CONSERVED RECEPTOR-BINDING DOMAINS OF LAKE VICTORIA MARBURGVIRUS AND ZAIRE EBOLAVIRUS BIND A COMMON RECEPTOR

The GP(1,2) envelope glycoproteins (GP) of filoviruses (marburg- and ebolaviruses) mediate cell-surface attachment, membrane fusion, and entry into permissive cells. Here we show that a 151-amino acid fragment of the Lake Victoria marburgvirus GP1 subunit bound filovirus-permissive cell lines more efficiently than full-length GP1. An homologous 148-amino acid fragment of the Zaire ebolavirus GP1 subunit similarly bound the same cell lines more efficiently than a series of longer GP1 truncation variants. Neither the marburgvirus GP1 fragment nor that of ebolavirus bound a nonpermissive lymphocyte cell line. Both fragments specifically inhibited replication of infectious Zaire ebolavirus, as well as entry of retroviruses pseudotyped with either Lake Victoria marburgvirus or Zaire ebolavirus GP(1,2). These studies identify the receptor-binding domains of both viruses, indicate that these viruses utilize a common receptor, and suggest that a single small molecule or vaccine can be developed to inhibit infection of all filoviruses.

The GP 1,2 envelope glycoproteins of filoviruses (marburg-and ebolaviruses) mediate cell-surface attachment, membrane fusion, and entry into permissive cells. Here we show that a 151-amino acid fragment of the Lake Victoria marburgvirus GP 1 subunit bound filovirus-permissive cell lines more efficiently than full-length GP 1 . A homologous 148-amino acid fragment of the Zaire ebolavirus GP 1 subunit similarly bound the same cell lines more efficiently than a series of longer GP 1 -truncation variants. Neither the marburgvirus GP 1 fragment, nor that of ebolavirus, bound a non-permissive lymphocyte cell line. Both fragments specifically inhibited replication of infectious Zaire ebolavirus, as well as entry of retroviruses pseudotyped with either Lake Victoria marburgvirus or Zaire ebolavirus GP 1,2 . These studies identify the receptor-binding domains of both viruses, indicate that these viruses utilize a common receptor, and suggest that a single small molecule or vaccine can be developed to inhibit infection of all filoviruses.
The filoviral GP 1 subunit mediates cell-surface receptor binding (8,11). Approximately half of the molecular weight of GP 1 is due to N-and Oglycans, many of which are located at the subunit's C-terminus, in a region described as the mucin-like domain (12,13). This domain contributes to cytopathicity observed in GP 1,2expressing cell lines, and has been suggested to play a critical role in pathogenesis of filoviral disease (14)(15)(16). However, its deletion enhances rather than decreases the efficiency of GP 1,2mediated infection (13,(16)(17)(18). Receptor binding is followed by endocytosis of the virions (19), acidification of the endocytotic vesicle (4,5,20), and proteolytic processing of GP 1 by endosomal cathepsins (18,21). Conformational changes in the filoviral GP 2 subunit facilitate lipid mixing and fusion of the viral and cellular membranes, in a sequence of steps thought similar to those mediated by orthomyxo-and retroviral transmembrane proteins (22)(23)(24)(25).
Here we identify fragments of the Lake Victoria marburgvirus (Musoke strain; MARV-Mus) and Zaire ebolavirus (Mayinga stain; ZEBOV-May) GP 1 subunit that efficiently bound cells permissive to filovirus infection, but not a non-permissive lymphocyte cell line. Each fragment inhibited infection of retroviruses pseudotyped with either marburgvirus or ebolavirus GP 1,2 . Both fragments also inhibited replication of infectious Zaire ebolavirus. Our data define homologous regions of otherwise divergent filoviruses that mediate association with a common receptor. Similarities in these receptorbinding domains may provide insight into the nature of this receptor and suggest vaccine and therapeutic approaches effective against all filoviruses.

EXPERIMENTAL PROCEDURES
Cells and culture conditions-African green monkey kidney (Vero E6) cells and Jurkat lymphocytes were obtained from the American Type Culture Collection (ATCC numbers CRL-1586 and TIB-152, respectively). Human embryonic kidney 293T cells are a derivative of 293 cells (ATCC CRL1573) created by S. Haase and described originally as 293/tsA1609neo (34). Adherent cells (Vero E6 and 293T) were maintained in DMEM (GIBCO-Invitrogen), and Jurkat lymphocytes in RPMI Medium 1640 (GIBCO-Invitrogen).
Cell-binding assays-293T cells and Vero E6 cells were detached with PBS/5mM EDTA (GIBCO-Invitrogen) 48 h after plating, resuspended in an equal volume of PBS/5mM MgCl 2 (SIGMA-Aldrich), and washed twice in PBS/2% goat serum (SIGMA-Aldrich). Jurkat lymphocytes were harvested and washed twice in PBS/2% goat serum. GP 1 -Fc constructs, truncation variants thereof, and control proteins were added to 5x10 5 cells to a final concentration of 100 nM, and incubated on ice for 1.5 h. Cells were washed twice in PBS/2% goat serum, and incubated for 45 min. on ice with a 1:40 dilution of goat Fc-specific fluorescein isothiocyanate (FITC) conjugated antihuman IgG antibody (SIGMA-Aldrich) in PBS/2% goat serum. Cells were washed three times with PBS/2% goat serum, once in PBS, and fixed with PBS/2% formaldehyde (SIGMA-Aldrich). Cell-surface binding of constructs was detected by flow cytometry with 10,000 events counted per sample. Baseline fluorescence was determined by measuring cells treated only with goat Fc-specific FITC-conjugate anti-human IgG antibody, which was then subtracted from binding values of the tested constructs and control proteins.
Infection assay with filovirus envelope glycoprotein-pseudotyped retroviruses-To generate retroviral pseudotypes, 293T cells were transfected by the calcium phosphate method with plasmid encoding MARV-Mus GP 1,2 , ZEBOV-May GP 1,2∆309-489 , or vesicular stomatitis Indiana virus (VSV) G protein, together with the pQCXIX vector (BD Biosciences) expressing green fluorescent protein (GFP), and plasmid encoding the Moloney murine leukemia virus (MLV) gag and pol genes (38) using equal concentrations of each plasmid. Cell supernatants were harvested 48 h post transfection, cleared of cellular debris by centrifugation and filtration through a 0.45 µm-pore size filter (Corning) and stored at 4°C. Supernatants containing pseudotyped viruses were added to 293T or Vero E6 cells in the presence or absence of the indicated concentrations of filovirus Fc truncation variants, or control proteins. After 5 h, cells were washed once in PBS, and replenished with fresh media After 48 h, cells were imaged by fluorescent microscopy, and detached with trypsin for analysis by flow cytometry.
Infection assay with recombinant greenfluorescent-protein-expressing Zaire ebolavirus-All experiments with infectious filovirus were performed under biosafety level 4 conditions. Vero E6 cells were infected with a GFPexpressing ZEBOV-May created by reverse genetics (39). Virus was incubated with cells at a multiplicity of infection equal to 1 for 1 h in the presence or absence of 800 nM of filovirus truncation variants or control protein. Virus was removed, cells were washed in PBS, and media and protein were replenished. After 48 h. cells were fixed in 10% neutral-buffered formalin. After 3 days of fixation, cells were removed from the biosafety level 4 suite in and the percent of GFPexpressing cells was measured with a Discovery-1 automated microscope (Molecular Devices Corp., Sunnyvale, CA) by measuring 9 individual spots per well.

MARV-Mus GP 1 truncation variant 38-188-Fc efficiently binds to filovirus-permissive cells-
The envelope glycoproteins of a number of viruses include discrete, independently folded domains that bind cellular receptors as efficiently as their entire ectodomain regions. We sought to identify similar receptor-binding domains (RBDs) of MARV-Mus and ZEBOV-May. To determine the location of the MARV-Mus GP 1 RBD, we synthesized a codon-optimized gene encoding the full-length mature MARV-Mus GP 1 protein fused to the Fc region of human immunoglobulin G1 at the C-terminus (17-432-Fc). Four sets of seven truncation variants were created, starting at Nterminal residues 17, 38, 61, or 87, and ending at C-terminal residues 432, 308, 265, 230, 188, 167, or 134 (Fig. 1A). All 28 constructs expressed efficiently in 293T cells as Fc-fusion proteins (Fig.  1B). Equivalent concentrations of each variant were incubated with MARV-Mus-permissive African green monkey kidney Vero E6 and human embryonic kidney 293T cells, and with nonpermissive Jurkat lymphocytes (5), and cellsurface association was determined by flow cytometry ( Fig. 2A-C). The RBD of the severe acute respiratory syndrome coronavirus (SARS-CoV) S protein (residues 318-510) and HIV-1 gp120, expressed as Fc-fusion proteins (SARS-CoV RBD-Fc, gp120-Fc), were used as controls (36,37). As previously reported, SARS-CoV RBD-Fc efficiently bound SARS-CoV-permissive Vero E6 cells but not 293T cells or Jurkat lymphocytes (40). Also expectedly, gp120-Fc bound CD4expressing Jurkat lymphocytes, but not Vero E6 or 293T cells. All 28 MARV-Mus proteins bound to Vero E6 and 293T cells, with varying efficiencies, whereas little or no association was observed with Jurkat lymphocytes in most cases. Successive truncation of the C-termini of MARV-Mus GP 1 variants initiated with residues 17, 38, 61, or 87 led to successively increased cell-surface binding to Vero E6 cells, up through the C-terminal truncation at residue 188 ( Fig. 2A). Further truncation beyond residue 188 decreased cell association. A single exception to this trend was observed with the 87-432-Fc variant, which bound Vero E6 cells with higher affinity than 87-308-Fc and 87-265-Fc. Variants initiated with residues 38, 61, and 87 bound more efficiently than those initiated with residues 17, with MARV-Mus 38-188-Fc consistently binding most efficiently to Vero E6 and 293T cells (Fig. 2B). These data identify a cell-binding region of MARV-Mus, located between GP 1 residues 38 and 188.

ZEBOV-May GP 1 truncation variant 54-201-Fc efficiently binds to filovirus-permissive cells-
Deletion of the mucin-like domain has been demonstrated to markedly increase efficiency of ZEBOV GP 1,2 -mediated infection (13,(16)(17)(18). To determine the location of the ZEBOV-May GP 1 RBD, we synthesized a codon-optimized gene encoding the mature ZEBOV GP 1 protein, lacking its mucin-like domain, and fused to the IgG1 Fc region (33-308-Fc). Three sets of four truncation variants were created, starting at N-terminal residues 33, 54, or 76, and ending at C-terminal residues 308, 201, 172, or 156 (Fig. 1C). With the exception of variant 76-172-Fc, all variants expressed efficiently (Fig. 1D). As with the MARV-Mus variants, equivalent concentrations of each variant were incubated with ZEBOV-Maypermissive Vero E6 and 293T cells, and with non-permissive Jurkat lymphocytes, and cell association was again assayed by flow cytometry. All 11 ZEBOV-May GP 1 variants bound to Vero E6 and 293T cells, whereas binding to Jurkat lymphocytes was negligible in all cases (Fig. 2D-F). ZEBOV-May GP 1 truncation variants showed a pattern of association to Vero E6 and 293T cells similar to that observed with MARV-Mus variants. In particular, 54-201-Fc and 76-201-Fc bound more efficiently than all other ZEBOV-May GP 1 variants assayed, with 54-201-Fc binding slightly but consistently better than 76-201-Fc to Vero E6 cells (Fig. 2D-E). These data identify a cell-binding region of ZEBOV-May, located between GP 1 residues 54 and 201, which corresponds to the cell-binding region of MARV-Mus (see Fig. 7 for alignment).
MARV strains Angola and Musoke GP 1 truncation variants bind to filovirus-permissive cells with comparable efficiency-The largest and most severe marburgvirus disease outbreak to date occurred in Angola in early 2005 (41,42). The envelope glycoprotein amino acid sequence of the strain responsible for this outbreak, MARV Angola (MARV-Ang), is homologous to that of the MARV-Mus strain (43). In particular, a comparison between MARV-Mus GP 1 amino acid residues 38-188 with the corresponding region of MARV-Ang yielded only one amino acid change, threonine 74 to alanine (T74A). This alteration was introduced into four MARV-Mus GP 1 truncation variants (MARV-Ang GP 1 38-188-Fc, 38-167-Fc, 61-188-Fc, and 61-167-Fc; Fig. 3A). Cell association of each of these variants was compared with those of MARV-Mus. Each MARV-Ang variant bound Vero E6 cells slightly less efficiently than its MARV-Mus counterpart (Fig. 3B). These data largely exclude the possibility that more efficient cellular association of the MARV-Ang cell-binding region contributes to increased severity of disease. MARV-Mus 38-188 Fc inhibits MARV/MLV entry more efficiently than other GP 1 truncation variants-We investigated whether the cell-binding efficiency of MARV-Mus and MARV-Ang GP 1 truncation variants correlated with their ability to inhibit entry of pseudotyped retroviruses (Fig. 5). Vero E6 cells were incubated with the indicated GP 1 variants together with VSV/MLV or MARV/MLV. None of the GP 1 variants inhibited VSV/MLV entry, whereas most of the MARV-Mus GP 1 variants assayed inhibited that of MARV/MLV (Fig. 5). Some variation between entry inhibition and cell-binding was observed. Notably, full-length MARV-Mus GP 1  inhibited MARV/MLV entry as efficiently as the defined receptor-binding domains of MARV-Mus and MARV-Ang (38-188-Fc). Apart from this interesting exception, the MARV-Mus RBD inhibited entry more efficiently than any other GP 1 variant assayed (Fig. 5). We speculate that the mucin-like domain of full-length GP 1 mediates a lower affinity interaction with Vero E6 cells which may contribute to inhibition of entry, but which may be more susceptible to the wash steps of the binding assay shown in Fig. 2. Alternatively, partial misfolding of the longer truncation variants may impair cell surface association. Our data show that variants of the MARV-Mus RBD that are slightly longer or shorter inhibit MARV/MLV less efficiently, consistent with their relatively lower affinity for filovirus-permissive cell lines.

MARV-Mus 38-188-Fc and ZEBOV-May 54-201 inhibit replication of infectious
Zaire ebolavirus-To determine if the filovirus RBDs also inhibited infectious filovirus, Vero E6 cells were incubated with an infectious Zaire ebolavirus modified to express GFP (39), at a multiplicity of infection of 1, together with MARV-Mus 38-188-Fc, ZEBOV-May 54-201-Fc, or SARS-CoV RBD-Fc. As expected, viral replication, measured as percentage of infected cells, was specifically inhibited by both filovirus RBDs, but not by that of SARS-CoV (Fig. 6). Higher concentrations were required to inhibit infectious filovirus than those used to inhibit pseudotyped retroviruses ( Fig. 4 and 6). These higher concentrations may be necessary to interfere with the greater number of GP 1,2 molecules present on the filamentous filoviruses, compared to the significantly smaller retroviral pseudotypes. As observed with pseudotyped retroviruses, the MARV-Mus RBD inhibited infectious Zaire ebolavirus more efficiently than the ZEBOV-May RBD (Fig. 6). Similar inhibition of Zaire ebolavirus replication was observed in primary monocyte-derived human dendritic cells treated with ZEBOV-May or MARV-Mus RBDs (data not shown). The efficiency with which the MARV-Mus RBD inhibited ebolavirus replication is consistent with the utilization of a common entry factor by both marburg-and ebolaviruses.

DISCUSSION
Enveloped viruses require specific proteins on the virion surface that mediate cell attachment and fusion of the viral and cellular membranes. Viral class I fusion proteins are typically comprised of two functionally distinct domains or subunits (44,45). The N-terminal domain, GP 1 in the case of filoviruses, mediates cell attachment and receptor association (8,11). Viral entry proteins attach to a number of cell-surface molecules including glycosoaminoglycans and C-type lectins, and these attachments frequently make substantial contributions to the efficiency of viral entry (46)(47)(48)(49). More critically, most enveloped viruses require one or more cellular receptors to initiate membrane fusion. Receptor-binding regions of viral fusion proteins are typically the most important antibody-neutralizing eptitopes on the virion, due to the functional importance of and limited variation in this region (44,45). In some cases, such as murine and feline leukemia viruses and SARS coronavirus, the receptor-binding region is localized to a discrete, independently folded domain that can efficiently bind the cellular receptor and inhibit infection (37,50,51). These domains themselves also can be sufficient to elicit protective neutralizing antibodies (45,52).
Here we defined small domains of the GP 1 proteins of two divergent filoviruses that bind filovirus-permissive cells. Several lines of evidence suggest that these domains bind a cellular receptor rather than a less specific attachment factor. First, these domains do not associate with a cell line refractory to filovirus infection. Second, they associate with filoviruspermissive cells more efficiently than larger and more heavily glycosylated GP 1 variants. Indeed, ZEBOV-May 54-201-Fc includes no Nglycosylation sites that could associate with a cellsurface lectin-like molecule (MARV-Mus 38-188-Fc has two potential N-glycosylation sites). Third, each domain efficiently inhibits entry mediated by their respective GP 1,2 at 50-200 nM, indicating that they associate with moderately high affinity and specifically with a factor critical to entry. Finally, they include the most highly conserved region of filovirus GP 1 (17). The conservation of this region among all marburg-and ebolaviruses raises the possibility that ZEBOV-May 54-201-Fc and MARV-Mus 38-188-Fc can be used to elicit antibodies that protect against most filoviruses.
Previous studies of Zaire ebolavirus GP 1,2 are also consistent with association of these domains with a specific cellular receptor. Medina (18,21).
Although the genomic organization of marburg-and ebolaviruses is similar, and although they cause similar diseases of comparable severity, it has not been clear whether all filoviruses utilize a common receptor. Several observations in the literature raised the possibility that their receptors or entry mechanisms are distinct. Lake Victoria marburgvirus has been reported to be less susceptible than Zaire ebolavirus to treatment of target cells with proteases and glycosidases (5). Electron micrographs of the virus entering cells have been used to suggest that Lake Victoria marburgvirus enters cells differently than Zaire ebolavirus (54), although earlier work suggests otherwise (19). Some variation in the relative efficiencies with which Lake Victoria marburgand Zaire ebolavirus GP 1,2 mediated entry in different cell lines also raised the possibility of distinct receptors (5).
Despite these observations, our data indicate that at least one of the receptors required by each filovirus is common to both. This situation is not unprecedented. For example SARS coronavirus and human coronavirus NL63 enter cells by distinct mechanisms although angiotensinconverting enzyme 2 is an obligate receptor for both (55,56). Further study will be necessary to clarify if the down-stream entry processes of marburg-and ebolaviruses are similarly distinct.
The conservation of the filovirus receptorbinding domains and their utilization of a common receptor raise the possibility that a vaccine could elicit antibodies that neutralize both marburg-and ebolaviruses, although cross-protective antibodies have not been described to date. Our observations also indicate that small molecules could be designed to inhibit entry of all filoviruses. Such cross-protection would be useful in the rapid containment of a novel filovirus epidemic.