Keyword

1 Introduction

The first cytochrome P450 structure, P450cam or CYP101A1, was solved in the early 1980s [1, 2], followed by the second, P450BM3, in 1993 [3]. At the time of the 3rd edition of this book published in 2004, there were a total of 13 unique P450 crystal structures deposited in the Protein Data Bank (PDB). As of April 2014, the PDB lists 449 entries with the name P450 in the title and of these about 54 are unique structures. The many new structures solved since the 3rd edition include various substrate/ligand complexes, P450s in various conformational states, and a few new P450-redox protein complexes. This wealth of new structural information has been particularly useful in a better understanding of P450 dynamics and how the P450 active site adapts to substrates of diverse sizes and shapes.

2 Overall Architecture

There now are a sufficient number of structures to safely state that the overall P450-fold is quite conservative. While it remains the case that there are no nonheme proteins that exhibit the P450-fold, there now are a small handful of examples of enzymes that exhibit the P450-fold but do not catalyze traditional P450 chemistry. These include the NO reductase, P450nor [4, 5], prostacyclin synthase [6–8], allene oxide synthase [8–11], P450BSβ [12], and a related peroxygenase , CYP152L1 [13], which hydroxylates fatty acids but does so using H2O2 as the oxidant.

The structures of six P450s are shown in Fig. 1.1, while Fig. 1.2 highlights some of the key secondary structural elements. Although the overall fold is maintained, the precise positioning of various structural elements differs substantially. In general, the closer to the heme, the more conserved the structure, especially helices I and L, which directly contact the heme. As expected, those regions controlling substrate specificity differ the most, especially the B′ helix. For example, in P450eryF, the B′ helix is oriented about 90° from the orientation observed in P450cam. The effect is a substantial change in local environment, which is required for substrate selectivity.

Fig. 1.1
figure 1

A representative example of known P450 structures illustrating the common three-dimensional fold

Full size image
Fig. 1.2
figure 2

The structure of P450cam (PDB: 5CP4) with key helical segments labeled. PDB Protein Data Bank

Full size image

Not too surprisingly, the most conserved elements of the P450 structure center on the heme–thiolate oxygen activation chemistry. The most noteworthy is the β-bulge segment housing the Cys ligand (Fig. 1.3), just prior to the L helix. This rigid architecture is required to both protect that Cys ligand and hold it in place in order to be within H-bonding distances of two peptide NH groups, although the H-bonding geometry is good for only one H-bond. This arrangement is not only found in all P450s but also in two closely related enzymes, nitric oxide synthase (NOS) and chloroperoxidase (CPO). Both NOS and CPO are heme–thiolate enzymes that, like P450s, catalyze monooxygenation reactions. Like P450, the Cys ligand in CPO is near peptide bond NH groups [14]. NOS is similar, except that an H-bond is provided by the indole ring N atom of a conserved Trp residue [15–17]. Such an H-bonding arrangement is not unique to heme–thiolate proteins, but is a characteristic feature of proteins containing Cys–Fe ligation, and was first observed in the ferredoxins [18]. These H-bonds aid in regulating the heme iron redox potential [19, 20]. Without such H-bonds, the redox potential would be too low for reduction by redox partner s. Thus, it appears that the protein must provide a suitable electrostatic environment around the Cys ligand in order to maintain the redox potential in a physiologically accessible range. The same is true for a close cousin to P450, the peroxidases. Here histidine (His) serves as the axial ligand, but, in this case, it is necessary to increase rather than decrease the redox potential [21]. As a result, the His ligand H-bonds with a buried Asp residue that imparts greater imidazolate character to the His, thus lowering the heme iron redox potential [22–26].

Fig. 1.3
figure 3

The Cys ligand “loop” in P450cam. The dashed lines indicate key hydrogen bonding interactions that aid in stabilizing the Cys ligand. Cys cysteine

Full size image

The other highly conserved region involved in O2 activation is the portion of helix I near the heme Fe (Fig. 1.4). Thr252 is involved in a local helical distortion in P450cam such that the threonine (Thr) side-chain OH donates an H-bond to a peptide carbonyl oxygen that would normally be involved in an α-helical H-bond. This Thr is not strictly conserved. For example, P450eryF contains an Ala instead of a Thr [27] and P450cin has an Asn [28]. Even so, these outliers also exhibit a similar distortion in the I helix. This arrangement is thought to be quite important for the proper delivery of protons to the iron-linked oxygen required for cleavage of the O–O bond , thus generating the active Fe–O hydroxylating species. The growing consensus is that ordered solvent at the active site serves as the direct proton donor to the iron-linked dioxygen [29–32]. P450–oxy complexes tend to be rather unstable, which is why there are only two crystal structures of P450–oxy complexes: P450cam [31, 33] and P450eryF [34]. In the P450cam–oxy complex, the I helix opens up slightly which provides sufficient room for two new waters to move into the active site. These waters form an H-bonded network that is thought to be important for the proper delivery of protons to dioxygen in order to promote heterolytic cleavage of the O–O bond (Fig. 1.4). While the positioning of new waters in the active site requires changes in the I helix, there are no changes in the P450eryF–oxy complex except for the repositioning of a water molecule. Since the conserved Thr252 found in P450cam is replaced by Ala in P450eryF (Fig. 1.5), the I helix is already in an open conformation similar to that of P450cam–oxy. It appears that P450eryF uses a substrate-assisted mechanism [35] since a substrate OH anchors the key water in place via H-bonding and is essential for activity. While the details of the proton shuttle machinery may differ from one P450 to the next, the surrounding protein groups and, in at least one case, the substrate, generally position solvent in the active site for proton delivery to dioxygen resulting in cleavage of the O–O bond.

Fig. 1.4
figure 4

A comparison of the I helix region in ferric and oxy-P450cam (PDB: 2A1M). When O2 binds, the I helix opens up and the H-bond between Thr252 and Glu248 is broken. This opening enables additional waters to move into the active site that are thought to be critical for completing a protein relay network required for O2 activation

Full size image
Fig. 1.5
figure 5

A comparison of the solvent-mediated hydrogen bonding network in oxy-P450eryF (PDB: 1Z8O) and the oxy complex of P450cam. Unlike in P450cam, there is very little movement of the I helix in P450eryF when O2 binds. This is probably because Thr252 in P450cam is replaced by Ala245 in P450eryF. As a result, the I helix is already more open in P450eryF

Full size image

3 Structural Features for Membrane Binding

In contrast to prokaryotic P450s, eukaryotic P450s are generally membrane-bound proteins. Most eukaryotic P450s are incorporated into the endoplasmic reticulum. However, several mammalian P450s that participate in the synthesis of sterols, steroids, and bile acids are located on the matrix side of the mitochondrial inner membrane. A longer N-terminal polypeptide chain of roughly 30–50 amino acids precedes the catalytic domain in eukaryotic P450s and mediates membrane targeting. In the case of mitochondrial P450 s, the targeting sequences are cleaved during import of the protein into the mitochondrion [36]. In contrast, the leader sequences of microsomal P450s are retained and inserted into the endoplasmic reticulum during protein synthesis [37]. The insertion process stops at the end of a hydrophobic stretch of roughly 20 amino acid residues, which are likely to form a helix in order to reduce the energetic costs of placing the polar peptide backbone in the nonpolar core of the bilayer [38]. A short linker region of about ten amino acids, which often includes positively charged amino acid residues, connects the transmembrane helix (TMH) to a generally conserved proline at the N-terminus of the structurally conserved P450-fold. The length of the 20 amino acid TMH corresponds roughly to the 3-nm width of the hydrocarbon core of the bilayer [39]. Additionally, the polar head groups of the phospholipids add another 1 nm outer layer on each side of the hydrophobic core, suggesting that a portion of the linker region resides in the polar head group layer.

The TMH is not required for function, as illustrated by the expression and successful reconstitution of several P450 monooxygenases in which this region was deleted [40–44]. Almost all of the currently available crystal structures have been determined for microsomal P450s expressed and crystallized without the TMH [45]. Initial structures of the human aromatase, CYP19A1, are an exception. Although the full-length aromatase was crystallized, the TMH and linker regions were disordered in the crystal [46]. Subsequently, engineered mutants of aromatase were expressed in Escherichia coli without the TMH, and these structures were not significantly affected by the absence of TMH [47].

Recently, additional evidence for the helical nature of the TMH was obtained from solid-state nuclear magnetic resonance (NMR) studies of rabbit microsomal CYP2B4 incorporated into magnetically oriented bicelles [48] and from a crystal structure reported for full-length, Saccharomyces cerevisiae CYP51A1, a sterol 14α-demethylase [49]. This crystal structure includes the linker region, TMH, and an additional amphipathic helix at the N-terminus. Interactions of the latter with a neighboring molecule in the crystal lattice contributed to a well-ordered N-terminus for structure determination. As a result, the predicted helical secondary structure of the TMH was confirmed, and a role for the additional amphipathic N-terminal helix in membrane binding to the distal leaflet of the bilayer was proposed, as illustrated in Fig. 1.6. Although the N-terminal amphipathic helix is not a general feature of microsomal P450s, this structure suggests that some P450s with extended N-terminal sequences could exhibit additional membrane interactions with the distal leaflet of the bilayer. In the S. cerevisiae CYP51A1 structure, the C-terminal end of the 24-residue TMH lies along the surface of the catalytic domain and passes from the proximal face to the distal face of the P450 along a trajectory that is roughly parallel with β-sheet 1, Fig. 1.6. The C-terminal end of the TMH helix corresponds to the linker region and is amphipathic with polar residues exhibiting hydrogen-bonding interactions with the catalytic domain and hydrophobic residues on the outer surface. This suggests that the observed trajectory is likely to be maintained when the enzyme is bound to the membrane [49]. As depicted in Fig. 1.6, a portion of the catalytic domain is likely to protrude into the membrane when the TMH resides in the lipid core. The surface of the catalytic domain surrounding this region is relatively hydrophobic for CYP51A1 (Fig. 1.7) as well as other membrane P450s [45, 50], which is likely to facilitate interactions with the lipid core of the bilayer.

Fig. 1.6
figure 6

Hypothetical model for the membrane binding of microsomal P450s. The cartoon depicts the experimentally determined fold of full-length, Saccharomyces cerevisiae CYP51A1 (PDB: 4KOF). For reference, the structure of the TMH is flanked by modeled arrays of phospholipid molecules depicted as CPK atoms. The amphipathic N-terminal helix is positioned at the transition of the polar head group and the hydrophobic layers with its hydrophobic surface oriented toward the lipid layer. The heme and bound inhibitor, itraconazole, are also rendered as CPK atoms. Itraconazole passes out of the access channel between helices A′ and F′, which are oriented toward the lipid portion of the bilayer. TMH transmembrane helix

Full size image
Fig. 1.7
figure 7

Surface rendering of full-length, Saccharomyces cerevisiae CYP51A1, (PDB: 4KOF) with acidic and basic residues colored black and gray, respectively. Note the relative absence of charged residues on the surface of the catalytic domain surrounding the entrance channel and the TMH. Itraconazole is depicted as a stick figure in the entrance channel. The hydrophobic surface surrounding the entrance channel is oriented toward the membrane in Fig. 1.6. efj1

Full size image

This hydrophobic surface is formed by the N-terminal portion of the catalytic domain together with the helix F–G region, and there are distinctive structural differences between mammalian P450s and soluble prokaryotic P450s for this portion of the catalytic domain. The initial comparison of the first structure of a microsomal P450, CYP2C5, with structures of CYP102 and CYP101 [50, 51] indicated that the N-terminal portion of the catalytic domain from the linker region and to the helix B–C loop of CYP2C5 is shifted significantly toward the proximal face when compared to structures of prokaryotic P450s (Fig. 1.8). The generality of this observation was established by a retrospective analysis of a larger number of structures from diverse eukaryotic and soluble prokaryotic P450s by Denisov et al. [52]. As a result of this shift in position, the heme A-ring propionate is oriented toward the proximal side of the heme plane in most mammalian membrane P450s, where it often interacts with basic amino acid side chains. Notable exceptions are eukaryotic CYP51A1 and the non-monooxygenases, CYP8A1, a prostacyclin synthase and CYP74A1, a plant allene oxide synthetase. The heme A-ring propionate resides more typically on the distal side of the heme plane in prokaryotic P450s, with some exceptions. This shifted N-terminal/β-sheet domain resides near the connector between helices F and G, which is typically longer in eukaryotic P450s than in soluble, prokaryotic P450s. The structure of the F–G helical region varies extensively between mammalian P450s and often exhibits two short helices, F′ and G′. Together these elements form the hydrophobic surface near the N-terminus of the catalytic domain that is likely to be inserted into the membrane [45, 50].

Fig. 1.8
figure 8

Superposition of CYP101A1 ( light gray) and CYP2C8 ( dark gray). The hemes are shown as stick figures with oxygen atoms colored black. The heme iron is depicted by a sphere. Although helices I through K superimpose well, the N-terminal region is shifted outward for CYP2C8 relative CYP101. Different orientations are evident for the heme A-ring propionates

Full size image

The orientation of the hydrophobic surface toward the membrane is supported by studies indicating that antibody epitopes in this region are inaccessible to the antibody when CYP2B4 is in its native membrane [53], whereas epitopes on other portions of the molecule react with their respective antibodies. These and other epitope -mapping studies indicate that extensive portions of the surfaces of drug-metabolizing P450s are accessible to the antibodies when bound to membranes, as shown in Fig. 1.9 and reviewed in more detail [54]. Atomic force microscopy experiments estimate that the height of microsomal CYP2B4 above a model phospholipid membrane is roughly 35–45 nm [55]. This would require a portion of the protein to be buried in the membrane, which is likely to be the hydrophobic region near the N-terminus of the catalytic domain . Additionally, studies of the association of CYP2B4 with Langmuir–Blodgett phospholipid monolayers indicate that the protein displaces an area that is larger than a single TMH [56]. This result would be consistent with the penetration of the hydrophobic surface of the protein into the adjacent leaflet of the lipid bilayer.

Fig. 1.9
figure 9

CPK rendering of the proximal (a) and distal surfaces (b) of CYP2C5 (PDB: 1N6B). Antibody epitopes recognized when the P450s are bound to microsomal membranes are colored dark gray, as reviewed [54]. Several conserved amino acid side chains that have been implicated in P450 reductase interactions with CYP2B4 [156] are colored medium gray. The orientation of the protein is similar to that depicted in Fig. 1.6 efj-1 with the N-terminus of the catalytic domain positioned toward the bottom of the figure. (Reproduced from Cytochrome P450, Third Edition with permission from Springer Science+Business Media)

Full size image

Molecular dynamics (MD) simulations of the binding of human microsomal P450s with phospholipid bilayers, reported initially for CYP2C9 [57–59] and CYP3A4 [52, 60], observed stable binding orientations for the catalytic domain s with the hydrophobic surface of the catalytic domain immersed in the proximal leaflet of the phospholipid bilayer, Fig. 1.10. The structure of the catalytic domain was reported to be stable and exhibiting dynamic motion with root-mean-square deviation (RMSD) values of less than 2.5 Å from the starting structures.

Fig. 1.10
figure 10

Immersion of CYP2C9 in a dioleoylphosphatidylcholine (DOPC) lipid bilayer. ( Left) Overlaid snapshots of CYP2C9 taken at 0.1 and 1 μs molecular dynamics (MD) simulations showing that the catalytic domain is immersed in a membrane depression framed by lipid phosphate groups (shown as orange spheres). Water molecules are not shown for clarity. The N-terminal helix shows precessional movement about the bilayer normal. The fold of the catalytic domain is conserved and agrees with that observed in X-ray crystallography experiments. ( Right) Snapshot taken at 1 μs of MD simulation showing positions of active site access and egress channels computed from the heme moiety using MOLE 2.0.20. The water channel ( white) points toward the cytosolic environment, whereas solvent channel S ( blue) points above the lipid head groups. All other channels point inside the bilayer. Channels 2e, 2c, and 3 point into the lipid head group region, whereas channels 4 and 2ac point below the lipid head groups. The heme tilt angle θ (between the heme plane and the bilayer normal z, i.e., defined according to Baylon et al. [60] is depicted. (Reprinted with permission from [59], copyright 2013 American Chemical Society. The channels are designated as described [93])

Full size image

The maximum heights of the catalytic domain s above the membrane surface in these MD simulations are similar to that of 35–45 nm determined by atomic force microscopy for CYP2B4 [55]. Additionally, the tilt of the heme plane relative to the membrane normal, Fig. 1.10, in the models of membrane-binding interactions can be compared to results from biophysical studies for this angle. This tilt angle has been estimated for CYP17A1 and CYP21A2 based on the anisotropic decay of the absorption spectrum following photodissociation of carbon monoxide complexes by polarized light. This approach gives two solutions for the angle of the orientation of the heme plane relative to the membrane normal of either 43° or 27° and 52° or 12°, respectively for the two enzymes [61]. The larger values are similar to a single value for tilt angle of 59.7 ± 4.1° estimated from the dichroic ratio observed for the absorption of visible light by the heme chromophore of P450 3A4 bound to nanodisc membranes. Tilt angles for the heme in the initial MD simulations for P450 2C9 were reported to be 55 ± 5° [58], and in additional MD simulations for P450s 1A2, 2A6, 2C9, 2D6, 2E1, and 3A4, using similar conditions, the heme-tilt angles differed between P450s and ranged from 56 ± 5° for CYP3A4 to 72 ± 6° for CYP2D6 [59]. Differences between P450s are not unexpected, as the distal surfaces of microsomal P450s differ significantly, and these differences are likely to affect the angle tilt and extent of membrane insertion.

Heme-tilt angles observed for CYP3A4 in MD simulations from two different studies were reported to be 68.7°–75.9° [60] and 56 ± 5° [59]. The reported differences between the two MD simulations could reflect differences in the model membranes used in the simulations, as well as different initial models for the N-terminus used in the MD simulations. As structures for the native N-terminal domains were not available for CYP3A4, and the other proteins characterized in these studies, they were modeled de novo with the hydrophobic portion of the N-terminus modeled as a TMH. The structure of the linker region in these proteins is less certain, and is likely to vary between P450s. X-ray crystal structures of mammalian microsomal P450s have generally been determined for proteins without their N-terminal TMH, and, in most cases, the native linker regions of family 2 P450s were modified to correspond to the linker region of CYP2C3, as described for CYP2C5, [40, 41, 50]. Moreover, the structures of these short N-terminal regions have not been defined for many P450s. CYP3A4 is an exception [62, 63], and the native linker region exhibits an Aʺ helix following a turn that directs the polypeptide chain along β-sheet 1 from the N-terminus of the catalytic domain near the hydrophilic proximal face toward the hydrophobic distal surface. This trajectory is similar, but not identical, to that observed more recently for the structure of full-length S. cerevisiae CYP51A1, Fig. 1.6. The initial model used by Baylon et al. [60] incorporated a flexible link between helix Aʺ and the TMH, which provides some flexibility for the orientation of the TMH independently of the catalytic domain during the MD simulation. The initial model used by Berka et al. [59] for the N-terminus CYP3A4 was based on their earlier equilibrated CYP2C9 model obtained following a 0.25 μs MD simulation [58]. Interestingly, the helix Aʺ region and the TMH of the initial model of P450 2C9 were built as a continuous helix, but a kink developed between helix Aʺ region and the TMH during the MD simulation that allowed the polar Arg side chains in the linker region to reside in the polar region of the bilayer, and the TMH to span the lipid core of the membrane, as illustrated in Fig. 1.10 by a 1 μs equilibrated model from a later study [59]. Helix Aʺ may not be a generally conserved feature for linker regions, as the same segment of the native linker regions does not exhibit an Aʺ helix in the structure of human CYP1A2 [64], and is not evident in the MD simulation model of CYP1A2 [59]. Both CYP1A2 and CYP3A4 exhibit short N-terminal helices that are roughly orthogonal to the TMH and that are positioned at the interface between the polar head group and lipid layers of the distal leaflet of the bilayer in the MD simulations of Berka et al. [59]. These models are similar to the bimodal membrane binding proposed for full-length S. cerevisiae CYP51A1 [49].

Although the MD simulations generally support the notion that a portion of the distal surface is embedded in the membrane surface, the results of biophysical experiments and topology studies often show differences that are difficult to reconcile with a single model. Fluorescent quenching of tryptophan residues introduced on the surface of CYP2C2 by site-directed mutagenesis suggested that residues 36 and 69 flanking helix A and 380 in β-sheet 2 of CYP2C2 are inserted into the fatty acyl core of the bilayer, while residue 80 on helix B and 225 at the turn between helices F′ and G′ are in the polar region of the phospholipid bilayer [65], leading the authors to propose a more vertical orientation for CYP2C2 than was observed in the MD simulations for the closely related CYP2C9 [57, 58]. Experimental evidence indicates that P450s are present as both monomers and dimers in membranes [66, 67], and a more vertical orientation relative to the membrane surface would be consistent with models for the dimerization of the catalytic domain of N-terminally truncated P450 2C8 in aqueous solution that involve interactions of the helix F–G loop region [63]. This model for the dimerization of 2C8 is supported by cross-linking studies for the membrane-bound full-length CYP2C8 [68]. Additionally, these cross-linking studies implicated the linker region and TMH in the dimerization of membrane-bound, full-length CYP2C8 expressed in mammalian cells or in E. coli membranes. Cys-scanning mutagenesis indicated that reactive cysteines reside on a single side of the TMH, whereas several consecutive residues were reactive in the linker region suggestive of a more flexible structure. This flexibility is necessary for reorientation of the proximal faces relative to TMH in order to form a P450 dimer through interactions of the helix F–G region. P450 dimerization in membranes is thought, in some cases, to inhibit reduction by the microsomal cytochrome P450 reductase, so the monomer is likely to be the predominant functional form of the enzyme [66].

As mitochondrial P450 s lack the N-terminal TMHs found in microsomal P450s, the interactions of the catalytic domain with the matrix side of the inner membrane are likely to be the predominant membrane interaction. Consistent with a role for the helix A′, F′, and G′ regions in membrane binding , these regions exhibit nonpolar, exterior surfaces in structures of mitochondrial P450s 11A1 [69, 70], 11B1 [71], and 24A1 [72]. Moreover, substitutions of more polar residues for hydrophobic residues on the F′ and G′ surfaces increase salt extractability and solubility of mitochondrial P450 27A1 [73]. The helix F–G region of mitochondrial P450 11A1 is also protected from chemical modification by membrane association [74]. Similarly, microsomal P450s expressed without their TMH retain capacities to bind to phospholipid membranes, and mutations made to the helix F′ and G′ regions of microsomal P450s 2C5 [41], 2D6 [75], and 7A1 [76] facilitate extraction in high salt buffers. These observations suggest the extended loop between helices F and G in eukaryotic P450s contributes to membrane binding for both mitochondrial and microsomal P450s.

4 Conformational Dynamics for Substrate Access

Many P450 structures are in the so-called closed state with no obvious way that substrates can gain access to the active site . As a result, substrate entry and product egress may involve rather large conformational changes. Once the P450cam structure became available, an immediate puzzle was how camphor gains access to the active site since the substrate is buried, and there is no obvious opening. The substrate-free and bound structures showed no differences, although substrate-free P450cam exhibited higher thermal motion in the B′, F, and G helices, suggesting that these regions must move to allow substrate to enter the active site [77]. The first clear indication that conformational changes are important in substrate binding was the structure of palmitoleic acid bound to P450BM3 [78], which was followed by a higher-resolution structure [79]. Interestingly, the experimentally observed conformational change was correctly predicted based on computational methods [80, 81] before the substrate-bound crystal structure was solved. The main motion involves the F and G helices sliding over the surface of the I helix. This motion closes off the entry channel, indicating that substrates enter near the F/G loop region that is similar to that of P450cam.

There now are a handful of P450 structures in the open and closed forms and in all of them, the F and G helices and the F/G loop undergo large changes. Not surprisingly, the most extensive analysis has been with P450cam. In the open form [82], the F and G helices move, and the B′ helix region becomes disordered. It also has been possible to trap the P450cam access channel using a series of tethered compounds where the substrate is attached to a long linker that extends out of the active site [83, 84]. A principal component analysis of 30 different tethered compound structures indicates that there are three dominant conformational states available to P450cam: closed, partially open, and fully open [84].

Two close homologues to P450cam with about 46 % sequence identity with P450cam, CYP101D1 [85] and CYP101D2 [86], have now been characterized. Both catalyze exactly the same reaction as P450cam, but there are substantial differences with respect to the open and closed states and the relationship between spin-state and substrate binding. For example, CYP101D2 has been crystallized only in the open state, but camphor can be soaked into the crystals and binds in the active site [86]. The camphor, however, does not bind in a productive mode, but instead the carbonyl O atom of the substrate H-bonds with the water coordinated to the heme iron (Fig. 1.11). MD simulations of CYP101D2 show that this P450 can adopt various conformational states, mainly by motions of the F/G helical substrate access channel, and provides a dynamic picture of substrate binding consistent with other P450s [87]. Perhaps the most unexpected difference between P450cam and its close cousins is that camphor binding to CYP101D1 gives only about 40 % high spin even with excess substrate. In addition, the Fe2S2 ferredoxin that supports CYP101D1 catalysis, Arx , is able to reduce substrate-free 100 % low-spin CYP101D1, while only high-spin substrate-bound P450cam can be reduced by its redox partner , Pdx . In addition, Pdx can support CYP101D1 catalysis, while only Pdx can support P450cam catalysis [88]. There is nothing obvious in the structures that can explain these differences other than the fact that in CYP101D2 the substrate can bind to the low-spin open state, albeit in a nonproductive binding mode (Fig. 1.11). One simple way of rationalizing these differences is to hypothesize that CYP101D1 can bind camphor in various orientations that are consistent with a water molecule remaining coordinated to the heme iron, as in CYP101D2, thus giving a substrate-bound mostly low-spin complex. Upon reduction of the heme iron, the water ligand is displaced and the substrate can “relax” to a productive binding mode. This hypothesis requires that CYP101D1 is “looser” than P450cam and can more readily adopt the open conformation . The static X-ray structures do not reveal anything obvious to support this scenario, and proof one way or the other must await other approaches more in tune with measuring dynamic differences.

Fig. 1.11
figure 11

The open substrate-binding channel in CYP101D2 (PDB: 3NV6) [86]. The substrate camphor binds but is not oriented in the productive binding mode. Instead, the camphor carbonyl O atom H-bonds with the water coordinated to the iron

Full size image

5 Substrate Access to Membrane P450s

Similar to prokaryotic P450s, membrane P450s have been crystallized in both open and closed conformations. For example, rabbit microsomal CYP2B4 and human CYP2B6 have been crystallized in closed forms, as illustrated in Fig. 1.12, by a CYP2B6 4-(4-chlorophenyl)-imidazole complex [89] and in open forms, as illustrated by a complex of CYP2B6 with one molecule of amlodipine coordinated to the heme iron and a second molecule bound in the entry channel and protruding between helix F′ and A′ [90]. These two conformations of CYP2B6 differ in the positions of the helices A′, A, B′, F, F′, and G. Open forms of rabbit CYP2B4 have also been determined where the helix F′–G′ and helix B–C regions are displaced to a much greater extent by ligand and detergent interactions [91, 92]. Mammalian drug-metabolizing enzymes such as CYP2B4 bind a wide-range of compounds, and conformational changes are often associated with the capacities of these enzymes to facilitate the metabolic clearance of many compounds by accommodating large compounds in an open access channel [45]. Mitochondrial P450s also exhibit open and closed structures. CYP11A1 and CYP11B1 exhibit closed structures for substrate complexes with the helix F and F′ region blocking the substrate access channel described earlier for CYP101, Fig. 1.13. In contrast, mitochondrial CYP24A1 was crystallized in an open conformation with a large cleft between helices A′ and helices F′–G′. As discussed in the previous section, these helices are likely to bind to the membrane, and the hydrophobic substrates cholesterol and vitamin D 3, respectively, could enter each enzyme from the membrane. Most P450 substrates exhibit partition coefficients that favor the hydrophobic environment of the bilayer over the aqueous phase, which suggests that the concentration of substrate in the membrane may be higher than in solution under physiological conditions.

Fig. 1.12
figure 12

Open (PDB: 3UA5) and closed (PDB: 3IBD) conformations of human CYP2B6. The open structure has two molecules of amlodipine ( spheres) with one molecule of amlodipine bound to the heme iron via nitrogen coordination and the second amlodipine in the open-substrate access channel. The closed structure has one molecule of 4-(4-chlorophenyl)imidazole ( spheres) coordinated to the heme iron with a closed substrate entrance channel. The heme is rendered as a stick figure with the iron shown as a sphere. Nitrogen and oxygen atoms are colored light gray and black, respectively

Full size image
Fig. 1.13
figure 13

Open (PDB: 3KNV) and closed (PDB: 3NAO) conformations of mitochondrial rat CYP24A1 and bovine CYP11A1 , respectively. The heme is rendered as a stick figure with the iron shown as a sphere. Nitrogen and oxygen atoms are colored light gray and black, respectively

Full size image

Structures obtained with bound ligands are often closed, and substrate access channels remain closed during MD simulations that are of short duration compared to substrate dissociation rates. Nevertheless, a number of solvent channel s have been identified in X-ray crystal structures and during MD simulations in an aqueous medium for soluble and truncated membrane P450s [93]. As shown in Fig. 1.10, several of these channels are oriented into the lipid portion of the bilayer in MD simulations. Comparisons of the duration and extent of opening during MD simulations, for the catalytic domain s in an aqueous environment and bound to membranes, are qualitatively similar and reveal differences in the frequency and duration of channel opening that reflect interactions between the catalytic domain and the membrane [57–60]. These solvent channels are thought to open and coalesce to form substrate access channels as seen for open conformations of soluble and membrane P450s determined by X-ray crystallography [93].

6 Substrate Complexes: Specific P450s

A fascinating structural feature of P450s is the ability to adapt to substrates of various sizes and shapes, yet retain the overall P450-fold and P450 electron transfer and O2 activation chemistries. Most of our detailed understanding of protein–substrate interactions derives from highly specific P450s that bind their respective substrates tightly and thus generate crystals that diffract well. Several substrates for various specific P450s are shown in Fig. 1.14. The size and shape of the various substrates shown in Fig. 1.14 are sufficiently diverse that the structural basis for what controls substrate specificity can, at least in part, be understood. As expected, all substrates are situated such that the atom to be hydroxylated is within 4–5 Å of the heme iron. Thus, regio- and stereoselective hydroxylation by the Fe(IV)-O species is achieved by specific protein–substrate interactions that hold the substrate in the correct position. The exception is P450BM3. The structure of the P450BM3 heme domain with palmitoleic acid [78] and N-palmitoylglycine [79] show that the fatty acid substrate is ≈ 7–8 Å from the iron which is too far for hydroxylation. However, NMR results indicate that the substrate moves to be within 3 Å of the iron upon reduction from Fe(III) to Fe(II) [94]. Precisely how reduction is linked to such a large repositioning of the substrate remains unknown.

Fig. 1.14
figure 14

Substrates bound to the active site of various P450s

Full size image

P450cam and P450epoK [95] represent the two extremes of substrate size and shape. Hence, a comparison between these two structures provides some insights on which regions of the structure change most in response to the requirements of substrate specificity . The two regions that differ the most between P450epoK and P450cam are the F, G, B′ helices, and the F/G loop (Fig. 1.15). The B′ helix is rotated 90° in P450epoK compared to P450cam. This reorientation opens the substrate-binding pocket, thus making room for the thiazole ring of the substrate. The F and G helices do not superimpose well, and the F/G loop adopts a substantially different conformation . There also are examples where a second substrate molecule is trapped in the access channel possibly because crystallization favors a partially open active site , thus leaving room for an additional molecule. Anecdotal observations not usually published show that E. coli “mystery” molecules will sometimes bind in the access channel or active site. This likely reflects the general hydrophobic nature of P450 active sites and the open/close dynamics that might make it possible for even specific P450s to bind different molecules present in the growth media.

Fig. 1.15
figure 15

A comparison of the P450cam and P450epoK (PDB: 1PKF) active site s. The very different size and shape of the substrates illustrate how the active site substantially differs from one P450 to the next

Full size image

An unusual example of a P450–substrate interaction is CYP107H1 (P450BioI) . P450s participate in polyketide biosynthesis, and these pathways involve multiple enzymatic steps that process a growing fatty acid-like chain into the array of complex and well-known antibiotics and other natural products. In many of these systems, an acyl carrier protein (ACP) forms a covalent bond with the substrate and transfers the substrate from one enzyme to the next. Where hydroxylation reactions are required, P450s often are involved, which means that in some of these systems the substrate is delivered to the P450 by the carrier protein. One well-characterized system is from the biotin biosynthetic pathway in B. subtilis [96]. P450BioI catalyzes the formation of pimelic acid through the oxidative cleavage of a fatty acid carbon–carbon bond, which then proceeds on to biotin [97, 98]. There is now a crystal structure of such fatty acid acylated ACP protein complexed with the P450 (Fig. 1.16) [99]. Structurally, P450BioI is a typical P450, yet here the substrate entry pocket has been adapted to bind ACP. Note that the substrate enters the active site near the connection between the F and G helices that is the main entry point for substrates in many P450s.

Fig. 1.16
figure 16

The crystal structure of P450BioI (PDB: 3EJB) [99]. ACP ( darker molecule) binds such that the fatty acid substrate attached to ACP extends into the active site of the P450. The opening near the F/G loop region that enables substrate entry is the same as observed in many other P450s. ACP acyl carrier protein

Full size image

There is one final example of P450 substrate adaptability, but in this case there may be two different active site s and two enzyme activities. CYP170A1 from Streptomyces coelicolor catalyzes the oxidation of epi-isozizaene to an epimeric mix of 5-albaflavenol (Fig. 1.17). The structure shows that there are two substrate molecules bound, one in the expected location just above the heme and a second in the substrate access channel [100]. What was most unexpected is the finding that the conversion of farnesyl diphosphate to epi-isozizaene is catalyzed by CYP170A1. Sequence comparisons between known sesquiterpene synthase enzymes pointed toward a particular region of CYP170A1 that might be involved (arrow in Fig. 1.17). Subsequent mutagenesis in this region eliminated the synthase activity but not the P450 activity [100]. Given that we are accustomed to viewing enzymes as requiring a relatively large size to properly form the active site, it might at first seem odd that such a small region of a P450, or any enzyme, could serve a catalytic function. However, sesquiterpene synthase enzymes appear not to operate by typical acid–base catalysis requiring suitably positioned active site groups to move protons [101]. Instead, it appears that metal ions and the substrate diphosphate are the keys to catalysis and that the enzyme may serve a more passive role, providing a template for substrate and metal ion binding.

Fig. 1.17
figure 17

The CYP170A1 (PDB: 3DBG) crystal structure [100] and reaction. Substrate 1 binds near the heme as expected, while substrate molecule 2 binds in the open access channel. The site thought to be responsible for the sesquiterpene cyclase activity is indicated by the arrow

Full size image

7 Active Site Diversity of Mammalian P450s

As with prokaryotic P450s, active site diversity underlies the unique roles of P450s in mammalian physiology. Structures now are available for several of the enzymes that hydroxylate the aliphatic side chains of cholesterol and vitamin D 3. P450 11A1 catalyzes three successive oxygenation reactions to produce sequentially 22R-hydroxycholesterol , 22R,20R-dihydroxycholesterol , and an unstable product that undergoes carbon–carbon bond scission to produce the 21-carbon steroid, pregnenolone , and isocaproaldehyde. It is thought that the peroxyanion intermediate that precedes formation of the oxene is the reactive intermediate for the third reaction [102]. The crystal structure of human mitochondrial CYP11A1 with cholesterol bound [70] indicates that the tetracyclic sterol moiety is bound in the entrance channel to the substrate-binding cavity under the helix F–F′ region and above β-sheet 1 with C22 of the aliphatic side chain positioned closest to the heme iron, Fig. 1.18a. Additional structures of P450 11A1 [69, 70] with the first and second products of the reaction, 22R-hydroxycholesterol and 22R,20R-dihydroxycholesterol, indicate that the tetracyclic sterol moiety is positioned similarly to that of cholesterol in each case, with changes in the dihedral angles of the side chain positioning the appropriate site of metabolism close to the heme iron. The structure of CYP46A1 [103] indicates that cholesterol sulfate binds similarly, but the aliphatic side chain is positioned with C24 closest to the heme iron, Fig. 1.18b. In contrast, an X-ray crystal structure of P450 2R1 [104] reveals that the sterol moiety of vitamin D3 is located under helix G near helices I and B′ with the site of metabolism, C25, positioned closest to the heme iron, Fig. 1.18c.

Fig. 1.18
figure 18

Substrate and inhibitor binding to human P450s that catalyze key steps in steroid metabolism . The substrates, inhibitor, and heme are shown as stick figures with the heme iron depicted as a sphere. Nitrogen and oxygen atoms are colored light gray and black, respectively. The dotted lines represent the distance from the heme iron for sites of metabolism labeled with the identity of the site of metabolism and the distance. The CYP17A1 inhibitor abiraterone binds directly to the heme iron. For reference, a portion of helix I and the helix B–C loop are shown. The topology and length of the helix B–C loops exhibit significant variation between proteins

Full size image

Other steroid biosynthetic enzymes catalyze reactions that modify the rigid tetracyclic steroid ring system. Three enzymes, P450s 7A1, 7B1, and 39A1, insert an oxygen atom into the 7α C–H bond to produce 7α-hydroxylated intermediates in the formation of bile acids . A structure of P450 7A1 with the cholesterol analog, cholest-4-en-3-one (PDB code 3SN5), indicates that the 7α C–H bond is positioned closest to the heme iron and that the plane of the sterol ring is parallel to the plane of the heme, Fig. 1.18d. The aliphatic side–chain passes out of the substrate-binding cavity between helix I and the helix B′–C loop. Structures of P450s 19A1 [105] and 11B1 [71] also place the tetracyclic steroid ring system of androst-4-ene-3,20-dione and 21-hydroxyprogesterone in a similar location, but with the C19 methyl group, Fig. 1.18e, and the 11β C–H bond, respectively, oriented toward the heme iron. P450 19A1 catalyzes three successive oxidations of the 19-methyl group with the product rearranging to produce formic acid and the unsaturated A ring of the estrogen , estrone.

Other reactions catalyzed by steroid biosynthetic enzymes target the ends of the steroid ring system. The structure of the adrenal 21-hydroxylase [106] with 17α-hydroxyprogesterone bound reveals that the tetracyclic steroid is oriented almost perpendicular to the plane of the heme with the 17β-side chain positioned near the heme iron. P450 17A1, which catalyzes the 17α-hydroxylation of progesterone and cleavage of the 17β-side chain of the pregnenolone to form androstenedione, has been crystallized with abiraterone [107] in the active site , Fig. 1.18f. Abiraterone is used clinically for the treatment of prostate cancer via inhibition of androgen formation catalyzed by P450 17A1. The steroid moiety of abiraterone is oriented similarly to 17α-hydroxyprogesterone in the P450 21A2 structure, with abiraterone coordinated to the heme iron through a heterocyclic nitrogen group. Structures of human CYP51A1 with inhibitors bound in the active site are also available to aid in the development of CYP51A1 inhibitors that will target these enzymes in pathogens without inhibiting the human enzyme [108, 109].

As is evident in Fig. 1.18, P450’s have evolved to catalyze these reactions by positioning the substrates for site-selective metabolism, and, in doing so, different portions of the P450 structure are utilized for substrate binding. This, in turn, reflects differences in the sizes and properties of the amino acids that occupy the active site cavity as well as changes in protein conformation . Examples of these conformational differences are readily apparent when comparing the helix B–C loop regions depicted in the six panels of Fig. 1.18.

In contrast, P450s in families 1A, 2A, 2B, 2C, 2D, 2E, 2J, and 3A frequently contribute to the metabolic clearance of drugs and other xenobiotics. In the absence of evolutionary selection to optimize the binding of these compounds, many xenobiotic substrates are likely to exhibit relatively poor fits in P450 active site s and several isoenergetic binding poses may be possible, as suggested by the formation of multiple metabolites. Reaction rates are likely to reflect probabilities for binding to specific enzymes, relative reactivity of potential sites of reaction, and probabilities for placement of the sites of reaction near the oxene intermediate, leading to uncoupling , multiple metabolites and poor catalytic efficiencies. Fortunately, the enzymes that catalyze these reactions exhibit significant active site diversity that provides protection from a wide range of structurally diverse xenobiotics.

Family 1 and 3 enzymes exhibit very different active-site cavities. The enzymes in family 1 typically metabolize polynuclear aromatic hydrocarbons, and the structures of human CYP1A1 [110], 1A2 [64], and 1B1 [111] exhibit narrow active-site cavities that complement the size and planarity of polynuclear aromatic hydrocarbons, as illustrated for CYP1A2 in Fig. 1.19a. These narrow active-site cavities are reinforced by a kink in helix F, which directs a portion of helix F and helix F′ between the active site and the N-terminal domain. This is likely to add rigidity to the narrow active site cavity. In contrast, CYP3A4 exhibits a large and open active site cavity (Fig. 1.19b) with a much larger exposure of the heme surface to substrates than seen in other xenobiotic metabolizing enzymes [62, 112]. This difference underlies the capacity of 3A4 to catalyze oxygenation of the steroids at carbons 6 or 7 in the center of the ring system as seen for steroidogenic CYP7A1 in Fig. 1.18. CYP3A4 is also unusual because helix F is short and does not cross above the active site. As a result, the active site can expand and contract by the flexible motion of the long connector between helix F and F′ and changes in the positions of helices F′ and G′ [113]. The active-site cavities of human family 2 P450s range from small for P450s 2E1 [114, 115], 2A6 [116] Fig. 1.19c, 2A13 [117], and 2B6 [89, 90] to large for 2C8 [63], Fig. 1.2C9 [118, 119], 2C19 [120], and 2D6 [75, 121], and they can vary due to conformational changes associated with ligand access and binding [45]. As such, these enzymes contribute diverse capacities for xenobiotic metabolism.

Fig. 1.19
figure 19

Portions of the structures of the complex of CYP1A2 with α-naphthoflavone (PDB:2HI4), CYP3A4 with ritonavir (PDB:3NXU), CYP2A6 with coumarin (PDB:1Z10), and CYP2C8 with montelukast (PDB:2NNI) are shown as cartoons displaying secondary structures. The heme and ligands are depicted as stick figures with the heme iron shown as a sphere. The surfaces of the active-site cavities were calculated using VOIDOO [157] and rendered as a transparent surface. Nitrogen and oxygen atoms are colored light gray and black, respectively

Full size image

8 Electron Transfer Complexes

P450s do not operate alone but must form a complex with a redox partner for electron transfer . Protein redox complexes, including those involving P450s, are designed not to be very tight or long-lived. A complex that is too tight will have a slow dissociation rate, which precludes rapid turnover. Nature thus must strike a balance between specificity , affinity, and high turnover. Such complexes have proven quite difficult to crystallize, which is why there are very few protein–protein redox complexes in the PDB and, to date, there are only three crystal structures of a P450 complexed with a redox partner.

The first structure of a redox complex to be solved was that between the heme and FMN domain s of P450BM3. Although P450BM3 is a bacterial enzyme, P450BM3 is more closely related in sequence, structure, activity, and redox partner to microsomal P450s than to other bacterial P450s. The unique feature of P450BM3 is that the diflavin P450 reductase is linked to the C-terminal end of the heme domain, thus giving a catalytically self-sufficient enzyme. Crystals were obtained by removing the FAD domain [122]. The structure (Fig. 1.20) shows that the FMN domain docks on the proximal surface of the P450, which was expected, based on complementary electrostatic surfaces and mutagenesis studies. The linker connecting the heme and FMN domains had been proteolyzed during crystallization, thus raising the possibility that the structure is an artifact of crystallization. Further experiments were carried out to test the functional validity of the model. Residues found at the interface were probed by mutagenesis [123]. Replacing Leu104 of P450BM3 with a Cys (Fig. 1.20) at the interface should not alter binding or electron transfer because replacing Leu with a smaller side chain should not cause any steric problems in forming the proper complex. However, covalent modification of the mutant Cys104 side chain with a large fluorophore should interfere with electron transfer. For these studies, laser flash photolysis was used wherein a laser flash photoreduced a potent reductant, deazariboflavin, which in turn reduces the FMN in the complex. The reduced FMN semiquinone then reduces the P450 heme. As predicted, mutation of Leu104 to Cys had no effect, while chemical modification of Cys104 dramatically decreased the FMN-to-heme electron transfer rate, thus implicating Leu104 as an important residue in forming the proper electron transfer complex.

Fig. 1.20
figure 20

Crystal structure of the P450BM3 electron-transfer complex (PDB: 1BVY) [122]. The closest contact at the interface is between Gln387 in the heme domain and the FMN

Full size image

A second prediction from the P450BM3 electron-transfer complex structure that can be tested is the electron transfer path. The heme–FMN domain interface is shown in Fig. 1.20. The closest point of contact between the two domains places the FMN about 4 Å from the peptide backbone of Gln387. The peptide chain from Gln387 to the heme ligand, Cys400, could constitute an electron transfer path. To test this hypothesis, Gln387 was converted to Cys and modified with (4-bromomethyl-4′-methylbipyridine)[bis(bipyridine)]ruthenium(II) [124]. The covalently attached Ru(II) is photoreduced, and the rate of reduction of the heme Fe(III) to Fe(II) by the photo-generated Ru(I) was followed. The same experiment was carried out with Ru(II) attached to Cys62. Both Cys62 and Cys387 are about the same distance from the heme, but electron transfer from Cys60-Ru(II) must make “through-space” jumps, while there is a continuous covalent connection between Cys386-Ru(II) and the heme ligand, Cys400. In the case of Cys387-Ru(II), the heme iron was reduced at a rate of 4.6 × 105 s− 1, while Cys60-Ru(II) did not reduce the heme iron. These results indicate that if the crystal structure of P450BM3 electron-transfer complex is functionally relevant, then the electron-transfer reaction can readily proceed along the direct point of contact between the FMN and heme domain.

The structure of the complex formed between adrenodoxin (Adx) and P45011A1, which converts cholesterol to pregnenolone , also has been solved. The crystal structure of the complex was solved by fusing adrenodoxin to the N-terminal end of CYP11A1 [70]. Although a good part of the Adx was disordered and not visible in electron-density maps, the interface with CYP11A1 was well defined (Fig. 1.21). The interface is dominated by electrostatic interactions and those residues involved are consistent with mutagenesis and chemical modifications studies [125–127]. A comparison between the free enzyme [69] and the enzyme complexed with Adx are essentially identical, so Adx binding does not result in any significant structural change.

Fig. 1.21
figure 21

Crystal structure of the complex formed between CYP11A1 and adrenodoxin (Adx; PDB: 3N9Y) [70]. Only part of the Adx is visible in electron-density maps. The interface is dominated by ionic interactions. Adx binding does not result in any major structural change in CYP11A1

Full size image

The most recent structure to be determined is the P450cam–Pdx complex crystal [128, 129] and NMR structures [128]. The P450cam–Pdx complex has received considerable attention especially since it was established some time ago that P450cam is not only very selective for Pdx but Pdx also plays an effector role by inducing structural changes required for electron transfer and O2 activation [130–132]. Prior to the recent crystal structure of the P450cam–Pdx complex, Pochapsky et al. developed a model of the P450cam–Pdx complex using NMR and molecular modeling [133] that is supported by mutagenesis data [134–138]. A wealth of spectral data shows that when Pdx binds on the proximal side of the heme, spectral changes ensue that are associated with the opposite distal substrate-binding pocket. These changes include resonance Raman [139], infrared [138, 140], and NMR [141–143]. NMR studies [144–146] showed that Pdx binding results in changes in the B′, C, F, and G helices that are well removed from where Pdx binds (Fig. 1.22). The B′ helix provides key contacts with the substrate, while large movements of the F and G helices are the main features of the open/close transition [82]. Pdx binding to oxy-P450cam decreases the stability of the oxy complex 150-fold [147], while oxidized Pdx shifts oxidized P450cam to the low-spin state [148]. All these observations point to significant structural changes in P450cam when Pdx binds.

Fig. 1.22
figure 22

Structure of the P450cam–Pdx complex (PDB: 4JX1) [129]. A key interaction is between PdxTrp106 and the C helix in P450cam. The C helix moves “up” in order to optimize interactions with PdxTrp106. This motion is coupled to an opening of the active site access channel on the opposite side of the protein (F/G helical region)

Full size image

The crystal structure of the P450–Pdx complex [128, 129] shows that P450cam adopts the open conformation , which is consistent with previous spectroscopic studies. The structure of the reduced form of the complex has four P450–Pdx molecules in the asymmetric unit and in three of these, the product, hydroxycamphor, is bound [129]. This means that the open form in the complex is active in O2 activation and hydroxylation. Interactions at the interface are consistent with earlier NMR studies [145] and mutagenesis data [43, 137, 138, 149–153]. PdxAsp38 interacts with P450camArg112 (Fig. 1.22), which requires little movement in either protein in the vicinity of the ion pair. However, interactions involving PdxTrp106, which has been known for some time to be a critical residue [131], require movement of the C helix (Fig. 1.22). In effect, the C helix moves “up” about 2–3 Å in order to form nonpolar and H-bonding interactions with PdxTrp106. This motion of the C helix is coupled to movements in the B′, I, F, and G helices, all of which are involved with substrate access or direct contacts with both substrates, camphor and O2. This motion results in a large movement of the F and G helices and the F/G loop, which effectively opens the active site to bulk solvent. This open conformation is the same as observed by Lee et al. [82] The main difference is that in the structure solved by Lee et al. [82] the B′ helix is disordered, while in a complex with Pdx the entire P450cam is highly ordered and the key interactions between camphor and the local environment remain, by and large, unchanged from the closed conformation. The main driving force for the conformational change appears to be PdxTrp106, which could not form tight interactions with P450cam without the structural switch.

The central question is why such Pdx -induced changes are important for activity. A possibly important part of the Pdx-induced structural change centers on the I helix near the O2-binding site. The switch in the I helix in going from the closed to open state results in opening of the I helix similar to what happens when O2 binds (Fig. 1.5). The closed to oxy-complex opening of the I helix is about midway between the extremes of the closed to fully open switch. This opening of the I helix is required to enable the catalytic waters to move into place for proton transfer to dioxygen [31, 33]. Thus, Pdx binding helps to stabilize the more oxy-like conformation of the I helix. However, the oxy-P450cam structure probably does not represent the final active state since Pdx binding perturbs the oxy-P450cam spectrum and results in a 150-fold destabilization of the oxy complex [147]. It has been argued that Pdx “pushes” the oxy complex more toward the active form that is probably the more open conformation. Pdx also alters the electronic properties of the thiolate ligand [147, 148], which could be due to a shortening of the peptide NH-thiolate H-bond observed in the P450cam–Pdx crystal structure.

Another large change that occurs when Pdx binds involves Asp251. Asp251 is part of the I helix and is usually Asp or Glu in many other P450s. Asp251 is essential for activity in P450cam [154], P450cin [155], and CYP101D1 [88]. The Asp251Asn mutant in P450cam exhibits a two-orders-of-magnitude decrease in activity, yet remains tightly coupled [154]. That is, nearly all the electrons funneled into the P450cam mutant are utilized for substrate hydroxylation and not the wasteful production of water or peroxide. This mutant also exhibits a kinetic solvent isotope effect of 10 compared to 1.8 for wild-type P450cam [32]. This strongly implicates Asp251 as being intimately involved with the proper delivery of protons to dioxygen required for heterolytic cleavage of the O–O bond . The problem with this view, however, is that Asp251 is tied up with Arg187 and Lys178 in two strong ion pairs (Fig. 1.23). However, in the Pdx complex, these ion pairs are broken, which frees Asp251 to serve its proposed role in shuttling protons from bulk solvent into the active site . It thus appears that an important part of Pdx binding may be to “arm” the proton delivery machinery required for proton-coupled electron transfer .

Fig. 1.23
figure 23

The region around Asp251 in P450cam with and without Pdx bound. In the Pdx-free closed state, Asp251 is tied up in strong ion pairs with Arg186 and Lys178. When Pdx binds these ionic interactions are broken, thereby releasing Asp251 for its role in shuttling solvent protons to the iron-linked O2 molecule required for O–O bond cleavage and thus, O2 activation

Full size image

The next obvious question is whether or not this sort of redox partner -mediated conformational change required for activity is a general property of all P450s or is limited to P450cam. The weight of the evidence so far indicates that P450cam may be an outlier. A number of P450s are known to be supported by nonphysiological redox partners and some redox partners, such as P450 reductase, service a large number of P450s. The only structural comparisons that can be made to address this question are the P450cam–Pdx and CYP11A1 –Adx complexes [69]. CYP11A1 does not change to the open form in the complex but remains closed [69, 70]. However, Asp290 (corresponds to Asp251 in P450cam) is not tied up in ion pairs and is exposed to bulk solvent. Hence, no structural changes are required to free Asp290 for catalysis, although it has yet to be established if Asp290 is essential for CYP11A1 catalysis. Given that Nature has so many P450s, it is doubtful that P450cam is the only P450 where selective redox partner binding coupled with conformational selection is required for activity. It should only be a matter of time before similar P450s are uncovered and analyzed in depth. Just as interesting a question is the biological basis for such control. What is the evolutionary advantage, if any, of P450cam exhibiting such specificity , while very closely related P450s do not?

9 Conclusion

The large increase in P450 crystal structures over the past few years is due in large part to technological advances in protein expression and purification. Just as important are the increasing genome databases which now makes it relatively easy to “discover” new P450s. The sophistication of user-friendly crystallization robots, software, and synchrotron data collection has opened up crystallography to the nonexpert which also has been a major contributor to the ever-expanding number of structures deposited in the protein database. In fact, the field is now at the stage where expression, purification, characterization, and crystal structure determination can outpace functional and biological studies. Many structures now are being solved before one knows much about function. We thus must start using structural information to guide functional and biological studies. This could be particularly important with orphan P450s that will continue to increase in number as more and more P450s are discovered in new and interesting places. Such advances coupled with powerful computational resources that can be used for molecular modeling and in silico screening of potential substrates can significantly contribute to a better understanding of function. Recent advances in defining various conformational states also is quite important since which conformational state one uses for virtual screening of substrate/inhibitors is obviously quite important. Now, however, we have a better idea on the various conformational states available to P450s which will further sharpen predictive computational tools. We thus anticipate that P450 structural biology will continue to move quickly but that much less time and energy will be devoted to the actual structure determination and instead, will be focused on function.