Characterization of biopartitioning micellar chromatography system using monolithic column by linear solvation energy relationship and application to predict blood–brain barrier penetration

Journal of Chromatography A, 1216 (2009) 5190–5198

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Characterization of biopartitioning micellar chromatography system using monolithic column by linear solvation energy relationship and application to predict blood–brain barrier penetration Rong Lu, Jin Sun, Yongjun Wang, Haiyan Li, Jianfang Liu, Liang Fang, Zhonggui He ∗ Department of Biopharmaceutics, School of Pharmacy, Shenyang Pharmaceutical University, Shenyang 110016, China

a r t i c l e

i n f o

Article history: Received 18 March 2009 Received in revised form 26 April 2009 Accepted 4 May 2009 Available online 13 May 2009 Keywords: Monolithic column Biopartitioning micellar chromatography Blood–brain barrier Linear solvation energy relationship

a b s t r a c t The linear solvation energy relationship (LSER) was applied to characterize biopartitioning micellar chromatography (BMC) system using monolithic column, and was utilized to compare the above system with other physicochemical and biological processes in this study. The solute volume and HB basicity had the maximum influence on the retention of the solutes, and an increase in the dipolarity/polarizability, HB basicity, HB acidity or excess molar refraction of the solutes decreased the retention. Principal component analysis of LSER coefficients showed that the system had certain similarity to drug biomembrane transport processes, such as blood–brain barrier penetration, transdermal and oral absorption. The quantitative retention–activity relationship (QRAR) of drug penetration across blood–brain barrier was established and its predictive capability for this biological process was evaluated. With the aid of the high flow rate, the monolithic column significantly facilitated the high-throughput analysis of large compounds’ bank without changing the mechanism of the retention in BMC and without impairing good predictive capability of the biological processes. Accordingly, the BMC system, together with monolithic column, allows for high-throughput profiling the biological processes, such as blood–brain barrier penetration. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Biopartitioning micellar chromatography (BMC), usually comprised of a C18 reversed stationary phase and polyoxyethylene (23) lauryl ether (Brij 35) mobile phase, is a kind of biopartitioning chromatography superior in describing the biological behavior of different kinds of drugs [1]. It has been successfully applied to mimic many biological processes, such as blood–brain barrier penetration, skin permeability and intestinal absorption [2–7]. The usefulness of BMC in constructing good predictive models can be attributed to the fact that the characteristics of the BMC system are similar to biological barriers and extracellular fluids [8]. The main disadvantage of this system is that the analysis rate of the strong hydrophobic solutes may be slow and usually a small amount of organic solvent needs to be added. Monolithic columns, prepared by sol–gel technology and composed of a single rod of silica-based material, have attracted considerable attention in liquid chromatography as they allow for

achieving good separation faster than the conventional particlepacked columns [9]. The highly porous monolithic rods of silica have revolutionary bimodal pore structures, that is, they can provide unique combination of macropores (2 ␮m) and mesopores (12 nm). The former allows rapid flow of the mobile phase at low pressure, while the latter creates the large uniform surface on which adsorption takes place, thereby enabling high performance chromatographic separation. With the emergence of commercial monolithic columns, the reports concerning practical applications of monolithic column in HPLC separations are increasing rapidly and its excellent performance has been confirmed [10–14]. In order to overcome the above-mentioned shortcomings and achieve a high-throughput screening in terms of BMC system, the monolithic column was introduced in this study for the first time. Linear solvation energy relationship (LSER) of Abraham assists in the interpretation and prediction of retention data in diverse chromatographic modes [15]. LSER is expressed as follows [16–18]:

SP = c + vV + sS + bB + aA + eE ∗ Corresponding author at: Mailbox 59#, Department of Biopharmaceutics, School of Pharmacy, Shenyang Pharmaceutical University, 103 Wenhua Road, Shenyang 110016, China. Tel.: +86 24 23986321; fax: +86 24 23986321. E-mail address: [email protected] (Z. He). 0021-9673/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.chroma.2009.05.007

where SP is the dependent solute property in a given system such as the logarithm of chromatographic retention factors (log k), the logarithm of octanol–water partition coefficients (log P), etc. The independent variables are solute descriptors: V is the solute

R. Lu et al. / J. Chromatogr. A 1216 (2009) 5190–5198

5191

Table 1 The Abraham descriptors and retention data of the compounds studied. No. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

Compounds a

Anisole Anthraceneb Benzenea Benzophenonea Biphenylc Bromobenzenea Bromonaphthalenec 4-Bromophenolc Catechola 3-Chloroanilinea 4-Chloroanilinea Chlorobenzanea 4-Chlorophenola 1,4-Dibromobenzened 1,2-Dichlorobenzened Diethyl phthalatec Ethylbenzenea Isopropylbenzened 2-Nitroanilinea 3-Nitroanilinea 4-Nitroanilinea Phenola Tetrabutylbenzened Thioureaa Toluenea p-Xylenea

log k1

log k2

log k3

V

S

B

A

E

1.00 1.34 1.03 1.18 1.34 1.21 1.30 0.95 0.48 0.85 0.84 1.20 0.92 1.36 1.23 1.04 1.35 1.44 0.80 0.75 0.68 0.66 1.50 −0.37 1.24 1.36

0.99 1.33 1.02 1.18 1.34 1.21 1.31 0.98 0.49 0.88 0.84 1.18 0.93 1.36 1.22 1.04 1.35 1.43 0.81 0.74 0.68 0.65 1.51 −0.36 1.22 1.36

1.01 1.35 1.02 1.20 1.36 1.23 1.31 0.97 0.51 0.90 0.86 1.20 0.94 1.37 1.24 1.06 1.35 1.45 0.83 0.76 0.70 0.68 1.52 −0.34 1.22 1.36

0.916 1.454 0.716 1.4808 1.324 0.891 1.26 0.95 0.8388 0.939 0.939 0.839 0.898 1.0664 0.9612 1.711 0.998 1.1391 0.9904 0.9904 0.9904 0.775 1.28 0.5696 0.857 0.998

0.75 1.34 0.52 1.5 0.99 0.73 1.13 1.17 1.1 1.1 1.1 0.65 1.08 0.86 0.78 1.4 0.51 0.49 1.37 1.71 1.91 0.89 0.49 0.82 0.52 0.52

0.29 0.28 0.14 0.5 0.26 0.09 0.13 0.2 0.47 0.36 0.35 0.07 0.2 0.04 0.04 0.88 0.15 0.16 0.36 0.35 0.38 0.3 0.18 0.87 0.14 0.16

0 0 0 0 0 0 0 0.67 0.88 0.3 0.3 0 0.67 0 0 0 0 0 0.3 0.4 0.42 0.6 0 0.77 0 0

0.708 2.29 0.61 1.447 1.36 0.882 1.598 1.08 0.97 1.05 1.06 0.718 0.915 1.15 0.872 0.729 0.613 0.602 1.18 1.2 1.22 0.805 0.619 0.84 0.601 0.613

log k1 : flow rate 1.0 mL min−1 ; log k2 : flow rate 2.0 mL min−1 ; log k3 : flow rate 3.0 mL min−1 . a Values taken from Ref. [43]. b Values taken from Ref. [44]. c Values taken from Ref. [45]. d Values provided kindly by Prof. Abraham.

McGowan volume in units of cm3 mol−1 /100, S is the polarizability/dipolarity, B is the overall hydrogen-bond basicity, A is the overall hydrogen-bond acidity, and E is an excess molar refraction. The coefficients v, s, b, a, e reflect the differences in the two phases between which the compound is transferred. v is a measure of the relative ease of forming a cavity; s is a measure of the difference in the capacity to take part in dipole–dipole and dipole-induced dipole interactions; b and a represent the differences in hydrogen-bond acidity and basicity, respectively; and e depends on the difference in the capacity to interact with solute n- or ␲-electrons for the solute. The term c comprised the constant contribution from the solutes and the system under consideration. The LSER has not only been applied to characterize diverse chromatographic systems, but also been utilized to study many physicochemical processes and biopartitioning processes [19–24]. However, only few LSER studies were found to be applied to characterize BMC in the literature [25]. In this study, the LSER was applied to characterize BMC system using monolithic column, and utilized to compare the system with other physicochemical and biological processes. In addition, the quantitative retention–activity relationship (QRAR) of drug penetration across blood–brain barrier by BMC system with monolithic column was established and its predictive capability for this biological process was originally evaluated and compared to the previous reports.

ratories of Shenyang Pharmaceutical University (Shenyang, China). The drugs used in this study were obtained from different sources: acetaminophen, acetylsalicylic acid, amidazophen, antipyrine, atenolol, atropine, caffeine, clonidine, ibuprofen, phenylbutazone, phenytoin, propranolol, salicylic acid, theophylline and verapamil were crude drug and provided by the pharmaceutical laboratories of Shenyang Pharmaceutical University (Shenyang, China); alprazolam, amitriptyline, amobarbital, carbamazepine, chlorpromazine, diazepam, midazolam, oxazepam and pentobarbita were standard substances and purchased from National Institute for the Control of Pharmaceutical and Biological Products (Beijing, China); imipramine, physostigmine and quinidine were standard substances and purchased from Sigma (St. Louis, MO, USA); haloperidol, mianserin, risperidone, tibolone, triazolam were the principal components in preparations available and extracted by methanol before use. The logarithm of the retention factor of 26 neutral compounds studied and their molecular descriptors are listed in Table 1. The logarithm of the brain–blood distribution coefficient (log BB), logarithm of the retention factor in BMC (log k) and physicochemical and structural descriptor values tested for log BB modeling are listed in Table 5. All stock solutions of the analytes were prepared with the concentration of about 1 mg mL−1 in methanol or water and diluted to about 0.1 mg mL−1 before injection with the corresponding mobile phase.

2. Experimental 2.1. Reagent and materials

Table 2 Correlation matrix for Abraham descriptors of the compounds studied.

Polyoxyethylene (23) lauryl ether (Brij 35) was of high purity grade and purchased from Sinopharm Chemical Reagent Co. (Shanghai, China). Sodium hydroxide and sodium dihydrogen phosphate were all of analytical grade and purchased from Bodi Chemicals Co. (Tianjin, China). The 26 neutral compounds were of reagent grade and generously donated by the pharmaceutical labo-

Descriptors

V

S

B

A

E

V S B A E

1.000 0.344 0.189 −0.490 0.446

1.000 0.466 0.348 0.643

1.000 0.414 0.077

1.000 0.016

1.000

5192

R. Lu et al. / J. Chromatogr. A 1216 (2009) 5190–5198

Table 3 LSER coefficients and the statistic of different flow rate (n = 26). Flow rate 1.0 2.0 3.0

C log k C log k C log k

V

S

B

A

E

Intercept

n

R2

F

P

1.080 (0.071) 1.104 (0.072) 1.094 (0.070)

−0.087 (0.047) −0.093 (0.048) −0.087 (0.046)

−1.284 (0.069) −1.290 (0.070) −1.270 (0.068)

−0.191 (0.062) −0.157 (0.063) −0.165 (0.061)

−0.165 (0.042) −0.163 (0.042) −0.161 (0.041)

0.564 (0.060) 0.539 (0.061) 0.551 (0.059)

26 26 26

0.985 (0.055) 0.984 (0.056) 0.985 (0.054)

256 241 256

<0.0001 <0.0001 <0.0001

2.2. Apparatus and conditions The chromatographic system was equipped with a LC-10AT constant flow pump (Shimadzu, Suzhou, China), a SPD-10A UV–Vis detector (Shimadzu) operating between 200 and 400 nm, a 7725i manual injection valve (Rheodyne, USA) and a HT-130 column heater (Hengao Instrument Co., Tianjin, China). Chromatographic signals were acquired and processed by Anastar chromatography data system (version 5.2) (Autoscience Instrument Co., Tianjin, China). A Chromolith Performance RP-18e (100 mm × 4.6 mm) column equipped with a Chromolith RP-18e guard cartridge (10 mm × 4.6 mm) (Merck, Germany) was used for separation. The mobile phases consisted of 0.04 mol L−1 polyoxyethylene (23) lauryl ether (Brij 35) with 0.01 mol L−1 sodium dihydrogen phosphate and were adjust to pH 7.4 by sodium hydroxide. The mobile phase was filtered through 0.45 ␮m nylon membranes before use. The mobile phase flow rates were 1.0, 2.0 and 3.0 mL min−1 with the logarithm of the retention factor expressed as log k1 , log k2 and log k3 . The detection wavelength was set at 210 nm. The column temperature was maintained at 40 ◦ C. Each value was determined in duplicate. 2.3. Statistical analysis

Fig. 1. Plot of residuals of log k values of 26 neutral solutes.

cating that the dipolarity/polarizability, HB acidity, HB basicity and excess molar refraction of the modified stationary phase were lower than those of the bulk mobile phase and an increase in the dipolarity/polarizability, HB basicity, HB acidity or excess molar refraction of solutes decreased the retention. The characteristic of the mono-

Excel 2003 and SAS software (version 9.0, SAS Institute, Cary, NC, USA) were used to perform data calculations and statistic analysis. 3. Results and discussion 3.1. Establishment and chemical interpretations of the LSER models The LSER equations of Abraham were established for BMC using monolithic column. All the test solutes were neutral and chemically diverse, and the values of log k and all Abraham descriptors spanned a wide range (Table 1). The correlation matrix is listed in Table 2. No significant covariances were found between the Abraham descriptors of the test solutes although a moderate covariance existed between the descriptor S and E. So the data sets were considered to be valid for LSER analysis. The coefficients of the LSER equations obtained at each flow rate are listed in Table 3. It was observed that the models fitted well for each system (R2 > 0.98), that is, the developed LSER models were able to reproduce accurately the experimental log k values for the solutes at all studied flow rates. The plots of residuals of log k versus the solute number showed a random distribution of the residuals with an average value practically equal to 0 (Fig. 1), proving the validity of each LSER model. The coefficients of LSER equations represent the characteristic of the chromatographic system. As shown in Table 3, the LSER equations at different flow rates were nearly the same, indicating that the flow rate did not influence the retention behavior of solutes in BMC system with monolithic column. The absolute values of coefficients v and b were the largest, suggesting the solute volume and HB basicity generally had the maximum influence on the retention of the solutes. The positive coefficients of v meant that creating a cavity was easier in stationary phase than in mobile phase and the increase of solute volume facilitated the retention. The coefficients s, b, a and e were all negative in all cases, indi-

Fig. 2. The score plot of PC2 against PC1 for the principal component analysis of normalized coefficients (pu ) of all studied systems.

Table 4 The original coefficients and normalized coefficients of all studied systems. Systems

v

s

b

a

e

s/v

b/v

a/v

e/v

vu

su

bu

au

eu

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

BMCmonolithic C18 log P [26] IAM [27] MELC [35] LEKC [28] MEKCSDS [29] MEKCSDS-Brij 35 [30] MLCSDS C8 [31] MLCCTAB C8 [31] log BB1 [32] log BB2 [21] log Kp [20] Human-%Abs [19] Rat-%Abs [33] CHI [34] log k50 [34] log kw [34]

1.080 3.814 2.468 0.901 2.580 2.860 2.829 1.720 1.010 0.995 0.861 2.296 10.600 1.920 68.40 1.207 2.069

−0.087 −1.054 −0.455 −0.645 −0.260 −0.310 −0.516 −0.300 −0.180 −0.687 −0.886 −0.473 4.100 10.400 −14.12 −0.280 −0.450

−1.284 −3.460 −2.303 −0.951 −3.090 −1.700 −2.699 −1.320 −1.060 −0.698 −0.666 −3.000 −21.100 −9.750 −70.19 −1.223 −1.955

−0.191 0.034 0.583 −0.922 0.330 −0.150 0.308 −0.340 0.140 −0.715 −0.724 −0.473 −21.700 −25.000 −25.40 −0.470 −0.290

−0.165 0.562 0.901 0.101 0.200 0.250 0.622 0.180 0.090 0.198 0.511 −0.106 2.940 −13.200 5.39 0.092 −0.029

−0.081 −0.276 −0.184 −0.716 −0.101 −0.108 −0.182 −0.174 −0.178 −0.690 −1.029 −0.206 0.387 5.417 −0.206 −0.232 −0.217

−1.189 −0.907 −0.933 −1.055 −1.198 −0.594 −0.954 −0.767 −1.050 −0.702 −0.774 −1.307 −1.991 −5.078 −1.026 −1.013 −0.945

−0.177 0.009 0.236 −1.023 0.128 −0.052 0.109 −0.198 0.139 −0.719 −0.841 −0.206 −2.047 −13.021 −0.371 −0.389 −0.140

−0.153 0.147 0.365 0.112 0.078 0.087 0.220 0.105 0.089 0.199 0.593 −0.046 0.277 −6.875 0.079 0.076 −0.014

0.636 0.721 0.691 0.521 0.637 0.853 0.706 0.774 0.680 0.629 0.519 0.598 0.327 0.061 0.668 0.668 0.714

−0.051 −0.199 −0.127 −0.373 −0.064 −0.092 −0.129 −0.135 −0.121 −0.435 −0.534 −0.123 0.126 0.328 −0.138 −0.155 −0.155

−0.756 −0.655 −0.645 −0.550 −0.763 −0.507 −0.674 −0.594 −0.714 −0.441 −0.401 −0.782 −0.650 −0.307 −0.686 −0.677 −0.675

−0.112 0.006 0.163 −0.533 0.081 −0.045 0.077 −0.153 0.094 −0.452 −0.436 −0.123 −0.668 −0.788 −0.248 −0.260 −0.100

−0.097 0.106 0.252 0.058 0.049 0.075 0.155 0.081 0.061 0.125 0.308 −0.028 0.091 −0.416 0.053 0.051 −0.010

PC1

Proportion of total variance

PC2

0.01 1.06 1.53 −0.60 0.82 0.70 1.21 0.59 0.98 −0.11 0.19 0.27 −1.95 −5.72 0.27 0.25 0.49

1.40 0.13 0.11 −1.47 1.24 −0.08 0.39 −0.04 0.85 −2.17 −3.16 1.07 0.21 0.58 0.26 0.16 0.53

56%

29%

BMC: biopartitioning micellar chromatography; log P: logarithm of the n-octanol/water partition coefficient; IAM: immobilized artificial membrane chromatography; MELC: microemulsion liquid chromatography; LEKC: liposome electrokinetic chromatography; MEKC: micellar electrokinetic chromatography; SDS: sodium dodecyl sulphate; CTAB: cetyltrimethyl ammonium bromide; MLC: micellar liquid chromatogramphy; log BB: logarithm of the brain–blood distribution coefficient; log Kp : logarithm of skin permeability coefficient; Human-%Abs: human intestinal absorption percentage; Rat-%Abs: rat intestinal absorption percentage; CHI: RP-HPLC, gradient chromatographic hydrophobicity index obtained from gradient retention time by calibration as close to the ϕ0 scale as possible; ϕ0 : isocratic chromatographic hydrophobicity index, defined as the volume percent of organic solvent at



which log k = 0; log kw : RP-HPLC, value of log k extrapolated to 0% organic modifier concentration; log k50 : RP-HPLC, value of log k measured by 50% (v/v) organic phase in the mobile phase; and pu = p/ where p presents any of the LSER coefficient and v, s, b, a, e are the specific coefficients of LSER equation for the considered system; the subscript u indicates the coefficients have been normalized.

R. Lu et al. / J. Chromatogr. A 1216 (2009) 5190–5198

No.

v2 + s2 + b2 + a2 + e2 ,

5193

5194

R. Lu et al. / J. Chromatogr. A 1216 (2009) 5190–5198

lithic column allowing for high flow rate facilitated the analysis rate of a large number of compounds without changing the mechanism of the retention for BMC system. 3.2. Comparison of BMC system using monolithic column with other physicochemical and biological processes The LSER model has been applied to characterize many biological and physicochemical processes since biological activities and chromatographic retention of solutes are based on the same basic intermolecular interaction forces, such as hydrophobic, electronic, steric effects and hydrogen bond. Setting up the relationships among these processes by LSER coefficients will benefit choosing the suitable high-throughput chromatographic systems to model some biological processes well. The LSERs of Abraham for sixteen systems were obtained from the literature [19–21,26–34]. The original and normalized coefficients of LSER of BMC system using monolithic column and obtained systems are listed in Table 4. The closeness of the LSER coefficients was able to indicate the ‘closeness’ of the chemical properties of the systems [17]. And principal component analysis (PCA) with the normalized coefficients (pu ) was performed in our study to estimate the similarities among different systems. The first two principal components, PC1 and PC2, accounted for 56% and 29% of the total variance of the LSER coefficients, respectively. A plot of the scores of PC2 against PC1 is shown in Fig. 2. It provided information of the relationship among different systems. Except for human-%Abs (13) and rat-%Abs (14) locating at left-side in Fig. 2, other systems located at the right-side indicating these systems were similar to each other. However, there were two distinct clusters observed in this domain, marked with circles. Cluster A included the BMC system in our studies (1), log P (2), IAM (3), LEKC (5), MEKC (6, 7), MLC (8, 9), log Kp (12) and RP-HPLC (15–17) systems. Cluster B included MELC (4) and two log BB (10, 11) systems. According to PC1, BMC was more similar to MLC and MELC than IAM among the studied chromatographic systems. This is not unexpected since BMC is a kind of MLC specified to the Brij 35 mobile phase and MELC can be considered as MLC in which some oils are added. The similar trend was also founded by Liu et al. [35]. BMC differs from IAM in that its surface is not modified by monolayers of phospholipid molecules but by absorbed surfactant monolayer. And IAM behaved more like to LEKC since phospholipid analog molecules are the main separation functional components in both systems. It should be pointed out that the BMC system using monolithic column was similar to the RP-HPLC systems both isocratic and gradient based on PCA. A fast gradient elution with C18 reversed phase chromatographic systems could have a comparable throughput with the monolithic columns. However, the BMC systems using monolithic columns are superior in modeling the diverse biological process of different kinds of drugs and environment friendliness to gradient elution with RP-HPLC systems [1]. These systems were similar to biological and physicochemical processes such as skin permeation, blood–brain barrier permeation and n-octanol/water partition. Therefore, they were utilized to model these processes instead of traditional animal models or other time-consuming methods and were developed into many high-throughput screening methods [3,5,36–38]. 3.3. Predicting drug penetration across the blood–brain barrier by BMC using monolithic column BMC using monolithic column was applied to predict drug penetration across the blood–brain barrier. Different molecular descriptors (logarithm of the n-octanol/water partition coefficient (log P), molar volume (MV), molecular weight (MW), polar surface area (PSA)) were introduced to obtain a better predictive equa-

tion (Table 5). Some compounds, which were reported to transport across cell membrane with the participation of some transporting proteins, were excluded from the drug set (for example verapamil, quinidine and risperidone). They were also proved to be outlier in the established models. Stepwise regression was utilized to select the suitable molecular descriptors during model development. When dealing with a large number of independent variables, it is of significant importance to determine the best combination of these variables to predict the dependent variable. Stepwise regression serves as a robust tool for the selection of the best subset models, i.e. the best combination of independent variables that best fits the dependent variable with considerably less computation [38]. In the present paper, the stepwise method was applied, and the significance level as entry and removal criteria was 0.30 and 0.10, respectively. 3.3.1. Establishment of model equations The retention of a drug in BMC involves the hydrophobic, electronic, hydrogen bond, and steric effects, and reflects adequately the extension of the biopartitioning process. So first, the correlation between log BB and log k1 was established and the equation was listed as follows: log BB = 0.655 (±0.064) log k1 − 0.330 (±0.069) N = 33, R2 = 0.769, SE = 0.296, F = 103, P < 0.0001

(1)

The result showed that the predictability of the equation was comparable to the results of Escuder-Gilabert et al. from BMC using Kromasil octadecyl-silane C18 columns (N = 41, R2 = 0.74, SE = 0.39, F = 109.6, P < 0.0001) [5]. Then stepwise regression was applied to introduce other molecular descriptors into the equations to obtain better predictability of log BB by BMC using monolithic column. Four descriptors were selected finally contributing to drug blood–brain barrier penetration. The equation was as follows: log BB = 0.751 (±0.091) log k1 − 0.088 (±0.043)HD −0.126 (±0.048) log P +

 0. 197 (±0.056)MW  100 2

−0.492 (±0.138) N = 33, R = 0.859, SE = 0.222, F = 43, P < 0.0001

(2)

The predictability was improved significantly compared with Eq. (1). The molecular electronic factor was not included in the equation, consistent with the results of Escuder-Gilabert et al. (the total molar charge added to the predicting equation did not improve the predictability much (N = 42, R2 = 0.75, SE = 0.39, F = 60, P < 0.0001)) [5]. The introduction of HD, log P and MW improved modeling blood–brain barrier penetration, since factors affecting BMC retention and blood–brain barrier penetration may differ to a certain extent. 3.3.2. Validation of the predictive equation The data set was split into three groups based on the variables of log k, HD, C log P and MW by cluster analysis using SAS software. 75% drugs of each group were randomly selected as training set; the others were assigned as test set. The grouping results are listed in Table 5. The regression equations of training set were: log BB = 0.657 (±0.074) log k1 − 0.356 (±0.081) N = 25, R2 = 0.776, SE = 0.284, F = 80, P < 0.0001

(3)

R. Lu et al. / J. Chromatogr. A 1216 (2009) 5190–5198

5195

Table 5 The logarithm of the brain–blood distribution coefficient (log BB), logarithm of the retention factor (log k) and physicochemical and structural descriptor values tested for log BB modeling. No.

Name

log BB

log k1

log k2

log k3

FRB

HA

HD

HB

log P

MV

MW

PSA

ıa

ıb

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

Acetaminophen Acetylsalicylic acid Alprazolam Amidazophen Amitriptyline Amobarbital Antipyrine Atenolol Atropine Benzene Caffeine Carbamazepine Chlorpromazine Clonidine Diazepam Ethylbenzene Haloperidol Ibuprofen Imipramine Mianserin Midazolam Dimethyl benzene Oxazepam Pentobarbita Phenylbutazone Phenytoin Physostigmine Propranolol Salicylic acid Theophylline Tibolone Toluene Triazolam Quinidine Risperidone Verapamil

−0.31 −0.5 0.044 0 0.886 0.04 −0.097 −0.87 −0.06 0.37 −0.055 −0.14 1.06 0.11 0.52 0.197 1.34 −0.18 1.06 0.99 0.36 0.295 0.61 0.12 −0.52 −0.04 0.079 0.64 −1.1 −0.29 0.4 0.37 0.74 −0.46 −0.02 −0.7

0.06 −0.71 0.89 0.26 2.12 0.88 0.15 −0.28 0.29 1.03 0.02 0.62 2.12 0.52 1.05 1.35 1.62 0.31 1.96 1.78 1.30 1.67 0.80 0.86 0.06 0.85 0.46 1.33 −0.37 −0.25 1.07 1.24 0.84 1.32 1.36 1.40

0.06 −0.70 0.90 0.24 2.01 0.89 0.10 −0.27 0.30 1.02 −0.03 0.63 2.05 0.51 1.05 1.35 1.66 0.34 1.96 1.74 1.28 1.36 0.81 0.87 0.05 0.85 0.42 1.30 −0.39 −0.26 1.07 1.22 0.85 1.32 1.38 1.39

0.06 −0.66 0.90 0.25 2.06 0.88 0.09 −0.25 0.28 1.02 0.00 0.64 2.08 0.51 1.06 1.35 1.66 0.36 1.94 1.78 1.30 1.36 0.82 0.88 0.08 0.87 0.45 1.33 −0.36 −0.22 1.09 1.22 0.87 1.35 1.45 1.46

2 3 1 2 3 4 1 9 6 0 0 0 4 1 1 1 7 4 4 0 1 0 2 4 5 2 2 7 2 0 1 0 1 5 4 13

3 4 4 4 1 5 3 5 4 0 6 3 2 3 3 0 3 2 2 2 3 0 4 5 4 4 5 3 3 6 2 0 4 4 6 6

2 1 0 0 0 2 0 4 1 0 0 2 0 2 0 0 1 1 0 0 0 0 2 2 0 2 1 2 2 1 1 0 0 1 0 0

5 5 4 4 1 7 3 9 5 0 6 5 2 5 3 0 4 3 2 2 3 0 6 7 4 6 6 5 5 7 3 0 4 5 6 6

0.339 1.19 2.499 0.758 4.92 2.053 0.268 0.097 1.528 2.218 −0.131 2.673 5.203 1.412 2.96 3.21 3.014 3.722 4.8 3.668 3.932 3.138 2.307 2.053 3.16 2.524 1.219 3.097 2.061 −0.175 4.025 2.678 2.659 3.44 2.885 3.899

120.9 139.5 225.5 196 257.7 211.4 162.7 236.6 242.2 89.4 133.3 186.5 262.9 153.1 225.8 122.2 303.2 200.3 269.2 223.6 239.8 121.9 201.8 209.1 262.7 200.5 236 237.1 100.3 122.9 274.1 105.7 234.8 266.3 296.8 429.3

151.16 180.16 308.76 231.29 277.4 226.27 188.23 266.34 289.37 78.11 194.19 236.27 318.86 230.09 284.74 106.17 375.86 206.28 280.41 264.36 325.77 106.17 286.71 226.27 308.37 252.27 275.35 259.34 138.12 180.16 312.45 92.14 343.21 324.42 410.48 454.6

49.3 63.6 38.1 26.8 3.24 75.3 23.5 84.6 49.8 0 53.5 46.3 31.8 36.4 32.7 0 40.5 37.3 6.48 6.48 25.3 0 61.7 75.3 40.6 58.2 44.8 41.5 57.5 69.3 37.3 0 38.1 45.6 61.9 64

0.003 1.000 0.000 0.000 0.000 0.224 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.999 0.000 0.000 0.000 0.000 0.000 0.249 0.999 0.105 0.000 0.000 1.000 0.059 0.000 0.000 0.000 0.000 0.000 0.000

0.000 0.000 0.000 0.001 0.984 0.000 0.000 0.983 0.997 0.000 0.000 0.000 0.990 0.834 0.000 0.000 0.876 0.000 0.992 0.876 0.014 0.000 0.000 0.000 0.000 0.000 0.893 0.982 0.000 0.000 0.000 0.000 0.000 0.987 0.756 0.974

pKa (A)

pKa (B)

9.86 3.48

1.72 2.28 4.5 9.18

7.94 13.88 14.11

13.94

13.85 4.41

0.7 9.16 9.98 0.73 −0.49 9.41 8.1 3.4 8.25 9.49 8.25 5.56 1.68

7.88 4.29 8.33 12.23 13.84 3.01 8.6 13.1

12.8

−0.29 −2.81 8.32 9.14 1.45

2.2 9.28 7.89 8.97

Set Test Train Test Test Train Train Test Train Train Test Train Train Train Train Train Train Train Train Test Train Train Train Train Test Train Train Train Train Train Train Train Train Test

FRB: free rotation bond; HA: hydrogen bond acceptor; HD: hydrogen bond donor; HB: hydrogen bond; log P: logarithm of the n-octanol/water partition coefficient; MV: molar volume; MW: molecular weight; PSA polar surface area; pKa (A) and pKa (B): dissociation constant for acidic and basic form, respectively. All the value above obtained from Scifinder database; ıa : mean net charge per molecule for acidic form; ıb : mean net charge per molecule for basic form; ıa = 1/(1 + 10(pKa -pH) ); ıb = 10(pKa -pH) /(1 + 10(pKa -pH) ). The log BB values taken from Ref. [46]. Table 6 Comparison of the model presented in this study with other chromatographic models presented in the literature. Method

BMC using monolithic column

BMC MELC IAM LEKC

Model

Solute set

Throughput

Reference

n = 33, structurally diverse compounds

High throughput

Current study

log BB = −0.87 (±0.12) + 0.84 (±0.08) log kBMC R2 = 0.74, SE = 0.39, F = 109.6, P < 0.0001 log BB = −0.19 (±0.08) + 2.42 (±0.24) log kMELC R2 = 0.772, SE = 0.399, F = 102, P < 0.0001 log BB = 0.58 log kIAM + 0.89I − 0.01V + 1.28 R2 = 0.848, SE = 0.27, F = 31.5

n = 41, structurally diverse compounds n = 32, structurally diverse compounds n = 21, structurally diverse compounds

Not been proved suitable for medium- to high-throughput operation

log BB = 0.827 (±0.085) log kLEKC + 0.163 (±0.066) R2 = 0.811, SE = 0.3103, F = 94.18, P < 0.0001

n = 24, structurally diverse compounds

High throughput

log BB = 0.655 (±0.064) log k1 − 0.330 (±0.069) R2 = 0.769, SE = 0.296, F = 103, P < 0.0001 log BB = 0.678 (±0.063) log k2 − 0.331 (±0.066) R2 = 0.788, SE = 0.258, F = 115, P < 0.0001 log BB = 0.681 (±0.063) log k3 − 0.346 (±0.067) R2 = 0.788, SE = 0.256, F = 116, P < 0.0001 log BB = 0.751 (±0.091) log k1 − 0.088 (±0.043)HD − 0.126 (±0.048) log P + 0.00197 (±0.00056)MW − 0.492 (±0.138) R2 = 0.859, SE = 0.222, F = 43, P < 0.0001 log BB = 0.793 (±0.089) log k2 − 0.095 (±0.040)HD − 0.136 (±0.045) log P + 0.00175 (±0.00052)MW − 0.426 (±0.131) R2 = 0.874, SE = 0.210, F = 49, P < 0.0001 log BB = 0.806 (±0.090) log k3 − 0.100 (±0.053)HD − 0.140 (±0.055) log P + 0.00172 (±0.00052)MW − 0.429 (±0.130) R2 = 0.876, SE = 0.228, F = 49, P < 0.0001

[5] [36] [39] [37]

5196

R. Lu et al. / J. Chromatogr. A 1216 (2009) 5190–5198

and log BB = 0.742 (±0.096) log k1 − 0.100 (±0.053)HD −0.140 (±0.055) log P +

 0.233 (±0.068)MW  100 2

−0.526 (±0.175) N = 25, R = 0.875, SE = 0.228, F = 35, P < 0.0001

(4)

Eqs. (3) and (4) were also similar to the Eqs. (1) and (2), respectively. The plots of experimental log BB versus predicted log BB for the training and test set are shown in Figs. 3 and 4. The predictability of both models was still good to the test set and so the developed models were robust, although the value of R2 decreased slightly. 3.3.3. Comparison of BMC system using monolithic column with other chromatographic systems and octanol/water partition system The log BB prediction equations were also established by BMC systems using monolithic column at higher flow rate. They are listed in Table 6 and compared with other systems usually used to model blood–brain barrier penetration. So far, lack of a simple and cost-effective in vitro model has been one of the outstanding issues. From Table 6, BMC, MELC, IAM, LEKC all had been applied to the prediction of drug blood–brain barrier penetration [5,36,37,39]. The models’ predictability was comparable with each other, except for IAM. In fact, the log BB correlated weakly with the log kIAM (r = 0.576) when I and V were excluded. Meanwhile, for

Fig. 4. Plot of predicted log BB versus experimental log BB for test set and training set (Eq. (4)).

Fig. 3. Plot of predicted log BB versus experimental log BB for test set and training set (Eq. (3)).

some strong hydrophobic compounds, extrapolation was needed to determine the k at 0% organic solvent using at least three different percentages of organic solvent in IAM. LEKC behaved the best as shown in Table 6, but the preparation of liposome was complex and time-consuming, and the reproducibility was still a question. The fast-screening and the similarity of the liposome’s composition to biomembrane were the evident strength of LEKC. MELC contains organic solvents, butanol, octanol, heptane, and so on, which were not environment friendly. The preparation of the microemulsion phase usually needed to stand still for a long time to achieve uniformity; the evaporation of the organic solvent usually made the retention of the compounds not stable along the time and the correction of retention factor with the proper internal standards was needed for MELC [36]. BMC with particle-based octadecyl-silane C18 columns took a long time when analyzing some strong hydrophobic compounds. However, the applications of monolithic column in BMC system can resolve this problem. In our study, three flow rates were studied and the good reproducibility of prediction equation was achieved. Detroyer et al. have evaluated “fast” micellar monolithic liquid chromatography for high-throughput quantitative structure retention relationship screening and found the retention mechanisms at very high flow rate, even at 9 mL min−1 , were not different from their application on classic particle-packed columns [40]. So the good ability of modeling many biological processes by BMC, in combination with the high-throughput analysis of monolithic column, will benefit the screening process of a range of new chemical entities to a large extent. The RP-HPLC using monolithic columns also benefit the high-throughput screening of lipophilicity of compounds but the

R. Lu et al. / J. Chromatogr. A 1216 (2009) 5190–5198

5197

Table 7 log P values and log k values in different chromatography systems of the test drugs. Compound

log BB [46]

log k1

log kIAM

log kBMC [5]

log kMELC [36]

log kLEKC [37]

log P

Alprazolam Amitriptyline Amobarbital Atenolol Carbamazepine Chlorpromazine Diazepam Ibuprofen Phenytoin Propranolol

0.044 0.886 0.040 −0.870 −0.140 1.060 0.520 −0.180 −0.040 0.640

0.889 2.123 0.880 −0.280 0.620 2.120 1.050 0.310 0.850 1.330

1.800 2.850 1.130 0.650 1.530 3.300 1.970 1.350 1.500 2.200

1.440 2.370 1.580 −0.340 1.210 2.430 1.650 1.190 1.540 1.620

0.025 0.438 0.150 −0.391 −0.044 0.439 0.199 0.050 0.119 0.274

0.090 0.710 −0.490 −1.000 −0.060 0.510 0.330 −0.500 0.130 0.770

2.499 4.920 2.053 0.097 2.673 5.203 2.960 3.722 2.524 3.097

log kIAM were determined in our laboratory on an IAM.PC.DD2 column with ammonium acetate buffer (pH 7.0) as the mobile phase.

other chromatographic methods, BMC using monolithic column will be a promising high-throughput screening method to model drug penetration across the blood–brain barrier and other biological membranes. Acknowledgement We are grateful to financial support from the National Natural Science Foundation of China, Nos. 30672556 and 30801443. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.chroma.2009.05.007. Fig. 5. The loading plot of PC2 against PC1 for the principal component analysis of studied systems listed in Table 7.

higher solvents consumption is a certain limitation of this approach [41]. A principal component analysis was also conducted with log BB, log kBMC-monolithic , log kBMC , log kMELC , log kIAM , log kLEKC and log P values of 10 drugs as variables (Table 7). A plot of the loadings of PC2 against PC1 is shown in Fig. 5. The first two principal components, PC1 and PC2, accounted for 90.9% and 4.2% of the total variance of the variables, respectively. The value of the loading reflected the importance of each original variable and the similar loadings indicated the main retention mechanisms of the systems might be the same [42]. As shown in Fig. 5, all the five systems had similar loadings to log P along PC1, indicating the partitions were mainly based on the solute’s hydrophobicity. Based on PC1 and PC2, BMC using monolithic column was closer to log BB, indicating the rationality for predicting log BB. Because the number of the objects (10, in this case the drug) was relatively small compared to the number of variables (7, here the systems), these PCA results were rough and suggestive. 4. Conclusions BMC using monolithic column was firstly characterized by LSER model and was compared against the other chromatographic systems and physicochemical or biological processes. The solute volume and HB basicity had the maximum influence on the retention of the solutes. The increase of the flow rate would not influence the mechanism of the retention. Principal component analysis with LSER coefficients indicated the possibility of BMC using monolithic column to simulate the in vivo biological processes. The promising BMC method using monolithic column to model blood–brain penetration was optimized and evaluated. The results indicated that the retention of drugs in this system was involved in reasonably hydrophobic, electronic and hydrogen bond contributions during the blood–brain penetration of drugs. In comparison with

References [1] J. Sun, X. Wu, R. Lu, J. Liu, Y. Wang, Z. He, Curr. Drug Metab. 9 (2008) 152. ˜ [2] M. Molero-Monfort, Y. Martín-Biosca, S. Sagrado, R.M. Villanueva-Camanas, M.J. Medina-Hernández, J. Chromatogr. A 870 (2000) 1. ˜ [3] J.J. Martínez-Pla, Y. Martín-Biosca, S. Sagrado, R.M. Villanueva-Camanas, M.J. Medina-Hernández, Biomed. Chromatogr. 17 (2003) 530. ˜ [4] J.J. Martínez-Pla, Y. Martín-Biosca, S. Sagrado, R.M. Villanueva-Camanas, M.J. Medina-Hernández, J. Chromatogr. A 1047 (2004) 255. ˜ [5] L. Escuder-Gilabert, M. Molero-Monfort, R.M. Villanueva-Camanas, S. Sagrado, M.J. Medina-Hernández, J. Chromatogr. B 807 (2004) 193. ˜ [6] M. Molero-Monfort, L. Escuder-Gilabert, R.M. Villanueva-Camanas, S. Sagrado, M.J. Medina-Hernández, J. Chromatogr. B 753 (2001) 225. ˜ [7] L. Escuder-Gilabert, J.J. Martínez-Pla, S. Sagrado, R.M. Villanueva-Camanas, M.J. Medina-Hernández, J. Chromatogr. B 797 (2003) 21. ˜ ˜ [8] C. Quinones-Torrelo, S. Sagrado, R.M. Villanueva-Camanas, M.J. MedinaHernández, J. Chromatogr. B 761 (2001) 13. [9] N. Tanak, H. Kobayashi, N. Ishizuka, H. Minakuchi, K. Nakanishi, K. Hosoya, T. Ikegami, J. Chromatogr. A 965 (2002) 35. [10] J. Liu, J. Sun, W. Zhang, K. Gao, Z. He, J. Pharm. Biomed. Anal. 46 (2008) 405. [11] A.H. Schmidt, J. Chromatogr. A 1073 (2005) 377. [12] M. Cledera-Castro, A. Santos-Montes, R. Izquierdo-Hornillos, J. Chromatogr. A 1087 (2005) 57. [13] L. Nováková, L. Matysová, D. Solichová, M.A. Koupparis, P. Solich, J. Chromatogr. B 813 (2004) 191. [14] K. Mc Fadden, J. Gillespie, B. Carney, D. O’Driscoll, J. Chromatogr. A 1120 (2006) 54. [15] J.R. Torres-Lapasió, M.J. Ruiz-Angel, M.C. García-Alvarez-Coque, M.H. Abraham, J. Chromatogr. A 1182 (2008) 176. [16] M.H. Abraham, A. Ibrahim, A.M. Zissimos, J. Chromatogr. A 1037 (2004) 29. [17] M. Gil-Agustí, J. Esteve-Romero, M.H. Abraham, J. Chromatogr. A 1117 (2006) 47. [18] M. Vitha, P.W. Carr, J. Chromatogr. A 1126 (2006) 143. [19] Y.H. Zhao, J. Le, M.H. Abraham, A. Hersey, P.J. Eddershaw, C.N. Luscombe, D. Butina, G. Beck, B. Sherborne, I. Cooper, J. Pharm. Sci. 90 (2001) 749. [20] M.H. Abraham, F. Martins, J. Pharm. Sci. 93 (2004) 1508. [21] J.A. Platts, M.H. Abraham, Y.H. Zhao, A. Hersey, L. Ijaz, D. Butina, Eur. J. Med. Chem. 36 (2001) 719. [22] M.H. Abraham, K. Takács-Novák, R.C. Mitchell, J. Pharm. Sci. 86 (1997) 310. [23] M.H. Abraham, Eur. J. Med. Chem. 39 (2004) 235. [24] Y.H. Zhao, M.H. Abraham, J. Le, A. Hersey, C.N. Luscombe, G. Beck, B. Sherborne, I. Cooper, Eur. J. Med. Chem. 38 (2003) 233. [25] R. Lu, J. Sun, Y. Wang, Z. He, Chromatographia, 2009, doi:10.1365/s10337-0091150-8. [26] M.H. Abraham, H.S. Chadha, G.S. Whiting, R.C. Mitchell, J. Pharm. Sci. 83 (1994) 1085. [27] J. Li, J. Sun, S. Cui, Z. He, J. Chromatogr. A 1132 (2006) 174. [28] Y. Wang, J. Sun, H. Liu, Y. Wang, Z. He, Chromatographia 65 (2007) 173.

5198

R. Lu et al. / J. Chromatogr. A 1216 (2009) 5190–5198

[29] M.D. Trone, M.G. Khaledi, Anal. Chem. 71 (1999) 1270. [30] M. Rosés, C. Ràfols, E. Bosch, A.M. Martínez, M.H. Abraham, J. Chromatogr. A 845 (1999) 217. [31] M.A. García, M.F. Vitha, J. Sandquist, K. Mulville, M.L. Marina, J. Chromatogr. A 918 (2001) 1. [32] M.H. Abraham, H.S. Chadha, R.C. Mitchell, J. Pharm. Sci. 83 (1994) 1257. [33] Y.H. Zhao, M.H. Abraham, A. Hersey, C.N. Luscombe, Eur. J. Med. Chem. 38 (2003) 939. [34] C.M. Du, K. Valko, C. Bevan, D. Reynolds, M.H. Abraham, Anal. Chem. 70 (1998) 4228. [35] J. Liu, J. Sun, Y. Wang, X. Liu, Y. Sun, H. Xu, Z. He, J. Chromatogr. A 1164 (2007) 129. [36] J. Liu, J. Sun, X. Sui, Y. Wang, Y. Hou, Z. He, J. Chromatogr. A 1198–1199 (2008) 164.

[37] [38] [39] [40] [41] [42] [43] [44] [45] [46]

Y. Wang, J. Sun, H. Liu, Z. He, Electrophoresis 28 (2007) 2391. Y. Wang, J. Sun, H. Liu, J. Liu, L. Zhang, K. Liu, Z. He, Analyst 134 (2009) 267. T. Salminen, A. Pulli, J. Taskinen, J. Pharm. Biomed. Anal. 15 (1997) 469. A. Detroyer, Y. Vander Heyden, K. Reynaert, D.L. Massart, Anal. Chem. 76 (2004) 1903. Z. Mrkvicková, P. Kovaríková, S. Balíková, J. Klimes, J. Pharm. Biomed. Anal. 48 (2008) 310. A. Detroyer, Y. Vander Heyden, I. Cambré, D.L. Massart, J. Chromatogr. A 986 (2003) 227. E. Lázaro, C. Ràfols, M.H. Abraham, M. Rosés, J. Med. Chem. 49 (2006) 4861. Y. Ishihama, N. Asakawa, J. Pharm. Sci. 88 (1999) 1305. L. Sprunger, B.H. Blake-Taylor, A. Wairegi, W.E. Acree Jr., M.H. Abraham, J. Chromatogr. A 1160 (2007) 235. P. Garg, J. Verma, J. Chem. Inf. Model 46 (2006) 289.