ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2016

Digital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 1378

Towards Understanding ofSelectivity & Enantioconvergenceof an Epoxide Hydrolase

ÅSA JANFALK CARLSSON

ISSN 1651-6214ISBN 978-91-554-9586-2urn:nbn:se:uu:diva-286557

Dissertation presented at Uppsala University to be publicly examined in A1:107a, BMC,Husargatan 3, Uppsala, Friday, 10 June 2016 at 09:00 for the degree of Doctor of Philosophy.The examination will be conducted in English. Faculty examiner: Prof. Uwe T. Bornscheuer(University of Greifswald).

AbstractJanfalk Carlsson, Å. 2016. Towards Understanding of Selectivity & Enantioconvergence ofan Epoxide Hydrolase. Digital Comprehensive Summaries of Uppsala Dissertations fromthe Faculty of Science and Technology 1378. 90 pp. Uppsala: Acta Universitatis Upsaliensis.ISBN 978-91-554-9586-2.

Epoxide hydrolase I from Solanum tuberosum (StEH1) and isolated variants thereof has beenstudied for mapping structure-function relationships with the ultimate goal of being able to insilico predict modifications needed for a certain activity or selectivity. To solve this, directedevoultion using CASTing and an ISM approach was applied to improve selectivity towardseither of the enantiomeric product diols from (2,3-epoxypropyl)benzene (1).

A set of variants showing a range of activites and selectivities was isolated and characterizedto show that both enantio- and regioselectivity was changed thus the enrichment in productpurity was not solely due to kinetic resolution but also enantioconvergence. Chosen libraryresidues do also influence selectivity and activity for other structurally similar epoxides styreneoxide (2), trans-2-methyl styrene oxide (3) and trans-stilbene oxide (5), despite these not beingselected for.

The isolated hits were used to study varying selectivity and activity with different epoxides.The complex kinetic behaviour observed was combined with X-ray crystallization and QM/MMstudies, powerful tools in trying to explain structure-function relationships. Crystal structureswere solved for all isolated variants adding accuracy to the EVB calculations and the theoreticalmodels did successfully reproduce experimental data for activities and selectivities in most casesfor 2 and 5. Major findings from calculations were that regioselectivity is not always determinedin the alkylation step and for smaller and more flexible epoxides additional binding modes arepossible, complicating predictions and the reaction scheme further. Involved residues for thecatalytic mechanism were confirmed and a highly conserved histidine was found to have majorinfluence on activity thus suggesting an expansion of the catalytic triad to also include H104.

Docking of 1 into the active site of the solved crystal structures was performed in an attempt torationalize regioselectivity from binding. This was indeed successful and an additional bindingmode was identified, involving F33 and F189, both residues targeted for engineering.

For biocatalytic purpose the enzyme were was successfully immobilized on aluminaoxide membranes to function in a two-step biocatalytic reaction with immobilizedalcoholdehydrogenase A from Rhodococcus ruber, producing 2-hydroxyacetophenone fromracemic 2.

Keywords: Epoxide hydrolase, Epoxide, Enantioselectivity, Regioselectivity,Enantioconvergence, Crystal structures, Biocatalysis, Immobilization, Transient kinetics,CASTing, Directed evolution

Åsa Janfalk Carlsson, Department of Chemistry - BMC, Box 576, Uppsala University,SE-75123 Uppsala, Sweden.

© Åsa Janfalk Carlsson 2016

ISSN 1651-6214ISBN 978-91-554-9586-2urn:nbn:se:uu:diva-286557 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-286557)

Ju mer man tänker, ju mer inser manatt det inte finns något enkelt svar

Nalle Puh

List of Papers

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

I Janfalk Carlsson, Å., Bauer, P., Ma, H. and Widersten M. Obtaining

optical purity in product diols in enzyme catalyzed hydrolysis: con-tributions from changes in both enantio- and regioselectivity. Bio-chemistry 2012 vol. 51 7627-7637

II Amrein, B. A., Bauer, P., Duarte, F., Janfalk Carlsson, Å., Nawory-ta, A., Mowbray, S. L., Widersten, M. and Kamerlin, S. C. L. Ex-panding the catalytic triad in epoxide hydrolase and related enzymes. ACS Catalysis 2015 vol. 10 5702-5713

III Bauer, P., Janfalk Carlsson, Å., Amrein, B. A., Dobritzsch, D., Widersten, M. and Kamerlin, S. C. L. Conformational diversity and enantioconvergence in potato epoxide hydrolase I. Organic & Biomolecular Chemistry 2016. DOI: 10.1039/c6ob00060f

IV Janfalk Carlsson, Å., Bauer, P., Nilsson, M., Dobritzsch, D. and

Widersten, M. Laboratory evolved enzymes provide snapshots of the development of enantioconvergence in enzyme-catalyzed epoxide hydrolysis. Submitted

V Janfalk Carlsson, Å., Bauer, Paul., Kamerlin, S. C. L. and Widersten,

M. Complex Kinetic Schemes are Required to Describe Epoxide Hydrolase Catalyzed Production of Diols. Manuscript

VI Billinger, E.,† Janfalk Carlsson, Å.,† Widersten, M. and Johansson,

G. Integrated action of Solanum tuberosum epoxide hydrolase I and Rhodococcus ruber alcohol dehydrogenase A in a two-step biocata-lytic reactor. Manuscript

†These authors contributed equally.

Paper I reprinted with permission from Biochemistry http://pubs.acs.org/doi/abs/10.1021/bi3007725 and for addition/corrections http://pubs.acs.org/doi/abs/10.1021/acs.biochem.6b00197.

Paper II reprinted with permission from ACS Catalysis http://pubs.acs.org/doi/abs/10.1021/acscatal.5b01639. Paper III reprinted with permission from the Royal Society of Chemistry http://pubs.rsc.org/en/content/articlelanding/2016/ob/c6ob00060f#!divAbstract

Contribution report

I. Performed most of the experimental work, contributed to the writing of the manuscript.

II. Constructed E35Q/H300N and performed kinetic studies for this variant, wrote the corresponding parts for the manuscript.

III. Performed laboratory work to obtain crystals for R-C1B1 and parts

of the kinetics, contributed to the structuring of the manuscript and wrote the corresponding parts of the work.

IV. Performed laboratory work to obtain crystals for all variants present-

ed, contributed to the writing of the manuscript.

V. Performed all experimental work, contributed to the writing of the manuscript.

VI. Performed part of the experimental work, contributed to the writing

of the manuscript.

Contents

Introduction to my work ................................................................................... 13Green chemistry ........................................................................................... 13Biocatalysis ................................................................................................... 14

Why enzymes are suitable catalysts ........................................................ 14Industrial Biocatalysis ............................................................................. 15Enzymes’ role in catalysis ....................................................................... 16Enzyme selectivity ................................................................................... 17

Improving enzymes to be better catalysts .................................................... 18Stability .................................................................................................... 18Activity, selectivity and substrate scope ................................................. 19Challenges & bottlenecks ........................................................................ 19

Epoxides ....................................................................................................... 20Epoxides studied here ................................................................................... 21Epoxide hydrolases ...................................................................................... 22

Potential industrial applications for epoxide hydrolases ........................ 25Epoxide hydrolase from potato ............................................................... 25

The workflow .................................................................................................... 27Enzyme engineering ..................................................................................... 27Iterative Saturation Mutagenesis (ISM) ....................................................... 28Screening and selection ................................................................................ 29Characterization ............................................................................................ 30

Enzyme kinetics ....................................................................................... 30Selectivity & Specificity ......................................................................... 32The power of structure determinations ................................................... 33Theoretical modeling for experimentalists ............................................. 35

Towards multistep biocatalysis ......................................................................... 37

My work ............................................................................................................ 38Aim and strategy .......................................................................................... 38Paper I ........................................................................................................... 40

An attempt to reduce the screening effort ............................................... 40Drawbacks with an incomplete ISM approach ....................................... 41Chaperonins necessary for tolerance of more modifications .................. 42“You get what you screen for” ................................................................ 43E-values do not tell the truth about product outcome ............................. 43Changes in selectivity also for structurally similar epoxides ................. 44Important residues in deciding regioselectivity ...................................... 45

Paper II ......................................................................................................... 46trans-stilbene oxide - a suitable model substrate .................................... 46Crystal structure of H300N shows no catalytic water ............................ 46Support of E35 acting as a back-up base for H300 ................................. 47Product outcome is decided in the hydrolytic step ................................. 47Expanding the catalytic triad ................................................................... 48

Paper III ........................................................................................................ 50Smaller substrate allowing for more than one possible binding mode ... 50Solved crystal structure add to accuracy ................................................. 51Enantio- and regioselectivity is reproduced by calculations .................. 51A change in preferred binding mode for the enzyme variant with varied regioselectivity ......................................................................................... 52The importance of W106 in regioselectivity .......................................... 52

Paper IV ........................................................................................................ 54Crystal structures show a step-wise increase in active-site volume ....... 54Slightly changed position of lid tyrosines ............................................... 55Increased flexibility of the substrate lead to a new binding mode ......... 55Pre-steady state kinetics suggests a change in rate-limiting step and binding mode .................................................................................... 56Altered regiopreferences could partly be rationalized from docking ..... 56

Paper V ......................................................................................................... 59Non-strict regioselectivity motivates the use of complex reaction schemes .................................................................................................... 59Different binding modes demands a further complex reaction scheme ..................................................................................................... 60Pre-steady state kinetics do not reveal product outcome ........................ 60R-C1 behaves similar to the wild-type enzyme ...................................... 61RC1B1 deviate more from wild-type and R-C1 ..................................... 61Single-turnover experiments suggest a change in rate-limiting step ...... 63Different binding modes lead to possible problems in computational reproduction of experimental data for pre-steady state kinetics ............. 63

Paper VI ........................................................................................................ 65Optimization of reaction conditions of free enzymes ............................. 66Co-factor regeneration & Inhibition problems ....................................... 66Characterization of immobilized enzymes .............................................. 68

Conclusions & Future perspective .................................................................... 69

Populärvetenskaplig sammanfattning ............................................................... 71

Acknowledgement ............................................................................................ 76

References ......................................................................................................... 82

Abbreviations

ADH-A Alcohol dehydrogenase A from Rhodococcus ruber

CASTing Combinatorial active site testing

DNA Deoxyribonucleic acid

E. coli Escherichia coli

EH Epoxide hydrolase

ESI/MS Electrospray ionization mass spectrometry

HPLC High-performance liquid chromatography

ISM Iterative saturation mutagenesis

LDH L-lactate dehydrogenase

NADH Nicotinamide adenine dinucleotide

NOX NADH oxidase from Lactobacillus sanfranciscencis

PDB Protein data bank

QM/MM Quantum mechanics/molecular mechanics

StEH1 Epoxide hydrolase 1 from Solanum tuberosum

* Indicates a chiral carbon

13

Introduction to my work

Welcome to join in on my journey towards understanding of

the structure-function relationship of epoxide hydrolase I from Solanum tuberosum. This enzyme has previously been thoroughly

studied and characterized, with a well-understood reaction mechanism and solved crystal structure, thus fulfilling many of the criteria wished for when setting out for a task like this. The aim for this study was to construct and characterize modified enzyme variants with improved enantio- or regioselec-tivity. Together with theoretical chemists and crystallographers we took quite a few steps towards understanding the underlying mechanisms. This will aid in future work of being able to foresee modifications needed for a certain selectivity or activity with also other epoxides. As a far-end goal we want to use this enzyme together with other well-characterized enzymes to perform multi-step biocatalytic reactions as an alternative to man-made catalysts used in industry today.

Green chemistry To save our planet we need to treat it in a sustainable and greener way and to think more about the environment. This is true for chemical industry in par-ticular where organic chemists work hard on trying to fulfill the twelve prin-ciples of green chemistry, a list created already in 1998 (Box I). The princi-ples can be summed up as to: conserve energy and resources and avoid waste and hazardous material in industrial processes. Parameters to describe pro-cesses has arisen and amongst them E-factor (mass ratio of waste to desired product), atom economy (molecular weight of desired product divided by the sum of molecular weight of all substances produced), and step economy are commonly used, all serving to increase yield and reduce waste. Earlier, effi-ciency was only discussed in high yields and said nothing about the waste where today process optimization also involves the use of raw material and elimination of waste and toxic compounds. When it comes to waste, not only the actual amount but also the nature of it is important and various factors have been introduced (e.g. the unfriendliness factor1) to be used in combina-tion with the E-factor.

Sol

14

Box I. The twelve principles of green chemistry that serve as points of optimization for chemical processes.

• Prevent waste

• Atom economy

• Less hazardous syntheses

• Design safer chemicals

• Safer solvents and auxiliaries

• Design for energy efficiency

• Use of renewable feedstocks

• Reduce derivatives

• Catalysis

• Design for degradation

• Real-time analysis for pollution prevention

• Inherently safer chemistry accident prevention

Biocatalysis Biocatalysis is nowadays broadly defined as the use of enzymes or whole cells as catalysts for industrial synthetic chemistry. If using whole cells, the cell membrane needs to be permeable to the substrate or the enzyme needs to be expressed on the surface.2,3 Whole cells are easy and cheap to produce and co-factor regeneration is solved by the endogenous metabolic network. This network, however, could also be a true drawback due to other pathways acting to disturb and regulate the target reaction. This is avoided if synthetic pathways are created, by combining free enzymes. Here, drawbacks are in-stead the need of high stability, for the enzyme not to be degraded outside the cell, and purification, adding a cost in both time and money.

Why enzymes are suitable catalysts Biocatalytic reaction pathways are great alternatives for making industrial chemical processes greener* and more environmentally friendly. What bio-catalysts often do, compared to traditional metallo- or organocatalysts, is reducing the total waste, eliminating the use of transition metals and some-

* Green biotechnology refers to the application in agricultural processes and for industrial processes the term white biotechnology is appropriate.

15

times-toxic compounds and improving yield and productivity. Enzymes, by their inherent properties, are meeting many of the twelve principles of green chemistry in being biodegradable and harm-free to both environment and humans, functioning well at ambient temperatures and pressures in water-based solutions. They are efficient and often highly chemo- regio- and enan-tioselective thus increasing product quality and simplifying downstream processing. Most enzymes function under similar conditions and can be combined to co-function in one pot without the need for group activation or protection/deprotection, leading to less waste and a more atom- and step-economic processes. Previous drawbacks with enzymes as catalysts included high production cost and low stability, something that has been overcome by the enormous pro-gress in DNA technology, where today it is possible to express almost any enzyme in large amounts using simple E. coli systems. The stability issue is routinely solved by various immobilization techniques or protein engineer-ing.

Industrial Biocatalysis What the chemical manufacturing industry desire is stable, selective and productive catalysts functioning well under the conditions presently used in the process. Thus, the talk about enzymes being environmentally friendly due to functioning in water is indeed true but to be implemented in existing processes they need to be active also in the organic solvents used. There is on-going research in trying to perform organic synthesis in water-based systems4 but focus is also on the nature of the organic solvents (to be less harmful according to FDA) or alternatives like ionic liquids.5–7 Most en-zymes do not withstand high concentration of organic solvents and therefore stability needs to be improved. Additionally, if there is a need for co-factors, (compounds that can be expensive, especially in larger scale) either whole cells or co-factor regeneration should be considered for a cost-efficient pro-cess. Enzymes will not necessarily replace all steps in present industrial processes but rather be used for key steps where high selectivity is of importance. En-zymes come in especially handy when working with chiral molecules, with the wish for enantiopure products, due to enzymes’ chiral nature and their possibility to show high selectivity. Most synthetic reactions in pharmaceuti-cal industry include steps with chiral molecules and despite the fact that two enantiomers are highly similar (Figure 1), being mirror images of each other, their biological activities can be completely different in nature. When it comes to pharmaceuticals, usually only one of the enantiomers has the tar-geted effect while the other could be not only inefficient but also detri-mental. One such classic example is thalidomide, a tranquilizer recommend-

16

ed for pregnant women where only one enantiomer gave the calming effect while the other gave severe birth defects. From an economical and green point of view, also in cases where the inefficient enantiomer is harmless, it is beneficial to produce only the bioactive one.

Figure 1. Two stereoisomers that are mirror images thus called enantiomers. Enzymes are remarkable catalysts and many enzymes are used in chemical industry today,8–12 however, they could be implemented even more in indus-trial processes.11 There is still a knowledge gap between the two fields of biocatalysis and organic synthesis, reducing the transfer of enzymes into this industry, the situation is improving (one successful example being the recent contract between Codexis and Merck).13 In most cases, enzymes need to be engineered to fulfill the criteria of stability, activity, selectivity and substrate scope where companies like BASF, Codexis and Novozymes all have plat-forms with enzymes and high-throughput techniques to obtain tailored vari-ants for a specific need.

The importance of selectivity has been presented. Even though enzymes in theory look like the ideal catalyst they often need to

be engineered to carry the preferred activity, selectivity, stability or substrate scope wished for. Before going there, I will talk a bit about en-zyme selectivity and catalysis.

Enzymes’ role in catalysis Enzymes catalyze reactions by lowering the activation energy for the transi-tion-state intermediate by the tight fit allowing for electrostatic and hydro-phobic interactions between the enzyme and the substrate.14 Additionally, the structure of the active site allow for selection of substrates by steric hin-drance and also for positioning of substrates to be close to the catalytic resi-dues of the enzyme. This is especially important for reactions with more than one substrate, or reactions needing co-factors, where the enzyme keeps the components close in space, thereby increasing the probability for the reac-tion to occur. Furthermore, in the active site of an enzyme, reactions occur in a hydrophobic pocket allowing for more controlled reaction conditions. The low water content means that even hydrolysis can occur under controlled and

C C

be

17

selective conditions. The low water content in combination with various interactions of closely positioned amino acids often lead to abnormal pKa values and the positioning of amino acids residues also lead to coupled pro-tonation states from e.g. charge repulsion. These are the fundamentals of enzyme catalysis, but what is actually making enzymes as remarkable as they are is still not completely understood, and might only be for nature to know. Another factor that has arisen to explain enzyme catalysis is dynam-ics. Molecular dynamics show that enzymes are not rigid but in constant motion where some parts are more rigid, e.g. the core, while others are more flexible, e.g. the loops. Dynamic motions range on the scale from femto-seconds to minutes and their contributions to the catalytic powers of en-zymes is an on-going debate.15–22

Enzyme selectivity First, it is important to clarify the difference between specificity and selec-tivity due to these two terms being sometimes used as synonyms. Selectivity is when one product is favored over the other and thus mostly one product is formed. Specificity, on the other hand, refers to complete selectivity mean-ing that only one product is formed. In other words, complete selectivity is the same as specificity (IUPAC). Enzymes are often described as being both specific and highly chemo-, re-gio- and enantioselective (Figure 2), which they indeed can be, but in many cases they are promiscuous and function for a range of substrates.23 For bio-catalytic purposes perfect selectivities are aimed for in order to obtain pure products and low amount of waste. Thus in most cases protein engineering needs to be applied.

Figure 2. Schematic figures of (from left to right) chemo-, enantio- and regioselec-tivity. When working with chiral catalysis (also named enantioselective catalysis or asymmetric catalysis) and a racemic mixture of substrate both enantio- and regioselectivity is important. Formation of enantiopure product is either met by kinetic resolution (Figure 3, A) or by enantioconvergence (Figure 3, B). For kinetic resolution only one enantiomer is reacted while the other is left

O

R

O

R

O

R

O

R

O

R

18

unreacted thus leading to a limit of 50 % conversion. This limit is overcome by enzymes that show enantioconvergence, where there is a change in regi-oselectivity for the different substrate enantiomers. Enantioconvergence is sometimes seen naturally in enzymes but could otherwise be obtained from protein engineering or by using a combination of enzymes with complemen-tary selectivity.

Figure 3. (A) Kinetic resolution of two enantiomers resulting in maximum 50 % enantiopure diol and the non-reacted epoxide enantiomer and (B) complete enantio-convergence resulting in 100 % enantiopure product from a racemic substrate.

To be able to obtain a successful biocatalyst, in most cases the catalytic properties and/or the substrate selectivity needs to be

improved. Thus a stable enzyme susceptible to modifications is wished for.

Improving enzymes to be better catalysts Stability The wish to improve enzyme stability has already been touched upon when discussing tolerance towards organic solvents and stability was the first property to be targeted for both enzymes and whole cells, solved by immobi-lization.24 Immobilization of enzymes does not only improve stability but also adds the possibility of re-use of the catalyst without complicated down-stream processing. A drawback with immobilization is a possible decrease in activity thus an optimal method needs to be found in each case. Different techniques allow for more or less controlled enzyme immobilization where a totally random approach will result in low efficiency due to binding also occurring to the active site. An alternative or additional method for improv-ing stability is protein engineering where modifications known to be benefi-cial for stability (e.g. di-sulphide bridges and residues allowing for salt bridges within the enzyme) is introduced.25,26 If the structure of the protein is

A

B

imp

O

R

O

R

O

R

R

OH

OH

O

R

O

R

R

OH

OH

19

not known, random mutagenesis can be applied targeting the whole se-quence, thus a screen for variants with improved stability and retained activi-ty is needed.

Activity, selectivity and substrate scope Directed evolution is process where a rapid version of nature’s own evolu-tion is copied in a test tube and this method has emerged as an effective pro-cess for tuning selectivity and activity and to broaden the substrate scope for enzymes.27–30 At first, this procedure was performed completely randomly but upon more protein structures being solved and made available, more structure-based approaches were allowed and also more rational design. Pro-tein engineering is nowadays rather straightforward and in most cases vari-ants with improved selectivity and activity are indeed isolated.31–34 More focused approaches often result in larger improvement in selectivity compared to random alternatives, thus motivating the use of methods focus-ing on residues lining the active site.35 However, also modifications far away from the active site, in so called second or third shell residues, have shown to be beneficial, most likely due to compensating to stability loss upon bind-ing to a new substrate but possibly also by influencing enzyme dynamics.36 This brings a challenge for both experiments and computations, as the whole protein needs to be targeted for engineering purposes and during computa-tional modeling.37,38 The capacity for calculations is increasing and in silico studies are used more and more, allowing also for larger systems to be in-cluded in calculations to describe kinetics.39 Now it is even possible to make de novo design of enzymes with decent and also completely new activities.40–42 In many cases, however, de novo design result in poor activi-ties, suggested to be from lack of dynamics, but computational power in theoretical modeling is nevertheless aiding in directed evolution, making it more rational.43–45

Challenges & bottlenecks For the directed evolution, mutations can either be introduced one by one or in combination of two or more residues. If introduced one at the time the direct effect is seen while one miss out on possible cooperative effects. Also, a large number of modifications might be needed and since enzymes are evolved to function to perfection in its natural habitat, introducing too many modifications normally comes with a loss in activity. This could be met by first introducing stabilizing mutations for the enzyme to be more susceptible to further engineering and possible loss in binding energy.46 This could ei-ther be made by introducing random or site-directed mutations or by recon-structing an ancestral protein.36,47,48 These common ancestors are thought to be generalists, with a broader substrate scope, that has evolved into different

20

specialists. Thus despite the search for a highly selective catalyst, starting out with a promiscuous enzyme is in many cases preferred, allowing for reactions with a broader range of substrates and to stepwise create different specialists. The use of the one-by-one approach is supported by the fact that a property is not unique but can occur from a number of different modifica-tions.49,50 When instead introducing two or more residues simultaneously synergistic effects are allowed for. A common bottleneck in directed evolution is library construction and screening. Enzyme libraries consisting of more than one residue is often constructed, due to multiple modifications not always being additive but one by one could have no or detrimental effect but in concert they are highly beneficial (epistasis).36A drawback with modifying too many residues at the same time is the cost in screening effort and a high-throughput method is crucial. Examples of fluorescent cell sorting, colony screens51,52 and ESI-MS for screening enantioselectivity53,54 have been reported as well as a solid-phase gene synthesis.55 The fast improvement in DNA technology has re-sulted in the possibility to order any gene of interest and to codon-optimize it for expression in an organism of choice. Thus the time from idea to isolation of protein is significantly reduced. Even whole pre-made libraries can be ordered, saving not only time for construction but also for screening and sequencing due the fact that geno- and phenotype is linked. Thus, there are ways to overcome the bottleneck of library construction and the need of ex-tensive oversampling.

Nowadays, with increasing power of DNA technology, enzyme cost and availability is in most cases not an issue but instead

the limit lies in engineering the enzyme to more efficiently cata-lyse the reaction of choice. To reduce this bottle neck there is a need for understanding structure-function relationships to be able to predict modifi-cations needed for a certain property. The target system for this thesis is epoxide hydrolysis by epoxide hydrolases.

Epoxides Epoxides are a group of ethers being reactive due to the strained angles of the three-membered epoxide ring and the partial positive charge on the oxirane carbons, making them strong electrophiles. Since there are two of these carbons and due to these substrates, in many cases, being chiral, both regio- and enantioselectivity is important to avoid multiple products. Hydrolysis of epoxides can be either acid or base catalyzed. At neutral or basic pH, steric factors usually dominate and the reaction occurs via SN2 attack on the least substituted epoxide carbon. Under acidic conditions, the

thl th

21

epoxide is protonated and reaction can occur via both SN1 and SN2 at the most substituted carbon due to electron donating effects stabilizing the posi-tive charge of the oxirane carbon attacked.56 In other words, regioselectivity can be directed by simply changing the pH, adding complexity is that the nature of the substituents also influences reactivity of the carbons. Here, electron-donating groups allow for attack at the substituted carbon also at neutral and basic pH57. An example of this is for styrene oxide where under basic conditions both carbons are attacked to the same extent.58 In addition, when starting from cheaper racemic substrates, also enantioselectivity is highly important and this is poor in non-enzymatic systems since enantio-mers often appear identical to free nucleophiles. Instead, chiral catalysts like enzymes can be used for kinetic resolution or, in combination with changed regioselectivity, enantioconvergence. The reason why epoxide hydrolases are of interest for industrial purposes and biocatalysis is due to the chiral nature of both the substrate epoxides and the product diols and their presence as common key building blocks in or-ganic synthesis59–62. In current processes including epoxide hydrolysis with organic catalysts the main issue is the need of high amount of expensive catalyst and a limited efficiency with sometimes varying regioselectivity and lack of enantioselectivity.63 It is here enzymes come in handy in fulfilling all these criteria.

Epoxides studied here The environment in the active site of an enzyme allows for other reactions than those predicted to occur in solution. The epoxides studied here are pre-sented in Figure 4. For epoxides 2 and 3 the attack is assumed to occur at the phenyl-substituted carbon but wild-type StEH1 did previously show not to obey this rule. However, for epoxide 1 the phenyl-ring is positioned one carbon away from the oxirane carbons thus steering of the nucleophilic at-tack would be even harder, but assumed to be at the carbon of least steric hindrance. Epoxide 5 is bulky and does not allow for different binding modes in the active site, also only one product is formed independent of where the nucleophilic attack occurs thus only enantioselectivity could be monitored experimentally.

22

Figure 4. The epoxides and the hydrolysis product of the main substrate, studied during this work; (2,3-epoxypropyl)benzene (1), styrene oxide (2), trans-2-methylstyrene oxide (3), 3-phenylpropane-1,2-diol (4) and trans-stilbene oxide (5) where the numbering of the carbons are presented as well as the chiral carbons. A problem with phenyl substituted epoxide substrates, when it comes to enzymatic hydrolysis, is their low solubility in water and if solvents is added the enzyme have to be active and stable in such environment. To improve stability, immobilization could be used or an alternative two-phase system either by adding an organic phase or by simply adding more epoxide than can be dissolved thus forming droplets of substrate in the buffer solution, keeping the concentration of free epoxide low enough. Another problem could be spontaneous non-selective hydrolysis thus a sufficiently fast reac-tion is important to be able to control product outcome.

Epoxide hydrolases Epoxide hydrolases are widespread in both host and function62,64,65 and is therefore covering hydrolysis of a range of epoxide substrates with different activities and selectivities. Most of the epoxide hydrolases belong to the same superfamily of α/β-hydrolase fold, sharing a common structure and catalytic triad and therefore also mechanism. This leads to the fact that un-derstanding the structure-function relationships for one enzyme could also be applicable to related ones. As mentioned, there are epoxide hydrolases active with different epoxides and with different enantioselectivities and they do function to give enantioconvergent conversions, either alone66,67 or in tandem in one-pot reactions68–70. This α/β-hydrolase fold family71 consists not only of epoxide hydrolases but also lipases, esterases, peptidases, proteases, peroxidases and dehalogenases. These enzymes have diverged from a common ancestor with a preserved

23

catalytic triad but a varying binding site, leading to enzymes covering a wide range of substrates. The family is named after the typical structural fold they share which is built up by β–sheets surrounded by α-helices and where the epoxide hydrolases of this superfamily all have a lid domain covering the core domain, together forming the active site pocket. The catalytic triad con-sists of a nucleophile, a histidine base and an acid for charge relay. The his-tidine is completely conserved within this superfamily and for epoxide hy-drolases also the aspartic nucleophile while for other members this can be a serine or a cysteine. In the active site there is a so-called oxyanion hole built up by backbone amides to stabilize the negative reaction intermediates formed. For epoxide hydrolases there are two conserved tyrosine residues, necessary for positioning of the epoxide for nucleophilic attack71–73 The ki-netic mechanism is well understood and shared within the α/β-hydrolase family and proceeds in two steps, first alkyl-enzyme formation followed by hydrolysis, as described in Figure 5 where numbering in brackets refers to the residues in StEH1.64,74–76 For most epoxide hydrolases the hydrolysis step has been shown to be rate limiting, allowing for build up of alkylenzyme suggesting that its presence could be detected in the pre-steady state kinet-ics.74,77,78

Figure 5. The reaction mechanism for epoxide hydrolases of the α/β-hydrolase fami-ly is a two-step reaction where an alkylenzyme formation is followed by a hydrolytic

24

step. The two tyrosines (Y154 and Y235) of the lid domain aid in forming the Mich-aelis complex and positioning the epoxide ring for nucelophilic attack. In the first step where the nucleophile (D105), activated by H300 that is charge-relayed by D265, attacks either of the oxirane carbons to form a covalent negatively charged alkyl-enzyme intermediate where again the lid tyrosines play an important role in charge stabilization. In the second step, a hydrolytic water molecule, positioned by E3575 and activated by H300, attacks the alkyl enzyme to form a tetrahedral inter-mediate stabilized by backbone amides (Figure from Paper I). Numerous protein-engineering studies have been performed for this class of enzymes to increase both substrate scope and selectivity. There are many successful cases where enantioselectivity was increased, sometimes referred to as the selectivity between substrates and sometimes for the actual product outcome. Methods used have been both random processes like error-prone PCR and DNA-shuffling,79–81 often motivated by the findings that also muta-tions far from the active site could influence selectivity, but also more ra-tional approaches and site-directed mutagenesis based on structural infor-mation82–84. Examples of residues at the entrance of the active site and fur-ther out have been pinpointed but most influence in selectivity is obtained from modifications of residues surrounding the active site, especially the residue next to the nucleophilic aspartate79,81,85 and the lid tyrosines86–88. Due to structural similarities of this class of enzymes this is valuable information for future engineering of other epoxide hydrolases. To help in translating an amino acid position in the enzyme of one organism to another there are available databases made up from structures for all α/β-hydrolases that are crystallized.89,90 These engineering strategies have been applied for a range of different epox-ide hydrolases, amongst these are different limonene epoxide hydrolases (LEH) and EH from Agrobacterium radiobacter (ArEH) and Aspergillus niger (ArEH), all having known crystal structures.91,92 From these, it can be seen that the limonene epoxide hydrolase is not an α/β-hydrolase enzyme thus the mechanism is different. Here, the hydrolysis is acid catalyzed where the epoxide is protonated before nucleophilic attack by a hydroxide ion, thus no covalent intermediate is formed.91 All of these epoxide hydrolases, as most of the known ones, are homo-dimers, and they have rather small active sites and thus show rather narrow substrate scope. There have been numerous QM/MM studies for systems related to those shown in this thesis93–95 and it has also been possible to predict the increase in enantioselectivity for LEH from in silico studies when compared to labor-atory evolved enzymes.50 Interestingly, the modifications were not at all the same, possibly due to the large number of amino-acid residues modified simultaneously for the computational approach allowing for introduction of compensating mutations to allow for larger changes of the active site.

25

Potential industrial applications for epoxide hydrolases Epoxide hydrolases are interesting for industrial purposes due to their asymmetric synthesis of vicinal diols, as both them and the chiral epoxides are important building blocks in pharmaceutical synthesis.63,96–102 The fact that there are so many different epoxide hydrolases with varying activities and selectivities allow for combination of the two to perform an all in all enantioconvergent process. There are also examples of where the enzymes themselves show enantioconvergence, something that is not seen in (m)any other enzymes. Epoxide hydrolases, including StEH1, have been implement-ed in industrial processes of different scales, for production of a number of known drugs.68–70,103–109 A drawback is these enzymes inability to function in organic solvents110,111 where it needs to be stabilized, possibly by immobilization, for implementa-tion in present industrial chemical processes. There are attempts where epox-ide hydrolases has been successfully immobilized to function also in organic solvents.112–114 but a problem is the need for water in order to function in hydrolysis.115 Alternatives to organic solvents are ionic liquids (ILs) that constitute a major research field within green chemistry where a range of epoxide hydrolases show retained activity and selectivity despite the activity being completely abolished in some organic solvents.116–120

Epoxide hydrolase from potato The epoxide hydrolase from Solanum tuberosum (StEH1) is a promising candidate for biocatalytic purposes due to it being a rather small (36.1 kDa) and monomeric enzyme without any need for co-factors. Its native role is not completely clear but the preferred substrates are fatty-acid epoxides and the enzyme is mostly expressed in the leaves of the potato plant suggesting a role in formation of cutin.65,121,122 This enzyme have a rather broad substrate scope due to a relatively large active site making it a good candidate for en-gineering purposes towards increased activity for a range of epoxides. The wild-type structure was solved to high resolution123 (Figure 6) allowing for site-directed mutagenesis to confirm residues involved in the catalytic mech-anism75,124,125 and also for more rational approaches for enzyme engineering.126 Further, StEH1 is easy to express in large amounts and is stable over time thus making it easy to work with and this enzyme has previ-ously shown to be susceptible to modifications in both enantio- and regiose-lectivity126–128 where it is also known to show the remarkable property of enantioconvergence with some epoxides.66

26

Figure 6. To the left is StEH1 (PDB entry 2cjp) with the catalytic triad and the two tyrosines involved in positioning the epoxide. To the right, a close-up of the active site.

Now, we have an enzyme that looks good in theory. The next step is to create modified variants and if successful these will be fur-

ther characterized in order to map structure-function relationships. th

27

The workflow

Figure 7. An overview of the workflow used for this study, from library construc-tion via screening and hit characterization. Everything is performed in an iterative manner where the best hit is further characterized and template the next round of modifaction and screening. Each part is described in the following paragraphs of this booklet.

Enzyme engineering Enzymes can be engineered in different ways depending on what the aim is and what knowledge is available (Table 1). To perform rational design structural information is needed, if not the crystal structure so at least the amino-acid sequence for the possibility to create a homology model, and knowledge about function of the enzyme. This information is used to per-form site-directed mutagenesis where a certain residue is modified to a set one. This is often the method of choice to introduce e.g. disulphide bridges or study the importance of a certain residue and verify its involvement in the catalytic mechanism, as mentioned earlier.

28

In the other end of the spectrum is directed evolution that is instead a ran-dom process in an attempt to simulate a rapid version of nature’s own evolu-tion. Here, no structural information is needed and some of the most com-monly used methods include error-prone PCR, DNA shuffling and circular permutation. Table 1. Different methods applicable for protein engineering. Error-prone PCR

Random

DNA shuffling

Random

Circular permutation

Random/Target flexibility and dynamics

Saturation mutagenesis

Specific

Directed evolution can also be used in a semi-rational way by restricting the area to modify. One way of identifying important residues is CASTing (Combinatorial Active-Site Test) where small focused libraries surrounding the active site is chosen, each of them being subject to random mutagenesis.35 One benefit with this method is the allowance for synergistic effects otherwise not seen. This method demands a known 3D-structure of the enzyme, with preferably the substrate bound, to identify residues possi-bly interacting with the substrate or having influence of the structure of the active site as such.

Once the method for directed evolution has been decided upon and as well as the residues to include in each library these sites

need to be examined in a controlled manner.

Iterative Saturation Mutagenesis (ISM) A commonly used method is ISM where iterative cycles of saturation muta-genesis are performed.129,130 The resulting hit from each round is used to template mutagenesis of another library and this process is repeated until satisfaction. ISM allows for doing this systematically to make sure all com-binations in all orders are visited. If used in complete, ISM allows for epista-sis where the effect of one modification is dependent on the presence of an-other or simply that there is a non-additive effect in combination compared to individual modifications. In short the ISM method involves: i) identifica-tion and ranking of important residues, ii) selection and grouping of target residues, iii) choice of diversification method, screening effort and amino acid alphabet.

ne

29

For ranking the importance of a residue the priority goes: for residues re-ported to enhance activity > influence substrate scope > act in determining stereoselectivity > enhance stability. The ranking is needed to keep the num-ber of colonies to screen manageable. At this point it is important to consult available data regarding conserved residues. To reduce library sizes, the residues included in each library should be kept low (typically not more than 3). To further reduce screening effort a degen-erate codon set could be considered, a common alternative being NDT al-lowing for 12 out of 20 amino acids covering most of the different character-istics of the natural amino acids (Table 2). Despite this being considered a reasonable trade-off, the reduced library size comes with a sacrifice in possi-ble missed potential hits. The use of NNK codon allows for all 20 natural amino acids but involves 32 possible codons, thus several amino acids are overrepresented leading to a bias in the method and also more mutants to screen. When applying NDT for libraries consisting of two amino acids only 430 clones need to be screened for a 95 % coverage as compared to 3068 with NNK.131 Table 2. The amino acids coded for using degenerate codon set NDT.

NDTa

Nature of sidechain

Amino acids

F L I V Y

H R D N S C

G

Hydrophobic

Charged or polar

Small a N = all bases, D = A, G, T To overcome the problem with biased codon-set and having to sacrifice the number of amino acids included, small intelligent libraries are constructed by using a combination of codon sets resulting in only one codon per amino acid.132,133 Here, all twenty natural amino acids are allowed for while reduc-ing the screening effort to 1450 clones for the same coverage as discussed above. The decision about how restricted the libraries need to be is often a matter of the potential throughput of the screening method in combination with the knowledge about the surroundings of the site for modification.

Screening and selection All efforts in trying to reduce library size are aimed towards having a man-ageable set of variants for screening and hit identification, something that is often the bottleneck in directed evolution. To allow for cooperativity, often more than one residue is modified at the time, thus the number of clones to

30

screen demands a method of relatively high throughput. The screen is de-signed dependent on the property searched for where stability, activity or enantioselectivity are common alternatives. Often libraries are transferred to 96-well microtiterplates where a screen for the lysate is applied. Examples of fast methods are colorimetric screens or the use of substrates where product formation can be followed by UV/Vis spectrophotometry. In these cases the screen is used to directly get activity and data can be gathered also for vari-ants that do not show improved behavior. There is a lot of information to get from these variants where also modifications detrimental to the sought-after property can be pin-pointed. Often an initial, less complicated screen is used and the hits isolated from here are further analyzed to confirm potential hits. The colorimetric method used here show different shades of blue depending on how much unreacted epoxide is left.134,135 If the same lysate is used for screening of two enantiomers in parallel, enantioselectivity can be selected for. This method has been developed to also function for whole cells and colonies directly from the plate, something that would increase the through-put.52 The identified top enzyme variants will be sent for sequencing and then be analyzed in more detail to confirm the hit.

Now when a set of enzyme variants has been selected from a primary screen and these need to be characterized in further

detail to describe their actual activity and selectivity.

Characterization All isolated hits need to be characterized to assure and map their properties. Decisions have to be taken concerning parameters to examine based on what property is searched for. A covering characterization can be time-consuming and in some cases only the very best variant can be documented. Parameters of interest are normally describing kinetics and selectivity but also stability and of course the actual sequencing. It can include measurements with a whole spectrum of substrates under different pH and temperatures, solvents and buffer conditions.

Enzyme kinetics The most important property for characterization of enzyme function is ki-netics, describing the enzyme’s catalytic efficiency. There are two different time-scales to measure on, the longer resulting in steady-state and the shorter for pre-steady-state kinetics.

pde

31

Steady-state kinetics Many enzymes have complicated reaction mechanisms but most obey, more or less strictly, Michaelis-Menten kinetics, in its simplest form shown in Scheme 1.

Scheme 1. The simplest form of kinetic mechanism where substrate and enzyme form a Michaelis complex (ES) and subsequently broken down either back to free enzyme and substrate or to form product. This model is described by Equation 1 where it is important to make sure that initial velocities are measured such that the change in substrate concen-tration can be considered constant and the product concentration neglected. In other words, the method involves two assumptions; k-2 is negligible and the concentration of Michaelis complex (ES) is constant.

(1) where the initial rates’ (v0) dependence on substrate concentration ([S]) is shown, allowing for determination of the maximal rate (Vmax) and the Mich-aelis constant (KM). Any positive cooperativity adds sigmoidality to the saturation curve being it from allosteric regulation or conformational changes. Complexity is also added by the existence of more than one enzyme bound intermediate (Scheme 2) where expression for kcat and KM gets more complicated.

Scheme 2. The kinetic mechanism an additional enzyme intermediate (EA) is formed. Steady-state measurements describes the “equilibrium” reaction giving the combined constants kcat and KM. kcat is the turnover number, the amount of product formed per enzyme molecule per time unit when all enzyme are bound to substrate and include the rate constants for all steps of the reaction after substrate binding. KM is an apparent dissociation constant of all en-zyme-bound substrate complexes where substrate turnover is not included.

k1

E + S ES E + P k-1

k2

E + S ES EA E + P k-1

k1

k-2

k2 k3

32

KM is often described as the affinity for the substrate, something that is only true for the simple reaction in scheme 1 if k2 is negligible in comparison to k-1. The affinity is instead described by the dissociation constant KS (k-1/k1). KM and kcat are complex expressions built up of a number of rate constants that can be determined by pre-steady state kinetic measurements. These rates are important for attempts to pinpoint rate-determining steps and explaining product outcome. In steady-state kinetic measurements activity is often detected spectropho-tometrically, demanding different absorbance spectra for substrate and prod-uct. If there is no such difference, endpoint measurements must instead be performed, taking samples at different time points and analyze the product formation e.g. by HPLC or mass spectrometry.

Pre-steady-state kinetics To reveal the transient rate constants building up KM and kcat pre-steady-state kinetics needs to be analyzed where an instrument able to measure on a shorter time scale (typically ms) is used. An example is stopped-flow spec-tro/flurorophotometry where one needs to have a detectable signal upon e.g. binding of substrate in absorbance or fluorescence. In the case of StEH1, tryptophan fluorescence is quenched upon alkylenzyme formation allowing for detection of rates for formation and decay thereof. Equation 2 can be applied to reveal the different intrinsic rate constants. For multiple-turnover measurements, enzyme concentration must be well below substrate concen-tration to allow for assumption of this to be constant during the detection. What is measured is the difference in rates of decay plus formation of al-kylenzyme.

(2)

The observed rates are substrate dependent and the intrinsic rate constants can be obtained. Single-turnover experiments are only possible for reactions with low KS due to the need for substrate concentrations well above this val-ue to assure saturation and enzyme concentration to be well above this con-centration to assume all substrate to be bound and consumed during one catalytic cycle.

Selectivity & Specificity How could specificity be described and compared between different enzyme variants? kcat does not include substrate binding and is therefore not a suitable measure whereas KM does not include substrate turnover thus neither of these parameters are alone a good measure. Instead a combined quota is used the

33

so-called specific activity, kcat/KM. This parameter is used to compare the specificity of two substrates or two enantiomers of one substrate. In the latter case this gives the enantioselectivity thus how the enzyme distinguishes be-tween the two enantiomers of a substrate, often presented by the enantio-meric ratio (E) described by (kcat/KM)R/ (kcat/KM)S.136 Another important parameter for describing enzymes is regioselectivity. In the case of epoxides this describes the enzyme’s ability to distinguish be-tween the two oxirane carbons for nucleophilic attack (Figure 8). When var-ying regioselectivities are involved the E-value is not suitable for describing product outcome but enantiomeric excess of product or product ratio should be used. Figure 8. The product outcome from an enantiopure substrate upon different regi-oselectivity.

Once the enzyme is characterized and well described there is a wish to try to understand structure-function relationships in

order to be able to predict mutations needed for a certain catalytic property. In an attempt to solve this question experimental data can be com-bined with crystallization and theoretical modeling.

The power of structure determinations The number of protein structures deposited in PDB has exploded during the last decades with X-ray crystallography still being the by far most commonly used method. Determination of the tertiary structure of proteins adds invalu-able information for understanding of biological systems. X-ray crystallog-raphy provides atomic resolution and thus the possibility to study proteins in detail. Crystallization can be time-consuming and for some proteins not achievable. Conditions for a successful crystallization cannot be foreseen but are deter-mined by trial and error. When crystallizing a range of modified variants of the same enzyme a good starting point is to use similar conditions but this does not necessarily have to be successful. Methods used for crystallization are normally hanging or sitting drop vapor diffusion, where in both cases the difference in precipitant concentration in the drop and the container solution

ord

O

R

R

HO

OH

R

HO

OH

C1

C21

2

34

will make water diffuse from the drop thus causing an increase in protein concentration that initiates precipitation. When comparing X-ray crystallization data for small organic molecules and large proteins the accuracy is much lower for the latter, a consequence of proteins being dynamic and in constant motion. If a few conformers are overrepresented these could be resolved one by one from the same scattering data but if there are more (which is often the case) then they cannot be dis-tinguished and thus an average structure resembling the most commonly occupied one will be obtained. This average conformer does not necessarily resemble the catalytically efficient one since crystallographic conditions differ from the native enzyme habitat in demanding high protein concentra-tions, cryo protectants, precipitants and in many cases absence of substrate. One should also be careful when tuning the pH used since this will contrib-ute to shifts in the structure depending on different protonation states of pos-sibly interacting residues. Crystal structures do nevertheless contribute a lot to understanding of protein structure, and can identify residues putatively contributing to catalysis or substrate binding, especially in cases where there is a possibility to co-crystallize the enzyme with a ligand or substrate. To trap a covalently bound substrate in the enzyme normally demands modifica-tions of the enzyme to be catalytically impaired or the design of an inhibitor or transition-state analog that will allow binding but not product formation. When dealing with proteins that do not crystallize, multidimensional nuclear magnetic resonance (NMR) is an alternative method. Here, enzyme dynam-ics and conformational flexibility is studied at near-physiological conditions (in solution and at lower enzyme concentration). With this method differ-ences in structure upon binding of substrate can be monitored, however, on behalf of resolution. Other possible drawbacks are the size limit of 35 kDa for the studied protein, and that the protein needs to be stable at room tem-perature during the run. A third alternative for protein structure determination is cryo-electron mi-croscopy (cryo-EM). This method has made an enormous development in recent years and thus the number of structures determined by cryo-EM to date (about 1/100 of the number of crystal structures) will for sure increase rapidly, especially with achieved resolutions getting better where today 2.2 Å is possible.137 One drawback of cryo-EM is that the protein studied needs to have a certain size: thus so far it is mostly used to study large complexes and membrane proteins. For the present work X-ray crystallization was the method of choice where the structural information obtained allows for site-directed mutagenesis and also a possibility to perform more rational approaches for directed evolution. The crystal structure for wild-type StEH1 was solved previously123 and pro-

35

vided invaluable information that led to a better understanding of the catalyt-ic mechanism74,75,125 and to identification of functionally important residues to be targeted by various engineering approaches.126 Having a solved crystal structure for the wild-type enzyme also makes crystallization of slightly modified variants easier as the conditions under which they crystallize may be identical or very similar and also facilitates solving the crystal structures of the variants by providing a search model for molecular replacement.

Once structures are solved these can be used to improve theo-retical models of the reaction mechanism, not having to manu-

ally modify the residues. However, one needs to remember that these are rigid structures without ligands bound.

Theoretical modeling for experimentalists Theoretical modeling is increasing in use and importance for enzyme sys-tems due to an improvement in computing powers now allowing for model-ing of also bond-making and breaking in large systems like proteins. This also leads to improvement in accuracy making the theoretical models able to reproduce subtle effects like enantio- and regioselectivity. What is used to compare calculations with experiments is Equation 3 based on the transition state theory,

(3) where k is the rate constant (s-1), kB is Boltzmann’s constant (1.38 × 10-23 J/K), T is the temperature (K), h is Planck’s constant (6.626 × 10-34 Js), ΔG≠ is Gibbs free energy difference between the reactant and the transition state, and R is the gas constant (8.314 J/K). From this equation it can be seen that a large rate enhancement of 10 will result in an energy difference of only 1.4 kcal/mol, often within the error of the calculations. This means that theoretical modeling is by no doubt a pow-erful instrument in trying to explain behaviors seen from experiments and to suggest amino acid residues contributing to rate enhancement, however, not yet a tool for determining rates of the enzymatic reaction. Different methods for theoretical modeling demands different computational power and can therefore be applied for larger or smaller systems (Table 3). Molecular mechanics (MM) and molecular dynamics (MD) can be used to study all atoms in a protein where MM is used for docking experiments. In MD, sampling according to Newtonian physics allows the investigation of equilibrium properties, but the increasing complexity of the calculations also increases the computational demand. If the interest is in the actual chemistry,

alh

36

thus to study the kinetics of the enzyme-substrate system, quantum mechan-ics (QM) needs to be involved. Here, the computational demand is very high and this system is limited to study only a few atoms. Thus, QM cannot be used to study the whole protein but often a combination of QM/MM is used where QM is applied for the substrate and the closest atoms of the enzyme which is involved in catalysis, and MM is applied for the rest of the protein. For the QM part the empirical valence bond theory (EVB) was applied. This model is based on the introduction of individual valence bond states for all reacting species and states. Using a well described reference state, this meth-od allows the investigation of changes to the electrostatic environment of the reaction, as might be experienced for shifting a reaction from solvent to a protein active site.138–140 Table 3. Different methods for calculations and what can be obtained for these.

Parameters included

What system could Be applied

Molecular Mechanics MM Bonded and non-bonded interactions

Whole protein

Molecular Dynamics

MD

Dynamics included

Whole protein

Quantum Mechanics

QM

Bond breaking/making

A limited number of atoms

37

Towards multistep biocatalysis

Now enzyme variants and methods fill up the toolbox from

which one can pick and match to combine them in different modes depending on the reaction of choice. Structural infor-

mation from crystallization is combined with known preferences and activi-ties from characterization and calculations to get a more educated guess. And this is where my work ended. What is important when combining two or more different enzymes is to find optimal reaction condition for buffer choice and pH. In cases where co-factors are needed these should be re-generated in order to keep the costs down. Further, the more compounds you include the more complex the reaction gets in inhibition and downstream processing. More choices to make are whether to use free enzymes or whole cells, if to immobilize them and in that case by what method.

mmation from

38

My work

Aim and strategy My thesis is based on two parts: A) Directed evolution of StEH1 and under-standing of structure-function relationships and B) Towards multi-step bio-catalysis. In the major part (A), directed evolution was applied for creation of a num-ber of enzyme variants with different enantio- and regioselectivity to try to understand structure-function relationships. Experimental data was together with solved crystal structures used to train mathematical models with the aim of being able to in silico predict mutations needed to get a desired catalytic function. Modifying an enzyme to perform a certain reaction is often a bottleneck in various applications, to overcome this, understanding why an enzyme be-haves like it does is of great importance. If the structure-function relation-ship for a set of enzyme variants is mapped there is a possibility of foresee-ing which mutations to introduce a certain property to be achieved. In this way one can create a platform of enzyme variants where one could relatively easy and fast find a suitable candidate, with the long-term aim of building up a toolbox consisting enzyme variants with a range of different properties. To be able to create a platform like this, understanding the enzymes prefer-ences for one or the other stereoisomer of a substrate and why the attack is on a certain carbon of an epoxide ring, is needed. Thus we have studied en-antio- and regioselectivity of different enzyme variants and through collabo-ration with crystallographers and theoretical chemists we are approaching an understanding of structure-function relationships. In all my work the enzyme used has been an epoxide hydrolase from Sola-num tuberosum and variants thereof. This enzyme has been thoroughly stud-ied over the past decade resulting in a large amount of experimental data for a range of epoxide substrates. In addition, the crystal structure for the wild-type enzyme was solved and thus the potential for finding a structure-function relationship seemed promising. The enzyme had also been shown to be susceptible to changes in both enantio- and regioselectivity.

39

In the second part (B) engineered enzymes are put together in vitro to per-form a two-step biocatalytic reaction. Thus, the long-term goal will be a toolbox constituting not only enzyme variants with a range of properties but also techniques and a platform for optimization of reaction conditions.

40

Paper I Obtaining Optical Purity for Product Diols in Enzyme-Catalyzed Epoxide Hydrolysis: Contributions from Changes in both Enantio- and Regioselectivity At the start of this project we had the crystal structure for the wild-type en-zyme, a suggested and well-studied mechanism and an enzyme variant show-ing a reverted enantioselectivity compared to the wild-type enzyme. Directed evolution was performed to obtain enzyme variants with further increased product purity from racemic (2,3-epoxypropyl)benzene. In addition, activity and selectivity with two structurally similar epoxides were studied to see if any patterns in structure-activity relationships could be observed. The wild-type enzyme shows a modest 2.5-fold preference for the (S)- com-pared to (R)-(2,3-epoxypropyl)benzene (1), the main substrate of this work. However, previous studies do describe isolated variants with reverted enan-tioselectivity.126 This finding was the stepping-stone here, with the aim of trying to improve product purity even further. In addition, two structurally related substrates, styrene oxide (2) and trans-2-methylstyrene oxide (3), were examined to see if any conclusions on selectivity could be drawn using the same enzyme hits. If patterns could be seen, one might use these sub-strates for screening instead, leading to higher throughput due to product formation being possible to detect by spectrophotometrical measurements. In addition, one would also be able to use the same enzyme for different bio-catalytic reactions.

An attempt to reduce the screening effort To reduce screening effort CASTing was applied where residues were picked for their structural involvement in the active site and the anticipated involvement in positioning of the substrate but without touching the catalytic residues. Library residues were previously decided upon (A-D) with the exception of library E consisting of L266 and V267. These two residues are positioned at the back of the substrate-binding pocket possibly affecting substrate positioning by steric constraints and interaction with the phenyl ring of the substrates (Figure 9).

41

Figure 9. The residues for libraries A (yellow), B (pink), C (green), D (petroleum) and E (orange) are shown as sticks together with the catalytically active residues (wheat). In an attempt to further reduce the library size a degenerate codon set (NDT) was used, allowing for 12 out of 20 amino acids. For libraries of 2 residues 144 instead of 400 different combinations are possible and for 95 %-coverage only 430 instead of 3068 clones need to be analyzed.141 The 12 amino acids do represent different physicochemical properties and were considered to be a reasonable trade-off, and indeed a hit rate of 8 % was obtained for the first round of screening. Notable is that NDT does not in-clude a tryptophan codon thus W106 was by default modified for variants visiting library B.

Drawbacks with an incomplete ISM approach The ISM approach was not strictly followed in that not all library combina-tions were visited. Instead the best variant found in another study (R-C1B1) was used as template for the A and D libraries so that in total all sites had been visited (Figure 10). When running incomplete ISM, potential hits might be missed due to some modifications only being beneficial in the presence of others, a phenomenon called epistasis. When moving from library D to E the number of hits decrease from 42 to 8, screening the same number of variants. This indicates a need to insert stabilizing mutations for future variants to be susceptible to further modifications. Yet, a range of hits with varying enan-tio- and regiopreferences was identified, serving as a good base for under-standing the underlying mechanism behind selectivity for this epoxide hy-drolase.

42

Figure 10. The ISM tree of previous (grey) and present (black) work. For the resi-dues involved in each library, see Figure 9.

Chaperonins necessary for tolerance of more modifications No hits were isolated for the next generation using R-C1B1 as template without co-expression of chaperonins GroEL/ES. Thus strains also carrying a plasmid with genes coding for expression of chaperons were used from here on. For library A all isolated hits showed silent mutations, thus, retain-ing F33 indicates the importance of this residue for function. For the D li-brary a number of hits with increased selectivity were isolated thus the best variant here (R-C1B1D33) was selected as template for the E library. Table 4 show that there is a decrease in number of hits when going from D to E library. It was also found that modifications occurring more than once are not necessarily the ones with the best selectivity, a clear-cut example being R-C1B1D33 with a unique modification (Table SI2, Paper I). The explana-tion to some modifications occurring more often could be that they are being less detrimental to protein folding. If this is the case, introduction of com-pensating/stabilizing modifications might allow for other, beneficial, combi-nations of amino-acid substitutions. Table 4. Information about library residues, library size and number of hits for the R-selected variants. Template Library Residues

Library size Number of

hits

R-C1B1 A F33 168 2

R-C1B1 D I180 F189 504 42

R-C1B1D33 E L266 V267 504 8

CBD

DB

DC

DA

CBA

BC

StEH1 wt

CB

D

C B

A

CA BA

CBDE

43

“You get what you screen for” The colorimetric method used for screening is rather fast with acceptable throughput where lysate from each colony picked was examined for activity with both R-1 and S-1. Hits were selected based on their ratio of activity for the two enantiomers, i.e. enantioselectivity. In addition, also lysate showing high activity with both enantiomers were considered hits. In this way, possi-ble hits showing a changed regioselectivity are also isolated. Protein concen-tration in the screen is not normalized for, but since the screen is compara-tive, using the same lysate for both enantiomer reactions, this would not influence the outcome. There is however a risk of missing out on the ones having too low activity, which might be due to low expression levels. On the other hand, low expression is also a sign of instability, which is not a desired property. You get what you screen for† is an expression often heard and truly valid for this work. The screen used for hit identification involves high epox-ide concentration (10 mM) thus low KM was not selected for, something that is clearly seen in the results. Hits characterized did all (except R-C1B1D33E3) show an increase in KM for both enantiomers of 1. Except this increase in KM also kcat was reduced, more for S-1 than for R-1 resulting in a higher kcat/KM for the latter. E-values ((kcat/KM)R/(kcat/KM)S) for these R-selected variants increased from 0.4 for wild type to 14 for R-C1B1D33E3.

E-values do not tell the truth about product outcome The actual product outcome from reaction with racemic 1 was studied over time and presented as product ratios, the ratio of one enantiomer over the other. However, when comparing these product ratios to the E-values ob-tained for some of the enzyme variants, the data did not agree, suggesting a change in regioselectivity, from the strict behavior seen for the wild type. This was indeed the case, where variants, to a certain degree, started prefer-ring attack at C1 for S-1. However, the attack for R-1 was still strictly at C2. This enantioconvergent behavior, although not complete, is a sought-after property due to the possibility of overcoming the otherwise 50 %-conversion limit for kinetic resolution of a racemic substrate mixture. So product outcome does not solely depend on enantiopreference for the epoxide substrate, but also on the regioselectivity of the nucleophilic attack and thus describing selectivities by E-value would therefore be highly mis-leading. Since, in many cases, two products are formed from one substrate enantio-mer this motivates the use of the branched reaction scheme also for reactions with pure enantiomeric substrate (Scheme 3). This means that the steady-

† Frances Arnold 1999

44

state kinetics does in fact measure the sum of two, not necessarily equal, rates. Some sigmoidality for several substrates was indeed seen in experi-ments. The sigmoidality was however not to an extent that addition of an extra parameter was statistically justified. To get the two individual rates of formation, samples from the steady state kinetics could be analyzed with chiral HPLC to get the product ratio for each time point.

Scheme 3. The reaction scheme chosen to describe the reactions involves a substrate-independent conformational step and reversible isomerizations of the two Michaelis complexes (ES) while the two alkylenzymes (EA) are considered to not be intercon-vertable.

Changes in selectivity also for structurally similar epoxides The behavior of enzyme variants with epoxides 2 and 3 did not follow that of 1, thus none of these would successfully serve as substitute substrates for screening, something that would otherwise present a faster throughput. The primary screen used here has a high throughput while further measurement for activity is a true bottleneck in having to use HPLC for endpoint meas-urements for different time points to build up a saturation curve. Due to this effort in identification of potential hits obtaining activities also for variants not showing enantio- or regioselectivity were not prioritized. To also map modifications that are detrimental to activity and selectivity would add im-portant information in trying to understand the mechanism behind selectivi-ty. Despite no clear pattern between activities with the different epoxides the isolated variants did show varying enantio- and regioselectivities also for substrates 2 and 3 (Figure 4). The most dramatic effect in enantioselectivity is seen for R-C1B1D33 with S-2, going from an E-value of 69 for wild type to 5800. Concerning regioselectivity the S enantiomers is attacked at the benzylic carbon as would be predicted for the reaction in solution while for R-2 this is only true for wild type and R-C1B1D33E6. Also, for R,R-3 a range of different preferences are seen thus an indication a successful choice of library residues suggesting this set of enzyme variants to be applicable also for other epoxide substrates.

45

Important residues in deciding regioselectivity Due to lack of crystals structures of the isolated enzyme variants all conclu-sions about structure/function relationships are based on simple dockings of substrates to the wild-type structure. Thus one should bear in mind that the discussion here is mostly speculative. The most pronounced effect is traced to the modification of residues W106 and L109, both interacting with the substrate and the former also contributing to the oxyanion hole. W106 is a highly conserved residue previously found to influence enzyme selectivity in related epoxide hydrolases, supporting the theory of its involvement also here.81,85,142 However, since this modification was not introduced alone but in tandem with position L109 the effect might not be solely due to modifica-tions of W106. We isolated enzyme variants showing increased enantioselectivity towards both R-4 and S-4. The underlying cause was shown to be not only due to a change in enantio- but also regioselectivity, thus showing tendency of enan-tioconvergence. Deeper understanding of this was, at the time, mostly specu-lative and to get further insight one would wish to get crystal structures of the newly constructed enzyme variants, preferably with an intermediate bound. In addition, data for transient kinetics together with calculations would also add valuable information. No patterns of selectivity could be seen between epoxides 1, 2 and 3 but different product outcomes were ob-tained also for the epoxides not primarily selected for, indicating that the targeted residues did indeed play an important role in enantio- and regiose-lectivity among this set of structurally similar epoxides.

46

Paper II Expanding the Catalytic Triad in Epoxide Hydrolases and Related Enzymes Wild-type epoxide hydrolase together with a set of modified variants, con-structed to pin-point residues involved in catalysis, were examined to theo-retically repeat experimental values for selectivities and activities. Here, activity with a bulky substrate, allowing for few binding modes, was studied.

trans-stilbene oxide - a suitable model substrate The symmetric substrate, trans-stilbene oxide (5), is a suitable model sub-strate, large enough to fill up the active site thus allowing for only one possi-ble binding mode resulting in less expensive calculations. Once a reliable set-up for the calculations was found the same set of parameters could be used also for smaller and more flexible substrates. This substrate is not ex-pected to show any enantio- or regioselectivity for the uncatalyzed reaction in solution. For the enzymatic reaction, on the other hand, a preference to-wards R,R-5 is seen from experimental data while regioselectivity cannot be monitored since attack at either carbon, for either enantiomer, will result in the same meso-hydrobenzoin product. QM/MM calculations were applied to see if experimental data for enantiose-lectivity could be reproduced and regioselectivity explained. A more detailed understanding of the catalytic mechanism is needed to obtain further evolved variants and towards the aim of reliably and efficiently predict the effects of mutations on function in silico. Thus, activity for enzyme variants, previous-ly constructed to confirm specific residues’ involvement in catalysis (E35Q, H300N, Y149F, Y154F and Y253F), was confirmed by calculations to vali-date the reaction mechanism. In addition, a variant carrying two of previous deleterious modifications in combination (E35Q/H300N) was constructed with the aim of getting a variant truly inactive in the hydrolytic half-reaction.

Crystal structure of H300N shows no catalytic water To aid calculations available crystal structures for wild type and Y149F was used together with the structure for H300N that was solved here (PDB entry 4Y9S, resolution 2 Å). For the rest of the enzyme variants studied, modifica-tions of amino acid were added manually to the wild-type structure thus add-ing some inaccuracy. When comparing the crystal structures of wild type and H300N, the side chain of N300 has a different orientation and there is no density where the hydrolytic water normally is positioned resulting in a

47

stronger interaction between H104 and E35 that is not seen in the wild-type enzyme.

Support of E35 acting as a back-up base for H300 Steady-state kinetics for E35Q/H300N showed no product formation after five minutes of incubation with either of the enantiomers of trans-stilbene oxide, thus this enzyme variant was considered inactive. This does however only assure that hydrolysis/product release does not take place whereas al-kylenzyme formation could occur. H300 and E35 are thought to mainly be involved in activation and positioning of the hydrolytic water but are sug-gested also to help in activating D105 via charge relay and proton transporta-tion out from the active site. To investigate the impaired activity further, stopped-flow measurements were performed. Here, alkylenzyme-formation is monitored by a decrease in tryptophan fluorescence and the signal is re-gained upon hydrolysis and product release (or possibly decay of alkylen-zyme back to free enzyme and substrate). A clear quenching of fluorescence for both enantiomers of 5 was detected but no regained signal (after more than five minutes). Thus no, or undetectable, hydrolysis occur. Supporting the theory of E35 acting as a back-up base in H300N where the hydrolytic step could not be modeled due to the absence of the catalytic water in the crystal structure.

Product outcome is decided in the hydrolytic step Experimental values for catalytic activities were reproduced within 1-2 kcal/mol by calculations. For wild-type, enantioselectivity is reproduced for the alkylation step, showing a lower energy for the preferred R,R-5 while hydrolysis is lower for S,S-5 thought to be coupled to an error in getting the correct reference reaction for the R,R- enantiomer. Regioselectivity cannot be seen experimentally since attack of either of the two ring carbons of this symmetric epoxide will result in the same product and thus no data is availa-ble to support calculations. When only energies for alkylenzyme formation is studied the preferred attacks differ between the two enantiomers, C1 for S,S-5 and C2 for R,R-5. However, the only pathway that leads to product for-mation is via C2 attack also for S,S-5 due to a too high energy barrier for the hydrolysis of C1-alkylenzyme. The residues that seem to influence regioselectivity mostly in the hydrolysis step is H104 (destabilizing), W106 and D265 (both stabilizing) where the sum show a favored hydrolysis for alkylenzymes formed from attack at C2 for both enantiomers of 5. From electrostatic interactions it can also be seen that both residues of the E library (L266 and V267) show up as important residues for electrostatic contributions in the active site. This suggests the

48

inclusion of these residues to be a good choice, supporting continued work in introducing stabilizing mutations and re-visit library E.

Expanding the catalytic triad A new, previously ignored but highly conserved, residue (H104) was found to be also important for catalysis. Further support can be seen from sequence alignments where a replacement of H104 is coupled to a replacement of also the adjacent E35, to compensate for the loss of positive charge. Thus H104 and E35 most likely form an ion pair and H104 is also close in space to the hydrolytic water and H300, suggesting the protonation states of these two histidines to be coupled (Figure 11). Calculations for wild-type enzyme veri-fy the deprotonated state of H300 allowing it to function as a base. In order to provide charge balance in the active site H104 must be protonated. This is further supported by calculations where experimental data were reproducible for this form. Interesting is that this residue seems to have different protona-tion states for different enzyme variants, being neutral for both E35Q and H300N. Figure 11. The active site with the additional residues E35 and H104 included. The involvement of H104 in catalysis might be the explanation of the acidic shift in pH dependence seen for E35Q and R,R-5 where the pKa for this resi-due drops several units in calculations. For the wild-type enzyme there is a strong interaction between E35 and H104 that blocks solvent access to the active site. A similar behavior is seen for E35Q variant with S,S-5 where Q35 interacts with the tetrahedral intermediate formed after hydroxide at-tack. For R,R-5 the substrate pushes Q35 out of the active site thus allowing solvent access to the active site. This could be a clue to the peculiar pH pro-

49

file for E35Q where R,R-5 is affected in both alkylation and hydrolysis while the negative effects are smaller for S,S-5. Calculations pointed out the involvement of yet another residue (H104) im-portant for the mechanism of epoxide hydrolysis and its tight relationship with E35. Regioselectivity was decided in the hydrolytic step, thus, even though the preferred carbon for alkylation varied for the two enantiomers, product formation will in both cases only happen for C2 attack due to a ki-netically blocked hydrolytic step for C1.

50

Paper III Conformational Diversity and Enantioconvergence in Potato Epoxide Hydrolase 1 Calculations were performed to explain the enantioconvergent behavior for the wild-type enzyme with a smaller epoxide substrate, allowing for more than one binding mode. In addition, a modified variant with changed regi-oselectivity was included to see if also this behavior could be pin-pointed. The same procedure as in Paper II was applied here for the QM/MM calcula-tions with the smaller and non-symmetrical substrate styrene oxide (2). Also, regioselectivity is experimentally determined where the wild-type enzyme shows almost complete enantioconvergence. Calculations were performed to reproduce experimental data for both enantio- and regioselectivity. In an attempt to try to explain the underlying mechanism, something that will aid in construction of new enzyme variants possibly showing enantioconver-gence also for other epoxides (Table 5). In addition to wild-type enzyme a modified variant (R-C1B1) was included due to its loss in regioselectivity with one of the enantiomers. Table 5. Regio- and enantioselectivity shown by StEH1 wt and R-C1B1 with 2.

Regioselectivity

Enantioselectivity

R-2

S-2

ES/R

(%) Preferred carbon

(%) Preferred carbon

(fold)

StEH1 wt

89±0.8

C2

99±0.1

C1

69±6

R-C1B1

55±10

C2

95±0.4

C1

42±3

Smaller substrate allowing for more than one possible binding mode Two stable binding modes were found (stable as the substrate staying in the active site) where the phenyl ring of 2 interacts either with the imidazole ring of H300 (Mode1) or with the indole of W106 (Mode2) (Figure 12). The in-volvement of W106 in regioselectivity was previously suggested by simply studying the crystal structure with the substrate manually docked and the idea of different binding conformers to be involved in selectivity was dis-cussed in earlier work. To verify this speculation calculations for nucleo-

51

philic attack by D105 at both carbons and enantiomers with substrate bound in either of the binding modes were performed.

Figure 12. The two binding modes found to be occupied for styrene oxide. On the left, Mode1 where the phenyl ring of the epoxide interacts with H300 and to the right, Mode2 where the interaction is instead with W106.

Solved crystal structure add to accuracy The accuracy of the calculations is aided by the solved crystal structure for enzyme variant R-C1B1. Compared to the wild-type structure there are no major differences except for the substituted residues W106L, L109Y, V141K and I155V where the former two seems to influence selectivity the most. Y109 is oriented away from the binding site thus increasing the vol-ume of the active site while W106 is involved in Mode2 binding thus this mode is for L106 expected to be less stabilized. In addition, V141K, located at the entrance could affect availability of the active site but due to the low density of this side chain this cannot be stated.

Enantio- and regioselectivity is reproduced by calculations Both wild type and R-C1B1 prefer S-2 to R-2, which was reproduced by calculations where in both cases energies were significantly lower for reac-tion with the (S)-enantiomer. The explanation lies in the change of preferred binding mode for the different enantiomers, where stabilization from amino acids lining the active site and the availability of the respective carbons for ring opening are important. Mapping regioselectivity is a more complicated issue due to the different possible binding modes. Despite the alkylation being favored for ring open-ing at one carbon the energy for the subsequent hydrolytic step might be sufficiently high to be considered blocked. For wild-type enzyme and R-2 Mode2 is preferred where both alkylation and hydrolysis being energetically

52

favorable for attack at C2. For S-2 on the other hand Mode1 is preferred but essentially blocked for hydrolysis. Thus, also here Mode2 is the productive binding mode. Ring opening at C1 will be preferred due to lower energy for the alkylation despite the hydrolytic step being less favorable. Thus, here regioselectivity can be reproduced already in the alkylation step and not in the hydrolytic step as for epoxide 5.

A change in preferred binding mode for the enzyme variant with varied regioselectivity For R-C1B1 and R-2, Mode1 was instead preferred where alkylation is more energetically favorable for attack at C2 while hydrolysis is equal with either of the alkyl enzymes. For S-2 binding of either Mode1 or 2 is possible with a slight preference for the former. In both cases, ring opening at C1 is pre-ferred for alkylation but not for hydrolysis but this is still the most likely reaction path. Thus, regioselectivity was reproduced from calculated ener-gies for S-2 but not for the R enantiomer. What is seen for R-C1B1 is a shift from Mode2, seen for the wild type, to Mode1. This is most likely due to the loss of interactions between the phenyl ring of the epoxide and L106 after substitution. The active site is larger for R-C1B1 where additional binding modes might be available and indeed a third binding mode did appear during modeling however not stable enough to be included.

The importance of W106 in regioselectivity When studying electrostatic and van der Waals contribution for wild-type enzyme W106 show major influence for the (R)-enantiomer in C2 attack, a possible explanation to the different regioselectivities for the two variants. Since this residue is mutated for R-C1B1 this might be the reason why a different mode is preferred. However, calculated energies still suggest C2 attack to be highly preferred while experimental data show attack at both carbons to be equally probable. Despite this failure in reproducing regiose-lectivity for R-C1B1 it is important to remember that simulations used for calculations are only studied for 10 ns and does not allow for the real reac-tion to occur. Also, the energy difference in attack at C1 and C2 is very small; a ratio of 100 in relative rates giving an energy difference of about 2 kcal/mol, comparable to the error in the calculations. Based on this, pin-pointing the origin of this small effect is challenging. The new findings, that different binding modes are preferred for different enantiomers and enzyme variants need to be considered when designing libraries for directed evolution. Also, even though one conformer is pre-

53

ferred for binding and alkylation the subsequent hydrolytic step might be too high in energy to be feasible. Two stable binding modes were found for styrene oxide in these enzymes where different modes were preferred for different enantiomers and for dif-ferent enzyme variants. Calculations were able to reproduce experimental data for the enantioselectivity of both wild-type enzyme and a modified vari-ant thereof. For wild type also regioselectivity and the concomitant enantio-convergence could be described via calculated energy differences for each step.

54

Paper IV Laboratory Evolved Enzymes Provide Snapshots of the De-velopment of Enantioconvergence in Enzyme-Catalyzed Epoxide Hydrolysis Here we set out to try to explain the regioselectivities seen with (2,3-epoxypropyl)benzene (1) for the set of enzyme variants isolated in Paper I. Crystals structures were solved for all these variants and substrate was docked into each active site to see if the preference in carbon to attack could be rationalized. From Paper I, regioselectivities showed a strict attack at C2 with R-1 for all variants, while for S-1 there was a stepwise change from C2 for wild-type enzyme to C1 for R-C1, R-C1B1 and R-C1B1D33. This preference is re-duced again for variant R-C1B1D33E6 that is back to wild-type behavior. Thus the high product purity is in this case from kinetic resolution while for R-C1B1D33 this is due to enantioconvergence. Residues W106L and L109Y were suggested to have the most influence in regioselectivity from structural estimations, based on speculations from what was seen upon manual modifi-cations of wild-type structure. This involvement was further supported by work with epoxide hydrolase from another source where selectivity was altered upon modifications of this site.81,85,142 Since then, in Paper III, calcu-lations had been performed for a structurally similar epoxide (2) where W106 was pointed out as highly stabilizing for one of the possible binding modes, adding to this hypothesis. Here, crystal structures for all enzyme variants have been solved and used for docking experiments thus adding more accuracy to speculations.

Crystal structures show a step-wise increase in active-site volume Here we present three solved crystal structures (R-C1, R-C1B1D33 and R-C1B1D33E6). In addition, the inactive variant E35Q/H300N, described ear-lier, was co-crystallized with S-1 in an attempt to obtain a structure with a bound intermediate. Crystal structures were solved where some density in the active site could be seen, however, not to the extent that any conclusions could be drawn. Also, no covalent bond is seen between this density and D105 and the overall structure do not differ much from that of the wild type, suggesting that crystallization conditions need to be further optimized. It is worth mentioning that in the active site of R-C1 there is a glycerol, from the crystallization medium, positioned in between the tyrosines and the nucleo-philic aspartate possibly indicating product binding due to structural similari-ties.

55

When studying the structures, the backbone carbons do not differ significant-ly in modified variants compared to wild-type enzyme. Further, there is no large difference in active site volume for R-C1 compared to wild type and these two variants are also similar in activity and selectivity. The R-C1B1 structure on the other hand shows an increased active-site volume due to Y109 being oriented away from the binding site upon interaction with N241. Upon modification F189L for R-C1B1D33, even more space is introduced in the active site. For R-C1B1D33E6, L266 is modified to a glycine, increasing the active-site volume even further. This increase in active site is not surpris-ing since we see an increase in KM values for the modified variants with this substrate compared to wild type. This increase in KM goes hand in hand with the screen used for identification of these hits where a high concentration of substrate was used (as described in Paper I). Thus low KM was not selected for but rather a decent kcat.

Slightly changed position of lid tyrosines Other changes that could be seen for the modified enzyme variants are vary-ing positions for the lid tyrosines Y154 and Y235. I155 is in close proximity to these (and H153) where for R-C1 the I155V exchange could potentially contribute to the positional flexibility observed for the lid Tyr phenol rings. This difference is slightly more pronounced for Y235 where in R-C1B1 it differs by 1.2 Å, compared to wild-type StEH1. This movement of Y235 is suggested to be due to W106L modification, allowing for an increased flexi-bility otherwise prevented by the bulky side chain. It should be mentioned that a dioxane molecule is present in the binding site, possibly influencing side-chain locations. Also for R-C1B1D33, the tyrosine positions are changed as the observed distances between the hydroxyl groups of Y235 and Y154 is measured to 5.0 Å compared to 4.4 Å in R-C1B1. This positional flexibility is rather minor in extent, but since the lid Tyr position the epoxide ring and facilitate its opening, even a small change may affect regio- and stereoselectivity as well as microscopic parameters of the catalyzed reaction, something that has been seen in other studies.86–88

Increased flexibility of the substrate lead to a new binding mode As mentioned, W106 was thought to have major contributions to the changed changed regioselectivies due to its involvement in one of the bind-ing modes for styrene oxide (Paper II). However, the docking studies here revealed that this binding mode was rarely adapted for the epoxide studied here, not even for wild type. Instead a new binding mode was revealed (Mode3) involving interactions with F33 and F189. This involvement of F33 could be an explanation as to why this residue was conserved when targeted for modification (Paper I) where both hits found showed silent mutations indicating this residue’s importance for activity. The new binding mode is

56

probably due to the increased flexibility from the methylene carbon connect-ing the epoxide and the phenyl ring of 1. When comparing KM for this sub-strate to the isomer trans-2-methylstyrene oxide the values are significantly higher for the former, suggesting that free rotation of the epoxide and the phenyl ring contributes to less binding stability. This suggests that it is actu-ally the dissociation constant (KS) that is increased, further supported by MD simulations where neither of the substrate enantiomers of 1 could be trapped in the active site without adding restraints. Also activity is lower with 1 compared to 3 probably also due to the increased flexibility.

Pre-steady state kinetics suggests a change in rate-limiting step and binding mode To pinpoint the varying regioselectivities further, transient kinetic data was wished for to possibly decide the rate-limiting step for this set of enzyme variants. Since the residue thought to have the strongest effect in tryptophan fluorescence (W106, due to its position in the active site close to negatively charged D105) was modified for R-C1B1 and further evolved variants, tyro-sine fluorescence was measured instead. Tyrosine fluorescence has a lower quantum yield and thus weaker intensities were expected. However, R-C1B1D33 did show detectable amplitudes and therefore tyrosine fluores-cence was considered a good alternative also for R-C1B1. Despite clear am-plitudes for wild type and R-C1B1D33 the rate of alkylenzyme build-up was too fast (within the dead-time of the instrument) to measure. For R-C1 and R-C1B1 instead no rates could be determined, probably due to inefficient build up of alkylenzyme. The different behavior in pre-steady state meas-urement could indicate a change in rate limiting step between the different mutants, from limiting hydrolysis in the wild type to limiting alkylation in variants R-C1 and R-C1B1, and back to limiting hydrolysis in the R-C1B1D33 variant.

Altered regiopreferences could partly be rationalized from docking The low efficiency in binding and speculations about fast chemistry suggest docking to be a suitable instrument for visualizing regioselectivity, once a favorable binding mode is met the rest of the reaction pathway will proceed leading to product formation. This motivated the use of simple docking tools to predict preferred attack for alkylation to possibly reflect product outcome. Having real crystal structures add in accuracy for the dockings despite the lack of bound intermediates. The data in Table 6 shows that in most cases the regiopreference matches the shortest distances between the nucleophilic D105 and the preferred carbon for attack. It should however be kept in mind that the preferred attack for alkylenzyme formation do not always lead to

57

product formation, as seen in Paper II and III where hydrolysis of the al-kylenzyme were in some cases blocked and product formation would only be allowed after attack at the least preferred carbon. In such cases conclusions of regioselectivity from substrate docking would fail. It should also be men-tioned that what is seen from docking is that the preferred binding mode differs between R-C1B1 and R-C1B1D33 where the latter adapt Mode2, edge on Y109, possibly the explanation to why there is a detectable fluores-cence signal here and not for R-C1B1. However, speculations like these should be confirmed by QM/MM calculations. Docking also proposed pre-viously neglected residues F191 and F158 to interact with R-1 and Y183 and F301 with S-1, suggested to be targeted for future rounds of modifications. Table 6. Regioselectivity and distances between nucleophile (D105) and the two carbons of the epoxide ring for the preferred binding mode seen in docking experi-ments.

R-1

S-1

Enzyme

C1a

(%)

rC1

b

(Å)

C2a

(%)

rC2

b

(Å)

Modec

C1a

(%)

rC1

b

(Å)

C2a

(%)

rC2

b

(Å)

Modec

StEH1 (wt) 5 3.6 95 2.7 3 7 3.0 93 3.0 1’

R-C1 3 2.8 97 2.7 1’ 15 3.0 85 2.6 3

R-C1B1 2 2.8 98 2.8 1’ 60 2.6 40 3.0 1’

R-C1B1D33 2 2.6 98 2.8 2 62 2.7 38 2.7 2

R-

C1B1D33E6

2 2.6 98 2.8 2 9 2.7 91 2.9 1’

aExperimental data from Paper I. bCalculated distance between D105 and C1/C2. cMode1’ involves interaction of the phenyl ring and H300, Mode2 W106 and Mode3 F33 and F189. The overall crystal structures for the variants do not differ significantly form the wild-type where most changes, except for the gradually larger active site, was the slight movement of the lid tyrosines, possibly influencing the posi-tioning of the epoxide ring and thus also possible effect on regioselectivity. The larger volume of the active site goes hand in hand with the fact that KM is increased, which is an effect from the increased flexibility of the substrate and the screening set-up, where high epoxide concentration was used. Pre-viously identified binding modes for styrene oxide was not occupied for this more flexible epoxide thus the direct involvement of W106 could not be seen. Instead a new binding mode including interactions with F33 and F189 was adapted. Further, docking experiments were in most cases able to point out the preferred carbon for alkylation to reproduce experimentally determined

58

regioselectivities but also here other steps on the reaction pathway could be involved in deciding product outcome. What docking does show are the fea-tures from having a more flexible substrate thus aiding in getting infor-mation out from the solved crystal structures. To try to map the reason for the changed regiopreferences seen there is a need to apply QM/MM calcula-tions.

59

Paper V Complex Kinetic Schemes are Required to Describe Epox-ide Hydrolase Catalyzed Production of Diols Pre-steady state kinetics were performed for studying wild-type enzyme, R-C1 and R-C1B1 (previously isolated for their ability of producing enriched R-4 from a racemic mixture of 1) with trans-stilbene oxide (5) and the small-er trans-2-methylstyrene oxide (3) in order to reveal the microscopic rates to explain product outcome. Crystal structures for all enzyme variants were reported earlier (Paper III and IV) and regioselectivites were analyzed in Paper I for epoxide 3 (Table 7). For epoxide 5, attack at either of the two carbons will result in the same me-so-hydrobenzoin product and thus experimental data for regioselectivity cannot be obtained. Instead, calculations for wild-type enzyme in Paper II show that for alkyl-enzyme formation the preferred attack is C1 for S,S-5 and C2 for R,R-5. However, the only pathway that leads to product for-mation is via C2 attack also for S,S-5 due to hydrolysis of C1-alkylenzyme being essentially kinetically blocked. During this work complementing pre-steady state kinetic measurements were performed with both 3 and 5. Calcu-lations did further show two different binding modes for more flexible sub-strates, structurally similar to 3 (Paper III), suggesting that they may also be present here. Table 7. The regioselectivities for 3, shown in preferred carbon for attack and the percentage thereof.

S,S-3

R,R-3

Preferred carbon

(%) Preferred carbon

(%)

StEH1 wt C1 >99 C1 55±0.3

R-C1 C1 >99 C1 66±0.6

R-C1B1 C1 >99 C2 89±0.4

Non-strict regioselectivity motivates the use of complex reaction schemes As can be seen from both product outcome and calculations regioselectivity is not strict, thus a branched reaction scheme is for sure needed to explain enzyme behavior (Scheme 3). The proposed formation of two different al-kylenzymes for (amongst others) S,S-5 should ideally be seen from pre-

60

steady state kinetics. However, here only one rate was detected. Since the preferred attack through C1 is kinetically blocked this would allow for quenching of the signal due to build up of alkylenzyme. The productive pathway, via C2 alkylation, on the other hand is followed by a fast hydrolyt-ic step thus possibly leading to less build up. This suggests that what we measure in the stopped flow could be not product formation, at least not alone, but dead-end alkylation. To complicate the situation even more the decay of alkylenzyme back to Michaelis complex for C1 attack is much low-er than the one for C2, thus this might reduce the signal again. To sum up, what is seen in the stopped flow could be a mix of the different microscopic rates while we assume the numbers belong to one single pathway and this is important to remember when analyzing data.

Different binding modes demands a further complex reaction scheme Preliminary simulation results together with literature143,144 suggest that the two binding modes from Paper III can also be available here for epoxide 3. The use of the branched reaction (Scheme 3) is convenient in trying to visu-alize different scenarios and explain product outcome. The detected new binding modes, however, may prove that this model should be revised to include more branches to cover also these. The different possible confor-mations for substrate binding do play a role in this case, something that is not seen with the bulky epoxide 5.

Pre-steady state kinetics do not reveal product outcome For some of the multiple-turnover measurements the curves looked nice although the intercept was put to zero and no reliable values could be ob-tained, most likely due to the fact that a mix of different transient rates are measured. There are different scenarios that can occur, that need calculations to be confirmed and here I try to show the problems one stumble upon dur-ing these measurements using the branched reaction scheme mentioned ear-lier and presented in Figure 13. In case (A) where there are two rates, two allowed alkylenzymes and two products formed the outcome is intuitive. If instead (B) two rates are seen but only one product is formed the explanation could be one of the hydrolyt-ic pathways being kinetically blocked. Due to the fact that different binding modes can result in two different rates for attack at the same carbon yet an-other case (C) occur where two rates are seen despite only one alkylenzyme being energetically allowed for. This means that a branched scheme (D) like in all cases is needed for one production pathway where each branch repre-sent one binding mode.

61

Figure 13. The branched reaction schemes used to explain various reactions scenarios that might arise. (A) two rates and two products are formed, (B) two rates despite only one product, (C) two rates and two alkylenzymes but only one carbon for attack meaning that two different modes are available that react at different rates. The last scenario (D) suggests one branched reaction scheme for each branch of the reaction in A, B and C due to two different binding modes forming the same alkylenzyme and product. For R,R-3 we do know that we have varying degrees of regioselectivity for most of the enzyme variants studied, resulting in two products, thus already adding to complexity when measuring multi-turnover kinetics. In addition, noisy signals were obtained and in some cases clear signals, indicating a sub-stantial build-up of alkylenzyme, but troubles in determining the actual rates. In order to pinpoint these behaviors there is a true need for QM/MM calcula-tions. Some systems could, however, still be explained as seen in the following section, where R-C1 and R-C1B1 are compared to wild-type enzyme.

R-C1 behaves similar to the wild-type enzyme R-C1 behaves similar to wild type and also in kcat/KM for R,R-3 due to a 5-fold increase in both kcat and KM (all kinetic parameters are presented in Ta-ble 8). The intrinsic rate constants of alkylenzyme formation is increased for both S,S-3 and 5, while rates for decay are similar where the energy barrier for alkylenzyme formation is lowered by around 1 kcal/mol. Also regioselec-tivity for S,S-3 is similar but for R-C1 only accumulation for one alkylen-zyme is suggesting a linear reaction pathway and if not, the other rate for alkylenzyme formation is low enough to be hidden..

RC1B1 deviate more from wild-type and R-C1 R-C1B1 differ more from the wild type where kcat/KM was decreased 30-fold for R,R-3, 200-fold for R,R-5, and a smaller decrease could be seen also for the (S)-enantiomers that could be explained by the increased active-site volume. Effects were expected also due to modifications of W106 where a main-chain amide interacts to stabilizing the tetrahedral intermediate and for a similar epoxide this residue was involved in stabilizing of one binding modes. This residue is therefore involved in both alkylenzyme formation and hydrolysis.

A

C D

B

Tabl

e 8.

Kin

etic

par

amet

ers f

or h

ydro

lysi

s of 3

and

5.

Enzy

me

k cat

(s-1

) K

M

(µM

) k c

at/K

M

(s-1×m

M-1

) K

S (µ

M)

k 2

(s-1

) k -

2 (s

-1)

k 3

(s-1

) k 5

+ k

-5

(s-1

) R

egio

- pr

efer

ence

g

R

,R-3

WT12

4 4.

7±0.

7 49

0±10

0 9.

7±2

- -

- -

8.1

± 3

55 %

C-1

R-C

1 25

±4

3300

±10c

7.7±

0.2

- -

- -

66

% C

-1

R-C

1B1

0.92

±0.4

34

00±2

00c

0.26

±0.0

1 -

- -

-

89 %

C-2

S,

S-3

WT12

4 63

±3

77±1

0 82

0±10

0 47

0±10

0 37

0±20

17

0±20

a 11

0±10

a 32

±2

99 %

C-1

R-C

1 65

±2

110±

10

560±

40

1500

±400

19

00±3

00

120±

30a

72±2

0a 44

±7

99 %

C-1

R-C

1B1

31±1

24

0±20

13

0±6

2700

±200

0 14

00±8

00

10±3

0a 32

±20a

99

% C

-1

R

,R-5

WT

17±0

.5

16±1

11

00±5

0 7.

8±13

18

0±10

0 13

±10d

42±0

.2c

-f

R-C

1 12

±0.5

20

±2

610±

40

13±1

7 85

±3

22±4

0d 28

±0.2

c

-

R-C

1B1

0.51

±0.0

9 17

0±40

3.

1±0.

1 -

- -

-

-

S,

S-5

WT

4.0±

0.06

1.

0±0.

08

3900

±200

5.

5±0.

9 23

±0.7

4.

5±0.

8d 3.

9±0.

005c

-f

R-C

1 4.

4±0.

09

1.3±

0.09

33

00±2

0 16

±5

120±

9 4.

2±6d

5.4±

0.01

c

-

R-C

1B1

4.0±

0.09

16

±0.9

25

0±0.

09

21±1

0 8.

9±1

5.1±

0.08

d 3.

0±0.

01e

-

a Calc

ulat

ed fr

om th

e ste

ady-

state

exp

ress

ion

for k

cat a

nd th

e de

term

ined

val

ues o

f kca

t and

(k-2+k

3). b Th

e pr

oduc

t is a

mes

o di

ol a

nd, h

ence

, the

diff

eren

t pro

duct

ena

ntio

mer

s are

in

disti

ngui

shab

le. c Ex

perim

enta

lly d

eter

min

ed in

sin

gle-

turn

over

exp

erim

ent.

d Val

ue c

alcu

late

d fro

m th

e ex

trapo

late

d va

lue

of (k

-2 +

k3)

from

mul

tiple

-turn

over

exp

erim

ent a

nd

the

expe

rimen

tally

esti

mat

ed k 3

from

sing

le-tu

rnov

er e

xper

imen

ts. e Ex

perim

enta

lly d

eter

min

ed in

sing

le-tu

rnov

er e

xper

imen

t but

with

not

fully

satu

rate

d en

zym

e. T

he d

eter

min

ed

valu

e is t

here

fore

≤ k 3

. f The p

rodu

ct is

a m

eso

diol

and,

hen

ce, t

he d

iffer

ent p

rodu

ct en

antio

mer

s are

indi

sting

uish

able

. g Dat

a fro

m P

aper

I.

63

When comparing microscopic rates of R-C1B1 to those of R-C1 the alkylenzyme formation rate for S,S-3 is similar while both rates of decay (k-2 and k3) are lower here. This suggests a more stabilized alkylenzyme and a lower turnover. For S,S-5 the alkylation rate is 15-fold lower while the decay is similar that could explain the 10-fold increase in KM. R,R-3 was modeled into the active site and the docking pose suggests that its interaction with the side chain of L109 changed upon modification to Y. This tyrosine side chain is turned away from the active site, allowing for R,R-3 to bind also in Mode1, pointing towards H300. This conformer allows for attack on C2, which can explain the shift in regioselectivity seen for R-C1B1. The modeling also indicate that the side-chain of L106 is shifted towards the substrate binding site, which could cause steric clashes with some of the binding modes and also hinder attack at C1.

Single-turnover experiments suggest a change in rate-limiting step Single-turnover experiments result in the most well-determined rate con-stants for k3. However, experiments like these demands a low dissociation constant (KS) to allow for saturation of the enzyme where substrate concen-tration should be at least 10-fold higher than KS and enzyme concentration a somewhat higher than substrate concentration. We were able to run single-turnover experiments for both enantiomers of 5 with wild-type enzyme and for S,S-5 with R-C1 and R-C1B1. For S,S-5 the values for kcat and k3 are similar and therefore hydrolysis is clearly the rate-limiting step. For R,R-5 on the other hand k3 values are substantially higher than kcat instead suggesting another step to also be rate limiting. What is seen from calculations (Paper II) is that there is one clearly preferred carbon for alkylation of about the same energy as the subsequent hydrolysis thus the rate-limiting step must be earlier on the reaction pathway.

Different binding modes lead to possible problems in computational reproduction of experimental data for pre-steady state kinetics QM/MM calculations are suggested as a solution for many of the challenges stumbled upon here, to get a clearer picture of the microscopic rates and dependencies. It is important to keep in mind that the build-up of alkylen-zyme for pre-steady state measurements relies on the formation being much faster than the decay. When trying to reproduce experimental pre-steady state data by theoretical modeling this will lead to difficulties, one example being the already discussed wild-type reaction with S,S-5, where the produc-tive route does not include the preferred alkylation and thus the signal we measure in the stopped-flow apparatus may not necessarily be coupled to

64

product outcome. Also, experimentally it is not possible to distinguish be-tween two different k-2 or k3 rate constants and what is monitored is a mix-ture if more than one contributes to the observed decay rate. Pre-steady state kinetics fail in describing the actual product outcome due to a mixture of reaction pathways and formation of two products. In some cases rate of alkylenzyme formation is too fast leading to troubles in catching the real rate. In others, signals are too weak probably due to insufficient build-up of alkylenzyme suggesting that hydrolysis might no longer be the rate-limiting step or what we measure is not purely product formation. This com-plicated story would for sure benefit from QM/MM calculations where there is a possibility to map both the different binding modes and the preferred carbons for attack but also steps that are kinetically blocked.

65

Paper VI Integrated action of Solanum tuberosum epoxide hydrolase I and Rhodococcus ruber alcohol dehydrogenase A in a two-step biocatalytic reactor Two enzymes were immobilized on separate aluminum-oxide membranes to perform a two-step biocatalytic reaction. This was done to study the poten-tial use in an industrial process, where immobilization is a beneficial ap-proach both for stabilization of enzymes in organic solvents and for re-cycling of the catalysts. My work is focused on one enzymatic step included in a larger reaction net-work (Figure 14) with the aim combining different enzymes to co-function in one pot. The product diols are useful chiral building block with both hy-droxyls having similar reactivity thus making selectivity somewhat hard to direct. Figure 14. The reaction starting from an epoxide that is hydrolyzed by epoxide hy-drolase to a vicinal diol. The next step involves oxidation by an alcohol dehydrogen-ase, either specific for primary hydroxyl groups forming aldehydes or for secondary hydroxyl group forming ketones. The latter is the one studied here, involving ADH-A. For this study R = phenyl group. The two-step reaction studied here starts from styrene oxide (2) that is hy-drolyzed to phenyl-1,2-ethanediol (6) by the engineered StEH1 variant R-C1B1D33. An aldoholdehydrogenase, selective for secondary alcohols was used for the second step, to form 2-hydroxyacetophenone (7) (Scheme 4). In addition to serve as a base for optimization of this two-step biocatalytic pathway the reaction also refines the cheap starting material in racemic sty-rene oxide to the more valuable product 2-hydroxyacetophenone.

66

Scheme 4. The reaction from styrene oxide (2) to phenyl-1,2-ethanediol (6) and further to 2-hydroxyacetophenone (7). The first step catalyzed by epoxide hydrolase and the second by alcohol dehydrogenase. The alcohol dehydrogenase of choice is ADH-A from Rhodococcus ruber and catalyzes the reversible reduction/oxidation of ketones and secondary alcohols. It belongs to the zinc-dependent group I alcohol dehydrogenases and is a suitable enzyme for biocatalysis due to its tolerance to high concen-trations of isopropanol and acetone.145 This tolerance makes it possible to use acetone for co-factor regeneration, however, the formed isopropanol is a preferred substrate as compared to the diol formed in step I (Scheme 4) thus engineering is required to reduce its activity with isopropanol. ADH-A was also shown to oxidize bulky substrates that are of importance for this project.

Optimization of reaction conditions of free enzymes First, the reaction conditions were optimized for the two enzymes to co-function free in solution. Green thinking and theoretical scale up was kept in mind, thus ammonium bicarbonate buffer was used that can be evaporated. pH optima differ for the enzymes and one has to make certain that the cor-rect conditions for the oxidation/reduction reaction are present. Since epox-ide hydrolase shows high activity for a rather broad pH range while ADH-A displays rather low activity with this substrate already at its pH optimum of 8, this was the chosen pH.

Co-factor regeneration & Inhibition problems For co-factor re-generation two different systems were studied and are pre-sented in Figure 15 where (A) reduction of acetone by the internal ADH-A and (B) reduction of pyruvate by lactate dehydrogenase (LDH). When using an internal enzyme in ADH-A for co-factor regeneration, a clean system is obtained, however, with the drawback of competition for the active site, where in this case both acetone and isopropanol are preferred substrates over diol 6. The alternative of an external enzyme, which possibly avoid the competition for the ADH-A active site but adds the problem of an additional pH optimum

67

to take into consideration and additional components added that could possi-bly complicate downstream processing.

Figure 15. The alternative systems for regeneration of NAD+ studied during this work: (A) ADH-A reduction of acetone to isopropanol and (B) LDH reduction of pyruvate to lactate. The second alternative includes an external enzyme, LDH, where pyruvate is reduced to lactate. Lactate showed to be a minor inhibitor of ADH-A and, to a larger extent, also for LDH, leading to difficulties in obtaining high ketone concentrations. Also, the product of the main reaction itself, 7, was previous-ly shown to inhibit LDH.146 To be able to use this system for co-factor re-generation the ketone concentration needs to be kept low by constantly with-drawing product from the reaction mixture. So far, the LDH alternative is the best method for ketone production where a maximum yield of 40 % was seen for a reaction with 33 mM 2 as starting material (Figure 16).

Figure 16. Ketone (7) formation with co-factor regeneration for acetone as external electron acceptor (dark grey) and LDH and pyruvate (black). Since NADH is known to inhibit ADH-A it is important to include a co-factor regeneration system to keep this concentration low. Also 7 is a poten-tial inhibitor thus a two-phase system where ketone can be removed upon formation is probably beneficial, particularly since this product is suggested to react with NAD+. Despite only two or three enzymes being included to co-

68

function there is already a challenge in finding optimal reaction conditions due to inhibition and different pH optima, resulting in a rather complex de-pendence network. When thinking of the environment and the cost-efficiency of the process acetone would be the best alternative, out of these two, since the system tolerates high acetone concentrations, acetone is cheap and an internal enzyme is used.

Characterization of immobilized enzymes

Enzymes were immobilized to alumina oxide membranes (Figure 17) and characterized one by one before being placed in a sequential mode.

Figure 17. The nanoporous alumina-oxide membrane and a schematic view of the coupling of StEH1 to the linker via a lysine residue of the enzyme. The same reaction conditions were applied for the immobilized system ex-cept that for initial studies no co-factor regeneration was applied. This al-lowed for product outcome to be monitored by following the formation of NADH by UV/VIS spectrophotometric measurements at 340 nm. What was seen from the results when comparing free and immobilized enzymes was a reduced activity for both R-C1B1D33 and ADH-A. A larger decrease for ADH-A could be expected due fuction being dependent on a dimeric en-zyme. The two enzymes, R-C1B1D33 and ADH-A, were able to co-function also in a sequential mode although at lower rates. 6 is known to be a poor substrate for ADH-A thus enzyme engineering should be applied to increase activity. Both epoxide hydrolase and ADH-A are active upon immobilization and the two enzymes did also function in sequential mode to generate product. Im-mobilization is not only beneficial for a possible increase in stability and the easy re-use of the catalyst and downstream processing but also to allow for tuning the reaction conditions in between the enzymes. Thus problems with optimization is possibly reduced where in the case of using acetone for co-factor regeneration this can be added after the epoxide hydrolase.

69

Conclusions & Future perspective

Great progress has been made towards understanding of the underlying mechanism of enantio- and regioselectivity shown by StEH1 and variants thereof. Theoretical models showed to be successful in reproducing experi-mental values for activity and selectivity but we are still far from the ulti-mate goal of being able to predict which modifications to introduce for a certain selectivity. What calculations did produce was suggestion of various binding modes and information of possible kinetically blocked steps of a reaction pathway. Thus, the combination of experimental data, crystal struc-tures and theoretical modeling has been crucial for trying to understand these complicated systems. We managed to get crystal structures for most of the enzyme variants isolat-ed, which together with experimental kinetic data have improved the theoret-ical models used to simulate the catalytic reaction. Yet, assumptions still have to be made due to lack of structures with reaction intermediates bound, something that is desirable in order to get closer to the actual structural background to selectivity. An enzyme variant with inactive hydrolytic ability (E35Q/H300N) was constructed to serve this purpose, but so far no solved structure did contain a clearly bound intermediate. Structures are one im-portant key for understanding structure-function relationships thus further optimization for crystallization would be useful. In addition, this set of modi-fications was introduced also to the other isolated enzyme variants from Paper I, and these should also be characterized and used in co-crystallization experiments. Another approach could be to modify H104, a residue previ-ously not considered to be involved in catalysis and therefore not well stud-ied. Enzyme variants isolated here did show a clear enrichment of product enan-tiomers but for an enzyme to be useful in e.g. pharmaceutical industry the purity needs to get higher. Further in vitro evolution to search for variants with improved but also varying selectivity are wished for. For the furthest evolved variant (R-C1B1D33E6) six mutations were introduced. The hit rate was decreased and before moving forward to additional mutagenesis at the active site, stabilizing mutations should be introduced. It could be favorable to do an inventory of the introduced mutations where V141K has shown to have only minor effects on product outcome and this should be investigated in more detail. Going back in the evolutionary tree and restart from a revert-

70

ed variant would be interesting, to see if other mutations would be possible. Yet another benefit from reverting this mutation is to avoid the trouble this side chain cause for structure solving, due to its high flexibility. To minimize library size, a degenerated codon set (NDT) was used allowing for twelve out of twenty amino acids. It would be interesting to also include the remaining eight by using an unbiased set of primers for all 20 amino acids.132,133 The number of variants would increase from 122 to 202 for librar-ies of two residues, which would also be possible to screen. However, an alternative screening method should be investigated allowing for a higher throughput and possibly to directly getting the selectivity and activity out. This would allow us to identify and characterize also variants with decreased activity or selectivity something that is highly informative in mapping struc-ture-function relationships. A suggestion could be to couple the screening reaction to ADH-A to be able to measure NADH formation by following absorbance at 340 nm giving directly a measure of the efficiency of the two-step reaction. In industry, the use of the same enzyme for several reactions would be bene-ficial. As a step towards this the use of the same libraries for screening also other substrates could be helpful as well as to measure activity for the en-zyme variants isolated as hits here with a range of epoxides, to see what could be fished out. Here, the immobilized enzymes could be used in a se-quential mode to detect epoxide hydrolase activity indirectly by measuring NADH formation for ADH-A. I would also like to continue the work with the coupling of StEH1 and ADH-A, to optimize co-factor regeneration and other reaction conditions and try other combinations of variants and substrates. Further, it would be interest-ing to continue the work with co-expression of the enzyme in E. coli or to express them on the surface to avoid possible problems with transporting the epoxide in to the cell. I would like to do more research about present indus-trial processes and what parameters to consider when scaling up. I would like to find potential industrial processes including a step of epoxide hydrol-ysis where StEH1 could replace existing catalyst to improve enantiomeric excess of product and thereby the yield.

71

Populärvetenskaplig sammanfattning Mot förståelse av selektivitet & enantiokonvergens hos ett epoxidhydrolas

”Jag får enzymer att göra saker de egentligen inte kan så att de skulle kunna användas för framställning av rena produkter i exempelvis läkemedelsindu-strin.” Detta är vad min beskrivning har kokat ner till med åren. Att förklara att man jobbar med biokatalys säger inte folk i allmänhet vidare mycket. Vad är bi-okatalys? En katalysator vet många vad det är, man har det i bilen. Det är något som snabbar på reaktioner utan att själv förbrukas – mycket effektivt! Bio syftar till att det är naturens egna katalysatorer, i form av hela celler eller rena enzymer, och de kan snabba på reaktioner upp till miljarder gånger genom att positionera molekyler (så kallade substrat) perfekt för reaktion och genom att stabilisera vissa labila mellanprodukter (Figur 18). Enzymer är kedjor av aminosyror och är fantastiska små varelser som finns överallt; i naturen, i djur och människor där de ser till att reaktioner sker, lagom fort och ofta. Det fina är att man kan använda dessa utanför deras naturliga habi-tat för att snabba på reaktioner av olika slag inom exempelvis läkemedelsin-dustrin.

Figur 18. En schematisk bild över hur enzymet epoxid hydrolas omvandlar substrat till produkt. Varför är biokatalys viktigt? Vi blir mer och mer medvetna om att vi måste värna om miljön och hushålla med jordens resurser. En stor industri som definitivt behöver göras mer miljövänlig är kemikalieindustrin där nuva-

O

OH

OH

72

rande processer genererar miljöfarligt avfall och ofta kräver extrema tempe-raturer och tryck. Man försöker på många vis göra dessa processer grönare genom att exempelvis försöka utföra reaktioner i vatten istället för i orga-niska lösningsmedel. Här kommer biokatalys och enzymer väl till pass då dessa redan arbetar optimalt i vattenlösningar och vid skonsamma tempera-turer samt pH. Utöver detta är de biologiskt nedbrytbara och de flesta enzy-mer har liknande optimala arbetsmiljöer så att flera kan samverka i ett och samma kärl, något som reducerar mängden avfall samt antalet reaktionssteg. Enzymer är även, som titeln säger, selektiva i att de ofta är utvecklade för att utföra en viss reaktion med ett visst substrat. Detta kommer ifrån att enzymer har ett så kallat aktivt säte där reaktionen sker och detta är oftast skräddarsytt för ett visst substrat vilket innebär att de kan välja ut ett visst molekylslag från en blandning. Denna egenskap gör processerna mer effektiva då den till och med möjliggör framställning av enbart den ena av två spegelbildsmole-kyler, något som är eftertraktat i läkemedelsindustrin (Figur 19). Figur 19. Två spegelbildsmolekyler eller enantiomerer som de kallas. I de fall då enzymet enbart katalyserar reaktion för den ena substratmoleky-len (kinetisk upplösning, övre delen i Figur 20) så erhålls ett utbyte på max-imalt 50 % från en substratblandning. Detta innebär i och för sig att en ren produkt erhålls men även att hälften av ursprungsmaterialet slösas bort. Man kan blanda olika enzymer med kompletterande egenskaper för att över-komma detta men det finns även enzymer (däribland huvudrollsinnehavaren i denna avhandling) som har den fantastiska egenskapen att kunna göra en och samma spegelbildsprodukt från en blandning av två spegelbildssubstrat, ett fenomen som kallas enantiokonvergens (nedre delen i Figur 20). Denna förmåga är viktig då de två produkterna kan, även om de egentligen är väldigt lika, ha helt olika egenskaper. Ett exempel är smakerna citron och apelsin (S- och R-limonen). Särskilt viktig är selektivitet när det kommer till läkemedel där det oftast bara är den ena av de två spegelbildsmolekylerna som har den önskade effekten medan den andra kan vara rent av skadlig. Så var fallet med neurosedynincidenten på 1960-talet, då gravida kvinnor re-kommenderades en lugnande medicin som bestod av en blandning av två spegelbildsmolekyler där den ena visade sig ha fosterskadande effekter. Nu för tiden har man kännedom om de ofta skilda egenskaperna och ser till att

C C

73

rena upp de två spegelbildsmolekylerna ifrån varandra. Alternativt kan man försäkra sig om att den andra är ofarlig, men av ekonomiska skäl önskar man så klart att framställa enbart den önskvärda molekylen, och särskilt på ett enantiokonvergent vis. Figur 20. Överst visas kinetisk upplösning och under enantiokonvergens, pilen indi-kerar var attacken för reaktion sker. Förr i tiden extraherades enzymer ur sina naturliga källor, i mitt fall potatis-blasten, vilket kunde vara både tidsödande och ge ett lågt utbyte. Men med dagens rekombinanta genteknik kan man konstruera genen som kodar för det enzym man önskar och få en bakterie att uttrycka detta för att sedan rena upp enorma mängder enzym på ett enkelt vis - mycket revolutionerande! Följakt-ligen utgör själva tillgången på enzym oftast inte längre något hinder utan vad som är svårt är istället att hitta ett enzym som gör precis det man önskar. Naturens enzymer har utsatts för selektionstryck under miljontals år i syfte att effektivt utföra en viss reaktion och detta försöker vi återskapa under kontrollerade former genom att utföra evolution i provrör. Vi använder så kallad riktad evolution där vi snabbar på evolutionen genom att modifiera ett enzym i syfte att skapa nya egenskaper. Hur får man då enzymerna att göra det man vill? Hur får man enzymer att göra något de egentligen inte kan? Det finns tredimensionella bilder av en-zymerna som erhålls genom att skjuta röntgenstrålar på enzymkristaller. Med tillämpning av denna metod får man data för att kunna studera hur en-zymet ser ut och hur molekyler binder där inuti. Härifrån görs kvalificerade antaganden avseende modifieringar som exempelvis tillåter större eller mindre molekyler att passa in eller styr bildandet av den ena eller andra spe-gelbildsprodukten. Det är dock mycket svårt att förutse hur enzymet kom-mer att reagera på förändringar varför stora bibliotek innehållande varianter med olika modifieringar framställs. För att hitta de varianter som har just den eftersökta egenskapen behövs en snabb metod för att mäta hur bra de är, en så kallad screen. När vinnaren har identifierats används denna för att skapa nya varianter och så vidare tills ett enzym med de rätta egenskaperna isole-rats. Skapandet av bibliotek och screening är en tidskrävande process som man inte vill upprepa varje gång en ny egenskap efterfrågas. Det är därför

O

R

O

R

R

OH

OH

O

R

O

R

O

R

R

OH

OH

74

viktigt att försöka förstå mekanismen bakom den nyvunna egenskapen, nå-got som kartläggs med den typ av grundforskning som bedrivs i akademin. I syfte att kartlägga förekommande struktur-beteende-samband har vi kom-binerat experimentella data och kristallstrukturer av enzymer med beräk-ningar. I detta sammanhang har beräkningskemisterna till uppgift att ta fram teoretiska modeller som beskriver reaktionen där de utgår ifrån tillgängliga enzymstrukturer och experimentella data för att träna modellerna. Under förutsättning att detta lyckas för en mängd olika enzym-substrat-kombinationer är tanken att, med hjälp av datorn, kunna förutsäga vilka mo-difieringar som krävs för att få en viss aktivitet eller selektivitet även för andra, liknande molekyler. I mitt arbete började jag med att göra tusentals varianter av epoxidhydrolas från potatis (StEH1), ett enzym som består av 321 aminosyror. Målet var att skapa en modifierad variant som endast framställer den ena av två spegel-bildsmolekyler. I detta avseende lyckades jag identifiera flertalet nya en-zymvarianter med nya selektiviteter genom att bara byta ut upp till sex stycken aminosyror. Vissa var selektiva genom att bara reagera med det ena av de två spegelbildssubstraten medan andra visade sig vara enantiokonver-genta som beskrivits ovan. Med andra ord är det möjligt att producera en och samma spegelbildsmolekyl från båda substraten. Härvid kan tilläggas att dessa enzymvarianter visade sig fungera bra även med andra, liknande sub-strat vilket är önskvärt om enzymet ska användas i industrin. Vi vet sedan tidigare hur strukturen för epoxidhydrolas från potatis ser ut och lyckades här även få strukturer för de modifierade varianterna (Figur 21). Utifrån dessa var det möjligt att härleda hur substratet band och vilka aminosyror som eventuellt bidrog mest till de nya egenskaperna (Figur 21). För att få en djupare förståelse för detta så tillämpades ovannämnda beräk-ningar, där beräkningskemister tillhandahölls underlag i form av kristall-strukturer och siffror för enzymernas beteende med olika substrat med avse-ende på katalytisk förmåga och selektivitet. Genom dessa beräkningar identi-fierades ytterligare några aminosyror som är involverade i mekanismen, vilket har bidragit till en djupare förståelse för hur reaktionen går till och hur enzymet bidrar till processen. Därutöver har vi, i vissa fall, kunnat identifiera vilket steg i reaktionen som bestämmer selektiviteten och vilka aminosyror som har störst inverkan på både selektivitet och reaktivitet. Så det är möjligt att med hjälpa av datorn reproducera de data som erhålls i laboratoriet men ännu är det långt kvar till det långsiktiga målet att underlätta det praktiska arbetet genom att tillämpa teoretiska beräkningar för att förutspå vilka modi-fieringar som krävs för att få en viss egenskap hos ett enzym men vi är på god väg. De teoretiska modellerna behöver tränas mycket mer, med mer data för fler varianter och fler substrat. Dessutom försöker vi att få fram struk-turer med substrat bundet i för att se hur det verkligen ser ut när detta sker.

75

Detta skulle inte bara rent visuellt bidraga till förståelsen utan också under-lätta än mer för beräkningskemisterna och på så vis leda oss än närmare en kartläggning av struktur-funktions-sambandet. Figur 21. Till vänster ses strukturen för StEH1 där de aminosyrorna som är involve-rade i katalysen är färgade svarta, de visas även förstorade till höger i bild. Man kan härifrån se att de katalytiska aminosyrorna är väldigt få i jämförelse med hela prote-inet (321 stycken). Sålunda har vi skapat en verktygslåda med en mängd strukturer, teoretiska modeller och enzymvarianter att välja ifrån för att utföra reaktioner med olika aktivitet och selektivitet för olika substrat. Enzymerna har modifierats och studerats i detalj (ett och ett) för att sedan kunna kopplas ihop med andra enzymer i syfte att utföra flerstegs-syntes. Fördelen med att ha rena enzymer är att man själv sätter ihop dessa för att få just den reaktion man önskar och på så vis slipper andra störande metabola vägar som uppstår vid användning av hela celler. Fördelen, å andra sidan, med att ha hela celler är att cellen själv reglerar produktion av andra molekyler som kan vara nödvändiga (och dyra att framställa) för en effektiv reaktion. Slutligen så tror jag inte att enzymer kommer att ersätta alla steg i kemisk industri utan att man låter de två alternativen samverka. Enzymer kan med fördel användas för de steg där selektivitet är ytterst nödvändig medan tradit-ionell organisk syntes kan behållas för de reaktionssteg som fungerar väl utan användning av höga temperaturer eller kemikalier med negativ påver-kan på miljön. Det bedrivs mycket forskning i syfte att göra den kemiska industrin mer miljövänlig men man använder fortfarande huvudsakligen organiska lösningsmedel så för att enzymer ska kunna implementeras i de existerande processerna krävs att de modifieras för att överleva i denna ona-turliga miljö. Ett stabiliseringsalternativ innebär bindning av enzymet till tunna membraner, en smart metod då det enkelt går att återanvända systemet för andra reaktioner. Jag tror på att vi måste utnyttja det bästa av båda värl-dar, ett samarbete mellan handgjorda och naturens egna katalysatorer, för en hållbar industri.

76

Acknowledgement

There are so many people I would like to thank, both inside and outside the BMC-building. Thank you all! Now I am ready for a life in the real world and hopefully people will benefit from what I have learnt here. Yey! First I would like to thank my supervisor M i k a e l W i d e r s t e n for al-lowing me to think my life through before I jumped on this train. For trying to guide an engineer with a what-could-we-use-this-for mind into the fasci-nating world of enzyme kinetics. For your unbelievable fascination for eve-rything we do, being it a perfect SDS-PAGE gel or listening to the sound of stopped-flow for a whole afternoon. For your ability to always see possibili-ties in all results even though me myself sometimes lost all hope. Your en-thusiasm has been invaluable. I would also like to thank my co-supervisor H e l e n a D a n i e l s o n for always encouraging me to keep my engineering-hunger. For helping me conclude what I want and what I need to do to fulfill that when my own mind was sometimes a bit blurry. Thanks for pushing me forward in your own slightly brutal but caring way. I would like to thank my co-workers L y n n , P a u l , B e a t , F e r n a n -d a , D o r e e n , M i k a e l N , H u a n , G u n n a r & E r i k a without you guys this thesis would have been awfully thin. I could not thank you enough! Thank you Å F o r s k & L i l j e v a l c h s & W a l l e n b e r g s for allow-ing me to go to America, Japan and Hong Kong to present my research (and to do some sightseeing and cute-stuff shopping). I have learned a lot from interacting with researches from all over the world. I am also grateful for the help in the proofreading of this piece of work by H u a n , R i k a r d & C h r i s t o f f e r your comments highly improved the quality of the story.

77

and now i will completely skip grammar and spelling and first give a huge thank you to the widersten group. my second family for the last five or so years m i c k e för att du lärde mig hur man labbar. eller hur man inte labbar. för att du alltid får mig att skratta. trasiga glas(ögon) och soppa i väskan. för att du äntligen köpte dig en hjälm. på rea. för att du aldrig släp-per taget om din kaffekopp. forskningsidéer. favoritskjortan. nyfikenheten. radarbilderna. för att du alltid är positiv. för dina stora visioner (trädäck, JACS, babyhalkstrumpor till hundar). för att du skapar en fantastisk atmosfär i gruppen. som får en att känna sig som hemma från första stund. för att du höll dig mer informerad om min fotboll än jag själv. och för att du hjälpte mig att ro detta i hamn a n n & d i a n a för att ha banat väg. lite svårt att axla era mantlar bara då det ställdes riktigt höga krav på mig. diana, sättet du presenterar en på är förundrande. tack c i s s i för klättringsup-plevelser. och för alla resor. dödsturer på manhattan. 4th of july-firande i hålor som gud glömde (rosetten är fortfarande den finaste present jag nånsin fått och sura tanten kan fortfarande ge mig rysningar). färglada flingor på super8. nakentur i japansk taxibil (fortfarande det syndigaste jag gjort). det kändes stort att släppa iväg dig till amerika. utan att bära vare sig ditt pass. din packning. eller dig h u a n min glada häst. for all unforgettable mo-ments. always with a cute touch. for afro classes. for being my tea supplier. wherever you go i will come visit you. for all etsy inspiration. for your fighting spirit. for always being there. in sad and happy times. future is soon here. with awesome jobs homes and whatever you wish for. for you and your petit prince emilien merci for all post-it decorations in our office

e m i l för att du är lika noggrann och eftertänksam som jag. skönt att inte vara ensam. kanske lite väl överdrivet att köra om mig med avhandling-splaneringen här på sluttampen bara. tack för inspirationen (stressen) att skaffa unge. men jag går inte på det en gång till ☺ n i s s e för strukturen. både på lab och de faktiska kristallerna för att du uppmuntrade min ord-ningsamma ådra och för att du bjöd hem oss till din lilla familj. för din hu-mor. och för att du var som en storebror på labbet. som tog hand om allt. visste allt. och retade cissi. era duster var tuffa i början men jag härdades ☺ saknar er båda! t h i l a k sorry for being in a stress bubble trying to finish everything off instead of getting to know you. exam & project workers

h u a n p a u l m a h a h e n r i k for all the useful data you collected for me. and thank you for letting me be the boss. for once a n d r e a s r o b i n v i t a l i for keeping the lab occupied and waking me up from the computer every now and then. i hope the stopped-flow machine will be okey soon. it is probably just sad. sing for it – and it will sing for you ♪ ♫ ♪ ♫ ♪ ♪ ♫ ♪ ♫ ♪ ♪ i would really like to thank my co-workers again for doing a lot of hard work while i was abscent for a few months. without you guys this thesis would not

78

have been what it is today. thank you d o r e e n for teaching me and guid-ing me all the way through the crystallization experiments and for always having time and showing enthusiasm. and for baking awesome cakes. you will become a supermom d i r k for taking care of my monster crystals now, let us make them get you a laboratory engineer. everything will be fine. you will have a kick-ass introductional part ☺ g u n n a r j för att du vet allt. institutionens levande uppslagsverk. tack för all ovärderlig hjälp med immobiliseringsmanuskriptet. och e r i k a min stjärna. allt jobb du gjort för vårt lilla manuscript. tack! för att jag hittade min smörkompis i japan. jag kommer aldrig att glömma vår fantastiska resa. för all spinning och pump och andra campusevent. för alla viktiga samtal och sms-trådar. tack för allt engagemang i jobbjakten. hej linus! tack för leenden. födelsedagspresenter. och glöm inte att fortsätta med piff även efter minis ankomst. ser fram emot dagslånga fikor med våra kids. arbetslösheten kommer att vara en dröm ☺

l y n n k a m e r l i n & g r o u p for the endless hours of calculations and re-calculations. and re-re-calculations. a special thank you to

p a u l i e (i know you have always loved your nick name) for answering my questions over and over again. and again. hang in there ☺ i would also like to thank all the past and present members of the HD group. in particular t o n y for introducing me to climbing. and the hilarious world of german-swedish. haha. for all nice moments in uppsala and during trips. for your easy-going lifestyle. for laughing at me for not having friends here when I first arrived e l d a r tjenamors din gamla galosch! sorry for leaving you hanging but you turned out just fine. new job and all. commut-ing is fun! c h r i s t i a n s you and your muffins and LateX j o h a n w för den hemliga posten. för att du förgyllde tillvaron tillsammans med royalisten s o f i a och all-time-high-skrattets ägare a n g e l i c a

s a r a s för alla snickesnack i stopped flow-rummet. hoppas att ni har det bra i danmark. thanks to the small but naggande goda YI group g u s t a v så ung. du har lååång tid kvar. tack för att du inte övergav oss vid lunchbordet. de ibland något opassande samtalsämnena till trots. men många gånger var det faktiskt du som började. joho! y l v a cool som få. å nyfiken. skönt att det kom-mer in lite unga forskare med livsgnistan i behåll ☺ å tack för att du tar på dig frågeställarkoftan. i would also like to post a big thank you to all past and present organic chemists from the dark and the bright side. for all hej hejs in the corridor

r i k a r d för alla tips och goda råd. för alla lunchsamtal och milslånga facebook-konversationer med dig och @johanna. för att du tittar förbi då och då för att se att allt går som det ska. för att du hejade mig hela vägen över mållinjen. nu kan du läsa den svenska sammanfattningen ☺ när ska jag komma till nya boet med skumpa? a n n a för att allt är så enkelt. aldrig

79

några problem. för att du gör det du tror på och struntar i vad andra tycker. vi får ta en playdate allihopa snart! s a r a för all inspiration. sagolika foton. unge. allt. kanske att du tar det lite lugnare nu och inte hetsar upp dig för att någon cyklar en liten bit in i cykelbanan? tack för alla tips för att inte bli ruinerad under mammaledigheten. verkligen. tack m a g n u s för att du var i samma sits som jag. på många vis. skönt att ha någon att prata med. mitt hjärta är för evigt ristat av våra du-vet-vad. så mycket skitsnack har jag aldrig fått ur mig gällande en och samma sak. vad hände med vårt samar-bete? jag har fortfarande en box med photoswitchisar i min frys. men fort-farande ingen signerad avhandling à la blom j o h a n v välkommen hit. välkommen hem. ditt ständiga leende och fniss var saknat. känns bara lite ovant att vi inte kan bestämma över ditt huvud längre. men man vänjer sig nog. i sinom tid. du borde alltid ha med dig lite fairy dust i fickan

c h r i s t i a n vår egen cover-boy. måtte du hitta en ny pizzeria. å det är alls inget fel på åttiotalsmat. kasslerlåda. köttfärslimpa. å lite after eight-päron på det. fint t o b i a s för alla lunchspinningpass. för pendlings-förståelse. för konferenssällskap på andra sidan jorden och för jobbmedlande

a l e x a n d r a for sharing the stress and reminding me that life is good and everything will turn out just fine c a r i n a for nice walks to ångström and excellent teaching company t h o m a s n tack för världens finaste cykel och ved och för att du förstått vikten av fika och kärleken till råsunda. sushi & choklad – det håller magen glad. For all people who made teaching a true pleasure f r a n ç o i s e alltid positiv och glad. och omtänksam som få. ledsen för ditt första möte med ni får göra upp det där någon dag. merci för att du finns. som de säger i chokladreklamen d a v i d r för att du gjorde undervisningen av grundläggande kemi till ett rent nöje. när ska vi göra raketbjörnar igen? jag fixar scones. efter recept l i l i a n för att du gjorde undervisningen till en dans på rosor. hoppas att du har det lika bra där borta på andra sidan villorna

j o h a n n a a för all hjälp med att hitta lilians gömställen samt stöd när man behöver prata av sig efter en tung dag på labbet. uppsala är en mysig stad, glöm inte det. tack alla administratörer som hjälper till med allt måste-arbete även om ni försöker outsourca det mesta på gemensam lön. och ibland tar er lite väl stora friheter som i att ordna bokningssystem (gurra. visst är du fort-farande arg?) vad är det för fel på penna och post-it? j o h a n n a j jojo. biffen. johanna sektreterare. master of coins. kärt barn har många namn. för att du lyser upp en grå institution med rosa mössa, filt och stövlar. tack för all hemnetinspo. spinning-fighting-spirit (föredrar dig framåtlutad med mördarblick där framför på cykelbanan). skönt att ha någon att diskutera gäddhäng å fettvolanger med. någon som förstår. den första som fått mig att tvivla på mitt motto att ärlighet vara längst. utan våra sms-trådar med erika vore den här avhandlingen på tok för lång. så tack! i n g e r för att du all-

80

tid hade tid när man kom och knackade på. och hade access till $ kreditkortet $ hoppas att du har det bra även utan oss l i n a j för din sköna inställning till livet å ditt luriga leende. för att du kom utifrån och berättade att det är viktigt att ha semester och så. så att vi lite fick känslan av att vi har ett riktigt jobb e v a för att man äntligen fick ett ansikte på den där mailutsändaren. tack för allt fix med mp. allt finns på mp. det har jojo lärt oss. eller hur eri-ka? g u n n a r s för alla snack. inte minst under min första sommar här då alla andra verkade ha semester. för all glädje du sprider som julafton varje dag! b o s s e för all sponsring till fotbollen och all hjälp med pro-gram hit och dit. även om du höll dig på säkerhetsavstånd från a n n a t för kakboden. eller ja. egentligen gillade jag den inte alls men att ha en egen vending machine som tar rosa post-it är ren lyx. tack till alla som bidrog till att jag valde att följa den här vägen. tack

v a l t e r & n i l s - g u s t a v för osynlighetsskrift och klurigheter. till alla som gjorde lundatiden till en lek p101 & c:o för tegelbyggnadshäng och våra försök till att ta död på både KRUT och overallstraditionen. gjorde vårt karaoketält verkligen succé? ett särskilt tack till m a r i a & r o b b a n för att jag fick bo på kontoret trots att det var ämnat åt alla demoprylar från su-perföretaget. tack för alla skratt. för tighta tv-tablåer. för nintendomaraton. för bakmaskinsuppvaknandet. tack m a u r i z i o för att du var en fantas-tisk handledare som fick mig att vilja doktorera. för alla jobbtips. jag kom-mer om några år! u p p s a l a tack för att du är en så fin stad. för alla små jurister som bidrog till den mysiga fritidsgårdskänslan på kantorsgatan. tack

f o t b o l l e n som tvingade mig att strukturera arbetet för att kunna cykla från den ena änden av stan till den andra och bara få vara. få lite skäll av max och tro att försvarsarbete plötsligt är det enda viktiga. och att skratta. för alla kompisar där. särskilt de som även fanns utanför plan m i a m a d d e c o r n e l i a t i l d a h a n n a l i n a l tack h e l e n a för att du alltid säger att det kommer att gå bra. och sedan erkänner att du inte har en aning ☺ för att man kan prata med dig om allt. för att du bryr dig om. så glad för att vi träffades den där julen. i den där alpbyn j e n n i e för att du alltid är så vettig och genombäst. min stöttepelare och vapendragare. att bli ihop-blandad med dig under hela gymnasietiden var ett rent nöje. jag ska se till att vi kommer till göteborg så att det kan hända igen b r u n b e r g & p e -t e r z o n mottot “det ordnar sig” har jag haft med mig under hela skri-vandet. mycket lugnande m a m m a d a g i s denna brokiga skara som fann varann ändå. så skönt att ta en paus från allt vetenskapligt dravel å läsa ikapp lite logiska mammafunderingar i chatten. snart står jag åter till ert för-fogande ☺ d a v i d g för att du vet allt. å kan allt. å hjälper till med allt. för att du berättar att man kan jobba i morgon och att man inte ska skicka lamptips mitt i natten före jobbdag l a d i ladan . för att du alltid är så peppande och förstående. och för all involvering i min jobbjakt

j o n s s o n den bästa gudmodern och vännen man kan få tack för att du älskar ruth som vi gör. för att du fyller min garderob när jag inte har tid att

81

tänka på sånt. för att du får mig att skratta. med dig. fast kanske mest åt dig. för att du föreslår singelbidrag när möbler blir för dyra. nu blir det utflykter igen. tack ♫ tidy tidy tidy up. I’d like to sit down on my potty. no monsters living in our home. five little ducks. wind the bobbin up för att ni fick mig att slappna av och andas justin bieber. jag hoppas innerligt att min kärlek till din love yourself är av det övergående slaget. sist men inte minst vill jag tacka min familj. tack för att ni inte är som alla andra. m a m m a ett gott skratt förlänger livet så det har vi dig att tacka för “eller hur. pappa?” för att du då och då ringer för att fråga vad det är du ska säga att jag jobbar med igen. det var verkligen lättare när jag höll på med bagerijäst. ska tänka på det nu inför nästa jobb. tack för all envishet. att du visat att allt går. knutar är till för att lösas. och kryddburkar kan man klämma sig i p a p p a för att du kan allt. och hjälper till med allt. fixar-frasse. med en egen MacGyver i familjen är det klart att man lär sig lösa diverse tekniska problem på labbet. gräddvispsklubben är utökad! k i m för alla in-spirerande samtal om andra saker än forskning. som filmer. serier. politik (Z z z Z Z Z z z ). för att du ger ruth det vi inte kan. intellekt och skägg.

t o n y a n n i k a & a d r i a n för den familjeförebild ni är. vi må ha skaffat unge först men vi kommer aldrig ikapp. hus. robotgräsklippare. och utspädd juice j a n & g u n i l l a för allt ni gör för oss. för att ni tycker att vi är viktigast i världen m o r m o r för alla telefonsamtal om allt och inget ℡ för din vassa tunga och rappa humor. alla korsordsstunder som gjort mig till den klurare jag är idag. morfar skulle bara veta. tack för lingon i mjölken och aromat på bordet. tack f a r f a r för allt du och farmor har gett mig. alla kortspelsstunder. våfflor. kolaglassar från källaren. för att du alltid visar enorm glädje när man hör av sig även om du har hundra andra barnbarn och barnbarnsbarn. för att farmor änglavaktade mig när jag körde av vägen

m o g g e & i n g e r å minifamiljen i minihuset. för all trivsel ni bidrar med. för alla uppmuntrande ord. och skrattfester.

c h r i s t o f f e r & r u t h för att ni finns. snarkar när jag åker hemifrån. så glada att ni inte kan stå still när jag kommer hem. tack g o s s i p g i r l & a b l o g t o w a t c h för att ni tar hand om c h r i s s e som i sin tur tar hand om allt. nästan. tack för att du introducerade detfinnsingarenaklä-dervimåsteköpanya-begreppet. trevligt ändå. tack för att du säger att jag inte ska curla mina studenter. för dina pappa-skills och rut(h)iner. för köttbullar och falukorv. för pyttipanna och kebab. vilken hemmaman! r u t h tack för timingen som gjorde att artiklar blev publicerade medan jag hängde i lekhagar och käkade majskrokar. utan dig hade denna avhandling inte blivit vad den blev. nu blir jag skojig igen. lovar ! xoxo

82

References

1. Sheldon, R. A. Fundamentals of green chemistry: efficiency in reaction design. Chem. Soc. Rev. 41, 1437–51 (2012).

2. Kondo, A. & Ueda, M. Yeast cell-surface display - applications of molecular display. Appl. Microbiol. Biotechnol. 64, 28–40 (2004).

3. Jose, J., Maas, R. M. & Teese, M. G. Autodisplay of enzymes - molecular basis and perspectives. J. Biotechnol. 161, 92–103 (2012).

4. Chanda, A. & Fokin, V. V. Organic synthesis ‘on water’. Chem. Rev. 109, 725–748 (2009).

5. Park, S. & Kazlauskas, R. J. Biocatalysis in ionic liquids – advantages beyond green technology. Curr. Opin. Biotechnol. 14, 432–437 (2003).

6. van Rantwijk, F. & Sheldon, R. A. Biocatalysis in ionic liquids. Chem. Rev. 107, 2757–85 (2007).

7. Gorke, J., Srienc, F. & Kazlauskas, R. Toward advanced ionic liquids. Polar, enzyme-friendly solvents for biocatalysis. Biotechnol. Bioprocess Eng. 15, 40–53 (2010).

8. Reetz, M. T. Biocatalysis in Organic Chemistry and Biotechnology : Past , Present , and Future. J. Am. Chem. Soc. (2013). doi:10.1021/ja405051f

9. Schmid, A., Dordick, J. S., Hauer, B., Kiener, A., Wubbolts, M. & Witholt, B. Industrial biocatalysis today and tomorrow. Nature 409, 258–68 (2001).

10. Strohmeier, G. A., Pichler, H., May, O. & Gruber-Khadjawi, M. Application of designed enzymes in organic synthesis. Chem. Rev. 111, 4141–4164 (2011).

11. Tao, J. & Xu, J. H. Biocatalysis in development of green pharmaceutical processes. Curr. Opin. Chem. Biol. 13, 43–50 (2009).

12. Schoemaker, H. E., Mink, D. & Wubbolts, M. G. Dispelling the myths - biocatalysis in industrial synthesis. Science 299, 1694–7 (2003).

13. Clouthierz, C. M. & Pelletier, J. N. Expanding the organic toolbox: a guide to integrating biocatalysis in synthesisw. Chem. Soc. Rev. Chem. Soc. Rev 41, 1585–1605 (2012).

14. Kraut, D. a, Carroll, K. S. & Herschlag, D. Challenges in enzyme mechanism and energetics. Annu. Rev. Biochem. 72, 517–71 (2003).

15. Karplus, M. & McCammon, J. A. Molecular dynamics simulations of biomolecules. Nat. Struct. Biol. 9, 646–652 (2002).

16. Hay, S. & Scrutton, N. S. Good vibrations in enzyme-catalysed reactions. Nat. Chem. 4, 161–168 (2012).

17. Callender, R. & Dyer, R. B. The dynamical nature of enzymatic catalysis. Acc. Chem. Res. 48, 407–413 (2015).

18. Kohen, A. Role of dynamics in enzyme catalysis: Substantial versus semantic controversies. Acc. Chem. Res. 48, 466–473 (2015).

19. Agarwal, P. K., Billeter, S. R., Rajagopalan, P. T. R., Benkovic, S. J. & Hammes-Schiffer, S. Network of coupled promoting motions in enzyme catalysis. Proc. Natl. Acad. Sci. U. S. A. 99, 2794–2799 (2002).

83

20. Ma, H., Szeler, K., Kamerlin, S. C. L. & Widersten, M. Linking Coupled Motions and Entropic Effects to the Catalytic Activity of 2-deoxyribose-5-phosphate Aldolase (DERA). Chem. Sci. 00, 1–7 (2015).

21. Boehr, D. D., Nussinov, R. & Wright, P. E. The role of dynamic conformational ensembles in biomolecular recognition. Nat. Chem. Biol. 5, 789–96 (2009).

22. Henzler-Wildman, K. & Kern, D. Dynamic personalities of proteins. Nature 450, 964–972 (2007).

23. Gatti-Lafranconi, P. & Hollfelder, F. Flexibility and reactivity in promiscuous enzymes. Chembiochem 14, 285–92 (2013).

24. Sheldon, R. A. & Van Pelt, S. Enzyme immobilisation in biocatalysis: why, what and how. Chem. Soc. Rev. Chem. Soc. Rev 42, 6223–6235 (2013).

25. Kumar, S., Tsai, C.-J. & Nussinov, R. Factors enhancing protein thermostability. Protein Eng. 13, 179–191 (2000).

26. Dombkowski, A. A., Sultana, K. Z. & Craig, D. B. Protein disulfide engineering. FEBS Lett. 588, 206–212 (2014).

27. Turner, N. J. Directed evolution of enzymes for applied biocatalysis. Trends Biotechnol. 21, 474–478 (2003).

28. Turner, N. J. Directed evolution drives the next generation of biocatalysts. Nat. Chem. Biol. 5, 567–573 (2009).

29. Bornscheuer, U. T., Huisman, G. W., Kazlauskas, R. J., Lutz, S., Moore, J. C. & Robins, K. Engineering the third wave of biocatalysis. Nature 485, 185–94 (2012).

30. Porter, J. L., Rusli, R. A. & Ollis, D. L. Directed Evolution of Enzymes for Industrial Biocatalysis. ChemBioChem 2601, 197–203 (2015).

31. Jaeger, K. E. & Reetz, M. T. Directed evolution of enantioselective enzymes for organic chemistry. Curr. Opin. Chem. Biol. 4, 68–73 (2000).

32. Bloom, J. D. & Arnold, F. H. In the light of directed evolution: pathways of adaptive protein evolution. Proc. Natl. Acad. Sci. 106, 9995–10000 (2009).

33. Dougherty, M. J. & Arnold, F. H. Directed evolution: new parts and optimized function. Curr. Opin. Biotechnol. 20, 486–491 (2009).

34. Kazlauskas, R. J. & Bornscheuer, U. T. Finding better protein engineering strategies. Nat. Chem. Biol. 5, 526–9 (2009).

35. Reetz, M. T., Bocola, M., Carballeira, J. D., Zha, D. & Vogel, A. Expanding the range of substrate acceptance of enzymes: Combinatorial active-site saturation test. Angew. Chemie - Int. Ed. 44, 4192–4196 (2005).

36. Tokuriki, N. & Tawfik, D. S. Stability effects of mutations and protein evolvability. Curr. Opin. Struct. Biol. 19, 596–604 (2009).

37. Barrozo, A., Borstnar, R., Marloie, G. & Kamerlin, S. C. L. Computational Protein Engineering: Bridging the Gap between Rational Design and Laboratory Evolution. Int. J. Mol. Sci. 13, 12428–60 (2012).

38. Arnold, F. H. Combinatorial and computational challenges for biocatalyst design. Nature 409, 253–7 (2001).

39. Warshel, A. Computer simulations of enzyme catalysis: methods, progress, and insights. Annu. Rev. Biophys. Biomol. Struct. 32, 425–43 (2003).

40. Kries, H., Blomberg, R. & Hilvert, D. De novo enzymes by computational design. Curr. Opin. Chem. Biol. 17, 221–228 (2013).

41. Jiang, L., Althoff, E. A., Clemente, F. R., Doyle, L., Rothlisberger, D., Zanghellini, A., Gallaher, J. L., Betker, J. L, Tanaka, F., Barbas, C. F., Hilvert, D., Houk, K. N., Stoddard, B. L. & Baker, D. De Novo Computational Design of Retro-Aldol Enzymes. Science (80-. ). 319, 1387–1391 (2008).

84

42. Röthlisberger, D., Khersonsky, O., Wollacott, A. M., Jiang, L., Dechancie, J., Betker, J., Gallaher, J. L., Althott, E. A., Zanghellini, A., Dym, O., Albeck, S., Houk, K. N., Tawfik, D. S. & Baker, D. Kemp elimination catalysts by computational enzyme design. Nature 453, 190–5 (2008).

43. Turner, N. J. & Truppo, M. D. Biocatalysis enters a new era. Curr. Opin. Chem. Biol. 17, 212–214 (2013).

44. Currin, A., Swainston, N., Day, P. J. & Kell, D. B. Synthetic biology for the directed evolution of protein biocatalysts: navigating sequence space intelligently. Chem. Soc. Rev. 44, 1172–239 (2015).

45. Damborsky, J. & Brezovsky, J. Computational tools for designing and engineering enzymes. Curr. Opin. Chem. Biol. 19, 8–16 (2014).

46. Bloom, J. D., Labthavikul, S. T., Otey, C. R. & Arnold, F. H. Protein stability promotes evolvability. Proc. Natl. Acad. Sci. U. S. A. 103, 5869–5874 (2006).

47. Soskine, M. & Tawfik, D. S. Mutational effects and the evolution of new protein functions. Nat. Rev. Genet. 11, 572–82 (2010).

48. Gerlt, J. a & Babbitt, P. C. Divergent Evolution of Enzymatic Function: Mechanistically Diverse Superfamilies and Functionally Distinct Suprafamilies. Biochemistry 70, 209–246 (2001).

49. Tracewell, C. A. & Arnold, F. H. Directed enzyme evolution: climbing fitness peaks one amino acid at a time. Curr. Opin. Chem. Biol. 13, 3–9 (2009).

50. Wijma, H. J., Floor, R. J., Bjelic, S., Marrink, S. J., Baker, D. & Janssen, D. B. Enantioselective enzymes by computational design and in silico screening. Angew. Chemie - Int. Ed. 54, 3726–3730 (2015).

51. Cohen, N., Abramov, S., Dror, Y. & Freeman, A. In vitro enzyme evolution: the screening challenge of isolating the one in a million. Trends Biotechnol. 19, 507–10 (2001).

52. Zocher, F., Enzelberger, M. M., Bornscheuer, U. T., Hauer, B. & D. Schmid, R. A colorimetric assay suitable for screening epoxide hydrolase activity. Anal. Chim. Acta 391, 345–351 (1999).

53. Gordes, D. & Thurow, K. High-Throughput Screening Application For the Determination of Enantiomeric Excess Using ESI-MS. J. Assoc. Lab. Autom. 11, 128–133 (2006).

54. Reetz, M. T. et al. A GC-based method for high-throughput screening of enantioselective catalysts. Catal. Today 67, 389–396 (2001).

55. Hoebenreich, S., Zilly, F. E., Acevedo-Rocha, C. G., Zilly, M. & Reetz, M. T. Speeding up Directed Evolution: Combining the Advantages of Solid-Phase Combinatorial Gene Synthesis with Statistically Guided Reduction of Screening Effort. ACS Synth. Biol. (2014). doi:10.1021/sb5002399

56. March, J. Advanced Organic Chemistry: Reactions, Mechanism and Structure (John Wiley & Sons, New york, 1992).

57. Parker, R. E. & Isaacs, N. S. Mechanisms of epoxide reactions. Chem. Rev. 59, (1959).

58. Blumenstein, J. J., Ukachukwu, V. C., Mohan, R. S. & Whalen, D. L. Effects of para-substituents on the mechanisms of solvolysis of styrene oxides. J. Org. Chem. 58, 924 (1993).

59. Widersten, M., Gurell, A. & Lindberg, D. Structure-function relationships of epoxide hydrolases and their potential use in biocatalysis. Biochim. Biophys. Acta 1800, 316–26 (2010).

60. Archelas, A. & Furstoss, R. Synthesis of enantiopure epoxides through biocatalytic approaches. Annu. Rev. Microbiol. 51, 491–525 (1997).

61. Widersten, M. Protein engineering for development of new hydrolytic biocatalysts. Curr. Opin. Chem. Biol. 21, 42–47 (2014).

85

62. Archelas, A. & Furstoss, R. Epoxide hydrolases: new tools for the synthesis of fine organic chemicals. Trends Biotechnol. 16, 108–16 (1998).

63. Choi, W. J. & Choi, C. Y. Production of chiral epoxides: Epoxide hydrolase-catalyzed enantioselective hydrolysis. Biotechnol. Bioprocess Eng. 10, 167–179

64. Morisseau, C. & Hammock, B. D. Epoxide Hydrolases: Mechanisms, Inhibitor Designs, and Biological Roles. Annu. Rev. Pharmacol. Toxicol. 45, 311–333 (2005).

65. Newman, J. W., Morisseau, C. & Hammock, B. D. Epoxide hydrolases: their roles and interactions with lipid metabolism. Prog. Lipid Res. 44, 1–51 (2005).

66. Monterde, M. I. et al. Enzymatic transformations. Part 58: Enantioconvergent biohydrolysis of styrene oxide derivatives catalysed by the Solanum tuberosum epoxide hydrolase. Tetrahedron: Asymmetry 15, 2801–2805 (2004).

67. Schober, M. & Faber, K. Inverting hydrolases and their use in enantioconvergent biotransformations. Trends Biotechnol. 31, 468–78 (2013).

68. Cao, L., Lee, J., Chen, W. & Wood, T. K. Enantioconvergent Production of ( R ) -1-Phenyl-1 , 2-Ethanediol From Styrene Oxide by Combining the Solanum tuberosum and an Evolved Agrobacterium radiobacter AD1 Epoxide Hydrolases. (2006). doi:10.1002/bit

69. Simeó, Y. & Faber, K. Selectivity enhancement of enantio- and stereo-complementary epoxide hydrolases and chemo-enzymatic deracemization of (±)-2-methylglycidyl benzyl ether. Tetrahedron Asymmetry 17, 402–409 (2006).

70. Manoj, K. M., Archelas, A., Baratti, J. & Furstoss, R. Microbiological transformations. Part 45: A green chemistry preparative scale synthesis of enantiopure building blocks of Eliprodil: Elaboration of a high substrate concentration epoxide hydrolase-catalyzed hydrolytic kinetic resolution process. Tetrahedron 57, 695–701 (2001).

71. Ollis, D. L., Chea, E., Cyglerl, M., Djikstra, B., Frolow, F., Franken, S. M., Harel, M., Remington, S. J., Silman, I., Schragl, J., Sussman, J. L. & Goldmans, A. The α/β-hydrolase fold. 5, 197–211 (1992).

72. Barth, S., Fischer, M., Schmid, R. D. & Pleiss, J. Sequence and structure of epoxide hydrolases: A systematic analysis. Proteins Struct. Funct. Genet. 55, 846–855 (2004).

73. Nardini, M. & Djikstra B. α/β Hydrolase fold enzymes: the family keeps growing. Current Opinion in Structural Biology. 9, 732–737 (1999).

74. Elfström, L. T. & Widersten, M. Catalysis of potato epoxide hydrolase, StEH1. Biochem. J. 390, 633–40 (2005).

75. Thomaeus, A., Carlsson, J., Åqvist, J. & Widersten, M. Active site of epoxide hydrolases revisited: A noncanonical residue in potato StEH1 promotes both formation and breakdown of the alkylenzyme intermediate. Biochemistry 46, 2466–2479 (2007).

76. Armstrong, R. N. & Cassidy, C. S. New Structural and Chemical Insight Into the Catalytic Mechanism of Epoxide Hydrolases. Drug Metab. Rev. 32, 327–338 (2000).

77. Tzeng, H., Laughlin, L. T., Lin, S. & Armstrong, R. N. The Catalytic Mechanism of Microsomal Epoxide Hydrolase Involves Reversible Formation and Rate-Limiting Hydrolysis of the Alkyl - Enzyme Intermediate. 7863, 9436–9437 (1996).

78. Rink, R. & Janssen, D. B. Kinetic mechanism of the enantioselective conversion of styrene oxide by epoxide hydrolase from Agrobacterium radiobacter AD1. Biochemistry 37, 18119–18127 (1998).

86

79. van Loo, B., Kingma, J., Heyman, G., Wittenaar, A., Lutje Spelberg, J. H., Sonke, T. & Janssen, D. B. Improved enantioselective conversion of styrene epoxides and meso-epoxides through epoxide hydrolases with a mutated nucleophile-flanking residue. Enzyme Microb. Technol. 44, 145–153 (2009).

80. Reetz, M. T., Torre, C., Eipper, A., Lohmer, R., Hermes, M., Brunner, B., Maichele, A., Bocola, M., Arand, M., Cronin, A., Genzel, Y., Archelas, A. & Furstoss, R. Enhancing the enantioselectivity of an epoxide hydrolase by directed evolution. Org. Lett. 6, 177–80 (2004).

81. Rui, L., Cao, L., Chen, W., Reardon, K. F. & Wood, T. K. Protein Engineering of Epoxide Hydrolase from Agrobacterium radiobacter AD1 for Enhanced Activity and Enantioselective Production of (R)-1-Phenylethane-1,2-Diol. 71, 3995–4003 (2005).

82. Rui, L., Cao, L., Chen, W., Reardon, K. F. & Wood, T. K. Active site engineering of the epoxide hydrolase from Agrobacterium radiobacter AD1 to enhance aerobic mineralization of cis-1,2-dichloroethylene in cells expressing an evolved toluene ortho-monooxygenase. J. Biol. Chem. 279, 46810–46817 (2004).

83. Reetz, M. T., Bocola, M., Wang, L., Sanchis, J., Cronin, A., Arand, M., Zou, J., Archelas, A., Bottalla, A., Naworyta, A. & Mowbray, S. L. Directed evolution of an enantioselective epoxide hydrolase: uncovering the source of enantioselectivity at each evolutionary stage. J. Am. Chem. Soc. 131, 7334–43 (2009).

84. Sun, Z., Lonsdale, R., Kong, X. D., Xu, J. H., Zhou, J. & Reetz, M. T. Reshaping an Enzyme Binding Pocket for Enhanced and Inverted Stereoselectivity: Use of Smallest Amino Acid Alphabets in Directed Evolution. Angew. Chemie - Int. Ed. 54, 12410–12415 (2015).

85. Laughlin, L. T., Tzeng, H. F., Lin, S. & Armstrong, R. N. Mechanism of microsomal epoxide hydrolase. Semifunctional site-specific mutants affecting the alkylation half-reaction. Biochemistry 37, 2897–904 (1998).

86. Rink, R., Lutje Spelberg, J. H., Pieters, R. J., Kingma, J., Nardini, M., Kellogg, R. M., Djikstra, B. W. & Janssen, D. B. Mutation of tyrosine residues involved in the alkylation half reaction of epoxide hydrolase from Agrobacterium radiobacter AD1 results in improved enantioselectivity [7]. J. Am. Chem. Soc. 121, 7417–7418 (1999).

87. Genzel, Y., Archelas, A., Lutje Spelberg, J. H., Janssen, D. B. & Furstoss, R. Microbiological transformations. Part 48: Enantioselective biohydrolysis of 2-,3- and 4-pyridyloxirane at high substrate concentration using the Agrobacterium radiobacter AD1 epoxide hydrolase and its Tyr215Phe mutant. Tetrahedron 57, 2775–2779 (2001).

88. Lutje Spelberg, J. H., Rink, R., Archelas, A., Furstoss, R. & Janssen, D. B. Biocatalytic Potential of the Epoxide Hydrolase from Agrobacterium radiobacter AD1 and a Mutant with Enhanced Enantioselectivity. Adv. Synth. Catal. 344, 980–985 (2002).

89. Kourist, R., Jochens, H., Bartsch, S., Kuipers, R., Padhi, S. K., Gall, M., Böttcher, D., Joosten, H. & Bornscheuer, U. T. The alpha/beta-hydrolase fold 3DM database (ABHDB) as a tool for protein engineering. Chembiochem 11, 1635–43 (2010).

90. Lenfant, N., Velluet, E., Bourne, Y., Marchot, P. & Chatonnet, A. ESTHER, the database of the alpha/beta-hydrolase fold superfamily of proteins: Tools to explore diversity of functions. Nucleic Acids Res. 41, 1–7 (2013).

87

91. Arand, M., Hallberg, B. M., Zou, J., Bergfors, T., Oesch, F., van der Werf, M. J., de Bont, J. A. M., Jones, T. A. & Mowbray, S. L. Structure of Rhodococcus erythropolis limonene-1,2-epoxide hydrolase reveals a novel active site. EMBO J. 22, 2583–2592 (2003).

92. Nardini, M., Ridder, I. S., Rozeboom, H. J., Kalk, K. H., Rink, R., Janssen, D. B. & Djikstra, B. W. The X-ray Structure of Epoxide Hydrolase from Agrobacterium radiobacter AD1: An Enzyme to Detoxify harmful epoxides. J. Biol. Chem. 274, 14579–14586 (1999).

93. Lonsdale, R., Hoyle, S., Grey, D. T., Ridder, L. & Mulholland, A. J. Determinants of reactivity and selectivity in soluble epoxide hydrolase from quantum mechanics/molecular mechanics modeling. Biochemistry 51, 1774–86 (2012).

94. Hou, Q. Q., Sheng, X., Wang, J. H., Liu, Y. J. & Liu, C. B. QM/MM study of the mechanism of enzymatic limonene 1,2-epoxide hydrolysis. Biochim. Biophys. Acta - Proteins Proteomics 1824, 263–268 (2012).

95. Lind, M. E. S. & Himo, F. Quantum chemistry as a tool in asymmetric biocatalysis: Limonene epoxide hydrolase test case. Angew. Chemie - Int. Ed. 52, 4563–4567 (2013).

96. Steinreiber, A. & Faber, K. Microbial epoxide hydrolases for preparative biotransformations. Curr. Opin. Biotechnol. 12, 552–558 (2001).

97. Lee, E. Y. & Shuler, M. L. Molecular Engineering of Epoxide Hydrolase and Its Application to Asymmetric and Enantioconvergent Hydrolysis. 98, 318–327 (2007).

98. de Vries, E. J. & Janssen, D. B. Biocatalytic conversion of epoxides. Curr. Opin. Biotechnol. 14, 414–420 (2003).

99. Breuer, M., Ditrich, K., Habicher, T., Hauer, B., Kesseler, M., Stürmer, R. & Zelinski, T. Industrial methods for the production of optically active intermediates. Angew. Chem. Int. Ed. Engl. 43, 788–824 (2004).

100. Orru, R. V & Faber, K. Stereoselectivities of microbial epoxide hydrolases. Curr. Opin. Chem. Biol. 3, 16–21 (1999).

101. Lee, E. Y. Epoxide hydrolase-mediated enantioconvergent bioconversions to prepare chiral epoxides and alcohols. Biotechnol. Lett. 30, 1509–14 (2008).

102. Kotik, M., Archelas, A. & Wohlgemuth, R. Epoxide hydrolases and their application in organic synthesis. Curr. Org. Chem. 16, 451–482 (2012).

103. Cleij, M., Archelas, A. & Furstoss, R. Microbiological transformations 43. Epoxide hydrolases as tools for the synthesis of enantiopure α-methylstyrene oxides: A new and efficient synthesis of (S)-ibuprofen. J. Org. Chem. 64, 5029–5035 (1999).

104. Orru, R. V. A., Osprian, I., Kroutil, W. & Faber, K. An efficient large-scale synthesis of (R)-(–)-Mevalonolactone using simple biological and chemical catalysts. Synthesis (Stuttg). 9, 1259–1263 (1998).

105. Hwang, S., Choi, C. Y. & Lee, E. Y. One-pot biotransformation of racemic styrene oxide into (R)-1,2-phenylethandiol by two recombinant microbial epoxide hydrolases. Biotechnol. Bioprocess Eng. 13, 453–457 (2008).

106. Kim, H. S., Lee, O. K., Hwang, S., Kim, B. J. & Lee, E. Y. Biosynthesis of (R)-phenyl-1,2-ethanediol from racemic styrene oxide by using bacterial and marine fish epoxide hydrolases. Biotechnol. Lett. 30, 127–133 (2008).

107. Kong, X. D., Ma, Q., Zhou, J., Zeng, B. B. & Xu, J. H. A smart library of epoxide hydrolase variants and the top hits for synthesis of (S)-β-blocker precursors. Angew. Chemie - Int. Ed. 53, 6641–6644 (2014).

108. Archelas, A. & Furstoss, R. Synthetic applications of epoxide hydrolases. Curr. Opin. Chem. Biol. 5, 112–9 (2001).

88

109. Botes, B. A. L. Epoxide Hydrolases : A Biocatalytic Technology Platform for the Production of Chiral Pharmaceutical Intermediates. 86–89 (2006).

110. Choi, W. J., Choi, C. Y., De Bont, J. a & Weijers, C. a. Continuous production of enantiopure 1,2-epoxyhexane by yeast epoxide hydrolase in a two-phase membrane bioreactor. Appl. Microbiol. Biotechnol. 54, 641–646 (2000).

111. Baldascini, H. & Janssen, D. B. Interfacial inactivation of epoxide hydrolase in a two-liquid-phase system. Enzyme Microb. Technol. 36, 285–293 (2005).

112. Petri, A., Marconcini, P. & Salvadori, P. Efficient immobilization of epoxide hydrolase onto silica gel and use in the enantioselective hydrolysis of racemic para-nitrostyrene oxide. J. Mol. Catal. B Enzym. 32, 219–224 (2005).

113. Karboune, S., Archelas, A., Furstoss, R. & Baratti, J. Immobilization of epoxide hydrolase from Aspergillus niger onto DEAE-cellulose: Enzymatic properties and application for the enantioselective resolution of a racemic epoxide. J. Mol. Catal. B Enzym. 32, 175–183 (2005).

114. Mateo, C., Archelas, A., Fernandez-Lafuente, R., Guisan, J. M. & Furstoss, R. Enzymatic transformations. Immobilized A. niger epoxide hydrolase as a novel biocatalytic tool for repeated-batch hydrolytic kinetic resolution of epoxides. Org. Biomol. Chem. 1, 2739–2743 (2003).

115. Cassimjee, K. E., Kourist, R., Lindberg, D., Wittrup Larsen, M., Thanh, N. H., Widersten, M., Bornscheuer, U. T. & Berglund, P. One-step enzyme extraction and immobilization for biocatalysis applications. Biotechnol. J. 6, 463–9 (2011).

116. Chiappe, C., Leandri, E., Lucchesi, S., Pieraccini, D., Hammock, B. D. & Morisseau, C. Biocatalysis in ionic liquids: the stereoconvergent hydrolysis of trans-β-methylstyrene oxide catalyzed by soluble epoxide hydrolase. J. Mol. Catal. B Enzym. 27, 243–248 (2004).

117. Chiappe, C., Leandri, E., Hammock, B. D. & Morisseau, C. Effect of ionic liquids on epoxide hydrolase-catalyzed synthesis of chiral 1,2-diols. Green Chem. 2007, 162–168 (2007).

118. Lindberg, D., de la Fuente Revenga, M. & Widersten, M. Deep eutectic solvents (DESs) are viable cosolvents for enzyme-catalyzed epoxide hydrolysis. J. Biotechnol. 147, 169–71 (2010).

119. Chen, W.-J., Lou, W.-Y. & Zong, M.-H. Efficient asymmetric hydrolysis of styrene oxide catalyzed by Mung bean epoxide hydrolases in ionic liquid-based biphasic systems. Bioresour. Technol. 115, 58–62 (2012).

120. Yu, C.-Y., Wei, P., Li, X.-F., Zong, M.-H. & Lou, W.-Y. Using Ionic Liquid in a Biphasic System to Improve Asymmetric Hydrolysis of Styrene Oxide Catalyzed by Cross-Linked Enzyme Aggregates (CLEAs) of Mung Bean Epoxide Hydrolases. Ind. Eng. Chem. Res. 53, 7923–7930 (2014).

121. Morisseau, C., Beetham, J. K., Pinot, F., Debernard, S., Newman, J. W. & Hammock, B. D. Cress and potato soluble epoxide hydrolases: purification, biochemical characterization, and comparison to mammalian enzymes. Arch. Biochem. Biophys. 378, 321–332 (2000).

122. Stapleton, A., Beetham, J. K., Pinot, F., Garbarino, J. E., Rockhold, D. R., Friedman, M., Hammock, B. D. & Belknap, W. R. Cloning and expression of soluble epoxide hydrolase from potato. Plant J. 6, 251–258 (1994).

123. Ahlgren, K. M., Mowbray, S. L. & Elfström, L. T. X-ray structure of potato epoxide hydrolase sheds light on substrate specificity in plant enzymes. 3, 1628–1637 (2006).

89

124. Lindberg, D., Ahmad, S. & Widersten, M. Mutations in salt-bridging residues at the interface of the core and lid domains of epoxide hydrolase StEH1 affect regioselectivity, protein stability and hysteresis. Arch. Biochem. Biophys. 495, 165–73 (2010).

125. Thomaeus, A., Naworyta, A. & Mowbray, S. L. Removal of distal protein – water hydrogen bonds in a plant epoxide hydrolase increases catalytic turnover but decreases thermostability. 1275–1284 (2008). doi:10.1110/ps.034173.107.corresponding

126. Gurell, A. & Widersten, M. Modification of substrate specificity resulting in an epoxide hydrolase with shifted enantiopreference for (2,3-Epoxypropyl)benzene. ChemBioChem 11, 1422–1429 (2010).

127. Lindberg, D., De La Revenga, M. F. & Widersten, M. Temperature and pH dependence of Enzyme-catalyzed hydrolysis of trans-methylstyrene oxide. A unifying kinetic model for observed hysteresis, Cooperativity, and regioselectivity. Biochemistry 49, 2297–2304 (2010).

128. Lindberg, D., Gogoll, A. & Widersten, M. Substrate-dependent hysteretic behavior in StEH1-catalyzed hydrolysis of styrene oxide derivatives. FEBS J. 275, 6309–6320 (2008).

129. Reetz, M. T. & Carballeira, J. D. Iterative saturation mutagenesis (ISM) for rapid directed evolution of functional enzymes. Nat. Protoc. 2, 891–903 (2007).

130. Zaugg, J., Gumulya, Y., Gillam, E. M. J. & Bodén, M. Directed Evolution Library Creation. Methods Mol. Biol. 1179, 315–33 (2014).

131. Reetz, M. T., Kahakeaw, D. & Lohmer, R. Addressing the numbers problem in directed evolution. Chembiochem 9, 1797–804 (2008).

132. Kille, S., Acevedo-Rocha, C. G., Parra, L. P., Zhang, Z., Opperman, D. J., Reetz, M. T. & Acevedo, J. P. Reducing codon redundancy and screening effort of combinatorial protein libraries created by saturation mutagenesis. ACS Synth. Biol. 2, 83–92 (2013).

133. Tang, L., Gao, H., Zhu, X., Wang, X., Zhou, M. & Jiang, R. Construction of ‘small-intelligent’ focused mutagenesis libraries using well-designed combinatorial degenerate primers. Biotechniques 52, 149–158 (2012).

134. Cedrone, F., Bhatnagar, T. & Baratti, J. C. Colorimetric assays for quantitative analysis and screening of epoxide hydrolase activity. Biotechnol. Lett. 27, 1921–7 (2005).

135. Agarwal, S. C., Van Duuren, B. L. & Kneip, T. J. Detection of epoxides with 4-(p-nitrobenzyl) pyridine. Bull. Environ. Contam. Toxicol. 23, 825–9 (1979).

136. Straathof, A. J. J. & Jongejan, J. A. The enantiomeric ratio: Origin, determination and prediction. Enzyme Microb. Technol. 21, 559–571 (1997).

137. Milne, J. L. S. & Subramaniam, S. 2.2 Å resolution cryo-EM structure of β-galactosidase in complex with a cell-permeant inhibitor. Science 348, 1147–1152 (2015).

138. Warshel, A. & Weiss, R. M. An empirical valence bond approach for comparing reactions in solutions and in enzymes. J. Am. Chem. Soc. 102, 6218–6226 (1980).

139. Warshel, A. & Weiss, R. M. Empirical valence bond calculations of enzyme catalysis. Ann. N. Y. Acad. Sci. 367, 370–382 (1981).

140. Kamerlin, S. C. L. & Warshel, A. The empirical valence bond model: Theory and applications. Wiley Interdiscip. Rev. Comput. Mol. Sci. 1, 30–45 (2011).

141. Reetz, M. T., Kahakeaw, D. & Lohmer, R. Addressing the numbers problem in directed evolution. ChemBioChem 9, 1797–1804 (2008).

90

142. van Loo, B., Kingma, J., Heyman, G., Wittenaar, A., Lutje, J. H., Sonke, T & Janssen, D. B. Enzyme and Microbial Technology Improved enantioselective conversion of styrene epoxides and meso-epoxides through epoxide hydrolases with a mutated nucleophile-flanking residue. 44, 145–153 (2009).

143. Lau, E. Y., Newby, Z. E. & Bruice, T. C. A theoretical examination of the acid-catalyzed and noncatalyzed ring-opening reaction of an oxirane by nucleophilic addition of acetate. Implications to epoxide hydrolases. J. Am. Chem. Soc. 123, 3350–3357 (2001).

144. Schiøtt, B. & Bruice, T. C. Reaction mechanism of soluble epoxide hydrolase: insights from molecular dynamics simulations. J. Am. Chem. Soc. 124, 14558–14570 (2002).

145. Hamnevik, E., Blikstad, C., Norrehed, S. & Widersten, M. Kinetic characterization of Rhodococcus ruber DSM 44541 alcohol dehydrogenase A. J. Mol. Catal. B Enzym. 99, 68–78 (2014).

146. Liese, a, Karutz, M., Kamphuis, J., Wandrey, C. & Kragl, U. Enzymatic resolution of 1-phenyl-1,2-ethanediol by enantioselective oxidation: overcoming product inhibition by continuous extraction. Biotechnol. Bioeng. 51, 544–50 (1996).

Acta Universitatis UpsaliensisDigital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 1378

Editor: The Dean of the Faculty of Science and Technology

A doctoral dissertation from the Faculty of Science andTechnology, Uppsala University, is usually a summary of anumber of papers. A few copies of the complete dissertationare kept at major Swedish research libraries, while thesummary alone is distributed internationally throughthe series Digital Comprehensive Summaries of UppsalaDissertations from the Faculty of Science and Technology.(Prior to January, 2005, the series was published under thetitle “Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology”.)

Distribution: publications.uu.seurn:nbn:se:uu:diva-286557

ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2016