Structural insight into substrate and inhibitor discrimination by human P-glycoprotein

See allHide authors and affiliations

Science  15 Feb 2019:
Vol. 363, Issue 6428, pp. 753-756
DOI: 10.1126/science.aav7102

To transport or not to transport

Therapeutic drug delivery into cells is complicated by membrane proteins like ABCB1 (also termed P-glycoprotein) that shuttle diverse compounds out of cells. Alam et al. determined high-resolution cryo–electron microscopy structures of ABCB1 bound either to a substrate, the cancer drug Taxol, or to the ABCB1 inhibitor zosuquidar. The conformational changes that facilitate drug transport are caused by hydrolysis of adenosine triphosphate (ATP). The structures show that, although Taxol and zosquidar bind to the same site, subtle structural differences lead to altered conformations of the nucleotide binding domains that are responsible for ATP hydrolysis.

Science, this issue p. 753


ABCB1, also known as P-glycoprotein, actively extrudes xenobiotic compounds across the plasma membrane of diverse cells, which contributes to cellular drug resistance and interferes with therapeutic drug delivery. We determined the 3.5-angstrom cryo–electron microscopy structure of substrate-bound human ABCB1 reconstituted in lipidic nanodiscs, revealing a single molecule of the chemotherapeutic compound paclitaxel (Taxol) bound in a central, occluded pocket. A second structure of inhibited, human-mouse chimeric ABCB1 revealed two molecules of zosuquidar occupying the same drug-binding pocket. Minor structural differences between substrate- and inhibitor-bound ABCB1 sites are amplified toward the nucleotide-binding domains (NBDs), revealing how the plasticity of the drug-binding site controls the dynamics of the adenosine triphosphate–hydrolyzing NBDs. Ordered cholesterol and phospholipid molecules suggest how the membrane modulates the conformational changes associated with drug binding and transport.

ABCB1, or P-glycoprotein, is an ATP-binding cassette (ABC) transporter of physiological and clinical importance. Its nucleotide-binding domains (NBDs) harness the energy of ATP hydrolysis to generate conformational changes in the transmembrane domains (TMDs) that facilitate the shuttling of chemically diverse compounds across many blood-organ barriers (14). Consequently, ABCB1 activity can confer multidrug resistance to cancer cells and prevent drugs from reaching therapeutic concentrations in target cells or organs, which complicates chemotherapy and/or the treatment of certain neurological disorders. Despite showing promise in model systems (57), chemosensitization of multidrug-resistant cells through the simultaneous delivery of ABCB1 inhibitors (e.g., the third-generation inhibitor zosuquidar) and chemotherapeutic drugs (e.g., Taxol/paclitaxel) has so far been clinically unsuccessful (8, 9). To understand the interaction of ABCB1 with small-molecule compounds, to rationalize its substrate specificity and the discrimination of substrates and inhibitors, and to facilitate the development of more specific or potent inhibitors for clinical use, structural insight into drug and inhibitor binding to ABCB1 is essential. No structures of ABCB1 bound to transport substrates are available at present, and inhibitor-bound and apo structures are only available for detergent-solubilized ABCB1 and remain controversial because proper ABCB1 function is strongly dependent on the membrane.

We reconstituted ABCB1 in nanodiscs comprising a mixture of brain polar lipids and cholesterol and determined near-atomic resolution cryo–electron microscopy (cryo-EM) structures in complex with Taxol (3.6-Å resolution) or zosuquidar (3.9-Å resolution). In both cases, the antigen-binding fragment (Fab) of the inhibitory antibody UIC2 (10), shown to be compatible with inward-open and occluded conformations (11), was added (complex mass: ~200 kDa) to facilitate higher-resolution structure determination.

Nanodisc-reconstituted wild-type human ABCB1 (ABCB1H) displayed ATPase activity in the range of 200 to 400 nmol ATP mg−1 min−1, which was mildly stimulated by Taxol and inhibited by zosuquidar (Fig. 1A), in agreement with earlier observations (12, 13). This suggested that at 10 μM, the Taxol concentration chosen for structural studies, a sufficiently large fraction of ABCB1H molecules should contain bound drug. We observed two main conformations in our single-particle cryo-EM analysis (fig. S1). The highest-resolution structure (Fig. 1B) revealed an occluded conformation with density covering a single Taxol molecule (Fig. 1C) in a central cavity formed by the closing of a gate region consisting of TM4 and TM10 (Fig. 1D). The NBDs were closer together than in previously determined, inward-open apo structures of mouse ABCB1 (11, 1416) and more closely resembled those of disulfide-trapped human-mouse chimeric ABCB1 (ABCB1HM) structures (11) despite the absence of nucleotides or disulfide cross-linking. The second conformation revealed a slightly larger separation of the NBDs and poorly ordered TM4 and TM10 segments. In this conformation, the cytoplasmic gate to the drug-binding cavity is open. Our results demonstrate that binding of Taxol to ABCB1 induces an occluded conformation and a concomitant closure of the inter-NBD gap, which is in line with earlier mutagenesis and biochemical work (17, 18). The central pocket of Taxol-bound ABCB1 is lined by amino acid residues from all 12 TM helices. Whereas the density for interacting residues was well defined, that of the Taxol molecule was less clear, suggesting the possibility of multiple binding modes. The orientation of Taxol shown in Fig. 1, C and E, had the strongest density assigned to the tetracyclic core (baccatin III) with the cyclooctane ring in a crown conformation. The peripheral moieties displayed conformational heterogeneity, and their placement was guided by fitting the Y-shaped tail of the molecule to avoid steric clashes with neighboring side chains. Given its volume, only one Taxol molecule can bind to the central cavity of ABCB1, and occlusion of the drug-binding pocket is triggered irrespective of which binding mode the molecule adopts. The drug-binding cavity of ABCB1 is globular in shape, in contrast to the flatter, slit-like drug-binding pocket previously visualized in the human multidrug transporter ABCG2 (19, 20). This is in line with the finding that Taxol cannot bind to ABCG2 or modulate its activity (21, 22). A comparison of the substrate- and inhibitor-bound structures of these two key human multidrug exporters therefore allows us to rationalize their divergent substrate specificities.

Fig. 1 In vitro function and structure of nanodisc-reconstituted ABCB1.

(A) Taxol- and zosuquidar-modulated ATPase activity. Data points analyzed represent means of three independent measurements. Error bars indicate SD. (B) Ribbon diagram of human ABCB1 bound to Taxol (green spheres). The N- and C-terminal halves of ABCB1 are colored yellow and orange, respectively, with the UIC2 Fab shown in blue. (C) Close-up of binding site showing side chains of residues within 5 Å of bound Taxol (green sticks), viewed parallel to the membrane plane. EM density is shown as a blue mesh, contoured at 6σ. (D) Ribbon representation of TM4 (yellow) and TM10 (orange) adopting kinked conformation, with Taxol located in the center of the occluded cavity. EM density after nanodisc subtraction is contoured at 8σ. (E) Interactions between Taxol and ABCB1 side chains. Nonbonded interactions are represented by spoked arcs and hydrogen bonds are indicated by dashed green lines. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; E, Glu; F, Phe; I, Ile; L, Leu; M, Met; Q, Gln; S, Ser; W, Trp; and Y, Tyr.

For the zosuquidar-bound structure, we used a hydrolysis-deficient variant of ABCB1HM harboring an E→Q mutation in the walker-B motif (ABCB1HM-EQ) and added ATP to ensure that zosuquidar had indeed trapped ABCB1 in an inhibited state, given that, in the presence of ATP and Mg2+, this ABCB1 variant had been shown to adopt a closed NBD conformation coupled to a closed TMD conformation with a collapsed translocation pathway (23). Our zosuquidar-inhibited ABCB1HM-EQ structure (Fig. 2A and figs. S3 to S5) displayed a conformation similar to that of Taxol-bound ABCB1—with the TMDs forming an analogous occluded cavity—but containing two bound zosuquidar molecules. We again observed two main ABCB1 conformations, characterized by a distinct degree of NBD opening. However, unlike for Taxol, both conformations revealed bound zosuquidar molecules. The density for zosuquidar was better defined than that for Taxol and allowed unambiguous placement in a specific binding mode (figs. S4 and S5, B and C). This likely stems from the increased contact area (buried surface) between the two zosuquidar molecules and ABCB1 (~1000 Å2) compared with Taxol (~800 Å2) as well as the intermolecular contact between the two zosuquidar molecules themselves (interface surface area of ~190 Å2). The orientation and binding interactions of the zosuquidar molecules are similar to those previously observed in the disulfide-trapped, detergent-solubilized ABCB1HM structure, suggesting that the opposite effect of zosuquidar on the ATPase rate of ABCB1 in lipid bilayers (reduction of ATPase activity) or detergent solution (stimulation) is not attributable to distinct binding sites of zosuquidar in these lipidic environments.

Fig. 2 Comparison of zosuquidar- and Taxol-bound ABCB1.

(A) Cartoon of zosuquidar-bound ABCB1HM-EQ structure with zosuquidar molecules shown as yellow and magenta spheres. The N- and C-terminal halves of ABCB1 are colored pink and blue, respectively. (B to D) Superposition of Taxol-bound human ABCB1 (green ribbon) and zosuquidar-bound ABCB1HM-EQ (magenta ribbon) with intracellular helices interacting with NBDs shown as cylinders. Zosuquidar molecules are shown as spheres.

Zosuquidar and Taxol binding to the same pocket raises the key questions of how ABCB1 distinguishes transport substrates from inhibitors and how these compounds exert opposite effects on ATPase activity. To explain the polyspecificity of ABCB1, plasticity of the drug-binding pocket in terms of side-chain and backbone rearrangements to accommodate distinct substrates has often been invoked (24, 25). We therefore superimposed Taxol-bound and zosuquidar-bound ABCB1 and found that the main-chain and side-chain conformations of the residues surrounding bound drugs are largely similar (Fig. 2, B to D, and fig. S6). However, there were a number of small but notable structural differences localized primarily in the second half of the transporter. As seen in Fig. 2B, the subtle changes originate at the drug-binding site and are transmitted and amplified via the helix pair TM7-TM8 and TM12 to NBD2 (Fig. 2, C and D). They lead to an outward shift of intracellular helix 2 (coupling helix 1) in zosuquidar-bound ABCB1, increasing the distance between the NBDs and offering a plausible explanation for the reduced ATPase activity in the presence of zosuquidar. Thus, our results demonstrate that plasticity occurring within the confines of the occluded drug-binding pocket can be linked to NBD movement and ATPase activity of ABCB1.

The function of ABCB1 is known to be modulated by the lipidic membrane. As was observed in the structures of nanodisc-reconstituted ABCG2 (19), we found ordered cholesterol and phospholipid molecules bound to the transmembrane region of ABCB1 (Fig. 3 and fig. S2). At the level of the outer membrane leaflet, a ring of ordered cholesterol molecules is bound to surface grooves on ABCB1, and specific interactions include hydrogen bonds with the hydroxyl group of cholesterol and stacking interactions with aromatic side chains, as observed for other cholesterol-protein interactions (26). At the level of the inner leaflet, density compatible with a bound phospholipid and cholesterol was observed in a membrane-exposed pocket near TM3, TM4, and TM6 (Fig. 3, left panel) and at the pseudosymmetrically related site near TM9, TM10, and TM12 (Fig. 3, right panel). Because the phospholipid- and cholesterol-binding sites are formed by the kinking of TM4 and TM10 in response to drug and inhibitor binding, our observation suggests a direct mechanism of ABCB1 modulation by inner leaflet lipids.

Fig. 3 Phospholipids and cholesterol bound to ABCB1.

(Center) Surface representation of Taxol-bound human ABCB1 showing bound lipid (phosphotidylethanolamine, magenta spheres) and cholesterol molecules (purple spheres). Zoom-in panels show details of binding sites of (left) phospholipid (magenta sticks) and (right) cholesterol (purple sticks) at the level of the inner leaflet, near the kinking TM helices TM4 and TM10. EM density (blue mesh) is contoured at 6σ.

When combined with the previously reported structure of ATP-bound ABCB1EQ (23), our results offer a structural mechanism both for drug extrusion in a transport cycle and competitive inhibition of this reaction by small-molecule inhibitors. Formation of the closed ATP-bound NBD dimer triggers conformational changes in TM4 and TM6 from TMD1 as well as the symmetrically related TM10 and TM12 from TMD2, generating a steric clash with bound drugs (Fig. 4, A and B). This suggests that a peristaltic mechanism contributes to the extrusion of bound substrate during the transport cycle (Fig. 4C). Competitive inhibitors such as zosuquidar likely function by arresting the transporter in an occluded conformation, thus (i) restricting access to the substrate binding site, (ii) preventing NBD closure and consequently inhibiting ATPase activity, and (iii) preventing a transition to an outward open state. Our data are in line with proposed mechanisms of relaying long-range structural changes upon substrate and inhibitor binding to the NBDs (17), as well as subtle induced-fit type rearrangements of a dynamic binding pocket (18, 24). In contrast to transport substrates, where multiple holo and apo forms of ABCB1 likely coexist, inhibitors have fewer or a single binding mode, fill the drug-binding cavity more completely, and form a larger number of contacts with ABCB1. This suggests that, in a continuum of substrates and inhibitors, three-dimensional shape complementarity and the strength of the contacts rather than distinct binding sites are key determinants distinguishing substrates from inhibitors. Although the functional modulation of ABCB1 by cholesterol and lipids has previously been demonstrated (12, 2729), our results illustrate specific interactions and binding sites in cholesterol-containing lipid bilayers.

Fig. 4 Proposed mechanism of P-glycoprotein/ABCB1 substrate transport and small-molecule inhibition.

(A) Side view of select TM helices of ABCB1 after superposition of the drug-bound, occluded conformation (green and red ribbons for Taxol- and zosuquidar-bound structures, respectively) with the previously reported ATP-bound posttranslocation state (black ribbons). Taxol and zosuquidar molecules are shown as green and red spheres, respectively. (B) Same as (A) but viewed from the cytoplasm. (C) Schematic of proposed ABCB1 transport cycle in the presence of substrate (Taxol, green star) and inhibitor (zosuquidar, red L-shape). ATP and adenosine diphosphate are indicated by T and D, respectively, whereas the dashed lines in the NBDs represent ATP-binding elements required for NBD dimerization and ATP hydrolysis. Major conformational states are represented by circled numbers. State 1: apo state [Protein Data Bank (PDB) IDs 4M1M and 4QNH, among others]. States 2 and 2′: drug- and inhibitor-bound, this study. State 3: proposed outward-facing conformation based on Sav1866 structure (30). State 4: collapsed posttranslocation state (PDB ID 6C0V). Pi, inorganic phosphate.

Finally, given that the structures of Taxol- and zosuquidar-bound ABCB1 reveal critical interactions in a key state of the transporter, our results may allow medicinal chemists and computational biologists to exploit structural insight into ABCB1 to guide the design of drugs or inhibitors.

Supplementary Materials

Materials and Methods

Figs. S1 to S6

Table S1

References (3145)

References and Notes

Acknowledgments: We thank the staff at the Scientific Center for Optical and Electron Microscopy (SCOPEM) at ETH Zürich. We also acknowledge N. Tremp for help with protein expression and cell culture work. Funding: This work was funded by the European Molecular Biology Organization long-term postdoctoral fellowship to A.A., grants from the Swiss Cancer League to K.P.L., the Swiss National Science Foundation through NCCR Structural Biology and TransCure, the Swiss Cancer League, and U.S. National Institutes of Health grant P20GM109091 to E.B. and I.R. Author contributions: A.A. carried out all experimental procedures related to protein expression and purification and EM sample preparation. A.A. prepared cryo-EM grids and A.A. and J.K. collected and processed EM data. E.B. and I.R. provided UIC2-producing hybridoma cells and helped analyze data. K.P.L. and A.A. conceived of the project and planned the experiments. A.A. and K.P.L. wrote the manuscript with input from all authors. Competing interests: The authors declare no competing interests. Data and materials availability: All data are available within the main text or supplementary materials. Cryo-EM maps are available at the Electron Microscopy Data Bank with accession codes EMD-4539, EMD-4540, and EMD-4541 (maps 1, 2, and 3 of Taxol-bound ABCB1, respectively) and EMD-4536 (zosuquidar-bound ABCB1). Coordinates for deposited models are available at the Protein Data Bank with IDs 6QEX (Taxol-bound ABCB1) and 6QEE (zosuquidar-bound ABCB1).

Stay Connected to Science

Navigate This Article