Filovirus receptor NPC1 contributes to species-specific patterns of ebolavirus susceptibility in bats

1Ebola virus (EBOV) and some of its relatives in the family Filoviridae (filoviruses) cause sporadic outbreaks of a highly lethal disease. These outbreaks are thought to be initiated by viral spillover from an animal reservoir to a highly susceptible accidental host, such as a human or nonhuman primate. Recent work suggests that some filoviruses infect bats in nature, and that these viruses may be distributed more widely than previously recognized. Very short RNA fragments corresponding to portions of ebolavirus genomes were detected in several frugivorous bats of the family Pteropodidae (‘Old World fruit bats’) in both Africa and Asia, and longer filovirus RNA fragments and near-complete RNA genomes were isolated from insectivorous Schreibers's long-fingered bats in Asia and Europe, respectively. However, despite considerable efforts, infectious ebolaviruses have never been recovered from bats. By contrast, Marburg (MARV) and Ravn (RAVV) viruses were found to circulate in Egyptian rousettes (Rousettus aegyptiacus), indicating that these bats are susceptible to MARV/RAVV and encounter them frequently in nature. Egyptian rousettes have been proposed as natural hosts for these viruses. This progress notwithstanding, many key questions remain. For example, the biological factors that influence filovirus host range and interspecies transmission are still poorly understood, as are the virus-host relationships that determine which species of bats are susceptible to infection by EBOV and other filoviruses.

2Viral entry receptors are key determinants of tissue tropism and host range. Niemann-Pick C1 (NPC1), a highly conserved endo/lysosomal protein involved in cellular cholesterol trafficking, was recently identified to be an essential entry receptor for all known filoviruses. In this study, we uncover a pattern of virus and host species specificity in the filovirus susceptibility of bat cells, which can be explained by changes in the affinity of the essential interaction between NPC1 and the filovirus entry glycoprotein, GP. Crucially, genetic analyses reveal that NPC1 is under positive selection in bats, with a strong signature of selection at precisely the same residue that influences the filovirus-receptor interaction. Our findings suggest that amino acid sequence changes in NPC1 at these positively-selected sites represent host adaptations to resist filovirus infection, and reveal one pathway by which a filovirus could escape from receptor control. In sum, our results support the hypothesis that bats and filoviruses have been engaged in a long-term co-evolutionary relationship, one facet of which is a molecular arms race between the viral glycoprotein and its entry receptor, NPC1.

3We first explored the possibility that there exist virus- and/or bat species-dependent differences in the cellular host range of filoviruses. Kidney fibroblast cell lines derived from three African pteropodids whose ranges overlap the locations of known African filovirus disease outbreaks (Figure 1A,B) were exposed to authentic EBOV and MARV (Figure 1C). We observed a large EBOV infection defect in African straw-colored fruit bat (Eidolon helvum) cells but not in cells from Büttikofer’s epauletted fruit bats (Epomops buettikoferi) and Egyptian rousettes. By contrast, cells from bats of all three species were similarly susceptible to infection by MARV (Figure 1C). Thus, cells from African straw-colored fruit bats appear to be selectively refractory to EBOV infection.

4The viral spike glycoprotein, GP1,2 (herein termed GP) mediates all steps of filovirus entry into the cytoplasm of host cells. Vesicular stomatitis viruses bearing filovirus GP proteins (VSV pseudotypes) provide a highly validated surrogate system to recapitulate filovirus entry under biosafety level 2 containment. To assess whether the EBOV infection defect in the African straw-colored fruit bat cells occurs at the viral entry step, we exposed an expanded panel of kidney fibroblast cell lines from four African pteropodids to VSV pseudotypes bearing GP spikes (VSV-GP) from seven filoviruses, including two non-African viruses, Reston virus (RESTV) and Lloviu virus (LLOV) (Figure 1D). As observed with authentic EBOV, VSV-EBOV GP infection was substantially reduced in the African straw-colored fruit bat cells; however, this virus could efficiently infect cells derived from the other pteropodids, including those of a proposed EBOV host, the hammer-headed fruit bat (Hypsignathus monstrosus). Strikingly, only VSVs bearing EBOV GP, and to a lesser degree, those bearing BDBV and TAFV GP, were deficient at infecting African straw-colored fruit bat fibroblasts. Similar strong but EBOV-specific reductions in infection were measured in two kidney and lung cell lines derived from additional African straw-colored fruit bats (Figure 1—figure supplement 1). Therefore, reduced infection of these bat cells by EBOV reflects a virus- and host species-specific restriction at the cell entry step.

5We surmised that the filovirus receptor, NPC1, might explain the selective resistance of the African straw-colored fruit bat cells to EBOV entry and infection. Accordingly, we engineered these cells to stably express human NPC1 (HsNPC1) (Figure 2—figure supplements 1,2), and then exposed them to EBOV (Figure 2A). Provision of HsNPC1 substantially enhanced authentic EBOV infection in the African straw-colored fruit bat cells. By contrast, we found no evidence that either MARV infection in these cells, or EBOV/MARV infection in permissive Büttikofer’s epauletted fruit bat cells was limited by receptor availability (Figure 2A). Finally, similar results were obtained with VSVs bearing filovirus glycoproteins (Figure 2B). Taken together, these findings indicate that EBOV infection is reduced in African straw-colored fruit bat cells because of a specific molecular incompatibility between the EBOV glycoprotein and the filovirus entry receptor.

6Although EBOV entry and infection in African straw-colored fruit bat cells was consistently reduced to 0.1–1% relative to that in cells from the other pteropodids, we noted that infection was not completely blocked. To determine if EBOV could inefficiently infect these bat cells via an NPC1-independent mechanism, we used CRISPR/Cas9 genome engineering to derive an African straw-colored fruit bat cell line fully deficient in NPC1. We identified a single cell clone (Eidolon helvum NPC1–#1 [EhNPC1–#1]) in which all NPC1 alleles bore insertions or deletions (indels) at the expected site (Figure 3A). These indels were predicted to frameshift the NPC1 open reading frame at amino acid position 81 (Homo sapiens HsNPC1 numbering), generating truncated polypeptides of 82, 83, and 109 residues that lacked the majority of the 1278-amino acid NPC1 sequence. EhNPC1–#1 cells were deficient in clearance of lysosomal cholesterol, a well-established cellular function of NPC1, but could be rescued by ectopic HsNPC1 expression, confirming that NPC1 had indeed been disrupted in these cells (Figure 3B).

7We next exposed wild-type (WT) and EhNPC1–#1 fibroblasts to VSVs bearing EBOV or MARV GP. No detectable infection was obtained with either virus in NPC1-deficient cells, indicating that filovirus entry into these cells is absolutely dependent on the E. helvum NPC1 ortholog (Figure 3C). Moreover, EBOV GP-dependent infection in EhNPC1–#1 cells reconstituted with HsNPC1 was dramatically enhanced over that observed in WT cells, whereas MARV GP-dependent infection was rescued by HsNPC1 expression to a level resembling that in WT cells (Figure 3C). Therefore, the low levels of EBOV infection in African straw-colored fruit bat cells likely arise from the weak, but non-zero, activity of EhNPC1 as an EBOV entry receptor.

8Filovirus GPs must directly engage the second luminal domain of NPC1, domain C, during cell entry. Accordingly, we postulated that the African straw-colored fruit bat NPC1 ortholog is poorly recognized by EBOV GP. To test that hypothesis, we generated and sequenced NPC1 cDNAs from all four pteropodid cell lines. Alignment of their domain C amino acid sequences with that of HsNPC1 revealed a high degree of conservation (>90%), with identical arrangements of cysteine residues and similar predicted secondary structures suggestive of a similar overall fold (Figure 3—figure supplement 1).

9To examine GP-NPC1 binding, we engineered and expressed soluble forms of the four pteropodid NPC1 domain Cs, as described for HsNPC1 (Figure 4—figure supplement 1). A cleaved form of EBOV GP could capture HsNPC1 domain C in an ELISA, as shown previously. EBOV GP bound with similar avidity to NPC1 domain Cs derived from Egyptian rousettes (RaNPC1), hammer-headed fruit bats (HmNPC1) and Büttikofer’s epauletted fruit bats (EbNPC1), but poorly or not at all to that of African straw-colored fruit bats (EhNPC1) (Figure 4A). Like the infection defect in African straw-colored fruit bat cells, this receptor binding defect was selective for EBOV GP, since GPs derived from MARV and the European filovirus, LLOV, bound equivalently to all four pteropodid domain Cs (Figure 4A). These findings strongly suggest that African straw-colored fruit bat cells are poorly susceptible to EBOV infection because EBOV GP poorly recognizes their ortholog of the filovirus receptor, NPC1.

10To define the molecular basis of the defect in interaction between EBOV and EhNPC1, we generated a panel of NPC1 domain C chimeras comprising sequences from permissive RaNPC1 and nonpermissive EhNPC1, and tested them in the GP-binding ELISA. A single chimera, EhNPC1 domain C containing four EhNPC1→RaNPC1 amino acid residue changes, regained the capacity to efficiently recognize EBOV GP (Figure 4B). Further dissection revealed that only a single amino acid change, F502D, in a central region of NPC1 domain C was needed to effect this complete restoration in GP-NPC1 binding (Figure 4C). The presence of residue F or D at position 502 for all four pteropodid NPC1 proteins tested was fully concordant with EBOV GP-NPC1 binding and NPC1 receptor function, and was also true for HsNPC1 (D, permissive; F, restrictive) (Figures 1,2, Figure 3—figure supplement 1). Finally, analysis of NPC1 sequences derived from nine additional wild-caught African straw-colored fruit bats confirmed that the presence of residue F at position 502 is a conserved feature of EhNPC1 (Figure 4—figure supplement 2). We conclude that a species-specific defect in virus-receptor interaction, caused by a single amino acid residue change in EhNPC1 relative to other, permissive African pteropodid NPC1 orthologs, reduces EBOV infection in African straw-colored fruit bat cells. Moreover, because residues in the NPC1-binding site are conserved among all available EBOV GP sequences (Supplementary file 1), this restriction is almost certain to be encountered by all known EBOV variants and their isolates, including those detected in EBOV disease patients during the recent epidemics in Western and Middle Africa.

11Previous work has led to the hypothesis that bats in equatorial Africa and elsewhere harbor filoviruses (reviewed in). These results, together with our findings for virus- and host species-specific differences in cellular susceptibility to filovirus infection, hinted at the possibility of a deeper co-evolutionary relationship between filoviruses and bats. One hallmark of such a relationship between a virus and its host is the evolution, under selective pressure to resist infection, of host genes encoding proviral and antiviral factors. To evaluate whether the NPC1 gene has evolved under positive selection in bats, we combined the NPC1 sequences obtained in this study with those of bats from six other species (two non-African pteropodids, two phyllostomids, and two vespertilionids) compiled through assembly of publicly available RNAseq data (Figure 5, Figure 5—figure supplement 1, Supplementary files 2–4). We then analyzed this NPC1 multiple alignment (Supplementary file 4) for codon positions enriched for nonsynonymous substitutions relative to synonymous substitutions (dN/dS>1), indicative of positive, or diversifying, selection in favor of amino-acid altering mutations. Using four common tests, we found strong evidence for positive selection in bat NPC1 (Figure 5—figure supplement 1).

12Examination of the specific NPC1 codons with dN/dS>1 shed light on the nature of this selective pressure—18 codons were enriched but only one codon, position 502, was identified by all four tests (Figure 5B,C, Figure 5—figure supplement 1). Strikingly, this is the same position in NPC1 at which a mutation in EhNPC1 reduces receptor binding to EBOV GP and viral infection, a phenotype that could reasonably produce a selective advantage (Figure 4). Other codons identified in only some of the tests for dN/dS>1, or at slightly lower significance levels, may still have functional significance. For example, additional codons were identified in two regions of domain C that may form a part of the recognition surface for EBOV GP (Figure 5C). Our finding that signatures of accelerated sequence evolution localize to structural features in NPC1 that are important for virus binding (domain C and position 502) leads us to postulate that mutations at these sites can protect bats from infection or severe disease caused by filoviruses and/or other intracellular microbes.

13Co-evolutionary arms races between hosts and pathogens are thought to be driven by cycles of genetic adaptation and counter-adaptation. In this context, we postulated that mutation of residue 502 in EhNPC1 could be countered by viral mutation. To identify such putative compensatory viral changes, we screened a panel of point mutants in the NPC1-binding site of EBOV GP by ELISA for enhanced binders to EhNPC1 domain C. While no single point mutant bound to EhNPC1 as well as it did to the other pteropodid NPC1s or to HsNPC1, GP(V141A) partially restored EhNPC1 binding (Figure 6A). Infection by VSV particles bearing EBOV GP(V141A) was substantially enhanced in African straw-colored fruit bat cells, commensurate with this mutant GP’s increased binding affinity for EhNPC1 (Figure 6B). Examination of the X-ray crystal structure of EBOV GP revealed that V141 is located at the edge of the putative NPC1-binding site, where it forms part of a raised rim (Figure 6C). The V141A mutation likely creates a more sterically favorable (open) NPC1-binding site that can overcome the structural mismatch at the GP-NPC1 binding interface (Figure 6C).

14Although no known EBOV isolate contains the V141A mutation, we observed that LLOV and Sudan virus (SUDV) GP naturally possess A141 (Figure 6D). Because both GP proteins could mediate efficient viral entry into African straw-colored fruit bat cells (Figure 1D) and bind to EhNPC1 (Figure 4A data not shown for SUDV), we postulated that amino acid changes at position 141 in the GP receptor-binding site broadly influence the capacity of filovirus glycoproteins to utilize EhNPC1 for viral entry. Accordingly, we exposed pteropodid kidney fibroblasts to VSV pseudotypes bearing SUDV GP(WT) or SUDV GP(A141V) (Figure 6E). Consistent with our hypothesis, the A141V mutation substantially reduced SUDV GP-dependent infection in African straw-colored fruit bat cells. Unexpectedly, this mutant virus also infected Egyptian rousette cells significantly less well than WT, pointing to the existence of sequence context-dependent effects that selectively affect SUDV GP(A141V) binding to RaNPC1 (Figure 6E). These findings provide evidence that GP residue 141 can influence cellular susceptibility to infection by modulating NPC1 recognition in a manner that depends on the sequences of both proteins. We speculate that sequence variation at residue 141 and potentially other positions in the receptor-binding site of filovirus glycoproteins has been shaped by selective pressure to utilize restrictive NPC1 receptors, with potential consequences for viral host range and virulence.

15The ongoing, unprecedented Ebola virus disease epidemic in Western Africa highlights the urgent need to uncover the biological and ecological factors that underlie the distribution, evolution, and emergence of filoviruses. While a full answer to this question will require the integration of knowledge across multiple levels of biological organization, from genes to populations to ecosystems, previous work has shown that studies of molecular interactions between viruses and their host cells can contribute important pieces to this puzzle. The essential interactions between viruses and their entry receptors provide particularly cogent examples. A switch in receptor binding from the feline to the canine ortholog of the transferrin receptor drove the emergence of a new virus, canine parvovirus, and fueled a global disease pandemic in dogs. Analyses of interactions of SARS-like coronaviruses with their receptor ACE2 have helped to trace the emergence of SARS coronavirus from bats to humans, and its use of civets as intermediate amplifying hosts.

16In this study, we show that interactions between filoviruses and their entry receptor NPC1 can influence the cellular susceptibility of bats to infection. This observation is especially striking in light of previous findings that filoviruses could efficiently infect a broad range of mammalian cells, including some derived from bats. Indeed, this prior work and the results of experimental infection studies in rodents and bats have led to the hypothesis that interactions between viral components and those of the host innate and adaptive immune systems constitute the primary molecular variables influencing filovirus host range in nature.

17Here, we propose that NPC1 is also a genetic determinant of filovirus susceptibility in bats. The essential nature of NPC1 for infection in cells derived from mammals of multiple species, including bats (Figure 3), and for infection and in vivo pathogenesis in lethal EBOV infection mouse models argues against the existence of alternative filovirus entry receptors. Therefore, strong reductions in the affinity of virus-NPC1 recognition are predicted to reduce or eliminate infection in whole bat hosts, as observed in NPC1-deficient mice, barring viral mutation to enhance this affinity. It is conceivable that even modest defects or delays in viral multiplication through such a mechanism could help determine host range by accelerating viral immune clearance, as recently observed in NPC1-heterozygous mice, or by synergizing with other host-virus barriers. The highly virus- and host species-specific nature of the virus-receptor mismatch uncovered in this study warrants the determination of more bat NPC1 sequences for inclusion in genetic analyses (see below), and a more comprehensive phenotypic examination of virus-bat pairs. Such studies maydiscover additional interesting bat-filovirus dynamics, including incompatibilities between filoviruses and NPC1 or other proviral/antiviral host factors. Such discoveries have potential implications for our understanding of the molecular basis of filovirus infection, virulence, and host range.

18We found that a single amino acid change, at residue 502, in the African straw-colored fruit bat ortholog of NPC1 (EhNPC1) greatly diminished the susceptibility of cells from multiple tissues and individuals to EBOV. These migratory pteropodids are widely distributed across sub-Saharan Africa (Figure 1A), roost in large colonies near human settlements, and host other RNA viruses with zoonotic potential. Moreover, they are extensively hunted for bushmeat in Western Africa, making them ideal candidates to transmit viruses directly to humans. Unfortunately, there is little information currently available on the susceptibility of African straw-colored fruit bats to EBOV or their potential role as filovirus hosts. Serologic surveys have found some evidence for exposure to one or more filovirus; however, neither infectious virus nor coding-complete or full viral genomes—the gold standards—have been successfully obtained from these bats, indicating they may only have been exposed to filoviruses, rather than being productively infected (reviewed in). While more extensive wildlife sampling and, if feasible, experimental infections of African straw-colored fruit bats will be required to clarify this picture, we can extrapolate to several possible scenarios. First, these bats are fully resistant to EBOV, and therefore cannot be the source of this virus in the 2013–present EBOV disease outbreak in Western Africa or the 2014 outbreak in Middle Africa. Second, because African straw-colored fruit bat cells do remain weakly susceptible to EBOV (Figure 3C), it is conceivable that they support EBOV replication at low levels. Indeed, this is one hallmark of a sustaining viral reservoir. Third, the filoviruses circulating in these bats, whether EBOV or otherwise, bear one or more GP mutations (e.g., V141A) that circumvent the infection barrier imposed by EhNPC1. Assessing this last hypothesis and understanding the nature of the selection pressures that drive GP evolution in vivo will require the isolation of ebolavirus GP sequences from bats—there are none currently available.

19Although these results suggest that African straw-colored fruit bats are selectively refractory to EBOV, our genetic findings indicate that this is not merely a special relationship between one host and one virus. Rather, we used a diverse set of bat NPC1 sequences, only one of which is from African straw-colored fruit bats, to show that a number of codons, including residue 502, have evolved under recurrent positive selection. This is a process in which resistant NPC1 variants are ‘serially replaced’ in response to compensating viral mutations that restore susceptibility. We provide evidence that the filovirus GP interaction surface in the second luminal domain of NPC1, domain C, is a hotspot for such positive selection (Figure 5). By contrast, the vast majority of codons in mammalian NPC1 have evolved under purifying selection. We propose that this pattern of selection is the signature of a long-term genetic conflict between filoviruses and NPC1 in bats, superimposed over the normal evolutionary signature of a housekeeping gene with a critical role in cellular cholesterol trafficking. Similar signatures of recurrent positive selection have been identified in other housekeeping genes that encode viral receptors, including the transferrin receptor (TfR; receptor for New World arenaviruses, the betaretrovirus murine mammary tumor virus, and parvoviruses), bat angiotensin-converting enzyme-2 (ACE2; receptor for SARS-like coronaviruses), and mammalian dipeptidyl peptidase-4 (DPP4; receptor for MERS-like coronaviruses). In these cases as well, the preponderance of positively-selected residues localize to virus-receptor interfaces. Interestingly, the sequence polymorphism at NPC1 residue 502 did not impair cholesterol clearance from lysosomes (Figure 3), and none of the residues under positive selection were found to be mutated in Niemann-Pick type C disease patients. Thus, despite being constrained by its housekeeping function, NPC1 appears to retains a sizeable sequence space accessible to adaptive mutation.

20It is tempting to speculate that sequence variation at residue 141 (Figure 6) and potentially other positions in the receptor-binding site of filovirus glycoproteins represents the other half of the genetic arms race, shaped by selective pressure to utilize restrictive NPC1 receptors. Although more data, especially filovirus sequences from bats, are needed, our findings raise the tantalizing possibility that filoviruses, including those yet undiscovered, are each adapted to specific bat hosts, with co-evolved virus-receptor interactions constituting one potential biological barrier to interspecies viral transmission. Alternatively, it is conceivable that repeated contacts between unknown (non-bat) reservoir hosts carrying specific filoviruses, and bats of particular species, have driven positive selection in bat NPC1 to limit infection (and selection of filoviruses with compensating sequence changes in GP). In this scenario, detection of anti-filovirus antibodies or filovirus genome-derived oligonucleotides may reflect a type of spillover event from the actual filovirus reservoir hosts into bats.

21Our hypothesis that NPC1 in bats has been genetically sculpted by filoviruses (and vice versa) presupposes not only a long-term coevolutionary relationship, but also one in which these viruses have imposed selective pressure on bats to limit or eliminate infection. The discovery of filovirus NP- and VP35-related endogenous viral elements (EVEs) in bat genomes is consistent with such a long-term relationship. To further investigate the deeper origins of filoviruses in bats, we screened all available bat genomes for filovirus-related EVEs. We obtained evidence for synteny between a filovirus nucleoprotein (NP)-like EVE in the genome of the big brown bat (Eptesicus fuscus) and those previously identified in three, more distantly-related, myotis bats (Figure 7 and Supplementary file 5). This new discovery strongly suggests that all four EVEs resulted from a single insertion event prior to the divergence of the Myotis and Eptesicus genera, ≈25 million years ago. Therefore, bats may have been exposed to filovirus-like agents for far longer than previously recognized (≈13 million years ago).

22Available experimental exposure studies, although limited in number and scope, suggest that some filoviruses isolated from humans can replicate in bats without causing substantial host pathology (e.g., MARV and RAVV in Egyptian rousettes). These observations therefore prompt a key question: what is the origin and nature of the selective pressure that has driven accelerated NPC1 evolution in bats? Our scant understanding admits a number of possibilities. First, it is conceivable that some filoviruses do indeed replicate in a manner that is deleterious to their specific bat hosts—we may simply not have identified the viruses and hosts in question. Indeed, the filovirus LLOV, discovered in Schreibers's long-fingered bat carcasses in Spain and Portugal, may exemplify this possibility. Alternatively, in some cases (e.g., ebolaviruses and Egyptian rousettes), the human viral isolates used in challenge studies may differ from these bat isolates in important respects due to human adaptation (human EBOV, BDBV, TAFV, RESTV, and SUDV isolates do not infect Egyptian rousettes). Second, filoviruses may have been more virulent in bats in the past. Thus, the positive selection signatures observed in bat NPC1, which cannot be accurately dated, may represent fixed alleles that are the consequence of a selective process driven by ancient filoviruses with properties distinct from their modern counterparts. Indeed, the lack of virulence observed in some bats may reflect a détente that was shaped by precisely these historic genetic conflicts between filoviruses and bats. Third, we cannot rule out the (unlikely) possibility that the evolution of NPC1 in bats was driven by an entirely different infectious agent that also utilizes (or utilized) NPC1 to multiply in its hosts. Regardless of the mechanisms that genetically shaped NPC1, we propose that polymorphisms in this gene nevertheless impose host barriers that impede the colonization and spread of present-day filoviruses in bats in Africa and elsewhere. Our findings set the stage for broader explorations of species-specificity in filovirus interactions with proviral and antiviral host factors, with an eye to uncovering new molecular arms races between filoviruses and bats and new genetic determinants of filovirus host range and host switching.

23The following immortalized pteropodid fibroblast cell lines were used: RoNi/7.1 (kidney; Rousettus aegyptiacus), HypNi/1.1 (kidney; Hypsignathus monstrosus), EpoNi/22.1 (kidney; Epomops buettikoferi), EidNi/41.2, EidNi/41.3 (kidney; Eidolon helvum), and EidLu/20 (lung; Eidolon helvum). The species origin of each cell line was confirmed in the publication in which it was first described.

24Bat cell populations stably expressing human NPC1 (HsNPC1) were generated as described previously. Briefly, subconfluent monolayers of cells were transduced with a retroviral vector expressing HsNPC1 modified at the C–terminus with a triple flag epitope tag. Transduced cells were selected by puromycin treatment (10 µg/mL).

25Licenses for capturing and export of bats, as well as ethical review and clearances of animal handling procedures were obtained from the Ghana Forestry Commission of the Ministry of Food and Agriculture. Bat organ samples were obtained as described. Bats were caught, anesthetised with ketamine/xylazine and exsanguinated by heart puncture. Carcasses were transported on ice to a nearby laboratory facility, and organs were dissected and immediately snap-frozen for long-term storage. Animals were typed morphologically and genetically as described previously.

26Vero African grivet kidney cells and 293T human embryonic kidney fibroblast cells were obtained from ATCC. Cell lines were maintained in Dulbecco’s modified Eagle medium (DMEM) (Life Technologies, Grand Island, NY) and supplemented with 10% fetal bovine serum (Atlanta Biologicals, Flowery Branch, GA), and 1% penicillin-streptomycin (Life Technologies). All cell lines were maintained in a humidified 37°C, 5% CO2 incubator.

27We knocked out the NPC1 gene in the EidNi/41.3 cell line by CRISPR-Cas9-mediated genome editing as described previously. A CRISPR guide RNA (gRNA) sequence to target 5’-GTTGTGATGTTCAGCAGCTTCGG-3’ in the E. helvum NPC1 mRNA was cloned into the gRNA cloning vector (Addgene Plasmid #41824). EidNi/41.3 cells were co-transfected with plasmid encoding human codon-optimized endonuclease Cas9 (hCas9, Addgene Plasmid #41815), gRNA cloning vector encoding the E. helvum NPC1-specific gRNA, a monomeric red fluorescent protein (mRFP1) expression plasmid (to monitor transfection efficiency), and pMX-IRES-Blasti (confers blasticidin resistance to transfected cells) using Lipofectamine 2000 (Life Technologies). At 24 hr post-transfection, transfected cells were selected with 50 µg/ml of blasticidin for 24 hr and then allowed to recover in the absence of the selection agent.

28Total RNA was isolated from surviving cells with the RNAeasy Mini kit (Qiagen, Valencia, CA) as per the manufacturer’s directions. The E helvum NPC1 mRNA sequence flanking the gRNA target site was amplified with the One-Step RT-PCR kit (Qiagen) and the following primers: Forward: 5’-ATTCTGGACTACCAAAATCTTTGCC-3’, and Reverse: 5’- ACATGGCATCCAAGCCCAAG-3’. Thermocycling conditions used for the RT-PCR were: 50°C for 30 min (reverse transcription), followed by 95°C for 15 min (initial PCR activation), then 30 cycles of 94°C for 30 sec, 60°C for 30 sec, 72°C for 1 min, then a final extension of 72°C for 10 min. Amplified PCR products were tested for indels at the target site with the SURVEYOR mutation detection kit for standard gel electrophoresis (Transgenomic, Omaha, NE), as per the manufacturer’s instructions. Once indels were confirmed, amplified PCR products from single cell clones were cloned into a TOPO-TA vector (Life Technologies). Multiple clones for each single cell population were sequenced to confirm disruption of NPC1 alleles.

29Recombinant vesicular stomatitis Indiana viruses (VSVs) expressing eGFP, and EBOV, MARV, or LLOV GP in place of VSV G have been described previously. VSV pseudotypes bearing glycoproteins derived from VSV, EBOV, BDBV, TAFV, SUDV, and MARV were generated essentially as described previously. VSV particles containing GPCL were generated by incubating rVSV-GP-EBOV with thermolysin (200 μg/mL) (Sigma-Aldrich, St. Louis, MO) for 1 hr at 37°C. The protease was inactivated by addition of phosphoramidon (1 mM) (Sigma-Aldrich), and reaction mixtures were used immediately. Infectivities of VSV pseudotypes were measured by manual counting of eGFP-positive cells using fluorescence microscopy at 16–24 hr post-infection, as described. Infectivities of rVSVs were measured in a similar manner, except that NH4Cl (20 mM) was added to infected cell cultures at 1–2 hr post-infection to block viral spread, and individual eGFP-positive cells were manually counted at 12–14 hr post-infection.

30The wild-type filoviruses Ebola virus/H.sapiens-tc/COD/1995/Kikwit-9510621 (EBOV/Kik-9510621; “EBOV-Zaire 1995”) and Marburg virus/H.sapiens-tc/DEU/1967/Hesse-Ci67 (MARV/Ci67) used in this study were described previously. Cells were exposed to virus at an MOI of 1 pfu/cell (Figure 1C) or 3 pfu/cell (Figure 2A) for 1 hr. Viral inoculum was then removed, and fresh culture media was added. At 48 hr (Figure 2A) or 72 hr (Figure 1C) post-infection, cells were fixed with formalin and blocked with 1% bovine serum albumin (BSA). EBOV-infected cells and uninfected controls were incubated with EBOV GP-specific monoclonal antibody KZ52. MARV-infected cells and uninfected controls were incubated with MARV GP-specific monoclonal antibody 9G4. Cells were washed with PBS prior to incubation with either goat anti-mouse IgG or goat anti-human IgG conjugated to Alexa 488. Cells were counterstained with Hoechst stain (Invitrogen, Carlsbad, CA), washed with phosphate-buffered saline (PBS), and stored at 4°C. Infected cells were quantitated by fluorescence microscopy and automated image analysis. Images were acquired at 20 fields/well with a 20× objective lens on an Operetta high content device (PerkinElmer, Waltham, NY). Operetta images were analyzed with a customized scheme built from image analysis functions available in Harmony software.

31From bats of four species (Hypsignathus monstrosus, Eidolon helvum, Epomops buettikoferi, and, Rousettus aegyptiacus), mRNA was collected from cell lines (or spleen samples for additional Eidolon helvum NPC1 domain C sequences; Figure 4—figure supplement 2), cDNA libraries were constructed, and the NPC1 transcript was sequenced (see Supplementary file 3 for primers). Using available RNAseq read data (Supplementary file 2), we assembled bat transcriptomes and identified NPC1 sequences in bats of six additional species (Myotis brandtii, Artibeus jamaicensis, Cynopterus sphinx, Myotis lucifugus, Pteropus alecto, and Desmodus rotundus). Transcriptome data were cleaned with Trimmomatic and assembled using Trinity and Trans-ABySS. The 10-species NPC1 alignment (Supplementary file 4) was analyzed for positive selection using the M8 codon model in the codeml package in PAML, REL, and FEL, and MEME available at http://datamonkey.org/. All evolutionary analyses were done using both the NPC1 gene tree and a species tree (Figure 5—figure supplement 1). The species tree represents the accepted relationships amongst the bats analyzed.

32To identify orthologous filovirus-related EVE insertions, we screened bat genomes in silico for EVEs. A representative set of filovirus protein sequences was obtained from GenBank, supplemented by the putative protein sequences of previously identified filovirus EVEs. These sequences were used as ‘probes’ in tBLASTn screens of whole genome shotgun (WGS) sequence data derived from bats of ten species (Eidolon helvum, Eptesicus fuscus, Myotis brandtii, Myotis davidii, Myotis lucifugus, Pteropus alecto, Pteropus vampyrus, Megaderma lyra, Pteronotus parnellii, and Rhinolophus ferrumequinum). Statistically significant matches to filovirus probes were extracted, conceptually translated, and aligned with homologous filovirus proteins. Orthologous flanking sequences were identified by BLAST comparison of EVE-containing contigs. An alignment of the identified EVEs, along with the flanking information in the relevant bat genomes, is shown in Supplementary file 5.

33A construct engineered to encode HsNPC1 domain C (residues 372–622) flanked by sequences that form a stable, antiparallel coiled coil, and fused to a preprotrypsin signal sequence with flag and hexahistidine tags at its N–terminus has been described. Similar constructs bearing bat NPC1 domain Cs were generated by replacing the human domain C sequence with a sequence encoding domain C from each bat NPC1 ortholog. Soluble domain C was expressed in human 293-Freestyle cells (Invitrogen) and purified from supernatants by nickel affinity chromatography, as described previously. Alternatively, cell supernatants containing soluble domain C were used directly for GP-NPC1 binding ELISAs following calibration for domain C concentration (see below).

34NPC1 domain C concentrations used in the ELISAs were normalized as follows. Proteins were resolved by SDS-PAGE followed by immunoblotting with an anti-flag antibody followed by an anti-mouse Alexa-680 secondary antibody (Invitrogen). Blots were visualized using the LI-COR Odyssey Imager, and the domain C band was quantified using the LI-COR Image Studio package (LI-COR Biosciences, Lincoln, NE).

35Thermolysin-cleaved VSV-EBOV GP particles were captured onto high-binding 96-well ELISA plates (Corning, Corning, NY) using KZ52, a conformation-specific anti-EBOV GP monoclonal antibody. Plates were blocked with PBS containing 3% BSA, and serial dilutions of NPC1 domain C protein were then added. Bound domain C was detected with an anti-flag antibody conjugated to horseradish peroxidase (Sigma-Aldrich) and Ultra-TMB substrate (ThermoFisher, Grand Island, NY). All binding steps were carried out at 37°C for 1 hr or at 4°C overnight. ELISAs with VSVs bearing LLOV and MARV GP were performed as above, but with the following modifications. VSV-LLOV GP particles were cleaved by incubation with a reduced concentration of thermolysin (12.5 µg/mL, 37°C, 1 hr) due to its enhanced protease sensitivity relative to ebolavirus GPs, as described. The viral envelope was then labeled with biotin using a function-spacer-lipid construct (FSL-biotin) (Sigma-Aldrich), as described previously. Biotinylated viral particles were captured onto streptavidin-coated ELISA plates (0.65 µg/mL). The remainder of the steps in the ELISA were performed as described above for VSV-EBOV GP. VSV-MARV GP particles were cleaved by incubation with trypsin (150 µg/mL, 37°C, 5 min; Sigma-Aldrich), modified as above using FSL-biotin, and captured onto streptavidin-coated magnetic beads (Spherotech, Lake Forest, IL). Beads were then aliquotted into a 96-well round-bottomed plate for the remaining ELISA steps. PBS containing 5% nonfat dry milk was used for blocking and washing steps. Binding curves were generated by nonlinear regression analysis using Prism (4-parameter logistic equation; GraphPad Software, La Jolla, CA).

36To detect NPC1 in primate or bat kidney fibroblasts, whole cell lysates were prepared as previously described. Briefly, cells were washed with PBS and lysed in NTE-CHAPS buffer (10mM Tris [pH 7.5], 140mM NaCl, 1mM EDTA, 0.5% vol/vol 3-[(3-cholamidopropyl)dimethylammonio]-1- propanesulfonate) (Sigma-Aldrich) containing a protease inhibitor cocktail (Roche, Basel, Switzerland), and placed on ice for 30 min. To assist in cell lysis, cell suspensions were vortexed every 10 min, and then placed on ice for 30 min. Samples were spun at 14,000 ×g for 10 min, and supernatants harvested for western blot. In some experiments, proteins were deglycosylated with protein N–glycosidase F (New England Biolabs, Ipswich, MA) according to the manufacturer’s instructions.

37Proteins were resolved in 8% sodium dodecyl sulfate (SDS)-polyacrylamide gels and transferred to nitrocellulose membranes. Endogenous NPC1 was detected using an anti-Niemann Pick C1 polyclonal antibody (1:1,000 dilution; ab36983, Abcam, Cambridge, MA), followed by incubation with a donkey anti-rabbit antibody conjugated to horseradish peroxidase (1:5,000 dilution, Santa Cruz Biotechnology, Dallas, TX). Endogenous cyclin-dependent kinase 4 (CDK4; loading control) was detected with a rabbit polyclonal antibody (1:1,000 dilution; sc-260, Santa Cruz Biotechnology). Ectopic expression of HsNPC1-flag was detected with an anti-flag antibody conjugated to horseradish peroxidase (Sigma-Aldrich). Bands were visualized by incubation with an enhanced chemiluminescence reagent (ThermoFisher) followed by exposure to X-ray film.

38In Figure 3, cells were visualized using an inverted fluorescence microscope under illumination with a 63X high-numerical aperture oil objective (Figure 3B) or a 10X air objective (Figure 3C). Images were captured with an Axiocam MRm CCD camera using AxioVision software (Zeiss USA, Thornwood, NY), and imported into Photoshop (Adobe Systems, San Jose, CA) for processing. Images were cropped, inverted (Figure 3B), and subjected to linear adjustment for overall brightness and contrast using the Levels tool. Developed X-ray films were digitized with a flatbed scanner and processed in Photoshop as described above.

39Statistical comparison of means among multiple independent groups was carried out by one-way analysis of variance (ANOVA) with Tukey’s post hoc test for multiple comparisons. In some figures (see Figure Legends), an unpaired two-tailed Student’s t-test with Welch’s correction for unequal variances was used for pairwise comparison of independent groups. All statistical analyses were performed in GraphPad Prism.