Molecular Interactions in Lipophilic Environments Studied by

Anal. Chem. 2008, 80, 203-208

Molecular Interactions in Lipophilic Environments Studied by Electrochemistry at Interfaces between Immiscible Electrolyte Solutions Maria A. Deryabina, Steen H. Hansen, and Henrik Jensen*

Department of Pharmaceutics and Analytical Chemistry, Faculty of Pharmaceutical Sciences, University of Copenhagen, DK-2100 Copenhagen, Denmark

The description and understanding of absorption and distribution of potential new drug compounds in the organism is of paramount importance for the successful development of new drugs. However, the currently used physical chemical parameters such as oil-water distribution coefficients and ionization constants frequently fall short when it comes to a detailed description of the highly heterogeneous environments of both lipophilic and hydrophilic characters through which the drug compound passes. In this work, a new procedure based on electrochemistry at the interface between immiscible electrolyte solutions for addressing drug compound-ligand interactions in lipophilic environments as well as nonspecific ligand effects on distribution behavior has been developed. An attractive feature of the method is that it can simultaneously provide data for oil-water partition coefficients and ionization constants. The new procedure is demonstrated using five drug compounds with different physical chemical parameters and cholesterol as the oilphase ligand. The use of ligand shift ion partition diagrams in the data presentation allows a quick visualization and comparison of a series of related drug compounds. In order to accurately describe absorption, distribution, and function of drug compounds in the human organism, it is customary to rely on a series of well-established physical chemical parameters such as oil-water distribution coefficients and ionization constants.1,2 However, considering the complex and diverse routes of drug compound absorption, these simple parameters often fall short when it comes to a more detailed description and understanding. Frequently, drug-ligand interactions in heterogeneous environments (as, for example, cell membranes, the blood brain barrier, or human skin) play an important role in the overall picture. Although a number of procedures are now available that can effectively address noncovalent drug-ligand interactions in aqueous solutions,3-7 there appears to be no * To whom correspondence should be addressed. E-mail: [email protected]. Fax: +45-35306010. (1) Avdeef, A.; Testa, B. Cell. Mol. Life Sci. 2002, 59, 1681-1689. (2) Pagliara, A.; Reist, M.; Geinoz, S.; Carrupt, P. A.; Testa, B. J. Pharm. Pharmacol. 1999, 51, 1339-1357. (3) Schalley, C. A. Analytical methods in supramolecular chemistry, 1st ed.; WileyVCH Verlag GmbH & Co.: Weinheim, 2007. (4) Ostergaard, J.; Heegaard, N. H. H. Electrophoresis 2006, 27, 2590-2608. 10.1021/ac071276t CCC: $40.75 Published on Web 12/04/2007

© 2008 American Chemical Society

efficient general methodologies when it comes to nonpolar waterimmiscible solvents. This point is rather important as, for instance, hydrogen bonds and ion-ion interactions are highly medium dependent. Ion-transfer reactions at interfaces between immiscible electrolyte solutions (ITIES) can be studied using electrochemical methods.8-13 In the case of ionizable drug compounds, electrochemistry at ITIES has proven particularly useful as it can provide information on the partition coefficient of the charged and neutral forms of the drug compound as well as pKa values.14-16 In practice, the transfer potential is measured at different pH of the aqueous phase. The results are then presented in an ionic partition diagram,14,16-18 which gives a convenient overview of the partition behavior of the various forms of the drug compounds at different pH. This knowledge is essential to predict and describe how drug compounds are absorbed and distributed in the organism. However, a simple homogeneous oil phase or organic solvent is a rather crude measure of the highly diverse and heterogeneous hydrophobic barriers that drug compounds may pass. In this paper, a general method based on electrochemistry at ITIES that can be used to investigate drug compound-ligand interactions in lipophilic environments as well as nonspecific ligand effects on drug distribution is described. As test system, choles(5) Ostergaard, J.; Khanbolouki, A.; Jensen, H.; Larsen, C. Electrophoresis 2004, 25, 3168-3175. (6) Ostergaard, J.; Heegaard, N. H. H. Electrophoresis 2003, 24, 2903-2913. (7) Ostergaard, J.; Schou, C.; Larsen, C.; Heegaard, N. H. H. Electrophoresis 2002, 23, 2842-2853. (8) Koryta, J. Electrochim. Acta 1988, 33, 189-197. (9) Koryta, J. Electrochim. Acta 1984, 29, 445-452. (10) Koryta, J. Electrochim. Acta 1979, 24, 293-300. (11) Reymond, F.; Fermin, D.; Lee, H. J.; Girault, H. H. Electrochim. Acta 2000, 45, 2647-2662. (12) Samec, Z. Pure Appl. Chem. 2004, 76, 2147-2180. (13) Reymond, F.; Gobry, V.; Bouchard, G.; Girault, H. H. Electrochemical Aspects of Drug Partitioning. In Pharmacokinetic Optimization in Drug Research; Testa, B., van der Waterbeemd, H., Folkers, G., Guy, R., Eds.; Wiley-VCH: Zurich, 2001; pp 327-349. (14) Reymond, F.; Chopineaux-Court, V.; Steyaert, G.; Bouchard, G.; Carrupt, P. A.; Testa, B.; Girault, H. H. J. Electroanal. Chem. 1999, 462, 235-250. (15) Reymond, F.; Steyaert, G.; Carrupt, P. A.; Testa, B.; Girault, H. J. Am. Chem. Soc. 1996, 118, 11951-11957. (16) Ulmeanu, S. M.; Jensen, H.; Bouchard, G.; Carrupt, P. A.; Girault, H. H. Pharm. Res. 2003, 20, 1317-1322. (17) Chopineaux-Court, V.; Reymond, F.; Bouchard, G.; Carrupt, P. A.; Testa, B.; Girault, H. H. J. Am. Chem. Soc. 1999, 121, 1743-1747. (18) Gobry, V.; Ulmeanu, S.; Reymond, F.; Bouchard, G.; Carrupt, P. A.; Testa, B.; Girault, H. H. J. Am. Chem. Soc. 2001, 123, 10684-10690.

Analytical Chemistry, Vol. 80, No. 1, January 1, 2008 203

Figure 2. Electrochemical cell and corresponding cell diagram.

Figure 1. Chemical structures of the investigated compounds: lidocaine hydrochloride (LidHCl), prilocaine hydrochloride (PriHCl), propranolol hydrochloride (ProHCl), diflunisal (DifH), warfarin (WarfH), and cholesterol (Chol).

terol (oil-phase ligand) and a total of five drug compounds having different physical chemical parameters and molecular structures have been studied (Figure 1). The data are presented in an easily interpretable manner using a new type of ionic partition diagram termed “ligand-shift partition diagram”. From the visual representation, it is directly evident which form of the drug compound interacts with the ligand as well as the magnitude of the interaction. EXPERIMENTAL SECTION Chemicals. All aqueous solutions were prepared using deionized water from a Milli-Q system. 1,2-Dichloroethane (1,2DCE) was HPLC grade (Aldrich) and was handled with all necessary precautions. The organic-phase supporting electrolyte salt bis(triphenylphosphoranylidene) ammonium tetrakis(pentafluorophenyl) borate (BTPPATPFB) was prepared by metathesis of equimolar quantities of the corresponding salts, lithium tetrakis(pentafluorophenyl) borate (Boulder Scientific Co.) and bis(triphenylphosphoranylidene) ammonium chloride (BTPPACl) (Fluka) in a minimum amount of a 2:1 methanol/water mixture. The resulting BTPPATPFB precipitate was filtered and recrystallized from acetone before use. Lidocaine hydrochloride, prilocaine hydrochloride, diflunisal, warfarin, and propranolol hydrochloride (all purchased from Sigma-Aldrich) were of analytical grade. Tetramethylammonium chloride (TMACl) and tetrapropylammonium chloride (TPACl) were supplied by Fluka. All inorganic salts were provided by Merck. Instrumentation and Experimental Procedure. Cyclic voltammograms were recorded using a four-electrode potentiostat (Autolab, PGSTAT30, Eco-Chemie) using IR drop compensation in order to compensate for the solution resistance. An aqueous 204 Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

buffer solution with an ionic strength of 0.010 mol/L (adjusted with NaCl) was used as the aqueous electrolyte and 2 × 10-3 mol/L acetate, phosphate, or borate buffers was used to adjust the pH to the desired level. The concentration of the 1,2-DCE supporting electrolyte BTPPATPFB was 5 × 10-3 mol/L. The concentration of BTPPACl in the aqueous reference phase was 1 × 10-3 mol/L. The volume of each phase was 2 mL, the interface was 1 cm2, and the experiments were carried out at room temperature (20 ( 2 °C). The analytes were placed in the water phase; the concentration of the studied drugs was 0.1 × 10-3 mol/L. For studies on drug-cholesterol interactions, cholesterol was dissolved in the oil phase in known concentrations. A schematic of the electrochemical cell and the corresponding cell diagram is shown in Figure 2. The electrochemical glass cell was custom-made. All ion-transfer half-wave potentials were referred to the half-wave potential of the TMA+ and TPA+ ions added to the aqueous phase in the form of TMACl and TPACl, at the end of each experiment. The half-wave potential were converted into standard transfer potentials using standard transfer potentials of 0.160 and -0.093 V for the TMA+ and TPA+ ions, respectively. Additional details on the procedure have been reported in the literature.14 RESULTS AND DISCUSSION A new procedure for the determination of physicochemical characteristics of drug substances with relevance to their absorption and distribution in vivo has been developed. As model drug compounds, a number of bases (lidocaine (Lid), prilocaine (Pri), and propranolol (Pro)) and acids (diflunisal (DifH) and warfarin (WarfH)) were chosen (structures are shown in Figure 1). Using the developed procedure, ionic partition diagrams may be constructed and from those the overall distribution behavior at different pH as well as quantitative information on drug ligand effects on distribution can be obtained. The shape of the ionic partition diagram is dependent on the charge of the basic as well as acidic form of the drug molecule; the two groups of compounds shall therefore be discussed separately. The details of ionic partition diagrams have been described in the literature,13,15,16 but it will briefly be reviewed in the case of lidocaine. In electrochemistry at ITIES, the interfaces between the immiscible electrolyte solutions are polarized, thus enabling ion-

and negative peak potential) pertaining to lidocaine is independent of pH and can be expressed as

∆wo φ1/2 ) ∆wo φoi ′

(2)

where ∆wo φoi ′ is the formal transfer potential related to ∆wo φoi according to

Figure 3. Cyclic voltammogram corresponding to the ion transfer of LidH+ between water and 1,2-DCE at pH 4.0.

∆wo φoi ′ ) ∆wo φio +

o RT γi ln w zi F γ

(3)

i

where γoi and γwi are the activity coefficients of i in oil and water, respectively. The standard partition coefficient of i, log P oi , can be obtained according to13,20

zi F ∆wφo′ log Poi ) RT ln 10 o i

(4)

When the pH of the water phase approaches the pKa (the second boundary line, separating B (oil) and BH+ (oil)), the signal becomes dependent on pH according to eq 5 valid for lipophilic B and assuming identical diffusion coefficients of B and BH+:16 Figure 4. Ion partition diagram for LidHCl obtained at the aqueous buffer-1,2-DCE interface.

transfer phenomena between the two phases to be studied. The distribution of an ion, i, is determined by the Nernst-Donnan equation for the ITIES: o

∆wo φ ) ∆wo φoi +

RT ai ln z iF a w i

(1)

where ∆wo φoi is the standard transfer potential of the ion and aoi and awi are the activities of i in oil and water, respectively. The protonated form of lidocaine (LidH+) has a pKa of 7.9. At low pH, lidocaine is thus in the form LidH+ and at high pH it has the form of the free base, Lid. In Figure 3 is shown the cyclic voltammogram (CV) corresponding to the transfer of lidocaine at pH 4 between water and oil. At low potentials, the CV is limited by the transfer of the negative ions of the aqueous-phase electrolyte to the oil phase, and at high potentials, the cation of the aqueous-phase electrolyte limits the potential window. The signal corresponding to LidH+ appears in the middle of the potential window. The partition diagram for lidocaine is shown in Figure 4. At low pH, the first boundary line separates the conditions corresponding to LidH+ in oil and LidH+ in water. Under these conditions and assuming that the water and 1,2-DCE diffusion coefficients are identical (it is possible to correct for this effect;19 it is, however, insignificant in the present set of experiments), the half wave potential (obtained as the average of the positive (19) Samec, Z.; Langmaier, J.; Trojanek, A. J. Electroanal. Chem. 1996, 409, 1-7.

∆wo φ1/2 ) ∆wo φoi ′ +

RT ln 10 (log PN + pH - pKa) (5) zi F

where log PN is the partition coefficient of the neutral free base. The intercept of the two lines corresponding to eqs 2 and 5 is given by

pH ) pKa - log PN

(6)

Equation 6 defines the third boundary line, which is parallel to the ordinate axis. It is thus possible to obtain information on the pKa value, log PN of the free base of lidocaine, and the partition behavior of the protonated form of lidocaine from the partition diagram. Often the pKa value is known from other sources,21 leading to a direct determination of the log PN values. Further, from the measured partition coefficients of the ionized and neutral forms of the drug compound, dissociation constants in 1,2-DCE may be obtained.14 Ionic partition diagrams are valuable tools for quick visualization of the conditions under which the drug compound partition from an aqueous phase into a lipophilic environment (such as, for example, a cell membrane, the blood brain barrier, or through skin). However, the oil-water partition coefficient is a rather crude model for drug transport in biological systems, as biorelevant lipophilic environments have a highly heterogeneous composition. It has previously been demonstrated that interactions between metal ions and an ionophore can be quantified from the electro(20) Reymond, F.; Chopineaux-Court, V.; Steyaert, G.; Bouchard, G.; Carrupt, P. A.; Testa, B.; Girault, H. H. J. Electroanal. Chem. 1999, 462, 235-250. (21) Hansch, C. Comprehensive medicinal chemistry; Pergamon Press: New York, 1990; Vol. 6.

Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

205

Figure 5. Ligand shift ion partition diagram for the LidHCl/Chol system obtained at the aqueous buffer-1,2-DCE interface.

Figure 7. Ion partition diagram for ProHCl obtained at the aqueous buffer-1,2-DCE interface in the presence and absence of 30 × 10-3 mol/L cholesterol in 1,2-DCE.

insoluble in water),27 which simplifies data analysis significantly. Using eq 4, the partition coefficient of the charged form in the presence and absense of 30 × 10-3 mol/L cholesterol can be calculated (Table 1). According to the results presented in Figure 5, it can be assumed that cholesterol assists the transfer of charged lidocaine from water to oil by stabilizing the drug compound in the oil phase. According to eq 7, an association constant, K1:1, (assuming a 1-1 complex) of LidH+-cholesterol can be obtained (Table 1)23

∆wo φ1/2 ) ∆wo φoi ′ Figure 6. Ligand shift ion partition diagram for the PriHCl/Chol system obtained at the aqueous buffer-1,2-DCE interface.

chemical signal.8-10,22-25 However, to our knowledge, there have never been systematic comparative studies on drug compoundligand interactions in organic solvents. In the present investigation, cholesterol was chosen as a model oil-phase ligand. Cholesterol is an important constituent of skin and cell membranes, where it plays a key role in transport processes.26 Case 1. Bases. When cholesterol was introduced in the oil phase, a dependence of the formal transfer potential upon the cholesterol concentration was observed. In the case of the studied bases, the maximum concentration used was 30 × 10-3 mol/L cholesterol. In Figure 5 is shown the ligand shift diagram corresponding to the lidocaine-cholesterol system. It is immediately apparent that the signal corresponding to lidocaine is shifted in the presence of 30 × 10-3 mol/L cholesterol in the oil phase. However, the intercept between the oblique and horizontal lines remains the same, indicating that cholesterol only interacts to a measurable extend with the protonated form of lidocaine. The pKa of lidocaine in the aqueous phase is not dependent on the presence of cholesterol in the oil phase (cholesterol is virtually (22) Reymond, F.; Carrupt, P. A.; Girault, H. H. J. Electroanal. Chem. 1998, 449, 49-65. (23) Matsuda, H.; Yamada, Y.; Kanamori, K.; Kudo, Y.; Takeda, Y. Bull. Chem. Soc. Jpn. 1991, 64, 1497-1508. (24) Beattie, P. D.; Willington, R. G.; Girault, H. H. J. Electroanal. Chem. 1995, 396, 317-323. (25) Reymond, F.; Brevet, P. F.; Carrupt, P. A.; Girault, H. J. Electroanal. Chem. 1997, 424, 121-139. (26) Bastiaanse, E. M. L.; Hold, K. M.; van der Laarse, A. Cardiovasc. Res. 1997, 33, 272-283.

206

Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

RT ln(K1:1CoL) z iF

(7)

where K1:1 is the association constant and CoL is the concentration of ligand (cholesterol) in the oil phase. Interestingly, as shown in Figure 6, the related compound prilocaine behaves differently. The intercept is affected by the presence of cholesterol in the oil phase, indicating an interaction between cholesterol and the neutral form of the drug compound. In this case, cholesterol thus interacts with both forms of prilocaine in the oil phase. On the other hand, the partition behavior of the drug compound propranolol (Figure 7) is not affected by cholesterol (i.e., neither the charged nor neutral forms). The results pertaining to the three compounds may be interpreted from their respective molecular structures. The important difference between lidocaine and prilocaine is that lidocaine is a tertiary amine whereas prilocaine is a secondary amine. The steric hindrance at the tertiary amino group in lidocaine prevents a strong hydrogen bond with the hydroxyl group of cholesterol to be formed. In prilocaine, on the other hand, the steric hindrance at the secondary amino group is much less, thus allowing stabilization by cholesterol. This interpretation is in agreement with studies on hydrogen bond formation between alcohols and secondary and tertiary amines in solvents of low polarity, showing the weakest hydrogen bond formation by tertiary amines.28,29 Furthermore, the fact that prilocaine can donate a hydrogen bond to the hydroxyl group in cholesterol also leads to (27) Madan, D. K.; Cadwallader, D. E. J. Pharm. Sci. 1973, 62, 1567-1569. (28) Lin, M. L.; Scott, R. M. J. Phys. Chem. 1972, 76, 587-591. (29) Goralski, P. Thermochim. Acta 1994, 235, 31-38.

Table 1. Experimental Results for the Investigated Compounds Obtained by Employing Electrochemistry at ITIES compound

∆wo φ°′a (V)

K1:1b (L/mol)

pKa c

log PoI d

∆log PoI e

log PN f

∆log PN g

LidH+ PriH+ ProH+ DifH WarfH

0.120 0.201 0.115 -0.096 -0.112

135 125

7.9 7.9 9.5 3.0 5.1

-2.0 -3.4 -1.9 -1.6 -1.9

0.73 0.39 0.0 0.0 (-0.22)h 0.0 (-0.37)h

1.7 2.3 3.1 2.6 2.7

0.0 0.8 0.0 0.0 (-0.6)h 0.0 (-0.3)h

a Formal transfer potential obtained at the aqueous buffer-1,2-DCE interface. b A 1:1 association constant in 1,2-DCE between cholesterol and the charged drug compound obtained from eq 7 by a data-fitting procedure of data points corresponding to cholesterol concentrations of 0, 10, 20, and 30 × 10-3 mol/L. c Literature values.21 d 1,2-DCE water partition coefficients of the ionized form of the drug compound obtained from the formal standard transfer potentials according to eq 4. e Effect of cholesterol (30 × 10-3 mol/L unless otherwise noted) on the 1,2-DCE-water partition coefficients of the ionized form of the drug compound. f 1,2-DCE water partition coefficients of the neutral form of the drug compound obtained from the ion partition diagrams. g Effect of cholesterol (30 × 10-3 mol/L unless otherwise noted) on the 1,2-DCE-water partition coefficients of the neutral form of the drug compound. h Obtained using a cholesterol concentration of 50 × 10-3 mol/L.

Figure 8. Ligand shift ion partition diagram for the DifH/Chol system obtained at the aqueous buffer-1,2-DCE interface.

Figure 9. Ligand shift ion partition diagram for the WarfH/Chol system obtained at the aqueous buffer-1,2-DCE interface.

a stronger interaction.30 Finally, the compound propranolol has a hydroxyl group in the β position from the amine group. The fact that no stabilizing effect of cholesterol was observed for this compound indicates that the β-hydroxy group is capable of forming an intramolecular hydrogen bond with the amine group of the free base and thereby blocks for any appreciable interactions with cholesterol to take place. It is also conceivable that the lone pair of the hydroxy group interacts with the ammonium ion of the protonated form of propranolol thus preventing additional stabilization by cholesterol. Case 2. Acids. In contrast to the results pertaining to lidocaine and prilocaine, no substantial effect of cholesterol on the partition diagrams of diflunisal and warfarin was observed for concentrations less than 30 × 10-3 mol/L. At concentrations higher than 30 × 10-3 mol/L, a small gradual shift in the formal transfer potential of the acids was observed. In Figures 8 and 9 are presented ligand shift diagrams in the presence and absence of 50 × 10-3 mol/L cholesterol for the two acids, diflunisal and warfarin. The partition diagrams of acids are inverted with respect to bases according to the sign of the charge of the ionized species. In the presence of cholesterol, the signal corresponding to the deprotonated acid is shifted. The intercept between these two lines is also changed, suggesting that cholesterol affects the distribution of both the acid and its corresponding base. The variation of the half-wave potential with cholesterol concentration appears only at high concentrations of cholesterol (more than 30 × 10-3

mol/L, Table 1). This finding is different from what was observed for lidocaine and prilocaine, where a substantial effect was observed even for cholesterol concentrations as low as 10 × 10-3 mol/L. The results pertaining to these two acids may thus be interpreted as a general medium effect rather than a relatively strong specific interaction is observed for some of the bases. Furthermore, the effect of cholesterol is to decrease the oil-water distribution coefficient (albeit to a small extend), which does not comply with a specific (stabilizing) interaction. A possible explanation for this phenomenon is that the introduction of cholesterol makes the oil phase more lipophilic compared to neat 1,2-DCE, which should render the transfer of the charged drug compounds more difficult.

(30) Spencer, J. N.; Wolbach, W. S.; Hovick, J. W.; Ansel, L.; Modarress, K. J. J. Solution Chem. 1985, 14, 805-814.

CONCLUSION Lead structures for potential new drugs are routinely screened for their affinity for specific receptor or target sites. Later in the development phase, their physical chemical parameters are determined. Today a relatively limited number of parameters (notably oil-water distribution coefficients and ionization constants) are used for physical chemical characterization prior to clinical tests. However, for most drug compounds, the actual site of action and distribution in the organism is far more complex and additional parameters are therefore required for a detailed description and understanding. In this work, it was described how noncovalent interactions in a lipophilic environment (i.e., an organic solvent) can be quantified and used to explain the relative partition behavior of a group of model drug compounds. The methodology is generally applicable and is able to distinguish in Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

207

which form the drug compounds interact with an oil-phase ligand. The relatively simple data representation makes it straightforward to visually compare different drug compounds as well as different ligands without having to rely on complicated interpretations and calculations. Of particular relevance to the present work, physical chemical characterization of drug delivery to membrane-bound drug targets can be mentioned, but the developed methodology is generally applicable for studies on ligand effects on drug transport and distribution.

208

Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

ACKNOWLEDGMENT The Drug Research Academy at the Faculty of Pharmaceutical Sciences, University of Copenhagen, is acknowledged for financial support for this project.

Received for review June 18, 2007. Accepted October 6, 2007. AC071276T