Effects of a detergent micelle environment on P-glycoprotein (ABCB1)-ligand interactions

Abstract

P-glycoprotein (P-gp) is a multidrug transporter that uses energy from ATP hydrolysis to export many structurally dissimilar hydrophobic and amphipathic compounds, including anticancer drugs from cells. Several structural studies on purified P-gp have been reported, but only limited and sometimes conflicting information is available on ligand interactions with the isolated transporter in a dodecyl-maltoside detergent environment. In this report we compared the biochemical properties of P-gp in native membranes, detergent micelles, and when reconstituted in artificial membranes. We found that the modulators zosuquidar, tariquidar, and elacridar stimulated the ATPase activity of purified human or mouse P-gp in a detergent micelle environment. In contrast, these drugs inhibited ATPase activity in native membranes or in proteoliposomes, with IC₅₀ values in the 10–40 nm range. Similarly, a 30–150-fold decrease in the apparent affinity for verapamil and cyclic peptide inhibitor QZ59-SSS was observed in detergent micelles compared with native or artificial membranes. Together, these findings demonstrate that the high-affinity site is inaccessible because of either a conformational change or binding of detergent at the binding site in a detergent micelle environment. The ligands bind to a low-affinity site, resulting in altered modulation of P-gp ATPase activity. We, therefore, recommend studying structural and functional aspects of ligand interactions with purified P-gp and other ATP-binding cassette transporters that transport amphipathic or hydrophobic substrates in a detergent-free native or artificial membrane environment.

Introduction

P-glycoprotein (P-gp;³ ABCB1) is a member of the ATP-binding cassette (ABC) transporter superfamily. There are 48 known members in humans divided into seven classes denoted A through G (1). P-gp is a member of the B subfamily of ABC transporters, designated ABCB1, and is best characterized as a single polypeptide containing two symmetrical halves that are 65% homologous to one another (2, 3). Each half is composed of a transmembrane domain (TMD) containing six transmembrane helices and a cytoplasmic nucleotide-binding domain (NBD). The NBDs of P-gp consist of a catalytic core domain and an α-helical subdomain specific to ABC transporters. The catalytic cores of NBD1 and NBD2 each contain a Walker A, Walker B, and signature C motif that mediates ATP hydrolysis by the transporter.

We observed earlier that the interaction of modulators with their binding sites in the TMDs of human P-gp, as measured by the ligand's effect on ATP hydrolysis and modulator binding, depends on their interactions with amino acids present in the substrate-binding pocket of the protein (4). Specifically, we discovered that certain drugs such as zosuquidar, elacridar, and tariquidar that normally inhibit ATP hydrolysis mediated by human P-gp switch to stimulation of ATP hydrolysis when a pair of phenylalanine-tyrosine structural motifs in the substrate-binding pocket was mutated (4). Molecular modeling studies further led us to infer that formation of hydrogen bonds between these inhibitors and tyrosine residues of the structural motifs in the substrate-binding pocket determine the affinity of modulators. These studies suggested that ligand-amino acid interaction at the substrate-binding pocket of P-gp is a crucial determinant of the modulation of ATP hydrolysis by P-gp.

Structural studies on P-gp by X-ray crystallography (5–10), EM (11, 12), and with cryo-EM (13, 14) have been carried out using purified protein in n-dodecyl β-d-maltoside (DDM) micelles. In a majority of cases, high concentrations of substrates or modulators had to be used to observe their effects on the ATPase activity of purified transporter in detergent micelles. In addition, attempts to co-crystallize mouse or human P-gp in detergent solution with modulators including zosuquidar, tariquidar, and elacridar, which exhibit very high affinity in the 10–40 nm range in native membranes, have been unsuccessful.

In the present study we evaluated the effect of changing the normal lipid environment of purified P-gp to a detergent environment on the interaction of ligands and modulators with the transporter. We show that the detergent micelle environment used to purify the protein leads to a change from inhibition by high-affinity inhibitors to stimulation of ATP hydrolysis of P-gp due to their interaction at an alternate site with decreased affinity. Ligand binding affinity in the substrate-binding pocket is known to modulate the ATPase activity by influencing the release of ADP, a rate-limiting step in the catalytic cycle of P-gp (15). Similarly, there is a 30–150-fold decrease in the apparent affinity for verapamil and cyclic peptide inhibitor QZ59-SSS, which stimulate ATPase activity even in detergent micelles. Interestingly, the addition of excess phospholipid to the detergent-protein mixture alone does not overcome the detergent effect, but removal of detergent with Bio-Beads SM-2 in the presence of excess phospholipids and generation of proteoliposomes restores the apparent affinities of ligands for the transporter to similar levels as observed in native membranes. These findings in aggregate demonstrate that due to the inaccessibility of the high affinity site for interactions in a detergent micelle environment, ligands bind to a low affinity site, resulting in altered modulation of P-gp ATPase activity. We suggest that the structural and functional properties of ligand interactions with purified P-gp and other ABC transporters that transport amphipathic or hydrophobic substrates should be studied in a detergent-free artificial membrane (proteoliposomal or possibly nanodisc) environment.

Results

The detergent DDM was used for crystallization of mouse and Caenorhabditis elegans P-gp, and the ATPase activity of P-gp was measured in DDM micelles in the presence of added lipid (16, 17). Until now, a systematic comparison of the effect of substrates and modulators on the ATPase activity of P-gp in native membranes, after its purification in DDM micelles and after reconstitution in proteoliposomes or nanodiscs, had not been reported. In this report we characterize the ligand interactions with mouse and human P-gp (in the following sections, mouse P-gp is referred as mP-gp and human P-gp as hP-gp) in native membranes and after its purification in DDM micelles and reconstitution in nanodiscs and proteoliposomes under similar conditions to those used for cryo-EM and generation of crystals for X-ray crystallography studies.

Differential modulation of ATPase activity of P-gp by tariquidar, elacridar, and zosuquidar when present in native membranes or detergent micelles

To determine the effect of changing the environment from membranes to detergent micelles, ATP hydrolysis by mP-gp in native membranes and after its purification in DDM micelles in the presence of three known and well characterized inhibitors (tariquidar, elacridar, and zosuquidar) was measured as detailed under “Experimental procedures.” It was observed that whereas all three drugs inhibited the basal (in the presence of DMSO solvent only) ATP hydrolysis mediated by mP-gp in native membranes with IC₅₀ values in the low nanomolar range (Fig. 1, A–C), these inhibitors instead stimulated ATP hydrolysis of mP-gp in DDM-detergent micelles in a concentration-dependent manner with EC₅₀ values 0.51, 2.18, and 11.82 μm (Fig. 1, D–F). In addition, we also evaluated the effect of verapamil and QZ59-SSS, two compounds that are known to stimulate ATP hydrolysis by P-gp (18, 19), on the ATPase activity of purified mP-gp in DDM micelles. Both verapamil and QZ59-SSS also stimulated the ATP hydrolysis of mP-gp, as they do in native membranes (Fig. 2). However, the EC₅₀ values for stimulation of ATP hydrolysis were 150-fold and 30-fold higher in detergent micelles than in native membranes in the presence of verapamil (Fig. 2, A and B) and QZ59-SSS (Fig. 2, D and E), respectively. These data show that the affinity of compounds that stimulate the ATPase activity of P-gp is also decreased in detergent micelles.

Figure 1.

Effect of tariquidar, elacridar, or zosuquidar on ATP hydrolysis by mP-gp depends on its presence in either native membranes or detergent micelles. Mouse P-gp-expressing native membranes of High-Five insect cells (50–100 μg of protein/ml; panels A-C) or purified P-gp in gel filtration buffer (1.5–3 μg of protein; panels D–F) was incubated with the indicated concentrations of tariquidar (A and D), elacridar (B and E), and zosuquidar (C and F) at 37 °C for 5 min. The ATP hydrolysis was then determined by adding 5 mm ATP as described under “Experimental procedures.” The basal specific ATPase activity in native membranes (for panels A–C) and in detergent micelles (for panels D–F) ranged from 42 to 54 and 79 to 83 nmol of P_i/min/mg of protein, respectively. The basal activity in the presence of DMSO alone was subtracted, and the percent stimulation or inhibition of ATP hydrolysis (y axis) was plotted as a function of varying concentrations of the given modulator (x axis). The curves were fitted using GraphPad Prism 7.0 with nonlinear one-phase exponential decay or association analysis. The curves represent the mean ± S.D. from three independent experiments. The IC₅₀ and EC₅₀ values derived from the curves are given in Table 1.

Figure 2.

Effect of verapamil and QZ59-SSS on the ATP hydrolysis of P-gp in native membranes, detergent micelles, or proteoliposomes. Mouse P-gp-expressing membranes from High-Five insect cells (A and D), concentrated purified mP-gp in gel filtration buffer (detergent micelles) (B and E), or proteoliposomes reconstituted with purified P-gp (C and F) were incubated with the indicated concentrations of verapamil (A–C) and QZ59-SSS (D–F) at 37 °C for 5 min. The ATP hydrolysis was then initiated by adding 5 mm ATP as described under “Experimental procedures.” The basal-specific ATPase activity in native membranes (panels A and D) and in detergent micelles (panels B and E) and proteoliposomes (panels C and F) ranged from 42 to 54, 79 to 83, and 716 to 730 nmole of P_i/min/mg of protein, respectively. The percent stimulation of the ATP hydrolysis (y axis) was plotted as a function of varying concentrations of the compound (x axis). The curves were fitted using GraphPad Prism version 7.0 with nonlinear one-phase association analysis. The curves represent the mean ± S.D. values from three independent experiments. The EC₅₀ values derived from the curves are given in Table 1.

Significant decrease in the ability of tariquidar to inhibit photolabeling of purified mP-gp with ¹²⁵Iodoarylazidoprazosin (IAAP) in detergent micelles

Based on the above data, it appeared that purified mP-gp in detergent micelles has properties that are different when the transporter is present in native membranes. There is a possibility that the conformation of the substrate-binding pocket of the protein is altered in a detergent environment, thereby resulting in a change of behavior when interacting with ligands. To further study this change, we evaluated the ability of tariquidar, a representative inhibitor from this class, to inhibit the binding of another ligand, IAAP, which is transported by P-gp (20). As shown in Fig. 3A, although pretreatment of the native membranes with tariquidar in a concentration-dependent manner inhibited the photolabeling of mP-gp with IAAP with IC₅₀ 0.1 ± 0.01 μm, there was only partial inhibition (40–45%) even at a 10 μm drug concentration of the binding of IAAP to purified mP-gp in DDM micelles (Fig. 3B), with apparent IC₅₀ > 10 μm. Thus, photolabeling with IAAP provided further evidence for the significantly decreased affinity of tariquidar for purified mP-gp in detergent micelles.

Figure 3.

Tariquidar's ability to inhibit photocross-linking of mP-gp with IAAP was significantly decreased in detergent micelles. Native membranes from mP-gp-expressing High-Five insect cells (25–50 μg of protein) (A) or purified mP-gp (2–5 μg of protein) (B) were incubated with increasing concentrations of tariquidar at room temperature for 5 min followed by photocross-linking with ¹²⁵I-AAP (5 nm) using 366-nm UV light as described under “Experimental procedures.” The photocross-linked P-gp samples were separated on 7% Tris acetate SDS-PAGE gels. The gels were dried and exposed to X-ray films. Representative autoradiograms of IAAP-labeled P-gp bands in native membranes (panel A) and detergent micelles (panel B) are shown at the top. The concentration of tariquidar (Tar) is given above each lane. The curves of IAAP incorporation (% of control) as a function of tariquidar concentration were plotted using GraphPad Prism 7.0 with nonlinear one-phase exponential decay analysis. Similar results were obtained in three independent experiments.

Purified hP-gp in detergent micelles has ATP hydrolysis properties similar to those of mP-gp

As the above studies were done with mP-gp, we wanted to investigate its relevance to hP-gp, which shares 87% sequence identity and 93% sequence similarity with mouse mdr1a and has functional similarities. However, by X-ray crystallography, resolution of the structure of hP-gp has been proven to be a very challenging task. hP-gp was purified from High-Five cell membranes as described under “Experimental procedures” in the presence of DDM, and the effect of the inhibitors tariquidar, elacridar, and zosuquidar on its ATPase activity in DDM micelles was evaluated. Similar to mP-gp, these inhibitors also showed a concentration-dependent stimulation of ATP hydrolysis, with EC₅₀ values of 6.68 μm (tariquidar), 8.86 μm (elacridar), and 10.22 μm (zosuquidar), respectively (Fig. 4, A–C). Therefore, stimulation of ATP hydrolysis-mediated by hP-gp by these modulators in detergent micelles was the opposite effect as observed in native membranes (4). Stimulation of the ATPase activity of purified hP-gp up to 10-fold by tariquidar in DDM micelles in the presence of lipids was also observed earlier (21).

Figure 4.

Effect of tariquidar (A), elacridar (B), and zosuquidar (C) on purified hP-gp-mediated ATP hydrolysis. Purified and concentrated hP-gp in gel filtration buffer with ∼67.5× CMC DDM (2.5–3.5 μg of protein) was incubated with the indicated concentrations of the inhibitors at 37 °C for 5 min. ATP hydrolysis mediated by purified hP-gp in the presence of these modulators was determined as described under “Experimental procedures.” The basal specific ATPase activity was in the range of 70–80 nmole of P_i/min/mg of protein. The analysis was done as described in the legend to Fig. 1. Shown here are graphs with mean ± S.D. values of three independent experiments. The EC₅₀ values derived from the curves were 6.68 ± 1.32, 8.86 ± 1.53, and 10.22 ± 2.75 μm for tariquidar, elacridar, and zosuquidar, respectively.

Addition of excess phospholipids to purified P-gp in detergent micelles did not change ATP hydrolysis properties

As stated above, when P-gp is in a detergent micelle environment, its interaction with inhibitors is altered as well as its apparent affinity for verapamil and the cyclic peptide inhibitor QZ59-SSS. Therefore, we sought to determine whether replacing or diluting the detergent by adding excess phospholipids would lead to a change in its biochemical properties. In this regard, sonicated Escherichia coli total phospholipid mixture was added to purified mP-gp and hP-gp in excess concentration (P-gp: phospholipids 1:25 w/w; 1:4208 mol/mol ratio). As shown in Fig. 5, A and B, the inhibitors tariquidar, elacridar, and zosuquidar at 10 μm concentration still stimulated the ATP hydrolysis mediated by both mP-gp and hP-gp. These observations suggest that due to the use of high concentrations of DDM (40 mm or 333× critical micelle concentration (CMC) during solubilization and 1.35 mm or 11.25× CMC during purification), the addition of phospholipids did not dilute the detergent sufficiently or that the primary (high-affinity) binding site of P-gp was occupied by DDM-detergent molecules, and therefore, the addition of phospholipids could not displace DDM. Furthermore, when we increased the P-gp:phospholipid ratio to 1:100 (w/w) or 1:16,832 (mol/mol) in the ATPase assay, similar stimulation of ATPase activity was observed (data not shown), suggesting the latter possibility is most likely.

Figure 5.

Effect of the addition of excess E. coli total phospholipids to purified protein in a detergent buffer on ATP hydrolysis mediated by mP-gp and hP-gp. Sonicated E. coli total phospholipids were added to purified mP-gp (A) or hP-gp (B) in a 25:1 w/w ratio to purified P-gp in gel filtration buffer. The reaction mixture was incubated at 37 °C for 10 min followed by the addition of 10 μm tariquidar, elacridar, and zosuquidar as indicated. The ATP hydrolysis was evaluated as described under “Experimental procedures.” The histograms show the ATPase-specific activity (y axis) in the presence of these modulators (x axis). Mean ± S.D. values from three independent experiments are given.

Reconstitution of purified mP-gp into proteoliposomes restores the inhibitory effect of tariquidar and elacridar on ATPase activity

The above observations demonstrated that the addition of excess phospholipids to the detergent micelles did not restore the ligand-P-gp interactions to those observed in native membrane environment. A well established method of replacing DDM is by forming proteoliposomes through slow removal of detergent in the presence of Bio-Beads SM2 and phospholipids. We therefore reconstituted mP-gp into proteoliposomes using an E. coli acetone-ether-washed phospholipid mixture, as described under “Experimental procedures.” The resulting proteoliposomes were collected by ultracentrifugation and used for ATPase assays, as detailed under “Experimental procedures.” The vanadate-sensitive ATP hydrolysis of the reconstituted mP-gp was then evaluated in the presence of tariquidar and elacridar. As shown in Fig. 6, A and B, when the detergent was removed by Bio-Beads SM2 in mP-gp proteoliposomes, both tariquidar and elacridar inhibited ATP hydrolysis in a concentration-dependent manner similar to their observed effect in native membranes (Fig. 1). It should be noted that the IC₅₀ values of these inhibitors for inhibition of ATP hydrolysis in proteoliposomes were comparable with those observed in native membranes (Table 1). We further evaluated the effect of tariquidar on the photolabeling of mP-gp with IAAP in proteoliposomes (Fig. 6C). As observed in native membranes, tariquidar at 1 μm in this case inhibited >90% of the photolabeling of mP-gp with IAAP, demonstrating that replacing DDM micelles with membrane lipid bilayer allows tariquidar access to its high-affinity-binding pocket (Fig. 6C).

Figure 6.

Interaction of mP-gp with inhibitors in proteoliposomes is similar to that in native membranes. A and B, tariquidar and elacridar inhibit basal ATPase activity of purified mP-gp after reconstitution in proteoliposomes. Purified mP-gp was reconstituted into proteoliposomes using E. coli total phospholipids as described under “Experimental procedures.” Proteoliposomes containing 2 μg of purified P-gp were incubated with indicated concentrations of tariquidar (A) or elacridar (B), and the ATP hydrolysis was measured as described under “Experimental procedures.” The specific basal ATP activity ranged from 716 to 730 nanomoles P_i/min/mg of protein. The percent inhibition of basal activity was plotted (y axis) as a function of concentration of the inhibitors (x axis) as described in the legend to Fig. 1. Values shown represent the mean ± S.D. from at least three independent experiments. The IC₅₀ values derived from the curves are given in Table 1. C, inhibition of IAAP incorporation in P-gp by tariquidar. Proteoliposomes reconstituted with mP-gp (2.5–5 μg of protein) were incubated with DMSO (−) or 1 μm tariquidar (+) for 5 min at room temperature. The samples were photocross-linked with IAAP and separated on a 7% Tris acetate SDS-PAGE gel, and the autoradiogram was developed as described in the legend to Fig. 3. Shown here is a representative autoradiogram from three independent experiments.

Table 1.

Summary of IC₅₀ and EC₅₀ values of tariquidar, elacridar, zosuquidar, verapamil, and QZ59-SSS for ATP hydrolysis of mouse P-gp in native membranes, DDM micelles, and proteoliposomes

The ATPase activity was measured in native membranes, DDM micelles, and proteoliposomes as described under “Experimental procedures.” These data are shown as graphs in Figs. 1, 2, and 6. Values represent the mean ± S.D. from three or more independent experiments. ND = not determined.

Compound	P-gp in native membranes	Purified P-gp in DDM micelles	Purified P-gp in proteoliposomes
	μm	μm	μm
Tariquidar	IC50 = 0.01 ± 0.01	EC50 = 0.51 ± 0.07	IC50 = 0.04 ± 0.01
Elacridar	IC50 = 0.02 ± 0.01	EC50 = 2.18 ± 0.40	IC50 = 0.04 ± 0.01
Zosuquidar	IC50 = 0.19 ± 0.06	EC50 = 11.82 ± 2.34	ND
Verapamil	EC50 = 2.86 ± 0.05	EC50 = 421.8 ± 93.30	EC50 = 1.13 ± 0.18
QZ59-SSS	EC50 = 0.07 ± 0.01	EC50 = 2.08 ± 0.48	EC50 = 0.04 ± 0.01

Removal of detergent by reconstitution of P-gp in nanodiscs also restores inhibition of ATP hydrolysis by tariquidar and elacridar

Membrane proteins including ABC transporters can also be reconstituted in a membrane-like environment using a combination of membrane scaffold protein (MSP) and lipids, thereby forming nanodiscs (22). In addition, nanodiscs offer an advantage of accessibility to both extracellular and intracellular regions of the transporter. The resultant nanodiscs thus keep membrane proteins in a defined native-like phospholipid bilayer environment with ligand access from both cytoplasmic and extracellular sides, which provides stability to the protein compared with detergent micelles. In view of the above advantages of nanodiscs over proteoliposomes, mP-gp was reconstituted in nanodiscs as described under “Experimental procedures.” Purified mP-gp in DDM micelles and nanodiscs reconstituted with purified protein was resolved on 7% Tris acetate gel and stained with InstantBlue. As shown in Fig. 7A, lane 2, the nanodiscs fraction contains both mP-gp and MSP protein. The effect of tariquidar and elacridar on the ATP hydrolysis of mP-gp reconstituted in nanodiscs was then evaluated. As expected, contrary to their effect on mP-gp in detergent micelles, tariquidar and elacridar inhibited the ATP hydrolysis of mP-gp in nanodiscs (Fig. 7, B and C). However, the maximum inhibition was observed to be only 60–65%, and the IC₅₀ values of 1.55 ± 0.38 and 1.18 ± 0.22 μm for tariquidar and elacridar were significantly higher compared with those observed with proteoliposomes (Fig. 6), indicating that the affinity for these drugs is lower in nanodiscs. Because we used the same E. coli acetone-ether-washed phospholipid mixture for reconstitution of purified P-gp in nanodiscs and proteoliposomes, the difference is not due to the lipid. Although it is likely that the MSP belt protein influences the access of tariquidar to its binding site, we do not have any definitive explanation at present for the observed differences. However, it is very clear that the inhibitory effect of tariquidar and elacridar on ATPase activity is recovered in nanodiscs.

Figure 7.

Effect of tariquidar and elacridar on the ATPase activity of mP-gp reconstituted in nanodiscs. A, SDS-PAGE of purified mP-gp in DDM micelles and after its reconstitution in nanodiscs. mP-gp-containing nanodiscs with membrane scaffold protein (MSP1E3D1) were prepared as described under “Experimental procedures.” Purified mP-gp in DDM micelles (lane 1; 0.5 μg of protein) and in nanodiscs (lane 2; 0.5 μg of protein) was resolved on 7% Tris acetate gel and stained with InstantBlue (Coomassie-based sensitive stain was from Expedeon Inc., San Diego, CA). Please note that P-gp and MSP, due to their hydrophobic nature, travel to lower positions instead of traveling to the normal positions corresponding to their true molecular sizes. B and C, ATP hydrolysis mediated by mP-gp in these nanodiscs was measured in the presence of indicated concentration of tariquidar (B) and elacridar (C) as described under “Experimental procedures.” The basal ATPase activity in nanodiscs was in the range of 298–322 nanomoles of P_i/min/mg of protein. The percent inhibition of ATP hydrolysis (considering basal activity as 100%) and IC ₅₀ values were calculated as described in the legend to Fig. 1. The IC₅₀ values for inhibition by tariquidar and elacridar were 1.55 ± 0.38 and 1.18 ± 0.22 μm, respectively.

Molecular modeling studies

Docking studies of tariquidar and DDM at the substrate-binding pocket of P-gp were carried out to find an explanation at the molecular level for the remarkable change in the biochemical behavior of P-gp in lipid membranes versus detergent micelles. As previously reported, the program AutoDock Vina finds many high score poses for tariquidar in the substrate-binding pocket of hP-gp using a homology model (4). The 10 poses with the best scores for hP-gp are shown in Fig. 8A. The same docking analysis of tariquidar was carried out for mP-gp using the 3.4 Å resolution X-ray structure 4Q9H.pdb (9); the poses with the highest scores are shown in Fig. 8B.

Figure 8.

Docking of tariquidar and DDM in the binding pocket of P-gp. Exhaustive ligand docking in a homology model of hP-gp based on mP-gp structure 4M2T.pdb (A) and in the mP-gp structure 4Q9H.pdb (B) was carried out using the AutoDock Vina program with a receptor grid centered at the position of the QZ59-RRR molecule, with flexible side chains (26 residues) and a search box of dimensions 40 Å × 35 Å × 35 Å. The structure 4Q9H was previously aligned to 4M2T to use the same receptor grid in both cases. The first 10 modes with the highest docking scores (given in the figure) were clustered and shown as blue (tariquidar) and magenta (DDM) stick models. N-terminal and C-terminal domains are shown in different colors (green and gray for PDB code 4M2T and light green and gray for 4Q9H, respectively). TM helices 10 and 12 are not shown for clarity. The figure was prepared with PyMOL 1.4.1.

Docking of DDM was also carried out in hP-gp and mP-gp using the same models and program settings. The detergent poses with the highest scores are shown in Fig. 8 on the right for both models along with the docking of tariquidar and scores on the left. These modeling studies suggest that both ligands, tariquidar and DDM, bind to P-gp at the same site when tariquidar and DDM poses are superimposed (compare positions of tariquidar and DDM in panels A and B). The docking scores of DDM (−7 to −9 kcal/mol) were numerically higher than the scores of tariquidar (−11 to −13 kcal/mol), reflecting the fact that tariquidar binds P-gp with higher affinity than DDM. It is very likely that due to a high concentration of detergent used to solubilize the membranes and to maintain the purified protein in soluble form, DDM occupies the binding site and cannot be displaced by tariquidar.

One of the poses shown in Fig. 8A fully matches the mutagenesis and biochemical data of the interaction of tariquidar with hP-gp, as studied in isolated native membranes (4). This pose of tariquidar shown in Fig. 9, panel A (native membranes), represents the site where tariquidar binds and inhibits the basal ATP hydrolysis of hP-gp. As schematically represented in Fig. 9B, tariquidar thus binds to a suggested secondary or alternative low affinity site (compare IC₅₀ and EC₅₀ values in Table 1), where it produces stimulation of the basal ATP hydrolysis of P-gp. The residues lining the secondary low affinity site are not known at present. The presence of DDM surrounding the hydrophobic region, including the drug-binding pocket of hP-gp, is supported by cryo-EM studies with hP-gp (13), where DDM micelles are clearly shown to surround the transmembrane region of P-gp, as depicted in Fig. 9B.

Figure 9.

Schematic of binding of tariquidar in the substrate-binding pocket of P-gp in membranes and detergent micelles. A, representation of tariquidar binding that leads to inhibition of the basal ATPase activity of P-gp in native membranes. The figure shows the docking of tariquidar in a homology model of hP-gp based on mP-gp X-ray structure 4M2T (4). B, schematic representation of the binding of tariquidar at a different site when the preferred site is occupied by DDM. Based on cryo-EM data, the DDM molecules surround the hydrophobic transmembrane area of P-gp (13). The consequent binding of tariquidar to another (low-affinity) site leads to stimulation of the basal ATP hydrolysis of P-gp in detergent micelles (Fig. 1D). The residues lining this site are not yet known. When P-gp is reconstituted in proteoliposomes, tariquidar binding leads to inhibition of the basal ATP hydrolysis similar to that observed with native membranes (panel A). The N-terminal half of P-gp is shown in green, and the C-terminal half is in gray. Transmembrane helices 10 and 12 were removed to better visualize the substrate-binding pocket. Tariquidar (blue) and DDM (magenta) are shown as spheres. The position of the lipid bilayer (insect cell membrane in A) is determined by the visible hydrophobic region of the protein surface. The structures were created with PyMOL version 1.4.1.

Discussion

ABC transporters represent one of the largest transport protein families, with diverse physiological functions that are clinically important for certain disease conditions. Therefore, understanding the structure and function of these transporters in their native environment is critical for advancing our understanding of their physiological and pathological roles. The approach adopted to study the structure of the proteins has always been to crystallize the proteins after purification. Eukaryotic ABC transporters are normally expressed in heterologous systems including Pichia- and baculovirus-infected insect cells and purified using a high concentration of detergent, significantly above the CMC, the concentration above which micelles are formed spontaneously to maintain the protein in active conformations. These conditions are optimal to obtain structural information but do not allow the protein to remain in its native membrane-like environment.

As both hP-gp and mP-gp recognize and transport a variety of chemically dissimilar amphipathic or hydrophobic compounds including toxins and anticancer agents (2, 23, 24), and a number of detergents have been reported to interact as ligands or “allocrits” with this multidrug transporter when tested in native membranes (21, 25, 26), we investigated the interactions of selected modulators and substrates with purified mP-gp and hP-gp in the presence of DDM micelles. In addition, we compared the differences in the activity of these transporters when they are present in native or artificial membranes and in detergent micelles.

We demonstrated that altering the membrane environment of P-gp has profound effects on its biochemical properties. Although tariquidar, elacridar, and zosuquidar inhibit P-gp-mediated ATP hydrolysis when the protein is present in native membranes, the same inhibitors actually stimulate the ATP hydrolysis mediated by purified protein in detergent micelles (Fig. 1). This change in interaction with the ligands was also demonstrated by tariquidar's significantly decreased ability (at least 100-fold lower compared with native membranes) to compete for photoaffinity labeling of purified mP-gp with IAAP (Fig. 3).

Our observations with modulators and stimulators of ATPase activity (verapamil and QZ59 SSS), which also exhibit significantly reduced the affinity for mP-gp in detergent micelles (Fig. 2 and Table 1), clearly show that somehow a detergent environment alters the conformation of the substrate-binding pocket of the transporter, thereby altering the access of inhibitors or substrates to the pocket. One likely possibility is that the DDM molecules occupy the primary substrate-binding site in purified P-gp, thereby blocking access of tariquidar to its preferred binding site. Tariquidar then binds to a secondary site with altered (decreased) affinity, resulting in a switch from inhibition to stimulation of ATPase activity of P-gp. A similar switch from inhibition to stimulation of hP-gp-ATPase activity in native membranes by tariquidar, elacridar, and zosuquidar was recently observed by us when the three residues Tyr-953, Gln-725, and Tyr-307 were substituted with alanine in the substrate-binding pocket (4). We found that substitution of two tyrosine residues and one glutamate residue with alanine disrupts the high-affinity interaction of these three modulators with P-gp, which is similar to our observations reported here with purified wild-type P-gp in a DDM micelle environment. In addition, P-gp has been shown to interact with alkyl phospholipids, edelfosine, and ilmofosine (27), suggesting that lipids or detergents can occupy the substrate-binding site in P-gp. Direct proof of these observations was provided in a recent study that showed lipid and/or detergent molecules in what may possibly be the substrate-binding pocket of mP-gp using mass spectrometry analysis (28). Thus, lipid molecules could interact with the mP-gp in the substrate-binding pocket as well as at the interface between the protein and the lipid bilayer. Occupation of the substrate-binding pocket by lipids might also explain the basal ATPase activity observed for P-gp, as has been reported earlier (29). In addition, several point mutations in the substrate-binding pocket of P-gp were found to result in a significant decrease in the basal ATPase activity (30).

X-ray crystallographic studies have also shown that detergent molecules can bind in the central cavity of P-gp (7) in a similar way to what docking studies of DDM in hP-gp and mP-gp have indicated (Fig. 8). The X-ray structure of C. elegans P-gp revealed two molecules of the detergent undecyl 4-O-α-d-glucopyranosyl-1-thio-β-d-glucopyranoside, with their disaccharide head groups inside the central cavity of the protein (7). One of these detergent molecules is in close proximity to some of the DDM poses generated in our molecular modeling studies, thereby providing strong evidence that detergent molecules can occupy the substrate-binding pocket of P-gp. Additionally, the presence of two molecules of another detergent, nonyl-glucopyranoside, was reported in the ligand-binding cavity of the antibacterial peptide transporter McjD (31), and DDM was found bound to the transmembrane helices and a portal between transmembrane helices 1 and 2 in the X-ray crystal structure of the mitochondrial ABCB10 transporter (32).

Based on our experimental and docking results, we propose a model schematically shown in Fig. 9. When a high affinity ligand such as tariquidar binds to the substrate-binding pocket of P-gp in the native membrane environment, it inhibits ATP hydrolysis. In the case of purified protein, the detergent, which is used at very high concentrations during membrane solubilization and purification, may occupy the high-affinity site in the substrate-binding pocket. Therefore, tariquidar cannot access its primary binding site in the substrate-binding pocket. Instead, it binds to an alternative site (referred to as a secondary site; Ref. 30), resulting in concentration-dependent stimulation of ATP hydrolysis. Such modulation of P-gp ATP hydrolysis by detergents in the absence of any ligand was studied earlier by Orlowski et al. (33). Their work showed that detergents present at concentrations above their CMC during purification had a significant effect on the ATP hydrolysis of P-gp. Their report, however, did not address the effect of ligands in the detergent micelle environment on the ATP hydrolysis of the protein. Our data show that the substrate-binding pocket of purified P-gp when present in DDM micelles does not represent a native conformation (Fig. 9B). The native conformation of this pocket is restored when P-gp is reconstituted into proteoliposomes with similar properties as in native membranes (Fig. 9A), where the detergent molecules are replaced by lipids. This is demonstrated by restoration of tariquidar's ability to inhibit the ATP hydrolysis by P-gp and inhibition of IAAP binding to the substrate-binding pocket of the transporter. We further demonstrated the importance of lipids in maintaining the native state of P-gp and its interaction with ligands by incorporating P-gp into nanodiscs. As shown in Fig. 7, B and C, tariquidar was able to inhibit the ATPase activity of the purified protein in nanodiscs, although with a lower affinity.

Clearly, studying ligand interactions using purified P-gp in DDM micelles and excess phospholipids might not provide information about high-affinity binding sites in the substrate-binding pocket of this transporter. Loo et al. (21) first reported stimulation of purified P-gp ATPase activity by tariquidar in the presence of DDM micelles and lipids, but this observation was not consistent with other published work using native membranes. They also mapped tariquidar's binding sites (34) on P-gp, which are most likely low affinity binding sites, as the ATPase activity of purified mutant proteins was measured in a buffer containing DDM micelles and lipids. Our findings thus demonstrate that it is critical to study ligand interactions with P-gp in a detergent-free environment using native or artificial membranes. Consistent with our results, significant differences in conformations and biophysical properties of the bacterial MsbA transporter in DDM micelles and nanodiscs were recently reported (35).

In summary, our data demonstrate that the environment of the substrate-binding pocket of P-gp determines its interaction with various substrates, and the preferred binding sites for ligands in a detergent micelle environment are not accessible. The structural information about the ligand binding sites obtained with protein in DDM micelles by X-ray crystallography alone should be interpreted with caution. Similarly, the recently proposed role of individual helical movements in the determination of the polyspecificity of mP-gp (6) will have to be evaluated again using purified protein reconstituted in artificial lipid membranes. The inaccessibility of high-affinity sites on purified P-gp for modulators and substrates in detergent micelles can also explain the failure to co-crystallize these transporters with high-affinity modulators or substrates. We suggest that new methods to keep P-gp and other ABC drug transporters in a detergent-free environment (much closer to that found under physiological conditions) are required for correlation of structural, biochemical, and biophysical studies.

Experimental procedures

Materials

Anagrade DDM and other detergents were obtained from Anatrace (Maumee, OH). E. coli acetone-ether-washed phospholipids were purchased from Avanti Polar Lipids, Inc. (Alabaster, Al), and nickel-nitrilotriacetic acid resin was from (Qiagen Inc., Valencia, CA). IAAP (2200Ci/mmol) was obtained from PerkinElmer Life Sciences. Bio-Beads-SM-2 were obtained from Bio-Rad. All other chemicals were purchased from Sigma.

Expression of human and mouse P-gp in baculovirus-infected High-Five insect cells and preparation of membranes

High-Five insect cells (Thermo Fisher Scientific, Waltham, MA) were infected with baculovirus carrying wild-type hMDR1 (36) or M107L mmdr1a (37) with a 6-histidine tag at the C-terminal end, and total membranes from infected cells were prepared after hypotonic lysis, as described previously (15), with minor modification. Aliquots of membranes were quickly frozen in dry-ice and stored at −80 °C.

Purification of hP-gp and mP-gp

The protein was purified as described earlier (13). Briefly, the membranes (150–250 mg of protein) were solubilized by using 2% DDM in solubilization buffer (50 mm Tris-HCl (pH 7.5), 200 mm NaCl, 15% glycerol, 5 mm β-mercaptoethanol, 20 mm imidazole, and UltraCruz EDTA-free protease inhibitor mixture tablets (Santa Cruz Biotechnology, Dallas, TX)). The supernatant was separated from unsolubilized membranes by centrifugation at 38,000 rpm (Ti-45 rotor; Beckman Coulter, Brea, CA) for 45 min at 4 °C. The supernatant containing solubilized protein was incubated with nickel-nitrilotriacetic acid resin (Qiagen) pre-equilibrated with column buffer (50 mm Tris-HCl (pH 7.5), 200 mm NaCl, 15% glycerol, 5 mm β-mercaptoethanol, 20 mm imidazole, 0.04% sodium cholate, 0.0675% DDM) for 14–16 h at 4 °C. The metal resin was washed 5× with column buffer, and P-gp was eluted with the same buffer containing 300 mm imidazole. The eluted protein was concentrated and further purified using size exclusion chromatography using Superdex 200 10/300 GL in gel filtration buffer (50 mm Tris-HCl (pH 7.5), 2% glycerol for mP-gp or 15% for hP-gp, 200 mm NaCl, 5 mm β-mercaptoethanol, 0.0675% DDM, 0.04% sodium cholate). The fractions containing monomer P-gp were collected and concentrated using Amicon 100-KDa concentrators (Thermo Fisher Scientific). 10 mm DTT was added to the concentrated protein fraction before quick-freezing in dry-ice and storage at −80 °C.

Reconstitution of P-gp in nanodiscs and proteoliposomes

Purified hP-gp or mP-gp was reconstituted into nanodiscs using a recombinant membrane scaffold protein (MSP1E3D1) and E. coli total phospholipids, as described earlier with minor modifications (38). Briefly, purified hP-gp or mP-gp was mixed with purified MSP1E3D1 (a gift from Van Que and Andrew Stephen, Leidos Inc., Frederick, MD) and sonicated E. coli phospholipids in 20 mm sodium cholate in a molar ratio of 1:10:110 (P-gp:MSP1:phospholipids). The mixture was then incubated at room temperature with constant agitation for 1 h. To initiate removal of DDM and assembly of nanodiscs, Bio-Beads SM-2 (0.6 g/ml) were added and incubated at 4 °C for 12–14 h with constant agitation. Reconstituted nanodiscs were separated from the Bio-Beads with a 25-gauge needle and further purified with size exclusion chromatography using Superdex 200 10/300 GL in a buffer (50 mm Tris-HCl, pH 7.5, 200 mm NaCl, and 5 mm DTT). The peak fractions were collected, concentrated with 100-kDa cut-off concentrators, and stored at 4 °C until used. In addition, empty nanodiscs (without P-gp) were also prepared and employed as controls in protein estimation and ATP hydrolysis assays.

For reconstitution in proteoliposomes, the concentrated purified P-gp was combined with a sonicated E. coli acetone-ether-washed phospholipid mixture in the ratio of P-gp:phospholipids (1:50 (w/w); 1:8559 mol/mol ratio). The reconstitution mixture was then incubated at 21–22 °C for 30 min. The detergent was removed by incubating the mixture with prewashed Bio-Beads at a concentration of 0.6 g/ml, and the mixture was kept on a rotating shaker at 4 °C for 14–16 h. After removing Bio-Beads, P-gp proteoliposomes were collected by ultracentrifugation at 140,000 × g for 1 h at 4 °C and resuspended in buffer containing 50 mm Tris-HCl (pH 7.5), 200 mm NaCl, and 5 mm DTT. Total protein in membranes, concentrated fractions, nanodiscs. and proteoliposomes was quantified using the Schaffner and Weissmann method (39).

ATPase assays

The ATPase activity of P-gp in High-Five insect cell membranes, in purified protein in DDM micelles, and after its reconstitution in nanodiscs or proteoliposomes was measured by the end point P_i release assay, as described earlier with minor modifications (40). P-gp specific activity was measured as vanadate-sensitive ATPase activity. Native membrane protein (4.5 μg), purified P-gp in DDM micelles (1.5–3 μg) or reconstituted P-gp in nanodiscs or proteoliposomes (1–3 μg) per 100 μl, was preincubated with the indicated inhibitor at 37 °C for 5 min in the presence and absence of sodium orthovanadate (0.3 mm) in a 50 mm MES-Tris (pH 6.8) buffer containing 50 mm KCl, 10 mm MgCl₂, 5 mm NaN₃, 1 mm EGTA, 1 mm ouabain, and 2 mm DTT. The reaction was initiated by the addition of 5 mm ATP and quenched with SDS (final concentration, 5%); the amount of inorganic phosphate released over 20 min for native membranes or reconstituted P-gp and over 10 min for purified P-gp in DDM solution at 37 °C was measured using a colorimetric method as previously described (40).

Photolabeling of P-gp IAAP

The native membranes of High-Five insect cells expressing mP-gp (1.0 mg/ml), purified P-gp in DDM micelles (25–50 μg/ml), and reconstituted P-gp proteoliposomes (25–50 μg/ml) were incubated at room temperature with the indicated concentrations of the inhibitor for 5 min. IAAP (5–7 nm) was later added, and the reaction mixture was further incubated for 5 min under subdued light. The samples were photocross-linked with 366-nm UV light for 10 min at room temperature followed by electrophoresis and quantification as described previously (41).

Docking of tariquidar and DDM in the drug-binding pocket of P-gp

Docking of tariquidar in hP-gp has been previously reported (4). Docking of tariquidar in mP-gp and DDM in hP-gp and mP-gp was carried out in a similar fashion. The binding of DDM in hP-gp was modeled using a homology model of hP-gp based on the refined X-ray structure of mP-gp 4M2T.pdb (8), whereas the modeling of both tariquidar and DDM in mP-gp was carried out with the X-ray structure of mP-gp 4Q9H.pdb (9). The receptor (protein) and ligand structures were prepared with the MGL Tools software package (Scripps Research Institute) (42), and the actual docking studies were executed with the program AutoDock Vina (Scripps) (43). The search for the best ligand poses was performed in an inner box of dimensions 40 Å × 35 Å × 35 Å, centered at the position of the QZ59-RRR molecule in the X-ray structure of mP-gp 4M2S.pdb (x = 19.317, y = 52.588, and z = −0.676). Twenty-six residues of the drug-binding pocket of P-gp were set as flexible residues (see Ref. 4 for the list of the residues), and the exhaustiveness level was set to 40.

Author contributions

S. S., B. A., and S. V. A. designed the study. S. S. and B. A. performed the experiments. E. E. C. performed the docking studies. S. V. A. supervised the study. All authors analyzed the data and contributed to writing the paper.