Towards a novel explanation of Pseudomonas cepacia lipase enantioselectivity via molecular modelling of the enantiomer trajectory into the active site

Article abstract of DOI:10.1016/S0957-4166(03)00374-4

In the transesterification reaction between (RS)-2-bromophenyl acetic acid ethyl ester and 1-octanol in n-octane, Pseudomonas cepacia lipase enantioselectivity towards the (R)-isomer is 57. Two strategies are described to investigate the structural basis involved in this enzyme enantioselectivity. Molecular modelling of the tetrahedral intermediate mimicking the transition state enables the identification of two potentially productive substrate-binding modes for each enantiomer. However, the conformations obtained with the faster and slower-reacting enantiomers have equivalent potential energies and most of them possess the hydrogen bonds essential for catalysis. On this basis, it is not possible to distinguish the diastereomeric complexes. The second approach is original and consists in a simple but robust protocol of pseudomolecular dynamics simulations under constraints to map the probable trajectory of the enantiomers in the active site. Enzyme/substrate interaction energy is always found to be lower for the faster-reacting enantiomer, which satisfactorily corroborates the experimental results. Energy differences are attributed to specific interactions of these substrates with a network of hydrophobic residues lining the access path. Furthermore, mechanistic details suggest that the pivoting side chains of the hydrophobic residues act in a concerted step-tooth gear motion whose basic role is to select and guide the substrates towards the active site. With this type of lipase, such dynamic features could be the key explanation of this as yet unexplored enantiorecognition. For the slower-reacting enantiomer, it appears that the concerted motion of the side chains is perturbed when the substrate passes through a bottleneck formed by Val266 and Leu17. The enantioselectivity of mutant Val266Leu with a more bulky side chain at this position supports our assumption: by narrowing the bottleneck, the enantioselectivity was considerably enhanced as much as up to 200.

Full text of DOI:10.1016/S0957-4166(03)00374-4

TETRAHEDRON:
ASYMMETRY
Pergamon
Tetrahedron: Asymmetry 14 (2003) 1807–1817
Towards a novel explanation of Pseudomonas cepacia lipase
enantioselectivity via molecular modelling of the enantiomer
trajectory into the active site
David Guieysse,a Christophe Salagnad,b Pierre Monsan,a Magali Remaud-Simeona,* and Vinh Tranc
aCentre de Bioinge´nierie Gilbert Durand, De´partement de Ge´nie Biochimique et Alimentaire, UMR CNRS 5504,
UMR INRA 792, INSA, 135 Avenue de Rangueil, F-31077 Toulouse Cedex 4, France
bAventis Pharma, Process Development Biotechnology, 9, quai Jules Guesde, F-94400 Vitry sur Seine, France
cCentre de Nantes, Unite´ de Physicochimie des Macromole´cules, INRA, BP 71627, F-44316 Nantes Cedex 3, France
Received 11 March 2003; accepted 7 May 2003
Abstract—In the transesterification reaction between (RS)-2-bromophenyl acetic acid ethyl ester and 1-octanol in n-octane,
Pseudomonas cepacia lipase enantioselectivity towards the (R)-isomer is 57. Two strategies are described to investigate the
structural basis involved in this enzyme enantioselectivity. Molecular modelling of the tetrahedral intermediate mimicking the
transition state enables the identification of two potentially productive substrate-binding modes for each enantiomer. However,
the conformations obtained with the faster and slower-reacting enantiomers have equivalent potential energies and most of them
possess the hydrogen bonds essential for catalysis. On this basis, it is not possible to distinguish the diastereomeric complexes. The
second approach is original and consists in a simple but robust protocol of pseudomolecular dynamics simulations under
constraints to map the probable trajectory of the enantiomers in the active site. Enzyme/substrate interaction energy is always
found to be lower for the faster-reacting enantiomer, which satisfactorily corroborates the experimental results. Energy differences
are attributed to specific interactions of these substrates with a network of hydrophobic residues lining the access path.
Furthermore, mechanistic details suggest that the pivoting side chains of the hydrophobic residues act in a concerted step–tooth
gear motion whose basic role is to select and guide the substrates towards the active site. With this type of lipase, such dynamic
features could be the key explanation of this as yet unexplored enantiorecognition. For the slower-reacting enantiomer, it appears
that the concerted motion of the side chains is perturbed when the substrate passes through a bottleneck formed by Val266 and
Leu17. The enantioselectivity of mutant Val266Leu with a more bulky side chain at this position supports our assumption: by
narrowing the bottleneck, the enantioselectivity was considerably enhanced as much as up to 200. © 2003 Elsevier Science Ltd.
All rights reserved.
1. Introduction
tion, and aminolysis processes.1–6 Among them, Pseu-
domonas cepacia (recently renamed Burkholderia
cepacia) lipase (PCL) has turned out to be very useful
To reduce the cost of development of enantiopure
molecules such as pharmaceuticals, many efforts are
focused on the understanding of the molecular features
that control enzyme enantioselectivity. The aim is to
develop reliable predictive methods and to guide sub-
strate and/or enzyme modifications to enhance kinetic
resolutions. Lipases are serine hydrolases (EC 3.1.1.3)
frequently employed in organic synthesis to catalyse
kinetic resolutions of a wide range of substrates yield-
ing enantiomerically pure compounds. These enzymes
have been used to catalyse esterification transesterifica-
for the resolution of racemic mixtures of primary and
7–10
secondary alcohols
and carboxylic acids.11–19
The methods used to investigate the molecular factors
of enantioselectivity generally combine kinetic results,
structural data analyses and molecular modelling.20,21
22–27
Many serine hydrolase structures are available
and
the mechanistic details of the catalytic reaction are well
known.28–31 Proton transfer between the side-chain oxy-
gen Og of the catalytic serine and HisꢀNo of the
catalytic histidine enables nucleophilic attack by the
oxygen of the catalytic serine on the substrate carbonyl
function. This reaction yields a tetrahedral intermediate
stabilized by a network of hydrogen bonds followed by
* Corresponding author. Tel.: +33-56-155-9446; fax: +33-56-155-9400;
e-mail: remaud@insa-tlse.fr
0957-4166/03/$ - see front matter © 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0957-4166(03)00374-4

1808
D. Guieysse et al. / Tetrahedron: Asymmetry 14 (2003) 1807–1817
the release of the alcohol and the formation of the
acyl-enzyme. Then, the hydrolysis (or alcoholysis) of
the acyl-enzyme proceeds via the same mechanism
through the formation of a second tetrahedral interme-
diate followed by product release.
slower-reacting enantiomer and the amino acids of the
active site38,42–44 but none of them concerned P. cepacia
lipase.
In the present study, our aim was to determine the
structural basis of the P. cepacia lipase enantioselectiv-
ity in transesterification of an important building block
used in drug synthesis: (RS)-2-bromophenyl acetic acid
ethyl ester. We modelled the first tetrahedral intermedi-
ate obtained with each enantiomer assuming, from
experimental information, that this step is responsible
for the enantioselectivity of the overall reaction. Unfor-
tunately, this approach did not provide a satisfactory
explanation of the experimental data with the faster
and slower-reacting enantiomers. Thus, we built a sec-
ond protocol to tackle the probable trajectory of each
enantiomer through the access path.
To tackle the problem of enantioselectivity, X-ray
structures of lipase/enantiomer inhibitor complexes and
comparison of the complexes obtained with the (R)-
and (S)-inhibitors are very informative.32–34 Lang et
al.33 crystallised P. cepacia lipase with an enantiopure
analogue of triacylglycerol and showed that the model
obtained with the slower-reacting enantiomer revealed
unfavourable interactions between the carbonyl oxygen
(O7) of the sn-2 chain and Cd1 L287 and Cd1 I290
atoms. The recently reported crystal structure of P.
cepacia lipase complexed with a transition state ana-
logue of a secondary alcohol also showed that slower-
reacting enantiomer binding could not be productive
because of its orientation.32
From a methodological point of view, this protocol can
be assimilated to dynamics simulations under con-
straints. Analyses of the enzyme–substrate interactions
obtained along the path towards the active site for each
enantiomer are discussed with regards to enantioselec-
tivity and to the final goal, which is to design more
enantioselective mutants.
The methods combining kinetic results and structural
analyses also provide some empirical rules predicting
the enantiopreference of several lipases toward sec-
ondary alcohols35 or chiral carboxylic acids.36 These
rules establish a correlation between the size of the
substituents at the stereocentre and enantioselectivity.
Although useful, they do not precisely identify the
molecular determinants of enantiorecognition. The best
described approach consists in modelling the tetra-
hedral intermediates formed with each enantiomer
assuming that they are a good mimic of the transition
state.28 A lower potential energy is generally expected
for the faster-reacting enantiomer, but there is not
always a good correlation with experimental enantiose-
lectivity.37–39 Schulz et al.40,41 developed a quantitative
model to predict the enantioselectivity of Pseudomonas
cepacia lipase41 and of Candida rugosa lipase40 towards
a broad range of secondary alcohols. They established a
correlation between the HisNoꢀOalc distance of the
slower-reacting enantiomer and the experimental enan-
tioselectivity values.
2. Results and discussion
2.1. P. cepacia lipase enantioselectivity towards (RS)-2-
bromophenyl acetic acid ethyl ester
The PCL-catalysed transesterification between (RS)-2-
bromophenyl acetic acid ethyl ester (50 mM) and 1-
octanol at various concentrations (50–500 mM) was
first performed in n-octane, with the water activity
being controlled at 0.04. The transesterification rate
was maximal at 100 mM and decreased slightly above
300 mM, and no significant variation of enantioselectiv-
ity was observed between 50 and 500 mM (Figures 1a
and b). The transesterification reaction was then carried
out with six aliphatic alcohols of different chain
lengths. As reported in Table 1, all measurements
revealed the same preference for the (R)-enantiomer. In
addition, the enantioselectivity decreased slightly with
the length of the alcohol chain but this effect was not
very pronounced. From these results, it was assumed
Few studies address the problem of the enantioselectiv-
ity of lipases toward 2-substituted carboxylic acids.
Most computer simulations of lipase-catalysed resolu-
tions have revealed steric interactions between the
Figure 1. Effect of 1-octanol concentration (a) on initial rate and (b) on P. cepacia lipase enantioselectivity (E) for transesterifi-
cation between a-bromophenyl acetic acid ethyl ester (50 mM) and 1-octanol (50–500 mM) in n-octane at a water activity of 0.04.

D. Guieysse et al. / Tetrahedron: Asymmetry 14 (2003) 1807–1817
1809
Table 1. PCL-catalysed transesterification of (RS)-2-bromo-
phenyl acetic acid ethyl ester in n-octane. Effect of
aliphatic alcohol chain length on P. cepacia lipase enan-
tioselectivity
2.2.1.2. Description of binding modes. Among the four
tetrahedral intermediates that can be constructed
(named RR, SR, RS and SS), only two solutions (RR
and SR) were found to be productive on the basis of
the hydrogen bonding network necessary for catalysis.
Alcohols
E (viR/viS)
% Conversion
e.e.s (%)
e.e.p (%)
Two binding modes could exist (named I and II) in
both tetrahedral intermediates RR and SR as shown in
Figures 3a, b, c and d. Figures 3a and c (left side) show
the first binding mode for the RR and SR intermediate
types, respectively. The bromine atom points down
towards pocket HA, whereas the aromatic ring points
up towards pocket HB. Figures 3b and d (right side)
show the second binding mode with reverse orientations
for the bromine atom and the aromatic ring. For each
binding mode, the alkyl part of the substrate (-OCH2-
CH3) lies in pocket HH (bottom part of Fig. 3).
Butanol-1 70 ( 3)
Pentanol-1 63 ( 6)
Hexanol-1 64 ( 2)
Octanol-1 57 ( 7)
Decanol-1 53 ( 7)
43 (96 h)
44 (96 h)
46 (96 h)
47 (84 h)
49 (96 h)
72
76
77
77
76
87
86
86
84
77
E: enantioselectivity factor. e.e.s and e.e.p: enantiomeric excess of
substrate and product, respectively.
that the enzyme enantioselectivity is almost indepen-
dent on the deacylation step of the reaction. Conse-
quently, our modelling protocols were focused on the
primary stage of the reaction up to the formation of the
first tetrahedral intermediate.
The first criterion, which characterizes the productive
binding modes of the substrate in the active site is the
potential energy. The relative energies are 0, 4, 4 and 5
2.2. Molecular modelling
kcal/mol for SRmode I, SRmode II, RRmode I and RRmode II
,
respectively. Thus, SRmode I should be the preferred
solution because all others are about 5 kcal/mol higher.
But these results are unable to explain the R enantio-
preference of P. cepacia lipase. The enantioselectivity
could also be analysed according to the network of five
hydrogen bonds required for the catalysis as already
described in the literature.30,31,45 Table 2 reports all
hydrogen bonds shown in Figure 3. Within a maximum
2.2.1. Static molecular modelling of the tetrahedral
intermediates
2.2.1.1. Active site environment. The P. cepacia lipase
catalytic cavity (the active site plus the access path) has
three pockets33 as shown in Figures 2a and b. A first
inner hydrophobic pocket (HA), anchoring the acyl
part of the substrate, another inner hydrophobic/
hydrophilic pocket (HH), anchoring the alkyl part of
the substrate and finally, along the main axis of the
enzyme cavity, a third outer pocket (HB) can be seen
beyond Leu17 (Fig. 2a, above the plane of the paper).
The active site is deep inside the enzyme with a contigu-
ous access path like a funnel but with a pronounced
bottleneck in the middle, formed by Leu17 and Val266.
,
distance of 3.1 A, all intermediate conformations have
the five expected linkages except for SRmode I with the
missing His286[No2H···Og]Ser87 bond necessary for the
activation of the catalytic serine Og. In the latter case,
the absence of an essential H-bond seems to be incon-
sistent with the lowest potential energy.
Figure 2. Representation of the catalytic cavity with Connolly surfaces. (a) Hydrophobic pocket (HA, in orange), anchoring the
acyl part of the substrate, is lined by Leu17, Pro113, Phe119, Leu164, Leu167, Val266 and Val267. Hydrophobic/hydrophilic
pocket (HH, in blue), anchoring the alkyl part of the substrate, is lined by Gln16, Thr18, Tyr23, His86, Leu287, Ile290, Gln292
and Leu293. Residues Leu17, Ser87 and His286 (yellow, red, purple, respectively) separate pockets HA and HH. (b) Perpendicular
cross section of PCL cavity (roughly at level Ser87) showing the funnel shape of the access path (Ser87 at the bottom and His286,
Asp264 at the left). The bottleneck formed by Val266 and Leu17 delimits the inner sites (HA and HH) and the outer one (HB,
at the top). Ser87, His286 and Leu17 are in the plane of this figure and pockets HA and HH are in the background and in the
foreground, respectively. The bottleneck formed by Val266 and Leu17 delimits the inner sites (HA and HH) and the outer one
(HB, at the top).

1810
D. Guieysse et al. / Tetrahedron: Asymmetry 14 (2003) 1807–1817
Figure 3. For the sake of clarity, the enzyme orientation is identical to that of Figure 2a. Pockets HA and HB are at the top of
the schema (below and above plane of the figure, respectively) and pocket HH is at the bottom. (a) RRmode I, (b) RRmode II, (c)
SRmode I, (d) SRmode II. Residues in orange (Leu267, Pro113 and Leu167) represent the HA pocket (at the top and in the
background). The HB pocket is in above the plane of the paper while Tyr29 denotes the HH pocket (below). In binding mode
I, the bromine atom points down toward the HA pocket and the aromatic ring points up toward the HB pocket. Binding mode
II has the opposite orientation.
Table 2. Network of hydrogen bonds in tetrahedral intermediate geometries
a
,
Distances between heavy atoms (A)
Donor
Acceptor
RRmode I
RRmode II
SRmode I
SRmode II
His 286 Nd1H
His 286 No2H
His 286 No2H
Gln 88 NH
Od2 Asp 264
Og Ser 87
O-Et substrate
O− substrate
O− substrate
2.7
3.0
3.0
2.8
2.7
2.7
3.0
3.1
2.8
2.7
2.7
3.5
3.0
2.9
2.7
2.7
3.1
3.0
2.9
2.7
Leu 17 NH
a The hydrogen bond, can be detected with geometrical criteria: short distance between the donor and the acceptor atom and likely a linear
arrangement between the donor atom, hydrogen and the acceptor atom.45 These criteria are implemented in the InsightII package to
automatically visualize H-bonds.
Once more, this fact enhances the discrepancy between
this modelling protocol and the experimental results of
enantioselectivity. This means that despite our efforts
to detect, as exhaustively as possible, all the probable
intermediate geometries, those found were unable to
satisfactorily explain the PCL preference for the (R)-
enantiomer. Nevertheless, it must be mentioned that
modelling of this tetrahedral intermediate had to be
attempted first because, in several studies, this approach
was found appropriate to predict the lipase
enantioselectivity.38,40,41,45–48
2.2.2. Molecular modelling of the substrate trajectories.
As a consequence of the failure of the above model, we
assumed that the enantiomer discrimination possibly
occurred before the formation of the covalent bond
between the catalytic serine and the substrate during
the transport of the enantiomers to the active site
located at the bottom of a funnel. Thus, the preliminary
trajectory of each enantiomer in the active site and for
each mode was analysed. Indeed, if the access to the
active site is facilitated for one enantiomer, that will
affect the dissociation constant and the kinetic resolu-

D. Guieysse et al. / Tetrahedron: Asymmetry 14 (2003) 1807–1817
1811
tion. As already discussed, the funnel-like shape of the
access path technically yielded non-chronological
dynamics simulations and it was assumed that all tra-
jectories can be reversibly interpreted. Furthermore,
since there was no way to discriminate one of the two
binding modes, at this stage all initial SRmode I, SRmode
II, RRmode I and RRmode II conformations were subjected
to this dynamic protocol.
preliminary result is consistent with the experimental
preference for the (R)-enantiomer.
All along these trajectories, the potential energies of
both molecular entities were also checked. No interest-
ing information could be derived from the truncated
enzyme energy. On the contrary, those of the substrate
(Fig. 5) gave relevant information to discriminate the
two binding modes. The substrate potential energies are
always higher in binding mode II regardless of the
enantiomer type. This feature is particularly true for
low D(B1–B2) values (near the formation of the covalent
bond), which means that to produce binding mode II,
the substrate must be progressively distorted to adopt
the final geometry. These artificial distortions are corre-
lated with the [Fx,Fy (Scheme 1)] values corresponding
to high energy domains on the maps drawn at the first
stage of the former modelling protocol (maps and data
not shown). In contrast, for binding mode I, the
(Fx,Fy) values correspond to low minima regions. Fur-
thermore, as expected for enantiomeric forms, the (R)-
and (S)-substrate energies are similar within the small
fluctuations necessary to adapt their geometries to the
access path context. Thus, binding mode I must be the
only process leading to the tetrahedral intermediate.
2.2.2.1. Interaction energies along the trajectory. The
curves of the potential energy of the enzyme/substrate
interactions versus the D(B1–B2) values (described in
Section 4) are reported in Figure 4. The minimum
,
D(B1–B2) value (4.55 A) corresponds solely to the virtual
disruption of the OꢀC bond between the catalytic serine
and the substrate without any movement away of the
substrate. Due to the inherent incapacity for molecular
mechanics to properly model covalent bond disruption
(only a quantum scheme would be able to do that), an
initial abnormally high interaction energy is observed
for all starting geometries (the OꢀC bond being now
considered as too short a non-bonded distance between
the two entities). Therefore, the first part of these
curves must be discarded from the analysis.
As concerns the rest of the curves, two phenomena are
observed. The first general tendency is the slow but
continuous increase of the energy whatever the enan-
tiomer type and the binding mode. In the reverse
direction, this must be interpreted as a global regular
gradient of substrate attraction through the access path.
To sum up, at this stage in the dynamics protocol, these
calculations clearly show the (R)-enantiopreference,
which supports the more realistic binding mode I with
the bromine atom located in pocket HA and the aro-
matic ring in pocket HB.
,
More interestingly, approximately from 5.6 to 9.0 A,
the (R)-enantiomer trajectories always yield lower inter-
action energies compared to the (S)-enantiomer what-
ever the binding mode. It seems reasonable to consider
that this part of the curve is much more meaningful
2.2.2.2. Correction and analysis of the trajectories.
The R and S energy curves have several significant
variations. It appears that some other oscillating phe-
nomena are superimposed on the background continu-
,
than the other one (from 9.0 to 12.5 A) and this
Figure 5. Curves of potential energy of (RS)-2-bromophenyl
acetic acid ethyl ester versus D(B1–B2)
.
Scheme 1. Serine/2-bromophenyl acetic acid ethyl ester. Dihe-
drals angles FX(C1ꢀC2ꢀC7ꢀC8) and FY(C2ꢀC7ꢀC8ꢀO9).
Figure 4. Curves of non-corrected interaction energy between
the enzyme and the substrate versus D(B1–B2)
.

1812
D. Guieysse et al. / Tetrahedron: Asymmetry 14 (2003) 1807–1817
ous variation of the energy. For example, points A, B,
C and D were added to the RRmode I curve of Figure 4a
to illustrate a complete cycle of oscillations comprised
between two minima (points A and D) with a local
peak (point B) and a transitory state (point C). These
intriguing fluctuations can neither be correlated to the
potential energy of the enzyme nor to that of the
substrate and must be attributed to local unrelieved
steric conflicts between the two entities. The best way to
identify the molecular origin of these conflicts, was to
(elementary displacement of the substrate), the
algorithm was not able to properly diffuse this con-
straint toward the other entity and the frustrated rota-
tion of the Leu17 side chain, not exactly in phase with
the imposed substrate location at the end of the min-
imisation process, yielded a momentary energy increase.
This frustration can last for other subsequent increases
of D(B1–B2) (point C) until the local conflict becomes too
important inducing a flip of the Leu17 side chain. This
finally occurred at point D where the new Leu17 side
chain position returned back in phase with the sub-
strate location. It is important to mention that this
unrealistic fluctuation of energy certainly occurs if the
reverse trajectory is created because this artefact is
basically attached to the minimisation protocol.
,
superimpose two subsequent snapshots (at 0.3 A inter-
val) of the whole system geometry with significantly
different energies. As an example, Figure 6a shows such
a superposition of the C and D points of the RRmode I
.
Except for the imposed displacement of the substrate,
the only differences come from the side chain orienta-
tions of Leu287 and Leu17. In the latter case, the
enzyme residue is very close to the substrate.
To corroborate these technical and physical explana-
tions, we tried to circumvent this behaviour by locally
exploring this trajectory in the reverse direction. In so
doing, it was expected that, starting from the next
relaxed state (point D), it could be possible to relieve
uphill frustrations generated during the former propa-
gation direction and therefore to decrease the interac-
tion energy (especially for point B). Finally this local
correction was successfully carried out and new lower
energies replaced former ones to better represent the
substrate trajectory. These non-automatic corrections
were systematically performed on both RRmode I and
SRmode I trajectories from each local minima detected
on the previous energy curves (Fig. 4a). These manual
corrections were very efficient and yielded much better
smoothed curves assumed to represent the realistic sub-
strate trajectories (Fig. 7).
From a mechanical point of view, on the one hand, the
backbone of Leu17 is basically fixed and the only
degree of freedom is its side chain rotation. On the
other hand, due to the very close contacts, any tiny
displacement imposed on the substrate would require a
larger Leu17 side chain rotation to avoid steric clashes.
In other words, this figure suggests that the turnstile-
like motion of Leu17 could be especially designed to
monitor the substrate displacement along the access
path. Now, with this schematic mechanism in mind,
one must go back to the protocol background consist-
ing in successive small displacements for the substrate.
It happens that, despite our efforts to fully relax the
system after each distance constraint (combination of
pseudo-dynamics simulations and long minimisation
process), it was technically impossible to relieve all local
steric conflicts as in any optimisation algorithm applied
to a large system with a huge number of internal
parameters. Once more, this is the clear manifestation
of the local false minima problem. To understand this
feature better, a whole cycle of energy fluctuations will
be described in relation with the local side chain rota-
tion. The starting point A is a relaxed energy position
(Leu17 and the substrate are not yet in conflict). At
point B, with a new constraint imposed on one entity
These curves confirm the enzyme enantiopreference
whose mode I is the only acceptable binding mode
because the energy for the faster-reacting enantiomer
(R) is always lower than that of the slower one (S) and
above all, at any step of the trajectory. The energy
difference fluctuates between approximately 3 and 5
kcal/mol. On this basis, PCL appears to be designed to
attract the (R)-enantiomer to form the first tetrahedral
intermediate.
Figure 6. Network of hydrophobic bulky residues. (a) Superimposition of the enzyme/substrate complexes for two subsequent
,
snapshots [points C and D; D(B1–B2): 8.25 and 8.55 A, respectively]. Several side chain orientations of residues Leu287 and Leu17.
(b) Representation of their potentially pivoting side chains with stick bonds. The substrate is shown in CPK representation.

D. Guieysse et al. / Tetrahedron: Asymmetry 14 (2003) 1807–1817
1813
the access path. In other words, even if it is difficult to
precisely describe the action of all the elementary pieces
of this machinery, it clearly appears that this network
of hydrophobic side chains acts in a concerted step–
tooth gear motion. This common network architecture
must be a sophisticated but above all a very efficient
dynamic guiding system.
This simultaneous rotation of several side chains con-
siderably perturbs the minimisation algorithm. It fully
explains the frustrated rotation of one member of this
network when the relaxation process is technically
insufficiently diffused to the other members yielding a
local increase of energy.
Figure 7. Curves of corrected interaction energy between the
enzyme and the substrate versus D(B1–B2)
.
2.2.2.4. Selecting role of the hydrophobic network.
According to binding mode I, the (R)- and (S)-enan-
tiomers must approach the catalytic site with the
bromine atom directed towards the right and left sides,
respectively. Access for the (R)-enantiomer seems much
easier than for the (S) one. Due to its relative orienta-
tion during the final steps just before covalent bond
formation, there is enough room for the displacement
of the (R)-substrate without additional significant inter-
nal distortions of both entities. As seen in Figure 8a,
except for the imposed translation toward the active
site, only small global rotations of the substrate were
observed with limited fluctuations of the hydrophobic
side chains on the enzyme. On the contrary, the bottle-
neck formed by Leu17 and Val266 seemed to hinder the
access of the (S)-enantiomer more severely. As seen in
Figure 8b, to pass through this obstacle with the
bromine atom pointed toward Val266, the (S)-enan-
tiomer trajectory must shift left but much closer to
Leu17. This overcrowded context led to more severe
conflicts with the enzyme, obviously yielding higher
interaction energies that were only partially reduced by
larger enzyme side chain motions (especially for Leu17
as seen in Figure 8b) and also by more significant
substrate internal distortions (although not taken into
account in the interaction energy curve). These two
features are clearly seen in Figure 8b and the steric
obstacle corresponding to Val266 and Leu17 seems to
2.2.2.3. Substrate guiding system. Leu17 is not the
only residue in charge of guiding the substrate from
outside toward the catalytic site. From a static point of
view, examination of the access path of P. cepacia
lipase reveals a dense network of bulky hydrophobic
residues such as valine, leucine or isoleucine. As shown
in Figure 6b, Leu17, Leu167, Leu248, Val266, Leu287,
Leu293 are well distributed all around the path and
could act similarly to Leu17 with their pivoting side
chain. From a dynamic point of view, the turnstile-like
motions of their side chain are illustrated in Figure 8 by
,
superimposing several snapshots (ꢀ5 A<D(B1–B2)<ꢀ8
,
A). For the sake of clarity, only some of them are
shown. Figure 8 has the same orientation as Figure 2a
and the path axis (or B1–B2 axis) is roughly perpendic-
ular to its plane. Finally, along this axis, the side chain
motions occur at different levels. For example Leu17
,
and Val266 interact with the substrate at 5–7 A on the
D(B1–B2) abscissa, and Leu248, Leu287 and Leu293 at
,
7–12 A.
More interestingly, these hydrophobic residues, which
are very close to each other, must have concerted side
chain rotations to avoid intra-molecular steric hin-
drances in addition to inter-molecular ones with the
substrate. Furthermore, all these local motions must be
in phase to monitor the substrate displacement through
Figure 8. Superimposition of the different structures of P. cepacia lipase obtained for the corrected trajectories for the B1–B2
,
distance of 4.75–7.95 A. Left: the (R)-enantiomer, right the (S)-enantiomer.

1814
D. Guieysse et al. / Tetrahedron: Asymmetry 14 (2003) 1807–1817
4. Materials and methods
4.1. Biotransformation
play a key role in the discrimination of the (R)- and
(S)-enantiomers. Thus, in addition to their guiding
role, some of these hydrophobic residues can also
have a selecting role and control which substrate is
allowed to access to the inner catalytic site.
The synthesis of (RS)-2-bromophenyl acetic acid ethyl
ester and the kinetic resolution by P. cepacia lipase
were performed as previously described.11
2.2.3. Enantioselectivity of the Ps-FVL mutant. To fur-
ther highlight the role of the concerted motion of
hydrophobic residues with pivoting side chains and
the importance of the barrier created by Leu17 and
Val266 in the enantiomer discrimination, we tested
the enantioselectivity of the mutant of P. cepacia
lipase: Ps-Phe221Leu-Val266Leu-Leu287Ile.49 Residue
Phe221 is very far from the active site thus, it sounds
reasonable to assume that the F221L mutation has no
influence on the enantioselectivity. On the other hand,
the Val266Leu and Leu287Ile mutations are much
more important because of the substitution of the ini-
tial hydrophobic residues by more bulky ones. This is
especially true for Leu266 facing Leu17 in this geo-
metrical context. Consequently, the size of the bottle-
neck was significantly reduced yielding a dramatic
change in enantioselectivity. The Ps-FVL enantioselec-
tivity toward (RS)-2-bromophenyl acetic acid ethyl
ester was found to be exceptionally high (>200). This
result is thus fully consistent with our explanations
and supports the possible role of the hydrophobic
residues with their pivoting side chains in the discrim-
ination of the (R)- and (S)-enantiomers.
4.1.1. Determination of enantiomeric excess (e.e.), con-
version rate and enantioselectivity (E). HPLC analyses
(carried out as previously described11) were used to
determine the rate of production of each enantiomer
at various concentrations of racemic. From these
analyses, enantiomeric excess (e.e.) was calculated as
defined below: e.e.s={[R]–[S]}s/{[R]+[S]}s (s=sub-
strate) and the conversion rate: C=1–[(R+S)t/(R+
S)t=0]×100. Enantioselectivity value was determined
as the ratio of the initial rate of (R)-enantiomer pro-
duction (viR) versus the initial rate of (S)-enantiomer
production (viS): E=(viR/viS). Initial rates were
determined, before 10% of substrate conversion, by
linear regression over at least five points.
4.2. Molecular modelling
4.2.1. Equipment and software. Calculations, molecular
constructions and graphic displays were performed on
O2 (R10000) Silicon Graphics workstations. All calcu-
lations were done within the InsightII environment
(Accelrys, San Diego, California) using the Discover
package. Potential energies were expressed with the
CFF91 force field implemented in Discover (version
2.97). All minimisations were performed with the
steepest descent algorithm.
3. Conclusions and perspectives
The aim of this study was to understand the enan-
tioselectivity of P. cepacia for (RS)-2-bromophenyl
acetic acid ethyl ester, and to propose a methodology
to explain and eventually predict the enantioselectivity
of P. cepacia lipase using a molecular modelling
approach.
4.2.2. Initial lipase geometry. The X-ray structure of
Pseudomonas cepacia lipase complexed with diethyl
phosphate inhibitor were kindly provided by M.
Cygler and J. Schrag (NRC Biotechnology Research
Center, Montreal). The enzyme coordinates were
compared to those of the native P. cepacia lipase23
(Brookhaven Data Base code: 3LIP). Since no signifi-
cant differences were observed, the coordinates of the
inhibitor/lipase complex were used as template for the
construction of the models.
The concerted side chain rotations of hydrophobic
residues organized in a specific network could be a
general scheme for the enantioselectivity of some
other lipases because of its sophisticated architecture
and at the same time, its efficiency. As a matter of
fact, except for R. miehei lipase, this structural feature
also concerns lipases from C. antarctica B and C.
rugosa. Indeed, these lipases also possess a hindered
active site with a contiguous funnel-like or tunnel-like
access path. As in PCL, a very dense network of
hydrophobic residues lines the sides of the path. Simi-
larly, these close-packed residues could play a major
role in activity and enantioselectivity with their pivot-
ing side chains. The protocol described here, once
adapted to each catalytic environment, should enable
a straightforward examination of the role of these
4.2.3. Construction of the starting geometry. The crys-
tallographic water molecules were deleted from the
X-ray structure, and the hydrogen atoms were added
using the Biopolymer module. The catalytic histidine
š
(His286) was protonated (His ). A covalent bond
was created between the Ser87 oxygen atom and the
phosphorus atom of the inhibitor. The hydrogen
atom positions were first refined (all the other atoms
,
fixed; derivative convergence of 0.01 kcal/A mol).
Then heavy atom positions of the enzyme were
refined using the ‘‘tether’’ option of Discover (con-
straint force of 200 kcal/mol; derivative convergence
hydrophobic residues. More generally, if such
a
,
of 0.5 kcal/A mol).
dynamic scheme was finally found to be an essential
key of enantiorecognition, it would be necessary to
reanalyse all lipase cases, even those for which static
modelling of the tetrahedral intermediate previously
gave results consistent with experiment.
To spare CPU time, calculations were further per-
formed with a subset of amino acids (Table 3) accord-
ing to the following rules: (i) All the 43 residues
delimiting the active site are kept. (ii) If one of these 43

D. Guieysse et al. / Tetrahedron: Asymmetry 14 (2003) 1807–1817
1815
Table 3. Definition of the truncated Pseudomonas cepacia
lipase
the substrate and the second letter to the stereogenic
centre created by the intermediate formation. For each
of
these
combinations,
dihedral
angles
Segments of residues lining the Amino acids selecteda
active cavity (catalytic
site+access path)
FX(C1ꢀC2ꢀC7ꢀC8) and FY(C2ꢀC7ꢀC8ꢀO9) represent
the two major degrees of flexibility (Scheme 1). There-
fore, they were systematically scanned by 10° incre-
ments, fixing all the other internal parameters (ROTOR
option of Discover) to roughly locate energy minima.
These minima solutions were further fully refined (no
constraints; derivative convergence of 0.01 kcal/A mol)
in vacuo and without any steric consideration of the
active site topography.
*G16, L17, T18, Y23, L27,
E28, Y29
-14-15-…-21-22-23-24-25-26-27-28
-29-…-32-33
*F52
-46-47-…-60-61-62
,
*H86, S87, Q88
-85-86-87-88-89-90-91
-110-111-…-117-118-…-127-128-…
-133-134-…-150
*G111, T112, P113, G116,
S117, F119, A120, V123,
L127, I139, F142, V143,
F146, T150
Second stage. All these core geometries were then
‘‘transposed’’ into the active site by superimposing the
Ser87 cartesian coordinates of modelled cores to those
of the crystallized complex to obtain the truncated
tetrahedral intermediate geometries. Clearly, this brutal
technical transposition generally provokes initial huge
(and somewhat artificial) steric clashes that could be
solved in some cases by the local rearrangement of a
limited number of atoms. Therefore, each tetrahedral
intermediate geometry was submitted to three succes-
sive minimisations with a decreasing number of con-
straints combined with an increasing number of
iterations to manage, as properly as possible, the side
chain atom relaxation. The first minimisation consisted
of 2000 iterations with all heavy atoms fixed except
those of the catalytic core. In so doing, it was expected
(if necessary or possible) to reposition the core slightly.
During the second minimisation (10,000 iterations), the
protein backbone was fixed and the enzyme side chains
were allowed to vary in order to remove local conflicts
between the protein and the branched core. Finally,
during the third minimisation (50,000 iterations), only
the backbone of the residues belonging to secondary
elements was fixed. This corresponds to the largest
relaxation preserving the general shape of this trun-
cated system. Finally, after the successive minimisation
only two solutions were kept according the geometric
features and the energy criterion.
*A160, L161, A163, L164,
L167
*P243, S244, A247, L248,
T251, V266, V267
*H286, L287, I290, Q292,
L293, L294
-160-…-166-167-168-169-170
-241-242-243-244-…-257-258-…
-268-269
-282-283-…-287-288-289-290-291-
…-303-304
a See the rules defined in the text; the backbone of the amino acids
belonging to the secondary element (in bold) is kept fixed during the
calculations.
residues belongs to a secondary element, one or two
adjacent residues along the primary structure are also
selected. (iii) If the amino acids forming the active site
are localized inside a loop, all the residues of this loop
are conserved as well as the extremities of the loop
belonging to the adjacent secondary elements. In total,
this subset represents 46% of the whole protein struc-
ture. Consequently, this artefactual truncation imposed,
for all following calculations, a basic set of constraints
with fixed backbone atoms for residues belonging to
secondary elements to preserve the global folding of the
protein, regardless of other additional constraints nec-
essary for the selected strategy.
4.2.4. First protocol: modelling of the first tetrahedral
intermediate. Basically, the tetrahedral intermediate
geometries were built in two distinct stages. Primarily, a
systematic conformational search of the core entity (the
catalytic serine Ser87 covalently bound with (R)- and
(S)-2-bromophenyl acetic acid ethyl ester) was per-
formed, regardless of the enzyme active site context.
Then, all acceptable geometries found in this former
stage were ‘transposed’ within the active site environ-
ment. The goal of this protocol is to describe, as
systematically as possible, all conformations with
regards to energy as well as catalytic topography
criteria.
4.2.5. Second protocol: modelling of the substrate path in
the catalytic cavity. Starting geometries. As in the first
approach, the truncated tetrahedral intermediates previ-
ously built were used to spare CPU time. Firstly, it was
necessary to break the covalent bond between the cata-
lytic serine and the carbon of the substrate carbonyl
function to create two molecular entities with distinct
topologies: the truncated enzyme (encompassing Ser87)
and the substrate. Two virtual points (local barycen-
ters) were defined to represent these entities. The first
one, B1, based on the backbone atoms of the catalytic
serine is attached to the heaviest entity (truncated
enzyme). The second one B2, is based on atoms (C2,
C7, C8, O9) and represents the substrate (Scheme 1).
First stage. The diethyl phosphate inhibitor coordinates
and those of S87 in the complex crystallographic struc-
ture were initially used to build the catalytic core. The
formation of the tetrahedral intermediate leads to the
generation of a second stereogenic carbon, thus with
each enantiomers, two configurations can be obtained.
A total of four combinations are thus possible and were
referred as: RR, RS, SR and SS in the text. The first
letter (R or S) refers to the initial stereogenic centre of
Trajectory protocol. PCL active site is a well defined
contiguous channel. The particular shape of this chan-
nel, like a funnel, yielded a specific trajectory protocol.
Formally, the absolute location of the substrate in the
enzyme reference should need 6 parameters (e.g. 3

1816
D. Guieysse et al. / Tetrahedron: Asymmetry 14 (2003) 1807–1817
rotations and 3 translations). But in the present con-
text, the simplest way to monitor the relative displace-
ment of these two entities was to impose regular
incremental variations of B1ꢀB2 distance (DB1–B2) all
through a series of calculations to describe the accessi-
bility of the substrate to the active site. It must be
pointed out that, in this scheme, the 6 positioning
parameters were replaced by DB1–B2. This drastic reduc-
tion of the conformation space to explore is acceptable
only because of the narrow enzyme path, especially
near the catalytic site. The use of a similar dynamic
protocol in another geometrical context (e.g. a large
access path with enough room for substrates to have
internal rotations) would certainly lead to useless and
random results.
vokes huge deformations of the smallest entity during
the calculations (dynamics simulations and minimisa-
tion as well). Therefore, additional constraints on dihe-
drals FX and FY were also imposed to preserve the
global shape of the substrate and to technically coun-
terbalance the high mass of the enzyme and the con-
straints imposed on it.
All along the trajectory, the interaction energy between
the enzyme and the substrate was the first criterion
taken into account to explain the enantioselectivity as
in the following section. However, the potential energy
variations of the enzyme and substrate were constantly
checked to ensure a quite realistic representation of the
trajectory. In practice, a momentary increase of one of
these two terms would generally mean an artificial
deformation of the corresponding entity due to insuffi-
cient relaxation of the whole system. Finally, for the
sake of clarity, the interaction energies were reported
In our protocol, only the final coordinates are precisely
known (formation of the tetrahedral intermediate)
which corresponds to the shortest D(B1–B2) value (4.55
,
,
only every 0.3 A although calculations were performed
A). Due to the large entrance of the access path, any
,
with increments of 0.1 A.
arbitrary choice of initial substrate orientation would
undoubtedly guide the subsequent trajectory performed
in a chronological model (from outside into the cata-
lytic site) without any guarantee that such simulations
are the most significant. Thus, it was technically easier
to generate ‘‘reverse’’ substrate trajectories by progres-
sively increasing D(B1–B2). From visual considerations of
the enzyme access path, we chose a maximum distance
Acknowledgements
The authors thank Dr. Hirose (Amano Co., Japan) for
providing the Ps-FVL mutant of Pseudomonas cepacia;
Dr. Miroslaw Cygler and Dr. Joseph Schrag (NRC
Biotechnology Research Center, Montreal) for access
to the X-ray crystallographic coordinates of PCL com-
plex; and Dr. Lorenzo Di Bari and Dr. Gennaro
Pescitelli (Dipartimento di Chimica e Chimica Industri-
ale, Pisa, Italy) for the determination of the absolute
configurations of a-bromo phenyl acetic acid by the
Circular Dichroism Exciton method.
,
of 12.55 A corresponding to a substrate position inside
the access path with enough room to potentially avoid
steric conflicts with the enzyme. In fact, the first steps
of this increasing D(B1–B2) protocol are much more
physically meaningful (up to a certain distance to be
defined further) since it can specifically reveal the
residues responsible for the steric conflicts. Then, due to
the funnel-like shape of the access path, steric hin-
drances will progressively vanish for larger D(B1–B2)
values.
References
,
Within the total range of variation (7.5 A), regular
increments of 0.1 A were imposed. This increment was
,
1. Bornscheuer, U. T.; Bessler, C.; Srinivas, R.; Krishna, S.
H. Trends Biotechnol. 2002, 20, 433–437.
chosen short enough to temporarily relax the system
until the following increment step. In relation with this
short substrate displacement, a sophisticated elemen-
tary sequence of calculations was chosen to relax, as
much as possible, the total molecular system. At each
distance constraint, molecular dynamics simulations
were performed (300K during 1 ps) and the enzyme/
substrate entity coordinates were stored every 100 fs.
These calculations were used as an exploration tech-
nique of all internal parameters of the system (anneal-
ing simulations but at room temperature). Finally, the
best of each 10 conformations was selected for subse-
quent minimisation (10,000 iterations). During the
dynamics simulations and the minimisation stage, D(B1–
B2) was highly constrained with a specific harmonic
term a(r−r0) with a constraint force of 1000 kcal/A .
Initially, only the enzyme side chains were allowed to
fluctuate (to preserve the general shape of the access
path) while no constraint was imposed on the substrate.
Unfortunately, spurious substrate deformations were
always observed whose origin is the mass ratio between
the enzyme and the substrate which artificially pro-
2. Reetz, M. T. Curr. Opin. Chem. Biol. 2002, 6, 145–150.
3. Schmid, A.; Dordick, J. S.; Hauer, B.; Kiener, A.; Wub-
bolts, M.; Witholt, B. Nature 2001, 409, 258–268.
4. Klibanov, A. M. Nature 2001, 409, 241–246.
5. Berglund, P. Biomol. Eng. 2001, 18, 13–22.
6. Schmidt, R. D.; Verger, R. Angew. Chem., Int. Ed. Engl.
1998, 37, 1608–1633.
7. Kamal, A.; Sandbhor, M.; Ramana, K. V. Tetrahedron:
Asymmetry 2002, 13, 815–820.
8. Salunkhe, M. M.; Nair, R. V. Enzyme Microb. Technol.
2001, 28, 333–338.
9. Pchelka, B. K.; Loupy, A.; Plenkiewicz, J.; Petit, A.;
Blanco, L. Tetrahedron: Asymmetry 2001, 12, 2109–2119.
10. Lundell, K.; Raijola, T.; Kanerva, L. T. Enzyme Microb.
Technol. 1998, 22, 86–93.
11. Guieysse, D.; Salagnad, C.; Monsan, P.; Remaud-
Simeon, M. Tetrahedron: Asymmetry 2003, 14, 317–323.
12. Haughton, L.; Williams, J. M. J. Synthesis 2001, 6,
943–946.
2
2
,
13. Tranel, F.; Haufe, G. Tetrahedron: Asymmetry 2000, 11,
889–896.

D. Guieysse et al. / Tetrahedron: Asymmetry 14 (2003) 1807–1817
1817
Sunjic, V.; Vitale, L.; Saenger, W.; Kojic-Prodic, B. Eur.
J. Biochem. 2001, 268, 3964–3973.
14. Miyazawa, T.; Kurita, S.; Ueji, S.; Yamada, T. Biocatal.
Biotrans. 2000, 17, 459–473.
33. Lang, D. A.; Mannesse, M. L. M.; De Haas, G. H.;
Verheij, H. M.; Dijkstra, B. W. Eur. J. Biochem. 1998,
254, 333–340.
15. Sundholm, O.; Kanerva, L. T. Models Chem. 1998, 135,
625–640.
16. Jones, M. M.; Williams, M. J. Chem. Commun. 1998,
2519–2520.
17. Palomer, A.; Cabre, M.; Ginesta, J.; Mauleon, D.; Car-
ganico, G. Chirality 1993, 5, 320–328.
18. O’Hagan, D.; Rzepa, H. S. J. Chem. Soc., Perkin Trans.
2 1994, 3–4.
19. Kalaritis, P.; Regenye, R. W.; Partridge, J. J.; Coffen, D.
L. J. Org. Chem. 1990, 55, 812–815.
20. Kazlauskas, R. J. Science 2001, 293, 2277–2279.
21. Kazlauskas, R. J. Curr. Opin. Chem. Biol. 2000, 4, 81–88.
22. Kim, K. K.; Song, H. K.; Shin, D. H.; Hwang, K. Y.;
Suh, S. W. Structure 1997, 5, 173–185.
23. Schrag, J. D.; Li, Y.; Cygler, M.; Lang, D.; Burgdorf, T.;
Hecht, H.-J.; Schmid, R.; Schomburg, D.; Rydel, T. J.;
Oliver, J. D.; Strickland, L. C.; Dunaway, C. M.; Larson,
S. B.; Day, J.; McPherson, A. Structure 1997, 5, 187–202.
24. Grochulski, P.; Bouthillier, F.; Kazlauskas, R. J.; Serreqi,
A. N.; Schrag, J. D.; Ziomek, E.; Cygler, M. Biochemistry
1994, 33, 3494–3500.
25. Uppenberg, J.; Ohrner, N.; Norin, M.; Hult, K.; Kley-
wegt, G. J.; Patkar, S.; Waagen, V.; Anthonsen, T.;
Jones, T. A. Biochemistry 1995, 34, 16838–16851.
26. Uppenberg, J.; Hansen, M. T.; Patkar, S.; Jones, T. A.
Structure 1994, 2, 293–308.
34. Cygler, M.; Grochulski, P.; Kazlauskas, R. J.; Schrag, J.
D.; Bouthillier, F.; Rubin, B.; Serreqi, A. N.; Gupta, A.
K. J. Am. Chem. Soc. 1994, 116, 3180–3186.
35. Kazlauskas, R. J.; Weissfloch, A. N. E.; Rappaport, A.
T.; Cuccia, L. A. J. Org. Chem. 1994, 56, 2656–2665.
36. Ahmed, S. N.; Kazlauskas, R. J.; Morinville, A. H.;
Grochulski, P.; Schrag, J. D.; Cygler, M. Biocatalysis
1994, 9, 209–225.
37. Haeffner, F.; Norin, T.; Hult, K. Biophys. J. 1998, 74,
1251–1262.
38. Orrenius, C.; van Heusden, C.; van Ruiten, J.; Over-
beeke, P. L. A.; Kierkels, H.; Duine, J. A.; Jongejan, J.
A. Protein Eng. 1998, 11, 1147–1153.
39. Norin, M.; Haeffner, F.; Adnane, A.; Norin, T.; Hult, K.
Protein Sci. 1994, 3, 1493–1503.
40. Schulz, T.; Pleiss, J.; Schmid, R. D. J. Mol. Model. 2001,
7, 265–270.
41. Schulz, T.; Pleiss, J.; Schmid, R. D. Protein Sci. 2000, 9,
1053–1062.
42. Berglund, P.; Vallikivi, I.; Fransson, L.; Dannacher, H.;
Holmquist, M.; Martinelle, M.; Bjo¨rkling, F.; Parve, O.;
Hult, K. Tetrahedron: Asymmetry 1999, 10, 4191–4202.
43. Berglund, P.; Holmquist, M.; Hult, K. J. Mol. Catal. B:
Enzym. 1998, 5, 283–287.
27. Brady, L.; Brzozowski, A. M.; Derewenda, Z. S.; Dod-
son, E.; Dodson, G.; Tolley, S.; Turkenburg, J. P.; Chris-
tiansen, L.; Huge-Jensen, B.; Norskov, L.; Thim, L.;
Menge, U. Nature 1990, 343, 767–770.
44. Holmquist, M.; Haeffner, F.; Norin, T.; Hult, K. Protein
Sci. 1996, 5, 83–88.
45. Tuomi, W. V.; Kazlauskas, R. J. J. Org. Chem. 1999, 64,
2638–2647.
28. Hu, C. H.; Brinck, T.; Hult, K. Int. J. Quantum Chem.
1997, 69, 89–103.
29. Warshel, A.; Naray-Szabo, G.; Sussman, F.; Hwang,
J.-K. Biochemistry 1989, 28, 3629–3637.
30. Chapus, C.; Se´me´riva, M. Biochemistry 1976, 15, 4988–
4991.
31. Chapus, C.; Se´me´riva, M.; Bovier-Lapierre, C.;
Desnuelle, P. Biochemistry 1976, 15, 4980–4987.
32. Luic, M.; Tomic, S.; Lescic, I.; Ljubovic, E.; Sepac, D.;
46. Tafi, A.; van Almsick, A.; Corelli, F.; Crusco, M.; Lau-
men, K. E.; Schneider, M. P.; Botta, M. J. Org. Chem.
2000, 65, 3659–3665.
47. Haeffner, F.; Norin, T. Chem. Pharm. Bull. 1999, 47,
591–600.
48. Yagnik, A. T.; Littlechild, J. A.; Turner, N. I. J. J.
Computer-Aided Mol. Design 1997, 11, 256–264.
49. Hirose, Y.; Kariya, K.; Nakanishi, Y.; Kurono, Y.;
Achiwa, K. Tetrahedron Lett. 1995, 36, 1063–1066.