The cellular lectins DC-SIGN, DC-SIGNR, LSECtin, ASPGR-1 and hMGL [41,85,86,87,88,89] can promote filovirus entry into transfected cells (Table 1). An initial study suggested that expression of some of these lectins might be sufficient to render target cells susceptible to filovirus infection [85]. However, subsequent work showed that lectins merely facilitate usage of low levels of so far unidentified molecules important for filovirus entry [81,90,91]. Augmentation of filovirus infection by lectins depends on binding of these proteins to glycans on GP_1,2, and the respective lectins exhibit differences in their carbohydrate specificities, which are reflected by differential requirements for GP_1,2 binding. Thus, the extensively O-glycosylated MLR is dispensable for GP_1,2 binding to DC‑SIGN and DC‑SIGNR [92], which mainly recognize high-mannose glycans, while deletion of the MLR impedes interactions with hMGL, which recognizes terminal galactose-/N-acetylgalactosamine [41,50,93]. In addition, differences in lectin engagement among the ebolavirus species and MARV have been noted [41,92]. The expression of the GP-binding lectins on relevant filovirus target cells, including tissue macrophages and dendritic cells (hMGL, DC-SIGN) [94,95], hepatocytes (ASGPR-1) [96] and liver and lymph node sinusoidal endothelial cells (LSECtin, DC-SIGNR) [97,98] suggests that they could modulate filovirus spread in the infected host. However, there is currently little evidence that endogenous lectin expression appreciably augments filovirus entry into primary target cells [90], and the role of lectins in filovirus spread in animal models has not been examined. In the context of HIV-1 infection, it has been demonstrated that viral binding to DC-SIGN on dendritic cells triggers signal transduction and thereby commandeers the cell to undergo aberrant maturation and to produce immunosuppressive cytokines [99], which is believed to promote viral spread. Whether a similar mechanism is operative during filovirus infection is unknown. Finally, it is noteworthy that secreted lectins can modulate filovirus infection. Thus, recombinant mannose-binding C-type lectin (MBL) was shown to protect mice from a lethal EBOV infection potentially by targeting the virus for phagocytosis and complement-directed lysis [100].

The human T cell Ig mucin 1 (TIM-1) surface molecule was initially found to be a cellular receptor for hepatitis A virus [104]. By a bioinformatics-based correlation analysis between gene expression profiles and susceptibility of cell lines to EBOV-GP_1,2-driven infection, TIM-1 was recently also identified as an entry factor for filoviruses [105]. TIM-1 is a type I membrane glycoprotein with an extracellular IgV domain and a mucin-like domain predicted to be heavily O-glycosylated [106]. The IgV domain allows highly specific recognition of phosphatidylserine exposed on the surface of apoptotic cells, and TIM-1 was shown to be involved in the clearance of apoptotic cells [107,108]. TIM-1 is expressed on activated T-cells, epithelial cells, conjunctiva and renal tissue [105,109,110] as well as certain cell lines including the liver cell line Huh7 [105] (Table 2), which is permissive for filovirus replication [111].

A physical interaction between the receptor binding domain of EBOV-GP₁ and soluble TIM-1 could be demonstrated [105], and soluble TIM-1 was shown to block EBOV-GP_1,2-driven infection, indicating that TIM-1-dependent filovirus entry requires GP_1,2 interactions with this protein. RNA knock-down of endogenous TIM-1 inhibited pseudovirus entry into otherwise susceptible cell lines, and ectopic expression of TIM-1 on TIM-1-negative cell lines enhanced infection [105], indicating that TIM-1 indeed promotes filovirus infection of certain cell lines. However, macrophages and dendritic cells, which are important targets of filovirus infection, lack TIM-1 expression [110], suggesting that so far unidentified cellular factors facilitate filovirus entry into these cells. Finally, it would be interesting to determine whether TIM-1 intracellular signaling [112,113] might contribute to filovirus GP-mediated cellular uptake.

Another interesting group of molecules shown to be involved in ebolavirus infectious entry are members of the Tyro3/Axl/Mer (TAM) family of receptor tyrosine kinases [119,131]. These proteins contain a kinase domain and an adhesion-molecule-like extracellular domain and play important roles in diverse biological processes including cell proliferation and survival, cell adhesion and cytokine release [131]. Members of the TAM family are evolutionarily conserved between vertebrate species and the respective proteins are widely expressed in adult tissues, most prominently in the brain, lung, kidney, lymphatic tissue and the vascular system [118,119] (Table 2). The natural ligands for TAM receptors are the structurally homologous proteins Gas6 and protein S [132,133]. These secreted proteins bind to phosphatidylserine residues exposed by apoptotic cells [134,135], and ligand-activated TAM receptors have been found to inhibit inflammation pathways in macrophages and dendritic cells [120].

By introduction of a cDNA library derived from permissive VeroE6 cells into non-susceptible Jurkat lymphocytes, Axl was identified as an EBOV entry factor expressed at the cell surface [117]. The ectopic expression of Axl and the TAM family members DTK and MER on lymphoid cells allowed for transduction of GP_1,2-harboring pseudotypes, which could be blocked by Axl-specific antibodies, confirming that TAM-proteins could promote filovirus entry [117], although the mechanism underlying the antibody-mediated blockade of viral entry is currently unclear. Mutational analysis revealed that both the extracellular ligand binding domain and the cytoplasmic tail of Axl were required for efficient GP_1,2-mediated entry [116,117]. A role of TAM family proteins in filovirus entry was confirmed by an independent study, which showed that expression of an mRNA encoding for Axl correlates with susceptibility of cell lines to EBOV infection [114]. The use of TAM protein‑specific antibodies and siRNA knockdown indeed identified several cell lines, in which EBOV-GP_1,2-faciliated entry was dependent on Axl expression. In contrast, down-regulation of Axl‑expression in other cell lines did not compromise GP_1,2-driven entry [114,115,117]. So far, all studies failed to detect a direct interaction between EBOV-GP₁ and Axl. However, Axl expression was shown to augment internalization of EBOV-GP_1,2-bearing pseudovirions and virus-cell fusion [114], and this activity correlated with enhanced macropinocytosis in Axl-expressing cells [115]. Thus, Axl might promote filovirus uptake by macropinocytosis, which was previously shown to be a pathway exploited by filoviruses for cellular uptake [136,137]. Indeed, inhibition of Axl and blockade of PIK3, which is important for macropinocytosis and Axl-dependent signaling, both inhibit filovirus entry in a cell type-dependent fashion [21,125]. Finally, it is noteworthy that Gas6 can promote Sindbis virus entry by bridging phosphatidylserine present in the viral envelope to Axl localized on target cells [138]. Whether a similar mechanism operates in the context of filovirus entry remains to be investigated. The observation that recombinant Gas6 inhibits Axl- and DTK-dependent GP_1,2-mediated entry [117] might argue against this hypothesis.

Integrins are cell surface expressed heterodimeric type I transmembrane glycoproteins, which are composed of two non-covalently linked subunits (α and β) [122]. The integrin family in mammals comprises eighteen α and eight β subunits which can assemble into 24 different heterodimers [139]. These heterodimers, in turn, convey specificity to cell-cell and cell-extracellular matrix adhesion, immune cell recruitment, extravasation, and signaling events [121,140]. Members of the integrin family have been discovered as attachment factors or receptors for a large number of enveloped and non-enveloped viruses, including herpesviruses [141,142,143], adenoviruses [144,145], hantaviruses [146,147,148,149], picornaviruses [150,151,152], and reoviruses [153,154]. The association with viral surface proteins and microbial pathogens are followed by multiple signaling events, some of which promote cytoskeletal reorganization and thus facilitate receptor-mediated endocytosis (as reviewed in [155,156] and references therein). It is not surprising that numerous viruses hijack integrins for infectious entry, as they are widely expressed in various tissues throughout the body (Table 2).

Expression of EBOV-GP_1,2 was initially shown to interfere with surface expression of various cellular membrane proteins, including α3 and ß1 integrins [40,124]. However, a more recent study suggested that GP_1,2 expression does not reduce cell surface levels of integrins but rather sterically occludes epitopes in these proteins otherwise recognized by antibodies [157]. Experiments with EBOV-GP_1,2-bearing pseudotypes demonstrated that soluble recombinant ß1 integrin or ß1-reactive antibodies diminish GP_1,2 driven entry, suggesting that GP_1,2 might need to engage ß1 integrins for infectious entry [124]. However, a direct interaction between EBOV-GP_1,2 and integrins remains to be demonstrated. Work by Schornberg and colleagues provided evidence that α5β1-integrin is required for expression of the double chain forms of cathepsin B and L and for full cathepsin L activity [123], an endosomal protease involved in priming of GP_1,2 for membrane fusion, as discussed below. In contrast, α5β1-integrin was dispensable for GP_1,2-mediated binding and uptake into target cells [123]. These observations suggest that α5β1-integrin indirectly promotes GP_1,2-driven entry by ensuring activity of GP_1,2-priming cysteine proteases or by stimulating the protease maturation pathway, which might be required for viral entry. Whether physical interactions of GP_1,2 with α5β1-integrin also contribute to filovirus entry remains to be determined.

Cathepsins comprise serine, aspartic and cysteine proteases and carry out diverse biological functions, including antigen processing for MHCII presentation [168,169,170]. Cysteine proteases of the papain family, some of which are localized in endosomes, are expressed as preproenzymes and are activated by proteolysis in the endoplasmatic reticulum and the late endosome/lysosome [168,169,170]. The low pH environment present in the latter compartment is essential for cathepsin enzymatic activity.

Chandran and colleagues demonstrated that two lysosomal cathepsins, cathepsin B and L, cleave filovirus GP_1,2 and that cathepsin activity is essential for GP_1,2-driven host cell entry [57]. This report showed that cathepsins B and L prime the filovirus GP_1,2 for membrane fusion and proposed that cathepsin cleavage of virion-associated GP_1,2 occurs in a sequential fashion: First, cathepsin L and/or B cleave GP_1,2 into an 18 kDa form, which is fully infectious but still requires cathepsin B activity for infectious entry. Subsequently, the 18 kDa form is processed by cathepsin B and cleavage might be sufficient to trigger membrane fusion [57]. An alternative model for GP_1,2 activation has been proposed by subsequent studies. Thus, Schornberg and colleagues demonstrated that processing of virion-inserted GP_1,2 by recombinant cathepsin B and L or the bacterial protease thermolysin yielded a 19 kDa form of GP₁ and was associated with a notable increase in infectivity [56], a finding confirmed by others [126]. Processing of GP₁ into the 19 kDa from proceeded via 50 kDa and 20 kDa intermediates and its was speculated that the 20 kDa form might differ from the 18 kDa form observed by Chandran and colleagues only in the presence of a N-linked glycan [56]. Virions bearing the 19 kDa form were largely resistant to cathepsin B but not L inhibitors but remained sensitive to a lysosomotropic agent and a cysteine protease inhibitor. On the basis of these findings, a two-step model was proposed, suggesting that GP_1,2 must first be processed by cathepsins B and L before the activity of a third lysosomal factor, potentially a thiol reductase, triggers GP_1,2-dependent membrane fusion [56,171]. Indeed, subsequent studies provided evidence that the 19 kDa form represents a metastable conformation in which the fusion machinery is not yet exposed [37,172,173,174] and which can be triggered for membrane fusion by low pH and reduction [173]. In addition, it was demonstrated that cleavage of GP_1,2 removes a glycan cap and the MLR, while the N-terminal RBR and GP₂ remain in the molecule [172,175,176]. In sum, proteolytic processing by cathepsins B and L primes GP_1,2 for membrane fusion and exposes the RBR. Subsequently, an incompletely understood stimulus triggers membrane fusion and these final steps of the lysosomal escape of filoviruses critically depend on GP_1,2 binding to NPC1 (Figure 3), as discussed below.

Despite the importance of cathepsin B and L in priming EBOV-GP_1,2 for membrane fusion in several cell lines, the dependence on these particular proteases for viral entry is not universal among filoviruses. A requirement for cathepsin B activity during entry of EBOV-, TAFV- and BDBV-but not SUDV-, RESTV and MARV-GP-bearing pseudotypes has been described, and the same group showed that particles harboring the GP_1,2 of EBOV, SUDV and MARV exhibited enhanced transduction efficiency when cathepsin L was active in concert with cathepsin B. In contrast, entry of RESTV was dependent on a cysteine protease distinct from cathepsins B and L [125]. Furthermore, it has been reported that cathepsin L activity is dispensable for ebolavirus GP_1,2-driven entry into Vero cells and mouse embryonic fibroblasts [57,125] as well as human monocyte-derived dendritic cells [128]. Moreover, the observation of the failure of EBOV-GP_1,2-bearing pseudotypes to transduce CatB⁻/₋ CatL⁻/₋ mouse embryonic fibroblasts can be overcome by ectopic expression of CatB, suggests that a protease other than CatL is required for a post-CatB cleavage step necessary for membrane fusion [125]. In addition, transduction of primary human macrophages by EBOV-GP_1,2-carrying pseudotypes was shown to be dependent on both cathepsin B and L, whereas MARV-GP_1,2-facilitated entry was not blocked efficiently by cathepsin B/L inhibitors, suggesting that MARV-GP_1,2 might employ a so far unknown protease for priming in macrophages [177]. It is also noteworthy that many studies investigating the role of cathepsins in filovirus host cell entry were performed with GP_1,2‑bearing vectors and not with authentic filoviruses. It would thus be interesting to examine the effect of cathepsin B and L knock-out on filoviral spread and pathogenicity, particularly in the light of efforts to develop cathepsin inhibitors as treatment for SARS-coronavirus [178] and filovirus infection. The respective knock-out mice required for such studies have been described [179,180,181]. A study by Wong and colleagues indicates that blockade of cathepsin activity might result in the development of resistant viruses. Resistance was associated with mutations at the border between GP₁ and GP₂, which increased susceptibility to proteolytic cleavage and allowed GP_1,2 priming by cysteine cathepsins other than cathepsin B and L [182]. Finally, it is noteworthy that expression of integrins previously suggested to promote filovirus entry is required for activity of cathepsin B and L [123], indicating an intricate interplay between some of the host cell molecules exploited by filoviruses for cellular entry.

An endosomal factor required for filovirus entry after GP_1,2 priming by cathepsins has recently been discovered by two independent studies as the Niemann-Pick C1 (NPC1) protein. Cote and colleagues discovered that filoviruses depend on NPC1 for cellular entry by screening a library of chemical compounds for entry inhibitors [20], while Carette and coworkers found in a screen of haploid human cells that mutations in NPC1 are not compatible with filovirus GP_1,2-driven entry [129]. Finally, a recent study found that CHO cells selected for resistance to EBOV-GP_1,2-dependent entry harbored a defect in the NPC1 gene [183]. The NPC1 protein is highly conserved among species and is ubiquitously expressed in human tissues, with the highest expression in the liver [184,185] (Table 2). The protein is an integral membrane protein of late endosomes and lysosomes and exhibits a polytopic orientation, forming several luminal and cytoplasmatic loops. NPC1 is a cholesterol transporter and mutations in the NPC1 gene result in fatal, progressive neurodegenerative disorder, Niemann-Pick C1 disease, due to a defect in the export of cholesterol from lysosomes [186]. The abnormal accumulation of cholesterol in turn leads to altered protein and lipid trafficking [187,188].Which lines of evidence suggest that NPC1 facilitates filovirus entry? NPC1 interacts with primed GP₁ [20,130] and the contribution of NPC1 to infectious entry of filoviruses can be separated from its cholesterol transport activity [20,129], indicating that the protein directly facilitates entry. NPC1-deficient cells, including primary fibroblasts derived from NPC1 patients, were resistant to filovirus infection, but still allowed for efficient cellular entry of several other viruses [20,129]. Furthermore, transduction of the wild type NPC1 gene into NPC1-defective, patient-derived cells or NPC1-negative CHO cells fully restored infection [20,129] and directed expression of NPC1 in non-susceptible reptilian cells or haploid hamster CHO-K1 cell clones was sufficient to render these cells susceptible to GP_1,2-mediated infection [130]. Additionally, siRNA knockdown of NPC-1 in HeLa cells resulted in reduced virus uptake [20]. Finally, heterozygous NPC1^−/+ mice were protected against filovirus infection in sharp contrast to wild type mice [129]. In sum, these results show that NPC1 plays a key role in cellular entry of ebolaviruses.

As the NPC1 protein is localized on the endosomal and lysosomal membranes, it was proposed to act downstream of filovirus GP_1,2 engagement of attachment and signaling factors at the cell surface. Indeed, GP_1,2-mediated viral uptake was readily detectable in NPC1-deficient cells, where the virions accumulated in early endosomes, indicating that membrane fusion was not triggered [129]. Proteolytic processing of GP_1,2 and thus exposure of the RBR was a prerequisite of NPC1 binding, as only the cleaved 19 kDa form was able to physically interact with NPC1 [20,130]. Mapping studies revealed that the 19 kDa form binds to the second luminal domain of NPC1 [130] and cell surface presentation of this domain in the context of an artificial receptor molecule was sufficient to allow entry of pseudotypes carrying thermolysin-primed filovirus GP_1,2 [130]. It can be speculated that the interaction between the 19 kDa form of GP_1,2 and NPC1 might expose the GP₂ residues involved in membrane fusion. However, fusion of pseudotypes bearing the primed 19 kDa GP_1,2 with the plasma membrane of target cells expressing the second loop of NPC1 at their surface could not be induced by low pH treatment. Thus, binding of primed GP_1,2 to NPC1 is not sufficient to trigger membrane fusion [130].

Besides filoviruses, flaviviruses have also been found to depend on molecules involved in cholesterol transport for infectious entry into host cells. The cholesterol-transport inhibitor U18666A interfered with hepatitis C and Dengue virus (strain TSV01) infection [189,190], and a factor with 40% amino acid identity to NPC1 (named NPC1-like 1, which is only expressed in human hepatocytes and the intestine) was shown to be involved in uptake of hepatitis C virus [191]. In contrast to filovirus infection, however, hepatitis C virus entry mediated by the NPC1-like 1 protein was dependent on cholesterol uptake and no evidence for an interaction between hepatitis C virus and NPC1-like 1 protein has been reported so far [191].

Taken together, the NPC1 protein is a filovirus receptor with unexpected characteristics: In comparison to cell surface receptors used by other viruses, NPC1 is unique in recognizing the filoviral glycoprotein only after uptake into the cell and after proteolytic processing. The identification of small molecule inhibitors directly or indirectly inhibiting NPC1-usage by GP_1,2 for host cell entry [20,129] makes it an attractive candidate for antiviral therapy.