Intrinsic Selectivity in Capillary Electrophoresis for Chiral Separations

Laboratoire de Chimie Organique 2, Université René Descartes, 4, Avenue de l'Observatoire, 75270 Paris Cedex 06, France. Defined as the ratio of the...
0 downloads 0 Views 321KB Size
Download PDF
Anal. Chem. 1997, 69, 393-401

Intrinsic Selectivity in Capillary Electrophoresis for Chiral Separations with Dual Cyclodextrin Systems Fre´de´ric Lelie`vre and P. Gareil*

Laboratoire d’Electrochimie et de Chimie Analytique (URA CNRS 216), Ecole Nationale Supe´ rieure de Chimie de Paris, 11 rue Pierre et Marie Curie, 75231 Paris Cedex 05, France Y. Bahaddi and H. Galons

Laboratoire de Chimie Organique 2, Universite´ Rene´ Descartes, 4, Avenue de l’Observatoire, 75270 Paris Cedex 06, France

Defined as the ratio of the affinity factors of the analytes for a complexing agent, the intrinsic selectivity is representative of the very nature of the complexing agent. When more than one complexing agent are present in the background electrolyte, it is possible to define several intrinsic selectivities according to whether complexing agents are considered separately or all together. A theoretical model with respect to selectivity is presented for separations that involve two complexing agents, using the concept of apparent constant for complex formation. When only independent complexation occurs (absence of mixed complexes), then the intrinsic selectivity of a complexing agent X in the presence of a complexing agent Y can be easily related to the intrinsic selectivity of each complexing agent and to complex formation constants. Dual systems of cyclodextrins (CDs), implementing the cationic mono(6-amino-6-deoxy)-β-cyclodextrin (β-CDNH2) and a neutral CD (trimethyl-β-CD (TM-β-CD) or dimethyl-β-CD (DM-β-CD)), were studied to illustrate this model and to offer an alternative to the separation of neutral enantiomers when β-CD-NH2 shows no or insufficient stereoselectivity. With a dual β-CD-NH2/TM-β-CD system at pH 2.3, arylpropionic acid enantiomers were baseline resolved and benzoin derivatives were partially resolved. For the arylpropionic acids, β-CD-NH2, which is not stereoselective, confers on them a nonzero mobility, while TM-β-CD allows the chiral recognition. A study of the respective influence of ΤM-β-CD and β-CD-NH2 concentrations was performed to determine the optimal conditions with respect to resolution. This theoretical approach allowed characterization of the intrinsic selectivity of neutral CDs for pairs of neutral enantiomers and therefore identification of the potential of neutral chiral agents for neutral enantiomers. Capillary electroseparation techniques offer high separation efficiencies, the possibility of carrying out separations with pure aqueous or hydroorganic medium, rapid analysis times, low reagent consumption, and effluent generation and have been therefore largely studied as alternative or complementary techniques of the conventional chromatographic techniques. One area that have received a lot of attention is the field of chiral separation because of the increasing need for characterization of the optical S0003-2700(96)00607-5 CCC: $14.00

© 1997 American Chemical Society

purity of drugs. Several reviews, describing the analytical applications and principles, have been published.1-5 Most studies have been carried out by capillary zone electrophoresis (CZE) with a chiral selector dissolved in the background electrolyte (BGE). The separation relies then on the formation of in situ diastereoisomeric complexes between the enantiomers and the complexing chiral agent. Resolution of two enantiomers arises from (1) the difference of formation constants and/or (2) the difference in mobilities of the enantiomer-chiral agent complexes. A further requirement is that the mobilities of the free and complexed enantiomers are different. Cyclodextrins (CDs), crown ethers, oligosaccharides, chiral metal chelates, proteins,3-5 and macrocyclic antibiotics6 have been shown to be excellent chiral selectors. Different models have been proposed to describe and predict the influence of the main parameters such as concentration of chiral agent,7-16 pH,13-17 electroosmotic flow velocity,16 nature of the BGE co-ion,17,18 and organic solvent composition.8 In the previous paper,19 we discussed the selectivity concept as used in capillary electrophoresis to characterize separations obtained in the presence of a complexing agent. According to the authors, selectivity for a pair of analytes has been defined as the ratio of migration times, apparent electrophoretic mobilities, effective mobilities, or binding constants between the analytes and the complexing agent. When the analytes are separated on the basis (1) Snopek, J.; Jelı´nek, I.; Smolkova´-Keulemansova´, E. J. Chromatogr. 1992, 609, 1-17. (2) Kuhn, R.; Hoffstetter-Kuhn, S. Chromatographia 1992, 34, 505-512. (3) Terabe, S.; Otsuka, K.; Nishi, H. J. Chromatogr. 1994, 666, 295-319. (4) Nishi, H.; Terabe, S. J. Chromatogr. 1995, 694, 245-276. (5) Lelie`vre, F.; Gareil, P.; Caude, M. Analusis 1994, 22, 413-429. (6) Armstrong, D. W.; Rundlett, K. L.; Chen, J.-R. Chirality 1994, 6, 496-509. (7) Wren, S. A.; Rowe, R. C. J. Chromatogr. 1992, 603, 235-241. (8) Wren, S. A.; Rowe, R. C. J. Chromatogr. 1992, 609, 363-367. (9) Wren, S. A.; Rowe, R. C. J. Chromatogr. 1993, 635, 113-118. (10) Wren, S. A. J. Chromatogr. 1993, 636, 57-62. (11) Wren, S. A.; Rowe, R. C.; Payne, R. S. Electrophoresis 1994, 15, 804-807. (12) Penn, S. G.; Bergstro¨m, E. T.; Goodall, D. M.; Loran, J. S. Anal. Chem. 1994, 66, 2866-2873. (13) Rawjee, Y. Y.; Staerk, D. U.; Vigh, G. J. Chromatogr. 1993, 635, 291-306. (14) Rawjee, Y. Y.; Williams, R. L.; Vigh, G. J. Chromatogr. 1993, 652, 233245. (15) Rawjee, Y. Y.; Vigh, G. Anal. Chem. 1994, 66, 619-627. (16) Rawjee, Y. Y.; Williams, R. L.; Vigh, G. J. Chromatogr. 1994, 680, 559607. (17) Rawjee, Y. Y.; Williams, R. L.; Vigh, G. Anal. Chem. 1994, 66, 3777-3781. (18) Bechet, I.; Paques, P.; Fillet, M.; Hubert, P.; Crommen, J. Electrophoresis 1994, 15, 818-823. (19) Lelie`vre, F.; Gareil, P.; Jardy, A. Anal. Chem. 1997, 69, 385-392.

Analytical Chemistry, Vol. 69, No. 3, February 1, 1997 393

of differences of affinity with a complexing agent, each of these definitions, but for the latter, is not completely representative of the very nature of the agent. We proposed to define the intrinsic selectivity of a complexing agent C for a pair of analytes A and B as the ratio of their affinity factors, the affinity factor being defined as

k)

number of moles of complexed analyte number of moles of free analyte

(1)

Concurrently, the affinity coefficient, Da, is defined as the ratio of the concentration of the analyte interacting with the separating agent to the concentration of the free analyte (in capillary electrophoresis, k ) Da). We then proposed to define a K′ parameter as

K′ ) k/[C] ) Da/[C]

(2)

K′ (in M-1) depends on the equilibrium constants for selectoranalyte binding and on the complexing agent concentration. Then,

R ) kB/kA ) K′B/K′A

(3)

When only 1:1 complexation occurs, then the K′ parameters are the equilibrium constants (or apparent equilibrium constants) that characterize the complex formation and the selectivity is independent of the concentration of the complexing agent. This approach allows the laying of the stress on the very nature of the complexing agent, and for the case of chiral separations, it is characteristic of the chiral recognition that occurs.19 The intrinsic selectivity definition is coherent with the one developed in conventional chromatography and micellar electrokinetic capillary electrochromatography (MECC). Thus, MECC formalism can be developed. When the analytes have the same mobility in their free and complexed forms, MECC equation resolution is valid and optimum conditions (affinity factor, complexing agent concentration) can be predicted.19 Native and modified neutral CDs have been the most commonly used chiral resolving agents. CDs are chiral cyclic oligosaccharides with a shape similar to a truncated cone with a relatively hydrophobic cavity. The formation of inclusion complexes between enantiomers and CDs is influenced by the hydrophobic interaction in the cavity and bondings between the hydroxyl groups (or other substituents) on the rim of CDs and substituent groups of the asymmetric center of the analytes. These neutral CDs are well suited for the analysis of charged analytes but cannot be used directly for the separation of neutral enantiomers. For these separations, one approach (direct approach) consists of using charged CDs that are chemically modified with a substituent that carries a permanent charge (sulfobutyl-ether-β-CD (SBE-β-CD)20,21) or an acid-base group (carboxymethyl-β-CD (CM-β-CD),22 mono(6-amino-6-deoxy)-β-CD (20) Chankvetadze, B.; Endresz, G.; Blaschke, G. Electrophoresis 1994, 15, 804807. (21) Tait, R. J.; Thompson, D. O.; Stella, V. J.; Stobaugh, J. F. Anal. Chem. 1994, 66, 4013-4018. (22) Schmitt, T.; Engelhardt, H. J. High Resolut. Chromatogr. 1993, 16, 525529.

394

Analytical Chemistry, Vol. 69, No. 3, February 1, 1997

(β-CD-NH2).19,23,24 The other approach (indirect approach) involves the simultaneous use of an achiral agent that confers a nonzero mobility to the analyte and of neutral CDs that permit the chiral recognition. Neutral chiral hydrophobic analytes have been thus separated with CD-MECC systems that involve partitioning between the bulky phase and the micellar pseudophase and complexation with the CDs.25,26 Carbohydrate enantiomers form complexes with borate ions and thereby have been resolved with a borate-CD system.27 Recently, Anigbogu et al.28 obtained the separation of neutral aminoglutethimide enantiomers in the presence of a charged CD (CM-β-CD) and a neutral CD (β-CD). This approach can be considered as indirect in the sense where the enantiomers were not resolved when only CM-β-CD was present in the background electrolyte. Resolution of the enantiomers with a CM-β-CD/β-CD system was improved when methanol was used (50% v/v). This approach has been called dual CDCE. Systems that involve only CDs offer the advantage over CDMECC systems of being more compatible with organic solvents and more predictible since surfactant monomers include themselves in the CD cavity and thereby intervene in the complexation of the analyte.28,29 The aim of this paper is to show how the concept of intrinsic selectivity19 applies to systems that contain two complexing agents and to develop the potential of charged CDs, when these are not or weakly enantioselective toward neutral enantiomers, using dual CD systems. Separation of arylpropionic acids, benzoin, and methyl ether benzoin enantiomers were studied at pH 2.3 with (β-CD-NH2/trimethyl-β-CD (TM-β-CD) or dimethyl-β-CD (DM-βCD) dual systems. A study of the respective influence of ΤM-βCD and β-CD-NH2 concentrations was performed to determine the optimal conditions with respect to resolution. Assuming that independent complexation occurs (no mixed complex), selectivity study was used to identify the potential of neutral chiral agents for neutral enantiomers. THEORY Intrinsic Selectivity of a Complexing Agent Y in the Presence of a Given Concentration of a Complexing Agent X. The theoretical study of the selectivity for a pair of analytes, A and B, in the presence of two complexing agents, X and Y, was carried out considering as hypotheses that only 1:1 complexation occurs and that the two complexing agents lead to independent complexation (absence of mixed complex). These hypotheses are appropriate if X and Y are CDs since inclusion complexes are generally 1:1 host-guest complex. (A more general study of the selectivity when analytes A and B are both complexed by X only, Y only, and X and Y simultaneously can be formally derived from the presentation given in ref 19 for the case of a weak acid in the presence of a CD, the second complexing agent taking the place of the hydronium ion.) According to the present hypotheses, the complex formation reactions are (M ) A, B) (23) Nardi, A.; Eliseev, A.; Bocek, P.; Fanali, S. J. Chromatogr. 1993, 638, 247253. (24) Fanali, S.; Aturki, Z. J. Chromatogr. 1995, 694, 297-305. (25) Otsuka, K.; Terabe, S. J. Liq. Chromatogr. 1993, 16, 945-953. (26) Ueda, T.; Kitamura, F.; Mitchell, R.; Metcalf, T. Anal. Chem. 1991, 63, 29792981. (27) Stefansson, M.; Novotny, M. J. Am. Chem. Soc. 1993, 115, 11573-11580. (28) Anigbogu, V. C.; Copper, C. L.; Sepaniak, M. J. J. Chromatogr. 1995, 705, 343-349. (29) Sepaniak, M. J.; Copper, C. L.; Whitaker, K. W.; Anigbogu, V. C. Anal. Chem. 1995, 67, 2037-2041.

M + X S MX

KX ) [MX]/[M][X]

(4)

M + Y S MY

KY ) [MY]/[M][Y]

(5)

This system can be described by an apparent equilibrium valid at a given concentration of X

M′ + Y S MY

(6)

and by an apparent equilibrium constant K′Y

K′Y )

KY [MY] ) [M′][Y] 1 + KX[X]

(7)

with

[M′] ) [M] + [MX]

(8)

K′Y denotes the apparent constant for complex formation with complexing agent Y, valid at a given concentration of X. Defined as the ratio of the affinity factors of the analytes A and B, the intrinsic selectivity of X (Y, respectively) is equal to the ratio of equilibrium constants KX (KY, respectively):

RX ) KBX /KAX

(9)

RY ) KBY /KAY

(10)

(a) If all three selectivity values R, RX, and RY are known for a pair of analytes, then an inequality between R and the ratio RY/ RX will indicate the presence of a mixed complex such as a 1:1:1 complex. (b) If one of the selectivities cannot be determined directly by an electrophoresis experiment (for example, intrinsic selectivity of a neutral complexing agent for a pair of neutral analytes), then it can be determined from eq 14 by assuming that no mixed complex is present. (c) A system with two complexing agents is beneficial to the separation only if the affinity order of the analytes for each complexing agent is opposite. (d) If one of the complexing agent is not selective (for example KAX ) KBX ), then R ) RY whatever the concentration of X is. It is to be noted that in this theoretical part, no consideration was given to the order of complexation, but numerical applications below are presented so that selectivity values are superior to 1. Determination of Intrinsic Selectivity. The determination of the effective mobility values of the analytes in the presence of X only, Y only, and Y in the presence of a given concentration of X upon the concentration of the complexing agent allows the calculation of RX, RY, and R, respectively. In effect, the effective mobility of an analyte M is a function of its mobility in the absence of complexing agent, µf, and in excess of complexing agent, µc (where the analyte is totally complexed), of the concentration of the complexing agent, and of the equilibrium constant or apparent equilibrium constant. Especially, the effective mobility of analyte M in the presence of Y and a given concentration of X is

and

The affinity factor of analyte M for complexing agent Y in the presence of a given concentration of X is

K′Y[Y] 1 µf + µ 1 + K′Y[Y] 1 + K′Y[Y] c

(15)

KX[X] 1 µM + µ 1 + KX[X] 1 + KX[X] MX

(16)

µM )

with

k)

µf )

number of moles of M complexed by Y number of moles of free M + number of moles of M complexed by X

(11) and

(17)

and thereby the intrinsic selectivity of Y in the presence of a given concentration of X is

µc ) µMY

R ) kBY/kAY ) K′BY /K′AY

where µM, µMX, and µMY are the absolute mobilities of free M and complexes MX and MY, respectively. Combining eqs 12 and 15 yields selectivity R

Combining eqs 7 and 12 gives

( )(

KBY 1 + KAX[X]

R)

KAY

1+

)

KBX [X]

(12)

R) (13)

R is independent of the concentration of Y and is clearly representative of the influence of the nature of Y on the separation for a given concentration of X. When KX[X] . 1, i.e., the free form of M is negligible relative to the form complexed by X, then

R ) RY/RX It ensues from this approach that the following apply:

(14)

(

)(

)

µf,B - µB µA - µc,A µf,A - µA µB - µc,B

(18)

The determination of equilibrium constant or apparent equilibrium constant can be done in various ways from eq 15 and mobility data obtained as a function of the complexing agent concentration by exploiting linear30 or nonlinear12,31-33 curve-fitting procedures or by determining the inflection point of the curve of mobility (30) Rundlett, K. L.; Armstrong, D. W. J. Chromatogr. 1996, 721, 173-186. (31) Shibukawa, A.; Lloyd, D. K.; Wainer, I. W. Chromatographia 1993, 35, 419429. (32) Rogan, M. M.; Altria, K. D.; Goodall, D. M. Electrophoresis 1994, 15, 808817. (33) Pen, S. G.; Goodall, D. M.; Loran, J. S. J. Chromatogr. 1993, 636, 149-152.

Analytical Chemistry, Vol. 69, No. 3, February 1, 1997

395

Figure 1). If z varies and if RX < RY, then R3 varies between RX and RY. R3 is the selectivity of the two-complexing agent system for which the concentration ratio is equal to z. When studying the influence of the concentration of the two complexing agents while keeping z ratio constant (dotted line), the effective mobility of the analytes can be expressed as a linear combination of the mobility of the free analyte, µM, and the mobility of the analyte when x and y tend toward infinity (z constant), µzc. µzc is equal to

µzc ) Figure 1. Experimental design for the study of the influence of the concentration ratio between β-CD-NH2 and TM-β-CD (x and y represent the concentration values of β-CD-NH2 and TM-β-CD respectively).

plotted against the logarithm of the complexing agent concentration.34,35 At this point, half-complexation occurs and the constant is equal to the reciprocal of the concentration of the complexing agent. In this work, constants were roughly estimated from the determination of the concentration at half-complexation. Influence of the Concentration Ratio of the Two Complexing Agents. To clearly understand the influence of the concentration ratio of the two complexing agents, a study was carried out according to the experimental design presented in Figure 1. In the above theoretical part, we showed that it is possible to define two selectivity parameters: the selectivity of X in the presence of a given concentration of Y ([Y] ) y), R1 (horizontal direction), and the selectivity of Y in the presence of a given concentration of X ([X] ) x), R2 (vertical direction).

( )( )

(19)

( )( )

(20)

KBX 1 + yKAY

R1 )

and

KAX 1 + yKBY

KBY 1 + xKAX

R2 )

KAY 1 + xKBX

In general, R1 is different from R2. Another approach is to consider the system composed on the one side of the analyte complexed either by X or Y and on the other side by the free analyte. It is possible to define a global affinity factor, k, as

k)

number of moles of M complexed by X and by Y number of moles of free M (21)

The selectivity defined as the ratio of the global affinity factors is

xKBX + yKBY R3 )

KBX + zKBY ) xKAX + yKAY KAX + zKAY

(22)

with z ) y/x. R3 is constant when z is constant (dotted line in (34) Gareil, P.; Pernin, D.; Gramond, J.-P.; Guyon, F. J. High. Resolut. Chromatogr. 1993, 16, 195-197. (35) Lelie`vre, F.; Gareil, P. J. Chromatogr. 1996, 735, 311-320.

396

Analytical Chemistry, Vol. 69, No. 3, February 1, 1997

KX zKY µMX + µ KX + zKY KX + zKY MY

(23)

At constant value of z, there is a threshold for x and y values beyond which the analytes are completely complexed by X and Y. Thereby above this threshold, the resolution is constant, this one depending on the difference of mobility µzc between the two analytes and on the efficiency. Thus, increasing X and Y concentrations beyond this threshold is useless, except perhaps for improving the solubility of some analytes. EXPERIMENTAL SECTION Chemicals. Mono(6-amino-6-deoxy)-β-cyclodextrin (β-CDNH2) was synthesized by Y.B. and H.G. in the Laboratory of Organic Chemistry 2 of University Rene´ Descartes. Dimethyl-βCD with a substitution degree of 1.8 (DM-β-CD) was a gift of Wacker-Chemie (Wacker-Chemie, Munich, Germany and Lyon, France). Rhoˆne-Poulenc Rorer (Vitry-Alfortville, France) graciously supplied the arylpropionic acids analytes [carprofen (Car), flurbiprofen (Flu), ketoprofen (Ket), naproxen (Nap), suprofen (Sup)] and Dr. A. Jutand (Ecole Normale Supe´rieure, Paris) supplied the racemic naproxen sample. Hepta(2,3,6-tri-O-methyl)β-CD (TM-β-CD), benzoin (Be), and methyl ether benzoin (MeBe), were obtained from Sigma-Aldrich-Fluka (L’Isle d’Abeau Chesnes, France). pH 2.3 buffers were prepared with orthophosphoric acid (85% by weight), ammediol (2-amino-2-methyl-1,3propanediol), and β-CD-NH2. All buffers were prepared using water from an Alpha-Q water purification system (Millipore, Bedford, MA). Buffers were filtered and thoroughly degassed prior to use. β-CD-NH2 basicity (pKa ≈ 8.2, fully ionized at pH 2.3) and the presence of sodium ions as an impurity of the β-CDNH2 lot (0.4% by weight) were taken into account in the preparation of the 24 mM ionic strength buffers. Therefore, 40.8 mM phosphoric acid/24 mM ammediol and 34 mM phosphoric acid/ 20 mM β-CD-NH2 (4 mM Na+) pH 2.3 buffers were prepared, and buffers of intermediate concentrations of β-CD-NH2 (1, 5, and 10 mM) were obtained by mixing these two buffers. β-CD-NH2/TMβ-CD dual systems were prepared by addition of the relevant amount of TM-β-CD to these buffers. Apparatus and Conditions. A HP 3DCE capillary electrophoresis system (Hewlett-Packard, Waldbronn, Germany) equipped with a diode array detector was used for the capillary electrophoresis experiments. All experiments were carried out using the following conditions: untreated fused-silica capillaries, 50 µm i.d. × 38.5 cm (30 cm to the detection window) from Supelco (Bellefonte, PA); capillary thermostated at 25 °C; pressure injection: 4 s at 25 mbar. The viscosity of 34 mM phosphate/20 mM β-CD-NH2/TM-βCD electrolytes was measured using a Haake microviscometer (Haake, Karlsruhe, Germany) thermostated at 20 °C.

Figure 3. Effective electrophoretic mobility of the arylpropionic acids studied as a function of β-CD-NH2 concentration. Conditions: untreated fused-silica capillary, 50 µm i.d. × 38.5 cm (30 cm to the detection window); phosphoric acid/ammediol/β-CD-NH2 buffer, pH 2.3 (ionic strength, 24 mM); V ) 20 kV (I ) 33 µA); UV detection at 240 (Car and Flu), 200 (Ket), 230 (Nap), or 254 nm (Sup); hydrodynamic injection, 4 s at 25 mbar.

Figure 2. Influence of sample preparation on the electrophoretic profile of ketoprofen at pH 2.3 in the presence of 20 mM β-CD-NH2. Conditions: untreated fused-silica capillary, 50 µm i.d. × 38.5 cm (30 cm to the detection window); 34 mM phosphoric acid/20 mM β-CD-NH2 buffer, pH 2.3 (ionic strength 24 mM); V ) 20 kV; UV detection at 254 nm; hydrodynamic injection, 4 s at 25 mbar; sample 0.5 mM ketoprofen prepared in a methanol/water mixture (a) 10:90 and (b) 50:50 (v/v); eo, electroosmosis signal.

RESULTS AND DISCUSSION β-CD-NH2 Enantioselectivity. At pH below 6, β-CD-NH2 is cationic and is directly suitable for the chiral separation of neutral enantiomers. In our previous study, baseline resolution was shown for the neutral enantiomers of chlorthalidon and hydrobenzoin and partial resolution was obtained for MeBe and Be at pH 2.3, the respective values of the intrinsic selectivity being equal to 1.43, 1.35, 1.2 and 1.15.19 β-CD-NH2 enantioselectivity with respect to some chiral arylpropionic acids (Car, Flu, Nap, Sup, Ket) was also studied at pH 2.3, the pH at which they are nearly completely protonated and therefore very hydrophobic. In the first experiments, these analytes were prepared in a 50:50 (v/v) water/methanol mixture. The analysis of these samples with pH 2.3 buffers in the presence of β-CD-NH2 at concentrations as high as 20 mM was characterized by baseline perturbations or the appearence of extraneous peaks (Figure 2). This phenomenon is similar to the one observed in MECC for hydrophobic compounds: the higher the organic solvent content in the sample is, the more perturbed the analytical system is due to the solubilizing effect of the organic solvent. In the presence of β-CD-NH2, the arylpropionic acids form inclusion complexes, which confers on them a nonzero mobility. The analytes have, however, a strong affinity for methanol, which migrates at the velocity of the electroosmotic flow. This leads to a splitting of the analyte zones. Whatever the percentage of methanol (from 10 to 50%), the mobility of the first recorded peak on the electropherograms remained constant, which shows that this peak corresponds to the fraction of analyte that was the less

influenced by methanol and the more easily complexed by the CD. The lower the organic content was, the less the system was perturbed (Figure 2). This study shows the importance of sample preparation protocols. The analytes were then prepared in the minimum amount of organic solvent. The effective mobility of these analytes first increased upon increasing the concentration of β-CD-NH2 (1-5 mM) and then plateaued at higher CD concentrations (5-20 mM) (Figure 3). These results indicate that nearly complete complexation was reached at 20 mM β-CD-NH2 for all the analytes but for Ket. No chiral separation of the arylpropionic acids was observed within this range of β-CD-NH2 concentrations at this pH. This was expected for Car, Flu, Sup, and Ket since some previous studies have shown that the native β-CD was not enantioselective with regard to these enantiomers24,35 and β-CD-NH2 enantioselectivity is likely to be similar to that of β-CD, the amino substituent of β-CD-NH2 being located on one of the primary hydroxyl groups. However, the absence of resolution for Nap enantiomers in the presence of β-CD-NH2 contrasts with their separation at pH 4-6 in the presence of hydroxypropyl-β-cyclodextrin (HP-β-CD) and (hydroxypropyl)methylcellulose obtained by Rawjee et al.,16 the enantioselectivity of HP-β-CD being expected to also be similar to that of β-CD, if only its primary hydroxyl groups are derivatized. From the results of the experiments performed at pH 2.3 with β-CD-NH2 as the single complexing agent, it was possible to rapidly estimate the equilibrium constant for the complexation with β-CD-NH2, Kβ-CD-NH2. The constant was simply evaluated as the reciprocal concentration of β-CD-NH2 leading to an effective electrophoretic mobility equal to half the complex absolute mobility, µc. For Be and MeBe, µc values were taken from Table 2 in our preceding paper.19 For the studied APAs, µc values were roughly estimated from the inspection of the plateau displayed in Figure 3 at high β-CD-NH2 concentration. The Kβ-CD-NH2 values obtained were 330, 300, and 400 M-1 for Ket, Be, and MeBe, respectively, of the order of 1000 M-1 for Nap and Car, and in excess of 1000 M-1 for Flu and Sup. This indicates that in the presence of 20 mM β-CD-NH2, ∼85% of Ket, Be, and MeBe and at least 95% of Car, Flu, Nap, and Sup are complexed (e.g., for Kβ-CD-NH2 ) 350 M-1, the ratio of complexed to free forms is 75, so the species is 87.5% complexed). β-CD-NH2/TM-β-CD Dual System. A previous study of arylpropionic acid enantiomers in buffers of various pH (pH 4, 6, Analytical Chemistry, Vol. 69, No. 3, February 1, 1997

397

Figure 4. Effective electrophoretic mobility of the enantiomers of ketoprofen and naproxen as a function of TM-β-CD concentration in the presence of 20 mM β-CD-NH2 at pH 2.3. Conditions: untreated fused-silica capillary, 50 µm i.d. × 38.5 cm (30 cm to the detection window); 34 mM phosphoric acid/20 mM β-CD-NH2/TM-β-CD buffer, pH 2.3 (ionic strength 24 mM); V ) 20 kV; UV detection at 200 (Ket) or 230 nm (Nap); 0.5 mM sample prepared in 34 mM phosphoric acid/20 mM β-CD-NH2 buffer containing 2% methanol and 1% DMF; hydrodynamic injection, 4 s at 25 mbar.

8, and 10; the pH at which the acids are partially or fully ionized) with various neutral CDs (β-CD, hydroxypropyl-β-CD, dimethylβ-CD, trimethyl-β-CD, hydroxypropyl-γ-CD) revealed that ΤM-βCD was the most selective CD.35 Thus, the enantioselectivity of a β-CD-NH2/TM-β-CD dual system was investigated for these enantiomers and also for Be and MeBe at pH 2.3. The influence of TM-β-CD concentration (between 0 and 40 mM for Car, Flu, Nap, and Sup and 0 and 125 mM for Ket, Be, and MeBe) was studied with a 34 mM phosphate buffer, pH 2.3, containing 20 mM β-CD-NH2. As stated before, the analytes are nearly fully complexed in the presence of solely 20 mM β-CD-NH2. Analytes (0.5 mM) were prepared in the 34 mM phosphate/20 mM β-CDNH2 buffer, pH 2.3, supplemented with 2% methanol and 1% dimethylformamide (DMF). The presence of β-CD-NH2 in the sample permits the solubilization of the analytes with a minimum of organic solvent and thus minimizes effects due to sample preparation. DMF was used as a marker of the electroosmotic flow. The variations of the effective mobility of Nap and Ket enantiomers with TM-β-CD concentration are shown in Figure 4. As expected, analyte complexation with neutral TM-β-CD led to a decrease of its mobility. The curves obtained for Car, Flu, Sup, Be, and MeBe exhibited a similar decline (not shown). The higher the TM-β-CD concentration, the more the mobility is close to the mobility of the analyte-TM-β-CD complex, this latter being equal to zero for neutral analytes and located between 0 and -0.5 × 10-5 cm2/V‚s for the weakly ionized analytes such that the arylpropionic acids at pH 2.3. This study enables one to estimate the apparent constant of complex formation with TM-β-CD in the presence of 20 mM β-CDNH2 at pH 2.3 using the determination of the complexing agent concentration at half-complexation. Measured effective mobilities were corrected for viscosity variations according to the experimentally established relationship η ) 1.05 + 3.64[TM-β-CD] (where η denotes viscosity in mPa‚s). K′TM-β-CD estimated values were 30, 15, and 10 M-1 for Ket, Be, and MeBe, respectively. These low values are well representative of the competition between the two complexing agents. Applying eq 7, the inclusion constant of these analytes with TM-β-CD, KTM-β-CD, could easily 398

Analytical Chemistry, Vol. 69, No. 3, February 1, 1997

Figure 5. Resolution of the enantiomers of arylpropionic acids, benzoin, and methyl ether benzoin as a function of TM-β-CD concentration in a β-CD-NH2/TM-β-CD dual system at pH 2.3. Conditions are as in Figure 4.

Figure 6. Separation of the enantiomers of naproxen with a β-CDNH2/TM-β-CD dual system at pH 2.3. Conditions are as in Figure 4 except for buffer, 34 mM phosphoric acid/20 mM β-CD-NH2/10 mM TM-β-CD, and detection wavelength, 230 nm.

be derived. The estimated values were 230, 100, and 90 M-1 for Ket, Be, and MeBe, respectively. Baseline resolution was obtained for the arylpropionic acid enantiomers with the 20 mM β-CD-NH2/ΤM-β-CD dual system at pH 2.3 (Figure 5). Assuming that there is no mixed complex, this dual system is such that β-CD-NH2 confers a nonzero mobility to the enantiomers while ΤM-β-CD permits their chiral recognition. The chiral separation of these enantiomers at pH 2.3 with TM-β-CD confirms the fact that the selectivity relies on the difference of stability of the complexes formed between the protonated forms of these enantiomers and the neutral CD.13,15,35 The chiral separation of the enantiomers of naproxen is presented in Figure 6. MeBe and Be enantiomers were also partially resolved with this dual cyclodextrin system (Figure 5). As expected, resolution initially increases with increasing TM-β-CD concentration and then reaches a maximum before declining at higher concentrations. Intrinsic Selectivity Determination in Dual Cyclodextrin Systems. In this study, the inclusion complexes migrate slower than the free analytes. If by convention index 2 is attributed to the slowest enantiomer, then selectivity should be written

R)

(

)(

)

k2 µf,2 - µ2 µ1 - µc,1 ) k1 µf,1 - µ1 µ2 - µc,2

(23)

Table 1. Intrinsic Selectivity of TM-β-CD in the Presence of 20 mM β-CD-NH2 for the Enantiomers of Arylpropionic Acids, Benzoin, and Methyl Ether Benzoin, and Optimal Resolution Obtained with 20 mM β-CD-NH2/TM-β-CD Dual Systems at pH 2.3a

R RS(opt)

Car

Flu

Ket

Nap

Sup

Be

MeBe

1.13 1.35

1.16 g1.7

1.21 2.1

1.42 3.1

1.14 g1.6

1.10 g1.2

1.10 1.1

a The analytical conditions are the same as in Figure 4. R was S calculated using the formula RS ) 1.177((t2 - t1)/(δ1 + δ2)), where t is the migration time and δ the width at half-height.

In the presence of 20 mM β-CD-NH2 alone, µf,1 ) µf,2 for the studied enantiomers. Assuming that µc,1 ) µc,2 (excess of TM-βCD and in the presence of 20 mM β-CD-NH2), which is generally the case for inclusion complexes with CDs, then selectivity can be more simply expressed

R)

(

)(

)

µf - µ2 µ1 - µc µf - µ1 µ2 - µc

The determination of the intrinsic selectivity of TM-β-CD in the presence of 20 mM β-CD-NH2 from this equation requires knowing µf and µc with great precision. This difficulty can be overcome by considering that the intrinsic selectivity should be independent of TM-β-CD concentration and thereby, by trying to make the ratio [(µf - µ2)/(µf - µ1)][(µ1 - µc)/(µ2 - µc)] converge toward a constant value, whatever the CD concentration, giving different values to the pair (µf, µc). For all the studied analytes, it was possible to find such a pair with the µf value remaining between the measured value µexp and µexp + 0.5 × 10-5 cm2/V‚s f f -5 and the µc value between 0 and -0.5 × 10 cm2/V‚s (µc ) 0 for Be and MeBe). The resulting selectivity values and the obtained optimal resolutions are shown in Table 1. Note that prior to the determination of R values, the measured mobilities were corrected for viscosity variations and for slight inacurracy on electroosmosis velocity measurement caused by the slight inclusion of DMF inside the CD cavity. It appears from Table 1 and Figure 5 that the selectivity orders correspond to the optimal resolution order. This result was expected since the migration window was the same for all the analytes and differences in efficiency (N between 20 000 and 35 000) were not of primary importance. In the presence of 20 mM β-CD-NH2 alone, the arylpropionic acid enantiomers are nearly fully complexed. Thus, assuming that mixed complexation does not occur, the intrinsic selectivity of TM-β-CD in the presence of 20 mM β-CD-NH2 is equal, according to eq 14, to the ratio of the intrinsic selectivity of each complexing agent. Since β-CD-NH2 is not enantioselective for these analytes, the measured selectivity with the 20 mM β-CD-NH2/TM-β-CD dual system is equal to the intrinsic selectivity of TM-β-CD alone. The obtained values confirm that TM-β-CD is a good chiral agent for these enantiomers. Enantiomer assignments for Nap and Ket were made by spiking the racemates with a pure optical isomer [(S)-(+)-Nap, (S)-(+)-, or (R)-(-)-Ket]. (S)-(+)-Nap and (R)-()-Ket appeared to be more retained and therefore more complexed by TM-β-CD in this dual system than (R)-(-)-Nap and (S)-(+)Ket. Since β-CD-NH2 is not stereoselective, this affinity order corresponds to the one for TM-β-CD. The migration order

obtained with this β-CD-NH2/TM-β-CD dual system using a fusedsilica capillary is the opposite of the one observed when the separation is performed with a buffer containing solely TM-β-CD and of a pH such that the acids are ionized.35 That migration order would be the same as the one obtained under conditions of suppressed electroosmosis and in the presence of TM-β-CD alone. This depicts the versatility of capillary electrophoresis with regard to the control of the migration order, this aspect being of paramount importance for optical purity determinations. The intrinsic selectivity of TM-β-CD in the presence of 20 mM β-CD-NH2 that was calculated for the enantiomers of Be (R ) 1.10) and MeBe (R )1.10) differs from the one derived when β-CDNH2 was solely present (Rβ-CD-NH2 ) 1.15 for Be and 1.20 for MeBe19). This difference indicates that ΤM-β-CD is enantioselective with regard to these analytes. The β-CD-NH2 stereoselectivity study showed that the (R)-(-)-Be enantiomer migrates first, i.e., has a higher affinity for β-CD-NH2 than its antipode and the experiment with the 20 mM β-CD-NH2/TM-β-CD dual system shows that the (R)-(-)-Be enantiomer still migrates first. Knowing these two orders of migration and considering that in the presence of 20 mM β-CD-NH2 alone these enantiomers are nearly fully complexed (see above), then eq 14 applies and it can be inferred that the intrinsic selectivity of TM-β-CD is equal to Be R-Be S-Be RTM-β-CD ) KTM-β-CD /KTM-β-CD ) Rβ-CDNH2/R ) 1.05

and the (R)-(-)-Be enantiomer has a higher affinity for TM-β-CD than its antipode. Considering the same order of affinity for the enantiomers of MeBe, then the intrinsic selectivity of TM-β-CD for these enantiomers would be equal to 1.09. In fact, in the presence of 20 mM β-CD-NH2, Be and MeBe enantiomers are ∼85% complexed. Therefore eq 13 should be used for a more precise calculation of the intrinsic selectivity of TM-β-CD. For R-Be Be, knowing that Kβ-CD-NH2 is equal to ∼300 M-1 (see above), then RTM-β-CD ) 1.02. Similarly, for MeBe, Kβ-CD-NH2 ) 400 M-1 and then RTM-β-CD ) 1.07. The examples of Be and MeBe are representative of the situation in which the enantiomers have the same order of affinity for the two complexing agents, and thereby, the intrinsic selectivity obtained with the dual system is lower than the selectivity produced by the best stereoselective agent. This result expresses that neutral TM-β-CD is less stereoselective than β-CD-NH2 for Be enantiomers. The presence of secondary hydroxyl groups at the rim of β-CD-NH2 cavity renders possible the formation of hydrogen bonds with the analytes and may thereby allow a better chiral recognition. Similarly, the stereoselectivity of neutral DM-β-CD with regard to all of these analytes was studied with a 20 mM β-CD-NH2/ DM-β-CD dual system at pH 2.3. Among the arylpropionic acids, only the naproxen enantiomers were separated. The intrinsic selectivity of DM-β-CD for these enantiomers at this pH was considered to be equal to the intrinsic selectivity of DM-β-CD in the presence of 20 mM β-CD-NH2 since β-CD-NH2 alone was not enantioselective. This leads to a value of 1.18. This shows that DM-β-CD is less stereoselective for the Nap enantiomers than TM-β-CD. These results are in good agreement with previous ones obtained at pH 4 with DM-β-CD or TM-β-CD as single complexing agent.35 Slight resolutions (RS ≈ 0.5) were also obtained for Be and MeBe enantiomers with the 20 mM β-CD-NH2/DM-β-CD dual Analytical Chemistry, Vol. 69, No. 3, February 1, 1997

399

Figure 7. Experimental design for the study of naproxen enantiomers with a β-CD-NH2/TM-β-CD dual system at pH 2.3. The CD concentrations are given in millimolar, and the values between parentheses correspond to the resolution: phosphoric acid/ammediol/ β-CD-NH2/TM-β-CD buffer; 0.5 mM sample prepared in 37.4 mM phosphoric acid/12 mM ammediol/10 mM β-CD-NH2/10 mM TM-βCD medium containing 2% methanol and 1% DMF.

system. Using the same approach as that developed with β-CDNH2/TM-β-CD dual system (see above), the intrinsic selectivity of DM-β-CD was estimated to 1.07 and 1.11 for Be and MeBe, respectively (considering that the analytes are nearly fully complexed in the presence of 20 mM β-CD-NH2 alone). This value indicates that DM-β-CD is more stereoselective than TM-β-CD for Be and MeBe enantiomers but this enantioselectivity remains less than the intrinsic selectivity of β-CD-NH2. This behavior suggests that, contrary to the case of APAs, hydrogen bonding is involved in the chiral recognition mechanism of Be and MeBe. This could be explained by the presence of the acceptor carbonyl group adjacent to the chiral center. This study clearly shows the interest of dual CD systems. They can permit not only the separation of neutral enantiomers but also the enantioselectivity evaluation of neutral CDs for neutral enantiomers. Using this approach, it is also possible to predict the experimental conditions for improving resolution. In effect, when one neutral CD is identified as a good potential chiral agent for the separation of some neutral enantiomers, then two separation strategies can be contemplated: the first one would be the use of a dual complexing agent system in which a charged complexing agent would lead to an affinity order inverse to the one obtained with the neutral CD; the second strategy would be to synthetize a CD that is substituted with an ionic group on at least one of its primary hydroxyl groups while having the same substituents at the wider rim of the cavity, as for the neutral stereoselective CD, this charged CD being likely to have a selectivity similar to the neutral one. Influence of the Concentration Ratio of the Two Complexing Agents. Naproxen enantiomers were studied according to the experimental design presented in Figure 7. The central point was chosen such as [TM-β-CD] ) [β-CD-NH2] ) 10 mM. To ascertain that there was no drift in the obtained results, the central point experiment was replicated every other experiment. The sample was prepared in a 37 mM phosphate/12 mM ammediol/ 10 mM β-CD-NH2/10 mM TM-β-CD medium, pH 2.3. Obtained resolutions are given in Figure 6. The intrinsic selectivity of β-CD-NH2 in the presence of 10 mM TM-β-CD, R1, turns out to be equal to the intrinsic selectivity of TM-β-CD in the presence of 10 mM β-CD-NH2, R2, (R1 ) R2 ) 1.42). This was expected since naproxen enantiomers are nearly fully complexed in the presence of 10 mM β-CD-NH2 or 10 mM 400 Analytical Chemistry, Vol. 69, No. 3, February 1, 1997

TM-β-CD and since β-CD-NH2 is not stereoselective with regard to these enantiomers (Rβ-CD-NH2 ) 1). The experiments realized with the same TM-β-CD/β-CD-NH2 concentration ratio, z, i.e., z ) 2 for 10:5 and 20:10 and z ) 0.5 for 5:10 and 10:20, lead to similar resolutions, i.e., 6.6 and 6.8 for z ) 2 and 5.4 and 5.6 for z ) 0.5. This is caused by the fact that, under these conditions, the analytes are completely complexed by the CDs and each enantiomer is characterized by its mobility µzc (see eq 23). As expected, the analysis times for the z ) 0.5 experiments are identical. An increase in the migration times was nevertheless noticed for the z ) 2 experiments when the CD concentration increases, likely due to the influence of TM-β-CD on electrolyte viscosity. As already stressed, the studied concentration zone is such that the enantiomers are fully complexed by the CDs and the resolution is constant for the experiments carried out at z constant. Thus, there is an optimal value for z maximizing the resolution in this zone, which corresponds to the conditions yielding the highest difference in enantiomer mobilities µzc (eq 23). The results observed show that the β-CD-NH2 (TM-β-CD, respectively) optimum concentration in the presence of TM-β-CD (β-CD-NH2, respectively) should be located between 5 and 10 mM (10 and 20 mM, respectively) (Figure 7). Therefore, the optimal value for z is between 1 and 2. In practice, it will be desirable to decrease the concentration of the two complexing agents to the threshold of full complexation while the concentration ratio is maintained constant and equal to the optimal z value, in order to decrease the chiral agent consumption and to minimize the baseline absorbance as well as the electrolyte viscosity. Conversely, the possibility of obtaining chiral separations at high concentrations with dual CD systems may be of great interest for analytes that are highly hydrophobic, the CDs being able to increase their solubility. CONCLUSION This study has extended the concept of intrinsic selectivity in capillary electrophoresis to analytical systems in which two complexing agents are involved. Using the concept of apparent equilibrium of complex formation and related apparent constant, the intrinsic selectivity of a complexing agent Y in the presence of a given concentration of a complexing agent X can be introduced. This selectivity parameter is independent of the concentration of Y and is representative of the effect of Y in these analytical conditions. When only independent complexation occurs (absence of mixed complex), then the intrinsic selectivity of a complexing agent Y in the presence of a given concentration of a complexing agent X is related to the intrinsic selectivity of each complexing agent and to complex formation constants. Dual systems of cyclodextrins were shown to constitute an interesting alternative strategy for the separation of neutral enantiomers. Arylpropionic acid enantiomers were baseline resolved in their protonated form with a β-CD-NH2/TM-β-CD dual system at pH 2.3. β-CD-NH2, which is not enantioselective in this case, confers a nonzero mobility to the analytes while TM-β-CD permits the chiral recognition of the enantiomers. As expected from previous experiments at higher pH with a neutral CD solely, a β-CD-NH2/DM-β-CD dual system was not as stereoselective as a β-CD-NH2/TM-β-CD one. Besides, benzoin and methyl ether benzoin enantiomers, which are partially resolved with β-CD-NH2 alone, were also partially resolved with a β-CD-NH2/DM-β-CD dual system and with a β-CD-NH2/TM-β-CD one.

Using the concept of intrinsic selectivity and assuming that no mixed complex occurred, it was possible to determine the intrinsic selectivity of neutral β-CD derivatives for the studied neutral enantiomers. For arylpropionic acids, the intrinsic selectivities of TM-β-CD and DM-β-CD in the presence of β-CD-NH2 are equal to the intrinsic selectivities of TM-β-CD and DM-β-CD, respectively, since β-CD-NH2 is not stereoselective. For benzoin and methyl ether benzoin enantiomers, TM-β-CD and DM-β-CD are less stereoselective than β-CD-NH2. Furthermore, these latter enantiomers have the same order of affinity for β-CD-NH2, TMβ-CD, and DM-β-CD. Thereby β-CD-NH2/TM-β-CD and β-CDNH2/DM-β-CD dual systems are less stereoselective than the most stereoselective CD (β-CD-NH2). This approach is of great interest to identify the potential of neutral chiral agents for neutral enantiomers and thereby, develop some new analytical strategies (for example, synthesis of a charged CD that otherwise retains the main characteristics of the most stereoselective neutral CD). In the concentration range where the enantiomers are fully complexed by the CDs, it was shown that the resolution is constant for the experiments carried out at constant CD concentration ratio. The existence of an optimum value for this ratio with respect to resolution was also evidenced. In practice it is therefore desirable to meet this condition while choosing the

lowest concentrations corresponding to the threshold for which full complexation of the analytes is reached. The use of higher CD concentrations may, however, remain beneficial when sample solubility in the bulk aqueous buffer is critical. ACKNOWLEDGMENT The authors acknowledge Rhoˆne-Poulenc Rorer (RPR), Centre de Recherche de Vitry-Alfortville, France, for funding this work and providing F.L.’s fellowship, Dr. A. Brun, RPR, and Professor J. Crommen, Institute of Pharmacy, Liege, Belgium, for their interest in this study, Dr. A. Jardy, Ecole Supe´rieure de Physique et de Chimie Industrielles de la Ville de Paris, Paris, France, for fruitful discussions, Dr. A. Jutand, Ecole Normale Supe´rieure, Paris, France, for the gift of racemic naproxen, and Dr. E. Kolossa and Dr. Angleys, Wacker-Chemie, for the gift of DM-β-CD. Dedicated to Professor B. Tre´millon on the occasion of his 65th birthday. Received for review June 19, 1996. Accepted November 7, 1996.X AC960607R X

Abstract published in Advance ACS Abstracts, December 15, 1996.

Analytical Chemistry, Vol. 69, No. 3, February 1, 1997

401