Directed evolution of an enantioselective enoate-reductase: Testing the utility of iterative saturation mutagenesis

Article abstract of DOI:10.1002/adsc.200900644

Directed evolution utilizing iterative saturation mutagenesis (ISM) has been applied to the old yellow enzyme homologue YqjM in the quest to broaden its substrate scope, while controlling the enantioselectivity in the bioreduction of a set of substituted cyclopentenone and cyclohexenone derivatives. Guided by the known crystal structure of YqjM, 20 residues were selected as sites for saturation mutagenesis, a pooling strategy based on the method of Phizicky [M. R. Martzen, S. M. McCraith, S. L. Spinelli, F. M. Torres, S. Fields, E. J. Grayhack, E. M. Phizicky, Science 1999, 286, 1153-1155] being used in the GC screening process. The genes of some of the hits were subsequently employed as templates for randomization experiments at the other putative hot spots. Both (R)-and (S)-selective variants were evolved using 3-methylcyclohexenone as the model substrate in the asymmetric bioreduction of the olefinic functionality, only small mutant libraries and thus minimal screening effort being necessary. Some of the best mutants also proved to be excellent catalysts when testing other prochiral substrates without resorting to additional mutagenesis/screening experiments. Thus, the results constitute an important step forward in generalizing the utility of ISM as an efficient method in laboratory evolution of enzymes as catalysts in organic chemistry.

Full text of DOI:10.1002/adsc.200900644

FULL PAPERS
DOI: 10.1002/adsc.200900644
Directed Evolution of an Enantioselective Enoate-Reductase:
Testing the Utility of Iterative Saturation Mutagenesis
Despina J. Bougioukou,a Sabrina Kille,a Andreas Taglieber,a
and Manfred T. Reetza,*
a
Max-Planck-Institut fꢀr Kohlenforschung, Kaiser-Wilhelm-Platz 1, 45470 Mꢀlheim/Ruhr, Germany
Fax : (+49)-208-306-2985; e-mail: reetz@mpi-muelheim.mpg.de
Received: September 18, 2009; Published online: December 10, 2009
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/adsc.200900644.
Abstract: Directed evolution utilizing iterative satu- hot spots. Both (R)- and (S)-selective variants were
ration mutagenesis (ISM) has been applied to the old evolved using 3-methylcyclohexenone as the model
yellow enzyme homologue YqjM in the quest to substrate in the asymmetric bioreduction of the ole-
broaden its substrate scope, while controlling the finic functionality, only small mutant libraries and
enantioselectivity in the bioreduction of a set of sub- thus minimal screening effort being necessary. Some
stituted cyclopentenone and cyclohexenone deriva- of the best mutants also proved to be excellent cata-
tives. Guided by the known crystal structure of lysts when testing other prochiral substrates without
YqjM, 20 residues were selected as sites for satura- resorting to additional mutagenesis/screening experi-
tion mutagenesis, a pooling strategy based on the ments. Thus, the results constitute an important step
method of Phizicky [M. R. Martzen, S. M. McCraith, forward in generalizing the utility of ISM as an effi-
S. L. Spinelli, F. M. Torres, S. Fields, E. J. Grayhack, cient method in laboratory evolution of enzymes as
E. M. Phizicky, Science 1999, 286, 1153–1155] being catalysts in organic chemistry.
used in the GC screening process. The genes of some
of the hits were subsequently employed as templates Keywords: asymmetric reduction; directed evolution;
for randomization experiments at the other putative enantioselectivity; enoate-reductase; screening assay
Introduction
Our contribution to efficacy in directed evolution is
iterative saturation mutagenesis (ISM), which can be
The concept of directed evolution of enantioselective used to control enantioselectivity and/or substrate
enzymes[1] provides a way to eliminate many of the scope (rate)[11,12] as well as thermostability.[13] Accord-
traditional limitations of biocatalysis[2] in synthetic or- ingly, appropriate sites in an enzyme, labeled A, B, C,
ganic chemistry and complements the traditional ap- etc., are randomized by saturation mutagenesis with
proaches to asymmetric catalysis[3] based on chiral formation of a set of initial libraries which are
transition metal complexes[4] or organocatalysts.[5] Fol- screened for enhanced enantioselectivity or thermo-
lowing proof-of-concept more than a decade ago,[6] stability. A given site can be composed of one or
many different enzymes have been engineered for en- more amino acid positions in the enzyme. The genes
hanced enantioselectivity and broadened substrate of the hits are then utilized as templates for perform-
scope (rate) using a variety of different gene muta- ing saturation mutagenesis at the respective other
genesis methods, including epPCR, saturation muta- sites, the process being repeated until the desired
genesis and DNA shuffling.[1] In most of these early degree of catalyst improvement has been achieved.
studies and in other directed evolution reports regard- Choosing the appropriate sites is crucial, but it is not
ing other parameters such as protein thermostability, necessary to explore all theoretically possible upward
the question of efficiency was not a central issue.[7] pathways. In the case of enantioselectivity and sub-
However, this is a pressing issue, as to this day the strate acceptance, residues aligning or near the bind-
screening effort constitutes the bottleneck of directed ing pocket are chosen as randomization sites in what
evolution.[7–10]
has been termed combinatorial active-site saturation
test (CAST).[14] Iterative CASTing was first shown to
Adv. Synth. Catal. 2009, 351, 3287 – 3305
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
3287

Despina J. Bougioukou et al.
FULL PAPERS
be unusually effective in the enhancement of enantio- crease activity, while controlling (R)- and (S)-selectiv-
selectivity of an epoxide hydrolase,[11] and was subse- ity on an optional basis. Subsequently, we wanted to
quently applied successfully to other enzymes.[15] test other prochiral substrates using the (R)- and (S)-
However, claiming wide applicability requires further selective mutants evolved for substrate 1a without
research, especially since different ways to apply the performing additional mutagenesis/screening experi-
iterative process exist.
ments.
Relevant to our investigation is the recent report
Here we report that directed evolution based on
iterative CASTing is a viable way to control the enan- by Stewart and co-workers, who performed site-spe-
tioselectivity of the 37.4 kDa enoate-reductase YqjM cific mutagenesis at position Trp116 of the OYE from
from Bacillus subtilis,[16] a homologue of the old Saccharomyces pastorianus in order to test the 19 re-
yellow enzyme (OYE; EC 1.6.99.1),[17] serving as a spective mutants, the surprising result of which was
catalyst in the asymmetric reduction of cyclic enones. reversal of stereoselectivity of some mutants.[22] Our
We demonstrate that (R)- and (S)-selectivity is possi- present contribution is the first report of directed evo-
ble on an optional basis, and that only small mutant lution of an OYE.
libraries are needed to achieve this challenging goal.
In order to reduce the screening effort even more, we
have developed a pooling strategy. Apart from the Results and Discussion
importance of methodology development in directed
evolution,[8,10,12] our results are relevant to synthetic Choice of Sites for Saturation Mutagenesis
organic chemistry. The enantioselective reduction of
prochiral a,b-unsaturated ketones using such different In order to choose appropriate randomization sites,
catalyst systems as enoate-reductases,[2,18] chiral transi- we were guided by the published X-ray structure of
tion metal complexes[19] or organocatalysts[20] is of YqjM containing the inhibitor pHBA.[21] The catalytic
considerable current interest.
active site of the enzyme is quite open (Figure 1, a). It
The 339 amino acid enzyme YqjM is a flavin-con- is covered mainly by water molecules and partially by
taining, NADPH-dependent enoate-reductase which the amino acids comprising the C-terminus of chain
can be expressed efficiently in E. coli (161 mg/ B. Twenty residues were selected for saturation muta-
liter).[16] The 1.3 ꢂ crystal structure of the homote- genesis, based on their distance to the inhibitor
tramer (dimer of active dimers) containing the inhibi- pHBA (Figure 1, b). The triad His164, His167 and
tor p-hydroxybenzaldehyde (pHBA) was reported in Tyr169 was excluded as these residues participate di-
2005, revealing a typical (ab)8-barrel with active site rectly in the catalytic process.[21,23] The selected CAST
residues His164, His167 and Tyr169 as well as the residues can be separated roughly into three groups:
FMN cofactor.[21] Mechanistically, the carbonyl func- 1) a “hot zone” composed of amino acids Cys26,
tion of such substrates as cyclohexenone or deriva- Tyr28, Ile69, Asp73 and Thr70 and to a smaller extent
tives thereof is activated by two hydrogen bonds orig- Met25, Met27, and Glu111, the respective side-chains
inating from the two histidines. The reduced flavin de- most likely being closest to the b-C-atom of the sub-
livers the hydride from one p-face of the olefinic strate; 2) a presumably “soft zone” comprising Ala60,
double bond and Tyr169 provides the proton on the Ala104 and Ala106 and to a lesser extent Lys109, the
opposite side in an overall trans-addition. The oxi- three Ala residues presumably pointing to the a-C-
dized flavin is then reduced by NADPH, allowing the atom of the substrate; 3) a presumably “indifferent
catalytic cycle to begin once more.[16,21]
zone” composed of Leu103, His105, Ala166, Ser249,
In the present directed evolution study we chose Gly251, Ala258 and Arg’336, with the latter belonging
the asymmetric reduction of 3-methylcyclohexenone to chain B. It is worth mentioning that the “west” of
(1a) with formation of 3-methylcyclohexanone (2a) as the flavin is rather open with the amino acid residues
the model reaction (Scheme 1). The goal was to test belonging to a loop (black arrow in Figure 1, a), and
the virtue of iterative CASTing in the quest to in- is more than 12 ꢂ away from the catalytic center.
Bringing this loop closer to the binding pocket would
require a mutationally induced allosteric effect, which
was not considered in the present investigation.
In any study based on ISM, the number of chosen
amino acid positions (n) can be randomized by satu-
ration mutagenesis individually with formation of n li-
braries of mutants. In this case the degree of oversam-
pling,[10] was calculated on the basis of known algo-
rithms assuming the absence of amino acid bias.[25]
Scheme 1. Model reaction used in screening libraries of
Ensuring 95% coverage of relevant sequences when
using NNK degeneracy, only about 100 transformants
YqjM mutants. NADPH was regenerated by glucose dehy-
drogenase (GDH)/glucose.
3288
asc.wiley-vch.de
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2009, 351, 3287 – 3305

Directed Evolution of an Enantioselective Enoate-Reductase
FULL PAPERS
Figure 1. The 20 amino acid positions near the binding pocket of the YqjM flavoprotein considered for saturation mutagene-
sis (potential sites for CASTing). a) Space-filling model of YqjM (Chain A, light gray; Chain B, light blue; red spheres,
water molecules; yellow spheres, FMN; blue spheres, p-hydrobenzaldehyde as inhibitor). Certain water molecules at the en-
trance of catalytic cavity were removed for clarity. Black arrow indicates the distant loop. b) Close-up view of the catalytic
active site (amino acids in green responded positively, gray ones negatively to saturation test.)[24]
have to be evaluated which can be handled by auto- amino acids,[10] the conventional procedure could not
mated GC or HPLC. However, if a large number of be applied because one set of primers cannot be used
such small libraries need to be considered, then the when targeting distal sites. Although we have recently
total screening effort may prove to be formidable. Al- developed an improved PCR method for creating sat-
ternatively, the operator can consider sites comprised uration mutagenesis libraries for difficult-to-amplify
of more than a single residue by a grouping process. templates,[28] we opted to add an alternative strategy
This entails a smaller number of libraries and the ad- based on a recently developed method for efficient in
vantage of opening the door for potential cooperative vitro recombination cloning of multiple gene inserts
effects within a site (more than additivity), but the directly in E. coli, the so-called sequence and ligation
disadvantage of increased oversampling in the screen- independent cloning (SLIC).[29] Briefly, a set of pri-
ing process. In the present study we opted for the first mers encompassing the desired degenerated codons is
choice, which nevertheless posed a significant screen- used in order to amplify a part of the YqjM gene
ing problem due to the high number of saturation using the pET21a-YqjM as a template, Figure S3
sites.
(Supporting Information). The rest of the YqjM gene
and the vector are amplified with a second set of pri-
mers which shares ca. 30 nucleotides identity with the
first pair. After the elimination of the template, the
two PCR products are treated separately with T4
Generation of Saturation Mutagenesis Libraries
The 20 single site saturation mutagenesis libraries are DNA polymerase. After a short time treatment, the
schematically shown in Figure S2 (Supporting Infor- vector and the pool of degenerated inserts are mixed
mation), indicating the positions in the 1017 bp YqjM and incubated with E. coli RecA which facilitates the
gene at which the genetic information is to be altered, assembly of the homologous single strands in vitro, to
representing all of the 20 common proteinogenic be followed by transformation into E. coli DH5a
amino acids. All initial randomization experiments as strain.
well as the iterative processes were performed using
Using this approach, we obtained up to 700 colo-
nies per transformation. A protocol for covering the
the QuikChange protocol of Stratagene.[26]
In an additional experiment involving simultaneous complete sequence space at positions 26/60 (20ꢃ20=
randomization of two amino positions, Cys26 and 400 amino acid combinations) was developed as sche-
Ala60 using NDT[27] codon degeneracy encoding 12 matized in Figure S4 (Supporting Information).
Adv. Synth. Catal. 2009, 351, 3287 – 3305
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
3289

Despina J. Bougioukou et al.
FULL PAPERS
Table 1. Degeneracy of YqjM saturation mutagenesis librar-
Quality Control of Saturation Mutagenesis Libraries
ies.[a]
The quality of a mutant library has been defined em-
pirically in terms of the observed frequency of hits
and their respective degree of catalyst improve-
ment.[10] In this section we address “library quality” in
a different way, which bears a relationship to the em-
pirical results. It concerns undesired amino acid bias,
which needs to be checked when generating mutant
libraries irrespective of the mutagenesis method. In
directed evolution studies this type of quality has
rarely been analyzed, primarily due to the high exper-
imental effort and the costs.[7,8,10,25] Since hits are gen-
erally found regardless of statistical considerations re-
lating to oversampling,[10,12,25] quality control of libra-
ries appeared not to be a pressing issue thus far. How-
ever, efficiency in laboratory evolution has become an
important focus of recent research.[8,10] Indeed, with-
out such checks, screening efforts can be wasted, i.e.,
“you should not search for something that does not
exist!” Libraries in which the circular template has
not been efficiently eliminated will require considera-
ble oversampling. Excessive appearance of wild-type
(WT) clones reduces library quality. Even worse, in
those cases in which certain codons are under-repre-
sented or even completely missing from the library,
even a high degree of oversampling will not solve the
problem. Therefore, at least some quality control is
required.
Since we wanted to test the utility of ISM,[11–13,15]
the question of quality control in the present context
was addressed. It can be argued that in general at
least 10% of the created clones should be sequenced
for a meaningful statistical analysis. Thus, even for a
library of 5,000 members, which is currently consid-
ered to be small in directed evolution, 500 plasmids
would have to be extracted and analyzed. In the pres-
ent study we applied a “short-cut” quality control as
reported earlier.[28] The quality was checked by per-
forming sequence analyses of pooled plasmids for
[a]
The % of each base was calculated from the sequencing
data of the pooled plasmids for each library. Results
from the simultaneous randomization at two positions
(Library 263) were derived from sequence analysis of 11
individual clones (Table S1, Supporting Information).
The following color code is used: green: A; red: T;
black: G; blue: C.
each library prior to transformation into the expres- forming simultaneous randomization at two distal po-
sion strain (Experimental Section). The results of sitions in YqjM (26/60), individual colonies were
such an analysis are shown in Table 1.
picked for sequencing, since this was the first time
It is apparent that even for this small-scale quality that we had employed the SLIC method.[29] The re-
control we find that the four bases are not equally sults show, inter alia, that almost complete elimination
represented as they ideally should be for A/T/C/G de- of the pET21a-YqjM WT template has been accom-
generacy. A small degree of bias (5–10%) can be ex- plished, yet a small degree of bias for certain bases/
pected to occur due to possible imbalance at the syn- codons still remains (Table S1, Supporting Informa-
thesis level of the primers. Yet another problem is the tion).
presence of a small but efficiently transformed circu-
lar template which survives digestion. Finally, some
bias due to different annealing efficiency of the Pooling as a Strategy for Reduced Screening Effort
pooled primers also needs to be taken into account.[30]
In the case of libraries A and B, generated by using Screening remains the bottleneck of directed evolu-
NNK codon degeneracy, we also created two ANK tion.[1,7,9] As delineated in the introduction, we have
degeneracy libraries in order to compensate for the developed iterative saturation mutagenesis (ISM) in
small percent of adenine in the first base. When per- order to address this problem. Coupled with the use
3290
asc.wiley-vch.de
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2009, 351, 3287 – 3305

Directed Evolution of an Enantioselective Enoate-Reductase
FULL PAPERS
of reduced amino acid alphabets,[10,12] this structure- case the scale of pooling is in a much lower range,
based or informatics-driven approach generally pro- namely only 8 clones per pool versus 96 clones per
vides high-quality libraries requiring considerably less pool, we suspected that this step can be skipped. In
screening. Nevertheless, any further reduction in the all cases automated GC was used in the screening
screening effort would add to the overall efficacy of process.
the method. We reasoned that a viable approach
The model compound 1a is hardly reduced by WT
could be the concept of pooling mutant libraries, as YqjM, only 0.6–1.2% conversion occurs under the 96
for example in the Bommarius approach.[31] In the deep-well plate conditions (Experimental Section).
present study we developed a different pooling strat- Therefore, the screening system was designed for
egy based on the original concept of Phizicky regard- identifying YqjM variants with enhanced activities.
ing the screening of predefined pools of proteins de- Libraries A, B, C and G were initially screened by
rived from genome sequencing as a method for link- performing bioreductions in 96-well microtiter plates
ing defined biological activities with previously un- (24 h incubation at 308C), extracting the 96 reactions
characterized proteins.[32] A possible pitfall in this ap- per plate with ethyl acetate and transferring the or-
proach is the low signal to noise ratio of the detected ganic layers into a 96-format glass microtiter plate for
parameter due to the high background; this relates to automated GC analysis. Careful analysis of the con-
high amounts of cells without activity, thereby mask- version data after 800 GC runs, i.e., tabulation of the
ing the sought-after enzyme property. The problem total conversion that corresponds to each column by
was subsequently solved by pooling followed by pro- adding the conversion of each of the eight wells from
tein purification by fusion to a GST-tag. Since in our which a 96 well plate consists, in combination with
Figure 2. Simulation of pooling experiments for libraries A, B, C and G. The X-axis shows the different columns in each
plate, the Y-axis the conversion that corresponds to each column. Dark grey bars indicate where positive YqjM variants
were found, and black dashed lines denote a potential threshold. Library G did not contain any hit.
Adv. Synth. Catal. 2009, 351, 3287 – 3305
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
3291

Despina J. Bougioukou et al.
FULL PAPERS
Table 2. Summary of screening effort.
the sequencing data, revealed that the positive clones
belong most often to the columns showing higher con-
version. Importantly, the results demonstrate that had
a “column” pre-screening of eight pooled wells been
performed, less than 50% of the columns would have
had to be analyzed (Figure 2).
A control experiment in which a positive variant
identified in library C was added from 0 to 8 times
per column in six different columns showed a good
correlation between conversion and the number of
hits (Figure S5, Supporting Information). Therefore,
all other libraries were screened using this kind of
pooling and subsequent deconvolution of positive col-
umns as illustrated schematically in Figure 3.
Library
Number of
Number of
GC samples
colonies picked
A
250
150[a]
190[a]
40[a]
B
250
C
200
G
200
25[a]
Other 16 sites
ISM
1920
920
1300
5040
18 days
960[b]
400[b]
410[b]
2175
8 days
C26/A60[c]
All libraries
Total GC time[d]
[a]
Based on the simulation experiments of pooling.
Based on pooling strategy experiments.
Library C26 A60 is the sum of four libraries as depicted
in Figure S4.
[b]
[c]
The complete results and analysis of a specific pool-
ing example can be found in the Supporting Informa-
tion (Figure S6 to Figure S11). The total screening
effort concerning the 20 single residue libraries han-
dled in this study is summarized in Table 2. It is ap-
[d]
Based on 5 min per GC run.
parent that the pooling strategy shortens the GC anal-
ysis time by a factor of at least two.
Screening Results
In the first round of saturation mutagenesis with for-
mation of 20 libraries, we aimed for increased activity
in the model reaction involving substrate 1a, without
setting any particular (high) threshold. Thus, we col-
lected and sequenced all YqjM mutants which show
any measurable conversion against the model com-
pound. Under the standardized screening conditions
(high load cell mass bioreductions, Experimental Sec-
tion), WT YqjM shows ca. 3% conversion with this
substrate, and moderate enantioselectivity (ee=76%)
in favor of (R)-2a. This allowed us to choose from a
broad set of variants regarding enantioselectivity,
which was subsequently measured for each hit. As
shown in Figure 4a, the results are remarkable for
several reasons. Firstly, not all libraries contain hits
(nine out of twenty), which means that the respective
amino acid positions are either not hot spots, or they
do represent such positions which however require at
least one further point mutation to unfold cooperative
effects. Secondly, both (R)- and (S)-selective active
Figure 3. Schematic protocol for screening by pooling de-
fined-cell cultures overexpressing YqjM variants. a) Pick
and inoculate individual colonies. b) Induce expression of mutants are present in one and the same library
YqjM variants. c) Recover by centrifuging individual cell
pellets and combine all eight cell pellets belonging to the
same column. d) Lyse cells and incubate 12 biotransforma-
tions per plate by adding the appropriate reagents. e) Ex-
tract product with organic solvent and analyze organic layer
by GC. Column hits will be identified in this step by setting
an appropriate threshold. f) Using as a reference the plate
from step e) return to master plate in step a) and re-exam-
ine the columns of interest, if there is any, by adding re-
agents to each well separately. For the complete experimen-
tal protocol see Experimental Section.
(Figure 4, b), which means that the nature of the
newly introduced amino acid at a given position de-
termines the sense of enantioselectivity. Thirdly, nota-
bly high ee values of up to 91% (R) and 84% (S) are
achieved in the first round of mutagenesis, although
saturation mutagenesis was performed at single amino
acid sites which exclude the possibility of cooperative
effects.
A limited number of iterative saturation mutagene-
sis (ISM) experiments was performed in the second
3292
asc.wiley-vch.de
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2009, 351, 3287 – 3305

Directed Evolution of an Enantioselective Enoate-Reductase
FULL PAPERS
Figure 4. Performance of individual sites after the first and the second round of saturation test, the model reaction being
1a!2a. In both graphs, X-axis shows the sites that responded positively to saturation test. The letter X is used when a partic-
ular residue can be replaced by more than one amino acid. a) Y-axis represents the mean conversion of the hits found in
each site (light gray bars, first round; dark gray bars, second round). Error bars are shown the upper and lower limits from
the mean conversion value. When only one amino acid could replace the site of interest, the error bars indicate the maxi-
mum experimental error on the conversion values (i.e., ꢁ20%). C26D/I69X mutants have shown 100% conversion even at
twice higher concentration of 1a. b) Y-axis represents the average enantioselectivity that each site exhibits (R and S values
were treated separately). Error bars are shown the upper and lower limits from the mean enantioselectivity value. When the
observed enantioselectivity is represented by only one amino acid, the error bars indicate the maximum experimental error
(i.e, ꢁ2.0%). Conversion was calculated by GC integration of the substrate and product area peak, after 24 h incubation at
308C, using the high-load cell mass protocol, described in Experimental Section [i.e., 2.5 mM substrate, 0.7ꢁ0.2 g (wet
weight mass) cells, 100 mM glucose, 0.5 mM NADP+; 50 U glucose dehydrogenase, in 5 mL of 100 mM phosphate buffer,
pH 7.0]. Enantioselectivity was determined by chiral GC column.
round by choosing the genes of some of the “best the iterative process, i.e., only the double mutants
hits” as templates for randomizing at another amino lead to excellent results.
acid position that had responded positively in the first
Figure 6 shows the data of single and double mu-
round. The quality of each hit is defined by two num- tants arranged according to enantioselectivity, activity
bers, % conversion (0–100) and % ee (0–100) and one as measured by % conversion being included in each
symbol (R or S). Choosing the “best hits” to be used case. Some interesting trends emerge. In the case of
in subsequent ISM was guided by multiplying the two the most (R)-selective mutants (ee=94–97%), activity
numbers and relying on those that have the highest is generally also high, as in the apparently most active
score (Table 3), i.e., not by picking those of highest one Cys26Asp/Ile69Thr with ee=94%. In this case a
activity. Complete systematization was not strived for, 130-fold rate improvement in converting substrate 1a
because this limited exploration comprising less than was observed, even at higher concentrations up to
10 libraries already led to the identification of a 10 mM (Supporting Information). A similar result was
number of double mutants with significantly im- observed in the case of mutant Cys26Asp/Ile69Val
proved activities, which were then checked for enan- having a somewhat lower enantioselectivity ee=84%.
tioselectivity. Some of the best variants were subse- In the (S)-selective regime, a correlation between
quently characterized by separate kinetic experiments enantioselectivity and activity is hardly discernable.
(see below). The pathway scheme and the results in For example, although mutant Cys26Gly/Ala60Val
terms of enantioselectivity are summarized in has both excellent enantioselectivity and good activi-
Figure 5. It can be seen that several mutants were ty, other highly enantioselective variants show rela-
identified in the (R)- and (S)-selective regimes, re- tively low activity, e.g., Tyr70His/Ala60Cys with ee=
spectively, all with ee >95%. This requires in all cases 94% and Cys26Gly/Ala60Ile with eeꢀ99%.
Adv. Synth. Catal. 2009, 351, 3287 – 3305
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
3293

Despina J. Bougioukou et al.
FULL PAPERS
Table 3. Evaluation of the active mutants after the first round of directed evolution, the model reaction being 1a!2a.
Entry[a] Active mutant Total Impact[b] Entry Active mutant Total Impact Entry Active mutant Total Impact
1
2
3
4
5
6
7
8
A104W
A104H
A104Y
I69T
A60V
I69V
A104F
D73I
C26D
Y28A
C26W
C26A
ꢂ4086
ꢂ2498
ꢂ2184
ꢂ2122
+1395
ꢂ1242
ꢂ1190
+1117
+1049
+987
13
14
15
16
17
18
19
20
21
22
23
24
Y28S
+873
+761
+722
+673
ꢂ623
+589
+476
+469
ꢂ454
+442
ꢂ378
+280
25
26
27
28
29
30
31
32
33
34
35
E111F
A106C
WT
D73Q
D73L
E111A
D73V
T70G
A106S
T70C
+278
ꢂ266
ꢂ233
+228
ꢂ204
+179
+168
+87
+66
+14
ꢂ12
T70H
A6ßC
Y28P
D111G
C26G
Y28V
Y28C
C26N
C26V
A106G
Y28I
9
10
11
12
ꢂ983
A60W
ꢂ890
[a]
Variants are sorted based on the absolute value of their total impact which reflects their impact on conversion, Ic and
enantioselectivity, Ie (total impact=IcꢀIe). The chosen starting points for the second round of evolution are denoted in
bold.
[b]
Positive values indicate (S)-selectivity and negative values denote (R)-selectivity. (See also Table S2, Supporting Informa-
tion).
Figure 5. Scheme for iterative saturation mutagenesis ISM applied to YqjM, the bioreduction of substrate 1a being the
model reaction. a) Two-dimensional map of the amino acids selected for saturation mutagenesis (gray shells illustrate posi-
tive and white negative response). Distances were measured from the Ca carbon of the amino acid of interest to the closest
carbon of the pPHB ligand, using the 1Z42 PDB file as a reference. b) Chosen ISM pathways. c) Pathways leading to the
highest enantioselectivities. Light gray dashed lines indicate the first round of mutagenesis, black dashed lines the second
round. Filled diamonds indicate the starting point and filled cycles indicate the ending point of the ISM pathway.
3294
asc.wiley-vch.de
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2009, 351, 3287 – 3305

Directed Evolution of an Enantioselective Enoate-Reductase
FULL PAPERS
Figure 6. Best YqjM mutants from the first and second round of saturation mutagenesis, the model reaction being the biore-
duction of substrate 1a. The variants are arranged according to their respective enantioselectivity. Gray bars: (R)-selectivity;
open bars: (S)-selectivity; black bars: % conversion under standardized reaction conditions as specified in Figure 4.
Lastly, we simultaneously randomized residues
Cys26 and Ala60, applying the SLIC-based method
for library creation. This double residue site was eval-
uated to see if any better double combinations were
missed when applying the iterative strategy based on
randomizing single amino acid positions. Significantly
improved variants were not found.
Substrate Scope of Best Mutants
The well-known credo in directed evolution, “you get
what you screen for”, is certainly correct,[7] but for or-
ganic chemists it is discouraging because evolving an
enantioselective catalyst for a single compound in a
specific and useful case does not advance the field of
asymmetric catalysis significantly enough. Rather, a
Scheme 2. Bioreduction of additional substrates using the
best YqjM mutants evolved for the model reaction 1a!2a.
given catalyst should cover at least a small range of
different substrates. Therefore, we tested some of the
best (R)- and (S)-selective YqjM mutants evolved for amples are recorded in Table 4. Similarly to what we
the model reaction 1a!2a as catalysts in the asym- observed for the model compound 1a, enantiocom-
metric reduction of a collection of other 3-substituted plemetary variants emerged from this small pool of
cyclohexenones 1b–e and the cyclopentenone ana- hits for most of the compounds tested in this study.
logues 3a, b (Scheme 2) without performing addition- Unfortunately, (S)-selective variants are significantly
al mutagenesis/screening experiments.
less active compared to their (R) counterparts,
Compounds 1b–d are not substrates for the WT Figure 7. Alkene reductases able to deliver the (S)-3-
YqjM. After screening our YqjM variants, in a 0.5- substituted alkyl ketones have already been identified
mL reaction scale (low-load cell mass protocol in Ex- by other research groups.[33]
perimental Section), we were delighted to learn that
However, enoate reductases able to produce the
our mutants efficiently catalyze the asymmetric trans- (R)-enantiomer, at least for this class of compounds,
formation of compounds 1b–d. Some prominent ex- have not been reported prior to this study. Compound
Adv. Synth. Catal. 2009, 351, 3287 – 3305
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
3295

Despina J. Bougioukou et al.
FULL PAPERS
Table 4. Substrate scope of some YqjM mutants.
several YqjM variants delivered the same product in
excellent enantiomeric purity (Table 4). Most impor-
tantly, a single amino acid mutation (i.e., C26G) was
found to induce complete reversal of the enantiose-
lectivity of YqjM mediated reductions. Notably, three
more (S)-selective double mutants performed equally
well under the reaction conditions for both com-
pounds (Table 4 and Figure 8).
Lastly, it is worth noting that during the course of
the above experiments, we observed the formation of
significant amounts of cyclic saturated alcohols (espe-
cially for compounds 1b–d). The regeneration system
used during this study (i.e., GDH/glucose) is not re-
sponsible for the ketone reduction since replacement
of this system by phosphite dehydrogenase/phosphite
or by stoichiometric amount of NADPH did not solve
the problem (Figure S13, Supporting Information).
Therefore, an endogenous E. coli keto-reductase is
most likely acting on the carbonyl moiety after the
action of the enoate reductase. By adding EDTA to
the buffer solution, the latter reaction can be sup-
pressed to synthetically acceptable levels, but not
completely eliminated (see Table 7, below). More-
over, the keto-reductase activity does not jeopardize
the enantioselectivity of the YqjM variants which, de-
spite the percentage of formed alcohol, remained un-
Substrate
Product 2 conversion [%]/ee
Enzyme YqjM
1b
1b
1b
1c
1c
1c
1d
1d
1e
1e
1e
–/–[a]
100/99 (R)
9/96 (S)
–/–[a]
33/99 (R)
2/91 (S)
–/–[a]
92/99 (R)
85/99 (R)
92/99 (R)
100/98 (S)
WT
C26D/I69T
C26G/A60 V
WT
C26D/I69T
C26G/A60 V
WT
C26D/I69T
WT
C26D/I69T
C26G
Substrate
Product 4 conversion [%]/ee
Enzyme YqjM
3a
3a
3a
3b
3b
3b
1[b]/nd[c]
WT
5[b]/50 (R)
36[b]/93 (S)
100/96 (R)
100/99(R)
100/93 (S)
C26D/I69T
C26D
WT
I69T
C26G
[a]
No conversion.
Under high-load cell mass conditions.
Not determined.
[b]
[c]
3a was and remains a particularly difficult to reduce changed during the course of the biotransformations.
substrate for WT YqjM and mutants described
herein.
Compounds 1e and 3b are of special interest since Purification and Catalytic Activity of YqjM and
a second functionality, namely an ester group, has Selected Mutants
been introduced at the b-C-atom of the cyclic enones.
WT YqjM converts both substrates to the corre- When purifying YqjM, it has been reported that the
sponding (R)-products very efficiently. In addition, use of GST- or His-tags significantly impairs catalytic
Figure 7. Performance of YqjM mutants from the first and second round of saturation mutagenesis with compounds 1b–d.
Mutants are sorted based on decreased conversion (as a sum of saturated ketone and alcohol) for the (R)- enantiomer
(black bars).
3296
asc.wiley-vch.de
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2009, 351, 3287 – 3305

Directed Evolution of an Enantioselective Enoate-Reductase
FULL PAPERS
Figure 8. Performance of YqjM mutants from the first and second round of saturation mutagenesis with compounds 1e
(upper panel) and 3b (lower panel). Mutants are sorted based on decreased conversion towards the (R)-enantiomer.
activity relative to the native forms.[34] Therefore, we a reference which is often employed in OYE research.
utilized a special affinity matrix comprising N-(4-hy- Table 5 summarizes the results regarding specific ac-
droxybenzoyl)aminohexyl-agarose
developed
for tivity. For the model substrate 1a, which was used in
OYEs by Massey, which had been applied to the pu- the mutagenesis/screening experiments, the specific
rification of the OYE from brewersꢄ bottom yeast.[35] activity of the mutants increases by a factor ranging
The respective affinity of the oxidized form of OYE between 2 and 48. The situation with the additionally
to this particular matrix results from the tight binding tested substrates 1e, 3b and 5 is different because
of aromatic and especially phenolic ligands. Following some variants are more active than WT, others less.
purification of WT YqjM and six variants (Fig-
We also measured the apparent steady state kinetic
ure S14, Supporting Information), kinetic studies parameters of the same enzymes using the model
were performed using substrates 1a, 1e and 3b togeth- compound 1a and cyclohexenone (5). The data assem-
er with the parent compound 2-cyclohexenone (5) as bled in Table 6 (for Michaelis–Menten plots see Sup-
Adv. Synth. Catal. 2009, 351, 3287 – 3305
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
3297

Despina J. Bougioukou et al.
FULL PAPERS
Table 5. Summary of specific activity data in mU/mg (average of three measurements).
Substrate
Enzyme YqjM
1a
1e
3b
5
Wild-Type
4ꢁ3
1614ꢁ202
2163ꢁ183
1711ꢁ223
2538ꢁ14
753ꢁ25
2628ꢁ78
1489ꢁ148
2491ꢁ94
2262ꢁ8
928ꢁ98
351ꢁ7
2560ꢁ98
2669ꢁ171
1117ꢁ113
1827ꢁ97
828ꢁ17
C26D/I69T
C26D/A104F
A60C/C26G
A60C/C26D
C26G/A60V
C26G/A60I
190ꢁ5
19ꢁ2
23ꢁ2
7ꢁ2
20ꢁ1
25ꢁ1
510ꢁ18
537ꢁ10
337ꢁ64
311ꢁ55
383ꢁ3
Table 6. Apparent steady-state kinetic parameters.
Substrate
1a
5
kcat [sꢂ1]
Km [mM]
kcat/Km [sꢂ1 mꢂ1]
kcat [sꢂ1]
Km [mM]
kcat/Km [sꢂ1 mꢂ1]
Enzyme YqjM
[a]
Wild-Type
–
–
–
2.2
2.6
0.05
1.3
0.3
0.5
0.3
0.28
ꢀ1.5
ꢀ2.3
ꢀ2.0
0.09
0.43
0.49
7857
ꢃ1733
ꢃ22
C26D/I69T
C26D/A104F
A60C/C26G
A60C/C26D
C26G/A60 V
C26G/A60I
0.16
ꢀ4.7
ꢀ3.0
–
ꢃ35
ꢃ11
–
~0.04
[a]
–
–
ꢃ650
3333
1162
612
[a]
–
–
~0.03
ꢀ1.2
ꢃ27
[a]
–
–
–
[a]
We were unable to determine values for this case.
porting Information) likewise show that the (R)-selec- (5), it becomes apparent that all mutants are less
tive mutant Cys26Asp/Ile69Thr surpasses the other active than WT YqjM, which is not surprising.
tested mutants in terms of kcat/Km values. In the case
of the model compound 1a, WT YqjM shows such
low activity that it was not possible to obtain steady- Testing the Applicability of YqjM Mutants
state kinetic parameters, in contrast to mutant
Cys26Asp/Ile69Thr which is characterized by a kcat/Km It is often stated that mere kinetic parameters are not
value of about 35. This represents a notable rate ac- sufficient for evaluating the usefulness of enzyme in a
celeration, but in absolute terms activity is fairly low. biocatalytic process.[2b,36] Therefore, based on the data
It should be noted that since the enzyme operates by we collected in the two previous sections, we attempt-
a ping-pong mechanism, the value of such data is lim- ed to scale-up some bioreductions. The results of
ited. When analyzing the results for cyclohexenone these efforts are summarized in Table 7. We used ca.
Table 7. Scale-up bioreductions.
Substrate
YqjM Variant[a]
Substrate Loading[b]
Conversion, Time
% ee
Yield[c]
1a
1a
1b
1b
1d
1e
1e
C26D/A104F
C26G/A60V
C26D/A104F
C26G/A60V
C26D/A104F
C26D/A104F
C26G/A60C
2.0
2.0
2.5
2.5
2.8
15
99%, 5 h
90%, 5 h
99%, 5 h
63%, 5 h
93%, 5 h
97%, 1 h
95%, 1 h
96.6 (R)
97.4 (S)
40%
59%
80%[d]
72%
71%
90%
63%
ꢀ99.0 (R)
91.6 (S)
ꢀ99.0 (R)
ꢀ99.0 (R)
97.0 (S)
15
[a]
Biocatalyst loading: 5 gLꢂ1.
[b]
[c]
[d]
In gLꢂ1.
Sum of recovered material including starting material.
Mixture of ketone and alcohol in 6:1 ratio.
3298
asc.wiley-vch.de
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2009, 351, 3287 – 3305

Directed Evolution of an Enantioselective Enoate-Reductase
FULL PAPERS
Figure 9. Induced fit docking (IFD) poses. a) Compound 1e docked into wild-type YqjM. b) Compound 1e docked into
YqjM-C26G mutant. c) Compound 3b docked into wild-type YqjM. d) Compound 3b docked into YqjM-C26G mutant. In
all poses thin lines represent the wild-type structure before the optimization with the IFD protocol; grey tubes represent the
position of the amino acids after the IFD. Cys26 shown as blue tubes in panel a) and b) has been replaced by Gly in panel
b) and d). Dashed lines represent H-bond interactions.[24]
50 mg of flavoprotein in 10 mL final reaction volume ing the source of enhanced activity and stereoselectiv-
for all reactions (Experimental Section/Scale-Up Bio- ity of some of the YqjM mutants. The gross mechanis-
reductions).
tic features of YqjM and of other enoate-reductases
Overall, the results presented in Table 7 are in line are well known.
with the specific activity values reported in Table 5,
His164 and His167 appear to be essential for car-
proving the excellent performance of two double mu- bonyl activation by H-bonding, while Tyr169 probably
tants with substrate 1e, whereas the results for com- delivers the proton following hydride transfer from
pounds 1a–1d are, as expected, moderated. However, reduced flavin (Figure 1). Based on kinetic measure-
there is still room for improvement, at least for the ments with other substrates, it has been proposed that
production of the (R)-ketones, if the Cys26Asp/ YqjM prefers substrates with a second carbonyl group
Ile69Thr variant is used instead (data not shown).
opposite to the first one, in such cases Tyr28 and
Arg’336 defining a second binding site.[21] In order to
see if this holds for the two substrates 1e and 3b con-
taining an additional ester function, induced docking
experiments were performed. Indeed, this additional
Molecular Modeling
Although the primary thrust of this study was testing binding motif was clearly identified (Figure 9). More-
the utility of ISM as a strategy in directed evolution, over, the preferred flipped binding mode above flavin
we nevertheless performed induced fit docking ex- in the case of reversed enantioselectivity is indicated.
periments in order to gain some initial insight regard-
Adv. Synth. Catal. 2009, 351, 3287 – 3305
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
3299

Despina J. Bougioukou et al.
FULL PAPERS
Conclusions
study, in the present case activity and enantioselectivi-
ty. There are several ways to approach this problem,
The present study aimed at testing the viability of e.g., to evolve one property first, followed by the opti-
iterative saturation mutagenesis (ISM) in the form of mization of the other catalytic parameter. In the pres-
CASTing as an efficient strategy in directed evolution ent study we opted for another approach which in-
of enantioselective enzymes. Increasing the activity volves simultaneous evolution of both parameters. In
and enhancing as well as inverting the enantioselectiv- doing so, it was crucial not to apply too stringent con-
ity of the OYE homologue YqjM served as the exper- straints, i.e., YqjM variants were selected in the
imental platform. Our efforts proved to be successful, screening process with a fairly broad range of activity,
requiring the screening of less than 5,000 transform- in contrast to considering only the most active ones.
ants for evolving active (R)- and (S)-selective mutants Especially when choosing single residue saturation
in the model bioreduction of 3-methylcyclohexenone mutagenesis as performed here, it appears wise to ac-
(1a). Generalizing ISM is important, because this ap- cumulate a panel of medium- and higher-quality hits
proach to directed evolution appears to require only activity, which are then screened for the second cata-
small mutant libraries and therefore minimal screen- lytic property (enantioselectivity) before going into
ing effort, which is otherwise the bottleneck of labora- the next mutagenesis/screening round. These kinds of
tory evolution.
non-discarded variants can be related to neutral
In the process of reaching the goals defined in this drifts,[37] although the term “lateral hits” may be more
project, several problems were solved which deserve appropriate. We also note a relationship to the Eigen/
special mention. The CAST analysis based on the X- Schuster notion of quasi-species[38] which has been in-
ray structure of inhibitor-bound YqjM pointed to 20 voked by Mannervik in other directed evolution
amino acid positions as potential saturation mutagen- work.[39] The approach utilized here is illustrated in
esis sites. We had to make a decision whether to treat Figure 10, the upper right section of the scheme har-
them separately as single residue sites or to group boring the desired variants with improved properties
them according to some scheme in which a lower of both catalytic parameters A and B.
number of sites are considered, each composed of
more than one amino acid position. The latter would
allow for cooperative effects in the initial libraries,
but we nevertheless chose the first option. The posi-
tive results demonstrate that even without the possi-
bility of initial cooperative effects, success may well
set in. Nevertheless, the relative utility of grouping
single residues when performing ISM in the YqjM
system needs to be investigated in future studies.
Whereas choosing single residue sites means less
screening in each individual library, the high number
of libraries designed in the present case does consti-
tute a formidable screening effort. In order to solve
this problem, a pooling strategy based on the method
of Phizicky was developed, which reduced the overall
screening effort by more than 50%. The benefits of
the pooling approach are especially pronounced when
a saturation mutagenesis site is incorrectly chosen
(not a hot spot), because in such cases the “incorrect”
site is rapidly eliminated. For example, instead of 96
GC runs per plate, 12 are sufficient because there is
no need for deconvolution.
Finally, the question of substrate scope of the
Figure 10. Preferred approach for the simultaneous optimi-
evolved (R)- and (S)-selective variants was addressed
zation of two catalyst properties A and B. Black star indi-
by testing them as catalysts in the asymmetric biore-
cates the desired variant; blue and green dashed lines: strin-
duction of several other substrates. It was discovered
gent thresholds; blue and green rectangles: relaxed thresh-
that a few, but not all, substrates were reduced with
olds; blue and green filled circles: best mutant for property
enantioselectivities of >95% ee without the need to
A and B, respectively, which are not used in further muta-
perform any new mutagenesis/screening experiments.
genesis; red-crossed blue and green circles: variants with im-
The important question in directed evolution re-
garding the simultaneous improvement of two differ-
proved property A or B; red-crossed black circles: mutants
with improved A and B property. Black dashed arrows:
ent catalytic properties was also addressed in this second round of mutagenesis.
3300
asc.wiley-vch.de
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2009, 351, 3287 – 3305

Directed Evolution of an Enantioselective Enoate-Reductase
FULL PAPERS
template (pET21a-YqjM, 2.5 ng) and 0.5 mL of KOD. The
PCR reaction started at 948C (3 min), continued with thirty
cycles of 948C (15 sec), 568C (1 min) and 688C (1 min)
without final extension and storage at 48C. Each PCR prod-
uct (2ꢃ50 mL) was digested with DpnI and purified as de-
scribed above. The 2ꢃ50 mL PCR reactions for amplification
of pET21a-YqjM backbone contained 5 mL of 10 ꢃ KOD
hot start polymerase buffer, 5 mL of dNTPs (2 mM each),
5 mL of YqjM_A60_FW primer (2.5 mM), 5 mL of
YqjM_C26_RV primer (2.5 mM), 2 mL MgSO4 (25 mM),
template (pET21a-YqjM, 2.5 ng) and 0.5 mL of KOD. Each
PCR product (2ꢃ50 mL) was digested with DpnI and puri-
fied as described above. One mg of insert and one mg of
vector were treated, separately, with 0.5 U of T4 DNA poly-
merase (228C, 40 min) in a 30 mL and 20 mL reaction
volume, respectively. containing the appropriate amount of
T4 polymerase buffer and BSA. The reactions were termi-
nated by adding 1/10 volume of dCTP (10 mM) and stored
on ice. In the reaction tube containing the T4-treated insert
(33 mL) one mL of the T4-treated vector (ca. 50 ng) was
added followed by 4 mL of 10ꢃT4 ligase buffer (NEB), 1 mL
of distilled water and 1 mL of RecA (2 mg). The annealing
reaction mixture was incubated at 378C for 30 min and
stored at 48C. An aliquot of 20 mL of the annealing reaction
mixture was used to transform 200 mL of E. coli DH5a
chemically “ultra-competent” cells. The transformation mix-
ture was incubated with 1 mL of SOC medium at 378C with
shaking and spread on LB-agar plates containing 100 mg/mL
carbenicillin. We obtained between 300 and 800 colonies per
transformation. Plasmid DNA was extracted from 32 ran-
domly picked colonies grown overnight in LBCB (100 mg/mL)
medium and sequenced.
Experimental Section
General
Restriction enzymes, T4 DNA Polymerase, and RecA were
purchased from New England Biolabs. KOD hot start Poly-
merase was obtained from Novagen. Primers and dCTP
were obtained from Invitrogen. Glucose dehydrogenase,
phosphite dehydrogenase, NADP+ and NADPH were pur-
chased from Codexis. Glucose oxidase from Aspergillus
niger (type VII) was obtained from Sigma. E. coli strain
BL21 Gold (DE3) was obtained from Stratagene. TB
medium contained 24 g/L yeast extract, 12 g/L peptone,
4 mL/L glycerol, 0.017M KH2PO4 and 0.072MK2HPO4.
SOC medium contained 20 g/L yeast extract, 5 g/L peptone,
20 mM glucose, 10 mM NaCl; 2.5 mM KCl, 10 mM MgCl2
and 10 mM MgSO4. SOB medium was prepared as the SOC
medium excluding glucose. Inoue transformation buffer con-
tained 50 g/L sucrose (optional), 10 mM PIPES (10 mL/L of
a 0.5M PIPES solution, pH 6.7), 50 mM KCl, 15 mM CaCl2
and 55 mM MnCl2. All chemicals were obtained from com-
mercial suppliers and used without further purification. GC
analysis was conducted with one of the following columns:
HP-5, Lipodex A, BGB-178/OV-1701, Hydrodex-BDBDAC.
Library Creation based on the QuikChange Protocol
The PCR reactions contained 5 mL of 10ꢃ KOD hot start
polymerase buffer, 5 mL of dNTPs (2 mM each), 10 mL of the
appropriate forward and reverse primers (2.5 mM each), 2 mL
of MgSO4 (25 mM), the appropriate template (20 ng) and
0.5 mL of KOD polymerase in a final volume of 50 mL dis-
tilled water. The PCR reaction started at 958C (3 min), con-
tinued with twenty cycles of 958C (1 min), 52–608C (anneal-
ing temperature depending on the particular set of primers,
1 min) and 728C (8 min) and finished with 10 min at 728C
and storage at 48C. To ensure elimination of the circular
polymethylated template plasmid, 20 mL of PCR reaction
mixture were mixed with 1 mL DpnI (10 U/mL) and incubated
overnight at 378C, followed by an additional 1 mL of DpnI
for 3 h. Upon purification of the DpnI-digested product with
a QIAquick PCR purification spin column, an aliquot of 3 mL
was used to transform 100 mL of E. coli DH5a chemically
“ultra-competent” cells, prepared in house base on the Inoue
protocol (see Supporting Information). The transformation
mixture was incubated with 700 mL of SOC medium at 378C
with shaking and spread on LB-agar plates containing 100 mg/
mL carbenicillin (CB). More than 300 colonies were obtained
per transformation. For higher transformation efficiency elec-
tro-competent cells can be used instead. All colonies were
scraped from the plate with the help of 5 mL LB medium.
The corresponding plasmid pool was extracted with a QIAp-
rep spin column and sequenced in order to evaluate the
codon degeneracy of the saturated position.
Screening of Libraries
One mL of the plasmid pool from each library was used to
transform 30 mL of in house prepared, electro-competent E.
coli BL21 Gold (DE3) cells. The transformation mixture
was incubated with 700 mL of LB medium at 378C with
shaking and spread on LB-agar plates containing 100 mg/mL
carbenicillin. Individual colonies were randomly picked and
used to inoculate 96 deep-well plates containing 900 mL
LBCB (100 mg/mL) medium per well. Cultures were grown over-
night at 378C with shaking (800 rpm). Normally, one column
of a plate was assigned for culturing the wild-type YjqM.
Glycerol (20% v/v) stock solutions, in sterile 96-well micro-
titer plates, were prepared at this stage and kept at ꢂ808C.
Aliquots of 100 mL overnight cultures were used to inoculate
new 96 deep-well plates containing 700 mL of TB medium
supplemented with CB (100 mg/mL) and isopropyl thio-b-d-
galactoside (IPTG, 100 mM final concentration). After five
hours incubation at 308C with shaking (800 rpm) the plates
were centrifuged and the supernatant was discarded. Each
pellet was resuspended in 110 mL glycerol (10% v/v) and the
cells (from eight wells per column) were pooled. The glycer-
ol solution was removed by centrifugation and the pellets
were kept at ꢂ808C for at least one hour. Aliquots of
250 mL potassium phosphate buffer (100 mM, pH 7.0) sup-
plemented with 2 mg/mL lysozyme were added to each well
of the plate and the cell pellets resuspended by shaking
(1000 rpm). The plates were incubated at 378C for one hour
with shaking (200 rpm). The biotransformations were started
Library Creation based on in vitro Homologous
Recombination Protocol
The insert encompassing the Cys26 and Ala60 saturation
codons was amplified in 2ꢃ50 mL PCR reactions containing
5 mL of 10ꢃ KOD hot start polymerase buffer, 5 mL of
dNTPs (2 mM each), 10 mL of the appropriate forward and
reverse primers (2.5 mM each), 2 mL of MgSO4 (25 mM),
Adv. Synth. Catal. 2009, 351, 3287 – 3305
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
3301

Despina J. Bougioukou et al.
FULL PAPERS
by addition of 250 mL reaction solution (100 mM potassium
phosphate buffer, pH 7.0, 200 mM glucose, 2 mM NADP+;
2–3 U glucose dehydrogenase, 10 mM 3-methyl-2-cyclohexe-
none) per well. The incubation was continued for 24 h at
308C with shaking (150 rpm). The reaction mixture of each
well was extracted (vigorous extraction was not necessary)
with 800 mL of ethyl acetate, the plate was centrifuged, and
a portion (ca. 500 mL) of the organic layer was transferred
to a GC vial or to a glass 96 well plate for analysis. Individu-
al cell pellets were screened as described above by simply
omitting the pooling step and the addition of glycerol solu-
tion after the centrifugation of the expression culture plates.
Plasmids from positive variants were extracted from over-
night cultures inoculated with the appropriate clone from
the ꢂ808C glycerol stock plates, and sequenced.
tion, 250 mL of the organic layer were transferred to a GC
vial for achiral and chiral GC analysis.
Protein Isolation of YqjM Wild-Type and Mutants
An overnight culture of the appropriate overexpression
strain grown in LBCB (100 mg/mL) medium was diluted 1:100 into
2ꢃ200 mL of TBCB (100 mg/mL) in 2ꢃ1-L flasks. The culture was
shaken at 378C until the optical density at 600 nm reached
0.7, then IPTG was added to a final concentration of
100 mM and the culture was shaken for additional 5 h at
308C. Cells were collected by centrifuging at 4,000ꢃg for
8 min at 48C. The cell pellet (ca. 4 g wet mass) was washed
once with 30 mL of 100 mM Tris-Cl pH 8.0 containing
100 mM (NH4)2SO4 (Buffer A), resuspended in 15 mL of
the same buffer and kept at ꢂ808C for at least 1 hour. The
solution was thawed and frozen twice before addition of
phenylmethanesulfonyl fluoride (PMSF) 0.1 mM (final con-
centration) and disruption by sonication (Bandelin,15ꢃ10
sec with 10 sec interval, 40% pulse, on water-ice bath). Cell
debris was removed by centrifugation at 18,000ꢃg for
45 min at 108C. The supernatant (15 mL) was saturated
(80%) with 8.0 g of solid ammonium sulfate, added over a
period of 15 min with constant stirring at 48C. The solution
was stirred for additional 45 min before centrifugation at
18,000ꢃg for 45 min at 108C. The supernatant was discarded
and the precipitated proteins were redissolved in the mini-
mum amount of Buffer A (ca. 7.5 mL final volume). The so-
lution was desalted by passing it (3ꢃ2.5 mL) through a PD-
10 Desalting column (GE Healthcare) pre-equilibrated with
Buffer A. An aliquot of 5 mL desalted sample was incubat-
ed with excess of FMN (1 mg/mL) overnight at 48C. The
excess of free flavin was removed by passing the protein so-
lution through a PD-10 Desalting column twice. The sample
was concentrated using an ultrafiltration centrifugal filter
(10 kD cut-off membrane, Amicon) to a final volume of
5 mL and was applied (4.5 mL) to an N-(4-ydroxybenzoyl)a-
minohexyl-agarose affinity column pre-equilibrated with
Buffer A. The column was washed with Buffer A until the
absorbance at 280 nm was less than 0.4. The bound flavopro-
tein was eluted with 20 mL of Buffer A containing 5 mM
sodium dithionite. The eluted protein was concentrated
using an ultrafiltration centrifugal filter (10 kD cut-off mem-
brane, Amicon), then was diluted again with 15 mL of
Buffer A and reconcentrated by ultrafiltration. The dilution-
concentration step was repeated twice in order to remove
the excess of sodium dithionite. During this process, the pro-
tein solution retained in the Amicon filter turned yellow.
Mutant Cys26Asp/Ile69Thr was not retained on the top of
the affinity column and eluted from the column after the
first 15 mL of Buffer A flowed through. Two fractions of
eluted flavoprotein (5 mL each) were collected. The purified
protein samples were stored in aliquots at ꢂ208C. The puri-
fied mutant Cys26Asp/Ala104Phe was found to lose activity
after repeated cycles of freezing and thawing. The purity of
the protein samples (ꢀ 85%) was evaluated by SDS-PAGE
(12.5%).
High-Load Cell Mass Biotransformation
A 5 mL LBCB (100 mg/mL) overnight preculture of the desired
YqjM variant was diluted 1:100 into 200 mL of TBCB (100 mg/mL)
medium in a 1-L flask. This was shaken at 378C until the
O.D.600 was 0.7 (ꢁ0.2). IPTG was added to a final concen-
tration of 100 mM and the culture was shaken for additional
5 h at 308C. The liquid culture was transferred to four 50-
mL Falcon tubes; cells were collected by centrifugation
(5,000ꢃg for 10 min at 48C) and kept at ꢂ808C. Into the
50-mL Falcon tube containing 0.7ꢁ0.2 g (wet mass) of cells,
2.5 mL of potassium phosphate buffer (100 mM, pH 7.0)
supplemented with 2 mg/mL lysozyme were added and the
solution was incubated at 378C for one hour with shaking
(150 rpm). The biotransformation was started by addition of
2.5 mL reaction solution (100 mM potassium phosphate
buffer, pH 7.0, 200 mM glucose, 1 mM NADP+; 50 U glu-
cose dehydrogenase, 5 mM of the appropriate substrate).
The reaction mixture was incubated at 308C for 24 h with
shaking (120 rpm). The pH was adjusted to 7.0 with 1M
KOH after 4 h incubation. The reaction mixture was extract-
ed with 4 mL of ethyl acetate, centrifuged for 5 min at
5,000ꢃg and the organic phase was analyzed by chiral GC
column. As a negative control, an E. coli BL21 Gold (DE3)
strain harboring a pET22b plasmid was cultured and treated
in the same manner. In the latter case, GC analysis showed
no detectable amount of ketone 2.
Low-Load Cell Mass Biotransformation
YqjM wild type and YqjM mutants were cultured in a 96
deep-well plate at conditions identical to those described in
the library screening section. Cell pellets (ca. 10 mg wet
weight mass) were kept at ꢂ808C until used. Aliquots of
250 mL potassium phosphate buffer (100 mM, pH 7.0) sup-
plemented with 2 mg/mL lysozyme were added to each well
of the plate and the cell pellets were resuspended by shak-
ing (1000 rpm). The plates were incubated at 378C for one
hour with shaking (200 rpm). The biotransformations were
started by addition of 250 mL reaction solution (100 mM po-
tassium phosphate buffer pH 7.0, 200 mM glucose, 2 mM
NADP+, 2–3 U glucose dehydrogenase, 10 mM of the appro-
priate substrate) per well. The incubation was continued for
24 h at 308C with shaking (150 rpm). The reaction mixture
of each well was transferred into a micro-centrifuge tube
and extracted with 500 mL of ethyl acetate. After centrifuga-
Enzyme Activity Assays
Enzyme activity was determined spectrophotometrically at
258C by following the decrease in absorbance of NADPH
(molar extinction coefficient 6,220Mꢂ1 cmꢂ1) at 340 nm for
3302
asc.wiley-vch.de
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2009, 351, 3287 – 3305

Directed Evolution of an Enantioselective Enoate-Reductase
FULL PAPERS
five min. The 500 mL standard assay mixture contained: References
100 mM NADPH, 20 mM glucose, 20 U glucose oxidase,
15 mM of the appropriate substrate (1M stock solution in
ethanol) in 100 mM Tris-Cl, pH 7.0 and the appropriate
amount of enzyme. One unit of enzyme activity was defined
as the quantity sufficient to oxidize 1 mmol of NADPH per
minute in the mixture described above. Apparent steady-
state kinetic parameters were determined by varying the
substrate concentration. The concentration of the enzymes
[1] Recent reviews of directed evolution of enantioselec-
tive enzymes: a) M. T. Reetz, in: Asymmetric Organic
Synthesis with Enzymes, (Eds.: V. Gotor, I. Alfonso, E.
Garcꢅa-Urdiales), Wiley-VCH, Weinheim, 2008, pp 21–
63; b) M. T. Reetz, J. Org. Chem. 2009, 74, 5767–5778.
[2] a) K. Drauz, H. Waldmann, Enzyme Catalysis in Or-
ganic Synthesis: A Comprehensive Handbook, Vols. I–
III, 2nd edn., Wiley-VCH, Weinheim, 2002; b) A.
Liese, K. Seelbach, C. Wandrey, Industrial Biotransfor-
mations, 2nd edn., Wiley-VCH, Weinheim, 2006.
[3] a) P. J. Walsh, M. C. Kozlowski, Fundamentals of Asym-
metric Catalysis, University Science Books, Sausalito,
CA, 2009; b) H.-U. Blaser, E. Schmidt, Asymmetric
Catalysis on Industrial Scales, Wiley-VCH, Weinheim,
2004.
[4] E. N. Jacobsen, A. Pfaltz, H. Yamamoto, Comprehen-
sive Asymmetric Catalysis, Vols. I–III, Springer, Berlin,
1999.
[5] a) G. Lelais, D. W. C. MacMillan, Aldrichimica Acta
2006, 39, 79–87; b) J. Seayad, B. List, Org. Biomol.
Chem. 2005, 3, 719–724; c) A. Berkessel, H. Grçger,
Asymmetric Organocatalysis, Wiley-VCH, Weinheim,
2004.
was estimated using
a molar extinction coefficient of
11,600Mꢂ1 cmꢂ1 at 455 nm. The spectrophotometric data
were fitted to Michaelis–Menten equation using the soft-
ware KaleidaGraph 4.0.
Scale-Up Bioreductions with YqjM Mutants
The cell pellet was prepared as described in the protein iso-
lation section, washed once with 30 mL of 100 mM Tris-Cl
pH 8.0 containing 100 mM (NH4)2SO4 and 5 mM EDTA
(Buffer B), resuspended in 20 mL of the same buffer and
kept at ꢂ808C for at least 1 hour. Cells were lysed by soni-
cation (Bandelin, 15ꢃ10 sec with 10 sec intervals, 40%
pulse, on water-ice bath). Cell debris was removed by centri-
fugation at 18,000ꢃg for 45 min at 108C. 9.0 g of solid am-
monium sulfate were added to the supernatant (20 mL) over
a period of 20 min with constant stirring at 48C. The solu-
tion was stirred for additional 45 min before centrifugation
at 18,000ꢃg for 45 min at 108C. The supernatant was dis-
carded and the precipitated proteins were redissolved in the
minimum amount of Buffer B (ca. 10 mL, final volume).
The solution was desalted as described before and was incu-
bated with excess of FMN (1 mg/mL) overnight at 48C. The
excess of free flavin was removed by passing the protein so-
lution twice through a PD-10 Desalting column equilibrated
with 100 mM potassium phosphate buffer, pH 7.0 containing
5 mM EDTA (Buffer C). The reconstituted sample was con-
centrated using an ultrafiltration centrifugal filter (10 kD
cut-off membrane, Amicon) to a final volume of ca. 7.5 mL
containing ca. 50 mg of flavoprotein. The flavoprotein con-
centration was estimated using a molar extinction coefficient
of 11,600Mꢂ1 cmꢂ1 at 455 nm.
The bioreduction mixture contained ca. 0.001 mmol re-
constituted flavoprotein, 200 U glucose dehydrogenase,
200 mM glucose, 2 mM NADP+ and the appropriate amount
of substrate (0.18 mmol of 1a–d or 0.97 mmol of 1e (150 mg)
dissolved in 100–500 mL of ethanol) in a final volume of
10 mL Buffer C. The reaction was terminated by adding 4 g
of ammonium sulfate and extraction of the mixture with 3ꢃ
12 mL MTBE. The combined organic layers were dried over
magnesium sulfate and the solvent was removed in a rotary
evaporator. Samples were analyzed by GC for determina-
tion of conversion and enantiomeric purity as described in
Table S4, Supporting Information.
[6] a) M. T. Reetz, A. Zonta, K. Schimossek, K. Liebeton,
K.-E. Jaeger, Angew. Chem. 1997, 109, 2961–2963;
Angew. Chem. Int. Ed. Engl. 1997, 36, 2830–2832;
b) M. T. Reetz, Pure Appl. Chem. 2000, 72, 1615–1622;
c) M. T. Reetz, Proc. Natl. Acad. Sci. USA 2004, 101,
5716–5722.
[7] Recent reviews of directed evolution:[1] a) S. Lutz, U. T.
Bornscheuer, Protein Engineering Protocols, Vols. I and
II, Wiley-VCH, Weinheim, 2009; b) K. M. Arndt, K. M.
Mꢀller, Protein Engineering Protocols (Methods in Mo-
lecular Biology), Vol. 352, Humana Press, Totowa, NJ,
2007; c) N. J. Turner, Nat. Chem. Biol. 2009, 5, 567–
573; d) C. Jꢆckel, P. Kast, D. Hilvert, Annu. Rev. Bio-
phys. Biomol. Struct. 2008, 37, 153–173; e) S. Bershtein,
D. S. Tawfik, Curr. Opin. Chem. Biol. 2008, 12, 151–
158; f) P. A. Romero, F. H. Arnold, Nat. Rev. Mol. Cell
Biol. 2009, 10, 866–876.
[8] See, for example: a) S. Lutz, W. M. Patrick, Curr. Opin.
Biotechnol. 2004, 15, 291–297; b) R. J. Fox, G. W. Huis-
man, Trends Biotechnol. 2008, 26, 132–138; c) A.
Herman, D. S. Tawfik, Protein Eng. Des. Sel. 2007, 20,
219–226; d) J. D. Bloom, M. M. Meyer, P. Meinhold,
C. R. Otey, D. MacMillan, F. H. Arnold, Curr. Opin.
Struct. Biol. 2005, 15, 447–452; e) T. S. Wong, D. Roc-
catano, M. Zacharias, U. Schwaneberg, J. Mol. Biol.
2006, 355, 858–871.
[9] a) M. T. Reetz, in: Evolutionary Methods in Biotech-
nology, (Eds.: S. Brakmann, A. Schwienhorst), Wiley-
VCH, Weinheim, 2004, pp 113–141; b) J.-L. Reymond,
Enzyme Assays – High-throughput Screening, Genetic
Selection and Fingerprinting, Wiley-VCH, Weinheim,
2006.
Acknowledgements
We thank Prof. Dr. P. Macheroux (Graz) for providing the
YqjM plasmid, P. Wedemann for the synthesis of not com-
mercially available compounds, N. J. A. Martin for a refer-
ence sample of (R)-4a, and J. Rosentreter for GC analyses.
We are also grateful to Dr. Dirk Opperman for his valuable
suggestions.
[10] M. T. Reetz, D. Kahakeaw, R. Lohmer, ChemBioChem
2008, 9, 1797–1804.
[11] M. T. Reetz, L.-W. Wang, in part M. Bocola, Angew.
Chem. 2006, 118, 1258–1263; Erratum 2556; Angew.
Chem. Int. Ed. 2006, 45, 1236–1241; Erratum 2494.
Adv. Synth. Catal. 2009, 351, 3287 – 3305
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
3303

Despina J. Bougioukou et al.
FULL PAPERS
[12] a) M. T. Reetz, D. Kahakeaw, J. Sanchis, Mol. BioSyst.
2009, 5, 115–122; b) M. T. Reetz, in: Protein Engineer-
ing Handbook, Vol. 2 (Eds.: S. Lutz, U. T. Born-
Scheuermann nꢇe Taylor, C. Jaekel, Adv. Synth. Catal.
2008, 350, 2708–2714; e) J. G. de Vries, C. J. Elsevier,
Handbook of Homogeneous Hydrogenation, Vols. I–III,
Wiley-VCH, Weinheim, 2006.
ACHTUNGTRENNUNGscheuer), Wiley-VCH, Weinheim, 2009, pp 409–439.
[13] a) M. T. Reetz, J. D. Carballeira, A. Vogel, Angew.
Chem. 2006, 118, 7909–7915; Angew. Chem. Int. Ed.
2006, 45, 7745–7751; b) M. T. Reetz, J. D. Carballeira,
Nat. Protoc. 2007, 2, 891–903.
[20] a) J. B. Tuttle, S. G. Ouellet, D. W. C. MacMillan, J.
Am. Chem. Soc. 2006, 128, 12662–12663; b) N. J. A.
Martin, B. List, J. Am. Chem. Soc. 2006, 128, 13368–
13369.
[14] a) M. T. Reetz, M. Bocola, J. D. Carballeira, D. Zha, A.
Vogel, Angew. Chem. 2005, 117, 4264–4268; Angew.
Chem. Int. Ed. 2005, 44, 4192–4196; b) C. M. Clouthier,
M. M. Kayser, M. T. Reetz, J. Org. Chem. 2006, 71,
8431–8437; c) J. D. Carballeira, P. Krumlinde, M.
Bocola, A. Vogel, M. T. Reetz, J.-E. Bꢆckvall, Chem.
Commun. 2007, 1913–1915.
[15] a) S. Bartsch, R. Kourist, U. T. Bornscheuer, Angew.
Chem. 2008, 120, 1531–1534; Angew. Chem. Int. Ed.
2008, 47, 1508–1511; b) L. Liang, J. Zhang, Z. Lin,
Microb. Cell Fact. 2007, 6, 36; c) R. Hawwa; S. D.
Larsen, K. Ratia, A. D. Mesecar, J. Mol. Biol. 2009,
393, 36–57; d) M. R. M. De Groeve, M. De Baere, L.
Hoflack, T. Desmet, E. J. Vandamme, W. Soetaert, Pro-
tein Eng. Des. Sel. 2009, 22, 393–399; e) K. Okrasa, C.
Levy, M. Wilding, M. Goodhall, N. Baudendistel, B.
Hauer, D. Leys, J. Micklefield, Angew. Chem. 2009,
121, 7827–7830; Angew. Chem. Int. Ed. 2009, 48, 7691–
7694; f) A. G. Sandstrçm, K. Engstrçm, J. Nyhlen, A.
Kasrayan, J.-E. Bꢆckvall, Protein Eng. Des. Sel. 2009,
22, 413–420.
[21] K. Kitzing, T. B. Fitzpatrick, C. Wilken, J. Sawa, G. P.
Bourenkov, P. Macheroux, T. Clausen, J. Biol. Chem.
2005, 280, 27904–27913.
[22] S. K. Padhi, D. J. Bougioukou, J. D. Stewart, J. Am.
Chem. Soc. 2009, 131, 3271–3280.
[23] a) B. J. Brown, Z. Deng, P. A. Karplus, V. Massey, J.
Biol. Chem. 1998, 273, 32753–32762; b) R. M. Kohli, V.
Massey, J. Biol. Chem. 1998, 273, 32763–32770; c) D.
Xu, R. M. Kohli, V. Massey, Proc. Natl. Acad. Sci. USA
1999, 96, 3556–3561.
[24] Figure were created by using PyMOL: W. L. DeLano,
The PyMOL Molecular Graphics System, 2008,
DeLano Scientific LLC, Palo, CA, USA http://www.
pymol.org.
[25] a) L. Rui, Y. M. Kwon, A. Fishman, K. F. Reardon,
T. K. Wood, Appl. Environ. Microbiol. 2004, 70, 3246–
3252; b) A. D. Bosley, M. Ostermeier, Biomol. Eng.
2005, 22, 57–61; c) M. A. Mena, P. S. Daugherty, Pro-
tein Eng. Des. Sel. 2005, 18, 559–561; d) M. Denault,
J. N. Pelletier, in: Protein Engineering Protocols, Vol.
352, (Eds.: K. M. Arndt, K. M. Mꢀller), Humana Press,
Totowa, NJ, 2007, pp 127–154.
[16] T. B. Fitzpatrick, N. Amrhein, P. Macheroux, J. Biol.
Chem. 2003, 278, 19891–19897.
[17] a) O. Warburg, W. Christian, Biochem. Z. 1933, 266,
377–411; b) Y. S. Niino, S. Chakraborty, B. J. Brown, V.
Massey, J. Biol. Chem. 1995, 270, 1983–1991.
[26] H. H. Hogrefe, J. Cline, G. L. Youngblood, R. M.
Allen, BioTechniques 2002, 33, 1158–1165.
[27] N refers to an equal mixture of nucleotides A, T, G,
and C; D refers to an equal mixture of A, T and G; K
refers to an equal mixture of G and T.
[18] a) R. E. Williams, N. C. Bruce, Microbiology 2002, 148,
1607–1614; b) R. Stuermer, B. Hauer, M. Hall, K.
Faber, Curr. Opin. Chem. Biol. 2007, 11, 203–213;
c) J. F. Chaparro-Riggers, T. A. Rogers, E. Vazquez-Fig-
ueroa, K. M. Polizzi, A. S. Bommarius, Adv. Synth.
Catal. 2007, 349, 1521–1531; d) M. Hall, C. Stueckler,
W. Kroutil, P. Macheroux, K. Faber, Angew. Chem.
2007, 119, 4008–4011; Angew. Chem. Int. Ed. 2007, 46,
3934–3937; e) A. Muller, B. Hauer, B. Rosche, Biotech-
nol. Bioeng. 2007, 98, 22–29; f) A. Mꢀller, R. Sturmer,
B. Hauer, B. Rosche, Angew. Chem. 2007, 119, 3380–
3382; Angew. Chem. Int. Ed. 2007, 46, 3316–3318;
g) A. Fryszkowska, K. Fisher; J. M. Gardiner, G. M.
Stephens, J. Org. Chem. 2008, 73, 4295–4298; h) M.
Hall, C. Stueckler, H. Ehammer, E. Pointner, G. Ober-
dorfer, K. Gruber, B. Hauer, R. Stuermer, W. Kroutil,
P. Macheroux, K. Faber, Adv. Synth. Catal. 2008, 350,
411–418; i) B. Kosjek, F. J. Fleitz, P. G. Dormer; J. T.
Kuethe; P. N. Devine, Tetrahedron: Asymmetry 2008,
19, 1403–1406; j) H. S. Toogood, A. Fryszkowska, V.
Hare; K. Fisher, A. Roujeinikova, D. Leys, J. M.
Gardiner, G. M. Stephens, N. S. Scrutton, Adv. Synth.
Catal. 2008, 350, 2789–2803.
[28] J. Sanchis, L. Fernꢈndez, J. D. Carballeira, J. Drone, Y.
Gumulya, H. Hçbenreich, D. Kahakeaw, S. Kille, R.
Lohmer, J. J.-P. Peyralans, J. Podtetenieff, S. Prasad, P.
Soni, A. Taglieber, S. Wu, F. E. Zilly, M. T. Reetz,
Appl. Microbiol. Biotechnol. 2008, 81, 387–397.
[29] M. Z. Li, S. J. Elledge, Nat. Methods 2007, 4, 251–256.
[30] A. Airaksinen, T. Hovi, Nucleic Acids Res. 1998, 26,
576–581.
[31] a) K. M. Polizzi, M. Parikh, C. U. Spencer, I. Matsu-
mura, J. H. Lee, M. J. Realff, A. S. Bommarius, Biotech-
nol. 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. Screening 2005, 10,
856–864.
[32] M. R. Martzen, S. M. McCraith, S. L. Spinelli, F. M.
Torres, S. Fields, E. J. Grayhack, E. M. Phizicky, Science
1999, 286, 1153–1155.
[33] a) K. Shimoda, N. Kubota; H. Hamada, Tetrahedron:
Asymmetry 2004, 15, 2443–2446; b) M. A. Swiderska;
J. D. Stewart, J. Mol. Catal. B: Enzym. 2006, 42, 52–54.
[34] T. B. Fitzpatrick, S. Auweter, K. Kitzing; T. Clausen, N.
Amrhein, P. Macheroux, Protein Expression Purif.
2004, 36, 280–291.
[19] a) S.-M. Lu, C. Bolm, Chem. Eur. J. 2008, 14, 7513–
7516; b) T. Ohshima, H. Tadaoka, K. Hori, N. Sayo, K.
Mashima, Chem. Eur. J. 2008, 14, 2060–2066; c) S.-M.
Lu, C. Bolm, Angew. Chem. 2008, 120, 9052–9055;
Angew. Chem. Int. Ed. 2008, 47, 8920–8923; d) C. J.
[35] A. S. Abramovitz; V. Massey, J. Biol. Chem. 1976, 251,
5321–5326.
3304
asc.wiley-vch.de
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2009, 351, 3287 – 3305

Directed Evolution of an Enantioselective Enoate-Reductase
FULL PAPERS
[36] a) S. Luetz, L. Giver, J. Lalonde, Biotechnol. Bioeng.
2008, 101, 647–653; b) R. J. Fox, M. D. Clay, Trends
Biotechnol. 2009, 27, 137–140.
[37] a) S. G. Peisajovich, D. S. Tawfik, Nat. Methods 2007, 4,
991–994; b) M. A. DePristo, HFSP Journal 2007, 1,
94–98; c) J. D. Bloom, F. H. Arnold, Proc. Natl. Acad.
Sci. USA 2009, 106, 9995–10000.
[38] M. Eigen, J. McCaskill, P. Schuster, J. Phys. Chem.
1988, 92, 6881–6891.
[39] S. Kurtovic, B. Mannervik, Biochemistry 2009, 48,
9330–9339.
Adv. Synth. Catal. 2009, 351, 3287 – 3305
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
3305