Application of structure-based thermodynamic calculations to the rationalization of the enantioselectivity of subtilisin in organic solvents

Article abstract of DOI:10.1016/S0957-4166(98)00086-X

The effect of organic solvents on the selectivity of lyophilized or CLEC (cross-linked enzyme crystals)-subtilisin in the resolution of sec-phenethyl alcohol and trans-sobrerol was studied. A theoretical model, that tries to predict solvent effects on ena

Full text of DOI:10.1016/S0957-4166(98)00086-X

TETRAHEDRON:
ASYMMETRY
Pergamon
Tetrahedron: Asymmetry 9 (1998) 1205–1214
Application of structure-based thermodynamic calculations to the
rationalization of the enantioselectivity of subtilisin in organic
solvents
Giorgio Colombo,a Gianluca Ottolina,a,∗ Giacomo Carrea,a Anna Bernardi b and
Carlo Scolastico b
aIstituto di Chimica degli Ormoni, CNR, Via Mario Bianco 9, 20131, Milano, Italy
bDipartimento di Chimica Organica e Industriale, Università degli Studi di Milano, Via Venezian 21, 20133, Milano, Italy
Received 2 February 1998; accepted 25 February 1998
Abstract
The effect of organic solvents on the selectivity of lyophilized or CLEC (cross-linked enzyme crystals)-subtilisin
in the resolution of sec-phenethyl alcohol and trans-sobrerol was studied. A theoretical model, that tries to predict
solvent effects on enantioselectivity only as a function of the activity coefficients of the desolvated part of the
substrate in the relevant transition state of the reaction (Ke, T.; Wescott, C. R.; Klibanov, A. M. J. Am. Chem. Soc.,
1996, 118, 3366) was examined and shown to agree poorly with the experimental data. The tetrahedral intermediate
was studied with MonteCarlo molecular mechanics, and the activity coefficients were calculated with UNIFAC.
© 1998 Elsevier Science Ltd. All rights reserved.
1. Introduction
The study of enzyme selectivity in nearly anhydrous organic solvents is a particularly interesting and
intriguing field of research.1 In these media, the absence of a continuous aqueous phase around the
enzymes allows their direct interaction with the solvents, and confers to the biocatalysts new properties
with respect to stability, activity and specificity-selectivity.2 Furthermore, in organic media, enzymes
such as hydrolases can catalyze esterifications and transesterifications readily, with high product yields.2
A deep understanding of enzyme–substrate–solvent interactions is necessary to improve further
enzyme performance. Thus several theories have been proposed to explain the mechanisms by which
solvents influence enzymatic properties. For instance, the selectivity could be altered if the solvent
molecules could bind within the active site and alter either the interaction between enzyme and substrate
∗
Corresponding author. E-mail: ottolina@ico.mi.cnr.it
0957-4166/98/$19.00 © 1998 Elsevier Science Ltd. All rights reserved.
PII: S0957-4166(98)00086-X

1206
G. Colombo et al. / Tetrahedron: Asymmetry 9 (1998) 1205–1214
Fig. 1. Subtilisin-catalyzed transesterification reactions of sec-phenethyl alcohol 1 and trans-sobrerol 2 with vinyl-acetate
or the structural characteristics of the complex.3 According to another theory, the solvent could modify
the enzyme conformation and thus influence the selectivity by altering the molecular recognition process
between substrate and enzyme.4 A third model introduces a dependence of selectivity on the energetics
of substrate solvation, regardless of the presence of other mechanisms.5 Based on this last rationale,
Klibanov and coworkers have proposed a quantitative model to rationalize the solvent dependence of
enzymatic selectivity solely on the basis of the thermodynamics of substrate solvation.6 The validity of
this model is examined in the present paper in the case of two different substrates and two different forms
(crystalline and lyophilized) of subtilisin Carlsberg.
2. Results and discussion
The quantitative model that tries to rationalize the solvent dependence of enzyme selectivity solely on
the basis of the thermodynamics of substrate solvation was elaborated by Klibanov and coworkers,6 and
is represented by this simple equation:
log E = log(γ0 /γ0 ) + constant
(1)
S
R
where E is the enantiomeric ratio, and γ0S and γ0R are defined as the activity coefficients of the desolvated
substrate moiety in the transition state for the S- and R-enantiomer. This model predicts that the logarithm
of enzyme selectivity (E) should be proportional to the logarithm of the ratio of the thermodynamic
activity coefficients of the desolvated portions of the substrates in the transition state of the reaction,
plus a constant. We have applied this model to the subtilisin-catalyzed transesterification reactions of
sec-phenethyl alcohol 1 and trans-sobrerol 2 with vinyl-acetate or butyrate, carried out in several organic
solvents (Fig. 1).
Table 1 shows the enantiomeric ratio (E) and the relative velocity (Vrel) of our model reactions with the
organic solvents most commonly used for enzymatic transesterifications. Regarding the acylating agent,
the enantioselectivity was higher with vinyl butyrate than with vinyl-acetate in all the solvents tested.
It can also be seen that the enzyme was markedly more selective with trans-sobrerol than with sec-
phenethyl alcohol, and that the enantioselectivities of lyophilized and CLEC-subtilisin were comparable.
The transesterification rates in the various solvents don’t show a clear trend: the selectivity was found to

G. Colombo et al. / Tetrahedron: Asymmetry 9 (1998) 1205–1214
1207
Table 1
Enantiomeric ratio (E) and relative velocity (Vrel) of subtilisin catalyzed transesterification in various
organic solvents
vary erraticly and to depend on the acylating agent and enzyme form. It should be emphasized that all the
experiments were carried out at a constant water activity (<0.1) by working in the presence of molecular
sieves or by using reactants pre-equilibrated with molecular sieves.7,14
In order to generate a model of the solvated and desolvated portions of the substrates in the transition
state, a conformational analysis of the tetrahedral intermediates TS1 and TS2 was carried out. The
intermediates were generated starting from the X-ray structure of the subtilisin Carlsberg–Eglin complex,
as described in the computational section (vide infra). After an initial minimization of the entire structure,
the conformational space of the R- and S-acetyl intermediates was investigated by MC/EM technique8 on
a substructure which included the substrate and all atoms within 15 Å of the tetrahedral carbon. All the
minima found within 50 kJ/mol of the global minimum were fully reminimized to allow a more accurate
determination of the relative energies.
In qualitative agreement with the experimental data, the complexes formed by the S-substrates were
found to be more stable than those formed by the R-enantiomers. The calculated ∆ES−R were 29 kJ/mol
for 1 and 180 kJ/mol for 2. In both cases, the predicted selectivity was one order of magnitutide larger

1208
G. Colombo et al. / Tetrahedron: Asymmetry 9 (1998) 1205–1214
Fig. 2. Schematic representation of binding of S- and R-enantiomers of sec-phenethyl alcohol (a), and trans-sobrerol (b) in the
active site of subtilisin. The aminoacids reported are those of the catalytic triad and of the oxyanion hole
than the experimentally observed one in all the solvents examined. However, transesterification of 2 was
correctly predicted to be more selective than the transesterification of 1.
The lowest minimum conformation of the sec-phenethyl alcohol complex S0 showed that the hydrogen
of Nε of His64 was directed towards the alcoholic oxygen of the substrate and the Oγ of Ser221. This
arrangement seemed to favor a stacking interaction between the benzene ring of the substrate and the
imidazole ring of His64. In the first structure above the global minimum (S1) the methyl group present in
the alcoholic moiety was directed outside the catalytic pocket.
The stabilizing hydrogen bond network and the stacking interaction were not found in the R-complex
model, and the lack of these favorable interactions can be invoked as a reason for the preference of
subtilisin for the S-enantiomer (Fig. 2a) (∆ES−R=29 kJ/mol).
In the case of trans-sobrerol, the S0-enantiomer was much more stable than the R0-enantiomer because
it was less sterically congested. In the S0-complex, the substrate was directed outside the protein and was
stabilized by a network of H-bonds, which were absent in the case of the R0-complex (Fig. 2b).
Once the structures of the intermediates were determined, UNIFAC was applied to calculate the activity
coefficients.9 The solvent accessible areas of the tetrahedral intermediate were evaluated both in the
enzyme-bound and free form and, in the latter case, the enzyme was replaced by a hydrogen atom. The
surface areas for each UNIFAC group were calculated by the Connolly algorithm12 and the results are
reported in Table 2 and Table 3. The upper part of the table shows the desolvation areas of the alcoholic
moiety of the intermediate and the lower part, the data for the acetyl moiety. The linear regression of
Eq. 1 between the experimentally determined E in the various solvents and the calculated γ0 /γ0 in the
S
R
same solvents gave the slopes reported in Table 4.
According to the literature,6 two different strategies were employed for the calculation of the γ0
values. The groups with areas greater than 50% desolvated were considered completely desolvated and
the groups with areas less than 50% desolvated were considered completely solvated (Table 4, all or
nothing column). Alternatively, the precise percentages of desolvation were used for the calculation of

G. Colombo et al. / Tetrahedron: Asymmetry 9 (1998) 1205–1214
1209
Table 2
Solvent accessible areas of R- and S-sec-phenethyl alcohol transition states
Table 3
Solvent accessible areas of R- and S-trans-sobrerol transition states

1210
G. Colombo et al. / Tetrahedron: Asymmetry 9 (1998) 1205–1214
Table 4
Slopes and correlation coefficients obtained from the linear regression of Eq. 1 applied to the
experimentally determined E values, for subtilisin catalyzed transesterification in different organic
solvents, and calculated γ0 /γ0
S
R
γ0 (Table 4, fractional column). In the original Klibanov procedure the γ0 /γ0R ratios were calculated on
S
the alcohol moiety only. We followed this procedure (entries 7 and 14) and extended the calculation to
the entire intermediate (Table 4, entries 1–6 and 8–13). For sec-phenethyl alcohol the two global minima
R0 and S0 were indistinguishable in terms of desolvation surface area (Table 2) and thus by UNIFAC
calculations (Table 4, entry 1). The calculation was therefore extended to all the minima found for the
tetrahedral intermediate in an energy range of 25 kJ/mol from the global minimum, so six structures for
the R-complex (R0–R5) and two structures for the S-complex (S0–S1) were taken into account (Table 2).
In terms of desolvation of different groups, the first two structures for the S-complex differed in the
position of the methyl group in the alcoholic moiety, which in S1 is more exposed to the solvent. For
the R-complex the six structures (R0–R5) fell into three clusters: one containing structures R0, R3 and R5
(Table 2, R0-complex), one containing R2 and R4 (Table 2, R2-complex), and one containing R1 alone
(Table 2, R1-complex).
Each R- and S-structure were compared to one another for vinyl acetate (entries 1–7) and vinyl butyrate
(entries 8–14) transesterifications and linear regressions are reported in Table 4. In all cases the slopes
obtained differed greatly from unity, which is the theoretical value, and spanned from −2.51 to 3.43.
This basically means that the model in some cases tends to underestimate the influence of the solvent
(slope>1, i.e., the variations of E values were higher than those of γ0 /γ0 ), and in others to overestimate
S
R

G. Colombo et al. / Tetrahedron: Asymmetry 9 (1998) 1205–1214
1211
Fig. 3. Dependence of the enantiomeric ratio (E) of lyophilized (open squares, — ) and CLEC-subtilisin (filled squares, ······)
on the ratio of the activity coefficients, calculated with the all or nothing approach, of the desolvated parts of sec-phenethyl
alcohol in various organic solvents in the transesterification with vinyl butyrate
it (slope<1, i.e., the variations of E values were lower than those of γ0 /γ0 ). The case of negative
S
R
slope values is even worse because the predicted solvent effects on enantioselectivity were opposite to
the experimental ones. The use of the fractional instead of the all or nothing model did not change
the essence of the results, because in both cases the slopes differed greatly from unity. These data also
show that this model was unsuitable for both lyophilized10 and CLEC-subtilisin, since both enzyme
forms experimentally displayed identical trends with the same solvents. For instance, Fig. 3 depicts the
dependence of the enantiomeric ratio of both lyophilized and CLEC-subtilisin on the ratio of activity
coefficients, using the all or nothing approach in the transesterification of sec-phenethyl alcohol with
vinyl butyrate.
The same considerations apply to the transesterification of trans-sobrerol with vinyl acetate (Table 4,
entries 15 and 16). In this case, clear differences in terms of desolvation were observed in the S0- and
R0-complexes (Table 3 and Fig. 2b). However no correlation was found between the UNIFAC calculation
and experimental selectivities either by considering the alcohol moiety or by extending the calculation.
It could be argued that these results were a consequence of an inaccurate determination of E values.
However, these values were obtained through simple and reliable chiral chromatographic tecniques14
and, indeed, the standard deviations were less than 20%. One point to be taken into account is that
all calculations were run in vacuo, which could have led to some artifacts in the determination of the
structure of the global minimum, that could have been different for the different solvents if their presence
had been considered explicitly.11 Such an approach, however, would be extremely demanding if a large
number of organic solvents were to be taken into consideration. One more point is that UNIFAC is unable
to distinguish between slightly different structures in terms of desolvation (e.g. in the case of R0- and S0-
complexes of sec-phenthyl alcohol) and it may be suitable only for markedly different complexes.

1212
G. Colombo et al. / Tetrahedron: Asymmetry 9 (1998) 1205–1214
Finally, it must be considered that the solvent-dependence of enzyme-selectivity could not be entirely
due to substrate desolvation phenomena, but also to the presence of solvent molecules at the active site,
that could change the molecular recognition process by steric, electrostatic and conformational factors.3,4
Thus, even though the method developed by Klibanov and coworkers6 to rationalize and predict enzyme
selectivity is very appealing for its simplicity, and gives excellent results for some systems, it does not
seem to be generally applicable to enzymatic resolutions in organic solvents.6
3. Materials and methods
(ꢀ)-sec-Phenethyl alcohol and (ꢀ)-trans-p-menth-6-ene-2,8-diol (trans-sobrerol), vinyl acetate and
vinyl butyrate were purchased from Aldrich. All organic solvents were distilled and equilibrated on 4
Å molecular sieves before use. Subtilisin Carlsberg (protease Type VIII) was purchased from Sigma
and, before use, dissolved in 20 mM potassium buffer (pH 7.8) and lyophilized.13 Cross-linked enzyme
ꢁ
crystals (CLEC) of subtilisin (dry powder, ChiroCLEC™-BL) was a generous gift from Altus Biologics,
Cambridge, MA, USA.
3.1. Enzyme catalyzed transesterifications
sec-Phenethyl alcohol and lyophilized subtilisin: To a 3 mL screw-capped vial were added 1 mL
of organic solvent, 0.01 mmol of sec-phenethyl alcohol, 0.2 mmol of acylating agent (vinyl acetate or
butyrate), 50 mg of molecular sieves and 10 mg of subtilisin; the reaction mixture was shaken at 180 rpm,
45°C and, at different times, aliquots were withdrawn and the degree of conversion and the enantiomeric
excess of the product determined by chiral GC as already described.14a
sec-Phenethyl alcohol and CLEC-subtilisin: The reaction was carried out as described above using
0.083 mmol of sec-phenethyl alcohol, 0.35 mmol of vinyl butyrate and 1–5 mg of CLEC-subtilisin
without adding molecular sieves to the solution. Prior to mixing, all components were brought to aw<0.1
by equilibration (2 days) with molecular sieves in sealed containers.14
trans-Sobrerol and lyophilized subtilisin: The reaction was carried out as above using 0.05 mmol of
sobrerol, 0.07 mmol of vinyl acetate and 30 mg of subtilisin. The degree of conversion and the enan-
tiomeric excess of the products were determined by chiral GC as already described.14a The enantiomeric
ratio was calculated according to Chen et al.15
3.2. Molecular modeling
Calculations and graphic simulations were performed on a Silicon Graphics Indigo2 R10000. Starting
geometries of the studied compounds were generated and subjected to simulation using the program
Macro Model-BatchMin (version 5.5). The initial coordinates of subtilisin Carlsberg complexed with the
polypeptidic inhibitor Eglin C (2sec.pdb) were obtained the from Brookhaven Data Bank. The inhibitor,
except for the residues Leu59 and Asp60, and all water molecules except one, which is placed close to
Leu217, His67 and Pro225 and is highly conserved and present in a large number of subtilisins from
different sources, were removed from the enzyme structure. The incorrect bond lengths in His39 and
His67 were corrected and the acidic hydrogens added. Only polar hydrogens and substrate hydrogens
were considered explicitly, while for the rest of the protein a united atom model was used. The enzyme
was completely minimized under vacuum using the AMBER* force field,16 Polak–Ribiere conjugated
gradient (PRCG) as minimization algorithm, and 0.05 kJ/Å mol as gradient convergence criterion.

G. Colombo et al. / Tetrahedron: Asymmetry 9 (1998) 1205–1214
1213
The residual part of the inhibitor (Leu-Asp dimer) was used to create the structures of the tetrahedral
intermediates of our substrates inside the active site and then deleted.
In the case of sec-phenethyl alcohol, both R- and S-acyl-intermediates were subjected to conforma-
tional analysis using the Montecarlo/Energy Minimization technique, PRCG, AMBER*.17 The study was
limited to the atoms included in a 15 Å range from the carbonyl carbon in the tetrahedral intermediate.18
During the energy minimization step the atoms in this range are kept fixed in their crystallographic
positions with a distance-dependent harmonic force constant that varies as a function of the distance
from the tetrahedral intermediate. To better reproduce the H-bond network in the active site, no restraint
was imposed on the coordinates of polar hydrogens. The rest of the protein was completely neglected
in this phase of the calculation. From the conformational analysis, 184 structures for the S-enantiomer
and 224 structures for the R-enantiomer were obtained in 50 kJ/mol. These were fully minimized using
PRCG, AMBER* until a gradient of 0.05 kJ/mol was reached, giving 3 structures for the S-enantiomer
complex and 10 structures for the R-enantiomer complex in 41 kJ/mol. trans-Sobrerol was first studied
as a free uncomplexed molecule to find the most stable conformation, performing three systematic-
pseudo-MonteCarlo/Energy Minimization conformational searches (2000 steps each)19 to obtain the
global minimum. This structure was then used to build the transition state intermediate that was placed
inside the active site as already described for sec-phenethyl alcohol, and the conformational search on
this intermediate was run using the same restraints as described before.17 From the conformational
analysis, 90 structures were obtained for the S-enantiomer and 28 structures for the R-enantiomer in
50 kJ/mol. These were fully minimized using PRCG, AMBER* until a gradient of 0.05 kJ/mol was
reached, giving in 41 kJ/mol two structures for the S-enantiomer complex and two structures for the
R-enantiomer complex. Solvent accessible surface areas for our substrates were calculated using the
Connolly algorithm12 with a probe radius of 1.4 Å and the ‘free’ tetrahedral intermediates were obtained
by replacing the subtilisin with an atom of hydrogen. The percentages of desolvation were calculated
according to Eq. 2.
˚
˚
˚
% = 100 ∗ (Afree − Aenzyme bound)/Afree
(2)
3.3. UNIFAC application
All the activity coefficients (γ) were calculated using the UNIFAC computer algorithm.9 The desol-
vated fraction of the substrate was determined measuring the Connolly surface area12 for each group.
For sec-phenethyl alcohol the desolvated portions were calculated on the global minimum for the R-
enantiomer, while for the S-enantiomer they were calculated on the two lowest minima. For trans-
sobrerol we compared the two global minima structures. For the calculation of γ values the concentrations
of acylating agents and substrates reported above (enzyme catalyzed transesterifications) were used.
Acknowledgements
We thank the Biotechnology Programme of the European Commission (BI04-CT95-0231) and the
CNR Target Project on Biotechnology for their financial support.
References
1. Wescott, C. R.; Klibanov, A. M. Biochim. Biophys. Acta, 1994, 1206, 1. Carrea, G.; Ottolina, G.; Riva S. Trends Biotechnol.,
1995, 13, 63.

1214
G. Colombo et al. / Tetrahedron: Asymmetry 9 (1998) 1205–1214
2. Biocatalysis in Non-Conventional Media; Tramper, J.; Vermue, M. H.; Beeftink, H. H.; von Stockar, U.; Eds. Elsevier;
Amsterdam, 1992. Enzymatic Reaction in Organic Media; Koskinen, A. M. P.; Klibanov, A. M.; Eds. Blakie; London,
1996. Klibanov, A. M. Trends Biotechnol., 1997, 15, 97.
3. Nakamura, K.; Takebe, J.; Kitayama, T.; Ohno, A. Tetrahedron Lett., 1991, 32, 4941. Secundo, F.; Riva S.; Carrea G.
Tetrahedron: Asymmetry, 1992, 3, 267. Hirose, Y.; Kariya K.; Sasaky, J.; Kurono, Y.; Ebike, H.; Achiwa K. Tetrahedron
Lett., 1992, 33, 7157.
4. Fitzpatrick, P. A.; Klibanov, A. M. J. Am. Chem. Soc., 1991, 113, 3166; Wu, H. S.; Chu, F. Y.; Wang, K. Bioorg. Med. Chem.
Lett., 1991, 1, 339. Ueji, S.; Fujio, R.; Okubo, N.; Miyazawa, T.; Kurita, S.; Kitadani, M.; Muromatsu, A. Biotechnol. Lett.,
1992, 14, 163.
5. Wescott, C. R.; Klibanov, A. M. J. Am. Chem. Soc., 1993, 115, 1629. Wescott, C. R.; Klibanov, A. M. J. Am. Chem. Soc.,
1993, 115, 10362.
6. Ke, T.; Wescott, C. R.; Klibanov, A. M. J. Am. Chem. Soc., 1996, 118, 3366. Wescott, C. R.; Noritomi, H.; Klibanov, A. M.
J. Am. Chem. Soc., 1996, 118, 10365.
7. Valivety, R. H.; Halling, P. J. Biochim. Biophys. Acta, 1992, 1118, 218.
8. Chang, R.; Guida, W. C.; Still, W. C. J. Am. Chem. Soc., 1989, 111, 4379. Chang, R.; Guida, W. C.; Still, W. C. J. Am.
Chem. Soc., 1990, 112, 1419.
9. Skjold-Jorgensen, S.; Kolbe, N.; Gmehling, J.; Rasmussen, P. Ind. Eng. Chem. Proc. Des. Dev., 1979, 18, 714. Gmehling,
J.; Rasmussen, P.; Fredenslund, A. Ind. Eng. Chem. Proc. Des. Dev., 1982, 21, 118. Macedo, E. A.; Weidlich, U.; Gmehling,
J.; Rasmussen, P. Ind. Eng. Chem. Proc. Des. Dev., 1983, 22, 676. Tiegs, D.; Gmehling, J.; Rasmussen, P.; Fredenslund, A.
Ind. Eng. Chem. Proc. Des. Dev., 1987, 26, 159. Hansen, H. K.; Rasmussen, P.; Schiller, M.; Gmehling, J. Ind. Eng. Chem.
Res., 1991, 30, 2355. van Tol, J. B. A.; Stevens, R. M. M.; Veldhuizen, W. J.; Jongejan, J. A.; Duine, J. A. Biotechnol.
Bioeng., 1995, 47, 71.
10. Wescott, C. R.; Klibanov, A. M.; Biotechnol. Bioeng., 1997, 56, 340.
11. Toba, S.; Merz, K. M. Jr. J. Am. Chem. Soc., 1997, 119, 9939. Gilson, M. K.; Given, J. A.; Bush, B. L.; McCammon J. A.
Biophys. J., 1997, 72, 1047. Norin, M.; Haeffner, F.; Hult, K.; Edholm, O. Biophys. J., 1994, 67, 548.
12. Connolly, M. L. Science, 1983, 221, 709.
13. Riva, S.; Chopineau, J.; Kieboom, P. G.; Klibanov, A. M. J. Am. Chem. Soc., 1988, 110, 584.
14. (a) Ottolina, G.; Bovara, R.; Riva, S.; Carrea, G. Biotechnol. Lett., 1994, 16, 923. (b) Orrenius, C.; Norin, T.; Hult, K.;
Carrea, G. Tetrahedron: Asymmetry, 1995, 6, 3023.
15. Chen, C.-S.; Fujimoto, Y.; Gidaukas, G.; Sih, C. J. J. Am. Chem. Soc., 1982, 104, 7294.
16. Kollmann, P. A.; Case, P.; Profeta, S.; Wiener, P. J. Am. Chem. Soc., 1984, 106, 765.
17. The torsion angles considered for the analysis are represented:
18. Guida, W. C.; Bohacek, R. S.; Erion, M. D. J. Compt. Chem., 1992, 13, 214.
19. Goodman, J.; Still, W. C. J. Compt. Chem., 1991, 12, 1110.