Expanding the toolbox of organic chemists: directed evolution of P450 monooxygenases as catalysts in regio- and stereoselective oxidative hydroxylation

Gheorghe-Doru Roiban a and Manfred T. Reetz *bc
aGlaxoSmithKline R&D Ltd, Medicines Research Centre, Gunnels Wood Road, Stevenage, Hertfordshire, UK
bDepartment of Chemistry, Philipps-Universität Marburg, Hans-Meerwein Str., 35032 Marburg, Germany. E-mail: reetz@mpi-muelheim.mpg.de; Fax: +49-64212825620
cMax-Planck-Institut für Kohlenforschung, Kaiser-Wilhelm-Platz 1, 45470 Mülheim an der Ruhr, Germany

Received 18th November 2014 , Accepted 28th November 2014

First published on 8th December 2014

Cytochrome P450 enzymes (CYPs) have been used for more than six decades as catalysts for the CH-activating oxidative hydroxylation of organic compounds with formation of added-value products. However, it has not been possible to control regio- and stereoselectivity in a general manner, which is necessary for wide (industrial) applications. Especially directed evolution as a Darwinian approach to protein engineering has changed the situation for the better during the last 3–4 years leading to extensive progress, which is summarized and analysed in this Feature Article.

image file: c4cc09218j-p1.tif

Gheorghe-Doru Roiban

Gheorghe-Doru Roiban obtained his doctoral degree in organic chemistry in 2009 at Babeş-Bolyai University, Cluj-Napoca, Romania, following a codirected thesis with University of Zaragoza, Spain. He then joined the Marie Curie Excellence Team of Dr Daniel P. Funeriu at Technical University München where he was involved in an EU funded project with focus on enzyme microarrays. From January 2011 until June 2014 he was a postdoctoral fellow in the group of Manfred T. Reetz at the Max-Planck-Institut für Kohlenforschung (Germany), focusing on enzyme catalyzed asymmetric reactions. In September 2014 he joined the synthetic biochemistry team at GlaxoSmithKline, Stevenage, UK. The Syn-Bio group is working on the development of more sustainable manufacturing processes for APIs, the group having a focus on biocatalysis and synthetic biology.

image file: c4cc09218j-p2.tif

Manfred T. Reetz

Manfred T. Reetz obtained his doctoral degree in 1969 with Ulrich Schöllkopf at Göttingen University and then performed postdoctoral work in the group of Reinhard W. Hoffmann at Marburg University. Following an appointment at Bonn University, he became Full Professor of organic chemistry back at Marburg University (1980–1991) before joining the Max-Planck-Institut für Kohlenforschung in Mülheim/Germany, where he served as Managing Director from 1993 to 2002. Following formal retirement as Director in 2011, he accepted the offer to become the first Hans-Meerwein-Research-Professor in the Chemistry Department of Marburg University while also being external (emeritus) group leader of the Mülheim Max-Planck-Institute.


The development of catalysts for the CH-activating oxidative hydroxylation RH → ROH of simple and complex organic compounds with high levels of regio- and stereoselectivity constitutes an area of intense current interest, synthetic transition metal catalysts1 and biocatalysts of the type cytochrome P450 monooxygenases (CYPs)2 being the main options. No catalyst can ever be truly general, and therefore synthetic organic chemists welcome a toolbox which features a variety of different man-made catalysts as well as biocatalysts,3 the two often being complementary. More than six decades ago chemists at Upjohn, Merck, Pfizer and Schering utilized various organisms harboring CYPs in order to oxidatively hydroxylate steroids regio- and stereoselectively, the reaction of readily available progesterone with formation of the 11α-product being a prime example in which a fungus strain served as the catalyst.4 The 11α-alcohol was easily converted by a few chemical steps into cortisone or hydrocortisone. This is a seminal example of a commercial process that combines the virtues of biocatalysis and synthetic methods, and a viable alternative to Woodward's impressive 40-step total synthesis of hydrocortisone.5

In the years that followed, further fungal and bacterial strains were screened for the hydroxylation of many other compounds. In parallel research, CYPs were isolated and characterized structurally and mechanistically.2 The basic feature of the mechanism involves abstraction of an H-atom by the catalytically active high-spin heme-Fe[double bond, length as m-dash]O intermediate (compound I, Scheme 1) with formation of the respective radical, followed by rapid C–O bond formation.2

image file: c4cc09218j-s1.tif
Scheme 1 Compound I [Ferryl (FeIV)-oxo-π porphorin cation radical].2

Currently, whenever a given compound needs to be oxidatively hydroxylated in a biocatalytic manner, it is best to first test the available wild-type (WT) CYPs.2,6 Unfortunately, this is not routinely successful. Either so-called rational design utilizing site-specific mutagenesis,7 or directed evolution can then be invoked.8 In rational design, structural and computational data is utilized in order to make reasonable choices for amino acid exchanges around the enzyme's binding pocket. This may require a trial and error strategy involving not just the generation of a single mutant, but also the formation of several likewise “designed” variants. The method has been applied to CYPs for a long time, especially by the groups of Negishi,9 Poulos,10 Wong,11 Munro12 Vermeulen,13 Commandeur,13 Pleiss,7,14 Urlacher,15 Schmid,15 Schwaneberg,16 Bernhardt17 and Hauer,18 among others.2 For example, on the basis of structural comparisons of different CYPs, Negishi showed 20 years ago that mutation L209N in mouse P450-(2a-4) induces androstenedione hydroxylase activity with high selectivity for 15α.9 A more recent contribution is due to Wong et al., who designed the P450cam variant F87A/Y96F/L244A/V247L, which is an excellent catalyst for the regio- and diastereoselective hydroxylation of (+)-valencene with formation of (+)-trans-nootkatol (86% selectivity).19 This type of research has led to the accumulation of a great deal of information regarding the correlation of mutant structures and catalytic properties,2f which is useful in the expanding area of protein engineering of CYPs.

Directed evolution is a widely applied protein engineering method comprising iterative steps of gene mutagenesis, expression and screening (or selection) of mutant enzyme libraries, each cycle exerting “evolutionary pressure”.8 Due to the Darwinian character of the method, it is currently the most reliable approach to protein engineering. Random mutagenesis was first applied by Sligar to a CYP in 1997 using a mutator strain which led to a mutant showing altered regioselectivity in the oxidative hydroxylation of fatty acids, but enantioselectivity was not reported.20 In a more systematic approach, Arnold et al. applied directed evolution to a CYP as the catalyst in the oxidation of long-chain (protected) alcohols regioselectively at terminal positions,21a which was followed by an investigation using linear alkanes such as octane or nonane.21b The latter study resulted in some improvement of regioselectivity, but enantioselectivity at non-terminal positions was low to moderate. Later, some of the mutants were screened for structurally very different substrates without performing any further mutagenesis, the oxidative hydroxylation of phenyl acetic acid esters being an example (up to 93% ee).21c

Until 2011 directed evolution had not been exploited for obtaining CYP mutants showing both high regioselectivity and high stereoselectivity (each >95%) which is needed for meaningful applications. An additional challenge is to evolve both (R)- and (S)-selective mutants. This article features recent highlights in the field of directed evolution of selective CYPs with emphasis on the increasingly important role of such selective biocatalysts in synthetic organic chemistry. In order to illuminate the molecular biological problems in this endeavor, sufficient attention is paid to methodology development in directed evolution as analyzed in selected case studies; for details the interested reader is referred to recent reviews.8

Short introduction to methods and strategies in laboratory evolution

The most commonly used gene mutagenesis techniques are error-prone polymerase chain reaction (epPCR) (a shotgun method which addresses the whole gene/protein), saturation mutagenesis (randomization of canonical amino acids at predetermined sites), and DNA shuffling (mimic of sexual evolution in nature).8 Since screening is the bottleneck of directed evolution, much work has gone into the establishment of medium- and high-throughput assays,22 in addition to the development of efficient mutagenesis strategies which enable the production of high-quality libraries requiring less screening.8 Iterative saturation mutagenesis (ISM) has emerged as a particularly effective approach which delivers small and “smart” mutant libraries.8a,23 Sites labeled A, B, C, D, etc. around the binding pocket of an enzyme according to the Combinatorial Active-Site Saturation Test (CAST) are first identified using X-ray data or homology models, each comprising one or more amino acid positions (Fig. 1A).23,24 Following screening of the initial saturation mutagenesis libraries, the best mutants serve as templates for randomization at the other sites, and the evolutionary process is continued until the desired degree of enzyme improvement has been achieved. Fig. 1B shows 2-, 3- and 4-site ISM systems. It is not necessary to explore all pathways, but some may be more proficient than others. Details and guidelines for applying ISM have been published elsewhere.23 In order to minimize the screening effort so that automated GC or HPLC can be considered, reduced amino acid alphabets have been used, e.g., NDT codon degeneracy encoding 12 instead of the normal 20 canonical amino acids as building blocks at the randomization sites. Keeping Emil Fischer's lock-and-key hypothesis in mind, this Darwinian approach is perhaps the most logical way to reshape the binding pocket of an enzyme in the quest to manipulate regio- and stereoselectivity. Prerequisite is an X-ray structure or a homology model. In the absence of structural data, epPCR or older methods such as mutator strains20,25 constitute alternatives.
image file: c4cc09218j-f1.tif
Fig. 1 (A) Sites A, B, C, D, etc. surrounding the binding pocket of an enzyme chosen for iterative saturation mutagenesis (ISM); (B) 2-, 3- and 4-site ISM systems.23

In addition to the creation of “smarter” mutant libraries, efficient screening systems have also been developed.22 Whereas some methods such as multiplexing mass spectrometry (MS) in combination with isotopically labelled substrates require expensive instrumentation,26a automated GC and HPLC robotically taking samples directly from the wells of 96-format (or larger) microtiter plates have emerged as the preferred analytical techniques.26b,c They allow the screening of about 300–600 transformants per day. On-plate UV/Vis pretests assessing CYP activity are also recommended, monitoring NADPH-consumption being one option.22 Along a different line, the known concept of enzyme activity profiling using multiple substrates for “fingerprinting” in order to identify and classify enzymes has been adapted by Fasan and coworkers for protein engineering of CYPs.27 Accordingly, five methoxy-bearing chromogenic probes of different structure and size are subjected to P450-catalysis, oxidation of the methyl group at the [–OCH3] moiety leading to formaldehyde which is rapidly detected by the Purpald-based colorimetric assay.

Initial examples of CYP mutants with notable regio- and stereoselectivity generated by directed evolution

In 2010/2011 two studies appeared describing directed evolution for generating P450 mutants with notable degrees of regio- and stereoselectivity. Zhao et al. reported protein engineering of P450pyr in the oxidative hydroxylation of N-benzylpyrrolidine (1), WT leading to the regioselective formation of (S)-2 with poor enantioselectivity (43% ee) (Scheme 2).28 A colorimetric high-throughput plate-test using two complementary alcohol dehydrogenases (ADHs) was developed for screening several thousand transformants within a short time. The goal was inversion of enantioselectivity, which was achieved by applying saturation mutagenesis at 17 amino acid positions within 5 Å of the heme-docked substrate. Following the identification of an improved (S)-selective variant (65% ee) and one showing reversed enantioselectivity (42% ee in favor of (R)-2), ISM was applied at the respective remaining 1-residue sites. This provided the best (R)-selective mutant N100S/T186I (83% ee). Improved (S)-selective mutants were not found. For further improvements, randomization at multiple residue sites employing reduced amino acid alphabets and techniques for escaping from local minima (if encountered) need to be applied.23,29
image file: c4cc09218j-s2.tif
Scheme 2 Screening system used in the P450pyr catalyzed model reaction 12 utilizing two complementary ADHs.28

Based on our long-standing interest in developing directed evolution not just for thermostabilization but also for stereoselectivity as the catalytic parameter,30 we likewise had an interest in P450 enzymes.31 We chose P450-BM3 as the catalyst, a well-known self-sufficient CYP from Bacillus megaterium,2,21,32 and testosterone (4) as the substrate which is not accepted by WT (Scheme 3).33 In previous work it had been demonstrated that due to steric reasons the single mutant F87A accepts a larger range of compounds than WT P450-BM3.2,32 In the reaction of 4, this mutant proved to be active, but unselective with formation of a 1[thin space (1/6-em)]:[thin space (1/6-em)]1 mixture of the 2β- and the 15β-isomer in addition to small amounts of other alcohols.33

image file: c4cc09218j-s3.tif
Scheme 3 P450-BM3 (F87A) catalysed oxidative hydroxylation of testosterone (4).33

The demanding problem of evolving 2β- and 15β-selective mutants separately without any erosion of diastereoselectivity was solved by applying ISM. CAST residues were chosen, which can be treated as single residue sites for saturation mutagenesis, or they can be grouped into a variety of different multiple-residue sites for randomization. The latter option entails higher genetic diversity, but also increased screening effort. In several recent ISM studies involving other types of enzymes, the use of multiple residue randomization sites has proven to be more productive than saturation mutagenesis at 1-residue sites,23 but in the case of CYPs this issue has not been resolved with certainty (and indeed both strategies can be successful). Guided by the crystal structure of P450-BM3 reported by Peterson,32d,e we grouped 20 single sites into 9 multi-residue randomization sites as shown in Fig. 2.33

image file: c4cc09218j-f2.tif
Fig. 2 Overview of the 9 defined CAST sites33 A (red), B (dark blue), C (orange), D, E, F, G, H and I (all in white) in N-palmitoylglycine (dark grey) bound P450-BM3 WT (PDB ID: 1JPZ).32d,e The heme cofactor is shown in dark grey and helix I is given as a point of reference. The inner surface of the active site cavity is represented in light grey and the potential substrate/exit lies south west, close to site A.

It was not necessary to test all 9 randomization sites, because A, B and C sufficed. In order to reduce the screening effort of libraries A and C, a reduced amino acid alphabet23 based on NCD codon degeneracy was chosen which encodes 12 amino acids (Gly, Val, Leu, Ile, Phe, Ser, Cys, Tyr, His, Asp, Asn and Arg).33 In the case of library B, NNK codon degeneracy encoding all 20 canonical amino acids was employed. In the initial round of mutagenesis, 8700 transformants were screened by automated HPLC-analysis (400 transformants per day). Table 1 features the best mutants arising from these initial libraries, which shows that the synthetic problem was essentially solved, up to 97% 2β-selectivity favoring 5 and 94% 15β-selectivity with formation of 6 being achieved with complete β-diastereoselectivity. In order to boost 15β-selectivity, one round of ISM was applied, leading to an improved variant R47Y/T49F/V78L/A82M/F87A (96% regioselectivity and 100% diastereoselectivity). A total of only 9300 transformants were screened in this study. Some of the point mutations had been described in earlier studies utilizing other types of substrates.2

Table 1 Typical P450-BM3 mutants as catalysts in the oxidative hydroxylation of testosterone (4)33
Mutant Library Mutations Conv.a (%) 5 6 Othersc
a Conditions: 1 mM testosterone, 24 h, 25 °C, resting cells containing expressed P450 mutants. For detailed expression and reaction conditions. b Mutant KSA-4 was used as a parent and site B was addressed. Mean values are given, standard deviation for conversion is ±5% and for selectivity ±1%. c Minor side products were identified by HPLC as 16β-hydroxytestosterone and 6β-hydroxytestosterone. d Wild-type amino acid was retained in at least one position; mutant KSA-9 contains a silent mutation V78V (GTA → GTG).
Starting enzyme F87A 21 52 45 3
KSA-1 Lib 330NNK A330W/F87A 79 97 3 0
KSA-2 Lib A R47I/T49I/Y51I/F87A 67 94 6 0
KSA-3 Lib A R47I/T49I/Y51V/F87A 53 94 6 0
KSA-4 Lib A R47Y/T49F/F87Ad 84 37 62 1
KSA-5 Lib B V78L/A82F/F87A 86 3 91 6
KSA-6 Lib B V78T/A82F/F87A 58 4 91 6
KSA-7 Lib B V78I/A82F/F87A 78 3 89 8
KSA-8 Lib B V78I/A82D/F87A 69 84 16 0
KSA-9 Lib B V78V/A82N/F87Ad 32 58 40 2
KSA-10 Lib B V78C/A82G/F87A 55 44 55 1
KSA-11 Lib C M185N/L188N/F87A 65 28 68 4
KSA-12 Lib C M185S/L188C/F87A 74 49 50 1
KSA-13 Lib C M185G/L188G/F87A 73 47 52 1
KSA-14 Lib AB R47Y/T49F/V78L/A82M/F87Ab 85 3 96 1
KSA-15 Lib AB R47Y/T49F/V78I/A82M/F87Ab 91 3 94 3
KSA-16 Lib AB R47Y/T49F/V78T/A82F/F87Ab 60 2 90 8

Some of the best mutants were then used in the oxidative hydroxylation of other steroids such as progesterone (7) without performing additional mutagenesis experiments. In this case the starting enzyme P450-BM3 (F87A) shifts the point of attack from position 15 to position 16, the 2β[thin space (1/6-em)]:[thin space (1/6-em)]16β ratio amounting to 18[thin space (1/6-em)]:[thin space (1/6-em)]82. It was found that variant A82N/F87A is >99% 2β-selective in favor of 8, while F87A/A330W results in reversal of regioselectivity (9[thin space (1/6-em)]:[thin space (1/6-em)]91 in favor of 9) (Scheme 4).33

image file: c4cc09218j-s4.tif
Scheme 4 P450-BM3 mutants evolved for testosterone (4) as catalysts in the oxidation of progesterone (7).33

We have discovered that these libraries constitute valuable collections of mutants which are excellent catalysts not just for steroids, but also for structurally very different substrates. For industrial purposes, we have pooled the most active hits from several studies with formation of a commercially available library of about 350 variants which can be used as a starting point for hydroxylating any new compound.

The experimental data resulting from the reactions of testosterone (4) called for a theoretical analysis of the origin of 2β- and 15β-selectivity. Since CYP-catalyzed oxidative hydroxylations constitute high-energy radical abstraction reactions by the high-spin catalytically active intermediate heme-Fe[double bond, length as m-dash]O (compound I) in the rate-determining step,2 it is unlikely that the protein environment stabilizes the transition state of these enzymatic processes to a significant extent. The crucial feature of a selective mutant must involve the optimal positioning of the substrate in the binding pocket so that the respective H-atom of the methylene moiety points toward catalytically active heme-Fe[double bond, length as m-dash]O. Mulholland has estimated that the optimal O–H–C angle is about 130°.34 Using molecular dynamics (MD) simulations and induced fit docking as provided by the Schrödinger software package35a and the QM-derived force fields for compound I,35b we were able to propose a reasonable model.33 In the case of the non-selective starting mutant F87A, two energetically similar poses for substrate 1 in close vicinity to heme-Fe[double bond, length as m-dash]O were identified, one poised for 2β-hydrogen abstraction and the other for reaction at the 15β-hydrogen atom. When considering the 2β- and 15β-selective mutants R47I/T49I/F87A and R47Y/T49F/V78L/A82M/F87A, respectively, the situation is very different because the binding pocket has been reshaped considerably. Testosterone is now forced to adopt quite different poses in the two mutants, one leading to the 2β- and the other to the 15β-product (Fig. 3).33 The induced fit Schrödinger software readily allows the visualization of amino acid side-chain movements at the binding pocket upon interaction of the substrate with the X-ray structure, which appears to be more cumbersome using other programs.

image file: c4cc09218j-f3.tif
Fig. 3 Computed pose of testosterone (4) explaining 2β-selectivity (mutant R47I/T49I/F87A) and the respective pose leading to 15β-selectivity (mutant R47Y/T49F/V78L/A82M/F87A).33

Commandeur, Vermeulen et al. also obtained a 2β-selective mutant F87N by generating all 19 mutants at this hot spot.36a Although an explanation for the origin of selectivity was not presented, the result shows once more that 19 mutants at position 87 of P450-BM3 should be considered as potential candidates when new substrates are tested. Yun et al. have also combined rational design and random mutagenesis successfully for steroid hydroxylation.36b Combinatorial alanine substitution at selected hot spots of P450-BM3 has been used to generate variants that demethylate methyl-ether and methyl-amine functions of large substrates, but this approach has not worked well in regio- and stereoselective hydroxylation of steroids.36c

Generalizing the process of directed evolution of P450 enzymes with high regio- and stereoselectivity

Following the above studies,28,33 it remained to be demonstrated that directed evolution can be applied to generate CYP mutants displaying >95% regio- and >95% enantioselectivity for both (R)- and (S)-products in the oxidative hydroxylation of a given substrate. Using once again structure-guided ISM as the mutagenesis strategy, ester 10 as the substrate and P450-BM3 as the catalyst, this goal was reached (Scheme 5).37 WT P450-BM3 is only 84% regioselective (16% other oxidation products) with poor enantioselectivity in slight favor of (R)-11 (34% ee).
image file: c4cc09218j-s5.tif
Scheme 5 Model reactions used in evolving regio- and enantioselective P450-BM3 mutants with the goal of achieving selectivities of at least 95%.37

In this study a different CAST/ISM strategy was followed. Based on the crystal structure of WT P450-BM3,32d,e 24 single residues were chosen for potential saturation mutagenesis, which can be assigned to three structural categories:37 (1) residues closest to the catalytically active heme-Fe[double bond, length as m-dash]O; (2) residues still around the binding pocket but significantly farther away from heme-Fe[double bond, length as m-dash]O; (3) residues at the entrance to the binding pocket but fairly far (15 Å) from heme-Fe[double bond, length as m-dash]O (Fig. 4).

image file: c4cc09218j-f4.tif
Fig. 4 The 24 residues in P450-BM3 considered for saturation mutagenesis,37 guided by the X-ray structure of the heme domain (PDB ID: 1JPZ).32d,e These residues were assigned to three categories marked green (residues closest to heme-Fe[double bond, length as m-dash]O), blue (residues relatively far from the heme-Fe[double bond, length as m-dash]O but still next to the binding pocket), and yellow (residues at entrance to the large binding pocket); the heme cofactor is marked in red.

Saturation mutagenesis was first performed individually at 23 of the 24 single residues (position 329 was not considered at this point), and at selected 2-residue sites F87/A328, A328/F329 and V78/L437. In the case of single-residue sites, NNK codon degeneracy was applied requiring the screening of only 100 transformants for 95% library coverage. This number increases to 3000 when randomizing a 2-residue site, which is the reason why we chose NDT codon degeneracy encoding only 12 amino acids (Phe, Leu, Ile, Val, Tyr, His, Asn, Asp, Cys, Arg, Ser and Gly), but requiring much less screening for 95% library coverage (430 transformants).23 Pre-screening for activity was performed by the traditional UV/Vis assay monitoring NADPH consumption (which is somewhat crude but fast), followed by GC-based ee-analysis of active hits. In this case taking samples robotically directly from microtiter plates for automated GC-based ee-analysis was not employed which would be most efficient, because at the time we had not optimized the analytical setup (today we can screen one sample of this reaction mixture in less than 7 minutes). Mutants showing ee-values ≥40% are listed in Table 2, which reveals some remarkable trends. Double mutants originating from randomization at 2-residue sites are generally more selective than the single mutants generated by saturation mutagenesis at single-residue sites. An exception is the (R)-selective variant A82M (80% ee). All of the double mutants are (R)-selective, the three best ones leading to 96% ee with more than 95% conversion in the model reaction (10 mM, 0.1 mmol scale, 8 h). Mutations closer to the heme-Fe[double bond, length as m-dash]O seem to have a higher impact on enantioselectivity.

Table 2 P450-BM3 mutants obtained in the initial saturation mutagenesis libraries as catalysts in the model reaction 1011 with ≥40% ee37
Entry Site Mutations Enantio.a (%)
a Average of at least two different time points. b Values observed following several scaled up experiments under different conditions.
1 WT 34, (R)
2 47 R47Y 56, (R)
3 47 R47G 52, (R)
4 78 V78T 54, (R)
5 78 V78S 45, (R)
6 82 A82M 80, (R)
7 82 A82L 57, (R)
8 87 F87D 64, (S)
9 87 F87I 48, (R)
10 87 F87G 46, (S)
11 87 F87P 42, (R)
12 260 T260L 47, (R)
13 260 T260V 44, (R)
14 263 I263G 66, (S)
15 263 I263C 55, (S)
16 328 A328V 60, (R)
18 78/181 V78C/L181I 60, (R)
19 87/328 F87V/A328V 96, (R) (94–99)b
20 87/328 F87V/A328N 96, (R) (94–96)b
21 87/328 F87I/A328V 96, (R) (86–99)b
22 87/328 F87I/A328N 86, (R)
23 328/329 A328V/P329Y 85, (R)
24 328/329 A328N/P329Y 84, (R)
25 328/329 A328V/P329V 84, (R)
26 328/329 A328I/P329C 81, (R)
27 328/329 A328V/P329H 81, (R)
28 328/329 A328I/P329I 80, (R)
29 328/329 A328V/P329G 77, (R)
30 328/329 A328V/P329C 77, (R)
31 328/329 A328V/P329S 73, (R)
32 328/329 A328V/P329I 72, (R)

In an initial attempt to enhance (S)-selectivity,37 the single mutations of some of the moderately (S)-selective variants were combined to form double mutants, a strategy that has been used in the evolution of stereoselective lipase mutants.38 In the present case such a traditional approach failed to provide notably improved variants because 70% were essentially inactive, and only one improved variant was identified (I63C/A328L; 79% ee). While these results may appear odd, recent studies regarding the question of additive versus non-additive mutational effects in protein engineering show that combining positive mutations need not lead to improved variants.39 For example, when combining two (S)-selective single mutations of an enzyme to the respective double mutant, the result may be reversed (R)-selectivity.39 At this point we opted for the previously engineered Escherichia coli strain BOU730 capable of NADPH regeneration, which has the advantage that the usual requirement of adding glucose dehydrogenase for cofactor regeneration is not necessary.40 With this host, ISM was invoked employing the best variant I263G as the template for randomization at a limited number of single sites F81, F87 and A328 using NNK, NDT and NNK codon degeneracies, respectively.

Robotic GC with samples being taken directly from the wells of 96-format microtiter plates was employed for these mini-libraries. This provided notably improved (S)-selective mutants, namely F87Y/I263G (90% ee) from library at residue 87 and I263G/A328S (96% ee; >95% conversion at 2.5 mM final substrate concentration; 3 μmol scale, 20 h) from library at position 328. Upon performing scale-up to 1.1 mmol, the (R)-selective mutant F87V/A328N led to 94–96% ee favoring (S)-11 (7% other oxidation side-products at 99% conversion). Under similar conditions the (S)-selective variant I263G/A328S led to slightly reduced enantioselectivity (94% ee; 76% conversion; 3% other oxidation products). Since selectivity may vary somewhat depending upon the substrate concentration, a syringe pump for gradual addition was used in further scale-up. Bioprocess optimization41 was not performed here nor in most of the other protein engineering studies cited in this Feature Article, which would be necessary for large-scale (industrial) applications.

Since the initial libraries were (crudely) pre-screened for activity, some highly (S)- and/or (R)-selective variants may have been missed. Therefore, the F87/A238 library was generated once more and screened directly by chiral GC. This led to the re-discovery of most of the previous variants, but also to several new ones: F87L/A328V (98% ee/R), A328S (97% ee/S), F87N/A328I (96% ee/R) and F87L/A328I (95% ee/R). All of these variants ensure >95% conversion (at 2.5 mM; 3 μmol scale; 20 h). The lessons learned from this study are twofold: (1) use the engineered E. coli strain BOU730 with built-in NADPH regeneration; (2) vary and control the ratio of substrate to P450 enzyme, and if this influences stereoselectivity, employ a syringe pump for continuous addition of substrate in scale-up experiments. A total of 4600 transformants were tested in this study, most by the UV-Vis-based pre-screen followed by GC-based ee-analysis.

Upon subjecting structurally related 12 (readily accessible by Diels–Alder reaction) to oxidative hydroxylation catalyzed by P450-BM3 variants F87V/A328N and I263G/A328S, enantioselectivities of 84% ee in favor of (R)-13 and 93% ee in favor of (S)-13 were observed at conversions of 99% and 89%, respectively, (Scheme 6).37 Of particular interest is (S)-13, because it had been prepared as the O-silyl protected form in 11 synthetic steps and used as a pivotal intermediate in the preparation of a number of biologically active molecules.42 Thus, the present two-step sequence is a viable alternative.

image file: c4cc09218j-s6.tif
Scheme 6 Selective oxidative hydroxylations catalyzed by P450-BM3 mutants.37

Many P450-catalysed hydroxylation products are biologically interesting compounds, as in the case of numerous metabolites,2 but they can also serve as synthetic intermediates in subsequent transformations with formation of value-added products.4 Recent examples include deoxyfluorination reactions (R–OH → R–F) reported by Arnold et al.43 In the present study, (R)-11 was exploited as an intermediate enroute to structurally quite different products. Hydroxy-controlled diastereoselective epoxidation44 with formation of 14 characterized by three contiguous centers of chirality, and Pd-catalysed regioselective allylic nucleophilic substitution45 of 15 leading to the GABA-analog 16 are illustrative examples (Scheme 7).37

image file: c4cc09218j-s7.tif
Scheme 7 Chemical modification of compound (R)-11.37 Reagents and conditions: (a) m-CPBA, RT, 19 h, CH2Cl2, 82%; (b) (CF3CO)2O, NEt3, 30 min, 87%; (c) BnNH2, Pd(PPh3)4, PPh3, RT, 1 h, PhMe, Ar, 85%.

In another study, Li et al. considered once more the already discussed P450pyr-catalysed hydroxylation of N-benzylpyrrolidine (1) with formation of (R)- and (S)-2 (Scheme 2).46a In this case an MS-based high-throughput screening system26a,d was applied, deuterium-labeled (R)-D-1 and (S)-D-1 serving as substrates (Fig. 5). In the previous ISM-based study,28 the best (S)-selective was only moderately enantioselective (53% ee). In order to enhance (S)-selectivity, 20 residues in the vicinity of the active site were chosen for saturation mutagenesis guided by the crystal structure, namely A77, I82, I83, L98, P99, N100, I102, A103, S182, D183, T185, T186, L251, V254, G255, D258, T259, L302, M305 and F403, each library requiring the screening of only 188 transformants.46a Some of the best mutants were used as templates for saturation mutagenesis at selected single sites, and the ISM process was continued in selected upward pathways (total of 6204 transformants screened). The best variant favoring (S)-2 with 98% ee proved to be I83H/M305Q/A77S having a respectable specific activity of 5.2 U g−1 cdw 81% activity of WT P450pyr (Fig. 5).46a (R)-Selective mutants for improving the moderate 83% ee previously evolved28 were not mentioned in this paper.

image file: c4cc09218j-f5.tif
Fig. 5 (a) High-throughput ee assay based on the use of a deuterated substrate and MS detection;46a (b) example of screening an enantioselective mutant (I83H). (i) MS spectra for fast ee determination, (ii) chiral HPLC chromatogram for accurate ee analysis, (iii) reversed phase HPLC chromatogram for conversion analysis; (c) chiral HPLC chromatogram of the purified product 2 from biohydroxylation of N-benzylpyrrolidine 1 with best mutant I83H/M305Q/A77S.

When faced with the daunting problem of controlling non-terminal hydroxylation of alkanes lacking functional groups with high regio- and enantioselectivity, the group of Li relied again on ISM.46b Using P450pyr as catalyst and n-octane as substrate, six rounds of ISM provided variant A77Q/I83F/N100S/F403I/T186I/L302V leading to (S)-2-octanol (98% ee) and complete regioselectivity. Other compounds such as propylbenzene also reacted selectively.46b

In a noteworthy report by Fasan and coworkers, P450-BM3 was subjected to directed evolution in order to generate mutants which enable late-stage hydroxylation of the anti-malaria therapeutic drug artemisinin (17), the site of attack not being defined at the start of the study (Scheme 8).47 Since WT P450-BM3 does not accept 17, the most active mutant generated earlier by the fingerprinting method, FL#62, was used as the starting gene. It is characterized by 16 point mutations: V78A, F81S, A82V, F87A, P142S, T175I, A180T, A184V, A197V, F205C, S226R, H236Q, E252G, R255S, A290V, and L353V.27 This variant accepts substrate 17 with formation of three main products 18, 19 and 20 in a ratio of 83[thin space (1/6-em)]:[thin space (1/6-em)]10[thin space (1/6-em)]:[thin space (1/6-em)]7. Further improvements were therefore necessary.47

image file: c4cc09218j-s8.tif
Scheme 8 P450-BM3 mutants as catalysts in the oxidative hydroxylation of artemisinin (17).47

As in previous studies,33,37,46 saturation mutagenesis was applied at sites surrounding or near the binding pocket. Using NNK codon degeneracy encoding all 20 canonical amino acids, eight randomization libraries were constructed at the following sites: 78/87, 78/81/87, 78/87/181, 78/87/184, 78/81/82/87, 81/82/87/184, 78/81/82/87/181/184 and 74/81/82/87/181/184. The authors estimate that this diversity includes more than 105 library members,47 which would entail huge screening efforts if 95% library coverage were to be strived for. In fact, based on the Patrick/Firth algorithm48 for calculating the number of transformants needed to be screened for 95% library coverage, NNK-based randomization at a 6-residue site would require the screening of >3 × 109 transformants as calculated by the CASTER computer aid,49 which is obviously unrealistic. Instead of employing reduced amino acid alphabets as shown to be successful previous P450-BM3 studies33,37 and in directed evolution of other enzyme types as well,23 a different strategy was followed based on the fingerprinting method27 (see Introduction above). In a first step, 12[thin space (1/6-em)]500 transformants were screened for activity in 96-well plates using five semisynthetic chromogenic probes, the liberated formaldehyde being detected by the UV/Vis Purpald-based assay at 550 nm (Fig. 6a).47

image file: c4cc09218j-f6.tif
Fig. 6 Fingerprint-based strategy to guide the discovery of regio-and stereoselective artemisinin-hydroxylating P450 catalysts.47 (a) Schematic summary of the process of identification of the functionally diverse P450 variants from the active site libraries via high-throughput fingerprinting followed by fingerprint comparative analysis (tier 2). (b) Outline of the method applied to predict ART reactivity via multivariate fingerprint analysis (tier 3). FP: fingerprint. MLR: multiple linear regression.

This screening step delivered about 1950 active variants (criterion: “>10% of parent enzyme activity on at least one of the fingerprint probes”), and 522 variants functionally unique (criterion: “larger than 20% variation on at least one of the 5 fingerprint components compared to the parent enzyme and any other member of the library”).47 Instead of testing all 522 variants by automated HPLC (which may not have been available), the authors developed a guide for searching for mutants that react with the actual substrate of interest (17). They realized that this compound had little similarity with the chromogenic fingerprint probes, and therefore developed a further guide by correlating the P450-BM3 fingerprints with experimental 17-reactivities across a small subset of the enzyme's mutants (Fig. 6b). The reader is referred to the original study for details,47 suffice it to say that at the end of the overall process, 75 variants were inspected by HPLC analysis, one (IV-H4) showing complete selectivity for 18, and the other (II-H10) being fully selective for its diastereomer 19 (Table 3).

Table 3 Best P450-BM3 mutants as catalysts in the oxidative hydroxylation of artemisinin47 (17)a
Amino acid substitutionsb Prod. distribution (%) TTN KD (μM) Prod. formation ratec CEd (%)
Variant 74 78 81 82 87 181 184 18 19 20
a Mean values and standard deviations are calculated from triplicate experiments. b Mutations in FL#62 vs. P450-BM3 are V78A, F81S, A82V, F87A, P142S, T175I, A180T, A184V, A197V, F205C, S226R, H236Q, E252G, R255S, A290V, L353V. c Moles of product per mole of P450 per minute. Rates are measured over initial 30 s. d CE: coupling efficiency; ratio between product formation rate and NADPH oxidation rate in the presence of artemisinin.
FL#62 A A S V A L V 83 10 7 339 ± 12 29 ± 5 316 ± 20 41.9
IV-H4 S V A 100 0 0 362 ± 15 53 ± 11 100 ± 4 71.4
V-H2 S I A 96 0 4 434 ± 21 45 ± 10 41 ± 2 23.5
II-H10 N F T F F T 0 100 0 270 ± 08 61 ± 2 32 ± 2 12.8
III-B1 F F A 19 81 0 403 ± 17 38 ± 2 72 ± 2 23.3
II-E2 N F A 22 30 48 393 ± 25 164 ± 18 148 ± 3 38.2
X-E12 V N F A A T 4 2 94 113 ± 12 300 ± 24 72 ± 3 45.1
X-F11 T N F A S 0 8 92 376 ± 19 234 ± 29 56 ± 2 56.3

From the 75-membered pool, another variant (II-E2) was identified which leads to 48%-selectivity in favor of attack at the C6 methyl group with formation of 20 (Scheme 8). In order to improve this variant, ISM was invoked using II-E2 as the template and performing saturation mutagenesis at positions 74/181/184.47 For 95% library coverage, this would require the screening of about 98[thin space (1/6-em)]000 transformants. The authors randomly picked 3000 transformants for screening not using the actual substrate 17, but employing the same fingerprinting procedure involving the five chromogenic probes used earlier.27 This finally provided variant X-E12 showing 94% selectivity for product 20. The essential results together with kinetic data are summarized in Table 3. Since FL#62 was used as the starting enzyme, the final three variants are characterized by 19 to 22 point mutations, which are relatively high numbers in directed. Thermostability data was not reported.

Large-scale reactions (400 mg of 17) utilizing the best variants led to ≈90% yields of the three products 18, 19 and 20, respectively. Compound 18 was deoxyfluorinated to provide 21 in stereochemically pure form, which was subsequently transformed into 23 and 24 (Scheme 9).47 The transformations starting from 17 nicely demonstrate the complementary nature of biocatalysis and modern synthetic organic methods.

image file: c4cc09218j-s9.tif
Scheme 9 Added-value reactions of compound 21.47

In another recent study, Pietruszka and coworkers took a different approach in the successful attempt to evolve P450-BM3 mutants for regio- and enantioselective oxidative hydroxylation of ω-alkene carboxylic acid esters.50 Rather than applying saturation mutagenesis at known hot spots around the binding pocket of the enzyme using NNK codon degeneracy, a 65-membered mutant library was created as follows: a P450-BM3 library was constructed by specific exchanges of the amino acid codons for R47, Y51 and L188, in addition to the generation of single and double mutant combinations A74G, F87V, A74G/F87V (GV), A74G/L188Q (GQ), F87V/L188Q (VQ) and A74G/F87V/L188Q (GVQ). All of the 65 members of this library had been described previously in the literature. The library was screened by GC analysis using 6-heptene carboxylic acid ethyl ester (25) as the model compound (Scheme 10).50

image file: c4cc09218j-s10.tif
Scheme 10 P450-BM3 catalysed oxidation of ester 25.50

Table 4 shows that especially the double mutant A74G/L188Q is an excellent catalyst, leading to 90% regioselectivity in favor of allylic hydroxylation (26) versus epoxidation (27), and 95% ee favoring the (S)-alcohol 26. Several homologs of 25 led to similar results. Although MD or docking computations were not performed, the authors speculate that the ester function can serve as an anchor point in the large binding pocket of P450-BM3, since similar alkenes lacking this functional moiety react with significantly lower degrees of selectivity. The synthetic results are significant, because the chiral allylic alcohols serve as intermediates in the synthesis of a variety of different biologically active products.

Table 4 Screening and characterization of P450-BM3 monooxygenase mutants for the stereoselective allylic hydroxylation of ethyl 6-heptenoate (25)50
Entry P450-BM3 mutant Conv. (%) 26[thin space (1/6-em)]:[thin space (1/6-em)]27 Enantio. (%) CE (%)
1 Enzyme library 0–59 65[thin space (1/6-em)]:[thin space (1/6-em)]35–97[thin space (1/6-em)]:[thin space (1/6-em)]3 87–98
2 A74G 47 ± 3 89(±0)[thin space (1/6-em)]:[thin space (1/6-em)]11 96 ± 1 36 ± 1
3 L188Q 37 ± 2 85(±0)[thin space (1/6-em)]:[thin space (1/6-em)]15 93 ± 1 51 ± 1
4 A74G/L188Q 49 ± 1 90(±0)[thin space (1/6-em)]:[thin space (1/6-em)]10 95 ± 1 52 ± 1

The approach taken in this study does not mean that such small collections of previously generated mutants will routinely lead to success when targeting other substrates. But it does show that it is not always necessary to generate new saturation mutagenesis libraries. Indeed, depending upon the structure of a new substrate, our group either starts by screening small previously generated CAST libraries, or design new small libraries guided by structural biology.

In the attempt to generate selective variants for the oxidation of 1-methylcyclohexene (28) (Scheme 11),51 we first tested WT P450-BM3 as well as some of the best hits that had been obtained in the earlier ISM-based study using ester 10 as substrate.37 Since olefin 28 lacks functional groups with lone electron pairs that can anchor in the enzyme's binding pocket by means of H-bonds, we did not expect high selectivities. Indeed, fairly indiscriminate oxidation reactions occurred which include hydroxylation and epoxidation with formation of about a dozen different products of known structure. Upon testing the double mutant F87I/A328V, which in the reaction of the ester 10 ensured 95% regioselectivity and 96% ee in favor of (R)-11,37 the fingerprint profile of the product mixture changed significantly. The only product showing high enantiopurity is the hydroxy-epoxide 29 with the trans-(1S,2R,6R) configuration (Scheme 11).51 It is the product of a synthetically interesting tandem reaction, which unfortunately occurs only to a small extent, and which needs to be optimized by directed evolution. Protein engineering of CYPs enabling sequential oxidations with creation of multiple centers of chirality is a challenging research area for the coming years.

image file: c4cc09218j-s11.tif
Scheme 11 P450-BM3 F87I/A328V catalysed oxidation of 1-methylcyclohexene 28.51

In a more recent project, we considered CYP-catalysed regioselective hydroxylation of tetralones 30a–c because the 4-hydroxy derivatives 31a–c are known to be valuable components and/or building blocks of a variety of biologically active natural products and pharmaceuticals (Scheme 12).52 These include Asian and European folk medicines such as glucosides from Juglans mandshurica containing (S)-31a which have been used in China to treat certain cancers, dermatosis and pain, as well as the fresh pericarps of Juglans sigillata used in the treatment of various illnesses.53a,b Industrially developed 8MAPK inhibitors which act as anti-inflammatory agents for treating respiratory diseases likewise deserve mention.53c

image file: c4cc09218j-s12.tif
Scheme 12 P450-BM3 catalysed hydroxylation of tetralones 30a–c with regioselective formation of the 4-hydroxy derivatives 31a–c.52

Using the parent compound 30a as the model substrate, WT P450-BM3 was found to be surprisingly regioselective with sole formation of the desired 4-hydroxy derivative 31a, but enantioselectivity proved to be poor (33% ee) in slight favor of the (S)-enantiomer. We first tested 40 previously generated variants which had shown activity toward methylcyclohexane in another study.54 As summarized in Table 5, several mutants were discovered that display >98% regioselectivity and 96–99% ee in favor of (S)-31a, all of them being single mutants at position 328, a hot spot reported in other studies.2 (R)-Selective variants were not found in this small collection. In the case of substrate 30b, the best mutant is A328F which is (R)-selective but at significantly lower regioselectivity (48%) (Table 5, entry 7). Therefore, saturation mutagenesis at position 328 was performed using NNK codon degeneracy, requiring the screening of only ≈100 transformants for 95% library coverage. This resulted in the identification of variant A328P which shows 94% ee in favor of (R)-31b at slight expense of regioselectivity (85%) (Table 5, entry 8). This small library was then screened in the hydroxylation of substrate 30c, because WT P450-BM3 is essentially stereorandom (1% ee in minute favor of (S)-31c). Two improved variants were identified (Table 5, entries 10 and 11). Structurally related substrates such as indanone, indane and tetralin were also tested using some of the already generated variants, which resulted in high selectivities. All in all, less than 200 transformants had to be screened in this study.

Table 5 P450-BM3 catalysed oxidative hydroxylation of tetralones 30a–ca[thin space (1/6-em)]52
Entry Substrate P450-BM3 Product Regio. (%) Enantio. (%) TOFb [min−1] Conv.b,c (%)
a Values were obtained by averaging at least three independent experiments performed with resting cells at 5 mM. b TOF and conversions were calculated for WT and the best mutants. c Conversion calculated after 20 h. d Not determined.
1 30a WT 31a 99 33, (S) 1.9 86
2 30a A328F 31a 98 99, (S) 3.8 >99
3 30a A328K 31a 99 96, (S) d 56
4 30a A328R 31a 99 88, (S) 59
5 30a A328Y 31a 99 97, (S) 39
6 30b WT 31b 97 82, (R) 2.2 86
7 30b A328F 31b 48 95, (R) 1.3 64
8 30b A328P 31b 85 94, (R) 71
9 30c WT 31c 91 1, (S) 6.2 88
10 30c A328F 31c 50 99, (S) 3.0 75
11 30c A328I 31c 84 86, (S) 92

In order to understand the origin of regio- and enantioselectivity of the hydroxylation of the parent compound tetralone (30a), docking experiments were performed. Several docking poses of similar energy were found, but only one of them fulfilled the requirement of close proximity to the high spin catalytically active heme-Fe[double bond, length as m-dash]O with the H-atom abstraction occurring at an angle of about 130° as predicted to be ideal for such a reaction. This pose clearly suggests that the pro-S atom of the benzylic CH2-group should react preferentially, in accord with the observed absolute configuration of product 31a.52 The conformational dynamics of 1-tetralone was then investigated by performing two unrestrained MD simulations on the docked structure. Although the substrate rarely gets close enough to heme-Fe[double bond, length as m-dash]O, when it does, it is positioned in a way that causes the pro-S hydrogen-atom to point toward the catalytically active species (Fig. 7).52 The question why 6-methoxytetralone 30b causes reversal of enantioselectivity was also answered.

image file: c4cc09218j-f7.tif
Fig. 7 Structure obtained from an unrestrained molecular dynamics simulation of 1-tetralone (30a) in WT P450-BM3 (after 34[thin space (1/6-em)]940 ps).52 The O–H distance between the ferryl oxygen of compound I and the pro-S hydrogen attached to C4 of 1-tetralone is highlighted by the blue dashed line. The F87 and A328 residues are also highlighted in yellow stick form.

Recently, directed evolution of P450-BM3 was used to generate variants that induce enantioselective α-hydroxylation of ketones,55 a biocatalytic route to chiral acyloins which are useful intermediates in the synthesis of various biologically active products. In a final study to be highlighted here, we discuss our attempts to devise systems starting from achiral substrates which upon a single oxidative hydroxylation lead to the creation of two or more new centers of chirality.54a If achiral substrates of the type I shown in Scheme 13 are considered in which regioselectivity operates in favor of the two methylene groups next to the X-carrying C-atom, then hydroxylation can occur at four different stereotopic H-atoms, each leading to a different stereoisomer (two diastereomers and their two mirror images). Finding catalysts that enable regio-, diastereo- and enantioselectivity is the challenge in this endeavor. We chose methylcyclohexane (32) as the model compound, a “difficult” substrate lacking functional groups (Scheme 14).54a

image file: c4cc09218j-s13.tif
Scheme 13 Stereochemical consequences of oxidative hydroxylation of prochiral compounds of the type I.54a

image file: c4cc09218j-s14.tif
Scheme 14 Possible products of P450-BM3 catalyzed oxidative hydroxylation of methylcyclohexane (32).54a

WT P450-BM3 proved to be surprisingly regio-(78%), diastereo-(>97%) and enantioselective (82% ee) in favor of cis-(1S,2R)-33, in addition to 34 (17%), 35 (4%) and 36 (1%).54a In order to boost selectivity, a limited number of saturation mutagenesis experiments were carried out, specifically randomization at CAST sites V78/T88, V78/L181, and T268/A328 using a highly reduced amino acid alphabet encoding only six building blocks (Phe, Tyr, Trp, Lys, Arg and His) and the corresponding WT amino acids, and site F87 using NNK codon degeneracy. Upon screening a total of 760 transformants by automated GC, the known variant F87A was identified which led to an undesired shift of regioselectivity in preference of hydroxylation at the weakest CH-bond with formation of 34 (71%), while variant A328F increased enantioselectivity at slight expense of regioselectivity (33/71%; d.r. = 97[thin space (1/6-em)]:[thin space (1/6-em)]3 (cis); 92% ee).54a

Seven mono-substituted cyclohexane derivatives were also tested with WT and the two variants F87A and A328F, which led to similar selectivities. An example is the oxidation of cyclohexyl methyl ketone (38) with preferential formation of cis-(1R,2S)-39 (78% regioselectivity; d.r. = 98[thin space (1/6-em)]:[thin space (1/6-em)]2 and 96% ee at 85% conversion) (Scheme 15).54a WT is only 58% regioselective and moderately enantioselective (54% ee).

image file: c4cc09218j-s15.tif
Scheme 15 Regio-, diastereo- and enantioselective oxidative hydroxylation of methyl cyclohexyl ketone (38).54a

This strategy for desymmetrizing prochiral compounds leading to the simultaneous creation of two chirality centers in a single hydroxylation event constitutes a unique way to transform simple achiral starting materials into value-added products. To date, chiral synthetic catalysts have not been developed for such transformations. By choosing appropriate substrates, it is possible to perform similar single hydroxylations with simultaneous creation of three or more chirality centers with high regioselectivity, but thus far it has not been possible to control enantioselectivity.54b

In addition to the substrates featured thus far in this Feature Article, other types of compounds have also been subjected to CYP-catalysed hydroxylation, generally (but not always) utilizing protein engineered mutants to control selectivity. Recent examples are summarized in Table 6. Included are selected studies of epoxidation, although this type of CYP-catalysed reaction is not systematically covered here. The question when a CYP induces epoxidation rather than hydroxylation has not been answered satisfactorily.2

Table 6 Selected recent examples of selective hydroxylation and epoxidation reactions catalysed by P450s
Entry P450 Mutations Reaction Regio. (%) Enantio. (%) Ref.
a Racemic product detected. b 0.41 g L−1, TON = 6531 (referring to the P450 monooxygenase). c Enantiomeric excess not given.
1 P450tol WT image file: c4cc09218j-u1.tif 99 (R)-99 56
2 P450terp WT image file: c4cc09218j-u2.tif 98 (R)-96 56
3 P450tol WT image file: c4cc09218j-u3.tif 95 (S)-99 56
4 P450-BM3 WT/L407C-Ru1 image file: c4cc09218j-u4.tif 100 (R)-85 57
5 P450pyr A77Q/I83F/N100S/T186I/L302V/F403I image file: c4cc09218j-u5.tif >99 (S)-98 58
6 P450-BM3 R47L/E64G/F81I/F87V/E143G/L188Q/E267V/G415S/V87F image file: c4cc09218j-u6.tif 89 59
7 P450 Pseudomonas plecoglossicida WT image file: c4cc09218j-u7.tif 99 (R)-98 60
8 P450cam F87A/Y96F image file: c4cc09218j-u8.tif >90 61
9 P450-BM3 F87A image file: c4cc09218j-u9.tif 100 62
10 P450-BM3 R47S/Y51W/I401M image file: c4cc09218j-u10.tif >99 63
11 P450-BM3 V78A/H138Y/T175I/V178I/A184V/H236Q/E252G/R255S/A290V/A295T/L353V image file: c4cc09218j-u11.tif 96 64
12 P450 from Thermobifida fusca YX WT image file: c4cc09218j-u12.tif 100 65
13 P450-BM3 WT image file: c4cc09218j-u13.tif 100 a 66
14 P450-BM3 & ADH from L. kefir R47C/V78F/A82S/K94I/P142S/T175I/A184V/F205C/S226R/H236Q/E252G/R255S/A290V/A328F/L353V/E464G/I710T/L52I/L188P/I366V image file: c4cc09218j-u14.tif b 67
15 P450-BM3 A264V/A238V/L437F image file: c4cc09218j-u15.tif 97 18
16 P450 from Sorangium cellulosum WT image file: c4cc09218j-u16.tif 100 c 68
17 P450-BM3 D251G/G307H image file: c4cc09218j-u17.tif 100 69
18 P450sca-2/Pdx/Pdr G52S/F89I/P159A/D269E/T323A/E370V/T85F/T119S/V194N/N363Y image file: c4cc09218j-u18.tif 100 70

Example 18 in Table 6 deserves special attention because it involves an elegant strategy for converting mevastatin into pravastatin (a cholesterol-lowering drug). Lin et al. applied ISM to the novel hybrid system P450sca-2/Pdx/Pdr in order to increase activity and enhance electron transfer.70 Upon screening a total of only 1100 transformants, a mutant was identified which is about one order of magnitude more active than non-engineered enzyme (whole-cell biotransformation activity of 378 mg L−1).

Other contributions relating to P450 enzymes as catalysts in organic chemistry

It has already been noted that rational design using site-specific mutagenesis continues to contribute notably to the general field of CYPs as catalysts in organic chemistry. In yet another study, Commandeur and collaborators generated a minimal library of 25 combinatorial single and double mutants starting from a template (R47L/F87V/L188Q/E267V) which they had previously produced by rational design, leading to two additional point mutations A82W/S72I.71 This novel mutant led to regioselective hydroxylation of testosterone at the 16-position with 81% α-selectivity, the first case of inversion of diastereoselectivity at this position. Merging rational design and directed evolution constitutes an attractive strategy. In a fundamentally different approach, unnatural amino acids have been used in protein engineering of P450-BM3.72

Rational design was also practiced by Keasling, Baker and coworkers in their attempt to produce P450-BM3 mutants as catalysts in the oxidation of amorpha-4,11-diene.73 One of the mutants (F87A/A328L) led to the formation of artemisinic-11S,12-epoxide at titers >250 mg L−1. In polyketide metabolic pathway engineering, Schulz, et al. developed minimally invasive mutagenesis as a unique way to generate biosynthetic polyketide compound libraries, but thus far CYPs have not been targeted.74a Brück et al. have reviewed the field of microbial factory engineering which includes CYP-catalysis in the production of cyclic sesqui- and diterpenes.74b CYPs have also been employed in cascade reactions, as in the conversion of a methylene group into the respective ketone function.67,75,76 Sieber et al. converted the unactivated C5-methylene group of camphor into 2,5-diketobornane by employing a mixture of P450cam enzymes and the dehydrogenase FdeH on a preparative scale (followed by transition metal catalyzed reductive amination with formation of the diamine).75

Our group was interested in creating designer cells for stereocomplementary de novo enzymatic cascade reactions in which cells harboring P450-BM3 mutant genes and mutant genes of the enoate reductase YqjM form the catalytic system. The goal was to construct two different strain systems, one providing (R)-41, the other the enantiomeric product (S)-41 (Scheme 16).76 The difficulty in this cascade reaction scheme is the necessity of the P450-BM3 mutants not only to be regioselective, but also substrate-selective (e.g., neither intermediate 40 nor product 41 should be accepted!). This was achieved with three different versions, the best delivering (R)- or (S)-41 (both 99% ee) with yields of 72–75%.

image file: c4cc09218j-s16.tif
Scheme 16 Enzymatic cascade reactions envisioned for the construction of stereocomplementary designer cells harboring three different enzymes.76

An under-explored “substrate engineering” strategy in CYP-catalyzed transformations, originally demonstrated by Griengl,77a has been elegantly extended by Sherman et al.77b In order to hydroxylate certain carbocyclic compounds regioselectively, the substrates were attached to a desosamine glycoside via an acetal linkage with formation of a compound that binds specifically to the CYP PikC (PikCD50N-RhFRED). The aminosugar-derived anchoring group allowed for regioselective oxidation reactions.77b

The use of chemically inert activating additives such as per-fluoro fatty acids which enter the large binding pocket of P450-BM3 allows small alkanes such as propane (but not methane) to be oxidatively hydroxylated with high activity and pronounced coupling efficiency.78a,b Other additives have also been reported.78c In another contribution, light-driven stereoselective CYP-catalyzed hydroxylations have been described, a novel technique which replaces standard NADPH-regeneration systems.79

Along a different line, Drone et al. developed a reliable expression system for one of the members of the CYP153 (AH153) family, CYP153A13a, in E. coli by fusing it to RhFred and to a histidine affinity tag.80 This self-sufficient soluble CYP is an excellent catalyst for terminal oxidation of linear alkanes, and also accepts cyclic alkanes such as cyclohexane. Finally, protein engineering of CYPS has been applied for enhancing stability,81 while immobilization82 studies likewise contribute to the applicability of these enzymes.

Conclusions and perspectives

More than six decades after the original industrial work based on various fungal and microbial strains for the regioselective hydroxylation of steroids,4 the use of cytochrome P450 monooxygenases (CYPs) has been extended to include a wide variety of substrates. However, regio- and stereoselectivity of WT CYPs is not routinely satisfactory. During the last 3–4 years, the application of directed evolution8 as an alternative to rational design7 has emerged as a powerful tool in protein engineering of CYPs. The most reliable strategy in directed evolution is structure-based iterative saturation mutagenesis (ISM) at sites surrounding the binding pocket, because it delivers high-quality libraries which can be screened on-plate using automated GC or HPLC.8a,23 We expect this trend to continue with inclusion of numerous synthetically useful substrates, including those in which a single hydroxylation event creates two or more chirality centers with formation of value-added products.54 The regio- and stereochemical control of sequential hydroxylations in one-pot reactions likewise constitutes a challenging goal for enriching the toolbox of organic chemists.51

Biocatalytic retrosynthesis will play a greater role in the future,83 which can be expected to include catalysis by CYPs. In terms of large-scale industrial applications of these enzymes, bioprocess engineering remains essential, as in any commercial enzymatic reaction.41 Combining the benefits of additive effects78 with directed evolution also needs to be considered. As Hollmann has pointed out, improvements in electron-transfer within CYPs is also necessary for industrial applications.84 The ultimate goal would be the development of reliable mutagenesis methods and strategies with which CYP-catalysed oxidative hydroxylation can be directed regio- and stereoselectively to any CH-position of a given compound that the researcher may be interested in. Presently, natural product chemists hesitate to design synthetic sequences which rely on late-stage CH-activating hydroxylation. As more mutational and structural information emerges from laboratory evolution of CYPs and from the catalytic profiles of new P450 enzymes, the situation can be expected to change. Such data serve as a basis for enhancing the quality of new mutant libraries requiring less screening. For example, ISM can be focused on two or more hot spots identified in earlier studies. Some CYP mutant libraries are commercially available. If a mutant produces at least some of the desired product in a reaction mixture resulting from the hydroxylation of a particular compound, then it can be used as the template for tuning the catalytic profile in subsequent structure-guided directed evolution.


This work was supported by the LOEWE Research Cluster SynChemBio and the Max-Planck-Society.

Notes and references

  1. Recent reviews of oxidative hydroxylation using synthetic catalysts: (a) T. Newhouse and P. S. Baran, Angew. Chem., Int. Ed., 2011, 50, 3362–3374 CrossRef CAS PubMed ; (b) M. C. White, Science, 2012, 335, 807–809 CrossRef CAS PubMed ; (c) S. R. Neufeldt and M. S. Sanford, Acc. Chem. Res., 2012, 45, 936–946 CrossRef CAS PubMed ; (d) M. Sun, J. Zhang, P. Putaj, V. Caps, F. Lefebvre, J. Pelletier and J. M. Basset, Chem. Rev., 2014, 114, 981–1019 CrossRef CAS PubMed ; (e) E. Roduner, W. Kaim, B. Sarkar, V. B. Urlacher, J. Pleiss, R. Gläser, W.-D. Einicke, G. A. Sprenger, U. Beifuß, E. Klemm, C. Liebner, H. Hieronymus, S.-F. Hsu, B. Plietker and S. Laschat, ChemCatChem, 2013, 5, 82–112 CrossRef CAS ; (f) M. Canta, D. Font, L. Gomez, X. Ribas and M. Costas, Adv. Synth. Catal., 2014, 356, 818–830 CrossRef CAS .
  2. Recent reviews of CYPs: (a) P. R. Ortiz de Montellano, Cytochrome P450: Structure, Mechanism, and Biochemistry, Springer, Berlin, 3rd edn, 2005 Search PubMed ; (b) E. M. Isin and F. P. Guengerich, Biochim. Biophys. Acta, Gen. Subj., 2007, 1770, 314–329 CrossRef CAS PubMed ; (c) A. W. Munro, H. M. Girvan and K. J. McLean, Nat. Prod. Rep., 2007, 24, 585–609 RSC ; (d) P. R. Ortiz de Montellano, Chem. Rev., 2010, 110, 932–948 CrossRef CAS PubMed ; (e) S. Shaik, S. Cohen, Y. Wang, H. Chen, D. Kumar and W. Thiel, Chem. Rev., 2010, 110, 949–1017 CrossRef CAS PubMed ; (f) C. J. C. Whitehouse, S. G. Bell and L.-L. Wong, Chem. Soc. Rev., 2012, 41, 1218–1260 RSC ; (g) J. C. Lewis, P. S. Coelho and F. H. Arnold, Chem. Soc. Rev., 2011, 40, 2003–2021 RSC ; (h) V. B. Urlacher and M. Girhard, in Enzyme Catalysis in Organic Synthesis, ed. K. Drauz, H. Gröger and O. May, Wiley-VCH, Weinheim, 3rd edn, 2012, vol. 3, pp. 1227–1267 Search PubMed ; (i) E. O'Reilly, V. Köhler, S. L. Flitsch and N. J. Turner, Chem. Commun., 2011, 47, 2490–2501 RSC ; (j) R. Fasan, ACS Catal., 2012, 2, 647–666 CrossRef CAS ; (k) Y. Khatri, F. Hannemann, M. Girhard, R. Kappl, A. Meme, M. Ringle, S. Janocha, E. Leize-Wagner, V. B. Urlacher and R. Bernhardt, Biotechnol. Appl. Biochem., 2013, 60, 18–29 CrossRef CAS PubMed ; (l) D. Holtmann, M. W. Fraaije, I. W. C. E. Arends, D. J. Opperman and F. Hollmann, Chem. Commun., 2014, 50, 13180–13200 RSC .
  3. Reviews of enzymes as catalysts in organic chemistry: (a) K. Drauz, H. Gröger and O. May, Enzyme Catalysis in Organic Synthesis, Wiley-VCH, Weinheim, 3rd edn, 2012 Search PubMed; (b) K. Faber, Biotransformations in Organic Chemistry, Springer, Heidelberg, 6th edn, 2011 Search PubMed; (c) H. Pellissier and M. Santelli, Org. Prep. Proced. Int., 2001, 33, 1–58 CrossRef CAS ; (d) M. T. Reetz, J. Am. Chem. Soc., 135, 12480–12496 CrossRef CAS PubMed ; (e) B. M. Nestl, S. C. Hammer, B. A. Nebel and B. Hauer, Angew. Chem., Int. Ed., 2014, 53, 3070–3095 CrossRef CAS PubMed ; (f) C. M. Clouthier and J. N. Pelletier, Chem. Soc. Rev., 2012, 41, 1585–1605 RSC .
  4. Review of this transformation reported in 1952 and other early steroid oxidations practiced by Upjohn Company: J. A. Hogg, Steroids, 1992, 57, 593–616 CrossRef CAS .
  5. R. B. Woodward, F. Sondheimer, D. Taub, K. Heusler and W. M. MacLamore, J. Am. Chem. Soc., 1952, 74, 4223–4251 CrossRef CAS .
  6. Recent examples of testing new CYPs as catalysts in organic chemistry: (a) E. O'Reilly, M. Corbett, S. Hussain, P. P. Kelly, D. Richardson, S. L. Flitsch and N. J. Turner, Catal. Sci. Technol., 2013, 3, 1490–1492 RSC ; (b) L. M. Podust and D. M. Sherman, Nat. Prod. Rep., 2012, 29, 1251–1266 RSC ; (c) C. J. von Bühler and V. B. Urlacher, Chem. Commun., 2014, 50, 4089–4091 RSC ; (d) Y.-C. Yin, H.-L. Yu, Z.-J. Luan, R.-J. Li, P.-F. Ouyang, J. Liu and J.-H. Xu, ChemBioChem, 2014, 15, 2443–2449 CrossRef CAS PubMed .
  7. Review of rational design in protein engineering: J. Pleiss, in Enzyme Catalysis in Organic Synthesis, ed. K. Drauz, H. Gröger and O. May, Wiley-VCH, Weinheim, 2012, vol. 1, pp. 89–117 Search PubMed .
  8. Recent overviews of directed evolution: (a) M. T. Reetz, Directed Evolution of Enzymes, in Enzyme Catalysis in Organic Synthesis, ed. K. Drauz, H. Gröger and O. May, Wiley-VCH, Weinheim, 3rd edn, 2012, vol. 1–3, pp. 119–190 Search PubMed ; (b) S. Lutz and U. T. Bornscheuer, Protein Engineering Handbook, Wiley-VCH, Weinheim, 2009, vol. 1–2 Search PubMed ; (c) N. J. Turner, Nat. Chem. Biol., 2009, 5, 567–573 CrossRef CAS PubMed ; (d) P. A. Romero and F. H. Arnold, Nat. Rev. Mol. Cell Biol., 2009, 10, 866–876 CrossRef CAS PubMed ; (e) L. G. Otten, F. Hollmann and I. W. C. E. Arends, Trends Biotechnol., 2010, 28, 46–54 CrossRef CAS PubMed ; (f) G. A. Strohmeier, H. Pichler, O. May and M. Gruber-Khadjawi, Chem. Rev., 2011, 111, 4141–4164 CrossRef CAS PubMed .
  9. M. Iwasaki, T. A. Darden, L. G. Pedersen and M. Negishi, Biochemistry, 1995, 34, 5054–5059 CrossRef CAS .
  10. (a) H. Furuya, T. Shimizu, K. Hirano, M. Hatano, Y. Fujii-Kuriyama, R. Raag and T. L. Poulos, Biochemistry, 1989, 28, 6848–6857 CrossRef CAS ; (b) T. L. Poulos, Nature, 1989, 339, 580–581 CrossRef CAS PubMed .
  11. (a) L.-L. Wong, Curr. Opin. Chem. Biol., 1989, 2, 263–268 CrossRef ; (b) A. B. Carmichael and L.-L. Wong, Eur. J. Biochem., 2001, 268, 3117–3125 CrossRef CAS ; (c) C. J. C. Whitehouse, S. G. Bell, W. Yang, J. A. Yorke, C. F. Blanford, A. J. F. Strong, E. J. Morse, M. Bartlam, Z. Rao and L.-L. Wong, ChemBioChem, 2009, 10, 1654–1656 CrossRef CAS PubMed .
  12. A. W. Munro, J. R. Coggins and J. G. Lindsay, Biochem. Soc. Trans., 1993, 21, 409S CAS .
  13. (a) V. Rea, A. J. Kolkman, E. Vottero, E. J. Stronks, K. A. M. Ampt, M. Honing, N. P. E. Vermeulen, S. S. Wijmenga and J. N. M. Commandeur, Biochemistry, 2012, 51, 750–760 CrossRef CAS PubMed ; (b) B. M. A. Lussenburg, L. Babel, N. P. E. Vermeulen and J. N. M. Commandeur, Drug Metab. Rev., 2005, 37, 20–21 Search PubMed ; (c) H. Venkataraman, S. B. A. de Beer, D. P. Geerke, N. P. E. Vermeulen and J. N. M. Commandeur, Adv. Synth. Catal., 2012, 354, 2172–2184 CrossRef CAS .
  14. (a) Ł. Gricman, C. Vogel and J. Pleiss, Proteins, 2014, 82, 491–504 CrossRef PubMed ; (b) A. Seifert and J. Pleiss, Curr. Drug Metab., 2012, 13, 197–202 CrossRef CAS .
  15. (a) C. von Bühler, P. Le-Huu and V. B. Urlacher, ChemBioChem, 2013, 14, 2189–2198 CrossRef PubMed ; (b) E. Weber, A. Seifert, M. Antonovici, C. Geinitz, J. Pleiss and V. B. Urlacher, Chem. Commun., 2011, 47, 944–946 RSC ; (c) M. Dietrich, T. A. Do, R. D. Schmid, J. Pleiss and V. B. Urlacher, J. Biotechnol., 2009, 139, 115–117 CrossRef CAS PubMed .
  16. Q. S. Li, U. Schwaneberg, M. Fischer, J. Schmitt, J. Pleiss, S. Lutz-Wahl and R. D. Schmid, Biochim. Biophys. Acta., 2001, 1545, 114–121 CrossRef CAS .
  17. (a) M. Lisurek, B. Simgen, I. Antes and R. Bernhardt, ChemBioChem, 2008, 9, 1439–1449 CrossRef CAS PubMed ; (b) K. T. Nguyen, C. Virus, N. Günnewich, F. Hannemann and R. Bernhardt, ChemBioChem, 2012, 13, 1161–1166 CrossRef CAS PubMed .
  18. A. Seifert, M. Antonovici, B. Hauer and J. Pleiss, ChemBioChem, 2011, 12, 1346–1351 CrossRef CAS PubMed .
  19. R. J. Sowden, S. Yasmin, N. H. Rees, S. G. Bell and L.-L. Wong, Org. Biomol. Chem., 2005, 3, 57–64 CAS .
  20. S. A. Maves, H. Yeom, M. A. McLean and S. G. Sligar, FEBS Lett., 1997, 414, 213–218 CrossRef CAS .
  21. (a) A. Glieder, E. T. Farinas and F. H. Arnold, Nat. Biotechnol., 2002, 20, 1135–1139 CrossRef CAS PubMed ; (b) M. W. Peters, P. Meinhold, A. Glieder and F. H. Arnold, J. Am. Chem. Soc., 2003, 125, 13442–13450 CrossRef CAS PubMed ; (c) M. Landwehr, L. Hochrein, C. R. Otey, A. Kasrayan, J.-E. Bäckvall and F. H. Arnold, J. Am. Chem. Soc., 2006, 128, 6058–6059 CrossRef CAS PubMed .
  22. Enzyme Assays – High-Throughput Screening, Genetic Selection and Fingerprinting, ed. J.-L. Reymond, Wiley-VCH, Weinheim, 2006 Search PubMed .
  23. (a) M. T. Reetz, Angew. Chem., Int. Ed., 2011, 50, 138–174 CrossRef CAS PubMed ; (b) C. G. Acevedo-Rocha, S. Kille and M. T. Reetz, Methods Mol. Biol., 2014, 1179, 103–128 CrossRef .
  24. Saturation mutagenesis at sites lining the binding pocket of an enzyme in order to evolve mutants with enhanced enantioselectivity was first applied to a lipase, although the acronym CAST was not used at the time: M. T. Reetz, S. Wilensek, D. Zha and K.-E. Jaeger, Angew. Chem., Int. Ed., 2001, 40, 3589–3591 CrossRef CAS .
  25. R. Carr, M. Alexeeva, A. Enright, T. S. C. Eve, M. J. Dawson and N. J. Turner, Angew. Chem., Int. Ed., 2003, 42, 4807–4810 CrossRef CAS PubMed .
  26. (a) M. T. Reetz, M. H. Becker, H.-W. Klein and D. Stöckigt, Angew. Chem., Int. Ed., 1999, 38, 1758–1761 CrossRef CAS ; (b) M. T. Reetz, K. M. Kühling, S. Wilensek, H. Husmann, U. W. Häusig and M. Hermes, Catal. Today, 2001, 67, 389–396 CrossRef CAS ; (c) M. T. Reetz, F. Daligault, B. Brunner, H. Hinrichs and A. Deege, Angew. Chem., Int. Ed., 2004, 43, 4078–4081 CrossRef CAS PubMed ; (d) Y. Chen, W. L. Tang, J. Mou and Z. Li, Angew. Chem., Int. Ed., 2010, 49, 5278–5283 CrossRef CAS PubMed .
  27. K. Zhang, S. E. Damaty and R. Fasan, J. Am. Chem. Soc., 2011, 133, 3242–3245 CrossRef CAS PubMed .
  28. W. L. Tang, Z. Li and H. Zhao, Chem. Commun., 2010, 46, 5461–5463 RSC .
  29. Y. Gumulya, J. Sanchis and M. T. Reetz, ChemBioChem, 2012, 13, 1060–1066 CrossRef CAS PubMed .
  30. M. T. Reetz, A. Zonta, K. Schimossek, K. Liebeton and K.-E. Jaeger, Angew. Chem., Int. Ed. Engl., 1997, 36, 2830–2832 CrossRef CAS .
  31. F. Schulz, PhD dissertation, Ruhr-Universität, Bochum, 2007 .
  32. (a) L. O. Narhi and A. J. Fulco, J. Biol. Chem., 1986, 261, 7160–7169 CAS ; (b) A. W. Munro, D. J. Leys, K. J. McLean, K. R. Marshall, T. W. B. Ost, S. Daff, C. S. Miles, S. K. Chapman, D. A. Lysek, C. C. Moser, C. C. Page and P. L. Dutton, Trends Biochem. Sci., 2002, 27, 250–257 CrossRef CAS ; (c) K. H. Clodfelter, D. J. Waxman and S. Vajda, Biochemistry, 2006, 45, 9393–9407 CrossRef CAS PubMed ; (d) D. C. Haines, D. R. Tomchick, M. Machius and J. A. Peterson, Biochemistry, 2001, 40, 13456–13465 CrossRef CAS PubMed ; (e) D. C. Haines, B. Chen, D. R. Tomchick, M. Bondlela, A. Hegde, M. Machius and J. A. Peterson, Biochemistry, 2008, 47, 3662–3670 CrossRef CAS PubMed ; (f) H. Li and T. L. Poulos, Nat. Struct. Biol., 1997, 4, 140–146 CrossRef CAS PubMed .
  33. S. Kille, F. E. Zilly, J. P. Acevedo and M. T. Reetz, Nat. Chem., 2011, 3, 738–743 CrossRef CAS PubMed .
  34. (a) R. Lonsdale, J. N. Harvey and A. J. Mulholland, J. Phys. Chem. Lett., 2010, 1, 3232–3237 CrossRef CAS ; (b) R. Lonsdale, J. N. Harvey and A. J. Mulholland, J. Phys. Chem. B, 2010, 114, 1156–1162 CrossRef CAS PubMed .
  35. (a) LLC. Schrödinger, Maestro 8.5 User Manual, Schrödinger Press, 2008 Search PubMed ; (b) A. Seifert, S. Tatzel, R. D. Schmid and J. Pleiss, Proteins: Struct., Funct., Bioinf., 2006, 64, 147–155 CrossRef CAS PubMed .
  36. (a) E. Vottero, V. Rea, J. Lastdrager, M. Honing, N. P. E. Vermeulen and J. N. M. Commandeur, JBIC, J. Biol. Inorg. Chem., 2011, 16, 899–912 CrossRef CAS PubMed ; (b) G. S. Cha, S. H. Ryu, T. Ahn and C.-H. Yun, Biotechnol. Lett., 2014, 36, 2501–2506 CrossRef CAS PubMed  ; and references cited therein; (c) J. C. Lewis, S. M. Mantovani, Y. Fu, C. D. Snow, R. S. Komor, C.-W. Wong and F. H. Arnold, ChemBioChem, 2010, 11, 2502–2505 CrossRef CAS PubMed .
  37. R. Agudo, G.-D. Roiban and M. T. Reetz, ChemBioChem, 2012, 13, 1465–1473 CrossRef CAS PubMed .
  38. M. T. Reetz, J. D. Carballeira, J. Peyralans, H. Höbenreich, A. Maichele and A. Vogel, Chem. – Eur. J., 2006, 12, 6031–6038 CrossRef CAS PubMed .
  39. M. T. Reetz, Angew. Chem., Int. Ed., 2013, 52, 2658–2666 CrossRef CAS PubMed .
  40. D. J. Bougioukou, S. Kille, A. Taglieber and M. T. Reetz, Adv. Synth. Catal., 2009, 351, 3287–3305 CrossRef CAS .
  41. (a) P. Tufvesson, J. Lima-Ramos, M. Nordblad and J. M. Woodley, Org. Process Res. Dev., 2011, 15, 266–274 CrossRef CAS ; (b) ed. A. Liese, K. Seelbach and C. Wandrey, Industrial Biotransformations, Wiley-VCH, Weinheim, 2006 Search PubMed ; (c) M. Schrewe, M. K. Julsing, B. Bühler and A. Schmid, Chem. Soc. Rev., 2013, 42, 6346–6377 RSC .
  42. J. L. Pawlak and G. A. Berchtold, J. Org. Chem., 1987, 52, 1765–1771 CrossRef CAS .
  43. A. Rentmeister, F. H Arnold and R. Fasan, Nat. Chem. Biol., 2009, 5, 26–28 CrossRef CAS PubMed .
  44. D. A. Evans and G. C. Fu, Chem. Rev., 1993, 93, 1307–1370 CrossRef .
  45. B. M. Trost and M. L. Crawley, Chem. Rev., 2003, 103, 2921–2943 CrossRef CAS PubMed .
  46. (a) S. Q. Pham, G. Pompidor, J. Liu, X.-D. Li and Z. Li, Chem. Commun., 2012, 48, 4618–4620 RSC ; (b) Y. Yang, J. Liu and Z. Li, Angew. Chem., Int. Ed., 2014, 53, 3120–3124 CrossRef CAS PubMed .
  47. K. Zhang, B. M. Shafer, M. D. Demars II, H. A. Stern and R. Fasan, J. Am. Chem. Soc., 2012, 134, 18695–18704 CrossRef CAS PubMed .
  48. W. M. Patrick and A. E. Firth, Biomol. Eng., 2005, 22, 105–112 CrossRef CAS PubMed .
  49. M. T. Reetz and J. D. Carballeira, Nat. Protoc., 2007, 2, 891–903 CrossRef CAS PubMed .
  50. K. Neufeld, B. Henßen and J. Pietruszka, Angew. Chem., Int. Ed., 2014, 53, 13253–13257 CrossRef CAS PubMed .
  51. G.-D. Roiban, R. Agudo and M. T. Reetz, Tetrahedron, 2013, 69, 5306–5311 CrossRef CAS PubMed .
  52. G.-D. Roiban, R. Agudo, A. Ilie, R. Lonsdale and M. T. Reetz, Chem. Commun., 2014, 50, 14310–14313 RSC .
  53. (a) L. Liu, W. Li, K. Koike, S. Zhang and T. Nikaido, Chem. Pharm. Bull., 2004, 52, 566–569 CrossRef CAS  , and references therein; (b) Q. Liu, P. Zhao, X.-C. Li, M. R. Jacob, C.-R. Yang and Y.-J. Zhang, Helv. Chim. Acta, 2010, 93, 265–271 CrossRef CAS PubMed ; (c) C.-K. Woo and M. Bodil Van Niel, US patent application, 0143914 A1, 2013 Search PubMed .
  54. (a) G.-D. Roiban, R. Agudo and M. T. Reetz, Angew. Chem., Int. Ed., 2014, 53, 8659–8663 CrossRef CAS PubMed ; (b) A. Ilie, R. Agudo, G.-D. Roiban and M. T. Reetz, Tetrahedron DOI:10.1016/j.tet.2014.11.067 .
  55. R. Agudo, G.-D. Roiban, R. Lonsdale, A. Ilie and M. T. Reetz, J. Org. Chem. Search PubMed  , in revision.
  56. A. Li, S. Wu, J. P. Adams, R. Snajdrova and Z. Li, Chem. Commun., 2014, 50, 8771–8774 RSC .
  57. M. Kato, D. Nguyen, M. Gonzalez, A. Cortez, S. E. Mullen and L. E. Cheruzel, Bioorg. Med. Chem., 2014, 22, 5687–5691 CrossRef CAS PubMed .
  58. Y. Yang, J. Liu and Z. Li, Angew. Chem., Int. Ed., 2014, 53, 3120–3124 CrossRef CAS PubMed .
  59. H. Venkataraman, M. C. A. Verkade-Vreeker, L. Capoferri, D. P. Geerke, N. P. E. Vermeulen and J. N. M. Commandeur, Bioorg. Med. Chem., 2014, 22, 5613–5620 CrossRef CAS PubMed .
  60. D. Zheng, M. Yang, J. Zhuo, K. Li, H. Zhang, J. Yang, B. Cui and Y. Chen, J. Mol. Catal. B: Enzym., 2014, 110, 87–91 CrossRef CAS PubMed .
  61. S. G. Bell, J. T. J. Spence, S. Liu, J. H. George and L.-L. Wong, Org. Biomol. Chem., 2014, 12, 2479–2488 CAS .
  62. K. Neufeld, J. Marienhagen, U. Schwaneberg and J. Pietruszka, Green Chem., 2013, 15, 2408–2421 RSC .
  63. A. Dennig, N. Lülsdorf, H. Liu and U. Schwaneberg, Angew. Chem., Int. Ed., 2013, 52, 8459–8462 CrossRef CAS PubMed .
  64. J. N. Kolev, J. M. Zaengle, R. Ravikumar and R. Fasan, ChemBioChem, 2014, 15, 1001–1010 CrossRef CAS PubMed .
  65. A. M. Bogazkaya, C. J. von Bühler, S. Kriening, A. Busch, A. Seifert, J. Pleiss, S. Laschat and V. B. Urlacher, Beilstein J. Org. Chem., 2014, 10, 1347–1353 CrossRef CAS PubMed .
  66. C.-H. Chiang, R. Ramu, Y.-J. Tu, C.-L. Yang, K. Y. Ng, W.-I. Luo, C. H. Chen, Y.-Y. Lu, C.-L. Liu and S. S.-F. Yu, Chem. – Eur. J., 2013, 19, 13680–13691 CrossRef CAS PubMed .
  67. S. Staudt, E. Burda, C. Giese, C. A. Müller, J. Marienhagen, U. Schwaneberg, W. Hummel, K. Drauz and H. Gröger, Angew. Chem., Int. Ed., 2013, 52, 2359–2363 CrossRef CAS PubMed .
  68. Y. Khatri, M. Girhard, A. Romankiewicz, M. Ringle, F. Hannemann, V. B. Urlacher, M. C. Hutter and R. Bernhardt, Appl. Microbiol. Biotechnol., 2010, 88, 485–495 CrossRef CAS PubMed .
  69. G. E. Tsotsou, A. Sideri, A. Goyal, G. Di Nardo and G. Gilardi, Chem. – Eur. J., 2012, 18, 3582–3588 CrossRef CAS PubMed .
  70. L. Ba, P. Li, H. Zhang, Y. Duan and Z. Lin, Biotechnol. Bioeng., 2013, 110, 2815–2825 CrossRef CAS PubMed .
  71. H. Venkataraman, S. B. A. de Beer, L. A. H. van Bergen, N. van Essen, D. P. Geerke, N. P. E. Vermeulen and J. N. M. Commandeur, ChemBioChem, 2012, 13, 520–523 CrossRef CAS PubMed .
  72. J. N. Kolev, J. M. Zaengle, R. Ravikumar and R. Fasan, ChemBioChem, 2014, 15, 1001–1010 CrossRef CAS PubMed .
  73. J. A. Dietrich, Y. Yoshikuni, K. L. Fisher, F. X. Woolard, D. Ockey, D. J. McPhee, N. S. Renninger, M. C. Y. Chang, D. Baker and J. D. Keasling, ACS Chem. Biol., 2009, 4, 261–267 CrossRef CAS PubMed .
  74. (a) S. Kushnir, U. Sundermann, S. Yahiaoui, A. Brockmeyer, P. Janning and F. Schulz, Angew. Chem., Int. Ed., 2012, 51, 10664–10669 CrossRef CAS PubMed ; (b) T. Brück, R. Kourist and B. Loll, ChemCatChem, 2014, 6, 1142–1165 Search PubMed .
  75. M. Hofer, H. Strittmatter and V. Sieber, ChemCatChem, 2013, 5, 3351–3357 CrossRef CAS .
  76. R. Agudo and M. T. Reetz, Chem. Commun., 2013, 49, 10914–10916 RSC .
  77. (a) D. F. Münzer, H. Griengl, A. Moumtzi, R. Saf, T. Terzani and A. de Raadt, Eur. J. Org. Chem., 2005, 793–796 CrossRef ; (b) S. Li, M. J. Chaulagain, A. R. Knauff, L. M. Podust, J. Montgomery and D. H. Sherman, Proc. Natl. Acad. Sci. U. S. A., 2009, 106, 18463–18468 CrossRef CAS PubMed .
  78. (a) F. E. Zilly, J. P. Acevedo, W. Augustyniak, A. Deege, U. W. Häusig and M. T. Reetz, Angew. Chem., Int. Ed., 2011, 50, 2720–2724 CrossRef CAS PubMed  ; corrigendum Angew. Chem., Int. Ed., 2013, 52, 13503–2724; (b) N. Kawakami, O. Shoji and Y. Watanabe, Angew. Chem., Int. Ed., 2011, 123, 5427–5430 CrossRef ; (c) V. Polic and K. Auclair, Bioorg. Med. Chem., 2014, 22, 5547–5554 CrossRef CAS PubMed .
  79. (a) A. Taglieber, F. Schulz, F. Hollmann, M. Rusek and M. T. Reetz, ChemBioChem, 2008, 9, 565–572 CrossRef CAS PubMed ; (b) M. Girhard, E. Kunigk, S. Tihovsky, V. V. Shumyantseva and V. B. Urlacher, Biotechnol. Biochem., 2013, 60, 111–118 CrossRef CAS PubMed .
  80. M. Bordeaux, A. Galarneau, F. Fajula and J. Drone, Angew. Chem., Int. Ed., 2011, 50, 2075–2079 CrossRef CAS PubMed .
  81. (a) T. Mandai, S. Fujiwara and S. Imaoka, Biochem. Biophys. Res. Commun., 2009, 384, 61–65 CrossRef CAS PubMed ; (b) R. Fasan, M. M. Chen, N. C. Crook and F. H. Arnold, Angew. Chem., Int. Ed., 2007, 46, 8414–8418 CrossRef CAS PubMed .
  82. (a) L. A. Wollenberg, J. L. Kabulski, M. J. Powell, J. Chen, D. R. Flora, T. S. Tracy and P. M. Gannett, Appl. Biochem. Biotechnol., 2014, 172, 1293–1306 CrossRef CAS PubMed ; (b) L. Zhao, G. Güven, Y. Li and U. Schwaneberg, Appl. Microbiol. Biotechnol., 2011, 91, 989–999 CrossRef CAS PubMed ; (c) I. Wynant, A. Dirieux, H. Bohets, K. Lavrijsen, G. Horvath and J.-P. Simon, Biocatal. Biotrans., 2009, 27, 348359 CrossRef .
  83. N. J. Turner and E. O'Reilly, Nat. Chem. Biol., 2013, 9, 285–288 CrossRef CAS PubMed .
  84. F. Hollmann, I. W. C. E. Arends and K. Buehler, ChemCatChem, 2010, 2, 762–782 CrossRef CAS .

This journal is © The Royal Society of Chemistry 2015