Expression and Induction of p-Glycoprotein-1 on Cultured Human Brain Endothelium

Introduction

The ABC-transporter, p-glycoprotein-1 (pgp-1), is an important element of the blood–brain barrier. It is expressed primarily on the lumenal surface of brain endothelium and it actively prevents many toxic molecules and potentially useful drugs from entering the CNS (Löscher and Potschka, 2005). Expression is constitutive on brain endothelium but is generally absent from vascular endothelium that lacks barrier functions and continuous tight junctions (Holloway et al, 2007). Pgp-1 is also reported to be transcriptionally induced by a wide variety of toxic molecules, as well as cytokines, steroids, cellular stressors, heat-shock, and ultraviolet irradiation (Sarkadi et al, 2006; Uchiumi et al, 1993). However, most of these studies were carried out using tumor cells with induced pgp-1, or other cell types that constitutively express it, such as gut epithelium. It is sometimes assumed that transcriptional activators that act on one cell type will act in a similar way on other cell types, but this is often not true.

Many of the inducers of pgp-1 are also substrates for pgp-1, for example, puromycin and verapamil. If such drugs increase pgp-1 on brain endothelium, it potentially creates problems in achieving sustained delivery of the drugs to the CNS, because a drug used for treatment induces the transporter that prevents it from entering the CNS. Moreover, because the substrate specificity of pgp-1 is low, induction of the pgp-1 gene (MDR1) by one drug could affect permeability of other drugs (Begley, 2004).

While studying transcriptional controls of the brain-endothelium phenotype, we aimed to identify tissue-specific elements in the pgp-1 promoter and distinguish them from elements that control induced expression. The study used the human brain endothelial cell line hCMEC/D3 (Weksler et al, 2005) and puromycin, a drug that is both a proposed inducer and substrate of pgp-1, and which has been extensively used to select and purify brain endothelium (Perriere et al, 2005). We compared pgp-1 induction by puromycin with the results produced by verapamil a calcium channel blocker, which is a substrate and functional activator of pgp-1.

Materials and methods

The human brain endothelial cell line, hCMEC/D3, was used at passages 23 to 33 (Weksler et al, 2005). Cells were plated onto collagen-coated 96-well plates for cell surface ELISA or 6-well plates for fluorescence activated cell sorting (FACS) and grown to confluence in EBM2 medium (Cambrex, Wokingham, UK), containing half of the standard growth factors. At this stage the cells were switched into EBM2 without growth factors or antibiotics, but retaining hydrocortisone and sodium ascorbate, for the remainder of the experiment. Cells were rested for 48 h before treatment with puromycin, verapamil, or UV light. Puromycin or verapamil were either present throughout the experiment, with medium changes every third day, or the concentration was increased at each medium change for ramped treatments.

Cell surface ELISA was performed as described earlier (Male et al, 1987), using 70 μL per well of 3 μg/mL anti-pgp-1 as primary antibody (IgG2a clone MRK16, Kamiya Biomedical Co. Seattle, WA, USA). This antibody recognizes an external epitope of pgp-1; preliminary experiments using clone JSB1, which recognizes an internal epitope, produced similar results with permeabilised cells, but staining was much weaker. Intercellular adhesion molecule-2 (ICAM-2) (Serotec, Oxford, UK, clone MCA 1140) was used as an additional control, as a cell surface molecule that is not induced by pgp-1 substrates. ELISA data are shown as mean and s.e.m. of triplicate wells. Results were analyzed by ANOVA, followed by Dunnett's multiple comparison test, comparing all treatments with the untreated control.

FACS analysis was performed by collecting cells with trypsin/EDTA, washing twice in PBS and fixing in 2% paraformaldehyde for 30 mins. Washed cells (2 × 10⁵ in 100 μL) were stained overnight at 4°C with 25 μg/mL anti-pgp-1, or IgG2a isotype control. The secondary antibody was 100 μL of 1/100 fluorescein isothiocyanate (FITC)-labelled sheep antimouse IgG (Sigma-Aldrich, Poole, UK). All incubations were performed in PBS containing 2 mg/mL bovine serum albumin and 1% normal sheep serum. Ten thousand cells of each treatment were analyzed on a Facscan (Becton Dickinson, Franklin Lakes, NJ, USA), with detectors set so that >90% of isotype-control cells registered <10 fluorescence units. Cell numbers on the wells were obtained by counting the cells immediately after they had been collected and before washing and fixation. The ELISA and FACS experiments showing pgp-1 expression were performed on three independent occasions, with similar results and representative plots are shown.

Results

To assess the induction of pgp-1, hCMEC/D3 were initially cultured for 3 days in 0 to 1,000 ng/mL puromycin, 0 to 200 μmol/L verapamil, or after treatment with 0 to 16 J/M² UV irradiation, and then assayed by ELISA. There was no significant induction of pgp-1 (ANOVA P>0.05) at day 3 (not shown). However, cells treated with puromycin for 5 days showed a small but significant increase in pgp-1 expression, which was dose-dependent and further increased by day 8 (Figure 1). Ramping up the puromycin concentration, during the 8-day culture period produced additional increases in pgp-1 expression (Figure 1). Cells subjected to higher levels of irradiation (>3 J/M²) or puromycin (>0.6 μg/mL) at the start of the experiment, appeared to have fewer cells by day 3, although the monolayers were still confluent at the time of assay.

OPEN IN VIEWER

Figure 1

ELISA of pgp-1 and ICAM-2 expression at day 8, after treatment of hCMEC/D3 endothelial cells with the indicated doses of puromycin. Results show mean and s.e.m. (controls n=6, treatments n=3), and show a substantial increase in pgp-1 expression (ANOVA, P<0.0001). There was a small decrease in ICAM-2 expression (ANOVA, P=0.015), which was only significant at the highest dose of puromycin (P<0.001).

At first sight, the increased expression of pgp-1 suggested an increase in transcription or translation, but the induction was surprisingly slow—8 days. Consequently, pgp-1 expression was examined in more detail by FACS. Figure 2A shows that the range of pgp-1 expression on resting cells was very variable. Moreover, a ramped puromycin treatment over 8 days (0.4 to 1.0 μg/mL), resulted in the disappearance of the low-expressing subpopulation. However, there was no overall increase in expression on the high-expressing cells. Effectively, puromycin seemed to eliminate the subset of cells that expressed the lowest levels of pgp-1—an example of selective pressure acting on a cell population in vitro.

OPEN IN VIEWER

Figure 2

FACS analysis of pgp-1 expression on hCMEC/D3 cells. Untreated cells (filled histograms) were compared with puromycin- or verapamil-treated cells (unfilled histograms). Cells stained with isotype-control antibody (Con) are indicated. (A) Cells were treated with an increasing concentration (0.4 to 1.0 μg/mL) of puromycin over days 0 to 6 and stained at day 8. (B) Cells were treated with increasing concentrations of (0.4 to 1.4 μg/mL) of puromycin on days 0 to 11. Surviving cells (visually estimated as <5%) were regrown in EBM2 medium without puromycin until confluent, and stained for pgp-1 at day 28. (C) Cells were treated with 100 to 200 μmol/L verapamil over days 0 to 6 and stained at day 8. The data from independent experiments (mean±s.e.m., n=3) are summarized in the histograms, which show the number of cells recovered (D) and pgp-1 expression (E) of cells treated with either puromycin or verapamil for 8 days, compared with untreated cells (=100%).

Further investigation of the cells at day 8 showed that the number of cells recovered from puromycin-treated wells was only 14 to 28% (range) of that on matched, untreated wells, even though the monolayers remained confluent. The puromycin-treated cells were significantly larger when analyzed by flow cytometry (forward scatter 654±23, compared with 472±33 in untreated cells—mean and coefficient of variance, P<0.001 KS-test). Moreover, there was a small reduction in expression of ICAM-2 (a control, noninducible endothelial surface molecule) measured by ELISA, although this was only significant at the highest doses of puromycin (Figure 1). These results confirm that puromycin causes the loss of a subpopulation of endothelium, but that the remaining cells, spread and become larger, to maintain the integrity of the monolayer.

To determine whether the higher expression of pgp-1 after puromycin treatment was a stable characteristic, puromycin-treated cells were regrown in medium without puromycin and pgp-1 expression was reanalyzed by FACS. Cells treated with ramped puromycin (0.4 to 1.0 μg/mL) reverted to the same level of pgp-1 expression as control cells after 8 days in medium lacking puromycin (data not shown). Cells that had been treated with higher doses of puromycin (0.4 to 1.4 μg/mL, over 8 days), and which had suffered >95% cell death, could be recovered and regrown in EBM2, and these cells retained a higher level of pgp-1 expression, than control cells, at day 28 after several rounds of cell division needed to reform the monolayers (Figure 2B).

Treatment of the cells with 100 to 200 μmol/L verapamil over 8 days also resulted in an increase in pgp-1 expression detectable by FACS (Figure 2C), which was comparable to the increase seen on puromycin-treated cells (Figure 2E). However, verapamil was much less toxic than puromycin, and the cell number on verapamil-treated monolayers was not significantly lower than on untreated wells (Figure 2D). Moreover, the profile of the FACS plots on verapamil-treated cells indicated an overall increase in pgp-1 expression (Figure 2C), rather than the selective loss of low expressors.

Discussion

Many pgp-1 substrates are thought to induce pgp-1 expression, based on results seen in tumor cells (Scotto, 2003). However, transcriptional controls in cells that constitutively express a protein are generally different from controls on induced expression. This study suggests that brain endothelial cells, which constitutively express high levels of pgp-1, are not further induced by puromycin. Rather, the apparent slow induction of pgp-1 in vitro by puromycin, which has also been observed with rat brain endothelial lines (Demeuse et al, 2004), seems to be due to selective cell death of low-expressing populations. Interestingly, it has been noted that the development of multidrug resistance in tumors is often associated with an increase in pgp-1 expression and it has been hypothesized that the increase in pgp-1 expression is due to selective survival of a subset of tumor cells that originally expressed high pgp-1 levels. This mechanism is analagous to the one described here for induction of pgp-1 on brain endothelium in vitro, by puromycin.

Verapamil was originally developed as a calcium channel blocker to treat cardiac conditions, but was subsequently found to act as a substrate for pgp-1, and has been used as a competitive inhibitor of pgp-1 activity in vitro. Verapamil is much less cytotoxic than puromycin in the concentrations used in this study. It is therefore less likely that verapamil increases pgp-1 by causing selective cell death of low expressors. The mechanism by which verapamil increases pgp-1 expression is uncertain, but could involve mobilization of intracellular pgp-1 to the cell surface or increased transcription/translation of pgp-1.

The transcriptional control of pgp-1 is complex. The MDR1 promoter seems to require activation through transcription factors interacting with a CCAAT-box and a GC-rich box (Sundseth et al, 1997), which are modulated by several other elements as part of the MDR1 enhanceosome (Scotto, 2003). On rodent brain endothelium, pgp-1 can be further induced through the pregnane-X receptor (Bauer et al, 2004) — cell surface expression was induced over 3 to 6 h by dexamethasone or pregnenolone-16α-carbonitrile in vitro and after 3 days treatment in vivo. Induction appeared to be at least partly because of increased transcription, together with early mobilization of pgp-1. However, although the effect of verapamil on pgp-1 expression is similar to that of dexamethasone, there is no evidence that verapamil activates the MDR1 gene through the same transcriptional controls.

Drug resistance presents problems for drug delivery to the CNS, particularly for tumors and epilepsy (Löscher and Potschka, 2005), and it has been proposed that increased expression of multidrug transporters on the brain endothelium underlies resistance. In support of this view, an increase in pgp-1 expression has been noted on brain capillaries in epileptic subjects (Tishler et al, 1995), and it has been debated whether this is due to the condition itself, or induction by drugs. However, because the increase is local rather than global, it suggests that it is not because of the treatment (Sisodiya et al, 2002). The induction of pgp-1 on brain endothelium by its substrates in vitro has also raised concerns about therapeutic drugs inducing drug resistance in vivo (Bauer et al, 2005). This study shows that increased pgp-1 expression in vitro may be related to selective cytotoxicity rather than increased pgp-1 synthesis, in conditions that are unlikely to occur in vivo. Hence, the concerns about drug-induced pgp-1 induction in vivo are, in some cases, unfounded.