www.elsevier.com/locate/toxinvit

Toxicology in Vitro 19 (2005) 675–684

Correlation of in vitro cytotoxicity with paracellular permeabilityin Caco-2 cells

Roula Konsoula, Frank A. Barile *

Department of Pharmaceutical Sciences, College of Pharmacy and Allied Health Professions, St. John’s University, 8000 Utopia

Parkway, Jamaica, NY 11439, USA

Received 27 January 2005; accepted 22 March 2005

Abstract

This in vitro study aims to develop a cell culture model that compares paracellular permeability (PP) with acute cytotoxicity

(AC). Caco-2 cells were seeded in 96-well plates and on polycarbonate filter inserts. Confluent monolayers were exposed to increas-

ing concentrations of 20 reference chemicals for 24-h and 72-h. Cytotoxicity was determined using MTT and NRU cell viability

assays in 96-well plates. PP was measured using transepithelial electrical resistance (TEER) measurements, as well as passage of luci-

fer yellow (LY), [3H]-mannitol (both low mw indicators), and FITC-dextran (higher mw indicator) in culture inserts. Inhibitory con-

centrations 50% (IC50s) suggest that there were good correlations between 24-h and 72-h exposures. NRU IC50 values correlated

better with TEER, which is consistent with the Registry of Cytotoxicity (RC; ICCVAM) database report. Both cell viability assays

indicate that cytotoxicity occurs before TEER is compromised. In addition, 24-h and 72-h NRU assays, and 72-h TEER measure-

ments, displayed the highest correlations with established rodent LD50s. PP experiments showed that passive paracellular transport

of the tight junction markers, especially [3H]-mannitol, correlates with the IC50s determined with the viability assays and TEER

measurements. Our AC/TEER/PP model thus allows for the differentiation between the concentrations necessary for AC and those

needed to interfere with PP. We propose that the in vitro AC, TEER and PP results be used to compute a formula which can ‘‘nor-

malize’’ and improve the predictive ability of in vitro acute cytotoxicity assays for in vivo lethality.

� 2005 Elsevier Ltd. All rights reserved.

Keywords: In vitro cytotoxicity; Paracellular permeability (PP); Caco-2; Neutral red uptake assay (NRU); MTT viability assay; Transepithelial

electrical resistance (TEER)

1. Introduction

In a recent workshop, the Interagency Coordinating

Committee on the Validation of Alternative Methods(ICCVAM) and the National Toxicology Program

(NTP) Interagency Center for the Evaluation of Alter-

native Toxicological Methods (NICEATM) forwarded

recommendations on the development of candidate pro-

cedures and their use in validation study designs (ICC-

VAM publications 01-4499 and 01-4500, 2001). In its

0887-2333/$ - see front matter � 2005 Elsevier Ltd. All rights reserved.

doi:10.1016/j.tiv.2005.03.006

* Corresponding author. Tel.: +1 718 990 2640; fax: +1 718 990 1877.

E-mail address: barilef@stjohns.edu (F.A. Barile).

report, the workshop summarized the validation status

and current potential uses of in vitro methods as predic-

tors of acute in vivo toxicity. Among the major recom-

mendations, the report concluded that in order tofurther reduce the use of animals in acute lethality as-

says, it is necessary to develop simple predictive models

for human acute cytotoxicity (AC). The group encour-

aged the optimization of simple systems that mimic gas-

trointestinal absorption, among other methods, to

improve the ability of in vitro cytotoxicity assays to pre-

dict rodent LD50 values, or any in vivo toxic effects.

Consequently, we sought to apply a cell culture model,previously used for pharmacokinetic evaluation of in

676 R. Konsoula, F.A. Barile / Toxicology in Vitro 19 (2005) 675–684

vitro absorption, as a toxicokinetic tool for predicting

acute systemic toxicity. The intention is to correlate

well-documented AC methods with paracellular perme-

ability (PP), using a reference set of chemicals.

Several investigators have recommended a strategy to

reduce the number of animals required for acute oraltoxicity testing by using in vitro cytotoxicity data to

determine the starting dose for in vivo testing (Halle

et al., 2000; Spielmann et al., 1999). Also known as

the ZEBET approach, this method is based on the stan-

dard regression between mean IC50s and corresponding

acute oral LD50 data included in the Registry of Cyto-

toxicity (RC; ICCVAM publication 01-4499, 2001).

[The ZEBET database, of which the RC is part of, pro-vides acute oral LD50 data from rats and mice and aver-

age IC50 values of chemicals and drugs from a variety of

in vitro cytotoxicity assays and cell types, based on 347

chemicals. The LD50 data is derived from the NIOSH

Registry of Toxic Effects of Chemical Substances

(RTECS)]. The calculated regression could then be used

to estimate the LD50 value of a new compound as the in

vivo starting dose of a study. Others, however, have rec-ognized that the application of this technique is limited

by the lack of information on in vitro models for gastro-

intestinal uptake, as well as blood–brain barrier passage

and biotransformation (Curren et al., 1998). Concerning

the former, monolayers of intestinal and colonic epithe-

lial cells have, in fact, been used for many years as cell

culture models for detecting transepithelial transport

of drugs, and other intestinal responses to xenobiotics(Carriere et al., 2001). The connection between the effect

of chemicals on in vitro models for PP and AC testing,

however, has not been solidified.

Artursson (1990) and Artursson and Karlsson (1991)

first described the calculation of a good correlation be-

tween oral drug absorption in humans and drug perme-

ability coefficients in Caco-2 cells. They went on to

conclude that paracellular absorption in humans canbe studied mechanistically in in vitro models of rat intes-

tinal segments and Caco-2 cells (Artursson et al., 2001).

While studies have focused on the biopharmaceutical

development of Caco-2 cell culture models as effective

drug delivery systems and high-throughput screening

methods for a variety of compounds (Biganzoli et al.,

1999; Yamashita et al., 2000, 2002), only a few have

hinted at the relationship between paracellular transportand cell viability. For instance, Duizer et al. (1998) re-

port that enhancement of transport of palmitoyl carni-

tine chloride, a commonly used absorption enhancer,

in Caco-2 cells, correlated with reduced cell viability.

Karlsson et al. (1999) confirmed these findings by sug-

gesting that a pronounced disruption of the tight junc-

tion barrier is required for efficient enhancement of

paracellular intestinal drug transport.Together with well documented in vitro AC methods,

a system for identifying the effects of chemicals on PP

could yield a precise scheme for estimating in vivo toxic

doses. The current study aims at the development of an

in vitro test system for PP, which in combination with

AC procedures, can improve the predictive ability of

in vitro acute cytotoxicity assays for in vivo lethality.

2. Materials and methods

2.1. Cell culture

Cell culture supplies were obtained from Life Tech-

nologies (Carlsbad, CA, USA) or VWR (Bridgeport,

NJ, USA). Chemicals (>99.9% purity) were obtainedfrom Sigma Chemical Co., (St. Louis, MO, USA) and

from Alfa Products (Ward Hill, MA, USA). Immortal

human colon epithelial cells (Caco-2, HTB-37, Ameri-

can Type Culture Collection, Rockville, MD, USA),

passage numbers 22-40, were subcultured and seeded

at 104 cells/cm2 in either 96-well plates or onto 12-well

plates fitted with Isopore PCF polycarbonate Millicell�

culture plate inserts (5 · 104 cells/insert). Cultures weregrown in Dulbecco�s modified Eagle�s medium supple-

mented with 10% fetal bovine serum (DMEM-10), 1%

non-essential amino acids (NEAA), 1% glutamine,

50 U/ml penicillin and 50 lg/ml streptomycin in an

atmosphere of 7.5% CO2 and 100% humidity in air.

The chemicals used in the studies were suggested by

the Registry of Cytotoxicity (RC; Halle, 2003); they

were selected based on the verification of the data set(RC-II), and for their validity in establishing a regres-

sion model between oral LD50 values and single

mammalian cell line IC50 values (ICCVAM publication

01-4500, 2001).

2.2. Assay procedures

2.2.1. MTT cell viability assay

In 96-well plates confluent monolayers of Caco-2 cells

were achieved in 7-days and were incubated with increas-

ing concentrations of each chemical for 24-h and 72-h

(chemicals are listed in Table 1). The MTT assay (Mos-

mann, 1983; Dolbeare and Vanderlaan, 1994) was mod-

ified as previously described (Barile and Cardona, 1998;

Schmidt et al., 2004). Briefly, cells are exposed to increas-

ing concentrations of the chemical (12 wells per concen-tration-group plus 1 control group) for 24-h or 72-h at

37 �C in an atmosphere of 7.5% CO2 in air. Control

groups consist of cells in media (minus chemical) which

are processed identically and incubated simultaneously

as treated groups. Incubation medium consists of

DMEM-10 supplemented as above. In the last hour of

incubation, 10-ll MTT solution (5 mg/ml in DMEM) is

added to each well. The medium is then replaced with100-ll dimethylsulfoxide (DMSO), agitated for 5 min

at 25 �C, and the absorbance is read at 550 nm on the

Table 1

IC50s (mmol/l) for Caco-2 cells using MTT and NRU assays, and TEER measurements, at 24-h and 72-h exposures

Chemical number Chemical compounds MTT 24-h MTT 72-h NRU 24-h NRU 72-h TEER 24-h TEER 72-h

1 Acrylamide 13.8 6.83 14.3 23.2 11.5 3.6

2 Actinomycin 0.028 0.009 0.007 0.0045 0.028 0.0085

3 Antipyrine 38.0 10.6 32.3 46.8 107 75.9

4 Cadmium chloride 0.12 0.04 1.0 0.16 0.05 0.03

5 Cupric sulfate 1.0 0.66 3.35 1.0 1.8 1.7

6 Dimethyl formamide 193 208 660 407 900 741

7 Doxorubicin 0.010 0.009 0.028 0.014 0.034 0.018

8 Glycerol 100 93 1098 821 430 707

9 Ibuprofen 2.2 2.3 7.2 3.1 0.075 0.038

10 Lithium sulfate 10.0 26 14.7 9.7 145 57

11 Manganese chloride 9.6 7.8 3.5 20.9 3.1 3.6

12 Niacinamide 26 23 19 13.3 189 239

13 Nickel chloride 1.78 1.74 1.07 7.4 8.4 7.8

14 q-Phenylenediamine a a 1.24 0.50 20 11.7

15 Propranolol 0.41 0.22 1.69 1.3 0.80 0.50

16 Quinine HCl 0.12 0.12 0.12 0.44 0.80 0.30

17 Salicylic acid 33.8 21.8 64 44.1 0.034 0.19

18 Sodium dichromateb 0.33 0.024 0.18 0.14 0.05 0.01

19 Trichlorforon 0.95 0.27 11.2 0.90 0.90 0.48

20 Verapamil HCl 0.19 0.13 1.8 0.87 0.58 0.60

a q-Phenylenediamine interfered with the MTT assay.b Dihydrate salt. Statistical analysis revealed that, with the exception of MTT 72-h vs. TEER 72-h (P < 0.05), none of the group comparisons were

significantly different from each other (paired students� t-test, P > 0.05; see Table 2).

R. Konsoula, F.A. Barile / Toxicology in Vitro 19 (2005) 675–684 677

BioTek FL600� fluorescence/absorbance plate reader.

Cell viability is expressed as a percentage of the control

group. The same plate contained additional wells with

media and chemical only (without cells) and processed

in parallel as reference blanks and to test for chemically

induced reduction of MTT.

2.2.2. Neutral red uptake (NRU) cell viability assay

The NRU cytotoxicity assay was performed as de-

scribed by Borenfreund and Puerner (1986). In 96-well

plates confluent monolayers of Caco-2 cells are incu-

bated with increasing concentrations of each chemical

for 24-h and 72-h as with the MTT assay. NRU is deter-

mined for each treatment concentration as follows:

monolayers are incubated with test chemical for 24-h

or 72-h, over a range of eight concentrations (includingcontrol group), at 37 �C in an atmosphere of 7.5% CO2

in air (100% humidity). At 21-h of incubation, the med-

ium is aspirated, monolayers are rinsed with 150 ll PBS,and 100 ll NR medium (1:80 dilution of 0.4% stock

solution) is added for the remaining 3-h. The medium

is aspirated, monolayers are rinsed with PBS, and

150 ll of NR desorbing fixative (ethanol/acetic acid) is

added. After shaking the cultures for 10 min, absor-bance values are read at 540 nm on the plate reader.

Absorbance values for treatment groups are compared

to that determined in control cultures. Control blanks

(medium without cells containing chemical plus NR)

are used to screen for background and to monitor pH

changes. Relative cell viability is expressed as percent

NRU of untreated control groups.

2.2.3. Transepithelial electrical resistance (TEER)

measurements

For TEER measurements, Caco-2 cells are seeded

onto 12-well plates fitted with Isopore PCF polycarbon-

ate Millicell� culture plate inserts, and cultured as

described by Biganzoli et al. (1999). DMEM-10, supple-

mented as above, is added to the apical and basolateral

chambers and replenished three times a week. Cultureswere confluent at 7-days, but maximum resistance val-

ues (at least 1000 X cm2) were reached in intact mono-

layers after 14-days. Transmembrane specific resistance

was measured using the Millicell-ERS� resistance sys-

tem (Millipore) before and after 24-h and 72-h incuba-

tion with test chemicals. As with the AC assays,

blanks (inserts without cells containing media and chem-

ical) are used to determine baseline values. Values aremeasured as X cm2, and expressed as percent TEER of

untreated control groups.

For all assays, dosage range-finding experiments were

performed. The IC50s were extrapolated from concen-

tration-effect curves using linear regression analysis.

When the IC50 was not bracketed in the initial dosage

range used for the chemical, the experiments were re-

peated and the concentrations adjusted as necessary.After the determination of the IC50, each experiment

was repeated at least two more times. Values in figures

are expressed as percentages of control groups. Also,

AC assays and PP assays were performed during the

same passage numbers to prevent further differentiation

of Caco-2 cells, thus maintaining the cultures at an early

stage of differentiation. In addition, PP experiments

678 R. Konsoula, F.A. Barile / Toxicology in Vitro 19 (2005) 675–684

were initiated when absolute TEER measurements

reached 1000 X cm2 (�2 weeks after seeding). This con-

sistent manipulation also kept the cultures as high-resis-

tant variants.

2.2.4. Paracellular permeability (PP) assays

For PP assays, Caco-2 cells were seeded onto 12-well

plates fitted with Isopore PCF polycarbonate Millicell�

culture plate inserts, as described above for TEER mea-

surements. Cultures were incubated with the test chem-

icals for 24-h and indicators were introduced in the last

90 min of incubation. Low and high molecular weight

indicators were used to determine the effect of test chem-

icals on PP. [3H]-mannitol (mw 182) has low lipophilic-ity; FITC-dextran (mw 40 K–50 K) and lucifer yellow

(LY, mw 450) are more hydrophobic; all permeate

paracellularly.

Mannitol (0.1% w/v) is dissolved in DMEM plus

HEPES, spiked with [3H]-mannitol (30 Ci/mM), and

added to the apical chamber to a final concentration

of 1 mCi/l (Liu et al., 1999). Simultaneously, cold-

DMEM (without radioactivity) is dispensed into thebasolateral chamber. At end of the incubation period,

passage of radiolabeled marker is determined by dissolv-

ing an aliquot of the basolateral medium in Budget-

solve� and measuring the cpm by liquid scintillation

(Beckman LS5801 counter). Blanks (inserts without

cells) and control cultures (minus chemical) are moni-

tored simultaneously. Samples are also taken from the

apical solutions to measure the initial donor concentra-tions. Background radioactivity is determined using

DMEM, and dpm is calculated based on the instru-

ment�s counting efficiency for [3H] (�45%).

Fluorescent marker molecules are used at concentra-

tions of 1 mg/ml in DMEM and applied to the apical

chamber. At the end of the exposure period, the basolat-

eral medium is collected and the concentration of LY

and FITC-dextran is determined by measuring fluores-cence intensity with the BioTek FL600 fluorescence/

absorbance microplate reader. Blanks (inserts without

cells) are also monitored simultaneously. The excitation

and emission wavelengths for LY are 430 nm and

540 nm, respectively, and for FITC-dextran, 487 nm

and 518 nm, respectively. As with the above assays, fluo-

rescence values for treatment groups are compared to

that determined in control cell cultures. Relative cellpermeability is expressed as a percent of untreated con-

trol groups.

2.3. Solubilization or incompatibility of chemicals

Most of the experimental problems involved the sol-

ubilization or miscibility of some of the chemicals with

the media. Solid chemicals that were insoluble in waterpresented with dissolution problems, especially at the

higher dosage levels. The solubility of these chemicals

was improved by the addition of a solubilizer to the

incubating medium as follows: salicylic acid and ibupro-

fen—1 N NaOH; propranolol, verapamil, doxorubicin,

actinomycin—1% DMSO; quinine sulfate—1% ethanol.

In separate experiments we determined that none of the

additives influenced control group cytotoxicity or para-cellular permeability of [3H]-mannitol.

q-Phenylenediamine was excluded from the MTT

assay data because of its propensity for reducing the sub-

strate and interfering with absorbance measurements.

2.4. Statistical analysis

IC50 values are extrapolated from concentration-effect curves using linear regression analysis for the aver-

age of at least three experiments. The coefficient of

determination (R2), regression analysis, slopes, inter-

cepts, and the t-statistic (two-tailed paired students� t-test with the more stringent equal variances assumption)

are calculated for all the chemicals in a particular assay

as previously described, using Microsoft Excel� (Ekwall

and Barile, 1994).In addition, the regressions calculated for the AC

studies are compared with the RC regression for the

20 chemicals tested, to contrast our model cytotoxicity

test with the RC prediction model. The RC regression

equation is based on the following formula:

logðLD50Þ ¼ 0.435� logðIC50Þ þ 0.625

where r = 0.67 for 347 chemicals in the RC database

(Halle et al., 2000) and 0.97 for 11 of the 20 chemicals

(ICCVAM publication 01-4499, 2001) used in this study.

If the regression line obtained with our model parallels

the RC regression and is within ±log5 interval, then

the test is considered suitable to generate IC50 data to

use with the RC regression for estimating starting doses.

3. Results

3.1. Comparison of cytotoxicity viability data and TEER

measurements

Table 1 shows the IC50 data generated from the AC

and TEER studies. The data summarizes the experi-ments for 24-h and 72-h. Cytotoxicity was determined

using MTT and NRU cell viability assays (AC studies)

on confluent Caco-2 cells in 96-well plates, and TEER

was measured in confluent monolayers on Millipore

polycarbonate filter membranes, as described above.

IC50s were calculated from regression analyses (not

shown). Table 2 summarizes the statistical analysis of

the data in Table 1. MTT, NRU, and TEER assays werecompared to each other at corresponding times (24-h

and 72-h). The coefficient of determination (R2) and

slope (m) of the comparisons are indicated. R2 measures

Table 2

Statistical comparison of IC50 data from Table 1 for Caco-2 cells using

MTT and NRU assays, and TEER measurements at 24-h and 72-h

exposures for the 20 reference chemicals according to regression

IC50 Data (X vs. Y) Number of chemicals R2 m

NRU 24-h vs. NRU 72-h 20 0.99 0.71

MTT 24-h vs. MTT 72-h 19 0.96 0.87

MTT 24-h vs. TEER 24-h 19 0.95 0.20

TEER 24-h vs. TEER 72-h 20 0.89 0.91

MTT 72-h vs. TEER 72-ha 19 0.84 0.20

NRU 72-h vs. TEER 72-h 20 0.82 0.80

MTT 24-h vs. NRU 24-h 19 0.69 0.13

NRU 24-h vs. TEER 24-h 20 0.62 1.00

MTT 72-h vs. NRU 72-h 19 0.53 0.18

a With the exception of MTT 72-h vs. TEER 72-h (P < 0.05), none of

the group comparisons were significantly different from each other

(paired students� t-test, P > 0.05). R2 = coefficient of determination of

regression analysis; m = slope of line of best fit.

R. Konsoula, F.A. Barile / Toxicology in Vitro 19 (2005) 675–684 679

the degree of correlation between the sets of data, while

the slope is an indication of the deviation of the plot of

experimental values from a 1:1 (mM:mM) relationship.

With the exception of MTT 72-h vs. TEER 72-h, none

of the sets of data were significantly different from each

other (P > 0.05, Tables 1 and 2).

The regressions indicate that there was a high corre-

lation between 24-h and 72-h intra-assay cytotoxicityfor NRU (R2 = 0.99), MTT (R2 = 0.96), and TEER

(R2 = 0.89; Table 2). There was also a high correlation

between the MTT 24-h and TEER 24-h (R2 = 0.95).

Although there were no significant differences among

all assays (with one exception), the flatter slopes of the

Table 3

IC50 data for in vitro assays with the highest correlations with the RC LD50s f

the same chemicals

Chemical

number

Chemical

compounds

RC database

LD50 (mmol/kg)aRC data

IC50 (mm

1 Acrylamide 2.39 1.61

2 Actinomycin 0.0057 0.0000

3 Antipyrine 9.56 11.6

4 Cadmium chloride 0.48 0.0064

5 Cupric sulfate 1.2 0.33

6 Dimethyl formamide 38.3 114

7 Doxorubicin 1.2 0.0003

8 Glycerol 137 624.0

9 Ibuprofen 4.89 0.52

10 Lithium sulfate 10.8 34

11 Manganese chloride 7.5 0.13

12 Niacinamide 28.7 44

13 Nickel chloride 0.81 0.27

14 q-Phenylenediamine 0.74 0.05

15 Propranolol 1.59 0.12

16 Quinine HCl 1.72 0.075

17 Salicylic acid 6.45 3.380

18 Sodium dichromateb 0.19 0.0009

19 Trichlorforon 1.75 0.27

20 Verapamil HCl 0.22 0.10

a Values obtained from ICCVAM guidance document Registry of Cytotoxb Dihydrate salt.

lines suggest that the 24-h assays, in general, are the

more sensitive procedures (in the calculation of slope

on log scale, when m is less than 1.0, the line is shifted

to the right and the y-values are lower than correspond-

ing x-values). High correlations were also obtained with

the 72-h MTT and NRU assays vs. 72-h TEER(R2 = 0.84 and 0.82, respectively). Both assays were also

more sensitive than the TEER measurements (m = 0.20

and 0.80, respectively). Our AC/TEER analysis thus al-

lows for the differentiation between the concentrations

necessary for AC and those needed to interfere with

TEER. These results suggest that the MTT assay is a

more sensitive indicator of chemical exposure than

NRU or TEER measurements (m = 0.20, 0.20, 0.13,0.18; Table 2). In fact, the model thus far reveals that

at equivalent IC50s, mitochondrial activity is more likely

to be altered before paracellular permeability is

compromised.

3.2. Comparison of cytotoxicity viability data and RC

database

Table 3 compares the IC50s obtained from the cyto-

toxicity assays and resistance measurements against

the LD50s for the same representative chemicals from

the RC database. Statistical analysis of the data is pre-

sented in Table 4. The regressions indicate that the RC

IC50s and both 24-h and 72-h NRU assays have the

highest correlations to the RC LD50s (R2 = 0.97, 0.92,

and 0.88, respectively). In addition, the 72-h TEER

or the 20 reference chemicals, as well as IC50s from the RC database for

base

ol/l)aNRU 72-h

IC50 (mmol/l)

NRU 24-h

IC50 (mmol/l)

TEER 72-h

IC50 (mmol/l)

23.2 14.3 3.6

081 0.0045 0.007 0.0085

46.8 32.3 75.9

0.16 1.0 0.03

1.0 3.35 1.7

407 660 741

3 0.014 0.028 0.018

821 1098 707

3.1 7.2 0.038

9.7 14.7 57

20.9 3.5 3.6

13.3 19 239

7.4 1.07 7.8

0.5 1.24 11.7

1.3 1.69 0.50

0.44 0.12 0.30

44.1 64 0.19

3 0.14 0.18 0.01

0.90 11.2 0.48

0.87 1.8 0.60

icity (RC) database, 2001.

Table 4

Statistical comparison of IC50 data from Table 3 with LD50 data from

the RC database for the 20 reference chemicals according to regression

RC LD50a data (X)

vs. IC50s (Y)

Number of

chemicals

R2 m b

RC LD50sa vs. RC IC50s

a 20 0.97 0.22 3.7

vs. NRU 72-h 20 0.92 0.15 2.3

vs. NRU 24-h 20 0.88 0.10 2.7

vs. TEER 72-h 20 0.70 0.12 2.0

vs. MTT 24-h 19 0.42 0.46 3.3

vs. TEER 24-h 20 0.37 0.09 4.8

vs. MTT 72-h 19 0.34 0.37 5.7

a From the Registry of Cytotoxicity (RC) database (ICCVAM

guidance document, 2001). None of the group comparisons were sig-

nificantly different from each other (paired students� t-test, P > 0.05).

R2 = correlation coefficient of regression analysis; m = slope of line of

best fit; b = y-intercept, used to ‘‘normalize’’ the data (see Section 4).

680 R. Konsoula, F.A. Barile / Toxicology in Vitro 19 (2005) 675–684

measurements also rank with a good correlation with

the LD50s (R2 = 0.70). The low values of the slopes indi-

cate that in all cases, LD50s are more sensitive than all

IC50 values. Only the 24-h MTT assay approaches the

sensitivity of the LD50 values (m = 0.46), although the

data does not correlate well (R2 = 0.42). It was not unex-

pected, therefore, that the 72-h NRU protocol would

demonstrate comparable correlation with RC LD50

data, as the RC IC50s, since the RC IC50s are based

on the NRU assay. In fact, our NRU IC50s in Caco-2

Fig. 1. NRU and TEER IC50s in Caco-2 cells vs. RC IC50s. Graphs of (a) 24

IC50s obtained for the same chemicals from the RC database. The regression

theoretical 1:1 correlation.

cells mirrored the RC NRU correlation performed in

BALB/c 3T3 mouse fibroblasts for 72-h (Table 4).

Fig. 1 compares IC50s derived from NRU and TEER

assays with IC50s for the same representative chemicals

from the RC database. The regression plots (a–c) show

good correlations, suggesting that AC and TEER mea-surements together are consistent with the RC IC50s.

Although the correlations of the three procedures with

the RC IC50s have high correlations and parallel the

LD50s (Table 4), the slopes of the lines are systematically

higher (the data is shifted to the left, rendering the slopes

greater than 1.0). Consequently, on a logarithmic scale,

this suggests that the IC50 concentrations for the indi-

cated assays are slightly higher than those reported inthe RC database.

3.3. Comparison of PP data and TEER measurements

Figs. 2–4 illustrate the data obtained from PP exper-

iments using passive paracellular transport markers. The

concentrations used are based on the IC50s determined

for the TEER experiments (Table 1) as well as the upperand lower limits (data not shown). Fig. 2 groups four

representative chemicals with highest transport of [3H]-

mannitol at IC50s determined for the viability assays.

This figure illustrates that paracellular transport of

[3H]-mannitol, as a tight junction marker, more closely

-h NRU, (b) 72-h NRU, and (c) 72-h TEER IC50s for Caco-2 cells vs.

s (R2) of the solid lines are indicated. The dotted line (� � �) represents a

Fig. 2. Effect of (a) acrylamide, (b) manganese chloride, (c) glycerol, and (d) q-phenylenediamine on paracellular transport in Caco-2 cells.

Concentrations correspond to IC50s generated in the TEER studies. Scale for % of control for PP markers (bars) are on left axis; for TEER

measurements (solid line), scale is on right axis. All control values are set at 100%. FITC = fluorescein isothiocyanate-dextran, LY = lucifer yellow,3H = [3H]-mannitol, TEER = transepithelial electrical resistance.

Fig. 3. Effect of (a) niacinamide, (b) doxorubicin, (c) dimethylformamide, and (d) lithium sulfate on paracellular transport in Caco-2 cells.

Concentrations correspond to IC50s generated in the TEER studies. Scale for % of control for PP markers (bars) are on left axis; for TEER

measurements (solid line), scale is on right axis. All control values are set at 100%. FITC = fluorescein isothiocyanate-dextran, LY = lucifer yellow,3H = [3H]-mannitol, TEER = transepithelial electrical resistance.

R. Konsoula, F.A. Barile / Toxicology in Vitro 19 (2005) 675–684 681

Fig. 4. Effect of (a) cadmium chloride, (b) antipyrine, (c) quinine, and (d) verapamil on paracellular transport in Caco-2 cells. Concentrations

correspond to IC50s generated in the TEER studies. Scale for % of control for PP markers (bars) are on left axis; for TEER measurements (solid line),

scale is on right axis. All control values are set at 100%. FITC = fluorescein isothiocyanate-dextran, LY = lucifer yellow, 3H = [3H]-mannitol,

TEER = transepithelial electrical resistance.

682 R. Konsoula, F.A. Barile / Toxicology in Vitro 19 (2005) 675–684

parallels the fall in TEER than LY or FITC-dextran.

Fig. 3 groups four representative chemicals that show

an increase in PP of all tight junction markers more uni-

formly as TEER decreases. Fig. 4 shows chemicals that

influence PP of FITC-dextran as well as [3H]-mannitolas TEER decreases. In every instance, transport of

[3H]-mannitol to the basolateral surface correlates better

than FITC with a decrease in resistance [Data from ini-

tial experiments using [14C]-mannitol showed no differ-

ences from [3H]-mannitol]. PP data for the rest of the

20 chemicals is not shown but mimics the graphs in

Fig. 2.

4. Discussion

Currently available AC methods are limited by the

inadequate availability of in vitro models for paracellu-

lar or transcellular transport, blood–brain barrier pas-

sage, and biotransformation. Recently, ICCVAM and

the National Toxicology Program (NTP) NICEATMconvened a workshop to address several specific objec-

tives: 1. review the status of in vitro methods for assess-

ing acute systemic toxicity; 2. recommend candidate

methods for further validation studies, and; 3. identify

reference chemicals for use in the in vitro methods. Only

a few groups have attempted to deviate from the tradi-

tional investigations of paracellular and transcellular

transport on Caco-2 cells and incorporate cytotoxicity

testing procedures. Chao et al. (1998) have shown that

long-chain acylcarnitines increase intestinal absorptionof small hydrophilic compounds through the paracellu-

lar pathway by dilating paracellular spaces in Caco-2

monolayers. By comparing the potency index of a novel

absorption enhancer with palmitoyl carnitine, Liu et al.

(1999) show that dodecylphosphocholine is significantly

safer. This index is a ratio between the IC50 value,

obtained using the MTT test, and EC50 value, the effec-

tive concentration at which TEER drops to 50%. Duizeret al. (1999) detected a decrease in TEER with 5- to 30-

lM cadmium chloride (CdCl2) exposure in Caco-2 cells,

with a corresponding disruption of the paracellular bar-

rier, yet without a significant loss of viability. Boveri

et al. (2004) report that CdCl2 toxicity can be detected

in Caco-2 cells at very low concentrations using LDH

release and HSP70 as toxicological endpoints. Okada

et al. (2000) compared changes in TEER in Caco-2 cellmonolayers with cell viability using the LDH release as-

say. They showed that a decrease in TEER, associated

with increased transepithelial permeability, preceded

LDH release. It is interesting to note that although the

authors recommend the TEER method and LDH re-

R. Konsoula, F.A. Barile / Toxicology in Vitro 19 (2005) 675–684 683

lease assay as suitable for detecting a variety of marine

toxicants, the procedure may not be useful for agents

which inhibit protein synthesis or that require incuba-

tion times beyond 3 h. Shah et al. (2004) evaluated the

cytotoxicity of class specific compounds (enzyme inhib-

itors and absorption enhancers) in Caco-2 cells using avariety of permeability and viability assays, including a

tetrazolium-based assay. Toxicity varied with time of

incubation; correlations between absorption and viabil-

ity, however, were not reported.

We obtained high regression values between AC

IC50s and TEER measurements (Table 2), indicating

that PP varies directly with AC; i.e. as AC increases,

transepithelial resistance decreases proportionately.According to the regression standards for the RC data-

base, cytotoxicity is predicted by our AC/PP model.

However, almost all of the slopes for the AC assays

vs. TEER measurements are less than 1.0, suggesting

that AC assays are more sensitive than TEER. Conse-

quently, determination of R2 is not sufficient for deter-

mining sensitivity of in vitro systems. In addition,

paracellular transport of tight junction markers, espe-cially the low-molecular weight [3H]-mannitol, corre-

lates precisely with decreases in TEER. Thus the data

suggests that interference with cell viability is more

likely to occur before PP is compromised. (It is impor-

tant to note that although some chemicals may selec-

tively interfere with the paracellular route, overall, the

experiments are not designed to address the mechanisms

of acute toxicity. Nor will the protocol differentiate be-tween passive transcellular and paracellular transport.

The study is constructed, however, to determine the

association between PP and cell viability. As a result,

calculation of a standardized formula from a larger set

of compounds would dilute any specific mechanism of

toxicity).

Comparison of the regressions between 72-h and 24-h

NRU, and 72-h TEER IC50s, with RC LD50s for thesame chemicals (Table 4), indicates that the model sys-

tem correlates well with the LD50 data. Together with

the regression analysis, computation of the slopes of

all plots further qualifies the relationships. These calcu-

lations (0.22, 0.15, 0.10, and 0.12; Table 4) suggest that

in vitro AC methods correlate well with animal LD50s.

Because the slopes are less than 1.0 however, the LD50

values (mmol/kg) are lower than the in vitro IC50s(mmol/l) (slopes less than 1.0 convey a flatter line of best

fit or a shift-to-the-right, both of which compute to

higher values of X).

PP experiments showed that passive paracellular

transport of tight junction markers correlates with the

IC50s determined with the TEER measurements. Since

the IC50s for the 24-h MTT and NRU experiments were

generally lower than the 24-h TEER measurements, andPP of tight junction markers increases systematically

with a fall in TEER, it is reasonable to conclude that cel-

lular viability is compromised before PP is affected. In

addition, as a low molecular weight substance, [3H]-

mannitol, is the most sensitive indicator for PP. Our

AC/TEER/PP model thus allows for the differentiation

between the concentrations necessary for AC and thoseneeded to interfere with PP.

Our study represents a systematic approach to corre-

lating a traditional pharmacokinetic system with in vitro

cytotoxicity testing procedures. Based on these results,

we recommend that the AC/PP model be used together

as part of a battery of in vitro tests to effectively screen

for in vivo toxicity. In fact, we propose that a ‘‘normal-

ization’’ of the results of the methods should be used forpredicting starting doses in toxicity testing. Based on the

calculations of the averages of the slopes (m) and y-

intercepts (b) for the best regressions obtained in these

studies, the following formula can be used to estimate

an LD50 or starting dose from the line of best fit:

logðmmol=kg LD50Þ ¼ 0.15½logðmM IC50Þ þ 2.7

where m = 0.15 and b = 2.7 are computed as the mean

for the first 4 assays with the highest regressions (Table

4). This standardized approach for using the AC/PP

model improves the ability to predict acute oral toxicity.

Interestingly, knowledge of a compound�s intestinal per-meability or absorption, however, may not necessarily

be useful to further classify cytotoxicity. Comparison

of our data with available permeability indexes for someof the compounds tested (Ingels et al., 2004; Ruell et al.,

2003) revealed no correlation (data not shown), suggest-

ing that permeability indexes and cytotoxicity are not

necessarily related. It is also important to note that vari-

ability in expression of transporter/efflux systems has

been shown to alter intracellular concentrations of toxic

agents, and should be taken into account in the valida-

tion of specific classes of compounds (Behrens andKissel, 2003). However, since almost all of the chemicals

in Table 1 are not effluxed compounds, P-glycoprotein

(P-gp) or non-P-gp carriers would not affect most chem-

icals. Furthermore, maintenance of consistent cell cul-

ture conditions limits the influence of carrier-mediated

transport systems in the monolayers (Behrens and Kis-

sel, 2003). Also, since our model incorporates 2 exposure

times and 3 assays, a single method is not relied upon toeffectively identify a toxic substance. Consequently the

case has been repeatedly advanced that a selective bat-

tery of validated (and normalized) tests would be to

the advantage of any in vitro toxicology program.

Acknowledgement

This work was supported in part by grants from the

NIH, NIEHS (R15 ES012170-01), USA.

684 R. Konsoula, F.A. Barile / Toxicology in Vitro 19 (2005) 675–684

References

Artursson, P., 1990. Epithelial transport of drugs in cell culture. I: A

model for studying the passive diffusion of drugs over intestinal

absorptive (Caco-2) cells. Journal of Pharmaceutical Sciences 79,

476–482.

Artursson, P., Karlsson, J., 1991. Correlation between oral drug

absorption in humans and apparent drug permeability coefficients

in human intestinal epithelial (Caco-2) cells. Biochemical Biophys-

ical Research Communications 175, 880–885.

Artursson, P., Palm, K., Luthman, K., 2001. Caco-2 monolayers in

experimental and theoretical predictions of drug transport.

Advances in Drug Delivery Reviews 46, 27–43.

Barile, F.A., Cardona, M., 1998. Acute cytotoxicity testing with

cultured human lung and dermal cells. In vitro Cell and Develop-

mental Biology 34, 631–635.

Behrens, I., Kissel, T., 2003. Do cell culture conditions influence the

carrier-mediated transport of peptides in Caco-2 monolayers?

European Journal of Pharmaceutical Sciences 19, 433–442.

Biganzoli, E., Cavenaghi, L.A., Rossi, R., Brunati, M.C., Nolli, M.L.,

1999. Use of a Caco-2 cell culture model for the characterization of

intestinal absorption of antibiotics. Il Farmaco 54, 594–599.

Borenfreund, E., Puerner, J.A.., 1986. Cytotoxicity of metals, metal–

metal and metal–chelator combinations assayed in vitro. Toxicol-

ogy 39, 121–124.

Boveri, M., Pazos, P., Gennari, A., Casado, J., Hartungh, T., Prieto,

P., 2004. Comparison of the sensitivity of different toxicological

endpoints in Caco-2 cells after cadmium chloride treatment.

Archives of Toxicology 78, 201–206.

Carriere, V., Chambaz, J., Rousset, M., 2001. Intestinal responses to

xenobiotics. Toxicology in Vitro 15, 373–378.

Chao, A.C., Taylor, M.T., Daddona, P.E., Broughall, M., Fix, J.A.,

1998. Molecular weight-dependent paracellular transport of fluo-

rescent model compounds induced by palmitoylcarnitine chloride

across the human intestinal epithelial cell line Caco-2. Journal of

Drug Target 6, 37–43.

Curren, R., Bruner, L., Goldberg, A., Walum, E., 1998. 13th meeting

of the Scientific Group on Methodologies for the safety evaluation

of chemicals (SGOMSEC): Validation and acute toxicity testing.

Environmental Health Perspectives 106, 419–425.

Dolbeare, F., Vanderlaan, M., 1994. Techniques for measuring cell

proliferation. In: Tyson, C.A., Frazier, J.M. (Eds.), In Vitro

Toxicity Indicators, Methods in Enzymology, vol. 1B. Academic

Press, San Diego, USA, pp. 178–200.

Duizer, E., Gilde, A.J., Versantvoort, C.H., Groten, J.P., 1999. Effects

of cadmium chloride on the paracellular barrier function of

intestinal epithelial cell lines. Toxicology and Applied Pharmacol-

ogy 155, 117–126.

Duizer, E., van der Wulp, C., Versantvoort, C.H., Groten, J.P., 1998.

Absorption enhancement, structural changes in tight junctions and

cytotoxicity caused by palmitoyl carnitine in Caco-2 and IEC-18

cells. Journal of Pharmacology and Experimental Therapeutics

287, 395–402.

Ekwall, B., Barile, F.A., 1994. Standardization and validation. In:

Barile, F.A. (Ed.), Introduction to in Vitro Cytotoxicology:

Mechanisms and Methods. CRC Press Inc., Boca Raton, FL,

USA, pp. 189–208.

Halle, W., 2003. The registry of cytotoxicity: Toxicity testing in cell

cultures to predict acute toxicity (LD50) and to reduce testing in

animals. Alternatives to Laboratory Animals 31, 89–198.

Halle, W., Spielmann, H., Liebsch, M., 2000. Prediction of human

lethal concentrations by cytotoxicity data from 50 MEIC chemi-

cals. ALTEX: Alternativen zu Tierexperimenten 17, 75–79.

ICCVAM (Interagency Coordinating Committee on the Validation of

Alternative Methods), 2001. Report of the International Workshop

on In Vitro Methods for Assessing Acute Systemic Toxicity. NIH

Publication 01-4499. National Institute of Environmental Health

Sciences, Research Triangle Park, NC, USA. Available from:

<http://iccvam.niehs.nih.gov/docs/guidelines/subguide.htm>

(accessed 3/21/05).

ICCVAM (Interagency Coordinating Committee on the Validation of

Alternative Methods), 2001. Guidance document on using in vitro

data to estimate in vivo starting doses for acute toxicity. NIH

Publication 01-4500. National Institute of Environmental Health

Sciences, Research Triangle Park, NC, USA. Available from:

<http://iccvam.niehs.nih.gov/docs/guidelines/subguide.htm>

(accessed 3/21/05).

Ingels, F., Beck, B., Oth, M., Augustijns, P., 2004. Effect of simulated

intestinal fluid on drug permeability estimation across Caco-2

monolayers. International Journal of Pharmaceutics 274, 221–

232.

Karlsson, J., Ungell, A., Grasjo, J., Artursson, P., 1999. Paracellular

drug transport across intestinal epithelia: influence of charge and

induced water flux. European Journal of Pharmaceutical Sciences

9, 47–56.

Liu, D.Z., LeCluyse, E.L., Thakker, D.R., 1999. Dodecylphosphoch-

oline-mediated enhancement of paracellular permeability and

cytotoxicity in Caco-2 cell monolayers. Journal of Pharmaceutical

Sciences 88, 1161–1168.

Mosmann, T., 1983. Rapid colorimetric assay for cellular growth and

survival: Application to proliferation and cytotoxicity assays.

Journal of Immunological Methods 65, 55–63.

Okada, T., Narai, A., Matsunaga, S., Fusetani, N., Shimizu, M., 2000.

Assessment of the marine toxins by monitoring the integrity of

human intestinal Caco-2 cell monolayers. Toxicology in Vitro 14,

219–226.

Ruell, J.A., Tsinman, K.L., Avdeef, A., 2003. PAMPA—a drug

absorption in vitro model-5. Unstirred water layer in iso-pH

mapping assays and pK fluxa —optimized design (pOD-PAMPA).

European Journal of Pharmaceutical Sciences 20, 393–402.

Schmidt, C.M., Cheng, C.N., Marino, A., Konsoula, R., Barile, F.A.,

2004. Hormesis effect of trace metals on cultured normal and

immortal human mammary cells. Toxicology and Industrial Health

20, 57–68.

Shah, R.B., Palamakula, A., Khan, M.A., 2004. Cytotoxicity evalu-

ation of enzyme inhibitors and absorption enhancers in Caco-2

cells for oral delivery of salmon calcitonin. Journal of Pharmaceu-

tical Sciences 93, 1070–1082.

Spielmann, H., Genschow, E., Leibsch, M., Halle, W., 1999. Deter-

mination of the starting dose for acute oral toxicity (lD50) testing

in the up-and-down procedure (UDP) from cytotoxicity data.

Alternatives to Laboratory Animals 27, 957–966.

Yamashita, S., Furubayashi, T., Kataoka, M., Sakane, T., Sezaki, H.,

Tokuda, H., 2000. Optimized conditions for prediction of intestinal

drug permeability using Caco-2 cells. European Journal of Phar-

maceutical Sciences 10, 195–204.

Yamashita, S., Konishi, K., Yamazaki, Y., Taki, Y., Sakane, T.,

Sezaki, H., Furuyama, Y., 2002. New and better protocols for a

short-term Caco-2 cell culture system. Journal of Pharmaceutical

Sciences 91, 669–679.