Article

pubs.acs.org/molecularpharmaceutics

pH-Dependent Solubility and Permeability Criteria for Provisional

Biopharmaceutics Classification (BCS and BDDCS) in Early Drug
Discovery
Manthena V. Varma,† Iain Gardner,‡ Stefanus J. Steyn,† Paul Nkansah,† Charles J. Rotter,†
Carrie Whitney-Pickett,† Hui Zhang,† Li Di,† Michael Cram,‡ Katherine S. Fenner,‡
and Ayman F. El-Kattan*,†
†
Pfizer Global Research and Development, Pfizer Inc., Groton, Connecticut 06340, United States
‡
Pfizer Global Research and Development, Pfizer Inc., Sandwich, Kent, CT13 9NJ, U.K.

ABSTRACT: The Biopharmaceutics Classification System (BCS) is a scientific

framework that provides a basis for predicting the oral absorption of drugs. These
concepts have been extended in the Biopharmaceutics Drug Disposition Classification
System (BDDCS) to explain the potential mechanism of drug clearance and
understand the effects of uptake and efflux transporters on absorption, distribution,
metabolism, and elimination. The objective of present work is to establish criteria for
provisional biopharmaceutics classification using pH-dependent passive permeability
and aqueous solubility data generated from high throughput screening methodologies
in drug discovery settings. The apparent permeability across monolayers of clonal cell
line of Madin−Darby canine kidney cells, selected for low endogenous efflux
transporter expression, was measured for a set of 105 drugs, with known BCS and
BDDCS class. The permeability at apical pH 6.5 for acidic drugs and at pH 7.4 for
nonacidic drugs showed a good correlation with the fraction absorbed in human (Fa).
Receiver operating characteristic (ROC) curve analysis was utilized to define the permeability class boundary. At permeability ≥5
× 10−6 cm/s, the accuracy of predicting Fa of ≥0.90 was 87%. Also, this cutoff showed more than 80% sensitivity and specificity
in predicting the literature permeability classes (BCS), and the metabolism classes (BDDCS). The equilibrium solubility of a
subset of 49 drugs was measured in pH 1.2 medium, pH 6.5 phosphate buffer, and in FaSSIF medium (pH 6.5). Although dose
was not considered, good concordance of the measured solubility with BCS and BDDCS solubility class was achieved, when
solubility at pH 1.2 was used for acidic compounds and FaSSIF solubility was used for basic, neutral, and zwitterionic
compounds. Using a cutoff of 200 μg/mL, the data set suggested a 93% sensitivity and 86% specificity in predicting both the BCS
and BDDCS solubility classes. In conclusion, this study identified pH-dependent permeability and solubility criteria that can be
used to assign provisional biopharmaceutics class at early stage of the drug discovery process. Additionally, such a classification
system will enable discovery scientists to assess the potential limiting factors to oral absorption, as well as help predict the drug
disposition mechanisms and potential drug−drug interactions.
KEYWORDS: Biopharmaceutics Classification System (BCS), Biopharmaceutics Drug Disposition Classification System (BDDCS),
solubility, intestinal permeability, high throughput screening

■ INTRODUCTION
Solubility and permeability are considered to be pivotal
water. A drug substance is considered highly permeable when
the extent of absorption (Fa) is ≥90% of oral dose.
properties that determine drug absorption following oral Permeability class can be determined by estimating the extent
administration. Based on these fundamental properties, of drug absorption in human oral pharmacokinetic studies in
Amidon et al. proposed a biopharmaceutics classification comparison to intravenous reference dose or mass-balance
system (BCS), which serves as a guide for regulatory and studies or by measuring the human effective permeability (Peff)
industrial purposes to waive conducting expensive bioequiva-
across the jejunum membrane.4,5 Alternatively, tools like in situ
lence clinical studies (BE) for high solubility-high permeability
(class I) drugs.1−3 According to the US Food and Drug rat intestinal perfusion and in vitro epithelial cell culture
Administration (FDA) guidance, 2 a drug substance is models, that are appropriately validated to predict the extent of
considered highly soluble when the highest dose strength is
soluble in 250 mL or less of aqueous medium over the pH Received: September 29, 2011
range of 1−7.5. The volume estimate of 250 mL is derived from Revised: March 23, 2012
typical BE study protocols that prescribe administration of a Accepted: April 10, 2012
drug product to fasting human volunteers with 250 mL of Published: April 10, 2012

© 2012 American Chemical Society 1199 dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212
Molecular Pharmaceutics Article

drug absorption in humans, can be used for permeability appropriate chemistry strategies during lead optimization and
classification.6 candidate selection. We determined pH-dependent solubility
Several high throughput (HT) screening methods for and permeability in a high throughput setting for a set of over
measuring permeability and solubility have been previously 100 drugs with known BCS and BDDCS class. Apparent
reported for BCS classification and estimating the oral permeability determined across monolayers of low-efflux
absorption of druglike molecules.7 Cell-based permeability Madin−Darby canine kidney (MDCK-LE) cells was validated
assays utilizing Caco-2, Madin−Darby canine kidney (MDCK), against human jejunum effective permeability (human Peff); and
and recently the low efflux-MDCK cell lines have been a pH-dependent permeability criterion was defined for
employed in discovery settings.8−11 Similarly, the parallel provisional biopharmaceutics classification, based on the
artificial membrane permeation assay (PAMPA) has been relationship between permeability and human intestinal
established as an alternative to cell-based assays for predicting absorption (human Fa). Equilibrium solubility in pH 1.2 for
oral absorption.12,13 Also, several HT solubility assays are used acids and in fasted-state simulated intestinal fluid (FaSSIF)
to screen large numbers of druglike molecules in early drug medium for nonacids (bases, neutrals, and zwitterions) was
discovery stage.14−17 In most cases, compounds are introduced used to define solubility boundary. The provisional biopharma-
as the dimethyl sulfoxide (DMSO) stock solution and the ceutics classification based on the current pH-dependent
nonthermodynamic solubility is estimated typically in pH buffer permeability and solubility criteria showed good agreement
at 7.4.7 However, the solubility data generated under with both BCS and BDDCS.
thermodynamic equilibrium conditions represents the best
case scenario.14 The solubility is quantified utilizing a UV plate
reader and liquid chromatography with UV or mass
■ MATERIALS AND METHODS
Materials. All drugs or test compounds were obtained from
spectrometry, after equilibration and filtration, or determined Pfizer Global Material Management (Groton, CT) or
in situ by turbidimetric or nephelometric methods.16−18 purchased from Sigma-Aldrich (St. Louis, MO). Minimum
Although solubility and permeability data are often generated essential medium (MEM) with L-glutamine, ribo/deoxyribo
in HT format, a clear strategy of how to use the data to assign nucleosides, heat inactivated fetal bovine serum (FBS),
compounds into biopharmaceutics classes has not been nonessential amino acids (NEAA), penicillin−streptomycin
presented in discovery settings. (Pen/Strep), L-glutamine and 0.25% trypsin−EDTA, Hanks
While the pharmaceutical industry has taken advantage of balanced salt solution with CaCl2, D-glucose, HEPES, and
BCS-based biowaivers, its principles are used throughout drug MgCl2 were purchased from Gibco Laboratories (Grand Island,
discovery and development to drive oral active programs.6 On NY). The 96-Transwell insert plates with polyethylene
the basis of the apparent correlation between intestinal terephthalate membrane (1 μm pore size, including 96-well
permeability and extent of drug metabolism, Benet and co- membrane inserts and feeder tray), 96-well angled bottom
workers proposed the Biopharmaceutics Drug Disposition collection plates, and velocity V11 peelable seals were
Classification System (BDDCS), where drugs are categorized in purchased from BD Falcon (Bedford, MA). All other reagents
terms of the extent of metabolism and solubility.19−21 The were of analytical grade.
group noted that the major route of elimination in humans for a Cell Culture and High Throughput Permeability
majority of high-permeability class I and class II drugs was Assay. The cell culturing conditions for in-house low efflux
metabolism, predominantly cytochrome P450-mediated, with transporter MDCK (MDCK-LE) cell line (clonal cells isolated
an extent of metabolism ≥70%; while the major route of from Madin−Darby canine kidney cells, selected for low
elimination for the poorly permeable (classes III and IV) drugs endogenous efflux transporter expression) were discussed
was renal and/or biliary excretion of unchanged drug with an elsewhere.10 Briefly, MDCK-LE cells were cultured at 37 °C,
overall extent of metabolism ≤30%. Interestingly, most drugs 5% CO2, 95% relative humidity in minimum essential medium
are either very highly metabolized or very poorly metabolized, that contained 10% FBS, 1% NEAA, 100 U/mL penicillin, 100
and a relatively few drugs showed the extent of metabolism μg/mL streptomycin, and 1% L-glutamine. Cells were passaged
between 30% and 70%. Based on the established concordance each week at about 90%, confluency. Cells were trypsinized and
between the intestinal permeability and the extent of drug resuspended in complete media to obtain a cell suspension of
metabolism, if an efficient measure of intestinal permeability 2.5 × 105 cells/mL, and then plated onto 96-well membrane
rate is identified and the relationship established, it would be inserts, with each insert receiving a volume of 75 μL. The
possible to use these permeability values to predict the major inserts were placed into a feeder tray containing complete
route of elimination for new molecular entities (NME) in growth medium. Plates were used on day 4 for transport
humans in early discovery. studies.
The objective of the present study was to determine if the Cell inserts were washed with prewarmed transport buffer
data generated using HT permeability and solubility assays before the experiment. Monodirectional transport studies were
could be used early in the drug discovery process to provide performed with 2 μM drug solution in transport buffer with
provisional biopharmaceutics classification for new molecular 0.1% DMSO. Drug solution was added to the donor wells and
entities (NMEs). This will be pivotal in enabling discovery buffer was added the receiver wells to initiate the transport
teams to flag potential liabilities for oral absorption and predict assay (in-house screening code: RRCKG_01). The plates were
the clearance mechanism and the potential for drug−drug incubated at 37 °C, and samples from both the donor and
interaction. Even though there are many of unknowns at the receiver were taken at time 0 min and 90 min for analysis.
early stage of drug discovery, such as dose, dissolution rate, final Permeability was determined at donor pH 6.5 and 7.4, with
solid form, clearance, route of elimination, etc., the provisional receiver medium pH always 7.4.
biopharmaceutics classification may help project teams calibrate Drugs were quantified in the samples using LC−MS/MS
NME against BCS and BDDCS. Additionally, such a methodology. Pfizer research compound (CP-628374) was
classification scheme can be used to guide teams to develop used as an internal standard for LC−MS in both positive and
1200 dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212
Molecular Pharmaceutics Article

negative ionization modes. LC−MS/MS analysis was con- where area is the surface area of the cell monolayer (0.0625
ducted on a Sciex Triple Quad 4000 mass spectrometer cm2), CD(0) the initial concentration of compound applied to
(turbospray ionization source) with a Shimadzu LC-10 HPLC the donor chamber, and Mr the mass of drug in the receiver
system and Gilson 215 autosampler. The mass spectrometer compartment, at time t.
was controlled by Analyst 1.4.2 software. The Gilson The complete radial mixing (parallel tube) model, as
autosampler was independently controlled by Gilson 735 suggested by Fagerholm et al., was used to predict the
software and synchronized to Analyst via contact closure. The human Fa from human effective permeability (Peff) and
LC method consisted of a step gradient with 25 μL samples apparent MDCK-LE permeability (Papp).22,23
loaded onto a 1.5 × 5 mm Showadenko ODP 13 μm particle Tres
size column using 95% ammonium acetate buffer (2 mM), 2.5% Fa = 1 − e−(2APapp r ·2.8) (2)
methanol and 2.5% acetonitrile. Samples were eluted with 10% where Tres and r are the average human small intestinal transit
ammonium acetate buffer (2 mM), 45% methanol, and 45% time and radius, and are assumed to be 3 h and 1.75 cm,
acetonitrile. MS/MS parameters utilized either negative or respectively. A is a constant, defining the relationship between
positive ionization mode depending on the compound. Peak human effective permeability and the measured apparent
area counts of analyte compound and internal standard were permeability, obtained by fitting the data. Correction factor, f-
integrated using DiscoveryQuant Analyze as an add-on to value was set at 2.8.22
Analyst 1.4.2. Binary Classification and Receiver Operating Charac-
High Throughput Equilibrium Solubility Assay. Equi- teristic (ROC) Analysis. Sensitivity, specificity, and accuracy
librium solubility of 49 drugs was measured in simulated gastric are statistical measures of the performance of a binary
fluid without pepsin (SGF) at pH 1.2, in 50 mM phosphate classification test and are measured as shown in eqs 3 and 4.
buffer (PBS) at pH 6.5, and in fasted-state simulated intestinal Sensitivity measures the ratio of actual positives that are
fluid (FaSSIF) at pH 6.5. An in-house fully automated HT correctly identified, whereas specificity is defined as the ratio of
method (screen code: SW_BIORE_SOL) was used to perform negatives selected by the test to the true negatives.
all solubility and pH measurements. The HT platform true positives
consisted of an integrated robotic system with a powder sensitivity =
true positives + false negatives (3)
dispensing and liquid handling of compounds, a robotic arm for
plate transfer between robots and peripherals such as heater/ true negatives
cooler shakers, vacuum filtration unit, plate sealer, GR4 specificity =
true negatives + false positives (4)
centrifuge, and a customized Microlab probe for pH measure-
ment. The protocol used for the solubility studies involved ROC curve analysis was used to determine a permeability cutoff
weighing out each solid compound in triplicate onto sample for high human Fa and test the predictability of the obtained
plates using the powder dispenser. To each sample one of three cutoff value. The ROC curve depends on the calculation of
media (SGF pH 1.2, PBS pH 6.5 and FeSSIF pH 6.5) was sensitivity and specificity.24 The area under the curve (AUC) of
added. Sample plates were then sealed and shaken for 18 h at a ROC curve can be used as a diagnostic of the performance of
room temperature. After incubation, the insoluble portion was the test, where an AUC of 1 would mean that the test is 100%
removed through 0.5 μm filtration plates. The sample plates are specific and sensitive, while an AUC of 0.5 or below indicates
unsealed, and the final pH of samples was measured. The random prediction. In general, an AUC value equal to or larger
filtrates were diluted, and an autosampler delivered samples to a than 0.8 is usually regarded as acceptable.24,25 For obtaining
LC−MS single-quadrupole (SQ) mass detector to determine permeability boundary, we defined human Fa greater than or
the concentrations of dissolved drug. The equilibrium solubility equal to 80% or 90% as true positives and human Fa less than
of the test compound was determined against a calibration 80% or 90% as true negatives. The point that has the shortest
curve. For versatility and to increase the throughput the LC was distance to zero on x-axis and one on y-axis (i.e., the point
in-turned coupled with multiwavelength UV (λ = 210−400 having both sensitivity and specificity of 1) represented the
cutoff value.

■
nm) plate reader for recording the full UV absorption spectra of
samples passing through the LC detector. Crystallinity of the
undissolved drug after incubation was determined using
RESULTS
polarized light microscopy (PLM). In order to balance sample Data Set. A list of 130 drugs and compounds, for which
consumption, throughput, analytical detection, and data quality, BCS class was assessed and listed by Custodio et al.,26 was the
the current HT equilibrium solubility screen was developed and initial source for the current work. When a drug was listed in
validated with a focus on low soluble compounds typically more than one class, the most appropriate class was assigned
encountered in early drug discovery space. Consequently the based on further literature assessment. Except for compounds
screen was designed for a dynamic range of 0.3−300 μg/mL in like glucose and phenylalanine, the majority of the drugs were
all media. Therefore, in cases where the measured solubility evaluated for permeability using the in-house MDCK-LE
exceeded the limits of the assay, they were reported with the model.10 Permeability data for 105 drugs was available after
qualifier, “≤0.3 μg/mL” or “≥300 μg/mL”. considering the assay recovery (70−120%) and reproducibility
(coefficient of variance <30%). Of these drugs, human Fa and
Data Analysis. The absorptive (AP-to-BL) transepithelial
human Peff values were available for 97 and 21 drugs,
transport was represented as permeability value (Papp, cm/s)
respectively.4,27,28 BDDCS class for 101 compounds was
calculated using the following equation:10
adopted from Benet et al.;29 while data on extent of parent
1 dM r excreted in urine and bile was taken from the published
Papp = × compilations.30,31 A subset of 50 drugs was selected from the
area × C D(0) dt (1) above list of 105 drugs for solubility assessment. While the
1201 dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212
 Table 1. Summary of MDCK-LE Apparent Permeability, Equilibrium Solubility, and Provisional Class, along with the Literature Data on the Human Fa, Renal and Biliary
Excretion, and BCS and BDDCS Classes of 105 Drugs
class MDCK-LE Papp solubility % parent excreted
ionic pH max. dose human % human in urine in bile in bile
name BCSa BDDCSa provisional state 6.5 pH 7.4 sel.b cryst.c pH 1.2 pH 6.5 FaSSIF sel.d strength Peffa Faa (human)a (human)a (rat)a
acetaminophen I I N 11.8 12.3 12.3 1000 82 3.0 0.7 1.0
acyclovir III IV III N 0.6 0.3 0.3 −1 ≥300 ≥300 ≥300 ≥300 800 22 76.0
amiloride III III III N 4.5 3.8 3.8 1 ≥300 200 268 268 5 1.6 90 49.0
amitriptyline I I III B 2.2 4.1 4.1 150 95 2.0
Molecular Pharmaceutics

−1 ≥300 ≥300 ≥300 ≥300

antipyrine I I I N 28.8 31.1 31.1 −1 ≥300 ≥300 ≥300 ≥300 500 5.6 97 3.8
atenolol III III III B 1.5 1.5 −1 ≥300 ≥300 ≥300 ≥300 100 0.15 50 100.0
atorvastatin II II IV A 4.8 2.0 4.8 1 3.7 271 300 3.7 80
atropine I III B 3.1 6.1 6.1 0.4 98 57.0
azithromycin II III B 0.5 0.7 0.7 600 37 4.5 9.6
bidisomide III B 0.9 0.8 0.8
buspirone I II B 19.3 27.6 27.6 10 80
caffeine I I N 23.8 31.2 31.2 65 100 1.2
carbamazepine II II II N 38.2 38.7 38.7 1 94 169 198 198 300 4.3 100 1.4
carvedilol II II B 0.7 1.2 1.2 25 65 1.3
cetirizine III III Z 3.8 2.9 3.0 10 70
chloroquine I III I B 1.0 5.1 5.1 0 ≥300 ≥300 ≥300 ≥300 500 100 56.0
chlorothiazide IV IV IV A 1.1 1.3 1.1 1 193 ≥300 ≥300 193 500 13

1202
chlorpheniramine I I I B 16.3 28.2 28.2 −1 ≥300 ≥300 ≥300 ≥300 4 94 17.0
chlorpromazine II I B 0.5 10.2 10.2 200 96 1.0
cimetidine III III III B 0.8 0.8 0.8 −1 ≥300 ≥300 ≥300 ≥300 800 0.26 68 97.5 1.8
ciprofloxacin III IV Z 0.9 1.4 1.4 750 69 83.1 0.5 9.9
cisapride II II B 3.4 7.6 7.6 20 99
danazol II II II N 4.7 7.6 7.6 1 ≤0.3 ≤0.3 6.4 6.4 200
desipramine I I I B 2.7 7.0 7.0 0 ≥300 ≥300 ≥300 ≥300 200 4.5 100 1.8
diazepam I I N 36.0 36.0 10 100 0.0 15.0 0.0
diclofenac II I A 23.4 17.6 23.4 50 100 1.0 1.0 1.3
dicloxacillin III III A 2.1 1.4 2.1 500 100 60.0
diflunisal II II A 31.4 22.6 31.4 500 100
digoxin II III IV N 1.3 1.5 1.5 1 2.3 29 29 29 0.25 81 100.0 30.6 31.0
diltiazem I I I B 19.2 25.6 25.6 −1 ≥300 ≥300 ≥300 ≥300 120 90 0.7
diphenhydramine I I B 31.0 31.0 50 100 2.0
disopyramide I III B 1.5 2.4 2.4 150 83 55.0
doxycycline I III III Z 3.7 1.8 1.8 0 ≥300 ≥300 ≥300 ≥300 40 100 41.0
enalapril I I A 2.7 1.2 2.7 20 63
ephedrine I B 6.0 22.5 22.5
ergonovine BDDCS I I B 5.1 20.1 20.1 0.2
erythromycin III III III B 0.6 0.7 0.7 −1 ≤0.3 ≥300 ≥300 ≥300 500 35 6.0 12.6 34.0
ethambutol III III III B 0.2 0.8 0.8 −1 298 ≥300 ≥300 ≥300 400 80
ethinylestradiol I I N 28.6 28.6 1 100
Article

famotidine III III III B 1.1 0.7 0.7 −1 50 ≥300 ≥300 ≥300 40 50 67.0 0.1

dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212

 Table 1. continued
class MDCK-LE Papp solubility % parent excreted
ionic pH max. dose human % human in urine in bile in bile
name BCSa BDDCSa provisional state 6.5 pH 7.4 sel.b cryst.c pH 1.2 pH 6.5 FaSSIF sel.d strength Peffa Faa (human)a (human)a (rat)a
fexofenadine III III Z 0.5 0.5 180 0.07 30 40.3
fluoxetine I I B 1.1 5.0 5.0 20 80
flurbiprofen II II II A 27.4 15.8 27.4 1 7.3 ≥300 ≥300 7.3 100 95
fluvastatin I I A 15.0 11.5 15.0 40 2.4 100 20.0
furosemide IV IV IV A 1.9 0.7 1.1 1 7.7 7.7 80 0.3 50 70.8 1.2
Molecular Pharmaceutics

≥300 ≥300
ganciclovir III III N 0.6 0.5 0.5 500 9 90.0
glipizide II II II A 14.4 5.6 14.4 1 0.8 31 275 0.8 10 100 5.0
glyburide II II II A 27.1 17.1 27.1 1 ≤0.3 18 ≥300 0.3 6 100 0.0
griseofulvin II II II N 25.2 30.4 30.4 1 10.5 11.8 17.4 17.4 500
hydrochlorothiazide III III III N 1.3 0.9 0.9 1 137 ≥300 ≥300 ≥300 50 0.04 55
ibuprofen II II A 35.2 30.5 35.2 800 95 1.0
imipramine I I B 8.5 18.3 18.3 50 100 2.0
indinavir II II N 2.9 3.4 3.4 400 63 8.5
indomethacin II II II A 32.0 19.7 32.0 1 0.4 234 ≥300 0.4 50 100 15.0 2.0
itraconazole II II B 15.0 15.0 100 85 0.0
ketoconazole II II IV B 2.8 4.2 4.3 −1 ≥300 71 25 25 200 75
ketoprofen II II II A 32.4 24.3 32.4 1 74 ≥300 ≥300 74 75 8.5 92 3.0
labetalol I I B 13.0 13.0 300 95 4.0

1203
lansoprazole II II N 23.4 23.4 30 85 1.0
levodopa I I I Z 166.1 166.8 −1 ≥300 ≥300 ≥300 ≥300 250 3.4 100 1.0
levofloxacin I III Z 2.7 2.7 750 100 74.0 10.1
lidocaine I I B 21.3 44.7 44.7 20 2.0
lisinopril III III III Z 0.3 0.5 0.5 −1 ≥300 ≥300 ≥300 ≥300 40 0.33 35 100.0
losartan III II III A 1.8 0.9 1.8 −1 ≥300 ≥300 ≥300 ≥300 100 1.15 90 11.0
lovastatin II II N 2.6 10.7 10.7 40 31 10.0
mebendazole IV II N 23.0 24.8 24.8 100 100 1.1
methotrexate IV III III A 1.6 2.1 1.6 0 ≥300 ≥300 ≥300 ≥300 15 65 81.0 14.0 72.0
methyldopa III III III Z 1.2 0.9 1.2 0 ≥300 ≥300 ≥300 ≥300 500 0.1 43 60.0
metoclopromide III III B 10.1 23.5 23.5 10 100 20.0
metoprolol I I I B 8.4 25.5 19.0 −1 ≥300 ≥300 ≥300 ≥300 100 1.5 98 5.0
metronidazole I I N 18.4 11.4 18.4 500 100 22.4
midazolam I I N 27.0 27.0 2 90 0.0
minocycline I I Z 4.9 1.8 4.9 100 100
nadolol III III III B 1.5 1.5 0 ≥300 ≥300 ≥300 ≥300 160 57 62.1
naproxen II II II A 28.0 36.0 28.0 1 7.4 ≥300 ≥300 7.4 500 8.5 100
nifedipine II II II N 18.1 88.9 88.9 1 4.6 10.6 16.1 16.1 20 95 0.0
ofloxacin II III Z 4.6 4.6 400 100 77.1
oxaprozin II II A 36.4 23.9 36.4 600 98
phenazopyridine II IV N 20.0 20.1 20.2 200 90
phenobarbital I I A 50.0 21.5 50.0 60 100 24.0
Article

phenytoin II II II N 24.8 28.2 27.2 1 20 33 82 82 300 90 0.3

dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212

 Table 1. continued
class MDCK-LE Papp solubility % parent excreted
ionic pH max. dose human % human in urine in bile in bile
name BCSa BDDCSa provisional state 6.5 pH 7.4 sel.b cryst.c pH 1.2 pH 6.5 FaSSIF sel.d strength Peffa Faa (human)a (human)a (rat)a
piroxicam II II II A 66.4 32.5 66.4 1 107 120 ≥300 107 20 10.2 100
pravastatin (acid) III III A 3.8 2.1 3.8 80 34 45.0 83.0
prednisolone I I I N 5.2 5.5 5.5 1 273 243 ≥300 ≥300 5 99 20.7
prednisone I II II N 11.7 14.9 14.9 1 160 175 117 117 50 95 1.8
primaquine I I I B 5.4 10.9 16.0 15 96
Molecular Pharmaceutics

−1 ≥300 ≥300 ≥300 ≥300

promazine I I B 3.1 3.9 3.9 100 40
propranolol I I I B 15.0 15.0 −1 ≥300 ≥300 ≥300 ≥300 80 2.91 99 10.0
quinidine I I B 6.8 17.4 17.4 300 81 18.0
qunidine I I B 7.9 17.2 17.2 260 81 12.0
ranitidine III III III B 0.8 0.7 0.7 −1 ≥300 ≥300 ≥300 ≥300 300 0.27 55 77.1 1.5
rifampcin II II Z 12.0 12.0 600
ritonavir II II II N 4.9 8.4 8.4 1 ≥300 4.1 6.6 6.6 100 68 1.5
rosiglitazone I I II Z 34.3 30.3 34.3 −1 ≥300 6.3 11.2 11.2 8 100 0.0
saquinavir II II B 0.8 0.9 0.9 500 80
tacrolimus II II N 4.9 6.1 6.1 5 15 1.0
talinolol II III III B 0.6 0.8 0.9 0 ≥300 ≥300 ≥300 ≥300 100 65 53.0
tamoxifen BDDCS I II B 94.4 69.9 69.9 20 100
tetracycline III III Z 0.7 0.7 25 90 60.0

1204
theophylline I I N 17.1 11.4 17.1 600 100 10.0
trimethoprim III III B 9.6 12.6 12.6 160 97 44.0
valsartan III IV A 0.5 0.6 0.6 320 55 29.0 88.0 70.0
verapamil I I I B 13.0 13.0 −1 169 ≥300 ≥300 ≥300 120 6.8 100 3.0
warfarin II II II A 46.5 29.7 46.5 1 5.7 223 ≥300 5.7 10 98 2.0
zidovudine I I N 6.1 6.1 300 100 22.7
a
Refer to Data Set under Results for the source of the data. bMDCK-LE Papp selected (×10−6 cm/s) represents apparent permeability values at apical pH 6.5 and pH 7.4 for acidic and nonacidic drugs,
respectively. cCrystallinity (cryst.) represents solid form of the undissolved compound after 18 h incubation, as determine using polarized light microscopy. (Key: +1, crystalline; 0, mixture of amorphous
and crystalline; −1, amorphous.) dSolubility selected (sel., μg/mL) represents equilibrium solubility in pH 1.2 (SGF) and FaSSIF media for acidic and nonacidic drugs, respectively.
Article

dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212

Molecular Pharmaceutics Article

Figure 1. Correlation between human fraction absorbed vs MDCK-LE apparent permeability (A, B) and human effective permeability (C, D). Data
set included all BCS I−IV class drugs (A, C); or included only BCS I and III class drugs, where solubility is not limiting human Fa (B, D). Solid line
is data fit to the complete radial mixing model, Fa = 1 − e−(2APapp)(Tres/r)2.8). A-value is constraint to 1.0 in case of human effective permeability and Fa
relationships.22 R2, regression coefficient; A, constant; n, number of drugs in the data set.

major criterion for inclusion into the subset included the a significant log−log correlation (p < 0.01, R2 = 0.75) with the
availability of drugs in the solid form, care was taken to cover a human effective permeability. The regression fit between
wide range of Fa (9−100%), human Peff (0.04−10.2 × 10−4 MDCK-LE permeability and human Fa of 97 compounds in
cm/s), oral dose (0.2−1000 mg) and the in-house MDCK-LE the data set suggested that Fa can be predicted from eq 2, with
permeability (0.3−166.8 × 10−6 cm/s) (Table 1). Finally, a correlation scalar “A-value” of 22.7 ± 2.4 (R2 = 0.38; n = 97; p
mebendazole was excluded due to related analytical sensitivity < 0.01). A similar A-value (24.0 ± 2.7) with improved fit to the
issues in the HT settings, and pH-dependent equilibrium data (R2 = 0.53; n = 61; p < 0.01) was noted when the drugs of
solubility data was available for 49 drugs. BCS classes II and IV, with possible solubility rate-limited
pH-Dependent Permeability Criterion. The apparent absorption, were excluded. Overall, the quantitative predict-
permeability values at apical pH 6.5 ranged from 0.2 to 94.4 × ability of human Fa utilizing the pH-dependent MDCK-LE
10−6 cm/s with ethambutol showing the least and tamoxifen permeability (Figure 1A,B) was comparable to the predictability
the maximum permeability, while the permeability values at of human effective permeability (Figure 1C,D).
apical pH 7.4 ranged from 0.3 to 166.1 × 10−6 cm/s, for Receiver operating characteristic (ROC) curve analysis of 97
acylovir and levodopa, respectively (Table 1). In general, acidic compounds, for which human Fa is available, suggested that a
drugs showed higher permeability at donor pH 6.5, whereas MDCK-LE permeability cutoff of 5 × 10−6 cm/s can be used to
bases, neutrals, and zwitterions showed higher or comparable predict drugs with high Fa (Figure 2). The area under the ROC
permeability at pH 7.4. A good correlation between apparent curve was about 0.87 with human Fa higher limit defined at
permeability and human Fa was observed, when the either ≥90% or ≥80%, indicating a good relationship between
permeability values at apical pH 6.5 and pH 7.4 were used the MDCK-LE permeability and human Fa. It is interesting to
for acidic and nonacidic drugs, respectively (Figure 1). Under note that both the limits of human Fa yielded the apparent
these conditions, the apparent permeability values also showed permeability boundary of 5 × 10−6 cm/s. Good concordance
1205 dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212
Molecular Pharmaceutics Article

49 compounds were suggested different class in BDDCS, in

comparison to BCS.
Provisional Biopharmaceutics Classification. Drugs
listed in the data set were classified on the basis of pH-
dependent MDCK-LE permeability and equilibrium solubility
criteria (Figure 5) and compared to the literature BCS and
BDDCS classes (Figure 6). Of the 49 drugs included in this
study, 40 (82%) and 41 (84%) drugs were correctly classified in
comparison to BCS and BDDCS, respectively. The provisional
classes I through IV showed prediction accuracy of 73%, 78%,
100%, and 67%, respectively, with respect to BCS classes
(Figure 5). However, the prediction accuracy of the provisional
classification in comparison to BDDCS was 83%, 83%, 88%,
and 67%, for classes I through IV. While there are no drugs of
BCS/BDDCS class I falling into class IV of provisional
classification, and vice versa, two distinct outliers in the
provisional classification are rosiglitazone and amitriptyline
(BCS/BDDCS class I), which showed low solubility and low
Figure 2. The ROC analysis of MDCK-LE apparent permeability vs
human Fa. Human Fa ≥90% or ≥80% (horizontal dotted lines) were
permeability, respectively (Figure 6). The BCS and BDDCS
set as true positives. Inset shows the ROC curves. The apparent class II drugs atorvastatin and ketoconazole were misclassified
permeability cutoff was found to be 5 × 10−6 cm/s, based on the as provisional class IV. Digoxin, which showed low solubility
shortest distance to the point of (0,1) on the y-axis of the ROC curve. and permeability in the current study, is a BCS II and BDDCS
MDCK-LE apparent permeability values represent transport at pH 6.5 III drug. On the other hand, the BCS class I drug prednisone
for acids and pH 7.4 for bases, neutrals, and zwitterions. Key: Data showed low solubility and high permeability, while the class II
points are categorized based on BCS classes I−IV. drug talinolol showed high solubility and low permeability in
our provisional classification. It is interesting to note that
with literature BCS and the current permeability classes (80% prednisone and talinolol are categorized as BDDCS classes II
sensitivity and 89% specificity) was observed for 105 drugs, and III, respectively. Nevertheless, the provisional classification
using high permeability cutoff of ≥5 × 10−6 cm/s (Figure 3A). based on the proposed criteria show good agreement with both
Although other cutoff values were evaluated for permeability the BCS and BDDCS.
classification, no improvement over the concordance achieved MDCK-LE Permeability vs Percent Dose Excreted in
with 5 × 10−6 cm/s was observed. Furthermore, this Urine or Bile. Figure 7 shows the relationship between
permeability criterion was able to predict metabolism classes, permeability and extent of urinary excretion in human, or with
as defined by BDDCS, with 85% sensitivity and 86% specificity. the percentage of parent excreted in bile in human and rat.
Evidently, BCS and BDDCS class I and II compounds showed Clearly, urinary excretion showed an inverse relationship with
significantly (p < 0.001) higher permeability over the permeability, where drugs with extensive urinary elimination
compounds of classes III and IV. The drugs in the BCS classes are low permeable (MDCK-LE permeability <5 × 10−6 cm/s).
I through IV have an average apparent permeability of 21.2 × There was no sufficient human biliary excretion data to draw a
10−6 cm/s (n = 41), 22.2 × 10−6 cm/s (n = 36), 2.7 × 10−6 cm/ conclusion; nevertheless, the relationship between permeability
s (n = 24), and 7.1 × 10−6 cm/s (n = 4); while BDDCS classes and rat biliary excretion suggests that high biliary excretion is
showed an average permeability of 23.5 × 10−6 cm/s (n = 34), less likely when passive permeability is higher. In general,
23.0 × 10−6 cm/s (n = 32), 3.0 × 10−6 cm/s (n = 29), and 4.1 BDDCS and provisional class III and IV drugs are over-
× 10−6 cm/s (n = 6). represented in drugs primarily excreted in urine and bile.
pH-Dependent Solubility Criterion. Equilibrium solubil- Overall, this data also substantiates the applicability of
ity of a set of 49 drugs in pH 1.2, pH 6.5 PBS, and FaSSIF was permeability cutoff and the provisional classification in
predicting clearance pathways, as suggested with BDDCS.21,26

■
obtained using in-house HT solubility assay (Table 1). To
accurately classify acidic drugs, we adopted an approach similar
to the BCS regulatory criteria of lowest solubility in pH 1−7.5;2 DISCUSSION
and the solubility measured at pH 1.2 was used. In contrast, an Based on extensive studies, it was determined that a good
analysis of the pH 6.5 PBS and FaSSIF solubility resulted in correlation existed between the human Peff measured using
misclassification of a large number of acids, presumably due to intestinal perfusion and the fraction of oral dose absorbed in
higher solubility. However, for basic, neutral, and zwitterionic humans, suggesting human Peff as “reference-standard” for
drugs the solubility measured in pH 6.5 PBS and FaSSIF media permeability assessment.1,2,4 While human Peff data is not easily
showed good concordance with the literature solubility accessible, reliable permeability data useful for permeability
classification. Interestingly, no major differences were observed classification and oral absorption prediction purposes can be
in equilibrium solubility between pH 6.5 buffer and FaSSIF obtained by validated permeability assays. Here we used an in-
medium for the bases and neutrals in this data set (Table 1). house HT permeability model, based on MDCK-LE cell
Good agreement with literature BCS high and low solubility monolayers,10 and defined a pH-dependent permeability
classification (93% sensitivity and 86% specificity) was criterion that provided a good relationship with human Fa as
observed, using the current pH-dependent criterion and an well as proposed cutoff value for provisional classification.
equilibrium solubility cutoff of 200 μg/mL (Figure 4). Similar A good sigmoidal relationship, as described by the complete
statistical analysis showing a good agreement between BDDCS radial mixing (parallel tube) model,22,23,32 was observed
and current solubility classification was noted, although 8 of the between apparent permeability and human Fa, when the
1206 dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212
Molecular Pharmaceutics Article

Figure 3. MDCK-LE apparent permeability vs BCS classes (A) and BDDCS classes (B). Inset shows the truth tables indicating the permeability class
predictability of the MDCK-LE permeability cutoff, 5 × 10−6 cm/s. MDCK-LE apparent permeability values represent transport at pH 6.5 for acids
and pH 7.4 for bases, neutrals, and zwitterions. Data points are slightly jittered for clarity.

permeability values at apical pH 6.5 and pH 7.4 were used for higher than the apparent permeability determined across
acidic and nonacidic drugs, respectively (Figure 1). The MDCK-LE cell monolayers (Table 1 and Figure 1). Major
outcome of this assessment is consistent with other reports reason for the fold differences between the models could be
published earlier. For example, Artursson and Karlsson showed due to the larger effective absorptive area of the human
a good relationship between the transport across Caco-2 intestinal perfusion segment provided by the villi and microvilli,
monolayers and human Fa.32 Irvine et al. using a set of 55 although differences in the diffusion coefficient, membrane
compounds showed an approximately sigmoidal correlation thickness, membrane bilayer composition, luminal content, or
between permeability and human Fa and also demonstrated a extracellular mucus layer affecting drug partitioning into the
linear correlation of MDCK and Caco-2 permeability.11 Also, membrane may also contribute.22,32
the current data set of over 100 drugs further substantiates our On the basis of ROC analysis of the relationship between
previous report that MDCK-LE can be a useful tool for human MDCK-LE permeability and human Fa, a cutoff for highly
Fa predictions and permeability classification.10 On average, the permeable drugs, Papp ≥ 5 × 10−6 cm/s, ensuring >90% oral
human effective permeability estimates were about 24 times absorption was identified. Of the data available, about 78% of
1207 dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212
Molecular Pharmaceutics Article

Figure 4. Solubility vs BCS classes (A) and BDDCS classes (B). Inset shows the truth tables indicating the solubility class predictability of the
solubility cutoff, 200 μg/mL. Solubility values represent solubility in pH 1.2 for acids and in FaSSIF for bases, neutrals, and zwitterions. Data points
are slightly jittered for clarity.

drugs with apparent permeability <5 × 10−6 cm/s are provide guidance for project teams on the absorption
incompletely absorbed (<90% human Fa). On the other potential.10,36
hand, only a couple of drugs (e.g., lovastatin and tacrolimus) It is generally believed that the proximal intestinal segment
with high apparent permeability are incompletely absorbed (duodenum and upper jejunum) is the major site of absorption
(<70% human Fa), which may be attributed to a rate-limiting for immediate-release formulations.4,27 The average luminal pH
solubility process or involvement of efflux transporters in their in this segment is about 6.5, suggesting that permeability
oral absorption.33,34 It may be expected that the cutoff value in measured at pH 6.5 is physiologically relevant for estimating
particular for permeability may differ from one laboratory to oral absorption.37 However, due to lower extent of ionization,
another. Similar to the Caco-2 cell line, MDCK-LE is a basic drugs usually possess relatively higher permeability at pH
7.4. Since the average luminal pH of the lower half of the small
heterogeneous cell population, which would be exposed to
intestine is about 7.4, the higher permeability in the distal
different selection pressures in different laboratories.35 There- segment may compensate for the low permeability of basic
fore, it would be prudent to define the permeability cutoff value drugs in the proximal intestine. Dahan et al. studied pH-
in the respective laboratories conducting similar profiling. dependent permeability of a basic drug, sotalol, to investigate
Furthermore, one may be cautious in adopting a single-value the disparity between its apparent low permeability and high
cutoff to differentiate high and low permeability, as the human Fa.38 While the permeability of sotalol is much higher at
compounds around the boundary may be misclassified due to pH 7.5 compared to pH 6.5, the report suggested that a
inherent inter- and intraday variability with cell-based models. compound with high permeability at some point along the
A conservative range of <2.5 × 10−6 (low), 2.5−10 × 10−6 relevant intestinal regions, not necessarily in the jejunum, may
(medium), and 10 × 10−6 cm/s (high) was adopted at Pfizer to show high human Fa. Similarly, in the current data set, several
1208 dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212
Molecular Pharmaceutics Article

Figure 5. Provisional biopharmaceutics classification of a set of 49

drugs using pH-dependent solubility and permeability. (A) Compar-
ison to literature BCS classes. (B) Comparison to literature BDDCS
classes. The solubility criterion was based on solubility in pH 1.2 for
acids and in FaSSIF for nonacids, and was set as <200 μg/mL for low
solubility and ≥200 μg/mL for high solubility. The permeability
criterion was based on permeability across MDCK-LE monolayers at
pH 6.5 or 7.4, and the cutoff was set at 5 × 10−6 cm/s. Accuracy is
defined as the percentage ratio of number of compounds rightly
classified using current criteria to the number of compounds in that Figure 7. Relationship of MDCK-LE permeability and (A) percentage
particular class, according to literature BCS or BDDCS. Data points of parent excreted in urine in human, (B) percentage of parent
are slightly jittered for clarity. Key: Data points are categorized based excreted in bile in human, and (C) percentage of parent excreted in
on BCS or BDDCS classes I−IV. bile in rat. Key: Data points are categorized based on BDDCS classes
I−IV as in Figure 5B.
basic drugs including atropine, chlorpromazine, cisapride, and
desipramine, which are completely absorbed in human (human

Figure 6. Concordance of provisional biopharmaceutics classification with (A) BCS and (B) BDDCS. Of 49 compounds, provisional classification
correctly predicted BCS and BDDCS class of 40 and 41 compounds, respectively.

1209 dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212

Molecular Pharmaceutics Article

Fa >90%), showed low permeability (<5 × 10−6 cm/s) at pH excretion are also identified as class III and class IV of
6.5 (Table 1). The permeability of these compounds is much BDDCS.19,20,26,29 While this paradigm shift brings with it the
above the defined cutoff (5 × 10−6 cm/s) at pH 7.4, indicating reduction of in vivo toxicity findings and better potency toward
that their major site of absorption is the lower small intestine. certain biological targets, it also presents a more difficult
Although a majority of the neutral compounds showed similar chemical space for oral absorption. In addition, these classes of
permeability at both pH conditions, few compounds tended compounds would be associated with higher liability in terms of
toward higher values at higher pH; prompting the use of transporter-mediated DDI and potential impact of involved
permeability values at pH 7.4 for all nonacidic compounds. transporter(s) pharmacogenomics on the overall exposure,
Overall, the pH-dependent criterion proposed in the current efficacy, and toxicity.40 Therefore, having a discovery strategy
analysis, where permeability at pH 6.5 and pH 7.4 was that would enable projection of the route of elimination and
considered for acids and nonacids, respectively, is in line with therefore proactive assessment and management of relevant
the physiologically relevant pH at the intestinal region that is potential liabilities is pivotal to ensure the successful develop-
likely to handle majority of absorption for a given drug.38 ment of these new therapeutic agents. The current study
The current in-house HT solubility assay balances sample showed good agreement between the provisional classification
consumption, throughput, analytical detection, and data quality and BDDCS, suggesting the applicability of HT permeability
and was developed and validated with a focus on low soluble and solubility screens in defining BDDCS class, which
compounds (ca. 10 μg/mL or less) typically encountered in otherwise is derived from extent of metabolism and solubility.
early drug discovery space. Therefore, the solubility data Therefore, it is expected that the provisional classification will
obtained here inherited a dynamic range of 0.3−300 μg/mL in predict the predominant clearance mechanism and provide
all media. According to the regulatory guidelines, the boundary guidance on the transporter-mediated DDI liability.
between high and low solubility is defined by the dose number We noted a distinct inverse relationship between MDCK-LE
(Do, maximum dose/solubility × 250 mL) over the pH range permeability and the percentage of parent excreted in the urine
of 1−7.5.2 A high solubility drug should show Do < 1 in this (Figure 7A), as well as with human renal clearance (data not
pH range. However, to strictly classify the drugs according to shown). Renal clearance is the net result of three processes:
the BCS, thermodynamic solubility measurements over pH 1− glomerular filtration, active secretion, and reabsorption.
7.5 are required. These requirements are not practical at the However, compounds with sufficient passive permeability can
early drug discovery stage due to the lack of availability of the be efficiently reabsorbed as the passive reabsorption process
material, resources involved, and feasibility on high throughput occurs throughout the length of the nephron, against the
systems. Furthermore, any agreement around the salt form and
secretion predominantly occurring at the proximal tubule.30
maximum oral dose in the clinic is not attained at this stage.
With the exception of the two outliers (atropine and
Therefore, we propose criteria based on the pH-dependent
trimethoprim), all drugs with permeability higher than 5 ×
solubility values, rather than dose number, which potentially
10−6 cm/s showed less than 30% urinary excretion, suggesting
can assist in provisional biopharmaceutics classification at the
that the drugs with permeability above this value are not
early discovery stage. Based on this data set, equilibrium
predominantly cleared by kidney, and also are unlikely to be
solubility in pH 1.2 (SGF) medium provides a basis for defining
solubility boundary for acidic drugs, while, for nonacidic drugs, victims of renal DDI, irrespective of their interaction with renal
equilibrium solubility in the FaSSIF medium is a required transporters. Apparently, almost all drugs that show clinical
measure. On the basis of this pH-dependent solubility, an renal DDI are hydrophilic and often ionized at physiological pH
arbitrary solubility boundary drawn at 200 μg/mL was able to with a cLog D <1.0,41 although several lipophilic drugs are also
differentiate solubility classes of 49 drugs tested with a known to be transported by secretory transporters like OAT1,
sensitivity of 93% and a specificity of 86%. Observed good OAT3, and OCT2.42,43 Correspondingly, we note that several
agreement with the solubility classification, in spite of not drugs like cimetidine, famotidine, furosemide, ranitidine,
considering dose in the equation, may offer confidence in methotrexate, etc., which showed significant renal DDI in
suggesting that NMEs with solubility >200 μg/mL throughout clinic,44 are less permeable across MDCK-LE cells and fall in
the GI pH are less likely to show solubility-limited absorption. class III or IV of both the BDDCS and provisional classification
Admittedly, further investigation into this cutoff value is systems.
warranted, especially due to the sparse data points in the Lack of sufficient human biliary excretion data limited
range of 100−250 μg/mL, and also to put dose strength in drawing any trends; nevertheless, percentage biliary excretion in
context. We believe that, while the current criterion is helpful at rat showed inverse relationship with permeability (Figure 7C).
the early discovery stage, detailed solubility assessment and It is also evident that the class III and IV drugs may be excreted
solid-state characterization are required due to likely change in in bile, while class I and II drugs are less likely to show biliary
solid or salt form as the compound progress into development. excretion. Furthermore, with a set of over 200 discovery
Over the last two decades, the pharmaceutical industry made compounds, we observed that the substrates of hepatic
major strides in optimizing the physicochemical space to sinusoidal uptake transporters (hOATP1B1, hOATP1B3, and
minimize the extent of metabolism contribution to the overall hOATP2B1) that play a primary role in hepatobiliary transport
clearance and the CYP-mediated DDI of NMEs. Evidently, are markedly overrepresented in the low permeability class
reducing NMEs' lipophilicity and increasing polarity would lead (MDCK-LE permeability <5 × 10−6 cm/s) compared to the
to a reduced extent of metabolism and potential for CYP- high permeability class.45 While transporters are the significant
mediated drug−drug interactions (DDI). This strategy resulted drivers in the biliary and renal elimination, the provisional
in a greater number of compounds presenting with low passive classification scheme combined with the in vitro tools for
permeability characteristics, the classes of compounds identified transporter characterization (example: transfected cell lines,
as class III and class IV according to the BCS.39,40 Indeed, these membrane vesicles) should be useful in predicting the clearance
compounds with predominant clearance via renal and/or biliary mechanism and the clinical relevance of transporter-based DDI.
1210 dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212
Molecular Pharmaceutics Article

In conclusion, we proposed pH-dependent solubility and cells: A tool for membrane permeability screening. J. Pharm. Sci. 1999,
permeability criteria for provisional biopharmaceutics classi- 88 (1), 28−33.
fication using the high throughput data available early in the (12) Avdeef, A. The rise of PAMPA. Expert Opin. Drug Metab.
discovery stage. The simplicity and efficient nature of the tools Toxicol. 2005, 1 (2), 325−42.
not only enable discovery teams to predict oral absorption of (13) Avdeef, A.; Artursson, P.; Neuhoff, S.; Lazorova, L.; Grasjo, J.;
Tavelin, S. Caco-2 permeability of weakly basic drugs predicted with
NMEs but also aid in understanding drug disposition and the
the double-sink PAMPA pKa(flux) method. Eur. J. Pharm. Sci. 2005,
relevance of enzymes and/or transporters in absorption and 24 (4), 333−49.
elimination, early on. Furthermore, such a provisional (14) Alsenz, J.; Kansy, M. High throughput solubility measurement in
classification system may provide valuable insight into the drug discovery and development. Adv. Drug Delivery Rev. 2007, 59 (7),
clinical relevance of transporter-mediated drug−drug inter- 546−67.
actions.

■
(15) Alsenz, J.; Meister, E.; Haenel, E. Development of a partially
automated solubility screening (PASS) assay for early drug develop-
AUTHOR INFORMATION ment. J. Pharm. Sci. 2007, 96 (7), 1748−62.
Corresponding Author (16) Bevan, C. D.; Lloyd, R. S. A high-throughput screening method
*Pfizer Inc., Pfizer Global Research and Development, Groton, for the determination of aqueous drug solubility using laser
nephelometry in microtiter plates. Anal. Chem. 2000, 72 (8), 1781−7.
CT 06340. Phone: +1-860-686-9311. Fax: +1-860-441-6402. E-
(17) Roy, D.; Ducher, F.; Laumain, A.; Legendre, J. Y. Determination
mail: Ayman.El-Kattan@pfizer.com. of the aqueous solubility of drugs using a convenient 96-well plate-
Notes based assay. Drug Dev. Ind. Pharm. 2001, 27 (1), 107−9.
The authors declare no competing financial interest. (18) Stegemann, S.; Leveiller, F.; Franchi, D.; de Jong, H.; Linden, H.

■ ACKNOWLEDGMENTS
All authors were employees of Pfizer, Inc., when the study was
When poor solubility becomes an issue: from early stage to proof of
concept. Eur. J. Pharm. Sci. 2007, 31 (5), 249−61.
(19) Benet, L. Z.; Amidon, G. L.; Barends, D. M.; Lennernas, H.;
Polli, J. E.; Shah, V. P.; Stavchansky, S. A.; Yu, L. X. The use of
conducted.

■
BDDCS in classifying the permeability of marketed drugs. Pharm. Res.
2008, 25 (3), 483−8.
REFERENCES (20) Chen, M. L.; Amidon, G. L.; Benet, L. Z.; Lennernas, H.; Yu, L.
(1) Amidon, G. L.; Lennernas, H.; Shah, V. P.; Crison, J. R. A X. The BCS, BDDCS, and regulatory guidances. Pharm. Res. 2011, 28
theoretical basis for a biopharmaceutic drug classification: the (7), 1774−8.
correlation of in vitro drug product dissolution and in vivo (21) Wu, C. Y.; Benet, L. Z. Predicting drug disposition via
bioavailability. Pharm. Res. 1995, 12 (3), 413−20. application of BCS: transport/absorption/ elimination interplay and
(2) CDER/FDA. Guidance for industry: Waiver of in vivo development of a biopharmaceutics drug disposition classification
bioavailability and bioequivalence studies for imidiate-release solid system. Pharm. Res. 2005, 22 (1), 11−23.
oral dosage forms based on a biopharmaceutics classification system. (22) Fagerholm, U.; Johansson, M.; Lennernas, H. Comparison
Center for Drug Evaluation and Research, 2000. between permeability coefficients in rat and human jejunum. Pharm.
(3) Yu, L. X.; Amidon, G. L.; Polli, J. E.; Zhao, H.; Mehta, M. U.; Res. 1996, 13 (9), 1336−42.
Conner, D. P.; Shah, V. P.; Lesko, L. J.; Chen, M. L.; Lee, V. H.; (23) Varma, M. V.; Panchagnula, R. Prediction of in vivo intestinal
Hussain, A. S. Biopharmaceutics classification system: the scientific absorption enhancement on P-glycoprotein inhibition, from rat in situ
basis for biowaiver extensions. Pharm. Res. 2002, 19 (7), 921−5. permeability. J. Pharm. Sci. 2005, 94 (8), 1694−704.
(4) Lennernas, H. Human intestinal permeability. J. Pharm. Sci. 1998, (24) Bewick, V.; Cheek, L.; Ball, J. Statistics review 13: receiver
87 (4), 403−10. operating characteristic curves. Crit. Care 2004, 8 (6), 508−12.
(5) Sandstrom, R.; Karlsson, A.; Knutson, L.; Lennernas, H. Jejunal (25) Strik, J. J.; Honig, A.; Lousberg, R.; Denollet, J. Sensitivity and
absorption and metabolism of R/S-verapamil in humans. Pharm. Res. specificity of observer and self-report questionnaires in major and
1998, 15 (6), 856−62. minor depression following myocardial infarction. Psychosomatics
(6) Varma, M. V.; Khandavilli, S.; Ashokraj, Y.; Jain, A.; Dhanikula,
2001, 42 (5), 423−8.
A.; Sood, A.; Thomas, N. S.; Pillai, O.; Sharma, P.; Gandhi, R.;
(26) Custodio, J. M.; Wu, C. Y.; Benet, L. Z. Predicting drug
Agrawal, S.; Nair, V.; Panchagnula, R. Biopharmaceutic classification
disposition, absorption/elimination/transporter interplay and the role
system: a scientific framework for pharmacokinetic optimization in
of food on drug absorption. Adv. Drug Delivery Rev. 2008, 60 (6),
drug research. Curr. Drug Metab. 2004, 5 (5), 375−88.
(7) Obata, K.; Sugano, K.; Machida, M.; Aso, Y. Biopharmaceutics 717−33.
classification by high throughput solubility assay and PAMPA. Drug (27) Lennernas, H. Intestinal permeability and its relevance for
Dev. Ind. Pharm. 2004, 30 (2), 181−5. absorption and elimination. Xenobiotica 2007, 37 (10−11), 1015−51.
(8) Artursson, P. Epithelial transport of drugs in cell culture. I: A (28) Varma, M. V.; Obach, R. S.; Rotter, C.; Miller, H. R.; Chang, G.;
model for studying the passive diffusion of drugs over intestinal Steyn, S. J.; El-Kattan, A.; Troutman, M. D. Physicochemical space for
absorptive (Caco-2) cells. J. Pharm. Sci. 1990, 79 (6), 476−82. optimum oral bioavailability: contribution of human intestinal
(9) Artursson, P.; Magnusson, C. Epithelial transport of drugs in cell absorption and first-pass elimination. J. Med. Chem. 2010, 53 (3),
culture. II: Effect of extracellular calcium concentration on the 1098−108.
paracellular transport of drugs of different lipophilicities across (29) Benet, L. Z.; Broccatelli, F.; Oprea, T. I. BDDCS Applied to
monolayers of intestinal epithelial (Caco-2) cells. J. Pharm. Sci. Over 900 Drugs. AAPS J. 2011, 13 (4), 519−47.
1990, 79 (7), 595−600. (30) Varma, M. V.; Feng, B.; Obach, R. S.; Troutman, M. D.;
(10) Di, L.; Whitney-Pickett, C.; Umland, J. P.; Zhang, H.; Zhang, X.; Chupka, J.; Miller, H. R.; El-Kattan, A. Physicochemical determinants
Gebhard, D. F.; Lai, Y.; Federico, J. J., 3rd; Davidson, R. E.; Smith, R.; of human renal clearance. J. Med. Chem. 2009, 52 (15), 4844−52.
Reyner, E. L.; Lee, C.; Feng, B.; Rotter, C.; Varma, M. V.; Kempshall, (31) Yang, X.; Gandhi, Y. A.; Duignan, D. B.; Morris, M. E.
S.; Fenner, K.; El-Kattan, A. F.; Liston, T. E.; Troutman, M. D. Prediction of biliary excretion in rats and humans using molecular
Development of a new permeability assay using low-efflux MDCKII weight and quantitative structure-pharmacokinetic relationships. AAPS
cells. J. Pharm. Sci. 2011, 100 (11), 4974−85. J. 2009, 11 (3), 511−25.
(11) Irvine, J. D.; Takahashi, L.; Lockhart, K.; Cheong, J.; Tolan, J. (32) Artursson, P.; Karlsson, J. Correlation between oral drug
W.; Selick, H. E.; Grove, J. R. MDCK (Madin-Darby canine kidney) absorption in humans and apparent drug permeability coefficients in

1211 dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212

Molecular Pharmaceutics Article

human intestinal epithelial (Caco-2) cells. Biochem. Biophys. Res.

Commun. 1991, 175 (3), 880−5.
(33) Kino, T.; Hatanaka, H.; Hashimoto, M.; Nishiyama, M.; Goto,
T.; Okuhara, M.; Kohsaka, M.; Aoki, H.; Imanaka, H. FK-506, a novel
immunosuppressant isolated from a Streptomyces. I. Fermentation,
isolation, and physico-chemical and biological characteristics. J.
Antibiot. (Tokyo) 1987, 40 (9), 1249−55.
(34) Patel, M.; Tekade, A.; Gattani, S.; Surana, S. Solubility
enhancement of lovastatin by modified locust bean gum using solid
dispersion techniques. AAPS PharmSciTech 2008, 9 (4), 1262−9.
(35) Artursson, P.; Palm, K.; Luthman, K. Caco-2 monolayers in
experimental and theoretical predictions of drug transport. Adv. Drug
Delivery Rev. 2001, 46 (1−3), 27−43.
(36) Gupta, R. R.; Gifford, E. M.; Liston, T.; Waller, C. L.; Hohman,
M.; Bunin, B. A.; Ekins, S. Using open source computational tools for
predicting human metabolic stability and additional absorption,
distribution, metabolism, excretion, and toxicity properties. Drug
Metab. Dispos. 2010, 38 (11), 2083−90.
(37) Avdeef, A. Physicochemical profiling (solubility, permeability
and charge state). Curr. Top. Med. Chem. 2001, 1 (4), 277−351.
(38) Dahan, A.; Miller, J. M.; Hilfinger, J. M.; Yamashita, S.; Yu, L. X.;
Lennernas, H.; Amidon, G. L. High-Permeability Criterion for BCS
Classification: Segmental/pH Dependent Permeability Considerations.
Mol. Pharmaceutics 2010, 7 (5), 1827−1834.
(39) Price, D. A.; Blagg, J.; Jones, L.; Greene, N.; Wager, T.
Physicochemical drug properties associated with in vivo toxicological
outcomes: a review. Expert Opin. Drug Metab. Toxicol. 2009, 5 (8),
921−31.
(40) Varma, M. V.; Ambler, C. M.; Ullah, M.; Rotter, C. J.; Sun, H.;
Litchfield, J.; Fenner, K. S.; El-Kattan, A. F. Targeting intestinal
transporters for optimizing oral drug absorption. Curr. Drug Metab.
2010, 11 (9), 730−42.
(41) Feng, B.; LaPerle, J. L.; Chang, G.; Varma, M. V. Renal clearance
in drug discovery and development: molecular descriptors, drug
transporters and disease state. Expert Opin. Drug Metab. Toxicol. 2010,
6 (8), 939−52.
(42) Dantzler, W. H.; Wright, S. H. The molecular and cellular
physiology of basolateral organic anion transport in mammalian renal
tubules. Biochim. Biophys. Acta 2003, 1618 (2), 185−93.
(43) Kido, Y.; Matsson, P.; Giacomini, K. M. Profiling of a
prescription drug library for potential renal drug-drug interactions
mediated by the organic cation transporter 2. J. Med. Chem. 2011, 54
(13), 4548−58.
(44) Masereeuw, R.; Russel, F. G. Mechanisms and clinical
implications of renal drug excretion. Drug Metab. Rev. 2001, 33 (3−
4), 299−351.
(45) Varma, M. V.; Chang, G.; Lai, Y.; Feng, B.; El-Kattan, A. F.;
Litchfield, J.; Goosen, T. C. Physicochemical property space of hepato-
biliary transport and computational models for predicting rat biliary
excretion, unpublished.

1212 dx.doi.org/10.1021/mp2004912 | Mol. Pharmaceutics 2012, 9, 1199−1212