Prediction of the solvent dependence of enzymatic prochiral selectivity by means of structure-based thermodynamic calculations

Article abstract of DOI:10.1021/ja952674t

A new, quantitative model is elaborated to rationalize the solvent dependence of enzymatic selectivity solely on the basis of the thermodynamics of substrate solvation. The model predicts that any type of the selectivity (defined as the ratio of k(cat)/K(

Full text of DOI:10.1021/ja952674t

3366
J. Am. Chem. Soc. 1996, 118, 3366-3374
Prediction of the Solvent Dependence of Enzymatic Prochiral
Selectivity by Means of Structure-Based Thermodynamic
Calculations
Tao Ke, Charles R. Wescott,† and Alexander M. Klibanov*
Contribution from the Department of Chemistry, Massachusetts Institute of Technology,
Cambridge, Massachusetts 02139
ReceiVed August 7, 1995X
Abstract: A new, quantitative model is elaborated to rationalize the solvent dependence of enzymatic selectivity
solely on the basis of the thermodynamics of substrate solvation. The model predicts that any type of the selectivity
(defined as the ratio of kcat/KM values) should be proportional to the ratio of the thermodynamic activity coefficients
of the desolvated portions of the substrate(s) in the relevant transition state of the enzymatic reaction. The latter
ratio is calculated by (i) determining the desolvated portion of the substrate in the transition state using molecular
modeling based on the crystal structure of the enzyme, (ii) approximating this desolvated portion of the substrate by
a distinct model compound, and (iii) calculating the activity coefficient of this model compound using the UNIFAC
computer algorithm. In this study, the developed general model has been applied to, and verified with, prochiral
selectivity of enzymes. Crystals (lightly cross-linked with glutaraldehyde) of γ-chymotrypsin or subtilisin Carlsberg
used as asymmetric catalysts in organic solvents almost quantitatively adhere to the predictions of the model in the
acetylation of 2-substituted 1,3-propanediols. In contrast, little agreement between the predicted and observed solvent
dependences of the prochiral selectivity has been obtained with lyophilized or acetone-precipitated chymotrypsin,
thus confirming that the enzyme structure in such preparations (but not in crystals) is non-native.
Introduction
transition state. This assumption precludes the extension of the
proposed model to enantioselectivity, since the partition coef-
ficients for different enantiomers of the same compound are
identical. Likewise, prochiral, regio-, and chemo-selectivities
cannot be analyzed either, because in all these instances the
same substrate molecule (just different parts of it) reacts with
the enzyme.
In the present study, we further develop and broaden our
thermodynamic treatment to eliminate the aforementioned
limitations. The resultant model, tested herein with prochiral
selectivity, takes into account variations in substrate desolvation
in the transition states for pro-R and pro-S orientations.
Thermodynamic activity coefficients of the desolvated portions
of the substrate, calculated using computer-generated, transition-
state structures and UNIFAC, correctly predict the solvent
dependence of prochiral selectivity of crystalline chymotrypsin
and subtilisin.
One of the most profound revelations arisen from nonaqueous
enzymology1 is the discovery that the specificity of an enzyme
strongly depends on the solvent.2 Of all the types of enzyme
specificity found to be controlled by the solvent2senantiose-
lectivity, prochiral selectivity, substrate specificity, regioselec-
tivity, and chemoselectivitysthe first two are particularly
important for synthetic applications. Indeed, if generalized and
understood, solvent control of enzymatic stereoselectivity should
enhance the utility of biocatalysis in organic chemistry by
allowing the rational manipulation of the stereochemical
outcome of asymmetric transformations simply by altering the
reaction medium. The ultimate challenge in this regard is to
learn how to predict enzyme selectivity as a function of the
solvent.
As a first step toward this goal, we have recently elaborated
a thermodynamic model which explains the substrate specificity
of the protease subtilisin Carlsberg in organic solvents on the
basis of solvent-to-water partition coefficients of the substrates.3
These partition coefficients can be either measured experimen-
tally or calculated using the UNIFAC computer algorithm.4 An
explicit assumption of our analysis is that the substrates are fully
desolvated, i.e., inaccessible to the solvent, in the enzyme-bound
Theory
The solvent may influence enzymatic selectivity through
several distinct mechanisms. For instance, it could change the
(4) (a) Fredenslund, A.; Gmehling, J.; Rasmussen, P. Vapor-Liquid
Equilibria Using UNIFAC; Elsevier: New York, 1977. (b) Steen, S.-J.;
Ba¨rbel, K.; Gmehling, J.; Rasmussen, P. Ind. Eng. Chem. Process Des.
DeV. 1979, 18, 714. (c) Rasmussen, P.; Fredenslund, A. Ind. Eng. Chem.
Process Des. DeV. 1982, 21, 118. (d) Macedo, E. A.; Weidlich, U.;
Gmehling, J.; Rasmussen, P. Ind. Eng. Chem. Process Des. DeV. 1983, 22,
678. (e) Teigs, D.; Gmehling, J.; Rasmussen, P.; Fredenslund, A. Ind. Eng.
Chem. Res. 1987, 26, 159. (f) Hansen, H. K.; Rasmussen, P.; Schiller, M.;
Gmehling, J. Ind. Eng. Chem. Res. 1991, 30, 2355. Recently, the use of
UNIFAC for estimating thermodynamic activity coefficients has been called
into question by van Tol et al. (van Tol, J. B. A.; Stevens, R. M. M.;
Veldhuizen, W. J.; Jongejan, J. A.; Duine, J. A. Biotechnol. Bioeng. 1995,
47, 71) on the basis of a small number of experiments. However, references
(a)-(f) above provide ample validation for UNIFAC for most solutes/
solvents examined. Also, van Tol et al. used partition coefficients combined
with solubility measurements (plagued with problems outlined in ref 3a) to
determine the activity coefficients, while a rigorous treatment requires the
use of vapor-liquid equilibrium measurements.33
† An NIH Biotechnology Predoctoral Trainee.
X Abstract published in AdVance ACS Abstracts, March 15, 1996.
(1) (a) Klibanov, A. M. Trends Biochem. Sci. 1989, 14, 141. (b) Chen,
C.-S.; Sih, C. J. Angew. Chem., Int. Ed. Engl. 1989, 28, 695. (c) Dordick,
J. S. Enzyme Microb. Technol. 1989, 11, 194. (d) Klibanov, A. M. Acc.
Chem. Res. 1990, 23, 114. (e) Gupta, M. N. Eur. J. Biochem. 1992, 203,
25. (f) Faber, K.; Riva, S. Synthesis 1992, 895. (g) Halling, P. J. Enzyme
Microb. Technol. 1994, 16, 178. (h) Enzymatic Reactions in Organic Media;
Koskinen, A. M. P.; Klibanov, A. M., Eds.; Blackie: London, 1996.
(2) For a review, see (a) Wescott, C. R.; Klibanov, A. M. Biochim.
Biophys. Acta 1994, 1206, 1. (b) Carrea, G.; Ottolina, G.; Riva, S. Trends
Biotechnol. 1995, 13, 63.
(3) (a) Wescott, C. R.; Klibanov, A. M. J. Am. Chem. Soc. 1993, 115,
1629. (b) Wescott, C. R.; Klibanov, A. M. J. Am. Chem. Soc. 1993, 115,
10362.
0002-7863/96/1518-3366$12.00/0 © 1996 American Chemical Society

SolVent Dependence of Enzymatic Prochiral SelectiVity
J. Am. Chem. Soc., Vol. 118, No. 14, 1996 3367
Scheme 1
unsolvated in the transition state (SU). Similarly, the transition
state is regarded as the sum of SS and another portion which
includes only the enzyme and the unsolvated substrate moiety
in the transition state (ESqU). ∆GqA can be expressed as the sum
of the energetic terms of the alternative path9
∆GAq ) ∆GBq + ∆GtEr + ∆GtSr + ∆GStr
-
U
S
∆Gtr - ∆GtSr (1)
Scheme 2
q
ESU
S
where ∆GqB is the activation free energy for the reaction in
solvent B, and ∆Gtr is the free energy of transfer of the moiety
indicated in the subscript from solvent A to B. Assuming that
the solvated surfaces of E and ESqU for low-molecular-weight
tr
substrates are identical,10 ∆GtEr ) ∆G . Consequently, eq 1
ESqU
can be simplified to:
enzyme conformation and thus affect the selectivity of the
reaction by altering enzyme-substrate interactions.5,6 Alterna-
tively, solvent molecules could bind within the enzyme active
site and block the normal binding mode of the substrate.7,8 While
these two possible mechanisms do not necessarily influence
selectivity, a third, driven by the energetics of substrate
solvation, must do so, regardless of the presence of other
mechanisms. Indeed, it has been demonstrated that the energet-
ics of substrate solvation is the dominant means by which the
solvent influences the substrate specificity of subtilisin.3
The contribution of solvation energies to the solvent depen-
dence of enzyme kinetics is demonstrated by the thermodynamic
cycle in Scheme 1. The lower horizontal arrow represents the
enzyme (E) and the substrate (S) reacting in solvent A to form
the transition state (ESq)A with an activation free energy of ∆
GqA. This transition state spontaneously decomposes to ulti-
mately form the free enzyme and products. An alternative,
hypothetical path exists leading to the same transition state
complex. In this path, the substrate and enzyme are separately
transferred from solvent A to solvent B, where they react to
form the transition state (ESq)B, which is subsequently trans-
ferred back to solvent A. For thermodynamic purposes, the
substrate can be represented as the sum of two portions (see
Scheme 2), that which is solvated in the transition state (SS)
and that which is enveloped by the enzyme and is thus
∆GqA ) ∆GqB + ∆GtSr
(2)
U
∆Gq is related to kcat/KM as:11
∆Gq ) -RT ln
kcat
h
(3)
(
)
(
[
)
]
KM κT
where h, κ, R, kcat, KM, and T are the Planck, Boltzmann, gas,
catalytic, and Michaelis constants and temperature, respectively.
∆Gtr can be expressed in terms of thermodynamic activity
coefficients as RT ln (xBγB/xAγA), where γ and x are the activity
coefficient and mole fraction, respectively, of the solute in the
indicated solvent.12 If γ′ is defined as the activity coefficient
of the unsolvated substrate moiety (SU), then
∆GtSr ) RT ln (xBγ′B/xAγ′A)
(4)
U
Substituting eqs 3 and 4 into 2 yields
(kcat/KM)A ) (γ′A/γ′B)(xA/xB)(kcat/KM)B
(5)
Note that for dilute solutions xA/xB depends only on the molar
volume of the solvents when the transfer is done at constant
molar concentration.13 The mole fraction ratio is thus the same
for any substrate and cancels out when eq 5 is expressed for
two substrates, I and II, and solved for the logarithm of the
selectivity in solvent A:
(5) This hypothesis has been proposed to explain solvent-induced
variations in enzymatic selectivity: (a) Wu, H.-S.; Chu, F.-Y.; Wang, K.-
T. Bioorg. Med. Chem. Lett. 1991, 1, 339. (b) Ueji, S.; Fujio, R.; Okubo,
N.; Miyazawa, T.; Kurita, S.; Kitadani, M.; Muromatsu, A. Biotechnol. Lett.
1992, 14, 163.
(6) In the present work, this mechanism is selected against through the
use of crystalline chymotrypsin and crystalline subtilisin, whose conforma-
tions have been shown to be essentially unaffected by replacement of water
by organic solvents as the medium. For example, it is reasoned that if the
structure of chymotrypsin is nearly the same in water and in hexane, then
it also should be the same in other organic solvents because the differences
in physicochemical properties among them are less than those between water
and hexane.
(9) Each of the steps of the cycle in Scheme 1 is reversible. Single, rather
than double, arrows are used in the scheme solely to illustrate the
directionality in the definition of the energetic terms.
(10) Because the conformations of serine proteases (and most other
enzymes) are not altered in the transition state, the solvent-accessible surface
of the enzyme is assumed to be the same in ground and transition states.
(7) This hypothesis has been proposed to explain solvent-induced
variations in enzymatic selectivity: (a) Nakamura, K.; Takebe, Y.; Kitayama,
T.; Ohno, A. Tetrahedron Lett. 1991, 32, 4941. (b) Secundo, F.; Riva, S.;
Carrea, G. Tetrahedron: Asymmetry 1992, 3, 267.
The only difference between E and ESqU is, apart from displacement of the
solvent from the active site, the addition of SU, which is unsolvated and
thus does not contribute to the solvated surface of the complex.
(11) Fersht, A. Enzyme Structure and Mechanism, 2nd ed.; Freeman:
New York, 1985.
(8) It is likely16,29 that the active center of the enzyme is occupied by at
least a few solvent molecules when a substrate molecule is not bound. These
bound solvent molecules would be in a dynamic equilibrium between the
enzyme active center and the bulk solvent, described by a binding constant
(effectively an inhibition constant, KI). If 1/(KIVM) , [S]/KM (where VM is
the molar volume of the solvent), the bound solvent would exert little effect
on the catalytic properties of the system. If, on the other hand, the opposite
were true, the solvent would act as an effective inhibitor. If such a tightly
bound solvent molecule affects each substrate binding mode equally, it
would not influence the prochiral selectivity. If the solvent molecule is
tightly bound in such a way that it only hinders one substrate binding mode,
the prochiral selectivity would be affected in a manner that could not be
predicted by the treatment used in the present work.
(12) The free energy of a solute dissolved in a solvent is described by G
) G° + RT ln (xγ). Because the solvent is the primary variable in our
work, the standard state is chosen as the pure liquid solute. Thus G° is
independent of the solvent, and the free energy of transfer of the solute
from solvent A to B (∆Gtr) is RT ln (xBγB/xAγA).
(13) For dilute solutions, xsolute ≈ nsolute/nsolvent ) [solute]VM, where VM
is the molar volume of the solvent which is reciprocal to its molar
concentration. For solvents A and B, xA/xB ) [solute]VMA/[solute]VMB. If
the solute concentration is kept the same in both solvents (i.e., its transfer
from A to B is done at constant molar concentration), [solute] cancels out,
yielding xA/xB ) VMA/VMB, i.e., the ratio of the molar volumes of the
solvents.

3368 J. Am. Chem. Soc., Vol. 118, No. 14, 1996
Ke et al.
Scheme 3
(kcat/KM)I
(kcat/KM)II
γ′I
γ′II(kcat/KM)I
γ′I(kcat/KM)II
log
) log
+ log
(6)
( )
[
]
[
]
γ′II
A
A
B
Equation 6 expresses the general relationship between
enzymatic selectivity and solvent-transition-state interactions.
In the present work, we specifically test eq 6 with respect to
prochiral selectivity. While the same substrate leads to both
the R and S products, the reaction proceeds through conforma-
tionally distinct transition states for the production of each
enantiomer (a pro-R transition state results in the R product,
while a pro-S transition state forms the S product). Thus,
differences between γ′I and γ′II arise from variations in transition
state solvation, not from chemical differences between two
substrates. The parameters for substrates I and II in eq 6 can
therefore be replaced with those for the pro-R and pro-S reaction
pathways to describe prochiral selectivity. Also, if B is fixed
as a reference solvent,14 the final term in eq 6 will be a constant.
Consequently, one arrives at the following equation describing
the solvent dependence of the prochiral selectivity of an enzyme
in terms of a γ′ ratio:
solvent has been recently determined by X-ray crystallography16
and found to be nearly the same as in water, thus enabling
structure-based computer modeling.
The first step in the implementation of the model is the
determination of the desolvated portions of 1 in the transition
state for each product enantiomer. Since serine proteases
suspended in organic solvents act via the ping-pong bi-bi
mechanism,17 at least two transition states are involved in each
enzyme turnover, one for the acylation by vinyl acetate to form
the acyl-enzyme intermediate and the other for the regeneration
of the free enzyme by the nucleophile. The latter transition
state leads to the production of the chiral product. Thus, two
structural models were constructed (see Methods for details) of
the transition state for the reaction of 1 with the acyl-
(kcat/KM)pro-S
(kcat/KM)pro-R
γ′pro-S
γ′pro-R
log
) log
+ const
(7)
(
)
[
]
Unlike the situation for substrate specificity, where solvent-
dependent variation in the activity coefficient ratio for the two
substrates is primarily driven by chemical differences between
the substrates, γ′ for prochiral selectivity differs for each reaction
pathway only due to differences in transition state solvation. In
this work, we calculate γ′ for both the pro-R and pro-S transition
states using a three-step procedure. First, the desolvated portion
of the substrate in the transition state is determined using
molecular modeling based on the crystal structure of the enzyme.
Second, this desolvated moiety is expressed in terms of distinct
chemical groups to yield a model compound which approximates
the portion of the substrate removed from the solvent in the
transition state. Finally, the thermodynamic activity coefficient
of this model compound is calculated using UNIFAC and then
equated to γ′. According to eq 7, knowing only γ′pro-R and γ′pro-S
for a series of solvents, it should be possible to predict the
solvent dependence of prochiral selectivity.
chymotrypsin. In the first (Figure 1A), 1 binds to the enzyme
in such a way that the S product is formed. Due to steric
constraints, the dimethoxyphenyl group in the pro-S transition
state is unable to enter the S1 binding site of chymotrypsin and
hence must extend away from the enzyme into the solvent. The
second model (Figure 1B) depicts the transition state which leads
to the R product. Unlike in the pro-S situation, the pro-R
transition state adopts a conformation in which the dimethoxy-
phenyl group is buried in the S1 binding site of the enzyme.
The determination of the solvent-accessible surface areas of 1
in each of the transition state models (Figure 2) reveals that the
entire substrate is unsolvated in the pro-R transition state, while
it is largely solvated in the pro-S transition state.
With the extent of solvation of the substrates elucidated, it is
possible to proceed with the second phase of our strategy: the
selection of model compounds which mimic the unsolvated
portion of 1 in the pro-S and pro-R transition states. Table 1
quantitatively examines the extent to which each chemical group
of the substrate is desolvated. The 2-methyl-1,3-propanediol
(3) moiety is almost equally desolvated in the pro-S and pro-R
transition states and consequently should be a constituent of
both model compounds. The methoxy and phenyl groups are
completely desolvated only in the pro-R transition state and are
Results and Discussion
As an initial test of the ability of the model described above
to predict the solvent dependence of prochiral selectivity, we
have examined the enzymatic acylation of a designed prochiral
diol, 2-(3,5-dimethoxybenzyl)-1,3-propanediol (1), by vinyl
acetate to produce the chiral monoester 3-hydroxy-2-(3,5-
dimethoxybenzyl)propyl acetate (2) (see Scheme 3). The
crystalline chymotrypsin, lightly cross-linked with glutaralde-
hyde, has been selected as a catalyst.15 Its structure in an organic
(14) This reference solvent should not be confused with the standard
state of the thermodynamic activity coefficients.12 It should further be noted
that consideration of this constant in terms of a “corrected substrate
specificity” will depend on the standard state chosen for the activity
coefficients (Janssen, A. E. M.; Halling, P. J. J. Am. Chem. Soc. 1994,
116, 9827).
(15) Such cross-linked enzyme crystals (CLECs) have been employed
as robust catalysts in synthetic transformations. (a) St. Clair, N. L.; Navia,
M. A. J. Am. Chem. Soc. 1992, 114, 7314. (b) Sobolov, S. B.; Bartoszko-
Malik, A.; Oeschger, T. R.; Montelbano, M. M. Tetrahedron Lett. 1994,
35, 7751. (c) Persichetti, R. A.; St. Clair, N. L.; Griffith, J. P.; Navia, M.
A.; Margolin, A. L. J. Am. Chem. Soc. 1995, 117, 2732. (d) Lalonde, J. J.;
Govardhan, C.; Khalaf, N.; Martinez, A. G.; Visuri, K.; Margolin, A. L. J.
Am. Chem. Soc. 1995, 117, 6845.
(16) (a) Yennawar, N. H.; Yennawar, H. P.; Farber, G. K. Biochemistry
1994, 33, 7326. (b) Yennawar, H. P; Yennawar, N. H.; Farber, G. K. J.
Am. Chem. Soc. 1995, 117, 577.
(17) (a) Zaks, A.; Klibanov, A. M. J. Biol. Chem. 1988, 263, 3194. (b)
Chatterjee, S.; Russell, A. J. Enzyme Microb. Technol. 1993, 15, 1022.

SolVent Dependence of Enzymatic Prochiral SelectiVity
J. Am. Chem. Soc., Vol. 118, No. 14, 1996 3369
Figure 1. Conformation of substrate 1 in the pro-S (A) and pro-R (B) transition states with γ-chymotrypsin. See the Methods section for details
on the construction of the molecular models.
Figure 2. Solvent-accessible surface areas of substrate 1 in the pro-S (left) and pro-R (right) transition states with γ-chymotrypsin. The solvent-
accessible surfaces (represented by dot surfaces) were calculated using the Connoly method (see the Methods section for details).
Table 1. Extent of Desolvation of 1 in the Transition State with
γ-Chymotrypsin
7. Because γ′ is defined as the activity coefficient of the
unsolvated substrate moiety in the transition state, it is ap-
proximated by the activity coefficient of the appropriate model
compound. Consequently, UNIFAC-calculated activity coef-
percentage of
desolvationa
compound
group
methoxy 1
methoxy 2
phenyl
2-methyl-1,3-propanediol
pro-S (%) pro-R (%)
ficients of 3 and 1 are reported in Table 2 as γ′pro-S and γ′pro-R
respectively.
,
35
49
14
78
100
100
100
71
A profound solvent effect on the activity coefficient ratio is
evident in Table 2: γ′pro-S/γ′pro-R varies over an order of
magnitude when the solvent is changed from diisopropyl ether
to acetonitrile. Equation 7 predicts that this variation in the
activity coefficient ratio will be reflected in the same change in
prochiral selectivity. As a test of this prediction, the prochiral
selectivity of cross-linked crystalline γ-chymotrypsin toward 1
was measured experimentally in a variety of solvents (Table 3,
first data column). In such solvents as diisopropyl ether and
cyclohexane, the enzyme strongly favors production of the
R-product. As predicted by eq 7, switching to solvents with
higher γ′pro-S/γ′pro-R ratios, such as dioxane or tetrahydrofuran,
brings about a concomitant change in prochiral selectivity,
eventually resulting in the preferential formation of the S-
product, e.g., in acetonitrile and methyl acetate (i.e., a solvent-
a The desolvated surface area of each group of the substrate in the
enzyme-bound transition state was determined using molecular models
(see Figures 1 and 2) and expressed as a percentage of the solvent-
accessible surface area of the group in the unbound substrate.
therefore only included in the pro-R model compound; their
partial desolvation in the pro-S transition state is disregarded
(see below). Thus analysis of the data in Table 1 leads to the
choice of 3 and 1 as the pro-S and pro-R model compounds,
respectively.
The final phase in the prediction of the solvent dependence
of enzymatic prochiral selectivity is the employment of the
model compounds to calculate γ′pro-R and γ′pro-S for use in eq

3370 J. Am. Chem. Soc., Vol. 118, No. 14, 1996
Ke et al.
Table 2. Thermodynamic Activity Coefficients of Compounds
Modeling the Desolvated Fractions of Substrates 1 and 4 in the
Transition States with γ-Chymotrypsin Calculated Using UNIFAC
substrate 1
substrate 4
γ′pro-S
γ′pro-R γ′pro-R γ′pro-S γ′pro-R γ′pro-R
γ′pro-S
solvent
γ′pro-S
diisopropyl ether 73.5
99.6
69.4
266
2.37
4.59
2.65
16.1
11.9
0.738 73.8
0.741
36.2
2.04
dibutyl ether
cyclohexane
dioxane
51.4
480
5.09
1.80
2.15
2.94
3.07
4.39
5.62
5.09
13.5
3.36 1.51
5.16 2.62
tert-butyl acetate 13.5
tetrahydrofuran
p-xylene
toluene
8.13
70.7
66.9
methyl acetate
propionitrile
benzene
7.25
3.10
67.6
2.71
0.986 7.35
0.401 7.73
7.25
67.6
2.14 3.39
7.28
9.29
14.6
4.63
acetonitrile
0.278 9.75
induced inversion of chymotrypsin’s prochiral selectivity is
observed). Indeed, the prochiral selectivity as a function of the
solvent varies over the order of magnitude range predicted by
eq 7.
Equation 7 further predicts that a double-logarithmic plot of
(kcat/KM)pro-S/(kcat/KM)pro-R vs γ′pro-S/γ′pro-R will produce a linear
correlation. Such a plot, presented in Figure 3A, does indeed
yield a reasonable linear dependence (R2 ) 0.94). The slope
of the regression line of Figure 3A is 1.1, close to the predicted
value of 1.0.
Several sources of error are inherent to this treatment and
may contribute to the 10% deviation of the slope from unity.
First, the entire dimethoxyphenyl group of 1 in the pro-S
transition state is treated as solvated. In reality, a fraction of
its surface is partially desolvated (see Table 1), making it
impossible to exactly represent the desolvated substrate moiety
as a complete molecule. Second, there are unavoidable errors
associated with UNIFAC, which produces a mere statistical
estimate of the activity coefficient.4
Because most enzymes used in organic solvents to date have
been prepared in the form of lyophilized powders,1 it is
important to ascertain whether the thermodynamic analytical
methodology developed herein is applicable to such and other
enzyme preparations besides cross-linked crystals. The deter-
mination of γ′ is contingent on the ability to predict the specific
interactions between the substrate and the enzyme. Hence, even
a minor conformational perturbation of the enzyme active site
could change the solvation of the transition state and thus
invalidate the calculation of γ′. Recent hydrogen isotope
exchange/NMR18 and FTIR19 experiments indicate that revers-
ible conformational changes occur in proteins upon dehydration.
Thus, we reasoned that lyophilized or precipitated (from water
with cold acetone) chymotrypsin may not adhere to the solvent
dependence of prochiral selectivity predicted on the basis of
the enzyme crystal structure.
Figure 3. Dependence of the prochiral selectivity of different prepara-
tions of γ-chymotrypsin toward substrate 1 in various organic solvents
on the ratio of the activity coefficients of the desolvated portions of
the substrate in the pro-S and pro-R transition states (see eq 7).
Ascross-linked enzyme crystals; Bslyophilized enzyme powder; Cs
acetone-precipitated enzyme powder. Solvents: asdiisopropyl ether;
bsdibutyl ether; cscyclohexane; dstert-butyl acetate; esdioxane;
fstetrahydrofuran; gsp-xylene; hstoluene; ispropionitrile; jsmethyl
acetate; ksacetonitrile. The straight lines are drawn using linear
regression; see text for discussion. For experimental details, see the
Methods section.
plots of the prochiral selectivity of lyophilized (Figure 3B) and
acetone-precipitated (Figure 3C) chymotrypsin vs γ′pro-S/γ′pro-R
do not yield the correlations obtained for structurally intact
chymotrypsin: while theory predicts that the slopes of the
aforementioned plots should be unity, those determined by linear
regression are 0.32 and 0.24, respectively. Additionally, only
a poor correlation between the prochiral selectivity of these
enzyme powders and γ′pro-S/γ′pro-R is reflected by R2 values of
0.42 for lyophilized and 0.68 for acetone-precipitated chymo-
trypsin (as compared to 0.94 for the cross-linked crystals). These
findings underscore the importance of dealing with structurally
defined enzyme catalysts if a quantitative rationale involving
stereoselectivity is sought after.
While our model requires the enzyme to be in a native (or at
least a known) conformation in organic solvents, it places no
further constraints on the catalyst. Therefore, the proposed
analysis should be applicable to enzymes other than chymo-
trypsin. To test this prediction, the acylation of 1 by vinyl
acetate, catalyzed by cross-linked crystalline subtilisin Carls-
berg20 (henceforth referred to as subtilisin), has been investigated
using the same methodology as with the chymotrypsin-catalyzed
reaction.
When the prochiral selectivity of lyophilized or acetone-
precipitated chymotrypsin is measured for substrate 1 in several
solvents (Table 3), the selectivity varies less than 3-fold, while
the solvent effect predicted by eq 7 assuming an intact enzyme
structure spans a factor of 13. Moreover, double-logarithmic
(18) (a) Desai, U. R.; Osterhout, J. J.; Klibanov, A. M. J. Am. Chem.
Soc. 1994, 116, 9420. (b) Desai, U. R.; Klibanov, A. M. J. Am. Chem. Soc.
1995, 117, 3940.
(19) (a) Prestrelski, S. J.; Tedischi, N.; Arakawa, T.; Carpenter, J. F.
Biophys. J. 1993, 65, 661. (b) Prestrelski, S. J.; Tedischi, N.; Arakawa, T.;
Carpenter, J. F. Arch. Biochem. Biophys. 1993, 303, 465. (c) Costantino,
H. R.; Griebenow, K.; Mishra, P.; Langer, R.; Klibanov, A. M. Biochim.
Biophys. Acta 1995, 1253, 69. (d) Griebenow, K.; Klibanov, A. M. Proc.
Natl. Acad. Sci. U.S.A. 1995, 92, 10969.
(20) Whose X-ray crystal structure in an anhydrous solvent was found
to be virtually indistinguishable from that in water.

SolVent Dependence of Enzymatic Prochiral SelectiVity
J. Am. Chem. Soc., Vol. 118, No. 14, 1996 3371
Table 3. Solvent Dependence of the Prochiral Selectivity of
γ-Chymotrypsin, Prepared by Different Methods, toward Substrate 1
Table 5. Solvent Dependence of Prochiral Selectivity of
Cross-Linked Crystalline Subtilisin Carlsberg for Substrate 1
a
a
(kcat/KM)pro-S/(kcat/KM)pro-R
(kcat/KM)pro-S
cross-linked
crystallineb
acetone-
solvent
(kcat/KM)pro-R
solvent
lyophilizedc
precipitatedd
diisopropyl ether
cyclohexane
tert-butyl acetate
acetonitrile
1.2
1.2
1.6
2.2
2.3
diisopropyl ether
dibutyl ether
cyclohexane
tert-butyl acetate
dioxane
tetrahydrofuran
p-xylene
toluene
propionitrile
acetonitrile
0.10
0.17
0.28
0.32
0.38
0.45
1.0
1.3
1.4
1.9
2.2
0.98
0.67
0.91
1.7
2.3
1.6
1.3
1.2
2.1
2.9
0.63
0.77
0.67
toluene
a (kcat/KM)pro-S/(kcat/KM)pro-R values were determined from the ratio of
initial velocities for the production of each enantiomer as described in
the Methods section. The initial velocities, vS and vR, respectively,
observed for 2.5 mg/mL crystalline subtilisin Carlsberg were (in µM/
h) as follows: diisopropyl ethers28 and 23; cyclohexanes20 and 17;
tert-butyl acetates4.3 and 2.7; acetonitriles3.1 and 1.4; toluenes18
and 7.8.
1.2
1.0
1.5
methyl acetate
1.2
a (kcat/KM)pro-S/(kcat/KM)pro-R values were determined from the ratio of
initial velocities for the production of each enantiomer as described in
the Methods section. b The initial velocities, vS and vR, respectively,
observed for 5 mg/mL crystalline γ-chymotrypsin were (in µM/h) as
follows: diisopropyl ethers0.52 and 5.2; dibutyl ethers0.34 and 2.0;
cyclohexanes0.74 and 2.6; tert-butyl acetates0.068 and 0.21; diox-
anes0.63 and 1.6; tetrahydrofurans1.4 and 3.1; p-xylenes5.6 and 5.6;
toluenes31 and 24; propionitriles4.7 and 3.3; acetonitriles2.5 and
1.3; methyl acetates1.8 and 0.82. c The initial velocities, vS and vR,
respectively, observed for 15 mg/mL lyophilized γ-chymotrypsin were
(in µM/h) as follows: diisopropyl ethers44 and 45; dibutyl ethers15
and 23; cyclohexanes4.7 and 5.4; tert-butyl acetates14 and 24;
dioxanes3.8 and 8.6; tetrahydrofurans11 and 18; p-xylenes120 and
96; toluenes53 and 45; propionitriles6.7 and 3.2; acetonitriles2.4
and 8.1; methyl acetates11 and 9.2. d The initial velocities, vS and vR,
respectively, observed for 15 mg/mL acetone-precipitated γ-chymo-
trypsin were (in µM/h) as follows: diisopropyl ethers32 and 53; dibutyl
ethers19 and 24; cyclohexanes7.9 and 12; toluenes73 and 61;
acetonitriles28 and 27; methyl acetates18 and 12.
Table 6. Solvent Dependence of Prochiral Selectivity of
Cross-Linked Crystalline γ-Chymotrypsin for Substrate 4
a
(kcat/KM)pro-S
solvent
dioxane
diisopropyl ether
tert-butyl acetate
methyl acetate
benzene
(kcat/KM)pro-R
0.58
0.64
0.88
1.2
1.3
a (kcat/KM)pro-S/(kcat/KM)pro-R values were determined from the ratio of
initial velocities for the production of each enantiomer as described in
the Methods section. The initial velocities, vS and vR, respectively,
observed for 5 mg/mL crystalline γ-chymotrypsin were (in µM/h) as
follows: dioxanes1.0 and 1.8; diisopropyl ethers2.3 and 3.6; tert-
butyl acetates3.0 and 3.4; methyl acetates11 and 9.5; benzenes16
and 12.
Table 4. Extent of Desolvation of 1 in the Transition State with
Subtilisin Carlsberg
chymotrypsin. Modeling of the pro-R and pro-S transition states
and subsequent calculation of the unsolvated areas of the
substrates reveals that, as was the case with 1, while the
2-methyl-1,3-propanediol fragment is unsolvated in both transi-
tion states, the phenyl group is only protected from the solvent
in the pro-R. Consequently, the model compounds representing
the unsolvated portions of 4 in the pro-S and pro-R transition
states are 3 and 4, respectively.
percentage of
desolvationa
compound
group
methoxy 1
methoxy 2
phenyl
2-methyl-1,3-propanediol
pro-S (%) pro-R (%)
9
54
55
56
14
60
52
62
Table 2 (last column) lists the γ′ ratios for substrate 4 in some
of the same solvents used for 1. Using these γ′ ratios, our model
predicts that 4 will behave differently than substrate 1 in two
respects. First, the γ′ ratio varies only 3-fold for 4, whereas it
spans a 12-fold range for 1. Thus the prochiral selectivity for
substrate 4 should be similarly less solvent-sensitive. Second,
the γ′ ratio for 4 is lower in dioxane than in diisopropyl ether,
while the opposite is true for 1. An analogous inversion in the
ranking of these solvents with respect to prochiral selectivity is
predicted by eq 7 upon switching from substrate 1 to 4.
The measured prochiral selectivities of chymotrypsin in the
acetylation of 4 are presented in Table 6. As predicted, the
solvent effect on selectivity is much smaller for substrate 4 than
for 1, spanning less than 3 fold vs 22 fold. Also, the predicted
inversion of the ranking of dioxane and diisopropyl ether with
respect to prochiral selectivity is indeed observed. In further
corroboration of eq 7, the double-logarithmic plot of prochiral
selectivity vs γ′ ratio (Figure 4) is linear (with a satisfactory
correlation coefficient of 0.92) and has a slope of 0.82.
a The desolvated surface area of each group of the substrate in the
enzyme-bound transition state was determined using molecular models
(similar to those in Figures 1 and 2) and expressed as a percentage of
the solvent-accessible surface area of the group in the unbound substrate.
Because the S1 binding site of subtilisin is much more shallow
than chymotrypsin’s, no portion of substrate 1 in the either pro-S
or pro-R transition state is fully shielded from the solvent.
Furthermore, calculation of the unsolvated surfaces of the groups
which make up the substrate reveals that there is little difference
in the extent of desolvation of the two transition states (Table
4). Due to the analogous solvation patterns, the model
compounds for the calculation of γ′pro-R and γ′pro-S, whatever
they may be, should be identical, resulting in the prediction by
eq 7 that the prochiral selectivity of subtilisin in the acetylation
of 1 will be independent of the solvent. In support of this
prediction, when (kcat/KM)pro-S/(kcat/KM)pro-R for subtilisin is
measured in a range of solvents which span a 19-fold change
in the prochiral selectivity of chymotrypsin for the same
substrate, less than a 2-fold variation is observed (Table 5).
In addition to being applicable to various enzymes, our
treatment contains no restraints on the substrate. To verify its
applicability to other substrates, we have examined the acylation
of 2-benzyl-1,3-propanediol (4) (i.e., 1 devoid of both methoxy
groups) by vinyl acetate catalyzed by cross-linked crystalline
Concluding Remarks
When crystalline enzymes are used as asymmetric catalysts
in anhydrous organic solvents, the solvent dependence of
enzymatic prochiral selectivity can be attributed primarily to
changes in the relative solvation energies for the pro-R and pro-S

3372 J. Am. Chem. Soc., Vol. 118, No. 14, 1996
Ke et al.
NMR (CDCl3) δ 6.2-6.3 (3 H, m), δ 3.71 (6 H, s), δ 3.67 (6 H, s), δ
3.64 (1 H, t, J ) 7.4 Hz), δ 3.11 (2 H, d, J ) 7.4 Hz).
1 was prepared by dissolving 0.56 g (2 mmol) of dimethyl 2-(3,5-
dimethoxybenzyl)malonate in 10 mL of dry THF, followed by a
dropwise addition to 8 mL of a 1.0 M LiAlH4 solution in ether at 0
°C. The mixture was stirred under argon overnight at room temperature,
and then 0.5 mL of water was added to quench the reaction. The
product was extracted using ether and, after rotary evaporation, purified
by flash column chromatography (3:1 (v/v) ethyl acetate:hexane). The
yield of 1 was 0.27 g (60% of theor.). 1H NMR (CDCl3) δ 6.2-6.4 (3
H, m), δ 3.76 (2 H, dd, J ) 10.2, 3.9 Hz), δ 3.73 (6 H, s), δ 3.65 (2
H, dd, J ) 10.2, 6.9 Hz), δ 2.55 (2 H, d, J ) 7.6 Hz), δ 2.06 (1 H, m),
δ 2.03 (2 H, s). Anal. Calcd for C12H18O4: C, 63.70; H, 8.02; O,
28.28. Found: C, 62.60; H, 8.10; O, 29.63.
2 (a racemate used to calibrate the HPLC instrument) was prepared
by dissolving 0.11 g (0.5 mmol) of 1 in 10 mL of dry ether, followed
by addition of 0.05 mL of triethylamine at 0 °C and 0.05 mL of acetyl
chloride dissolved in 5 mL of dry ether. The reaction was monitored
by TLC and quenched by addition of 1 mL of water. The product was
extracted with ethyl acetate and, after rotary evaporation, purified by
flash column chromatography (1:1 (v/v) ethyl acetate:hexane). The
yield was 54 mg (40% of theor.). 1H NMR (CDCl3) δ 6.2-6.3 (3 H,
m), δ 4.15 (1 H, dd, J ) 12, 4.2 Hz), δ 4.04 (1 H, dd, J)12, 6.4 Hz),
δ 3.73 (6 H, s), δ 3.56 (1 H, dd, J ) 12, 4.2 Hz), δ 3.47 (1 H, dd, J
) 12, 6.6 Hz), δ 2.57 (1 H, dd, 13, 7.3 Hz), δ 2.52 (1 H, dd, J ) 13,
6.6 Hz), δ 2.1(1 H, m), δ 2.05 (3 H, s), δ 1.66 (1 H, m). Anal. Calcd
for C14H20O5: C, 62.67; H, 7.51; O, 29.82. Found: C, 62.03; H, 7.62;
O, 29.92.
4 was prepared by dissolving 0.50 g (2 mmol) of diethyl benzyl-
malonate in 10 mL of dry THF, and adding this solution dropwise to
8 mL of 1.0 M LiAlH4 dissolved in ether at 0 °C. The mixture was
stirred under argon overnight at room temperature, and then 0.5 mL of
water was added to quench the reaction. The product was extracted
using ether and, after rotary evaporation, purified by flash column
chromatography (3:1 (v/v) ethyl acetate:hexane). The yield was 0.20
g (60% of theor.). 1H NMR (CDCl3) δ 7.1-7.4 (5 H, m), δ 3.85 (2 H,
dd, J ) 11, 3.6 Hz), δ 3.72 (2 H, dd, J ) 11, 6.7 Hz), δ 2.67 (2 H, d,
J ) 7.6 Hz), δ 2.1 (1 H, m), δ 2.08 (2 H, s). Anal. Calcd for
C10H14O2: C, 72.26; H, 8.49; O, 19.25. Found: C, 72.50; H, 8.85; O,
17.77.
Figure 4. Dependence of the prochiral selectivity of γ-chymotrypsin
toward substrate 4 in various organic solvents on the ratio of the activity
coefficients of the desolvated portions of the substrate in the pro-S
and pro-R transition states. Solvents: asdioxane; bsdiisopropyl ether;
cstert-butyl acetate; dsmethyl acetate; esbenzene. See the legend to
Figure 3 for other details.
binding modes of the substrate in the transition state. This work
presents a quantitative model which satisfactorily predicts the
solvent effect on prochiral selectivity solely on the basis of these
solvation energies. Thus other factors not considered by the
model, e.g., the effect of the solvent on the enzyme or
displacement of bound solvent molecules from the active site
by the substrate, are deemed relatively unimportant. While this
model performs reasonably well with crystalline enzymes, its
implementation with respect to amorphous (lyophilized or
acetone-precipitated) enzyme powders is not possible due to
the ill-defined, reversibly denatured structure of proteins in such
dehydrated states. Because no specific assumptions are made
regarding the enzyme or the substrate in the derivation of the
equation, the model should be generally applicable. Moreover,
our treatment should be similarly applicable to other types of
enzymatic selectivity, such as enantioselectivity, chemoselec-
tivity, regioselectivity, and substrate selectivity.
5 (a racemate used to calibrate the HPLC instrument) was prepared
by dissolving 83 mg (0.5 mmol) of 4 in 10 mL of dry ether, followed
by addition of 0.05 mL of triethylamine at 0 °C and 0.05 mL of acetyl
chloride dissolved in 5 mL of dry ether. The reaction was followed
by TLC and quenched after 20 min with 1 mL of water. The product
was extracted with ethyl acetate and purified by flash column
chromatography (1:1 (v/v) ethyl acetate:hexane). The yield of 5 was
45 mg (43% of theor.). 1H NMR (CDCl3) δ 7.35 (5 H, m), δ 4.28 (1
H, dd, J ) 4.6, 11.3 Hz), δ 4.18 (1 H, dd, J ) 6.4, 11.2 Hz), δ 3.70
(1 H, dd, J ) 4.6, 11.2 Hz), δ 3.60 Hz (1 H, dd, J ) 6.2, 11.3 Hz), δ
2.77 (1H, dd, J ) 7.8, 11.5 Hz), δ 2.73 (1 H, dd, J ) 7.8, 11.5 Hz),
δ 2.2 (1 H, m), δ 2.18 (3H, s), δ 1.94 (1H, m). Anal. Calcd for
C12H16O3: C, 69.21; H, 7.74; O, 23.05. Found: C, 68.21; H, 7.88; O,
24.01.
Materials and Methods
Enzymes. Rhizomucor miehei lipase (EC 3.1.1.3) was obtained from
Fluka. Subtilisin Carlsberg (alkaline protease from Bacillus licheni-
formis, EC 3.4.21.14) and bovine pancreatic γ- and R-chymotrypsins
(EC 3.4.21.1) were purchased from Sigma Chemical Co. Note that
the R- and γ-forms of chymotrypsin are covalently identical and
interconvertable in a pH-dependent manner, differing primarily in their
crystallization properties.21 γ-Chymotrypsin crystals were created from
the R- form of the enzyme, following the method of Stoddard et al.
(see the enzyme crystallization section for further details).
Chemicals and Solvents. All chemicals and solvents were pur-
chased from Aldrich Chemical Co. The organic solvents were of the
highest purity available from that vendor (analytical grade or better)
and were dried prior to use to a water content below 0.01% (as
determined by the Karl Fischer titration22) by shaking with Linde’s
3-Å molecular sieves.
Enzymatically Prepared (R)-2. In a 20-mL screw-cap scintillation
vial, 45 mg (0.2 mmol) of 1 was dissolved in 10 mL of dried diisopropyl
ether. Then, 20 mg of cross-linked γ-chymotrypsin crystals (see below)
was added to the vial, followed by addition of 20 µL of deionized
water and 0.4 mL (10 mmol) of vinyl acetate . The vial was shaken
at 45 °C and 300 rpm for 12 h. The mixture was filtered, and the
crystalline enzyme was washed 3 times with 10 mL of diisopropyl ether.
The washings were combined with the filtrate, and the subsequent
workup of the mixture was the same as described above for the
Dimethyl 2-(3,5-dimethoxybenzyl)malonate was prepared by dis-
solving 0.12 g (5 mmol) of metallic sodium in 20 mL of dry methanol,
followed by addition of 0.6 g (5 mmol) of dimethyl malonate. The
solution was cooled to 0 °C, and 0.76 g (4 mmol) of 3,5-dimethoxy-
benzyl chloride in 20 mL of dry THF was added dropwise over 30
min. The mixture was refluxed overnight under argon and cooled to
room temperature, and then 20 mL of cold water was added to quench
the reaction. The product was extracted with ethyl acetate and purified
by distillation. The yield of the product was 1.0 g (70% of theor.). 1H
28
chemically synthesized racemic 2. The yield of 2 ([R]D +20.3˚
(CHCl3), 82% ee by chiral HPLC) was 36 mg (71% of theor.);
determination of its absolute configuration is described below. 1H NMR
(CDCl3) δ 6.2-6.3 (3 H, m), δ 4.15 (1 H, dd, J ) 12, 4.2 Hz), δ 4.04
(1 H, dd, J)12, 6.4 Hz), δ 3.73 (6 H, s), δ 3.56 (1 H, dd, J ) 12, 4.2
Hz), δ 3.47 (1 H, dd, J ) 12, 6.6 Hz), δ 2.57 (1 H, dd, J ) 13, 7.3
Hz), δ 2.52 (1 H, dd, J ) 13, 6.6 Hz), δ 2.1(1 H, m), δ 2.05 (3 H, s),
δ 1.66 (1 H, m). Anal. Calcd for C14H20O5: C, 62.67; H, 7.51; O,
29.82. Found: C, 62.03; H, 7.62; O, 29.92.
(21) Cohen, G. H.; Silverton, W. E.; Davies, D. R. J. Mol. Biol. 1981,
148, 449.
(22) Laitinen, H. A.; Harris, W. E. Chemical Analysis, 2nd ed.; McGraw-
Hill: New York, 1975; pp 361-363.

SolVent Dependence of Enzymatic Prochiral SelectiVity
J. Am. Chem. Soc., Vol. 118, No. 14, 1996 3373
Enzymatically Prepared (R)-5. In a 20-mL screw-cap scintillation
vial, 62 mg (0.3 mol) of 4 was dissolved in 10 mL of dry diisopropyl
ether. Then 30 mg of lyophilized R. miehei lipase powder was added,
followed by 0.4 mL (4 mmol) of vinyl acetate. The vial was shaken
at 45 °C and 300 rpm for 12 h. The mixture was filtered, and the
enzyme powder was washed 3 times with 10 mL of diisopropyl ether.
The washings were combined with the filtrate, and the subsequent
workup of the mixture was the same as described above for the
chemically synthesized racemic 5. The yield of (R)-5 was 34 mg (55%
chromophore, an additional one was added to facilitate the use of CD.
To this end, the free hydroxyl group in 2 was acylated by 9-anthroyl
chloride. The absolute configuration of the enzymatically prepared 5
was assigned to be R by comparison of optical rotation data with the
literature value, and (R)-5 was used as a reference compound for 2.
The CD spectrum of the anthroate derivative of (R)-5 was measured in
hexane to validate this method.
The molecular models of the anthroate derivatives28 of (R)-5 and 2
were constructed using the Insight II and Discover programs. The initial
structures were energy-minimized using the steepest descent method
for 100 iterations, followed by conjugate gradient minimization until
the maximum derivative was less than 0.001 kcal/Å. In the low energy
conformation of the derivative of (R)-5, the alignment of the electronic
transitions in the two chromophores (1Bb transition for the anthroate
chromophore and CT transition for the benzyl group) shows positive
chirality, which stands for positive first and negative second Cotton
effects in the CD spectrum. The situation is the same for the derivative
of (R)-2 and the opposite for that of (S)-2.
28
of theor.). The observed [R]D +24.4˚ (CHCl3) (83% ee by chiral
24
HPLC) is in agreement with the value of [R]D +24.2° (86% ee)
reported by Tsuji et al.23 1H NMR (CDCl3) δ 7.35 (5 H, m), δ 4.28 (1
H, dd, J ) 4.6, 11.3 Hz), δ 4.18 (1 H, dd, J ) 6.4, 11.2 Hz), δ 3.70
(1 H, dd, J ) 4.6, 11.2 Hz), δ 3.60 Hz (1 H, dd, J ) 6.2, 11.3 Hz), δ
2.77 (1H, dd, J ) 7.8, 11.5 Hz), δ 2.73 (1 H, dd, J ) 7.8, 11.5 Hz),
δ 2.2 (1 H, m), δ 2.18 (3H, s), δ 1.94 (1H, m). Anal. Calcd for
C12H16O3: C, 69.21; H, 7.74; O, 23.05. Found: C, 68.71; H, 7.88; O,
23.51.
The CD spectrum of the derivative of (R)-5 features a positive first
Cotton effect around 250 nm and a negative second Cotton effect around
210 nm, indicating positive chirality alignment of the two chro-
mophores. This result is consistent with the known absolute config-
uration of (R)-5. In the CD spectrum of the derivative of 2, there is a
strong positive first Cotton effect at 252 nm and a negative second
Cotton effect at 212 nm, indicating that the absolute configuration of
enzymatically prepared 2 is R.
Structural Modeling. The enzyme crystal structures used were
those of γ-chymotrypsin in hexane16 (Brookhaven entry 1GMD) and
subtilisin Carlsberg in acetonitrile29 (Brookhaven entry 1SCB). Because
the transition state for the acylation or deacylation of a serine protease
is structurally similar to the corresponding tetrahedral intermediate for
the reaction,30 transition states were modeled as the tetrahedral
intermediates for the reactions. Such models were produced using a
two-step procedure. First, potential binding modes of the chiral
products were generated by performing molecular dynamics simulations,
followed by energy minimization. The carbonyl oxygen of the product
was tethered to the oxyanion binding site using a harmonic potential
with a force constant selected to allow widely different conformations
to be explored, while preventing the product from diffusing too far
from the enzyme. Second, each product binding mode thus identified
was used as a template for creating an initial model of the tetrahedral
intermediate. The low-energy conformation of each of these starting
models was found using molecular dynamics simulations and energy
minimizations. The lowest-energy conformer of the tetrahedral inter-
mediate was selected as the model of the transition state.
The first step (the product binding mode search) is necessary because
the covalently bound tetrahedral intermediate is sufficiently sterically
constrained that molecular dynamics simulations do not sample highly
different conformations separated by large energetic barriers. For
example, in the case of the pro-R transition state for the deacylation of
acetyl-chymotrypsin by 1, an initial tetrahedral intermediate model
which starts with the dimethoxyphenyl group bound in the S1 pocket
is unable to span the energetic barrier to sample conformations in the
S1′ binding pocket during molecular dynamics simulation. The product
binding mode study identifies both these, as well as other potential
starting structures, allowing each of these types of conformations to
be examined in the modeling of the tetrahedral intermediate.
Molecular modeling and dynamics simulations were performed with
the Insight II and Discover programs31 as follows: The initial structures
were energy-minimized using the steepest descent method for 50
iterations, followed by conjugate gradient minimization until the
maximum derivative was less than 0.001 kcal/Å. The minimized
structure was then subjected to 40 ps of molecular dynamics at 900 K
with steps of 1 fs. After each simulated picosecond, the atomic
Anthroate derivatives of 2 and (R)-5 were prepared as described
by Wiesler and Nakanishi.24 The products were purified by flash
column chromatography (1:3 (v/v) ethyl acetate:hexane) and character-
ized by 1H NMR (CDCl3). For the derivative of 2: δ 8.55 (1 H, s), δ
8.3 (2 H, m), δ 8.1 (2 H, m), δ 7.5 (4 H, m), δ 6.3 (3 H, m), δ 4.36
(1 H, dd, J ) 4.7, 11.2 Hz), δ 4.29 (1 H, dd, J ) 6.5, 11.4 Hz), δ 4.14
(1 H, dd, J ) 4.6, 11.3 Hz), δ 4.06 (1 H, dd, J ) 6.4, 11.2 Hz), δ 3.74
(6 H, s), δ 2.57 (1 H, dd, J ) 8.0, 11.6 Hz), δ 2.54 (1 H, dd, J ) 7.8,
11.4 Hz), δ 2.1 (1 H, m), δ 2.05 (3 H, s). For the derivative of (R)-5:
δ 8.57 (1 H, s), δ 8.3 (2 H, m), δ 8.0 (2 H, m), δ 7.5 (4 H, m), δ 7.3
(5 H, m), δ 4.51 (1 H, dd, J ) 4.6, 11.3 Hz), δ 4.40 (1 H, dd, J ) 6.2,
11.3 Hz), δ 4.28 (1 H, d, J ) 4.6, 11.3 Hz), δ 4.18 (1 H, dd, J ) 6.5
Hz, 11.2 Hz), δ 2.78 (1 H, dd, J ) 7.8, 11.5 Hz), δ 2.72 (1 H, dd, J
) 7.7, 11.5 Hz), δ 2.2 (1 H, m), δ 2.18 (3 H, s). Both of the derivatives
were further purified by HPLC prior to CD spectrum measurement.
Enzyme Crystallization. Crystallization of γ-chymotrypsin fol-
lowed the method described by Stoddard et al.25 except that it was
scaled up 5-fold. Crystals typically appeared within 1 week and were
harvested after 2-3 weeks. The crystals were transferred from the
mother liquor to 1.5-mL microcentrifuge tubes, with some 2 mg of
crystals placed in each tube. One milliliter of cross-linking solution
(1.5% (v/v) glutaraldehyde, 17% (w/v) Na2SO4, 30 mM sodium
cacodylate adjusted to pH 7.5 by 1 M HCl) was added to each tube.
After a brief shaking and a 20-min incubation at room temperature,
each tube was centrifuged, and the supernate was discarded. The
crystals were washed with deionized water (5 times), with 20 mM
phosphate buffer, pH 7.8 (5 times), and stored in this buffer for 18 h
at 4 °C before use. Subtilisin Carlsberg was crystallized following
the procedure of Niedhart and Petsko.26 Cross-linking of subtilisin was
accomplished in the same manner as described above for chymotrypsin.
The cross-linked enzyme crystals were washed with water, which was
subsequently removed by vacuum filtration prior to placement in organic
solvents.
Enzyme Lyophilization. γ-Chymotrypsin and R. miehei lipase were
lyophilized from 5 mg/mL solutions in 20 mM sodium phosphate buffer
(pH 7.8 for chymotrypsin, pH 7.0 for lipase). Both enzymes were
freeze-dried for at least 24 h.
Enzyme Precipitation. One hundred milligrams of γ-chymotrypsin
was dissolved in 1 mL of 20 mM Na2HPO4 solution at 4 °C. The pH
was adjusted to 7.8, followed by addition of 10 mL of acetone at 0 °C
to precipitate the enzyme. The precipitated protein was incubated at 4
°C for 20 min, centrifuged, washed three times with cold acetone, and
dried in a vacuum desiccator at 4 °C before use.
Determination of the Absolute Configuration of 2. The CD
chirality27 method was used to determine the absolute configuration at
C-2 in the enzymatically prepared 2. Because 2 has a single
(28) Harada, N.; Ono, H.; Uda, H.; Parveen, M.; Khan, N. U.-D.; Achari,
B.; Dutta, K. P. J. Am. Chem. Soc. 1992, 114, 7687.
(29) (a) Fitzpatrick, P. A.; Steinmetz, A. C. U.; Ringe, D.; Klibanov, A.
M. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 8653. (b) Fitzpatrick, P. A.;
Ringe, D.; Klibanov, A.M. Biochem. Biophys. Res. Commun. 1994, 198,
675.
(30) Warshel, A.; Naray-Szabo, G.; Sussman, F.; Hwang, J.-K. Bio-
chemistry 1989, 28, 3629.
(31) Biosym Inc.; San Diego: CA.
(23) Tsuji, K.; Terao, Y.; Achiwa, K. Tetrahedron Lett. 1989, 30, 6189.
(24) Wiesler, W. T.; Nakanishi, K. J. Am. Chem. Soc. 1990, 112, 5574.
(25) Stoddard, B. L.; Bruhnke, J.; Porter, N.; Ringe, D.; Petsko, G. A.
Biochemistry 1990, 29, 4871.
(26) Neidhart, D. J.; Petsko, G. A. Protein Eng. 1988, 2, 271.
(27) Harada, N.; Nakanishi, K. Circular Dichroic Spectroscopy Excitation
Coupling in Organic Stereochemistry; University Science Books: Mill
Valley, CA, 1983.

3374 J. Am. Chem. Soc., Vol. 118, No. 14, 1996
Ke et al.
coordinates were saved, resulting in 40 independent structures with
different conformations. The resulting structures were then minimized
as outlined above, except the minimization proceeded until the
maximum derivative was less than 0.0001 kcal/Å. During all mini-
mizations and molecular dynamics simulations, only the atoms of the
substrate and those of the catalytic triad’s serine were allowed to move,
and a cutoff distance of 11 Å was used with the CVFF force field
provided with the Discover program. Because solvent molecules and
counterions were not included in the simulations, all protein residues
were modeled in their un-ionized forms. Of the 40 minimized
structures, the lowest energy conformer was selected, and the solvent-
accessible surface area was calculated using the Connoly algorithm,
as implemented in the Insight software package.
dicted by UNIFAC are underestimated by about 30%, the activity
coefficient ratio (the quantity used in our work) is estimated to within
6%.
Activity coefficients reported in Table 2 include the effects of 100
mM vinyl acetate and 0.2% (v/v) water.
Kinetic Measurements. One milliliter of solvent containing 100
mM vinyl acetate and 10 mM prochiral diol was added to 5 mg of
crystals or 15 mg of acetone-precipitated or lyophilized enzyme. Then
0.2% (v/v) water was added to the system to enhance the rate of
enzymatic transesterification.17a In the presence of the dissolved
substrates, the amount of added water was soluble in each of the
solvents used. The vinyl acetate hydrolysis product, acetic acid, was
not detected during any of the reactions studied. Note that any
competing hydrolysis would merely reduce the concentration of acetyl-
enzyme available for reaction with the prochiral diol, equally reducing
the rate of production of both enantiomers of the chiral monoester
product, and thus leaving the prochiral selectivity unaffected. The
suspensions were shaken at 45 °C and 300 rpm. Periodically, a 10-µL
sample was withdrawn and assayed by chiral HPLC. Because the
reactions which lead to the R- and S-products take place in the same
vial, and compete for the same population of free enzyme, the ratio of
In support of the validity of the structural modeling methods
described above, we were able to use this procedure to correctly predict
the conformation of N-acetyl-L-phenylalanine trifluoromethyl ketone
in its hemiketal complex with chymotrypsin.32
Activity Coefficient Calculation. All activity coefficients were
calculated using the UNIFAC method.4 Because UNIFAC is a group
contribution method, it allows the estimation of activity coefficients in
systems for which there is no experimental data by assessing the
individual contribution of each group which makes up the system. Use
of this method requires three types of parameters for each group in the
system: the group’s surface area, the volume of the group, and
empirically determined parameters which reflect the free energy of
interaction between a given group and every other group in the system.
the initial velocities of the reactions is equal to (kcat/KM)pro-S/(kcat/KM)pro-
2a,3a
.
R
Mass transfer constraints cannot alter the measured initial velocity
ratio because both products are generated from the same substrate. Initial
velocity ratios were measured two to four times to assure reproduc-
ibility. Standard deviations for the ratios were within 13% of the mean
values.
As a test of the accuracy of the UNIFAC calculations, we compared
some activity coefficients derived from published vapor-liquid equi-
librium (VLE) data to those calculated using UNIFAC. The types of
systems for which such data are available are quite limited, but we
were able to find VLE data for two compounds (3-methylphenol and
2-methyl-1-propanol) which represent some of the functional groups
present in our model molecules in the most nonideal solvent observed
in the present work, cyclohexane.33 Interpreting the VLE data using
the Wilson equation of state, for 298 K and a mole fraction of 0.001,
the activity coefficients for 3-methylphenol and 2-methyl-1-propanol
are 47 and 29, respectively. Under identical conditions, UNIFAC
predicts an activity coefficient of 31 for 3-methylphenol, and 21 for
2-methyl-1-propanol. While the individual activity coefficients pre-
Chiral HPLC separations were performed using a Chiralcel OD-H
column and a mobile phase of 95:5 (v/v) hexane:2-propanol. A flow
rate of 0.8 mL/min separated the enantiomers of 2 with retention times
of 29 and 32 min for the R and S enantiomers, respectively. Chiral
resolution of 5 was achieved with a flow rate of 0.5 mL/min, resulting
in 30 and 32 min retention times for the respective R and S enantiomers.
The products were quantified using a UV absorbance detector tuned
to 220 nm.
Acknowledgment. This work was financially supported by
NIH Grant GM39794. We thank Ms. Jennifer L. Schmitke for
providing the cross-linked crystals of subtilisin and Prof.
Gregory Farber and Ms. Anke Steinmetz for helpful advice on
the preparation of crystalline γ-chymotrypsin.
(32) Brady, K.; Wei, A.; Ringe, D.; Abeles, R. H. Biochemistry, 1990,
29, 7600.
(33) Gmehling, J.; Onken, U.; Weidlich, U. Vapor-Liquid Equilibrium
Data Collection, Supplement 2d. DECHEMA: Frankfurt, 1982.
JA952674T