The ABCG2 multidrug transporter is a pump gated by a valve and an extracellular lid

The human ATP-binding cassette transporter ABCG2 is a key to anticancer resistance and physiological detoxification. However, the molecular mechanism of substrate transport remains enigmatic. A hydrophobic di-leucine motif in the ABCG2 core separates a large intracellular cavity from a smaller upper cavity. We show that the di-leucine motif acts as a valve that controls drug extrusion. Moreover, the extracellular structure engages the re-entry helix and all extracellular loops to form a roof architecture on top of the upper cavity. Disulfide bridges and a salt bridge limit roof flexibility, but provide a lid-like function to control drug release. We propose that drug translocation from the central to the upper cavities through the valve is driven by a squeezing motion, suggesting that ABCG2 operates similar to a peristaltic pump. Finally, the roof contains essential residues, offering therapeutic options to block ABCG2 by either targeting the valve or essential residues in the roof.

(a) The structure of ABCG2 shows the N-terminal nucleotide binding domain NBD (green) connected to the elbow helix (pink) followed by transmembrane helix TMH 1-6 (grey), the first intracellular loop ICL1 (orange), cytoplasmic distal TMH2, TMH3 helical part (yellow). The extracellular elements are ECL1 (golden), ECL2 (old rose), the short loop after TMH5 (red), the re-entry helix (cyan) and the large ECL3 (violet). Amino acid residues are shown as colored balls, positive (blue), negative (red), leucine (violet) and proline (yellow). The salt bridge interactions are indicated by blue dotted lines. (b) The cryo-EM structure of the homo-dimeric human ABCG2 transporter (ABCG2-MZ29-Fab (PDB ID: 6ETI)) is shown in space-filling mode (left) and ribbon representation (right). In the ATP-free state, the inward-open conformation shows a compact extracellular region, colored as in panel a. (c) The transmembrane domains (TMDs) of ABCG2 form a large water drop-like central cavity (blue), separated from an upper cavity (blue oval) by a molecular valve structure ranging from residue G553 to T559, a part of the short linker after TMH5 (red ribbon). ECL3, located on the top of the upper cavity depicted in polarity surface mode; polar residues (green), negatively charged residues (red), positively charged residues (blue), non-polar residues (white). (d) Top view of the outer ABCG2 architecture highlighting the polar roof. (e) Zoom-in membrane-side view rotated by 90° of ABCG2 in ribbon mode to show the extracellular membrane regions; positions of three cysteine residues (white balls-andsticks) and asparagine (green balls-and-sticks) in ECL3 are given. A kinked re-entry helix (cyan ribbon) submerges into the outer leaflet, reaching into the membrane lipid bilayer.

Supplementary Figure 2. Sequence alignments of mammalian ABCGs and yeast PDRs.
The alignments were conducted using clustalX2. The conserved residues are highlighted and the conservation score shown at the bottom, indicated by the height of grey bars. The extracellular regions are marked by boxes: (a) ECL1 (golden), (b) ECL2 (old rose) and (c) short link after TMH5 (red), re-entry helix (cyan) and ECL3 (violet), respectively.

Supplementary Figure 3. Immunodetection of ABCG2 mutants in stable HEK293 cells.
(a) Proteolytic stability of ABCG2 mutants were analyzed in stable HEK293 cells. Cycloheximide was treated at indicated time point. Cells were lysed and subjected to immunoblotting using the monoclonal mouse anti-ABCG2 (BXP-21) antibody. The arrows indicate mature glycosylated protein at approximately 72 kDa, while arrow heads point to the immature bands. β-actin was used as an internal loading control. (b) Immunodetection of ABCG2 variants, stably expressed in HEK293 cells in reducing and non-reducing conditions and detected by the anti-ABCG2 (BXP-21) antibody. ABCG2 monomers migrate at approximately 72 kDa, while dimeric ABCG2 is located above. β-actin was used as an internal loading control.

Supplementary Figure 4. Intracellular mitoxantrone accumulation in stable HEK293 cells.
Mitoxantrone accumulation was quantified in the presence (black bars) and absence (grey bars) of Ko143 after incubation for 20 min at 37 °C. Mitoxantrone accumulation is given as percentage full inhibition relative to WT. Data are from several independent experiments (n = 7) and represented as means with + SEM; **P < 0.01; *P < 0.1 vs. K86M as negative control.

Supplementary Figure 5. Water movement through the di-leucine valve.
(a) Spatial density of water was calculated in a 4x4x6 nm box centered at the di-leucine valve (magenta box). Two blue boxes of 2x2x2 nm sides were used to determine water flow through the di-leucine valve by counting the number of exchanged water molecules in consecutive simulation frames of 10 ps separation. (b) Waters passing through the di-leucine valve were counted and plotted versus time. The counting was based on two selection regions (see blue boxes) and a value was picked if a water molecule left one selection and entered another at the same time.

Supplementary Fig. 6: Conformational changes between inward-and outward-facing states.
(a) Overlay of dimeric ABCG2 between inward-facing (pale orange and pale pink) and outwardfacing (dark orange and dark pink) states (PDB ID: 6ETI and 6HBU, respectively). (b) Top-view and bottom-view at both membrane leaflets of the TMD indicate movements of six TMHs at the membrane border during the catalytic cycle (NBD-and extracellular region-removed for clarity). (c) Conformational changes of the TMDs during switch between the inward-(grey) and the outward-facing states (green). The ABCG2 dimer was fitted onto TMD1. The movement of TMH5 that stretch along the central cavity was measured in both states and highlighted as the angle in red line.