The mechanistic dissection of the plunge in enzymatic activity upon transition from water to anhydrous solvents

Article abstract of DOI:10.1021/ja9539958

Subtilisin Carlsberg dissolved in aqueous solution is several orders of magnitude more active than the enzyme suspended in anhydrous acetonitrile. In order to ascertain why, we employed crystalline subtilisin lightly cross-linked with glutaraldehyde as a catalyst in both aqueous and organic media. The structure of this crystalline enzyme in acetonitrile had been previously found to be virtually identical to that in water, thus ruling out solvent induced conformational changes as the cause of the enzymatic activity drop. Titration of the competent active centers of subtilisin revealed that k(cat)/K(M) is solely responsible for this activity plunge. Quantitative mechanistic analysis of the 7-order-of-magnitude difference in (cat)/K(M) values between subtilisin dissolved in water and cross-linked subtilisin crystals suspended in anhydrous acetonitrile allowed accounting for at least 5.6 orders of magnitude. This drastic decline is due to (i) a marked shift in the activity vs pH profile of the cross-linked crystalline enzyme compared to its soluble counterpart; (ii) different (far less favorable in acetonitrile than in water) energetics of substrate desolvation; and (iii) very low thermodynamic activity of water in anhydrous acetonitrile resulting in a much more rigid and thus less active enzyme.

Full text of DOI:10.1021/ja9539958

3
360
J. Am. Chem. Soc. 1996, 118, 3360-3365
The Mechanistic Dissection of the Plunge in Enzymatic
Activity upon Transition from Water to Anhydrous Solvents
Jennifer L. Schmitke, Charles R. Wescott, and Alexander M. Klibanov*
Contribution from the Department of Chemistry, Massachusetts Institute of Technology,
Cambridge, Massachusetts 02139
X
ReceiVed NoVember 28, 1995
Abstract: Subtilisin Carlsberg dissolved in aqueous solution is several orders of magnitude more active than the
enzyme suspended in anhydrous acetonitrile. In order to ascertain why, we employed crystalline subtilisin lightly
cross-linked with glutaraldehyde as a catalyst in both aqueous and organic media. The structure of this crystalline
enzyme in acetonitrile had been previously found to be virtually identical to that in water, thus ruling out solvent-
induced conformational changes as the cause of the enzymatic activity drop. Titration of the competent active centers
of subtilisin revealed that kcat/KM is solely responsible for this activity plunge. Quantitative mechanistic analysis of
the 7-order-of-magnitude difference in kcat/KM values between subtilisin dissolved in water and cross-linked subtilisin
crystals suspended in anhydrous acetonitrile allowed accounting for at least 5.6 orders of magnitude. This drastic
decline is due to (i) a marked shift in the activity Vs pH profile of the cross-linked crystalline enzyme compared to
its soluble counterpart; (ii) different (far less favorable in acetonitrile than in water) energetics of substrate desolvation;
and (iii) very low thermodynamic activity of water in anhydrous acetonitrile resulting in a much more rigid and thus
less active enzyme.
6
Introduction
and aqueous-organic mixtures, they may turn out to be the
future catalyst of choice in both aqueous and nonaqueous
enzymology.
Enzymes have been found to possess unique properties in
nonaqueous media, e.g., the ability to catalyze transformations
virtually impossible in aqueous solution. However, a major
1
Results and Discussion
drawback limiting the utility of enzymes is that their catalytic
activity in anhydrous organic solvents is drastically diminished
relative to that in water. For example, lyophilized powder of
the serine protease subtilisin Carlsberg is some 6 orders of
magnitude less active when suspended in acetonitrile than the
soluble enzyme is in water.2 A tempting explanation for this
drop in activity is that a conformational change inflicted upon
the enzyme by lyophilization and/or subsequent placement in
the solvent is responsible.3 Recently, however, it was demon-
strated that this could be no more than a minor part of the
explanation because the crystal structure of cross-linked crystals
In order to approach the explanation of the difference in
catalytic activity of enzymes in organic solvents Vs in water,
we selected a model system on which to focus our study,
namely, the difference in the catalytic activity of soluble
subtilisin in aqueous solution at pH 7.8 for the hydrolysis of
N-Ac-L-Phe-OEt and that of the CLCs of subtilisin in anhydrous
acetonitrile for the transesterification of this ester with propanol.
We found that Vmax/KM plummets more than 7 orders of
7
magnitude upon transition from the aqueous to organic system,
-
1
-8 -1
from 2.1 s to 4.5 × 10
s
(1 mg/mL of enzyme).
(
CLCs) of subtilisin in anhydrous acetonitrile was found to be
Two factors could contribute to the observed reduction in
4,5
indistinguishable from that in water, and yet the CLCs are
Vmax/K , namely, a reduction in the catalytic efficiency of the
M
4
still far less catalytically active in acetonitrile than in water.
enzyme (k /K ) or in the concentration of catalytically
cat
M
In the present study, we addressed this issue mechanistically
using the aforementioned CLCs of subtilisin as a model system.
Since the structure of this crystalline enzyme preparation is the
same in acetonitrile and in water,4 the solvent-induced con-
formational change hypothesis, while viable with the amorphous
enzyme, may not be invoked. In addition, because CLCs are
highly stable and possess other attractive properties in aqueous
competent enzyme ([E]0). Although the conformation of cross-
linked crystalline subtilisin suspended in acetonitrile is virtually
4,5
identical to that in water (indicating that subtilisin in the CLCs
is native), it is possible that only a small fraction of the active
centers in the CLCs is accessible to the substrate because of
hindering protein-protein contacts or cross-links within the
crystal, resulting in a reduced [E]0. To test for such an effect,
the percentage of catalytically competent active centers was
measured by the direct titration of Ser 221 of subtilisin’s
catalytic triad (see Experimental Section for details). The
,5
*
To whom correspondence should be addressed.
Abstract published in AdVance ACS Abstracts, March 15, 1996.
X
(1) Koskinen, A. M. P.; Klibanov, A. M., Eds. Enzymatic Reactions in
Organic Media; Blakie Academic & Professional: London, 1996. San-
taniello, E.; Ferraboschi, P.; Grisenti, P. Enzyme Microb. Technol. 1993,
(6) Lalonde, J. J.; Govardhan, C.; Khalaf, N.; Martinez, A. G.; Visuri,
K.; Margolin, A. L. J. Am. Chem. Soc. 1995, 117, 6845-6852. Persichetti,
R. A.; St. Clair, N. L.; Griffith, J. P.; Navia, M. A.; Margolin, A. L. J. Am.
Chem. Soc. 1995, 117, 2732-2737. Sobolov, S. B.; Bartoszko-Malik, A.;
Oeschger, T. R.; Montelbano, M. M. Tetrahedron Lett. 1994, 35, 7751-
7754. St. Clair, N. L.; Navia, M. A. J. Am. Chem. Soc. 1992, 114, 7314-
7316.
1
5, 367-382. Faber, K. Biotransformations in Organic Chemistry; Springer-
Verlag: Berlin, 1992. Klibanov, A. M. Acc. Chem. Res. 1990, 23, 114-
20. Dordick, J. S. Enzyme Microb. Technol. 1989, 11, 194-211. Chen,
1
C.-S.; Sih, C. J. Angew. Chem., Int. Ed. Engl. 1989, 28, 695-707.
(2) Zaks, A.; Klibanov, A. M. J. Biol. Chem. 1988, 263, 3194-3201.
3) Desai, U. R.; Osterhout, J. J.; Klibanov, A. M. J. Am. Chem. Soc.
(
1
994, 116, 9420-9422.
4) Fitzpatrick, P. A.; Steinmetz, A. C. U.; Ringe, D.; Klibanov, A. M.
Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 8653-8657.
5) Fitzpatrick, P. A.; Ringe, D.; Klibanov, A. M. Biochem. Biophys.
Res. Commun. 1994, 198, 675-681.
(7) Vmax/KM in the two systems can be directly compared as long as they
involve the same ester substrate. This is because (kcat/KM)N-Ac-L-Phe-OEt, the
bimolecular rate constant for the reaction of the free enzyme with this ester
substrate, is independent of the nucleophile substrateswhether water (in
aqueous solution) or an alcohol (in organic solvents).2
(
(
0
002-7863/96/1518-3360$12.00/0 © 1996 American Chemical Society

Enzymatic ActiVity in Organic SolVents Vs Water
J. Am. Chem. Soc., Vol. 118, No. 14, 1996 3361
M
Table 1. Values of (kcat/K )N-Ac-L-Phe-OEt for the Subtilisin-Catalyzed
Hydrolysis of N-Ac-L-Phe-OEt or Its Transesterification with
Propanol
form of subtilisina
solvent
k
cat/K
b
(M-1 s-1
)
M
4
dissolved
CLCs
CLCs
CLCs
CLCs
water (pH 7.8)
water (pH 7.8)
(9.1 ( 1.3) × 10
(3.4 ( 0.5) × 10
40 ( 20
3
octane (a
octane (a
acetonitrile (a
acetonitrile (a
w
) 1)
w
) 0.006)
0.93 ( 0.40
-
-
2
2
w
) 0.006) (2.3 ( 0.5) × 10
CLCs
w
e 0.002) (1.1 ( 0.4) × 10
a
All organic solvents contained 100 mM propanol as the nucleophile.
The values of the portion of the CLCs that is active subtilisin, used
b
to calculate [E]
14% in acetonitrile. These values were in turn used to calculate
from Vmax/K
0
, are 57 ( 20% in water, 45 ( 24% in octane, and 33
(
k
cat/K
M
M
.
fractions of active enzyme dissolved in aqueous solution and
in CLCs suspended in acetonitrile were found to be similar, 63
(
5% and 33 ( 14%, respectively, thus ruling out the possibility
that the 7-order-of-magnitude drop in activity could be due to
a decrease in [E]0. When the [E]0 values determined from the
active center titrations are used to convert Vmax/KM to kcat/KM,
the diminished activity of CLCs in acetonitrile compared to that
of the dissolved enzyme in water is unequivocally attributable
to decreased catalytic efficiency of the enzyme (Table 1).
When one compares soluble subtilisin in water to CLCs of
subtilisin in acetonitrile, two parameters have been changed
simultaneously, namely the solvent and the form of the enzyme.
In order to determine the contribution of each factor individually,
we examined the behavior of the CLCs relative to the soluble
enzyme in water, thus changing only the form of the enzyme,
while keeping the solvent the same. The kcat/KM value of the
CLCs in water at pH 7.8 was found to be 1.4 orders of
magnitude lower than that of soluble subtilisin (Table 1).
Therefore, the first question to be addressed was why this is
so.
Figure 1. Dependence of the catalytic activity of CLCs of subtilisin
in the hydrolysis of N-Ac-L-Phe-OEt in water (pH 7.8) on the fraction
(f) of the active subtilisin in the CLCs. The value of f was manipulated
by co-crystallizing mixtures having different ratios of active subtilisin
and PMSF-inactivated subtilisin. For conditions, see the Experimental
Section.
One plausible explanation is that the reaction catalyzed by
the CLCs is slowed by internal diffusion of the substrate within
them. To test this, we co-crystallized active and inactivated
(with phenylmethylsulfonyl fluoride) subtilisin at various f
values (where f is the fraction of the active subtilisin in the
CLCs) and studied the dependence of the enzymatic activity
on f. If the reaction is limited by internal diffusion, this
dependence should be convex.8 If, however, the reaction is free
of diffusional limitations (internal or external), a linear depen-
8
dence should result. As seen in Figure 1, the dependence of
the activity of the CLCs on f is essentially linear, thus ruling
out diffusional limitations as the cause of the 27-fold drop in
activity of the CLCs in water compared to the soluble enzyme.
Note that if the enzymatic activity of the CLCs of subtilisin is
not limited by diffusion in water, it certainly will not be limited
by diffusion in organic solvents, where the reaction is far slower
Figure 2. The pH dependence of the initial rate (V) of the enzymatic
hydrolysis of N-Ac-L-Phe-OEt in water for dissolved subtilisin (b) and
the CLCs of subtilisin (9). The data were fitted to the theoretical
ionization curve (- - -) by nonlinear regression, where V ) (A ×
(pH-pKa)
(pH-pKa)
1
0
)/(1 + 10
), where A is a constant. For conditions, see
(
Table 1) and diffusion faster (because the solvents’ viscosities
the Experimental Section.
8
are lower than water’s).
Another tenable reason for the difference in activity of the
CLCs compared to soluble subtilisin in water at pH 7.8 is that
the activity Vs pH profiles of the two forms of the enzyme are
distinct. Indeed, we found (Figure 2) that the pH dependence
of the activity of the CLCs is markedly shifted (by 3.7 pH units)
aqueous solution at pH 7.8 to the CLCs in dry acetonitrile can
be explained by the shift of the activity Vs pH profile of the
CLCs relative to that of the enzyme in solution.
Having accounted for 1.4 orders of magnitude, there is still
a 5.6-order-of-magnitude decrease in activity to be rationalized
that transpires upon moving the CLCs from water to anhydrous
acetonitrile. Two factors, which have previously been shown
9
compared to that of the dissolved enzyme. This shift can
readily account for the 1.4-order-of-magnitude difference in the
kcat/KM values between the cross-linked crystalline and dissolved
subtilisins at pH 7.8. Therefore, the first drop in enzymatic
activity that occurs upon transition from soluble subtilisin in
(9) Although the exact cause of the shift in the activity Vs pH profile is
not known, at least one possible explanation seems likely. In the crystalline
state, the subtilisin molecules are held in a close proximity and fixed
orientation to each other (packed in the crystal lattice). The ensuing
intermolecular electrostatic interactions could affect the pKas of the active
center’s catalytic residues, leading to the observed effect.
(8) Boudart, M.; Burwell, R. L. Techniques in Chemistry, 3rd ed.;
Wiley: New York, 1973; Vol. 6, Chapter 12. Crank, J. The Mathematics
of Diffusion; Oxford University Press: London, 1957; Chapter 5.

3
362 J. Am. Chem. Soc., Vol. 118, No. 14, 1996
Schmitke et al.
Scheme 1
Table 2. Thermodynamic Activity Coefficients γ of
N-Ac-L-Phe-OEt in Various Solvent Systems
solventa
γ
b
N-Ac-L-Phe-OEt
3
2
2
water (pH 7.8)
(2.6 ( 0.2) × 10
(8.2 ( 0.2) × 10
(8.2 ( 0.3) × 10
13 ( 1
hydrous octane (a
octane (a
w
) 1)
w
) 0.006)
acetonitrile (a
acetonitrile (a
w
) 0.006)
e 0.002)
w
13 ( 1
to influence the kinetics of enzymes suspended in organic
solvents, are likely to be responsible. The first is the difference
between the desolvation free energies of the substrate, N-Ac-
a
All organic solvents contained 100 mM propanol. b Calculated from
the solubility of the ester in the appropriate solvent system (see
Experimental Section).
L-Phe-OEt, in acetonitrile and water. Because the binding
energy of a substrate is utilized for enzyme catalysis,10 and this
the sum of the energetic terms of this alternative path produces
binding of a substrate molecule to an enzyme active center
requires the desolvation of the substrate, kcat/KM depends on
the free energy of substrate desolvation.11 The second likely
contributor is the dehydration of the CLCs, i.e., their exposure
to a much lower thermodynamic water activity (aw). For
lyophilized powders of enzymes in general, and of subtilisin
in particular,13 increasing the aw in organic solvents leads to
enhanced enzymatic activity, presumably because water acts as
a molecular lubricant,14 increasing the conformational mobility
of the enzyme. We endeavored to isolate these two factors and
elucidate their individual contributions to the decrease in
catalytic efficiency of the CLCs upon transition from water to
dry acetonitrile.
The two effects can be segregated using the following four-
staged stepladder (Table 1). Because in anhydrous acetonitrile
the aw is less than 0.002 but not known exactly, the first move
is to acetonitrile at a still low but defined aw (step 1). Salt
hydrate pairs can be used in organic solvents (where they are
insoluble) to buffer the aw to a constant value.15 We selected
the pair BaBr2‚1H2O/BaBr2 which buffers the aw to 0.006. The
next transition is from acetonitrile with aw ) 0.006 to octane
with the same aw to isolate the contribution of the substrate
desolvation energy difference (step 2); the introduction of this
intermediate solvent is necessary because it is impossible to
attain the same water activity, aw ) 1, in acetonitrile (infinitely
miscible with water) and in water. The subsequent shift from
octane with aw ) 0.006 to that with aw ) 1 addresses the water
activity effect alone (step 3). The final move from octane with
aw ) 1 to water should again indicate the role of substrate
desolvation in these solvents (step 4).
q
A
tr
E
tr
S
q
B
tr
ES
∆
G ) ∆G + ∆G + ∆G - ∆G
q
(1)
Assuming that the entire substrate is removed from the solvent
in the transition state, the solvation of E and ES is essentially
identical, provided that there is no substrate-induced confor-
mational change in the enzyme and that desolvation of the
q
12
16
tr
E
tr
ES
active center can be neglected. Thus, ∆G ≈ ∆G
q
. (If the
assumption does not hold, then the effect described below will
be less.)
Canceling the equal terms yields
q
q
B
tr
S
∆
G ) ∆G + ∆G
(2)
(3)
A
q
A
10
∆G is related to (k /K ) by
cat M A
kcat
q
A
h
κT
∆G
) -RT ln
[
( )
(
)
]
K
M A
where h, κ, and R are Planck’s, the Boltzmann, and the gas
constants, respectively, and T is the absolute temperature. The
analogous expression can be written for solvent B. ∆G can
in turn be expressed in terms of the thermodynamic activity
tr
S
1
7
coefficient of the substrate (γ):
tr
S
∆
G ) RT ln(γ /γ )
(4)
B
A
Substitution of eqs 3 and 4 into eq 2 yields:
(k /K )
γA
γB
cat M A
)
(5)
(k /K )
cat M B
The aforementioned thermodynamic effect, which arises from
the difference in the substrate desolvation energy between two
solvents and the resulting difference in reactivity toward the
This predictive model, culminating in eq 5, was tested in step
2 of the ladder. In this step, when the situation in octane with
aw ) 0.006 is compared to acetonitrile with aw ) 0.006, the
experimentally obtained kcat/KM ratio of 40 ( 19 (Table 1) was
similar to the value predicted by the activity coefficient ratio
of N-Ac-L-Phe-OEt in these two solvents, 63 ( 6 (Table 2).
These data support the thermodynamic model that predicts
changes in kcat/KM based on variations in the substrate desol-
vation energetics.
11
enzyme, is depicted by the cycle in Scheme 1. The lower
horizontal arrow represents the enzyme (E) reacting with the
q
substrate (S) in solvent A to form the transition state (ES ). In
a hypothetical, equivalent path, E and S separately partition from
solvent A into solvent B, and the transition state is formed there
and is transferred from solvent B back into A. In the cycle,
q
A
q
B
∆G
and ∆G represent the free energies of activation in
tr
tr
tr
solvents A and B, respectively, and ∆G , ∆G , and ∆G
q
are
(16) Serine proteases do not act via an induced-fit mechanism.10 The
E
S
ES
the free energies of transfer of the enzyme, substrate, and
X-ray crystal structures of subtilisin Carlsberg in its native state and that
inhibited by eglin c are indistinguishable.4
q
transition state, respectively, from A to B. Expressing ∆G as
A
(
17) The free energy of a solute dissolved in a solvent is described by
the equation G ) G° + RT ln(xγ), where x and γ are the solute mole fraction
and activity coefficient, respectively. Because the solvent is the primary
variable in our work, the standard state is chosen as the pure liquid solute.
(
10) Fersht, A. Enzyme Structure and Mechanism, 2nd ed.; Freeman:
New York, 1985; Chapter 12.
11) Wescott, C. R.; Klibanov, A. M. J. Am. Chem. Soc. 1993, 115,
629-1631.
(
Thus G° is independent of the solvent, and the free energy of transfer of
1
tr
S
the solute from solvent A to solvent B (∆G ) is RT ln(xBγB/xAγA). If the
(
12) Zaks, A.; Klibanov, A. M. J. Biol. Chem. 1988, 263, 8017-8021.
Halling, P. J. Trends Biotechnol. 1989, 7, 50-52. Halling, P. J. Enzyme
transfer is made at constant mole fraction (i.e., the volumes of the solvents
tr
S
Microb. Technol. 1994, 16, 178-206.
are directly proportional to their molar volumes), this simplifies to ∆G )
(
13) Yang, Z.; Zacherl, D.; Russell, A. J. J. Am. Chem. Soc. 1993, 115,
RT ln(γB/γA). For dilute solutions, the concentration dependence of the
activity coefficient can be disregarded.26 Thus, the variation in x which
occurs in our experiments (e.g., due to concentration variation in the
measurement of kcat/KM or differences in the molar volumes of the solvent)
is neglected.
1
2251-12257.
14) Poole, P. L.; Finney, J. L. Int. J. Biol. Macromol. 1983, 5, 308-
10.
15) Halling, P. J. Biotechnol. Tech. 1992, 6, 271-276.
(
3
(

Enzymatic ActiVity in Organic SolVents Vs Water
J. Am. Chem. Soc., Vol. 118, No. 14, 1996 3363
In addition to describing the solvent dependence of the
acylation step of subtilisin-catalyzed transesterifications, Scheme
between water and acetonitrile. (This value is obtained by
adding the effects of going from aw < 0.002 to aw ) 0.006
(0.28, step 1) and from aw ) 0.006 to aw ) 1 (1.6, step 3) in
Table 1.)
1
is equally applicable to the deacylation step. The model
predicts that a difference in the desolvation energy of the
nucleophile substrate (as opposed to the ester substrate) between
two solvents will also result in a kcat/KM difference. In
agreement with this, when a small activity coefficient ratio was
calculated, e.g., for propanol in acetonitrile Vs in tert-amyl
alcohol (2.7), the corresponding experimentally determined kcat/
KM ratio was also small (2.2). When a larger difference in (kcat/
KM)nucleophile was predicted by the activity coefficient ratio, as
is the case for 2-methoxyethanol in cumene Vs in acetonitrile
Thus only 1.4 out of the original 7 orders of magnitude are
now left unexplained. This gap is seen in Table 1 as the
difference between CLCs in water and CLCs in octane with aw
) 1, where the relative substrate desolvation energies (Table
2) predict only a 3.2-fold drop in catalytic efficiency as opposed
to the 85-fold observed. The exact origin of this disparity is
still unclear. One possibility is that the activity Vs pH profile
in organic solvents is responsible, i.e., the protonation state of
the catalytic triad’s histidine is different for CLCs suspended
in organic solvents than it is in water at pH 7.8. In fact, we
did find the ratio of (kcat/KM)N-Ac-L-Phe-OEt for CLCs at pH 7.8 Vs
pH 4.5 in water to be 34-fold; whereas the (kcat/KM)N-Ac-L-Phe-OEt
ratio in octane for CLCs that were removed from aqueous
buffers of pH 7.8 Vs pH 4.5 is only 7-fold. Alternatively, the
solvent binding in the active center could compete with the
substrate and thus inhibit subtilisin in organic solvents. Four
organic solvent molecules bind in subtilisin’s active center
(
14), the experimental ratio of the specificity constants was also
18
higher (7.6).
On the basis of all these data, one can conclude that the
difference in the desolvation energy of N-Ac-L-Phe-OEt in water
compared to dry acetonitrile can account, as indicated by γwater/
γacetonitrile (Table 2), for 2.3 orders of magnitude of the difference
in kcat/KM of the CLCs in these solvents. Hence only 3.3 orders
of magnitude out of the original 7 are now left to explain.
As discussed earlier, the water activity may have a profound
effect on the catalytic efficiency of the CLCs in organic solvents.
Indeed, as seen in step 3 of the ladder (Table 1), kcat/KM of the
CLCs in octane at aw ) 1 is 43-fold higher than that in the
same solvent at aw ) 0.006. In addition, a doubling in the
activity of the CLCs is observed in step 1, i.e., in acetonitrile
with aw ) 0.006 compared to this solvent with aw < 0.002. As
mentioned above, one of the likely explanations for the increase
in enzymatic activity with aw is that water acts as a molecular
lubricant and affords a higher conformational flexibility of the
4
,5
region in the CLCs in acetonitrile.
Also, in the γ-chymo-
trypsin crystal structure in hexane (a solvent homologous to
octane), two hexane molecules bind in the vicinity of the active
2
3
center.
Conclusions
Using a systematic approach to explain the 7-order-of-
magnitude drop in catalytic activity of CLCs of subtilisin in
anhydrous acetonitrile relative to the enzyme dissolved in
aqueous solution, we have been able to account for 5.6 of those
by the following: (i) a shift in the activity Vs pH profile of the
CLCs in water relative to the aqueous enzyme solution; (ii) the
unfavorable desolvation energy of the substrate in the organic
solvent relative to water; and (iii) the diminished conformational
flexibility of the CLCs in the solvent due to dehydration.
Consequently, this study has provided the rationale for avoiding
much of this activity loss by optimizing the pH of the solution
from which the CLCs are collected, judiciously selecting the
organic solvent to minimize the unfavorable substrate desolva-
tion energetics, and maintaining a high water activity in the
system.
1
9
enzyme resulting in enhanced catalysis. Another possibility
is that the added water increases the dielectric constant (polarity)
of the solvent and thus in the enzyme active center, with an
ensuing stabilization of the charged transition state and hence
an elevation in kcat/KM.
To distinguish between these two possibilities, we investigated
the effects of two selected additives on kcat/KM in octane. The
first, formamide, is a good molecular lubricant due to its high
propensity to form hydrogen bonds20 and the second, N-
methylacetamide, while a poorer molecular lubricant (because
it can form fewer hydrogen bonds), has a higher dielectric
constant than formamide (191 Vs 111).21 If the flexibility
hypothesis is correct, formamide should increase the observed
catalytic activity in octane to a greater extent than N-methyl-
acetamide. If the polarity hypothesis is correct, the opposite
Experimental Section
Cross-Linked Enzyme Crystals (CLCs) of Subtilisin. Subtilisin
Carlsberg (serine protease from Bacillus licheniformis, E.C. 3.4.21.14)
was purchased from Sigma Chemical Co. Crystals were grown at 30
2
2
will be the case. While the addition of formamide (at a ) 1)
raises kcat/KM by a factor of 13 (from 0.93 ( 0.40 to 12 ( 5
-
1
-1
M
s ), no enhancement occurs when N-methylacetamide is
°
C from an aqueous 330 mM cacodylate buffer, pH 5.6, saturated with
-
1
-1
added to the same activityskcat/KM ) 0.79 ( 0.30 M s .
The value of γN-Ac-L-Phe-OEt in octane is unaffected by these
Na SO .24 The crystals were cross-linked with a 1.5% glutaraldehyde
2
4
(
solution (pH 7.5, 30 mM cacodylate buffer, 13% Na
2 4
SO ), followed
additives.) These data support the hypothesis that an elevated
aw leads to a greater conformational mobility of the enzyme
that translates into a higher kcat/KM. Therefore, the depressed
enzymatic flexibility of the CLCs in dry acetonitrile (aw <
by washing with the buffer (containing no glutaraldehyde) and distilled
4
,5
water, and stored in a 20 mM phosphate buffer, pH 7.8, at 10 °C.
The average dimensions of the needle-like CLCs were (100 ( 40) µm
×
(15 ( 5) µm × (15 ( 5) µm. Note that the size of the CLCs did
not change upon transition from water to acetonitrile and octane, as
evidenced by microscopic examination. Furthermore, the unit cell
dimensions of the subtilisin CLCs are essentially the same in acetonitrile
as in water.4
0
.002) can account for 1.9 orders of magnitude of activity lost
(18) The water activity was not buffered in the solvents in which (kcat/
,5
KM)nucleophile was determined.
(19) Xu, Z.; Affleck, R.; Wangikar, P.; Suzawa, V.; Dordick, J. S.; Clark,
Substrates, Products, and Solvents. N-Ac-L-Phe-OEt, N-trans-
cinnamoylimidazole, and phenylmethylsulfonyl fluoride (PMSF) were
purchased from Sigma. The organic solvents, including alcohols,
utilized were reagent grade and were dried over 3-Å molecular (Linde)
D. S. Biotechnol. Bioeng. 1994, 43, 515-520.
(
20) Kitaguchi, H.; Klibanov, A. M. J. Am. Chem. Soc. 1989, 111, 9272-
273. Ray, A. Nature 1971, 231, 313-315.
21) Reichardt, C. SolVents and SolVent Effects in Organic Chemistry,
nd ed.; VCH: New York, 1988, Appendix Table A-1.
22) The dielectric constants for these solvent systems are 2.0, 2.1, 2.2,
9
2
(
sieves prior to use to a water content below 0.01%. Anhydrous BaBr
was purchased from Aldrich Chemical Co. The mixture of BaBr ‚
2
(
2
and 2.6 for pure octane, octane saturated with water, octane saturated with
formamide, and octane saturated with N-methylacetamide, respectively.
These values were calculated as described by: Onsager, L. J. Am. Chem.
Soc. 1936, 58, 1486-1493.
(23) Yennawar, N. H.; Yennawar, H. P.; Farber, K. Biochemistry 1994,
33, 7326-7336.
(24) Niedhart, D. J.; Petsko, G. A. Protein Eng. 1988, 2, 271-276.

3
364 J. Am. Chem. Soc., Vol. 118, No. 14, 1996
O/BaBr was prepared by placing the anhydrous salt (5 g/16.8
Schmitke et al.
1
H
2
2
and propanol and the kcat/K
M
value of propanol, where (kcat/K
M 2-
)
mmol) over pure water (75.8 mg/4.2 mmol) in a sealed vessel saturated
with water vapor.
One of the product esters, N-acetyl-L-phenylalanine 2-methoxyethyl
ester (N-Ac-L-Phe-OEtOMe), used for GC calibration, was synthesized
as follows. To a mixture of 2 g of N-Ac-L-Phe (10 mmol) in 45 mL
of benzene were added 4.5 mL of 2-methoxyethanol (60 mmol) and 4
M
methoxyethanol ) (V2-methoxyethanol/Vpropanol)(kcat/K )
propanol since both nucleo-
11
philes were present in equal concentrations.
The measurement of (kcat/K
where the a was buffered to 0.006 was carried out as follows. The
M
)N-Ac-L-Phe-OEt in octane and acetonitrile
w
solvent containing the ester (0.6-6 mM in octane; 10-100 mM in
acetonitrile) and 100 mM propanol was added to 100 mg of the mixture
2 4
drops of 95% H SO . The reaction mixture was refluxed for 3 h and
2 2 2
of BaBr ‚1H O/BaBr ; the suspension was equilibrated by shaking at
then transferred to a separatory funnel containing 100 mL of ether,
where it was washed 5 times with 50 mL of a 5% aqueous solution of
NaHCO . The ether phase was dried over anhydrous MgSO and
3 4
30 °C and 300 rpm for 90 min. One milligram of the CLCs was added,
and the resulting mixture was shaken and the reaction monitored as
described above. Again the kinetic data were fitted using nonlinear
concentrated by rotary evaporation; the product ester crystallized out
of the solution. The purity (99%) of the product N-Ac-L-Phe-OEtOMe
was verified by GC. N-Acetyl-L-phenylalanine propyl ester was
similarly synthesized from the free acid and propanol, and its purity
regression to obtain kcat/K
M
.
The (kcat/K N-Ac-L-Phe-OEt values in octane with a ) 0.006 and aadditive
M
)
w
) 1 (where the additive was formamide or N-methylacetamide) were
determined as follows. The ester (0.6-6 mM) and 100 mM propanol
(99%) was likewise confirmed.
were dissolved in octane containing 100 mg/mL of BaBr
at 30 °C. The additive (containing 100 mM propanol) was added until
the appearance of a second phase. The octane phase (1 mL) was then
added to 1 mg of the CLCs. Kinetic measurements and kcat/K
calculations were performed as above.
Finally, the determination of kcat/K
2
2 2
‚1H O/BaBr
Phenylmethylsulfonic acid (PMSOH) was synthesized from PMSF.
One gram of the latter (6 mmol) was dissolved in 10 mL of propanol,
to which 10 mL of 5 M NaOH (50 mmol) was added. The reaction
mixture was refluxed overnight. The solution was then cooled and
M
acidified with H
mL of ether. The ether phase was washed thrice with 50 mL of a 10%
solution of H SO , dried over anhydrous MgSO , and subjected to rotary
2
SO
4
, and the resultant PMSOH was extracted into 50
M
of N-Ac-L-Phe-OEt in “hydrous”
octane (a ) 1) was performed as follows. The ester (0.6-6 mM)
w
and 100 mM propanol were dissolved in 2 mL of octane, and the
solution was allowed to equilibrate with water (containing propanol)
through the vapor phase at 30 °C until the activity of water reached
unity (3 days). One milliliter of the resulting octane solution was
withdrawn and added to 1 mg of CLCs. The initial rate of the reaction
2
4
4
evaporation. The resultant solid PMSOH was 99% pure, as verified
by HPLC.
Active Center Titration. The percentage of the catalytically
competent subtilisin molecules (used to calculate [E]
0
) of the soluble
shaken at 30 °C and 300 rpm was monitored by GC, and the kcat/K
value was obtained by nonlinear curve fitting of the kinetic data.
M
enzyme preparation was determined by titration with N-trans-cin-
2
5
namoylimidazole. The concentration of the catalytically competent
subtilisin in the CLCs which is accessible to the substrate in water and
in acetonitrile was determined by titration of the active centers with
the irreversible serine protease inhibitor PMSF (to avoid multiple
turnovers due to long titration times) in two independent experiments,
each done in triplicate. CLCs (20 mg/mL) were placed in 2 mL of
either water (pH 5.0; 10 mM acetate buffer) or acetonitrile, both
containing 1 mM PMSF, and the suspension was shaken at 30 °C and
Diffusion Limitation Assay. To test whether the hydrolysis of
N-Ac-L-Phe-OEt catalyzed by the CLCs in water at pH 7.8 was limited
by diffusion, active and PMSF-inactivated subtilisin were co-crystallized
at various f values. The dependence of the activity of the resulting
partially inactive crystals as a function of f was found to be linear
(Figure 1).
(a) Inactivation of Subtilisin. To a 58-mL subtilisin solution (10
mg/mL, pH 7.8, 20 mM phosphate buffer) was added 2.0 mL of a
PMSF solution (43 mM; 6-fold molar excess) in propanol. The mixture
was incubated for 4 h at room temperature. The remaining activity of
subtilisin, assayed by measuring the initial rate of the enzymatic
hydrolysis of 1.0 mM N-Ac-L-Phe-OEt in water (pH 7.8, 100 mM KCl,
3
00 rpm. The disappearance of PMSF, as well as any PMSOH
produced by spontaneous hydrolysis, was monitored by HPLC. The
validity of this titration method was verified with a subtilisin solution
of an independently determined [E]
0
.
Kinetic Measurements. The kcat/K
M
values in water were measured
3
0 °C), was found to be less than 0.1%. The excess PMSF was then
potentiometrically for the subtilisin-catalyzed hydrolysis of N-Ac-L-
Phe-OEt (1-12 mM ester; 79 µg/L and 29 mg/L dissolved and
suspended CLCs of subtilisin, respectively; pH 7.8; 30 °C; 100 mM
KCl). Initial rate data were fitted to the Michaelis-Menten equation
using the nonlinear curve-fitting function of SigmaPlot (Jandel Scien-
tific). The pH dependence of the soluble subtilisin (1.3-17 µg/L) and
of the CLCs (15-190 mg/L) was also measured potentiometrically for
the same reaction (1.0 mM ester substrate).
removed by centrifugation/ultrafiltration (10 °C), followed by a 15-
fold dilution with deionized water (thrice); the PMSF-inactivated
subtilisin solution was then concentrated to 15 mg/mL.
(b) Co-crystallization of Active and PMSF-Inactivated Subtilisin.
Varying volumes of solutions of the PMSF-inactivated and active
subtilisins were mixed to achieve a range of values of f from 0.05 to
0
.75. The crystals were then grown in the same manner as described
above. To verify that the active and inactive subtilisin had co-
crystallized, i.e., a single crystal was comprised of both active and
inactive subtilisin, a single crystal was removed from a given batch of
crystals and dissolved in aqueous solution (0.63 mg/L, pH 7.8, 100
mM KCl, 30 °C), and its activity was then measured in the hydrolysis
of 1.0 mM N-Ac-L-Phe-OEt. The activity obtained was compared to
that of crystals of completely active subtilisin (f ) 1), and the f value
of the given batch of crystals was calculated.
In anhydrous organic solvents, the ester and alcohol kcat/K
M
values
were determined in the following manner. The CLCs of subtilisin (5
mg) were recovered from the pH 7.8 phosphate buffer and washed with
anhydrous acetonitrile (2 × 1 mL) and then by the organic solvent in
which the kinetics were studied (3 × 1 mL). For (kcat/K
M N-Ac-L-Phe-OEt
)
the appropriate solvent containing a solution of the ester (0.6-6 mM
in octane; 10-100 mM in acetonitrile) and 100 mM propanol was
prepared immediately before kinetic measurements. In the determi-
(c) Activity Ws f Studies. The crystals of inactive/active subtilisin,
M
nation of (kcat/K )propanol, the solutions of 100 mM N-Ac-L-Phe-OEt and
propanol (2.5-100 mM) in tert-amyl alcohol and acetonitrile were
prepared immediately before use, as was the case for the measurement
of a given f, were cross-linked as described above. The activity of the
CLCs (0.05 mg/mL) was measured in the hydrolysis of 1.0 mM N-Ac-
L-Phe-OEt in water (pH 7.8, 100 mM KCl, 30 °C) and plotted against
f (Figure 1).
M
of (kcat/K )2-methoxyethanol where the solution was comprised of N-Ac-L-
Phe-OEt (100 mM in acetonitrile; 50 mM in cumene) and alcohol (50
mM in acetonitrile; 2.5-100 mM in cumene). One milliliter of the
appropriate solution was then added to 1 mg of the CLCs. The reaction
mixture was shaken at 30 °C and 300 rpm. Periodically, a 1-µL sample
Activity Coefficient Calculations. The activity coefficient of N-Ac-
26
L-Phe-OEt was obtained from its solubility in the appropriate solvent,
and those of the nucleophiles propanol and 2-methoxyethanol in organic
solvents were calculated using the UNIFAC algorithm.27
was withdrawn and assayed by GC. All kcat/K
-methoxyethanol in acetonitrile, were determined by fitting the kinetic
data as described above. The value of (kcat/K
nitrile was calculated via the initial rate (V) ratios of 2-methoxyethanol
M
values, except that for
(a) Determining the Solubility (S). In anhydrous solvents, an
2
excess of N-Ac-L-Phe-OEt was placed in the solvent containing 100
mM propanol. In hydrous solvents, where the water activity was
buffered to 0.006, the excess of ester was placed in 1 mL of solvent
M
)2-methoxyethanol in aceto-
containing 100 mM propanol and 100 mg of the BaBr
mixture. For S in water, the ester was placed in a 100 mM KCl solution,
2
‚1H
2 2
O/BaBr
(
25) Schonbaum, G. R.; Zerner, B.; Bender, M. L. J. Biol. Chem. 1961,
2
36, 2930-2935.

Enzymatic ActiVity in Organic SolVents Vs Water
pH 7.8. For the octane systems with a
J. Am. Chem. Soc., Vol. 118, No. 14, 1996 3365
w
) 0.006 and aadditive ) 1, an
mg/mL of BaBr
its solubility was reached. The resulting suspensions were shaken at
0 °C and 300 rpm. In octane with a ) 1, the ester was placed in a
2 2 2
‚1H O/BaBr , 30 °C) and the additive was added until
excess of the ester was added to the octane (100 mM propanol, 100
3
w
(26) The relationship between solubility and activity coefficient can be
vial containing octane (and 100 mM propanol) in contact through the
vapor phase with water containing propanol. The system was allowed
to equilibrate at 30 °C until the water activity reached unity. Periodi-
cally, the ester concentration in each system was assayed by GC; the
final concentration was denoted S.
described as follows. If the standard state is the pure liquid solute, the activity
of the solute is 1, n is the number of moles of solute in solution, nS is the
number of moles of solvent in solution, γ is the solution phase activity
coefficient of the solute, and x is the solution phase mole fraction of the
solute, then the solute will dissolve in the solvent until γx ) 1. If the solute
is sparingly soluble, then x ) n/nS, where n ) SV (S is the molar solubility
of the solute and V is the system volume) and nS ) V/VM, where VM is the
molar volume of the solvent. Then x ) SVM and γ ) 1/SVM. The value of
γ, which is a function of solute concentration, was calculated at the
concentration equaling S. At low mole fractions, where solute-solute
interactions are negligible, γ becomes constant with respect to concentration.
This is true for both Henry’s and Raoult’s law activity coefficients.
JA9539958
(27) Wescott, C. R.; Klibanov, A. M. J. Am. Chem. Soc. 1993, 115,
10362-10363.
(28) This work was financially supported by NIH Grant No. GM39794.
C.R.W. is an NIH Biotechnology Predoctoral Trainee.