Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

P-glycoprotein (Pgp) is a member of the ATP-binding cassette superfamily of membrane transport proteins, responsible for the efflux of many drugs. It represents a major component of the blood-brain barrier (Schinkel et al., 1994) and the intestinal barrier (van Asperen et al., 1998), and it contributes to renal and biliary elimination of drugs (Kusuhara et al., 1998; Chiou et al., 2000). At the blood-brain barrier Pgp is localized in the apical membrane of brain capillary endothelial cells and transports substrates toward the blood compartment (Cordon-Cardo et al., 1989; van Asperen et al., 1997). Therefore, Pgp can limit the penetration into and retention within the brain and thus modulate effectiveness and central nervous system toxicity of numerous compounds. In contrast, the absence of active Pgp as observed in mdr-1 knockout mice lacking Pgp and thus exhibiting unrestricted access of Pgp substrates to the brain yields significantly increased central nervous system concentrations often exceeding those observed in wild-type mice by orders of magnitude (Schinkel et al., 1994, 1996). Pgp is also highly expressed in the apical membrane of epithelial cells in the small and large intestine, where it transports drugs out of the cells into the intestinal lumen (Cordon-Cardo et al., 1989; van Asperen et al., 1998), thus limiting bioavailability of compounds such as paclitaxel and human immunodeficiency virus protease inhibitors (Sparreboom et al., 1997; Kim et al., 1998).

Established antidepressants, in particular tricyclic antidepressants, have a significant potential of inducing adverse drug reactions, exhibit a limited effectiveness even in patients with optimum compliance (Bollini et al., 1999), and are subject to numerous drug-drug interactions (Stockley, 1999). Newer antidepressants, which have been developed and introduced to supplement the established antidepressants, include the selective inhibitors of serotonin reuptake (SSRIs, e.g., sertraline) or norepinephrine reuptake (e.g., reboxetine) or compounds inhibiting reuptake of both neurotransmitters (e.g., venlafaxine). They are almost completely biotransformed before excretion (Caccia, 1998) and some of the numerous known metabolites (e.g., norfluoxetine, N- and O-desmethylvenlafaxine) are active with significant antidepressant effects (Preskorn, 1997; Caccia, 1998).

Because it is impossible to perform drug-drug interaction studies with all combinations prescribed to patients, it is important to elucidate possible mechanisms of interaction in vitro to guide initial dosage adaptations and/or to prompt focused interaction studies in vivo and efficient monitoring of adverse drug reactions in patients.

The role of Pgp in causing clinically relevant drug interactions is becoming more and more obvious (Yu, 1999). Concentrations of paroxetine and venlafaxine increased significantly (1.7- to 3-fold) in the brain of mdr1ab (/) knockout mice after single dose administration and after treatment for 11 days (Uhr, 2002), suggesting that these antidepressants are Pgp substrates and that their pharmacokinetics might be influenced by coadministered Pgp inhibitors. In contrast, for fluoxetine the absence of Pgp substrate characteristics was reported (Uhr et al., 2000), indicating that not all class members share these properties. For citalopram, the results are contradictory (Rochat et al., 1999; Uhr et al., 2000). Hitherto, nothing is known about the inhibitory potential of the newer antidepressants on Pgp and thus their potential to modulate the access of Pgp substrates into the brain.

We therefore characterized the inhibitory potencies of the newer antidepressants and some of their available main metabolites in an in vitro assay by using calcein-acetoxymethylester (calcein-AM) as Pgp substrate. Two different cell systems expressing Pgp were used and compared concerning their suitability for a Pgp inhibition assay: L-MDR1 (porcine kidney cells overexpressing the human isoform of the transporter Pgp) (Schinkel et al., 1995) and primary cell cultures of porcine brain capillary endothelial cells (pBCECs), expressing porcine pgp1A (Török et al., 1999), as a model for the blood-brain barrier.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Materials. Culture media, fetal calf serum, medium supplements, antibiotics, and Hanks' balanced salt solution were purchased from Invitrogen (Karlsruhe, Germany); collagenase/dispase and dispase were from Roche Diagnostics (Mannheim, Germany); collagen-R was from Serva (Heidelberg, Germany); DMSO and Triton X-100 were from AppliChem (Darmstadt, Germany); dextran was from Sigma-Aldrich (Taufkirchen, Germany); Percoll was from Amersham Biosciences, Inc. (Freiburg, Germany); calcein-AM was from MoBiTec (Göttingen, Germany); vincristine was from Calbiochem (Darmstadt, Germany); and 96-well microtiter plates were from NUNC GmbH & Co. KG (Wiesbaden, Germany).

Drugs. Citalopram hydrobromide and desmethylcitalopram hydrochloride were kind gifts from Lundbeck (Valby, Denmark); fluvoxamine maleate was from Solvay (Hannover, Germany); LY335979 was obtained from Eli Lilly & Co. (Bad Homburg, Germany); paroxetine hydrochloride hemihydrate and paroxetine metabolite were from GlaxoSmithKline (Stevenage, UK), SDZ-PSC833 was from Novartis (Basel, Switzerland); reboxetine methanesulfonate was from Pharmacia (Kalamazoo, MI); sertraline hydrochloride and desmethylsertraline from were Pfizer (Karlsruhe, Germany); and venlafaxine hydrochloride, N-desmethylvenlafaxine hydrochloride, and O-desmethylvenlafaxine were from Wyeth (Münster, Germany). Fluoxetine hydrochloride, norfluoxetine hydrochloride, and verapamil hydrochloride were purchased from Sigma-Aldrich and quinidine was from Roth (Karlsruhe, Germany).

LLC-PK1 and L-MDR1 Cells. As model for human Pgp we used L-MDR1 cells, a cell line generated by transfection of the porcine kidney epithelial cell line LLC-PK1 with the human MDR1 gene (Schinkel et al., 1996) and the parental cell line LLC-PK1 (American Type Culture Collection, Manassas, VA) as a control. The L-MDR1 cell line was kindly provided by Dr. A. H. Schinkel (The Netherlands Cancer Institute, Division of Experimental Therapy, Amsterdam, The Netherlands). The cells were cultured under standard cell culture conditions with medium M199 supplemented with 10% heat inactivated fetal calf serum, 2 mM glutamine, 100 U/ml penicillin, and 100 µg/ml streptomycin sulfate. To maintain Pgp expression, the culture medium for L-MDR1 was supplemented with 0.64 µM vincristine. For the calcein assay, cells were seeded on collagen-coated microtiter plates in a density of 10,000 cells/cm² and cultured for 3 days. One day before the assay, both cell lines were fed with vincristine-free culture medium.

Isolation of Porcine Brain Capillary Endothelial Cells. The isolation of pBCECs was essentially based on the method described by Audus et al. (1996). Medium M199 supplemented with L-glutamine (0.7 mM), streptomycin sulfate (100 µg/ml), penicillin G (100 U/ml), gentamycin (100 µg/ml), and HEPES (10 mM) was used for all preparation steps. Briefly, cortical gray matter from seven to eight fresh porcine brains, which were obtained from a local slaughterhouse, was isolated, cut into very small pieces, and digested enzymatically using 0.5% dispase (in preparation medium). After centrifugation (1,000g, 4°C, 10 min), the pellets were resuspended in 13% dextran solution in preparation medium and centrifuged again (5800g, 4°C, 12 min). The pellet containing the cerebral microvessels was subsequently incubated in preparation medium containing 0.1% (w/v) collagenase/dispase. The resulting cell suspension was filtered, and the brain capillary endothelial cells were separated on a discontinuous Percoll gradient (densities 1.03 and 1.07, centrifugation at 1,250g, 5 min, 20°C), washed, and filtered again before being seeded on collagen-coated microtiter plates in a density of 100,000 cells/cm². Cells were cultured under standard cell culture conditions with medium M199 containing L-glutamine (0.7 mM), streptomycin sulfate (100 µg/ml), penicillin G (100 U/ml), HEPES (10 mM), and 10% heat-inactivated horse serum. Eight days after seeding the confluent monolayers were used for the calcein assay.

Reverse Transcription-Polymerase Chain Reaction (RT-PCR) for the Detection of Pgp Expression at the mRNA Level. Expression of human Pgp and/or porcine pgp1A (ppgp1A) at the mRNA level in the cell lines used was verified by reverse transcription of RNA followed by polymerase chain reaction. RNA was isolated using the RNeasy mini kit (QIAGEN, Hilden, Germany). First-strand cDNA synthesis was performed with the First-Strand cDNA synthesis kit for RT-PCR (Roche Diagnostics) with random hexamer primers according to the manufacturer's instructions.

Stock Solutions. Stock solutions of test compounds were prepared strictly following the manufacturers' instructions. Most compounds were soluble in aqua bidest. Only sertraline, desmethylsertraline, O-desmethylvenlafaxine, quinidine, verapamil hydrochloride, SDZ-PSC833, and LY335979 were dissolved in DMSO. The DMSO concentration in the assays never exceeded 1% (v/v), a concentration that was found not to influence the results of the assay in pilot experiments.

Calcein Uptake Assay. Calcein-AM is a fluorogenic, highly lipid-soluble dye that rapidly penetrates the plasma membrane. Inside the cell, endogenous esterases cleave the ester bonds, producing the hydrophilic and fluorescent dye calcein, which cannot leave the cell via the plasma membrane (Hollo et al., 1996). Whereas calcein-AM is a substrate of Pgp, calcein is not (Homolya et al., 1993; Hollo et al., 1996). Cells expressing high levels of Pgp rapidly extrude nonfluorescent calcein-AM from the plasma membrane, thus preventing accumulation of fluorescent calcein in the cytosol. Because the transport capacity of Pgp is inversely proportional to the accumulation of intracellular calcein fluorescence, inhibition of Pgp will lead to intracellular calcein accumulation.

Quenching Test. Each test compound was analyzed for possible quenching effects on the calcein fluorescence. Because calcein-AM is nonfluorescent and cannot be used for a quenching test, we generated the fluorescent dye calcein by incubating LLC-PK1 cells with 1 µM calcein-AM for 60 min at 37°C on a rotary shaker at 450 rpm. Increasing concentrations of the test compounds were added to aliquots of the cell lysate and the fluorescence was compared with control wells without test compounds.

Cytotoxicity Assay. Each test compound was screened for possible cytotoxic effects with the cytotoxicity detection kit (Roche Diagnostics), a colorimetric assay for the quantification of lactate dehydrogenase activity released from the cytosol of damaged cells into the supernatant.

Statistical Analysis. For calculation of the inhibitor effects, a nonlinear four-parameter fit was used (Grafit, version 4; Erithacus Software, Middlesex, UK) according to the sigmoidal I_max model with the following formula: y = ((I_max  background)/(1 + {x/IC₅₀)^s)) + background, where I_max is the maximal inhibition, IC₅₀ is the concentration leading to half-maximal inhibition of the calcein-AM transport, and s is the slope factor.

View larger version (21K): [in this window] [in a new window]   Fig. 1.   Calcein assay. Evaluation of the inhibitory effect of a compound by calculation of the concentration required to increase baseline calcein fluorescence 2-fold (example: paroxetine in pBCECs, n = 8 wells).

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Method Validation. RT-PCR demonstrated the expression of mRNA of human Pgp and/or porcine pgp1A (Fig. 2). These results were verified by Western blot and immunohistochemistry (data not shown). The different Pgp levels of L-MDR1 and LLC-PK1 were also confirmed in functional experiments by differences in calcein accumulation and its inhibition by verapamil, a typical Pgp inhibitor (Fig. 3).

View larger version (40K): [in this window] [in a new window]   Fig. 2.   Expression of human Pgp (hPgp) and ppgp1A at the mRNA level in L-MDR1, LLC-PK1, and pBCECs detected by RT-PCR. For 1.5% agarose gel electrophoresis, lanes 1, 6, and 11, molecular size marker pUC19/MspI; lanes 2 and 7, no template control; lanes 3 to 5, detection of hPgp in L-MDR1 (3), LLC-PK1 (4), and pBCECs (5); and lanes 8 to 10, detection of ppgp1A in L-MDR1 (8), LLC-PK1 (9), and pBCECs (10).

View larger version (17K): [in this window] [in a new window]   Fig. 3.   Calcein-AM uptake in LLC-PK1 and L-MDR1 cells and influence of the Pgp inhibitor verapamil (VPL, 200 µM). Data are given as means ± S.D. for n = 14 experiments performed in octuplet. p values are determined by analysis of variance with Dunnett's multiple comparison test for post hoc pairwise comparison of the results with the LLC-PK1 cells without inhibitor. *, significant difference compared with LLC-PK1 (p < 0.01).

Evaluation of the Pgp Inhibitory Potency of the Newer Antidepressants. All concentration-response curves reaching a plateau and thus enabling the calculation of IC₅₀ values are shown in Fig. 4. In addition to three of the four control compounds (Fig. 4a), this applied to sertraline, paroxetine, and fluoxetine in L-MDR1 cells (Fig. 4b; Table 1). The potency of sertraline and paroxetine was comparable with the potency of quinidine (Tables 1 and 2). For most compounds plateau effects were not reached either because of cytotoxicity or limited dissolution. To permit a comparison between all compounds tested, IP_50rel and f2 values were calculated (Tables 1 and 2). Only venlafaxine and its metabolites (in both cell lines) and fluoxetine (in pBCECs) did not reach f2. Nevertheless, except for O-desmethylvenlafaxine the highest concentrations analyzed (100 µM for fluoxetine and O-desmethylvenlafaxine, 500 µM for venlafaxine and N-desmethylvenlafaxine) produced significant increases in baseline fluorescence (p < 0.0001, Wilcoxon matched pairs test), thus confirming Pgp inhibition.

View larger version (19K): [in this window] [in a new window]   Fig. 4.   Calcein assay. Concentration-dependent effect of control compounds (A) and the three antidepressants tested reaching a plateau (B) on the calcein accumulation in L-MDR1 cells. Each curve depicts one representative experiment of a series of three to five. Data are expressed as mean ± S.D. for n = 8 wells.

View larger version (20K): [in this window] [in a new window]   Fig. 5.   Calcein accumulation in relation to verapamil in L-MDR1 cells. Values were calculated by dividing f2 of the control verapamil by f2 of the respective test compound. For venlafaxine and O-desmethylvenlafaxine this value was not definable (Table 2). Control compounds are drawn in black, antidepressants in gray. N-dm-, N-desmethyl-.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

The activity of the efflux transporter Pgp affects the pharmacokinetic parameters of many drugs and contributes to numerous pharmacokinetic drug-drug interactions (Yu, 1999). Hitherto, the role of Pgp for the bioavailability, distribution, and excretion of the newer antidepressants and for their interaction with coadministered drugs has not been elucidated thoroughly. For paroxetine, venlafaxine, and fluoxetine the data indicate that they might be Pgp substrates; for citalopram the data are conflicting (Rochat et al., 1999; Uhr et al., 2000; Uhr, 2002).

Even less is known about potential inhibitory characteristics of newer antidepressants on Pgp. In theory, Pgp inhibition by drugs may play an important role in drug safety, because it may increase plasma and brain concentrations of coadministered drugs and thus cause adverse drug reactions. Thus far, only fluoxetine has been tested in this regard. In line with the absence of Pgp-substrate characteristics (Uhr et al., 2000) and in agreement with our results no evidence for a potent interaction was found (Ekins et al., 2002). So far, there are no studies that systematically examine possible interactions with the newer antidepressants at the level of Pgp.

The aim of the present study was to clarify whether the widely used newer antidepressants and their major metabolites inhibit Pgp in vitro as a marker of potential drug-drug interactions in vivo. Another objective was the comparison of the pBCECs and L-MDR1 cells concerning their suitability for a Pgp inhibition assay with calcein-AM as Pgp substrate.

We used an in vitro assay to characterize compounds concerning their ability to inhibit the transport of the Pgp substrate calcein-AM. The fact that all four well characterized Pgp inhibitors (LY335979, SDZ-PSC833, verapamil, and quinidine) were also potent Pgp inhibitors in these assays confirms the applicability of the calcein assays for the evaluation of Pgp-modulating drug effects. The two cell systems used offer different advantages. L-MDR1 cells overexpress human Pgp and can be compared with their parental cell line LLC-PK1. Effects only seen in the transfected cell line can thus be attributed to functional human Pgp. Effects observed in both cell lines can be ascribed to porcine pgp1A, which is expressed in both cells or to other shared characteristics. Indeed, the substantial difference in baseline fluorescence and the fact that verapamil had nearly no effect in LLC-PK1 cells confirms the expected difference in the Pgp activity between parental and transfected cell line (Fig. 2) and emphasizes the suitability of this cell system.

For all drugs tested, the influence on calcein fluorescence in LLC-PK1 cells was either absent or much less pronounced than in the Pgp-overexpressing cell line, indicating that the enhancement of the calcein fluorescence was based on inhibition of human Pgp. This finding indicates that MRP1 and 2 do not play a substantial role in this assay, particularly because the MRP inhibitors probenecid and MK571 had no effects in L-MDR1 cells and only minor effects in pBCECs on the calcein accumulation. This conclusion was also drawn by Eneroth et al. (2001), evaluating a Pgp-overexpressing Caco-2 cell line in a calcein assay.

Interestingly, the maximal fluorescence obtained in L-MDR1 cells was greater than in LLC-PK1 cells (Fig. 2), most likely because L-MDR1 cells are roughly 60% thicker than LLC-PK1 cells and thus accumulation of calcein in the transfected cell line may be greater. Another possibility for such a finding might be differences within the intracellular milieu, e.g., involving different esterase activities.

For comparison primary cell cultures of (porcine) pBCECs as a second, independent cell system were studied. These cells are sumptuous to isolate, differ slightly from preparation to preparation, and are more sensitive to cytotoxic effects. They exhibit a constant pgp1A expression (Hegmann et al., 1992; Huwyler et al., 1996) and due to the lower Pgp expression level compared with L-MDR1 cells they seem particularly suited to detect minor effects of weak inhibitors.

The similarity of the ranking order of inhibition in both cell systems suggests that both are suitable for the evaluation of Pgp-modulating effects. However, it is common experience that absolute values cannot be compared between different cell systems (Tables 1 and 2), especially if Pgp expression levels are different. Therefore, it is conceivable that different concentrations of a compound are needed in the two cell systems to reach similar effects.

Whenever possible, IC₅₀ values were used to compare inhibitory characteristics. If no plateau effect was reached and solubility or cytotoxic effects precluded further increases of the concentration, we also calculated IP_50rel (Bogman et al., 2001) and the concentration needed to increase basal fluorescence 2-fold (f2). The IP_50rel features the advantage of normalizing all values to a control (e.g., verapamil) and thus compensates for interassay variability. However, the concentration at which the effect is compared is predefined arbitrarily, neglecting the fact that the slope factors of different concentration-effect curves may differ and that the potency of different compounds may vary by orders of magnitude. Moreover, the meaningful comparison of compounds with substantially differing maximum effects (efficacy) is not possible, and the IP_50rel does not give clear evidence at which concentration range a compound is active.

Hence, we introduced another assessment method (determination of f2) that also takes the different shapes of the respective concentration-response curves into consideration. Only for very weak inhibitors, which do not lead to a 2-fold increase in basal fluorescence, this method is not suitable. However, such minor effects are normally negligible and may not have importance for the in vivo situation. As an example, in the calcein assay the prototype Pgp substrate digoxin has only minor effects. This is perfectly in line with extensive clinical experience with this drug with only insignificant and rare alteration of the pharmacokinetics of coadministered substrates (Rameis, 1985). Despite the individual advantages and disadvantages of the different cell systems applied and independent of the calculation methods and the cell line used, this series of experiments for the first time shows that the newer antidepressant are inhibitors of Pgp. Indeed, sertraline, desmethylsertraline, and paroxetine had an effect similar to one of the most potent inhibitors (quinidine), whereas citalopram, desmethylcitalopram, venlafaxine, and N-desmethylvenlafaxine exerted only very weak inhibition.

Based on these in vitro data, sertraline and paroxetine bear the apparently largest potential to influence the pharmacokinetics of coadministered drugs at the level of Pgp. However, at usual therapeutic doses, the IC₅₀ value for inhibition of Pgp is around 250-fold higher than the plasma concentration for paroxetine and around 500-fold higher for sertraline (Preskorn, 1997). Von Moltke et al. (1998) recently suggested a model to predict in vivo drug interactions based on in vitro data, which they applied to estimate the inhibitor potential of SSRI on cytochrome P450 CYP2D6. Provided that blood/liver concentration ratios were considered, which amount to 1:36 for sertraline, and not the unbound SSRI plasma concentration, the model yielded good predictions of the in vivo situation. For sertraline, with effective plasma levels of about 65 nM (Preskorn, 1996), this would implicate concurrent liver concentrations of about 2 µM. However, these concentrations are roughly 1 order of magnitude below the concentrations that were found to inhibit Pgp in our assays, suggesting that even if the accumulation of the drugs within the cell (e.g., in the biliary or renal system) is taken into account, the Pgp inhibition observed in vitro might not be clinically relevant.

This is substantiated by the fact that neither sertraline (Rapeport et al., 1996) nor fluvoxamine (Ochs et al., 1989) nor citalopram (Larsen et al., 2001) had a clinically relevant influence on the pharmacokinetic parameters of digoxin, a Pgp prototype substrate. There is only one case report describing a possible increase of serum digoxin levels after treatment with fluoxetine (Leibovitz et al., 1998). On the other hand, in addition to being an inhibitor of CYP2D6, paroxetine is a substrate of this isozyme, whose activity is regulated by a genetic polymorphism. In the absence of active enzyme (poor metabolizer) plasma paroxetine concentrations are up to 25-fold higher than in extensive metabolizers (Sindrup et al., 1992). Accordingly, it cannot be excluded that in poor metabolizer patients administration of high paroxetine doses may translate into clinically relevant modulation of the pharmacokinetics of concomitantly administered Pgp substrates.

In conclusion, the present study demonstrates that not only the widely used L-MDR1 cells are well suited to evaluate drug-induced Pgp inhibition with calcein-AM but also the pBCEC primary cell cultures, which are a well established model to assess the pharmacological properties of the blood-brain barrier. This is the first study that comprehensively quantified inhibitory effects of the newer antidepressants, some of which exerted substantial Pgp inhibition. It remains to be investigated whether this property of the newer antidepressants might lead to drug-drug interactions in patients.

Such interactions might, for instance, be relevant when drugs with low oral bioavailability due to substantial transport back into the gut lumen are to be coadministered, as it has be shown for loperamide when given in combination with quinidine (Sadeque et al., 2000).