Theoretical and Experimental Studies of Chiral Recognition in Charged Pirkle Phases

Article abstract of DOI:10.1246/cl.2001.232

The role of electrostatic interactions in enantioselective separation was demonstrated. Enantioselective separation of N-(3,5-dinitrobenzoyl)leucine and its esterified analogue was investigated by (S)-phenylglycine-based HPLC under an intermediate pH, and examined with a relaxed scan calculation combined with a Monte Carlo conformation search.

Full text of DOI:10.1246/cl.2001.232

232
Chemistry Letters 2001
Theoretical and Experimental Studies of Chiral Recognition in Charged Pirkle Phases
One-Sun Lee, Yun Hee Jang, Young Gi Cho,† Myung Ho Hyun,† Hie-Joon Kim, and Doo Soo Chung*
School of Chemistry, Seoul National University, Seoul 151-747, Korea
†Department of Chemistry, Pusan National University, Pusan 609-753, Korea
(Received December 13, 2000; CL-001117)
The role of electrostatic interactions in enantioselective
subsequently used to covalently attach the host molecule to the sili-
ca stationary phase.5 All organic solvents used as eluents were of
HPLC grade.
separation was demonstrated. Enantioselective separation of N-
(3,5-dinitrobenzoyl)leucine and its esterified analogue was
investigated by (S)-phenylglycine-based HPLC under an inter-
mediate pH, and examined with a relaxed scan calculation com-
bined with a Monte Carlo conformation search.
Results obtained using five different mobile phases are summa-
rized in Table 1. Enantioselectivity was lost in the presence of water,
probably because water screens electrostatic interactions. In less
polar eluents, (S)-G was eluted prior to (R)-G. It is speculated that
stronger electrostatic interactions between –COO– and –NH3+ prima-
rily determined the configuration of the complex. With this con-
straint, the rest of the complex, particularly the two benzene rings,
would be configured in a different manner in the two diastereomeric
complexes, allowing different extents of secondary interactions, par-
ticularly π–π interactions. Among those solvents, the highest resolu-
tion (Rs) was obtained with dichloromethane. The effect of acetic
acid concentration on enantioselective separation was investigated
Enantioselective separation has become increasingly important
in many areas of chemistry including pharmacology, agrochemistry,
and catalysis.1,2 One of the most successful methods of enantiose-
lective separation was developed by Pirkle and collaborators.3
However, an important aspect that has not been addressed yet in the
Pirkle-type host–guest interaction is the electrostatic interaction
between charged species. If a separation is attempted under a pH
where the carboxylic acid group in the analyte is deprotonated and
the amine group in the chiral stationary phase (CSP) is protonated,
the structure of the complex would be determined primarily by elec-
trostatic interactions between these oppositely charged groups and
secondarily by the π–π interaction or hydrogen bonding.
Optimization of conditions affecting the electrostatic interaction
could improve enantioselectivity.4
In this letter, we present a case where the role of electrostatic
interaction in enantioselective separation can be demonstrated
experimentally and theoretically: Enantioselective high perform-
ance liquid chromatography (HPLC) separation of N-(3,5-dini-
trobenzoyl)leucine enantiomers (abbreviated as (R)-G and (S)-G)
using an (S)-phenylglycine derivative as the CSP under an interme-
diate pH. Since there would be a stronger influence of the solvent
on charged molecules than neutral molecules, simulations were per-
formed both in the gas phase and in solvent, and the results were
compared.
~
further in dichloromethane (Table 1), and 0.3 vol% acetic acid (pH
=
6.5) gave the best result. At higher acid concentrations, the carboxy-
late group would become more protonated and the charge interaction
would decrease. Under the optimized condition, the enantioselectiv-
ity factor (α) was 1.20 and Rs 2.15 (Figure 2).
The role of the charge carried by the guest was investigated
further by attempting the enantioselective separation of the esteri-
fied analogue (Figure 1d). The enantiomers of this neutral analogue
were not separated under the same HPLC conditions, indicating the
important role of the electrostatic interaction between charged
groups in the enantioselective separation.
Gas-phase calculations. An approach suggested by Lipkowitz
et al. has been used to obtain ∆∆G in Pirkle-type host–guest interac-
tions.6 In this approach, the Boltzmann distributions of the host and
the guest structures are obtained by calculating their energies vary-
ing all the significant torsion angles incrementally. Significant
structures are selected based on the Boltzmann factor and then treat-
ed as rigid bodies in calculating the Boltzmann-averaged energy of
the host–guest complex by a Monte Carlo (MC) search.
HPLC experiments. (S)-Phenylglycine and N-(3,5-dinitroben-
zoyl)leucine were purchased from Aldrich (Milwaukee, WI,
U.S.A.). (S)-Phenylglycine was endowed with a linker, which was
The structure of the model host, (S)-phenylglycine (Figure
1b), was built and energy-minimized in MacroModel 6.07 with the
Copyright © 2001 The Chemical Society of Japan

Chemistry Letters 2001
233
MM2 force field which is adequate for describing Pirkle-type
host–guest interactions.6 The potential energy surface was obtained
by varying its torsion angles, β and γ , by 10° (Figure 3a). Only one
minimum was located at (β, γ ) = (130°, 130°) and further mini-
mization led to a minimum at more refined angles (134°, 127°). A
separate MC conformation search7 led to the same result. Thus,
only this conformer (Hg) was considered for the complex. The
same relaxed scan minimization was performed for the guest in
three independent variables (δ, ψ and ω) by a total of 42 875 calcu-
lations [(35 for δ) × (35 for ψ) × (35 for ω)]. Figure 3b shows a
potential energy curve as a function of δ. There are two minima,
one at δ = 20° (syn) and another at δ = 170° (anti). The anti con-
former was 14 kJ/mol less stable than the syn conformer, probably
due to larger repulsion between the two carbonyl oxygens. About
99% of the guest molecules should exist as syn and thus only the
syn conformer was considered in the simulation. Figure 3c shows a
contour diagram of potential energies of this syn conformer as a
function of ψ and ω. Five stable conformers (G1–G5) were locat-
ed. Since the energy differences were small and the energy barriers
between them were low, all of these conformers were included in
the calculations. Using these selected conformers, five host–guest
complexes (Hg:G1, ···, Hg:G5) were built, assuming a 1:1 ratio of
the host to the guest. The host and the guest were treated as rigid
bodies.6 Minimum-energy configurations of the complexes were
identified by the MC search method by rotating and translating the
guest relative to the host randomly. Ten thousand host–guest com-
plex structures were tried. The Boltzmann-averaged ∆∆G was cal-
culated6 over the stable configurations of the complexes.
repulsion between the two carbonyl oxygens. However, the syn
conformer is still dominant (97% population) and was considered in
the complex simulation in solvent. The potential energy surface of
ψ and ω identified seven local-minima (S1–S7; Figure 3d) and all
of them were used for the host–guest complex study in solvent.
The global minimum energy structures are in Figure 4. The free
energy of the (S)-H:(R)-G was 0.6 ± 0.4 kJ/mol lower than that of
the (S)-H:(S)-G, leading to an enantioselectivity of 1.3 ± 0.2.
Overestimation of ∆∆G in the gas phase calculation was largely
resolved by including the solvent effect.
The free energy of the (S)-H:(R)-G was 9.2 ± 0.7 kJ/mol lower
than that of the (S)-H:(S)-G. This indicates that (R)-G would be
retained longer than (S)-G. The relation
This work was supported by KOSEF (No. 96-0501-08-01-3)
and CRI, the Ministry of Science and Technology, Korea. OSL
thanks the Ministry of Education for the BK 21 fellowship.
∆∆G = –RT ln α
(1)
leads to a very high enantioselectivity factor α = 40 ± 10. The elu-
tion order was correctly predicted but the enantioselectivity was
severely overestimated.
References
1
“Chiral Separations: Applications and Technology,” ed. by S. Ahuja,
American Chemical Society, Washington, D.C., (1997).
E. Thall, J. Chem. Educ., 73, 481 (1996).
W. H. Pirkle and T. C. Pochapsky, Chem. Rev., 89, 347 (1989).
S. Hwang, O.-S. Lee, and D. S. Chung, Chem. Lett., 2000, 1002.
M. H. Hyun, Y. J. Cho, J.-J. Ryoo, K. K. Jyung, and G. S. Heo, J.
Chromatogr. A, 696, 173 (1995).
Solution-phase calculations. The same approach used in the
gas-phase calculation was used in solvent. A continuum solvent
model was used with a dielectric constant (ε) of 10 (for
dichloromethane, ε = 8.93). Again, only one minimum was located
for the host (β = 125° and γ = 121°; Hs) and this was used for the
host–guest complex calculation in solvent. For the guest, the rela-
tive energy of the anti conformer with respect to the syn con-
former and the barrier between them decrease as the dielectric con-
stant increases (Figure 3b), probably because the solvent screens the
2
3
4
5
6
7
K. B. Lipkowitz, R. Larter, T. Darden, D. A. Demeter, and R.
Zegarra, J. Am. Chem. Soc., 110, 3446 (1988).
F. Mohamadi, N. G. J. Richards, W. C. Guida, R. Liskamp, M.
Lipton, C. Caufield, G. Chang, T. Hendrickson, and W. C. Still, J.
Comput. Chem., 11, 440 (1990).