Human P-glycoprotein Transports Cortisol, Aldosterone, and Dexamethasone, but Not Progesterone*

We expressed human MDRl cDNA isolated from the human adrenal gland in porcine LLC-PK1 cells. A highly polarized epithelium formed by LLC-GA5-COL300 cells that expressed human P-glycoprotein specifically on the apical surface showed a multidrug-resistant phenotype and had 8.3-, 3.4-, and 6.5-fold higher net basal to apical transport of 3H-labeled cor- tisol, aldosterone, and dexamethasone, respectively, compared with host cells. But progesterone was not transported, although it inhibited azidopine photoaffinity labeling of human P-glycoprotein and increased the sensitivity of multidrug-resistant cells to vinblastine. An excess of progesterone inhibited the transepithelial transport of cortisol by P-glycoprotein. These results suggest that cortisol and aldosterone are physiological substrates for P-glycoprotein in the hu- man adrenal cortex and that substances that efficiently bind to P-glycoprotein are not necessarily transported by P-glycoprotein.

(MDR)' cells (6)(7)(8), 2 ) inhibiting ['Hlvinblastine or vincristine binding to the plasma membrane of MDR cells (7-9), 3) enhancing the reduced uptake of vinblastine in MDR cells (8), and 4) increasing the sensitivity of MDR cells to vinblastine (7,8 ) and because 5 ) progesterone itself photoaffinity labels P-glycoprotein (10). But it is not clear whether Pglycoprotein transports steroids because verapamil-sensitive transport of progesterone could not be measured in NIH-3T3 transformants expressing human P-glycoprotein' and the steady-state level of progesterone accumulation has been suggested to be very similar in drug-sensitive and -resistant rodent cells (8), and apparently no efflux of progesterone is observed in rodent MDR cells (7).
In this study, we expressed human MDRl cDNA isolated from the human adrenal gland in porcine LLC-PK1 cells and found that P-glycoprotein transepithelially transports cortisol, aldosterone, and dexamethasone, but not progesterone. Our results suggest that these steroid hormones could be physiological substrates for P-glycoprotein in the human adrenal cortex. Cell Culture and Transfection-LLC-PK1 cells were obtained from American Type Culture Collection and propagated in M199 medium supplemented with 10% fetal calf serum. A human P-glycoprotein expression vector (pSKGA) was constructed by replacing the MDRl cDNA isolated from human MDR cell line KBX2.5 (11) in plasmid pSK1.MDR (12) with the MDRl cDNA isolated from a normal adrenal gland (13). LLC-PK1 cells were transfected with pSKGA by the calcium phosphate coprecipitation method (14). Cells were first selected in 10 nM 2'-deoxycoformycin and 4 p M 9-~-D-XylOfUranOSyladenine (15) for 2 weeks. One transformant (LLC-GA5) was further selected in stepwise increasing concentrations (20,40,80,150, and 300 ng/ml) of colchicine. LLC-GA5-COLBOO cells were selected at 300 ng/ml colchicine and maintained at 300 ng/ml colchicine.

Human P-glycoprotein
as a Steroid Transporter 24249 done with the enhanced chemiluminescence method from Amersham Corp. Photoaffinity labeling was done as described previously (17) with minor modifications. Briefly, membrane vesicles corresponding to 10 pg of protein were reacted with 0.2 pM ["Hlazidopine in the presence or absence of a 1000-fold excess (200 p~) of steroids or vinblastine at room temperature for 20 min. The reaction mixture was then irradiated with a UV lamp (Ultra-Violet Products, Blak Kay Type XX-15L) for 30 min on ice. Labeled proteins were separated by SDS-polyacrylamide gel electrophoresis on 7% gels. Electron Microscopic Immunocytochemistry-Monolayers of LLC-GA5-COL300 cells were primarily fixed in 4% paraformaldehyde, washed in phosphate-buffered saline containing 2 mg/ml bovine serum albumin, and scraped from the dish; and the floating cell sheets were incubated with monoclonal antibody MRK16 (10 pg/ml), followed by incubation with antimouse IgG-gold (10 nm) conjugate (BioCell). Then the cell sheets were post-fixed in glutaraldehyde and Os04 and embedded primarily in 1% low melting point agarose and then in Epon 812. Sections (200 nm thick) were analyzed at 100 kV in a Hitachi H-7100 electron microscope.
Transepithelial Transport of "H-Labeled Vinblastine and Steroids "Cells were seeded on microporous polycarbonate membrane filters (3.0-pm pore size, 24.5-mm diameter, Transwell'"" 3414, Coster") at a cell density of 4 x 1 0 and 5 X IO5 cells/cm2 for LLC-PK1 and LLC-GA5-COL300, respectively. The cells were grown for 3 days; and 6 h before transport experiments, the culture medium was replaced with fresh medium without colchicine. For mesurement of transepithelial transport, the medium in either the basal or apical side of the monolayers was replaced with 2 ml of medium containing "H-labeled vinblastine or steroid and 0.66 p~ 14C-labeled inulin. The cells were incubated at 37 "C; an aliquot (25 pl) of the medium in the other side was taken at 1,2, and 3 h, and the appearance of radioactivities in the other side was measured and presented as the fraction percent of the total radioactivity. Each directional transport was measured with more than three filter-bottomed cups and presented with standard error. The paracellular fluxes monitored by the appearance of ["Clinulin in the other side were 1.5% of the total radioactivity per h.
Effect of Progesterone on Drug Resistance-The IC, (the drug concentration that inhibits cell growth by 50% after 72 h) for LLC-GA5-COL300 cells was determined by a colorimetric assay using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (18) from dose-response curves using increasing concentrations of vinblastine in the presence and absence of steroid.

RESULTS
Polarized Expression of Human P-glycoprotein in Porcine LLC-PKl Cells-Because steroids are very lipophilic so that accumulation data are unreliable due to the nonspecific binding to plasma membrane and filter, we used a transepithelial transport system (19) to identify the physiological substrates transported by human P-glycoprotein. We expressed MDRl cDNA isolated from the human normal adrenal gland (13) in LLC-PK1 cells, which are derived from the porcine kidney proximal tubule. LLC-PK1 cells form a highly polarized epithelium, and the polarized transport of nutrients and xenobiotic compounds across cultured epithelia has been well documented (20)(21)(22)(23). LLC-PK1 cells were transfected with a human P-glycoprotein expression vector and were first selected for expression of the adenosine deaminase gene, used as a selectable marker. Several clones resistant to 9-/3-~xylofuranosyladenine were examined for insertion of human MDRl cDNA in the host chromosome by Southern hybridization (data not shown). One transformant (LLC-GA5) was further selected in medium containing colchicine to isolate cells expressing more P-glycoprotein. P-glycoprotein expressed in the plasma membrane of LLC-GA5-COL300, which was selected in 300 ng/ml colchicine, was detected as a 170-kDa mature form by Western blot analysis using anti-Pglycoprotein monoclonal antibody C219 (Fig. lA) and was photoaffinity-labeled by ["Hlazidopine ( Fig. 1B). LLC-PK1 host cells express a marginal level of porcine P-glycoprotein (Fig. 1A). LLC-GA5-COL300 showed 54-, 150-, and 167-fold resistance to colchicine, vinblastine, and adriamycin, respectively, compared to LLC-PK1 cells. Because we did not observe any colchicine-resistant clones in nontransfected controls during the stepwise selection with increasing concentrations of colchicine and because no amplification of porcine mdr gene(s) was observed in LLC-GA5-COL300 by Southern hybridization with the human MDRl probe, which could detect porcine mdr gene(s) by cross-hybridization (data not shown), the P-glycoprotein expressed in LLC-GA5-COL300 cells reacted with C219 is considered to be mainly the human one. Immunostaining using monoclonal antibody MRK16, which reacts selectively with human P-glycoprotein, confirmed the high expression of human P-glycoprotein in the plasma membrane of LLC-GA5-COL300 (data not shown), and electron microscopic immunocytochemistry using MRK16 showed that human P-glycoprotein is specifically expressed on the apical surface of the cells (Fig. 2), as previously reported using Madin-Darby canine kidney cells as host cells (24).
Transepithelial Transport of Vinblastine in Monolayers of LLC-PKI and LLC-GA5-COL300-The substrates for P-glycoprotein are highly lipophilic and so enter cells freely through both surfaces. If there is no specific transport system, basal to apical and apical to basal transport will be the same. However, because human P-glycoprotein localizes specifically in the apical plasma membrane in LLC-GA5-COL300 cells, substrates for P-glycoprotein entering the cell from the basal surface will be expelled from the apical surface, and those entering the apical plasma membrane are likely to be caught and pumped back to the medium by P-glycoprotein, resulting in a net basal to apical flow of substrates as shown in the model (Fig. 3A). In LLC-GA5-COL300 cells, basal to apical transport of vinblastine increased -2-fold, and apical to basal transport decreased 1.5-2-fold compared to transport in LLC-PK1 cells, resulting in a 6-10-fold higher net basal to apical transport (Fig. 3B).
Transepithelial Transport of Steroid Hormones-By using this transepithelial transport system in epithelia formed by LLC-PK1 and LLC-GA5-COL300, we measured transport of 'H-labeled cortisol, aldosterone, progesterone, and dexamethasone (Fig. 4). In LLC-GA5-COLSOO cells, basal to apical transport of cortisol, aldosterone, and dexamethasone increased and apical to basal transport decreased compared to transport in LLC-PK1 cells, resulting in 8.3-, 3.4-, and 6.5fold higher net basal to apical transport of cortisol, aldosterone, and dexamethasone, respectively, than host cells. Transepithelial transport of cortisol, aldosterone, and dexamethasone was inhibited by verapamil (Fig. 5 ) , a competitive inhibitor of P-glycoprotein (25), which confirmed that steroid transport is mediated by human P-glycoprotein. But transepithelial transport of progesterone in LLC-GA5-COL300 was not higher than that in host cells. We examined transepithelial transport of progesterone ranging in concentrations from 1.2 nM (one-tenth the concentration of that used in Fig. 4) to 50 PM, a concentration which increased the sensitivity of LLC-GA5-COL300 cells to vinblastine (Fig. 6A) and inhibited the transepithelial transport of cortisol (Fig. 7). But transepithelial transport of progesterone in LLC-GA5-COL300 and host cells was indistinguishable from that shown in Fig. 4; thus, progesterone in a range of physiological concentrations cannot be transported by P-glycoprotein. This is consistent with the lack of apparent efflux of progesterone from rodent MDR cells reported by Yang et al. (7). These results suggest that human P-glycoprotein transports cortisol, aldosterone, and dexamethasone, but does not transport progesterone.
Interaction of Steroid Hormones with Human P-glycoprotein -Progesterone was considered to be a physiological x 9.000 X 50,000 substrate because it binds to P-glycoprotein (8, lo), inhibits vincristine binding to P-glycoprotein (9), and inhibits transport of vinblastine by P-glycoprotein in MDR cells (7, 8). However, P-glycoprotein expressed in the apical membranes of LLC-GA5-COL300 cells did not transport progesterone. The inhibition of [3H]azidopine photoaffinity labeling and the increase of sensitivity to vinblastine by progesterone were shown with mouse mdrl b-encoded P-glycoprotein and mouse MDR cells (7, 8). The inhibition of the ATP-dependent binding of vincristine was examined with the plasma membrane of adriamycin-resistant variants of human myelogenous leukemia K562 cells (9). Mouse mdrlb-encoded P-glycoprotein and human MDRl-encoded P-glycoprotein expressed in the plasma membrane of human myelogenous leukemia cells and those in the apical membrane of porcine kidney cells could behave differently when interacting with progesterone. We therefore examined the effects of steroids on azidopine photoaffinity labeling and sensitivity to vinblastine using LLC-GA5-COL300 cells.
Progesterone strongly inhibited azidopine photoaffinity labeling of human P-glycoprotein in the apical membrane of porcine cells, and aldosterone and testosterone weakly inhibited it, whereas cortisol and dexamethasone, even at 1000fold, scarcely inhibited it (Fig. 1B). These results were consistent with the apparent inhibitory constants of steroid hormones on the ATP-dependent binding of vincristine reported by Naito et al. (9). The apparent inhibitory constants of progesterone, testosterone, and cortisol were reported to be 7, 20, and >200 PM, respectively. These results indicate that progesterone has a high affinity for human P-glycoprotein whereas cortisol and dexamethasone have very low affinities for human P-glycoprotein.
The ability of progesterone or cortisol to reverse vinblastine resistance in LLC-GAS-COL3OO cells was examined (Fig. 6A). Progesterone increased the sensitivity of LLC-GAS-COL300 cells to vinblastine, whereas cortisol scarcely increased it, consisting with the effect of these steroid hormones on vinblastine accumulation and on vinblastine resistance in mouse Human P-glycoprotein as a Steroid Transporter  affinity for human P-glycoprotein whereas cortisol has a very low affinity for human P-glycoprotein.

DISCUSSION
In this paper, we showed that human P-glycoprotein transports cortisol and aldosterone, but not progesterone. Human MDR cells (8). Progesterone had a small influence (<2-fold) on the IC5,, for vinblastine in LLC-PK1 host cells (Fig. 6B). The concentration (50 PM) of progesterone and cortisol allowed >90% of normal cell growth in the absence of vinblastine. These results also indicate that progesterone has a high Human P-glycoprotein as a Steroid Transporter P-glycoprotein is expressed in the zona glomerulosa, zona fasciculata, and zona reticularis (3). Cortisol, a glucocorticoid, is the main steroid produced in the human adrenal gland, particularly in the zona fasciculata (26). Aldosterone is the main mineralocorticoid produced in the zona glomerulosa of the human adrenal gland (27). Results of transport experiments and the location of expression strongly suggest that Pglycoprotein aids in the secretion of cortisol and aldosterone out of cells in the human adrenal cortex. During human pregnancy, estrogens and progesterone are produced in the fetoplacental unit, and metabolic intermediates move among the fetus, placenta, and mother. Because Pglycoprotein is expressed in the pregnant uterus and placenta (6), it may be important in transporting metabolic intermediates of estrogens and progesterone in the fetoplacental unit. Because an excess of progesterone inhibited the transepithelial transport of cortisol (Fig. 7), it may regulate the function of P-glycoprotein in the fetoplacental unit and adrenal cortex.
Cortisol is transported by human P-glycoprotein, although it scarcely inhibits azidopine photoaffinity labeling (Fig. 1B) or scarcely affects sensitivity to vinblastine (Fig. 6). Progesterone is not transported by human P-glycoprotein, although it inhibits azidopine photoaffinity labeling and increases the sensitivity of LLC-GA5-COL300 cells to vinblastine. These results suggest that a slight difference in the structure of steroids determines whether it is transported by P-glycoprotein and that agents that efficiently compete with azidopine photoaffinity labeling and vinblastine transport of P-glycoprotein are not necessarily transported by P-glycoprotein. Safa et al. (28) reported that an amino acid substitution (Gly to Val) a t position 185 of human P-glycoprotein increases the efficiency of P-glycoprotein in the efflux of colchicine, but decreases the binding of a photoactive colchicine analog. In contrast, the Gly to Val substitution decreases its efficiency in the efflux of vinblastine, but increases the binding of a photoactive vinblastine analog (28). These results suggest that efficiency in competing with azidopine photoaffinity labeling does not parallel the efficiency in the transport by P-glycoprotein. It is proposed that the Gly to Val substitution at position 185 affects not the initial binding site of P-glycoprotein, but another site, associated with the release of P-glycoprotein-bound drugs to the outside of the cell (28). Naito and Tsuruo (25) reported that progesterone has an apparent inhibitory constant similar to that of adriamycin for the ATPdependent binding of vincristine to human MDR cells. Adriamycin is actively transported out of the cells by P-glycoprotein, whereas progesterone is not transported. These results suggest that the binding efficiency of a substrate is not the factor that determines whether P-glycoprotein transports it.
It has not been determined how P-glycoprotein recognizes many different hydrophobic molecules that share no obvious structural similarity and how P-glycoprotein actively transports them out of the cells. We are now examining which steroid metabolites are transported by human P-glycoprotein and which are not. These experiments may shed light on the structural specificity of substrates for P-glycoprotein to transport. Using this transepithelial transport system, we may not only be able to identify the physiological functions of human P-glycoprotein in more detail, but also find a clue to understand the mechanism through which substrates are recognized and transported by P-glycoprotein.