Human P-glycoprotein Transports Cyclosporin A and FK506*

Cyclosporin A, a cyclic undecapeptide, and FK506 are efficient immunosuppressive agents. They also at-tract attention as effective P-glycoprotein modulators that inhibit P-glycoprotein from binding to anticancer drugs and overcome multidrug resistance. Cyclosporin A itself interacts with a common binding site of P-glycoprotein to which Vinca alkaloids and verapamil bind. We were interested to determine whether cyclosporin A and FKSO6 are substrates for P-glycoprotein to transport, and we studied their transcellular transport. In LLC-PKI cells, derived from porcine kidney proximal tubule and forming a highly polarized epithelium, cyclosporin A was transported in a saturable manner. LLC-GAS-COLSOO, a transformant cell line derived by transfecting LLC-PK, with human MDRI cDNA isolated from normal adrenal gland, expresses P-glycoprotein specifically on the apical surface and shows a typical multidrug-resistant phenotype. LLC-GAS-COL300 cells showed increased transport of cy- closporin A from the basal to the apical side. Kinetic analysis showed that this transport was a typical sat- urable transport with the calculated apparent Michaelis constant (KY) and the maximum flux ( Vm=) as 8.4 p~ and 2.4

duction of human P-glycoprotein confers resistance against the peptide ionophore antibiotics gramicidin D (7) and valinomycin.* Chinese hamster and human P-glycoproteins were reported to be capable of transporting a tripeptide, N-acetylleucyl-leucyl-norleucinal, which is an inhibitor of various intracellular proteases (8). It was also reported that the mouse mdr3 gene but not the human MDRl gene could complement yeast STEG, which is a homologue of mammalian m d r genes and mediates export of a-factor mating peptide in Saccharomyces cerevisiue (9,10).
In in vitro experiments, some lipophilic compounds are able to reverse the MDR phenotype. These compounds include natural alkaloids, some calcium channel blockers, protein kinase inhibitors, and some immunosuppressive agents. They are thought to modulate MDR by interacting with drugbinding site(s) of P-glycoprotein competitively (11). Among them, cyclosporin A, a cyclic undecapeptide with highly immunosuppressive effects, efficiently prevents P-glycoprotein from binding with anticancer drugs and overcome MDR, and attracts a great deal of attention as an effective P-glycoprotein modulator (12)(13)(14). A photoreactive analogue of cyclosporin A was shown to label P-glycoprotein as well as cyclophilin in living cells, and this labeling was inhibited by cyclosporin A, cyclosporin H, diltiazem, and verapamil, but not by colchicine (15). A binding study using membrane vesicles of MDR cells indicated that cyclosporin A competitively interacts with a common drug-binding site of P-glycoprotein, to which Vinca alkaloids and verapamil bind (16). It was also reported that MDR Chinese hamster cells accumulated less cyclosporin A than the corresponding drug-sensitive cells (16, 17), but it remains to be answered if cyclosporin A is a substrate for human P-glycoprotein to transport. This question is important not only for understanding the exact mechanism by which effective modulators reverse MDR, but also for understanding the pharmacokinetics of immunosuppressive agents.
We have reported about transcellular transport by human P-glycoprotein expressed in LLC-PK1 cells, derived from the epithelial cells of porcine kidney proximal tubule (18, 19). Human P-glycoprotein was expressed specifically on the apical surface, and transported a cardiac glycoside, digoxin (19), vinblastine, and the steroid hormones aldosterone and cortisol (18). Because this transcellular transport system is useful to investigate transport by P-glycoprotein without annoyance by nonspecific binding of lipophilic compounds, we attempted to show if cyclosporin A and the more potent immunosuppressive drug FK506 (20) are transported by P-glycoprotein. Here we report that cyclosporin A and FK506 were substrates for P-glycoprotein to transport and P-glycoprotein reduced the accumulation of cyclosporin A and FK506 in cells.

6077
Cell Lines-LLC-PK1 is a cell line derived from porcine kidney proximal tubules. LLC-PKl cells were transfected with human MDRl cDNA (21), and a clone designated as LLC-GA5-COL300 was isolated by selecting with 300 ng/ml colchicine (18). LLC-GA5-COL300 cells express P-glycoprotein specifically in the apical membrane.
Transcellular Tramport of Cyclosporin A and FK506-LLC-PK1 cells were planted on bottom-filtered cups at a density of 4 X lo6 cells/cm2. LLC-GA5-COL300 cells were planted at a density of 5 X 10' cells/cm2, and colchicine was added at a final concentration of 300 ng/ml. Cells were incubated over 3 nights, and media were changed for fresh and colchicine-free medium 6 h before experiments. For measurement of transcellular transport, the medium of either the basal or apical side of the monolayer was replaced with medium containing 3.7 kBq/ml (8.6 nM) [3H]cyclosporin A and 3.7 kBq/ml (24.8 p~) ["C]sucrose, or 2.9 kBq/ml (6. LLC-PKI cells derived from epithelial cells of porcine kidney proximal tubule form an epithelium consisting of a monolayer of highly polarized cells. Some lipophilic compounds, such as vinblastine, and some sterols can pass across the epithelium by simple diffusion, and the amount passed from the basal to the apical side and that from the apical to the basal side are almost equal (18). LLC-GAB-COL3OO is a transformant cell line that was derived by transfection with human MDRl cDNA isolated from normal adrenal gland (21) and by selection with colchicine. LLC-GAS-COL3OO expresses P-glycoprotein specifically on the apical surface and shows a typical MDR phenotype. We demonstrated that in LLC-GA5-COL300 cells the amounts of vinblastine, aldosterone, cortisol (18), and digoxin (19) passed from the basal to the apical side were increased and the amounts from the apical to the basal side were decreased, indicating that these compounds are substrates for P-glycoprotein to transport. Using this system, we measured the time course of transcellular transport of cyclosporin A (Fig. lA). After 3 h, the amount passed across the LLC-PK1 epithelium from the basal to the apical side and that from the apical to the basal side were 4.5 and 3.5%, respectively. These amounts are quite low compared to that of vinblastine and sterols, since the amount of vinblastine and sterols passed across the LLC-PKl monolayer in both directions reached 20-30% in 3 h (18). However, because the amounts of ["C]sucrose moved across the epithelia, which represent paracellular leak of LLC-PK1 monolayer, were only 1-2% in 3 h (Fig. lA, lower panel), we considered that a fraction of cyclosporin A passed through the transcellular route.
The amount of cyclosporin A passed across LLC-GA5-COL300 epithelium from the basal to the apical side reached 10% in 3 h, although that in the opposite direction remained 3.9% (Fig. lA). This increased amount of cyclosporin A passed across the epithelium was suppressed by adding a 1000-fold excess of unlabeled cyclosporin A (Fig. 1D). These results suggest that P-glycoprotein transports cyclosporin A and that this transport could be saturable. The amount of ['4C]sucrose moved across the LLC-GA5-COL300 monolayer was signifi- cantly higher than that in LLC-PK1 (Fig. lA, lower panel), indicating that the tight junction of LLC-GA5-COL3OO cells was slightly loose for a small aqueous solute compared to that of LLC-PK1 cells. However, it might not be the case with lipophilic substances, because almost equal amounts of [3H] nitrendipine, which is a calcium channel blocker and was reported not to bind to the membrane vesicles of MDR cells (22), were transported in host and transformant cells in both directions (Fig. 1B). Furthermore, the difference in the amount of transported [3H]staurosporine, a protein kinase inhibitor, between LLC-GA5-COL300 cells and LLC-PK1 cells was smaller than ["C]sucrose and was less than l%/h in both directions (Fig. IC). Because the molecular weight of cyclosporin A (1201) is significantly higher than that of staurosporine (495), the paracellular leak across LLC-GA5-COL300 monolayer is expected to be lower than that of staurosporine. These results suggest that the tight junction of LLC-GAS-COL3OO cells was as tight for lipophilic substances such as cyclosporin A as that of host cells and that it is worth comparing the transport of lipophilic substances across LLC-GA5-COL300 monolayers with that across LLC-PK1 monolayer.
Cells were lysed in 0.3 N NaOH after 3 h, and the amounts of cyclosporin A accumulated in cells were measured (Fig. 2). When cyclosporin A was added to the apical side, the amount of cyclosporin A accumulated in LLC-GA5-COL300 was onesixth of that in LLC-PK1, although no decrease of the apical- to-basal transport in transformant cells was detected in transcellular transport experiments (Fig. 1A). This is probably because the amount passed from the apical to the basal side of host cells was very low and the decrease from that was undetectable. On the other hand, when cyclosporin A was added to the basal side, the difference in accumulation between the transformant and the host cells was not significant, although LLC-GA5-COL300 cells accumulated less cyclosporin A than LLC-PKl. These results could be explained as follows. When cyclosporin A was added to the apical side, Pglycoprotein localized in the apical membrane efficiently caught cyclosporin A during the process of entering cells, and pumped it back to the apical medium. When cyclosporin A was added to the basal side, considerable amount of cyclosporin A was adsorbed by cellular components before it reached to the apical membrane. These results indicate that P-glycoprotein mediates the efflux of cyclosporin A from cells, and are consistent with the polarized localization of P-glycoprotein in the apical plasma membrane of LLC-GA5-COL300. We further investigated the mode of interaction of cyclosporin A with P-glycoprotein by a kinetic study. Various concentrations of cyclosporin A were added in the media of basal side, and the radioactivities appeared in the apical side were measured (Fig. 3A). Transport of cyclosporin A across LLC-PK1 as well as LLC-GA5-COL300 monolayer was saturated at the concentration of 20-40 p~. Net transport, calculated by subtracting the amount of cyclosporin A passed across the LLC-PKl monolayer from that across the LLC-GA5-COL300 monolayer (Fig. 3A), representing the P-glycoprotein-mediated transcellular transport, was plotted in Fig. 3B. This subtraction should cancel the nonspecific transport, including the simple diffusion and the paracellular leak, provided that these effects are similar between two kind of cells. The transport of cyclosporin A associated with P-glycoprotein showed a typical pattern of saturable transport. The double-reciprocal plot of the rate of transport against the cyclosporin A concentration at the basal side clearly showed a linear relationship with the correlation coefficient of 0.99 (Fig. 3C). The apparent Michaelis constant and the maximum transport (V-) were calculated as 8.4 pM and 2.4 nmol/mg protein/h, respectively. These parameters reveal the overall characteristics of transcellular transport, which was accelerated by P-glycoprotein because we used the concentration at the basal side of media instead of intracellular concentration. It has been reported that the concentration of vincristine necessary to inhibit the growth of multidrug-resistant K562/ADM cells was shifted from 680 nM to 84 nM and to 15 nM in the presence of 3 p M and 10 p~ cyclosporin A, respectively, whereas that of the drug-sensitive parent K562 cells was 0.8 nM (20). Kzp of 8.4 p M for the transcellular transport of cyclosporin A seems to be reasonable, if the overall characteristics of cyclosporin A transcellular transport and the effect of cyclosporin A to overcome MDR phenotype can be directly compared. Although the values are apparent, our data above do indicate that P-glycoprotein transports cyclic peptide cyclosporin A by a saturable mechanism. Because cyclosporin A, an efficient and clinically used immunosuppressive agent, was found to be a substrate for Pglycoprotein to transport, we were interested to find if another potent immunosuppressive agent, FK506, is also transported by P-glycoprotein. It was reported that FK506 modulated Pglycoprotein functions more effectively than cyclosporin A (20). FK506 overcame resistance to vincristine and doxorubicine of various MDR cells in uitro and increased the chemotherapeutic effect of vincristine for MDR tumor cell-bearing mice in uiuo (20). The amount of FK506 passed from the basal to the apical side and that from the apical to the basal side of LLC-PK1 epithelium were 27 and 26%, respectively (Fig. 4A), indicating that FK506 passed across LLC-PKI monolayer probably by simple diffusion. The amount passed across LLC-GA5-COL300 monolayer from the basal to the apical side was increased to reach 48% in 3 h, and the amount from the apical to the basal side was decreased to lo%, representing the typical pattern of oriented transport by P-glycoprotein observed in our previous study (18, 19). Accumulated FK506 in LLC-GA5-COL300 after 3 h decreased to one-fourth and onefifth of that in LLC-PK1, when the donor side was the basal and the apical side, respectively. These results indicate that P-glycoprotein mediated transport of FK506.
Our data described above indicate that cyclosporin A and FK506, which are clinically useful immunosuppressive agents, are substrates for P-glycoprotein to transport. Because cyclosporin A was reported to interact with the common drugbinding site of P-glycoprotein (16), we examined the inhibitory effects of 1000-fold excesses of verapamil and vinblastine (8.6 pM) on transcellular transport of cyclosporin A (Fig. 1B).
However, no obvious effect was observed, although 1000-fold excess of cyclosporin A inhibited it. It was reported that photoaffinity labeling of rodent P-glycoprotein by a cyclosporin A analogue was inhibited by 30 p~ verapamil(l5) and that reduced accumulation of cyclosporin A in Chinese hamster MDR cells were reversed by 100 p~ verapamil (17). One possible reason that verapamil or vinblastine did not affect the transport under our conditions is that the 8.6 p~ (1000fold excess) verapamil and vinblastine we used was not enough to compete with cyclosporin A on human P-glycoprotein. However, even 20 p~ verapamil failed to inhibit transcellular transport of cyclosporin A (data not shown). We could not use higher concentrations of verapamil and vinblastine, because those caused remarkable paracellular leaks. Alternatively, the competition of photoaffinity labeling may not reflect the competition of the transport by P-glycoprotein. We reported that cortisol, which scarcely inhibited azidopine photoaffinity labeling, was transported by P-glycoprotein, and that progesterone, which efficiently inhibited azidopine binding, was not transported by P-glycoprotein (18). Safa et al.
(23) also reported that efficiency in competing with azidopine photoaffinity labeling of mutant P-glycoprotein did not par-allel the efficiency in the transport by P-glycoprotein. These results suggest that the binding of a substrate to P-glycoprotein does not reflect the transport by P-glycoprotein. Cyclosporin A transport was not inhibited by verapamil or vinblastine in this study, although cyclosporin A and its nonimmunosuppressive analogues have been reported to increase intracellular accumulations of anticancer drugs efficiently (12,14,20). Cyclosporin A may interact with a key site of Pglycoprotein for transport with a high affinity. This may be one of the reasons that cyclosporin A is an efficient modulator of P-glycoprotein. Actually we found that cyclosporin A efficiently inhibited the excretion of digoxin, which is a substrate transported by P-glycoprotein (19), into urine in U~U O . ~ Because one of the physiological roles of P-glycoprotein is the secretion of metabolites and natural toxic substances into urine at the proximal tubules of kidney (19), the inhibition of functions of P-glycoprotein there may cause the nephrotoxicity of cyclosporin A (24,25). Cyclosporin A and FK506 were transported by human P-glycoprotein, and both agents were expelled from cells expressing P-glycoprotein and intracellular concentrations were markedly reduced. These results should be important for understanding the pharmacokinetics of both immunosuppressive agents.