Ebola Viral Glycoprotein Bound to Its Endosomal Receptor Niemann-Pick C1

Abstract

Filoviruses, including Ebola and Marburg, cause fatal hemorrhagic fever in humans and primates. Understanding how these viruses enter host cells could help to develop effective therapeutics. An endosomal protein, Niemann-Pick C1 (NPC1), has been identified as a necessary entry receptor for this process, and priming of the viral glycoprotein (GP) to a fusion-competent state is a prerequisite for NPC1 binding. Here, we have determined the crystal structure of the primed GP (GPcl) of Ebola virus bound to domain C of NPC1 (NPC1-C) at a resolution of 2.3 Å. NPC1-C utilizes two protruding loops to engage a hydrophobic cavity on head of GPcl. Upon enzymatic cleavage and NPC1-C binding, conformational change in the GPcl further affects the state of the internal fusion loop, triggering membrane fusion. Our data therefore provide structural insights into filovirus entry in the late endosome and the molecular basis for design of therapeutic inhibitors of viral entry.

Graphical abstract

Figure 1

Model of NPC1 Bound to the Primed EBOV GP and Overall Structure of NPC1-C (A) The full-length EBOV GP was primed by cathepsin L and B in the late endosome, and the highly glycosylated 130 kDa GP1 subunit was cleaved into a 19 kDa GP1, generating a primed form of GP (GPcl). Then, the GPcl binds to the domain C of NPC1 molecule (NPC1-C), which is an endosomal 13-transmembrane protein with three large luminal domains (domains A, C, and I). (B) Crystal structure of NPC1-C. The helices (α and η) are colored in cyan, the β strands are colored in magenta, and the loops are colored in light pink. The disulfide bonds are colored in yellow. NPC1-C displays a helical core structure surrounded by several β strands with two extended loops. (C) Topological secondary structure of NPC1-C. Secondary elements are shown as cylinders and arrows for helices and β strands. See also Figures S1, S2, and S3 and Table S1.

Figure S1

Characterization of Refolded NPC1-C Protein and Analysis of Potential N-Link Glycosylation Sites, Related to Figure 1 (A) The refolded NPC1-C protein was eluted as a single sharp peak in the gel filtration on Superdex 200^® column (GE Healthcare). The NPC1-C protein shows perfect purity, which is ran as a single band in the SDS-PAGE without any extra bands. Irrelevant lanes were removed during preparation of the figure. (B) The NPC1-C has seven potential N-link glycosylation sites, six of which locate in the loops except one (N452) at the tip of α4 helix. The sites are shown in red spheres in the NPC1-C structure.

Figure S2

Sequence Alignment of NPC1-C and NPC1, Related to Figure 1 The intact luminal domain C consists of residues 372 to 640. In our crystal structure, only residues 384 to 624 are visible, and the secondary structures are labeled. We cannot see the electron density for the N-terminal 12 amino acids and C-terminal 16 amino acids, indicating these disordered regions might be flexible loops.

Figure S3

Protruding Model of NPC1-C, Related to Figure 1 We cannot see the density for both ends immediately adjacent to the TM segments. The missing residues (in dash lines) of both the N-terminal and C-terminal could be flexible loops, which will not make the visible domain C structure to be tightly attached to the endosomal membrane. It is assumed that, with these flexible N-terminal and C-terminal loops, the domain C can protrude from the endosomal membrane and be easily bound by the primed GP molecule.

Figure S4

Characterization of GP-ectoΔmucin and GPcl Proteins and Binding Affinities of 2G4 Bound to GP and GPcl, Related to Figure 2 (A) The GPectoΔmucin protein was eluted as a single peak (the elution volume represents a molecular weight of ∼180 kDa, homotrimer) in the gel filtration of Superdex^® 200 column (GE Healthcare). The primed GP protein (GPcl) was also eluted as a single peak (the elution volume represents a molecular weight of ∼120 kDa). In the SDS-PAGE, after thermolycin digestion, the GP1 subunit was cleaved into a 19 kDa band, the molecular weight of which is similar to the GP2 subunit. (B) The GP-ectoΔmucin and GPcl proteins were immobilized on the CM5 chip with about 2000 response units. Then serial concentrations of 2G4 Fab were flow through the chip. The BIAcore^® diagrams reveal that 2G4 Fab binds to GP or GPcl with slow on and off rates. Moreover, 2G4 Fab binds to GP or GPcl with similar binding affinities (8.2 nM versus 3.8 nM). The K_D values were calculated by the BIAcore^® 3000 analysis software (BIAevaluation Version 4.1).

Figure 2

Binding Affinity of GPcl Bound to NPC1-C (A) BIAcore diagram of refolded E. coli-expressed NPC1-C protein bound to the GP-ectoΔmucin protein. The refolded NPC1-C shows no binding to the GP-ectoΔmucin. (B and C) BIAcore diagram and saturation curve of refolded NPC1-C protein bound to the GPcl protein. The refolded NPC1-C binds to the GPcl with a low affinity and a fast kinetics. (D) BIAcore diagram of glycosylated 293T-expressed NPC1-C protein bound to the GP-ectoΔmucin protein. The glycosylated NPC1-C also shows no binding to the GP-ectoΔmucin. (E and F) BIAcore diagram and saturation curve of glycosylated NPC1-C protein bound to the GPcl protein. The glycosylated NPC1-C binds to the GPcl with a similar low affinity and a fast kinetics. Response units were plotted against protein concentrations. The K_D values were calculated by the BIAcore 3000 analysis software (BIAevaluation Version 4.1). See also Figures S4 and S5.

Figure S5

Characterization of Glycosylated NPC1-C Protein Expressed in Mammalian 293T Cells, Related to Figure 2 The NPC1-C protein was expressed in 293T mammalian cells as secreted protein. After purification with affinity chromatography using HisTrap FF column (GE Healthcare), the glycosylated NPC1-C protein was eluted as a single peak in the gel filtration on Superdex 200 column (GE Healthcare). The NPC1-C protein shows good purity, which is run as a dispersing single band in the SDS-PAGE, with an average molecular weight of 55 kDa. The protein was further confirmed by Western blot assay with anti-histag antibody. Irrelevant lanes were removed during preparation of the figure.

Figure 3

Complex Structure of NPC1-C Bound to GPcl (A) The bound GPcl structure. The visible portions of GP1 and GP2 subunits are colored in green and yellow, respectively, and the other invisible or cleavage-removed portions are colored in gray. (B) Overall complex structure. The NPC1-C uses its two protruding loops to bind to the head of GPcl at a perpendicular angle. (C) Surface representations of interacting residues in the NPC1-C and GPcl. The residues responsible for the binding in the NPC1-C (top) are colored in blue, and the interacting residues in the GPcl (bottom) are colored in orange. The NPC1-C mainly utilizes aromatic residues to engage the cavity of the GPcl, which mainly consists of hydrophobic residues. See also Table S1.

Figure 4

NPC1-C Mimics Similar but Stronger Interactions like the Glycan Cap in the Unprimed GP (A) In the unprimed GP structure, the glycan cap utilizes two aromatic residues, F225 and Y232, to insert into the hydrophobic receptor binding cavity in the head of GPcl. (B) NPC1-C takes advantage of four aromatic residues (Y423, F503, F504, and Y506) and one hydrophobic residue P424 to interact with the hydrophobic cavity. In particular, the residue F503 inserts into the cavity at a deeper position. Thus, the binding of NPC1-C mimics the interaction like the glycan cap but likely with a better binding capacity by providing more aromatic residues. (C) An MARV-neutralizing antibody MR79 targets the hydrophobic binding cavity of MARV GP through aromatic residues (F111.2 and Y112.2).

Figure 5

Conformational Changes in the GPcl upon Enzymatic Cleavage and NPC1-C Engagement (A) Comparison of bound GPcl (GP1 in green and GP2 in yellow) and unbound GP-ectoΔmucin (gray glycan cap is not shown here). Three conformational changes are observed in the bound GPcl, including sites I, II, and III. At site I, the β7-β8 loop moves downward; at site II, the α1 helix raises upward. Accompanied with the raising α1 helix, an upward movement of the 3¹⁰ helix in the β3-α1 loop occurs at site III. (B) Zoomed view of the site III in the unbound GP-ectoΔmucin. The residue N73 in the 3¹⁰ helix could form a hydrogen bond with the residue K510 in the N-terminal portion of the IFL. This interaction could separate the electric repulsion between K510 and R559 (in the HR1 helix of GP2). (C) Zoomed view of the site III in the bound GPcl. The upward-moved residue N73 does not form a hydrogen bond with the residue K510, and the electric repulsion between the K510 and R539 may help to release the IFL and trigger the membrane fusion.

Figure S6

Gel Filtration Curves of Refolded Alanine(A)-Substituted NPC1-C Mutant Proteins, Related to Figure 6 Different groups of A-substituted NPC1-C mutant proteins were design at the five sites (Y423, P424, F503, F504 and F506), which contribute most of interaction between NPC1-C and GPcl. The quintuple-A and Y423A&P424A mutants cannot be refolded, as there are only aggregates peaks in the gel filtration curves on Superdex 200 (GE Healthcare). For other mutant proteins, we can see monomer peak with different refolding efficiencies.

Figure S7

Gel Filtration Curves of Refolded Glycine(G)-Substituted NPC1-C Mutant Proteins, Related to Figure 6 Different groups of A-substituted NPC1-C mutant proteins were design at the five sites (Y423, P424, F503, F504 and F506), which contribute most of interaction between NPC1-C and GPcl. The Y423G and Y506G mutants cannot be refolded, as there are only aggregates peaks in the gel filtration curves on Superdex 200 (GE Healthcare). For other mutant proteins, we can see monomer peak with different refolding efficiencies.

Figure 6

Alanine/Glycine-Scanning Mutagenesis Experiments on NPC1-C (A–I) BIAcore diagrams of quintuple-G, Y423G&P424G, F503G&F504G, F503A&F504A, F503G, F504G, F503A, F504A, and Y506A mutants of NPC1-C binding to the GPcl. These nine mutants abolish the binding to the GPcl. (J) BIAcore diagram and saturation curve of P424G mutant of NPC1-C binding to the GPcl. The P424G mutant can keep binding to the GPcl with an affinity of 213 μM, which is slightly lower than that of wild-type NPC1-C. (K) BIAcore diagram and saturation curve of Y423A mutant of NPC1-C binding to the GPcl. The Y423A mutant can keep binding to the GPcl with an affinity of 166 μM, which is slightly lower than that of wild-type NPC1-C. (L) BIAcore diagram and saturation curve of P424A mutant of NPC1-C binding to the GPcl. The P424A mutant can keep binding to the GPcl with an affinity of 281 μM, which is slightly lower than that of wild-type NPC1-C. Response units were plotted against protein concentrations. The K_D values were calculated by the BIAcore 3000 analysis software (BIAevaluation Version 4.1). See also Figures S6 and S7.