Measurement of Dissociation Constants (pKa Values) of Organic Compounds by Multiplexed Capillary Electrophoresis Using Aqueous and Cosolvent Buffers MARINA SHALAEVA,1 JEREMY KENSETH,2 FRANCO LOMBARDO,1 ANDREA BASTIN2 1

Pfizer Global Research and Development, Groton Laboratories, Groton, Connecticut 06340

2

CombiSep, 2711 South Loop Drive, Suite 4200, Ames, Iowa 50010

Received 7 June 2007; revised 15 October 2007; accepted 15 November 2007 Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.21287

ABSTRACT: Evaluation of a multiplexed capillary electrophoresis (CE) method for pKa measurements of organic compounds, including low solubility compounds, is presented. The method is validated on a set of 105 diverse compounds, mostly drugs, and results are compared to literature values obtained from multiple references. Two versions of the instrument in two different labs were used to collect data over a period of 3 years and inter-laboratory and inter-instrument variations are discussed. Twenty-four point aqueous and mixed cosolvent buffer systems were employed to improve the accuracy of pKa measurements. It has been demonstrated that the method allows direct pKa measurements in aqueous buffers for many compounds of low solubility, often unattainable by other methods. The pKa measurements of compounds with extremely low solubility using multiplexed CE with methanol/water cosolvent buffers are presented. ß 2008 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci Keywords: pKa measurements; multiplexed capillary electrophoresis; low solubility compounds; mixed cosolvent buffers; aqueous buffers; dissociation constants of drug molecules

INTRODUCTION The acid–base dissociation constant of substances (pKa value) is a very important parameter in drug design and optimization. The degree of ionization strongly affects solubility, permeability, and drug

This article contains supplementary material, available at www.interscience.wiley.com/jpages/0022-3549/suppmat. Advanced Analytical Technologies, Inc. (Formerly CombiSep), 2711 South Loop Drive, Suite 4200, Ames, IA 50010. Franco Lombardo’s present address is Novartis Institutes for Biomedical Research, 250 Massachusetts Avenue, Cambridge, MA 02139. Correspondence to: Franco Lombardo (Telephone: 617-8714003; Fax: 617-871-3078.; E-mail: franco.lombardo@novartis. com) Journal of Pharmaceutical Sciences ß 2008 Wiley-Liss, Inc. and the American Pharmacists Association

disposition properties—absorption, distribution, metabolism and excretion (ADME).1 For example, our recent work on the prediction of volume of distribution in humans (VDss)2 has shown that the fraction of compound ionized at pH ¼ 7.4, together with the fraction of free drug in plasma are the largest contributors to the prediction of VDss, via the fraction unbound in tissues ( fut). In recent years high throughput drug discovery operations have brought to focus the need for the rapid evaluation of various physicochemical properties of newly synthesized compounds using minute quantities3–6 and significant efforts have been devoted to developing techniques adapted to these challenges. There have been a number of methods employed for pKa measurements based on solubility, potentiometric titration,7–10 spectrophotometry,11–14 HPLC15–17 and, most recently, JOURNAL OF PHARMACEUTICAL SCIENCES

1

2

SHALAEVA ET AL.

on capillary electrophoresis (CE).18–33 The solubility method is of limited accuracy and potentiometric titrations typically require mg amounts of pure sample with a throughput of 20–40 min per compound.34 In addition, the sample concentrations required for potentiometry often result in precipitation of poorly soluble compounds, necessitating mixed solvent extrapolation methods and further impacting sample throughput. A rapid method for pKa measurement employing a mixed-buffer linear pH gradient and spectrophotometric detection was recently demonstrated.35 The spectral gradient analysis (SGA) technique performs pKa measurements sequentially from a 96-well plate format using 10 mM DMSO stock solutions and samples are assayed in about 4 min. Compounds, however, must possess good chromophores close enough to the center of ionization for the detection of a spectral shift with changing pH; otherwise, the pKa values may go undetected. Precipitation of low solubility compounds is also a limitation of the method, although cosolvents could be employed, but data analysis often requires user intervention and sample impurities, degradants or counter ions possessing similar UV characteristics may also potentially interfere with the measurement. The application of CE for medium to high throughput pKa measurements in support of drug discovery projects has steadily increased in recent years. The many advantages of CE for pKa measurement were highlighted in several recent reviews34,36–38 including: minimal sample requirement with extremely low sample consumption, the ability to separate impurities and/or degradants from the target compound, sensitive on-line UV detection, and automated operation. In addition, precise knowledge of sample concentration is not required because only compound migration times are required for analysis, and no special demands are placed upon the purity of buffer solutions. The effective mobility (meff) of a solute, in a field of voltage V, is easily calculated from the difference in migration time between the compound (ta) and a neutral marker (tm), commonly DMSO, through Eq. (1):    Ld Lt 1 1 meff ¼
(1) ta tm V and this technique is based on the variation of the solute mobility with the variation of buffer pH. In Eq. (1) Ld is the length from the capillary inlet to the detection window and Lt is the total JOURNAL OF PHARMACEUTICAL SCIENCES

capillary length in cm, and V is the applied voltage. A plot of compound meff as a function of pH yields a sigmoidal curve, from which the inflection point(s) corresponds to the apparent pKa value(s) and thus the mobility is correlated with the pKa. Equations relating the measured meff and pH values to the pKa value via nonlinear regression analysis have been previously well described in the literature for up to three ionizable groups, which covers the majority of pharmaceutical compounds.22,28 Theoretically, there is no restriction on the number of ionization equilibria that can be considered.34 In Pharmaceutical Discovery operations it is often of interest to discriminate differences in pKa values among structural analogs and it is not unusual to encounter molecules with two or more very close pKas. For the method to be suitable for the simultaneous measurement of the variety of possible pKa combinations, in a high throughput mode, it is necessary to have buffers spaced with relatively small increments and also covering as wide a pH range as possible. For example, Ishihama et al.22 demonstrated the successful measurement of up to six pKa values for angiotensin by CE using 19 different buffers from pH 1.8 to 12.0. One approach to increasing sample throughput involves the application of multiplexed capillary array electrophoresis with UV absorption detection.39,40 Recently, the use of a 96-capillary array for the rapid measurement of pKa values for 96 different compounds (mostly monoprotic acids and bases) was demonstrated.33 A measurement of 128–168 compounds in an 8 h period was achieved by simultaneously analyzing eight compounds over 12 pH values in a single CE run. A significant challenge for all of the aforementioned pKa measurement techniques including CE is the precipitation of low aqueous solubility compounds. Unfortunately, the current trend in drug discovery is toward compounds possessing a relatively high lipophilicity and a fairly low aqueous solubility.41 Detection limits using UV spectrophotometry are reported to be 1–10 mM as long as the compound possesses a suitable pH-sensitive chromophore,5 while CE methods employing low UV wavelength detection allow measurements at similar concentrations without restrictions regarding the position of the chromophore.22 A potential approach for increasing the detection sensitivity and expanding the scope of detection involves the integration of mass spectroDOI 10.1002/jps

MULTIPLEXED CE pKa DETERMINATION

metry with CE (CE-MS).42,43 Wan et al.43 measured the pKa values for 60 different compounds including some that were sparingly soluble or of low UV absorbance by CE-MS, and sample throughput was increased by pooling up to 56 compounds. Although promising, the day-to-day reliability and reproducibility of the method remains to be assessed and commercially available software for performing mass deconvolution and data processing would need to be developed to enable routine pKa measurement. A common approach for the pKa measurement of aqueous insoluble compounds involves the use of different percentages of mixed cosolvent buffers and extrapolation to 0% cosolvent. Typically, at least three and up to six different percentages of cosolvent are recommended. Methanol is the most commonly employed cosolvent and is considered to yield the least deviation from a completely aqueous environment as its effects on pKa have been extensively studied.9,44,45 When working in mixed methanol/water solutions two different pH scales are commonly used, depending upon how the pH meter is calibrated prior to pH measurement. The notation describing the different pH scales used within this work is that recommended by IUPAC for pH quantities46 and that employed by others.45,47–49 The lowercase left-hand superscript indicates the solvent (w for water or s for mixed solvent) in which the measurement is made, while the lower-case lefthand subscript indicates the solvent where the ionic activity coefficient is referred to unity at infinite dilution (i.e., how the meter was calibrated). Therefore, the measurement of pH values following calibration of the pH meter with standard buffers prepared, in the same methanol/water composition as the sample, yields ss pH values. If the pH meter is calibrated with aqueous standard buffers, measurements in the mixed methanol/water solutions are in the intersolvental scale and referred to as sw pH values. It has been rigorously demonstrated in the literature that the two pH scales are related by means of Eq. (2): s s pH

¼ sw pH þ d

(2)

where d is a constant that is dependent strictly upon the composition of the mixed solvent.47 Eq. (2) assumes that the liquid junction potential of the potentiometric system used for pH measurement is negligible, as was previously demonstrated for the same type of combination glass electrode used in this work.47 Values for d have DOI 10.1002/jps

3

been previously determined over a wide range of methanol/water compositions.48 Measurements of meff as a function of sw pH value using CE yield a compound’s apparent sw pKa value(s) for that particular solvent composition and ionic strength. The sw pKa value can be converted to a ss pKa value via Eq. (3): s s pKa

¼ sw pKa þ d

(3)

Alternatively, the ss pKa value can be obtained directly for a compound working in the ss pH scale by employing CE with cosolvent buffers prepared from equimolar mixtures of an acid and its corresponding salt50–52 or by using pH meter calibration standards with known ss pKa values (e.g., potassium hydrogenphthalate).53,54 However, use of the sw pH scale combined with Eqs. (2) and (3) provides a much more practical approach when working with mixed solvent systems. The most popular extrapolation method is the Yasuda–Shedlovsky (Y–S) method, which relates s s pKa to the inverse of the dielectric constant of the binary solvent (1/e) by Eq. (4): s s pKa

þ log½H2 O ¼

a þb "

(4)

where a and b are constants. Extrapolation to e ¼ 78.3 and log H2O ¼ 55.5 (the dielectric constant and molar concentration of pure water, respectively) yields the apparent w w pKa value. Previous studies have demonstrated that Y–S extrapolation generally yields a linear relationship and accurate w w pKa values when using solvent mixtures possessing e > 50, which for methanol corresponds to