The domain architecture of cytochrome P450BM-3.

Cytochromes P450 utilize redox partners to deliver electrons from NADPH/NADH to the P450 heme center. Microsomal P450s utilize an FAD/FMN reductase. The bacterial fatty acid hydroxylase, P450BM-3, is similar except the P450 heme and FAD/FMN proteins are linked together in a single polypeptide chain arranged as heme-FMN-FAD. Sequence comparisons indicate that the P450BM-3 FMN and FAD domains are similar to flavodoxin and ferredoxin reductase, respectively. Previous work has shown that the heme and FMN/FAD domains can be separately expressed and purified. In this study we have expressed, purified, and characterized the following additional domains: heme-FMN, FMN, and FAD. Each domain retains their prosthetic groups although the FMN domain is more labile. The FAD domain retains a high level of ferricyanide reductase activity but no cytochrome c reductase activity. In addition, we have deleted a 110-residue stretch in the FAD domain that is not present in ferredoxin reductase. This protein retains both FAD and heme but not FMN. We also have investigated the dimerization pattern of the individual domains that lead to the following conclusions. Holo-P450BM-3 appears to dimerize via interactions that do not involve disulfide bond formation, whereas the reductase and FAD domains form intermolecular disulfides. This indicates that the Cys residues not available for dimerization in holo-P450BM-3 are unmasked in the individual domains.

P450s can be classified according to the type of electron transfer system utilized to deliver electrons from NADPH or NADH to the P450 heme. Type I bacterial and mitochondrial P450s normally require two proteins to transport electrons from reduced pyridine nucleotide to the P450 heme as follows.
FAD protein 3 ͑Fe Ϫ S͒ 2 protein 3 P450 REACTION 2 In contrast, type II microsomal P450s utilize a single FAD/ FMN reductase to deliver electrons to the P450. The fatty acid monooxygenase, cytochrome P450BM-3 from Bacillus megaterium, is the first bacterial P450 found to belong to the microsomal class of P450s in that P450BM-3 uses a FAD/FMN reductase to deliver electrons to the P450. However, in P450BM-3 the P450 and FAD/FMN reductase are fused together in a single polypeptide chain of 119,000 daltons (1), whereas in microsomal P450s the reductase and P450 are separate polypeptides. Additionally, the sequence homology between the reductase domain of P450BM-3 and microsomal P450 reductase is high, whereas the heme domain more closely resembles in sequence certain classes of microsomal fatty acid P450 monooxygenases than other bacterial P450s (2,3). Owing to these similarities, P450BM-3 has become an excellent model system for microsomal P450s with the distinct advantage that large amounts of P450BM-3 can be prepared from Escherichia coli recombinant expression systems (4) thus providing sufficient material for detailed biophysical studies.
Porter and Kasper (5) first noted that the N-terminal part of the diflavin P450 reductase exhibits good homology with FMNcontaining bacterial flavodoxins, whereas the C-terminal half is homologous to the FAD-containing spinach ferredoxin reductase. This suggests that P450 reductase was constructed by the fusion of flavodoxin-like and ferredoxin reductase-like genes. This homology with flavodoxin and ferredoxin reductase extends to the P450BM-3 reductase domain. Such comparisons indicate that holo-P450BM-3 was pieced together by the fusion of at least three separate folding units: heme, FAD, and FMN. The architecture for nitric oxide synthase is remarkably similar (6). Apparently, once Nature discovers a useful functional unit, the various units are used in multiple ways by covalently tethering the domains via a gene fusion mechanism resulting in redox proteins with novel activities. If, as this view suggests, each of the P450BM-3 domains can fold as semi-autonomous units, it should be possible to recombinantly express the individual domains that retain both prosthetic groups and some functional activities. With P450BM-3 the heme and FAD/FMN reductase domains already have been separately expressed in recombinant systems, and with the heme domain, the crystal structures in the substrate-free (7) and -bound forms are known (8). The recombinantly expressed FMN and FAD domains of microsomal P450 reductase also have been characterized (9). In this report we show that the FMN and FAD reductase domains of P450BM-3 also behave as autonomous folding units. We have found that the FAD domain can be eliminated leaving behind the heme/FMN subdomain. In addition, we have probed the role of a large insertion in the reductase domain that is not present in either flavodoxin or ferredoxin reductase. Fig. 1 is a schematic diagram showing the regions that were cloned, expressed, and purified. The various domains were cloned by polymerase chain reaction using the pT7 BM3 as the template (10). The 5Ј-oligonucleotides were synthesized with an overhanging BamHI restriction site, whereas the 3Ј-oligonucleotides contained an EcoRI site. The polymerase chain reaction product was digested with EcoRI and BamHI and ligated to the pT7-7 vector digested with BamHI and EcoRI. The 110-amino acid deletion mutant (⌬110) was constructed by the method of Kunkel et al. (11) using the single-stranded DNA from the wild type P-450BM3 gene in the pT7-7 system as the template. All the mutants generated were confirmed by Promega fmol DNA sequencing system using polymerase chain reaction. The plasmids were used to transform E. coli BL21 (DE3) for the expression of the mutant enzymes and for making plasmids for sequencing.

Constructs-
Purification-In the following, each of the domains will be indicated by the residue number and the prosthetic group expected to be associated with the domain with the exception of the holo-P450BM-3 that is missing residues 711-821. This mutant will be referred to as ⌬110 (Fig.  1). The purification of the FAD-(654 -1048), FAD-(483-1048) reductase, and FAD/FMN-(471-1048) followed the following protocol. Transformant colonies picked from LB/agar plates with 100 g/ml ampicillin were grown at 37°C in 1.5 ϫ LB/ampicillin medium. Overnight cultures were used to inoculate 1.5 liters of LB/ampicillin in 3-liter flasks and were incubated at 37°C. When the absorbance at 600 nm reached 1.0, another 100 g/ml ampicillin was added, and T7-RNA polymerase expression was induced by adding isopropyl-␤-D-thiogalactopyranoside to 0.6 mM. Growth was allowed to continue for another 12-15 h at 30°C. Five hundred ml of the culture was used for plasmid preparation, and the rest of the cells were harvested and stored at Ϫ70°C as a cell pellet until further use. All further handling of cells or enzyme preparations was performed in the cold. Frozen cell paste was suspended in 3-4 volumes of Buffer A (30 mM phosphate, 1 mM DTT, 1 0.1 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride, pH 7.4). The cells were lysed using a French press and centrifuged at 30,000 rpm for 60 min at 4°C. The supernatant in Buffer A was loaded on to a 2Ј,5Ј-ADP-Sepharose affinity column (2.5 ϫ 4 cm). Elution of the protein from the affinity column was performed as described by Black et al. (12). The purified fractions from the affinity column was loaded onto Sephadex G-25 column (2.5 ϫ 22 cm) to remove the 2Ј-AMP.
The ⌬110 deletion mutant and the wild type P450BM-3 were purified using the DEAE-Sephacel column (2.6 ϫ 70 cm). The cell lysate in 0.1 M P i with the above protease inhibitors from a 6-liter culture was loaded onto a DEAE-Sephacel column, and after washing with 1 liter of 0.1 M phosphate buffer, pH 7.4, the proteins were eluted using a linear gradient from 0.1 to 0.5 M potassium phosphate buffer in the presence of protease inhibitors. The FMN-(471-664) domain, heme-(1-625), and heme/FMN-(1-664) were partially purified using DEAE-Sephacel, and the column was further purified using Sephacryl S-100 (1.8 ϫ 75 cm) gel filtration column. SDS-polyacrylamide gel electrophoresis of the various purified domains is shown in Fig. 2A.
Enzyme Assays-All protein concentrations were estimated by the Bio-Rad protein estimation method. The amount of heme in wild type P450BM-3 and the ⌬110 mutant were determined by reduced pyridine hemochromogen method (13). Purified FAD/FMN-(471-1048), FAD-(483-1048), FAD-(654 -1048), and⌬110 were checked for their ability to carry out NADPH-dependent electron transfer to the artificial electron acceptors, ferricyanide and cytochrome c. All spectrophotometric assays were carried out using a Cary 3 spectrophotometer. Cytochrome c reductase, ferricyanide reductase, and NADPH oxidation activities were determined as described (14). Substrate binding was estimated using spectrophotometric titration by following the characteristic low to high spin transition as indicated by the shift of the main Soret absorp- ͓Free myristate͔ ϭ ͓total myristate͔ Ϫ ͓P450͔⌬absorbance/⌬absorbance max (Eq. 1) Flavin Content-Flavin content was determined both by reverse phase high performance liquid chromatography and by measuring the absorbance of the released flavin at 450 nm using extinction coefficients of 12.2 mM Ϫ1 for FMN and 11.3 mM Ϫ1 cm Ϫ1 for FAD. Twenty M enzyme in 200 l of 100 mM Tris-HCl buffer, pH 8.0, was placed in a boiling water bath for 10 min, chilled on ice, and centrifuged at 14,000 rpm for 20 min, and the supernatant was injected onto the C-18 reverse phase high performance liquid chromatography column (0.46 ϫ 25 cm). The flavins were eluted from the column by washing for 5 min with 50 mM ammonium acetate, pH 4.2, followed by a gradient from 50 mM ammonium acetate to 25% acetonitrile in 50 mM ammonium acetate over a period of 45 min at a flow rate of 1.0 ml/min. The retention times were 27.3 min for FAD and 28.6 min for FMN. Using Sigma FAD and FMN as standards, peak areas were used to set up the standard curves. The flavin content for the different domains presented is an average of the values measured using both the techniques.

RESULTS
FMN-(471-664) Domain-Sequence alignments of the FAD/ FMN reductase domain of P450BM-3 with clostridial flavodoxin and spinach ferredoxin NADP ϩ reductase were carried out using the FASTA module in the Biosym INSIGHT II package (Fig. 3). The FMN and FAD binding regions identified from the known crystal structures (15,16) were manually aligned and held fixed prior to automated alignment for the rest of the sequence. These alignments then were used as guides for preparing the various constructs needed to express the domains. Initially we attempted to express the heme/FMN domain consisting of residues 1-625. A protein of the correct molecular weight was expressed but did not contain FMN. Expression of residues 1-664, however, did give a soluble protein that contained 0.5 Ϯ 0.05 mol of FMN per mol of protein (Table I) indicating that the additional 39 residues were essential for FMN incorporation. The decrease in absorbance near 500 nm in heme/FMN-(1-664) is indicative of flavin reduction (Fig. 4) (Fig. 5).
The sequence comparisons indicated that the FMN domain of holo-P450BM-3 contains an extra 12 residues at the N terminus not present in flavodoxin. To test the importance of these extra residues, a construct of the FAD/FMN domain was prepared consisting of residues 483-1048 rather than the usual 471-1048. During purification of FAD/FMN-(483-1048), most of the protein bound to the affinity column very tightly, and only a small amount of protein was obtained using 50 mM 2Ј-AMP in 0.5 M phosphate buffer. The purified protein also showed increased susceptibility to proteases when stored at 4°C. Analysis of flavins showed 0.77 eq of FAD but no FMN (Table I), and the FAD was reduced to semiquinone form by NADPH as shown in Fig. 5. This indicates that the additional 12 residues, 471-482, are important for FMN binding.
FAD Reductase Domain-Spinach ferredoxin-NADP ϩ reductase is a FAD containing protein that catalyzes electron transfer from reduced ferredoxin to NADP ϩ during photosynthesis (17). Comparison of the P450BM-3 reductase to ferredoxin-NADP ϩ reductase reveals striking similarity over stretches of amino acids in the C-terminal region of P450BM3 reductase especially the key residues involved in the binding of FAD and NADPH. Fig. 3 shows the alignment of the C-terminal segment of P450BM3 reductase residues 625-1048 with the complete sequence of ferredoxin-NADP ϩ reductase. Careful alignment of the two proteins indicates an overall shift of sequence toward the C terminus is due to an insertion of 110 residues in the P450BM-3 reductase sequence. This also is the case with microsomal cytochrome P450 reductase (18). The purified FAD-(654 -1048) gave absorption maxima at 454 and 381 nm (Fig. 5) and has 0.83 eq of FAD (Table I). NADPH was able to partially reduce the FAD, probably to the semiquinone, whereas dithionite completely reduced FAD to the hydroquinone (Fig. 5).
Purified FAD-(654 -1048) gave more than one band under nonreducing conditions on SDS-polyacrylamide gel electrophoresis (Fig. 2B). With reference to the Pharmacia molecular mass standards, these two protein bands corresponded to molecular mass of 110.5 kDa (Ϸ20% dimer) and 45.4 kDa (Ϸ80% monomer). Purified P450BM3 reductase run under identical conditions migrates as two bands under nonreducing conditions corresponding to molecular mass of 150 kDa (Ϸ60% dimer) and 63.9 kDa (Ϸ40% monomer).
The apparent dimerization of the FAD reductase domain was further analyzed by gel filtration chromatography using FPLC gel filtration chromatography (Superdex 200 HR 10/30). In 0.1 M phosphate buffer, pH 7.4, with 0.1 M NaCl, the protein was found to elute as two peaks (Fig. 6). In comparison to the calibration standards, these two peaks corresponded to molecular mass of 120 kDa (40%) and 42.9 kDa (60%). However, when the FAD reductase was incubated with 50 mM DTT for 5 min and run in the presence of 10 mM DTT in the same buffer, most of the protein eluted as a monomer. We also studied P450BM3 reductase, DTT at 4°C and run in the presence of 10 mM DTT chromatographed mostly as one peak at 90 kDa. This is higher than the expected size of 63 kDa possibly due to the shape of the molecule. These results indicate that both the FAD-(654 -1048) and FAD/FMN-(471-1048) dimerize via formation of intermolecular disulfide bonds. Holo-P450BM-3 dimerizes via some other mechanism. With or without the preincubation with 50 mM DTT, holo-P450BM-3-(1-1048) eluted as a dimer (230 kDa) even when 10 mM DTT was included in the elution buffer (data not shown). Moreover, in nonreducing gels, holo-P450BM-3(1-1048) behaves as a monomer further indicating that disulfide bridges are not responsible for the dimeric chormatographic behavior of holo-P450BM-3. Fig. 4, only the heme/FMN-(1-664) contained FMN indicating that the additional 39 residues are essential for FMN incorporation. The FMN in the heme/FMN-(1-664) could not be reduced with NADPH, but dithionite was able to reduce the FMN to the fully reduced hydroquinone form.

Heme-(1-625) and Heme/FMN-(1-664)-As evident from
⌬110 Deletion Mutant-As noted earlier, the FAD-(654 -1048) domain of P540BM-3 and microsomal P450 reductase have a Ϸ110-residue insertion relative to spinach ferredoxin reductase (Fig. 3). To investigate the significance of this insertion, the 110 residues were deleted from holo-P450BM3 (Fig.  1), and the deletion mutant was designated ⌬110. Despite such a large internal deletion, the protein was overexpressed and could be purified. However, the ⌬110 bound much more tightly to the 2Ј,5Ј-ADP-Sepharose affinity column than wild type P450BM-3 and could not be eluted using 50 mM AMP. Therefore, we switched to ion exchange chromatography for purification, and the resulting purified protein contained 0.5 eq of heme, 0.5 eq of FAD but no FMN (Table I). Reconstitution of the ⌬ 110 mutant by incubating the enzyme with 5-fold excess of free FMN on ice overnight did not improve the FMN content of the enzyme. The K D value estimated using the 14-carbon fatty acid, myristic acid, was 0.588 M which is very close to the K D for the wild type (0.48 M).
Functional Activities of the Different Domains-The purified domains were tested for their ability to carry out NADPH-dependent electron transfer to ferricyanide and cytochrome c. In comparing % activities, the FAD/FMN-(471-1048) reductase domain was taken as 100%. A comparison of such activities is shown in Table I. FAD-(654 -1048) retained 25% of the ferricyanide reductase activity but very low cytochrome c reductase activity. Similar studies carried out with FAD domain from microsomal P450 reductase showed 50% ferricyanide reductase activity and no cytochrome c reductase activity (9). Ferricyanide reduction by FAD-(654 -1048) was ionic strength-dependent as ferricyanide reductase activity by FAD-(654 -1048) decreases to 30 and 10% in 10 mM phosphate and 1 mM phosphate buffers, respectively, in comparison to 50 mM phosphate buffer (100%). Addition of FMN-(471-664) to FAD-(654 -1048) did not alter the ferricyanide reductase activity significantly, but the  cytochrome c reductase activity increased slightly in comparison to FAD-(654 -1048) alone.
Since the FMN-(471-664) lacks the NADPH domain, little NADPH-dependent reductase activity was expected. As shown in Table I, this domain does not have significant levels of either ferricyanide or cytochrome c reductase activities.
The FAD/FMN-(483-1048), which has FAD but no FMN, retained about 25-30% ferricyanide reductase activity but very little cytochrome c reductase activity. This is in keeping with the expectation that FMN is required for cytochrome c reduction but not ferricyanide reduction (19). Preincubation of this domain with excess FMN did not restore any cytochrome c reductase activity. This indicates that the presence of residues 471-483 at the N terminus of the FAD/FMN reductase domain is important for FMN binding. P450 reductase domain of P450BM-3 where the N-terminal 120 amino acids cleaved by trypsin showed similar kinetics (20).
⌬110 (contains FAD but not FMN) exhibited only very low levels of cytochrome c reductase activity, but ferricyanide reductase activities were slightly higher than that of FAD-(654 -1048) domain. Addition of excess FMN did not restore cytochrome c reductase activity. The heme domain is not significantly affected by the 110-residue deletion. Addition of substrate gave the same low to high spin spectral transition observed in holo-P450BM-3 and the recombinant heme domain. In addition, upon reduction in the presence of CO, the characteristic peak at 450 nm was observed.
Both the heme-(1-625) and heme/FMN-(1-664) domains did not show any cytochrome c or ferricyanide reductase activity using NADPH as the electron donor which is expected since the NADPH-binding site is close to the C terminus of the holo-P450BM-3, and only FAD can accept electron from NADPH. Reconstitution of heme/FMN-(1-664) with 5-fold excess of FAD reductase domain did not improve the cytochrome c or the ferricyanide reductase activity of the heme /FMN-(1-664). Myristate hydroxylase activity could not be reconstituted using the heme/FMN-(1-664) and FAD reductase domains. From the data in hand we cannot tell if the inability to bind FMN is due to the loss of some key residues required for binding FMN or a more drastic unfolding or loosening of the FMN domain. An examination of the Clostridium Mp. flavodoxin x-ray structure (3FXN, see Ref. 15) shows that the N-terminal region is involved in a ␤-sheet structure and that the C-terminal region is helical and close to the N terminus. Disruption of either end might well lead to folding problems or a loosening of the FMN binding. This could explain the sensitivity of FMN binding to manipulation of the FMN domain's N and C termini in P450BM-3. It also should be noted that it is relatively easy to remove FMN from P450 reductase (19), whereas the FAD is more rigidly held in place. This, too, is evident from the x-ray structures since a single Trp residue sits between the FMN and bulk solvent in flavodoxin. In contrast, the larger FAD with two rings, isoalloxazine and adenine, in spinach ferredoxin reductase is more deeply embedded in the protein and forms a more extensive array of interactions with the protein (16). Why FMN binding should be so sensitive to deletion of residues 711-821 is less clear. However, the answer will be forthcoming shortly since a microsomal P450 reductase has been crystallized in a form suitable for high resolution crystal structure determination (22).
We can be more definitive about the integrity of the FAD and heme domains. The simplest criterion demonstrating that the FAD domain remains intact is the ability to adhere to the affinity column used for purification. This property requires a correctly folded NADP ϩ site. Moreover, the ability to carry out NADPH-dependent ferricyanide reduction is a good indication that the FAD domain is correctly folded. The robustness of the heme domain is evidenced by determination of the heme domain crystal structure with (8) and without substrate (7) and by retention of the many spectral properties expected of a functional P450. Therefore, P450BM-3 consists of at least three domains (heme, FAD, and FMN) that can independently fold. This is consistent with a similar study on P450BM-3 recently published (21) and previous work on microsomal P450 reductase (9).
Enzymatic Activities-Each of the domains retains the expected catalytic activities. Nevertheless, intact holo-P450BM-3 is required for full fatty acid hydroxylation activity. Reconstitution of some activity has been achieved by mixing the heme and FAD/FMN domains (23), but no combination of domains gives the very high levels of fatty acid hydroxylation activity (k cat Ͼ 1,000 min Ϫ1 ) found with holo-P450BM-3. Some of our previous work indicates that this has less to do with the structure of the individual domains but rather the way in which the FMN and/or FAD domains interacts with the heme domain. The flow of electrons is NADPH-to-FAD-to-FMN-to-heme. The C terminus of the heme domain is attached to the N terminus of the FMN domain by a linker consisting of residues that may include residues 456 -470. We say "may include" because it is not certain where the FMN domain begins. It does appear that the important part of the heme domain may end at residue 455 where the last ␤-sheet structure in the heme domain terminates. Residues immediately following 455 may belong to the linker between the heme and reductase domains. This might be why these residues are disordered in the crystal structure of the 1-471 heme domain (7). We have analyzed the role of this linker by making both amino acid substitution, deletion, and insertion mutants. Enzyme activity is insensitive to replacement of residues in the linker including a 6 Pro or Gly substitution (24) but is very sensitive to shortening the linker (14). The main conclusion drawn from this previous work is that the linker is critical for correctly orienting the reductase domain relative to the heme domain for efficient electron transfer. Although somewhat speculative, it is interesting to consider that Nature spent most of her evolutionary energy on how to link the various domains together rather than modifying the domains themselves. It appears that the correct length of linker had to be discovered in order to allow the reductase to sample the correct orientation for electron transfer. The importance of the linker is further evidenced by the example of nitric oxide synthase which has the same heme-FMN-FAD architecture as P450BM-3. One major difference is that the nitric oxide synthase linker binds calmodulin which switches on the FMNto-heme electron transfer reaction. In this case, it is very likely that the linker must adopt a helical structure, the expected conformation for calmodulin binding (25). Nevertheless, the same theme seems to be repeating itself in both nitric oxide synthase and P450BM-3 in that the FAD/FMN reductase domains have been relatively unaffected throughout evolution but that linker has been tailored for controlling the FMN-to- heme electron transfer reaction.
Oligomeric Structure-The FPLC and electrophoresis data indicate that the FAD/FMN-(471-1048) and FAD-(654 -1048) dimerize through disulfide bonds. This suggests that removal of the heme domain makes available Cys residues that cannot readily form disulfides in holo-P450BM-3. This could be due to a physical unmasking of Cys residues when the heme domain is removed or be due to steric restrictions that are absent in the FAD/FMN and FAD domains. In contrast, holo-P450BM-3 appears not to dimerize via disulfides (27). Some hint as to what causes holo-P450BM-3 to dimerize stems from the crystal structure of the substrate-free (26) and -bound (8) heme domain. Each form of the heme domain, with and without substrate, crystallizes in different space groups with a different number of molecules in the asymmetric unit. Nevertheless, in each crystal form, the heme domain forms dimers at the same interface via an array of noncovalent contacts near Tyr-166. Dimerization of holo-P450BM-3 at the same interface could explain the differences in dimerization patterns observed in the various domains.
In addition to gaining some insights into the domain architecture of P450BM-3, an additional reason for undertaking this study was to develop simpler functional units for further detailed biophysical studies. One of the more interesting and important is intramolecular electron transfer between prosthetic groups which becomes a complex problem in holo-P450BM-3. The heme/FMN-(1-664) should be particularly useful for studying the FMN-to-heme electron transfer reaction which is the topic of the accompanying manuscript (28).