To identify chemical probes that target EboV host factors, we screened a library of small molecules and identified a novel benzylpiperazine adamantane diamide, 3.0, that inhibits infection of Vero cells by vesicular stomatitis virus particles (VSV) pseudotyped with EboV Zaire GP, but not with VSV G or Lassa fever virus (LFV) GP (). To verify that 3.0 is a bona fide inhibitor, we measured EboV growth on Vero cells for 96 hours and found it was reduced by >99% in the presence of 3.0 (Supplementary Fig. 1a
). We synthesized and tested >50 analogs of 3.0 and found that the addition of a (methoxycarbonyl) benzyl group at the ortho position of the benzene ring (compound 3.47) increased the potency as measured by a single cycle of EboV GP-dependent infection and efficacy as measured by growth of EboV on Vero cells ().
Structure and function of ebolavirus entry inhibitors
Previous studies revealed that the endosomal protease cathepsin B is essential for EboV infection because it cleaves the GP1 subunit of GP3,4
. To address the possibility that 3.0 and 3.47 target this step, we measured cathepsin B activity in the presence of these compounds and found no effect in vitro
or in cells (data not shown). Moreover, 3.0 and 3.47 inhibited infection by VSV EboV particles treated with thermolysin, a metalloprotease that faithfully mimics cathepsin cleavage of the GP1 subunit of GP (Supplementary Fig. 1b
. These findings demonstrate that cathepsin B is not the target of 3.0 and 3.47.
HeLa cells treated with 3.0 or 3.47 for more than 18 hours developed cytoplasmic vacuoles that were labeled by cholesterol-avid filipin (). The induction of filipin-stained vacuoles by the compounds suggested that they target one or more proteins involved in regulation of cholesterol uptake in cells. To test this hypothesis, we used mutant cell lines and cells treated with siRNA to analyze proteins for which loss of activity had been previously associated with cholesterol accumulation in late endosomes10–12
. We found that EboV GP infection is dependent on the expression of Niemann-Pick C1 (NPC1), but not Niemann-Pick C2 (NPC2), acid sphingomyelinase (ASM), ALG-2-interacting protein X (Alix), or oxysterol binding protein 5 (ORP5) (, Supplementary Fig. 2a–c
). NPC1 is a polytopic protein that resides in the limiting membrane of late endosomes and lysosomes (LE/LY) and mediates distribution of lipoprotein-derived cholesterol in cells10,13
. To analyze the role of NPC1 in infection, we studied Chinese hamster ovary (CHO)-derived cell lines that differ in expression of NPC1. We found that the titer of a murine leukemia virus (MLV) vector pseudotyped with EboV GP on wild type CHO cells (CHOwt
) exceeded 106
infectious units/ml (). Importantly, CHO cells lacking NPC1 (CHOnull
) were completely resistant to infection by this virus and infection of these cells was fully restored when NPC1 was expressed (CHONPC1
). Thus, NPC1 expression is essential for EboV infection.
NPC1 is essential for ebolavirus infection
cells, LE/LY are enlarged and contain excess cholesterol (Supplementary Fig. 3
. To determine if EboV infection is inhibited by endosome dysfunction secondary to the absence of NPC1, we studied a well-characterized NPC1 mutant P692S that is defective in cholesterol uptake and NPC1-dependent membrane trafficking13–15
and found that expression of NPC1 P692S fully supports infection of CHOnull
cells (). Conversely, gain-of-function mutants NPC1 L657F and NPC1 D787N14
did not enhance EboV GP infection. Thus, EboV entry is strictly dependent on NPC1 expression but not NPC1-dependent cholesterol transport activity. Consistent with the conclusion that NPC1 expression is essential for EboV GP-dependent entry, we found that ebolavirus did not grow on CHOnull
cells (). In addition, we tested a single round of infection by MLV particles bearing GPs from the filoviruses EboV Sudan, EboV Côte d'Ivoire, EboV Bundibugyo, EboV Reston and marburgvirus and found that all are strictly NPC1-dependent (Supplementary Fig. 4
). Since these viruses are not closely related16
, these findings suggest that the requirement for NPC1 as an entry factor is conserved among viruses in the Filoviridae
Since NPC1 and cathepsin B are both essential host factors, we analyzed their relationship during infection. In our initial experiment, we compared cathepsin B activity in CHOnull
cells and found it was not significantly different from CHOwt
cells (Supplementary Fig. 5
). To determine if NPC1 is required for virus processing by cathepsin B, we tested whether thermolysin-cleaved particles are dependent on NPC1. As expected, we found that thermolysin-cleaved particles are infectious and resistant to inactivation of cathepsin B when NPC1 is present (). However, thermolysin cleavage did not bypass the barrier to virus infection of NPC1 deficient cells. Taken together, these findings indicate that cathepsin B and NPC1 mediate distinct steps in infection.
Previous studies suggest that the product of cathepsin B cleavage of the GP1 subunit of EboV GP is a ligand for a host factor6,17–20
. To test this hypothesis, we performed a series of experiments measuring binding of EboV GP to LE/LY membranes from CHOnull
cells ( left panel). The source of EboV GP is a purified recombinant protein that is truncated just before the transmembrane domain (EboV GPΔ™
). EboV GPΔ™
is a trimer that is faithfully cleaved by thermolysin (, right panel). We found that binding of EboV GPΔ™
to LE/LY membranes is concentration dependent, saturable, and strictly dependent on both thermolysin cleavage of GP1 and membrane expression of NPC1 or NPC1 P692S ( and Supplementary Fig. 6a,b
). To determine if cleaved GP binds to NPC1, we performed a co-immunoprecipitation experiment. LE/LY membranes were incubated with EboV GPΔ™
and then solubilized in detergent. NPC1 was recovered from the lysate by immunoprecipitation and the immune complexes were analyzed for GP1. The findings indicate that cleaved EboV GPΔ™
binds to NPC1 and that uncleaved EboV GPΔ™
does not ().
Protease-cleaved EboV GP binds to NPC1
Since the small molecules 3.0 and 3.47 inhibit infection of thermolysin-treated VSV EboV GP particles (Supplementary Fig. 1b
) and inhibit cholesterol uptake from LE/LY into cells (), both of which require NPC1, this suggests the possibility that these compounds directly target NPC1. To test this hypothesis, we synthesized the 3.47 derivative 3.98. This compound has anti-EboV activity and contains two additional functional moieties: an aryl-azide for photo-affinity labeling of target proteins and an alkyne for click conjugation with biotin21
(Supplementary Fig. 7
). Compound 3.98 was incubated with LE/LY membranes, activated by UV light and coupled to biotin. NPC1 was then isolated by immunoprecipitation and analyzed using streptavidin-HRP. The findings show that NPC1 is cross-linked to 3.98 and that cross-linking is inhibited by the presence of 3.47 but not by the closely-related analog 3.18, which has weak anti-viral activity (, Supplementary Fig. 7
). In addition, we observed that overexpression of NPC1 conferred resistance to the anti-viral activity of 3.0 and 3.47 (Supplementary Fig. 8
), thus providing additional functional evidence supporting the conclusion based on the results of the cross-linking experiment using 3.98 that NPC1 is a direct target of the anti-viral compounds.
NPC1 is a target of the small molecule inhibitors
The evidence that NPC1 is the target of the 3.0-derived small molecules selected for anti-EboV suggested that these compounds interfere with binding of cleaved GP to NPC1. Consistent with this hypothesis, we found that 3.0 and 3.47 inhibited binding of cleaved EboV GPΔ™
to NPC1 membranes in a concentration-dependent manner (). Importantly, we observed a direct correlation between the potency of 3.47, 3.0, and 3.18 in inhibiting binding (, left panel) and in inhibiting EboV infection (Supplementary Fig. 7
). We also tested U18666A, a small molecule inhibitor of LE/LY cholesterol transport and membrane trafficking22,23
, and found that it does not inhibit binding of cleaved EboV GP to NPC1 membranes (, right panel). These results support the conclusion that the 3.0-derived compounds inhibit EboV infection by interfering with binding of cleaved GP to NPC1.
Previous studies show that cleavage of GP by endosomal cathepsin proteases removes heavily-glycosylated domains in the GP1 subunit and exposes the N-terminal domain3–7
. It has been proposed that binding of this domain to a host factor is essential for infection6,17–20
. The most straight forward interpretation of the findings in this report is that NPC1 is this host factor. This conclusion is based on the observations that NPC1 is strictly required for infection, that cleaved GP1 binds to NPC1, and that small molecules that target NPC1 are potent inhibitors of binding and infection.
Analysis of the EboV GP structure reveals that the residues in the N-terminal domain of GP1 that mediate binding to NPC1 are interspersed with the residues that make stabilizing contacts with GP25
. This structural feature is consistent with the possibility that binding of cleaved GP1 to NPC1 relieves the GP1-imposed constraints on GP2 and promotes virus fusion to the limiting membrane (). The role of cathepsin proteases in cleavage of GP1 to expose the NPC1 binding site during EboV infection is analogous to the role of CD4 in inducing a conformational change in gp120 to expose the co-receptor binding site during human immunodeficiency virus infection8
. An alternative possibility is that binding of protease-cleaved GP1 to NPC1 is an essential step in infection, but virus membrane fusion is not completed until an additional signal is received, possibly including further cleavage of GP by cathepsin proteases, as has been proposed3,4,9
. These studies provide an example of how small molecules identified by screening and medicinal chemistry optimization can be used as molecular probes to analyze virus-host interactions.