Molecular basis for enantioselectivity of lipase from Pseudomonas cepacia toward primary alcohols. Modeling, kinetics, and chemical modification of Tyr29 to increase or decrease enantioselectivity

Article abstract of DOI:10.1021/jo981783y

Lipase from Pseudomonas cepacia (PCL) shows good enantioselectivity toward primary alcohols. An empirical rule can predict which enantiomer of a primary alcohol reacts faster, but there is no reliable strategy to increase the enantioselectivity. We used a combination of molecular modeling of lipase-transition state analogue complexes and kinetic measurements to identify the molecular basis of the enantioselectivity toward two primary alcohols: 2-methyl-3-phenyl-1-propanol, 1, and 2-phenoxy-1-propanol, 2. In hydrolysis of the acetate esters, PCL favors the (S)-enantiomer of both substrates (E = 16 and 17, respectively), but, due to changes in priorities of the substituents, the (S)-enantiomers of 1 and 2 have opposite shapes. Computer modeling of transition state analogues bound to PCL show that primary alcohols bind to PCL differently than secondary alcohols. Modeling and kinetics suggest that the enantioselectivity of PCL toward 1 comes from the binding of the methyl group at the stereocenter within a hydrophobic pocket for the fast-reacting enantiomer, but not for the slow-reacting enantiomer. On the other hand, the enantioselectivity toward 2 comes from an extra hydrogen bond between the phenoxy oxygen of the substrate to the phenolic OH of Tyr29. This hydrogen bond may slow release of the (R)-alcohol and thus account for the reversal of enantioselectvity. To decrease the enantioselectivity of PCL toward 2-acetate by a factor of 2 to E = 8, we eliminated the hydrogen bond by acetylation of the tyrosyl residues with N- acetylimidazole. To increase the enantioselectivity of PCL toward 2-acetate by a factor of 2 to E = 36, we increased the strength of the hydrogen bond by nitration of the tyrosyl residues with tetranitromethane. This is one of the first examples of a rationally designed modification of a lipase to increase enantioselectivity.

Full text of DOI:10.1021/jo981783y

2638
J . Org. Chem. 1999, 64, 2638-2647
Molecu la r Ba sis for En a n tioselectivity of Lip a se fr om
P seu d om on a s cepa cia tow a r d P r im a r y Alcoh ols. Mod elin g,
Kin etics, a n d Ch em ica l Mod ifica tion of Tyr 29 to In cr ea se or
Decr ea se En a n tioselectivity
W. Victor Tuomi and Romas J . Kazlauskas*
McGill University, Department of Chemistry, 801 Sherbrooke Street West,
Montre´al, Que´bec H3A 2K6 Canada
Received September 1, 1998 (Revised Manuscript Received J anuary 25, 1999)
Lipase from Pseudomonas cepacia (PCL) shows good enantioselectivity toward primary alcohols.
An empirical rule can predict which enantiomer of a primary alcohol reacts faster, but there is no
reliable strategy to increase the enantioselectivity. We used a combination of molecular modeling
of lipase-transition state analogue complexes and kinetic measurements to identify the molecular
basis of the enantioselectivity toward two primary alcohols: 2-methyl-3-phenyl-1-propanol, 1, and
2-phenoxy-1-propanol, 2. In hydrolysis of the acetate esters, PCL favors the (S)-enantiomer of both
substrates (E ) 16 and 17, respectively), but, due to changes in priorities of the substituents, the
(S)-enantiomers of 1 and 2 have opposite shapes. Computer modeling of transition state analogues
bound to PCL show that primary alcohols bind to PCL differently than secondary alcohols. Modeling
and kinetics suggest that the enantioselectivity of PCL toward 1 comes from the binding of the
methyl group at the stereocenter within a hydrophobic pocket for the fast-reacting enantiomer,
but not for the slow-reacting enantiomer. On the other hand, the enantioselectivity toward 2 comes
from an extra hydrogen bond between the phenoxy oxygen of the substrate to the phenolic OH of
Tyr29. This hydrogen bond may slow release of the (R)-alcohol and thus account for the reversal
of enantioselectvity. To decrease the enantioselectivity of PCL toward 2-acetate by a factor of 2 to
E ) 8, we eliminated the hydrogen bond by acetylation of the tyrosyl residues with N-
acetylimidazole. To increase the enantioselectivity of PCL toward 2-acetate by a factor of 2 to E )
36, we increased the strength of the hydrogen bond by nitration of the tyrosyl residues with
tetranitromethane. This is one of the first examples of a rationally designed modification of a lipase
to increase enantioselectivity.
In tr od u ction
secondary alcohols, Figure 1, predicts which enantiomer
reacts faster based on the size of the substituents at the
stereocenter.2 This rule implies that lipases recognize
enantiomers based on the size of the substituents. Indeed,
researchers have increased the enantioselectivity of a
reaction by increasing the size of the large substituent.3
As shown in Figure 1, this rule is only qualitative, but
researchers have also more precisely defined the sizes of
the medium and large substituents using either box
models4 or comparative molecular field analysis (CoM-
FA).5
X-ray crystallographers have solved the structures of
more than 10 different lipases, including some containing
bound transition state analogues.6 These structures show
an alcohol binding site with two pockets consistent with
the empirical rule in Figure 1. One pocket is large, lined
Molecular recognition is the key to current problems
in chemistry and biology. Stereoselective reagents rec-
ognize correct substrate orientation in the transition
state. Drug action, cell-cell recognition, and receptor
target recognition all rely on molecular recognition by
proteins. Stereoselective and regioselective enzyme-
catalyzed reactions also rely on molecular recognition and
selective stabilization of transition states by proteins.
This paper focuses on the ability of a lipase from
Pseudomonas cepacia (PCL) to distinguish between enan-
tiomers of primary alcohols. Enantioselectivity is easier
to study than diastereoselectivity or other types of
selectivity because the physical properties of enantiomers
are identical. Thus, enantioselectivity stems directly from
molecular recognition, not from differences in solubility,
reactivity, etc., which may be the case for diastereomers.
Organic chemists use lipases such as PCL to resolve
enantiomers and to selectively protect and deprotect
synthetic intermediates.1 To guide synthetic applications,
chemists often use empirical rules, which summarize the
known selectivities of lipases. An empirical rule for
(2) Kazlauskas, R. J .; Weissfloch, A. N. E.; Rappaport, A. T.; Cuccia,
L. A. J . Org. Chem. 1991, 56, 2656-2665 and references therein.
(3) Scilimati, A.; Ngooi, T. K.; Sih, C. J . Tetrahedron Lett. 1988, 29,
4927-4930. J ohnson, C. R.; Ple`, P. A.; Adams, J . P. J . Chem. Soc.,
Chem. Commun. 1991, 1006-1007. Goergens, U., Schneider, M. P. J .
Chem. Soc., Chem. Commun. 1991, 1064-1066. Ibid. 1066-1068. Kim,
M. J .; Choi, Y. K. J . Org. Chem. 1992, 57, 1605-1607. Gupta, A. K.;
Kazlauskas, R. J . Tetrahedron: Asymmetry 1993, 4, 879-888. Rotticci,
D., Orrenius, C., Hult, K., Norin, T. Tetrahedron: Asymmetry 1997,
8, 359-362.
(4) Pseudomonas fluorescens lipase: Burgess, K.; J ennings, L. D.
J . Am. Chem. Soc. 1991, 113, 6129-6139. Naemura, K.; Fukuda, R.;
Konishi, M.; Hirose, K.; Tobe, Y. J . Chem. Soc., Perkin Trans. 1 1994,
1253-1256. Naemura, K.; Fukuda, R.; Murata, M.; Konishi, M.; Hirose,
K.; Tobe, Y. Tetrahedron: Asymmetry 1995, 6, 2385-2394.
(1) Kazlauskas, R. J .; Bornscheuer, U. T. Biotechnology, 2nd ed.;
Kelly, D. R., Ed.; Wiley-VCH: Weinheim, 1998; Vol. 8a, pp 37-191;
Faber, K. Biotransformations in Organic Chemistry, 3rd ed.; Springer:
Berlin, 1997. Wong, C.-H.; Whitesides, G. M. Enzymes in Synthetic
Organic Chemistry; Pergamon: New York, 1994. Roberts, S. M., Ed.
Preparative Biotransformations, Wiley: New York, 1992-1998.
10.1021/jo981783y CCC: $18.00 © 1999 American Chemical Society
Published on Web 03/31/1999

Enantioselectivity of Pseudomonas cepacia Lipase
J . Org. Chem., Vol. 64, No. 8, 1999 2639
In this paper, we use computer modeling to investigate
the molecular basis of enantiomer recognition by PCL.
We will address three questions. First, the models for
primary alcohols and secondary alcohols suggest a re-
versal of enantiopreference for PCL. In the fast-reacting
enantiomer of secondary alcohols, the hydroxyl group
faces toward the reader, but for the fast-reacting enan-
tiomer of primary alcohols, the CH2OH group points into
the plane of the paper. Computer modeling will show that
the large substituents of primary and secondary alcohols
bind in different regions of PCL. This difference in
binding accounts for the reversal of enantiopreference.
F igu r e 1. Empirical rules summarize the enantiopreference
of PCL toward chiral alcohols. (a) Shape of the favored
enantiomer of secondary alcohols. M represents a medium-
sized substituent, e.g., CH3, and L represents a large substitu-
ent, e.g., Ph. (b) Shape of the favored enantiomer of primary
alcohols. This rule for primary alcohols is reliable only when
the stereocenter lacks an oxygen atom. Note that the OH of
secondary alcohols points toward the reader, while the CH2OH
of primary alcohols points away from the reader.
The second question is why primary alcohols with an
oxygen at the stereocenter do not follow the empirical
rule. Of the 27 substrates collected in an earlier paper,
only 10 fit the empirical rule; 31% agreement.8 This is
worse than guessing which gives 50% and suggests that
this class of primary alcohols may follow an opposite
empirical rule. The computer modeling will identify a
hydrogen bond that can reverse the enantioselectivity for
some primary alcohols with an oxygen at the stereo-
center.
with hydrophobic side chains and open to the solvent.
The other pocket is medium-sized and contains polar as
well as hydrophobic side chains.
Researchers used computer modeling of transition
state analogues bound to lipases to confirm the molecular
recognition model suggested in Figure 1.7,5b This model-
ing starts with the X-ray structure of the lipase, docks
into it a tetrahedral intermediate or transition state
analogue, and minimizes the energy of the complex. A
productive complex is one that has all the hydrogen bonds
required for catalysis and a good fit within the active site.
Such modeling usually correctly predicts the fast-reacting
enantiomer, but quantitative prediction of enantioselec-
tivity remains difficult. The slow enantiomer either does
not fit as well into the active site or fits in a nonproduc-
tive manner.
Understanding the molecular recognition of primary
alcohols has been more difficult. First, most lipases show
low enantioselectivity toward primary alcohols. Only
lipase from Pseudomonas cepacia (PCL) and lipase from
porcine pancreas (PPL) show moderate to high enanti-
oselectivity toward a wide range of primary alcohols, but
even for these the enantioselectivity is usually lower than
toward secondary alcohols. Second, a simple rule based
on the size of the substituents cannot predict the favored
enantiomer for all primary alcohols. For PPL, researchers
working with different substrates proposed opposite,
enantiomeric rules! Of course, neither rule predicted the
favored enantiomer for all substrates. For PCL, we found
that a simple rule works if we exclude primary alcohols
with an oxygen at the stereocenter, Figure 1.8
The last and most challenging question is how to
increase the enantioselectivity of PCL toward primary
alcohols. For secondary alcohols, increasing the difference
in size of the substituents usually increases the enanti-
oselectivity of lipase. However, previous work showed
that this approach was not reliable for PCL-catalyzed
reactions of primary alcohols. Sometimes it had no effect,
sometimes it increased, and sometimes it decreased the
enantioselectivity. The modeling below suggests a way
to increase the enantioselectivity toward some alcohols
by chemical modification to increase the strength of a key
hydrogen bond.
Although the goal of this paper is to understand the
molecular recognition of enantiomers of 1 and 2 by PCL,
pure enantiomers of 1 and 2 are also useful for synthesis.
For example, Delnick and Margolin noted that pure
enantiomers of 1 would be useful for the synthesis of
adenosine receptor agonists and antagonists,9 while
Heathcock and co-workers used (R)-1 in the synthesis of
the side chain of zaragozic acid A (squalestatin S1).10
Wimmer et al. used pure enantiomers of 2 to prepare
analogues of insect juvenile hormones,11 and Dirlam et
al. used a p-fluoro derivative to make sorbinil, an aldose
reductase inhibitor.12
(5) (a) Rhizomucor miehei lipase: Ranghino, G., Battistel, E.,
Giovenco, S. Trends QSAR Mol. Modell. 92, Proc. Eur. Symp. Struct.-
Act. Relat.: QSAR Mol. Modell., 9th, 1992 (1993) (Wermuth, C.-G.,
Ed.; ESCOM: Leiden, NL) 1992, 373-378. (b) Candida rugosa
lipase: Faber, K.; Griengl, H.; Hoenig, H.; Zuegg, J . Biocatalysis 1994,
9, 227-239. (c) Similar approach for Pseudomonas cepacia lipase:
Lemke, K., Lemke, M., Theil, F. J . Org. Chem. 1997, 62, 6268-6273.
(6) Reviews: Dodson, G. G.; Lawson, D. M.; Winkler, F. K. Faraday
Discuss. 1992, 95-105. Derewenda, Z. S. Adv. Prot. Chem. 1994, 45,
1-52. Cygler, M.; Grochulski, P.; Schrag, J . D. Can. J . Microbiol. 1995,
41, 289-96. Kazlauskas, R. J . Trends Biotechnol. 1994, 12, 464-72
(errata 1994, 13, 195).
All computer modeling in this paper begins with the
X-ray crystal structure of the open conformation of PCL,
which was recently reported by four groups.13 Like other
lipases, PCL is an R/â hydrolase.14 Ser87, His286, and
Asp264 form the catalytic triad, while the amide hydro-
gens of Leu17 and Gln88 also contribute to catalysis by
hydrogen bonding to the oxyanion intermediate.
(7) Secondary alcohols, HLL, RML, CRL: Norin, M.; Haeffner, F.;
Achour, A.; Norin, T.; Hult, K. Prot. Sci. 1994, 3, 1493-1503. Sainz-
D´ıaz, C. I., Wohlfahrt, G., Nogoceke, E., Herna´ndez-Laguna, A.,
Smeyers, Y. G., Menge, U. Theochem, J . Mol. Struct. 1997, 390, 225-
237. Parve, O., Vallikivi, I., Metsala, A., Lille, U., Tougu, V., Sikk, P.,
Kaambre, T., Vija, H., Pehk, T. Tetrahedron 1997, 53, 4889-4900.
Ema, T., Kobayashi, J ., Maeno, S., Sakai, T., Utaka, M. Bull. Chem.
Soc. J pn. 1998, 71, 443-453. Faber, K.; Griengl, H.; Hoenig, H.; Zuegg,
J . Biocatalysis 1994, 9, 227-239. CAL-B: Uppenberg, J .; O¨ hrner, N.;
Norin, M.; Hult, K.; Patkar, S.; Waagen V.; Anthonsen, T.; J ones, T.
A. Biochemistry 1995, 34, 16838-16851. Haeffner, F., Norin, T., Hult,
K. Biophys. J . 1998, 74, 1251-1262.
(8) Weissfloch, A. N. E.; Kazlauskas, R. J . J . Org. Chem. 1995, 60,
6959-6969.
(9) Delinck, D. L., Margolin, A. L. Tetrahedron Lett. 1990, 31, 6797-
6798.
(10) Stoermer, D., Caron, S., Heathcock, C. H. J . Org. Chem. 1996,
61, 9115-9125.
(11) Wimmer, Z., Saman, D., Francke, W. Helv. Chim. Acta 1994,
77, 502-508.
(12) Dirlam, N. L., Moore, B. S., Urban, F. J . J . Org. Chem. 1987,
52, 3587-3591.

2640 J . Org. Chem., Vol. 64, No. 8, 1999
Tuomi and Kazlauskas
Sch em e 1
Ta ble 1. En a n tioselectivity of P CL a n d Acetyla ted P CL
tow a r d Aceta tes of 1 a n d 2
substrate
enzyme
eep, %
ees, %
Ea
1-acetate
2-acetate
2-acetate
2-acetate
2-acetate
PCL
PCL
79
73
40
50
94
nd
86
92
89
24
16 (S)b
17 (S)c
36 (S)
8 (S)
PCL-nitratedd
PCL-acetylatede
PCL-deacetylatedf
42 (S)
a
The enantiomeric ratio measures the relative rate of hydrolysis
of the fast enantiomer as compared to the slow enantiomer as
defined by Sih (Chen, C. S.; Fujimoto, Y.; Girdaukas, G.; Sih, C.
J . J . Am. Chem. Soc. 1982, 104, 7294-7299). The absolute
configuration of the favored enantiomer is indicated in parenthe-
ses. b Data from ref 8. c The absolute configuration was determined
by chemical correlation. The product alcohol was oxidized using
J ones reagent, and the HPLC chromatogram (Chiracel OD column)
of the resulting acid was compared to that previously reported (ref
F igu r e 2. Intermediates and their analogues in the lipase-
catalyzed hydrolysis of esters. Amino acids correspond to those
in PCL. (a) In the accepted mechanism for hydrolysis of esters,
the catalytic serine (Ser 87) attacks the ester once it binds in
the active site. This attack forms the tetrahedral intermediate
marked 1. Hydrogen bonds from two amide N-H's stabilize
the oxyanion in this intermediate. Collapse of tetrahedral
intermediate 1 releases the alcohol and forms an acyl enzyme
intermediate. The formation and collapse of tetrahedral in-
termediate 1 determine the selectivity of a lipase toward the
alcohol moiety of an ester. Attack of water on the acyl enzyme
yields tetrahedral intermediate 2. Collapse releases the acid
and regenerates the free enzyme. (b) A phosphonate mimics
the transition states for formation and collapse of tetrahedral
intermediate 1. Five hydrogen bonds, marked a-e, stabilize
the charges. To be judged catalytically competent, the mini-
mized structures must contain all five hydrogen bonds.
d
35). PCL was chemically modified by nitration with tetrani-
tromethane. e PCL was chemically modified by acetylation with
N-acetylimidazole. We estimate that 10 of the 14 tyrosine residues
were acetylated. f The sample of acetylated PCL was treated with
hydroxylamine to remove all O-acetyl groups.
Resu lts
En a n tioselectivity a n d Rever sed En a n tiop r efer -
en ce. Lipase from Pseudomonas cepacia (PCL)15 cata-
lyzes the enantioselective hydrolysis of the acetate esters
of 1 and 2, Scheme 1, Table 1. The enantioselectivity
(relative rate of hydrolysis of the fast enantiomer as
compared to the slow enantiomer) is moderate for both
acetates: 16 for 1-acetate and 17 for 2-acetate. In both
cases, PCL favors the (S)-enantiomer, but, due to changes
in the priorities of the substituents, the (S)-enantiomers
of 1 and 2 have the opposite shapes. The fast-reacting
enantiomer of 1 has the shape predicted by the rule in
Figure 1b. The shape of the fast-reacting enantiomer of
2 does not fit the rule in Figure 1b. This lack of
agreement for 2 is not surprising since we previously
reported that the rule is not reliable for primary alcohols
such as 2 that contain an oxygen at the stereocenter.
Con for m a tion a l An a lysis of Un bou n d Su bstr a tes.
To find the reason for the opposite enantiopreference of
PCL toward 1- and 2-acetate, we first compared the
ground state conformations of the two substrates (details
in Supporting Information). These calculations show that
both 1- and 2-acetate are flexible without a well-defined
conformation in solution. Both substrates easily adopt
similar conformations. Thus, conformational preferences
of the substrates do not cause the reversal in enantiose-
lectivity of PCL. Different interactions of 1- and 2-acetate
with the active site of PCL likely cause this reversal.
Molecu la r Mod elin g of Tr a n sition Sta te An a -
logu es Bou n d to P CL. The first chemical step in lipase-
catalyzed ester hydrolysis is the attack of the active site
serine at the ester carbonyl forming a tetrahedral inter-
mediate, Figure 2a. Collapse of this tetrahedral inter-
mediate releases the alcohol. The transition state in-
volved in formation or collapse of this first tetrahedral
intermediate defines the selectivity of PCL toward alco-
hols. To mimic this transition state for hydrolysis of
butyrate esters of 1 or 2, we used the phosphonates
linked to PCL shown in Figure 2b. Indeed, researchers
have shown that phosphonates mimic well the transition
state for ester hydrolysis.16
(13) Kim, K. K., Song, H. K., Shin, D. H., Hwang, K. Y., Suh, S. W.
Structure 1997, 5, 173-185. Schrag, J . D., Li, Y. G., Cygler, M., Lang,
D. M., 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.
(14) Ollis, D. L.; Cheah, E.; Cygler, M.; Dijkstra, B.; Frolow, F.;
Franken, S. M.; Harel, M.; Remington, S. J .; Silman, I.; Schrag, J .
Sussman, J . L.; Verschueren, K. H. G.; Goldman, A. Prot. Engineer.
1992, 5, 197-211.
(15) Microbiologists recently renamed the bacterial species Pseudomo-
nas cepacia as Burkholderia cepacia, but we continue to use the older
name in this paper.
To search for stable conformations, we systematically
minimized all nine possible conformations along bonds
b and c (see structures in Tables 2 and 3.).17 We did not
vary the dihedral angle along bond a because the active

Enantioselectivity of Pseudomonas cepacia Lipase
J . Org. Chem., Vol. 64, No. 8, 1999 2641
Ta ble 2. Sta ble Con for m a tion s of Tr a n sition Sta te An a logs for th e P CL-Ca ta lyzed Hyd r olysis of 1-Bu tyr a te
dihedral angles,a deg
(bonds b, c)
Eb
no. of
structure name
(kcal/mol)
H-bondsc
acidd
productive
(S)-1-PCL-a
(S)-1-PCL-b
(S)-1-PCL-c
(S)-1-PCL-d
(S)-1-PCL-e
(S)-1-PCL-f
-175, 176
85, 78
-60, 176
75, -110
-164, -59
-85, 53
0, 8
6
2, 3, 6
17
21
9
5
Glu289
Asp264, Glu289
Glu289
yes
yes
6e
4f
3h
4f
1i
maybeg
no
Asp264
Asp264
no
no
-
(R)-1-PCL-a
(R)-1-PCL-b
(R)-1-PCL-c
(R)-1-PCL-d
(R)-1-PCL-e
(R)-1-PCL-f
(R)-1-PCL-g
(R)-1-PCL-h
-177, -169
58, -160
158, 76
13
20
28
0, 7
5
7
17
19
5
5
5
Asp264
Glu289
Glu289
Glu289
Glu289
Asp264
Glu289
Asp264
yes
yes
yes
no
no
no
-63, -158
4f
3h
3h
3h
3h
-67, 93
52, -76
58, 56
no
no
175, -48
a
b
Dihedral angle for bond b is defined by O-C1-C2-C3. The dihedral angle for bond c is defined by C1-C2-C3-C5. Different initial
conformations sometimes converged to the same final structure, but with different energy values. Each energy value is listed. c The transition
d
state in Figure 2b requires five hydrogen bonds. PCL contains redundant acid moieties for the catalytic triad: Asp264 and Glu289.
e His 286 forms a bifurcated hydrogen bond to both Asp264 and Glu289. f The hydrogen bond between His286 and the substrate oxygen
g
h
was missing. The missing hydrogen bond may be a weak hydrogen bond. Hydrogen bonds between His286 and Ser87 and between
His286 and the substrate oxygen are missing. All hydrogen bonds are missing except for one between phosphonyl oxygen and Leu17.
i
site fixed this angle near 180°. A productive conformation
requires the phosphonyl oxygen (oxyanion mimic) in the
oxyanion hole and the alcohol oxygen within hydrogen
bonding distance of His286. These requirements fix the
dihedral angle along bond a near 180°. The dihedral
angle along bond d varied significantly during minimiza-
tion, so we did not vary it manually. In addition, we
manually constructed several reasonable structures and
minimized them. Ideally, all starting conformations
should converge to one, lowest energy minimum. In
practice, some conformations converge, but others do not.
The results of the calculations are several local minima,
each of which must be judged for catalytic competence.
Tables 2 and 3 summarize the stable conformations.
One conclusion was obvious. The large substituent of
primary alcohols did not bind as expected in the large
hydrophobic pocket. Previous work on secondary alcohols
identified a large and a medium hydrophobic pocket in
the alcohol-binding site of lipases.18,7 The large pocket
lies above the active site serine in the view shown in
Figures 3 and 4, while the medium pocket lies below the
active site serine. As expected the medium substituents
of 1 and 2, CH3, lies in, or near, the medium-sized binding
pocket analogous to that postulated for secondary alco-
hols. However, the large substituent of the primary
alcohols (CH2Ph or OPh) did not bind in the large
hydrophobic pocket. In all calculated minima, this large
pocket is empty or nearly empty. Instead, the CH2Ph or
OPh groups usually bind in a narrow groove at the base
of the active site. We call this groove the alternate
hydrophobic pocket. It is also lined with hydrophobic
amino acidssTyr23, Leu27, Tyr29, Phe146, Ile290, and
Leu293. (In a few nonproductive structures, the CH2Ph
or OPh groups lie outside this pocket, but no structure
showed the CH2Ph or OPh groups within the large
hydrophobic pocket.) Thus, the large substituents of
primary and secondary alcohols bind to different pockets
in PCL. Other researchers have also suggested that PCL
contains two hydrophobic pockets on the basis of sub-
strate mapping.19
Meth od for Id en tifyin g Best P r od u ctive Con for -
m a tion s. We used two criteria to identify which local
minimima were catalytically relevant. First, we looked
for the five hydrogen bonds required for catalysis, see
Figure 2b. To assign a hydrogen bond, we looked, first,
for a donor to acceptor atom distance of less than 3.1 Å.
Second, we looked for a nearly linear arrangement of
donor atom, hydrogen, and acceptor atom. We considered
an angle of 120° or more as nearly linear. These are
relatively generous limits for a hydrogen bond. We
classified any structure lacking two or more of the five
hydrogen bonds as nonproductive. However, if a structure
lacked only one hydrogen bond and it was only slightly
outside the limits set above, then we classified it as
maybe productive.
(16) Phosphonates strongly inhibit hydrolases; for example, Hanson,
J . E.; Kaplan, A. P.; Bartlett, P. A. Biochemistry 1989, 28, 6294-6305.
Antibodies to phosphonates often catalyze hydrolysis of the corre-
sponding esters; for example, J acobs, J .; Schultz, P. G.; Sugasawara,
R.; Powell, M. J . Am. Chem. Soc. 1987, 109, 2174-2176.
(17) Although some researchers used molecular dynamics to search
conformations of ligand-protein complexes, we found that molecular
dynamics searched too small a region of conformational space. For
example, starting from a structure of (S)-1-PCL with bond angles of
-74° and 75° (bonds b and c, respectively), a long molecular dynamics
run (300 K for 1 ps for the equilibration period and 25 ps for the
simulation) shifted torsion angles by less than 20°. However, upon
energy minimization, this conformation relaxed to torsion angles of
-60° and 176° (structure (S)-1-PCL-c in Table 1). For this reason,
we used a systematic search to more completely search for productive
conformations.
PCL contains redundant acids for the catalytic triad,
Asp264 and Glu289. The catalytic histidine, His286, lies
near both acids and can hydrogen bond to either one or
both acids. The closely related lipase from Pseudomonas
glumae also contains redundant acids: Asp263 and
(18) 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.
(19) (a) Hof, R. P.; Kellogg, R. M.J . Chem. Soc., Perkin Trans. 1 1996,
2051-2060. (b) Lemke, K., Lemke, M., Theil, F. J . Org. Chem. 1997,
62, 6268-6273.

2642 J . Org. Chem., Vol. 64, No. 8, 1999
Tuomi and Kazlauskas
Ta ble 3. Sta ble Con for m a tion s of Tr a n sition Sta te An a logs for th e P CL-Ca ta lyzed Hyd r olysis of 2-Bu tyr a te
dihedral angles,a deg
(bonds b, c)
Eb
no. of
structure name
(kcal/mol)
H-bondsc
acidd
productive
(S)-2-PCL-a
(S)-2-PCL-b
(S)-2-PCL-c
(S)-2-PCL-d
(S)-2-PCL-e
(S)-2-PCL-f
(S)-2-PCL-g
-161, -153
-44, -65
48, -168
-77, -166
-179, -78
159, 71
5
18
8, 13
0
22
16
2
4e,f
4f,g
4f,h
3i
Asp264
Asp264
Asp264
Glu289
Asp264
-
maybe
maybe
maybe
no
no
no
3i
2j
61, 79
1k
-
no
(R)-2-PCL-a
(R)-2-PCL-b
(R)-2-PCL-c
(R)-2-PCL-d
(R)-2-PCL-e
(R)-2-PCL-f
-155, 176
83, 86
-61, 145
-159, -64
79, 59
7, 11
19
2, 3, 9
13
19
0
6l
5
Asp264
Asp264
Asp264
Asp264
Glu289
Asp264
yes
yes
maybe
maybe
no
4e,f
4e,f
4e
-53, -76
2m
no
a
b
Dihedral angle for bond b is defined by O-C1-C2-O. The dihedral angle for bond c is defined by C1-C2-O-C4. Different initial
conformations sometimes converged to the same final structure, but with different energy values. Each energy value is listed. c The transition
d
state in Figure 2b requires five hydrogen bonds. PCL contains redundant acid moieties for the catalytic triad: Asp264 and Glu289.
eThe hydrogen bond between His286 and Ser87 is missing. f The missing hydrogen bond may be a weak hydrogen bond. The hydrogen
g
h
bond between His286 and the substrate oxygen was missing. Two minimizations of this conformation gave slightly different hydrogen
bonding paterns. In one minimization, the hydrogen bond between His286 and the alcohol oxygen was weak, but the hydrogen bond
i
between His286 and Asp264 was present. In the other minimization, both of these bonds were weak. Hydrogen bonds between His286
j
and Ser87 and between His286 and the substrate oxygen are missing. All hydrogen bonds are missing except for the two bonds between
k
the phosphonyl oxygen and Leu17 and the phosphonyl oxygen Gln88. All hydrogen bonds are missing except for one between phosphonyl
l
oxygen and Leu17. In addition to the five catalytically essential hydrogen bonds, this structure also shows an additional hydrogen bond
m
between the phenoxy oxygen of the substrate and the phenolic hydroxyl of Tyr29. Only hydrogen bonds between His286 and Asp264
and between the phosphonyl oxygen and Leu17.
Glu288.20 Mutation of Asp263 to Ala yielded a lipase with
25% of the original activity, showing that Glu288 can
serve as the acid in the catalytic triad. For this reason,
we accepted a hydrogen bond from His286 to either of
Asp264 or Glu289 as productive.
The second requirement for catalytically relevant
minima was good binding of the alcohol moiety in the
active site. We made this judgment qualitatively and
picked minima that bound the methyl and phenyl groups
of the alcohol more completely. In addition, we avoided
binding modes that required extensive rearrangement of
the side chains within the active site.
We did not use calculated energies to judge whether a
conformation was catalytically relevant because we found
the calculated energies unreliable. With large structures
such as an enzyme, the sum of small differences, even
those far from the active site, yields large differences in
calculated energy. For example, in the minimization of
(S)-1-PCL, three initial structures converged to the
“same” final structure, labeled (S)-1-PCL-c in Table 2.
The geometry of the transition state analogue was
identical in all three, but the calculated energies differed
by 4 kcal/mol. Unnoticed differences in the protein
residues or solvent molecules caused this difference in
energy. Similarly, for (S)-1-PCL-a in Table 2 the calcu-
lated energy of two apparently identical structures dif-
fered by 8 kcal/mol.
(S)-1-PCL-a and -b, we favor conformation a as the
catalytically competent one, Figure 3, because it has
qualitatively better hydrophobic interactions. The methyl
substituent lies in a shallow pocket between His86 and
His286, while the phenyl ring lies in alternate hydro-
phobic pocket mentioned above. Conformation b is similar
to a with the following differences. His286 turns between
Asp264 and Glu289 and hydrogen bonds to both acid
residues. The methyl points out of the active site toward
the solvent and the phenyl ring is oriented deeper into
the alternate hydrophobic pocket. To accommodate the
phenyl ring, residues Ile290, Leu287, and Trp30 move
from their original positions in the X-ray crystal struc-
ture. In conformation c, the methyl lies in the same
pocket as in conformation a, while phenyl ring points out
of the pocket, exposing one face to solvent. In addition,
conformation c contains a weak hydrogen bond between
His286 and the substrate oxygen. The nitrogen-oxygen
distance of 3.05 Å is within the accepted range, but the
N-H-O angle of 104° is too acute.
Conformations d and f lack four and two of the five
catalytically important hydrogen bonds, respectively, and
are thus nonproductive. Conformation e lacks only one
hydrogen bond, but the nitrogen-oxygen distance of 3.42
Å is larger than expected for even a weak hydrogen bond.
Thus, conformation e is also nonproductive. The N-H-O
angle of 135° is within the accepted range.
For (R)-1, we found three productive conformations and
five nonproductive conformations. As with (S)-1, all
conformations bind the benzyl group, not in the large
hydrophobic pocket, but in the alternate hydrophobic
pocket. The three productive conformations, a-c, all
contain the five catalytically essential bonds, but none
of the three makes good hydrophobic contacts between
the lipase and both the methyl and benzyl substituents.
Ca ta lytica lly Releva n t Con for m a tion s of 1-P CL.
Among the six calculated minima for (S)-1-PCL, we
identified two productive conformations, one possibly
productive conformation, and three nonproductive con-
formations, Table 2. Of the two productive conformations,
(20) Noble, M. E. M.; Cleasby, A.; J ohnson, L. N.; Egmond, M. R.;
Frenken, L. G. J . FEBS Lett. 1993, 331, 123-128.

Enantioselectivity of Pseudomonas cepacia Lipase
J . Org. Chem., Vol. 64, No. 8, 1999 2643
F igu r e 3. Catalytically competent orientation of the phos-
phonate transition state analogue for the PCL-catalyzed
hydrolysis of (S)-1 butanoate as identified by molecular
modeling. Atoms from the lipase are shown in light gray, while
atoms from the phosphonate are in light gray (phosphorus),
dark gray (carbons), and darkest gray (oxygens). For clarity,
all hydrogen atoms and water molecules are hidden. On the
basis of previous work with secondary alcohols, we expected
the large substituent of this primary alcohol (CH2Ph) to bind
in the large hydrophobic pocket. However, modeling found an
alternate binding pocket. This binding mode maintains all the
catalytically essential hydrogen bonds identified in Figure 2b.
In the fast-reacting enantiomer, shown above, the methyl
group nestles into a hydrophobic pocket created in part by the
catalytic histidine (His 286). The slow reacting enantiomer (not
shown) shows an identical orientation, except the hydrogen
and methyl groups at the stereocenter have exchanged posi-
tions. This exchange removes the methyl group from the
hydrophobic pocket and leaves it solvent-exposed. Thus,
modeling predicts that the fast-reacting enantiomer binds
better to PCL.
F igu r e 4. Catalytically competent orientation of the phos-
phonate transition state analogue for the PCL-catalyzed
hydrolysis of (S)-2 butanoate as identified by molecular
modeling. Atoms from the lipase are shown in light gray, while
atoms from the phosphonate are in light gray (phosphorus),
dark gray (carbons), and darkest gray (oxygens). For clarity,
all hydrogen atoms and water molecules are hidden, and two
amino acid residues, Tyr 23 and Tyr 29, are shown in the stick
representation. In addition, the phenolic oxygen atom of Tyr
29 is colored darkest gray. This binding mode maintains all
the catalytically essential hydrogen bonds identified in Figure
2b. The binding mode of the fast-reacting enantiomer, shown
above, is similar to that shown for (S)-1 butanoate in Figure
3. The large substituent (OPh) binds in the alternate hydro-
phobic pocket. The oxygen of the OPh group lies near the
phenolic oxygen of Tyr 29 (O-O distance 2.84 Å), but does not
form a hydrogen bond since the O-H-O bond angle (98°) is
too acute. The slow-reacting enantiomer (not shown) shows
two differences. First, the hydrogen and methyl at the stereo-
center have exchanged positions. This exchange nestles the
methyl group into the hydrophobic pocket. Thus, modeling
predicts that the slow enantiomer bind better. Second, a
hydrogen bond forms between the oxygen of OPh and the
phenolic oxygen of Tyr 29 (O-O distance 3.04 Å, O-H-O
angle ) 131°) due to subtle rearrangements of the structure.
The hydrogen bond probably slows down the release of the
alcohol.
Of the three conformations, conformation a makes the
best hydrophobic contacts. It binds the phenyl ring well
within the alternate binding pocket, but the methyl group
lies without hydrophobic contacts and points toward the
opening of the active site pocket. In conformation c, the
methyl group lies without hydrophobic contacts as in
conformation a. The benzyl group twists within the
alternate hydrophobic pocket, thereby losing hydrophobic
contacts to Tyr23 and Tyr29. In conformation b, the
methyl lies in the alternate hydrophobic pocket and the
benzyl substituent points toward the medium sized
pocket, but is too large to bind in it. The phenyl ring loses
all hydrophobic contacts. Since conformation a makes the
best hydrophobic contacts, it will be considered the
catalytically relevant conformations.
Conformation d is nonproductive because it lacks the
hydrogen bond between His286 and the substrate oxygen.
The distances are too long with N-O distance of 3.58 Å
and the N-H-O angle of 89° is too acute. Conformations
e-h lack two hydrogen bonds and are therfore nonpro-
ductive.
Thus, the major difference between transition states
for esters of (S)-1 and (R)-1 is the binding of the methyl
group (the medium substituent). For esters of preferred
enantiomer, (S)-1, the methyl group binds in a hydro-
phobic pocket, but for esters of (R)-1 it lies outside this
pocket.
Ca ta lytica lly Releva n t Con for m a tion of 2-P CL.
For (S)-2, we found three possibly productive and four
nonproductive conformations, Table 3. The possibly
productive conformations, a-c, contained one weak hy-
drogen bond. For conformation a, the hydrogen bond
between His286 and Ser87 is weak: the N-O distance
is 3.32 Å and the N-H-O bond angle is 123°. For
conformation b, the hydrogen bond between His286 and
alcohol oxygen was weak: the N-O distance is 3.01 Å
and the N-H-O bond angle is 140°. Conformation c gave
slightly different hydrogen bonding patterns in two
different minimizations. In one minimization, the hydro-
gen bond between His286 and the alcohol oxygen was
weak (the N-O distance is 3.33 Å and the N-H-O bond
angle is 141°). In the other minimization, hydrogen bonds
between His286 and the alcohol oxygen and between
His286 and Asp264 were both weak.
Of the three possibly productive conformations, (S)-
2-PCL-a shows the best hydrophobic interactions with
the substrate. This conformation resembles (R)-1-PCL-a
above. The benzyl group binds in the alternate hydro-

2644 J . Org. Chem., Vol. 64, No. 8, 1999
Tuomi and Kazlauskas
Ta ble 4. Kin etic P a r a m eter s for P CL-Ca ta lyzed
phobic pocket while the methyl group lies without
hydrophobic contacts and points toward the opening of
the active site pocket. In conformation b the methyl group
binds in the medium pocket, while the phenyl group
points toward the large hydrophobic pocket. The phenyl
group does not reach the hydrophobic pocket and has no
hydrophobic interactions with PCL. In conformation c the
methyl lies in the alternate hydrophobic pocket and the
benzyl substituent points toward the medium sized
pocket, but is too large to bind in it. The phenyl ring loses
all hydrophobic contacts. None of these three conforma-
tions is clearly best. The hydrogen bond appears stronger
for b and c, while the hydrophobic interactions are
definitely stronger for a. We chose conformation a as the
catalytically relevant conformation, but most of the
interpretation below would remain the same if we had
chosen conformations b or c.
Hyd r olysis of En a n tiom er s of 1-Aceta te a n d 2-Aceta tea
substrate
kcat (min-1
)
Km (mM)
kcat/Km (M-1 min-1
)
(S)-1-acetate
(R)-1-acetate
(S)-2-acetate
(R)-2-acetate
4.7 ( 2.1
3.2 ( 0.4
480 ( 150
5.0 ( 0.8
28 ( 1
300 ( 140
33 ( 12
6 ( 4
170
11
15000
830
a
Kinetic constants were determined by the method of Hwang
et al. (ref 36). This method fits the observed degree of conversion
as a function of time with the previously measured enantiomeric
ratios (Table 1). Since the substrate was not completely dissolved
in the reaction mixture, the values of Km are apparent values.
the methyl group and the “extra” hydrogen bond. For
esters of preferred enantiomer, (S)-2, the methyl group
lies outside the hydrophobic pocket, and no hydrogen
bond forms between the phenoxy oxygen and the phenol
OH of Tyr29. In contrast, for esters of (R)-2, the methyl
group binds in the hydrophobic pocket, and a hydrogen
bond does form between the phenoxy oxygen and the
phenol OH of Tyr29.
Conformations d-g lack two to four hydrogen bonds
and are therefore nonproductive.
For (R)-2, we found two productive conformations, two
possibly productive conformations and two nonproductive
conformations. The two productive conformations, a and
b, have all five catalytically essential hydrogen bonds.
The two possibly productive conformations, c and d, have
a weaker hydrogen bond between His286 and Ser87 (c:
N-O distance is 3.27 Å, N-H-O angle is 131°; d: 3.18
Å and 124°, respectively).
Of these four conformations, (R)-2-PCL-a shows the
best hydrophobic interactions. It binds the methyl group
of (R)-2 in the medium-sized pocket and the benzyl group
in the alternate hydrophobic pocket. Conformation b
leaves the methyl group outside the hydrophobic pocket.
Accommodating the benzyl group also required rear-
rangement of the amino acid side chains. Conformation
c has the medium-sized substituent oriented toward the
alternate hydrophobic pocket and the large substituent
partly in the medium-sized pocket. One face of the phenyl
ring is open to the solvent. Conformation d has the
methyl substituent bound in the medium-sized pocket.
The large substituent is bound within the alternate
hydrophobic pocket, but in a manner that causes signifi-
cant rearrangement of the amino acid side chains. Thus,
we believe conformation a is the catalytically relevant
conformation because it contains all necessary hydrogen
bonds and qualitatively shows the best hydrophobic
interactions in the active site.
Further, conformation a of (R)-2-PCL shows an ad-
ditional hydrogen bond between the phenolic hydroxyl
of Tyr29 and the phenoxy oxygen of the substrate (O-O
distance of 3.04 Å, O-H-O angle of 131°). None of the
other conformations for either (S)-2 or (R)-2 formed this
hydrogen bond. This bond may explain the slower hy-
drolysis of esters of (R)-2, see below. The catalytically
relevant conformation for the fast-reacting enantiomer,
(S)-2-PCL-a, lacks this hydrogen bond, but there may
be a weak interaction. The distances are suitable (the
O-O distance is 2.84 Å), but the angle is too acute (O-
H-O angle is 98°).
Deter m in a tion of Kin etic Con sta n ts. To help in-
terpret the results from molecular modeling, we mea-
sured the kinetic constants for 1-acetate and 2-acetate,
Table 4. For 1-acetate, the values of kcat are similar for
both enantiomers, but the values for Km differ by ap-
proximately a factor of 10. In most cases, researchers
treat Km as a binding constant. Thus, PCL favors the (S)-
enantiomer of 1-acetate because it binds better than the
(R)-enantiomer. Modeling suggests that in the catalyti-
cally relevant conformation, the methyl group in (S)-1
binds in the medium pocket, while the methyl group of
(R)-1 does not. For these reasons, we propose that binding
of the methyl group within the medium pocket deter-
mines the enantiopreference of PCL for esters of (S)-1.
For 2-acetate, both kcat and Km differ significantly for
the enantiomers. The value of kcat is approximately 100
times larger for the preferred (S)-2-acetate than for the
(R)-2-acetate. On the other hand, the preferred (S)-
enantiomer binds less tightly to PCL; the value of Km is
approximately five times higher for (S)-2-acetate than for
(R)-2-acetate. Thus, binding differences favor the (R)-
enantiomer while reaction rate differences favor the (S)-
enantiomer. Since the differences in reaction rate are
larger than differences in binding constant, PCL favors
(S)-2-acetate. Nishizawa et al.21 measured similar kinetic
constants for a related substrate, 2-methyl-2-(4-phenoxy-
phenoxy)ethyl acetate, with PCL. The structure is the
same as 2-acetate, but with an added p-phenoxy group.
The measured value of kcat is eight times larger for the
(S)-enantiomer (18 vs 2.2 min-1 for (R)), but the value of
Km is also one and a half times larger for the (S)-
enantiomer (6.8 vs 4.9 mM). Thus, as for 2, kcat favors
the (S)-enantiomer, while Km favors the (R) enantiomer.
The overall enantioselectivity was 5.8 favoring the (S)-
enantiomer.
The values of kcat for the preferred enantiomer of 2 are
much higher than for 1. It is not readily apparent why 2
should react much faster than 1.
Conformation (R)-2-PCL-e is nonproductive because
it lacks the hydrogen bond between His286 and Ser87.
The N-O distance is 3.42 Å, and the N-H-O angle is
136°. These distances are too far for a hydrogen bond.
Conformation (R)-2-PCL-f is nonproductive because it
lacks three hydrogen bonds.
Thus, kinetic measurements show that esters of (S)-2
react faster, but bind more poorly. Modeling easily
accounts for the poorer binding. The methyl group of (S)-2
lies outside the hydrophobic pocket, but the methyl group
of (R)-2 lies inside this pocket. Modeling also suggests
Thus, the two major differences between transition
states for esters of (S)-2 and (R)-2 are the orientation of
(21) Nishizawa, K.; Ohgami, Y.; Matsuo, N.; Kisida, H.; Hirohara,
H. J . Chem. Soc., Perkin Trans. 2 1997, 1293-1298.

Enantioselectivity of Pseudomonas cepacia Lipase
J . Org. Chem., Vol. 64, No. 8, 1999 2645
enantioselectivity upon acetylation and nitration are
consistent with the proposed role of Tyr29, but, given the
nonselective nature of chemical modification, they do not
prove that the proposed role is correct.
Discu ssion
Molecular modeling has identified an alternate binding
pocket in PCL for the large substituent of alcohols 1 and
2. The catalytically productive conformations of transition
state analogues for hydrolysis of esters of (S)-1, (R)-1,
(S)-2, and (R)-2 all position the phenyl group within this
alternate binding pocket. The X-ray structure of PCL
clearly shows this pocket, and two previous substrate
selectivity studies also suggest such a pocket.19 The
identification of an alternate binding pocket explains why
PCL shows opposite enantiopreference toward primary
and secondary alcohols. Primary and secondary alcohols
bind differently. In secondary alcohols, the large sub-
stituent binds in the large pocket, but in primary
alcohols, the large substituent binds in the alternate
binding pocket.
Modeling and kinetics suggest that PCL recognizes the
enantiomers of 1 by the binding of the methyl group. In
the favored enantiomer, PCL binds the methyl group in
a hydrophobic pocket, while in the unfavored enantiomer,
the methyl group lies exposed to solvent. The maximum
hydrophobic binding energy of a methyl group is 3.2 kcal/
mol.23 An enantiomeric ratio of 16 corresponds to an
energy difference at 25 °C of 1.7 kcal/mol in the transition
state for the two enantiomers. This value is consistent
with partial hydrophobic binding for one enantiomer, but
not the other. One way to increase the enantioselectivity
of PCL toward 1 would be to make mutations that more
completely bind the methyl group.
Modeling and kinetics also reveal why the enantiose-
lectivity reverses for 2. Although the binding of the
methyl group favors (R)-2, this structure also contains
an “extra” hydrogen bond between the phenoxy oxygen
and Tyr29. The rate-determining step of serine esterase-
catalyzed hydrolysis of esters is probably breakdown of
the tetrahedral intermediate.24-26 Slowing the release of
alcohol would increase the chances for the reverse
reactionsreturn to the starting ester. On the basis of the
structure of a related lipase, lipase from Chromobacte-
rium viscosum, without a bound transition state ana-
logue, Lang et al. suggested that Tyr29 stabilizes the
conformation of Glu288 (part of the catalytic triad) by
hydrogen bonding through His86.27 Loss of this stabiliza-
tion when Tyr29 makes a hydrogen bond to the substrate
may also explain the slower reaction of PCL with esters
of (R)-2.
F igu r e 5. Chemical modification of tyrosyl residues. Acetyl-
ation of tyrosine residues disrupts any hydrogen bonds formed
by the phenolic hydroxyl. Subsequent reaction with hydroxy-
lamine removes the O-acetyl groups. Nitration of the tyrosyl
residues increases the acidity of the phenolic hydroxyl and
strengthens any hydrogen bonds involving this group. Chemi-
cal modification of enzymes is not selective for a particular
position. All accessible tyrosines will be modified.
an explanation for the faster reaction of (S)-2. The “extra”
hydrogen bond between phenolic OH of Tyr29 and the
phenoxy oxygen of (R)-2 may hold the leaving alcohol in
place and promote the reverse reaction: breakdown to
the starting ester. The overall effect would be to slow the
forward reaction.
Ch em ica l Mod ifica tion of Tyr 29. To test the im-
portance of the phenolic OH of Tyr29 in the enantiose-
lectivity of PCL toward 2, we selectively modified the
tyrosine hydroxyl groups by acetylation with N-acetyl-
imidazole,22 Figure 5. Acetylation will remove the “extra”
hydrogen bond mentioned above. We predict that break-
ing this bond should lower the enantioselectivity of PCL
toward 2. However, acetylation also introduces a large
group in the binding pocket. We cannot predict the effect
of changing the size of the pocket. Another complication
is that acetylation modified not only Tyr29, but an
average of 10 of the 14 tyrosine residues in PCL. We
cannot predict the effect of acetylating tyrosines other
than Tyr29. With these cautions in mind, we observed a
decrease in the enantioselectivity toward 2-acetate from
17 (unmodified PCL) to 8 (acetylated PCL), Table 1 above.
We expected that the enantioselectivity of PCL toward
2-acetate would return to 17 upon treatment with hy-
droxylamine to remove the O-acetyl groups from acetyl-
ated tyrosines.23 However, the enantiomeric ratio in-
creased more than expected to 42. This increase suggests
an irreversible change occurred during chemical modi-
fication, perhaps acetylation of lysine residues or the
N-terminal amino group.
We also modified PCL by nitration of the tyrosine
residues with tetranitromethane to give 3-nitrotyrosyl
residues, Figure 5. Nitration of tyrosine increases the
acidity of the phenolic hydroxyl (pKa drops from ap-
proximately 10.3 to 7.3). The increase acidity should
increase the hydrogen bond-donating ability of Tyr29.
Consistent with this expectation, the enantioselectivity
toward 2-acetate from 17 (unmodified PCL) to 37 (ni-
trated PCL), Table 1 above. Nitration modified on aver-
age 11 of the 14 tyrosine residues. The changes in
The identification of a hydrogen bond involved in
enantiomer recognition of 2 also suggested a strategy to
increase the enantioselectivity. Increasing the strength
of this hydrogen bond should increase enantioselectivity,
while decreasing the strength should decrease enanti-
oselectivity. Indeed, chemical modification of PCL by
nitration to increase the strength of the hydrogen bond
(24) Nishizawa, K., Ohgami, Y., Matsuo, N., Kisida, H., Hirohara,
H. J . Chem. Soc., Perkin Trans. 2 1997, 1293-1298.
(25) Hirohara, H.; Philipp, M.; Bender, M. L. Biochemistry 1977,
16, 1573-1580.
(26) Hunkapiller, M. W.; Forgac, M. D.; Richards, J . H. Biochemistry
1976, 15, 5581-5588.
(27) Lang, D. Hofmann, B. Haalck, L. Hecht, H. J . Spener, F.
Schmid, R. D. Schomburg, D. J . Mol. Biol. 1996, 259, 704-717.
(22) Lundblad, R. L. Techniques in Protein Modification; CRC: Boca
Raton, FL, 1995; pp 209-231.
(23) Fersht, A. Enzyme Structure and Mechanism, 2nd ed.; Free-
man: New York, 1985; p 304.

2646 J . Org. Chem., Vol. 64, No. 8, 1999
Tuomi and Kazlauskas
lamine (0.94 mL, 6.75 mmol) in dichloromethane (5 mL). The
reaction was stirred overnight under an atmosphere of nitro-
gen. The reaction mixture was washed with water (2 × 5 mL)
and dried over MgSO4. Evaporation of the solvent yielded pure
product in 92% yield.
and acetylation to decrease the strength of the hydrogen
bond changed the enantioselectivity in the expected
direction. Although the magnitude of the change was
modest (a factor of 2, approximately 0.4 kcal/mol), this
is the first time that researchers have rationally in-
creased the enantioselectivity of a lipase. Gu and Sih
previously increased the enantioselectivity (typically from
E ) 2 to 38) of lipase from Candida rugosa toward chiral
acids upon nitration of the tyrosyl residues,28 but they
discovered this effect empirically. A number of other
attempts to increase enantioselectivity by chemical modi-
fication were less successful. Modification of arginine in
lipase from Rhizomucor miehei decreased or slightly
increased (from E ) 2.9 to 4.4) the enantioselectivity
toward esters of 2-methyldecanoic acid.29 Acetylation of
lysine residues in PCL either had no effect or decreased
the enantioselectivity toward alcohols.30
These conclusions may hold for PCL-catalyzed reac-
tions of other primary alcohols. For alcohols without an
oxygen at the stereocenter, the alternate binding pocket
would account for the reversed enantiopreference be-
tween primary and secondary alcohols. However, Zuegg
et al. modeled several other primary alcohols in PCL,31
but did not identify the “alternate binding pocket”
discussed in this paper. Since primary alcohols are more
flexible than secondary alcohols, there may be more than
one explanation for the reversed enantiopreference.
When the primary alcohol contains an oxygen at the
stereocenter, the situation is even more complex. De-
pending on the detailed structure, primary alcohols with
an oxygen at the stereocenter may or may not make a
hydrogen bond and thus may or may not follow the rule
in Figure 1b. This “extra” hydrogen bond explains the
difficulties researchers encountered in predicting the
enantiopreference of PCL toward this class of alcohols.
The “extra” hydrogen bond identified in this paper may
also contribute to the high selectivity of lipase from
Pseudomonas fluorescens toward 1,2-diols of the type
HOCH2C(OH)RR′, where reaction occurred at the pri-
mary alcohol.32,20a Researchers found that Pseudomonas
lipase AK from Amano efficiently resolves a number of
alcohols with that general structure. Although crystal-
lographers have not yet solved the structure of this lipase,
an alignment of its amino acid sequence with that of PCL
suggest a tyrosine in the same position.33 The OH at the
stereocenter may also form a hydrogen bond to Tyr50 in
lipase AK.
En a n tiom er ic Ra tio of P CL tow a r d 2-Aceta te. A solu-
tion of PCL in phosphate buffer (2 mL, 10 mM, pH 7.5 or 8.0)
was added to a mixture of distilled water (7 mL) and 2 (48
mg) dissolved in tert-butyl methyl ether (5 mL). The pH was
maintained at either 7.5 or 8.0 with the addition of 0.0981 M
NaOH controlled by a Radiometer RTS 822 pH stat. The
reaction was stopped by adding 1.0 N HCl until the pH
dropped to 1.8. The mixture was extracted with ethyl acetate
(3 × 10 mL), and the combined organic extracts were dried
with MgSO4. The alcohol and acetate were separated by
column chromatography on silica gel eluted with 1:4 ethyl
acetate/hexane. The enantiomeric purity of the alcohol was
measured by gas chromatography on a 30 m Chiraldex G-TA
capillary column (Astec, Whippany, NJ ): 2-phenoxypropanol:
90 °C 33.3 min (R), 34.0 min (S). The acetate was not well
resolved on this column, so it was converted to the trifluoro-
acetate derivative for analysis: 2-phenoxypropanol trifluoro-
acetate: 80 °C, 22.1 min (R), 22.9 min (S). In some experiments
a Chiralsil-Dex CB column (Chrompack, Middeburg, The
Netherlands) was used (25 m, 110 °C) which separated
enantiomers of both the alcohol and the acetate.
Absolu te Con figu r a tion of F a vor ed En a n tiom er of 2.
J ones reagent was added dropwise, until the yellow color
persisted, to an acetone solution of 2-phenoxypropanol (36 mg)
isolated from a PCL-catalyzed hydrolysis. After stirring for 1.5
h, the acetone was evaporated in vacuo and the residue was
triturated with ethyl ether. The ether extract was washed with
water. The aqueous layer was reextracted with ethyl ether,
and the combined ethereal extracts were dried over magne-
sium sulfate. Purification of a portion by preparative TLC (8:
2 cyclohexane/ethyl acetate, Rf ) 0.12) yielded 16 mg. HPLC
analysis using a Chiracel OD column (90:10:1 hexane/2-
propanol/formic acid) showed that the major enantiomer eluted
first (major, 10.0 min; minor, 16.1 min) and thus was assigned
the (S) configuration based on previous reports.34
Kin etic Con sta n ts. Values of Km and kcat were determined
using the method of Hwang et al.35 Briefly, we measured the
degree of hydrolysis of the racemate as a function of time and,
using the previously measured enantiomeric ratio, fit these
data (nine to twelve points for each substrate) to the theoretical
curve using Kaleidegraph software. Lipase solution was added
to a suspension of substrate ((()-1-acetate or (()-2-acetate)
in phosphate buffer (1 mL, 0.2 M, pH 7.0, 25 °C) stirred
vigorously at 24 °C. At different reaction times, the reaction
was stopped by lowering the pH to 2 with HCl (1 N). The
reaction mixture was extracted with ethyl acetate (5 × 0.8 mL),
and the combined organic extracts were dried over MgSO4. The
degree of hydrolysis was measured by GC (25 m Chirasil-Dex
CB column, 140 °C, FID detector) and corrected for response
factors of the alcohol and acetate. Enantiomers were not
separated at this temperature. Since the substrate was not
completely dissolved in the reaction mixure, the values of Km
are apparent values. Although true values of Km could be
calculated if we knew the partion coeffficient of the substrate,
both enantiomers have the same partion coefficient, so both
the true and apparent Km’s have the same relative values.
Acetyla tion of P CL. Solid N-acetylimidazole (300-fold
molar excess) was added to a solution of PCL (100 mg in 20
mL in 0.010 M phosphate buffer, pH 7.5 with 5 vol % hexane
to promote opening of the lipase lid). After stirring for 1 h, a
2.5 mL aliquot was desalted by passing through a gel filtration
column (Pharmacia PD-10 column equilibrated with 0.01 M
phosphate buffer, pH 7.5). Another 2.5 mL aliquot was
deacetylated by adding hydroxylamine (1 M final concentra-
Exp er im en ta l Section
Lipase from Pseudomonas cepacia was purchased from
Amano Enzyme Company (Lombard, IL) as LPL-200. This is
a purified form of the lipase containing only protein and a
small amount of glycine. For synthetic applications, less pure
forms are suitable.
2-P h en oxy-1-p r op a n ol Aceta te, (()-2 Aceta te. Acetic
anhydride (0.60 mL, 6.75 mmol) was slowly added to a solution
of (() 2-phenoxy-1-propanol (1.00 g, 6.57 mmol) and triethy-
(28) Gu, Q.-M.; Sih, C. J . Biocatalysis 1992, 6, 115-126.
(29) Holmquist, M.; Martinelle, M.; Berglund, P.; Clausen I. G.;
Patkar, S.; Svendsen, A.; Hult, K. J . Prot. Chem. 1993, 12, 749-757.
(30) Bianchi, D.; Battistel, E.; Bosetti, A.; Cesti, P.; Fekete, Z.
Tetrahedron: Asymmetry 1993, 4, 777-82.
(31) Zuegg, J ., Honig, H., Schrag, J . D., Cygler, M.J . Mol. Catal. B
Enzymol. 1997, 3, 83-98.
(32) Chen, S. T., Fang, J . M. J . Org. Chem. 1997, 62, 4349-4357.
(33) We used SWISS-PROT: P25275 for the amino acid sequence
of lipase from Pseudomonas fluorescens.
(34) Wu, S.-H.; Lai, S.-Y.; Lin, S.-L.; Chu, F.-Y.; Wang, K.-T.
Chirality 1991, 3, 67-70.
(35) Hwang, S. Y., Brown, K. S., Gilvarg, C. Anal. Biochem. 1988,
170, 161-167. Wu, S.-H., Guo, Z.-W., Sih, C. J . J . Am. Chem. Soc.
1990, 112, 1990-1995.

Enantioselectivity of Pseudomonas cepacia Lipase
J . Org. Chem., Vol. 64, No. 8, 1999 2647
tion), stirring at room temperature for 30 min and desalting
(Pharmacia PD-10 column). The protein content of both the
acetylated and the deacetylated samples were determined with
the Coomassie Blue-binding assay from Bio-Rad using bovine
serum albumin as the standard. The degree of acetylation (N)
was estimated spectrophotometrically using the equation N
) ∆ꢀ278M/1160c. ∆ꢀ278 is the change in the extinction coefficient
at 278 nm, M is the molecular weight of the enzyme (33 kDa),
1160 is the difference in the extinction coefficients for O-
acetyltyrosine and tyrosine, and c is the protein concentration
of the enzyme (mg/mL).36 We estimate that an average of 10
of the 14 tyrosine residues in PCL were acetylated.
Nitr a tion of P CL. Tetranitromethane (300-fold molar
excess) was added to a solution of PCL (100 mg in 20 mL in
0.05 M Tris buffer, pH 8 with 5 vol % hexane to promote
opening of the lipase lid). After stirring in the dark for various
times, a 2.5 mL aliquot was desalted by passing through a gel
filtration column (Pharmacia PD-10 column equilibrated with
0.05 M Tris buffer, pH 8). The protein content of the nitrated
samples were determined with the Coomassie Blue-binding
assay from Bio-Rad using bovine serum albumin as the
standard. The degree of nitration was estimated from the
increase in extinction coefficient at 275 nm using the equation
above. The difference extinction coefficient for 3-nitrotyrosine
and tyrosine is 2640 cm-1 M-1. We estimate that an average
of 11 of the 14 tyrosine residues in PCL were nitrated.
hydrogen bonds. Crystallographic water molecules were in-
cluded in all minimizations.
Transition state analogues for each alcohol were constructed
and manually adjusted to each of the nine staggered local
minima possible along bonds b and c. Steric interactions with
the enzyme sometimes prevented the starting exactly at the
local minimum, but the starting positions were within 15° of
the local minima for the free substrate. Energy minimization
proceeded in four stages. First, 200 of iterations steepest
descent algorithm, all protein atoms constrained with a force
constant of 10 kcal mol-1 Å-2; second, 200 iterations of
conjugate gradients algorithm with the same constraints; and
third, 500 iterations of conjugate gradients algorithm with only
the backbone constrained by a 10 kcal mol-1 Å-2 force constant.
For the fourth stage, minimization was continued using
conjugate gradients algorithm without any constraints until
the rms deviation reached less than 0.005 Å mol-1. Water
molecules and the substrate were not constrained through any
of the minimization cycles.
Molecu la r Dyn a m ics. Starting from a transition state
analogue derived from (S)-1 bound to the active site of PCL,
we manually set the torsion angles to -74° and 75° (bonds b
and c, respectively, in Table 1). These angles place the phenyl
in an unhindered region of the active site. To remove any
strong steric repulsions, we minimized the structure first using
the steepest descent algorithm (200 iterations, all protein
atoms constrained with a force constant of 10 kcal mol-1 Å-2
)
Mod elin g of Tr a n sition Sta te An a logu es in P CL. All
modeling was done with Discover, version 2.9.7 (Biosym/MSI,
San Diego, CA), using the Amber force field except where
noted. We used a distance dependent dielectric constant of 4.0
and scaled the 1-4 van der Waals interactions scaled by 50%.
The distance dependent dielectric constant damps long-range
electrostatic interactions and thus compensates for the lack
of explicit solvation.37 Results were displayed using Insight
II, version 95.0 (Biosym/MSI). Protein structures in Figures 3
and 4 were created using RasMac v2.6.38 Miroslaw Cygler and
J oseph Schrag of the NRC Biotechnology Research Institute,
Montre´al, kindly provided the coordinates for PCL. This
structure contains a phosphate inhibitor that is not included
in the PDB data file (accession code: 3LIP), but is otherwise
very similar. Using the Biopolymer module of Insight II,
hydrogen atoms were added to correspond to pH 7.0. Histidines
were uncharged (singly protonated), aspartates and glutamates
were negatively charged, and arginines and lysines were
positively charged. The catalytic histidine (His286) was pro-
tonated. The inhibitor found in the crystal structure was
covalently linked to Ser87. The phosphate inhibitor was first
modified to a phosphonate by replacing an oxygen with carbon.
Minimization of this structure maintained all five essential
and then using conjugate gradient algorithm (500 iterations,
only protein backbone constrained with a force constant of 10
kcal mol-1 Å-2). We imposed no constraints on the water
molecules or substrate during this initial minimization. We
ran the molecular dynamics for 1 ps at 300 K for the
equilibration period and 25 ps at 300 K for the simulation
using a time step of 1 fs with a history file generated every 10
fs. Both torsion angles remained within 20° of their initial
settings throughout the simulation. Because this long simula-
tion sampled such a small region of conformational space, we
used a systematic search to find the catalytically productive
conformations.
Ack n ow led gm en t is made to the donors of the
Petroleum Research Fund, administered by the Ameri-
can Chemical Society, for support of this research. We
thank NSERC (Canada) for an equipment grant for the
modeling computer and software. We thank Alexandra
N. E. Weissfloch for determination of E of 2 and
determining the absolute configuration, J essamine Ng
for help with the acetylation and nitration of PCL, and
Miroslaw Cygler and J oseph Schrag (NRC Biotechnol-
ogy Research Center, Montre´al) for access to the X-ray
crystallographic coordinates for PCL.
(36) Riordan, J . F., Vallee, B. L. Methods Enzymol. 1967, 11, 570-
576.
(37) Weiner, S. J .; Kollman, P. A.; Case, D. A.; Singh, U. C.; Ghio,
C.; Alagona, G.; Profeta, S., J r.; Weiner, P. J . Am. Chem. Soc. 1984,
106, 765-784.
(38) Sayle, R. For information, see http://www.umass.edu/microbio/
rasmol/.
Su p p or tin g In for m a tion Ava ila ble: Details of the con-
formational analysis of the free substrates. This material is
available free of charge via the Internet at http://pubs.acs.org.
J O981783Y