Ligand-mediated Tertiary Structure Changes of Reconstituted P-glycoprotein

Ligand-dependent changes in accessibility of purified P-glycoprotein, functionally reconstituted in liposomes, were investigated by fluorescence measurements. Trp quenching experiments provided evidence that P-glycoprotein adopts different tertiary structures upon binding of drug substrates in the absence and presence of MgATP and its nonhydrolyzable analog, MgATPγS. Five anthracycline derivatives were tested as drug substrates: daunorubicin, 4′-epi-doxorubicin, iododoxorubicin, 4-demethoxy-daunorubicin, and methoxy-morpholino-doxorubicin. Among them, daunorubicin and 4′-epi-doxorubicin have been shown to be rejected outside the multidrug-resistant cells, whereas the three others have been shown to accumulate in multidrug-resistant cells overexpressing P-glycoprotein and therefore retain their cytotoxic activity. A small conformational change was associated with nucleotide binding and amplified after nucleotide hydrolysis. Different conformational states were adopted by P-glycoprotein upon the addition of the anthracycline derivatives in the absence and presence of MgATP or MgATPγS. These conformational changes are shown to be related to the nature of the antitumor agents and more precisely to their capacity to accumulate in resistant cells. These data also suggest that the cytotoxicity of iododoxorubicin and 4-demethoxy-daunorubicin is related to the fact they are not transported by P-glycoprotein. On the contrary, methoxy-morpholino-doxorubicin cytotoxicity may be explained in terms of its rapid reincorporation into the plasma membrane after being transported by P-glycoprotein.

with characteristic Walker motifs A and B. However, experimental studies concerning this proposed topology (2)(3)(4)(5) remain controversial. According to its sequence, P-glycoprotein is classified as a member of a large family of membrane transporters known as the ATP-binding cassette superfamily that includes yeast, bacteria, and mammalian transporters (6,7). P-glycoprotein is proposed to function as an ATP-driven efflux pump, transporting through the plasma membrane an unusually broad but well defined spectrum of structurally unrelated cytotoxic drugs, including the Vinca alkaloids, anthracyclines, epipodophyllotoxins, and taxanes (8 -10).
A large body of evidence suggests that the transmembrane domains of the P-glycoprotein participate in the recognition of substrates, whereas the ATP hydrolysis necessary for transport is carried out by both NBD regions with a similar efficiency in an alternating fashion (11)(12)(13)(14)(15)(16). The mechanism of coupling ATP hydrolysis at the two cytoplasmic nucleotidebinding sites to drug transport by the intramembrane drugbinding site(s) is likely to involve substantial conformational changes in the P-glycoprotein structure (17). Different tertiary conformational changes have previously been shown to take place upon addition of MgATP and MgATP-verapamil, an actively transported chemosensitizer (18). The comprehension of the mechanism of interaction between the drug-binding site(s) and the NBD domains is essential for understanding how Pglycoprotein transports its substrates. The aim of this study was to further investigate the different conformations adopted by the protein in the presence of nucleotide ligands and drugs.
Purified P-glycoprotein from CHO cells was reconstituted into lipid vesicles with an "inside-out" orientation, exposing its cytoplasmic region, which contains the NBD domains, to the external medium. The resulting proteoliposomes have previously been shown to exhibit both ATP-dependent drug transport and drug-stimulated ATPase activity (18).
Acrylamide quenching of Trp fluorescence was used to monitor ligand-dependent changes in the accessibility of reconstituted P-glycoprotein in the presence of MgATP, MgATP␥S, a nonhydrolyzable analog of MgATP, and of five anthracycline antitumor agents (daunorubicin, 4Ј-epi-doxorubicin, iododoxorubicin, 4-demethoxy-daunorubicin, and FCE) previously used in pharmacokinetics studies on multidrug-resistant and sensitive K562 cells (19). Three of these agents (iododoxorubicin, 4-demethoxy-daunorubicin, and FCE) have been shown to accumulate within multidrug-resistant cells overexpressing P-glycoprotein and to retain their cytotoxic activity. On the contrary, daunorubicin and 4Ј-epi-doxorubicin were rejected outside the multidrug-resistant cells. We analyze here how the P-glycoprotein conformational changes are related to the nature of the antitumor agents and, more precisely, to their capacity to accumulate into resistant cells.
ATP Hydrolysis-ATP hydrolysis was measured as described by Shapiro and Ling (20) in the absence and presence of increasing concentrations (2, 10, and 50 M) of each anthracycline derivative. Protein determination was performed according to Peterson (21).
Fluorescence Quenching Experiments-Acrylamide quenching experiments were carried out on a SLM Aminco 8000 fluorimeter at an excitation wavelength of 295 nm instead of 280 nm to reduce the absorbance by acrylamide. Control experiments established that no emission of the anthracycline derivatives occurred at 340 nm if excited at 295 nm (see Ref. 22 for excitation spectrum, ϳ390 -540 nm, of daunomycin). Acrylamide aliquots were added from a 3 M solution to the proteoliposome suspension (1 ml in water) containing 8 g of reconstituted P-glycoprotein and the various ligands. The final concentration was 3 mM for MgATP and MgATP␥S and 10 M for the various anthracycline derivatives. Fluorescence intensities were measured at 340 nm after each addition of quencher. All measurements were carried out at 25°C. Acrylamide quenching data were subjected to a linear fit up to 80 mM acrylamide. Above this concentration, the static quenching by acrylamide is responsible for the deviation from linearity in Stern-Volmer plots.
Azidopine Photolabeling-Plasma membrane vesicles were prepared from CH R B30 cells as described previously (23). The plasma membrane vesicles are typically about 50% inside-out, exposing the cytoplasmic ATP-binding sites, and are sealed (24). The vesicles were incubated in darkness at room temperature for 1 h with 0.3 M [ 3 H] azidopine and the indicated concentrations of anthracycline derivatives. After incubation, the samples were exposed to a UV lamp for 10 min while being kept on ice. Laemmli's buffer was added, and the samples were analyzed by SDS-polyacrylamide gel electrophoresis and by autoradiography of the dried gel.

RESULTS
Fluorescence Experiments-Fluorescence experiments were performed to investigate changes in P-glycoprotein structure that occur upon binding of drug and nucleotide substrates. Five anthracycline derivatives, cytotoxic (iododoxorubicin, 4-demethoxy-daunorubicin, and FCE) and noncytotoxic (daunorubicin and 4Ј-epi-doxorubicin) to resistant cells, were used (Fig. 1). These were added to a final concentration of 10 M to Pglycoprotein-containing proteoliposomes in the presence and absence of MgATP (3 mM) or MgATP␥S (3 mM), a nonhydrolyzable MgATP analog. This analog allows discrimination between the influence of nucleotide binding and nucleotide hydrolysis on P-glycoprotein structure. The exposure of P-glycoprotein Trp residues to the external solvent was subsequently determined by continuous monitoring of P-glycoprotein fluorescence intensity in the presence of increasing concentrations of acrylamide (0 -0.1 M), a neutral aqueous quencher. In the presence of drugs noncytotoxic to resistant cells (daunorubicin and 4Ј-epi-doxorubicin) known to be transported by P-glycoprotein, addition of MgATP resulted in the stabilization of the enzyme in a conformational state intermediate between that observed in the presence of MgATP alone and that observed upon binding of the five drugs (Fig. 4A). Replacement of MgATP by its nonhydrolyzable analog, MgATP␥S, had no effect on the levels of quenching observed (Fig. 4B). This indicates that nucleotide binding was necessary and sufficient to modify the accessibility of the protein to the water phase.
Upon addition of drugs cytotoxic to resistant cells (iododoxorubicin, 4-demethoxy-daunorubicin, and FCE) and MgATP or MgATP␥S, two distinct situations were observed (Fig. 5): 1) in the presence of iododoxorubicin and 4-demethoxy-daunorubicin, no fluorescence quenching was detected even after addition of MgATP or MgATP␥S and 2) in the presence of the morpholino derivative (FCE), no fluorescence quenching was detected after addition of MgATP␥S, but addition of MgATP led to substantial quenching of the protein fluorescence. In the presence of MgATP, the accessibility of the protein was identical to that observed with transported noncytotoxic molecules. Stern-Volmer constants were calculated in each case (Table I).
ATPase Activity Measurements in the Presence of Anthracycline Derivatives-Upon binding of anthracycline derivatives, except for the FCE derivative, P-glycoprotein adopted the same conformations when MgATP was replaced by its nonhydrolyzable analog MgATP␥S. Furthermore, MgATP had no influence on P-glycoprotein conformation when iododoxorubicin and 4-demethoxy-daunorubicin were bound to the protein, suggesting that the nucleotide could not access the ATP-binding sites in this conformation. P-glycoprotein ATPase activity was therefore measured in the absence and presence of each of the  anthracycline derivatives to determine whether ATP binding and hydrolysis still occurred. In the absence of anthracycline derivatives, an ATPase activity of 65 nmol/min/mg of protein was measured for the reconstituted P-glycoprotein. According to Table II, P-glycoprotein exhibited ATPase activity in the presence of all the derivatives tested, demonstrating that significant inhibition of ATP binding and hydrolysis did not occur when the drugs were bound to the protein.

Inhibition of [ 3 H]Azidopine Photolabeling of P-glycoprotein by Anthracycline
Derivatives-Azidopine, an analog of verapamil, is able to specifically label P-glycoprotein (25). The ability to inhibit photoaffinity labeling has frequently been used as an indicator of whether a particular drug interacts with Pglycoprotein (26,27). Therefore, to demonstrate the binding of the anthracycline derivatives to P-glycoprotein and to determine whether differences in azidopine competition could be identified between cytotoxic and noncytotoxic anthracycline derivatives, photolabeling experiments were performed on CH R B30 plasma membrane vesicles in which about 15% of the protein is P-glycoprotein. Fig. 6 shows the densitometric analysis of the inhibition of [ 3 H]azidopine labeling in the presence of increasing concentrations (10, 30, and 100 M) of each anthracycline derivative. Although all the compounds were able to compete with [ 3 H]azidopine labeling, the observed rate of inhibition varied as a function of the drug tested. In fact, at final concentrations of 30 and 100 M, the three cytotoxic compounds (iododoxorubicin, 4-demethoxy-daunorubicin, and FCE) exhibited a stronger inhibition of [ 3 H] azidopine labeling than noncytotoxic derivatives (daunorubicin and 4Ј-epi-doxorubicin). This suggests a higher affinity of these cytotoxic agents for P-glycoprotein or the binding of drugs to several P-glycoprotein sites. DISCUSSION Previous experiments, including infrared spectroscopy (18), enzymatic proteolysis (28,29), fluorescence labeling in the NBD domains (17), and immunoreactivity experiments (30), have suggested that P-glycoprotein may exist in different conformational states during its catalytic cycle. Our data provide strong evidence that the protein undergoes tertiary conformational changes depending on the nature of the ligands.
Our experiments demonstrated that upon addition of MgATP, the enzyme adopts a different tertiary structure, resulting in a significantly increased solvent accessibility, according to previous data (18). MgATP␥S, a nonhydrolyzable analog of MgATP, was used to investigate the structural changes associated with nucleotide binding. Our fluorescence measurements provide evidence that MgATP␥S binding increases, slightly but significantly, the accessibility of some P-glycoprotein domains. The structural change observed upon ATP bind-  ing (P-glycoprotein ϩ MgATP␥S) was much more pronounced when ATP hydrolysis occurred (P-glycoprotein ϩ MgATP).
In this study, we also demonstrated the influence of the binding of five anthracycline derivatives to P-glycoprotein. Binding of the five anthracycline derivatives to P-glycoprotein reduces accessibility of the protein to solvent. The enzyme therefore undergoes a first conformational change. However, in the presence of MgATP or MgATP␥S, addition of cytotoxic and noncytotoxic derivatives resulted in various degrees of Trp fluorescence quenching. In the presence of noncytotoxic derivatives (daunorubicin and 4Ј-epi-doxorubicin), binding of MgATP␥S was necessary and sufficient for the protein to undergo a conformational change. The protein was stabilized in a conformation state intermediate between the "opened" structure observed in the presence of MgATP alone and the "closed" structure observed upon binding of the drug. In the presence of the cytotoxic iododoxorubicin and 4-demethoxy-daunorubicin, neither ATP binding nor ATP hydrolysis were capable of modifying the tertiary structure of P-glycoprotein. The protein was maintained in the closed conformational state induced by the binding of drug derivative. However, the ATPase activity measurements described in this paper clearly indicated that ATP is still able to bind to P-glycoprotein and is hydrolyzed in the presence of the five anthracycline derivatives tested. In the presence of noncytotoxic agents, the protein probably undergoes a conformational change after ATP binding and hydrolysis, which might be an important step in the catalytic cycle of drug transport. In contrast, the cytotoxic derivatives iododoxorubicin and 4-demethoxy-daunorubicin probably inhibit one or more steps in the catalytic cycle of P-glycoprotein. This impairs their transport by the protein and results in an improved accumulation of these compounds within multidrug-resistant cells. In the presence of the cytotoxic FCE derivative, a third situation was observed: only hydrolysis of ATP increased accessibility of the protein. Binding of MgATP␥S was not sufficient to induce a conformational change. After hydrolysis of ATP, FCE behaves like transported noncytotoxic molecules, and this argues in favor of transport of FCE by P-glycoprotein. It follows that cytotoxicity of this derivative may be due to its large hydrophobicity, which would favor rapid reincorporation into the plasma membrane after transport. Such a property has already been proposed for verapamil, a well known chemosensitizer that is thought to trap P-glycoprotein in a futile cycle by rapid rediffusion through the membrane (31,32). Briefly, these data suggest that the step of the catalytic cycle stabilizing the P-glycoprotein depends on the nature of the drug.
The existence of such different conformational states could be due to the involvement of distinct binding sites for the anthracycline derivatives tested. Azidopine photolabeling experiments also suggest the possibility for the drugs to bind to different P-glycoprotein sites. These experiments, performed in the presence of varying amounts of each anthracycline derivative, revealed a higher ability of the three cytotoxic derivatives to compete with azidopine, suggesting a stronger affinity of the cytotoxic drugs for P-glycoprotein or the presence of several different drug-binding sites. Independent experimental data also support the hypothesis that P-glycoprotein contains several substrate sites (33). Shapiro and Ling (34) previously showed the existence of two different and positively cooperative sites for Hoechst 33342 and rhodamine 123 binding and transport. Furthermore, unpublished data 2 suggest the possibility of binding of cytotoxic and noncytotoxic anthracycline derivatives to distinct but overlapping P-glycoprotein domains. The initial rates of rhodamine 123 and Hoechst 33342 transport in Pglycoprotein-containing plasma membrane vesicles isolated from CH R B30 cells were followed in the presence of varying The gel was dried, amplified, and exposed to x-ray film (Kodak) for 1 day at Ϫ70°C. Approximately half of the sample was lost. concentrations of each anthracycline derivative by fluorescence monitoring. According to these data, noncytotoxic daunorubicin and 4Ј-epi-doxorubicin inhibit rhodamine 123 transport but stimulate Hoechst 33342 transport. Cytotoxic iododoxorubicin and FCE seem to bind to the Hoechst 33342 site as well as the rhodamine 123 site, whereas 4-demethoxy-daunorubicin binds more weakly to the Hoechst site. All together, these data suggest distinct binding sites for the cytotoxic and noncytotoxic drugs.
In conclusion, this study describes distinct P-glycoprotein conformations corresponding to different drug and ATP binding and hydrolysis states. Separate or simultaneous addition of MgATP and drug substrates give rise to distinct conformational changes in the P-glycoprotein molecule, confirming that coupling between the drug-binding site(s) and the catalytic sites for ATP hydrolysis occurs.
Interestingly, a difference in the protein comportment during these stages has been identified in the presence of cytotoxic and noncytotoxic derivatives. It is currently assumed that Pglycoprotein binds and transports substrates with a low specificity via a relatively nonspecific hydrophobic binding pocket (35). However, our results demonstrate that minor changes in the anthracycline structure cause major modifications in the conformational states adopted by the enzyme, suggesting that if P-glycoprotein effectively binds its substrates with a low specificity, the following steps in its catalytic cycle leading to transport are much more dependent on the drug structure.