The roles of ebolavirus glycoproteins in viral pathogenesis

Ebolaviruses are highly dangerous pathogens exhibiting extreme virulence in humans and nonhuman primates. The majority of ebolavirus species, most notably Zaire ebolavirus, can cause Ebola virus disease (EVD), formerly known as Ebola hemorrhagic fever, in humans. EVD is associated with case-fatality rates as high as 90%, and there is currently no specific treatment or licensed vaccine available against EVD. Understanding the molecular biology and pathogenesis of ebolaviruses is important for the development of antiviral therapeutics. Ebolavirus encodes several forms of glycoproteins (GPs), which have some interesting characteristics, including the transcriptional editing coding strategy and extensive O-glycosylation modification, clustered in the mucin-like domain of GP1, full-length GP (GP1,2), and shed GP. In addition to the canonical role of the spike protein, GP1,2, in viral entry, ebolavirus GPs appear to have multiple additional functions, likely contributing to the complex pathogenesis of the virus. Here, we review the roles of ebolavirus GPs in viral pathogenesis.


INTRODUCTION
The 2014 Ebola epidemic in West Africa was caused by the most pathogenic ebolavirus species, EBOV (Zaire ebolavirus), and resulted in 28, 652 suspected, probable, and confirmed cases, with 11, 325 deaths as of April 13, 2016 (U.S. Centers for Disease control and Prevention, 2016a). As before the 2014 outbreak, epidemics of highly pathogenic ebolaviruses had primarily occurred sporadically in central Africa, the study of these viruses had been somewhat neglected for many years. The unprecedented and massive damage to human health and the social economy caused by the recent outbreak raised the alarm regarding the potential large-scale re-emergence of such deadly viral diseases. To (Bukreyev et al., 2014) have been identified. In addition to EBOV, Sudan virus, Bundibugyo virus, and Taï Forest virus are also pathogenic to humans, while human infections with Reston virus are thought to be asymptomatic (Peters and Leduc, 1999;Feldmann and Geisbert, 2011;Rougeron et al., 2015;. Although many studies of ebolavirus (particularly EBOV) have been performed using animal infection models (Nakayama and Saijo, 2013;Shurtleff and Bavari, 2015), knowledge of the pathology and pathogenesis of human ebolavirus infections remains limited. Highly pathogenic ebolavirus infections in humans and non-human primates are characterized by immune suppression and dysregulated inflammatory responses which can cause uncontrolled virus replication, immune system impairment, and tissue damage and thus likely result in the manifestations of Ebola virus disease (EVD), such as high fever, hemorrhage, disseminated intravascular coagulation, multiorgan failure, and shock (Zaki and Gold-smith, 1999;Mahanty and Bray, 2004;Feldmann and Geisbert, 2011;Kortepeter et al., 2011;Ansari, 2014;Chertow et al., 2014;Singh et al., 2015).
GP 1,2 is the virus surface protein and forms trimeric spikes (Sanchez et al., 1998) that facilitate virus entry by mediating receptor binding and membrane fusion (Lee and Saphire, 2009). In addition to the essential role of GP 1,2 as the structural protein mediating virus entry, GP 1,2 as well as other GP variants have also been shown to, or are suspected to, have multiple functions in the virus life cycle and virus-host interactions, thus likely making a versatile contribution to the complex pathogenesis of ebolavirus.
Knowledge of viral protein functions and their contributions to viral pathogenesis will contribute to the development of prophylactic treatment to protect against ebolavirus infections. In this review, the function of ebolavirus GPs is briefly summarized, with a particular focus on their defined, or potential, roles in viral pathogenesis.
The roles of full-length GP (GP 1,2 ) in the ebolavirus life cycle As the virion surface spikes, trimeric GP 1,2 complexes mediate ebolavirus entry processes, the initial steps of viral infection, with GP1 acting as the receptor-binding subunit and GP2 as the membrane fusion subunit (Jiang et al., 2009;Lee and Saphire, 2009;Wang et al., 2011). The GP 1,2 -mediated entry processes and entry-associated structural and functional characterization of GP 1,2 have been intensively investigated in ebolavirus and are discussed extensively elsewhere (Lee et al., 2008;Lee and Saphire, 2009;Falzarano and Feldmann, 2015;Moller-Tank and Maury, 2015;Gong et al., 2016;Tang, 2016;Wang et al., 2016;White and Whittaker, 2016). Due to its ability to bind to a variety of cell surface molecules, GP 1,2 largely determines the broad cellular tropism of ebolaviruses, which can infect a wide range of cell types, including immune cells (monocytes, macrophages, and dendritic cells) and many other tissue cell types (endothelial cells, hepatocytes, adrenal cortical cells, and some epithelial cells, among others) (Feldmann and Geisbert, 2011;Takada, 2012;Martines et al., 2015;Singh et al., 2015). The infection of these cells by ebolaviruses and the resultant disturbance of normal cell physiology undoubtedly have important roles in the complicated pathogenesis of ebolavirus. For example, aside from the disorders of host immune/inflammatory responses caused by the infection of immune cells, infected monocytes, macrophages, and dendritic cells likely contribute to viral spread from the initial site of infection to regional lymph nodes, and the liver, spleen, and other tissues/organs, through the blood and lymphatic systems, especially as these immune cells are early and preferred replication sites of ebolaviruses (Schnittler and Feldmann, 1998;Geisbert et al., 2003;Bray and Geisbert, 2005). Moreover, infections in the liver, and the resulting hepatocellular necrosis, could impair synthesis of coagulation factors and other plasma proteins, contributing to the hemorrhagic tendencies observed in infected individuals, while adrenocortical infection and necrosis could damage adrenocortical function and inhibit steroid synthesis, promoting the development of hypotension, hypovolaemia, and shock that are often noted in EVD cases. The innate immune system, and particularly the type I interferon (IFN) response, provides a stubborn line of defense in resistance to viral infections. IFN induction and signaling lead to the establishment of the host antiviral state by inducing the expression of more than 300 IFNstimulated genes which can limit viral replication and spread by targeting multiple aspects of the viral life cycle (Stark, 2007;Randall and Goodbourn, 2008;Sadler and Williams, 2008;Schneider et al., 2014;Errett and Gale, 2015). However, viruses, especially highly pathogenic hemorrhagic fever viruses, have evolved various strategies to antagonize these antiviral responses (Elliott and , among which the GP gene encodes three GP precursors, resulting in multiple GP protein products. The primary product of the GP gene is pre-sGP, which is expressed from the majority of RNA transcripts (e.g., approximately 71% of the total transcripts in EBOV-infected Vero E6 cells) with no shift of ORF and can be cleaved by furin at its C-terminus, yielding an Nglycosylated sGP dimer and an O-glycosylated Δ-peptide. Transcriptional editing can occur at a series of seven uridine residues in the GP gene, resulting in corresponding changes in the number of adenosine (A) residues in the transcripts. A + 1 shift results in an extended ORF, encoding the full-length GP (approximately 24% of transcripts), while a + 2 shift leads to the synthesis of the ssGP from the truncated ORF (approximately 5% of transcripts). Note that the addition of more A residues or the deletion of a single A nucleotide has also been observed, although transcripts containing 7A, 8A, or 9A at the transcriptional editing site are the most common, and encode sGP, full-length GP, or ssGP, respectively. The full-length GP precursor is cleaved by furin to form a disulfide-linked GP1-GP2 dimer, which subsequently assembles into the GP 1,2 trimer and locates to the plasma or viral membrane. GP 1,2 trimers on the viral membrane function as virion surface spikes, facilitating virus entry. A feature of GP 1,2 is that it contains not only N-glycans, but also extensive O-glycans, clustered in the mucin-like domain (MLD) of the GP1 subunit. As reflected in the diagram, in the native structure of GP1, the MLD (red) sits above the N-terminal domain. Some surface GP 1,2 can be further cleaved by the TNF-α converting enzyme (TACE) at the membrane-proximal external region to release shed GP. Like sGP, mature ssGP also forms a dimer and is N-glycosylated. ssGP, full-length GP (specifically GP1), and sGP share a common Nterminus (green), but differ in their C-termini, thus exhibiting diverse functions. Weber, 2009;Borrow et al., 2010;Ye et al., 2013;Ning et al., 2014;Ma and Suthar, 2015;Messaoudi et al., 2015;Ning et al., 2015). VP35 and VP24 are well-known IFN antagonists of ebolaviruses (Audet and Kobinger, 2015;Basler, 2015;Messaoudi et al., 2015), and the fulllength GP also exhibits an antagonistic activity against the innate immune system by targeting tetherin, an IFNinduced antiviral factor (Van damme et al., 2008;Kaletsky et al., 2009;Tokarev et al., 2009). Tetherin is a typeⅡtransmembrane protein containing an N-terminal TMD and a C-terminal glycosylphosphatidylinositol (GPI) membrane anchor, which is considered an unusual topology (Kupzig et al., 2003). Tetherin restricts the release of some enveloped viruses (including ebolaviruses) from infected cells by retaining virions on the cell surface through simultaneous anchoring to both the plasma membrane and the virus envelope (Jouvenet et al., 2009;Perez-Caballero et al., 2009;Tokarev et al., 2009;Fitzpatrick et al., 2010;Hammonds et al., 2010;Hinz et al., 2010;Le tortorec et al., 2011). Ebolavirus full-length GP can counteract the antiviral activity of tetherin, facilitating the completion of virus budding (Kaletsky et al., 2009). Thus, in addition to its essential roles in viral entry, full-length GP also contributes to viral budding, not only by participating in virus assembly as the structural spike glycoprotein, but also by antagonizing tetherinmediated antiviral activity. Although the mechanism underlying the antagonism of tetherin by full-length GP has not yet been clearly defined, it was reported that the ebolavirus GP does not remarkably effect the expression level of tetherin on the cell surface, or prevent the association of tetherin with lipid rafts, but may block the interaction of tetherin with VP40, a primary matrix protein with a key role in virus assembly and budding, and this may contribute to the observed antagonistic activity (Kaletsky et al., 2009;Lopez et al., 2010;Kuhl et al., 2011;Lopez et al., 2012;Gustin et al., 2015;Vande Burgt et al., 2015).

The pathogenicity of full-length GP (GP 1,2 ) and its MLD
In addition to its pivotal contributions to the virus life cycle, the GP 1,2 protein itself also exhibits notable direct pathogenicity to host cells. Expression of EBOV GP 1,2 causes evident rounding and detachment of adherent cells, such as cultured human or non-human endothelial cells, epithelial cells, and macrophages (Chan et al., 2000;Takada et al., 2000;Yang et al., 2000;Simmons et al., 2002). GP 1,2 from other ebolavirus species can also exert similar effects on cells, albeit at lower levels than EBOV GP 1,2 , paralleling the differences in virulence observed among ebolavirus species (Simmons et al., 2002). Importantly, GP 1,2 expression from a replication-defective adenoviral vector induces massive endothelial cell detachment in explanted blood vessels and thus substantially increases vascular permeability . Interestingly, EBOV GP 1,2 causes endothelial disruption in both human and non-human primate blood vessels, whereas GP 1,2 from Reston virus damages non-human primate but not human blood vessels, further indicating the role of GP 1,2 as a virulence factor mediating vascular damage .
The MLD is required for GP 1,2 -induced toxicity to both cultured adherent cells and blood vessel explants Simmons et al., 2002). In addition, cell surface expression of full-length GP, including a TMD, is needed for the GP 1,2 activity; however, the requirement for a TMD does not appear to be for the specific TMD of GP2, as the toxicity of GP 1,2 is maintained when the GP2 TMD is substituted with other TMDs (including those derived from influenza virus hemagglutinin, and Moloney murine leukemia virus envelope protein) (Takada et al., 2000;Yang et al., 2000). Moreover, surface expression of the MLD fused with an exogenous TMD is sufficient to induce cytopathic effects (CPE), including cell rounding and detachment Francica et al., 2009).
Mechanistic studies have demonstrated that the detectable levels of cellular surface molecules, such as integrins, intercellular cell adhesion molecule-1(ICAM-1), platelet/endothelial cell adhesion molecule-1 (PECAM-1), vascular cell adhesion molecule-1 (VCAM-1), epidermal growth factor receptor (EGFR), and major histocompatibility complex class I (MHC-I) appear to decrease in GP 1,2 -expressing cells (Takada et al., 2000;Simmons et al., 2002). Initially, it was considered that the surface expression of these molecules was down-regulated by GP 1,2 /MLD; however, further studies suggested that MLD, acting as a "glycan umbrella", sterically shields the epitopes and functions of cellular surface proteins, rather than affecting their abundance (Reynard et al., 2009;Francica et al., 2010). Considering the significant roles of these surface proteins in physiological functions of cells, such as adhesion and immune/inflammatory signaling, GP 1,2 /MLD-mediated cytotoxicity may contribute significantly to the inflammatory dysregulation, immune suppression, and vascular damage characteristics of ebolavirus pathogenesis.
As noted above, it is reasonable to assume that the shielding of surface molecules of endothelial cells explains GP 1,2 -induced endothelial CPE and vascular damage, while the functional impairment of immune cell surface proteins would interfere with the immune and inflammatory responses. Moreover, it has indeed been shown that GP 1,2 expression blocks CD8+ T cell-recognition of MHC-I on antigen-presenting cells, impairing MHC-I antigen presentation (Francica et al., 2010). Steric shielding of cell surface immune molecules by the massive glycans of MLD is a novel mechanism of viral immune evasion. Furthermore, although antibodies can be raised against the MLD, this region is unnecessary for ebolavirus entry and highly variable compared with the other parts of the full-length GP (Wilson et al., 2000;Jeffers et al., 2002). Interestingly, MLD glycans mask antigenic epitopes of the core structure (including the receptor binding domain) of GP 1,2 itself, beyond the MLD, and thereby likely blocking the activities of neutralizing antibodies, perhaps representing another MLD-mediated immune escape strategy of ebolaviruses (Lee et al., 2008;Reynard et al., 2009;Francica et al., 2010;Dias et al., 2011;Martinez et al., 2011;Cook and Lee, 2013;Misasi et al., 2016;Pallesen et al., 2016).
In addition to the MLD, a putative immunosuppressive domain (ISD) close to the C-terminus of GP 1,2 may also contribute to viral immune suppression by inhibiting T cell cycle progression and inducing T cell apoptosis, thus likely promoting the lymphocyte depletion observed in severe infections (Volchkov et al., 1992;Becker, 1995;Baize et al., 1999;Zaki and Goldsmith, 1999;Geisbert et al., 2000;Yaddanapudi et al., 2006).

SHED GP
Shed GP is a product of the proteolytic cleavage of cell surface GP 1,2 by TNF-α-converting enzyme (TACE) (Dolnik et al., 2004) (Figure 1). Shed GP is released from the cell surface in significant amounts in a soluble trimeric form (Dolnik et al., 2004). The proteolytic cleavage results in the removal of a short transmembrane anchor from the GP2 subunit of GP 1,2 (Figure 1). Due to the preservation of its antigenic properties, shed GP can be recognized by anti-GP 1,2 antibodies and exhibits a decoy function, sequestering anti-GP 1,2 antibodies that would otherwise bind to GP 1,2 , and potentially induce the elimination of viruses and infected cells (Dolnik et al., 2004).
Recently, shed GP was shown to trigger the activation of non-infected dendritic cells and macrophages, leading to profuse expression of inflammatory cytokines, in a similar manner to GP 1,2 , (Escudero-Perez et al., 2014). Interestingly, anti-TLR4 antibodies can efficiently block shed GP-mediated activation of immune cells; moreover, treatment of shed GP with either deglycosylases or mannose-binding lectin can inhibit the activity of shed GP (Escudero-Perez et al., 2014). These results suggest that cellular TLR4 signaling and shed GP surface glycans are likely involved in the inflammatory activation associated with ebolavirus infection. In addition, like GP 1,2 , shed GP may also activate LSECtin signaling and trigger pro-inflammatory cytokine secretion by dendritic cells .
Furthermore, treatment of HUVEC monolayers with shed GP increases their permeability, indicating a direct effect of shed GP in damage of the endothelial barrier . Combined with the significant activities of some inflammatory cytokines, particularly TNF-α (Wahl-Jensen et al., 2005b), in impairment of endothelial barrier integrity, these data suggest that the disruption of endothelial barriers can be mediated directly by shed GP itself, by an unknown mechanism, or indirectly by shed GP-induced inflammatory cytokines (Escudero-Perez et al., 2014).
Additionally, aside from the RNA-editing strategy for regulation of GP expression (Volchkov et al., 2001), the release of shed GP may be another mechanism to modulate the abundance of cell surface GP 1,2 , and hence virion GP 1,2 content and cytotoxicity, and this process likely has a role in orchestrating optimal ebolavirus infectivity and spread (Dolnik et al., 2015;Mohan et al., 2015).
Previous studies have suggested that sGP may exhibit versatile immunomodulatory functions (Kindzelskii et al., 2000;Ito et al., 2001;Sui and Marasco, 2002;Wahl-Jensen et al., 2005b;Mohan et al., 2012;de La Vega et al., 2015). Since sGP is the main product of the GP gene and shares a common N-terminus (295 aa residues) with fulllength GP, it was hypothesized that, like shed GP, sGP acts as a decoy antigen by adsorbing the antibodies against GP 1,2 , thus counteracting antibody-mediated clearance of viral infection (Wilson et al., 2000;Ito et al., 2001). Furthermore, a model of sGP-mediated "antigenic subversion" was proposed, which postulated that the much higher quantity of sGP may lead to sGP domination of host humoral responses, and divert the immune response away from GP 1,2 (Mohan et al., 2012). sGP may also exhibit some anti-inflammatory activities, including inactivation of neutrophils and partial restoration of the TNF-α-induced decrease of the barrier function of endothelial cells (Kindzelskii et al., 2000;Sui and Marasco, 2002;Wahl-Jensen et al., 2005b), although the molecular mechanisms underlying these phenomena have not yet been determined.
In addition to being extracellularly secreted as a nonstructural protein, sGP appears to be able to assemble with GP2 as a substitute for GP1 and, intriguingly, the sGP-GP2 complex can also mediate infection in a vesicular stomatitis virus (VSV) pseudotype model, indicating a potential role for sGP as a structural protein (Iwasa et al., 2011). However, the biological significance of the replacement of GP1 by sGP in GP 1,2 has yet to be determined.

Δ-PEPTIDE
Δ-peptide was first described by Volchkova et al. during the course of an investigation of the processing and maturation of sGP (Volchkova et al., 1999). sGP is N-glycosylated, while Δ-peptide is O-glycosylated (Volchkova et al., 1999) (Figure 1). The production of Δ-peptide is conserved across all ebolavirus species, suggesting an important role (or roles) for the peptide. Radoshitzky et al. reported that Δ-peptides of several pathogenic ebolaviruses (EBOV, Sudan virus, and Taï Forest virus) can efficiently bind to filovirus-permissive cells and inhibit Marburg virus (MARV, another filovirus) and ebolavirus GP 1,2 -mediated cell entry (Radoshitzky et al., 2011), even though MARV does not produce Δ-peptides. In contrast, the Δ-peptide of Reston virus exhibits less-efficient cell binding and no obvious inhibitory effect on cell entry of filoviruses, indicating a potential role for Δ-peptide in determining viral virulence (Radoshitzky et al., 2011). Suppression of ebolavirus entry by Δ-peptide, which is a post-infection product, may prevent virus superinfection and thereby benefit viral spread and systematic infection. Although the mechanism of Δpeptide-mediated inhibition of filovirus infection remains unclear, the impairment of cell entry of both MARV and ebolaviruses by Δ-peptides suggests that these peptides probably interfere with a common pathway involved in filovirus entry (Radoshitzky et al., 2011).
Recently, Gallaher and Garry performed sequence analyses and computational modeling to identify a conserved amphipathic region in the Δ-peptides of filoviruses (particularly EBOV) with high similarity to the cytolytic peptide motif of rotavirus nonstructural protein 4 (NSP4) (Gallaher and Garry, 2015). The authors hypothesized that Δ-peptides may serve as membrane-damaging viroporins, thus contributing to ebolavirus pathogenesis (Gallaher and Garry, 2015). The membrane-association potential of Δ-peptides may explain the retention of Δ-peptides in cells after secretion of the bulk of sGP has occurred (Gallaher and Garry, 2015); nevertheless, whether Δ-peptides exhibit such biological functions requires experimental verification.

SMALL SOLUBLE GLYCOPROTEIN (ssGP)
The existence of ssGP, which is encoded by the + 2-shift reading frame of the GP gene, had long been postulated before being experimentally verified by Mehedi et al. (Mehedi et al., 2011). The similar molecular masses of sGP and ssGP, the low abundance of ssGP (encoded by < 5% of GP transcripts) (Figure 1), and the lack of ssGPspecific antibodies make the detection of ssGP expression difficult by conventional methodology (Mehedi et al., 2011). To date, the function of ssGP is unknown. ssGP is essentially a truncated form of sGP and the two proteins share an identical N-terminal sequence (295 aa) (Figure 1). Moreover, mature ssGP is extensively Nglycosylated and forms homodimers, in the same way as sGP (Mehedi et al., 2011) (Figure 1). However, despite similarities in the biochemical properties and primary sequences of sGP and ssGP, ssGP does not demonstrate the anti-inflammatory activity exhibited by sGP that reverses TNF-α-induced loss of endothelial barrier function (Mehedi et al., 2011;Wahl-Jensen et al., 2005b).

CONCLUSION
Underestimating highly virulent pathogens, such as EBOV, would be immensely costly to society. The 2014 Ebola outbreak not only resulted in devastating health effects and pronounced socio-economic impacts in West Africa, but also caused a global panic, particularly with the occurrence of imported Ebola cases in countries out- Table 1. Summary of the known or potential functions of ebolavirus GPs

GPs
Functions References Full-length GP (GP 1,2 ) Mediates virus entry as the virion surface spike Lee and Saphire., 2009 Promotes virus budding by antagonizing tetherin Kaletsky et al., 2009 Sterically shields the epitopes and functions of cellular surface proteins via the MLD, causing rounding and detachment of cultured cells, endothelial cell damage, leakage of explanted blood vessels, and loss of cell physiological functions (such as antigen presentation by MHC-I) Chan et al., 2000;Takada et al., 2000;Yang et al., 2000;Simmons et al., 2002;Reynard et al., 2009;Francica et al., 2010 Sterically shields the epitopes of the GP 1,2 core via the MLD, blocking recognition by neutralizing antibodies (?) Reynard et al., 2009;Francica et al., 2010 Activates MΦ/DCs and triggers the secretion of inflammatory cytokines by the MLD (?), likely contributing to the excessive inflammation in EVD Wahl-Jensen et al., 2005a;Ye et al., 2006;Martinez et al., 2007 Contains a putative ISD, mediating T cell dysfunction/apoptosis (?) Volchkov et al., 1992;Yaddanapudi et al., 2006 Shed GP Functions as a decoy for anti-GP 1,2 antibodies, contributing to viral immune evasion Dolnik et al., 2004 Activates MΦ/DCs leading to the secretion of inflammatory cytokines; increases the permeability of HUVEC monolayers Escudero-Perez et al., 2014 Its release modulates the abundance of surface GP 1,2 , likely orchestrating virus cytotoxicity, infectivity, and spread (?) Dolnik et al., 2015 Contains a putative ISD, mediating T cell dysfunction/apoptosis (?) Volchkov et al., 1992;Yaddanapudi et al., 2006 sGP Functions as a decoy of anti-GP 1,2 antibodies, or mediates "antigenic subversion", diverting the immune response away from GP 1,2 (?) Wilson et al., 2000;Ito et al., 2001;Mohan et al., 2012 Inactivates neutrophils and reverses TNF-α-induced injury of endothelial barriers, playing anti-inflammatory roles (?) Kindzelskii et al., 2000;Sui and Marasco., 2002;Wahl-Jensen et al., 2005b Assembles with GP2 as a substitute for GP1, perhaps as a structural protein (?) Iwasa et al., 2011 Δ-peptide Binds to filovirus-permissive cells and inhibits filovirus GP 1,2 -mediated cell entry Radoshitzky et al., 2011 Contains an amphipathic region similar to the cytolytic peptide motif of rotavirus NSP4 and may serve as a membranedamaging viroporin (?) GPs (especially GP 1,2 ) of ebolaviruses are important, and potentially preferred targets, for the prevention and treatment of ebolavirus infections, for example as targets for potential therapeutic antibodies Bornholdt et al., 2016;Corti et al., 2016;Furuyama et al., 2016;Howell et al., 2016;Misasi et al., 2016;Pallesen et al., 2016;Wec et al., 2016). Among the GP variants, sGP, GP 1,2 , and shed GP are the main viral GP products and their functional roles are more fully characterized. We propose a model for the critical roles of GP 1,2 , shed GP, and sGP in virus infection and pathogenesis, which is summarized in Figure 2. In addition to the direct roles of GP 1,2 in the virus life cycle, the three GP variants are all believed to contribute to ebolavirus immune evasion by multiple strategies, thus contributing to uncontrolled virus infection and spread ( Figure 2). Moreover, GP 1,2 and shed GP may also promote pathological lesions and the development of clinical manifestations of EVD such as high fever, hemorrhage, shock, and organ dysfunction, by their direct effects (such as cytotoxicity) and the indirect effects induced by excessive inflammation (Figure 2). Despite the important achievements of studies of ebolavirus GPs, many key facets of the functions of these proteins remain to be elucidated. For example, although many investigations have indicated that GP 1,2 can induce cytotoxicity and vascular damage, to date there is no in vivo data verifying these roles of GP 1,2 . The physiological relevance of the observed functions of GPs in vitro needs to be further investigated in vivo using animal models, along with various GP gene-engineered recombinant ebolaviruses (Groseth et al., 2012). Functional and structural studies of the soluble GPs, especially ssGP and Δ-peptide, are relatively scarce, compared with those of GP 1,2 . In addition to GP 1,2 and the soluble GPs discussed above, other products of the GP gene have also been identified (Sanchez et al., 1998;Volchkov et al., 1998b); these include the short transmembrane anchor remaining after the release of shed GP, secreted GP1 (separate from GP2), and lone GP2. Whether these pro-teins are merely by-products of the GP gene or have significant biological functions requires further investigation.
Since the majority of studies of ebolaviruses have focused on EBOV, the molecular biology and pathogenesis of ebolaviruses are best characterized for this species; however, we consider that the notable variation in the virulence of different ebolavirus species provides an opportunity to unravel viral pathogenesis, and that comparative studies of different species and diverse hosts would be of interest and are merited. Additionally, insights from ebolaviruses will be valuable in enabling better understanding of the biology and pathogenesis of other pathogens causing hemorrhagic fever, such as MARV and Crimean-Congo hemorrhagic fever virus (a bunyavirus), which also encode GPs with a heavily O-glycosylated MLD (Sanchez et al., 2006;Wertheim and Worobey, 2009). It is reasonable and interesting to hypothesize that these GPs and MLDs may play similar significant roles in the virulence and pathogenesis of these various pathogens.

ACKNOWLEDGMENTS
This work was supported by the National Natural Science Foundation of China (No. 31125003 and No. 31321001) and the Basic Work Program of the Ministry of Science and Technology of China (2013FY113500).

COMPLIANCE WITH ETHICS GUIDELINES
The authors declare that they have no conflicts of interest. This article does not contain any studies with human or animal subjects performed by any of the authors. Figure 2. Model for the roles of GP 1,2 , shed GP, and sGP in ebolavirus pathogenesis. GP 1,2 , shed GP, and sGP constitute the main GP products of ebolaviruses and are more fully characterized functionally than other GPs. Severe ebolavirus infections are typified by uncontrolled virus infection and spread, and systemic inflammation, to which GP 1,2 , shed GP, and sGP likely contribute in various ways. In addition to the essential roles of GP 1,2 in the virus life cycle of facilitating virus entry and assembly, all three GPs are involved in virus immune escape, leading to uncontrolled disseminated virus replication and consequent increased GP production. GP 1,2 antagonizes tetherin to promote virus budding, mediates steric blocking of cell surface immune molecules (such as MHC-I) and perhaps the GP 1,2 core via the extensively O-glycosylated MLD, and may induce T cell dysfunction/apoptosis via the putative immunosuppressive domain (ISD) adjacent to the TACE cleavage site (indicated by a box). Shed GP retains the ISD and thus has the potential capacity to target T cells. Shed GP and sGP may both inhibit anti-GP 1,2 antibody-mediated clearance of viral infection by acting as decoy antigens. Moreover, sGP may also exert the same effect by inducing "antigenic subversion". On the other hand, GP 1,2 and shed GP both activate macrophages (MΦ) and dendritic cells (DCs) and induce inflammatory cytokine secretion by triggering TLR4 and LSECtin signaling pathways. In addition, GP 1,2 directly induces CPE, including cell rounding and detachment, and impairs cell physiological functions by blocking cell surface molecules with the MLD, leading to endothelial damage and vascular leakage and other tissue/organ injury. Shed GP can also directly cause increased endothelial permeability by an undefined mechanism. Both the direct toxicity of GP 1,2 and shed GP, and their inflammation-mediated pathogenic characteristics, may eventually contribute to the clinical manifestations of EVD, such as fever, hemorrhage, shock, and multi-organ dysfunction.