Systematic investigation of Saccharomyces cerevisiae enzymes catalyzing carbonyl reductions

Article abstract of DOI:10.1021/ja0469479

Eighteen key reductases from baker's yeast (Saccharomyces cerevisiae) have been overproduced in Escherichia coli as glutathione S-transferase fusion proteins. A representative set of α- and β-keto esters was tested as substrates (11 total) for each purified fusion protein. The stereoselectivities of β-keto ester reductions depended both on the identity of the enzyme and the substrate structure, and some reductases yielded both L- and D-alcohols with high stereoselectivities. While α-keto esters were generally reduced with lower enantioselectivities, it was possible in all but one case to identify pairs of yeast reductases that delivered both alcohol antipodes in optically pure form. Taken together, the results demonstrate not only that individual yeast reductases can be used to supply important chiral building blocks, but that GST-fusion proteins allow rapid identification of synthetically useful biocatalysts (along with their corresponding genes).

Full text of DOI:10.1021/ja0469479

Published on Web 09/15/2004
Systematic Investigation of Saccharomyces cerevisiae
Enzymes Catalyzing Carbonyl Reductions
†
Iwona A. Kaluzna, Tomoko Matsuda, Aileen K. Sewell, and Jon D. Stewart*
Contribution from the Department of Chemistry, UniVersity of Florida,
GainesVille, Florida 32611
Received May 24, 2004; E-mail: jds2@chem.ufl.edu
Abstract: Eighteen key reductases from baker’s yeast (Saccharomyces cerevisiae) have been overproduced
in Escherichia coli as glutathione S-transferase fusion proteins. A representative set of R- and â-keto esters
was tested as substrates (11 total) for each purified fusion protein. The stereoselectivities of â-keto ester
reductions depended both on the identity of the enzyme and the substrate structure, and some reductases
yielded both L- and D-alcohols with high stereoselectivities. While R-keto esters were generally reduced
with lower enantioselectivities, it was possible in all but one case to identify pairs of yeast reductases that
delivered both alcohol antipodes in optically pure form. Taken together, the results demonstrate not only
that individual yeast reductases can be used to supply important chiral building blocks, but that GST-
fusion proteins allow rapid identification of synthetically useful biocatalysts (along with their corresponding
genes).
selectivities operate simultaneously.5 Brute-force screening
-7
Introduction
of additional microbial strains in hopes of identifying those
Enantioselective ketone reductions are one of the most
popular strategies for producing homochiral alcohol building
blocks for organic synthesis. The ideal reagent or catalyst for
lacking competing reductases is the most common solution to
8
this problem. Unfortunately, there are three problems with this
1
approach. First, it is time-consuming, and there is no simple
way to direct the search rationally. In addition, a diverse range
of microorganisms results from this type of screening, and each
has specific nutritional and growth requirements that must be
worked out individually for each case. This increases process
development times. Finally, even after a suitable strain has been
identified, obtaining the gene encoding the key enzyme in order
to produce it at higher levels in a more-easily handled host is
often a time-consuming procedure, although recent progress in
this area has been reported by Kim and co-workers.9
this conversion would offer broad substrate acceptance, predict-
able access to all product enantiomers and/or diastereomers,
rapid reaction rates, low cost, and minimal environmental
burden. While no existing methodology meets all of these
criteria, the properties of enzymes make them strong contenders,
and this has prompted intensive efforts to expand and improve
the range and selectivities of biocatalytic carbonyl reductions.
Such studies have often shown that identifying the most
2
appropriate biocatalyst is a critical bottleneck. High-throughput
methods for evaluating “chemical” reagents and catalysts have
Here, we describe an alternative to whole-cell microbial
screening that rapidly uncovers the most valuable enzyme(s)
and directly provides the corresponding gene(s). The method
takes advantage of the wealth of data that has accrued from
sequencing whole genomes. Martzen et al. pioneered the creation
of genome-wide expression libraries in which every potential
open reading frame is expressed as a fusion protein with
glutathione S-transferase (GST), allowing one to isolate every
catalyst produced by an organism by a common, one-step
been developed;3,4 however, this task is usually much slower
in biocatalysis because of the time required to grow microbial
cells before they can be tested for their ability to carry out the
desired conversion. Moreover, while “chemical” processes are
typically carried out in the presence of only a single reduction
catalyst or reagent, whole microbial cells contain a wealth of
enzymes. When one or a few reductases with identical stereo-
selectivities dominate within a cell, single products can be
obtained from these reactions. Often, however, mixtures of
stereoisomeric alcohols are formed because multiple enzymes
with overlapping substrate specificities but conflicting stereo-
1
0
affinity purification. The catalytic activity of every fusion
protein can then be assessed individually without interference
(5) Shieh, W.-R.; Gopalin, A. S.; Sih, C. J. J. Am. Chem. Soc. 1985, 107,
2
993-2994.
(
6) Shieh, W.-R.; Sih, C. J. Tetrahedron: Asymmetry 1993, 4, 1259-1269.
†
Current address: Ryukoku University, Otsu, Shiga, Japan.
(7) Nakamura, K.; Kawai, Y.; Nakajima, N.; Ohno, A. J. Org. Chem. 1991,
56, 4778-4783.
(
1) Wills, M.; Hannedouche, J. Curr. Opin. Drug Discuss. DeVel. 2002, 5,
8
81-891.
(8) Nakamura, K.; Yamanaka, R.; Matsuda, T.; Harada, T. Tetrahedron:
Asymmetry 2003, 14, 2659-2681.
(
2) Demirjian, D. C.; Shah, P. C.; Mor ´ı s-Varas, F. Top. Curr. Chem. 1999,
2
00, 1-29.
(9) Kim, J.-Y.; Choi, G.-S.; Jung, I.-S.; Ryu, Y.-W.; Kim, G.-J. Protein Eng.
2003, 16, 357-364.
(
(
3) Senkan, S. M. Nature 1998, 394, 350-353.
4) Maxwell, I. E.; van den Brink, P.; Downing, R. S.; Sijpkes, A. H.; Gomez,
S.; Maschmeyer, T. Top. Catal. 2003, 24, 125-135.
(10) Martzen, M. R.; McCraith, S. M.; Spinelli, S. L.; Torres, F. M.; Fields, S.;
Grayhack, E. J.; Phizicky, E. M. Science 1999, 286, 1153-1155.
10.1021/ja0469479 CCC: $27.50 © 2004 American Chemical Society
J. AM. CHEM. SOC. 2004, 126, 12827-12832
9
12827

A R T I C L E S
Kaluzna et al.
from competing enzymes. Since each fusion protein is produced
from the corresponding cloned gene, relationships between
genotype and phenotype are established directly. In this report,
we describe the application of the methodology to the systematic
study of reductases produced by baker’s yeast (Saccharomyces
Scheme 1
11
cereVisiae).
Since the seminal studies by Neuberg,12 baker’s yeast cells
have been the most popular biocatalyst for asymmetric ketone
reductions (for recent reviews, see 13-15). The substrate
acceptance of this organism is very broad, and using com-
mercially available whole cells is both experimentally simple
and inexpensive. Unfortunately, such reductions are often
bedeviled by poor stereoselectivities, which can be traced to
the presence of multiple enzymes with divergent enantio- and
diastereomeric preferences. Notable efforts to overcome this
problem have included adjusting the substrate concentration and
16
structure (to favor acceptance by a subset of yeast reductases),
including additives to poison one or more competing en-
7,17,18
19,20
zymes,
and genetic knockout/overexpression approaches.
In practice, these strategies have often proven only partially
successful. The reason for these difficulties became clear from
our analysis of the complete yeast genome, which revealed that
the protein products of ca. fifty open reading frames might
catalyze ketone reductions.21 Clearly, it would be very difficult
to devise a set of selective inhibitors for such a large collection
of reductases. Likewise, our earlier strategy based on creating
targeted gene knockouts in hopes of improving the stereose-
lectivities of yeast reductions was also not viable, given the
number of yeast genes that would have to be inactivated
simultaneously.20 By contrast, screening individual, cloned
fusion proteins not only allows their substrate- and stereose-
lectivities to be determined cleanly, but also provides a rapid
entry into heterologous overexpression systems (such as engi-
neered Escherichia coli cells) that often offer both high
volumetric productivities and freedom from competing reduc-
of these candidates were eliminated, either because they were
known or suspected to have narrow substrate specificities (zinc-
containing yeast alcohol dehydrogenases, for example) or their
sequence similarities to bona fide reductases was low. The
winnowing process left 22 open reading frames that included
23
2
4
25
members of the aldose reductase, short-chain, medium-
2
6
chain, and D-hydroxyacid dehydrogenase superfamilies. We
considered these the most promising candidates for discovering
synthetically useful biocatalysts.
The original method for preparing the yeast GST-fusion
1
0
proteins relied on overexpression in baker’s yeast. While this
approach most closely mimics the native environment for these
proteins, the choice of host cell limited protein overexpression
to modest levels. Preliminary studies showed that the level of
GST-fusion protein production in yeast cells was not sufficient
to allow their activities toward nonnatural substrates to be
assessed reliably. We therefore used standard molecular biology
techniques to create E. coli overexpression plasmids for each
of the twenty-two GST fusion proteins of interest for this study.
These constructs retained the identical protein coding regions
but utilized the strong T7 promoter for bacterial overexpression.
The fusion proteins were purified by affinity chromatography
on glutathione-agarose and they appeared to retain their catalytic
activities for at least a year when stored at -20 °C in 50%
2
2
tases.
Results and Discussion
This study explored the properties of twenty two S. cereVisiae
reductases. This group was selected from our earlier analysis
of the yeast genome, which revealed that 49 open reading frames
encoded known or putative reductases.21 Approximately half
(
11) Kaluzna, I.; Andrew, A. A.; Bonilla, M.; Martzen, M. R.; Stewart, J. D. J.
Mol. Catal. B: Enzymatic 2002, 17, 101-105.
(
12) Neuberg, C.; F., N. F. Biochem. Z. 1914, 62, 482-488.
13) Sybesma, W. F. H.; Straathof, A. J. J.; Jongejan, J. A.; Pronk, J. T.; Heignen,
J. J. Biocatal. Biotrans. 1998, 16, 95-134.
27
glycerol. This allowed us to keep the library of yeast reductases
(
on hand for rapid screening.
(
14) Stewart, J. D. Curr. Opin. Drug DiscoVery DeVelopment 1998, 1, 278-
A representative panel of â- and R-keto esters was selected
to profile the substrate- and enantioselectivities of the isolated
yeast proteins (Scheme 1). These compounds were selected so
that data from homologous structures might reveal useful trends
that could be applied to guide future synthetic applications. In
addition, several of the alcohol products from these reductions
have been used as chiral building blocks. For example, Corr eˆ a
and co-workers have employed (R)-3d in their synthesis of (S)-
2
89.
(
15) Santaniello, E.; Ferraboschi, P.; Manzocchi, A. In Enzymes in Action. Green
Solutions for Chemical Problems; Zwanenburg, B., Mikolajczyk, M. and
Kielbasinski, P., Eds.; Kluwer Academic Publishers: Dordrecht, 2000, pp
9
5-115.
(
16) Sih, C. J.; Zhou, B.-N.; Gopalin, A. S.; Shieh, W.-R.; VanMiddlesworth,
F. Sel., Goal. Synth. Effic., Proc. Workshop Conf. Hoechst. 14th 1983, 184,
2
51-261.
(
(
(
(
(
17) Nakamura, K.; Inoue, K.; Oshio, K.; Oka, S.; Ohno, A. Chem. Lett. 1987,
79-682.
18) Nakamura, K.; Kawai, Y.; Oka, S.; Ohno, A. Bull. Chem. Soc. Jpn. 1989,
2, 875-879.
19) Rodriguez, S.; Kayser, M. M.; Stewart, J. D. Org. Lett. 1999, 1, 1153-
155.
20) Rodriguez, S.; Kayser, M. M.; Stewart, J. D. J. Am. Chem. Soc. 2001,
6
6
1
(23) Leskovac, V.; Trivic, S.; Pericin, D. FEMS Yeast Res. 2002, 2, 481-494.
(24) Jez, J. M.; Penning, T. M. Chemico-Biol. Interactions 2001, 130-132, 499-
525.
(25) Krozowski, Z. J. Steroid Biochem. Mol. Biol. 1994, 51, 125-130.
(26) Persson, B.; Zigler, J. S.; J o¨ rnvall, H. Eur. J. Biochem. 1994, 226, 15-22.
(27) Four of the overexpression strains yielded the fusion proteins as insoluble
inclusion bodies, and these were eliminated from further study (YOR388c,
YMR318c, YCR105w, and YJR159w).
1
23, 1547-1555.
21) Stewart, J. D.; Rodriguez, S.; Kayser, M. M. In Enzyme Technologies for
Pharmaceutical and Biotechnological Applications; Kirst, H. A., Yeh, W.-
K. and Zmijewski, M. J., Eds.; Marcel Dekker: New York, 2001, pp 175-
2
08.
(22) Walton, A. Z.; Stewart, J. D. Biotechnol. Prog. 2004, 20, 403-411.
1
2828 J. AM. CHEM. SOC. VOL. 126, NO. 40, 2004
9

Investigation of S. cerevisiae reductases
A R T I C L E S
Figure 1. Correlations between the absolute configurations of reduction products and colors for â- and R-keto esters depicted in Tables 1 - 3.
Table 1. Biocatalytic Reductions of Representative â-keto Esters.
2
-methyl-4-octanol, an aggregation pheromone of Curculionidae
Yeast Enzymes Are Referred to by Their Genetic Codes and
Grouped by Superfamily (aldose reductases, D-hydroxyacid
dehydrogenases, medium-chain dehydrogenases and short-chain
dehydrogenases). Product Compositions from Reactions that
Proceeded to g 20% Conversion within 24 h Are Shown in Pie
Charts (legend in Figure 1)
28
29
species. L-alcohol 2e has been used to synthesize L-carnitine.
The (R)-alcohol derived from R-keto ester 4c is a key chiral
building block in a variety homophenylalanine-containing
pharmaceuticals (see 11 and references therein).
Conversions were carried out at pH 7.0 with initial substrate
+
concentrations fixed at 5 mM. A catalytic amount of NADP
was included along with 1.3 equivalents of glucose-6-phosphate
and glucose-6-phosphate dehydrogenase.30 Reactions were
sampled periodically and analyzed by GC to determine both
the extent and stereoselectivity of product formation. All of the
alcohol products from these reactions are known compounds
whose absolute configurations have been established. We
therefore used authentic standards to match structures with
chiral-phase GC elution patterns. In some cases, these standards
were available from our prior studies;2
0,31
the remaining ones
were obtained by isolating alcohols from reductions that afforded
single products and comparing their spectral data and optical
rotations with literature values. Data for enzyme/â-keto ester
pairs where only enantiomeric products are possible are shown
in Table 1 (substrates 1a-e). The reductases are grouped by
superfamilies, and data for reactions carried out under standard
conditions with commercial yeast cells are also included for
comparison.1
6,20,28
From these data, it is clear that all 18 of the
enzymes examined in this study are indeed reductases, accepting
at least one of the substrates examined. In addition, several
trends can be discerned. First, there is a definite correlation
between substrate size and the number of enzymes that accept
simple â-keto esters: while 15 of the 18 enzymes tested
catalyzed the reduction of ethyl acetoacetate 1a; this fraction
dropped to 9 of 18 for 1b and 1c and only 3 of 18 for 1d. In
addition, the stereoselectivities of individual reductases were
not determined solely by the enzyme, but also depended
critically on substrate structure. For example, straight-chain
â-keto esters 1a-c were all reduced primarily (often exclu-
sively) to l-alcohols; however, branched homologue 1d gave
only the D-product from the three enzymes that accepted this
a
20% conversion after 24 h. b Reference 20. c Reference 16. d Refer-
<
ence 28.
substrate, even though the same enzymes were completely
L-selective for 1a-c. These results make it clear that the former
practice of referring to yeast reductases as “L-“ or “D-specific”
enzymes must be discontinued.
The enzymatic reduction of ethyl 4-chloroacetoacetate 1e
provided several interesting results. The most striking feature
is that the outcomes of these reactions bore a complex
relationship to those of homologues 1a-c for all reductase
superfamilies except short-chain dehydrogenases. In this latter
case, reactions of 1e afforded virtually the same stereoselec-
tivities as 1b and 1c, which make similar steric demands,
suggesting that substrate binding by these enzymes does not
involve special interactions with the chlorine atom. Reductases
(
(
(
(
28) Baraldi, P. T.; Zarbin, P. H. G.; Vieira, P. C.; Correa, A. G. Tetrahedron:
Asymmetry 2002, 13, 621-624.
29) Zhou, B.-N.; Gopalin, A. S.; VanMiddlesworth, F.; Shieh, W.-R.; Sih, C.
J. J. Am. Chem. Soc. 1983, 105, 5925-5926.
30) Preliminary studies showed that all of the reductases studied here were
active with NADPH. Parallel reactions using NADH were therefore omitted.
31) Rodriguez, S.; Schroeder, K. T.; Kayser, M. M.; Stewart, J. D. J. Org.
Chem. 2000, 65, 2586-2587.
J. AM. CHEM. SOC.
9
VOL. 126, NO. 40, 2004 12829

A R T I C L E S
Kaluzna et al.
Table 3. Biocatalytic Reductions of R-substituted â-keto Esters.
Yeast Enzymes Are Referred to by Their Genetic Codes and
Grouped by Superfamily (aldose reductases, D-hydroxyacid
dehydrogenases, medium-chain dehydrogenases and short-chain
dehydrogenases). Product Compositions from Reactions that
Proceeded to g 20% Conversion within 24 h Are Shown in Pie
Charts (legend in Figure 1)
Table 2. Biocatalytic Reductions of Representative R-keto Esters.
Yeast Enzymes Are Referred to by Their Genetic Codes and
Grouped by Superfamily (aldose reductases, D-hydroxyacid
dehydrogenases, medium-chain dehydrogenases, and short-chain
dehydrogenases). Product Compositions from Reactions that
Proceeded to g20% Conversion within 24 h Are Shown in Pie
Charts (legend in Figure 1)
a
20% conversion after 24 h. b Reference 20.
<
a
20% conversion after 24 h. b Reference 32. c Reference 33.
<
values for these reductions. On the other hand, the alcohol
products were stable to the reaction conditions, so that stereo-
selectivities could be determined with confidence.
from the other three superfamilies do not follow the same
pattern, however. Instead, the outcomes of these reactions fall
into one of two categories. For those enzymes that accept 1b
and 1c (YOR120w, YDR368w, YPL113c and YAL060w), 1e
is reduced in a similar manner. By contrast, reductases that do
not accept 1b and 1c (YJR096w, YDL124w, YBR149w,
YHR104w, YGL185c, YNL274c, and YLR070c) convert 1e to
primarily (or exclusively) the D-alcohol. These differing out-
comes must signal differences in substrate binding orientation
and underscore the difficulties in extrapolating stereoselectivity
patterns to novel substrates. Indeed, it was for this reason that
we selected the GST-fusion protein methodology so that
empirical screening could be carried out rapidly in preference
to developing active site models for each reductase.
In contrast to the complex results obtained from simple â-keto
esters, reductions of R-keto esters generally followed similar
patterns across the homologous substrate series, although there
were a few exceptions. For example, aldose reductase YDR368w
reduced 4a to the corresponding L-alcohol with >98% ee;
however, it displayed opposite selectivity for 4b and 4c.
D-hydroxyacid dehydrogenase YPL275w reduced only 4a and
did not accept the larger homologues . Medium-chain dehy-
drogenase YLR070c reduced only the largest substrate tested,
4c, but not the smaller analogues. In general, R-keto esters were
reduced with lower stereoselectivities than the corresponding
â-keto esters, although it was still possible in most cases to
identify pairs of enantiocomplementary enzymes that afforded
both alcohol antipodes.
Data from reductions of representative R-keto esters using
yeast GST-fusion proteins are summarized in Table 2. As in
the previous examples, the enzymes are grouped by superfamily
and the substrates are arranged in order of increasing steric bulk.
Data from reductions using whole S. cereVisiae cells are
Data for representative R-substituted â-keto ester substrates
that can afford diastereomeric alcohols (1f-h) are depicted in
Table 3. The high acidity of the R-protons ensures that all of
these conversions are dynamic kinetic resolutions. Interestingly,
all of these reductions resulted in exclusive formation of
L-alcohols (except for ca.. 5% D-product formed by YAL060w
from 1g). This behavior is consistent with results obtained from
1a (Table 1). Three classes of behavior are evident from these
results. Certain reductasesYNL274c and YPL275wsare re-
stricted to converting only the smallest substrate (1f). The
included for comparison.3
2,33
32
In line with earlier studies, we
found that all three R-keto esters decomposed spontaneously at
neutral pH, making it difficult to assess fractional conversion
(
32) Nakamura, K.; Kondo, S.-i.; Kawai, Y.; Ohno, A. Bull. Chem. Soc. Jpn.
1
993, 66, 2738-2743.
(33) Dao, D. H.; Okamura, M.; Akasaka, T.; Kawai, Y.; Hida, K.; Ohno, A.
Tetrahedron: Asymmetry 1998, 9, 2725-2737.
12830 J. AM. CHEM. SOC.
9
VOL. 126, NO. 40, 2004

Investigation of S. cerevisiae reductases
A R T I C L E S
remaining enzymes accept a larger variety of substrates, although
they show divergent behavior with respect to stereoselectivities.
For some, the preferences remain the same throughout the series
of compounds. On the other hand, for five reductases, the major
product from the smallest substrate is syn-2f while anti-2h
dominates the reduction of 1h. As in the previous cases, the
varied results demonstrate that directly determining reaction
outcomes will likely be preferable to computer modeling
strategies, at least soon.
catalysts from a diverse range of organisms that may be difficult
to employ directly in bioconversions, along with the ability to
discover the most appropriate biocatalyst(s) rapidly, are par-
ticularly powerful features of our fusion protein library strategy.
The other key advantage of our method is that once a suitable
reductase catalyst has been identified by screening GST-fusion
proteins, preparative-scale reactions can be carried out directly
by whole cells of the same engineered E. coli strain. This
provides a common platform for process development. More-
over, since each E. coli strain expresses only a single yeast
reductase, competition by other enzymes is minimized. In this
way, the benefits of rapid GST-fusion protein screening can be
quickly realized in organic synthesis. In one example, we have
produced optically pure (S)-ethyl-3-hydroxybutyrate with a
space-time yield of 2.0 g/L‚h and a final product titer of 16 g/L
using an E. coli strain that overexpressed the yeast YOL151w
Conclusions
Our data demonstrate clearly that the lack of stereoselectivity
commonly observed for reductions by whole S. cereVisiae cells
is not due to inferior properties of the individual enzymes. This
had been suspected, but heretofore, there had been little direct
experimental data to support this contention. Many reductions
with single yeast enzymes proceed with g90% ee and g90%
de. The problem is that a single cell contains multiple catalysts
with conflicting stereopreferences and the net result is a mixture
of products. Our genetic strategy effectively “purifies” the
individual reductases so that their intrinsic properties can be
discovered without interference by other enzymes. This approach
also allows the properties of yeast reductases that may be poorly
expressed in native cells to be assessed. It is also rapid: we
routinely carry out all of the screening reactions for an individual
substrate in a total of 48 h. These are significant advantages
2
2
protein. We expect that other reactions discovered here will
be equally amenable to scale-up.
Experimental Section
General. Recombinant DNA procedures were carried out
3
5
using standard procedures. Enzymes required for molecular
biology were purchased from commercial suppliers and used
as recommended. Standard media and techniques for growth
and maintenance of E. coli were used and LB medium contained
1
% Bacto-Tryptone, 0.5% Bacto-Yeast Extract and 1% NaCl.
Glucose-6-phosphate dehydrogenase (Sigma type XV from
baker’s yeast) was used for cofactor regeneration.
2
0,34
over earlier methods based on whole S. cereVisiae cells.
Does the current set of 19 S. cereVisiae reductases encompass
all of the enzymes that contribute to carbonyl reductions by
the native organisms? The answer is certainly no. For example,
the major product from the yeast-mediated reduction of â-keto
ester 1d is L-alcohol 2d; however, none of the reductases
surveyed in our study afforded even a trace of this enantiomer.
Likewise, while D-alcohol 2c is essentially the sole product when
whole yeast cells reduce 1c, none of the reductases examined
here gave predominantly this enantiomer. Discovering the
enzyme(s) with these stereoselectivities will require augmenting
the current collection of GST-fusion proteins with additional
yeast open reading frames.
Ketones 1a-c, e-h, and 4c were obtained from commercial
suppliers and used as received. â-Keto ester 1d was prepared
2
8
by a literature method and R-keto esters 4a and 4b were
prepared by Fischer esterification of the commercially available
acids. NMR spectra were obtained from instruments operating
1
13
at 300 or 500 MHz for H and 75 MHz for C, respectively,
and peaks were referenced to residual protonated solvent. IR
spectra were recorded from thin films. Mass spectra (EI, 70
eV) were obtained from a benchtop GC/MS system equipped
with a 0.32 mm × 30 m DB-17 column. Optical rotations were
measured from CHCl3 solutions using a Perkin-Elmer 241 or
3
41 polarimeter operating at room temperature. Analytical GC
For synthetic purposes, the ability to prepare both enantiomers
of alcohol building blocks in homochiral form is essential. We
therefore designed our strategy to discover the correct enzyme
pairs for each substrate rapidly, and we hoped that basing our
reductase library on the S. cereVisiae genome would provide
sufficient genetic diversity to solve a range of synthetic
problems. For R-keto esters 4a-c, this goal has been largely
realized (Table 2) and only L-6b is not available in optically
pure form in this series. The situation is more complex for â-keto
esters. On one hand, both L- and D-alcohols derived from 1e
can be produced in homochiral form (Table 1). By constrast,
L-alcohols derived from 1a-c and the D-alcohol from 1d are
available in >98% ee, but the antipodes are not. Only one of
the four possible diastereomers derived from â-keto esters 1f-h
can be produced in optically pure form (syn-2f-h). Overcoming
these problems will require adding additional reductases to the
collection. For these purposes, the additional enzymes could
be derived from S. cereVisiae or from any other organism whose
genome has been sequenced. This ability to access potential
analyses were carried out with a 0.32 mm × 30 m DB-17
column for nonchiral separations and a 0.25 mm × 25 m
Chirasil-Dex CB or a 0.25 × 25 m Chirasil-L-Val column for
enantiomer separations. Samples for GC analysis of biotrans-
formation reactions were prepared by mixing 200 µL of the
reaction mixture with an equal volume of Et2O. After vortex
mixing, the organic layer was removed for analysis. Racemic
alcohols were prepared from ketones 1a-h and 4a-c by
reduction with NaBH4 and conditions that gave baseline
resolution of all products were used for analyzing products from
enzymatic reductions. Alcohols derived from â-keto ester 1 g
required acetylation prior to GC analysis for resolution of the
optical isomers, which was carried out with excess Ac2O in
DMF catalyzed by DMAP overnight at room temperature. After
derivatization, the reaction mixture was extracted with Et2O (2
×
1 mL), then the combined organic layers were concentrated
and analyzed by GC.
Creation of Overexpression Strains for Yeast GST-Fusion
Proteins. An E. coli plasmid (pIK2) was constructed to allow
(
34) Katz, M.; Hahn-H a¨ gerdal, B.; Gorwa-Grauslund, M. F. Enzyme Microb.
(35) Sambrook, J.; Fritsch, E. F.; Maniatis, T. Molecular Cloning. A Laboratory
Manual, 2nd ed.; Cold Spring Harbor: Cold Spring Harbor, 1989.
Technol. 2003, 33, 163-172.
J. AM. CHEM. SOC.
9
VOL. 126, NO. 40, 2004 12831

A R T I C L E S
Kaluzna et al.
overexpression of yeast reductase genes flanked by common
restriction sites under control of a T7 promoter. The GST coding
region from pYEX 4T-1 (Clontech) was PCR-amplified using
a pair of primers that incorporated AseI and NcoI restriction
sites at the 5′- and 3′-ends of the gene, respectively (Forward,
supernatant was mixed with an equal volume of 50 mM Tris-
Cl, 4 mM MgCl2, 1 mM DTT, 10% glycerol, pH 7.5 and loaded
onto a 2.4 × 5.0 cm column of glutathione resin (Clontech) at
a flow rate of 0.5 mL/min. at 4 °C that had been equilibrated
with 50 mM Tris-Cl, 4 mM MgCl2, 1 mM DTT, 500 mM NaCl,
10% glycerol, pH 7.5 (wash buffer). The flow-through was
discarded and the resin was washed twice with 20 mL wash
buffer. Material eluted by wash buffer was also discarded.
Essentially pure GST-fusion proteins were eluted with 40 mL
of freshly prepared elution buffer (wash buffer (39.6 mL) plus
2 M NaOH (0.40 mL) and solid glutathione (0.31 g)). The eluant
was dialyzed against 20 mM Tris-Cl, 2 mM EDTA, 4 mM
MgCl2, 1 mM DTT, 55 mM NaCl, 50% glycerol, pH 7.5 prior
to storage at -20 °C. The glutathione agarose resin was
regenerated by washing with 20 column volumes of phosphate-
buffered saline supplemented with 3 M NaCl (10 mM Na2HPO4,
1.8 mM KH2PO4, 2.7 mM KCl, 3.14 M NaCl, pH 7.4) followed
by 10 column volumes of wash buffer.
5′-ATT AAT GAC CAA GTT ACC TAT ACT AGG TTA T-3′
and Reverse, 5′-CCA TGG GCA TAT GAC GCG GAA CCA
GAT GAT CCG ATT-3′; restriction sites underlined). The PCR
product was inserted into a commercially available plasmid
(pCR 2.1-Topo, Invitrogen), then the GST coding region was
excised by digestion with AseI and NcoI and ligated with
pET26b (Novagen) that had been cut with NdeI and NcoI to
yield pIK2. Note that joining the compatible overhangs of AseI
and NdeI eliminated the NdeI site that otherwise would have
been present at the 5′-end of the GST coding region. This
allowed pIK2 to retain only a single NdeI site in addition to
the unique NcoI site at the 3′-end of the GST coding region.
Genes encoding known and putative reductases were sub-
cloned into expression plasmid pIK2 using standard methods.
General Procedure for Ketone Reductions Using Yeast
20,31
+
For reductase genes available from our earlier studies,
DNA
GST-Fusion Proteins. Reaction mixtures contained NADP
fragments were excised directly from existing plasmids and
incorporated into pIK2 using compatible restriction sites (NdeI
at the 5′- end and BamHI, EcoRI, HindIII, SacI, or SalI at the
(0.20 µmoles, 0.15 mg), glucose-6-phosphate (14 µmoles, 4.3
mg), glucose-6-phosphate dehydrogenase (5 µg), ketone sub-
strate (5 mM) and purified GST-fusion protein (10-100 µL,
containing 5-50 µg) in 1.0 mL of 100 mM KPi, pH 7.0.
Reaction mixtures were incubated at 30 °C and sampled for
GC analysis periodically.
When product isolation was required in order to determine
absolute configurations of alcohols, the bioconversions described
above were scaled up 10- or 20-fold. After nearly all of the
substrate had been consumed, the reaction mixture was extracted
with Et2O (3 × (5 × reaction volume)). The combined organic
extracts were washed with brine (1 volume) and water (1
volume), then dried with MgSO4 and concentrated in vacuo. If
required, the alcohol product was purified by flash column
chromatography prior to spectral analysis.
3
′-end). The remaining reductase genes were PCR-amplified
from S. cereVisiae genomic DNA using appropriate primer pairs
and the amplification products were inserted into pCR2.1-Topo
or PCR TopoII-Blunt vectors. After transformation into E. coli
Top10 cells, plasmid DNA was extracted from randomly picked
colonies and examined by restriction enzyme mapping. Plasmids
with the correct structures were prepared on large scales by CsCl
density gradient ultracentrifugation and the inserts were se-
quenced completely to ensure that no spurious mutations were
present. The verified genes were then inserted into pIK2 as
above. Plasmids with the desired structures were then used to
transform E. coli BL21(DE3) cells to create the final overex-
pression strains.
Acknowledgment. We thank the National Science Foundation
for generous financial support of this work (CHE-0130315).
Support by Ryukoku University (visit of T.M.) and the
University of Florida (Ruegamer Fellowship for I.A.K.) is also
gratefully acknowledged. We thank Dr. Mark R. Martzen for
providing the original clones for the yeast GST-fusion proteins.
Isolation of Yeast GST-Fusion Proteins. An overnight
culture of the appropriate overexpression strain grown in LB
medium containing 25 µg/mL kanamycin was diluted 1:100 into
5
00 mL of the same medium in a 2 L baffled flask. The culture
was shaken at 37 °C until the optical density at 600 nm reached
.5-1.0, then isopropylthio-â-D-galactoside was added to a final
0
Supporting Information Available: Restriction map and
sequence of pIK2, the vector used to overexpress GST-fusion
proteins in E. coli and alternative versions of Tables 1-3, in
which the data are presented in numerical form (7 pages, print/
PDF). This material is available free of charge via the Internet
at http://pubs.acs.org.
concentration of 100 µM and the culture was shaken for an
additional 6 h at room temperature. The cells were collected
by centrifugation, washed twice with cold sterile water, then
resuspended in 25 mL of 100 mM KPi, pH 7.0. All purification
steps were carried out at 0-4 °C. The cells were lysed by
passage through a French pressure cell and debris was removed
by centrifugation at 10 000 × g for 10 min at 4 °C. The
JA0469479
12832 J. AM. CHEM. SOC.
9
VOL. 126, NO. 40, 2004