The FMN to Heme Electron Transfer in Cytochrome P450BM-3

The crystal structure of the complex between the heme and FMN-containing domains of Bacillus megaterium cytochrome P450BM-3 (Sevrioukova, I. F., Li, H., Zhang, H., Peterson, J. A., and Poulos, T. L. (1999)Proc. Natl. Acad. Sci. U. S. A. 96, 1863–1868) indicates that the proximal side of the heme domain molecule is the docking site for the FMN domain and that the Pro382—Gln387peptide may provide an electron transfer (ET) path from the FMN to the heme iron. In order to evaluate whether ET complexes formed in solution by the heme and FMN domains are structurally relevant to that seen in the crystal structure, we utilized site-directed mutagenesis to introduce Cys residues at positions 104 and 387, which are sites of close contact between the domains in the crystal structure and at position 372 as a control. Cys residues were modified with a bulky sulfhydryl reagent, 1-dimethylaminonaphthalene-5-sulfonate-l-cystine (dansylcystine (DC)), to prevent the FMN domain from binding at the site seen in the crystal structure. The DC modification of Cys372 and Cys387 resulted in a 2-fold decrease in the rates of interdomain ET in the reconstituted system consisting of the separate heme and FMN domains and had no effect on heme reduction in the intact heme/FMN-binding fragment of P450BM-3. DC modification of Cys104 caused a 10–20-fold decrease in the interdomain ET reaction rate in both the reconstituted system and the intact heme/FMN domain. This indicates that the proximal side of the heme domain molecule represents the FMN domain binding site in both the crystallized and solution complexes, with the area around residue 104 being the most critical for the redox partner docking.

alignment of donor and acceptor molecules to achieve rapid electron flow. Despite the wealth of kinetic, thermodynamic, site-directed mutagenesis, and structural data on electron transferring enzymes, many questions in the area of biological ET remain unclear. For example, which chemical forces are most important for maintaining a functional protein-protein complex, how are electrons transferred from one redox center to another, and are the ET complexes observed in crystal structures functionally relevant?
Crystallization of redox partner complexes, such as cytochrome c peroxidase (CcP) and cytochrome c (cyt c) (1), methylamine dehydrogenase and amicyanin (2), and the latter complex and cytochrome c 551i (3), has provided valuable structural information about the interacting surfaces, orientation of cofactors, and possible pathways for the electron flow. The x-ray crystal structure of 1:1 complexes between yeast CcP and yeast and horse cyt c has revealed that the crystallographic CcP⅐cyt c complex (1) differs significantly from the proposed model based on the complementarity of surface charges (4). With only a few hydrogen bonds at the interface, hydrophobic and van der Waals interactions are the predominant forces holding the two proteins together. The functional relevance of the binding orientation observed in the crystal structure is supported by mutagenesis and kinetic and chemical modification studies (5)(6)(7).
A single crystal polarized absorption study on binary and ternary complexes of methylamine dehydrogenase with its redox partners demonstrated that both complexes are competent for substrate oxidation and ET (8). Also it was shown that the orientation of methylamine dehydrogenase and amicyanin in the solution complex is the same as the one seen in the crystal structure (9). However, amicyanin binding to cytochrome c 551i in solution appeared to occur at different sites when amicyanin is free or when it is in complex with methylamine dehydrogenase (10).
Multidomain or multisubunit electron transferring proteins serve as examples of protein complexes where the interaction between partners is optimized by the covalent linkage of the domains. Several structures of multidomain electron transfer proteins are available (11)(12)(13)(14)(15). Among them, flavocytochrome b 2 is the most intensively studied. Analysis of structural information on this enzyme has allowed the identification of structural elements and amino acid residues that are important in controlling ET from FMN to heme. The hinge region and one particular interface residue, Tyr 143 , were found to play an important role in modulation of interdomain ET (16 -19).
Flavocytochrome P450BM-3, a catalytically self-sufficient fatty acid monooxygenase from Bacillus megaterium (20,21), is an example of a multidomain protein in the superfamily of cytochrome P450 enzymes. Unlike other P450s, the heme domain of P450BM-3 is naturally fused with an FMN/FAD-containing reductase domain, which provides electrons required for P450 catalysis. Being a structural and functional analog to the microsomal P450s and P450 reductase, P450BM-3 is a very attractive model system for studying structure/function relationships and mechanism of ET in the P450-dependent monooxygenase system. Recombinant expression of functional domains of P450BM-3, including heme-(BMP), FMN/FAD-, FMN-, FAD-, and heme/FMN-binding fragments (22)(23)(24)(25)(26)(27), has provided simplified systems and facilitated studies on domaindomain interaction in this complex enzyme. Since the FMN is the site donating electrons to the heme of P450, the heme/ FMN-binding domain (BMP/FMN) of P450BM-3 represents the simplest model for the electron transfer complex (28). The heme domain of P450BM-3 has been structurally characterized. Both substrate-free and substrate-bound x-ray structures are available (29,30). Recent crystallographic studies on BMP/FMN have provided information about the structure of the FMN domain and defined the interaction site between the flavin and heme domains (31). Although the linkage between the two domains was proteolyzed during crystallization, the manner of domain-domain interaction was consistent with previous studies on P450s. Moreover, the precise positioning of the FMN toward the heme-binding loop led to the conclusion that the complex between the FMN and heme domains is specific and that the Pro 382 -Cys 400 peptide of BMP can provide an electron pathway from the flavin to the heme iron.
The present study is the first attempt to test the physiological relevance of the crystallographic complex. We have utilized laser flash photolysis techniques to examine the effect of chemical modification of specifically engineered cysteine residues of the heme domain of P450BM-3 on interdomain ET in both the reconstituted system consisting of the separate heme and FMN domains and intact BMP/FMN. The sterically bulky group of dansylcystine (DC) was covalently attached to the cysteines, which were introduced into the proximal side of the heme domain molecule to cause steric hindrance in the domain-domain interaction. The results indicate that the proximal side of the heme domain molecule in both separate BMP and intact BMP/FMN is the interaction site for the FMN domain and that the structure of the solution complex is consistent with that seen in the BMP/FMN crystal structure (31).

EXPERIMENTAL PROCEDURES
Materials-Reagents for bacterial growth were purchased from Difco. pProEX TM HTb vector, rTEV protease, and Pwo DNA polymerase were from Life Technologies, Inc. Restriction endonucleases and other modifying enzymes were purchased from Roche Molecular Biochemicals. Didansyl-L-cystine was from Molecular Probes. The Ni 2ϩ -nitrilotriacetic acid resin was obtained from Qiagen. Other chemicals were purchased from Sigma.
Cloning and Site-directed Mutagenesis-The 6-histidine tag fused cytochrome P450BM-3 holoenzyme and its heme-, FMN-, and heme/ FMN-binding domains were cloned by PCR using the pT7 BM3 as the template (32). The 5Ј-oligonucleotide primers for the holoenzyme, the heme/FMN-and heme-binding domains were synthesized with an overhanging EcoRI restriction site, whereas the 5Ј-oligonucleotide for the FMN domain contained an overhanging EheI restriction site. The 3Јoligonucleotides for all fragments of P450BM-3 contained a stop codon followed by a KpnI restriction site. The PCR products were ligated to a pNEB vector (New England Biolabs), and the sequences were confirmed. After digestion of pNEB-BM3, pNEB-BMP, and pNEB-BMP/ FMN with EcoRI and KpnI, and pNEB-FMN with EheI and KpnI, the DNA fragments containing coding regions of the desirable domains were ligated into unique EcoRI/KpnI and EheI/KpnI sites of the pPro-EX TM HTb vector, resulting in pProEX-BM3, pProEX-BMP, pProEx-BMP/FMN, and pProEX-FMN expression plasmids. Mutations were introduced into the heme domain coding region using the pProEX-BMP plasmid as a template for PCR. The 5Ј-oligonucleotide primers GGGA-GACGATGTGTGCGAGTCCGTCCAG (E372C) and CCAAGTGCGAT-TCCGTGCCATGCGTTTAAACC (Q387C) together with 3Ј-nucleotide primer corresponding to the carboxyl terminus of the heme domain, and 3Ј-oligonucleotide primers CTGAAGCTTGGACATAAGATATTATG (L104C) and GTTAAAGCCGCTAAGACCAATTGT (C156S) with the 5Ј-nucleotide primer corresponding to the amino terminus of the heme domain were used in the first PCR amplification (underlines indicate mutated codons). Synthesized 0.2-0.5-kilobase pair fragments served as 5Ј-or 3Ј-primers for the second PCR. The 1.4-kilobase pair fragments of DNA encoding the heme domain mutants were cloned into pNEB vector. Correct generation of desired mutations was confirmed by DNA sequencing. The PinAI/PmeI fragments of pNEB-BMP plasmids containing the mutations were ligated into PinAI/PmeI-digested expression plasmid pProEx-BMP. To build double mutants C156S/L104C, C156S/E372C, and C156S/Q387C of the heme domain of P450BM-3, the MunI/PstI fragment of pProEx-BMP(C156S) was exchanged with corresponding fragments of pProEx-BMP plasmids carrying single mutations. In order to introduce mutations into the holoenzyme of P450BM-3 and the heme/FMN-binding domain, the PinAI/PmeI fragments of pProEx-BM3 and pProEx-BMP/FMN were exchanged with corresponding fragments of pProEx-BMP carrying single and double mutants.
Protein Purification-Transformation, cell growth, and purification of the 6-histidine tag fused wild type (WT) and mutants of BMP, BMP/FMN, and P450BM-3 was performed as described previously for the 1-649 heme/FMN-containing fragment of P450BM-3 (27). Sephacryl S-200HR was used for gel filtration of the holoenzyme of P450BM-3 as the last purification step. In final preparations of WT and mutants of BMP, BMP/FMN, and P450BM-3, A 418/280 was not less than 1.62, 1.03, and 0.68, respectively. The recombinant FMN domain of P450BM-3 was expressed and purified as described elsewhere (25).
Spectral Studies-All UV-visible spectroscopy was performed using a Cary 3 spectrophotometer. P450 content was measured by the reduced CO difference spectrum (33). The concentration of the FMN domain and the FMN content were determined as described previously (25). The concentration of didansyl-L-cystine (DDC) was measured spectrophotometrically using ⑀ 328 nm ϭ 8.4 mM Ϫ1 cm Ϫ1 . The rate of NADPH oxidation by P450BM-3 was measured at 25°C by monitoring the absorbance change at 340 nm using ⑀ ϭ 6.22 mM Ϫ1 cm Ϫ1 . Reactions were carried in 50 mM sodium phosphate buffer, pH 7.4, in the presence of 100 M palmitic acid as described elsewhere (27). Cyt c reductase activity was measured in 50 mM sodium phosphate buffer in the presence of 100 M cyt c employing ⑀ 550 nm ϭ 21 mM Ϫ1 cm Ϫ1 .
Sulfhydryl Group Modification-Sulfhydryl-specific fluorescent labeling with DDC was carried out as described previously (34,35). After reduction of 100 M solutions of the mutants of BMP or BMP/FMN in 20 mM sodium phosphate buffer, pH 7.4, 0.1 M NaCl, with 20 mM dithiothreitol (DTT) at 37°C for 1.5 h, the excess of DTT was separated from the proteins by chromatography on Sephadex G-25 in the same buffer. A 40-fold excess of DDC was added to the proteins and incubated overnight at room temperature in the dark. Solutions were concentrated by ultrafiltration in a Centricon 50 (Amicon) and passed through a Sephadex G-25 column to remove unbound DDC. The fluorescence intensity was measured at 495 and 535 nm (excitation at 330 nm) for protein-bound and protein-free label, respectively, using a Hitachi F-4500 fluorescence spectrophotometer. Fluorescence measurements were carried out in 20 mM sodium phosphate buffer, pH 7.4, 0.1 M NaCl, at 25°C.
Laser Flash Photolysis Experiments-Laser flash photolysis experiments were performed anaerobically at room temperature as described elsewhere (18,28). The technique involves the generation of a strong reductant, 5-deazariboflavin semiquinone (dRFH ⅐ ), in situ through the excitation of 5-deazariboflavin (dRF) by a laser pulse, followed by rapid reduction of the protein heme or FMN by dRFH ⅐ . If a complex of redox partners is present in the solution and one partner is reduced preferentially, the rate of intermolecular ET can subsequently be followed spectroscopically. A 10 or 100 mM phosphate buffer solution, pH 7.4, contained 100 M dRF and 2 mM semicarbazide as a sacrificial electron donor. The solutions (minus enzyme) were made anaerobic by bubbling for 1 h with either argon or CO/argon mixtures that had been passed through two 1.5 ϫ 100-cm columns containing an oxygen removing catalyst (R3-11, Chemical Dynamics). The absence of oxygen was monitored by the amplitude and decay of the dRFH ⅐ transient signal obtained at 500 nm upon laser excitation of the reaction prior to addition of the enzyme(s). In order to maintain anaerobiosis, all aliquots of added protein or substrate were subjected to a flow of the oxygen-free gas prior to mixing with the bulk reaction solution. The protein concen-tration was always larger than the concentration of dRFH ⅐ generated by the laser flash, so that pseudo-first order conditions applied and no more than a single electron could enter each protein molecule. Transient kinetic data were collected using a Tektronix TDS 410A digitizing oscilloscope and analyzed on a personal computer using KINFIT (OLIS Co, Jefferson, GA). Fig. 1 represents the x-ray structure of the complex between the heme and FMN domains of P450BM-3 (31). Residues 479 -630 of the flavin domain were found to be well defined in the structure and were thought to represent the core of the flavodoxin-like molecule. For this reason, in order to clone the FMN-containing fragment of the minimal length, Asn 479 and Leu 630 were chosen as the amino-and carboxyl-terminal residues to design primers for the PCR reaction. Surprisingly, the expression of the 479 -630 construct was very low. A five amino acid extension of the carboxyl terminus to Ser 635 , however, resulted in a high expression of the properly folded 479 -635 FMN domain, whose spectral and redox properties were similar to those of previously expressed proteins (25,26). Ser 635 also was chosen as the carboxyl-terminal residue for cloning and expression of the shorter 1-635 BMP/FMN domain of P450BM-3. Fourteen amino acid carboxyl-terminal truncation did not affect either spectral or substrate binding properties of BMP/FMN. Similar to the construct of Li and Poulos (36), we chose Leu 455 as the carboxyl-terminal residue to produce the 6-histidine tag fused 1-455 heme domain of P450BM-3.

Design of Recombinant Domains of P450BM-3-
Characterization of the Cysteine Mutants of P450BM-3-The heme domain of P450BM-3 contains three natural cysteines at positions 62, 156, and 400. Cys 400 is the proximal ligand for the heme iron and is buried inside the molecule in the solventinaccessible hydrophobic heme pocket (29). Cys 62 is located close to the surface of the heme domain and has 8.62 Å 2 and 0.96 Å 2 solvent-accessible surfaces in BMP and BMP/FMN, respectively. However, Cys 62 has never been found to interact with the heavy atoms during preparation of the heavy atom derivatives of the crystals of either the heme domain or BMP/ FMN. 2 In contrast, Cys 156 , whose solvent-accessible surfaces are equal to 2.3 Å 2 and 1.7 Å 2 in BMP and BMP/FMN, respectively, was found to be easily derivatized with mercury compounds during short time soaks of both BMP and BMP/FMN crystals. For this reason, the highly reactive cysteine 156 was replaced by serine in BMP, BMP/FMN, and the holoenzyme of P450BM-3. The C156S mutation did not affect the spectral and substrate binding properties and rates of NADPH oxidation and cyt c reduction by the holoenzyme (data are not shown). Therefore, C156S mutants can be considered as control proteins in the studies described below.
Three single cysteines were introduced in the heme-binding domains of different constructs at positions 104, 372, and 387 to produce the double mutants C156S/L104C, C156S/E372C, and C156S/Q387C. In the x-ray structure ( Fig. 1), residues Leu 104 and Gln 387 were found to be at the interface between the heme and flavin domains of P450BM-3. Glu 372 , located on the side of the heme domain that does not contact the FMN domain, was replaced by cysteine as a control. The double mutants of BMP, BMP/FMN, and the holoenzyme of P450BM-3 incorporated heme and had UV-visible absorption spectra similar to those of the WT proteins. The addition of an excess of arachidonic acid to the mutated hemoproteins resulted in a near complete conversion of the heme iron to the high spin form. The NADPH oxidation and cytochrome c reduction activities of the double mutants of P450BM-3 were similar to those of the WT holoenzyme (data are not shown).
Modification of C156S/L104C, C156S/E372C, and C156S/ Q387C Mutants with Didansylcystine-Optimal conditions for sulfhydryl group modification were found by varying the molar ratio of DDC to protein and incubation time. Fluorescence spectra of free DC and DC-modified double mutants of BMP and BMP/FMN are shown in Fig. 2. Binding to the proteins was accompanied by an enhancement of the dansyl group fluorescence and a characteristic blue shift in the excitation from 535 to 495 nm. The 2-fold higher fluorescence yield of DC bound to BMP/FMN compared with that bound to BMP (Fig. 2B) is indicative of a more hydrophobic environment of the fluorescent probe in the intact heme/FMN domain, most likely because of the presence of the covalently linked flavin domain.
The specificity and stoichiometry of SH group modification were determined by addition of DTT to the labeled proteins. Loss of more than 85% of fluorescence within a few minutes due to bound DC upon DTT treatment (Fig. 2, spectra 5-7) indicates that fluorescent label binds specifically to the sulfhydryl groups of the proteins. The estimation of the fluorescence intensity of free DC released from the proteins after DTT reduction enabled us to determine the molar ratio of the fluorescent probe to protein, which was near stoichiometric and varied from 0.8 to 1.2 for many experiments with various mutants. To check the possibility of interaction of DDC with cysteine 62, C156S mutants of BMP and BMP/FMN were incubated with the label under the same conditions as the double mutants. There was partial labeling of Cys 62 in both proteins (Fig. 2, spectra 4). Calculation of the amount of the released label after addition of DTT to C156S mutants of BMP and BMP/FMN showed that less than 15% of the cysteines reacted with DDC. DC-labeled C156S mutants were used as controls in all experiments. The sulfhydryl group modification did not affect the substrate bind-2 H. Li and I. Sevrioukova, personal observations. ing properties of the proteins. The spectra of DC-labeled BMP/ FMN mutants in the presence of arachidonic acid are shown in Fig. 3. The C156S/L104C BMP/FMN produced slightly smaller low to high spin conversion than other mutants. The dansylated preparations retained the dansyl group after a prolonged period.
Laser Flash-induced Reduction of the Individual Heme and FMN Domains-Illumination of a solution of dRF by a laser flash in the presence of semicarbazide as a sacrificial electron donor results in a rapid formation of a 1-electron reduced dRFH ⅐ species that absorbs in the 450 -550 nm region similar to other neutral flavin semiquinones (37). dRFH ⅐ is very unstable in solution and, in the absence of protein, is lost by both a self-recombination pathway producing the non-reactive dRF dimer and recombination with the oxidized radical of semicarbazide (38). In the presence of an electron-accepting protein that can react with dRFH ⅐ , the absorption measured at different wavelengths decays below or above the preflash base line depending on the spectral properties of the species generated in solution. In our initial experiments, we have studied the reduction of the FMN domain by dRFH ⅐ . The reduced minus oxidized difference spectrum of the FMN domain is shown in Fig. 4 Because of the long time scale (Ͼ2 s) and the weak signal at 380 nm, the disproportional reaction of FMN . to form the fully oxidized and fully reduced FMN was difficult to study. However, it should be pointed out that, under all studied conditions, there was no detectable protonation of the FMN . and consequent formation of the blue, neutral FMN semiquinone, FMNH ⅐ , in the reaction mixture. This is in agreement with the structural data (31) that indicate that the FMN-binding site of the FMN domain of P450BM-3 is different from that of flavodoxins and microsomal P450 reductase in that it does not have structural elements required for the formation and stabilization of FMNH ⅐ . Taken together, we conclude that FMN . , the only semiquinone form produced upon 1-electron reduction of the flavin in the FMN domain under physiological conditions, is the species that donates electrons to the heme iron Reaction of the heme domain of P450BM-3 with dRFH ⅐ was carried out in the presence of arachidonic acid and carbon monoxide. The CO complex formation was monitored at 460 nm. CO binding by reduced BMP was previously shown to be fast and proceed with a second order rate constant of 4 ϫ 10 6 M Ϫ1 s Ϫ1 (40). In our reaction system with saturating concentrations of carbon monoxide (approximately 1 mM at room temperature), the rate of CO complex formation was limited by the slower BMP reduction and, thus, reflected the rates of the inter-or intramolecular ET between FMN . and oxidized heme. dRFH ⅐ was capable of reducing the separate heme domain of P450BM-3 with k obs values ranging from 1,000 to 2,400 s Ϫ1 within the studied concentration range. The non-linear plot of k obs versus the BMP concentration was fit to a hyperbolic equation to obtain a limiting value of 2,660 s Ϫ1 for k obs and a K d value of 1.03 M for the interaction of the free dRFH ⅐ with the heme domain (Fig. 5B). It is important to point out that the K d and k obs limiting values for direct reduction of BMP by dRFH ⅐ are an order of magnitude lower than those for the FMN domain.
Laser Flash-induced Reduction of the Mixture of the WT Heme and FMN Domains-In the system containing separate heme and FMN domains of P450BM-3, the flavin domain was present in excess and was reduced by dRFH ⅐ first. The FMN reduction was monitored at 475 nm, an isosbestic point for the CO-bound ferrous and ferric forms of the heme. This was followed by an intermolecular ET from the FMN to the heme iron of BMP, whose reduction and CO complex formation was monitored by an absorbance increase at 460 nm. In order to find optimal conditions for domain-domain interaction, we studied the effect of ionic strength on the intermolecular ET. The presence of the heme domain did not affect the rate of the FMN reduction by dRFH ⅐ in either 10 or 100 mM phosphate buffer (Table I). Compared with the spectral changes observed during the reaction in 10 mM phosphate, the absorbance decrease at 475 nm due to flavin reduction was two times larger in 100 mM phosphate buffer, indicating that twice as much FMN semiquinone was produced in the reconstituted system at the higher ionic strength (data are not shown). Nevertheless, at high ionic strength, the BMP heme reduction in the presence of the FMN domain was slow, monophasic, and concentration-independent within the studied concentration range with the rate constant of the reaction equal to 1.3 Ϯ 0.2 s Ϫ1 . In 10 mM phosphate buffer, heme reduction appeared to be biphasic with the rate constant for the fast phase an order of magnitude larger than that measured in 100 mM phosphate buffer (Table I) (Fig. 6). It should be noted that the rate constants for BMP reduction were comparable with those for FMN reoxidation. In addition, the reduction of BMP in the reconstituted system was 2 orders of magnitude slower than its direct reduction by dRFH ⅐ . Taken together, these results demonstrate that in the reconstituted system BMP receives electrons from the reduced FMN domain and not directly from dRFH ⅐ . The higher rates of the intermolecular ET between the FMN and heme domains at low ionic strength suggest that electrostatic interactions are important for complex formation and ET between the two proteins. For this reason, the study of the effect of DC modification on the interdomain ET was carried out in 10 mM phosphate buffer.
Effect of DC Modification on the Heme Domain Reduction in the Reconstituted System-Kinetic parameters of the reactions of FMN reoxidation and heme reduction in the reconstituted system consisting of the separate FMN domain and the WT and DC-labeled cysteine mutants of BMP are compared in Table II. Similar to WT and C156S BMP, up to 90% of DC-modified C156S/E372C and C156S/Q387C mutants of BMP were reduced in the fast phase. However, the rate constants for reduction of these proteins were two times smaller than WT. The attachment of a dansyl group to cysteine 104 had the most dramatic effect on the reduction of the hemoprotein by the FMN domain. Less than 60% of the mutant was reduced in the fast phase of the reaction. Moreover, the rate constants of the reduction of C156S/L104C BMP in both phases were an order of magnitude smaller than those for WT and C156S BMP. It is important that the unmodified C156S/L104C mutant was reduced by the flavin domain in a manner similar to that of WT, indicating that the mutation itself does not affect interdomain ET. The plots of k obs for the fast phase of the heme domain reduction in the reconstituted system versus BMP concentration were linear for all heme domain mutants within the concentration range studied, indicating a bimolecular process (Fig. 7).
Effect of DC Modification on the Heme Reduction in the Intact Heme/FMN Domain-Similar to the nonlinear plot of k obs versus the concentration of the separate FMN domain, the analogous plots for the flavin reduction in BMP/FMN in the absence and presence of substrate appeared to be hyperbolic (Fig. 5A). In the absence of substrate, the accessibility of the FMN for interaction with dRFH ⅐ in BMP/FMN was decreased by the covalently tethered heme domain. At high protein concentrations, compared with the separate FMN domain, the FMN reduction rates in BMP/FMN were approximately 30% slower. A limiting rate constant of 1. dRFH ⅐ but also promoted ET. The limiting rate constant was increased almost 2-fold to 2.3 ϫ 10 4 s Ϫ1 , whereas K d , 5.1 M, was not significantly different from that in the absence of arachidonate. The data indicate that the conformational change caused by substrate binding to the heme domain alters the solvent exposure of the FMN in BMP/FMN and the mode of interaction between the domains. Thus, the solution complex between the flavin and heme domains is likely to be flexible and may undergo conformational changes during catalysis. Furthermore, the possibility should not be ruled out that binding of arachidonic acid to BMP/FMN also could promote dRFH ⅐ to FMN ET by affecting the FMN redox potential.
The effect of DC modification on the interdomain ET in BMP/FMN mutants was studied in the presence of arachidonic acid. It is important to note that the reduction of the FMN by dRFH ⅐ was not significantly affected by DC modification in any of the BMP/FMN mutants (data are not shown). The kinetics of the CO complex formation with the reduced heme iron in the intact heme/FMN domain were similar for WT, the DC-modified C156S, C156S/E372C, and C156S/Q387C mutants, and the unmodified C156S/L104C mutant of BMP/FMN (Fig. 8). Furthermore, the ET reactions from the FMN to the heme iron were found to be monophasic. Although an intramolecular ET reaction is expected to be independent of protein concentration, the plot of k obs versus [BMP/FMN] for the heme reduction in BMP/FMN is hyperbolic. Considering the hyperbolic character of the kinetics of the FMN reduction by dRFH ⅐ in either the separate FMN domain or in the intact BMP/FMN (Fig. 5A), we assume that the hyperbolic dependence of k obs of the intramolecular FMN to heme ET is a consequence of the preceding reaction, i.e. FMN reduction. It should be mentioned that there was a direct correlation between the rate constants of FMN . oxidation and the heme reduction/CO binding which demonstrates the occurrence of an intramolecular ET.
As seen from the Fig. 8, the rate of interdomain ET was dramatically diminished in the DC-modified C156S/L104C mutant of BMP/FMN. Both the reactions of FMN reoxidation and heme reduction of this protein appeared to be biphasic (Table  III) and were 15 to 20 times slower than those observed for the WT and the other modified cysteine mutants of BMP/FMN. These results are in accord with the kinetic data on the reduction of the separate DC-labeled C156S/L104C BMP by the FMN domain in the reconstituted system and demonstrate that residue 104 is a part of the interface in the ET complex between the heme and flavin domains of P450BM-3 formed in solution. DISCUSSION A recent determination of the structure of the complex between the heme and FMN domains of P450BM-3 (31) was  Second, the crystallographic complex between the heme and FMN domains of P450BM-3 ( Fig. 1) provided detailed information about the arrangement of cofactors and the domain-domain interface. It also provides the first structural insights into how P450s may interact with their redox partner, P450 reductase, and what could be the electron pathway from the FMN to the heme iron. In the complex, the flavin domain interacts with the proximal side of the BMP molecule and forms contacts with the residues from the C-and L-helices, His 100 and Asn 101 , and the Pro 382 -Gln 387 peptide that precedes the heme-binding loop of BMP. The manner of domain-domain interaction, consistent with previous studies on P450s, and the precise positioning of the methyl groups of the FMN toward the heme-binding loop, led to the conclusion that the crystallographic complex between the FMN and heme domains of P450BM-3 is specific (31).
Designing of Cysteine Mutants-To test this model, we attempted to modify the complex interface by introducing a sterically bulky group on the surface of the heme domain molecule to hinder sterically complex formation. For this purpose, we  Replacement of selected residues with cysteines followed by chemical modification had several advantages over a simple introduction of bulky amino acids like Phe or Trp. First, this allowed us to establish that substitution of either hydrophobic, negatively charged, or polar side chains of Leu 104 , Glu 372 , and Gln 387 , respectively, for a short nonpolar side chain of Cys did not affect spectral and substrate binding properties of the heme domain and BMP/FMN and the catalytic activity of the holoen-zyme of P450BM-3. Second, the cysteine residues are highly reactive and can be easily modified with a variety of chemical reagents to produce derivatives suitable for different studies. In the present study, DDC was chosen for sulfhydryl group modification. Compared with the side chains of Phe or Trp, the dansyl group ( Fig. 2A) is more bulky and is likely to cause larger steric hindrance. In addition, its fluorescent properties provide a simple method to follow and control the proteinlabeling process. BMP and BMP/FMN mutants with DC-modified sulfhydryl groups were capable of binding substrates as well as the WT proteins. In accord with modification studies on other proteins (34,35,(42)(43)(44), dansylation of the P450 mutants occurred specifically through introduced SH groups. The residual labeling of the natural Cys 62 in double mutants was accounted for by using C156S mutants of BMP and BMP/FMN as controls in all experiments. The specificity and simplicity of DDC modification and ease of monitoring and evaluating the extent of DC labeling make dansylation an attractive tool for introducing a sterically bulky fluorescent group at desired sites on the protein surface.
Redox Active Moiety of FMN-Although the kinetics and thermodynamics of the transient intermediates in the catalytic cycle of P450BM-3 have been reported, there is no general agreement on the precise catalytic mechanism (45)(46)(47)(48). On the basis of a multiwavelength analysis of the reduction of the reductase domain of P450BM3 by stopped-flow spectrophotometry, it was proposed that the anionic FMN semiquinone is the species that donates electrons to the heme iron (46). Later, the red, anionic flavin semiquinone formed during the catalytic cycle of the holoenzyme of P450BM-3 was concluded to be FAD . (48). However, studies of the redox properties of the individually expressed FMN-and FAD-binding domains of P450BM-3 (25,26) disproves the latter conclusion by showing that the FAD domain was producing exclusively the blue, neutral semiquinone upon reduction by either NADPH or sodium dithionite. Titration of the FMN domain with sodium dithionite resulted in the conversion of the protein to the fully reduced FMN without accumulation of the intermediate semiquinone forms (25,26). The reduction potentials of the FMN in the separate FMN domain and the holoenzyme of P450BM-3 were estimated, and the positive reduction potential difference in the FMN semiquinone/hydroquinone couple was detected (49). This confirmed that the FMN hydroquinone is more thermodynamically stable than the semiquinone form. The thermodynamic instability of the semiquinone form and the lack of 1-electron reductants for the FMN precluded studies on the characterization of the intermediate species produced during reduction of the FMN domain. Laser flash photolysis studies on the 1-664-truncated heme/FMN-containing domain (BM t ) of P450BM-3 (28) have demonstrated that dRFH ⅐ can efficiently reduce FMN to the semiquinone state. Spectral changes observed upon reduction were consistent with the formation of FMN . but not FMNH ⅐ . However, the strong absorbance of the heme in the 400 nm region did not allow the flavin-reduced minus oxidized difference spectrum for BM t in the 350 -450 nm range to be measured. In the present paper, for the first time, we have obtained the redox difference spectrum of the FMN domain (Fig. 4) which clearly shows that upon 1-electron reduction this protein forms a red, anionic FMN semiquinone. The formation of FMN . was found to be pH-independent within the pH range from 6 to 8. The inability of the FMN domain of P450BM-3 to produce and stabilize the FMNH ⅐ , a unique feature among the flavodoxin-like proteins, has been found to be defined by the structure of the FMN-binding site of the protein (31).
Effect of DC Modification on the Interdomain ET-The effect of DC modification of specific cysteine residues of the heme domain of P450BM-3 on the domain-domain interaction was evaluated by measuring the kinetics of ET from the FMN to the heme iron in the system consisting of either separate heme and FMN domains or intact BMP/FMN, utilizing the laser flash photolysis technique. This methodology has proven to be an excellent tool for studying both intermolecular ET between different redox partners and intramolecular ET in multicenter redox proteins (50 -57).
Studies on the separate FMN-and heme-binding domains of P450BM-3 have shown that both proteins can be easily reduced by dRFH ⅐ . Direct second order reduction of the FMN or heme by dRFH ⅐ was expected to yield linear plots of k obs versus [protein]. However, this was not the case for either the FMN domain or BMP (Fig. 5). The approach of k obs to limiting values suggests that, at higher protein concentrations, dRFH ⅐ was capable of forming complexes with both domains of P450BM-3. It has been shown previously that FMN could directly interact with the heme domain and, in the presence of NADPH, transferred electrons and supported fatty acid hydroxylation (58). It is likely that dRFH ⅐ interacts with the same site of BMP. In the flavin domain, the FMN-binding site could be a potential site for interaction with dRFH ⅐ whose isoalloxazine ring could form a complex with the coplanar and partly exposed aromatic rings of the FMN and Trp 574 (31). Indeed, in the absence of substrate in BMP/FMN, where the FMN-binding site is buried in the interface between the two domains, both the interaction with the protein and the ET from dRFH ⅐ to FMN were perturbed.
There was no apparent complex formation between dRFH ⅐ and the heme domain in the system reconstituted with the FMN domain. In these experiments, the flavin domain was present in excess and was reduced first. The biphasic kinetics of BMP reduction by the FMN domain in the reconstituted system at low ionic strength might be due to the existence of multiple FMN domain-binding sites on the surface of the heme domain. Appearance of the fast phase in the reaction of the heme reduction at lower ionic strength indicates that electrostatic forces are important for complex formation between the heme and FMN domains, and that the electrostatically stabilized complex is more effective in the FMN to heme ET. Although only one salt bridge was found in the 967-Å 2 area of interface between the two domains in the x-ray structure, the surfaces that interact were electrostatically complementary with the overall molecular dipoles oriented for maximum stability (31). Even at saturating concentrations of the FMN domain, reduction of the heme domain in the reconstituted system was an order of magnitude slower than in the intact heme/FMN domain. This indicates that the interaction between separate domains is not strong and is greatly facilitated by the covalent linkage in BMP/FMN. A similar conclusion was made during fluorometric studies on domain-domain interactions in P450BM-3 (59) and during analysis of the interaction site in the crystal structure of the complex (31).
The rate constants for the heme reduction in BMP/FMN measured in the present study were 2-fold higher than those determined previously for the 1-664-heme/FMN-containing fragment of P450BM-3 (28). The reasons for this could be the shorter (⌬29) length of the present heme/FMN-binding domain, which excludes the interference of the carboxyl-terminal peptide with the domain-domain interaction, and the lower ionic strength of the reaction mixture, which appeared to facilitate the interaction of the two domains. Also, using arachidonic, rather than myristic (28) acid as a substrate results in a larger conversion of the low to high spin state of the heme iron.
According to the analysis of intraprotein ET developed from ET measurements both in biological and in chemical systems, the interdomain ET rate constant of 100 -500 s Ϫ1 observed in BMP/FMN in the present study corresponds to an approximately 20-Å edge-to-edge distance between the FMN and heme (60). This is in a good agreement with the crystallographic complex, where the distance between the flavin and the heme iron was shown to be 18.4 Å. Therefore, the conformational changes in BMP/FMN that occur upon substrate binding probably are not large and do not result in an appreciable change in the separation of the cofactors.
The covalent attachment of the hydrophobic, sterically bulky dansyl group to cysteines of the BMP mutants altered to different extents the ability of the heme domain to interact with and accept electrons from the FMN domain (Fig. 7). Dansylation of Cys 387 on the proximal side of BMP had the same effect as modification of the control Cys 372 . The reduction of both these mutants by the FMN domain was 2-fold slower than the WT. In contrast, DC modification of analogous cysteines in BMP/FMN did not perturb the interdomain ET rate. Modification of cysteine 104 had the most dramatic effect on the rates of the FMN to heme ET in both the reconstituted system and BMP/FMN. Heme reduction of the dansylated C156S/L104C mutant was 10 -20 times slower compared with the WT. In accord with the structural data, these results demonstrate that the proximal side of the BMP molecule is the interaction site with the FMN domain. Although the backbone of Gln 387 is the closest approach for the flavin in the crystals, the side chain of this residue is solvent-accessible. It is, therefore, possible that the dansyl group on Cys 387 is not oriented toward the FMN domain surface and does not cause steric hindrance for the FMN domain binding. In contrast, residue 104 is buried at the interface, and its dansyl group cannot be accommodated at the interprotein interface without causing a significant conformational change.
In summary, a structural model of the ET complex between TABLE III Kinetic parameters of the reduction of the DC-modified C156S/L104C mutant of BMP/FMN by dRFH ⅐ Rate constants were determined by the laser flash photolysis technique at room temperature. The reaction mixture contained 100 M dRF, 2 mM semicarbazide, 100 M arachidonic acid, CO-saturated 10 mM phosphate buffer, pH 7.0, and different concentrations of DC-modified C156S/L104C mutant of BMP. The FMN reoxidation and the CO-reduced BMP complex formation were followed by an absorbance change at 475 and 460 nm, respectively. Data were analyzed using KINFIT (OLIS Co., Jefferson, GA). the heme and FMN domains of P450BM-3 has been tested. Three residues on the heme domain surface were mutated to cysteines and chemically modified. DC modification of Cys 104 was the most critical for the interdomain ET in both the reconstituted system and the intact heme/FMN domain. Dansylation of residues 372 and 387 had only moderate effects on the BMP reduction and did not affect ET in BMP/FMN. Although the results do not prove that the orientation of domains in the crystal structure is exactly the same as in ET complex in solution, the present study does demonstrate that the proximal side of the heme domain molecule in both separate BMP and intact BMP/FMN is the interaction site for the FMN domain and that the structure of the ET complex in solution is consistent with that seen in the crystalline complex.