D. Leys, Faculty of Life Sciences, University of Manchester, 131 Princess Street, Manchester M1 7DN, UK Fax: +44 16 306 8917 Tel. +44 161 306 5150 E-mail: david.leys@ manchester.ac.uk A. Munro, Faculty of Life Sciences, University of Manchester, 131 Princess Street, Manchester M1 7DN, UK Fax: +44 16 306 8917 Tel: +44 161 306 5151 E-mail: email@example.com
We report the crystal structure of the FAD/NADPH-binding domain (FAD domain) of the biotechnologically important Bacillus megaterium flavocytochrome P450 BM3, the last domain of the enzyme to be structurally resolved. The structure was solved in both the absence and presence of the ligand NADP+, identifying important protein interactions with the NADPH 2′-phosphate that helps to dictate specificity for NADPH over NADH, and involving residues Tyr974, Arg966, Lys972 and Ser965. The Trp1046 side chain shields the FAD isoalloxazine ring from NADPH, and motion of this residue is required to enable NADPH-dependent FAD reduction. Multiple binding interactions stabilize the FAD cofactor, including aromatic stacking with the adenine group from the side chains of Tyr860 and Trp854, and several interactions with FAD pyrophosphate oxygens, including bonding to tyrosines 828, 829 and 860. Mutagenesis of C773 and C999 to alanine was required for successful crystallization, with C773A predicted to disfavour intramolecular and intermolecular disulfide bonding. Multiangle laser light scattering analysis showed wild-type FAD domain to be near-exclusively dimeric, with dimer disruption achieved on treatment with the reducing agent dithiothreitol. By contrast, light scattering showed that the C773A/C999A FAD domain was monomeric. The C773A/C999A FAD domain structure confirms that Ala773 is surface exposed and in close proximity to Cys810, with this region of the enzyme’s connecting domain (that links the FAD domain to the FMN-binding domain in P450 BM3) located at a crystal contact interface between FAD domains. The FAD domain crystal structure enables molecular modelling of its interactions with its cognate FMN (flavodoxin-like) domain within the BM3 reductase module.
Bacillus megaterium flavocytochrome P450 BM3 (BM3) is a pivotal member of the cytochrome P450 (P450) superfamily of monooxygenase enzymes [1,2]. The P450s catalyse regio- and stereo-selective oxygenation of a wide variety of organic molecules in organisms from all of the major domains of life. The P450s require two electrons for reductive scission of heme iron-bound dioxygen and for the oxygenation (frequently hydroxylation) of a substrate bound proximal to the heme. Typically, these electrons are donated from NAD(P)H via a membranous diflavin enzyme (cytochrome P450 reductase, or CPR) in eukaryotes (a class II redox system) ; or from a FAD-containing ferredoxin/flavodoxin reductase via a flavodoxin or a ferredoxin in prokaryotic systems (a class I redox system) . However, the P450 superfamily is diverse and a number of variations in enzyme electron-delivery systems (or even their complete absence) are also observed in nature [5–7]. For some time, it was assumed that the class II-type P450 redox system was restricted to eukaryotes. However, discovery of the B. megaterium flavocytochrome P450 BM3 system (CYP102A1) by Fulco and coworkers in the 1980s, and its subsequent biochemical characterization and resolution of its multidomain construction, demonstrated clearly that CPRs were natural redox partners for certain prokaryotic P450s [8–10]. In P450 BM3, the fatty acid hydroxylase P450 is fused to its CPR in a soluble 119 kDa polypeptide. The fusion arrangement enables rapid electron transfer between the partner enzymes and allows P450 BM3 to have the highest rate of monooxygenase activity yet reported for a P450 . Recent studies indicate that the enzyme is functional as a fatty acid hydroxylase in the dimeric form, with electron transfer between monomers enabling oxidation of fatty acids at ω-1, ω-2 and ω-3 positions for most saturated lipid substrates [12–16].
The structure of the full-length flavocytochrome P450 BM3 enzyme has yet to be determined, but genetic dissection of the enzyme has enabled the production of the component heme (P450) and diflavin reductase (CPR) domains, and subsequently the further dissection of the CPR to enable expression and purification of the FAD/NADPH-binding (FAD) and FMN-binding (FMN) domains of P450 BM3 [9,10,17]. These studies confirmed that the individual domains retained natural properties (i.e. substrate binding, redox and catalytic activity) and provided important evidence that: (a) P450 BM3 was a natural fusion of P450 and CPR enzymes, and (b) that CPR itself was the result of an ancestral fusion between ferredoxin reductase-like (FAD domain) and flavodoxin-like (FMN domain) modules . Individual P450 BM3 domain preparations enabled the crystallization and structural determination of the heme (P450) domain of the enzyme in both substrate-free and substrate (fatty acid)-bound forms, revealing a major conformational change on binding substrate [19–22]. The structure of the flavodoxin-like FMN domain has also been determined . The only domain of the BM3 enzyme for which a structure is not yet available is the ferredoxin reductase-like FAD/NADPH-binding module. The structure of the FAD domain of BM3 is important in terms of understanding its interactions with its cognate FMN domain, and the nature of its FAD- and NADP(H)-binding sites.
In this article, we present crystal structures of the FAD/NADPH-binding domain of BM3 in both its ligand-free and NADP+-bound forms, to resolutions of 2.4 and 2.15 Å, respectively. These structures reveal the nature of flavin cofactor and NADP(H)-binding sites, and structural determinants that facilitate the strong selectivity for NADPH over NADH in this enzyme. The structural data enable rationalization of results of preceding mutagenesis strategies aimed at perturbing cofactor and NADPH-binding properties [24–28]. They also enable the reconstruction of the FAD and FMN flavin interaction site in a BM3 CPR model, helping to explain the efficiency of the electron transfer process within the reductase that enables BM3 to have the highest reported rate of substrate oxidation within the P450 superfamily . Crystallization of the BM3 FAD domain and its subsequent structural elucidation was facilitated by the use of a mutant designed to remove superficial cysteine residues in order to prevent inter- and/or intramolecular disulfide bond formation.
Multiangle laser light scattering (MALLS) analysis of the BM3 FAD domain and preliminary protein crystallization studies
Black and Martin’s early HPLC-gel filtration studies on the BM3 FAD domain indicated the heterogeneity of the protein, with monomer and various aggregated states of the protein reported . Subsequent size-exclusion chromatography studies by Kitazume et al.  reported a FAD domain elution profile essentially identical to that of the larger CPR domain of BM3, suggesting that a trimeric form might predominate. Our studies of the wild-type FAD domain using MALLS were consistent with the findings of Black and Martin, indicating that both monomer, dimer and some higher molecular mass aggregates were present (with monomer/dimer in an ∼ 60%:40% ratio). When the peak fractions for the dimeric form were separated, pooled and reloaded on this system, a single dimeric species was obtained with apparent average molecular mass of 86.4 ± 3.8 kDa, close to that predicted for the FAD domain dimer (88.4 kDa in the absence of the initiator methionine). However, analysis of the same dimer peak fraction sample pretreated with the reductant dithiothreitol and then with iodoacetamide (in order to covalently modify any exposed cysteine residues) provided a different result, with both monomer (44.0 ± 2.1 kDa, ∼ 25%) and dimer (86.3 ± 5.6 kDa, ∼ 75%) present (Fig. 1). These data are consistent with the solution-state heterogeneity of the FAD domain, and with the presence of one or more intermolecular disulfide bonds that are at least partially accessible for reduction by dithiothreitol. Preliminary crystallographic studies of the wild-type FAD domain failed to produce crystals and, in light of our MALLS analysis, we considered routes to decreasing the heterogeneity and molecular aggregation in the FAD domain samples. Inspection of the molecular surface of the domain modelled with reference to the homologous domain of the rat CPR structure suggested that a pair of surface-exposed cysteine residues (C773 and C810, using numbering for the full-length BM3 enzyme) might be able to form inter- or intramolecular disulfide bonds. It was previously reported that both the wild-type BM3 FAD domain and the CPR domain (residues 471–1048) formed dimers under oxidizing conditions, and that treatment of intact P450 BM3 with dithiothreitol decreased the sample heterogeneity and led to lower proportions of higher molecular mass species . In view of potential intra- and/or intermolecular disulfide bridges in the BM3 FAD domain, and in attempts to improve the solution-state properties of the BM3 FAD domain to facilitate its crystallization, we generated a point mutant of one of the relevant cysteine residues, C773A, as described in the Experimental Procedures. In addition, we generated the double mutant C773A/C999A, placing the potential disulfide-bond-disrupting mutation in a background where a second surface-located cysteine residue is also changed to alanine. In a previous study, we reported the catalytic properties of the C999A mutants of BM3 FAD and CPR domains, and their effects on catalytic efficiency .
Production and properties of the P450 BM3 C773A and C773A/C999A FAD domain mutants
The FAD domain construct comprises amino acid residues 653–1048 of the 1048 amino acid flavocytochrome [29,30]. C773A and C773A/C999A mutant enzymes were generated, expressed and purified as described in Experimental Procedures, and in a similar yield to the previously reported wild-type domain . UV–visible absorption studies demonstrated that both the C773A and C773A/C999A mutant FAD domains bound flavin stoichiometrically, had oxidized spectral features similar to those reported previously for the wild-type FAD domain, and could be reduced to the blue semiquinone form with NADPH and to the hydroquinone form by sodium dithionite, as described previously for the wild-type and C999A FAD domain mutant proteins . Thus, FAD binding and cofactor reduction properties using physiological and nonphysiological reductants were as previously reported for the wild-type FAD domain, as might be expected for peripheral mutations that do not impact on the immediate environment of the FAD cofactor. The C773A/C999A FAD domain had oxidized flavin maxima at 384/457 nm and was found to crystallize readily. It was also shown to be catalytically active in potassium ferricyanide reduction, with a kcat value of 16.3 ± 1.3 s−1 and Km (NADPH) = 160.3 ± 31.0 μm. The affinity for NADPH and the kcat value (as also reported previously for the C999A FAD domain point mutant) are diminished compared with the wild-type FAD domain as a consequence of the replacement of a cysteine implicated in NADPH binding and the regulation of electron transfer . MALLS studies were repeated for the C773A/C999A FAD domain, and revealed that the protein was monomeric with an apparent molecular mass of 42.0 ± 2.5 kDa.
Crystallization and structural resolution of the C773A/C999A FAD domain of P450 BM3
The FAD domain C773A/C999A mutant was crystallized in the ligand-free form, as described in the Experimental Procedures. The structure of the ligand-free FAD domain was solved by molecular replacement using a modified model of the rat CPR enzyme (PDB code 1AMO) . The NADP+-bound complex structure was obtained by soaking crystals in mother liquor supplemented with 10 mm NADP+. Omit maps calculated with the ligand-free structure revealed clear electron density corresponding to the NADP+ ligand.
The final refined model of the ligand-free BM3 FAD domain C773A/C999A mutant contains two molecules per asymmetric unit, and the structure was resolved to 2.4 Å. There is no electron density corresponding to residues 655–657, 731–736 and 1048 for molecule A, while molecule B lacks residues 655–658, 730–741 and 1048. This model gives a final R-factor of 0.211 and a free R-factor of 0.256. The final refined model of the C773A/C999A BM3 FAD domain NADP+ complex is very similar, with a final R-factor of 0.191 and a free R-factor at 0.225 (data to 2.15 Å). Full crystallographic statistics for the ligand-free and NADP+-bound FAD domain structures are given in Table 1.
Table 1. Crystallographic statistics for BM3 FAD domain structures. Structures were determined for the BM3 FAD domain C773A/C999A mutant, in both ligand-free and NADP+-bound forms. Structures have been deposited in the PDB with codes 4DQK and 4DQL, respectively.
BM3 FAD domain
Native structure + NADP+
No. of observations
rmsd from ideal
Global structure of the P450 BM3 FAD-binding domain
Crystal structures have been determined for the wild-type BM3 heme domain (residues 1–472 of the flavocytochrome) in substrate-free and various substrate/ligand-bound forms, as well as for various mutant forms of the heme domain [19–21]. The structure of the flavodoxin-like FMN-binding domain of BM3 has also been determined – solved as part of a larger construct encompassing both the heme and FMN domains of the enzyme . The structure of the FAD domain of P450 BM3 thus provides the last piece of structural information for the flavocytochrome P450 BM3 enzyme. The BM3 FAD domain (residues 653–1048) is a member of the ferredoxin–NADP+ reductase (FNR) enzyme family  and is clearly comprised of three individual domains: (a) the NADP(H)-binding domain (residues 888–1048) consisting of five α helices and five β strands, (b) the FAD cofactor-binding domain (residues 660–705 and 826–887), and (c) the ‘connecting domain’ (residues 705–825) that links the FAD/NADPH-binding domain to the FMN domain of the enzyme to form the CPR domain of P450 BM3 (residues 471-1048 of the flavocytochrome). The CPR domain has also been expressed and characterized [9,30,32]. This CPR domain is structurally related to the eukaryotic CPR enzymes, with the structure of the rat CPR providing the structural prototype for this enzyme family .
The overall fold of the BM3 FAD domain is similar to that adopted by the FAD/NADPH domain of rat CPR and by other members of the FNR family, for example the Escherichia coli flavodoxin reductase and spinach ferredoxin reductase [33–35]. Figure 2 shows the structure of the BM3 FAD domain highlighting the positions of cysteine residues (and the C773A and C999A mutations) and the relative positioning of the two molecules in the asymmetric unit. There is little difference between the overall structures of molecules A and B in the ligand-free or NADP+-bound FAD domains, although in both structures there is a small displacement of the NADPH/FAD domain relative to the connecting domain between molecules A and B in the asymmetric unit. The C773A mutation is clearly defined at the protein surface of the connecting domain module of the FAD domain. It is placed in a loop region between helices C and D of this domain. Ala773 occurs at a crystal contact between two connecting FAD domains, with the Cys810 in close proximity to Ala773 (at 7.5 Å Cα–Cα distance) (Figure 2). Under oxidizing conditions, it is clearly possible that an intramolecular disulfide bridge could be formed in the wild-type FAD domain involving these cysteines. Because of the small size of this interface (716 Å2, representing only ∼ 4% of the total accessible surface), it remains unclear whether the observed lattice contact is in fact representative of the dimer interface in solution. However, if this is the case, the presence of C773/C810 pair in this interface suggests the possibility of intramolecular disulfide bonding disrupting the wild-type FAD dimer interface domain, which would lead to further heterogeneity that might compromise crystallogenesis.
Figure 3 shows an overlay of the NADP+-bound form of the BM3 FAD domain with the rat CPR structure. The structure of the BM3 FMN domain is also superimposed on the rat CPR structure . The extended nature of the FAD domain is clear, as is the similarity in structure and relative positions of the FAD- and NADP(H)-binding domains between both reductases. There is a minor difference in the relative positioning of the connecting domains. The position of the connecting domain with respect to the cofactor-binding domains is considered to be flexible (as exemplified by the difference in position between molecules A and B for the BM3 FAD domain structure), enabling separation of the FAD and FMN domains in CPR-like enzymes (following electron transfer from NADPH-reduced FAD onto the FMN) to allow communication of the reduced FMN cofactor with the P450 partner . This hypothesis is consistent with data on domain mobility from NMR and small angle X-ray scattering studies [31,37]. The proximity of the FAD and FMN cofactors in the rat CPR structure points to direct inter-cofactor electron transfer, and a similar situation is envisaged for the BM3 reductase domain (see below).
The BM3 FAD-binding site in detail
The FAD cofactor is bound in an extended conformation by the FAD-binding domain, with adenine dinucleotide and isoalloxazine ring components bound by distinct parts of the FAD domain structure. (Fig. 4). Key interactions include hydrogen bonding of the adenine base to the peptide oxygen of Glu852, and of an adenine imidazole ring nitrogen to the peptide NH of Trp854, as well as van der Waals’ interactions with Arg797, Val849 and Ala853. The aromatic groups of Tyr860 and Trp854 stack either side of the adenine group. Polar interactions with the pyrophosphate oxygens are established by Gln756, Arg827 and Tyr860, as well as by the backbone nitrogens of Tyr828, Tyr829, Ile863 and Ala864. The pyrophosphate-binding motif was recognized as the most strongly conserved structural element in FAD-binding proteins, pointing to an important role in cofactor recognition and in binding energy .
Hydrogen-bonding interactions with FAD ribityl hydroxyl groups are made by Tyr828 and Tyr829. Hydrogen bonding between the FAD isoalloxazine ring and the protein occurs through residues Ser830 (side chain hydroxyl to flavin N5 and peptide NH to flavin C4 carbonyl oxygen), Thr845 (peptide oxygen to flavin N4) and Ser847 (peptide NH to flavin C2 carbonyl). A network of water-mediated hydrogen-bonding interactions is also present. It is proposed that water molecules involved in this network play a role in the acid/base protonation/deprotonation of the N5 atom involved in the electron transfer mechanism. Many of these interactions are very similar to those seen in the structure of rat CPR .
A conserved ‘catalytic triad’ of residues (Ser830, Cys999 and Asp1044) is found across the CPRs and other members of the diflavin reductase family, and is recognized as important in the binding of NADP(H) and regulation of electron transfer from NADPH to the FAD cofactor . In a previous study, we demonstrated that C999A mutant reductase and FAD domains of BM3 had substantially diminished rates of hydride transfer from NADPH to FAD, and less efficient dissociation of NADP+ product compared with wild-type enzymes . As discussed above, Ser830 hydrogen bonds to the FAD isoalloxazine ring, and is also hydrogen bonded via its side chain hydroxyl to a side chain carboxylate oxygen of Asp1044, which in turn is in close proximity to the Ala999 of the C999A mutation. Finally, the aromatic side chain of Trp1046 stacks over the central portion of the FAD isoalloxazine ring on its re face, protecting it from solvent and preventing access of the NADPH-reduced nicotinamide ring for electron transfer. In the NADP+-bound BM3 FAD domain crystal structure, the Trp1046 side chain remains in the ‘closed’ position over the FAD.
The BM3 NADP(H)-binding site in detail
P450 BM3 displays a strong selectivity for NADPH over NADH as its reducing substrate (Km values are 6.5 and 3030 μm respectively in ferricyanide reduction assays) . The structure of the FAD domain NADP(H)-binding site provides insight into the basis for this selectivity, with polar interactions between the adenine ribosyl 2′-phosphate and Arg966, Lys 972 and Tyr974. The side chain hydroxyl of Ser965 also interacts with both the 2′-phosphate and the adenine ribosyl hydroxyl group. This set of interactions is highly conserved in the FNR and diflavin reductase family members [38,39] (Fig. 5).
In addition to the interactions conferring specificity for NADP(H), several other interactions are made with non-2′-phosphate moieties of the NADP+ molecule, many of these similar to those seen in other diflavin reductase family members. The nicotinamide mononucleotide moiety is disordered in the complex, similar to other NADP+ complex structures [30,35]. The lack of electron density around the nicotinamide mononucleotide group indicates that there is extensive mobility of the nicotinamide when Trp1046 is stacked with the FAD isoalloxazine ring. Comparison of the ligand-free and NADP+-bound structure reveals that there is no significant structural rearrangement upon NADP+ binding. The rmsd between the native and NADP+-complexed protein structure is 0.206 Å, which is of a similar order as the rmsd between molecules A and B in the asymmetric unit.
A model for the BM3 reductase based on the rat CPR structure
The structure of the FAD domain of P450 BM3 enables a reconstruction of the likely mode of molecular interaction between the FAD/NADPH- and FMN-binding domains of flavocytochrome P450 BM3. Using the crystal structure of the rat CPR, a model is readily constructed for the interaction of the BM3 FAD and FMN domains. Superimposition of the FAD and FMN domain structures using rat CPR as a template (PDB code 1AMO) reveals no significant clashes between both BM3 domains when taking the small difference in connecting domain position into account (Fig. 6). The FAD domain of BM3 aligns with a rmsd of 1.55 Å (361 Cα; 36% identity) and the FMN domain of BM3 aligns with a rmsd of 1.30 Å (147 Cα, 32% identify) with the rat CPR structure . The dimethyl groups of the isoalloxazine rings of the FAD and FMN flavins are closely juxtaposed in the resulting model with only ∼ 4 Å between the flavins, consistent with efficient and direct electron transfer between the flavin centres [40,41].
Surface-charge analysis for the individual BM3 reductase domains indicates that both are predominantly negatively charged, but in the case of the FAD domain a region of positive charge is present at both the NADP(H)-binding site and at the predicted FMN domain docking site, thus providing some electrostatic complementarity to enable interdomain docking (Fig. 6). The overall charge distributions are similar to those seen in rat CPR, suggestive of an evolutionary conservation of surface charges to enable productive molecular docking, orientation and inter-cofactor electron transfer.
In the absence of the BM3 FMN domain, a single loop within the connecting domain (residues 729–743) remains unstructured in both molecules of the FAD domain within the asymmetric unit. This stretch of residues is highly charged, and has potential to be involved in interdomain interactions. In the rat CPR structure, the corresponding loop region (residues 348–364) establishes additional contacts with the FMN domain .
The structure of the FAD/NADPH-binding domain of flavocytochrome P450 BM3 completes the collection of individual domain structures of this intensively studied flavocytochrome, and leads logically to further studies on determining the relative orientation of the domains within the full-length enzyme. The oxidase enzyme is active as a dimer, with electron transfer to P450 heme occurring from the reductase of the opposite monomer [12,15,16]. The NADP+-bound FAD domain structure provides important insights into BM3’s strict selectivity for NADPH over NADH. Important interactions with the adenine ribosyl 2′-phosphate group are made by Ser965, Arg966 and Lys972, and this 2′-phosphate binding ‘motif’ is highly conserved in various members of the FNR and diflavin reductase enzyme family [31,42]. In recent work, we demonstrated substantial decreases in NADPH affinity and catalytic efficiency of NADPH-dependent ferricyanide reduction for S965A, R966A and K972A mutants of the BM3 FAD domain . Another residue shown to be important in discriminating between NADPH/NADH is Trp1046, which has aromatic amino acid counterparts in numerous FNR and diflavin reductase enzymes. Mutation of the corresponding residue in human CPR (Trp676) caused substantial changes in NADPH/NADH selectivity in W676A/H mutants, with catalytic efficiency (kcat/Km ratio) switched ∼ 103-fold towards NADH in the W676A mutant . In the comparable P450 BM3 FAD domain W1046A mutant, catalytic efficiency of ferricyanide reduction was switched by ∼ 3.3 × 103-fold towards NADH . This conserved aromatic residue places a hydrophobic ‘lid’ over the FAD in the FNRs and diflavin reductases, and access of the NAD(P)H nicotinamide to the FAD for electron transfer is gated by the motion of the side chain, identified by stopped-flow tryptophan fluorescence studies of human CPR FAD reduction . Removal of the aromatic barrier to the FAD explains the diminished discrimination between NADPH/NADH in mutants such as the BM3 W1046A variant. However, the NADP+-bound FAD domain structure shows that the Trp1046 ‘lid’ remains closed over the FAD and that the nicotinamide end of the NADP+ is disordered in the crystal, consistent with previous data for rat CPR . In pea FNR, a productive mode of NADP(H) binding was observed in crystallographic studies of the respective Y308S mutant, where the nicotinamide was able to approach the FAD isoalloxazine in a near-parallel geometry, with the flavin N5 atom close to the nicotinamide C4 in an appropriate configuration for an electron transfer event . Studies of rat CPR W677G and W677X mutants (the latter involving deletion of the final two residues of the protein) also revealed ordering of the nicotinamide end of the NADP+ and close proximity of the nicotinamide to the FAD isoalloxazine .
The global structure of the BM3 FAD domain is similar to that of various FNR and diflavin reductase FAD/NADPH domain modules. Superimposition of the FAD domain onto the rat CPR crystal structure (shown in Fig. 3) demonstrates this similarity in structure and conformation of the three major modules of the BM3 FAD domain [NADP(H)-binding, FAD-binding and connecting modules]. The BM3 FAD domain is also similar to the structure of the respective domain of rat neuronal nitric oxide synthase reductase , and strong similarities extend to FNR-like enzymes such as Acinetobacter benzoate 1,2-dioxygenase reductase and E. coli flavodoxin reductase, although these types of FNR lack the connecting domain essential for joining/orientating the FAD/FMN domains in the diflavin reductase family [33,48]. Molecular modelling of the interaction of the BM3 FMN domain with the FAD/NADPH domain indicates that these domains likely interact in similar fashion as seen in rat CPR, and that the FAD and FMN cofactors are spatially close, consistent with direct and efficient interflavin electron transfer, and with the rapid electron transfer reactions observed previously . Electrostatic guidance is likely involved in the correct orientation of the FAD/NADPH and FMN domains for efficient electron transfer, and may also be important for the relocation of the reduced (anionic semiquinone) form of the FMN domain to interact with the BM3 heme domain, where there is a basic region on the heme domain surface around the ‘pocket’ encompassing the cysteinate ligand to the heme iron . However, a simple model of the FMN ‘swinging’ from FAD/NADPH domain to heme domain on a hinge provided by the connecting domain is complicated by findings that the enzyme is functional in the dimeric state, and that the electron transfer likely occurs from the FMN domain of one monomer to the heme domain of the other, as also observed in nitric oxide synthase [12,16,49]. A major challenge in the field is the determination of the structure of the catalytically relevant dimeric form of a P450 BM3-like (CYP102A family) or nitric oxide synthase flavocytochrome enzyme.
The crystallization and structural elucidation of the P450 BM3 FAD/NADPH-binding domain was facilitated only following mutagenesis of both the catalytic triad cysteine C999 and a surface-exposed cysteine residue, C773A, following failures to obtain suitable crystals for the wild-type FAD domain. This mutant FAD domain crystal structure showed that Ala773 was surface exposed in a loop region between helices of the connecting domain, at a crystal contact region and in close proximity to another cysteine (Cys810). Thus, disulfide linkages both within (C773-to-C810) and between monomers (involving Cys773) and/or structural influences on crystal packing are likely reasons for the failure of the wild-type FAD domain to form useful crystals. This conclusion is borne out by MALLS data on wild-type and C773A/C999A FAD domains, and by the proximity of Cys(Ala)773 and Cys810 in the structure. Alignment of the BM3 FAD domain amino acid sequence with homologous domains from other diflavin reductases shows that a cysteine corresponding to Cys773 is absent in various members of the nitric oxide synthase, CPR and methionine synthase reductase families, and that only in selected methionine synthase reductase proteins might a cysteine be positioned similarly to Cys810 in the BM3 FAD domain. Thus, the C773A mutation that proved important in enabling FAD domain crystallization is specific to the BM3 system among the diflavin reductase family.
To date, the crystallization and structural determination of the intact flavocytochrome P450 BM3 has not proven possible. BM3 is a dimer in its functional form, an arrangement that likely enables efficient inter-monomer electron transfer for its high turnover rate. However, conformational flexibility and heterogeneity of the aggregation state may be factors underlying the difficulties encountered in BM3 crystallization. Regardless of these issues, there are important reasons to determine the structure of this 238 kDa dimeric enzyme. BM3 is an important model system for engineering of altered substrate selectivity in P450s, with notable successes achieved (by both rational mutagenesis and directed evolution approaches) in converting substrate selectivity from oxidation of long-chain fatty acids towards, for example, oxidation of short-chain alkanoates and alkanes, desaturation of alkylbenzenes and enhanced binding of dopamine for biosensor development [50–54]. Rationalizing its overall structural arrangement is thus important to understand further the nature of the interface between the reductase and heme domains (there is, as yet, no crystal structure of a stoichiometric P450:redox partner complex) and to facilitate further engineering that might stabilize the enzyme and/or lead to further novel activities (particularly if the heme domain conformational equilibria are sensitive to interactions with the FMN domain in the dimeric flavocytochrome form). The FAD domain structure is an important step forward, because it provides the final structural piece in the BM3 jigsaw.
Generation of expression clones for the wild-type, C999A and C773A/C999A mutants of the P450 BM3 FAD/NADPH-binding and reductase domains
The reductase domain of P450 BM3 (residues 471–1048) was amplified by PCR from the construct encoding full-length wild-type (WT) P450 BM3 (pBM23) using primers pNcoI, 5′-GAACAGTCTGCTAAAACCATGGCAAAAAAGGCAGAAAACGCTC-3′ and pBamHI, 5′-ACTAAACTACTTTTATCGGATCCTCTTTTTAAT-3′, as previously described, and where the underlined nucleotides indicate sites for the relevant restriction enzymes [9,27,28]. The amplified DNA for the wild-type reductase was cloned into plasmid vector pET11d (Merck Chemicals Ltd, Nottingham, UK) pre-digested with the same enzymes, generating reductase domain expression construct pREDWT. The FAD-binding domain (residues 653–1048) from wild-type reductase was amplified from the pREDWT using the oligonucleotide primers pNdeI, 5′-GCGCCGCGCATATGCCGCTTGCG-3′ and pBamHI, using PCR conditions described previously [27,28]. The FAD domain gene was excised at the underlined sites using the indicated restriction enzymes and then cloned into vector pET11a (Novagen) pre-cut with the same enzymes to generate FAD domain expression construct pFADWT. The QuikChange site directed mutagenesis kit (Stratagene, La Jolla, CA, USA) was used to generate the C999A mutant in the BM3 FAD domain, using a primer pair described previously . The C773A mutation was then generated in both the wild-type and C999A BM3 FAD domains, using the method described previously and oligonucleotide primers C773AF, 5′-GCTAAAACGGTCGCCCCGCCGCATAAAGTAG-3′ and C773AR, 5′-CTACTTTATGCGGCGGGGCGACCGTTTTAGC-3′ . Gene sequences were verified by sequencing of the plasmids at the PNACL facility (University of Leicester, UK).
Purification of the wild-type, C773A and C773A/C999A P450 BM3 FAD domains, and crystallization of the C773A/C999A domain
The wild-type, C773A and C773A/C999A mutants of the P450 BM3 FAD domain were expressed in E. coli strain BL21 (DE3), using growth and induction conditions described previously, and purified to homogeneity as in previous studies [27–30,40]. The purified flavoproteins were quantified using an extinction coefficient of 11 300 m−1·cm−1, at the oxidized flavin absorbance maximum, as previously reported [27,28]. The BM3 C773A/C999A FAD domain was crystallized using the sitting drop method at room temperature. Sitting drops were prepared by adding 2 μL of mother liquor to 2 μL of 12 mg·mL−1 FAD domain. Crystals were obtained using a well solution of 28% poly(ethylene glycol) 8000, 0.3 m ammonium sulfate, 100 mm cacodylate buffer pH 6.5. Crystals of dimensions 70 × 70 × 900 μm formed after 4–7 days. In order to form a complex with NADP+, C773A/C999A FAD domain crystals were soaked in a 10 mm NADP+ solution for 10 min. Native crystals and crystals soaked with NADP+ were immersed in 10% poly(ethylene glycol) in mother liquor to act as cryoprotectant, prior to mounting on a nylon loop and flash cooling in liquid nitrogen.
Data collection, structure elucidation and refinement
The data used for refinement were collected on a Mar CCD detector on ID14-EH3 at the European Synchrotron Radiation Facility (Grenoble, France). Crystals were cooled at 100 K and diffraction data were collected in 0.5° oscillations on a Mar 165 mm CCD detector. Data were processed and scaled using the HKL package programs denzo and scalepack . The crystal structure of the C773A/C999A BM3 FAD domain was solved using the molecular replacement program amore and the FAD domain of rat CPR (Protein Data Bank code 1AMO) as the search model [31,56]. Positional and B-factor refinement were carried out using refmac5, with manual rebuilding of the model in turbo-frodo [57,58]. In the final stages of refinement, addition of solvent molecules was carried out using ARP/wARP . Analysis of the stereochemical quality of the protein model was accomplished using the procheck  and whatif  structure validation programs. The NADP+-bound BM3 C773A/C999A FAD structure was then solved using difference Fourier methods. Data collection and final refinement statistics are given in Table 1. The atomic coordinates and structure factors for both the native and NADP+-bound FAD domain structures are deposited in the Protein Data Bank with accession codes 4DQK and 4DQL.
Spectroscopic and kinetic analysis of wild-type and C773A/C999A FAD domains of P450 BM3
UV–visible absorbance spectra for wild-type and C773A/C999A FAD domains were collected using a Cary 50 UV-visible scanning spectrophotometer (Agilent Life Sciences, Wokingham, UK). Steady-state kinetic analysis of potassium ferricyanide reduction was done as described previously .
Analysis of the oligomeric state of the wild-type BM3 FAD domain was done using a size-exclusion column (Superdex 200 10/300 GL) in series with an on-line MALLS detector, a quasi-elastic light scattering detector, refractive index and UV detectors. The FAD domain (2.7 mg·mL−1 in 50 mm potassium phosphate, 100 mm KCl, pH 7.0) was injected onto the column. A Dionex GS50 gradient pump (Dionex, Sunnyvale, CA, USA) was used with a mobile phase flow rate of 0.71 mL·min−1, and eluent was channelled to a Jasco UV 2077 plus visible spectrophotometer (Jasco Inc, Easton, MD, USA), a Dawn Heleos-II 18-angle light scattering detector for measurements of the intensity of scattered light for absolute molecular weight characterization, an Optilab rEX refractive index detector, and a quasi-elastic light scattering detector (Wyatt Technology, Santa Barbara, CA, USA) for size determination from measurement of the scattered light intensity autocorrelation function of the intensity of scattered light. Sample concentration was measured using the Optilab rEX and a refractive index increment (dn/dc) of 0.186. Wyatt astra 126.96.36.199 software was used for data collection and analysis. Prior to measurement, the FAD domain was applied to a Superdex 200 10/300 GL column to remove any aggregates. The purified dimeric fraction was split into two. The first fraction was subjected to MALLS analysis immediately without any further modifications. Dithiothreitol (10 mm) was added to the second fraction, the sample mixed and incubated at 37 °C for 1 h. The sample was then cooled to room temperature, mixed with 50 mm iodoacetamide and incubated in the dark at room temperature for 30 min. The modified sample was then centrifuged and subjected to MALLS analysis as above. MALLS analysis of the C773A/C999A BM3 FAD domain (at concentrations up to 10 mg·mL−1) was done using the same method and shown to be completely monomeric, obviating the necessity for further analysis of dithiothreitol-treated FAD domain.
NADPH and NADP+, isopropyl thio-β-d-galactoside and bacterial growth media (yeast extract and tryptone) were purchased from Melford Laboratories (Ipswich, UK). All other reagents were purchased from Sigma (Poole, UK) and were of the highest grade available.
The work was funded by the UK Biotechnology and Biological Sciences Research Council (grant numbers BB/F00252/1 and BB/F00883X/1).