Anal. Chem. 1996, 68, 3460-3467
Chiral Separations by Nonaqueous Capillary Electrophoresis Fang Wang and Morteza G. Khaledi*
Department of Chemistry, P.O. Box 8204, North Carolina State University, Raleigh, North Carolina 27695-8204.
This paper reports successful chiral separations of pharmaceutical racemic amines by nonaqueous capillary electrophoresis (NACE) using β- and γ- cyclodextrins (CDs) and various derivatives of β-CDs. The results in three organic solvents, formamide (FA), N-methylformamide (NMF), and N,N-dimethylformamide (DMF) were compared to those in pure water and in 6 M urea in water systems. The binding constants of trimipramine, mianserin, and thioridazine with β-CD were determined in the following five solvent systems: water, 6 M urea in water, FA, NMF, and DMF. The binding constants decreased systematically from ∼104 in water to ∼10 in FA and ∼10-2 in DMF. As a result, the optimum CD concentration in the aqueous media is in the high micromolar range, while that in the FA is around 100 mM. In the aqueous media, the occurrence of the optimum at very low concentrations and the rapid changes in enantioselectivity with CD concentration would make it difficult to develop methods based on trial and error. Nevertheless, it is shown that, even under nonoptimum concentrations of the chiral selector, other experimental parameters such as ionic strength, addition of tetraalkylammonium (TAA+), and temperature can be adjusted to achieve acceptable resolutions. This, however, is often achieved at the expense of longer analysis times. In addition, the effects of apparent pH (pH*) and type of cyclodextrin on chiral separations in NACE are studied. The application of negatively charged β-CD in FA is also reported. Chiral separation of trimipramine was achieved at lower concentration of anionic CD due to the additional Coulombic interactions. Capillary electrophoresis (CE) separations in purely nonaqueous solvents (NACE) have recently received attention.1-5 The use of organic solvents extends the application range of CE to charged hydrophobic compounds. In certain organic solvents, higher electric field strengths and/or higher ionic strengths can be used that lead to higher efficiency and/or better peak stacking and detectability as compared to that in aqueous media.2 In addition, various acid-base chemistry and ion-solvation effects in nonaqueous solvents can lead to different selectivity in electrophoretic migration.6 However, the present knowledge of the proton donor-acceptor chemistry and solvation of charged (1) Walbroehal, Y.; Jorgenson, J. W. J. Chromatogr. 1984, 315, 135. (2) Sahota, R. S.; Khaledi, M. G. Anal. Chem. 1994, 66, 1141. (3) Tomlinson, A. J.; Benson, L. M.; Naylor, S. LC-GC 1994, 12, 122. (4) Ng, C. L.; Lee, H. K.; Li, S. F. Y. J. Liq. Chromatogr. 1994, 17, 3847. (5) Salimi-Moosavi, H.; Cassidy, R. M. Anal. Chem. 1995, 67, 1067. (6) Ye, B.; Khaledi, M. G. Presented at the 1993 Pittsburgh Conference, Chicago, IL; Abstract No. 199.
3460 Analytical Chemistry, Vol. 68, No. 19, October 1, 1996
organic molecules in nonaqueous solvents is very limited. Likewise, there is little or no information about the interaction of solutes with organized media such as cyclodextrins (CDs) in nonaqueous media. A better understanding of these issues is essential for the assessment of the true capabilities and limitations of NACE. Chiral separation by CE offers a number of distinct advantages over HPLC, such as high efficiency, speed of analysis, flexibility of rapid incorporation of various chiral selectors, and feasibility of method development.7 Resolution in CE can be described by the following equation:8
Rs )
[
]
xN ∆µ 4 µav + µeo
(1)
where N is the number of theoretical plates, ∆µ is the difference in electrophoretic mobility of two enantiomers, µav is the average electrophoretic mobility of two enantiomers, and µeo is the mobility of electroosmotic flow. The electrophoretic mobility of an enantiomer is expressed in terms of CD concentration, [CD], and analyte binding constant to CD as9
µ)
µf + K[CD]µc 1 + K[CD]
(2)
According to Wren and Rowe,10,11 enantioselectivity between two enantiomers in CE can be described as
∆µ )
(µf - µc)∆K[CD] (1 + K1[CD])(1 + K2[CD])
(3)
where µf and µc are the electrophoretic mobilities of an enantiomer at CD concentrations equal to zero and infinity, [CD] is the equilibrium concentration of CD, K1 and K2 are binding constants between CD and enantiomers 1 and 2, respectively, and ∆K is the difference in binding constants. Selection of optimum parameters is crucial in achieving high resolution of enantiomeric mixtures. The type of chiral selector is the most important parameter, as it controls the term ∆K and consequently enantioselectivity. No chiral separation is achieved if the two enantiomers bind to a chiral selector to the same extent. (7) Nishi, H.; Terabe, S. J. Chromatogr. 1995, 694, 245. (8) Jorgenson, J.; Lukacs, K. D. Anal. Chem. 1981, 53, 1298. (9) Guttman, A.; Paulus, A.; Cohen, A. S.; Grinberg, N.; Karger, B. L. J. Chromatogr., 1988, 448, 41. (10) Wren, S. A. C.; Rowe, R. C. J. Chromatogr. 1992, 603, 235. (11) Wren, S. A. C.; Rowe, R. C. J. Chromatogr. 1992, 609, 363. S0003-2700(96)00537-9 CCC: $12.00
© 1996 American Chemical Society
However, different values of binding constants for two isomers would not automatically translate into high resolution. Other factors, such as concentration of the chiral selector, pH, electroosmotic flow, and solvent composition, can also play significant roles. The majority of CE chiral separations have been performed in neat aqueous media and some in mixed hydroorganic solvents.7 Addition of organic modifiers influences the electroosmotic flow as well as the binding constants between the enantiomers and the chiral selector.11-14 Ye and Khaledi reported the first successful chiral separation of trimipramine with β-CD in NACE.15 Terabe et al., as well as Riekkola and co-workers, recently reported chiral separations by NACE.16,17 In this paper, the use of CDmediated separations of enantiomeric compounds by NACE is systematically examined. The binding constants of some basic pharmaceutical racemates in different solvents were determined. Since NACE is still a relatively unexplored area, the effects of CD types, CD concentrations, buffer ionic strength, the apparent pH (pH*), control of electroosmotic flow, and operating temperature on chiral separations were also studied. EXPERIMENTAL SECTION Apparatus. Experiments were carried out on a home-built unit and a P/ACE 5500 system with a photodiode array detector (Beckman, Fullerton, CA). The home-built unit consists of a (30 kV high-voltage power supply (Series EH, Glassman High Voltage Inc., Whitehouse, NJ), a UV-visible detector (Model 200, Linear Instruments, Reno, NV) operating at 254 nm, and an electronic integrator (Hewlett-Packard, Avondale, PA). Untreated fused-silica capillary tubes (Polymicro Technologies, Phoenix, AZ) with 50 µm i.d., 363 µm o.d. were used. The total length of the capillary was 42 cm, and the length of the capillary to the detector was 31 cm. The capillary temperature was maintained at 25 °C by jacketing it in light mineral oil using a constant-temperature circulator (Type K2-R, Lauda, Germany). All the experiments were run at +30 kV. The injection was done by gravity injection. Viscosity was determined by an Ostwald capillary viscometer, which was held in thermostated water at 25 °C, with temperature variation of e(0.05 °C. Chemicals. R-Cyclodextrin (R-CD) was purchased from American Maize. γ-Cyclodextrin (γ-CD) was purchased from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). β-Cyclodextrin (β-CD), methyl-β-cyclodextrin (M-β-CD), hydroxypropylβ-cyclodextrin (HP-β-CD), and sulfated β-cyclodextrin (β-CD(SO4-)4) were gifts from American Maize Products Co. (Hammond, IN). Citric acid, tris(hydroxymethyl)aminomethane (Tris), tetramethylammonium bromide (TMABr), and tetrabutylammonium bromide (TBABr) were purchased from Aldrich (Milwaukee, WI). All the racemic samples were purchased from Sigma (St. Louis, MO), and their chemical structures are listed in Figure 1. Formamide (FA) and N-methylformamide (NMF) were purchased from Fluka Chemika-Biochemika (Buchs, Switzerland). N,N(12) Fanali, S. J. Chromatogr. 1991, 545, 437. (13) Nishi, H.; Fukuyama, T.; Terabe, S. J. Chromatogr. 1991, 553, 503. (14) Armstrong, D. W.; Rundlett, K.; Reid, G. L., III. Anal. Chem. 1994, 66, 1690. (15) Ye, B.; Khaledi, M. G. Presented at the 1994 HPCE Meeting, San Diego, CA; Abstract No. 133. (16) Bjoernsdottir, I.; Terabe, S.; Hansen, S. H. Presented at the 1996 HPCE Meeting, Orlando, FL; Abstract No. 118. (17) Valko, I. E.; Siren, H.; Riekkola, M.-L. Presented at the 1996 HPCE Meeting, Orlando, FL; Abstract No. 367.
Dimethylformamide (DMF) was purchased from Fisher Scientific (Pittsburgh, PA). Safety. FA and its derivatives decompose at boiling point to poisonous gases such as NH3, CO, HCN, and H2O at a rate of ∼0.5%/min, and both acids and bases accelerate the decomposition.18 A hood should be used for the heating of FA series media. According to the manufacturer, these solvents are irritating to the eyes and skin. Therefore, proper eye and skin protection is suggested. Procedures. The background electrolyte was prepared by mixing the proper amount of citric acid (or acetic acid) and Tris stock solutions in the running solvent. The appropriate amount of CD was dissolved in the background electrolyte, and then the solution was filtered with a 0.45 µm filter before use. The capillary was conditioned by first rinsing with running solvent, followed by water, 1 M sodium hydroxide, and water, each for 10 min, respectively, and finally with running buffer for 20 min with the running voltage applied. Binding constants were determined from a nonlinear least-squares curve-fitting of eq 2 (µ vs [CD]) using a SAS (Research Triangle Park, NC) program. RESULTS AND DISCUSSION Binding Constants and Concentration of CDs. The binding constant of an enantiomer with CD can be determined by measuring the electrophoretic mobility of an enantiomer at different CD concentrations and by nonlinear regression of the data according to eq 2. The electrophoretic mobilities of enantiomers decreased with the increase of concentration of β-CD in all five solvent systems. The decrease of mobilities for basic solutes can be attributed to inclusion complexation with CD as well as the dramatic increase in the viscosity with the increase in CD concentration. No viscosity correction was made for the pure aqueous medium due to the strong binding between the test solutes and β-CD and the minor viscosity change (3%) that occurred when the concentration of β-CD was increased from 0 to 16 mM.19 The decrease of the mobilities due to the viscosity effect in the other four media was corrected relative to viscosity at [CD] ) 0, η0, by20
µ′ ) µη/η0
(4)
where µ is the measured electrophoretic mobility, µ′ is the corrected electrophoretic mobility, η is the viscosity at a given CD concentration, η0 is the viscosity of the medium at [CD] ) 0, and η/η0 is the relative viscosity. The viscosity was determined at different CD concentrations by a viscometer. Figure 2 shows the variation in mobility with CD concentration that was used for binding constant measurement of thioridazine (11) in water, 6 M urea in water, FA, NMF, and DMF buffer media. The peaks could not be identified due to the lack of availability of pure enantiomers, thus they were assigned according to their elution sequences. The markers in the figure represent the experimental data, and the lines represent the (18) Eberling, C. L. Encyclopedia of Chemical Technology, 3rd ed.; John Wiley & Sons: New York, 1978; Vol. 11, pp 258-263. (19) Paduana, L.; Sartorio, R.; Vitagliano, V.; Constantino, L. J. Solution Chem. 1990, 19, 31. (20) Moelwyn-Hughes, E. A. Physical Chemistry, 2nd ed.; Pergamon Press: Oxford, 1961.
Analytical Chemistry, Vol. 68, No. 19, October 1, 1996
3461
Figure 1. Structures of chiral compounds. Table 1. Binding Constants Determined Using Eq 2 for Mianserin (4), Trimipramine (10), and Thioridazine (11) with β-CD in Different Solvent Media mianserin (4)
trimipramine (10)
thioridazine (11)
mediuma,b
pHc
water 6 M urea FA NMF DMF
3.02 (1.35 ( 0.51) × (2.59 ( 0.73) × (5.02 ( 1.33) × (7.20 ( 2.40) × (2.91 ( 1.43) × (3.54 ( 3.02) × 104 3.90 (2.83 ( 0.47) × 101 (4.66 ( 1.08) × 101 (1.17 ( 0.21) × 103 (1.36 ( 0.36) × 103 (2.98 ( 0.40) × 103 (3.82 ( 0.42) × 103 5.4 2.20 ( 2.00 2.31 ( 1.29 8.57 ( 1.21 (1.03 ( 0.14) × 101 5.39 ( 0.80 6.98 ( 0.88 6.3 0.11 ( 0.25 0.17 ( 0.12 0.34 ( 0.77 0.45 ( 0.46 0.59 ( 0.72 0.59 ( 0.72 5.8 (2.50 ( 0.05) × 10-2 (2.50 ( 0.05) × 10-2 0.52 ( 0.07 0.52 ( 0.07 (5.90 ( 0.02) × 10-2 (5.90 ( 0.02) × 10-2
K1
K2 102
K1 102
K2 103
K1 103
K2 104
a The supporting electrolyte was 50 mM citric acid-25 mM Tris, except in DMF, where it was 200 mM citric acid-50 mM Tris. b Relative viscosity, η/ηo, was solved by multilinear regression fitting at different CD concentration: η/ηo ) 1 + 6.44[CD] - 28.6[CD]2 + 137[CD]3 in 6 M urea; η/ηo ) 1 + 10.1[CD] - 55.7[CD]2 + 209[CD]3 in FA; η/ηo )1 + 6.28[CD] - 3.66[CD]2 + 141[CD]3 in NMF; η/ηo ) 1 + 9.60[CD] 4.38[CD]2 + 348[CD]3 in DMF (r2 > 0.98 for all cases). c The apparent pH (pH*) was used in organic media.
regression results from eq 2. Chiral separation of the enantiomers of 11 was achieved in the β-CD concentration range from 0.04 to 0.5 mM in aqueous buffer (Figure 2A). The two enantiomers eluted at the same time when the concentration of the CD was further increased. This is due to the fact that the binding constants of the enantiomers are large and their binding constant difference is only about 20%. In 6 M urea medium, chiral separation was achieved in the β-CD concentration range from 0.25 to 2.0 mM (Figure 2B). Parts C and D of Figure 2 show the change of corrected mobilities with the β-CD concentration in FA, NMF, and DMF. Chiral separation was obtained in the FA medium when the concentration of β-CD was larger than 20 mM. When the concentration of β-CD was larger than 300 mM, the viscosity of the medium increased dramatically, and the current across the capillary was very small, causing the separation time 3462 Analytical Chemistry, Vol. 68, No. 19, October 1, 1996
window to be very long. In NMF and DMF media, no chiral separation was observed for 11. A linear relationship between corrected mobilities and β-CD concentration was obtained in these two media (Figure 2D), due to the fact that the term K[CD] in the denominator of eq 2 is much smaller than 1. Table 1 shows the binding constants of three solutes with β-CD in five different solvent systems. The influence of organic solvents on the binding constants was compared to those in pure aqueous and in 6 M urea in aqueous buffers. The general trend observed when the medium was changed from pure aqueous to DMF was that the binding constants for the three test compounds decreased with the decrease in the polarity of the solvents. Although the apparent pH (pH*) in different media is not the same, the analytes are fully protonated under the experimental conditions, as
A
B
Figure 3. Relationship between ∆µ′ and β-CD concentration for thioridazine (11) in different solvent systems. Conditions were the same as Figure 2. Curves were calculated from eq 3 using the data from Table 1. CD equilibrium concentration: (A) 0-0.20 M; (B) 0-0.001 M. Curve: (A) aqueous buffer, (B) 6 M urea in aqueous buffer, and (C) FA.
Figure 2. Relationship between mobilities of thioridazine (11) and β-CD concentration in different media. Buffer: (A) 50 mM citric acid, 25 mM Tris in water, pH 3.02; (B) 50 mM citric acid, 25 mM Tris in 6 M urea in water, pH 3.90; (C) 50 mM citric acid, 25 mM Tris in FA, pH* 5.4; (D) 50 mM citric acid, 25 mM Tris in NMF, pH* 6.3 (rectangle with plus), and 200 mM citric acid, 50 mM Tris in DMF, pH* 5.8 (solid triangle). All markers stand for the experimental data and lines for theoretical calculation from eq 2.
evidenced by a lack of variation in mobility in the low pH range.21 Thus, one can conclude that the decrease in the binding constants in Table 1 was caused by the decrease in solvent polarity. Among
the compounds studied, mianserin (4) has the lowest binding constants with β-CD in each medium. The binding abilities for trimipramine (10) and thioridazine (11) with β-CD become much smaller when the medium is changed from water to FA; nevertheless, baseline separations were achieved in FA at higher concentrations of β-CD. For 4, the separation degraded by media changes from water, to 6 M urea in water, to FA. Figure 3 shows the dependence of enantioselectivity for thioridazine (11) in pure water (curve A), in 6 M urea in water (curve B), and in FA (curve C). These curves were calculated according to eq 3 using the binding constant values reported in Table 1. As can be seen in Figure 3, there exists an optimum CD concentration at which enantioselectivity is maximum. In the aqueous buffers, enantioselectivity passes through maxima and changes rapidly with CD concentration. Therefore, it is crucial to operate at the optimum CD concentration for solutes with large binding constants. The optimum concentration of CD is inversely related to the enantiomers’ binding constants to CD as [CD]opt ) (K1K2)-1/2.22 Consequently, for solutes that bind strongly to CD (such as trimipramine (10) and thioridazine (11)), the optimum occurs at very small concentration (e.g., [CD]opt ) 3.1 × 10-5 for 11). The occurrence of the optimum at very low concentration and the rapid changes in enantioselectivity with CD concentration in the aqueous buffers would make it difficult to develop methods based on trial and error. It points out the significance of systematic method development in chiral separations. (21) Williams, R. L.; Vigh, G. J. Chromatogr. 1995, 716, 197. (22) Penn, S. G.; Goodall, D. M.; Loran, J. S. J. Chromatogr. 1993, 636, 149.
Analytical Chemistry, Vol. 68, No. 19, October 1, 1996
3463
Table 2. Effect of Type of CD on Separation in FAa Type of CD β-CD compound chlophedianol (1)
tR
M-β-CD
γ-CD
R
Rs
tR
R
Rs
HP-β-CD
tR
R
Rs
tR
R
9.20
1.000
ns
9.62
1.000
ns
10.35 10.37 9.28
1.007
0.34
11.14
1.000
ns
1.000
ns
1.016
0.83
8.34 8.36 9.08 9.13 8.93 8.96 9.74
1.006
0.42
1.025
1.61
1.016
0.97
1.039
1.82
1.009
0.59
9.92 9.98 8.46 8.56 10.48 10.62 9.46
1.000
ns
1.000
ns
10.37
1.000
ns
10.86 10.98 9.41
1.050
1.94
11.40
1.000
ns
1.000
ns
10.18 10.29 8.44 8.54 10.20
1.030
1.47
1.025
1.61
1.000
ns
1.024
0.88
12.25
1.000
ns
chlorcyclizine (2)
8.73 8.81 10.13
1.000
ns
1.017
0.69
ethopropazine (3)
9.07
1.000
ns
11.23 11.31 10.25
1.000
ns
8.14 8.30 10.01 10.18 9.73
1.047
1.88
1.014
0.68
1.059
1.63
1.014
0.68
1.000
ns
9.99 10.06 9.98 10.05 10.03
1.000
ns
9.00 9.03 11.18
1.009
0.32
10.85
1.000
ns
1.000
ns
10.86
1.000
ns
9.64 9.78 9.03 9.18 10.61 10.85
1.046
1.40
10.87
1.000
ns
1.046
1.60
1.031
1.50
8.25
1.000
ns
1.094
2.17
9.63 9.78 11.33 11.43
1.021
0.85
9.51 9.56
1.016
0.93
mianserin (4) nefopam (5) primaquine (6) propiomazine (7) trihexyphenidyl (8) trimepazine (9) trimipramine (10) thioridazine (11) a
Rs
150 mM citric acid-100 mM Tris, 100 mM CD, pH* ) 5.1. ns, no separation. R ) µ1/µ2.
As the binding constants of the solute decrease upon changing the solvent system to 6 M urea in water (curve B in Figure 3B) and FA (curve C in Figure 3A), the binding constants are reduced, and the optimum concentration shifts to higher values. In FA, however, the enantioselectivity has an asymptotic behavior as a function of CD concentration and maximum enantioselectivity occurs over a wide range of CD concentration; thus, selection of the optimum concentration is not as critical as compared to that in aqueous media. Figure 4 shows the chiral separation of trimipramine (10) using β-CD both in water and in FA. Chiral separation of the solute was observed in aqueous solution at β-CD concentration range of 0.08-1 mM (Figure 4A). No chiral separation was observed as the concentration of β-CD was increased to >1 mM. A baseline separation of the solute was achieved in the FA system at 100 mM β-CD (Figure 4B). Previously, Quang and Khaledi reported the result of the chiral separation of 10 in 20 mM HPβ-CD at pH 2.50 with 50 mM TMA+.23 The baseline separation of this compound was achieved at the cost of a long separation time (∼1 h). According to the binding constant values for 10 in Table 1, the [CD]opt values are 0.17 mM in aqueous buffer and 106 mM in FA. Note that the [CD]opt value represents the optimum equilibrium CD concentration (rather than the analytical concentration). For compounds with large binding constants (such as 10 in aqueous buffer), the optimum CD concentration value is small and is comparable to solute concentration in the sample zone. Therefore, the difference between the equilibrium and analytical concentrations is significant. In such a case, the equilibrium CD concentration would have to be calculated from the analytical CD concentration and analyte concentration. On the other hand, for cases where the binding constants are not too large (such as in FA systems), the optimum CD concentration is significantly larger than solute concentration; thus, the analytical (23) Quang, C.; Khaledi, M. G. J. Chromatogr. 1995, 692, 253.
3464 Analytical Chemistry, Vol. 68, No. 19, October 1, 1996
Figure 4. Chiral separation of trimipramine (10) in aqueous and FA media: (A) 0.2 mM β-CD in 50 mM citric acid, 25 mM Tris in water, pH 3.02; (B) 100 mM β-CD in 150 mM citric, 100 mM Tris in FA, pH* 5.1.
CD concentration is nearly identical to the equilibrium CD concentration. Another important issue is the differences in mechanisms in chiral separations in the aqueous and nonaqueous media. According to the three-point interaction theory for chiral separation,
in aqueous buffer, the inclusion complexation of enantiomers into the CD cavity, driven by hydrophobic interaction, plays an important role in the separation process. The hydrogen-bonding interactions with the hydroxyl groups on the CD rims would provide additional points of interactions. It is not clear whether the same mechanism is applicable for the chiral separation in the amide solvents. Obviously, the hydrophobic interaction between the solutes and CD does not exist in the NACE systems. It is then quite logical to conclude that the inclusion complexation occurs to a small extent, if at all. It seems that the NACE systems are mainly effective for the tricyclic compounds (Figure 1). The best chiral separations were achieved using β-CD and its derivatives, while partial or no separation was achieved using γ-CD (Table 2). This is quite interesting, since tricyclic compounds are typically too large for β-CD and fit γ-CD much better. This observation might suggest that inclusion complexation is not the primary mechanism. Instead, solutes might “lay flat on the mouth” of the CD and interact with the hydroxyl groups on the rims through polar interaction.24 Types of CD. In aqueous chiral separation, three major factors influence selectivity: pH, type of CD, and concentration of CD. Changing the type of CD will change the enantioselectivity. Different CDs were selected for chiral separation in FA (Table 2). Under the experimental conditions, none of the test solutes were resolved by the R-CD, while many of them could be separated by the rest of the CDs. Among the CDs, β-CD could resolve the enantiomers of seven solutes, while HP-β-CD could resolve five of them. As compared with β-CD and its derivatives, γ-CD has the smallest enantioselectivity. Chiral isomers of mianserin (4) and nefopam (5) could be resolved by all of the CDs listed in Table 2, while the enantiomers of trimipramine (10) and thioridazine (11) could be separated by three of the CDs. Chiral separations of chlophedinol (1), ethopropazine (3), propiomezine (7), and trihexyphenidyl (8) could only be achieved by one type of CDs. All these amines are tertiary amines except primaquine (6). From the structural point of view, the position of chiral centers is very important for chiral separation. If a chiral carbon atom is part of a ring, chiral separation can be easily achieved. Compounds such as 4, 5, and 11 are examples. The highest enantioselectivity was achieved when these three compounds were separated by β-CD. The position of the chiral center relative to some polar functional groups is also important. According to HPLC chiral separation results in the polar-organic solvent mode, the presence of hydrogen-bonding groups located at either R or β positions relative to the stereogenic center is an important factor for chiral resolution.25,26 It is not surprising, then, that trimeprazine (9) and trimipramine (10) have similar results, because they have the same side chain, and the chiral center is in the β position relative to the amine group. These two compounds have the second highest enantioselectivity with β-CD. Other functional groups that are directly connected to the chiral center, such as phenyl group, are also important for chiral separation. Chlophedianol (1), chlorcyclizine (2), and trihexyphenidyl (8) contain one or more aromatic rings and/or polar groups at their chiral centers. (24) Armstrong, D. W.; Chen, S.; Chang, C.; Chang, S. J. Liq. Chromatogr. 1992, 15, 545. (25) Zukowski, J.; Pawlowska, M.; Nagatkina, M.; Armstrong, D. W. J. Chromatogr. 1993, 629, 169. (26) Chang, S. C.; Reid, G. L., III; Chen, S.; Chang, C. D.; Armstrong, D. W. Trends Anal. Chem. 1993, 12, 144.
Figure 5. Influence of ionic strength on chiral separation of trimipramine (10). Concentration ratio of citric acid to Tris: (A) 10/5, (B) 50/25, and (C) 100/50. The running electrolyte consisted of 250 mM β-CD in NMF at pH* 5.7.
Common to the structures of 1-3 and 6-8 is that a polar group is directly linked to the chiral centers of the compounds. In our case, the compounds with polar groups right next to the chiral centers are resolved least. Ionic Strength of Media. Figure 5 shows the influence of ionic strength on chiral separation in NMF. The increase of the ionic strength of the system upon increasing the concentration of citric acid to Tris from 10/5 to 100/50 enhanced chiral separation. No further improvement in chiral separation was observed at ratios larger than 100/50. The mobilities of test Analytical Chemistry, Vol. 68, No. 19, October 1, 1996
3465
Figure 6. Influence of pH* on chiral selectivity (2) of mianserin (4) and mobility of the first-eluted enantiomer (9). Conditions: 250 mM β-CD in NMF at a total concentration of acetic acid and Tris equal to 250 mM.
solutes and the electroosmotic flow decreased with the increase of ionic strength. The decrease in the electrophoretic mobilities of the enantiomers and EOF caused an increase of the chiral resolution. The electrophoretic mobility of a solute in NACE is inversely proportional to the ionic strength of the medium, which is the same behavior as reported in the aqueous buffer.27 pH* Effect. It has been recognized that pH has an influence on the EOF and charge status of a solute in aqueous systems. Rawjee and Vigh28,29 showed that different pH values could give totally different enantioselectivity for a given system in three different cases. One drawback with NACE is that the knowledge about acid-base chemistry in organic media is very limited, and it is difficult to directly relate the apparent pH in organic media, pH*, to aqueous pH. The purpose of the measurement of pH* was to monitor the relative value of the pH in organic media to achieve reproducible results. pH* was adjusted by varying the concentrations of acetic acid and Tris while keeping the total concentration of acetic acid and Tris constant. Figure 6 shows the influence of pH* on the electrophoretic mobility and enantioselectivity of mianserin (4) in NMF. It was observed that the mobility of 4 was almost zero when pH* was larger than 9.0, and no chiral separation was observed unless pH* was less than 8.2. This means that, above pH* ) 9.0, 4 is neutral and coeluted with EOF. When pH* was less than 9.0, the mobility increased with a decrease of pH* because of the increase of the effective charge on the solute. The effect of pH* on chiral selectivity is the same as that observed for basic solutes in aqueous media.28,29 Control of EOF. In chiral separations in aqueous CE, both long-chain cationic surfactants and short-chain surfactants have been used to enhance separation.30,31 In a previous study, it was shown that short-chain surfactants have some advantages, such as better capillary wall coverage at low pH, no micelle formation, and less inclusion complexation with CDs.31 Tetramethylammonium (TMA+) and tetrabutylammonium (TBA+) cations were used in FA to test the possibility of controlling EOF. Since tetraalkylammonium (TAA+) cation competes with enantiomers to complex with β-CD, little chiral separation was observed when 100 mM (27) Janini, G. M.; Issaq, H. J. Capillary Electrophoresis Technology; Marcel Dekker: New York, 1993; p 119. (28) Rawjee, Y. Y.; Staerk, D. U.; Vigh, G. J. Chromatogr. 1993, 635, 291. (29) Rawjee, Y. Y.; Vigh, G. Anal. Chem. 1994, 66, 619. (30) Soini, H.; Reikkola, M.; Novotny, M. V. J. Chromatogr. 1992, 608, 265. (31) Quang, C.; Khaledi, M. G. Anal. Chem. 1993, 65, 3354.
3466 Analytical Chemistry, Vol. 68, No. 19, October 1, 1996
Figure 7. Influence of TAA+ on chiral separation of trimipramine (10): (A) without TAA+, (B) with 100 mM TBA+, and (C) with 100 mM TMA+. The running electrolyte was 150 mM citric acid, 100 mM Tris, 250 mM β-CD in FA at pH* 5.1.
β-CD and 100 mM TMA+ coexisted in FA. Higher concentrations of β-CD were used with TAA+ to monitor the influence on chiral separation. Figure 7 shows the effect of addition of TAA+ on chiral separation of trimipramine (10) in 250 mM β-CD. Partial separation was achieved in the absence of short-chain cationic surfactants due to the higher than optimum concentration of chiral selector used (Figure 7A). The addition of 100 mM TBA+ and TMA+ (Figure 7B,C) caused the migration of the test solute to increase from 23 to 50 min. Baseline separations were achieved in TMA+ solution. TMA+ was more effective than TBA+ for control of EOF. The EOF could not be reversed in FA at any TBA+ or TMA+ concentrations. The addition of short-chain surfactants to NMF had very little influence on both migration time and chiral separation and almost no influence in DMF. Temperature Effect. As shown in Figure 8, the migration time of trimipramine (10) in NMF increased from 20 to 30 min when the temperature changed from 40 to 25 °C. As the
A
Figure 8. Influence of temperature on chiral separation of trimipramine (10). (A) 40 °C and (B) 25 °C. The running electrolyte was 150 mM citric acid, 100 mM Tris, 250 mM β-CD in NMF at pH* 5.7.
temperature decreased, better separation was achieved. This was due to the fact that the decrease in temperature increased the viscosity of the medium, and therefore increased the migration time. The change in temperature can also change the binding constants and enantioselectivity. Charged Cyclodextrins. By using charged CD, one can incorporate electrostatic interactions in chiral separation. The use of charged CD in aqueous media has been recently reported by several workers.32-34 In nonaqueous media, the positively charged amines interact more strongly with the negatively charged sulfated β-CD as compared to the neutral β-CD. As a result of the increased binding of chiral solutes, the optimum CD concentration would shift to lower values. This is shown in Figure 9, where trimipramine (10) was successfully separated at much lower concentration of the charged CD than in neutral β-CD. The results of chiral separations by NACE using charged chiral selectors will be published elsewhere.35 CONCLUSIONS This study has shown the possibility of chiral separation by NACE. In varying the solvent of a chiral medium, it is possible to change the binding constants and their difference between chiral selector and enantiomers. The general trend is that for CD-type chiral selectors, the more apolar the solvent, the smaller (32) Schmitt, T.; Engelhardt, H. Chromatographia 1993, 37, 475. (33) Anigbogu, V. C.; Copper, C. L.; Sepaniak, M. J. J. Chromatogr. 1995, 705, 343. (34) Tait, R. J.; Thompson, D. O.; Stella, V. J.; Stogaugh, J. F. Anal. Chem. 1994, 66, 4013. (35) Wang, F.; Khaledi, M. G. Manuscript in preparation for J. Microcolumn Sep.
Figure 9. Chiral separation of trimipramine (10) in FA with β-CDSO4: (A) 5.0 mM β-CD-SO4 in 150 mM citric acid, 100 mM Tris, pH* 5.1; (B) 10 mM β-CD-SO4 in 150 mM citric, 100 mM Tris in FA, pH* 5.1.
the binding constants. It is essential to determine the binding constants in a medium in order to optimize the concentration of the chiral selector. The results have demonstrated that the influences of type of CD, ionic strength, and temperature show the same trends as those in chiral separations by aqueous CE. The effects of pH* on test solutes were the same as those in aqueous CE. EOF could be slowed down by addition of short-chain surfactants; however, it could not be reversed in FA. The anionic β-CD has a higher binding ability with the basic solutes in FA than that of neutral β-CD; thus, chiral separations could be achieved at lower CD concentrations. ACKNOWLEDGMENT A research grant from the U.S. National Institutes of Health (GM 38738) is gratefully acknowledged. The authors are grateful to G. A. Reed of American Maize Products for providing the cyclodextrins and their derivative samples for this study. We thank Dr. C. Brownie of Department of Statistics, North Carolina State University, for providing the SAS computer program for the calculation of binding constants. Received for review June 3, 1996. Accepted July 22, 1996.X AC960537O X
Abstract published in Advance ACS Abstracts, August 15, 1996.
Analytical Chemistry, Vol. 68, No. 19, October 1, 1996
3467