BIOTECHNOLOGICALLY RELEVANT ENZYMES AND PROTEINS Directed evolution of a Baeyer���Villiger monooxygenase to enhance enantioselectivity Anett Kirschner & Uwe T. Bornscheuer Received: 1 July 2008 /Revised: 30 July 2008 /Accepted: 2 August 2008 / Published online: 22 August 2008 # Springer-Verlag 2008 Abstract The Baeyer���Villiger monooxygenase (BVMO) BmoF1 from Pseudomonas fluorescens DSM 50106 was shown before to enantioselectively oxidize different 4- hydroxy-2-ketones to the corresponding hydroxyalkyl acetates, being the first example of a BVMO-catalyzed kinetic resolution of aliphatic acyclic ketones. However, the wild-type enzyme exhibited only moderate E values (E���55). Thus, the enantioselectivity was enhanced by means of directed evolution and optimization of reaction conditions since it was found that higher E values (E���70 for wild-type BmoF1) could already be obtained when performing biotransformations in shake flasks rather than small tubes. In a first step, random mutations were introduced by error-prone polymerase chain reaction, and BmoF1 mutants (3,500 clones) were screened for im- proved activity and enantioselectivity using a microtiter- plate-based screening method. Mutations S136L and L252Q were found to increase conversion compared to wild type, while several mutations (H51L, F225Y, S305C, and E308V) were identified enhancing the enantioselec- tivity to a varying extent (E���75���90). In a second step, beneficial mutations were recombined by consecutive cycles of QuikChange�� site-directed mutagenesis resulting in a double mutant (H51L/S136L) showing both improved conversion and enantioselectivity (E���86). Keywords Baeyer���Villiger monooxygenase . Directed evolution . Enzyme catalysis . Enantioselectivity. ��-hydroxyketones Introduction Baeyer���Villiger monooxygenases (BVMOs, E.C. 1.14.13.x) belong to the class of oxidoreductases and convert aliphatic, arylaliphatic, and cyclic ketones into esters and lactones, respectively, using molecular oxygen (Kamerbeek et al. 2003a, b Mihovilovic 2006 Mihovilovic et al. 2002 Walsh and Chen 1988). Thus, they mimic the chemical Baeyer���Villiger oxidation (Baeyer and Villiger 1899) which is usually peracid-catalyzed and proceeds by a two- step process as initially proposed by Criegee (Criegee 1948). Additionally, they are capable of aldehyde (Moonen et al. 2005) and heteroatom oxidation including sulphur, nitrogen, or phosphorous groups (Carrea et al. 1992 de Gonzalo et al. 2006). Since it is getting more and more of interest to perform Baeyer���Villiger oxidations in an enantioselective manner (Mihovilovic et al. 2004), BVMOs represent a valuable alternative to metal-based chiral catalysts (Strukul 1998). Until now, stereoselective Baeyer���Villiger oxidations using enzymes were described for prochiral or racemic mono- and bicyclic ketones (Mihovilovic 2006) as well as racemic arylaliphatic ketones (Alphand and Furstoss 2000 Kamerbeek et al. 2003a, b). Recently, we could show that also aliphatic acyclic ketones such as 4-hydroxy-2-ketones were enantioselectively Appl Microbiol Biotechnol (2008) 81:465���472 DOI 10.1007/s00253-008-1646-4 A. Kirschner : U. T. Bornscheuer (*) Institute of Biochemistry, Department of Biotechnology and Enzyme Catalysis, Greifswald University, Felix-Hausdorff-Str. 4, 17487 Greifswald, Germany e-mail: uwe.bornscheuer@uni-greifswald.de Present address: A. Kirschner Biochemistry Laboratory, Groningen Biomolecular Science and Biotechnology Institute, University of Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands

converted by a BVMO from Pseudomonas fluorescens DSM 50106 (Kirschner and Bornscheuer 2006), which was recombinantly expressed in Escherichia coli (Kirschner et al. 2007). The corresponding (S)-hydroxyalkylacetates are formed as products, which after hydrolysis give access to synthetically useful aliphatic (S)-1,2-diols. However, the enantioselectivities were only moderate (E���50), and higher values would be advantageous for preparative-scale Baeyer��� Villiger oxidations. Directed evolution and rational protein design have been proven to be powerful tools for the generation of enzyme mutants with enhanced activity and enantioselectivity (May et al. 2000 Reetz et al. 1997 Schmidt et al. 2006). While information on the 3D structure as well as the catalytic mechanism of the protein is necessary for rational redesign of the active site pocket of an enzyme to improve enantioselectivity, directed evolution can be applied without knowledge of either characteristic. Although several BVMOs have already been recombinantly expressed and characterized (Brzostowicz et al. 2002, 2003 Kamerbeek et al. 2001 Kostichka et al. 2001 van Beilen et al. 2003), so far directed evolution and rational protein design have been applied to only three of them to alter their substrate specificity and enantioselectivity. Consecutive rounds of error-prone polymerase chain reaction (epPCR) were used to improve the enantioselectivity of cyclohexanone mono- oxygenase from Acinetobacter calcoaceticus NCIMB 9871 in the conversion of 4-hydroxycyclohexanone (the corresponding (R)-lactone was produced with 90%ee by a mutant showing four amino acid exchanges compared to wild type giving the (R)-lactone with only 9%ee) and methyl-p-methylbenzylsulfide (improvement in enantio- meric excess from 14%ee to 99%ee for various mutants showing (R)- and (S)-selectivity Reetz et al. 2004a, b). In contrast, the substrate range of phenylacetone monooxy- genase (PAMO) from Thermobifida fusca was broadened by means of rational protein design due to its known 3D structure (Bocola et al. 2005 Torres Pazmino et al. 2007). Additionally, the enantioselectivity of cyclopentanone monooxygenase from Comamonas sp. NCIMB 9872 in the conversion of prochiral p-substituted cyclohexanones was altered by restricted CASTing, a combination of ra- tional protein design and saturation mutagenesis (Clouthier et al. 2006). In all examples, identification of improved mutants within the libraries was performed by rather low-throughput gas chromatography (GC) or high-performance liquid chromatography analysis. Although an agar plate screening based on a decrease in pH after hydrolysis of the ester, which is formed by Baeyer���Villiger oxidation of the substrate ketone, was suggested (Watts et al. 2002), pH assays have the disadvantage that the released acid is usually not strong enough to give a clear signal and can hardly be used in a reliable screening of whole cells and cannot be applied for an enantioselective screen. In 2003, Furstoss and coworkers reported the first fluorescence assay for BVMOs based on the release of umbelliferone after a cascade of three enzymatic and a spontaneous reaction step (Guti��rrez et al. 2003), which is rather cumbersome in terms of high-throughput screening. Recently, further chromophoric ketone substrates were published based on umbelliferone or p-nitrophenol release (Sicard et al. 2005), but this requires the multistep synthesis of starting materials and the acceptance of the bulky substrates by the BVMO. Here we report on the creation of variants of the BVMO BmoF1 from P. fluorescens DSM 50106 by directed evolution showing enhanced activity and enantioselectivity in the conversion of 4-hydroxy-2-decanone (1). For this, the basic principle of the adrenalin assay originally published by Wahler and Reymond (2002) was adapted to allow fast and reliable screening of mutant libraries. Materials and methods Chemical syntheses Racemic 4-hydroxy-2-decanone (1) was synthesized as described previously (Kirschner and Bornscheuer 2006). Synthesis of (R)- and (S)-1 was carried out by lipase-catalyzed kinetic resolution of the racemic hydroxyketone through transesterification of the free hydroxyl group using vinyl butyrate (Scheme 2). In two 100 mL flasks, immobilized Candida antarctica lipase B (250 mg each, Chirazyme L-2, C2, Roche) was mixed with dry hexane (10 mL), vinyl butyrate (1 mL), and racemic 1 (500 mg, 2.9 mmol). The reactions were stirred at 25��C for 20 or 32 h (to isolate either highly pure product or substrate), and the enzyme was removed by filtration and the solvent evaporated in vacuo. The crude products were separated by silica gel chromatography using hexane/ethyl acetate=2:1. (R)-1 was obtained in 47% yield (230 mg, 1.36 mmol) having 94%ee and (S)-butyric acid-1-(2- ketopropyl)-heptyl ester (S)-4 in 55% yield (386 mg, 1.5 mmol) with 93%ee. This was then hydrolyzed using recombinant Bacillus subtilis esterase (Schmidt et al. 2007). The ester was dissolved in dimethyl sulfoxide (DMSO 2 mL) and added to the enzyme (400 mg lyophilisate) in phosphate buffer (25 mL). The reaction was stirred for 3.5 h at room temperature and afterwards extracted four times using ethyl acetate (10 to 20 mL each). For better phase separation, the mixture was centrifuged for 5 min at 4,400��g after each extraction step. The combined organic layers were dried over anhydrous sodium sulphate and filtrated, and the solvent was removed in vacuo. The crude product was purified by silica gel chromatography using hexane/ethyl acetate=2:1. Thus, (S)-1 was obtained in 43% yield (134 mg, 0.78 mmol) with 87%ee. 466 Appl Microbiol Biotechnol (2008) 81:465���472