Characterization of Influenza Virus Pseudotyped with Ebolavirus Glycoprotein

ABSTRACT We have produced a new Ebola virus pseudotype, E-S-FLU, that can be handled in biosafety level 1/2 containment for laboratory analysis. The E-S-FLU virus is a single-cycle influenza virus coated with Ebolavirus glycoprotein, and it encodes enhanced green fluorescence protein as a reporter that replaces the influenza virus hemagglutinin. MDCK-SIAT1 cells were transduced to express Ebolavirus glycoprotein as a stable transmembrane protein for E-S-FLU virus production. Infection of cells with the E-S-FLU virus was dependent on the Niemann-Pick C1 protein, which is the well-characterized receptor for Ebola virus entry at the late endosome/lysosome membrane. The E-S-FLU virus was neutralized specifically by an anti-Ebolavirus glycoprotein antibody and a variety of small drug molecules that are known to inhibit the entry of wild-type Ebola virus. To demonstrate the application of this new Ebola virus pseudotype, we show that a single laboratory batch was sufficient to screen a library (LOPAC1280; Sigma) of 1,280 pharmacologically active compounds for inhibition of virus entry. A total of 215 compounds inhibited E-S-FLU virus infection, while only 22 inhibited the control H5-S-FLU virus coated in H5 hemagglutinin. These inhibitory compounds have very dispersed targets and mechanisms of action, e.g., calcium channel blockers, estrogen receptor antagonists, antihistamines, serotonin uptake inhibitors, etc., and this correlates with inhibitor screening results obtained with other pseudotypes or wild-type Ebola virus in the literature. The E-S-FLU virus is a new tool for Ebola virus cell entry studies and is easily applied to high-throughput screening assays for small-molecule inhibitors or antibodies. IMPORTANCE Ebola virus is in the Filoviridae family and is a biosafety level 4 pathogen. There are no FDA-approved therapeutics for Ebola virus. These characteristics warrant the development of surrogates for Ebola virus that can be handled in more convenient laboratory containment to study the biology of the virus and screen for inhibitors. Here we characterized a new surrogate, named E-S-FLU virus, that is based on a disabled influenza virus core coated with the Ebola virus surface protein but does not contain any genetic information from the Ebola virus itself. We show that E-S-FLU virus uses the same cell entry pathway as wild-type Ebola virus. As an example of the ease of use of E-S-FLU virus in biosafety level 1/2 containment, we showed that a single production batch could provide enough surrogate virus to screen a standard small-molecule library of 1,280 candidates for inhibitors of viral entry.

wt/GIN/2014/Kissidougou-C15 [Zaire C15]). Transduced cells were healthy in morphology (Fig. 1). EBOV-GP was expressed at the cell surface and can be detected with GP-specific monoclonal antibody (MAb) KZ52. Cells with high expression of EBOV-GP were sorted and stored as the E-SIAT cell line. GP surface expression was not toxic to MDCK-SIAT1 cells and was stable and uniform up to at least the ninth passage (Fig. 1).
Cloned seed S-FLU virus that is coated in influenza virus HA was added to propagate in E-SIAT cells to produce the EBOV-GP-pseudotyped S-FLU virus. After 48 h, culture supernatants were harvested and titrated on indicator MDCK-SIAT1 cells for infection. Infected MDCK-SIAT1 cells express eGFP that is encoded in the virus. We named the new pseudotype E-S-FLU. The E-S-FLU virus reaches a lower titer than viruses grown in H5 HA-transduced producer cells and forms smaller, more diffuse plaques than the H5-S-FLU virus (Fig. 2C). The typical 50% effective concentration (EC 50 ) dilutions of the E-S-FLU virus at 48 h were ϳ1:8, compared to ϳ1:500 for the H5-S-FLU virus ( Fig. 2A). From the EC 50 dilution and the number of cells per well (3 ϫ 10 4 ), the 50% cell infective dose (CID 50 )/ml was calculated as ϳ2 ϫ 10 6 in the E-S-FLU virus batch and ϳ1.6 ϫ 10 8 in the H5 S-FLU virus batch. However, the harvested culture supernatant from E-SIAT cells contains adequate pseudovirions for 50 l of a 1:4 dilution to give close to full infection of the 3 ϫ 10 4 cells in a well of a 96-well plate. No further purification or concentration steps are needed for the screening of inhibitory drugs or antibodies.
With the same protocol, we were also able to pseudotype the S-FLU virus with Bundibugyo, Sudan, and Mayinga-Zaire EBOV-GP to similar viral titers (data not shown).
It has been shown previously that the transmembrane and cytoplasmic domains of influenza virus HA are important for influenza viral particle assembly (23)(24)(25)(26)(27)(28)(29)(30). To potentially improve the viral titer of the E-S-FLU virus, we designed a GP-HA hybrid protein that contains the extracellular domain of the EBOV-GP and the transmembrane and cytoplasmic domains of HA. After transduction and sorting, the GP-HA protein can be detected with MAb KZ52 at the surface of EH-SIAT cells at a level similar to that on  Viruses were preincubated for 2 h with antibody KZ52 (anti-GP) or 21D85A (anti-H5) before MDCK-SIAT1 indicator cells were added. Antibodies were titrated in 2-fold dilutions from 10 g/ml. After 24 h, infection was quantified by measuring eGFP expression (influenza virus) or luminescence (lentivirus). Titration curves are shown in panels A to C. A summary of the EC 50 s of MAbs KZ52 and 21D85A for the three pseudotyped viruses is shown in panel D. Multiple EC 50 s from repeat experiments are shown with the geometric means and 95% CIs. A value of 10,000 ng/ml was assigned to antibodies that did not neutralize. (E and F) Microscopy (10ϫ objective) of assays showing the specificity of MN by antibodies. For "Virus only," the E-S-FLU and H5-S-FLU viruses were added directly to MDCK-SIAT1 cells. Infected cells express eGFP, shown as green cells. In the MN assay, virus was preincubated with 10 g/ml MAb KZ52 (anti-GP) or 21D85A (anti-H5) before the infection of MDCK-SIAT1 cells. Inhibition is visualized as a significantly reduced number of green fluorescent cells. and requires this molecule to infect the MDCK-SIAT1 indicator cells. These results show that the E-S-FLU virus can be used to screen for neutralizing antibodies to Ebola virus.

Inhibition of E-S-FLU virus infection by small molecules.
In the literature, several drug screening studies have been done with wild-type Ebola virus (37)(38)(39)(40)(41) and Ebola virus surrogates such as Ebola virus-like particles (42), EBOV-GP-pseudotyped recombinant VSV (32), and lentivirus (43,44). It was shown that drug molecules with diverse pharmacological functions have inhibited Ebola virus infection in vitro. To verify the use of the E-S-FLU virus in drug screening assays, we selected several types of drug molecules that have been tested repeatedly in the literature and screened them against the E-S-FLU virus in our drug inhibition assay. Table 1 summarizes the 13 initial inhibitors we tested. E-S-FLU virus infection was effectively blocked by different groups of drugs, and the drug concentrations at which eGFP fluorescence was reduced by 50% (IC 50 s) were in the same range as those in the literature.
All of the screenings in Table 1 were repeated at least three times to calculate the geometric mean IC 50 and 95% confidence interval (CI). We also screened these drugs against the H5-S-FLU virus (see Table S1 in the supplemental material). Most of them did not inhibit the H5-S-FLU virus, except niclosamide at 2.64 M. Figure 5 shows an example of drug inhibition by tetrandrine. The infection level was quantified by measuring eGFP expression. We also estimated the toxicity of each drug by staining the cells with fluorescently labeled wheat germ agglutinin (WGA) after infection, fixation, and washing. WGA is a lectin that binds to cell surface N-acetyl-D-glucosamine and sialic acid. Cells in good condition will remain adhered to the plate during analysis and can be detected with WGA. Dead or unhealthy cells will detach and be washed away. WGA staining allows quick indication of drug toxicity in high-throughput plate screening assays. This is verified by light microscopy ( Fig. 5C and D).
LOPAC 1280 library screening. Having established that infection with the E-S-FLU virus was dependent on the NPC1 receptor and was inhibited by specific antibodies and established small molecules, we proceeded to a formal drug screening. We screened 1,280 drug molecules from the LOPAC 1280 library (Sigma-Aldrich) in our drug inhibition assay to test their effects on E-S-FLU and H5-S-FLU virus infections. This library contains 1,280 pharmacologically active small molecules with different molec- ular targets. All of the drugs were screened in the first round at 100 and 5 M. Selected drugs were then titrated once in a drug inhibition assay to calculate their IC 50 s. In total, 215 out of 1,280 molecules inhibited E-S-FLU virus infection to some level before reaching toxicity in vitro ( Fig. 6 and 7; see Table S1). Out of the 215 that inhibited the E-S-FLU virus, 58 had an IC 50 (Table 2) also inhibited the E-S-FLU virus and thus were likely to be acting either on both EBOV-GP and influenza virus HA or at a postentry stage that is independent of the surface glycoprotein. We did not see any molecule that inhibited the H5-S-FLU virus but did not inhibit the E-S-FLU virus. A detailed profile of all the inhibitors is shown in Table S1. These 215 inhibitors belong to a variety of pharmacological groups according to the description on Sigma's website (Fig. 6). Table S1 also includes 13 small molecules tested in addition to the LOPAC 1280 library, to give a total of 228 E-S-FLU virus inhibitors, and of these, 25 inhibited both the E-S-FLU and H5-S-FLU viruses ( Fig. 7 and Table 2). In our screening system, we quantified infection by measuring the overall fluorescence level of cells. Under a fluorescence microscope, it was clear that most drugs reduced the number of fluorescent cells (i.e., reduced the proportion of infected cells) as the drug concentration increased. Twelve drugs, however, caused a reduced overall fluorescence level of infected cells without reducing the number of fluorescent cells significantly. All 12 of these drugs also affected H5-S-FLU virus infection. This suggests that they acted at a postentry phase by limiting the expression of viral proteins. In Table  2, they are annotated as "dimmer" drugs. The other 13 molecules that inhibited both the E-S-FLU and H5-S-FLU viruses reduced the number of infected cells and thus may have affected the function of both glycoproteins.
The IC 50 s of all 228 molecules that inhibited the E-S-FLU virus are plotted in Fig. 7. These include the 215 inhibitors from the LOPAC 1280 library and the additional 13 individual drugs we tested in addition to this library.   Fig. 5). For detailed information on the inhibitors, see Table S1.  Table 3). The results obtained with 442 of them agreed; 23 of these were inhibitors, and 419 had no effect. Of the rest of the drugs, 70 were positive in the LOPAC 1280 screening only and 26 were positive in the screening done by Johansen et al. only. We used a two-tailed Fisher exact test for statistical analysis of the comparison, and the correlation was significant (P Ͻ 0.0001).
One of the main differences between the two screenings is the range of drug concentrations tested. The highest concentration tested in the screening done by Johansen et al. was 13 M, whereas in our screening, most of the drugs were screened at concentrations of Ͼ50 M if toxicity was not reached. For this reason, the great majority of "agreements" occurred in the drugs with an IC 50 of Ͻ10 M in our assay (Table S1). For the drugs that were positive in the LOPAC 1280 screening only, we selected ebastine (E9531), imipramine (I7379), and verapamil (V4629) and ordered them individually from Sigma to confirm their inhibitory effects. All three drugs had an inhibitory effect on the E-S-FLU virus with IC 50 s comparable to those in the LOPAC 1280 library screening. All three drugs have previously been reported as inhibitors (32,37,38,40). Table S1 LOPAC plate no.  Of the 26 drugs that were positive in the screening done by Johansen et al. but negative in our LOPAC 1280 screening, we ordered individually from Sigma and further tested 14, auranofin (Sigma code A6733), bepridil (B5016), calcimycin (C7522), D609 (T8543), diphenyleneiodonium (D2926), doxylamine (D3775), linezolid (PZ0014), naproxen (M1275), nocodazole (M1404), perphenazine (P6402), rotenone (R8875), sanguinarine (S5890), serotonin (H9523), and thapsigargin (T9033).

No. in
Bepridil (a calcium channel inhibitor) and perphenazine (a cationic amphiphile) from the LOPAC 1280 library did not inhibit the E-S-FLU virus, but the individually ordered samples showed specific inhibition of the E-S-FLU virus (bepridil IC 50 ϭ 2.7 M, perphenazine IC 50 ϭ 8.2 M). This indicates some variation due to false negatives in the library preparation from Sigma.
D609, doxylamine, disopyramide, linezolid, naproxen, rotenone, and sanguinarine were confirmed as noninhibitors of the E-S-FLU virus in our assay up to 100 M. Auranofin, calcimycin, diphenyleneiodonium, nocodazole, and thapsigargin were toxic to MDCK-SIAT1 cells even at 0.2 M, yet some cells were still infected. None of the drugs that were further tested inhibited the H5-S-FLU virus.

DISCUSSION
The design of the E-S-FLU virus was based on a pseudotyped influenza virus, the S-FLU virus, that was previously developed by our group and is similar to other single-cycle influenza virus pseudotypes (12,22,51). The E-S-FLU virus is a nonreplicating pseudotyped virus and can be manipulated at biosafety level 1/2, as it does not contain any genetic information from Ebola virus. For ease of production, we found that the MDCK-SIAT1 cell line (21) can stably express full-length EBOV-GP without toxicity. Although replication of the S-FLU virus core in the EBOV-GP-transduced cell line is 2 to 3 orders of magnitude less efficient than that in H5 HA-expressing cells (Fig. 2), production of the pseudotype at a scale sufficient for high-throughput assays is simple and efficient. The E-S-FLU virus encodes eGFP as a reporter in preference to firefly luciferase or beta-galactosidase (9,52,53). Enzyme reporters are much more sensitive than eGFP, and very low-level infection can be amplified and detected. However, cells have to be lysed to release the enzymes and then substrate is added for a one-round readout only. In comparison, for drug inhibition and neutralization assays, the S-FLU virus can be added to indicator cells at a multiplicity of infection (MOI) of Ն1 to produce bright eGFP fluorescence in most cells that is stable over months after fixation and can be measured in a plate reader or observed by microscope without further manipulation ( Fig. 4 and 5).
To validate the E-S-FLU virus as a useful surrogate for Ebola virus, we showed that infection is fully dependent on the NPC1 receptor (Fig. 3), is inhibited by the EBOV-GP-specific MAb KZ52 (35, 36) (Fig. 4), and is also inhibited by a well-characterized set of 13 drugs that are thought to act at various points during viral entry (37-42) ( Table  1 and Fig. 8).
Several laboratories have observed that a wide variety of FDA-approved drugs can inhibit infection of cell lines in vitro by Ebola virus (40,42). We have screened 1,280 drugs from the LOPAC 1280 library to identify pharmacologically active small molecules that inhibit E-S-FLU virus cell entry. The H5-S-FLU virus was tested in parallel with the E-S-FLU virus to differentiate inhibition specific to EBOV-GP or to the S-FLU virus core. We identified 215 molecules from the library that inhibited the E-S-FLU virus, whereas only 22 inhibited the H5-S-FLU virus, all of which also inhibited the E-S-FLU virus. Therefore, the inhibitory effects of 193 drugs were EBOV-GP dependent.
We compared our screening results with the list of drugs screened by Johansen et al. (40), who screened 2,635 small molecules on GFP-expressing Ebola virus. A total of 538 small molecules were common to the two libraries (Johansen et al. versus LOPAC 1280 ). Out of these, 442 molecules had concordant results in the two screenings. The correlation was statistically significant (P Ͻ 0.0001). It is noteworthy that we identified at least 17 calcium channel inhibitors as specific inhibitors of E-S-FLU virus infection. Independent evidence exists for a possible role for calcium channels in the lysosomal membrane in viral entry (37,38), perhaps through an effect on vesicle fusion (32,54,55). Recent studies have suggested that two-pore channel 2 (TPC-2), a lysosomal calcium channel, is involved in the Ebola virus entry process, and this could be the target of the calcium channel inhibitors (38,56).
A large group of inhibitors share a cationic amphiphilic feature, i.e., a hydrophilic end, usually a weak base amine group (high pKa) and a hydrophobic end (high logP), and are therefore likely to concentrate in lysosomes (57)(58)(59). The mechanism by which these drugs inhibit Ebola virus entry has not been resolved (37,39). Some cationic amphiphilic drugs can cause an increase in cholesterol accumulation in endosomes and lysosomes, where EBOV-GP membrane fusion occurs. Others can influence lysosomal membrane stability and indirectly inhibit acid sphingomyelinase, which are also known as functional inhibitors of acid sphingomyelinase (60,61), and may have an indirect effect on calcium mobilization from the lysosome through the accumulation of sphingosine. In addition, the most powerful member of the cationic amphiphilic group, toremifene, has been shown to bind to a hydrophobic pocket of the EBOV-GP and destabilize the molecule in vitro (62) and so may have two sites of action.
Many of these drugs have been found to have a risk of causing long-QT syndrome, such as amiodarone, toremifene, and clarithromycin (63,64). Inhibition of the potas- Journal of Virology sium channel hERG (human ether-a-go-go) in cardiac myocytes is commonly associated with long-QT syndrome (65). By comparing the 228 drugs that inhibited the E-S-FLU virus in our studies with the hERG screening literature (65)(66)(67), we identified at least 46 (20.2%; indicated in Table S1) as confirmed inhibitors of the hERG channel and hence may have this dangerous side effect on cardiac conduction. Most of these hERG binders are typical cationic amphiphilic drugs. It is unclear whether the features that inhibit Ebola virus entry also predispose to the effect on hERG binding and lengthening of the QT interval. Further dissection of their mechanisms of action is therefore necessary before they can be administered to Ebola virus-infected patients, although they are FDA-approved drugs (68,69). As more potential mechanisms of action of these drugs are uncovered, testing of combinations that act at different sites for synergistic effects may be valuable to reduce this risk (70,71). One advantage of our S-FLU virus pseudotype is that infected cells express eGFP and fluoresce brightly. Visual inspection of cells in assay plates by fluorescence microscopy revealed that two forms of inhibition could be distinguished. In most cases, the drugs reduced the number of infected cells. All of the 203 drugs that acted in an EBOV-GPdependent manner reduced the number of infected cells, which is consistent with inhibition of viral entry (Table S1). In contrast, in some cases, the drugs did not reduce the number of infected cells but reduced the intensity of fluorescence expression in the infected cell population. Of the 228 drugs tested, 12 reduced the fluorescence intensity ( Table 2). All of these also inhibited the H5-S-FLU virus. In general, the IC 50 s of these agents were similar for the E-S-FLU and H5-S-FLU viruses, which is consistent with their acting on functions controlled by the influenza virus core, which is common to both. We suggest that they affect the virus at a postentry level like faviparivir, a specific inhibitor of RNA-dependent RNA polymerase (72)(73)(74)(75) that was initially synthesized to inhibit influenza virus but later found to be effective against Ebola virus in vitro. In addition aminopterin, a dihydrofolate reductase inhibitor; brequinar, a dihydroorotate dehydrogenase inhibitor (76)(77)(78)(79); and ribavirin, a guanosine analogue nucleoside inhibitor (71,80,81), may affect transcription indirectly by blocking the biosynthesis of nucleotides. It is interesting that spiranalactone and triamterene both have the dimmer effect. These drugs are used clinically as diuretics that block sodium channels. It is unclear how they can affect the functions of the influenza virus core at a postentry level.
Of the 25 drugs that inhibited both the H5-S-FLU and E-S-FLU viruses, 13 reduced the number of infected cells and thus may have inhibited cell entry by both H5 HA and EBOV-GP. Of these, chloroquine (a lysosomotropic drug) and niclosamide may act by reducing the lysosomal proton concentration, which can influence the efficiency of cathepsin in endosome-digesting EBOV-GP (82). Niclosamide is a proton ionophore (83,84) and had similar IC 50 s on H5-S-FLU virus (3.3 M) and E-S-FLU virus (1.7 M) entry. Arbidol may inhibit the fusion of viral and lysosome membranes in a manner related to its effect on influenza virus HA (85)(86)(87). Although we did not find inhibition by arbidol of the H5-coated S-FLU virus, we have since confirmed the inhibition of S-FLU virus coated with H7 HAs from A/Netherlands/219/2003 and A/Shanghai/1/2013/H7, as well as H1 HAs from A/PR/8/1934 and A/England/195/2009 (data not shown). A recent structure revealed arbidol bound to H7 HA from influenza virus (88); it may be interesting to analyze the binding of arbidol to EBOV-GP.
In summary, we have developed a new surrogate E-S-FLU virus for Ebola virus. It mimics the Ebola virus at the level of cell entry and can be a valuable tool for the screening of antibodies and therapeutic drugs, as well as studying the fundamental biology of the Ebola virus entry process.
Transduced cells were stained with EBOV-GP-specific MAb KZ52, which is specific to Zaire (35,36), or 66-4-C12, which is cross-reactive with all Ebola virus species (P. Rijal, unpublished data), and with a second layer of fluorescein isothiocyanate (FITC)-conjugated goat anti-human antibody (Life Technologies reference no. H10301). Stained cells were bulk sorted with a fluorescence-activated cell sorter (FACS) to achieve maximal expression of EBOV-GP (E-SIAT cell line, shown in Fig. 1). EBOV-GP was detected on Ͼ99% of the sorted cells and was stable and uniform up to at least to the ninth passage after sorting.
The E-S-FLU virus was generated on the basis of a nonreplicating pseudotyped influenza virus (S-FLU) that was previously described (12,90). We used the version that expresses enhanced green fluorescent protein (S-eGFP), which has its S-HA coding sequence replaced with an eGFP reporter gene. We isolated subclones of the original S-FLU virus that showed brighter fluorescence and found a single T60C mutation in the A/PR/8/34 HA signal sequence (NCBI reference sequence accession no. NC_002017.1) that abrogates an out-of-frame ATG, which had interfered with downstream eGFP expression. All subsequent versions of the S-FLU virus contain this enhancement, which gives a sufficiently bright fluorescence signal from infected cells to read on the CLARIOstar fluorescence plate reader. Seed viruses were doubly cloned by limiting dilution before the seeding of E-SIAT or H5-SIAT producer cells.
Seed S-FLU viruses coated in H5, H7, or H1 HA were titrated by standard 50% tissue culture infective dose (TCID 50 ) assay on MDCK-SIAT1 cells by fluorescence detection (as described below). Titration experiments revealed that a relatively high MOI of approximately 1 TCID 50 of seed S-FLU virus per cell added to infect E-SIAT cells produced the highest titers of the E-S-FLU virus, whereas a lower MOI (ϳ0.01) was adequate for seeding of HA-SIAT cell lines (12). The viral growth medium (VGM) used was Dulbecco's modified Eagle's medium (DMEM) with 0.1% bovine serum albumin, 10 mM HEPES, 2 mM glutamine, 100 U/ml penicillin, and 100 g/ml streptomycin. After 2 h of incubation of E-SIAT cells with seed S-FLU virus, any remaining seed virus was removed by washing with phosphate-buffered saline (PBS). Cells were then incubated in VGM for 48 h at 37°C in a 5% CO 2 incubator without added trypsin. Culture supernatant was harvested and stored in aliquots at Ϫ80°C. Replication of HA-pseudotyped viruses required the addition of trypsin as described previously (12).
Seed virus TCID 50 . Seed S-FLU viruses coated in influenza virus HA were titrated for the concentration of replication-competent clones (TCID 50 ) by a standard limiting-dilution method as described by Powell et al. (12). Wells containing clones of eGFP-expressing S-FLU virus Ͼ4 standard deviations above 16 control wells were detected with a plate reader (described below), and the TCID 50 /ml was calculated by the Reed-Muench method (91).

S-FLU virus batch CID 50 .
Batches of the E-S-FLU and H5-S-FLU viruses were titrated for infectivity of MDCK-SIAT1 indicator cells after overnight infection. Harvested culture supernatants (100 l) containing the E-S-FLU or H5 S-FLU virus were titrated in 2-fold dilutions in VGM across a flat-bottom 96-well plate ( Fig. 2A). MDCK-SIAT1 indicator cells were washed in PBS, and 3 ϫ 10 4 cells in 100 l of VGM were added to each well and incubated for 18 h (37°C, 5% CO 2 ). Medium was removed from the wells, and cells were fixed in 10% formalin (in PBS) for 30 min at 4°C. The virus titer was assessed by eGFP fluorescence analyzed with a plate reader (CLARIOstar). The dilution of virus giving 50% of the maximum plateau fluorescence signal (EC 50 ) was calculated by linear interpolation. The EC 50 dilution obtained by best fit of the titration curve to the Poisson distribution was very similar (data not shown). The typical EC 50 dilutions for the E-S-FLU virus harvested after 48 h were ϳ1:8, compared to ϳ1:500 for the H5-S-FLU virus (Fig. 2). From the EC 50 dilution and the number of cells per well (3 ϫ 10 4 ), the CID 50 /ml was calculated as ϳ2 ϫ 10 6 in the E-S-FLU virus batch and ϳ1.6 ϫ 10 8 in the H5 S-FLU virus batch. For inhibition assays, the H5-S-FLU and E-S-FLU viruses were used at a concentration that gave close-to-maximum plateau fluorescence in infected MDCK-SIAT1 cells (MOI, ϳ4 CID 50 /cell). This value was used to normalize the fluorescence readout to a percentage of the maximum in the figures (Fig. 4A and B and 5A).
In this study, all of the evaluation and screening was done with the Zaire (C15) E-S-FLU and H5-S-FLU viruses in parallel.
CLARIOstar plate reader setting. Fluorescence plate readout was performed with Costar 96-well cell culture plates (Costar no. 3799). We chose the COSTAR 96 microplate mode in the CLARIOstar software. For measurement of eGFP levels, fluorescence excitation was at 483 nm (bandwidth of 8 nm) and emission was at 515 nm (bandwidth of 8 nm) with a 499-nm dichroic filter. Each well had 50 flashes at a 4-mm diameter and fluorescence readout from the top optic, giving the orbital averaging value. For measurement of WGA (Alexa Fluor 647 conjugate) in the drug inhibition assay, fluorescence excitation was at 630 nm (bandwidth of 30 nm) and emission was at 678 nm (bandwidth of 20 nm). Each well had 53 flashes at 4 mm and fluorescence readout from the top optic, giving the orbital averaging value. The focal length and gain were adjusted before each plate was read. The generally optimal focal length was 5.0 mm, the gain of the eGFP channel was adjusted to 3,000, and the gain for WGA-647 was adjusted to 2,500.

E-S-FLU virus infection of NPC1 KO cells.
The NPC1 protein is the key receptor for Ebola virus cell entry at the level of the lysosomal membrane (31)(32)(33). Two NPC1 KO HeLa cell lines (ex2 NPC1-KO and ex4 NPC1-KO, named for the exon that was deleted) were generated by the CRISPR-cas9 technique. Generation and verification of the NPC1 KO cell lines have been described previously by Tharkeshwar et al. (34).
A total of 1 ϫ 10 5 wild-type, ex2 NPC1-KO, or ex4 NPC1-KO HeLa cells were seeded into a 24-well tissue culture plate in 500 l of D10 medium (DMEM with 10% fetal calf serum, 2 mM glutamine, 100 U/ml penicillin, and 100 g/ml streptomycin) the night before infection. Before infection, the medium was removed and cells were washed once with PBS. Cells were infected with maximal doses of 1 ϫ 10 6 CID 50 of the E-S-FLU virus (MOI, ϳ10 CID 50 /cell) or 8 ϫ 10 7 CID 50 of the H5-S-FLU virus (MOI, ϳ800 CID 50 /cell) and incubated overnight. After infection, cells were trypsinized and transferred to 5-ml polypropylene tubes. Percent infection was analyzed by counting fluorescent cells in a flow cytometer (CyAn ADP analyzer).
MN assay. The MN assay used measures the inhibition titer of antibodies against virus entry in vitro. The assay was done as described by Powell et al. (12,92), with minor modifications. The E-S-FLU (MOI, ϳ4 CID 50 /cell) and H5-S-FLU (MOI, ϳ4 CID 50 /cell) viruses were diluted in VGM to give plateau expression of eGFP in 3 ϫ 10 4 MDCK-SIAT1 cells after overnight infection in 96-well flat-bottom plates. A 50-l volume of virus was preincubated with 50 l of antibody at 2-fold dilutions for 2 h. A total of 3 ϫ 10 4 MDCK-SIAT1 cells in 100 l of VGM were then seeded into each well, and overnight infection was allowed (37°C, 5% CO 2 ). After infection, the medium was removed from the wells and the cells were fixed with 100 l of 10% formalin (in PBS) for 30 min at 4°C; the formalin was then replaced with 100 l of PBS. The plates were analyzed with a fluorescence plate reader (CLARIOstar) and by fluorescence microscopy. The IC 50 was calculated by linear interpolation. KZ52 is a human MAb that specifically binds Zaire EBOV-GP and neutralizes wild-type EBOV in vitro (35,36,93). Our version was reconstituted from the sequence published in the crystal structure PDB file (PDB code 3CSY) (36). 21D85A is a human MAb made in-house that specifically binds to H5 HA and neutralizes the H5-S-FLU virus.
The EBOV-GP-pseudotyped lentivirus (GP-lenti) we used in this study is coated with the same EBOV-GP from Zaire C15 as the E-S-FLU virus. GP-lenti carries a firefly luciferase reporter gene. Fiftymicroliter volumes of GP-lenti pseudotyped virus were preincubated with 50 l of antibodies at 2-fold dilutions for 2 h, and 3 ϫ 10 4 MDCK-SIAT1 cells were added. Cells were incubated with virus for infection in VGM for 48 h. After 48 h, infection was quantified by measuring luciferase reporter expression. Medium was removed from the wells, and cells were lysed with 50 l of Glo-lysis buffer (Promega) for 5 min, and then 50 l of Bright-Glo (Promega) substrate was added for a 5-min enzymatic reaction. Solutions were transferred to opaque plates for luminescence measurement with the GloMax-Multi detection system (Promega).
Drug inhibition assay. All small inhibitor molecules, including the LOPAC 1280 library used in this study, were ordered from Sigma-Aldrich and stored in 10 mM dimethyl sulfoxide at Ϫ20°C, with the exception of AY9944, which was ordered from Merck (catalog no. 190080), and trans-Ned19 (provided by our collaborator Antony Galione). Drugs were diluted in VGM to give a starting concentration of 100 M (unless stated otherwise) and then titrated in 50 l of VGM at 2-fold dilutions across a 96-well plate. A total of 3 ϫ 10 4 MDCK-SIAT1 cells were seeded in 50 l of VGM and preincubated with drugs for 3 h (37°C, 5% CO 2 ). A 100-l volume of the E-S-FLU (MOI, ϳ4 CID 50 /cell) or H5-S-FLU (MOI, ϳ4 CID 50 /cell) virus diluted in VGM to give plateau expression of eGFP in 3 ϫ 10 4 MDCK-SIAT1 cells was then added for overnight infection. Cells were fixed with 100 l of 10% formalin (in PBS) for 30 min at 4°C, the formalin was then replaced with 100 l of PBS, and the plates were analyzed with a fluorescence plate reader (CLARIOstar) as for the MN assay. Infection and inhibition were quantified by measurement of the eGFP expression level. When more than three measurements were made for a drug, the 95% CI was calculated ( Table 1).
Because of the toxic effect on cells of many drug molecules at high concentrations, dead cells were washed away before plate reading. This may give a falsely low eGFP fluorescence reading in the plate reader but can be detected by visual inspection in a fluorescence microscope. To estimate the number of cells remaining in each well, plates were stained with WGA after cells were fixed and washed once with PBS. WGA is a lectin that binds to sialic acid and N-acetylglucosaminyl residues at the cell surface. A 50-l volume of 5 g/ml WGA (Alexa Fluor 647 conjugate; ThermoFisher catalog no. W32466) was added to each well and incubated at room temperature for 15 min. Plates were then washed twice with PBS before being read (see the plate reader setting mentioned above).  Table S1, they listed all of the experimental data for the small molecules they screened in their study. Of the 2,635 molecules they tested, 538 were also found in the sigma LOPAC 1280 library and were screened against the E-S-FLU virus in our study. We compared the inhibitory effects of these 538 compounds in both screenings (i.e., each molecule was either positive [inhibited infection] or negative [did not inhibit infection in the concentration range tested]) and performed a Fisher exact test to measure the significance of the association.
LogP and pKa calculations. For the drug molecules listed in Table 1, logP (partition parameter) and pKa values (strongest basic pKa) were estimated on the basis of their chemical structures by using the MarvinSketch program.