Enantioselective Alcohol Synthesis Using Ketoreductases, Lipases Or an Aldolase

Enantioselective Alcohol Synthesis using Ketoreductases, Lipases or an Aldolase

Menno J. Sorgedrager

Cover: Representing “Diversity in Parameter Space” Part of a screenprint by J.C. van de Griendt; in honour of my father

Enantioselective Alcohol Synthesis using Ketoreductases, Lipases or an Aldolase

Proefschrift

ter verkrijging van de graad van doctor aan de Technische Universiteit Delft, op gezag van Rector Magnificus, prof. dr. ir. J.T. Fokkema, voorzitter van het College van Promoties, in het openbaar te verdedigen op

29 mei 2006 om 12.30 uur

door

Menno Jort SORGEDRAGER

ingenieur in de bioprocestechnologie geboren te Groningen

Dit proefschrift is goedgekeurd door de promotor: Prof. dr. R.A. Sheldon

Toegevoegd promotor: Dr. ir. F. van Rantwijk

Samenstelling promotiecommissie:

Rector Magnificus Voorzitter Prof. dr. R.A. Sheldon Technische Universiteit Delft, promotor Dr. ir. F. van Rantwijk Technische Universiteit Delft, toegevoegd promotor Prof. dr. W.R. Hagen Technische Universiteit Delft Prof. dr. J.A.M. de Bont Technische Universiteit Delft Prof. dr. A. Liese Technische Universiteit Hamburg Prof. dr. ir. A.P.G. Kieboom Universiteit van Leiden Dr. G. Huisman Codexis Inc. (USA, CA)

The research described in this thesis was financially supported and performed in cooperation with Codexis inc. (Redwood City, USA).

ISBN: 90-9020702-3

All rights reserved. No part of the material protected by this copyright notice may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without written approval from the author.

Contents

INTRODUCTION

Chapter 1 Introduction 1

PART I: LIPASE CATALYSED RESOLUTION OF ALDOL ADDUCTS

Chapter 2 Lipase catalysed Resolution of Nitro aldol Adducts 21

Chapter 3 Optimising the Deracemisation of Nitro aldol Adducts 33

PART II ENANTIOSELECTIVE CARBONYL REDUCTION

Chapter 4 Asymmetric Reduction with Candida Magnoliae Ketoreductase S1 47

Chapter 5 Asymmetric Carbonyl-Reductions with Microbial Ketoreductases 63

PART III ENANTIOSELECTIVE ALDOL REACTION: DERA

Chapter 6 Production and Optimisation of 2-deoxyribose-5-phosphate Aldolase 79

Chapter 7 DERA as Catalyst for Statin Precursors 97

Chapter 8 Cross-linked Enzyme Aggregates of DERA 113

Summary 125

Samenvatting 127

Dankwoord 129

Curriculum vitae 132

Introduction

Introduction The carbonyl group is probably the most important functional group in organic chemistry. Besides aldehydes and ketones, many functional groups contain the C=O bond, to which they own much of their behaviour in chemistry, such as carboxylic acids, acidhalides, acid anhydrides, esters and amides. When the molecule is bearing at least one α-hydrogen, which is acidic due to the electron-withdrawing effect of the carbonyl group and therefore can be abstracted by a strong base, these compounds are able to react via the resulting enolate ion. This forms the basis for a variety of synthetically usefull C-C bond forming reactions. Michael addition, aldol and Claissen reactions are examples of this reaction pathway.

One of the most important of these reactions is probably the aldol reaction. Besides taking a central role in synthetic organic chemistry, this reaction is vital for all living organisms in nature. Aldol-type reactions are the key step in the metabolic pathways of micro-organisms, where it takes part in the breakdown of carbohydrates. Besides its role in anabolism, the aldol reaction is often applied in catabolism as well. The aldol reaction is initiated by the formation of the enolate by a base, followed by nucleophilic attack hereof on another carbonyl carbon (Figure 1.1). Depending on the reaction conditions it is possible to isolate the β-Hydroxy carbonyl compound or the unsaturated product that is formed by subsequent dehydration. In the latter case it is referred to as aldol condensation.

β-Hydroxy carbonyl compounds are versatile and interesting intermediates. They can easily be synthesised directly via an aldol reaction of the corresponding aldehydes. The β-ketoesters, later described in this Chapter in connection with their reduction into β-Hydroxy esters, are easily accessible via the aldol reaction analogue for esters known as the Claisen condensation.

Base H H O OH dehyration O - - O H2C O H2C O O

Figure 1.1 The aldol reaction and subsequent condensation of acetaldehyde

2 Chapter 1

The demand for optically pure intermediates has grown rapidly in recent years. Therefore it is of growing importance to control chirality during a reaction. Although there are various chemical methods available1 biocatalysts are more and more applied to achieve this goal.1,2 Enzymes from the classes hydrolases, oxidoreductases and lyases are synthetically interesting for this purpose. Lipases, able to hydrolyze esterbonds; Ketoreductases, able to reduce carbonyl groups and aldolases that perform aldol reactions are well suited to synthesize enantiopure secondary alcohols.

Biocatalysis: Kinetic resolution vs asymmetric synthesis Two strategies are possible to obtain homochiral compounds. One is a kinetic resolution of a racemic mixture, in which the enzyme catalyses the reaction of the two enantiomers of the substrate with different rates. The lipase mediated esterification of racemic secondary alcohols (Chapters 2 and 3) is a well known example of applying a kinetic resolution strategy. The other is an asymmetric synthesis starting from a prochiral substrate. In an asymmetric synthesis a new chiral centre is introduced into the substrate molecule. Examples of asymmetric syntheses, which are reported in this thesis, are the asymmetric reduction mediated by ketoreductases (Chapters 4 and 5) and the DERA catalysed aldol reaction (Chapters 6-8).

The efficiency of a resolution process is dependent on the difference in rate in which the two enantiomers are converted. This is indicated with the enantiomeric ratio (E), which is the ratio of the two pseudo-first order kinetic rate constants (eq. 1.1). The enantiomeric excess of a compound (ee) is the excess amount of one enantiomer compared to the total amount of both enantiomers (eq. 1.2).

The enantiomeric ratio:

R RR k ()VKmax / m E ≡= (eq. 1.1) S SS k ()VKmax / m

3 Introduction

The enantiomeric excess: ()cc− ee = RS (eq. 1.2) ()ccRS+

cx: Concentration enantiomer x

The E value can be determined from experimental data via the conversion and either the ee of the substrate or the ee of the product.3

ln{() 1−−ξ ( 1 ees )} Via substrate: E = (eq. 1.3) ln{}()() 1−+ξ 1 ees

ln{ 1−+ξ ( 1 eep )} Via product: E = (eq. 1.4) ln{} 1−−ξ () 1 eep

E: Enantiomeric ratio ees: Enantiomeric excess of the substrate eep: Enantiomeric excess of the product ξ: Conversion

An alternative method for calculating the enantiomeric ratio (E) directly from the enantiomeric excess of the substrate (ees) and the product (eep) is given by equation 1.5. The conversion of an irreversible reaction without substrate inhibition (up to 40% conversion) can then be calculated directly from the optical purities of the substrate (ees) and the product (eep) according to equation 1.6.3

()11−+ee() ee Direct from ee’s: E = lnss ln (eq. 1.5) 11++ee ee ee ee ()sp() sp ee Conversion: ξ = s (eq. 1.6) ()ees + eep

In a classical kinetic resolution two enantiomers of a racemic mixture are transformed to product with different reaction rates. In the ideal situation only one of the enantiomers reacts to product and thus yields a maximum conversion of 50 % of the total amount of substrate. To overcome this limitation many efforts have been made to combine the racemisation of the substrate in the resolution proces4 in a so-called dynamic kinetic resolution

4 Chapter 1

(DKR, Figure 1.2) The biggest challenge here is to find conditions that are compatible with both processes.

Fast Fast (S)-Substrate (S)-Product (S)-Substrate (S)-Product in Situ Racemisation

Slow Slow (R)-Substrate (R)-Product (R)-Substrate (R)-Product

Classical kinetic resolution Dynamic kinetic resolution

Figure 1.2 General schemes for a classical kinetic resolution and a dynamic kinetic resolution

Lipase resolution Lipases are enzymes of the hydrolase group. In nature they catalyse the hydrolysis of triglycerides and other long chain fats and oils. Besides their normal substrates most lipases accept a very broad range of acyl donors and acyl acceptors5. Since lipases generally exhibit mostly a high stability, accept a wide variety of substrates and have a wide occurrence in nature, they have become readily available and industrially applicable enzymes. They have current industrial applications as detergent enzymes, in food and paper technology and in the pharmaceutical and speciality chemicals industry5.

In their natural reaction lipases operate at an oil/water interface and convert sparingly water soluble triglycerides. Therefore, it is perhaps not surprising that they exhibit high stability and activity in organic media. Whereas in aqueous media ester hydrolysis is the main reaction of a lipase, in non- aqueous media the reaction with other nucleophiles such as alcohols, esters, hydrogenperoxides, ammonia and amines becomes possible.

5 Introduction

Asp

O O H His N

N H O Asp - Asp O O O O O Ser O O NH HN H H His O N O His N R N N O O Asp H O H O O Ser O O Ser O NH HN H HN His NH O N O O O N R Free ROAc H O Acyl enzyme enzyme intermediate - (with acyldonor) O O (with alcohol) Ser NH HN O O

Tetrahedral intermediates

Figure 1.3 Catalytic machanism of a lipase catalyzed esterification of a chiral alcohol with vinyl acetate as model acyl donor.

The active site of a lipase consists of a catalytic triad of serine, histidine and aspartate. In most lipases a flexible lid of one or more short α helices covers this active site. In the open form the substrate (acyl donor) can enter the active site and bind with the serine to form the tetrahedral intermediate. The negative charge on the oxygen atom in the tetrahedral intermediate is highly stabilised by hydrogen bonding with peptide backbone NHs of the so-called oxyanion hole and electrostatic effects of the protonated imidazole ring of histidine. The release of an alcohol or water molecule (in Figure 1.3 given as acetaldehyde, which is the tautomeric product of the released alcohol when vinylacetate is used as acyl donor) results in the acyl-enzyme intermediate. In

6 Chapter 1

the next step this reacts with a nucleophile (in Figure 1.3 an alcohol) to result in the product that is released and the enzyme that is ready for the next catalytic cycle5,6,7.

Often the enantioselective catalytic properties of lipases are used in a kinetic resolution to resolve chiral substrates such as alcohols or amines. Recently these processes are more often designed as a dynamic kinetic resolution (see previous paragraph).

Keto reductases The asymmetric reduction of ketones is one of the most efficient methods for the production of chiral secondary alcohols. In nature such reactions are important steps in many metabolic pathways and therefore the source of carbonyl reducing enzymes from nature is abundant. Keto reductases belong to the enzyme class of the oxidoreductases. The aldo-keto reductase (AKR) superfamily and the short chain dehydrogenase/reductase (SDR) superfamily, to which keto reductases belong, are subdivisions of this class of enzymes. Based on crystal structure analysis, sequence information and point mutation studies a common catalytic pathway for SDR enzymes has been proposed. The majority of SDR enzymes have an active site that consists of a catalytic triad of tyrosine, lysine and serine (Figure 1.4). 8,9

The first step in the mechanism is binding of the cofactor NAD(P)H to the apo enzyme to form the binary state of the enzyme with the reduced form of the cofactor in the active site (ER). The lysine residue in the active site plays a role as stabilizer for the position of the cofactor. Subsequently the keto substrate binds to form a ternary state of the enzyme (ERK). The hydride is transferred from the cofactor to the carbonyl carbon of the substrate to form the ternary alcohol bound oxidized state (EOA). Tyrosine donates a proton in this process and thus functions as a general acid. Release of the alcohol brings the enzyme back in a binary state with the oxidized form of the cofactor still present (EO). The last step is the release of the oxidized cofactor and

7 Introduction

protonation of the catalytic tyrosine residue by bulk water to bring the enzyme back in its apo state. (Figure 1.4).

Tyr Tyr Ser OH Ser OH HO HO O O H H O R R O H H O H H

H2N H2N N N NAD(P)H OH O OH O

+ + H N Lys H3N Lys R OH 3 R OH NAD(P)H NAD(P)H

ER: Binary reduced state ERK: Ternary keto-bound reduced state Tyr

Ser OH HO

+ H3N Lys

E: Apo state Tyr Tyr Ser OH Ser OH O O H O O H H R NAD(P)+ H O H O H OH- H N H2N 2 N N OH OH O OH O R H + +H N Lys H3N Lys R OH 3 R OH + NAD(P)+ NAD(P)

EO: Binary oxidized state EOA: Ternary alcohol-bound oxidized state

Figure 1.4: Proposed catalytic mechanism for SDR enzymes

Enzymes of the SDR superfamily use four different pathways for transferring the hydride from the cofactor NAD(P)H to the carbonyl carbon of the substrate. Either the pro-(S) (for E2 and E4 enzymes) or the pro-(R) hydride from the cofactor (for E1 and E3 enzymes) can be transferred to either the Re-face or the Si-face of the carbonyl carbon (figure 1.5)10. Hydride attack on the Re-face is according to Prelog’s rule for the prediction of the stereoselectivity of biological ketoreductions11, while attack on the Si-face will be anti-Prelog.

8 Chapter 1

E2 E3

O H H E1 E4 O H H S R O S R H2N H2N L Si face s N N Re face R R NAD(P)H NAD(P)H

Figure 1.5 Stereospecificity of SDR enzymes

Most of the ketoreductases have a preference for one of the two possible cofactors: NADH, NADPH. The origin of SDRs having different coenzyme specificity is related to their natural function. The intracellular ratio of NADPH over NADP+ is relatively high. Hence NADP(H) dependent enzymes function in vitro mainly as reductases while, due to a significantly low NADH over NAD+ ratio enzymes dependent on NAD(H) will mainly work as dehydrogenases.12 The specificity of the SDR enzymes towards NAD(H) or NADP(H) as coenzyme seems to be dominated by electrostatic effects.13,14,15,16,17 The amino acids located around the binding position of the adenine ribose moiety of the coenzyme are typically different in nature for NAD(H) preferring enzymes than for NADP(H) preferring enzymes.

Aldolases: 2-Deoxy-D-ribose-5-phosphate aldolase (DERA) Aldolases are enzymes that belong to the enzyme class of the lyases. These are present in nearly all living organisms where they take a central role in catabolic metabolism. Aldolases catalyse carbon-carbon bond formation (or cleavage) through aldol reactions, typically with high enantioselectivity. There are two types of aldolases known: Type I, which uses an imine intermediate between a Lysine in the active site and the donor substrate in the catalytic mechanism and type II which requires a metal cofactor (Zn2+) in the catalytic mechanism18 (Figure 1.6).

9 Introduction

Glu His

CO2 OH His Zn H 2- His O3PO OH O O N 2- O3PO OH 2- O OPO3 Lys OH H2N HO OH Asp His CO2H Glu OH His 2- Zn O3PO OH CO2H NH Type I Type II O His

Lys 2- O OPO3 OH H2N Glu Asp HO CO H CO2H 2 OH OH O 2- O3PO O His OH His OH Zn O OH O His 2-O PO OH 3 2-O PO NH 3 R 2- O 2- OPO3 R=OPO3 R=H OH OH OH H2N Lys HO OH

Figure 1.6 Mechanistic scheme of the two types of aldolases represented by fructose-1,6- bisphosphate aldolase (Type I) and fuculose-1-phosphate aldolase (type II).

Most aldolases are very specific for and limited by their natural substrates. Further characterization of aldolases is therefore done by classification with the donor substrate they are dependent on (Table 1.1).18 This specificity limits the range of applications but is beneficial in cross-aldol reactions. Chemically catalyzed cross-aldol reactions mostly yield mixtures comprised of all possible product combinations of the substrate aldehydes and or ketones. Aldolases, owing to their often high selectivity in accepted donor substrates, generally yield a single product.

Table 1.1 Aldolase classification by donor dependence18

Classification Donor substrates Example aldolase Pyruvate dependent pyruvate NeuAc aldolase Phosphoenolpyruvate phosphoenolpyruvate NeuAc syntetase dependent Glycine dependent glycine l-threonine aldolase Dihydroxyacetone dependent dihydroxyacetone phosphate FruA, RAMA Acetaldehyde dependent acetaldehyde DERA

10 Chapter 1

Due to the position of most aldolases in metabolism the donor or the acceptor substrate is a phosphorylated compound. The cell has phosphorylated these molecules somewhere in the preceding metabolic pathway, to make them more reactive, but more important that they cannot leak out of the cell through the cell membrane after prior energy investment. For organic synthetic purposes the necessity for these phosphorylated substrates is a drawback because they are costly to synthesize and after the reaction the phosphate group has to be removed. This has precluded a widespread application of aldolases untill today.

2-Deoxy-D-ribose-5-phosphate aldolase (DERA) is a type I aldolase, which takes a crucial role in the anabolism of a cell. DERA’s natural reaction is the aldol reaction between acetaldehyde and D-glyceraldehyde-3-phosphate to form 2-deoxy-D-ribose-5-phosphate, which is the sugar moiety of the DNA backbone. (Figure1.7)

O O O OH 2- O3PO DERA 2- 2- CH3 + OPO3 OPO3 O OH OH OH OH

Figure 1.7 Natural reaction of DERA: the aldol reaction of acetaldehyde and D- glyceraldehyde-3-phosphate to 2-deoxy-D-ribose-5-phosphate

The crystal structure and the full catalytic mechanism of this aldolase have recently been elucidated19,20 (Figure 1.8) Two lysine residues and an aspartate are identified to be crucial for catalytic activity. Due to electrostatic interactions with the environment of Lys167 has a slightly lower pKa. Therefore it is uncharged and able to act as a nucleophile to attack the carbonyl donor. Asp102 and Lys201 together with an active site water molecule comprise a very fast proton relay system responsible for shuttling protons during catalysis.

11 Introduction

Lys 201 Lys 201 Lys 201 Lys 201

NH 3 NH2 NH3 NH2 O O H H O H H O H H H H O O O O O H H H2O H O HO HO Asp 102 Me OH O OH H NH2 Asp 102 HN Me Asp 102 HN Me Asp 102 N C H2

Lys 167 Lys 167 Lys 167 Lys 167

O R R O O

Me Lys 201

Lys 201 Lys 201

NH3 NH3 NH3 O O O H H O H H H H O O H OH H O H OH H2O O O HO Asp 102 O Asp 102 Asp 102 HN R N R HN CH2 R

Lys 167 Lys 167 Lys 167 Figure 1.8 Catalytic machanism of DERA

While other aldolases only accept certain ketones as aldol donor, DERA is the only known aldolase that accepts aldehydes and ketones as donor substrates. The accepted aldol donors are: acetaldehyde, propanal, acetone and fluoroacetone21. A wide variety of aldol acceptors are possible, albeit with relatively low reaction rates for non-natural acceptors21. With these characteristics DERA has a wide and interesting product scope, like β-hydroxy aldehydes, deoxy sugars, azido sugars or trideoxyhexoses22,23. Recent reviews give a good overview about the possibilities.18,24

Precursors for statins The most investigated enantiopure β-hydroxy carbonyl compounds are without doubt the side chain intermediates of statins. During the 1950s and 1960s, it became apparent that elevated concentrations of cholesterol in the bloodstream were a major risk factor for the development of coronary heart disease. This initiated the search for drugs that could reduce plasma cholesterol. One logical possibility to reduce plasma cholesterol was to reduce cholesterol biosynthesis. To achieve this, the rate-limiting enzyme in the

12 Chapter 1

cholesterol biosynthetic pathway, 3-hydroxy-3-methyl-glutaryl-CoA (HMG- CoA) reductase, was a good target for investigation.25

Compactin and lovastatin, natural products with a powerful inhibitory effect on HMG-CoA reductase, were discovered in the 1970s. These substances were taken into clinical development as potential drugs for lowering LDL cholesterol. The use of compactin was discarded along the way due to serious side effects but large-scale trials confirmed the effectiveness of lovastatin and the observed tolerability of this statin continued to be excellent. Lovastatin was approved by the US FDA in 1987. Several other HMG-CoA reductase inhibitors, now widely known as statins, subsequently became available for prescription.25 Statin medicines that are now on the pharmacy shelves in the U.S. are shown in Figure 1.9.

OH OH O OH OH O NH

N O- OH 2+ O Ca N O O S N N F 2 Atorvastatin Rosuvastatin (Lipitor®) (Crestor®)

HO O HO O HO O

O O O O O O

O O O H H H

HO Lovastatin Simvastatin Pravastatin (Mevacor®) (Zocor®) (Pravachol®)

F F

OH OH O OH OH O

ONa – ++ O )2Ca

N N

Fluvastatin Pitavastatin (Lescol®) (Lipalo®) Figure 1.9 Statins with 3,5-dihydroxy hexanoic acid based side chain26,27

Statins dominate the over US$ 20 billion market of cholesterol lowering drugs, which is the largest in the pharmaceutical sector. Atorvastatin and simvastatin are the two best-selling drugs in the world. Revenues in 2003 of US$ 9.2

13 Introduction

billion (2002: US$ 8.0 billion) and US$ 6.1 billion (2002: US$ 6.2 billion) are recorded for atorvastatin and simvastatin, respectively. Simvastatin, pravastatin, and lovastatin are produced by fermentation or partial synthesis. Atorvastatin, fluvastatin and rosuvastatin are fully synthetic statins that are rapidly increasing in market value. Pitavastatin is the next synthetic statin, which is in phase 3 of its clinical trials.27

The statins have in common that they have the same homochiral side chain which has the form of a 3,5-dihydroxy acid. Variations are in the hetero aromatic or cyclic residue.27 Since the synthetic statins have such an extremely high market value and require high chemical and stereochemical purity (> 99.5% ee, > 99% de) competition between different research and development groups to find methods to produce the homochiral 3,5-dihydroxy acid side chain precursor, is fierce.27 This makes that these molecules are the most important homochiral β-hydroxy carbonyl compounds regarding annual sales. Hence various chemo-enzymatic routes for their synthesis utilising the three major classes of synthetically useful enzymes (e.g. hydrolases, oxidoreductases and lyases) have been developed (Figure 1.10).

O O ADH OH O X X OEt OEt 1. ADH OH O O HO 2. Lipase OtBu OH OH O OR' OR O O O R'O ADH OEt O OR'' O O OH O Lipase EtO OEt HO OEt

OH OH O Nitrilase NC CN NC OH

O O 1. KRED OH O Cl NC OEt 2. HHDH OEt X O O Aldolase O OH X + 2 H CH3 H OH Figure 1.10 Biocatalytic transformations in the synthesis route towards (3R, 5S)- dihydroxyhexanoates27

14 Chapter 1

Directed evolution Enzymes in general catalyse their native reaction with very high specificity and enatioselectivity. As a result of natural evolution they are optimized to meet the cell’s demands to perform a specific biological role efficiently. High levels of sophistication have been established over several millions of years of evolution for this specific role. However, the enzyme’s activity, selectivity and stability often do not meet the synthetic chemist’s needs in non-natural reactions and environments. Therefore, there is a great need to adapt and alter the enzyme to meet these demands. Several rational methods are developed, such as site directed mutagenesis28, which targets a specific amino acid residue and exchanges it with another and saturation mutagenesis, where one target amino acid is exchanged by all other natural amino acids. These methods need a high level of knowledge about structure, mechanism, function, etc. Following the principle of natural evolution, that generates large numbers of variants and selects the “fittest” candidate, a lab scale mimic, called directed evolution, has been developed29,30. Directed evolution is a powerful method to adapt enzymes to the needs of the synthetic chemist without the requirement of detailed knowledge about its crystal structure or the catalytic mechanism.31,32 Directed evolution is an iterative process in which generation of diversity is followed by screening of the resulting library for a desired function and using the positive hits in the next cycle (Figure 1.11)

Diversity can be generated in non-recombinative ways33, like error prone PCR, site saturation mutagenesis, cassette mutagenesis or in recombinative ways33 like DNA-shuffling29,34, staggered extension process35, random-priming recombination36, heteroduplex recombination.37 Regardless of which strategy is chosen the resulting gene-variants are ligated to form plasmids, which are used to transform the host cells. These are grown and induced to start the production of the required protein. After a selection procedure, in which the best candidates for the target-purpose are selected, this procedure is repeated until the target performance is reached.

15 Introduction

Non-recombinative method Recombinative method wild type gene -ePCR - DNA-shuffling™ - site saturation mutagenesis -StEP - casete mutagenesis - random-priming recombination -hetroduples recombination -ITCHY

Generation of Diversity

Expression of variants

Selection

Repeat cycle starting with improved variant gene

Figure 1.11 General scheme for directed evolution using non-recombinative methods (left) or recombinative methods (right)

Scope of this thesis The work described in this thesis deals with biocatalytic routes to produce homochiral β-hydroxy carbonyl compounds. This type of building block is important in the production of a wide variety of fine chemicals. Three different biocatalytic routes were investigated: a lipase catalysed resolution of the secondary alcohols, the ketoreductase catalysed reduction of the corresponding ketones and a DERA catalysed aldol reaction. Directed evolution aimed at altering the enzyme’s specificity and improving its characteristics to meet the organic synthetic needs takes a central role in the routes that utilize ketoreductases and the aldolase DERA. Analysis of enzyme structure relationships for improved variants produced by directed evolution

16 Chapter 1

can provide a basis for understanding mechanistic details and for further improvemend in substrate specificity, activity, selectivity, etc.

References

1 Sheldon, R.A. Chirotechnology: Industrial synthesis of optically active compounds, 1993, Marcel Dekker Inc., New York 2 Panke, S.; Held, M.; Wubbolts, M. Curr. Opin. Biotechnol. 2004, 15, 272-279 3 Straathof, A.J.J.; Jongejan, J.A. Enzym. Microb. Technol. 1997, 21, 559-571 4 Pellissier, H. Tetrahedron 2003, 59, 8291-8327 5 Schmid, R.D.; Verger, R. Angew. Chem. Int. Ed. 1998, 37, 1608-1633 6 Ema, T. Curr. Org. Chem. 2004, 8, 1009-1025 7 Pleiss, J.; Fisher, M.; Schmid, R.D. Chem. Phys. Lipids 1998, 93, 67-80 8 Varughesem K.I.; Xuong, N.H.; Kiefer, P.M.; Matthews, D.A.; Whiteley, J.M. Proc. Natl. Acad. Sci 1994, 91, 5582-5586 9 Jörnvall, H.; Persson, B.; Krook, M.; Penning, T.M. Biochemistry 1995, 34, 6003-6013 10 Bradshaw, C.W.; Fu, H.; Shen, G.J.; Wong, C.H. J. Org. Chem. 1992, 57, 1526-1532 11 Prelog, V. Pure Appl. Chem. 1964, 9, 119 12 Tanaka, N.; Nonaka, T.; Nakamura, K.T.; Hara, A. Curr. Org. Chem. 2001, 5, 89-111 13 Tanaka, N.; Nonaka, T.; Nakanishi, M.; Deyashiki, Y.; Hara, A.; Mitsui, Y. Structure 1996, 4, 33-45 14 Tsigelny, I.; Baker, M.E. Biochem. Biophys. Res. Commun. 1996, 226, 118-127 15 Nakanishi, M.; Kakumoto, M.; Matsuura, K.; Deyashiki, Y.; Tanaka, N.; Nonaka, T.; Mitsui, Y.; Hara, A. J. Biochem. 1996, 120, 527-263 16 Nakanishi, M.; Matsuura, K.; kaibe, H.; Tanaka, N.; Nonaka, T.; Mitsui, Y.; Hara, A. J. Biol. Chem. 1997, 272, 2218-2222 17 Mazza, C.; Breton, R.; Housset, D.; Fontecilla-Camps, J.C. J. Biol. Chem. 1998, 273, 8145-8152 18 Machajewski, T.D. Angew. Chem. Int. Ed. 2000, 39, 1352-1374 19 Heine, A.; DeSantis, G.; Luz, J.G.; Mitchell, M.; Wong, C.H. Wilson, I.A. Science 2001, 294, 369-374 20 Liu, J.J.; DeSantis, G.; Wong, C.H. Can. J. Chem./Rev. Can. Chim. 2002, 80, 643-645 21 Barbas, C.F.; Wang, Y.F.; Wong, C.H. J. Am. Chem. Soc., 1990, 112, 2013-2014 22 Gijsen, H.J.M.; Wong, C.H. J. Am. Chem. Soc. 1995, 117, 7585-7591 23 Liu, J.J.; Hsu, C.C.; Wong, C.H. Tetrahedron Lett. 2004, 45, 2439-2441 24 Gijsen, H.J.M.; Qiao, L.; Fitz, W.; Wong, C.H. Chem. Rev. 1996, 96, 443-473

17 Introduction

25 Tolbert, J.A. Nat. Rev. Drug Discov. 2003, 2, 517-526 26 Öhrlein, R.; Baisch, G. Adv. Synth. Catal. 2003, 345, 713-715

27 Müller, M. Angew. Chem. Int. Ed. 2005, 44, 362-365 28 Trower, M.K. In vitro mutagenesis protocols 1996, Humana Press., New jersey 29 Stemmer, W.P.C. Nature 1994, 370, 389-391 30 Arnold, F.H. Chem. Eng. Sci. 1996, 51, 5091-5102 31 Jaeger, K.; Reetz, M.T. Curr. Opin. Chem. Biol. 2000, 4, 68-73 32 Petrounia, I.P.; Arnold, F.H. Curr. Opin. Biotechnol. 2000, 11, 325-330 33 Jaeger, K.E.; Eggert, T.; Eipper, A.; Reetz, M.T. Appl. Microbiol. Biotechnol. 2001, 55, 519-530 34 Stemmer, W.P.C. Proc. Natl. Acad. Sci.1994, 91, 10747-10751 35 Zhao, H.; Giver, L.; Shao, H.; Affholter, J.A.; Arnold, F.H. Nat. Biotechnol. 1998, 16, 258-261 36 Shao, Z.; Zhao, H.; Giver, L.; Arnold, F.H. Nucleic Acids Res. 1998, 26, 681-683 37 Volkov, A.A.; Shao, Z.; Arnold, F.H. Nucleic Acids Res. 1999, 27, 18

Part I Lipase catalysed Resolution of Aldol Adducts

Lipase catalysed Resolution of Nitro aldol Adducts

Abstract The kinetic resolution of a range of 1-nitro-2-alkanols by lipase-catalysed esterification using various lipases and succinic anhydride as acyl donor was studied. E values up to 100 were obtained with Novozym 435 in the resolution of 1-nitro-2-pentanol with succinic anhydride in TBME. Acylation with succinic anhydride was much more enantioselective than with vinyl acetate .

The content of this chapter has been published in: M.J. Sorgedrager, R. Malpique, F. van Rantwijk, R.A. Sheldon, Tetrahedron: Asym., 2004, 15, 1295-1299.

Lipase catalysed Resolution of Nitro-aldol Adducts

Introduction The aldol reaction is one of the most important methods for C-C bond formation.1 In the related nitro-aldol reaction, the Henry reaction, the coupling of a nitro alkane to an aldehyde or ketone results in the corresponding β-nitro alcohol. Such chiral nitro alcohols are interesting building blocks in organic synthesis. They can be converted into the chiral β-hydroxy amine by reduction of the nitro group. Alternatively, further carbon-carbon bond formation on the α-carbon of the nitro group will lead to a wide variety of other useful intermediates.2,3 Control of the stereochemistry is of importance for many synthetic purposes, especially for pharmaceutical and agricultural applications. The products of Henry reactions are secondary alcohols. Hence, they are suitable substrates for resolution by lipase-catalysed enantioselective acylation (Figure 2.1). The lipase-mediated kinetic resolution of secondary alcohols has been widely studied over the past 20 years4 and has become a common synthetic and industrial methodology for producing chiral compounds as pure enantiomers.5,6

OH NO2 Base + NO O H3C 2

1c 2c O

OH Lipase OH O R NO2 NO2 + NO2 Acyl donor 2c (R)-2c (S)-3c

R = CH3 C2H4COOH Figure 2.1 Nitro-aldol reaction of 1-nitro-2-pentanol followed by lipase-catalysed resolution

The choice of the acyl donor in such kinetic resolutions requires careful consideration because the transesterification equilibrium should be entirely on the side of the product.4 The often-used vinyl esters, which react irreversibly because the liberated vinyl alcohol isomerises into acetaldehyde, satisfy this requirement.4

22 Chapter 2

The resolution of some β-nitro alcohols via transesterification with vinyl acetate as the acyl donor has been reported previously,3,7,8 but the separation of the enantiomerically enriched ester and alcohol is often laborious. Acylation with a cyclic anhydride, which also reacts irreversibly, results in a half ester that can be readily extracted from the reaction mixture. This potential benefit for reaction work-up procedures has been reported for the resolution of several secondary alcohols.9,10,11,12,13,14

We have studied the applicability of succinic anhydride as acyl donor in the lipase-mediated resolution of a number of alkyl- and phenylalkyl substituted nitro alcohols. The effects of the lipase and the reaction medium will be discussed.

Results and discussion

Resolution of nitro alcohols: lipases, acyl donors, solvents

The acylation of 1-nitro-2-pentanol (2c, Figure 2.1), which we selected as a suitable test reactant, was performed in the presence of a range of microbial lipases (Table 2.1). The reaction was fast and enantioselective when performed in the presence of Novozym 435, an immobilized preparation of Candida antarctica lipase B (CaLB). Two cross-linked preparations of CaLB, the cross-linked enzyme aggregate (CLEA) and the cross-linked enzyme crystal (ChiroCLEC CaB) were much less active, although comparable amounts of units were used,∗ and the enantioselectivity was low. In these hydrophilic particles, with a high density of active protein, the reaction is probably no longer kinetically controlled but limited by the rate of diffusion in and out of the particles. This could lead to lower conversion and enantioselectivity.

∗ 1 unit will liberate 1 µmol of acetic acid per minute from triacetin

23 Lipase catalysed Resolution of Nitro-aldol Adducts

Table 2.1 The performance of different lipases in the resolution of 1-nitro-2-pentanola

Enzyme Conversion ees eep E (%, 24 h) (%) (%) Novozym 435 54 92 93 82 (S) CaLB CLEC 11 8 67 5 (S) CaLB CLEA 7 0 5 1 (S) CaLA SP 526 10 11 82 12 (R) BcL (Amano 1) 45 34 46 4 (R) BcL CLEC 54 59 37 4 (R) CrL 3 2 14 1 (R) CrL CLEC 8 1 3 1 (S) ClL nra - - - RmL SP 524 nr - - - CaLB: Candida anctarctica lipase B; CaLA: Candida antarctica lipase A; RmL: Rhizomucor miehei lipase; CrL: Candida rugosa lipase; ClL: Candida lipolytica lipase, BcL: Burkholderia cepacia lipase Reaction conditions: 1 ml diisopropyl ether, 100 mM 1-nitro-2-pentanol, 100 mM succinic anhydride, 25 mg enzyme or 5 mg CLEC or CLEA. a No reaction

In the presence of CaLB, the (S)-enantiomer was preferentially converted, as predicted by the Kazlauskas rule.15 Two preparations of Burkholderia cepacia lipase (Amano 1 and ChiroCLEC PC), in contrast, showed a slight preference for the (R)-enantiomer of the reactant. The activity of the other lipases tested was much lower and, apart from Candida antarctica lipase A (CaLA), the enantioselectivity was poor.

In view of the above results, we selected Novozym 435 for our further investigations. We next compared the performance of the acyl donors vinyl acetate and succinic anhydride in the resolution of 2c. The reactions were performed in a range of solvents because it is known that the outcome often critically depends on the nature of the solvent. We found (Table 2.2) that the acylation with succinic anhydride took place with high enantioselectivity in tert- butyl methyl ether (TBME) or diisopropyl ether (DIPE) as reaction medium. When the reaction was performed in 1,2-dimethoxyethane (DME), in contrast, it proceeded sluggishly and with low E.

24 Chapter 2

Nitromethane and acetonitrile likewise proved to be unsuitable for this reaction. No adequate resolution of 2c upon acylation with vinyl acetate could be obtained in any solvent.

Table 2.2 Resolution of 1-nitro-2-pentanol in the presence of CaLB; effect of the acyl donor and the solventa

Solvent Succinic anhydride Vinyl acetate

conv eeS eeP E conv eeS eeP E (%) (%) (%) (%) (%) (%)

MeNO2 <1 0 1 - 3 1 45 3 ACN 8 2 100 2 nda nd nd nd DME 4 3 76 7 37 38 52 5 TBME 42 70 95 100 46 41 35 3 DIPE 54 92 93 82 37 15 25 2 Reaction conditions: 1 ml diisopropyl ether, 100 mM 1-nitro-2-pentanol, 100 mM acyl donor, 25 mg Novozym 435 a Not determined

Substrate specificity and enantiodiscrimination

A range of nitro alcohols, resulting from aldol reaction of aliphatic (1a-c) and aromatic (1d-f) aldehydes with nitromethane, was studied in the lipase- mediated resolution with succinic anhydride. The selectivity of Novozym 435 for these substrates follows the Kazlauskas rule15 (Figure 2.2). H OH

M L

Figure 2.2 Steric model for the preferentially converted enantiomer of secondary alcohols by a lipase

This model intrinsically implies that the enantiodiscrimination of secondary alcohols by a lipase is dominated by steric interactions. The presence of a nitro group, which also interacts electrostatically, might influence the enantiodiscrimination.

25 Lipase catalysed Resolution of Nitro-aldol Adducts

Such effects would become visible when both enantiotopic groups are similar in size. We found, however, that the enantiomer discrimination of 2b (R=Et) was much lower that that of 2a or 2c (R=Me or Pr, respectively, see Table 3). We conclude from these results that enantiodiscrimination of these compounds by CaLB is dominated by steric interactions and that electrostatic effects are at best minor.

The longer alkyl- (2c) and the phenyl alkyl substituted (2d-f) substrates all show (S)-selectivity in accordance with the predictive model shown in Figure 2.2. Surprisingly, there was no reaction observed with 3-phenyl-1-nitro-2- propanol (2e) when succinic anhydride is used as acyl donor. We have no plausible explanation for this observation. During the resolution of the aromatic nitro alcohols some spontaneous elimination of carboxylic acid into the corresponding nitroalkene was observed. Elimination of acetic acid occurred more readily than that of succinic acid. No detectable alkene formation from the aliphatic nitroalcohols and their esters was observed under the reaction conditions. Overall, much higher enantioselectivities were found in the resolution of 1- nitro-2-alkanols when succinic anhydride was used as the donor compared with vinyl acetate.

Conclusion Succinic anhydride, besides having potential benefits for reaction workup was shown to be an efficient acyl donor for resolutions of β-nitro alcohols. Much higher E values can be achieved compared to other common acyl donors such as vinyl acetate. The best results, with regard to both rate and enantioselectivity, were observed in tert-butyl methyl ether and diisopropyl ether.

26 Chapter 2

O OH Lipase OH OH O NO2 + NO NO + O R H C 2 Succinic 2 3 R R NO O anhydride R 2 O O OH OH O O

NO2 O NO2 O

(R)-3a (S)-3d

O O OH OH O O

NO2 O NO2 O (R)-3b (S)-3e

O O OH OH O O

NO2 O NO2 O

(S)-3c (S)-3f Figure 2.3 Preferentially formed enantiomers of 1-nitro-2-alkanols by Novozym 435

Table 2.3 Resolution of 1-nitro-2-alkanols with succinic anhydride and vinyl acetate in diisopropyl ethera

Succinic anhydride Vinyl acetate

Conv ees eep Conv ees eep R E E (%, 24 h) (%) (%) (%, 24h) (%) (%)

a CH3 39 57 67 28 (R) 30 2 5 1 (R) a c c C2H5 47 44 43 4 (R) 72 1 nd 1 (R)

C3H7 54 92 93 82 (S) 37 15 25 2 (S) d d c C6H5 4 3 75 7 (S) 13 13 100 20 (S) d b c (C6H5)CH2 nr - - - 53 34 nd 2 (S) d c (C6H5)C2H4 42 71 97 96 (S) 56 21 100 2 (S) Reaction conditions: 1 ml diisopropyl ether, 100 mM substrate, 100 mM succinic anhydride, 25 mg Novozym 435 a Reaction finished after 5h c Not determined c Calculated only with ees. d Substrate and product degradation to the corresponding alkene.

27 Lipase catalysed Resolution of Nitro-aldol Adducts

Experimental part- Material and methods

Instruments and materials

HPLC analyses were performed on a Chiralcel OD column with a flow of 0.6 ml/min and an eluent consisting of hexane-isopropyl alcohol-trifluoroacetic acid (95:5:0.1) for the aliphatic nitro alkanols and hexane-isopropyl alcohol- trifluoroacetic acid (80:20:0.1) for the aromatic nitro alkanols. Detection was performed with a Waters 486 UV detector at 215 nm. All chemicals were of analytical purity and obtained from Sigma-Aldrich. The lipase from Candida rugosa was obtained from Sigma-Aldrich. Candida lypolitica lipase was bought from Fluka. Novozym 435 (Candida antarctica lipase B on Lewatit E), SP524 (Rhizomucor miehei lipase), SP526 (Candida antarctica lipase A) were kindly donated by Novozymes. Cross-linked enzyme crystals (CLECs) were donated by Altus Biologics; CaLB CLEA was donated by CLEA Technologies. The 1-nitro-2-alkanols were synthesized as described below.

Methods

General synthesis of aliphatic nitro alcohols 2a-c

Aldehyde 1a-c (100 mmol) and KOH (1 ml, 1M solution) were added to nitromethane (10 ml) at 0-5 °C. After 1 h the mixture was brought to room temperature and left for another 2-4 h. The reactions were followed by TLC using ether-petroleum ether 3:2 as the eluent. Dichloromethane was added and this mixture was washed successively with 5% aqueous HCl and saturated aqueous NaHCO3. The organic phase was dried with MgSO4 and the solvent was evaporated. The crude product was purified by distillation, giving a colourless liquid.

28 Chapter 2

1-Nitro-2-propanol 2a 1 Isolated yield 46%. 300 MHz H NMR (CDCl3) δ (ppm): 1.3 (3H, d), 3.2 (1H, 13 s), 4.4(2H, m), 4.5 (1H, m). 75 MHz C NMR (CDCl3) δ (ppm): 19.8, 65.0, 81.6

1-Nitro-2-butanol 2b 13 Isolated yield 48%. 75 MHz C NMR (CDCl3) δ (ppm): 5.6, 26.9, 70.1, 80.5

1-Nitro-2-pentanol 2c The reaction was carried out on 1M scale. Isolated yield 69%. 300 MHz 1H

NMR (CDCl3) δ (ppm): 1.0 (3H, t), 1.6 (2H, m), 3.0(1H, s), 4.3 (1H, m), 4.5 (2H, m).

General synthesis of aromatic nitro alcohols 2d-f

Aldehyde 1d-f (100 mmol) and KOH (1 ml, 1M solution) were added to nitromethane (20 ml) at 0-5°C. After 1 h the mixture was brought to room temperature and left stirring for another 2-4 h. The reactions were followed by TLC using ether-petroleum ether 1:1 as the eluent. The pH was adjusted to 7 and the excess of nitromethane was evaporated. Ether was added and washed successively with acidic water, saturated NaHCO3 solution and water.

The organic phase was dried with MgSO4 and the solvent evaporated to yield the crude product.

2-Phenyl-1-nitro-2-ethanol 2d The crude product was purified by distillation giving a yellow liquid. Isolated 1 yield 57%. 300 MHz H NMR (CDCl3) δ (ppm): 3.2 (1H, s), 4.5 (2H, m), 5.4 13 (1H, q), 7.3 (5H, m). 75 MHz C NMR (CDCl3) δ (ppm): 70.9, 81.1, 125.9, 128.9, 129, 138.2.

3-Phenyl-1-nitro-2-propanol 2e The crude product was purified by distillation giving a yellow liquid. Isolated 1 yield 23%. 300 MHz H NMR (CDCl3) δ (ppm): 2.8 (2H, m), 4.4 (2H, m), 4.5

29 Lipase catalysed Resolution of Nitro-aldol Adducts

13 (1H, q), 7.3 (5H, m). 75 MHz C NMR (CDCl3) δ (ppm): 40.4, 69.5, 79.7, 127.3, 128.9, 129.4, 135.9.

4-Phenyl-1-nitro-2-butanol 2f The product was recrystallised from diisopropyl ether resulting in a white solid. 1 Isolated yield 56%. 300 MHz H NMR (CDCl3) δ (ppm): 1.8 (2H, m), 2.6 (1H, d), 2.85 (2H, m), 4.35 (1H, m), 4.45 (2H, m), 7.3 (5H, m). 75 MHz 13C NMR

(CDCl3) δ (ppm): 31.3, 35.1, 67.7, 80.5, 126.3, 128.4, 128.6, 140.6.

Lipase-catalysed resolutions

The lipase-catalysed acylation reactions were performed in 1 ml solvent at room temperature, with 100 mM nitro alcohol, an equivalent amount of acyl donor and 25 mg of the various enzyme preparations. When CLECs or CLEAs were used 5 mg was added instead to ensure the use of the same amount of active protein. Except for the enzyme screening, all reactions were carried out with Novozym 435. Trimethoxybenzene (2.5 g/l) was used as the internal standard. The reactions were monitored by chiral HPLC.

References

1 Machajewski, D.; Wong, C.H. Angew. Chem. Int. Ed. Engl. 2000, 39, 1352-1374 2 Itayama, T. Tetrahedron 1996, 52, 6139-6148 3 Nakamura, K.; Kitayama, T.; Inoue, Y.; Ono, A. Tetrahedron 1990, 46, 7471-7481 4 Bornscheuer, U.T.; Kazlauskas, R.J. Hydrolases in Organic Synthesis; Wiley-VCH: Weinheim, 1999 5 Jaeger, K.; Eggert, T. Curr. Opin. Biotechnol 2002, 13, 390-397 6 Schmid, A.; Hollman, F.; Park, J.B.; Bühler, B. Curr. Opin. Biotechnol. 2002, 13, 359- 366 7 Kitayama, T.; Rokutanzono, T.; Nagao, R., et al. J. Mol. Catal. B: Enzym. 1999, 7, 291- 297

30 Chapter 2

8 Morgan, B. et al. Enzymes in non-aqueous solvents, Vulfson, E.N.; Halling, P.J.; Holland, H.L., eds, Humana press: New York, 2001, p. 444-451 9 Gutman, A.L.; Brenner, D.; Boltanski, A. Tetrahedron: Asymmetry, 1993, 4, 839-844 10 Hyatt, J.A.; Skelton, C. Tetrahedron: Asymmetry, 1997, 8(4), 523-526 11 Fiaud, J.; Gil, R.; Legros, J.; Aribi-Zouioueche, L.; König, W.A. Tetrahedron Lett. 1992, 33, 6967-6970 12 Nakamura, K.; Takenaka, K.; Ohno, A. Tetrahedron: Asymmetry 1998, 9, 4429-4439 13 Terao, Y.; Tsuij, K.; Murata, M.; Achiwa, K.; Nishio, T.; Watanabe, N.; Seto, K. Chem. Pharm. Bull. 1989, 37, 1653-1655 14 Patel, R.N.; Banerjee, A.; Nanduri, V.; Goswami, A.; Comezoglu, F.T. J.Am.Oil Chem. Soc. 2000, 77, 1015-1019 15 Kazlauskas, R.J.; Weissfloch, A.N.E.; Rappaport, A.T.; Cuccie, L.A. J. Org. Chem. 1991, 56, 2656

Optimizing the Deracemisation of Nitro aldol Adducts

Abstract Various techniques have been studied to increase the yield of the lipase- catalysed resolution of 1-nitro-2-alkanols beyond the theoretical 50 %. In-situ racemisation via a prochiral precursor using the reversible Henry reaction could not be applied due to the instability of the produced ester under these conditions. The resolution carried out in a subtractive way, converting the ester ex-situ via the corresponding alkene to the racemic alcohol, exhibited a slightly increased yield but proved not to be efficient enough to be attractive.

Optimizing the Deracemisation of Nitro-aldol Adducts

Introduction In Chapter 2, a viable resolution method for β-nitro alcohols has been demonstrated, employing esterification with succinic anhydride catalysed by Novozym 435.1 The acyl donor was found to have a profound influence on the success of this resolution (Figure 3.1). A 50 times higher E value could be achieved when succinic anhydride (E 96) was used as acyl donor instead of the commonly employed vinyl acetate (E 2).

OH OH OH O Novozyme 435 NO2 NO NO2 O 2 + Succinic anhydride

Figure 3.1 Lipase catalysed resolution of 4-phenyl-1-nitro-2-butanol with succinic anhydride

To overcome the limitation of a 50% maximum yield typical for a classical kinetic resolution the possibility of a dynamic kinetic resolution (DKR) for β- nitro alcohols was investigated. A dynamic kinetic resolution combines the enantioselective resolution with an in-situ racemisation of the slow reacting enantiomer, making full conversion of a racemic substrate into one single enantiomer possible.2 Various methods have been applied and studied to achieve this goal. A recent promising method for secondary alcohols combines an enzymatic resolution with a ruthenium-complex catalysed racemisation of the remaining enantiomer (Figure 1.2).3,4,5 Another approach yielding the same result is to keep the starting material racemic by using a fast reversible reaction via a prochiral precursor. The latter approach has been effective in the synthesis of enantiopure cyanohydrins, by using the equilibrium of the hydrocyanation of carbonyl compounds.6,7,8 Aldol reactions can be used in a similar fashion when the reversibility of the aldol reaction is used to racemise the alcohol (Figure 3.2).

The previously described methods lose their applicability if the produced ester is unstable and eliminates carboxylic acid easily to yield the corresponding alkene. Under most conditions needed for racemisation of the alcohol in the preceding methods, the ester formed during resolution of β-nitro-alcohols is

34 Chapter 3

unstable. Therefore it is impossible to isolate it and an alternative procedure is required. O OH Novozym 435 OH O NO2 Succinic NO2 O anhydride (S) O Base + OH

NO2 Me-NO2 (R) Figure 3.2 A dynamic kinetic resolution using the equilibrium of the aldol reaction

A resolution can be performed in subtractive manner, isolating the fast reacting enantiomer as the ester after resolution or in an attractive manner, isolating the slow reacting enantiomer of the substrate.9 The latter method is more preferable because it consists of fewer steps to gain the pure homochiral alcohol. Besides it is easier to gain high ee even with moderate enzymatic selectivity by controling how far the reaction is allowed to proceed. The previous described in-situ racemisation methods for enhancing the efficiency of a resolution need to be combined with a subtractive resolution due to the fact that the alcohol is used for racemisation.

Alternatively the resolution can be performed in an attractive manner. By combining this attractive resolution with a racemisation method for the unstable ester the yield of the deracemisation of β-nitro alcohols can be increased as well. Decarboxylation of the ester will result in the corresponding alkene. This occurs partly spontaneously and can be promoted to full conversion by a slightly basic environment. The alkene can be converted into the racemic alcohol by hydration of the double bond (Figure 3.3). Such an approach would overcome the problem of product instability because the remaining alcohol is isolated instead of the ester and it would demonstrate an alternative process for deracemisation of nitro alcohols where a one pot DKR is not possible.

35 Optimizing the Deracemisation of Nitro-aldol Adducts

O OH OH Novozym 435 OH O NO2 NO2 + NO2 O Succinic anhydride (R) (S)

H+ Base

H2O

NO2 Figure 3.3 Reaction scheme for the subtractive resolution and combined racemisation of β- nitro alcohols

In this chapter the feasibility of the in situ and ex situ racemisation methods to improve the deracemisation of β-nitro alcohols is reported.

Results and discussion

Dynamic kinetic resolution and instability of the ester

The kinetic resolution of the test substrate 4-phenyl-1nitro-2-butanol has been studied and was reported previously in Chapter 2. All trials for racemising the slow-reacting enantiomer of the alcohol, by using the equilibrium of the aldol reaction, were unsuccessful. Catalytic amounts of Amberlyst A21, NaOH, pyridine or triethylamine were used as aldol reaction catalysts for the purpose of racemisation during the resolution reaction. Instead of gaining better yields in ester formation and a racemic alcohol throughout the full course of the resolution, the main product was the corresponding alkene. The instability of the ester can be ascribed to the acidic α-proton next to the nitro-group, which enables elimination via a cyclic mechanism (Figure 3.4). When elimination occurs according to this mechanism it will result in an alkene with cis configuration. On the other hand, dissociation via standard E1 elimination will favour the trans configuration. To verify the hypothesis that instability of the ester can be ascribed to the acidic α-proton next to the nitro-group, the stereochemistry around the double bond has been elucidated on the basis of 1 1 1 its H NMR spectrum and H NOESY NMR. Although in the H NMR ja-b = 13.5 Hz is not typically indicating cis- nor trans-coupling, strong cross over was

36 Chapter 3

found in the NOESY experiment for Ha and Hb (Figure 3.4). This is strong evidence for a cis configuration of the alkene and supports the proposed mechanism of elimination.

R R O H H H O O e d b H H OH a H c NO2 NO2 He

R: Restgroup of the acyl donor Figure 3.4 Dissociation mechanism of the esters of β-nitro alcohols

Recycling of the unstable ester

To overcome the problem of product instability the resolution is performed subtractively, isolating the slow reacting enantiomer of the alcohol. Following the reaction scheme shown in Figure 3.3 the ester that is formed from the fast reacting enantiomer can than easily be converted in the cis-alkene with Amberlyst A21. The main challenge is the subsequent hydration of this alkene to yield the racemic nitro alcohol. A standard procedure to achieve water addition to a double bond is by using concentrated sulfuric acid. The alkene reacts with concentrated sulfuric acid to give an alkyl hydrogen sulfate. The alkyl hydrogen sulfate can be converted by SN1 substitution to an alcohol by boiling in water. Due to the electron withdrawing effect of the nitro group present in the alkene, and therefore a more stable carbocation on the β- position than on the α-position during the formation of the alkyl hydrogen sulfate, this will lead to the correct position of the alcohol.

Table 3.1 Hydration of 4-phenyl-1-nitrobut-1-ene

ConcentrationH2SO4 Conversion 1 h (%) Conversion 2 h (%) 5 M 17 24 6 M 13 20 7 M 6 21 8 M 1 22 Reaction conditions: 30 min at room temperature followed by reflux at 100°C

37 Optimizing the Deracemisation of Nitro-aldol Adducts

Initially (1h of reaction) low concentrations of sulphuric acid seem beneficial, but final conversion after 2h is similar for all conditions (Table 3.1). Due to the hydration being exothermic together with a decrease in number of molecules, the equilibrium of this reaction is thermodynamically not favourable for hydration. Low temperatures would be beneficial, but due to the low reactivity of the alkene more severe reaction conditions are often required.10 Lower and higher temperatures than 100°C (75°C, 130°C) did not show any improvement and conversions lowered drastically into the few percent range. A cycle combining the resolution and the hydration with intermediate alcohol isolation (Figure 3.5) resulted in an isolated total yield of 48 %. Substantial enantiomeric enrichment of the produced alcohol was achieved during crystallisation raising the ee from 70 % to >99 %.

Alcohol 0.61 mmol ee > 99% Ester s Ester 0.45 mmol 0.45 mmol

Alcohol Alcohol eep = 97 % Alcohol eep = 97 % 2.1 mmol 1.22 mmol Alkene 0.61 mmol Alkene for startup Kinetic ee = 71% 0.44 mmol ee = 42% 0.44 mmol Alkene s Precipitation s Resolution Formation = 50% Alcohol ξ = 42% η ξ = 100% 1.28 mmol E = 96 for 2nd round

Recycle nd Alcohol Alkene Alcohol Alkene for 2 round 0.82 mmol 0.12 mmol 0.61 mmol 0.89 mmol ees = 32% (55% loss) Alkene ees = 42% Hydration ξ = 24% Figure 3.5 Integrated deracemisation cycle

Attention has to be paid that the conversion in the kinetic resolution was only 42 % and the alcohol was only partially isolated by precipitation. If the resolution would be performed till completion (50 %) and the separation is done more efficiently the maximum theoretical efficiency would still only be 60 %.

38 Chapter 3

Dynamic kinetic resolution of α-proton deficient substrates

To study the intrinsic possibility to use the reversible Henry reaction for designing a dynamic kinetic resolution as shown in Figure 3.2, substrates without the acidic α-protons where studied. The esters that will be formed during lipase-catalysed resolution will be much more stable because they cannot dissociate according to the mechanism shown in Figure 3.4.

OH O R' Hydrolase NO2 NO R R 2 Acyl donor

R= CH3 Hydrolases Novozym 435 C2H5 Candida antarctica lipase B CLEA C6H5 -C2H4 Candida antarctica lipase A Pseudomonas cepacia lipase Candida rugosa lipase R'= C2H4COOH Acylase 1 (esterase) CH3 Figure 3.6 Resolution experiments with α-proton deficient substrates

Ethyl esters of the three α-proton deficient substrates proved to be much more stable indeed under the conditions needed for the Henry reaction (Figure 3.6). Unfortunately these substrates, of which the α-protons were replaced by methyl groups, were not accepted by any of the lipases that were tested with succinic anhydride and vinylacetate as acyl donor. Also the esterase in Acylase 1 did not show any activity. The presence of a quaternary carbon atom directly next to the chiral centre apposes too much steric constrains for these substrates to be able to bind effectively in the active site.

Conclusion Studies to increase the efficiency of deracemisation of nitro-aldol adducts by lipase-catalysed resolution did not lead to a satisfying success. Using the reversibility of the Henry reaction for in-situ alcohol racemisation during the resolution could not be applied due to the instability of the ester under aldol

39 Optimizing the Deracemisation of Nitro-aldol Adducts

conditions. The more stable α-proton deficient substrates, to prove the principle, were not accepted as substrate by the lipases tested. The yield of alkene hydration, for recycling the dissociated ester towards racemic alcohol is too low to be efficient.

Experimental part: Materials and Methods

Instruments and materials

All chemicals were of analytical purity and obtained from Sigma-Aldrich. Novozym 435 (Candida antarctica lipase B on Lewatit E) was kindly donated by Novozymes. CaLB-CLEA was made available by CLEA Technologies. The nitro alcohols were synthesized via Henry reaction as described below. HPLC analyses were performed on a Chiralcel OD column with a flow of 0.6 ml/min and an eluent consisting of hexane-isopropyl alcohol-trifluoroacetic acid (80:20:0.1). Detection was performed with a Waters 486 UV detector at 215 nm.

Methods

Synthesis of 4-phenyl-1-nitro-2-butanol

3-Phenylpropionaldehyde (1 M), nitromethane (2 M) and KOH (1 ml, 5M solution) were added to 1,2-dimetyhoxyethane (200 ml) at 0-5°C. After one hour the mixture was brought to room temperature and left stirring for another 2-4 hours. The reaction was followed by TLC using ether-petroleum ether 1:1 as the eluent. After the reaction was finished the pH was neutralized and the excess of nitromethane and solvent were evaporated. Ether was added and washed sequentially with acidic water, saturated NaHCO3 solution and water.

The organic phase was dried with MgSO4 and the solvent evaporated to yield the crude product. The product was recrystallised from diisopropyl ether 1 resulting in a white solid. Isolated yield 56%. 300 MHz H NMR (CDCl3) δ

40 Chapter 3

(ppm): 1.8 (2H, m), 2.6 (1H, d), 2.85 (2H, m), 4.35 (1H, m), 4.45 (2H, m), 7.3 13 (5H, m). 75 MHz C NMR (CDCl3) δ (ppm): 31.3, 35.1, 67.7, 80.5, 126.3, 128.4, 128.6, 140.6.

Synthesis of 4-phenyl-1-nitrobut-1-ene

4-phenyl-1-nitro-2-butanol (20g) and acetic anhydride (9.6 ml) were dissolved in 20 ml of ether. Pyridine (9.2 ml) was added and the reaction mixture was stirred for 4 h at room temperature. Ester formation was followed by HPLC and TLC using ether-petroleum ether 1:1 as the eluent. The reaction mixture was washed sequentially with acidic water, saturated NaHCO3 solution and water. The organic phase was dried with MgSO4. The formation of 4-phenyl-1- nitrobut-1-ene was induced by refluxing the resulting solution in ether for 2 h. After evaporation of the solvent the reaction mixture was distilled and the product was collected at 140 °C and 5⋅10-3 mbar. Isolated yield 61 %. 300 1 MHz H NMR (CDCl3) δ (ppm): 2.6 (2H, m), 2.9 (2H, t), 6.9 (1H, m), 7.1-7.3 (5H, m). The phase-sensitive NOESY 1H NMR spectra were obtained with 512 fn1 increments and 2048 data points. 4 free induction decays were collected for each value of fn1. The mixing time used was 0.1 s.

Synthesis of 1-phenyl-4-methyl-4-nitro-3-pentanol

3-Phenylpropionaldehyde (1.3 ml, 10 mmol), 2-nitropropane (1.8 ml, 20 mmol) and Amberlyst A21 in OH- form (1 g) were added to ethanol (100 %, 25 ml) at 0-5°C. After one hour the mixture was brought to room temperature and left stirring for another 20 hours. The reaction was followed by TLC using ether- petroleum ether 1:1 as the eluent. After the reaction was finished the Amberlyst A21 was filtered and washed with ethanol. The organic filtrate was washed sequentially with acidic water, saturated NaHCO3 solution and water.

The organic phase was dried with MgSO4 and the solvent and the excess of 2-nitropropane evaporated to yield the crude product. The product was further purified by destillation and collected at 80 °C and 4.10-2 mbar. Isolated yield 41%.

41 Optimizing the Deracemisation of Nitro-aldol Adducts

Synthesis of 2-methyl-2-nitro-3-pentanol

Propionaldehyde (7.3 ml, 100 mmol), 2-nitropropane (9 ml, 100 mmol) and - Amberlyst A21 in OH form (4 g) were added to dichloromethane (30 ml) at 0- 5°C. After one hour the mixture was brought to room temperature and left stirring for another 20 hours. The reaction was followed by TLC using ether- petroleum ether 1:1 as the eluent. After the reaction was finished the Amberlyst A21 was filtered and washed with dichloromethane. The organic filtrate was washed sequentially with acidic water, saturated NaHCO3 solution and water. The organic phase was dried with MgSO4 and the solvent and the excess of 2-nitropropane evaporated to yield the crude product. The product was further purified by distillation. Isolated yield 10 %.

Synthesis of 2-dimethyl-2-nitro-3-hexanol

Butyraldehyde (4.4 ml, 50 mmol), 2-nitropropane (4.5 ml, 50 mmol) and Amberlist A21 in OH- form (2 g) were added to ethanol (100%, 25 ml) at 0- 5°C. After one hour the mixture was brought to room temperature and left stirring for another 20 hours. The reaction was followed by TLC using ether- petroleum ether 1:1 as the eluent. After the reaction was finished the Amberlyst A21 was filtered and washed with dichloromethane. The organic filtrate was washed sequentially with acidic water, saturated NaHCO3 solution and water. The organic phase was dried with MgSO4 and the solvent and the excess of 2-nitropropane evaporated to yield the crude product. The product was further purified by distillation. Isolated yield 26 %.

Lipase-catalysed resolutions

The lipase-catalysed acylation was performed in diisopropylether (1 ml) as solvent at room temperature, with 100 mM nitro alcohol, an equivalent amount of succinic anhydride and 25 mg of Novozyme 435. Trimethoxybenzene (2.5 g/l) was used as the internal standard. The reactions were monitored by chiral HPLC. For the dimethylated nitro alcohols Novozym 435 (25 mg), Pseudomonas cepacia lipase (PS-C, 25 mg), Candida antarctica Lipase A

42 Chapter 3

(CaLA, 25 mg), Candida rugosa lipase (CR, 25 mg), Candida antarctica Lipase B CLEA (CaLB-CLEA, 5 mg) were screened.

Conversion of the ester into racemic alcohol

- The resulting mixture from the resolution was treated with Amberlist A21 (OH form, 100 mg) for 1 h to completely convert the succinate ester of 4-phenyl-1- nitro-2-butanol into the 4-phenyl-1-nitrobut-1-ene. After separation of the solids diisopropyl ether was evaporated and a sulfuric acid solution (various concentrations, amount sufficient to get nitro alcohol concentration to be 1 M) was added. After incubation for 30 minutes at room temperature the mixture was heated till 80 °C for 2 hours and after cooling, and neutralizing the pH with NaHCO3 extracted with diisopropylether. Analysis of the samples was performed by chiral HPLC.

References

1 Sorgedrager, M.J.; Malpique, R.; van Rantwijk, F.; Sheldon, R.A.Tertraherdron: Asym. 2004, 15, 1295-1299 2 Pellissier, H.; Tetrahedron, 2003, 59, 8291-8327 3 Kim, M.J.; Ahn, Y.; Park, J. Curr. Opin. Biotechnol. 2002, 13, 578-587 4 Huarta, F.F.; Minidis, A.B.E.; Bäckvall, J.E. Chem. Soc. Rev. 2001, 30, 321-331 5 Pàmies, O.; Bäckvall, J.E. Curr. Opin. Biotechnol. 2004, 14, 407-413 6 Inagaki, M.; Hiratake, J.; Nishioka, T; Oda, J. J. Org. Chem. 1992, 57, 5643-5649 7 Li, Y.X.; Straathof, J.J.; Hanefeld, U. Tetrahedron: Asym. 2002, 13, 739-743 8 Paizs, C.; Tosa, M.; Majdik, C.; Tähtinen, P.; Irimie, F.D.; Kanerva, L.T. Tetrahedron: Asym. 2002, 14, 619-627 9 Sheldon, R.A. Chirotechnology 1993, Marcel Dekker Inc., New York, p. 91-92 10 Izymi, Y. Catalysis Today 1997, 33, 371-409

Part II Enantioselective Carbonyl Reduction

Asymmetric Reduction with Candida magnoliae Ketoreductase S1

Abstract The product scope of ketoreductase S1 from Candida magnoliae has been studied throughout 10 generations of directed evolution. The activity and selectivity and changes herein through evolution have been studied for ethyl- 4-chloroacetoacetate and similar β-ketoesters. The activity increased by a factor 16 for the substrate the enzyme was evolved for, while the activities for similar ketoesters remained more or less the same. In nearly all cases the enantioselectivity was very high (>99 % ee) and not affected by directed evolution.

Asymmetric Reduction with Candida magnoliae Ketoreductase S1

Introduction Biocatalytic reduction of ketones is a convenient route towards homochiral secondary alcohols. Especially the synthesis of the enantiopure side chain of the statins (Chapter 1) has gained a remarkable amount of interest over the past years. Various synthetic approaches have been studied (Figure 1.10) of which the biocatalytic reduction of the corresponding β-ketoesters and β,δ- diketoesters is a promising method. This approach has been the target of various research projects in which the reduction of the halogenated ethyl-4- chloroacetoacetate (1a) gained most interest. Besides being interesting for the synthesis of the statins, this compound is also an intermediate in the synthesis of carnitine and 1,4-dihydropyridine type β-blockers.1 Baker’s yeast reductions have been performed with these type of ketoesters yielding the (S)- enantiomer.2 The stereochemistry of these reductions with Saccharomyces cerevisiae could be controlled by the size of the ester group.2,3 However the cheap and readily available baker’s yeast did not always give the right configuration or optical purity. In the case of ethyl-4-chloroacetoacetate (1a) the major product had the desired (S)-configuration but only with 55% ee.2

Over 400 other microorganisms have been screened for their ability to reduce 4-chloroacetoacetate ethyl ester.4 The yeast Candida magnoliae performed best in this reduction. Heat-treated cells produced 90 g/l (S)-ethyl-4-chloro-3- hydroxybutanoate ((S)-2a) with 99% ee in the presence of a cofactor regeneration system based on glucose dehydrogenase (Figure 4.1).5 The enzyme involved in this latter reaction, the NADPH dependent ketoreductase S1, has been isolated, purified and characterised.6 Subsequently it has been overexpressed in E. coli.7 A combined overexpression of the genes encoding for C. magnoliae ketoreductase S1 and the glucose dehydrogenase of Bacillus megaterium has been performed successfully.8 The reported turnover of NADP+ obtained with this combined overexpression system was 21.600 mol/mol. With this system (S)-2a was produced in a fed-batch reaction yielding 208 g/l with 100% ee.

48 Chapter 4

Ketoreductase S1 is originally NADPH dependent. Regeneration systems for this cofactor are less attractive than for NADH, due to the costs of the cofactor and the lower stability of NADPH. There is also a paucity of enzymes that mediate the reduction of NADPH by suitable sacrificial reductants. The popular regeneration system for NADH, based on formate dehydrogenase

(FDH), produces CO2 gas, while in the regeneration system for NADPH gluconate is produced (Figure 4.1). In the latter case a more laborious downstream processing is required. Therefore, attempts have been made to alter the cofactor specificity of ketoreductase S1.9,10 Computer modelling of homology models for S1 have been used to identify amino acid residues in the adenosine binding pocket that could be suitable targets for site-directed mutagenesis. The most efficient mutant in those studies produced 163 g/l (S)- 2a with 99% ee using NADH as cofactor combined with a regeneration system based on formate dehydrogenase.9

FDH CO2 HCO2H

NADH NAD+ O O OH O R R C.m. R R 1 O 3 1 O 3 KRED R2 R2 NADPH NADP+ 1a-d 2a-d

Gluconate Glucose GDH

a: R1=Cl; R2=H; R3=C2H5 c: R1=H; R2=H; R3=C(CH3)3 b: R1=H; R2=H; R3=C2H5 d: R1=H; R2=CH3; R3=C2H5 Figure 4.1 Asymmetric reduction of β-ketoesters with common used cofactor regeneration systems

The relative substrate specificity and enantioselectivity of C. magnoliae ketoreductase S1 was investigated with a number of substituted β-ketoesters 11,12 (Table 4.1). The R1 group has a huge influence on the activity. Comparing the halogenated ethyl-3-oxobutanoates (Table 4.1, entry 1a, 1h, 1i) the activity decreased in the order of decreasing electronegativity (Cl>Br>I). An OH group at the same position (1k) had the same effect on the relative

49 Asymmetric Reduction with Candida magnoliae Ketoreductase S1

activity. The presence of electron-withdrawing substituents at R1 apparently activates the carbonyl group for hydride attack.

The enzyme also seems to have a very pronounced preference for ethyl esters. The activity with the smaller methyl ester of 4-chloroacetoacetate (1f) was only 11% of that of the ethyl ester (1a). The much larger and more hydrophobic octyl ester was also much less reactive (36% relative to 1a) and the bulky tert-butyl ester was reported to be unreactive.11,12

Table 4.1 Enantioselective reduction of various β-ketoesters

R1 R2 R3 Relative ee (%) Ref activity (%) b 1a Cl H C2H5 100 n.d 11

1b H H C2H5 7 >99 R 12

1c H H C(CH3)3 0 n.d 12 a 1d H CH3 C2H5 14 (2S,3R) 19;(2S,3S) 78 12 (2R,3R) 3 ;(2R,3S) <1 a 1e H Cl C2H5 11 (2S,3R) 2 ;(2S,3S) 56 12 (2R,3R) 41;(2R,3S) <1 1f Cl H CH3 11 n.d 11

1g Cl H C8H17 36 n.d 11

1h Br H C2H5 72 >99 S 12

1i I H C2H5 16 >99 S 12

1j N3 H C2H5 0 n.d 12

1k OH H C2H5 80 >99 S 12

1l C6H5CH2O H C2H5 21 21 S 12

1m C2H5 H C2H5 0.5 n.d 12 a Product distribution in %; preferential syn product = (2S,3R); preferential anti product = (2S,3S) b Not determined

The β-ketoester of most interest (1a) has a limited solubility and stability in water and the enzyme does not tolerate high concentrations of both the product and the substrate.1. Hence a biphasic system has been developed for the biocatalytic reduction with a combined cofactor regeneration system, to meet the needs of an industrial procedure. Candida magnoliae ketoreductase

50 Chapter 4

S1 can tolerate ethyl acetate, butyl acetate and diisopropyl ether as co- solvents in a biphasic system.10,13 We have been employing directed evolution using DNA shuffling technology, to increase the activity of the ketoreductase S1 from Candida magnoliae for ethyl-4-chloroacetoacetate (1a).13 In this chapter we report the behaviour of various mutants selected from 10 rounds of shuffling according to their activity with 1a, as well as related β-ketoesters (1b-1d). The influence of the cofactor and of co solvents has been investigated.

Results and discussion The variants that were studied originated from various generations of directed evolution. CmKr1, CmKr2A, CmKr2B, CmKr4, CmKr8, CmKr9, CmKr10 are selected variants of Candida magnoliae ketoreductase S1 from the library of round 1, 2, 4, 8, 9 and 10 of DNA shuffling respectively.

Reduction of β-ketoesters

The first substrate (1a) is the substrate the enzyme is evolved for. A 16-fold increase in activity was observed in 10 rounds of DNA shuffling without any loss of enantioselectivity. Only the variant CmKr2A showed a slight decrease in enantioselectivity (Table 4.2, entry 2). This variant was from a library where it was attempted to incorporate a change of cofactor dependency towards NADH into the development of this enzyme. Although the data shown in Table 4.2 is for NADPH, the same decrease in ee was found using NADH as cofactor. Due to the decreased enantioselectivity this strategy was abandoned and further evolution was continued from the library of round 1 with NADPH as cofactor. The variant CmKr2B is part of the resulting second generation library. The later libraries are based on that one.

The trend in activity improvement that was found with compound 1a was not observed with the other β-ketoesters. Thus, without the chlorine substituent (1b) no upward trend was observed through the generations. The highest

51 Asymmetric Reduction with Candida magnoliae Ketoreductase S1

improvement (5-fold) was found with CmKr2A, which originated from the library that was developed co-targeting NADH dependency.

The activity decreased drastically when replacing the ethyl ester with the more bulky tert-butyl one (1c). Although in previous studies no activity was found with this substrate in the wild type enzyme (Table 4.1), we found low activities in all variants with in general high enantioselectivity (ee >99%). Since at the moment of writing no crystallographic data is available for this enzyme, no structural explanation can be given. We suggest that steric congestion due to the tert-butyl group made it difficult for the ketoester to bind in the active site with the keto group in the right position for hydride transfer.

Table 4.2 Asymmetric reduction of β-ketoesters catalysed by variants from 10 rounds of directed evolution on Candida magnoliae ketoreductase S1

OH O OH O OH O Cl O O O 2a 2b 2c Vini ee Vini ee Vini ee Enzyme µmol⋅min-1⋅ g-1 (%) µmol⋅min-1⋅ g-1 (%) µmol⋅min-1⋅ g-1 (%) CmKr1 481 >99 (S) 43 >99 (R) 92 >99 (R) CmKr2A 1865 98.5 (S) 311 >99 (R) 17 >99 (R) CmKr2B 551 >99 (S) 12 >99 (R) 11 73 (R) CmKr4 1700 >99 (S) 149 >99 (R) 20 98 (R) CmKr8 5397 >99 (S) 229 >99 (R) 53 >99 (R) CmKr9 4615 >99 (S) 94 >99 (R) 24 >99 (R) CmKr10 7763 >99 (S) 214 >99 (R) 41 >99 (R)

Activity measurement conditions: 2 ml Tris-HCl buffer (50 mM, pH 7.6), 0.16 mM NADPH, 10 mM substrate, 0.1 mg enzyme Enantioselectivity measurement conditions: 0.2 ml Tris-HCl buffer (50 mM, pH 7.6), 11 mM NADPH, 10 mM substrate, 0.2 mg enzyme

The stereoselectivity of reduction of the model substrates 1a-1c with C. magnoliae ketoreductase S1 was not as predicted by Prelog’s rule for stereochemistry in biocatalytic reductions (Chapter 1). When applying the Cahn-Ingold-Prelog convention,14 the anti-Prelog reduction with ketoreductase S1 resulted in (S)-configuration for 1a and (R)-configuration for the others.

52 Chapter 4

The mutations that resulted in the very considerable rate improvement in the reduction of 1a did not lead to similar improvements with the related compounds 1b-1d. Neither did these narrow down the substrate specificity, as the activity for the non-chlorinated substrates (1b-1d) did not decrease through the generations of evolution. The enantioselectivity remained high in all variants for all model substrates.

Reduction of α substituted ketoesters

Ethyl-3-oxobutanoates bearing a substituent on C-2 have a chiral center at position 2 and a prochiral carbon at position 3. It is interesting to study how these substrates are handled by ketoreductase S1 from Candida magnoliae. Baker’s yeast,15,16 Geotrichum candidum17and Candida magnoliae12 are known to reduce these substrates. Whole cells of baker’s yeast reduce the ethyl-2-methyl-3-oxobutanoate (1d) exclusively yielding the (2R,3S)- configuration,19 whereas the 2-chloro substituted compound was reduced into a 1:1 mixture of (2S,3S) and (2R,3S) product. For a number of 2-substituted ethyl-3-oxobutanoates whole cells of baker’s yeast as well as a large number of its purified reductases give exclusively (3S) products.16,18,19 The selectivity for the configuration on C-2 varies and is dependent on the substrate and on which particular enzyme is used.

Wild type Candida magnoliae ketoreductase S1 is reported to reduce these substrates with relatively low selectivities for both the configuration at C-2 and C-3.12 In the case of 1d the configuration of the main product was (2S,3S). Minor amounts of (2S,3R) and (2R,3S) were formed and hardly any (2R,3R) product could be detected. With a chlorine as substituent on C-2 the major products were (2S,3S) and (2R,3R) with only a slight excess of (2S,3S).

We studied the 2-methyl-3-oxobutanoate ethyl ester (1d) and the influence of directed evolution on the activity and selectivity for this substrate. We found that the activity with 1d was in general surprisingly higher than that observed with the non-substituted equivalent (1b), although the steric and electrostatic

53 Asymmetric Reduction with Candida magnoliae Ketoreductase S1

effects of the methyl group seemingly would not be in favor of this substrate (Table 4.2 and 4.3). Hence, other effects should predominate here. The partitioning between the active site and the bulk water might be better with 1d due to its less hydrophilic character compared to 1b. Besides, there might be a hydrophobic region in the active site that is favorably filled by the methyl group with expulsion of bulk water.

It is interesting to note that the stereoselectivity of the reduction at C-3 is strongly affected by the presence of the methyl substituent at C-2. We observed that the main product of the reduction of ethyl-2-methylacetoacetate (1d) had the (3S)-configuration (Table 4.3). This implies that the hydride is transferred in a Prelog fashion, whereas in the case of 1b the hydride is transferred in an anti-Prelog manner. This opposite enantioselectivity is very interesting, but hard to explain well without detailed knowledge of the active site and the three dimensional structure of the enzyme.

Table 4.3 Activities and product distribution for ethyl-2-methylacetoacetate from variants selected out of 10 rounds of directed evolution on Candida magnoliae ketoreductase S1

(2S)-products (2R)-products OH O OH O OH O OH O (S) (S) (R) (R) Vini (S) O (R) O (S) O (R) O µmol⋅min-1⋅ g-1 Enzyme % % % % CmKr1 192 90.2 2.9 0.7 6.2 CmKr2A 334 64.8 2.8 3.7 28.7 CmKr2B 27 94.2 3.2 0.5 2.1 CmKr4 253 81.5 3 1.2 14.3 CmKr8 363 77.6 2.8 1.8 17.7 CmKr9 383 81.5 3.3 1.6 13.6 CmKr10 469 78.3 2.4 1.6 17.8 Activity measurement conditions: 2 ml Tris-HCl buffer (50 mM, pH 7.6), 0.16 mM NADPH, 10 mM substrate, 0.1 mg enzyme Enantioselectivity measurement conditions: 0.2 ml Tris-HCl buffer (50 mM, pH 7.6), 11 mM NADPH, 10 mM substrate, 0.2 mg enzyme

Looking in more detail at the stereoselectivity of this reaction, all variants of ketoreductase S1 had a preference for the (2S) enantiomer over the (2R)

54 Chapter 4

enantiomer of the substrate. The (2S)-enantiomer of 1d is preferentially reduced to get (S)-configuration at C-3 of the product, while the (2R)- enantiomer is preferentially reduced to yield (R)-configuration at C-3 of the product (Table 4.3). These results indicate that the enzyme not only handles the substrates 1b and 1d in a different manner, but also the two enantiomers of 1d are reacting differently.

We suppose that (S)-1d is positioned in the active site in the opposite way of 1b, interchanging the ester-side and the keto-side of the molecule. This should facilitate a favorable positioning of the methyl group at C-2 in a hydrophobic part of the active site. The conformation of the methyl group in the (R) enantiomer is such that binding in the same way as 1a-1c is preferred. A slight decrease in the preference of the mutants for the (2S)-enantiomer was observed through the generations of evolution. With this a slight decrease in the selectivity for the (S)-configuration at C-3 was observed as well. The highest selectivity (ee 97%, de 93%) was found with CmKr2B, of which the activity is very low compared to the other variants. The worst selectivity was observed with CmKr2A (39% ee, de 87%).

Activities with acetophenone

The activity of the variants has been determined with acetophenone (3) as aromatic model substrate, using both the native cofactor NADPH and NADH. The activities for this substrate were generally very low (Table 4.4). With some mutants, acetophenone was, surprisingly reduced faster by NADH than by NADPH, although this latter cofactor is the natural one of ketoreductase S1. It is worth noting that CmKr2A was more active with NADPH than with NADH, although this variant had been selected from a library co-targeting NADH dependency. The enantioselectivity, determined for the variant from round 10 only, was very high (over 99%, (R)-selective).

55 Asymmetric Reduction with Candida magnoliae Ketoreductase S1

Table 4.4 Activity with acetophenone of the mutants of ketoreductase S1 using both NADH and NADPH as the cofactor

clone NADPH as cofactor NADH as cofactor (µmol⋅min-1⋅ g-1) (µmol⋅min-1⋅ g-1) CmKr1 5 18 CmKr2A 17 4 CmKr2B 2 3 CmKr4 13 20 CmKr8 15 16 CmKr9 14 35 CmKr10 16 (>99 R) 15 Reaction conditions: 2 ml Tris-HCl buffer (50 mM, pH 7.6), 0.16 mM cofactor, 10 mM substrate, 0.1 mg enzyme

Cofactor preference

Native Candida magnoliae ketoreductase S1 requires NADPH as electron donor, but we found activity with NADH as cofactor in all the variants from directed evolution and with all model substrates (Figure 4.2). It is worth noting that 1a and 1d, for which the variants exhibited a good activity with NADPH (Tables 4.2 and 4.3), reacted comparatively sluggishly with NADH (2-10%). An increasing preference of the enzymes for using their natural cofactor (NADPH) with these latter substrates was observed when going down in evolution. This is due to an increasing rate with NADPH, while the rate using NADH as electron donor remains more or less the same. One could be led to conclude that as these substrates are reduced faster by NADPH than by NADH, that the binding of NADH becomes rate-determining. The substrates 1c and 3 were the more difficult substrates for the Candida magnoliae KREDs to handle (Tables 4.2 and 4.4). Most of the variants converted these latter substrates with comparable or higher rates using NADH instead of NADPH as the cofactor (Figure 4.2 c and e). Remarkable higher activities with these substrates using NADH instead of NADPH were found with CmKr4 and CmKr9. These results indicate that for slow reacting non- natural substrates the cofactor to be used is not strictly dictated by the natural preference of the enzyme, but also by the substrate itself.

56 Chapter 4

9000 350 O O O O 8000 a b Cl 300 O O 7000 250 6000 5000 200

4000 15 0 3000 10 0 2000 50 1000 0 0 CmKr1 CmKr2A CmKr2B CmKr4 CmKr8 CmKr9 CmKr10 CmKr1 CmKr2A CmKr2B CmKr4 CmKr8 CmKr9 CmKr10

10 0 500 O O O O 90 c 450 d 80 O 400 O

70 350

60 300

50 250 40 200 30 15 0 20 10 0 10 50 0 0 CmKr1 CmKr2A CmKr2B CmKr4 CmKr8 CmKr9 CmKr10 CmKr1 CmKr2A CmKr2B CmKr4 CmKr8 CmKr9 CmKr10

40 e O 35 -1 -1 Activity = Vini (µmol⋅min ⋅g ) 30

25 Vini with NADH 20

15 Vini with NADPH

10 a-d: Substrate 1a – 1d 5 e: Substrate 3

0 CmKr1 CmKr2A CmKr2B CmKr4 CmKr8 CmKr9 CmKr10

Figure 4.2 Activity of the variants from directed evolution on Candida magnoliae ketoreductase S1 using NADH compared to the activity using NADPH (%)

Activities in organic solvent

Previous studies have shown that Candida magnoliae ketoreductase S1 tolerates ethyl acetate, butyl acetate and diisopropyl ether as co-solvent in aqueous biphasic systems.10 We studied the tolerance of the variants from directed evolution against some water-miscible organic solvents. With few exceptions, the activity of Candida magnoliae ketoreductase S1 mutants in such monophasic aqueous mixtures was low (Table 4.6).

57 Asymmetric Reduction with Candida magnoliae Ketoreductase S1

Table 4.6 Residual activities (%) of the variants of C.m. ketoreductase S1 using various amounts of water miscible organic solvents

2-propanol tert-butyl alcohol 1,2-dimethoxyethaan 20 % (v/v) 80 % (v/v) 20 % (v/v) 80 % (v/v) 20 % (v/v) 80 % (v/v) CmKr1 38.3 7.9 29.8 83.3 6.4 0 CmKr2A 21.8 1.7 9.7 7.5 0.8 0.1 CmKr2B 51.6 44.2 46.7 130.3 0 0.3 CmKr4 24.7 1.6 12.5 17.5 0.4 0.2 CmKr8 21.9 3.2 9.3 13.2 0.2 0 CmKr9 18.1 4.3 10.5 32.2 0 0 CmKr10 19.5 26.7 11.6 10.1 0.6 0 Reaction conditions: Tris-HCl buffer (50 mM, pH 7.6)/organic solvent total 2 ml reaction volume, 0.16 mM cofactor, 10 mM substrate, 0.1 mg enzyme Activities were compared with those in pure Tris-HCl buffer (50 mM, pH 7.6) as solvent.

1,2-Dimethoxyethane caused nearly complete deactivation of the enzymes, both at high and low concentrations. 2-Propanol was tolerated slightly better by the mutants, although more than 50% of the activity in water was lost in aqueous mixtures containing 20% (v/v) of the latter organic solvent. Higher concentrations of 2-propanol render most mutants nearly completely inactive. The tolerance towards tert-butyl alcohol was in general comparable to 2- propanol, with the exception that high concentrations (80% v/v) were better tolerated than low concentrations. With the mutants CmKr1, CmKr2B, CmKr9 we observed in 80% tert-butyl alcohol a residual activity up to three times higher than that in 20% mixtures. In general Candida magnoliae ketoreductase S1 exhibits low tolerance towards the studied water-miscible solvents, but has a high tolerance towards some non water-miscible ones.10,13

Conclusion The reactant scope of ketoreductase S1 from Candida magnoliae was not narrowed down by directed evolution. The activity for ethyl-4-chloroacetoacetate increased by a factor 16, while the activities with similar β- ketoesters remained more or less the same. The enantioselectivity remained

58 Chapter 4

high and was not affected by directed evolution. This enzyme, although being NADPH dependent, also showed activity with NADH as the cofactor. In some cases the activity with NADH was higher than with NADPH.

Experimental part- Material and methods

Instruments and materials

Ethyl-4-chloroacetoacetate (1a), ethyl-acetoacetate (1b), tert-butyl- acetoacetate (1c), ethyl-2-methylacetoacetate (1d) and acetophenone (3) were purchased from Sigma-Aldrich. The corresponding alcohols were obtained by reduction with sodium borohydride in methanol. The cofactors NADH and NADPH were bought from Jülich Fine Chemicals. Codexis made the variants that resulted from directed evolution available. Tris(hydroxymethyl)aminoethane (99.8%) was purchased from Fluka for the preparation of buffers. All purchased reagents were used without additional purification. UV activity measurements were carried out on a Varian Cary 3 Bio UV-visible spectrophotometer at a wavelength of 340 nm. Chiral HPLC analyses were performed on a Chiralcel AD-H column with a flow of 0.6 ml/min and an eluent consisting of hexane-isopropyl alcohol (9:1) Detection was performed with a Waters 486 UV detector at 215 nm. GC analyses were performed using a beta cyclodex CB 25 M x 0.32 mm DF 0.25 column with He as carrier gas and flame ionisation detection.

Methods

Activity essay

59 Asymmetric Reduction with Candida magnoliae Ketoreductase S1

mM). The substrates used were: ethyl-4-chloroacetoacetate, ethyl- acetoacetate, tert-butyl-acetoacetate, ethyl-2-methylacetoacetate. The assay was started by adding 10 µl of a diluted sample containing +/- 10 mg/ml enzyme and monitoring the UV absorption at 340 nm and 20 °C. The molar absorption coefficient used for NAD(P)H was 6.22 l.mmol-1.cm-1.

Determination of the absolute chiral configuration

The conformation around the chiral centers when using ethyl-4-chloro acetoacetate (1a), ethyl acetoacetate (1b), tert-butyl acetoacetate (1c) as the substrate has been ascribed according to known data from the literature for Candida magnoliae S112 and with the resulting data for the products of reduction of these substrate by Saccharomyces cerevisiae.19 Analysis has been performed with chiral GC in the case of using ethyl-4-chloro acetoacetate (1a), ethyl acetoacetate (1b), tert-butyl acetoacetate (1c) as the substrate. For 1a a isothermal temperature program of 95°C was used. For 1b and 1c an isothermal temperature program was used of 85°C. In the case of ethyl-2-methylacetoacetate (1d) HPLC was used and the peaks were compared with the reduction products of this substrate by Saccharomyces cerevisiae.19 After derivatisation with 3,5-dinitrobenzoylchloride of the product resulting from Candida magnoliae S1 reduction the retention times were compared with the literature values found in the same analysis.12 Both data support the following elution order of the isomers: (2S,3S), (2R,3S), (2S,3R), (2R,3R). Analysis has been performed with an AD-H column and an elluent consisting of hexane:isopropyl alcohol (9:1) at a flowrate of 0.6 ml/min.

Determination of enantioselectivity

Reactions were performed for each enzyme and each substrate under the following conditions. To Tris-buffer (63 µl, 50 mM, pH 7.6) in a 1 ml vial were added a solution of NAD(P)H (67 µl, 32 mM) and a solution of substrate (20 µl, 100 mM). 50 µl of a 10 mg/ml stock solution of each enzyme was added to start the reaction. The reaction was left shaking overnight before extraction of

60 Chapter 4

the reaction products with ethylacetate (500 µl). The organic layer was dryed and injected in the chiral GC. For 1d extraction was performed with hexane (500 µl), which was injected on the chiral hplc using the AD-H column.

Stability towards organic solvents

The stability towards organic solvents has been performed according to the procedure for the activity assay in which the amount of buffer is reduced and the various amounts of organic solvents are substituting this to add up to a total reaction volume of 2 ml. The model substrate used for the comparison was 1b.

References

1 Shimizu, S.; Kataoka, M.; Kita, K. Ann. NY. Acad. Sci. 1998, 864, 87-95 2 Zhou, B.; Gopalan, A.S.; VanMiddlesworth, F.; Shieh, W.; Sih, C.J. J. Am. Chem. Soc. 1983, 105, 5925-5936 3 Shieh, W.; Gopalan, A.S.; Sih, C.J. J. Am. Chem. Soc. 1985, 107, 2993-2994 4 Kita, K.; Katoaka, M.; Shimizu, S. J. Biosci. Bioeng. 1999, 88, 591-598 5 Yasohara, Y.; Kizaki, N.; Hasegawa, J.; Takahashi, S. Wada, M.; Kataoka, M.; Shimizu, A. Appl. Microbiol. Technol. 1999, 51, 847-851 6 Wada, M.; Kataoka, M.; Kawabata, H.; Yasohara, Y.; Kizaki, N.; Hasegawa, J.; Shimizu, S. Biosci. Biotechnol. Biochem. 1998, 62, 280-285 7 Yasohara, Y.; Kizaki, N.; Hasegawa, J.; Wada, M.; Kataoka, M.; Shimizu, S. Biosci. Biotechnol. Biochem. 2000, 64, 1430-1436 8 Kizaki, N.; Yasohara, Y.; Hasegawa, J.; Wada, M.; Kataoka, M.; Shimizu, S. Appl. Microbiol. Biotechnol. 2001, 55, 590-595 9 Morikawa, S.; Nakai, T.; Yasohara, Y.; Nanba, H.; Kizaki, N.; Hasegawa, J. Biosci. Biotechnol. Biochem. 2005, 69, 544-552 10 Nakai, T.; Morikawa, S.; Kizaki, N.; Yasohara, Y. EP1416050, 2004 11 Shimizu, S.; Katoaka, M.; Kita, K. J. Mol. Catal. B: Enzym. 1998, 5, 321-325 12 Yasohara, Y.; Kizaki, N.; Hasegawa, J.; Wada, M.; Kataoka, M.; Shimizu, S. Tetrahedron: asymmetry, 2001, 12, 1713-1718

61 Asymmetric Reduction with Candida magnoliae Ketoreductase S1

13 Davis, C.S.; Grate, J.H.; Gray, D.R.; Gruber, J.M.; Huisman, G.W.; Ma, S.K.; Newman, L.M.; Sheldon, R.A. WO2005/018579, 2005 14 Cahn, R.S.; Ingold, C.K.; Prelog, V. Angew. Chem. Int. Ed. Engl. 1966, 5, 385-415 15 Nakamura, K.; Kawai, Y.; Miyai, T.; Honda, S.; Nakajima, N.; Ohno, A. Bull. Chem. Soc Jpn. 1991, 64, 1467-1470 16 Rodriguez, Y.; Schroeder, K.T.; Kayser, M.M.; Stewart, J.D. J. Org. Chem. 2000, 65, 2586-2887 17 Kawai, Y.; Tananobe, K.; Ohno. Bull. Chem. Soc. Jpn. 1997, 70, 1683-1686 18 Kaluzna, I.A.; Feske, B.D.; Wittayanan, W. Ghiviriga, I.; Stewart, J.D. J. Org. Chem. 2005, 70, 342-345 19 Kaluzna, I.A.; Matsuda, T.; Sewell, A.K.; Stewart, J.D. J. Am. Chem. Soc. 2004, 126, 12827

Asymmetric Carbonyl- Reductions with Microbial Ketoreductases

Abstract A variety of ketoreductases from different origin were studied in the biocatalytic reduction of β-ketoesters and some aromatic ketones. Their tolerance towards organic co-solvents was studied, together with the possibility of using 2-propanol as sacrificial substrate to regenerate the cofactor. The enantioselectivities and reaction rates were dependent both on the enzyme as well as the substrate. This Chapter presents an overview of a range of synthetically useful biocatalysts for the reduction of carbonyl compounds.

Asymmetric Carbonyl-Reductions with Microbial Ketoreductases

Introduction Due to the often strict substrate recognition of reducing enzymes, which results in very high chemo-, regio- and enantioselectivities, biocatalytic reduction of carbonyl compounds is a very attractive synthetic route towards homochiral secondary alcohols. The biocatalytic reduction of carbonyl compounds using whole cells of Saccharomices cerevisiae has been known and employed widely since the pioneering work in 1914.1 The major drawback of the application of this commercially available, inexpensive organism, which is moreover simple to apply, is that the enantioselectivity is generally poor. This is ascribed to the presence of multiple reductases, with conflicting enantiomeric preference.2 Numerous approaches to overcome this problem have been modestly successful.3,4,5,6

With the more advanced biochemical tools that exist nowadays, many ketoreductases have been isolated from their parent organism and overexpressed in a suitable host. Due to the high interest in synthetic strategies for the production of homochiral β-hydroxy esters and β,δ-dihydroxy esters, a thorough study of the reduction of the corresponding ketoesters with whole cells of bakers’ yeast and a number of its purified enzymes has been performed.2,7,8,9 Additional to the enzymes from bakers yeast, many ketoreductases from a wide variety of other organisms have been studied in the reduction of these substrates.10,11 Besides being of interest as a route to β-hydroxy esters, asymmetric bioreduction of ketones is generally considered as an attractive route to many valuable chiral alcohol intermediates. An overview of the latest achievements in the field of asymmetric biocatalytic carbonyl reductions can be found in recent reviews.12,13

In our studies we used ketoreductases isolated from six different microorganisms (Table 5.1). The ketoreductases from Candida magnoliae and Rhodococcus erythropolis were subjected to directed evolution. The other enzymes are wild type ketoreductases.

64 Chapter 5

Table 5.1 The studied ketoreductases, their source and cofactor preference

Source organism ID Cofactor dependence Candida magnoliae CmKrxa NADPH Saccharomyces cerevisiae ScE1 NADPH ScE2 NADH ScE3 NADPH ScE4 NADH, NADPH Rhodococcus erythropolis RhoC,Rhhoxxa NADH Sporobolomyces salmonicolor SpbC NADH Streptomyces coelicolor StaC NADH Lactobacillus kefir KefC NADPH

a x Indicates an identification number of a variant from directed evolution

In this Chapter we report the performance of the microbial ketoreductases presented in Table 5.1 in the reduction of a number of ketoesters and aromatic ketones. The influence of organic co-solvents and in-situ cofactor regeneration using 2-propanol as a cosubstrate were also studied.

Results and discussion

Reduction of β-ketoesters

The activity with a series of non-branched β-ketoesters was determined for all the microbial ketoreductases (KREDs) shown in Table 5.2. The Candida magnoliae ketoreductase S1 and its variants from directed evolution, targeting the reduction of ethyl 4-chloroacetoacetate (1a), have been reported in detail in Chapter 4. The results with the variant selected from round 10 of evolution are included for comparison. From Rhodococcus erythropolis, wild type ketoreductase (RhoC) and three selected variants from directed evolution (Rhh001, Rhh004, Rhh014) were studied. Furthermore, wild type ketoreductases from Saccharomyces cerevisiae, Sporobolomyces salmonicolor, Streptomyces coelicolor and Lactobacillus kefir were studied.

65 Asymmetric Carbonyl-Reductions with Microbial Ketoreductases

Table 5.2 Asymmetric reduction of β-ketoesters catalysed by microbial ketoreductases

O O OH O a: R1=Cl; R2=H; R3=C2H5 R R KRED R R b: R1=H; R2=H; R3=C2H5 1 O 3 1 O 3 c: R1=H; R2=H; R3=C(CH3)3 R2 R2 + d: R =H; R =CH ; R =C H 1a-dNAD(P)H NAD(P) 2a-d 1 2 3 3 2 5

OH O OH O OH O Cl O O O (S)-2a (R)-2b (R)-2c Vini ee Vini ee Vini ee Enzyme µmol⋅min-1⋅ g-1 (%) µmol⋅min-1⋅ g-1 (%) µmol⋅min-1⋅ g-1 (%) CmKr10 7763 >99 (S) 214 >99 (R) 41 >99 (R) RhoC 60 >99 (R) 587 >99 (S) 122 >99 (S) Rhh001a 51 >99 (R) 341 >99 (S) 135 >99 (S) Rhh004a 55 >99 (R) 314 >99 (S) 126 >99 (S) Rhh014a 45 >99 (R) 599 >99 (S) 120 >99 (S) ScE1 48 >99 (S) 17 >99 (R) 5 36 (R) ScE2 62 77 (R) 60 >99 (S) 42 >99 (S) ScE3 53 >99 (R) 969 >99 (S) 240 >99 (S) ScE4 30 >99 (R) 76 >99 (S) 21 >99 (S) StaC 29 >99 (R) 47 50 (S) 26 >99 (S) SpbC 56 >99 (R) 66 19 (S) 37 >99 (S) KefC 46 >99 (S) 590 >99 (R) 105 >99 (R)

Activity measurement conditions: 2 ml Tris-HCl buffer (50 mM, pH 7.6), 0.16 mM NAD(P)H, 10 mM substrate, 0.1 mg enzyme Enantioselectivity measurement conditions: 0.2 ml Tris-HCl buffer (50 mM, pH 7.6), 11 mM NAD(P)H, 10 mM substrate, 0.2 mg enzyme a Numbers do not indicate the generation in evolution

The activity of all KREDs with 1a was low, compared with the exception of variants of Candida magnoliae that had been evolved specifically for this substrate (CmKr10, Chapter 4). When the substrate lacks the chlorine atom, as in 1b, the activity is considerably higher with the Rhodococcus and Lactobacillus enzymes as well as ScE3 from Saccharomyces cerevisiae, which showed the best activity with this latter substrate. The other enzymes from baker’s yeast and the KREDs from Sporobolomyces salmonicolor and Streptomyces coelicolor remained low in activity. The same trend was observed for substrate 1c, albeit with activities that were in general lower than with 1b. This was expected due to the bulky tert-butyl ester group.

66 Chapter 5

The stereoselectivity of reduction of the model substrates 1a-1c, mediated by the ketoreductases from C. magnoliae, L. kefir and ScE1 from S. cerevisiae, was the opposite of what would be predicted by Prelog’s rule. In Contrast, the enzymes from R. erythropolis, S. coelicolor, S. salmonicolor and ScE2, ScE3 and ScE4 from S. cerevisiae reduced these substrates according to Prelog’s rule (Table 5.2). Applying the Cahn-Ingold-Prelog convention,14 reduction according to Prelog’s rule would result in the (R)-configuration for 1a and (S)- configuration for the others. Anti-Prelog reduction would yield the opposite enantiomers. The enzymes KefC from Lactobacillus kefir and ScE3 from Saccharomyces cerevisiae both exhibited good reaction rates and absolute but opposite stereoselectivity. Hence, these enzymes gave full access to both pure enantiomers of these substrates.

Reduction of α-substituted ketoesters

The reduction of the α-substituted ketoester 1d involves a kinetic resolution (2R vs. 2S) as well as prochiral selectivity. Most enzymes preferentially reduced (2R)-1d (Table 5.3). CmKr10 and KefC, in contrast, showed a modest kinetic bias towards (2S)-1d and the enantiomeric bias of ScE2 and ScE3 was low. Most of the enzymes exhibited the same prochiral selectivity in the reduction of 1d as with 1b (Tables 5.2 and 5.3). Exceptions are ScE1, which formed approximately equal amounts of (2S, 3R)-2d and (2R, 3S)-2d, indicating that the enantiomers of 1d bind in different ways. CmKr10 produced mostly (2S, 3S)-2d, while it is producing (3R)-2b. This behavior of Candida magnoliae ketoreductase S1 and its variants from directed evolution was discussed in detail in Chapter 4.

The enzymes ScE3 from Saccharomyces cerevisiae and KefC from Lactobacillus kefir, showed a relatively high activity for ethyl-2- methylacetoacetate (1d). This activity was comparable with the activity they exhibited for the unsubstituted equivalent 1b. CmKr10 from Candida magnoliae (Chapter 4) showed a slightly higher activity for 1d than for 1b. The other enzymes had a moderately lower activity for 1d compared to 1b.

67 Asymmetric Carbonyl-Reductions with Microbial Ketoreductases

Table 5.3 Asymmetric reduction of ethyl-2-methylacetoacetate catalysed by various microbial ketoreductases

(2S)-products (2R)-products OH O OH O OH O OH O (S) (S) (R) (R) Vini (S) O (R) O (S) O (R) O µmol⋅min-1⋅ g-1 Enzyme % % % % CmKr10 469 78.3 2.4 1.6 17.8 RhoC 123 19.9 0 80.1 0 Rhh001a 68 6.0 0 94.0 0 Rhh004a 117 20.0 0 80.0 0 Rhh014a 222 14.2 0 85.8 0 ScE1 8 7.3 46.4 46.3 0 ScE2 72 54.7 0 45.3 0 ScE3 962 0 0 100 0 ScE4 25 13.5 0.2 86.0 0.2 StaC 27 0.7 0 99.3 0 SpbC 42 0 0 100 0 KefC 675 36.1 27.4 0 36.5 Activity measurement conditions: 2 ml Tris-HCl buffer (50 mM, pH 7.6), 0.16 mM NADPH, 10 mM substrate, 0.1 mg enzyme Enantioselectivity measurement conditions: 0.2 ml Tris-HCl buffer (50 mM, pH 7.6), 11 mM NADPH, 10 mM substrate, 0.2 mg enzyme a Numbers do not indicate the generation in evolution

Reduction of aromatic carbonyl compounds

The reduction of some aromatic ketones using the ketoreductases mentioned above were studied. The model substrates studied were: acetophenone (3), α-tetralone (4), β-tetralone (5) and deoxybenzoin (6). This latter substrate (6) proved to be completely unreactive with all of the ketoreductases studied. The tetralones reacted very sluggishly, with α-tetralone being the better substrate of the two isomers (Table 5.4). The reduction rate of β-tetralone (5) was in many cases too low to quantify and the enantioselectivity could not be determined reliably. Of the studied aromatic substrates, acetophenone (3) was the best-accepted substrate by the ketoreductases used in these studies.

68 Chapter 5

Table 5.4 Asymmetric reduction of aromatic carbonyl compounds catalysed by various microbial ketoreductases

O O

3 4 5 Vini ee Vini ee Vini ee Enzyme µmol⋅min-1⋅ g-1 % µmol⋅min-1⋅ g-1 % µmol⋅min-1⋅ g-1 % CmKr10 9 >99 (R) 3 91 (S) <0.5 nd RhoC 80 >99 (S) 3 93 (S) <0.5 nd Rhh001 35 >99 (S) 2 94 (S) nr nd Rhh004 9 >99 (S) 3 93 (S) <0.5 nd Rhh014 66 >99 (S) 3 95 (S) <0.5 nd ScE1 19 10 (S) 9 53 (R) 1 nd ScE2 12 >99 (S) 3 97 (S) <0.5 nd ScE3 14 97 (S) 4 97 (S) <0.5 nd ScE4 4 >99 (S) 4 95 (S) <0.5 nd StaC 2 >99 (S) 4 95 (S) <0.5 nd SpbC 2 31 (R) 3 17 (R) 0.5 nd KefC 256 >99 (R) 3 62 (S) 2 nd

Reaction conditions: 0.2 ml Tris-HCl buffer (50 mM, pH 7.6), 11 mM NAD(P)H, 10 mM substrate, 0.1 mg enzyme. Activity determined over first hour of reaction nd: Not determined nr: No reaction

The ketoreductases that had an (S) prochiral selectivity with 1b, were (S)- selective with the studied aromatic substrates. According to Prelog’s prediction model for the preferentially formed enantiomer during carbonyl reductions (Chapter 1), the similar prochiral preference with 3-5 was expected as with 1b. An exception was SpbC, which showed a modest (R)-selectivity for the aromatic substrates but was (S)-selective with 1b. CmKr10 and KefC, which were (R)-selective with 1b, produced (R)-3, but (S)-4. ScE1, also being (R)-selective with 1b, produced 3 and 4 with (S)-configuration.

69 Asymmetric Carbonyl-Reductions with Microbial Ketoreductases

Tolerance towards organic solvents

Tolerance towards organic co-solvents is an important characteristic for enzymes if substrates that are sparingly soluble in water need to be converted. We studied the tolerance of the various ketoreductases towards some water-miscible organic solvents. With few exceptions, the activity of Candida magnoliae ketoreductase S1 mutants was low in such monophasic aqueous mixtures (Chapter 4). The same holds true for the ketoreductases of interest in this Chapter as in many cases most of the activity was lost when water-miscible organic solvents were added to the reaction medium.

Table 5.5 Residual activities (%) of the microbial ketoreductases using various amounts of water miscible organic solvents

2-propanol tert-butyl alcohol 1,2-dimethoxyethaan 20 % (v/v) 80 % (v/v) 20 % (v/v) 80 % (v/v) 20 % (v/v) 80 % (v/v) CmKr10 19.5 26.7 11.6 10.1 0.6 0 Rhoc 16.0 7.5 9.0 3.7 1.4 0 Rhho01 6.6 0.1 4.1 4.9 0 2.5 Rhho04 6.3 0 0 0.3 0 4.2 Rhho14 12.6 0 7.5 2.4 0 17.5 ScE1 6.9 1.1 0 40.3 0 2.4 ScE2 45.5 5.5 34.6 152.8 9.7 6.4 ScE3 11.7 0.2 18.3 0.1 52.9 13.2 ScE4 98.3 0 0.9 11.9 2.3 0 StaC 1.8 1.8 0.6 28.4 0 0 SpbC 2.0 0.8 0.8 19.1 0 0 KefC 1.3 0.2 57.7 2.7 77.9 0 Reaction conditions: Tris-HCl buffer (50 mM, pH 7.6)/organic solvent total 2 ml reaction volume, 0.16 mM cofactor, 10 mM substrate 1b, 0.1 mg enzyme Activities were compared with those in pure Tris-HCl buffer (50 mM, pH 7.6) as solvent.

Except for KefC and ScE3, which tolerated low concentrations of 1,2- dimethoxyethane quite well, this latter solvent was highly deactivating (Table 5.5). tert-Butyl alcohol was a better co-solvent. Higher residual activities were obtained in the more concentrated biphasic systems (80 % v/v) than in diluted systems (20 % v/v) with this solvent. In one case hyperactivation was observed (ScE2, Table 5.5). 2-Propanol at low concentration resulted in low to

70 Chapter 5

medium residual activities, while at high concentration nearly complete deactivation was observed in most cases.

Cofactor regeneration

The addition of a water miscible organic solvent to the aqueous medium is necessary with reducing substrates that are sparingly water-soluble. The addition of 2-propanol can fulfil this purpose. Besides, it can be used as a sacrificial substrate to regenerate the cofactor, obviating the need for a second enzyme to regenerate the cofactor. We studied the ability of the ketoreductases of interest to use 2-propanol to regenerate their own cofactor (Table 5.6). The Candida magnoliae KREDs did not show any reduction of NADP+ to NADPH by using 2-propanol. The bakers’ yeast enzymes were able to reduce their oxidized cofactor, albeit with very poor rates. The enzymes from Rhodococcus erythropolis and Lactobacillus kefir exhibited good reduction rates of their oxidized cofactors. Therefore cofactor regeneration can easily be achieved by the addition of 2-propanol to the reaction medium with these enzymes.

Table 5.6 Reduction rate with various microbial ketoreductases of the oxidized cofactors (NAD+, NADP+) in the presence of 5% (v/v) 2-propanol

Enzyme Cofactor reduction rate Enzyme Cofactor reduction rate µmol⋅min-1⋅ g-1 µmol⋅min-1⋅ g-1 CmKr10 0 ScE2 0 RhoC 194 ScE3 2 Rhh001a 59 ScE4 3 Rhh004a 100 StaC 3 Rhh014a 185 SpbC 5 ScE1 3 KefC 585 a Numbers do not indicate the generation in evolution

Conclusion The reactant scope of the ketoreductases studied in this chapter is not limited to ketoesters. Some aromatic ketones were accepted as well, depending on

71 Asymmetric Carbonyl-Reductions with Microbial Ketoreductases

the enzymes used. Small amounts of water miscible organic solvents were in many cases tolerated, while in some cases high concentrations were prefered over the lower range. The KREDs from Rhodococcus erythropolis and Lactobacillus kefir can efficiently regenerate their own cofactor with the use of 2-propanol as sacrificial substrate.

Experimental part- Material and methods

Instruments and materials

Ethyl-4-chloroacetoacetate (1a), ethyl-acetoacetate (1b), tert-butyl- acetoacetate (1c), ethyl-2-methylacetoacetate (1d), acetophenone (3), α- tetralone (4) and β-tetralone (5) were purchased from Sigma-Aldrich. The corresponding alcohols were obtained by reduction with sodium borohydride in methanol. The cofactors NADH and NADPH were bought from Jülich Fine Chemicals. Codexis made the ketoreductases from various microbial organisms available. Tris(hydroxymethyl)aminoethane (99.8%) was purchased from Fluka for the preparation of buffers. All purchased reagents were used without additional purification. UV activity measurements were carried out on a Varian Cary 3 Bio UV-visible spectrophotometer at a wavelength of 340 nm. HPLC analysis of aromatic substances was performed on 2 in series placed Chromolith SpeedROD® RP-18e 50-4.6 mm columns with water-MeOH 75:25 as eluent and a flow rate of 1 ml/min. UV detection was performed with a Shimadzu SPD-36A at 215 nm. Trimethoxybenzene was used as internal standard. Chiral HPLC analyses of 1d were performed on a Chiralcel AD-H column with a flow of 0.6 ml/min and an eluent consisting of hexane-isopropyl alcohol (9:1) Detection was performed with a Waters 486 UV detector at 215 nm. Chiral HPLC analyses of 3-5 were performed on a Chiralcel OD-H column with a flow of 0.6 ml/min and an eluent consisting of hexane-isopropyl alcohol (98:2) Detection was performed with a Waters 486 UV detector at 215 nm. GC

72 Chapter 5

analyses were performed using a beta cyclodex CB 25 M x 0.32 mm DF 0.25 column with He as carrier gas and flame ionisation detection.

Methods

Activity essay for aliphatic ketoesters

The activity of the various enzymes for a particular substrate was measured by following the consumption of NAD(P)H by means of its UV absorption. To Tris-buffer (1.78 ml, 50 mM, pH 7.6) in a 2 ml cuvette were added: a solution of NAD(P)H (10 µl, 32 mM) and a solution of substrate (200 µl, 100 mM). The substrates used were: ethyl-4-chloroacetoacetate, ethyl- acetoacetate, tert-butyl-acetoacetate, ethyl-2-methylacetoacetate. The assay was started by adding 10 µl of a diluted sample containing +/- 10 mg/ml enzyme and monitoring the UV absorption at 340 nm and 20 °C. The molar absorption coefficient used for the consumed NAD(P)H was 6.22 l.mmol-1.cm- 1.

Activity essay for aromatic ketoesters

To Tris-buffer (103 µl, 50 mM, pH 7.6) in a 1 ml flask were added: a solution of NAD(P)H (67 µl, 32 mM) and a solution of substrate in 2-propanol (20 µl, 100 mM). A solution of the enzyme (10 µl of a 10 mg/ml stock) was added to start the reaction. The substrates used were: acetophenone (3), α-tetralon (4) and β-tetralon (5) and deoxybenzoin (6). After 1 h the reaction was stopped by quenching with acetone (400 µl). After centrifugation and enzyme removal using membrane filtration with Millipore centripreps with a 10 kD cut off to remove all dissolved protein, this was injected on the HPLC. Analysis was performed on 2 serial placed Chromolith SpeedROD® RP-18e 50-4.6 mm columns with water-methanol 75:25 as eluent and a flow rate of 1 ml/min.

73 Asymmetric Carbonyl-Reductions with Microbial Ketoreductases

Determination of absolute chiral configuration

The configuration around the chiral centers when using ethyl-4- chloroacetoacetate (1a), ethyl acetoacetate (1b), tert-butyl acetoacetate (1c) as the substrate has been ascribed according to known data from the literature for Candida magnoliae S115 and with the resulting data for the products of reduction of these substrate by Saccharomyces cerevisiae.16 Analysis has been performed with chiral GC in the case of using ethyl-4- chloroacetoacetate (1a), ethyl acetoacetate (1b), tert-butyl acetoacetate (1c) as the substrate. For 1a a isothermal temperature program of 95°C was used. For 1b and 1c an isothermal temperature program was used of 85°C. In the case of ethyl-2-methylacetoacetate (1d) HPLC was used and the peaks were compared with the reduction products of this substrate by Saccharomyces cerevisiae.19 After derivatisation with 3,5-dinitrobenzoylchloride of the product resulting from Candida magnoliae S1 reduction the retention times were compared with the literature values found in the same analysis.15 Both data support the following elution order of the isomers: (2S,3S), (2R,3S), (2S,3R), (2R,3R). Analysis has been performed with an AD-H column and an elluent consisting of hexane-isopropyl alcohol (9:1) at a flowrate of 0.6 ml/min. The absolute configuration of the reduction product of phenylacetone has been determined by comparing the HPLC results on an OD-H column with the results of the Burkholdria cepacia lipase-catalysed esterification with vinyl acetate, which is known to be (R)-selective.17 The configuration of the reduction products of the tetralons was determined using the same (R)- selective esterification by Burkholdria cepacia lipase of these substrates.18 Analysis was performed with HPLC on the OD-H column.

Determination of enantioselectivity

74 Chapter 5

substrates 1a-1c the reaction products were extracted from the reaction mixture with ethylacetate (500 µl). The organic layer was dried and injected in the chiral GC. For 1d extraction was performed with hexane (500 µl), which was injected on the chiral HPLC using the AD-H column. Reactions with the aromatic substrates were performed under similar conditions, only with 10% (v/v) 2-propanol as co-solvent present. Samples were taken by extraction with hexane (500 µl), which were injected on the chiral HPLC using the OD-H column.

Stability towards organic solvents

The stability towards organic solvents was studied according to the procedure for the activity assay in which the amount of buffer is reduced and the various amounts of organic solvents are substituting this to add up to a total reaction volume of 2 ml. The model substrate used for the comparison was 1b.

Cofactor regeneration

The ability of the various enzymes to use 2-propanol to regenerate their cofactor was measured by following the production of NAD(P)H from NAD(P)+ by means of its UV absorption. To Tris-buffer (1.58 ml, 50 mM, pH 7.6) in a 2 ml cuvette were added: 2-propanol (200 µl) and a solution of NAD(P)+ (10 µl, 32 mM). The assay was started by adding 10 µl of a diluted sample containing +/- 10 mg/ml enzyme and monitoring the UV absorption at 340 nm and 20 °C. The molar absorption coefficient used for the consumed NAD(P)H was 6.22 l.mmol-1.cm-1.

75 Asymmetric Carbonyl-Reductions with Microbial Ketoreductases

References

1 Neuberg, C.; F., N.F. Biochem Z. 1914, 62, 482-488 2 Kaluzna, I.; Matsuda, T.; Sewell, A.L.; Stewart, J.D. J. Am. Chem. Soc. 2005, 126, 12827-12832 3 Zhou, B.; Gopalan, A.S.; van Middlesworth, F.; Shieh, W. R.; Hih, C.J. J. Am. Chem. Soc. 1983, 105, 5925-5926 4 Kawai, Y.; Kondo, S.; Tsujimoto, M.; Nakamura, K.; Ohno, A. J. Bull. Chem. Soc. Jpn. 1994, 67, 2244-2247 5 D’Arrigo, P.; Pedrocchi-Fantoni, G.; Servi, S.; Strini, A. Tetrahedron: Asymmetry 1997, 8, 2375-2379 6 Rodriguez, S.; Kayser, M.; Steward, J.D. Org. Lett. 1999, 1, 1153-1155 7 Rodriguez, S.; Kayser, M.M.; Stewart, J.D. J. Am. Chem. Soc. 2001, 123, 1547-1555 8 Wolberg, M.; Kaluzna, I.A.; Müller, M.; Stewart, J.D. Tetrahedron: Asymmetry 2004, 15, 2825-2828 9 Kaluzna,0 I.A.; Feske, B.D.; Wittayanan, W.; Ghiviriga, I.; Stewart, J.D. J. Org. Chem. 2005, 70, 342-345 10 Kita, K.; Kataoka, M.; Shimizu, S. J. Biosci. Bioeng. 1999, 88, 591-598 11 Shimizu, S.; Kataoka, M.; Kita, K. J. Mol. Catal. B: Enzym. 1998, 5, 321-325 12 Nakamura, K.; Yamanaka, R.; Matsuda, T.; Harada, T. Tetrahedron: Asymmetry 2003, 14, 2659-2681 13 Kroutil, W.; Mang, H.; Edegger, K.; Faber, K. Adv. Synth. Catal. 2004, 346, 125-142 14 Cahn, R.S.; Ingold, C.K.; Prelog, V. Angew. Chem. Int. Ed. Engl. 1966, 5, 385-415 15 Yasohara, Y.; Kizaki, N.; Hasegawa, J.; Wada, M.; Kataoka, M.; Shimizu, S. Tetrahedron: asymmetry, 2001, 12, 1713-1718 16 Kaluzna, I.A.; Matsuda, T.; Sewell, A.K.; Stewart, J.D. J. Am. Chem. Soc. 2004, 126, 12827 17 Ema, T.; Kageyama, M.; Korenaga, T.; Sakai, T. Tetrahedron: Asymmetry 2003, 14, 3943-3947 18 Merlic, C.; Walsh, J. J. Org. Chem. 2001, 66, 2265-2274

Part III Enantioselective Aldol Reaction: DERA

Optimization and Production of 2- Deoxyribose-5-phosphate Aldolase

Abstract Directed evolution has been applied to alter the substrate specificity of DERA aiming to increase the acceptance of non-phosphorylated compounds. Improved expression methods have been applied increasing the protein yield from fermentation substantially. The processes of gene isolation and directed evolution have been studied and the resulting variants have been characterized. In two rounds of DNA shuffling an 8-fold increase in activity for the non-phosphorylated substrate 2-deoxy-D-ribose was obtained. The mutations that cause this activity improvement did not affect the chiral preference of the enzyme.

Production and Optimization of 2-deoxyribose-5-phosphate Aldolase

Introduction In the past two decades, major advances in molecular biological techniques have been achieved. Methodologies for gene isolation, expression and enzyme production became, to an increasing extent, “off the shelf” technologies. Additionally biological information databases and gene libraries are growing rapidly. These information sources, where genetic codes and originating organisms for certain proteins can be found, together with the well- developed gene technology, has drastically improved the access towards various enzymes and the ease of their production.

A very important issue in the efficient large-scale production and evolution of any biocatalyst, which is often overlooked as being a minor detail, is the availability of an efficient overexpression system in a host that is well known in its behavior during fermentation and handling. A recombinant Escherichia coli strain DH5α that overexpresses the gene for 2-deoxy-D-ribose-5-phosphate aldolase1,2,3 is commercially available*. However, we required the isolated gene for mutation studies and better overexpression systems are available nowadays. Hence, the gene from E. coli encoding for DERA was isolated and overexpressed in an improved, proprietary recombinant production strain.

DERA is the only known aldolase that accepts aldehydes and ketones as donor substrates. It has a broad aldol acceptor tolerance, albeit with relatively low reaction rates for non-phosphorylated compounds1,2. In contrast, other aldolases that depend on phosphorylated substrates, such as fructose-1,6- bisphosphate aldolase, lose all activity when the phosphate group is not present in the substrate. The interesting product scope and the fact that a minor reactivity with non-phosphorylated compounds is already present in wild type DERA, makes it interesting to improve this phosphate independence, which is of utmost importance for its synthetical- and industrial applicability.

* Available from the ATCC under culture number ATCC 86963

80 Chapter 6

In studies with the goal of changing the substrate specificity of wild type DERA from the negatively charged 2-deoxy-D-ribose-5-phosphate (DRP) to the neutral 2-deoxy-D-ribose (DR), several side chains in the phosphate binding pocket (consisting of residues Gly171, Lys172, Gly204, Gly205, Val206, Arg207, Gly236, Ser238, Ser239; Figure 6.1) were targeted by site directed mutagenesis.4 Three neutral side chains in the phosphate binding pocket were replaced with acidic ones, resulting in the mutants: Gly205Glu, Ser238Asp, Ser239Glu.

Asp102

Lys201 Lys167

Gly236

Ser238

Gly171

Lys172

Gly205 Ser239 Figure 6.1 The active site of DERA (Asp 102, Lys167 and Lys201) and the residues of the phosphate binding pocket (represented as balls)

The residue Gly205 is highly conserved in close homologues of DERA4 and exchange of this residue resulted in complete inactivity towards DR as well as DRP. Both these facts indicate that Gly205 is a structurally important residue. The exchange of Ser238 with the acidic Asp should retain the hydrophilic nature of the binding pocket, but neutral and positively charged groups become preferred over negatively charged groups. A variant with the mutation

Ser238Asp accordingly showed a 2.5 times higher activity (kcat/kM) towards DR than the wild type and a much lower activity towards DRP due to

81 Production and Optimization of 2-deoxyribose-5-phosphate Aldolase

electrostatic interactions.5,6 The same holds true for the more remote mutation Ser239Glu. Hence the increased activity for DR is much less evident, but the activity towards DRP is still substantially lower than in the wild type.

Compared with wild type DERA (WT), the Ser238Asp mutant also showed improvement in activity toward the non-natural sequential aldol reaction between two molecules of acetaldehyde and some β-substituted propionaldehydes (Table 6.1).7,8 Moreover, it exhibits activity for 3- azidopropionaldehyde that the wild type lacks.

Table 6.1 Sequential aldol reaction of non phosphorylated substrates catalysed by DERA variant Ser238Asp and the wild type enzyme R R DERA O OH O O R O + 2 O 6 d Br2, BaCO3 OH OH R WT yield (%) Ser238Asp Yield (%) a b CH2CH2 N3 nr 35 b b CH2CH2 Cl 25 43

CH2CH2NO2 nr nr a No reaction b Yield based on the final lacton production

The above rational mutagenesis strategies have led to an overall increase in activity for non-phosphorylated substrates and broadening of the substrate scope. The residue Ser238 seems to be playing an important role in the enzyme’s preference for phosphorylated aldol acceptors. The influence of other residues, especially the ones further removed from the active site, is hard to predict. Therefore, the utilisation of non-rational methods to further increase the phosphate independence of DERA is the obvious choice. Directed evolution is a very powerful tool to adjust and improve enzymatic activities, especially when there is already some initial activity present in the wild type enzyme. For example, directed evolution has been effectively employed to increase the catalytic efficiency (Kcat/KM) for non- phosphorylated substrates of E.coli KDPG aldolase by a factor of 68.9

82 Chapter 6

Besides improving the activity for 2-deoxy-D-ribose, the activity for its mirror image 2-deoxy-L-ribose is of interest, to monitor the combined effects of charge and mutagenisis on the stereorecognition at C-2 and C-3. Moreover 2- deoxy-L-ribose is an interesting compound itself for studies with mirror image 10 DNA. We note that the aldol reaction of either D- or L-glyceraldehyde with acetaldehyde is catalysed by wild type DERA with the same rate.1 In this latter case the selectivity of DERA for the configuration at C-3 has not been studied. Depending on (S)- or (R)-selectivity either 2-deoxyribose or 2-deoxyxylose will be formed. From D-glyceraldehyde the D enantiomers are accessible, from L- glyceraldehyde the L sugars are accessible (Figure 6.2).

CHO CHO H OH HO H

CH2OH CH2OH D-Glyceraldehyde L-Glyceraldehyde

O O O O (S)-sel (R)-sel (S)-sel (R)-sel

CHO CHO CHO CHO H H H H H H H H H OH HO H H OH HO H H OH H OH HO H HO H

CH2OH CH2OH CH2OH CH2OH

-2-deoxy-D-ribose2-deoxy-D-xylose 2-deoxy-L-xylose 2-deoxy-L-ribose

Figure 6.2 Stereochemistry of the possible C5 sugars produced by DERA catalysed aldol reaction with glyceraldehyde and acetaldehyde

In this Chapter, the directed evolution of DERA to increase its activity for non- phosphorylated substrates and the production of such proteins by means of fermentation are described. The resulting variants are characterized with respect to their activity with phosphorylated and non-phosphorylated sugars. Moreover the product selectivity of the variants is studied in the aldol reaction of rac-glyceraldehyde or D-glyceraldehyde and acetaldehyde

83 Production and Optimization of 2-deoxyribose-5-phosphate Aldolase

Results and discussion

Isolation, cloning and expression of the gene encoding for DERA

The gene encoding for DERA was succesfully cut out and amplified from purified genomic DNA from E. coli. A clear band at around 800 base pairs, corresponding with the DERA gene length, was found in the electrophoretic analysis (Figure 6.3A). After gel purification of this band, digestion, ligation with three different vectors and transformation into E. coli, the presence of the DERA gene was proven by electrophoretic analysis of the PCR product from these transformed cells (Figure 6.3B). Sequence analysis of the 800 bp. fragments identified the correct sequence. The overexpression level of the gene was tested by growth of the transformed cells followed by various methods of induction. A clear very strong band around 28 kD was observed in the SDS-PAGE of the lysate from these induced cells (Figure 6.4), but was much less evident in the control line. Therefore a successful isolation of the gene as well as an efficient expression system for the production of DERA has been obtained.

800 bp 800 bp

A B

Figure 6.3 Electrophoresis of the amplified DERA gene (A) and transformed E.coli cells (B)

84 Chapter 6

Figure 6.4 SDS- PAGE of the expression of the WT-DERA gene after induction with 1 mM IPTG (line 1: induced 30°C 4 h.; line 2: control; line 3: induced 37°C 4 h 37 °C 4 h; line 4 induced 37°C overnight)

Production, fermentation and purification of wild type DERA

The fermentation of the cells overexpressing wild type E. coli DERA performed on a 11 L scale yielded 950 g. of wet cells. After cell disruption by means of a French-press the lysate had a total activity of 1,080,000 units* (98,000 units/liter of culture). The crude cell lysate exhibited a profound instability at room temperature and storage in the fridge (4 °C). Therefore lyophilisation was applied, yielding 100 g of solid with a total activity of 700,000 units (65 % of the activity of the lysate). This preparation has not shown any deactivation over 2 years of storage at -18 °C. The previously reported fermentation of DERA using the commercially available E. coli DH5α

* 1 unit (U) will cleave1 µmol of DRP into glyceraldehyde-3-phosphate and acetaldehyde per minute

85 Production and Optimization of 2-deoxyribose-5-phosphate Aldolase

strain yielded 124,000 units out of a 6L fermentation (21,000 units/liter of culture),3 which was, moreover, difficult to reproduce.11

Directed Evolution

The first generation library contained 4500 variants of shuffled semi-synthetic

DERA genes. By screening for activity in the retro-aldol reaction of 2-deoxy-D- ribose (D-DR) towards glyceraldehyde and acetaldehyde, 71 variants (1.5 % of the population) with substantial improved activity over the wild type enzyme were identified. These hits were used for DNA shuffling to create the second generation library which contained 1500 variants. Upon screening on activity in the same reaction, this second generation library yielded 90 hits (6 % of the population). All of the variants identified as hits in this screening were retested in triplo and compared with negative and positive controls. On the basis of these results two variants from the first library (DERA1-1 and DERA1-2) and four from the second (DERA2-1 to DERA2-4) were selected. The selected variants were produced by fermentation and further characterized. No sequence information is available for these variants at the moment of writing.

Characterisation of the DERA variants

Wild type DERA and the selected variants from directed evolution were characterised according to their activities in the retro-aldol reaction of 3 different substrates. Activities were determined for: 2-deoxy-D-ribose-5- phosphate (DRP) being the natural substrate of DERA, 2-deoxy-D-ribose (D- DR) being the substrate that is used for screening during directed evolution and 2-deoxy-L-ribose (L-DR) being the opposite enantiomer (Table 6.2).

The main goal of the directed evolution experiments with DERA was to increase its activity for non-phosphorylated substrates with D-DR as model compound. After one round of DNA shuffling a variant of DERA was obtained with a 5.6 fold increase in activity over the wild type enzyme for this substrate (Table 6.2 entry 2). After a second round of DNA shuffling a variant was

86 Chapter 6

obtained with an activity that was nearly 8 times higher than that of the wild type enzyme (Table 6.2 entry 5). The previously reported rational mutagenesis efforts with the mutation Ser238Asp in the phosphate binding pocket resulted in a 2.5 fold increase in activity for D-DR over the wild type enzyme.5,6 By using directed evolution a higher improvement in activity was reached in only one round.

Table 6.2 Activities in the retro-aldol reaction of 2-deoxy-D-ribose-5-phosphate (DRP), 2- deoxy-D-ribose (D-DR) and 2-deoxy-L-ribose (L-DR) of wild type DERA and the selected variants from 2 generations of directed evolution

Activities (µmol/min/g enzyme) DERA Variant D-DRPa D-DRb L-DRb Wild type 6483 18 5.0 DERA1-1 8398 103 8 DERA1-2 1369 59 4 DERA2-1 935 104 93 DERA2-2 344 144 51 DERA2-3 3321 122 37 DERA2-4 7132 96 35 a: Reaction conditions: Tris buffer pH 7.6, 1.9 mM substrate, 5 U/ml ADH, 0.16 mM NADH b Reaction conditions: Tris buffer pH 7.6, 60 mM substrate, 5 U/ml ADH, 0.16 mM NADH

The increased activity of the variants for the non-phosphorylated substrate D- deoxyribose does not automatically imply a decrease in activity for the natural substrate D-DRP. Although the activity of wild type DERA for its natural substrate was already very high, improvement was still observed in DERA1-1. This enzyme was also the best variant in the first generation of evolution in activity for D-DR. In the second generation a variant is present (DERA2-4) with also a 5-fold improved activity for D-DR and with a comparable activity for D-DRP to the wild type. With all the other variants the activity for the natural substrate decreased drastically whereas the activity for D-DR was improved in general. This observation was most pronounced for DERA2-2, which was the best variant for D-DR (8-fold improvement) and had the lowest activity for D- DRP (20-fold decreased).

87 Production and Optimization of 2-deoxyribose-5-phosphate Aldolase

The wild type enzyme and the selected variants from both generations also showed activity for L-DR. This is expected, since the aldol reaction of L- glyceraldehyde with acetaldehyde is known to be catalysed by wild type DERA. The activities for the L enantiomer were significantly higher in the second generation variants. This can be due to an improved acceptance of L- glyceraldehyde or by a change in the enantioselectivity of the newly-formed stereogenic centre from mainly S configuration to partly R (Figure 6.2). This will be clarified in more detail in the next paragraph.

Product distribution of the DERA variants

The product distribution, and therefore the selectivity of the enzyme, was studied in more detail by investigating the ratio between 2-deoxyribose and 2- deoxyxylose produced from the DERA-catalysed aldol reaction with acetaldehyde and rac-glyceraldehyde or D-glyceraldehyde (Table 6.3 and

Figure 6.2). For both C5 sugars the ratio of the α to β anomer, determined by 1 H NMR, is 1. 2-deoxy-D-ribose is for 72% present in the pyranose conformation. For 2-deoxy-D-xylose the furanose conformation could not be detected by 1H NMR.

Table 6.3 Product selectivity of the DERA catalysed aldol reaction with rac-glyceraldehyde or D-glyceraldehyde and acetaldehyde

DL-Glyceraldehyde (%) D-Glyceraldehyde (%) D-DR D-DX L-DX L-DR D-DX D-DR Wild type 49.5 0.5 37.5 12.5 0.9 99.1 DERA1-1 49.7 0.3 37.7 12.3 0.5 99.5 DERA1-2 49.7 0.3 37.7 12.3 0.6 99.4 DERA2-1 49.7 0.3 34.7 15.3 0.6 99.4 DERA2-2 49.6 0.4 37.6 12.4 0.8 99.2 DERA2-3 49.8 0.2 40.8 9.2 0.3 99.7 DERA2-4 49.8 0.2 42.8 7.2 0.4 99.6

Reaction conditions: 0.5 ml Tris-HCl (50 mM, pH 7.6), 100mM glyceraldehyde, 100mM acetaldehyde, 25 mg enzyme, reaction time 24 h, quantitative yield L-DX: 2-deoxy-L-xylose D-DX: 2-deoxy-D-xylose L-DR: 2-deoxy-L-ribose D-DR: 2-deoxy-D-ribose

88 Chapter 6

When DL-glyceraldehyde is used as aldol acceptor more or less 40% of the product is 2-deoxyxylose and 60% 2-deoxyribose. There was no significant difference in this product distribution between the DERA mutants. In the case of D-glyceraldehyde, in contrast hardly any 2-deoxy-D-xylose was formed (Table 6.3). The newly generated chiral center of wild type DERA catalysed aldol reactions is known to have the (S) configuration and therefore 2-deoxy-

D-ribose is the expected product. This chiral preference did not show any significant change between the variants resulting from directed evolution. Hence the selectivity of the enzyme is not affected by the mutations that were beneficial for activity.

When the ratio between 2-deoxyribose and 2-deoxyxylose found with D- glyceraldehyde is combined with the same ratio using DL-glyceraldehyde as substrate (at full conversion), the distribution of enantiomers in this case can be elucidated (Table 6.3). Because there is hardly any 2-deoxyxylose formed out of D-glyceraldehyde, the 2-deoxyxylose formed in the case of DL- glyceraldehyde is formed from the L enantiomer and is therefore mainly 2- deoxyx-L-xylose. The observed amount of 2-deoxyribose exceeded 50% when using the racemic substrate, which had to originate from L-glyceraldehyde.

This indicates that the (S) selectivity using the L enantiomer of glyceraldehyde is lower than for the D enantiomer. For D-glyceraldehyde the enzymes are

>99% (S) selective for the configuration at C-3, while for L-glyceraldehyde this varies from 69% for DERA2-1 to 86% for DERA2-4.

Conclusion The use of directed evolution as a tool to improve the substrate specificity of DERA for non-phosphorylated compounds has proven successful. Improved expression methods have been applied to increase the protein yield from fermentation substantially. In two rounds of DNA shuffling a 8-fold increase in the activity for the non-phosphorylated substrate 2-deoxy-D-ribose was obtained. All variants still had a high selectivity (>99%) for creating the new

89 Production and Optimization of 2-deoxyribose-5-phosphate Aldolase

stereogenic center in (S)-configuration when using D-glyceraldehyde as substrate. For L-glyceraldehyde this selectivity ranged from 69-86%.

Experimental part- Material and methods

Instruments and materials

All dilutions and liquid handling in the high throughput screening (HTS), if not performed with automated pipettes, were robotically automated with the Beckman Coulter Multimek 96/384-Channel automated pipettor and the Clontech TECAN Genesis RSP Robot. Colony picking was robotically automated by using a Genetix Q-bot 240V colony picker. Fluorometric activity assays were performed using a SpectraMax Gemini plate reader. Molecular biology kits have been used from Qiagen, Invitrogen and Zymoresearch. 1H and 13C NMR spectra were obtained on a Varian Unity Inova-300 instrument or on a Varian VXR-400S instrument. UV activity measurements were carried out on a Varian Cary 3 Bio UV-Visible Spectrophotometer at a wavelength of 340 nm. HPLC analysis was performed with two serial placed Phenomenex® 8 x 300 mm 8µ Rezex Monosaccharide column with water as eluent. A flow rate of 0.6 ml/min and a column temperature of 85 oC were used. RI detection was performed on a Showa Denko K.K Shodex RI SE-51. Before the samples were analysed by HPLC, precipitated enzyme was first spinned down and the supernatant was filtered through a 10kDa filter (Amicon Millipore Microcon®) to ensure removal of all enzyme. After that, the reaction mixture was directly injected on the HPLC.

Polymixin B sulfate was purchased from VWR; 2-deoxy-L-ribose was obtained from the Slovak Academy of Sciences. All other chemicals were purchased from Sigma-Aldrich.

90 Chapter 6

Methods

Gene isolation and overexpression

The 780 bp gene encoding for DERA (28kD) was isolated and amplified by means of standard PCR techniques from the purified genomic DNA of E. coli K12, with the primers ald01fw (5’-TCT AGA GGC CAG CCT GGC CAT AAG GAG ATA T), ald01in-fw (5’-AGA TAT ACA TAT GAC TGA TCT GAA AGC AAG C), ald03rev (5’-CTA ATG GTG ATG GTG ATG GTG GCC AGT TTG), ald04in-rev (5’-AGT TTG GCC TGA TAA GCG CAG CGC ATA C). The isolated fragment encoding for DERA was ligated into three different Codexis Inc. proprietary vectors (pCK110200, pCK110700, pCK110900) after digesting both first. Transformation of E. coli K12 (w 3110) with the resulting plasmids was achieved by electroporation as described further on. Sequence analysis of the insert indicated the mutation Ile83Phe.

Directed evolution

Directed evolution was performed using DNA shuffling techniques.12 Semi- synthetic libraries were generated using the wild type DERA gene from E. coli K12 as a backbone. Following PCR amplification with uracil and fragmentation of the backbone, diversity from homologues was spiked in using synthetic oligonucleotides in assembly PCR without primers. Full-length gene variants were rescued using nested primers. These genes were then digested with SfiI and cloned into the Codexis’ proprietary vector pCK110900. This is a low copy-number vector, with a P15A E. coli replication origin, a pACYC derived chloramphenicol (Cam) resistance, and a Lac(13)P/O promoter and lacI repressor. The gene library was used to transform E. coli W3110 for screening.

All hits identified in round one (71 variants) were used for shuffling as parents in round 2. Round 1 hits were grown overnight in a 96-well plate format in 1.2 ml LB medium (peptone 10%, yeast extract 5%, and NaCl 10% in deionized water at pH 7) with glucose (1%) and of Cam (30 µg/ml). The plasmid was

91 Production and Optimization of 2-deoxyribose-5-phosphate Aldolase

isolated from each clone using QiaPrep 96 Turbo BioRobot kit (Qiagen Inc.). Plasmid DNA was used as template for PCR amplification with uracil for each parent individually. PCR reactions were run at 50µl volume in 96-well plate format and purified using Zymoclean ZR-96 DNA Clean-up Kit (ZymoResearch Inc.). After purification, PCR products were run on 96-well E- gel (agarose 1%, Invitrogen Inc.) to verify amplification. Equal amounts of each PCR product were pooled and used for fragmentation and assembly. Full-length variants were rescued using nested primers. These variants were digested with SfiI and cloned into the vector pCK110900. The gene library was used to transform the E. coli W3110 for screening.

Transformation of the E. coli W3110 with the library genes was performed by electroporation. Electroporation cuvettes with a 2 mm gap were chilled on ice. E. coli strain W3110 (125 µl) and the mutated plasmids (1 µl), that resulted from DNA shuffling, were transferred to the bottom of the cuvette. The formation of bubbles was avoided. The electroporation was carried out at 200 Ω, 25 mF and 2.4 kV with a time constant in the range of 3-5 ms. 1 ml of SOC medium (tryptose 2%, yeast extract 0.5%, glucose 0.4%, NaCl 10 mM, KCl

2.5 mM, MgCl2 5 mM and MgSO4 5 mM) was added immediately after the electroporation and the resulting mixture was transferred to a culture tube. The tubes were shaken at 225 rpm at 37 oC for 30-60 min. The resulting cultures were then plated out on agar plates containing LB with Cam (30 µg/ml) and incubated for 20 h at 37 oC.

High throughput screening

Colonies from the agar plates after transformation were picked by means of the colony picker and transferred to the 96 well V-bottom library plates. An additional line of positive and negative controls was added afterwards. In the screening of the first generation the positive control was the previously generated E. coli mutant over expressing WT DERA, and the negative control was E. coli containing the plasmid without the insert of the DERA gene. In the

92 Chapter 6

screening of the second generation the positive controls were now two selected hits from the first library; Negative controls were the same. The library plates containing the E. coli variants in LB medium with Cam (30 µg/ml) and glucose (1%) were grown overnight at 37 oC, after which they were replicated in fresh LB with Cam (30 µg/ml) and allowed to grow for 4 hours at the same conditions. Afterwards the cells were induced with Isopropyl-β-D- thiogalactopyranoside (IPTG) for 4 hours at 30 0C. The concentration of IPTG in the wells was 1mM. Cells were collected by centrifugation of the 96 well plates for 10 min at RT (room temperature) and the supernatant was discarded. Lysis buffer (220 µl: Tris-HCl buffer (50mM, pH 7.6) containing lysozyme 1 mg/ml, polymixin B sulfate 2 µg/ml) was added to the plates and they were gently shaken for 1 h at RT. Cell debris was removed by centrifugation at 4,000 rpm for 10 min. Black 96-well assay plates suitable for fluorometric measurements were filled with assay fluid (100 µl: Tris-HCl buffer

(50mM, pH 7.6) containing either D or L 2-deoxyribose 120mM, alcohol dehydrogenase (ADH) 2.5 units/ml, NADH 0.32 mM). 90 µl of the cell lysate was used to test activity on 2-deoxy-D-ribose and 90 µl to test on 2-deoxy-l- ribose. Hits in the first generation library were defined as all variants that showed more than 1.5-fold increase in activity for the retro-aldol reaction with

2-deoxy-D-ribose over the wild type enzyme. Hits in the second generation library were defined as all the variants that showed an activity higher that 1.3 times the average activity of the positive controls for the retro-aldol reaction with 2-deoxy-D-ribose. 96 Well plates containing every identified hit in triple and with the previously described controls were retested for activity to confirm the HTS results. Library and hit plates containing the E. coli mutants were stored at –80 oC in medium supplemented containing glycerol (20%).

Fermentation and purification

Transformed E. coli K12, containing the overexpressed gene for DERA, was grown on minimal LB medium containing chloroamphenicol (1mg/l) on an 11 L scale. This provided 1 kg cell paste after centrifugation. After cell disruption by means of French press, and separation of cell debris the lysate was

93 Production and Optimization of 2-deoxyribose-5-phosphate Aldolase

lyophilized yielding a very stable preparation with a total of 700,000 units. No activity loss has been observed over a period of 2 years when stored at -18 °C.

Activity essay

The activity of the enzyme for a particular substrate was measured with the retro-aldol reaction of 2-deoxy-D-ribose-5-phosphate, 2-deoxy-D-ribose or 2- deoxy-L-ribose coupled to the reduction of the hereby-formed acetaldehyde with alcohol dehydrogenase (ADH) from yeast. To Tris-buffer (1.65 ml, 50 mM, pH 7.6) containing NADH (0.16 mM) in a 2 ml cuvette were added: a solution of alcohol dehydrogenase (100 µl, 100 U/ml) and a solution of the substrate (for 2-deoxy-D-ribose and 2-deoxy-L-ribose: 200 µl, 600 mM and for

2-deoxy-D-ribose-5-phosphate 200 µl, 19 mM). The assay was started by adding 50 µl of diluted sample containing the enzyme and monitoring the UV absorption at 340 nm and 20 °C. The molar absorption coefficient used for the consumed NADH was 6.22 l.mmol-1.cm-1.

Determination of product selectivity

To Tris-buffer (450 µl, 50 mM, pH 7.6) acetone (50 µl of 1M stock solution) and D-glyceraldehyde (4.5 mg) or DL-glyceraldehyde (4.5 mg) were added in a 2 ml GC flask. 25 mg DERA variant was added to the reaction. The reaction was monitored by HPLC on 2 sequentially placed Phenomenex® 8 x 300 mm 8µ Rezex Ca Monosaccharide column with water as eluent. A flow rate of 0.6 ml/min and a column temperature of 85 oC were used.

Identification of the products from the aldol reaction between glyceraldehyde and acetaldehyde

The product 2-deoxy-D-ribose was identified by comparison with the commercial standard and its NMR. The second product, showing at a lower retention time as the D-DR peak in the HPLC, was repetitively collected from

94 Chapter 6

the HPLC effluent until enough material was obtained for NMR analysis. This compound was identified as 2-deoxyxylose.

2-deoxy-D-ribose: 1 300 MHz H NMR (D2O) δ (ppm): 1.60-2.50 (many peaks, 2H(α+β)6-ring; 2H(α+β)5-ring), 3.5-4.5 (many peaks, (4H(α+β)6-ring; 4H(α+β)5-ring), 4.75 (dd) + 5.3 (t) (1H(α+β)6-ring), 5.5-5.6 (dd + t, 1H(α+β)5-ring). The α : β ratio is 1. The ratio of the 6 membered ring : the 5 membered ring is 3.5. 13 100 MHz C NMR APT (D2O) δ (ppm): 35.2; 36.5 (α+β CH2, 6-ring); 42.5;

42.6 (α+β CH2, 5-ring); 62.9; 64.1 (α+β CH2, 5-ring); 64.3; 67.5 (α+β CH2, 6- ring); 66.1; 67.9 (α+β CH, 6-ring); 72.4; 72.7 (α+β CH, 5-ring); 86.7; 87.3 (α+β CH, 5-ring); 93.1; 95.3 (α+β CH, 6-ring); 99.5; 99.6 (α+β CH, 5-ring). 2-deoxyxylose: 1 300 MHz H NMR (D2O) δ (ppm): 1.40-1.70 (many peaks, 1H(α+β); 1.9-2.3 (many peaks, 1H(α+β); 3.2-4.0 (many peaks, (4H(α+β)), 4.8 (dd) + 5.2 (t) (1H(α+β)). The α : β ratio is 1. The ratio of the 6 membered ring : the 5 membered ring is nearly undetectable. 13 100 MHz C NMR APT (D2O) δ (ppm): 37.9; 40.3 (α+β CH2); 62.9; 64.1

(α+β CH2); 64.0; 66.5.5 (α+β CH2); 71.2; 71.3 (α+β CH); 71.7; 71.8 (α+β CH); 93.1; 95.6 (α+β CH).

Synthesis of D-glyceraldehyde

D-glyceraldehyde was synthesized from 1,2:5,6-diisopropylidene-D-mannitol, which was obtained by protection of D-mannitol. D-mannitol (1g) was dissovled in acetone (5 ml) and 2,2-dimethoxypropane (1.6 ml) was added. The reaction was stirred and brought to reflux. Then stannous chloride (0.1 g) was added and the reaction was left stirring for 1 h. After filtration the solvent was removed and the resulting solids extracted with dichloromethane. This crude

1,2:5,6-diisopropylidene-D-mannitol was directly used in the next step.

The dichloromethane containing 1,2:5,6-diisopropylidene-D-mannitol was concentrated on the rotary evaporator till 15 ml. To this mixture was added saturated aqueous solution of sodium bicarbonate (400 µl), sodium

95 Production and Optimization of 2-deoxyribose-5-phosphate Aldolase

metaperiodate (1.3 g). The reaction temperature was controlled below 35 °C with a water bath. After 2 h of reaction time magnesium sulfate (500 mg) was added and stirring is continued for 20 min. The slurry was filtered and the residue washed thoroughly with dichloromethane. The organic layer was concentrated with a rotary evaporator and the remaining oil distilled in a kugl- rohr apparatus. The 1,2-O-isopropylidene-D-glyceraldehyde was collected at 60 °C at 25 mbar. Overall yield 35 %.

The above produce 1,2-O-isopropylidene-D-glyceraldehyde was deprotected by acid catalyzed hydrolysis in 5 ml methanol:water 1:1 and Dowex W50-X8 in H+ form (1g). The product was identified, using HPLC, by comparison with the commercial racemic glyceraldehyde.

References

1 Barbas, C.F.; Wang, Y.F.; Wong, C.H. J. Am. Chem. Soc. 1990, 112, 2013-2014 2 Chen, L.; Dumas D.P.; Wong, C.H. J. Am. Chem. Soc. 1992, 114, 741-748 3 Wong, C.H.; Garcia-Junceda, E.; Chen, L.; Blanco, O.; Gijsen, H.J.M.; Steensma, D.H. J. Am. Chem. Soc. 1995, 117, 3333-3339 4 Heine, A.; DeSantis, G.; Lutz, J.G.; Mitchell, M.; Wong, C.H. Science 2001, 294, 369 5 Liu, J.; DeSantis, G.; Wong, C.H. Can. J. Chem. 2002, 80, 643-646 6 DeSantis, G.; Liu, J.; Clark, P.; Heine, A.; Wilson, I.A.; Wong, C.H. Bioorg. Med. Chem. 2003, 11, 43-52 7 Liu, J.; Hsu, C.C.; Wong, C.H. Tetrahedron Lett. 2004, 45, 2439-2441 8 Wong, C.H.; Lui, J.; DeSantis, G. WO 03077868 9 Fong, S.; Machajewski, T.D.; Mak, S.S.; Wong, C.H. Chemistry and Biolog 2000, 7, 873-883 10 Urata, H.; Ogura, E.; Shinohara, K.; Ueda, Y.; Akagi, M. Nucleic Acids Res. 1992, 20, 3325-3332 11 Schoevaart, R. Applications of aldolases in organic synthesis, chapter 7, Thesis, TUDelft, 2000 12 Minshull, J.; Govindarajan, S.; Cox, T.; Ness, J.E.; Gustafsson, C. Methods 2004, 32, 416-427

DERA as Catalyst for Statin Precursors

Abstract The behaviour of DERA and its variants resulting from two rounds of directed evolution was studied in the commercially interesting sequential aldol reaction of chloroacetaldehyde and two molecules of acetaldehyde to yield 6-chloro-

2,4,6-trideoxy-D-allose. The activity of the best variant for this reaction was improved by a factor 4.2 over the wild type enzyme. The stability towards chloroacetaldehyde, which is an important parameter in this reaction, was lower with all of the variants than with the wild type enzyme.

DERA as Catalyst for Statin Precursors

Introduction The enzyme 2-deoxyribose-5-phosphate aldolase (DERA, EC 4.1.2.4) catalyses the aldol reaction between D-glyceraldehyde-3-phosphate as acceptor and acetaldehyde as donor substrate. Besides its natural substrates it is active with a wide variety of aldol acceptors and acetaldehyde, propanal, acetone and fluoroacetone as only known donors.1 The first reports of DERA catalysing a two step sequential aldol reaction of three aldehyde molecules to form 2,4,6-trideoxyhexoses date back from 1994.2 This type of structure can be identified in the side chain of all known statins, which are applied as HMG- CoA inhibitor in cholesterol lowering drugs (Figure 1.x, and 7.2).3,4 A number of substrates have been shown to give this 6 membered ring structure as final product in sequential DERA catalysed aldol reactions (Table 7.x).2,5,6

Table 7.1 Sequential aldol reactions catalysed by DERA R O OH O DERA OH O DERA OH OH O R R C2H4O R C2H4O 12 3 OH R Yield (%)

1a CH3 22

1b CH2OH <3

1c CH2OMe 65

1d CH2OMOM 25

1e CH2Cl 70

1f CH2N3 23

1g CH(CH3)2 13

1h CH2CH2COOH 80 Reaction conditions: 100 mM acceptor, 300 mM donor, 20 ml TEA buffer (100 mM, pH 7.3), 1000 U* DERA for 6 days under N2

Due to the formation of a cyclic hemiacetal, which is thermodynamically stablised, the reaction is driven towards this 6 membered ring and at the same time further polymerisation is prevented. The disappointing result for glycolaldehyde (1b, yield <3 %) is mainly due to the formation of a stabilised 5

* 1 unit (U) will cleave 1 µmol DRP into D-glyceraldehyde-3-phosphate and acetaldehyde per minute

98 Chapter 7

membered cyclic hemiacetal after the first aldol reaction step. Moreover, glycolaldehyde exists in aqueous solution in at least four different species of which only a small amount (4%) is free aldehyde.2,7 Protection of the alcohol group with a methylmethoxy group (MOM, 1d) increased the yield to 25 %, which is still low compared to the similar methoxyacetaldehyde (1c, 65%). In general electron withdrawing groups at the α-position of the aldehyde have a positive effect on the final yield of the trideoxyhexoses in these sequential aldol reactions. Chloroacetaldehyde, which is utilised in the DERA catalysed synthesis of the statin side chain, therefore has a good overall yield.

In Chapter 6, the reasons and methods for increasing the activity of DERA with non phosphorylated substrates have been discussed. A rational mutagenesis approach indicated that mainly the mutation Ser238Asp was beneficial for the activity towards acceptor substrates without a phosphate group present. This mutant has proven itself also more effective in the sequential aldol reactions of interest here than the wild type enzyme (Table 6.1).8,9,10,11 Besides having a higher activity for the substrates used in the sequential synthesis of trideoxyhexoses this mutation also exhibits activity for 3-azidopropionaldehyde that the wild type lacks. Due to a higher efficiency in the subsequent steps a more effective route towards the statin intermediate is possible when this substrate is used (Figure 7.1, step 3).

Recent screening of large mixed genomic libraries originating from environmental DNA provided a new DERA from an unknown source organism.13 This enzyme showed improved activity for the sequential aldol reaction of interest and also increased substrate tolerance over the E. coli DERA (specially towards chloroacetaldehyde). The protein sequence of this enzyme has close homology to DERAs from other Gram-positive organisms, but less than 30 % sequence identity to E. coli DERA was observed.

99 DERA as Catalyst for Statin Precursors

Cl Cl O O O OH O O Aldolase 1 Cl + 2 H H H2O OH OH

O O O OH + 2 S238D N 3 H H H2O OH 1: NaOCl, HOAc, H2O, RT, 3h1 2 0 2a: NaCN, DMF, 5%, H2O, 40 C

2b: Dimethoxypropane, DMF, catalytic H2SO4, then trimethylsilysdiazomethane

2c: Ra-Ni, MeOH, 50 PSIG H2 1 3a: MeONa, MeOH 3b: Camphorsulfonic acid, dimethoxypropane N3 3c: Ph3P, 3days O O H3C CH3 4: Knoevenagel condensation: benzaldenyde 3 5: Stetter reaction: methyl thiazolium catalyst, O O O

4-F-benzaldehyde, NEt3 6: pivalic acid, toluene/heptane/THF OH H2N OMe 7a: HCl, MeOH than NaOH 7b: Ca(OAc)2 +

O O O O O O CH3 5 4 F H3C H3C NHPh NHPh CH3 CONHPh CH3 CH3 Ph

F 2+ Ca F

H3C CH3 3H O OH OH O 2 O O O

N O- 7 N OMe

CH3 PhHNOC CH3 PhHNOC H3C 2 H3C

Figure 7.1 Synthetic route towards atorvastatin using wild type DERA and mutant S238D as catalyst for the homochiral side chain12,13,9

Huge improvements due to optimisation of the process parameters and the use of this latter enzyme for the production of the statin intermediate are reported (Table 7.2).13 The process optimisation is mainly focused on keeping the chloroacetaldehyde concentration as low as possible. This potent

100 Chapter 7

electrophile, which is able to couple with residues on the enzyme surface or in the active site, is responsible for a strong and irreversible enzyme deactivation.

The strong inhibitory effect of this substrate has been overcome by fed batch operation in which the concentration is kept as low as possible throughout the course of reaction. The application of the DERA obtained from genomic library screening increased the space-time yield further and lowered the catalyst load by a factor 2.

Table 7.2 Space-time yields for the production of (3R,5S)-6-chloro-2,4,6-trideoxy-erythro- hexose

Space-time yield Catalyst loading (g/l/h) (%) WT E. coli DERA, Batch2,5 0.08 20 WT E. coli DERA, Fed-batch13 25.3 4.8 Gen. Lib. DERA, Fed-batch13 30.6 2

Directed evolution experiments carried out by DSM to enhance the stability towards chloroacetaldehyde of E. Coli DERA resulted in variants that were up to 12 times more resistant to 100 mM chloroacetaldehyde and even showed 2 times the activity at 400 mM as wild type DERA did at 0 mM chloroacetaldehyde. 14 Mutation hotspots for improved stability were identified to be: Tyr49, Ala71, Asn80, Asp84, Gly97, Glu127, Ala128, Lys146, Lys160, Met185, Lys198. Met185 is close to the active site residue Lys167; Ala71 and Tyr49 are within 6-8 Å from the substrate; Asp84, Gly97 and Tyr49 might have an influence on the surface accessible to the solvent.

We applied directed evolution by DNA shuffling technology with the goal of enhancing the activity for non-phosphorylated aldol acceptors. This yielded a couple of mutants with up to seven times improved activity for the non- phosphorylated deoxyribose moiety compared to the wild type enzyme (Chapter 6). Their behaviour in the sequential aldol reaction for producing the homochiral side chain for statins is reported in this chapter.

101 DERA as Catalyst for Statin Precursors

Results and discussion

Reaction course with wild type DERA as catalyst

The final lactitol product (3e, Table 7.1) is formed through an initially formed C4 intermediate (2e). The intermediate product concentration increases to the point were the rate of its formation becomes lower than the rate of its conversion into the main product. Under the given conditions this point is reached after about 30 min reaction time (Figure 7.2). After about 22 h, the reaction levelled off at concentrations of 3 mmol/L chloroacetaldehyde (1e), 5.4 mmol/L 1-chloro-2-hydroxybutanal (2e) and 60.7 mmol/L 6-chloro-2,4,6- trideoxy-D-allose (3e). This corresponds with 92 % conversion of chloroacetaldehyde into the main product.

70.00 60.00 50.00 40.00 30.00 mmol/L 20.00 10.00 0.00 0 20 40 60 80 100 120 140 160 180 200 Time [min]

chloroacetaldehyde 1-chloro-2-hydroxybutanal 6-chloro-2,4,6-trideoxyhexose

Figure 7.2 Concentration profiles of chloroacetaldehyde, 1-chloro-2-hydroxybutanal and 6- chloro-2,4,6-trideoxyhexose during the sequential aldol reaction with chloroacetaldehyde (66 mmol/L) and acetaldehyde (132 mmol/L), with DERA WT (11 g/L) as catalyst

When the reaction was performed with one third of the catalyst load (3 g/L) it stopped after 2 h of reaction at approximately 30 % conversion (Figure 7.3). Upon the addition of a fresh portion of enzyme the reaction immediately resumed to approximately 95 % conversion of chloroacetaldehyde. Addition of more enzyme did not result in a further increase in conversion, indicating that equilibrium was reached at this stage.

102 Chapter 7

A higher conversion of the chloroacetaldehyde (95%) was obtained than reported earlier (70 %, Table 7.1), without any additional adjustments to the reaction system. In the reported experiments all aldolase has probably been deactivated before reaching equilibrium. The equilibrium conversion of 1e was high compared to similar reactants (Table 7.1), which we ascribe to product stabilisation through the formation of a pyranose hemiacetal. The electron withdrawing behaviour of the chlorine at the same time had a positive effect on the rate of reaction. The α/β anomeric ratio, determined by 1H NMR, is 0.25.

70.00 30 mg enzyme added to start 60.00

50.00 60 mg enzyme

40.00

30.00

20.00 50 mg enzyme

10.00

Chloroacetaldehyde [mmol/L] Chloroacetaldehyde 0.00 0 500 1000 1500 2000 2500 3000 3500 Time [min]

Figure 7.3 Determination of the equilibrium conversion of chloroacetaldehyde in the sequential aldol reaction between chloroacetaldehyde (66 mmol/L) and acetaldehyde (132 mmol/L) with DERA as catalyst

Activities of the DERA variants in the sequential aldol reaction

The activities in the sequential aldol reaction of chloroacetaldehyde and two molecules of acetaldehyde were determined for wild type DERA and the six variants originating from two rounds of directed evolution (Chapter 6, Table

7.3). The activities for 2-deoxy-D-ribose (D-DR) are presented as well, because this was the substrate for which the variants were selected during evolution.

103 DERA as Catalyst for Statin Precursors

Table 7.3 Activities of DERA variants in the sequential aldol reaction of chloroacetaldehyde and two molecules of acetaldehyde

Activity (µmol/min/g enzyme) DERA Variant Sequential aldol 2-deoxy-D-ribose Wild Type 49 18.3 DERA1-1 211 103.0 DERA1-2 103 59.3 DERA2-1 130 104.3 DERA2-2 <10 144.6 DERA2-3 180 122.1 DERA2-4 150 95.8 Reaction conditions: 66 mM chloroacetaldehyde, 132 mM acetaldehyde, RT, Tris buffer pH 7.6, enzyme: 3 g/L

Wild type DERA and the variants from the first generation of directed evolution (DERA1-1 and DERA1-2) showed roughly similar trends in improvement in the sequential aldol reaction when compared to the retro-aldol reaction of D- DR. Apparently, the protein modifications in these first generation variants benefit the activity of DERA towards both non-phosphorylated substrates in a similar fashion. This might lead to the assumption that general phosphate independence can be achieved by evolving DERA with its natural substrate minus the phosphate group. This does not hold though for the variants of the second generation. Although DERA2-1, 2-2 and 2-3 showed improved activity over DERA1-1 in the retro-aldol reaction of D-DR, they did not show the same improvement in the sequential aldol reaction. The most active DERA variant in the sequential aldol reaction with a 4.2 fold improvement over the wild type was found in the first generation (DERA1-1).

Stability towards chloroacetaldehyde

Previous research13 showed that chloroacetaldehyde, being a potent electrophile, is the main cause of the deactivation of DERA in the sequential aldol reaction. Because this is a very important parameter in designing the optimum reaction conditions and process operation for the sequential aldol reaction we studied the stability towards this compound for all variants resulting from directed evolution.

104 Chapter 7

The effect of the concentration of chloroacetaldehyde on the activity of wild type DERA became evident when monitoring its residual activity during exposure to 4 different concentrations of chloroacetetaldehyde in time (Figure 7.4). A concentration of 25 mM chloroacetaldehyde reduced the enzyme’s activity to approximately half of its original value in 3 h. Exposure to 50 mM chloroacetaldehyde rendered the enzyme almost completely inactive after only 2 h. At concentrations of 75 and 100 mM, DERA’s life time was reduced to less than 0.5h. These results indicate that the concentration of chloroacetaldehyde in the reaction has a serious effect on the lifetime of the enzyme and it should be kept at a minimum.

120

100

Residual activity (%)Residual activity 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 -20 Time of exposure [h] 25 mM 50 mM 75 mM 100 mM

Figure 7.4 Residual activities (%) of WT DERA (3 g/l) in chloroacetaldehyde concentrations of 25, 50, 75 and 100 mM after 1, 2 and 3 hours of exposure

The DERA variants that were obtained from the first generation of directed evolution for phosphate independence were incubated in 25 mM of chloroacetaldehyde. The residual activities were followed in time and compared with wild type DERA (Figure 7.5, Table 7.4). The deactivation of wild type DERA and the first generation variants in 25 mM chloroacetaldehyde showed a linear decreasing trend in time. Activity losses per hour in 25 mM chloroacetaldehyde were: 16% for the wild type, 19% for DERA1-2 and 24% for DERA1-1.

105 DERA as Catalyst for Statin Precursors

120

100

20 Residual Activity (%) Activity Residual

0 00.511.522.533.5 Time of Exposure [h] Wild Type Variant 2 Variant 1

Figure 7.5 Residual activities (%) of WT DERA and the 2 first generation variants in a chloroacetaldehyde concentration of 25 mM after 1, 2 and 3 hours of exposure

All variants from directed evolution were compared on the basis of their residual activity for the sequential aldol reaction after 2 h of exposure to 25 mM chloroacetaldehyde (Table 7.4) as well as the conversion that is reached after the enzyme load was completely deactivated.

Table 7.4 Sequential aldol reaction activities, residual activities and final conversions of DERA WT and its variants

DERA Variant Activity Residual Activity Final Conversion (µmol/min/g enzym) [%]a [%]b Wild Type 49 69 17.6 DERA1-1 211 52 48.5 DERA1-2 103 61 27.1 DERA2-1 130 32 31.3 DERA2-2 <10 n.d. <1 DERA2-3 180 26 39.1 DERA2-4 150 40 21.3

Reaction conditions: first 2 h of exposure to 25 mM of chloroacetaldehyde, than concentrations adjusted to 66 mM chloroacetaldehyde, 132 mM acetaldehyde, RT, Tris buffer pH 7.6, enzyme: 3 g/L a After 2 h of exposure to 25 mM of chloroacetaldehyde b Conversion obtained till total enzyme deactivation was reached

Wild Type DERA was more stable in 25 mM chloroacetaldehyde than all of its variants. But, due to its relatively low activity, the resulting final conversion

106 Chapter 7

that can be reached under the studied conditions is only 17 percent. The variants from the first generation still showed a reasonable residual activity whereas the variants from the second generation displayed especially poor stabilities (Table7.4). Though being less stable, owing to their higher activity all variants reached a higher conversion before an identical amount of enzyme was completely deactivated compared to the wild type enzyme under the same conditions. While activities for the sequential aldol reaction increased with evolution towards phosphate independence, the stability of the variants towards chloroacetaldehyde decreased rapidly the further away they were in evolution from the wild type. To obtain variants that have this characteristic of stability towards chloroacetaldehyde it should be included in the screening procedure. A combined approach could also be efficient where “hotspot mutations” for stability are combined with “hotspot mutations” for activity as a starting point for further evolution.14

Improvement of process operation

The optimum conditions in which the reaction should take place are hard to determine. When a batch mode reaction process is studied an optimum has to be found between two extremes. Either a low concentration of chloroacetaldehyde should be used, resulting in low product concentrations but a relatively good catalyst load of enzyme to product (wt/wt), or a high concentration of chloroacetaldehyde is used resulting in high product concentrations but with a relatively high catalyst load. For this type of substrate deactivation fed batch operation would be more suitable. A fed-batch reaction was performed with DERA WT. Acetaldehyde and chloroacetaldehyde were added to the reaction mixture with a feed based on 0.75 times the initial activity of DERA WT for 3 hours. A longer life-time of DERA was expected in this reaction because the chloroacetaldehyde concentration was kept at a minimum. However, the chloroacetaldehyde concentration started to increase from the beginning of the reaction and after 90 min all enzyme was deactivated. This behaviour could be caused by the

107 DERA as Catalyst for Statin Precursors

formation of a hotspot in chloroacetaldehyde concentration at the end of the addition funnel. A membrane reactor or addition of less concentrated feed could be a solution for better performance.

A second approach is to use an immobilised and hopefully stabilised enzyme preparation. A cross-linked enzyme aggregate of wild type DERA (Chapter 8) was studied under the same conditions as the free enzyme in the sequential aldol reaction. The CLEA had an activity of 66% (wt/wt) of the free enzyme and showed basically the same deactivation behaviour. After a first reaction cycle and separation of the catalyst it was reused in a second reaction cycle. The CLEA was totally inactive from the beginning of this reaction. Using a CLEA of DERA apparently did not solve the problem of instability towards chloroacetaldehyde.

Conlusion Regardless of a possible optimum concentration of chloroacetaldehyde for the reaction to take place in, improving the stability of DERA in chloroacetaldehyde is by far the most important issue for the economics of the sequential aldol reaction using this enzyme. High substrate tolerance and activity of the enzyme, ease of enzyme and product removal and fed batch operation are necessary for economic viability.

Experimental part- Material and methods

Instruments and materials

A 50 wt. % chloroacetaldehyde solution in water was purchased from Aldrich and acetaldehyde (99.5%) was purchased from Acros. Celite 521, purchased from Aldrich was used as a filter agent. Tris(hydroxymethyl)aminoethane (99.8%) was purchased from Fluka for the preparation of buffers. The retro-

108 Chapter 7

aldol activity assays were carried out with D-2-deoxyribose (97%), NADH (~98%), and ADH from Bakers Yeast (440 U/mg solid), purchased from Sigma. L-deoxyribose was purchased from the Slovak Academy of Sciences. Chemically pure D-sorbitol, purchased from Sigma, was used as an internal standard for HPLC analysis. All purchased reagents were used without additional purification. 1H and 13C NMR spectra were obtained on a Varian Unity Inova-300 instrument or on a Varian VXR-400S instrument. UV activity measurements were carried out on a Varian Cary 3 Bio UV-Visible Spectrophotometer at a wavelength of 340 nm. HPLC analysis was performed on a Phenomenex® 8 x 300 mm 8µ Rezex Monosaccharide column was with water as eluent. A flow rate of 0.5 – 0.6 ml/min and a column temperature of 60 – 75 oC were used. RI detection was performed on a Showa Denko K.K Shodex RI SE-51 and UV detection with a Shimadzu SPD-36A at 215 nm. Before the samples were analysed by HPLC, precipitated enzyme was first spinned down and the supernatant was filtered through a 10kDa filter (Amicon Millipore Microcon®) to ensure removal of all enzyme. After that, the reaction mixture was directly injected on the HPLC.

Methods

Synthesis of 6-chloro-2,4,6-trideoxyhexose

In a 100 mL round bottom flask with magnetic stirrer lyophilized WT DERA (1.25 g) was dispersed in an aqueous solution (100 ml) containing chloroacetaldehyde (66 mM) and acetaldehyde (132 mM). The reaction was allowed to proceed for 3 hours at RT. The enzyme was precipitated by the addition of acetone (200 ml) and subsequent stirring on ice for an hour. The solids were separated by centrifugation at 3000 rpm and the supernatant was filtered through celite. The filtrate was concentrated at reduced pressure to a volume of about 60 ml and extracted three times with ethylacetate (85 ml). The organic layer was dried over anhydrous sodium sulphate and concentrated under reduced pressure to provide 973 mg of crude lactol as

109 DERA as Catalyst for Statin Precursors

yellow oil. TLC analysis was performed on silica gel 60 F254 plates with an eluent of EtOAc : PE(40/60) in a ratio of 2:1. Products were visualized as blue spots by staining with acidic cerium molybdate, made by mixing 235 mL of water with 31 mL of concentrated sulphuric acid, 21 g of ammonium molybdate(VI) tetrahydrate and 1 g of ammonium cerium(IV) nitrate. The Rf of

1-chloro-2-hydroxybutanal was 0.36 and the Rf of 1-chloro-2,4,6- trideoxyhexose was 0.15. After column chromatography 821 mg of a colourless viscous liquid was 1 obtained as pure product. 300 MHz H NMR (D2O) δ (ppm): 1.49-1.95 (m, 4H), 3.54-3.73 (m, 2H), 4.07-4.16 (m, 1H, β), 4.18-4.24 (m, 1H, α), 4.29-4.35 (m, 1H, β), 4.39-4.48 (m, 1H, α), 5.11 (dd J=2.2, 9.7 1H, β), 5.25 (br t, J=3.9 13 1H, α). Ratio of α : β anomer = 1 : 4. 100 MHz C NMR APT (D2O) δ (ppm): α-anomer: 36.659 (E), 40.201 (E), 48.724 (E), 64.801 (U), 66.297 (U), 93.292 (U); β-anomer: 35.341 (E), 39.538 (E), 48.619 (E), 66.063 (U), 72.338 (U), 93.745 (U)

Sequential aldol reaction

In a 10 ml screw cap bottle acetaldehyde (1.32 ml, 1 M solution), chloroacetaldehyde (0.0838 ml, 50 wt% in H2O) and of D-sorbitol (40 mg) as internal standard were added to 8.6 ml of Tris buffer (50 mM, pH 7.6). The reaction was initiated by adding 110 mg of lyophilized WT DERA at RT. In the case of determining the initial activity of all the DERA preparations 25 mg of certain variant was added instead and samples were taken after 0, 20, 40 and 60 minutes. The reaction was continued overnight, after which the final sample was taken. The activities of DERA WT and the 6 variants were determined by calculating the consumption of chloroacetaldehyde in the first 20 min of the reaction. All reaction components except for acetaldehyde were followed by HPLC analysis on a Rezex Phenomenex Ca2+ column operating at 60°C and a flow rate of 0.6 ml/min H2O. Prior to injection samples were subjected to membrane filtration with Millipore centriprep with a 10 kD cut off to remove all dissolved protein.

110 Chapter 7

Stability of the DERA preparations

In a 10 mL screw cap bottle, a mixture was prepared containing: chloroacetaldehyde (25 mM), sorbitol (4 g/L) and the chosen variant of lyophilized DERA (30 mg). Separate reaction mixtures were prepared for 0, 1, 2, and 3 h of exposure time. After the indicated exposure time finished, acetaldehyde from a stock solution (1 M) was added to the reaction mixture to a final concentration of 50 mM. Residual activity for the sequential aldol reaction was measured by monitoring this reaction for 1 h by HPLC. The same method was used for determining the residual activity of the DERA preparations after exposure to chloroacetaldehyde concentrations of 50, 75, and 100 mM.

Fed batch operation

The fed batch reaction was carried out using a Metrohm 665 Dosimat system with a 1 mL dispenser. In a 100 mL round-bottom flask, WT lyophilized DERA lysate (1 g) was dispersed in 70 mL of water. For 3 h, chloroacetaldehyde (1.5 M) and acetaldehyde (3.1 M) were fed to the reaction mixture (Φ = 1.5 ml/h,

75% of Vini). The reaction was carried out at room temperature and followed for 4 hours by HPLC.

References

1 Machajewski, T.D.; Wong, C.H. Angew. Chem. Int. Ed. 2000, 39, 1352-1374 2 Gijsen, H.J.M.; Wong, C.H. J. Am. Chem. Soc. 1994, 116, 8422-8423 3 Öhrlein, R.; Baisch, G. Adv. Synth. Catal. 2003, 345, 713-715 4 Müller, M. Angew. Chem. Int. Ed. 2005, 44, 362-365 5 Wong, C.H.; Garcia-Junceda, E.; Chen, L.; Blanco, O.; Gijsen, H.J.M.; Steensma, D.H. J. Am. Chem. Soc. 1995, 117, 3333-3339 6 Gijsen, H.J.M.; Wong, C.H. J. Am. Chem. Soc 1995, 117, 7585-7591 7 Yaylayan, V.; Harty-majors, S.; Ismail, A.A. Carbohydr. Res. 1998, 309, 31-38 8 Liu, J.; DeSantis, G.; Wong, C.H. Can. J. Chem. 2002, 80, 643-646

111 DERA as Catalyst for Statin Precursors

9 Liu, J.; Hsu, C.C.; Wong, C.H. Tetrahedron Lett. 2004, 45, 2439-2441 10 DeSantis, G.; Liu, J.; Clark, P.; Heine, A.; Wilson, I.A.; Wong, C.H. Bioorg. Med. Chem. 2003, 11, 43-52 11 Wong, C.H.; Lui, J.; DeSantis, G. WO 03077868 12 Baumann, K.L.; Butler, D.E.; Deering, C.F.; Mennen, K.E.; Millar, A.; Nanninga, T.N.; Palmer, C.W.; Roth, B.D. Tetrahedron Lett. 1992, 33, 2283-2284 13 Greenber, W.A.; Varvak, A.; Hanson, S.R.; Wong, K.; Huang, H.; Chen, P.; Burk, M.J. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 5788-5793 14 Jennewein, s. (DSM), lecture at COST D25 workshop entitled “Reaction intermediates and Products, 2004

112

Cross-linked Enzyme Aggregates of DERA

Abstract The parameters that influence the production of an immobilised DERA preparation using CLEA technology were optimised. The optimum conditions were found to be: enzyme precipitation in a solution of ammonium sulphate (final saturation 80 %, pH 8), cross-linking with 10 % (v/v) of a polyaldehyde- dextran solution and reduction with 0.5 mg/mL of sodium borohydride at pH 8.5. In this way 86% of the initial activity of the soluble enzyme was recovered. The CLEAs produced with this method showed a much better stability against temperature and organic solvents than the free enzyme.

Cross-linked Enzyme Aggregates of DERA

Introduction A major challenge in industrial biocatalysis is the development of stable, robust and preferably insoluble biocatalysts. Besides the ease of separating an immobilised (insoluble) biocatalyst from the reaction medium, stability over a broad pH and temperature range as well as tolerance towards various organic solvents is highly desirable in industrial biocatalysis. These performance improvements as well as the separation benefits can often be gained by immobilisation.1,2,3 Initially the main interest in immobilisation of enzymes was to facilitate separation and reuse of the catalyst.4 Over the years the influence of the immobilisation techniques on the catalytic properties (e.g. activity, selectivity, stability etc.) has gained more and more attention as well.

Immobilisation of enzymes can be divided into two main classes: carrier- bound and carrier-free immobilisates.5 Initially most efforts were put into the first category, where the enzymes are bound to a, usually polymeric, carrier. There are many methods to immobilize biocatalysts on rigid supports such as, e.g., covalent immobilization, ionic immobilization or hydrophobic adsorption.6,7,8 One of the major drawbacks of these methods is that a large part (often over 90 %) of the biocatalyst consists of non-catalytic carrier material, decreasing the productivity (g product/g catalyst) drastically. Additionally a large loss of activity (frequently over 50 %) compared to the native enzyme is often observed.9,10 Many commercially used carriers are not sufficiently mechanically stable and therefore less suitable for stirred tank operations.

Nowadays, methods of immobilization of enzymes without the use of supports are gaining interest. All examples of carrier-free immobilised enzymes are based upon intramolecular cross-linking of proteins to form a macromolecular insoluble particle. Since these do not need any solid supports, the catalyst obtained only consists of protein and a small amount of cross-linking agent. Therefore, there is a much higher content of catalytic mass present in the immobilised enzyme preparation. Although the first reports about cross-linking

114 Chapter 8

proteins with retention of activity date back from the 1960’s, their application in biocatalysis dates only from the early 1990’s.

First reported were cross-linked enzyme crystals (CLECs)11,12,13 and cross- linked dissolved enzymes (CLEs).14,15 The major disadvantages of CLEs were a low activity recovery, low mechanical stability and a high dependence on a delicate balance of parameters during their preparation. CLECs turned out to be more attractive. They reached higher mechanical stability and also their stability under extreme conditions was much better.13,16,17,18 By choosing the right crystallisation, cross-linking and reaction conditions the performance of the CLECs could be engineered.19,20 The main disadvantage of CLECs is that the first step in their production, the crystallisation of proteins, is not an easy straightforward task and can be quite laborious. The more recently developed cross-linked enzyme aggregates (CLEAs) have overcome this latter problem. In this method, the enzyme is precipitated from an aqueous solution by adding a salt, water miscible organic solvent or a polymer such as polyethylene glycol (PEG). In a subsequent step, the aggregates of protein molecules are cross-linked with a bifunctional agent to form the CLEA.21 Studies towards the commercially attractive penicillin G acylase, used in semi-synthetic antibiotic synthesis, show comparable activity and stability for CLEAs compared to CLECs.21,22 A study with several lipases showed that by varying precipitation and cross-linking conditions the properties of CLEAs can be engineered and preparations can be obtained that are more active than the native enzyme.23 The effect of additives during aggregation and cross-linking can be used to beneficially modify the standard procedure in some cases to get more active or stable CLEAs.23,24 The influence of various precipitants, the glutaraldehyde concentration as commonly used cross-linker and the enzyme concentration and manner of addition on the final activity of the CLEAs of a wide variety of enzymes has been reported.25 Precipitation agents that showed a good recovered activity for all 12 enzymes studied were tert-butyl alcohol, PEG and saturated

(NH4)2SO4. Organic solvents that generally showed low activity recovery were

115 Cross-linked Enzyme Aggregates of DERA

methanol, DMF, DMSO. In some cases CLEAs were obtained with higher activities than the free enzyme they were derived from (hyperactivation). 23,25

With some enzymes, the conventionally used cross-linker glutaraldehyde causes a big loss in activity during the cross-linking step. This inactivation could be caused by the high reactivity of glutaraldehyde and its relatively small size, which makes it possible to access the amino acid residues in and near the active site. Enzymes that possess lysine residues that are essential for their catalytic activity seem to be more prone to this kind of deactivation. Recently milder cross-linking methods have been developed that result in high activity recovery with enzymes that deactivate when cross-linked with glutaraldehyde.26,27 A dextran-derived polyaldehyde proved a successful mild cross-linker for nitrilases, oxynitrilases and alcohol dehydrogenases.26 Other carbohydrate based glutaraldehyde analogues have been studied for cross- linking phytase, galactose oxidase and Candida antarctica lipase B.27

The CLEA approach is a highly attractive immobilisation procedure due to its simplicity and robustness. Due to DERA’s interesting product scope, but relatively low reaction rates with non-natural substrates, stability and reusability are very important with regard to its industrial applicability. Hence, immobilisation of DERA using the CLEA methodology was our preferred approach to develop a stable, reusable and robust biocatalyst with a high specific activity. In the present chapter, the optimisation of all the parameters that influence the development of this type of biocatalyst will be reported.

Results and discussion

Precipitation of DERA

The first step in the process towards a cross-linked enzyme aggregate is the precipitation of the enzyme, preferably without any loss of activity. The precipitation of DERA was performed with a series of known aggregating

116 Chapter 8

agents for proteins (Figure 8.1). The water miscible organic solvents: 1,2- dimethoxyethane (DME) and ethanol (EtOH), resulted in a low recovered activity in the aggregate, combined with little activity in the supernatant. Hence, these agents tend to irreversibly denature the enzyme. Acetone and tert-butyl alcohol were not able to precipitate the protein completely resulting in high activities in the supernatant (almost 60% for tert-butyl alcohol). Polyethylene glycol is reasonably efficient, but best is to use an aqueous solution of ammonium sulphate for which nearly full activity recovery was observed in the aggregate

Ammonium sulfate (66% saturated)

PEG

Acetone

t-BuOH

Ethanol

Dimethoxyethane

0 102030405060708090100 Recovered activity (%)

supernatant aggregate

Figure 8.1 Recovered activity after precipitation of “wild type” DERA

The pH has a slight influence on the precipitation with ammonium sulphate. Therefore it is advisable to adjust the pH if possible close to the pH optimum of the enzyme, which for this enzyme is 7.8.

117 Cross-linked Enzyme Aggregates of DERA

Cross-linking of the Aggregates

After aggregation, which was performed with an 80 % saturated ammonium sulphate solution at pH 8, the cross-linking step was studied. The often-used cross-linker glutaraldehyde and the recently reported dextran polyaldehyde26 were studied at different concentrations. When glutaraldehyde was used an optimum concentration of 100 mM was found. When lower concentrations were used activity was found in the supernatant, indicating incomplete cross- linking. The use of higher concentrations was at a penalty of 10-20 % activity loss (Table 8.1). This loss of activity may be caused by the reaction between glutaraldehyde and the lysine present in the active site of the protein (Figure 1.8). When dextran polyaldehyde was used as the cross-linking agent, nearly all of the activity was found in the cross-linked aggregates and none in the supernatant (Table 8.1).

Table 8.1 Recovered activity in “wild type” DERA CLEAs with different cross-linking agent concentrations

Glutaraldehyde concentration Supernatant CLEA (mM) (%) (%) 25 0.6 40 50 0 38 100 0.1 61 200 0 45 400 0.1 41 Dextran polyaldehyde Supernatant CLEA (µl per ml suspension) (%) (%) 100 0 100 200 0 100 500 0 87

The reduction of the Schiff’s’ base bonds

After cross-linking the aggregates with a dextran polyaldehyde solution, a reduction step is necessary to reduce the Schiff’s base bonds between the aldehyde groups and the lysine residues. Sodium borohydride or the relatively milder cyanoborohydride were used to perform this reduction. Table 8.2 shows the obtained activities for the preparations after different reduction reactions. The best procedure consists of reduction with a 0.5 mg/mL solution

118 Chapter 8

of sodium borohydride during 30 minutes at 4°C. After using this methodology the overall recovered activity was 86%. Keeping the temperature low during the reduction is of utmost importance. At room temperature only a total recovered activity of only 44 % could be obtained (Table 8.2).

Table 8.2 Residual activity in “wild type” DERA CLEA’s after different reduction conditions

Conditions Sodium Sodium Borohydride Cyanoborohydride 0.5 mg/ml at 4°C for 30 min 82 82 1 mg/ml at 4°C for 30 min 86 83 0.5 mg/ml at 20°C for 30 min 44 nda a Not performed

Stability of the biocatalyst

The 2-deoxyribose-5-phosphate aldolase from E. coli, used in these studies, has a low tolerance towards organic solvents and it deactivates rapidly at 60 °C.

t-BuOH 50%

Ethanol 50 % Co-solvents

Water

0 102030405060708090100

WT DERA1-1 DERA1-2 CLEA WT CLEA D1-1 CLEA D1-2

Figure 8.2 Residual activities after incubation for 24 h in various co-solvent systems

119 Cross-linked Enzyme Aggregates of DERA

It is shown that in the presence of 50 % organic co-solvent the cross-linked preparations have retained a moderate residual activity after 24 h incubation (Figure 8.2). The free enzyme preparations have become nearly inactive under the same conditions.

Free DERA was nearly completely deactivated after 6 h at 60 °C, whereas the CLEAs were much more stable and 20-50 % of the starting activity was recovered (Table 8.3).

Table 8.3 Residual activities of the DERA preparations after incubation at 60 °C for 6 h.

DERA preperation Residual activity (%) Wild Type 1 DERA1-1 2 DERA1-2 0 CLEA Wild Type 22 DERA1-1 51 DERA1-2 40 DERA1-1 and DERA1-2 are two variants resulting from directed evolution (Chapter 6)

Conclusion We have developed a new immobilised DERA catalyst using the CLEA technology as immobilization method. The parameters that influence the properties of the resulting CLEAs were optimised. The optimum conditions found to make the immobilisate are: enzyme precipitation in a solution of ammonium sulphate (final saturation 80 %, pH 8), cross-linking with 10 % (v/v) of a dextran polyaldehyde solution and reduction with 0.5 mg/mL of sodium borohydride at pH 8.5, 86% of the initial activity of the soluble enzyme can be recovered. The CLEAs produced with this method showed a much better stability against temperature and organic solvents.

120 Chapter 8

Experimental part- Material and methods

Instruments and materials

All chemicals were of analytical purity and obtained from Sigma-Aldrich. Alcohol dehydrogenase from yeast was obtained from Sigma-Aldrich as well. Wild type DERA (from E. coli) was made available by Codexis via over expression and fermentation. The two mutant preparations were obtained by directed evolution using gene shuffling followed by fermentation. Dextran with a molecular weight of 100-200 KD was purchased from Serva Feinbiochemica, sodium metaperiodate was obtained from Janssen Chimica.

Methods

Activity assay for DERA

The activity of the enzyme in the preparations was assayed with the retro- aldol reaction of 2-deoxy-D-ribose coupled to the reduction of the hereby- formed aldehyde with alcohol dehydrogenase (ADH) from yeast. To Tris- buffer (1.65 ml, 50 mM, pH 7.6) containing 0.16 mM NADH in a 2 ml cuvette were added: a solution of alcohol dehydrogenase (100 µl, 100 U/ml) and a solution of 2-deoxy-D-ribose (200 µl, 600 mM). The assay was started by adding 50 µl of sample containing the enzyme and monitoring the UV absorption at 20 °C and 340 nm. The molar absorption coefficient used for the consumed NADH was 6.22 l.mmol-1.cm-1.

Oxidation of dextran

1.65 g of dextran (100-200 KD) was dissolved in 50 mL water and 3.85 g sodium periodate was added. The final solution was stirred at room temperature during 90 minutes. After this time, the solution was dialysed 5 times, using a tube that retains molecules higher than 12 KD, using 5 litres of

121 Cross-linked Enzyme Aggregates of DERA

water each time at room temperature during more than 2 hours and under stirring.28

Preparation of the CLEAs

Precipitation of the aggregates. For these studies different precipitating agents such as organic solvents, polyethyleneglycol (PEG 10000 in water (500 mg/L)) or saturated solutions of ammonium sulphate at different pHs (adjusted by addition of NaOH or HCl), were assayed. A solution of DERA (100 µl, 28 mg/ml) was slowly added to the precipitant (400 µl). The resulting mixture was stirred in ice for 15 minutes. Precipitant and supernatant were separated by centrifugation. The precipitant was resuspended in Tris-buffer (50 mM, pH 7.6) and the activity of the supernatant and the resuspended precipitate was measured using the described activity assay.

Cross-Linking of the aggregates. To 1 mL of the previously formed aggregates of DERA different amounts of cross-linkers (glutaraldehyde or dextran polyaldehyde prepared as described before) were added. The suspension was left under magnetic stirring during 16 hours at 4° C. After this, supernatant and suspension were separated. The resulting CLEA was washed 3 times with 500 µL of Tris-buffer (50 mM, pH 7.6). The activity of both, supernatant and suspension was assayed.

Reduction of the CLEAs When dextran polyaldehyde was used as cross-linker the reduction of the Schiff’s bases formed have to be promoted. For this, different amounts of solid reducing agent (sodium borohydride and cyanoborohydride) were dissolved in 100 mM bicarbonate buffer at pH 8.5. 1 mL of these different solutions was added to the obtained solid CLEA. The suspension was left under stirring during 30 minutes at the specified temperature. After this time the suspension was washed 3 times with Tris-buffer (50 mM, pH 7.6).

122 Chapter 8

Stability Studies

Stability in organic Co-Solvents To 250 µl of a solution of DERA (20 mg in 1 ml of Tris-HCl buffer, 50 mM, pH 7.6) an equal amount of Tris-HCl buffer (50 mM, pH 7.6), ethanol or tert-butyl alcohol was added. These mixtures were left for 24 h shaking at room temperature. The residual activity of DERA in these mixtures was determined with the described activity essay.

Temperature stability A solution of DERA (500 µl of: 20 mg DERA in 1 ml of Tris-HCl buffer, 50 mM, pH 7.6) was incubated stirring at 60 °C for 6 hours. Afterwards the residual activity was determined with the described activity essay.

References

1 Clark, D.S. Trends Biotechnol. 1994, 12, 439-443 2 Cabral, J.M.S.; Kennedy, J.F. Thermostability of Enzymes, Edited by Gupta, M.N., Berlin: Springer Verlag, 1993, 163-179 3 Rocchietti, S.; Urrutia, A.S.V.; Pregnolato, M.; Tagliani, A.; Guisan, J.M.; Fenrnandez- Lafuente, R.; Terreni, M. Enzyme Microb. Technol., 2002, 31, 88-93 4 Messing, P.A. Immobilized Enzymes for Industrial Reactors, London, academic Press, 1975 5 Cao, L.; van Langen, L.; Sheldon, R.A. Curr. Opin. Biotechnol. 2003, 14, 387-394 6 Guisan, J.M. Enzyme Microb. Technol. 1998, 10, 375-382 7 Mateo, C.; Abian, O.; Fernandez-Lafuente, R.; Guisan, J.M. Biotechnol. Gioeng. 2000, 68, 98-105 8 Palomo, J.M.; Munoz, G.; Fernandez-Lorente, G.; Mateo, C.; Fernandez-Lafuente, R.; Guisan, J.M. J. Mol. Cat. B:Enzym. 2002, 19, 279-286 9 Bryjak, J.; Kolarz, B.N. Biochemistry 1998, 33, 409-417 10 Janssen, M.H.A.; van Langen, L.M.; Pereira, S.R.M.; van RAntwijk, F.; Sheldon, R.A. Biotechnol. Bioeng. 2002, 78, 425-432 11 Quiocho, F.A.; Richards, F.M. Proc. Natl. Acad. Sci 1964, 52, 833-839 12 Alter, G.M.; Leussing, D.L.; Neurath, H.; Vallee, B.L. Biochemistry 1977, 16, 3663-3668

123 Cross-linked Enzyme Aggregates of DERA

13 St Clair, N.L.; Navia, M.A. J. Am. Chem. Soc. 1992, 114, 7314-7316 14 Habeeb, A.F.S.A. Arch. Biochem. Biophys. 1967, 119, 264-268 15 Jansen, E.F.; Olson, A.C. Arch. Biochem. Biophys. 1969, 129, 221-228 16 Quiocho, F.A.; Richards, F.M. Biochemistry 1966, 5, 4062-4076 17 Tüchsen, E.; Ottesen, M. Carlsberg Res. Commun. 1977, 42, 407-420 18 Lee, K.M.; Blaghen, M.; Samama, J.P.; Bielman, J.F. Bioorg. Chem. 1986, 14, 202-210 19 Lalonde, J. Chemtech. 1997, 27, 38-45 20 Marnolin, A.L. Trends Biotechnol. 1996, 14, 223-230 21 Cao, L; van Rantwijk, F.; Sheldon, R.A. Org. Lett. 2000, 2, 1361-1364 22 Cao, L.; van Langen, L.M.; van Rantwijk, F.; Sheldon, R.A. J. Mol. Cat. B:Enzym. 2001, 11, 665-670 23 Lopez-Serrano, P.; Cao, L.; van Rantwijk, F.; Sheldon, R.A. Biotechnol. Lett. 2002, 24, 1379-1383 24 Wilson, L.; Illanes, A.; Abian, O.; Pessela, B.C.C.; Fernandez-lafuente, R, Guisan, J.M. Biomacromolecules 2004, 5, 852-857 25 Schoevaart, R.; Wolbers, M.W.; Golubovic, M.; Ottens, M.; Kieboom, A.P.G.;van Rantwijk, F.; ven der Wielen, L.A.M.; Sheldon, R.A. Biotech. bioeng. 2004, 87, 754-762 26 Mateo, C.; Palomo, J.M.; van Langen, L.M.; van Rantwijk, F.; Sheldon, R.A. Biotech. Bioeng. 2004, 86, 273-276 27 Schoevaart, R.; Siebum, A.; van Rantwijk, F.; Sheldon, R.A.; Kieboom, T. Starch/Stärke 2005, 57, 161-165 28 Drobchenko, S.N.; Isaevaivanova, L.S.; Kleiner, A.R.; Lomakin, A.V.; Kolker, A.R.; Noskin, V.A. Carbohydr. Res. 1993, 241, 189-199

124 Summary

Summary The demand for optically pure secondary alcohols, which has grown rapidly in recent years, has spurred the development of adequate enantioselective synthetic procedures. Although there are various chemical methods available, biocatalysts are increasingly applied due to their natural characteristic to discriminate between enantiomers. This thesis describes the synthesis of enantiopure chiral alcohols using enzymes from the three major classes that are synthetically useful: hydrolases, oxidoreductases and a lyase. Directed evolution aimed at altering the enzyme’s specificity and improving its characteristics to meet the requirements of organic synthesis takes a central role in some of these biocatalytic routes. Chapter 1 provides an overview of the three enzyme groups: lipases, ketoreductases and the aldolase 2- deoxyribose-5-phosphate aldolase (DERA). Directed evolution as an iterative method to alter protein structure and character randomly is described. A group of major pharmaceutical products, the statins, is reviewed.

In Part I the lipase catalysed esterification to resolve nitro-aldol adducts is described. In Chapter 2 a series of lipases is studied with various aliphatic and aromatic nitro aldol adducts. Acylation with succinic anhydride was much more enantioselective than commonly used acyl donors, such as vinyl acetate. A Study to increase the efficiency of the deracemisation procedure, by racemising the unwanted enantiomer, is described in Chapter 3. The unsatisfactory results are ascribed to the instability of the produced ester, which is prone to elimination.

In part II enantioselective carbonyl reductions are described using ketoreductases from different origin. In Chapter 4 the ketoreductase S1 from Candida Magnoliae is studied in the reduction of various β-ketoesters. The activity and selectivity and changes herein through 10 generations of directed evolution have been studied for ethyl-4-chloroacetoacetate and similar substrates. The activity for the substrate the enzyme was evolved for increased, by a factor 16 whereas the activities for closely related substrates

125 Summary

changed little. The enantioselectivity was high and not affected by directed evolution. In Chapter 5 a variety of ketoreductases from different microbial source were studied in the reduction of various β-ketoesters and aromatic ketones. Their tolerance towards organic co-solvents was studied together with the possibility of using 2-propanol as sacrificial substrate to regenerate the cofactor without the need of a second enzyme.

In Part III enantioselective aldol reactions using 2-deoxyribose-5-phosphate aldolase (DERA) are described. Chapter 6 describes the optimisation of the enzyme by means of directed evolution aiming for higher activities with non- phosphorylated substrates. The production of the wild type and its variants from directed evolution is described. The enzymes (wild type and variants) are characterised according to their activity for the natural substrate and both enantiomers of the non-phosphorylated analogue. The product distribution of the enzymes was investigated for both enantiomers of glyceraldehyde in the reaction with acetaldehyde. All variants remained (S)-selective, albeit with much lower selectivity in the case of L-glyceraldehyde compared to its natural substrate D-glyceraldehyde. Chapter 7 presents the results of the variants of DERA from directed evolution targeting phosphate independence in the synthesis of the atorvastatin side chain. Activities, course of reaction and the stability towards chloroacetaldehyde are reported. In Chapter 8 the production of an immobilised preparation, using CLEA (Cross Linked Enzyme Aggregate) technology, of DERA is reported. The parameters that influence the production of such immobilisates are optimised and this way 86% of the initial activity could be retained in the CLEAs.

126 Samenvatting

Samenvatting De vraag naar optisch zuivere secondaire alcoholen, die de laatste jaren snel is gegroeid, heeft de ontwikkeling van efficiënte enantioselectieve synthese procedures bevorderd. Hoewel er diverse chemische methodes beschikbaar zijn, worden biokatalysatoren in toenemende mate toegepast vanwege hun natuurlijke eigenschap onderscheid te kunnen maken tussen enantiomeren. Dit proefschrift beschrijft de synthese van optisch zuivere chirale alcoholen, gebruik makend van de drie belangrijkste enzymklassen, die synthetisch interessant zijn: hydrolasen, oxidoreductasen en een lyase. Directed evolution, met als doel het veranderen van de enzymspecificiteit en het verbeteren van zijn eigenschappen om tegemoet te komen aan de wensen van de organisch chemicus, neemt een centrale rol in in enkele van deze biokatalytische routes. Hoofdstuk 1 geeft een overzicht van de drie enzymgroepen: lipasen, ketoreductasen en de aldolase 2-deoxyribose-5-phosphate aldolase (DERA). Directed evolution als iteratieve methode om eiwitstructuur en eiwiteigenschap willekeurig te veranderen is beschreven. Een belangrijke groep farmaceutische producten, statines, is onder de loep genomen.

In Deel I wordt de lipase gekatalyseerde verestering van nitro-aldol adducten beschreven. In Hoofdstuk 2 is een serie lipasen bestudeerd met verscheidene aliphatische en aromatische nitro-aldol adducten. Acylering met barnsteenzuur anhydride bleek veel enantioselectiever te verlopen dan met veel gebruikte acyldonoren, zoals vinyl acetaat. Een studie ter verhoging van de efficiëntie van de deracemisatieprocedure, door rasemisatie van de ongewenste enantiomeer, is beschreven in Hoofdstuk 3. De tegenvallende resultaten worden toegewezen aan de instabiliteit van de gevormde ester, welke geneigd is te elimineren.

In Deel II worden enantioselectieve carbonyl reducties beschreven, gebruik makend van ketoreductasen van verschillende oorsprong. In Hoofdstuk 4 wordt de ketoreductase S1 van Candida magnoliae bestudeerd in de reductie van verscheidene β-ketoesters. De activiteit, selectiviteit en veranderingen

127 Samenvatting

hierin door 10 generaties van directed evolution, zijn bestudeerd voor ethyl-4- chlooracetoacetaat en gelijksoortige substraten. De activiteit voor het substraat waarvoor het enzym is ontwikkeld, is met een factor 16 verbeterd, terwijl de activiteit voor de gelijksoortige substraten weinig veranderde. De enantioselectiviteit was hoog en veranderde niet als gevolg van directed evolution. In Hoofdstuk 5 is een reeks ketorreductasen met een verschillende microbiële oorsprong bestudeerd in de reductie van verscheidene β-ketoesters en aromatische ketonen. Hun verdraagzaamheid voor organische oplosmiddelen is bestudeerd. Tevens is de mogelijkheid om 2-propanol te gebruiken als opofferingssubstraat ten behoeve van cofactorregeneratie zonder een tweede enzym bestudeerd.

In Deel III zijn enantioselectieve aldol reacties met DERA beschreven. Hoofdstuk 6 beschrijft de optimalisatie van het enzym door directed evolution, met als doel hogere activiteiten voor niet-gefosforyleerde substraten. De productie van het wildtype en de varianten uit directed evolution is beschreven. De enzymen zijn gekarakteriseerd ten opzichte van hun activiteit met het natuurlijke substraat en beide enantiomeren van het niet gefosforyleerde equivalent. De productverdeling van de enzymen is bestudeerd voor beide enantiomeren van glyceraldehyde in de reactie met acetaldehyde. Alle varianten bleven (S)-selectief, weliswaar met een veel lagere selectiviteit met L-glyceraldehyde als substraat vergeleken met D- glyceraldehyde. Hoofdstuk 7 geeft de resultaten weer van de varianten van DERA uit directed evolution in de synthese van de zijketen van atorvastatin. De directed evolution experimenten waren gericht op verhoging van de activiteit voor niet- gefosforyleerde substraten. Beschreven zijn: activiteit, reactie verloop en stabiliteit voor chlooracetaldehyde. In Hoofdstuk 8 wordt de vervaardiging van een geïmmobiliseerd preparaat van DERA beschreven, gebruik makend van CLEA (Cross Linked Enzyme Aggregate) technologie. De parameters die invloed hebben op de productie

128 Samenvatting

van een dergelijk geïmmobiliseerde biokatalysator zijn geoptimaliseerd. Deswege werd 86% van de initiële activiteit behouden.

129 Dankwoord

Dankwoord Aangekomen aan het einde van mijn promotieonderzoek zou ik graag iedereen willen bedanken, die op enigerlei wijze hun steun of een bijdrage hebben geleverd. Tevens wil ik iedereen bedanken die de afgelopen vier jaar hebben gezorgd voor jaren met veel goede herinneringen en hilarische momenten. Roger, bedankt voor de mogelijkheid mijn onderzoek in jouw groep te kunnen doen. Fred, ik wil jou bijzonder bedanken voor de hulp bij alle dagelijkse zaken: chemisch, technisch en organisatorisch. Er was geen chemisch probleem waar je geen oplossing voor had. Heel erg bedankt voor je bijstand in raad en daad tijdens de afgelopen jaren. Verder wil ik de gehele vaste staf bedanken voor alle noodzakelijke ondersteunende zaken. Mieke, bedankt voor het regelen van alles wat anders de soep in was gelopen. Adrie, bedankt voor de hulp met de vele niet- meewerkende LCMS-experimenten. Kristina, bedankt voor de hulp bij het opnemen van mijn iedere keer weer onmogelijke NMR-samples (tja, 50 keer een piek uit een hplc vissen levert nog niks op). Remco, bedankt voor de grote behulpzaamheid bij allerhande technische ellende, gecrashte pompen, GC’s etc. (en voor het goede gezelschap bij meerdere peukie-pauzes).

A vital part of my research project has been done in cooperation with Codexis (USA, CA). Without their support, technology and financing this thesis would have been impossible to exist in its current form. Most enzymes have been made available by means of technology and production facilities from Codexis. I am extremely grateful for this fruitful cooperation and support and have great memories of my visits to California (both work wise and social; those maxyhours are a great invention by the way). I want to thank all people at codexis for my pleasant visits and the efforts, patience and enormous help I got from everybody during my struggles with biochemistry and directed evolution. In particular I would like to thank Gjalt Huisman for the fruitful teleconferences, supervision and organisation of my work in the USA. Furthermore I would like to thank Vesna and Ken for their great supervision

130 Dankwoord

and support, which made the directed evolution experiments to a success. Without your help I would not have been able to get any enzyme or variants out of those bugs. Furthermore I thank everybody who helped during the screening of variants, making the robots dance and fermenting the organisms.

Ik wil iedereen die in de afgelopen 5 jaar heeft rondgesjouwd bij de vakgroep Biocatalysis and Organic Chemistry bedanken voor bijdragen variërend van slap geouwehoer rond de koffietafel tot ludieke nieuwe chemische ideeën. In het bijzonder wil ik alle studenten uit binnen en buitenland bedanken voor hun bijdrage aan verschillende delen van dit onderzoek. Rita thanks a lot for your hard struggle and persistence in the lipase work, which has led to a nice article in Tetrahedron Asymmetry. Chris bedankt voor het voortzetten van dit werk en met even grote vasthoudendheid te hebben getracht de DKR aan de praat te krijgen. Tevens waren jou verhalen over je taxi avonturen uiterst sappig om de dag een beetje mee op te leuken. Marisol bedankt voor het per ongeluk stuiten op de oplossing om een CLEA van DERA te maken. Martina thanks for the help in optimising and characerising the DERA CLEA’s and for a nice Italian influence. (ie. the homemade tiramisu was deliscious). Sander, ben ik bijzonder dankbaar voor je werk aan DERA als katalysator de lipitor intermediate synthese. Je uitstap naar California om de tweede generatie gedirigeerd te evolueren heeft een grote bijdrage geleverd aan de Hoofstukken 6 en 7. Kortom mooi werk en erg prettig samenwerken in ons laatste jaar. Furthermore a special thanks to the people I shared the office with: Luuk (the Force) van Langen, Rute, Alexandra and Anne for creating a nice atmosphere in a small, dark place where fresh air could not find a way in. Pedro thanks a lot for some interesting years, support, jam sessions, views of life and philosophical talks over coffebreak and at our “supertetsende” shared lab bench. Antonio, your visits to the “lab of Doom” were great. Some wild parties, Nederland-Duitsland in the beerhouse and our stay at the capital of Murcia (owned by you of course), were absolutely great times Dorian and me will never forget. See you soon, amigo! Zo zijn er ook veel niet-scheiko mensen die ik wil bedanken. Allereerst mijn ouders die het door hun financiële en emotionele steun mogelijk hebben

131 Dankwoord

gemaakt om het hele universitaire traject af te laten leggen en altijd enthousiast en geïnteresseerd zijn geweest in mijn werk en bezigheden. Mam ontzettend bedankt voor je onvoorwaardelijke steun en interesse de afgelopen 4 jaar (en in alle jaren daarvoor). Pap, jammer dat je mijn promotie niet hebt mogen meemaken en me ook zo’n Dr. titel hebt zien bemachtigen, maar ik weet zeker dat je onnoemelijk trots bent. Derk en Jelle, de vele blaatsessies in de keuken, een zelf gemaakte kaasbalfondue en alle andere hilarische uitspattingen, onder het genot van een pilsje, waren geweldig en daarnaast zeer waardevol voor het vergeten van de nodige mislukte experimenten en allerlei ander geneuzel. Een geweldige tijd dankzij jullie op de PB15. Dorian, bedankt voor al je liefde en steun en de interesse die je op kon brengen voor mijn nogal “abstracte” onderwerp gedurende de laatste en voor velen de meest stresserende periode van de promotie. Het laatste jaar van mijn promotie zou niet zo ongestressed zijn verlopen zonder jouw enthousiasme, reflexievermogen en de vele verhalen uit de “normale” wereld. We gaan nog een mooie tijd tegemoet. Bedankt, ik hou van je.

Iedereen bedankt,

132 Curriculum vitae

Curriculum Vitae Menno Jort Sorgedrager werd op 6 november 1975 geboren te groningen. In 1994 behaalde hij het Atheneum diploma aan het Maartenscollege te Haren (gr). In het jaar 1994 werd begonnen aan de opleiding Scheikundige Technology aan de Technische Universiteit Delft, waar werd gekozen voor de richting Bioproces Technologie. Deze studie werd afgerond met een onderzoek naar het gedrag van enzymen in ionische vloeistoffen in de vakgroep Biokatalyse en Organische Chemie onder supervisie van prof. dr. R.A. Sheldon. Vanaf september 2001 werd een promotieonderzoek verricht in dezelfde vakgroep onder supervisie van dr. Ir. F. van Rantwijk en prof. dr. R.A. Sheldon en in samenwerking met Codexis Inc. (USA, CA). Een deel van dit onderzoek is uitgevoerd bij Codexis in California, USA. De resultaten van dit onderzoek zijn in dit proefschrift beschreven.

133