Markedly enhancing enzymatic enantioselectivity in organic solvents by forming substrate salts

Article abstract of DOI:10.1021/ja984283v

A new approach is proposed to enhance enzymatic enantioselectivity in organic solvents. It is based on the presumption that the less reactive substrate enantiomer experiences greater steric hindrances in the enzyme- bound transition state than the more re

Full text of DOI:10.1021/ja984283v

3334
J. Am. Chem. Soc. 1999, 121, 3334-3340
Markedly Enhancing Enzymatic Enantioselectivity in Organic
Solvents by Forming Substrate Salts
Tao Ke and Alexander M. Klibanov*
Contribution from the Department of Chemistry, Massachusetts Institute of Technology,
Cambridge, Massachusetts 02139
ReceiVed December 11, 1998
Abstract: A new approach is proposed to enhance enzymatic enantioselectivity in organic solvents. It is based
on the presumption that the less reactive substrate enantiomer experiences greater steric hindrances in the
enzyme-bound transition state than the more reactive one; enlarging the substrate by forming a salt with a
bulky counterion should exacerbate these hindrances disproportionately for the less reactive enantiomer, thus
increasing the enantioselectivity. This strategy was implemented and verified with several structurally diverse
chiral compounds converted to salts by treatment with numerous Brønsted-Lowry acids or bases. Enzymatic
transesterifications or hydrolyses of the resultant racemic salts were much more enantioselective than those of
the free substrates. This effect was observed in various organic solvents (but not in water, where the salts
apparently dissociate to regenerate the free substrates), using both crystalline and lyophilized enzymes
(Pseudomonas cepacia lipase and the protease subtilisin Carlsberg). As demonstrated both analytically and
preparatively, in some instances an enzyme would exhibit virtually no enantiopreference toward a free substrate
but a profound one toward its salts. All these effects were rationalized by using molecular modeling to derive
enzyme-bound transition-state structures of R and S enantiomers of both free substrates and their salts. The
areas of substrate surfaces overlapping with the enzyme calculated from these models turned out to be a
remarkably good predictor of enzymatic enantioselectivity (E)sthe greater the difference in the area of surface
overlap between the enantiomers, the higher the E value.
Introduction
selectivity. For example, in addition to the aforementioned
solvent control of enzyme enantioselectivity,8 the latter is also
a function of enzyme history.9 In this study, we propose a new
strategy for increasing the enantioselectivity of enzymes in
organic solvents. It is based on the idea that discrimination
between the two substrate enantiomers by a given enzyme can
be enhanced by temporarily enlarging the substrate via salt
formation. A bulky counterion in such a salt will exacerbate
the steric hindrances encountered by the less reactive (and hence
more sterically constrained) substrate enantiomer in the enzyme-
bound transition state, thus increasing the enantioselectivity E10
and, in turn, product e.e. Following the enzymatic resolution
and enzyme removal, the reaction mixture is separated and
treated with an acid or base to dissociate the salt and recover
the free substrate and product.
Spurred by the ever-growing demand for enantiopure phar-
maceuticals,1 as well as agricultural and other specialty chemi-
cals, the use of enzymes as asymmetric practical catalysts
continues to expand.2 A major obstacle to this expansion is that
an enzyme from a given source is often insufficiently enantio-
selective toward a particular (e.g., commercially valuable)
substrate, thus leading to its incomplete enzymatic resolution
and a poor e.e. of the resultant product.3 A traditional remedy
for this problem is to screen enzymes from other sources in the
hope of finding a more enantioselecive one.4 This empirical and
laborious approach is unappealing to organic chemists, and
therefore alternative strategies have been sought. Among such
means3,5 identified to manipulate enzymatic enantioselectivity,
temperature6 and, in the case of nonaqueous enzymology,7
solvent8 are notable for their broad applicability.
Results and Discussion
The realization that enzymes can function in organic solvents
and the current widespread use of this phenomenon7 have
offered novel opportunities for improving enzymatic enantio-
Our overall goal is to discover and exploit new, facile
approaches to enhancing the enantioselectivity of a given
(1) Stinson, S. C. Chem. Eng. News 1998, 76, No. 38.
(6) Phillips, R. S. Trends Biotechnol. 1996, 14, 13-16.
(7) Koskinen, A. M. P.; Klibanov, A. M., Eds. Enzymatic Reactions in
Organic Media; Blackie: London, 1996.
(8) Wescott, C. R.; Klibanov, A. M. Biochim. Biophys. Acta 1994, 1206,
1-9. Carrea, G.; Ottolina, G.; Riva, S. Trends Biotechnol. 1995, 13, 63-
70. Wescott, C. R.; Noritomi, H.; Klibanov, A. M. J. Am. Chem. Soc. 1996,
118, 10365-10370.
(9) Stahl, M.; Jeppsson-Wistrand, U.; Mansson, M.-O.; Mosbach, K. J.
Am. Chem. Soc. 1991, 113, 9366-9368. Noritomi, H.; Almarsson, O¨ .;
Barletta, G. L.; Klibanov, A. M. Biotechnol. Bioeng. 1996, 51, 95-98. Ke,
T.; Klibanov, A. M. Biotechnol. Bioeng. 1998, 57, 746-750.
(10) Chen, C.-S.; Fujimoto, Y.; Girdaukas, G.; Sih, C. J. J. Am. Chem.
Soc. 1982, 104, 7294-7299. Straathof, A. J. J.; Jongejan, J. A. Enzyme
Microb. Technol. 1997, 21, 559-571.
(2) Sheldon, R. A. Chirotechnology: Industrial Synthesis of Optically
ActiVe Compounds; M. Dekker: New York, 1993. Wong, C.-H.; Whitesides,
G. M. Enzymes in Synthetic Organic Chemistry; Pergamon Press: Oxford,
1994. Drauz, K.; Waldmann, H. Enzyme Catalysis in Organic Synthesis,
2nd ed.; Springer-Verlag: Berlin, 1996. Zaks, A.; Dodds, D. R. Drug DeV.
Trends 1997, 2, 513-531. Faber, K. Biotransformations in Organic
Chemistry, 3rd ed.; Springer-Verlag: Berlin, 1997.
(3) Faber, K.; Ottolina, G.; Riva, S. Biocatalysis 1993, 8, 91-132.
(4) Roberts, S. M.; Turner, N. J.; Willetts, A. J.; Turner, M. K.
Introduction to Biocatalysis Using Enzymes and Microorganisms; Cam-
bridge University Press: New York, 1995.
(5) Chen, C.-S.; Sih, C. J. Angew. Chem., Ind. Ed. Engl. 1989, 28, 695-
707. Hult, K.; Norin, T. Pure Appl. Chem. 1992, 64, 1129-1134.
10.1021/ja984283v CCC: $18.00 © 1999 American Chemical Society
Published on Web 03/17/1999

Enhancement of Enzymatic EnantioselectiVity
J. Am. Chem. Soc., Vol. 121, No. 14, 1999 3335
Table 1. Enantioselectivity of Cross-Linked Crystalline Pseudomonas cepacia Lipase toward 1 and Its Salts with Various Carboxylic Acids in
the Transesterification Depicted in Scheme 1 in Different Anhydrous Solvents and in the Hydrolysis in Watera
E (S/R)
substrate
acetonitrile
tetrahydrofuran
methyl tert-butyl ether
9.4 ( 1.2
tert-butyl acetate
20 ( 2
water
1
5.7 ( 1.0
4.8 ( 1.1b
24 ( 1b
10 ( 1
28 ( 2
1‚chloroacetic acid
1‚benzoic acid
27 ( 2b
28 ( 1b
1‚3,4,5-trimethoxycinnamic acid
38 ( 2b
40 ( 2b
37 ( 3b
> 50b,c
a The conditions for the transesterification reaction were the following: 40 mM racemic 1, 200 mM propanol, 10 mg/mL CLC lipase, and
shaking at 30 °C and 300 rpm. For the hydrolysis reaction (the last column), the conditions were the following: 10 mM aqueous phosphate buffer
(pH 7.0), 10 mM racemic 1, 0.2 mg/mL enzyme, and shaking at 4 °C and 300 rpm. Under these conditions, no appreciable reaction was detected
without enzyme in any of the solvents. The enantioselectivity E, defined as (kcat/KM)S/(kcat/KM)R, was measured by means of chiral HPLC with
Chiralcel OD-H (for transesterification) or Crownpak CR-(+) (for hydrolysis) columns. The E values presented are the mean values obtained from
at least three independent measurements, with the standard deviations indicated. The salts were obtained by mixing 1 and the corresponding acid
in water (pH 7.0), followed by extraction with ethyl acetate, rotary evaporation of the solvent, and subsequent redissolution in one of the solvents
listed in the table. For other experimental conditions, see Methods. b The initial reaction rate for the faster (S) enantiomer of 1 decreased due to salt
formation by 40%, 32%, and 46% in acetonitrile (for the salts given in the table order), and by 53%, 43%, and 50% for the 3,4,5-trimethoxycinnamic
salt in tetrahydrofuran, methyl tert-butyl ether, and tert-butyl acetate, respectively. There was no appreciable rate reduction in water. c Estimated
based on the detection limit of our chiral HPLC technique.
Scheme 1
Candida rugosa and Pseudomonas cepacia,14 the proteases were
too enantioselective (E > 50) to be accurately assayed using
our methodology, and the first lipase was unreactive. On the
other hand, P. cepacia lipase was found to be both reasonably
reactive and enantioselective not only in acetonitrile but also
in other solvents, both organic and aqueous, with E values
varying from 5 to 30 (first line in Table 1).
The E value of 5.7 observed in acetonitrile (Table 1), while
clearly reflecting the enzyme’s preference for the S enantiomer
of 1, would be insufficient for an effective resolution of the
racemate. Therefore we endeavored to improve it by using the
rationale outlined above. Specifically, a salt was formed between
the amino group of 1 and the carboxyl group of a common
organic acid, benzoic acid.16 This salt was then subjected to
the enzymatic transesterification (Scheme 1) in acetonitrile under
the same conditions as the free 1. While the reactivity of the
more reactive, S enantiomer upon salt formation with benzoic
acid decreased by less than a third, that of the R enantiomer
slumped, thereby resulting in more than a 4-fold jump in the E
(Table 1).
It was interesting to examine how the E value of a 1 salt
depends on the size of the Brønsted-Lowry acid. To this end,
we selected one carboxylic acid smaller (chloroacetic), and
another larger (3,4,5-trimethoxycinnamic), than benzoic. The
data presented in the first column of Table 1 show that the
bulkier the counterion, the greater the enzymatic enantioselec-
tivity with the resultant 1 salt. The highest enantioselectivity,
E ) 38, was observed with the salt of 3,4,5-trimethoxycinnamic
acid whose S enantiomer was still more than half as reactive as
that of 1, while the reactivity of the R enantiomer plummeted
almost 12-fold compared to that of the free base.
enzyme toward a given substrate. In doing so, as mentioned in
the Introduction, we take advantage of the opportunities
stemming from conducting enzymatic resolutions in organic
solvents instead of water.
When one substrate enantiomer is preferred by an enzyme
over the other, it means that the Gibbs free energy of the
former’s enzyme-bound transition state Gq is lower than that
of the latter. In the simplest case, this difference in the free
energies is due to greater encumbrances suffered by the less
reactive enantiomer in the enzyme active site. If so, then
complexing the substrate with a bulky agent should exaggerate
these steric difficulties (presumably to a larger extent than for
the more reactive enantiomer), disparately raise the Gq values,
and thus increase enzymatic enantioselectivity. In the present
work, we explored and validated this rationale by forming salts
between various substrates and suitable Brønsted-Lowry acids
or bases. Note that in contrast to covalent modifications of the
substrate, a salt can be readily decomposed to regenerate the
free reactants.
The first asymmetric transformation examined by us herein
was enzymatic transesterification of phenylalanine methyl ester
(1) depicted in Scheme 1. Initially, we employed cross-linked
crystalline (CLC) enzymes as catalysts11 because their structure
in organic solvents has been shown to be essentially native12
(unlike that of lyophilized enzymes13), and it was our intention
in this study to perform structure-based computer modeling of
enzyme-bound transition states of chiral substrates such as 1.
Of the four different CLC enzymes tested by us as catalysts
of the propanolysis of 1 in anhydrous acetonitrile, namely bovine
pancreatic γ-chymotrypsin, subtilisin Carlsberg, and lipases from
To obtain further insights into why P. cepacia lipase is more
enantioselective toward the salts of 1 than toward the free
substrate (Table 1), we employed the means of molecular
modeling involving molecular dynamics, energy minimization,
and van der Waals surface calculations.15 The structure of the
enzyme-bound transition state of 1 in the propanolysis reaction
depicted in Scheme 1, approximated by the corresponding
tetrahedral intermediate (see Methods), was built from the
(11) Margolin, A. L. Trends Biotechnol. 1996, 18, 223-230.
(12) Fitzpatrick, P. A.; Steinmetz, A. C. U.; Ringe, D.; Klibanov, A. M.
Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 8653-8657. Fitzpatrick, P. A.; Ringe,
D.; Klibanov, A. M. Biochem. Biophys. Res. Commun. 1994, 198, 675-
681. Schmitke, J. L.; Stern, L. J.; Klibanov, A. M. Proc. Natl. Acad. Sci.
U.S.A. 1997, 94, 4250-4255. Schmitke, J. L.; Stern, L. J.; Klibanov, A.
M. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 12918-12923.
(13) Griebenow, K.; Klibanov, A. M. Proc. Natl. Acad. Sci. U.S.A. 1995,
92, 10969-10976.
(14) Chymotrypsin was crystallized and cross-linked by us as described
previously.15 The other three cross-linked crystalline enzymes, commercially
available from Altus Biologics, Inc., were generously donated by that
company.
(15) Ke, T.; Wescott, C. R.; Klibanov, A. M. J. Am. Chem. Soc. 1996,
118, 3366-3374. Ke, T.; Klibanov, A. M. J. Am. Chem. Soc. 1998, 120,
4259-4263.
(16) The structure of all Brønsted-Lowry salts examined in this work
was confirmed by FTIR spectroscopy, as described in the Methods section.

3336 J. Am. Chem. Soc., Vol. 121, No. 14, 1999
Ke and KlibanoV
Table 2. Comparison of the Experimentally Determined
Enantioselectivities of Cross-linked Crystalline Pseudomonas
cepacia Lipase toward 1 and Its Salts with Molecular Modeling
Dataa
area of surface overlap (ASO),b Å2
substrate
E (S/R)
S
R
R-S
1
5.7 ( 1.0
4.8 ( 1.1
24 ( 1
40.1
45.6
50.9
63.7
45.4
51.4
63.7
78.1
5.3
5.8
12.8
14.4
1‚chloroacetic acid
1‚benzoic acid
1‚3,4,5-trimethoxy- 38 ( 2
cinnamic acid
a The E values are for the transesterification depicted in Scheme 1
in anhydrous acetonitrile. The molecular modeling (Figure 1) data
include the areas of the van der Waals surfaces of the substrate transition
states that overlap with those of the enzyme, as well as the difference
in that parameter between the R and S enantiomers. The E values are
taken from the first column of Table 1. The ASO values were calculated
by using the following protocol: (i) molecular models of the transition
state of the enzyme-bound substrate were constructed for both 1 and
its salts; (ii) using the models, the ASO for each enantiomer of the
substrate was derived from the difference in the van der Waals surfaces
between its unbound and enzyme-bound states. For details, see Methods.
b In addition to the ASO values, we also calculated the ∆∆Gq values
between the S and R transition state models. In the order in the table,
they were -0.36, -0.10, -1.37, and -1.24 kcal/mol. Comparison of
these numbers with the experimental ∆∆Gq values derived from the E
values in the table (-1.05, -0.94, -1.91, and -2.18 kcal/mol) revealed
only a qualitative agreement in terms of the overall trend: there was
an unexplainable systematic upward shift in all the calculated values.
Therefore, we adopted the ASO as the means to rationalize the effect
of salt formation on enzymatic enantioselectivity.
Figure 1. Molecular models of the enzyme-bound transition states of
S-1 (A) and R-1 (B), as well as of their 3,4,5-trimethoxycinnamic acid
salts (C and D, respectively). The transition states, approximated by
the relevant tetrahedral intermediates, correspond to the acylation phase
of the transesterification reaction depicted in Scheme 1 catalyzed by
Pseudomonas cepacia lipase. The molecular models were built on the
basis of the X-ray crystal structure of the lipase with use of molecular
dynamics, energy minimization, and van der Waals surface calculations,
as described in the Methods section. For clarity, only the substrate
portions of the transition states are shown; the position of the enzyme
is identically fixed in all instances (A through D). The dots correspond
to the van der Waals surfaces of the substrate transition states that
overlap with those of the enzyme. In C and D, the darkened structures
represent the 3,4,5-trimethoxycinnamic counterion of the salt, and the
dashed lines represent the putative salt bridge between the counterion
and 1.
While we used crystalline P. cepacia lipase to catalyze the
transesterification of 1 (to know the enzyme structure and thus
be able to carry out molecular modeling depicted in Figure 1),
lyophilized enzymes are still most frequently used as catalysts
in organic solvents.7 Upon lyophilization, the enzyme structure
usually (reversibly) changes,13 giving rise to altered enantiose-
lectivity when assayed in organic solvents9,15 (in water, the
reversibly denatured conformation reverts back to the native).
We found that the enantioselectivity of lyophilized lipase in
the reaction shown in Scheme 1 in acetonitrile is 3.2 ( 0.9,
i.e., lower than that of its CLC counterpart. However, this E
value can be improved to 7.8 ( 1.1 by forming the salt of 1
with 3,4,5-trimethoxycinnamic acid.
The salts of 1 listed in Table 1 were preformed in water,
extracted with ethyl acetate, and then dissolved in one of the
solvents following rotary evaporation of the ethyl acetate (see
Methods for details). Alternatively, a salt can be prepared
directly in the solvent of interest by mixing 1 and a desired
acid, followed by a 15-min refluxing. This simpler procedure
gave the same result in terms of both the structure of the salt
formed16 and the enhanced E value obtained (36 ( 2 Vs 38 (
2).
recently published crystal structure of the enzyme.17 Figure 1
shows such transition-state structures for the S and R enanti-
omers of 1 (A and B, respectively). The dotted surfaces are the
parts of the substrate transition states that overlap with the
enzyme. Figures 1C and 1D show such structures deduced for
the enantiomers of the salt of 1 and 3,4,5-trimethoxycinnamic
acid (the darkened fragment). One can see that the counterion
in the R transition state (Figure 1D) has a greater surface overlap
with the enzyme (i.e., greater steric hindrances) than the S
enantiomer.
Similar molecular models were generated for the other salts
listed in Table 1, and the calculated areas of the surface overlap
(ASO) for both free 1 and its salts are presented in Table 2.
These data show quantitatively that there is a correlation between
the lipase enantioselectivity and the difference in the ASO
between the two enantiomers (the last column in Table 2).
To ensure that the observed enhancement of the lipase
enantioselectivity toward 1 due to salt formation is not limited
to a single reaction medium, we examined this effect in
additional anhydrous solvents. It is seen in Table 1 that 3,4,5-
trimethoxycinnamic acid, which afforded the greatest E en-
hancement in acetonitrile, also increases several-fold the
enzymatic enantioselectivity in tetrahydrofuran, methyl tert-butyl
ether, and tert-butyl acetate. In contrast, there is no enantio-
selectivity enhancement (or rate changes) in water (the last
column in Table 1), presumably because the salts dissociate in
aqueous solution, and hence the chemical species reacting with
the enzyme is the protonated 1 regardless of whether its initial
form is the free base or a salt (such a dissociation should not
occur in organic solvents).
It was essential to ascertain the generality of the proposed
approach to enhancing enzymatic enantioselectivity with respect
to the substrate. To this end, we selected another substrate, tropic
acid (3); the S enantiomers of its many derivatives are potent
anti-cholinergic drugs.18 In addition to being structurally distinct
from 1, 3 acted as a Brønsted-Lowry acid (instead of a base)
in the salt formation and as a nucleophile (instead of an acyl
donor) in an enzymatic transesterification (Scheme 2).
(17) Schrag, J. D.; Li, Y.; Cygler, M.; Lang, D.; Burgdorf, T.; Hecht,
H.-J.; Schmid, R.; Schomburg, D.; Rydel, T. J.; Oliver, J. D.; Strickland,
L. C.; Dunaway, C. M.; Larson, S. B.; Day, J.; McPherson, A. Structure
1997, 5, 187-202.
(18) These include atropine, hyoscyamine, and scopolamine (Reynolds,
J. E. F., Ed. Martindale, the Extra Pharmacopoeia, 28th ed.; Pharmaceutical
Press: London, 1982; p 304).

Enhancement of Enzymatic EnantioselectiVity
J. Am. Chem. Soc., Vol. 121, No. 14, 1999 3337
Scheme 2
Table 3. Enantioselectivity of Cross-linked Crystalline
Pseudomonas cepacia Lipase toward 3 and Its Salts with Various
Amines in the Transesterification Depicted in Scheme 2 in Different
Anhydrous Solventsa
E (R/S)
substrate
acetonitrile tetrahydrofuran
toluene
3
1.1 ( 0.1
1.0 ( 0.2
1.6 ( 0.1
3‚tetrabutylammonium 2.4 ( 0.2b
Figure 2. Molecular models of the enzyme-bound transition states of
R-2 (A) and S-2 (B), as well as of their quinuclidine salts (C and D,
respectively). The transition states correspond to the deacylation phase
of the transesterification reaction depicted in Scheme 2 catalyzed by
Pseudomonas cepacia lipase. For other details, see the legend to Figure
1 (with quinuclidine being the counterion).
3‚tribenzylamine
3‚3,5-lutidine
3‚isoquinoline
3‚4-phenylpyridine
3‚1-adamantanamine
3‚quinuclidine
3.5 ( 0.3b
3.8 ( 0.3b
4.2 ( 0.4b
5.8 ( 0.4b
5.9 ( 0.4b
7.0 ( 0.6b
7.6 ( 0.5b
6.2 ( 0.5b
a The conditions for the transesterification reactions were the
following: 40 mM racemic 3, 200 mM vinyl acetate, 2 mg/mL CLC
lipase, and shaking at 30 °C and 300 rpm. Under these conditions, no
appreciable reaction was detected without enzyme in any of the solvents.
The enantioselectivity E, defined as (kcat/KM)R/(kcat/KM)S, was measured
by means of chiral HPLC with a Chiralcel OD-H column. The E values
presented are the mean values obtained from at least two independent
measurements, with the errors indicated. The salts were obtained by
mixing 3 and the corresponding equimolar amine in acetonitrile,
followed by a 15-min refluxing, rotary evaporation of the solvent, and
subsequent redissolution of the salt in one of the solvents listed in the
table. For other experimental conditions, see Methods. b The initial
reaction rate for the faster (R) enantiomer of 3 decreased due to salt
formation by 72%, 83%, 23%, 34%, 41%, 43%, and 35% in acetonitrile
(for the salts given in the table order), and by 33% and 39% for the
quinuclidine salts in tetrahydrofuran and toluene, respectively.
Table 4. Comparison of the Experimentally Determined
Enantioselectivities of Cross-linked Crystalline Pseudomonas
cepacia Lipase toward 3 and Its Salts with Molecular Modeling
Dataa
area of surface overlap (ASO), Å2
substrate
E (R/S)
R
S
S-R
3
1.1 ( 0.1
42.3
70.1
76.2
54.5
53.2
55.6
58.4
57.6
42.9
73.1
79.6
61.4
60.8
63.5
66.2
67.0
0.6
3.0
3.4
6.9
7.5
7.9
7.8
9.3
3‚tetrabutylammonium 2.4 ( 0.2
3‚tribenzylamine
3‚3,5-lutidine
3.5 ( 0.3
3.8 ( 0.3
4.2 ( 0.4
5.8 ( 0.4
5.9 ( 0.4
7.0 ( 0.6
3‚isoquinoline
3‚4-phenylpyridine
3‚1-adamantanamine
3‚quinuclidine
a The E values are for the transesterification depicted in Scheme 2
in anhydrous acetonitrile. The molecular modeling (Figure 2) data
include the areas of the van der Waals surfaces of the substrate transition
states that overlap with those of the enzyme, as well as the difference
in that parameter between the S and R enantiomers. The E values were
taken from the first column of Table 3. See footnotes to Table 2 and
Methods for more details.
P. cepacia lipase exhibited no appreciable enantioselectivity
in the acylation of 3 with vinyl acetate in acetonitrile (E ) 1.1
( 0.1, Table 3).19 We attempted to make the enzyme enantio-
selective toward 3 by using the substrate’s carboxyl group to
form salts with tertiary and quaternary amines. The data obtained
for seven such amines of diverse structures are presented in the
first column of Table 3. One can see that while all of them
cause the lipase to prefer the R enantiomer of the substrate, the
magnitude of the effect is strongly influenced by the nature of
the base. Bulky but flexible tetrabutylammonium and triben-
zylamine afford only modest E values (2.4 and 3.5, respectively).
Among structurally rigid amines, the enantioselectivity enhance-
ment grows with size to reach 7.0 for quinuclidine, which is
quite impressive given enzyme’s lack of enantiopreference for
free 3.
As with 1 (Figure 1), molecular modeling was employed to
rationalize these observations. One can see in Figure 2 (parts
A and B) and in the first line of Table 4 that the lipase-bound
R and S transition states of 3 have very similar areas of surface
overlap, which is evidently responsible for the absence of
enantioselectivity. The situation is conspicuously different with
the quinuclidine salts, where the counterion in the more reactive
R enantiomer (Figure 2C) is enveloped with far fewer dots than
the S one (Figure 2D); the quantitative ASO data (the last line
in Table 4) support these visual impressions. In addition, for
all the salts examined there is a correlation between the observed
E and ∆ASO (ASOS - ASOR) (the first and last columns in
Table 4)sthe greater the steric hindrances endured dispropor-
tionately by the less reactive enantiomer, the greater the
enzymatic enantioselectivity.
As in the case of 1, the enhancement of enantioselectivity
brought about by salt formation of 3 was manifested in various
solvents: the E rose from 1.0 to 7.6 in tetrahydrofuran and from
1.6 to 6.2 in toluene upon transition from free 3 to its
quinuclidine salt (Table 3). The results obtained with different
modes of salt formation and with lyophilized (rather than
crystalline) P. cepacia lipase also resembled those with 1. The
quinuclidine salt of 3 preformed in water, followed by extraction
(rather than by direct mixing in the nonaqueous solvent,
followed by refluxing, as in Table 3), gave the E value of 6.8
( 0.7 in acetonitrile. Likewise, the enantioselectivity of the
lyophilized lipase in that solvent rose from 1.3 ( 0.1 for the
free substrate to 4.8 ( 0.4 for its quinuclidine salt.
To verify the generality of our strategy further, we next
extended its scope not only to another, dissimilar substrate but
also to a distinct type of a reaction catalyzed by a different
enzyme. Specifically, we examine the hydrolysis (instead of
transesterification) of 2-benzylsuccinic acid 1-monomethyl ester
(5) catalyzed by the CLC protease subtilisin Carlsberg (instead
(19) Subtilisin and C. rugosa lipase were found to be catalytically inactive
in this reaction, while chymotrypsin was both active and very enantio-
selective (E ) 12.2 ( 0.2).

3338 J. Am. Chem. Soc., Vol. 121, No. 14, 1999
Ke and KlibanoV
Table 5. Enantioselectivity of Cross-linked Crystalline Subtilisin Carlsberg toward 5 and Its Salts with Various Amines in the Hydrolysis
Depicted in Scheme 3 in Different Organic Solvents Containing 1% Water and in Watera
E (S/R)
substrate
tert-amyl alcohol
methyl tert-butyl ether
2.0 ( 0.1
acetonitrile
tert-butanol
2.0 ( 0.2
2-propanol
water
5
1.5 ( 0.1
2.8 ( 0.2b
3.1 ( 0.3b
3.6 ( 0.3b
4.3 ( 0.4b
6.9 ( 0.5b
8.1 ( 0.6b
1.8 ( 0.1
2.1 ( 0.2
9.8 ( 0.9
5‚3,5-lutidine
5‚isoquinoline
5‚quinuclidine
5‚1-adamantanamine
5‚4-phenylpyridine
5‚4-(4-chlorobenzoyl)pyridine
5.9 ( 0.4b
7.0 ( 0.5b
7.3 ( 0.6b
7.8 ( 0.5b
8.4 ( 0.7b
6.5 ( 0.5b
9.6 ( 0.6b
a The conditions for the hydrolysis reaction in organic solvents were the following: 40 mM racemic 5, 1% (v/v) water, 10 mg/mL CLC subtilisin,
and shaking at 30 °C and 300 rpm. For the hydrolysis reaction in water (the last column), the conditions were the following: 10 mM aqueous
phosphate buffer (pH 7.0), 10 mM racemic 5, 0.4 mg/mL enzyme, and shaking at 4 °C and 300 rpm. Under these conditions, no appreciable
reaction was detected without enzyme in any of the solvents. The enantioselectivity E, defined as (kcat/KM)S/(kcat/KM)R, was measured by means of
chiral HPLC with a Chiralcel OD-H column. The E values presented are the mean values obtained from at least three independent measurements,
with the standard deviations indicated. The salts were obtained by mixing 5 and the corresponding equimolar amine in acetonitrile, followed by a
15-min refluxing, rotary evaporation of the solvent, and subsequent redissolution of the salt in one of the solvents listed in the table. For other
experimental conditions, see Methods. b The initial reaction rate for the faster (S) enantiomer of 5 decreased due to salt formation by 32%, 41%,
28%, 34%, 41%, and 55% in tert-amyl alcohol (for the salts given in the table order), by 33% and 39% for the 4-phenylpyridine salt in methyl
tert-butyl ether and acetonitrile, respectively, and by 53%, 45%, 60%, and 61% for the 4-(4-chlorobenzoyl)pyridine salt in methyl tert-butyl ether,
acetonitrile, tert-butanol, and 2-propanol, respectively. There was no detectable rate reduction in water.
Scheme 3
of P. cepacia lipase) as shown in Scheme 3. The enantioselec-
tivity in this reaction carried out in tert-amyl alcohol containing
1% water (to allow hydrolysis) was 1.5 (Table 5), i.e., subtilisin
exhibited a slight preference for S-5.
Six tertiary amines, either from Table 3 or new, were selected
as salt-forming bases with the carboxyl group of 5 and hence
as potential E-enhancing agents. The data obtained, presented
in Table 5, reveal that all of them indeed raise subtilisin’s
enantioselectivity, with, e.g., 4-(4-chlorobenzoyl)pyridine exert-
ing more than a 5-fold effect. Note that the order of the amines
Figure 3. Molecular models of the enzyme-bound transition states of
S-3 (A) and R-3 (B), as well as of their 4-(4-chlorobenzoyl)pyridine
in terms of their enantioselectivity enhancement in Table 5 is
distinct from that in Table 3: for example, quinuclidine, which
was the most potent agent with 3, is in the bottom half with 5.
Molecular modeling (Figure 3) proved insightful in explaining
this trend. As can be seen in Table 6, the order of the ascending
E values unmistakably correlates with the order of the increasing
∆ASO values.
Inspection of Table 5 reveals that, as with 1 and 3, salts of
5 are better enantiodifferentiated enzymatically than the free
substrate in a variety of organic solvents (all containing 1%
water). On the other hand, as with 1 (Table 1), in water the
enzymatic enantioselectivity against even the 4-(4-chloro-
benzoyl)pyridine salt was the same as that against the free 5.
The observed enzymatic enantioselectivity against a given
substrate salt was, once again, independent of how the latter
was preparedswhether via direct mixing in a solvent followed
by refluxing (Table 5), or via formation in water followed by
extraction. Likewise, the enantioselectivity of lyophilized sub-
tilisin toward 5 in tert-amyl alcohol was raised by forming the
4-(4-chlorobenzoyl)pyridine salt (from 1.7 ( 0.1 to 5.2 ( 0.4),
although to a lesser extent than in the case of the CLC enzyme
(Table 5).
salts (C and D, respectively). The transition states correspond to the
acylation phase of the hydrolysis reaction depicted in Scheme 3
catalyzed by the protease subtilisin Carlsberg. For other details, see
the legend to Figure 1 (with 4-(4-chlorobenzoyl)pyridine being the
counterion).
was enzymatically acetylated (Scheme 2) in acetonitrile,20 the
e.e. value of the acetyl-3 obtained after a 35% conversion was
an insignificant 6%. However, when the resolution was con-
ducted with the quinuclidine salt of 3 under otherwise the same
conditions, the e.e. of the resultant acetyl-3 was a much more
impressive 65%.
In closing, in this study we proposed a new approach to
making a given enzyme more enantioselective in organic
solvents toward a given substrate, with no covalent alterations,
by converting the latter into a readily reversible salt. This
approach was validated by using several structurally diverse
substrates, salt-forming Brønsted-Lowry acids and bases, and
(20) In 30 mL of anhydrous acetonitrile, 1.2 mmol of R,S-3 (or its
quinuclidine salt) and 6 mmol of vinyl acetate were dissolved. The enzymatic
acetylation was initiated by adding 60 mg of CLC P. cepacia lipase. The
reaction mixture was shaken at 300 rpm and 30 °C for 8.3 h (free substrate)
or 12.1 h (substrate salt). Then the enzyme was removed by filtration, and
the remaining solution was worked up by rotary evaporation, redissolution
in 60% ethyl acetate/40% hexane/0.1% trifluoroacetic acid (which decom-
poses the salt), and using this solvent as a mobile phase in flash silica gel
column chromatography (see Methods for details).
To illustrate the synthetic utility of our enantioselectivity
enhancement strategy, we carried out preparative resolution of
racemic 3 catalyzed by P. cepacia lipase. When the free acid

Enhancement of Enzymatic EnantioselectiVity
J. Am. Chem. Soc., Vol. 121, No. 14, 1999 3339
Table 6. Comparison of the Experimentally Determined
Enantioselectivities of Cross-linked Crystalline Subtilisin Carlsberg
toward 5 and Its Salts with Molecular Modeling Dataa
enantiomers in chiral HPLC). Using a mobile phase of 99% hexane/
1% 2-propanol/0.1% diethylamine (henceforth, all such percentages are
v/v/v) with a flow rate of 0.80 mL/min and a Chiralcel OD-H column,
R- and S-1 were separated with retention times of 34.4 and 38.6 min,
respectively.
area of surface overlap (ASO), Å2
substrate
E (S/R)
S
R
R-S
Racemic 2 was synthesized21 by dissolving 200 mg of (()-
phenylalanine in 20 mL of anhydrous propanol, followed by refluxing
with simultaneous purging with dry HCl gas. The resulting reaction
mixture was concentrated in a vacuum and cooled, and the crystals
were filtered off and recrystallized from propanol (yield of 130 mg,
5
1.5 ( 0.1
2.8 ( 0.2
3.1 ( 0.3
3.6 ( 0.3
4.3 ( 0.4
6.9 ( 0.5
27.0
34.3
40.1
36.3
39.8
42.5
45.3
27.9
36.8
43.1
40.6
44.2
47.4
50.7
0.9
2.5
3.0
4.3
4.4
4.9
5.4
5‚3,5-lutidine
5‚isoquinoline
5‚quinuclidine
5‚1-adamantanamine
5‚4-phenylpyridine
1
60% of theoretical). H NMR (CDCl3) δ 7.37 (5H, m), δ 4.20 (1H,
m), 3.84 (2H, t, J ) 7.2 Hz), 3.30 (1H, dd, J ) 6.2, 11.5 Hz), 3.19
(1H, dd, J ) 4.6, 11.4 Hz), 1.64 (2H, m), 7.3 (3H, t, J ) 7.7 Hz). To
determine the elution sequence of the enantiomers, R-2 was synthesized
the same way except for using R-Phe as a starting material. Using the
same chiral HPLC conditions as for 1, R- and S-2 had retention times
of 23.2 and 25.6 min, respectively.
Using the procedure of McKenzie and Wood,22 (()-3 was recrystal-
lized three times with equimolar quinine in ethanol to give R-3 with
an e.e. of 96%, as determined by chiral HPLC. Using a mobile phase
of 96% hexane/4% 2-propanol/0.1% trifluroacetic acid with a flow rate
of 0.50 mL/min and a Chiralcel OD-H column, R- and S-3 had retention
times of 40.7 and 46.0 min, respectively.
Racemic 4 was synthesized23 by dissolving 250 mg of (()-3 in 20
mL of dry tetrahydrofuran, and 200 mg of acetyl chloride was added
dropwise at 4 °C. The resulting reaction mixture was stirred at room
temperature overnight and then quenched by addition of 20 mL of water.
After extraction with ethyl acetate three times, the product was dried
over MgSO4, concentrated by rotary evaporation, and purified by flash
silica gel column chromatography (yield of 213 mg, 68% of theoretical).
1H NMR (CDCl3) δ 7.32 (5H, s), 4.57 (dd, 1H, J ) 9.5, 10.9 Hz),
4.34 (dd, 1H, J ) 9.5, 10.9 Hz), 3.95 (dd, 1H, J ) 5.5, 9.5 Hz), 2.03
(s, 3H). R-4 was prepared the same way from R-3 in order to determine
the HPLC elution sequence of 4. Using the same chiral HPLC
conditions as for 3, R- and S-4 had retention times of 22.6 and 20.5
min, respectively.
5‚4-(4-chlorobenzoyl)- 8.1 ( 0.6
pyridine
a The E values are for the hydrolysis depicted in Scheme 3 in tert-
amyl alcohol containing 1% water. The molecular modeling (Figure
3) data include the areas of the van der Waals surfaces of the substrate
transition states that overlap with those of the enzyme, as well as the
difference in that parameter between the R and S enantiomers. The E
values were taken from the first column of Table 5. For more details,
see footnotes to Table 2 and Methods.
solvents, as well as different enzymes and reactions catalyzed
by them. The results suggest that, at least for the instances
examined, the steric hindrances suffered by the less reactive
substrate enantiomer in the enzyme-bound transition state are
the predominant reason for enzymatic enantiodiscrimination;
when these hindrances are exacerbated further by enlarging the
substrate due to salt formation, the enantioselectivity rises. It is
also clear that the substrate salts remain intact even in the
enzyme-active sites. Molecular modeling of R and S enzyme-
bound transition states, for both the free substrates and their
various salts, reveals a surprisingly good correlation in each
substrate series between experimentally determined E values
and the calculated differential areas of surface overlap (ASO
for the less reactive enantiomer minus that for the more reactive
one). The proposed enantioselectivity enhancement strategy is
effective in organic solvents but not in water (where the salts
evidently dissociate thus reverting to the free substrates). Thus
the present work points to novel synthetic opportunities afforded
by nonaqueous enzymatic catalysis. In addition to the salt
formation explored herein, these presumably may also include
other reversible complexes stable in organic solvents but not in
water.
R-4 was also prepared enzymatically from (()-3‚quinuclidine salt
(formed by refluxing, as described below). The salt (3 mg, 1.20 mmol)
and 520 mg of vinyl acetate (6 mmol) were dissolved in 30 mL of
acetonitrile, followed by addition of 60 mg of CLC P. cepacia lipase.
The reaction mixture was shaken at 30 °C and 300 rpm. Once the overall
conversion reached 35% (12.1 h), as judged by chiral HPLC, the
reaction was stopped by filtering off the enzyme, and the filtrate was
concentrated by rotary evaporation. The residue was redissolved in a
small amount of the mobile phase (60% ethyl acetate/40% hexane/
0.1% trifluroacetic acid) to decompose the salt, purified with flash silica
gel column chromatography, and yielded 75 mg of R-4 (30% of
Materials and Methods
1
theoretical yield, 65% e.e. as determined by chiral HPLC). H NMR
(CDCl3) δ 7.34 (5H, s), 4.60 (dd, 1H, J ) 9.5, 11.1 Hz), 4.37 (dd, 1H,
J ) 9.4, 11.0 Hz), 3.95 (dd, 1H, J ) 5.5, 9.5 Hz), 2.06 (s, 3H). Using
the same procedure, the acetylation of 3 was carried out for 8.4 h, which
yielded 78 mg of R-4 (32% of theoretical yield, 6% e.e. as determined
by chiral HPLC). 1H NMR (CDCl3) δ 7.21 (5H, s), 4.61 (dd, 1H, J )
9.9, 11.4 Hz), 4.31 (dd, 1H, J ) 9.3, 11.8 Hz), 3.90 (dd, 1H, J ) 5.7,
9.2 Hz), 2.01 (s, 3H).
Using a mobile phase of 97.5% hexane/2.5% 2-propanol/0.1%
trifluroacetic acid with a flow rate of 0.80 mL/min and a Chiralcel
OD-H column, R-5, S-5, R-6, and S-6 had retention times of 27.4, 32.3,
40.7, and 54.5 min, respectively.
Enzymes. Cross-linked crystals of P. cepacia lipase, C. rugosa
lipase, and subtilisin Carlsberg were gifts from Altus Biologics, Inc.
γ-Chymotrypsin was crystallized and cross-linked with glutaraldehyde
using the previously described procedure.15 Prior to use, enzyme crystals
were filtered, washed twice with dry acetonitrile and then twice with
the corresponding solvent.
Lyophilized enzyme powders were prepared as follows: the enzyme
was dissolved (5 mg/mL) in 20 mM aqueous phosphate buffer of pH
7.0 or 7.8 (for P. cepacia lipase and subtilisin Carlsberg, respectively)
at 4 °C, then the solution was frozen using liquid N2, and lyophilized
for 48 h.
Kinetic Measurements. In the propanolysis of 1 in anhydrous
solvents (Scheme 1), 10 mg/mL of lipase, 40 mM racemic 1, and 200
mM propanol were used, while for the hydrolysis of 1 in 10 mM
aqueous phosphate buffer (pH 7.0), 10 mM racemic 1 and 0.2 mg/mL
of enzyme were used. For the acetylation of 3 (Scheme 2), 2 mg/mL
of lipase, 40 mM racemic 3, and 200 mM vinyl acetate were used. In
the hydrolysis of 5 in organic solvents (Scheme 3), 10 mg/mL of
subtilisin, 40 mM racemic 5, and 1% (v/v) of deionized water were
Chemicals and Solvents. Most of the chemicals and solvents used
in this work were purchased from Aldrich Chemical Co. and Sigma
Chemical Co. R- and S-5 were from Chirotech Technology Ltd., and
(()-6 was from Narchem Corp. Organic solvents were of the highest
purity available from Aldrich (analytical grade or better) and were dried
prior to use by shaking with Linde’s 3-Å molecular sieves.
(()-1 was prepared by dissolving 1.0 g of its hydrochloride in 40
mL of deionized water and adjusting the pH to 10. This solution was
extracted with 50 mL of ethyl acetate thrice, followed by combining
the organic layers and drying over anhydrous MgSO4. Subsequent rotary
evaporation resulted in 0.74 g of (()-1 (90% of theoretical yield). R-1
was prepared the same way (to determine the elution sequence of the
(21) Vanderhaeghe, H. J. Pharm. Pharmacol. 1954, 6, 57-59.
(22) McKenzie, A.; Wood, J. K. J. Chem. Soc. 1919, 828-834.
(23) Wolffenstein, M. Chem. Ber. 1908, 41, 730-736. Schmidt, R. J.
Pharm. Sci. 1968, 57, 443-452.

3340 J. Am. Chem. Soc., Vol. 121, No. 14, 1999
Ke and KlibanoV
used, while for the hydrolysis of 5 in 10 mM aqueous phosphate buffer
(pH 7.8), 0.4 mg/mL of enzyme and 40 mM racemic 5 were used. All
kinetic experiments in organic solvents were carried out at 30 °C, while
the hydrolysis reactions in water were carried out at 4 °C. All reaction
mixtures were shaken at 300 rpm, and the blank reactions were
monitored under identical conditions without enzyme (no appreciable
reaction was ever detected).
Periodically, 25-µL aliquots were withdrawn from reaction mixtures
and assayed by chiral HPLC with a UV detector tuned to 258 nm. A
Chiralcel OD-H column was used for most separations, except that a
Crownpak CR-(+) was used for the separation of R- and S-phenyl-
alanines. In initial rate measurements, the product conversion never
exceeded 15%. Because in the transesterifications (or hydrolyses) of
racemates the R and S substrate enantiomers compete for the same
population of the free enzyme, the ratio of their initial rates is equal to
(kcat/KM)R/(kcat/KM)S,24 i.e., to E.9 Each measurement was carried out at
least in duplicate, and the typical deviation from the mean did not
exceed 15%.
FTIR Measurements. The salt formation between the substrates
and the corresponding Brønsted-Lowry acids or bases was verified
by FTIR spectroscopy.25 Under identical conditions, the FTIR spectra
of the salt, the substrate, and the corresponding Brønsted-Lowry acid
or base were recorded. Upon salt formation, the IR absorbance peak at
around 1740 cm-1 of the participating carboxyl group typically either
disappeared completely or dropped at least 2-3-fold.25 This is due to
the disruption of the CdO bond caused by the C-O-H-N bond
network formation.25 In the case of pyridine derivatives as salt-forming
agents (some entries in Tables 3 and 5), a small peak around 1640
cm-1 was also observed assigned to aromatic N-H stretching.25c,e
The FTIR measurements were carried out in a Nicolet Magna-IR
System 550 optical bench as described previously.13 Samples were
prepared in acetonitrile the same way as in the corresponding kinetic
measurements. For 5, tert-amyl alcohol solutions were also examined
and gave the spectra identical with those in acetonitrile. In all
measurements, 100-µm spacers were used; blank solvents were used
as a background.
Salt Preparation. Typically, substrate salts were prepared as
follows: 50 mM substrate and the corresponding Brønsted-Lowry acid
or base were dissolved in acetonitrile and the solution was refluxed
for 15 min, cooled, concentrated by rotary evaporation, and dried under
vacuum overnight. Certain salts were also prepared via an alternative,
extraction procedure. For example, 50 mM 1 was dissolved in deionized
water, and the pH was adjusted to 7.0 by adding a saturated aqueous
solution of 3,4,5-trimethoxycinnamic acid. The resulting solution was
extracted thrice with equal volumes of ethyl acetate whose layers were
combined, dried over MgSO4, and concentrated in a vacuum overnight.
Similar procedures were carried out for the salts of 1 with benzoic and
chloroacetic acids, 3 with quinuclidine, and 5 with 4-phenylpyridine.
In all FTIR and kinetic experiments, the salts prepared with this method
behave identically to those prepared via the direct mixing/refluxing
method.
Structural Modeling. Molecular models were built by using the
crystal structures of P. cepacia lipase (Brookhaven Data Bank entry
2LIP) in water17 and of subtilisin Carlsberg in acetonitrile (Brookhaven
Data Bank entry 1SCB).12 The transition states for the enzymatic
acylations and deacylations were approximated by the corresponding
tetrahedral intermediates.26 Such models were obtained by using the
two-step procedure described previously15 with minor modification.
First, potential binding modes of each enantiomer of the substrate were
generated by performing molecular dynamics simulations, followed by
energy minimization. The carbonyl oxygen of the substrate was tethered
to the oxyanion binding site by using a harmonic potential with a force
constant selected to allow widely different conformations to be explored,
while preventing the substrate from diffusing too far from the enzyme.
This substrate 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. Each substrate-
binding mode thus identified was used as a template for creating an
initial model of the tetrahedral intermediate. The low-energy conforma-
tion of each of these starting models was found by using molecular
dynamics simulations and energy minimizations. The lowest-energy
conformer of the tetrahedral intermediate was selected as the model of
the transition state. In the case of modeling substrate salts, the salt-
forming group (amino or carboxyl) in the corresponding Brønsted-
Lowry acid or base was tethered to the vicinity of the substrate by
using a harmonic potential (50 kcal/(mol‚Å2)). A salt bridge was
approximated by a hydrogen bond between the substrate and the
corresponding acid or base with the hydrogen bond length of 2.0 Å
and bond angle of 180° used as starting positions.27 Each salt was
subjected to the same two-step procedure as above, and the final lowest-
energy conformer was selected as the transition state model.
The area of surface overlap (ASO) was calculated as follows: (i)
using the final conformation of the transition state model, the van der
Waals surfaces of the substrate and its salt partner were calculated
individually in the absence of enzyme using Connolly’s method28 with
a probe size of 0 Å; (ii) the same operation was repeated with the
enzyme-bound transition state of the substrate salt; and (iii) the ASO
(Tables 2, 4, and 6) was calculated as the difference between the sum
of the surface areas calculated in step (i) (no enzyme) and that calculated
in step (ii) (for the enzyme-bound situation). This parameter was used
as a measure of steric hindrances encountered by the substrate (or its
salt) in the enzyme-bound transition state (similar to the previously
used29 number of close contacts).
Acknowledgment. This work was financially supported by
the National Science Foundation (Grant No. BES-9712497).
JA984283V
(24) Fersht, A. Enzyme Structure and Mechanism, 2nd ed.; Freeman:
New York, 1985; p 112. Wescott, C. R.; Klibanov, A. M. J. Am. Chem.
Soc. 1993, 115, 1629-1631.
(25) (a) Barrow, G. M.; Yerger, E. A. J. Am. Chem. Soc. 1954, 76, 5211-
5216. (b) Yerger, E. A.; Barrow, G. M. J. Am. Chem. Soc. 1955, 77, 4474-
4478. (c) Barrow, G. M. J. Am. Chem. Soc. 1956, 78, 5802-5814. (d)
Yerger, E. A.; Barrow, G. M. J. Am. Chem. Soc. 1955, 77, 6206-6207. (e)
Johnson, S. L.; Rumon, K. A. J. Phys. Chem. 1965, 69, 74-85.
(26) Warshel, A.; Naray-Szabo, G.; Sussman, F.; Hwang, J.-K. Bio-
chemistry 1989, 28, 3629-3635.
(27) Donohue, J. J. Phys. Chem. 1952, 56, 502-510. Fuller, W. J. Phys.
Chem. 1959, 63, 1705-1717.
(28) Connolly, M. L. Science 1983, 221, 709-715.
(29) Fitzpatrick, P. A.; Ringe, D.; Klibanov, A. M. Biotechnol. Bioeng.
1992, 40, 735-740.