Rational Redesign of Candida Antarctica Lipase B

Rational redesign of Candida antarctica lipase B

Anders Magnusson

Royal Institute of Technology School of Biotechnology Department of Biochemistry Stockholm, May 2005

Cover illustration: Thr40 (red), Trp104 (green), and Ser105 (yellow) of Candida antarctica lipase B, as positioned in the active site.

©Anders Magnusson

Royal Institute of Technology School of Biotechnology Department of Biochemistry AlbaNova University Center SE-106 91 Stockholm Sweden

Printed at Universitetsservice US AB Box 700 14 100 44 Stockholm Sweden

ISBN 91-7178-012-2

Anders Magnusson (2005). Rational redesign of Candida antarctica lipase B. Doctoral dissertation from the School of Biotechnology, Department of Biochemistry, Royal Institute of Technology, KTH, AlbaNova University Center, Stockholm, Sweden.

ISBN 91-7178-012-2

Abstract This thesis describes the use of rational redesign to modify the properties of the enzyme Candida antarctica lipase B. Through carefully selected single-point mutations, we were able to introduce substrate-assisted catalysis and to alter the reaction specificity. Other single-point mutations afforded variants with greatly changed substrate selectivity and enantioselectivity. Mutation of the catalytic serine changed the hydrolase activity into an aldolase activity. The mutation decreased the activation energy for aldol addition by 4 kJ×mol-1, while the activation energy increased so much for hydrolysis that no hydrolysis activity could be detected. This mutant can catalyze aldol additions that no natural aldolases can catalyze. Mutation of the threonine in the oxyanion hole proved the great importance of its hydroxyl group in the transition-state stabilization. The lost transition- state stabilization was partly replaced through substrate-assisted catalysis with substrates carrying a hydroxyl group. The poor selectivity of the wild-type lipase for ethyl 2-hydroxypropanoate (E=1.6) was greatly improved in the mutant (E=22), since only one enantiomer could perform substrate-assisted catalysis. The redesign of the size of the stereospecificity pocket was very successful. Mutation of the tryptophan at the bottom of this pocket removed steric interactions with secondary alcohols that have to position a substituent larger than an ethyl in this pocket. This mutation increased the activity 5 500 times towards 5-nonanol and 130 000 times towards (S)-1-phenylethanol. The acceptance of such large substituents (butyl and phenyl) in the redesigned stereospecificity pocket increases the utility of lipases in biocatalysis. The improved activity with (S)-1-phenylethanol strongly contributed to the 8 300 000 times change in enantioselectivity towards 1-phenylethanol; example of such a large change was not found in the literature. The S-selectivity of the mutant is unique for lipases. Its enantioselectivity increases strongly with temperature reaching a useful S-selectivity (E=44) at 69 °C. Thermodynamics analysis of the enantioselectivity showed that the mutation in the stereospecificity pocket mainly changed the entropic term, while the enthalpic term was only slightly affected. This pinpoints the importance of entropy in enzyme catalysis and entropy should not be neglected in rational redesign.

Keywords: Candida antarctica lipase B, rational redesign, secondary alcohols, substrate-assisted catalysis, S-selective, entropy, aldolase, stereospecificity pocket, oxyanion hole

©Anders Magnusson

Anders Magnusson (2005). Rational redesign of Candida antarctica lipase B. Doktorandavhandling från skolan för Bioteknik, Institutionen för Biokemi, Kungliga Tekniska Högskolan, KTH, AlbaNova Universitetscenter, Stockholm, Sverige.

ISBN 91-7178-012-2

Sammanfattning Den här avhandlingen beskriver rationell förändring av enzymet Candida antarctica lipas B. Genom väl utvalda punktmutationer införde vi substratassisterad katalys och ändrade reaktionsspecificiteten. Med andra mutationer ändrades substratselektiviteten och enantioselektiviteten drastiskt. Genom att mutera den katalytiska serinen ändrades hydrolysaktiviteten till aldolasaktivitet. Mutationen sänkte aktiveringsenergin för aldoladdition med 4 kJ×mol-1, samtidigt som aktiveringsenergin för hydrolys ökade så mycket att ingen hydrolysaktivitet kunde detekteras. Den här mutanten kunde katalysera aldoladditionsreaktioner som naturliga aldolaser inte kan katalysera. Mutation av treoninen i oxyanjonhålet visade hur viktig dess hydroxylgrupp var för stabiliseringen av övergångstillståndet. Den förlorade stabiliseringen kunde delvis återfås genom införandet av substratassisterad katalys. Den mycket låga enantioselektiviteten hos vildtypsenzymet mot etyl-2-hydroxy- propanoat (E=1,6) var klart förbättrad i mutanten (E=22), vilket berodde på att bara ena enantiomeren kunde utföra substratassisterad katalys. Designen av en mutant med en stor stereospecificitetsficka var mycket lyckad. En mutation av tryptofanen som sitter i botten av stereospecificitetsfickan skapade utrymme för sekundära alkoholer som måste placera en substituent större än en etylgrupp där. Mutationen ökade specificitets- konstanten 5 500 gånger för 5-nonanol och över 130 000 gånger för (S)-1-fenyletanol. Möjligheten att katalysera reaktioner med sekundära alkoholer med så stora substituenter som butyl och fenyl har ökat den biokatalytiska användbarheten av lipaser. Den ökade aktiviteten för (S)-1-fenyletanol bidrog starkt till att enantioselektiviteten mot 1-fenyletanol ändrades 8 300 000 gånger. En sådan stor förändring i enantioselektivitet finns tidigare inte beskriven i litteraturen. Den här mutanten är S-selektiv, vilket är unikt för lipaser. Dess enantioselektivitet ökade dessutom starkt med temperaturen och nådde E=44 vid 69 °C. Termodynamiska analyser av enantioselektiviteten avslöjade att det var entropibidraget som hade påverkats av mutationen i stereospecificitetsfickan, medan entalpibidraget var nästan opåverkat. Detta visar hur viktig entropin kan vara i enzymkatalys och att den bör tas med i beräkningen vid rationell design av mutanter.

Nyckelord: Candida antarctica lipas B, rationell design, sekundära alkoholer, substratassisterad katalys, S-selektiv, entropi, aldolas, stereospecificitetsficka, oxyanjonhål

Nothing shocks me. I’m a scientist.

Harrison Ford (1942-) as Indiana Jones

LIST OF PUBLICATIONS

This thesis is based on the following publications, which in the text are referred to by their Roman numerals:

I Creation of an Enantioselective Hydrolase by Engineered Substrate- Assisted Catalysis A. Magnusson, K. Hult, M. Holmquist Journal of the American Chemical Society 2001, 123: 4354-4355

II Carbon-Carbon Bonds by Hydrolytic Enzymes C. Branneby, P. Carlqvist, A. Magnusson, K. Hult, T. Brinck, P. Berglund Journal of the American Chemical Society 2003, 125: 874-875

III Creating Space for Large Secondary Alcohols by Rational Redesign of Candida antarctica Lipase B A. O. Magnusson, J. Rottici-Mulder, A. Santagostino, K. Hult ChemBioChem 2005, in press

IV An S-Selective Lipase was Created by Rational Redesign and the Enantioselectivity Increased with Temperature A. O. Magnusson, M. Takwa, A. Hamberg, K. Hult Angewandte Chemie 2005, accepted

TABLE OF CONTENTS

LIST OF PUBLICATIONS

INTRODUCTION 1 1-Biocatalysis 2 2-Optimizing enzyme catalysis 5 2.1 Optimizing the reaction conditions 6 2.2 Screening for new enzymes 7 2.3 Modifying the enzyme 8 2.3.1 Random methods 8 2.3.2 Rational redesign 10 3-Candida antarctica lipase B 12

RESULTS AND DISCUSSION 17 4-General methods 18 4.1 Mutagenesis 18 4.2 Protein expression and purification 19 4.3 Protein immobilization and active-site titration 20 4.4 Activity measurements 20 4.5 Enzyme kinetics 21 5-Reengineering the reaction mechanism 24 5.1 New reaction mechanism 25 5.2 Substrate-assisted catalysis 29 6-Reengineering the substrate selectivity 34 6.1 Broadening the substrate specificity 35 6.2 Redesigning the enantioselectivity 38

CONCLUDING REMARKS 44

ACKNOWLEDGEMENTS 46

REFERENCES 48

INTRODUCTION

Biocatalysis involves the use of enzymes to catalyze chemical conversions and is increasingly used in both industry and academia. Attractive properties of enzymes are their high substrate specificity, high enantioselectivity, and high activity in mild reaction conditions. The possibility of recombinant protein expression has increased their utility by decreasing their production cost. Biomolecular methods have also conferred the opportunity to improve the properties of enzymes when they do not meet the process requirements. Enzymes can even be altered to catalyze reactions unnatural to the native protein. Rational redesign is a knowledge-based approach to modify the catalytic performance of enzymes. The knowledge about specific enzymes and enzyme catalysis in general is constantly growing, increasing the scope of rational redesign. This thesis describes rational redesign of Candida antarctica lipase B (CALB) aiming at reengineering the reaction mechanism (Paper I and II) and at reengineering the substrate selectivity (Paper III and IV). The research was performed with the following goals; to increase the understanding of enzyme catalysis in general, to further elucidate the factors controlling the properties of CALB, to alter the catalytic performance of CALB, and to challenge our hypotheses for rational redesign.

1- Biocatalysis

The name enzyme was created in 1878 from the Greek words “en” and “zyme” meaning “in yeast”. Enzymes were identified during the studies of fermentation processes performed by Liebig, Pasteur, Fisher, and others (Ball, 2001). Enzymes catalyze chemical reactions in living systems, which means that they act in aqueous solutions and usually at neutral pH and mild temperature. Enzymes have been evolved by nature to be efficient catalysts in their natural environment and usually show very high reaction selectivity, chemoselectivity, substrate selectivity, regioselectivity, and enantioselectivity. More than hundred years ago, it was shown that enzymes can be used in organic solvents (Hill, 1898; Kastle and Loevenhart, 1900). Their high selectivity and activity in organic solvents make them ideal catalysts for chemical conversions. However, the interest for biocatalysis did not boost until the 1980’s with the general acceptance that enzymes can catalyze unnatural reactions efficiently in organic solvents (Klibanov, 1986). The development of recombinant protein expression has decreased the production cost of enzymes. The utility of enzymes has also increased thanks to biomolecular methods offering the possibility to improve their properties. Biocatalytic processes have been developed to meet the increased demand of enantiopure compounds from the pharmaceutical industry. Presently, enzyme catalysis is considered as an important tool for industrial synthesis (Schoemaker et al., 2003; Koeller and Wong, 2001). The development of biotransformations is mainly performed with hydrolases, followed by oxidoreductases, while the four remaining enzyme classes are involved in only 15% of the research (Figure 1-1) (Faber, 2004). The world market of 1.7 billion euros for industrial enzymes in 2004 was shared by a few main producers (Figure 1-1) (The Novozymes Report 2004).

1.Oxidoreductases 25% 6.Ligases 1% Novozymes 2.Transferases 5% DSM 5.Isomerases 2% (Denmark) (The Netherlands) 4.Lyases 7% 44% BASF 5% (Germany) 5% Genencor Intl. 3.Hydrolases 60% Others (USA) 28% 18%

Figure 1-1. Left: Percentage of biotransformation development performed with enzymes from a given class for the 1987-2003 period (Modified from Faber, 2004). Right: Distribution of the world enzyme market of 1.7 billion euros among the main producers in 2004 (Modified from The Novozymes Report 2004, www.novozymes.dk). 2 Biocatalysis

An extensive list of large-scale commercial biocatalytic processes can be found in a review by Schmid et al. and some examples are described below (Schmid et al., 2002). Examples of biotransformations using lipases can be found in chapter 3 of this thesis. A perfect example of a very selective biotransformation is the production of the low-calorie sweetener aspartame (Schmid et al., 2001). The process involves two enzymatic steps, aspartase catalyzes the addition of ammonia to fumaric acid with high enantioselectivity and thermolysin couples N-protected aspartic acid with phenylalanine with both high enantioselectivity and regioselectivity (Figure 1-2). In addition, both enzymes are used in the reversed direction compared to their role in nature. Aspartame is produced in thousand tons annually at DSM/Tosoh, using this process.

HO O HO O HO O O O O NH2 N O Aspartase Cl Protection H O NH OH 3 O OH O OH Fumaric Aspartic N-Protected aspartic acid acid acid

N-Protected aspartic acid H N O H + Thermolysin Deprotection N O NH2 O O O O O N O O H NH2 O O OH O OH Methyl phenylalanine N-Protected aspartame Aspartame

Figure 1-2. The biocatalytic process to produce the low-calorie sweetener aspartame involves two very selective enzymatic steps, one catalyzed by aspartase and the other by thermolysin. The figure is modified from Schmid et al (Schmid et al., 2001).

Biocatalytic processes produce 300 000 tons of acrylamide worldwide per year. The conversion of acrylonitrile to acrylamide is catalyzed by a nitrile hydratase (Figure 1-3). O

Nitrile hydratase N H O NH2 Acrylonitrile 2 Acrylamide

Figure 1-3. Biocatalytic conversion of acrylonitrile to acrylamide catalyzed by a nitrile hydratase. The enzyme is produced in Rhodococcus rhodochrous J1 and immobilized whole cells, with cell-associated nitrile hydratase, are used for the catalysis.

The nitrile hydratase is produced in Rhodococcus rhodochrous J1. Immobilized whole cells with cell-associated enzyme are used to catalyze the reaction (Tramper, 1996). All acrylonitrile is converted so no separation of substrate and product is needed. Acrylamide is widely used as a cement binder and 3 Biocatalysis solidification agent. It is the precursor of polyacrylamide, used for polymer and flocculant applications. The largest-scale biotransformation is the production of high-fructose corn syrup, which is produced in millions of tons (Cheetham, 2000). Glucose is isomerized to fructose by a glucose isomerase (Figure 1-4).

CHO CH2OH

H OH O

HO H HO H Glucose isomerase H OH H OH

H OH H OH

CH2OH CH2OH D-Glucose D-Fructose

Figure 1-4. Glucose isomerase catalyzes the isomerization of glucose to fructose in the production of high-fructose corn syrup.

In this process, the enzyme is also produced in cells (Streptomyces murinus) that are immobilized in the reactor. The high-fructose corn syrup is significantly sweeter than glucose and is used in soft drinks. There is no existing chemical catalyst that can perform this reaction. Enzymes are used in biocatalytic processes for their high selectivity and for their high activity in mild reaction conditions (pH, temperature, and solvent). Mild reaction conditions are crucial when the process involves sensitive substrates or products and are beneficial for the environment. Given the large expansion in the area of biocatalysis, there are still a vast number of catalytic processes that lack a suitable enzyme. This calls for the discovery of new enzymes and the improvement of enzymes currently not fulfilling the process requirements

4 2- Optimizing enzyme catalysis

Even if a lot of biocatalytic processes exist today, many of the chemical conversions in synthetic routes are not catalyzed by enzymes. To develop a biocatalytic process for a desired conversion many factors have to be considered, as illustrated in Figure 2-1.

Substrate Production Product Process Economics, Biocatalyst Product Integration Recovery Screening Reaction constraints, Downstream processing, Biodiversity, In situ product recovery High throughput screening

Biocatalyst Biocatalyst Application Characterization Stability, Immobilization, Reaction conditions, Cofactor regeneration, Sequence, Structure, Recycling Kinetics

Biocatalyst Biocatalyst Production Engineering Production method, Protein, cell, and Carbon source, metabolic engineering, Purification Modeling

Figure 2-1. Factors influencing the optimization of the biocatalytic process for a desired chemical reaction. Figure modified from Schmid et al. and van Beilen and Li (Schmid et al., 2001; van Beilen and Li, 2002).

The optimization of a biocatalytic process is obviously a very complex task that involves many steps. The work presented in this thesis deals with the modification of the performance of an enzyme. For many processes, the enzyme does not fully meet the requirements. The stability of the enzymes can be too low in the solvent and at the pH or temperature required for the reaction, and oxidizing or reducing conditions can degrade the enzyme. The performance of the enzyme can be insufficient regarding the activity, enantioselectivity, or substrate selectivity. For other processes there might not even exist an enzyme that can catalyze the desired reaction. Here, different approaches aiming to improve the performance of the enzymatic step and to maintain this performance over time (i.e. stability) will be discussed; the performance can be improved by optimizing the reaction conditions without direct modification of

5 Optimizing enzyme catalysis the catalyst, by screening the metagenome for novel enzymes, or by modifying the enzyme through random methods or rational redesign.

2.1 Optimizing the reaction conditions

The traditional way to improve biocatalysis has been to adjust the reaction conditions to be optimal for the enzyme. Both enzyme performance and stability can be improved by many factors including substrate engineering, solvent engineering, and immobilization techniques. The reaction conditions are rather simple to change and although the effects are hard to predict, they can have a large effect on the outcome of the biocatalytic step. The possibility to improve the catalysis by optimizing the reaction conditions is restricted when the enzymatic step is a part of a larger process. To be economically feasible, the conditions of the enzymatic step have to be compatible with the preceding and following steps. One example of an important parameter in the optimization of the reaction conditions is the choice of solvent. If hydrolysis activity is to be avoided, an organic solvent has to be selected. The solubility of the different reactants will have an effect on both activity and selectivity of the enzyme. The stability of the enzyme is also affected by the solvent; it usually decreases with increasing polarity of the organic solvent. The water activity (aw) of the solvent is very important for both the activity and stability of the enzyme (Halling, 1994). The stability of porcine pancreatic lipase at 100°C was demonstrated to be greatly influenced by the water activity (Carrea and Riva, 2000). The half-life was only seconds in an aqueous buffer and more than 12 hours in the near absence of water. Two rather intriguing properties of enzymes are the pH memory and molecular memory. When enzymes are lyophilized from an aqueous buffer they are believed to keep the protonation state when suspended in an anhydrous solvent. Similarly, when the enzyme is lyophilized with a molecule bound in the active site, which is later removed, the active site will keep that conformation in an anhydrous solvent (Klibanov, 2001). The imprinted conformation is easily lost if the enzyme is used in a solvent allowing protein mobility. The enzyme is more evenly distributed in an organic solvent and more easily recovered from the solvent if it is immobilized on a carrier. Many different carriers and immobilization techniques are available and the choice will influence the stability as well as the selectivity of the enzyme. The enantioselectivity of the lipase from Candida rugosa varied from 1.2 to over 200 depending on the choice of carrier (Fernandez-Lorente et al., 2001). This was explained by different conformations of the enzyme depending on the immobilization technique used. Diffusion limitation is sometimes lowering the performance of the biocatalytic step. The diffusion is influenced by many factors such as solvent, substrate concentration, solubility, and mixing. When the enzyme is immobilized on a carrier, the accessibility and concentration of the enzyme can 6 Optimizing enzyme catalysis influence the selectivity (Rotticci et al., 2000a). In commercial preparations of immobilized enzymes, the carrier is often loaded with a lot of enzyme. This will give the preparation a high activity, and the activity will remain the same even when the outer enzyme layer has been degraded. Unfortunately this will also give a lower selectivity as only the enzyme layer closest to the surface will “experience” the true substrate concentration, the enzyme molecules immobilized in the inner layers will only “see” the slow-reacting substrate. Temperature is a parameter that affects the biocatalysis in many aspects. An increased temperature usually increases substrate solubility and enzyme activity, but it decreases stability and often enantioselectivity of the enzyme. An unusual example of increased enantioselectivity obtained by changing the solvent and raising the temperature is given in Paper IV.

2.2 Screening for novel enzymes

Considering the huge amount of living organisms and all the different niches they strive in, it is believed that any imaginable chemical conversion can be catalyzed by an existing enzyme. In only one gram of soil, over 10 millions genes can be found. This is 100 times more than the number of genes sequenced from bacterial genome during the past ten years (Streit et al., 2004). The screening of the metagenome for the optimal enzyme is limited by the inability to cultivate most of the existing microorganisms and by the difficulty to select the enzyme with the desired properties. The traditional route to find new enzymes is to cultivate organisms and then screen pure strains for the desired enzymatic properties. A more recent method consists of the two following steps; firstly DNA is extracted from the relatively small part of the metagenome carried by cultivable organisms or DNA of poor quality is extracted from crude microbial samples, secondly this DNA is transformed into a host system and the enzymes are expressed and screened for the desired properties or the DNA is sequenced and candidates are selected based on homology modeling (Lorenz et al., 2002). In a nice piece of work, Rondon et al. identified several enzymes with different activities (Rondon et al., 2000). DNA was extracted from soil samples and over 1 Gbp of DNA were inserted into a bacterial host, resulting in over 28 000 clones. Screening of the clones for different activities was performed on a solid medium and mainly with substrates that would produce a chromophore upon reaction. The screening resulted in clones with DNase (one), antibacterial (one), lipase (two), and amylase (eight) activity, but no cellulase, chitinase, esterase, keratinase, protease, or hemolytic activity was detected. A new nitrilase was identified for the production of (R)-mandelic acid from hydroxy-phenyl-acetonitrile (Figure 2-2) (Schmid et al., 2001). The nitrile spontaneously racemizes in aqueous solution through the formation of benzaldehyde and hydrogen cyanide. The conversion of the (R)-nitrile to (R)- mandelic acid is catalyzed by the novel nitrilase. (R)-Mandelic acid is produced in thousands of tons annually at BASF using this process and is used as a chiral building block in pharmaceuticals production. 7 Optimizing enzyme catalysis

OH O OH OH O + HCN Nitrilase N N +2H2O NH3 OH (S)-hydroxy- (R)-hydroxy- benzaldehyde (R)-mandelic acid phenyl-acetonitrile phenyl-acetonitrile

Figure 2-2. The production of (R)-mandelic acid using a novel nitrilase.

About 5 000 strains from enriched cultures were screened for nitrilase activity, three novel nitrilases were found and one is used for the conversion shown in Figure 2-2. Enrichment of cultures showing nitrilase activity can be done using nitriles as the sole carbon or nitrogen source. In the search of a novel catalyst, many potential enzymes will be lost during the DNA collection, protein expression, and screening. This might seem as a big disadvantage, but not all candidates have to be found, not even the best one, as long as a sufficiently good enzyme is found.

2.3 Modifying the enzyme

By modifying an enzyme it is theoretically possible to create a suitable catalyst for any given process. This is done by random or rational mutations of the gene coding for the enzyme of interest. The mutations can improve the stability of the enzyme, the performance towards the natural reaction, or even change the enzyme to catalyze a new reaction. The introduction of new reaction activity in an enzyme has been called the holy grail of enzyme redesign (Penning and Jez, 2001). Even if the new catalytic activity was undetectable in the wild-type enzyme it is unlikely that it is completely unable to catalyze the reaction, unless the active site was totally rebuilt. The introduction of new reaction activity is rather an enhancement of a proficient activity. The improvement of an alternative activity, proficient activity, has been performed for several enzymes as reviewed by O’Brien and Herschlag (O'Brien and Herschlag, 1999). The random and rational approaches are often used iteratively. The random approach can yield knowledge to guide further improvement of the enzyme through rational alterations, and rational improvements are often followed by random mutations to fine-tune the catalyst.

2.3.1 Random methods The improvement of the enzyme properties through random alterations requires the gene (or genes) of interest, but essentially no knowledge about the catalyst. The random alterations in the gene can be introduced by error-prone PCR (polymerase chain reaction) (Cadwell and Joyce, 1992). In this method the gene is copied in vitro with an error rate that usually is selected to introduce, in average, one or a few mutations in the gene. Gene shuffling is another method to create random alterations (Crameri et al., 1998). This requires several related

8 Optimizing enzyme catalysis genes or several variants of one gene. The genes are cut in pieces and the pieces from different origins are randomly recombined to form new hybrid genes. The random alteration of the gene(s) usually creates thousands of variants. These genes are expressed and the proteins are screened for the desired properties. The selected variants can be subjected to consecutive rounds of random alterations and screening, until an enzyme with the required properties is created. This process is referred to as “directed evolution” or more accurately “molecular breeding”. The crucial step in this approach is the screening, which should be fast, simple, easy, and cheap. The screening for enzymatic activity is often performed with a substrate that will produce a chromophore or fluorophore upon reaction. While this is a great advantage for high-throughput screening, it directs the molecular breeding toward such substrates. When a screening process is developed one has to remember the first law of directed evolution “you only get what you screen for” (Arnold and Moore, 1997). The random mutagenesis by the error-prone PCR technique very rarely creates more than one change per codon. As a consequence not all the amino acids can be introduced at a given position, possibly preventing the creation of an improved variant. Random methods often find improved enzyme variants with mutations far from the active site. Enzymes have most of their amino acids far from the active site and random mutagenesis will consequently target those amino acids with a higher frequency. That explains why few variants with mutations close to the active-site are found, even if such mutations affect the performance of the enzyme to a larger extent (Park et al., 2005). One example of random mutagenesis was performed by Zhao and Arnold (Zhao and Arnold, 1999). The thermostability and temperature optimum of subtilisin E was increased by five rounds of molecular breeding, alternating random mutagenesis and gene shuffling. The hydrolytic activity was screened towards succinyl-Ala-Ala-Pro-Phe-p-nitroanilide by measuring the absorbance of the released chromophore. A total of 20 000 clones were screened and a mutant with a temperature optimum that was 17°C higher and a half-life at 65°C that was more than 200 times longer, compared to that of the wild-type protease, was found. Another example is the increased activity of a haloalkane dehalogenase, achieved by performing only two rounds of random mutagenesis and screening (Bosma et al., 2002). In the production of epichlorohydrin, which is used in the manufacture of epoxy resins and elastomers, the highly toxic waste product 1,2,3-trichloropropane is formed. It is desirable to have a microbial degradation of this toxic compound in environmental samples. Such degradation is limited by the first dehalogenation step, shown in Figure 2-3.

Cl Cl Haloalkane dehalogenase Cl Cl H2O Cl OH HCl

Figure 2-3. The optimized haloalkane dehalogenase catalyze the critical step in the degradation of the toxic compound 1,2,3-trichloropropane.

9 Optimizing enzyme catalysis

After the two rounds of molecular breeding, the specificity constant of the enzyme was increased by a factor of eight. The improved haloalkane dehalogenase was expressed in Agrobacterium radiobacter AD1. This bacterium strain was able to live on 1,2,3-trichloropropane as the sole carbon source and could be used for ground-water purification.

2.3.2 Rational redesign The improvement of an enzyme through rational redesign is a knowledge- based approach. The structure of the enzyme has to be known and the success rate increases if more information about the enzyme is available, such as reaction mechanism, catalytic residues, active-site conformation, and substrate specificity. In this approach, one or a few amino acids believed to affect the property to be improved are targeted. These are altered by site-directed mutagenesis to create one or a few mutants. Two methods to introduce a mutation of a specific amino acid are briefly described in General methods, section 4.1. The mutants are expressed and the enzyme variants are characterized to determine the success of the optimization of the catalyst. No screening is needed and the enzyme variants can be tested towards the desired reaction directly. Random methods often identify improved variants with mutations far from the active site. On the other hand, distant mutations are rarely made in rational redesign to improve the catalytic performance. It is too difficult to predict the effect of such mutations. To restrict the mutations to the vicinity of the active site is sometimes considered as a weakness of the rational redesign approach. An example of rational redesign is the conversion of an aspartate aminotransferase into an L-aspartate β-decarboxylase by three active-site mutations (Graber et al., 1999). The two reactions, transamination and decarboxylation, are described in Figure 2-4.

NH2 L-aspartate O NH2 O O CO2 + O β-decarboxylase O O + H+ O O O O aspartate O L-alanine L-aspartate oxaloacetate aminotransferase + + O NH2

O O O O

O O O O 2-oxoglutarate L-glutamate

Figure 2-4. Aspartate aminotransferase and L-aspartate β-decarboxylase activities of the wild- type and triple mutant of aspartate aminotransferase.

The wild-type enzyme has a transaminase activity that is 3 000 000 times larger than the decarboxylase activity. On the other hand, the triple mutant has a selectivity of 8 for the decarboxylation over the transamination reaction. The decarboxylation activity increased over 1000 times and the transamination

10 Optimizing enzyme catalysis activity decreased 20 000 times by the three mutations. This gives an impressing change of the reaction selectivity of 24 000 000 times. A very stable mutant of a thermolysin-like protease from Bacillus stearothermophilus was created by rational redesign (van den Burg et al., 1998). Eight amino acids were mutated to create rigidifying mutations such as Ala to Pro and Gly to Ala as well as one disulfide bridge. Thermostable variants were used as models when the mutations were selected. The half-life at 100 °C increased from less than half a minute for the wild-type to 170 minutes for the rationally designed mutant. The stability also increased towards denaturing agents, while the activity at 37 °C remained as high as that of the wild-type protease. The stabilizing effect of the rigidifying mutations was explained by reduced entropy of the denatured enzyme.

3- Candida antarctica lipase B

Lipase B from Candida antarctica (CALB) belongs to the enzyme class of hydrolases (E.C.3). It acts on ester bonds (E.C.3.1) of carboxylic esters (E.C.3.1.1). It is a triacylglycerol lipase (E.C.3.1.1.3) and hydrolyzes triacylglycerols to fatty acids, diacylglycerol, monoacylglycerol, and glycerol. Hydrolases are the enzymes most used in biocatalysis and lipases form an important part of this class. The largest user of lipases is the detergent industry, which contributes to 30% of the total enzyme market (Houde et al., 2004). A large number of the detergents sold all over the world contain the lipase from Thermomyces lanuginosa (formerly Humicola lanuginosa), which is produced by Novozymes and sold under several trade names. This lipase has an optimal activity at pH 10.5-11 and 40°C, and is stable in the harsh laundry conditions. It has broad substrate specificity and can degrade various kinds of fat stains found on fabrics. Lipases are also widely used in the food industry for processing fats and oils (Sharma et al., 2001). They can transform cheap substrates into high-value fats, for example in the production of cocoa butter. Cocoa butter is a desirable but expensive ingredient in chocolate, which gives it a nice gloss, melting temperature, and consistence (Houde et al., 2004). Cocoa butter substitutes having the desirable properties are produced through acyl-transfer reactions between cheaper fats catalyzed by the lipase from Rhizopus niveus (Amano). Lipases are also used in the pulp and paper, cosmetics, and pharmaceuticals industry and for the synthesis of fine chemicals (Sharma et al., 2001). CALB is usually mentioned among the most used lipases and the review by Anderson et al. is cited as reference for its applications (Anderson et al., 1998). Its interesting properties and great potential as catalyst are described there and in many other reviews, but there are few examples in the literature describing the use of CALB in industrial biocatalysis. One reason for the absence of reports is the rather expensive price of the available preparations of CALB; Novozym (Novozymes A/S) and Chirazyme (Roche Molecular Biochemicals). These preparations are mainly intended for the production of high-priced specialty chemicals (Kirk and Christensen, 2002). Another reason could be the reluctance of the companies to reveal their biocatalytic processes. One industrial process using CALB is the production of isopropyl myristate, which is used as a component in cosmetics (Houde et al., 2004).

12 Candida antarctica lipase B

The yeast Candida antarctica was originally isolated in Antarctica and was found to produce two lipase variants (CALA and CALB) (Kirk and Christensen, 2002). Most lipases show interfacial activation; their activity is much higher when acting on substrates at a water-micelle interface compared to dissolved substrates. CALA shows interfacial activation, while CALB does not show such behavior and is therefore not considered to be a true lipase (Martinelle et al., 1995). The interfacial activation can be explained by the opening of a lid (flap) structure of the enzyme at an interface. The lipase with open lid is the active form of the enzyme and gives the substrate access to the active site. This lid structure covering the active site of true lipases is absent or very small in CALB. The structure of CALB was solved in 1994 and is shown in Figure 3-1 (Uppenberg et al., 1994; Uppenberg et al., 1995).

Figure 3-1. Structure of Candida antarctica lipase B that belongs to the α/β-hydrolase-fold superfamily. The α-helices are shown in red, the β-sheets in pale green, and the enzyme surface is in dark green. The ends of the acyl and alcohol chains of the substrate are visible in the narrow entrance of the active site. The catalytic amino acids and the other parts of the substrate are buried in the active site. A close-up of the active site is presented in Figure 3-2.

CALB belongs to the α/β-hydrolase-fold superfamily (Ollis et al., 1992), which contains enzymes that have evolved from a common ancestor (divergent evolution) to catalyze reactions as various as hydrolysis of esters, thioesters, peptides, epoxides, and alkyl halides or cleavage of carbon bonds in hydroxynitriles (Holmquist, 2000). CALB is built up of 317 amino acids and has a molecular weight of 33 kDa. The active site of CALB is illustrated in Figure 3-2. It contains the catalytic triad, Ser105-His224-Asp187, common to all serine hydrolases. Such a catalytic triad exists in enzymes with different folding, including trypsin and subtilisin, and is an example of convergent evolution (Holmquist, 2000). The active site of 13 Candida antarctica lipase B

CALB possesses an oxyanion hole that stabilizes the transition state and the oxyanion in the reaction intermediate. This oxyanion hole is a spatial arrangement of three hydrogen-bond donors, one from the side chain of Thr40 and two from the back-bone amides of Thr40 and Gln106. The active site also contains a small cavity called the stereospecificity pocket (Uppenberg et al., 1995), in which secondary alcohols have to orient one substituent during catalysis (Orrenius et al., 1998; Rotticci et al., 1998). This gives CALB a high enantioselectivity towards chiral secondary alcohols.

Figure 3-2. Close-up of the active site of CALB (the insert of Figure 3-1 indicates the zoom-in area). The tetrahedral intermediate of (R)-3-hexyl butanoate is covalently bound to the catalytic serine. The ends of the acyl and alcohol chains are visible in the narrow entrance of the active site. The hydrogen bonds in the oxyanion hole that stabilize the oxyanion are visible through the entrance. The rest of the active-site can be seen through the half-transparent amino acids closer to the surface. The catalytically essential hydrogen bonds from His224 are formed (center). The medium-sized substituent (ethyl) of the alcohol points into the stereospecificity pocket and it can be seen that it reaches the bottom of this pocket (the tryptophan in the top right corner).

The fast reacting enantiomer of secondary alcohols can simply and accurately be predicted by Kazlauskas rule (Figure 3-3) (Kazlauskas et al., 1991). Kazlauskas rule generally predicts the R-enantiomer as the fast reacting. This empirical rule has later been rationalized for CALB at a molecular level (Hæffner et al., 1998).

14 Candida antarctica lipase B

M L

Figure 3-3. The conformation of the large (L) and medium (M) substituent and the hydroxyl group of the fast reacting enantiomer of secondary alcohols as predicted by Kazlauskas rule (Kazlauskas et al., 1991).

The enantioselectivity of CALB towards secondary alcohols is determined by the steric requirements of the stereospecificity pocket. The fast-reacting enantiomer of the secondary alcohols orients its large substituent towards the active-site entrance and its medium-sized substituent in the stereospecificity pocket. To react, the slow-reacting enantiomer has to have the opposite orientation for its substituents compared to the fast-reacting enantiomer. The large substituent is not easily accommodated in the stereospecificity pocket, which explains the low reaction rate of this enantiomer (Rotticci et al., 1998). CALB catalyzes acyl-transfer reactions and follows the ping-pong bi-bi mechanism illustrated in Figure 3-4.

Ser105 O Gln106 O O 2 R Ser105 N 1 1 R O R H H H N N O O Asp187 O O H H H N N Asp187 O His224 O H O Free enzyme R2 N O His224

R3 First tetrahedral intermediate R1 O Thr40 R2 Gln106 O N Ser105 R3 H R1 H O O O Ser105 R1 O H O O H H H N N Asp187 O O H O H O R3 N N Asp187 O N His224 Second tetrahedral intermediate His224 Acyl enzyme Thr40

Figure 3-4. Reactions catalyzed by CALB follow a ping-pong bi-bi mechanism. The first substrate enters the active site and the first tetrahedral intermediate is formed (top right). The first product leaves the active site and the acyl-enzyme is formed (bottom right). The second substrate enters the active site and the second tetrahedral intermediate is formed (bottom left). The second product leaves the active site and enzyme is ready for another catalytic cycle (top left).

CALB can catalyze acyl-transfer reactions between various compounds. Depending on the R-groups in Figure 3-4, the reaction can be an ester 15 Candida antarctica lipase B hydrolysis, an esterification, or a transesterification. The second substrate can also be an amine, which would result in an aminolysis. In hydrolysis or transesterification, the first substrate (the acyl donor) can be a thioester. Some potential applications for CALB are in the resolution of chiral secondary alcohols (Ohtani et al., 1998; Rotticci et al., 2001), in the production of polylactones (Córdova et al., 1998; Uyama et al., 1997), and polyesters (Binns et al., 1998). CALB is a stable enzyme that has been used at 150 °C (Lozano et al., 2003), in organic solvents of high polarity such as acetonitrile and dimethyl sulfoxide (Paper IV), in ionic liquids (Park et al., 2003), in solid/gas systems (Lamare et al., 2001), and in supercritical carbon dioxide (Ottosson et al., 2002; Lozano et al., 2002). The literature about CALB together with the long and valuable experience acquired in our research group forms a large knowledge base about this enzyme. Therefore CALB was an obvious candidate for the rational redesign of enzymatic performance.

RESULTS AND DISCUSSION

To improve the properties of an enzyme through rational redesign, strong knowledge about catalysis and the catalyst is required. Our research group has a long experience and a high understanding of the molecular basis for the substrate specificity of lipase B from Candida antarctica (CALB). This information together with the structure of the enzyme formed the basis for the rational redesign of CALB presented in this thesis. We wanted to change the performance of CALB through carefully selected single-point mutations. The mutants were created and characterized to test if we; i) could change the catalytic properties of the enzyme, ii) had enough knowledge to successfully redesign the enzyme, and iii) could increase the knowledge about enzyme catalysis. Some general methods will be briefly described before discussing the results from the rational design, aiming to reengineer the reaction mechanism and the substrate selectivity.

4- General methods

We designed mutants aiming to change the performance of CALB. The gene coding for CALB (a generous gift by Novo Nordisk) had already been placed in a plasmid vector suitable for mutational work and consecutive protein expression in the yeast Pichia pastoris (Rotticci-Mulder et al., 2001). The desired mutations were introduced in the gene coding for CALB. The wild-type and mutated genes were transferred into the genome of Pichia pastoris. The lipase variants were expressed, purified, and often immobilized. Wild-type CALB and the mutant enzymes were characterized by acyl-transfer or hydrolysis reactions. The experimental data were analyzed and fitted to equations for enzyme kinetics to compare the mutants with CALB wild-type and also to compare the mutants with each other.

4.1 Mutagenesis

The desired point mutations in the CALB gene were introduced by two different methods. The overlap extension polymerase chain reaction technique is presented in Figure 4-1 (Ho et al., 1989).

1 (1a) (2a) (1a) 2 3 (1b) 1

Figure 4-1. Illustration of the steps to introduce specific point mutations with the overlap extension polymerase chain reaction technique (Ho et al., 1989). Two parallel reactions are performed, each containing the gene of interest, one primer carrying the mismatch ( ) and one complementary primer (__). 1. Several cycles (20-30) with denaturation of the double-stranded gene (part of a plasmid), annealing of primers, and DNA polymerisation reaction. 1a and 1b: Several copies of the first or second half of the gene (corresponding to the two different pairs of primers) overlapping at the region carrying the desired mutations. 2. Mixing of the gene fragments, denaturation, and annealing. 2a: Two sets of half genes annealed at the overlapping region that carries the mutation. 3. DNA polymerization reaction. 3a: The full gene with the desired mutation.

18 General methods

The method offered by the TransformerTM site-directed mutagenesis kit from Clontech is presented in Figure 4-2 (www.clontech.com).

Cloned Parental 2 Gene Plasmid 1 ● Target (1b) (2b) Parental + Plasmid + Parental 4 ● 1 Plasmid ● 3 ● Unique ● Mutated- (3b) (4b) Restriction ssParental 2 Parental Site Hybrid ● ● 3 Mutated Mutated Mutated (1a) (2a) 4 5 Plasmid Plasmid Plasmid

(3a) (4a) (5a)

Figure 4-2. The steps to introduce specific point mutations with the TransformerTM site-directed mutagenesis kit from Clontech (Figure modified from www.clontech.com). 1. Denaturation of double-stranded plasmids and addition of the primers carrying a mismatch: S to introduce the desired mutation in the gene and V to remove a unique restriction site (●). 1a: Single-stranded plasmids with annealed primers. 1b: Not denatured or re-annealed plasmids. 2. Synthesis of 2nd strand with T4 DNA polymerase, sealing of gaps with T4 DNA ligase, and addition of selection restriction enzyme. 2a: Not digested mutated-parental hybrid plasmids. 2b: Digested parental plasmids. 3. Transformation into mutS E. coli (unable to repair DNA mismatches). Both strands of the mutated-parental hybrid are copied in the cultivation of E. coli. Isolation of the plasmids from the transformant pool. 3a: Mutated plasmids originating from the mutated strand of the hybrid. 3b: Parental plasmids originating from the original strand of the hybrid. 4. Addition of selection enzyme. 4a: Not digested mutated plasmids. 4b: Digested parental plasmids. 5. Transformation into E. coli, cultivation of single colonies, plasmid isolation, and sequencing to confirm the presence of the desired mutation.

Both methods require that the gene of interest is inserted into a plasmid vector. We started with the gene coding for wild-type CALB inserted in pPIC9 or pPIC9k (Rotticci-Mulder et al., 2001). The oligonucleotides (primers) needed to introduce the mutations in the gene coding for CALB were obtained from Interactiva (Ulm, Germany). The primers have to be complementary to sequences inside and outside of the gene coding for CALB, except where the mutations are to be introduced. All the mutations introduced into the gene coding for CALB were confirmed by DNA sequencing.

4.2 Protein expression and purification

The plasmids containing the CALB gene with the desired mutation were linearized and transformed into electro-competent cells of the yeast Pichia 19 General methods pastoris (SMD1168, Invitrogen). These yeast cells have a defect gene making them unable to survive without supplementation of histidine in the growth medium. The cells that have integrated the linearized plasmid into the yeast genome can survive in the absence of histidine. Yeast clones surviving in the absence of histidine were screened for lipase excretion. The selected clones were used for production of the lipase variants. The expression was induced by growing the yeast Pichia pastoris on methanol as the sole carbon source. CALB wild-type and mutants were produced in shake-flask cultivations with an expression level of 10-20 mg×L-1 and in small-scale fermentations (8 L) with an expression level of 1-5 g×L-1 (Paper I and III)(Jahic et al., 2002). The lipase was secreted in the cultivation medium and the purification was started by separating medium from the yeast cells through centrifugation. Hydrophobic interaction chromatography on Butyl Sepharose 4 Fast Flow gel (Pharmacia Biotech) was used as a following purification step to yield high purity of the CALB variant. For some protein samples, we also performed gel filtration on a SuperdexTM 75 gel (Pharmacia Biotech) (Rotticci-Mulder et al., 2001). The buffer of the lipase solution was then changed to 20 mM MOPS-KOH (3-(N-morpholino)propanesulfonic acid-potassium hydroxide) of pH 7.2. Such aqueous solutions of different CALB variants were used for enzyme characterization towards ester hydrolysis (Paper I).

4.3 Protein immobilization and active-site titration

CALB variants dissolved in the MOPS buffer were immobilized on polypropylene beads (Holmquist et al., 1993). The enzyme immobilized on the carrier was washed with an ammonium hydrogen carbonate buffer (20 mM, pH 7), lyophilized to remove water and the volatile buffer, and equilibrated to a water activity of 0.1 against a saturated lithium chloride solution. The immobilized CALB variants were used for the enzyme characterization towards reactions catalyzed in organic solvents (Paper II-IV). In order to determine the specificity constants of the CALB variants, the active-site concentration has to be known. The synthesis of the irreversible inhibitor methyl 4- methylumbelliferyl hexylphosphonate was developed from the inhibitor described by Fujii et al. and synthetic route described by Rotticci et al. (Paper III; Fujii et al., 2003; Rotticci et al., 2000b). The inhibitor was purified and added to the lipase variants in an organic solvent. It reacts with the catalytic serine (Ser105) to form a stable acyl enzyme and releases 4-methylumbelliferone. The active-site concentration was determined from the fluorescence intensity (Ex. 360 nm, Em. 445 nm) of 4-methylumbelliferonate.

4.4 Activity measurements

The reaction rates of ester hydrolysis were determined spectrophotometrically at 405 nm. The chromophore para-nitrophenolate was added to the reaction mixture. When the enzyme hydrolyzes an ester, an 20 General methods alcohol and an acid are formed. The produced acid will release a proton and when the chromophore gets protonated, it will lose its absorbance at 405 nm. The decrease in absorbance versus time monitors the reaction rate in real-time. Acyl-transfer reactions were performed in organic solvents. Samples were taken at regular time intervals, separated on a gas chromatograph equipped with a chiral or achiral column, and the substrate and product concentrations were determined with a flame ionization detector. This does not allow for real- time monitoring of the reaction rate, but the reaction rate of several substrates can be determined simultaneously.

4.5 Enzyme kinetics

Acyl-transfer reactions catalyzed by CALB follow a ping-pong bi-bi mechanism. This gives rise to two-substrate kinetics, but if one substrate concentration is kept constant it, can be treated with pseudo-one-substrate kinetics (Martinelle and Hult, 1995). In the studied acyl-transfer reactions, either the acyl-donor or acyl-acceptor concentration was kept constant. The reaction rate, v, depends on the catalytic constant, kcat, the Michaelis constant, KM, the total enzyme concentration, [E], and the substrate concentration, [S], as described by the Michaelis-Menten equation (Equation 4-1).

k × []E × [S] v = cat Equation 4-1 K M + []S

The catalytic constant is a measure of the maximal activity of the enzyme and the Michaelis constant is an apparent equilibrium constant. The ratio kcat/KM is called the specificity constant and is an apparent second-order rate constant. It gives a measure of how good the substrate is for the enzyme. The apparent kinetic constants (kcat and KM) in Paper III and IV were determined by measuring the reaction rate in acyl-transfer reactions and fitting the data to the Michaelis-Menten equation (Equation 4-1). The concentration of the acyl donor was kept constant (and high) while the concentration of the acyl acceptor was varied. Specificity constants were calculated from the ratio of the catalytic constants. The specificity constants determined in Paper I were calculated from the initial hydrolysis rates in water at acyl-donor concentrations much lower than KM. In Paper III we determined relative specificity constants (substrate selectivities) by measuring the initial reaction rates of two or more substrates simultaneously. From the reaction rates (vx) and substrate concentrations (Sx) for the substrates to be compared, the relative specificity constant is calculated according to Equation 4-2.

⎛ k ⎞ cat v1 ⎜ ⎟ ⎝ K M ⎠ S S1 = 1 Equation 4-2 v ⎛ kcat ⎞ 2 ⎜ ⎟ K S2 ⎝ M ⎠ S 2 21 General methods

The enantioselectivity can be determined the same way as the substrate selectivity. It can also be calculated from the enantiomeric excess of the substrate and product, which can be accurately determined. This method (used in Paper IV) does not require low conversion, which the determination of initial reaction rates does, and only the relative concentrations of the substrates and of the products are required. If the enantioselectivity favors the S-enantiomer (S) over the R-enantiomer (R), the enantiomeric excess of the substrates (eeS) and of the product (eeP) is calculated according to Equation 4-3 and 4-4, respectively.

R − S ee = Equation 4-3 S S + R

S − R ee = Equation 4-4 P S + R

The enantioselectivity (E) is calculated from the enantiomeric excesses according to Equation 4-5 (Rakels et al., 1993).

⎛ ⎞ ⎜ ⎟ 1− ee ln⎜ S ⎟ ⎜ ee ⎟ ⎜1+ S ⎟ ⎜ ee ⎟ E = ⎝ P ⎠ Equation 4-5 ⎛ ⎞ ⎜ ⎟ ⎜ 1+ eeS ⎟ ln⎜ ⎟ eeS ⎜1+ ⎟ ⎝ eeP ⎠

The energy levels in an enzyme-catalyzed reaction can be presented as in Figure 4-3. It is assumed that [S]>>KM and that the rate-limiting step is the transition from the enzyme-substrate complex (ES) to the transition state (ES#). The energy ∆G(ES) depends on the Michaelis constant and ∆G(ES#) depends on the specificity constant.

∆G(ES#)

∆G(kcat)

∆G(ES)

KS kcat E + S ES ES# E + P Reaction coordinate

Figure 4-3. The energy profile diagram for a simplified enzyme-catalyzed reaction, where [S]>>KM and the rate-limiting step is the transition from the enzyme-substrate complex (ES) to the transition state (ES#). E is the enzyme, S the substrate, ES the enzyme-substrate complex, ES# the transition state, and P the product.

22 General methods

The activation energy (∆G#) for the enzyme-catalyzed reaction is related to kcat and KM as described by Equation 4-6, where R is the gas constant and T the temperature in Kelvin.

# ⎛ kcat ⎞ ∆G = −RT × ln⎜ ⎟ Equation 4-6 ⎝ K M ⎠

To compare the specificity constants of the enzyme variants, we have calculated various ratios of specificity constants and the corresponding difference in the activation energy (Paper I-IV). The substrate selectivity is equal to the ratio of the specificity constants towards the two substrates. The # difference in activation energy (∆S1-S2∆G ) between the reactions involving the two substrates (S1 and S2) is calculated using Equation 4-7.

⎛ ⎛ k ⎞ ⎞ ⎜ ⎜ cat ⎟ ⎟ ⎜ ⎜ K ⎟ ⎟ # ⎝ M ⎠ S1 ∆ S1−S 2 ∆G = −RT ×ln⎜ ⎟ Equation 4-7 ⎛ k ⎞ ⎜ ⎜ cat ⎟ ⎟ ⎜ ⎜ K ⎟ ⎟ ⎝ ⎝ M ⎠ S 2 ⎠

If the two substrates are enantiomers, the ratio of the specificity constants is called enantioselectivity, E. The difference in activation energy between a mutant (mut) and the wild-type (wt) enzyme towards the same substrate (∆mut- # wt∆G ) is calculated in the same way as the substrate selectivity (Equation 4-7). If the substrate selectivity (or enantioselectivity) of a mutant and the wild-type # enzyme are compared, the differential activation energy (∆mut-wt∆S1-S2∆G ) is calculated according to Equation 4-8.

⎛ ⎛ k ⎞ ⎞ ⎜ cat ⎟ ⎜ ⎟ ⎜ ⎝ K M ⎠ ⎟ ⎜ S1 ⎟ ⎛ k ⎞ ⎜ ⎜ cat ⎟ ⎟ ⎜ ⎜ ⎟ ⎟ K M # ⎝ ⎝ ⎠S 2 ⎠mut ∆ mut−wt ∆ S1−S 2∆G = −RT × ln Equation 4-8 ⎛ ⎛ k ⎞ ⎞ ⎜ cat ⎟ ⎜ ⎟ ⎜ ⎝ K M ⎠ ⎟ ⎜ S1 ⎟ ⎛ k ⎞ ⎜ ⎜ cat ⎟ ⎟ ⎜ ⎜ ⎟ ⎟ K M ⎝ ⎝ ⎠S 2 ⎠wt

5- Reengineering the reaction mechanism

Candida antarctica lipase B naturally catalyzes various acyl-transfer reactions. The catalytic triad and the oxyanion hole are located in the active site and are very important for the catalytic efficiency. Their interactions with the substrate during catalysis are highlighted in the reaction mechanism for acyl-transfer reactions catalyzed by CALB (Figure 5-1).

Ser105 R2 O O O H Asp187 H O O N N R1 His224 Gln106

N Ser105 2 R H O O O H H H Asp187 O O N N O R1 H His224 N

Thr40 Ser105 R2 O O

Asp187 H H O O N N O

1 His224 R

Figure 5-1. The reactions catalyzed by CALB follow a ping-pong bi-bi mechanism. The first substrate enters the active site (top). In the catalytic triad, Ser105 donates a proton to the basic Asp187-His224 pair, and performs a nucleophilic attack on the substrate to form the first reaction intermediate. In the reaction intermediate, the substrate is covalently bound to the catalytic Ser and the oxyanion is stabilized by hydrogen bonds from Gln106 and Thr40 (middle). The reaction intermediate collapses to the acyl-enzyme and the first product leaves the active site (bottom). A second substrate enters the active site (bottom), the enzyme goes through the second reaction intermediate (middle), the second product is formed, leaves the active site and the enzyme is ready for another cycle (top). 24 Reengineering the reaction mechanism

The catalytic triad is made of Asp187, His224, and Ser105. The aspartate- histidine pair forms a base that transfers protons between the catalytic serine and the substrate, and the serine performs the nucleophilic attack on the substrate. The oxyanion hole is a spatial arrangement of three hydrogen-bond donors that stabilize the transition state. Two of the hydrogen-bond donors are the backbone amides of Thr40 and Gln106 and the third is the hydroxyl group in the side-chain of Thr40. Different approaches have been used to modify the reaction mechanism and change the substrate specificity of enzymes (Berglund and Park, 2005). They can be divided into; random changes by directed evolution, replacement of key amino acids according to a model enzyme, introduction of a catalytic machinery in non-enzymatic proteins, and modification of the active site without guidance from other enzymes. Our approach to reengineer the reaction mechanism of CALB falls into the last group. The two hypotheses to alter the reaction mechanism tested and presented in this thesis were that:

- The hydrolytic activity of the enzyme could be deleted by replacing a catalytic amino acid. The remaining parts of the active-site were believed to show aldolase activity following a new reaction mechanism (Paper II).

- The transition-state stabilization could be decreased by removing a functional group in the enzyme by a point mutation. Substrates carrying this functional group were expected to introduce a new form of substrate-assisted catalysis (Paper I).

5.1 New reaction mechanism

To introduce a new reaction mechanism in an enzyme scaffold, the catalytic machinery has to be modified or completely replaced. Examples of changed catalytic machinery leading to altered substrate specificity and the approaches used to achieve it are described in a recent review article (Berglund and Park, 2005). In our work, the reaction mechanism in CALB was changed by mutating one of the catalytic amino acids (Paper II). The nucleophilic Ser was replaced by an Ala or Gly (Ser105Ala and Ser105Gly) to remove the hydrolytic activity of the enzyme. The active site of the mutants still contains the oxyanion hole that can stabilize a negative charge and the Asp-His pair of the former catalytic triad that can function as a base. These properties should make the CALB mutants able to perform aldol additions following the reaction mechanism presented in Figure 5-2.

25 Reengineering the reaction mechanism

Gln106

Ser105Ala N CH H O 3 H Asp187 H O N N O H O His224 N R1 Gln106 Thr40 O Ser105Ala N CH H O 3 R1 O H OH Asp187 H H O N N O R1 H 2 R His224 N Gln106 Thr40 Ser105Ala N CH H O 3 H O H Asp187 H O N N O O R1 H His224 N Gln106 R2 Thr40 Ser105Ala N CH H R2 O 3 O H O H Asp187 H O N N O R1 H His224 N

Thr40

Figure 5-2. Proposed reaction mechanism for the aldol addition catalyzed by the Ser105Ala mutant of Candida antarctica lipase B. The first substrate enters the active site and the oxyanion hole increases the acidity of the α-proton (top right). The α-proton is removed by His224 and the enolate formed is stabilized by the oxyanion hole (bottom right). The second substrate enters the active site (bottom left). One of the electron pairs in the double bond of the enolate attacks the carbonyl carbon of the second substrate. The proton on His224 is taken by the carbonyl oxygen of the second substrate. The product leaves the active site and the enzyme is free for another catalytic cycle (top left).

The mutants had no detectable hydrolytic activity, while aldolase activity with both propanal and hexanal could be quantified. The addition products of the two substrates were produced with similar rates by both mutants. The specific activity for the Ser105Ala mutant towards hexanal was 0.0013 µmol×min-1×mg-1 or 65 day-1, assuming that all immobilized CALB mutant is active. This is similar to the aldolase activity that has been evolved for antibodies (Hoffmann et al., 1998). The mutants catalyzed the reaction 4 times more efficiently than the wild-type, 300 times faster than albumin or empty carrier. The mutants had a diastereoselectivity different from the uncatalyzed

26 Reengineering the reaction mechanism reaction and the reaction rate of inhibited wild-type enzyme was equal to that of empty carrier, showing that the active site is needed for catalysis. Natural aldolases are divided into two classes depending on the reaction mechanisms. Class I aldolases have a reaction mechanism involving the formation of a Schiff-base intermediate. Class II aldolases follow a reaction mechanism similar to that proposed for the CALB mutants, they contain a divalent cation (usually Zn2+ or Fe2+) that has a function similar to that of the oxyanion hole. In both classes, a basic residue (i.e. Tyr, Asp, Glu, or Lys) has a role similar to that of His224 in the Ser105 mutants of CALB. Recently, an aldolase from Arabidopsis thaliana was proposed to have a reaction mechanism not belonging to any of these well-known classes, see Figure 5-3 (Bauer et al., 2004). It neither involves a metal cofactor nor a Schiff-base formation, but the substrate has an internal structure similar to a Schiff-base that facilitates the catalysis. The substrate performs substrate-assisted catalysis that involves an electron-withdrawing group.

Asn70

O Lys99 Glu21 H N H H O H O O Glu73 O H OH O Asn70 O OH H N O N OH O Lys99 H O Glu21 H H N N N N H H H O H O H H Glu73 O H OH O O Asn70 H N O N

O Lys99 H N N N Glu21 H H N H H O H H O

Glu73 O H OH O O H N O N H N N N H H Figure 5-3. Proposed reaction mechanism for a novel aldolase class (Bauer et al., 2004). The acid- base catalysis is performed by Lys99. The substrate has an imine functionality that behaves like the Schiff base in Class I aldolases and forms a type of substrate-assisted catalysis.

Aldolases form carbon-carbon bonds, and such enzymes are very interesting for synthetic chemistry (Breuer and Hauer, 2003). Enzymes that form carbon- carbon bonds originate from the following enzyme classes; transferases (E.C.2), lyases (E.C.4), and ligases (E.C.6). In living systems they usually act on activated substrates with high specificity. Such substrates are expensive and limit the use of aldolases in synthetic processes. Still, aldolases are the enzymes most used for carbon-carbon coupling. They have broad substrate specificity for the electrophile, but their use is limited by the very narrow substrate specificity 27 Reengineering the reaction mechanism for the nucleophile. Our Ser105-mutants of CALB has increased the substrate range available for aldol-addition reactions and demonstrated the possibility offered by rational redesign to increase utility of biocatalysis. Experiments and theoretical calculations were performed to further examine the aldolase activity of the CALB Ser105-mutants. Aldolase activity was quantified towards several other substrates (both aldehydes and ketones), and the relatively low activity was explained by a too long distance between the α- proton on the substrate and the proton-accepting nitrogen on the catalytic histidine (Branneby et al., 2004). The reengineered active site of the Ser105Ala mutant also catalyzed Michael-type additions of various thiols and α,β- unstaurated carbonyl compounds (Carlqvist et al., 2005). The turn-over numbers ranged from 10-3 to 4 min-1 and the mutant had a catalytic proficiency up to 107. In the proposed reaction mechanism, the oxyanion hole increases the electrophilicity of the β-carbon in the carbonyl compound. His224 takes the proton from the thiol substrate, the thiolate performs a nucleophilic attack on the β-carbon, and the enolate formed is stabilized in the oxyanion hole. The His224 proton is added to the α-carbon and the product is released from the active site. Other types of reactions that could be catalyzed by the CALB Ser105-mutants are racemization, Baeyer-Villiger oxidation, and epoxidation. These reactions also involve proton transfer by an acid/base and activation of a carbonyl group and/or stabilization of an oxyanion. A rational work to redesign the reaction mechanism of alanine racemase was published recently (Yow et al., 2003). Yow et al. made two point mutations, which resulted in an undetectable racemization activity and introduced a half transamination activity (from D-alanine to pyridoxal 5´-phosphate) of 3.3 h-1. This activity is of the same magnitude as the aldolase activity of our Ser105Ala mutant of CALB. Another interesting example is the cysteine protease papain that was rationally redesigned by one point mutation to introduce nitrile hydratase activity (Dufour et al., 1995). The wild-type enzyme hydrolyzes peptide bonds and shows a low nitrile hydratase activity. Via one carefully selected point mutation, Gln19Glu, the nitrile hydratase activity was increased by a factor of 1000-3000. This corresponds to a difference in activation energy of 17-20 kJ×mol-1. The reaction mechanism proposed for the nitrile hydratase activity of the Gln19Glu of papain is presented in Figure 5-4.

28 Reengineering the reaction mechanism

Gln19Glu Gln19Glu O O

O O H H H R N N R N

O R S H S N NH N NH H2N R Cys25 Cys25 His159 His159

H2O

Gln19Glu Gln19Glu Gln19Glu O O O

O O O H H H H H N H N R N

R S R S O S H N NH H N NH O N NH O H Cys25 Cys25 Cys25 H H His159 His159 His159

Figure 5-4. Proposed reaction mechanism for the nitrile hydratase activity of the Gln19Glu mutant of papain (Dufour et al., 1995). The carbon atom of the nitrile in the substrate is attacked by Cys25. His159 and the redesigned Gln19Glu perform acid-base catalysis. A water molecule is added to the nitrile bond and the final product is an amide.

5.2 Substrate-assisted catalysis

In substrate-assisted catalysis (SAC), a functional group in the substrate contributes to the catalysis. Several examples of SAC have been identified in natural enzymes and created by reengineering enzymes. These enzymes belong to different enzyme classes including serine proteases, type II restriction nucleases, and GTPases (Dall'acqua and Carter, 2000). The enzymes naturally performing SAC, have very high substrate specificity towards substrates bearing the functional group that participates in the catalysis. The introduction of SAC could serve as a powerful route to create such high substrate selectivity for reactions catalyzed by other enzymes. Our strategy to introduce SAC in CALB was to remove the hydroxyl group in the enzyme that stabilizes the transition state and to replace it by a hydroxyl group in the substrate (Paper I). The oxyanion hole of CALB has three hydrogen-bond donors that stabilize the transition state and the reaction intermediate; the two backbone amides from Thr40 and Gln106 and the hydroxyl group on the side chain of Thr40 (Figure 5-5, top). Two mutants were created, Thr40Ala and Thr40Val, to remove the possibility of the hydrogen bond from the side-chain of amino acid 40 to be formed. Substrates bearing a hydroxyl group could replace the missing hydrogen-bond donor in the mutants and perform SAC (Figure 5-5, bottom).

29 Reengineering the reaction mechanism

Gln106

Wild-type N Ser105 H O O O H H H Asp187 O O N N O H

His224 N

Thr40 Gln106

Thr40Val O N Ser105 H H O O O H H Asp187 N N O O H

His224 N

Thr40Val

Figure 5-5. The tetrahedral reaction intermediate in CALB wild-type (top) and Thr40Val (bottom) showing the catalytic triad (Ser105-His224-Asp187), the substrate in bold, and the oxyanion hole (Gln106 and Thr40). The substrate is covalently bound to the catalytic serine. In the wild-type (top), the oxyanion in the substrate is stabilized by three hydrogen bonds, two from the backbone amides of Thr40 and Gln106, and one from the side-chain of Thr40. In the Thr40Val mutant (bottom), no hydrogen bond can be formed from the side-chain of Thr40Val. It is replaced by the hydroxyl group of the substrate, resulting in substrate-assisted catalysis.

The hydrolytic activity of the created mutants (Thr40Ala and Thr40Val) towards straight–chain esters (ethyl propanoate and ethyl butanoate) was much lower than that of the wild-type lipase (Table 5-1), corresponding to 18-21 kJ×mol-1 higher activation energy. The role of the oxyanion hole has also been studied in subtilisin (Wells et al., 1986), papain (Ménard et al., 1991), cutinase (Nicolas et al., 1996), and Esherichia coli type I signal peptidase (Carlos et al., 2000). Point mutations were performed in the oxyanion hole of these enzymes to remove one of the hydrogen-bond donors stabilizing the transition state. This increased the activation energy by 9-21 kJ×mol-1. It was shown to mainly depend on a lowered transition-state stabilization ( kcat ) and rather than an increased Michaelis constant ( K M ).

30 Reengineering the reaction mechanism

Table 5-1. Specificity constants ( kcat / K M ) for the enzyme catalyzed and second-order [a] rate constants ( k2 ) for the base catalyzed hydrolysis of achiral and chiral ethyl esters. Substrate Catalyst Ethyl ester -R wild-type Thr40Ala Thr40Val OH- -1 -1 -1 -1 kcat / K M (s M ) k2 (s M ) O -H 830 0.33 0.60 0.22 R O -(S)-OH 190 2.0 5.0 2.0 -(R)-OH 120 0.20 0.23 2.0

O R -H 6700 1.7 2.0 0.15 -(S)-OH 54 0.14 0.059 0.36 O -(R)-OH 1700 1.1 0.12 0.36 [a] Specificity constants were obtained from initial reaction rates determined with 1-20 mM substrate, 0.125 mM p-nitrophenol, and 2.5 mM MOPS buffer at pH 7.2 and 25°C. Enzyme activities were measured spectrophotometrically at 400 nm in triplicates or more. Second-order rate constants for the base catalyzed hydrolysis were determined at pH 10. For original data and standard deviations see Paper I.

The side-chain of Thr40 in CALB was shown to be very important for the transition-state stabilization. The next step was to replace the missing hydrogen-bond donor in the Thr40 mutants with a hydroxyl group in the substrate. Both mutants showed higher activity towards ethyl (S)-2-hydroxypropanoate compared to ethyl propanoate, while the activity of the wild-type decreased. This showed that SAC had been introduced in the reengineered active sites. The steric requirements of the active site prevented the R- enantiomer from participating in SAC, leading to higher enantioselectivity compared to that of the almost non-selective wild-type lipase. The differential activation energies and the corresponding E-values are presented for CALB wild-type, Thr40Ala, and Thr40Val towards ethyl 2-hydroxypropanoate and ethyl 3-hydroxybutanoate in Figure 5-6.

EthylEthyl 3-hydroxybutanoate

EthylEtyl 2-hydroxypropanoate 2-hydroxypropanoate 10 E =22 E =9.8 S S ] (kJ/mol) R )

M 5

/K E =1.6

cat S 0 / (k S ) M

/K E =2.0 -5 R cat E =7.9 R -10 E =31 R RT ln [ (k Wild type Thr40Ala Thr40Val

Figure 5-6. Comparison of the differential activation energy and enantioselectivity of wild type CALB and oxyanion hole mutants for the hydrolysis of ethyl esters (Paper I). 31 Reengineering the reaction mechanism

Both mutants favored the S-enantiomer compared to the wild-type enzyme. The enantioselectivity increased from 1.6 for the wild-type to 22 for the Thr40Val mutant towards ethyl 2-hydroxypropanoate. For both substrates, the S-enantiomer is stabilized by 6-7 kJ×mol-1 more in the Thr40Val mutant than in the wild-type enzyme. An interesting observation is that the difference in differential activation energy between the two enantiomers and the two # # # substrates ( ∆∆∆G or (∆ S−R∆G )2−hydroxypropanoate − (∆ S−R∆G )3−hydroxybu tan oate ) is very similar for the three CALB variants (wild-type, Thr40Ala, and Thr40Val), corresponding to 9.4 to 11 kJ×mol-1. We created SAC by rational redesign of the catalytically important oxyanion hole in CALB. The oxyanion hole is a feature common to many enzymes and several of them have a side-chain contributing to the transition-state stabilization. These enzymes could also be redesigned to create the same kind of SAC as in the Thr40 mutants of CALB. Substrates bearing a free hydroxyl group on the alcohol part of the substrate (diols and triols) and substrates with other hydrogen-bond donors could also be used with such mutants to introduce SAC. An example of natural SAC that is present in and crucial for all leaving organisms is the hydrolysis of ribonucleic acid (RNA) catalyzed by RNase (Figure5-7) (Voet et al., 1999). It is the hydroxyl group in the 2´-position on the ribose unit that takes part in the catalysis. This gives the RNase its very high specificity for RNA over DNA.

R1 R1 R1

O Base His12 O Base His12 O Base His12

NNH NNH NNH H O O H O O O OH H O P O P O 2 O P O R2 O O O O O H H H H H N N N

HO NH R2 NH NH His119 His119 His119

Figure 5-7. Reaction mechanism for hydrolysis of RNA catalyzed by RNase. The catalytic amino acids are numbered as in bovine pancreatic RNase A. Two histidine residues perform acid-base catalysis and the substrate performs substrate-assisted catalysis. Figure modified from Voet et al. (Voet et al., 1999).

The first engineered SAC was done on subtilisin BPN´ in 1987 (Carter and Wells, 1987). Subtilisin has the same reaction mechanism as CALB (and other lipases) and their active sites are close to mirror images. This enzyme is an endoprotease that cleaves off the last amino acid in a peptide chain with broad substrate specificity. The histidine in the catalytic triad was replaced with an alanine. This decreased the activity up to a million fold, but it was partially restored for substrates containing a histidine. The histidine in the substrate replaces the function of the original histidine in the enzyme as a proton donor

32 Reengineering the reaction mechanism and acceptor. The histidine mutant showed a selectivity of 170 for an Ala over a His in the substrate and this selectivity was changed 670 times compared to the enzyme with the intact catalytic triad. These changes correspond to a difference in activation energy of 13 and 16 kJ×mol-1. This is almost two times larger than the effects of the Thr40 mutations of CALB. The mutations of Thr40 in CALB changed the substrate selectivity corresponding to a difference in activation energy of 8-9 kJ×mol-1.

6- Reengineering the substrate selectivity

Lipase B from Candida antarctica (CALB) shows high substrate selectivity for acyl-transfer reactions to or from secondary alcohols. This high selectivity is explained by the geometrical requirements of the active site. During catalysis, secondary alcohols have to position one of their substituents in a small cavity in the active site, the stereospecificity pocket (Orrenius et al., 1998; Rotticci et al., 1998). The enantioselectivity follows Kazlauskas rule (Figure 3-3) (Kazlauskas et al., 1991), which generally predicts the R-substrate as the fast reacting enantiomer. The fast-reacting enantiomer of secondary alcohols positions its medium-sized substituent in the stereospecificity pocket and the large substituent towards the active site entrance (Figure 6-1), while the slow-reacting enantiomer positions its large substituent in the stereospecificity pocket and the medium-sized substituent towards the active-site entrance. The large substituent is not comfortably accommodated in the limited space of the stereospecificity pocket, but this is the only conformation where the slow- reacting enantiomer can form all the hydrogen bonds essential for catalysis.

Ser105 Gln106 O H

H His224 H R O Thr40 O H

4 0 1 p r T

Figure 6-1. The reaction intermediate of CALB with (R)-1-phenylethyl butanoate covalently bound to the catalytic Ser. The molecular model of the structure (left) and a schematic representation viewed from the right side of the molecular model (right). The secondary alcohol positions its medium-sized substituent (methyl) in the stereospecificity pocket and orients its large substituent (phenyl) towards the active-site entrance. The oxyanion is stabilized by hydrogen bonds from Thr40 and Gln106 in the oxyanion hole. The catalytically essential hydrogen bonds from the catalytic His are formed. 34 Reengineering the substrate selectivity

The stereospecificity pocket is only large enough to comfortably accommodate an ethyl group or smaller. This gives CALB a very high enantioselectivity towards secondary alcohols with a medium-sized substituent smaller or equal to an ethyl and a large substituent larger than an ethyl (Rotticci et al., 2001). In accordance with this size limitation, secondary alcohols with two substituents larger than an ethyl are poor substrates for CALB. We increased the size of the stereospecificity pocket of CALB by rational redesign (Paper III and IV). The size of this pocket is confined by back-bone atoms and by the side chains of Thr42, Ser47, and Trp104 (Figure 6-1). The size of the stereospecificity pocket was increased by replacing the large Trp104, located at the bottom of this pocket, with smaller amino acids. The large stereospecificity pocket of the mutants should accommodate larger substituents than that of the wild-type lipase. This was expected to have two consequences:

- Secondary alcohols with both substituents larger than an ethyl group could be good substrates for the CALB mutants, broadening the substrate specificity and thereby the utility of the enzyme (Paper III).

- The enantioselectivity of the mutated CALB should be very different from that of the wild-type (Paper IV).

6.1 Broadening the substrate specificity

The number of secondary alcohols that are good substrates for CALB is limited by the size of the stereospecificity pocket. If the secondary alcohol has two substituents larger than an ethyl, it is not readily accepted by CALB. We wanted to overcome this limitation and broaden the substrate specificity of the lipase (Paper III). Having identified the size of the stereospecificity pocket as the crucial parameter, we performed point mutations of amino acids with side chains lining this pocket. To increase the size of the stereospecificity pocket, Trp104 was replaced by three amino acids of relative decreasing size; a histidine (Trp104His), a glutamine (Trp104Gln), and an alanine (Trp104Ala). To confirm that the size of the stereospecificity pocket is the important factor rather than electrostatic effects, we created three other mutants with nearly isosteric but non-polar amino acids replacing polar ones (Thr42Val, Ser47Ala, and Thr42Val/Ser47Ala). To map the size of the stereospecificity pocket of the mutants, acylation reactions of secondary alcohols were performed. During catalysis, secondary alcohols have one substituent in the stereospecificity pocket and the other towards the active-site entrance. By choosing secondary alcohols with two identical substituents as substrates, the size of the substituent positioned in the stereospecificity pocket is known. The reaction rates of two or more alcohols were determined simultaneously and the relative specificity constants were calculated using Equation 4-2. The specificity constants relative to that of 4- heptanol are presented in Table 6-1

35 Reengineering the substrate selectivity

Table 6-1. Specificity constants relative to that of 4-heptanol for the acylation of symmetrical secondary alcohols determined for the mutants with nearly isosteric mutations in the stereospecificity pocket, the wild-type lipase, and the Trp104 mutants (Paper III). CALB Acyl acceptor variant 2-propanol 3-pentanol 4-heptanol 5-nonanol cyclohexanol

Thr42Val 12000 100 1.0 n.d. n.d. Ser47Ala 12000 110 1.0 n.d. n.d. Thr42Val/Ser47Ala 10000 110 1.0 n.d. n.d.

Wild-type 12000 100 1.0 0.036 200

Trp104His 2.9 0.36 1.0 0.12 0.23 Trp104Gln 0.20 0.10 1.0 0.26 0.088 Trp104Ala 0.54 0.067 1.0 0.74 0.079 n.d. The relative specificity constant was not determined.

The substrate specificity of CALB wild-type decreased sharply with increasing size of the alcohol (Table 6-1, middle). The substrate selectivity for 2- propanol over 5-nonanol was very high, over 330 000. The three mutants with close to isosteric mutations in the stereospecificity pocket had nearly identical substrate selectivity compared to that of the wild-type lipase (Table 6-1, top). This showed that the polarity of these side chains had no significant effect on the substrate selectivity towards the tested substrates. On the other hand, the mutants with redesigned size of the stereospecificity pocket (Trp104 mutants) behaved completely different compared to the wild-type lipase (Table 6-1, bottom). They showed equal substrate selectivity between all the tested substrates. The relative specificity constants allow us to compare the substrate selectivity between the variants. They do not reveal the magnitude of the individual specificity constants and the efficiency of each enzyme variant for the acyl- transfer reactions can not be compared. We therefore determined the apparent kinetic constants towards 4-heptanol for CALB wild-type and the Trp104 mutants at a fixed concentration of the acyl donor (Table 6-2).

Table 6-2. Kinetic constants of the CALB wild-type and Trp104 mutants for the acylation of 4-heptanol at 500 mM vinyl butanoate (Paper III). app app [a] CALB kcat K M kcat / K M variant (s-1) (mM) (s-1M-1) Wild-type 0.53 130 4.2 Trp104His 12 21 550 Trp104Gln 10 18 560 Trp104Ala 28 24 1100 app app [a] Calculated from kcat and K M .

The mutations of Trp104 to increase the size of the stereospecificity pocket app improved the apparent Michaelis constant ( K M ). These mutants have an

36 Reengineering the substrate selectivity apparent Michaelis constant 5-7 times smaller than that of the wild-type lipase. A larger effect of the mutations was seen on the apparent catalytic constant app ( kcat ), with an increase of 20-50 times compared to that of the wild-type. The specificity constant ( kcat / K M ) towards 4-heptanol was calculated from the ratio of the apparent catalytic constant and the apparent Michaelis constant, and was then multiplied by the relative specificity constants. This yielded the specificity constant towards each substrate for CALB wild-type and the Trp104 mutants (Table 6-3).

Table 6-3. Specificity constants of the CALB wild-type and Trp104 mutants for the acylation of achiral secondary alcohols (Paper III). CALB Acyl acceptor variant 2-propanol 3-pentanol 4-heptanol 5-nonanol cyclohexanol -1 -1 kcat / K M (s M ) Wild-type 50000 420 4.2 0.15 840 Trp104His 1600 200 550 67 130 Trp104Gln 110 59 560 150 49 Trp104Ala 610 76 1100 830 89

The mutations performed to increase the size of the stereospecificity pocket had a large effect on the specificity constants. The mutants had much higher specificity constants towards the large secondary alcohols compared to that of CALB wild-type. A trend among the mutants can be seen towards 4-heptanol and even more towards 5-nonanol; the specificity constant increased with decreasing size of amino acid 104. This resulted in the very good specificity constants of 1100 and 830 s-1M-1 for the Trp104Ala mutant towards 4-heptanol and 5-nonanol, respectively. This is 260 and 5500 times higher than the specificity constant of the wild-type lipase. To further compare the mutants with the wild-type lipase, we calculated the difference in activation free energy # ( ∆ mut−wt ∆G ) between each mutant (mut) and the wild-type (wt) for the 5 tested alcohols (Figure 6-2).

20 15 15 11 8.7 10 OH OH 7.1 5.7 5.0 4.3 4.8 5 1.9 # -1 ∆∆G /kJ*mol 4-heptanol 5-nonanol 0 2-propanol 3-pentanol cyclohexanol -5 OH OH OH -10 -12 -12 -15 -14 Se∆rie1 ∆G# -15 (Trp104His)-(wild-type) -17 -20 Serie2 # ∆(Trp104Gln)-(wild-type)∆G Serie3 # -22 -25 ∆(Trp104Ala)-(wild-type)∆G

Figure 6-2. Difference in activation free energy between each Trp104 mutant and the wild-type lipase, for the acylation of achiral secondary alcohols (Paper III). 37 Reengineering the substrate selectivity

The activation energy was 12-14 kJ×mol-1 lower towards 4-heptanol and 15- 22 kJ×mol-1 lower towards 5-nonanol for the mutants compared to CALB wild- type (Figure 6-2). The difference in substrate selectivity, for 5-nonanol over 2- propanol, was over 400 000 times for both the Trp104Ala and Trp104Gln mutation compared to the wild-type enzyme. This corresponds to a change in activation energy of 33 kJ×mol-1. From the kinetic constants determined towards 4-heptanol, it was seen that only a smaller part of the decreased activation energy can be attributed to the improved Michaelis constant. The largest effect was seen from the increased catalytic constant, which can be explained by a higher ratio of productive binding in the redesigned stereospecificity pockets than in that of the wild-type lipase. Our rational redesign of the size of the stereospecificity pocket in CALB can be compared to the redesign of a binding pocket in subtilisin (Rheinnecker et al., 1994). Subtilisin is an endoprotease that hydrolyses peptides with broad substrate specificity. Its active site contains the catalytic triad Ser-His-Asp and an oxyanion hole, like the active site of CALB. Subtilisin has two specificity pockets (S1 and S4), each bind one amino acid of the peptide that is hydrolyzed. The substrate specificity was redesigned by increasing the size of the S4 binding pocket. The substrate selectivity changed 200 times compared to that of the wild-type, favoring substrates with a leucine instead of an alanine.

6.2 Redesigning the enantioselectivity

Due to the steric requirements of the stereospecificity pocket in Candida antarctica lipase B, this enzyme has a high enantioselectivity towards secondary alcohols (Hæffner et al., 1998; Rotticci et al., 1998). During catalysis, a secondary alcohol has to position one of its substituents in the stereospecificity pocket, to allow for all the catalytically essential hydrogen bonds to be formed. The stereospecificity pocket in the wild-type lipase is only large enough to comfortably accommodate an ethyl group or smaller. We redesigned the size of the stereospecificity pocket through point mutations of Trp104, situated at the bottom of this pocket. The stereospecificity pocket of the Trp104Ala mutant was found to comfortably accommodate much larger substituents than that of the wild-type lipase (Paper III). As a consequence, the Trp104Ala mutant was expected to show an enantioselectivity different from that of CALB wild-type. The enantioselectivity should be low towards chiral secondary alcohols that can accommodate their large substituent in the redesigned stereospecificity pocket. Molecular modeling was performed with CALB wild-type and the Trp104Ala mutant to compare their stereospecificity pockets. The tetrahedral reaction intermediate in the acylation of the R- and S-enantiomer of 1- phenylethanol was covalently bound to the catalytic serine (Figure 6-3). Energy minimization was performed on the reaction intermediates to find the optimal orientation of the substrate in the active site and the stereospecificity pocket.

38 Reengineering the substrate selectivity

Figure 6-3. The active site of CALB wild-type (left) and Trp104Ala (right) with the tetrahedral reaction intermediate in the acylation of (R)-1-phenylethanol (top) and (S)-1-phenylethanol (bottom) covalently bound to the catalytic serine. The substrate is presented in stick model and amino acid 104 in white space fill. The R-enantiomer has similar conformation in the wild-type and Trp104Ala mutant, the large substituent (phenyl) points towards the active-site entrance and the medium-sized substituent (methyl) is positioned in the stereospecificity pocket. In the wild-type, the S-enantiomer can not position its phenyl group in the stereospecificity pocket, and not all the essential hydrogen bonds to the catalytic His can be formed. In the Trp104Ala mutant is the phenyl group comfortably accommodated in the space liberated by the mutation in the stereospecificity pocket (Paper IV).

In the modeling studies, the R-enantiomer had similar orientation in both CALB wild-type and Trp104Ala. The medium-sized substituent (methyl) points into the stereospecificity pocket, the large substituent (phenyl) is oriented towards the active-site entrance, and the hydrogen bonds essential for catalysis are formed (Figure 6-3, top). The S-enantiomer has to orient its phenyl group in the stereospecificity pocket to allow for the formation of the catalytically essential hydrogen bonds. This is not possible with the wild-type enzyme (Figure 6-3, bottom left). On the other hand, the redesigned stereospecificity 39 Reengineering the substrate selectivity pocket of the Trp104Ala mutant comfortably accommodates the phenyl group of the S-enantiomer (Figure 6-3, bottom right). The absence of a catalytically essential hydrogen bond in CALB wild-type with the S-enantiomer predicts a very high R-selectivity. Observing the reaction intermediates in the Trp104Ala mutant does not reveal any obvious reason for a high enantioselectivity. We decided to experimentally determine the enantioselectivity of CALB wild-type and Trp104Ala for the acylation of 1-phenylethanol. The enantioselectivity of CALB wild-type was too high to be determined by measuring the enantiomeric excess during kinetic resolution. The product of the S-enantiomer could not be quantified before all of the R-enantiomer had app app reacted. Instead, the apparent kinetic constants ( kcat and K M ) were determined towards the pure enantiomers of 1-phenylethanol for both CALB wild-type and Trp104Ala (Table 6-4).

Table 6-4. Apparent kinetic constants, specificity constants, preferred enantiomer (Pe), and E-value of the CALB wild-type and Trp104Ala mutant for the acylation of 1- phenylethanol in cyclohexane at 30°C with 500 mM vinyl butanoate (Paper IV). app app [a] [a] CALB Substrate kcat K M kcat / K M Pe E variant enantiomer (s-1) (mM) (s-1M-1) R 570 61 9300 Wild-type OH R 1300000 S 0.00053 71 0.0075

R 4.4 29 150 Trp104Ala S 6.6 S 34 34 1000 app app [a] Calculated from kcat and K M .

The enantioselectivity was very high for CALB wild-type, 1 300 000 in favor of the R-enantiomer. The single point mutation in the Trp104Ala mutant changed this 8 300 000 times, resulting in an enantioselectivity of 6.6 for the S- enantiomer. This great change was mainly achieved by the large increase of the app catalytic constant ( kcat ) towards the S-enantiomer. The Michaelis constants app ( K M ) are close to equal towards the two enantiomers, and those of the mutant are only half of those of the wild-type lipase. The specificity constant ( kcat / K M ) towards the R-enantiomer was 63 times lower for the Trp104Ala mutant compared to that of the wild-type, while the specificity constant towards the S- enantiomer increased 130 000 times with the mutation. This large increase can not be explained by the minor change of the Michaelis constant and the Trp104Ala mutation should not have changed the transition-state stabilization. The increase of the specificity constant must be due to a higher ratio of productive binding in the mutant compared to the wild-type enzyme. The redesigned stereospecificity pocket allows for the substrate to be positioned in an orientation optimal for catalysis. The Trp104Ala mutant is a good catalyst towards the S-enantiomer with a specificity constant of 1000 s-1M-1, only 9 times lower than the specificity constant of CALB wild-type towards the R- enantiomer. The enantioselectivity of the Trp104Ala mutant for the acylation of 1- phenylethanol was studied in three solvents as a function of the temperature 40 Reengineering the substrate selectivity

(Figure 6-4). The Trp104Ala mutant was S-selective in all solvents, but the solvent had a large effect on the enantioselectivity. The mutant showed the unusual behavior of an increasing enantioselectivity with temperature, and the temperature effect was also unusually large. The enantioselectivity of the Trp104Ala mutant towards 1-phenylethanol reached a value of 44, at 69°C in cis-decalin (Figure 6-4).

20 E

0 10 20 30 40 50 60 70 T/°C

Figure 6-4. The Trp104Ala mutant of CALB shows an S-selectivity increasing with temperature for the resolution of 1-phenylethanol in acetonitrile (●), cyclohexane (▲), and cis-decalin (■). Trend lines are plotted according to Equation 6-1 using the values of the entropy and enthalpy terms shown in Table 6-5 (Paper IV).

The enantioselectivity determined towards 1-phenylethanol for the Trp104Ala mutant at several temperatures (Figure 6-4) allow the calculation of # # the thermodynamic components ( ∆ S−R ∆S and ∆ S−R ∆H ) controlling the selectivity. The enantioselectivity is related to the thermodynamic components as described by Equation 6-1.

# # # −∆S −R∆G ()−∆S−R∆H +T∆S−R∆S E = e RT = e RT Equation 6-1

Taking the natural logarithm and rearranging Equation 6-1 gives a linear correlation between lnE and T-1 (Equation 6-2).

∆ ∆G # ∆ ∆H # R × ln E = − S−R = − S −R + ∆ ∆S # Equation 6-2 T T S−R

The intercept of Equation 6-2 corresponds to the differential entropy # ( ∆ S−R ∆S ), the slope corresponds to the negative differential enthalpy # ( ∆ S−R ∆H ), and the standard errors are given by the linear regression (Table 6-5, Entry 1-3). The thermodynamic components for the Trp104Ala mutant were

41 Reengineering the substrate selectivity also determined for the resolution of three other substrates (1-phenylethanol, 2- hexanol, and 3-methyl-2-butanol), with cyclohexane as solvent (Table 6-5, Entry 4-6). 3-Methyl-2-butanol was chosen as a substrate to be able to directly compare the Trp104Ala mutant with CALB wild-type in resolution reactions. The wild-type enzyme has a too high enantioselectivity towards the other substrates to allow for an accurate determination of the enantiomeric excess of the substrate and product. For 3-methyl-2-butanol the thermodynamic components could be determined for CALB wild-type (Table 6-5, Entry 7). The preferred enantiomer (Pe) and the enantioselectivity at 303 K are also presented # in Table 6-5, together with the entropic contribution at 303 K (T∆ S−R ∆S ) and the racemic temperature (TR ). The racemic temperature is defined as the ratio of the enthalpic and entropic terms (Phillips, 1996). Assuming that the thermodynamic terms are constant, E equals 1 at this temperature.

Table 6-5. Thermodynamic components for acyl-transfer reactions from vinyl butanoate to secondary alcohols in various solvents catalyzed by the CALB wild-type or Trp104Ala # # mutant. Differential entropy ( ∆ S −R ∆S ) and enthalpy ( ∆ S −R ∆H ) were determined from the −1 # linear regression of R×ln E versus T . Entropic contribution (T∆ S−R ∆S ) and the enantioselectivity ( E ) were calculated for 303 K from the linear relation of Equation 6-2. The

preferred enantiomer (Pe) and the racemic temperature (TR ) are also presented. Paper IV also contains the standard errors for the enthalpic and entropic terms. # # # Entry CALB Secondary Solvent Pe E T∆ S −R ∆S ∆ S −R ∆H ∆ S −R ∆S TR variant alcohol (kJ×mol-1) (J×K-1×mol-1) (K) 1 Trp104Ala AcN[a] S 3.8 21 18 70 250 2 Trp104Ala Deca[b] S 13 34 27.5 112 250 OH 3 Trp104Ala cHex[c] S 7.9 35 30 116 260

4 Trp104Ala cHex[c] S 12 26 20 87 230 OH

5 Trp104Ala cHex[c] S 2.2 23 21 77 280 OH

6 Trp104Ala cHex[c] R 2.8 26 28 84 330 OH 7 Wild-type cHex[c] R 470 7.8 23 26 900 [a] acetonitrile, [b] cis-decalin, [c] cyclohexane

The enthalpic term favored the R-enantiomer and the entropic term favored the S-enantiomer for both CALB wild-type and Trp104Ala. The enthalpic term was in the same range for both enzyme variants, while the entropic term was much larger for the Trp104Ala mutant compared to that of the wild-type lipase. This caused the very different enantioselectivity of the Trp104Ala mutant, compared to that of the wild-type lipase. At 303 K the enthalpy contribution is dominant for the wild-type lipase, making it strongly R-selective. For the resolutions of 3-methyl-2-butanol with the Trp104Ala mutant, the racemic temperature was observed within the experimental range and an inversion of the enantioselectivity occurred at 330 K (Table 6-5, Entry 6). With the other substrates the Trp104Ala mutant had an entropic contribution larger than the enthalpic contribution, making the mutant S-selective. The ratio between the entropic term and the enthalpic term is 3 to 4 times larger for the Trp104Ala 42 Reengineering the substrate selectivity mutant compared to the wild-type, which also can be seen in the much lower racemic temperatures (Table 6-5). The rational redesign of the stereospecificity pocket created an S-selective lipase. The enantioselectivity changed 8 300 000 times towards 1-phenylethanol, mainly by increasing the catalytic constant towards the S-enantiomer. The enantioselectivity increased to a large extent with the temperature and was strongly influenced by the choice of solvent, resulting in an S-selectivity of 44 at 69°C in cis-decalin. Two recent review articles describe the improvement of enzyme activity and stability, as well as their substrate and reaction selectivity (Hult and Berglund, 2003; Bornscheuer, 2002), but examples of altered enantioselectivity through random and rational methods are sparse (Bornscheuer and Pohl, 2001). The most successful rational and random approaches to change the enantioselectivity, found in the literature, are presented for comparison with our change in enantioselectivity of CALB. An impressing example of rationally redesigned enantioselectivity was done on the phosphotriesterase from Pseudomonas diminuta (Chen-Goodspeed et al., 2001). The substrate is positioned in three binding pockets (small, large, and leaving group) located in the active site, and the enzyme hydrolyzes the phosphoester bond (P-O) of the leaving group. The enantioselectivity of the phosphotriesterase was changed by increasing the size of the small and reducing the size of the large binding pocket. The S-selectivity of 35 for the wild-type esterase was increased to 15 000 by one mutation and inverted to an R-selectivity of 460 by 4 mutations, creating two mutants with a difference of 6 900 000 in enantioselectivity. The largest change in enantioselectivity compared to the wild-type esterase was 18 000 times (by 4 mutations) and the largest effect by a single mutation was 520 times. A famous example of improved enantioselectivity through directed evolution was performed with the lipase from Pseudomonas aeruginosa (Zha et al., 2001). The almost non-selective wild-type lipase (S-selectivity of 1.1) was altered to yield one mutant with an S-selectivity of 51 and another mutant with an R- selectivity of 30. This gives a difference in enantioselectivity between the mutants of 1 500 times. In this thesis we presented a change in enantioselectivity of 8 300 000 compared to the wild-type enzyme, which is much larger than the examples above. The change in CALB was also achieved by a single point mutation, as compared to several point mutations or extensive screening.

CONCLUDING REMARKS

The rational redesign of Candida antarctica lipase B successfully introduced substrate-assisted catalysis and changed the lipase into an aldolase by altering the reaction mechanism. Furthermore, both the substrate selectivity and the enantioselectivity were greatly modified. The created CALB mutants have unique specificities that could be useful for chemical synthesis. For example, the mutant showing aldolase activity catalyzes the aldol addition of completely different substrates compared to known aldolases. New lipase variants made it possible to catalyze reactions involving secondary alcohols bearing two large substituents. We created the first lipase that is selective for the S-enantiomer of secondary alcohols. Our knowledge about enzyme catalysis allowed us to efficiently redesign CALB, but a lot more is still to be learnt. Some results were explained by an altered ratio of productive and non-productive binding of the substrate in the active site, which is a concept that needs to be further studied. The entropy was shown to have a large effect on the enantioselectivity and the entropy term was greatly changed by the rational redesign. This shows that the entropy can play an important role in enzyme catalysis; it can not be neglected when an enzyme is redesigned, and should be used as an important tool to create a better catalyst. The constantly increasing knowledge about enzyme catalysis governs the approach of rational redesign to improve the performance of enzymes. In the industry it is important to have an enzyme that is stable, reusable, cheap, and easy to produce. A concept used in a near future could be to have a set of favorite enzymes that fulfill these key requirements, but show different substrate and reaction selectivities. When a suitable catalyst would not exist for a desired application, the best candidate(s) of the favorite enzymes could be redesigned according to what is wished for.

ACKNOWLEDGEMENTS

I would like to acknowledge everyone who has contributed to this thesis.

I am especially grateful to Karl Hult for all the interesting, fascinating, and rewarding discussions about biocatalysis and science in general. Thank you for also having such large understanding about chemistry outside the lab.

I am also very grateful to Mats Holmquist, who first accepted me as a diploma worker and then guided me through the first years of my PhD. You gave me a good start, and I think you passed some of your great sense for organization (even if it does not always appear so).

Per Berglund, the guru of nomenclature and our link to organic chemistry, for all the useful discussions, especially about reaction mechanisms.

Romas Kazlauskas for giving me the chance to visit his lab in Montreal and spend a lot of nice time in this beautiful city.

Linda for your great generosity and all the nice pictures. Joke, coauthor and co-Francophile, for your patience (most of the time) when you were questioned about all the standard approaches at the lab bench. Eva, for your kindness and encouragement over the years and for the cover picture. Cecilia B, coauthor and competing author. I will finish before you. Park, the chemist specialized in all fields, for a lot of good advises and discussions. Stuart Denman, the constantly joking wizard of molecular biology. All “my” diploma workers, Jens, Andreas, Anders, Alberto, and Mohamad, for interesting and fruitful collaborations. All the former and present members of the biocatalysis group for a nice time together. All members of floor 2 for enjoyable lunch discussions and interesting microbiological experiments in the fridge. Alla innebandyspelarna för mycket nöje och många tuffa matcher.

Mor och far, tack för all er uppmuntran och tilltro. Ni har stött mig i mina val och gett mig en trygg grund att utvecklas ifrån. Lena, min allra bästa syster, för att du ”uppskattat” alla mina hyss när vi var små och för din länk till det skotska guldet.

Verner och Gwna, söndagsbruncher, fotbollskickning och antidans. Christian, min gamla granne (trots att jag är ung och du är ännu yngre?). Mycket hyss och en massa sport. Redan som unga utforskade vi ”directed evolution” (svaga äppelträd överlever ej fotbollen), det var ej uppskattat och jag fick senare ägna mig åt ”rationell redesign”. Grabbarna i Långbacken för många och långa pokernätter och ”tidiga” hemfärder på surpuchen.

Michel et Monique, merci de m’avoir adopté dans votre famille, bien que je vole votre fille et votre bon vin. Anne, merci de faire travailler la marmotte pour nous.

Valérie, my love, inspiration, and happiness. You are the best that could happen to me. I love you crazy much.

REFERENCES

Anderson E.M., Larsson K.M., Kirk O. (1998) One biocatalyst - many applications: the use of Candida antarctica B-lipase in organic synthesis. Biocatalysis and Biotransformation, 16: 181-204.

Arnold F.H., Moore J.C. (1997) Optimizing industrial enzymes by directed evolution. Advances in Biochemical Engineering/Biotechnology, 58: 1-14.

Ball P. (2001) Biocatalysis: Synthesis methods that exploit enzymatic activities. Nature Insight, 409: 225.

Bauer S., Schott A.-K., Illarionova V., Bacher A., Huber R., Fischer M. (2004) Biosynthesis of tetrahydrofolate in plants: Crystal structure of 7,8- dihydroneopterin aldolase from Arabidopsis thaliana reveals a novel aldolase class. Journal of Molecular Biology, 339: 967-979.

Berglund P., Park S. (2005) Strategies for altering enzyme reaction specificity for applied biocatalysis. Current Organic Chemistry, 9: 325-336.

Binns F., Harffey P., Roberts S.M., Taylor A. (1998) Studies of lipase- catalyzed polyesterification of an unactivated diacid/diol system. Journal of Polymer Science: Part A: Polymer Chemistry, 36: 2069-2080.

Bornscheuer U.T., Pohl M. (2001) Improved biocatalysts by directed evolution and rational protein design. Current opinion in chemical biology, 5: 137- 143.

Bornscheuer U.T. (2002) Methods to increase enantioselectivity of lipases and esterases. Current Opinion in Biotechnology, 13: 543-547.

Bosma T., Damborský J., Stucki G., Janssen D.B. (2002) Biodegradation of 1,2,3-trichloropropane through directed evolution and heterologous expression of a haloalkane dehalogenase gene. Applied and Environmental Microbiology, 68: 3582-3587.

Branneby C., Carlqvist P., Hult K., Brinck T., Berglund P. (2004) Aldol additions with mutant lipase: analysis by experiments and theoretical calculations. Journal of Molecular Catalysis B: Enzymatic, 31: 123-128.

Breuer M., Hauer B. (2003) Carbon-carbon coupling in biotransformation. Current Opinion in Biotechnology, 14: 570-576.

Cadwell R.C., Joyce G.F. (1992) Randomization of genes by PCR mutagenesis. PCR Methods and Applications, 2: 28-33.

Carlos J.L., Klenotic P.A., Paetzel M., Strynadka N.C.J., Dalbey R.E. (2000) Mutational evidence of transition state stabilization by serine 88 in Escherichia coli type I signal peptidase. Biochemistry, 39: 7276-7283.

Carlqvist P., Svedendahl M., Branneby C., Hult K., Brinck T., Berglund P. (2005) Exploring the active-site of a rationally redesigned lipase for catalysis of Michael-type additions. ChemBioChem, 6: 331-336.

Carrea G., Riva S. (2000) Properties and synthetic applications of enzymes in organic solvents. Angewandte Chemie, International Edition, 39: 2226-2254.

Carter P., Wells J.A. (1987) Engineering enzyme specificity by "substrate- assisted catalysis". Science, 237: 394-399.

Cheetham P.S.J. (2000) Case studies in the application of biocatalysts for the production of (bio)chemicals. Applied biocatalysis, 2nd edition. Straathof A.J.J., Adlercreutz P., Eds., Harwood Academic Publishers: 93-151.

Chen-Goodspeed M., Sogorb M.A., Wu F., Raushel F.M. (2001) Enhancement, relaxation, and reversal of the stereoselectivity for phosphotriesterase by rational evolution of active site residues. Biochemistry, 40: 1332-1339.

Córdova A., Hult A., Hult K., Ihre H., Iversen T., Malmström E. (1998) Synthesis of a poly(є-caprolactone) monosubstituted first generation dendrimer by lipase catalysis. Journal of the American Chemical Society, 120: 13521-13522.

Crameri A., Raillard S.-A., Bermudez E., Stemmer W.P.C. (1998) DNA shuffling of a family of genes from diverse species accelerates directed evolution. Nature, 391: 288-291.

Dall'acqua W., Carter P. (2000) Substrate-assisted catalysis: molecular basis and biological significance. Protein Science, 9: 1-9.

Dufour E., Storer A.C., Ménard R. (1995) Engineering nitrile hydratase activity into a cysteine protease by a single mutation. Biochemistry, 34: 16382- 16388.

Faber K. (2004) Biotransformations in organic chemistry, 5th edition. Springer.: 22.

Fernandez-Lorente G., Fernandez-Lafuente R., Palomo J.M., Mateo C., Bastida A., Coca J., Haramboure T., Hernandez-Justiz O., Terreni M., Guisán J.M. (2001) Biocatalyst engineering exerts a dramatic effect on selectivity of hydrolysis catalyzed by immobilized lipases in aqueous medium. Journal of Molecular Catalysis B: Enzymatic, 11: 649-656.

Fujii R., Utsunomiya Y., Hiratake J., Sogabe A., Sakata K. (2003) Highly sensitive active-site titration of lipase in microscale culture media using fluorescent organophosphorus ester. Biochimica et Biophysica Acta, 1631: 197-205.

Graber R., Kasper P., Malashkevich V.N., Strop P., Gehring H., Jansonius J.N., Christen P. (1999) Conversion of aspartate aminotransferase into an L- aspartate β-decarboxylase by a triple active-site mutation. Journal of Biological Chemistry, 274: 31203-31208.

Hæffner F., Norin T., Hult K. (1998) Molecular modeling of the enantioselectivity in lipase-catalyzed transesterification reactions. Biophysical Journal, 74: 1251-1262.

Halling P.J. (1994) Thermodynamic predictions for biocatalysis in nonconventional media: theory, tests, and recommendations for experimental design and analysis. Enzyme and Microbial Technology, 16: 178-206.

Hill A.C. (1898) Reversible zymohydrolysis. Journal of the Chemical Society, 73: 634-658.

Ho S.N., Hunt H.D., Horton R.M., Pullen J.K., Pease L.R. (1989) Site- directed mutagenesis by overlap extension using the polymerase chain reaction. Gene, 77: 51-59.

Hoffmann T., Zhong G., List B., Shabat D., Anderson J., Gramatikova S., Lerner R.A., Barbas C.F., III (1998) Aldolase Antibodies of Remarkable Scope. Journal of the American Chemical Society, 120: 2768-2779.

Holmquist M., Martinelle M., Berglund P., Clausen I.G., Patkar S., Svendsen A., Hult K. (1993) Lipases from Rhizomucor miehei and Humicola lanuginosa: modification of the lid covering the active site alters enantioselectivity. Journal of Protein Chemistry, 12: 749-757.

Holmquist M. (2000) Alpha/beta-hydrolase fold enzymes: structures, functions and mechanisms. Current Protein and Peptide Science, 1: 209-235.

Houde A., Kademi A., Leblanc D. (2004) Lipases and their industrial applications. An overview. Applied Biochemistry and Biotechnology, 118: 155-170.

Hult K., Berglund P. (2003) Engineered enzymes for improved organic synthesis. Current Opinion in Biotechnology, 14: 395-400.

Jahic M., Rotticci-Mulder J.C., Martinelle M., Hult K., Enfors S.-O. (2002) Modeling of growth and energy metabolism of Pichia pastoris producing a fusion protein. Bioprocess and Biosystems Engineering, 24: 385-393.

Kastle J.H., Loevenhart A.S. (1900) Concerning lipase, the fat-splitting enzyme, and the reversibility of its action. American Chemical Journal, 24: 491- 525.

Kazlauskas R.J., Weissfloch A.N.E., Rappaport A.T., Cuccia L.A. (1991) A rule to predict which enantiomer of a secondary alcohol reacts faster in reactions catalyzed by cholesterol esterase, lipase from Pseudomonas cepacia, and lipase from Candida rugosa. Journal of Organic Chemistry, 56: 2656-2665.

Kirk O., Christensen M.W. (2002) Lipases from Candida antarctica: Unique biocatalysts from a unique origin. Organic Process Research & Development, 6: 446-451.

Klibanov A.M. (1986) Enzymes that work in organic solvents. Chemtech, 16: 354-359.

Klibanov A.M. (2001) Improving enzymes by using them in organic solvents. Nature, 409: 241-246.

Koeller K.M., Wong C.-H. (2001) Enzymes for chemical synthesis. Nature, 409: 232-240.

Lamare S., Caillaud B., Roule K., Goubet I., Legoy M.D. (2001) Production of natural esters at the pre-industrial scale by solid/gas biocatalysis. Biocatalysis and Biotransformations, 19: 361-377.

Lorenz P., Liebeton K., Niehaus F., Eck J. (2002) Screening for novel enzymes for biocatalytic processes: accessing the metagenome as a resource of novel functional sequence space. Current Opinion in Biotechnology, 13: 572-577.

Lozano P., de Diego T., Carrié D., Vaultier M., Iborra J.L. (2002) Continuous green biocatalytic processes using ionic liquids and supercritical carbon dioxide. Chemical Communications, 692-693.

Lozano P., De Diego T., Carrié D., Vaultier M., Iborra J.L. (2003) Lipase catalysis in ionic liquids and supercritical carbon dioxide at 150ºC. Biotechnology Progress, 19: 380-382.

Martinelle M., Holmquist M., Hult K. (1995) On the interfacial activation of Candida antarctica lipase A and B as compared with Humicola lanuginosa lipase. Biochimica et Biophysica Acta, 1258: 272-276.

Martinelle M., Hult K. (1995) Kinetics of acyl transfer reactions in organic media catalyzed by Candida antarctica lipase B. Biochimica et Biophysica Acta, 1251: 191-197. 51

Ménard R., Carrière J., Laflamme P., Plouffe C., Khouri H.E., Vernet T., Tessier D.C., Thomas D.Y., Storer A.C. (1991) Contribution of the glutamine 19 side chain to transition-state stabilization in the oxyanion hole of papain. Biochemistry, 30: 8924-8928.

Nicolas A., Egmond M., Verrips C.T., de Vlieg J., Longhi S., Cambillau C., Martinez C. (1996) Contribution of cutinase serine 42 side chain to the stabilization of the oxyanion transition state. Biochemistry, 35: 398-410.

O'Brien P.J., Herschlag D. (1999) Catalytic promiscuity and the evolution of new enzymatic activities. Chemistry & Biology, 6: R91-R105.

Ohtani T., Nakatsukasa H., Kamezawa M., Tachibana H., Naoshima Y. (1998) Enantioselectivity of Candida antarctica lipase for some synthetic substrates including aliphatic secondary alcohols. Journal of Molecular Catalysis B: Enzymatic, 4: 53-60.

Ollis D.L., Cheah E., Cygler M., Dijkstra B., Frolow F., Franken S.M., Harel M., Remington S.J., Silman I., Schrag J., Sussman J.L., Verschueren K.H.G., Goldman A. (1992) The α/β hydrolase fold. Protein Engineering, 5: 197-211.

Orrenius C., Hæffner F., Rotticci D., Öhrner N., Norin T., Hult K. (1998) Chiral recognition of alcohol enantiomers in acyl transfer reactions catalysed by Candida antarctica lipase B. Biocatalysis and Biotransformation, 16: 1-15.

Ottosson J., Fransson L., King J.W., Hult K. (2002) Size as a parameter for solvent effects on Candida antarctica lipase B enantioselectivity. Biochimica et Biophysica Acta, 1594: 325-334.

Park S., Viklund F., Hult K., Kazlauskas R.J. (2003) Vacuum-driven lipase- catalysed direct condensation of L-ascorbic acid and fatty acids in ionic liquids: synthesis of a natural surface active antioxidant. Green Chemistry, 5: 715-719.

Park S., Morley K.L., Horsman G.P., Holmquist M., Hult K., Kazlauskas R.J. (2005) Focusing mutations into the P. fluorescens esterase binding site increases enantioselectivity more effectively than distant mutations. Chemistry & Biology, 12: 45-54.

Penning T.M., Jez J.M. (2001) Enzyme redesign. Chemical Reviews, 101: 3027- 3046.

Phillips R.S. (1996) Temperature modulation of the stereochemistry of enzymic catalysis: prospects for exploitation. Trends in Biotechnology, 14: 13-16.

Rakels J.L.L., Straathof A.J.J., Heijnen J.J. (1993) A simple method to determine the enantiomeric ratio in enantioselective biocatalysis. Enzyme and Microbial Technology, 15: 1051-1056.

Rheinnecker M., Eder J., Pandey P.S., Fersht A.R. (1994) Variants of subtilisin BPN' with altered specificity profiles. Biochemistry, 33: 221-225.

Rondon M.R., August P.R., Bettermann A.D., Brady S.F., Grossman T.H., Liles M.R., Loiacono K.A., Lynch B.A., MacNeil I.A., Minor C., Tiong C.L., Gilman M., Osburne M.S., Clardy J., Handelsman J., Goodman R.M. (2000) Cloning the soil metagenome: a strategy for accessing the genetic and functional diversity of uncultured microorganisms. Applied and Environmental Microbiology, 66: 2541-2547.

Rotticci D., Hæffner F., Orrenius C., Norin T., Hult K. (1998) Molecular recognition of sec-alcohol enantiomers by Candida antarctica lipase B. Journal of Molecular Catalysis B: Enzymatic, 5: 267-272.

Rotticci D., Norin T., Hult K. (2000a) Mass transport limitations reduce the effective stereospecificity in enzyme-catalyzed kinetic resolution. Organic Letters, 2: 1373-1376.

Rotticci D., Norin T., Hult K., Martinelle M. (2000b) An active-site titration method for lipases. Biochimica et Biophysica Acta, 1483: 132-140.

Rotticci D., Ottosson J., Norin T., Hult K. (2001) Candida antarctica lipase B: A tool for the preparation of optically active alcohols. Methods in Biotechnology, 15: 261-276.

Rotticci-Mulder J.C., Gustavsson M., Holmquist M., Hult K., Martinelle M. (2001) Expression in Pichia pastoris of Candida antarctica lipase B and lipase B fused to a cellulose-binding domain. Protein Expression and Purification, 21: 386- 392.

Schmid A., Dordick J.S., Hauer B., Kiener A., Wubbolts M., Witholt B. (2001) Industrial biocatalysis today and tomorrow. Nature, 409: 258-268.

Schmid A., Hollmann F., Park J.B., Bühler B. (2002) The use of enzymes in the chemical industry in Europe. Current Opinion in Biotechnology, 13: 359-366.

Schoemaker H.E., Mink D., Wubbolts M.G. (2003) Dispelling the myths- Biocatalysis in industrial synthesis. Science, 299: 1694-1697.

Sharma R., Chisti Y., Banerjee U.C. (2001) Production, purification, characterization, and applications of lipases. Biotechnology Advances, 19: 627-662.

Streit W.R., Daniel R., Jaeger K.-E. (2004) Prospecting for biocatalysts and drugs in the genomes of non-cultured microorganisms. Current Opinion in Biotechnology, 15: 285-290.

Tramper J. (1996) Chemical versus biochemical conversion: when and how to use biocatalysts. Biotechnology and Bioengineering, 52: 290-295.

Uppenberg J., Patkar S., Bergfors T., Jones T.A. (1994) Crystallization and preliminary X-ray studies of lipase B from Candida antarctica. Journal of molecular biology, 235: 790-792.

Uppenberg J., Öehrner N., Norin M., Hult K., Kleywegt G.J., Patkar S., Waagen V., Anthonsen T., Jones T.A. (1995) Crystallographic and molecular- modeling studies of lipase B from Candida antarctica reveal a stereospecificity pocket for secondary alcohols. Biochemistry, 34: 16838-16851.

Uyama H., Suda S., Kikuchi H., Kobayashi S. (1997) Extremely efficient catalysis of immobilized lipase in ring-opening polymerization of lactones. Chemistry Letters, 1109-1110.

van Beilen J.B., Li Z. (2002) Enzyme technology: an overview. Current Opinion in Biotechnology, 13: 338-344.

van den Burg B., Vriend G., Veltman O.R., Venema G., Eijsink V.G.H. (1998) Engineering an enzyme to resist boiling. Proceedings of the National Academy of Sciences of the United States of America, 95: 2056-2060.

Voet D., Voet J.G., Pratt C.W. (1999) Enzymatic catalysis. Fundamentals of biochemistry. John Wiley and sons, Eds.: 292-293.

Wells J.A., Cunningham B.C., Graycar T.P., Estell D.A. (1986) Importance of hydrogen-bond formation in stabilizing the transition state of subtilisin. Philosophical Transactions of the Royal Society of London, Series A: Mathematical, Physical and Engineering Sciences, 317: 415-423.

Yow G.-Y., Watanabe A., Yoshimura T., Esaki N. (2003) Conversion of the catalytic specificity of alanine racemase to a D-amino acid aminotransferase activity by a double active-site mutation. Journal of Molecular Catalysis B: Enzymatic, 23: 311-319.

Zha D., Wilensek S., Hermes M., Jaeger K.-E., Reetz M.T. (2001) Complete reversal of enantioselectivity of an enzyme-catalyzed reaction by directed evolution. Chemical Communications, 2664-2665.

Zhao H., Arnold F.H. (1999) Directed evolution converts subtilisin E into a functional equivalent of thermitase. Protein engineering, 12: 47-53.