Anal. Chem. 1995,67, 755-762
Membrane Partition Coefficients Chromatographically Measured Using Immobilized Artificial Membrane Surfaces Shaowei Ong,t Hanlan Liu,t Xiaoxing Qiu,+Ganapati Bhat,* and Charles Pidgeon*st
Department of Medicinal Chemistty, School of Pharmacy, Purdue University, West Lafayette, Indiana 47907, and Regis Technologies Inc., 8210 Austin Avenue, Morton Grove, Illinois 60053
Immobilized artificial membranes (IAMs)are chromatographic surfaces prepared by covalently immobilizing cell membrane phospholipids. IAM surfaces mimic fluid cell membranes. Solute capacity factors ( k ’ w )measured on IAM columns correlate very well with solute equilibrium partition coefficients (Km’) measured in fluid liposome systems. For 23 structurally unrelated compounds, log ( k ’ m ) correlates with lOg(Km’) with a linear correlation coefEcientr = 0.907. This indicates that solute partitioning between the IAM bonded phase and the aqueous mobile phase is similar to the solute partitioning between liposomes and the aqueous phase. Although both IAM chromatography and liposome partitioning can be used as in vitro methods to predict solute partitioning into cell membranes, IAM chromatography is experimentally convenient compared to liposome systems. To study the effect of lipid structure on drug binding to IAMs, LAMS were prepared from three different phosphatidylcholine ligands: (i) a diacylated phosphatidylcholine ligand, (ii) a single chain ether phosphatidylcholine ligand, and (iii) a single chain phosphatidylcholine ligand that lacks a glycerol backbone. Solute retention data were identical for all of these IAMs, and consequently, predictions of solute binding to fluid membranes were also identical. This indicates that the structure of the phosphatidylcholine ligand that is immobilized is not critical for the binding of solutes. Since the structure is not important, the binding of solutes to membranes is a bulk phase property, Le., it is the interface created by the ligands that determines the solute binding properties, not the ligands themselves. Solute partitioning using OctanoVwater systems does not correlate with k ’ m unless a homologous series of hydrophobic solutes is being evaluated. Solute partitioning between two phases is a general phenomenon. The two phases involved in the partitioning process may be two fluid phases, as in organic/aqueous solvent mixtures, or a suspension of particles where the solute partitions between the continuous phase of the solvent and the surface of the particles. Other partitioning systems exist but are not of interest in this work. When there is no net mass transfer between the two phases, solute distribution is at equilibrium, and the partition coefficient is calculated as the ratio of the solute concentration in each phase. The partition coefficient between membrane suspen+
Purdue University. Regis Technologies, Inc.
0003-2700/95/0367-0755$9.00/0 0 1995 American Chemical Society
sions and an aqueous phase is defined as the membrane partition coefficient (Kd. The membrane partition coefficient provides direct experimental information about the solute’s free energy of interaction (AGd with the fluid membrane,
AGm = -RT In (Km)
(1)
In addition, enthalpy (AH) and entropy (AS) of the solutemembrane interaction can be obtained by measurements of the membrane partition coefficient at various temperatures.’+ Membrane partition coefficients thus provide critical insight into solute-membrane interactions. One highly important observation is that K, values for various drugs correlate with the permeability (Pdof drug transport across membranes. The linear relationship between P, and K, is given by6
Pm= DmKm/L
(2)
where D, is the membrane diffusion coefficient of the solute and L is the thickness of the fluid bilayer membrane. Thus, the two most important parameters used to evaluate P,,, are D, and K,. The membrane permeability coefficient of a drug (or solute) is actually the linear velocity of drug movement through the membrane and therefore not only is of fundamental importance in membrane biology but also has practical significance in the pharmaceutical sciences in that P (cm/s) reflects the rate of drug transport across the membrane. Drug permeability is very difficult to measure, and therefore methods to estimate P based on measurements of K, and D, are needed. Because of the biological importance of K , , measurements of K, have been the subject of many studies for the last several decades. However, Km is difficult to measure in vivo, and therefore three in vitro membrane systems have been developed to model solute partitioning into membranes. The three models include simple organic solvent/aqueous partitioning systems,7-15 chro(1) Wimley, W. C.; White, S. H. Biochemistry 1993,32, 6307-6312. 12813-12818. (2) Wimley, W. C.; White, S. H. Biochemistry 1992,31, (3) Katz, Y.; Diamond, J. M.J. Membr. Biol. 1974,17,69-86. (4) Diamond, J. M.; Katz, Y.J. Membr. Biol. 1974,17,101-120. (5) Betageri, G. V.; Rogers, J. A Int. 1.Pharm. 1987,36,165-173. (6)Stein, W. D. Transport and Diffusion across Cell Membranes; Academic Press: Orlando, FL, 1986. (7) Meyer, H. Arch. Exp. Pathol. Pharmakol. 1899,42, 110. (8)Overton, E. Studien mer Die Narkose; Fisher: Jena, Germany, 1901. (9) Hansch, C.; Muri, R M.; Fujita, T.; Maloney, P. P.; Geiger, F.; Streich, M. /. Am. Chem. Soc. 1963,85, 2817-2824.
Analytical Chemistry, Vol. 67,No. 4, February 15, 1995 755
tion, numerous QSAR studies have successfully correlated drug matographic partitioning systems using octadecyl silica (ODS) as with drug liposome partition coefficient~.5>2~-28 a stationary p h a ~ e , ' ~and J ~ liposome partitioning s y ~ t e m s . ~ ~ ~ J activities ~J~ Simple organic solvent/aqueous partitioning systems have Although liposomes can model both polar and nonpolar solute-membrane binding interactions, this model has the limitabeen used as models for biological membranes since the turn of the century, when Meyer7 and Overtons showed that the narcotic tion that it is experimentally laborious. Liposome preparation, followed by solute equilibrium in the liposome suspension, activities of drugs parallel the drugs, oil/water partition coeffollowed by quantitation of the free solute in the presence of ficients. Among many organic solvent systems used to date, octanol/water has been the most often used model solvent liposomes and the correction for the amount of drug that has ~ystem.~-ljHowever, it is well recognized that simple organic partitioned into the aqueous space of the liposomes make the liposome method time consuming and tedious. If many comsolvent/water systems are good models for solute-membrane pounds are under evaluation, then it is very time consuming to partitioning only when polar group interactions between the solute and the phospholipid bilayer are minimal or absent.20 In other measure the equilibrium liposome binding constants for all of the compounds. To circumvent this experimental difficulty, our words, for a group of relatively hydrophobic compounds, octanol/ laboratory has developed a solid phase membrane m ~ d e P ~ of - ~ ~ water partitioning correlates well with solute partitioning into fluid membranes; however, if the compounds are polar, the correlations the fluid membrane found in liposomes. In this work we demonstrate that our model, which is an immobilized artificial are not good. In fact, because of this problem, Hansch et al. membrane, can substitute for liposomes as a rapid, simple, and suggested that octanol is a rational model solvent because it accurate method to determine the membrane partition coefficients models polar molecular interactions between solutes and memwith little experimental effort. b r a n e ~ . ~ - 'For ~ quantitative structure-activity relationships (QSARs),the equilibrium partition coefficient (KO,+) measured in Immobilized artificial membranes @AMs) are solid phase octanol/aqueous systems has been the standard in the field for membrane mimetics whereby cell membrane phospholipid molthe last few decades. This is in spite of the observation that ecules are covalently bonded to silica particles at high molecular logarithm of KO,,often exhibits poor correlations with the surface densities. IAMs are typically used as a chromatography logarithm of drug activities, even for a homologous series of stationary phases. IAMs were invented in this laboratory a few years a g ~ and drug^.^-'^ It is not surprising that KO,,or the partition coefficient ~ have - ~ been ~ used to purify membrane pr0teins,3~-3~ for other aqueous/organic phases does not model the subtle polar immobilize e n z y m e ~ , 3obtain ~ * ~ ~enzyme-ligand binding constants and nonpolar molecular interactions between solutes and memfor drugs,36 and obtain hydrophobic parameter^.^^^^^ We note that in this work we refer to drug binding to IAMs as a membrane branes, and therefore better models are needed. binding constant instead of a hydrophobic parameter as defined Chromatography with octadecyl silica (ODS) as a stationary by other^.^^,^^ phase has been used as an alternative method for determination Although IAMs used in this work contained only phosphatiof Koct.16J7 Although ODS chromatography is a relatively simple measurement compared to the octanol/buffer system, it inherits dylcholine (PC), we have prepared mixed lipid IAMs containing the most serious limitation of the octanol/buffer system in that it PC and a minor phospholipid constituent such as phosphatidyllacks structural similarities to biological membranes. Unlike ethanolamine @E), phosphatidylglycerol (PG),phosphatidylserine octanol, which contains a polar hydroxyl and a nonpolar hydrc(PS), or phosphatidic acid (PA).4O Drug partitioning into mixed carbon chain, ODS chromatography surfaces contain only nonlipid liposomes has recently been compared to the drug KW' polar hydrocarbon chains. Therefore, ODS chromatographycan measured on mixed lipid IAMs.*O model only the partitioning process associated with the hydroEXPERIMENTAL SECTION phobic components of lipid bilayers; the interactions between Chemicals. The following chemicals were purchased from solutes and the polar lipid head groups are not modeled. Sigma: xylometazoline, oxymetazoline, naphazoline, tetrahydroIn contrast to simple organic solvents or ODS chromatography (22) Jacobs, R E.; White, S. H. Biochemisty 1989,28, 3421-3437. surfaces, only liposome suspensionsprepared from phospholipids (23) De Young, L.; Dill, K. A.Biochemisty 1988,27, 5281-5289. exhibit structural similarities to the phospholipid bilayer found (24) Reig, F.; Busquets, M. A; Haro, I.; Rabanal, F.; Alsina, M. A]. Pharm. Sci. in cell membranes.21 Consequently, immediately after liposomes 1992,81,546-550. (25) Betageri, G. V.; Rogers, J. A. Pharm. Res. 1989,6,399-403. were developed, Katz and Diamond quickly established that solute (26) Betageri, G. V.; Rogers, J. A Pharm. Res. 1993,10, 913-917. partition coefficients measured in aqueous liposome suspensions (27) Choi, Y. W.; Rogers, J. A Pharm. Res. 1990,7,508-512. (28) Castelli, F.; Rosario, P.; Sarpietro, M. G.; Mazzone, P.; Raciti, G.; Mazzone, were similar to the partition coefficients measured in endogenous G . J. Pharm. Sci. 1994,83,362-366. membranes (Kd.3,43J9 Liposome partition coefficients have been ,229) Pidgeon, C. U S . Patent 4,927,879, 1990. used to study many solute-membrane intera~tions.2~~~~ In addi(30) Pidgeon, C. U S . Patent 4,931,498, 1990. (10) Hansch, C.; Fujita, T. J. Am. Chem. SOC.1964,86,1616-1626. (11) Hansch, C. Acc. Chem. Res. 1969,2, 232-239. (12) Hansch, C.; Dunn, W. J. J. Pharm. Sci. 1972,61,1-19. (13) Hansch, C.; Clayton, J. M. /. Pharm. Sci. 1973,62, 1-21. (14) Leo, A; Hansch, C.; Elkins, k Chem. Rev. 1971,71,525-605. (15) Smith, R N.; Hansch, C.; Ames, M. M.J. Pharm. Sci. 1975,64,599-606. (16) Mirrlees, M. S.; Moulton, S. J.; Murphy, C. T.; Taylor, P. J. ]. Med. Chem. 1976,19,288-296. (17) Schmidt, D.; Votaw, J. H.; Kessler, R M.; De Paulis, T. J. Pharm. Sci. 1994, 83,305-315. (18) Katz, Y.; Diamond, J. M. 1 .Membr, Bid. 1974,17,67-86. (19) Diamond, J. M.; Katz, Y. J. Membr. B i d . 1974,17,101-120. (20) Rogers, J. A.; Wong, A. Int. J. Pharm. 1980,6, 339-348. (21) Bangham, A. D. Prog. Biophys. Mol. Bid. 1968,18,19.
756 Analytical Chemistry, Vol. 67, No. 4, February 15, 1995
(31) Pidgeon, C.; Venkatamm, U. V. Anal. Biochem. 1989,176,36-47. (32) Pidgeon, C.; Stevens, J.; Otto, S.; Jefcoate, C.; Marcus, C. Anal. Biochem. 1991,194,163-173. (33) Thumhofer, H.; Schnabel, J.; Betz, M.; Lipka, G.; Pidgeon, C.; Hauser, H. Biochim. Biophys. Acta 1991,1064,275-286. (34) Otto, S.; Marcus, C.; Pidgeon, C.; Jefcoate, C. Endocrinology 1991,229(2), 970- 982. (35) Pidgeon, C.; Cai, S.; Bemal, C. Anal. Biochem., submitted. (36) Chui, W. IC;Wainer, I. Anal. Biochem. 1992,201, 237-245. (37) Zhang, X.-M.; Wainer, I. W. Tetrahedron Lett. 1993,34,4731-4734. (38) Kallury, K. M. R; Lee, W. E.; Thompson, M. Anal. Chem. 1992,64,10621068. (39) Kallury, K M. R; Lee, W. E.; Thompson, M. Anal. Chem. 1992,64,10621068. (40) Pidgeon, C.;Ong, S.; Choi, H.; Liu, H. Anal. Chem. 1994,66,2701-2709.
-
Scheme 1. Synthetic Route for Preparing salAM.PCc10*3
0
.
1 POCI, Et3N
II
C H3 OC (C &)IOCH20H
0-
b
b
(CH&N+C H2C H2 0 OC 4 (C 4
2. choline OTs, py 3. H2O
1
0
0-
0 I
II
II 1 7 OC0C H3
KOH
C H30H:H20
0 2
1. CDI , CHC13
IAM.Pe'O/C3
0-
I
0
II
(CH3)3N C H2CH20 0C y (C I+),oC0 H _____) (CH,)3 N+CH2C H20 POC y(C H2),oC-N H m S i l i c a II 2. SPA 0 3. c10 & C3 endcapping +
!
3
zoline, clonidine, propranolol hydrochloride, alprenolol, oxprenolol, metoprolol, pindolol, nadolol, atenolol, tolazoline, and choline p-toluenesulfonatesalt. The following drugs were kindly provided by Boehringer-Ingelheim: STH2224, STH2100, ST606, ST476, sT585, ST590, ST608, ST475, ST603, ST600, and tramazoline. Phenethylamjne derivatives, including MESC, ESC, PROSC, ISOPROSC, BROSC, 2C-T, DOT, DON, DOB, DOM, DOET, DOPR, DOBU, and DOAM, were kindly provided by Dr. D. Nichols of the Department of Medicinal Chemistry at Purdue University. Phosphorus oxychloride and alcohol-free CHC4 were purchased from Aldrich. Silica propylamine (SPA) was kindly provided by Regis Technologies Inc. Synthesis of IAM Surfaces. For the study of the effect of lipid structure on drug binding to IAMs, IAMs were prepared from three different phosphatidylcholine ligands: PC/ligand-1, a diacylated phosphatidylcholine ligand; PC/ligand-2, a single chain ether phosphatidylcholine ligand; and PC/ligand-3, a single chain phosphatidylcholine ligand that lacks a glycerol backbone. Detailed synthetic procedures have been described for PC/ligand1,31 and the IAM surface is denoted as eskrIAM.PC. The sythesis and bonding of PC/ligand-2 was described?l and the IAM surfaces are denoted as etherIAM.PCc10/c3.PC/ligand-3, which has the glycerol backbone removed, has not been prepared prior to this work. PC/ligand3 is denoted as 6cPC,where the superscript 6G denotes the deletion of the glycerol backbone. Scheme 1shows (i) the synthetic pathway to prepare the desired single chain 6GPC ligand (3)and (ii) the immobilization strategy used to tether the 6GPC ligand to SPA to prepare 6GIAM.PCc10/c3.Methyl 12hydroxydodecanoate (1)was synthesized as d e s ~ r i b e dand , ~ ~the syntheses of the remaining compounds (2 and 3) are described below. Synthesis of 1l-(MethoxycarbonyI)undecylphosphocholine (2). To a solution of triethylamine (6.85 mL, 0.049 mol) and phosphorus oxychloride (4.68 mL, 0.049 mol) in 120 mL of alcoholfree CHC13 cooled to -10 "C in an ice bath was added a solution containing 9.12 g of 1 (0.039 mol) in 100 mL of alcohol-freeCHC1, slowly over 30 min. The reaction mixture was then warmed to room temperature. After the mixture was stirred for another 30 min, choline p-toluenesulfonate salt (16.35 g, 0.039 mol) and pyridine (32.0 mL, 0.396 mol) were added. The reaction mixture was then stirred for 20 h at room temperature. The reaction was quenched with 1.0 mL of water and stirred for 30 min. The solvent (41) Rhee, D.; Markovich, R.; Chae, W. G.; Qiu, X.; Pidgeon, C. Anal. Chim. Acta 1994,297, 377-386. (42) Qiu, X.; Ong, S.; Bemal, C.; Rhee, D.; Pidgeon, C.J. Org. Chem. 1994,59, 537-543.
'7A M.PCc10'c3 was removed under vacuum, and the residue was dissolved in 100 mL of dichloromethane-toluene (1:l v/v), filtered, and concentrated under vacuum. The residue was dissolved in 60 mL of THF-HZO (91 v/v) and chromatographed on an AG 501-X8 D column using THF-H20 (9:l v/v). The fractions containing 2 were pooled, and the solvent was evaporated to concentrate the crude product, which was relatively pure and was used in the next reaction without further purification. Synthesis of 11-Carboxylundecylphosphocholine(3). A MeOH/H20 (486.8 v/v) solution of 2 (3.5 g) and KOH (2.988 g) was stirred at room temperature for 2 h. At this time, 5 mL of HzO was added to the reaction mixture, followed by acidification to pH 3 using formic acid. This mixed solvent of the reaction was dried under vacuum until -5 mL of HzO remained. The crude product was desalted on a C18 reversed phase column by loading the acidified product on the C18 column, washing with 500 mL of HzO, and then eluting the product with MeOH. Fractions (-250 mL) were collected and checked for phosphate using phospray; the column was eluted with MeOH until no phospray positive spots were detected by Tu=.After the solvent was removed under vacuum, 1.50 g of 3 was obtained as a white solid. The procedure for immobilization of the 6cPCocarboxyl ligand (3) on SPA, followed by C10 and C3 end-capping to form 6cIAM.PCc10/c3 (4), was exactly the same as that for the synthesis of etherIAM.PC10/C3 described previously>' LAM Chromatography and ODS Chromatography. All HPLC columns containing IAM stationary phases were packed at Regis Technologies Inc. Analytical size HPLC columns were 15 cm x 0.46 cm and have a void volume 6of -1.85 mL. For all studies, the injection volume was -15 pL of a solute aqueous solution (-1 ,ug/pL) in 0.01 M phosphate buffered saline (PBS), buffered at pH 7.4. The flow rate was 2 mL/min, and solute detection was at 220 nm. Chromatograms were obtained using a Rainin HPLC pumping system equipped with a Knauer Model 87 detector and interfaced with a Macintosh computer. Rainin Dynamax software was used to record the chromatograms on the computer. The retention times (tr)of solute molecules on IAM chromatography columns were used to calculate the solute capacity factors (k'dusing the following equation:
where tr is the retention time in minutes of the test compound and to corresponds to the column dead time or void volume. In Analytical Chemistry, Vol. 67, No. 4, February 15, 7995
757
3.5 1
2 M
s
3.0
-
2.5
-
Table 1. Structures of Compounds Used To Compare Solute Binding to Liposomes with Solute Binding to I A M Columns
/
0
P-Blockers &cHzcH=cHz
2.0-
4
X = OCH2CH(OH)CH2NHCH(CH3)2 &ocHzcH=cHz
1.51.00.5
F
CHpCHzCOCH3
r = 0.907 slope = 0.994 intercept = 0.496
a
-
propranolol
alpranolol
oxprenolol
metoprolol
OCHzCH(OH)CHzNHC(CH3)3
0.0 , 4q 0.0
0.5
1.0
1.5
2.0
HO
u
2.5
n
log (k’iAM ) Figure 1. Correlation of solute binding to DMPC liposomes [log(Km’)]with solute binding etherlAM.PCC10’C3 surfaces [ ~ ~ ~ ( K I A forM ) ] seven P-blockers (0),six imidazoline derivatives (O), and 10 imidazolidine derivatives (A). The liposome partition coefficients of these 23 solutes were measured using DMPC liposomes dispersed in 0.01 M PBS buffer (pH 7.4).5,26,27IAM capacity factors, KIAM,were measured on a 15 cm x 0.46 cm etherlAM.PCC10’C3 column using a mobile phase of 0.01 M PBS (pH 7.4).
pindolol
atenolol
nadolol
N H
oxymetazoline
our lab, solute retention times exhibit a day to day variation of less than 4%. A 15 cm x 0.46 cm IAM column typically contains about 1.2 g of IAM packing materials with -80 mg of immobilized phospholipid.
< ’3 N
Imidazoline Derivatives X =
xylometazoline
tramazoline
gCHz-’ 8’ q
naphazoline
N
=
X
clonidine
tettyzoline
RESULTS AND DISCUSSION
The capacity factor, k’, is linearly related to the equilibrium partition coefficient, K, of a solute that partitions between the stationary phase and the mobile phase:
where V ,is the total volume of solvent within the HPLC column and V, is the volume of the IAM interphase created by the immobilized pho~pholipids.4~ The phase ratio VJV, in eq 4 is a constant for a given IAM column, and therefore it is not necessary to make the difficult experimental measurement of V,. In addition, since VJV, is a constant, k’ is directly proportional to K. Thus, for IAM chromatography, eq 4 can be rewritten as
k’Kv = (VJVJKM
(5)
where k’m is the solute capacity factor on the IAM column and K m is the IAM partition coefficient of the solute on the IAM column. It is clear from eq 4 that by measuring k’m one can determine the solute partition coefficient in the IAM interphase; the IAM interphase is physically an immobilized liquid formed by the conglomerate of the bonded lipids. The key concept in using IAM chromatography to predict partitioning of solutes into fluid membranes is that IAMs are physically similar to, and therefore mimic, fluid phospholipid bilayers. IAMs have monolayer surface densities of immobilized phospholipids similar to those of liposome membra ne^.^^^^^ IAMs (43) Snyder, L. R.; Kirkland, J. J. Introduction to Modern Liquid Chromatography; John Wiley & Sons: New York, 1979. (44) Ong, S.; Cai, S. J.; Bernal, C.; Rhee, D.; Qiu, X.; Pidgeon, C. Anal. Chem. 1994,66,782-792.
758 Analytical Chemistry, Vol. 67, No. 4, February 15, 7995
Imidazolidine Derivatives
tpax
STH2224 (X = 2,3-di-Br) STH2100 (X = 2-CI, 3-Br) ST606 (X = 2-CH3, 3-Br) ST476 (X = 2,3-di-CI) ST565 (X = 2-CH3, 5-Cl) ST590 (X = 2-CH3, 4-Br) ST608 (X = 2-CL 3-CH3) ST475 (X = 2,5-di-CI) ST603 ( x = 2-cI, 5-Br) ST600 (X = 2-CH3,5-F)
also exhibit interfacial motional properties similar to the motional properties of the mobile lipids in fluid liposomes, as revealed by our recent 31P NMR s t ~ d i e s . 4 ~These 3 ~ ~ physical similarities between IAMs and fluid membranes encouraged us to correlate IAM partition coefficients, KIAM,with the equilibrium membrane partition coefficient, Km, of solute binding to liposome membranes. Although K m can be measured by equilibrium binding studies using loose IAM powders suspended in aqueous media containing solute we measured K k v in HPLC columns because K m is linearly proportional to k’m according to eq 5. To validate the idea that IAM chromatography can be used to measure solute-membrane partition coefficients, k’m values of 23 solutes were compared to the liposome partition coefficients, Km’, measured by Rogers5aZ6 using dmyristoylphosphatidylcholine @MPC) liposomes Figure 1). The structures of these 23 solutes are given in Table 1. It is clear from Table 1that these 23 solutes are structurally unrelated molecules, and therefore many different solute-membrane interactions are involved in the solutemembrane partitioning process. As shown in Figure 1,Iog(k’m) vs log(K,? exhibits an excellent linear correlation, with r = 0.907. It should be noted that the slope of the plot of log(k’lAM) vs log(K”) is very close to 1 (slope = 0.994),which indicates that k’m is linearly proportional to K,’. (45) Qiu, X.; Pidgeon, C. J. Phys. Chem. 1993,97, 12399-12407. (46) Ong, S.; Qiu, X.; Pidgeon, C. J. Phys. Chem. 1994,98, 10189-10199
r = 0.883 slope = 1.014 intercept = -0.345
r = 0.957 slope = 0.978 intercept = -1.021
sM
0.0
1
-0.5
-0.5
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.0
log (K,,,’)
0.5
slope = 0.783
0
I
A
I
1.5
2.0
2.5
r = 0.419 sloDe = 0.743
2 . 0 ~ r=0.483
- 1 .o 7
1.0
log ( ~ ’ I A M )
I
1
I
r
-1.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.0
1% (&It)
a
I
I
1
I
1
0.5
1.0
1.5
2.0
2.5
log @‘,AM)
Figure 2. Comparison of solute partitioning in an octanoVbuffer system [log(&)] and solute binding to etherlAM.PCC10/C3 surfaces [ l o g ( K l ~ ~ ) ] for seven B-blockers (0),six imidazoline derivatives (O), and 10 imidazolidine derivatives (A). The octanol/buffer partition coefficients of these 23 solute molecules were measured by Rogers and co-workers using a 0.01 M PES buffer (pH 7.4).5s26827 The liposome partition coefficients of these 23 solutes were measured using DMPC liposomes in 0.01 M PES buffer (pH 7.4).5,26,27 IAM capacity factors, KIAM,were measured on a 15 cm x 0.46 cm e*erlAM.PCC10/C3 column using a mobile phase of 0.01 M PES (pH 7.4). ~~~~
~
Table 2. Structures of Phenethyiamine Derivatives
:i 4 1
phenethylamine derivatives
0
r = 0.985 slope = 1.700 intercept = 0.256
1
I-
0.0
I
I
I
I
1
0.5
1.0
1.5
2.0
2.5
log @’,AM) Figure 3. Correlation of solute partitioning in an octanol/buffer [~~~(KIAM)] system [log(&a)] with solute binding to etherlAM.PCC10/C3 for 15 phenethylamine derivatives. The octanoVbuffer partition coefficients were measured using an aqueous phase buffered at pH 8.0.49 IAM capacity factors, KIAM,were measured on a 15 cm x 0.46 cm *he‘IAM.PCC10~C3 column using a mobile phase of 0.01 M PES at pH 7.4. The structures of the phenethylamine derivatives are shown in Table 2.
Based on the correlation in Figure 1, the liposome partition coefficient (K,? can be expressed in terms of KIAMas
K,’
=
= 3.M”
where 4 is the intercept of the plot of log(&& vs lOg(K”) (4 = 0.496). Since 4 is a constant, KIAMmeasured by IAM chroma-
MESC ESC PROSC ISOPROSC BROSC 2C-T DOT DON DOB DOM
DOET DOPR DOBU
DOAM DOTB
tography allows K, to be calculated for fluid membranes according to eq 6. Solute partitioning into liposomes was next compared to solute partitioning using conventionaloctanol/water systems. Figure 2A shows that the correlation between log(Km’)and log(Kod is very poor, with a linear correlation coefficient r = 0.483. The correlation between log@‘& and log(K& is also very poor, with a linear correlation coefficient r = 0.419 Figure 2B). Since the liposome partitioning system did not correlate with the octanol/aqueous Analytical Chemistry, Vol. 67,No. 4, February 15, 1995
759
Chart I.Structures of Three Different IAM.PC Surfaces Used To Determine Membrane Partition Coefficients e 8 t e r ~ ~ ~ . ~ t~ ,N(CH3)3
f
+
t
?N(cH3)3
endcapplng
c’o
partitioning system Figure 2A), the IAM partitioning system was also not expected to correlate (Figure 2B), because, as shown in Figure 1, solute partitioning into liposomes and IAMs is virtually identical for the 23 compounds tested. It should be emphasized that when nonpolar interactions between solutes and membranes dominate the membrane binding energy, both K‘, and k’m are expected to correlate with KO,,.Using only the hydrophobic P-blockers as a subset of the 23 solutes, this can be clearly seen in Figure 2C and D: log(Km’)correlates with log(KocJwith Y = 0.987 (Figure 2C), and log(k’m) correlates with log(K,,J with r = 0.986 (Figure 2D). The key concept here is that when hydrophobic interactions dominate the membrane binding energy, octanol/water partition methods give the same results as IAM methods. This was verified on another set of compounds, phenylethylamine derivatives, as shown in Figure 3: log(k’lAM_)correlates with log(KocJwith a linear correlation coefficient Y = 0.985. As shown in Table 2, these 15 phenethylamine derivatives are all very hydrophobic, and they are expected to interact predominantly with the nonpolar hydrophobic chains of the lipid bilayer. In summary,for a homologous series of hydrophobic solutes which interact mainly with the nonpolar part of the lipid bilayer, KO,, usually correlates well with Km or k’m. However, for chemical structures that have polar functional groups that interact with lipid head group during solute partitioning, K,, does not correlate well with K, or k’m (as shown in Figure 2A and B). ODS bonded phases are models of the octanol/water partition system,16J7and we attempted to compare the ~ ’ O D Swith the liposome membrane binding of the 23 solutes. However, ODS bonded phases are nonpolar, and aqueous mobile phases frequently do not have a strong enough eluotropic strength to elute hydrophobic molecules, even if very short ODS columns are used for the measurements. For instance, most of the solutes listed 760 Analytical Chemisfry, Vol. 67, No. 4, February 75, 7995
?
/N(CH3’3
I
‘OhH3
in Table 1 would not elute from ODS columns using only an aqueous mobile phase. Experimental measurements of the retention times of these 23 solutes using ODS chromatography were not performed because (i) very long retention times and organic phase modifiers are required in the mobile phase and (ii) previous results (Figure 2) showed that only hydrophobic compounds give correlation between KO,,and K,. Drug Binding to LAM.PC Columns. Chart 1 shows three IAM.PC stationary phases that have been prepared in our laboratory for the purpose of evaluating drug binding to membranes and drug transport through membranes. The firstgeneration IAM column is denoted as eSterIAM.PCand was prepared by immobilizing a double chain PC ligand containing ester-linked fatty acyl chains at the glycerol ba~kbone.~’ The key structural feature of this commercially available column is that it is non-end-capped, and consequently the silica floor contains residual amines. We have end-capped esterIAM.PCwith C10 and C3 alkyl groups to make esterIAM.PCC10/C3, but this column was not tested during this work. The second generation of IAM.PC, denoted as etherIAM.PCC10/C3, was prepared by immobilizing a single chain PC ligand with only one fatty acid linked to the glycerol backbone with an ether bond.41 The third generation of IAM.PC is denoted as dGIAM.PCc10/c3, in which the glycerol backbone was eliminated from the single chain ether PC. The data given in Figures 1-3 were performed on e*erIAM.PCC10/C3 columns, but to study the effect of PC structure on drug binding, the three IAM.PC bonded phases shown in Chart 1 were compared. As shown in Figure 4A, solute binding to esterIAM.PC[log(k’d] Correlatesvery well with solute binding to e*erIAM.PCC10/C3 [log(k’&], with a linear correlation coefficient Y = 0.990. Furthermore, a near unit slope (0.935) and a zero intercept (0.01) of log(k’m) (esterIAM.PC)vs log(k’m) (e*erIAM.PCC10/c3) indicate
2.0 4 n
u
2
1.5
s
L!
1.0
2
8 M 0.5 0
L .
0.0
-.--
slope = 0.935 intercept = 0.01
0.0
0.5
1.0
1.5
2.0
1% (k*IAM ?IAM.PC~’~’CJ)
I
intercept = -0.01
0.0
0.0
0.5
1.5
1.0
2.0
IOg (k*IAM ) (%M.PCC’w9
Figure 4. Comparison of solute binding to three different IAM.PC columns: eSterlAM.PC, e*erlAM.PCC10/C3, and nGIAM.PCC101C3 using seven six imidazoline derivatives (O), and 10 imidazolidine derivatives (A).All three IAM.PC columns were 15 cm x 0.46 cm using a /3-blockers (0), mobile phase of 0.01 M PBS (pH 7.4).
that the
IAM capacity factors ( E m ) measured on either the
esterIAM.PC or the etherIAM.PCC10/C3 column are virtually the same for the 23 molecules. Most interesting is that removing the glycerol backbone was not critical for drug binding of the 23 solutes. Thus, as shown in Figure 4B, etherIAM.PCC10/C3 and ncIAM.PCc10/c3 columns give essentially identical E m values for each compound. To summarize the data in Figure 4, for 23 test molecules, the IAM capacity factors ( E m ) measured on esterIAM.PC, etherIAM.PCC10/C3, and ncIAM.PCc10/c3 columns are the same. Although the 23 test molecules exhibited virtually equal binding constants for esterIAM.PC and etherIAM.PCC10/C3 columns, this is not the case for other sets of compounds that we have recently tested (data not shown). More recent studies in our laboratory have shown that the eSterIAM.PC column requires endcapping with C10 and then C3 alkyl groups (i.e., conversion of esterIAM.PC into esterIAM.PCC10/C3) to behave chromatographically identically to etherIAM.PCC10/C3 in the prediction of drug binding. In other words, the residual amines on the floor of the IAM surface may participate in drug retention and drug binding to the immobilized memb1-ane.4~Residual amines thus lower the correlations obtained for drug binding to IAMs and liposomes when the surface amines are drug binding sites. The main conclusion from Figure 4 is that, when the IAM surfaces are endcapped, the exact structure of the immobilized phospholipid is not critical for predicting the binding of solutes to membranes. All three of the IAM.PC bonded phases contain the same critical chemical elements responsible for solute binding, even though the structures are quite different. All of the IAM.PC phases contain a PC head group and -15 A of hydrophobic fatty acid chains. It is clear from Figure 4 that the glycerol backbone, the linkage between the glycerol backbone and the acyl chain (ether linkage or ester linkage), and the number of acyl chains are not important in evaluating solute partitioning into membranes containing phosphocholine analogs. We have demonstrated that drug partitioning into IAMs correlates very well with drug intestinal permeability, drug intestinal absorption, and oral drug absorption in The IAMs used were e*erIAM.PCC10/C3 and nGIAM.PCC101C3 columns. (47)Markovich, R J.; Qiu. X;Invergo. B.; Nichols, D. E.; Alvarez, F. M.; Pidgeon, C.A n d . Chem. 1991,63, 1851-18f30.
Scheme 2. Solute Binding to Fluid Membranes (A), Modeled by Solute Binding to Immobilized Membranes (B) by Measuring the Solute Capacity Factor by Retention Time Data (Cp solute
6
8 solute
(A)
(R)
Fluid Membrane Bilayers
ImmobiI ized membrane chromatography surfaces
t
time
-
(C) Solute Capacity factor measurements on IAMs a See eqs 5 and 6 for the relationships between and Km.
~’IAM,
KAM,
CONCLUSIONS
Solute-membrane partitioning is a critical factor in the permeability or transport of solutes across biological membranes. Solute partitioning into fluid liposome membranes can be modeled by solute partitioning into IAMs,as depicted in Scheme 2. ACKNOWLEDGMENT
We are very grateful for support from Eli Lilly and Co. This work was also supported by NSF (CIS9214794),NIH (AI33031), (48) Pidgeon, C.; Ong, S.: Liu, H.;Qiu, X;Pidgeon, M.; Dantzig, A H.; Monroe, J.; Glunz, L;Szczerba, T., manuscript submitted for publication. (49) Nichols, D. E.; Shulgin. A T.; Dyer, D. C. L$e Sci. 1977.21. 569-576.
Analytical Chemistry, Vol. 67, No. 4, February 15, 1995
761
and Regis Technologies Inc. (2R446M3022-02). We are also very grateful for the gifts of imidazolidine derivatives (STdrugs) from Boehringer-Ingelheim Co. and the gifts phenethylamine derivatives from Dr. D. Nichols of the Department of Medicinal Chemistry at Purdue University. ABBREVIATIONS
c3 c10 D
D, DMPC HPLC IAM 6) k’ VlAM KODS
K KIAM
Km
propionyl decanoyl diffusion coefficient membrane diffusion coefficient dimyristoylphosphatidylcholine high-performance liquid chromatography immobilized artiticial membrane(s) capacity factor capacity factor on IAM column capacity factor on ODS column partition coefficient IAM partition coefficient membrane partition coefficient
762 Analytical Chemistty, Vol. 67, No. 4, Februaty 15, 1995
Km’ Koct
ODS P PA PBS PC PE PG PS QSAR SPA
liposome partition coefficient octanol/water partition coefficient octadecyl silica permeability coefficient phosphatidic acid phosphate buffered saline phosphatidylcholine phosphatidylethanolamine phosphatidylglycerol phosphatidylserine quantitative structure-activity relationship silica propylamine
Received for review August 24, 1994. Accepted November 29, 1994.@ AC940842E @
Abstract published in Adoance ACS Abstracts, January 1, 1995.
