i

ω-Transaminase in Biocatalysis Methods, Reactions and Engineering

Karim Engelmark Cassimjee

Doctoral thesis

KTH Royal Institute of Technology

School of Biotechnology

Stockholm 2012

ii

Cover picture: The Enzyme at the Heart of the Green Chemical Factory

© David Karim Engelmark Cassimjee 2012 KTH Royal Institute of Technology School of Biotechnology AlbaNova University Centre SE-106 91 Stockholm Sweden ISBN 978-91-7501-242-1 TRITA-BIO Report 2012:13 ISSN 1654-2312 Printed in Stockholm, March 2012 Universitetsservice US-AB Drottning Kristinas väg 53 B SE-100 44 Stockholm Sweden

iii

For Nadia, my one year old daughter, who I suspect will enjoy tearing out the pages (especially the ones with figures).

iv

Abstract

Biocatalysis offers an alternative to classic chemistry by using enzymes, the protein catalysts of Nature, for production of fine chemicals. Evolution has created enzymes capable of catalysis at moderate temperature of a specific reaction in the presence of a plethora of compounds in the aqueous cell environment. The focal point of biocatalysis is to utilise these traits in vitro, for creation of valuable molecules. The ω-transaminase is an enzyme capable of producing chiral amines, compounds used to great extent in pharmaceuticals. Much effort has in recent years been invested in the research and engineering of this enzyme type since the catalysed reaction offers an advantageous alternative to classical techniques. Nevertheless, there is a need for method development, adaptation of the enzyme and increased understanding of the catalytic mechanism for feasibility as an effective biocatalyst for unnatural substrates. This thesis addresses a chosen set of obstacles as a contribution to meeting the demands at hand. ω-Transaminase from Chromobacterium violaceum and Arthrobacter citreus was used. Many homologous ω-transaminases are available, which are also subject to engineering where variants are produced. To accurately compare their kinetic constants an active site quantification method is required but has not been available. Here such a method is presented (Paper 1) which encompasses a virtually irreversible half transamination reaction. In stereoselective synthesis the ω-transaminase catalysed equilibrium reaction inherently results in incomplete conversion. An equilibrium displacement system is presented (Paper II) where isopropylamine is the amino donor for transamination of acetophenone and derivatives thereof, coupled to an enzymatic cascade reaction. For many unnatural substrates the specificity and enantiospecificity is insufficient. Rationally redesigned variants were produced with improved properties for chosen substrates (Paper III and IV). The catalytic contributions of field and resonance of a variant compared to the wild type were investigated (Paper IV) for increased knowledge of the mechanism. For rational redesign of an enzyme the three-dimensional structure is required, of which only a few are available for the ω-transaminases. X-ray crystallographic structures of the holo and apo form of Chromobacterium violaceum ω-transaminase were made (Paper V) which revealed significant structural rearrangements upon coenzyme binding which may be of consequence for future engineering.

v

Sammanfattning

Biokatalys utgör ett alternativ till klassisk kemi genom att finkemikalier produceras med hjälp av enzymer, naturens katalysatorer. Naturen har genom evolution skapat enzymer med förmåga att vid måttlig temperatur katalysera en specifik reaktion i närvaro av ett överflöd av föreningar i vattenmiljön hos den levande cellen. Huvudsyftet med biokatalys är att utnyttja dessa egenskaper in vitro, för att producera värdefulla molekyler. ω-Transaminaset är ett enzym med förmåga att producera kirala aminer, föreningar som används i läkemedelsutveckling i stor omfattning. Mycket forskning har under de senaste åren investerats i detta enzymslag eftersom den katalyserade reaktionen utgör ett fördelaktigt alternativ till klassiska tekniker. Dock finns det behov av metodutveckling, anpassning av enzymet och ökad förståelse för den katalytiska mekanismen för dess effektiva användning som biokatalysator för onaturliga substrat. Denna avhandling behandlar valda begränsande faktorer, ett bidrag till att möta de krav som ställs. ω-Transaminaser från Chromobacterium violaceum och Arthrobacter citreus användes. Många homologa ω-transaminaser finns tillgängliga, vilka också är föremål för utveckling av enzymvarianter. För att jämföra kinetiska data mellan dessa är en metod för kvantifiering av aktiva säten nödvändig. En sådan metod presenteras (Artikel I) som innefattar en synbart irreversibel halv-transamineringsreaktion. I stereoselektiv syntes med ω-transaminas resulterar den katalyserade jämviktsreaktionen i ofullständig omsättning. Ett jämviktsförskjutningssystem presenteras här (Artikel II) där iso-propylamin är aminodonator för transamineringen av acetofenon och dess derivat, kopplat till en enzymatisk kaskadreaktion. För många onaturliga substrat är specificiteten och enantio-specificiteten otillräckliga. Rationellt designade varianter produ-cerades med förbättrade egenskaper för utvalda substrat (Artikel III och IV). De katalytiska bidragen från fält- och resonanseffekter hos en variant jämfört med vildtypen undersöktes (Artikel IV) med syfte att öka kunskapen om reaktionsmekanismen. Rationell design av ett enzym fordrar dess tredimensionella struktur, endast ett fåtal sådana är tillgängliga för ω-transaminaser. Röntgenkristallografiska strukturer av holo- och apo-former av Chromobacterium violaceum ω-transaminas togs fram (Artikel V), som visade tydliga strukturella omorganiseringar vid koenzym-bindning som kan vara av betydelse för framtida utveckling.

vi

This thesis is based on the following scientific publications: I Active Site Quantification of an ω-Transaminase by Performing

a Half Transamination Reaction Karim Engelmark Cassimjee1, Maria Svedendahl Humble1, Valentina Miceli, Carla Granados Colomina, and Per Berglund ACS Catalysis 2011, 1, 1051-1055

II Transaminations with Isopropyl Amine: Equilibrium Displacement with Yeast Alcohol Dehydrogenase Coupled to in situ Cofactor Regeneration Karim Engelmark Cassimjee, Cecilia Branneby, Vahak Abedi, Andrew Wells and Per Berglund Chemical Communications 2010, 46, 5569-5571

III Key Amino Acid Residues for Reversed or Improved Enantiospecificity of an ω-Transaminase Maria Svedendahl Humble1, Karim Engelmark Cassimjee1, Vahak Abedi, Hans-Jürgen Federsel and Per Berglund ChemCatChem 2012, In press, doi: 10.1002/cctc.201100487

IV Chromobacterium violaceum ω-Transaminase Variant Trp60Cys Shows Increased Specificity for (S)-1-Phenylethylamine and 4’-Substituted Acetophenones, and Follows Swain-Lupton Parameterisation Karim Engelmark Cassimjee, Maria Svedendahl Humble, Henrik Land, Vahak Abedi and Per Berglund Submitted manuscript

V Crystal Structures of the Chromobacterium violaceum ω-Transaminase Reveal Major Structural Rearrangements Upon Binding of Coenzyme PLP Maria Svedendahl Humble, Karim Engelmark Cassimjee, Maria Håkansson, Yengo Kimbung, Björn Walse, Vahak Abedi, Hans-Jürgen Federsel, Per Berglund and Derek T. Logan FEBS Journal 2012, 279, 779-792

The above publications will be referred to by their assigned roman numerals.

1 Shared first authorship

vii

Summary of the contributions of Karim E. Cassimjee (KC) to Papers I to V Paper I: KC supplied the idea of using a virtually irreversible half transamination reaction for active site quantification, designed and performed the active site quantification method, designed the molecular cloning, and performed the majority of the writing.

Paper II: KC supplied the idea of utilizing the selectivity of yeast alcohol dehydrogenase for equilibrium displacement, carried out the molecular cloning, reactions and analysis, and did the majority of the writing.

Paper III: KC designed the molecular cloning and the experimental procedures for enantioselectivity determination, and took part in the rational redesign and the writing.

Paper IV: KC performed the molecular cloning, reactions and analysis, reaction parameterisation, took part in the finding of the key point mutation, and performed the majority of the writing.

Paper V: KC designed the molecular cloning, performed stability experiments, and took part in the writing.

Publications by KC, not included in this thesis • One-Step Enzyme Extraction and Immobilization for

Biocatalysis Applications Karim Engelmark Cassimjee, Robert Kourist, Diana Lindberg, Marianne Wittrup Larsen, Nguyen Hong Thanh, Mikael Widersten, Uwe T Bornscheuer and Per Berglund Biotechnology Journal 2011, 6, 463-469

• Silica-Immobilized His6-Tagged Enzyme: Alanine Racemase in Hydrophobic Solvent Karim Engelmark Cassimjee, Martin Trummer, Cecilia Branneby and Per Berglund Biotechnology and Bioengineering 2008, 99, 712-716

• Synthesis of Cyclic Polyamines by Enzymatic Generation of an Amino Aldehyde in situ Karim Engelmark Cassimjee, Sílvia Rodríguez Marín and Per Berglund Submitted manuscript

• Process for the Production of Polyamines Bernhard Hauer, Karim Engelmark Cassimjee, Per Berglund 2011 Patent publication number: WO2011051433

viii

List of Abbreviations 1-PEA 1-phenylethylamine Cv-ωTA Chromobacterium violaceum ω-transaminase E-value ratio of specificity constants (kcat/KM) of enantiomers ee enantiomeric excess Enz enzyme eq. equation GC gas chromatography HPLC high pressure liquid chromatography IMAC immobilised metal affinity chromatography IPA isopropylamine PLP pyridoxal-5’-phosphate PMP pyridoxamine-5’-phosphate RMSD root-mean-square deviation U enzyme activity unit, defined as (µmol/min)

Amino acid (residue) abbreviations Ala, A alanine Arg, R arginine Asn, N asparagine Asp, D aspartate Cys, C cysteine Glu, E glutamate Gln, Q glutamine Gly, G glycine His, H histidine Ile, I isoleucine Leu, L leucine Lys, K lysine Met, M methionine Phe, F phenylalanine Pro, P proline Ser, S serine Thr, T threonine Trp, W tryptophan Tyr, Y tyrosine Val, V valine

ix

”A tidy laboratory means a lazy chemist.”

Jöns Jacob Berzelius (1779-1848)

Note: If the above statement is valid in the opposite sense, i.e. a messy laboratory means a devoted chemist; my co-workers and I are possibly the most energetic chemists in the country.

x

Contents

1 Introduction 1 

1.1 Enzymes, proteins with catalytic ability 1 1.2 Enantiospecificity, an enzymatic trait 4 1.3 Enzyme kinetics 6 1.4 Biocatalysis 8 1.4.1 Enzyme engineering 9 

1.5 ω-Transaminase, a PLP-dependent enzyme 10 

2 Aims and Scope 15 

3 Active Site Quantification of Cv-ωTA (Paper I) 17 

4 Equilibrium Displacement of ω-Transaminase Reactions with Isopropylamine as Amino Donor (Paper II) 21 

5 Rational Redesign of Cv-ωTA (Paper III and IV) 27 

5.1 Improved enantioselectivity and reversed enantiomeric preference (Paper III) 29 5.2 Increased specificity (Paper IV) 32 

6 Structure of Cv-ωTA (Paper V) 39 

7 Concluding Discussion 44 

7.1 Reactions and engineering with ω-transaminase, summarising discussion of Paper I-V 45 7.2 Future prospects 48 7.2.1 Dynamic kinetic resolution of an amine 48 7.2.2 Mechanistic studies 49 

8 Acknowledgements 50 

9 References 51 

Appendix 57 

1

1 Introduction This thesis is a contribution to the field of biocatalysis1 in which methods, reactions and engineering with ω-transaminase are presented. Biocatalysis is an integral part of industrial or white biotechnology.2,3

Fine chemicals, especially pharmaceuticals, are undoubtedly important products which have become essential for our society. Their fabrication is a continuously developing area where there is a demand for new and improved synthesis routes for target compounds. A frequent motif in pharmaceuticals is the chiral amine,4 a biofunctional group also present in natural metabolism.

In this section enzymology (with related concepts in chemistry), the field of biocatalysis and the ω-transaminase enzyme are briefly described. The concepts and observed traits of enzymes are put into the context of how they may be applied for engineering and synthesis of fine chemicals, chiral amines in particular.

1.1 Enzymes, proteins with catalytic ability In Nature, living organisms depend on a functioning metabolism. Cellular structures are created from simpler building blocks. Compounds are formed and broken down in what is described as controlled metabolic pathways in which energy is rearranged and utilised for vital processes. The lion’s share of the chemical reactions in these processes takes place in an aqueous environment and are catalysed by enzymes. The name, enzyme, originates from ancient Greek, meaning “in sourdough”. Enzymes are the foundation of chemical life; they uphold essentially all chemical reactions at a viable rate within a living organism.1,5

Proteins are polymers (polyamides) formed by the condensation of α-amino acids (molecules which contain an amino group and a carboxylic acid group with one methylene) which are found in all known forms of life. Generally, residues of twenty types of amino acid are present in proteins. The amino acids are joined in a defined

2

Figure 1: Graphical representation of an α‐helix (left) and β‐sheets (right), common motifs in proteins

order based on a gene sequence, in which each amino acid residue contributes with a distinct side chain with properties elegantly chosen by evolution. The proteins fold to form ordered vital structures, biologically functional units. The structures formed by the amino acid residues are many; the most common motifs are the α-helix and the β-sheet which are usually depicted in three-dimensional structures, shown in figure 1. When folded, the protein molecule can form a globular structure with a hydrophobic core, a result of minimization of hydrophobic exposure in an aqueous environment. The individual protein molecules may exist alone or form complexes with others. A protein which is capable of catalysing a chemical reaction is called an enzyme.

Observed at macro scale proteins behave in unexpected ways. E.g. with rising temperature a protein solution (such as egg white) will turn into a semi solid (an omelette6) whereas many other com-pounds or solutions thereof typically melt or lower in viscosity. This is a result of unfolding and subsequent aggregation of the protein chains. Looking at proteins in micro scale also yields fascinating findings. An enzyme is able to bind one or more distinct molecules or atoms, its substrates, and provide an environment which is more favourable for a chemical reaction of the substrate(s) than the surrounding solution. The location in the enzyme where this takes place is called the active site. According to the transition state theory,7 applicable for all types of catalysis, the reaction rate increase can be attributed to a lowering of the activation energy.

3

The mechanisms vary between enzymes and remain a matter of debate. The theory of near attack conformers (NAC)8 presented by Bruice and co-workers attributes the lowering of the activation energy to the ability of an enzyme to place its substrate(s) in positions which are favourable for reaction. Another theory, presented by Warshel and co-workers, assigns the catalytic ability to the presence of favourable electrostatic interactions9 in the active site which stabilise the transition state. The enzyme macromolecule is a flexible structure, all known enzymes display this feature.10 The ability of conformational adaptation to favourable conformations during a catalytic process has been presented by Hammes-Schiffer and co-workers as being of paramount importance.10 In this view the substrate binding, transition state stabilisation, product release, and all transitions in-between, require different conformations of the enzyme for effective catalysis.

Some enzymatic reactions require a non-protein moiety, a specific molecule bound in the active site, for the catalysis to take place. Such moieties are called coenzymes.11 When the coenzyme is bound the complex is denoted as the holo enzyme, whereas the non-catalytic unit without coenzyme is called the apo enzyme. Coenzymes are distinguishable from substrates since they are regenerated, not consumed in the reaction. If the holo enzyme is stable and the coenzyme is not degraded no more than one equivalent of coenzyme to apo enzyme is needed for complete activation, i.e. complete formation of holo enzyme, in a given reaction. When non-natural conditions or non-natural substrates are applied, arguably, there is a possibility of increased instability of the enzyme-coenzyme complex; this should be taken into consideration when enzymes are used in chemical synthesis. When an enzyme catalyses an oxidation reaction a cosubstrate is used for a complementary reduction reaction, e.g. NAD+ can be reduced to NADH. The reversed reduction reaction will, naturally, yield NAD+ from NADH as a complementary oxidation.5 For in vitro synthesis applications the cosubstrate may be regenerated by an auxiliary reaction to circumvent the need for higher concentrations.

Inhibition of enzyme catalysis is a common observation, the reaction rate may drop in the presence of a substance (an inhibitor) or higher substrate or product concentrations.12,13 Inhibiting and activating

4

compounds are used by a living cell to control the flow of metabolism.5 When substrate concentrations are significantly higher than what is observed in the cell environment, arguably, inhibition arises as a consequence of the non-natural situation. The active site is in essence blocked by the presence of more than one substrate. In ordered multi substrate reactions high concentrations may lead to blockage by substrates binding in incorrect order. For in vitro applications catalysis at high concentrations may be desired, but in nature no evolutionary pressure existed to achieve this ability.

1.2 Enantiospecificity, an enzymatic trait A molecule is defined as chiral if it cannot be superimposed upon its mirror image, much alike a human hand. A molecule is not chiral (achiral) if it has symmetry elements of the second kind, i.e. a mirror plane, a centre of inversion, or a rotation-reflection axis. Chiral versions of otherwise identical compounds are termed enantiomers. In nature such molecules are predominantly existent as one of the possible versions. In chemistry, the common way of denoting the enantiomers are with an (S) (left handed) or (R) (right handed). Amino acids are usually classified under a different nomenclature which allows the majority of the in nature existing ones to be designated as L-enantiomers, as opposed to the corresponding D-enantiomers.14 An equal mix of enantiomers is referred to as a racemate.

An enzyme is composed only of L-amino acid residues and is thereby chiral and enantiopure. The interaction of an enzyme with chiral substrates can differ with the enantiomers. It may possess a high catalytic effect for one enantiomer of the substrate while the turnover of the other is low, sometimes virtually undetectable. The interaction of enantiomers with an enzyme can be envisioned as analogous to hands in gloves, depicted in Figure 2.

The enantiomers may be viewed upon as different substrates. Specificity is the term used to describe the ability of an enzyme to distinguish between substrates, enantiospecificity when the substrates are enantiomers. The term substrate specificity is generally used to describe different catalytic activity for similar substrates (not enantiomers). Enzymes also exhibit chemospecificity

5

NH2 H2N

H H

Figure 2: A pair of human hands (the author’s) next to a right hand glove compared to the (S)‐ and (R)‐forms of 1‐phenylethylamine next to a ribbon representation of an ω‐transaminase enzyme. However rotated, either hand or molecule cannot be superimposed on the other. The right hand fits in the glove, analogous to how one of the enantiomers fits in the active site of the enzyme. 

for similar reactions, and regiospecificity for analogous reactions on different regions of a substrate. Figure 3 shows examples of substrates subject to specificity. An examination of the reactions and compounds of metabolic pathways reveals the necessity of these traits in order to uphold the order of metabolism. In essence, the enzyme is evolved to catalyse a reaction specifically enough for the remaining molecules in its environment to be sufficiently unperturbed for continued metabolism to function sufficiently. In this way a cell can produce one type of sugar, protein or lipid while consuming another.

NH2 NH2

OH OH

NH

O

O

O

Figure 3: Examples of substrates which are subject to chemo‐, regio‐ or substrate specificity with examples of enzyme types.

substrate specificity ω‐transaminase 

chemospecificity amidase or esterase 

regiospecificity alcohol dehydrogenase 

6

1.3 Enzyme kinetics When steady-state conditions can be assumed, and the enzyme concentration is significantly lower than that of the substrate (the usual case) the reaction rate of an enzyme catalyzed reaction can be described by the Michaelis-Menten equation.15-19 The equation in its simplest form, for a reaction with one substrate without inhibition, gives:

E A

M A (eq. 1)

where v is the observed reaction rate, kcat the turnover number, KM the Michaelis-Menten constant, Enz the enzyme, and A the substrate. From the equation follows that when the substrate concentration is significantly lower than KM the rate is given by:

M

Enz A (eq. 2)

by ignoring [A] in the denominator. This is the usual situation in a living cell. When the substrate concentration is significantly higher than KM which can then be ignored, the rate can be described by:

Enz (eq. 3)

This is a consequence of saturation of the active sites; hence the maximum rate denoted as Vmax. The KM is a combination of rate constants for substrate binding, substrate dissociation and product formation. Since the rate of product formation is generally slower than substrate binding and dissociation, the KM can be viewed upon as an inverse measure of the affinity of a substrate; a higher value signifies a lower affinity. The division of the turnover number with the Michaelis-Menten constant, kcat/KM, is a measure of the catalytic efficiency, termed the specificity constant.12 In many cases the specificity constant is the chosen number for comparison, rather than the turnover number or the Michaelis-Menten constant. The specificity constant may well have been deemed the main enzymatic constant in the original works by Henri,15 and Michaelis and Menten,16 where it was in fact circuitously reported as a central factor.17 If it were, the common nomenclature would probably have

7

designated the specificity constant as the Michaelis-Menten constant.

For reactions with competing substrates, multi-substrate reactions, reactions with inhibition, etc, the rate equations become more complex. The enzyme described in section 1.5, an ω-transaminase, catalyses a two-substrate reaction with an ordered ping-pong bi-bi mechanism.20 This means that two substrates enter the active site in a defined order and the first product leaves before the second substrate enters. The rate equation for such a situation is written as:

E A B

MA B M

B A A B (eq. 4)

where B is the second substrate, remaining letters follow the definitions above. To graphically describe equation 4 a three-dimensional plot is needed; experimentally many rates need to be recorded where the two substrates are individually varied. This is not always of practical importance; instead a pseudo one-substrate model may be employed. Equation 4 may be rewritten as follows:

E A

MA A M

B

B

(eq. 5)

By keeping the concentration of one of the substrates at a constant level, e.g. B in equation 5, apparent kinetic constants can be experimentally deduced which are then valid at this chosen concentration. The bracketed expression in the denominator of equation 5 will be treated as a constant. This simplification is valuable when comparisons of different substrates are made, with the second substrate in common. When comparing enzymes or enzyme variants pseudo one-substrate kinetics may also be applied but the results cannot directly be regarded as general since the affinity of the constant substrate (B in equation 5) may vary and give arbitrary results only valid at the given reaction conditions. However, equation 5 will simplify to equation 2 if [A] is small enough. This means that the true specificity constant of one substrate of an ordered ping-pong bi-bi reaction may be deduced by studying the linear concentration range of the rate dependence with the other substrate at a fixed concentration.

8

In a situation where two substrates are allowed to compete for an enzyme, in a one-substrate reaction, the rate equation for an individual substrate is given by:

AA E M

B A

MB A M

A B MA

MB (eq. 6)

The ratio of the reaction rates, given by the division of equation 6 for e.g. substrates A and B, will simplify to the division of the specificity constants if the substrate concentrations are equal:

A

B

M

A

M

B (eq. 7)

When the substrate is a racemate the enantiomers can be regarded as competing substrates. The level of enantiospecificity, denoted as the E-value, is defined as the reaction rate of the faster reacting enantiomer divided by the rate of the slower reacting, hence the ratio of their specificity constants as in equation 7:21,22

⁄

⁄. (eq. 8)

The E-value can be used for assessing the usefulness of an enzyme for synthesizing an enantiopure compound. Henceforth, the word enantiospecificity will be used to describe the E-value. The word enantiopreference (enantiomeric preference) will be used to describe which enantiomer is favoured, without regard for the degree of specificity.

1.4 Biocatalysis The field of biocatalysis is a merge of biology, chemistry, enzymology and biotechnology, with an aim of using enzymes for chemical synthesis. Theoretically, the enzymatic traits described in section 1.2 can be used for production of fine chemicals; beneficially when other methods fail or are inefficient. Enzymes may provide the chemist with a means of creating enantiopure compounds, performing specific reactions in the presence of other reactive compounds and/or steering of reactions to specific groups, in an environmentally compatible setting.23 An enantiospecific enzyme

9

may e.g. be used for the production of an enantiopure pharmaceutically active compound; the corresponding enantiomer may be toxic or have an undesired effect in the human body.5,24

Classic synthetic chemistry often employs organic solvents (i.e. non-aqueous environments), high temperatures and metal catalysts. Numerous efficient processes have been developed in this field, and it is undoubtedly the foundation of today’s chemical industry. Still, nature has been performing advanced chemical transformations in aqueous environments at moderate temperature by employment of enzymes since the beginning of life. The harnessing, mimicking and modification of these natural processes remain an engineering challenge, but with great promise of reward by an environmentally sustainable chemical industry.25-27

1.4.1 Enzyme engineering The modification of enzymes for adaptation to the environment in a chemical process is often considered necessary, for which many strategies have been developed.28 Often the compounds in the process are poor substrates, the enantiospecificity may be insufficient, or a more heat stable or co-solvent tolerable enzyme is desired. By using an experimentally determined structure of the enzyme (determined by X-ray chrystallography or NMR) the enzyme engineer can locate specific amino acid residues suitable for mutation. This is called rational redesign; arguably, the amount of knowledge which is available regarding the function of a given enzyme dictates the likelihood of success with this strategy. When such attempts are unsuccessful, or a structure is unavailable, the amino acid chain can be altered in a randomised fashion and a library of mutants can be screened for the desired properties. An improved clone can be further randomised and a new library can be screened. The iterative process is repeated for the potential discovery of the required enzyme. This strategy is called directed evolution.29-33 Further, to increase the probability of finding improved variants, semi-rational28 approaches have been devised, such as CASTing34,35 and saturation mutagenesis. The more recent scientific publications involving enzyme engineering contain examples of rational and semi-rational approaches with a higher frequency than previous years. The method advancements, increased understanding of function and examples of successful enzyme variants have made researchers

10

less inclined to resorting to directed evolution, which most often involves a tedious screening procedure. The choice of method is usually not straightforward, and remains a challenge for the enzyme engineer.36

Notwithstanding the effectiveness of specificity (section 1.2), enzymes may also display promiscuity,37-40 the ability to function in a different way than for which the enzyme appear to be evolved. Promiscuity is theorised to be a foundation for protein evolvability,41 a starting point for the emergence of beneficial properties.

When displaying condition promiscuity an enzyme may be employed in e.g. non-aqueous environment. Substrate promiscuity can be applied when reactions of non-natural substrates are desired. Catalytic promiscuity, the ability to catalyse several types of similar reactions, can be used for new desired reactions.37,38

Ample promiscuous effects arise predominantly when an enzyme is introduced to an environment which is different from that in which it was evolved; enzyme promiscuity is obviously not significantly detrimental to a living cell in its natural environment. The enzyme engineer may utilize and reinforce a desired promiscuous effect by simply changing the solvent, or creating enzyme variants with the methods described above.37-40,42

1.5 ω-Transaminase, a PLP-dependent enzyme The coenzyme pyridoxal-5’-phosphate (PLP, Figure 4), a derivative of vitamin B6, is utilized by a broad range of enzymes for catalysis of e.g. decarboxylation, racemisation, retro-aldol cleavage and transamination.43-45 PLP readily forms Schiff bases,14 condensation products of amines and ketones. This feature is utilized in many enzymes with amino substrates.43,44,46 The Schiff base structure enables resonance stabilisation of the formed carbanion after abstraction of, most commonly, a proton from the carbon bound to the Schiff base nitrogen.43,44 The aldehyde in Figure 4 is the free form of PLP, when bound to an enzyme PLP generally forms a Schiff base with a lysine residue in the active site.44,45 The catalytic mechanism of a given PLP-dependent enzyme can therefore be depicted as a reaction of a Schiff base complex, the holo enzyme,

11

N

OH

O

PO

OHO

HO

Figure 4: The coenzyme PLP (pyridoxal‐5’‐phosphate), derivative of vitamin B6.

with the amino substrate in question; the free aldehyde in Figure 4 and its reaction with the apo enzyme is not considered a part of the catalytic mechanism per se.

In the substrate-PLP Schiff base structure, the proton bound to the Schiff base nitrogen is most favourably abstracted if it is positioned anti-periplanar to the conjugated π-electron system.43,44 Therefore, effective catalysis cannot be performed solely by the presence of enough space for the substrate, but by the ability of the substrate to be positioned in a catalytically productive manner. This was demonstrated by Blackburn and co-workers by rational redesign of an alanine racemase which was able to accept serine as substrate when a variant enabled anti-periplanar positioning of the proton in question.47 The situation may well be described as the formation of a NAC.

NH2O ONH2ω-TA

Figure 5: A transamination equilibrium reaction where (S)‐1‐phenylethylamine and acetone is produced from acetophenone and isopropylamine, catalysed by an ω‐transaminase (ω‐TA).

Transamination is the process in which an amine and a ketone react to interchange their functional groups, producing the corresponding ketone and amine (Figure 5). In a living cell an amino acid can in this way be formed from the corresponding keto acid, by PLP-dependent transaminase enzymes.5 In this case the transaminase is labelled by the Greek letter α, having an α-amino acids and α-keto acids as substrates and products (not the reaction depicted in Figure 5).

Transaminases which are able to accept aliphatic ketones and amines as their substrates, i.e. not only keto acids and amino acids, are labelled with the Greek letter omega (ω-transaminase, EC

12

NH

OHPO O

O

NH

OHPO

OO

OLys

NH2NH

H

NH

OHPO

OO

OLys

NH2NH

NH

OHPO

OO

O

NH

NH

OHPO

OO

O

NH2

O

NH

OHPO

OO

O

NH

Lys

NH2

O

O

H

NH

OHPO

OO

O

NH

Lys

NH2

NH

OHPO

OO

O

NH

Lys

NH2

O

ONH2

O

O

H2O

H2O

ONH

Lys

Planar quinonoid

Planar quinonoid

PMP

NH2

O

O

OO

O

Lys

NH2

Lys

NH3

H

H

Figure 6: Proposed reaction mechanism of an (S)‐selective ω‐transaminase. PLP is, in the holo enzyme, covalently bound to a lysine residue (top left) as an internal aldimine. An amino donor, here (S)‐1‐phenylethylamine, reacts with the internal aldimine to form PMP and a keto compound (here acetophenone). Then, a keto substrate, here pyruvate, reacts with PMP to form alanine. This equilibrium reaction in this case strongly favours the products. Adapted from the mechanism deduced for aspartate α‐transaminase.20,46

2.6.1.18).48 Figure 6 depicts the putative reaction mechanism, adapted from the mechanism of aspartate α-transaminase.20,46

In the proposed mechanism, the amino donor substrate forms a Schiff base with PLP and lowers the pKa of a proton which thereby may be abstracted to form a planar quinonoid moiety. Thereafter electron and proton rearrangements and a hydrolysis reaction yields the corresponding ketone and PMP (Figure 6). When another ketone enters the active site essentially the same reaction steps in reverse will result in the reformation of the holo enzyme and the corresponding amine. The enantiospecificity is presumably a result of unfavoured coordination of one enantiomer in a catalytically

13

productive conformation, due to lack of space. In the ω-trans-aminase from Vibrio fluvialis the active site can be modelled as exhibiting a large and a small pocket in which the groups bound to the nitrogen binding carbon on the substrate are positioned after reaction with PLP/PMP.49 Since the proton is abstracted by the catalytic lysine, which also binds PLP in the holo enzyme, positioned on one side of the Schiff base structure one distinct conformation will be favoured.49 The same rational is assumed to apply for the ω-transaminase from Chromobacterium violaceum and Arthrobacter citreus which are used in this work.

Amines are found in living cells. Dopamine, epinephrine (adrenaline) and serotonin are examples of amines present as signal substances in mammals. Amines are also common motifs in pharmaceuticals. Synthesis of a wide range of enantiopure amines are of great interest for the pharmaceutical industry.50-52 The use of ω-transaminases for this purpose has been shown to be feasible and is an attractive option since the enzymes have been found to display substrate promiscuity and high enantiospecificity.48,50-67 Still, effective synthesis requires engineering for addressing inherent obstacles in the enzymes function. The equilibrium reaction results in low product yield. Also, many substrates and products inhibit the enzyme which practically halts the reaction at high concentrations. These are unwanted limiting factors in large scale synthesis.

ω-Transaminase reactions can be categorised by distinguishing between stereoselective synthesis and kinetic resolution reactions (see Figure 7). The latter is referring to a process in which the racemate of an amine is used. One enantiomer is removed by the enantiospecific transaminase leaving the other, which can be extracted in enantiopure form. In this case the maximum yield is fifty percent. In synthesis mode the enantiopure amine is synthesised from the corresponding prochiral ketone. This reaction can in theory reach a yield of one hundred percent if the equilibrium is displaced; in many examples of stereoselective synthesis reactions the equilibrium favours the substrates.48,61

The impediment of the transamination reaction by substrate and product inhibition has been tackled by means of directed evolution with some success.64,68,69 Equilibrium displacement has been done

14

A kinetic resolution ONH3 NH2NH2 Oω-TA

B stereoselective synthesis NH2O ONH2ω-TA

Figure 7: Examples of two types of reactions with an (S)‐selective ω‐transaminase (ω‐TA). In A, a reaction defined as performed in resolution mode to obtain the (R)‐enantiomer with a possible theoretical yield of 50%, and in B, a reaction in synthesis mode to obtain the (S)‐enantiomer with a 100% possible theoretical yield if the equilibrium can be displaced. 

by removal of the products or large excess of amino donor.48,66,70-75 When using isopropylamine, a cost efficient chemical, as amino donor acetone is produced beside the enantiopure amine. Acetone is volatile and can be removed by distillation.71 But, in larger scale distillation is not always feasible since it requires efficient mixing and heating; also, the distillation procedure will result in evaporation of isopropylamine to a varying degree depending on pH (analogous to 2-butylamine65). Raising the temperature also requires a heat stable enzyme.

15

2 Aims and Scope The overall aim of this work was the development of methods, creation of improved enzyme variants and gathering of information; to facilitate the use of ω-transaminases for chemical synthesis of enantiopure amines. Synthesis thereof with classical methods has limitations.52 Development of ω-transaminase for biocatalysis applications has therefore become attractive. Nevertheless, the ω-transaminase reaction has, from a biocatalytic perspective, inherent boundaries with opportunity for engineering and method develop-ment. The main obstacles targeted here can be summarised as follows:

• The equilibrium reaction results in poor product yields. • There is a lack of R-selective homologues or variants. • The specificity and enantiospecificity for a variety of desired

substrates are insufficient. • Correct kinetic comparisons of homologues and variants are

impossible due to a lack of an active site titration method. • Only a small number of three-dimensional structures have

been published, which limits the possibilities for rational redesign.

At present there are examples of successful biocatalysis with ω-transaminases where the said boundaries have been broken by reaction and enzyme engineering. Still, additions to this field are indeed needed; issues of importance were here selected and resolved. ω-Transaminase enzyme from Chromobacterium violaceum (Cv-ωTA), and from Arthrobacter citreus in one case, was used.

An alternative solution for displacing the equilibrium in stereoselective synthesis was needed after identifying isopropyl-amine as a cost efficient amino donor. An enzymatic cascade reaction where enzyme selectivity was used to form an equilibrium displacement system was investigated.

In a rational redesign approach, creation of variants with increased activity, shifted enantiopreference and increased enantiospecificity were pursued. The goal, apart from creation of enzyme variants

16

suitable for future synthesis applications, was to find key residues in the active site.

To correctly assess and compare the activity of the engineered enzyme variants a method for quantifying the amount of active sites is required. Therefore the establishment of such a method was considered important and part of the aim.

Since no crystal structure of the enzyme was available a homology model was first made on which the engineering work was based. At a later stage a crystal structure was included in the scope of the project after viewing the homology model as insufficient for describing experimental findings.

17

3 Active Site Quantification of Cv-ωTA (Paper I) Active sites can be titrated with e.g. a suicide inhibitor,76 the purpose is to determine the actual amount of active enzyme. The misfolded or for other reasons inactive enzyme fraction in a given cultivation batch may vary. A given preparation may also contain proteins and other contaminants which will influence concentration determination with total protein quantification methods. Once the molar amount of active sites has been determined the turnover number (number of converted substrate molecules per time unit and active site, usually given in units of s-1) for a given reaction can be calculated. To accurately compare homologous enzymes, or enzyme variants, the turnover number is required. If the specific activity (i.e. the activity of total enzyme or total protein) is instead used for the comparisons the result may be greatly misleading since the true enzyme amount is unknown.13

PLP-dependent enzymes, including some transaminases, have shown to be irreversibly inhibited by D-cycloserine and gabaculine,77,78 which therefore may possibly serve as active site titrants for Cv-ωTA. Using a derivative of PLP, namely 1-deazapyridoxal-5'-phosphate, proved to yield a stable aldimine in an aspartate transaminase.79 The Cv-ωTA was partially inhibited by gabaculine, shown by Ward and coworkers.80 Instead of testing the putative usefulness of inhibitors the strategy in this work was to use a reaction which was known to be catalyzed by the enzyme for active site quantification, not titration.

For the Cv-ωTA a reaction which was virtually irreversible was selected; the half transamination reaction of 1-phenylethylamine to form acetophenone (Figure 8). The hypothesis, after studying the proposed reaction mechanism (see section 1.5, Figure 6), was that the Enz-PLP complex will be converted to Enz:PMP by an amino donor, but not be regenerated to a significant amount, if the formed amino acceptor was a poor substrate or thermodynamically favoured. When employing 1-phenylethylamine as the sole substrate (without an amino acceptor) the formation of acetophenone was observed together with the consumption of Enz-PLP. An excess of

18

H2O

NH2

PLP

O

PMP

Figure 8: The half transamination reaction of 1‐phenylethylamine.

Figure 9: The amount of consumed (S)‐1‐phenylethylamine ((S)‐1‐PEA) versus the total enzyme amount in half transamination reactions with Cv‐ωTA.

1-phenylethylamine was used with different enzyme amounts. The amount of consumed substrate could be plotted versus the enzyme amount to produce a linear relationship which coincided with origo (Figure 9). The reaction includes the consumption of a water molecule which, as the reaction is performed in aqueous buffer, will greatly shift the equilibrium towards products.

According to Figure 9 the molar amount of active sites can be determined by the consumption of consumed substrate. The active fraction can be deduced from the slope of the linear regression, in this case 81%. The consumed amount of (S)-1-phenylethylamine was quantified by HPLC. Since the enantiospecificity for this substrate was high, E=150 (S), the racemic substrate was used and the peak corresponding to the (R)-enantiomer was used as internal standard.

The observation that the Enz-PLP complex was spectrophoto-metrically detectable at 395 nm (yellow colour), while the Enz:PMP complex was not, allowed further simplification of the measurement method. By correlating the absorbance at 395 nm, where a

y = 0.81x + 0.00005R² = 0.99

0

0.005

0.01

0.015

0.02

0.025

0.03

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035Consum

ed (S)‐1‐PEA (m

M)

[Enz] mM

19

Figure 10: The absorbance of the holo form of Cv‐ωTA at 395 nm versus the amount of consumed (S)‐1‐PEA in a half transamination reaction.

maximum was found, to the amount of consumed substrate a direct measurement method was made. The method requires that all active enzyme molecules are in Enz-PLP form when the absorbance is recorded, and the assumption that all Enz-PLP is active enzyme. Figure 10 shows the absorbance at 395 nm versus the amount of consumed substrate, the regressed line can be said to pass through origo. The graph shows that the active enzyme amount can be directly quantified by absorbance with an apparent extinction coefficient of 8.1 mM-1cm-1, with given assumptions.

The enzyme was incubated with an excess of PLP after purification (by IMAC) and subsequent desalting to ensure that the enzyme was in the holo form. To rule out the putative presence of Enz:PMP emanating from the cultivation an effective amino acceptor such as pyruvate could be added to the incubation with PLP. This would purge the presence of PMP, leaving only Enz-PLP. In this case the enzyme was treated with a high excess of PLP which also binds covalently, the presence of Enz:PMP is therefore unlikely. Still, if present the Enz:PMP may be active enzyme which is not detected in the method described here. Therefore the method should include the use of pyruvate. Although the method was not tested with homologous enzymes it is seemingly feasible for all ω-transaminases, given that the reaction is virtually irreversible in all cases and an equal amount of PLP to enzyme can be ensured.

y = 8.1x + 0.0021R² = 0.99

0

0.05

0.1

0.15

0.2

0.25

0 0.005 0.01 0.015 0.02 0.025 0.03

Abs 395

 nm 

Consumed (S)‐1‐PEA (mM) 

20

During the assay design a few noteworthy characteristics of the enzyme were found. Phosphate buffer is not a good choice since its presence results in poor activity and instability of the enzyme. Phosphate may compete with PLP, which binds to the enzyme with a phosphate group, for the same binding site. It was also found that excess PLP was detrimental to the initial rate, i.e. the presence of more PLP than enzyme was unfavourable (a graph is presented in section 5.2, Figure 16). A similar effect was found with an ω-transaminase from Arthrobacter citreus (see section 4).

When incubated with PLP a time dependent activation was observed, the initial reaction rate increased by 129% without additional addition of PLP. This effect may be a result of dimerisation for formation of active enzyme, if the IMAC-purification resulted in monomerisation upon binding to the used cobalt(II) containing resin. Or, and perhaps more likely, the result is a consequence of structural rearrangements, described in section 6.

21

4 Equilibrium Displacement of ω-Trans-aminase Reactions with Isopropylamine as Amino Donor (Paper II) Since the ω-transaminases catalyze an equilibrium reaction complete conversion is inherently impossible without equilibrium displacement. Virtually complete reactions can be achieved in kinetic resolution reactions with an excess of amino acceptor; or a catalytic amount of pyruvate as amino acceptor if combined with an amino acid oxidase,66 thus regenerating the pyruvate from formed alanine. Stereoselective synthesis reactions have generally proven to reach meagre conversions although excess amino donor is used, and are therefore subject to reaction engineering. Theoretically it is possible to reach a hundred percent yield with equilibrium displacement. Enzymatic cascade reactions have been published where alanine was the amino donor; the formed pyruvate was further reacted by lactate dehydrogenase48,71,72,81,82 or pyruvate decarboxylase.70 When using isopropylamine as amino donor the equilibrium can be displaced by distillation of the formed acetone,71 but this method is not practical in large scale as mentioned in section 1.5. Employment of two-phase,61,81,83 or even three-phase (solvent bridge84) systems have proven useful for product removal, thus limiting product inhibition; as well as use of ion-exchange resin,85 and continuous extraction procedures.63

As a model substance 1-Phenylethylamine was chosen, being a small chiral aromatic primary amine. The prochiral ketone acetophenone was therefore used to produce (S)-1-phenyl-ethylamine; thereafter chosen derivatives thereof were produced from their corresponding amines. Isopropylamine was the chosen amino donor, a cost efficient industrial chemical.

An alternative method from distillation for removing acetone in a transamination reaction in synthesis mode with isopropylamine as donor was explored; the acetone was converted to isopropanol with the enzyme yeast alcohol dehydrogenase (Saccharomyces cerevisiae, YADH). This alcohol dehydrogenase is selective for small alcohols and ketones.86 The amine acceptor is most often a larger ketone, hence unaffected by the presence of YADH. This is an example of

22

NH2O O

OH O-HO

C OONH2

YADH FDH

ω-TA

NAD+

NADH,H+

H+

Figure 11: A transamination reaction with isopropylamine as donor and acetophenone as acceptor, catalysed by an ω‐transaminase. The equilibrium is displaced by selectively removing the formed acetone with YADH and regenerating the cofactor NADH with FDH.

utilisation of enzyme selectivity for practical use in an enzymatic cascade reaction. The cosubstrate NADH was regenerated with Candida boidinii formate dehydrogenase (FDH), the production of carbon dioxide effectively drives the reaction to completion. Figure 11 shows the used reaction scheme.

The gene for an Arthrobacter citreus ω-transaminase variant69,87 was cloned and successfully overexpressed in E. coli BL21(DE3), and purified by IMAC. The influence of excess concentration of the coenzyme PLP was examined; it was found that exceeding the concentration beyond that of the enzyme was detrimental to the reaction. Figure 12 shows the initial reaction rate versus the excess concentration of PLP, it is clear that the rate is greatly decreased with increasing amount of coenzyme. Since the enzyme concentration is usually very low, in the micromolar range, this effect is of importance. Already at millimolar concentrations PLP will practically completely discontinue the reaction.

Other examples of inhibition by PLP have been published, e.g. of RuBisCO,88 and of thymidylate synthase89 where a putative thio-hemiacetal was formed by PLP and a cysteine residue. Although these are related examples it is unlikely to be a similar mechanism of inhibition as for the ω-transaminase here. Arguably, enzymes with a lysine residues in the active site will be inhibited by PLP to some extent if a Schiff base with PLP can be formed, regardless if they are PLP-dependent or not. The mechanism for the inhibition of Cv-ωTA observed here remains to be investigated.

23

Figure 12: Activity versus excess PLP, more PLP than enzyme results in lowering of the initial rate.

A balanced ratio of PLP to enzyme was obtained by dissolving an excess of PLP in the purified enzyme solution and then changing the buffer. Assuming that no enzyme dissociates from the PLP during this process this provides an enzyme ready for use with the correct amount of coenzyme. No coenzyme was thereafter added in the reactions.

Acetophenone is a poor substrate (i.e. has a comparatively low specificity constant) but is advantageous since it is spectrophoto-metrically detectable while the corresponding amine shows a significantly lower extinction coefficient.90 A continuous spectro-photometric measurement verified the equilibrium displacement system. Figure 13 shows a comparison of the displaced versus the non-displaced reaction as the consumption of acetophenone over time. The displaced reaction continued towards completion while the non-displaced virtually stopped. Here the initial concentration of acetophenone was low, only 1.8 mM. This concentration was chosen to keep the starting absorbance at a spectrophotometrically measurable level and to have a practical conversion point for comparison. Even with a large excess of amino donor, 564 mM of isopropylamine was used, the conversion (without displacement) only reached 82%. The amount of formed product, (S)-1-phenylethylamine, and the enantiomeric excess was confirmed by HPLC. Indeed, the displaced reaction had virtually reached completion and the product amine was enantiomerically pure.

0

0.2

0.4

0.6

0.8

1

0 100 200 300 400 500

Rate (U

/mg)

(excess [PLP])/[Enz] 

24

Figure 13: Comparison of a displaced and a non‐displaced transamination reaction, followed by continuous absorbance measurement of the consumption of acetophenone.

Table 1: Results from stereoselective synthesis reactions performed witha  variant  of  A.  citreus  ω‐transaminase  with  isopropylamine  as  aminodonor (564 mM), with and without the displacement system. 

Product amine  ee (%) Conversion (%)  no displacement 

Conversion (%)  with displacement  

NH2

>99.9  82 a  >99 a 

NH2

O2N>99.9  63 b  >99 b 

NH2

>99.9  89 c  >99 c 

NH2

HO OH >99.9  68 c  >99 c 

a 1.8 mM acetophenone. Conversions were calculated based on formed product. b 20 mM 4‐nitroacetophenone (two phase system as saturation is reached at ~2 mM). Conversions were calculated based on formed product.  c 20 mM ketone substrate. Conversions were calculated based on consumed substrate. 

The displacement system was then explored for production of derivatives of (S)-1-phenylethylamine. Table 1 summarises the results; with 20 mM substrate concentration the displaced reactions also reached completion. No workup was performed, conversions were measured by HPLC.

0

1

2

0 6 12 18

[acetoph

enon

e] mM

Time (h)

Equilibrium reactionDisplaced reaction

25

Enz-PLP

NH2

Enz NH

Enz PMP

O

O

Enz NHPLPNH2

H2O

H2O

PLP

Figure 14: Schematic representation of the ω‐transaminase catalysed reaction as half transamination reactions, of isopropylamine/acetone (top) and acetophenone/(S)‐1‐phenylethylamine (bottom). 

Higher product concentrations (20 mM was achieved here) were not feasible. It was later observed that the specificity constant for the chiral amino product was many times higher than for isopropylamine, similar to published results with homologous enzymes48,62,81 and the results with Cv-ωTA in section 5.2. It is therefore unlikely that the product concentration limitation is a result of an unfavourable equilibrium, i.e. a thermodynamic effect, since displacement was in fact used. Rather, it is more likely a kinetic effect due to the specificity of the enzyme; although the thermodynamic equilibrium distribution was not measured.

Figure 14 shows a simplified scheme of the reaction mechanism with the half transamination reactions separated. If the specificity constant for the formed product, (S)-1-phenylethylamine, is higher than for the amino donor, isopropylamine, the enzyme will practically seize to catalyse the half reaction of isopropylamine as the concentration of (S)-1-phenylethylamine increases. The virtual discontinuation of the reaction is a result of a competing substrate situation arising from the formation of the chiral product which can function as an amino donor, in which case the isopropylamine will be disfavoured. At a given point an excess of isopropylamine will be insufficient for upholding continued production of the product amine due to the competing substrate situation, which will be observed as the termination of product formation. The bottom half of

26

the reaction scheme in Figure 14 may still be catalysed but will not be detected. Displacement of the equilibrium alone will not resolve this issue. Higher product amounts may be achieved by continuous removal thereof, which would enable the enzyme to react with isopropylamine as amino donor.

27

5 Rational Redesign of Cv-ωTA (Paper III and IV) Due to the lack of crystal structures protein engineering of ω-transaminases has to the most part been based on directed evolution approaches, where the activity, stability and enantio-selectivity has been targeted.64,69,74,91

In a rational redesign approach, however, a tryptophane residue was targeted in the ω-transaminase from Vibrio fluvialis.92 Kim and co-workers state that when changed to a glycine residue the enzyme displays higher activity and enantioselectivity for a number of aliphatic amines due to increased space in the active site. The same rational was applied for Caulobacter crescentus ω-transaminase93. In these cases homology models were applied.

Hughes and co-workers also applied a homology model on an enzyme variant from Arthrobacter spp. where a semi-rational substrate walking approach yielded an enzyme suitable for synthesis of sitagliptin.60 This is an example of application of an ω-transaminase for the industrial production of a pharmaceutical.

(S)-enantiopreference appears to be most common in nature; still, a number of ω-transaminases with (R)-preference were found by Bornscheuer and co-workers through computational search methods.94 High yield stereoselective synthesis of (R)-amines have been published by Kroutil and co-workers,95 and by Bornscheuer and coworkers.96 Although rational redesign has been used to achieve a substrate dependent shift of enantiopreference from (S) to (R),97 it still is a challenging engineering task which may lead to increased understanding of the enzyme when successfully achieved.

Other PLP-enzymes have been extensively studied and a number of successful examples of rational redesign are published. E.g. alanine racemase from Geobacillus stearothermophilus, as described in section 1.5, accepted serine as a substrate when more space was made in the active site which enabled a catalytically productive conformation.47

28

N

OHPO

OO

O

NH

D259

O O

HN

OHPO

OO

O

NH

D259

O O

H

Pro-(S)

G'319

H'318

F'320T'321

F'88

A231

W60

Pro-(R)

G'319

H'318

F'320T'321

F'88

A231

W60

A B

Figure 15: Schematic representation of the active site, deduced by a homology structure, with planar quinonoid structures resulting from the predicted reaction with (S)‐1‐phenylethylamine in A and (R)‐1‐phenylethylamine in B.

The Cv-ωTA was found to have good enantioselectivity for e.g. 1-phenylethylamine; but for a variety of substrates the E-values were not high enough for biocatalytic application. The possibility of creating a general enzyme, i.e. good activity and high E-value for a vast number of substrates, is unlikely. Such an enzyme would need to be able to accommodate large substrates, i.e. large groups on either side of the amino/keto group, and still bind small substrates in only one defined conformation. This is unreasonable since smaller substrates would be able to bind in a catalytically productive manner in mirrored conformations in an enlarged active site which would result in poor enantiospecificity, as also mentioned by Kroutil and co-workers.98

Here the goal was to increase the enantiospecificity and activity for chosen substrates by rational redesign, and in doing so, finding key amino acid residues in the active site for future engineering efforts. Also, the possibility of shifting the enantiopreference was explored.

A crystal structure was not available. Instead a homology model was constructed in accordance with a previously published method.97 A transaminase from Silicibacter pomeroyi with available crystal structure with sequence similarity of 51% was used as a fold template for Cv-ωTA. The rational redesign was then based on this homology model, at a later stage a crystal structure was made which is described in section 6.

29

5.1 Improved enantioselectivity and reversed enantiomeric preference (Paper III) A schematic visualization of the active site is presented in Figure 15 with planar pro-(S) and pro-(R) quinonoid structures. The engi-neering approach was based on the assumption that binding of these planar quinonoid structures are essential for the resulting enantiospecificity, i.e. the difference of stability of the enzyme-planar quinonoid complex for the pro-(S) versus the pro-(R) quinonoid structure is a determinant for the difference in activation energy of the enantiomers in the transamination reaction.

Several mutants were made (unpublished results) based on the schematic picture in Figure 15, with the straightforward assumption that the stability of the binding of the quinonoid structures from reactions with (S)- or (R)-forms of unnatural substrates is to the most part dependent on space in the active site. Hypothetically, if the enantioselectivity is a result of unfavoured coordination of e.g. the pro-(R) quinonoid, thereby favoring the pro-(S), the enantioselectivity may be reversed if the amino acid residues are changed and more space is made for the pro-(R) and less for the pro-(S). In silico docking experiments were performed to guide the choice of mutation points. The structures in Figure 15 are the reaction intermediates formed after abstraction of the proton from the Schiff base. The probability for proton abstraction, which is plausibly a rate determining step, is in this model assumed to be proportional to the coordination for the resulting structures. This model is therefore seen as a simplified guide, not an attempt to predict the actual activation energies for reactions of the enantiomers.

One variant made according to this hypothesis was a success, with a substrate dependent shift of enantiopreference. Table 2 displays E-values for the chosen substrates in five chosen variants.

A variant with two amino acid residue substitutions, phenylalanine 88 to alanine and alanine 231 to phenylalanine, showed a switch of enantiopreference for 2-aminotetraline with good enantiospecificity. Practically, the variant consists of a replacement of a phenyl group. For 1-aminotetraline this variant also showed a shifted enantio-preference, but with a low E-value. For 1-phenylethylamine the

30

Table  2:  E‐values  for  chosen  substrates  in  Cv‐ωTA  wild  type  and variants, with specified enantiopreference.* 

 Enzyme 

wild type  150  (S)  6.7  (S)  3.9  (S) 

Trp60Ala  1.5  (S)  1.0   ‐  5.3  (S) 

Trp60Cys  340  (S)  60  (S)  59  (S) 

Phe88Ala  7.1  (S)  1.2  (R)  no reaction   ‐ 

Ala231Phe  15  (S)  6.0  (S)  1.0   ‐ 

Phe88Ala, Ala231Phe 

6.3  (S)  1.9  (R)  63  (R) 

*Reactions run with pyruvate (1 mM) as amino acceptor.

enantiopreference was not shifted, but a lowered E-value was observed. This substrate dependent shift of enantiopreference remains to be explored for other substrates. It is possible that (R)-selectivity will be observed for aromatic amines with one methylene or more to separate the amino moiety from the phenyl group, such as amphetamine and derivatives thereof. According to the hypothesis (described in section 1.5) of the enantioselectivity largely being dependent of a large and a small pocket in the active site, the enzyme would have higher enantioselectivity for amines with larger groups. This is consistent with the variant having higher enantioselectivity for 2-aminotetraline than 1-aminotetraline. Kinetic resolution or stereoselective synthesis of (S)- and (R)-2-aminotetraline is interesting from a pharmacological standpoint. The substances and their derivatives are central nervous system (CNS)-active drug compounds.99 1-phenylethylamine is here used as a model substrate since it is a small aromatic chiral amine. However, it may not be a satisfactory representative for the enantiospecificity; there is a possibility that the corresponding planar quinonoid structure displays rigidity for rotation of the phenyl ring due to the π-electron system.

To increase the enantiospecificity of the (S)-selective Cv-ωTA the strategy was, analogous to the rational for shifting enantio-preference, to favour coordination of the pro-(S) quinonoid. The tryptophan residue on position 60 was targeted since the homology model predicted this amino acid residue to significantly influence the space in the active site for the phenyl group from (S)-1-

NH2 NH2NH2

31

phenylethylamine in the quinonoid; if changed for a smaller residue the pro-(S) conformation should be more favoured than in the wild type. Contradictory to this basic hypothesis was the finding of the variant with an alanine residue on this position (Trp60Ala) to have lower enantioselectivity, as well as lower activity. The Trp60Ala was also an unstable enzyme which made experimental procedures complicated.

Instead, when a cysteine residue was introduced on position 60 a stable and effective enzyme was obtained. The choice of cystein originated from comparison with a literature variant found by directed evolution of Arthrobacter citreus ω-transaminase;69,87 this variant has a cysteine residue in the corresponding position which could be identified with the homology model. Introduction of cysteine residues is not otherwise a common choice since it is a reactive species.

As displayed in Table 2 the increase of the E-values for the tested substrates of the Trp60Cys variant are significant. The E-value of this variant for 1-phenylethylamine, 1-aminotetraline and 2-aminotetraline are high enough for use in kinetic resolution reactions, whereas the E-value for the wild type is only adequate for 1-phenylethylamine.

32

5.2 Increased specificity (Paper IV) The increased E-values of the variant Trp60Cys presented in section 5.1 was a result of increased specificity for the (S)-enantiomer, not a lowering for the (R). A variant with a lowered catalytic activity may have an improved E-value if the specificity for the slower reacting enantiomer is fractionally more decreased than the specificity for the faster. This is an undesired result which was not the case for the variant Trp60Cys. The variant showed increased reaction rates for the (S)-enantiomers compared to wild type.

To further explore the function of the variant specificity constants for (S)-1-phenylethylamine, isopropylamine and a number of 4’-substituted acetophenones were determined. The active site quantification method (section 3) was used and turnover numbers could be calculated. As described in section 1.3, the apparent (pseudo one-substrate) specificity constants can be regarded as the true constants in the case of an ordered ping-pong bi-bi reaction, which applies to the ω-transaminase reaction.

The pH optimum was measured to 7.0 for Trp60Cys and 8.3 for the wild type. A study of the coenzyme dependence showed that the wild type displayed lowered rates with excess PLP whereas the variant showed enhanced rates and was dependent on excess PLP to reach maximum rate, see Figure 16. The reactions with the variant and the wild type described below were performed at the found optimum conditions, i.e. not equal pH and coenzyme concentrations. HEPES buffer and 37 °C was used in both cases. As described in section 3, phosphate inhibited and destabilised the enzymes, presumably due to competition with the phosphate group of PLP.

Kinetic constants were experimentally determined for (S)-1-phenylethylamine by spectrophotometrically following the formation of acetophenone,90 and for isopropylamine by quantifying the amount of formed alanine after five minutes with ninhydrin. Isopropylamine does not produce a colour reaction with ninhydrin,100 whereas alanine is readily detectable. A fixed concentration of pyruvate (1.0 mM) as amino acceptor was used. The values are shown in Table 3. The specificity constant for (S)-1-phenylethylamine was increased 29-fold for the variant Trp60Cys,

33

Figure 16: The effect of excess PLP on the initial rate of Cv‐ωTA wild type (blue dots) and variant Trp60Cys (red diamonds). The wild type enzyme was inhibited while the variant did not reach maximum rate below a two‐fold excess of the coenzyme. Upon studying the continuation of the reaction, the variant required 3‐fold or more excess of coenzyme in order to maintain the observed initial rate.

Table 3: Kinetic constants for (S)‐1‐phenylethylamine and isopropylamine for Cv‐ωTA wild type and variant Trp60Cys.* 

NH2

NH2

 

M 

M  M   

(s‐1∙mM‐1)    (s‐1∙mM‐1)  (mM)  (s‐1) 

Cv‐ωTA wild type 

0.68  0.20±0.004  73  15±0.3 

Cv‐ωTA Trp60Cys 

20  0.024±0.002  190  4.6±0.4 

*Reactions performed with pyruvate (1 mM) as amino acceptor.

compared to the wild type. Isopropylamine, on the other hand, was a poorer substrate for the variant in terms of turnover number and affinity. The KM value for isopropylamine had increased more than two-fold and the turnover number was more than three times lower.

By allowing the chosen 4’-substituted acetophenones to compete for the enzyme, with isopropylamine as amino donor (50 mM for the wild type and 200 mM for the variant), relative specificity constants were obtained. Three ketones (0.2 mM each) were employed

0

20

40

60

80

100

0 2 4 6 8

activity

 (%)

(excess [PLP])/[Enz]

wild type

Trp60Cys

34

Table 4: Specificity constants for chosen 4’‐substituted acetophenones for Cv‐ωTA wild type and variant Trp60Cys.  

OX

 

X: M (s‐1∙mM‐1) 

Cv‐ωTA wild type  Cv‐ωTA Trp60Cys NO2  0.030  0.16 Cl  0.029  n.d.c Br  0.028  0.039 CN  0.026  0.10 F  0.020  n.d.c H  0.0082a  0.034a 

Me  0.0050  0.024 OH  0.0024  0.013 

MeO  n.d.b  0.011 aExperimentally determined values  from which the remaining values were calculated from their relative values, IPA (50 mM for wild type, 200 mM for Trp60Cys) was used as amino donor.   bNot determined, no observed reaction (<0.0001). cNot  determined,  background  reaction  higher  than  the  enzyme catalysed reaction.

simultaneously in each reaction; inter-comparison was done by common ketones to link the reactions. The rates were obtained by studying consumption over time by HPLC. This is an effective and precise way of obtaining the relative specificity constants. As stated in section 1.3, the ratio of rates for competing substrates simplifies to the ratio of specificity constants. Dilution errors for enzyme concentration and time measurement errors cancel out from the division of the equations. The rate for acetophenone was then spectrophotometrically determined90 and used for calculating the remaining rates from their relative values. In Table 4 the chosen 4’-substituted acetophenones are listed with their corresponding specificity constants, for the variant and the wild type enzyme.

An average 4-fold increase of the specificity constant for the 4’-substituted acetophenones was observed for the variant compared to the wild type. Overall, an electron withdrawing group as 4’-substituent increased the specificity constant while electron donating groups led to a decrease. This observation warranted a study of what effects governed the observed reactivity of the acetophenone derivatives, a parameterisation approach was chosen.

35

For reactions of benzoic acid, and derivatives thereof, the Swain-Lupton equation101-103 can often be used for finding the electronic effect of substituents on the reaction rate, in terms of the field and resonance contributions separately. The Swain-Lupton equation is a refinement of the Hammett equation,104 which does not distinguish between field and resonance effects; an empirical constant for a given substituent is derived based on the difference of pKa of substituted benzoic acids. The Swain-Lupton equation assumes field (and inductive) effects to be independent of resonance. By linearly combining the terms the equation gives:

(eq. 9)

where k is the rate constant for a substituted species, k0 is the rate constant for the unsubstituted species, f and r are reaction specific constants, and F and R are deduced values for each substituent which are shown in Table 5.

The specificity constants in Table 4 were treated as first-order rate constants for parameterization and least squares-fitting to the Swain-Lupton equation. In this model the ordered ping-pong bi-bi two-substrate enzyme reaction has been drastically simplified to a first order reaction; the enzyme and amino donor are treated as constants. For determination of the field and resonance effects alone, and for comparison of the variant to the wild type, the model is applicable. The reaction conditions differed for wild type and variant. The pH, coenzyme concentration and amino donor concentration were optimised for each enzyme. There is a possibility of the results being biased or condition specific, but the alternative would be to evaluate the field and resonance effects at suboptimal conditions. The comparison made above regarding the specificity constants in Table 4 are indeed done in absolute terms, but from a practical perspective. In the parameterisation comparisons of the fractional contribution of field and resonance for each enzyme are made, not the absolute values. The Swain-Lupton plots are shown in Figure 17 together with the fitted equations.

36

Figure  17:  Swain‐Lupton  plots  for  reactions  of  chosen  4’‐substituted acetophenones from the rate constants listed in Table 4, for Cv‐ωTA wild type in A  and  variant  Trp60Cys  in  B.  The  4’‐substitution  of  the  acetophenones  are shown  in  the graph. The values of  the  coefficients  f and  r were  calculated  to 0.90 and 0.50 for the wild type, 0.28 and 0.34 for Trp60Cys. The contribution of resonance is 36% in the wild type and 55% in the variant. 

y = 1.0001x ‐ 9E‐05R² = 0.999

‐0.6

‐0.4

‐0.2

0

0.2

0.4

0.6

0.8

‐1 ‐0.5 0 0.5 1 1.5

log (k/k

0)

Swain‐Lupton

A Cv‐ωTA  wild type

OH‐

Me‐

F‐

Cl‐

Br‐CN‐

NO2‐

y = 1.0002x ‐ 0.0014R² = 0.9839

‐0.6

‐0.4

‐0.2

0

0.2

0.4

0.6

0.8

‐1 ‐0.5 0 0.5 1

log (k/k

0)

Swain‐Lupton

MeO‐

OH‐

Me‐

Br‐

CN‐

NO2‐

B Cv‐ωTA  Trp60Cys

log 0.90 0.50

log 0.28 0.34

37

Table 5: Swain‐Lupton parameters101‐103

for chosen 4’‐substituents.   Field 

parameter (F) Resonance 

parameter (R) NO2  1.0  1.0 CN  0.9  0.71 F  0.74  ‐0.6 Br  0.72  ‐0.18 Cl  0.72  ‐0.24 H  0  0 

Me  ‐0.01  ‐0.41 OH  0.46  ‐1.89 

MeO  0.54  ‐1.68 

     

For the wild type the fitting was excellent except for the reactions of 4’-cyano- and 4’-nitro-acetophenone, in the variant the fitting was good over the full range of chosen reactions. This can be interpreted as the catalytic contribution by the variant being equal for all these substrates, the rate differences can be attributed to the activation or deactivation by the substrates alone. In the wild type the same can be said except for the cyano and nitro substituents for which the Swain-Lupton predicted reaction rates are faster than the experi-mental values. From this point of view the variant may be seen as a more general catalyst. The contribution of resonance effects are governing the specificity in the variant (55%), compared to the wild type (36%) where field and inductive effects appear to be dominant.

The variant is not expected to have altered electronic, coenzyme and pH dependence without a change in the catalytic mechanism; e.g. a shift of rate determining step. It is conceivable that the introduced cysteine residue, a reactive species, can function as a Lewis base and thereby significantly influence the reaction. The thiol group may provide a field effect which lessens the importance of the effect by the substituent. In cannot, however, provide a resonance contribution if is not bound to the substrate-PLP complex.

38

The reason for the lowered E-value of the Trp60Ala variant (section 5, Table 2) is possibly a result of destabilization; it is possible that the specificity constant for the (S)-enantiomer was lowered by a larger fraction than for the (R). The low expression yield is consistent with a destabilized enzyme, possibly the enzyme produced inclusion bodies. The Trp60Ala variant was not considered a subject for deeper investigation since the practical applications were limited and the low activity made testing complicated.

39

6 Structure of Cv-ωTA (Paper V) The discovery of Cv-ωTA as an ω-transaminase was done by sequence identification with the ω-transaminase from Vibrio fluvialis as template, by Ward and co-workers.105 PLP-dependent enzymes are classified in fold types (I-V) by their tertiary structure and into subgroups (I-IV) by their sequence identity.106-108 Cv-ωTA belongs to fold type I and subgroup III. Aspartate transaminase also belongs to fold type I, but subgroup I. Cv-ωTA is (S)-selective (L-amino acid) and the catalysed reaction is assumed to follow the same reaction steps as an aspartate transaminase;20 the sequence dissimilarity, however, indicates that the enzyme has different structure. Crystal structures of ω-transaminases may be a necessary basis for further exploration.

Although the rational redesign described in section 4 was performed with use of a homology model, a more complete structure of Cv-ωTA was desired. In the model, coordinated water molecules were not included and amino acid residue side chain rotational positions could not be assured. Additionally, the homology model could not serve as a base for a hypothesis about the experimental findings described in section 2; when incubated with PLP the enzyme displayed a time dependent activation. A crystal structure is the preferred base for further engineering, notwithstanding the success with the homology model.

There are examples of successful homology model based engineering,60,92,93 described in section 5. Still, it is conceivable that more precise structures of both apo and holo forms of an ω-transaminase may reveal influential functions for further understanding and development. There are available crystal structures of ω-transaminases; the apo enzyme from Vibrio fluvialis by Jang and co-workers (PDB ID: 3NUI) and one monomer of the holo enzyme from Psudomonas putida by Sasaki and co-workers (PDB ID: 3A8U). These structures are presented without a published structural analysis. Experimental details regarding the crystallisation of Cv-ωTA have been published by Littlechild and co-workers,109 the structure itself was published shortly after Paper V (PDB ID: 4AH3).

40

X-ray crystallographic structures of both apo and holo forms of Cv-ωTA were produced at resolutions between 1.35 and 2.4 Å, one of the holo form (PDB ID: 4A6U), two in the apo form (PDB IDs: 4A6R and 4A6T) and one in a mixed form (PDB ID: 4A72). Figure 18 shows the enzyme in α-helix/β-sheet/flexible loop representation.

The enzyme crystallised as a homodimer. Each monomer had a PLP-binding site at the subunit interface. The homodimer thereby has two active sites composed by elements from both subunits, similar to homologous enzymes.

The structures revealed significant differences between the holo and apo forms, also stated by Littlechild.109 This was interpreted as structural rearrangements upon coenzyme binding. With bound PLP the enzyme appears more ordered while the apo enzyme has flexible parts on one side of the active site. The observed time dependant activation may be a result of these rearrangements.

The holo enzyme crystal structure was rather similar to the homology model. Out of the 916 Cα-positions 896 could be aligned with an RMSD of 1.01 Å, with a seemingly identical active site apart from a few side-chain rotamer positions. The active site of the holo enzyme is shown in Figure 19.

The structural rearrangements upon coenzyme binding are of significance for the form of the active site. The parts coloured in yellow in Figure 19 are part of a structure which may be envisioned as a lid which partly covers the active site when PLP is bound. The reordered section consists mainly of residues 317-322 of the subunit which is not binding PLP. The threonine residue 321 and tyrosine residue 322 are repositioned in proximity to the internal aldimine complex and are likely to be of significance for the coenzyme and substrate binding. When repositioned, the coenzyme is coordinated by more hydrogen bonds; the phosphate group of PLP appears to be fixed and the otherwise flexible loop of residues 287-290 is positioned so that the PLP binding lysine residue 288 is projected into the active site. The described conformations of the apo and holo forms of the active site region are shown in Figure 20.

41

 

Figure 18: Overall fold of Cv‐ωTA in the holo form with the large domain coloured pink, the two small domains coloured light and dark blue and the linker containing an α‐helix in grey. The monomer is depicted in A, and the native dimer in B. (Courtesy of Assoc. prof. Derek Logan, Lund University)

A 

B 

42

Figure 19: Graphical representation of the active site of Cv‐ωTA in the holo form, composed of elements from both subunits of the homodimer distinguishable by grey and yellow colour. Amino acid residues within 4 Å of PLP are shown in stick representation together with the PLP‐internal aldimine complex inside a semitransparent grey surface. Hydrogen bonds are depicted as dotted lines, water molecules participating in hydrogen bonding with the aldimine are shown as red spheres. (Courtesy of Assoc. prof. Derek Logan, Lund University)

43

Figure 20: Graphical representations of the active site in the apo (A) and holo (B) form with the PLP‐binding subunit in light blue, the other subunit in beige and the rearranging structures in dark red. (Courtesy of Assoc. prof. Derek Logan, Lund University)

44

7 Concluding Discussion Evolution of enzymes has proven to be a success story for life, all living organisms can sustain their metabolic needs by enzyme catalysis of which the qualities remain to be fully utilised in chemical engineering. Many enzymes are presently being utilised in industry and academic research, but the current exploitation does not by any means reach the full potential.

When a chemist is presented with a task of synthesising a given chemical the available methods commonly suggest a classical approach where organic solvents and metal catalysts rule. If the chemist encounters a reaction step which cannot be accomplished with this approach, an enzyme as an alternative method might be sought after for this specific step in the process. Thereby the enzyme is hoped to cope with, or to be adapted to, the from an enzymologist’s perspective rough conditions. The fundamental strategy should be to devise an entire process based on biocatalysis, in aqueous conditions. This approach has been discussed as the development of cascade chemo-enzymatic processes or enzyme initiated domino reactions.40,110-112 This requires further method development and demonstrations of successful large scale reactions.

The choice of product may play a paramount role. A compound for a desired application may be selected within a range of suitable products for a complete biocatalytic synthesis method, assuming that not only one specific compound will suit the application. Today the choices may be based on available classical synthesis methods, which may not be beneficial in a long term perspective. Biocatalysis is currently in a sense only competing with the reactions which are not available or not satisfactory with other methods. In a life cycle analysis, a calculation of the total production cost of a chemical, where waist disposal and environmental impact is included a biocatalytic synthesis may be the most cost efficient;113 particularly if environmental damages are to a larger extent subject to economical recompense enforced by legislation.

45

7.1 Reactions and engineering with ω-transaminase, summarising discussion of Paper I-V The reactions catalyzed by ω-transaminases readily compete with classical methods. The enzyme can form a chiral amine from a prochiral ketone, the amination is in itself an advantage and the enantioselectivity has in many cases been shown to be high enough for production of virtually enantiopure compounds.58

The described active site quantification method (Paper I) provides a means for correct assessments and comparisons of engineered variants and homologues. The turnover number and finding optimal reaction conditions is many times overlooked, and arbitrary comparisons in given conditions with an unknown active enzyme fraction are then made. This cannot be considered a satisfactory procedure since it stands the risk of giving greatly misleading values. Active site quantification is also beneficial when employing an enzyme in a chemical process to find the expected catalytic efficiency. The method presented is a straightforward procedure; a simple spectrophotometric measurement will yield the amount of active sites. The measurement can be done on an enzyme batch before employment in a reaction, no enzyme is consumed. An alternative method for measuring the active sites, a titration procedure, may include the use of cycloserine which is a suicide inhibitor for many PLP-enzymes, or 1-deazapyridoxal-5′-phosphate which can create a stable external aldimine in an aspartate transaminase.79 A titration method may be employed to re-verify the obtained apparent extinction coefficient, which arguably provides an effective measurement procedure.

The use of alcohol dehydrogenase from yeast for displacing the equilibrium of a transamination reaction with isopropylamine as the amino donor (Paper II) was shown to be efficient, at relatively low substrate concentrations (20 mM ketone). YADH is used for selectively removing a compound in a cascade reaction, with cosubstrate regeneration. Enantiopure amines were produced in virtually complete reactions. Further development is needed to overcome the large difference of specificity which is many orders of magnitude lower for the amino donor (isopropylamine) than the desired product amine, to achieve a higher total product

46

concentration. The work is an example of the possibilities that biocatalysis can provide in terms of applications of existing knowledge. Development of computational search tools may enable finding of other more complex cascade reactions which otherwise would require extensive book mining.

Rational redesign, based on a homology structure, was used to produce ω-transaminase variants with shifted enantiopreference, higher enantiospecificity and improved specificity for chosen substrates than Cv-ωTA wild type (Paper III and IV). At a later stage a crystal structure was obtained which verified the accuracy of the used homology structure (Paper V). This successful example, together with other published works,60,92,93 shows that it is possible to perform rational redesign without a crystal structure on ω-transaminase by use of homology structures, although the examples may be circumstantial. The putative reaction mechanism was based on that of aspartic transaminase,20 another feature which was assumed to be similar to homologous enzymes. Further engineering on this enzyme group may benefit from continued exploration of the reaction mechanism; especially since the mechanism possibly is altered in the variant (Cv-ωTA Trp60Cys) with increased specificity constants. Such experiments may be of greater value than directed evolution approaches for further engineering by increasing the knowledge. The results in Paper IV indicate that the reaction rate is dependent of electronic effects which can be adapted to enhance activity. Further engineering efforts may include a model for alteration of the charge distribution of the active site. Reaction rates of poor substrates may be elevated by e.g. counteracting the deactivation shown by the electron donating 4’-substituents of acetophenone. If a less reactive amino acid residue can be introduced without destabilising the enzyme (Cv-ωTA Trp60Ala was not stable) Swain-Lupton parameterisation of its reactions may show if the enhanced specificity constants of Cv-ωTA Trp60Cys were due to the introduced cysteine residue, and to what extent the effect depends on increased space.

The variant Cv-ωTA Trp60Cys showed significantly increased activity for the aromatic substrates tested compared to the wild type, but significantly lower for isopropylamine. This may be a result of the increased space in the active site which may allow more degrees of

47

freedom for the smaller substrate and thereby disfavouring a catalytically productive conformation. This would imply that the specificity for smaller substrates is inherently lowered if the enzyme is able to accommodate larger substrates. This is contradictory to the published data on the variant of ω-transaminase from Vibrio fluvialis with the corresponding position changed for a glycine residue (on which the Cv-ωTA Trp60Cys variant was partly based).92 In this case, however, the activity was measured at the same pH for both wild type and variant, no active site quantification was performed and the coenzyme dependence was not explored. It is possible that the V. fluvialis enzyme variant has similar properties to Cv-ωTA Trp60Cys regarding pH optimum and PLP dependence, and the observed increase of activity for isopropylamine is only valid at the given reaction conditions.

The structural rearrangements upon cofactor binding revealed by crystal structures (Paper V) is a quality which was not elucidated by the homology modelling, which is likely to be of importance for further engineering. One might speculate that an unnatural substrate may induce an open form upon reaction with the holo enzyme, thus rendering the enzyme inactive. Time dependent activation of the enzyme was observed when incubated with the coenzyme. The results from the crystallization experiments indicate that the structural rearrangements are a result of coenzyme binding, or even hydrogen bonding to inorganic phosphate in the “phosphate group binding cup”.114 The rearrangements can be a slow process and may be then be observed as time dependent activation. This hypothesis is consistent with the observed deactivation of the enzyme by phosphate buffer (sulphate and borate as well). The mechanism behind the change in coenzyme dependence by the variant (Trp60Cys), which was not inhibited but activated by excess PLP, may be elucidated by a crystal structure, and could be important information for further engineering.

It is possible that the coenzyme binding is a sequential process in which PLP first forms a Schiff-base with the catalytic lysine residue, and then binds to the phosphate binding cup which induces a structural rearrangement which fixes the enzyme in the active form. Another possibility is that PLP first binds to the phosphate binding cup, which then induces a structural rearrangement which favours

48

formation of the internal aldimine. In either case, if the concentration of PLP is high this process may be inhibited by the presence of two PLP moieties, one bound to the lysine residue and another to the phosphate binding cup, thereby preventing the formation of the active form. If this is the case, the mode of coenzyme binding must be different in the variant which was not affected by excess PLP. The apo enzyme of the variant did not show a disulphide bridge with the introduced cystein residue, which is located adjacent to another cystein. This was observed in a preliminary crystallographic study.115 It is unlikely that a disulphide bond should be present in the holo form since the apo enzyme had previously been incubated with PLP, in which case the enzyme would have undergone oxidation subsequent to coenzyme dissociation.

7.2 Future prospects This section discusses ideas for possible continuation of the present investigation.

7.2.1 Dynamic kinetic resolution of an amine A lipase, such as Candida antarctica lipase B (CalB), can enantioselectively acylate an amine in an organic solvent.116 If an amine can be racemised effectively, combined with enantioselective acylation of either enantiomer, the reactions would result in an enantiopure amide. This has been shown by Bäckvall et al. for 1-phenylethylamine with racemisation by a ruthenium catalyst and acylation with CalB.117

Amine racemisation may possibly be achieved by using an ω-transaminase with poor E-value, or combining two enzymes with opposite enantiopreference, and a catalytic amount of amino acceptor (e.g. pyruvate). The dynamic kinetic resolution reaction may require a two phase-system where the ω-transaminase is dissolved in an aqueous phase and the immobilised CalB is situated in an organic solvent phase; this would require adequate mass transfer to function. Another approach is to attempt the complete reaction in aqueous medium; CalB, and a variant thereof, has been shown to perform acylation in buffer.118

49

There is also the option of testing the ω-transaminase for activity in organic solvents, after immobilisation. In related works we immobilised enzymes on modified silica oxide to achieve condition promiscuity. The His6-tag was used for binding to metal ions on a chelating surface.119,120 Initial tests with modified controlled pore glass in place of the silica oxide were performed with ω-transaminase, activity was found in toluene, but the system remains to be explored and optimised. Active ω-transaminase in organic solvent will presumably display altered substrate specificity, and is likely to provide possibilities for alternative solutions for equilibrium displacement, although it arguably is a less environmentally sustainable approach compared to aqueous systems.

7.2.2 Mechanistic studies The reaction mechanism of the ω-transaminases is assumed to follow the same pattern as proposed for aspartic α-transaminases.46 In the proposed reaction mechanism (section 1.5, Figure 6) a proton is abstracted to form a planar quinonoid structure. When an α-amino acid forms a Schiff-base with PLP the pKa is lowered for the proton to be abstracted for formation of the quinonoid intermediate, the presumed rate determining step. When the substrate is an aliphatic amine this effect is less significant, the conjugated base is stabilised by resonance in the pyridoxal moiety but not by a carboxylate. Preliminary quantum mechanical modelling predicted the pKa to be 14 units higher for an isopropylamine-PLP complex compared to the alanine-PLP complex when modelled in Geobacillus stearothermophilus alanine racemase.121 It is still unclear as to why the ω-transaminase is able to abstract this proton, notwithstanding having a relatively low turnover number. The charge distribution in the active site of the ω-transaminases may lower the barrier in question; this effect is possibly enhanced in Cv-ωTA Trp60Cys. Crystallization experiments with 1-deazapyridoxal-5'-phosphate to form a stabile external aldimine with chosen substrates, as demonstrated by Toney et al. for an aspartate transaminase,79 may be a suitable extension of this study for mapping the structure and substrate interactions of the wild type and variant Trp60Cys. A comparative study with a homologous amino acid racemase may suggest feasible mutation points for creating an amino racemase with a suitable PLP-dependent enzyme as template.

50

8 Acknowledgements To my supervisor and mentor, Prof. Per Berglund, thank you for half a decade (at least) of great teaching! The same goes for Prof. Karl Hult, my enzyme guru, who never turned down an opportunity for discussion. I will miss stumbling into your offices with questions out of the blue.

To all my colleagues; office mates and co-workers, students and teachers, co-authors and collaborators, I thank you for help and friendship inside and outside of the lab. I don not think I have seen the last of you...

To my family and extended family, thank you for all your help and support! You made it possible to work at those so important odd hours before a deadline. To all my friends (not least my band mates), thanks for much needed distractions from science.

To my life companion Julia and my daughter Nadia, thank you for encouragement, understanding and being more important to me than anything else in the world!

Financial support from The Swedish Governmental Agency for Innovation Systems (VINNOVA) is gratefully acknowledged. EU-COST Action CM0701 is acknowledged for travel funds and for the opportunity to join the COST Training School in Cascade Chemoenzymatic Processes – New Synergies between Chemistry and Biochemistry - in Italy 2009. KTH is acknowledged for travelling scholarships in 2006, 2007 and 2008 that enabled conference participation.

51

9 References 1.  P. Grunwald, Biocatalysis: Biochemical Fundamentals and 

Applications, Imperial College Press, 2009. 2.  C. Burel, Industrial or White Biotechnology ‐ Research for Europe, 

EuropaBio, ESAB. 3.  D. Carrez and W. Soetaert, Industrial or White Biotechnology ‐ A 

driver of sustainable growth in Europe, EuropaBio, ESAB. 4.  D. J. Mack, M. Brichacek, A. Plichta and J. T. Njarðarson, 

http://cbc.arizona.edu/njardarson/group/sites/default/files/Top200PharmaceuticalProductsByWorldwideSalesin2009.pdf, 2009. 

5.  D. Voet, J. G. Voet and C. W. Pratt, Fundamentals of Biochemistry: Life at the Molecular Level, 2nd edn., Wiley, 2006. 

6.  T. Golson and B. Fink, The Farmstead Egg Cookbook, St. Martin's Press, 2006. 

7.  R. Chang, Physical Chemistry for the Chemical and Biological Sciences, 3rd edn., University Science Books, 2000. 

8.  T. C. Bruice, Chem Rev, 2006, 106, 3119‐3139. 9.  A. Warshel, P. K. Sharma, M. Kato, Y. Xiang, H. Liu and M. H. 

Olsson, Chem Rev, 2006, 106, 3210‐3235. 10.  G. G. Hammes, S. J. Benkovic and S. Hammes‐Schiffer, 

Biochemistry, 2011, 50, 10422‐10430. 11.  M. Nic, J. Jirat and B. Kosata, International Union of Pure and 

Applied Chemistry, 2011, http://goldbook.iupac.org/PDF/goldbook.pdf. 

12.  H. Bisswanger, Enzyme Kinetics: Principles and Methods, Wiley‐VCH, 2002. 

13.  R. A. Copeland, Enzymes: A Practical Introduction to Structure, Mechanism and Data Analysis, 2nd edn., Wiley‐VCH, 2000. 

14.  M. Jones, Organic Chemistry, 2nd edn., W. W. Norton & Company, Inc., New York, 2000. 

15.  V. Henri, 1903, Hermann, Paris. 16.  L. Michaelis and M. L. Menten, Biochem Z, 1913, 49, 333‐369. 17.  L. Michaelis, M. L. Menten, K. A. Johnson and R. S. Goody, 

Biochemistry, 2011, 50, 8264‐8269. 18.  G. P. Moss, 

http://www.chem.qmul.ac.uk/iubmb/kinetics/index.html, 1981, Symbolism and Terminology in Enzyme Kinetics, Nomenclature Committee of the International Union of Biochemistry (NC‐IUB). 

52

19.  G. E. Briggs and J. B. Haldane, Biochem J, 1925, 19, 338‐339. 20.  G. G. Hammes and J. L. Haslam, Biochemistry, 1969, 8, 1591‐1598. 21.  C. S. Chen, Y. Fujimoto, G. Girdaukas and C. J. Sih, J Am Chem Soc, 

1982, 104, 7294‐7299. 22.  C. S. Chen, S. H. Wu, G. Girdaukas and C. J. Sih, J Am Chem Soc, 

1987, 109, 2812‐2817. 23.  N. Ran, E. Rui, J. Liu and J. Tao, Curr Pharm Des, 2009, 15, 134‐152. 24.  Y. S. Ding and J. S. Fowler, Drug Dev Res, 2003, 59, 227‐239. 25.  C. Khosla and P. B. Harbury, Nature, 2001, 409, 247‐252. 26.  F. H. Arnold, Nature, 2001, 409, 253‐257. 27.  A. Schmid, J. S. Dordick, B. Hauer, A. Kiener, M. Wubbolts and B. 

Witholt, Nature, 2001, 409, 258‐268. 28.  S. Lutz, Curr Opin Biotechnol, 21, 734‐743. 29.  W. P. C. Stemmer, Proc Natl Acad Sci U S A, 1994, 91, 10747‐10751. 30.  A. Koltermann and U. Kettling, Biophys Chem, 1997, 66, 159‐177. 31.  F. H. Arnold, Acc Chem Res, 1998, 31, 125‐131. 32.  N. J. Turner, Trends Biotechnol, 2003, 21, 474‐478. 33.  F. H. Arnold and A. A. Volkov, Curr Opin Chem Biol, 1999, 3, 54‐59. 34.  M. T. Reetz, L. W. Wang and M. Bocola, Angew Chem Int Ed Engl, 

2006, 45, 1236‐1241. 35.  M. T. Reetz, M. Bocola, J. D. Carballeira, D. Zha and A. Vogel, 

Angew Chem Int Ed Engl, 2005, 44, 4192‐4196. 36.  R. J. Kazlauskas and U. T. Bornscheuer, Nat Chem Biol, 2009, 5, 

526‐529. 37.  M. S. Humble and P. Berglund, Eur J Org Chem, 2011, 3391‐3401. 38.  K. Hult and P. Berglund, Trends Biotechnol, 2007, 25, 231‐238. 39.  U. T. Bornscheuer and R. J. Kazlauskas, Angew Chem Int Ed Engl, 

2004, 43, 6032‐6040. 40.  Q. Wu, B.‐K. Liu and X.‐F. Lin, Curr Org Chem, 2010, 14, 1966‐1988. 41.  N. Tokuriki and D. S. Tawfik, Science, 2009, 324, 203‐207. 42.  C. Branneby, Doctoral Thesis, Royal Institute of Technology, 2005. 43.  M. D. Toney, Arch Biochem Biophys, 2005, 433, 279‐287. 44.  M. D. Toney, Biochim Biophys Acta, 2011, 1814, 1407‐1418. 45.  E. F. Oliveira, N. M. Cerqueira, P. A. Fernandes and M. J. Ramos, J 

Am Chem Soc, 2011, 133, 15496‐15505. 46.  R. B. Silverman, The Organic Chemistry of Enzyme‐Catalysed 

Reactions, Academic Press, Elsevier Science, 2000. 47.  W. M. Patrick, J. Weisner and J. M. Blackburn, ChemBioChem, 

2002, 3, 789‐792. 

53

48.  J. S. Shin and B. G. Kim, Biotechnol Bioeng, 1999, 65, 206‐211. 49.  J. S. Shin and B. G. Kim, J Org Chem, 2002, 67, 2848‐2853. 50.  J. Ward and R. Wohlgemuth, Curr Org Chem, 2010, 14, 1914‐1927. 51.  B. M. Nestl, B. A. Nebel and B. Hauer, Curr Opin Chem Biol, 2011, 

15, 187‐193. 52.  T. C. Nugent and M. El‐Shazly, Adv Synth Catal, 2010, 352, 753‐819. 53.  M. Höhne and U. T. Bornscheuer, ChemCatChem, 2009, 1, 42‐51. 54.  D. Koszelewski, D. Pressnitz, D. Clay and W. Kroutil, Org Lett, 2009, 

11, 4810‐4812. 55.  D. Koszelewski, I. Lavandera, D. Clay, G. M. Guebitz, D. Rozzell and 

W. Kroutil, Angew Chem Int Ed Engl, 2008, 47, 9337‐9340. 56.  D. Koszelewski, M. Göritzer, D. Clay, B. Seisser and W. Kroutil, 

ChemCatChem, 2010, 2, 73‐77. 57.  D. Koszelewski, K. Tauber, K. Faber and W. Kroutil, Trends 

Biotechnol, 2010, 28, 324‐332. 58.  K. O. Lidström, PharmaChem, 2007, 6, 5‐7. 59.  J. D. Stewart, Curr Opin Chem Biol, 2001, 5, 120‐129. 60.  C. K. Savile, J. M. Janey, E. C. Mundorff, J. C. Moore, S. Tam, W. R. 

Jarvis, J. C. Colbeck, A. Krebber, F. J. Fleitz, J. Brands, P. N. Devine, G. W. Huisman and G. J. Hughes, Science, 2010, 329, 305‐309. 

61.  J. S. Shin and B. G. Kim, Biotechnol Bioeng, 1998, 60, 534‐540. 62.  J. S. Shin and B. G. Kim, Biosci Biotechnol Biochem, 2001, 65, 1782‐

1788. 63.  J. S. Shin, B. G. Kim, A. Liese and C. Wandrey, Biotechnol Bioeng, 

2001, 73, 179‐187. 64.  H. Yun, B. Y. Hwang, J. H. Lee and B. G. Kim, Appl Environ Microbiol, 

2005, 71, 4220‐4224. 65.  H. Yun, B. K. Cho and B. G. Kim, Biotechnol Bioeng, 2004, 87, 772‐

778. 66.  M. D. Truppo, N. J. Turner and J. D. Rozzell, Chem Commun (Camb), 

2009, 2127‐2129. 67.  P. Tufvesson, J. Lima‐Ramos, J. S. Jensen, N. Al‐Haque, W. Neto and 

J. M. Woodley, Biotechnol Bioeng, 2011, 108, 1479‐1493. 68.  B. Y. Hwang and B. G. Kim, Enzyme Microb Technol, 2004, 34, 429‐

436. 69.  A. R. Martin, R. DiSanto, I. Plotnikov, S. Kamat, D. Shonnard and S. 

Pannuri, Biochem Eng J, 2007, 37, 246‐255. 70.  M. Höhne, S. Kuhl, K. Robins and U. T. Bornscheuer, ChemBioChem, 

2008, 9, 363‐365. 

54

71.  M. D. Truppo, J. D. Rozzell, J. C. Moore and N. J. Turner, Org Biomol Chem, 2009, 7, 395‐398. 

72.  D. Koszelewski, I. Lavandera, D. Clay, D. Rozzell and W. Kroutil, Adv Synth Catal, 2008, 350, 2761‐2766. 

73.  P. P. Taylor, D. P. Pantaleone, R. F. Senkpeil and I. G. Fotheringham, Trends Biotechnol, 1998, 16, 412‐418. 

74.  G. Matcham, M. Bhatia, W. Lang, C. Lewis, R. Nelson, A. Wang and W. Wu, Chimia, 1999, 53, 584‐589. 

75.  W. Wu, M. Bhatia, C. M. Lewis, W. Lang, A. L. Wang and G. W. Matcham, US Pat., 1999, 6 133 018. 

76.  G. R. Schonbaum, B. Zerner and M. L. Bender, J Biol Chem, 1961, 236, 2930‐2935. 

77.  T. S. Soper and J. M. Manning, J Biol Chem, 1981, 256, 4263‐4268. 78.  D. Peisach, D. M. Chipman, P. W. Van Ophem, J. M. Manning and D. 

Ringe, J Am Chem Soc, 1998, 120, 2268‐2274. 79.  W. R. Griswold, A. J. Fisher and M. D. Toney, Biochemistry, 2011, 

50, 5918‐5924. 80.  U. Schell, R. Wohlgemuth and J. M. Ward, J Mol Catal B Enzym, 

2009, 59, 279‐285. 81.  J. H. Seo, D. Kyung, K. Joo, J. Lee and B. G. Kim, Biotechnol Bioeng, 

2011, 108, 253‐263. 82.  D. Koszelewski, D. Clay, D. Rozzell and W. Kroutil, Eur J Org Chem, 

2009, 2009, 2289‐2292. 83.  B. K. Cho, H. J. Cho, S. H. Park, H. Yun and B. G. Kim, Biotechnol 

Bioeng, 2003, 81, 783‐789. 84.  H. Yun and B. G. Kim, Biosci Biotechnol Biochem, 2008, 72, 3030‐

3033. 85.  M. D. Truppo, J. D. Rozzell and N. J. Turner, Org Process Res Dev, 

2010, 14, 234‐237. 86.  V. Leskovac, S. Trivic and D. Pericin, FEMS Yeast Res, 2002, 2, 481‐

494. 87.  CambrexNorthBrunswickInc, S. Pannuri, S. Kamat and A. R. M. 

Garcia, World Pat, 2006, WO 2006/063336 A063332. 88.  W. Whitman and F. R. Tabita, Biochem Biophys Res Commun, 1976, 

71, 1034‐1039. 89.  S. C. Chen, H. H. Daron and J. L. Aull, Int J Biochem, 1989, 21, 1217‐

1221. 90.  S. Schätzle, M. Höhne, E. Redestad, K. Robins and U. T. 

Bornscheuer, Anal Chem, 2009, 81, 8244‐8248. 

55

91.  A. R. Martin, D. Shonnard, S. Pannuri and S. Kamat, Appl Microbiol Biotechnol, 2007, 76, 843‐851. 

92.  B. K. Cho, H. Y. Park, J. H. Seo, J. Kim, T. J. Kang, B. S. Lee and B. G. Kim, Biotechnol Bioeng, 2008, 99, 275‐284. 

93.  B. Y. Hwang, S. H. Ko, H. Y. Park, J. H. Seo, B. S. Lee and B. G. Kim, J Microbiol Biotechnol, 2008, 18, 48‐54. 

94.  M. Höhne, S. Schätzle, H. Jochens, K. Robins and U. T. Bornscheuer, Nat Chem Biol, 2010, 6, 807‐813. 

95.  F. G. Mutti, C. S. Fuchs, D. Pressnitz, J. H. Sattler and W. Kroutil, Adv Synth Catal, 2011, 353, 3227‐3233. 

96.  S. Schätzle, F. Steffen‐Munsberg, A. Thontowi, M. Höhne, K. Robins and U. T. Bornscheuer, Adv Synth Catal, 2011, 353, 2439‐2445. 

97.  M. Svedendahl, C. Branneby, L. Lindberg and P. Berglund, ChemCatChem, 2010, 2, 976‐980. 

98.  F. G. Mutti, J. Sattler, K. Tauber and W. Kroutil, ChemCatChem, 2011, 3, 109‐111. 

99.  H.‐J. Federsel, M. Hedberg, F. R. Qvarnström, M. P. T. Sjögren and W. Tian, Acc Chem Res, 2007, 40, 1377–1384. 

100.  V. J. Harding and R. M. MacLean, J Biol Chem, 1916, 25, 337‐350. 101.  C. G. Swain, J Org Chem, 1984, 49, 2005‐2010. 102.  C. G. Swain and E. C. Lupton, J Am Chem Soc, 1968, 90, 4328‐4337. 103.  C. G. Swain, S. H. Unger, N. R. Rosenquist and M. S. Swain, J Am 

Chem Soc, 1983, 105, 492‐502. 104.  L. P. Hammett, J Am Chem Soc, 1937, 59, 96‐103. 105.  U. Kaulmann, K. Smithies, M. E. B. Smith, H. C. Hailes and J. M. 

Ward, Enzyme Microb Technol, 2007, 41, 628‐637. 106.  P. K. Mehta, T. I. Hale and P. Christen, Eur J Biochem, 1993, 214, 

549‐561. 107.  J. N. Jansonius, Curr Opin Struct Biol, 1998, 8, 759‐769. 108.  P. Berglund, M. Svedendahl Humble and C. Branneby, in 

Comprehensive Chirality, eds. H. Yamamoto, E. Carreira and N. J. Turner, Elsevier, Oxford, UK, 2012, vol. 7, in press. 

109.  C. Sayer, M. N. Isupov and J. A. Littlechild, Acta Crystallogr Sect F Struct Biol Cryst Commun, 2007, 63, 117‐119. 

110.  S. F. Mayer, W. Kroutil and F. Kurt, Chem Soc Rev, 2001, 30, 332‐339. 

111.  R. A. Sheldon and F. van Rantwijk, Aust J Chem, 2004, 57, 281‐289. 112.  COST European Cooperation in the field of Scientific and Technical 

Research, 2012, CASCAT COST ACTION CM0701 Cascade Chemo‐

56

Enzymatic Processes: New Synergies between Chemistry and Biochemistry, http://www.cost‐cascat.polimi.it/. 

113.  O. Thum and K. M. Oxenbøll, SOFW J, 2008, 134. 114.  A. I. Denesyuk, K. A. Denessiouk, T. Korpela and M. S. Johnson, J 

Mol Biol, 2002, 316, 155‐172. 115.  D. Logan, Personal communication, 2011. 116.  F. van Rantwijk and R. A. Sheldon, Tetrahedron, 2004, 60, 501‐519. 117.  L. K. Thalén and J. E. Bäckvall, Beilstein J Org Chem, 2010, 6, 823‐

829. 118.  M. W. Larsen, D. F. Zielinska, M. Martinelle, A. Hidalgo, L. J. Jensen, 

U. T. Bornscheuer and K. Hult, ChemBioChem, 2010, 11, 796‐801. 119.  K. E. Cassimjee, M. Trummer, C. Branneby and P. Berglund, 

Biotechnol Bioeng, 2008, 99, 712‐716. 120.  K. E. Cassimjee, R. Kourist, D. Lindberg, M. Wittrup Larsen, N. H. 

Thanh, M. Widersten, U. T. Bornscheuer and P. Berglund, Biotechnol J, 2011, 6, 463‐469. 

121.  L. Rhong‐Zhen and F. Himo, Personal communication, 2008.  

57

Appendix Papers I to V