Microbial Alcohol Dehydrogenases: Identification, Characterization and Engineering

Ronnie Machielsen

Promotoren: Prof. dr. J. van der Oost Hoogleraar Microbiologie en Biochemie Wageningen Universiteit

Prof. dr. W.M. de Vos Hoogleraar Microbiologie Wageningen Universiteit

Leden van de Promotiecommissie: Prof. dr. P.C. Wright University of Sheffield

Prof. dr. R. Wever Universiteit van Amsterdam

Prof. dr. M.A. Huynen Radboud Universiteit Nijmegen

Prof. dr. H. Gruppen Wageningen Universiteit

Dit onderzoek is uitgevoerd binnen de onderzoekschool VLAG

Microbial Alcohol Dehydrogenases: Identification, Characterization and Engineering

Ronnie Machielsen

Proefschrift Ter verkrijging van de graad van doctor Op gezag van de rector magnificus van Wageningen Universiteit, Prof. dr. M.J. Kropff, in het openbaar te verdedigen op dinsdag 16 oktober 2007 des namiddags te half twee in de Aula

Microbial Alcohol Dehydrogenases: Identification, Characterization and Engineering Ronnie Machielsen

Ph.D. thesis Wageningen University, Wageningen, The Netherlands (2007) 136 p., with summary in Dutch

ISBN 978 90 8504 710 0 Abstract

Machielsen, R. (2007). Microbial Alcohol Dehydrogenases: Identification, Characterization and Engineering. PhD thesis. Laboratory of Microbiology, Wageningen University, The Netherlands.

Alcohol dehydrogeases (ADHs) catalyze the interconversion of alcohols, aldehydes and ketones. They display a wide variety of substrate specificities and are involved in an astonishingly wide range of metabolic processes, in all living organisms. Besides the scientific interest in ADHs, they are also attractive biocatalysts for industrial applications. Especially, their capability to catalyze chemo-, stereo-, and region-selective reactions endows them with considerable potential for a range of applications in food, pharmaceutical and fine chemicals industries. Natural enzymes are, however, optimized and often highly specialized for certain biological functions; hence, in many instances natural enzymes do not perform optimally in an industrial process. Laboratory evolution (random methods) and/or rationale- based protein engineering (design methods) can be applied to obtain enzyme variants, which are suitable for a particular industrial application. This study describes the identification, characterization and engineering of microbial ADHs. Because protein stability is an important issue for in vitro applications, thermophiles are considered a potential source of thermozymes. The genes coding for sixteen putative ADHs were identified in the genome of the hyperthermophilic archaeon Pyrococcus furiosus and a more detailed level of function prediction was attempted for all identified genes. Subsequently, fifteen of the putative ADHs were heterologously produced and an initial screening for activities was performed in which two of the putative ADHs showed activities. Next, these two ADHs, AdhD and Pf-TDH, were purified and characterized. AdhD is a member of the aldo-keto reductase superfamily and Pf- TDH is a L-threonine dehydrogenase. In addition, either laboratory evolution or rationale protein engineering was applied to improve the properties of two selected ADHs to extend their potential for industrial applications. AdhA, a thermostable ADH of P. furiosus, was engineered by laboratory evolution to improve the production of (2S,5S)-hexanediol at moderate temperatures. High-resolution 3D structures are required to perform the directed engineering of proteins. The ADH of Lactobacillus brevis (Lb-ADH) was selected because a 1.0 Å resolution structure was available, and because of its potential for industrial application. Structure-based computational protein engineering was applied to adjust its cofactor specificity.

Keywords: alcohol dehydrogenase, laboratory evolution, rational protein engineering, Pyrococcus furiosus, biocatalysis, characterization, computational design, thermostability.

Table of contents

Chapter 1 Aim and outline of the thesis 9

Chapter 2 General introduction 13

Chapter 3 Identification and heterologous production of NAD(P)-dependent 43 alcohol dehydrogenases from the hyperthermophilic archaeon Pyrococcus furiosus

Chapter 4 Production and characterization of a thermostable alcohol 57 dehydrogenase that belongs to the aldo-keto reductase superfamily

Chapter 5 Production and characterization of a thermostable L-threonine 71 dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus

Chapter 6 Laboratory evolution of Pyrococcus furiosus alcohol dehydrogenase to 85 improve the production of (2S,5S)-hexanediol at moderate temperatures

Chapter 7 Cofactor engineering of Lactobacillus brevis alcohol dehydrogenase 101 by computational design

Chapter 8 Summary and concluding remarks 115

Samenvatting en conclusies 121

Dankwoord / Acknowledgements 129

Curriculum vitae 131

List of publications 133

Activities in the frame of VLAG research school 135

Chapter 1

Aim and outline of the thesis Chapter 1

Alcohol dehydrogenases (ADHs) are present in virtually all organisms. They catalyze the interconversion of alcohols, aldehydes and ketones. ADHs require a cofactor, display a wide variety of substrate specificities, and play an important role in a broad range of physiological processes (e.g. alcohol and alkane metabolism, and cell defense towards exogenous alcohols and aldehydes). These features make them very interesting from a scientific perspective. Stable and enantioselective ADHs (e.g. Pyrococcus furiosus AdhA and Lactobacillus brevis ADH, mentioned further on) have also considerable potential for a range of applications in food, pharmaceutical and fine chemicals industries. The production of enantio- and diastereomerically pure diols is particularly desired because these are important chemical building blocks. Biocatalysis is gradually taking over from chemical catalysis in many industrial applications. Enzymes are environmentally friendly, biodegradable, efficient, and low cost in terms of resource requirements; as such they provide benefits compared to traditional chemical approaches in various industrial processes. In many instances, however, natural enzymes do not perform optimally in a particular unnatural process, and as such can be unsuitable for large-scale industrial applications. Industry needs enzymes that have (i) a high activity over longer periods of time, (ii) a high stability under harsh physical (high temperature) and chemical (non-aqueous solvents) conditions, and (iii) a high specificity and selectivity that does allow the enzyme to efficiently generate specific products that are not necessarily present in nature. There are three major and principally different routes to obtain enzyme variants with improved features: (1) isolating enzymes from organisms living in environments that correspond as much as possible to actual process conditions, (2) rationale- based protein engineering, and (3) laboratory evolution (random methods). The aim of the research described in this thesis was to identify and characterize novel and stable alcohol dehydrogenases for the production of fine chemicals, and to improve the properties of selected alcohol dehydrogenases to extend their potential for industrial applications. The latter has been done by laboratory evolution (random methods) and rationale protein engineering. In search for (thermo)stable ADHs for the production of aldehydes, ketones and (chiral) alcohols the hyperthermophilic archaeon Pyrococcus furiosus was chosen. Enzymes from hyperthermophiles, i.e., microorganisms that grow optimally above 80°C, generally display an extreme stability at high temperature, high pressure, as well as high concentrations of chemical denaturants; this feature generally makes hyperthermophiles a very good source of robust enzymes. In addition, the genome sequence of P. furiosus and of several closely related micro-organisms were available. Furthermore, the research described in this thesis was part of the EU project PYRED, which aimed to exploit the very high reducing power of P. furiosus (including possibly engineered strains) for the production of aldehydes, ketones and (chiral) alcohols. Therefore, identification, production and characterization of ADHs from P. furiosus, as described in this thesis, were important for the possible in vivo and in vitro production of valuable fine chemicals.

10 Aim and outline

To extend their potential for industrial applications AdhA, an alcohol dehydrogenase of P. furiosus, and the alcohol dehydrogenase of Lactobacillus brevis (Lb-ADH) were selected to be improved. AdhA was selected to improve its low-temperature activity for the production of (2S,5S)-hexanediol, while retaining the stability of the enzyme. Lb-ADH is R- specific, relatively stable and has a broad substrate specificity. These features have already made the enzyme very attractive for industrial applications. However, a drawback is its preference for NADP(H) as a cofactor, which is more expensive and labile than NAD(H). As the atomic resolution (1.0 Å) structure of Lb-ADH is available, it was possible to perform structure-based computational protein engineering. In both projects a clear engineering challenge, either improving the low-temperature activity or adjusting the cofactor specificity, could be dealt with. In addition, both laboratory evolution and rationale protein engineering could be utilized.

Chapter 2 gives a general introduction and a review on improving enzyme performance by the random approach of laboratory evolution as well as by rational and computational design. The available techniques for the engineering of enzymes, sub-divided into directed (rational protein engineering) and random (laboratory evolution) techniques, are described. Additionally, different selection and (high-throughput) screening methods are described.

Chapter 3 reports on the identification and heterologous production of sixteen predicted alcohol dehydrogenases and two previously characterized ADHs from the hyperthermophilic archaeon Pyrococcus furiosus. Based on the general genome annotations, a more detailed level of function prediction was attempted for all identified genes. Subsequently an initial screening for activities has been performed.

Chapter 4 describes the heterologous production, purification and characterization of a novel, thermostable alcohol dehydrogenase (AdhD) from P. furiosus. The physiological role of AdhD, a member of the aldo-keto reductase superfamily, is discussed.

In Chapter 5 the heterologous production, purification and characterization of the L-threonine dehydrogenase (Pf-TDH) from P. furiosus is described. Pf-TDH is most likely involved in threonine catabolism, i.e. the oxidation of L-threonine to 2-amino-3-oxobutyrate.

Chapter 6 presents the engineering of a thermostable P. furiosus alcohol dehydrogenase (AdhA) to improve the production of (2S,5S)-hexanediol at moderate temperatures. As there is no high-resolution crystal structure available for AdhA the random method of laboratory evolution was chosen as method to improve the low-temperature activity of the enzyme. Improved AdhA variants were analyzed, compared to wild-type enzyme and the results were interpreted by using a structural model of the enzyme.

11 Chapter 1

Chapter 7 deals with the engineering of the R-specific alcohol dehydrogenase from Lactobacillus brevis (Lb-ADH). Structure-based computational protein engineering was used to predict mutations to alter the cofactor specificity of Lb-ADH. Mutations were introduced into Lb-ADH and subsequently the mutants were tested for alteration of the cofactor specificity. An improved Lb-ADH variant was analyzed and compared to wild-type enzyme.

Chapter 8 summarizes the results from the previous chapters and briefly discusses the impact of the work described in this thesis.

12

Chapter 2

General introduction

Parts of this chapter have been published: Machielsen R, Dijkhuizen S, Kaper T, Looger LL & Van der Oost J. (2007) Improving enzyme performance in food applications. In: Novel enzyme technology for food applications (Ed, Rastall R) Woodhead Publishing Limited, Cambridge. Chapter 2

Biocatalysis

Biocatalysts (enzymes) are attractive for industry because of their exquisite chemo-, region-, and stereoselectivity, their impressive catalytic effiency, their reactivity in organic solvents as well as aqueous environments, and the fact that they are environmentally friendly. The excellent chiral (enantio-) and positional (region-) selectivities enzymes exhibit and the wide array of complex substrates that they can accept make it possible to use them in both simple and complex transformations without the need for the tedious blocking and deblocking steps that are common in enantio- and regioselective organic synthesis. The high selectivity and catalytic efficiency also enables efficient reactions with few by-products, thereby making enzymes an environmentally friendly alternative to chemical catalysts. These attributes have resulted in a number of applications, particularly in the food, fine chemical and pharmaceutical industries where high reaction selectivity on complex substrates is critical (Powell et al., 2001; Schmid et al., 2001). Oxidoreductases, including alcohol dehydrogenases (e.g. AdhA and Lb-ADH; Van der Oost et al., 2001; Schlieben et al., 2005), catalyze a myriad of chemo-, stereo-, and regioselective reactions and consequently they can be used for the production of a variety of pharmaceutical and chemical products or chemical building blocks which are further synthesized. Whole-cell biocatalytic processess are often used for oxidoreductases because they are cofactor dependent enzymes. Cofactors such as NAD(P)H are expensive and must be recycled. Recycling of the cofactor is most easily achieved with intact and metabolically active cells as biocatalysts. However, the use of whole cells for biocatalysis can have undesired side effects, for example the formation of unwanted by-products. Isolated enzymes are, however, increasingly used in in vitro biocatalytic processess, due to the development of highly efficient in vitro cofactor recycling systems (Powell et al., 2001; Schmid et al., 2001). Substrate-coupled NADPH regeneration is a nice example of such an in vitro cofactor recycling system, which has been described for ADHs. In substrate-coupled NADPH regeneration a single enzyme (ADH) is involved in both the main and recycling reactions. In the oxidation of 2-propanol to acetone the cofactor is regenerated in the same active site that carries out the reductive conversion of the substrate of interest. To achieve high conversions the 2-propanol should be present in excess (Bastos et al., 1999; Eckstein et al., 2004).

Alcohol dehydrogenases

Interconversions of alcohols, aldehydes and ketones are essential processes in all organisms (Fig. 2.1). The oxidoreductases catalyzing these reactions, alcohol dehydrogenases (ADHs), use a variety of different cofactors and can be divided into three classes based on their cofactor specificity: (i) NAD(P), (ii) pyrollo-quinoline quinone (PQQ), heme or cofactor F420, and (iii) flavin adenine dinucleotide (FAD). The NAD(P)-dependent alcohol dehydrogenases are the best studied class and are generally subdivided in three major groups:

14 General introduction short-chain ADHs, iron-activated ADHs, and the zinc-dependent ADHs (Reid & Fewson, 1994). In addition, some of the NAD(P)-dependent dehydrogenases are member of yet another group, the aldo-keto reductase (AKR) superfamily (Seery et al., 1998; Jez & Penning, 2001; Hyndman et al., 2003). The ADHs that use the cofactors PQQ, heme or cofactor F420 are sometimes referred to as the NAD(P)-independent ADHs. The FAD-dependent alcohol oxidases catalyze the irreversible oxidation of alcohols (Reid & Fewson, 1994). ADHs are present in virtually all organisms, they are capable of catalyzing a wide variety of substrates (e.g. normal and branched-chain aliphatic and aromatic alcohols, both primary and secondary alcohols, corresponding aldehydes and ketones, polyols) and they are involved in an astonishingly wide range of metabolic processess; they are for instance involved in alcohol, alkane, sugar and lipid metabolism, and cell defense towards exogenous alcohols and aldehydes (Park & Plapp, 1991; Ma et al., 1995; Tani et al., 2000; Burdette et al., 2002; Vangnai et al., 2002; Yoon et al., 2002). Many organisms comprise multiple ADHs, and it can often be difficult to unravel their physiological roles (Reid & Fewson, 1994).

Figure 2.1 Interconversion of alcohols, aldehydes and ketones catalyzed by NAD(P)-dependent alcohol dehydrogenases. (A) The reversible conversion of alcohols into aldehydes (R is a hydrogen, an alkyl group or an aryl group). (B) The reversible conversion of alcohols into ketones (R and R’ are an alkyl group or an aryl group).

15 Chapter 2

Improving enzyme performance

As mentioned above, there are many advantages of using enzymes in stead of traditional chemical approaches in various industrial processes. In many instances, however, natural enzymes do not perform optimally in a particular unnatural process, and as such can be unsuitable for large-scale industrial applications (Schoemaker et al., 2003). Reflecting their participation in complex metabolic networks inside living cells, enzymes are often inhibited by their own substrates or products, either of which may severely limit the productivity of an industrial biocatalytic process. During natural evolution enzymes are optimized and often highly specialized for certain biological functions within the context of a living organism. In contrast, biotechnology needs enzymes that have (i) a high activity over longer periods of time (a feature that might clash with the need for rapid protein turnover inside a cell), (ii) a high stability under harsh physical (high temperature) and chemical (non-aqueous solvents) conditions, and (iii) a high specificity and selectivity that does allow the enzyme to efficiently generate specific products that are not necessarily present in nature. There are three major and principally different routes to obtain enzyme variants with improved features: (1) isolating enzyme variants from organisms living in appropriate environments (e.g. (thermo)stable enzymes from hyperthermophiles), (2) rationale-based mutagenesis, and (3) laboratory evolution. The first option assumes that the desired enzyme is around and has been generated by natural evolution. To allow rationale-based engineering, a high-resolution model of the three-dimensional fold and insight in the structure-function relations of the biocatalyst of interest are required. Laboratory evolution offers a way to randomly optimize enzymes in the absence of structural or mechanistic information (Bornscheuer & Pohl, 2001). In this chapter, the available techniques for the engineering of enzymes are described. The engineering approaches are sub-divided into directed (rational protein engineering) and random (laboratory evolution) techniques. In addition, different selection and (high- throughput) screening methods are described, a crucial development that allows screening of large mutant libraries.

Laboratory evolution

Laboratory evolution has emerged as a powerful tool to improve biocatalysts as well as to broaden our understanding of the underlying principles for substrate specificity, stereoselectivity, and the responsible catalytic mechanism. In contrast to rational protein design (discussed on page 27), laboratory evolution does not require knowledge of the three- dimensional structure of a given enzyme or about the relationship among structure, sequence and mechanism.

16 General introduction

Laboratory evolution experiments implement a simple, iterative Darwinian optimization algorithm. Molecular diversity is typically created by random mutagenesis (e.g. error-prone PCR) and/or recombination of a target gene or a family of related target genes. Improved variants are identified in a screen (or selection) that accurately reflects the properties of interest. The gene(s) encoding those improved enzymes are if necessary used as parents for the next round of evolution. The basis of laboratory evolution, also referred to as directed evolution, was laid by Pim Stemmer in 1994. He proposed a recombination system for genes (“gene shuffling”, or “molecular breeding”) by using random fragmentation of two or more genes and their subsequent reassembly (Stemmer, 1994a,b). Compared to the error-prone PCR method in which few point mutations are introduced, the gene-shuffling rather results in blocks of mutations. Since then, numerous variations on this theme have been developed, each specific for certain types of proteins or desired outcomes (Yuan et al., 2005). Currently, laboratory evolution principles have been used to improve a variety of enzyme properties; enantioselectivity (Reetz, et al., 2004), catalytic efficiency or rate (Van der Veen et al., 2004, 2006), enzyme stability (Kim et al., 2003; Eijsink et al., 2005), pH activity profile (Bessler et al., 2003), enzyme functionality in organic solvents (Castro & Knubovets, 2003), product inhibition (Rothman et al., 2004), and substrate specificity (Zhang et al., 1997; Yano et al., 1998; Oue et al., 1999). Further developments in the field of laboratory evolution take the procedures to a different level. Enzymatic pathways (Masip et al., 2004; Umeno et al., 2005) and genomes (Patnaik et al., 2002; Zhang et al., 2002; Dai & Copley, 2004) are now subjected to various shuffling protocols, with various outcomes. Unlike natural recombination, the genetic material (gene, operon, genome) of more than two parents may be shuffled in a single laboratory evolution experiment.

Techniques used in laboratory evolution The major steps in a typical laboratory evolution experiment are outlined in Figure 2.2. The genetic diversity for evolution is created by mutagenesis and/or recombination of one or more parent sequences. These altered genes are cloned back into a plasmid for expression in a suitable host organism. Clones expressing improved enzymes are identified in a high- throughput screen, or in some cases, by selection, and the gene(s) encoding those improved enzymes are isolated and if necessary recycled to the next round of laboratory evolution. The most widely used approaches for generating diversity are random point mutagenesis and in vitro recombination. The commonly used technique to create random point mutants is error-prone PCR. The fidelity of the DNA polymerase is decreased (i) by adding divalent cations (e.g. Mn2+) to substitute for the optimal Mg2+ in the PCR mixture, or (ii) by employing high error-rate DNA polymerases (e.g. Taq polymerase lacking proofreading activity, Mutazyme which is designed to make mistakes) that will incorporate mismatching bases at a controllable rate during gene amplification. A low error-rate (2-3 base substitutions or ~1 amino acid changes) accumulates mostly adaptive mutations, whereas

17 Chapter 2

Figure 2.2 Key steps in a typical laboratory evolution experiment. higher error-rates merely generate neutral and deleterious mutations (Arnold et al., 2001). DNA shuffling methods enable the in vitro recombination of DNA sequences that are more or less related, either orthologous genes or error-prone PCR variants (Fig. 2.3). DNA shuffling requires a degree of identity >55-65%. Techniques such as incremental truncation for the creation of hybrid enzymes (ITCHY) are variant methods that allow for recombination of sequences with a lower degree of homology. In the last decade a multitude of methods have been developed to enable shuffling of genes with lower homology, to improve mutant libraries by negative selection for wild-type sequences, or to get a less biased library (Kurtzman et al., 2001). Most new methods have been invented to solve drawbacks of the original protocol or to circumvent patent limitations. Although these alternative procedures aim to solve one problem, they usually appear to create another one. For example, it is difficult to start with highly different parent genes and still reach a high recombination frequency. Methods aiming at higher recombination frequencies start with more homologous parents, thereby introducing fewer mutations. On the other hand, methods have been developed to recombine genes without any homology. These processes result mainly in only one crossover, although some methods do generate multiple crossovers at fixed places. Another way of obtaining a high recombination frequency is starting from synthetic oligonucleotides. The advantage of this method is that any mutant can be constructed and thereby the largest possible sequence space can be explored. An additional advantage is the

18 General introduction possibility to use other codons than the original ones in order to obtain more homology. Furthermore, the preferred codon usage for the expression host can be applied in the synthetic oligonucleotides. The largest disadvantages of synthetic methods are the high costs and the large size of the library, which quickly exceeds the most elaborate screening and selection methods (Ostermeier, 2003). Other conflicting parameters seem to be speed and bias. Recombination methods that aim at the generation of unbiased libraries all consist of numerous steps in order to achieve this, resulting in more elaborate procedures, while quick protocols normally result in more wild-type background and a biased library. Most of the mentioned methods can result in good mutant libraries. Therefore, the choice for one or another strategy is usually led by the size of the protein, the goal of the research, the existence of homologous proteins, the selection and/or screening capacity and practical issues like the equipment and expertise in the research group. Table 2.1 summarizes some of the methods that have been successfully utilized for laboratory evolution of a variety of proteins. This is not a complete list, as techniques and strategies for laboratory evolution are constantly arising.

Figure 2.3 The basic scheme of DNA shuffling. (A) The starting pool of homologous genes can be either a library of random mutants of a parental gene (e.g. by error-prone PCR) or a family of related genes. The pool of genes is fragmented with DNaseI. (B) Pool of random DNA fragments. (C) The fragments are reassembled into full-length genes by repeated cycles of PCR without added primers. During this step the fragments prime each other on the homologous regions, resulting in recombination when fragments derived from one parental gene prime on another one, causing a template switch (crossovers shown with the black line). Reassembly of the random fragments into full-length genes results in frequent template switching and recombination. (D) Selected pool of improved recombinants provides the starting point for another round of mutation and/or recombination.

19 Chapter 2

Table 2.1 Laboratory evolution techniques Technique Description Literature Error-prone PCR Introduces random point mutations by imposing Leung, 1989 imperfect, and thus mutagenic, reaction conditions. Cadwell, 1992 In vivo random Mutagenesis in vivo is performed by transforming a Greener, 1997 mutagenesis plasmid containing the gene to be mutagenized to a Camps, 2003 mutator E. coli strain. These strains lack DNA repair mechanisms or contain a modified polymerase with lower fidelity. Saturation mutagenesis With this approach, it is possible to create a library of Miyazaki, 1999 mutants containing all possible mutations at one or Zheng, 2004 more predetermined target positions in a gene Wong, 2004 sequence. Wong, 2005 DNA shuffling DNA is randomly digested and allowed to recombine Stemmer, 1994a,b to form novel sequences. Minshull, 1999 Synthetic shuffling Using degenerate oligonucleotides, every amino acid Ness, 2002 from a set of parents is allowed to recombine Ostermeier, 2003 independently of every other amino acid. Physical starting genes are unnecessary, and additional design criteria such as optimal codon usage can also be incorporated. Staggered extension StEP consists of priming the template sequence(s) Zhao, 1998 process (StEP) in vitro followed by repeated cycles of denaturation and Zhao, 2004 recombination extremely abbreviated annealing/polymerase- Aguinaldo, 2003 catalyzed extension. In each cycle the growing fragments anneal to different templates based on sequence complementarity and extent further. This is repeated until full-length sequences form. Due to template switching, most of the polynucleotides contain sequence information from different parental sequences. Random DNA shuffling method for generating highly Coco, 2001 chimeragenesis on recombined genes. The approach relies on the Coco, 2003 transient templates ordering, trimming and joining of randomly cleaved (RACHITT) parental DNA fragments annealed to a transient polynucleotide scaffold. Sequence independent By using inserted marker tags this technique allows at Hiraga, 2003 site-directed discrete sites for the recombination of sequences that chimeragenesis are not related at all. (SISDC)

20 General introduction

Structure based A semi-rational protein engineering approach that O’maille, 2002 combinatorial protein uses information from protein structure coupled with O’maille, 2004 engineering (SCOPE) established DNA manipulation techniques to design and create multiple crossover libraries from non- homologous genes. Incremental truncation Incremental truncation, a method for creating a library Ostermeier, 1999 for creation of hybrid of every one base truncation of dsDNA, creates Horswill, 2004 enzymes (ITCHY) diversity by changing the length of a gene. The combination of two incremental truncation libraries creates diversity by fusing two gene fragments. Performing ITCHY between two different genes generates libraries of fusion proteins in a DNA- homology independent fashion. Creating multiple- This technique is based on the ITCHY technique, but Lutz, 2001 crossover DNA includes an extra round of gene shuffling. Kawarasaki, 2003 libraries independent of sequence identity (SCRATCHY) Sequence homology- This procedure maintains alignment between two Sieber, 2001 independent protein parental genes, and produces crossovers mainly at Udit, 2003 recombination structurally related sites along the sequences. (SHIPREC) Combinatorial libraries A combination of PCR reassembly and in vivo Abecassis, 2000 enhanced by recombination in yeast produces highly shuffled recombination in yeast libraries. (CLERY) Degenerate Procedure for gene shuffling using degenerate primers Gibbs, 2001 oligonucleotide gene that allows control of the relative levels of shuffling (DOGS) recombination between the genes that are shuffled and reduces the regeneration of unshuffled parental genes. Degenerate Polymorphisms are randomly swapped between a Coco, 2002 homoduplex gene collection of polymorphic genes. This technique is family recombination different to shuffle techniques in that random (DHR) segments of genes are not recombined, but homologous segments containing point mutations.

21 Chapter 2

Selection and Screening The success of a laboratory evolution experiment highly depends on the method that is used to select the best mutant enzyme. Since most laboratory evolution experiments generate a huge mutant library, it is very important to develop an efficient method to screen this library for the desired property. Both selection and screening strategies have been developed for all kinds of enzyme functions (Boersma et al., 2007). The big challenge in these strategies is generally to make the improved function quantifiable, i.e. differentiate signal from noice. Enzymatic assays have to be sufficiently sensitive and specific to identify positive mutants (Zhao & Arnold, 1997). Selection is based on the fact that mutants with the desired enzyme function have an advantage over wild-type enzymes; variants are selected because, under certain conditions, they enable the host to grow. Selection in the lab mimics the natural survival-of-the-fittest strategy, and is the most efficient method to find the best mutant, since only mutants of interest will appear. Unfortunately, this approach is not possible for all enzymatic activities. For in vivo selection this means that only enzyme activities with a growth or survival advantage can be used. Only a few industrially interesting enzymes are essential for the bacterial cell themselves, so most selection methods are based on enzymatic activities that lead to the generation of a product that is essential for growth of the expression host. The coupling of the desired enzymatic reaction to survival in the selection step often requires the development of complex, nontrivial and intelligent assays (Taylor et al., 2001). Sometimes, this means that the substrates in these selection systems are not the desired substrates, but analogues thereof. This may result in the selection of undesired mutants with activity towards the analogue and not towards the wanted substrate. It is, therefore, very important to carefully choose the selection substrate, since the first law of laboratory evolution is: “you get what you select/screen for” (Schmidt-Dannert & Arnold, 1999). In vitro selection is usually based on the binding of the enzyme to the desired substrate or a transition state analogue, although strategies in which catalytic properties are used for selection are also described (Pelletier & Sidhu, 2001; Boersma et al., 2007). These methods are mostly based on a physical linkage between phenotype and genotype. The first established and most used technique is phage display, which has been successfully used to identify improved enzymes. In this system the enzyme of interest is fused to a coat protein of a filamentous phage and as such displayed on the outside of the phage, where in principle it is able to retain enzymatic activity. Since the gene encoding the displayed protein is present in the phage particle, the gene of the mutant enzyme with the desired property is linked to its phenotype (Fernandez-Gacio et al., 2003). When the displayed proteins may be toxic to filamentous phage assembly or incompatible with the bacterial secretion pathway, lytic phages can be used that allow displayed sequences to minimize negative selection. Other in vitro selection methods with a physical phenotype-genotype linkage are cell-surface display, ribosome display, plasmid display and mRNA-protein fusion (Lin & Cornish, 2002; Becker et al., 2004). Recently, a different approach was described to maintain a linkage between

22 General introduction genotype and phenotype. In vitro compartmentalisation is a method in which compartments are formed as aqueous droplets in water-in-oil emulsions which contain only one gene and a complete transcription/translation machinery (Tawfik & Griffiths, 1998). These droplets mimic a bacterial cell by keeping the gene and its product together. The droplets containing an enzyme with the desired activity can be selected by fluorescence activated cell sorting (FACS) or, when the gene is physically bound to the substrate, by breaking the droplets and fishing out the desired product (Griffiths & Tawfik, 2000). The advantages of in vitro over in vivo selection are the larger sample size of a mutant library that can be handled, and as such the larger amount of possible enzyme variants to be tested. A drawback is that making the right water-in-oil emulsions with only one gene per droplet is tricky, and that the efficiency of the in vitro transcription & translation can be a bottleneck. Another way of finding the desired mutant enzyme is by screening. In screening methods all mutants have to be tested for the desired enzymatic reaction, even those that might not be active or accurately folded. The advantage is, however, that almost every enzymatic reaction can be tested, since the activity does not have to be dependent on growth rate or the formation of essential products. This can be done in a qualitative way by relatively simple visual screens such as the formation of coloured or fluorescent products or halos around a colony on a plate. For protein functions such as catalysis of a specific reaction or substrate specificity this is very difficult or even impossible. Quantitative methods are better suited to screen for these enzymatic activities, but are usually more labor intensive. This means that on a normal time scale only small libraries can be tested, or that high throughput screening (HTS) has to be employed. HTS demands miniaturisation and automation of enzymatic assays. In the past decade a lot of research has been focussed on finding better, cheaper, quicker and more accurate HTS assays (Cohen et al., 2001; Aharoni et al., 2005). This has made HTS feasible for many laboratories all over the world now, resulting in a lot of smart enzymatic screening methods. A screening method is used to find the best enzyme from a large pool of mutants. The resulting amount of clones depends on the accuracy of the assay and should be optimised for every laboratory evolution experiment. For this optimisation one should consider (i) the size of the library, (ii) the amount of (false) positive and false negative colonies that can be allowed, (iii) the costs of the assay, and (iv) the possibility of performing a more accurate assay with the best mutants. Most screening assays are based on spectrophotometric methods in 96 or 384 wells plates (most often colorimetric detection) or fluorescent methods. Usually raw cell extracts are incubated with the substrate or an analogue thereof, which is followed by measuring either substrate consumption, cofactor conversion or product generation. Another possibility is the use of a discontinuous assay in which the product is used by a second enzyme, which allows for the indirect analysis of product formation of the enzyme of interest. More HTS methods are developed every day, both for specific enzymatic reactions and general applications, making screening the method of choice for many researchers.

23 Chapter 2

Examples of improving enzyme stability and ability by laboratory evolution

• Enzyme stability Naturally available enzymes are often not optimally suited for industrial applications. This incompatibility generally relates to the stability of the enzymes under process conditions. Although it sometimes is beneficial to adapt industrial processes to mild and environmentally friendly conditions favoured by the enzyme, the use of more extreme conditions is often desirable. Despite many successful efforts to understand the structural basis of protein stability, there is still no universal strategy to stabilise any protein by a limited number of rationally designed mutations (Eijsink et al., 2004). It is concluded that protein stability appears to be the result of many small stabilizing features. Although the rational design (see section: rational protein engineering) of enzyme stabilization has been successful in several instances (e.g. Nielsen & Borchert, 2000 - α−amylase; Van den Burg et al., 1998 - protease), it relies heavily on the availability of a high-resolution 3D structure. Therefore, the random techniques used in laboratory evolution can be a powerful alternative to improve enzyme stability. The method used for screening the libraries obtained during laboratory evolution is extremely important. With respect to thermal stability for instance, the distinction between thermal tolerance and real thermal stability is important. The first term refers to the ability to withstand incubation at elevated temperatures, without necessarily being active at those temperatures. Real thermal stability refers to enzymes that not only withstand elevated temperatures, but that also retain activity at these temperatures. Clearly, these two types of properties need different screening regimes. It is important to ensure that the screening procedure accounts for all the properties that one wishes to improve, while ensuring that other qualities that are important for a certain process (for instance activity) are preserved. Screening directly for activity under denaturing conditions (e.g. high temperature, extreme pH, organic solvents) is a simple and good method, but this may pose some practical problems. For example, in the case of thermal stability one faces the limitation that most high- throughput microplate readers can only reach temperatures of about 50°C. Most strategies for screening for enzyme stability rely on measuring residual enzyme activity after exposure to a denaturing challenge on a filter or in microtiter plate wells (Eijsink et al., 2005). There are alternative methods for stability selection based on phage-display (e.g. the Proside method, a phage-based method for selecting thermostable proteins; Sieber et al., 1998; Martin et al., 2003) or the use of extremophilic microorganisms as expression host, e.g. Thermus thermophilus has been used for the selection of thermostable selection markers (e.g. Hoseki et al., 1999 - kanamycin; Brouns et al., 2005 - bleomycin resistance) and enzymes (e.g. Tamakoshi et al., 2001 – 3-isopropylmalate dehydrogenase). The value of laboratory evolution in stabilisation of enzymes was illustrated by Richardson et al. (2002), who used it for the development of a stable amylase for starch liquefaction in corn wet milling. Corn wet milling is an example of a multistep industrial process in which improvement of enzyme performance is desired. The process would benefit

24 General introduction from an α-amylase capable of liquefying starch at pH 4.5 and 105°C without the addition of calcium. High-throughput screening of microbial DNA libraries from various environments was used to identify α-amylases with activity at low pH, and high temperature in the absence of calcium. This large screening effort yielded 15 primary hits and three variants having good properties with respect to at least one of the desired phenotypes were selected for optimization by DNA shuffling. This next step was performed by gene reassembly, random ligation of a pool of restriction site-determined fragments of ca. 150 bp., off all three selected genes. The resulting library of 21,000 clones was first screened for activity at pH 4.5 and high temperature. Subsequently, improved variants resulting from this screening (<100) were tested for thermal stability, liquefaction ability, expression levels and dependence on calcium. After this biochemical and process-specific characterization of the best variants, one α- amylase with exceptional process compatibility was identified (Richardson et al., 2002). Xylanases catalyze the hydrolysis of xylan, a major constituent of hemicellulose. The enzyme is the focus of much attention because of its potential for use in industrial processes, including paper and pulp industries and food and feed industries. Many efforts have been made to improve the properties of xylanase to handle industrial tasks. In particular, thermostability is a major target of such modifications. Recently, Miyazaki et al. used laboratory evolution to enhance the thermostability of glycosyl hydrolase family-11 xylanase from Bacillus subtilis. At first, random mutagenesis (error-prone PCR) was carried out for the entire sequence of the xylanase gene and the mutant library was screened for thermostability. Next site-saturation mutagenesis was employed to optimize the sequence at three positions, which were identified as positions that are involved in thermostability by the random mutagenesis approach. Improved variants were then recombined by DNA shuffling. This combination of random mutagenesis, saturation mutagenesis, and DNA shuffling yielded a thermostable variant, which contained three amino acid substitutions. The half-inactivation temperature (the midpoint of the melting curves) for the xylanase variant compared with the wild-type enzyme after incubation for 10 minutes was elevated from 58 to 68°C and at 60°C the thermostable variant retained full activity for more than 2 h, whereas the wild-type enzyme inactivated within 5 minutes (Miyazaki et al., 2006).

• Enantioselectivity By inverting the enantioselectivity of a key enzyme in a multi-enzyme pathway, a process for the production of L-methionine in E. coli has been improved. Through random mutagenesis, saturation mutagenesis and screening the enantioselectivity of a D-hydantoinase was inverted (from enantiomeric excess eeD = 40% to eeL = 20%) and its total activity with D,L-5-(2- methylthioethyl)hydantoin (D,L-MTEH) as substrate was increased five-fold. Introduction of the evolved L-hydantoinase into a recombinant whole-cell catalyst increased productivity for L-methionine and decreased accumulation of an undesired intermediate, compared to cells with the wild-type pathway. It is worth noting that a single amino acid substitution was sufficient to invert the hydantoinase enantioselectivity. Highly D-selective variants (eeD =

25 Chapter 2

90%) containing a single amino acid substitution were also found (May et al., 2000). Until recently the number of studies on the laboratory evolution of enantioselective biocatalysts is limited, one reason being the lack of generally applicable high-throughput screening methods. However, several efficient high-throughput ee assays have been developed lately, so that today the analytical bottleneck of improving the enantioselectivity of many enzymes can be solved. The most efficient ee assays are based on mass spectrometry (MS), nuclear magnetic resonance spectroscopy (NMR) and infrared spectroscopy IR which allow 1,000 – 10,000 samples to be evaluated per day (Reetz, 2004).

Genome shuffling Current methods for the improvement of industrial microorganisms range from the random approach of classical strain improvement (CSI; sequential random mutagenesis and screening) to the highly rational methods of metabolic engineering. The first strategy is resource intensive and time-consuming, requiring many generations of mutation and selection to allow accumulation of multiple beneficial mutations in a single strain. The efficacy of the second strategy is limited by the ability to predict which mutations will improve a particular phenotype. Thus, it is not possible to take advantage of mutations in genes that are not obviously related to the phenotype of interest but may nevertheless improve microbial fitness or performance under a particular set of conditions. Genome shuffling is a random method that is much more efficient for evolution of strains of microbes with desirable phenotypes. Genome shuffling is a process that combines the advantage of multi-parental crossing allowed by family DNA shuffling with the combination of entire genomes normally associated with conventional breeding. The driving force for the accelerated evolution is the recombination of multiple parents in a recursive manner. The advantage of genome shuffling over CSI has been demonstrated with Streptomyces fradiae, a commonly used strain for commercial production of the complex polyketide antibiotic tylosin (Zhang et al., 2002). To generate a diverse population for genome shuffling, a low-production parental strain was subjected to one round of CSI using nitrosoguanidine as mutagen. Two rounds of genome shuffling based on recursive protoplast fusion of mixed populations and screening for tylosin production resulted in mutant strains with productivities similar to that of the commercial strain SF21. However, while it took 20 years and about 1,000,000 assays for the 20 rounds of CSI required to obtain SF21, similar results were produced with 24,000 assays in 1 year of genome shuffling. Patniak et al. (2002) demonstrated the use of genome shuffling for improved acid tolerance in production of lactic acid by lactobacilli. Lactobacilli are of interest in the commercial production of lactic acid, a compound that has long been used as a food additive and as an important feedstock for the production of other chemicals such as 2,3-pentadione. Improving the growth and lactic acid production of lactobacilli at low pH could decrease waste and the cost of product purification. Growth at low pH involves multiple widely distributed loci and without an understanding of the mechanism of pH tolerance in bacteria, a

26 General introduction rational approach to engineering pH tolerance was impractical. In addition, the Lactobacillus strain used in their study, is of commercial interest, but has not been genetically characterized. Lactobacillus strains with improved low-pH tolerance were first obtained by CSI in order to generate the initial biodiversity pool and then shuffled for five rounds by recursive protoplast fusion. The shuffled population contained new strains that grew at substantially lower pH (pH 3.8) than the wild-type strain does and improved strains that produced threefold more lactic acid than the wild-type strain at pH 4.0 (Patnaik et al., 2002). Genome shuffling was also used to improve the degradation of pentachlorophenol (PCP) by the Gram-negative bacterium Sphingobium chlorophenolicum (Dai & Copley, 2004). Degradation of pentachlorophenol, a highly toxic anthropogenic pesticide, is slow and S. chlorophenolicum cannot tolerate high levels of PCP. To generate a diverse population for genome shuffling, S. chlorophenolicum was subjected to one round of CSI using nitrosoguanidine as mutagen. Three successive rounds of protoplast fusion were carried out, and after each round, the concentration of PCP in the plates used for selection was increased. Several strains obtained after the third round of shuffling grew on plates containing 6 to 8 mM PCP, while the original strain cannot grow in the presences of concentrations higher than 0.6 mM. Some mutant strains were able to completely degrade 3mM PCP in liquid media, whereas no degradation could be achieved by the wild-type strain. The examples described all use recursive protoplast fusion as a useful method for the genome shuffling, but other mechanisms of genetic exchange (such as conjugation and phage- mediated transduction) are also useful for the genomic shuffling of bacteria, and lower and higher eukaryotes. As most of these approaches rely on natural homologous recombination, organisms modified by these means are not considered “genetically modified”. Genome shuffling thus represents a practical method for the rapid manipulation of the complex phenotypes of whole cells and organisms (Stephanopoulos, 2002; Zhang et al., 2002).

Rational and computational protein engineering

The “holy grail” of rational protein engineering and computational design is the accurate de novo design of proteins with specified characteristics. However, the astronomical number of possible interactions in an average protein presents computational problems and currently limits de novo prediction of protein structure to small proteins (Baker, 2006). As most proteins of interest are fairly large, this implies that proteins of known structure (or which are highly homologous to a protein of known structure) will be the most amenable to rational and computational engineering. This is in contrast to the aforementioned laboratory evolution methods, which require no knowledge of protein structure. A high-resolution structure is most preferable, but a reliable model (supplemented by sequences of homologues, mutant characterization, reaction mechanism, and the like) may also facilitate design. Rational molecular engineering requires the establishment of a set of desired design goals, and a paradigm with which to rank proposed sequences and structures. For

27 Chapter 2 computational design, this takes the form of a quantitative molecular fitness function for the system. This typically involves “first principles” biophysical components (Kraemer-Pecore et al., 2001), e.g. ‘van der Waals’ steric contacts, hydrogen bonds, electrostatics, and cavity formation, but may also include indirect terms such as statistical preponderance of side-chains in given combinations or protein micro-environments. Depending on the level of sophistication of the engineering experiment, terms describing bond length stretching, bond angle bending, and quantum mechanical terms may be added to the fitness function. Recent developments in computational protein engineering include the repeated cycling between different optimization techniques (e.g. Monte Carlo and gradient minimization (Baker, 2006), dead-end elimination and hydrogen bond inventorying (Looger et al., 2003)), design of sequence libraries instead of individual sequences (Saraf et al., 2006), and the simultaneous design of multiple molecular components. Recent design goals of rational and computational engineering have been thermostabilization (Korkegian et al., 2005), altering of the specificity of a superfamily of bacterial binding proteins (Looger et al., 2003), re-engineering of protein-protein interactions (Kortemme et al., 2004), alteration of enzyme substrate specificity (Ashworth et al., 2006; Park et al., 2006), and de novo design of activity (Dwyer et al., 2004). It is likely that the greatest advances will be made by a judicious combination of rational and evolutionary techniques. Both methodologies are employed to select particular regions of sequence space for experimental analysis. The metaphor of climbing a mountain range has been proposed for sequence space optimization; laboratory evolution may be thought of as ascending to the highest peak within a given mountain range (path A in Fig. 2.4). It seems likely that rational design will facilitate more dramatic sequence space movements, as >20 simultaneous mutations may be required for a novel function within a given scaffold. It thus seems appropriate to describe rational design as being capable of discovering a new island in sequence space and landing on the beach (path B in Fig. 2.4), after which laboratory evolution may discover the peak of the island (path C in Fig. 2.4).

Laboratory techniques for the construction of designed sequences Once individual sequences or sequence libraries have been designed, they must be fabricated in the laboratory and tested. QuikChange mutagenesis (Stratagene) is a reliable method for introducing mutations, but is limited to a single mutagenic oligonucleotide per round, whereas Kunkel mutagenesis (Kunkel, 1985) efficiently introduces multiple mutagenic primers simultaneously. “Inside-out” PCR techniques have also been employed for the assembly of target sequences. Finally, rational design may be employed to design sequence libraries, which can be input to laboratory evolution methods.

28 General introduction

Figure 2.4 Fitness of an enzyme for functions 1 and 2 versus sequence space.

Examples of improving enzyme stability and ability by rational protein engineering

• Enzyme stability Proteins exhibit marginal stabilities equivalent to only a small number of weak intermolecular interactions (Jaenicke, 2000). Therefore, a single successfully engineered stabilising interaction can lead to a significant increase in overall protein stability. Factors that contribute to protein stability are packing of the hydrophobic core, disulfide and salt bridges, hydrogen bonding, and secondary structure propensities (Strickler et al., 2006). The mobility of a single loop can determine the overall stability of a protein. Eight designed mutations mainly in a 14- residue surface loop increased the kinetic stability of the moderately stable metalloprotease from Bacillus stearothermophilus CU21 more than 350-fold at 100 ºC while retaining wild- type catalytic properties (Van den Burg et al., 1998). Since inactivation of the enzyme was thought principally to occur by autodegradation, the mutations were designed to reduce the mobility of the surface loop. The stabilized protein contained an engineered disulfide bridge, a rigidifying Xaa → Pro mutation, and five residues present in thermolysine, a close thermostable homologue of the protease. The study exemplified that protein stability is acquired by a combination of multiple stabilizing interactions and that low-temperature activity and enzyme stability can be independent from each other. The importance of efficient packing of the hydrophobic core in protein stability was shown by computational redesign of the labile, 153-amino-acid, dimeric yeast cytosine deaminase (Korkegian et al., 2005). Using the RosettaDesign scoring function and a Metropolis Monte Carlo search algorithm, target sequences were threaded onto the backbone of the available 3D-structure of the enzyme and mutated towards a lower energy. Residues in

29 Chapter 2 and directly surrounding the active site as well as those involved in oligomerization were excluded, subjecting about 40% of the residues to the design. Half of those emerged as wild- type after the design. Of the 33 predicted substitutions, those at the protein surface were ignored. Finally, three mutations, all resulting in a more efficient packing of the hydrophobic core, were found to be stabilizing (Fig. 2.5 A). The triple mutant resulted in a 30-fold increase of kinetic stability at 50ºC. One remaining challenge for computational redesign algorithms is the modelling of the interactions of buried polar and charged side chains (Korkegian et al., 2005). Interactions at the protein surface contribute to thermostability as well, as protein structures of proteins from hyperthermophilic origin have indicated (Vieille & Zeikus, 2001), and provide another possibility for improving the stability of enzymes. The contribution of surface salt bridges to stability depends on the spatial orientation of the involved residues, their desolvation upon salt bridge formation, and the local context of the salt bridge (Makhatadze et al., 2003). By combining a genetic algorithm for evolution of protein sequences with a computational evaluation of surface charge-charge interactions, five proteins ranging 72-100 amino acids in size were stabilized up to 18 kJ/mol resulting from the introduction of only a few salt bridges (Fig. 2.5 B; Strickler et al., 2006). Another computational technique, which has been experimentally validated, is the design of particular stabilizing interactions such as disulfide bonds. Recently the computational design of a disulfide-stabilized double mutant of the penicillin acylase from Alcaligenes faecali has been reported (Wang et al., 2006). The mutant exhibits a 50%- increased half-life at 55°C and will be useful for antibiotic biosynthesis.

• Reaction mechanism The glycoside hydrolases present a diverse group of enzymes that are of great significance for food industry. Consequently, many studies have been undertaken to engineer their activity. Glycoside hydrolases that use the retaining mechanism for cleavage of glycosidic bonds have been engineered into ’glycosynthases’ and ’thioglycoligases’ for the exclusive synthesis of oligosaccharides and thioglycosides by engineering of the catalytic residues in the active site. To obtain glycosynthases, the catalytic nucleophile of retaining glycosidases was removed (Mackenzie et al., 1998). These inactivated enzymes are able to synthesize oligosaccharides from glycosyl fluorides with inverted anomeric stereochemistry and suitable acceptors in high yields. This approach has been extended to several other glycoside hydrolases like exo-β- glycosidases (Trincone et al., 2000), endo-β-glycosidases (Fairweather et al., 2002, 2003; Van Lieshout et al., 2004), endo-β-xylanases (Sugimura et al., 2006), exo-α-glucosidases (Okuyama et al., 2002), and exo-α-xylosidases (Moracci et al., 2001). Thioglycoligases lack the catalytic general acid/base residue of retaining glycosidases and use dinitrophenyl glycosides as donors and deoxy-thiosugars as acceptors (Mullegger et al., 2005). Cyclodextrin glucano transferases (CGTases) of the α-amylase family convert starch into a mixture of α, β , and γ-cyclodextrins, circular donut-shaped dextrin molecules of 6,7 or

30 General introduction

8 glucose molecules with a hydrophilic exterior and a hydrophobic core that can be used for micro-encapsulation of hydrophobic molecules, thus protecting it from an aqueous environment. One of the applications of cyclodextrins in food is as a flavor preservative. Based on comparison of crystal structures, substrate binding sites have been probed by site- directed mutagenesis and this resulted in optimized CGTase variants with tailored α, β, and γ cyclodextrin yields (Wind et al., 1998). In addition, CGTase could be converted from a glucano transferase into a starch hydrolase by a single site-directed mutation in acceptor substrate binding site +2 (Leemhuis et al., 2002). An additional substitution in acceptor subsite +1 resulting in the same reversal of reaction specificity was identified by a laboratory evolution approach (Leemhuis et al., 2003).

• Ligand and substrate specificity There have been several successful alterations of the specificity of receptors and enzymes. A number of members of the bacterial periplasmic binding protein (PBP) superfamily have been structurally solved in the ligand-free open and/or ligand-bound closed form (Dwyer et al., 2004), allowing careful study of this general hinge-bending mechanism of ligand binding. The modularity of this conformational change has allowed the computational diversification of these binding proteins for a variety of ligands. Metal-binding sites (Marvin & Hellinga, 2001) and nascent metalloenzyme active sites (Benson et al., 2002) have been designed into family members. Subsequently, binding sites for trinitrotoluene, serotonin, lactate, and a soman nerve gas analogue were designed into proteins with a cognate specificity for sugars and amino acids (Fig. 2.5 C; Looger et al., 2003; Allert et al., 2004). In several cases, the proteins were produced in bacteria as part of a synthetic signal transduction cascade, harnessing the designed binding event to initiate expression of a reporter gene. The interaction of proteins with other proteins (Kortemme et al., 2004) and with nucleic acids (Ashworth et al., 2006) has been systematically altered by computational design, as well. A recent experiment of particular note is the in silico affinity maturation of a therapeutic antibody by computational protein design (Clark et al., 2006), in which the already-tight binding affinity of a clinically relevant antibody was computationally improved by an order of magnitude. The DNA-binding and -cleaving specificity of the homing endonuclease I-MsoI has been altered by computational design (Ashworth et al., 2006). A single base-pair substitution known not to be well-recognized by the wild-type enzyme was modelled in complex with the enzyme in silico, and interfering side-chains were identified. Computational design was used to repack the protein-nucleic acid interface, with particular emphasis on the formation of precise hydrogen bonds with the novel DNA bases introduced. A double mutant was predicted, shown to cleave the target DNA with 10,000-fold increased efficiency relative to the wild-type enzyme, and was crystallized and shown to adopt the target interface structure (Fig. 2.5 D).

31 Chapter 2

Wild-type protein Designed protein

A)

B)

C)

D)

E)

Figure 2.5 Examples of designed proteins discussed in the text. (A) Improved packing of hydrophobic core in yeast cytosine deaminase (Korkegian et al., 2005). (B) Optimized surface charge distribution of ubiquitin (Strickler, 2006). (C) Designed trinitrotoluene-binding protein (Looger et al., 2003). (D) Designed affinity of homing endonuclease I-MsoI for a novel recognition sequence (Ashworth et al., 2006). (E) Designed cefotaxime hydrolase (β-lactamase) (Park et al., 2006).

32 General introduction

In a demonstration of complementary methods for modelling and design, the known crystal structure of human butyrylcholinesterase was used to create a model of cocaine docked into the active site (Pan et al., 2005). Molecular dynamics was then employed to model the ensemble of conformations, which the substrate can assume in the binding pocket, and in silico mutagenesis was used to select a quadruple mutant predicted to stabilize the hydrolytic transition state of cocaine better than the wild-type protein. The mutant was experimentally characterized and shown to catalyze the breakdown of cocaine roughly 500-fold faster than the wild-type enzyme. An enzyme useful for “green synthesis” of the flavour vanille, the vanillyl-alcohol oxidase (VAO), has been improved by alteration of its stereospecificity (van den Heuvel et al., 2000). The architechture of the active site of VAO has been perturbed by its rational engineering; transfer of the reactive side chains to the opposite face of the active site cavity leads to a significant increase in specificity (>80% enantiomeric excess) for the desired substrate enantiomer. Crystal structure determination of the designed double mutant indeed has verified the proposed active site structure.

Examples of combined laboratory evolution and computational design As stated above, the most successful designs are likely to result from the combined efforts of rational/computational design and laboratory evolution (Chica et al., 2005). Computational methods may be used to directly design sequence libraries (Voigt et al., 2001); alternatively, designed sequences may be optimized by evolutionary methods. Feedback of experimental results into modelling steps may increase the accuracy of fitness functions used in the initial design steps. The following examples demonstrate the power of combining rational protein engineering and laboratory evolution. The E. coli ribose-binding protein, a member of the PBP superfamily and catalytically inert, has been converted into a triose phosphate isomerase (TIM) enzyme via computational design followed by laboratory evolution (Dwyer et al., 2004). A model of the target active site geometry was constructed by analysis of the crystal structure complex of a transition state analogue in wild-type TIM. Binding pocket side-chains and a transition state model were simultaneously optimized in silico to produce a set of putative active sites, which were then repacked to produce a set of final designs. The resulting designs contained 18 to 22 mutations, exhibited a rate enhancement of more than 105 over that of the uncatalyzed reaction, and are biologically active. The best design exhibited a Kcat/Km ratio for the conversion of dihydroxyacetophenone phosphate to glyceraldehyde-3-phosphate, which was still about three orders of magnitude less than the ratio for wild-type triose phosphate isomerase.

Subsequently, a laboratory evolution approach has been used to improve the Kcat/Km ratio of the designed enzyme. Km is a constant that is equal to the substrate concentration at which an enzyme reaction proceeds at half the maximum velocity. This represents (for enzyme reactions exhibiting Michaelis–Menten kinetics) the affinity for the substrate. The larger the

Km, the weaker the binding affinity of the enzyme for the substrate. Kcat (catalytic constant) is

33 Chapter 2 also called the turnover number of the enzyme, that is, the number of reaction processes

(turnovers) that each active site catalyzes per unit time. The Kcat/Km ratio is a measure of the enzyme’s catalytic efficiency. Final protein variants (~3-5 additional mutations) were selected with sufficient activity to complement a TIM-deficient strain. As is often the case, many of the accumulated changes identified by laboratory evolution were localized at the protein surface, in regions distant from the active site, and their effect on activity is therefore difficult to rationalize and with the present knowledge impossible to predict (Dwyer et al., 2004).

A similar approach was used to alter the substrate specificity of the (α/β)8 hydrolytic enzyme glyoxalase II from S-D-lactoylglutathione to the antibiotic cefotaxime (Park et al., 2006). Rational design was used to graft the active site loops from β-lactamase into glyoxalase, with insertions and deletions both used to select optimal grafts. Initial designs were subjected to numerous rounds of PCR shuffling followed by increasingly-stringent antibiotic selection. Final designs were 59% identical to the parent glyoxalase scaffold, and produced a 100-fold increased cefotaxime resistance in transformed bacteria (Fig. 2.5 E). Combined rational and evolutionary protein engineering has been applied to enzymes of industrial relevance, as well. The resistance of α-amylase to both thermal (Nielsen & Borchert, 2000) and pH-induced (Nielsen et al., 2001) unfolding has been increased by both techniques. Rational methods, such as the introduction of prolines, optimization of a metal- binding site, and improvement of electrostatic interactions, were used to render the α-amylase scaffold more robust. Subsequent random mutagenesis identified further beneficial mutations not predicted by any of the rational improvements.

Discussion and future trends

Literature contains now a large number of studies in which enzymes have been improved by rational or random approaches. It has been shown that both approaches have strengths, but also weaknesses. Rationale-based engineering on the one hand requires for instance a high- resolution model of the three-dimensional fold and insight in the structure-function relations of the biocatalyst of interest. Laboratory evolution on the other hand requires the ability to screen large libraries; when major adjustments are desired, the leap in sequence space may be too big to bridge. In other words: the chance is extremely small to successfully obtain a protein variant with more than five specific amino acid substitutions - this will only work when extremely large libraries can be screened (205 = 3.2 x 106 combinations). For that reason the random approach heavily depends on the starting material (the gap in sequence space should be relatively limited), and the efficiency of screening and selection. Progress in the random approach of laboratory evolution will be focused on improved techniques to generate diversity, but more importantly on the development of efficient (simple and cheap) methods for the screening of large mutant libraries for a large number of enzyme features. At the same time rational protein engineering will take advantage of (i) the increased computer power and optimized combinatorial algorithms, and (ii) the increasing knowledge of protein structures

34 General introduction that is due to the fast growing PDB (not in the least because of structural genomics initiatives) and to the insight from previous design and laboratory evolution studies. Clearly, these developments will have decisive influence on the future use of the independent rational and random approaches in enzyme engineering. Moreover, the complementary use of both rational design and laboratory evolution appears to be a very promising path towards the production of proteins with new and improved properties.

References

Abecassis V, Pompon D & Truan G (2000) High efficiency family shuffling based on multi-step PCR and in vivo DNA recombination in yeast: statistical and functional analysis of a combinatorial library between human cytochrome P450 1A1 and 1A2. Nucl Acids Res 28(20): e88. Aguinaldo AM & Arnold FH (2003) Staggered extension process (StEP) in vitro recombination. Methods Mol Biol 231: 105-110. Aharoni A, Griffiths A D & Tawfik DS (2005) High-throughput screens and selections of enzyme- encoding genes. Curr Opin Chem Biol 9(2): 210-216. Allert M, Rizk SS, Looger LL & Hellinga HW (2004) Computational design of receptors for an organophosphate surrogate of the nerve agent soman. Proc Natl Acad Sci USA 101(21): 7907-12. Arnold FH, Wintrode PL, Miyazaki K & Gershenson A (2001) How enzymes adapt: lessons from directed evolution. Trends Biochem Sci 26(2): 100-106. Ashworth J, Havranek JJ, Duarte CM, Sussman D, Monnat RJ Jr., Stoddard BL & Baker D (2006) Computational redesign of endonuclease DNA binding and cleavage specificity. Nature 441(7093): 656-659. Baker D (2006) Prediction and design of macromolecular structures and interactions. Philos Trans R Soc Lond B Biol Sci 361(1467): 459-463. Bastos FM, Dos Santos AG, Jones J, Oestreicher EG, Pinto GF & Paiva LMC (1999) Three different coupled enzymatic systems for in situ regeneration of NADPH. Biotechnol Tech 13: 661-664. Becker S, Schmoldt HU, Adams TM, Wilhelm S & Kolmar H (2004) Ultra-high-throughput screening based on cell-surface display and fluorescence-activated cell sorting for the identification of novel biocatalysts. Curr Opin Biotechnol 15(4): 323-329. Benson DE, Haddy AE & Hellinga HW (2002) Converting a maltose receptor into a nascent binuclear copper oxygenase by computational design. Biochemistry 41(9): 3262-3269. Bessler C, Schmitt J, Maurer KH & Schmid RD (2003) Directed evolution of a bacterial α- amylase: toward enhanced pH-performance and higher specific activity. Protein Sci 12(10): 2141- 2149. Boersma YL, Melloney JD & Quax WJ (2007) Selection strategies for improved biocatalysts. FEBS J 274: 2181-2195. Bornscheuer UT & Pohl M (2001) Improved biocatalysts by directed evolution and rational protein design. Curr Opin Chem Biol 5(2): 137-143. Brouns SJJ, Wu H, Akerboom J, Turnbull AP, De Vos WM & Van der Oost J (2005) Engineering a selectable marker for hyperthermophiles. J Biol Chem 280(12): 11422-11431.

35 Chapter 2

Burdette DS, Jung SH, Shen GJ, Hollingsworth RI & Zeikus JG (2002) Physiological function of alcohol dehydrogenases and long-chain (C30) fatty acids in alcohol tolerance of Thermoanaerobacter ethanolicus. Appl Environ Microbiol 68(4): 1914-1918. Cadwell RC & Joyce GF (1992) Randomization of genes by PCR mutagenesis. PCR Methods Appl 2(1): 28-33. Camps M, Naukkarinen J, Johnson BP & Loeb LA (2003) Targeted gene evolution in Escherichia coli using a highly error-prone DNA polymerase I. PNAS 100(17): 9727-9732. Castro GR & Knubovets T (2003) Homogeneous biocatalysis in organic solvents and water-organic mixtures. Crit Rev Biotechnol 23(3): 195-231. Chica RA, Doucet N & Pelletier JN (2005) Semi-rational approaches to engineering enzyme activity: combining the benefits of directed evolution and rational design. Curr Opin Biotechnol 16(4): 378-384. Clark LA, Boriack-Sjodin PA, Eldredge J, Fitch C, Friedman B, Hanf KJ, Jarpe M, Liparoto SF, Li Y, Lugovskoy A, Miller S, Rushe M, Sherman W, Simon K & Van Vlijmen H (2006) Affinity enhancement of an in vivo matured therapeutic antibody using structure-based computational design. Protein Sci 15(5): 949-960. Coco WM, Levinson WE, Crist MJ, Hektor HJ, Darzins A, Pienkos PT, Squires CH & Monticello DJ (2001) DNA shuffling method for generating highly recombined genes and evolved enzymes. Nat Biotechnol 19(4): 354-359. Coco WM, Encell LP, Levinson WE, Crist MJ, Loomis AK, Licato LL, Arensdorf JJ, Sica N, Pienkos PT & Monticello DJ (2002) Growth factor engineering by degenerate homoduplex gene family recombination. Nat Biotechnol 20(12): 1246-1250. Coco WM (2003) RACHITT: Gene family shuffling by random chimeragenesis on transient templates. Methods Mol Biol 231: 111-127. Cohen N, Abramov S, Dror Y & Freeman A (2001) In vitro enzyme evolution: the screening challenge of isolating the one in a million. Trends Biotechnol 19(12): 507-510. Dai M & Copley SD (2004) Genome shuffling improves degradation of the anthropogenic pesticide pentachlorophenol by Sphingobium chlorophenolicum ATCC 39723. Appl Environ Microbiol 70(4): 2391-2397. Dwyer MA, Looger LL & Hellinga HW (2004) Computational design of a biologically active enzyme. Science 304(5679): 1967-1971. Dwyer MA & Hellinga HW (2004) Periplasmic binding proteins: a versatile superfamily for protein engineering. Curr Opin Struct Biol 14(4): 495-504. Eckstein M, Dauβmann T & Kragl U (2004) Recent Developments in NAD(P)H regeneration for enzymatic reductions in one- and two-phase systems. Biocatal Biotransform 22: 89-96. Eijsink VG, Bjork A, Gaseidnes S, Sirevag R, Synstad B, Van den Burg B & Vriend G (2004) Rational engineering of enzyme stability. J Biotechnol 113(1-3): 105-120. Eijsink VG, Gaseidnes S, Borchert TV & Van den Burg B (2005) Directed evolution of enzyme stability. Biomol Eng 22(1-3): 21-30. Fairweather JK, Faijes M, Driguez H & Planas A (2002) Specificity studies of bacillus 1,3-1,4- beta-glucanases and application to glycosynthase-catalyzed transglycosylation. Chembiochem 3(9): 866-873.

36 General introduction

Fairweather JK, Hrmova M, Rutten SJ, Fincher GB & Driguez H (2003) Synthesis of complex oligosaccharides by using a mutated (1,3)-beta-D-glucan endohydrolase from barley. Chemistry 9(11): 2603-2610. Fernandez-Gacio A, Uguen M & Fastrez J (2003) Phage display as a tool for the directed evolution of enzymes. Trends Biotechnol 21(9): 408-414. Gibbs MD, Nevalainen KMH & Bergquist PL (2001) Degenerate oligonucleotide gene shuffling (DOGS): a method for enhancing the frequency of recombination with family shuffling. Gene 271(1): 13-20. Greener A, Callahan M & Jerpseth B (1997) An efficient random mutagenesis technique using an E. coli mutator strain. Mol Biotechnol 7(2): 189-195. Griffiths AD & Tawfik DS (2000) Man-made enzymes - from design to in vitro compartmentalisation. Curr Opin Biotechnol 11(4): 338-353. Hiraga K & Arnold FH (2003) General method for sequence-independent site-directed chimeragenesis. J Mol Biol 330(2): 287-296. Horswill AR, Naumann TA & Benkovic SJ (2004) Using incremental truncation to create libraries of hybrid enzymes. Methods Enzymol 388: 50-60. Hoseki J, Yano T, Koyama Y, Kuramitsu S & Kagamiyama H (1999) Directed evolution of thermostable kanamycin-resistance gene: a convenient selection marker for Thermus thermophilus. J Biochem 126(5): 951-956. Hyndman D, Bauman DR, Heredia VV & Penning TM (2003) The aldo-keto reductase superfamily homepage. Chem Biol Interact 143-144: 621-631. Jaenicke R (2000) Stability and stabilization of globular proteins in solution. J Biotechnol 79(3): 193- 203. Jez JM & Penning TM (2001) The aldo-keto reductase (AKR) superfamily: an update. Chem Biol Interact 130-132(1-3): 499-525. Kawarasaki Y, Griswold KE, Stevenson JD, Selzer T, Benkovic SJ, Iverson BL & Georgiou G (2003) Enhanced crossover SCRATCHY: construction and high-throughput screening of a combinatorial library containing multiple non-homologous crossovers. Nucleic Acids Res 31(21): e126. Kim YW, Choi JH, Kim JW, Park C, Kim JW, Cha H, Lee SB, Oh BH, Moon TW & Park KH (2003) Directed evolution of Thermus maltogenic amylase toward enhanced thermal resistance. Appl Environ Microbiol 69(8): 4866-4874. Korkegian A, Black ME, Baker D & Stoddard BL (2005) Computational thermostabilization of an enzyme. Science 308(5723): 857-860. Kortemme T, Joachimiak LA, Bullock AN, Schuler AD, Stoddard BL & Baker D (2004) Computational redesign of protein-protein interaction specificity. Nat Struct Mol Biol 11(4): 371-379. Kraemer-Pecore CM, Wollacott AM & Desjarlais JR (2001) Computational protein design. Curr Opin Chem Biol 5(6): 690-695. Kunkel TA (1985) Rapid and efficient site-specific mutagenesis without phenotypic selection. Proc Natl Acad Sci USA 82: 488-492. Kurtzman AL, Govindarajan S, Vahle K, Jones JT, Heinrichs V & Patten PA (2001) Advances in directed protein evolution by recursive genetic recombination: applications to therapeutic proteins. Curr Opin Biotechnol 12(4): 361-370.

37 Chapter 2

Leemhuis H, Dijkstra BW & Dijkhuizen L (2002) Mutations converting cyclodextrin glycosyltransferase from a transglycosylase into a starch hydrolase. FEBS Lett 514(2-3): 189-192. Leemhuis H, Rozeboom HJ, Wilbrink M, Euverink GJ, Dijkstra BW & Dijkhuizen L (2003) Conversion of cyclodextrin glycosyltransferase into a starch hydrolase by directed evolution: the role of alanine 230 in acceptor subsite +1. Biochemistry 42(24): 7518-7526. Leung DW, Chen E & Goeddel DV (1989) A method for random mutagenesis of a defined DNA segment using a modified polymerase chain reaction. Technique 1: 11-15. Lin H & Cornish VW (2002) Screening and selection methods for large-scale analysis of protein function. Angew Chem Int Ed Engl 41(23): 4402-4425. Looger LL, Dwyer MA & Hellinga HW (2003) Computational design of receptor and sensor proteins with novel functions. Nature 423(6936): 185-190. Lutz S, Ostermeier M, Moore GL, Maranas CD & Benkovic SJ (2001) Creating multiple- crossover DNA libraries independent of sequence identity. Proc Natl Acad Sci USA 98(20): 11248- 11253. Ma K, Loessner H, Heider J, Johnson MK & Adams MW (1995) Effects of elemental sulfur on the metabolism of the deep-sea hyperthermophilic archaeon Thermococcus strain ES-1: characterization of a sulfur-regulated, non-heme iron alcohol dehydrogenase. J Bacteriol 177(16): 4748-4756. Mackenzie LF, Wang Q, Warren RAJ & Withers SG (1998) Glycosynthases: Mutant glycosidases for oligosaccharide synthesis. J Am Chem Soc 120: 5583-5584. Makhatadze GI, Loladze VV, Ermolenko DN, Chen X & Thomas ST (2003) Contribution of surface salt bridges to protein stability: guidelines for protein engineering. J Mol Biol 327(5): 1135- 1148. Martin A, Schmid FX & Sieber V (2003) Proside: a phage-based method for selecting thermostable proteins. Methods Mol Biol 230: 57-70. Marvin JS & Hellinga HW (2001) Conversion of a maltose receptor into a zinc biosensor by computational design. Proc Natl Acad Sci USA 98(9): 4955-4960. Masip L, Pan JL, Haldar S, Penner-Hahn JE, DeLisa MP, Georgiou G, Bardwell JCA & Collet JF (2004) An engineered pathway for the formation of protein disulfide bonds. Science 303(5661): 1185-1189. May O, Nguyen PT & Arnold FH (2000) Inverting enantioselectivity by directed evolution of hydantoinase for improved production of L-methionine. Nat Biotechnol 18(3): 317-320. Minshull J & Stemmer WPC (1999) Protein evolution by molecular breeding. Curr Opin Chem Biol 3(3): 284-290. Miyazaki K & Arnold FH (1999) Exploring nonnatural evolutionary pathways by saturation mutagenesis: rapid improvement of protein function. J Mol Evol 49(6): 716-720. Miyazaki K, Takenouchi M, Kondo H, Noro N,Suzuki M & Tsuda S (2006) Thermal stabilization of Bacillus subtilis family-11 xylanase by directed evolution. J Biol Chem 281(15): 10236-10242. Moracci M, Trincone A, Cobucci-Ponzano B, Perugino G, Ciaramella M & Rossi M (2001) Enzymatic synthesis of oligosaccharides by two glycosyl hydrolases of Sulfolobus solfataricus. Extremophiles 5(3): 145-152. Mullegger J, Jahn M, Chen HM, Warren RA & Withers SG (2005) Engineering of a thioglycoligase: randomized mutagenesis of the acid-base residue leads to the identification of improved catalysts. Protein Eng Des Sel 18(1): 33-40.

38 General introduction

Ness JE, Kim S, Gottman A, Pak R, Krebber A, Borchert TV, Govindarajan S, Mundorff EC & Minshull J (2002) Synthetic shuffling expands functional protein diversity by allowing amino acids to recombine independently. Nat Biotechnol 20(12): 1251-1255. Nielsen JE, Borchert TV & Vriend G (2001) The determinants of alpha-amylase pH-activity profiles. Protein Eng 14(7): 505-512. Nielsen JE & Borchert TV (2000) Protein engineering of bacterial α-amylases. Biochim Biophys Acta 1543(2): 253-274. Mullegger J, Jahn M, Chen HM, Warren RA & Withers SG (2005) Engineering of a thioglycoligase: randomized mutagenesis of the acid-base residue leads to the identification of improved catalysts. Protein Eng Des Sel 18(1): 33-40. Okuyama M, Mori H, Watanabe K, Kimura A & Chiba S (2002) α-Glucosidase mutant catalyzes “α-glycosynthase”-type reaction. Biosci Biotechnol Biochem 66(4): 928-933. O'Maille PE, Bakhtina M & Tsai MD (2002) Structure-based combinatorial protein engineering (SCOPE). J Mol Biol 321(4): 677-691. O'Maille PE, Tsai MD, Greenhagen BT, Chappell J & Noel JP (2004) Gene library synthesis by structure-based combinatorial protein engineering. Methods Enzymol 388: 75-91. Ostermeier M, Shim JH & Benkovic SJ (1999) A combinatorial approach to hybrid enzymes independent of DNA homology. Nat Biotechnol 17(12): 1205-1209. Ostermeier M (2003) Synthetic gene libraries: in search of the optimal diversity. Trends Biotechnol 21(6): 244-247. Oue S, Okamoto A, Yano T & Kagamiyama H (1999) Redesigning the substrate specificity of an enzyme by cumulative effects of the mutations of non-active site residues. J Biol Chem 274(4): 2344- 2349. Pan Y, Gao D, Yang W, Cho H, Yang G, Tai HH & Zhan CG (2005) Computational redesign of human butyrylcholinesterase for anticocaine medication. Proc Natl Acad Sci USA 102(46): 16656- 16661. Park DH & Plapp BV (1991) Isoenzymes of horse liver alcohol dehydrogenase active on ethanol and steroids. cDNA cloning, expression, and comparison of active sites. J Biol Chem 266(20): 13296- 13302. Park HS, Nam SH, Lee JK, Yoon CN, Mannervik B, Benkovic SJ & Kim HS (2006) Design and evolution of new catalytic activity with an existing protein scaffold. Science 311(5760): 535-538. Patnaik R, Louie S, Gavrilovic V, Perry K, Stemmer WPC, Ryan CM & Del Cardayre S (2002) Genome shuffling of Lactobacillus for improved acid tolerance. Nat Biotechnol 20(7): 707-712. Pelletier J & Sidhu S (2001) Mapping protein-protein interactions with combinatorial biology methods. Curr Opin Biotechnol 12(4): 340-347. Powell KA, Ramer SW, Del Cardayre SB, Stemmer WP, Tobin MB, Longchamp PF & Huisman GW (2001) Directed evolution and biocatalysis. Angew Chem Int Ed 40(21): 3948-3959. Reetz MT (2004) Controlling the enantioselectivity of enzymes by directed evolution: practical and theoretical ramifications. Proc Natl Acad Sci USA 101(16): 5716-5722. Reetz MT, Torre C, Eipper A, Lohmer R, Hermes M, Brunner B, Maichele A, Bocola M, Arand M, Cronin A, Genzel Y, Archelas A & Furstoss R (2004) Enhancing the enantioselectivity of an epoxide hydrolase by directed evolution. Org Lett 6(2): 177-180. Richardson TH, Tan X, Frey G, Callen W, Cabell M, Lam D, Macomber J, Short JM, Robertson DE & Miller C (2002), A novel, high performance enzyme for starch liquefaction.

39 Chapter 2

Discovery and optimization of a low pH, thermostable α-amylase. J Biol Chem 277(29): 26501- 26507. Rothman SC, Voorhies M & Kirsch JF (2004) Directed evolution relieves product inhibition and confers in vivo function to a rationally designed tyrosine aminotransferase. Protein Sci 13(3): 763-772. Saraf MC, Moore GL, Goodey NM, Cao VY, Benkovic SJ & Maranas CD (2006) IPRO: an iterative computational protein library redesign and optimization procedure. Biophys J 90(11): 4167- 4180. Schlieben NH, Niefind K, Muller J, Riebel B, Hummel W & Schomburg D (2005) Atomic resolution structures of R-specific alcohol dehydrogenase from Lactobacillus brevis provide the structural bases of its substrate and co-substrate specificity. J Mol Biol 349(4): 801-813. Schmid A, Dordick JS, Hauer B, Kiener A, Wubbolts M & Witholt B (2001) Industrial biocatalysis today and tomorrow. Nature 409(6817): 258-268. Schmidt-Dannert C & Arnold FH (1999) Directed evolution of industrial enzymes. Trends Biotechnol 17(4): 135-136. Schoemaker HE, Mink D & Wubbolts MG (2003) Dispelling the myths - Biocatalysis in industrial synthesis. Science 299(5613): 1694-1697. Seery LT, Nestor PV & FitzGerald GA (1998) Molecular evolution of the aldo-keto reductase gene superfamily. J Mol Evol 46(2): 139-146. Sieber V, Pluckthun A & Schmid FX (1998) Selecting proteins with improved stability by a phage- based method. Nat Biotechnol 16(10): 955-960. Sieber V, Martinez CA & Arnold FH (2001) Libraries of hybrid proteins from distantly related sequences. Nat Biotechnol 19(5): 456-460. Stemmer WPC (1994a) Rapid evolution of a protein in vitro by DNA shuffling. Nature 370(6488): 389-391. Stemmer WPC (1994b) DNA shuffling by random fragmentation and reassembly: in vitro recombination for molecular evolution. Proc Natl Acad Sci USA 91(22): 10747-10751. Stephanopoulos G (2002), Metabolic engineering by genome shuffling. Nat Biotechnol 20(7): 666- 668. Strickler SS, Gribenko AV, Keiffer TR, Tomlinson J, Reihle T, Loladze VV & Makhatadze GI (2006) Protein stability and surface electrostatics: a charged relationship. Biochemistry 45(9): 2761- 2766. Sugimura M, Nishimoto M & Kitaoka M (2006) Characterization of glycosynthase mutants derived from glycoside hydrolase family 10 xylanases. Biosci Biotechnol Biochem 70(5): 1210-1217. Tamakoshi M, Nakano Y, Kakizawa S, Yamagishi A & Oshima T (2001) Selection of stabilized 3- isopropylmalate dehydrogenase of Saccharomyces cerevisiae using the host-vector system of an extreme thermophile, Thermus thermophilus. Extremophiles 5(1): 17-22. Tani A, Sakai Y, Ishige T & Kato N (2000) Thermostable NADP+-dependent medium-chain alcohol dehydrogenase from Acinetobacter sp. strain M-1: purification and characterization and gene expression in Escherichia coli. Appl Environ Microbiol 66(12): 5231-5235. Tawfik DS & Griffiths AD (1998) Man-made cell-like compartments for molecular evolution. Nat Biotechnol 16(7): 652-656. Taylor SV, Kast P & Hilvert D (2001) Investigating and engineering enzymes by genetic selection. Angew Chem Int Ed Engl 40(18): 3310-3335.

40 General introduction

Trincone A, Perugino G, Rossi M & Moracci M (2000) A novel thermophilic glycosynthase that effects branching glycosylation. Bioorg Med Chem Lett 10: 365-368. Udit AK, Silberg JJ & Sieber V (2003) Sequence homology-independent protein recombination (SHIPREC). Methods Mol Biol 231: 153-163. Umeno D, Tobias AV & Arnold FH (2005) Diversifying carotenoid biosynthetic pathways by directed evolution. Microbiol Mol Biol Rev 69(1): 51-78. Van den Burg B, Vriend G, Veltman OR, Venema G & Eijsink VG (1998) Engineering an enzyme to resist boiling. Proc Natl Acad Sci USA 95(5): 2056-2060. Van den Heuvel RHH, Fraaije MW, Ferrer M, Mattevi A & Van Berkel WJH (2000) Inversion of stereospecificity of vanillyl-alcohol oxidase. Proc Nat Acad Sci 97(17): 9455-9460. Van der Oost J, Voorhorst WG, Kengen SWM, Geerling ACM, Wittenhorst V, Gueguen Y & De Vos WM (2001) Genetic and biochemical characterization of a short-chain alcohol dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus. Eur J Biochem 268: 3062-3068. Van der Veen BA, Potocki-Veronese G, Albenne C, Joucla G, Monsan P & Remaud-Simeon M (2004) Combinatorial engineering to enhance amylosucrase performance: construction, selection, and screening of variant libraries for increased activity. FEBS Lett 560(1-3): 91-97. Van der Veen BA, Skov LK, Potocki-Veronese G, Gajhede M, Monsan P & Remaud-Simeon M (2006) Increased amylosucrase activity and specificity, and identification of regions important for activity, specificity and stability through molecular evolution. FEBS Journal 273(4): 673-681. Vangnai AS, Arp DJ & Sayavedra-Soto LA (2002) Two distinct alcohol dehydrogenases participate in butane metabolism by Pseudomonas butanovora. J Bacteriol 184(7): 1916-1924. Van Lieshout J, Faijes M, Nieto J, Van der Oost J & Planas A (2004) Hydrolase and glycosynthase activity of endo-1,3-β-glucanase from the thermophile Pyrococcus furiosus. Archaea 1: 285-292. Vieille C & Zeikus GJ (2001) Hyperthermophilic enzymes: sources, uses, and molecular mechanisms for thermostability. Microbiol Mol Biol Rev 65(1): 1-43. Voigt CA, Mayo SL, Arnold FH & Wang ZG (2001) Computational method to reduce the search space for directed protein evolution. Proc Natl Acad Sci USA 98(7): 3778-3783. Wang T, Zhu H, Ma X, Ma Y & Wei D (2006) Structure-based stabilization of an enzyme: the case of penicillin acylase from Alcaligenes faecalis. Protein Pept Lett 13(2): 177-183. Wind RD, Uitdehaag JCM, Buitelaar RM, Dijkstra BW & Dijkhuizen L (1998) Engineering of cyclodextrin product specificity and pH optima of the thermostable cyclodextrin glycosyltransferase from Thermoanaerobacterium thermosulfurigenes EM1. J Biol Chem 273(10): 5771-5779. Wong TS, Tee KL, Hauer B & Schwaneberg U (2004) Sequence saturation mutagenesis (SeSaM): a novel method for directed evolution. Nucleic Acids Res 32(3): e26. Wong TS, Tee KL, Hauer B & Schwaneberg U (2005) Sequence saturation mutagenesis with tunable mutation frequencies. Anal Biochem 341(1): 187-189. Yano T, Oue S & Kagamiyama H (1998) Directed evolution of an aspartate aminotransferase with new substrate specificities. PNAS 95(10): 5511-5515. Yoon SY, Noh HS, Kim EH & Kong KH (2002) The highly stable alcohol dehydrogenase of Thermomicrobium roseum: purification and molecular characterization. Comp Biochem Physiol B Biochem Mol Biol 132(2): 415-422. Yuan L, Kurek I, English J & Keenan R (2005) Laboratory-directed protein evolution. Microbiol Mol Biol Rev 69(3): 373-392.

41 Chapter 2

Zhang JH, Dawes G & Stemmer WPC (1997) Directed evolution of a fucosidase from a galactosidase by DNA shuffling and screening. PNAS 94(9): 4504-4509. Zhang YX, Perry K, Vinci VA, Powell K, Stemmer WPC & del Cardayre SB (2002), Genome shuffling leads to rapid phenotypic improvement in bacteria. Nature 415(6872): 644-646. Zhao H & Arnold FH (1997) Combinatorial protein design: strategies for screening protein libraries. Curr Opin Struct Biol 7(4): 480-485. Zhao H, Giver L, Shao Z, Affholter JA & Arnold FH (1998) Molecular evolution by staggered extension process (StEP) in vitro recombination. Nat Biotechnol 16(3): 258-261. Zhao H (2004) Staggered extension process in vitro DNA recombination. Methods Enzymol 388: 42- 49. Zheng L, Baumann U & Reymond JL (2004) An efficient one-step site-directed and site-saturation mutagenesis protocol. Nucleic Acids Res 32(14): e115.

42

Chapter 3

Identification and heterologous production of NAD(P)-dependent alcohol dehydrogenases from the hyperthermophilic archaeon Pyrococcus furiosus

Machielsen R, Dutilh BE, Huynen MA & Van der Oost J

Parts of this chapter will be published as a review Chapter 3

Abstract

The Pyrococcus furiosus genome has been analyzed for genes encoding putative NAD(P)- dependent alcohol dehydrogenases. This has resulted in the identification of two previously characterized genes (adhA, adhB) and 16 new genes that potentially encode alcohol dehydrogenases. Because all identified genes were annotated in general terms only (Genbank), attempts were made to reach a more detailed level of function prediction by using the STRING genomic context analysis, as well as the Pfam and the COG classification. After successful production in Escherichia coli, an initial screening for activity was performed in which AdhA, AdhB and two of the putative alcohol dehydrogenases showed significant activities.

Introduction

Alcohol dehydrogenases (ADHs) catalyze the interconversion of alcohols and aldehydes or ketones. Reductive reactions result in the production of an alcohol and an oxidized cofactor (e.g. NAD+), and oxidation of an alcohol results in the production of aldehydes or ketones and a reduced cofactor (e.g. NADH). ADHs can be classified into different classes based on their cofactor specificity: (i) NAD(P), (ii) pyrollo-quinoline quinone, heme or cofactor F420, and (iii) flavin adenine dinucleotide (FAD). The NAD- or NADP-dependent alcohol dehydrogenases are generally subdivided into three major groups: short-chain ADHs, iron- activated ADHs, and the zinc-dependent ADHs (Reid & Fewson, 1994; Nordling et al., 2002; Oppermann et al., 2003; Riveros-Rosas et al., 2003). Moreover, some of the NAD(P)- dependent dehydrogenases are member of yet another group, the aldo-keto reductase (AKR) superfamily (Seery et al., 1998; Jez & Penning, 2001; Hyndman et al., 2003). Virtually all organisms contain alcohol dehydrogenases and many organisms possess multiple ADHs. They exhibit a wide spectrum of substrate specificities and perform an important role in a broad range of physiological processes (Reid & Fewson, 1994). ADHs are for example involved in alcohol and alkane metabolism, cell defense towards exogenous alcohols and aldehydes, modification of molecules (e.g. retinol, steroids), and sugar, lipid and cholesterol metabolism (Park & Plapp, 1991; Ma et al., 1995; Tani et al., 2000; Burdette et al., 2002; Vangnai et al., 2002; Yoon et al., 2002; Hoog et al., 2003). However, it is often difficult to establish their physiological function (Reid & Fewson, 1994). There exists considerable interest in stable alcohol dehydrogenases for a range of applications in food, pharmaceutical and fine chemicals industries. In particular, the production of enantio- and diastereomerically pure diols is desired because these are important chemical building blocks (Hummel, 1999; Patel, 2001; Okamura-Matsui et al., 2003; Radianingtyas & Wright, 2003). For certain applications robust biocatalysts are desired. Enzymes from hyperthermophiles, i.e., microorganisms that grow optimally above 80°C, generally display

44 Identification of ADHs from Pyrococcus furiosus an extreme stability at high temperature, high pressure, as well as high concentrations of chemical denaturants (Zeikus et al., 1998; Vieille & Zeikus, 2001; Haki & Rakshit, 2003). These features make hyperthermophilic enzymes very interesting from both scientific and industrial perspectives. The hyperthermophilic archaeon Pyrococcus furiosus grows optimally at 100°C by the fermentation of peptides and carbohydrates to produce acetate, CO2, alanine and H2, together with minor amounts of ethanol. The organism will also generate H2S if elemental sulphur is present (Fiala & Stetter, 1986; Kengen et al., 1994, 1996). Three different alcohol dehydrogenases have previously been identified in Pyrococcus furiosus. A short-chain AdhA and an iron-containing AdhB encoded by the lamA operon (Van der Oost et al., 2001), and an oxygen-sensitive, iron- and zinc-containing alcohol dehydrogenase that has been purified from cell extracts of P. furiosus (Ma & Adams, 1999). The work reported here describes the advanced analysis of the Pyrococcus furiosus genome for novel NAD(P)-dependent alcohol dehydrogenases, which resulted in the identification of sixteen genes that potentially encode alcohol dehydrogenases. For all identified genes a more detailed level of function prediction was attempted using STRING genomic context analysis, Pfam annotation and COG annotation. The genes for AdhA, AdhB and fifteen putative alcohol dehydrogenases from P. furiosus were successfully cloned and overexpressed in E. coli. Subsequently, an initial screening for activity was performed.

Materials and Methods

In silico analyses The genomes of the closely related hyperthermophilic archaea, Pyrococcus furiosus (Fiala & Stetter, 1986; Robb et al., 2001), Pyrococcus abyssi (Erauso et al., 1993; Cohen et al., 2003) and Pyrococcus horikoshii (Gonzalez et al., 1998; Kawarabayasi et al., 1998), were screened, using HMM searches, for the protein families in the Pfam database (http://www.sanger.ac.uk/software/Pfam). For every protein family, all the homologous proteins in the investigated species were listed in a web-accessible database, Pyrobase (http://www.cmbi.ru.nl/pyrobase/). Next, Pyrobase was scanned for homologs of domains relevant for the conversion between alcohols and aldehydes or ketones. A function prediction was attempted for all identified genes using Pfam annotation, COG (cluster of orthologous groups of proteins; http://www.ncbi.nlm.nih.gov/COG) annotation and STRING genomic context analysis (http://string.embl.de/).

Chemicals and plasmids All chemicals (analytical grade) were purchased from Sigma-Aldrich or Acros Organics. The restriction enzymes were obtained from Invitrogen and New England Biolabs. Pfu Turbo and T4 DNA ligase were purchased from Invitrogen and Stratagene, respectively. For

45 Chapter 3 heterologous expression the vector pET-24d (Novagen) and the tRNA helper plasmid pSJS1244 (Kim et al., 1998; Sorensen et al., 2003) were used.

Organisms and growth conditions Escherichia coli XL1 Blue (Stratagene) was used as a host for the construction of pET24d derivatives. E. coli BL21(DE3) (Novagen) harbouring the tRNA helper plasmid pSJS1244 was used as an expression host. Both strains were grown under standard conditions (Sambrook et al., 1989) following the instructions of the manufacturer.

Cloning and sequencing of the alcohol dehydrogenase encoding genes The identification of the genes encoding a putative alcohol dehydrogenase were based on automated function prediction by Pfam annotation (protein family) and COG annotation (cluster of orthologous groups of genes), and significant sequence similarity to several known alcohol dehydrogenases. The P. furiosus adh genes were identified in the P. furiosus database (http://www.genome.utah.edu). The adh genes (except PF0799) were PCR amplified from chromosomal DNA of P. furiosus using the primers shown in Table 3.1. The fragments generated were purified using Qiaquick PCR purification kit (Qiagen). The purified genes were digested with the appropriate restriction enzymes and cloned into E. coli XL1-Blue using a pET24d vector digested with the appropriate restriction enzymes. Subsequently, the resulting plasmids (pWUR72 through pWUR85, pWUR102, pWUR103 and pWUR104) were transformed into E. coli BL21(DE3) harbouring the tRNA helper plasmid pSJS1244. The sequence of the expression clones was confirmed by sequence analysis of both DNA strands.

Production of putative ADHs E. coli BL21(DE3) harboring pSJS1244 and expression plasmids containing P. furiosus genes predicted to code for ADHs were grown in 5 ml Luria-Bertani medium supplemented with 50

µg/ml kanamycin and spectinomycin until a cell density of OD600 nm = 0.6 was reached. The cultures were then induced with 0.1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) and incubation of the culture at 37 °C was continued overnight. Cells were harvested, resuspended in 20 mM Tris/HCl buffer (pH 7.8) and disrupted by sonication. The crude cell extract was centrifuged for 10 min. at 15.000 × g. The resulting supernatant (cell free extract; CFE) was heated for 30 min at 80°C and subsequently centrifuged for 10 min at 15.000 × g. The resulting supernatant (heat stable cell free extract; HSCFE) was used to screen the potential of the putative ADHs.

46 Identification of ADHs from Pyrococcus furiosus

Table 3.1 Primers used to create P. furiosus adh constructs. Gene Primer Primer sequencea PF0074 BG1394 5’-GCGCGCCATGGCAAAGGTTGCCGTAATTACTGGG BG1396 5’-GCGCGGGATCCTCAATACTCAGGTTTTTGATAAATTGAG PF0075 BG1270 5’-GCGCGCCATGGCATTTGAGATATCAATTTATCTTCCC BG1288 5’-GCGCGGGATCCTTAGTATGCTCTTCTATAGATCTCTTC PF0149 BG1271 5’-GCGCGCCATGGCATTGAAGATAGACTTATCTGGAAAG BG1289 5’-GCGCGGAATTCCTAAAATACTGAGTTTAGTCTACCTC PF0402 BG1272 5’-GCGCGCCATGGCAAAGGTATTAGTCACTGGAGGG BG1290 5’-GCGCGGGATCCTCAGGCCTTTAGAAGCTCCCTAAC PF0557 BG1273 5’-GCGCGCCATGGCACCATGCTATAACCTGAAGAATAAA BG1291 5’-GCGCGGGATCCTTAGTCTGCATATATCATTTTGACTGT PF0608 BG1274 5’-GCGCGCCATGGTTATGTTTTTCCTAAAGACTAAGA BG1292 5’-GCGCGGGATCCTCAATCTCCATAGAAGGCCCTCTC PF0765b BG1276 5’-GCGCGCCATGGCAAAGCTCCTGGGATTAAGTAGGG BG1293 5’-CGTTAGCTCCGTGGTCGGCGAA BG1275 5’-TTCGCCGACCACGGAGCTAACG BG1294 5’-GCGCGGGATCCTTAAAACACTCTCCCAACCCTTAAAT PF0771 BG1277 5’-GCGCGCCATGGCAATAAATAATGCACAACCTCCAATT BG1295 5’-GCGCGGGATCCTCACCAGCACACCCCCTCATAAAT PF0991 BG1279 5’-GCGCGCCATGGCATCCGAGAAGATGGTTGCTATCA BG1297 5’-GCGCGGGATCCTCATTTAAGCATGAAAACAACTTTGCC PF1355 BG1280 5’-GCGCGTCATGAGAGTTTCGGTTGTTGGCTCT BG1298 5’-GCGCGGGATCCTCACCAGCACACCCCCTCGTAA PF1357 BG1281 5’-GCGCGCCATGGCAATGGACCCGATAAAGTCAGATG BG1299 5’-GCGCGGGATCCTTAAATTTTCCCCCCAAACCATTTTAT PF1382 BG1282 5’-GCGCGCCATGGCACACATAATGGAGTTCCCGCGA BG1300 5’-GCGCGGGATCCTCAAATTACACCAGTGATTTTAGCAG PF1631 BG1283 5’-GCGCGCCATGGCAAGAATCCTTAAATTTTGTGTTCTAT BG1301 5’-GCGCGGTCGACTCATACGTCCCCTCTCCCCACT PF1788 BG1284 5’-GCGCGCCATGGCAAAGAATAAGTTAGTTGTTGTAACG BG1302 5’-GCGCGGAATTCTTAAAAGTATTTCCATTTTTTAAACCATTC PF1853 BG1285 5’-GCGCGCCATGGCAAACATTTTAATAACAGCTTCTTCAA BG1303 5’-GCGCGGGATCCTTAAAGAAGAACGCTCCTAGTCATTG PF1899 BG1286 5’-GCGCGCCATGGCAAAACCCATATGTGAGTTGATTAG BG1304 5’-GCGCGGGATCCCTAGGCTGACAGAAAGCCACCAT PF1960 BG1287 5’-GCGCGCCATGGCAAAAAGGGTAAATGCATTCAACGA BG1305 5’-GCGCGGGATCCTCACACACACCTCCTTGCCATCT aRestriction sites (NcoI, BspHI, BamHI, EcoRI and SalI) are underlined in the sequences. In order to introduce an NcoI restriction site often an extra alanine codon (GCA) was introduced in the gene by the forward primer (boldface in the sequences). bA two step PCR was performed to remove a NcoI site in the sequence. First, the primer pairs BG1276 / BG1293 and BG1275 / BG1294 were used to introduce the mutation (boldface in the sequences). Next, the two fragments were used as template with the primer pair BG1276 / BG1294 to obtain the whole gene.

47 Chapter 3

SDS-PAGE electrophoresis Protein composition was analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE, 10%) (Sambrook et al., 1989), using a Mini-Protean 3 system (Biorad). Protein samples for SDS-PAGE were prepared by heating for 30 min at 100°C in the presence of sample buffer (0.1 M sodium phosphate buffer, 4% SDS, 10% 2- mercaptoethanol, 20% glycerol, pH6.8). A broad range protein marker (Biorad) was used to estimate the molecular mass of the proteins.

Screening An initial screening for alcohol oxidation and/or aldehyde reduction was performed in microtiter plates (total volume 200 μl). Alcohol oxidation and aldehyde reduction were determined by following either the reduction of NAD(P)+ or the oxidation of NAD(P)H at 340 nm. Each oxidation reaction mixture contained 50 mM glycine (pH 9.5) or 50 mM Tris buffer (pH 7.4), 150 mM of alcohol and 0.28 mM NAD(P)+. The reduction reaction mixture contained 50 mM Tris buffer (pH 7.4 or pH 6.4), 150 mM aldehyde or ketone and 0.28 mM NAD(P)H. In all assays the reaction was initiated by addition of an appropriate amount of HSCFE. The heat stable cell free extract and assay buffer were mixed, and incubated at 70°C. After different time points of incubation (5 to 45 minutes) the absorbance at 340 nm was measured in each well using an iEMS reader MF (Thermo Electron Corporation, Breda, The Netherlands). The following controls were included: (i) HSCFE of E. coli BL21(DE3) harboring pSJS1244 and an empty pET24d vector, (ii) blank with no HSCFE, (iii) blank with no substrate. The temperature dependent spontaneous degradation of NAD(P)H was corrected for.

Activity assays Rates of alcohol oxidation and aldehyde reduction were also determined at 70 °C by following either the reduction of NAD(P)+ or the oxidation of NAD(P)H at 340 nm using a Hitachi U2010 spectrophotometer, with a temperature controlled cuvette holder. Each oxidation reaction mixture contained 50 mM glycine (pH 9.5), 150 mM of alcohol and 0.28 mM NAD(P)+. The reduction reaction mixture contained 50 mM Tris buffer (pH 7.4 or pH 6.4), 150 mM aldehyde or ketone and 0.28 mM NAD(P)H. In all assays the reaction was initiated by addition of an appropriate amount of HSCFE. One unit of ADH was defined as the oxidation or reduction of 1 μmol of NAD(P)H or NAD(P)+ per min, respectively. Protein concentration was determined using Bradford reagents (Bio-Rad) with bovine serum albumin as a standard (Bradford, 1976). The temperature dependent spontaneous degradation of NAD(P)H was corrected for.

48 Identification of ADHs from Pyrococcus furiosus

Results

Identification and function prediction Automated function predictions for all genes in the genomes of the related species P. abyssi (Erauso et al., 1993; Cohen et al., 2003), P. furiosus (Fiala & Stetter, 1986; Robb et al., 2001) and P. horikoshii (Gonzalez et al., 1998; Kawarabayasi et al., 1998) were carried out by classification using Pfam (protein family; http://www.sanger.ac.uk/software/Pfam) and COG (cluster of orthologous groups of proteins; http://www.ncbi.nlm.nih.gov/COG). The results were presented in a web-accessible database, Pyrobase (http://www.cmbi.ru.nl/pyrobase/). Pyrobase was scanned for homologs of domains relevant for the conversion between alcohols and aldehydes or ketones. Four promising candidate alcohol dehydrogenase families were selected, containing 17 genes in P. furiosus: iron-containing alcohol dehydrogenase (Fe- ADH), short chain dehydrogenase (adh_short), zinc-dependent dehydrogenase (adh_zinc) and UDP-glucose/GDP-mannose dehydrogenase family (UDPG_MGDP_dh). In addition, the genome of P. furiosus was searched for homologs of canonical NAD(P)-dependent alcohol dehydrogenases by BLAST search analysis (http://www.ncbi.nlm.nih.gov/BLAST/). This analysis of the P. furiosus genome for genes encoding putative alcohol dehydrogenases resulted in the identification of adhA, adhB and 16 genes that potentially encode alcohol dehydrogenases (Table 3.2). For all identified genes a more detailed level of function prediction was attempted using STRING genomic context analysis (http://string.embl.de/) and Pfam and COG annotation (Table 3.2).

Production and Screening of putative ADHs The genes for AdhA, AdhB and fifteen of the putative ADHs from P. furiosus were successfully cloned and overexpressed in Escherichia coli. From analysis of HSCFE on SDS- PAGE electrophoresis it could be estimated that over 90% of the protein content is due to the overexpression of the putative ADHs (data not shown). An initial screening for alcohol oxidation and/or aldehyde reduction by the putative ADHs was performed in microtiter plates. The oxidation reaction is known to occur often at a higher pH. So, the screening for alcohol oxidation was performed at a more neutral pH of 7.4 and a higher pH of 9.5. The screening was performed with both cofactors, NAD+ and NADP+. A range of alcohols (primary alcohols, secondary alcohols, different number of C-atoms, more than one hydroxyl group, etc.) were used in the screening: ethanol, 1-propanol, 2- propanol, 2-propen-1-ol, 3-phenyl-1-propanol, 1-hexanol, 2-hexanol, cis-4-hexenol, 6-phenyl- 1-hexanol, 1-decanol, 2-decanol, cis-4-decenol, 2,3-butanediol and glycerol. Screening for the reduction reaction was also performed at two different pH (6.4 and 7.4) and with both cofactors, NADH and NADPH. Several aldehydes and ketones were included in the screening: propanal, 2-propanone, 3-phenylpropionaldehyde, hexanal, 2-hexanone, decanal, 2-decanone.

49 Chapter 3

Table 3.2 Putative alcohol dehydrogenases in Pyrococcus furiosus Pfam domain COG Gene Plasmid Annotation PF0577 pWUR75 Alcohol dehydrogenase (EC 1.-.-.-) PF1899 pWUR84 3-oxoacyl-ACP reductase (EC 1.1.1.100) PF0149 pWUR73 Enoyl-ACP reductase (EC 1.3.1.9) COG1028 PF0074 Alcohol dehydrogenase (EC 1.1.1.1) pWUR104 (AdhA) (Van der Oost et al., 2001) adh_short PF1853 pWUR83 Glucose 1-dehydrogenase (EC 1.1.1.47) PF1788 pWUR82 UDP-glucose 4-epimerase (EC 5.1.3.2) PF0402 pWUR74 UDP-glucose 4-epimerase (EC 5.1.3.2) COG0451 dTDP-glucose 4,6-dehydratase PF1357 pWUR79 (EC 4.2.1.46) PF0991 L-threonine dehydrogenase (EC 1.1.1.103) adh_zinc COG1063 (PfTDH pWUR78 (Machielsen & Van der Oost, 2006) - AdhC) not PF0075 Alcohol dehydrogenase (EC 1.1.1.1) pWUR72 assigned (AdhB) (Van der Oost et al., 2001) Fe-ADH COG1454 PF0608 pWUR76 Alcohol dehydrogenase (EC 1.1.1.1) Glycerol-1-phosphate dehydrogenase COG0371 PF1382 pWUR80 (EC 1.1.1.261) PF0765 pWUR102 NDP-sugar dehydrogenase PF0799 - Gene fragment COG0677 N-acylmannosamine dehydrogenase PF1631 pWUR81 UDPG_ (EC 1.1.1.233) MDPG_dh UDP-glucose dehydrogenase PF0771 pWUR77 (EC 1.1.1.22) COG1004 UDP-glucose dehydrogenase PF1355 pWUR103 (EC 1.1.1.22) Aldo-keto PF1960 Alcohol dehydrogenase (EC 1.1.1.1) COG0656 pWUR85 reductase (AdhD) (Machielsen et al., 2006)

Significant activities were mainly found in the screening for alcohol oxidation. AdhA (PF0074), which has already been characterized (Van der Oost et al., 2001), showed activity with most of the substrates in the oxidation and reduction reaction. Only the cofactor NADP(H) was used by AdhA and in the oxidation reaction it showed highest activity at pH 9.5. It showed low activity towards the primary alcohols and high activity with 2,3-butanediol and in particular the secondary alcohols 2-propanol and 2-hexanol. The large number of C-

50 Identification of ADHs from Pyrococcus furiosus atoms in 2-decanol clearly reduced activity of AdhA. The reduction reaction could be performed at both pH tested. AdhB (PF0075), which was identified and purified previously (Van der Oost et al., 2001), exhibited low activities in the oxidation reaction at pH 9.5. It could only use NADP+ as cofactor and highest activities were determined with the primary alcohols. AdhB showed no activity with secondary alcohols. Previously, Van der Oost et al. observed rapid loss of AdhB activity after one day of storage at 4°C, which can account for the low AdhB activity determined in the screening. Even more interesting were the activities observed with two of the putative alcohol dehydrogenases, PF0991 and PF1960. Both of them showed activity in the oxidation reaction at pH 9.5 and only with the cofactor NAD+. For PF0991 no activity was observed with primary and secondary alcohols, but a high activity was observed with 2,3-butanediol as substrate. Furthermore, it displayed a low activity with glycerol as substrate. Another putative alcohol dehydrogenase, PF1960, showed activity towards a broad range of alcohols. It showed a clear preference for secondary alcohols and a low activity towards primary alcohols. However, the highest activity was again observed with 2,3-butanediol as substrate. The activity of PF1960 with 2,3-butanediol as substrate was the highest activity observed in the whole screening. Unfortunately, the other putative ADHs showed no significant activity in this screening.

Discussion

In hyperthermophilic bacteria and archaea, several types of alcohol dehydrogenases have been discovered. Almost all are NAD(P)-dependent (Radianingtyas & Wright, 2003). The here described analysis of the genome of the hyperthermophilic archaeon P. furiosus resulted in the identification of adhA, adhB and sixteen genes that potentially encode alcohol dehydrogenases. Previously, Ma and Adams (1999) identified an additional oxygen-sensitive, iron- and zinc-containing alcohol dehydrogenase. The amino-terminal sequence of this unusual ADH was compared to the putative ADHs identified in P. furiosus. None of the putative ADHs matched the amino-terminal sequence of this oxygen-sensitive, iron- and zinc- containing ADH. The putative ADHs show similarity with the UDP-glucose/GDP-mannose dehydrogenase family or the major groups of the NAD(P)-dependent alcohol dehydrogenases: short-chain ADHs, zinc-dependent ADHs, iron-activated ADHs and the aldo-keto reductase superfamily. The UDP-glucose/GDP-mannose dehydrogenases are a small group of enzymes which possesses the ability to catalyze the NAD-dependent 2-fold oxidation of an alcohol to an acid without the release of an aldehyde intermediate (Roychoudhury et al., 1989; Campbell et al., 1997, 2000; Lokanath et al., 2007). The other four groups catalyze the reversible conversion of either alcohols to aldehydes, or alcohols to ketones. The short-chain alcohol dehydrogenases are part of the short-chain dehydrogenases/reductases (SDR) family (Jornvall et al., 1995; Oppermann et al., 2003). These enzymes have a well conserved NAD(P)H binding site at the N-terminus, as well as several residues that are directly or indirectly

51 Chapter 3 involved in catalysis. In general they have no metal requirement (Filling et al., 2002). Zinc- dependent ADHs are member of the medium-chain dehydrogenase/reductase (MDR) superfamily (Nordling et al., 2002; Riveros-Rosas et al., 2003). These enzymes utilize NAD(P)H as cofactor, are homotetramers or homodimers and usually contain one or two zinc atom(s) per subunit with catalytic and/or structural function (Jornvall et al., 1987). The Fe- activated/containing alcohol dehydrogenases require a divalent metal ion for catalysis. This is generally Fe2+, but there are also members which contain another divalent metal ion (e.g. Zn2+). On the basis of this it has been proposed to rename this group of metal-dependent alcohol dehydrogenases (Scopes, 1983; Ruzheinikov et al., 2001; Montella et al., 2005). Members of the last group, the AKR superfamily, share a common fold and are mostly monomeric proteins that bind the nicotinamide cofactor without a Rossmann-fold motif (Rondeau et al., 1992; Hyndman et al., 2003). An improved function prediction has been provided for all identified genes. This demonstrates the broad range of physiological processes in which they can play an important role and the wide spectrum of substrate specificities. However, experimental evidence is necessary to confirm the proposed substrate specificities and actual physiological functions. Therefore an initial screening was performed in which two of the putative alcohol dehydrogenases (PF0991 and PF1960) showed high activities. These two enzymes were selected for a more detailed study (published elsewhere). The substrates and conditions tested in the initial screening were limited (e.g. pH range, anaerobic conditions, substrate and cofactor concentrations, metal requirements) and a more elaborate screening is required to gain insight into the physiological function of the other putative alcohol dehydrogenases. In this respect it is promising that all predicted alcohol dehydrogenases from P. furiosus are now incorporated in an industrial high-throughput screening platform for oxidoreductases. Within this screening facility oxidoreductases (including ADHs) are screened for the conversion of industrial-relevant substrates. Up to now this has resulted in the detection of low activities at 25°C for PF0557 with 1-phenylethanol, 1-naphtol and 1-chromanol.

Acknowledgements

This work was supported by the EU 5th framework program PYRED (QLK3-CT-2001- 01676).

52 Identification of ADHs from Pyrococcus furiosus

References

Bradford MM (1976) A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 72: 248-254. Burdette DS, Jung SH, Shen GJ, Hollingsworth RI & Zeikus JG (2002) Physiological function of alcohol dehydrogenases and long-chain (C30) fatty acids in alcohol tolerance of Thermoanaerobacter ethanolicus. Appl Environ Microbiol 68(4): 1914-1918. Campbell RE, Sala RF, Van de Rijn I & Tanner ME (1997) Properties and kinetic analysis of UDP-glucose dehydrogenase from group A Streptococci. Irreversible inhibition by UDP-chloroacetol. J Biol Chem 272(6): 3416-3422. Campbell RE, Mosimann SC, Van De Rijn I, Tanner ME & Strynadka NC (2000) The first structure of UDP-glucose dehydrogenase reveals the catalytic residues necessary for the two-fold oxidation. Biochemistry 39(23): 7012-7023. Cohen GN, Barbe V, Flament D, Galperin M, Heilig R, Lecompte O, Poch O, Prieur D, Querellou J, Ripp R, Thierry JC, Van der Oost J, Weissenbach J, Zivanovic Y & Forterre P (2003) An integrated analysis of the genome of the hyperthermophilic archaeon Pyrococcus abyssi. Mol Microbiol 47(6): 1495-1512. Erauso G, Reysenbach AL, Godfroy A, Meunier JR, Crump B, Partensky F, Baross JA, Marteinsson V, Barbier G, Pace NR & Prieur D (1993) Pyrococcus abyssi sp. nov., a new hyperthermophilic archaeon isolated from a deep-sea hydrothermal vent. Arch Microbiol 160: 338- 349. Fiala G & Stetter KO (1986) Pyrococcus furiosus sp. nov. represents a novel genus of marine heterotrophic archaebacteria growing optimally at 100oC. Arch Microbiol 145: 56-61. Filling C, Berndt KD, Benach J, Knapp S, Prozorovski T, Nordling E, Ladenstein R, Jornvall H & Oppermann U (2002) Critical residues for structure and catalysis in short-chain dehydrogenases/reductases. J Biol Chem 277(28): 25677-25684. Gonzalez JM, Masuchi Y, Robb FT, Ammerman JW, Maeder DL, Yanagibayashi M, Tamaoka J & Kato C (1998) Pyrococcus horikoshii sp. nov., a hyperthermophilic archaeon isolated from a hydrothermal vent at the Okinawa Trough. Extremophiles 2(2): 123-130. Haki GD & Rakshit SK (2003) Developments in industrially important thermostable enzymes: a review. Bioresour Technol 89(1): 17-34. Hoog JO, Stromberg P, Hedberg JJ & Griffiths WJ (2003) The mammalian alcohol dehydrogenases interact in several metabolic pathways. Chem Biol Interact 143-144: 175-181. Hummel W (1999) Large-scale applications of NAD(P)-dependent oxidoreductases: recent developments. Trends Biotechnol 17(12): 487-492. Hyndman D, Bauman DR, Heredia VV & Penning TM (2003) The aldo-keto reductase superfamily homepage. Chem Biol Interact 143-144: 621-631. Jez JM & Penning TM (2001) The aldo-keto reductase (AKR) superfamily: an update. Chem Biol Interact 130-132(1-3): 499-525. Jornvall H, Persson B & Jeffery J (1987) Characteristics of alcohol/polyol dehydrogenases. The zinc-containing long-chain alcohol dehydrogenases. Eur J Biochem 167(2): 195-201. Jornvall H, Persson B, Krook M, Atrian S, Gonzalez-Duarte R, Jeffery J & Ghosh D (1995) Short-chain dehydrogenases/reductases (SDR). Biochemistry 34(18): 6003-6013.

53 Chapter 3

Kawarabayasi Y, Sawada M, Horikawa H, Haikawa Y, Hino Y, Yamamoto S, Sekine M, Baba S, Kosugi H, Hosoyama A, Nagai Y, Sakai M, Ogura K, Otsuka R, Nakazawa H, Takamiya M, Ohfuku Y, Funahashi T, Tanaka T, Kudoh Y, Yamazaki J, Kushida N, Oguchi A, Aoki K & Kikuchi H (1998) Complete sequence and gene organization of the genome of a hyperthermophilic archaebacterium, Pyrococcus horikoshii OT3. DNA Res 5(2): 55-76. Kengen SWM, de Bok FA, Van Loo ND, Dijkema C, Stams AJM & De Vos WM (1994) Evidence for the operation of a novel Embden-Meyerhof pathway that involves ADP-dependent kinases during sugar fermentation by Pyrococcus furiosus. J Biol Chem 269(26): 17537-17541. Kengen SWM, Stams AJM & De Vos WM (1996) Sugar metabolism of hyperthermophiles. FEMS Microbiol Rev 18(2-3): 119-137. Kim R, Sandler SJ, Goldman S, Yokota H, Clark AJ & Kim SH (1998) Overexpression of archaeal proteins in Escherichia coli. Biotechnol Lett 20(3): 207-210. Lokanath NK, Pampa KJ, Kamiya T & Kunishima N (2007) Purification, crystallization and preliminary X-ray diffraction studies of a putative UDP-N-acetyl-d-mannosamine dehydrogenase from Pyrococcus horikoshii OT3. Acta Crystallogr Sect F 63(5): 412-414. Ma K, Loessner H, Heider J, Johnson MK & Adams MW (1995) Effects of elemental sulfur on the metabolism of the deep-sea hyperthermophilic archaeon Thermococcus strain ES-1: characterization of a sulfur-regulated, non-heme iron alcohol dehydrogenase. J Bacteriol 177(16): 4748-4756. Ma K & Adams MW (1999) An unusual oxygen-sensitive, iron- and zinc-containing alcohol dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus. J Bacteriol 181(4): 1163- 1170. Machielsen R, Uria AR, Kengen SWM & Van der Oost J (2006) Production and characterization of a thermostable alcohol dehydrogenase that belongs to the aldo-keto reductase superfamily. Appl Environ Microbiol 72(1): 233-238. Machielsen R & Van der Oost J (2006) Production and characterization of a thermostable L- threonine dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus. FEBS J 273(12): 2722-2729. Montella C, Bellsolell L, Perez-Luque R, Badia J, Baldoma L, Coll M & Aguilar J (2005) Crystal structure of an iron-dependent group III dehydrogenase that interconverts L-lactaldehyde and L-1,2- propanediol in Escherichia coli. J Bacteriol 187(14): 4957-4966. Nordling E, Jornvall H & Persson B (2002) Medium-chain dehydrogenases/reductases (MDR): Family characterizations including genome comparisons and active site modelling. Eur J Biochem 269(17): 4267-4276. Okamura-Matsui T, Tomoda T, Fukuda S & Ohsugi M (2003) Discovery of alcohol dehydrogenase from mushrooms and application to alcoholic beverages. J Mol Catal B: Enzymatic 23: 133-144. Oppermann U, Filling C, Hult M, Shafqat N, Wu X, Lindh M, Shafqat J, Nordling E, Kallberg Y, Persson B & Jornvall H (2003) Short-chain dehydrogenases/reductases (SDR): the 2002 update. Chem Biol Interact 143-144: 247-253. Park DH & Plapp BV (1991) Isoenzymes of horse liver alcohol dehydrogenase active on ethanol and steroids. cDNA cloning, expression, and comparison of active sites. J Biol Chem 266(20): 13296- 13302. Patel RN (2001) Biocatalytic synthesis of intermediates for the synthesis of chiral drug substances. Curr Opin Biotechnol 12(6): 587-604.

54 Identification of ADHs from Pyrococcus furiosus

Radianingtyas H & Wright PC (2003) Alcohol dehydrogenases from thermophilic and hyperthermophilic archaea and bacteria. FEMS Microbiol Rev 27(5): 593-616. Reid MF & Fewson CA (1994) Molecular characterization of microbial alcohol dehydrogenases. Crit Rev Microbiol 20(1): 13-56. Riveros-Rosas H, Julian-Sanchez A, Villalobos-Molina R, Pardo JP & Pina E (2003) Diversity, taxonomy and evolution of medium-chain dehydrogenase/reductase superfamily. Eur J Biochem 270(16): 3309-3334. Robb FT, Maeder DL, Brown JR, DiRuggiero J, Stump MD, Yeh RK, Weiss RB & Dunn DM (2001) Genomic sequence of hyperthermophile, Pyrococcus furiosus: implications for physiology and enzymology. Methods Enzymol 330: 134-157. Rondeau JM, Tete-Favier F, Podjarny A, Reymann JM, Barth P, Biellmann JF & Moras D (1992) Novel NADPH-binding domain revealed by the crystal structure of aldose reductase. Nature 355(6359): 469-472. Roychoudhury S, May TB, Gill JF, Singh SK, Feingold DS & Chakrabarty AM (1989) Purification and characterization of guanosine diphospho-D-mannose dehydrogenase. A key enzyme in the biosynthesis of alginate by Pseudomonas aeruginosa. J Biol Chem 264(16): 9380-9385. Ruzheinikov SN, Burke J, Sedelnikova S, Baker PJ, Taylor R, Bullough PA, Muir NM, Gore MG & Rice DW (2001) Glycerol dehydrogenase: structure, specificity, and mechanism of a family III polyol dehydrogenase. Structure 9(9): 789-802. Sambrook J, Fritsch EF & Maniatis T (1989) Molecular cloning: A laboratory manual, 2nd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Scopes RK (1983) An iron-activated alcohol dehydrogenase. FEBS Lett 156(2): 303-306. Seery LT, Nestor PV & FitzGerald GA (1998) Molecular evolution of the aldo-keto reductase gene superfamily. J Mol Evol 46(2): 139-146. Sorensen HP, Sperling-Petersen HU & Mortensen KK (2003) Production of recombinant thermostable proteins expressed in Escherichia coli: completion of protein synthesis is the bottleneck. J Chromatogr B Analyt Technol Biomed Life Sci 786(1-2): 207-214. Tani A, Sakai Y, Ishige T & Kato N (2000) Thermostable NADP+-dependent medium-chain alcohol dehydrogenase from Acinetobacter sp. strain M-1: purification and characterization and gene expression in Escherichia coli. Appl Environ Microbiol 66(12): 5231-5235. Van der Oost J, Voorhorst WG, Kengen SWM, Geerling ACM, Wittenhorst V, Gueguen Y & De Vos WM (2001) Genetic and biochemical characterization of a short-chain alcohol dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus. Eur J Biochem 268(10): 3062-3068. Vangnai AS, Arp DJ & Sayavedra-Soto LA (2002) Two distinct alcohol dehydrogenases participate in butane metabolism by Pseudomonas butanovora. J Bacteriol 184(7): 1916-1924. Vieille C & Zeikus GJ (2001) Hyperthermophilic enzymes: sources, uses, and molecular mechanisms for thermostability. Microbiol Mol Biol Rev 65(1): 1-43. Yoon SY, Noh HS, Kim EH & Kong KH (2002) The highly stable alcohol dehydrogenase of Thermomicrobium roseum: purification and molecular characterization. Comp Biochem Physiol B Biochem Mol Biol 132(2): 415-422. Zeikus JG, Vieille C & Savchenko A (1998) Thermozymes: biotechnology and structure-function relationships. Extremophiles 2(3): 179-183.

55

Chapter 4

Production and characterization of a thermostable alcohol dehydrogenase that belongs to the aldo-keto reductase superfamily

Machielsen R, Uria AR, Kengen SWM & Van der Oost J

Applied and Environmental Microbiology (2006) 72(1): 233-238 Chapter 4

Abstract

The gene encoding a novel alcohol dehydrogenase that belongs to the aldo-keto reductase superfamily has been identified in the hyperthermophilic archaeon Pyrococcus furiosus. The gene, referred to as adhD, was functionally expressed in Escherichia coli and subsequently purified to homogeneity. The enzyme has a monomeric conformation with a molecular mass of 32 kDa. The catalytic activity of the enzyme increases up to 100°C, and a half life value of 130 min at this temperature indicates its high thermostability. AdhD exhibits a broad substrate specificity, with in general a preference for the reduction of ketones (pH optimum 6.1), and the oxidation of secondary alcohols (pH optimum 8.8). Maximal specific activities were detected with 2,3-butanediol (108.3 U/mg) and diacetyl / acetoin (22.5 U/mg) in the oxidative and reductive reactions, respectively. GC-analysis indicated that AdhD mainly produced (S)- 2-pentanol (ee 89%) when 2-pentanone was used as substrate. The physiological role of AdhD is discussed.

Introduction

Alcohol dehydrogenases (ADHs) are present in all organisms. They display a wide variety of substrate specificities and play an important role in a broad range of physiological processes. However, it is often difficult to prove their physiological function (Reid & Fewson, 1994). Alcohol dehydrogenases can be divided into different classes based on their cofactor specificity: (i) NAD(P), (ii) pyrrolo-quinoline quinone, heme or cofactor F420 and (iii) FAD. The NAD- or NADP-dependent alcohol dehydrogenases are often subdivided into three major groups: the zinc-dependent ADHs, short-chain ADHs and iron-activated ADHs (Reid & Fewson, 1994). In addition, some of the NAD(P)-dependent dehydrogenases are related to yet another group, the aldo-keto reductase (AKR) superfamily (Seery et al., 1998; Jez & Penning, 2001; Hyndman et al., 2003). This enzyme superfamily consists of NAD(P)-dependent oxidoreductases, which share a common fold and are mostly monomeric proteins that bind nicotinamide cofactor without a Rossmann-fold motif (Wierenga et al., 1986; Borhani et al., 1992; Rondeau et al., 1992; Khurana et al., 1998). There is considerable interest in the use of stable alcohol dehydrogenases in food, pharmaceutical and fine chemicals industries for the production of aldehydes, ketones and chiral alcohols. The production of chiral synthons is particularly desired because this is an increasingly important step in the synthesis of chirally pure pharmaceutical agents (Zeikus et al., 1998; Hummel, 1999; Ellis, 2002; Radianingtyas & Wright, 2003). For certain applications robust biocatalysts are desired. Enzymes from hyperthermophiles, i.e., microorganisms that grow optimally above 80°C, generally display an extreme stability at high temperature, high pressure, as well as high concentrations of chemical denaturants

58 Thermostable ADH of the AKR superfamily

(Vieille & Zeikus, 2001). These features make hyperthermophilic enzymes very interesting from both scientific and industrial perspectives. The hyperthermophilic archaeon Pyrococcus furiosus grows optimally at 100°C by the fermentation of peptides and carbohydrates to produce acetate, CO2, alanine and H2, together with minor amounts of ethanol. The organism will also generate H2S if elemental sulphur is present (Fiala & Stetter, 1986; Kengen et al., 1994, 1996). Three different alcohol dehydrogenases have previously been identified in Pyrococcus furiosus. A short-chain AdhA and an iron-containing AdhB encoded by the lamA operon (Van der Oost et al., 2001), and an oxygen-sensitive, iron- and zinc-containing alcohol dehydrogenase that has been purified from cell extracts of P. furiosus (Ma & Adams, 1999). By careful analysis of the P. furiosus genome, sixteen additional genes have been identified that potentially encode alcohol dehydrogenases (Machielsen, unpublished results). The work reported here describes the functional production of one of the newly identified alcohol dehydrogenases, AdhD, in Escherichia coli. The enzyme was purified to homogeneity, and characterized with respect to substrate specificity, kinetics and stability. Since AdhD was found to exhibit high thermostability, high enantioselectivity and broad substrate specificity, it is an attractive candidate for industrial utilization.

Materials and Methods

Chemicals and plasmids All chemicals (analytical grade) were purchased from Sigma-Aldrich or Acros Organics. The restriction enzymes were obtained from Invitrogen and New England Biolabs. Pfu Turbo and T4 DNA ligase were purchased from Invitrogen and Stratagene, respectively. For heterologous expression the vector pET-24d (Novagen) and the tRNA helper plasmid pSJS1244 (Kim et al., 1998; Sorensen et al., 2003) were used.

Organisms and growth conditions Escherichia coli XL1 Blue (Stratagene) was used as a host for the construction of pET24d derivatives. E. coli BL21(DE3) (Novagen) harbouring the tRNA helper plasmid pSJS1244 was used as an expression host. Both strains were grown under standard conditions (Sambrook et al., 1989) following the instructions of the manufacturer.

Cloning and sequencing of the alcohol dehydrogenase encoding gene The identification of the gene encoding an alcohol dehydrogenase was based on significant sequence similarity to several known alcohol dehydrogenases. The P. furiosus adhD gene (PF1960, GenBank accession number AE010289 region: 7356 - 8192, NCBI) was identified in the P. furiosus database (http://www.genome.utah.edu). The adhD gene (837 bp) was PCR amplified from chromosomal DNA of P. furiosus using the primers BG1287 (5’- GCGCGCCATGGCAAAAAGGGTAAATGCATTCAACGA, sense) and BG1305 (5’-

59 Chapter 4

GCGCGGGATCCTCACACACACCTCCTTGCCATCT, antisense), containing NcoI and BamHI sites (underlined in the sequences). In order to introduce an NcoI restriction site an extra alanine codon (GCA) was introduced in the adhD gene by the forward primer BG 1287 (bold in the sequence). The fragment generated was purified using QIAquick PCR purification kit (Qiagen). The purified gene was digested with NcoI/BamHI and cloned into E. coli XL1-Blue using an NcoI/BamHI-digested pET24d vector. Subsequently, the resulting plasmid pWUR85 was transformed into E. coli BL21(DE3) harbouring the tRNA helper plasmid pSJS1244. The sequence of the expression clone was confirmed by sequence analysis of both DNA strands.

Production and purification of ADH E. coli BL21(DE3) harbouring pSJS1244 was transformed with pWUR85 and a single colony was used to inoculate 5 ml Luria-Bertani medium with kanamycin and spectinomycin (both 50 μg⋅ml-1) and incubated overnight in a rotary shaker at 37 °C. Next, 1 ml of the preculture was used to inoculate 1 L Luria-Bertani medium with kanamycin and spectinomycin (both 50 mg⋅L-1) in a 2-L conical flask and incubated in a rotary shaker at 37 °C until a cell density of

OD600 nm = 0.6 was reached. The culture was then induced with 0.2 mM isopropyl thio-β-D- galactoside (IPTG) and incubation of the culture was continued at 37 °C for 18 h. Cells were harvested, resuspended in 20 mM Tris/HCl buffer (pH 7.5) and passed twice through a French press at 110 MPa. The crude cell extract was centrifuged for 20 min at 10.000 × g. The resulting supernatant (cell free extract) was heated for 30 min at 80°C and subsequently centrifuged for 20 min at 10.000 × g. The supernatant (heat stable cell free extract) was filtered (0.45 μm) and applied to a Q-sepharose high performance (Amersham Biosciences) column (1.6 x 10 cm) equilibrated in 20 mM Tris/HCl buffer (pH 7.8). Proteins were eluted with a linear 560 ml gradient from 0.0 to 1.0 M NaCl, in the same buffer.

Size-exclusion chromatography Molecular mass was determined by size-exclusion chromatography on a Superdex 200 HR 10/30 column (24 ml, Amersham Biosciences) equilibrated in 50 mM Tris/HCl (pH 7.8) containing 100 mM NaCl. Two hundred fifty μl of enzyme solution in 20 mM Tris/HCl buffer (pH 7.8), containing 872 μg enzyme, was injected on the column. Proteins used for calibration were Blue dextran 2000 (>2000 kDa), aldolase (158 kDa), bovine serum albumin (67 kDa), ovalbumin (43 kDa), chymotrypsinogen (25 kDa) and ribonuclease A (13.7 kDa).

SDS-PAGE electrophoresis Protein composition was analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE, 10%) (Sambrook et al., 1989), using a Mini-Protean 3 system (Biorad). Protein samples for SDS-PAGE were prepared by heating for 30 min at 100°C in the presence of sample buffer (0.1 M sodium phosphate buffer, 4% SDS, 10% 2-

60 Thermostable ADH of the AKR superfamily mercaptoethanol, 20% glycerol, pH6.8). A broad range protein marker (Biorad) was used to estimate the molecular mass of the proteins.

Activity assays Rates of alcohol oxidation and aldehyde reduction were determined at 70 °C, unless stated otherwise, by following either the reduction of NAD+ or the oxidation of NADH at 340 nm using a Hitachi U2010 spectrophotometer, with a temperature controlled cuvette holder. Each oxidation reaction mixture contained 50 mM glycine (pH 8.8), 100 mM of alcohol and 0.28 mM NAD+. The reduction reaction mixture contained 0.1 M sodium phosphate buffer (pH 6.1), 100 mM aldehyde or ketone and 0.28 mM NADH. In all assays the reaction was initiated by addition of an appropriate amount of enzyme. One unit of ADH was defined as the oxidation or reduction of 1 μmol of NADH or NAD+ per min, respectively. Protein concentration was determined using Bradford reagents (Bio-Rad) with bovine serum albumin as a standard (Bradford, 1976). The temperature dependent spontaneous degradation of NADH was corrected for. pH optimum The pH optimum for alcohol oxidation was determined in a sodium phosphate buffer (100 mM, pH range 5.4 – 7.9) and a glycine buffer (50 mM, pH-range 7.9 – 10.3), whereas the pH optimum for aldehyde reduction was determined in a sodium phosphate buffer (100 mM, pH range 5.4 – 7.9). The pH of the buffers was set at 25°C and temperature corrections were made using their temperature coefficients (-0.025 pH/°C for glycine buffer and -0.0028 pH/°C for sodium phosphate buffer).

Optimum temperature and thermostability The thermostability of AdhD (enzyme concentration: 0.17 mg·mL-1 in 20 mM Tris buffer pH 7.8) was determined by measuring the residual activity (2,3-butanediol oxidation according to the standard assay) after incubation of a time series at 100°C. The temperature optimum was determined in 50 mM glycine buffer pH 8.8 by analysis of initial rates of 2,3-butanediol oxidation in the range of 30 -100°C.

Kinetics

The AdhD kinetic parameters Km and Vmax were calculated from multiple measurements (at least 8 measurements) using the Michaelis-Menten equation and the program Tablecurve (Tablecurve 2D, version 5.0). All the reactions followed Michaelis-Menten type kinetics. The -1 turnover number (kcat, s ) was calculated as: Vmax ∗ subunit molecular mass (32 kDa) / 60.

Salts, metals and inhibitors The effect of several salts, metals (K+, Mg2+, Mn2+, Na+, Fe2+, Fe3+, Li2+, Ni2+, Co2+, Zn2+, Ca2+) and inhibitors (Ethylenediaminetetraacetic acid, dithiothreitol, 2-iodoacetamide) on the

61 Chapter 4 initial activity of AdhD was checked using 2,3-butandiol as substrate in the oxidation reaction and acetoin in the reduction reaction. Concentrations ranging from 1 - 25 mM were tested.

Enantioselectivity The enantioselectivity of AdhD was determined by reduction of 2-pentanone with cofactor regeneration at 60°C for 24 h. The reaction mixture contained 1.0 mM NAD+ or NADP+, 250 mM 2-pentanone, 300 mM glucose, 350 mM CaCO3, 10 U glucose dehydrogenase from Bacillus megaterium (for cofactor regeneration) and 4 nmol AdhD in 2 ml 50 mM Tris-HCl buffer (pH 7.0). Chiral gas chromatography was used to determine the enantiomeric excess (ee). All the samples were extracted with chloroform (1:1) and 1 μl was applied on a Chirasil Dex Column (injector temperature was 250°C).

Results

Analysis of nucleotide and amino acid sequences The P. furiosus genome has been analyzed for genes encoding putative alcohol dehydrogenases, which resulted in the identification of sixteen genes that potentially encode alcohol dehydrogenases. After successful production in E.coli, an initial screening for activity was performed in which two of the putative alcohol dehydrogenases, including AdhD, showed high activities (Machielsen, unpublished results). The two enzymes were selected for a more detailed study; with respect to the other putative alcohol dehydrogenases, a more elaborate screening is currently being performed to get insight in their substrate specificity, and possibly in their physiological function. Here we describe the production and characterization of the novel alcohol dehydrogenase, AdhD (PF1960). Analysis of the adhD locus revealed that the start codon of a gene coding for a tungsten-containing aldehyde reductase, wor4 (PF1961), is directly adjacent to the adhD stop codon (Roy & Adams, 2002), suggesting an operon-like organization. Interestingly, conserved context analysis with STRING (http://string.embl.de/) reveals that the clustering of these two genes is also observed in the related species Pyrococcus abyssi and Pyrococcus horikoshii; manual inspection identified a similar conserved gene pair in Thermococcus kodakaraensis. The adhD gene encodes a protein of 278 amino acids and a calculated molecular mass of 31.794 kDa. The sequence belongs to the COG0656 (“aldo-keto reductases; related to diketogulonate reductase”; http://www.ncbi.nlm.nih.gov/COG/). BLAST-P analysis (http://www.ncbi.nlm.nih.gov/blast/) reveals the highest similarity with hypothetical oxidoreductases and putative members of the aldo-keto reductase superfamily from hyperthermophilic archaea and bacteria. Some of these most significant hits of a BLAST search analysis were a hypothetical oxidoreductase/aldo-keto reductase of Thermococcus kodakaraensis KOD1 (85% identity, TK0845), a putative 2,5-diketo-D-gluconic acid reductase of Pyrococcus abyssi (84% identity, PAB2329), a putative dehydrogenase of

62 Thermostable ADH of the AKR superfamily

Sulfolobus solfataricus P2 (53% identity, SSO2779) and an oxidoreductase/aldo-keto reductase of Thermotoga maritima (47% identity, TM1743). Together with an aldose reductase of Sus scrofa (pig; Rondeau et al., 1992; Urzhumtsev et al., 1997) for which a structure has been determined (32% identity, ALR2, pdb: 1AH4) an alignment was made

(Fig. 4.1). The AKR superfamily shares a common (α/β)8-barrel fold, a catalytic tetrad (Asp, Tyr, Lys and His) and their members bind the cofactor NAD(P)H in an extended conformation without a Rossmann fold motif. Highly conserved residues within the AKR superfamily are indicated with an asterisk (Fig. 4.1) and they are presumably involved in the three shared properties: catalytic tetrad (Asp58, Tyr63, Lys89 and His121), cofactor binding (Asp58, Ser152, Asn153, Gln175, Glu257, Asn258) and the remaining conserved residues probably play a structural role (P. furiosus numbering) (Wilson et al., 1992; Sanli & Blaber, 2001; Jez et al., 1997ab; Ellis, 2002).

Figure 4.1 Multiple sequence alignment of the P. furiosus AdhD with (hypothetical) members of the aldo-keto reductase superfamily. The following symbols are used: Pyrfu, P. furiosus; Theko, T. kodakaraensis; Pyrab, P. abyssi; Sulso, S. solfataricus; Thema, T. maritima; Suscr, S. scrofa. The sequences were aligned using the CLUSTAL program. Asterisks indicate highly conserved residues within the AKR superfamily.

Purification of recombinant AdhD The pyrococcal AdhD was purified to homogeneity from heat-treated cell-free extracts of E. coli BL21(DE3)/pSJS1244/pWUR85 by anion-exchange chromatography. Active AdhD eluted between 0.35 and 0.52 M NaCl (peak at 0.41 M NaCl). Fractions containing the purified enzyme were pooled. The migration of AdhD on SDS/PAGE confirms a molecular subunit mass of 32 kDa (Fig. 4.2). The molecular mass of the native AdhD was estimated to be 30 kDa by size-exclusion chromatography, which indicated a monomeric structure.

63 Chapter 4

kDa 1 2 3 4 200 116.25 97.4 66.2

45

AdhD 31

Figure 4.2 SDS-PAGE analysis of heterologous produced AdhD in each purification step. Lane 1, broad range protein marker; lane 2, cell free extract (CFE); lane 3, heat-stable cell free extract (HSCFE); and lane 4, pooled Q-sepharose fractions.

Substrate and cofactor specificity The substrate specificity of AdhD in the oxidation reaction was analyzed using a range of alcohols, including primary alcohols (methanol to dodecanol, C1-C12), secondary alcohols (2- propanol to 2-decanol, C3-C10), aromatic alcohols and alcohols containing more than one hydroxyl group (Table 4.1). AdhD showed low activity towards primary alcohols and a broad range of secondary alcohols could be oxidized, this included substrates with 3-10 carbon atoms (Fig. 4.3). The highest specific activity of AdhD in the oxidative reaction was found -1 with 2,3-butanediol (Vmax 108.3 U· mg ). The substrate specificity of the reduction reaction was analyzed by using aldehydes, ketones and aldoses as substrate (Table 4.2). The highest reduction rate of AdhD was -1 observed with diacetyl / acetoin (3-hydroxy-2-butanone, Vmax 22.5 U· mg ). AdhD could use both NAD(H) and NADP(H) as cofactor, but has a clear preference for NAD(H). Measurements with 2,3-butanediol as substrate showed that the activity with NADP+ as cofactor was only 14% of that found with NAD+ as cofactor. Measurements with other substrates exhibited different ratios, but always a preference for NAD(H).

Enantioselectivity The enantioselectivity of AdhD was tested using 2-pentanone as substrate and after 24 hours of conversion the products formed were measured by GC analysis. This enzyme preferably reduced 2-pentanone to (S)-2-pentanol. When NAD+ was used as cofactor an ee-value of 89.4% was obtained and with NADP+ an ee-value of 84.8%.

64 Thermostable ADH of the AKR superfamily

2.5

2

1.5

1

0.5

specific activity (U/mg) 0 12345678910 number C-atoms primary alcohol secondary alcohol

Figure 4.3 Specific activity of P. furiosus AdhD towards primary and secondary alcohols.

Table 4.1 Substrate specificity of P. furiosus Table 4.2. Substrate specificity of P. furiosus AdhD in the oxidation reaction AdhD in the reduction reaction Relative Relative Substratea activity (%) Substratea activity (%) 1-Pentanol 1 Acetone 0.1 2-Pentanol 9 Dihydroxyacetone 36 1,2-Propanediol 3 Dihydroxyacetonephosphate 82a Glycerol 1 Hexanal 16 3-Phenyl-1-propanol 1 2-Hexanone 0.2 2-Methyl-1-propanol 2 Decanal 1 2-Propen-1-ol 1 Acetoin 100 1-butanol 1 Diacetyl ∼150b 2-butanol 7 DL-Glyceraldehyde 31 1,2-Butanediol 2 Pyruvic aldehyde 81 1,3-Butanediol 3 D-Arabinose 25 2,3-Butanediol 100 D-Xylose 2 Acetoin 18 D-glucose 0 3-Phenyl-1-butanol 0.4 D-Ribose 2 3-Methyl-1-butanol 0.4 D-Mannose 2 D-Arabinose 66 Cyclohexanone 38 L-Arabinose 0 a Substrates were present in 100 mM D-Xylose 3 concentrations except for D-Glucose 2 dihydroxyacetonephosphate which had a Cyclohexanol 12 concentration of 50 mM. Benzylalcohol 3 b Due to experimental difficulties no kinetic a Substrates were present in 100 mM data could be obtained and maximal activity concentrations. was not reached.

65 Chapter 4

Effect of salts, metals and inhibitors The salts, metals and inhibitors tested caused no significant inhibition or activation. In the oxidation reaction only high concentrations of Mg2+ caused slightly lower activities and in the reduction reaction 2 mM of Ni2+, Co2+ and Ca2+ had the same small inhibitory effect. There is no inhibition when DTT, EDTA or 2-iodoacetamide is added to the reaction, confirming that the protein does not require metals for its activity and that no essential disulfide bridges or thiol groups are involved.

Thermostability and pH optima The oxidation reaction catalyzed by AdhD showed a pH optimum of 8.8 and the aldehyde reduction by AdhD showed a high level of activity over a wide range of pH with maximal activity at pH 6.1. The reaction rate of AdhD increased with increasing temperature from 30°C (6.3 U/mg) up to 100°C (38.1), but due to instability of the cofactors at that temperature all other activity measurements were performed at 70°C. At this temperature the activity was only 15% lower than at 100°C. AdhD has a high resistance to thermal inactivation, which was shown by a half-life value of 130 minutes at 100°C.

Enzyme kinetics The kinetic properties of AdhD were determined for the substrates that were converted with relatively high rates in the oxidation and reduction reaction, as well as for the cofactors used in these reactions. This showed that AdhD has a relative high affinity for acetoin (Km 6.5 mM, -1 -1 Vmax 22.5 U/mg, kcat/Km 1.8 s • mM ) and NADH (Km 97 µM, Vmax 22.5 U/mg) in the reduction reaction and clearly a lower affinity for 2,3-butanediol (Km 86.8 mM, Vmax 108.3 -1 -1 + U/mg, kcat/Km 0.7 s • mM ) and NAD (Km 600 µM, Vmax 108.3 U/mg) in the oxidation reaction.

Discussion

In (hyper)thermophilic archaea, several types of alcohol dehydrogenases have been discovered. Almost all are NAD(P)-dependent and are classified into three different groups, the zinc-dependent ADHs, the short-chain ADHs and the iron-activated ADHs (Radianingtyas & Wright, 2003). The enzyme described here, AdhD, represents a novel NAD(P)H-dependent alcohol dehydrogenase from Pyrococcus furiosus that does not show similarity with these groups, but rather with the aldo-keto reductase superfamily. AdhD was functionally produced in E. coli and due to its stability at high temperature only two steps were needed for purification. It showed a preference for NAD(H) as cofactor and had a broad substrate specificity in the oxidation and reduction reaction. A preference was observed for secondary alcohols when compared with the primary alcohols, but highest activities were detected when polyols were used as substrate.

66 Thermostable ADH of the AKR superfamily

Alcohol dehydrogenases can be involved in a wide range of metabolic processes, which makes it often difficult to determine their physiological role. The presence of the wor4 gene, encoding a tungsten-containing aldehyde reductase (Roy & Adams, 2002), directly adjacent to the adhD gene suggests a role in which AdhD might use or produce aldehydes/ ketones associated with the WOR4 activity. The fact that the clustering of these two genes is also observed in the related species P. abyssi, P. horikoshii and Thermococcus kodakaraensis strengthens this suggestion. WOR4 was purified from P. furiosus cells that were grown in the presence of elemental sulphur (S0) (Roy & Adams, 2002), and DNA microarray analyses showed that wor4 and adhD were both moderately upregulated, respectively three- and fourfold, in maltose grown P. furiosus when S0 was present (Schut et al., 2001). This is in contrast to an iron alcohol dehydrogenase from Thermococcus strain ES-1 (ES-1 ADH), which is down-regulated by the presence of S0 (Ma et al., 1995). It was proposed that, under S0 limitation, ES-1 ADH reduces (toxic) aldehydes that are generated by fermentation, thereby disposing some of the excess reductant as alcohol. This function is also assigned to an iron/zinc-containing alcohol dehydrogenase from P. furiosus (Ma & Adams, 1999). Kinetic data showed high catalytic efficiency for the reduction reaction and high affinity for the substrate and cofactor involved in this reaction, which suggest that the physiological function of AdhD is the reduction of aldehydes or ketones. Although it cannot be ruled out that these substrates are formed by WOR4, this appears unlikely because all described tungsten-containing aldehyde oxidoreductases (AOR, FOR, GAPOR) are known to catalyze the unidirectional oxidation of aldehydes (Mukund & Adams, 1991, 1995; Heider et al., 1995; Johnson et al., 1996; Roy et al., 1999). Assuming a physiological link between AdhD and WOR4, based on the conserved context, this gives rise to two other options. First of all it is conceivable that AdhD oxidizes an alcohol to produce an aldehyde or ketone which is subsequently oxidized further by WOR4. However because of the apparent preference of AdhD for the reduction reaction and WOR4 for the oxidation reaction, another putative scenario can be envisaged as well. It might be possible that an unidentified aldolase produces the substrates for AdhD and WOR4, for instance a ketone and an aldehyde. Subsequently, the ketone could be reduced to an alcohol by AdhD and the aldehyde could be oxidized by WOR4. However, since no activity has yet been ascribed for WOR4, more experiments are required to establish the physiological role of both enzymes. Apart from the scientific interest in the physiological role of alcohol dehydrogenases, there is also interest from an industrial perspective. AdhD can be produced in relatively high amounts by heterologous expression in E. coli and if necessary it can be easily purified. It is extremely thermostable, which is shown by a half-life value of 130 minutes at 100°C, and it is highly S-enantioselective. These properties together with a broad substrate specificity and a preference for the more inexpensive cofactor NAD(H) make this enzyme a potential catalyst for industry, especially for the production of chiral compounds. Further study to test AdhD for the production of interesting chiral compounds is in progress.

67 Chapter 4

Acknowledgements

This work was supported by the EU 5th framework program PYRED (QLK3-CT-2001- 01676). We thank H. Hennemann and T. Dauβmann (Jülich Fine Chemicals, Germany) for enantioselectivity analysis.

References

Borhani DW, Harter TM & Petrash JM (1992) The crystal structure of the aldose reductase•NADPH binary complex. J Biol Chem 267: 24841-24847. Bradford MM (1976) A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 72: 248-254. Ellis EM (2002) Microbial aldo-keto reductases. FEMS Microbiol Lett 216: 123-131. Fiala G & Stetter KO (1986) Pyrococcus furiosus sp. nov. represents a novel genus of marine heterotrophic archaebacteria growing optimally at 100oC. Arch Microbiol 145: 56-61. Heider J, Ma K & Adams MWW (1995) Purification, characterization, and metabolic function of tungsten-containing aldehyde ferredoxin oxidoreductase from the hyperthermophilic and proteolytic archaeon Thermococcus strain ES-1. J Bacteriol 177: 4757-4764. Hummel W (1999) Large-scale applications of NAD(P)-dependent oxidoreductases: recent developments. Trends Biotechnol 17: 487-492. Hyndman D, Bauman DR, Heredia VV & Penning TM (2003) The aldo-keto reductase superfamily homepage. Chem Biol Interact 143-144: 621-631. Jez JM, Bennett MJ, Schlegel BP, Lewis M & Penning TM (1997a) Comparative anatomy of the aldo-keto reductase superfamily. Biochem J 326 (Pt 3): 625-636. Jez JM, Flynn TG & Penning TM (1997b) A new nomenclature for the aldo-keto reductase superfamily. Biochem Pharmacol 54: 639-647. Jez JM & Penning TM (2001) The aldo-keto reductase (AKR) superfamily: an update. Chem Biol Interact 130-132: 499-525. Johnson MK, Rees DC & Adams MWW (1996) Tungstoenzymes. Chem Rev 96: 2817-2840. Kengen SWM, De Bok FA, Van Loo ND, Dijkema C, Stams AJM & De Vos WM (1994) Evidence for the operation of a novel Embden-Meyerhof pathway that involves ADP-dependent kinases during sugar fermentation by Pyrococcus furiosus. J Biol Chem 269: 17537-17541. Kengen SWM, Stams AJM & De Vos WM (1996) Sugar metabolism of hyperthermophiles. FEMS Microbiol Rev 18: 119-137. Khurana S, Powers DB, Anderson S & Blaber M (1998) Crystal structure of 2,5-diketo-D-gluconic acid reductase A complexed with NADPH at 2.1-A resolution. PNAS 95: 6768-6773. Kim R, Sandler SJ, Goldman S, Yokota H, Clark AJ & Kim SH (1998) Overexpression of archaeal proteins in Escherichia coli. Biotechnol Lett 20: 207-210. Ma K, Loessner H, Heider J, Johnson MK & Adams MWW (1995) Effects of elemental sulfur on the metabolism of the deep-sea hyperthermophilic archaeon Thermococcus strain ES-1: characterization of a sulfur-regulated, non-heme iron alcohol dehydrogenase. J Bacteriol 177: 4748- 4756.

68 Thermostable ADH of the AKR superfamily

Ma K & Adams MWW (1999) An unusual oxygen-sensitive, iron- and zinc-containing alcohol dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus. J Bacteriol 181: 1163- 1170. Mukund S & Adams MWW (1991) The novel tungsten-iron-sulfur protein of the hyperthermophilic archaebacterium, Pyrococcus furiosus, is an aldehyde ferredoxin oxidoreductase. Evidence for its participation in a unique glycolytic pathway. J Biol Chem 266: 14208-14216. Mukund S & Adams MWW (1995) Glyceraldehyde-3-phosphate Ferredoxin Oxidoreductase, a Novel Tungsten-containing Enzyme with a Potential Glycolytic Role in the Hyperthermophilic Archaeon Pyrococcus furiosus. J Biol Chem 270: 8389-8392. Radianingtyas H & Wright PC (2003) Alcohol dehydrogenases from thermophilic and hyperthermophilic archaea and bacteria. FEMS Microbiol Rev 27: 593-616. Reid MF & Fewson CA (1994) Molecular characterization of microbial alcohol dehydrogenases. Crit Rev Microbiol 20: 13-56. Rondeau JM, Tete-Favier F, Podjarny A, Reymann JM, Barth P, Biellmann JF & Moras D (1992) Novel NADPH-binding domain revealed by the crystal structure of aldose reductase. Nature 355: 469-472. Roy R, Mukund S, Schut GJ, Dunn DM, Weiss R & Adams MWW (1999) Purification and molecular characterization of the tungsten-containing formaldehyde ferredoxin oxidoreductase from the hyperthermophilic archaeon Pyrococcus furiosus: the third of a putative five-member tungstoenzyme family. J Bacteriol 181: 1171-1180. Roy R & Adams MWW (2002) Characterization of a fourth tungsten-containing enzyme from the hyperthermophilic archaeon Pyrococcus furiosus. J Bacteriol 184: 6952-6956. Sambrook J, Fritsch EF & Maniatis T (1989) Molecular cloning: A laboratory manual, 2nd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Sanli G & Blaber M (2001) Structural assembly of the active site in an aldo-keto reductase by NADPH cofactor. J Mol Biol 309: 1209-1218. Schut GJ, Zhou J & Adams MWW (2001) DNA microarray analysis of the hyperthermophilic archaeon Pyrococcus furiosus: evidence for a new type of sulfur-reducing enzyme complex. J Bacteriol 183: 7027-7036. Seery LT, Nestor PV & FitzGerald GA (1998) Molecular evolution of the aldo-keto reductase gene superfamily. J Mol Evol 46: 139-146. Sorensen HP, Sperling-Petersen HU & Mortensen KK (2003) Production of recombinant thermostable proteins expressed in Escherichia coli: completion of protein synthesis is the bottleneck. J Chromatogr B Analyt Technol Biomed Life Sci 786: 207-214. Urzhumtsev A, Tete-Favier F, Mitschler A, Barbanton J, Barth P, Urzhumtseva L, Biellmann JF, Podjarny A & Moras D (1997) A 'specificity' pocket inferred from the crystal structures of the complexes of aldose reductase with the pharmaceutically important inhibitors tolrestat and sorbinil. Structure 5: 601-612. Van der Oost J, Voorhorst WG, Kengen SWM, Geerling ACM, Wittenhorst V, Gueguen Y & De Vos WM (2001) Genetic and biochemical characterization of a short-chain alcohol dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus. Eur J Biochem 268: 3062-3068. Vieille C & Zeikus GJ (2001) Hyperthermophilic enzymes: sources, uses, and molecular mechanisms for thermostability. Microbiol Mol Biol Rev 65: 1-43.

69 Chapter 4

Wierenga RK, Terpstra P & Hol WG (1986) Prediction of the occurrence of the ADP-binding beta alpha beta-fold in proteins, using an amino acid sequence fingerprint. J Mol Biol 187: 101-107. Wilson DK, Bohren KM, Gabbay KH & Quiocho FA (1992) An unlikely sugar substrate site in the 1.65 Å structure of the human aldose reductase holoenzyme implicated in diabetic complications. Science 257: 81-84. Zeikus JG, Vieille C & Savchenko A (1998) Thermozymes: biotechnology and structure-function relationships. Extremophiles 2: 179-183.

70

Chapter 5

Production and characterization of a thermostable L-threonine dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus

Machielsen R & Van der Oost J

FEBS Journal (2006) 273(12): 2722-2729 Chapter 5

Abstract

The gene encoding a threonine dehydrogenase (TDH) has been identified in the hyperthermophilic archaeon Pyrococcus furiosus. The Pf-TDH protein has been functionally produced in Escherichia coli and purified to homogeneity. The enzyme has a tetrameric conformation with a molecular mass of approximately 155 kDa. The catalytic activity of the enzyme increases up to 100°C, and a half life value of 11 minutes at this temperature indicates its thermostability. The enzyme is specific for NAD(H), and maximal specific activities were detected with L-threonine (10.3 U/mg) and acetoin (3.9 U/mg) in the oxidative and reductive reactions, respectively. Pf-TDH utilizes also L-serine and D-threonine as substrate, but could not oxidize other L-amino acids. The enzyme requires divalent cations 2+ 2+ such as Zn and Co for the activity and contains at least one zinc atom/subunit. The Km + values for L-threonine and NAD at 70°C were 1.5 mM and 0.055 mM, respectively.

Introduction

L-Threonine dehydrogenase (TDH; EC 1.1.1.103) plays an important role in L-threonine + catabolism. TDH catalyzes the NAD(P) -dependent oxidation of L-threonine to 2-amino-3- oxobutyrate, which spontaneously decarboxylates to aminoacetone and CO2 or is cleaved in a CoA-dependent reaction by 2-amino-3-ketobutyrate coenzyme A lyase (EC 2.3.1.29) to glycine and acetyl coenzyme A (Bell & Turner, 1976; Newman et al., 1976; Potter et al., 1977). Most threonine dehydrogenases are closely related to the zinc-dependent alcohol dehydrogenases and member of the medium-chain dehydrogenase/reductase (MDR) superfamily. The superfamily is classified into eight families based on amino acid sequence alignment and the structural similarity of substrates. TDH belongs to the polyol dehydrogenase (PDH) family (Nordling et al., 2002; Riveros-Rosas et al., 2003). These enzymes utilize NAD(P)(H) as cofactor, are homotetramers or homodimers and usually contain one or two zinc atom(s)/subunit with catalytic and/or structural function. Enzymes from hyperthermophiles, microorganisms that grow optimally above 80°C, display an extreme stability at high temperature, high pressure, as well as high concentrations of chemical denaturants (Vieille & Zeikus, 2001). These features make hyperthermophilic enzymes very interesting from both scientific and industrial perspective. The hyperthermophilic archaeon Pyrococcus furiosus grows optimally at 100°C by the fermentation of peptides and carbohydrates to produce acetate, CO2, alanine and H2, together with minor amounts of ethanol. The organism will also generate H2S if elemental sulphur is present (Fiala & Stetter, 1986; Kengen et al., 1994, 1996). Three different alcohol dehydrogenases have previously been identified in Pyrococcus furiosus. A short-chain AdhA and an iron-containing AdhB encoded by the lamA operon (Van der Oost et al., 2001), and an oxygen-sensitive, iron and zinc-containing alcohol dehydrogenase that has been purified from

72 L-Threonine dehydrogenase from Pyrococcus furiosus cell extracts of P. furiosus (Ma & Adams, 1999). By careful analysis of the P. furiosus genome, sixteen additional genes have been identified that potentially encode alcohol dehydrogenases (Machielsen, unpublished results). The work reported here describes the functional production of one of the newly identified putative alcohol dehydrogenases, a threonine dehydrogenase (Pf-TDH, at first it was named AdhC), in Escherichia coli. The enzyme was purified to homogeneity, and characterized with respect to substrate specificity, metal requirement, kinetics and stability.

Materials and Methods

Chemicals and plasmids All chemicals (analytical grade) were purchased from Sigma-Aldrich or Acros Organics. The restriction enzymes were obtained from Invitrogen and New England Biolabs. Pfu Turbo and T4 DNA ligase were purchased from Invitrogen and Stratagene, respectively. For heterologous expression the vector pET-24d (KanR; Novagen) and the tRNA helper plasmid pSJS1244 (SpecR) (Kim et al., 1998; Sorensen et al., 2003) were used.

Organisms and growth conditions Escherichia coli XL1 Blue (Stratagene) was used as a host for the construction of pET24d derivatives. E. coli BL21(DE3) (Novagen) harbouring the tRNA helper plasmid pSJS1244 was used as an expression host. Both strains were grown under standard conditions (Sambrook et al., 1989) following the instructions of the manufacturer.

Cloning and sequencing of the alcohol dehydrogenase encoding gene The identification of the gene encoding an alcohol dehydrogenase was based on significant sequence similarity to several known alcohol dehydrogenases. The P. furiosus tdh gene (PF0991, GenBank accession number AE010211 region: 3490 - 4536, NCBI) was identified in the P. furiosus database (http://www.genome.utah.edu). The tdh gene (1047 bp) was PCR amplified from chromosomal DNA of P. furiosus using the primers BG1279 (5’- GCGCGCCATGGCATCCGAGAAGATGGTTGCTATCA, sense) and BG1297 (5’- GCGCGGGATCCTCATTTAAGCATGAAAACAACTTTGCC, antisense), containing NcoI and BamHI sites (underlined in the sequences). In order to introduce an NcoI restriction site an extra alanine codon (GCA) was introduced in the tdh gene by the forward primer BG 1279 (bold in the sequence). The fragment generated was purified using QIAquick PCR purification kit (Qiagen). The purified gene was digested with NcoI/BamHI and cloned into E. coli XL1-Blue using an NcoI/BamHI-digested pET24d vector. Subsequently, the resulting plasmid pWUR78 was transformed into E. coli BL21(DE3) harbouring the tRNA helper plasmid pSJS1244. The sequence of the expression clone was confirmed by sequence analysis of both DNA strands.

73 Chapter 5

Production and purification of ADH E. coli BL21(DE3) harbouring pSJS1244 was transformed with pWUR78 and a single colony was used to inoculate 5 ml Luria-Bertani medium with kanamycin and spectinomycin (both 50 μg⋅ml-1) and incubated overnight in a rotary shaker at 37 °C. Next, 1 ml of the preculture was used to inoculate 1 L Luria-Bertani medium with kanamycin and spectinomycin (both 50 mg⋅L-1) in a 2-L conical flask and incubated in a rotary shaker at 37 °C until a cell density of

OD600 nm = 0.6 was reached. The culture was then induced with 0.2 mM isopropyl thio-β-D- galactoside (IPTG) and incubation of the culture was continued at 37 °C for 18 h. Cells were harvested, resuspended in 20 mM Tris/HCl buffer (pH 7.5) and passed twice through a French press at 110 MPa. The crude cell extract was centrifuged for 20 min at 10.000 × g. The resulting supernatant (cell free extract) was heated for 30 min at 80°C and subsequently centrifuged for 20 min at 10.000 × g. The supernatant (heat stable cell free extract) was filtered (0.45 μm) and applied to a Q-sepharose high performance (GE Healthcare) column (1.6 x 10 cm) equilibrated in 20 mM Tris/HCl buffer (pH 7.8). Proteins were eluted with a linear 560 ml gradient from 0.0 to 1.0 M NaCl, in the same buffer.

Size-exclusion chromatography Molecular mass was determined by size-exclusion chromatography on a Superdex 200 HR 10/30 column (24 ml, GE Healthcare) equilibrated in 50 mM Tris/HCl (pH 7.8) containing 100 mM NaCl. Two hundred fifty μl of enzyme solution in 20 mM Tris/HCl buffer (pH 7.8) was injected on the column. Blue dextran 2000 (>2000 kDa), aldolase (158 kDa), bovine serum albumin (67 kDa), ovalbumin (43 kDa), chymotrypsinogen (25 kDa) and ribonuclease A (13.7 kDa) were used for calibration.

SDS-PAGE electrophoresis Protein composition was analyzed by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE, 10%) (Sambrook et al., 1989), using a Mini-Protean 3 system (Biorad). Protein samples for SDS-PAGE were prepared by heating for 30 min at 100°C in the presence of sample buffer (0.1 M sodium phosphate buffer, 4% SDS, 10% 2- mercaptoethanol, 20% glycerol, pH6.8). A broad range protein marker (Biorad) was used to estimate the molecular mass of the proteins.

Activity assays Rates of alcohol oxidation and aldehyde reduction were determined at 70 °C, unless stated otherwise, by following either the reduction of NAD+ or the oxidation of NADH at 340 nm using a Hitachi U2010 spectrophotometer, with a temperature controlled cuvette holder. Each oxidation reaction mixture contained 50 mM glycine (pH 10.0), 25 - 100 mM of alcohol and 0.28 mM NAD+. The reduction reaction mixture contained 0.1 M sodium phosphate buffer (pH 6.6), 100 mM aldehyde or ketone and 0.28 mM NADH. In all assays the reaction was initiated by addition of an appropriate amount of enzyme. One unit of ADH was defined as

74 L-Threonine dehydrogenase from Pyrococcus furiosus the oxidation or reduction of 1 μmol of NADH or NAD+ per min, respectively. Protein concentration was determined using Bradford reagents (Bio-Rad) with bovine serum albumin as a standard (Bradford, 1976). The temperature dependent spontaneous degradation of NADH was corrected for. pH optimum The pH optimum for alcohol oxidation was determined in a sodium phosphate buffer (100 mM, pH range 5.4 – 7.9) and a glycine buffer (50 mM, pH-range 7.9 – 11.5), whereas the pH optimum for aldehyde reduction was determined in a sodium phosphate buffer (100 mM, pH range 5.4 – 7.9). The pH of the buffers was set at 25°C and temperature corrections were made using their temperature coefficients (-0.025 pH/°C for glycine buffer and -0.0028 pH/°C for the sodium phosphate buffer).

Optimum temperature and thermostability The thermostability of Pf-TDH (enzyme concentration: 0.31 mg·mL-1 in 20 mM Tris buffer pH 7.8) was determined by measuring the residual activity (2,3-butanediol oxidation according to the standard assay) after incubation of a time series at 80, 90 or 100°C. The temperature optimum was determined in 50 mM glycine buffer pH 10.0 by analysis of initial rates of 2,3-butanediol oxidation in the range of 30-100°C.

Kinetics

The Pf-TDH kinetic parameters Km and Vmax were calculated from multiple measurements (at least 8 measurements) using the Michaelis-Menten equation and the program Tablecurve 2D (version 5.0). All the reactions followed Michaelis-Menten type kinetics. The turnover -1 number (kcat, s ) was calculated as: Vmax ∗ subunit molecular mass (38 kDa) / 60.

Salts, metals and inhibitors The effect of several salts, metals (K+, Mg2+, Mn2+, Na+, Fe2+, Fe3+, Li2+, Ni2+, Co2+, Zn2+, Ca2+) and inhibitors (Ethylenediaminetetraacetic acid - EDTA, dithiothreitol - DTT, 2- iodoacetamide) on the initial activity of Pf-TDH was checked using 2,3-butandiol as substrate in the oxidation reaction and acetoin in the reduction reaction. Concentrations ranging from 1 - 25 mM were tested. To determine the metal ion requirement, the enzyme solution was incubated for 30 minutes with 10 mM EDTA at 80°C. Subsequently, the treated enzyme solution was applied to a PD-10 desalting column (GE Healthcare) to remove the EDTA. The reactivity of the different divalent cations was tested by the addition of 2 mM ZnCl2, CoCl2, MnCl2, MgCl2,

NiCl2 or LiCl2 to the reaction mixture (2,3-butanediol oxidation according to the standard assay).

75 Chapter 5

The metal content (assayed for Ni, Mg, Zn, Cr, Co, Cu and Fe) of the purified enzyme was determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES) using 20 mM Tris/HCl buffer (pH 7.8) as a blank.

Results

Analysis of nucleotide and amino acid sequences The P. furiosus genome has been analyzed for genes encoding putative alcohol dehydrogenases, which resulted in the identification of sixteen genes that potentially encode alcohol dehydrogenases. After successful production in E. coli, an initial screening for activity was performed in which two of the putative alcohol dehydrogenases, including Pf- TDH, showed relatively high activities (Machielsen, unpublished results). The two enzymes were selected for a more detailed study; with respect to the other putative alcohol dehydrogenases, a more elaborate screening is currently being performed to get insight in their substrate specificity, and possibly in their physiological function. Here we describe the production and characterization of one of the selected enzymes, a novel L-threonine dehydrogenase, Pf-TDH (PF0991). The P. furiosus tdh gene encodes a protein of 348 amino acids and a calculated molecular mass of 37.823 kDa. The sequence belongs to the cluster of orthologous groups of proteins 1063 (threonine dehydrogenase and related Zn-dependent dehydrogenases; http://www.ncbi.nlm.nih.gov/COG/). BLAST-P analysis (http://www.ncbi.nlm.nih.gov/blast/) reveals the highest similarity with (putative) threonine dehydrogenases and zinc-containing alcohol dehydrogenases from archaea and bacteria. Some of these most significant hits of a BLAST search analysis were a L-threonine dehydrogenase of Pyrococcus horikoshii (95% identity, PH0655) (Higashi et al., 2005ab; Shimizu et al., 2005), a putative L-threonine dehydrogenase of Thermococcus kodakaraensis KOD1 (88% identity, TK0916), a hypothetical threonine or Zn-dependent dehydrogenase of Thermoanaerobacter tengcongensis (53% identity, TTE2405) and a L-threonine 3-dehydrogenase of Escherichia coli (44% identity, tdh) (Boylan & Dekker, 1981; Aronson et al., 1989). These sequences were used to make an alignment (Fig. 5.1). Highly conserved residues within the MDR superfamily, especially the PDH family, are indicated with an asterisk (Fig. 5.1, P. furiosus numbering). Members of the PDH family bind the cofactor NAD(P) with a Rossmann-fold motif of which the residues Gly168, Gly175, Gly177, Gly180 and Gly212 are highly conserved (Wierenga et al., 1986; Jornvall et al., 1987). Residues that are necessary to bind the catalytic zinc ion and modulate its electrostatic environment, Cys42, Asp45, His67, Glu68 and Asp/Glu 152 (Sun & Plapp, 1992; Chen et al., 1995; Johnson et al., 1998), and residues responsible for binding the structural zinc ion, the cysteines 97, 100, 103 and 111 (Sun & Plapp, 1992; Clark-Baldwin et al., 1998), are also completely conserved. The other conserved residues are a probable base catalyst for alcohol oxidation (His47), as well as residues involved in substrate binding (Gly66, Gly71, Gly77 and Val80) and in facilitating

76 L-Threonine dehydrogenase from Pyrococcus furiosus proton removal from the substrate (Thr44) (Sun & Plapp, 1992). In addition, His94 is suggested to be an active-site residue, which modulates the substrate specificity of L-threonine dehydrogenase (Marcus & Dekker, 1995; Johnson & Dekker, 1998). Conserved context analysis with STRING (http://string.embl.de/) reveals no functional link in the genome neighbourhood of Pf-TDH, although manual inspection identified that the genome neighbourhood of the tdh homologs in the related species P. furiosus, P. abyssi and P. horikoshii is highly conserved. Interestingly, this analysis revealed that the hypothetical threonine dehydrogenase of T. tengcongensis was followed directly by a gene (TTE2406) encoding 2-amino-3-ketobutyrate coenzyme A ligase, the enzyme which converts 2-amino-3- oxobutyrate to glycine. BLAST-P analysis showed that there is also a homolog of this enzyme in P. furiosus (PF0265, 37% identity).

Figure 5.1 Multiple sequence alignment of the P. furiosus TDH with (hypothetical) threonine dehydrogenases and related Zn-dependent dehydrogenases. The following symbols are used: Pyrfu, P. furiosus; Pyrho, P. horikoshii; Theko, T. kodakaraensis; Thete, T. tengcongensis; Escco, E. coli. The sequences were aligned using the CLUSTAL program. Asterisks indicate highly conserved residues within the MDR superfamily.

77 Chapter 5

Purification of recombinant Pf-TDH The pyrococcal TDH was purified to homogeneity from heat-treated cell-free extracts of E. coli BL21(DE3)/pSJS1244/pWUR78 by anion-exchange chromatography (Table 5.1). Active Pf-TDH eluted between 0.32 and 0.46 M NaCl (peak at 0.40 M NaCl). Fractions containing the purified enzyme were pooled. The migration of Pf-TDH on SDS-PAGE reveals a molecular subunit mass of approximately 40 kDa, which is in fair agreement with the molecular weight (38 kDa) calculated from the amino acid sequence. The molecular mass of the native Pf-TDH was estimated to be 156 kDa by size-exclusion chromatography, which indicated a homotetrameric structure.

Table 5.1 Pf-TDH purification table Total Specific Purification Protein activity activity Yield Purification step (mg) (U) (U/mg) (%) (fold) Cell extract 806.9 78.3 0.097 100 1 Heat treatment 44.8 62.7 1.40 80 14 Q-Sepharose 10.5 49.4 4.70 63 48

Substrate and cofactor specificity The substrate specificity of Pf-TDH in the oxidation reaction was analyzed using primary alcohols (methanol to dodecanol, C1-C12), secondary alcohols (2-propanol to 2-decanol, C3-

C10), alcohols containing more than one hydroxyl group and L-amino acids. Pf-TDH showed no activity towards primary alcohols and secondary alcohols. The highest specific activity of -1 Pf-TDH in the oxidative reaction was found with L-threonine (Vmax 10.3 U· mg ). The enzyme exhibited also activity with D-threonine, L-serine, L-glycerate, 3-hydroxybutyrate, lactate, 2,3-butanediol, 1,2-butanediol, 1,2-propanediol and glycerol (Table 5.2), but many other L-amino acids, including L-aspartate, L-glutamine, L-alanine, L-arginine, L-cysteine, L- proline, L-phenylalanine, L-lysine, L-tryptophan, L-isoleucine, L-tyrosine, L-histidine, L- leucine, L-valine, L-methionine, L-glutamate and glycine, could not be oxidized by Pf-TDH. The substrate specificity of the reduction reaction was analyzed by using aldehydes, ketones and aldoses as substrate. Unfortunately, the substrate 2-amino-3-oxobutyrate could not be tested because of its instability, and activities were only observed with diacetyl and -1 acetoin (3-hydroxy-2-butanone, Vmax 3.9 U· mg ). Pf-TDH could use NAD(H) as cofactor, but could not utilize NADP(H).

78 L-Threonine dehydrogenase from Pyrococcus furiosus

Table 5.2 Substrate specificity of P. furiosus Pf-TDH in the oxidation reaction Relative Substrate activity (%) L-threonine 100 D-threonine 5 L-serine 15 L-Glycerate 6 3-hydroxybutyrate 3 Lactate 1 2,3-butanediol 94 1,3-butanediol 0 1,2-butanediol 52 1-butanol 0 2-butanol 0 1,2-propanediol 55 glycerol 4

Metals and inhibitors The effect of several salts, metals and inhibitors on the initial activity of Pf-TDH was checked using 2,3-butandiol as substrate in the standard oxidation reaction, and acetoin in the reduction reaction. The activity of Pf-TDH was significantly increased by addition of 2 mM

CoCl2 (relative activity to that of the standard reaction: 170%) and not by the addition of 2 mM ZnCl2 or one of the other metals/salts tested. The enzyme was inhibited by addition of 5 mM DTT (relative activity to that of the standard reaction: 24%) and 2 mM 2-iodoacetamide (74%). Inhibition by the thiol reducing agent DTT, and by the alkylating sulfhydryl reagent 2- iodoacetamide, suggests that disulfide bridges and/or thiol groups play an important role in Pf-TDH. The activity was completely lost when the enzyme was incubated for 30 minutes with the chelating agent EDTA (10 mM) at 80°C. However, EDTA did not inhibit the enzyme when it was added to the standard reaction without the incubation at 80°C. After removal of

EDTA, the full activity of the enzyme could be recovered by the addition of 2 mM ZnCl2 or

CoCl2. The activity could be partially restored by addition of MgCl2 (69%) and NiCl2 (27%). Metal analysis of the purified Pf-TDH by ICP-AES revealed that the enzyme contains 0.64 mol of Zn2+ per mol of enzyme subunit. This result strongly suggests that the enzyme has (at least) one zinc atom per subunit, which is similar to the TDH of E. coli (Epperly & Dekker, 1991; Clark-Baldwin et al., 1998).

79 Chapter 5

Thermostability and pH optima The oxidation reaction catalyzed by Pf-TDH showed a pH optimum of 10.0, and the reduction reaction by Pf-TDH showed a high level of activity over a wide range of pH with maximal activity at pH 6.6. The reaction rate of Pf-TDH increased with increasing temperature from 37°C (0.55 U/mg) up to 100°C (6.43 U/mg), but due to instability of the cofactors at that temperature all other activity measurements were performed at 70°C. At this temperature the activity was 28% lower than at 100°C. Pf-TDH has an extreme resistance to thermal inactivation, which was shown by half-life values of 100 minutes at 80°C, 36 minutes at 90°C and 11 minutes at 100°C.

Enzyme kinetics The kinetic properties of Pf-TDH were determined for the substrates that were converted with relatively high rates in the oxidation and reduction reaction, as well as for the cofactors used in these reactions. It was found that in the oxidation reaction Pf-TDH has a relative high -1 -1 affinity for L-threonine (Km 1.5 mM, Vmax 10.3 U/mg, kcat/Km 4.3 s • mM ) and NAD (Km 55

µM, Vmax 10.3 U/mg) and clearly a lower affinity for 2,3-butanediol (Km 25.9 mM, Vmax 9.7 -1 -1 U/mg, kcat/Km 0.24 s • mM ). In the reduction reaction Pf-TDH showed a high affinity for the cofactor NADH (Km 10.8 µM, Vmax 3.9 U/mg), but a very low affinity for the substrate -1 -1 acetoin (Km 231.7 mM, Vmax 3.9 U/mg, kcat/Km 0.011 s • mM ).

Discussion

Three pathways for threonine degradation are known. Threonine aldolase (EC 4.1.2.5) is responsible for the conversion of threonine to acetaldehyde and glycine. The threonine dehydratase (EC 4.3.1.19) catalyzed reaction leads to formation of 2-oxobutanoate (and NH3), which can be further converted to propionate or isoleucine. Alternatively, threonine dehydrogenase (EC 1.1.1.103) catalyzes the NAD(P)+-dependent conversion of threonine to

2-amino-3-oxobutyrate, which spontaneously decarboxylates to aminoacetone and CO2, or is cleaved in a CoA-dependent reaction by 2-amino-3-ketobutyrate coenzyme A lyase (EC 2.3.1.29) to glycine and acetyl-CoA. Aminoacetone can be further converted to 1-amino-2- propanol, or via methylglyoxal to pyruvate (Bell & Turner, 1976; Newman et al., 1976). Threonine dehydrogenases have been found in eukaryotes, bacteria and recently also in archaea (Boylan & Dekker, 1981; Yuan & Austic, 2001; Shimizu et al., 2005). The enzyme described here, Pf-TDH, is a L-threonine dehydrogenase from the archaeon Pyrococcus furiosus. Pf-TDH was functionally produced in E. coli and due to its stability at high temperature only two steps were needed for purification. It could only use NAD(H) as cofactor and showed highest activity with L-threonine. Pf-TDH utilized also L-serine and D- threonine as substrate, but could not oxidize other L-amino acids. The Km values for L- threonine and NAD+ at 70°C were 1.5 mM and 0.055 mM, respectively, which resembles the

80 L-Threonine dehydrogenase from Pyrococcus furiosus

values reported for L-threonine dehydrogenase of E. coli (Boylan & Dekker, 1981). The substrate specificity shown in Table 5.2 reveals that Pf-TDH does require neither the amino group nor the carboxyl group of L-threonine for activity, but the enzyme kinetics clearly show the preference for L-threonine when compared to 2,3-butanediol. Determinants of the Pf-TDH substrate specificity are shown in Figure 5.2. The specific configuration of the substrate is clearly important as demonstrated by the difference in activity with L-threonine and D- threonine (Fig. 5.2 A). The activity is significantly higher when the oxidisable substrate possesses a methyl group at C4 (Fig. 5.2 B, L-threonine versus L-serine), and when it possesses either an amino or a hydroxyl group at C2, which is probably involved in correct positioning of the substrate molecule through hydrogen bonding (Fig. 5.2 C, L-threonine and 2,3-butanediol versus 2-butanol). Although the carboxyl group is not required for activity, it is obvious from the comparison between 3-hydroxybutyrate and 2-butanol as substrate that it can have a distinct influence on the activity (Fig. 5.2 D). Like most threonine dehydrogenases, Pf-TDH belongs to the polyol dehydrogenase (PDH) family, which is part of the medium-chain dehydrogenase/reductase (MDR) superfamily. Members of this superfamily have either a dimeric or tetrameric structure and contain one or two zinc atoms/subunit, a catalytic and/or structural zinc atom. Size-exclusion chromatography indicated a homotetrameric structure for Pf-TDH and metal analysis by ICP- AES revealed that Pf-TDH contains at least one zinc atom per subunit, which is similar to the TDH of E. coli (Epperly & Dekker, 1991; Clark-Baldwin et al., 1998). However, the alignment reveals that both enzymes contain the conserved residues which are (potentially) involved in binding of both the catalytic and the structural zinc atom. Incubation with EDTA at 80°C abolished the activity of Pf-TDH completely and addition of Zn2+ or Co2+ could restore the full activity of the enzyme. Although this indicates that the metal ion is essential for activity, further research is needed to establish if the zinc atom is catalytic or structural. This has been done for the TDH of E. coli and X-ray absorption spectroscopic studies have shown that its zinc atom is probably liganded by four cysteine residues, which suggests a structural role for Zn2+ (Clark-Baldwin et al., 1998). However, additional studies have resulted in the speculation that, in vivo, the enzyme has not only the structural 4-Cys Zn2+ binding site, but also a second divalent metal ion that is responsible for the relatively high affinity for L-threonine (Epperly & Dekker, 1991; Johnson et al., 1998; Johnson & Dekker, 1998). Since Pf-TDH is stimulated by addition of Co2+ (and not by Zn2+), it is possible that in vivo Co2+ is the second catalytic metal ion of each Pf-TDH subunit, which would then contain one structural Zn2+, as well as one Co2+ involved in substrate binding. Conserved context analysis followed by a blast search identified a possible 2-amino-3- ketobutyrate coenzyme A lyase in P. furiosus. Studies with TDH and 2-amino-3-ketobutyrate coenzyme A lyase from mammalian source and from E. coli have shown that together these enzymes catalyze the 2-step conversion of L-threonine to glycine (Tressel et al., 1986; Ravnikar & Somerville, 1987). In addition, it has been shown in E. coli that these enzymes are responsible for the formation of threonine from glycine in vitro and in vivo (Marcus &

81 Chapter 5

Dekker, 1993), however, the primary role of this pathway is believed to be threonine catabolism. We suggest that the physiological role of Pf-TDH is the oxidation of L-threonine to 2-amino-3-oxobutyrate, which is probably converted to glycine by a 2-amino-3- ketobutyrate coenzyme A lyase.

Figure 5.2 Determinants of Pf-TDH substrate specificity. (A) configuration of the substrate, (B) methyl group, (C) additional amino group (threonine) or hydroxyl group (2,3-butanediol) for H-bonding, (D) carboxyl group. * racemic mixtures have been used in activity measurements.

82 L-Threonine dehydrogenase from Pyrococcus furiosus

Acknowledgements

This work was supported by the EU 5th framework program PYRED (QLK3-CT-2001- 01676). We thank Dr. F.A. de Bok (Wageningen) for metal analysis by ICP-AES.

References

Aronson BD, Somerville RL, Epperly BR & Dekker EE (1989) The primary structure of Escherichia coli L-threonine dehydrogenase. J Biol Chem 264(9): 5226-5232. Bell SC & Turner JM (1976) Bacterial catabolism of threonine. Threonine degradation initiated by L-threonine-NAD+ oxidoreductase. Biochem J 156(2): 449-458. Boylan SA & Dekker EE (1981) L-Threonine dehydrogenase. Purification and properties of the homogeneous enzyme from Escherichia coli K-12. J Biol Chem 256(4): 1809-1815. Bradford MM (1976) A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 72: 248-254. Chen YW, Dekker EE & Somerville RL (1995) Functional analysis of E. coli threonine dehydrogenase by means of mutant isolation and characterization. Biochim Biophys Acta 1253(2): 208-214. Clark-Baldwin K, Johnson AR, Chen Y-W, Dekker EE & Penner-Hahn JE (1998) Structural characterization of the zinc site in Escherichia coli L-threonine dehydrogenase using extended X-ray absorption fine structure spectroscopy. Inorg Chim Acta 275-276: 215-221. Epperly BR & Dekker EE (1991) L-Threonine dehydrogenase from Escherichia coli. Identification of an active site cysteine residue and metal ion studies. J Biol Chem 266(10): 6086-6092. Fiala G & Stetter KO (1986) Pyrococcus furiosus sp. nov. represents a novel genus of marine heterotrophic archaebacteria growing optimally at 100oC. Arch Microbiol 145: 56-61. Higashi N, Fukada H & Ishikawa K (2005a) Kinetic study of thermostable L-threonine dehydrogenase from an archaeon Pyrococcus horikoshii. J Biosci Bioeng 99(2): 175-180. Higashi N, Matsuura T, Nakagawa A & Ishikawa K (2005b) Crystallization and preliminary X-ray analysis of hyperthermophilic L-threonine dehydrogenase from the archaeon Pyrococcus horikoshii. Acta Crystallogr Sect F 61(4): 432-434. Johnson AR, Chen YW & Dekker EE (1998) Investigation of a catalytic zinc binding site in Escherichia coli L-threonine dehydrogenase by site-directed mutagenesis of cysteine-38. Arch Biochem Biophys 358(2): 211-221. Johnson AR & Dekker EE (1998) Site-directed Mutagenesis of histidine-90 in Escherichia coli L- threonine dehydrogenase alters its substrate specificity. Arch Biochem Biophys 351(1): 8-16. Jornvall H, Persson B & Jeffery J (1987) Characteristics of alcohol/polyol dehydrogenases. The zinc-containing long-chain alcohol dehydrogenases. Eur J Biochem 167(2): 195-201. Kengen SW, de Bok FA, van Loo ND, Dijkema C, Stams AJM & de Vos WM (1994) Evidence for the operation of a novel Embden-Meyerhof pathway that involves ADP-dependent kinases during sugar fermentation by Pyrococcus furiosus. J Biol Chem 269(26): 17537-17541. Kengen SWM, Stams AJM & de Vos WM (1996) Sugar metabolism of hyperthermophiles. FEMS Microbiol Rev 18(2-3): 119-137.

83 Chapter 5

Kim R, Sandler SJ, Goldman S, Yokota H, Clark AJ & Kim SH (1998) Overexpression of archaeal proteins in Escherichia coli. Biotechnol Lett 20(3): 207-210. Ma K & Adams MW (1999) An unusual oxygen-sensitive, iron- and zinc-containing alcohol dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus. J Bacteriol 181(4): 1163- 1170. Marcus JP & Dekker EE (1993) Threonine formation via the coupled activity of 2-amino-3- ketobutyrate coenzyme A lyase and threonine dehydrogenase. J Bacteriol 175(20): 6505-6511. Marcus JP & Dekker EE (1995) Identification of a second active site residue in Escherichia coli L- threonine dehydrogenase: methylation of histidine-90 with methyl p-nitrobenzenesulfonate. Arch Biochem Biophys 316(1): 413-420. Newman EB, Kapoor V & Potter R (1976) Role of L-threonine dehydrogenase in the catabolism of threonine and synthesis of glycine by Escherichia coli. J Bacteriol 126(3): 1245-1249. Nordling E, Jornvall H & Persson B (2002) Medium-chain dehydrogenases/reductases (MDR): Family characterizations including genome comparisons and active site modelling. Eur J Biochem 269(17): 4267-4276. Potter R, Kapoor V & Newman EB (1977) Role of threonine dehydrogenase in Escherichia coli threonine degradation. J Bacteriol 132(2): 385-391. Ravnikar PD & Somerville RL (1987) Genetic characterization of a highly efficient alternate pathway of serine biosynthesis in Escherichia coli. J Bacteriol 169(6): 2611-2617. Riveros-Rosas H, Julian-Sanchez A, Villalobos-Molina R, Pardo JP & Pina E (2003) Diversity, taxonomy and evolution of medium-chain dehydrogenase/reductase superfamily. Eur J Biochem 270(16): 3309-3334. Sambrook J, Fritsch EF & Maniatis T (1989). Molecular cloning: A laboratory manual, 2nd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Shimizu Y, Sakuraba H, Kawakami R, Goda S, Kawarabayasi Y & Ohshima T (2005) L- Threonine dehydrogenase from the hyperthermophilic archaeon Pyrococcus horikoshii OT3: gene cloning and enzymatic characterization. Extremophiles 9(4): 317-324. Sorensen HP, Sperling-Petersen HU & Mortensen KK (2003) Production of recombinant thermostable proteins expressed in Escherichia coli: completion of protein synthesis is the bottleneck. J Chromatogr B Analyt Technol Biomed Life Sci 786(1-2): 207-214. Sun HW & Plapp BV (1992) Progressive sequence alignment and molecular evolution of the Zn- containing alcohol dehydrogenase family. J Mol Evol 34(6): 522-535. Tressel T, Thompson R, Zieske LR, Menendez MI & Davis L (1986) Interaction between L- threonine dehydrogenase and aminoacetone synthetase and mechanism of aminoacetone production. J Biol Chem 261(35): 16428-16437. Van der Oost J, Voorhorst WG, Kengen SWM, Geerling ACM, Wittenhorst V, Gueguen Y & de Vos WM (2001) Genetic and biochemical characterization of a short-chain alcohol dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus. Eur J Biochem 268(10): 3062-3068. Vieille C & Zeikus GJ (2001) Hyperthermophilic enzymes: sources, uses, and molecular mechanisms for thermostability. Microbiol Mol Biol Rev 65(1): 1-43. Wierenga RK, Terpstra P & Hol WG (1986) Prediction of the occurrence of the ADP-binding beta alpha beta-fold in proteins, using an amino acid sequence fingerprint. J Mol Biol 187(1): 101-107. Yuan JH & E. Austic R (2001) Characterization of hepatic -threonine dehydrogenase of chicken. Comp Biochem Physiol B - Biochem Mol Biol 130(1): 65-73.

84

Chapter 6

Laboratory evolution of Pyrococcus furiosus alcohol dehydrogenase to improve the production of (2S,5S)-hexanediol at moderate temperatures

Machielsen R, Leferink NGH, Hendriks A, Brouns SJJ, Hennemann H, Dauβmann T & Van der Oost J

Submitted Chapter 6

Abstract

There is considerable interest in the use of enantioselective alcohol dehydrogenases for the production of enantio- and diastereomerically pure diols, which are important building blocks for pharmaceuticals, agrochemicals and fine chemicals. Due to the need for a stable alcohol dehydrogenase with activity at low-temperature process conditions (30°C) for the production of (2S,5S)-hexanediol, we have improved an alcohol dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus (AdhA). A stable S-selective alcohol dehydrogenase with increased activity at 30°C on the substrate 2,5-hexanedione was generated by laboratory evolution on the thermostable alcohol dehydrogenase AdhA. One round of error-prone PCR and screening of ∼1500 mutants was performed. The maximum specific activity of the best performing mutant with 2,5-hexanedione at 30°C was 10-fold higher compared to the activity of the wild type enzyme. A 3D-model of AdhA revealed that this mutant has one mutation in the well conserved NADP(H)-binding site (R11L), and a second mutation (A180V) near the catalytic and highly conserved threonine at position 183.

Introduction

Alcohol dehydrogenases (ADHs) are present in all organisms. They display a wide variety of substrate specificities and play an important role in a broad range of physiological processes (Reid & Fewson, 1994). There is considerable interest in stable alcohol dehydrogenases for a range of applications in food, pharmaceutical and fine chemicals industries. The production of enantio- and diastereomerically pure diols is particularly desired because these are important chemical building blocks (Zeikus et al., 1998; Hummel, 1999; Haberland et al., 2002a; Radianingtyas & Wright, 2003). (2S,5S)-Hexanediol is such a versatile building block for the synthesis of various fine chemicals, pharmaceuticals and chiral phosphine ligands (Haberland et al., 2002b; Brunel & Faure, 2004). For most applications robust biocatalysts are desired. Enzymes from hyperthermophiles, i.e., microorganisms that grow optimally above 80°C, generally display an extreme stability at high temperature, high pressure, as well as high concentrations of chemical denaturants (Vieille & Zeikus, 2001). Enzymes are environmentally friendly, biodegradable, efficient, and low cost in terms of resource requirements; as such they provide benefits compared to traditional chemical approaches in various industrial processes. In many instances, however, natural enzymes do not perform optimally in a particular unnatural process, and as such can be unsuitable for large-scale industrial applications (Schmid et al., 2001; Schoemaker et al., 2003). The production of (2S,5S)-hexanediol, can be achieved by chemical synthesis or biocatalytic processes. Chemical synthesis of (2,5)-hexanediol results in a racemic/meso mixture of (2,5)-hexanediol with a theoretical yield of 25% for (2S,5S)-hexanediol. This diol

86 Random mutagenesis P. furiosus AdhA can also be obtained by the enzymatic reduction of the cheap 2,5-hexanedione with baker's yeast or an enantioselective alcohol dehydrogenase (Lieser, 1983). The enzymatic conversion of 2,5-hexanedione to (2S,5S)-hexanediol is preferably performed with low cost process conditions such as a process temperature of 30°C, using a stable and enantioselective alcohol dehydrogenase and an efficient cofactor regeneration system which also allows easy purification of the end product. Previously, the adhA gene was cloned, sequenced and functionally overexpressed in Escherichia coli (Van der Oost et al., 2001). AdhA was purified, characterized and found to be a homodimer of 26 kDa subunits. The enzyme is very well adapted to the high growth temperature of the organism (70-103°C) (Fiala & Stetter, 1986) with an optimum temperature for activity of 90°C, and a high resistance to thermal inactivation which was shown by half life values of 150 h at 80°C, 22.5 h at 90°C and 25 min. at 100°C. AdhA is an NADP+- dependent alcohol dehydrogenase that converts alcohols to the corresponding aldehydes/ ketones and vice versa, with a rather broad substrate specificity. Maximal specific activities were observed with 2-pentanol (46 U⋅mg-1) and pyruvaldehyde (32 U⋅mg-1) in the oxidative and reductive reaction, respectively (Van der Oost et al., 2001). We here report that the activity of AdhA at 30°C does not exceed 5% of its optimal activity. In addition, AdhA has been demonstrated to be S-selective: it stereoselectively reduces 2,5-hexanedione, to (2S,5S)-hexanediol. Interestingly, it has been found that in the presence of excess 2-propanol, AdhA can perform substrate-coupled NADPH regeneration, as has been described for other ADHs. In substrate-coupled NADPH regeneration a single enzyme (AdhA) is involved in both the main and recycling reactions. In the oxidation of 2- propanol to acetone the coenzyme is regenerated in the same active site that carries out the reductive conversion of the substrate of interest, 2,5-hexanedione (Fig. 6.1). To achieve high conversions the 2-propanol should be present in excess (Bastos et al., 1999; Eckstein et al., 2004). The main goal has been the improvement of the low-temperature activity of AdhA from the hyperthermophilic archaeon Pyrococcus furiosus. There are two main routes to improve an enzyme: (1) rationale-based mutagenesis, and (2) laboratory evolution. To allow rationale-based engineering, a high-resolution crystal structure and insight into the structure- function relations of the biocatalyst of interest are required. Unfortunately, no crystal structure of AdhA is available at present. Laboratory evolution on the other hand offers a way to optimize enzymes by random mutagenesis in the absence of structural or mechanistic information (Lebbink et al., 2000; Bornscheuer & Pohl, 2001; Yuan et al., 2005). The latter approach has been selected as the method to improve the low-temperature activity of AdhA on 2,5-hexanedione, while retaining its stability. The obtained improved AdhA variants have been analyzed, and the results are interpreted by using a structural model of the enzyme.

87 Chapter 6

Figure 6.1 Substrate coupled approach for the cofactor regeneration. In substrate coupled NADPH regeneration a single enzyme (AdhA) is involved in both the main and recycling reactions.

Materials and Methods

Chemicals and plasmids 2,5-Hexanedione and (2S,5S)-hexanediol were kindly provided by Julich Chiral Solutions (Jülich, Germany). All other chemicals (analytical grade) were purchased from Sigma-Aldrich (Munich, Germany) or Acros Organics (Geel, Belgium). The restriction enzymes were obtained from Invitrogen (Paisley, UK) and New England Biolabs (Ipswich, MA, USA). T4 DNA ligase and Genemorph random mutagenesis kit were purchased from Stratagene (Amsterdam, The Netherlands) and REDTaq DNA polymerase from Sigma-Aldrich. For heterologous expression the vector pET-24d (KanR; Novagen, Darmstadt, Germany) and the tRNA helper plasmid pSJS1244 (SpecR) (Kim et al., 1998; Sorensen et al., 2003) were used.

Organisms and growth conditions Escherichia coli XL1-Blue (Stratagene) was used as a host for the construction of pET24d derivatives. E. coli BL21(DE3) (Novagen) harboring the tRNA helper plasmid pSJS1244 was used as an expression host. Both strains were grown under standard conditions (Sambrook et al., 1989) following the instructions from the manufacturer.

Construction of a random mutant AdhA library The adhA gene (705 bp) was PCR amplified from the chromosomal DNA of P. furiosus using the primers BG1394 (sense, 5’- GCGCGCCATGGCAAAGGTTGCCGTAATTACTGGG) and BG1396 (antisense, 5’-GCGCGGGATCCTCAATACTCAGGTTTTTGATAAATTGA G), containing the NcoI and BamHI sites (underlined in the sequences). In order to introduce an NcoI restriction site an extra alanine codon (GCA) was introduced in the adhA gene by the forward primer BG1394 (boldface in the sequence). The fragment generated was purified

88 Random mutagenesis P. furiosus AdhA using the QIAquick PCR purification kit (Qiagen, Hilden, Germany). The purified gene was digested with NcoI-BamHI and cloned into E. coli XL1-Blue using an NcoI-BamHI digested pET24d vector. The resulting plasmid was named pWUR104 and the sequence was confirmed by sequence analysis of both DNA strands (Baseclear, Leiden, The Netherlands). Random mutations were introduced into the adhA gene on pWUR104 by PCR amplification with primers BG417 (sense, 5’-CTTTAAGAAGGAGATATACCATG; creates a NcoI restriction site in PCR product) and BG1740 (antisense, 5’- GGAGCTCGAATTCGGATCCTCA; BamHI restriction site underlined). The error-prone PCR was conducted with REDTaq DNA polymerase under mutagenic reaction conditions (50 µl reaction mixture containing 10 ng template DNA, 200 ng primer BG417, 200 ng primer

BG1740, 0.2 mM dATP and dGTP, 1 mM dCTP and dTTP, 0.5 mM MnCl2 and 5 units REDTaq DNA polymerase) and using the Genemorph random mutagenesis kit with the Mutazyme DNA polymerase (reaction conditions according to the manufacturer; Stratagene). The resulting PCR products were digested with NcoI-BamHI and subsequently cloned into expression vector pET24d. The resulting plasmids were used to transform E. coli XL1- Blue. Transformation mixtures were plated onto selective Luria-Bertani agar plates containing 50 µg/ml kanamycin and incubated overnight at 37°C. Next the mutant library was harvested from each Luria-Bertani agar plate by adding 1.5 ml Luria-Bertani medium and gently resuspending the colonies from the plate. The QIAprep spin miniprep kit (Qiagen) was used for the isolation of plasmid and subsequently plasmid of the mutant library was transformed to E. coli BL21(DE3) harboring the tRNA helper plasmid pSJS1244. Transformation mixtures were plated onto selective Luria-Bertani agar plates containing 50 µg/ml kanamycin and 50 µg/ml spectinomycin. Single colonies were transferred to microtiter plates containing 200 µl/well Luria-Bertani medium supplemented with 10% (w/v) glycerol, 50 µg/ml kanamycin and spectinomycin. In each microtiter plate 3 reference wells were present: (i) E. coli BL21(DE3) harboring pSJS1244 and pWUR104 (expression of wild type AdhA), (ii) E. coli BL21(DE3) harboring pSJS1244 and an empty pET24d vector, (iii) blank (not inoculated).

Screening for increased activity on 2,5-hexanedione at 30°C A replica of the random AdhA library was prepared in microtiter plates containing 200 µl/well Luria-Bertani medium supplemented with 0.1 mM isopropyl-β-d- thiogalactopyranoside (IPTG), 50 µg/ml kanamycin and spectinomycin. After overnight growth at 37°C the E. coli cells were harvested by centrifugation at 4000 rpm for 10 minutes in a Hermle 383 Z centrifuge and subsequently lysed by addition of 40 µl Bacterial Protein Extraction Reagent (B-PER; Pierce, Rockford, IL, USA). No negative effect of B-PER on AdhA activity was detected in control experiments. The lysed cells were centrifuged (4000 rpm, 15 min.) and 40 µl of the resulting supernatant (cell free extract) was transferred to a second set of microtiter plates containing 160 µl/well 0.1 M sodium phosphate (pH 7.0), 125 mM 2,5-hexanedione and 0.35 mM NADPH. The cell free extract and assay buffer were

89 Chapter 6 mixed, incubated at 30°C and the absorbance at 340 nm was measured in each well using an iEMS reader MF (Thermo Electron Corporation, Breda, The Netherlands). Potential high- performance mutants were selected by a significant faster decrease of absorbance at 340 nm (due to conversion of NADPH to NADP+) compared to the wild type enzyme. The temperature dependent spontaneous degradation of NADPH was corrected for. These mutants were regrown in 5 ml Luria-Bertani medium supplemented with 50 µg/ml kanamycin and spectinomycin until a cell density of OD600 nm = 0.6 was reached. The culture was then induced with 0.1 mM IPTG and incubation of the culture at 37 °C was continued overnight. Cells were harvested, resuspended in 20 mM Tris/HCl buffer (pH 7.8) and disrupted by sonication. The crude cell extract was centrifuged for 10 min. at 15.000 × g. The resulting supernatant (cell free extract) was heated for 30 min at 80°C and subsequently centrifuged for 10 min at 15.000 × g. The resulting supernatant (heat stable cell free extract) was used to rescreen the potential high-performance mutants.

Production and purification of ADH Five ml Luria-Bertani medium with kanamycin and spectinomycin (both 50 μg⋅ml-1) was inoculated with E. coli BL21(DE3) harboring pSJS1244 and expression plasmids containing either wild type or mutant adhA genes and incubated overnight in a rotary shaker at 37 °C. Next, 1 ml of the preculture was used to inoculate 1 L Luria-Bertani medium with kanamycin and spectinomycin (both 50 mg⋅L-1) in a 2-L conical flask and incubated in a rotary shaker at

37 °C until a cell density of OD600 nm = 0.6 was reached. The culture was then induced with 0.2 mM IPTG and incubation of the culture was continued at 37 °C for 18 h. Cells were harvested, resuspended in 20 mM Tris/HCl buffer (pH 7.5) and passed twice through a French press at 110 MPa. The crude cell extract was centrifuged for 20 min at 10.000 × g. The resulting supernatant (cell free extract) was heated for 30 min at 80°C and subsequently centrifuged for 20 min at 10.000 × g. The supernatant (heat stable cell free extract) was filtered (0.45 μm) and applied to a Q-sepharose high performance (GE Healthcare, Chalfont, St. Giles, UK) column (1.6 x 10 cm) equilibrated in 20 mM Tris/HCl buffer (pH 7.8). Proteins were eluted with a linear 560 ml gradient from 0.0 to 1.0 M NaCl, in the same buffer. Wild type and potential high-performance mutants AdhA were produced and purified for further characterization.

SDS-PAGE electrophoresis Protein composition was analyzed by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE, 10%) (Sambrook et al., 1989), using a Mini-Protean 3 system (Biorad, Hercules, CA, USA). Protein samples for SDS-PAGE were prepared by heating for 5 min at 100°C in the presence of sample buffer (0.1 M sodium phosphate buffer, 4% SDS, 10% 2-mercaptoethanol, 20% glycerol, pH6.8). A broad range protein marker (Biorad) was used to estimate the molecular mass of the proteins.

90 Random mutagenesis P. furiosus AdhA

Characterization of high-performance mutants

Activity assays Rates of alcohol oxidation and aldehyde reduction were determined at 30°C, unless stated otherwise, by following either the reduction of NADP+ or the oxidation of NADPH at 340 nm using a Hitachi U2010 spectrophotometer, with a temperature controlled cuvette holder. Each oxidation reaction mixture contained 50 mM glycine (pH 10.5), 125 mM of alcohol and 0.35 mM NADP+. The reduction reaction mixture contained 0.1 M sodium phosphate buffer (pH 7.0), 125 mM aldehyde or ketone and 0.35 mM NADPH. In all assays the reaction was initiated by addition of an appropriate amount of enzyme. One unit of ADH was defined as the oxidation or reduction of 1 μmol of NADPH or NADP+ per min, respectively. Protein concentration was determined using Bradford reagents (Bio-Rad) with bovine serum albumin as a standard (Bradford, 1976). The temperature dependent spontaneous degradation of NADPH was corrected for.

Optimum temperature and thermostability The thermostability of AdhA mutants (enzyme concentration: 0.36 mg·mL-1 in 20 mM Tris buffer pH 7.8) was determined by measuring the residual activity (2,5-hexanedione reduction according to the standard assay) after incubation of a time series at 100°C. The temperature optimum was determined in 0.1 M sodium phosphate buffer pH 7.0 by analysis of initial rates of 2,5-hexanedione reduction in the range of 30 -100°C.

Kinetics

The kinetic parameters Km and Vmax were calculated from multiple measurements (at least 8 measurements, which were done in duplo) using the Michaelis-Menten equation and the program Tablecurve 2D (version 5.0). All the reactions followed Michaelis-Menten type -1 kinetics. The turnover number (kcat, s ) was calculated as: Vmax ∗ subunit molecular mass (26 kDa) / 60.

DNA sequencing and three-dimensional structure analysis Plasmid DNA was isolated from 5 ml cultures by using the QIAprep spin miniprep kit (Qiagen) and the sequence of the different clones was determined by sequence analysis of both DNA strands (BaseClear). Amino acid substitutions were deduced from identified mutations in the DNA sequence of the adhA gene. A structural model of AdhA was constructed by homology modelling using the crystal structures of the NADP+-dependent clavulanic acid dehydrogenase from Streptomyces clavuligerus as a template (PDB-ID: 2JAP and 2JAH) (MacKenzie et al., 2007). The primary sequence of this enzyme aligns with the complete AdhA sequence while sharing 30% sequence identity. The aligned sequences were modelled using SwissModel (Schwede et al., 2003), an automated comparative protein modeling server. The validity of the model was checked by WhatCheck (Hooft et al., 1996).

91 Chapter 6

The model was compared to several homologous short chain ADH structures, i.e. to the sepiapterin reductase of Chlorobium tepidum (26% identity, CT0609, PDB-ID 2BD0) (Supangat et al., 2006), the gluconate-5-dehydrogenase of Thermotoga maritima (22% identity, TM0441, PDB-ID 1VL8), and the human 17-β-hydroxysteroid dehydrogenase (23% identity, PDB-ID 1A27) (Breton et al., 1996; Mazza et al., 1998).

Results

Construction of a random mutant AdhA library and screening for increased activity on 2,5-hexanedione at 30°C The hyperthermostable short-chain alcohol dehydrogenase AdhA from P. furiosus is optimally active at 90°C and its activity at 30°C does not exceed 5 % of its optimal activity. AdhA has activity on the substrate 2,5-hexanedione and reduces it stereoselectively to (2S,5S)-hexanediol. However, the activity on 2,5-hexanedione at 30°C, the preferred temperature for production of (2S,5S)-hexanediol in an industrial setting, is very low. To study the possibility of increasing this low activity at 30°C, while retaining the stability of the enzyme, AdhA was subjected to random mutagenesis and low-temperature activity screening. Random mutations were introduced into the adhA gene by error-prone PCR (Leung et al., 1989; Cadwell & Joyce, 1992). The mutated adhA genes were cloned in an E. coli expression host and a library of approximately 1500 clones was constructed. Complete DNA sequence analysis of nine randomly selected mutants revealed an average mutation frequency of 2.75 base pairs / adhA gene, indicating that on average 2 or 3 amino acids will be changed in each AdhA enzyme (data not shown). The mutant AdhA library was screened for increased activity on 2,5-hexanedione at 30°C. This resulted in the identification of 30 mutants with significantly higher activity than the wild type controls. Heat-stable cell free extracts of the 30 mutants were further analyzed for increased low-temperature activity on 2,5-hexanedione. Eventually, the 7 most active mutants (harboring plasmids pWUR314 – pWUR320) were selected. Wild type and high- performance AdhA mutants were produced and purified for further characterization.

Characterization of high-performance mutants To gain more insight in the performance of the mutants at low and high temperature their activity with different substrates was tested at 30°C and 70°C (Table 6.1). The AdhA mutant 13F2 performed the best at 30°C with all substrates. The maximum specific activity of 13F2 with 2,5-hexanedione at 30°C was 10-fold higher compared to the activity of the wild type enzyme. At 70°C the wild type enzyme still performed the best. All mutants, with the exception of 16H2, also had an increased activity with 2-propanol at 30°C, which is important for the cofactor regeneration system. Previous experiments have shown that wild type AdhA can perform substrate coupled NADPH regeneration making use of the oxidation of 2-

92 Random mutagenesis P. furiosus AdhA propanol to acetone (Fig. 6.1). Further experiments with 2-propanol showed that all mutants had retained their activity in the presence of an excess of 2-propanol (data not shown). The temperature optimum was determined for the mutants 13F2, 5E12 and 6G6, the three mutants with the highest activity at 30°C with 2,5-hexanedione, and the wild type enzyme (Fig. 6.2). The temperature optimum of the wild type enzyme was determined to be between 90°C and 100°C, which is comparable to the optimum of 90°C found previously (Van der Oost et al., 2001). The temperature optimum of the three mutants is not changed (90°C), but their activity at lower temperature (60°C and below) is clearly increased when compared to the wild type enzyme. Next, the stability of the best performing mutant, 13F2, was compared with the wild type enzyme. Therefore, the thermostability of both enzymes was tested at 100°C. The wild type enzyme had an extremely high resistance to thermal inactivation, which was shown by a half-life value of 178 min. The mutant 13F2 with a half life value of 44 min. was clearly less stable than the wild type enzyme, but it still exhibits a very high resistance to thermal inactivation. The kinetic properties at 70°C of wild type enzyme and mutant 13F2 were determined for the substrates 2,5-hexanedione, pyruvaldehyde, (2S,5S)-hexanediol and 2-pentanol, as well as for the cofactors used in these reactions (Table 6.2). The wild type enzyme showed a higher affinity for all substrates and mutant 13F2 had merely a higher affinity for the cofactor NADPH, which is used in the reduction reaction.

Table 6.1 Specific activity (U·mg-1) of wild type AdhA and high-performance mutants Substrate Temperature Wild (°C) type 4A2 16H3 13F2 5E12 6G6 16H2 17A3

Reduction 2,5- 30 0.09 0.14 0.12 0.95 0.20 0.18 0.12 0.08 Hexanedione 70 4.55 1.26 2.45 3.46 2.17 1.87 0.50 1.84 Pyruval- 30 0.73 1.22 3.37 7.97 1.07 3.00 0.78 1.12 dehyde 70 42.1 30.6 15.6 57.0 18.0 22.0 17.3 15.8

Oxidation (2S,5S)- 30 0.91 1.09 0.71 3.23 1.24 1.31 0.07 1.10 Hexanediol 70 14.6 13.5 10.7 10.6 8.00 12.2 0.62 8.74 2-Propanol 30 0.11 0.19 0.20 1.11 0.43 0.17 0.07 0.15 30 1.47 1.30 1.11 3.47 0.73 0.86 0.06 0.65 2-Pentanol 70 10.1 14.3 7.69 9.28 2.79 10.5 0.10 7.86 30 0.16 0.33 0.60 1.98 0.57 0.24 0.09 0.35 2-Butanol 70 5.93 3.47 1.25 6.65 1.48 3.18 0.003 3.00 All measurements were performed at least in duplo.

93 Chapter 6

Figure 6.2 Optimum temperature curve for wild type AdhA (○), and the mutants 5E12 (×), 6G6 (∗) and 13F2 (□).

Table 6.2 The kinetic properties at 70°C of wild type enzyme and mutant 13F2 Wild type AdhA 13F2

Substrate / Km Vmax kcat/Km Km Vmax kcat/Km cofactor (mM) (U⋅mg-1) (s-1⋅mM-1) (mM) (U⋅mg-1) (s-1⋅mM-1) 2,5- hexanedione 56.4 ± 0.6 1.9 ± 0.1 0.01 79.0 ± 1.7 2.9 ± 0.4 0.02 pyruvaldehyde 1.2 ± 0.03 10.4 ± 0.2 3.8 4.2 ± 0.1 52.7 ± 0.5 5.4 (2S,5S)- hexanediol 5.4 ± 0.3 12.8 ± 0.4 1.0 327.6 ± 9.6 28.1 ± 1.1 0.04

2-pentanol 30.3 ± 0.4 13.1 ± 0.8 0.1 51.1 ± 0.9 24.7 ± 2.1 0.2

NADPHa 0.4 ± 0.02 10.3 ± 0.07 - 0.1 ± 0.01 48.3 ± 0.9 -

NADP+ 0.2 ± 0.03 12.8 ± 0.2 - 0.7 ± 0.08 37.3 ± 1.0 - All measurements were performed in duplo. aThe kinetic parameters for NADPH and NADP+ were determined using respectively pyruvaldehyde and 2-pentanol as substrate.

94 Random mutagenesis P. furiosus AdhA

Table 6.3 Amino acid substitutions in high-performance mutants Plasmid AdhA Amino acid mutant substitution pWUR314 4A2 V66A L176P Y229H

pWUR315 16H3 A18P K40N Q165H D182H E202G R219M

pWUR316 13F2 R11L A180V

pWUR317 5E12 L176P

pWUR318 6G6 T153A

pWUR319 16H2 N86D R213I

pWUR320 17A3 K209T

Amino acid substitutions in high-performance mutants As with all members of the short-chain dehydrogenase/reductase (SDR) family, AdhA from P. furiosus contains a well conserved NADP(H)-binding site at the N-terminus (Gly8-X-X-X- Gly12-X-Gly14), as well as several residues that are directly or indirectly involved in catalysis, including Asp59, Asn110, Ser137, Tyr150, Lys154, Thr183 (Van der Oost et al., 2001; Filling et al., 2002). A three-dimensional model of AdhA was constructed using the crystal structure of the NADP+-dependent clavulanic acid dehydrogenase from Streptomyces clavuligerus (MacKenzie et al., 2007). The 7 most active mutants were sequenced and amino acid substitutions were deduced from the changes in the DNA sequence (Table 6.3). This revealed that the high performance mutants (13F2, 5E12 and 6G6) contained 1 or 2 amino acid substitutions. The mutant 13F2 has the amino acid substitution R11L, which is located in the NADP(H)-binding site (Fig.

95 Chapter 6

6.3). Although the mutation is not at a conserved position of the NADP(H)-binding motif, the kinetic data supports the hypothesis that the mutation enhances cofactor binding affinity and turnover rate, as shown by a lowered Km and increased Vmax. Moreover, according to Wierenga et al. (1986) a hydrophobic residue such as leucine is preferred at this position in the NADP(H)-binding site. The second substitution of mutant 13F2, A180V, is located in the substrate binding pocket near the conserved Thr183, which is involved in catalysis (Fig. 6.3). A valine at this position is also observed in several closely related AdhA homologs from mesophilic organisms. Mutation L176P is observed in two independent mutants (5E12 and 4A2) and the residue is located in the β-strand preceding an active site loop containing Thr183 (Fig. 6.3). Proline is most commonly found in turns, loops, the edge strands of β- sheets and the beginning or end of α-helices. Possibly this substitution is involved in a repositioning of the subsequent active site loop. While the introduction of proline residues at certain positions (in loop regions) has been associated with increased (thermo)stability and rigidification (Vieille & Zeikus, 2001), in AdhA it appears to enhance low-temperature activity. The amino acid substitution in mutant 6G6, T153A, is located near the residues Tyr150 and Lys154, which are both members of the conserved catalytic tetrad in short-chain alcohol dehydrogenases (Fig. 6.3) (Filling et al., 2002). This mutation appears to have a conformational effect close to the active site, which is supported by the observation that this mutant can accept more bulky substrates than the wild type enzyme (T. Dauβmann, personal communication). Mutation N86D that occurs in mutant 16H2 is also located in the cofactor binding site, adding negative charge to the bottom of the pocket. Finally, the mutation R213I and K209T that occur in two different mutants (16H2 and 17A3) are spatially very close in the tertiary structure. This suggests that the enhanced activity of these mutants at low- temperature is due to a loss of positive charge at this site. Interestingly, close inspection of the model reveals two carboxyl residues (E23 and D24) in the C-terminal part of the neighbouring α-helix within ion pair forming range of the R213 and K209. Most likely the increased activity at low-temperature is due to the loss of this ionic interaction, which possibly makes the enzyme more flexible at lower temperature. This observation correlates well with the increased occurrence of ion pairs in proteins from hyperthermophiles compared to mesophiles (Karshikoff & Ladenstein, 2001). The stabilizing feature is apparently rapidly lost when the selection pressure for extreme thermostability is relieved, and low-temperature activity is screened for.

96 Random mutagenesis P. furiosus AdhA

Figure 6.3 Ribbon representation of the AdhA structural model based on the clavulanic acid dehydrogenase (PDB-ID 2JAP and 2JAH) from Streptomyces clavuligerus (MacKenzie et al., 2007). Mutated amino acids and the NADP+-cofactor are indicated as sticks.

97 Chapter 6

Discussion

Our results show that the probed sequence space in the AdhA library was sufficient and the screening method suitable to detect mutants with an increase in low-temperature activity. The high-performance mutant 13F2 has a 10-fold increase in low-temperature activity on 2,5- hexanedione. Its stability compared to the wild type enzyme has been slightly compromised, but it still exhibits a high resistance to thermal inactivation. The decrease in stability could be the result of increased flexibility, which might be the reason for enhanced low-temperature activity (Shiraki et al., 2001; Vieille & Zeikus, 2001; Liang et al., 2004). It was also shown that the substrate coupled NADPH regeneration using 2-propanol could still be used. The amino acid substitutions in the most enhanced mutants revealed that there are several adaptations in different regions of the enzyme, which can enhance the low- temperature activity. Regions of the enzyme where the mutations were found to affect the polypeptide chain include the substrate binding pocket, the cofactor binding site and near the catalytic site. Lebbink et al. (2000) reported a similar pattern when they improved the low- temperature catalysis of the β-glucosidase (CelB) from P. furiosus. Characterization of the latter CelB mutants showed that the increase in low-temperature activity was achieved in different ways, including altered substrate specificity and increased flexibility. In conclusion, the laboratory evolution approach has again been successful in engineering an improved enzyme. The best mutant (13F2) has an increased low-temperature activity, retained a high level of stability and can still use the efficient cofactor regeneration process, substrate coupled NADPH regeneration, at the lower temperature. This high- performance mutant can be used for the enzymatic conversion of 2,5-hexanedione to (2S,5S)- hexanediol or as a new starting point for further improvement.

Acknowledgements

This work was supported by the EU 5th framework program PYRED (QLK3-CT-2001- 01676).

References

Bastos FM, Dos Santos AG, Jones J, Oestreicher EG, Pinto GF & Paiva LMC (1999) Three different coupled enzymatic systems for in situ regeneration of NADPH. Biotechnol Tech 13: 661-664. Bornscheuer UT & Pohl M (2001) Improved biocatalysts by directed evolution and rational protein design. Curr Opin Chem Biol 5: 137-143. Bradford MM (1976) A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 72: 248-254.

98 Random mutagenesis P. furiosus AdhA

Breton R, Housset D, Mazza C & Fontecilla-Camps JC (1996) The structure of a complex of human 17-β-hydroxysteroid dehydrogenase with estradiol and NADP+ identifies two principal targets for the design of inhibitors. Structure 4: 905-915. Brunel JM & Faure B (2004) Enantioselective palladium catalyzed allylic substitution with a new phosphite ligand issued from (2S,5S)-hexanediol. J Mol Catal A: Chem 212: 61-64. Cadwell RC & Joyce GF (1992) Randomization of genes by PCR mutagenesis. PCR Methods Appl 2: 28-33. Eckstein M, Dauβmann T & Kragl U (2004) Recent Developments in NAD(P)H regeneration for enzymatic reductions in one- and two-phase systems. Biocatal Biotransform 22: 89-96. Fiala G & Stetter KO (1986) Pyrococcus furiosus sp. nov. represents a novel genus of marine heterotrophic archaebacteria growing optimally at 100oC. Arch Microbiol 145: 56-61. Filling C, Berndt KD, Benach J, Knapp S, Prozorovski T, Nordling E, Ladenstein R, Jörnvall H & Oppermann U (2002) Critical residues for structure and catalysis in short-chain dehydrogenases/reductases. J Biol Chem 277; 25677-25684. Haberland J, Hummel W, Daussmann T & Liese A (2002a) New continuous production for enantiopure (2R,5R)-hexanediol. Organic Process Research & Development 6: 458-462. Haberland J, Kriegesmann A, Wolfram E, Hummel W & Liese A (2002b) Diastereoselective synthesis of optically active (2R,5R)-hexanediol. Appl Microbiol Biotechnol, 58: 595-599. Hooft RW, Vriend G, Sander C & Abola EE (1996) Errors in protein structures. Nature 381(6580): 272. Hummel W (1999) Large-scale applications of NAD(P)-dependent oxidoreductases: recent developments. Trends Biotechnol 17: 487-492. Karshikoff A & Ladenstein R (2001) Ion pairs and the thermotolerance of proteins from hyperthermophiles: a "traffic rule" for hot roads. Trends Biochem Sci 26: 550-556. Kim R, Sandler SJ, Goldman S, Yokota H, Clark AJ & Kim SH (1998) Overexpression of archaeal proteins in Escherichia coli. Biotechnol Lett 20: 207-210. Lebbink JHG, Kaper T, Bron P, Van der Oost J & De Vos WM (2000) Improving low- temperature catalysis in the hyperthermostable Pyrococcus furiosus β-glucosidase CelB by directed evolution. Biochemistry 39: 3656-3665. Leung DW, Chen E & Goeddel DV (1989) A method for random mutagenesis of a defined DNA segment using a modified polymerase chain reaction. Technique 1: 11-15. Liang ZX, Tsigos I, Bouriotis V & Klinman JP (2004) Impact of protein flexibility on hydride- transfer parameters in thermophilic and psychrophilic alcohol dehydrogenases. J Am Chem Soc 126: 9500-9501. Lieser JK (1983) A simple synthesis of (S,S)-2,5-hexanediol. Synth Commun 13: 765-767. MacKenzie AK, Kershaw NJ, Hernandez H, Robinson CV, Schofield CJ & Andersson I (2007) Clavulanic acid dehydrogenase: structural and biochemical analysis of the final step in the biosynthesis of the beta-lactamase inhibitor clavulanic acid. Biochemistry 46: 1523-1533. Mazza C, Breton R, Housset D & Fontecilla-Camps JC (1998) Unusual charge stabilization of NADP+ in 17-β-hydroxysteroid dehydrogenase. J Biol Chem 273: 8145-815. Radianingtyas H & Wright PC (2003) Alcohol dehydrogenases from thermophilic and hyperthermophilic archaea and bacteria. FEMS Microbiol Rev 27: 593-616. Reid MF & Fewson CA (1994) Molecular characterization of microbial alcohol dehydrogenases. Crit Rev Microbiol 20: 13-56.

99 Chapter 6

Sambrook J, Fritsch EF & Maniatis T (1989) Molecular cloning: A laboratory manual, 2nd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Schmid A, Dordick JS, Hauer B, Kiener A, Wubbolts M & Witholt B (2001) Industrial biocatalysis today and tomorrow. Nature 409: 258-268. Schoemaker HE, Mink D & Wubbolts MG (2003) Dispelling the myths - biocatalysis in industrial synthesis. Science 299: 1694-1697. Schwede T, Kopp J, Guex N & Peitsch MC (2003) SWISS-MODEL: An automated protein homology-modeling server. Nucleic Acids Res 31: 3381-3385. Shiraki K, Nishikori S, Fujiwara S, Hashimoto H, Kai Y, Takagi M & Imanaka T (2001) Comparative analyses of the conformational stability of a hyperthermophilic protein and its mesophilic counterpart. Eur J Biochem 268: 4144-4150. Sorensen HP, Sperling-Petersen HU & Mortensen KK (2003) Production of recombinant thermostable proteins expressed in Escherichia coli: completion of protein synthesis is the bottleneck. J Chromatogr B Analyt Technol Biomed Life Sci 786: 207-214. Supangat S, Seo KH, Choi YK, Park YS, Son D, Han CD & Lee KH (2006) Structure of Chlorobium tepidum Sepiapterin Reductase complex reveals the novel substrate binding mode for stereospecific production of L-threo-tetrahydrobiopterin. J Biol Chem 281: 2249-2256. Van der Oost J, Voorhorst WG, Kengen SWM, Geerling ACM, Wittenhorst V, Gueguen Y & De Vos WM (2001) Genetic and biochemical characterization of a short-chain alcohol dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus. Eur J Biochem 268: 3062-3068. Vieille C & Zeikus GJ (2001) Hyperthermophilic enzymes: sources, uses, and molecular mechanisms for thermostability. Microbiol Mol Biol Rev 65: 1-43. Wierenga RK, Terpstra P & Hol WG (1986) Prediction of the occurrence of the ADP-binding beta alpha beta-fold in proteins, using an amino acid sequence fingerprint. J Mol Biol 187: 101-107. Yuan L, Kurek I, English J & Keenan R (2005) Laboratory-directed protein evolution. Microbiol Mol Biol Rev 69: 373-392. Zeikus JG, Vieille C & Savchenko A (1998) Thermozymes: biotechnology and structure-function relationships. Extremophiles, 2: 179-183.

100

Chapter 7

Cofactor engineering of Lactobacillus brevis alcohol dehydrogenase by computational design

Machielsen R, Looger LL, Raedts J, Dijkhuizen S, Hummel W, Hennemann H, Dauβmann T & Van der Oost J

Submitted Chapter 7

Abstract

The R-specific alcohol dehydrogenase from Lactobacillus brevis (Lb-ADH) catalyses the enantioselective reduction of prochiral ketones to the corresponding secondary alcohols. It is stable and has broad substrate specificity. These features make this enzyme an attractive candidate for biotechnological applications. A drawback is its preference for NADP(H) as a cofactor, which is more expensive and labile than NAD(H). Structure-based computational protein engineering was used to predict mutations to alter the cofactor specificity of Lb-ADH. Mutations were introduced into Lb-ADH by site-directed mutagenesis and then tested against the substrate acetophenone, with either NAD(H) or NADP(H) as cofactor. The mutant Arg38Pro showed four-fold increased activity with acetophenone and NAD(H) relative to wild-type. Both Arg38Pro and wild-type exhibit a pH optimum of 5.5 with NAD(H) as cofactor, significantly more acidic than with NADP(H). These and related Lb-ADH mutants may prove useful for green synthesis of pharmaceutical precursors.

Introduction

Alcohol dehydrogenases (ADHs) are present in all organisms. They display a wide spectrum of substrate specificities and play important physiological roles (Reid & Fewson, 1994). There is considerable interest in alcohol dehydrogenases for a range of applications in the food, pharmaceutical and fine chemicals industries. The production of enantio- and diastereo- merically pure diols is particularly desirable because these are important fine chemical building blocks (Hummel, 1999). The R-specific alcohol dehydrogenase from Lactobacillus brevis (Lb-ADH) was previously identified, purified and characterized (Riebel, 1996). Moreover, the structure of Lb-ADH has been solved to atomic resolution (Niefind et al., 2000; Niefind et al., 2003; Schlieben et al., 2005). Lb-ADH, a member of the superfamily of short-chain dehydrogenases/reductases (SDR; Jornvall et al., 1995; Filling et al., 2002), is a homotetramer with a molecular mass of 26.6 kDa per subunit. Lb-ADH catalyses the enantioselective reduction of prochiral ketones to the corresponding secondary alcohols. It has a broad substrate specificity, even accepting ketones with bulky side chains. Therefore it can be used for the production of valuable synthons for asymmetric organic synthesis. Lb-ADH has a relatively high, Mg2+-dependent stability, and has a clear cofactor preference for NADP(H) over NAD(H) (Niefind et al., 2000; Niefind et al., 2003; Bommarius & Riebel, 2004; Schlieben et al., 2005). Lb-ADH is very attractive for industrial applications because of the above-mentioned stability, stereoselectivity and substrate specificity (Wolberg et al., 2001; Ernst et al., 2005; Hildebrand & Lutz, 2006). For industrial utility, however, NAD(H) cofactor specificity would be preferable to NADP(H), due to the greater stability and five- to ten-fold lower price of the former (Bommarius & Riebel, 2004). Protein engineering may be accomplished by rational,

102 Cofactor engineering of Lactobacillus brevis ADH typically structure-guided mutagenesis, by random laboratory evolution, or by combinations of both methods (Bornscheuer & Pohl, 2001; Yuan et al., 2005). Rational protein engineering has been used to improve many protein properties, e.g. stability, enantioselectivity, ligand and substrate specificity (Van den Burg et al., 1998; Van den Heuvel et al., 2000). During the past decade the available tools for rational engineering have been complemented with computational protein design (Looger et al., 2003; Dwyer et al., 2004; Korkegian et al., 2005). The atomic resolution (1.0 Å) structure of Lb-ADH (Schlieben et al., 2005) makes structure-guided protein engineering an attractive method for systematic enzyme improvement.

Materials and Methods

Chemicals and plasmids All chemicals (analytical grade) were purchased from Sigma-Aldrich (Munich, Germany) or Acros Organics (Geel, Belgium). The restriction enzymes were obtained from Invitrogen (Paisley, UK) and New England Biolabs (Ipswich, MA, USA). T4 DNA ligase, Pfu Turbo DNA polymerase and QuikChange Site-Directed Mutagenesis kit were purchased from Stratagene (Amsterdam, The Netherlands) and REDTaq DNA polymerase from Sigma- Aldrich. The pGEM-T Easy Vector System I (AmpR; Promega, Madison, WI, USA) was used for cloning. For heterologous expression the vector pET-24d (KanR; Novagen, Darmstadt, Germany) and the tRNA helper plasmid pSJS1244 (SpecR) (Kim et al., 1998; Sorensen et al., 2003) were used.

Organisms and growth conditions Escherichia coli XL1-Blue (Stratagene) was used as a host for the construction of pGEM-T and pET24d derivatives. E. coli BL21(DE3) (Novagen) harboring the tRNA helper plasmid pSJS1244 was used as an expression host. Both strains were grown under standard conditions (Sambrook et al., 1989) following the instructions from the manufacturer.

Modelling Lb-ADH in complex with NAD(H) cofactor and acetophenone substrate was modelled from the NADP(H)-containing crystal structure 1ZK4 (REF) by in silico replacement of the phosphate group with a hydroxyl, using the Chameleon software program (Looger, unpublished results). Molecular mechanical parameters for the NAD(H) cofactor were taken from Pavelites et al. (1997). Protein positions in proximity to the NAD(H) in the modelled complex were computationally mutated to all twenty naturally-occurring amino acids, and side-chains which were estimated to preferentially interact with NAD(H) over NADP(H) were selected for experimental characterization. The highest-scoring mutant at each of four protein positions (38, 90, 112, and 140; mutations were made separately) was constructed and tested.

103 Chapter 7

Site-directed mutagenesis The Lactobacillus brevis adh gene (759 bp) was PCR amplified from the plasmid pBTAC-R- ADH (Julich Chiral Solutions, Jülich, Germany) using the primers BG1858 (sense, 5’- GCGCGCCATGGCATCTAACCGTTTGGATGGTAAGG) and BG1859 (antisense, 5’- GCGCGGGATCCCTATTGAGCAGTGTAGCCACCGT), containing the NcoI and BamHI sites (underlined). In order to introduce an NcoI restriction site an extra alanine codon (GCA) was introduced into the adh gene by the forward primer BG1858 (boldface). The PCR fragment was purified using the QIAquick PCR purification kit (Qiagen, Hilden, Germany). A-tail cloning into pGEM-T was facilitated by addition of A-tails to the blunt PfuTurbo PCR product by REDTaq DNA polymerase. The sequence of the insert was verified by sequence analysis of both DNA strands (Baseclear, Leiden, The Netherlands). Subsequently, the pGEM-T-LbADH construct was digested with NcoI-BamHI and the resulting 764-bp fragment was cloned into E. coli XL1-Blue using an NcoI-BamHI digested pET24d vector. Next, the resulting plasmid pWUR321 was transformed into E. coli BL21(DE3) harbouring the tRNA helper plasmid pSJS1244. The Arg38Pro, Ala90Ser, Val112Asp, Met140Ile and Gly37Asp/Arg38Pro mutants were constructed using the Stratagene Quikchange site-directed mutagenesis kit, resulting in the plasmids pWUR328, pWUR325, pWUR326, pWUR327 and pWUR329 respectively. The plasmid pWUR321 was used as template for the construction of the Arg38Pro, Ala90Ser, Val112Asp, Met140Ile mutants, and pWUR328 was used as template for the construction of the double mutant Gly37Asp/Arg38Pro. Table 7.1 shows the primers used. Next, the plasmids were transformed into E. coli XL1-Blue and E. coli BL21(DE3) harbouring the tRNA helper plasmid pSJS1244.

Table 7.1 Primers used to construct Lb-ADH mutants. Bold residues denote the mutation site. Mutation Primer Primer sequence Arg38Pro BG1994 5’-CATGATTACCGGCCCGCACAGCGATGTTGG-3’ BG1995 5’-CCAACATCGCTGTGCGGGCCGGTAATCATG-3’ Ala90Ser BG1988 5’-GTTTCTACATTAGTTAATAACTCTGGGATCGCG GTTAACAAG-3’ BG1989 5’-CTTGTTAACCGCGATCCCAGAGTTATTAACTAA TGTAGAAAC-3’ Val112Asp BG1990 5’-GGCGTAAATTATTAGCCGACAACCTTGATGGTGTCT-3’ BG1991 5’-AGACACCATCAAGGTTGTCGGCTAATAATTTACGCC-3’ Met140Ile BG1992 5’-GGCTTCCATCATCAACATATCTTCGATCGAAGGCTTTG-3’ BG1993 5’-CAAAGCCTTCGATCGAAGATATGTTGATGATGGAAGCC-3’ Gly37Asp BG2035 5’-GGTCATGATTACCGACCCGCACAGCGATGTTGG-3’ & Arg38Pro BG2036 5’-CCAACATCGCTGTGCGGGTCGGTAATCATGACC-3’

104 Cofactor engineering of Lactobacillus brevis ADH

Screening of cell extracts E. coli BL21(DE3) harboring pSJS1244 and expression plasmids containing genes coding for either wild-type or mutant Lb-ADH genes Arg38Pro, Ala90Ser, Val112Asp, Met140Ile were grown in 5 ml Luria-Bertani medium supplemented with 50 µg/ml kanamycin and spectinomycin until a cell density of OD600 nm = 0.6 was reached. The cultures were then induced with 0.1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) and incubated at 37 °C for 3 hours. Cells were harvested, resuspended in 500 µl 20 mM Tris/HCl buffer (pH 7.8) with 10 mM MgCl2 and disrupted by sonication. Crude cell extract was centrifuged for 10 min. at 15.000 × g. The resulting supernatant (cell free extract) was used to determine the activity on acetophenone with NAD(P)H as cofactor (see below).

Production and purification of Lb-ADH Test tubes with 5 ml Luria-Bertani medium with kanamycin and spectinomycin (both 50 μg⋅ml-1) was inoculated with E. coli BL21(DE3) harboring pSJS1244 and expression plasmids and incubated overnight in a rotary shaker at 37 °C. Next, 1 ml of the preculture was used to inoculate 1 L Luria-Bertani medium with kanamycin and spectinomycin (both 50 mg⋅L-1) in a 2-L conical flask and incubated in a rotary shaker at 30 °C until a cell density of

OD600 nm = 0.6 was reached. The culture was then induced with 0.2 mM IPTG and incubated at 30 °C for 18 h. Cells were harvested, resuspended in 20 mM Tris/HCl buffer (pH 7.8) with

10 mM MgCl2 and passed twice through a French press at 110 MPa. The crude cell extract was centrifuged for 20 min at 10.000 × g. The resulting supernatant (cell free extract) was filtered (0.45 μm) and applied to a Q-sepharose high performance (GE Healthcare, Chalfont, St. Giles, UK) column (1.6 x 10 cm) equilibrated in 20 mM Tris/HCl buffer (pH 7.8) with 10 mM MgCl2. Proteins were eluted with a linear 560 ml gradient from 0.0 to 1.0 M NaCl, in the same buffer. Wild-type and Arg38Pro Lb-ADH were produced and purified for further characterization.

SDS-PAGE electrophoresis Protein composition was analyzed by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE, 10%) (Sambrook et al., 1989), using a Mini-Protean 3 system (Biorad, Hercules, CA, USA). Protein samples for SDS-PAGE were prepared by heating for 5 min at 100°C in the presence of sample buffer (0.1 M sodium phosphate buffer, 4% SDS, 10% 2-mercaptoethanol, 20% glycerol, pH 6.8). A broad range protein marker (Biorad) was used to estimate the molecular mass of the proteins.

Activity assays Rates of ketone reduction were determined at 30°C, unless stated otherwise, by following the oxidation of NAD(P)H at 340 nm using a Hitachi U2010 spectrophotometer, with a temperature controlled cuvette holder. Each reduction reaction mixture contained 0.1 M sodium phosphate buffer (pH 6.5), 1 mM MgCl2, 1.5 mM acetophenone and 0.3 mM

105 Chapter 7

NAD(P)H. In all assays the reaction was initiated by addition of an appropriate amount of enzyme. One unit of ADH was defined as the oxidation of 1 μmol of NAD(P)H per min. Protein concentration was determined using Bradford reagents (Bio-Rad) with bovine serum albumin as a standard (Bradford, 1976). The temperature-dependent spontaneous degradation of NAD(P)H was corrected for. pH dependence The pH dependence of ketone reduction was determined in a sodium phosphate buffer (100 mM, pH range 5.0 – 8.0) with 1 mM MgCl2. It was determined for Lb-ADH wild-type and mutant Arg38Pro with the cofactor NADH and NADPH.

Temperature dependence and thermostability The thermostability of Lb-ADH wild-type and mutant Arg38Pro (enzyme concentration: 0.11 mg·mL-1 in 20 mM Tris buffer pH 7.2) was determined by measuring the residual activity (acetophenone reduction according to the standard assay) after incubation of a time series at 30°C and 42°C. The temperature optimum was determined in 0.1 M sodium phosphate buffer pH 6.5 by analysis of initial rates of acetophenone reduction with NADH as cofactor in the range 30-60°C (according to the standard assay).

Kinetics

The kinetic parameters Km and Vmax were calculated from multiple measurements (at least 8 separate dilutions, each sampled twice) using the Michaelis-Menten equation and the program Tablecurve 2D (version 5.0). All the reactions followed Michaelis-Menten kinetics. The -1 turnover number (kcat, s ) was calculated as: Vmax ∗ subunit molecular mass (26.6 kDa) / 60.

Results

Modelling The R-specific alcohol dehydrogenase form Lactobacillus brevis has a clear preference for NADP(H) as cofactor. The phosphate group of NADP(H) interacts with both the side-chains and main-chain amides of Arg38 and Thr15 (Fig. 7.1 A), stabilizing this cofactor significantly better than NAD(H), which would leave a large cavity and unsatisfied hydrogen bonding capacity (Schlieben et al., 2005). In the present study, the aim was to obtain NAD(H)- dependent variants of the Lb-ADH. For that purpose, automated computational protein design was applied to predict mutations which would pack the cavity, and remove hydrogen bonding groups left unsatisfied in the absence of the phosphate group. The modelling resulted in four mutations: Arg38Pro, Ala90Ser, Val112Asp, Met140Ile (Table 7.1). Arg38Pro is modelled to remove both the side-chain guanidinium and main-chain amide groups to which the NADP(H) phosphate group hydrogen bonds in the wild-type structure, packing in the cavity left by the removal of the phosphate group and rigidifying the γ-turn

106 Cofactor engineering of Lactobacillus brevis ADH

A)

B)

C)

Figure 7.1 Structures of the wild-type enzyme/NADP, Gly37Asp/NAD, and Arg38Pro/NAD complexes. (A) Crystal structure of the wild-type enzyme with NADP cofactor. The cofactor phosphate moiety hydrogen bonds with two backbone amide hydrogens and Thr15; favourable electrostatic contact is made with Arg38. (B) Crystal structure of Gly37Asp with NAD cofactor. The aspartate carboxylate hydrogen bonds with a backbone amide and with the NAD, partially filling the phosphate cavity. Favourable electrostatic contact is restored with Arg38. (C) Structural model of Arg38Pro with NAD cofactor. The proline side-chain replaces the unsatisfied arginine side-chain and backbone amide hydrogen, and partially fills the phosphate cavity.

107 Chapter 7 from which the side-chain extends (Fig. 7.1 C). Ala90Ser is predicted to improve a hydrogen bonding network with the NAD(H) 2’-hydroxyl and the side-chains Thr12 and Ser63. The mutation Val112Asp is modelled to hydrogen bond with the adenosine base of NAD(H) and to electrostatically repel the NADP(H) phosphate group, increasing NAD(H) specificity. Met140Ile is predicted to repack the NAD(P)H cofactor more closely than the wild-type side- chain, preferentially creating steric conflict in the NADP(H) complex versus that of NAD(H).

Site-directed mutagenesis and initial screening The activity of wild-type Lb-ADH and mutants Arg38Pro, Ala90Ser, Val112Asp, Met140Ile (cell free extract) on acetophenone with NAD(H) and NADP(H) as cofactor was determined. Activity of the mutants Arg38Pro, Val112Asp, and Met140Ile was virtually abolished with NADP(H) as cofactor; activity of Ala90Ser with NADP(H) was approximately half of wild- type. Although wild-type Lb-ADH was originally reported to have no detectable activity at neutral pH against acetophenone with NAD(H) (Schlieben et al., 2005), the initial screening of the present study did reveal activity. Activity of Ala90Ser, Val112Asp, and Met140Ile with NAD(H) was diminished almost to background levels; only Arg38Pro showed a four-fold increase over wild-type Lb-ADH.

Characterization of Lb-ADH mutant Arg38Pro After purification to homogeneity, the pH optima of wild-type Lb-ADH and mutant Arg38Pro with NAD(H) and NADP(H) as cofactor were determined. Wild-type Lb-ADH showed a pH optimum between 6.0 and 6.5 for the reduction of acetophenone with NADP(H); optimal reduction with NAD(H) was shifted to pH 5.5 (Fig. 7.2; Bommarius & Riebel, 2004). The mutant Arg38Pro showed a pH optimum of 5.5 for the reduction reaction with both cofactors (Fig. 7.2). To further characterize the mutant Arg38Pro, the kinetic properties (at 30°C and optimal pH) of the mutant and wild-type enzyme were determined for the cofactors NAD(H) and NADP(H) using acetophenone as the substrate. Wild-type Lb-ADH and mutant Arg38Pro both have a relatively high affinity for NADP(H) (Km = 17 μM and 28 μM, respectively). Wild-type Lb-ADH and mutant Arg38Pro reached a comparable maximum specific activity, -1 -1 -1 -1 respectively 355 U/mg (kcat/Km = 9300 s · mM ) and 323 U/mg (kcat/Km = 5100 s · mM ).

Both wild-type and mutant enzyme have considerably lower affinity for NAD(H) (Km = 130 μM and 120 μM, respectively). In agreement with the analysis of the cell free extracts, the maximum specific activity with NAD(H) as cofactor is approximately four-fold higher with -1 -1 mutant Arg38Pro (Vmax = 80.3 U/mg, kcat/Km = 300 s · mM ) than wild-type Lb-ADH (Vmax -1 -1 = 22.0 U/mg, kcat/Km = 75 s · mM ). Both wild-type and mutant Lb-ADH showed a high level of activity over a wide temperature range with a temperature optimum of approximately 45°C. A temperature optimum of wild-type Lb-ADH of 55°C was previously reported (Bommarius & Riebel, 2004). The thermostability of both enzymes was determined at 30°C and 42°C (wild-type

108 Cofactor engineering of Lactobacillus brevis ADH with NADP(H); Arg38Pro with NAD(H)). Arg38Pro/NAD(H) showed higher thermostability at both temperatures (half-life 220 hours (30°C) and 3 hours (42°C)) relative to wild- type/NADP(H) (half-life 160 hours (30°C) and 1.25 hours (42°C)).

A 100 80 60 Arg38Pro 40 Lb-ADH 20

relative activity (%) activity relative 0 4.5 5 5.5 6 6.5 7 7.5 8 pH

B

100 80 60 Arg38Pro 40 Lb-ADH 20

relative activity (%) 0 4.5 5 5.5 6 6.5 7 7.5 8 pH

Figure 7.2 The pH dependence of ketone reduction for Lb-ADH wild-type and mutant Arg38Pro with the cofactor NADPH (A) and NADH (B).

Double mutant Gly37Asp/Arg38Pro Previously, a Lb-ADH mutant Gly37Asp with altered cofactor specificity has been reported (Bommarius & Riebel, 2004). An aspartate residue at position 37 introduces a negatively charged carboxylate group at the position in the Lb-ADH model that superimposes with the phosphate group in the Lb-ADH/NADP(H) structure (Fig. 7.1 B); steric and electrostatic forces lead to preference for the NAD(H) cofactor. Although the relative specificity for NAD(H) was improved, the specific activity of the enzyme decreased (Bommarius & Riebel,

109 Chapter 7

2004; Schlieben et al., 2005). In the present study, the NAD(H)-favoring mutations Gly37Asp and Arg38Pro were combined together into a double mutant; activity with NAD(H) was decreased relative to Arg38Pro alone, but is still higher than that of the wild-type enzyme.

Discussion

Previously, it has been shown that the cofactor specificity of Lb-ADH could be changed to favour NAD(H) over NADP(H) as cofactor. However, the Gly37Asp mutant described was insufficient for industrial applications, as the reaction rate of Gly37Asp/NAD(H) is lower than wild-type/NAD(H) (Bommarius & Riebel, 2004; Schlieben et al., 2005). Analysis of the crystal structure of Gly37Asp/NAD(H) suggests that the mutation primarily creates unfavourable interactions with NADP(H), rather than forming favourable interactions with NAD(H); this explains the large increase in cofactor specificity with a concomitant drop in reaction rate. Our results show that while the affinity for the cofactor NAD(H) does not change, mutant Arg38Pro/NAD(H) has a four-fold increased reaction rate compared to wild- type/NAD(H). Molecular modelling suggests that the removal of both the guanidinium side- chain and the amide proton by mutation to proline relieves these two unsatisfied hydrogen bond donors in complex with NAD(H), in addition to packing the cavity left by the phosphate. Furthermore, the main-chain conformation of Lb-ADH position 38 is in a highly- favoured region for proline, and position 37 is in a favourable region for “pre-proline” (a residue before a proline) (Lovell et al., 2003). The mutation Arg38Leu was previously discovered in a screen for mutants of Lb-ADH active with NAD(H) (Bommarius & Riebel, 2004); this mutation is modelled to relieve the unsatisfied guanidinium hydrogen bond donor, but to leave the unsatisfied amide proton and to poorly pack the phosphate cavity. The thermostability of Arg38Pro/NAD(H) is significantly higher than wild- type/NADP(H) at both 30°C and 42°C. It is possible that the Arg38Pro mutation rigidifies the Lb-ADH active site, rendering it more resistant to thermal perturbation. The pH optimum of reduction of acetophenone by wild-type/NADP(H) is 6.5; reduction by wild-type/NAD(H) is optimal at pH 5.5 (Fig. 7.2), perhaps explaining why previous neutral-pH experiments showed wild-type Lb-ADH to have negligible activity with NAD(H) (Schlieben et al., 2005). The mutation Arg38Pro shifts the pH optimum for both reduction reactions to pH 5.5 (Fig. 7.2), similar to that reported for the mutant Gly37Asp (Bommarius & Riebel, 2004). Structural analysis suggests that the electrostatic interaction between Arg38 and the NADP(H) phosphate favours neutral pH to optimize the charge-charge interaction, whereas removal of this electrostatic interaction in Arg38Pro/NAD(H) facilitates the higher reaction rates of the proton-dependent acetophenone reduction at lower pH before overall protein stability is compromised by large-scale protonation of acidic groups. A double mutant Gly37Asp/Arg38Pro was made to combine the two mutations each shown to improve specificity for NAD(H) over NADP(H). The double mutant, although more

110 Cofactor engineering of Lactobacillus brevis ADH active in NAD(H)-dependent acetophenone reduction than wild-type, was less active than the single mutant Arg38Pro. Structural analysis of the modelled double mutant (not shown) suggests that the two mutations may be sterically incompatible. The computational design resulted in four mutations (Arg38Pro, Ala90Ser, Val112Asp and Met140Ile), which were suggested to improve activity with NAD(H) as cofactor. While three of the four mutations dramatically decrease NADP(H)-dependent cofactor reduction, only one (Arg38Pro) led to an increased rate of NAD(H)-dependent activity. Ala90Ser is modelled to improve a hydrogen bonding network adjacent to the adenosine moiety of NAD(H), but perhaps the polar nature of the side-chain disrupts hydrophobic packing against the aromatic ring; the reason for its differential effect on NAD(H) and NADP(H) is unclear. The mutation Val112Asp is modelled to form a hydrogen bond with the adenosine amino group; it is likely, however, that electrostatic repulsion with the adjacent Asp62 destabilizes the active site and abolishes activity with both NADP(H) and NAD(H). Finally, the mutation Met140Ile, predicted to improve packing against an NAD(H) ribose ring, potentially interferes with the hydrogen bonding capacity of Asn89, leading to dramatically decreased activity with both cofactors. The mutant Arg38Pro has an NAD(H)-dependent activity that is four-fold higher than wild-type/NAD(H); nevertheless, this activity is still approximately five times less than wild- type/NADP(H). This is similar to the five-fold difference in price of the two cofactors, and as such this variant is an interesting alternative. Moreover, the elevated thermostability of the Arg38Pro/NAD(H) enzyme could allow for prolonged operation times. Additional rounds of iterative improvement, both by rational design and by directed evolution, will give rise to the step-wise generation of mutants with the desired chemical and physical features, including not only facile use of NAD(H) cofactor, but also increased specific activity, altered substrate specificity, and improved stability.

Acknowledgements

This work was supported by the EU 5th framework program PYRED (QLK3-CT-2001- 01676).

111 Chapter 7

References

Arnold FH (2001) Combinatorial and computational challenges for biocatalyst design. Nature 409(6817): 253-257. Bommarius AS & Riebel BR (2004) Biocatalysis: fundamentals and applications, 1 edn. Wiley- VCH, Weinheim. Bornscheuer UT & Pohl M (2001) Improved biocatalysts by directed evolution and rational protein design. Curr Opin Chem Biol 5(2): 137-143. Bradford MM (1976) A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 72: 248-254. Dwyer MA, Looger LL & Hellinga HW (2004) Computational design of a biologically active enzyme. Science 304(5679): 1967-1971. Ernst M, Kaup B, Muller M, Bringer-Meyer S & Sahm H (2005) Enantioselective reduction of carbonyl compounds by whole-cell biotransformation, combining a formate dehydrogenase and a (R)- specific alcohol dehydrogenase. Appl Microbiol Biotechnol 66(6): 629-634. Filling C, Berndt KD, Benach J, Knapp S, Prozorovski T, Nordling E, Ladenstein R, Jornvall H & Oppermann U (2002) Critical residues for structure and catalysis in short-chain dehydrogenases/reductases. J Biol Chem 277(28): 25677-25684. Hildebrand F & Lutz S (2006) Immobilisation of alcohol dehydrogenase from Lactobacillus brevis and its application in a plug-flow reactor. Tetrahedron: Asymmetry 17(23): 3219-3225. Hummel W (1999) Large-scale applications of NAD(P)-dependent oxidoreductases: recent developments. Trends Biotechnol 17(12): 487-492. Jornvall H, Persson B, Krook M, Atrian S, Gonzalez-Duarte R, Jeffery J & Ghosh D (1995) Short-chain dehydrogenases/reductases (SDR). Biochemistry 34(18): 6003-6013. Kim R, Sandler SJ, Goldman S, Yokota H, Clark AJ & Kim SH (1998) Overexpression of archaeal proteins in Escherichia coli. Biotechnology Letters 20(3): 207-210. Korkegian A, Black ME, Baker D & Stoddard BL (2005) Computational thermostabilization of an enzyme. Science 308(5723): 857-860. Looger LL, Dwyer MA, Smith JJ & Hellinga HW (2003) Computational design of receptor and sensor proteins with novel functions. Nature 423(6936): 185-190. Lovell SC, Davis IW, Arendall WB, 3rd, de Bakker PI, Word JM, Prisant MG, Richardson JS & Richardson DC (2003) Structure validation by Calpha geometry: phi,psi and Cbeta deviation. Proteins 50(3): 437-450. Niefind K, Muller J, Riebel B, Hummel W & Schomburg D (2003) The crystal structure of R- specific alcohol dehydrogenase from Lactobacillus brevis suggests the structural basis of its metal dependency. J Mol Biol 327(2): 317-328. Niefind K, Riebel B, Muller J, Hummel W & Schomburg D (2000) Crystallization and preliminary characterization of crystals of R-alcohol dehydrogenase from Lactobacillus brevis. Acta Crystallogr D Biol Crystallogr 56 Pt 12: 1696-1698. Pavelites JJ, Gao J, Bash PA & Mackerell J, A.D. (1997) A molecular mechanics force field for NAD+, NADH, and the pyrophosphate groups of nucleotides. J Comput Chem 18(2): 221-239. Reid MF & Fewson CA (1994) Molecular characterization of microbial alcohol dehydrogenases. Crit Rev Microbiol 20(1): 13-56.

112 Cofactor engineering of Lactobacillus brevis ADH

Riebel, B. (1996) Biochemische und molekularbiologische charakterisierung neuer mikrobieller NAD(P)-abhängiger alkoholdehydrogenasen. PhD thesis, Heinrich-Heine Universität, Düsseldorf. Sambrook J, Fritsch EF & Maniatis T (1989) Molecular cloning: A laboratory manual, 2nd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Schlieben NH, Niefind K, Muller J, Riebel B, Hummel W & Schomburg D (2005) Atomic resolution structures of R-specific alcohol dehydrogenase from Lactobacillus brevis provide the structural bases of its substrate and co-substrate specificity. J Mol Biol 349(4): 801-813. Sorensen HP, Sperling-Petersen HU & Mortensen KK (2003) Production of recombinant thermostable proteins expressed in Escherichia coli: completion of protein synthesis is the bottleneck. J Chromatogr B Analyt Technol Biomed Life Sci 786(1-2): 207-214. Van den Burg B, Vriend G, Veltman OR, Venema G & Eijsink VG (1998) Engineering an enzyme to resist boiling. Proc Natl Acad Sci U S A 95(5): 2056-2060. van Den Heuvel RH, Fraaije MW, Ferrer M, Mattevi A & van Berkel WJ (2000) Inversion of stereospecificity of vanillyl-alcohol oxidase. Proc Natl Acad Sci U S A 97(17): 9455-9460. Wolberg M, Hummel W & Muller M (2001) Biocatalytic reduction of beta, delta-diketo esters: a highly stereoselective approach to all four stereoisomers of a chlorinated beta, delta-dihydroxy hexanoate. Chemistry 7(21): 4562-4571. Yuan L, Kurek I, English J & Keenan R (2005) Laboratory-directed protein evolution. Microbiol Mol Biol Rev 69(3): 373-392.

113

Chapter 8

Summary and concluding remarks Chapter 8

Biocatalysis is gradually taking over from chemical catalysis in many industrial applications. Enzymes are environmentally friendly, biodegradable, efficient, and low cost in terms of resource requirements; as such they provide benefits compared to traditional chemical approaches in various industrial processes. Alcohol dehydrogenases (ADHs) are an example of such enzymes which are interesting from an industrial perspective. They can be enantioselective and display a wide variety of substrate specificities. Especially the enantio- and diastereoselective production of pure diols is desired. In many instances, however, natural enzymes are not suitable for large- scale industrial applications, because they do not perform optimally in such an unnatural process. There are three major and principally different routes to obtain enzyme variants with improved features: (1) isolating enzymes from organisms living in environments that correspond as much as possible to actual process conditions, (2) rationale-based protein engineering, and (3) laboratory evolution (random methods). The research described in this thesis has been focused on the identification and characterization of stable alcohol dehydrogenases from the hyperthermophilic archaeon Pyrococcus furiosus (route 1) and improvement of properties of selected alcohol dehydrogenases either by laboratory evolution or, when a high-resolution 3D structure is available, by rationale-based protein engineering (routes 2 and 3). A brief outline of this study is given in Chapter 1. In Chapter 2 a general introduction and a review on improving enzyme performance by the random approach of laboratory evolution as well as by rational and computational design are given. The available techniques for the engineering of enzymes, subdivided into directed (rational protein engineering) and random (laboratory evolution) techniques, are described. In addition, different selection and (high-throughput) screening methods are described, a crucial development that allows screening of large mutant libraries. Recent studies on improving enzyme performance by laboratory evolution and/or rationale-based protein engineering are also discussed. In Chapter 3 the identification of sixteen putative alcohol dehydrogenases of the hyperthermophilic archaeon Pyrococcus furiosus is described. Enzymes from hyperthermophiles, i.e. microorganisms that grow optimally above 80°C, generally display an extreme stability at high temperature, high pressure, as well as high concentrations of chemical denaturants; this feature generally makes hyperthermophiles a very good source of robust enzymes. The P. furiosus genome was analyzed for genes encoding putative alcohol dehydrogenases, which resulted in the identification of adhA, adhB (encoding the two previously characterized ADHs from P. furiosus) and sixteen genes that potentially encode additional alcohol dehydrogenases. The putative ADHs showed similarity with the UDP- glucose/GDP-mannose dehydrogenase family or the major groups of the NAD(P)-dependent alcohol dehydrogenases: short-chain ADHs, zinc-dependent ADHs, iron-activated ADHs and the aldo-keto reductase superfamily. For all identified genes a more detailed level of function prediction was attempted using STRING genomic context analysis, Pfam annotation and COG annotation. This demonstrated the broad range of physiological processes in which they

116 Summary and concluding remarks can play an important role and the wide spectrum of substrate specificities. However, experimental evidence is necessary to confirm the proposed substrate specificities and actual physiological functions. The genes for AdhA, AdhB and fifteen putative ADHs were therefore successfully cloned and overexpressed in Escherichia coli. Subsequently, an initial screening was performed in which AdhA, AdhB and two of the putative alcohol dehydrogenases (PF0991 and PF1960) showed activities. The substrates and conditions tested in the initial screening were limited (e.g. pH range, anaerobic conditions, substrate and cofactor concentrations, metal requirements) and a more elaborate screening will be required to gain insight into the physiological function of all putative alcohol dehydrogenases. The two putative alcohol dehydrogenases that showed activities in the initial screening, PF0991 and PF1960, were selected for a more detailed study. Hence, Chapter 4 describes the heterologous production, purification and characterization of a novel, thermostable alcohol dehydrogenase (AdhD; PF1960) from P. furiosus. AdhD is a member of the aldo-keto reductase (AKR) superfamily that consists of NAD(P)-dependent oxidoreductases. The members of this superfamily share a common (α/β)8-barrel fold, and generally are monomeric proteins that bind nicotinamide cofactor without a Rossmann-fold motif. AdhD was functionally produced in E. coli, and due to its stability at high temperatures, only two steps were needed for purification: heat-treatment and anion-exchange chromatography. The enzyme had a monomeric conformation with a molecular mass of 32 kDa. It showed a preference for NAD(H) as cofactor and a broad substrate specificity in the oxidation (pH optimum 8.8) and reduction reaction (pH optimum 6.1). A clear preference was observed for secondary alcohols rather than primary alcohols, but the highest activities were detected when polyols were used as substrate. Maximal specific activities were detected with 2,3-butanediol (108 U·mg-1) and diacetyl-acetoin (22.5 U·mg-1) in the oxidative and reductive reactions, respectively. Gas chromatography analysis indicated that AdhD produced mainly (S)-2-pentanol (enantiomeric excess, 89%) when 2-pentanone was used as substrate. The catalytic activity of the enzyme increased up to 100°C and its extreme thermostability was shown by a half-life value of 130 min. at 100°C. Analysis of the adhD locus revealed that the start codon of a gene coding for a tungsten-containing aldehyde oxidoreductase, wor4, is directly adjacent to the adhD stop codon. The clustering of these two genes was also observed in the related species Pyrococcus abyssi, Pyrococcus horikoshii and Thermococcus kodakaraensis. This suggests a role in which AdhD might use or produce aldehydes/ketones associated with the WOR4 activity. Kinetic data of AdhD showed high catalytic efficiency for the reduction reaction and high affinity for the substrate and cofactor involved in this reaction, which suggests that the physiological function of AdhD is the reduction of aldehydes or ketones. Although it cannot be ruled out that these substrates are formed by WOR4, this appears unlikely because all described tungsten-containing aldehyde oxidoreductases (AOR, FOR, GAPOR) are known to catalyze the unidirectional oxidation of aldehydes. Assuming a physiological link between AdhD and WOR4, two scenarios are possible. First of all, it is conceivable that AdhD

117 Chapter 8 oxidizes an alcohol to produce an aldehyde or ketone which is subsequently oxidized further by WOR4. However, because of the apparent preference of AdhD for the reduction reaction and WOR4 for the oxidation reaction, another putative scenario can be envisaged as well. It might be possible that an unidentified aldolase produces the substrates for AdhD and WOR4, for instance, a ketone and an aldehyde. Subsequently, the ketone could be reduced to an alcohol by AdhD and the aldehyde could be oxidized by WOR4. The latter enzyme has been purified from P. furiosus cells that were grown in the presence of elemental sulfur, but unfortunately no WOR4 activity has yet been detected. More experiments will be required to establish the physiological roles of both enzymes. Chapter 5 describes the heterologous production, purification and characterization of the L-threonine dehydrogenase (Pf-TDH; PF0991) from P. furiosus. Pf-TDH belongs to the polyol dehydrogenase (PDH) family, which is part of the medium-chain dehydrogenase/reductase (MDR) superfamily. These enzymes utilize NAD(P)H as cofactors, are homotetramers or homodimers, and usually contain one or two zinc atom(s) per subunit with catalytic and/or structural function. Pf-TDH was functionally produced in E. coli, and due to its stability at high temperatures, once more only two steps were needed for purification: heat-treatment and anion-exchange chromatography. The enzyme has a tetrameric conformation with a molecular mass of ≈ 155 kDa. Pf-TDH is specific for NAD(H) -1 as cofactor, and maximal specific activities were detected with L-threonine (103 U·mg ) and acetoin (3.9 U·mg-1) in the oxidative (pH optimum 10.0) and reductive (pH optimum 6.6) reactions, respectively. It also utilized L-serine and D-threonine as substrate, but could not oxidize other L-amino acids, primary alcohols or secondary alcohols. The substrate specificity revealed that Pf-TDH requires neither the amino group nor the carboxy group of L-threonine for activity, but the enzyme kinetics clearly showed a preference for L-threonine. Determinants of the Pf-TDH substrate specificity were shown and this revealed that the specific configuration of the substrate is clearly important, as was demonstrated by the difference in activity with L-threonine and D-threonine. Activity was significantly higher when the oxidisable substrate possesses a methyl group at C4 (L-threonine vs L-serine), and when it possesses either an amino or a hydroxyl group at C2, which is probably involved in correct positioning of the substrate molecule in the active site through hydrogen bonding (L- threonine and 2,3-butanediol vs 2-butanol). Although the carboxy group is not required for activity, a comparison between 3-hydroxybutyrate and 2-butanol as substrate revealed that it can have a distinct influence on the activity. It was shown that the enzyme requires bivalent cations such as Zn2+ and Co2+ for activity and contains at least one zinc atom per subunit. L-Threonine dehydrogenase plays an important role in L-threonine catabolism. It catalyzes the NAD(P)-dependent oxidation of L-threonine to 2-amino-3-oxobutyrate, which spontaneously decarboxylates to aminoacetone and CO2 or is cleaved in a CoA-dependent reaction by 2-amino-3-ketobutyrate coenzyme A lyase to glycine and acetyl-CoA. In addition, it has been shown in E. coli that these enzymes are responsible for the formation of threonine from glycine in vitro and in vivo. However, the primary role of this pathway is believed to be

118 Summary and concluding remarks threonine catabolism. Conserved context analysis followed by a BLAST search identified a possible 2-amino-3-ketobutyrate coenzyme A lyase in P. furiosus. Therefore it was suggested that the physiological role of Pf-TDH is the oxidation of L-threonine to 2-amino-3- oxobutyrate, which is probably converted into glycine by a 2-amino-3-ketobutyrate coenzyme A lyase. The engineering of a thermostable P. furiosus alcohol dehydrogenase (AdhA) to improve the production of (2S,5S)-hexanediol at moderate temperatures is described in Chapter 6. It has been shown that the thermostable AdhA stereoselectively reduces 2,5- hexanedione to (2S,5S)-hexanediol. Moreover, it has been found that in the presence of excess 2-propanol, AdhA can perform substrate-coupled NADPH regeneration, as has been described for other ADHs. Due to the need for a stable alcohol dehydrogenase with activity at low-temperature process conditions (30°C) for the production of (2S,5S)-hexanediol, it was chosen to engineer AdhA to improve the low-temperature activity of the enzyme on 2,5- hexanedione, while retaining its stability. It was shown that AdhA is optimally active between 90°C and 100°C, while its activity at 30°C does not exceed 5% of its optimal activity. As there is no high-resolution crystal structure available for AdhA, the random method of laboratory evolution was chosen as method to improve the low-temperature activity of the enzyme. One round of error-prone PCR and screening of ∼1500 mutants for increased activity on 2,5-hexanedione at 30°C was performed. Eventually, the 7 most active mutants were selected. Next, wild type and high-performance mutants were produced and purified for further characterization. The maximum specific activity of the best performing mutant (13F2) with 2,5-hexanedione at 30°C was 10-fold higher compared to the activity of the wild type enzyme. Its stability compared to the wild type enzyme has been slightly compromised, but it still exhibits a high resistance to thermal inactivation. The decrease in stability could be the result of increased flexibility, which might be the reason for enhanced low-temperature activity. It was also shown that the substrate coupled NADPH regeneration using 2-propanol could still be used. A structural model of AdhA was constructed and this revealed that the best performing mutant had one mutation in the well conserved NADP(H)-binding site (R11L), and a second mutation (A180V) near a catalytic and highly conserved threonine at position 183. Furthermore, the amino acid substitutions in the 7 best performing mutants revealed that there are several adaptations in different regions of the enzyme, which can enhance the low-temperature activity. Regions of the enzyme where the mutations were identified included the substrate binding pocket, the cofactor binding site, and the region near the catalytic site. The selected variant with the highest activity at 30°C can be used for the conversion of 2,5-hexanedione to (2S,5S)-hexanediol, or as a new starting point for further improvement. In Chapter 7 the engineering of the R-specific alcohol dehydrogenase from Lactobacillus brevis (Lb-ADH) to alter its cofactor specificity is described. Lb-ADH is R- specific, relatively stable and has a broad substrate specificity; these features make this enzyme an attractive candidate for biotechnological applications. A drawback is its preference

119 Chapter 8 for NADP(H) as a cofactor, which is more expensive and labile than NAD(H). As the atomic resolution (1.0 Å) structure of Lb-ADH is available, structure-based computational protein engineering was used to predict mutations to alter the cofactor specificity of Lb-ADH. The computational design was applied to predict mutations which would pack the cavity, and remove hydrogen bonding groups left unsatisfied in the absence of the phosphate group. This resulted in four mutations (Arg38Pro, Ala90Ser, Val112Asp and Met140Ile), which were suggested to improve activity with NAD(H) as cofactor. These predicted mutations were introduced into Lb-ADH and tested with the cofactors NAD(H) and NADP(H). Only Arg38Pro led to an increased rate of NAD(H)-dependent activity. It showed four-fold increased activity with NAD(H) as cofactor compared to wild type. It was shown that the pH optimum of wild type Lb-ADH for the reduction reaction was shifted towards pH 5.5 when NAD(H) was used as cofactor, while it has a pH optimum between 6.0 and 6.5 when NADP(H) is used. The mutant Arg38Pro showed a pH optimum of 5.5 for the reduction reaction with both cofactors. The thermostability of Arg38Pro was significantly higher than wild type at both 30°C and 42°C. It is possible that the Arg38Pro mutation rigidifies the Lb- ADH active site, rendering it more resistant to thermal perturbation. The affinity for the cofactor NAD(H) was not changed in Arg38Pro relative to wild type. A previously reported mutation (Gly37Asp) that altered the cofactor specificity, but decreased the reaction rate, was combined with the mutation Arg38Pro. In this double mutant, activity with NAD(H) was decreased relative to Arg38Pro alone, but it is still higher than that of the wild type enzyme. The mutant Arg38Pro has an NAD(H)-dependent activity that is four-fold higher than wild type/NAD(H). Nevertheless, this activity is still approximately five times less than wild type/NADP(H) activity. Mutant Arg38Pro could serve as starting point for additional rounds of iterative improvement, either by rational design or by laboratory evolution.

In conclusion, this project has used al three routes to obtain enzyme variants with possibly improved features. It has resulted in the identification, production and characterization of stable and novel alcohol dehydrogenases from the hyperthermophilic archaeon P. furiosus. Furthermore, the engineering by both laboratory evolution and rationale-based protein engineering was applied to improve the properties of two selected alcohol dehydrogenases. It is clear that the use of all three routes and especially combinations of them will be vital in acquiring enzymes suitable for biocatalysis in (large-scale) industrial applications.

120

Samenvatting en conclusies

Biokatalyse neemt geleidelijk aan de taak over van chemische katalyse in vele industriële toepassingen. Enzymen zijn milieuvriendelijk, biologisch afbreekbaar, efficiënt en de kosten voor vereiste hulpmiddelen zijn laag. Dusdanig leveren enzymen voordelen op ten opzichte van de traditionele chemische benaderingen die in diverse industriële processen gebruikt worden. Alcohol dehydrogenases (ADH’s) zijn een duidelijk voorbeeld van enzymen die interessant zijn voor de industrie. Ze kunnen stereoselectief zijn en een grote verscheidenheid aan substraat specificiteit vertonen. De productie van zuivere enantiomeren en diastereomeren van alcoholen is voornamelijk gewenst. Vaak zijn de natuurlijke enzymen echter niet geschikt voor industriële toepassingen op grote schaal, omdat zij niet optimaal presteren in een dergelijk onnatuurlijk proces. Er zijn drie belangrijke en hoofdzakelijk verschillende routes om enzym varianten met betere eigenschappen te verkrijgen: (1) enzymen isoleren uit organismen die in een milieu leven dat zo veel mogelijk voldoet aan de daadwerkelijke condities van het industriële proces, (2) rationele eiwit engineering (b.v. gerichte mutaties die gebaseerd zijn op basis van de structuur van het enzym) en (3) laboratorium evolutie (“random” methoden, b.v. willekeurige mutaties aanbrengen in het gen waarna een screening van de mutanten voor de gewenste eigenschappen de juiste variant moet opleveren). Het onderzoek wat in dit proefschrift wordt beschreven richt zich op de identificatie en karakterisatie van stabiele alcohol dehydrogenases afkomstig van de hyperthermofiele archaeon Pyrococcus furiosus (route 1) en de verbetering van eigenschappen van geselecteerde alcohol dehydrogenases door middel van laboratorium evolutie en indien er een goede structuur beschikbaar is door middel van eiwit engineering (route 2 en 3). In hoofdstuk 1 is een kort overzicht gegeven van de studie beschreven in dit proefschrift. In hoofdstuk 2 wordt een algemene introductie gegeven en een review over het verbeteren van de prestatie van enzymen door middel van laboratorium evolutie en rationele eiwit engineering (ondersteund door computergestuurde modellering). De beschikbare technieken voor de engineering van enzymen, onderverdeeld in rationele en “random” technieken, worden beschreven. Bovendien worden er verschillende selectie en (“high- throughput”) screening methoden beschreven die essentieel zijn voor de screening van grote aantallen mutanten. Recent gepubliceerd onderzoek over het verbeteren van enzymen door middel van laboratorium evolutie en/of rationele eiwit engineering wordt ook bediscussieerd. In hoofdstuk 3 is de identificatie van zestien mogelijke alcohol dehydrogenases uit de hyperthermofiel P. furiosus beschreven. Enzymen afkomstig van hyperthermofiele micro- organismen (micro-organismen die optimaal groeien bij temperaturen boven de 80°C) zijn over het algemeen extreem stabiel bij hoge temperaturen, hoge druk en in hoge concentraties van chemische denaturanten. Dit maakt hyperthermofiele micro-organismen over het algemeen zeer geschikt voor onderzoek naar robuuste enzymen. Het genoom van P. furiosus werd geanalyseerd voor genen die coderen voor mogelijke alcohol dehydrogenases. Dit resulteerde in de identificatie van adhA, adhB (genen die coderen voor twee eerder gekarakteriseerde ADH’s van P. furiosus) en zestien genen die mogelijk coderen voor alcohol

122 Samenvatting en conclusies dehydrogenases. De mogelijke ADH’s vertoonden overeenkomsten met de UDP- glucose/GDP-mannose dehydrogenase familie of met één van de belangrijke familie’s van de NAD(P)-afhankelijke alcohol dehydrogenases: short-chain ADHs (“korte keten ADH’s”), zinc-dependent ADHs (“zink afhankelijke ADH’s”), iron-activated ADHs (“ijzer geactiveerde of bevattende ADH’s) en de aldo-keto reductase superfamilie. Voor alle geïdentificeerde genen is STRING genoom context analyse, Pfam annotatie en COG annotatie gebruikt om een gedetailleerdere voorspelling van de functie te kunnen geven. Het resultaat hiervan demonstreerde de vele fysiologische processen waarin ADH’s een belangrijke rol kunnen spelen en de grote verscheidenheid aan substraat specificiteit die ze kunnen vertonen. Experimenteel bewijs is echter nodig om de voorspelde substraat specificiteit en fysiologische functie te bevestigen. AdhA, AdhB en vijftien mogelijke ADH’s werden succesvol gekloneerd en tot overexpressie gebracht in Escherichia coli. Vervolgens werd er een initiële screening uitgevoerd waarin AdhA, AdhB en twee van de mogelijke ADH’s (PF0991 en PF1960) activiteiten vertoonden. Er is een gelimiteerd aantal substraten en condities getest in de initiële screening en een meer uitgebreide screening (b.v. groter pH bereik, anaerobe condities, metalen) is nodig om een beter inzicht te verkrijgen in de fysiologische functie van de andere mogelijke ADH’s. De twee mogelijke alcohol dehydrogenases die activiteiten vertoonden in de initiële screening (PF0991 en PF1960) werden geselecteerd voor verdere karakterisatie. Vandaar dat in hoofdstuk 4 de heterologe productie, purificatie en karakterisatie van een nieuw thermostabiel alcohol dehydrogenase (AdhD; PF1960) van P. furiosus wordt beschreven. AdhD behoort tot de aldo-keto reductase superfamilie die bestaat uit NAD(P)-afhankelijke oxidoreductases. Leden van deze superfamilie delen een gemeenschappelijke (α/β)8-barrel vouwing en zijn over het algemeen monomeren die de nicotinamide cofactor binden zonder een “Rossmann-fold” motief. AdhD werd functioneel geproduceerd in E. coli en door zijn stabiliteit bij hoge temperaturen waren slechts twee stappen nodig voor de purificatie, namelijk een hittebehandeling gevolgd door anion-exchange chromatografie. Het is een monomeer met een moleculair gewicht van 32 kDa. AdhD heeft een voorkeur voor NAD(H) als cofactor en een uitgebreide substraat specificiteit in de oxidatie (pH optimum 8.8) en reductie (pH optimum 6.1) reactie. Er werd een duidelijke voorkeur voor secundaire alcoholen in vergelijking tot primaire alcoholen geobserveerd. De beste activiteiten werden echter gedetecteerd met polyolen (alcoholen met meerdere OH-groepen) als substraat. Maximale specifieke activiteit werd vastgesteld met 2,3-butaandiol (108 U·mg-1) in de oxidatie reactie en met diacetyl-acetoin (22.5 U·mg-1) in de reductie reactie. Gas chromatografie (GC) analyse liet zien dat AdhD met 2-pentanone als substraat voornamelijk (S)-2-pentanol produceerde (enantiomere overmaat 89%). De activiteit van het enzym nam toe met de temperatuur tot aan 100°C en de extreme thermostabiliteit van AdhD werd aangetoond met een halfwaardetijd (de tijd die het duurt voordat het enzym de helft van zijn activiteit verloren heeft) van 130 minuten bij 100°C.

123

Analyse van het adhD locus onthulde dat het start codon van een gen, coderend voor een wolfraam bevattende aldehyde oxidoreductase (wor4), zich direct aangrenzend aan het stop codon van adhD bevind. De clustering van deze twee genen werd ook gedetecteerd in de gerelateerde soorten Pyrococcus abyssi, Pyrococcus horikoshii en Thermococcus kodakaraensis. Dit suggereert een rol voor AdhD waarin het de aldehyden/ketonen, die geassocieerd zijn met de activiteit van WOR4, zou kunnen gebruiken of produceren. De kinetische data van AdhD tonen een hoge katalytische efficiëntie voor de reductie reactie aan en een hoge affiniteit voor het substraat en de cofactor die betrokken zijn bij deze reactie. Dit suggereert dat de fysiologische functie van AdhD de reductie van aldehyden of ketonen is. Hoewel men niet kan uitsluiten dat deze substraten gevormd worden door WOR4 lijkt dit zeer onwaarschijnlijk, omdat bekend is dat alle wolfraam bevattende aldehyde oxidoreductases die beschreven zijn (AOR, FOR, GAPOR) de oxidatie van aldehyden katalyseren. Als we aannemen dat er een fysiologische connectie is tussen AdhD en WOR4 zijn er twee scenario’s mogelijk. Ten eerste is het denkbaar dat AdhD de oxidatie van een alcohol katalyseert en zodanig een aldehyde of keton produceert die vervolgens verder geoxideerd wordt door WOR4. Vanwege de duidelijke voorkeur van AdhD voor de reductie reactie en van WOR4 voor de oxidatie reactie kan men zich echter ook nog een ander mogelijk scenario voorstellen. Het kan zijn dat een ongeïdentificeerde aldolase de substraten (bijvoorbeeld een aldehyde en keton) voor AdhD en WOR4 produceert. De keton zou vervolgens gereduceerd worden tot een alcohol door AdhD en de aldehyde zou geoxideerd kunnen worden door WOR4. Eerder is aangetoond dat WOR4 gezuiverd kon worden uit P. furiosus cellen die gegroeid werden in de aanwezigheid van zwavel. Helaas is er tot nu toe nog geen activiteit gedetecteerd voor WOR4. Meer experimenten zijn dan ook benodigd om de fysiologische functies van beide enzymen vast te stellen. In hoofdstuk 5 wordt de heterologe productie, purificatie en karakterisatie van een L- threonine dehydrogenase (Pf-TDH; PF0991) van P. furiosus beschreven. Pf-TDH behoort tot de polyol (meerdere OH-groepen) dehydrogenase familie, die onderdeel zijn van de medium- chain (“gemiddelde keten”) dehydrogenase/reductase superfamilie. Deze enzymen gebruiken NAD(P)(H) als cofactor, zijn homotetrameren of homodimeren en bevatten één of twee zink atomen per subeenheid die een katalytische en/of structurele functie hebben. Pf-TDH werd functioneel geproduceerd in E. coli en door zijn stabiliteit bij hoge temperaturen waren slechts twee stappen nodig voor de purificatie, namelijk een hittebehandeling gevolgd door anion- exchange chromatografie. Het is een tetrameer met een moleculair gewicht van ongeveer 155 kDa. Pf-TDH kan alleen NAD(H) als cofactor gebruiken. Maximale specifieke activiteit werd -1 vastgesteld met L-threonine (103 U·mg ) in de oxidatie reactie (pH optimum 10.0) en met -1 acetoin (3.9 U·mg ) in de reductie reactie (pH optimum 6.6). L-Serine en D-threonine konden ook gebruikt worden als substraat, maar Pf-TDH kon geen primaire alcoholen, secundaire alcoholen en andere L-aminozuren oxideren. De substraat specificiteit van Pf-TDH toonde aan dat voor Pf-TDH noch de amino groep noch de carboxyl groep van L-threonine essentieel is voor activiteit. De kinetische data liet echter een duidelijke voorkeur voor L-threonine zien.

124 Samenvatting en conclusies

Determinanten van de substraat specificiteit werden bestudeerd. Het verschil in activiteit met L-threonine en D-threonine toonde duidelijk aan dat de configuratie van het substraat belangrijk is. De activiteit was ook significant hoger wanneer het substraat een methyl groep op positie C4 bezat (L-threonine versus L-serine) of een amino of hydroxyl groep op positie C2, die waarschijnlijk betrokken is bij de correcte positionering van het substraat molecuul in het actieve centrum (“active site”) door middel van waterstofbruggen (L-threonine en 2,3- butaandiol versus 2-butanol). Hoewel de carboxyl groep niet benodigd is voor activiteit laat een vergelijking tussen 3-hydroxybutyraat en 2-butanol als substraat zien dat het wel degelijk invloed heeft op de activiteit. Pf-TDH heeft bivalente (“tweewaardige”) kationen, zoals Zn2+ en Co2+, nodig voor activiteit en het enzym bevat tenminste één zink atoom per subeenheid. L-Threonine dehydrogenases spelen een belangrijke rol in het L-threonine katabolisme. Ze katalyseren de NAD(P)-afhankelijke oxidatie van L-threonine naar 2-amino-3-oxobutyraat, die vervolgens spontaan decarboxyleert tot aminoacetone en CO2 of door het enzym 2-amino- 3-ketobutyraat coenzym A lyase in een CoA-afhankelijke reactie wordt omgezet in glycine en acetyl-CoA. Bovendien is in E. coli in vitro en in vivo aangetoond dat deze enzymen verantwoordelijk zijn voor de vorming van threonine uit glycine. Desondanks, wordt verondersteld dat de primaire rol van deze enzymatische route threonine katabolisme is. Een mogelijk 2-amino-3-ketobutyraat coenzym A lyase werd geïdentificeerd in P. furiosus door middel van geconserveerde context analyse gecombineerd met BLAST onderzoek. Daarom is er voorgesteld dat de oxidatie van L-threonine naar 2-amino-3-oxobutyraat de fysiologische functie van Pf-TDH is en dat het gevormde 2-amino-3-oxobutyraat waarschijnlijk omgezet wordt in glycine door een 2-amino-3-ketobutyraat coenzym A lyase. De engineering van AdhA, een thermostabiel ADH van P. furiosus, om de productie van (2S,5S)-hexaandiol bij gematigde temperaturen te verbeteren wordt beschreven in hoofdstuk 6. Het is aangetoond dat AdhA 2,5-hexanedione stereoselectief reduceert tot (2S,5S)-hexaandiol. Bovendien is er geconstateerd dat AdhA in de aanwezigheid van een overmaat aan 2-propanol substraat gekoppelde NADPH regeneratie (een cofactor regeneratie systeem dat al eerder met succes gebruikt is voor andere ADH’s) kan uitvoeren. Wegens de behoefte aan een stabiel alcohol dehydrogenase dat bij lage procestemperaturen (30°C) een goede activiteit vertoont voor de productie van (2S,5S)-hexaandiol is ervoor gekozen om de activiteit van AdhA met 2,5-hexanedione bij lage temperatuur (30°C) te verbeteren, terwijl de stabiliteit van het enzym behouden blijft. AdhA heeft een optimale activiteit tussen 90°C en 100°C en de activiteit bij 30°C is niet meer dan 5% van de optimale activiteit. Aangezien er geen hoge resolutie kristalstructuur beschikbaar is voor AdhA is rationele eiwit engineering geen goede optie. Er is dan ook voor gekozen om laboratorium evolutie te gebruiken om de lage temperatuur activiteit van AdhA te verbeteren. Er is één ronde van “error-prone” PCR (hierbij worden willekeurige mutaties in het gen gecreëerd doordat er een PCR uitgevoerd wordt die fouten maakt) uitgevoerd en vervolgens zijn er ∼1500 mutanten gescreend voor verhoogde activiteit met 2,5-hexanedione bij 30°C. Tenslotte zijn de 7 meest actieve mutanten geselecteerd. Deze mutanten en het wild type (het originele AdhA enzym) werden vervolgens

125

op grotere schaal geproduceerd en gezuiverd voor verdere karakterisatie. De best presterende mutant (13F2) had vergeleken met het wild type een 10 keer hogere activiteit met 2,5- hexanedione bij 30°C. De stabiliteit van deze mutant was ten opzichte van het wild type enigszins afgenomen, maar de mutant vertoonde nog steeds een zeer hoge weerstand tegen thermische inactivering. De lichte afname in stabiliteit zou veroorzaakt kunnen worden door verhoogde flexibiliteit. Verhoogde flexibiliteit kan tevens de reden zijn voor de verbeterde activiteit bij lagere temperaturen. Het substraat gekoppelde NADPH regeneratie systeem met 2-propanol kon ook met de mutanten gebruikt worden. Een structureel model van AdhA werd geconstrueerd (op basis van een ander enzym dat overeenkomsten vertoont met AdhA en waar een kristalstructuur van bekend is) en dit openbaarde dat de best presterende mutant één mutatie heeft in de geconserveerde NADP(H) bindingsplaats (R11L) en een tweede mutatie (A180V) in de buurt van de katalytische en zeer goed geconserveerde threonine op positie 183. Verder lieten de aminozuur substituties in de 7 beste mutanten zien dat er verscheidene adaptaties in verschillende regio’s van het enzym zijn die de activiteit bij lage temperatuur verbeteren. Regio’s van het enzym waar de mutaties werden geïdentificeerd waren o.a. de substraat bindingsplaats (“binding pocket”), de cofactor bindingsplaats en de regio dichtbij het katalytische centrum. De geselecteerde variant met de hoogste activiteit bij 30°C zou gebruikt kunnen worden voor de conversie van 2,5-hexanedione naar (2S,5S)-hexaandiol of als startpunt kunnen dienen voor verdere verbetering. In hoofdstuk 7 wordt de engineering van de R-specifieke alcohol dehydrogenase van Lactobacillus brevis (Lb-ADH), om zijn cofactor specificiteit te veranderen, beschreven. Lb- ADH is R-specifiek, relatief stabiel en heeft een brede substraat specificiteit. Vanwege deze eigenschappen is dit enzym een aantrekkelijke kandidaat voor biotechnologische toepassingen. Een nadeel is echter zijn voorkeur voor NADP(H) als cofactor, want NADP(H) is duurder en minder stabiel dan NAD(H). Aangezien er een hoge resolutie (1.0 Å) structuur van Lb-ADH beschikbaar is, was het mogelijk om op structuur gebaseerde computergestuurde modellering te gebruiken om mutaties te voorspellen die de cofactor specificiteit zouden wijzigen (rationele eiwit engineering). De computergestuurde modellering werd toegepast om mutaties te voorspellen die de holte zouden opvullen die ontstaat in afwezigheid van de fosfaat groep (de P in NADP(H) staat voor een fosfaat groep; NAD(H) bezit deze fosfaat groep dus niet) en die waterstofbruggen vormende groepen verwijderen die door de afwezigheid van de fosfaat groep onbevredigd blijven. Dit resulteerde in 4 mutaties (Arg38Pro, Ala90Ser, Val112Asp en Met140Ile), die werden voorgesteld om de activiteit met NAD(H) te verbeteren. Deze 4 mutaties werden geïntroduceerd in Lb-ADH en vervolgens getest met de cofactors NAD(H) en NADP(H). Alleen de mutatie Arg38Pro leidde tot een verhoogde NAD(H)-afhankelijke activiteit. Deze mutant had in vergelijking met het wild type een vier keer zo’n hoge activiteit met NAD(H) als cofactor. Wild type Lb-ADH had een pH optimum tussen 6.0 en 6.5 voor de reductie reactie met NADP(H) als cofactor. Als de cofactor NAD(H) gebruikt werd verschoof het pH optimum echter naar pH 5.5. De mutant Arg38Pro had in de reductie reactie met beide cofactors een pH optimum van 5.5. De

126 Samenvatting en conclusies thermostabiliteit van Arg38Pro was in vergelijking met het wild type significant hoger bij zowel 30°C en 42°C. Het is mogelijk dat de Arg38Pro mutatie het actieve centrum van Lb- ADH meer rigide (“stijf”) maakt, waardoor de mutant thermostabieler is. De affiniteit voor de cofactor NAD(H) was niet veranderd in de mutant in vergelijking tot het wild type. In eerder onderzoek is een mutatie (Gly37Asp) beschreven die de cofactor specificiteit van Lb-ADH veranderde, maar deze mutatie verminderde ook de activiteit van het enzym. De mutaties Gly37Asp en Arg38Pro werden gecombineerd in een dubbel mutant. De activiteit van de dubbel mutant met NAD(H) was in vergelijking met de Arg38Pro mutant afgenomen, maar de dubbel mutant had nog altijd een hogere activiteit dan het wild type. De Arg38Pro mutant heeft in vergelijking met het wild type een vier keer zo’n hoge NAD(H)-afhankelijke activiteit. Niettemin, is deze activiteit nog ongeveer vijf keer minder dan de wild type activiteit met NADP(H) als cofactor. Mutant Arg38Pro zou kunnen dienen als een startpunt voor verdere verbetering door middel van rationele eiwit engineering en/of laboratorium evolutie.

Samenvattend, heeft dit project alle drie de routes gebruikt om enzym varianten met eventuele betere eigenschappen te verkrijgen. Dit heeft geresulteerd in de identificatie, productie en karakterisatie van nieuwe, stabiele ADH’s van de hyperthermofiel P. furiosus. Verder zijn rationele eiwit engineering en laboratorium evolutie toegepast om de eigenschappen van twee geselecteerde ADH’s te verbeteren. Het is duidelijk dat het gebruik van alle drie de routes en vooral combinaties van deze routes essentieel zijn voor het verkrijgen van enzymen die geschikt zijn voor biokatalyse in industriële toepassingen (op grote schaal).

127

Dankwoord / Acknowledgements

Het dankwoord van een proefschrift wordt waarschijnlijk het best gelezen van alle stukken die in het proefschrift staan. Je wilt dan ook zeker zijn dat je niemand te kort doet of zelfs vergeet. Als je hier zeker van wilt zijn is er één simpele oplossing, waar ik dan ook mee wil beginnen: ‘Heel veel dank aan iedereen die op welke manier dan ook een bijdrage heeft geleverd aan dit onderzoek en aan mijn proefschrift’. Toch wil ik graag een aantal mensen bij naam noemen. Zonder de inzet, hulp en steun van veel mensen was dit onderzoek namelijk niet mogelijk geweest en was mijn proefschrift nooit zo ver gekomen.

Ten eerste wil ik mijn begeleider en promotor, John, bedanken voor zijn inzet, tijd, ideeën en aanhoudende enthousiasme. Ik heb veel dankwoorden doorgenomen om inspiratie op te doen (inderdaad ik vond het ook niet makkelijk om een dankwoord te schrijven) en ook daar wordt enthousiasme vaak genoemd als het over John gaat. De gesprekken die ik met John had over het onderzoek zijn hier een goed voorbeeld van. Als AIO ben je er namelijk niet altijd zeker van dat je op de goede weg zit, maar na zo’n gesprek met John had ik opeens weer nieuwe aanknopingspunten of twee hoofdstukken extra te schrijven. Ga zo door kan ik alleen maar zeggen, want je zorgt er op deze manier voor dat BacGen een fantastische plek is om te werken. Willem, naast John, bedankt dat je me de mogelijkheid hebt gegeven om bij je te promoveren. De begeleiding heb je vooral overgelaten aan John, maar zonder jou zou Microbiologie er toch heel anders uitzien. Bedankt voor het vertrouwen en steun. Het aandeel van John in de BacGen groep is natuurlijk groot, maar er zijn nog veel meer mensen die het tot zo’n goede werkplek maken en die ik dan ook wil bedanken. Servé, als kamergenoot en expert (o.a. enzymen en Pyrococcus) was je vaak de eerste aan wie ik mijn vragen stelde, bedankt voor je geduld, hulp en natuurlijk alle antwoorden. Ans, je zorgt goed voor BacGen (soms zonder dat we het merken en soms tegen wil en dank meer als politieagent), dank je wel voor de hulp bij meerdere practica. Corné, bedankt voor de hulp bij het opstarten van het project en de lol in de AIO-kamer (frisbeeën en voetballen). Alle andere AIO-kamergenoten wil ik ook bedanken voor de gezellige drukte: Krisztina, Johan, Arjen, Suzanne, Mark, Marco, Hao, Colin, Odette, Bart, Fabian en Marcel. Natuurlijk wil ik ook de mensen van de andere AIO-kamer bedanken voor de gezelligheid en al hun hulp: Stan, Jasper A., Thijs K., Leon, Thijs E., Ana, Harmen, Jasper W. en Marke. Dit boekje zou heel wat minder pagina’s hebben als ik niet de hulp gehad zou hebben van een aantal studenten. Agustinus, Nicole, Sjoerd, Annemarie, John en Tom heel erg bedankt voor jullie inzet en enthousiasme. Ik vond het leuk om jullie te begeleiden. I also want to thank Gergo Maroti for his assistance during his stay at our lab. I hope it was a informative and nice stay.

129 Naast de mensen van BacGen wil ik graag de mensen bedanken die ervoor zorgen dat alles goed loopt binnen Microbiologie: Francis, Wim, Nees, Renee, Jannie en Ria. En de andere twee groepen in het gebouw, MolEco’s en de MicFysers (bedankt voor het tolereren van mijn ‘geurige’ experimenten bij de spectrophotometer), voor de hulp, steun en gezelligheid gedurende mijn AIO-periode. Ook diverse samenwerkingsverbanden hebben ten grondslag gelegen aan de totstandkoming van dit proefschrift. I would like to thank Hans-Georg and Thomas from Julich Fine Chemicals, and all members of the PYRED project, for the nice collaboration and helpful discussions. Familie en vrienden bedankt voor jullie interesse en steun, maar vooral ook voor de afleiding en dus voor alles wat niet met dit werk te maken heeft. Rick, Marjolein, Jasper, Zuuz en Silvan bedankt voor de gezellige avonden die we hebben en de spellen die ik altijd weer mag winnen. Rick dank je wel voor je welkome hulp als paranimf. Peter, jij natuurlijk ook bedankt en veel plezier met dit, gelukkig niet te dikke boek. Binnenkort moeten we maar weer eens een potje gaan voetballen. Pa en Ma bedankt voor alles. Het boekje zal misschien niet zo makkelijk te lezen zijn, maar vooral jullie steun, interesse en hulp hebben ervoor gezorgd dat ik hier ben gekomen. Als laatste wil ik jouw bedanken, Liesbeth. Je bijdrage aan dit proefschrift is enorm. Je hebt me de gehele periode van AIO zijn geholpen en gesteund ondanks de drukke en soms moeilijke tijden. Je hebt geluisterd naar problemen, experimenten die lukten of soms mislukten, presentaties bekeken, hoofdstukken gelezen en zo kan ik nog een hele tijd doorgaan. Bedankt voor je betrokkenheid en steun, je was er altijd voor me. Je bent fantastisch én geweldig.

Nogmaals, Iedereen Bedankt!

130 Curriculum vitae

Ronnie (Marinus Petrus) Machielsen werd op 10 maart 1978 in Breda geboren en groeide op in het nabij gelegen Prinsenbeek. In 1996 slaagde hij voor zijn eindexamen VWO op het Mencia de Mendoza lyceum te Breda. Daarna begon hij aan de Universiteit Wageningen zijn studie biologie. Hier werd de interesse voor celbiologie en in het bijzonder microbiologie en genetica aangewakkerd. Hij deed dan ook een afstudeervak Bacteriële Genetica bij de vakgroep Microbiologie, onder begeleiding van Hauke Smidt, waar hij onderzoek verrichtte naar reductieve dehalogenase coderende genen en het TAT-translocase systeem van Desulfitobacterium dehalogenans. Zijn afstudeervak werd gevolgd door een stage bij DSM, bij Noel van Peij, waar hij werkte aan verbetering van het expressiesysteem in Aspergillus Niger. In september 2001 studeerde hij af en begon Ronnie als assistent in opleiding (AIO) bij de vakgroep Microbiologie aan de Wageningen Universiteit. Zijn werk, waarvan de resultaten in dit proefschrift worden besproken, maakte deel uit van het EU-project PYRED, ‘Novel bioreductions by hyperthermophilic microorganisms for the natural, specific and on-line production of fine chemicals’. Sinds juli 2006 is hij werkzaam als postdoctoraal onderzoeker (Post-Doc) bij NIZO Food Research te Ede in het TI Food and Nutrition (WCFS) project ‘Biodiversity and mixed cultures’.

131 132 List of publications

Machielsen R, Uria AR, Kengen SWM & Van der Oost J (2006) Production and characterization of a thermostable alcohol dehydrogenase that belongs to the aldo-keto reductase superfamily. Appl Environ Microbiol 72(1): 233-238.

Machielsen R & Van der Oost J (2006) Production and characterization of a thermostable L- threonine dehydrogenase from the hyperthermophilic archaeon Pyrococcus furiosus. FEBS J 273(12): 2722-2729.

Kube J, Brokamp C, Machielsen R, Van der Oost J & Markl H (2006) Influence of temperature on the production of an archaeal thermoactive alcohol dehydrogenase from Pyrococcus furiosus with recombinant Escherichia coli. Extremophiles 10(3): 221-227.

Machielsen R, Dijkhuizen S, Kaper T, Looger LL & Van der Oost J (2007) Improving enzyme performance in food applications. In: Novel enzyme technology for food applications (Ed, Rastall R) Woodhead Publishing Limited, Cambridge.

Uria AR, Machielsen R, Dutilh BE, Huynen MA & Van der Oost J (2007) Alcohol dehydrogenases from marine hyperthermophilic microorganisms and their importance to the pharmaceutical industry. Proceedings of the international seminar and workshop on marine biodiversity and their potential for developing bio-pharmaceutical industry – in press.

Machielsen R, Leferink NGH, Hendriks A, Brouns SJJ, Hennemann H, Dauβmann T & Van der Oost J (2007) Laboratory evolution of Pyrococcus furiosus alcohol dehydrogenase to improve the production of (2S,5S)-hexanediol at moderate temperatures. submitted for publication.

Machielsen R, Looger LL, Raedts J, Dijkhuizen S, Hummel W, Hennemann H, Dauβmann T & Van der Oost J (2007) Cofactor engineering of Lactobacillus brevis alcohol dehydrogenase by computational design. submitted for publication.

133 134 Activities in the frame of VLAG research school

Courses • Safe handling with radioactive materials and sources, Larenstein, 2002 • Bioinformatics Technology 2, WUR, 2002 • Protein engineering, VLAG, 2004 • Food chemistry, VLAG, 2005

Meetings • 9th and 10th Netherlands Biotechnology Congress, NBV, 2002-2004 • International congress on extremophiles, Naples, 2002 • Annual meeting studiegroep eiwitonderzoek, NWO, 2002-2005 • Vaalsbroek scientific meeting, DSM, 2003 • Biotrans 2005, Delft, 2005

General Courses • VLAG PhD week, 2002 • Teaching and supervising thesis students, OWU, 2004 • Career orientation, WGS, 2005 • NWO Talent day, 2005

135 The research described in this thesis has been carried out with the financial support from the Commission of the European Communities, specific RTD programme ‘Quality of Life and Management of Living Resources’, QLK3-CT-2001-01676 ‘PYRED: Novel bioreductions by hyperthermophilic microorganisms for the natural, specific and on-line production of fine chemicals’.

Printing Grafisch bedrijf Ponsen & Looijen B.V., Wageningen, The Netherlands

Cover design Liesbeth van den Brink and Ronnie Machielsen

136