Diatomic Ligand Discrimination by the Heme Oxygenases from Neisseria meningitidis and Pseudomonas aeruginosa*

Heme oxygenases have an increased binding affinity for O2 relative to CO. Such discrimination is critical to the function of HO enzymes because one of the main products of heme catabolism is CO. Kinetic studies of mammalian and bacterial HO proteins reveal a significant decrease in the dissociation rate of O2 relative to other heme proteins such as myoglobin. Here we report the kinetic rate constants for the binding of O2 and CO by the heme oxygenases from Neisseria meningitidis (nmHO) and Pseudomonas aeruginosa (paHO). A combination of stopped-flow kinetic and laser flash photolysis experiments reveal that nmHO and paHO both maintain a similar degree of ligand discrimination as mammalian HO-1 and the HO from Corynebacterium diphtheriae. However, in addition to the observed decrease in dissociation rate for O2 by both nmHO and paHO, kinetic analyses show an increase in dissociation rate for CO by these two enzymes. The crystal structures of nmHO and paHO both contain significant differences from the mammalian HO-1 and bacterial C. diphtheriae HO structures, which suggests a structural basis for ligand discrimination in nmHO and paHO.

Heme oxygenase (HO) 2 catalyzes the oxidative degradation of heme to biliverdin, iron, and CO ( Fig. 1). HO has been identified in a wide array of organisms, including mammals (1, 2), insects (3,4), and photosynthetic organisms (5,6). Of particular interest, HO is present in many pathogenic bacteria (7)(8)(9)(10), including Neisseria meningitidis (11) and Pseudomonas aeruginosa (12). These pathogenic bacteria have developed sophisticated heme uptake systems that harness the iron from hemecontaining proteins present in the host (13)(14)(15). The HO enzymes from N. meningitidis (nmHO) and P. aeruginosa (paHO) are both essential for the utilization of iron from imported heme, and the crystal structures of these two bacterial HO enzymes have now been solved (16,17). Although most HOs hydroxylate exclusively the ␣-meso heme carbon (Fig. 1), paHO is unusual because the ␥-meso carbon is the predominate site of hydroxylation (18) even though the structure paHO is very much the same as other HOs. The difference in hydroxylation patterns is due to an ϳ100 o rotation of the heme in paHO relative to other HOs which places the ␥-meso carbon at the same position in the active site as the ␣-meso carbon in other HOs (16).
The Fe(II) atom of the heme prosthetic group of heme proteins is an efficient binder of O 2 , NO, and CO, and the binding by heme proteins of these diatomic molecules is of critical importance to physiological processes such as respiration, vasodilation, and neurotransmission (19 -21). A common feature shared by all heme proteins is the need to not only bind its target ligand but also to discriminate against the binding of heme ligands of similar size and shape. Fe(II) adducts of CO are normally linear because backbonding is optimized by the overlap of Fe(II) d-orbitals with the empty * orbitals of CO (22)(23)(24). In contrast, Fe(II)-NO and Fe(II)-O 2 are naturally bent in order to maximize overlap with the occupied * electrons of the ligand with the dz 2 iron orbitals. Thus, the intrinsic binding geometries of these ligands can be utilized by heme proteins to increase ligand discrimination (25,26).
It has been suggested that ligand discrimination by globins results from the preferred ligand geometries of bound O 2 and CO (27). The greater polarity of the bound O 2 ligand also plays a role in enhancing the affinity of O 2 relative to CO for both the globins and HO enzymes. The partition constant (M), defined as the ratio of the equilibrium association constants for CO and O 2 (K CO /K O2 ), is ϳ30,000 for free heme, whereas that for sperm whale myoglobin is only ϳ40 and is further reduced to ϳ4 for human heme oxygenase (28,29). The O 2 affinities of mammalian HO-1 and HO-2 are 30 -90-fold higher than those of mammalian myoglobins (28), whereas the O 2 affinity of the bacterial HO from C. diphtheriae is 20-fold higher than that of myoglobin (29). This increased O 2 affinity is largely due to a 100-fold slower rate of dissociation from HO relative to the globins. Thus, even though the inherent affinity of the ferrous heme for O 2 is lower than that for CO, myoglobin and heme oxygenase clearly alter the binding characteristics of the heme, dramatically increasing the O 2 affinity relative to CO. Here we report the kinetic rate constants for the binding of CO and O 2 by the bacterial heme oxygenases, nmHO and paHO, and provide a structural basis for ligand discrimination.

EXPERIMENTAL PROCEDURES
Preparation of Ligand-bound Heme Oxygenases-Expression, purification, and reconstitution of the recombinant nmHO and paHO with heme were carried out as described * 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. 1 To whom correspondence should be addressed. previously (11,12). Spectral characterization and information on the stability of nmHO and paHO have been reported (11,12). Ligand binding reactions were measured using stoppedflow and flash photolysis techniques as described previously (30,31). The CO forms of the enzymes were prepared by diluting the ferrous form of the heme⅐HO complex into anaerobic buffer solutions containing known CO concentrations. Because of the rapid rate of autooxidation of ferrous iron, the oxygenated forms of the enzymes were made as follows. First, the ferric heme⅐HO complex was reduced with sodium dithionite in deoxygenated buffer. Then, the reduced form of the complex was loaded onto a column of Sephadex G-25 and eluted with buffer containing known oxygen concentrations. Formation of the complex and sample integrity were confirmed by optical absorption spectra recorded before and after flash photolysis measurements. All of the kinetic measurements were carried out in 20 mM Tris-HCl buffer, pH 7.5, at 20°C. Stock solutions of O 2 , CO, and NO were prepared by equilibrating the buffer with 1 atm of the pure gas at room temperature. If required, saturated stock solutions (1300 M O 2 , 1000 M CO, and 1800 M NO) were diluted with degassed buffer in gas-tight Sample-Lock Hamilton syringes. Association Rate Constants-Oxygen association rate constants were measured by laser flash photolysis using a pulsed dye laser system fitted with a Quantel Brillian B Nd:YAG laser. Solutions were irradiated at ϭ 540 nm (␤-band of heme) with a pulse of 10 ns and monitored at ϭ 434 nm (Fe 2ϩ soret peak). CO association rate constants were measured by stopped-flow kinetics using an Applied Photophysics SX18MV-R. Solutions were also monitored at ϭ 434 nm. Final protein concentrations for each set of experiments were 10 M. Time courses are the average of five traces, and the observed rates for O 2 and CO binding by nmHO and paHO were obtained by fitting the average traces to a single exponential expression. Second-order association rate constants were obtained from the slopes of the linear plots of the observed pseudo-first-order rate constants versus ligand concentration.
Dissociation Rate Constants-Dissociation rate constants for both O 2 and CO were measured directly by carrying out replacement reactions with an Applied Photophysics SX18MV-R. The ligands of a 10 M solution of the O 2 -and CO-bound forms of the HO enzymes were replaced by a high concentration of displacing ligand (500 M CO and 900 M NO, respectively). Solutions were measured at ϭ 410 nm and ϭ 420 nm in order to follow the absorbance changes due to the dissociation of O 2 and CO, respectively. The dissociation rate constants were then calculated from the expressions k O2 ϭ k obs (1 ϩ kЈ O2 [O 2 ]/kЈ CO [CO]) and k CO ϭ k obs (1 ϩ kЈ CO [CO]/ kЈ NO [NO]). Because kЈ NO Ͼ Ͼ kЈ CO for all heme proteins, the observed replacement rate constant (k obs ) was directly equal to the CO dissociation rate constant (k CO ).
Molecular Dynamics-Molecular dynamics simulations were carried out with Amber 8.0. Heme parameters were provided by Dr. Dan Harris (Molecular Research Institute). Partial charges for the heme and CO were taken from the ferrous⅐CO complex provided by Dr. John Straub (Boston University). The crystal structures for CO complexes of rat HO-1 (1IX4) and nmHO (1P3V) were stripped of crystallographically defined water molecules except for those in the active site near the bound CO. The solvent-stripped structure was solvated with water molecules within a 30 Å radius of the CO molecule. The entire solvated protein was energy minimized as follows. First, water and H atoms were allowed to move in a short 10-ps low temperature (50°C) MD simulation followed by 500 cycles of energy minimization for both protein and solvent. Second, both protein and solvent were allowed to move in a 10-ps 50°C MD simulation followed by 500 cycles of energy minimization. Finally, a 2-ns MD simulation was carried out at 300°C. Resi- dues within a 20 Å sphere surrounding the CO were allowed to move. Coordinate sets were saved every 10 ps, giving a total of 200 coordinate sets. During all minimizations and dynamics runs the CO molecule was constrained to the crystallographic coordinates. This was done to avoid ambiguity in modeling the Fe-C-O bond angle and tilt.

RESULTS AND DISCUSSION
Ligand Binding and Dissociation Kinetics-The association and dissociation rate constants for the binding of CO and O 2 by nmHO and paHO are compared with those of myoglobin, mammalian HO-1, and bacterial cdHO, and protoheme in Table 1. The CO and O 2 equilibrium constants (K CO and K O2 ) and the ratios of equilibrium constants (K CO /K O2 ) for nmHO and paHO are also provided in Table 1.
Bimolecular rebinding of O 2 by nmHO and paHO after flash photolysis is monophasic, and the observed rates show a first-order dependence on the O 2 concentration. The association rate constants for O 2 binding by nmHO and paHO are 5.0 and 4.5 M Ϫ1 s Ϫ1 , respectively. These values are ϳ4-fold smaller than that of myoglobin but very similar to both the mammalian and bacterial HO enzymes. The CO association reactions of both the nmHO and paHO heme complexes are also monophasic, with the observed rates showing a linear dependence on the CO concentration. Although the time courses for the binding of CO by HO-1 and cdHO were moderately biphasic and could be fit to a two-exponential expression (29), the time courses for CO binding by nmHO and paHO were observed to be monophasic and were fit to a single exponential expression at several different CO concentrations. The association rate constants for CO binding by nmHO and paHO are 1.10 and 0.90 M Ϫ1 s Ϫ1 , respectively. These values are similar to that of myoglobin, as well as those of HO-1 and cdHO. Sample traces for the CO and O 2 association reactions are provided in Fig. 2, A and C, respectively. The second-order association rate constants for the binding of CO and O 2 by  nmHO and paHO were obtained from the slopes of the linear plots of the observed rate constants (k obs ) versus ligand concentration (Fig. 2, E and F). The O 2 dissociation rate constants for nmHO and paHO are 1.5 and 1.7 s Ϫ1 , respectively. These data indicate that the dissociation rate constants for O 2 binding by nmHO and paHO are ϳ14-fold slower than that of myoglobin but are ϳ7-fold faster than those of mammalian HO-1 and bacterial cdHO. The CO dissociation rate constants for nmHO and paHO are 0.16 and 0.12 s Ϫ1 , respectively. The dissociation rate constants for CO binding by nmHO and paHO are ϳ6-fold faster than that of myoglobin but are ϳ22-fold faster than those of HO-1 and cdHO. Sample traces for the CO and O 2 dissociation reactions are provided in Fig. 2, B and D, respectively.
The oxygen equilibrium constants (K O2 ) for nmHO and paHO are 3.3 and 2.6 M Ϫ1 , respectively, which are ϳ3-fold greater than that of myoglobin but ϳ8-fold smaller than those of HO-1 and cdHO. The CO equilibrium constants (K CO ) for nmHO and paHO are 6.9 and 7.5 M Ϫ1 , respectively, which are ϳ5-fold smaller than that of myoglobin but are ϳ17-fold smaller than those of HO-1 and cdHO. It should be noted that although the CO and O 2 equilibrium constants (K CO and K O2 ) for nmHO and paHO are smaller than those of HO-1 and cdHO, the ratios of equilibrium constants (K CO /K O2 ) for nmHO and paHO are 2 and 3, respectively, which are similar to those of HO-1 and cdHO. The smaller CO and O 2 affinities by nmHO and paHO are largely due to the much faster dissociation rate constants for these enzymes compared with HO-1 and cdHO.
Structural Basis for Ligand Discrimination-The problem in understanding ligand discrimination is illustrated in Table 1. Protoheme binds CO ϳ30,000-fold more tightly than O 2 but only 40-fold more tightly in myoglobin. Steric factors were long thought to be the predominant structural underpinning of such ligand discrimination in heme proteins. Early on, Collman et al. (32) proposed that a strategically placed distal histidine in myoglobin could accommodate a bent O 2 ligand but would inhibit the binding of a normally linear CO ligand. However, in recent years structural, spectroscopic, and theoretical studies have revealed the importance of electrostatic interactions for diatomic ligand discrimination (25,33,34). In particular, it has been shown that the H-bond between the distal His in myoglobin and O 2 is quite important in controlling O 2 affinity (35).
The HO crystal structures show that steric crowding is an important factor in HO although electrostatics also is quite important. Dioxygen-bound cdHO has two strong hydrogen bonds that can preferentially stabilize the highly polar Fe⅐O 2 complex (29). One interaction is with the amide NH of Gly-139, and the other is with the distal pocket water molecule. Mutagenesis studies have also demonstrated that increasing the strength and number of hydrogen bonds donated from the distal pocket amino acids to the iron-bound O 2 decreases the rate of O 2 dissociation from myoglobin (36 -38). Therefore, the more extensive H-bonding in HO-1 compared with myoglobin explains why k off of O 2 is slower in HO-1 and thus is a major contributing factor in ligand discrimination.
Whether or not this view holds with all HOs will require additional structures of oxy complexes, but unfortunately obtaining such structures is very difficult. However, NO is a good mimic of O 2 because, like O 2 , NO prefers a bent geometry. Moreover, the structure of the rat HO-1⅐NO (39) complex exhibits a very similar H-bonding pattern as the cdHO⅐O 2 complex (Fig. 3D). Thus, a comparison between various CO and NO complexes provides some insights on O 2 /CO discrimination. Currently, the only two heme oxygenases that have crystal structures available for both the CO-and NO-bound forms are those of nmHO and rat HO-1 (39,40). As shown in Fig. 3D, the NO has three H-bonding partners, the peptide NH of Gly-143 and a water molecule which, as already noted, is very similar to the cdHO⅐oxy complex (29). The situation with nmHO is slightly different. In this case the NO forms only one H-bond, with the active site water. The peptide NH group of Gly-120 (Fig. 3B) is too far and geometry not optimal for good H-bonding. If, as we expect, the oxy complex is similar, then the missing H-bond in nmHO compared with HO-1 helps to explain why k off for O 2 is faster in nmHO than in HO-1 or cdHO.
The more significant difference between nmHO and HO-1 or cdHO is that k off for CO is ϳ18 times faster, indicating that the nmHO⅐CO complex is less stable that in HO-1. HO-1 undergoes a fairly large structural change in response to CO binding, whereas nmHO does not. To accommodate the CO ligand, which prefers a linear Fe-C-O geometry, the distal helix in HO-1 shifts by ϳ1.0 Å and the CO-bound heme shifts by ϳ1.0 Å in the opposite direction. The bent NO, however, leads to very little change.
In sharp contrast, the nmHO structure does not change when CO binds. As a result, the CO may experience less favorable electrostatic interactions than in HO-1. More specifically, in HO-1⅐CO the active site water is 2.9 Å from the CO whereas this distance is 3.1 Å in nmHO (Fig. 3, A and C). In addition, the CO is very close, 2.7 Å, to the carbonyl oxygen of Ser-117 whereas the corresponding distance is 2.9 Å in HO-1 (Fig. 3, A and C). A 0.2 Å difference is small and probably within the margin of error in comparing two structures. Thus, it is difficult to conclude that differences in steric crowding are the structural basis for the differences in CO k off . Here is where the molecular dynamics simulations provided some insights that point toward electrostatics as being an important factor in controlling the stability of the CO complex. The average MD structure (average of 100 structures) taken over the last 1 ns of the simulation superimposed on the crystal structures is shown in Fig. 4. Also shown in panels B and D are the average MD structures together with key distances. There are two important differences between nmHO and HO-1. First, note that in HO-1 the peptide carbonyl O atom of Gly-139 is tied up in a helical H-bond with Gly-144 (Fig. 4B) whereas the corresponding carbonyl O atom in nmHO (Ser-117) is not (Fig. 4D). As a result, the partial negative charge on the carbonyl O atom of Ser-117 in nmHO is not attenuated by H-bonds, thus provid- ing greater electrostatic destabilization of the CO O atom than in HO-1. Second, the water near the CO in HO-1 is closer to the CO than in nmHO, thus providing greater electrostatic stabilization of the CO O atom. Over the 2-ns simulation the water-CO distance is 4.61 Ϯ 0.97 Å in HO-1 and 5.83 Ϯ 1.29 Å in nmHO-1. There is a fair amount of fluctuation owing to large movement of solvent as expected, but it is clear that the CO is better solvated in HO-1. Thus, less electrostatic stabilization of and greater electrostatic repulsion of CO in nmHO can help to explain the faster k off for CO.
One problem with this argument is that CO is normally considered to be the least polar of heme diatomic ligands and, as such, electrostatics should not play that important a role in CO binding. However, electrostatic control of CO binding is supported by model heme studies. One comprehensive study on CO-and O 2 -bound twin-coronet porphyrins (TCPs) established that CO/O 2 ligand discrimination may be controlled by the polar effects exerted by overhanging hydroxyl groups (41). Specifically, the TCP⅐O 2 complex was stabilized by hydrogen bonding with two overhanging hydroxyl groups, whereas the TCP⅐CO complex was destabilized because of suppression of backbonding from the iron atom to bound CO. The decrease in backbonding was caused by strong negative electrostatic interactions of bound CO with the lone pairs of the hydroxyl groups in the distal pocket, which resulted in an increase in the dissociation rate of CO from TCP compared with chelated protoheme (41). These model heme studies thus support the view that the reason nmHO exhibits a slower CO k off than HO-1 is due to less favorable electrostatic stabilization of the CO complex compared with HO-1.