Letter pubs.acs.org/ac
Measurement of Drug Lipophilicity and pKa Using Acoustics Xin Li† and Matthew A. Cooper*,‡ †
Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge, CB2 1EW, United Kingdom Institute for Molecular Bioscience, University of Queensland, Carmody Road, St Lucia, Old 4072, Australia
‡
S Supporting Information *
ABSTRACT: Lipophilicity of chemicals and drug candidates is normally described in terms of octanol/water partitioning and log P. We investigated an alternate approach to lipophilicity determination using a mimic of an alkyl alcohol with compound partitioning quantified using acoustic sensing. A self-assembled monolayer composed of HSC10(CH2CH2O)6C18 was formed on planar gold electrodes of a piezoelectric acoustic sensor. The system was challenged with compounds covering a 4-log range of log D values. As compounds partitioned in the interfacial layer, changes in sensor resonant frequency were found to correlate with compound partition coefficients (log P) and with distribution coefficients (log D). Linear concordance (R2 = 0.933) was established between log(−dF/Mwt) and log P and with log D in both water and biological buffers at variant pH (pH 5.2 to 7.8). In turn, drug pKa could be determined by profiling log D changes during pH titration. The lipophilicity/pH profile of a weakly basic drug (quinine; pKa = 7.95) was sigmoidal with respect to −dF/Mw values, with a profile inverse to that of a weakly acidic drug (naproxen; pKa = 4.15).
T
he partition coefficient, commonly referred to as P, of a small molecule drug is an important parameter in drug design and development.1 The logarithm form of partition coefficient, known as log P, has been widely recognized as a descriptor of the lipophilicity or hydrophobicity of a drug molecule. Hydrophobic interactions are one of the driving forces for binding between a drug candidate and its intended target molecule and, in part, define bioavailability in vivo.2 Thus, log P values have been extensively used in establishing quantitative structure−activity relationships (QSAR).3,4 Furthermore, drug absorption and distribution involve drug molecules being transported through several semipermeable cell membranes. For passive transport, a drug molecule must be lipophilic enough to partition into the lipid bilayer and be hydrophilic enough to partition back to the aqueous environment of the cell. Thus, log P is also a relevant parameter in prediction of drug passive transport across cell membranes. Measurement of the partition coefficient of a drug is carried out by adding the compound to a mixture of 1-octanol and water.1 Strong agitation is followed by settling and equilibrium partitioning after which various methods can be used to determine the concentration of the compound in the two immiscible phases, e.g., UV−vis spectroscopy, mass spectrometry, etc. The partition coefficient of the drug between 1octanol and water can then be defined as P or Po/w = [Coct]/ [Cwater]. This conventional shake-flask method is tedious and requires large amounts of compound. Other measurements, such as high performance liquid chromatography (HPLC), are based on partitioning small molecule analyte between an aqueous liquid and a solid surface.5−8 The lipophilicity of a © 2012 American Chemical Society
drug can be quantified by correlating the retention time on a HPLC column under certain conditions with similar compounds of known log P values,9−13 most often with octanol in the mobile phase.14,15 It has been reported that small molecule analytes can be observed to partition into immobilized liposomes or lipid multilayers on a biosensor chip surface, to generate signal changes directly reflecting the lipophilicity of the molecule. Danelian et al.16 reported a surface plasmon resonance (SPR) biosensor method attempting to build the correlation between the fraction of a drug absorbed from human intestine (Fa) and the quantity of the drug partitioned into the sensing surface immobilized liposome of POPC/GM1 (palmitory-oleoylphosphatidyl-choline/monosialo-tetrahexosyl-ganglioside). However, outliers were identified attributing to sensitivity of the sensor with low molecular weight analytes. In addition, optical biosensors are sensitive to bulk signal shifts induced by solvent mismatch (arising from the variant dielectric or refractive indices of solvents). This compromises the accuracy and dynamic range of optical biosensor detection of molecules. In contrast, acoustic biosensors are sensitive to the square root of the differences in the viscosity−density product of liquids, which are comparatively small for solvents such as water, DMSO, and octanol. Okahata et al. developed a quartz crystal microbalance (QCM) biosensor based method to explore the relationship between the intrinsic potency of bioactive Received: January 10, 2012 Accepted: February 29, 2012 Published: March 1, 2012 2609
dx.doi.org/10.1021/ac300087z | Anal. Chem. 2012, 84, 2609−2613
Analytical Chemistry
Letter
compounds, such as bitter substances, odorants,17,18 or local anesthetics,19 and the partition coefficient of these compounds in the sensing surface immobilized lipid multilayers of 2C18N+2C1/PSS−. However, no studies have been reported that correlate the partitioning of small molecule compounds into surface-immobilized liposomes or lipid multilayers with conventional log Po/w. Furthermore, liposome and lipid multilayers are neither easily nor rapidly fabricated in a highly reproducible manner to attain a reliable correlation between the surface partition coefficient and standard log Po/w. Herein, we describe a self-assembled monolayer (SAM) sensor surface of HSC10EG6OC18(EG = OCH2CH2) fabricated on gold-coated 17 MHz quartz sensor chip. The resultant SAM coating, which can be characterized by ellipsometry and cyclic voltammetry in a batch production process, represents a welldefined surface architecture comprising three distinct layers (Figure 1). A cutoff using cyclic voltammetry area under the
Figure 2. The microfluidic manifold that enables controlled delivery of drug solutions with minimal dispersion to the piezoelectric sensor cassette (right) consisting of two high frequency piezoelectric sensors mounted in a stress-free manner within an acrylic cassette.
crystal microbalance (QCM), whereby high-frequency alternating voltage was applied to a piezoelectric crystal inducing the crystal to resonate, and the resonance frequency was then monitored in real time using a proprietary network analyzer. Such sensors have found application in the study of ligand− receptor interactions,23 drug-target affinity determination,22,24 DNA hybridization,20 and bacteria detection.25 In the current application, 10 marketed drugs (Table 1) were assayed in a background running fluid of either degassed MilliTable 1. Drug Parameters and Log(−dF/Mw) Detected in MQ Water Figure 1. Amphipathic self-assembled monolayer on a gold surface with envisaged partitioning of a drug in the interfacial layer.
curve of 50 sensors a very consistent level of response for a given drug was observed with variance coefficients of 4% over multiple instruments and 3% when a single instrument was used. For an individual sensor, we evaluated 10 drug solutions with assays replicated 6 times. Between drug solutions, a detergent solution was used to clean the sensor surface. This cycle of assays could be repeated at least 60 times. For 60 repeat measurements, the %CV of the raw dF signal and the calculated cLog P was less than 3%. We did not see deterioration of assay performance during experiments within the pH range of 3−12. The instrument detection limit (IDL) of the machine was 0.3 Hz at the fundamental resonant frequency used, which leads to a lower limit of quantification (LLOQ) in the 10 μM range for this particular application. We have exemplified a 4-log dynamic range for log P between −1 and 3 with this data set. As the assay was conducted in a continuous flow mode and the sensing area was relatively small (less than 0.12 cm2), the change of drug concentration in bulk solution can be assumed to be negligible relative to the amount of drug partitioning at the sensor surface. The partition coefficient can hence be analyzed in terms of total drug molecules present in the interfacial coating matrix vs those in bulk solution, which according to the Sauerbrey equation26 is directly correlated to the frequency change of the resonator. The frequency change of
⎛
Δms Δf = Δfm + Δfl = −f02 ⎜ + ⎜ Fq ρ A el ⎝ q
⎞ ⎟ f0 πμqρq ⎟⎠ ηl ρ l
(3)
Thus, the acoustic signal shift depends on both the properties of the solution phase (in particular, liquid viscosity and density) and the degree of “added mass” at the sensor interface, related to putative compound partitioning. Added mass is quadratic with frequency shift, while the viscosity−density product is inverse quadratic (square root). Hence, there is a proportionally large signal potentially attributable to partitioning compared to solution effects resulting from compound log P or changing ionic state during a pH titration. At this stage, it is not possible to accurately assign which component contributes to the observed signal changes; however, note that surfaces of lower hydrophobicity (e.g., a C8 SAM (HSC10EG6OC8) vs a C18 SAM (HSC10EG6OC18)) gave rise to a proportionally lower signal when exposed to the same compound (data not shown), suggesting surface partitioning was primarily responsible for the observed signal. We also note that with molecule weights of the compounds tested within the range between 157 and 392, at a concentration of 500 μM, we are dealing with solutions of 0.02% w/v or less; such dilute solutions would not give rise to significant changes in density nor viscosity. Considering that the Sauerbrey equation is defined in units of mass sensitivity, but the partition coefficient P is in terms of the ratio of the number of drug molecules in 1-octanol phase and that in aqueous phase, we normalized the frequency changes against the molecular weights of the drug: −dF/Mw. Note that, in a compound library profiling scenario, the compound molecular weight would be known. A linear relationship with reasonable concordance (R2 = 0.93 across all log P and log D values) between log(−dF/Mw) and log P or log D was established over a variety of solutions at differing pH and water (Figure 3). 2611
dx.doi.org/10.1021/ac300087z | Anal. Chem. 2012, 84, 2609−2613
Analytical Chemistry
Letter
Strictly speaking, the partition coefficient P is defined as the ratio of concentration of neutral species in 1-octanol phase divided by the concentration of neutral species in water phase. In fact, many drugs ionize in aqueous solution, existing in a distribution state of neutral and charged species depending on the ionizing capability of the molecule (pKa) and the surrounding pH. Each portion of the human body, especially along the digestive tract, has a different environmental pH; normal blood is tightly regulated around pH 7.4, saliva ranges from pH 6.0 to 7.4, stomach is pH 2.0−3.0, duodenum is pH 6.0−6.5, small intestine is pH 7.8, and large intestine is pH 5.5−7.0. Representative lipophilicity of a molecule hence is determined by the population of neutral and ionized species that may exist at a given pH, a parameter commonly reflected by the distribution coefficient, referred to as D. Log D is hence defined as the ratio of the sum of all species (neutral and ionized) in 1-octanol divided by the concentration of all species in aqueous phase at a given pH. In this study, the 10 selected drugs were assayed in water and then at pH 7.8 (representing small intestine environment and blood plasma), pH 6.8 (duodenum), and pH 5.2 (large intestine). A linear relationship between log(−dF/Mw) and log D maintained good concordance at all three pHs over 4 orders of magnitude of log D from −1 to 3 (Figure.3). The optimal log D range for orally available drugs is generally accepted to lie between 0 and 3, as these compounds possess an optimal balance of solubility and permeability. Compounds with log D values less than 0, although highly soluble, are too hydrophilic to efficiently permeate cell membranes or penetrate the blood brain barrier in the absence of an active transport mechanism. Whereas compounds with log D values larger than 5 normally exhibit poor water solubility and high plasma protein binding capability, both can significantly reduce drug bioavailability. The lipophilicity pH profile of a drug, also known as pH-profile of log D achieved by titrating its log D values in a pH ladder, can provide not only an overview of lipophilicity change of a drug over the whole pH range but also a good indication about ionizing tendency of drug molecules, which is normally expressed by the aqueous ionization constant pKa. It is believed that aqueous solubility of drugs is pKa dependent. Thus, lipophilicity, pKa, and aqueous solubility are three key principal physicochemical parameters to be considered during compound profiling. Figure 4 shows the lipophilicity pH profiles of naproxen and quinine from pH 2 to 12 determined with acoustic sensing on the HSC10EG6OC18 SAM coated sensor chip. Molecules are normally more lipophilic when neutral than when charged; hence, the acidic drug naproxen changed from a nonionized lipophilic state to an ionized hydrophilic state as the pH was increased. The pKa value 4.15 of naproxen, at which 50% of
naproxen molecules are in deprotonated state, lies soon after the molecular weight normalized frequency signal changes (−dF/Mw) started to reduce. In contrast, the basic drug quinine experienced the reverse trend from an ionized hydrophilic state at low pH to a nonionized lipophilic state at high pH, with the pKa 7.95 lying near −dF/Mw inflection to plateau. These characteristics of lipophilicity pH profiles are in good agreement with those found in online databases of experimentally determined results (e.g., Sirius Analytical http://www.sirius-analytical.com/science/logP.shtml). In conclusion, we demonstrate the utilization of an acoustic piezoelectric sensing biosensor platform and an amphiphilic SAM surface to determine drug log P and log D values within the range of parameters of interest for “lead like” and “drug like” compounds. As is the case with HPLC and UPLC methods, the technique is non-destructive, label-free, rapid, and automated. It may provide an additional tool to facilitate the estimation of drug physicochemical properties in compound library profiling, screening samples, and hit to lead programs.
■
ASSOCIATED CONTENT
S Supporting Information *
1 H NMR spectrum of HSC10EG6OC18, details on the RAP-id 4 instrument and protocols, fabrication of HSC10EG6OC18 SAM sensor chips, detailed assay procedures; additional more log D correlation results at varying pH from 5.2 to 7.8. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected] (M.A.C.). Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS The authors would like to express their gratitude to the NIAID (AI-061243) and NHMRC (AF-511105) for financial support and Dr. Reena Halai for abstract artwork.
■
REFERENCES
(1) Leo, A.; Hansch, C.; Elkins, D. Chem. Rev. (Washington, DC, U. S.) 1971, 71, 525616. (2) Miyamoto, S.; Kollman, P. A. Proc. Natl. Acad. Sci. U. S. A. 1993, 90, 8402−8406. (3) Scherrer, R. A.; Howard, S. M. J. Med. Chem. 1977, 20, 53−58. (4) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J. Adv. Drug Delivery Rev. 1997, 23, 3−25. (5) Mirrlees, M. S.; Moulton, S. J.; Murphy, C. T.; Taylor, P. J. J. Med. Chem. 1976, 19, 615−619. (6) Minick, D. J.; Frenz, J. H.; Patrick, M. A.; Brent, D. A. J. Med. Chem. 1988, 31, 1923−1933. (7) Tomlinson, E. J. Pharm. Sci. 1982, 71, 602−604. (8) Faller, B.; Grimm, H. P.; Loeuillet-Ritzler, F.; Arnold, S.; Briand, X. J. Med. Chem. 2005, 48, 2571−2576. (9) Menges, R. A.; Armstrong, D. W. Anal. Chim. Acta 1991, 255, 157−162. (10) Balogh, G. T.; Szántó, Z.; Forrai, E.; Győ rffy, W.; Lopata, A. J. Pharm. Biomed. Anal. 2005, 39, 1057−1062. (11) Casoni, D.; Kot-Wasik, A.; Namiesnik, J. J. Chromatogr., A 2009, 1216, 2456−2465. (12) Guillot, A.; Henchoz, Y.; Moccand, C.; Guillarme, D.; Veuthey, J.-L.; Carrupt, P.-A.; Martel, S. Chem. Biodiversity 2009, 6, 1828−1836. (13) Pallicer, J. M.; Sales, J.; Rosés, M.; Ràfols, C.; Bosch, E. J. Chromatogr., A 2011, 1218, 6356−6368.
Figure 4. Lipophilicity pH profiles of naproxen (left) and quinine (right), each data point is an average of six repeats. 2612
dx.doi.org/10.1021/ac300087z | Anal. Chem. 2012, 84, 2609−2613
Analytical Chemistry
Letter
(14) Benhaim, D.; Grushka, E. J. Chromatogr., A 2008, 1209, 111− 119. (15) Liu, X.; Tanaka, H.; Yamauchi, A. J. Chromatogr., A 2005, 1091, 51−59. (16) Danelian, E.; Karlen, A.; Karlsson, R.; Winiwarter, S.; Hansson, A.; Lofas, S.; Lennernas, H.; Hamalainen, M. D. J. Med. Chem. 2000, 43, 2083−2086. (17) Okahata, Y.; Ye, X.; Shimizu, A.; Ebato, H. Thin Solid Films 1989, 180, 51−59. (18) Okahata, Y.; Ebato, H. TrAC, Trends Anal. Chem. 1992, 11, 344−354. (19) Okahata, Y.; Ebato, H. Anal. Chem. 1991, 63, 203−207. (20) Wink, T.; vanZuilen, S. J.; Bult, A.; vanBennekom, W. P. Analyst 1997, 122, R43−R50. (21) Cooper, M. A. Drug Discovery Today 2006, 11, 1068−1074. (22) Godber, B.; Thompson, K. S. J.; Rehak, M.; Uludag, Y.; Kelling, S.; Sleptsov, A.; Frogley, M.; Wiehler, K.; Whalen, C.; Cooper, M. A. Clin. Chem. (Washington, DC, U. S.) 2005, 51, 1962−1972. (23) Li, X.; Abell, C.; Cooper, M. A. Colloids Surf., B 2008, 61, 113− 117. (24) Li, X.; Thompson, K. S. J.; Godber, B.; Cooper, M. A. Assay Drug Dev. Technol. 2006, 4, 565−573. (25) Huang, L.; Cooper, M. A. Clin. Chem. (Washington, DC, U. S.) 2006, 52, 2148−2151. (26) Sauerbrey, G. Z. Phys. 1959, 155, 206−212.
2613
dx.doi.org/10.1021/ac300087z | Anal. Chem. 2012, 84, 2609−2613
