Involvement of NADP(H) in the Interaction between Heme Oxygenase-1 and Cytochrome P450 Reductase*

Heme oxygenase-1 (HO-1) catalyzes the physiological degradation of heme at the expense of molecular oxygen using electrons donated by NADPH-cytochrome P450 reductase (CPR). In this study, we investigated the effect of NADP(H) on the interaction of HO-1 with CPR by surface plasmon resonance. We found that HO-1 associated with CPR more tightly in the presence of NADP+ (KD = 0.5 μm) than in its absence (KD = 2.4 μm). The HO-1 mutants, K149A, K149A/K153A, and R185A, showed almost no heme degradation activity with NADPH-CPR, whereas they exhibited activity comparable to that of the wild type when sodium ascorbate was used. R185A showed a 100-fold decreased affinity for CPR compared with wild type, even in the presence of NADP+ (KD = 36.3 μm). The affinities of K149A and K149A/K153A for CPR were decreased 7- and 9-fold (KD = 16.8 and 21.8 μm), respectively. In contrast to R185A, the affinities of K149A and K149A/K153A were improved by the addition of NADP+ (KD = 5.2 and 9.6 μm, respectively), as was the case with wild type. Computer modeling of the HO-1/CPR complex showed that the guanidino group of Arg185 is located within the hydrogen bonding distance of 2′-phosphate of NADPH, suggesting that Arg185 contributes to the binding to CPR through an electrostatic interaction with the phosphate group. On the other hand, Lys149 is close to a cluster of acidic amino acids near the FMN binding site of CPR. Thus, Lys149 and Lys153 appear to interact with CPR in such a way as to orient the redox partners for optimal electron transfer from FMN of CPR to heme of HO-1.

Recently, the crystal structures of the heme complexes of human HO-1 (16), rat HO-1 (17), Neisseria meningitidis HO (18), and Corynebacterium diphtheriae HO (19) have been determined and found to have similar overall structures. The structure of rat HO-1 (17) consists of eight helices, A through H; heme is sandwiched between the proximal A helix (Leu 13 -Glu 29 ), the B helix (Glu 32 -Gln 38 ), and the kinked distal F helix (Leu 129 -Met 155 ). The His 25 side chain in the A helix contributes the proximal heme ligand. The distal heme pocket, a relatively narrow cavity, is responsible for the strict discrimination between the distal heme ligands (20 -22) and accommodates a hydrogen-bonded network that plays an important role in catalysis (23)(24)(25)(26). The distal and proximal helices are flexible or disordered in the heme-free state (27,28) and, to a lesser degree, in the product-bound state (29). These helices, along with the G helix (Pro 175 -Met 186 ) juxtaposed to the N-terminal half of the F helix, provide a positively charged protein surface surrounding the bound heme that seems important for electrostatic interactions of HO-1 with the redox partner, CPR.
CPR is a 78-kDa microsomal protein that contains one molecule each of FMN and FAD and mediates electron transfer from NADPH, via FAD and FMN, to HO and other physiological electron acceptors, including cytochrome P450s (30), cytochrome b 5 (31), and fatty-acid elongase (32). Electrons can also be transferred to cytochrome c in vitro. The crystal structure of rat CPR (33) has revealed that the molecule is composed of several structural domains, including the FAD/NADPH and the FMN domains homologous to ferredoxin-NADP ϩ reductase and to flavodoxin (34), respectively. The cofactors lie in the middle of the bowl-shaped surface. Cross-linking (35), chemical modification (36 -38), and site-directed mutagenesis (39 -42) studies have provided ample evidence for electrostatic interactions between CPR and cytochrome P450s, although some contradictions as to the role of the intermolecular charge pairing remain unsolved (43,44). A study of the crystal structure of the complex between the heme-and FMN-binding domains of bacterial cytochrome P450 BM-3, in which the heme-containing P450 domain and the FMN/FAD-containing reductase domain are linked together in a single polypeptide, has shown that the FMN domain is positioned at the proximal face of the heme domain (45). In addition, the hydrophobic membrane-anchoring region at the N terminus of CPR is essential for the recognition of cytochrome P450s (46 -48).
Less is known about interactions between HO and CPR. In contrast to cytochrome P450, the HO activity can be reconstructed with the soluble forms of HO-1 and CPR, which lack their respective membrane binding regions (49). Very recently Wang and Ortiz de Montellano (50), using mutagenesis of human HO-1, have specified some basic surface residues of HO-1 that are involved in association with CPR and have also found that the binding site for CPR overlaps that for BVR to some extent. BVR is known to accelerate the release of biliverdin from HO (51). These findings have prompted us to hypothesize that there might be some kind of regulation of the proteinprotein interactions between HO and its protein partners, CPR or BVR, either by substrate binding or by product release. In the present study, we investigated the effect of NADP(H) on the binding between the soluble forms of rat HO-1 and rat CPR using surface plasmon resonance (SPR). Our kinetic data clearly showed that the binding is enhanced by addition of NADP ϩ . Furthermore, using site-directed mutagenesis we have identified surface amino acids of HO-1 that are essential for binding to and/or electron transfer from CPR.
Enzymes-The soluble form of rat HO-1 lacking the 22-amino acid C-terminal hydrophobic stretch, called wild type HO-1 in this study, was expressed in E. coli and purified as described previously (52). The mutants were generated from the wild type with the Stratagene QuikChange site-directed mutagenesis kit (La Jolla, CA). The mutations were confirmed by DNA sequencing. The homogeneity of the purified enzymes was determined by SDS-PAGE. The ferric heme complexes of the wild type HO-1 or its mutants were purified by hydroxylapatite column chromatography after mixing 1.2 eq of heme with the enzyme (52). Heme binding studies were carried out by difference absorption spectroscopy as previously described (53). Aliquots of heme (0.01-2 M) were added to both sample (1 M HO-1 or its mutants) and reference cuvettes at 25°C. Spectra were recorded after each addition of heme. By plotting the absorbance at 405 nm against the amount of heme added, titration curves were constructed and analyzed using Deltagraph (DeltaPoint, Monterey, CA). A recombinant rat liver CPR that lacks the N-terminal 57 hydrophobic amino acids was expressed and purified as previously described (49). The recombinant CPR contained one molecule each of FAD and FMN (49). Thus CPR was quantified on the basis of the quantity of flavin released from the enzyme by boiling, using an extinction coefficient of 23.5 mM Ϫ1 for the absorbance at 450 nm (54). The purification of BVR from rat liver was carried out according to published procedures (55), including 2Ј,5Ј-ADP-Sepharose column chromatography. The HO activity was determined from the rate of bilirubin formation, which was monitored by the absorbance increase at 468 nm (49).
Binding Affinity Measurements-The interaction between the heme complex of HO-1 and CPR was determined by SPR technology using a BIAcore 1000 instrument (BIAcore AB, Uppsala, Sweden). Rat CPR protein possesses seven cysteine residues, and at least one of them, namely Cys 228 , is located on the molecular surface (33). CPR was immobilized through thiol groups of cysteine residues with the aid of a coupling reagent, 2-(2-pyridinyldithio)ethaneamine. NADPH was included during the immobilizing reaction to protect the cysteine residues near the NADPH binding site and to stabilize the enzyme. To minimize nonspecific binding and mass transfer, the amount of the coupled CPR on the sensor chip was controlled at 2,000 resonance units (2 ng/mm 2 ). Modification of CPR by 2-(2-pyridinyldithio)ethaneamine did not affect the enzyme activity.
Binding analysis was performed at a flow rate of 10 l/min at 25°C. Before loading the HO-1 protein, the chip was equilibrated for at least 30 min with a running buffer, which contained 10 mM HEPES, pH 7.4, plus 150 mM NaCl, 3 mM EDTA, and 0.005% of the surfactant P-20. To examine the effect of NADP ϩ or NAD ϩ , the cofactor at several concentrations (10,20,50, and 100 M) was included in the running buffer. The heme complex of the wild type HO-1 was diluted with the running buffer to 0.2, 1.0, and 2.0 M, and an aliquot (30 l) of each protein solution was injected over the chip surface. Binding analysis for the HO-1 mutants was also performed with 30 l of the 1.0 M enzyme solution. To regenerate the chip, a 3-min dissociation run was performed with the running buffer. Control experiments were performed on CPR-free surface. The blank sensorgram was subtracted from the assay curves using BIA-evaluation 3.1 software (BIAcore), which was also used for determination of the kinetic constants for the interaction. All kinetic parameters were determined from three independent experiments. A competition between CPR and BVR for binding to HO-1 was analyzed using a similar protocol. The heme complex of the wild type HO-1 (1 M) and BVR (50 and 100 M) were injected together onto the CPR-immobilized sensor chip equilibrated with the running buffer in the presence or absence of 100 M NADP ϩ .
Measurements of Single Turnover HO Reaction-Single turnover HO reactions of the wild type and the mutant enzymes were monitored by optical absorption changes with a JASCO V-560 UV-visible spectrophotometer at 25°C. In general, the standard reaction mixture (100 l) contained 5 M heme complex of the wild type or its mutant enzymes, 40 nM CPR and 25 M NADPH (5 molar equivalents to the heme complex) in 0.1 M potassium phosphate buffer (pH 7.4). Spectra were recorded over the range of 300 -900 nm. When ascorbate was employed as reductant, 20 mM sodium ascorbate was added in place of NADPH-CPR.

Measurement of the Reduction Rates of Ferric Heme Complex of HO-1 and Its Mutants-
The reaction mixture (100 l) for the CPR-supported reaction containing 5 M heme complex of the wild type HO-1 or its mutants, and 4 nM CPR in 0.1 M potassium phosphate buffer (pH 7.4) was placed in a sealed quartz cuvette. To stop the reaction completely at the stage of the ferrous heme state, the solution and the gas phase in the cuvette were saturated with 100% CO and then 2 l of 1.5 mM NADPH was added by syringe. The spectral changes were monitored at 10-s intervals. In the ascorbate-supported reaction, 2 l of 1 M sodium ascorbate was added in place of NADPH-CPR. The reduction rate constant was determined by fitting the spectral changes at 405 (ferric form) and 421 nm (CO-reduced form) with a single-exponential function using Cary WinUV software (Varian, Palo Alto, CA) Computer Modeling-The program Hex was used to predict the structure of the complex of rat HO-1 and rat CPR. This program is an interactive protein docking and molecular superposition program and supports a Fourier-based calculation of electrostatic interaction energies (56). The program was run on a Unix-based SGI octane R12000 computer workstation, which has a clock speed of 300 MHz. Briefly, a docking calculation was started from a randomly oriented rat HO-1 (Protein Data Bank code 1DVE) placed over rat CPR (Protein Data Bank code 1AMO). Roughly 72 million trial orientations were generated by spinning the molecules (5 rotational degrees of freedom), and by varying the intermolecular distance in 0.5-Å steps (1 degree of freedom). These orientations were scored using shape correlations to n ϭ 16, and the best 20,000 orientations were then passed to a high resolution scoring stage, which used both shape and electrostatic correlations at n ϭ 30. The best 1,000 of the n ϭ 30 orientations were then refined using a soft molecular mechanics rigid body minimization algorithm. After minimization, the lowest energy solution was selected from 283 docking solutions.

Effect of NADP ϩ on the Association of HO-1 with CPR-
The interaction between the ferric heme complex of HO-1 and CPR was examined using an SPR technique. CPR was immobilized on the dextran surface through the thiol groups of cysteine residues. Fig. 1 shows representative association and dissociation curves with three different concentrations (0.2, 1.0, and 2.0 M) of the HO-1 complex in the absence (Fig. 1A) and presence ( Fig. 1B) of 100 M NADP ϩ . In both conditions, typical association and dissociation phases were observed by real-time measurements of the interaction between HO-1 and CPR. These binding curves clearly indicated that HO-1 binds more tightly to CPR in the presence of NADP ϩ than in its absence. By analyzing the association and dissociation phases of the binding curves, the association (k a ) and dissociation (k d ) rate con-stants as well as the dissociation equilibrium constant (K D ) were estimated (Table I). The k a value in the presence of NADP ϩ was ϳ5-fold larger than that in the absence of NADP ϩ , whereas the k d values were almost identical under both conditions. Hence, the K D value for the binding of HO-1 to CPR in the presence and absence of NADP ϩ were 0.49 and 2.4 M, respectively. These results indicate that NADP ϩ increases the affinity of HO-1 for CPR by accelerating the association rate. Fig. 2 shows remarkable NADP ϩ -dependent changes in the sensorgrams. The affinity of HO-1 for CPR increased with increasing concentration of NADP ϩ (0 -50 M) and reached a maximal level with 100 M NADP ϩ ( Fig. 2A). In contrast, even 100 M NAD ϩ had little effect on the binding of HO-1 to CPR (Fig. 2B); the K D value for the binding of HO-1 to CPR in the presence of 100 M NAD ϩ was calculated to be 2.0 M, which was comparable to that in the absence of NADP ϩ . The enhancement effect of 2Ј,5Ј-ADP on the affinity of HO-1 for CPR was recognized although it was weak; indeed the K D value in the presence of 2.0 mM 2Ј,5Ј-ADP was 1.0 M. On the other hand, ADP that bears no 2Ј-phosphate group did not show any significant effect on the binding of the two proteins. Thus, the phosphate group of NADP ϩ must be crucial for the association between HO-1 and CPR. Furthermore, we investigated the effect of NADP ϩ on the interaction between HO-1 and CPR in the presence of BVR, a competitor of CPR binding to HO-1 (50). As shown in Fig. 3A, BVR (50 M) completely inhibited the binding of HO-1 to CPR in the absence of NADP ϩ . However, in the presence of NADP ϩ (Fig. 3B), BVR, even at a concentration of 100 M, caused only a partial inhibition of the binding. This result indicates that NADP ϩ is able to enhance the affinity of CPR for HO-1 irrespective of the presence or absence of BVR, suggesting a physiological role of NADP(H) in the interaction between HO-1 and CPR.
Properties of HO-1 Mutants-To search for surface ionic residues of the HO-1 protein that play key roles in binding to CPR, we prepared Ala mutants of Lys 18 , Lys 22 , Arg 27 , Arg 35 , Lys 39 , Lys 149 , Lys 153 , Lys 177 , and Arg 185 . These residues are 1) relatively conserved among species, 2) positively charged, and 3) exposed to the molecular surface. SDS-PAGE analysis indicated that the purified mutant enzymes were highly homogeneous and had the expected molecular weights (data not shown). The ratio of heme bound to the mutants was calculated by difference absorption spectroscopy; all purified HO-1 mutants formed ferric heme complexes by binding a single equivalent of heme. The absorption spectra of the heme complexes of all mutants were very similar to that of the wild type complex. The heme dissociation constants (K d ) for the mutants were comparable to that for the wild type (Table II), indicating the heme complexes of these mutants were folded properly.
We next compared the HO activities of the wild type and the mutants with two reducing systems: NADPH-CPR and sodium ascorbate (Table II). With NADPH-CPR, most of the Ala mutants exhibited enzyme activity substantially similar to that of the wild type (70 -103%) except for the Lys 149 , Lys 153 , and Arg 185 mutants. K149A and K153A retained only 3 and 21%, respectively, of the wild type activity, and the double mutant, K149A/K153A, showed no activity. R185A also had almost completely lost activity (4%). With ascorbate, however, all mutants exhibited HO activity comparable to that of the wild type. These results suggested that the mutations of Arg 185 , Lys 149 , and Lys 153 did not alter the active site structure of HO-1 but rather affected the interaction between HO-1 and CPR and/or the intermolecular electron transfer process.
Interactions between HO-1 Mutants and CPR in the Absence or Presence of NADP ϩ -To investigate the binding affinities of HO-1 mutants (K18A, K22A, K149A, K149A/K153A, and R185A) for CPR, we performed SPR analysis (Table I). K18A and K22A showed binding properties similar to that of the wild type, consistent with the result of the enzyme activity data. In the absence of NADP ϩ , the K149A, K149A/K153A, and R185A mutants, which exhibited almost no enzyme activity with NADPH-CPR, showed much lower affinities for CPR than that of the wild type; the respective K D values of K149A, K149A/ K153A, and R185A were 17, 22, and 44 M. These are 7-, 9-, and 18-fold larger than that of the wild type, mainly due to the markedly smaller association rate constants. In the presence of  NADP ϩ , the binding affinities of K149A and K149A/K153A for CPR were increased 2-to 3-fold (K D ϭ 5.2 and 9.6 M, respectively), due to acceleration of the association rates. It should be noted that the enhancement of the affinity of these mutants for CPR brought about by addition of NADP ϩ was roughly comparable to that observed with the wild type. Unlike K149A and K149A/K153A, the binding affinity of R185A for CPR was not changed by the addition of NADP ϩ ; the respective K D values in the absence and presence of NADP ϩ were 44 and 36 M. These results clearly indicated that Lys 149 , Lys 153 , and Arg 185 are crucial for the interaction of HO-1 with CPR and that, Arg 185 in particular, must be involved in the binding mechanism mediated by NADP ϩ .
Single Turnover Reactions of Heme Complexes of HO-1 Mutants-To further investigate the effect of these mutations on heme degradation, we measured absorption spectral changes during the HO reaction in air. The single turnover reaction of the wild type heme complex with NADPH-CPR is shown in Fig.  4A. The wild type heme complex changed immediately to the ferrous oxy-form upon the addition of NADPH (25 M, 5 eq to the heme complex), and then was transformed to biliverdin as indicated by the decrease in the Soret band and the increase in absorption around 670 nm. The heme bound to the wild type was completely degraded to biliverdin within 30 min.
The single turnover reactions of the heme complexes of K149A, K149A/K153A, and R185A with NADPH-CPR are shown in Fig. 4 (B-D, respectively). The spectrum of the heme complex of K149A changed rapidly to its oxy-form following addition of NADPH (25 M). Although a decrease in absorption at 340 nm indicated consumption of NADPH, the broad but distinct absorbance around 670 nm characteristic of biliverdin formation was not observed within 30 min (Fig. 4B). However, with excess NADPH (50 M, 10 eq to the heme complex), the heme in K149A was completely degraded to biliverdin within 30 min (data not shown). A more drastic consequence of mutation was found in the HO reaction of K149A/K153A (Fig. 4C). The spectrum of the heme complex of K149A/K153A changed more slowly to its oxy-form upon addition of NADPH (25 M), but no further heme degradation occurred. The spectral changes showed two clear isosbestic points, at 521 and 590 nm, suggesting that this reaction is a single process converting ferric heme to its oxy-form. Finally, about 5% of heme in K149A/K153A was decomposed by nonspecific reactions not  leading to biliverdin formation. The heme complex of R185A showed a reactivity similar to that seen with the heme complex of K149A (Fig. 4D). On the other hand, when ascorbate was used as a reductant, all the heme bound to the wild type, and the K149A, K149A/ K153A, and R185A mutants was converted into biliverdin (Fig.  5, A-D) as expected from their HO activities with ascorbate (Table II).
Reduction Rates of Heme Bound to HO-1 Mutants-To explore electron transfer from CPR to HO-1, we compared the rate of reduction of the ferric heme bound to the wild type with those of mutant enzymes (Table III). With NADPH-CPR, the rate of reduction of the heme in R185A was about half (44%) that in the wild type. The reduction rates for K149A and K149A/K153A were further decreased to 38 and 20% of that for the wild type, respectively. On the other hand, with ascorbate, the reduction rates for the mutants were comparable to that for the wild type. These findings again demonstrated crucial roles of these residues in the HO reaction with NADPH-CPR.
Computer Docking Model of HO-1 and CPR-Computational modeling analysis was attempted to identify the surface contact regions in the complex of HO-1 and CPR. The x-ray structure of rat HO-1 (Protein Data Bank code 1DVE) and rat CPR (Protein Data Bank code 1AMO) were used as input to the program Hex (56). The lowest energy solution was selected from the 283 candidate docking solutions (Fig. 6). In this model, the distance between the heme iron of HO-1 and the isoalloxazine ring of FMN of CPR is about 19 Å (Fig. 7A), which is similar to that seen in the crystal structure of cytochrome P450 BM-3 where the FMN ring is 18.4 Å away from the heme iron (45). The heme and flavin planes are nearly perpendicular. The docking surface of HO-1 consists of the A-, B-, F-, and G-helices. Interestingly, the guanidino group of Arg 185 of HO-1 is located 2.9 Å from the 2Ј-phosphate on the adenosine moiety of NADP ϩ (Fig. 7B). On the other hand, the ⑀-amino group of Lys 149 of HO-1 is located 4.9 Å from the ribityl moiety of FMN (Fig. 7A). Furthermore, the Lys 149 residue is surrounded by two acidic clusters of CPR, Asp 207 -Asp 208 -Asp 209 and Glu 213 -Glu 214 -Asp 215 . Notably, the side chain of Lys 149 of HO-1 is close enough to form a hydrogen-bond with the side chain of Asp 207 of CPR. These two residues, Lys 149 of HO-1 and Asp 207 of CPR, may interact with each other in the electrostatic interaction as well as in the intermolecular electron transfer process between CPR and HO-1.

DISCUSSION
By means of the SPR technique, we found that rat HO-1 associates with rat CPR more tightly in the presence of NADP ϩ (K D ϭ 0.5 M) than in its absence (K D ϭ 2.4 M). Thus NADP ϩ increases the binding affinity of HO-1 for CPR about 5-fold. Unlike NADP ϩ , NAD ϩ did not have the capability of enhancing the binding of HO-1 to CPR, indicating that the presence of the 2Ј-phosphate of NADP ϩ is likely to be important for the interaction between the two proteins. This is supported by the observation that 2Ј,5Ј-ADP but not ADP enhanced the binding of HO-1 to CPR. The phosphate group is essential for CPR to rigorously discriminate in favor of NADP(H) but against NAD(H) (57)(58)(59)(60). In the crystal structure of the CPR-NADP ϩ complex (33), the 2Ј-phosphate is surrounded by the side chains of Ser 596 , Arg 597 and Lys 602 and is exposed on an edge of the negatively charged CPR surface. It is thought that this surface interacts with the positively charged HO-1 surface (16,49). Hence, it is conceivable that NADP ϩ plays a pivotal role in the association of the two proteins at the interface. In this binding study, we employed NADP ϩ rather than NADPH, because the addition of NADPH produced an unusual sensorgram that did not fit any sensible models (data not shown). This was probably due to heme degradation occurring during the association phase in the SPR experiment. Our primary purpose was to characterize the interaction between HO-1 and NADPH-bound CPR. NADP ϩ , however, is known to be a competitive inhibitor for the NADPH-dependent CPR reaction. The K i value for NADP ϩ is 18.8 M, which is comparable to the K m value for NADPH of 6.3 M (61, 62). Therefore, the ability of NADP ϩ to enhance the interaction between HO-1 and CPR should also occur with NADPH.
The enhanced association between HO-1 and CPR induced by NADP ϩ suggests that NADP(H) binding to CPR initiates a conformational change leading to favorable interactions with HO-1. Although no crystal structure has yet been determined for CPR in the NADP(H)-free state, multiple conformations of CPR in the absence of the coenzyme have been suggested by kinetic studies (59,60). The CPR molecule has a connecting region between the FMN and FAD domains. Structural rearrangement involving domain-domain movement during catalysis has been proposed from crystallographic studies of CPR mutants (63). Moreover, relaxation kinetic studies of internal electron transfer in CPR have indicated that NADPH binding induces not only a local conformational change in the vicinity of the cofactors but also a significant domain movement that optimizes electron transfer between the flavin cofactors (64). On the other hand, the folding of HO-1 also seems to be important for complex formation with CPR, because the heme-free HO-1, whose heme binding site is disordered (27), shows no binding affinity for CPR even in the presence of NADP ϩ . 2 It should also be noted that the enhancement of the interaction between HO-1 and CPR by NADP ϩ is observed even in the presence of BVR, although it indeed competes with CPR for binding to HO-1.
Recently, using fluorescence resonance energy transfer (FRET), Wang and Ortiz de Montellano reported that the heme 2 Y. Higashimoto, H. Sakamoto, and M. Noguchi, unpublished data.  complex of HO-1 binds to NADP(H)-free CPR with a K D value of 0.4 M (50). This value is six times smaller than the K D value (2.44 M) that we obtained here in the absence of NADP ϩ using SPR (Table I). The difference in the two values is possibly due to the different techniques employed in the two studies. Indeed, a binding study of CPR to cytochrome P450 2B4 by FRET (37) has yielded K D values ranging from 0.01 to 0.09 M, which are an order of magnitude smaller than the K D value of 0.45 M obtained by SPR (65). The reason for the discrepancy in K D values that depends on the techniques for analyzing proteinprotein interactions is unclear. However, it may be that in SPR experiments immobilization of CPR to the sensor chip brings about such a steric hindrance as to restrict the accessibility of HO-1 to CPR, whereas the two proteins interact freely with each other in FRET. The mutation of Arg 185 of HO-1 to Ala (R185A) caused a 15to 18-fold decrease in the binding affinity for CPR, irrespective of the presence of NADP ϩ . Moreover, the HO activity and the heme reduction rate of R185A were drastically reduced when NADPH-CPR was used as an electron donating system. However, when ascorbate was used instead of NADPH-CPR, R185A behaved like wild type. These results rule out the possibility of alteration of the active site structure of HO-1 by this mutation and rather suggest that Arg 185 contributes to the binding to CPR through electrostatic interactions with charged moieties on the CPR surface, including the 2Ј-phosphate of NADPH. The observation that addition of NADP ϩ had no effect on the binding of R185A to CPR implies the absence of this interaction. The side chain of Arg 185 should be specifically required for this interaction, because replacement of Arg 185 with either Asp or Lys caused a loss of the heme degradation activity (data not shown). This idea is supported by the docking model of HO-1 and CPR, showing that the guanidino group of Arg 185 in HO-1 is located 2.9 Å from the 2Ј-phosphate of NADPH, a distance sufficient for electrostatic interaction.
The mutations of Lys 149 and Lys 153 to Ala, K149A and K149A/K153A, caused 7-and 9-fold decreases, respectively, in the affinity of HO-1 for CPR. In contrast to R185A, addition of NADP ϩ increased the affinities of K149A and K149A/K153A for CPR 2-to 3-fold, as observed with the wild type HO-1. Moreover, the HO activity and the heme reduction rate with NADPH-CPR of these mutants are significantly reduced, whereas they behave like the wild type with ascorbate. The docking model of the HO-1 and CPR complex shows that Lys 149 is very close to residues, including an acidic cluster, Asp 207 -Asp 208 -Asp 209 , that forms the FMN binding site of CPR. Notably, site-directed mutagenesis in this cluster only influenced electron transfer from CPR to cytochrome P450s without affecting the binding reaction (41,42) and green, respectively. The side chains of residues are shown as sticks (white, carbon; red, oxygen; light blue, nitrogen). Heme is shown as ball-and-sticks (orange). FMN and NADP ϩ are shown as sticks (magenta). The phosphorus atom in the 2Ј-phosphate group of NADP ϩ is highlighted in yellow. The nicotinamide moiety of NADP ϩ is omitted due to disordering (33). appear to interact with CPR in such a way as to orient the redox partners for optimal electron transfer from FMN of CPR to heme of HO-1.
Wang and Ortiz de Montellano (50) also demonstrated by FRET the contribution of Arg 185 to the interaction with CPR and showed that the residue was close to the NADPH binding domain of CPR in their docking model of HO-1 and CPR. However, they reported that the mutation of Lys 149 had no significant effect on binding. On the other hand, in the present study, we could confirm the contribution of Lys 149 for the association with CPR. Our result is consistent with an early report by Yoshida and coworkers; replacement of Lys 149 and Lys 153 with Ala reduced the reduction rate of the ferric heme bound to HO-1 (66). The same authors also showed a decrease in the reduction rate of the heme bound to R185A. In the structure of HO-1, Lys 149 and Lys 153 in the F helix are located at one edge of the heme binding site, whereas Arg 185 in the G helix is positioned at the other edge (Fig. 6). Interactions at these two separate sites would be critical for CPR to cover the smaller HO-1 molecule with its "bowl-shaped surface" (33), leading to efficient electron transfer to the heme center of HO-1. With mutation at either site, accumulation of the ferrous oxy-form takes place and normal heme degradation to biliverdin is hampered.
It has been widely accepted that basic residues on the proximal surface of cytochrome P450s are involved in binding to the negatively charged region on the CPR surface (40). Several acidic residues surrounding the FMN of CPR have been identified as participating in interactions with cytochrome P450s and cytochrome c (41,42). The FMN domain of CPR is of particular importance for interaction with redox partners, because FMN acts as the exit point of electrons to heme. Interestingly, the FRET study showed that Lys 18 and Lys 22 in the proximal A helix of HO-1 are involved in the interaction with CPR (50). However, we failed to confirm this, and rather found that these two residues affected neither the binding to CPR nor the NADPH-CPR-supported HO activity. Yoshida's group also showed that replacement of Lys 18 and Lys 22 with Ala did not affect the reduction rate of the heme bound to HO-1 (66). As described above, our docking model shows a close contact between Lys 149 in the distal F helix of HO-1 and the acidic clusters near the FMN site of CPR (Fig. 7A), suggesting that HO-1 interacts with CPR in a fashion somewhat different from that of cytochrome P450s.
Aromatic residues Tyr 140 , Tyr 178 , and Phe 181 of CPR, highly conserved in FMN-binding proteins, are located in the proximity of the isoalloxazine ring (33) and involved in the binding and stability of FMN (67,68). A similar rearrangement is found in the FMN-domain of P450 BM3 (45), wherein Trp 574 (corresponding to Tyr 178 of CPR) is less than 4 Å away from Pro 382 and Ser 383 residues that precede the heme-binding loop containing the proximal ligand, Cys 400 . This part of the heme-binding peptide could provide a through-bond electron-transfer pathway from FMN to the heme. It should be noted that in the docking model (Fig. 7A) Tyr 178 of CPR faces Pro 170 of HO-1 within 6 Å and that Pro 170 is adjacent to a cluster of aromatic amino acids, Phe 166 , Phe 167 , and Phe 169 of HO-1. The phenyl ring of Phe 169 is located 3.8 Å from the side chain of Asp 140 in the distal F helix that directly interacts with the distal ligand of the heme iron. If, in HO-1, electrons flow from the flavin to the heme through the distal residues, this difference in the electron-transfer pathways between cytochrome P450 and HO-1 may reflect the difference in the oxygen activation mechanism.
In summary, we have demonstrated that the binding affinity of HO-1 for CPR is increased by addition of NADP ϩ . Mutagenesis has revealed essential roles of Lys 149 and Arg 185 of HO-1 in the complex formation with CPR and thus the CPR-dependent enzyme activity. The proposed docking model has clearly shown distinct binding sites on the CPR surface for these two residues.