rapid-gradient hplc method for measuring drug interactions with immobilized artificial membrane:...

Rapid-Gradient HPLC Method for Measuring DrugInteractions with Immobilized Artificial Membrane:Comparison with Other Lipophilicity Measures

KLARA VALKO,1 CHAU MY DU,2 CHRISTOPHER D. BEVAN,1 DEREK P. REYNOLDS,1 MICHAEL H. ABRAHAM2

1 Hit Generation Sciences & Analytical Technologies, GlaxoWellcome Medicines Research Centre, Stevenage,Herts SG1 2NY, United Kingdom

2 Department of Chemistry, University College London, 20 Gordon Street, London WC1H 0AJ, United Kingdom

Received 12 May 1999; revised 25 April 2000; accepted 4 May 2000

ABSTRACT: A fast-gradient high-performance liquid chromatographic (HPLC) methodhas been suggested to characterize the interactions of drugs with an immobilized ar-tificial membrane (IAM). With a set of standards, the gradient retention times can beconverted to Chromatographic Hydrophobicity Index values referring to IAM chroma-tography (CHIIAM) that approximates an acetonitrile concentration with which theequal distribution of compound can be achieved between the mobile phase and IAM.The CHIIAM values are more suitable for interlaboratory comparison and for highthroughput screening of new molecular entities than the log kIAM values (isocraticretention factor on IAM). The fast-gradient method has been validated against theisocratic log kIAM values using the linear free energy relationship solvation equationsbased on the data from 48 compounds. The compound set was selected to provide a widerange and the least cross-correlation between the molecular descriptors in the solvationequation:

SP = c + r?R2 + s?p2H + a?Sa2

H + b?Sb20 + v?Vx (2)

where SP is a solute property (e.g., logarithm of partition coefficients, reversed-phase(RP)-HPLC retention parameters, such as log k, log kw, etc.) and the explanatory vari-ables are solute descriptors as follows: R2 is an excess molar refraction that can beobtained from the measured refractive index of a compound, p2

H is the solute dipolar-ity/polarizability, ∑a2

H and ∑b20 are the solute overall or effective hydrogen-bond acid-

ity and basicity, respectively, and Vx is the McGowan characteristic volume (in cm3/100mol) that can be calculated for any solute simply from molecular structure using a tableof atomic constants. It was found that the relative constants of the solvation equationwere very similar for the CHIIAM and for the log kIAM. The IAM lipophilicity scale wasquite similar to the octanol/water lipophilicity scale for neutral compounds. The effectof charge on the interaction with IAM was studied by varying the mobile phase pH. ©2000 Wiley-Liss, Inc. and the American Pharmaceutical Association J Pharm Sci 89: 1085–1096,2000Keywords: immobilized artificial membrane; IAM chromatographic hydrophobicityindex; CHI solvation equation; high throughput screening

INTRODUCTION

In the drug discovery process, the evaluation ofthe drug–membrane interaction is a critical step

Correspondence to: K. Valko (Telephone: 44-(0)1438763309; Fax: 44-(0)1438 763352; E-mail: [email protected])Journal of Pharmaceutical Sciences, Vol. 89, 1085–1096 (2000)© 2000 Wiley-Liss, Inc. and the American Pharmaceutical Association

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 89, NO. 8, AUGUST 2000 1085

because drug activity, toxicity, distribution, andabsorption depend on drug–membrane partition-ing. Pidgeon and co-workers1–3 first introducedan easy way to measure drug–membrane interac-tions by immobilized artificial membrane (IAM)chromatography. Initially, they bonded lecithinonto a chromatographic support. Later theychemically bonded phospholipids with aminopro-pylsilica stationary phases used in high-performance liquid chromatography (HPLC). Sev-eral types of IAM stationary phases are available(Regis Technologies, Inc., Morton Grove, IL) andhave been used to evaluate drug–membrane pas-sive transport properties.4–7 Many of them con-tain a phosphatidylcholine (PC) head group andester or an ether linkage between the two acylchains and the glycerol backbone of the PC mol-ecule. Some of them, denoted as drug discovery(DD) columns, are short (5 cm) and designed for aquick determination of the retention factor (k) ofthe compounds on the IAM stationary phase. TheIAM.PC.DD phase has a single acyl chain and noglycerol backbone that can yield different results,particularly when ionized compounds are consid-ered. In this study, we used an IAM − PC 15-cmlong column with double acyl chain and glycerolbackbone.

The logarithmic value of the retention factor(log kIAM) can be directly related to the equilib-rium IAM partition coefficient (KIAM)8 as is shownby eq 1:

log kIAM = log KIAM + log ~Vs/Vm! (1)

where Vs/Vm is a constant characteristic of thecolumn called the phase ratio. The log KIAM canbe regarded as a linear free energy parametersimilar to the logarithmic value of the membranepartition coefficient (log Km). The molecular ratio-nale for using IAM chromatography to predict sol-ute partitioning into fluid membranes is thatIAMs are physically and chemically similar andtherefore mimic fluid phospholipid bilayers.9,10

Log kIAM values have shown good correlationwith skin penetration, stratum corneum mem-brane permeability,11 blood/brain barrier distri-bution,12, 13 CaCo-2 cell permeability, and absorp-tion in the small intestine of the rat.8 The corre-lations were better than with octanol/waterpartition coefficients (log P), octanol/water distri-bution coefficients at pH 7.4 (log D), or lipophilic-ity measured on octadecylsilica HPLC (log kw).Receptor binding values from rat cortical brainpreparations were also successfully correlated

with log kIAM values for calcium-channel block-ers.4. Yang et al.15 reviewed many other success-ful correlations with solute partitioning into lipo-some, bile salt-membrane interactions, etc. An-other recent review16 summarizes the use of IAMchromatography for drug transport applications.

Kaliszan et. al.17,18 have compared the IAM li-pophilicity scale with other chromatographic lipo-philicity measures and with the octanol/waterpartition coefficients for a heterogeneous set ofcompounds including acidic and basic drug mol-ecules from various pharmacological classes (b-adrenolytics, phenothiazines, a-adrenomimetics,etc.). They concluded that each hydrophobicitymeasure reveals some specific aspects of the na-ture of the hydrophobic binding sites on receptorsor plasma proteins. Thus different hydrophobicitymodels may be required for predicting drug per-meation through biological membranes. The IAMchromatography, however, has a definite advan-tage in that it is easier to measure in a simple,fast, and reproducible manner.

Because there is a need for automated, highthroughput characterization of newly synthesizedcompounds (compound libraries) at early stages ofdrug development, we wanted to investigate thepossible advantages of using IAM − HPLC reten-tion data in relation to model building and pre-dicting biological transport properties.

The advantages of using HPLC are well knownfor this purpose: there is no need to determine theconcentration of compounds, minor impurities areseparated during the procedure and do not dis-turb the distribution measurement, and thewhole process can be easily automated. However,there are certain disadvantages of using retentionfactors when we want to build up a large databasebecause the interlaboratory comparison is diffi-cult. One disadvantage is that the application ofreference compounds is recommended for correc-tions of the column aging.13 Another disadvan-tage is that the isocratic mobile phase composi-tion has to be modified by using organic modifier(acetonitrile) for compounds that strongly inter-act with the IAM stationary phase. This proce-dure means that some compounds have to be mea-sured under various conditions and the log kIAMdata can only be obtained by extrapolation.

Recently we have proposed19,20 that a gradientretention time, linearly converted to a chromato-graphic hydrophobicity index (CHI), gives anequivalent lipophilicity scale as the isocratic re-tention factor. The absolute magnitude of the CHIparameter is dependent on the values assigned to

1086 VALKO ET AL.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 89, NO. 8, AUGUST 2000

the set of standards used to calibrate the gradi-ent, and it approximates to the organic phase con-centration by which equal distribution of the com-pound can be achieved between the mobile andthe stationary phases. The advantage of CHI isthat we can determine the retention parameterfrom a single fast-gradient run for all compoundswithout preliminary method development. Wehave used the solvation equation21 to compare theisocratic and the gradient lipophilicity data aswell as to compare the CHI scale with other lipo-philicity scales.

The general solvation equation model (eq 2)has been applied to numerous processes that in-volve transport of solutes between chromato-graphic phases,22–24 solvent/solvent partitions,25

blood–brain barrier distribution,26,27 and skinpenetration:28

SP = c + r?R2 + s?p2H + a?Sa2

H + b?Sb2H + v?Vx

(2)

where SP is a solute property (e.g., logarithm ofpartition coefficients, RP-HPLC retention param-eters, such as log k, log kw, etc.) and the explana-tory variables are solute descriptors as follows: R2is an excess molar refraction that can be obtainedfrom the measured refractive index of a com-pound, p2H is the solute dipolarity/polarizability,∑a2

H and ∑b2H are the solute overall or effective

hydrogen-bond acidity and basicity, respectively(note that the alternative ∑b2

0 parameter is usedhere, as required, for some special solutes), andVx is the McGowan characteristic volume (in cm3/100 mol) that can be calculated for any solute sim-ply from molecular structure using a table ofatomic constants.21

The purposes of this study were to develop ageneric gradient method for the determination ofdrug–IAM interaction without the need of severalisocratic measurements and extrapolation of theorganic phase concentration to zero, to comparethe data obtained with the traditional isocraticmethod and the new fast-gradient method, and tocompare the IAM lipophilicity scale with otherlipophilicity measures, like octanol/water parti-tion coefficients and RP-HPLC lipophilicity. Inthis paper we report the validation of the fast-gradient method in comparison to the isocraticmethods based on the log kIAM values and theCHIIAM values of 48 compounds. The solvationequation has been set up for the CHIODS and logP lipophilicity scales for the same set of com-

pounds, and the parameters of the equations havebeen compared.

EXPERIMENTAL SECTION

Reagents

The 48 compounds investigated in this study(shown in Table 1) were all commercially avail-able and were selected to represent a wide varietyof solute properties. They were dissolved in ace-tonitrile (Rathburn, Walkerburn, UK) and 50 mMaqueous ammonium acetate solutions (pH 7.4; Fi-sons, Loughborough, UK) at ∼0.5 mg/mL concen-tration. A 5-mL aliquot of the solutions was in-jected onto the HPLC system. The HPLC mobilephases were also 50 mM ammonium acetate andHPLC grade acetonitrile from the aforementionedsource. The concentrated phosphoric acid and ace-tic acid for the pH adjustment were obtained fromFisons Reagents (Loughborough, UK).

IAM Chromatography

A Hewlett Packard 1090 series high-performanceliquid chromatograph was used. Data acquisitionand processing were performed on a Viglen IBMcompatible PC with HP Chemstation software(Hewlett-Packard, Amsterdam, Netherlands). Weused a RexChrom IAM PC2 (CH2)12 column (S-12-300-IAM-PC) with the dimensions of 150 × 4.6mm obtained through Fisher Scientific UK,Loughborough, UK.

Isocratic Measurements of log kIAM

The mobile phase “A” contained 50 mM ammo-nium acetate. The pH of mobile phase A was ad-justed to pH 7.4 with concentrated ammonia so-lution. Mobile phase “B” was 100% acetonitrile.For the isocratic determination of the retentiontimes (tR), 0, 10, 20, 30, and 35% acetonitrile wasused. The dead time (t0) of the system was deter-mined by injecting sodium nitrate solution. Thelog kIAM values were obtained by log[(tR − t0)/t0].The mobile phase flow rate was 1.00 mL/min.Most of the compounds had retention times of >30min when only buffer was used as the mobilephase. Therefore, the log kIAM values referring tothe buffer-only mobile phase were extrapolatedby plotting the log kIAM values and the appliedacetonitrile concentration. The intercept of thestraight line was used as the extrapolated logkIAM to buffer-only mobile phase.

RAPID-GRADIENT HPLC TO MEASURE LIPOPHILICITY 1087


Table 1. The Investigated Compounds and Their log Poct, CHIIAM, CHIIn, and log kIAM Data

Compound Log PCHIIAM

neutralCHIIAM

at pH 7.4CHIIAM

at pH 5.7CHIIAM

at pH 6.1Log kIAM

20% pH 7.4Log kIAM

extrapol 0%CHIODS

at pH 7.4

n-Heptanophenone 4.32 45.49 45.49 45.39 45.36 0.957 3.04 112.79n-Hexanophenone 3.79 41.91 41.91 41.82 41.88 0.651 2.13 105.13n-Valerophenone 3.28 37.71 37.71 37.62 37.75 0.381 1.69 96.53n-Butyrophenone 2.66 32.55 32.55 32.57 32.68 0.094 1.32 87.14n-Propiophenone 2.19 26.39 26.39 26.56 26.54 −0.078 0.88 76.93Acetophenone 1.58 17.07 17.07 17.45 17.32 −0.424 0.75 63.26Acetanilide 1.16 11.57 11.57 11.62 11.56 −0.086 1.16 42.38Theophylline −0.02 −7.72 −8.74 −6.76 −6.90 −0.770 0.14 17.91Caffeine −0.07 −5.58 −5.58 0.10 −0.34 −0.760 0.25 24.45Indazole 1.77 24.08 24.08 23.01 23.16 −0.272 1.33 49.95Benzonitrile 1.56 15.64 15.64 16.25 16.28 −0.163 1.24 65.07Chlorobenzene 2.89 36.83 36.83 35.96 36.15 0.826 2.06 92.17Naphthalene 3.30 42.75 42.75 41.89 41.91 1.471 2.45 99.421,4-Dinitrobenzene 1.47 20.55 20.55 20.86 20.89 0.289 1.45 70.90Testosterone 3.31 38.22 38.22 38.52 38.56 1.162 2.86 75.67Hydrocortisone 1.55 26.03 26.03 26.54 26.69 0.473 1.92 50.83Cortisone-21-acetate 2.10 30.36 30.36 31.05 31.19 0.822 2.54 56.80Pyrene 5.00 60.52 60.52 58.55 58.54 1.650 3.39 124.10Aldosterone 1.08 22.09 22.09 22.74 23.03 −0.257 1.64 46.13Progesterone 3.70 41.29 41.29 42.08 42.14 1.466 3.38 100.38Anisole 2.11 25.43 25.43 24.95 25.22 0.365 1.42 78.41Benzamide 0.64 −1.96 −1.96 −1.55 −1.33 −0.372 0.39 29.23Adenine −0.09 −11.90 −11.90 −11.07 −10.32 −0.869 0.10 18.833,4-Dichlorophenol 3.33 46.12 47.72 45.29 45.54 1.377 2.79 77.36Phenol 1.50 18.68 18.68 16.04 16.45 0.146 1.01 48.394-Nitrophenol 1.91 32.78 26.85 31.19 31.37 0.458 1.47 56.294-Chlorophenol 2.15 32.90 32.90 31.11 31.27 0.594 1.68 62.644-I-Phenol 2.91 44.13 44.13 41.84 41.98 1.142 2.45 73.39Resorcinol 0.80 13.75 13.75 11.01 11.49 −0.064 0.70 25.214-CN-Phenol 1.60 23.50 22.34 22.16 22.48 0.306 1.38 47.974-Nitrobenzoic acid 1.89 41.53 −3.63 18.32 14.31 −0.229 1.16 56.28Benzoic acid 1.87 28.09 −16.34 3.42 −9.58 −0.535 0.73 50.224-OH-Benzyl alcohol 0.25 −3.28 −2.74 −4.61 −4.07 −0.432 0.45 18.91Salicylic acid 2.25 55.59 5.46 21.25 18.24 −0.159 1.25 56.87Phenylacetic acid 1.41 22.42 −16.68 2.59 −10.44 −0.519 0.68 50.814-Nitroaniline 1.39 25.46 25.62 25.19 25.51 0.366 1.46 53.48Propranolol 3.37 42.67 42.67 31.42 33.53 0.908 2.06 87.12p-Toludine 1.39 15.93 15.93 13.79 15.62 0.088 0.95 58.08Pyridine 0.65 −8.14 −8.14 −6.44 −1.98 −0.503 0.30 28.86Aniline 0.90 1.00 1.00 −0.89 0.27 −0.156 0.56 45.343-Nitroaniline 1.37 23.32 23.32 23.03 23.31 0.296 1.34 58.30Procaine 1.89 14.64 14.64 2.7 7.86 −0.233 0.87 61.31Nicotine 1.17 8.51 8.51 −12.02 3.58 −0.357 0.40 58.77Pyridazine −0.72 −20.17 −18.15 −19.13 −18.66 −1.155 −0.50 4.77Methyl-

4-hydroxybenzoate 1.96 25.95 25.95 25.05 25.28 0.394 1.56 52.00n-Ethyl

4-hydroxybenzoate 2.47 31.67 31.67 30.87 31.02 0.674 2.00 61.50n-Propyl

4-hydroxybenzoate 3.04 37.48 37.48 36.53 36.65 1.011 2.52 71.08n-Butyl

4-hydroxybenzoate 3.57 42.3 42.3 41.29 41.36 1.371 3.09 79.99

1088 VALKO ET AL.


Determination of CHIIAM

For the determination of CHIIAM values, the gra-dient retention times (tg) were measured underthe following gradient conditions: 0–1.5 min, 0%acetonitrile; 1.5–10.5 min, 0–100% acetonitrile;10.5–11.5 min, 100% acetonitrile; 11.5–12.0 min,100–0% acetonitrile; 12.0–15.0 min, 0% acetoni-trile. The mobile phase flow rate was 1.0 mL/min.The system was calibrated by injecting the follow-ing compounds and plotting their CHIIAM val-ues29 against the obtained gradient retentiontime (tg) values: octanophenone (CHIIAM 4 49.4),hetanophenone (CHIIAM 4 45.7), hexanophenone(CHIIAM 4 41.8), valerophenone (CHIIAM 437.3), butyrophenone (CHIIAM 4 32.00), propio-phenone (CHIIAM 4 25.9), acetophenone (CHIIAM4 17.2), acetanilide (CHIIAM 4 11.5), and para-cetamol (CHIIAM 42.9). The constants of thestraight line (slope and intercept) were used toconvert the gradient retention times to CHI val-ues (CHI 4 slope tg + intercept). We used severalA solvents; they are, 50 mM ammonium acetatebuffer (pH 7.4), the 50 mM ammonium acetatebuffer adjusted to pH 6.1 with acetic acid, 50 mMammonium acetate buffer adjusted to pH 5.7 withacetic acid, and 0.1% phosphoric acid solution (pH2). Thus, several CHIIAM values were obtainedreferring to various pHs. We preferred to use am-monium acetate buffer, which is compatible withmass spectrometry, rather than the originallysuggested phosphate buffer saline (PBS) mobilephase, for the IAM chromatography.

Determination of CHIODS

For the determination of CHI values on a tradi-tional RP-HPLC column (ODS2-IK5 Inertsil;Capital HPLC Ltd., Broxburn, Scotland), a col-umn with the dimensions of 150 × 4.6 mm wasused. The mobile phases and the gradient pro-grams were the same as is described for the IAMcolumn. The calibration mixture was slightly dif-ferent. The procedure has been described in de-tails in our previous papers.19,20

Octanol/Water Partition Coefficients (log P)

The measured log P values were retrieved fromthe Medicinal Chemistry Database (BioByteCorp., Claremont, CA), using software from Day-light CIS (Santa Fe, NM).

Calculations

The data analysis was carried out by using theMicrosoft™ Excel 5 software package. The mul-tiple linear regression analysis was carried outusing the Drugidea software package (ChemicroLtd., Budapest, Hungary) with built-in cross vali-dation.

RESULTS AND DISCUSSION

Table 1 shows the measured octanol/water parti-tion coefficients (log P) from the MedChem Data-base, the measured isocratic log kIAM data using20% acetonitrile in the mobile phase, the log kIAMdata extrapolated to the zero organic phase con-centration, and the CHIIAM data obtained at pH7.4 as well as the CHIIAM data for the neutralstate of the compounds. We used ammonium ac-etate buffers throughout the IAM and RP-HPLCmeasurements. The advantage of using ammo-nium acetate buffer instead of the suggested PBSfor IAM chromatography is its compatibility withmass spectrometry. When the high throughputscreens are running at an early stage of drug re-search, the compounds are not always very pureand liquid chromatography–mass spectrometry(LC–MS) is frequently used for the characteriza-tion of new libraries. Because the ammonium ac-etate is volatile, it is compatible with many ion-ization modes applied in mass spectrometry. Weare aware that the use of ammonium acetatebuffer instead of PBS can give rise to discrepan-cies with the results of other authors because of adifferent capability to form ion pairs.

Comparison of Isocratic and Gradient RetentionData on IAM

Figure 1 shows the plot of the isocratically deter-mined log kIAM values obtained using 20% aceto-nitrile against CHIIAM values for the 48 com-pounds. Both sets of values were obtained at pH7.4. It can be seen that a reasonably good corre-lation was obtained between the two sets of data.Higher scatter can be observed for compoundsthat had weak interaction with the IAM station-ary phases (short retention times). For stronglyretained compounds, the peaks were very wide inthe isocratic mode and the measurements tookmuch longer. Much narrower peaks were ob-served for strongly interacting compounds withthe gradient method, which makes the retention



time determination more precise. Because all thepeaks elute within the gradient time (15 min), thegradient method is much faster. For the compari-son of the two methods, the solvation equationwas set up for the isocratic log kIAM measured at20% acetonitrile in the mobile phase and for theextrapolated log kIAM values to the pure buffer aswell as for the gradient CHIIAM values (eqs 3–5)The four strong acids (salicylic acid, benzoic acid,4-nitrobenzoic acid, and phenylacetic acid) wereleft out of the equations because they are stronglyionized at pH 7.4 and the solvation equation isvalid only for neutral molecules:

log kIAM ~20%! = −0.83 + 0.23~50.08!?R2

− 0.20~50.08!?p2H

+ 0.22?~50.09!?Sa2H

− 2.03~50.17!?Sb20

+ 1.89~50.11!?Vxn = 44 r = 0.971 s = 0.17 F = 125 (3)

log kIAM ~0%! = −0.37 + 0.28~50.13!?R2

− 0.08~50.12!?p2H

+ 0.30?~50.13!?Sa2H

− 2.65~50.15!?Sb20

+ 2.53~50.17!?Vxn = 44 r = 0.964 s = 0.27 F = 101 (4)

CHIIAM~pH 7.4! = −7.42 + 8.02~52.06!?R2

− 8.81~52.16!?p2H

+ 5.97?~52.16!?Sa2H

− 52.38~52.42!?Sb20

+ 50.64~52.32!?Vxn = 44 r = 0.973 s = 4.42 F = 137 (5)

The n refers to the number of compounds takeninto consideration to establish the regression con-stants, r is the multiple correlation coefficient, s isthe standard error of the estimate, and F is thevalue of the Fisher test. Note that the CHI scaleranges from 0 to 100, whereas the log k scaleranges from −1.155 to 3.39; this variation ac-counts for the difference in the values of s forequations referring to gradient and isocratic data.We created another set of CHIIAM values calledCHIIAM,neutral when we included the CHIIAM val-ues obtained at pH 2 for all the compounds exceptthe basic ones. The data of the 4-nitrobenzoic acidand salicylic acid were outliers (they showedhigher CHI values than expected at low pH). Wehave included the data of benzoic acid and phe-nylacetic acid and eq 6 could be set up (the inde-pendent variable used included the numbers inthe second column in Table 1):

CHIIAM, neutral = −6.95 + 7.43~52.06!?R2

− 8.69~52.16!?p2H

+ 7.28?~52.16!?Sa2H

− 52.64~52.42!?Sb20

+ 50.73~52.32!?Vxn = 46 r = 0.969 s = 4.67 F = 123 (6)

For the easier comparison of the impact of thevarious molecular descriptors on the IAM reten-tion, the relative values of the coefficients to theVx coefficient (r/v, s/v, a/v, b/v) were calculatedand are tabulated in Table 2. Table 2 contains therelative coefficients of the solvation equations setup for all the measured isocratic and gradientdata as well as the log P and CHI values obtainedon regular ODS column with acetonitrile. It canbe seen in Table 2, that the relative coefficients ofthe solvation equations referring to the isocraticand gradient IAM data are very similar. They allhave positive coefficients for R2 (r/v is within therange 0.04–0.16), negative coefficients for ∑p2

H

(s/v is within the range −0.03–−0.17), small posi-tive coefficients for ∑a2

H (a/v is within the range0.07–0.14), and large negative coefficients for∑b20 (b/v is within the range ms1.03–−1.15). Com-paring with the relative coefficients obtained forthe octanol/water log P, we can see that none ofthem are sensitive to the H-bond acidity of thecompounds (they have relative coefficients posi-tive and close to zero). More exactly, the IAM in-teracts more strongly with H-bond acids. In com-parison with the ODS systems using a similaracetonitrile gradient, the relative retention (CHI)

Figure 1. The plot of isocratic log kIAM (20%) versusthe gradient CHIIAM values at pH7.4

1090 VALKO ET AL.


of H-bond acid compounds is smaller (the coeffi-cient for ∑a2

H is a significant negative value). Weshould mention that the CHI values obtained onthe IAM column are less than half that obtainedon the ODS column. For example, the CHI valuesof heptanophenone are CHIODS 4 112.8, andCHIIAM 4 45.5 (see the plot in Figure 2). Thosecompounds that are far from the trend line con-tain H-bond donor groups, so their ∑a2

H valuesare high. The two acidic compounds seem to haveless interaction with the IAM than what would beexpected based only on their charge. The IAM sys-tem is more sensitive to the highly dipolar or po-larizable compounds with acetonitrile in the mo-

bile system than without acetonitrile. However,the relative coefficient for ∑p2

H is still a smallernegative value for the IAM systems than for theoctanol/water system. The plot in Figure 3 showsthe relationship of CHIIAM,neutral values to log Pvalues. Comparing Figures 2 and 3, we can seethat the IAM lipophilicity scale is much closer tothe octanol/water lipophilicity scale than theODS–aqueous mobile phase lipophilicity scale.However, the two strong acids (4-nitrobenzoicacid and salicylic acid) seem to interact morestrongly with the IAM, as expected from the log Pvalue. These results are perfectly consistent withthe findings of Abraham et al.23 when they set up

Table 2. The Relative Regression Coefficients and the Statistical Parameters of the Solvation EquationsObtained for Isocratic and Gradient Retention Data on IAM, log P (Octanol/Water), and CHI Values on anInertsil ODS Column with Acetonitrile

Solute property r/v s/v a/v b/v N R S

Log kIAM (20%) 0.04 −0.11 0.12 −1.07 44 0.971 0.17Log kIAM extrapol to 0% 0.11 −0.03 0.12 −1.04 44 0.964 0.27CHIIAM (pH 7.4) 0.16 −0.17 0.12 −1.03 44 0.973 4.42CHIIAM (pH 6.1) 0.14 −0.11 0.07 −1.11 44 0.969 4.64CHIIAM (pH 5.7) 0.16 −0.10 0.09 −1.15 44 0.951 6.06CHIIAM (neutral) 0.15 −0.17 0.14 −1.04 46 0.969 4.67Log P 0.13 −0.26 0.02 −0.93 46 0.997 0.1CHIODS AcN 0.16 −0.24 −0.29 −1.01 44 0.985 5.07

Figure 2. The plot of CHIIAM values in the function of CHIIn, ODS values obtained at pH 7.4.



solvation equations for the isocratic retention onIAM column using pH 7.4 PBS as the mobilephase.

Based on the solvation equations, the IAM li-pophilicity scale represents a very similar lipo-philicity scale to the octanol/water partition forneutral compounds. This finding is in contradic-tion to the results of Pidgeon et al.6 However, thisresult corresponds to those of Salminen et al.,13

who say that IAM chromatography offers no ad-vantage over the octanol/water partition data forneutral compounds. Pidgeon et al.8 have alsopointed out that the IAM provides a better model

for the polar interactions with the charge sur-faces, whereas when nonpolar interactions domi-nate, the membrane bonding energy, log kIAM,and log P are expected to correlate. It also can beseen from our results that the ODS HPLC lipo-philicity scale is very different from log P or theIAM scale even for neutral molecules. Principalcomponent analysis has been carried out on theretention and partition data listed in Table 1. Theanalysis was carried out on the correlation ma-trix. Figure 4 shows the plot of the first and thesecond principal component loadings. The closertogether the points are, the closer the correlationof the retention data (i.e., the more similar thesystems). The correlation coefficients of the reten-tion data obtained in the seven systems repre-sented in Figure 4 are summarized in Table 3. Itcan be seen that the gradient and isocratic IAMretention data obtained at pH 7.4 represents avery similar lipophilicity scale to the log P scale.The gradient chromatographic hydrophobicityindex obtained on the Inertsil ODS column(CHIODS), however, represents a different scale aswas pointed out earlier by comparing Figures 2and 3. It also can be seen that by decreasing thepH of the mobile phase in IAM chromatography,the parameters of the equations become differentfrom those obtained at pH 7.4. Therefore, we in-vestigated the effect of pH of the mobile phase.

Figure 3. The plot of CHIIAM, neutral and log P values.

Figure 4. The plot of the first two principal components obtained from the data in Table 1.

1092 VALKO ET AL.


Comparison of CHIIAM Values Obtained atVarious pHs

When we plot the gradient retention data of theinvestigated compounds obtained at various pHs,we can reveal which compounds change their de-gree of interaction with the IAM stationary phaseand the direction of the change. From the changewe can recognize the basic and acidic compoundsin our set of compounds. Figure 5 shows the plotof CHIIAM values obtained at pH 2 and 7.4. It canbe seen that acidic compounds show much longerretention on IAM at pH 2 and basic compoundshave weaker interaction at pH 2 with the IAMstationary phase. This result can be expectedfrom other lipophilicity scales as well, where the

charge (either positive or negative) will decreasethe lipophilicity of the molecules. Among the in-vestigated molecules, the four acids (benzoic acid,salicylic acid, 4-nitrobenzoic acid, and phenylace-tic acid) can be expected to be charged at pH 7.4.However, there are also basic compounds thatshould be partially ionized at pH 7.4 (propranolol,procaine, and nicotine). These latter compoundswere not outliers when we set up the solvationequation, which is valid only for neutral com-pounds (see Figure 6). This phenomenon is inagreement with a previous study published byBarbato et al.30 By plotting the measured and re-calculated CHIIAM,neutral values with eq 6, onlythe two strong acids were outliers. The data of the

Table 3. The Correlation Coefficients of the Various Lipophilicity Data (Isocratic Retention on IAM, GradientCHIs on IAM and ODS and log P)

Solute property Log kIAM,20 Log kIAM,0 CHIIAM,7.4 CHIIAM,6.1 CHIIAM,5.7 Log P CHIODS

Log kIAM,20 1.00 0.98 0.96 0.97 0.96 0.94 0.86Log kIAM,0 0.98 1.00 0.93 0.93 0.93 0.91 0.81CHIIAM,7.4 0.96 0.93 1.00 0.99 0.88 0.95 0.89CHIIAM,6.1 0.97 0.93 0.99 1.00 0.99 0.94 0.89CHIIAM,5.7 0.96 0.93 0.98 0.99 1.00 0.91 0.86Log P 0.94 0.91 0.95 0.94 0.91 1.00 0.94CHIODS 0.86 0.81 0.89 0.89 0.86 0.94 1.00

Figure 5. The plot of the CHIIAM values obtained at acidic (pH 2) and neutral (pH 7.4) conditions.



charged basic compounds are on the line. Thisobservation suggests that ionized basic com-pounds strongly interact with the IAM stationaryphase and their retention becomes as strong as ifthey were uncharged. Austin et al.31 have also

found that protonated, charged amines are able topartition into a phospholipid bilayer, and theysuggest it is not a consequence of ion pairing.However, when we plot CHIIAM pH 7.4 againstCHIIAM pH 5.7 (Figure 7) we can see that these

Figure 7. The differences of CHIIAM values obtained at pH 5.7 and 7.4. (The data of the strong acids were omitted.)

Figure 6. The measured and the re-calculated CHIIAM,neutral values by eq 6.

1094 VALKO ET AL.


basic compounds have smaller CHI values atlower pH, which suggests that their ionization in-fluences their IAM lipophilicity as well. Ottiger etal.32 investigated thoroughly the pH dependenceof retention on IAM stationary phase. We shouldmention that when we change the mobile phasepH, we also change the charges on the IAM sur-face. Further study is needed to compare the ef-fect of ionization of basic compounds in the octa-nol/water system, simple HPLC ODS systems,and the IAM system. Barton at al.33 found a veryinteresting result when they studied in vivo dis-tribution of basic, acidic, and neutral compoundsin adipose tissue. They found that the differencebetween the octanol/water partition and IAM par-tition at pH 7.4 showed significant correlationwith the tissue distribution. According to our re-sults, these differences (dlog D octanol/water andmembrane/water) should arise only from the dif-ferences in the sensitivity towards charge.

CONCLUSIONS

The rapid gradient measurements of the interac-tion of compounds with the IAM reveal the sameinteractions as the isocratic measurements. Thesuggested method provides a fast and easily au-tomated screen for modeling the membrane inter-action of a large number of new molecular enti-ties. Based on the solvation equation, the IAMlipophilicity scale was more similar to the octa-nol/water lipophilicity scale than to the RP-HPLClipophilicity scale. We found that charged basiccompounds had stronger interactions with theIAM column, as expected. Therefore, the CHIIAMvalues for charged compounds could be signifi-cantly different from their behavior as deter-mined by other lipophilicity scales. A fast methodfor the determination of the interaction of com-pounds with IAM has a useful predictive role inthe drug discovery process.

REFERENCES

1. Pidgeon C, Venkataram UV. 1989. Immobilized ar-tificial membrane chromatography: Supports com-posed of membrane lipids. Anal Biochem 176:36–47.

2. Pidgeon C, Stevens J, Otto S, Jefcoate C, Marcus C.1991. Rapid purification of functional membraneproteins. Anal Biochem 194:163–173.

3. Pidgeon C, Marcus C, Alvarez F. 1991. Immobi-lized artificial membrane chromatography: Surface

chemistry and applications. In: Kelly JW, BaldwinTO, editors. Applications of enzyme biotechnology.New York: Plenum, pp 201–220.

4. Kansy M, Senner F, Gubernator K. 1998. Physico-chemical high throughput screening: Parallel arti-ficial membrane permeation assay in the descrip-tion of passive absorption processes. J Med Chem41:1007–1010.

5. Ong S, Liu H, Qui X, Bhat G, Pidgeon C. 1995.Membrane partition coefficients chromatographi-cally measured using immobilized artificial mem-brane surfaces. Anal Chem 67:755–762.

6. Pidgeon C, Ong S, Liu H, Qui X, Pidgeon M,Dantzig AH Munroe J, Hornback WJ, Kasher JS,Glunz L, Szczerba T. 1995. IAM chromatography:An in vitro screen for predicting drug membranepermeability. J Med Chem 38:590–594.

7. Caldwell GW, Masucci JA Evangelisto M, White R.1998. Evaluation of the immobilized artificialmembrane phosphatidylcholine. Drug discoverycolumn for high-performance liquid chromatogra-phy screening of drug-membrane interactions.J Chromatogr A 800:161–169.

8. Ong S, Liu H, Pidgeon C. 1996. Immobilized-artificial-membrane chromatography: measure-ments of membrane partition coefficient and pre-dicting drug membrane permeability. J Chro-matogr A 728:113–128.

9. Ong S, Cai SJ, Bernal C, Rhee D, Qui X, Pidgeon C.1994. Phospholipid immobilization on solid sur-faces. Anal Chem 66:782–792.

10. Rhee D, Markovich R, Chae WG, Qui X, Pidgeon C.1994. Chromatographic surfaces prepared fromlysy phosphatidylcholine ligands. Anal Chim Acta297:377–386.

11. Alvarez FM, Bottom CB, Chikhale P, PidgeonC.1993. Immobilized artificial membrane chroma-tography; Prediction of drug transport across bio-logical barriers. In: Ngo TT, editor. Molecular in-teractions in bioseparations. New York: Plenum,pp 151–166.

12. Reichel A, Begley DJ. 1998. Potential of immobi-lized artificial membranes for predicting drug pen-etration across the blood-brain barrier. Pharm Res15:1270–1274.

13. Salminen T, Pulli A, Taskinen J. 1997. Relation-ship between immobilised artificial membranechromatographic retention and the brain penetra-tion of structurally Diverse drugs. J Pharm BiomedAnal 15:469–477.

14. Barbato F, LaRotonda MI, Quaglia F. 1996. Chro-matographic indices determined on an immobilizedartificial membrane (IAM) column as descriptors oflipophilic and polar interactions of 4-phenyldihy-dropyridine calcium-channel blockers with bio-membranes. Eur J Med Chem 31:311–318.

15. Yang CY, Cai SJ, Liu HL, Pidgeon C. 1997. Immo-bilized artificial membranes—Screens for drug



membrane interactions. Adv Drug Delivery Rev 23:229–256.

16. Stewart BH, Chan OH. 1998. Use of immobilizedartificial membrane chromatography for drugtransport applications. J Pharm Sci 87:1471–1478.

17. Kaliszan R, Kaliszan A, Wainer IW.1993. Deacti-vated hydrocarbonaceous silica and immobilizedartificial membrane stationary phases in high-performance liquid chromatographic determina-tion of hydrophobicities of organic bases: Relation-ship to log P and CLOGP. J Pharm Biomed Anal11:505–511.

18. Kaliszan R, Nasal A, Bucinski A. 1994. Chromato-graphic hydrophobicity parameter determined onan immobilized artificial membrane column: Rela-tionships to standard measures of hydrophobicityand bioactivity. Eur J Med Chem 29:163–170.

19. Valko K, Bevan C, Reynolds D. 1997. Chromato-graphic hydrophobicity index by fast-gradient RP-HPLC: A high-throughput alternative to log P/logD. Anal Chem 69:2022–2029.

20. Du CM, Valko K, Bevan C, Reynold D, AbrahamMH. 1998. Rapid gradient RP-HPLC method forlipophilicity determination: A solvation equationbased comparison with isocratic methods. AnalChem 70:4228–4234.

21. Abraham MH. 1993. Scales of solute hydrogenbonding: Their Construction and application tophysicochemical and biochemical processes. ChemSoc Rev 22:73–83.

22. Abraham MH, Chadha HS, Leo A. 1994. Hydrogenbonding XXXV. Relationship between high-performance liquid chromatography capacity fac-tors and water-octanol partition coefficients.J Chromatogr 685:203–211.

23. Abraham MH, Chadha HS, Leitao RAE, MitchellRC, Lambert WJ, Kaliszan R, Nasal A, Haber P.1997. Determination of solute lipophilicity, as logP(octanol) and log P(alkane) using poly(styrene-divinylbenzene) and immobilized artificial mem-brane stationary phases in reversed-phase high-peformance liquid chromatography. J ChromatogrA 766:35–47.

24. Abraham MH, Roses M, Poole CF, Poole SK. 1997.

Hydrogen bonding. 42. characterization of re-versed-phase high-performance liquid chromato-graphic C18 stationary phases. J Phys Org Chem10:358–368.

25. Abraham MH, Roses M. 1994. Hydrogen bonding.XXXVIII. Effect of solute structure and mobilephase composition on reversed-phase high-performance Liquid chromatographic capacity fac-tors. J Phys Org Chem 7:672–684.

26. Chadha HS, Abraham MH, Mitchell RC. 1994.Physicochemical analysis of the factors governingdistribution of solutes between blood and brain.Bioorg Med Lett 4:2511–2516.

27. Gratton JA, Abraham MH, Bradbury MW, ChadhaHS. 1997. Molecular factors influencing drugtransfer across the blood-brain barrier. J PharmPharmacol 49:1211–1216.

28. Abraham MH, Martins F, Mitchell RC. 1997. Algo-rithm for skin permeability using hydrogen bonddescriptors: The problem of steroids. J PharmPharmacol 49:858–865.

29. Valko K, Plass M, Bevan C, Reynolds D, AbrahamMH. 1998. Relationships between the chromato-graphic hydrophobicity indices and solute descrip-tors obtained by using several reversed-phase, diol,nitrile, cyclodextrin and immobilised artificialmembrane bonded high-performance liquid chro-matography columns. J Chromatogr 797:41–55.

30. Barbato F, Larotonda MI, Quaglia F. 1997. PharmRes 14(12):1699–1705.

31. Austin RP, Davis AM, Manners CN. 1995. Parti-tioning of ionizing molecules between aqueousbuffers and phospholipid vesicles. J Pharm Sci 84:1180–1183.

32. Ottiger C, Wunderli-Allenspach H. 1999. Immobi-lized artificial membrane (IAM)-HPLC for parti-tion studies of neutral and ionized acids and basesin comparison with the liposomal partition system.J Pharm Sci 16:643–650.

33. Barton P, Davis AM, McCarthy DJ, Webborn PJH.1997. Drug-phospholipid interactions. 2. Predict-ing the sites of drug distribution using n-octanol/water and membrane/water distributioncoefficients. J Pharm Sci 86:1034–1039.

1096 VALKO ET AL.


rapid-gradient hplc method for measuring drug interactions with immobilized artificial membrane:...

Documents