg-183 appear to interact with the propionate carboxyl groups of the heme and may be important for proper orientation of the heme (47,48). We have therefore examined the binding of NO to the E29A, G139A, D140A, S142A, G143A, G143F, and K179A/R183A mutants, all of which were heterologously expressed in Escherichia coli, purified, and found to have appropriate Soret maxima (Table II). The values for halfsaturation of NO binding using NOC9 as the NO donor show that none of the mutations altered the NO binding affinity by more than a factor of ϳ2 (Table II). Thus, the high NO affinity of ferric hHO-1-heme is not very sensitive to the identities of these active site residues despite the fact that some of them interact strongly with the distal water ligand that is replaced by NO. Indeed, the high affinity for NO appears to be insensitive to the presence or absence of the distal water ligand, as the water ligand is absent in at least the G139A, G143A, and D140A mutants (36,39).
Inhibition of Bilirubin Formation-Incubation of a system consisting of ferric hHO-1-heme, biliverdin reductase, cytochrome P450 reductase, and hemin with increasing concentrations of NOC9 for various times, followed by addition of NADPH to assay bilirubin formation, clearly demonstrated that the protein is reversibly inhibited by NO (Fig. 6). Little inhibition was observed with a 1 M concentration of the NO donor at any time point, but inhibition was observed when the donor concentration was raised to 10 M. With a 50 M concentration of NOC9, bilirubin formation was completely suppressed when the assay was carried out without NOC9 preincubation, but activity was recovered when the assay was carried out after 10 min or longer of preincubation. Similar results were obtained with 100 -600 M NOC9, except that inhibition was observed after even longer preincubation periods (Fig. 6). A 100 M concentration of NOC9 is required to half-saturate the active site of ferric hHO-1-heme with NO, and this concentration therefore must roughly correspond to the K d of 1.4 M for NO itself. The fact that complete inhibition is observed with half of this NOC9 concentration, and some inhibition even at lower concentrations, suggests that inhibition may also reflect some binding of NO to the ferrous intermediate obtained when the enzyme complex is reduced by cytochrome P450 reductase. Furthermore, the inhibition, like formation of the spectroscopically determined Fe 3ϩ -NO complex, is reversible. Inhibition is therefore lost with time as the NO donor is exhausted and the NO concentration decays. This is most clearly seen in Fig. 7, in which the recovery of bilirubin forming activity has been measured as a function of the time a fixed concentration of the NO donor is preincubated with the enzyme prior to carrying out the activity assay. In this case, two different NO donors were employed, NOC9 and Dea/NO. As the figure shows, the activity of the enzyme is completely inhibited when there is no preincubation and for NOC9 even after a 20-min preincubation, but the activity with both NO donors recovers as the preincubation time is prolonged. However, only ϳ80% of the activity was recovered in these experiments. To determine whether stable NO donor decomposition products inhibit hHO-1 activity, ferric hHO-1 was incubated with solutions of decomposed NO donors for 40 min, and the bilirubin forming activity was then assayed. As shown in Fig. 8, the products of decomposition of the NO donors have some hHO-1 inhibitory activity. Only 80% of the control activity was observed in the presence of the decomposition products from a 1.0 mM concentration of the NO donors, readily explaining the recovery of only 80% of the enzyme activity in the incubations with the NO donors (Fig. 7). Control experiments in which NOC9 was added to an incubation only containing biliverdin reductase and P450 reductase, followed by addition of hHO-1, heme, and NADPH resulted in negligible inhibition, confirming that hHO-1-heme is the site of inhibition. High concentrations of NOC9 added to heme alone prior to addition of the other components of the catalytic system also inhibited turnover (results not shown), in accord with the finding that the spectrum of heme in solution was altered by the NO donors. DISCUSSION NO, unlike O 2 and CO, can bind to the iron atom of hemoproteins in both the ferric and ferrous states, although the binding affinity is generally much higher for the ferrous state (49,50). For example, at pH 7.4 and 25°C, the K d values for the binding of NO to ferrous deoxymyoglobin and ferric met-Mb are 7 ϫ 10 Ϫ6 and 905 M, respectively, a difference of roughly 10 8 (49,51). The binding of NO to ferric hemoproteins, however, occurs over a considerable range, the tightest binding reported being K d ϭ 0.5 M for catalase (45,46). The binding of NO to the ferric hHO-1-heme complex with K d ϭ 1.4 M thus approaches the tightest binding so far observed for any ferric hemoprotein. Detailed analysis shows that this high affinity, when compared with the ϳ500-fold lower affinity for met-Mb, is due to a 50 times faster k on and a 10 times slower k off for NO (Table I)  The basis for this difference in binding affinity is unclear. The overall vibrational characterization of these {Fe(NO)} 6 structures in hHO-1 and met-Mb demonstrates that these complexes share the same bonding geometry and strength. Resonance Raman studies show that the hHO-1-heme Fe 3ϩ -NO complex is 6-coordinated low spin (Fig. 2), as is the met-Mb complex (30). Furthermore, FTIR shows that the (N-O) band of the hHO-1 Fe 3ϩ -NO complex is at 1918 cm Ϫ1 , a value only 4 cm Ϫ1 lower than for the met-Mb complex (Fig.  3) (52). This suggests that coordination of the NO to the heme iron is similar in both hemoproteins, although the broader bandwidth observed with the hHO-1 complex suggests that the NO has greater mobility in that protein. Generally, as is the case for the binding of NO to ferric hemoproteins, these results strongly argue that the NO is bound perpendicular to the heme face rather than at an angle (53)(54)(55)(56), in contrast to its orientation when bound to ferrous hemoproteins. Furthermore, in both met-Mb and ferric hHO-1-heme, the iron in the absence of NO is coordinated to a water molecule. The pK a values for deprotonation of these iron-coordinated water molecules are similar in both proteins (30,57), again indicating similar coordination states and environment and suggesting that the energy cost of displacing the water ligand should be similar in both proteins.
In view of the similarities in coordination properties, the differences in the K d for binding of NO to ferric hNO-1-heme and met-Mb presumably stem from other differences in the active sites of the two proteins. The ϳ50-fold increase in k on rate for NO in hHO-1 compared with met-Mb may reflect a higher steric hindrance in the distal pocket of met-Mb. In both these ferric proteins the sixth iron coordination site is occupied by a water molecule that must be displaced for NO to bind, but displacement of the water molecule in hHO-1 may require relatively little protein side chain rearrangement relative to that which occurs in met-Mb (58). In an attempt to identify residues that might contribute to the unique binding affinity of NO for ferric hHO-1-heme, we have mutated seven residues that could contribute to this affinity. However, in no instance did the mutation cause more than a 2-3-fold change in the observed K d value. Mutations of Asp-140 and Gly-143 slightly decreased the affinity; those of Gly-139 slightly increased it; and those of Glu-29, Ser-142, and Lys-179/Arg-183 did not significantly alter it (Table II). Even though these residues, particularly Asp-140, Gly-139, Ser-142, and Gly-143, have been shown by mutagenesis and the crystal structure to interact with the distal water ligand (36,39,47), they do not appear to be key determinants of the high NO affinity of ferric hHO-1-heme.
The unusually low K d value for the binding of NO to ferric hHO-1-heme suggests that the catalytic turnover of the enzyme should be inhibited by NO. Furthermore, as the catalytic cycle of hHO-1 traverses the ferrous state, NO could bind not only to the ferric but also to the ferrous protein, again inhibiting the enzyme. The K d value for the binding of NO to the ferrous protein is expected to be much lower than that for the ferric enzyme, but the K d value for binding to the ferric enzyme is   (Fig. 7). These concentrations of NO donors provide a sufficient flux of NO to inhibit completely the enzyme at short periods (10 -20 min) of incubation, but the inhibition is lost at longer incubation times as the NO donors are exhausted and the NO concentration falls. Inhibition by NO is reversible (Fig. 7), although a full recovery of the hHO-1 activity is not observed due to a residual inhibition caused by persistent NO donor decomposition products distinct from NO (Fig. 8).
The inhibition of heme oxygenase by NO and NO donors has received little attention (32,33). A possible role for NO as an in vivo inhibitor of heme oxygenase is suggested by the report that L-arginine, the substrate of the nitric-oxide synthases, inhibits heme oxygenase activity when added to spleen or brain homogenates, whereas analogous addition of L-NAME, an inhibitor of the nitric-oxide synthases, stimulates the heme oxygenase activity (33). In the only molecular level study prior to this work, Ding et al. (32) reported that the NO donors 3-morpholinosydnonimine (SIN-1), S-nitroso-N-acetylpenicillamine (SNAP), and sodium nitroprusside inhibit HO-2 but not HO-1. The study focused on the possible role of the non-catalytic heme binding domains that are only present in HO-2. However, in that study the NO donors were removed by dialysis prior to measuring the residual catalytic activity, so that the reversible inhibition by NO reported here would not have been detected because the NO would be removed at the same time as the NO donors. Furthermore, Ding et al. (32) reported that enzyme inactivation was not observed with HO-2 when the cysteine residues within the two heme-binding motifs had been mutated to alanines. The irreversible inhibition they observed therefore reflects modifications related to the presence of the two cysteines, neither of which is present in HO-1. This is consistent with the present findings that HO-1 is not irreversibly inactivated by NO or NO donors.
In sum, the present results clearly establish that NO inhibits HO-1 by binding to the catalytic iron atom and show that this inhibition can involve binding to even the ferric heme complex due to its unusually high affinity for NO. The inhibition of heme oxygenase by NO occurs at concentrations of this agent that are pathologically and possibly physiologically accessible. The cross-talk between the nitric-oxide synthase and heme oxygenase systems may play some role in the pleiotropic re-FIG. 6. Effects of NO donors on the hHO-1 bilirubin forming activity. The reaction system consisting of hHO-1, ferric heme, biliverdin reductase, and cytochrome P450 reductase was preincubated with increasing concentrations of NOC9 for the indicated times, after which NADPH was added, and the formation of bilirubin was measured. The preincubation and activity assays were both carried out at room temperature. sponses that are associated with these two hemoprotein signaling systems.