articles
nature publishing group
Drug–Drug Interactions Mediated Through P-Glycoprotein: Clinical Relevance and In Vitro– In Vivo Correlation Using Digoxin as a Probe Drug KS Fenner1, MD Troutman2, S Kempshall1, JA Cook3, JA Ware4,6, DA Smith1 and CA Lee5 The clinical pharmacokinetics and in vitro inhibition of digoxin were examined to predict the P-glycoprotein (P-gp) component of drug–drug interactions. Coadministered drugs (co-meds) in clinical trials (N = 123) resulted in a small, ≤100% increase in digoxin pharmacokinetics. Digoxin is likely to show the highest perturbation, via inhibition of P-gp, because of the absence of metabolic clearance. In vitro inhibitory potency data (concentration of inhibitor to inhibit 50% P-gp activity; IC50) were generated using Caco-2 cells for 19 P-gp inhibitors. Maximum steady-state inhibitor systemic concentration [I], [I]/IC50 ratios, hypothetical gut concentration ([I2], dose/250 ml), and [I2]/IC50 ratios were calculated to simulate systemic and gut-based interactions and were compared with peak plasma concentration (Cmax),i,ss/Cmax,ss and area under the curve (AUC)i/AUC ratios from the clinical trials. [I]/IC50 < 0.1 shows high false-negative rates (24% AUC, 41% Cmax); however, to a limited extent, [I2]/IC50 < 10 is predictive of negative digoxin interaction for AUC, and [I]/IC50 > 0.1 is predictive of clinical digoxin interactions (AUC and Cmax). P-glycoprotein (P-gp) was discovered in the context of multidrug-resistant tumors in which P-gp overexpression correlated with multidrug-resistant phenotype expression that allowed cells to become resistant to a wide array of chemotherapeutics.1 P-gp is able to transport a wide range of compound structures,2,3 as broad as that for cytochrome P450 3A (CYP3A), which encompasses up to 50% of currently marketed drugs.4,5 The expression of P-gp in the body is high in epithelial and endothelial barrier– forming tissues such as the gut and blood–brain barrier, and in organs of xenobiotic clearance, such as the liver and kidney,6 thus making it of significance in the distribution and elimination of drugs.4,7–9 Transport activity mediated by P-gp is saturable and is subject to a variety of interactions with substrates and inhibitors.10 Although this has been demonstrated in animal and in vitro studies, few clinically relevant drug–drug interactions (DDIs) attributable solely to P-gp have been reported. One reason P-gp has not been more widely implicated in DDIs may be that many known P-gp substrates (notably loperamide, quinidine, cyclosporine, and vinblastine) are also metabolized by cytochrome P450 (CYP). The structure–activity overlap between
P-gp and CYP, especially the CYP3A4 isoform, has been widely reported.4 Our current understanding of the respective roles that metabolism and transport may play in DDIs is limited. Further confounding this understanding is the overlap in the inhibition and induction profiles of P-gp and CYP3A4. In broad terms, inhibition of CYP metabolism for these dual substrates is often considered the major cause of many of these DDIs, particularly when victim exposure levels are highly determined by metabolism. Two P-gp substrates without the confounding influence of metabolism are digoxin and talinolol. P-gp plays a role in the absorption and elimination of both compounds in vivo, and both have shown pharmacokinetic (PK) changes upon coadministration with P-gp inhibitors.9–12 The concomitant administration of erythromycin produced small increases in both the area under the curve (AUC; 52%) and peak plasma concentration (Cmax; 26%) of talinolol, whereas several other co-meds resulted in decreased talinolol PK parameters, which were not considered clinically relevant.12,13 Digoxin has a narrow therapeutic window and even slight exposure changes have been associated with adverse reactions; this has resulted in close monitoring of digoxin serum
1Department of Pharmacokinetics, Dynamics and Metabolism, Pfizer Global Research & Development, Sandwich, UK; 2Department of Pharmacokinetics, Dynamics
and Metabolism, Pfizer Global Research and Development, Groton/New London, Connecticut, USA; 3Department of Clinical Pharmacology, Pfizer Global Research and Development, Groton/New London, Connecticut, USA; 4Department of Pharmacokinetics, Dynamics and Metabolism, Pfizer Global Research and Development, Ann Arbor, Michigan, USA; 5Department of Pharmacokinetics, Dynamics and Metabolism, Pfizer Global Research & Development, La Jolla, California, USA; 6Current address: Genentech, Inc., 1-DNA Way MS-70, South San Francisco, California 94080, USA. Correspondence: CA Lee (
[email protected]) Received 17 April 2008; accepted 28 August 2008; advance online publication 5 November 2008. doi: 10.1038/clpt.2008.195 Clinical pharmacology & Therapeutics | VOLUME 85 NUMBER 2 | FEBRUARY 2009
173
articles levels.14–16 As a consequence of the relative safety concerns for each compound, far more digoxin DDI reports are available. Thus the remainder of this article focuses on digoxin for the following reasons: (i) overall relevance because of safety concerns, (ii) digoxin provides an unambiguous determination of P-gp-mediated DDI, and (iii) it is the recommended probe in the US Food and Drug Administration (FDA, http://www.fda.gov/cder/guidance/ index.htm) draft guidance for assessing P-gp inhibition. The FDA draft guidance document mentioned earlier provides recommendations for the in vitro evaluation of drug interaction potential for P-gp substrates and/or inhibitors to facilitate decisions regarding the need to conduct clinical trials. Regarding P-gp inhibition, the guidance states that a clinical DDI trial with digoxin should be performed if the systemic inhibitor concentration [I] to inhibitory potency measure (Ki or concentration of inhibitor to inhibit 50% P-gp activity; IC50) ratio is >0.1. It appears that the FDA guidance has applied the ratio introduced by Rowland et al.17 and Williams et al.,18 who showed that for a single route of clearance, the increase in the AUC in the presence of an inhibitor can be represented as 1 + [I]/Ki. When [I]/Ki is 0.1, the predicted increase in AUC ratio will be 10%. An alternative ratio using inhibitor gut concentration [I2], [I2]/IC50, aims to predict interactions that occur during absorption, but has been utilized less extensively. The goal of this work was to provide further insight as to when P-gp-related DDIs may be clinically relevant. We have analyzed a large number of digoxin DDI studies to better understand the incidence and magnitude of the event. Key in vitro IC50 data have been produced to allow an understanding of the relationship between [I], [I2], and the clinical effect. On the basis of these results, we make recommendations on in vitro–derived ratios for guidance and provide insight for future studies. Results
3.5 AUC Cmax
3 2.5 2 1.5
in
yc
Itr
da
id
C
po
ui n
Q
ls Va
e lo sp ac orin o C la naz rit hr ole om y R an cin ol az in C e ar ve d Ve ilo l ra Am pa m io i da l ro D ne ilt i a C on zen iv ap ta C ap n to p M ib ril Pr efr op ad il af en o C im ne et id in N ife e di pi R ne ito Te na lm vir is ar ta Ta n lin Fe olo l lo di At p or ine va s N itr tati en n d i O m pin e ep ra zo Is l ra e di p Se ine rtr a N ic line ar di pi Lo ne s a Tr og rtan lit az Va on e re ni cl in e
1 r
AUCi/AUC or Cmax,i/Cmax digoxin ratios
Analysis of 123 study reports that examined digoxin pharmacokinetics in the presence and absence of co-meds (many of which were known to be P-gp substrates or inhibitors) showed only small changes in PK parameters (a subset of the 123 study reports is depicted in Figure 1; references for Figure 1 are found in Table 1). In general, coadministration of potential
P-gp inhibitors did not increase digoxin AUC or Cmax ratios by >100%, with the vast majority (93 of 123, 76%) of the reported interactions exhibiting 25% increase or a ratio of ≥1.25. For some precipitants, digoxin data from more than one clinical trial were used in the IVIVC analyses. Figure 2 presents data from Table 3 in graphical form, where the solid horizontal line indicates a 25% change (ratio of 1.25) in digoxin AUC (Figure 2a) or Cmax,ss (Figure 2b) ratio and the dashed vertical line denotes the [I]/IC50 of 0.1. For [I]/ IC50 > 0.1, seven of nine inhibitors showed a correlation based on AUC ratios, and seven of eight inhibitors were consistent with a clinically relevant change based on Cmax,ss ratios. For [I]/ IC50 < 0.1, 4 of 17 inhibitors showed a clinically relevant digoxin AUC ratio (a 24% false-negative rate), whereas the false-negative rate increased to 41% for the digoxin Cmax,ss ratio, where 7 of 17 inhibitors revealed clinically relevant digoxin change. Similar to Figure 2, Figure 3 shows [I2]/IC50 data from Table 3; again the solid horizontal line indicates the 25% increase
(ratio of 1.25) in digoxin AUC (Figure 3a) or Cmax,ss (Figure 3b) ratios, and the dashed vertical line denotes the [I2]/IC50 of 10. For [I2]/IC50 > 10, 10 of 17 inhibitors (59%) were consistent with a clinically relevant change based on AUC ratios, whereas a higher correspondence was observed for Cmax,ss ratios (11 of 16, 69%). For [I2]/IC50 < 10, good IVIVC was observed, as seven of nine inhibitors (78%) showed no clinically relevant digoxin AUC ratio changes, whereas agreement diminished slightly for Cmax,ss ratios (six of nine inhibitors). Table 4 lists the drugs that show false-negatives and false-positives based on the proposed cutoff ratios shown in Figures 2 and 3. The [I]/IC50 values presented in Table 3 suggest that P-gp inhibition is most pronounced in the gut, as the vast majority of values are 300
0.03
1 × 10−3
22.14
0.1
70
The PK ratios were determined using the arithmetric mean values. For same precipitant from same reference, digoxin PK data reported for different doses of precipitant unless indicated by a footnote. AUC, area under the curve; Cmax, peak plasma concentration; IC50, concentration of inhibitor to inhibit 50% P-gp activity. Inhibitor concentration values were obtained from the cited reference or from a reference source with levels adjusted to dose assuming dose linearity (ref. 71). aDigoxin PK parameters determined in male subjects. bIC50 values were not obtained, but maximum tested concentration was used as IC50 to allow for determinations of [I]/IC50 and [I2]/IC50. cDigoxin PK parameters determined in female subjects.
administered intravenously and orally as Lanoxin. The effect of a given inhibitor on the digoxin AUC ratio is less following intravenous (IV) vs. oral administration of Lanoxin, indicating that fewer P-gp-mediated drug interactions are observed following IV administration. Discussion
Digoxin is devoid of CYP metabolism, and changes to digoxin absorption and elimination can be largely attributed to P-gp efflux activity in the gut and kidney, respectively.11,23–25 The narrow therapeutic window of digoxin requires close monitoring of patients for untoward signs of digitalis-mediated toxicity, and numerous co-meds, including well-characterized P-gp inhibitors such as quinidine, amiodarone, and verapamil, have been 176
reported to increase digoxin levels to exceed its upper limit of safety.21,22,26–28 In addition, digoxin DDI trials are routinely performed for new medications, leading to a large body of available study data. For these reasons, digoxin is considered the gold standard and most relevant probe for studying clinical P-gp–related DDIs, and we have therefore focused this work on digoxin. We have collected and analyzed 123 reported digoxin DDI studies to understand in vivo outcomes mediated by a variety of co-meds, many of which inhibit or are substrates for P-gp in vitro. Digoxin AUC or Cmax ratios were 2. Based on the modest change in digoxin pharmacokinetics in the presence VOLUME 85 NUMBER 2 | FEBRUARY 2009 | www.nature.com/cpt
articles a
a 3.00
2.50 AUCi/AUC ratio
2.50 AUCi/AUC ratio
3.00
2.00
2.00
1.50
1.50
1.00 0.00
0.20
0.40
1.00 0.01
0.60
0.10
10.00 1.00 [l 2]/IC50
100.00
0.10
10.00 1.00 [l 2]/IC50
100.00 1,000.00
3.00
2.50
2.00
1.50
1.00 0.00
0.20
0.40
b
3.00
Cmax,i,ss/Cmax,ss ratio
b Cmax,i,ss/Cmax,ss ratio
[l ]/IC50
2.50
2.00
1.50
1.00 0.01
0.60
[l ]/IC50
Figure 2 (a) In vivo AUCi/AUC ratio or (b) Cmax,i,ss/Cmax,ss ratio vs. [I]/IC50. Table 1 lists reference numbers for drug–drug interaction data shown in this figure. AUC, area under the curve; Cmax, peak plasma concentration; IC50, concentration of inhibitor to inhibit 50% P-gp activity.
1,000.00
Figure 3 Relationship between (a) in vivo AUCi/AUC or (b) Cmax,i,ss/Cmax,ss ratios vs. [I2]/IC50. Table 1 lists reference numbers for drug–drug interaction data shown in this figure. AUC, area under the curve; Cmax, peak plasma concentration; IC50, concentration of inhibitor to inhibit 50% P-gp activity.
Table 4 List of drugs showing false negative and false positive digoxin interactionsa False negatives [I]/IC50 < 0.1
False positives [I2]/IC50 < 10
[I]/IC50 > 0.1
[I2]/IC50 > 10
AUC ratio
Cmax ratio
AUC ratio
Cmax ratio
AUC ratio
Cmax ratio
AUC ratio
Cmax ratio
Carvedilolb
Carvedilolb
Conivaptan
Conivaptan
Telmisartan
Troglitazone
Telmisartan
Troglitazone
Conivaptan
Conivaptan
Captopril
Captopril
Troglitazone
Troglitazone
Nicardipine
Diltiazemc
Diltiazemc
Nicardipine
Sertraline
Captopril
Captopril
Sertraline
Diltiazem
Isradipine
Diltiazem
Carvedilolb
Felodipine
Carvedilolb
Nitrendipine
Isradipine
Felodipine
Nitrendipine AUC, area under the curve; Cmax, peak plasma concentration; IC50, concentration of inhibitor to inhibit 50% P-gp activity. aFalse positive and false negative assessments based on data presented in Table 3. bFalse negative for [I]/IC < 0.1 was observed and false positive for [I ]/IC was observed. 50 2 50 cFalse negative—ref. 53.
of various co-meds, many of which were potent P-gp inhibitors and expected to elicit a PK change, our analysis supports that DDIs related to changes in P-gp activity are generally not clinically significant (assuming that PK ratios ≤2 do not require dose adjustments). These findings with P-gp-mediated DDIs are in stark contrast to CYP-mediated PK ratios, which are often in the range of 3–5 (refs. 29,30) and as high as 60 (observed for saquinavir following coadministration with ritonavir31). For loperamide and quinidine, Cmax and AUC ratios >2 due to DDIs have been observed; however, these increases are likely attributable to inhibition of both P-gp and CYP3A4.32 In cases such as
these, CYP-mediated metabolism can directly influence a compound’s first-pass metabolism and elimination, which translates into a direct and proportional change in exposure. Conversely, P-gp activity most often influences rather than determines a compound’s disposition profile (e.g., by slowing absorption rather than attenuating it, although this may enhance first-pass metabolism), which is less influential on exposure levels. These differences may lead to vastly different consequences of CYPmediated vs. P-gp-mediated processes. Although it can be concluded that in general P-gp-mediated DDIs are not clinically relevant, P-gp–mediated DDI concerns
Clinical pharmacology & Therapeutics | VOLUME 85 NUMBER 2 | FEBRUARY 2009
177
articles do apply to any drugs such as digoxin that have a narrow therapeutic window for which P-gp plays a key role in determining disposition. For digoxin, a 25% increase in exposure is viewed as clinically relevant because untoward toxicity may occur as a result of increased drug levels.14–16 Given the ramifications of changing digoxin exposure, it is important to elucidate whether digoxin DDI events occur on the level of the intestine, kidney, or both. Indeed, several P-gp inhibitors have been shown to modulate digoxin exposure by improving absorption27,33 and/or decreasing renal elimination.34,35 Renal function plays a major role in determining digoxin dosing in patients36, and secretion plays a role in digoxin CLrenal.34 Evaluation of 25 digoxin DDI studies (Table 3) shows that the P-gp–related changes in digoxin steady-state kinetics reflected in Cmax and AUC ratios were highly correlated. Fewer compounds showed a clinically relevant change (>25%) with CLrenal ratio than Cmax ratio, as shown in Table 2. For a few compounds, P-gp interactions at the level of the kidney and the intestine were observed, but the interactions at the level of the kidney are more difficult to attribute to P-gp. This point is illustrated with amiodarone, with steady-state Cmax ratios of 1.84 and 1.72, whereas the CLrenal ratios were near unity.37 In the case Table 5 Comparison of intravenous (IV) digoxin pharmacokinetic (PK) parameter ratios in presence of P-glycoprotein inhibitor and digoxin alone IV digoxin dosing—mimics Lanoxicap (solution digoxin in capsule)
Precipitant
Inhibitor concentration AUCi/AUC In vitro IC50 (μmol/l) ratio (μmol/l) [I]/IC50 Reference
Carvedilol
0.52
0.98
Clarithromycin
1.54a
4.00
0.13
48
Losartan
0.64
1.19
66.00
0.02
50
1.01
144.00
0.004
57
Propafenone
0.28a
1.14
95.00
0.00
64
Propafenone
1.30
1.29
95.00
0.01
64
The PK ratios were determined using the arithmetric mean values. AUC, area under the curve; Cmax, peak plasma concentration; IC50, concentration of inhibitor to inhibit 50% P-gp activity. Inhibitor concentration values were obtained from the cited reference or from a reference source with levels adjusted to dose assuming dose linearity (ref. 71).
of ranolazine, coadministration resulted in a 46% decrease in digoxin CLrenal ratio, whereas reported circulating plasma levels of 8.4 μmol/l are much lower than the IC50 value of 49 μmol/l needed to inhibit P-gp.38 For valspodar, a 75% decrease in digoxin CLrenal ratio was observed, although little unchanged valspodar appears in urine.39 However, valspodar metabolites are excreted by the kidney and could contribute to the decrease in digoxin CLrenal; therefore, circulating metabolites could be one possible explanation. Another explanation could involve the recently identified OATP4C1 uptake transporter located on the basolateral membrane of the proximal tubules of the kidney.40 In vitro studies indicate that digoxin is a substrate for this transporter (Km 7.8 μmol/l), and inhibition of digoxin uptake into the proximal tubule by OATP4C1 could further contribute to a decrease in digoxin CLrenal. Although there are examples of changes in digoxin CLrenal following co-med administration, the degree of the changes is not always as expected (increase, decrease, or no change), possibly due to a multiplicity of unique mechanisms, as highlighted earlier. Given that the Cmax ratio of digoxin is straightforward and provides the most sensitivity, we conclude that it is the most appropriate parameter for assessing clinical digoxin DDI potential, and that intestinal rather than kidney P-gp plays a greater role toward P-gp–related DDIs. The FDA recently released a draft guidance document suggesting that [I]/IC50 > 0.1 (with IC50 derived from in vitro studies) would trigger running a clinical digoxin DDI study, presumably aimed at profiling a new drug candidate’s ability to inhibit P-gp clinically. It has also been recently proposed that [I2]/IC50 > 10 may also be used as a metric for establishing digoxin DDI IVIVC.41 To investigate the robustness of the [I]/IC50 approach with associated cutoff values, IC50 values for 19 P-gp inhibitors were generated; these span a reasonable range of potency, [I], and dose. A high concordance was observed when comparing [I]/IC50 > 0.1 with clinically significant digoxin AUC and Cmax,ss ratio changes (Figure 2, Table 3). However, a 41% false-negative rate (7 of 17) is evident when [I]/IC50 < 0.1, indicating that the in vitro data underpredict the in vivo digoxin P-gp–mediated DDIs. The high false-negative percentage raises concern for the utility of [I]/IC50 < 0.1, as it does not appear to provide adequate guidance or confidence that digoxin DDIs will not occur. A further complication in using a “global” cutoff limit could be the interlaboratory
Table 6 Comparison of oral digoxin pharmacokinetic (PK) parameter ratios in presence of P-glycoprotein inhibitor and digoxin alone Oral digoxin dosing—Lanoxin tablets Precipitant
Cmax,ss,i/Cmax,ss Inhibitor Hypothetical intestinal ratio concentration (μmol/l) concentration (μmol/l) AUCi/AUC ratio
In vitro IC50 (μmol/l)
[I]/IC50
[I2]/IC50
Reference
Carvedilol
0.13
61.50
1.56
1.38
4.00
0.03
15.38
48
Carvedilol
0.13
61.50
1.24
1.00
4.00
0.03
15.38
48
Carvedilol
0.52
246.00
1.20
1.60
4.00
0.13
61.50
47
Clarithromycin
1.54
1,337.00
1.64
1.83
66.00
0.02
20.26
50
1.05
1.01
144.00
0.00
3.01
57
—
1.33
95.00
0.01
18.50
63
Losartan
0.64
433.80
Propafenone
0.65
1,757.21
The PK ratios were determined using the arithmetric mean values. AUC, area under the curve; Cmax, peak plasma concentration; IC50, concentration of inhibitor to inhibit 50% P-gp activity. Inhibitor concentration values were obtained from the cited reference or from a reference source with levels adjusted to dose assuming dose linearity (ref. 71). 178
VOLUME 85 NUMBER 2 | FEBRUARY 2009 | www.nature.com/cpt
articles variation in any values derived from biological assays even when similar protocols are used.19–22 Looking at the list of compounds with [I]/IC50 < 0.1, other properties such as dose and circulating levels should also be considered to lessen the risk of potential digoxin interactions. Moreover, it is also important to recognize that the pharmacology of many drugs given as co-meds with digoxin may result in the direct alteration of cardiac output and/or kidney function independent of whether the compound interacts with P-gp (as observed with captopril). In considering the theoretical intestinal concentration [I2], a strong in vitro [I2]/IC50 and in vivo AUC or Cmax,ss ratio demarcation is not apparent (Figure 3). Following the proposed guidance of [I2]/IC50 > 10, only 10 of 17 inhibitors showed a clinically relevant change based on digoxin AUC ratio, and similar concordance (11 of 16) was observed using digoxin Cmax,ss ratios. It should be noted that some of the compounds that did not correlate (false-positives), such as troglitazone, nicardipine, and isradipine, may act as inhibitors as well as inducers of P-gp, resulting in digoxin PK ratios 0.1 is predictive of positive clinical digoxin DDIs related to P-gp and that, to a limited extent, [I2]/ IC50 < 10 is predictive of negative clinical digoxin DDIs related
to P-gp. However, the number of false-negatives observed with [I]/IC50, and false-positives with [I2]/IC50, is of concern. We conclude that there is a great need for more investigative analysis concerning both digoxin CLrenal and the role of other transporters and experimental protocols used in the various preclinical P-gp assays to generate IC50 values for establishing better IVIVC around digoxin DDIs. Methods [3H]-digoxin was obtained from NEN (Hounslow, UK). Mibefradil, amiodarone, quinidine, diltiazem, captopril, nicardipine, verapamil, cimetidine, isradipine, felodipine, clarithromycin, propafenone, ranolazine, sertraline, digoxin, and Lucifer yellow were obtained from Sigma-Aldrich (Poole, UK). Talinolol, losartan, omeprazole, telmisartan, carvedilol, conivaptan, nifedipine, nitrendipine, troglitazone, atorvastatin, varenicline, and paroxetine were obtained from Pfizer Global Research and Development (Sandwich, UK, and Milwaukee, WI). Caco-2 cells were obtained from American Type Culture Collection (Rockville, MD). Cell culture media, supplements, and buffers were purchased from Invitrogen (Paisley, UK). Caco-2 cell culture. Caco-2 cells were cultured at 37 °C in an atmosphere
of 95% humidity and 5% CO2 in minimum essential medium supplemented with 20% fetal calf serum, 1% nonessential amino acid, 2 mmol/l L-glutamine, and 2 mmol/l sodium pyruvate. Cells were maintained in flasks and subcultured weekly when they reached 70–80% confluency. Cells were used in studies up to 20 passages from the original passage number. They were seeded onto 24-well HTS multiwell membrane inserts (Becton Dickinson, Cowley, UK), 1-µm pore size at 1.2 × 105 cells/cm2. The medium was replaced three times per week, and cells were used 21–25 days after seeding. Before experiments, transepithelial electrical resistance values were measured. Values >3,000 Ω/cm2 were considered confluent, and these cells were used in experiments. Inhibition assays. Inhibition of [3H]-digoxin (5 µmol/l) efflux across Caco-2 cell monolayers was determined in the absence and presence of increasing concentrations of potential inhibitor in duplicate wells. Inhibitor was added to both apical and basolateral chambers along with the addition of the substrate to the donor chamber. After a 2-h incubation at 37 °C and 5% CO2 with 95% humidity, samples from both donor and acceptor were removed and added to liquid scintillant for counting. Net secretory flux (NSF) was calculated by subtracting the average absorptive flux from the average secretory flux of [3H]-digoxin at each inhibitor concentration. The percentage inhibition was then calculated using the following equation:
NSFi % Inhibition = 1 − × 100 NSFa where NSFi is the NSF in the presence of an inhibitor, and NSFa is the NSF in the absence of an inhibitor. This is further derived from the equation published by Choo et al.44 The integrity of the Caco-2 cell monolayer was confirmed at the end of the experiment by a further incubation for 1 h with Lucifer yellow (100 μmol/l) at 37 °C. The amount of Lucifer yellow in the receiver wells was determined using a fluorescent plate reader (Victor 2; Wallac) with λ excitation at 485 nm and λ emission at 535 nm, and apparent permeability values were determined using the equation: Papp =
δQ 1 1 × × δt Co SA
where δQ is the amount in the receiver chamber, δt is the length of the incubation with Lucifer yellow (3,600 s), Co is the starting concentration of Lucifer yellow (100 μmol/l), and SA is the surface area of the monolayer (0.33 cm2). Monolayers with Papp values
