The Crystal Structure of Cytochrome P450 4B1 (CYP4B1) Monooxygenase Complexed with Octane Discloses Several Structural Adaptations for ω-Hydroxylation*♦

P450 family 4 fatty acid ω-hydroxylases preferentially oxygenate primary C–H bonds over adjacent, energetically favored secondary C–H bonds, but the mechanism explaining this intriguing preference is unclear. To this end, the structure of rabbit P450 4B1 complexed with its substrate octane was determined by X-ray crystallography to define features of the active site that contribute to a preference for ω-hydroxylation. The structure indicated that octane is bound in a narrow active-site cavity that limits access of the secondary C–H bond to the reactive intermediate. A highly conserved sequence motif on helix I contributes to positioning the terminal carbon of octane for ω-hydroxylation. Glu-310 of this motif auto-catalytically forms an ester bond with the heme 5-methyl, and the immobilized Glu-310 contributes to substrate positioning. The preference for ω-hydroxylation was decreased in an E310A mutant having a shorter side chain, but the overall rates of metabolism were retained. E310D and E310Q substitutions having longer side chains exhibit lower overall rates, likely due to higher conformational entropy for these residues, but they retained high preferences for octane ω-hydroxylation. Sequence comparisons indicated that active-site residues constraining octane for ω-hydroxylation are conserved in family 4 P450s. Moreover, the heme 7-propionate is positioned in the active site and provides additional restraints on substrate binding. In conclusion, P450 4B1 exhibits structural adaptations for ω-hydroxylation that include changes in the conformation of the heme and changes in a highly conserved helix I motif that is associated with selective oxygenation of unactivated primary C–H bonds.

skin. Additionally, some cytochrome P450 -hydroxylases, such as human cytochrome P450s 4A11, 4F2, and 4F3b, catalyze the formation of 20-hydroxyeicosatetraenoic acid from arachidonic acid, which contributes to regulation of vascular tone and renal sodium absorption (10,11). Genetic association studies link variants of human CYP4A11 and CYP4F2 with increased risks for hypertension, which in the case of CYP4A11 is likely to reflect dysregulation of renal sodium uptake as the blood pressure increase is sensitive to levels of dietary salt (12,13).
Fatty acid -hydroxylation reflects the addition of oxygen to a primary C-H bond of the terminal carbon of the fatty acid. This reaction is generally considered to be the most difficult reaction catalyzed by cytochrome P450 monooxygenases because the enzyme needs to exclude access of neighboring and more reactive secondary C-H bonds to the iron-bound reactive oxygen intermediate. The reactive intermediate is formed by the reduction of oxygen bound to the heme iron followed by protonation and heterolytic scission of the dioxygen bond with elimination of the water molecule and formation of a highly electrophilic oxygen atom bound to the hypervalent heme iron.
Oxygenation of aliphatic C-H bonds is thought to proceed by abstraction of a hydrogen by the iron oxo intermediate followed by recombination of the resulting substrate radical with the transiently iron-bound hydroxyl radical (14). As the strength of primary C-H bonds is greater than that of secondary C-H bonds, oxygen addition at the -1 or -2 positions of fatty acids is typically seen for less specialized enzymes (15)(16)(17). These considerations suggest that the -hydroxylases have structural features that favor the approach to the reactive iron oxo species of the -carbon relative to the -1 carbon (18).
Within human family 4 cytochrome P450 enzymes, the capacity for -hydroxylation is associated with a substitution of glutamic acid for the first alanine or glycine in the canonical sequence motif, (A/G)GX(D/E)T (6). This sequence motif is located on helix I in the active site near the heme iron. The presence of the glutamic acid near the heme iron leads to autocatalytic formation of an ester bond between the glutamic acid and the 5-methyl group of the heme (19 -21). This covalent linkage is likely to contribute to an active site architecture that favors -hydroxylation over (-1)-hydroxylation by these enzymes because substitutions of alanine or glycine for the glutamic acid increase the propensity of the enzymes to catalyze -1 rather than -hydroxylation (21,22). Additionally, the related human enzymes 4F8 and 4F12 (which have glycine residues at this position) as well as human 4X1 and 4Z1 with the corresponding alanine have not been shown to exhibit a preference for -hydroxylation reactions. In contrast, human P450 4X1 catalyzes the 14,15-epoxidation of anandamide (23); P450 4Z1 catalyzes in-chain hydroxylation of lauric and myristic acids (24); and P450s 4F8 and 4F12 catalyze (-3)-hydroxylation of arachidonic acid (25).
To identify structural features that underlie the function of family 4 P450s, the structure of rabbit P450 4B1 was determined by X-ray crystallography. P450 4B1 is distinguished from -hydroxylases in the 4A, 4F, and 4V2 subfamilies by its increased capacity to catalyze -hydroxylation of smaller fatty acids and n-alkanes (C7-C12). P450 4B1 has also been recognized for its role in activation of 4-ipomeanol and perilla ketone in livestock, two pulmonary toxins, as well as activation of small aromatic amines to cytotoxins (26,27). Rabbit P450 4B1 has been successfully expressed at high levels in Escherichia coli and isolated as functionally active full-length and N-terminally truncated enzymes with covalently linked hemes (28). The expression vector used to express truncated CYP4B1 retains roughly half of the N-terminal transmembrane helix (TMH) of P450 4B1, which is predicted to encompass residues 7-29 using TMHMM (29). The truncated enzyme retained its characteristic preference for hydroxylation of the terminal primary C-H bond of lauric acid, albeit at lower rates than the full-length enzyme (28). The lower turnover number of the truncated enzyme could reflect altered interactions with P450 reductase when the P450 and P450 reductase are reconstituted. Additionally, the truncated enzyme exhibited 99% covalent linkage of the heme prosthetic group to the protein via an ester bond formed with the side chain of Glu-310 indicating that the enzyme was active in the expression host because the covalent linkage is formed auto-catalytically by the enzyme (28). Reduced flavodoxin is likely to drive reduction of the iron for this reaction as flavodoxin can support the monooxygenase activity of other mammalian microsomal P450s (30). In this study, we report successful crystallization of the N-terminally truncated enzyme in the presence of octane and determination of its structure to a 2.7-Å limiting resolution. In addition to the covalent linkage between the heme and the protein, the structure exhibits other adaptations that promote the preference of the enzyme for -hydroxylation. Comparisons of the amino acid sequence of P450 4B1 with other family 4 -hydroxylases indicate that amino acids that restrain octane for -hydroxylation are highly conserved in these enzymes.

Results and Discussion
Structure Determination-The plasmid, pCW-4B1 #7, was employed for the expression of the previously characterized truncated form of rabbit P450 4B1 in E. coli (28). The coding sequence was derived from the cDNA (Uniprot, M29852) cloned originally by Gasser and Philpot (31). This truncated P450 4B1 expression construct encodes amino acids 20 -506 at the C terminus of the native protein (UniProt, P15128). Codons encoding Met-Ala precede the native codon for Gly-20 to provide a translation initiation site and an expression optimized second codon. Additionally, codons for Ser-Thr-His-His-His-His-His-His were inserted following the native C terminus of 4B1 to facilitate purification. Although microsomal P450s have generally been expressed and crystallized without their N-terminal membrane-targeting regions, the pCW-4B1 #7 construct retains roughly half of the protein's hydrophobic TMH. The expressed protein binds tightly to membranes, and a mixture of cholic acid and Nonidet P-40 was used to extract the protein from membranes as described previously (28). Titration of the purified truncated enzyme with octane ( Fig. 1) indicated that the protein exhibits a high affinity for binding to octane as judged by conversion of the low spin enzyme to the high spin state (32) with an estimated dissociation constant of 0.34 M. This value is similar to a K d of 0.13 M determined for fulllength P450 4B1.
The isolated protein was crystallized in the presence of octane, and the structure of the P450 4B1-octane complex was determined using X-ray diffraction data collected from a single crystal at 100 K on Stanford Synchrotron Radiation Lightsource (SSRL) beam line 12-2. Initial phasing was obtained by molecular replacement using Phaser (33) as implemented in Phenix (34) with the structure of P450 46A1 (PDB code 2Q9F) as the search model. This search identified the enantiomeric space group as P 32 2 1 with a single protein chain in the asymmetric unit. P450 46A1 (an enzyme that hydroxylates the alkyl side chain of cholesterol) is the most closely related and structurally characterized P450 by sequence similarity. Although the sequence identity was low (Ͻ30% identity), detailed 2mF o Ϫ DF c and mF o Ϫ DF c maps provided evidence for the covalently bound heme prosthetic group, the presence of the substrate octane, and sequence-related side-chain differences. The Autobuild module in Phenix was able to generate an initial model with side chains for much of the core helix bundle as well as backbone traces for other regions of regular secondary structure. The final model was built and refined iteratively against the data to a limiting resolution of 2.70 Å using Coot (35) and Phenix, respectively. The model encompasses residues 20 -501 of the protein ( Fig. 2A). Residue numbers in the structure correspond to that of the native protein (UniProt, P15128). The five remaining residues of the native protein and the His tag at the C terminus as well as the two amino acids added at the N terminus to facilitate purification and initiate mRNA translation, respectively, could not be modeled. Additionally, residues 196 -200 in the connector between helices E and F and residues 272-276 in the connector between helices G and H could not be modeled as these surface loops were not well defined by the electron density maps. The overall fit of the model to the data and quality assessment statistics are documented in Table 1. Additional validation results are available from the PDB, code 5T6Q.
Two cysteine pairs were identified that could potentially form disulfide bonds. One pair, Cys-364 (helix K) and Cys-336  The truncated N-terminal TMH domain is colored dark gray, the polar linker region is colored light gray, and the catalytic domain is colored from red beginning with residue 42 to blue at the C terminus. The heme prosthetic group is depicted as a stick figure with the iron shown as a sphere. The semitransparent tubes illustrate tunnels rendered by Mole leading from the activesite cavity above the heme to the surface of the enzyme. B, linker region and TMH form a continuous AЉ helix. Polar interactions between the linker region and the catalytic domain are depicted by dashed lines between the labeled amino acid residues shown as stick figures. Nitrogen, oxygen, and iron atoms are colored blue, red, and orange, respectively. C, trajectory of the AЉ helix relative to the catalytic domain is compared with that of human P450 3A4 (PDB code 1TQN) and P450 51A1 from S. cerevisiae (PDB code 4LXJ). The 51A1 TMH is colored dark green; the relatively short polar linker is lime green, and a portion of the catalytic domain is depicted in pale green. Helix AЉ in the linker region of 3A4 is colored copper, and a portion of the 3A4 catalytic domain is colored pale yellow.
(helix J), with the sulfur atoms separated by 3.6 Å is conserved in other family 4 enzymes. A second pair, Cys-368 (helix K) and Cys-325 (helix I), with their sulfur atoms separated by 4.0 Å is not conserved in family 4B. These distances are too large to reflect covalent bonding between the two pairs of sulfur atoms. As synchrotron radiation can reduce disulfide bonds (36), it is possible that one or both pairs were bonded in the protein prior to data collection.
Structural Overview-The tertiary structure and major elements of secondary structure of the P450 4B1 catalytic domain resemble that of other mammalian microsomal P450s. The distal surface of the catalytic domain from the heme prosthetic group of the enzyme is shown in Fig. 2A. The polypeptide chain of the catalytic domain is traced from red to blue beginning with proline 42. Conserved secondary structures are labeled according to commonly observed conventions for P450s. Substrates bind in an internal cavity above the heme and below helix G. A solvent channel rendered with a transparent surface ( Fig. 2A) was identified by MOLE 2.0 (37). There is a constriction in the channel with a minimum radius of 1.6 Å that isolates the substrate-binding cavity from the protein surface, and the channel would need to open further for the passage of the substrate or product between the active site and the exterior. The channel branches before reaching the surface with the less restrictive and shorter channel exiting at the protein outer surface between helix F ( Fig. 2A, green), helix A (red), and ␤-sheet 4 (blue). The second and more constricted solvent channel splits from the larger channel and exits between the ␤-sheet 1 ( Fig. 2A, orange and cyan strands), the helix B-C loop (orangeyellow), and the helix F-GЈ loop (green).
The construct used for crystallization of P450 4B1 retains roughly half, residues 20 -29, of the predicted TMH, and residues 7-29 of the native protein. The truncated hydrophobic TMH ( Fig. 2A, dark gray) and most of the polar linker region (light gray), which connects the TMH to the catalytic domain, form a continuous AЉ helix (gray) that lies adjacent to ␤-sheet 3 (cyan) and ␤-sheet 1 (orange and cyan). The AЉ helix extends from the surface proximal to the heme beyond the distal surface along the side of the protein (Fig. 2, A-C). The linker region (light gray) comprising the C-terminal portion of helix AЉ is positioned similarly to the AЉ helix seen for the linker region in structures of P450 3A4 (Fig. 2C) expressed without its TMH. The linker region of helix AЉ interacts exclusively with the catalytic domain and does not exhibit interactions with other protein chains in the crystal lattice. In addition to extensive hydrophobic contacts with amino acid side chains on ␤-sheets 1 and 3, there are polar interactions between Arg-31 and Gln-82 on the first turn in ␤-sheet 1, and between Arg-30 and a pair of residues, Asp-108 on the helix B-C loop and Gln-381 on ␤-sheet 1 (Fig. 2B). After passing ␤-sheet 1, the surface of the catalytic domain diverges from the TMH portion of helix AЉ. The trajectory of the AЉ helix differs from that observed for Saccharomyces cerevisiae Cyp51A1 (Fig. 2C), which was crystallized with its complete N-terminal region (38). In the CYP51A1 structure, the shorter polar linker region makes a more acute turn between the catalytic domain and the hydrophobic TMH AЉ, which passes on the side of the first turn in ␤-sheet 1 that is opposite from the path of the AЉ helix seen in CYP4B1 (Fig. 2C). The interactions between the TMH and the catalytic domain observed in either structure may change when the proteins are bound to phospholipid bilayers with the TMH traversing the bilayer and the catalytic domains embedded in proximal leaflet of the bilayer as predicted by molecular dynamics simulations for other P450s (39,40).
Covalent Linkage of the Heme and Glu-310 -The heme cofactor is the catalytic center of the protein, and in P450 4B1, the 5-methyl group of the heme is linked covalently to protein by an ester bond with Glu-310 that forms auto-catalytically (21). The ester bond is clearly evident as defined by continuous electron density shown in Fig. 3A, which displays the refined structure with an unbiased 2mF o Ϫ DF c electron density omit map computed without the heme and Glu-310 in the phasing model used for calculation of the map. Although it was reported previously that P450 4B1 expressed in E. coli might incorporate 40% of the heme in an alternative conformation that placed the 8-methyl group near Glu-310 (41), there was no evidence for this in the crystallized protein based on well defined density for the location of the vinyl groups that define the asymmetry of the heme and the absence of additional electron density for the alternative conformation of the heme.
The covalent binding is associated with an out-of-plane distortion of the 5-methyl group and the attached heme pyrrole relative to plane of the macrocyclic ring of the heme (Fig. 3B). Mammalian peroxidases exhibit similar covalent linkages between acidic side chains of the peroxidases and the 5-methyl and 8-methyl carbons of their heme prosthetic groups, and similar out-of-plane distortions of their heme prosthetic groups are evident in structures of these enzymes (42). Resonance Raman studies suggest that the out-of-plane distortions of the heme in lactoperoxidase are relaxed when the enzyme is cleaved by limited proteolysis suggesting that protein interactions with the heme in the enzyme's active site induce the distortion (43). For the final refinement of the model, a semi-empirical quantum mechanical approach was employed to minimize the Glu-310/heme ester geometry in conjunction with maximizing of the fit with the X-ray diffraction data using the DivCon 6.1 module with Phenix 1.10 (44). The PM6 Hamiltonian, which includes optimized parameters for the iron, was used for this purpose. The main advantage of this approach is optimization of the bond lengths and angles that reflect the formation of the ester bond and the effects of out-of-plane deformations on electron delocalization in the macrocyclic structure of the heme. When compared with a starting structure refined with the conventional heme restraint file, refinement using the DivCon plugin with Phenix reduced the computed strain energy relative to the starting model and improved the fit to the electron density based on real space residuals and real space correlation coefficients.
Additionally, the electron density maps (Fig. 3A) indicated that the orientation of the heme 7-propionate was on the substrate side of the macrocyclic ring of the heme where it interacts with octane (Fig. 3, C and D) as it approaches the heme iron. Although this orientation of the 7-propionate is commonly seen in structures of prokaryotic P450s, it is unusual for struc-tures of mammalian microsomal P450s (45). Other mammalian P450 monooxygenases that exhibit this 7-propionate orientation are CYP51A1 (the sterol 14␣-demethylase) and CYP46A1, and a similar heme 7-propionate interaction with a tyrosine is evident in structures of CYP51A1 and CYP46A1. In contrast, the heme 7-propionate in structures of most mammalian microsomal P450s is oriented on the other side of the heme plane (45), where it interacts with basic residues corresponding to Arg-380 of 4B1 (Fig. 3D) that resides on the strand of ␤-sheet 1 nearest to the heme. In the structure of 4B1, there is a bridging water molecule between the 7-propionate and Arg-380. The water molecule exhibits additional hydrogen bonding interactions with the side-chain hydroxyl group of Ser-442 and the carbonyl oxygen of Arg-446. This arginine also exhibits hydrogen bond stabilized ionic interactions with the heme 6-propionate. The heme 7-propionate exhibits additional H-bonds with Lys-105 and Tyr-110. Tyr-110 together with the heme 7-propionate form part of the surface of the active-site cavity that contributes to positioning of the substrate relative to the heme iron ( Figs. 3 and 4).
Substrate-binding Site-Electron density maps indicated that a molecule of octane was bound in the active site with a terminal carbon close to the heme iron (3.4 Å) (Fig. 3C). The active-site cavity of P450 4B1 is highly optimized for -hydroxylation of aliphatic substrates by providing a hydrophobic cavity that constrains the linear n-octane in a plane. As indicated earlier, hydrogen abstraction is generally disfavored for primary C-H bonds due to their higher bond strengths relative to adjacent secondary or tertiary C-H bonds or sites where functionalization reduces the bond strength, e.g. benzyl groups or allylic carbons (15)(16)(17). As a consequence, P450 4B1 must restrict access of the adjacent secondary C-H bonds to reactive intermediate to display a preference for hydroxylation of the primary C-H bond in P450 4B1 (46) as well as in other -hydroxylases (18). The active-site cavity of the P450 4B1 octane complex exhibits a slot-like active site that is narrow near the heme iron and broadens near the solvent channel that exits the active site between Gln-218, Gln-377, Tyr-379, and Pro-376 (Fig. 4A). The narrowest aspect of the cavity largely constrains octane in a plane. The depth of the slot is greater and accommodates the alternating up and down orientation of the carbon atoms along the aliphatic chain (Fig. 4B) until the active-site cavity broadens near the solvent channel to the surface of the protein. Branched substrates such as 4-methylheptane, valprioic acid, and other short carboxylic acids (46) are likely to be accommodated in this narrow cavity without significant changes in the protein. As a result of these narrow constraints, the primary C-H bonds are more accessible to the reactive oxygen than the secondary C-H bonds of the C2 and C3 carbons. As shown in Fig. 4C, the amino acid residues lining the active-site cavity as well as adjacent residues of rabbit P450 4B1 are highly conserved for other characterized mammalian P450 4B1's, in sharp contrast to the high degree of active site divergence evident for family 2 paralogues that contribute to a broad capacity for xenobiotic metabolism (47). Moreover, the residues that position the substrate near the heme iron (labeled with bold text in Fig. 4, A and B) are highly conserved in human -hydroxylases (Fig. 5). Features Supporting -Hydroxylation-The importance of the Glu-310 ester bond with the 5-methyl of the heme can be seen as both the C␤ and C␥ carbons of Glu-310 contact the octane molecule, and when oxygen binds to the iron, octane will need to move outward from 3.14 to Ͼ4 Å from the iron to accommodate it. The presence of the Glu-310 C␥ contact will provide an increasing critical restraint for positioning a terminal C-H bond near the reactive iron at the larger distance (Fig.  4, D and B). Such a shift is likely to accommodate the binding of molecular oxygen to the heme iron during the catalytic cycle. Additionally, the ester bond formed by Glu-310 stabilizes this interaction because of increased rigidity of the side chain in relation to the heme prosthetic group, which prevents alterna-tive conformations of the Glu-310 side chain that might conflict with substrate binding in the active site. Moreover, the formation of the ester bond creates a less polar active site for the binding of the hydrocarbon. In contrast, the charged non-esterified glutamate might stabilize the binding of water in the active site or interact directly with the heme iron to compete with substrate binding as seen in a structure of the P450 BM3 A264E mutant (48). As Glu-310 is tethered to the heme, no unfavorable loss of Glu-310 side-chain conformational entropy is associated with substrate binding.
In addition to the glutamic acid residue 310, flanking residues on helix I are also conserved in human P450 family 4 -hydroxylases (6), as illustrated in Fig. 5, as well as in other family 4  (Fig. 5). C, portions of an amino acid sequence alignment for characterized P450 4B1 proteins from mammalian species with the active-site residues in rabbit 4B1 indicated by asterisks. Regions containing amino acids residues in rabbit 4B1 that form the substrate-binding cavity are shown with the active-site residues of rabbit 4B1 designated by asterisks. The regions correspond to five of the six SRS regions defined by Gotoh (60). The residue numbering corresponds to rabbit 4B1. A consensus sequence is shown for -hydroxylases. Glu-310 is not conserved in the four non--hydroxylases, and the latter show additional differences from the -hydroxylases for active-site residues. A complete sequence alignment is available as supplemental Fig. S1.
P450s in a variety of species (19), although this motif is less conserved in family members that catalyze other reactions, such as human P450s 4F12, 4F8, 4X1, and 4Z1 that do not have a glutamic acid residue corresponding to Glu-310 of 4B1. Two of these residues compensate for hydrogen bonds that are lost due to extension of the turn in helix I adjacent to the Glu-310. Conserved His-312 donates an H-bond to the carbonyl of Met-308 (Fig. 3, A and C). Similarly, the side-chain hydroxyl group of conserved Thr-314 forms an H-bond with the Glu-310 carbonyl oxygen in the peptide bond with Gly-311 (Fig. 3, A and B).
A third conserved residue, Phe-309, is oriented to cap the height of the cavity above the heme iron and the terminal carbon of octane. The cap over the substrate-binding cavity is propagated by the close packing of the Leu-485 (Fig. 4, dark blue carbons) on the turn in ␤-sheet 4 of the C-terminal loop ( Fig. 2A). Both residues are conserved in the human family 4 -hydroxylases. This constraint on the substrate is reinforced by a secondary layer of bulky amino acid side chains on helices BЈ, F, and G. Another conserved residue in the human -hydroxylases, Val-375, on the connector between helix K and ␤-sheet 1-3 (Fig. 4, cyan carbons) fills the space under the C-terminal loop and contacts Thr-314 to form the side of the cavity opposite Glu-310 (Fig. 4A). An additional residue conserved in the human -hydroxylases, Leu-122 (yellow carbons) on the ␤-turn between helices BЈ and C, fills the space between Glu-310 and Tyr-110 (Fig. 4, yellow-orange carbons). Together, these highly conserved residues form the narrow cavity for the approach of octane to the heme iron (Fig. 4). In the human family 4 enzymes that lack the glutamic acid corresponding to Glu-310, these additional active-site residues are less well conserved (Fig. 5) indicating that the catalytic divergence of these enzymes involves additional changes near the catalytic center of the active site. Although the substrate profiles for human -hydroxylases differ, these comparisons suggest that the human -hydroxylases exhibit very similar features for channeling the aliphatic ends of their substrates to the reactive oxygen as would be expected to achieve a preference for oxygenation of the primary C-H bonds.
Effects of Amino Acid Substitutions for Glu-310 -The effects of E310A, E310D, and E310Q substitutions on the regioselectivity of 4B1 catalysis for and -1 hydroxylation of heptane, octane, nonane, and decane were determined using reconstituted full-length enzymes ( Fig. 6 and Table 2). Of the three mutants, the E310A mutant exhibited the highest rates of total metabolite formation but with a reduction in the preference of relative to -1 hydroxylation. Nevertheless, a preference for -hydroxylation by the E310A enzyme was retained for n-heptane and n-octane but not for n-nonane and n-decane, which is consistent with an earlier report that substitution of an alanine for Glu-310 in 4B1 reduces the preference for -hydroxylation of lauric acid relative to hydroxylation of adjacent secondary C-H bonds (21). The retention of -hydroxylase activity by the E310A enzyme for the shortest chain length substrates is likely to reflect other conserved features of the active site that restrains substrates near the heme iron in the absence of the conserved glutamic acid and compensates for changes related to the loss of the covalent linkage of the heme to the enzyme. The E310D and E310Q mutants have longer side chains than the E310A mutant, which favors retention of the preference for -hydroxylation albeit with lower rates. This suggests that when the n-alkane is bound in a position for hydroxylation, the aspartate and glutamine side chains adopt a similar orientation as the ester-linked side chain of Glu-310 within the crowded active-site cavity. In the case of the aspartate substitution, the autocatalytic formation of the 5-methyl radical intermediate is likely to occur based on conversion of the 5-methyl to a 5-hydroxymethyl group, but the shorter reach of the aspartate does not permit formation of the ester. Isotope enrichment studies indicate that the proposed 5-methyl radical intermediate reacts with water to form a 5-hydroxymethyl group in the E310D mutant (49). Although the aspartate and glutamine side chains cannot form the ester bond, their greater conformational entropies relative to the tethered glutamate are likely to reduce the rate of reaction even though they restrain the substrate for -hydroxylation.
As reported previously for the wild-type enzyme expressed in insect cells, the ratio ofto -1 hydroxylation diminishes with the increase in substrate chain length from heptane (23-fold) to decane (1.6-fold), as does the total rate of metabolite formation (46). Similar values were obtained with the reconstituted fulllength 4B1 expressed and purified from E. coli (Table 2 and Fig.  6). This dependence of rates of formation on substrate chain length is also evident for the E310A, E310D, and E310Q mutants, although n-heptane is metabolized at lower rates than n-octane by these mutants. As seen in Fig. 4, octane almost completely fills the substrate-binding cavity suggesting that the binding of longer fatty acids and n-alkanes might effect changes in the cavity size. Consistent with this notion, the rates of hydroxylation and the ratio of to -1 hydroxylation decreased for n-alkanes with increasing chain length. Additionally, the rates for fatty acid hydroxylation are typically lower than for the corresponding n-alkane, and the ratios of to -1 hydroxylation are lower (46). As the active site of the octane complex does not contain a basic residue to complement a negatively charged carboxylate moiety of fatty acids, the enzyme may  Table 2.

Structure of the P450 4B1 -Hydroxylase
adopt a more open form with greater hydration of the entrance channel to facilitate binding of the fatty acid substrates as seen in P450 2C8 (50).
Overall, results of these mutagenesis studies are consistent with optimization of the 4B1 active-site cavity for the shorter hydrocarbons, and the binding of the longer compounds may force changes in the cavity to accommodate them albeit with less energetically favored binding modes and increased access of the secondary C-H bonds to the reactive intermediate. As shown in Fig. 4C, the residues that form this active-site cavity are highly conserved in family 4B1 enzymes from other mammalian species. The residues that form the distal portion of the substrate-binding cavity diverge more extensively (Fig. 5) and potentially underlie differences in substrate profiles for other family 4 enzymes in other subfamilies when compared with 4B1. Phe-113, which impinges on the substrate-binding cavity in 4B1 and which contacts both Phe-309 and Leu-485 to cap the distal active site, is a leucine in the functionally characterized human 4A and 4F -hydroxylases (Fig. 5). These enzymes generally prefer longer chain fatty acids than 4B1. The distal end of the cavity is delineated by Gln-377 and Tyr-379 in 4B1 (Fig. 4, A  and B, cyan carbons). Interestingly, the equivalent residues have shorter or absent side chains in the sequences of the human 4A and 4F -hydroxylases (Fig. 5) suggesting that the cavity is enlarged in the 4A and 4F enzymes. Consistent with this notion, a homology model for 4A11 reveals a larger substrate-binding cavity when compared with 4B1 that accommodates longer fatty acids with the side chain of Arg-96 positioned on the second strand of the ␤-sheet 1 to facilitate the binding of lauric acid (Fig. 7) and longer fatty acids, such as palmitic and arachidonic acid, which are substrates for human CYP4A11 (51). Arg-96 of 4A11 is conserved at this alignment position in rabbit 4A4 and 4A5, rat 4A2, 4A3, and 4A8, and mouse 4a12a, 4a12b, and 4a14. Interesting, this arginine is not conserved in rabbit 4A6 and 4A7, rat 4A1, and mouse 4a10 (Fig. 7B), but these P450s have an arginine that corresponds to His-88 of 4A11 on the first strand of ␤-sheet 1, which places the side chain in the active site in a similar position for interaction with the carboxylate moieties of fatty acids.
Summary-The structure of rabbit P450 4B1 is a prototype for a group of P450 enzymes with covalently bound heme pros-thetic groups. These include family 4 enzymes annotated in insects as well as annotated and unannotated P450s in other vertebrate and invertebrate animal species that have amino acid sequences with the conserved I-helix motif seen in P450 4B1 with a glutamic acid rather than the more common alanine or glycine present in other P450s. Sequence conservation of additional amino acids that form the proximal portion of active-site cavities in other family 4 enzymes suggests that these conserved residues are likely to be retained to facilitate -hydroxylation. In contrast, the sequence divergence in the distal portions of the substrate-binding cavity are likely to underlie distinct substrate profiles exhibited among family 4 -hydroxylases.

Experimental Procedures
Materials and Reagents-The following chemicals were purchased from Sigma: ␦-aminolevulinic acid, sodium cholate, L-histidine, DNase I, PMSF, and octane. IPTG and HEGA 8 were obtained from Anatrace (Maumee, OH); lysozyme was from Worthington; nickel-nitrilotriacetic acid-agarose was from Qiagen (Valencia, CA), and HA Ultragel was obtained from PALL Life Sciences (Port Washington, NY). Pierce protease inhibitor tablets (EDTA-free) and all other chemicals were purchased from Thermo Fisher Scientific.
Expression and Purification of Truncated Rabbit P450 4B1 for Crystallization-E. coli strain DH5␣ transformed with pCW-4B1 #7 (28) was used for protein expression. Single colonies were grown in super broth at 37°C overnight. The bacterial culture was pelleted at 2500 ϫ g for 10 min and then suspended in Terrific Broth. The culture media were shaken at 200 rpm at 37°C until the A 600 reached 0.4 -0.5. After 30 min of incubation at 27°C at 120 rpm, the 4B1 protein expression was induced by addition of 1 mM IPTG and 0.5 mM ␦-aminolevulinic acid. Cells were harvested after 44 -48 h. The expression level was routinely ϳ1200 nmol of P450 per liter of bacterial culture.
Truncated P450 4B1 was purified using nickel-nitrilotriacetic acid-agarose followed by hydroxylapatite chromatography as described (28) with modifications. Briefly, the harvested cells were suspended in 20 mM potassium phosphate buffer, pH 7.4, containing 20% glycerol (v/v), and protease inhibitors (0.5 mM PMSF and EDTA-free Pierce protease inhibitor tablets  profiles for hydroxylation of n-heptane, n-octane, n-nonane, and n-  tri-sodium citrate and 15% polyethylene glycol 3350. The well contained 0.5 ml of the precipitant solution, and the sealed plate was incubated at 24°C. Prior to freezing crystals for data collection, 2 l of a solution that was prepared by combining 3:7 (v/v) aqueous ethylene glycol and a 1:1 (v/v) solution of the protein buffer and the precipitant solution was added to the drop. Crystals were harvested and flash frozen in liquid nitrogen prior to shipment to SSRL. X-ray diffraction data collected from a single crystal at 100 K on SSRL line 12-2 were used for structure determination. XDS (52) was used for integration, and Aimless (53) was used for merging and scaling of the reflections to a limiting resolution of 2.70 Å. Initial phasing was obtained by molecular replacement using Phaser as implemented in Phenix (34) with the P450 46A1 structure PDB code 2Q9F as the search model. Phaser identified a single solution (LLG ϭ 127) with one protein molecule in the asymmetric unit in the P 32 2 1 space group. Autobuild generated an initial partial model with R/Rfree of 0.36/0.42. Model building and refinement of the atomic coordinates and B-values employed Coot (35) and Phenix, respectively (34). The protein model was completed for native residues 20 -501 with the exception of short gaps in the exterior loops connecting helices E with F and G with H. Electron density was evident for octane in the active site. The 2mF o Ϫ DF c and mF o Ϫ DF c weighted electron density maps also indicated that Glu-310 was covalently bonded to the 5-methyl group of the heme as expected and that the covalent linkage induced an out-of-plane distortion of the heme. For refinement of the covalently bound heme, Glu-310 and the heme were linked to create a single residue with a unique name. Stereochemical restraints for the new residue were generated by combining the respective restraint files and replacing the CMD HMD bond with the CMD-OE2 ester bond and removal of the HMD hydrogen. Ideal bond lengths and angles for the ester were based on idealized values for glutamic acid methyl ester in the PDB Chemical Component Library. This substitution increased the CD-OE2 ideal bond length to reflect the change in the delocalization of electrons in the carboxyl group and inclusion of an ideal bond length for the OE2-CMD bond. As the electron density maps indicated an out-of-plane distortion for the 5-methyl and attached pyrrole, restraints on the planarity of the heme were either relaxed or removed for refinement so that the pyrrole rings could lie in different planes as indicated by the electron density maps. Normal dihedral restraints were retained, and riding hydrogens were used during refinement. The resulting models were used for subsequent refinements employing the DivCon 6.1 plugin module for Phenix (44) that employed the PM6 Hamiltonian to provide stereochemical restraints derived from semi-empirical quantum mechanical geometry refinement to calculate gradients for xyz refinements of the combined residue as well as amino acids adjacent to Glu-310. Final refinements converged to similar strain energies from different starting models. The riding hydrogens were removed, and the combined residue was decomposed to the Glu-310 and heme radicals with a link record in the coordinate file indicating the covalent bond between the two residues. Data integration and merging statistics as well as statistics for model to data fit and model quality assessment are reported in Table 1. Molecular graphics were generated using PyMOL.
Expression and Purification of Rabbit P450 4B1 Glu-310 Mutants-Plasmids for expression of CYP4B1 Glu-310 mutants in E. coli were generated by PCR amplification of FastBac clones used previously to express the corresponding mutants in insects cells (21). The corresponding segment of pCW-4B1 construct #1 plasmid for expression of the full-length P450 4B1 protein in E. coli (28) was replaced with the mutated sequence from the FastBac clones. The wild-type and mutant proteins were expressed and purified as described previously (28).
Ligand Binding-Purified full-length CYP4B1 (0.5 M) was first reconstituted with extruded liposomes of dilauryl phosphatidylcholine (50 M). CYP4B1 was then placed into two separate cuvettes containing 50 mM potassium phosphate at pH 7.4 in a final volume of 0.9 ml. After allowing the sample and reference cuvettes to reach room temperature, octane was titrated in from 0 to 5 M. Difference spectra were obtained over the range 350 -500 nm using an Aminco DW2 double beam spectrometer. Estimates of the dissociation constant by non-linear regression using SigmaPlot 8.02 and the following function for the enzyme-substrate complex (⌬A(390 -422)) versus octane concentration were described by Morrison (54) and are shown in Equation 1, ͓ES͔ ϭ ͓͑E͔ t ϩ ͓S͔ t ϩ K s ͒ Ϫ ͱ͓͑E͔ t ϩ ͓S͔ t ϩ K s ͒ 2 Ϫ 4 ͓E͔ t ͓S͔ t 2 (Eq. 1) For ligand binding studies with the truncated P450 4B1, which was used for crystallization, the protein prepared in the presence of lauric acid or heptane was dialyzed using sequential rounds of dilution in ligand-free buffer followed by centrifugal ultrafiltration to concentrate the protein. The ligand-free protein exhibited a visible spectrum typical of a low spin ferric P450 with Soret maximum at 417 nm. Octane was dispersed in methanol and added in small volumes to the protein solution as described for the full-length enzyme. Addition of octane shifts the Soret maximum to the high spin ferric form with maximum absorbance at 394 nm. The concentration dependence of the response was analyzed using the quadratic form of the 1-site binding equation (54) by non-linear least squares fitting using SlideWrite with the assumption that the concentration of the bound ligand was proportional to the change in absorbance relative to the maximum change of absorbance (⌬A(388 -422)) for the 1:1 complex with protein.
Homology Modeling-MODELLER (55,56) was used to generate a homology model of human P450 4A11 using the structure of rabbit 4B1 as the template. The alignment of the two structures over residues 20 -504 of the 4B1 structure was straightforward based on the similarity of the sequences and single deletion of two residues in P450 4A11 corresponding to residues 429 and 430 of P450 4B1 (supplemental Fig. S1). MODELLER generated loops to span the two gaps in external loops of the 4B1 structure. Visual comparisons of the 4A11 model with the 4B1 structure in COOT indicated that the correspondence of the regular secondary structures flanking the loops overlaid well with the template structure. The planar heme generated by Modeler was replaced with the covalently bound heme in the template to link the heme to the Glu-323 of the model, which corresponds to Glu-310 of P450 4B1. The homology model is available as supplemental File P450 -4A11homology-model.pdb. Automated docking studies employed VINA (57). AUTODOCK TOOLS (58) was used to generate PDBQT files from the coordinates for the homology model and ligand and to define the search space. OPEN BABEL (59) was used to generate PDB files for the ligand from the VINA output. The PDBQT files define rotatable bonds and charges. VINA does not use the electrostatic interactions for scoring but does identify hydrogen bond donors based on polar hydrogens and acceptors based on atom type for scoring purposes. To better weight the strong hydrogen bonds between the fatty acid carboxylate and basic amino acids, the scoring weight for hydrogen bonds was doubled using the input parameter "-hydrogen_ weight-1.2" as described in the Vina FAQ.
Author Contributions-E. F. J. and M. H. crystallized the protein, collected X-ray diffraction data, determined the structure of CYP4B1 complexed with octane, and drafted the initial manuscript. A. E. R. and B. R. B. generated and characterized the enzymatic properties of CYP4B1 and selected mutants. A. E. R. provided the expression vector for truncated CYP4B1. All authors analyzed the results, contributed to writing the manuscript, and approved the final version of the manuscript.