Diversity of P450 Enzymes in the Biosynthesis of Natural Products

Abstract

Diverse oxygenation patterns of natural products generated by secondary metabolic pathways in microorganisms and plants are largely achieved through the tailoring reactions catalysed by cytochrome P450 enzymes (P450s). P450s are a large family of oxidative hemoproteins found in all life forms from prokaryotes to humans. Understanding the reactivity and selectivity of these fascinating C-H bond-activating catalysts will advance their use in generating valuable pharmaceuticals and products for medicine, agriculture and industry. A major strength of this P450 group is its set of established enzyme-substrate relationships, the source of the most detailed knowledge on how P450 enzymes work. Engineering microbial-derived P450 enzymes to accommodate alternative substrates and add new functions continues to be an important near- and long-term practical goal driving the structural characterization of these molecules. Understanding the natural evolution of P450 structure-function should accelerate metabolic engineering and directed evolutionary approaches to enhance diversification of natural product structures and other biosynthetic applications.

Historical perspective and overview

The chapter of P450 history that began in the mid-1980’s with the determination of the x-ray structure of P450cam^1–4, the camphor-metabolizing monooxygenase from the bacterium Pseudomonas putida, became a structural and mechanistic paradigm for this entire protein family. But it was not until 1995 that the first natural product P450, EryF, from the erythromycin biosynthetic gene cluster of Saccharopolyspora erythraea, was structurally characterized^5–9. In 2002 a second example was reported, the phenolic ring-coupling P450 OxyB from the vancomycin producer Amycolatopsis orientalis¹⁰. Following this nascent period, new structurally defined biosynthetic P450 monooxygenases began accumulating rapidly after 2005, bringing fresh insights to the field of structural biology, as well as new potential tools and applications for synthetic biology. To date, over twenty biosynthetic P450s have been characterized, both structurally and functionally (Table 1), that encompass different classes of natural products, including macrolides (EryF, EryK, PikC and MycG), polyenes (PimD, CYP105P1 and CYP105D6), glycopeptides (OxyB, OxyC, OxyD and CYP165D3), aromatic polyketides (AurH, CalO2, CYP158A1 and CYP158A2), hybrid polyketide/nonribosomal peptide systems (EpoK), sesquiterpenes (CYP170A1), alkaloids (StaP) and fatty acids (P450_BioI, CYP74A1 and CYP74A2). Over a dozen biosynthetic P450s have been characterized as complexes with their natural substrates.

Table 1.

Structurally characterised natural product P450s

Year	P450	Biosynthetic pathway	Reaction type	Substrate co-crystals	References
1995	EryFa	Erythromycin	Hydroxylation	6-deoxyerythronolide B	7, 8
2002	OxyB	Vancomycin	Phenolic ring-coupling	-	10
2003	OxyC	Vancomycin	Aryl ring-coupling	-	19
2003	EpoK	Epothilone	Epoxidation	Epothilones B and D	20
2004	CYP154A1	Methylenomycin	Paternò–Büchi cyclization	-	21, 22
2005	CYP158A2	Flaviolin polymer	Aryl-ring coupling	Flaviolin, 2-hydroxynaphthoquinone	23, 24
2006	PikC	Methymycin/neomethymycin/pikromycin	Hydroxylation	YC-17, narbomycin, unnatural semi-synthetic molecules	25–27
2007	StaP	Staurosporine	Aryl ring-coupling/decarboxylation	Chromopyrrolic acid	28
2007	CYP158A1	Flaviolin polymer	Aryl-ring coupling	Flaviolin	29
2008	P450BioI	Biotin	C-C oxidative cleavage	ACP-bound fatty acids	30
2008	CYP74A1	Oxylipin	Allene oxide synthase	13(S)-HOTb, 13(S)-HODc, vernolic acid	14
2008	CYP74A2	Oxilipin	Allene oxide synthase	13(S)-HOD	15
2009	CalO2	Caliheamicin	Hydroxylation	-	31
2009	CYP105P1	Filipin	Hydroxylation	Filipin	32, 33
2009	EryK	Erythromycin	Hydroxylation	Erythromycin D	34
2009	CYP170A1	Albaflavenone	Allylic oxidation	epi-Isozizaene	18
2010	CYP105D6	Filipin	Hydroxylation	-	33
2010	OxyD	Vancomycin	Tyrosine β-hydroxylation	-	35
2010	PimD	Pimaricin	Epoxidation	4,5-desepoxypimaricin	36
2011	CYP165D3	Teicoplanin	Phenolic ring-coupling	-	37, 38
2011	AurH	Aureothin	Tetrahydrofuran ring formation	-	39
2012	MycG	Mycinamicin	Hydroxylation/epoxidation	Mycinamicins III, IV and V	40

^aP450s are referred to by the more commonly used names

^b13(S)-hydroxyoctadecatrienoic acid;

^c13(S)-hydroxyoctadecadienoic acid.

Amino acid sequence alignments show ≤43% pair-wise sequence identity between structurally defined P450s involved in natural product assembly and tailoring, which is consistent with their assignments to distinct families according to generally accepted sequence identity-based P450 nomenclature¹¹ (Fig. 1). Most natural product P450s structurally characterized to date originate from Actinomycetes, particularly Streptomyces spp. This is in part because the Streptomyces secondary metabolome is the richest source of biologically active natural products developed as pharmaceuticals, accounting for over two-thirds of microbially derived antibiotics^{12, 13}. Bacterial P450s are more amenable to functional characterization and x-ray structure analysis than their fungal and plant counterparts. Currently, only two plant P450s, both from CYP74 family of allene oxide synthases, have been structurally characterized^{14, 15}. EryF, the first and most highly-characterized natural product P450, shares the highest overall similarity with the rest of the bacterial enzymes. By contrast, CYP170A1 of Streptomyces coelicolor A3(2), is the most remotely related member of the bacterial group (Fig. 1). In the typical heme-dependent active site, CYP170A1 carries out two sequential allylic oxidations to convert epi-isozizaene (1) to an epimeric mixture of albaflavenols (2) and then to the tricyclic sesquiterpene antibiotic albaflavenone (3)^{16, 17} (Fig. 2). Uniquely, CYP170A1 possesses an alternative terpene synthase activity associated with the “moonlighting” active site formed by the four-helix bundle composed of the C, H, I and L α-helices in the 3-D structure of the enzyme¹⁸, thus generating farnesene (4) from farnesyl diphosphate (5).

Fig. 1. Phylogenic diversity of structurally defined bacterial natural product P450s.

Tree is based on the sequence alignments in Fig. 5, and built using PhyloWidget, a web-based visualization tool⁴¹

Fig. 2. Albaflavenone and farnesene biosynthetic pathway in S. coelicolor A3(2).

In secondary metabolic pathways, P450 genes are integrated within biosynthetic gene clusters. Probing of these metabolic systems by gene disruption and bioconversion studies has allowed the reaction steps catalyzed by each enzyme to be identified and characterized. Thus, P450 enzymatic catalysis achieves aliphatic and aromatic bond hydroxylation, double bond epoxidation, heterocyclization, aryl and phenolic ring-coupling, oxidative rearrangement of carbon skeleton and C-C bond cleavage in diverse natural-product settings, all with high chemo-, regio-, and stereoselectivity. Structural diversification of natural products serves in nature to maintain the competitive advantages of organisms against environmental and developmental challenges⁴². Programmed manipulation of genes encoding enzymes in biosynthetic pathways offers the promise of redesigning antibiotic structures to create molecules with new biological activities^{43, 44}. Toward that goal, P450 monooxygenases have already proven to be an effective tool for introducing targeted structural variability in genetically engineered chemical products⁴⁵.

A salient recent example of the importance of P450 enzymes in advancing drug discovery comes from characterization of pladienolide natural products from the actinomycete Streptomyces platensis Mer 11107⁴⁶ (Fig. 3). Among the suite of seven 12-membered ring macrolactones produced by the bacterium, pladienolide B and D were shown to be promising anticancer agents targeting the 140 kDa protein in the SF3b splicing factor⁴⁷. In particular, pladienolide D (6) was found to have high potency against tumor cells, leading to the development of a natural product semi-synthetic analog, E7107 (7), by Eisai Pharmaceuticals. This molecule quickly advanced as a drug candidate showing promise in the human lung cancer LC-6-JCK xenograft model, inducing complete tumor remission with a wide therapeutic window. Based on these characteristics, E7107 progressed to human phase I cancer clinical trials.

Fig. 3. Pladienolide natural products.

Conversion of pladienolide A to semi-synthetic analog E7107 by an engineered strain of S. platensis bearing a P450 gene from S. bungoensis.

A key factor in developing E7107 was identifying a P450 enzyme that maximized the production of 6 in S. platensis. Although the most active of the series, this compound was found only in minor amounts in the wild-type microorganism, and for industrial purposes improved efficiency was required at the hydroxylation step at C16. After screening a series of actinomycete strains, Eisai found that Streptomyces bungoensis A-1544 specifically converted pladienolide A (8) to 6. The corresponding P450 gene was identified and engineered into S. platensis to achieve high-level production of the desired metabolite. A short series of synthetic chemical steps was employed to generate the urethane derivative E7107 for further development^{48, 49}.

P450 scaffold, conserved motifs and molecular mechanics

The topologies and molecular mechanics associated with the protein scaffold are common to the P450 protein family. The P450 structural core is composed of a four-helix bundle, D, E, I, and L (enclosed in red box in Fig. 4A), that carries the trigonal prism-shaped structure of the protein molecule. Substrate entry relies on the flexibility of the BC- and FG-loop and swinging of the F- and G-helices that transiently exposes the active site so that substrate can enter and product can exit^50–53. Spatial arrangement of the secondary structure elements subdivides the P450 scaffold in the α- and β-rich domains. The only absolutely invariant residue across the P450 family is cysteine, the heme iron proximal ligand situated at the N-terminus of the L-helix that coordinates to the iron via a thiolate bond. In bacterial natural product P450s, the G³⁴⁷XXXC³⁵¹ motif in the heme-binding loop harbouring cysteine is invariant (residue numbering here adheres to the EryF sequence).

Fig. 4. PikC structure in open and closed conformations.

A, P450 protein scaffold is built around a four-helix bundle core (enclosed in red box). Transition between open and closed conformations is approximated linearly by an ensemble of conformers generated computationally and superimposed with the experimentally obtained open and closed PikC structures. Molecule is viewed along the L-helix harbouring the invariant cysteine residue at the N-terminus facing the viewer. Ribbon representation of the open (B) and closed (C) conformations as observed in ligand-free PikC. The mobile F- and G-helices are highlighted in pink, I-helix in green, β-sheets in yellow; heme is shown as green sticks. Molecules in B and C are rotated ~90° toward the viewer along the horizontal axis in the plane of the figure as compared to A.

A prominent feature of the P450 scaffold is the long I-helix (green in Fig.4 B, C) running over the distal surface of the heme. In its central portion, the I-helix harbours the highly conserved A/Gⁿ⁻¹-Gⁿ-XX-Tⁿ⁺³ motif (Fig. 5A), which is present in virtually all P450 enzymes, although examples of non-conserved amino acid substitutions for each residue in the motif are known. Functionally, the role of Tⁿ⁺³ is critical in the H-bond network that enables protonation of the distal atom in the iron-bound oxygen molecule to promote heterolytic scission of the O-O bond^{55, 56}. Gⁿ is associated with bending of the I-helix, resulting in disruption of the I-helix H-bonding pattern, as observed in majority of structurally defined P450s. In the primary sequence, Gⁿ is preceded by either an alanine or glycine - A/Gⁿ⁻¹ - that H-bonds to an axial water ligand in the ferric hexa-coordinated low-spin P450 via a main chain carbonyl oxygen atom (Fig. 6). Release of the axial water associated with A/Gⁿ⁻¹ via an H-bond is a prerequisite for P450 catalysis. The invariant association of these two functions in the P450 structure suggests that the binding and release of the axial water may work in unison with the I-helix’s bending. A small size of the n-1 residue is dictated by proximity of the heme cofactor. The difference between the alanine or glycine modulates binding of axial water as alanine is more hydrophobic and thus “water repulsive”⁵⁷.

Fig. 5. Sequence alignments.

A, Sequence alignments encompassing the I-helix of natural product P450s. Alignment was performed using NCBI’s COBALT-constraint-based multiple alignment online tool⁵⁴. UniProt database (http://www.uniprot.org/) accession numbers are provided for each protein sequence. Residue numbering in (A) corresponds to EryF. B, Cartoon illustrating putative coupling mechanics of the substrate entry and release of the iron axial water ligand.

Fig. 6. P450 oxygen scission site.

Hydrophobicity of the active site in CYP130 increases as a result of the Gⁿ⁻¹ to alanine mutation (A) compared to the wild type enzyme⁵⁷ (B). Water molecules bound in the active site are shown in red spheres, heme is in yellow sticks and protein is in blue sticks. Interactions with an axial water ligand are indicated by dashed lines. G243A substitution in the active site also facilitates expulsion of axial water ligand by incoming inhibitors⁵⁷.

To visualize this coupling, it is useful to imagine the I-helix as a lever pivoting on the conserved Gⁿ with the force applied at the I-helix N-terminus (Fig. 5B). In this model, a large torque could be generated next to the fulcrum, at A/Gⁿ⁻¹, by a small force applied at the site of entry (FG-loop) by the incoming substrate and transformed to the swinging of the F- and G-helices and further transmitted through the system of flexible loops, to the I-helix N-terminus (Fig. 4A). The hypothetical coupling of substrate binding and axial water release may have evolved as an anti-oxidative stress measure in the ancestors of contemporary P450s which may have employed hydrogen peroxide and other peroxy compounds as oxygen donors to catalyse oxygenation of substrates when the terrestrial atmosphere contained little or no oxygen⁵⁸. This protective measure would prevent spontaneous heme reduction and binding of molecular oxygen to avoid the unproductive generation of active oxidants in the absence of substrate.

Consistent with the above hypothesis, P450s in the CYP152 family or plant allene oxide synthases that function as peroxygenases, lack A/Gⁿ⁻¹-Gⁿ-XX-Tⁿ⁺³ motifs. For instance, in Bacillus subtilis peroxygenase P450_BSβ (CYP152A1), which uses hydrogen peroxide as a source of oxygen to carry out α- and β-hydroxylation of the long-chain fatty acid substrates, all three conserved residues, A/Gⁿ⁻¹, Gⁿ and Tⁿ⁺³, are replaced with R²⁴², P²⁴³ and A²⁴⁶, respectively⁵⁹. Functionally, R²⁴² provides a salt-bridge interaction to the carboxylic group to stabilize a long-chain fatty acid substrate in the active site^{59, 60}. Similarly, plant allene oxide synthases (CYP74A), which utilize the peroxide shunt pathway to catalyse oxidative rearrangement of polyunsaturated fatty acid hydroperoxide substrates, also lack the A/Gⁿ⁻¹-Gⁿ-XX-Tⁿ⁺³ motif. Furthermore, the asparagine 321 side chain, a structural equivalent of the n-1 residue in allene oxide synthases, interferes with binding of a diatomic oxygen molecule at the heme iron^{14, 15}. Other examples of alternative amino acid residues at the n-1 position in natural product P450s include T²³³ in CalO2 and P²⁷⁴ in CYP170A1, although their functional roles have yet to be established.

P450 catalytic cycle

The iron of the heme prosthetic group, which is linked to the invariant cysteine via an axial thiolate bond, is central to activation of molecular oxygen. The generally accepted mechanism of P450-catalyzed substrate hydroxylation includes six consensus steps summarized in Figure 7: (i) entry of the substrate to the active site, which displaces solvent, including the iron axial water ligand; (ii) a one-electron reduction of the ferric heme iron followed by binding of molecular oxygen to form the ferrous dioxy (Fe²⁺-O₂) complex; (iii) transfer of a second electron and a proton to yield a ferric hydroperoxy (Fe³⁺-OOH) complex, also known as Compound 0; (iv) a second protonation and heterolytic cleavage of the O-O bond with concurrent production of a water molecule to form a reactive ferryl-oxo intermediate (Fe⁴⁺=O, porphyrin π-cation radical), known as Compound I⁶¹; (v) the abstraction of a hydrogen atom from the substrate followed by radical recombination to give the hydroxylated product; and (vi) dissociation of the product from the active site, with return of the enzyme to the initial ferric state⁶². Binding of hydrogen peroxide to ferric heme allows the first three catalytic steps to be bypassed, and is known as the peroxide shunt pathway.

Fig. 7. Conventional P450 catalytic cycle.

RH represents the substrate and ROH the resulting monooxygenated product. RO and ROOH depict singe oxygen atom donors, oxotransfer agents and organic peroxides, respectively. Shunt and uncoupling pathways are indicated with dashed-line arrows.

In consideration of the steps described above, the P450 reaction cycle requires an input of two electrons normally derived from NAD(P)H and delivered to the heme iron by accessory redox proteins^63–65, and two protons acquired from water. From structural studies on P450cam, the threonine residue Tⁿ⁺³ mediates interactions with the “catalytic” water molecule that serves as a direct proton donor to the iron-bound O₂^{55, 56}. Consistent with structural studies, most mutations of Tⁿ⁺³ to natural amino acids result in uncoupling of proton delivery steps and non-productive dissociation of hydrogen peroxide with consequent loss of catalytic activity^66–68 (Fig. 7). Efficient catalytic turnover of P450cam is supported by a threonine to asparagine mutation, suggesting that asparagine is capable of taking over its function⁶⁹. In a notable number of natural product P450s (see list in Fig. 5A), the n+3 position is also substituted with alanine (EryF, CYP158A1 and CYP158A2), serine (MycG and PimD), asparagine (OxyB) or glutamine (CYP165D3), indicating the possibility of alternative proton delivery routes. Prominent examples of the absence of threonine functionality have been reported in plant allene oxide synthases^{14, 15} and peroxygenases of CYP152 family, including Bacillus subtilis P450_BSβ^{59, 60} and homologs from other organisms^{70, 71}, which rearrange fatty acid substrates without involvement of molecular oxygen or NADPH-derived reducing equivalents.

Paternò–Büchi cyclization

Another example of P450-catalyzed rearrangement without involvement of external reducing equivalents is CYP154A1 of S. coelicolor A(3)2 that performs reaction of intramolecular cyclization to a Paternò–Büchi-like product⁷² without net oxidation-reduction. CYP154A1 known for flipped heme orientation observed in the x-ray structure²¹, converts dipentaenone (9) via reaction between the C5 carbonyl group and the C11–C12 double bond to the four-membered oxetane ring-containing product (10)²². No NADPH/NADP or redox partners are required for this rearrangement. It is not understood if the unique heme orientation is associated with this unusual in the field of P450 enzymology reaction. Although the biosynthetic origin of 9 is unknown, the subunit pentaenone ring (on the right) resembles the scaffold of methylenomycin C, an antibiotic produced by S. coelicolor A(3)2⁷³. In culture, CYP154A1 catalytic activity is essential for spore longevity and germ resistance to osmotic pressure²².

Substrate-assisted hydroxylation

While threonine mediates delivery of catalytically necessary protons in most P450 enzymes, an important mechanistic diversification that allows the Tⁿ⁺³ requirement to be bypassed involves substrate-assisted proton delivery. This was initially demonstrated for EryF, which performs the first hydroxylation reaction at C-6 of the 14-membered ring macrolactone 6-deoxyerythronolide B (11) in the erythromycin (12) producer Saccharopolyspora erythraea. The catalytically important peculiarity of EryF lies in its substitution of conserved threonine for alanine, A²⁴⁵. Based on x-ray structure analysis, the mechanism of substrate-assisted catalysis in EryF hinges on the substrate’s C5-OH group, which supplies a proton directly to the iron-bound molecular oxygen^{8, 9} (Fig. 8A). Accordingly, the substrate analog 5,6-dideoxy-5-oxoerythronolide, which lacks the OH group at C5 is not hydroxylated by EryF^{7, 74}. Additionally, while replacement of A²⁴⁵ by serine or threonine strongly decreases the rate of 6-deoxyerythronolide B hydroxylation⁷, it does enable low level oxidation of the alternative substrate testosterone⁷⁵. An example of indirect substrate-mediated proton delivery is also found in CYP158A1 and CYP158A2, involved in dimerization of flaviolin (15), both of which have an alanine at the n+3 position. Flaviolin C-5 and C-7 OH-groups are favourably positioned in the active site to serve as H-bond donors to the catalytically essential water^{24, 29}.

Fig. 8. Substrate assisted catalysis in EryF (A) and PimD (B).

A, 6-deoxyerythronolide B (yellow sticks) in the catalytic site of EryF. Oxygen molecule bound to the heme iron is shown as a red stick. B, 4,5-desepoxypimaricin (yellow sticks) in the catalytic site of PimD. A fragment of the electron density map (blue mesh) indicates inward rotation of the S²³⁸ side chain toward the I-helix groove. Fe axial water ligand is shown as red sphere. H-bonding interactions are indicated by dashed lines with distances in angstroms. Heme is in orange spheres. Fragments of the I-helix are shown as a grey ribbon or green sticks.

Substrate-assisted epoxidation

In contrast to EryF, the n+3 position in PimD, an epoxidase in the pimaricin producer Streptomyces natalensis, is occupied by serine. Although generally thought to be a conserved threonine substitution, S²³⁸ in PimD mimics alanine rather than threonine. Structurally, the serine side chain is rotated inward toward the I-helix groove where it H-bonds to the carbonyl oxygen of A²³⁴, thereby excluding itself from the proton delivery network (Fig. 8B)³⁶. As in EryF, the C7-OH group of the PimD substrate 4,5-desepoxypimaricin serves as the direct proton donor in the epoxidation reaction that leads to the polyene macrolide antibiotic pimaricin (13), which is characterized by a 26-membered macrolactone ring incorporating a chromophore formed by four conjugated double bonds and a six-membered hemi-ketal ring with an exocyclic carboxyl group. From the structural similarity of substrates and the conservation of S²³⁸, the PimD-related P450s AmphL and NysL from the amphotericin and nystatin biosynthetic pathways in Streptomyces nodosus and Streptomyces noursei, respectively, are predicted to employ substrate-assisted proton delivery, albeit for hydroxylation of their corresponding substrates³⁶. Despite the similarity in the proton delivery routes, an important functional difference between EryF and PimD is that the latter catalyses epoxidation of the double bond rather than hydroxylation of the C-H bond. A more complete understanding of this differential reactivity is a principal focus of interest, whose resolution will enable application of rational approaches to generate predetermined structural targets.

Distinct epoxidation mechanisms in EpoK and PimD

Together with the epothilone epoxidase, EpoK from the myxobacterium Sorangium cellulosum, PimD is one of two cytochromes P450 with native epoxidase activity that have been structurally characterized. According to the generally accepted mechanism, P450-mediated epoxidation should proceed via direct interaction between Compound I and the substrate olefin π-bond as the initial intermediate^76–81. In support of this route is the orientation of the substrate in EpoK (Fig. 9A), involved in epoxidation of epothilones C and D (to epoithone A and B, respectively (14)), which are 16-membered macrolactones with a pendant thiazole moiety²⁰. The orientation of the C12–C13 π-orbitals of the epothilone substrate is orthogonal to the heme plane. The ~5 Å offset relative to the Fe center suggests initial attack by Compound I at the C-13 atom. In contrast to EpoK, the orientation of the double bond in 4,5-desepoxypimaricin is inconsistent with epoxidation by Compound I and points instead to Compound 0 (Fig. 9B). The limits imposed by protein-substrate interactions and the strong intrinsic constraints on the molecular topology of pimaricin, which result in a single conformer for the macrolactone ring⁸² virtually eliminate the possibility of double bond reorientation. Catalytic evidence supports the view that the oxygen atom of Compound 0 distal to the heme iron inserts into the double bond to generate the epoxide ring (Fig. 9C). While hydrogen peroxide efficiently supports epoxidation, single oxygen atom donors such as peracetic acid and iodosobenzene have been shown to fail to epoxidize 4,5-desepoxypimaricin³⁶. Steric and stereoelectronic factors and an impaired proton delivery network may prevent Compound I from acting on the bulky rigid substrate in PimD. Accordingly, the substrate-assisted supply of protons in the transition state pushes the reaction into what is considered to be an energetically less favorable path^83–85.

Fig. 9. Distinct epoxidation mechanisms in EpoK and PimD.

Epoxidation product epothilone B (yellow) is shown bound in the EpoK active site (A) and a fragment of substrate desepoxypimaricin (yellow) in the PimD active site (B). Blue arrow points are collinear with the π-orbitals in the double bonds C12–C13 in epothilone and C4–C5 in desepoxypimaricin. Fragment of the I-helix is shown as a gray ribbon. Distances are in angstroms. C, PimD epoxidation reaction scheme. Substrate atoms are outlined in grey.

Molecular mechanism of ring-coupling

One of the hallmark reactions that exemplifies natural product P450 versatility is their ability to catalyze C-C bond formation between two aromatic ring systems. Two aryl ring-coupling P450s from Streptomyces coelicolor A(3)2, CYP158A1 and CYP158A2, both catalysing in vitro polymerization of flaviolin (15) but with different product profiles, were co-crystallized each with two flaviolin subunits in the active site. No common mechanistic pattern emerged from these studies as the flaviolin molecules are differently positioned in the active site^{24, 29}. Based on the crystal structure of CYP158A2, Zhao et al., proposed that the aryl-aryl coupling proceeds by direct hydrogen abstraction or direct bond formation with Compound I²³.

The detailed structure-based mechanism of aryl-aryl coupling was proposed for P450 StaP from Streptomyces sp. TR-A0274 that catalyses both intramolecular C-C bond formation and oxidative decarboxylation of chromopyrrolic acid (16) in staurosporine (17) biosynthesis resulting in formation of the indolocarbazole (18) skeleton²⁸ (Fig. 10). Compound I, formed in the catalytic cycle in the StaP-substrate complex, first removes one electron from the proximal indole ring to generate ferryl-oxo heme with the cation radical localized on the substrate indole ring, an equivalent to cytochrome c peroxidase Compound I⁸⁶. After relocation of the cation radical to the distal indole ring followed by ring deprotonation, the remaining ferryl-oxo heme oxidizing equivalent is spent to remove a second electron from the proximal indole. Intramolecular radical coupling then forms the C-C bond between the two indole rings. In the course of the reaction, the substrate’s twisted butterfly shape is transformed in the active site to the planar indolocarbazole scaffold (18). In the second turnover, StaP is believed to catalyze one–electron oxidation to decarboxylate the aryl-aryl coupled product; the indolocarbazole scaffold’s planarity may facilitate decarboxylation to avoid a steric clash with the carboxyl groups²⁸.

Fig. 10. Mechanism of aryl-aryl coupling of chromopyrrolic acid by StaP.

The x-ray structures of two other ring-coupling P450 monooxygenases, OxyB¹⁰ and OxyC¹⁹ in the vancomycin pathway have been determined in substrate-free form, as has CYP165D3^{37, 38}, which participates in the phenolic ring-coupling of the aromatic side chains in teicoplanin. These enzymes are responsible for the cross-linking of the aromatic side-chains of the heptapeptide precursors (19) but have limited activity toward free peptides, as they preferentially transform substrates bound to the non-ribosomal peptide synthetase (NRPS) peptidyl carrier protein⁸⁷. Due to the complex nature of the substrates, obtaining substrate-bound structures for these enzymes has proven challenging, and has yet to be accomplished. It is worthy of note that CYP165D3 lacks the conserved P450 motif A/Gⁿ⁻¹-Gⁿ-XX-Tⁿ⁺³. Here Gⁿ is mutated to alanine, while A/Gⁿ⁻¹ and Tⁿ⁺³, are mutated to methionine and glutamine, respectively. The presence of M²²⁶ at the n-1 position is particularly striking because its bulky side chain projects across the heme face to place the sulfur atom within 5.2 Å of the heme iron³⁷, potentially interfering with binding of the oxygen molecule.

Protein-assisted substrate delivery

Although binding of natural product P450s with their substrates has been achieved in a majority of reported crystal structures (Table 1), the accomplishment is a nontrivial pursuit. In some P450 enzymes, the assistance of acyl- or peptidyl carrier proteins in substrate delivery is required for binding to occur. A carrier protein is a four-helix bundle of ~80 amino acids that acts either in isolation or as domains of larger multi-domain assemblies, such as polyketide synthases (PKS) or NRPSs⁸⁸. In a recent mini-review⁸⁷, four classes of carrier protein-dependent P450 systems were described. They include (i) phenolic and aryl ring-coupling heptapeptides by OxyA, OxyB, OxyC and CYP165D3 in glycopeptide antibiotic biosynthesis; (ii) amino acid β-hydroxylating P450s, including OxyD, which produces β-hydroxytyrosine, an essential precursor for biosynthesis of the vancomycin-type aglycones; the related P450s NovI and NikQ catalyse tyrosine and histidine β-hydroxylation in novobiocin⁸⁹ and nikkomycin⁹⁰ antibiotic biosynthesis, respectively; (iii) the fatty acid oxidative cleavage enzyme P450_BioI in biotin biosynthesis, and possibly (iv) CalO2 in aromatic polyketide calicheamicin biosynthesis.

Currently, P450_BioI is the only P450 for which an acyl carrier protein complex has been structurally characterized³⁰ (Fig. 11A). Regiospecific long-chain fatty acid C-C bond cleavage is achieved in P450_BioI by the three-step in-chain oxidation reaction to yield seven-carbon pimelic acid (20), which constitutes the majority of the biotin carbon skeleton^{30, 91, 92}. The cleavage site is defined by the sharp U-shaped kink in the fatty acid chain between the C-7 and C-8 atoms over the heme iron, positioned there by interactions with the carrier protein and the prosthetic phosphopantetheine linker to which the long-chain fatty acid substrate is covalently attached (Fig. 11A). This spatial arrangement makes bond cleavage largely independent of the fatty acid length.

Fig. 11. Carrier protein-assisted (A) and chemical group-assisted (B) substrate delivery.

A, Complex between P450_BioI (cyan) and the acyl carrier protein (green). Fatty acid substrate (yellow sticks) assumes U-shape in the active site. For clarity, the P450 F-, G- and I-helices are highlighted in pink, blue and light green, respectively. B, Multiple binding modes of desosaminyl cycloalkane (pink and cyan) in the active site of PikC. PikC side chains are in green, heme in orange.

Substrate specificity of OxyD catalysing β-hydroxylation of tyrosine (21) also hinges on the specificity of the peptidyl carrier protein domain, as this P450 lacks specificity for loaded amino acid³⁵. Also, CalO2-catalysed step in calicheamicin biosynthesis (22) may require acyl carrier protein³¹. Selectivity for the delivering group coupled with a tolerance for loaded substrate makes these P450 systems of particular interest as bio-catalysts. For example, they might allow delivery of a range of molecules functionalized in specific way for selective oxidation in the P450 active site.

Chemical group-assisted substrate delivery

A major hurdle to broadening P450 applications in synthetic chemistry is narrow substrate specificity. This barrier might be partially overcome by utilizing a small molecule “anchor” binding specifically within the P450 to mediate delivery of unnatural substrates to the active site. This concept was first explored for P450 PikC, which has innate substrate flexibility toward macrolide systems²⁷. To deliver chemically synthesized molecules to the PikC active site, the desosamine sugar moiety of the natural macrolide substrates was linked via an acetal linkage to a series of carbocyclic rings ranging from 12 (23) to 15 carbons to generate the corresponding carbolides. The specificity of desosamine largely relies on the interactions of its protonated tertiary amine group with the triad of carboxyl-containing amino acid residues. As designed, desosamine indeed directed the unnatural molecules to the PikC active site for selective C-H bond hydroxylation. Attachment of desosamine to generate the carbolides led to productive enzyme binding and considerable regioselectivity in the hydroxylation reactions with impressive yields. However, the level of regio- and stereoselectivity observed for natural macrolide substrates was reduced due to alternative binding, as confirmed by x-ray structure analysis of these enzyme-substrate complexes²⁷ (Fig. 11B). Only when a rigid pyrene derivative consisting of four fused benzene rings was used as a substrate was a single oxidized product observed, albeit in low yield²⁷.

A similar approach was explored more recently to achieve controlled hydroxylation in the major drug-metabolizing human liver P450, CYP3A4, notorious for its substrate promiscuity⁹³. The analogs of the CYP3A4 substrate lisofylline (24) carrying a series of linear alkane and alkene substituents ranging from 3–6 carbon atoms linked to an alkaloid theobromine (25), were all hydroxylated or epoxidized at the fourth carbon position relative to theobromine. These results suggest an anchoring function for this achiral chromophore, although insights from co-crystal structures have not been reported. Enzymatic reactions occurred with predicted chemo-, regio-, and stereoselectivity and with high yield. Enhanced regio- and stereoselectivity achieved in this system compared to PikC is apparently due to the simplicity of the loaded molecules, and limited number of C-H bonds. As an added benefit, theobromine is easily cleaved in a post-enzymatic reaction to restore the free amino group.

Broad substrate specificity

Prototypical P450s involved in secondary metabolism are often associated with high substrate-, regio- and stereospecificities. Thus, the two OH-groups in erythromycin (12) are introduced by two distinct P450 enzymes, EryF and EryK. Similarly, the two OH-groups of the polyene macrolide antibiotic filipin are introduced by CYP105P1 and CYP105D6 (26). By sequence similarity, the filipin hydroxylases CYP105P1 and CYP105D6 group with other CYP105 family members known for the broad substrate specificity utilized in enzymatic defense functions, including degradation of xenobiotics and environmental pollutants^94–96. No apparent determinants of substrate promiscuity emerged from the crystal structures of the promiscuous members of the CYP105 family, MoxA⁹⁶ and P450SU-1^{94, 95}, suggesting lack of a clear distinction between the high and broad substrate specificity in P450s. Plausibly, a large hydrophobic active site lacking functional groups for specific substrate anchoring may be a prerequisite for broad substrate specificity.

For biotechnological applications, substrate promiscuity should combine with strict regio- and stereospecificity, which is only achieved through precise positioning of substrate relative to the catalytic iron center. Binding ambiguity may be manifested in distinct binding modes as evidenced by P450 Vdh from Pseudonocardia autotrophica. This enzyme of unknown biological function is capable of catalysing a two-step hydroxylation of vitamin D₃ to yield the 1α,25-dihydroxy derivative from two distinct anti-parallel substrate binding modes⁹⁷.

PikC is the most extensively studied biosynthetic P450, performing multiple hydroxylations of structurally variant macrolide products obtained from the pikromycin biosynthetic pathway in Streptomyces venezuelae. PikC hydroxylates both the 12-membered ring macrolactone of YC-17 (at C-10 and C-12) and the 14-membered ring of narbomycin (at C-12) at the sterically and electronically distinct carbon centers, leading to generation of the macrolides methymycin/neomethymycin (27) and the natural ketolide pikromycin (28), respectively⁹⁸.

Based on x-ray structure analysis^{25, 26}, the desosamine sugar moiety of native substrates explores two different binding pockets in the active site, suggesting a multistep substrate binding mechanism whereby substrate sequentially progresses toward a catalytically favorable binding mode. However, no alternative conformations have been observed for the macrolactone portion of the substrates. In all available PikC co-structures, hydroxylation of the methylene sites at C-12 (YC-17 ethyl side chain) and C-14 (narbomycin ethyl side chain) would appear to be favored based on the 5.3 Å distance to the heme iron^{25, 26}. However, hydroxylation of narbomycin at the C-14 position occurs at a very low level⁹⁹. Instead, the C-12 allylic center within 7.1 Å of the iron center is the major hydroxylation site in narbomycin, leading to pikromycin. By contrast, 50% of YC-17 hydroxylation is directed to C-12, while the remainder results in hydroxylation of the C-10 allylic center within 7.5 Å of the heme iron.

The original hypothesis that conformational dynamics allows PikC to accommodate different substrates in catalytically competent orientations by induced fit is not consistent with experimental observations. Although PikC does undergo significant conformational changes (Fig. 4), those sampled by the substrate-bound forms are virtually indistinguishable from the compact substrate-free form of the enzyme²⁵. Remarkably, unnatural carbocyclic substrates fitted with desosamine as the anchoring group (22) experience substantial loss of regio- and stereo-selectivity due to their wobbling in the active site instead of inducing a tighter fit²⁷ (Fig. 10B). Alternatively, the less reactive hydroperoxoferric intermediate, Compound 0, that reaches a few angstroms farther from the macrocycle plane than Compound I, could plausibly serve as an oxidant to modify the more reactive allylic carbon sites in both YC-17 and narbomycin.

Multifunctional P450 monooxygenases

P450 diversity extends to an emerging group of multifunctional biosynthetic P450s operating on substrates having widely varying chemical structures, including AurH (aureothin pathway)³⁹, MycG (mycinamicin pathway)⁴⁰, Gfs4 (macrolide antibiotic FD-891)¹⁰⁰ and TamI (tirandamycin pathway)^{101, 102}, whose reaction cascades have been recently characterized. A critical difference that sets multifunctional P450s apart from the substrate promiscuous enzymes is an apparent hierarchy in the reaction sequence, suggesting that each step is a prerequisite for the one that follows. The mechanisms that underlie the diverse reactivity of the multifunctional P450s remain to be explored. Ongoing structural and functional studies will undoubtedly provide more detailed insights into the molecular basis for sequential reactivity and pattern of oxidation in these systems.

AurH from Streptomyces thioluteus sequentially installs two C-O bonds into the polyketide backbone of deoxyaureothin (29) to yield a tetrahydrofuran ring of aureothin (31)³⁹. A hypothetical switch of function mechanism has been proposed based on computational docking of two consecutive substrates into x-ray structures of different AurH conformers. The key role is assigned to Q⁹¹, which upon completion of the hydroxylation step changes the conformation to provide an H-bond to the newly installed C7-OH group. In the course of mutual conformational adjustments, the hydroxylated intermediate (30) relocates deeper into the substrate binding pocket to enable the next attack by the activated oxygen species at C-9a, while the substrate backbone bends to facilitate tetrahydrofuran ring formation.

To address the switch of function in two other multifunctional P450s, TamI and MycG, the authors’ research team recently pursued characterization of these enzymes with their natural substrates. TamI, in the tirandamycin biosynthetic pathway of Streptomyces sp. 307-9, hydroxylates a bicyclic ketal moiety of tirandamycin C (32) to provide the substrate for the TamL-mediated oxidation to the ketone form, after which an exchange back to TamI enables successive epoxidation and hydroxylation to generate the terminal product tirandamycin B (33)^{101, 102} (Fig. 12). Reminiscent of AurH, TamI readily forms a crystallographic dimer via an engineered N-terminal overhang harboring a His-tag. To free access to the active site, the His-tag was subsequently placed at the C-terminus, albeit with no positive effect on co-crystallization.

Fig. 12. Oxidative reactions of P450 TamI.

Four consecutive oxidation steps catalyzed by P450 TamI and flavoprotein TamL in the biosynthesis of tirandamycin B in Streptomyces sp. 307-9.

MycG co-crystallization was more successful. MycG catalyses consecutive hydroxylation and epoxidation of the 16-membered ring macrolide mycinamicins M-IV (34) and M-V (35) to yield the final product M-II (36), produced by the actinomycete Micromonospora griseorubida⁴⁰. Similar to MycG, Gfs4 in Streptomyces graminofaciens operates on the 16-membered ring macrolide substrate to generate FD-891 (37) but has reverse order reactivities¹⁰⁰. Both consecutive mycinamicin substrates formed with MycG what we believe is an initial recognition complex with the hydroxylation and epoxidation sites positioned further away from the Fe center than would be required for catalysis¹⁰³ (Fig. 13A). A signature of the recognition complex is the hydrophobic interactions of the dimethoxylated mycinose sugar moiety with the heme macrocycle, which appear to be critical in discriminating between substrates and the catalytically inactive biosynthetic precursor M-III (38) bearing the monomethoxylated sugar, javose. The javose-decorated precursor penetrated deeper into the active site and assumed an orientation parallel to the heme, only to present the wrong side of the macrolactone ring to the cofactor (Fig. 13B).

Fig. 13. Substrate binding in MycG.

A, Substrates M-IV (34) and M-V (35) (yellow sticks) are bound orthogonal to the heme plane with mycinose methoxy groups at van der Waals distances to the heme (green spheres), preventing access of C-14 and C12–C13 double bond to the Fe centre. Distances between Fe and two reactive centres are in angstroms. B, Javose-decorated precursor, M-III (38), in the parallel orientation presenting the wrong side of the macrolactone ring to the heme. Electron density for M-III is shown in blue mesh.

Fungal natural product P450s

Both fungi and plants are characterized by their ability to generate natural products with high structural diversity and a broad range of biological activities^104–106. Cytochrome P450 enzymes are implicated as key enzymes in many of these processes, although the functional and particularly structural characterization of fungal and plant P450s can often be challenging¹⁰⁷. This is due to their association with intracellular membranes via a transmembrane spanning region and/or tight hydrophobic interactions via protein surfaces. The best studied fungal secondary metabolite pathways are those of mycotoxins (i.e. aflatoxins, trichothecenes and fumonisins) and higher plant hormones (i.e. gibberellins)¹⁰⁵. These oxidation pathways are often complex, and can include several oxygenation steps catalyzed by different enzymes. The list of functionally characterized fungal P450s was recently reviewed by Cresnar & Petric¹⁰⁵.

The current data reveal that many fungal P450s are multifunctional enzymes that catalyze up to four consecutive steps on the same substrate molecule. For instance, Tri4 (CYP58 family) in the plant pathogen Fusarium graminearum performs four consecutive oxygenation steps, including one epoxidation and three hydroxylations (Fig. 14) in the biosynthesis of trichothecenes, sesquiterpenoid secondary metabolites that are major mycotoxins of mold-contaminated cereal grains, which can cause serious health problems^{108, 109}. A more recent example involves the multifunctional P450 VerE from the verrucarin/roridin natural product pathway from the marine fungus Myrothecium verrucaria that performs at least three of the four predicted reactions enroute to verrucarol (Y. Ding and D.H. Sherman, unpublished). Another example of a multifunctional fungal P450 is Fum6 (CYP505 family) in the biosynthesis of fumonisins from the maize pathogen Fusarium verticillioides^110–112, which catalyses two consecutive hydroxylations at adjacent carbon atoms. The biosynthesis of gibberellins in the rice pathogen Fusarium fujikuroi involves four multifunctional P450 enzymes that catalyse 10 of the 15 biosynthetic steps^113–115. As fungal P450s are less amenable for structural and biophysical characterization, understanding the switch of function mechanism in the more accessible bacterial multifunctional P450s should bring considerable new insights to this versatile class of unexplored fungal monooxygenases.

Fig. 14. Oxidative reactions of fungal P450 Tir4.

Four consecutive oxidation steps catalysed by Tir4 in the biosynthesis of trichothecenes in fungi F. graminearum.

Plant P450s

Plant P450s represent the most diversified gene family, and contribute significantly to the phytochemical structural diversity through oxidative modifications of the carbon skeletons in these secondary metabolites¹¹⁶. Plant secondary metabolism is abundant in unusual P450-dependent reactions such as oxidative rearrangement of the carbon skeleton and oxidative C-C bond cleavage, which correlates with high numbers of amino acid substitutions in the conserved P450 motifs¹⁰⁶. Two plant allene oxide synthases, CYP74A1 and CYP74A2, from oxylipin pathways in Arabidopsis thaliana and Parthenium argentatum, respectively, have been structurally characterized^{14, 15}. CYP74 enzymes rearrange fatty acid hydroperoxides into structurally varied products. These include the prostaglandin-like jasmonates and green leaf volatiles, which are essential for plant development, host immunity, and biotic and abiotic stress management. The mechanism of reaction centre rearrangement is unusual in that substrate itself delivers dioxygen in the form of the peroxide group in a catalytically competent manner. It is speculated that the CYP74 family might have evolved prior to the transition from H₂O₂ to O₂ as oxygen donor in P450 catalysis^{14, 117}.

The unusual catalytic route in the allene oxide synthases is consistent with significant modifications in the conserved P450 motifs, including a nine-residue insert in the middle of the signature heme-binding loop and lack of A/Gⁿ⁻¹-Gⁿ-XX-Tⁿ⁺³ motif. Structurally, the N³²¹ carboxamide group positioned directly above the heme plane prevents the orthogonal binding of the diatomic oxygen molecule favoring instead the angled binding of the substrate peroxy group (39). Subsequent homolytic cleavage of the O-O bond, gives rise to the substrate alkoxyl radical and a protonated oxo-ferryl species (Fe⁴⁺-OH). An alkoxyl radical interacts intramolecularly with the C11=C12 double bond to form an epoxide ring and generate a C-centered radical stabilized by means of π-interactions with F¹³⁷ (40). Oxidation of the C-radical by (Fe⁴⁺-OH) followed by β-proton elimination, completes formation of the double bond adjacent to the epoxide to generate allene oxide (41). F¹³⁷ plays a key role in driving the reaction toward allene oxide. Aliphatic functionality at 137 promotes formation of the unstable hemiacetal, which spontaneously dissociates into short-chain aldehydes¹⁴. Elucidating the mechanisms of chemical reactions leading to a broad spectrum of molecules beneficial in agriculture and medicine opens new venues for genetic modifications in plants to enhance crop resistance, fine-tune flavors and pigments, and better understand biological functions of the P450-derived products.

Conclusions

The versatility of natural product P450s is a cumulative effect of (i) scaffold mechanics, (ii) the diverse reactivity of distinct activated oxygen species in the P450 catalytic cycle, (iii) contributions of substrate functional groups, and (iv) recruitment of carrier proteins for substrate delivery. The mobility of the F- and G-helices, the variability of the BC-loop region constructed of short helices and loops, and a matrix of individual amino acids together efficiently tailor the P450 active site to accommodate a continuum of molecular shapes and sizes. Being large and highly decorated chemical entities, natural product substrates deliver a variety of functional groups to the active site, giving P450s a degree of functional independence from normally conserved amino acid functionalities (e.g. hydroxyl groups) or even molecular oxygen and external reducing equivalents. In multifunctional P450s, newly introduced hydroxyl groups may trigger a switch of function mechanism to enable oxidative attack on the next reactive centre. Natural product substrates often accommodate multiple carbon sites with different electronic properties, including more reactive allylic carbons and double bonds. Plausibly, Compound 0, the intermediate that is less reactive than Compound I within the P450 catalytic cycle, can serve as an alternative oxidant with a slightly larger radius of action on electronically activated carbon sites when the transition state is favoured by binding geometries. Finally, the recruitment of carrier proteins enables regio- and stereospecific modifications of flexible substrates comprised of repetitive units (e.g. fatty acids and glycopeptides), or small aromatic molecules possessing a high degree of symmetry, which impose a challenge for specific recognition. Taken together these features significantly extend the repertoire of P450 catalyst versatility and provide multiple venues to directed evolution of both P450 and their substrates toward desired biosynthetic applications.