Infections by the Ebola (EboV) and Marburg (MarV) filoviruses cause a rapidly fatal hemorrhagic fever in humans for which no approved antivirals are available1. Filovirus entry is mediated by the viral spike glycoprotein (GP), which attaches viral particles to the cell surface, delivers them to endosomes, and catalyzes fusion between viral and endosomal membranes2. Additional host factors in the endosomal compartment are likely required for viral membrane fusion. However, despite considerable efforts, these critical host factors have defied molecular identification3,4,5. Here we describe a genome-wide haploid genetic screen in human cells to identify host factors required for EboV entry. Our screen uncovered 67 mutations disrupting all six members of the HOPS multisubunit tethering complex, which is involved in fusion of endosomes to lysosomes6, and 39 independent mutations that disrupt the endo/lysosomal cholesterol transporter protein Niemann-Pick C1 (NPC1)7. Cells defective for the HOPS complex or NPC1 function, including primary fibroblasts derived from human Niemann-Pick type C1 disease patients, are resistant to infection by EboV and MarV, but remain fully susceptible to a suite of unrelated viruses. We show that membrane fusion mediated by filovirus glycoproteins and viral escape from the vesicular compartment requires the NPC1 protein, independent of its known function in cholesterol transport. Our findings uncover unique features of the entry pathway used by filoviruses and suggest potential antiviral strategies to combat these deadly agents.
We have developed haploid genetic screens to gain insight into biological processes relevant to human disease8,9. Here we use this approach to explore the filovirus entry pathway at unprecedented level of detail. To interrogate millions of gene disruption events for defects in EboV entry, we used a replication-competent vesicular stomatitis virus bearing the EboV glycoprotein (rVSV-GP-EboV)10. Although this virus replicates in most cell lines, it inefficiently killed near-haploid KBM7 cells (Figure S1C). In an unsuccessful attempt to induce pluripotency in KBM7 cells by expression of OCT4, SOX-2, c-MYC and KLF411, we obtained HAP1 cells (Figure S1A). HAP1 cells grew adherently and no longer expressed hematopoietic markers (Figure S1B). The majority of these cells in early passage cultures were haploid for all chromosomes, including chromosome 8 (which is diploid in KBM7 cells). Unlike KBM7 cells, HAP1 cells were susceptible to rVSV-GP-EboV (Figure S1C) allowing screens for filovirus host factors.
We used a retroviral gene-trap vector9 to mutagenize early-passage HAP1 cells. To generate a control dataset, we mapped ~800,000 insertions using deep sequencing (Table S1). Next, we selected rVSV-GP-EboV-resistant cells, expanded them as a pool, and mapped insertion sites. Enrichment for mutations in genes was calculated by comparing a gene’s mutation frequency in resistant cells to that in the control dataset (Figure S2). We identified a set of genes enriched for mutations in the rVSV-GP-EboV-resistant cell population (Figure 1A, S3 and Table S2). Nearly all of these candidate host factors are involved in the architecture and trafficking of endo/lysosomal compartments. Gratifyingly, our screen identified cathepsin B (CatB), the only known host factor whose deletion inhibits EboV entry5. Further inspection showed that mutations were highly enriched in all 6 subunits of the homotypic fusion and vacuole protein-sorting (HOPS) complex (VPS11, VPS16, VPS18, VPS33A, VPS39 and VPS41), for which we identified 67 independent mutations. The HOPS complex mediates fusion of endosomes and lysosomes6 and affects endosome maturation12,13. The identification of all members of the HOPS complex demonstrates high, and possibly saturating, coverage of our screen. We also identified factors involved in biogenesis of endosomes (PIKFYVE, FIG4)14, lysosomes (BLOC1S1, BLOC1S2)15, and in targeting of luminal cargo to the endocytic pathway (GNPTAB)16. The strongest hit was the Niemann-Pick disease locus NPC1, encoding an endo/lysosomal cholesterol transporter7. NPC1 also affects endosome/lysosome fusion and fission17, calcium homeostasis18 and HIV-1 release19.
We subcloned the resistant cell population to obtain clones deficient for VPS11 and VPS33A, and NPC1 (Figure S4A, B and Figure 1B). These mutants displayed marked resistance to infection by rVSV-GP-EboV and VSV pseudotyped with EboV or MarV GP (Figure 1C and Figure S4C). Cells lacking a functional HOPS complex or NPC1 were nonetheless fully susceptible to infection by a large panel of other enveloped and nonenveloped viruses, including VSV and recombinant VSV bearing different viral glycoproteins (Figure 1D and S5). The susceptibility of HAP1 clones to rVSV-GP-EboV infection was restored by expression of the corresponding cDNAs (Figure S6A, B, C).
Loss of NPC1 causes Niemann-Pick disease, a neurovisceral disorder characterized by cholesterol and sphingolipid accumulation in lysosomes7. We tested susceptibility of patient primary fibroblasts to filovirus GP-dependent infection. NPC1-mutant cells were infected poorly or not at all by rVSV-GP-EboV and VSV pseudotyped with filovirus GP proteins (Figure 2A, B), and infection was restored by expression of wild type NPC1 (Figure 2C).
Mutations in NPC2 cause identical clinical symptoms and phenocopy defects in lipid transport20. Surprisingly, NPC2-mutant fibroblasts derived from different patients were susceptible to filovirus GP-dependent infection (Figure 2A and Figure S7), despite a similar accumulation of cholesterol in NPC2- and NPC1-mutant cells (Figure 2B). Moreover, cholesterol clearance from NPC1-null cells by cultivation in lipoprotein-depleted growth medium did not confer susceptibility (Figure S8). Therefore, resistance of NPC1-deficient cells to rVSV-GP-EboV is not caused by defects in cholesterol transport per se.
Filoviruses display broad mammalian host and tissue tropism21,22. To determine if NPC1 is generally required for filovirus GP-mediated infection, we used NPC1-null Chinese hamster ovary (CHO) cells. Loss of NPC1 conferred complete resistance to viral infection (Figure S6D) that was reversed by expression of human NPC1 (Figure S6E). Certain small molecules such as U18666A23 and the antidepressant imipramine24 cause a cellular phenotype similar to NPC1 deficiency possibly by targeting NPC123. Prolonged U18666A treatment was reported to modestly inhibit VSV25. However, we found that brief exposure of Vero cells and HAP1 cells to U18666A or imipramine potently inhibited viral infection mediated by EboV GP but not VSV or rabies virus G (Figure 2D, S9, and S10). Because U18666A inhibits rVSV-GP-EboV infection only when added at early time points, it likely affects entry rather than replication (Figure S10). Thus, NPC1 has a critical role in infection mediated by filovirus glycoproteins that is conserved in mammals and likely independent of NPC1’s role in cholesterol transport.
Filoviruses bind to one or more cell-surface molecules2,26,27 and are internalized by macropinocytosis28,29. In VPS33A- and NPC1-mutant cells, we observed no significant differences in binding or internalization of Alexa 647-labeled rVSV-GP-EboV (Figure 3A, Figure S11 and Figure S12A). Similar results were obtained by flow cytometry using fluorescent EboV virus-like particles (Figure S12B). Moreover, bullet-shaped VSV particles were readily observed by electron microscopy at the cell periphery and within plasma membrane invaginations resembling nascent macropinosomes (Figure 3B). Finally, VPS33A- and NPC1-null cells were fully susceptible to vaccinia virus entry by macropinocytosis (Figure S13). Thus, GP-mediated entry is not inhibited at viral attachment or early internalization steps in NPC1- or HOPS-defective cells, suggesting a downstream defect.
Cathepsin L (CatL)-assisted cleavage of EboV GP by CatB is required for viral membrane fusion3,5. Mutant HAP1 cells possess normal CatB/CatL activity (Figure S14B, C) and were fully susceptible to mammalian reoviruses, which utilize CatB or CatL for entry (Fig. S14D). Moreover, these cells remained refractory to in vitro-cleaved rVSV-GP-EboV particles (Figure 3C) that no longer required CatB/CatL activity within Vero cells (Figure S14A). Therefore the HOPS complex and NPC1 are likely required downstream of the initial GP proteolytic processing steps that generate a primed entry intermediate.
Finally, we used the intracellular distribution of the internal VSV M (matrix) protein as a marker for membrane fusion (Figure 3D). Cells were infected with native VSV or rVSV-GP-EboV and immunostained to visualize the incoming M protein. Endosomal acid pH-dependent entry of either virus into wild type HAP1 cells caused redistribution of the incoming viral M throughout the cytoplasm (Figure 3D) (Figure S15A). By contrast, only punctate, perinuclear M staining was obtained in drug-treated and mutant cells infected with rVSV-GP-EboV or rVSV-GP-MarV (Figure 3D and Figure S15B). Electron micrographs of mutant cells infected with rVSV-GP-EboV revealed agglomerations of viral particles within vesicular compartments (Figure 3E and S16A) containing LAMP-1 (Figure S16B), suggesting that fusion and uncoating of incoming virus is arrested. Similarly, U18666A treatment increased the number of viral particles in NPC1-and LAMP1-positive endosomes (Figure S17). Therefore, NPC1 and the HOPS complex are required for late step(s) in filovirus entry leading to viral membrane fusion and escape from the lysosomal compartment.
We next tested if infection by authentic EboV and MarV is affected in NPC1-mutant primary patient fibroblasts. Yields of viral progeny were profoundly reduced for both viruses in mutant cells (Figure 4A). Stark reductions in viral yield were also obtained in Vero cells treated with U18666A (Figure 4B). Moreover U18666A greatly reduced infection of human peripheral blood monocyte-derived dendritic cells and umbilical-vein endothelial cells (HUVEC) (Figure 4C), without affecting cell number or morphology (Figure S19). Finally, knockdown of NPC1 in HUVEC diminished infection by filoviruses (Figure 4D and S18). These findings indicate that NPC1 is critical for authentic filovirus infection.
We assessed the effect of NPC1 mutation in lethal mouse models of EboV and MarV infection. Heterozygous NPC1 (NPC1−/+) knockout mice and their wild type littermates were challenged with mouse-adapted EboV or MarV and monitored for 28 days. Whereas NPC1+/+ mice rapidly succumbed to infection with either filovirus, NPC1−/+ mice were largely protected (Figure 4E).
We have used global gene disruption in human cells to discover components of the unusual entry pathway used by filoviruses. Most of the identified genes affect aspects of lysosome function, suggesting that filoviruses exploit this organelle differently from all other viruses that we have tested (Figure 4F). The unanticipated role for the hereditary disease gene NPC1 in viral entry, infection, and pathogenesis may facilitate the development of anti-filovirus therapeutics.