Engineering Cytochrome P450 Enzymes - Chemical Research in

Dec 8, 2007 - The last 20 years have seen the widespread and routine application of methods in molecular biology such as molecular cloning, recombinan...
0 downloads 0 Views 301KB Size
220

Chem. Res. Toxicol. 2008, 21, 220–231

Engineering Cytochrome P450 Enzymes Elizabeth M. J. Gillam* School of Biomedical Sciences, The UniVersity of Queensland, St. Lucia, Brisbane, Australia 4072 ReceiVed August 7, 2007

The last 20 years have seen the widespread and routine application of methods in molecular biology such as molecular cloning, recombinant protein expression, and the polymerase chain reaction. This has had implications not only for the study of toxicological mechanisms but also for the exploitation of enzymes involved in xenobiotic clearance. The engineering of P450s has been performed with several purposes. The first and most fundamental has been to enable successful recombinant expression in host systems such as bacteria. This in turn has led to efforts to solubilize the proteins as a prerequisite to crystallization and structure determination. Lagging behind has been the engineering of enzyme activity, hampered in part by our still-meager comprehension of fundamental structure–function relationships in P450s. However, the emerging technique of directed evolution holds promise in delivering both engineered enzymes for use in biocatalysis and incidental improvements in our understanding of sequence–structure and sequence–function relationships, provided that data mining can extract the fundamental correlations underpinning the data. From the very first studies on recombinant P450s, efforts were directed toward constructing fusions between P450s and redox partners in the hope of generating more efficient enzymes. While this aim has been allowed to lie fallow for some time, this area merits further investigation as does the development of surface-displayed P450 systems for biocatalytic and biosensor applications. The final application of engineered P450s will require other aspects of their biology to be addressed, such as tolerance to heat, solvents, and high substrate and product concentrations. The most important application of these enzymes in toxicology in the near future is likely to be the biocatalytic generation of drug metabolites for the pharmaceutical industry. Further tailoring will be necessary for specific toxicological applications, such as in bioremediation. Contents 1. Introduction 2. Engineering Expression and Solubility––N-Terminal and Other Modifications 3. Engineering Activity 4. Engineering Electron Transport and Redox Partner Interactions 5. Engineering Other Properties 6. Factors Likely To Be Important for the Successful Implementation of Engineered Enzymes in Specific Applications 7. A Perspective for the Next 20 Years?

220 221 222 224 224 225

226

1. Introduction Cytochrome P450 enzymes (P450s)1 are central to the study of toxicology due to their fundamental roles in the clearance of xenobiotics in mammals and most other organisms studied to date. These are enzymes of unequalled catalytic versatility: While the most common reactions catalyzed can generally be classified as monooxygenations, this overall definition encompasses many distinct types of biotransformation (reviewed in 1 and 2), such as carbon hydroxylations within aliphatic chains and aromatic systems, heteroatom dealkylations and oxidations, epoxidations with or without group migration, and oxidation of aldehydes to acids. Alterations to ring systems may occur in * To whom correspondence should be addressed. Tel: +61-7-3365 1410. Fax: +61-7-3365 1766. E-mail: [email protected]. 1 Abbreviations: GDEPT, gene-directed enzyme prodrug therapy; MeIQ, 3,5-dimethylimidazo[4,5-f]quinoline; P450, cytochrome P450, heme-thiolate protein P450; SRS, substrate recognition sequence; StEP, staggered extension process.

some cases. In addition, cytochrome P450s (P450s) have been found to catalyze reductions of N-oxides, reductive dehalogenations, and isomerizations of various types. While many P450mediated reactions involve scission of a molecule (e.g., dealkylations), coupling reactions are also known (1, 2). P450s have been found in all kingdoms and in most organisms in which they have been sought (3). The multiplicity of P450s in some organisms, for example, Arabidopsis thaliana (with 273 P450s; http://drnelson.utmem.edu/Arablinks.html) and Oryza satiVa (rice, with at least 323 P450s; http://drnelson.utmem.edu/ rice.html) underscores their importance in many biosynthetic, catabolic, and signaling pathways. Currently, there are over 6800 P450 sequences known (http://drnelson.utmem.edu/ CytochromeP450.html), although far fewer forms have been characterized at the protein or functional level. As a group, P450s also show diversity in the substrates that they accept, in line with the divergent roles for which they have evolved in nature. Some individual forms, such as those involved in xenobiotic metabolism, show extreme versatility at the single enzyme level, acting on a broad range of substrates that may vary in size, structural features, and chemical properties, such as charge and lipophilicity. For example, eight P450s are responsible for the clearance of roughly three-quarters of all drugs in clinical use and P450 3A4 alone is believed to catalyze 40–50% of P450-mediated drug metabolism (4, 5). These three factors—the catalytic versatility, substrate diversity, and sheer number of P450s—have led to considerable interest over the last 20 years in the engineering of P450 enzymes for a variety of purposes (Figure 1). The purpose of this perspective is to explore the successes and failures in the attempts to tailor P450s to different purposes, for both practical

10.1021/tx7002849 CCC: $40.75  2008 American Chemical Society Published on Web 12/08/2007

PerspectiVe

Chem. Res. Toxicol., Vol. 21, No. 1, 2008 221

Figure 1

applications and the advancement of understanding of these unusual enzymes, with a particular focus on the xenobioticmetabolizing forms that are most relevant to toxicology. Four general areas will be addressed, namely, the engineering of recombinant expression, physical properties, catalytic activity, and interactions with redox partners. The scope will be limited to those studies designed intentionally to alter the enzyme for a defined application rather than pure analyses of structure– function relationships or elucidation of the roles of specific residues in particular activities. Also, studies in which the P450 has been used for an application without any sequence modification (i.e., engineering) will not be discussed at any length.

2. Engineering Expression and Solubility––N-Terminal and Other Modifications The study of P450s has depended heavily on the recombinant expression of individual forms in heterologous systems such as bacteria, yeast, baculovirus-infected insect cells, and mammalian cell culture. Because most organisms studied to date have multiple P450s, purification was required for characterization of individual forms from native tissues. While P450s have certain characteristics, such as the typical Fe(II)·CO vs Fe(II) difference spectrum, which facilitate quantification of the intact protein, the difficulty of discriminating between forms and problems with reconstituting activity in vitro made recombinant expression an attractive prospect. Therefore, it is not surprising that some of the first attempts at purposeful tailoring of P450s were done with the aim of achieving successful expression in a host system. Heterologous expression in bacteria has been the aim of a large number of engineering studies. While native coding sequences can often be expressed successfully in host systems such as mammalian and insect cell cultures and yeasts, it was discovered in the early 1990s that some modification of the P450 nucleotide sequence, and often also the amino acid sequence, was required to facilitate expression of significant amounts of recombinant hemoprotein in bacteria. Initial attempts at expression involved modification of simply the N-terminal sequence since this region (6), and especially the second codon (7), had been proposed to be especially important in determining protein yield. The first critical success was obtained in 1991 by Barnes et al. (8). Bovine P450 17A1 was expressed with the second codon altered to Ala and the

subsequent seven codons altered to enhance AT content (9), with the intention of minimizing secondary structure formation in the transcript and concurrently to match Escherichia coli codon preferences (10–14). Many subsequent studies employed the same approach with more or less success, depending on the form concerned [e.g., P450 2E1 (15), P450 4A4 (16), P450 6A1 (17), P450 1A1 (18), P450 27 (19), P450 2A6 (20), and P450 4A11 (21)]. Alternatively, the N-terminal sequence from the successful P450 17A1 recombinant was simply aligned to the N-terminal sequence of the form to be expressed and substituted at the appropriate position or attached directly to the N terminus [e.g., P450 1A2 (22, 23), P450 3A4 (24, 25), P450s 2C1, 2C2, 2C3, 2C4, 2C5, 2C8, 2C9, 2C16, 2C18, 2C19 (26, 27), P450 76A1 (tyrosine N-hydroxylase from Sorghum bicolor) (28), P450 3A5 (29), P450 2B1 (30), P450 73A4 (cinnamate 4-hydroxylase from E. coli) (31), P450 3A7 (32), P450s 4F4 and 4F5 (33), P450 2F3 (34), P450 2B6 (35), and P450 3A43 (36)]. An extension of this approach, namely, optimization of codons throughout the coding sequence to those preferred by the host organism by complete gene synthesis, has been recently used to good effect with P450s 2W1 and 27C1 (37, 38). In other studies, bacterial leader sequences have been fused in frame with the P450 in an attempt to direct expression to the outside of the cell or to the periplasmic space and, in some cases, to enable a fully native sequence to be delivered after cleavage of the leader sequence (39–41). An alternative approach was based on the hypothesis that the hydrophobicity of the N-terminal anchor of many P450s impeded efficient expression in bacteria. Thus, studies by the laboratories of M. J. Coon (42–44) and J. Chiang (45, 46) involved truncation of the N-terminal sequence to a greater or lesser extent. It was found that disruption of the proline-rich region was deleterious (47, 48), this region being essential for the correct folding of the nascent P450 but not the maintenance of the correct conformation (49). Truncation prior to that motif allowed the successful expression of many forms [e.g., P450 2E1 and P450 2B4 (42–44, 50), P450 7A1 (45, 46), P450 2C9 (51), P450 2D6 (52, 53), P450 2C11 (54), P450 1B1 (55), P450 46A1 (56), and P450 1A5 (57)]. Sueyoshi et al. employed an alternative approach to solubilization of the P450 in substituting an amphipathic leader sequence for the N terminus of P450 2a4 (58).

222 Chem. Res. Toxicol., Vol. 21, No. 1, 2008

The mechanisms by which N-terminal modifications enhance P450 expression in E. coli are poorly understood even after 16 years of successful bacterial expression. It is still not possible to predict which sequence modification will be most effective for a given form. However, changes made to culture media and conditions, such as supplementation with heme precursors (52, 53), coexpression with chaperones (59, 60), induction of a heat/cold shock response (61), and microaerobic culture (62),2 have enabled many forms to be successfully expressed despite an incomplete understanding of the mechanisms behind sequence effects on protein folding, heme insertion, and control of translation. Another important motivation for expression of a soluble P450 was the crystallization of a eukaryotic form, achieved initially for a modified form of the rabbit P450 2C5 and thereafter for several other P450s (63–73), leading to major shifts in our understanding of the structure and function of P450s. Again, solubilization of the P450 was a principal aim, since aggregation of lipophilic, membrane-bound proteins hinders crystallization and structure determination. In addition to N-terminal truncations, alteration of other residues believed to be involved in association with the membrane in vivo and aggregation of P450s in vitro has often been required to achieve monomers. Deletion of residues 3–20 of P450 2C3 plus mutation of other residues at the new N-terminus led to a protein that was soluble at high ionic strength (74). These alterations, plus substitution of residues from the F-G loop region of the more soluble P450 2C3, resulted in the engineering of a soluble, monomeric P450 2C5 (75), the first mammalian form to be crystallized. Again, both adaptation of the approach as well as simple substitution of the N termini from successfully expressed forms have proven useful for different P450s [e.g., various P450 2B forms (64, 76) and P450 2C9 (71)]. For some forms, N-terminal modification without any additional alterations to the internal sequence to reduce aggregation has led to generation of a soluble enzyme suitable for crystallization [e.g., P450 3A4 (67, 72), P450 2C8 (65), P450 2C9 (66), P450 2A6 (68), P450 2D6 (73), P450 2A13 (69), and P450 1A2 (70)]. The crystallization of mammalian P450s has revealed the essential similarity in overall fold between forms but also the subtle yet critical differences between related forms. The structure of P450 2B4 (64) provided evidence for open and closed conformations of the P450, supporting the idea that these enzymes underwent major conformational changes in solution. The P450 2D6 structure (73) allowed many years of speculation regarding substrate binding by this form to be rationalized and that of P450 1A2 more recently (70) provided an explanation of the characteristic predominance of the high-spin state form of this P450. However, the various structures of P450 3A4 with and without alternative ligands have revealed both the power and the limitations of crystal structures in explaining structure– function relationships in P450s (77). It is now widely appreciated that crystal structures are unlikely to reflect the variety of conformations adopted by a P450 when binding to its substrates, let alone different intermediates along the reaction pathway that may influence the efficiency with which a P450 acts on a given substrate.

3. Engineering Activity Mutagenesis studies on the bacterial P450s 101 and 102A1 have led to significant improvements in our understanding of 2 Johnston, W. A., Huang, W., Hayes, M. A., De Voss, J. J., and Gillam, E. M. J. (2007) Quantitative whole cell cytochrome P450 measurement suitable for high throughput application. J. Biomol. Screening, in press.

Gillam

general aspects of P450 catalytic function, such as the role of the conserved threonine residue (78, 79). Site-directed mutagenesis has been widely employed with mammalian, membranebound P450s, and especially those involved in human drug metabolism, for assessment of determinants of catalytic specificity and efficiency. Many of these studies have been built upon comparisons of the different activities of similar forms, such as members of the same subfamily (reviewed in ref 80). Incremental advances in our understanding of the determinants of substrate specificity have been made in this manner, with some notable early advances such as the conversion of testosterone hydroxylase specificity by a single amino acid change (81). Homology models based on existing crystal structures have also been used for hypothesis generation and to rationalize changes observed. Before the first successful crystallization of a mammalian P450, Jones and colleagues used homology modeling and engineering to obtain evidence for a conserved overall fold between mammalian and bacterial P450s, demonstrating activity in a chimera between P450 2C9 and P450 101 (82). One of the early outcomes of structure–function studies was the proposal of substrate recognition sequences (SRSs) by Gotoh (83) in which a comparison of available bacterial P450 structures, P450 family 2 sequences, and mutagenesis results revealed six regions of sequence that appeared to contain relatively more sequence diversity and in which mutations appeared to affect substrate binding. Subsequent examination of crystal structures showed SRSs to be located in proximity to the substrate binding site and putative access channels of the P450 (63, 67, 73). A recent area of interest is the design and development of novel P450s as biocatalysts of reactions of commercial interest, such as in pharmaceutical or fine chemical synthesis. Rational engineering has been carried out on P450 101 and P450 102A1 to good effect; P450 101 and P450 102 have been engineered to improve activity toward various alkanes, halogenated alkanes and benzenes, and polyaromatic hydrocarbons [P450 101 (84–92) and P450 102 (93–95)]. Rational approaches are based on some understanding of the determinants of function in P450s (96). While this understanding is generally satisfactory for the heavily studied, bacterial forms, structure–function relationships are less clear for mammalian, xenobiotic-metabolizing forms. The concept of SRSs appears to be valid; however, as more crystal structures are becoming available, the diversity in the binding modes shown by single enzymes in response to alternative substrates is becoming evident (77), confounding the identification of clear links between specific structural elements and substrate binding. Thus, random or semirandom methods (often focused on SRSs) have been employed on P450s from families 1–3 as well as bacterial forms. In particular, directed evolution has been used as both a technical tool and a means of generating information. The advantage here is that by using the inherent power of mutation and selection, combined with bulk screening of mutant libraries, mutants can be found with novel combinations of properties without any prior understanding of the mechanism by which the mutations introduced have conferred the altered characteristics. The Guengerich group has used saturation mutagenesis of SRSs (97) and subsequently random mutagenesis throughout the entire P450 open reading frame to generate libraries derived from P450 1A2 and P450 2A6 (97–100). Twenty-seven P450 1A2 mutants were found showing elevated activation of 3,5dimethylimidazo[4,5-f]quinoline (MeIQ) from 6000 screened colonies. Different relative activity profiles were seen toward a

PerspectiVe

range of substrates, and some mutants showed activities that increased 3–4-fold over those of the parental P450 1A2 (97). In a subsequent study using the same mutagenicity screen but with error-prone PCR, optimized to introduce 1–2 changes per clone throughout the entire open reading frame (98), enzyme efficiency was increased 12-fold after three generations of mutagenesis and screening of ∼12000 mutants per generation. A P450 1A2 triple mutant obtained using three rounds of the same mutagenesis strategy coupled with a fluorometric screening method showed a 5-fold improved kcat toward 7-methoxyresorufin (99). In each case, turnover was improved without any enhancement of binding affinity. Using indigo formation as a colorimetric screen, Nakamura et al. isolated mutants with improved activity toward indole and indole analogues, after mutagenesis of P450 2A6 by a combination of saturation mutagenesis and staggered extension process (StEP) (100). The best mutants showed increased kcat values for coumarin 7-hydroxylation and naphthol 1-hydroxylation but not 7-ethoxycoumarin deethylation activity, supporting the premise that activities can be improved independently by the engineering of P450s. A mutant isolated in this study was subsequently applied to the generation of novel indigoids for inhibition of cyclin-dependent kinases and in engineering-altered traits in tobacco (101, 102). Wu et al. set out to expand the activity of P450 2A6 toward bulky indole derivatives by random mutagenesis of the entire open reading frame of this mutant. From ∼3000 clones, a mutant was isolated containing a total of five mutations as compared to the wild type that turned blue in the presence of 5-hydroxybenzylindole, indicating that an expansion had been engineered into the typically compact P450 2A6 active site (103). Altered coumarin 7-hydroxylation activity was seen among random mutants of P450 2A6, but none showed enhancements in enzyme efficiency (104). Kumar, Halpert, and colleagues have focused on P450 2B1 and P450 3A4 in several studies of rational engineering and directed evolution. Rational engineering was performed on P450 2B1 based on the P450 2C5 crystal structure to alter the regioselectivity of progesterone metabolism from 16R- to 21hydroxylation (105). A P450 2B1 mutant was isolated with a 3-fold higher kcat than P450 2C5 for 21-hydroxylation and 80% regioselectivity, but the Km for progesterone was still higher than that of P450 2C5. A 6-fold improvement in mutant P450 2B1 activity toward 7-ethoxy-4-trimethylcoumarin was engineered by a combination of random and site-directed mutagenesis (106). The activity of this quadruple mutant (V183L/F202L/ L209A/S334P) was supported effectively by peroxide, whereas a triple mutant (F202L/L209A/S334P) showed elevated NADPHsupported activity as well as increased activity toward 7-benzyloxyresorufin, benzphetamine, and testosterone (106). P450 2B1 mutations were also found that conferred enhanced activation of the chemotherapeutic prodrugs, cyclophosphamide and ifosphamide, and could be used to rationally improve the activity of P450 2B11 toward these prodrugs (107). A similar approach of random and site-directed mutagenesis was used to improve the peroxide-supported activity of P450 3A4 toward benzyloxyquinoline (108). The Km for peroxide decreased, and the kcat/Km,HOOH increased, while the efficiency toward 7-benzyloxy-4-(trifluoromethyl)coumarin and testosterone also improved. Recombinatorial methods have been used in several studies on bacterial and mammalian P450s, with the advantage that many more mutations can be introduced simultaneously than by random and site-directed mutagenesis. Unfortunately, because of the complex changes introduced into such mutants, inter-

Chem. Res. Toxicol., Vol. 21, No. 1, 2008 223

pretation of the effect of individual mutations on the structure of the P450 is much more difficult. The technique of DNA shuffling (109, 110), involving homology-dependent recombination of fragments of orthologous P450 genes in a primerless PCR reassembly, has been exploited to the greatest effect. The Pompon group published the first results on DNA shuffling of mammalian P450s (111). In a mosaic library derived from P450 1A1 and P450 1A2, 12% of clones showed activity toward naphthalene, a common substrate of both P450 1A1 and P450 1A2 (111), and were catalytically diverse with respect to five different substrates (112, 113). In contrast to mutants derived by saturation mutagenesis, shuffled P450 1A mutants did not necessarily align with either P450 1A1 or 1A2 parent in a correlation analysis with methoxy- and ethoxyresorufins. With a third, structurally dissimilar substrate, mutants showing similar phenotypes with respect to the alkoxyresorufins could be further distinguished (112, 113). Other recent studies have confirmed the ability of DNA shuffling to generate libraries of functional diversity with a relatively high proportion of properly folded P450s (114, 115). Directed evolution has also been used to good effect with bacterial P450s, particularly P450 102A1 (P450BM3). In most studies, a combination of random mutagenesis and recombination has been used, such as in the engineering of activity toward small alkanes, polyaromatic hydrocarbons, and other commercially important molecules. Several excellent reviews have been published recently on this topic (116–118); therefore, a detailed discussion will not be provided here. However, one application of engineered bacterial P450s is especially relevant to toxicology. There has been particular interest in engineering P450 102A1 to better metabolize druglike molecules (119–122) for the generation of substantial amounts of authentic metabolites as well as in optimization of lead compounds. The Arnold laboratory showed that P450 102A1 mutants could be used to generate metabolites from a drug (propranolol), a highly attractive goal for the pharmaceutical industry. Rounds of saturation and random mutagenesis led to modest improvements in the peroxidesupported activity of a mutant P450 102A1 heme domain toward ring hydroxylation of propranolol (119). Yields obtained were comparable to those produced by recombinant human P450s in bacterial or baculovirus bioreactors (62, 123). However, the total number of turnovers achieved was much lower than that obtained with an alternative (nondrug) substrate, suggesting that further mutagenesis is required to optimize yields (119). While recombination of homologous sequences by DNA shuffling is effective in generating libraries of diverse and functional P450s, it fails when sequence identity falls below a critical level. The Arnold laboratory has developed the SCHEMA algorithm to facilitate construction of functional mosaic mutants from forms showing 73% of these were catalytically active (125). The folded P450s showed functional diversity and novel activities absent from the parents (126). SCHEMA has yet to be applied to mammalian P450s but may enable cross-family mosaics to be made, for example, between enzymes from families 2 and 3. The results of artificial chimeragenesis between forms such as P450 2D6, P450 2C9, and P450 3A4, which between them catalyze the majority of human oxidative drug metabolism, may be particularly informative. We may expect significant insights into the factors determining structure and catalytic specificity of these enzymes to flow from such experiments.

224 Chem. Res. Toxicol., Vol. 21, No. 1, 2008

4. Engineering Electron Transport and Redox Partner Interactions In almost all situations, P450s require a close interaction with auxiliary proteins to effect their monooxyygenase activity. The auxiliary proteins form an electron transport system to convey electrons from the reducing cofactor NAD(P)H to the heme center. Unfortunately, in many P450-dependent biotransformations, NADPH consumption is uncoupled from substrate monooxygenation to a significant extent. This is particularly true where P450s are not specialized to the substrate under investigation, but most P450s are uncoupled to at least some degree. In these cases, oxygen is reduced and released as superoxide, hydrogen peroxide, or water, resulting in poor efficiency of the catalytic cycle. One notable exception is P450 102A1 (P450BM3), a highly efficient fatty acid hydroxylase from E. coli, which exists as a fusion between a P450 domain and a FAD- and FMN-containing reductase domain. P450 102A1 has shown the highest turnover rates of any P450, and NADPH consumption is typically wellcoupled to the oxidation of preferred substrates so P450 102A1 has been held up as a model for optimizing P450 catalytic efficiency. Thus, fusions with NADPH-cytochrome P450 reductase and b5 have been explored as a way to engineer P450s with enhanced turnover and efficiency. Ohkawa and colleagues were the first to create artificial fusion proteins between mammalian P450s and their redox partners (127–131). Typically, the C terminus of the P450 domain was fused via a short linker of ∼2–3 residues to the trypsinsolubilized cytoplasmic domain of the reductase (therein losing the first ∼56 residues of the reductase including the membrane anchor). Estabrook and colleagues adapted this approach to engineer fusions between various mammalian P450s and their cognate redox partner, NADPH-P450 reductase (25, 132, 133). Other groups adopted the same approach with other forms, such as P450scc (134), P450 4A1 (135), P450 1A1 (136, 137), and P450 73A4 (31). Unfortunately, catalytic rates with fusion enzymes, while often comparable to or greater than rates achieved by reconstitution of the isolated enzymes, typically seemed to be suboptimal as evident from the enhancement seen with the supplementation with exogenous reductase (133, 138). This suggested that the arrangement of the two domains was not optimal and that intermolecular as well as intramolecular electron transfer occurred (133). Few studies (128, 130) have been reported on optimization of the linker between the domains, which has typically been designed to be much shorter than that proposed for P450 102A1 (139), significantly restricting the possible orientations of the two domains. Bicistronic expression systems, where the two proteins were expressed from a single plasmid but as discrete proteins (140–142), have largely supplanted fusion proteins as a system of choice for routine studies on the metabolic capacities of P450s. However, there is still significant interest in fusions for the purpose of engineering self-sufficient biocatalysts. Unusual fusions have been attempted, such as the P450 101–putidaredoxin–putidaredoxin reductase triple fusion (143), and fusions with nonmammalian reductase domains have received attention recently (144, 145). Unfortunately, no artificial fusion to date has approached the cataytic efficiency of P450 102A1. Coupling efficiencies of some forms, that is, the relative proportion of reducing equivalents expended in monooxygenation of product, may be only ∼10%, although coexpression or addition of cytochrome b5 can improve this (131, 133). From the very first studies of expression of P450s in bacteria, it has been recognized that endogenous E. coli redox proteins,

Gillam

such as the flavodoxin/flavodoxin reductase couple (146), have the potential to support the activity of heterologously expressed P450s (8). One exciting prospect is to engineer P450s to better couple with these redox partners. However, the effect on natural bacterial physiology is difficult to predict and requires investigation before a significant investment is made. An alternative approach is to support activity, not by a redox partner, but by exploiting the peroxide shunt. The Arnold laboratory first evolved P450 101 to efficiently use hydrogen peroxide in place of NADH (147) and then subsequently engineered a P450 102A1 with enhanced peroxygenase activity (148). Among mammalian P450s, Kumar et al. (108) isolated a double mutant of P450 3A4 with an 11-fold improvement in kcat/Km, CuOOH for 7-benzyloxyquinoline debenzylation supported by cumene hydroperoxide. Enhanced peroxide-supported activity was also seen for several other substrates. However, kcat values for activities supported by peroxides were typically an order of magnitude lower than those for the same enzymes using NADPH as the electron donor (108). Most peroxide-supported systems are confounded by the inevitable inactivation of the P450 by concentrations of peroxide used to support activity (149). However, Chefson et al. demonstrated that much lower peroxide concentrations could be used effectively by P450s 2D6 and 3A4 (150). Other means by which to transfer electrons have been examined, such as using mediators or direct electron transfer from electrodes (151–156), and the P450 has in some cases been engineered to better couple with these electron transfer systems (157). Such systems also avoid the cost of supplying NADPH to reactions using purified or semipurified enzymes. While it is difficult to draw direct comparisons, turnover rates obtained with such systems are generally 5–10-fold lower than those for the native NADPH-supported conditions (153, 155, 157, 158), coupling efficiencies are typically low (152, 156), and the addition of catalase may be needed to counter hydrogen peroxide formation (158). However, some studies have reported electrocatalytic turnover rates comparable to NADPH-supported rates (158, 159). Whether electrons are transferred directly from an electrode to the P450 (160) or using mediators and P450s in solution (152), typically only a small proportion of the P450 is electroactive. Coupling P450 activity to in situ NADPH regeneration also represents an effective way in which to avoid the addition of costly cofactor to isolated enzymes. Photosynthesis has been explored as the source of electrons using light-driven NADPH synthesis in chloroplasts and a fused NADPH P450 reductase–P450 1A1 monooxygenase system (161, 162). An alternative means by which to reduce cofactor costs has been to alter the cofactor specificity of the reductase toward NADH (163–165).

5. Engineering Other Properties P450 102A1 has been engineered by random and saturation mutagenesis to enhance solvent tolerance for delivery of hydrophobic substrates (166). Maurer et al. were able to use a biphasic cyclohexane/aqueous system to deliver cyclohexane, myristic acid, and octane (165). Auclair and colleagues have demonstrated that even typically unstable mammalian enzymes are compatible with biphasic organic/aqueous systems (167) and can be successfully lyophilized (168), facilitating their use in anhydrous conditions and adaptation to other process requirements for implementation as commercial biocatalysts. A quadruple mutant isolated by random mutagenesis of P450 2B1 (106) was optimized for both thermostability and DMSO tolerance (169). A targeted screen at elevated temperature and

PerspectiVe

DMSO content allowed the isolation of mutants with enhanced thermostability and activity in the presence of DMSO (169). Salazar et al. improved the thermostability of P450 102A1 mutants by several rounds of random mutagenesis and DNA shuffling (170). Urlacher and colleagues have recently improved the thermostability of P450 102A1 by exchanging the reductase domain with that of P450 102A3 (171) and with the sulfite reductase from Geobacillus stearothermophilus. Engineering of subcellular location is another goal of engineering for use of P450s as biocatalysts. In particular, it would be ideal if P450s and their redox partners could be displayed on the outside of the host cell to facilitate efficient delivery of substrate and recovery of the product to minimize processing costs (122). Surface display of P450 102A1 and rat NADPH-P450 reductase expression on the bacterial cell surface has been effected by attachment of a leader sequence from the ice nucleation protein (172). While the general applicability of this strategy to other P450s remains to be demonstrated, this result provides some promise for the successful engineering of whole cell biocatalysts in which workup costs will be substantially reduced.

6. Factors Likely To Be Important for the Successful Implementation of Engineered Enzymes in Specific Applications The current attempts to engineer the catalytic or physical properties of P450s are largely driven by the attractiveness of using P450s as catalysts in the fine chemical and pharmaceutical industries. As biological catalysts, P450s routinely impart a high degree of regio- and stereoselectivity to the transformations that they catalyze, a property that is particularly useful in drug and other fine chemical synthesis, where the substrates may be complicated organic molecules with many similar functional groups. However, for the economic viability of a biocatalytic process, P450s must be developed with sufficient efficiency toward a particular reaction to achieve satisfactory product yields and specificity, and as noted above, precursor and workup costs must be minimized (173). In addition, the successful implementation P450 biocatalysts in either bulk or fine chemical synthesis will require adaptation of these relatively unstable enzymes to the industrial environment. Factors such as thermostability and solvent tolerance, which have already been addressed in some studies as noted above, will need to be substantially improved. A cost-effective means of supplying electrons will be required. While this is most easily achieved by using intact, metabolically active cells, difficulties with substrate access to intracellularly expressed P450s or product recovery from whole cells may limit the usefulness of intact bacteria. Where product recovery issues necessitate the use of isolated enzymes, either in situ NAD(P)H regeneration or the peroxide shunt may prove cost-effective. The efficiency with which supplied electrons are used to generate monooxygenated product will also influence the yield and cost effectiveness of biocatalytic processes using P450s. Xenobiotic-metabolizing P450s show typically low rates of coupling, and whereas P450 102A1 is almost completely coupled when acting upon its natural fatty acid substrates, coupling efficiencies are reduced to