The relative partitioning of neutral and ionised compounds in sodium dodecyl sulfate micelles measured by micellar electrokinetic capillary chromatography

The relative partitioning of neutral and ionised compounds in sodium dodecyl sulfate micelles measured by micellar electrokinetic capillary chromatography

European Journal of Pharmaceutical Sciences 15 (2002) 225–234 www.elsevier.nl / locate / ejps The relative partitioning of neutral and ionised compou...

159KB Sizes 1 Downloads 24 Views

European Journal of Pharmaceutical Sciences 15 (2002) 225–234 www.elsevier.nl / locate / ejps

The relative partitioning of neutral and ionised compounds in sodium dodecyl sulfate micelles measured by micellar electrokinetic capillary chromatography Agnes Taillardat-Bertschinger, Pierre-Alain Carrupt, Bernard Testa* ´ Institut de Chimie Therapeutique , Section de Pharmacie, Universite´ de Lausanne, CH-1015 Lausanne, Switzerland Received 4 September 2001; received in revised form 21 December 2001; accepted 27 December 2001

Abstract The rational use of micelles in quantitative structure–activity and quantitative structure–permeation relationships implies a good knowledge of the nature of recognition forces underlying solute–micelle association. The aims of this study were to unravel the intermolecular interaction forces responsible for the association of neutral and ionised compounds with negatively charged sodium dodecyl sulfate (SDS) micelles, using micellar electrokinetic capillary chromatography (MEKC). The MEKC capacity factors (log k MEKC ) of 36 neutral model solutes were analysed by linear solvation free-energy relationships (LSERs). The results indicate that the size and H-bond acceptor strength of solutes are mainly responsible for their MEKC retention. Compared to n-octanol, the SDS micelles are more cohesive and stronger H-bond donors. Strong attractive electrostatic interactions govern solute–micelle association for positively charged compounds and micelles of the opposite charge, whereas repulsive electrostatic interactions occur between negatively charged solutes and micelles of the same charge. The capacity factors measured for the ionised forms of the acids and bases under study (log k IMEKC ) indeed lie on two distinct plateau, about 21.0 for the former and about 2.0 for the latter and depend on the solute’s charge –I more than on its chemical structure. Thus, the derivation of a diff(log k NMEKC ) value, defined as the difference between the log k MEKC values of the neutral and charged species, strongly correlates with the respective log k NMEKC value and does not afford additional information.  2002 Elsevier Science B.V. All rights reserved. Keywords: Micellar electrokinetic capillary chromatography (MEKC); Retention mechanisms; Solute–micelle interaction; Lipophilicity; Acids; Bases

1. Introduction The biological activity of drugs depends on their interaction with biomembranes both in a pharmacodynamic and in a pharmacokinetic context (Testa et al., 2000). The attraction between a drug and a lipid membrane is mainly attributed to the compound’s lipophilic features, which are traditionally described by its n-octanol–water partition coefficient (log Poct ). However, a number of recognition forces are not encoded in log P, namely ionic bonds and charge transfer interactions (Testa et al., 1996) which are of particular importance in the interactions between membranes and ionised compounds. A new approach to measure partition coefficients involves artificial membranes, i.e., liposomes (Betageri and Rogers, 1988; Pauletti and Wunderli-Allenspach, 1994), *Corresponding author. Tel.: 141-216-924-521; fax: 141-216-924525. E-mail address: [email protected] (B. Testa).

immobilised artificial membranes (IAMs) (Ong et al., 1995) and micelles (Tanaka et al., 1994; Treiner and Chattopadhyay, 1986). Micelles are aggregates of amphiphilic molecules and provide anisotropic microenvironments with hydrophobic and polar sites of interaction. Before the introduction of micellar electrokinetic capillary chromatography (MEKC), the monitoring of solute partitioning in micelles was tedious and time-consuming. In addition to offering comparable conditions to HPLC (i.e., no sample purity requirement, small sample size and short analysis times), MEKC allows the adjustment of the composition of the micellar pseudo-stationary phase by a simple change of the surfactant type to optimally model the intermolecular interactions found in biological systems. Anionic sodium dodecyl sulfate (SDS) micelles are most frequently used in MEKC and are of particular interest since about 10% of the inner lipid leaflets (of different membrane types, among them plasma membranes) are negatively charged (Yeagle, 1992). For the rational use of micelles in quantitative structure–

0928-0987 / 02 / $ – see front matter  2002 Elsevier Science B.V. All rights reserved. PII: S0928-0987( 02 )00004-0

226

A. Taillardat-Bertschinger et al. / European Journal of Pharmaceutical Sciences 15 (2002) 225 – 234

activity and quantitative structure–permeation relationships, it is indispensable to know the nature of the interactions underlying MEKC retention for both neutral and ionised compounds. Linear solvation free-energy relationships (LSERs) (Kamlet and Taft, 1976; Taft and Kamlet, 1976) based on the solvatochromic parameters (dipolarity / polarisability p *, hydrogen-bond donor acidity a and hydrogen-bond acceptor basicity b ) and the calculated molecular volume (Vw ) (Eq. (1)) are a powerful tool to unravel the structural properties (Sp ) governing the drug–micelle interactions of neutral compounds accounting for their MEKC retention. In Eq. (1), the symbols v, p, a and b represent the regression coefficients whereas c is the intercept. Sp 5 vVw 1 pp * 1 aa 1 bb 1 c

(1)

LSERs studies available to date indicate that the MEKC capacity factors using SDS micelles are primarily dependent on molecular size and hydrogen-bond acceptor basicity, whereas dipolarity / polarisability and hydrogen-bond donor acidity play minor roles (Vitha and Carr, 1998; Yang and Khaledi, 1995). However, a quantitative comparison with partitioning in n-octanol is difficult, as the series of solutes used to generate the respective LSER equations are not comparable and often include a limited number of compounds. The resulting statistical artefacts may mask or exaggerate some structural information. For this reason, Pagliara et al. (1995) defined a well-distributed set of 80 structurally diverse model compounds with known solvatochromic parameters allowing a regular and broad exploration of the property space defined by log P, V, a, b and p *. In the present study, the MEKC retention mechanisms for neutral compounds using SDS micelles were compared to partitioning in n-octanol using this well-distributed set of model solutes and the LSER approach. Recently, the migration behaviour of ionised solutes using anionic SDS micelles was also investigated (Khaledi et al., 1991; Otsuka et al., 1985; Strasters and Khaledi, 1991). This study confirmed the role of solute–micelle electrostatic interactions, such that positively charged solutes are attracted by the negatively charged micelles, whereas negatively charged solutes are repelled. The relative partitioning of the neutral and ionised form of a given solute in the same solvent system is characterised by its diff(log P N – I ) parameter, which is defined as the difference between the log P values of the neutral and charged species of an ionisable solute and contains important structural information, since it expresses the influence of ionisation on the compound’s intermolecular forces and intramolecular interactions (Caron et al., 1999). As shown by Kubinyi (1993) the diff(log P N – I ) of a monoprotic substance in n-octanol–water ranges from 3 to 4 depending on the nature and delocalisation of the charge. These indicative values differ in other solvent systems,

being for example around 5 in 1,2-dichloroethane–water (Reymond et al., 1999), and it was shown that the more similar to water the organic solvent, the smaller the diff(log P N – I ). Comparative MEKC studies dealing with the relative micelle association of both neutral and ionised species are not available. In this work, the MEKC migration behaviour of the neutral and ionised forms of 10 acidic and basic model compounds was investigated using anionic SDS micelles. The accessible pH range in MEKC is restricted to about pH 5–12. Below pH 5 the electroosmotic flow (EOF) is so much reduced that the anionic micelles no longer migrate to the cathode but to the anode (Otsuka and Terabe, 1989). For this reason, model compounds with aqueous pKa values between 7 and 9 had to be chosen. The choice was further constrained by the peak capacity, defined by the width of the elution window, allowing only a narrow range of lipophilicity to be assessed.

2. Theory

2.1. Partitioning of neutral solutes in anionic SDS micelles The MEKC migration time of a neutral solute (t r ) that interacts with the anionic SDS micelles will fall in a window of migration time between that of a very polar, non-interacting solute (t 0 ) and that of a solute concentrated entirely in the micelles (t mc ). The following equation derived by Terabe et al. (1985) is used to calculate MEKC capacity factors: tr 2 t0 k MEKC 5 ]]]] t 0 (1 2 t r /t mc )

(2)

The k MEKC of a solute is linearly related to its partition coefficient (PMEKC ) between the aqueous and the micellar phase: n mc k MEKC 5 ] 5 f PMEKC n aq

(3)

where n mc and n aq are the numbers of moles of solute in the micelles (pseudostationary phase) and in the bulk aqueous phase (the mobile phase); and f is the chromatographic phase ratio defined as the ratio of the volume of the micellar phase (Vmc ) to that of the aqueous phase (Vaq ).

2.2. Partitioning of acids in SDS micelles In contrast to uncharged compounds, the migration of a weak acid (and its anion) cannot be fully explained by interactions with the micellar pseudophase, since they have their proper negative electrophoretic mobility in the aqueous phase. The overall apparent mobility of a weak acid ( m ) is the weighted average of the mobility of the micellar

A. Taillardat-Bertschinger et al. / European Journal of Pharmaceutical Sciences 15 (2002) 225 – 234

phase and its own mobility. This can be accounted for by including the apparent mobility in the absence of micelles ( mion ), as shown in Eq. (4) (Khaledi et al., 1991):

S

D S

D

k MEKC 1 m 5 ]]] mmc 1 ]]] mion k MEKC 1 1 k MEKC 1 1

(4)

Two migrating fractions can be discerned: the acid–conjugated base pairs associated with the micelles and migrating at their mobility ( mmc ), and the anionic (dissociated) form in the aqueous phase moving with the mobility of mion . The following equations were derived from Eq. (4) to calculate k MEKC values for acidic solutes:

m 2 mion k MEKC 5 ]]] mmc 2 m

(5)

and t r 2 t ion k MEKC 5 ]]]] t ions1 2 t r /t mcd

(6)

where t ion is the retention time of the acid in absence of micelles. This equation contains several assumptions, that (a) the mobility of the micelles is not altered when associated with an analyte, (b) secondary chemical equilibria with buffer components do not occur, and (c) the possible effects of the ionisable analytes and their equilibria on the zeta potential and hence electroosmotic flow are insignificant. Although it is difficult to measure the electrophoretic migration of the ionised solute (t ion ) directly in the micellar solution, it is reasonable to assume that it is the same as in buffers having an SDS concentration just below the critical micelle concentration (Khaledi et al., 1991; Otsuka et al., 1985).

2.3. Partitioning of bases in SDS micelles The migration behaviour of a weak base (and its cation) is influenced by an additional equilibrium relative to a weak acid, due to the electrostatic interactions between the free surfactant ions in the aqueous phase and the cations (ion-pairing) (Strasters and Khaledi, 1991). However, Eq. (6) can also be used to calculate k MEKC of weak bases and cations because t ion accounts for the additional equilibrium, as it is measured using an aqueous phase containing surfactant monomers just below the critical micelle concentration.

3. Experimental section

3.1. Materials 3.1.1. Chemicals The small organic solutes employed for the LSER analysis were obtained from different commercial sources (Merck, Darmstadt, Germany; Fluka Chemie, Buchs, Swit-

227

zerland; Janssen, Beerse, Belgium; Aldrich, Steinheim, Germany) and were of the highest available purity. The acidic and basic model compounds under study (Fig. 1) were bought in the highest available purity. 2Chlorophenol (A), 3-chlorophenol (B), 3-nitrophenol (C) and N,N-dimethylbenzylamine (E) were purchased from Fluka, phenytoin (D), (S)-(2)-nicotine (F), clonidine (G) and lidocaine (H) from Sigma (Buchs, Switzerland), and procaine (I) from Aldrich, whereas amfepramone (J) was a gift from Orgamol (Evionnaz, CH). When not specified otherwise, the racemate of chiral drugs was used. Sodium dodecyl sulfate (SDS) of Microselect quality were obtained from Fluka, Sudan Red 7B from Aldrich and methanol of superpure quality for HPLC from Romil Chemicals (Cambridge, UK). Analytical-grade n-octanol was from Fluka and potassium chloride from Merck. All further chemicals were of analytical grade. The solutions for the MEKC experiments were prepared using demineralised and purified water obtained with a Seralpur PRO 90 C system (Seral, Renggli, Rotkreuz, Switzerland). Distilled water was used for the potentiometric determination of the ionisation constants and the noctanol–water partition coefficients.

3.1.2. Equipment The micellar electrokinetic capillary chromatography experiments were performed in a 50-mm i.d. (inner diameter) fused-silica capillary (eCAPE Capillary Tubing, Beckman Instruments, Fullerton, CA, USA), 57 cm in length (50 cm in effective length up to the detector) fitted in a capillary cartridge supplied by Beckman. Before a new capillary was filled with running buffer, it was hydrated by high pressure conditioning (20 p.s.i.). This was achieved by rinsing during 10 min with 1 M HCl, 5 min with H 2 O, 10 min with 1 M NaOH, 5 min with H 2 O and 20 min with the respective running buffer. Prior to use, the micellar background electrolyte solutions were filtered through a 2-mm syringe filter (Titan syringe filters, Infochroma, Zug, Switzerland). The capillary temperature was kept at 2560.5 8C during analysis, using a liquid thermostating system filled with a non-conductive fluorocarbon liquid (P/ACEE System 2000, Beckman). The samples were placed in the inlet try of the P/ACE 5510 system (Beckman), controlled by the P/ACEE Station software (Beckman), and introduced into the capillary by low pressure (0.5 p.s.i.) injection for 5 s, followed by injection of running buffer for 1 s. During the MEKC experiments a constant voltage was applied and the detection was carried out using a P/ACE UV absorbance detector connected to the cathodic end of the capillary. Between runs, the capillary was washed with 0.1 M NaOH, followed by reconditioning with running buffer during 5 min. The ionisation constants and n-octanol–water partition I coefficients (log P Noct and log P oct ) of the 10 model compounds were determined by the potentiometric pH-method

228

A. Taillardat-Bertschinger et al. / European Journal of Pharmaceutical Sciences 15 (2002) 225 – 234

Fig. 1. Chemical structures of the acidic and basic compounds used to study the relative partitioning of neutral and ionised forms in SDS micelles by MEKC.

using a PCA101 titrator (Sirius Analytical Instruments, Forrest Row, East Sussex, UK).

3.2. Multivariate statistical analysis The LSER models were generated by multivariate regression using both the TSAR program (Tsar 3.3, Oxford Molecular, Oxford, UK, 2000) and the QSAR module in the Sybyl software (Sybyl 6.5, Tripos Associates, St. Louis, MO, USA, 1998) running on Silicon Graphics Indy R4400 175 MHz, O 2 R5000 180 MHz and Origin 2000 R10000 195 MHz workstations. The relative contributions of each variable to the LSER models were obtained by Mager’s standardisation (Mager and Barth, 1979).

3.3. Optimal set for LSER analysis and calculation of molecular volumes The optimal set previously described by Pagliara et al. (1995) is composed of 80 compounds whose solvatochromic parameters are known (Kamlet et al., 1988). The original solvatochromic parameters (a, b and p *) were used for this study, while the van der Waals volumes (Vw ) were calculated with the standard software MOLSV (QCPE No 509) and the atomic radii of Gavezzotti (1983). The geometries used to generate them were optimised with

the MMFF94s force field (Halgren, 1996). Due to the non-availability of some compounds (66–70), the poor solubility of others (74, 79), experimental problems associated with basic (41, 65) and acidic solutes (42, 52–55, 58) and to the fact that only the log k MEKC values for UVdetectable compounds could be assessed, the optimal set had to be reduced to 36 compounds (Table 1).

3.4. Measurements of MEKC capacity factors The retention times for the 36 compounds in the optimal set used to derive the LSER model were measured using a running buffer prepared with 150 mM BH 3 O 3 , 10 mM Na 2 B 4 O 7 and 50 mM SDS and adjusted to pH 7.0. A constant voltage of 20 kV was applied during micellar electrokinetic capillary chromatography experiments; detection was carried out at 214 nm. The MEKC capacity factors of the neutral and ionised forms of the 10 acidic and basic model compounds (A–J) (Fig. 1) were measured at pH 6.0 and pH 11.0, respectively. The measurements at pH 6.0 were carried out using a running buffer prepared with 10 mM CH 3 CO 2 Na and 150 mM BH 3 O 3 , containing either 50 mM or 4 mM SDS. The detection was carried out at 254 nm and a voltage of 30 kV was applied during analysis. The running buffer used at pH 11.0 consisted of 25 mM BH 3 O 3 and either 50 mM or 3

A. Taillardat-Bertschinger et al. / European Journal of Pharmaceutical Sciences 15 (2002) 225 – 234

229

Table 1 Solutes in the optimal set used to establish the LSER model for MEKC retention using SDS micelles No.a

Solutes

Vw

p *b

bb

ab

log Poct

log k MEKC c

10 11 12 15 16 29 30 31 32 33 34 35 36 37 38 39 40 43 44 45 46 47 48 49 50 51 56 57 59 60 61 71 72 75 77 78

CH 3 COOCH 3 CH 3 COOC 2 H 5 CH 3 COOC 4 H 9 CH 3 –CO–N(CH 3 ) 2 CH 3 –CO–N(C 2 H 5 ) 2 C 6 H 5 CH 3 C 6 H 5 –CO–CH 3 C 6 H 5 NO 2 C 6 H 5 OCH 3 C 6 H 5 COOC 2 H 5 C 6 H 5 –CO–C 2 H 5 C 6 H 5 COOCH 2 C 6 H 5 2-Cl–C 6 H 4 NO 2 C 6 H 5 CH 2 CN C 6 H 5 CH 2 –CO–CH 3 C 6 H 5 (CH 2 ) 2 –O–CO–CH 3 Pyridine 2-Naphthylamine C 6 H 5 NH 2 C 6 H 5 NHC 2 H 5 2-Cl–C 6 H 4 NH 2 2-NH 2 –C 6 H 4 –C 6 H 5 4,49-(NH 2 ) 2 -Biphenyl 4-NO 2 –C 6 H 4 –NH 2 C 6 H 5 OH 3-Cl–C 6 H 4 OH C 6 H 5 CH 2 OH 4-Cl–C 6 H 4 CH 2 OH 1,3-C 6 H 4 Cl 2 Biphenyl CH 3 SOCH 3 Naphthalene 1,3,5-C 6 H 3 (CH 3 ) 3 1,2,4,5-C 6 H 2 Cl 4 C 6 H 5 (CH 2 ) 2 C 6 H 5 C 6 H(CH 3 ) 5

71.4 88.6 121.3 92.6 125.9 103.9 122.1 107.2 111.4 147.7 139.4 206.6 120.6 122.4 139.2 164.5 82.4 145.3 98.4 132.6 133.2 173.3 186.0 118.3 93.7 108.5 111.2 126.7 116.4 163.3 70.5 133.9 138.0 146.4 196.6 170.9

0.60 0.55 0.51 0.88 0.84 0.55 0.90 1.01 0.73 0.74 0.88 1.32 1.11 1.34 1.30 1.14 0.87 0.83 0.73 0.82 0.83 1.32 1.46 1.25 0.72 0.77 0.99 1.11 0.75 1.18 1.00 0.70 0.47 0.70 1.10 0.39

0.42 0.45 0.45 0.76 0.78 0.11 0.49 0.30 0.32 0.41 0.49 0.50 0.26 0.41 0.58 0.55 0.44 0.50 0.50 0.47 0.40 0.60 1.00 0.48 0.33 0.23 0.52 0.42 0.03 0.20 0.76 0.15 0.13 0.00 0.22 0.17

0.00 0.00 0.00 0.00 0.00 0.00 0.04 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.35 0.26 0.17 0.25 0.26 0.62 0.42 0.61 0.69 0.39 0.40 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0.18 0.73 1.82 20.77 0.34 2.69 1.58 1.85 2.11 2.64 2.20 3.97 2.24 1.56 1.44 2.30 0.65 2.28 0.90 2.16 1.91 2.84 1.53 d 1.39 1.49 2.49 1.08 1.96 3.48 3.90 21.35 3.35 3.84 4.51 4.80 4.56

21.00 20.63 0.23 21.04 20.44 0.45 0.24 0.12 0.23 0.93 0.56 2.12 0.51 0.19 0.17 0.77 20.52 0.83 20.29 0.39 0.22 1.13 0.69 0.09 20.28 0.37 20.24 0.36 1.11 1.69 21.58 1.15 1.31 2.01 2.16 1.86

a

The same numbering as in reference (Pagliara et al., 1995) was used. The solvatochromic parameters as well as the log Poct values were taken from reference (Kamlet et al., 1988). c MEKC capacity factors obtained using anionic SDS micelles. d Log Poct value determined by potentiometric titration in this work. b

mM SDS. The detection wavelength was set at 214 nm and a voltage of 20 kV was applied during analysis. In fact, the voltage was chosen as a function of analysis time and generated current and therefore adapted to the experimental conditions used at pH 6.0 and pH 11.0, respectively. SDS concentrations of 4 mM (at pH 6.0) and 3 mM (at pH 11.0), slightly below the CMC under the respective experimental conditions (previously determined according to Khaledi et al. (1991), results not shown) were used to measure the migration time of the ionised bases and acids in the absence of micelles (t ion ). Stock solutions of all compounds were prepared in running buffer (10 mg / ml). Prior to analysis, 10–500 ml of these solutions were mixed with 100 ml methanol saturated with Sudan Red 7B and completed with running buffer up to 1 ml total sample volume. The end-concentrations were chosen as a function of the solute’s absorption at 214 and

254 nm, respectively. Methanol and Sudan Red 7B were added to all samples in order to determine t 0 or t mc , respectively. Each analysis was done in triplicate. The MEKC capacity factors of neutral and ionised compounds were calculated according to Eqs. (2) and (6), respectively. In each case, the SD was #0.05. The day-to-day variability using freshly prepared buffers was within similar limits.

3.5. Potentiometric determination of pKa values and noctanol–water partition coefficients The pKa values and partition coefficients in n-octanol– water of both the neutral and ionised forms of the model compounds (A–J) were measured by potentiometry. This technique is based on successive pH-metric titrations and its principles have been explained elsewhere (Avdeef,

A. Taillardat-Bertschinger et al. / European Journal of Pharmaceutical Sciences 15 (2002) 225 – 234

230

1992). The usual procedure as described in (Caron et al., 1997) was applied for pKa and partitioning measurements. pKa values were also determined using an aqueous phase containing zero KCl (i.e., zero ionic strength) and standardised 0.5 M NaOH instead of 0.5 M KOH. This allowed a comparison of pKa values in the presence and absence of micelles (see below). The apparent pKa value (pKapp ) of the bases and acids (A–J) in the presence of micelles was measured by adding 50 mM SDS to the aqueous phase. Unfortunately, the presence of potassium ions normally used in potentiometry leads to precipitation of sodium dodecyl sulfate as an insoluble potassium salt (Kuhn and Hoffstetter-Kuhn, 1993). Therefore, standardised 0.5 M NaOH was used instead of 0.5 KOH and no KCl was added to the distilled water. The usual procedure (Caron et al., 1997) was used for electrode calibration and pKa measurements, but the stirring time within each titration step was increased when using micelles, to allow equilibrium to be reached.

4. Results and discussion

4.1. LSERs: structural properties governing the MEKC retention of neutral compounds In this study, due to experimental limitations, the optimal set of 80 compounds defined by Pagliara et al. (1995) had to be restricted to 36 solutes (Table 1). Before deriving the solvatochromic equation for the retention behaviour of neutral solutes in MEKC using anionic SDS micelles, it was necessary to check whether the smaller set still allowed a well balanced exploration of the 4D space defined by Vw , a, b and p *. The LSER of log Poct was used to test the relevance of the reduced set. The optimal set of 80 compounds was characterised by the LSER model of Eq. (7) with a relative contribution of 53% for Vw , 35% for b, 10% for p * and 2% for a (Mager and Barth, 1979): log Poct 53.22310 22 (60.13310 22 )Vw 20.67(60.12)p * 1 0.21(60.16)a 2 4.10(60.23)b 1 0.17(60.19) 2

n 5 80; r 5 0.99; q 2 5 0.98; s 5 0.18; F 5 1246

(7)

In this and the following equations, 95% confidence limits are given in parentheses; n is the number of compounds; r 2 the squared correlation coefficient; q 2 the cross-validated correlation coefficient; s the standard deviation; and F the Fischer’s test. The final set of 36 compounds yielded the following LSER model for their log Poct values (Eq. (8)): log Poct 53.21310 22 (60.17310 22 )Vw 20.44(60.17)p * 10.45(60.22)a 24.27(60.24)b 10.07(60.24) 2

n536; r 50.99; q 2 50.99; s50.12; F 51205

(8)

with the relative contribution of 48% for Vw , 42% for b, 6% for p * and 4% for a. Thus, Eqs. (7) and (8) have similar regression coefficients and relative contributions of the variables. Indeed, even the smaller set of 36 solutes yields a solvatochromic equation with the correct balance of intermolecular forces controlling n-octanol–water partitioning. The linear solvation free-energy relationships (LSERs) for the log k MEKC values of the neutral organic compounds listed in Table 1 is given in Eq. (9): log KMEKC 5 2.32 3 10 22 (60.14 3 10 22 )Vw 2 0.67(60.12) ? p * 1 0.10(60.23)a 2 2.05(60.27)b 1 0.61(60.24) n 5 36; r 2 5 0.98; q 2 5 0.98; s 5 0.12; F 5 495

(9)

Standardisation (Mager and Barth, 1979) of Eq. (9) yields the relative contribution of each variable to the LSER equation, namely 61% for Vw , 35% for b, 2% for p * and 2% for a. A comparison of the LSER models generated for partitioning in n-octanol (Eq. (8)) and retention in MEKC (Eq. (9)) shows that in both systems the size and hydrogen-bond acceptor basicity of the solutes are the two predominant factors, as already reported (Vitha and Carr, 1998; Yang et al., 1996; Yang and Khaledi, 1995). The positive v values and the negative b values reveal that retention and partitioning increase with the size of the compounds and decrease with their hydrogen-bond acceptor strength, whereas the effects of dipolarity / polarisability and hydrogen-bond donor acidity are minor. A closer look at Eqs. (8) and (9) shows that the positive v and the negative b coefficient are significantly smaller in the LSER model derived for MEKC retention. The smaller v value in Eq. (9) indicates that micelles are more cohesive than bulk n-octanol (Yang and Khaledi, 1995) and can be explained by the fact that micelles are ordered systems, being aggregates of amphiphilic molecules with polar head-groups. The smaller b coefficient in the LSER model obtained for the micellar system shows that the anionic SDS micelles provide a more favourable environment for the partitioning of strong H-bond acceptor solutes than wet n-octanol, micelles being more H-bond acidic than n-octanol. This implies that log k MEKC and log Poct yield slightly distinct lipophilicity scales, especially when H-bond acceptor solutes are considered. The existing differences between the two partitioning systems are also reflected in the balance between the intermolecular forces governing MEKC retention and partitioning in n-octanol, as shown by the relative contribution of Vw , a, b and p * to log k MEKC and log Poct . Indeed, in MEKC the relative contribution of the volume term (Vw ) is slightly higher, and that of b slightly decreased, compared to partitioning in n-octanol.

A. Taillardat-Bertschinger et al. / European Journal of Pharmaceutical Sciences 15 (2002) 225 – 234

4.2. The relative partitioning of neutral and ionised forms of acids and bases in SDS micelles compared to their partitioning in n-octanol

231

be neutral at pH 11.0 will in fact be ionised in MEKC experiments. Further analysis of Table 2 shows that the diff(log P Noct– I ) values for all compounds except (S)-(2)-nicotine (F) (diprotic) and clonidine (G), ranks from 3 to 4 as normally observed for monoprotic substances in n-octanol–water N–I (Kubinyi, 1993). In contrast, the diff(log k MEKC ) parameters measured by micellar electrokinetic capillary chromatography strongly depend on the basic or acidic characteristics of the compound. In fact, for acidic solutes the N–I diff(log k MEKC ) value is positive, whereas it is negative for bases. This means that due to repulsive electrostatic interaction forces between the negatively charged compounds and the micelles with the same charge, the N log k IMEKC of acids are smaller than the log k MEKC values. For basic compounds, due to additional attractive electrostatic interactions between the solutes and micelles, the log k IMEKC values are increased compared to the log k NMEKC values. Fig. 2 shows the correlation obtained for the neutral forms of the solutes between their n-octanol–water partition coefficients and MEKC capacity factors. In fact, the values for these compounds are shown together with those of the optimal set of solutes analysed in Section 4.1, a previous analysis having shown that the different experimental conditions gave comparable log k NMEKC values (results not shown). As expected, the neutral forms of the 10 acidic and basic solutes (A–J) belong to the same regression as the 36 compounds investigated in Section 4.1. A closer look however shows that lidocaine (H) and procaine (I) are above the correlation line. As shown above, strong H-bond acceptors favour the more acidic micelles over n-octanol. The strong H-bond acceptor properties of lidocaine (H) and procaine (I) have indeed been confirmed using a

The MEKC capacity factors for the neutral (log k NMEKC ) and ionised (log k IMEKC ) forms of the acidic and basic model compounds (A–J) were measured using a pH 6.0 N and pH 11.0 running buffer, respectively. The log P oct , I log P oct , pKa values in the absence of micelles and apparent pKa in their presence were determined by potentiometric titration. The physicochemical parameters of compounds A–J are listed in Table 2. The log P Noct values are seen to rank from 1.44 to 2.77. Unfortunately, the range of lipophilicity explorable by MEKC is restricted by the width of the elution window defined by t 0 and t mc . The ionisation constants determined at a ionic strength of 0.15 M KCl are only slightly distinct from those measured without KCl in the aqueous phase. The comparison of the pKa values determined in the presence of 50 mM SDS (pKapp ) and in its absence shows that pKapp is shifted to higher values for both acidic and basic compounds. This can be explained by the fact that for ionised basic compounds additional attractive interaction forces exist between the anionic micelles and the protonated solutes, whereas for acids repulsive electrostatic interaction forces dominate solute–micelle association. Thus, the basicity of the bases is enhanced in the presence of micelles, whereas the acidity of the acids is decreased (El Seoud, 1989; Garrone et al., 1992). Our results as well as literature data published for a large series of compounds (El Seoud, 1989) show clearly that the difference between aqueous pKa and the pKa measured in the presence of micelles can be as large as 3 pH units. Concerning our study, this shift implies that some of the bases assumed to

Table 2 Physicochemical parameters of the acidic and basic compounds under study No.a

pKa b

log P Noct

I log P oct

diff(log P Noct– I )

pKa c

pKapp d

pKapp 2pKa e

N log k MEKC

I log k MEKC

N–I diff(log k MEKC )

A B C D E F

8.33 8.90 8.20 7.94 9.04 8.09 / 3.31 8.11 7.94 9.03 8.77

2.18 2.67 2.00 2.68 2.01 1.44

21.42 n.d. 21.51 20.99 21.60 21.12 f

3.59 n.d. 3.51 3.67 3.67 2.55

0.95 1.14 0.79 1.94 21.13 21.22

2.35 3.15 3.24 4.10

0.22 0.50 0.44 1.13 1.73 1.36 / 1.60 2.53 1.25 1.21 1.55

20.77 20.81 20.73 20.87 1.75 1.56 f

20.76 20.78 21.21 21.32

8.67 9.49 8.73 9.24 10.69 9.30 / 4.62 10.60 9.12 10.18 10.27

0.18 0.34 0.06 1.07 0.62 0.34

1.59 2.37 2.03 2.77

8.45 8.99 8.29 8.11 8.96 7.94 / 3.02 8.06 7.87 8.97 8.73

0.24 1.31 1.39 1.29

2.18 2.10 1.92 2.36

21.94 20.79 20.53 21.08

G H I J a

See Fig. 1 for structures. pKa determined with a ionic strength of 0.15 M KCl. c pKa measured with a ionic strength of zero KCl. d pKapp determined in presence of 50 mM SDS (micelles) without adding KCl. e pKapp measured in presence of micelles minus the aqueous pKa measured using no KCl. f n-Octanol partition coefficient and MEKC capacity factor of the monoprotonated form.

b

232

A. Taillardat-Bertschinger et al. / European Journal of Pharmaceutical Sciences 15 (2002) 225 – 234

Fig. 2. Relationship between log P Noct and log k NMEKC . The values for the acidic (d) and basic (s) model compounds were added to the plot obtained in Section 4.1 for the 36 compounds in the optimal set (w).

recent computational system developed in our institute (Rey et al., 2001). Fig. 3 illustrates the relationship between log k NMEKC and log k IMEKC for acidic and basic solutes. It clearly appears I N that the log k MEKC values do not depend on the log k MEKC values, but lie on two distinct plateaux. In fact, all acids have log k IMEKC values of about 21.0 and the capacity factor of protonated bases equals about 2.0. These ‘plateau’ values, however, may be influenced by the limitations of the method. Indeed, compounds having capacity factors higher than about 2.5 in the log scale will co-elute with t mc , whereas too hydrophilic compounds cannot be distinguished from t 0 . Fig. 3 gives further information about the balance between hydrophobic and electrostatic interactions responsible for solute–micelle association. Whereas a solute’s hydrophobicity determines the affinity of its neutral form for micelles, attractive or repulsive electrostatic interaction

Fig. 3. Relationship obtained between the capacity factors measured for the ionised forms and those determined for the neutral species of the acidic (d) and basic (s) model compounds under study.

forces govern the solute–micelle association of its ionised species. A highly hydrophobic acidic compound, e.g., phenytoin (D), having a high affinity for micelles in its neutral form, will be repelled when ionised. In contrast, a rather polar base, e.g., clonidine (G), having a small affinity for micelles in its neutral form, will strongly interact with the negatively charged micelles when proton–I ated, resulting in negative diff(log k NMEKC ) values. I The constant values of log k MEKC obtained for both acidic and basic molecules indicate that micelle partitioning of ionised compounds fully depends on their charge. –I Thus, the derivation of a diff(log k NMEKC ) value is devoid of meaning as this parameter is strongly correlated with log k NMEKC , due to the fact that log k IMEKC values are nearly constant for all acids and bases and do not depend on hydrophobicity. In other words, the log k IMEKC values obtained by micellar electrokinetic capillary chromatography are not discriminative. The pKa shift (pKapp 2pKa ) induced by solute–micelle association is plotted against the log k NMEKC values in Fig. 4A for the acidic solutes and in Fig. 4B for the bases. A possible trend can be suspected for both acidic and basic

Fig. 4. Relationship between the pKa shift (pKapp 2pKa ) induced by solute–micelle interactions and MEKC retention measured for the neutral forms of the acidic (A) and basic (B) model compounds.

A. Taillardat-Bertschinger et al. / European Journal of Pharmaceutical Sciences 15 (2002) 225 – 234

compounds. For the acidic solutes the pKa shift seems to N increase with increasing log k MEKC values, whereas for bases the compounds with the largest log k NMEKC value shows the smallest pKa shift. In fact, the pKa shift induced by the presence of an organic phase is known to be directly ¨ proportional to the diff(log P N – I ) parameter (Kramer et al., 1998). As the log k IMEKC values do not depend on the log k NMEKC but lie on two distinct plateau, the pKa shift is N only influenced by the log k MEKC values and thus the two are linearly related. Considering the small number of compounds under study and their diverse structures, we note that our results are merely qualitative and need to be expanded to further compounds covering a broader range of lipophilicity and chemical diversity.

233

However, the charge density of purely anionic SDS micelles is much higher than that of biomembranes, where only about 10% of the lipids of the inner membrane leaflet are negatively charged. Nevertheless, the use of mixed micelles containing anionic and non-ionic surfactants allows the percentage of ionised amphiphilic molecules to be adapted; such micelles may be more appropriate to model biological membranes.

Acknowledgements P.-A.C. and B.T. are grateful for financial support by the Swiss National Science Foundation.

References 5. Conclusions The LSER model reported here (Eq. (9)) shows that size and hydrogen-bond acceptor basicity govern the retention behaviour of neutral compounds in MEKC when anionic SDS micelles are used. This seems to be a general trend among systems where hydrophobic interactions control solute partitioning between an aqueous and an organic phase. However, a closer look at our results shows that the pseudo-stationary phase formed by the SDS micelles is more cohesive and more H-bond donor than n-octanol. Thus, the two lipophilicity scales are not interchangeable for strong H-bond acceptor solutes, since these solutes partition better in SDS micelles than in n-octanol. This difference may be of interest when investigating structure– permeation relations where the H-bonding capacity of permeants often plays a determining role. Furthermore, it was shown that strong attractive electrostatic interactions govern drug–micelle association between protonated bases and anionic SDS micelles, whereas repulsive electrostatic interactions separate ionised acidic compounds and anionic micelles. Overall our results indicate that electrostatic interactions determine solute– micelle interactions of ionised compounds (the hydrophobic component becoming negligible), whereas overall hydrophobic interactions govern the interactions of neutral compounds. The attraction and repulsion of ionised bases and acids, respectively, is also reflected by the –I diff(log k NMEKC ) parameter, which is negative for acidic compounds and positive for basic solutes and varies with the corresponding log k NMEKC value. In fact, the –I diff(log k NMEKC ) parameter does not contain specific information since the log k IMEKC values obtained by micellar electrokinetic capillary chromatography do not depend on the solute’s chemical structure but only on its charge. The above results underline the importance of ionic interactions between charged amphiphilic molecules, as existing in biological membranes, and ionised compounds.

Avdeef, A., 1992. pH-Metric log P. Part 1. Difference plots for determining ion-pair octanol-water partition coefficients of multiprotic substances. Quant. Struct. -Act. Relatsh. 11, 510–517. Betageri, G., Rogers, J., 1988. The liposome as a distribution model in QSAR studies. Int. J. Pharm. 46, 95–102. Caron, G., Gaillard, P., Carrupt, P., Testa, B., 1997. Lipophilicity behavior of model and medicinal compounds containing a sulfide, sulfoxide, or sulfone moiety. Helv. Chim. Acta 80, 449–462. Caron, G., Reymond, F., Carrupt, P., Girault, H., Testa, B., 1999. Combined molecular lipophilicity descriptors and their role in understanding intramolecular effects. Pharm. Sci. Technol. Today 2, 327– 335. El Seoud, O., 1989. Effects of organised surfactant assemblies on acidbase equilibria. Adv. Colloid Interface Sci. 30, 1–30. Garrone, A., Marengo, E., Fornatto, E., Gasco, A., 1992. A study on pK app and partition coefficient of substituted benzoic acids in SDS a anionic micellar system. Quant. Struct. -Act. Relat. 11, 171–175. Gavezzotti, A., 1983. The calculation of molecular volumes and the use of volume analysis in the investigation of structured media and of solid-state organic reactivity. J. Am. Chem. Soc. 105, 5220–5225. Halgren, T., 1996. Merck molecular force field. II. MMFF94 van der Waals and electrostatic parameters for intermolecular interactions. J. Comput. Chem. 17, 520–552. Kamlet, M., Doherty, R., Abraham, M., Marcus, Y., Taft, R., 1988. Linear solvation energy relationships. 46. An improved equation for correlation and prediction of octanol / water partition coefficients of organic nonelectrolytes (including strong hydrogen bond donor solutes). J. Phys. Chem. 92, 5244–5255. Kamlet, M., Taft, R., 1976. The solvatochromic comparison method I. The b-scale of solvent hydrogen-bond acceptor (HBA) basicities. J. Am. Chem. Soc. 98, 377–383. Khaledi, M., Smith, S., Strasters, J., 1991. Micellar electrokinetic capillary chromatography of acidic solutes: migration behavior and optimisation strategies. Anal. Chem. 63, 1820–1830. ¨ Kramer, S., Gautier, J., Saudemon, P., 1998. Considerations on the potentiometric log P determinations. Pharm. Res. 15, 1310–1313. Kubinyi, H., 1993. In: QSAR: Hansch Analysis and Related Approaches. VCH, Weinheim. Kuhn, R., Hoffstetter-Kuhn, S., 1993. In: Capillary Electrophoresis: Principles and Practice. Springer, Berlin. Mager, H., Barth, A., 1979. Problems involved in the specification and interpretation of quantitative structure–activity relationships. Pharmazie 34, 557–559. Ong, S., Liu, H., Qiu, X., Bhat, G., Pidgeon, C., 1995. Membrane

234

A. Taillardat-Bertschinger et al. / European Journal of Pharmaceutical Sciences 15 (2002) 225 – 234

partition coefficients chromatographically measured using immobilized artificial membrane surfaces. Anal. Chem. 67, 755–762. Otsuka, K., Terabe, S., 1989. Effects of pH on electrokinetic velocities in micellar electrokinetic chromatography. J. Microcol. Sep. 1, 150–154. Otsuka, K., Terabe, S., Ando, T., 1985. Electrokinetic chromatography with micellar solutions retention behaviour and separation of chlorinated phenols. J. Chromatogr. 348, 39–47. Pagliara, A., Khamis, E., Trinh, A., Carrupt, P., Tsai, R., Testa, B., 1995. Structural properties governing retention mechanisms on RP-HPLC stationary phase used for lipophilicity measurements. J. Liquid Chromatogr. 18, 1721–1745. Pauletti, G., Wunderli-Allenspach, H., 1994. Partition coefficients in vitro: artificial membranes as a standardized distribution model. Eur. J. Pharm. Sci. 1, 273–282. Rey, S., Caron, G., Ermondi, G., Gaillard, P., Pagliara, A., Carrupt, P., Testa, B., 2001. Development of molecular hydrogen bonding potentials (MHBPs) and their application to structure permeation relations. J. Mol. Graphics Model 19, 521–535. Reymond, F., Carrupt, P., Testa, B., Girault, H., 1999. Charge and delocalisation effects on the lipophilicity of protonable drugs. Chem. Eur. J. 5, 39–47. Strasters, J., Khaledi, M., 1991. Migration behavior of cationic solutes in micellar electrokinetic capillary chromatography. Anal. Chem. 63, 2503–2508. Taft, R., Kamlet, M., 1976. The solvatochromic comparison method. 2 The a-scale of solvent hydrogen-bond donor (HBD) acidities. J. Am. Chem. Soc. 98, 2886–2894. Tanaka, A., Nakamura, K., Nakanishi, I., Fujiwara, H., 1994. A novel and useful descriptor for hydrophobicity, partition coefficient micellarwater, and its application to a QSAR study of antiplatelet agents. J. Med. Chem. 37, 4563–4566.

Terabe, S., Otsuka, K., Ando, T., 1985. Electrokinetic chromatography with micellar solution and open-tubular capillary. Anal. Chem. 57, 834–841. Testa, B., Caron, G., Crivori, P., Rey, S., Reist, M., Carrupt, P., 2000. Lipophilicity and related molecular properties as determinants of pharmacokinetic behaviour. Chimia 54, 672–677. Testa, B., Carrupt, P., Gaillard, P., Tsai, R., 1996. Intramolecular interactions encoded in lipophilicity: their nature and significance. In: Pliska, V., Testa, B., van de Waterbeemd, H. (Eds.), Lipophilicity in Drug Action and Toxicology. VCH, Weinheim, pp. 49–71. Treiner, C., Chattopadhyay, A., 1986. Correlation of partition coefficients for polar aromatic and aliphatic molecules between trimethyldodecylammonium bromide micelles1water and octanol1 water systems at 298.15 K. J. Colloid Interface Sci. 109, 101–108. Vitha, M., Carr, P., 1998. A linear solvation energy relationship study of the effects of surfactant chain length on the chemical interactions governing retention and selectivity in micellar electrokinetic capillary chromatography using sodium alkyl sulfate elution buffers. Sep. Sci. Techn. 33, 2075–2100. Yang, S., Bumgarner, J., Kruk, L., Khaledi, M., 1996. Quantitative structure–activity relationships studies with micellar electrokinetic chromatography. Influence of surfactant type and mixed micelles on estimation of hydrophobicity and bioavailability. J. Chromatogr. A 721, 323–335. Yang, S., Khaledi, M., 1995. Chemical selectivity in micellar electrokinetic chromatography: characterization of solute-micelle interactions for classification of surfactants. Anal. Chem. 67, 499–510. Yeagle, P., 1992. In: The Structure of Biological Membranes. CRC Press, Boca Raton, FL.