10
Catherine Foulon Cécile Danel Marie-Pierre Vaccher Jean-Paul Bonte Claude Vaccher Jean-François Goossens Laboratoire de Chimie Analytique, Faculté des Sciences Pharmaceutiques et Biologiques, Université de Lille, Lille, France Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition Baseline separation of ten new, substituted [1-(imidazo-1-yl)-1-phenylmethyl)] ben- zothiazolinone and benzoxazolinone derivatives with one chiral center was achieved using cyclodextrin-capillary zone electrophoresis (CD-CZE). A method for the enantio- meric resolution of these compounds was developed using neutral CDs (native a-, b-, g-CDs or a-, b-, g-hydroxypropyl (HP)-CDs) as chiral selectors. Operational parameters including the nature and concentration of the chiral selectors, pH, ionic strength, organic modifiers, temperature, and applied voltage were investigated. The use of neu- tral CDs provides enantiomeric resolution by inclusion of compounds in the CD cavity. The HP-a-CD and HP-b-CD were found to be the most effective complexing agents and allowed efficient enantiomeric resolutions. Optimal separation of N-imidazole deri- vatives was obtained using 50 mM phosphate buffer at pH 2.5 containing either HP-a- CD or HP-b-CD (7.5–12.5 mM) at 257C, with an applied field of 0.50 kV ? cm 21 giving resolution factors R s superior to 1.70 with migration times of the second enantiomer less than 13 min. The same enantiomer migration order observed for all molecules can be related to a close interaction mechanism with CDs. The influence of structural features of the solutes on R s and t m was studied. The lipophilic character (log k w ) of the solutes and the apparent and averaged association constants of inclusion complexes for four compounds with the six different CDs led us to rationalize the enantiosepara- tion mechanisms. The conclusions were corroborated with reversed-phase high-per- formance liquid chromatography (HPLC) on chiral stationary phases (CSPs) based on CDs. Keywords: Aromatase inhibitor / Chiral capillary electrophoresis / Hydrophobicity / Inclusion con- stant / Neutral cyclodextrin DOI 10.1002/elps.200405903 1 Introduction Since estrogens are the most important hormones [1, 2] involved in development of hormone-dependent breast tumors, inhibition of their biosynthesis is a potentially use- ful therapeutic option in hormone-sensitive breast cancer [2]. This can be accomplished by the use of antiestrogens that block estrogen receptors or by inhibiting P450 aro- matase, an enzyme complex that converts androgens to estrogens [2]. A number of steroïdal compounds like for- mestane [3] and nonsteroidal compounds like aminoglu- tethimide [4], letrozole, vorozole, anastrozole, and fadro- zole (Fig. 1) [4] have been developed as aromatase inhib- itors and are useful in second-line therapy of estrogen- dependent breast cancer in postmenopausal women [5]. But, because of the results obtained in vivo, the need for clinical drugs with increased specificity, clinical efficiency and tolerability remains a challenge in the development of new compounds [4, 6]. These considerations led us to investigate the aromatase inhibition properties of new 3-methyl-6-[1-(imidazo-1-yl)- 1-phenylmethyl)] benzothiazolinone (1, 2) and 3-methyl- 6-[1-(imidazo-1-yl)-1-phenylmethyl)] benzoxazolinone (3– 10) derivatives (Fig. 2). Many of them present higher in vitro activities than fadrozole. Pharmacological studies have shown that enantiomers of many drugs present dif- ferent activities or metabolism or toxicities (for example, (S)-fadrozole is the active enantiomer). Milligram amounts of the relevant compounds are generally required to investigate the pharmacodynamic properties of each Correspondence: Professor Claude Vaccher, Laboratoire de Chimie Analytique – Faculté des Sciences Pharmaceutiques et Biologiques, Université de Lille 2 – BP 83 – 3, rue du Pr. Laguesse, F-59006 Lille, France E-mail: [email protected] Fax: 133-3-2095-9009 Abbreviations: CSP , chiral stationary phase; DI, deionized; HP- -CD, hydroxypropyl-a-cyclodextrin; TEA, triethanolamine Electrophoresis 2004, 25, 2735–2744 2735 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim CE and CEC

Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition

Embed Size (px)

Citation preview

Page 1: Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition

Catherine FoulonCécile DanelMarie-Pierre VaccherJean-Paul BonteClaude VaccherJean-François Goossens

Laboratoire de ChimieAnalytique,Faculté des SciencesPharmaceutiques et Biologiques,Université de Lille,Lille, France

Chiral separation of N-imidazole derivatives,aromatase inhibitors, by cyclodextrin-capillary zoneelectrophoresis. Mechanism of enantioselectiverecognition

Baseline separation of ten new, substituted [1-(imidazo-1-yl)-1-phenylmethyl)] ben-zothiazolinone and benzoxazolinone derivatives with one chiral center was achievedusing cyclodextrin-capillary zone electrophoresis (CD-CZE). A method for the enantio-meric resolution of these compounds was developed using neutral CDs (native a-, b-,g-CDs or a-, b-, g-hydroxypropyl (HP)-CDs) as chiral selectors. Operational parametersincluding the nature and concentration of the chiral selectors, pH, ionic strength,organic modifiers, temperature, and applied voltage were investigated. The use of neu-tral CDs provides enantiomeric resolution by inclusion of compounds in the CD cavity.The HP-a-CD and HP-b-CD were found to be the most effective complexing agentsand allowed efficient enantiomeric resolutions. Optimal separation of N-imidazole deri-vatives was obtained using 50 mM phosphate buffer at pH 2.5 containing either HP-a-CD or HP-b-CD (7.5–12.5 mM) at 257C, with an applied field of 0.50 kV?cm21 givingresolution factors Rs superior to 1.70 with migration times of the second enantiomerless than 13 min. The same enantiomer migration order observed for all moleculescan be related to a close interaction mechanism with CDs. The influence of structuralfeatures of the solutes on Rs and tm was studied. The lipophilic character (log kw) of thesolutes and the apparent and averaged association constants of inclusion complexesfor four compounds with the six different CDs led us to rationalize the enantiosepara-tion mechanisms. The conclusions were corroborated with reversed-phase high-per-formance liquid chromatography (HPLC) on chiral stationary phases (CSPs) based onCDs.

Keywords: Aromatase inhibitor / Chiral capillary electrophoresis / Hydrophobicity / Inclusion con-stant / Neutral cyclodextrin DOI 10.1002/elps.200405903

1 Introduction

Since estrogens are the most important hormones [1, 2]involved in development of hormone-dependent breasttumors, inhibition of their biosynthesis is a potentially use-ful therapeutic option in hormone-sensitive breast cancer[2]. This can be accomplished by the use of antiestrogensthat block estrogen receptors or by inhibiting P450 aro-matase, an enzyme complex that converts androgens toestrogens [2]. A number of steroïdal compounds like for-mestane [3] and nonsteroidal compounds like aminoglu-

tethimide [4], letrozole, vorozole, anastrozole, and fadro-zole (Fig. 1) [4] have been developed as aromatase inhib-itors and are useful in second-line therapy of estrogen-dependent breast cancer in postmenopausal women [5].But, because of the results obtained in vivo, the need forclinical drugs with increased specificity, clinical efficiencyand tolerability remains a challenge in the development ofnew compounds [4, 6].

These considerations led us to investigate the aromataseinhibition properties of new 3-methyl-6-[1-(imidazo-1-yl)-1-phenylmethyl)] benzothiazolinone (1, 2) and 3-methyl-6-[1-(imidazo-1-yl)-1-phenylmethyl)] benzoxazolinone (3–10) derivatives (Fig. 2). Many of them present higher invitro activities than fadrozole. Pharmacological studieshave shown that enantiomers of many drugs present dif-ferent activities or metabolism or toxicities (for example,(S)-fadrozole is the active enantiomer). Milligram amountsof the relevant compounds are generally required toinvestigate the pharmacodynamic properties of each

Correspondence: Professor Claude Vaccher, Laboratoire deChimie Analytique – Faculté des Sciences Pharmaceutiqueset Biologiques, Université de Lille 2 – BP 83 – 3, rue du Pr.Laguesse, F-59006 Lille, FranceE-mail: [email protected]: 133-3-2095-9009

Abbreviations: CSP, chiral stationary phase; DI, deionized; HP-Æ-CD, hydroxypropyl-a-cyclodextrin; TEA, triethanolamine

Electrophoresis 2004, 25, 2735–2744 2735

2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

CE

and

CE

C

Page 2: Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition

2736 C. Foulon et al. Electrophoresis 2004, 25, 2735–2744

Figure 1. Nonsteroidal aroma-tase inhibitors.

Figure 2. Chemical structure ofsubstrates 1–10.

enantiomer. Chiral high-performance liquid chromatogra-phy (HPLC) has been recognized as a useful methodologyfor this purpose [7] since it furnishes directly both enantio-mers. Liquid chromatography using polysaccharides aschiral stationary phases (CSPs) for the enantiosepara-tions of imidazole derivatives is largely described in theliterature [8–15]. Our recent contribution in the chiralseparations of aromatase inhibitors included the use of anormal-phase methodology with two silica-based amy-lose or cellulose derivatives [16].

Capillary electrophoresis (CE) is recognized as one of themost efficient separation techniques owing to its high re-solving power. Additionally, chiral separations of racemiccompounds are promising topics in the field of CE. Thisseparation can be often achieved using cyclodextrins aschiral selector added to the electrolyte solution. Resolu-tions of chiral derivatives containing heterocycle as imida-zole or triazole analogs are reported in the literature withseveral CE method developments. Electrophoretic meth-ods have been performed using neutral CDs as pseudo-stationary phase. Since uncharged CDs migrate at thesame velocity as the electroosmotic flow (EOF), they onlyallow the separation of ionized analytes in a pH bufferbelow the pKa of azole derivatives. This capillary zone

electrophoresis mode was developed successfully forimidazole and triazole derivatives using a pH 3.0 phos-phate buffer with hydroxypropyl-b-CD (HP-b-CD) [17,18]. CD-modified micellar electrokinetic chromatography(CD-MEKC) is an alternative strategy used for a chiralseparation, performed at neutral pH buffer with achiralmicelles such as sodium dodecyl sulfate (SDS) and aneutral CD as chiral pseudostationary phase. Enantio-selective separation of tridimenol, a triazol derivative,has been achieved using HP-g-CD and SDS [19]. IonicCDs are interesting chiral selectors with electrophoreticproperty allowing high-resolution power for separation ofracemic neutral or ionic solutes without micelles in thebackground electrolyte (BGE). For the chiral separationof imidazole or triazole derivatives, electrokinetic chroma-tography using sulfated- or carboxymethyl-CD (CD-EKC)have been performed at acidic pH buffer. The enantiore-cognition was believed to result from the multiple interac-tions between both anionic CDs and the cationic analytewhich included inclusion effect, electrostatic interactionand hydrogen bonding [18–21].

In this paper, separation of enantiomers of compounds 1–10 is studied by CE using neutral CDs at acidic pH ; thesecationic compounds are slowing down by the CD. Thisstudy reports the optimization of enantiomeric separa-

2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Page 3: Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition

Electrophoresis 2004, 25, 2735–2744 Chiral separation of N-imidazole derivatives 2737

tion of the leading compound 3. The optimization ofvarious parameters (the nature and concentration ofthe chiral selectors, pH, ionic strength, organic modi-fiers, temperature, and applied voltage) has led to theanalysis of seven compounds derived from the leadingcompound 3 under optimal conditions and then tostudy the relations between analyte structure and elec-trophoretic behavior. The enantiomer migration orderfor the molecules investigated was determined. Appar-ent and averaged binding constants were calculatedfor the 24 complexes formed between the 6 neutralCDs and the 4 compounds 1, 2, 3, and 5 that offerthe highest pharmacological interest. The electropho-retic results were then compared to the chromato-graphic data obtained with Cyclobond columns. De-pendence between solute hydrophobicity and optimalCD concentration provided additional information re-garding to the enantiomer recognition mechanism.

2 Materials and methods

2.1 Apparatus

2.1.1 CE

CZE experiments were performed on a Beckman P/ACEMDQ Capillary Electrophoresis system (Beckman-CoulterFrance, Villepinte, France), including an on-columndiode-array UV detector. The whole system was drivenby a PC with the 32 Karat software (Beckman-CoulterFrance) package for system control, data collection andanalysis. It was equipped with a 50.2 cm (40.2 cm effec-tive length)650 mm ID untreated fused-silica capillary(Composite Metal Services, Worcestershere, UK). Thecapillary was mounted in a cartridge and thermostatedat 257C 6 0.17C. An hydrodynamic injection was madewith a 5 s injection time at 1.0 psi (anodic injection). Com-pounds were detected at 196 nm (or 220 nm for com-pounds 1 and 2). New capillaries were flushed for 20 minwith 0.1 M sodium hydroxide (NaOH) (P = 20 psi) and 5 minwith water (P = 20 psi). Each day it was flushed succes-sively with NaOH (5 min, 20 psi), water (1 min, 20 psi) andthen with BGE (3 min, 20 psi). Between each run, the cap-illary was treated with water (1 min, 20 psi) and BGE(3 min, 20 psi).

2.1.2 Chromatography (HPLC)

Determination of hydrophobicity was performed on a C18

Symetry column (15064.6 mm ID; 5 mm) (Waters, SaintQuentin-en-Yvelines, France) by studying isocratic ca-pacity factors log k, obtained for different organic modi-

fier/water ratios (range 40/60–60/40 v/v). Extrapola-tion to 100% water as a mobile phase gives log kW

values which are representative of the hydrophobicitiesof the solutes [22]. Log kW values were correlated to theRekker log P values [23]. Flow rate, temperature anddetection wavelength were, respectively, 1 mL?min21,307C and 205 nm. In addition, enantioseparation of com-pounds 1, 2, 3, and 5 was performed on Cyclobond I2000 (b-CD) and Cyclobond I 2000 RSP (HP-b-CD) col-umns (25064.6 mm ID; 5 mm) (Astec, Whippany, NJ,USA). A constant mobile-phase flow of 0.8 mL?min21

was provided by a gradient Waters 600E meteringpump model equipped with a 7125 Rheodyne injector(20 mL loop). Detection was achieved at 201 nm for 3,205 nm for 5 and 220 nm for 1 and 2 with a Waters 996photodiode-array spectrophotometer. Chromatographicdata were collected and processed on a computer run-ning with Millenium 2010 software. Mobile-phase elutionwas made isocratically using deionized (DI) water andorganic modifier (methanol or acetonitrile) at various per-centages. Chromatographic separations were performedat 257C.

2.2 Chemicals

a-CD, g-CD and HP-g-CD were purchased from WackerChimie (Lyon, France) and HP-a-CD from Cyclolab (Buda-pest, Hungary). b-CD and HP-b-CD were a gift of theRoquette Laboratories (Lestrem, France). The 2-hydroxy-propylated CDs, HP-a-CD, HP-b-CD and HP-g-CD,represent the multicomponent mixtures with molar sub-stitution (MS) 0.62–0.95, 0.75–0.95 and 0.5–0.8 per anhy-droglucose unit, respectively. HP-CDs concentrationswere calculated taking into account their averaged mo-lecular weight obtained with an MS of 0.79, 0.85 and0.65 for HP-a-CD, HP-b-CD and HP-g-CD, respectively.The molar concentration unit is used to permit the com-parison of the different selectors and the calculation ofapparent and averaged binding constants. NaOH wasobtained from Beckman. Phosphoric acid (85% w/w),triethanolamine (TEA), methanol, ethanol, isopropanol,and acetonitrile were purchased from Merck (Nogent-sur-Marne, France). DI water was obtained from a Milli-Qsystem (Millipore, Saint Quentin-en-Yvelines, France).Compounds 1–10 were prepared by some of us [16],leading to racemic mixtures of enantiomers. For CE anal-ysis, BGEs were prepared by solubilization of CDs in a50 mM phosphate buffer prepared from a phosphoricacid solution adjusted to pH 2.5 by addition of TEA. Sam-ple solutions at 100 mg?L21 in 2.5 mM phosphate buffer(pH 2.5) were obtained from ethanolic solutions at2 g?L21. For chromatographic analysis, 0.50 mM samplesolutions were prepared in methanol.

2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Page 4: Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition

2738 C. Foulon et al. Electrophoresis 2004, 25, 2735–2744

2.3 Calculations

The separation parameters in CE were calculated as fol-lows:

Apparent mobility:

mapp ¼ lLVt

(1)

Separation factor:

a ¼ t2

t1(2)

Resolution:

Rs ¼ 2ðt2 � t1Þðo1 þ o2Þ

(3)

where mapp is the apparent mobility, L and l are the totalcapillary length and the length to the detector, respec-tively, V is the run voltage, t is the enantiomer migrationtime, a and Rs are the selectivity and resolution, respec-tively, and o is the width of the peak at the baseline. Thesuffixes 1 and 2 refer to the first and the second detectedenantiomer.

3 Results and discussion

3.1 Optimization of chiral separations

The main parameters affecting enantiomeric separationby CD-CZE are the nature and concentration of the CDand the pH of the buffer. Secondary parameters are ionicstrength, presence of organic modifier, temperature, andvoltage. All enantioseparation optimizations were per-formed for compounds 1, 2, 3, and 5. Since similar resultswere obtained for all parameters and to clarify the presen-tation, only experimental data of analog 3 have been illus-trated.

3.1.1 Effect of the nature of the CD

According to pKa of N-methylimidazole (pKa = 7.2) [24],N-imidazole derivatives 1–10 are positively charged atthe working pH 2.5. Analyses were performed for CDconcentrations varying in the range of 1–20 mM. Migrationtimes, separation factors and resolutions obtained for thechiral separations of 3 with the six neutral CDs at 7.5 mM

are presented in Table 1. Migration times increase bychanging from a-CD to HP-a-CD, g-CD, HP-g-CD, b-CD,and then to HP-b-CD. Interactions between neutral CDsand solutes and CD-induced BGE viscosity cause a slow-ing down of the solute migration. The results suggest thatinteractions with a-CD, HP-a-CD, g-CD, and HP-g-CDare weaker compared to those established with b-CD

Table 1. Influence of the CD type on the enantiosepara-tion of compound 3

CD t1 (min) t2 (min) a Rs

a-CD 7.63 7.63 1.00 n.r.b-CD 10.28 10.44 1.02 0.78g-CD 8.14 8.14 1.00 n.r.

HP-a-CD 7.83 7.94 1.01 0.90HP-b-CD 11.05 11.38 1.03 1.45HP-g-CD 8.77 8.77 1.00 n.r.

Conditions : fused-silica capillary, 50.2 cm (40.2 cm effec-tive length)650 mm ID; BGE, 25 mM phosphate buffer,pH 2.5 (H3PO4 1 TEA), CD concentration, 7.5 mM; UVdetection at 196 nm; anodic injection, 1 psi pressure for5 s of 100 mg?L21 solution; temperature, 257C; appliedvoltage, 20 kV; n. r., no resolution

and HP-b-CD. A satisfactory chiral recognition wasachieved with b-CD but it was better with HP-a-CD andHP-b-CD. According to the literature [18, 25–27], thebest chiral recognition abilities of HP-CDs, compared tonative CDs, can be explained by two phenomena inducedby substitution: (i) an increase in the cavity size – thatfavors interactions, (ii) an improved chiral environment bypartial substitution of the glucose units. HP-b-CD seemsfinally to be the best-suited CD for the resolution of 3 andwas then chosen for the optimization of the separation.

3.1.2 Effect of the CD concentration

The CD concentration is an essential parameter for theoptimization of chiral separations [27–29]. The influenceof the HP-b-CD concentration, in the range of 2.5–15 mM,on the enantioseparation of 3 is shown in Fig. 3. Whereasmigration times always increase with the HP-b-CD con-centration, the resolution reaches a maximum value (Rs =1.45) for a concentration of 7.5 mM. An HP-b-CD concen-tration of 7.5 mM was selected for the final method andwas found to provide a good resolution and a relativelyshort run time with a lower CD consumption.

3.1.3 Effect of the pH buffer

The pH of the buffer is another important parameteraffecting chiral resolution. Effective charge and, conse-quently mobility of the analyte directly depends on thepH. Additionally, EOF is pH-dependent. Experimentswere performed using BGE at different pH values in therange of 2.5–5.5 obtained by addition of increased TEAamounts to H3PO4 solution. Ionic strength was kept con-stant (0.15 mM), by the use of various amounts of sodiumchloride in the BGE, to avoid its influence on the separa-

2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Page 5: Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition

Electrophoresis 2004, 25, 2735–2744 Chiral separation of N-imidazole derivatives 2739

Figure 3. Effect of the HP-b-CD concentration on (a)migration times (t1 and t2) and (b) resolution (Rs) of theenantiomers of compound 3. Conditions as in Table 1.(r), Migration time of the first eluted enantiomer;(j) migration time of the latest eluted enantiomer.

tion. The increase in pH values leads to a general de-crease in migration times and resolution. Separation pa-rameters t1, a, Rs of 3 are 17.92, 1.04, 1.88 and 11.31,1.00, n.r. (n.r., no resolution) for pH 2.5 and 5.5, respec-tively. This decrease of migration time is mainly due tothe increase of the EOF and unfavors chiral resolution.pH 2.5 appears finally to be the best-suited pH for theenantioseparation of this compound and was thenselected for the optimization of the separation.

3.1.4 Effect of the ionic strength

The influence of the ionic strength on the enantiosepara-tion was studied in the range of 0.02–0.07 mM, by use ofpH 2.5 phosphate buffer at different concentrations (25–100 mM). The use of high ionic strength buffer is limited byexcessive heat generation. An increase in the phosphateconcentration from 25 to 100 mM leads to an increase inmigration time almost equal to 70%. Separation parame-ters t1, a, Rs of 3 are for 25 and 100 mM phosphate con-centrations 10.04, 1.03, 1.46 and 17.94, 1.04, 1.21,respectively. These results are in accordance with Issaqet al. [30], concerning the proportional relation betweenthe electrophoretic mobility and the inverse of the squareroot of buffer concentration. The resolution increases for

buffer concentration up to 50 mM and then decreases.The increase of the resolution can be explained by thelarge improvement in peaks efficiency; the decrease ofresolution could be associated to the dispersion increasedue to the excessive migration times [31]. Taking intoaccount these results, a 50 mM phosphate concentrationwas selected for the optimal enantioseparation method.

3.1.5 Effect of the organic modifier

The addition of organic modifier in the BGE can play animportant role to obtain a complete separation of theenantiomers. In fact, the organic modifier may influenceseveral parameters, e.g., the inclusion complexation(competition between solutes and organic modifiers), thesolubility of enantiomers, the viscosity of the BGE, andthe EOF (variation of the dielectric constant and the zeta-potentiel of the capillary wall) [30]. Three organic modi-fiers, methanol, acetonitrile and isopropanol were thenadded to the BGE within a 0–20% range. Results are illus-trated in Fig. 4. In order to evaluate the effect of theorganic modifier on the BGE properties, migration timesof compound 3, without CD, were determined. New BGE

Figure 4. Effect of the presence of organic modifier on(a) migration times and (b) resolution (Rs) of the enantio-mers of compound 3. Conditions: fused-silica capillary,50.2 cm (40.2 cm effective length)650 mm ID; BGE,50 mM phosphate buffer, pH 2.5 (H3PO4 1 TEA), 7.5 mM

HP-b-CD; UV detection at 196 nm; anodic injection, 1 psipressure for 5 s of 100 mg?L21 solution ; temperature,257C; applied voltage, 20 kV. Migration time of the firsteluted enantiomer of compound 3 without (h) or with(n) CD in the BGE.

2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Page 6: Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition

2740 C. Foulon et al. Electrophoresis 2004, 25, 2735–2744

properties, when methanol or isopropanol are added tothe BGE, are clearly associated with an increase of migra-tion time of 3 in the absence of CD (Fig. 4). However, in thepresence of acetonitrile, the migration time remains con-stant. This polar organic solvent does not seem to modifythe BGE viscosity.

When CDs are used (7.5 mM, HP-b-CD), which contributesto increase the BGE viscosity, the migration times are notaffected by addition of methanol or isopropanol. Organicmodifier-induced BGE properties changes are then coun-terbalanced by two probable effects: a decrease of theanalyte inclusion in the CD cavities and an increase of thehydrophobic analytes solvatation in the organic modifier.These phenomena may account for the decreased resolu-tion observed when 20% of isopropanol is added to theBGE. Since isopropanol is highly hydrophobic and sol-vates the analytes, a competition between solutes and sol-vent occurs for the CD cavities, which leads to poor enan-tioseparation. When acetonitrile is used, the decrease ofthe migration times and resolution could also be explainedfrom the decrease of the 3-CD interactions. Finally, BGEwithout additive solvent was chosen.

3.1.6 Effect of the temperature

Compound migration times were determined between207C and 407C, by 57C steps. The temperature increaseleads to a reduction of both migration times and resolu-tion. Separation parameters t1, a, Rs of 3 are 15.66, 1.04,1.80 and 8.88, 1.02, 1.34 at 207C and 407C, respectively.These data show the influence of the temperatureincrease on the stability of the solute-CD complexes.Moreover, this temperature increase contributes to a de-crease of the buffer viscosity, which could enhance thediffusional band broadening and thus decrease the peakefficiency and resolution. The choice of temperature is acompromise between a lower migration time and a higherresolution. A temperature of 257C was finally chosen.

3.1.7 Effect of the voltage

A decrease of migration times was observed in therange of 10–30 kV whereas resolution remains relativelyconstant. Separation parameters t1, a, Rs of 3 are,respectively, 25.35, 1.03, 1.66 and 8.04, 1.03, 1.69 at10 and 30 kV. Since apparent mobility is almost con-stant in these experiments and the electrophoretic mo-bility of the analyte is independant of the applied volt-age, the electroosmotic mobility is invariant under ourexperimental conditions, i.e., at pH 2.5. Since Ohm’slaw is verified (linear relationship current – voltage : r2 =0.998), no excessive Joule heating was generated dur-

ing the electrophoresis process. To reduce the analysistime, a 25 kV voltage (field of 0.5 kV?cm21) was chosenfor the final method.

3.1.8 Results of enantioseparation optimizationfor compounds 1, 2, and 5

Optimization of the enantiomeric separation of 1, 2, and 5was investigated following the same strategy. The nature ofthe CD and their concentrations were the parameters toinvestigate. Optimal pH, ionic strength, organic modifieraddition, temperature, and voltage were the same for thesefourcompounds (1, 2, 5, and 3). Electropherogramsobtainedunder the optimized conditions are presented in Fig. 5.

Figure 5. Electropherograms of compounds 1, 2, 3, and5 under optimal conditions. Conditions, as in Fig. 4,except BGE, 50 mM phosphate buffer, pH 2.5 (H3PO4 1

TEA); type and concentration of the CD, 15 mM HP-a-CD, for 1; 12.5 mM HP-a-CD, for 2; 7.5 mM HP-b-CD, for3; 7.5 mM HP-a-CD, for 5; UV detection at 196 nm (for3 and 5) or 220 nm (for 1 and 2).

3.2 Mechanism of enantioselective recognition

3.2.1 Solute structure – enantioseparationrelationships

The effect of the solute structure on the enantioseparationwas investigated by studying the separation of sevencompounds derived from compound 3 by modifications

2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Page 7: Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition

Electrophoresis 2004, 25, 2735–2744 Chiral separation of N-imidazole derivatives 2741

of the phenyl ring (compounds 5, 6, 7, and 10) or modifi-cations of the benzoxazolinone moiety (compounds 4, 8and 9). This study was run under standard conditions cor-responding to the optimal enantioseparation of 3 (7.5 mM,HP-b-CD). Results are summarized in Table 2 and Fig. 6.

Table 2. Influence of the analyte structure on the enantio-separation

Compound log kW log P Dt (min) t1 (min) a Rs

3 3.33 3.86 6.00 13.67 1.05 1.974 3.11 3.33 9.79 17.67 1.03 1.005 2.70 3.46 3.45 12.12 1.00 n.r.6 2.64 3.46 2.11 10.41 1.02 0.927 3.99 4.59 9.64 18.35 1.04 1.648 3.62 4.29 8.87 17.35 1.03 1.299 3.41 3.86 3.40 11.62 1.01 0.90

10 3.11 2.50 0.27 4.81 1.00 n.r.

Conditions as in Table 1, except BGE, 50 mM phosphatebuffer, pH 2.5 (H3PO4 1 TEA), 7.5 mM HP-b-CD; Dt =t1(with CD)2 t(without CD) .

Furthermore, the hydrophobicity of all N-imidazole deri-vatives was evaluated by the chromatographic methoddescribed in the Section 2.1. The log kW values were wellcorrelated with the log P values, determined in accord-ance with the theoretical method of Rekker [23]. In orderto compare the affinity of the different compounds for theHP-b-CD, we have chosen to consider Dt (= t1(with CD) –t(without CD)) obtained for all compounds: the difference be-tween the solute migration times in the presence and inthe absence of CD derives both from their interactionwith the CD and the increased viscosity of the BGE.Since, under identical operational conditions, the CD-induced increase of BGE viscosity is the same for allsolutes, the comparison of the Dt obtained for differentsolutes roughly reflects the difference of the solute affinityvs. the HP-b -CD.

Concerning the modifications of the phenyl ring (Fig. 6a),the influence of cyano substitution of the phenyl ring wasstudied with compounds 5 (para-substitution) and 6(meta-substitution). For both compounds, a decrease inDt values – that may correspond to a decrease of thecomplexation with the HP-b-CD – is observed (Dt are6.00, 3.45 and 2.11 for 3, 5 and 6, respectively), whichmay be corroborated with their lower hydrophobicity (logkW are 3.33, 2.70 and 2.64 for compounds 3, 5 and 6,respectively). Furthermore, the symmetry loss of the phe-nyl ring, when substitution occurs in meta (6) rather than inpara (5), leads to an increase in resolution, i.e., in enantio-recognition process.

Figure 6. Electropherograms of compounds 3, 4, 5, 6, 7,8, 9, and 10: effect of the analyte structure by modifica-tions of (a) the benzoxazolinone moiety; (b) the phenylring. Conditions as in Fig. 4.

The para-chloro substitution (7) leads to an increase of Dt(Dt are 6.00 and 9.64 for 3 and 7, respectively), while res-olution remains quite constant. The greater complexationof 7 in comparison with 3 for the HP-b-CD may be relatedto its higher hydrophobicity (log kW of 3 and 7 are 3.33 and3.99, respectively). Lower Dt is observed when the phenylring is replaced by a pyridinium (Dt is 6.00 and 0.27 for

2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Page 8: Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition

2742 C. Foulon et al. Electrophoresis 2004, 25, 2735–2744

3 and 10, respectively). This small change of migrationtime in the presence of CD likely reflects the effect of theincreased viscosity rather than an interaction with the CD.The marked decrease of Dt (6.00–0.27) may correspondto a lack of affinity of 10 for HP-b-CD. We confirmed herethat the inclusion of a pyridinium ring in the hydrophobiccavity of the CD is not favorable. These structure-enantio-separation relationships confirm that inclusion into thehydrophobic cavity occurred through the phenyl ringwhich is a prerequisite and that this phenomenon (asso-ciated to other interactions with the CD rims) is essentialfor enantioselectivity.

Concerning the modifications of the benzoxazolinonemoiety (Fig. 6b) it is noteworthy that compound 4, whichonly differs from compound 3 by its R2 substituent, hasprobably a greater affinity for the HP-b-CD (Dt are 6.00and 9.79 for 3 and 4, respectively); indeed its ability tocreate hydrogen bondings with the hydrophilic side ofthe HP-b-CD seems to increase solute-CD interactionswhereas the resolution decreases. When the benzoxazo-linone is replaced with a benzoxazinone (compounds 3vs. 8), a greater affinity for the HP-b-CD is observed (Dtare 6.00 and 8.87 for 3 and 8, respectively) whereas theresolution decreases. Both examples cited above provethat strong binding does not a priori mean a better enan-tiorecognition. The enantiomer migration orders weredetermined for compounds 1, 2, 3, and 5 (Table 3) withCDs which permit enantioseparation (i.e., b-CD, HP-a-CDand HP-b-CD). The same migration order was observedfor all compounds (dextrogyre (1)-enantiomers have

Table 3. Apparent and averaged binding constants of thecomplexes formed between compounds 1, 2, 3,5, and the six neutral CDs

Compound a-CD b-CD g-CD

K1 (M21) K2 (M21) K1 (M21) K2 (M21) K1 (M21) K2 (M21)

1 49 n.r. 281 (2) 322 (1) 25 n.r.2 90 n.r. 133 (2) 146 (1) 16 n.r.3 48 n.r. 263 (2) 284 (1) 28 n.r.5 114 n.r. 181 (2) 209 (1) 16 n.r.

Compound HP-a-CD HP-b-CD HP-g-CD

K1 (M21) K2 (M21) K1 (M21) K2 (M21) K1 (M21) K2 (M21)

1 18 (2) 27 (1) 143 (2) 155 (1) 96 n.r.2 172 (2) 208 (1) 42 n.r. 93 n.r.3 30 (2) 38 (1) 140 (2) 161 (1) 92 n.r.5 104 (2) 127 (1) 38 n.r. 72 n.r.

Migration order of the enantiomers (1) and (2) are speci-fied. Conditions as in Table 1, except; BGE, 25 mM phos-phate buffer, pH 2.5 (H3PO4 1 TEA)

greater affinities for CDs than the corresponding levogyre(2)-analogues whatever the nature of the solutes andCDs). This can be related to a close interaction mechan-ism for the various solutes.

3.2.2 Determination of apparent and averagedbinding constants

The apparent binding constants are not determined forthermodynamic meaning. However, these calculationspermit to appreciate the extent of analyte inclusion andto compare the different analytes. Apparent and averagedbinding constants were calculated for the 24 complexesformed between the 6 neutral CDs and the 4 compounds1, 2, 3, and 5 having the highest pharmacological interest.A theoretical model relating the mobility to the concentra-tion of the CD selector was developed by Wren and Rowe[28, 29]:

½CD�mi � mf

¼ 1mc � mf

½CD� þ 1ðmc � mfÞK

(4)

where K is the apparent binding constant, [CD] is consid-ered to be the total concentration since the complexedCD concentration must be insignificant, mi is the experi-mental electrophoretic mobility observed of either enan-tiomer, mf and mc are the electrophoretic mobilities of theenantiomers in the free and complexed forms, respec-tively. In our conditions, i.e., at pH 2.5, as previouslydescribed [33, 34], it is possible to consider that the EOFis negligible. However, changes of the viscosity of theBGE when concentration of CD increases affect mobilityof solutes, and this may affect the results. Values of cur-rent in the absence and presence of CD (i0 and i) wereused to rectify the viscosity effect: experimental electro-phoretic mobilities (mi) were corrected by the ratio i0/i [28].

Table 3 lists the apparent and averaged binding constantsK determined by studying the variations of [CD]/(mi-mf) vs.[CD] at 2987K. Electrophoretic mobility of free solutes mf isquite constant for these four compounds, it varies from2.4761024 to 2.5961024 cm2?V21?s21. In accordancewith Eq. (4), linear plots are obtained for all complexes.The r2 values for the linear fit are greater than 0.998. It isclassically admitted that basic enantioseparation mech-anism of compounds presenting an aromatic group withCD is referred to an inclusion complexation. However,complexation also occurs through hydrogen bondings,dipole-dipole and van der Waals interactions with thehydrophilic surface of the CD. The comparison of the af-finity of the whole compounds could permit to envisagewhich is the predominant interaction involved with the dif-ferent CDs. These binding constants show that the affi-nities of the solutes for the CDs are strongly dependent

2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Page 9: Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition

Electrophoresis 2004, 25, 2735–2744 Chiral separation of N-imidazole derivatives 2743

on the nature of the phenyl substituents. Two groups ofcompounds are clearly defined: group I includes analogswithout substitution of the phenyl ring (1 and 3); group IIincludes analogs with para-cyano substitution of the phe-nyl ring (2 and 5).

Concerning the a-CD and HP-a-CD, binding constants(K1) with compounds 1, 2, 3, and 5 are 49, 90, 48, and114) M21 and 18, 172, 30, and 104 M21, respectively. Thegreatest binding constants are observed for the mosthydrophilic compounds (group II, log kW of 2 and 5 are3.16 and 2.70, respectively). These results suggest thatthe hydrophilic interactions should play a major role inthe interaction phenomenon between these four solutesand the a-CD or HP-a-CD. Binding constants (K1) ofsolutes 1, 2, 3, and 5 with b-CD and HP-b-CD are281,133, 263, and 181 M21 and 143, 42, 140, and 38 M21,respectively. The greatest binding constants are observedfor the most hydrophobic compounds (group I, log kW of1 and 3 are 3.78 and 3.33, respectively). Moreover, enan-tioseparation occurs with both CDs. These results sug-gest that the fit of the solutes to the b-CD and HP-b-CDcavities may be the leading interaction phenomenon withthese CDs.

Chromatographic experiments with Cyclobond columns(b-CD and HP-b-CD chiral selectors) were achieved inorder to rank the affinity of various solutes for these CDsand then to compare the results to those obtained withthe electrophoretic method. The solutes retention times,when a 80/20%v/v water/methanol mobile phase and ab-CD CSP are used, are 9.21, 10.62, 16.74, and 27.43min for compounds 2, 5, 1, and 3, respectively. Identicalsolute retention orders are obtained when HP-b-CD CSPis used. These results agree with those obtained by theelectrophoretic method: lower affinity of group II com-pounds (2 and 5) compared to group I compounds (1 and3). Here again, this can be explained by their lower hydro-phobicity. Concerning the g-CD and HP-g-CD, bindingconstants with compounds 1, 2, 3, and 5 are relativelyconstant for the various solutes; they are 25, 16, 28, and16 M21 and 96, 93, 92, and 72 M21, respectively. Moreover,no enantioseparation is observed. These results were at-tributed to the absence of a strong inclusion of the solutesas the cavity size of these CDs must be too large, unfavor-ing the inclusion maintenance of solutes in the CDs andtheir enantioresolution.

Furthermore, according to the model developed by Wrenand Rowe [28, 29], the binding constants permit to calcu-late the optimal concentration of CD, [CD]opt, i.e., concen-tration that gives a maximum electrophoretic mobility dif-ference (Dm) between enantiomers (Eq. 5).

½CD�opt ¼1ffiffiffiffiffiffiffiffiffiffiffiffi

K1K2p (5)

In a general manner, the calculated optimal concentra-tions are of the same order as those determined experi-mentally. When the predominant interaction phenomenonbetween the solute and the CD is the fit of the solutehydrophobic moiety in the CD cavity, according to Wrenand Rowe [29], optimal CD concentration is reduced forthe most hydrophobic compounds. Hydrophobicity andoptimal concentration of b-CD, HP-a-CD and HP-b-CDcalculated for each solute are summarized in Table 4.

Table 4. log kW and calculated optimal concentration inb-CD, HP-a-CD or HP-b-CD for compounds 1,2, 3, and 5

Compound log kW [b-CD]opt

(mM)[HP-a-CD]opt

(mM)[HP-b-CD]opt

[mM)

1 3.78 3 45 72 3.16 7 5 n.r.3 3.33 4 30 75 2.70 5 9 n.r.

log kW determined as specified in Section 2.1

When b-CD is used, for group I, log kW and optimal CDconcentrations are 3.78 and 3 mM for 1 and 3.33 and4 mM for 3, respectively; for group II, log kW and optimalCD concentrations are 3.16 and 7mM for 2 and (2.70 and5 mM for 5, respectively. Then, the decrease of solute hy-drophobicity when a cyano substitution occurs is relatedto greater optimal CD concentrations. These results con-firm the previous hypothesis: the interaction of soluteswith the b-CD should be dominated by the inclusion ofthe phenyl ring, substituted or not by a para-cyano group,in the hydrophobic cavity.

When HP-a-CD is used, an opposite behavior occurs:molecules in group II, corresponding to the less hydro-phobic compounds, require lower optimal CD concentra-tions than the most hydrophobic compounds (45, 30, 5,and 9 mM for compounds 1, 3, 2, and 5, respectively).With HP-a-CD, interactions with the hydrophilic surfaceof this CD seem to play a major role in the solute-CD inter-actions phenomenon. This relation can not be studiedwhen HP-b-CD is used due to no resolution observedwith group II compounds.

4 Concluding remarks

In conclusion, CE with neutral CDs can be used for theseparation of N-imidazole derivatives. The optimizationof various parameters leads to the analysis under optimalconditions of seven compounds derived from the leading

2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Page 10: Chiral separation of N-imidazole derivatives, aromatase inhibitors, by cyclodextrin-capillary zone electrophoresis. Mechanism of enantioselective recognition

2744 C. Foulon et al. Electrophoresis 2004, 25, 2735–2744

compound and then to study relations between analytestructure, hydrophobic character and electrophoretic be-havior. Structure-enantioseparation relationships haveindicated that enantioseparation mechanism with HP-b-CD is principally the inclusion of the phenyl ring in thehydrophobic cavity of the CD. Moreover, this inclusionphenomenon seems a prerequisite for enantioselectivity.It was established that greater CD affinity of solutes pos-sessing a higher hydrophobic character for the CDs is notcorrelated to a higher enantioresolution. On the basis ofthe binding constants, we have proposed an enantiosep-aration mechanism which takes principally into accountthe predominance of hydrophilic interactions or inclusionphenomenon when N-imidazole derivatives interact withCDs.

The authors are grateful to Roquette Laboratories for thegift of b-CD and HP-b-CD.

Received January 19, 2004

5 References

[1] Segaloff, A., in: McGuire, W. L. (Ed.), Hormones and Mam-mary Carcinogens. Advances in Research and Treatment,Experimental Biology Plenum, New York 1978.

[2] Persiani, S., Broutin, F., Cicioni, P., Stefanini, P., Strolin Ben-edetti, M., Eur. J. Pharm. Sci. 1996, 4, 331–340.

[3] Enguehard, C., Renou, J. L., Allouchi, H., Leger, J. M., Gueif-fier, A., Chem. Pharm. Bull. 2000, 48, 935–940.

[4] Auvray, P., Moslemi, S., Sourdaine, P., Galopin, S., Seralini,G. E., Enguehard, C., Dallemagne, P., Bureau, R., Sonnet, P.,Rault, S., Eur. J. Med. Chem. 1998, 33, 451–462.

[5] Njar, V. C. O., Brodie, A. M. H., Drugs 1999, 58, 233–255.[6] Feutrie, M. L., Bonneterre, J., Bull. Cancer 1999, 86, 821–

827.[7] Francotte, E., Junker-Buchheit, A., J. Chromatogr. 1992,

576, 1–45.[8] Shibukawa, A., Nakagawa, T., in: Krstulovic, A. M. (Ed.),

Chiral Separations by HPLC, Ellis Harwood, Chichester1989, UK.

[9] Okamoto, Y., Kaida, Y., J. Chromatogr. A 1994, 666, 403–419.

[10] Liu, G., Goodall, D. M., Hunter, A. T., Massey, P. R., Chirality1994, 6, 290–294.

[11] Daicel Chemical Industries, Instruction Manual for ChiralpakAS and AD Baker Franz.

[12] Kunath, A., Theil, F., Wagner, J., J. Chromatogr. A 1994, 684,162–167.

[13] Okamoto, Y., Yashima, E., Angew. Chem. 1998, 37, 1020–1043.

[14] Dingenen, J., in: Subramanian, G. (Ed.), A PracticalApproach to Chiral Separations by Liquid Chromatography,New York 1994, pp. 115–181.

[15] Francotte, E., J. Chromatogr. A 2001, 906, 379–397.[16] Danel, C., Foulon, C., Park, C., Yous, S., Bonte, J. P., Vac-

cher, C., Chromatographia 2004, 59, 181–188.[17] Van Eeckhaut, A., Boonkerd, S., Detaevernier, M. R.,

Michotte, Y., J. Chromatogr. A 2000, 903, 245–254.[18] Chankvetadze, B., Endresz, G., Blaschke, G., J. Chroma-

togr. A 1995, 700, 43–49.[19] Otsuka, K., Matsumura, M., Kim, J. B., Terabe, S., J. Pharm.

Biomed. Anal. 2003, 30, 1861–1868.[20] Wu, Y. S., Lee, H. K., Li, S. F. Y., Electrophoresis 2000, 21,

1611–1619.[21] Wu, Y. S., Lee, H. K., Li, S. F. Y., J. Chromatogr. A 2001, 912,

171–179.[22] Dun, W. J., Block, J. H., Pearlman, R. S., Partition Coefficient

Determination and Estimation, Pergamon Press, Elmsford,NY 1986.

[23] Rekker, R. F., The Hydrophobic Fragmental Constant Phar-macochemistry Library, Vol. 1,Elsevier, Amsterdam 1977.

[24] Kapinos, L. E., Song, B., Sigel, H., Inorg. Chim. Acta 1998,280, 50–56.

[25] Heuermann, M., Blaschke, G., J. Chromatogr. 1993, 648,267–274.

[26] Quang, C., Khaledi, M., J. High Resolut. Chromatogr. 1994,17, 99–107.

[27] Palmarsdottir, S., Edholm, L.-E., J. Chromatogr. A 1994,666, 337–350.

[28] Wren, S. A. C., Rowe, R. C., J. Chromatogr. 1992, 603, 235–241.

[29] Wren, S. A. C., Rowe, R. C., J. Chromatogr. 1993, 635, 113–118.

[30] Issaq, H. J., Atamna, I. Z., Muschik, G. M., Janini, G. M.,Chromatographia 1991, 32, 155–161.

[31] Chankvetadze, B., Capillary Electrophoresis in Chiral Analy-sis, John Wiley & Sons, Chichester 1997, pp. 5–72.

[32] Ferrera, G., Santagati, N. A., Aturki, Z., Fanali, S., Electro-phoresis 1999, 20, 2432–2437.

[33] Williams, B. A., Vigh, G., J. Chromatogr. A 1997, 777, 295–309.

[34] De Boer, T., De Zeeuw, R. A., De Jong, G. J., Ensing, K.,Electrophoresis 2000, 21, 3220–3229.

2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim