Identification, characterization and application

of novel (R)-selective amine transaminases

Inauguraldissertation

zur

Erlangung des akademischen Grades

doctor rerum naturalium (Dr. rer. nat.)

an der Mathematisch-Naturwissenschaftlichen Fakultät

der

Ernst-Moritz-Arndt-Universität Greifswald

vorgelegt von

Sebastian Schätzle

geboren am 12.1.1983

in Duisburg

Greifswald, November 2011

Dekan: Prof. Dr. Klaus Fesser

1. Gutachter: Prof. Dr. Uwe T. Bornscheuer

2. Gutachter: Prof. Dr. Per Berglund

Tag der Promotion: 11.01.2012

The Contents

I

TABLE OF CONTENTS

TABLE OF CONTENTS I

LIST OF ABBREVIATIONS AND SYMBOLS II

SCOPE AND OUTLINE OF THIS THESIS III

1 THE BACKGROUND - 1 -

1.1 AMINES IN NATURE AND PHARMACEUTICALS - 2 - 1.2 CHEMICAL SYNTHESIS OF CHIRAL AMINES - 3 - 1.3 ENZYMATIC SYNTHESIS OF CHIRAL AMINES - 4 - 1.4 TRANSAMINASES – ENZYMES OF NOBLE FAMILY - 5 - 1.5 ENZYME DISCOVERY – MANY WAYS LEAD TO ROME ARTICLE I - 7 -

2 THE ANALYTICS - 9 -

2.1 A FAST AND EASY ASSAY FOR SCREENING ARTICLE II - 11 - 2.2 A SLIGHTLY AMBITIOUS ASSAY FOR CHARACTERIZATION ARTICLE III - 12 -

3 EXPLORING NEW PATHS IN ENZYME DISCOVERY ARTICLE IV - 14 -

4 APPLICATION OF THE NEW BIOCATALYSTS ARTICLE V - 19 -

5 CONCLUSION - 23 -

6 REFERENCES - 24 -

ARTICLES - 28 -

AFFIRMATION - 83 -

CURRICULUM VITAE - 84 -

ACKNOWLEDGEMENTS - 85 -

The Abbreviations

II

LIST OF ABBREVIATIONS AND SYMBOLS

ATA amine transaminase

AspTer R-ATA from Aspergillus terreus

AspFum R-ATA from Aspergillus fumigatus

AspOry R-ATA from Aspergillus oryzea

-MBA -methylbenzylamine

-TA -amino acid transaminase

BCAT branched chain amino acid

aminotransferase

BLAST basic local alignment search tool

c conversion

CE capillary electrophoreses

cm centimeter

°C degree celsius

DATA D-amino acid aminotransferase

DMSO dimethylsulfoxide

DNA desoxyribonucleic acid

E. coli Escherichia coli

ee enantiomeric excess

e.g. for example

g gram

GC gas chromatography

GDH glucose dehydrogenase

GibZea R-ATA from Gibberella zeae

h hour

HPLC high performance liquid

chromatography

i.e. that is

l liter

LB lysogeny broth

LDH lactate dehydrogenase

wavelength

M mol per liter

mg milligram

min minute

ml milliliter

mM millimol per liter

MTP microtiter plate

S microsiemens

MycVan R-ATA from Mycobacterium

vanbaalenii

NAD(P)+ nicotinamide adenine dinucleotide

(phosphate), oxidized form

NAD(P)H nicotinamide adenine dinucleotide

(phosphate), reduced form

NeoFis R-ATA from Neosartorya fischeri

OD optical density

PAGE polyacrylamide gel electrophoresis

PCR polymerase chain reaction

PDB Brookhaven protein database

PDC pyruvate decarboxylase

PenChr R-ATA from Penicillium chrysogenum

pH pondus hydrogenii

pI isoelectric point

RhoSph S-ATA from Rhodobacter sphaeroides

sp. species

t time

TA transaminase

U unit

VibFlu S-ATA from Vibrio fluvialis

Furthermore, the usual codes for amino acids

were used.

The Outline

III

SCOPE AND OUTLINE OF THIS THESIS

This thesis is about the identification of novel (R)-selective amine transaminases and their application in

asymmetric synthesis of various chiral amines. Before a novel in silico search strategy (article IV) was

developed for the identification of these enzymes out of sequence databases, two specific assay systems

(article II & III) had been established, which allowed a fast activity screening and characterization of their

catalytic properties. Seven of 21 initially identified proteins were applied later on in the asymmetric

synthesis of enantiopure amines (article V). The review article I summarizes state-of-the-art methods for

enzyme discovery and protein engineering, illustrating this thesis in a comprehensive context.

Article I Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

G. A. Behrens*, A. Hummel*, S. K. Padhi*, S. Schätzle*, U. T. Bornscheuer*, Adv. Synth.

Catal. 2011, 353, 2191-2215

Modern tools for enzyme discovery and protein engineering substantially broadened the number of

applicable enzymes for biocatalysis and enable to alter their properties such as substrate scope and

enantioselectivity. Article I provides a summary of different concepts and technologies, which are

exemplified for various enzymes.

Article II Rapid and Sensitive Kinetic Assay for Characterization of ω-Transaminases

S. Schätzle, M. Höhne, E. Redestad, K. Robins, U. T. Bornscheuer, Anal. Chem. 2009, 81,

8244-8248

Article II describes the development of a fast kinetic assay for measuring the activity of amine

transaminases (ATAs) based on the conversion of the widely used model substrate (R)-

methylbenzylamine. The product of this reaction, acetophenone, can be detected

spectrophotometrically at 245 nm with high sensitivity. As all low-absorbing ketones, aldehydes or keto

acids can be used as cosubstrates, the amino acceptor specificity of a given ATA can be obtained quickly.

Furthermore, the assay allows the fast investigation of enzymatic properties like the optimum pH and

temperature as well as the stability of the catalyst.

Article III Conductometric Method for the Rapid Characterization of the Substrate Specificity of

Amine-Transaminases

S. Schätzle, M. Höhne, K. Robins, U. T. Bornscheuer, Anal. Chem. 2010, 82, 2082-2086

While article II describes a method for characterizing the specificity for different amino acceptors, in

article III a kinetic conductivity assay for investigation of the the amino donor specificity of a given ATA

was established. The course of an ATA-catalyzed reaction can be followed conductometrically – and

quantitatively evaluated – since the conducting substrates, a positively charged amine and a negatively

charged keto acid, are converted to non-conducting products, a non-charged ketone and a zwitterionic

amino acid.

The Outline

IV

Articel IV Rational Assignment of Key Motifs for Function guides in silico Enzyme Identification

M. Höhne*, S. Schätzle*, H. Jochens, K. Robins, U. T. Bornscheuer, Nature Chem. Biol.

2010, 6, 807–813

Biocatalysts are powerful tools for the synthesis of chiral compounds, but identifying novel enzymes that

are able to catalyze new reactions is challenging. Article IV presents a new strategy to find existing (R)-

selective amine transaminases (R-ATAs) based on rationally developed structural motifs, which were

used to search sequence space for suitable candidates, yielding 17 active enzymes with 100% correct

prediction of the enantiopreference.

Article V Enzymatic Asymmetric Synthesis of Enantiomerically Pure Aliphatic, Aromatic and

Arylaliphatic Amines with (R)-Selective Amine Transaminases

S. Schätzle, F. Steffen-Munsberg, A. Thontowi, M. Höhne, K. Robins, U. T. Bornscheuer,

Adv. Synth. Catal. 2011, 353, 2439-2445

In article V seven of the newly identified R-ATAs were applied in the asymmetric synthesis of twelve

aliphatic, aromatic and arylaliphatic (R)-amines starting from the corresponding ketones using a lactate

dehydrogenase/glucose dehydrogenase system for the necessary shift of the thermodynamic

equilibrium. For all ketones, at least one enzyme was found allowing complete conversion to the

corresponding chiral amine with excellent optical purities >99% ee. Variations in substrate profiles are

discussed based on the phylogenetic relationships between the seven R-ATAs.

* equal contribution

The Background

- 1 -

1 THE BACKGROUND

New biocatalytical processes can be based on the availability of new interesting enzymes, but more

often a desired product raises the demand for a suitable biocatalyst. Sometimes such an enzyme is

already commercially available or has been described in the literature. Alternatively, it will be necessary

to screen organisms or enzymes that allow conversion of available reactants.1 With promising candidates

at hand, a detailed biochemical characterization will help finding appropriate conditions for a successful

application as biocatalysts under the requirements of a specific process. If this characterization should

not yield satisfying results, several rounds of enzyme engineering might help creating and finding more

suitable enzyme variants.

Figure 1| The biocatalysis cycle1.

In our case, the aforementioned desired products were chiral amines with (R)-configuration and the

demand for novel biocatalysts initiated this project. The following chapters will give a brief introduction

to the importance of chiral amines as natural products and building blocks for pharmaceuticals, possible

methods for their synthesis – either chemical or enzymatic – and will introduce some general knowledge

about the investigated class of enzymes – amine transaminases. Furthermore, some-state-of-the-art

strategies for finding novel enzymes with desired properties will be summarized, placing this project in a

comprehensive context.

The Background

- 2 -

1.1 AMINES IN NATURE AND PHARMACEUTICALS

Amines and amino acids are ubiquitous in nature. They are not only essential parts of all proteins and

nucleic acids, but are also very important biologically active compounds themselves. Primary amines are

biogenetically produced by decarboxylation of the corresponding amino acids and they can be further N-

alkylated to yield secondary, tertiary and quaternary amines, which often results in heterocyclic

structures. Primary amines have great importance as neurotransmitter (e.g. adrenaline and histamine)

and as precursor of coenzymes (e.g. cysteamine of coenzyme A) and of complex lipids (e.g. ethanolamine

of phosphatidylethanolamine)2. Especially the higher substituted amines – pharmaceutically classified as

alkaloids – show an enormous variety of structures as well as biological effects found in all domains of

life (Figure 2).

Figure 2| Examples for amines in biologically active compounds. The displayed compounds are from human, bacterial or herbal origin. Coniin is the neurotoxic main alkaloid of Conium maculatum, Poison Hemlock, which was used to poison condemned prisoners in ancient Greece.

This huge variety of biological effects – from antibiotic to antiarrhytmic, analgesic and neurotoxic –

shows their potential as pharmaceuticals and thus makes them very promising candidates in the search

for new drugs. In fact, almost 80% of the 200 most prescribed brand name drugs in 2010 in the US

contain nitrogen, with the amino group [being] adjacent to a chiral carbon atom in more than 40% of

these compounds3. The absolute configuration of the stereocenters is crucial for the interaction with

biomolecules and thus for the type of effect on biological systems4. During the synthesis of the complex

target compounds (see Figure 3 for examples), the generation of chirality is most often the actual

challenge. Thus, optically pure amines gained an increased significance as building blocks in organic

chemistry5.

Figure 3| Examples for chiral amines in pharmaceuticals. The displayed compounds rank among the 200 most prescribed brand name drugs in 2010 in the US

3.

The Background

- 3 -

1.2 CHEMICAL SYNTHESIS OF CHIRAL AMINES

Most of the drugs mentioned above – and their precursors – are still produced chemically, and there are

a lot of different possibilities for the chemical synthesis of enantiopure amines. As this thesis is about

enzyme catalysis, the chemical methods are only introduced briefly and not discussed in detail. For

further aspects, the reviews written by Breuer et. al5 and Nugent et. al6 are highly recommended.

Traditionally, the chiral resolution of a racemic mixture of a given amine can be achieved by precipitating

one enantiomer as diastereomeric salt by addition of a chiral acid like (R)-mandelic acid or (R,R)-tartaric

acid5. Unfortunately, the theoretical yield of the desired enantiomer with this method is limited to 50%

(Figure 4). The better alternative with a theoretical yield of 100%, is to start from a prostereogenic

compound in an asymmetric synthesis to yield a chiral, enantiopure product.

Figure 4| General chemical routes to chiral amines. Besides the chiral resolution of a racemic amine with chiral

acids, asymmetric hydrogenation and addition are the two main strategies for the synthesis of -chiral primary amines. If chiral auxiliaries are used to generate enantioselectivity, R3 represents a chiral substituent.

The general approaches comprise asymmetric hydrogenation of enamides or ketimines and asymmetric

addition to aldimines or ketimines. The latter two are produced at first by a condensation reaction of a

ketone/aldehyde and an amine, whereas enamides can also be obtained from ketones by conversion to

the oxime and subsequent reduction, for instance with iron in the presence of acetic anhydride5. An

organocatalytic alternative is the proline-catalyzed direct asymmetric -amination of aldehydes with

azodicarboxylates7. It is worth mentioning that cyclic secondary amines can only be synthesized – in one

step – by asymmetric hydrogenation8. Furthermore, -tertiary amines with three different substituents

The Background

- 4 -

(unlike hydrogen) can only be produced by asymmetric addition to a ketimine9 as the chiral center is

generated by C-C coupling rather than by reduction, which always yields a C-H bond.

However, more and more companies replace their chemical routes by fermentation or biocatalysis – for

obvious reasons. Using biocatalysis, one is able to avoid toxic compounds like transition state catalysts,

save a lot of energy as the processes run at low temperature and pressure – in contrast to most chemical

processes, which run at rather high temperatures and often high pressure – and reduce waste, especially

during downstream processing and product isolation10. The companies Merck & Co. and Codexis, for

example, recently published an impressive case study on process chemistry, replacing the rhodium-

catalyzed Sitagliptin manufacture by a biocatalytic process using a highly-evolved amine transaminase11.

This process not only reduced the total waste (19%) and eliminated all need for heavy metals, but

even increased the overall yield by 13% and the productivity (kg/L per day) by 53% compared to the

chemical process. Both routes have recently been compared in detail12.

1.3 ENZYMATIC SYNTHESIS OF CHIRAL AMINES

For the biocatalytic production of chiral amines, there are in principle three options: the use of

hydrolases, oxidoreductases or transferases13.

Figure 5| Enzymatic routes to chiral amines. The displayed examples show (a) the kinetic resolution and (b) dynamic kinetic resolution with hydrolases, (c) deracemization with monoamine oxidases (MAO) and (d) asymmetric synthesis with amine dehydrogenases (amine-DH).

Proteases, amidases and lipases from the class of hydrolases are limited to (dynamic) kinetic resolutions

of a given racemic amine (Figure 5a and b) and monoamine oxidases (oxidoreductases) allow either a

The Background

- 5 -

kinetic resolution or a deracemization. The latter is an efficient one-pot reaction, wherein the produced

imine has to be reduced chemically to the racemic amine in order to accumulate the non-converted

enantiomer14 (Figure 5c). To complete the picture, there is another promising group of enzymes –

NAD(P)+ dependent amine dehydrogenases – which might be used for asymmetric syntheses in the

future (Figure 5d). This is because the only NAD(P)+ dependent amine dehydrogenase described so far

has indeed a broad substrate scope but is not yet applicable due to its low enantioselectivity15. Other

known amine dehydrogenases use different artificial electron acceptors such as phenazine

methosulfate16 or tryptophan tryptophylquinone17 and thus are not of biotechnological interest. Until

today, only amine transaminases (ATAs) can be used for kinetic resolutions, one pot-two step-

deracemizations and asymmetric syntheses.

Figure 6| Production of chiral amines using amine transaminases. In a kinetic resolution (a), the amine transaminase (ATA) converts in the ideal case only one of the amine enantiomers to the corresponding ketone. The remaining enantiomer can be isolated in high optical purity and a maximum yield of 50%. In an asymmetric synthesis (b), a prostereogenic ketone is aminated enantioselectively yielding directly the chiral amine. The most common co-substrates for ATAs are pyruvate/alanine. Since the equilibrium favors ketone formation, high yields in asymmetric synthesis can only be achieved by shifting the equilibrium, for example by enzymatic removal of the formed co-product pyruvate.

The kinetic resolution of the racemic amine with an (S)-selective transaminase produces the (R)-

enantiomer. The disadvantage of this approach is a theoretical maximum yield of only 50 % (Figure 6a).

The atom efficiency of such a process is low, as the ketone generated has little value and the recycling of

this compound requires chemical reductive amination to produce the racemic amine for subsequent

resolutions. Synthetically, an asymmetric synthesis is much more economical because yields of up to

100% are possible – in the case of an (S)-selective ATA the (S)-amine would be produced (Figure 6b). This

fact will be addressed in chapter 3 in more detail, but first, the protagonists of this project will be

introduced.

1.4 TRANSAMINASES – ENZYMES OF NOBLE FAMILY

Transaminases belong to the group of pyridoxal-5’-phosphate (PLP) dependent enzymes, whose

members most likely catalyze the largest scope of chemical reactions and belong to transferases, lyases,

isomerases and even hydrolases and oxidoreductases18. The different types of reactions can be

organized according to the position at which the reaction occurs, namely (transamination,

The Background

- 6 -

racemization, decarboxylation, elimination and substitution), and (both elimination and substitution).

However, initial attempts to organize the enzymes into an -, - and -class failed, so they were

classified with respect to their protein fold. Today, there are seven fold types of PLP-dependent enzymes

known and transaminases belong to fold class I (aspartate aminotransferase family) and IV (D-amino acid

aminotransferase family)19, 20. The transaminases of both fold types are homodimeric enzymes with two

catalytic sites at the interface of the subunits.

The catalytic cycle of the transamination reaction consists of two half reactions and follows a typical ping

pong bi bi kinetic. During the first half reaction the amino group of the amino donor is transferred to the

PLP to form pyridoxaminphosphate (PMP) and the keto product, which is released. The amino group is

then, in the second half reaction, transferred from the PMP to the amino acceptor, whereby the cofactor

is recycled and the product is released (Figure 7). The cofactor stabilizes the intermediate formed after

deprotonation of the external aldimine by delocalization of the negative charge through the system,

and thus initially enables the deprotonation. As mentioned above, PLP-dependent enzymes catalyze a

huge variety of reactions and so does the cofactor itself in the absence of enzyme, even though very

slowly21. The function of the protein apoenzyme, therefore, is to enhance this innate catalytic potential

and to enforce mechanistic and substrate selectivity22. The latter will be discussed in more detail in

chapter 3.

Figure 7| Simplified reaction scheme of the transamination. Illustrated is the first half reaction of the catalytic cycle. Sequentially, the PMP attacks the amino acceptor and finally the amine product is released and the PLP recycled

22.

Because of their crucial role in the amino acid metabolism,-amino acid aminotransferases (EC 2.6.1, -

transaminases) are ubiquitous enzymes found in all organisms. On the contrary, the small group of

amine-pyruvate transaminases (EC 2.6.1.18, amine transaminases) also converts substrates lacking an -

carboxylic acid moiety. The actual function of amine transaminases (ATAs) in nature so far remains

unknown; they probably take part in degrading biogenic amines like histamine or putrescin and enable

recycling of the nitrogen, in contrast to monoamine oxidases that release the nitrogen as ammonia.

Nevertheless, as the product range is not limited to -amino acids, ATAs are very attractive for organic

synthesis of optically active amines13, especially due to the possibility of asymmetric syntheses. Although

The Background

- 7 -

an asymmetric synthesis strategy is clearly favored over a kinetic resolution with respect to the atom

efficiency of the process, it has the major disadvantage that only one specific enantiomer can be

accessed with an enzyme having a distinct enantiopreference. As both enantiomers of a building block

are often needed for targets with different absolute configuration, an enzyme platform providing either

(S)- or (R)-specific enzymes is highly desired. Unfortunately, the majority of the more than a dozen

described amine transaminases in literature are (S)-selective. Only one commercially available enzyme

and two microorganisms were reported to show (R)-specificity23, but details about the proteins were

unknown. Consequently, many research groups – both from university and industry – have been looking

eagerly for new (R)-selective amine transaminases.

1.5 ENZYME DISCOVERY – MANY WAYS LEAD TO ROME ARTICLE I

The described lack of amine transaminases with the desired enantiopreference is a good example

showing that despite the enormous number of already described enzymes, there is a constant need for

novel biocatalysts with altered properties. Besides enantiopreference or –selectivity and substrate

scope, there are other properties, which often need to be altered regarding process conditions like

thermostability, pH profile and solvent tolerance.

Figure 8| Strategies for protein discovery and engineering. In addition to the classical way of screening strain collections or other environmental samples, structural information of known enzymes may be used to alter the properties by means of protein engineering. Hot spot positions for semi-random mutagenesis may be predicted or mutations suggested and modelled in silico. Instead of introducing mutations identified by rational design experimentally and further optimizing the enzyme by directed evolution, enzymes carrying these mutations may be identified in a database search. The bars represent the estimated amount of required information (green) and the necessary screening effort (blue) for the particular strategy.

The Background

- 8 -

To meet these demands, there are basically three strategies: (i) modification of already characterized

enzymes, (ii) search for new enzymes, or (iii) de novo design of completely new biocatalysts as a

potential, but not yet established route (Figure 8). For the modification of known enzymes there are two

main strategies in order to adapt their properties to the desired application: rational design and directed

(molecular) evolution24. Rational design uses structural and mechanistic information and/or molecular

modeling for the prediction of changes in the protein structure in order to alter or induce the desired

properties. In directed evolution on the other hand, mutant libraries are created by random changes and

screened/selected for the desired property. The variants showing promising results are subjected to

further rounds of evolution. As outlined in Figure 8, the choice of the proper method to solve a given

problem is limited on one end by the amount of information available and on the other end by the

disposable analytical methods. Without any structural information and deep comprehension of the

catalytic mechanism, rational design is a rather risky approach – especially if applied to alter global

protein properties like thermostability or pH dependency. Another limiting factor is the restricted quality

of computer simulations, which always is a compromise between informative value and time saving. On

the other hand, by creating large mutant libraries by methods of directed evolution, hundreds and

thousands of variants need to be screened. Without a fast and reliable assay system, the throughput is

so small that either the desired variant cannot be identified or the screening process takes very long

time. Nowadays, more and more researchers use combined methods of these two strategies, dubbed

focused directed evolution or semi-rational design25, 26. Here, information available from related protein

structures, families and mutants already identified is combined and then used for targeted

randomization of certain areas of the protein.

Access to new enzymes was traditionally gained through enrichment selection of microorganisms by

growth under limiting conditions or enzymes from plant or animal tissues were applied as crude extracts.

However, only a small number of microbes can successfully be cultivated in the laboratory

(approximately <1%)27. The metagenome approach circumvents this problem: the complete genomic

DNA is extracted from an environmental sample, fragmented and cloned yielding the corresponding

metagenome libraries. These libraries can then be screened either with a function-based assay for the

desired activity, or in a sequence-based approach in which genes are identified based on homology to

already described enzymes, or the entire DNA is sequenced. Besides the classical enrichment cultivation

and the metagenome approach, a third way to access new enzymes is called ‘in silico screening’. Protein

sequences obtained from sequencing of single enzymes, entire genomes or microbial consortia (e.g.

from the Sargasso Sea28), are deposited in public databases and currently >12 million protein sequences

are available (http://www.ncbi.nlm.nih.gov/RefSeq/). However, only a tiny fraction of all these

sequences corresponds to enzymes, which have been produced in the laboratory and their biochemical

properties and function have been experimentally confirmed. For the vast majority, a possible function is

postulated by sequence comparison, but this annotation can often be wrong or at least misleading, as

highlighted in a recent commentary29. Discovering novel biocatalysts in public databases is easier the

The Background

- 9 -

more information about the desired enzyme class exists. If several enzymes with a distinct activity are

already described in the literature, an alignment of their protein sequences together with experimentally

confirmed activities or specificities can reveal conserved regions within the sequences and a simple

BLAST-search can yield new and useful biocatalysts. There are numerous examples for the discovery of

new enzymes in silico including nitrilases30, a ligase31, an epoxide hydrolase32 and a reductase33. By

curtailing the sequences to be searched through, one is even able to increase the chances that the new

catalysts fulfil additional criteria. Scanning the genomes of (hyper-)thermophilic organisms for example

will increase the chances to find a thermostable enzyme. Thus, Machielsen et al. identified a

thermostable ADH in the genome of Pyrococcus furiosus with a half-life of 130 min at 100°C34 and Fraaije

et al. discovered a thermostable Baeyer-Villiger monooxygenase with a promiscuous sulphur oxidizing

activity35.

However, all these successful examples have one very important aspect in common: substantial

structural and especially sequence information about similar enzymes having the desired properties.

Unfortunately for the purpose of this project, there was no information about the type of enzyme we

were looking for, except that one commercially enzyme36 (Codexis ATA-117) and two microorganisms

with (R)-amine transaminase activity were known37, 38.

2 THE ANALYTICS

While working with large enzyme libraries, the fast, effective and reliable identification of enzymes with

the desired properties is often the bottleneck, since this is the most time-consuming step. Therefore a

large variety of fast and robust methods for determining enzyme activity have been developed for many

biocatalytic reactions, which often are capable of being performed in a high-throughput screening

format. But screening enzyme libraries for new or improved biocatalysts is not the only purpose of such

assays. Having found some promising candidates, an appropriate assay system will accelerate the

necessary biochemical characterization including substrate specificity, enantioselectivity and both

temperature and pH optima.

However, for the screening of amine transaminase (ATA) activity and said biochemical characterization

only a limited number of methods have been reported, due to the special type of reaction catalyzed: an

amine and a carbonyl substrate bearing a ketone or aldehyde group are converted into products by

exchanging their amino and carbonyl functional groups, without the stoichiometric consumption of a

cofactor. In case of other enzymatic reactions like e.g. hydrolysis of esters, the substrates (ester) and

products (alcohol and acid) can be easily discriminated39, or during reactions involving a cofactor like

NAD(P)H or side products like H2O2, these compounds can be detected specifically40, 41. For ATAs, these

approaches are not possible, and there are only a few methods to discriminate the similar substrates and

products without affecting enzyme activity42. Thus, only two high-throughput assays had been published

until recently.

The Analytics

- 10 -

Hwang et al. developed an assay based on the detection of alanine formed during ATA-catalyzed

reactions: after addition of a CuSO4/MeOH staining solution, the blue colored Cu-alanine complex could

be determined at 595 nm42. With this assay various aspects of the reaction can be monitored, like

investigating different amines and β-amino acids as substrates in combination with pyruvate as co-

substrate. As the assay depends solely on the conversion of the amino acceptor pyruvate, the amino

donor can be applied enantiomerically pure and thus the enantiopreference of a given enzyme can be

determined and an estimation of its enantioselectivity is possible (Figure 9a).

As an alternative, Truppo et al. published a multi enzyme cascade pH-indicator assay to monitor the

conversion in an asymmetric amine synthesis with ATAs43. The ketone is converted with alanine as co-

substrate, and NADH-dependent lactate or alanine dehydrogenases are used in order to remove the co-

product pyruvate. NADH is subsequently recycled with glucose dehydrogenase, and the arising gluconic

acid δ-lactone induces a decrease of the pH, which is visualized by the application of a pH-indicator

(Figure 9b).

Figure 9| Assays described in literature for measuring amine transaminase activity. ATA, amine transaminase; LDH, lactate dehydrogenase; GDH, glucose dehydrogenase; AAO, D- or L- amono acid oxidase; HRP, horse radish peroxidase; PGR, pyrogallol red.

Obviously, both methods show several drawbacks: since the copper staining solution in the first assay

inhibits the enzyme, the assay can only be used as an endpoint measurement and thus the

determination of kinetic parameters is rather difficult. Furthermore, the sensitivity is very low (ε ≈ 10 M-

1cm-1) and most commonly used buffers – e.g. phosphate and the Good’s buffer – form insoluble blue

Cu-complexes, which need to be centrifuged before the measurement. Cell extracts also give a blue

The Analytics

- 11 -

colored product, which, however, is soluble and thus cannot be removed. This is probably due to the

amino acids present in the cells. Although the sensitivity of the second assay is higher compared to the

CuSO4/MeOH assay, it works in the asymmetric synthesis mode and thus no information about

enantiopreference or -selectivity can be obtained. Furthermore, the second assay uses two additional

enzymes, which may be influenced by important parameters like temperature, pH and cosolvents.

Meanwhile, another method had been published using a combination of amino acid oxidase and

peroxidase for quantification of alanine via its oxidation and detection of the arising H2O244 (Figure 9c).

This assay has the disadvantage that it also depends on two additional enzyme activities, which must be

precisely synchronized.

Thus, our main target was to develop a reliable and easy-to-use assay system, which does not depend on

other enzyme activities and is resistant to external influences. In the end, we established two novel assay

systems for a fast and kinetic screening of ATA activity as well as for the characterization of the amino

acceptor and amino donor specificity of any given ATA.

2.1 A FAST AND EASY ASSAY FOR SCREENING ARTICLE II

In search of a better spectrophotometric assay we studied the absorption spectra of various compounds

and noticed that acetophenone, which is formed during the transamination of the standard substrate -

methylbenzylamine (-MBA), showed a high absorbance in the ultraviolet range of the spectrum with a

local maximum at 245 nm due to its carbonyl group conjugated to the aromatic system (Figure 10).

Compounds like pyruvate, alanine, alkylamines, -methylbenzylamine and non-aromatic ketones on the

other hand showed an absorption, which was less than 5% of acetophenone at this wavelength. Thus,

the reaction can be followed simply by monitoring the increasing absorption in the model reaction of

ATAs (Figure 10) neglecting absorption changes caused by other reaction partners. Since -MBA is

converted by most ATAs and all low absorbing ketones, aldehydes and keto acids may be used as

cosubstrates, the assay allows a fast characterization of the amino acceptor spectrum of any ATA,

determination of its enantiopreference as well as estimation of the enantioselectivity towards -MBA.

Furthermore, the pH and temperature optimum and stability can be characterized very easily. The

accuracy of the system was validated by measuring the absorption at 245 nm and detection of the

acetophenone formed by capillary electrophoresis in parallel. The results were comparable for both

purified enzyme and crude extract and the calculated activities differed only by 3%. The absorption

coefficient of acetophenone at 245 nm was determined to be ~12 mM-1cm-1, which results in a thousand

fold higher sensitivity compared to the CuSO4/MeOH based assay (10 M-1cm-1).

The Analytics

- 12 -

Figure 10| Principle of the developed acetophenone assay. Acetophenone (1 mM) shows a high absorbance (right)

compared to other reactants such as -methylbenzylamine, pyruvate or alanine (all 10 mM).

The only limitation of the assay comes along with the rather low wavelength of 245 nm, because in this

range the protein contributes to the initial absorbance. This limits the amount of protein that can be

applied, which eventually results in a decreased sensitivity at higher enzyme loads due to an increased

initial absorbance.

For verification, the assay was used to characterize an ATA identified in the genome of Rhodobacter

sphaeroides. After recombinant expression and affinity tag purification, the amino acceptor spectrum of

the biocatalyst was investigated. Furthermore, standard enzymatic properties like pH profile,

temperature profile and stability as well as the effect of additives or buffer compositions could be

investigated quickly and simply using this assay. Especially these last-mentioned properties could not

have been as easily determined with the previously described assay systems.

2.2 A SLIGHTLY AMBITIOUS ASSAY FOR CHARACTERIZATION ARTICLE III

As the spectophotometric assay is limited to α-MBA as amino donor, we developed a method for a fast

and easy characterization of the amino donor specificity, too. With this conductivity based approach

every amino donor and acceptor can be applied as substrate, as long as one of the substrates is an amino

or keto acid, respectively. In the course of an ATA-catalyzed reaction the conductivity of the reaction

medium changes since charged reactants – amine and keto acid – are converted to non-charged species

– ketone and a zwitterionic amino acid (Figure 11). For all experiments, the transamination was carried

out in the kinetic resolution mode. The rationale behind this approach is as follows: (i) the reaction

equilibrium favors product formation, (ii) both enantiomers of a given amine substrate can be used

separately in the assay, so that also information about enantioselectivity can be obtained, (iii) the amine

substrate usually is much better soluble than the corresponding ketone. Thus, even with higher substrate

The Analytics

- 13 -

concentrations a homogeneous assay solution is obtained rather than a biphasic system, which

otherwise might interfere with the measurements. The system allows a simple measurement of the

reaction progress, if some crucial requirements are fulfilled: first, to maximize sensitivity, the buffer

system should have low background conductivity. The pH of the buffer has to be kept in a pH range from

4 to 8, where the net charge of alanine is ≈0. Furthermore, for practical reasons or for high-throughput

screening purposes, the assay should not be sensitive towards variations in crude extract concentrations.

Figure 11| Principle and validation of the conductometric assay. A positively charged amine and a negatively charged keto acid are converted to zwitterionic amino acid and a non-charged ketone (left), thus the conductivity

of the solution decreases. The assay was validated by following the reaction of 10 mM -MBA and pyruvate (right) using both conductivity measurements (black symbols/bars) and detection of the acetophenone formed by gas chromatography (white bars).

Standard buffer systems like phosphate buffer or Tris-HCl could not be applied due to their very high

conductivity, whereby the relative change in conductivity caused by the enzymatic reaction is rather

small. Thus, five different Good’s buffers – BES, CHES, EPPS, HEPES and Tricine – were investigated

regarding the conductivity of 20 mM solutions at pH 7.5 and their effects on transaminase activity. Due

to their zwitterionic nature, solutions of these compounds show a much smaller conductivity if the pH is

kept within 2 units near the pI of the respective compound (isoelectric buffer). For CHES, EPPS and HEPES

we found significant inhibitory effects on the ATA from Rhodobacter sphaeroides (24% to 76% relative

activity compared to phosphate buffer) and though the BES buffer showed a very good performance

with purified enzyme, the activities of crude extract measured in this buffer unfortunately did not match

the results obtained by gas chromatography. Finally, the Tricine buffer system showed both a good

buffer capacity at pH 7.5 and a very good agreement of determined activities of purified enzyme and

crude extract by measuring the conductivity compared to GC analysis (Figure 11). For this reason the

Tricine buffer was used for all subsequent experiments. The inhibitory effect of the EPPS and CHES buffer

The Analytics

- 14 -

may vary for other transaminases, so these buffers should be considered for different enzymes due to

their really good performance in other respects.

To answer the second important question, whether the assay is sensible towards variations in crude

extract concentrations, we performed calibration experiments where different conversions were

simulated by varying the concentrations of all reactants according to a real reaction. As expected, the

background conductivity of the reaction medium increased proportional to the concentration of cell

extract used but the addition of different amounts of crude extract did not seem to affect the change of

conductivity significantly. This was surprising, because in theory the molar conductivity of a given

compound depends on the concentration of all electrolytes present in the reaction system for mainly

two reasons: on the one hand the protonation state and thus the netto charge of an analyte may be

affected by other compounds present in the solution. On the other hand and more importantly, ions or

even neutral molecules from the background electrolyte may affect the mobility and thus the molar

conductivity of an analyte significantly by intermolecular electrostatic or hydrophobic interactions. This

explains why the conductivity of a mixture comprised of different ions usually differs to some extent

from the sum of the contributions of the single ions to overall conductivity. For this reason, all

participating reactants and even the neutral species had to be included in the calibration of the assay. If

otherwise the change of the conductivity per millimolar reaction conversion [µS/mM] is calculated from

molar conductivities of the single compounds, the obtained value would be higher (49.4 µS/mM

conversion) compared to the measurement of the complete mixtures (44.1 µS/mM conversion).

In conclusion, the conductivity assay is an excellent complementary method to the spectrophotometric

assay and allows a fast screening of ATA variants for the reaction of any desired amine with a keto acid.

One limitation of the assay is that an appropriate buffer with low conductivity in the pH-range of 4–8 is

indispensable. For determining enzymatic properties like the pH- or temperature optimum, the

photometric assay is more flexible since every low absorbing buffer at any pH may be used and

absorbance is virtually not dependent on temperature, in contrast to conductivity. With both assays, the

complete characterization of the substrate specificity and estimation of enantioselectivity can be

performed quickly.

3 EXPLORING NEW PATHS IN ENZYME DISCOVERY ARTICLE IV

With two convenient assays at hand, we felt comfortable to start searching for the desired (R)-selective

amine transaminases (R-ATAs). We first considered protein engineering to create an R-ATA by inverting

the enantiopreference of an S-ATA. Inverting an enzyme’s enantiopreference is generally possible by

changing the binding pockets of the active site in a way that two substituents of the substrate’s chiral

carbon atom are bound inversely. This approach has already been applied successfully to many different

enzymes including e.g. a lyase45, a fluorophosphatase46 and an esterase47. Unfortunately, there were

several limitations, as no crystal structure of any ATA was available and thus rational protein design

The Search

- 15 -

would be very difficult to perform. Meanwhile, Svedendahl et al. succeeded in inverting the

enantiopreference of the S-ATA from Arthrobacter citreus at least substrate dependently: the mutant

V328A converted (R)-1-(4-fluorophenyl)propane-2-amine instead of the (S)-enantiomer, while it retained

its high (S)-selectivity for 1-(4-nitrophenyl)ethanamine48. Keeping in mind that the authors actually aimed

at increasing the enzymes enantioselectivity towards 1-(4-fluorophenyl)propane-2-amine – and were

successful with mutant Y331C – this example illustrates the tight boundaries of rational design, especially

without a proper 3D structure. However, there are crystal structures of (S)-selective amino acid TA

available. So we changed our plan and thought of reengineering the substrate-recognition site of an -TA

to create a variant that accepts substrates lacking the carboxylic function (Figure 12). This altered

substrate specificity would be accompanied by a formal switch in enantiopreference owing to the

changed priority according the Cahn-Ingold-Prelog (CIP) rule49, resulting in an (R)-selective amine TA.

Figure 12| Strategies for protein engineering. Possible ancestors of (R)-amine transaminases can be used to engineer the very same. This can be achieved by modification of the amino acids in the carboxyl group binding

pocket of an -TA such as an L-branched chain TA of the PLP fold type IV (left) or an aromatic amino acid TA from PLP fold type I (lower right) or by engineering of the binding pockets of an (S)-selective amine-TA (upper right) from PLP fold type I. It was assumed that according to the CIP-rules the large substituent (RL) has a higher priority than the small substituent (RS).

We explored this strategy with a well-investigated aromatic amino acid TA from Paracoccus

dentitrificans50 from PLP-fold class I. Unfortunately again, this enzyme performs a typical induced fit

mechanism and the coordination of the -carboxyl group of the substrate plays an important role for the

domain closure. The guanidinium group of Arg386 coordinates the carboxyl group in an end-on

geometry, while it is locked into position by several surrounding residues resulting in a complex network

of hydrogen bonds. To solve this challenge, many additional mutations might have been required, which

The Search

- 16 -

could not be predicted because of the complexity of the problem. After a close look at other (S)-selective

amino acid TAs from fold class I we realized that most of the enzymes showed an induced fit, too.

Therefore, we focused on fold class IV of PLP-dependent enzymes. In this fold class there are known two

types of transaminases with opposite enantiopreference, namely branched chain amino acid

transaminase51 (BCAT) and D-amino acid transaminase52 (DATA). This fact alone is very interesting since

it indicates a certain flexibility within this fold type, which apparently is not given in fold class I where all

enzymes are (S)-selective. The opposite enantiopreference of DATA and BCAT could be explained on the

basis of crystal structures by the differences in substrate coordination in the active site51, 52, which

consists of two binding pockets. If the -carboxyl group of the substrate is accommodated in binding

pocket A, this TA is (R)-selective and thus converts D-amino acids (Figure 12). If on the other hand the -

carboxyl group is positioned in binding pocket B, the enzyme converts L-amino acids and thus exhibits

(S)-preference (Figure 12). In other words, if (R)-selective amine TAs have evolved in this fold type, the

plausible ancestor must be an L-specific BCAT, since substitution of the carboxyl group of a L-amino acid

by a methyl group yields an (R)-amine – according to the CIP rule.

Working from the known crystal structures, we studied the coordination of the -carboxyl group in

BCATs and DATAs. This allowed us to predict the necessary differences in protein sequences within PLP-

fold class IV proteins that determine the desired switch in substrate specificity from -amino acids to

amines, in-line with the formal switch in enantiopreference. In contrast to the very rigid coordination in

aromatic amino acids TAs and DATAs, both BCATs from E. coli (PDB: 2IYD) and Thermus thermophilus

(PDB: 2EJ0) realize the coordination of the carboxyl group in a more subtle manner, without direct

contact of the carboxyl group oxygen atoms to any basic amino acid side chain such as an arginine

(Figure 13). Instead, the negative charge of one oxygen is compensated by a hydrogen bond to Tyr95,

which is activated by an adjacent Arg97. This pair of Tyr95 and Arg97 is highly conserved in all BCATs.

The other oxygen is bonded by two backbone amide nitrogen atoms of Thr262 and Ala263, which also

are activated by the coordination of their adjacent carbonyl groups by Arg40. Consequently, there are

two key mutations required to generate an (R)-selective amine TA from a BCAT. First, Tyr95 should be

exchanged by a hydrophobic residue, which is incapable of forming a hydrogen bond with the carboxyl

group. Secondly, Arg40 should be changed to a residue, which cannot activate the amide backbone

nitrogens of Thr262 and Ala263 by coordination of the neighbouring backbone carbonyl oxygens.

However, the situation regarding position 40 in PLP-fold class IV proteins is complex. DATA also has a

basic amino acid, Lys40, but in contrast to the Arg40 in BCAT, Lys40 in DATA adopts an altered

conformation so that its ε-amino group forms a hydrogen bond to the backbone of an adjacent loop.

That is why we included the absence of a basic residue at position 40 as a clear hint for R-ATA activity,

but we did not declare its presence as strict exclusion criterion.

The Search

- 17 -

Figure 13| Key amino acid motifs that allow prediction of function of PLP-dependent fold class IV proteins. Top: Schematic drawing of the external aldimines of the respective substrates in the active site of DATA, BCAT and ADCL. Blue colored amino acids are part of binding pocket A (Figure 12) and amino acids of binding pocket B (Figure 12) are shown in green. The grey shaded circle represents Gly or Val, respectively, which is located behind the carboxyl group. The red colored Thr36 is important during the catalytic mechanism for shuttling of a proton during the reaction transition state. Bottom: sequence motifs derived from multiple sequence alignments are given using the same color code, showing that amino acids important for substrate binding in the active site are rather conserved and can be used for a prediction of the substrate specificity of PLP-dependent fold class IV enzymes as well as for the enantiopreference of the transaminases within this fold class.

Instead of inserting these key mutations and performing several rounds of directed evolution to create

and identify an enzyme with the desired selectivity from a random mutant library, we developed a

strategy to search protein databases for proteins already carrying these mutations. These naturally

evolved “mutants” could have undergone selection over millions of years, which would have resulted in

highly optimized catalysts. Aside from the key residues considered important for amine TA activity, we

compared residues involved in substrate coordination in the different enzymes in order to exclude

undesirable specificities or activities – apart from BCATs and DATAs, the third enzyme class described in

fold class IV is 4-amino-4-deoxychorismate lyase53. Fortunately, the amino acids that are in direct contact

to the substrate in the active site are arranged in two relatively short sequence blocks. The first block is

located at positions 36–40 and the second block comprises six amino acids at positions 95-97 and 107-

109 (positions according to a multiple sequence alignment). Most of these amino acids fold into a -

sheet, only residues 107-109 are part of a loop. This information is important since during evolution

insertions or deletions take place preferentially in loop sequences without destroying enzyme activity

but not usually in α-helices or -sheets54. Alignments, which include all known BCAT, DATA and ADCL

proteins with experimentally verified enzyme activity, showed that the amino acids involved in substrate

recognition in the active site seem to be quite conserved in these three groups. This allowed us to

formulate different sequence motifs characteristic for DATA, BCAT and ADCL activity (Figure 13) Based

The Search

- 18 -

on these comparative considerations an annotation algorithm was developed using the amino acid

sequence motifs identified. This enabled an easy exclusion of all enzyme candidates, which could clearly

be assigned as BCAT, ADCL or DATA. The analysis of the remaining sequences aimed at identifying

transaminase sequences that fulfill the requirements for the desired (R)-selective amine-TA activity.

Using this algorithm, we analyzed about 5,700 sequences annotated as BCAT and 280 PLP-class IV-

annotated protein sequences from the NCBI database. This search identified 21 sequences, seven from

eukaryotes and 14 from prokaryotes. We ordered the synthetic genes, subcloned and expressed them in

E. coli BL21 and isolated the recombinant proteins via affinity tag purification. Seventeen of the 21

putative (R)-selective amine-TA genes that had been identified were found to be (R)-selective amine

transaminases. Three proteins could not be expressed in E. coli in sufficient amounts and one protein

was found to have very low activity on the substrates studied. Specific activity and substrate range

differed greatly between all of the enzymes investigated, which we find not surprising keeping in mind

that the protein sequences originate from various microorganisms, recombinant expression was never

reported before and specific activities within their natural function as well as the natural substrates are

unknown.

Figure 14| Characterization of the discovered (R)-selective amine transaminases. Specific activities toward (R) and

(S) enantiomers of amines 1-4 and amino acids 5 (D-alanine) and 6 (L-glutamate) are indicated by a color

gradient. -KG, -ketoglutarate.

The Search

- 19 -

Nevertheless, 10 out of the 21 proteins exhibited a specific activity >0.5 U/mg towards at least one of the

investigated amines (Figure 14), which is in the same activity range of known (S)-selective amine-TA55, 56.

Consequently, these newly identified (R)-selective transaminases were applied in the asymmetric

synthesis to yield optically pure (R)-amines (chapter 4).

The strength of this new in silico approach is that novel enzymes can be discovered very fast. On the

contrary, directed evolution requires several rounds of random mutagenesis or iterative saturation

mutagenesis in order to alter the enantioselectivity or substrate specificity, even if structural information

is available to identify hot spots for mutagenesis. Usually more than 103–104 variants have to be

screened for the identification of a mutant with desired properties. In contrast to this very low 'hit-rate',

about 50 % of the putative proteins identified in our study turned out to be useful biocatalysts, which are

an ideal starting point for further fine tuning and optimization by protein engineering. Hence, before

creating mutants or libraries based on rational predictions in the laboratory, it is worthwhile

investigating if nature has already designed the mutants and additionally optimized these catalysts over

million years of natural evolution.

4 APPLICATION OF THE NEW BIOCATALYSTS ARTICLE V

Before we applied the seven most promising candidates – with the highest specific activities towards

-methylbenzylamine (-MBA) and sufficient expression levels – in asymmetric syntheses of various

amines, we optimized the production of stable biocatalysts. During the initial biochemical

characterization we noticed that several enzymes were rather unstable after affinity tag purification in

the Tricine buffer used for the conductometric assay. Therefore, the crude extracts containing the seven

R-ATAs (AspTer from Aspergillus terreus, AspFum from Aspergillus fumigatus, AspOry from Aspergillus

oryzea, PenChry from Penicillium chrysogenum, NeoFis from Neosartorya fischeri, GibZea from Gibberella

zeae and MycVan from Mycobacterium vanbaalenii) were simply subjected to ammonium sulfate

precipitation and could be recovered fully active. Next, various additives – commonly used osmolytes

such as glycerol, ethylenglycol, PEG 1550, trehalose, saccharose and sarcosine as well as the surfactant

Tween 80 – were investigated for their stabilizing effect on the proteins during freeze-drying and storage

as lyophilizate and in solution. In most cases saccharose or glycerol had the best stabilizing effect on the

R-ATAs. All enzymes were stable for at least three months in lyophilized form and for at least two weeks

in solution at 4°C.

With stable and active biocatalysts at hand, the next important challenge for a successful asymmetric

synthesis of chiral amines was the identification of a good method to shift the equilibrium towards

product formation. This, unfortunately, is necessary because the equilibrium constant of e.g. the

standard reaction of -MBA and pyruvate is 8810057. Even a ten-fold excess of alanine in the reversed

reaction – the asymmetric syntheses of -MBA – would yield a theoretical yield of only <1%58. Luckily,

there have been some very convenient methods for shifting the equilibrium established. Most methods

The Application

- 20 -

remove the by-product pyruvate e.g. with a lactate dehydrogenase36 (LDH) or pyruvate decarboxylase59

(PDC) in order to pull the equilibrium to the product site. Another method uses alanine amino acid

dehydrogenase to recycle alanine from the arising pyruvate, thus pulling and pushing the equlibrium

towards product formation at the same time60. Maybe the most elegant method uses an excess of

isopropylamine as amino donor and the arising acetone is evaporated at low pressure61. The by-product

acetone can also be removed with an alcohol dehydrogenase62, but this requires – as all methods using

dehydrogenases – recycling of the cofactor NAD(P)H. Based on our good experiences in previous

projects, we investigated the use of PDC and the combination of LDH with glucose dehydrogenase

(GDH); the latter gave the most satisfying results in our preliminary experiments.

Next, several buffers were investigated to identify the best reaction system for biocatalysis. Two major

concerns led to these investigations: recent studies showed that the most frequently used phosphate

buffer is not always the best choice for transaminases because of inhibitory effects of the phosphate ions

– assumingly caused by blocking the binding site of the phosphate group of the PLP-cofactor. On the

other hand, we observed in previous studies that some Good’s buffers had an inhibitory effect on the S-

ATA from Rhodobacter sphaeroides. Therefore, the initial activity towards -MBA in four different buffer

systems (TRIS, HEPES, MOPS and BES) was compared to the activity in phosphate buffer. Altogether the

influence of the different buffer systems was only moderate and hence phosphate buffer was chosen for

subsequent experiments, as this buffer is known to be suitable for the LDH/GDH recycling system.

Eventually, we determined the pH profiles of the seven enzymes using the spectrophotometric assay.

The majority of the enzymes show a pH optimum between 8–9, which is also typical for S-ATAs such as

the enzyme from Vibrio fluvialis55; AspTer and PenChr show a sharp optimum at pH 8.5–9 while AspOry,

AspFum and NeoFis have a broader optimal range between 8–9. GibZea has a rather untypical pH

optimum of 7.5 and loses already 30 to 70% activity at pH 8.5 and 9 respectively, whereas MycVan has a

very broad optimum between pH 7.5–8.5 and still shows ~80% activity at pH 9. Nevertheless, in

combination with the LDH/GDH system preliminary experiments gave the highest yields at pH 7.5.

Finally, we applied the seven stabilized biocatalysts to asymmetric syntheses of various amines

(Figure 15) in the optimal reaction systems described above. For all substrates at least one R-ATA – and

for various compounds several enzymes – could be identified, which allowed complete conversion of the

ketone into the chiral amine with excellent enantioselectivity (>99% ee). This exceeded our expectations

as the ketones represent a rather diverse set of compounds. In only a few cases, lower conversions were

observed, but further optimization to increase the conversion was considered unnecessary in light of the

other very useful R-ATAs.

The Application

- 21 -

Figure 15| Conversion and enantiomeric excess of the amines obtained by asymmetric synthesis. Conversions are given in the order of the displayed products. The color gradient indicates the excess of the obtained (R) enantiomer, in case of VibFlu and RhoSph of the (S) enantiomer.

Although PenChr and AspTer show a rather high similarity score in a multiple sequence alignment (Figure

16), they showed very different performances. PenChr gave the lowest conversions, not exceeding 20% .

AspTer on the other hand is a pretty good candidate, especially for aliphatic amines. The two other

enzymes from Aspergillus species (AspOry and AspFum) turned out to be even better catalysts, as well as

NeoFis. According to their sequence identity these enzymes are much alike, especially AspFum and

NeoFis with a sequence identity of 96%. AspOry shares a similarity of 72% and 73%, respectively, and

gave very good to excellent conversions for the aliphatic substrates, considerably less conversion for 68

(probably due to the sterical hindrance in -position) and again very good conversions for the

arylaliphatic substrates. MycVan is the outsider of the investigated proteins, with similarity scores not

exceeding 40%. The phylogenetic tree (Figure 16) shows that MycVan has the highest degree of

relationship to the assumed ancestor, branched chain amino acid transaminase (BCAT). MycVan enabled

good yields for the aliphatic and arylaliphatic compounds, moderate yields for 6a and 7a and did not like

8a at all.

100% ee

95% ee

90% ee

70% ee

50% ee

30% ee

80% ee

99% ee

15% ee

The Application

- 22 -

Bringing together the observed performances during biocatalysis and the degree of relationship as

shown in the phylogenetic tree (Figure 16), several correlations attract attention. MycVan, for example,

is a good biocatalyst with a comparable performance as AspOry, AspFum and NeoFis, which show a high

degree of relationship. Although GibZea shows an even higher similarity towards these excellent

catalysts than MycVan, it showed only a rather poor performance. Interestingly, AspTer and PenChr are

also strongly related, but showed very different performances.

For comparative purposes we also applied two S-ATAs from Vibrio fluvialis and Rhodobacter sphaeroides

to asymmetric syntheses of the (S)-amines. We used the same reaction system as before and the yields

were good to moderate. But more importantly, the (S)-selective enzymes were significantly less

enantioselective as our (R)-selective enzymes (Figure 15). Just two amines (7 and 8) could be obtained

with >99% ee. In literature, only two of our substrates (7 and 9) were used for asymmetric syntheses

with the enzyme from V. fluvialis and the reported enantioselectivities match our results58, 59. In case of

aliphatic amines (15), the Trp57Gly mutant yielded excellent enantiomeric excesses in kinetic

resolutions57, but – on the contrary – did not show any activity in asymmetric syntheses63. The enzyme

from R. sphaeroides has not been used for asymmetric syntheses in literature. For some substrates, our

enzymes showed higher selectivities in comparison to the commercially available ATA-117, too. Although

this enzyme shows high enantioselectivities towards 9 and short aliphatic amines, the synthesis of 7

yielded only 96% ee and of 2-aminooctane just 90% ee36.

Figure 16| Phylogenetic tree of the investigated R-ATAs. The figure shows all 21 hits from the database search and the biocatalysts applied in asymmetric synthesis are named. DATA (D-amino acid transaminase), ADCL (4-amino-4-deoxychorismate lyase) and BCAT (branched chain amino acid transaminase) are the other three members of the PLP-fold type IV.

The Conclusion

- 23 -

5 CONCLUSION

In this thesis, two novel assay systems had been developed, which allow a fast and easy screening for

amine transaminase activity as well as the characterization of the amino donor and acceptor specificity

of a given amine transaminase. The assays overcome some limitations of previously described assays but

of course have some limitations themselves. The relatively low wavelength of 245 nm, at which the

production of acetophenone is detected with the spectrophotometric assay, limits the amount of

protein/crude extract that can be applied, which eventually results in a decreased sensitivity at higher

enzyme loads due to an increased initial absorbance. Otherwise, this assay can be used very easily for

the investigation of the amino acceptor specificity and both pH and temperature dependencies of amine

transaminases. The conductometric assay is – by its very nature – limited to low-conducting buffers, a

neutral pH and constant temperatures. In summary, the assays complement one another very well and

the complete characterization of the most important enzyme properties can be accomplished quickly.

Furthermore, we developed and applied a novel in silico search strategy for the identification of (R)-

selective amine transaminases in sequence databases. Structural information of probably related

proteins was used for rational protein design to predict key amino acid substitutions that indicate the

desired activity. We subsequently searched protein databases for proteins already carrying these

mutations instead of constructing the corresponding mutants in the laboratory. This methodology

exploits the fact that naturally evolved proteins have undergone selection over millions of years, which

has resulted in highly optimized catalysts. Using this in silico approach, we have discovered 17 (R)-

selective amine transaminases. In theory, this strategy can be applied to other enzyme classes and fold

types as well and for this reason constitutes a new concept for the identification of desired enzymes.

Finally, we applied the seven most promising candidates of the identified proteins to asymmetric

synthesis of various optical pure amines with (R)-configuration starting from the corresponding ketones.

We used a lactate dehydrogenase/glucose dehydrogenase system for the necessary shift of the

thermodynamic equilibrium. For all ketones at least one enzyme was found that allowed complete

conversion to the corresponding chiral amine with excellent optical purities >99% ee. Bearing in mind

that until last year there was only one (R)-selective amine transaminase commercially available and two

microorganisms with the corresponding activity described, the identification of numerous enzymes is a

breakthrough in asymmetric synthesis of chiral amines.

The References

- 24 -

6 REFERENCES

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

2. Teuscher, E., Melzig, M.F. & Lindequist, U. Biogene Arzneimittel. (Wissenschaftliche Verlagsgesellschaft, Stuttgart; 2004).

3. Mack, D.J., Weinrich, M.L., Vitaku, E. & Njardarson, J.T. (University of Arizona, Arizona; 2011). 4. Voet, D. & Voet, J.G. Biochemie. (VCH, Weinheim; 1994). 5. 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. 43, 788-824 (2004).

6. Nugent, T.C. & El-Shazly, M. Chiral amine synthesis – recent developments and trends for enamide reduction, reductive amination, and imine reduction. Adv. Synth. Catal. 352, 753-819 (2010).

7. List, B. Direct catalytic asymmetric α-amination of aldehydes. J. Am. Chem. Soc. 124, 5656-5657 (2002).

8. Li, C. & Xiao, J. Asymmetric hydrogenation of cyclic imines with an ionic Cp*Rh(III) catalyst. J. Am. Chem. Soc. 130, 13208-13209 (2008).

9. Cogan, D.A., Liu, G. & Ellman, J. Asymmetric synthesis of chiral amines by highly diastereoselective 1,2-additions of organometallic reagents to N-tert-butanesulfinyl imines. Tetrahedron 55, 8883-8904 (1999).

10. Hou, C.T. (ed.) Handbook of industrial biocatalysis. (CRC Press, Boca Raton; 2005). 11. Savile, C.K., Janey, J.M., Mundorff, E.C., Moore, J.C., Tam, S., Jarvis, W.R., Colbeck, J.C., Krebber,

A., Fleitz, F.J., Brands, J., Devine, P.N., Huisman, G.W. & Hughes, G.J. Biocatalytic asymmetric synthesis of chiral amines from ketones applied to Sitagliptin manufacture. Science 329, 305-309 (2010).

12. Desai, A.A. Sitagliptin manufacture: a compelling tale of green chemistry, process intensification, and industrial asymmetric catalysis. Angew. Chem. Int. Ed. 50, 1974-1976 (2011).

13. Höhne, M. & Bornscheuer, U. Biocatalytic routes to optically active amines. ChemCatChem 1, 42-51 (2009).

14. Alexeeva, M., Enright, A., Dawson, M.J., Mahmoudian, M. & Turner, N.J. Deracemization of α-methylbenzylamine using an enzyme obtained by in vitro evolution. Angew. Chem. Int. Ed. 114, 3309-3312 (2002).

15. Itoh, N., Yachi, C. & Kudome, T. Determining a novel NAD+-dependent amine dehydrogenase with a broad substrate range from Streptomyces virginiae IFO 12827: purification and characterization. J. Mol. Catal. B-Enzym. 10, 281-290 (2000).

16. Iwaki, M., Yagi, T., Horiike, K., Saeki, Y., Ushijima, T. & Nozaki, M. Crystallization and properties of aromatic amine dehydrogenase from Pseudomonas sp. Arch. Biochem. Biophys. 220, 253-262 (1983).

17. Govindaraj, S., Eisenstein, E., Jones, L.H., Sanders-Loehr, J., Chistoserdov, A.Y., Davidson, V.L. & Edwards, S.L. Aromatic amine dehydrogenase, a second tryptophan tryptophylquinone enzyme. J. Bacteriol. 176, 2922-2929 (1994).

18. Frey, P.A. & Hegemann, P. Enzymatic reaction mechanism. (Oxford University Press, New York; 2007).

19. Jansonius, J.N. Structure, evolution and action of vitamin B6-dependent enzymes. Curr. Opin. Struc. Biol. 8, 759-769 (1998).

The References

- 25 -

20. Percudani, R. & Peracchi, A. The B6 database: a tool for the description and classification of vitamin B6-dependent enzymatic activities and of the corresponding protein families. BMC Bioinformatics 10, 273 (2009).

21. Christen, P. & Metzler, D.E. (eds.) Transaminases. (Wiley, 1985). 22. Eliot, A.C. & Kirsch, J.F. Pyridoxal phosphate enzymes: mechanistic, structural, and evolutionary

considerations. Annu. Rev. Biochem. 73, 383-415 (2004). 23. Koszelewski, D., Tauber, K., Faber, K. & Kroutil, W. ω-Transaminases for the synthesis of non-

racemic α-chiral primary amines. Trends Biotechnol. 28, 324-332 (2010). 24. Kazlauskas, R.J. & Bornscheuer, U.T. Finding better protein engineering strategies. Nat. Chem.

Biol. 5, 526-529 (2009). 25. Reetz, M.T., Wang, L.-W. & Bocola, M. Directed evolution of enantioselective enzymes: iterative

cycles of CASTing for probing protein-sequence space. Angew. Chem. Int. Ed. 118, 1258-1263 (2006).

26. Fox, R.J., Davis, S.C., Mundorff, E.C., Newman, L.M., Gavrilovic, V., Ma, S.K., Chung, L.M., Ching, C., Tam, S., Muley, S., Grate, J., Gruber, J., Whitman, J.C., Sheldon, R.A. & Huisman, G.W. Improving catalytic function by ProSAR-driven enzyme evolution. Nat. Biotech. 25, 338-344 (2007).

27. Pace, N.R. A molecular view of microbial diversity and the biosphere. Science 276, 734-740 (1997).

28. Venter, J.C., Remington, K., Heidelberg, J.F., Halpern, A.L., Rusch, D., Eisen, J.A., Wu, D., Paulsen, I., Nelson, K.E., Nelson, W., Fouts, D.E., Levy, S., Knap, A.H., Lomas, M.W., Nealson, K., White, O., Peterson, J., Hoffman, J., Parsons, R., Baden-Tillson, H., Pfannkoch, C., Rogers, Y.-H. & Smith, H.O. Environmental genome shotgun sequencing of the Sargasso Sea. Science 304, 66-74 (2004).

29. Furnham, N., Garavelli, J.S., Apweiler, R. & Thornton, J.M. Missing in action: enzyme functional annotations in biological databases. Nat. Chem. Biol. 5, 521-525 (2009).

30. Kaplan, O., Bezouška, K., Malandra, A., Veselá, A., Petříčková, A., Felsberg, J., Rinágelová, A., Křen, V. & Martínková, L. Genome mining for the discovery of new nitrilases in filamentous fungi. Biotechnol. Lett. 33, 309-312 (2011).

31. Kino, K., Noguchi, A., Arai, T. & Yagasaki, M. Identification and characterization of a novel L-amino acid ligase from Photorhabdus luminescens subsp. laumondii TT01. J.Biosci. Bioeng. 110, 39-41 (2010).

32. van Loo, B., Kingma, J., Arand, M., Wubbolts, M.G. & Janssen, D.B. Diversity and biocatalytic potential of epoxide hydrolases identified by genome analysis. Appl. Environ. Microbiol. 72, 2905-2917 (2006).

33. Wang, L.-J., Li, C.-X., Ni, Y., Zhang, J., Liu, X. & Xu, J.-H. Highly efficient synthesis of chiral alcohols with a novel NADH-dependent reductase from Streptomyces coelicolor. Bioresource Technol. 102, 7023-7028 (2011).

34. Machielsen, R., Uria, A.R., Kengen, S.W.M. & van der Oost, J. Production and characterization of a thermostable alcohol dehydrogenase that belongs to the aldo-keto reductase superfamily. Appl. Environ. Microbiol. 72, 233-238 (2006).

35. Fraaije, M.W., Wu, J., Heuts, D.P.H.M., van Hellemond, E.W., Spelberg, J.H.L. & Janssen, D.B. Discovery of a thermostable Baeyer–Villiger monooxygenase by genome mining. Appl. Microbiol. Biot. 66, 393-400 (2005).

36. Koszelewski, D., Lavandera, I., Clay, D., Rozzell, D. & Kroutil, W. Asymmetric synthesis of optically pure pharmacologically relevant amines employing ω-transaminases. Appl. Environ. Microbiol. 350, 2761-2766 (2008).

37. Iwasaki, A., Yamada, Y., Ikenaka, Y. & Hasegawa, J. Microbial synthesis of (R)- and (S)-3,4-dimethoxyamphetamines through stereoselective transamination. Biotechnology Letters 25, 1843-1846 (2003).

The References

- 26 -

38. Matcham, G.W. & Bowen, A.R.S. Biocatalysis for chiral intermediates: meeting commercial and technical challenges. Chim. Oggi 14, 20-24 (1996).

39. Krebsfänger, N., Schierholz, K. & Bornscheuer, U.T. Enantioselectivity of a recombinant esterase from Pseudomonas fluorescens towards alcohols and carboxylic acids. J. Biotechnol. 60, 105-111 (1998).

40. Hinman, L.M. & Blass, J.P. An NADH-linked spectrophotometric assay for pyruvate dehydrogenase complex in crude tissue homogenates. J. Biol. Chem. 256, 6583-6586 (1981).

41. Holt, A. & Palcic, M.M. A peroxidase-coupled continuous absorbance plate-reader assay for flavin monoamine oxidases, copper-containing amine oxidases and related enzymes. Nat. Protocols 1, 2498-2505 (2006).

42. Hwang, B.-Y. & Kim, B.-G. High-throughput screening method for the identification of active and enantioselective ω-transaminases. Enzyme Micro. Tech. 34, 429-436 (2004).

43. Truppo, M.D., Rozzell, J.D., Moore, J.C. & Turner, N.J. Rapid screening and scale-up of transaminase catalysed reactions. Org. Biomol. Chem. 7, 395-398 (2009).

44. Hopwood, J., Truppo, M.D., Turner, N.J. & Lloyd, R.C. A fast and sensitive assay for measuring the activity and enantioselectivity of transaminases. Chem. Commun. 47, 773-775 (2011).

45. Williams, G.J., Woodhall, T., Farnsworth, L.M., Nelson, A. & Berry, A. Creation of a pair of stereochemically complementary biocatalysts. J. Am. Chem. Soc. 128, 16238-16247 (2006).

46. Melzer, M., Chen, J.C.H., Heidenreich, A., Gab, J., Koller, M., Kehe, K. & Blum, M.M. Reversed enantioselectivity of diisopropyl fluorophosphatase against organophosphorus nerve agents by rational design. J. Am. Chem. Soc. 131, 17226-17232 (2009).

47. Bartsch, S., Kourist, R. & Bornscheuer, U.T. Complete inversion of enantioselectivity towards acetylated tertiary alcohols by a double mutant of a bacillus subtilis esterase. Angew. Chem. Int. Ed. 47, 1508-1511 (2008).

48. Svedendahl, M., Branneby, C., Lindberg, L. & Berglund, P. Reversed enantiopreference of an ω-transaminase by a single-point mutation. ChemCatChem 2, 976-980 (2010).

49. Cahn, R.S., Ingold, C. & Prelog, V. Specification of molecular chirality. Angew. Chem. Int. Ed. 5, 385-415 (1966).

50. Okamoto, A., Ishii, S., Hirotsu, K. & Kagamiyama, H. The active site of Paracoccus denitrificans aromatic amino acid aminotransferase has contrary properties:  flexibility and rigidity. Biochemistry 38, 1176-1184 (1999).

51. Goto, M., Miyahara, I., Hayashi, H., Kagamiyama, H. & Hirotsu, K. Crystal structures of branched-chain amino acid aminotransferase complexed with glutamate and glutarate:  true reaction intermediate and double substrate recognition of the enzyme. Biochemistry 42, 3725-3733 (2003).

52. Sugio, S., Petsko, G.A., Manning, J.M., Soda, K. & Ringe, D. Crystal structure of a D-amino acid aminotransferase: how the protein controls stereoselectivity. Biochemistry 34, 9661-9669 (1995).

53. O'Rourke, P.E.F., Eadsforth, T.C., Fyfe, P.K., Shepherd, S.M. & Hunter, W.N. Pseudomonas aeruginosa 4-amino-4-deoxychorismate lyase: spatial conservation of an active site tyrosine and classification of two types of enzyme. PLoS One 6, e24158 (2011).

54. Wolf, Y., Madej, T., Babenko, V., Shoemaker, B. & Panchenko, A. Long-term trends in evolution of indels in protein sequences. BMC Evol. Biol. 7, 19 (2007).

55. Shin, J.S., Yun, H., Jang, J.W., Park, I. & Kim, B.G. Purification, characterization, and molecular cloning of a novel amine:pyruvate transaminase from Vibrio fluvialis JS17. Appl. Microbiol. Biot. 61, 463-471 (2003).

56. Hanson, R.L., Davis, B.L., Chen, Y., Goldberg, S.L., Parker, W.L., Tully, T.P., Montana, M.A. & Patel, R.N. Preparation of (R)-amines from racemic amines with an (S)-amine transaminase from Bacillus megaterium. Appl. Environ. Microbiol. 350, 1367-1375 (2008).

The References

- 27 -

57. Yun, H., Hwang, B.-Y., Lee, J.-H. & Kim, B.-G. Use of enrichment culture for directed evolution of the Vibrio fluvialis JS17 ω-transaminase, which is resistant to product inhibition by aliphatic ketones. Appl. Environ. Microbiol. 71, 4220-4224 (2005).

58. Shin, J.-S. & Kim, B.-G. Asymmetric synthesis of chiral amines with ω-transaminase. Biotechnol. Bioeng. 65, 206-211 (1999).

59. Höhne, M., Kühl, S., Robins, K. & Bornscheuer, U.T. Efficient asymmetric synthesis of chiral amines by combining transaminase and pyruvate decarboxylase. ChemBioChem 9, 363-365 (2008).

60. Koszelewski, D., Lavandera, I., Clay, D., Guebitz, G.M., Rozzell, D. & Kroutil, W. Formal asymmetric biocatalytic reductive amination. Angew. Chem. Int. Ed. 120, 9477-9480 (2008).

61. Matcham, G., Bhatia, M., Lang, W., Lewis, C., Nelson, R., Wang, A. & Wu, W. Enzyme and reaction engineering in biocatalysis: synthesis of (S)-methoxyisopropylamine (= (S)-1-methoxypropan-2-amine). Chimia 53, 584-589 (1999).

62. Cassimjee, K.E., Branneby, C., Abedi, V., Wells, A. & Berglund, P. Transaminations with isopropyl amine: equilibrium displacement with yeast alcohol dehydrogenase coupled to in situ cofactor regeneration. Chem. Commun. 46, 5569-5571 (2010).

63. Koszelewski, D., Göritzer, M., Clay, D., Seisser, B. & Kroutil, W. Synthesis of optically active amines employing recombinant ω-Transaminases in E. coli cells. ChemCatChem 2, 73-77 (2010).

The Articles

- 28 -

AUTHOR CONTRIBUTION

Article I Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

G. A. Behrens*, A. Hummel*, S. K. Padhi*, S. Schätzle*, U. T. Bornscheuer*, Adv. Synth.

Catal. 2011, 353, 2191-2215

* All authors contributed equally.

Article II Rapid and Sensitive Kinetic Assay for Characterization of ω-Transaminases

S. Schätzle, M. Höhne, E. Redestad, K. Robins, U. T. Bornscheuer, Anal. Chem. 2009, 81,

8244-8248

S.S. designed and performed the experiments with the help of M.H. and E.R. U.T.B, M.H. and S.S. wrote

the manuscript.

Article III Conductometric Method for the Rapid Characterization of the Substrate Specificity of

Amine-Transaminases

S. Schätzle, M. Höhne, K. Robins, U. T. Bornscheuer, Anal. Chem. 2010, 82, 2082-2086

S.S. designed and performed the experiments. M.H. conceived the original idea. U.T.B. and S.S. wrote the

manuscript.

Articel IV Rational Assignment of Key Motifs for Function guides in silico Enzyme Identification

M. Höhne*, S. Schätzle*, H. Jochens, K. Robins, U. T. Bornscheuer, Nature Chem. Biol.

2010, 6, 807–813

* These authors contributed equally. S.S. coordinated the comparative characterization of all proteins

and performed cloning, expression, purification, data collection and data analysis. M.H. designed the in

silico strategy, devised the annotation algorithm and performed the database search and identification

of the putative amine transaminases. M.H. expressed and confirmed amine transaminase activity and

(R)-selectivity for the first three proteins. H.J. contributed to gene cloning, protein expression and

activity measurements. K.R. and U.T.B. initiated the project. U.T.B. and M.H. cowrote the paper, and all

authors read and edited the manuscript.

Article V Enzymatic Asymmetric Synthesis of Enantiomerically Pure Aliphatic, Aromatic and

Arylaliphatic Amines with (R)-Selective Amine Transaminases

S. Schätzle, F. Steffen-Munsberg, A. Thontowi, M. Höhne, K. Robins, U. T. Bornscheuer,

Adv. Synth. Catal. 2011, 353, 2439-2445

S.S designed the experiments. S.S. and F.S.M performed the experiments with the help of A.T. U.T.B and

S.S. wrote the manuscript.

Confirmed, Greifswald, Nov. 22nd 2011 Uwe Bornscheuer

The Articles

- 29 -

ARTICLE I

DOI: 10.1002/adsc.201100446

Discovery and Protein Engineering of Biocatalysts for OrganicSynthesis

Geoffrey A. Behrens,a,b Anke Hummel,a,b Santosh K. Padhi,a,b

Sebastian Sch�tzle,a,b and Uwe T. Bornscheuera,b,*a Department of Biotechnology & Enzyme Catalysis, Institute of Biochemistry, Greifswald University, Felix-Hausdorff-Str.

4, 17487 Greifswald, GermanyFax: (+49)-3834-86-794367; e-mail: uwe.bornscheuer@uni-greifswald.de

b All authors contributed equally

Received: May 30, 2011; Revised: July 13, 2011; Published online: September 5, 2011

Abstract: Modern tools for enzyme discovery andprotein engineering substantially broadened thenumber of enzymes applicable for biocatalysis andhelped to alter their properties such as substraterange, enantioselectivity, and stability under processconditions. In addition, these methods also enabledone to explore reactions for organic synthesis forwhich no suitable enzymes were available until re-cently. This review provides a summary of the differ-ent concepts and technologies, which are exemplifiedfor various enzymes.

1 Introduction2 Tools for Enzyme Discovery2.1 Metagenome Approach2.2 Database Mining3 Tools for Protein Engineering3.1 Methods for Directed Evolution

3.2 Methods for Rational Protein Design3.3 Semi-Rational Design/Focused Directed Evolu-

tion4 Selected Examples4.1 Transaminases4.2 Enoate Reductases4.3 Esterases for Tertiary Alcohol Synthesis4.4 Monoamine Oxidases4.5 Dehalogenases4.6 Aldolases4.7 Cytochrome P450-Monooxygenases4.8 Baeyer–Villiger Monooxygenases5 Biocatalysts for New Reactions6 Conclusions

Keywords: biocatalysis; directed evolution; enantio-selectivity; enzyme catalysis; protein engineering; ra-tional protein design

1 Introduction

The earliest examples for the application of enzymesas biocatalysts for organic syntheses date back morethan a century. For instance, Rosenthaler used a plantextract for the synthesis of (R)-mandelonitrile frombenzaldehyde and hydrogen cyanide in 1908.[1] Still,until about 25 years ago, researchers who wished touse biocatalysts for organic synthesis had access toonly a handful of commercially available enzymes fora given reaction. For instance, about 5–10 lipasesfrom a few suppliers were usually at hand to find abiocatalyst with desired activity towards a certain sub-strate and if none of them showed sufficient activityand especially enantioselectivity, then variations suchas use of different solvents (medium engineering) oralternative acyl donors in transesterifications (oresters of different chain-length in hydrolysis; substrateengineering) were the means to find a solution for the

biocatalysis challenge. Similarly, for reduction of ke-tones to yield optically active chiral alcohols eitherbaker’s yeast [notably containing various ketoreduc-tases/alcohol dehydrogenases (ADH) with differentselectivities] or a few animal-derived (such as horseliver ADH) or microbial (such as Thermoanaerobiumbrockii ADH, Lactobacillus species ADH) enzymescould be investigated. For other enzyme classes, avail-ability was a major hindrance and only minuteamounts could be ordered at high price, as was thecase for, e.g., an aldolase from rabbit muscle(RAMA) before the enzyme could be cloned and re-combinantly expressed.[2] Although researchers coulddemonstrate the synthetic utility of RAMA, gram-scale synthesis was almost unaffordable, larger scalesynthesis impossible.

In order to find a new biocatalyst even for a similarreaction one had to screen many microbial sources inextensive screening programs. Altering important bio-

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 2191

REVIEWS

catalytic properties such as substrate range or selec-tivity, which are essential for using a biocatalyst in anorganic reaction, remained a challenge for a longtime. The target to find a biocatalyst with high robust-ness, thermostability, resistance to organic solventsand further process-related conditions often remainedunfulfilled, which prevented the area of enzyme catal-ysis from being competitive with chemicals routes fordecades. Although the discovery of recombinantDNA technology in the late 1970s was a major break-through, it still did not solve the issue of generatingan enzyme in the laboratory with desired biocatalyticproperties. Consequently, many attempts to establishbiocatalytic reactions suffered from the lack of avail-able enzymes and easy methods to alter their proper-ties. Of course, rational protein design based on X-ray

structures of proteins was then already possible, butmodelling for the prediction of positions for mutagen-esis was mostly limited to bioinformatic experts ableto program UNIX, and computers and software werevery expensive. Gene identification, cloning and re-combinant expression of proteins in microorganismswere also not so easy and mostly performed by expertmicrobiology groups.

This situation has changed dramatically in the pasttwo decades, fostered mostly by progress in molecularbiology and bioinformatics, but the reasons are many.For instance, sequencing of genomes and entire mi-crobial consortia now takes months not years; cheap,easy and fast access to synthetic genes circumventsmany time-consuming steps traditionally starting fromgrowth of a microorganism, protein purification, N-

Geoffrey A. Behrens (born 1985) studied biochem-istry in Greifswald, Germany. In 2010 after finishinghis diploma thesis, he started his PhD thesis onamino acid and amino acid derivative-converting en-zymes under supervision of Uwe Bornscheuer. Hismain focus is on protein engineering, by rationaldesign and by evolutive methods, of these enzymes.

Anke Hummel (born 1981) studied biochemistry inGreifswald, Germany. During her diploma thesis in2007 she identified and characterised novel isoen-zymes of pig liver esterase. Currently she is workingon her PhD thesis on the isolation of novel enzymesfrom metagenome sources under supervision ofUwe Bornscheuer. During her studies she stayedboth in Reykjav�k, Iceland and in Lund, Sweden fora half-year research internship.

Santosh Kumar Padhi (born 1977) studied chemistryand obtained his Ph.D. (2005) from the Indian Insti-tute of Technology Madras, India with supervisionby Anju Chadha. Between 2005 and 2009, he held

postdoctoral positions at the University of Florida,Gainesville, USA with Jon D. Stewart, and Universi-ty of Minnesota, Minneapolis, USA with Romas J.Kazlauskas. Currently he is a postdoctoral associateat the University of Greifswald in the group of UweBornscheuer as an Alexander von Humboldt re-search fellow. His research interests include enzymecatalysis and enzyme discovery by protein engineer-ing for organic synthesis, especially for asymmetricsynthesis.

Sebastian Sch�tzle (born 1983) studied Biochemistryin Greifswald. During his diploma thesis in 2008 hecharacterised several nitrile hydratases. Currently heis working on his PhD thesis on the identificationand application of novel (R)-selective amine transa-minases under supervision of Uwe Bornscheuer.During his studies, he stayed in Hat Yai, Thailandfor a short research internship.

Uwe T. Bornscheuer (born 1964) studied chemistryand completed his doctorate in 1993 at the Universi-ty of Hannover. He then was a postdoc at the Uni-versity of Nagoya, Japan. In 1998, he completed hisHabilitation at the University of Stuttgart at the In-stitute of Technical Biochemistry. He has been Pro-fessor at the Institute of Biochemistry at the Univer-sity of Greifswald since 1999. Bornscheuer hasedited and written several books, is Editor-in-Chiefof the Eur. J. Lipid Sci. Technol. and Co-Chairmanof the recently launched journal ChemCatChem. In2008, he received the BioCat2008 Award for his in-novative work on tailored biocatalysts for industrialapplications. His current research interest is focusingon protein engineering of enzymes from variousclasses with special emphasis on the synthesis ofchiral compounds and in lipid modification.

2192 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

terminal sequencing, genomic DNA (or mRNA) iso-lation, eventually succeeding in the identification ofthe protein encoding gene and its recombinant ex-pression. Consequently, the time scale to have the re-combinant protein at hand is now reduced from manymonths and often years to just a few weeks. Impres-sive examples are the discovery and characterizationof >150 nitrilases in metagenome libraries by DeSan-tis et al.[3] , while only a few nitrilases were foundbefore, during several decades of microbial screening.We have recently discovered 20 sequences encoding(R)-selective transaminases by an in silico screeningapproach[4] and the cloning, production and prelimi-nary characterization of these novel enzymes tookjust a few months.

Models of protein structure can nowadays be auto-matically generated (e.g., by Phyre[5] , SwissModell[6]

or Robetta[7]), software for molecular modelling nowruns on standard personal computers and is rathereasy to use (e.g., Yasara[8]), tools for sequence com-parison (e.g., BLAST: http://blast.ncbi.nlm.nih.gov/Blast.cgi[9]) and alignment (e.g., ClustalW: http://www.ebi.ac.uk/Tools/msa/clustalw2/[10]) enable quickidentification of novel sequences and conservedmotifs. A plethora of expression systems and chaper-one kits are available to remove the bottleneck of te-

dious identification of best conditions for recombi-nant enzyme production. Furthermore, protein engi-neering is not restricted any longer to rational designas the invention of directed (molecular) evolution inthe early 1990s based on simple random mutagenesisby methods such as error-prone PCR in combinationwith high-throughput screening makes creation of li-braries and identification of desired enzymes relative-ly easy. This now allows researchers to identify betterenzymes even if the knowledge about their function israther small.

Fostered by these important achievements on thetechnology side, modern biocatalysis using proteindiscovery and engineering now enables many academ-ic scientists as well as industrial researchers to estab-lish enzymatic routes for the synthesis of chemicalsthat are highly competitive to organic synthesis.Hence, numerous novel biocatalytic processes havebeen implemented on the multi-kilogram to ton scalessince then, often replacing existing chemical routes.The most recent and impressive example is the devel-opment of an engineered (R)-selective transaminaseand its implementation for the highly efficient andenantioselective synthesis[11] of the drug Sitagliptin,which replaced the transition metal-catalyzed hydro-genation reaction. Apart from the industrial exam-

Figure 1. Overview of approaches for protein discovery and engineering by rational, evolutionary or combined methods.

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2193

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

ples, a large number of commercial enzymes (listed inthe key references cited below) are available today,which allows many organic chemistry researchers toexplore their synthetic potential.

It would be impossible to give a full coverage ofthe achievements made in the past decades in thefield of protein engineering of enzymes for organicsynthesis in this article. Many examples are coveredtogether with applications in books[12] or (mostrecent) reviews.[13] This review rather aims to providean overview about principle approaches for proteindiscovery (Figure 1), their engineering by rational andevolutionary methods and hopes to provide insightsto facilitate success when facing future protein engi-neering problems. In addition, selected examples fordifferent enzyme classes and industrial processes aregiven to demonstrate the usefulness of biocatalystsfor organic synthesis.

2 Tools for Enzyme Discovery

Despite the high number of already described en-zymes, there is a constant need for novel biocatalystswith altered properties such as substrate scope, enan-tioselectivity or -preference, thermostability, pH pro-file, solvent tolerance, etc. To meet these demands,there are basically three strategies: (i) search for newenzymes, (ii) modification of already characterized en-zymes, or (iii) de novo design of completely new bio-catalysts as a potential, but not yet established route(Figure 1).

2.1 Metagenome Approach

Traditionally, access to new enzymes was gainedthrough enrichment selection of microorganisms bygrowth under limiting conditions or enzymes fromplant or animal tissues were applied as crude extracts.This has been overcome by the development of re-combinant DNA techniques, which have substantiallysimplified the access to enzymes and now allow theproduction of biocatalysts at stable quality as wasshown, for instance, for the hydroxynitrile lyase fromHevea brasiliensis[14] or pure isoenzymes of pig liveresterase.[15]

However, only a small number of microbes can suc-cessfully be cultivated in the laboratory. Dependingon the habitat, it was estimated that <1% of the mi-crobes present in an environmental sample can begrown under standard laboratory conditions.[16] Rea-sons include the facts that proper cultivation condi-tions are unknown, growth rates are too slow or thegrowth of a microbe is linked to other species provid-ing nutrients. The metagenome approach overcomesthese limitations by circumventing the cultivation

step. For this, the complete genomic DNA is extractedfrom an environmental sample, fragmented andcloned yielding the corresponding metagenome libra-ries.[17] These libraries can then be screened either forthe desired activity (with a function-based assay), orin a sequence-based approach in which genes areidentified based on homology to already describedenzymes, or the entire DNA is sequenced. Both ap-proaches have their advantages and disadvantages.The hit rate in the activity-based approach is usuallylower because the gene needs to be functionally ex-pressed in the host microorganism, which requirescorrect positioning of the metagenome sequence; pro-moter and codon usage must fit to the host and cor-rect post-translational modification of the enzymemust be ensured. If enzyme activity depends on sever-al subunits or requires additional proteins, e.g., forelectron transfer, the functional expression becomeseven more challenging. On the other hand, enzymesdiscovered by this activity-based approach are alreadycloned and accessible in a functional way. As thecosts and time required for DNA-sequencing substan-tially dropped by orders of magnitude in the past fewyears, entire metagenome libraries can now be se-quenced at reasonable budgets and timelines. This hasthe advantage that novel enzymes and even completepathways can be discovered quickly using bioinfor-matic tools. Together with sequence alignments fromdatabases, this approach facilitates the identificationof novel enzymes and the choice of an appropriate ex-pression system. Current reviews nicely summarize re-cently discovered enzymes using both approaches.[18]

In the activity-based approach the major limitationis the availability of reliable high-throughput assays.This is the reason why still most of the enzymes iso-lated from metagenome libraries are hydrolases.[19]

Although esterases, lipases and proteases form thelargest group, an increasing number of other enzymeclasses has been accessed from metagenome libraries,amongst them several industrially interesting polysac-charide-degrading enzymes like xylanases[20] , gluca-nases[21] , or cellulases[22] which are of interest for theproduction of bulk and fine chemicals from renewableresources or b-galactosidases for the food industry.[23]

2.2 Database Mining

Besides the classical enrichment cultivation and themetagenome approach, a third way to access new en-zymes is in silico screening. Protein sequences ob-tained from sequencing of single enzymes, entire ge-nomes or microbial consortia (i.e., from the SargassoSea[24]), are deposited in public databases and current-ly >12 million protein sequences are available (http://www.ncbi.nlm.nih.gov/RefSeq/). However, only a tinyfraction of all these sequences correspond to enzymes,

2194 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

which have been produced in the laboratory and theirbiochemical properties and function have been exper-imentally confirmed. For the vast majority, a possiblefunction is postulated by sequence comparison, butthis annotation can often be wrong or at least mis-leading, as shown in two recent examples from ourgroup for a monooxygenase[25] and transaminases.[4]

Discovering novel biocatalysts in public databases iseasier the more information about the desiredenzyme (class) exists. If several enzymes with a dis-tinct activity are already described in the literature,an alignment of their protein sequences together withexperimentally confirmed activities or specificities canreveal conserved regions within the sequences and asimple BLAST-search can yield new and useful bio-catalysts. There are numerous examples for the dis-covery of new enzymes in silico including nitrilases[26] ,a ligase[27] , an epoxide hydrolase[28] and a reductasefor the highly selective reduction yielding ethyl (S)-4-chloro-3-hydroxybutanoate with excellent opticalpurity (99% ee) at high substrate concentration(600 g/L).[29] By curtailing the sequences to besearched through, one is even able to increase thechances that the new catalysts fulfil additional crite-ria. Scanning the genomes of (hyper-)thermophilic or-ganisms, for example, will increase the chances to finda thermostable enzyme. Thus, Machielsen et al. identi-fied a thermostable ADH in the genome of Pyrococ-cus furiosus with a half-life of 130 min at 100 8C[30]

and Fraaije et al. discovered a thermostable Baeyer–Villiger monooxygenase with a promiscuous sulphuroxidizing activity.[31]

Many more examples can be found in recent re-views. Stewart, for example, illustrated the potentialof genome mining strategies exemplified for yeast de-hydrogenases[32] while Furuya et al. focused on aspectsof the genome mining approach that are relevant forthe discovery of novel P450 biocatalysts.[33] The iden-tification of biosynthetic gene clusters of indole alka-loids by this approach was summarized recently byLi.[34]

With increasing computing power and better soft-ware, scientists are now able to screen whole virtualmutant libraries in silico. Juhl et al. applied rationaldesign to the lipase CAL-B to expand the range ofcarboxylic acids converted, which is mainly limited tonon-branched fatty acids. After screening a total of2,400 mutants in silico with multiple docking experi-ments and expressing 11 variants in E. coli, theyfound one single mutant T138S with a 5-fold in-creased activity in the conversion of isononanoicacid.[35] A comparison between the power of in vivovs. in silico screening of mutants was recently beenmade by Barakat et al. as exemplified for the rationalredesign of an instable variant of Streptococcal pro-tein G.[36]

3 Tools for Protein Engineering

Two main strategies for protein engineering are usedto adjust enzyme properties towards the desired appli-cation: rational design and directed (molecular) evo-lution.[13d,37]

Rational design uses structural and mechanistic in-formation and molecular modelling for the predictionof changes in the protein structure in order to alter orinduce the desired properties. The advances in com-puter technology have helped in creating better pro-tein models to improve predictions for rationaldesign, but structure-activity relationships are still nottrivial. In directed evolution, mutant libraries are cre-ated by random changes, screened for the desiredproperty and the variants showing promising resultsare subjected to further rounds of evolution. Bothstrategies have their advantages and limitations.Nowadays, more researchers use combined methodsof these two strategies, dubbed focused-directed evo-lution or semi-rational design. Here, informationavailable from related protein structures, families andmutants already identified is combined and then usedfor targeted randomisation of certain areas of the pro-tein (Figure 1).

3.1 Methods for Directed Evolution

All methods used in directed evolution share acommon feature: they create random changes in theprotein sequence. Two different approaches are usedin directed evolution, recombining and non-recombin-ing. The first uses a set of related sequences, whichare randomly combined (e.g., gene shuffling[38])whereas the latter focuses on changing one single pro-tein sequence [e.g., error-prone polymerase chain re-action (epPCR[39]) or site saturation mutagenesis[40]] .Common methods for directed evolution of enzymesare covered in depth elsewhere.[41]

Directed evolution is a powerful tool in enzymedesign, but still with limitations. For example, therandom exchange of three amino acids in a 200 aminoacid protein leads to over 9 billion possible combina-tions. The screening of these huge libraries can hardlybe performed in microtiter plates, but there are meth-ods to access libraries that large, for example, fluores-cence-activated cell sorting (FACS) allows the screen-ing of up to 108 variants per day.[42] Furthermore, thesimultaneous measurement of pooled mutants andlater deconvolution of active variants can help toscreen more mutants per time frame. An interestingmethod using Monte-Carlo simulations has beenproven to successfully find active variants amongmany inactive clones,[43] but still a very sensitive activ-ity assay is needed.

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2195

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

Another way to reduce library size is to eliminateunfolded or unstable proteins. One strategy for thiselimination is to mimic the second major mechanismof biological evolution: genetic drift. Genetic drift isthe accumulation of random changes, while maintain-ing normal function. The assumption in this approachis that most variants with multiple substitutions areunstable. By screening for those that maintain theoriginal function, one ensures that the protein is stillproperly folded.[44]

3.2 Methods for Rational Protein Design

Rational design relies on protein structures to acquireinformation about possible and beneficial mutationsites. X-ray structures are not always at hand, thus aneed for more accurate models has arisen. Molecularmodelling uses force fields,[45] which are based on ex-perimental data to mimic protein behaviour as exactlyas possible,[6,8] while on the other hand using not toomuch computational power for large proteins. Hybridmodels are used to enhance modelling capabilitieswhile maintaining similar calculation speeds.[47] Thestructures are then evaluated for proposed mutationsand the promising changes identified in silico are thenactually inserted into the target protein by site-direct-ed mutagenesis.

3.3 Semi-Rational Design/Focused DirectedEvolution

Both strategies mentioned above have drawbacks thatlimit their practical use. Rational design can be re-stricted because quite often the information needed islacking or incomplete and extensive computationmight be necessary. On the other hand, directed evo-lution is limited by the size of the library to bescreened versus the availability of fast and reliablehigh-throughput screening methods (Figure 1, top).

A combination of both is often an alternative toovercome the limitations of each method. Focusing,for example, on certain enzyme areas that are knownto be linked to the desired property, can drasticallyreduce the screening effort. The combinatorial active-site saturation test (CAST) is a method pursuing thisstrategy. For this, a small set of amino acids in the vi-cinity of the active site is chosen and mutated ran-domly followed by screening for best hits. One singleCAST approach can result in an impressive enhance-ment of enzyme features as demonstrated for an ep-oxide hydrolase,[48] but this procedure is by definitionnot an evolutionary one. ISM, iterative saturation mu-tagenesis, introduces this evolutionary factor into theCASTing strategy. The best variants from a firstCASTing round serve as template for the next round.

Bearing in mind that mutations can have additive orcooperative effects, the order of positions to be mu-tated and screened might be crucial, leading to differ-ent pathways. Following all pathways is work inten-sive, as each step in a pathway includes a whole li-brary that needs to be screened.[49] Weinreich and co-workers tested all 120 paths to an improved variant ofa b-lactamase and found that only 18 paths improvedresistance to b-lactams at each stage.[50] Similarly, only8 paths increased the enantioselectivity of an epoxidehydrolase at each step (55 paths if one includes neu-tral mutations).[51]

Introduction of statistical methods can also greatlyimprove mutagenesis strategies. A computer programanalyses sets of sequence-activity data, preferentiallygained from different mutagenesis experiments andextracts information about the influence of individualmutations. The ProSAR-strategy (protein sequenceactivity relationships) developed by Fox et al.[52] wasused to guide the development of a halohydrin deha-logenase by the data obtained from different proteinengineering approaches (rational design, random mu-tagenesis, gene shuffling, saturation mutagenesis),sorting the single mutations into beneficial, neutraland deleterious ones and keeping only the beneficialmutations for further rounds of mutagenesis. The au-thors highlight this feature, as the problem withrandom directed evolution might be that certain mu-tations present in positive mutants will undoubtedlybe taken along into further mutation rounds, althoughthe mutation itself might not be beneficial. ProSARdetects and eliminates these. The final variant in thehalohydrin dehalogenase example contained 35 aminoacid substitutions leading to a 4,000-fold increasedvolumetric productivity in the synthesis of a Lipitor�

side chain.[52]

To increase enzyme stability against high tempera-ture or organic solvents, the “consensus approach”can be the key to success.[53] It is assumed that themost abundant amino acids at each position in a setof homologous enzymes contribute more than averageto protein stability. Comparison of sequences withinlarge enzyme families can identify conserved and dif-fering amino acids, which then guides the planning ofmutations to be introduced into the starting protein.This strategy was used to increase thermostability andorganic solvent tolerance of a glucose-1-dehydrogen-ase, raising the apparent melting point in the best var-iant to >80 8C and the half life in organic solvents by>2500-fold.[54] In another approach, a database(3DM) with 1,751 structurally related proteins fromthe a/b-hydrolase enzyme superfamily served as basisto create “small, but smart” libraries.[55] This 3DManalysis-based library was designed to cover onlyamino acids frequently occurring at a given positionin this superfamily. This concept enabled one to im-prove the enantioselectivity (Emutant =80, EWT =3.2) of

2196 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

an esterase from Pseudomonas fluorescens (PFE) bymutating four residues near the active site and alsoimproved the activity up to 240-fold.[56] This methodwas also used to increase the thermoactivity of PFEby 8 8C.[57] In another example, identification and mu-tagenesis of flexible residues using the B-FIT ap-proach in combination with ISM gave drastically im-proved thermoactivity (from 45 to 93 8C)[58] as well asenhanced stability in organic solvents for a lipasefrom Bacillus subtilis.[59]

4 Selected Examples

Enzymes such as lipases, esterases and oxidoreductas-es dominated the last two decades with many applica-tions in organic synthesis, especially targeting enantio-pure compounds. In the past decade several enzymesfrom other classes were discovered, characterised,cloned and engineered, to meet a range of challengesin organic synthesis and many are used already in in-dustry. The examples given below mostly refer toenzyme classes, which came into focus in the past fiveyears although they had been known and studiedsince several decades. We believe that the moderntools for protein discovery and engineering coveredabove were crucial to make them attractive biocata-lysts and to meet synthetic chemistry demands. Theexamples given below in more detail and surveyed inTable 1 provide an overview of enzymes improved byprotein engineering being important for organic syn-thesis.

4.1 Transaminases

Transaminases (TA, EC 2.6.1.x) catalyse the transferof an amino group from an amino donor (amine or a-amino acid) to an amino acceptor (ketone or a-ketoacid) utilising the cofactor pyridoxal 5’-phosphate(PLP). This reaction can be performed either as ki-netic resolution of a racemic amine (Scheme 1, left)or as an asymmetric synthesis starting from a proster-eogenic ketone (Scheme 1, right). By means of a com-bination of both strategies, deracemisation is also pos-sible: the ketone generated in a kinetic resolution cansubsequently be used as substrate in an asymmetricsynthesis employing a TA of opposite enantioprefer-ence.[119]

Whereas a-transaminases require the presence of acarboxylic acid group in the a-position to the keto oramine functionality and hence only allow the forma-tion of a-amino acids, amine transaminases (ATA)are much more versatile as they in principle acceptany ketone or amine.

Unfortunately, there is still no crystal structure ofan ATA available and hence rational protein design is

very difficult. However, directed evolution enabledthe identification of ATAs with strongly increasedspecific activity or reduced substrate and product in-hibition. In a very interesting directed evolutionstudy, Matcham and co-workers focused directly onavoiding product inhibition in the preparation of 2-amino-1-methoxypropane.[120] The wild-type transami-nase underwent significant product inhibition so thatwith 1M methoxyacetone and 1.5 equivalents of iso-propylamine only 12% conversion could be achieved,although this was not caused by an unfavourableequilibrium (the equilibrium constant was 8). Thus,the transaminase was improved by three rounds ofrandom mutagenesis and screening, and an enzymewas obtained from which, due to a higher Ki value forsubstrates and products, 65% conversion could be ob-tained without any need to shift the equilibrium. Anexplanation for the lowered inhibition constants maybe the increased KM values for both substrates. As anext step, five rounds of directed evolution were car-ried out yielding a variant with enhanced thermal andchemical stability so that the removal of acetone (gen-erated from the isopropylamine) at 50 8C under re-duced pressure was possible. The most effectivemutant afforded >99% optically pure (S)-2-amino-1-methoxypropane from 2 M methoxyacetone and 2.5 Misopropylamine at 93% yield after 7 h. Since the costof the biocatalyst was rather low (5% of the overallcost), it could be discarded after each productionbatch.

In a more recent study, a mesophilic (S)-selectiveamine transaminase from Arthrobacter species wasused to produce optically pure substituted aminotetra-line.[121] Although a high excess of isopropylamine asamino donor was used, the equilibrium still disfav-oured product formation. To allow for an efficientasymmetric synthesis, first, the catalyst had to be opti-mised to withstand high amounts of isopropylamineand reaction temperatures of 55 8C, so that acetonedistillation could shift the equilibrium, and, second,the specific activity for tetraline had to be increasedto reduce the reaction time. Five rounds of randommutagenesis resulted in a biocatalyst with a tempera-ture optimum at 55 8C and a 280-fold increased turn-over frequency while the Michaelis constant KM

stayed constant at about 10 mM. For a specific large-scale process the authors were able to reduce the re-action time from more than 48 h to less than 12 h andto reduce the enzyme loading 3-fold.

Apart from these successful examples for applyingdirected evolution to improve the performance oftransaminases, there are several examples for success-ful rational redesign of ATAs, too, although no crystalstructure of any amine transaminase has been solved(the first crystals could be obtained only recently).[122]

Cho et al. redesigned the substrate specificity of theextensively studied ATA from Vibrio fluvialis by a

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2197

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

Table 1. Survey of recent examples of enzymes optimized by protein engineering.[a]

Enzyme Methods Result and Comments Ref

ThermostabilityPapain RD Tmax for enzymatic activity increased by 15 8C for best triple mutant [60]

Xylanase RD Half-life at 50 8C and catalytic efficiency increased 15- and 1.3-times, respec-tively

[61]

Pyranose 2-oxidase SRD Melting temperature (TM) increased by 14 8C and improved catalytic proper-ties

[62]

Dioxygenase RD Optimum temperature and stability at basic pH increased by introduction ofdisulfide bond

[63]

Phytase epPCR 20% improvement of thermostability and 6–7 8C increased melting tempera-ture

[64]

Lipase RD Residual activity at 60 8C after 6 h increased from 20% to >70% by introduc-tion of disulfide bond

[65]

b-Keto ester reductase epPCR Residual activity at 45 8C after 2 h increased by single point mutation from 0to 64% accompanied by increased enantioselectivity

[66]

Parathion hydrolase RD TM increased by 3 8C and catalytic efficiency doubled by exchange of glycineto proline

[67]

b-Amylase RD Half-denaturation time at 60 8C increased from 6 to 35 min by the so called“ancestral mutation method“

[68]

Endoglucanase epPCR Triple mutant had 92% longer half-life at 60 8C on carboxymethyl cellulose [69]

Cellulase Hybrid. Half-life at 63 8C, pH 4.8 increased from 1.5 h to ~50 h [70]

Asparaginase StEP, SSM TM increased by 10 8C and half-life at 50 8C increased from 3 to 160 h [71]

Lipase B-FIT Increased thermolability with T50 value reduced from 72 8C to 36 8C [72]

General stabilityEndoxylanase RD Improved stability under acidic conditions (pH 3-4) [73]

Pancreatic lipase RM & RD Activity at acidic pH enhanced to 50% [74]

Mannanase RD & shuffling Activity and stability increased 3- and 7-fold, respectively [75]

Lipase ISSM Increased stability towards acetonitrile, dimethyl sulfoxide, dimethylforma-mide

[58]

Substrate specificityMonooxygenase RD 190-fold increased initial oxidation rate for 2-phenylethanol [76]

Peptide synthetase RD Specificity ratio Phe/Leu changed from 0.004 to 9.3 while increasing catalyticefficiency

[77]

DNA hemi-methylase RM & RD Specificity changed from GCGC to CGCG by drastically increasing the activ-ity towards (A)CGCG(T)

[78]

Glycine oxidase RD & SSM 15,000-fold increased specificity constant (glyphosate/glycine) and 210-foldincreased catalytic efficiency for glyphosate

[79]

Choline oxidase RD & DE Activity towards MTEA increased 5-fold and kcat/KM ACHTUNGTRENNUNG(MTEA)/kcat/KM-ACHTUNGTRENNUNG(choline) changed from 0.01 to 0.2

[80]

Styrene monooxyge-nase

RD Exchange of one amino acid led to acceptance of bulkier a-ethylstyrene [81]

Carboxymethylprolinesynthase

RD Production of substituted N-heterocycles with stereocontrolled enoate forma-tion

[82]

Transaldolase SSM Improved acceptance of non-phosphorylated substrates dihydroxyacetoneand glyceraldehyde

[83]

Xylose reductase SSM & epPCR Narrowing substrate acceptance to reduce unwanted formation of arabitolfrom arabinose while retaining xylose reductase activity

[84]

Toluene monooxyge-nase

Enantioselective oxidation of aromatic sulfides by point mutation yieldingmethyl phenyl sulfoxide with >98% ee (pro-S)

[85]

Galactose oxidase epPCR Catalyses oxidation of secondary alcohols [86]

Limonene epoxide hy-drolase

ISSM Catalyses the asymmetrisation of cyclopentene-oxide stereoselectively toform the (R,R)- or the (S,S)-diol

[87]

Lipase CAST Hydrolysis of a-substituted p-nitrophenyl esters (95–99% ee) [88]

Aniline dioxygenase SSM & epPCR Increased bioremediation activity up to 98-fold for aromatic amines [89]

Cofactor specificityAlcohol dehydrogen-ase

RD Improved preference for NADP(H) over NAD(H) [90]

Alcohol dehydrogen-ase

RD Single amino acid exchange broadened coenzyme acceptance from pureNAD(H) to both NAD(H) and NADP(H)

[91]

2198 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

single point mutation.[123] Several residues for poten-tial mutations were identified in a homology modelconstructed using the structure of a 2,2-dialkylglycinedecarboxylase (pdb-code: 1DGE). The wild-typeenzyme showed high specific activities towards aro-

matic amines, but had a more narrow substrate specif-icity towards aliphatic amines. The mutant W57Gwith the large binding pocket even widened showedsignificant changes in its substrate specificity withoutcompromising the original activities toward aromatic

Table 1. (Continued)

Enzyme Methods Result and Comments Ref

Lactate dehydrogenase RD & SSM Broadening of cofactor acceptance from NAD(H) to NADP(H) [92]

Putrescine oxidase RD Mutation leads to more stable binding of cofactor FAD over ADP [93]

Enantioselectivity/-preferenceLipase RD E value for bulky substrate increased from 5 (WT) to >200 [94]

Lipase CAST E=111 for kinetic resolution of an axially chiral allene, p-nitrophenyl 4-cy-clohexyl-2-methylbuta-2,3-dienoate

[95]

Diisopropyl fluoro-phosphatase

RD Reversed enantioselectivity and enhanced activity towards nerve agents [96]

N-Acetylneuraminicacid lyase

epPCR & SSM Two mutants with opposite stereoselectivity identified [97]

P450 monooxygenase ISSM Changed preference from 43% ee (S) for wt to 83% ee (R) [98]

Alcohol dehydrogen-ase

RD Reduction of benzylic and heteroarylic ketones with anti-prelog configurationby a single mutant

[99]

Ammonia lyase RD Stereoselectivity of 3-methylaspartate ammonia lyase enhanced by active sitemutation

[100]

Increased activityGlucose 6-oxidase RD 400-fold increased activity towards glucose [101]

Dehalogenase RD & SSM 32-fold higher activity towards 1,2,3-trichloropropane [102]

Cocaine hydrolase RD Computational design of mutant with 2,000-fold improved catalytic efficiency [103]

Serum paraoxonase RM Increased activity toward the coumarin analog of Sp-cyclosarin ~105-fold [104]

Aldolase epPCR, SSM& shuffling

60-fold higher kcat/KM for catalyzing the addition of pyruvate to d-erythrose4-phosphate to form DAHP

[105]

Phenylalanine ammo-nia lyase

DE 15-fold increased reaction rate [106]

Citramalate synthase epPCR Up to 22-fold higher production of 1-PrOH and 1-BuOH compared to wild-type

[107]

b-Glucuronidase epPCR & SSM 60-fold more active (kcat/KM) at pH 7.0 towards b-d-glucuronide [108]

Glycerol dehygroge-nase

RD & shuffling 26-fold improved activity for the oxidation of 1,3-butanediol to 4-hydroxy-2-butanone

[109]

Galactose oxidase SSM Mutations to optimize codon usage enable drastic enhancement of expression(>100,000 fold) and SSM in active site increased activity against galactose by60%

[110]

Nitrogenase RD Variants show increased production of H2 as side product during N2 fixation [111]

Promiscuous activityO-Glycosyltransferase RD O-glycosylating enzyme switched to C-glycosyltransferase [112]

Pyruvate decarboxy-lase

RD Aldehyde release vs. carboligation inverted (100-fold preference for thelatter)

[113]

Benzoylformate decar-boxylase

RD Single mutation converts benzaldehyde lyase into benzoylformate decarboxy-lase

[114]

Other propertiesb-Lactamase DE Introduction of allosteric regulation by metal ions [115]

P450 monooxygenase RD Construction of a hybrid metalloenzyme combining CYP450 and CytC withgood stability and activity

[116]

Histidine kinase Fusion proteins Combining histidine kinase with a light-oxygen-voltage photosensor domainto a light-regulated histidine kinase variant

[117]

b-Galactosidase SSM Removal of product inhibition (~10-times more galactose tolerated), thoughactivity and stability are reduced

[118]

[a] Abbreviations: (S)RD: (semi-)rational design; (I)SSM: (iterative) site-directed saturation mutagenesis; epPCR: error-prone PCR; DE: directed evolution; CAST: combinatorial active-site saturation test; MTEA: tris-(2-hydroxyethyl)-meth-ylammonium methyl sulfate; RM: random mutagenesis; wt: wild-type; Hybrid.: hybridization; B-FIT: B-factor iterativetest; TM =melting temperature.

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2199

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

amines and the enantioselectivity. The specific activitytowards 1-(1,2,3,4-tetrahydronaphthalen-2-yl)ethana-mine, for example, was increased 20-fold, towards 4-phenylbutan-1-amine 41-fold and towards isopropyla-mine 19-fold.

The ATA from Arthrobacter citreus has first beensubjected to directed evolution to increase thermosta-bility and isopropylamine acceptance resulting in themutant CNB05-01.[124] This mutant was later on ra-tionally redesigned by Svedendahl et al. for a furtherimprovement of the already high enantioselectivitytowards 4-fluorophenylacetone.[125] Mutant CNB05-01gave 22% yield with 98% ee in the asymmetric syn-thesis of (S)-1-(4-fluorophenyl)propan-2-amine. Withthe help of a homology model constructed using aclass III transaminase from Silicibacter pomeroyi(pdb-code: 3HMU) that showed only 27% sequenceidentity, they identified an active-site loop with sever-al residues close to the phenyl group of the substrateas well as to the phosphate group of the cofactor PLP.The single mutant Y331C actually showed a higherenantioselectivity with >99.5% ee and interestingly,mutant V328 A showed a reversed enantioselectivitywith 58% ee for the opposite enantiomer (R)-1-(4-flu-orophenyl)propan-2-amine while it retained its high(S)-selectivity for 1-(4-nitrophenyl)ethanamine.

The most extensive as well as successful redesign ofan ATA was recently reported by Savile et al. as men-tioned in the introduction.[11] They applied a combina-tion of different techniques starting with homologymodel-based rational design followed by ten roundsof directed evolution including random and site-satu-ration mutagenesis as well as DNA-shuffling. Thefinal 27 mutations of the best variant accumulatedmainly in or close to the active site and also at thesubunit interface of the homodimer. Starting fromATA-117 wild-type with hard to detect activity (0.2%conversion!) towards the target substrate, the finalvariant converted 200 g/L prositagliptin ketone to Si-tagliptin� with 99.95% ee at 92% yield. The biocata-lytic process not only reduced the total waste andeliminated all need for heavy metals, but even in-creased the overall yield by 10% and the productivity(kg/L per day) by 53% compared to the rhodium-cat-

alyzed process. Both routes have recently been com-pared.[126] Apart from the actual success in finding thedesired catalyst, the whole project brought in a hugevariety of mutants with a broad substrate range andvery good characteristics for application in organicsynthesis also useful for a-phenylethylamines withelectron-rich substituents and pyrrolidines.

We applied a different approach for the identifica-tion of so far unknown (R)-selective ATAs (R-ATAs).[127] During our project, the sequence of theATA-117 from Codexis as the only known isolated R-ATA was not available. Thus, we first had to hypothe-sise about possible relatives or ancestors of the targetenzyme in order to identify important residues re-sponsible for substrate coordination in aforesaid rela-tives. In the next step, comparable to rational design,we predicted key mutations for generating the desiredsubstrate specificity. Assuming that these mutationswould have generated at least a little of the desiredactivity, we then skipped the idea to go via directedevolution, but explored public databases for proteinswith these mutations, amongst other criteria. Our as-sumption was to circumvent the usually required sev-eral rounds of directed evolution for improving theimplanted activity as this should have occurred al-ready in nature during millions of years of naturalevolution. In the end, we identified 20 protein se-quences after filtering 5,000 sequences in public data-bases believed to be so far unknown (R)-selectiveATA, ordered the synthetic genes and expressedthem recombinantly in E. coli. The following charac-terisation resulted in ten enzymes with high R-ATAactivity (hit rate of 48%), and all 17 characterised mu-tants that could be expressed in E. coli showed thedesired activity and especially (R)-enantiopreference.Seven of these enzymes were recently applied suc-cessfully in the asymmetric synthesis of various (R)-amines having >99% ee at >99% yield.[128]

4.2 Enoate Reductases

Enoate reductases (EC 1.3.1.31) catalyse the asym-metric reduction of activated C=C bonds. The activat-

Scheme 1. Strategies for the synthesis of optically active amines using amine transaminases (ATA). In a kinetic resolution(left), the ATA converts one of the amine enantiomers to the corresponding ketone (�50% yield). In an asymmetric synthe-sis (right), a prostereogenic ketone is aminated enantioselectively, yielding directly the optically active amine at a theoreticalyield of 100%.

2200 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

ing group is usually an electron-withdrawing group(EWG), which includes aldehyde, ketone, carboxylicacid, carboxylic ester, anhydride, lactone, imide ornitro groups. This biocatalytic asymmetric reduction isuseful as it can produce two adjacent sp3 chiral cen-tres from a prostereogenic activated alkene in a singlestep. Furthermore, the reduction occurs by the net ad-dition of hydrogen (H2) in a trans-stereospecificmanner, which provides a synthetic route for asym-metric enzymatic trans-hydrogenation of activatedC=C bonds.

The synthetic application of this reaction has ex-plored a number of enoate reductases from variousyeasts and microbial sources, with the first enzymepurified from Saccharomyces pastorianus (OYE1).Toogood et al.[129] and St�rmer et al.[130] described var-ious other sources of enoate reductase in their recentreviews. Many of the enoate reductases were charac-terised and cloned to express the heterologous pro-teins, which were subsequently used for biocatalysis.Examples include the reduction of substituted andnon-substituted a,b-unsaturated cyclic and acyclic al-dehydes, ketones, imides, nitroalkenes, carboxylicacids, and esters, maleimides, cyclic and acyclicenones, terpenoids, cyanides, nitrate esters and alsothe nitro group reduction of nitroaromatics (TNT),glycerol trinitrate (GTN), and pentaerythritol tetrani-trate (PETN).

In spite of the broad substrate scope of OYE1 (Sac-charomyces pastorianus), it lacked higher activity to-wards the reduction of bulky 3-alkyl-substituted 2-cy-clohexenones. In order to increase substrate selectivi-ty, Padhi et al.[131] carried out site-saturation mutagen-esis at position 116 of OYE1 and discovered the var-iants W116I and W116F, which showed opposite

stereopreference for certain substrates. OYE1-W116Ireduced (R)- and (S)-carvone to enantiomers com-pared to diastereomers formed by the wild-typeenzyme (Scheme 2). The inverted stereoselectivitywas due to the substrate binding in a flipped orienta-tion in the active site followed by net trans H2 addi-tion. OYE1-W116I also showed opposite stereoprefer-ence for the C=C-reduction of (R)-perillaldehyde toproduce trans-dihydroperillaldehyde compared to thecis-form made by wild-type OYE1. However, anothervariant OYE1-W116F reduced neral to (R)-citronellalcompared to (S)-citronellal formed by the wild-type.

In the first report of directed evolution on theenoate-reductase YqjM from Bacillus subtilis, a ho-mologue of the old yellow enzyme,[132] a library of var-iants was generated, which showed increased activity,enhanced and inverted enantioselectivity in the reduc-tion of a number of cyclic enones. Compared to thelow activity of wild-type YqjM towards the reductionof 3-methylcyclohexenone [4 mU/mg, (R)-specific],the saturation mutagenesis library produced variantsthat showed higher activity and opposite enantioselec-tivity. Also broader selectivity for the C=C-reductionof other cyclic enones was found (Scheme 3, Table 2).

The site saturation mutagenesis on PETN reductaseat eight single positions with degenerate oligonucleo-tides (NNK) produced mutant W102F, which gave op-posite enantiopreference in the reduction of a-methyl-trans-cinnamaldehyde,[133] with 62% ee for the(R)-2-methyl-3-phenylpropionaldehyde compared to13% ee (S)-enantiomer formed by the wild-type.Mutant T26S also gave opposite enantiopreferencefor (R)-2-phenyl-1-nitropropane, 44% ee for the re-duction of (E)-2-phenyl-1-nitropropene, compared to(S)-product, 37% ee (Scheme 4). Although reversible

Scheme 2. OYE1-wild-type and W116I form products with opposite stereochemistry from (S)-carvone. With (R)-carvoneboth enzymes produced the same diastereomer.

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2201

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

isomerisation of (E)-2-phenyl-1-nitropropene to the(Z)-form could produce the (S)-product, the T26Svariant increased the relative population of the (E)-substrate by removing the possible clash between thesubstrate phenyl and the threonine and henceswitched the stereochemistry of the product.

Site-saturation mutagenesis on PETN reductase atH181 and H184 identified several single mutantsH184N/A/K/R. All four of them increased the enan-tioselectivity towards the reduction of (E)-2-phenyl-1-nitropropene (substrate of the second reaction ofScheme 4).[134] While the wild-type PETN reductasegave 42% ee, the variants produced 85–94% ee of the(S)-2-phenyl-1-nitropropane. In the library also var-iants were found, which switched the reactivity to pre-dominantly catalyze nitro-reduction. Variants H181A/C/N showed dominating oxime products by the reduc-tion of the nitro group in 52–75% yield as comparedto only 10% product by wild-type PETN reductase.

4.3 Esterases for Tertiary Alcohol Synthesis

Lipases and esterases are known to catalyse the hy-drolysis of tertiary alcohol esters to produce chiraltertiary alcohols, but with little selectivity.[135] In orderto improve the enantioselectivity, Henke et al. engi-neered an esterase from Bacillus subtilis (BS2) andthe first single mutant G105A gave higher enantiose-lectivity, that is, E=19 compared to E=3 for wild-type in the hydrolysis of 3-phenylbut-1-yn-3-yl ace-tate.[136] In a further medium engineering experiment,autohydrolysis of the substrate was suppressed by ad-dition of the water-miscible solvent DMSO, which in-creased E to 56 for variant BS2-G105A. Substrateand protein engineering guided by molecular model-ACHTUNGTRENNUNGling showed that BS2-G105A could catalyse the hy-drolysis of 4,4,4-trifluoro-3-phenylbut-1-yn-3-yl ace-tate with E>100 compared to E=42 for wild-type.[137] The high substrate selectivity and E value ofthe BS2-G105A variant were consistent with the pre-viously established concept that a GGG(A)X motif inhydrolases is a key determinant for enzymatic activityon tertiary alcohol substrates. Another variant (BS2-E188D) also showed high selectivity in the hydrolysisof 3-phenylbut-1-yn-3-yl acetate (E=46) and of 4,4,4-trifluoro-3-phenylbut-1-yn-3-yl acetate with E>100.Both of the above mentioned BS2 variants accepted a

Table 2. Reduction of cyclic enones catalyzed by YqjM and variants.[a]

Substrate n, R Enzyme Conversion/% eeP (Conf.) Substrate n, R Enzyme Conversion/% eeP (Conf.)

1, ethyl wild-type n.c.[b] 1, COOMe wild-type 85/99 (R)1, ethyl C26D/I69T 100/99 (R) 1, COOMe C26D/I69T 92/99 (R)1, ethyl C26G/A60V 9/96 (S) 1, COOMe C26G 100/98 (S)1, isopropyl wild-type n.c.[b] 0, methyl wild-type 1/n.d.[b]

1, isopropyl C26D/I69T 33/99 (R) 0, methyl C26D/I69T 5/50 (R)1, isopropyl C26G/A60V 2/91 (S) 0, methyl C26D 36/93 (S)1, n-butyl wild-type n.c.[b] 0, COOMe wild-type 100/96 (R)1, n-butyl C26D/I69T 92/99 (R) 0, COOMe I69T 100/99(R)

0, COOMe C26G 100/93 (S)

[a] n and R: see Scheme 3.[b] n.c.: no conversion, n.d.: not determined.

Scheme 4. Opposite enantiopreference of PETN variants towards C=C-bond reduction.[134]

Scheme 3. YqjM-catalyzed reduction of cyclic enones.

2202 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

broad range of trifluoromethyl arylaliphatic tertiaryalcohol acetates with E>100.[138]

Next, Bartsch et al. performed a focused directedevolution based on simultaneous saturation mutagen-esis at positions E188, A190 and M193 using the de-generate NNK codons with the aim to invert theenantiopreference of BS2.[48] After screening 2,800colonies with a high-throughput assay[139] , they identi-fied a key variant (BS2-E188D-M193C), whichshowed inverted enantioselectivity in the hydrolysisof 4,4,4-trifluoro-3-phenylbut-1-yn-3-yl acetate andother substrates with different aryl substituents. Thisvariant showed high (S)-selectivity, ES =64 comparedto ER =42 by wild-type BS2. Moreover, the engi-neered enzyme also catalysed the kinetic resolution ofa range of other trifluoromethyl derivatives with highenantiomeric ratio (ES up to >100) and hence itsbroad substrate scope was proven (Scheme 5).

Analysis of the role of E188 on enantioselectivityrevealed that increasing bulkiness of the residue dras-tically changed the enantiopreference from (R) to (S)(Scheme 6). High opposite enantiopreference wasonly found for the double mutant, as the singlemutant E188W showed ES =26 and M193C gave ER =16, which emphasises the importance of synergistic ef-fects of the residues.

4.4 Monoamine Oxidases

Monoamine oxidases (EC 1.4.3.4, MAO) catalyse theoxidative deamination of monoamines leading to theformation of the corresponding aldehydes. Mechanis-tic investigation of this reaction, however, revealedthat the reaction proceeds through the formation ofthe aldimine, which could also be isolated.[140] Turneret al. exploited this enzymatic transformation in thederacemisation of racemic amines to synthesise thecorresponding enantiopure products. The deracemisa-tion process involves stereoinversion of one enantio-mer to the other by repeated cycles of MAO-cata-lysed stereoselective oxidation to the imine combinedwith a non-selective chemical reduction of the imineto the amine, which notably works in a one-pot reac-tion (Scheme 7).

Using directed evolution Alexeeva et al.[141] im-proved the selectivity of the enzyme from Aspergillus

Scheme 5. Kinetic resolution of several arylaliphatic tertiary alcohol acetates by engineered BS2 enzymes showing oppositeenantiopreference.

Scheme 6. Effect of the size of amino acid substitutions atposition 188 in BS2 on enantioselectivity in the hydrolysis ofarylaliphatic tertiary alcohol acetates. Small residues lead topreferred conversion of the (R)-enantiomer while bulky res-idues gave (S)-selectivity.

Scheme 7. Deracemisation of a-methylbenzylamine (a-MBA) using an (S)-selective monoamine oxidase coupledwith ammonia-borane reduction of the intermediate imineto form racemic amine. In the first round, 50% (S)-amine isoxidised to the imine yielding 25% (R)-amine and 25% (S)-amine after chemical reduction, the latter being oxidisedagain. After 7–8 rounds, the racemic starting material iscompletely converted into optically pure (R)-a-MBA.

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2203

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

niger (MAO-N) towards the oxidation of a-MBA.The MAO-N mutant N336 showed 47-fold higher ac-tivity towards (S)-a-MBA and 5.8-fold higher selectiv-ity towards (S)-a-MBA over (R)-a-MBA. Combiningthis engineered protein with ammonia-borane as a re-ducing agent, (S)-a-MBA could be synthesised in77% yield and 93% ee from its racemic substrate. Themutant also showed broad substrate scope and highenantioselectivity towards a range of primaryamines.[142] Directed evolution of MAO-N producedanother variant, which catalysed the deracemisationof a structurally diverse range of secondaryamines.[143] Similarly, a further mutant enabled the(S)-selective oxidation of tertiary amines and hencecould be used in the deracemisation process for thesynthesis of enantiopure cyclic tertiary amines.[144]

One variant (MAO-N-5) was used in the chemo-enzy-matic synthesis of the alkaloid (+)-crispine and its de-rivatives (Scheme 8).[145]

MAO-N-5 also catalysed the desymmetrisation of3,4-substituted meso-pyrolidines to produce the corre-sponding D1-pyrolines.[146] The optically pure pyrolinesfrom the above process underwent one-pot condensa-tion in the presence of a carboxylic acid and an isocy-anide to afford chiral substituted prolyl peptides. Thisprovided a chemo-enzymatic route for the stereose-lective synthesis of highly functionalised, opticallypure 3,4-substituted prolyl peptides, a striking exam-

ple of a Ugi-type 3-component reaction(Scheme 9).[147]

4.5 Dehalogenases

Dehalogenases are enzymes which catalyse the hy-drolysis of a carbon-halogen bond by an SN2 mecha-nism. Haloalkane dehalogenases (HLD) and haloacid dehalogenases (HAD) act on haloalkane andhalo acid, respectively.[148] However, halohydrin deha-logenases (HHDH) act only on vicinal halo alcoholsto produce an epoxide (Scheme 10). Apart fromsimple chlorinated, brominated and iodinated alkanes,the HLDs also catalyse the hydrolysis of halogenatedalkenes, cycloalkanes, alcohols, carboxylic acids,esters, ethers, epoxides, amides and nitriles to producea range of corresponding halogen hydrolysed productsand hence provide access to various chiral intermedi-ates.[149] The halohydrin dehalogenases not only cata-lyse the formation of epoxides, but also the reversiblereaction to open the epoxide ring with different nu-cleophiles such as cyanide, nitro and azide[150] to pro-duce the corresponding multifunctional products.

A haloalkane dehalogenase from Rhodococcus rho-dochrous (DhaA) catalysed the hydrolysis of 1,2,3-tri-chloropropane (TCP) into 2,3-dichloropropanol.However, the half-life (t1/2) of the wild-type enzymewas only 11 min at 55 8C. Gray et al.[151] carried outsite saturation mutagenesis on DhaA and found a li-

Scheme 8. Chemo-enzymatic synthesis of (+)-crispine usingMAO-N-5-catalysed deracemisation.

Scheme 9. MAO-N catalysed oxidative desymmetrisation of meso-pyrolidines to optically pure D1-pyrolines and their subse-quent use in the Ugi-3-component reaction (Ugi-3-CR).

Scheme 10. Typical examples of three different types of re-actions catalysed by dehalogenases. DhlA and DhlB are hal-oalkane and haloacid dehalogenases from Xanthobacter au-totrophicus. HheC is a halohydrin dehalogenase from Agro-bacterium AD1.

2204 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

brary of variants with increased t1/2. The single sitevariants had half-lives from 5–45 h. A combined var-iant with five mutations had a half-life ~3000-foldhigher than wild-type DhaA. Combination of eightsingle-site mutations on DhaA increased the half-lifeby another factor of ten.

Random mutagenesis of the haloalkane dehaloge-nase from Rhodococcus sp. m15-3 (DhaA) followedby screening on eosin-methylene blue agar platesidentified a new variant that catalysed the dehaloge-nation of TCP into 2,3-dichloro-1-propanol nearlyeight times more efficient than the wild-type dehalo-genase (Scheme 11).[152] The resulting mutant C176Y/Y273F showed kcat/KM = 280M�1 s�1 compared to36M�1 s�1 by wild-type DhaA.

Further protein engineering on this enzyme pro-duced variants which showed up to 32-fold higher ac-tivity than the wild-type toward the degradation ofthe anthropogenic compound TCP.[102] Identificationof key residues in access tunnels connecting theburied active site with bulk solvent by rational designfollowed by random mutagenesis and directed evolu-tion produced improved variants. Three of themshowed up to 32-fold higher activity in the hydrolysisof TCP compared to wild-type DhaA (kcat =0.08 s�1).Table 3 shows the steady-state kinetic parameters ofthe variants and wild-type DhaA for this halide hy-drolysis.

DbjA, a haloalkane dehalogenase from Bradyrhi-zobium japonicum USDA110, showed E=132 and209 for the halide hydrolysis of 2-bromopentane andmethyl 2-bromobutyrate, respectively, with the (R)-enantiomer as predominant product. In comparisonto other HLDs, DbjA features an additional segment(140-His-Thr-Glu-Val-Ala-Glu-Glu-146) located onthe protein surface between the core a/b-hydrolaseand cap domains.[153] Deletion of this segment(DbjAD) altered the enantioselectivity for the twosubstrates mentioned above. The DbjAD and

DbjAD+ H189A variants showed E=33 and 100 inhalide hydrolysis of 2-bromopentane and E=290 and245 for methyl 2-bromobutyrate, respectively.

HheC was found to be unstable under oxidisingconditions, which could be prevented by addition ofb-mercaptoethanol. Tang et al.[154] carried out site di-rected mutagenesis on HheC and showed that the var-iant C153S had improved stability and also opticalpurity of the product was increased from 72% ee(wild-type) to 93% ee in the kinetic resolution of 2-chloro-1-phenylethanol.

In the case of HheC-catalysed dehalogenation of vi-cinal halo alcohols to epoxides, halide release wasfound to be the rate-limiting step during the catalyticcycle. Tang et al.[155] performed site-directed mutagen-esis and identified variants Y187F and W249F withhigher activity in the epoxidation of 1,3-dichloro-2-propanol and p-nitro-2-bromo-1-phenylethanol(Scheme 12). Mutant Y187F had a two-fold increasedkcat (18.7�1.1 s�1) and kcat/KM was raised to 1.6 �106 M�1 s�1 (wild-type 9.3 �105 M�1 s�1) for 1,3-di-chloro-2-propanol. Variant Y249F had increasedenantioselectivity with E=900 compared to 150 forthe wild-type and an improved kcat/KM.

In an impressive example, researchers at Codexis(USA) developed a method using engineered HheCto synthesise the side-chain of the cholesterol-lower-ing drug, atorvastatin (Lipitor�).[52] Starting fromethyl (S)-4-chloro-3-hydroxybutyrate (ECHB), the(S)-epoxide is formed first followed by formation ofethyl (R)-4-cyano-3-hydroxybutyrate (HN), the keybuilding block for the atorvastatin side chain(Scheme 13). (S)-ECHB and its corresponding epox-ide are poor substrates for wild-type HheC. Engineer-ing based on the ProSAR-driven directed evolution(see Section 3.3) led to variants which allowed a one-pot conversion yielding the product with 99.5% purityand >99.9% ee.

Scheme 11. Conversion of TCP into 2,3-dichloropropane-1-ol by a DhaA variant.

Table 3. Steady state kinetic parameters of wild-type DhaA and its variants.

Variant kcat [s�1] KM [mM] kcat/KM [M�1 s�1]

Wild-type 0.04�0.01 1.0�0.2 40�13I135L-W141F-C176Y-V245F-L246I-Y273F 0.55�0.04 1.2�0.2 458�83I135V-C176Y-V245F-L246I-Y273F 1.02�0.06 1.1�0.1 927�100I135F-C176Y-V245F-L246I-Y273F 1.26�0.07 1.2�0.1 1050�105

Scheme 12. HheC-catalysed conversion of a halohydrin intoan epoxide.

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2205

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

4.6 Aldolases

Aldolases (E.C. 4.1.2.x) constitute a group of C�Cbond forming lyases. The nucleophilic donor is usuallyinvariable and aldolases are grouped by their depend-ency on the donors dihydroxyacetone phosphate(DHAP), dihydroxyacetone (DHA), pyruvate/phos-phoenol pyruvate, acetaldehyde or glycine/alanine.On the other hand, they accept a broad range of ac-ceptor aldehydes enabling the formation of versatileproducts such as rare or non-natural sugars and theirderivatives. They also allow the introduction of differ-ent heteroatoms (amino, nitro, azido, phospho, sulfo,chloro, fluoro or deoxy groups). Especially attractiveis the fact that two stereocentres are formed duringthe aldolase reaction and that the absolute configura-tion of the product can be influenced by the choice ofthe enzyme. For instance, four different DHAP-de-pendent aldolases enable selective formation of allfour possible diastereomers from DHAP and glyceral-dehyde 3-phosphates[156] as shown in a review.[157] Toavoid the costly and rather unstable DHAP, variantswere created which accept DHA as alternative

donor.[158] Among the pyruvate-dependent aldolases,N-acetylneuraminic acid aldolase (NeuAc) is interest-ing for pharmaceutical application as it converts N-acetylmannosamine and pyruvate to d-sialic acid, abuilding block for the synthesis of neuraminidase in-hibitors for the treatment of flu.[159] The stereospeci-ficity of NeuAc could be inverted by directed evolu-tion using epPCR to form l-sialic acid in contrast tothe naturally occurring d-enantiomer.[160] This is im-portant because the unnatural l-form is more stableagainst enzymatic degradation (Scheme 14).

The group of Berry managed to modify the stereo-specificity of a tagatose 1,6-bisphosphate aldolase bygene shuffling providing interesting mechanistic in-sights.[161] The most prominent example for an indus-trially applied aldolase is the acetaldehyde-dependent2-deoxy-d-ribose-5-phosphate aldolase (DERA),which is used for the preparation of (3R,5S)-6-chloro-2,4,6-trideoxyhexose, a precursor in the synthesis ofthe side-chain of drugs like atorvastatin (Scheme 15).The major obstacles for this process were the narrowsubstrate range and the low stability of the aldolaseunder process conditions. This could be overcome for

Scheme 13. A HheC variant catalysed the highly selective formation of ethyl (R)-4-cyano-3-hydroxybutyrate from the (S)-chloro derivative, which can be subsequently used in the preparation of the atorvastatin side chain.

Scheme 14. Inversion of the stereospecificity of N-acetylneuraminic acid aldolase (NeuAc) towards sialic acid by directedevolution.[160]

2206 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

the E. coli DERA by directed evolution yielding avariant with much higher resistance and productivitytowards the chloroacetaldehyde substrate.[162] In an al-ternative approach, metagenome libraries werescreened for novel aldolases exhibiting higher activityand stability against chloroacetaldehyde.[163] The iso-lated enzyme had <30% sequence identity towardsthe E. coli DERA and would thus hardly have beenfound by sequence-based screening approaches. Incombination with process improvements, a highly effi-cient, scalable and enantioselective process could bedeveloped.

4.7 Cytochrome P450-Monooxygenases

P450-monooxygenases display a versatile superfamilyof heme-dependent oxygenases. Sequence databasescurrently contain >10,000 sequences of P450-like en-zymes from all domains of life, although just a smallfraction of them has been functionally expressed andcharacterised so far.[164] P450s catalyse a broad rangeof reactions including carbon hydroxylation, epoxida-tion of double bonds, C�C bond cleavage, ring forma-tion and oxidation of heteroatoms. The substratesrange from short-chain to long-chain alkanes/alkenesover fatty acids and alkaloids to complex moleculeslike steroids, amongst them various pharmaceuticallyinteresting compounds. In spite of their great poten-tial, there are only few examples for the industrial ap-plication of isolated P450 enzymes. Many P450s aredifficult to handle as they are often membrane bound,have low stability and activity or may consist of sever-al subunits, required for an efficient electron transfer.To circumvent these problems, industrial applicationsare often based on whole-cell systems, which alsoenable multi-step reactions. For the application of iso-lated enzymes, a great potential lies in the subfamilyCYP102, a group of soluble P450-monooxygenases.Its most prominent member is P450-BM3(CYP102A1) from Bacillus megaterium, in which oxy-genase and reductase subunits are fused to one singlepolypeptide chain. In numerous studies this enzymehas been subjected to engineering aiming to increase

activity, regioselectivity, thermal stability or to changecofactor-dependency or substrate preference. PositionF87 was identified in several studies to be a hotspotmediating substrate specificity and regioselectivity, forexample, in the rational design study by the Pleissgroup, where mutations at positions 87 and 328 ena-bled the conversion of the four terpene substrates ger-anylacetone, nerylacetone, (4R)-(+)-limonene and(+)-valencene (Scheme 16).[165] The triple mutantA74G/F87V/L188Q accepts polycyclic aromatic hy-drocarbons as substrates.[166] In another computermodelling-guided approach P450-BM3 was engi-neered to catalyse the oxidation of amorphadiene toartemisinic-11S,12-epoxide, a precursor of the anti-malaria drug artemisinin, which was achieved by theintroduction of four mutations into P450-BM3.[167]

In a molecular modelling-based study, the specifici-ty against alkoxyresorufin substrates could be alteredby exchange of five amino acids in the human P4501A1.[168] The study demonstrated that in the case ofP450 enzymes the prediction of the influence of cer-tain amino acid positions often works very well dueto the great structural information available for thisenzyme class. The thermal stability of P450-BM3could be significantly increased by a chimera-formingapproach by the Schmid group through combining theCYP102A1 with the reductase subunit of a sulfite re-ductase from the thermophilic strain Geobacillusstearothermophilus.[169] In several other studies thenatural fusion protein P450-BM3 was taken as inspi-

Scheme 15. Improvement of 2-deoxy-d-ribose-5-phosphate aldolase (DERA) towards higher stability and activity againstchloroacetaldehyde to form intermediates for the production of statin drugs like atorvastatin (Lipitor�).[162–163]

Scheme 16. P450-BM3 was modified by rational design toaccept the four terpenes, (4R)-(+)-limonene, (+)-valencene,geranylacetone and nerylacetone as substrates.[165]

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2207

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

ration to create unnatural fusion proteins combiningdifferent subunits, for example, in the study by Sher-man et al., in which the reductase domain of Rhodo-coccus sp. NCIMB 9784 P450 was combined with theoxygenase subunit of PikC cytochrome P450 fromStreptomyces venezuelae. The resulting fusion proteinshowed enhanced electron transfer efficiency and cat-alytic activity in the synthesis of pikromycin and me-thymycin.[170] The same reductase subunit was used toconstruct a vector, which allows the fusion of differentoxygenase subunits with this reductase and the systemwas validated with 24 different heme domains.[171] Be-sides P450-BM3, the camphor-hydroxylating P450cam

from Pseudomonas putida (CYP101) is another well-studied monooxygenase. For example, mutation of po-sition Y96 lead to increased oxidation of naphthaleneand pyrene by 1–2 orders of magnitude.[172] In arecent semi-rational approach, the substrate specifici-ty of P450cam was changed towards diphenylme-thane.[173] In another study the oxidation products of(+)-a-pinene were changed.[174]

Monooxygenases can also be applied for epoxida-tion of substrates. In P450-BM3 an epoxidising activi-ty could be created by saturation mutagenesis.[175] Be-sides the P450 type monooxygenases, there are alsoflavine-dependent styrene monooxygenases catalysingepoxidations.[176] The activity of the enzyme fromPseudomonas putida CA-3 could be improved byerror-prone PCR yielding an improved epoxidationactivity towards styrene and indene.[177]

4.8 Baeyer–Villiger Monooxygenases

Baeyer–Villiger monooxygenases (BVMO) catalysethe enzymatic counterpart to the chemical Baeyer–Villiger reaction introducing an oxygen atom next toa carbonyl function yielding esters or lactones. Thesubstrate spectrum ranges from linear and bulky ke-tones to complex molecules like steroids and sesqui-terpenoids. The most intensely studied BVMO is thecyclohexanone monooxygenase (CHMO) from Acine-tobacter calcoaceticus NCIMB 9871.[178] While onlyfew BMVOs were available for a long time, thenumber of enzymes increased immensely by the iden-tification of a conserved sequence motif to find genesin sequenced genomes. The phenylacetone monooxy-

genase (PAMO) from Thermobifida fusca, active onaromatic substrates, was isolated in this way.[31] Thegroup of Grogan cloned 29 new BVMOs by the se-quence-based approach from the genomes of the acti-nomycetes M. tuberculosis[179] and Rhodococcusjostii.[180] Remarkably, the conserved region used toidentify the genes turned out not to be in the activesite, but within a surface loop in the connection of theFAD and the NADPH domain.

The discovery of the first BVMO structure of thethermostable PAMO[181] in 2004 was a major break-through to enable rational protein engineering. Fiveyears later, a second BVMO structure of a novelCHMO in complex with NADP was resolved.[182]

Based on the PAMO structure, the Reetz groupevolved this BVMO in several studies yielding a var-iant, which now converts 2-phenylcyclohexanoneenantioselectively.[183] The exchange of two prolineresidues turned PAMO into a variant which acceptssubstrates commonly converted by CHMO-typeBVMOs.[184] Besides these rational design studies onPAMO, there are further examples in which the prop-erties of BVMOs were changed by engineering.Kirschner et al. were able to enhance the enantiose-lectivity of a BVMO from Pseudomonas fluorescensby error-prone PCR for the kinetic resolution of b-hy-droxy ketones (Scheme 17).[25b,185] The CHMO fromA. calcoaceticus was also redesigned in several studies.In a combined approach of epPCR and saturationmutagenesis, it was possible to enhance and invert theenantioselectivity in the conversion of 4-hydroxycy-clohexanone (Scheme 18).[186] The same group modi-fied CHMO to oxidise prostereogenic thioethersyielding chiral sulfoxides (Scheme 19).[187] By ex-change of methionine and cysteine residues on theprotein surface, it was possible to improve the stabili-ty of CHMO.[188]

One bottleneck for the application of BVMO on alarge scale is the dependency on the cofactor. MostBVMOs depend on NADPH as cofactor, which ismore expensive than NADH. Therefore the identifi-cation of the residues essential for the discriminationbetween both cofactors was useful for further studiesand the acceptance of NADH could be increased for4-hydroxyacetophenone monooxygenase andCHMO.[189] A very interesting concept to address thecofactor issue was developed by Hollmann et al.

Scheme 17. Directed evolution of a BVMO increased activity and enantioselectivity in the kinetic resolution of b-hydroxyketones yielding the monoacetate of a 1,2-diol as major product and the b-hydroxy acid methyl ester as minor by-prod-uct.[185]

2208 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

based on reduction of the flavin cofactor by (sun)lightin the presence of a mediator.[190] A completely differ-ent way to increase the catalytic efficiency of theBVMO was taken by the group of Fraaije fusing theBVMO with an NADPH regenerating phosphite de-hydrogenase yielding a self-sufficient enzyme.[191] Fur-ther examples for the application and engineering ofBVMOs can be found in comprehensive recent re-views.[192]

5 Biocatalysts for New Reactions

Although numerous enzymes have been identified,some important organic reactions seem to have nonatural enzymatic counterpart. In recent years signifi-cant advances have been made towards the design oftailor-made enzymes from scratch. Although the rela-tionship between an enzyme�s structure and its activi-ty is still poorly understood, examples have been pub-lished using extensive computational power to designenzymes that catalyse reactions not observed innature. However, these proteins still do not reach thehigh activity of natural biocatalysts.

The principle strategy in creating a new enzyme isbased on the fact that an enzyme can catalyse a chem-ical transformation because, amongst other reasons, itstabilises the transition state of the reaction. First ofall the basic mechanism of the reaction must beknown in detail, then, focusing on the transition state,amino acid residues that can stabilise or even interactwith the substrate are positioned in the relevantorder, thus forming the putative active site. This stepalready requires sophisticated computing using quan-tum mechanical (QM) calculations to find the rightamino acids and position them as exactly as possible.The next step consists of finding a natural scaffoldthat can be used to accommodate these residues intheir proper positioning. The group of Baker was ableto develop and enhance algorithms to perform thistask.[193]

Using these methods, it was possible to design aretro-aldolase, a Diels–Alderase and a Kemp-elimi-nase. The retro-aldolase reaction is catalysed by alysine residue via a Schiff base or an imine intermedi-ate. Since the proposed mechanism consisted of sever-al steps, the above-mentioned general method neededto be extended for the design of active sites that arecompatible with multiple transition states. In total, 32designs belonging to four different motifs were foundto show retro-aldolase activity with rate enhance-ments of up to 104 over the non-catalysed reaction,but still with low turnover rates of about 9 �10�3 min�1.[194]

The Kemp eliminase was proposed to use a generalbase, being either aspartate or glutamate or a histi-dine-aspartate/glutamate dyad. Of the 59 selected var-iants 39 were aspartate or glutamate dependent andtwenty had a dyad in the active site and eight designsshowed activity. A promising variant relied on a His-Asp dyad and showed a kcat of 2 �10�2 s�1. By randommutagenesis this could be further enhanced to1.3 s�1.[195] Another variant was further optimised byvarious techniques ending up with mutants showing aturnover number of 5–8 s�1 and catalytic efficienciesof 5 �104 M�1 s�1.[196]

Designing an enzyme capable of catalysing theDiels–Alder reaction was supposed to be quite chal-lenging, since bimolecular bond formation is involved.A carbonyl oxygen of a glutamine or asparagine resi-due was used to form a hydrogen bond to the dieneintermediate, whereas the dienophile was hydrogen-bonded by the hydroxy group of a serine, threonineor tyrosine. With these assumptions 1019 possibleactive site variants were computed and about 106

could be matched into protein scaffolds. These var-iants were further evaluated and in the end fifty var-iants were obtained as soluble proteins. Only two de-signs had Diels–Alderase activity and after furtheroptimisation by mutagenesis, a variant gave a rate ofkcat/kuncat =89 which is an effective turnover rate of

Scheme 18. Enantioselective conversion of 4-hydroxycyclo-hexanone to the corresponding lactones using engineeredCHMO mutants.[186a]

Scheme 19. CHMO variants created that were able to con-vert prostereogenic thioethers into chiral sulfoxides.[187]

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2209

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

about 2 s�1. Interestingly, the most active variant wasstrictly stereospecific, as intended by the design, andformed >97% of the desired stereoisomer.[197]

Discovering new enzymatic activity is for sure themost difficult challenge in protein design and engi-neering and hence it is much more common and rela-tively easy to alter the biocatalytic properties of anexisting enzyme.[198]

6 Conclusions

The exponential growth of the application of biocata-lysts since the last two decades was mostly due to sig-nificant achievements by several interdisciplinarybranches which all became very important for bioca-talysis, such as molecular biology, bioinformatics andmostly protein engineering with all its various ap-proaches and tools. This has substantially changed theway to discover or design biocatalysts compared totraditional methods. The many successful examplescovered in this review clearly underline that computa-tional methods to alter enzymes to catalyse new reac-tions or to change their selectivities has alreadyreached a mature level and it is to be expected thatthey will become even more important and easier touse in the future. De novo in silico design of novel en-zymatic activities is still far away from creating bio-catalysts with sufficiently high activities and proper-ties required for application in processes. However,within just a few years this technique was developedand thus it is likely that in the next decade our under-standing of the underlying principles to design moreefficient biocatalysts by computational means will bebetter and further so far undiscovered reactions willbe catalysed by designed enzymes. Protein engineer-ing also can contribute to the development of inte-grated processes where chemo- and enzyme catalysisare combined. A recent special issue of ChemCat-ACHTUNGTRENNUNGChem[199] already demonstrates that this is feasibleand has the advantage to create cleaner and greenerprocesses. Further, the tools for protein engineeringof biocatalysts covered in this review must not be con-sidered solely in a context of the use of isolated en-zymes as more and more processes are now being es-tablished using metabolic engineering; that is, the in-corporation of multi-step enzyme catalysis in an engi-neered microorganisms. An example is the synthesisof optically pure amino acids in the hydantoinase pro-cess, which was initially based on the separate use ofimmobilised hydantoinase, carbamoylase and race-mase in packed-bed column reactors and nowadayscan be performed using engineered E. coli cells.[200]

Most recently, the production of fatty acid alkylesters[201] or 1,4-butanediol[202] with whole cells was re-ported, both processes rely beside sophisticated meta-bolic engineering efforts on the tools for enzyme dis-

covery and protein engineering described in thisreview.

Acknowledgements

We are grateful to the Alexander von Humboldt Foundationfor a stipendium to SKP, the Deutsche BundesstiftungUmwelt for a stipendium to AH and the State Mecklenburg-Vorpommern for a stipendium to GAB.

References

[1] L. Rosenthaler, Biochem. Z. 1908, 14, 238–253.[2] D. R. Tolan, A. B. Amsden, S. D. Putney, M. S. Urdea,

E. E. Penhoet, J. Biol. Chem. 1984, 259, 1127–1131.[3] a) G. DeSantis, Z. Zhu, W. A. Greenberg, M. J. Burk,

J. Am. Chem. Soc. 2002, 124, 9024–9025; b) D. E.Robertson, J. A. Chaplin, G. DeSantis, M. Podar, M.Madden, E. Chi, T. Richardson, A. Milan, M. Miller,D. P. Weiner, K. Wong, J. McQuaid, B. Farwell, L. A.Preston, X. Tan, M. A. Snead, M. Keller, E. Mathur,P. L. Kretz, M. J. Burk, J. M. Short, Appl. Environ. Mi-crobiol. 2004, 70, 2429–2436.

[4] M. Hçhne, S. Sch�tzle, H. Jochens, K. Robins, U. T.Bornscheuer, Nat. Chem. Biol. 2010, 6, 807–813.

[5] L. A. Kelley, M. J. E. Sternberg, Nat. Protoc. 2009, 4,363–371.

[6] a) K. Arnold, L. Bordoli, J. Kopp, T. Schwede, Bioin-formatics 2006, 22, 195–201; b) F. Kiefer, K. Arnold,M. K�nzli, L. Bordoli, T. Schwede, Nucleic Acids Res.2009, 37, D387-D392; c) M. C. Peitsch, Bio/Technolo-gy 1995, 13, 658–660.

[7] D. E. Kim, D. Chivian, D. Baker, Nucleic Acids Res.2004, 32, W526-W531.

[8] E. Krieger, K. Joo, J. Lee, J. Lee, S. Raman, J. Thomp-son, M. Tyka, D. Baker, K. Karplus, Proteins Struct.Funct. Bioinf. 2009, 77, 114–122.

[9] S. F. Altschul, W. Gish, W. Miller, E. W. Myers, D. J.Lipman, J. Mol. Biol. 1990, 215, 403–410.

[10] a) M. Goujon, H. McWilliam, W. Li, F. Valentin, S.Squizzato, J. Paern, R. Lopez, Nucleic Acids Res.2010, 38 (Suppl.), W695–699; b) M. A. Larkin, G.Blackshields, N. P. Brown, R. Chenna, P. A. McGetti-gan, H. McWilliam, F. Valentin, I. M. Wallace, A.Wilm, R. Lopez, J. D. Thompson, T. J. Gibson, D. G.Higgins, Bioinformatics 2007, 23, 2947–2948.

[11] C. K. Savile, J. M. Janey, E. C. Mundorff, J. C. Moore,S. Tam, W. R. Jarvis, J. C. Colbeck, A. Krebber, F. J.Fleitz, J. Brands, P. N. Devine, G. W. Huisman, G. J.Hughes, Science 2010, 329, 305–309.

[12] a) A. Liese, K. Seelbach, C. Wandrey (Eds.), IndustrialBiotransformations, 2nd edn., Wiley-VCH, Weinheim,2006 ; b) R. N. Patel (Ed.), Biocatalysis in the Pharma-ceutical and Biotechnology Industries, CRC Press,Boca Raton, 2006 ; c) K. Buchholz, V. Kasche, U. T.Bornscheuer, Biocatalysts and Enzyme Technology,Wiley-VCH, Weinheim, 2005 ; d) K. Faber, Biotrans-formations in Organic Chemistry, 5th edn., Springer,Berlin, 2004 ; e) K. Drauz, H. Waldmann (Eds.),

2210 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

Enzyme Catalysis in Organic Synthesis, Vols. 1–3, 2ndedn., Wiley-VCH, Weinheim, 2002.

[13] a) M. T. Reetz, Angew. Chem. 2011, 123, 144–182;Angew. Chem. Int. Ed. 2011, 50, 138–174; b) A. S.Bommarius, J. K. Blum, M. J. Abrahamson, Curr.Opin. Chem. Biol. 2011, 15, 194–200; c) G. A. Stroh-meier, H. Pichler, O. May, M. Gruber-Khadjawi,Chem. Rev. 2011, DOI: 10.1021/cr100386u; d) R. J.Kazlauskas, U. T. Bornscheuer, Nat. Chem. Biol. 2009,5, 526–529.

[14] a) H. Griengl, H. Schwab, M. Fechter, Trends Biotech-nol. 2000, 18, 252–256; b) D. V. Johnson, A. A. Zabe-linskaja-Mackova, H. Griengl, Curr. Opin. Chem.Biol. 2000, 4, 103- 109.

[15] a) A. Musidlowska, S. Lange, U. T. Bornscheuer,Angew. Chem. 2001, 113, 2934–2936; Angew. Chem.Int. Ed. 2001, 40, 2851–2853; b) A. Hummel, E. Br�se-haber, D. Bçttcher, H. Trauthwein, K. Doderer, U. T.Bornscheuer, Angew. Chem. 2007, 119, 8644–8646;Angew. Chem. Int. Ed. 2007, 46, 8492–8494.

[16] a) R. I. Amann, W. Ludwig, K.-H. Schleifer, Micro-biol. Rev. 1995, 59, 143–169; b) J. T. Staley, A. Konop-ka, Annu. Rev. Microbiol. 1985, 39, 321–346; c) N. R.Pace, Science 1997, 276, 734–740; d) P. Hugenholtz,B. M. Goebel, N. R. Pace, J. Bacteriol. 1998, 180,4765–4774.

[17] W. R. Streit, R. Daniel, (Eds.), Metagenomics, Spring-er Protocols, Humana Press, New York, 2010.

[18] a) T. Uchiyama, K. Miyazaki, Curr. Opin. Biotechnol.2009, 20, 616–622; b) L. Fernndez-Arrojo, M.-E.Guazzaroni, N. Lpez-Cort�s, A. Beloqui, M. Ferrer,Curr. Opin. Biotechnol. 2010, 21, 725–733.

[19] a) E. Y. Yu, M. A. Kwon, M. Lee, J. Y. Oh, J. E. Choi,J. Y. Lee, B. K. Song, D. H. Hahm, J. K. Song, Appl.Microbiol. Biotechnol. 2011, 90, 573–581; b) S. Y.Park, H. J. Shin, G. J. Kim, J. Microbiol. 2011, 49, 7–14; c) Y. Okamura, T. Kimura, H. Yokouchi, M. Men-eses-Osorio, M. Katoh, T. Matsunaga, H. Takeyama,Mar. Biotechnol. 2010, 12, 395–402; d) R. P. Chen,L. Z. Guo, H. Y. Dang, World J. Microbiol. Biotech-nol. 2011, 27, 431–441; e) S. Bayer, A. Kunert, M.Ballschmiter, T. Greiner-Stoeffele, J. Mol. Microbiol.Biotechnol. 2010, 18, 181–187.

[20] a) G. Z. Wang, H. Y. Luo, Y. R. Wang, H. Q. Huang,P. J. Shi, P. L. Yang, K. Meng, Y. G. Bai, B. Yao, Biore-sour. Technol. 2011, 102, 3330–3336; b) X. C. Mo,C. L. Chen, H. Pang, Y. Feng, J. X. Feng, Appl. Micro-biol. Biotechnol. 2010, 87, 2137–2146.

[21] J. A. Liu, W. D. Liu, X. L. Zhao, W. J. Shen, H. Cao,Z. L. Cui, Appl. Microbiol. Biotechnol. 2011, 89,1083–1092.

[22] E. J. Kwon, Y. S. Jeong, Y. H. Kim, S. K. Kim, H. B.Na, J. Kim, H. D. Yun, H. Kim, J. Korean Soc. Appl.Biol. 2010, 53, 702–708.

[23] K. Wang, G. Li, S. Q. Yu, C. T. Zhang, Y. H. A. Liu,Appl. Microbiol. Biotechnol. 2010, 88, 155–165.

[24] J. C. Venter, K. Remington, J. F. Heidelberg, A. L.Halpern, D. Rusch, J. A. Eisen, D. Wu, I. Paulsen,K. E. Nelson, W. Nelson, D. E. Fouts, S. Levy, A. H.Knap, M. W. Lomas, K. Nealson, O. White, J. Peter-son, J. Hoffman, R. Parsons, H. Baden-Tillson, C.

Pfannkoch, Y.-H. Rogers, H. O. Smith, Science 2004,304, 66–74.

[25] a) A. Kirschner, J. Altenbuchner, U. T. Bornscheuer,Appl. Microbiol. Biotechnol. 2007, 73, 1065–1072;b) A. Kirschner, U. T. Bornscheuer, Angew. Chem.2006, 118, 7161–7163; Angew. Chem. Int. Ed. 2006, 45,7004–7006.

[26] O. Kaplan, K. Bezouska, A. Malandra, A. Vesel, A.Petr�ckov, J. Felsberg, A. Ringelov, V. Kren, L.Mart�nkov, Biotechnol. Lett. 2011, 33, 309–312.

[27] K. Kino, A. Noguchi, T. Arai, M. Yagasaki, J. Biosci.Bioeng. 2010, 110, 39–41.

[28] B. van Loo, J. Kingma, M. Arand, M. G. Wubbolts,D. B. Janssen, Appl. Environ. Microbiol. 2006, 72,2905–2917.

[29] L.-J. Wang, C.-X. Li, Y. Ni, J. Zhang, X. Liu, J.-H. Xu,Bioresour. Technol. 2011, 102, 7023–7028.

[30] R. Machielsen, A. R. Uria, S. W. M. Kengen, J. vander Oost, Appl. Environ. Microbiol. 2006, 72, 233–238.

[31] M. W. Fraaije, J. Wu, D. P. H. M. Heuts, E. W. vanHellemond, J. H. L. Spelberg, D. B. Janssen, Appl. Mi-crobiol. Biotechnol. 2005, 66, 393–400.

[32] J. D. Stewart, in: Advances in Microbiology, Vol. 59,(Eds.: S. S. Allen I. Laskin, M. G. Geoffrey), Academ-ic Press, San Diego, 2006, pp 31–52.

[33] T. Furuya, K. Kino, Appl. Microbiol. Biotechnol. 2010,86, 991–1002.

[34] S. M. Li, J. Antibiot. 2011, 64, 45–49.[35] P. B. Juhl, K. Doderer, F. Hollmann, O. Thum, J.

Pleiss, J. Biotechnol. 2010, 150, 474–480.[36] N. H. Barakat, N. H. Barakat, J. J. Love, Protein Eng.

Des. Sel. 2010, 23, 799–807.[37] S. Lutz, U. T. Bornscheuer, (Eds.), Protein Engineering

Handbook, Wiley-VCH, Weinheim, 2009.[38] W. P. C. Stemmer, Nature 1994, 370, 389–391.[39] R. C. Cadwell, G. F. Joyce, PCR Meth. Appl. 1992, 2,

28–33.[40] L. W. Chiang, in: In Vitro Mutagenesis Protocols, (Ed.:

M. K. Trower), Humana Press, Totowa, NJ, 1996,pp 311–321.

[41] a) F. H. Arnold, G. Georgiou, (Eds.), Directed Evolu-tion Library Creation: Methods and Protocols, Vol.231, Humana Press, Totawa, 2003 ; b) F. H. Arnold, G.Georgiou, (Eds.), Directed Enzyme Evolution: Screen-ing and Selection Methods, Vol. 230, Humana Press,Totawa, 2003 ; c) S. Brakmann, A. Schwienhorst,(Eds.), Evolutionary Methods in Biotechnology:Clever Tricks for Directed Evolution, Wiley-VCH,Weinheim, 2004.

[42] G. Y. Yang, S. G. Withers, ChemBioChem 2009, 10,2704–2715.

[43] a) K. M. Polizzi, M. Parikh, C. U. Spencer, I. Matsu-mura, J. H. Lee, M. J. Realff, A. S. Bommarius, Bio-technol. Prog. 2006, 22, 961–967; b) K. M. Polizzi,C. U. Spencer, A. Dubey, I. Matsumura, J. H. Lee,M. J. Realff, A. S. Bommarius, J. Biomol. Screening2005, 10, 856–864.

[44] a) J. D. Bloom, F. H. Arnold, Proc. Natl. Acad. Sci.USA 2009, 106, 9995–10000; b) R. D. Gupta, D. S.Tawfik, Nat. Methods 2008, 5, 939–942; c) S. G. Peisa-jovich, D. S. Tawfik, Nat. Methods 2007, 4, 991–994.

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2211

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

[45] E. Krieger, G. Koraimann, G. Vriend, Proteins Struct.Funct. Gen. 2002, 47, 393–402.

[46] E. Krieger, T. Darden, S. B. Nabuurs, A. Finkelstein,G. Vriend, Proteins Struct. Funct. Bioinf. 2004, 57,678–683.

[47] a) M. Mladenovic, K. Junold, R. F. Fink, W. Thiel, T.Schirmeister, B. Engels, J. Phys. Chem. B 2008, 112,5458–5469; b) H. M. Senn, W. Thiel, Atomistic Ap-proaches in Modern Biology: From Quantum Chemis-try to Molecular Simulations, 2007, 268, 173–290.

[48] S. Bartsch, R. Kourist, U. T. Bornscheuer, Angew.Chem. 2008, 120, 1531–1534; Angew. Chem. Int. Ed.2008, 47, 1508–1511.

[49] a) M. T. Reetz, L. W. Wang, M. Bocola, Angew. Chem.2006, 118, 1258–1263; Angew. Chem. Int. Ed. 2006, 45,1236–1241; b) M. T. Reetz, M. Bocola, J. D. Carbal-leira, D. X. Zha, A. Vogel, Angew. Chem. 2005, 117,4264–4268; Angew. Chem. Int. Ed. 2005, 44, 4192–4196.

[50] D. M. Weinreich, N. F. Delaney, M. A. Depristo, D. L.Hartl, Science 2006, 312, 111–114.

[51] M. T. Reetz, J. Sanchis, ChemBioChem 2008, 9, 2260–2267.

[52] R. J. Fox, S. C. Davis, E. C. Mundorff, L. M. Newman,V. Gavrilovic, S. K. Ma, L. M. Chung, C. Ching, S.Tam, S. Muley, J. Grate, J. Gruber, J. C. Whitman,R. A. Sheldon, G. W. Huisman, Nat. Biotechnol. 2007,25, 338–344.

[53] M. Lehmann, L. Pasamontes, S. F. Lassen, M. Wyss,Biochim. Biophys. Acta 2000, 1543, 408–415.

[54] E. Vazquez-Figueroa, V. Yeh, J. M. Broering, J. F.Chaparro-Riggers, A. S. Bommarius, Protein Eng. Des.Sel. 2008, 21, 673–680.

[55] a) R. K. Kuipers, H. J. Joosten, W. J. van Berkel, N. G.Leferink, E. Rooijen, E. Ittmann, F. van Zimmeren,H. Jochens, U. Bornscheuer, G. Vriend, V. A. dosSantos, P. J. Schaap, Proteins 2010, 78, 2101–2113;b) R. Kourist, H. Jochens, S. Bartsch, R. Kuipers, S. K.Padhi, M. Gall, D. Bçttcher, H. J. Joosten, U. T. Born-scheuer, ChemBioChem 2010, 11, 1635–1643.

[56] H. Jochens, U. T. Bornscheuer, ChemBioChem 2010,11, 1861–1866.

[57] H. Jochens, D. Aerts, U. T. Bornscheuer, Protein Eng.Des. Sel. 2010, 23, 903–909.

[58] M. T. Reetz, P. Soni, L. Fernandez, Y. Gumulya, J. D.Carballeira, Chem. Commun. 2010, 46, 8657–8658.

[59] M. T. Reetz, J. D Carballeira, A. Vogel, Angew.Chem. 2006, 118, 7909–7915; Angew. Chem. Int. Ed.2006, 45, 7745–7751.

[60] D. Choudhury, S. Biswas, S. Roy, J. K. Dattagupta,Protein Eng. Des. Sel. 2010, 23, 457–467.

[61] J. C. Joo, S. Pohkrel, S. P. Pack, Y. J. Yoo, J. Biotech-nol. 2010, 146, 31–39.

[62] O. Spadiut, C. Leitner, C. Salaheddin, B. Varga, B. G.Vertessy, T. C. Tan, C. Divne, D. Haltrich, FEBS J.2009, 276, 776–792.

[63] J. Wei, Y. Zhou, T. Xu, B. Lu, Appl. Biochem. Bio-technol. 2010, 162, 116–126.

[64] M.-S. Kim, X. G. Lei, Appl. Microbiol. Biotechnol.2008, 79, 69–75.

[65] Z. L. Han, S. Y. Han, S. P. Zheng, Y. Lin, Appl. Micro-biol. Biotechnol. 2009, 85, 117–126.

[66] H. Asako, M. Shimizu, N. Itoh, Appl. Microbiol. Bio-technol. 2009, 84, 397–405.

[67] J. A. Tian, P. Wang, S. Gao, X. Y. Chu, N. F. Wu, Y. L.Fan, FEBS J. 2010, 277, 4901–4908.

[68] K. Yamashiro, S. Yokobori, S. Koikeda, A. Yamagishi,Protein Eng. Des. Sel. 2010, 23, 519–528.

[69] W. J. Liu, X. Z. Zhang, Z. M. Zhang, Y. H. P. Zhang,Appl. Environ. Microbiol. 2010, 76, 4914–4917.

[70] P. Heinzelman, C. D. Snow, I. Wu, C. Nguyen, A. Vil-lalobos, S. Govindarajan, J. Minshull, F. H. Arnold,Proc. Natl. Acad. Sci. USA 2009, 106, 5610–5615.

[71] G. A. Kotzia, N. E. Labrou, FEBS J. 2009, 276, 1750–1761.

[72] M. T. Reetz, P. Soni, L. Fernandez, Biotechnol.Bioeng. 2009, 102, 1712–1717.

[73] T. Belien, I. J. Joye, J. A. Delcour, C. M. Courtin, Pro-tein Eng. Des. Sel. 2009, 22, 587–596.

[74] D. Y. Colin, P. Deprez-Beauclair, N. Silva, L. Infantes,B. Kerfelec, Protein Eng. Des. Sel. 2010, 23, 365–373.

[75] H. Y. Zhou, H. Y. Pan, L. Q. Rao, Y. Y. Wu, Appl.Biochem. Biotechnol. 2011, 163, 186–194.

[76] M. Brouk, Y. Nov, A. Fishman, Appl. Environ. Micro-biol. 2010, 76, 6397–6403.

[77] C. Y. Chen, I. Georgiev, A. C. Anderson, B. R.Donald, Proc. Natl. Acad. Sci. USA 2009, 106, 3764–3769.

[78] R. Gerasimaite, G. Vilkaitis, S. Klimasauskas, NucleicAcids Res. 2009, 37, 7332–7341.

[79] M. Pedotti, E. Rosini, G. Molla, T. Moschetti, C.Savino, B. Vallone, L. Pollegioni, J. Biol. Chem. 2009,284, 36415–36423.

[80] D. Ribitsch, S. Winkler, K. Gruber, W. Karl, E. Wehr-schutz-Sigl, I. Eiteljorg, P. Schratl, P. Remler, R. Stehr,C. Bessler, N. Mussmann, K. Sauter, K. H. Maurer, H.Schwab, Appl. Microbiol. Biotechnol. 2010, 87, 1743–1752.

[81] A. A. Qaed, H. Lin, D. F. Tang, Z. L. Wu, Biotechnol.Lett. 2011, 33, 611–616.

[82] R. B. Hamed, J. R. Gomez-Castellanos, A. Thalham-mer, D. Harding, C. Ducho, T. D. W. Claridge, C. J.Schofield, Nat. Chem. 2011, 3, 365–371.

[83] S. Schneider, M. Gutierrez, T. Sandalova, G. Schneid-er, P. Clapes, G. A. Sprenger, A. K. Samland, Chem-BioChem 2010, 11, 681–690.

[84] N. U. Nair, H. M. Zhao, ChemBioChem 2008, 9, 1213–1215.

[85] R. Feingersch, J. Shainsky, T. K. Wood, A. Fishman,Appl. Environ. Microbiol. 2008, 74, 1555–1566.

[86] F. Escalettes, N. J. Turner, ChemBioChem 2008, 9,857–860.

[87] H. B. Zheng, M. T. Reetz, J. Am. Chem. Soc. 2010,132, 15744–15751.

[88] K. Engstroem, J. Nyhlen, A. G. Sandstroem, J.-E.Baeckvall, J. Am. Chem. Soc. 2010, 132, 7038–7042.

[89] E. L. Ang, J. P. Obbard, H. Zhao, Appl. Microbiol.Biotechnol. 2009, 81, 1063–1070.

[90] E. Campbell, I. R. Wheeldon, S. Banta, Biotechnol.Bioeng. 2010, 107, 763–774.

[91] N. Stiti, I. O. Adewale, J. Petersen, D. Bartels, H. H.Kirch, Biochem. J. 2011, 434, 459–471.

[92] N. Richter, A. Zienert, W. Hummel, Eng. Life Sci.2011, 11, 26–36.

2212 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

[93] M. M. Kopacz, S. Rovida, E. van Duijn, M. W. Fraaije,A. Mattevi, Biochemistry 2011, 50, 4209–4217.

[94] T. Ema, S. Kamata, M. Takeda, Y. Nakano, T. Sakai,Chem. Commun. 2010, 46, 5440–5442.

[95] J. D. Carballeira, P. Krumlinde, M. Bocola, A. Vogel,M. T. Reetz, J.-E. Backvall, Chem. Commun. 2007,1913–1915.

[96] M. Melzer, J. C. H. Chen, A. Heidenreich, J. Gab, M.Koller, K. Kehe, M. M. Blum, J. Am. Chem. Soc.2009, 131, 17226–17232.

[97] G. J. Williams, T. Woodhall, L. M. Farnsworth, A.Nelson, A. Berry, J. Am. Chem. Soc. 2006, 128, 16238–16247.

[98] W. L. Tang, Z. Li, H. Zhao, Chem. Commun. 2010, 46,5461–5463.

[99] M. M. Musa, N. Lott, M. Laivenieks, L. Watanabe, C.Vieille, R. S. Phillips, ChemCatChem 2009, 1, 89–93.

[100] H. Raj, B. Weiner, V. P. Veetil, C. R. Reis, W. J. Quax,D. B. Janssen, B. L. Feringa, G. J. Poelarends, Chem-BioChem 2009, 10, 2236–2245.

[101] S. M. Lippow, T. S. Moon, S. Basu, S. H. Yoon, X. Li,B. A. Chapman, K. Robison, D. Lipovsek, K. L. J.Prather, Chem. Biol. 2010, 17, 1306–1315.

[102] M. Pavlova, M. Klvana, Z. Prokop, R. Chaloupkova,P. Banas, M. Otyepka, R. C. Wade, M. Tsuda, Y.Nagata, J. Damborsky, Nat. Chem. Biol. 2009, 5, 727–733.

[103] F. Zheng, W. C. Yang, M. C. Ko, J. J. Liu, H. Cho,D. Q. Gao, M. Tong, H. H. Tai, J. H. Woods, C. G.Zhan, J. Am. Chem. Soc. 2008, 130, 12148–12155.

[104] R. D. Gupta, M. Goldsmith, Y. Ashani, Y. Simo, G.Mullokandov, H. Bar, M. Ben-David, H. Leader, R.Margalit, I. Silman, J. L. Sussman, D. S. Tawfik, Nat.Chem. Biol. 2011, 7, 120–125.

[105] N. Ran, J. W. Frost, J. Am. Chem. Soc. 2007, 129,6130–6139.

[106] S. Bartsch, U. T. Bornscheuer, Protein Eng. Des. Sel.2010, 23, 929–933.

[107] S. Atsumi, J. C. Liao, Appl. Environ. Microbiol. 2008,74, 7802–7808.

[108] K.-C. Chen, C.-H. Wu, C.-Y. Chang, W.-C. Lu, Q.Tseng, Z. M. Prijovich, W. Schechinger, Y.-C. Liaw,Y.-L. Leu, S. R. Roffler, Chem. Biol. 2008, 15, 1277–1286.

[109] H. Zhang, G. T. Lountos, C. B. Ching, R. Jiang, Appl.Microbiol. Biotechnol. 2010, 88, 117–124.

[110] S. E. Deacon, M. J. McPherson, ChemBioChem 2011,12, 593–601.

[111] H. Masukawa, K. Inoue, H. Sakurai, C. P. Wolk, R. P.Hausinger, Appl. Environ. Microbiol. 2011, 76, 6741–6750.

[112] J. Harle, S. Gunther, B. Lauinger, M. Weber, B. Kam-merer, D. L. Zechel, A. Luzhetskyy, A. Bechthold,Chem. Biol. 2011, 18, 520–530.

[113] D. Meyer, L. Walter, G. Kolter, M. Pohl, M. Muller,K. Tittmann, J. Am. Chem. Soc. 2011, 133, 3609–3616.

[114] G. S. Brandt, M. M. Kneen, G. A. Petsko, D. Ringe,M. J. McLeish, J. Am. Chem. Soc. 2001, 123, 438–439.

[115] V. Mathieu, J. Fastrez, P. Soumillion, Protein Eng.Des. Sel. 2010, 23, 699–709.

[116] T. Ying, F. Zhong, Z.-H. Wang, W. Li, X. Tan, Z.-X.Huang, ChemBioChem 2011, 12, 707–710.

[117] A. Mçglich, R. A. Ayers, K. Moffat, J. Mol. Biol.2009, 385, 1433–1444.

[118] X. Hu, S. Robin, S. O�Connell, G. Walsh, J. G. Wall,Appl. Microbiol. Biotechnol. 2010, 87, 1773–1782.

[119] D. Koszelewski, D. Clay, D. Rozzell, W. Kroutil, Eur.J. Org. Chem. 2009, 2289–2292.

[120] G. B. Matcham, L. Mohit; W. Lang, C. Lewis, Nelson,A. Wang, W. Wu, Chimia 1999, 53, 584–589.

[121] A. R. Martin, R. DiSanto, I. Plotnikov, S. Kamat, D.Shonnard, S. Pannuri, Biochem. Eng. J. 2007, 37, 246–255.

[122] T.-H. Jang, B. Kim, O. K. Park, J. Y. Bae, B.-G. Kim,H. Yun, H. H. Park, Acta Crystallogr. F 2010, 66, 923–925.

[123] B.-K. Cho, H.-Y. Park, J.-H. Seo, J. Kim, T.-J. Kang,B.-S. Lee, B.-G. Kim, Biotechnol. Bioeng. 2008, 99,275–284.

[124] S. Pannuri, S. V. Kamat, A. R. M. Garcia, Patent WO2006/063336 A2, 2006.

[125] M. Svedendahl, C. Branneby, L. Lindberg, P. Ber-glund, ChemCatChem 2010, 2, 976–980.

[126] A. A. Desai, Angew. Chem. 2011, 123, 2018–2020;Angew. Chem. Int. Ed. 2011, 50, 1974–1976.

[127] M. Hçhne, S. Sch�tzle, H. Jochens, K. Robins, U. T.Bornscheuer, Nat. Chem. Biol. 2010, 6, 807–813.

[128] S. Sch�tzle, F. Steffen-Munsberg, M. Hçhne, K.Robins, U. T. Bornscheuer, Adv. Synth. Catal. 2011,353, 2439–2445.

[129] H. S. Toogood, J. M. Gardiner, N. S. Scrutton, Chem-CatChem 2010, 2, 892–914.

[130] R. Stuermer, B. Hauer, M. Hall, K. Faber, Curr. Opin.Chem. Biol. 2007, 11, 203–213.

[131] S. K. Padhi, D. J. Bougioukou, J. D. Stewart, J. Am.Chem. Soc. 2009, 131, 3271–3280.

[132] D. J. Bougioukou, S. Kille, A. Taglieber, M. T. Reetz,Adv. Synth. Catal. 2009, 351, 3287–3305.

[133] M. E. Hulley, H. S. Toogood, A. Fryszkowska, D.Mansell, G. M. Stephens, J. M. Gardiner, N. S. Scrut-ton, ChemBioChem 2010, 11, 2433–2447.

[134] H. S. Toogood, A. Fryszkowska, M. Hulley, M.Sakuma, D. Mansell, G. M. Stephens, J. M. Gardiner,N. S. Scrutton, ChemBioChem 2011, 12, 738–749.

[135] a) R. Kourist, U. T. Bornscheuer, Appl. Microbiol.Biotechnol. 2011, 91, 505–517; b) R. Kourist, P. Dom�-nguez de Mar�a, U. T. Bornscheuer, ChemBioChem2008, 9, 491–498.

[136] E. Henke, U. T. Bornscheuer, R. D. Schmid, J. Pleiss,ChemBioChem 2003, 4, 485–493.

[137] B. Heinze, R. Kourist, L. Fransson, K. Hult, U. T.Bornscheuer, Protein Eng. Des. Sel. 2007, 20, 125–131.

[138] R. Kourist, S. Bartsch, U. T. Bornscheuer, Adv. Synth.Catal. 2007, 349, 1393–1398.

[139] M. Baumann, R. St�rmer, U. T. Bornscheuer, Angew.Chem. 2001, 113, 4329–4333; Angew. Chem. Int. Ed.2001, 40, 4201–4204.

[140] J. C. G. Woo, R. B. Silverman, J. Am. Chem. Soc.1995, 117, 1663–1664.

[141] M. Alexeeva, A. Enright, M. J. Dawson, M. Mahmou-dian, N. J. Turner, Angew. Chem. 2002, 114, 3309–3312; Angew. Chem. Int. Ed. 2002, 41, 3177–3180.

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2213

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

[142] R. Carr, M. Alexeeva, A. Enright, T. S. C. Eve, M. J.Dawson, N. J. Turner, Angew. Chem. 2003, 115, 4955–4958; Angew. Chem. Int. Ed. 2003, 42, 4807–4810.

[143] R. Carr, M. Alexeeva, M. J. Dawson, V. Gotor-Fer-nandez, C. E. Humphrey, N. J. Turner, ChemBioChem2005, 6, 637–639.

[144] C. J. Dunsmore, R. Carr, T. Fleming, N. J. Turner, J.Am. Chem. Soc. 2006, 128, 2224–2225.

[145] K. R. Bailey, A. J. Ellis, R. Reiss, T. J. Snape, N. J.Turner, Chem. Commun. 2007, 44, 3640–3642.

[146] V. Kçhler, K. R. Bailey, A. Znabet, J. Raftery, M. Hel-liwell, N. J. Turner, Angew. Chem. 2010, 122, 2228–2230; Angew. Chem. Int. Ed. 2010, 49, 2182–2184.

[147] A. Znabet, E. Ruijter, F. J. J. de Kanter, V. Kçhler, M.Helliwell, N. J. Turner, R. V. A. Orru, Angew. Chem.2010, 122, 5417–5420; Angew. Chem. Int. Ed. 2010, 49,5289–5292.

[148] D. B. Janssen, Adv. Appl. Microbiol. 2007, 61, 233–252.

[149] T. Koudelakova, E. Chovancova, J. Brezovsky, M.Monincova, A. Fortova, J. Jarkovsky, J. Damborsky,Biochem. J. 2011, 354, 345–354.

[150] L. S. Campbell-Verduyn, W. Szymnski, C. P. Postema,R. A. Dierckx, P. H. Elsinga, D. B. Janssen, B. L. Fer-inga, Chem. Commun. 2010, 46, 898–900.

[151] K. A. Gray, T. H. Richardson, K. Kretz, J. M. Short, F.Bartnek, R. Knowles, L. Kan, P. E. Swanson, Dan E.Robertson, Adv. Synth. Catal. 2001, 343, 607–617.

[152] T. Bosma, G. Stucki, D. B. Janssen, Appl. Environ. Mi-crobiol. 2002, 68, 3582–3587.

[153] Z. Prokop, Y. Sato, J. Brezovsky, T. Mozga, R. Cha-loupkova, T. Koudelakova, P. Jerabek, V. Stepankova,R. Natsume, J. G. E. van Leeuwen, D. B. Janssen, J.Florian, Y. Nagata, T. Senda, J. Damborsky, Angew.Chem. 2010, 122, 6247–6251; Angew. Chem. Int. Ed.2010, 6111–6115.

[154] L. Tang, J. E. T. v. H. Vlieg, J. H. L. Spelberg, M. W.Fraaije, D. B. Janssen, Enzyme Microb. Technol. 2002,30, 251–258.

[155] L. Tang, D. E. Torres PazmiÇo, M. W. Fraaije, R. M.de Jong, B. W. Dijkstra, D. B. Janssen, Biochemistry2005, 44, 6609–6618.

[156] R. Schoevaart, F. van Rantwijk, R. A. Sheldon, Bio-technol. Bioeng. 2000, 70, 349–352.

[157] J.-Q. Liu, T. Dairi, N. Itoh, M. Kataoka, S. Shimizu,H. Yamada, J. Mol. Catal. B: Enzym. 2000, 10, 107–115.

[158] a) X. Garrabou, J. Joglar, T. Parella, R. Crehuet, J.Bujons, P. Clap�s, Adv. Synth. Catal. 2011, 353, 89–99;b) J. A. Castillo, C. Gu�rard-H�laine, M. Guti�rrez, X.Garrabou, M. Sancelme, M. Sch�rmann, T. Inoue, V.H�laine, F. Charmantray, T. Gefflaut, L. Hecquet, J.Joglar, P. Clap�s, G. A. Sprenger, M. Lemaire, Adv.Synth. Catal. 2010, 352, 1039–1046.

[159] M. von Itzstein, W.-Y. Wu, G. B. Kok, M. S. Pegg, J. C.Dyason, B. Jin, T. V. Phan, M. L. Smythe, H. F. White,S. W. Oliver, P. M. Colman, J. N. Varghese, D. M.Ryan, J. M. Woods, R. C. Bethell, V. J. Hotham, J. M.Cameron, C. R. Penn, Nature 1993, 363, 418–423.

[160] M. Wada, C.-C. Hsu, D. Franke, M. Mitchell, A.Heine, I. Wilson, C.-H. Wong, Bioorg. Med. Chem.2003, 11, 2091–2098.

[161] G. J. Williams, S. Domann, A. Nelson, A. Berry, Proc.Natl. Acad. Sci. USA 2003, 100, 3143–3148.

[162] S. Jennewein, M. Sch�rmann, M. Wolberg, I. Hilker,R. Luiten, M. Wubbolts, D. Mink, Biotechnol. J. 2006,1, 537–548.

[163] W. A. Greenberg, A. Varvak, S. R. Hanson, K. Wong,H. Huang, P. Chen, M. J. Burk, Proc. Natl. Acad. Sci.USA 2004, 101, 5788–5793.

[164] a) D. Sirim, F. Wagner, A. Lisitsa, J. Pleiss, BMC Bio-chem. 2009, 10, 27; b) D. R. Nelson, Human Genomics2009, 4, 1473–1479; c) M. Fischer, Q. K. Thai, M.Grieb, J. Pleiss, BMC Bioinformatics 2006, 7, 495;d) M. Fischer, M. Knoll, D. Sirim, F. Wagner, S.Funke, J. Pleiss, Bioinformatics 2007, 23, 2015–2017.

[165] A. Seifert, S. Vomund, K. Grohmann, S. Kriening,V. B. Urlacher, S. Laschat, J. Pleiss, ChemBioChem2009, 10, 853–861.

[166] Q.-S. Li, J. Ogawa, R. D. Schmid, S. Shimizu, Appl.Environ. Microbiol. 2001, 67, 5735–5739.

[167] J. A. Dietrich, Y. Yoshikuni, K. J. Fisher, F. X. Woo-lard, D. Ockey, D. J. McPhee, N. S. Renninger,M. C. Y. Chang, D. Baker, J. D. Keasling, ACS Chem.Biol. 2009, 4, 261–267.

[168] Y. Tu, R. Deshmukh, M. Sivaneri, G. D. Szklarz, DrugMetab. Dispos. 2008, 36, 2371–2380.

[169] S. Eiben, L. Kaysser, S. Maurer, K. K�hnel, V. B. Ur-lacher, R. D. Schmid, J. Biotechnol. 2006, 124, 662–669.

[170] S. Li, L. M. Podust, D. H. Sherman, J. Am. Chem. Soc.2007, 129, 12940–12941.

[171] F. Sabbadin, R. Hyde, A. Robin, E.-M. Hilgarth, M.Delenne, S. Flitsch, N. Turner, G. Grogan, N. C.Bruce, ChemBioChem 2010, 11, 987–994.

[172] P. A. England, C. F. Harford-Cross, J.-A. Stevenson,D. A. Rouch, L.-L. Wong, FEBS Lett. 1998, 424, 271–274.

[173] G. Hoffmann, K. Bçnsch, T. Greiner-Stçffele, M.Ballschmiter, Protein Eng. Des. Sel. 2011, 24, 439–446.

[174] S. G. Bell, X. Chen, R. J. Sowden, F. Xu, J. N. Wil-liams, L.-L. Wong, Z. Rao, J. Am. Chem. Soc. 2003,125, 705–714.

[175] T. Kubo, M. W. Peters, P. Meinhold, F. H. Arnold,Chem. Eur. J. 2006, 12, 1216–1220.

[176] a) S. Panke, B. Witholt, A. Schmid, M. G. Wubbolts,Appl. Environ. Microbiol. 1998, 64, 2032–2043; b) D.Tischler, D. Eulberg, S. Lakner, S. R. Kaschabek,W. J. H. van Berkel, M. Schlomann, J. Bacteriol. 2009,191, 4996–5009.

[177] L. Gursky, J. Nikodinovic-Runic, K. Feenstra, K.O�Connor, Appl. Microbiol. Biotechnol. 2010, 85, 995–1004.

[178] R. Gagnon, G. Grogan, M. S. Levitt, S. M. Robets,P. W. H. Wan, A. J. Willetts, J. Chem. Soc. PerkinTrans. 1 1994, 2537–2543.

[179] D. Bonsor, S. F. Butz, J. Solomons, S. Grant, I. J. S.Fairlamb, M. J. Fogg, G. Grogan, Org. Biomol. Chem.2006, 4, 1252–1260.

[180] C. Szolkowy, L. D. Eltis, N. C. Bruce, G. Grogan,ChemBioChem 2009, 10, 1208–1217.

[181] E. Malito, A. Alfieri, M. W. Fraaije, A. Mattevi, Proc.Natl. Acad. Sci. USA 2004, 101, 13157–13162.

2214 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2191 – 2215

REVIEWS Geoffrey A. Behrens et al.

[182] I. A. Mirza, B. J. Yachnin, S. Wang, S. Grosse, H. Ber-geron, A. Imura, H. Iwaki, Y. Hasegawa, P. C. K. Lau,A. M. Berghuis, J. Am. Chem. Soc. 2009, 131, 8848–8854.

[183] a) M. Bocola, F. Schulz, F. Leca, A. Vogel, M. W.Fraaije, M. T. Reetz, Adv. Synth. Catal. 2005, 347,979–986; b) M. T. Reetz, S. Wu, Chem. Commun.2008, 5499–5501.

[184] M. T. Reetz, S. Wu, J. Am. Chem. Soc. 2009, 131,15424–15432.

[185] A. Kirschner, U. T. Bornscheuer, Appl. Microbiol.Biotechnol. 2008, 81, 465–472.

[186] a) M. T. Reetz, B. Brunner, T. Schneider, F. Schulz,C. M. Clouthier, M. M. Kayser, Angew. Chem. 2004,116, 4167–4170; Angew. Chem. Int. Ed. 2004, 43,4075–4078; b) M. D. Mihovilovic, F. Rudroff, A. Win-ninger, T. Schneider, F. Schulz, Org. Lett. 2006, 8,1221–1224.

[187] M. T. Reetz, F. Daligault, B. Brunner, H. Hinrichs, A.Deege, Angew. Chem. 2004, 116, 4170–4173; Angew.Chem. Int. Ed. 2004, 43, 4078–4081.

[188] D. J. Opperman, M. T. Reetz, ChemBioChem 2010, 11,2589–2596.

[189] N. M. Kamerbeek, M. W. Fraaije, D. B. Janssen, Eur. J.Biochem. 2004, 271, 2107–2116.

[190] F. Hollmann, A. Taglieber, F. Schulz, M. T. ReetzAngew. Chem. 2007, 119, 2961–2964; Angew. Chem.Int. Ed. 2007, 46, 2903–2906.

[191] D. E. Torres PazmiÇo, R. Snajdrova, B.-J. Baas, M.Ghobrial, M. D. Mihovilovic, M. W. Fraaije, Angew.Chem. 2008, 120, 2307–2310; Angew. Chem. Int. Ed.2008, 47, 2275–2278.

[192] a) G. de Gonzalo, M. D. Mihovilovic, M. W. Fraaije,ChemBioChem 2010, 11, 2208–2231; b) J. Rehdorf,U. T. Bornscheuer, Monooxygenases, Baeyer–VilligerApplications in Organic Synthesis, John Wiley & Sons,Hoboken, 2009 ; c) H. Leisch, K. Morley, P. C. K. Lau,Chem. Rev. 2011, 111, 4165–4222.

[193] A. Leaver-Fay, M. Tyka, S. M. Lewis, O. F. Lange, J.Thompson, R. Jacak, K. Kaufman, P. D. Renfrew,

C. A. Smith, W. Sheffler, I. W. Davis, S. Cooper, A.Treuille, D. J. Mandell, F. Richter, Y. E. A. Ban, S. J.Fleishman, J. E. Corn, D. E. Kim, S. Lyskov, M. Ber-rondo, S. Mentzer, Z. Popovic, J. J. Havranek, J. Kara-nicolas, R. Das, J. Meiler, T. Kortemme, J. J. Gray, B.Kuhlman, D. Baker, P. Bradley, Methods Enzymol.2011, 545–574.

[194] L. Jiang, E. A. Althoff, F. R. Clemente, L. Doyle, D.Rothlisberger, A. Zanghellini, J. L. Gallaher, J. L.Betker, F. Tanaka, C. F. Barbas 3rd, D. Hilvert, K. N.Houk, B. L. Stoddard, D. Baker, Science 2008, 319,1387–1391.

[195] D. Rçthlisberger, O. Khersonsky, A. M. Wollacott, L.Jiang, J. DeChancie, J. Betker, J. L. Gallaher, E. A.Althoff, A. Zanghellini, O. Dym, S. Albeck, K. N.Houk, D. S. Tawfik, D. Baker, Nature 2008, 453, 190–195.

[196] O. Khersonsky, D. Rothlisberger, A. M. Wollacott, P.Murphy, O. Dym, S. Albeck, G. Kiss, K. N. Houk, D.Baker, D. S. Tawfik, J. Mol. Biol. 2011, 407, 391–412.

[197] J. B. Siegel, A. Zanghellini, H. M. Lovick, G. Kiss,A. R. Lambert, J. L. St Clair, J. L. Gallaher, D. Hil-vert, M. H. Gelb, B. L. Stoddard, K. N. Houk, F. E.Michael, D. Baker, Science 2010, 329, 309–313.

[198] N. U. Nair, C. A. Denard, H. Zhao, Curr. Org. Chem.2010, 14, 1870–1882.

[199] Special issue: “Chemoenzymatic Synthesis” Chem-CatChem 2011, 3, 237–419.

[200] B. Wilms, A. Wiese, C. Syldatk, R. Mattes, J. Alten-buchner, J. Biotechnol. 2001, 86, 19–30.

[201] E. J. Steen, Y. Kang, G. Bokinsky, Z. Hu, A. Schirmer,A. McClure, S. B. Del Cardayre, J. D. Keasling, Nature2010, 463, 559–562.

[202] H. Yim, R. Haselbeck, W. Niu, C. Pujol-Baxley, A.Burgard, J. Boldt, J. Khandurina, J. D. Trawick, R. E.Osterhout, R. Stephen, J. Estadilla, S. Teisan, H. B.Schreyer, S. Andrae, T. H. Yang, S. Y. Lee, M. J. Burk,S. van Dien, Nat. Chem. Biol. 2011, 7, 445–452.

Adv. Synth. Catal. 2011, 353, 2191 – 2215 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2215

Discovery and Protein Engineering of Biocatalysts for Organic Synthesis

The Articles

- 55 -

ARTICLE II

Rapid and Sensitive Kinetic Assay forCharacterization of ω-Transaminases

Sebastian Schatzle,† Matthias Hohne,† Erik Redestad,† Karen Robins,‡ andUwe T. Bornscheuer*,†

Institute of Biochemistry, Department of Biotechnology and Enzyme Catalysis, Greifswald University,Felix Hausdorff-Str. 4, 17487 Greifswald, Germany, and Lonza AG, Valais Works, Visp, Switzerland

For the biocatalytic preparation of optically active amines,ω-transaminases (ω-TA) are of special interest since theyallow the asymmetric synthesis starting from prostereo-genic ketones with 100% yield. To facilitate the purifica-tion and characterization of novel ω-TA, a fast kinetic assaywas developed based on the conversion of the widely usedmodel substrate r-methylbenzylamine, which is com-monly accepted by most of the known ω-TAs. The productfrom this reaction, acetophenone, can be detected spec-trophotometrically at 245 nm with high sensitivity (ε )12 mM-1 cm-1), since the other reactants show only alow absorbance. Besides the standard substrate pyru-vate, all low-absorbing ketones, aldehydes, or ketoacids can be used as cosubstrates, and thus the aminoacceptor specificity of a given ω-TA can be obtainedquickly. Furthermore, the assay allows the fast inves-tigation of enzymatic properties like pH and tempera-ture optimum and stability. This method was used forthe characterization of a novel ω-TA cloned fromRhodobacter sphaeroides, and the data obtained werein excellent accordance with a standard capillaryelectrophoresis assay.

ω-Transaminases (ω-TAs) are PLP-dependent enzymes thatcatalyze amino group transfer reactions. They have gainedincreased attention because of their potential for the asymmetricsynthesis of optically active amines, which are frequently used asbuilding blocks for the preparation of numerous pharmaceuticals.1

Although kinetic resolutions of racemic amines have beeninvestigated as well, the limitation to a maximum yield of 50%considerably hampers their application. On the other hand,asymmetric synthesis requires methods to shift the unfavorableequilibrium toward synthesis of single enantiomers of opticallypure amines for which several methods were developed.1-5 This

is one of the key prerequisites for efficient processes enablingthe use of transaminases on industrial scale. During processdevelopment or for the identification of novel ω-TA, purificationof the enzyme and characterization of enzymatic properties is ofgreat interest, but as ω-TA activity is usually determined with lowthroughput methods like HPLC or capillary electrophoresis (CE),the assay can become the limiting step.

This motivated us to develop a simple, fast and sensitive assaycomparable to, for example, the p-nitrophenyl ester assays foresterases and lipases,6 which can be used in a similar manner asa standard method for the determination of ω-TA activity. Kimand co-workers developed an assay based on the detection ofalanine formed during ω-TA-catalyzed reactions: after additionof a CuSO4/MeOH staining solution, the blue colored Cu-alaninecomplex could be determined at 595 nm.7 With this assayvarious aspects of the reaction can be monitored, like testingdifferent amines and �-amino acids as substrates in combinationwith pyruvate as the cosubstrate, the enantiopreference of theenzyme, and an estimation of enantioselectivity. There are,however, several drawbacks: Since the staining solution inhibitsthe enzyme, the assay can only be used as an end pointmeasurement and determination of kinetic parameters is notpossible. Furthermore, the sensitivity is very low (ε ≈ 10 M-1

cm-1), and the most commonly used buffers (e.g., phosphateand the Goods buffer) form insoluble blue Cu-complexes, whichneed to be centrifuged before the determination of the absor-bance value. Cell extracts also give a blue colored product,which is soluble and thus cannot be removed. This is probablydue to the amino acids present in the cells.

As an alternative, Truppo and co-workers recently publisheda multi-enzyme cascade pH-indicator assay to monitor the conver-sion in an asymmetric amine synthesis with ω-TAs.8 The ketoneis converted with alanine as cosubstrate, and NADH-dependent

* To whom correspondence should be addressed. Fax: (+49) 3834-86-80066.E-mail: uwe.bornscheuer@uni-greifswald.de.

† Greifswald University.‡ Valais Works.

(1) Hohne, M.; Bornscheuer, U. T. ChemCatChem 2009, 1, 42–51.(2) Matcham, G.; Bhatia, M.; Lang, W.; Lewis, C.; Nelson, R.; Wang, A.; Wu, W.

Chimia 1999, 53, 584–589.(3) Koszelewski, D.; Lavandera, I.; Clay, D.; Rozzell, D.; Kroutil, W. Adv. Synth.

Catal. 2008, 350, 2761–2766.(4) Koszelewski, D.; Lavandera, I.; Clay, D.; Guebitz, G. M.; Rozzell, D.; Kroutil,

W. Angew. Chem., Int. Ed. 2008, 47, 9337–9340.(5) Hohne, M.; Kuhl, S.; Robins, K.; Bornscheuer, U. T. ChemBioChem 2008,

9, 363–365.

(6) Krebsfanger, N.; Schierholz, K.; Bornscheuer, U. T. J. Biotechnol. 1998,60, 105–111.

(7) Hwang, B.-Y.; Kim, B.-G. Enzyme Microb. Technol. 2004, 34, 429–436.

Table 1. Purification Summary for the ω-TA fromRhodobacter sphaeroides

stepvolume(mL)

protein(mg)

activity(U/mL)

yield(%)

specificactivity(U/mg)

purificationfactor

crude cell extract 20 670 5.4 100 0.161 1IMAC/

ultrafiltration1 27 27.5 25.5 1.019 6.3

Anal. Chem. 2009, 81, 8244–8248

10.1021/ac901640q CCC: $40.75 2009 American Chemical Society8244 Analytical Chemistry, Vol. 81, No. 19, October 1, 2009Published on Web 09/09/2009

lactate or alanine dehydrogenases are used to remove thecoproduct pyruvate. NADH is then recycled with glucose dehy-drogenase, and the arising gluconic acid δ-lactone induces adecrease of the pH, which is visualized by the application of a pHindicator. Although the sensitivity is higher compared to theCuSO4/MeOH assay, no information about enantiopreferenceor enantioselectivity can be obtained.

EXPERIMENTAL SECTIONCloning, Expression, Purification of Rhodobacter sphaeroi-

des ω-TA. Genomic DNA of Rhodobacter sphaeroides 2.4.1 (DSM158) was amplified using the GenomePhi kit (GE healthcare) andused in a subsequent PCR to amplify the gene (forward-primer:

aggagatatacatatgcgtgacgatgcaccgaattcctg; reverse-primer: gatgat-gatgggatcctgaggcgacttcggcgaagaccttc; underlined: NdeI, BamHIrestriction sites). After 30 cycles consisting of 30 s denaturationat 95 °C, 30 s annealing at 65 °C and elongation at 74 °C for 1.5min, the fragment was purified, digested with BamH1 and NdeI,and ligated in pGASTON.9

Expression and Purification of Recombinant ω-TA in E.coli BL21 (DE3). Transformed E. coli BL21 (DE3) strains weregrown in 500 mL of LB medium supplemented with ampicillin(100 µg/mL). Cells were incubated initially at 37 °C on a gyratoryshaker until the OD600 reached 0.7. The cells were then inducedby the addition of 0.2% rhamnose, and at the same time theincubation temperature was decreased to 20 °C. After inductionthe incubation was continued for a further 18 h. Aliquots werewithdrawn at several points of time after induction to followthe expression. Crude cell extracts were prepared by sonicationand centrifugation (16 060g for 5 min) to separate the solubleand insoluble fractions.

The cell pellet (∼3 g wet weight) was washed twice with 50mM phosphate buffer (pH 8), containing 0.1 mM PLP at 4 °C.After disruption (french press) the cell suspension was centrifuged(4000 × g, 30 min), and the resulting supernatant was passedthrough a 0.5 µm filter prior to chromatography. Chromatography

Figure 2. Comparison of the acetophenone assay at differentwavelengths. The same reaction was measured at 245, 270, 275,280, and 290 nm, whereas the calculated activities at 245 and 270nm differed only by 1-7%. At wavelengths >270 nm, however,deviations of calculated activities in the range of 10-20% wereobserved. The considerably lower overall absorption at higherwavelengths allows measurements starting with higher substrateconcentrations up to higher product concentrations.

Table 2. Amino Acceptor Profile of Rhodobactersphaeroides ω-TA As Determined by the AcetophenoneAssay

amino acceptor relative activity [%]

pyruvate 100R-ketoglutarate n.d.glyoxylate 49succinic semialdehyde 78ethylpyruvate 52ethyl-3-oxobutanoate 0.1acetaldehyde 14butyraldehyde 38acetone n.d.2-heptanone 0.2acetoin 0.07N-Boc-3-aminopyrrolidone 0.07

Figure 1. Expression of Rhodobacter sphaeroides-ωTA in E. coli BL21. The formation of an overexpression band with a size of around 43kDa indicates the expression of the ω-TA in the soluble fraction. Also a minor amount of enzyme was found in the insoluble fraction. M: marker(Roti-Mark Standard, Roth: trypsin-inhibitor, 20 kDa; carboanhydrase, 29 kDa; ovalbumine, 43 kDa; serum albumine, 66 kDa; �-galactosidase,119 kDa; myosine, 200 kDa). Purified: enzyme after affinity chromatography (HisTag). Cell: crude extract before purification.

8245Analytical Chemistry, Vol. 81, No. 19, October 1, 2009

was performed using an AKTA Purifier (GE Healthcare). Thefiltered cellular extract was applied to a 5 mL column of IMACSepharose 6 Fast Flow (GE Healthcare). The column was washedat a flow rate of 5 mL min-1 with a 10 column volume of 50 mMphosphate buffer, pH 8, containing 100 mM NaCl, 0.1 mM PLP,and 30 mM imidazole (to avoid unspecific binding), and theω-TA activity was eluted with 10 column volumes of 50 mMphosphate buffer, pH 8, containing 100 mM NaCl, 0.1 mM PLP,and 300 mM imidazole (flow rate of 5 mL min-1). The activitycontaining fractions were pooled, concentrated and washedwith 0.1 mM PLP in phosphate buffer (pH 8) several times byultrafiltration and the purified enzyme was stored at 4 °C (Table1).

Sodium Dodecyl Sulfate-Polyacrylamide Gel Electro-phoresis (SDS-PAGE). Samples from cultivation after sonicationwere divided into soluble and insoluble fractions (10 µL) andanalyzed by SDS-PAGE on polyacrylamide gels (12.5 or 8%) witha stacking gel (4%). The low molecular weight protein standardmixture obtained from Roth (Karlsruhe, Germany) was used asreference. Gels were stained for protein detection with CoomassieBrilliant Blue (Figure 1).

Establishment and Validation of the SpectrophotometricAssay. All UV-photometric measurements were performed usinga Varioskan spectrophotometer (Thermo Electron Corporation,Langenselbold, Germany). Spectra of compounds in adequateconcentrations were taken in the range from 220 to 300 nm inBrand UV 384-well MTP (sample volume 114 µL) in five dupli-cates. Kinetic spectrophotometric measurements were conductedat 245 nm in Greiner UV 96-well plates with a reaction mixturecontaining 2.5 mM R-MBA and 2.5 mM pyruvate in 200 mLphosphate buffer (50 mM, pH 8), 0.25% DMSO, and an appropriateamount of enzyme (0.04 to 4 µg purified enzyme or 0.2 to 20 µLof crude extract with an OD600 of 10, depending on the respectiveactivity). The initial absorbance was below 0.5. If higherconcentrations of R-MBA or pyruvate are desired, for example,for determination of kinetic constants, measurements can beperformed at higher wavelengths, for example, 270-290 nm,or in a lower reaction volume to decrease the initial absorbance(Figure 2). Each measurement was repeated at least three times.The specific activity was expressed as units/mg purified protein.

One unit of activity was defined as the amount of enzyme thatproduced 1 µmol acetophenone from (S)-R-MBA per minute.Standard curves were measured with substrate ((S)-R-MBA/pyruvate) and product (acetophenone/alanine) concentrations of2 to 2.5 mM and 0 to 0.5 mM, respectively.

For validation of the spectrophotometric assay the reactionsolution was divided into different wells of a microtiter plate.During the spectrophotometric measurement, the reaction wasstopped successively in different wells by adding 1/10 of thevolume of 1 M HCl and centrifuged. The concentration ofacetophenone was analyzed by micellar electrokinetic chroma-tography (MEKC) in a capillary electrophoresis system using theCElixir SDS kit (MicroSolv Technology Corporation, Eatontown,U.S.A.). The separations were done on a Beckman PACE-MDQsystem equipped with a fused capillary and a photodiode arraydetector. The capillary length was 30 cm, with a distance of 10cm to the detector and an inner diameter of 50 µm. After injection(0.5 psi for 5 s) and injection of a water plug (0.1 psi for 10 s) thesamples were separated by applying 25 kV for 2.5 min. Themigration times were 1.4 min (acetophenone) and 1.8 min (R-MBA).

(8) Truppo, M. D.; Rozzell, J. D.; Moore, J. C.; Turner, N. J. Org. Biomol. Chem.2009, 7, 395–398.

(9) Henke, E. PhD thesis, University of Greifswald, 2001.

Figure 3. Principle of the acetophenone assay (left). Acetophenone (1 mM) shows a high absorbance (right) compared to other reactants suchas R-methylbenzylamine, pyruvate or alanine (all 10 mM).

Figure 4. Validation of the formation of acetophenone by thespectrophotometric assay (b) and by micellar electrokinetic chroma-tography (O). The calculated activities differed only by 3% for bothpurified enzyme (data not shown) and crude extract (data shown).

8246 Analytical Chemistry, Vol. 81, No. 19, October 1, 2009

Characterization of the ω-TA from Rhodobacter sphaeroi-des. Protein concentrations were determined using the BCA assaykit (Uptima, Montlucon, France). For determination of thetemperature and pH optimum of the ω-TA, the activity was assayedin the reaction mixture described above at various temperatures(4, 10, 20, 30, 40, 50, 60, 70 °C) and various pH values (5.5-7phosphate; 7.5-9 Tris; 9.5-11 glycine), respectively. For deter-mination of the stability, the purified enzyme was incubated atdifferent temperatures (see above), and aliquots were taken atcertain points. After dilution to 1:10 with phosphate buffer (pH 8)and incubation on ice for 5 min the activity was assayed at roomtemperature.

To determine the substrate specificity, the amino donorpyruvate was replaced by the same concentration of other ketones,aldehydes, and keto acids (Table 2), and activities were assayedas described above.

RESULTS AND DISCUSSIONWhile studying the absorption spectra of various compounds

we found that acetophenone, which is formed during the tran-samination of the standard substrate R-methylbenzylamine, showeda high absorbance in the ultraviolet range of the spectrum with a

local maximum at 245 nm due to its carbonyl group conjugatedto the aromatic ring system. On the other hand, compounds likepyruvate, alanine, alkylamines, R-methylbenzylamine, and ketonesshowed an absorption which was of less than 5% of acetophenoneat this wavelength. Thus, the reaction can be followed simply bymonitoring the increasing absorption in the model reaction of theω-TA (Figure 3) neglecting absorption changes caused by otherreaction partners. Since most ω-TA converts R-methylbenzylamine(R-MBA) well, all low absorbing ketones, aldehydes, and ketoacids may be used in the assay as cosubstrate. This allows thecharacterization of the amino acceptor spectrum of an ω-TA, aswell as the determination of the enantiopreference and anestimation of enantioselectivity for the substrate R-MBA. Addition-ally, characterization of pH and temperature optimum and tem-perature stability can be performed easily. We also investigatedother R-arylalkylketones like p-bromo- or p-nitroacetophenone,2-indanone, and benzaldehyde as amino donors and their respec-tive amine products, but the differences in absorbance of theketone and amine was not as high as found for the R-methylben-zylamine/acetophenone pair. Furthermore, the amines also showeda significant absorption itself, so that the reaction mixture wouldhave to be diluted prior to the measurement, since already at smallamine concentrations a high initial absorbance would prevent akinetic measurement of the reaction.

This assay was validated by following the reaction using bothmeasurement of the absorption at 245 nm and by detection ofthe acetophenone formed by capillary electrophoresis with mi-cellar electrokinetic chromatography (MEKC), yielding compa-rable results when using either purified enzyme or crude cellextract as enzyme source (Figure 4). The absorption coefficientof acetophenone at 245 nm was determined to be ≈12 mM-1 cm-1.

One limitation of the assay is that the protein content of thesample contributes to the initial absorbance. This limits theamount of protein to be used for the assay: the addition of a higheramount of enzyme preparation does not lead to an increasedsensitivity, because at a higher initial absorption the sensitivityfor small changes in the absorption decreases.

Application of the Assay in the Characterization of an ω-TAfrom Rhodobacter sphaeroides. To verify the methods de-scribed above, an ω-TA identified in the genome of Rhodobacter

Figure 5. Temperature profile (a) and pH-profile (b) of Rhodobacter sphaeroides ω-TA.

Figure 6. Temperature dependency of the stability of the Rhodo-bacter sphaeroides ω-TA.

8247Analytical Chemistry, Vol. 81, No. 19, October 1, 2009

sphaeroides 2.4.1 was characterized using this assay. The genegi3718887 was cloned from genomic DNA, and the enzyme wasexpressed in E. coli BL21. A C-terminal His-tag was introducedand allowed a simple purification of the enzyme to nearhomogeneity (Table 1). Investigation of the amino acceptorprofile revealed that beside pyruvate also glyoxylate, succinicsemialdehyde, and butyraldehyde are well accepted substrates.The ketones 2-heptanone and 1-N-Boc-3-aminopyrrolidine wereconverted only very slowly, and no conversion of 2-oxoglutarateor acetone was detected (Table 2).

Standard enzymatic properties like pH-profile (Figure 5b),temperature profile (Figure 5a), stability (Figure 6), and theeffect of additives or buffer composition on activity could beinvestigated very quickly and simply using this assay. The

enzyme is most active at pH 8.5 and 40 °C; however, the half-life of the enzyme decreases rapidly if it is incubated attemperatures above 20 °C.

CONCLUSIONSIn this paper we described a simple and sensitive kinetic assay

that can be used as a high throughput method for screeningprotein libraries and for the fast characterization of amino acceptorspecificity of ω-transaminases. This was achieved without the needfor any additional enzymes or staining solutions, which is themajor advantage compared to previously described procedures.

Received for review July 23, 2009. Accepted August 22,2009.

AC901640Q

8248 Analytical Chemistry, Vol. 81, No. 19, October 1, 2009

The Articles

- 61 -

ARTICLE III

Conductometric Method for the RapidCharacterization of the Substrate Specificity ofAmine-Transaminases

Sebastian Schatzle,† Matthias Hohne,† Karen Robins,‡ and Uwe T. Bornscheuer*,†

Institute of Biochemistry, Department of Biotechnology and Enzyme Catalysis, Greifswald University,Felix Hausdorff-Strasse 4, 17487 Greifswald, Germany, and Lonza AG, Valais Works, Visp, Switzerland

Amine-transaminases (ATAs, ω-transaminases, ω-TA) arePLP-dependent enzymes that catalyze amino group trans-fer reactions. In contrast to the widespread and well-known amino acid-transaminases, ATAs are able to con-vert substrates lacking an r-carboxylic functional group.They have gained increased attention because of theirpotential for the asymmetric synthesis of optically activeamines, which are frequently used as building blocks forthe preparation of numerous pharmaceuticals. Havingalready introduced a fast kinetic assay based on theconversion of the model substrate r-methylbenzylaminefor the characterization of the amino acceptor specificity,we now report on a kinetic conductivity assay for inves-tigating the amino donor specificity of a given ATA. Thecourse of an ATA-catalyzed reaction can be followedconductometrically since the conducting substrates, apositively charged amine and a negatively charged ketoacid, are converted to nonconducting products, a non-charged ketone and a zwitterionic amino acid. The de-crease of conductivity for the investigated reaction systemswere determined to be 33-52 µS mM-1. In contrast toother ATA-assays previously described, with this ap-proach all transamination reactions between any amineand any keto acid can be monitored without the needfor an additional enzyme or staining solutions. Theassay was used for the characterization of a ATA fromRhodobacter sphaeroides, and the data obtained werein excellent agreement with gas chromatographyanalysis.

The identification of novel biocatalysts and the optimizationof existing enzymes are key steps for developing highly efficientprocesses. Thus, protein engineering and the metagenome ap-proach have emerged as powerful tools that are used frequentlyto achieve these aims. While working with large enzyme librarieswhich are generated during these strategies, the fast, effective,and reliable identification of enzymes with the desired propertiesis often the bottleneck since this is the most time-consuming step.Therefore a large variety of fast methods for determining enzymeactivity have been developed for many biocatalysts, which often

are capable of being performed in a high-throughput screening(HTS) format. Examples includes solid-phase bound assays relatedto immunoassay,1 IR thermography,2 circular dichroism (CD)spectroscopy,3,4 fluorescence image analysis,5 and UV-vis-basedmethods.6,7

Amine-transaminases (ATA, ω-transaminases, ω-TA) gainedincreased attention in research in the last years because of theirpotential for the synthesis of optically active amines, which arefrequently used as building blocks for the preparation of numerousphysiologically active compounds. However, for the screening ofthe substrate specificities and enantioselectivities of ATA only alimited number of methods have been reported, due to the specialtype of reaction catalyzed (Figure 1): An amine and a carbonylsubstrate bearing a ketone or aldehyde group are converted intoproducts by exchanging their amino and carbonyl functionalgroups, without the stochiometric consumption of a cofactor. Incase of other enzymatic reactions like e.g. hydrolysis of esters oramides, the substrates (ester/amide) and products (alcohol/amine

* To whom correspondence should be addressed: Fax: (+49) 3834-86-80066.E-mail: uwe.bornscheuer@uni-greifswald.de.

† Greifswald University.‡ Lonza AG.

(1) Taran, F.; Gauchet, C.; Mohar, B.; Meunier, S.; Valleix, A.; Renard, P. Y.;et al. Angew. Chem., Int. Ed. 2002, 41, 124–127.

(2) Reetz, M. T.; Becker, M. H.; Kuhling, K. M.; Holzwarth, A. Angew. Chem.,Int. Ed. 1998, 37, 2647–2650.

(3) Ding, K.; Ishii, A.; Mikami, K. Angew. Chem., Int. Ed. 1999, 38, 497–501.(4) Reetz, M. T.; Kuhling, K. M.; Hinrichs, H.; Deege, A. Chirality 2000, 12,

479–482.(5) Soleihac, J.-M.; Cornille, F.; Martin, L.; Lenoir, C. Anal. Biochem. 1996,

241, 120–127.(6) Morıs-Varas, F.; Shah, A.; Aikens, J.; Nadkarni, N. P.; Rozzell, J. D.;

Demirjian, D. C. Bioorg. Med. Chem. 1999, 7, 2183–2188.(7) Baumann, M.; Hauer, B. H.; Bornscheuer, U. T. Tetrahedron: Asymmetry

2000, 11, 4781–4790.

Figure 1. Transaminase reaction and principle of the conductivityassay. A positively charged amine and a negatively charged keto acidare converted to a zwitterionic amino acid and a noncharged ketone,thus the conductivity of the reaction solution decreases.

Anal. Chem. 2010, 82, 2082–2086

10.1021/ac9028483 2010 American Chemical Society2082 Analytical Chemistry, Vol. 82, No. 5, March 1, 2010Published on Web 02/11/2010

and acid) can be discriminated very easily,8 or during reactionsinvolving a cofactor like NAD(P)H or side product like H2O2,these compounds can be detected specifically.9,10 For ATA, thisis not the case, and there are only a few special methods todiscriminate the similar substrates and products without affectingenzyme activity.11

In addition to ketones and aldehydes, ATA also convert theR-keto carboxylates pyruvate and glyoxylate. The methods forassaying aminotransferase activity published so far comprise (i)measurement of the generated amino acid (alanine, glycine) viadetection of the corresponding copper complex11 and (ii) a pHindicator based multienzyme cascade assay.12 Obviously, thereare several drawbacks: the copper staining solution in method(i) inhibits the enzyme, and nonspecific color formation withseveral buffers and crude cell extracts are observed. Thus, onlyend point measurements are possible, and appropriate measure-ments of the blank are necessary. Method (ii) works in theasymmetric synthesis mode (Figure 1, reaction from right to leftfor bottom partners), and thus no information about enantiopref-erence or -selectivity can be obtained.

Alternatively, we recently published an UV-spectrophotometricassay based on the conversion of the widely used model substrateR-methylbenzylamine.13 The product from this reaction, acetophe-none, can be detected spectrophotometrically at 245 nm with highsensitivity (ε ) 12 mM-1 cm-1), since the other reactantsshowed only a low absorbance. Besides the standard substratepyruvate, all low-absorbing ketones, aldehydes, or R-ketocarboxylates could be used as cosubstrates. Thus, the assayis a fast and easy method for determining transaminase activity.Additionally, the amino acceptor specificity of a given ATA canbe characterized very quickly.

As the acetophenone assay is limited to R-methylbenzylamineas amino donor, we now developed a method for a fast and easycharacterization of the amine donor specificity of a given ATA.With this conductivity based approach every amine donor andacceptor can be applied as substrate, as long as one of thesubstrates is an amino or keto acid, respectively.

EXPERIMENTAL SECTIONAll conductivity mesasurements were performed using either

a Qcond 2200 (VWR, Germany) or a CX-401 (Elmetron, Poland)multimeter with a graphite (L 12 mm) or a platinum (L 5 mm)electrode, respectively.

Buffer Preparation. All buffers had a concentration of 20 mMand were adjusted to pH 7.5. Tris was adjusted with HCl, the threebuffers containing BES, CHES, and HEPES were adjusted withBis-Tris (bis(2-hydroxyethyl)iminotris(hydroxymethyl)methane),and EPPS and tricine were adjusted with 1,8-diazabicyclo[5.4.0]-undec-7-en.

Calibration. For calibrating the assay, different standardsolutions with decreasing substrate and increasing productconcentrations of 0-10 mM were prepared, and the conductivityand pH were measured (Table 1). No change of the pH could bedetected. It is noteworthy that the accuracy of the assay wassignificantly higher starting from 10 mM substrates than startingfrom 5 mM substrates. Furthermore, it was important to use allreactants simultaneously, including the nonconducting ketone andalanine, for the calibration (see results). For analyzing theinfluence of crude extract on the conductivity measurements,standard curves with different amounts (0-18% (v/v)) of crudeextract (OD600 ≈ 10) of E. coli BL21 without ATA weremeasured in parallel. For the preparation of the crude extract,cells were washed twice and disrupted by sonication in thebuffer used for conductivity measurements.

Kinetic Measurements. The reactions were carried out atroom temperature (RT). Kinetic measurements were performedin reaction mixtures containing 10 mM substrates in buffer (20mM, pH 7.5), 10% dimethyl sulfoxide, and an appropriate amountof enzyme (purified enzyme or crude extract). Using the platinum(L 5 mm) electrode, the reaction can be carried out in a well ofa microtiter plate, and a reaction volume of 200 µL is sufficient.The course of the reaction was followed by measuring theconductivity of the solution continuously. Each measurement wasrepeated at least three times. For blank measurement, the reactionwas carried out with cell extract lacking an amine-transaminaseor with reaction mixtures containing only one of the substrates,but no significant change in conductivity could be detected. Thespecific activity was expressed as units per milligram protein. Oneunit of activity was defined as the amount of enzyme that produced1 µmol ketone product per minute.

Validation of the Assay with Gas Chromatography. Duringkinetic measurement, aliquots were taken at certain points forvalidation of the conductivity assay by gas chromatography. Thereaction was stopped by adding 1/10 of the volume of 1 M HCl.The ketones formed during the reaction were extracted with onevolume of ethyl acetate and determined by gas chromatography(Hewlett-Packard 5890 Series II Plus) using a forte BP-21 column(SGE, Griesheim, Germany) with benzaldehyde as internalstandard.

Characterization of the ATA from Rhodobacter sphaeroi-des. The ATA from Rhodobacter sphaeroides 2.4.1 (DSM 158) wasrecombinantly expressed in E. coli BL21 (DE3) and purified via

(8) Krebsfanger, N.; Schierholz, K.; Bornscheuer, U. T. J. Biotechnol. 1998,60, 105–111.

(9) Hinman, L. M.; Blass, J. P. J. Biol. Chem. 1981, 256, 6583–6586.(10) Holt, A.; Palcic, M. M. Nature Prot. 2006, 1, 2499–2505.(11) Hwang, B.-Y.; Kim, B.-G. Enzyme Microb. Technol. 2004, 34, 429–436.(12) Truppo, M. D.; Rozzell, J. D.; Moore, J. C.; Turner, N. J. Org. Biomol. Chem.

2009, 7, 395–398.(13) Schatzle, S.; Hohne, M.; Redestad, E.; Robins, K.; Bornscheuer, U. T. Anal.

Chem. 2009, 81, 8244–8248.

Table 1. Calibration of the Conductivity Assaya

substrate/product 1 & 6 2 & 6 3 & 6 4 & 6 5 & 6 3 & 7 3 & 8 3 & 9 3 & 10

conductivity [µs mM-1 cm-1] 36.6 39.0 44.0 34.6 32.9 37.1 40.0 49.5 52.1

a The values reflect the change in conductivity due to the conversion of 1 mM substrates to products (see Figure 4 for structures of substratesand products 1-10). The resolution of the applied conductometer was 0.1 µS.

2083Analytical Chemistry, Vol. 82, No. 5, March 1, 2010

affinity chromatography (IMAC sepharose) as reported previ-ously.13

Protein concentrations were determined using the BCA assaykit (Uptima, Montlucon, France).

To determine the substrate specificity of the enzyme, 10 mMof different substrates and an appropriate amount of purifiedenzyme (0.1-2 mg mL-1) were mixed with 1 mL of 20 mMtricine buffer pH 7.5. Dimethyl sulfoxide (10% (v/v)) was addedto the mixture to dissolve hydrophobic substrates. Solutionsof oxaloacetate and R-ketoglutarate were prepared freshly priorto measurements. The reaction was carried out at RT andfollowed by measuring the conductivity continuously. Forvalidation by GC, aliquots were taken at certain points, andthe concentration of the ketone products 1b-5b was analyzedas described above.

RESULTS AND DISCUSSIONIn the course of an ATA-catalyzed reaction the conductivity of

the reaction medium changes since charged reactants (amine andketo acid) are converted to noncharged species (the ketone anda zwitterionic amino acid) (Figure 1). For all experiments, thetransamination was carried out in kinetic resolution mode. Therationale behind this approach is as follows: (i) the reactionequilibrium favors product formation, (ii) both enantiomers of a

given amine substrate can be used separately in the assay, so thatalso information about enantioselectivity can be obtained, and (iii)the amine substrate usually is much better soluble than thecorresponding ketone. Thus, even with higher substrate concen-trations a homogeneous assay solution is obtained rather than abiphasic system, which otherwise might interfere with themeasurements.

This allows a simple measurement of the reaction progress, ifsome crucial requirements are fulfilled: First, to maximizesensitivity the buffer system should have low background con-ductivity. The pH of the buffer has to be kept in a pH range from4.0-8.0, where the net charge of alanine is ≈0. Furthermore, forpractical reasons or for high-throughput screening purposes, theassay should not be sensitive toward variations in crude extractconcentrations.

Standard buffer systems like phosphate buffer or Tris-HClcould not be applied due to their very high conductivity, wherebythe relative change in conductivity caused by the enzymaticreaction is rather small. Thus, Good’s buffers were investigatedregarding the conductivity of 20 mM solutions at pH 7.5 and effectson aminotransferase activity (Table 2). Due to their zwitterionicnature, solutions of these compounds show a much smallerconductivity if the pH is kept within 2 units near the pI of therespective compound (isoelectric buffers).

The first system investigated was BES-Bis-Tris. It showed avery good performance with purified enzyme, but unfortunatelyactivities of crude extract measured in this buffer did not matchthe results obtained by gas chromatography. Although the CHES-Bis-Tris system showed excellent performance with both purifiedenzyme and crude extract, due to its inhibitory effect (67% enzymeactivity compared to phosphate buffer) and its rather low buffercapacity at pH 7.5, we explored an alternative buffer system. Theinhibitory effect was the same for the EPPS buffer (76% rel activity)and even greater for the HEPES-Bis-tris system (24% rel activity).Finally, the tricine buffer showed both a good buffer capacity atpH 7.5 and a very good agreement of determined activities ofpurified enzyme and crude extract by measuring the conductivitycompared to GC analysis (Figure 2). For this reason the tricinebuffer was used for all subsequent experiments. The inhibitoryeffect of the EPPS and CHES buffer may vary for other transami-

Table 2. Summary of the Buffer Systems Investigatedc

buffera pKa

conductivity[µS cm-1]

relactivity [%]b

PB 7.2 2760 100BES 7.3 445 120HEPES 7.7 141 24EPPS 8.0 246 76Tris 8.1 1600 71tricine 8.3 240 98CHES 9.4 46 67A. dest - 3 0

a PB: phosphate buffer; BES: N,N-bis(2-hydroxyethyl)taurine; HEPES:4-(2-hydroxyethyl)piperazine-1-ethanesulfonic acid; EPPS: 4-(2-hydroxy-ethyl)-1-piperazinepropanesulfonic acid; Tris: 2-amino-2-(hydroxym-ethyl)-1,3-propanediol; tricine: N-[tris(hydroxymethyl)methyl]glycine;CHES: 2-(cyclohexylamino)ethanesulfonic acid. b Activities in thedifferent buffer systems were determined using the acetophenoneassay.13 c All buffers had a concentration of 20 mM and were adjustedto pH 7.5 (see the Experimental Section).

Figure 2. Validation of the assay by following the reaction of 10 mM (S)-R-methylbenzylamine and pyruvate using both conductivity measurements(black symbols/bar) and detection of the acetophenone formed by gas chromatography (white symbols/bar). The calculated activities differedonly by 5% for purified enzyme (left, 0 0.25 mg/mL, 3 0.5 mg/mL, O 1 mg/mL) and crude extract (10% (v/v)) (right, the curve displays theconductivity measurement, the bars show the calculated acetophenone concentrations).

2084 Analytical Chemistry, Vol. 82, No. 5, March 1, 2010

nases, so these buffers should be considered for furtherexperiments.

In practice and especially for screening of many samples, crudecell extract is used as enzyme preparation instead of purifiedproteins. Thus an important question was whether the presenceof different amounts of crude extract has an impact on theconductivity measurement. As expected, the background conduc-tivity of the reaction medium increased proportional to theconcentration of cell extract used. Next, calibration experimentswere performed where different conversions were simulated byvarying the concentrations of all reactants according to a realreaction. Fortunately, the addition of different amounts of crudeextract did not seem to affect the change of conductivitysignificantly (Figure 3). This was surprising, because in theorythe molar conductivity of a given compound depends on theconcentration of all electrolytes present in the reaction systemfor mainly two reasons: on the one hand, the protonation state

and thus the net charge of an analyte may be affected by othercompounds present in the solution. On the other hand, moreimportantly, ions or even neutral molecules from the backgroundelectrolyte may affect the mobility and thus the molar conductivityof an analyte significantly by intermolecular electrostatic orhydrophobic interactions. This explains why the conductivity ofa mixture comprised of different ions usually differs to some extentfrom the sum of the contributions of the single ions to overallconductivity. For this reason, all participating reactants and eventhe neutral species had to be included in the calibration of theassay. If otherwise the change of the conductivity per milimolarreaction conversion [µS/mM] is calculated from molar conductivi-ties of the single compounds, the obtained value would be higher(49.4 µS/mM conversion, O in Figure 3) compared with themeasurement of the complete mixtures (44.1 µS/mM conversion,b in Figure 3).

Application of the Assay in the Characterization of an ATAfrom Rhodobacter sphaeroides. To verify the method describedabove, an ATA identified in the genome of Rhodobacter sphaeroides2.4.1 was characterized using this assay. The enzyme wasexpressed recombinantly in E. coli BL21 and purified as de-scribed.13 Investigation of the amino donor profile revealed thatbesides the model substrate R-methylbenzylamine also benzyl-amine is a very well accepted substrate (Figure 4). Aliphatic andarylaliphatic compounds were converted only slowly, especiallythe cyclic aliphatic amine 1-N-Boc-3-aminopiperidine. The resultsobtained for the amino acceptor profile are in good agreementwith previously measured data using the acetophenone assay.13

Pyruvate is the best converted acceptor, glyoxylate and succinicsemialdehyde are converted moderately, and the keto-dicarboxylicacids oxalacetate and R-ketoglutarate are converted only veryslowly.

The conductivity assay is an excellent complementary methodto the acetophenone assay, since the usually more interestingamino donor specificity of an ATA can be investigated. With bothassays in hand, the complete characterization of the substratespecificity can be performed very rapidly. Furthermore, theconductivity assay allows a fast screening of ATA variants for thereaction of any desired amine with a keto acid. One limitation ofthe conductivity assay is that an appropriate low conductivity buffer

Figure 4. Substrate specificity of Rhodobacter sphaeroides ATA was determined using the conductivity assay (black) as well as gaschromatography (gray). All measurements were repeated three times.

Figure 3. Influence of different concentrations of crude extract (b0%, 1 5%, 9 10%,[ 14%, 2 18% (v/v)) on the change of conductivityduring the transaminase reaction. While measuring the standardcurves all reactants have to be present, as if measured separately,the calculated rate ∆µS mM-1 (O) is significantly higher. Contrary toother buffer systems investigated (and at substrate concentrationsof 5 mM), no significant influence of different crude extract concentra-tions were observed.

2085Analytical Chemistry, Vol. 82, No. 5, March 1, 2010

in the pH-range of 4-8 has to be used. For determining otherenzymatic properties like the pH- or temperature optimum of anATA, the acetophenone assay is more flexible since every lowabsorbing buffer at any pH may be used and absorbance isvirtually not dependent on temperature in contrast to conductivity,so that multiple calibrations can be avoided.

CONCLUSIONSIn this contribution we described a simple and sensitive

conductivity assay that can be used for the fast characterization

of the amino donor substrate profile of ATA. This was achievedwithout the need for any additional enzymes or staining solutions.With appropriate equipment, we envision an application of thisassay as a high-throughput method for screening enzyme libraries.

Received for review December 15, 2009. AcceptedJanuary 30, 2010.

AC9028483

2086 Analytical Chemistry, Vol. 82, No. 5, March 1, 2010

The Articles

- 67 -

ARTICLE IV

nature CHeMICaL BIOLOGY | vol 6 | NovEMBER 2010 | www.nature.com/naturechemicalbiology 807

articlepuBLIsHed OnLIne: 26 septeMBer 2010 | dOI: 10.1038/nCHeMBIO.447

Enzyme catalysis represents one cornerstone in the area of white (industrial) biotechnology1–5. The usually excellent chemo-, regio- and enantioselectivity of biocatalysts6,7 facili-

tates and simplifies many chemical processes for the production of a broad range of products. The production of optically pure building blocks for the pharmaceutical and fine chemicals industries are the most valuable areas8–12 of use for enzyme catalysis. These processes are environmentally friendlier than other options because addi-tional reaction steps can be avoided and the processes are usually performed under milder reaction conditions with fewer organic solvents and/or in the absence of hazardous materials13. The bio-catalytic synthesis of optically pure compounds can either be per-formed as a kinetic resolution or as an asymmetric synthesis14, as shown in Scheme 1 for transaminase-catalyzed reactions.

The kinetic resolution of the racemic amine with an (S)-selective transaminase produces the (R) enantiomer. The disadvantage of this approach is that a maximum yield of only 50% can be achieved (Scheme 1, left). The atom efficiency of such a process is low, as the ketone generated has little value and the recycling of this compound requires chemical reductive amination to produce the racemic amine for subsequent resolutions. Synthetically, an asymmetric synthesis is much more economical because yields of up to 100% are possible. However, in this case the (S)-amine would be produced (Scheme 1, right). An asymmetric synthesis strategy is thus clearly favored with respect to the atom efficiency of the process but has the major dis-advantage that only one specific enantiomer can be accessed with an enzyme having a distinct enantiopreference. As both enantiomers of a building block are often needed for targets with different abso-lute configuration, an enzyme platform providing either (R)- or (S)-specific enzymes is highly desired. Several strategies can be used to provide access to enzymes with complementary enantiopreference. This includes classical screening of strain collections for the identi-fication of complementary enzymes15. Indeed, (R)- and (S)-selective hydroxy nitrile lyases16, hydantoinases17 and keto reductases18 have been described. A much more promising option is the use of protein engineering tools19–22, with which changes in substrate specificity21, stability improvements23 and also switches in enantiopreference24–29 can be achieved. A recent example of the combination of rational protein design and directed evolution concepts is the inversion

of the enantiopreference of a Bacillus subtilis esterase by a double mutation identified in a library of only 2,800 variants30.

In this contribution, we have explored a third approach, in which we take advantage of the numerous protein sequences already deposited in databases. One practical advantage when compared with classical screening is that the complete nucleotide—and hence, protein—sequence is already known, and after identification of promising candidates, the gene can easily be cloned from the original source organism or obtained as a synthetic gene. The major chal-lenge is, however, the retrieval of desired sequences if no enzyme with desired enantiopreference has been described in the literature and the easy and ideal starting point for this in silico approach is thus missing. This is indeed the case for amine- pyruvate transaminases (referred to here as amine transaminases, EC 2.6.1.18): in contrast to various (S)-selective amine transaminases, only two enzymes with (R)-selectivity have been reported31–33, and at the submission of this study no information about nucleotide or amino acid sequences was available. In the meantime, the amino acid sequence of a previ-ously commercially available (R)-selective amine transaminase has been published34.

Transaminases belong to fold classes I and IV of pyridoxal-5′-phosphate (PLP)–dependent enzymes35,36. In contrast to α-amino acid aminotransferases (referred to throughout as α-transaminases, EC 2.6.1), which are ubiquitous enzymes found in all organisms, the small group of amine transaminases also converts substrates lack-ing an α-carboxylic acid moiety (Scheme 1). This makes them very attractive for organic synthesis of optically active amines37, especially as the product range is not limited to α-amino acids. Hence, there is an urgent need for the discovery of (R)-selective amine transami-nases to access a broad range of (R)-amines by efficient asymmetric synthesis (Scheme 1, right).

We first considered protein engineering to create an (R)-selective amine transaminase by inverting the enantiopreference of an (S)-selective amine transaminase (Fig. 1a), but there were several limi-tations, as no crystal structure of any amine transaminase is available and thus rational protein design is very difficult to perform.

One option would be to start from the crystal structure of an (S)-selective α-transaminase and reengineer the substrate-recognition site to create a variant that accepts substrates lacking the carboxylic

1Department of Biotechnology and Enzyme Catalysis, Institute of Biochemistry, Greifswald University, Greifswald, Germany. 2lonza AG, valais Works, visp, Switzerland. 3These authors contributed equally to this work. *e-mail: uwe.bornscheuer@uni-greifswald.de

rational assignment of key motifs for function guides in silico enzyme identificationMatthias Höhne1,3, sebastian schätzle1,3, Helge Jochens1, Karen robins2 & uwe t Bornscheuer1*

Biocatalysis has emerged as a powerful alternative to traditional chemistry, especially for asymmetric synthesis. One key requirement during process development is the discovery of a biocatalyst with an appropriate enantiopreference and enantio­selectivity, which can be achieved, for instance, by protein engineering or screening of metagenome libraries. We have devel­oped an in silico strategy for a sequence­based prediction of substrate specificity and enantiopreference. First, we used rational protein design to predict key amino acid substitutions that indicate the desired activity. Then, we searched protein databases for proteins already carrying these mutations instead of constructing the corresponding mutants in the laboratory. This meth­odology exploits the fact that naturally evolved proteins have undergone selection over millions of years, which has resulted in highly optimized catalysts. Using this in silico approach, we have discovered 17 (R)­selective amine transaminases, which catalyzed the synthesis of several (R)­amines with excellent optical purity up to >99% enantiomeric excess.

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

808 nature CHeMICaL BIOLOGY | vol 6 | NovEMBER 2010 | www.nature.com/naturechemicalbiology

article NaTUre chemical BiOlOgy dOI: 10.1038/nCHeMBIO.447

function (Fig. 1a), resulting in an (R)-selective amine transaminase (owing to the switch in priority according the Cahn-Ingold-Prelog (CIP) rule38). Unfortunately, the substrate’s α-carboxyl group plays an important role for the domain closure of the α-transaminase during substrate binding39,40. Thus, to solve this challenge, many additional mutations might be required, which could not be pre-dicted because of the complexity of the problem.

Instead of performing directed evolution to create and iden-tify an enzyme with the desired selectivity from a random mutant library, we developed a strategy to find enzymes with complemen-tary enantiopreference by searching in silico in protein databases (Fig. 1b). Essentially, the method is based on two steps: (i) identifi-cation and prediction of important amino acid residues on the basis of structural information from related enzymes and (ii) data mining

BA

BA

B

RSRLCOO–

NH2NH2 NH2

RLRS

a

Desired (R)-amine transaminase L-Branched chain transaminasePLP-dependent fold class IV

(S)-amine transaminasePLP-dependent fold class I

b Structural information

A Hypothesis of evolution

B Prediction of key mutations

C Annotation algorithm based on sequence motifs

E Cloning and expression ofidentified sequences

Desired enzyme

What would be a plausible ancestor for the evolution of an (R)-specific amine transaminase?

Which amino acid residues are involved in substrate recognition in the active site?Which amino acid exchanges are expected?

Can enzyme activities other than the desired be excluded?

Which sequences match the expected criteria?

Does the activity of the expressed protein match the predicted function?

D Database search

Putative (R)-selective amine transaminase

PLP-dependent fold class IV proteins

A

Figure 1 | Strategies for protein engineering. (a) Possible ancestors of amine transaminase with (R)-enantiopreference can be used to engineer an (R)-selective amine transaminase (center). This can be achieved by modification of the amino acids in the carboxyl group–binding pocket of an α-transaminase (left), such as an l-branched chain transaminase of the PlP-dependent fold class Iv, or by engineering of the binding pockets of an (S)-selective amine transaminase (right) from PlP-dependent fold class I. It was assumed that according to the CIP rule38, the large substituent (Rl) has a higher priority than the small substituent (RS). (b) Flow scheme of the in silico approach for the identification of transaminases with inverted enantiopreference, with steps A–E.

Scheme 1 | Strategies for the synthesis of optically active amines using amine transaminase. In a kinetic resolution (left), the amine transaminase converts in the ideal case only one of the amine enantiomers to the corresponding ketone. The remaining enantiomer can be isolated in high optical purity and at a (theoretical) maximum yield of 50%. In an asymmetric synthesis (right), a prostereogenic ketone is aminated enantioselectively, yielding directly the optically active amine. The most common cosubstrates for amine transaminases are pyruvate and alanine. As the equilibrium favors ketone formation, high yields in asymmetric synthesis can only be achieved by shifting the equilibrium, for example by enzymatic removal of the formed coproduct pyruvate49.

Kinetic resolution Asymmetric synthesis

O

RS

RS - small sized alkyl group

RL

RL - medium/large sized alkyl or aryl group

RS RL RS RL RS RL RS RL

NH2 NH2+

O

Amino acceptor,such as pyruvate

Coproduct,such as alanine

Amino donor,such as alanine

Coproduct,such as pyruvate

(S)-selective aminetransaminase

≤50% yield ≤100% yield

(S)-selective amine transaminase

(R)-amine (S)-amine

Coproduct removal

NH2

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nature CHeMICaL BIOLOGY | vol 6 | NovEMBER 2010 | www.nature.com/naturechemicalbiology 809

articleNaTUre chemical BiOlOgy dOI: 10.1038/nCHeMBIO.447

in protein sequences. This concept takes advantage of the currently available knowledge about the function of proteins combined with the experimentally uncharacterized huge diversity stored in protein sequence databases.

reSUlTSIn the first step, we carefully analyzed structural information of related enzymes to evaluate a possible hypothesis for the evolution of an (R)-selective amine transaminase (step A; see Fig. 1b for a schematic representation of the steps). Then, we predicted important positions and suggestions for amino acid exchanges (step B). Next, we per-formed identification of putative—and so far unknown—sequence motifs to exclude unwanted enzyme activities, and, as the key step, we developed an annotation algorithm on the basis of these sequence motifs (step C). Protein-coding sequences thus identi-fied in a database search (step D) as matching the predicted criteria were then cloned from synthetic genes. The resulting enzymes were investigated biochemically for transaminase activity and desired (R)-selectivity (step E).

analysis of structural informationTwo fold classes have been described for transaminases35. All α-transaminases belonging to fold class I of PLP-dependent enzymes known so far are L-selective, and no D-selective amino acid transami-nases have been described within this fold class. However, D-amino acid aminotransferases41 (DATAs) are members of fold class IV, and notably, L-branched chain amino acid aminotransferases (BCATs)42 belong to the same fold class43. This indicates that within this protein fold class there is a certain flexibility in the architecture of the active site with regard to substrate recognition, and therefore we focused our search on fold class IV aminotransferases. Apart from DATAs and BCATs, only 4-amino-4-deoxychorismate lyase (ADCL) is cur-rently known as a further member of fold class IV.

The opposite enantiopreference of DATAs and BCATs can be explained—via their crystal structures41,42—by the difference in sub-strate coordination in the active site, which consists of two binding

pockets. If the α-carboxyl group is positioned in binding pocket B (Fig. 1a), the enzyme converts L-amino acids and thus shows (S)-preference. If, however, the α-carboxyl group of the sub-strate is accommodated in binding pocket A, this transaminase is (R)-selective and thus converts D-amino acids.

This indicates that the desired diversity is perhaps already available in nature and that it only needs to be identified and exploited. Following this line of reasoning, an (R)-selective amine transaminase could already have been evolved from an (S)-selective α-transaminase if modifications of the α-carboxyl group–binding pocket had occurred, allowing small hydrophobic side chains to bind instead of the α-carboxyl functionality. The plausible ancestor of an (R)-specific amine transaminase must therefore be an L-(S)-selective BCAT, as, according to the CIP rule38, the substitution of the carboxyl group of an L-amino acid by a methyl group yields an (R)-amine.

Prediction of key features of the desired enzymeWorking from the known crystal structures, we studied the coordination of the α-carboxyl group in BCATs and DATAs. This allowed us to predict the necessary differences in protein sequences within PLP-dependent fold class IV proteins that determine the desired switch in substrate specificity from α-amino acids to amines, in line with the formal switch in enantiopreference. In the case of BCATs, the substrate binding in pocket B is realized in a more subtle manner than it is in DATAs and other α-transaminases (for more details see Supplementary Results and Supplementary Fig. 1). There is no direct contact of the carboxyl group oxygen atoms to any basic amino acid side chain such as that of an arginine (Fig. 2). Instead, one carboxyl oxygen atom is coordinated by the Tyr95 hydroxyl group, which is polarized by the coordination of an adjacent Arg97, and the other oxygen is bonded by two backbone amide nitrogen atoms of Thr262 and Ala263, which are activated by the coordination of their adjacent carbonyl groups by Arg40 (Fig. 2 and Supplementary Fig. 1; note that in related sequences a lysine is at position 40).

Thus, an amine transaminase should differ from BCATs in the following residues. First, Tyr95 should be exchanged with

Thr262

Asn127

O

N

RH

H

O

HN

O

Arg40

Met107

Val109

NH

HN NH

Arg97

O

Tyr95

H

Phe36

(S)-BCAT

O

Tyr97

H

Gln99

H2N

NH2

OX95

NH

NH

NH

+

++

+

+

O–

O–

–

–

–

+

–O OPO3H– PO3H– PO3H–

N+

RH

H

– O

O

Tyr36

Arg107

His109

X:Ile, Met, Ala, Ser, Thr, His

HN

O

H2N

Lys40

NHO

(R)-DATA

Gly38Val38

O O

NH

O

OO

O O

Arg107

Lys163

O

Thr38H

H

ADCL

31 95 107YGxxVFEGLK YIRxx…LGVA MR L M LS I V V

I

31 95 107FGDGIYEVIK XIYLQ…RxH S V VR L I Ax V M

F

31 95 107FGDGCFxTAx VxKVx…RGYYS L I G L V L I V

Lys97 Asn256

Thr262

Figure 2 | identification of key amino acid motifs that allow prediction of function of PlP­dependent fold class iV proteins. Top: schematic drawing of amino acids contributing to substrate binding in DATAs, BCATs and ADCls. The external aldimines, which are formed after binding and transimination of the respective substrates with the PlP cofactor of the enzymes, are shown. Blue amino acids are part of binding pocket A (Fig. 1a), and amino acids of binding pocket B (Fig. 1a) are shown in green. The gray shaded circle represents the glycine or valine, depending on the transaminase, that is located behind the carboxyl group. The red threonine residue is important in the catalytic mechanism for shuttling of a proton during the reaction transition state. Bottom: sequence motifs derived from multiple-sequence alignments are given using the same color code, showing that amino acids important for substrate binding in the active site are rather conserved and can be used for a prediction of the substrate specificity of PlP-dependent fold class Iv enzymes as well as for the enantiopreference of the transaminases within this fold class.

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

810 nature CHeMICaL BIOLOGY | vol 6 | NovEMBER 2010 | www.nature.com/naturechemicalbiology

article NaTUre chemical BiOlOgy dOI: 10.1038/nCHeMBIO.447

a hydrophobic residue that is incapable of forming a hydrogen bond with the carboxyl group. Second, Arg40 should be changed to a residue that cannot activate the amide backbone nitrogens of Thr262 and Ala263 by coordination of the neighboring backbone carbonyl oxygens. However, the situation regarding position 40 in PLP-dependent fold class IV proteins is complex. DATAs also have a basic amino acid, Lys40, but in contrast to the Arg40 in BCATs, Lys40 in DATAs adopts an altered conformation so that its ε-amino group forms a hydrogen bond to the backbone of an adjacent loop (Fig. 2 and Supplementary Fig. 1). Thus, in spite of the presence of Lys40 in DATAs, the activation of the amide backbone nitrogens, critical for carboxyl-group binding in binding pocket B, is pre-vented. From this observation, we concluded that if an Arg40 or Lys40 is found in one of the sequences in a fold class IV protein, whether this residue facilitates the binding of a substrate carrying an α-carboxyl group or not cannot be predicted. In summary, the pres-ence of a hydrophobic amino acid in position 95 and the absence of an arginine or lysine residue at position 40 would be a clear hint of altered substrate specificity toward amines.

Design and application of a sequence­based algorithmNext, we developed a sequence-based prediction of the substrate specificity—conversion of amines versus α-D- or α-L-amino acids—to identify putative (R)-amine transaminases within PLP-dependent fold class IV proteins. Aside from the key residues considered important for amine transaminase activity, we com-pared residues involved in substrate coordination in the different enzymes. To simplify the structural description, we introduced a general numbering scheme for the amino acid residues of the different PLP-dependent fold class IV proteins. This was based on a

multiple-sequence alignment (Supplementary Fig. 2). Fortunately, the amino acids that are in direct contact with the substrate in the active site are arranged in two relatively short sequence blocks. The first block is located at positions 36–40, and the second block comprises six amino acids at positions 95–97 and 107–109. Most of these amino acids fold into a β-sheet; only residues 107–109 are part of a loop. This information is important because during evolution, insertions or deletions take place preferentially in loop sequences without destroying enzyme activity but do not usually take place in α-helices or β-sheets44. Thus, it was observed that resi-dues contributing to substrate recognition of a PLP-dependent fold class IV enzyme aligned well within the different enzymes, except in the motif at positions 107–109, where we sometimes observed insertions or deletions of one to two amino acids. Alignments, which include all known BCAT, DATA and ADCL proteins with experimentally verified enzyme activity, showed that the amino acids involved in substrate recognition in the active site seem to be quite conserved in these three groups. This allowed us to formu-late different sequence motifs characteristic of DATA, BCAT and ADCL activity (Fig. 2; for a summary of the structure- function relationship of individual amino acids of the sequence motifs see Supplementary Tables 1 and 2; for the multiple-sequence align-ments see Supplementary Figs. 3–5). On the basis of these com-parative considerations, an annotation algorithm was developed (Supplementary Fig. 6) using the amino acid sequence motifs identified. This enabled an easy exclusion of all enzyme candidates that could clearly be designated as BCATs, ADCLs or DATAs. The analysis of the remaining sequences aimed at identifying trans-aminase sequences that fulfilled the requirements for the desired (R)-selective amine transaminase activity.

Table 1 | Section of a multiple­sequence alignment of putative (R)­selective amine transaminases. entry, GeneId

Organism

sequence motif 1 31 36 40

sequence motif 2 95 99 107

1 ecDATA Bacillus sp. 26 FGDGVYEVVKVYN 77 HIYFQVTRGTSPRAHQFP

2 ecBCAT Eschericha coli 26 YGTSVFEGIRCYD 86 YIRPLIFVGDVGMGVNPP

3 ecADCl Eschericha coli 21 FGDGCFTTARVID 78 VLKVVISRGSGGRGYSTL

4 115385557 Aspergillus terreus 55 HSDLTYDVPSVWD 106 FVELIVTRGLKGVRGTRP

5 211591081 Penicillium chrysogenum 53 HSDLTYDVPSVWD 104 FVEIIVTRGLKGVRGSRP

6 145258936 Aspergillus niger 53 RSDLTYDVISVWD 104 YVALIVTRGLQSVRGAKP

7 169768191 Aspergillus oryzae 53 HSDLTYDVPSVWD 104 FVELIVTRGLKGVRGNKP

8 70986662 Aspergillus fumigatus 53 HSDLTYDVISVWD 104 FVEVIVTRGLTGVRGSKP

9 119483224 Neosartorya fischeri 53 HGDLTYDVTTVWD 104 FVEVIVTRGLTGVRGSKP

10 46109768 Gibberella zeae 53 HGDLTYDVPAVWD 104 FVELIVTRGLKPVREAKP

11 114797240 Hyphomonas neptunium 53 HSDLTYDVPAVWN 104 YVEIIVTRGLKFLRGAQA

12 120405468 Mycobacterium vanbaalenii 69 HSDLTYTVAHVWH 120 FVNLTITRGYGKRKGEKD

13 13471580 Mesorhizobium loti 54 HSDATYDTVHVWN 105 YVEMLCTRGASPTFSRDP

14 20804076 Mesorhizobium loti 53 HSDATYDTVHVWE 104 YVEMICTRGGSPTFSRDP

15 86137542 Roseobacter sp. 45 HSDATYDVAHVWK 96 YVEFICTRGTSPTFSRDP

16 87122653 Marinomonas sp. 44 HSDATYDVVHVWQ 95 YVEMICTRGNSPDFSRDP

17 190895112 Rhizobium etli 38 RSDACQDTVSVWD 90 YVQIIMTRGRPPIGSRDL

18 89899273 Rhodoferax ferrireducens 34 RSDATYDVVTVWD 85 YVEMICTRGQPPWGSRDP

19 89053613 Jannaschia sp. 32 HSDIAYDVVPVWR 83 YVAMVAARGRNPVPGSRD

20 EEE43073 Labrenzia alexandrii 42 HSDITYDVVPVLD 93 YVAMVTSRGVNQVPGSRD

21 78059900 Burkholderia sp. 53 HADAAYDVVTVSR 104 YVWWCVTRGPLSVDRRDR

22 ABK12047 Burkholderia cenocepacia 48 HSDVTYDTVHVWN 99 YVEMLCTRGVSPTFSRDP

23 ZP_01448442 Alpha proteobacterium 22 HSDATYDVAHVWG 73 YVEFICTRGTSPNFSRDP

24 219677744 Gamma proteobacterium 30 LGDGVFDVVSAWK 81 SIRFIVTRGEPKGVVADP

The first three entries are the DATA, BCAT and ADCl sequences used for the identification of sequence motifs. Entries 4–24 refer to the newly discovered enzymes.

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nature CHeMICaL BIOLOGY | vol 6 | NovEMBER 2010 | www.nature.com/naturechemicalbiology 811

articleNaTUre chemical BiOlOgy dOI: 10.1038/nCHeMBIO.447

Using this algorithm, we analyzed about 5,700 sequences anno-tated as BCATs and 280 protein sequences annotated as PLP-dependent fold class IV proteins from the US National Center for Biotechnology Information (NCBI) Protein Database. This search identified 21 sequences, 7 from eukaryotes and 14 from prokaryotes (Table 1). These sequences matched our criteria for the selection of putative (R)-selective amine transaminases with the predicted enantiopreference: the absence of arginine or lysine at position 40 in all sequences and a Phe95 instead of a Tyr95 in 8 out of the 21 sequences indicated that substrates having a small alkyl group instead of a carboxyl group in the α-position might be bound to the active site. In the other 13 sequences, a putative BCAT activity as suggested by a Tyr95 residue is unlikely because of the presence of alanine, glutamine, aspartic acid or glutamic acid at residue 97 instead of the required polarizing arginine. Thus, the coordination of a small alkyl group in binding pocket B seemed much more likely than binding of an α-carboxyl group, which we considered strong evidence that the respective proteins are amine transaminases rather than BCATs. To convey (R)-selectivity, binding pocket B should only allow the binding of a small alkyl group. The fact that in posi-tion 95 a tyrosine or phenylalanine is found suggests that the size of the binding pocket is not large compared to BCATs, as this could be obtained by mutation of Tyr95 to a small hydrophobic amino acid, such as alanine or valine.

Thus, we considered the criteria for the desired switch in sub-strate specificity toward (R)-amines to be fulfilled. It is noteworthy that sometimes the annotated function of a protein included in a database did not match the prediction made by our sequence motif approach. For example, out of 26 proteins annotated in the curated

NCBI database as DATAs, 7 could clearly be identified as BCATs and one as an (R)-amine transaminase (Supplementary Fig. 7).

confirmation of predicted activity and enantiopreferenceIn the next step we ordered all the putative (R)-selective amine transaminase genes as codon-optimized sequences for expression in Escherichia coli. They were subcloned and expressed in E. coli BL21 and underwent His-tag purification. Then we investigated the puri-fied enzymes with respect to activity, enantioselectivity and enantio-preference toward a range of amines (1–4) (Fig. 3). Additionally, we also examined whether they possessed BCAT or DATA activity (conversion of d-alanine 5 or l-glutamate 6 with concomitant for-mation of valine). Seventeen of the 21 putative (R)-selective amine transaminase genes that had been identified were found to be (R)-selective amine transaminases (Fig. 3 and Supplementary Tables 3 and 4). Three proteins could not be expressed in E. coli in sufficient amounts, and one protein was found to have very low activity on the substrates studied. Specific activity and substrate range differed greatly among all of the enzymes investigated, which we find unsur-prising considering that the protein sequences originate from various microorganisms, that recombinant expression was never reported before and that specific activities within their natural functions as well as their natural substrates are unknown. Nevertheless, 10 out of the 21 proteins showed a specific activity >0.5 U mg−1 toward at least one of the investigated amines (Fig. 3), which is in the same activ-ity range of known (S)-selective amine transaminases45–47. (One unit (U) of activity was defined as the amount of enzyme that produced 1 μmol ketone product per minute.)

Consequently, these newly identified (R)-selective transami-nases can be used in asymmetric synthesis to yield optically pure (R)-amines. This was confirmed in preliminary experiments for the asymmetric synthesis of 2-aminohexane (2), 2-amino-4- phenylbutane (3), 1-N-Boc-3-aminopyrrolidine (4a) and 1-N-Boc-3-aminopiperidine (4b) from the corresponding ketones with three of the amine transaminases and resulted in low to moder-ate yields with excellent enantiomeric excess (ee) up to 99.6% (see Supplementary Table 5).

DiScUSSiONThe strength of this new in silico approach is that new enzymes can be discovered very quickly. In contrast, directed evolution requires several rounds of random mutagenesis or iterative saturation muta-genesis to alter the enantioselectivity or substrate specificity, even if structural information is available to identify hot spots for muta-genesis. Usually more than 103–104 variants have to be screened for the identification of a mutant with desired properties. In contrast to this very low ‘hit rate’, at least 50% of the putative proteins identified in our study turned out to be useful biocatalysts.

Hence, before creating mutants or libraries stemming from rational predictions in the laboratory, it is worthwhile to investigate whether nature has already designed variants and possibly opti-mized these catalysts over millions of years of natural evolution.

The newly identified (R)-selective amine transaminases are an ideal starting point for further fine-tuning and optimization by pro-tein engineering, as the requirements for industrial processes are often different from the enzyme’s original function in nature. This was demonstrated in a recent study in which the substrate specific-ity of an (R)-amine transaminase was broadened so that bulky sub-strates could also be converted34. Also, other important properties, such as stability at higher temperatures and tolerance of high cosol-vent, substrate and product concentration, can be improved signifi-cantly and can facilitate the application of amine transaminases for industrial-scale biotransformations.

In the example shown here, we took advantage of the enormous diversity of structures, elucidated mechanisms and substrate spec-ificities already reported for PLP-dependent proteins. Although

N

Boc

H2N

H2NH2NH2NH2N

O

PyruvatePyruvatePyruvateR S RSR SR S

+ +Pyruvate L-Glutamate

+ + + +

OHO HO O

Aspergillus terreus

Penicillium chrysogenum

Aspergillus oryzae

Aspergillus fumigatus

Neosartorya fischeri

Gibberella zeae

Hyphomonas neptunium

Mycobacterium vanbaalenii

Mesorhizobium loti

Mesorhizobium loti

Roseobacter sp.

Marimonas sp.

Rhizobium etli

Rhodoferax ferrireducens

Jannaschia sp.

Labrenzia alexandrii

Burkholderia sp.

Gamma proteobacterium

4

5

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

24

α-KG

1 2 3 4 5 6

Specific activity [U mg–1]: 5 1 0.5 0.2 0.1 0.05 0.01 0.002 0

Figure 3 | characterization of the discovered (R)­selective amine transaminases. Specific activities toward (R) and (S) enantiomers of amines 1–4 and α-amino acids 5 (d-alanine) and 6 (l-glutamate) are indicated by a color gradient. The numbering of the proteins corresponds to that given in Table 1 (details on specific activity and expression levels are given in Supplementary Tables 2 and 3). α-KG, α-ketoglutarate.

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

812 nature CHeMICaL BIOLOGY | vol 6 | NovEMBER 2010 | www.nature.com/naturechemicalbiology

article NaTUre chemical BiOlOgy dOI: 10.1038/nCHeMBIO.447

such diversity is not yet explored in depth for all enzyme classes, the possibilities for the discovery of new catalysts using this concept will steadily increase on a daily basis as the number of protein sequences and solved structures available in the databases grows.

A second important requirement for the successful application of in silico enzyme discovery is a reliable prediction of the nece ssary key amino acid differences that determine the desired substrate specificity. In this study, the molecular mechanism for the change in substrate specificity from binding a carboxyl group to a small alkyl group—and the related predicted, desired switch in enantioprefer-ence—could be easily rationalized. Method develop ment for under-standing the molecular basis and prediction of substrate specificity involving homology modeling and molecular docking are important research fields. Additionally, the application of more complex search algorithms and bioinformatics tools will help to refine such in silico enzyme discovery and facilitate its application in cases in which more complex solutions might be required.

meThODSAlignments. All pairwise and multiple amino acid sequence alignments were done with the computer program STRAP using the ClustalW3D algorithm with stand-ard parameters. The protein sequences annotated as BCATs or PLP-dependent fold class IV enzymes from the NCBI protein database were used for the database search and aligned to E. coli BCAT.

Cloning and expression of amine transaminase. The codon-optimized open read-ing frames (ORFs) encoding proteins 4, 5, 6, 7, 11, 12, 14, 16, 17, 18, 20, 21 and 24 (Table 1) were inserted into pGASTON between the NdeI and BamHI restriction sites. The codon-optimized ORFs encoding all other proteins were ordered subcloned in pET-22b. Transformed E. coli BL21 (DE3) strains were grown in 400 ml LB medium supplemented with ampicillin (100 μg ml−1). Cells were incubated initially at 37 °C on a gyratory shaker until the OD600 reached 0.7. The cells were then induced by addition of 0.2% (w/v) rhamnose (pGASTON) or 0.1 mM IPTG (pET-22b), respectively. At the same time the incubation temperature was decreased to 20 °C, and cultivation continued for 20 h.

Purification of proteins. The cell pellet (~3 g wet weight) was washed twice with phosphate buffer (pH 7.5, 50 mM) containing 0.1 mM PLP at 4 °C. After disrup-tion (French press), the cell suspension was centrifuged (10,000g, 30 min), and the resulting supernatant was passed through a 0.5-μm filter before chromatography. Chromatography was performed using an ÄKTA Purifier (GE Healthcare). The filtered cellular extract was applied to a 5-ml column of IMAC Sepharose 6 Fast Flow (GE Healthcare). The column was washed at a flow rate of 5 ml min−1 with 10 column-volumes of phosphate buffer (pH 7.5, 50 mM, containing 300 mM NaCl, 0.1 mM PLP and 30 mM imidazole to avoid nonspecific binding), and the active protein was eluted with 10 column-volumes of phosphate buffer (pH 7.5, 50 mM, containing 300 mM NaCl, 0.1 mM PLP and 300 mM imidazole at a flow rate of 5 ml min−1). The fractions with the desired protein were pooled and desalted via gel chromatography with a 20 mM tricine buffer, pH 7.5, containing 0.01 mM PLP. The purified enzymes were stored at 4 °C. Protein concentrations were determined using the BCA assay kit (Uptima) after gel chromatography.

Characterization of substrate specificity. For an initial confirmation of activity, α-methylbenzyl amine (1, α-MBA) served as substrate in an acetophenone-based microplate assay47: a solution of 2.5 mM (R)- or (S)-α-MBA and pyruvate was reacted in the presence of the purified enzyme, and the increase in absorbance at 245 nm was correlated to the formation of acetophenone. The conversions of the other amines (2–4) were monitored using a conductivity assay48: a solution containing 10 mM amine and pyruvate was reacted in the presence of the purified amine transaminase, and the decrease in conductivity was related to the conver-sion of substrate. For all measurements, either an appropriate amount of purified enzyme was applied, dependent on the specific activity, or the highest concentra-tion possible was applied (0.3–2.2 mg ml−1 purified enzyme, dependent on expres-sion and purification) while measuring low activities.

To verify DATA or BCAT activity, the decrease of NADH was measured spectrophotometrically at 340 nm using dehydrogenase-coupled microplate assays: a solution of 5 mM α-ketoglutaric acid and D-alanine 5 was reacted in the presence of the purified transaminase, and 1 U ml−1 lactate dehydrogenase and 0.5 mM NADH were used for measuring DATA activity. For measuring BCAT activity, a solution containing 5 mM 3-methyl-2-oxobutyric acid and L-glutamate 6, 10 mM ammonium chloride, 1 U ml−1 glutamate dehydrogenase and 0.5 mM NADH was used. All reactions took place in tricine buffer (pH 7.5, 20 mM) containing 0.01 mM PLP. The pH of the buffer was adjusted with 1,8-diazabicyclo[5.4.0]undec-7-ene. The specific activity was expressed as units per milligram protein, and one unit of activity was defined as the amount of enzyme that produced 1 μmol ketone product per minute.

Asymmetric synthesis of amines 1–4. Preliminary asymmetric syntheses were per-formed at 30 °C for 24 h in sodium phosphate buffer (100 mM, pH 7) containing PLP (1 mM) and NAD+ (1 mM) in 1.5-ml Eppendorf tubes. The reaction mixture contained 50 mM ketone, L-alanine (5 equiv., 250 mM), lactate dehydrogenase from bovine heart (90 U), glucose (150 mM) and glucose dehydrogenase (15 U). Amine transaminases from Aspergillus terreus, Mycobacterium vanbaalenii and Mesorhizobium loti (entries 4, 12 and 14 in Table 1) were expressed in E. coli BL21 as described above, frozen in aliquots and applied directly as whole-cell biocata-lysts (~0.05 g wet cell weight per ml) without further purification. The conversion was measured by detection of the formed amines (1, gas chromatography; 2–4, capillary electrophoresis). Chiral analysis of 2–4 was performed using capillary electrophoresis as reported previously49. The percent enantiomeric excess values for 1 were analyzed by gas chromatography. After extraction of the amine with ethyl acetate, derivatization to the trifluoroacetamide was performed by adding a 20-fold excess of trifluoroacetic acid anhydride. After purging with nitrogen to remove excess anhydride and residual trifluoroacetic acid, the derivatized compound was dissolved in ethyl acetate (50 μl) and baseline separated using a Shimadzu GC14A that was equipped with a heptakis-(2,3-di-O-acetyl-6-O-tert-butyldimethylsilyl)-β-cyclodextrin column (25 m by 0.25 mm). The retention times were 16.0 min ((S)-1) and 16.2 min ((S)-2) using the following oven temperature program: 80 °C for 10 min, heating with 20 °C per min to 180 °C, maintained at 180 °C for a further 10 min.

received 23 November 2009; accepted 23 august 2010; published online 26 September 2010

references1. Bornscheuer, U.T. & Kazlauskas, R.J. Hydrolases in Organic Synthesis

(Wiley-VCH, Weinheim, Germany, 2005).2. Breuer, M. et al. Industrial methods for the production of optically active

intermediates. Angew. Chem. Int. Edn Engl. 43, 788–824 (2004).3. Buchholz, K., Kasche, V. & Bornscheuer, U.T. Biocatalysis and Enzyme

Technology (Wiley-VCH, Weinheim, Germany, 2005).4. Schmid, A. et al. Industrial biocatalysis today and tomorrow. Nature 409,

258–268 (2001).5. Schoemaker, H.E., Mink, D. & Wubbolts, M.G. Dispelling the myths—Biocatalysis

in industrial synthesis. Science 299, 1694–1697 (2003).6. Bommarius, A.S. & Riebel, B.R. Biocatalysis: Fundamentals and Applications

(Wiley-VCH, Weinheim, Germany, 2004).7. Grunwald, P. Biocatalysis: Biochemical Fundamentals and Applications

(Imperial College Press, 2009).8. Liese, A., Seelbach, K. & Wandrey, C. (eds.) Industrial Biotransformations

(Wiley-VCH, Weinheim, London, UK, 2006).9. Patel, R.N. (ed.) Biocatalysis in the Pharmaceutical and Biotechnology

Industries (CRC Press, Boca Raton, Florida, USA, 2006).10. Pollard, D.J. & Woodley, J.M. Biocatalysis for pharmaceutical intermediates:

the future is now. Trends Biotechnol. 25, 66–73 (2007).11. Straathof, A.J.J., Panke, S. & Schmid, A. The production of fine chemicals by

biotransformations. Curr. Opin. Biotechnol. 13, 548–556 (2002).12. Tao, J., Lin, G.-Q. & Liese, A. (eds.) Biocatalysis for the Pharmaceutical

Industry: Discovery, Development, and Manufacturing (Wiley-VCH, Weinheim, Germany, 2009).

13. Hou, C.T. (ed.). Handbook of Industrial Biocatalysis (CRC Press, Boca Raton, Florida, USA, 2005).

14. Faber, K. Biotransformations in Organic Chemistry (Springer-Verlag, Berlin, 2004).15. Mugford, P.F., Wagner, U.G., Jiang, Y., Faber, K. & Kazlauskas, R.J.

Enantiocomplementary enzymes: classification, molecular basis for their enantiopreference, and prospects for mirror-image biotransformations. Angew. Chem. Int. Edn Engl. 47, 8782–8793 (2008).

16. Griengl, H., Schwab, H. & Fechter, M. The synthesis of chiral cyanohydrins by oxynitrilases. Trends Biotechnol. 18, 252–256 (2000).

17. Cheon, Y.-H. et al. Crystal structure of d-hydantoinase from Bacillus stearothermophilus: insight into the stereochemistry of enantioselectivity. Biochemistry 41, 9410–9417 (2002).

18. Gröger, H. et al. Enantioselective reduction of ketones with designer cells at high substrate concentrations: highly efficient access to functionalized optically active alcohols. Angew. Chem. Int. Edn Engl. 45, 5677–5681 (2006).

19. Fox, R.J. et al. Improving catalytic function by ProSAR-driven enzyme evolution. Nat. Biotechnol. 25, 338–344 (2007).

20. Kazlauskas, R.J. & Bornscheuer, U.T. Finding better protein engineering strategies. Nat. Chem. Biol. 5, 526–529 (2009).

21. Lutz, S. & Bornscheuer, U.T. (eds.) Protein Engineering Handbook (Wiley VCH, Weinheim, Germany, 2009).

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

23. Reetz, M.T., Carballeira, J.D. & Vogel, A. Iterative saturation mutagenesis on the basis of B factors as a strategy for increasing protein thermostability. Angew. Chem. Int. Edn Engl. 45, 7745–7751 (2006).

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nature CHeMICaL BIOLOGY | vol 6 | NovEMBER 2010 | www.nature.com/naturechemicalbiology 813

articleNaTUre chemical BiOlOgy dOI: 10.1038/nCHeMBIO.447

24. Ivancic, M., Valinger, G., Gruber, K. & Schwab, H. Inverting enantioselectivity of Burkholderia gladioli esterase EstB by directed and designed evolution. J. Biotechnol. 129, 109–122 (2007).

25. Koga, Y., Kato, K., Nakano, H. & Yamane, T. Inverting enantioselectivity of Burkholderia cepacia KWI-56 lipase by combinatorial mutation and high-throughput screening using single-molecule PCR and in vitro expression. J. Mol. Biol. 331, 585–592 (2003).

26. Magnusson, A.O., Takwa, M., Hamberg, A. & Hult, K. An S-selective lipase was created by rational redesign and the enantioselectivity increased with temperature. Angew. Chem. Int. Edn Engl. 44, 4582–4585 (2005).

27. May, O., Nguyen, P.T. & Arnold, F.H. Inverting enantioselectivity by directed evolution of hydantoinase for improved production of l-methionine. Nat. Biotechnol. 18, 317–320 (2000).

28. Williams, G.J., Woodhall, T., Farnsworth, L.M., Nelson, A. & Berry, A. Creation of a pair of stereochemically complementary biocatalysts. J. Am. Chem. Soc. 128, 16238–16247 (2006).

29. Zha, D.X., Wilensek, S., Hermes, M., Jaeger, K.E. & Reetz, M.T. Complete reversal of enantioselectivity of an enzyme-catalyzed reaction by directed evolution. Chem. Commun. (Camb.) 2664–2665 (2001).

30. Bartsch, S., Kourist, R. & Bornscheuer, U.T. Complete inversion of enantioselectivity towards acetylated tertiary alcohols by a double mutant of a Bacillus subtilis esterase. Angew. Chem. Int. Edn Engl. 47, 1508–1511 (2008).

31. Iwasaki, A., Yamada, Y., Ikenaka, Y. & Hasegawa, J. Microbial synthesis of (R)- and (S)-3,4-dimethoxyamphetamines through stereoselective transamination. Biotechnol. Lett. 25, 1843–1846 (2003).

32. Matcham, G.W. & Bowen, A.R.S. Biocatalysis for chiral intermediates: Meeting commercial and technical challenges. Chim. Oggi 14, 20–24 (1996).

33. Koszelewski, D., Lavandera, I., Clay, D., Rozzell, D. & Kroutil, W. Asymmetric synthesis of optically pure pharmacologically relevant amines employing ω-transaminases. Adv. Synth. Catal. 350, 2761–2766 (2008).

34. Savile, C.K. et al. Biocatalytic asymmetric synthesis of chiral amines from ketones applied to sitagliptin manufacture. Science 329, 305–309 (2010).

35. Jansonius, J.N. Structure, evolution and action of vitamin B6-dependent enzymes. Curr. Opin. Struct. Biol. 8, 759–769 (1998).

36. Percudani, R. & Peracchi, A. The B6 database: a tool for the description and classification of vitamin B6-dependent enzymatic activities and of the corresponding protein families. BMC Bioinformatics 10, 273 (2009).

37. Höhne, M. & Bornscheuer, U.T. Biocatalytic routes to optically active amines. ChemCatChem 1, 42–51 (2009).

38. Cahn, R.S., Ingold, C. & Prelog, V. Specification of molecular chirality. Angew. Chem. Int. Edn Engl. 5, 385–415 (1966).

39. Mizuguchi, H. et al. Strain is more important than electrostatic interaction in controlling the pK(a) of the catalytic group in aspartate aminotransferase. Biochemistry 40, 353–360 (2001).

40. Okamoto, A., Nakai, Y., Hayashi, H., Hirotsu, K. & Kagamiyama, H. Crystal structures of Paracoccus denitrificans aromatic amino acid aminotransferase: A substrate recognition site constructed by rearrangement of hydrogen bond network. J. Mol. Biol. 280, 443–461 (1998).

41. Sugio, S., Petsko, G.A., Manning, J.M., Soda, K. & Ringe, D. Crystal structure of a d-amino acid aminotransferase: how the protein controls stereoselectivity. Biochemistry 34, 9661–9669 (1995).

42. Goto, M., Miyahara, I., Hayashi, H., Kagamiyama, H. & Hirotsu, K. Crystal structures of branched-chain amino acid aminotransferase complexed with glutamate and glutarate: true reaction intermediate and double substrate recognition of the enzyme. Biochemistry 42, 3725–3733 (2003).

43. Mehta, P.K., Hale, T.I. & Christen, P. Aminotransferases: demonstration of homology and division into evolutionary subgroups. Eur. J. Biochem. 214, 549–561 (1993).

44. Wolf, Y., Madej, T., Babenko, V., Shoemaker, B. & Panchenko, A.R. Long-term trends in evolution of indels in protein sequences. BMC Evol. Biol. 7, 19 (2007).

45. Hanson, R.L. et al. Preparation of (R)-amines from racemic amines with an (S)-amine transaminase from Bacillus megaterium. Adv. Synth. Catal. 350, 1367–1375 (2008).

46. Shin, J.-S., Yun, H., Jang, J.-W., Park, I. & Kim, B.-G. Purification, characterization, and molecular cloning of a novel amine:pyruvate transaminase from Vibrio fluvialis JS17. Appl. Microbiol. Biotechnol. 61, 463–471 (2003).

47. Schätzle, S., Höhne, M., Redestad, E., Robins, K. & Bornscheuer, U.T. Rapid and sensitive kinetic assay for characterization of ω-transaminases. Anal. Chem. 81, 8244–8248 (2009).

48. Schätzle, S., Höhne, M., Robins, K. & Bornscheuer, U.T. A conductometric method for the rapid characterization of the substrate specificity of amine-transaminases. Anal. Chem. 82, 2082–2086 (2010).

49. Höhne, M., Kühl, S., Robins, K. & Bornscheuer, U.T. Efficient asymmetric synthesis of chiral amines by combining transaminase and pyruvate decarboxylase. ChemBioChem 9, 363–365 (2008).

author contributionsK.R. and U.T.B. initiated the project. M.H. designed the in silico strategy, devised the an-notation algorithm and performed the database search and identification of the putative amine transaminases. M.H. expressed and confirmed amine transaminase activity and (R)-selectivity for the first three proteins. S.S. coordinated the comparative characteriza-tion of all proteins and performed cloning, expression, purification, data collection and data analysis. H.J. contributed to gene cloning, protein expression and activity measure-ments. U.T.B. and M.H. cowrote the paper, and all authors read and edited the manuscript.

Competing financial interestsThe authors declare competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/naturechemicalbiology/.

additional informationSupplementary information is available online at http://www.nature.com/ naturechemicalbiology/. Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/. Correspondence and requests for materials should be addressed to U.T.B.

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

The Articles

- 75 -

ARTICLE V

DOI: 10.1002/adsc.201100435

Enzymatic Asymmetric Synthesis of Enantiomerically PureAliphatic, Aromatic and Arylaliphatic Amines with(R)-Selective Amine Transaminases

Sebastian Sch�tzle,a Fabian Steffen-Munsberg,a Ahmad Thontowi,a

Matthias Hçhne,a Karen Robins,b and Uwe T. Bornscheuera,*a Institute of Biochemistry, Department of Biotechnology and Enzyme Catalysis, Greifswald University, Felix

Hausdorff-Str. 4, 17487 Greifswald, GermanyFax: (+49)-3834-86-794367; e-mail: uwe.bornscheuer@uni-greifswald.de

b Lonza AG, Valais Works, 3930 Visp, Switzerland

Received: May 29, 2011; Published online: August 25, 2011

Supporting information for this article is available on the WWW underhttp://dx.doi.org/10.1002/adcs.201100435.

Abstract: Seven (R)-selective amine transaminases(R-ATAs) recently discovered by an in silico-basedapproach in sequence databases were produced re-combinantly in Escherichia coli and subjected to par-tial purification by ammonium sulfate precipitation.A range of additives and various buffers were inves-tigated to identify best conditions to ensure goodstorage stability and stable activity during biocataly-sis. All enzymes show pH optima between pH 7.5–9.These R-ATAs were then applied in the asymmetricsynthesis of twelve aliphatic, aromatic and arylali-phatic (R)-amines starting from the correspondingprochiral ketones using a lactate dehydrogenase/glu-

cose dehydrogenase system to shift the equilibrium.For all ketones, at least one enzyme was found thatallows complete conversion to the correspondingchiral amine having excellent optical purities >99%ee. Variations in substrate profiles are also discussedbased on the phylogenetic relationships between theseven R-ATAs. Thus, we have identified a versatiletoolbox of (R)-amine transaminases showing remark-able properties for application in biocatalysis.

Keywords: chiral amines; enantioselectivity; enzymecatalysis; transaminases

Introduction

Optically active amines play an important role in thepharmaceutical, agrochemical and chemical industry.They are frequently used as building blocks for thesynthesis of various biological active compounds usedas drugs or agrochemicals including, for example,drugs for the treatment of Alzheimer�s disease[1] andmalaria[2] as well as prostate drugs[3] and antitumorantibiotics.[4] As both enantiomers of a building blockare often needed for targets with different absoluteconfiguration, an enzyme platform for the productionof both (R)- and (S)-enantiomers is highly desired.

In principle, various enzymatic routes can be em-ployed for the synthesis of optically active primaryamines using hydrolases, oxidoreductases or transfer-ases (for further reading see review[5]). All routeshave their advantages and disadvantages, but onlyamine transaminases offer the unique possibility tosynthesize enantiomerically pure amines directly from

prostereogenic ketones. Furthermore, these enzymesusually show excellent enantioselectivity and thusthey are in great demand. In principle, both enantio-mers of a given amine can be produced with a singleenantioselective amine transaminase having, for in-stance, (S)-enantiopreference. A kinetic resolutionwith this enzyme would leave the (R)-enantiomerbehind, which can be obtained at maximum in 50%yield as the (S)-enantiomer is converted into theketone. In turn, asymmetric synthesis starting fromthe corresponding prochiral ketone would yield the(S)-amine as product in theoretically 100% yield ifthe equilibrium is efficiently shifted towards productformation for which a range of methods are avail-able.[6] Whereas many (S)-selective amine transami-nase (S-ATAs) had been discovered and extensivelyinvestitaged, only a few (R)-amine transaminases (R-ATAs) were known until recently and only oneenzyme (ATA-117, Codexis) was commercially avail-able. During an extensive project aiming for an R-

Adv. Synth. Catal. 2011, 353, 2439 – 2445 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 2439

FULL PAPERS

ATA for the industrial synthesis of the drug Sitaglip-tin, Codexis and Merck & Co created a variant ofATA-117, which now allows synthesis of the drug onindustrial scale.[7] Thus, an efficient asymmetric syn-thesis of (R)-amines was mostly hampered by the lim-ited access to R-ATAs. Recently, we identified 20(R)-amine transaminases by an in silico search strat-egy in which careful analysis of typical motifs for (S)-or (R)-selective amino acid transaminases (BCAT,branched chain amino acid transaminase; DATA, d-amino acid transaminase) and certain lyases from thesame fold class led to a complex algorithm.[8] This al-gorithm was then used in an alignment of 5,000 se-quences deposited in public databases to filter out allknown enzymes with BCAT, DATA or lyase activityand certain additional criteria to ensure that the re-maining sequences are true amine transaminases andthat they exhibit (R)-enantiopreference. The proteinsencoded by the remaining 20 sequences were allcloned, functionally expressed in E. coli and in pre-liminary analyses, we could demonstrate that 17 en-zymes were indeed amine transaminases and all had(R)-selectivity (three enzymes could not be expressedat sufficiently high levels) as confirmed by determin-ing specific activities towards different amine enantio-mers.

In this contribution, we now report further analysisof seven of these R-ATAs, which includes expressionoptimization to produce the enzyme in sufficientamounts, formulation to achieve stable biocatalystsand the determination of pH profiles. Special focus isgiven on the application of the enzymes in asymmet-ric synthesis to identify their potential in industrialbiocatalysis and hence suitable methods to shift theequilibrium needed to be identified too.

Results and Discussion

In the present study we utilized a toolbox consistingof seven (R)-selective amine transaminases (AspTerfrom Aspergillus terreus, AspFum from Aspergillus fu-migatus, AspOry from Aspergillus oryzae, PenChryfrom Penicillium chrysogenum, NeoFis from Neosar-torya fischeri, GibZea from Gibberella zeae and

MycVan from Mycobacterium vanbaalenii) for theasymmetric synthesis of a range of aliphatic, aromaticand arylaliphatic amines to elucidate the substratescope of these enzymes originating from various mi-croorganisms and for which no natural substrate isknown (Scheme 1). These enzymes were primarilychosen from the 17 recently discovered R-ATAs[8] be-cause of their high specific activities towards 2b, 7band 9b and sufficient expression levels in E. coli.

Enzyme Production and Formulation

Before the newly discovered R-ATAs could be bio-chemically characterized in more detail and used inasymmetric synthesis, methods for their expression inE. coli on sufficient scale and stability studies neededto be performed, especially as we have noticed beforethat several enzymes were rather unstable after His-tag purification in tricine buffer used for the conduc-tivity assay.[9] For application in biocatalysis purifiedenzymes are not needed as long as it is ensured thatthe crude cell lysate does not contain any interferingbackground activity, which was never observed for allR-ATAs expressed in E. coli. Thus, the crude extractscontaining the seven R-ATAs were only subjected toammonium sulfate precipitation at 60% saturationand could be recovered fully active. Some parts of thetotal protein could be discarded in the cases ofAspFum and GibZea by fractionated precipitation atan ammonium sulfate saturation of 40% beforehand.This step was not carried out for the other enzymes asit would have caused a reasonable loss of activity.Next, various additives – commonly used osmolytessuch as glycerol, ethylene glycol, PEG 1550, trehalose,sucrose and sarcosine as well as the surfactant Tween80 – were investigated for their stabilizing effect onthe proteins during freeze-drying and storage as lyo-philizate and in solution. The best conditions for eachR-ATA along with the activity and yield of lyophili-zate from 400-mL shake flask cultivations are sum-marized in Table 1. In most cases sucrose or glycerolhad the best stabilizing effect on the R-ATAs. All en-zymes were stable for at least three months in lyophi-

Scheme 1. Chiral (R)-amines obtained from the corresponding prochiral ketones using the amine transaminases.

2440 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2439 – 2445

FULL PAPERS Sebastian Sch�tzle et al.

lized form and for at least two weeks in solution at4 8C.

Finding the Right Reaction System

With stable and active biocatalysts available, the nextimportant challenge for a successful asymmetric syn-thesis of chiral amines was the identification of thebest method to shift the equilibrium towards productformation. Thus, the use of pyruvate decarboxylase[6b]

and the combination of lactate dehydrogenase (LDH)with glucose dehydrogenase (GDH) for cofactor recy-cling[6a] were preliminary investigated and the LDH/GDH system (Scheme 2) gave most satisfying results.

Next, several buffers were investigated to identifythe best system for the biocatalysis. Recent studiesshowed that the most often used phosphate buffer isnot always the best choice for transaminases becauseof inhibitory effects of the phosphate ions – assuming-ly by blocking the binding site of the phosphate group

of the PLP-cofactor. On the other hand, we showedin a previous study that TRIS, EPPS and especiallyHEPES had an inhibitory effect on the S-ATA fromRhodobacter sphaeroides compared to phosphatebuffer (7b was converted with 71%, 76% and 24%,respectively[9]). Therefore, the initial activity towards7b in four different buffer systems (TRIS, HEPES,MOPS and BES) was compared to the activity in50 mM and 100 mM phosphate buffer (Table 2).

Overall, the influence of the different buffer sys-tems was only moderate (Table 2). Significantly loweractivity was only observed for PenChr, in 100 mMphosphate or 50 mM TRIS buffer, whereas higher ac-tivity was found in tricine or HEPES buffer. Alto-gether the differences were rather small and hencephosphate buffer was chosen for subsequent experi-ments as this buffer is suitable for the LDH/GDH re-cycling system.

The pH profiles of the R-ATAs were determinedusing the photometric acetophenone assay.[10] The ma-jority of the enzymes show a pH optimum between 8–9, which is also typical for S-ATAs such as theenzyme from Vibrio fluvialis ;[11] AspTer and PenChrshow a sharp optimum at pH 8.5–9 while AspOry,AspFum and NeoFis have a broader optimal range

Table 1. Summary of parameters for the preparation of thelyophilized biocatalysts. Activities were measured with the ace-tophenone assay photometrically[10] and protein concentrationswere determined with the BC assay according to the supplier’smanual.

Enzyme Additive[a] Activity[b]

[U/mg]Yield[c]

[mg]Total ac-tivity[c]

[U]

Activity[U/mgprotein]

AspTer sucrose 0.74 956 711 3.94PenChr sucrose 0.15 789 122 1.01AspOry – 1.10 757 829 3.65AspFum sucrose 0.62 687 426 4.83NeoFis sarcosine 1.22 622 759 8.14GibZea sucrose 2.47 574 1415 18.63MycVan glycerol 0.44 580 258 1.95

[a] 2% (w/v) added after ammonium sulfate precipitation.[b] Activity of lyophilisate.[c] For lyophilisate obtained from 400 mL cultivation.

Scheme 2. Reaction scheme for the asymmetric synthesis ofchiral (R)-amines with the R-ATAs in combination with theLDH/GDH system to shift the equilibrium towards productformation.

Table 2. Relative activity of the R-ATAs in different buffer systems. The activity towards 7b was measured spectrophotomet-rically[10] in the different buffers at pH 8 and was normalized to the activity in 50 mM phosphate buffer (set to 100%). Allvalues are means from three individual measurements and standard deviations did not exceed 10%.

Relative activity [%]Phosphate(50 mM)

Phosphate(100 mM)

TRIS(100 mM)

HEPES(100 mM)

MOPS(100 mM)

BES(100 mM)

Tricine(100 mM)

AspTer 100 99 93 100 98 80 104PenChr 100 76 52 123 114 99 136AspOry 100 107 106 103 106 104 105AspFum 100 103 98 104 103 101 102NeoFis 100 101 103 102 101 97 101GibZea 100 112 107 122 123 122 119MycVan 100 95 103 104 104 109 108

Adv. Synth. Catal. 2011, 353, 2439 – 2445 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2441

Enzymatic Asymmetric Synthesis of Enantiomerically Pure Aliphatic, Aromatic and Arylaliphatic Amines

between 8–9. GibZea has a rather untypical pH opti-mum of 7.5 and loses already 30 to 70% activity atpH 8.5 and 9, respectively, whereas MycVan has avery broad optimum between pH 7.5–8.5 and stillshows ~80% activity at pH 9 (Figure 1). Nevertheless,in combination with the LDH/GDH system, prelimi-nary experiments gave the highest yields at pH 7.5.

Asymmetric Synthesis

Having found suitable ways for stabilizing the en-zymes and having elucidated the optimal reactionconditions, the R-ATAs were applied to asymmetricsynthesis to obtain the aliphatic, aromatic and arylali-phatic amines (Scheme 1) and the results are summar-ized in Table 3.

These data clearly show that all seven enzymes ex-hibited excellent selectivity (>99% ee) and in allcases the (R)-enantiopreference could be confirmed.Most importantly, for all substrates at least one R-ATA – and for various compounds several enzymes –could be identified, which allowed complete conver-sion of the ketone into the chiral amine. This exceed-ed our expectations as the ketones represent a ratherdiverse set of compounds. In only a few cases, lowerconversions were observed (i.e., 2b, 7b and 9b withPenChr or GibZea) but further optimization to ach-ieve higher conversion was considered unnecessary inlight of the other very useful R-ATAs.

Although PenChr and AspTer show a rather highsimilarity score in a multiple sequence alignment(Figure 2), the conversions achieved in biocatalysisdiffered considerably between these two R-ATA.AspTer enabled 100% conversion for 2-aminohexane2b, -heptane 3b and -nonane 4 b, but not for the short-er 2-aminopentane 1b (~40%) and 2-amino-4-methyl-pentane 5b (~60%). Substrates with sterically moredemanding groups than alkyl (cyclohexane or phenyl)next to the ketone function were converted evenworse. These differences will be analyzed in moredetail once a 3-D structure or homology model ofthese R-ATAs is available.

Another interesting result is the fact that AspTerdid not give high yields for methylphenylpropiona-mine 9b and the 4-methoxy derivative 10b (32% and20%), but good yields of around 80% could be ach-ieved with the hydroxy derivatives 11b and 12b. Thisphenomenon was only observed with AspTer. Thetwo other enzymes from Aspergillus species (AspOryand AspFum) turned out to be even better catalysts,as well as NeoFis. According to their sequence identi-ty these enzymes are much alike, especially AspFumand NeoFis with a sequence identity of 96%. AspOryshares a similarity of 72% and 73%, respectively, andgave very good to excellent conversions for the ali-phatic substrates, considerably less conversion for 6a,

Figure 1. pH profiles of the seven R-ATAs. Activities weremeasured with the acetophenone assay spectrophotometri-cally[10] in Davies buffer[12] , which is commonly used for pHprofile determinations.

2442 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2439 – 2445

FULL PAPERS Sebastian Sch�tzle et al.

7a and 8a and again very good conversions for the ar-ylaliphatic substrates 9–12a. MycVan is the outsiderof the investigated proteins, with similarity scores notexceeding 40%. The phylogenetic tree (Figure 2)shows that MycVan has the highest degree of relation-ship to the assumed ancestor BCAT. MycVan ach-ieved good yields for the aliphatic and arylaliphatic

compounds, moderate yields for 6a and 7a and didnot like 8a at all.

Bringing together the observed performancesduring biocatalysis and the degree of relationship asshown in the phylogenetic tree (Figure 2), several cor-relations attract attention. MycVan, for example, is agood biocatalysis not far from the performance ofAspOry, AspFum and NeoFis, which show a highdegree of relationship. Although GibZea shows ahigher similarity towards these good biocatalysts suchas MycVan, it showed only a rather poor perfor-mance. Interestingly, AspTer and PenChr are alsostrongly related, but showed very different perform-ances regarding the conversions achieved. However,they both prefer aliphatic substrates with a chainlength of at least six carbon atoms over shorter sub-strates, yielded only very small amounts of 7b, andboth achieved higher conversions with at least one hy-droxy derivative of 10b as with 10b itself

Conclusions

We have successfully applied the recently discoveredR-ATAs to asymmetric synthesis of various amineswith >99% conversion and excellent enantiomericpurities of >99% ee. Looking back to our first pre-liminary experiments with conversions of <40%,[8]

one may conclude that producing biocatalysts in theright formulation, optimization of the reaction systemas well as using a broad set of substrates is as impor-tant as finding new enzymes.

Table 3. Conversion and enantiomeric excess of the amines obtained by asymmetric synthesis from the corresponding ke-tones. In general, reactions were repeated in triplicates and standard deviation did not exceed 10%.

R-ATA AspTer PenChr AspOry AspFum NeoFis GibZea MycVanSubstrate c[a]

[%]eep

[b]ACHTUNGTRENNUNG[% ee]c[a]

[%]eep

[b]ACHTUNGTRENNUNG[%ee]c[a]

[%]eep

[b]ACHTUNGTRENNUNG[%ee]c[a]

[%]eep

[b]ACHTUNGTRENNUNG[%ee]c[a]

[%]eep

[b]ACHTUNGTRENNUNG[%ee]c[a]

[%]eep

[b]ACHTUNGTRENNUNG[%ee]c[a]

[%]eep

[b]ACHTUNGTRENNUNG[%ee]

1b 47 >99 1 – 92 >99 >99 >99 >99 >99 30 >99 66 >992b >99 >99 1 – >99 >99 >99 >99 >99 >99 14 >99 88 >993b >99 >99 15 >99 >99 >99 >99 >99 >99 >99 43 >99 69 >994b >99 >99 12 >99 77 >99 >99 >99 >99 >99 23 >99 65 >995b 64 >99 9 >99 79 >99 >99 >99 >99 >99 32 >99 68 >996b 30 >99 1 – 53 >99 93 >99 59 >99 16 >99 56 >997b 3 >99 0 – 39 >99 >99 >99 87 >99 4 >99 42 >998b 17 >99 2 – 8 99 68 >99 74 >99 3 99 6 999b 32 >99 2 – 92 >99 >99 >99 >99 >99 9 >99 89 >9910b 20 >99 1 – >99 >99 >99 >99 92 >99 7 >99 72 >9911b 83 >99 4 – >99 >99 >99 >99 >99 >99 5 >99 71 >9912b 77 >99 2 – 98 >99 94 >99 >99 >99 6 >99 85 >99

[a] Conversion as determined from the product analysis by HPLC.[b] % enantiomeric excess of product.

Figure 2. Phylogenetic tree of the investigated R-ATAs. Thefigure shows all 21 hits from the database search and thebiocatalysts used in this study are named. DATA, ADCL (4-amino-4-deoxychorismate lyase) and BCAT are the otherthree members of the PLP-fold type IV.

Adv. Synth. Catal. 2011, 353, 2439 – 2445 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2443

Enzymatic Asymmetric Synthesis of Enantiomerically Pure Aliphatic, Aromatic and Arylaliphatic Amines

Experimental Section

All chemicals were of analytical grade purity and obtainedfrom Sigma Aldrich (Munich, Germany), ABCR (Karlsruhe,Germany) or Acros (Geel, Belgium).

Functional Expression and Preparation ofLyophilized Transaminases

Expression of the transaminases was carried out as de-scribed elsewhere.[8] The crude extract was stirred on ice andsupplemented with 60% saturation of (NH4)2SO4 for precip-itation of R-ATAs and stirred for 30 min. In the cases ofAspFum and GibZea a fraction of the proteins from thecrude extract was precipitated by adding 40% saturation of(NH4)2SO4 from the supernatant, stirred and centrifuged be-forehand. The transaminase containing pellet was dissolvedin 10 mL sodium phosphate buffer (50 mM, pH 7.5) contain-ing 0.1 mM PLP and 2% additive (w/v) (Table 1) prior tolyophilization. Protein concentrations were determinedusing the BC assay kit (Uptima, Montlucon, France).

Transaminase Activity Assay

Activity measurements were performed in triplicate as de-scribed.[10] The pH profiles were determined using 100 mMDavies buffer and 10 mL (from a 10 mgmL�1 solution of lyo-philisate in deionized water) protein solution were added.The acetophenone assay was also used to evaluate thebuffer acceptance. For this, seven different buffers (sodiumphosphate 50 mM pH 8, phosphate 100 mM pH 8, Tris100 mM pH 8, BES 100 mM pH 8, HEPES 100 mM pH 8,MOPS 100 mM pH 8 and tricine 100 mM pH 8) were chosenand enzyme solution was added before measuring activity.

Asymmetric Synthesis

All reactions were carried out in 0.5 mL scale in 1.5 mL Ep-pendorf tubes containing 30 mgmL�1 transaminase lyophili-sate, 90 U lactate dehydrogenase, 15 U glucose dehydrogen-ase, 50 mM 1a–13a, 250 mM d-alanine, 150 mM d-glucose,1 mM NADH and 0.1 mM PLP in sodium phosphate buffer(100 mM, pH 7.5). While incubating at 30 8C and 1000 rpmin an Eppendorf shaker, samples (150 mL) were taken afterdistinct time periods and analyzed by HPLC for conversionand gas chromatography or capillary electrophoresis for de-termination of optical purity (% eeP). Promising reactionswere repeated in triplicate with 30 mgmL�1 TA lyophilisateand 150 mL were analyzed by HPLC whereas another150 mL were used for chiral analysis.

Analytical Methods

For quantitative analysis 150 mL samples of the reactionmixture were supplemented with 15 mL TFA, centrifuged(13000 g, 5 min) and 15 mL of the supernatant were directlyanalyzed by HPLC (Hitachi LaChrom) using a reversed-phase C18 column (LiChrospher�). Detection took placewith a UV detector set at 245 nm and an evaporating lightscattering detector (see Supporting Information).

The chiral analysis of 1–10b was performed by gas chro-matography. A 150 mL sample was supplemented with 15 mL10 M NaOH and extracted with 400 mL ethyl acetate. The or-

ganic phase was dried over Na2SO4 and the extractedamines were derivatized to the trifluoroacetamides byadding a 20-fold excess of trifluoroacetic anhydride. After5 min at room temperature and purging with nitrogen toremove excess anhydride, residual trifluoracetic acid andsolvent, the derivatized compounds were dissolved in 50 mLethyl acetate and 0.6 mL of this solution were injected into aHewlett Packard 5890 Series II gas chromatograph equippedwith an octakis-(2,3-di-O-acetyl-6-O-tert-butyldimethylsilyl)-g-cyclodextrin column (50 m� 0.25 mm) for analyzing 8b orwith a heptakis-(2,3-di-O-acetyl-6-O-tert-butyldimethyl-silyl)-b-cyclodextrin column (25 m � 0.25 mm) for 1–7b and9–10b.

Chiral analysis of 11–12b (Table 4) was performed by ca-pillary electrophoresis on a PACE-MDQ system equippedwith a fused silica capillary (50 mm inner diameter) as re-ported previously.[6b]

Acknowledgements

The authors thank the Enzymaticals AG (Greifswald, Ger-many) for the larger scale production of the amine transami-nases.

References

[1] M. Rçsler, R. Anand, A. Cicin-Sain, S. Gauthier, Y.Agid, P. Dal-Bianco, H. B. St�helin, R. Hartman, M.Gharabawi, T. Bayer, Br. Med. J. 1999, 318, 633–640.

[2] a) G. Bringmann, R. Weirich, H. Reuscher, J. R.Jansen, L. Kinzinger, T. Ortmann, Liebigs Ann. Chem.1993, 1993, 877–888; b) T. R. Hoye, M. Chen, Tetrahe-dron Lett. 1996, 37, 3099–3100; c) Y. F. Hallock, K. P.Manfredi, J. W. Blunt, J. H. Cardellina, M. Schaeffer,K.-P. Gulden, G. Bringmann, A. Y. Lee, J. Clardy, J.Org. Chem. 1994, 59, 6349–6355.

Table 4. Details of chiral analysis.

Amine[a] Temperature[8C]

Retention time(S) [min]

Retention time(R) [min]

1b 125 3.2 3.52b 130 3.3 3.83b 140 2.7 3.04b 130 6.3 8.55b 130 2.4 2.96b 150 2.3 2.47b 170 7.6 8.38b 220 7.0 6.39b 185 2.3 2.510b 185 5.6 6.111b CE 3.7 3.912b CE 3.5 3.7

[a] For 1b–10b the retention times refer to the trifluoraceticacid derivative.

2444 asc.wiley-vch.de � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Synth. Catal. 2011, 353, 2439 – 2445

FULL PAPERS Sebastian Sch�tzle et al.

[3] P. Abrams, M. Speakman, M. Stott, D. Arkell, R.Pocock, Br. J. Urol. 1997, 80, 587–596.

[4] X. Chen, J. Chen, M. De Paolis, J. Zhu, J. Org. Chem.2005, 70, 4397–4408.

[5] M. Hçhne, U. Bornscheuer, ChemCatChem 2009, 1, 42–51.

[6] a) D. Koszelewski, I. Lavandera, D. Clay, D. Rozzell,W. Kroutil, Adv. Synth. Catal. 2008, 350, 2761–2766;b) M. Hçhne, S. K�hl, K. Robins, U. T. Bornscheuer,ChemBioChem 2008, 9, 363–365; c) G. B. Matcham, L.Mohit, L. Wei, N. Craig, R. Nelson, A. Wang, W. Wu,Chimia 1999, 53, 584–589; d) M. D. Truppo, J. D. Roz-zell, N. J. Turner, Org. Process Res. Dev. 2010, 14, 234–237; e) D. Koszelewski, I. Lavandera, D. Clay, G. M.Guebitz, D. Rozzell, W. Kroutil, Angew. Chem. 2008,120, 9477–9480; Angew. Chem. Int. Ed. 2008, 47, 9337–

9340; f) K. E. Cassimjee, C. Branneby, V. Abedi, A.Wells, P. Berglund, Chem. Commun. 2010, 46, 5569–5571.

[7] C. K. Savile, J. M. Janey, E. C. Mundorff, J. C. Moore,S. Tam, W. R. Jarvis, J. C. Colbeck, A. Krebber, F. J.Fleitz, J. Brands, P. N. Devine, G. W. Huisman, G. J.Hughes, Science 2010, 329, 305–309.

[8] M. Hçhne, S. Sch�tzle, H. Jochens, K. Robins, U. T.Bornscheuer, Nat. Chem. Biol. 2010, 6, 807–813.

[9] S. Sch�tzle, M. Hçhne, K. Robins, U. T. Bornscheuer,Anal. Chem. 2010, 82, 2082–2086.

[10] S. Sch�tzle, M. Hçhne, E. Redestad, K. Robins, U. T.Bornscheuer, Anal. Chem. 2009, 81, 8244–8248.

[11] J. S. Shin, H. Yun, J. W. Jang, I. Park, B. G. Kim, Appl.Microbiol. Biotechnol. 2003, 61, 463–471.

[12] M. T. Davies, Analyst 1959, 84, 248–251.

Adv. Synth. Catal. 2011, 353, 2439 – 2445 � 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim asc.wiley-vch.de 2445

Enzymatic Asymmetric Synthesis of Enantiomerically Pure Aliphatic, Aromatic and Arylaliphatic Amines

The Affirmation

- 83 -

Hiermit erkläre ich, dass diese Arbeit bisher von mir weder an der Mathematisch-

Naturwissenschaftlichen Fakultät der Ernst-Moritz-Arndt-Universität Greifswald noch einer anderen

wissenschaftlichen Einrichtung zum Zwecke der Promotion eingereicht wurde.

Ferner erkläre ich, dass ich diese Arbeit selbständig verfasst und keine anderen als die darin

angegebenen Hilfsmittel und Hilfen benutzt und keine Textabschnitte eines Dritten ohne Kennzeichnung

übernommen habe.

My Vita

- 84 -

CURRICULUM VITA

PHD IN BIOCHEMISTRY

since 10/2008 PhD thesis in the group of Prof. Bornscheuer (Dept. of Biotechnology & Enzyme

Catalysis) at the Institute of Biochemistry, University of Greifswald, Germany

Identification, Characterization and Application of novel (R)-selective Amine

Transaminases

10/2010-12/2010 Research Internship in the group of Prof. Kittikun, Prince of Songkla University,

Hat Yai, Thailand

DIPLOMA IN BIOCHEMISTRY

10/2007-07/2008 Diploma thesis in the group of Prof. Bornscheuer at the Institute of

Biochemistry, University of Greifswald, Germany

Investigation of Nitrile Hydratases

04/2006-06/2007 Student assistant for Dr. Rainer Wardenga at the Dept. of Biotechnology &

Enzyme Catalysis” – Enzymatic Synthesis of amino acid surfactants

10/2006-11/2006 Industrial placement with Dr. Frank Hollmann, Goldschmidt GmbH (Evonik),

Essen, Germany – Organic Synthesis with esterases & lipases

10/2005-03/2006 Student assistant in the group of Inorganic Chemistry as supervisor of the

practical course Qualitative Inorganic Analytics

CIVILLIAN SERVICE

06/2002-03/2003 Caritasverband, City of Oberhausen, Franziskus-Haus

Caregiver for People with Disabilities

04/2003-08/2003 Temporary employee as Caregiver for People with Disabilities (see above)

SCHOOL

09/1993-05/2002 Heinrich-Heine-Gymnasium, Bottrop, Germany

The Acknowledgements

- 85 -

ACKNOWLEDGEMENTS

Mein erstes und größtes Dankeschön gilt Uwe. Vielen Dank für Dein Vertrauen in unsere Ideen und

unsere Arbeit sowie den damit verbundenen Freiheiten. Ich habe in den letzten Jahren sehr, sehr viel

von Dir gelernt und hoffe, dass mir möglichst viel davon noch lange in Erinnerung bleibt. Besonderen

Dank auch für meine tolle Zeit im Ausland und vor allem für die mit großem Fingerspitzengefühl

gesetzten dead-lines und all Deine Hilfe.

Das zweite und ähnlich große Dankeschön geht an Matthias. Vielen, lieben Dank für die hervorragende

Zusammenarbeit, Deine Ideen und Anregungen. Ich glaube das mit uns hat so prima funktioniert, weil

sich unsere doch recht unterschiedlichen Charaktere so vorzüglich ergänzt haben.

Das letzte Mitglied der “einflussreichen Drei” ist Onkel Rainer. Du hast mich ja als Hiwi nicht nur gequält

und schikaniert, sondern mir vor allem unglaublich viel beigebracht – sowohl praktisch als auch

theoretisch. Somit hast Du mich wahrscheinlich “wissenschaftlich“ gesehen am meisten geprägt. Noch

viel mehr möchte ich mich aber für die Zeit danach bedanken, nämlich dafür, dass Du mir wirklich immer

geholfen hast. Dankeschön.

Weiterhin bedanke ich mich bei allen Transaminase-Jungs, also bei Fabian, Hannes, Ahmad und Erik.

Vielen Dank für die tolle Zusammenarbeit und die schöne Zeit im Labor. Und natürlich für Eure Mühen

und vor allem die guten Ergebnisse. Erik, thank you so much for the great teamwork and – much more

important – for the occasional beer after work.

Bei der Lonza AG und insbesondere bei Karen bedanke ich mich für die sowohl finanzielle als auch

inhaltliche Unterstützung dieses Projekts. Vielen Dank für Dein großes Interesse an unserer Arbeit und

vor allem für Dein Vertrauen und all Deine Mühen. Und natürlich für die ganzen Gene.

Selbstverständlich möchte ich mich auch bei all den anderen netten Menschen aus dem Arbeitskreis

bedanken. Ganz besonders bei Howie, Marlen, Helge und dem Pückerer. Jedes Käffchen mit Euch war

mir eine Freude. Stefan, Maria und Hendrik danke ich für die schöne Atmosphäre im Labor. Ina, Angelika,

Dominique, Anke und Anita danke ich dafür, dass Ihr Euch immer um alles kümmert und deshalb alles so

wunderbar funktioniert.

Und zu guter Letzt: meine liebe Familie. Euch danke ich von ganzem Herzen für Euer Vertrauen und Eure

Geduld. Für die Geduld natürlich ganz besonders Dir, lieber Opa. Tut mir Leid, dass ich Dich so lange hab‘

warten lassen. Vielen Dank, dass Ihr immer für mich da seid und für Euer Verständnis, wenn ich mal nicht

da sein konnte.