Substrate-dependent stereospecificity of Tyl-KR1: An isolated polyketide synthase ketoreductase domain from Streptomyces fradiae

Article abstract of DOI:10.1002/chem.201300554

The stereospecificity of an enzymatic reaction depends on the way in which a substrate and its enantiomer bind to the active site. These binding modes cannot be easily predicted. We have studied the stereospecificity and stereoselectivity of the ketoreductase domain Tyl-KR1 of the tylactone polyketide synthase from Streptomyces fradiae by analysing the stereochemical outcome of the reduction of five different keto ester substrates. The absolute configuration of the Tyl-KR1 reduction products was determined by using vibrational circular dichroism (VCD) spectroscopy combined with quantum chemical calculations. The conversion of only one of the tested substrates, 2-methyl-3-oxovaleric acid N-acetylcysteamine thioester, afforded the expected anti-(2R,3R) configuration of the α-methyl-β-hydroxyl ester product, representing the stereochemistry observed for the physiological polyketide product tylactone. For all other substrates, which were modified with respect to the type of ester and/or the chain length (C₄ instead of C ₅), the opposite configuration (anti-(2S,3S)) was obtained with significant enantio- and diastereoselectivity. Inversion of both stereocentres suggests completely different binding modes invoked by only minor modifications of the substrate structure. Substrate dependence: The stereospecificity of an enzymatic reduction depends on the way a substrate binds to the active site. By using vibrational circular dichroism, it was shown that the stereospecificity and stereoselectivity of the ketoreductase Tyl-KR1 from a polyketide synthase complex in Streptomyces fradiae are inverted for different surrogates of polyketide-like substrates (see figure). Copyright

Full text of DOI:10.1002/chem.201300554

DOI: 10.1002/chem.201300554
Substrate-Dependent Stereospecificity of Tyl-KR1: An Isolated Polyketide
Synthase Ketoreductase Domain from Streptomyces fradiae
Matthias Hꢀckh, Michael Mꢁller, and Steffen Lꢁdeke*[a]
Abstract: The stereospecificity of an
enzymatic reaction depends on the way
in which a substrate and its enantiomer
bind to the active site. These binding
modes cannot be easily predicted. We
have studied the stereospecificity and
stereoselectivity of the ketoreductase
domain Tyl-KR1 of the tylactone poly-
ketide synthase from Streptomyces fra-
diae by analysing the stereochemical
outcome of the reduction of five differ-
ent keto ester substrates. The absolute
configuration of the Tyl-KR1 reduction
products was determined by using vi-
brational circular dichroism (VCD)
spectroscopy combined with quantum
chemical calculations. The conversion
of only one of the tested substrates,
2-methyl-3-oxovaleric acid N-acetylcys-
teamine thioester, afforded the expect-
ed anti-(2R,3R) configuration of the
a-methyl-b-hydroxyl ester product,
representing the stereochemistry ob-
served for the physiological polyketide
product tylactone. For all other sub-
strates, which were modified with re-
spect to the type of ester and/or the
chain length (C4 instead of C5), the op-
posite configuration (anti-(2S,3S)) was
obtained with significant enantio- and
diastereoselectivity. Inversion of both
stereocentres suggests completely dif-
ferent binding modes invoked by only
minor modifications of the substrate
structure.
Keywords: enzyme catalysis · poly-
ketides · stereospecificity · stereose-
lectivity · vibrational circular di-
chroism
Introduction
lide antibiotics is not established randomly but based on cer-
tain principles.[10] Knowledge of the underlying mechanisms
of stereochemical control by KR has significantly increased
in recent years.[11–14]
Tyl-KR1 from the first module of the tylactone PKS from
Streptomyces fradiae is involved in the biosynthesis of tylac-
tone, the aglycone precursor of the macrolide antibiotic ty-
losin (see the Supporting Information for the structure).[8,11]
Tyl-KR1 is known to reduce racemic 2-methyl-3-oxovaleric
acid N-acetylcysteamine thioester (1) enantioselectively to
Enzymes are often regarded as highly specific catalysts with
respect to a dedicated substrate and a selective outcome in
a biocatalytic process. However, enzymes are multifunction-
al and/or promiscuous[1] and even slight modifications of the
substrate may lead to altered transformations, apparent as
different stereospecificities and stereoselectivities for highly
related enzymes.[2] Carbonyl reductases ((CR), or alcohol
dehydrogenases (ADH)) are well known for their stereose-
lectivity in delivering chiral secondary alcohols from ketone
substrates. Their broad substrate range allows for their ap-
plication in the asymmetric synthesis of chiral building
blocks. The related ketoreductases (KR) can be found
among the basic catalytic domains that build up modular
polyketide synthases (PKS).[3,4] The use of KR with non-nat-
ural substrates provides new routes to polyketides and poly-
ketide-derived compounds.[5] The arrangement of hydroxyl
groups and their absolute configuration characterise the
skeleton of polyketides, such as macrolide antibiotics.[6] The
overall stereochemistry of these natural products depends
on the KR domains of the respective PKS.[7–9] In 1965,
Celmer proposed that the absolute configuration of macro-
the
corresponding
anti-(2R,3R)-a-methyl-b-hydroxyl
ester.[12] According to Prelogꢀs rule for the reduction of keto
groups, this mode can be called “anti-Prelog-type”.[15] The
resulting “anti-Prelog-type” anti configuration (also desig-
nated as d,d-configuration) is uncommon for natural prod-
ucts, although it has been recently identified in several poly-
ketides[16–18] and related secondary metabolites[19,20] of soil
bacteria.
The unusual stereoselectivity of Tyl-KR1 is surprising be-
cause the enzyme shows high sequential and structural ho-
mology to other KR, such as Ery-KR1; the difference in
stereoselectivity of these two particular KR has been eluci-
dated by docking studies of 2-methyl-3-oxovaleric acid
bound to acyl carrier protein (ACP).[11] The stereospecificity
and stereoselectivity of Tyl-KR1, however, may be altered
for smaller substrates.[21] Recently, it could be shown for a
fungal KR that the stereoselectivity was strongly correlated
to the chain length of the substrate.[22]
[a] M. Hꢁckh, Prof. Dr. M. Mꢂller, Dr. S. Lꢂdeke
Institute of Pharmaceutical Sciences
Albert-Ludwigs-Universitꢁt Freiburg
Albertstr. 25, 79104 Freiburg (Germany)
Fax : (+49)761-2036351
We have tested heterologously expressed Tyl-KR1 for its
stereospecificity for reduction of five different polyketide
analogues, including 1. After derivatisation of the resulting
E-mail: steffen.luedeke@pharmazie.uni-freiburg.de
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem.201300554.
8922
ꢃ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2013, 19, 8922 – 8928

FULL PAPER
b-hydroxyl esters, the stereochemical outcome was studied
by using vibrational circular dichroism (VCD) spectroscopy
for the determination of the absolute and relative configura-
tion.[23,24]
All enzymatic reductions were performed at pH 7.5 in po-
tassium phosphate buffer (KPi) by using lysate containing
Tyl-KR1 heterologously expressed in E. coli (Scheme 1).
The enzyme concentration was adjusted to 2.5 UmLÀ1 (ac-
tivity units refer to a photometric kinetic assay performed
with 5 as a standard substrate). For regeneration of the co-
factor nicotinamide adenine dinucleotide phosphate
(NADPH), glucose and glucose dehydrogenase (GDH)
were added.[25] Possible background reactions catalysed by
additionally expressed enzymes in Tyl-KR1-containing
lysate or GDH were identified from reactions with Tyl-
KR1-free lysate, and accounted for 2 (1), 4 (5) and less than
1% (2, 3 and 4). Incubation for 24 h resulted in virtually
complete reduction of 1 (99%), 4 (99%) and 5 (99%), and
reduced conversion of 2 (88%) and 3 (46%; Table 1).
Results and Discussion
The five tested substrates were 2-methyl-3-oxo esters, which
were modified with respect to the type of ester and/or the
chain length (Scheme 1). 2-Methyl-3-oxovaleric acid N-ace-
Table 1. Tyl-KR1-catalysed reduction of polyketide analogues 1–5.
Substrate
(product)
Conversion[a]
(Yield) [%]
d.r.[b] anti/syn
ee[c]
[%]
Absolute
G
G
configuration[d]
1 (1a)
2 (2a)
3 (3a)
4 (4a)
5 (5a)
>99 (79)
88 (69)
46 (33)
>99 (71)
99 (74)
>99:1
95:5
99:1
>99:1
97:3
84
36
54
43
92
2R,3R
2S,3S
2S,3S
2S,3S
2S,3S
[a] According to 1H NMR spectroscopy. [b] According to GC-MS analy-
sis of 1a–5a. [c] According to chiral-phase FID-GC analysis of deriva-
tives 8 and 9 (Figure 3). [d] According to VCD spectroscopy of deriva-
tives 8 and 9 (Figure 1 and 2).
Scheme 1. Regioselective enzymatic reduction of racemic b-keto esters
leads to four possible products. The substrates racemise quickly through
a tautomeric equilibrium, thus the reaction allows for dynamic kinetic
resolution (SNAC: N-acetylcysteamine thioester, KPi: potassium phos-
phate).
Stereochemical analyses, including the determination of
the relative and absolute configuration, diastereomeric ratio
(d.r.) and enantiomeric excess (ee), were performed on each
of the resulting products 1a–5a. Prior to analysis, the puri-
fied b-hydroxyl esters 1a–5a were derivatised. Thus, reduc-
tive ester cleavage by using lithium aluminum hydride[26]
gave the chiral diols 2-methylpentane-1,3-diol (6) and 2-
methylbutane-1,3-diol (7), which were converted into the
corresponding acetonides, 4-ethyl-2,2,5-trimethyl-1,3-diox-
ane (8) and 2,2,4,5-tetramethyl-1,3-dioxane (9; Scheme 2).[27]
This derivatisation proved to be useful for various rea-
sons: 1) LiAlH4 treatment of the enzymatic reduction prod-
ucts 1a–3a (esters of 3-hydroxy-2-methylvaleric acid) results
in constitutively identical diols 6 or, in the case of 4a and 5a
(esters of 3-hydroxy-2-methylbutyric acid), in diols 7. In this
way, the analytical data of structurally different substrates
become comparable. 2) Acetonides 8 and 9 facilitate chiral-
phase GC analysis and hence the precise determination of
ee values. 3) The rigid conformation of 8 and 9 allows for
tylcysteamine thioester (1) mimics the physiological ACP-
bound substrate in polyketide biosynthesis.[8] Tyl-KR1-cata-
lysed reduction of 1 selectively led to anti-(2R,3R)-3-hy-
droxy-2-methylvaleric acid N-acetylcysteamine thioester
(1a) with a high enantiomeric excess.[12] According to Pre-
logꢀs rule, the stereoselectivity in enzymatic carbonyl reduc-
tion depends on the preferred fit of a small and a large
group adjacent to the carbonyl in the active site of the
enzyme.[15] Thus, the (R) stereoselectivity of the Tyl-KR1-
catalysed keto reduction may be determined by the steric
demand of the N-acetylcysteamine thioester (SNAC)
moiety. To test this, we used the ethyl and tert-butyl esters, 2
and 3, of 2-methyl-3-oxovaleric acid representing variants of
the “large group” with a lower and a higher steric demand
than SNAC, respectively. Recently, a change in Tyl-KR1
stereoselectivity was reported for the shorter-chain substrate
2-methyl-3-oxobutyric acid SNAC (4),[21] implying that a
change of the “small group” from ethyl to methyl may also
have an impact. Therefore, in addition to the esters of 2-
methyl-3-oxovaleric acid 1–3, we also tested esters of 2-
methyl-3-oxobutyric acid. Besides SNAC ester 4, we chose
the ethyl ester 5 with both the “small group” and the “large
group” modified relative to 1.
Scheme 2. Derivatisation of enzymatic reduction products 1a–5a for
VCD and FID-GC analysis.
Chem. Eur. J. 2013, 19, 8922 – 8928
ꢃ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
8923

S. Lꢂdeke et al.
À
straightforward determination of the absolute and relative
configuration by VCD spectroscopy in combination with
quantum chemical calculations.
For assignment of the absolute and relative configurations
of 1a, 2a and 3a, we modelled trans-(4R,5S)-8 and cis-
(4R,5R)-8 by using SPARTAN (Wavefunction Inc., Irvine,
CA, USA). After conformer search at the molecular me-
chanics level (MMFF) in SPARTAN, we chose two conform-
ers each for trans-(4R,5S)-8 and cis-(4R,5R)-8, respectively,
and twisting vibrations linked to the C C stretching be-
tween C2 and the equatorial methyl group of the acetonide.
The opposite sign is probably due to the different relative
configuration with respect to the methyl group at C5 for the
two diastereomers and its implication on these modes. Twist-
ing and wagging modes of protons along the ring adjacent to
C4, C5, C6 and the ethyl group appear at 1290 and
1319 cmÀ1. Misinterpretation due to errors from inaccurate
Boltzmann weighting was excluded after thorough inspec-
tion of the spectral contributions from each calculated con-
former (see the Supporting Information for details). This is
consistent with observed axial–axial coupling of H4 and H5
for the DFT calculation (B3LYP/6-31+GACHTNUTRGNEUNG(d,p) level) of IR
and VCD spectra in Gaussian 09.[28] The two conformers
both exhibit ring puckering allowing for an equatorial ethyl
side chain and represent rotamers of the ethyl group with
minimal steric hindrance from the methyl groups at C2 and
C5 (Scheme 2 and the Supporting Information, Figure S2).
The measured VCD spectra (a–c) of acetonides 8 and VCD
spectra calculated for trans-(4R,5S)-8 (d) and cis-(4R,5R)-8
(e) are shown in Figure 1A. The theoretical spectra corre-
spond to the Boltzmann average of the spectra calculated
for the two conformers of the trans- and cis-acetonides
weighted with respect to relative energies obtained after
DFT geometry optimisation.
1
in the H NMR spectrum of 8, indicating trans configuration
(see the Supporting Information). From the determination
of trans-(4R,5S) configuration (a, Figure 1A), we deduced
anti-(2R,3R) configuration for SNAC ester 1a (Scheme 2).
This is in agreement with the stereoselectivity of the Tyl-
KR1-catalysed reduction of 1 previously reported in the lit-
erature.[12]
Surprisingly, the VCD spectra b and c appeared as mirror
images of spectrum a (Figure 1A), albeit with identical IR
absorbance (a–c, Figure 1B) and NMR spectra (see the Sup-
porting Information). From these data, the absolute configu-
ration of reduction products 2a and 3a can be unequivocally
assigned to the anti-(2S,3S) configuration. These findings
suggest that Tyl-KR1 stereoselectivity and thus the binding
of substrates 2 and 3 differ significantly from the binding of
1. Interestingly, both binding modes allow for dynamic ki-
netic resolution of the enantiomers with high diastereoselec-
tivity and moderate to good enantiomeric excess (Table 1,
Figure 3 and the Supporting Information, Figure S6).[29]
For assignment of the relative and absolute configurations
of 3-hydroxy-2-methylbutyric
The measured VCD spectrum a in Figure 1A shows good
agreement in sign and band shape with the calculated spec-
trum of trans-(4R,5S)-8 (spectrum d, Figure 1A). This is
most obvious for the bands at (À) 1475, (+) 1465, (À) 1386,
(+) 1369, (+) 1173, (À) 1059 and (+) 986 cmÀ1. To distin-
guish between trans-(4R,5S)-8 and its diastereomer cis-
(4R,5R)-8, we inspected the VCD bands at 1220, 1290 and
1319 cmÀ1 (indicated by asterisks), which are of opposite
sign in the calculated spectra (d and e, Figure 1A). The
band at around 1220 cmÀ1 corresponds to bending, wagging
acid esters 4a and 5a, we mod-
elled cis- and trans-9, diaster-
eomers of the corresponding
acetonide derivative. For both
cis- and trans-9, the calculated
spectra corresponded to
a
single conformer, each of
which accounted for over
99.9% of the Boltzmann popu-
lation after the B3LYP/6-31+
GACTHNUTRGNE(UNG d,p) calculation of the rela-
tive energies. Figure 2 shows
measured and calculated VCD
and IR spectra of acetonide
derivatives 9. Comparison of
the VCD spectra (Figure 2A)
indicates good overall agree-
ment with the spectrum calcu-
lated for trans-(4S,5R)-9, in
particular for prominent bands
at (+) 1383, (À) 1309, (À)
1243, (+) 1200 and (+)
994 cmÀ1. Furthermore, the
band pattern of (+) 1383, (À)
Figure 1. A) Comparison of recorded VCD spectra (corrected for CDCl3 background; a–c) of trans-8 to calcu-
lated spectra (B3LYP/6-31+GACTHNUTRGNE(NUG d,p); d, e) of trans-(4R,5S)-8 and cis-(4R,5R)-8 allows for deduction of the ab-
solute configuration of Tyl-KR1 reduction products. B) Simultaneously recorded IR spectra (a–c) of trans-8
enantiomers compared with calculated spectra (d and e).
8924
www.chemeurj.org
ꢃ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2013, 19, 8922 – 8928

Substrate-Dependent Stereospecificity of Tyl-KR1
FULL PAPER
used for ee determination, be-
cause the amount of syn dia-
stereomers (leading to cis-9),
formed upon Tyl-KR1-cata-
lysed reduction of 4 and 5 was
negligible according to GC-MS
and 1H NMR data (see the
Supporting Information for de-
tails).
The findings from VCD and
chiral phase GC analysis of 9
suggest that the (R,R) stereo-
selectivity supposedly deter-
mined by the SNAC ester is in
part outweighed by modifica-
tion of the “small group”. This
is in agreement with a recent
report on the loss of (R,R)
stereoselectivity for the short-
er-chain substrate 4 based on
chiral-phase HPLC studies.[21]
The interplay of two factors,
the SNAC ester on one hand,
increasing the propensity for
the anti-(2R,3R)-product, and
Figure 2. A) Deduction of the absolute configuration of Tyl-KR1 reduction products by comparison of record-
ed VCD spectra (corrected for CDCl3 background; a, b) of trans-9 to calculated spectra (B3LYP/6-31+GACTHNUTRGNEUG(N d,p);
c, d) of trans-(4S,5R)-9 and cis-(4S,5S)-9. B) Simultaneously recorded IR spectra (a, b) of trans-9 compared
with the calculated spectra (c, d).
1309 and (À) 1243 cmÀ1 (Figure 2A, asterisks) is characteris-
tic for trans-(4S,5R) stereochemistry: twisting and wagging
modes of protons adjacent to C4, C5 and C6 at 1309 cmÀ1
appear with opposite sign in the calculated VCD spectrum
of the corresponding cis species, whereas other modes corre-
sponding to those discussed for 8 are not sensitive to differ-
ent configurations at C5. The distinct band at 1309 cmÀ1,
however, is well reproduced in both experimental spectra
and allows for assignment of relative trans stereochemistry,
which was again corroborated by coupling constants of ring
protons in the 1H NMR spectrum of 9. The trans-(4S,5R)
stereochemistry of 9 indicates anti-(2S,3S) configuration for
Tyl-KR1 reduction products 4a and 5a.
the shorter chain length on the other hand, promoting the
formation of the anti-(2S,3S)-product, is reflected by the
moderate ee value of 43% for 4a (Table 1) and the low
molar absorptivity of the signals in the VCD spectrum (a,
Figure 2A). This is due to signal cancellation when both
enantiomers are present. Tyl-KR1-catalysed reduction of 5,
with modifications at both ends of the molecule (shorter
chain length, ethyl instead of SNAC), each of which may act
Figure 3 shows chromatograms of acetonides after deriva-
tisation of all the studied Tyl-KR1 reduction products 1a–5a
and of rac-cis/trans-8 and rac-cis/trans-9, the derivatives of
non-selective LiAlH4 reductions of 1 and 4, respectively.
The VCD results could now be used to determine the elu-
tion order of the respective isomers of 8 to the retention
times tR =40.0 min for trans-(4S,5R)-8, 41.8 min for trans-
(4R,5S)-8 and 45.4 and 48.8 min, respectively, for the two
enantiomers of cis-8 (Figure 3A, a). Comparison of the
chromatograms of derivatives of 1a–3a (Figure 3A, b–d)
supports the findings from VCD that the choice of ester de-
termines the enantioselectivity rather than the diastereose-
lectivity of the enzymatic reduction. The elution order of 9
could be assigned in a similar way, with trans-(4S,5R)-9 ap-
pearing at tR =41.4 min, trans-(4R,5S)-9 and one of the cis-9
enantiomers superimposed on each other at 45.0 min and
the second cis-9 enantiomer at 50.4 min (Figure 3B). Al-
though the four expected peaks are not fully resolved in the
reference trace (a), the chromatograms b and c could still be
Figure 3. A) Chiral phase GC chromatograms of 8 after derivatisation of
a) compound 1, leading to all four stereoisomers, b) compound 1a,
c) compound 2a and d) compound 3a. B) Chiral phase GC chromato-
grams of 9 after derivatisation of a) compound 4, b) compound 4a, and
c) compound 5a.
Chem. Eur. J. 2013, 19, 8922 – 8928
ꢃ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
8925

S. Lꢂdeke et al.
as an (S,S)-promoting factor, results in a higher ee value
(92%, Table 1) and a stronger VCD signal (spectrum b, Fig-
ure 2A).
In summary, these findings imply that the stereospecificity
and stereoselectivity of the bacterial ketoreductase Tyl-KR1
are not controlled by steric interactions with the “small
group” alone.
The comparison of X-ray crystal structures of Tyl-KR1
and other polyketide synthase-derived KR had previously
allowed for deduction of amino acid motifs to be correlated
with stereoselectivities and -specificities.[11] According to
modelling studies, the correct orientation of the physiologi-
cal substrate in KR of the Tyl-KR1 type depends on guid-
ance of the phosphopantetheinyl arm, connecting the sub-
strate to ACP, by a conserved Leu-Asp-Asp (LDD) motif.[11]
For the non-physiological substrates used in this study, this
shuttling mechanism is not possible but may be mimicked
by the pantetheinyl-like SNAC ester in 1 and 4 through
electrostatic interactions to other less remote residues than
the LDD motif. These interactions, however, may be less
pronounced for shorter-chain substrates like 4.
A detailed analysis of the mechanism of substrate-de-
pendent stereospecificity of ketoreductases has implications
in a broader sense on the use of enzymatic catalysis in ster-
eoselective synthesis. In previous studies, dedicated CR/
ADH have been used for the generation of polyketide-de-
rived chiral building blocks such as 2a and 5a,[32–34] and can
even reduce longer-chain g-methyl-b,d-diketo esters stereo-
selectively, leading to syn-(S,R)-,[35] syn-(R,S)- and anti-(S,S)-
configured products.[36] Yet a similar approach for accessing
anti-(R,R)-configured products remains difficult. Our results
suggest that enzymatic selectivity may be closely linked to
particular structural features in the substrate. This means
that stereospecificity and -selectivity of a CR cannot be reli-
ably predicted for different substrates and has thus to be de-
termined for every single reaction. We have shown that this,
in many cases, very tedious task can be significantly facilitat-
ed by the use of VCD rather than relying on comparison to
synthetic references.
Conclusion
Tyl-KR1 from Streptomyces fradiae is responsible for the
unusual anti-(R,R) stereochemical motif at positions 14 and
15 in the macrolide antibiotic tylosin.[8,12,30] Tyl-KR1, hetero-
logously expressed in E. coli, is able to retain this stereospe-
cificity for the physiological substrate mimic 1. We have
used VCD spectroscopy for the determination of the rela-
tive and absolute configuration of the reduction products of
surrogates of 1 with different steric properties. For qualita-
tive and quantitative analysis, the reduction products were
standardised by using straightforward derivatisation. We
have shown that the stereospecificity and stereoselectivity of
Tyl-KR1 strongly depends on both the type of ester and the
hydrocarbon tail of the b-keto ester substrate. The physio-
logical substrate mimic 1 is the only substrate for which Tyl-
KR1 had the naturally observed anti-(R,R) stereoselectivity;
other esters, regardless of lower (ethyl) or higher (tert-butyl)
steric demand, invoke a higher propensity for the anti-(S,S)-
product. Reduction of 2-methyl-3-oxobutyric acid esters led
to anti-(2S,3S) configuration, even for the SNAC ester,
which indicates a crucial role of the hydrocarbon tail for the
stereospecificity/stereoselectivity of Tyl-KR1. We, in analogy
to others, propose the existence of two distinct binding
modes leading to either anti-(R,R) or anti-(S,S) stereochem-
istry. By using substrate 1 as a standard, changing only one
of the two factors (ester moiety or hydrocarbon tail) shifts
the binding mode from anti-(R,R) to anti-(S,S) selective,
leading to a moderate ee value but still a high d.r. Changing
both factors, such as in the case of 5, leads to almost full in-
version of the stereospecificity and stereoselectivity.
This behaviour cannot be explained by a classic Prelog-
like model solely based on steric interactions, as previously
suggested for substrate-dependent reversal of anti-(2R,3R)
selectivity with an a-alkyl-b-keto ester reductase from Pau-
cimonas lemoignei.[31] A chain-length-dependent stereoselec-
tivity has recently been described for a KR from a fungal
PKS.[22] For Tyl-KR1, it has been reported that the absolute
configuration of the reduction products depends on the
length of the hydrocarbon tail adjacent to the keto end of
the substrate molecule,[21] which is supported by our VCD
and chromatography data. Whereas such a correlation with
higher steric demand of the “small group” would be in
agreement with a Prelog-like model for stereoselectivity,
this does not hold for the modification of the ester moiety
of substrates in Tyl-KR1-catalysed reduction. Introduction
of an ethyl ester eliminated anti-(R,R) selectivity. This could
not be recovered by using the sterically demanding tert-
butyl group, leaving SNAC as the only tested ester to be re-
duced with the indigenous “anti-Prelog-type” selectivity of
the enzyme.
The substrate dependence of stereoselective reactions
could be exploited for organic synthesis. We envisage that
synthetic routes could be tailored to meet the structural re-
quirements of a substrate to achieve a particular stereoselec-
tivity. This would allow for the synthesis of different stereo-
isomers through different synthetic routes, but still employ-
ing the same enzyme for the crucial enzymatic reduction
step.
Experimental Section
VCD measurements: The acetonide 8 or 9 was dissolved in anhydrous
CDCl3 to a concentration between 0.3 and 0.4m. IR spectra were meas-
ured on a Bruker Tensor 27 FT-IR spectrometer. For VCD measure-
ments, a Bruker PMA 50 VCD module (resolution: 4 cmÀ1) equipped
with a Stanford Research Systems SR 830 DSP lock-in amplifier and a
1
Hinds Instruments PEM-100 controller (retardation = l, optimum retar-
4
dation at 1400 cmÀ1, frequency: 50 kHz) was used. Solutions were placed
in a 110 mm BaF2 rotating cell. Experimental spectra represent the aver-
age over 6 h of measuring after solvent subtraction.
8926
www.chemeurj.org
ꢃ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2013, 19, 8922 – 8928

Substrate-Dependent Stereospecificity of Tyl-KR1
FULL PAPER
Quantum chemical calculations: For modelling of different conformers,
the MMFF-based conformer search algorithm in SPARTAN 08 software
(Wavefunction Inc., Irvine, CA, USA) was used. Two conformers each
for cis- and trans-8, both with equatorial ring puckering, representing ro-
tamers of the ethyl group at C4, were selected for geometry optimisation,
energy and IR/VCD calculations at the DFT level (B3LYP/6–31+G-
matographie Service, Langerwehe, for helpful discussions. We would like
to acknowledge the use of the computing resources provided by the
Black Forest Grid Initiative.
[1] S. D. Copley, Curr. Opin. Chem. Biol. 2003, 7, 265–272.
[2] A. Hummel, E. Brꢂsehaber, D. Bçttcher, H. Trauthwein, K. Doder-
er, U. T. Bornscheuer, Angew. Chem. 2007, 119, 8644–8646; Angew.
Chem. Int. Ed. 2007, 46, 8492–8494.
[3] S. Donadio, M. J. Staver, J. B. McAlpine, S. J. Swanson, L. Katz, Sci-
ence 1991, 252, 675–679.
[4] D. J. Bevitt, J. Cortes, S. F. Haydock, P. F. Leadlay, Eur. J. Biochem.
1992, 204, 39–49.
[5] M. Mꢂller, Angew. Chem. 2005, 117, 366–369; Angew. Chem. Int.
Ed. 2005, 44, 362–365.
[6] S. D. Rychnovsky, Chem. Rev. 1995, 95, 2021–2040.
[7] R. Castonguay, W. G. He, A. Y. Chen, C. Khosla, D. E. Cane, J. Am.
Chem. Soc. 2007, 129, 13758–13769.
[8] R. Castonguay, C. R. Valenzano, A. Y. Chen, A. Keatinge-Clay, C.
Khosla, D. E. Cane, J. Am. Chem. Soc. 2008, 130, 11598–11599.
[9] C. R. Valenzano, R. J. Lawson, A. Y. Chen, C. Khosla, D. E. Cane, J.
Am. Chem. Soc. 2009, 131, 18501–18511.
[10] W. D. Celmer, Antimicrob. Agents Chemother. 1965, 5, 144–156.
[11] A. T. Keatinge-Clay, Chem. Biol. 2007, 14, 898–908.
[12] A. P. Siskos, A. Baerga-Ortiz, S. Bali, V. Stein, H. Mamdani, D.
Spiteller, B. Popovic, J. B. Spencer, J. Staunton, K. J. Weissman, P. F.
Leadlay, Chem. Biol. 2005, 12, 1145–1153.
ACHTUNGTRENNUNG
(d,p)) in Gaussian 09.[28] In the case of 9, in which there is a methyl group
at C4, only a single conformer with equatorial ring puckering for both
cis- and trans-9 (Boltzmann weights: 99.9% with respect to relative ener-
gies from B3LYP/6–31+GACTHNUTRGENUG(N d,p) calculations) was chosen. Frequencies of
the calculated spectra were scaled by 0.97. Theoretical spectra were gen-
erated as the Boltzmann-weighted average of the single calculated spec-
tra with Lorentzian lineshapes of 6 cmÀ1 bandwidth around calculated in-
tensities.
Chiral gas chromatography: Chromatograms were recorded on a Shimad-
zu GC-2010 gas chromatograph equipped with a flame ionisation detec-
tor (FID) and using the chiral stationary phase FS-Cyclodex beta I/P
column (length: 50 m; inner diameter: 0.32 mm) from CS Chromatogra-
phie Service (Langerwehe, Germany). Samples of 8 and 9 were dissolved
in ethyl acetate. Parameters for separation of 8: split injection, pressure:
0.7 bar, carrier gas flow (helium): 7.9 mLminÀ1, temperature program:
708C/55 min, then 408C minÀ1 increase to 1808C/15 min. Parameters for
separation of 9: split injection, pressure: 0.7 bar, carrier gas flow
(helium): 8.1 mLminÀ1, temperature program: 608C/55 min, then 408C
minÀ1 increase to 1808C/15 min.
Enzymatic reductions: d-Glucose (5 equivalents) was dissolved in KPi
buffer (50 mm, pH 7.5). The substrate 1–5 (120 mg,
1 equivalent),
NADP+ solution (12.5 mm, 0.05 equivalents) and GDH (150 U, final con-
centration: 6 UmLÀ1) were added. Finally, Tyl-KR1-containing lysate
(62.5 U, final concentration: 2.5 UmLÀ1) was added, the total volume was
adjusted to 25 mL with KPi buffer and the reaction mixture was stirred
with a fish-clip for 24 h at room temperature. The reduction was termi-
nated by adding NaCl (3 g). After extraction with ethyl acetate (3ꢄ
60 mL), the pooled organic phases were filtered through coarse Celite
535 and dried over MgSO4. The solvent was removed under reduced
pressure. The crude products were purified by using flash column chro-
matography (see the Supporting Information for details).
[13] A. Baerga-Ortiz, B. Popovic, A. P. Siskos, H. M. OꢀHare, D. Spite-
ller, M. G. Williams, N. Campillo, J. B. Spencer, P. F. Leadlay, Chem.
Biol. 2006, 13, 277–285.
[14] P. Caffrey, ChemBioChem 2003, 4, 654–657.
[15] V. Prelog, Pure Appl. Chem. 1964, 9, 119–130.
[16] K. Ishida, T. Lincke, C. Hertweck, Angew. Chem. 2012, 124, 5566–
5570; Angew. Chem. Int. Ed. 2012, 51, 5470–5474.
[17] D. Janssen, D. Albert, R. Jansen, R. Mꢂller, M. Kalesse, Angew.
Chem. 2007, 119, 4985–4988; Angew. Chem. Int. Ed. 2007, 46, 4898–
4901.
Derivatisation reactions: Step 1: LiAlH4 reduction:[26] An ester 1, 4 or
1a–5a (1 equiv) was dissolved in distilled THF. This solution was slowly
added dropwise to a cooled suspension of LiAlH4 (2m in THF, 4 equiv)
in THF in an ice bath. The cooling bath was removed and the mixture
was stirred for 1 h at room temperature. Then, it was again cooled in an
ice bath, and water and 15% NaOH were successively added to quench
the remaining LiAlH4. This mixture was stirred for 1 h at room tempera-
ture. Solids were collected by filtration and washed with ethyl acetate
(5ꢄ2 mL). The organic phase was dried over MgSO4 and the solvent was
removed under reduced pressure. Flash column chromatography (ethyl
acetate/cyclohexane 7:3) afforded the 1,3-diol, 6 or 7.
[18] D. Pistorius, R. Mꢂller, ChemBioChem 2012, 13, 416–426.
[19] J. Barbier, R. Jansen, H. Irschik, S. Benson, K. Gerth, B. Bçhlen-
dorf, G. Hçfle, H. Reichenbach, J. Wegner, C. Zeilinger, A. Kirschn-
ing, R. Mꢂller, Angew. Chem. 2012, 124, 1282–1286; Angew. Chem.
Int. Ed. 2012, 51, 1256–1260.
[20] C. Jahns, T. Hoffmann, S. Mꢂller, K. Gerth, P. Washausen, G. Hçfle,
H. Reichenbach, M. Kalesse, R. Mꢂller, Angew. Chem. 2012, 124,
5330–5334; Angew. Chem. Int. Ed. 2012, 51, 5239–5243.
[21] S. K. Piasecki, C. A. Taylor, J. F. Detelich, J. Liu, J. Zheng, A. Kom-
soukaniants, D. R. Siegel, A. T. Keatinge-Clay, Chem. Biol. 2011, 18,
1331–1340.
[22] H. Zhou, Z. Gao, K. Qiao, J. Wang, J. C. Vederas, Y. Tang, Nat.
Chem. Biol. 2012, 8, 331–333.
[23] T. B. Freedman, X. L. Cao, R. K. Dukor, L. A. Nafie, Chirality 2003,
15, 743–758.
[24] P. J. Stephens, F. J. Devlin, J. J. Pan, Chirality 2008, 20, 643–663.
[25] C. H. Wong, D. G. Drueckhammer, H. M. Sweers, J. Am. Chem. Soc.
1985, 107, 4028–4031.
[26] D. V. Patel, F. Vanmiddlesworth, J. Donaubauer, P. Gannett, C. J.
Sih, J. Am. Chem. Soc. 1986, 108, 4603–4614.
[27] S. E. Bode, M. Mꢂller, M. Wolberg, Org. Lett. 2002, 4, 619–621.
[28] Gaussian 09, Revision B.01, M. J. Frisch, G. W. Trucks, H. B. Schle-
gel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V.
Barone, B. Mennucci, G. A. Petersson, H. Nakatsuji, M. Caricato, X.
Li, H. P. Hratchian, A. F. Izmaylov, J. Bloino, G. Zheng, J. L. Son-
nenberg, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa,
M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven,
J. A. Montgomery Jr., J. E. Peralta, F. Ogliaro, M. Bearpark, J. J.
Heyd, E. Brothers, K. N. Kudin, V. N. Staroverov, R. Kobayashi, J.
Normand, K. Raghavachari, A. Rendell, J. C. Burant, S. S. Iyengar,
J. Tomasi, M. Cossi, N. Rega, N. J. Millam, M. Klene, J. E. Knox,
J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E.
Step 2: Acetonide formation:[27] 2,2-Dimethoxypropane and acetone (each
ca. 1 mL per 50 mg diol) were added to 6 or 7 (1 equivalent) in a round-
bottom flask. Molecular sieves (pore size: 3 ꢅ) and p-toluenesulfonic
acid monohydrate (0.02 equivalents) were added and the mixture was
stirred for 2–2.5 h at room temperature (reaction monitored by TLC).
After the addition of distilled methyl tert-butyl ether (MTBE) and satu-
rated aqueous NaHCO3 solution, the aqueous phase was extracted with
MTBE (3ꢄ5 mL). The pooled organic phases were washed with brine
and dried over MgSO4. The solvent was carefully removed under reduced
pressure to afford the acetonide 8 or 9.
Acknowledgements
This work was funded by the DFG (IRTG 1038: Catalysts and Catalytic
Reactions for Organic Synthesis). We thank Prof. P. Leadlay, Cambridge,
for providing the tylactone PKS gene cluster sequence, W. Kornberger
for cloning of the Tyl-KR1 gene and transformation in E. coli, V. Brecht
and S. Ferlaino for NMR measurements, and Dr. V. Lorbach, CS Chro-
Chem. Eur. J. 2013, 19, 8922 – 8928
ꢃ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
8927

S. Lꢂdeke et al.
Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W.
Ochterski, R. L. Martin, K. Morokuma, V. G. Zakrzewski, G. A.
Voth, P. Salvador, J. J. Dannenberg, S. Dapprich, A. D. Daniels, ꢆ.
Farkas, J. B. Foresman, J. V. Ortiz, J. Cioslowski, D. J. Fox, Gaussian
Inc., Wallingford, CT, USA, 2009.
[32] W. A. Kaluzna, T. Matsuda, A. K. Sewell, J. D. Stewart, J. Am.
Chem. Soc. 2004, 126, 12827–12832.
[33] D. Kalaitzakis, S. Kambourakis, D. J. Rozzell, I. Smonou, Tetrahe-
dron: Asymmetry 2007, 18, 2418–2426.
[34] D. Kalaitzakis, J. D. Rozzell, S. Kambourakis, I. Smonou, Org. Lett.
2005, 7, 4799–4801.
[35] A. G. Ji, M. Wolberg, W. Hummel, C. Wandrey, M. Mꢂller, Chem.
Commun. 2001, 57–58.
[29] Enantiomeric excesses (ee) were determined by chiral gas chroma-
tography of acetonide derivatives. Particular care was taken on veri-
fication of reduction products to avoid adulterated ee values due to
the nonspecific LiAlH4 reduction of residual keto ester substrates.
[30] P. Caffrey, Chem. Biol. 2005, 12, 1060–1062.
[36] S. Lꢂdeke, M. Richter, M. Mꢂller, Adv. Synth. Catal. 2009, 351, 253–
259.
[31] S. Rodriguez, P. Panizza, S. Onetto, Biocatal. Biotransform. 2007, 25,
414–417.
Received: February 12, 2013
Published online: May 16, 2013
8928
www.chemeurj.org
ꢃ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2013, 19, 8922 – 8928