dmd 36:102–123, 2008 printed in u.s.a. permeability...

Permeability, Transport, and Metabolism of Solutes in Caco-2Cell Monolayers: A Theoretical Study

Huadong Sun and K. Sandy Pang

Department of Pharmaceutical Sciences, Leslie Dan Faculty of Pharmacy, University of Toronto, Toronto, Canada

Received February 23, 2007; accepted October 9, 2007

ABSTRACT:

We explored the properties of a catenary model that includes thebasolateral (B), apical (A), and cellular compartments via simula-tions under linear and nonlinear conditions to understand theasymmetric observations arising from transporters, enzymes, andpermeability in Caco-2 cells. The efflux ratio (EfR; Papp,B3A/Papp,A3B), obtained from the effective permeability from the A3Band B3A direction under linear conditions, was unity for passivelypermeable drugs whose transport does not involve transporters;the value was unaffected by cellular binding or metabolism, butincreased with apical efflux. Metabolism was asymmetric, showinglesser metabolite accrual for the B3A than A3B direction becauseof inherent differences in the volumes for A and B. Moreover, thenet flux (total � passive permeation) due to saturable apical efflux,absorption, or metabolism showed nonconformity to simple

Michaelis-Menten kinetics against CD,0, the loading donor concen-tration. EfR values differed with saturable apical efflux and metab-olism (>1), as well as apical absorption (EfRs <1), but approachedunity with high passive diffusive clearance (CLd) and increasingCD,0 at a higher degree of saturation of the process. The Jmax

(apparent Vmax estimated for the carrier system) and K�m [or the K�mbased on a modified equation with the Hill coefficient (�)] esti-mates from the Eadie-Hofstee plot revealed spurious correlationswith the assigned Vmax and Km. The sampling time, CLd, and pa-rameter space of Km and Vmax strongly influenced both the corre-lation and accuracy of estimates. Improved correlation was foundfor compounds with high CLd. These observations showed that thecatenary model is appropriate in the description of transport andmetabolic data in Caco-2 cells.

The majority of drugs available on the market are in oral dosageforms. For the assessment of permeability and oral drug absorption, insilico models (Stenberg et al., 2001) and high throughput systems,such as the parallel artificial membrane permeability assay (Kansy etal., 1998) and cell-based systems (Hidalgo et al., 1989) exist to relatedrug permeability to absorption, especially for compounds that do notundergo intestinal metabolism (Usansky and Sinko, 2005). The mostpopular high-throughput screening tool for drug permeability is hu-man colon carcinoma (Caco-2) (Hidalgo et al., 1989) or transfectedMadin-Darby canine kidney (Irvine et al., 1999) cells. Upon culture,Caco-2 cells differentiate and become confluent to form monolayers

with tight junctions and polarized apical/mucosal (A side) and baso-lateral/serosal (B side) membranes that are structurally and function-ally similar to those of enterocytes.

ABC efflux transporters such as P-glycoprotein (Pgp) multidrug resis-tance-associated protein 2 (MRP2), and the breast cancer-resistant protein(BCRP) are expressed on the mucosal membrane of Caco-2 (Hunter etal., 1993; Hirohashi et al., 2000), as are the absorptive transporters, suchas the proton-coupled oligopeptide transporter (PEPT1) (Guo et al., 1999)and the organic anion transporting polypeptide (OATP) (Kobayashi et al.,2003). Likewise, on the serosal membrane, basolateral efflux transporterssuch as MRP3 (Hirohashi et al., 2000) and organic solute transporters �

and � (OST�-OST�) (Okuwaki et al., 2007) are expressed in Caco-2. Inaddition, multiple metabolic enzymes such as the sulfotransferases, UDP-glucuronosyltransferases (H. Sun, L. Zhang, E. C. Chow, G. Lin, K. S.Pang, Z. Zuo, unpublished data), and the glutathione S-transferases (Pe-ters and Roelofs, 1989) reside in the Caco-2 cell monolayer. Besides,

This work was supported by the Canadian Institute for Health Research, GrantMOP64350.

Article, publication date, and citation information can be found athttp://dmd.aspetjournals.org.

doi:10.1124/dmd.107.015321.

ABBREVIATIONS: ABC, ATP-binding cassette; Pgp, P-glycoprotein; MRP, multidrug resistance-associated protein; BCRP, breast cancer-resistant protein; GF120918, N-(4-[2-(1,2,3,4-tetrahydro-6,7-dimethoxy-2-isoquinolinyl)ethyl]-phenyl)-9,10-dihydro-5-methoxy-9-oxo-4-acridinecarboxamide; A, apical; B, basolateral; CD,0, initial loading concentration in donor side at t � 0; Papp, effective permeability; EfR, efflux ratio;MK571, 3-[[3-[2-(7-chloroquinolin-2-yl)vinyl]phenyl]-(2-dimethylcarbamoylethylsulfanyl)methylsulfanyl] propionic acid; PPD, passive permeability;DR and Dcell, drug amounts in receiver and cell, respectively; AQ, absorptive quotient; SQ, secretory quotient; Mettotal, total amount of metabolite;ER, extraction ratio; CLint,sec, intrinsic clearance for apical efflux, mediated by transporters; CLint,met, metabolic intrinsic clearance; CLabs, intrinsicclearance for apical absorption, mediated by transporters; CLinflux and CLefflux, intrinsic clearances for basolateral influx and efflux, respectively,mediated by transporters; CLd, passive diffusion clearance common across apical and basolateral membranes; CLd4 and CLd1, passive diffusionclearance from the apical and basolateral compartments into the cell compartment, respectively; CLd3 and CLd2, passive diffusion clearance fromthe cell compartment into the apical and basolateral compartments, respectively; Vap and Vbaso, volumes of buffer solution in the apical andbasolateral compartments, respectively; Vcell, cellular volume; Pc, Jc, and Jmax, the permeability, flux, and apparent Vmax estimated for the carriersystem, respectively; K�m and K�m, the apparent Kms estimated from the Eadie-Hofstee plot, with and without the Hill coefficient, respectively; fmet,fraction of dose metabolized.

0090-9556/08/3601-102–123$20.00DRUG METABOLISM AND DISPOSITION Vol. 36, No. 1Copyright © 2008 by The American Society for Pharmacology and Experimental Therapeutics 15321/3287075DMD 36:102–123, 2008 Printed in U.S.A.

102

at ASPE

T Journals on A

ugust 31, 2018dm

d.aspetjournals.orgD

ownloaded from

http://dmd.aspetjournals.org/

CYP3A4, which is inherently poorly expressed in the cell monolayer,may be induced by incubation with 1�,25-dihydroxyvitamin D3 or ac-quired by transfection of the CYP3A4 gene (Cummins et al., 2001). TheCaco-2 system allows for paracellular transport (Adson et al., 1994), andCa2�-free buffer has often been used to disrupt paracellular transport offluorescein, a highly fluorescent probe, for assessment of the contributionof the pathway (Troutman and Thakker, 2003a). Furthermore, an un-stirred water layer that can impede the transport of highly lipophilic drugshas been described for the Caco-2 cells (Hidalgo et al., 1991).

Cell-based permeability models have become a necessary strategyin the drug discovery paradigm (Lentz et al., 2000; Polli et al., 2001).The protocol facilitates the drug-discovery scenario to define whethernew drug entities are Pgp substrates. Inhibitors such as GF120918 areused to obliterate apical efflux and identify Pgp involvement. Inparallel, computational models have evolved to describe Caco-2 trans-port. However, most of the interpretations express simplified viewsand provide approximate solutions for the passive permeability coef-ficient, Papp, by assuming that the system is a single barrier. Even fornonlinear cases, influx or efflux at the apical membrane was simpli-fied by viewing the system as a single barrier and assuming that drugmolecules present at either the apical or basolateral compartment havedirect access to intracellular enzymes or efflux transporters (Guo etal., 1999; Williams et al., 2002; Troutman and Thakker, 2003b; Tranet al., 2004).

We are of the view that the single-barrier model for the Caco-2 cellmonolayer is inadequate and that a catenary model comprising baso-lateral (B), apical (A), and cellular compartments (Ito et al., 1999;Tam et al., 2003; Irie et al., 2004; Gonzalez-Alvarez et al., 2005) ismore appropriate than what routine analyses subscribe, especiallyunder nonlinear conditions. Given the importance of the Caco-2system in drug discovery and development, we explored properties ofthe catenary model to gain a mechanistically based understanding ofthe asymmetric observations arising from transporters, enzymes, andeffective permeability under both linear and nonlinear conditions. Weexamined the definition of Papp under “sink” and “nonsink” condi-tions and when the parameter was based on cumulative or timedsampling. The definition of EfR and its dependence on transport,binding, and metabolic parameters was studied. Moreover, the appro-priateness of estimates of Michaelis-Menten parameters Vmax (asJmax) and Km (as K�m or K�m) was investigated.

Theory

The Caco-2 Cell Model. We hereby briefly revisit the terminol-ogies and concepts. Although we refer our terminology to the

Caco-2 system, the kinetics derived may also apply to transfectedcell lines or other similar, in vitro systems. The test compound isadministered into the donor side, the apical or basolateral com-partment, and samples are withdrawn at the alternate site (receiverside). Both sink and nonsink conditions for the Caco-2 system havebeen considered. Sink conditions refer to the situation in whichback diffusion of drug in the receiver side to cell monolayer isnegligible. Sink condition would be satisfied when the sampling isconducted within the time interval that drug concentration in thereceiver side remains �10% of the loading concentration, CD,0

(Troutman and Thakker, 2003a), or when drug in the serosal sideis removed rapidly and irreversibly, leaving no chance for the drugto return back to the enterocyte. In contrast, no assumption needsto be made for nonsink conditions; there is no need to assume thatdrug concentrations are maintained high in the donor side but lowin the receiver side and that drug molecules in the receiver side donot re-enter the cell compartment.

Definition of Papp and EfR. The most common approach is todocument drug appearance in the receiver side. This permeability

FIG. 1. Schematic presentation of the Caco-2 cell-based system by a catenary model comprising the basolateral (baso), cell, and apical (ap) compartments. V, f, and Cdenote volume, unbound fraction, and the concentration of drugs of each compartment, respectively; M denotes formed metabolite(s). CLd1 and CLd2, CLd4, and CLd3 denotethe influx and efflux and passive diffusion clearance on the basolateral and apical membrane, respectively. CLinflux and CLefflux represent the transporter-mediated influxand efflux intrinsic clearances on the basolateral membrane; CLabs and CLint,sec denote transporter-mediated intrinsic clearances of absorption and efflux on the apicalmembrane, respectively; CLint,met is the metabolic intrinsic clearance. Under nonlinear conditions, CLinflux, CLefflux, CLabs and CLint.sec, and CLint,met may be replaced bythe Vmax/(Km � Cu), where Cu is the unbound drug concentration.

FIG. 2. The effective permeabilities Papp according to eq. 1 for the A3B andB3A directions were identical when transporters and enzymes were absent: CLin-

flux � CLefflux � CLint,sec � CLabs � CLint,met � 0 and CLd1 � CLd2 � CLd3 �CLd4 � CLd. Note that, in the insets, Papp decreased with cellular binding (fcell

denotes the unbound fraction in the cell).

103PERMEABILITY, TRANSPORT, AND METABOLISM IN Caco-2

at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


term, Papp, defined in eq. 1, is commonly known as the apparent oreffective permeability (Hilgers et al., 1990),

Papp ��AR��t�S

CD,0(1)

where (�AR/�t) is the rate of drug appearance in the receiver side, Sis the surface area of the transwell (4.71 cm2 for transwell of 24-mm

insert diameter; Yamaguchi et al., 2000), and CD,0 is the initial drugconcentration in the donor side at time � 0. When the value is basedon cumulative sampling (total amount transported from time � 0 upto the last sampling point), the cumulative Papp is obtained. However,when the value is based on sampling at various time intervals, theincremental Papp is obtained, and CD,0 denotes the donor concentra-tion at the onset of the time interval (Youdim et al., 2003). Perme-

FIG. 3. The effective permeabilities Papp versus time for the A3B and B3A directions were different when CLd1 � CLd2 � CLd3 � CLd4 � CLd � 0.2 ml/min (A) or0.05 ml/min (B) and the corresponding EfRs at varying apical efflux activities (CLint,sec). All the unbound fractions were equal to unity; CLinflux � CLefflux � CLint,met �CLabs � 0.

FIG. 4. The effective permeabilities Papp versus time for the A3B and B3A directions were identical when CLd1 � CLd2 � CLd3 � CLd4 � CLd � 0.2 ml/min (A) or0.05 ml/min (B) at varying cellular metabolic activities (CLint,met). All the unbound fractions were equal to unity; CLinflux � CLefflux � CLint,sec � CLabs � 0.

104 SUN AND PANG

at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


ability, when based in absence of transporters, denotes the ability ofthe molecule to traverse membranes by passive means. However, highpassive permeability is not synonymous with high lipophilicity be-cause hydrogen bonding or the presence of unstirred water layer forthis group of compounds can delimit permeability to such an extentthat the compounds would exhibit a low permeability.

Usually, the purpose of transport studies in Caco-2 monolayer is toascertain whether the test compound is a substrate of the effluxtransporter, especially Pgp. The efflux ratio, or ratio of effectivepermeability for a drug given into B to appear in the A (B3A)direction to that in the A3B direction (eq. 2), is used.

EfR �Papp,B¡A

Papp,A¡B(2)

Generally speaking, a value of �1.2 or 1.5 for EfR infers that onlypassive diffusion is involved in drug transport, whereas EfR valuesthat greatly exceed unity suggest that test compounds are substrates ofefflux transporters at the apical membrane (Lentz et al., 2000; Polli etal., 2001). However, the inference is less certain for highly permeablesubstrates, even though the compounds are Pgp substrates; rapidinflux of drugs by passive diffusion or active uptake transporters onthe apical membrane can overcome the effect of apical efflux (Lentzet al., 2000).

For EfR values falling between approximately 1 and 1.5, thereexists the dilemma of identifying whether the test compound is asubstrate of Pgp or other efflux transporters. Methods involving use ofinhibitors are thus developed. The effective permeability (Papp) in thepresence of the specific inhibitors [GF120918 for Pgp and BCRP (denOuden et al., 1996), Ko143 for BCRP (Allen et al., 2002), and MK571for the MRPs (Gekeler et al., 1995)] would be reduced to approximatevalues of the passive permeability (PPD) when the transporter com-ponent is drastically reduced or totally obliterated. To avoid ambiguityin the definition of the concentration employed, the inhibitor shouldbe added to both A and B compartments.

The absorptive quotient (AQ) (eq. 3) and the secretory quotient(SQ) (eq. 4) are introduced with administration of the inhibitor intothe apical and basolateral sides (Troutman and Thakker, 2003c).These expressions are derived from Papp in absence and presence(PPD,A3B or PPD,B3A) of inhibition.

AQ �PPD,A¡B Papp,A¡B

PPD,A¡B(3)

SQ �Papp,B¡A PPD,B¡A

PPD,B¡A(4)

A change in Papp with inhibitors will alter the AQ and SQ values ifapical efflux transporters are involved. The enhancement of the asym-metric transport of drugs in A3B (absorptive) direction with atten-uation of transport in B3A (secretory) direction with Pgp inhibitorssuggests that Pgp is involved in drug transport. The unidirectionalmeasurements of AQ or SQ would reduce the workload in drugscreening while providing reliability to bidirectional study for identi-fication of Pgp substrates. It has been proposed that, if the AQ or SQvalue is more than 0.3, the drug is considered to be a Pgp substrate(Thiel-Demby et al., 2004).

Caco-2 Cells and Enzymes. Caco-2 also houses various drugmetabolizing enzymes, and the system has been used to study drugmetabolism as well as the interplay between transporters and enzymes(Cummins et al., 2001). Metabolism in Caco-2 monolayer couldmodulate Papp, EfR, and net efflux rates. The effects of metabolismhave not been properly addressed.

FIG. 5. Influence of CLint,sec, together with CLd (A), CLabs (B), and CLinflux/CLefflux

(C), on EfR under linear conditions, based on eq. 12. All passive clearances wereassumed to be equal (CLd1 � CLd2 � CLd3 � CLd4 � CLd), and unbound fractionsin basolateral and apical compartment were equal (fap � fbaso). For (A) and (B),CLinflux � CLefflux � 0 at the basolateral membrane; for (B) and (C), CLd wasassigned as 0.05 ml/min, respectively, and in (A) and (C), CLabs � 0.


at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


Various laboratories have used different terminologies to ex-press the cellular metabolism within the Caco-2 cell system. Oneof these is the extraction ratio (ER), given by the ratio of the totalamount of the metabolite (Mettotal) divided by the sum of theMettotal and the drug amount(s) in the receiving compartment (DR;eq. 5) (Fisher et al., 1999) or by the sum of Mettotal, DR, andthe drug amount(s) in the cell (Dcell) (eq. 6) (Cummins et al.,2001).

ER �Mettotal

Mettotal � DR(5)

ER �Mettotal

Mettotal � DR � Dcell(6)

These equations on the extraction ratio have been misconstrued todescribe the extents of metabolism because the equations have ne-

FIG. 6. Influence of Pgp-mediated efflux (CLint,sec, from 0 to 5 ml/min) on the asymmetry in metabolite formation at 1-h sampling expressed as a fraction of the dose ofmetabolite formed, fmet, from the B3A direction versus the A3B direction under linear conditions for solutes of (A) high (CLd1 � CLd2 � CLd3 � CLd4 � CLd � 0.2ml/min) or (B) low (CLd1 � CLd2 � CLd3 � CLd4 � CLd � 0.05 ml/min) passive permeability. In all simulations, CLint,met � 0.05 ml/min, CLabs � 0, CLinflux � CLefflux �0 at the basolateral membrane, and unbound fractions were equal to unity.

106 SUN AND PANG

at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


glected to consider the entire mass of the system, namely, drugs in thedonor and/or cell compartments in the system. The above equationshave led to the misconception that metabolism can be increased withincrease in efflux transport under linear conditions. A simple equation(eq. 7) that describes the fraction of dose metabolized (fmet) has beenused to correct for the misconception (Tam et al., 2003):

fmet �Mettotal

CD,0 � VD�

Mettotal

dose�

Mettotal

Mettotal � DR � DD � Dcell(7)

The parameter, fmet, expressed as a fraction of dose reflects themetabolic efficiency of the system and is more appropriate to describethe extent of drug metabolism in the Caco-2 monolayer.

Nonlinearity. To understand the transport mechanism of drugsacross the Caco-2 monolayer, the individual contribution of passivediffusion/permeability (PPD) and transporters (Pc) to the overall ortotal permeability (Papp) has to be distinguished under nonlinearsituations. The carrier-mediated permeability (Pc) has been defined asJmax/(K�m � CD,0) and Jmax/K�m for nonlinear and linear cases, respec-tively (Troutman et al., 2003b).

Pc � �Papp PPD� �Jmax�S

K�m � CD,0or

Jmax�SK�m

(8)

K�m and Jmax are estimates of the Michaelis-Menten constant (Km) andmaximum flux (Vmax) for the transporter, respectively. Usually, Pc, or

FIG. 7. Effects of saturable, apical efflux (Vmax � 50 nmol/min, Km � 10 �M) on the effect of permeability (Papp,A3B, Papp,B3A) and efflux ratio (EfR) for a solute ofhigh (A) and low (B) passive, diffusive clearance (CLd1 � CLd2 � CLd3 � CLd4 � CLd � 1 or 0.05 ml/min, respectively). In all simulations, CLabs � CLinflux � CLefflux �0, fap � fbaso � fcell � 1, and CLint,met � 0.1 ml/min; CD,0 was varied from 0.5 to 500 �M to reach greater degrees of saturation of apical efflux.


at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


the carrier-mediated permeability associated with the saturable pro-cess, is estimated from eq. 8 as Papp PPD. In the case of saturableapical efflux, PPD may be evaluated from use of inhibitors of effluxtransporters. The apical, absorptive transporter or enzyme activity isusually negligible in nontransfectants (Caco-2 cells), and transporter/metabolic activity shows up only in transfectants. Thus, Papp may beevaluated from transfectants and PPD from mock-transfected Caco-2cells (Cummins et al., 2001; Bhardwaj et al., 2005).

Analogously, the net flux due to carrier-mediated transport (Jc)is viewed as the difference between the total flux (J) and flux dueto passive diffusion (Jm) and was estimated as PcSCD,0. Theserelate to the estimated parameters, Jmax (for Vmax) and K�m (for Km).In most of the investigations, it has been assumed that apical effluxis the saturable process and that substrate concentration at thevicinity of the transport process is the initial loading concentration,CD,0.

FIG. 8. Effects of saturable, apical efflux on rates of metabolism expressed as a fraction of the dose metabolized (fmet,A3B) and (fmet,B3A) on the asymmetry of metabolismfor a solute of high or low passive, diffusive clearance (CLd1 � CLd2 � CLd3 � CLd4 � CLd � 1 or 0.05 ml/min, respectively); CLabs � CLinflux � CLefflux � 0, fap �fbaso � fcell � 1, and CLint,met � 0.1 ml/min. In this simulation, the Km (10 �M) for apical efflux was kept constant, whereas the Vmax was varied from 0 to 1000 nmol/minat CD,0 � 100 �M.

108 SUN AND PANG

at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


Jc � �J � Jm� �Jmax

K�m � CD,0CD,0 (9)

The strategies taken involved plotting Jc versus CD,0 in the Eadie-Hofstee plot for the appraisal of Jmax and K�m (Troutman and Thakker,

2003b). More often than not, curvature was observed in the plot of Jc

versus CD,0, since CD,0 is not the substrate concentration and theCaco-2 system is not a single barrier. To improve the fit, a Hillnumber or coefficient, �, an estimate of the number of binding sites onthe transporter, was added into eq. 10 (Stephens et al., 2001).

Jc �Jmax

K �m� � CD,0

�CD,0

� (10)

The equation was rearranged as shown below.

log �Jmax

Jc 1� � � log CD,0 � � log K �m (11)

In this “improvised method,” Jmax was first estimated from the y-intercept of the Eadie-Hofstee plot, and � and K�m were estimatedfrom refitting of data to eq. 11.

Materials and Methods

The Catenary Caco-2 Cell Model. A catenary model similar to those usedby others (Ito et al., 1999; Tam et al., 2003; Irie et al., 2004; Gonzalez-Alvarezet al., 2005) was employed to appraise various aspects of the effective per-meability, efflux ratio, and data interpretation on transporters and enzymes inCaco-2 monolayers. The model comprised the absorptive and ABC transport-ers of the apical membrane and influx and efflux barriers at the basolateralmembrane as two sets of potential barriers. The complexity of housing theABC transporters in a lipid bilayer was not considered (Tran et al., 2005)because the equilibrium of drug molecules between the aqueous compartmentand inner/outer plasma membrane would be achieved rapidly. In addition,paracellular transport was neglected, as this could be embedded in the param-eter for passive diffusion permeability.

The cell monolayer is viewed as a system that includes transporters andenzymes and allows for drug diffusion and cellular binding. The net fluxassociated with drug permeation in the catenary model involves two differentbarriers—one at the apical membrane, separating the apical compartment andcell, and the other at the basolateral membrane, separating the basolateralcompartment and cell (Fig. 1). The terms CLabs and CLd4 denote the absorptiveclearance ascribed to transporter-mediation and passive diffusion on the apicalmembrane, respectively; CLint,sec and CLd3 denote the secretory and passivetransport clearances of drug from the cell back to the apical compartment,respectively; CLd1 and CLinflux represent the passive and transporter-mediatedinflux clearances of drug between the basolateral compartment and cell mono-layer, respectively; CLd2 and CLefflux represent the passive and transporter-mediated efflux clearances of drug between cell monolayer and the basolateralcompartment, respectively. Unless specified, it was assumed that the clear-ances driven by passive diffusion are equal (CLd1 � CLd2 � CLd3 � CLd4 �CLd). These CLd terms relate to drug permeability by passive diffusion. A drugwith high CLd is highly permeable, whereas a drug with low CLd is poorlypermeable and requires transporters to facilitate transport.

In this study, simulations were conducted to produce profiles on transportand metabolism of solutes in the Caco-2 monolayer that was grown on atranswell of 24-mm insert diameter (six-well plate). Vap (1.5 ml) and Vbaso (2.5ml) were the volumes of buffer solution in the apical and basolateral compart-ments, respectively. One popular approach to estimate the cellular volume(Vcell) is by geometrical calculation, in which Vcell equals the insert area � cellheight. Blais et al. (1987) had measured the cell height by a pulse heightanalyzer and estimated Vcell to be 3.66 �l/mg protein. Since the protein contentof a confluent Caco-2 monolayer was 3 mg (Irie et al., 2004), the Vcell wascalculated to be 10.98 �l (3.66 �l/mg � 3 mg). In the second method,Yamaguchi et al. (2000) studied sulfanilamide, a poorly protein-bound com-pound whose transport was mediated only by passive diffusion, and estimatedVcell to be 12.15 �l from the amount of sulfanilamide cumulated in the cellmonolayer/external sulfanilamide concentration at equilibrium. Hence, in thepresent study, we took the average value of these Vcell values, 11.56 �l, for allsimulations.

Under nonlinear conditions, the rate of solute appearance in the receiver sidechanged with the initial concentration in the donor side, CD,0. These changes

FIG. 9. Effect of saturable, apical efflux (Vmax � 0.5 nmol/min and Km � 0.5 �M)on (A) carrier-mediated effective permeability, Pc, (B) the net flux associated withcarrier (Jc) at 1-h sampling, and (C) the resulting Eadie-Hofstee plots at differentloading concentrations (CD,0) in both A3B (open symbols) and B3A (solidsymbols) directions, when passive diffusion [CLd: E, F (0.01 ml/min), ƒ, � (0.1ml/min), and �, f (1 ml/min) (see inset)] was altered. The unbound fractions wereall set to unity; CLint,met was set to 0.01 ml/min; CLinflux � CLefflux � CLabs � 0.Note the nonconformity of some of the data to Michaelis-Menten kinetics. Usually,the early data points were used to estimate the Jmax (y-intercept at Jc/CD,0 � 0) thenthe K�m from the slope (K�m) or K�m with an equation associated with the Hillcoefficient (eq. 11).


at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


with concentration were observed with saturation of the apical efflux (Gonza-lez-Alvarez et al., 2005), apical influx (Tamai et al., 1997), basolateral influx(Irie et al., 2004), or metabolism (H. Sun, L. Zhang, E. C. Chow, G. Lin, K. S.Pang, and Z. Zuo, unpublished data). The linear model described in Fig. 1 wasmodified to include nonlinearity by substitution of the clearance term of theapical or basolateral membrane with Vmax/(Km � Cu), where Cu was theunbound drug concentration and Vmax and Km were the maximum velocity andMichaelis-Menten constant, respectively.

For simplicity, metabolism within the system was viewed as the aggregateamount of metabolite formed that was associated with the metabolic intrinsicclearance, CLint,met. Drug binding to proteins in the apical and basolateralcompartments would occur due to the presence of sloughed-off mucosal cells(unbound fractions, fap and fbaso, respectively) and within the cell (unboundfraction, fcell). Binding would affect the mass transfer and metabolic rates thatare based on unbound drug concentrations. The assumption was made that

formed metabolites did not compete with the parent drug for the transporters,enzymes, and drug-protein binding. The substrate was further assumed toundergo simple Michaelis-Menten kinetics for both transport and metabolismunder nonlinear conditions.

Solutions and Simulations. Mass balance equations (see Appendix) basedon the catenary model shown in Fig. 1 were developed. Under linear condition,the equations were solved for Papp and EfR with the program Maple 9.0(Waterloo Maple Inc., Waterloo, Canada). Computer simulations were per-formed under linear and nonlinear conditions with Scientist (Micromath, SaltLake City, Utah). Different CLd, CLint,met, CLabs, CLint,sec, Vmax, and Km wereassigned.

Results

Linear Conditions. Both sink and nonsink conditions were exam-ined. The Papp under nonsink conditions was best to describe drug

TABLE 1

Correlation between Vmax and Jmax, for apical efflux at various CLd, when CLint,met � 0.01 ml/min and Km � 0.5 �M

CLinflux � CLefflux � CLabs � 0 and CLd1 � CLd2 � CLd3 � CLd4 � CLd; all the unbound fractions were set to unity.

Assigned Vmax

A 3 B B 3 A

Jmax K�m* K�m* � Jmax K�m* K�m* �

nmol/min �M �M nmol/min �M �M

Case 1 (CLd � 0.01 ml/min)0.05 nmol/min 0.0134 1.75 3.75 1.20 0.0288 1.62 3.45 1.200.5 nmol/min 0.134 2.49 17.4 1.62 0.288 2.23 16.2 1.635 nmol/min 14.9 5.40 � 103 5.34 � 103 1.01 17.4 2.62 � 103 2.58 � 103 1.0125 nmol/min 7.45 � 103 2.78 � 106 2.78 � 106 1.00 1.56 � 104 2.52 � 106 2.52 � 106 1.0050 nmol/min 3.50 � 104 1.30 � 107 1.30 � 107 1.00 7.44 � 104 1.20 � 107 1.20 � 107 1.00

Case 2 (CLd � 0.1 ml/min)0.05 nmol/min 0.00641 1.66 1.71 1.01 0.00824 0.867 1.09 1.040.5 nmol/min 0.0640 1.60 3.57 1.21 0.0825 0.967 2.19 1.205 nmol/min 0.641 2.26 17.1 1.64 0.825 1.22 11.5 1.6925 nmol/min 3.90 113 161 1.18 4.18 8.34 49.7 1.5350 nmol/min 74.7 5.66 � 103 5.60 � 103 1.01 14.4 379 407 1.08

Case 3 (CLd � 1 ml/min)0.05 nmol/min 0.000671 1.58 1.59 1.00 0.000848 0.922 0.938 1.000.5 nmol/min 0.00671 1.58 1.77 1.03 0.00848 0.930 1.03 1.025 nmol/min 0.0671 1.63 3.63 1.21 0.0848 0.946 2.05 1.1925 nmol/min 0.336 1.89 10.6 1.53 0.424 1.03 6.43 1.5350 nmol/min 0.672 2.32 17.2 1.63 0.849 1.16 10.8 1.68

* True Km is 0.5 �M.

TABLE 2

Correlation between Km, K�m, and K�m for apical efflux at various CLd when CLint,met � 0.01 ml/min and Vmax � 0.5 nmol/min

CLinflux � CLefflux � CLabs � 0 and CLd1 � CLd2 � CLd3 � CLd4 � CLd; all the unbound fractions were set to unity.

Assigned Km

A 3 B B 3 A

Jmax* K�m K�m � Jmax* K�m K�m �


Case 1 (CLd � 0.01 ml/min)0.05 �M 0.134 0.252 12.0 2.10 0.288 0.226 11.2 2.120.5 �M 0.134 2.49 17.4 1.62 0.288 2.23 16.2 1.635 �M 0.135 23.9 39.0 1.20 0.290 21.5 36.0 1.2125 �M 0.138 109 116 1.04 0.296 99.3 107 1.0550 �M 0.139 205 209 1.02 0.298 188 192 1.02

Case 2 (CLd � 0.1 ml/min)0.05 �M 0.0640 0.173 1.49 1.49 0.0825 0.0874 0.857 1.460.5 �M 0.0640 1.60 3.57 1.21 0.0825 0.967 2.19 1.205 �M 0.0640 15.8 17.7 1.04 0.0825 9.78 11.2 1.0525 �M 0.0640 78.6 79.4 1.01 0.0826 48.8 49.6 1.0150 �M 0.0640 156 156 1.00 0.0826 97.4 97.9 1.00

Case 3 (CLd � 1 ml/min)0.05 �M 0.00671 0.160 0.233 1.06 0.00848 0.0908 0.123 1.040.5 �M 0.00671 1.58 1.78 1.03 0.00848 0.930 1.03 1.025 �M 0.00671 15.8 16.0 1.01 0.00848 9.22 9.40 1.0125 �M 0.00672 79.4 79.3 1.00 0.00849 46.4 46.5 1.0050 �M 0.00671 158 158 1.00 0.00849 92.8 92.8 1.00

* True Vmax � 0.5 nmol/min.

110 SUN AND PANG

at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


transport, and the incremental Papp, estimated at different time intervals,was less stable than the cumulative Papp [based on (amount transported)/(time 0 to time of sampling)] in the description of EfR (data not shown).Hence, further analyses only pertained to the cumulative Papp. Althoughan explicit solution was found for Papp by matrix inversion for nonsinkconditions, the solution was not readily presentable because the expres-sion was very lengthy. Instead, computer simulation was used to showpatterns of Papp and the factors that modulated Papp.

Papp and EfR in Absence of Metabolic Enzyme and TransporterActivity. When transporter and enzyme activities were absent, Papp

was found to be time- and permeability-dependent (Fig. 2). With the

assumption that CLd � CLd2 � CLd3 � CLd4 � CLd, Papp,A3B andPapp,B3A profiles were found to be identical (Fig. 2), resulting in EfRvalues that were constant (unity) and time-invariant. Binding withinthe cell further affected the Papp patterns (Fig. 2, insets), renderingdecreased Papp with decreasing unbound fraction, fcell. However,binding would not alter EfR, as predicted by eq. 12.

Papp and EfR in the Presence of Efflux or Metabolism. Papp wasfound to be modulated by CLint,sec (Fig. 3) at both high and lowpassive diffusion clearance, CLd. Papp,B3A was increased, whereasPapp,A3B was decreased with increasing CLint,sec; thus, EfRs wereincreased, and the values were higher for drugs with low CLd (Fig. 3).

FIG. 10. Correlation between estimated parameter Jmax at 1-h sampling versus the assigned Vmax (A, B, and C) and K�m and K�m versus the assigned Km (D, E, and F) forthe apical efflux transporter in the Caco-2 cell monolayer (data of Tables 1 and 2) when different CLd (0.01, 0.1, and 1 ml/min) were used for simulation. D–F, K�m (E,F) and K�m (‚, Œ) were represented against Km. Transport in A3B and B3A directions was denoted by open and solid symbols, respectively. For the simulations, allthe unbound fractions were set to unity; CLinflux � CLefflux � CLabs � 0; CLint,met � 0.01 ml/min.


at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


By contrast, Papp values were identical in both absorptive and secre-tory directions, and Papp decreased with increasing CLint,met (Fig. 4);the Papp values were lower for drugs of lower CLd. Values of theresultant EfR converged at unity, regardless of values of CLint,met.

Solution for EfR. The solution for EfR existed in a concise formunder linear conditions, and was similar to that reported by Adachi etal. (2001), with the exception of the unbound fractions.

EfR �fbaso�CLd1 � CLinflux�CLint,sec � CLd3

fap�CLd2 � CLefflux�CLabs � CLd4(12)

CLint,met and fcell were absent in eq. 12 for EfR. The equation impliedthat, under linear conditions, intracellular metabolism and the bindingof drugs to intracellular protein would not influence EfR. Whenprotein was absent (fap � fbaso � 1) or when binding was similar(fap � fbaso), the binding terms would disappear and the fbaso and fap

terms were canceled, and eq. 12 was further simplified.

EfR ��CLd1 � CLinflux�CLint,sec � CLd3

�CLd2 � CLefflux�CLabs � CLd4(13)

Based on eq. 13, the effect of the various clearances on EfR wasinvestigated. The results are summarized in three-dimensional graph-ical presentations. As shown in Fig. 5A, values of EfR fell between 1and 2 when the test compounds were of high CLd (�1 ml/min), evenwhen the efflux transporter activity (CLint,sec) was substantial (0.8ml/min). The same was observed for compounds that were activelytaken up by the apical transporters (CLabs CLd4), rendering valuesof EfR below unity (Fig. 5B). However, extremely high EfR valueswould result with increasing CLint,sec but decreasing CLd (Fig. 5A) orCLabs (Fig. 5B), which disallowed rapid equilibration of drug betweenthe cellular and the apical compartment. By contrast, decreasingvalues of CLinflux/CLefflux would furnish modest values of EfR (be-tween 0.5 and 2) even for substrates that were readily effluxed out atthe apical membrane (CLint,sec � 0.9 ml/min) (Fig. 5C).

With the assumption that CLd1 equals CLd2, CLd3 equals CLd4,and only apical efflux is carrier-mediated (i.e., CLabs � CLinflux �CLefflux � 0), eq. 13 was further simplified.

EfR � 1 �CLint,sec

CLd4(14)

Equation 14 inferred that EfR was dependent on the relative magni-tude of CLint,sec and CLd4 in absence of transporter-mediated apicalinflux, basolateral influx and efflux. For a Pgp substrate such asverapamil, which has a high CLd, CLint,sec � CLd and the efflux ratiowas close to 2. For other drugs of low CLd that are Pgp substrates(CLint,sec CLd), EfR was disproportionately higher (Lentz et al.,2000; Polli et al., 2001). Fortunately, the ambiguity may be removedwith the use of potent Pgp inhibitors that would drastically reduce theEfR, revealing AQ and SQ values that denote Pgp transport. There-fore, EfR values needed to be appraised and compared, together withother permeability data with Pgp inhibitors, such as AQ and SQ.

Asymmetry of Metabolism. The amount of metabolite formed(Mettotal) was normalized to the dose and expressed as fmet (eq. 7). Afaster metabolite formation rate was observed with apical over baso-lateral dosing, even with identical initial concentration administered.The observed asymmetry in metabolism resulting from administrationinto the apical or basolateral side was expected because of differencesin volumes (Vap and Vbaso). When identical permeability clearancesexisted across the apical and basolateral membranes (CLd1 � CLd2 �CLd3 � CLd4) and carrier-mediated transport was absent (CLabs �CLint,sec � CLinflux � CLefflux � 0), fmet,B3A/fmet,A3B was less than1 (Fig. 6). When Vap and Vbaso were equal, the metabolic asymmetrydisappeared (simulation not shown). With increasing incubation time,fmet,A3B, fmet,B3A, and fmet,B3A/fmet,A3B increased and finallyapproached 1 (when time approached infinity in this system, all drugsultimately became metabolized and fmet in both directions equaledunity). Again, the passive diffusion clearance (CLd) was important,and a greater permeability of the solute facilitated a faster entry ofdrug for cellular metabolism (Fig. 6, A and B). Carrier-mediated

FIG. 11. Effects of saturable, apical absorption (Vmax � 0.5 nmol/min and Km � 0.5�M) on (A) carrier-mediated permeability, Pc, (B) the net flux associated withcarrier Jc at 1-h sampling, and (C) the resulting Eadie-Hofstee plots at differentloading concentrations (CD,0) in both A3B (open symbols) and B 3A (solidsymbols) directions when passive diffusion [CLd: E, F (0.01 ml/min), ƒ, � (0.1ml/min), and �, f (1 ml/min) (see inset)] was altered. The unbound fractions wereall set to unity; CLint,met was set to 0.01 ml/min; CLinflux � CLefflux � CLint,sec �0. Note the nonconformity of some of the data to Michaelis-Menten kinetics.Usually, the early data points are used to estimate the Jmax (y-intercept at Jc/CD,0 �0) then the K�m from the slope (K�m) or K�m with an equation associated with the Hillcoefficient (eq. 11).

112 SUN AND PANG

at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


efflux on the apical side (CLint,sec) decreased values of fmet,B3A andfmet,A3B by removing cellular substrate for metabolism and alsoaffected the value of fmet,B3A/fmet,A3B, accentuating the asymmetryobserved for metabolite formation.

Nonlinear Conditions. Saturable terms, Vmax and Km, were used asreplacement of the intrinsic clearances in the catenary model (Fig. 1);CLd1, CLd2, CLd3, and CLd4 were set equal to CLd, and the unboundfractions were set to unity. For each simulation, only one saturableprocess (apical influx or efflux, basolateral influx or efflux, or intra-cellular metabolism) was considered in the Caco-2 system. Simula-tions were then performed to provide data for appraisal of the appro-

priateness of the estimates (K�m or K�m and Jmax) versus the assignedKm and Vmax for the saturable process. The strategy generated sets ofdata with assigned constants (Km and Vmax, CLd and CD,0), and thesewere further manipulated into forms used in data interpretation. Thetrue values and the derived estimates were then compared.

First, the manner in which saturation of apical efflux transportersaffected Papp and EfR at increasing CD,0 was examined. Solutes withhigh CLd entered the cell compartment rapidly and readily saturatedapical efflux with increasing CD,0. Values of Papp,A3B increased withCD,0 (Fig. 7, top panel), whereas Papp,B3A decreased with CD,0 (Fig.7, middle panel). These resulted in attenuated EfRs (Fig. 7, bottom

FIG. 12. Correlation between estimated parameter Jmax at 1-h sampling versus the assigned Vmax (A, B, and C) and K�m and K�m versus the assigned Km (D, E, and F) forthe absorptive, apical transporter in the Caco-2 cell monolayer (data of Tables 3 and 4) when different CLd (0.01, 0.1, and 1 ml/min) were used for simulation. D–F, K�m(EF) and K�m (ƒ�) estimates are plotted against the true Km. Transport in A3B and B3A directions was denoted by open and solid symbols, respectively. For thesimulations, all the unbound fractions were set to unity; CLinflux � CLefflux � CLint,sec � 0; CLint,met � 0.01 ml/min.


at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


panel). By contrast, less changes for Papp,A3B and Papp,B3A werefound for drugs of low CLd because these failed to enter the cellreadily to saturate apical efflux transporters. Second, the effect ofsaturable apical efflux on metabolism was considered. The presenceof apical efflux reduced the fraction of dose metabolized, fmet, sinceefflux was competing with metabolism. An increase in Vmax for apicalefflux also reduced fmet, regardless of whether the drug was given toA or B, and this existed when the drug equilibrated rapidly (Fig. 8A)or slowly (Fig. 8B).

Eadie-Hofstee Plot. Characteristic profiles were obtained uponplotting Pc (eq. 8) and Jc (eq. 9) against CD,0 (Figs. 9, A and B) whensaturable efflux existed. A lack of conformity to linearity was com-monly observed with the Eadie-Hofstee plot for the estimation of K�m

and Jmax (flux versus flux/CD,0) (Fig. 9C). Usually, the intercept andslope of the plot associated with the straight line component would beused for estimation of Jmax and K�m, respectively (see straight line ininset, Fig. 9C, for drug of high CLd). Poor estimates would resultwhen curvature existed in the Eadie-Hofstee plot for solutes of lowCLd (�1 ml/min). When eq. 10 was used to improve the fit with theHill coefficient, �, K�m, estimated from refitting of data to eq. 11, wasnot much improved compared with K�m (see below). The estimate ofJmax remained unchanged since the parameter was estimated from they-intercept of the Eadie-Hofstee plot.

Estimation of Parameters: Nonlinear Apical Efflux. With vary-ing CLd (0.01, 0.1, or 1 ml/min), further simulations were performedwith varying Vmax (0.05 to 50 nmol/min) at a fixed Km (0.5 �M) for

TABLE 3

Correlation between Vmax and Jmax, for apical influx at various CLd, when CLint,met � 0.01 ml/min and Km � 0.5 �M

CLinflux � CLefflux � CLint,sec � 0 and CLd1 � CLd2 � CLd3 � CLd4 � CLd; all the unbound fractions were set to unity.

Assigned Vmax

A 3 B B 3 A

Jmax K�m* K�m* � Jmax K�m* K�m* �


Case 1 (CLd � 0.01 ml/min)0.05 nmol/min 0.0134 0.581 0.864 1.08 0.0285 14.8 14.6 1.010.5 nmol/min 0.134 0.612 5.28 1.62 0.289 25.3 51.1 1.295 nmol/min 1.34 1.38 34.1 1.82 331 1.24 � 105 1.24 � 105 1.0025 nmol/min 86.4 1.02 � 104 N.A. N.A. 1.60 � 104 5.99 � 106 5.99 � 106 1.0050 nmol/min 157 1.87 � 104 N.A. N.A. 6.86 � 104 2.56 � 107 2.56 � 107 1.00

Case 2 (CLd � 0.1 ml/min)0.05 nmol/min 0.00640 1.15 1.66 1.08 0.00825 1.40 1.28 0.990.5 nmol/min 0.0640 1.34 5.22 1.39 0.0825 1.35 3.51 1.265 nmol/min 0.642 2.66 28.7 1.73 0.826 2.01 19.4 1.7025 nmol/min 38.3 6608 N.A. N.A. 6.91 343 374 1.0850 nmol/min 28.8 5229 N.A. N.A. 1199 9.36 � 104 9.35 � 104 1.00

Case 3 (CLd � 1 ml/min)0.05 nmol/min 0.000671 1.60 1.61 1.00 0.000848 0.91 0.95 1.010.5 nmol/min 0.00671 1.59 2.02 1.06 0.00848 0.93 1.15 1.045 nmol/min 0.0671 1.70 5.95 1.36 0.0848 0.97 3.33 1.3225 nmol/min 0.336 2.39 18.3 1.65 0.424 1.18 11.1 1.6950 nmol/min 0.674 4.18 32.7 1.61 0.849 1.54 18.5 1.75

N.A., not available due to negative Jmax values.* True Km is 0.5 �M.

TABLE 4

Corelation between Km (0.5 �M), K�m, and K�m for apical influx at various CLd when CLint,met � 0.01 ml/min and Vmax � 0.5 nmol/min

CLinflux � CLefflux � CLint,sec � 0 and CLd1 � CLd2 � CLd3 � CLd4 � CLd; all the unbound fractions are set to unity.

Assigned Km

A 3 B B 3 A



Case 1 (CLd � 0.01 ml/min)0.05 �M 0.134 0.0613 3.96 2.10 0.288 4.70 31.6 1.600.5 �M 0.134 0.612 5.28 1.62 0.289 25.3 51.1 1.295 �M 0.134 6.11 10.3 1.18 0.284 133 140 1.0425 �M 0.134 30.4 33.0 1.04 0.264 450 450 1.0050 �M 0.134 60.3 62.1 1.02 0.253 799 799 1.00

Case 2 (CLd � 0.1 ml/min)0.05 �M 0.0640 0.100 3.04 1.85 0.0825 0.180 1.78 1.550.5 �M 0.0640 1.34 5.22 1.39 0.0825 1.35 3.50 1.265 �M 0.0641 13.7 17.4 1.09 0.0824 12.0 13.8 1.0525 �M 0.0642 67.4 69.2 1.02 0.0822 56.6 57.2 1.0150 �M 0.0643 133 134 1.00 0.0820 110 111 1.00



114 SUN AND PANG

at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


apical efflux. The simulations were repeated by changing the Km (0.5to 50 �M) at a fixed Vmax (0.5 nmol/min) to arrive at different sets ofJc at varying sampling times (0–120 min). The plots Pc, the carrier-mediated flux Jc versus CD,0, and the Eadie-Hofstee plot (Figs. 9,A–C) revealed that CLd strongly influenced the curvature of theEadie-Hofstee plot with time. Higher values of Pc and Jc wereobtained for the B3A than the A3B direction, yielding asymmetricJmax and lower K�m estimates (Tables 1 and 2). The early data pointsof the Eadie-Hofstee plots were used to estimate the Jmax and K�m fromthe intercept and slope, respectively (Fig. 9C). A linear relationshipwas observed for the Eadie-Hofstee plot only with higher CLd (Fig.9C, inset). Upon presenting the entire set of estimates, it was apparentthat Jmax differed from the assigned Vmax and varied according to thevalue of CLd (Fig. 10; Table 1). The correlation was poor at low CLd

(0.01 ml/min), whereas an improved and linear correlation was iden-tified at increasing CLd (�1 ml/min) (Fig. 10, A–C). The estimatedK�m and K�m were less affected by CLd. Linear relationships were foundbetween K�m and K�m and the assigned Km (Fig. 10, D–F); the valuesbecame closest when CLd was highest (Table 2). However, no furtherimprovement was provided with use of eq. 10 or eq. 11 to arrive at K�m(Table 2; Fig. 10, D–F).

Estimation of Parameters: Nonlinear Apical Absorption. Withvarying CLd (0.01, 0.1, or 1 ml/min), simulations were performedwith varying Vmax (0.05 to 50 nmol/min), but at a fixed Km (0.5 �M),and at varying Km (from 0.5 to 50 �M) but at a fixed Vmax (0.5nmol/min) for the apical absorption transporter to arrive at sets of Pc

and Jc data (Fig. 11). With saturable apical absorption, higher Pc andJc values were obtained in the B3A and not A3B direction, exceptfor drugs of a very low CLd (Fig. 11, A and B). Again, curvatureswere observed for the Eadie-Hofstee plot, except for drugs of highCLd (Fig. 11C, inset). The estimated Jmax was poorly correlated to theVmax, especially when CLd was lower (Fig. 12, A–C; Table 3), but alinear correlation between Jmax and Vmax was attained when CLd

values were high (Fig. 12C). The correlation between K�m and K�m toKm was acceptable among the various CLd (Fig. 12, D–F; Table 4),and an improved correlation existed when the CLd values were high(Fig. 12F). Again, no further improvement was demonstrated with useof eq. 10 or eq. 11 to arrive at K�m (Table 4; Fig. 12, D–F).

Estimation of Parameters: Nonlinear Metabolism. With varyingCLd (0.01, 0.1, or 1 ml/min), simulations were performed with varyingVmax (0.05 to 50 nmol/min) at a fixed Km (0.5 �M) for the metabolismand at varying Km (0.05 to 50 �M) at a fixed Vmax (0.5 nmol/min) toarrive at sets of Pc and Jc data (Fig. 13). With saturable metabolism,values of Pc were higher with higher CLd (Fig. 13A). Higher Jc valueswere observed in the A3B but not B3A direction (Fig. 13B), yieldingcorrespondingly higher Vmax (y-intercept) (Fig. 13C). The correlationbetween Jmax and Vmax was poor regardless of values for CLd (Fig. 14,A–C; Table 5). By contrast, a linear correlation was found between K�mand K�m with Km among the various CLd (Fig. 14, D–F; Table 6); thecorrelation improved when CLd values were high. Again, no furtherimprovement was provided with use of eq. 10 or eq. 11 to arrive at K�m(Table 6; Fig. 14, D–F).

Sampling Time. Sampling in the Caco-2 system was often per-formed at 1- or 2-h intervals. However, the manner in which samplingtime affected the Jc estimates was seldom studied. We examined howsampling time affected estimation of Pc and Jc within the sets ofsimulations performed on saturation of apical efflux, apical absorp-tion, and metabolism. Indeed, the sampling time affected values of Jc,although saturation profiles were retained versus CD,0 in both A3Band B3A (Fig. 15). Hence, the general shapes were retained for allof the Eadie-Hofstee plots, but estimates for K�m and Jmax would

change when the sampling time, whether 0.5, 1, or 2 h, was used (datanot shown).

Effect of Nonlinear Metabolite Formation. Simulation was per-formed to examine how the various saturable processes affectedmetabolite formation. For drugs of high CLd (1 ml/min), saturableapical efflux (Fig. 16A), apical absorption (Fig. 16B), basolateralefflux (Fig. 16C), and basolateral influx (Fig. 16D) resulted in asym-metric metabolite formation, namely, higher for the secretory direc-

FIG. 13. Effect of saturable cellular metabolism (Vmax � 0.5 nmol/min, Km � 0.5�M) on (A) carrier-mediated effective permeability, Pc, (B) the net flux associatedwith carrier Jc at 1-h sampling, and (C) the resulting Eadie-Hofstee plots at differentloading concentrations (CD,0) in both A3B (open symbols) and B3A (solidsymbols) directions when passive diffusion [CLd: E, F (0.01 ml/min), ƒ, � (0.1ml/min), and �, f (1 ml/min)] was altered. The unbound fractions were all set tounity; CLinflux � CLefflux � CLabs � CLint,sec � 0. Note the nonconformity of someof the data to Michaelis-Menten kinetics. Usually, the early data points are used toestimate the Jmax (y-intercept at Jc/CD,0 � 0) then the K�m from the slope (K�m) orK�m with an equation associated with the Hill coefficient (eq. 11).


at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


tion over the absorptive direction. The trend was partially due to thehigher volume of the basolateral compartment; the dose VbasoCD,0

was larger than VapCD,0. Patterns for metabolite formation undersaturable apical efflux (Fig. 16A) and basolateral efflux (Fig. 16C)were similar; lesser metabolism was observed with higher Vmaxs tobring drug molecules out of the cell. The cumulative amount ofmetabolite formed was time-dependent (Fig. 16). In addition, similarpatterns existed between the profiles of metabolite formation forsaturable absorption or basolateral influx; greater extents of metabo-lism were found associated with higher Vmax, since this tended tobring more drug molecules into the cell (Figs. 16, B and D). For drugs

of poor diffusion clearance (CLd � 0.01 ml/min), greater amounts ofmetabolite formed were seen in the B3A direction, except when theapical transporter was present (Fig. 16F). At low CLd, drug transportinto and out of the cell would be nominal, unless there was enhancedapical absorption by transporters, rendering a greater extent of me-tabolism in the A3B direction; the trend intensified with increasedVmax for the absorptive transporter (Fig. 16F).

When metabolite formation (at 1 h) was further examined at in-creasing CD,0, a trend indicating higher metabolite formation for theB3A direction over the A3B direction was unilaterally observed(Fig. 17, A–D), regardless of the saturable process (apical efflux,

FIG. 14. Correlation between estimated parameter Jmax at 1-h sampling versus the assigned Vmax (A, B, and C) and K�m and K�m versus the assigned Km (D, E, and F) forsaturable cellular metabolic activity in the Caco-2 cell monolayer (data of Tables 5 and 6) when different CLd (0.01, 0.1, and 1 ml/min) were used for simulation. Transportin A3B and B3A directions was denoted by open and solid symbols, respectively. In (B), K�m and K�m were represented by E, F and ‚, Œ, respectively. For thesimulations, all the unbound fractions were set to unity; CLinflux � CLefflux � CLabs � CLint,sec � 0.

116 SUN AND PANG

at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


apical absorption, basolateral influx, or basolateral efflux). The cu-mulative amount of metabolite formed was less with higher Vmax

associated with saturable apical efflux (Fig. 17A) and saturable ba-solateral efflux (Fig. 17C), and opposite trends existed with saturationof apical absorption and basolateral influx (Fig. 17, B and D). Fordrugs of poor diffusion clearance (CLd � 0.01 ml/min), higher me-tabolite formation was again seen in the B3A direction, except whenapical absorptive transport was present (Fig. 17F). At low CLd,transport into and out of the cell would be nominal unless there wasenhanced absorption by transporters, rendering more metabolism inthe A3B direction. The trend intensified when the Vmax for theabsorptive transporter was increased (Fig. 17F).

Effect of Nonlinear Processes on EfR. Patterns of EfR undernonlinear conditions were examined. In these simulations, the satura-

ble component was assigned a Vmax of 0.5 nmol/min and Km of 0.5�M, and CLd1 � CLd2 � CLd3 � CLd4 � CLd. EfR values differedwith saturable apical efflux and metabolism ( 1) and apical absorp-tion (EfRs �1) but approached unity with high CLd and increasingCD,0 at a higher degree of saturation of the process. Saturable apicalefflux (Fig. 18A) resulted in EfR values that exceeded unity butapproached unity at high CD,0 when the apical efflux transporter wasbecoming saturated. Similar to the linear case, lower EfR values werefound at higher CLd. A reverse trend was observed for saturable apicalinflux, showing EfR values below unity (Fig. 18B); increasinglyhigher EfRs resulted with higher CLd and CD,0. With saturable me-tabolism, an upswing then downswing pattern was observed for EfRwith CD,0, and the changes were more pronounced for higher CLd

(Fig. 18C). For saturable apical efflux, influx, and metabolism, rapid

TABLE 5

Correlation between Vmax and Jmax for cellular metabolism (Km � 0.5 �M) at various CLd when other satiable components(apical/basolateral influx, efflux) were absent

CLinflux � CLefflux � CLint,sec � CLabs � 0 and CLd1 � CLd2 � CLd3 � CLd4 � CLd; all the unbound fractions were set to unity.

Assigned Vmax

A 3 B B 3 A

Jmax K�prime]m * K�m* � Jmax K�m * K�m* �


Case 1 (CLd � 0.01 ml/min)0.05 nmol/min 0.0256 1.13 3.40 1.29 0.0238 1.05 3.13 1.280.5 nmol/min 0.256 1.75 18.4 1.72 0.238 1.57 17.1 1.735 nmol/min 124 2.92 � 104 2.91 � 104 1.00 92.4 2.17 � 104 2.16 � 104 1.0025 nmol/min 2.81 � 104 6.82 � 106 6.82 � 106 1.00 2.65 � 104 6.27 � 106 6.27 � 106 1.0050 nmol/min 1.32 � 105 3.11 � 107 3.11 � 107 1.00 1.22 � 105 2.89 � 107 2.89 � 107 1.00

Case 2 (CLd � 0.1 ml/min)0.05 nmol/min 0.0293 1.23 1.87 1.10 0.0205 0.858 1.26 1.080.5 nmol/min 0.293 1.37 8.20 1.53 0.205 0.927 5.59 1.515 nmol/min 3.21 50.1 96.3 1.26 2.06 2.79 35.6 1.6925 nmol/min 199 1.32 � 104 1.31 � 104 1.00 267 1.77 � 104 1.76 � 104 1.0050 nmol/min 1.09 � 104 7.27 � 105 7.27 � 105 1.00 2.20 � 104 1.48 � 106 1.48 � 106 1.00

Case 3 (CLd � 1 ml/min)0.05 nmol/min 0.0310 1.34 1.80 1.07 0.0189 0.812 1.05 1.050.5 nmol/min 0.310 1.43 7.63 1.51 0.189 0.846 4.52 1.475 nmol/min 3.15 10.0 57.8 1.52 1.89 1.51 29.2 1.7925 nmol/min 1.34 � 107 8.60 � 108 8.60 � 108 1.00 1.08 � 105 6.93 � 106 6.93 � 106 1.0050 nmol/min 3.41 � 109 2.19 � 1011 2.19 � 1011 1.00 4.11 � 109 2.64 � 1011 2.64 � 1011 1.00

* True Km is 0.5 �M.

TABLE 6

Correlation between Km, K�m, and K�m for cellular metabolism at various CLd when other satiable components (apical/basolateral influx, efflux) were absent

CLinflux � CLefflux � CLint,sec � CLabs � 0 and CLd1 � CLd2 � CLd3 � CLd4 � CLd; all the unbound fractions were set to unity.

Assigned Km

A 3 B B 3 A



Case 1 (CLd � 0.01 ml/min)0.05 �M 0.256 0.179 13.4 2.19 0.238 0.160 12.6 2.200.5 �M 0.256 1.75 18.4 1.72 0.238 1.57 17.1 1.735 �M 0.258 16.7 35.2 1.29 0.240 15.0 32.6 1.2925 �M 0.263 75.9 87.7 1.08 0.244 69.2 80.8 1.0850 �M 0.266 142 149 1.04 0.247 130 137 1.04

Case 2 (CLd � 0.1 ml/min)0.05 �M 0.293 0.139 6.02 2.03 0.205 0.0951 3.93 1.980.5 �M 0.293 1.37 8.20 1.53 0.205 0.927 5.59 1.515 �M 0.293 13.6 20.0 1.14 0.206 9.21 13.9 1.1525 �M 0.295 66.7 70.0 1.03 0.206 45.6 48.6 1.0350 �M 0.295 131 133 1.01 0.207 90.3 92.1 1.01




at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


drug entry by passive diffusion (high CLd) counteracted the effects ofthe saturable process on EfR. Eventually, the EfR values returned tounity when the apical efflux, absorptive, or metabolic process failed tocontribute significantly to the system with increasing CD,0, as if onlypassive diffusion prevailed under linear conditions (Fig. 5; eq. 12).

Other Simulations. Other scenarios such as saturation of basolat-eral influx or basolateral efflux were not examined. This was becausethe patterns generated would be similar to those simulated for satu-ration of apical influx and apical secretion (mirror images for donorversus receiver side), although minor differences would exist due todifferences in volumes (Vap and Vbaso).

Discussion

The Caco-2 system has become the gold standard to relate drugpermeability to oral drug absorption (Hidalgo et al., 1989; Lentz et al.,2000; Polli et al., 2001; Thiel-Demby et al., 2004). Due to the prevalentuse, a thorough examination in revisiting the terminologies used inliterature and data interpretation in the Caco-2 monolayer is thereforejustified. Equation 1 was established under Fick’s law to define drugpermeability, Papp, with the assumptions that the driving force is theconcentration gradient difference and a single barrier exists for theCaco-2 monolayer (Hilgers et al., 1990). Although Papp is a time-depen-dent variable and higher Papp is associated with higher CLd, the ratio ofPapp (Papp,B3A/Papp,A3B) yields a meaningful EfR that is time-indepen-dent and CLd-invariant when transporter activity is absent.

The meaning of EfR under linear conditions was further clarified byeq. 12. As shown in eq. 12, EfR is dependent not only on the apical,secretory intrinsic clearance (CLint,sec) but also on the relative mag-

nitude of passive diffusion and transport clearances at the basolateralmembrane (CLd1 � CLinflux)/(CLd2 � CLefflux), CLd3 and CLd4, andthe apical absorptive clearance CLabs. Hence, the value of EfR maynot always identify whether a test compound is a Pgp substrate. Fordrugs of rapid uptake clearances at the apical membranes, values ofEfR are reduced by high CLd4 (eq. 14) or CLabs values (Fig. 5); theasymmetrical influx and efflux on basolateral membrane (CLinflux/CLefflux) can also modulate the EfR. The same comment was made byAdachi et al. (2001). Even with the simplification that CLabs � 0 andall diffusive clearances are identical (CLd1 � CLd2 � CLd3 � CLd4),values of EfR may still hover around unity for drugs whose diffusionclearance (CLd4) is high in relation to that of the secretory clearance(eq. 14) (Lentz et al., 2000).

In addition, EfR will be affected differentially by secretion andmetabolism under linear conditions. Pgp-, BCRP-, or MRP2-mediatedefflux tends to provide higher EfR values when CLint,sec CLd andwhen the passive diffusion clearance (CLd) is low, rendering asym-metry in permeability from the B3A direction over the A3B direc-tion (Fig. 3). For solutes of high CLd, ambiguous EfR values (close tounity) may result unless CLint,sec is extremely high (Fig. 3A). Thepossibility that the compound is a Pgp substrate needs to be clarifiedwith use of inhibitors by the AQ and SQ (eqs. 3 and 4). For soluteswith low CLd, higher EfR values will result as a consequence ofsecretion for Pgp substrates (Fig. 3B). By contrast, EfR is not affectedby metabolism; the metabolic intrinsic clearance affects Papp in theB3A and A3B directions to the same extent, thereby EfR remainsunchanged (Fig. 4). The extent of metabolite formation in the A3Bdirection is higher than that in the B3A direction. The pattern isinherently asymmetric simply because of unequal Vap and Vbaso, and

FIG. 15. Time- and concentration-dependent profiles of Jc in A3B (upper panel) and B3A (lower panel) directions when saturation of (A) apical efflux (Vmax � 0.5nmol/min, Km � 0.5 �M; CLinflux � CLefflux � CLabs � CLint,met � 0), (B) apical absorption (Vmax � 0.5 nmol/min, Km � 0.5 �M; CLinflux � CLefflux � CLint,sec �CLint,met � 0), and (C) metabolism (Vmax � 0.5 nmol/min, Km � 0.5 �M; CLinflux � CLefflux � CLint,sec � CLabs � 0) were examined. The simulations were set as follows:CLd1 � CLd2 � CLd3 � CLd4 � CLd � 0.1 ml/min; the unbound fractions were set to unity.

118 SUN AND PANG

at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


FIG. 16. Modulation of metabolite formation (CLint,met � 0.01 ml/min) by saturable efflux (A, E) or influx (B, F) on the apical membrane, or efflux (C, G) or influx (D, H) onthe basolateral membrane, for drugs of high CLd (CLd � CLd1 � CLd2 � CLd3 � CLd4 � 1 ml/min) (A–D) or low CLd (0.01 ml/min) (E–H). Simulation was performed at CD,0 �100 �M when only one saturable pathway existed (Km � 0.5 �M and Vmax was varied from 0.5 to 50 nmol/min); all the unbound fractions were set to unity.


at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


metabolite formation is reduced by apical secretion in both directions(Fig. 6).

Saturation of transport and metabolic processes, however, will beencountered with increasing drug loading concentration within the

Caco-2 cell system. The biggest disappointment in other existing datainterpretation strategies lies in treating the Caco-2 cell system as asingle barrier for parameter estimates of saturable processes. Evenwhen perfect data for Pc and Jc are attainable with use of “specific”

FIG. 17. Modulation of metabolite forma-tion by saturable efflux (A, E) or influx (B,F) on the apical membrane, or efflux (C, G)or influx (D, H) on the basolateral mem-brane, for drugs of high CLd (CLd �CLd1 � CLd2 � CLd3 � CLd4 � 1 ml/min)(A–D) or low CLd (0.01 ml/min) (E–H).Simulation was performed for CD,0 � 1 to500 �M, with sampling at 1 h, when onlyone saturable pathway existed (Km � 0.5 �Mand Vmax was varied from 0.5 to 50 nmol/min); CLint,met � 0.01 ml/min, and all theunbound fractions were set to unity.

120 SUN AND PANG

at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


inhibitors that completely block the secretory process, or when trans-port/metabolic activities are associated only with transfected but notmock-transfected cells to enable precise measurements, the Jmax andK�m estimates are far from the true values (Tables 1–6). The method,based on the assumption that the Caco-2 cell system is a single barrier,will yield inappropriate estimates.

In addition, the time of sampling (Fig. 15), the passive perme-ability of drug, tissue binding, and metabolism, and the parameterspace of Km and Vmax strongly influence both the correlation andaccuracy of prediction. Saturation in apical absorption, efflux, ormetabolism in the Caco-2 system results in similar Eadie-Hofstee

plots for the receiver side that show nonconformity to simpleMichaelis-Menten kinetics (Figs. 9C, 11C, and 13C). The net fluxis a consequence of multiple processes involving more than asingle barrier, and CD,0 cannot estimate the substrate concentrationin the vicinity of the transporter or enzyme. The estimated Jmax ispoorly correlated to the true Vmax, especially for substrates of lowdiffusive clearance (CLd � 1 ml/min); the observation persistswhen apical influx, secretion, or cellular metabolism is the satu-rable process (Figs. 10, 12, and 14, A–C; Tables 1, 3, and 5). Jmax

is correlated with the true Vmax only for solutes of high CLd forapical secretion and absorption but not for cellular metabolism(Figs. 10C, 12C, and 14C). An improved correlation is also ob-served with reduction of assigned values of Vmax (�20 nmol/min)or increased values of CD,0 for simulation (data not reported). Bycontrast, a linear correlation is found between the estimated K�m orK �m and Km when apical influx, secretion, or cellular metabolism isthe saturable process (Figs. 10, 12, and 14, D–F; Tables 2, 4, and6). The finding is in contrast to that of Bentz et al. (2005), whocommented that a better correlation existed between Jmax and Vmax

and not K�m or K �m and Km. The correlation is again dependent onvalues of CLd, since drugs with higher CLd equilibrate fasterbetween the donor, cellular, and receiving sites, rendering im-provement in the correlation (Figs. 10, 12, and 14; Tables 1– 6).K �m, the parameter obtained by fitting Pc to the Hill-equation-likeformula, fails to furnish additional physical meaning or improve-ment over K�m in the correlation with Km. The existence of corre-lation between effective and true parameters (Jmax versus Vmax andK�m or K �m versus Km) is useful (Figs. 10, 12, and 14), since changesin Jmax or K�m (e.g., induction or inhibition) would commensuratewith those of Vmax or Km, rendering useful interpretation oninhibition or induction of transporter or enzyme in the Caco-2 cellsystem.

For the description of added complexities such as carrier-medi-ated transport, intracellular metabolism, and saturation of theabove processes, we propose use of the catenary model (Fig. 1) forproper data interpretation in the Caco-2 system. The presentationof mass transfer (Appendix) in this kinetic model had allowed us tosimulate data over time to conduct a thorough theoretical exami-nation on transport and metabolism. Through this exercise, weobtained analytical solutions for EfR under linear conditions andexamined the effects of transport and metabolic intrinsic clear-ances (CLd1, CLd2, CLd3, CLd4, CLint,sec, CLabs, and CLint,met).Moreover, the model yielded data on Pc, PPD, and Jc under non-linear conditions for illustration with the Eadie-Hofstee plots.Although saturation of the basolateral influx or efflux process wasnot simulated, these patterns could be inferred because the datawould be mirror images of saturable apical influx and efflux. Forthe first time, trends in EfR according to CLd and CD,0 withvarying saturable processes were shown (Fig. 18); EfR valuesgreater than unity would not conclusively reflect involvement ofapical efflux transporters since the same trends were observed withsaturation in metabolism (Fig. 18C). Although we had not exploreda wide parameter space, the analyses showed that the single barriermodel for the Caco-2 cell monolayer was inadequate to providesound estimates of Vmax and Km of saturable systems.

The usefulness of the catenary model has been demonstratedfully by the present analysis. In this model, the timed samplescollected are utilized fully for data fitting. However, due to thelarge number of parameters to be ascertained and saturation of oneor more of the processes, the fit of the timed-metabolite data may

FIG. 18. Modulation of the efflux ratio estimated at 1 h for sampling by saturableapical efflux (A), apical influx (B), and intracellular metabolism (C). The value ofEfR � 1 is denoted by the dotted line. In (A) and (B), CLint,met � 0.01 ml/min. Onlyone saturable component (Vmax � 0.5 nmol/min and Km � 0.5 �M) was present ata time. For simulation, the following assumptions were made: CLd � CLd1 �CLd2 � CLd3 � CLd4 � 0.01, 0.1, or 1 ml/min; all the unbound fractions equaledunity.


at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


not always lead to fruitful outcomes. The usual strategy is to addthe total amount of formed metabolite (amount in cell � donor �receiver compartments) to describe metabolism in the fitting pro-cedure. Additional data on metabolite transport or metabolic dataare absolutely necessary and would greatly add to the accuracy ofthe fit. The model was able to describe saturable efflux (Ito et al.,1999), saturable apical absorption, and basolateral influx (Irie etal., 2004), and recently, saturable metabolism and substrate inhi-bition (H. Sun, L. Zhang, E. C. Chow, G. Lin, K. S. Pang, andZ. Zuo, unpublished data). Proper data interpretation with thecatenary model would definitely remove the bias in parameterestimates, avoid viewing the Caco-2 cell monolayer as a singlebarrier, and provide accurate estimates.

Appendix

Mass Balance Equations in Caco-2 Cell Monolayerfor Linear Conditions

In the apical compartment:

Vap

dCap

dt� �CLabs � CLd4fapCap � �CLint,sec � CLd3fcellCcell

(15)

In the cellular compartment:

Vcell

dCcell

dt� �CLabs � CLd4fapCap � �CLd1 � CLinfluxfbasoCbaso

�CLint,sec � CLd3 � CLint,met � CLd2 � CLeffluxfcellCcell (16)

In the basolateral compartment:

Vbaso

dCbaso

dt� �CLd2 � CLeffluxfcellCcell �CLd1 � CLinfluxfbasoCbaso

(17)

Mass Balance Equations in Caco-2 Cell Monolayerfor Nonlinear Conditions

For Saturation of Apical SecretionIn the apical compartment:

Vap

dCap

dt� �CLabs � CLd4fapCap

� � Vmax

Km � fcellCcell� CLd3�fcellCcell (18)


Vcell

dCcell

dt� �CLabs � CLd4fapCap � �CLd1 � CLinfluxfbaso Cbaso

� Vmax

Km � fcellCcell� CLd3 � CLint,met � CLd2 � CLefflux�fcellCcell

(19)


Vbaso

dCbaso


(20)

For Saturation of Apical AbsorptionIn the apical compartment:

Vap

dCap

dt

� � Vmax

Km � fapCap� CLd4�fapCap � �CLint,sec � CLd3fcellCcell

(21)


Vcell

dCcell

dt

� � Vmax

Km � fapCap� CLd4�fapCap � �CLd1 � CLinfluxfbasoCbaso

�CLint,sec � CLd3 � CLint,met � CLd2 � CLeffluxfcellCcell (22)


Vbaso

dCbaso


(23)

For Saturation of Cellular Metabolism

Vap

dCap

dt� �CLabs � CLd4fapCap � �CLint,sec � CLd3fcellCcell

(24)


Vcell

dCcell

dt� �CLabs � CLd4fapCap � �CLd1 � CLinfluxfbasoCbaso

�CLint,sec � CLd3 �Vmax

Km � fcellCcell� CLd2 � CLefflux�fcellCcell

(25)


Vbaso

dCbaso


(26)

References

Adachi Y, Suzuki H, and Sugiyama Y (2001) Comparative studies on in vitro methods forevaluating in vivo function of MDR1 P-glycoprotein. Pharm Res 18:1660–1668.

Adson A, Raub TJ, Burton PS, Barsuhn CL, Hilgers AR, Audus KL, and Ho NF (1994)Quantitative approaches to delineate paracellular diffusion in cultured epithelial cell mono-layers. J Pharm Sci 83:1529–1536.

Allen JD, van Loevezijn A, Lakhai JM, van der Valk M, van Tellingen O, Reid G, Schellens JH,Koomen GJ, and Schinkel AH (2002) Potent and specific inhibition of the breast cancerresistance protein multidrug transporter in vitro and in mouse intestine by a novel analogue offumitremorgin C. Mol Cancer Ther 1:417–425.

Bentz J, Tran TT, Polli JW, Ayrton A, and Ellens H (2005) The steady-state Michaelis-Mentenanalysis of P-glycoprotein mediated transport through a confluent cell monolayer cannotpredict the correct Michaelis constant Km. Pharm Res 22:1667–1677.

Bhardwaj RK, Herrera-Ruiz D, Sinko PJ, Gudmundsson OS, and Knipp G (2005) Delineation ofhuman peptide transporter 1 (hPepT1)- mediated uptake and transport of substrates transfectedhPepT1/Madin-Darby canine kidney clones and Caco-2 cells. J Pharmacol Exp Ther 314:1093–1100.

Blais A, Bissonnette P, and Berteloot A (1987) Common characteristics for Na�-dependentsugar transport in Caco-2 cells and human fetal colon. J Membr Biol 99:113–125.

Cummins CL, Mangravite LM, and Benet LZ (2001) Characterizing the expression of CYP3A4and efflux transporters (P-gp, MRP1, and MRP2) in CYP3A4-transfected Caco-2 cells after

122 SUN AND PANG

at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


induction with sodium butyrate and the phorbol ester 12-O- tetradecanoylphorbol-13-acetate.Pharm Res 18:1102–1109.

den Ouden D, van den Heuvel M, Schoester M, van Rens G, and Sonneveld P (1996) In vitroeffect of GF120918, a novel reversal agent of multidrug resistance, on acute leukemia andmultiple myeloma cells. Leukemia 10:1930–1936.

Fisher JM, Wrighton SA, Watkins PB, Schmiedlin-Ren P, Calamia JC, Shen DD, Kunze KL,Thummel KE (1999) First-pass midazolam metabolism catalyzed by 1�,25-dihydroxy vitaminD3-modified Caco-2 cell monolayers. J Pharmacol Exp Ther 289:1134–1142.

Gekeler V, Ise W, Sanders KH, Ulrich WR, and Beck J (1995) The leukotriene LTD4 receptorantagonist MK571 specifically modulates MRP associated multidrug resistance. BiochemBiophys Res Commun 208:345–352.

Gonzalez-Alvarez I, Fernandez-Teruel C, Garrigues TM, Casabo VG, Ruiz-Garcia A, andBermejo M (2005) Kinetic modeling of passive transport and active efflux of a fluoroquino-lone across Caco-2 cells using a compartmental approach. Xenobiotica 35:1067–1088.

Guo A, Hu P, Balimane PV, Leibach FH, and Sinko PJ (1999) Interactions of a nonpeptidic drug,valacyclovir, with the human intestinal peptide transporter (hPEPT1) expressed in a mamma-lian cell line. J Pharmacol Exp Ther 289:448–454.

Hidalgo IJ, Raub TJ, and Borchardt RT (1989) Characterization of the human colon carcinomacell line (Caco-2) as a model system for intestinal epithelial permeability. Gastroenterology96:736–749.

Hidalgo IJ, Hilgreen KM, Grass GM, and Borchardt RT (1991) Characterization of the unstirredwater layer in Caco-2 cell monolayers using a novel diffusion apparatus. Pharm Res 8:222–227.

Hilgers AR, Conradi RA, and Burton PS (1990) Caco-2 cell monolayers as a model for drugtransport across the intestinal mucosa. Pharm Res 7:902–910.

Hirohashi T, Suzuki H, Chu XY, Tamai I, Tsuji A, and Sugiyama Y (2000) Function andexpression of multidrug resistance-associated protein family in human colon adenocarcinomacells (Caco-2). J Pharmacol Exp Ther 292:265–270.

Hunter J, Jepson MA, Tsuruo T, Simmons NL, and Hirst BH (1993) Functional expression ofP-glycoprotein in apical membranes of human intestinal caco-2 cells. kinetics of vinblastinesecretion and interaction with modulators. J Biol Chem 268:14991–14997.

Irie M, Terada T, Okuda M, and Inui K-I (2004) Efflux properties of basolateral peptidetransporter in human intestinal cell line Caco-2. Pflugers Arch 449:186–194.

Irvine JD, Takahashi L, Lockhart K, Cheong J, Tolan JW, Selick HE, and Grove JR (1999)MDCK (madin-darby canine kidney) cells: A tool for membrane permeability screening.J Pharm Sci 88:28–33.

Ito S, Woodland C, Sarkadi B, Hockmann G, Walker SE, and Koren G (1999) Modeling ofP-glycoprotein-involved epithelial drug transport in MDCK cells. Am J Physiol 277:F84–96.

Kansy M, Senner F, and Gubernator K (1998) Physicochemical high throughput screening:Parallel artificial membrane permeation assay in the description of passive absorption pro-cesses. J Med Chem 41:1007–1010.

Kobayashi D, Nozawa T, Imai K, Nezu J, Tsuji A, Tamai I (2003) Involvement of human organicanion transporting polypeptide OATP-B (SLC21A9) in pH-dependent transport across intes-tinal apical membrane. J Pharmacol Exp Ther 306:703–708.

Lentz KA, Polli JW, Wring SA, Humphreys JE, and Polli JE (2000) Influence of passivepermeability on effective P-glycoprotein kinetics. Pharm Res 17:1456–1460.

Okuwaki M, Takada T, Iwayanagi Y, Koh S, Kariya Y, Fujii H, Suzuki H (2007) LXR alphatransactivates mouse organic solute transporter alpha and beta via IR-1 elements shared withFXR. Pharm Res 24:390–398.

Peters WH and Roelofs HM (1989) Time-dependent activity and expression of glutathioneS-transferases in the human colon adenocarcinoma cell line caco-2. Biochem J 264:613–616.

Polli JW, Wring SA, Humphreys JE, Huang L, Morgan JB, Webster LO, and Serabjit-Singh CS(2001) Rational use of in vitro P-glycoprotein assays in drug discovery. J Pharmacol Exp Ther299:620–628.

Stenberg P, Norinder U, Luthman K, Artursson P (2001) Experimental and computationalscreening methods for the prediction of intestinal drug absorption. J Med Chem 44:1927–1937.

Stephens RH, O’Neill CA, Warhurst A, Carlson GL, Rowland M, Warhurst G (2001) Kineticprofiling of P-glycoprotein-mediated drug efflux in rat and human intestinal epithelia. J Phar-macol Exp Ther 296:584–591.

Tam D, Sun H, and Pang KS (2003) Influence of P-glycoprotein, transfer clearances, and drugbinding on intestinal metabolism in Caco-2 cell monolayers or membrane preparations: atheoretical analysis. Drug Metab Dispos 31:1214–1226.

Tamai I, Saheki A, Saitoh R, Sai Y, Yamada I, and Tsuji A (1997) Nonlinear intestinal absorptionof 5-hydroxytryptamine receptor antagonist caused by absorptive and secretory transporters.J Pharmacol Exp Ther 283:108–115.

Thiel-Demby VE, Tippin TK, Humphreys JE, Serabjit-Singh CJ, and Polli JW (2004) In vitroabsorption and secretory quotients: practical criteria derived from a study of 331 compoundsto assess for the impact of P-glycoprotein-mediated efflux on drug candidates. J Pharm Sci93:2567–2572.

Tran TT, Mittal A, Gales T, Maleeff B, Aldinger T, Polli JW, Ayrton A, Ellens H, and Bentz J(2004) Exact kinetic analysis of passive transport across a polarized confluent MDCK cellmonolayer modeled as a single barrier. J Pharm Sci 93:2108–2123.

Tran TT, Mittal A, Aldinger T, Polli JW, Ayrton A, Ellens H, and Bentz J (2005) The elementarymass action rate constants of P-gp transport for a confluent monolayer of MDCKII-hMDR1cells. Biophys J 88:715–738.

Troutman MD and Thakker DR (2003a) Rhoadamine 123 requires carrier-mediated influx for itsactivity as a P-glycoprotein substrate in Caco-2 cells. Pharm Res 20:1192–1199.

Troutman MD and Thakker DR (2003b) Efflux ratio cannot assess P–glycoprotein- mediatedattenuation of absorptive transport: asymmetric effect of P-glycoprotein on absorptive andsecretory transport across Caco-2 cell monolayers. Pharm Res 20:1200–1209.

Troutman MD and Thakker DR (2003c) Novel experimental parameters to quantify the modu-lation of absorptive and secretory transport of compounds by P-glycoprotein in cell culturemodels of intestinal epithelium. Pharm Res 20:1210–1224.

Usansky HH and Sinko PJ (2005) Estimating human drug oral absorption kinetics from Caco-2permeability using an absorption-disposition model: model development and evaluation andderivation of analytical solutions for ka and Fa. J Pharmacol Exp Ther 314:391–399.

Williams GC, Liu A, Knipp G, and Sinko PJ (2002) Direct evidence that saquinavir is transportedby multi-drug resistance-associated protein (MRP1) and canalicular multispecific organicanion transporter (MRP2). Antimicrob Agents Chemother 46:3456–3462.

Yamaguchi H, Yano I, Hashimoto Y, and Inui KI (2000) Secretory mechanisms of grepafloxacinand levofloxacin in the human intestinal cell line caco-2. J Pharmacol Exp Ther 295:360–366.

Youdim KA, Avdeef A, Abbott NJ (2003) In vitro trans-monolayer permeability calculations:often forgotten assumptions. Drug Discov Today 8:997–1003.

Address correspondence to: Dr. K. Sandy Pang, Faculty of Pharmacy, Uni-versity of Toronto, 144 College Street, Toronto, ON, Canada M5S 3M2. E-mail:[email protected]


at ASPE

T Journals on A

ugust 31, 2018dm


ownloaded from


dmd 36:102–123, 2008 printed in u.s.a. permeability...

Documents