A Critical Domain of Ebolavirus Envelope Glycoprotein Determines Glycoform and Infectivity

Ebolaviruses comprises 5 species that exert varying degrees of mortality/infectivity in humans with Reston ebolaviruses (REBOV) showing the lowest and Zaire ebolaviruses (ZEBOV) showing the highest. However, the molecular basis of this differential mortality/infectivity remains unclear. Here, we report that the structural features of ebolavirus envelope glycoproteins (GPs) and one of their counter receptors, macrophage galactose-type calcium-type lectin (MGL/CD301), play crucial roles in determining viral infectivity. The low infectivity of REBOV mediated by the interaction between GPs and MGL/CD301 dramatically increased when the N-terminal 18 amino acids (33rd through 50th) of GPs were replaced with that of ZEBOV. Furthermore, structural analysis of glycans of GPs revealed that N-glycans were more extended in REBOV than in ZEBOV. N-glycan extension was reversed by the replacement of aforementioned N-terminal 18 amino acid residues. Therefore, these data strongly suggest that extended N-glycans on GPs reduce MGL/CD301-mediated viral infectivity by hindering the interaction between GPs and MGL/CD301 preferentially binds O-glycans.


MGL/CD301 plays an important role in the infection of VSV pseudotyped with RGP and ZGP
into monocyte-derived immature dendritic cells. Attachment of EBOV to host cells is known to be mediated by the so-called mucin-like domain of GPs, with surface lectins of hepatocytes, dendritic cells, macrophages, and endothelial cells apparently being engaged with viral attachment and subsequent viral entry, as stated above. Among several C-type lectins expressed on the surfaces of dendritic cells and macrophages, a predominant role was predicted for MGL/CD301 based on our previous findings showing that infection of VSV pseudotyped with ZGP was significantly enhanced in K562 cells expressing MGL/CD301 (K562-MGL/CD301 cells) 18 . We report here that a specific blocking monoclonal antibody against MGL/CD301, mAb MLD-1 21 , significantly reduced the infectivity of VSV pseudotyped with ZGP or RGP in human monocyte-derived immature DCs. As shown in Fig. 1a, these pseudotyped viruses infected human monocyte-derived immature DCs, with VSV pseudotyped with RGP being less infective than VSV pseudotyped with ZGP. Pre-incubation of immature DCs with mAb MLD-1 reduced the infection rate to ~40% compared to an isotype control antibody (Fig. 1a). According to these results, we judged MGL/CD301 on the surface of immature DCs plays a predominant role in infection of these cells with EBOV. The results also indicate that structural characteristics of EBOV GPs determine the differential viral infectivity, because the only difference between VSV pseudotyped with RGP and VSV pseudotyped with ZGP lies in the structures of their GPs. Therefore, we focused on MGL/CD301 and GPs to explore the molecular basis of the differential infectivity between REBOV and ZEBOV.
Infectivity of VSV pseudotyped with GPs correlates with the binding of MGL/CD301 to the GPs from VLPs. MGL/CD301-mediated infection of VSV pseudotyped with GPs (derived from ZEBOV, Sudan ebolavirus (SEBOV) and REBOV) was examined using K562-MGL/CD301 cells. All of the pseudotyped viruses infected K562-MGL/CD301 cells more efficiently than K562-mock cells (Fig. 1b). VSV pseudotyped with ZGP or SEBOV GP (SGP) showed significantly higher infectivity to cells expressing MGL/CD301 than VSV pseudotyped with RGP (Fig. 1b). Again, these results strongly suggest that the structural features of GPs determine the degree of the infection through their binding to MGL/CD301. To investigate this correlation further, the binding of recombinant MGL/CD301 to electrophoretically separated viral proteins of VSV pseudotyped with ZGP, ZGP lacking the mucin-like domain (Δmuc), SGP, or RGP was compared by immunoblotting. As shown in Fig. 1c, MGL/CD301 could bind to ZGP/SGP much stronger than ZGP Δmuc/RGP, whereas similar differences were not detected with the binding of anti-GP1 mAb 42/3.7, which recognizes a common epitope in RGP and ZGP 22 19,23,24 . Similar to the results of the immunoblotting analysis with MGL/CD301 or mAb 42/3.7 using VSV pseudotyped with GPs, the binding of MGL/CD301 to VLP-ZGP was higher than to VLP-RGP or VLP-ZGPΔmuc (Fig. 1d). Taken together, the binding intensity of MGL/CD301 to GPs from VLPs correlated with the infectivity of pseudotyped viruses.
The primary amino acid sequence of the so-called mucin-like domain of ZGP is not responsible for the differential infectivity and MGL/CD301 binding. The EBOV GP mucin-like domain (amino acid residues 311-462 of GP1) is rich in both N-and O-glycans 3 . The primary amino acid sequences of this domain are distinct between ZGP and RGP. The amino acid sequence homology between full-length ZGP and RGP is 58.1%, whereas that of the mucin-like domain is only 16.7% (see Fig. S1 for amino acid sequences). Thus, we thought that the primary amino acid sequence of the mucin-like domain determines the differential viral infectivity and differential MGL/CD301 binding. To test this hypothesis, we generated both VSV and VLP bearing a chimeric GP in which the mucin-like domain of ZGP was replaced with that of RGP. Unexpectedly, the infectivity of VSV pseudotyped with the chimeric GP was similar to that of VSV pseudotyped with ZGP in K562-MGL/CD301 cells ( Fig. 2a; Z311-462R). Also, the binding of MGL/CD301 to chimeric GP from VLP was similar to that of ZGP from VLP ( Fig. 2b; Z311-462R). These results indicate that the primary structure of the mucin-like domain is not responsible for the differences in MGL/CD301-mediated infectivity or the differential binding capacity to MGL/CD301 between ZGP and RGP. We could not test chimeric RGP carrying the mucin-like domain of ZGP, because the production of both pseudotyped VSV and VLP containing this domain was inefficient for unknown reasons, although the presence of this glycoprotein was confirmed on the cell surface of transfected HEK293T cells. Considering that Ser/Thr residues, representing potential O-glycosylation sites, are also present outside of the mucin-like domain, various lengths of polypeptide stretches around the mucin-like domain of ZGP were replaced with the corresponding region of RGP based on amino-acid sequence homology (Fig. 2a). Such chimeric GPs were examined for viral infectivity in K562-MGL/CD301 cells and MGL/ CD301's binding capacity to GPs. Even when 276 amino acids (amino acids 187 to 462) including the mucin-like domain (151 amino acids) of ZGP were replaced with the corresponding region of RGP, the infectivity of VSV pseudotyped with these chimeric GPs was very similar to that of VSV pseudotyped with ZGP (Fig. 2a). Similarly, Figure 1. A positive correlation was found between the infectivity of VSV pseudotyped with EBOV GPs and MGL/CD301 binding to the GPs. (a) Blockade of infection of VSV pseudotyped with ZGP or RGP to monocytederived dendritic cells by a monoclonal antibody specific for MGL1/CD301 (mAb MLD-1). In this calculation, the number of infected cells with VSV pseudotyped with ZGP treated with mouse IgG1 was set to 100%, and each relative infectivity was compared by the ratio to this value. For statistical analysis, Student's t-test was performed. *p < 0.05, **p < 0.01. (b) Effect of MGL/CD301 expression on the infection of VSV pseudotyped with ZGP, SGP, or RGP. All experiments were performed 3 times, and representative results are shown. In this calculation, the number of infected K562-mock cells with VSV pseudotyped with ZGP, RGP, or SGP was set to 100%, respectively, and the relative infectivity of K562-MGL/CD301 was shown by the ratio to this value. For statistical analysis, Student's t-test was performed. the binding capacity of MGL/CD301 to these chimeric GPs was almost identical to the binding capacity to ZGP (Fig. 2b). However, to our surprise, replacement of amino acid residues 1-462 or 33-462 from ZGP into RGP dramatically reduced both the infectivity to K562-MGL/CD301 cells and the binding capacity to MGL/CD301 compared to ZGP (Fig. 2a,b). These results made us hypothesize that the amino acid residues 33-186 of GP might regulate the infectivity to K562-MGL/CD301 cells and the binding capacity to MGL/CD301. The amino acid residues 33-50 of ZGP and RGP are critical to the differential infectivity and the differential MGL/CD301 binding between ZEBOV and REBOV. To further investigate the role of amino acid residues 33-186 of GP on viral infectivity and binding capacity to MGL/CD301, we generated VSVs pseudotyped with chimeric ZGP/RGP and VLPs, both containing the swapped 33-186 amino acid residues of RGP/ZGP (Fig. 3a). Then, we examined the viral infectivity in K562-MGL/CD301 cells and the binding capacity to MGL/CD301. The infectivity of VSV pseudotyped with chimeric ZGP (ZGP 33-186 swapped into RGP) was dramatically reduced compared to ZGP ( Fig. 3a; Z33-186R). Similarly, the binding capacity of MGL/CD301 to the chimeric GP was dramatically reduced compared to ZGP (Fig. 3b,c; Z33-186R). On the contrary, the infectivity of VSV pseudotyped with chimeric RGP (RGP 33-186 swapped into ZGP), and the binding capacity of MGL/ CD301 to chimeric RGP were dramatically increased compared to RGP (Fig. 3a-c; R33-186Z). These results suggest that the residues 33-186 of GP have a pivotal role in determining the differential infectivity and the differential binding to MGL/CD301 between ZEBOV and REBOV. To further investigate this issue, we focused on residues 33-50, because among residues 33-186 the homology of residues 33-50 is relatively low between ZGP and RGP. By swapping residues 33-50 of ZGP into RGP, both the infectivity to K562-MGL/CD301 cells and the binding capacity to MGL/CD301 were dramatically reduced to the level of RGP (Fig. 3a-c; Z33-50R). As we expected, the infectivity of VSV pseudotyped with chimeric GP containing swapped residues 33-50 from RGP to ZGP was significantly greater than that of VSV pseudotyped with RGP ( Fig. 3a; R33-50Z). Similarly, the binding capacity of MGL/CD301 to the chimeric GPs was dramatically increased compared with RGP ( Fig. 3b,c; R33-50Z). These results strongly suggest that the amino acid residues 33-50 near the N-terminus of GP are responsible for the biosynthesis of the GP structure that determines the interactions with MGL/CD301 and, in turn, viral infectivity in K562-MGL/CD301 cells.

Binding of EBOV GPs to MGL/CD301 is mediated by O-glycans, while N-glycans in RGP interfere with the interaction between MGL/CD301 and O-glycans. To examine whether
N-glycans or O-glycans of RGP and ZGP determine the interaction with MGL/CD301, GPs were treated with Peptide:N-glycanase F (PNGase F), and the binding of MGL/CD301 was examined. Removal of N-glycans was The infectivity of VSV pseudotyped with the indicated GPs was determined using K562-MGL/CD301 cells as described in the Material and methods. Data shown are means ± SD of triplicate experiments. For statistical analysis, Student's t-test was performed. *p < 0.05, **p < 0.01 (b,c) Lysates of VLPs with the indicated GPs were subjected to immunoblotting using MGL/CD301 (upper panels) or mAb 42/3.7 (lower panels). Quantification results of immunoblotting are shown in (c) (n = 3). In this quantification, peak areas of each band were measured and the relative ratio of MGL/CD301 to mAb 42/3.7 was calculated. The relative ratio of ZGP was set to 100%. (d,e) VLPs with ZGP or RGP were treated or untreated with PNGase F. Lysates of treated VLPs were electrophoretically separated and subjected to immunoblotting using MGL/CD301, PHA-L 4 or mAb 42/3.7. Quantification results of immunoblotting are shown in (e) (n = 4). In this quantification, peak areas of each band were measured and the relative ratio of MGL/CD301 to mAb 42/3.7 was calculated. The relative ratio of ZGP/RGP without PNGase F treatment was set to 100%. The equations used for this calculation were as follow; Relative ratio of MGL/CD301 to mAb 42/3.7 of ZGP = ZGP with PNGase F treatment/ZGP without PNGase F treatment x 100 (%), Relative ratio of MGL/ CD301 to mAb 42/3.7 of RGP = RGP with PNGase F treatment/RGP without PNGase F treatment x 100 (%). For confirmed by the loss of PHA-L 4 binding and by an alteration in electrophoretic migration distance. As shown in Fig. 3d,e, the binding of MGL/CD301 to ZGP did not change after PNGase F treatment. In contrast, the binding of MGL/CD301 to RGP increased after PNGase F treatment. These results suggest that the binding of MGL/ CD301 to GPs was mediated by O-glycans not by N-glycans. Additionally, the binding of MGL/CD301 to ZGP/ RGP became almost identical after PNGase F treatment, suggesting that N-glycans in RGP interfered with the interaction between MGL/CD301 and RGP. RGP has more extended N-glycans than ZGP, and the extended N-glycans seem to inhibit the interaction between GP and MGL/CD301. To compare the structural features of N-glycans present on RGP and ZGP, the structures of N-glycans released from RGP, ZGP, and chimeric GPs were analyzed. GPs were immunoprecipitated from dissolved VLPs using mAb 42/3.7. Immunoprecipitated GPs were subjected to in-gel PNGase F digestion followed by glycoblotting 25 and MALDI-TOF MS. Typical MS spectra of ZGP and RGP are shown in Fig. 4a (see Fig. S2 for other GPs), and the structural features estimated from the mass numbers and TOF/TOF analysis are shown in Fig. 4a (only a part of the complex type N-glycans are shown in symbols). The complete estimated compositions of N-glycans observed in MS analyses are shown in Table 1. The structures of the detected N-glycans were similar between RGP and ZGP with about 90% of them being complex-type N-glycans (Fig. 4b). In this calculation, we analyzed the MS peak area of all detected N-glycans (Table 1) and defined their sum as 100%. Based on the results of our MS analyses, we then focused on complex-type N-glycans on GPs. Comparison of the extension pattern of complex-type N-glycans between RGP and ZGP revealed that there were more extended complex-type N-glycans on RGP than ZGP (Fig. 4c). At the same time, the amount of non-extended complex-type N-glycans (agalacto-type glycans) in RGP decreased (Fig. 4c). For this comparison, we defined glycan 10, 14, 16, 22, 25, and 31 in Table 1 as non-extended complex-type N-glycans (i.e. agalacto-type N-glycans; N-glycans, which have only N-acetylglucosamine (GlcNAc) residues at the non-reducing terminal) and we difined galactosylated and sialylated N-glycans (such as glycan 20, 27, 44, 55 in Table 1 in case of bi-antennary N-glycans) as extended complex-type N-glycans (i.e. extended-type N-glycans). When N-glycans are maturated/elongated, their structures are known to become bulky and the steric hindrance around O-glycans are likely to increase. Therefore, we hypothesized that the steric hindrance of extended N-glycans of RGP inhibit the interaction between GP and MGL/CD301, and that N-terminal amino acid residues 33-50 or 33-186 of GP regulate the extension of N-glycans.

RGP binding to MGL/CD301 increases by trimming of N-glycans. To check this hypothesis, VLPs
bearing RGP were treated with neuraminidase and β-galactosidase, and the effect of N-glycan trimming on the interactions of GPs with MGL/CD301 was examined. The treated samples were separated by SDS-PAGE and bindings of mAb 42/3.7, Maackia amurensis leucoagglutinin (MAL), Ricinus communis agglutinin-120 (RCA-120), and recombinant MGL/CD301 were examined as shown in Fig. 5a. Bindings of mAb 42/3.7 revealed that the electrophoretic migration distance altered. Bindings of both MAL and RCA-120 completely diminished after treatment of RGP-bearing VLPs with these glycosidases, confirming the successful removal of sialic acid and galactose residues. The binding of MGL/CD301 to RGP increased after enzymatic treatments (Fig. 5a,b). These results strongly suggest that extension of N-glycans on RGP is responsible for the reduced interaction with MGL/ CD301 and the lower infectivity of VSV pseudotyped with RGP compared to VSV pseudotyped with ZGP.

N-terminal short peptide sequence regulates the extension pattern of N-glycosylation on GPs.
To further confirm that the differential extention of N-glycans between RGP and ZGP was regulated by N-terminal amino acid residues 33-50 or 33-186 of GP, we analyzed the glycan extension pattern of chimeric GPs (Z33-50R, Z33-186R, R33-50Z, R33-186Z). As shown in Fig. 4c, the relative ratio of extended N-glycans increased by swapping the N-terminal residues 33-50 or 33-186 from ZGP into RGP (Z33-50R, Z33-186R). In contrast, the swapping of the N-terminal residues 33-50 or 33-186 from RGP into ZGP (R33-50Z, R33-186Z) resulted in the reduction of the relative ratio of extended-type N-glycans. These results strongly suggest that the short peptide sequence near the N-terminus of the GPs regulates the extension of N-glycans.

Discussion
Glycan-lectin interactions play crucial roles in virus-host interactions, yet many of the underlying mechanisms have not been well-defined. In the present study, we focused on the interaction between glycans present on EBOV GPs and a C-type lectin, MGL/CD301 on the host cells, and aimed to answer the question why REBOV is less pathogenic than ZEBOV. We demonstrated that VSV pseudotyped with EBOV GP infected into DCs mediated by MGL/CD301 and that the N-terminal 33-50 amino acid sequence of GP determines the differential infectivity between VSV pseudotyped with RGP and ZGP. This differential infectivity seems to depend on the extension patterns of N-glycans on GPs, which in turn is regulated by the N-terminal 33-50 amino acid sequence of GP itself.
Other host cell molecules, including folate receptor-α 26 , Tyro3 family molecules 27 , T-cell immunoglobulin and mucin domain 1 28 , and other C-type lectins 12-18 expressed on the initial target cells of filovirus (e.g. monocyte/ macrophage, DCs, hepatocytes, sinusoidal endothelial cells) are reported to be involved in the initiation of infection. Those C-Type lectins, which were previously reported, have diverse carbohydrate specificities, for instance, DC-SIGN preferentially binds to mannose (Man). It is likely that C-type lectins with various carbohydrate specificities [29][30][31][32] act in concert to facilitate the attachment and the entry of EBOV into DCs. MGL/CD301 seems to play predominant roles among these lectins and non-lectin receptors, because the MGL/CD301 blocking monoclonal antibody blocked the infection of pseudotyped viruses into human immature DCs by more than 60% (Fig. 1a).
The filovirus envelope GP, which is the only surface molecule present on viral particles, is known to be involved in virus entry into cells 2,5-11 . Manicassamy and co-workers reported that the N-terminal 150 amino acid residues of GP1 were important for the maintenance of protein integrity, the formation of a pseudotyped virus, and the  viral entry into HEK293T cells, based on their single amino acid mutagenesis analysis 33 . They also claimed that several key amino-acid residues are crucial to the protein folding 33 . However, the importance of glycosylation of GP1 was not explored in their study. Marzi and co-workers reported that the signal sequence of GP (N-terminal 1-32 amino acid residues) affected the GP glycosylation that resulted in an enhancement of DC-SIGN/R-mediated viral infection 34 . They showed that high-mannose type glycosylation of SGP increased when the signal peptide of SGP was replaced with that of ZGP, though they did not study the glycan structures attached to GPs. Our structural analyses by MS revealed that N-glycans of RGP are more extended than those of ZGP, even though these GPs were prepared in the same cell line (Fig. 4c). The extension patterns apparently affect the interaction with MGL/CD301and the binding of MGL/CD301 to RGP increased to a level to that to ZGP after the removal of N-glycans (Fig. 3d,e). The binding of MGL/CD301 to RGP was also increased when their extended glycans were trimmed using glycosidases (Fig. 5b). These results support that the structures of N-glycans of GPs affect the interaction between MGL/CD301 and GPs. More extended N-glycans of RGP seem to lower the binding to MGL/CD301, which preferentially binds O-glycans, and to decrease the capacity of GPs to mediate the infection of DCs and macrophages. N-glycans of ZGP do not seem to have such capacity (Fig. 3d,e).
Using a molecular genetic approach, we demonstrated that limited numbers of amino acid residues of GPs were critical for the differential lectin binding between RGP and ZGP. As far as we examined, a stretch of 18 amino acid residues near the amino terminus of the GP (33-50) seemed to play a key role. We also showed that the same stretch of amino acid residues plays an important role in the regulation of the differential N-glycan extension between RGP and ZGP. The lectin-binding potential and glycan extension pattern of ZGP and RGP were reversed when the N-terminal 33-50 (or 33-186) amino acid residues of the GPs were swapped (Figs 3a, 4d). These results strongly suggest that there is an oligopeptide motif that can control N-glycan extension. How this extension is controlled, however, remains to be determined. It is also worthwhile to mention that, in contrast to a previous report 34 , the signal sequence (1-32 amino acid residues) of GP did not seem to be involved in this phenomenon. We did not observe any effect of the signal sequence on the differential infectivity of pseudotyped viruses nor MGL/CD301 binding between REBOV and ZEBOV GPs (Fig. 2, ZGP vs Z33-462R, or Z1-462R vs Z33-462R).
Regarding the potential mechanism that controls N-glycan extension, we assumed that intracellular localization, trafficking speed, or trafficking pathway of the GP might be influenced by this peptide motif. However, as far as we examined, the biosynthetic rate and the intracellular localization of GP did not seem to be affected. Pulse-chase experiments on 35 S-Methionine incorporation revealed almost identical biosynthetic rates between RGP and ZGP (Fig. S3a). The intracellular localizations of RGP and ZGP were very similar (Fig. S3b). We attempted to prepare VSV pseutotyped with RGP/ZGP lacking amino acid residues 33-50 (RGPΔ33-50, ZGPΔ33-50) and VLPs bearing RGPΔ33-50/ZGPΔ33-50 to confirm the importance of this oligopeptide motif. The expression of those GPs on VSV pseudotyped viruses or VLPs were confirmed by immunoblotting using mAb 42/3.7 (Fig. S6a,b), however, infectious viruses were not obtained (Fig. 6c) for unknown reasons. The band patterns of RGPΔ33-50 and ZGPΔ33-50 from VLPs were different from those of other GPs (Fig. S6b) suggesting that the amino acid residue 33rd through 50th of GP affect some processing of GPs. Further studies are needed to elucidate the mechanism by which this peptide sequence regulates N-glycan extension.
Collar and co-workers recently reported results of structural analyses of N-and O-glycans of recombinant EBOV GPs 35 . They concluded by comparing Zaire ebolavirus, Sudan ebolavirus, Bundibugyo ebolavirus, Tai Forest Figure 5. Binding of MGL/CD301 to RGP increased after glycosidases treatment. (a) Lysates of VLPs with RGP or ZGP were treated or untreated with sialidase and galactosidase, electrophoretically separated and subjected to immunoblotting with mAb 42/3.7, MGL/CD301, RCA-120, or MAL. The removal of sialic acid (NeuAc) and galactose by the enzymatic treatment was confirmed by MAL and RCA-120, respectively. Shown are cropped images from representative results (n = 4) and whole blot images are shown in Fig. S5. (b) The binding capacity of glysosidase-treated RGP/ZGP to MGL/CD301 was compared to that of untreated RGP/ZGP. Relative intensity of MGL/CD301 bindings to mAb 42/3.7 is shown (n = 4). In this quantification, relative intensity of untreated ZGP was set to 100%. For statistical analysis, Student's t-test was performed. *p < 0.05, n.s.: not significant.
SCIenTIFIC REPORTS | (2018) 8:5495 | DOI:10.1038/s41598-018-23357-8 ebolavirus that N-glycan compositions were similar but O-glycan structures varied. However, in contrast to our study, they analyzed recombinant GPs from the lysates of 293 T cells expressing GPs, while our study was on GPs from VLPs which closely represent viral particles. Furthermore, they did not analyze RGP which represent the molecular characteristics of a poorly infectious variant. Our results on the N-glycan structures, however, were consistent with their results in that most of the N-glycans were fucosylated and asialo-type glycans (Fig. S4) 35 . According to their results, ZGP and SGP contained lower amounts of O-glycans with sialic acid than GPs of Bundibugyo/Ivory Coast. Further studies are necessary to prove or disprove the hypothesis that decreased sialylations contribute to elevated interactions of these GPs with MGL/CD301. Previous reports showed that GP was useful as a therapeutic target for EBOV infection. GP-specific neutralizing antibodies protect experimental animals from lethal EBOV infection. Some antibodies were able to protect mice completely and guinea pigs partially 8,10 . Another neutralizing antibody was effective for guinea pigs but not for non-human primates 36,37 . Recently, several neutralizing antibodies that can protect macaques from lethal EBOV infection were also reported [38][39][40] . As shown in the present report, by enhancing N-glycan extension in cells infected with EBOV, infectivity of the virus produced by these cells may be reduced. Furthermore, structural analysis of GPs with extended N-glycans might assist to identify an epitope that could be used as a target of an effective therapeutic antibody against EBOV. Blockade of the infection should also be achieved by interfering with the interaction between GPs and MGL/CD301. The present study will also contribute to understand novel regulatory mechanisms of glycosylation in cells during infection.

Materials and Methods
Construction of chimeric GPs. Standard molecular biology technique was used to construct chimeric GPs containing pCAGGS vectors. The sequence of ZGP and RGP were obtained from MP1153-ZGP vector and pBluescript-RGP vetor respectively, and primers as described in Supplementary Table 1 were used.
Generation of recombinant VSVΔG*-pseudotyped virus. VSVΔG*-pseudotyped viruses were generated as described previously with some modifications 5 . In the VSVΔG* genome, the coding region of the VSV G protein gene was replaced with the coding region of a modified version of the GFP gene 5 . GPs of ZEBOV (Strain Mayinga), REBOV (Strain Pennsylvania), SEBOV (Strain Boniface), and chimeric GPs were cloned into pCAGGS vector. These plasmids were transfected into human embryonic kidney (HEK293T) cells by using Trans-IT LT1 (Takara). Twenty-four hours after transfection, cells were infected with VSVΔG*-VSVG (VSV pseudotyped with VSVG) at a multiplicity of infection (MOI) of 1 for 1 hour at 37 °C. Cells were then washed twice with Dulbecco's modified Eagle Medium-High Glucose (DMEM-HG) without fetal calf serum (FCS) and once with DMEM-HG with 10% FCS, and then medium was added. After 24 hours of incubation at 37 °C in a 5% CO 2 atmosphere, the culture fluid was collected and centrifuged to remove cells. Each virus stock was stored at −80 °C until further use. Regarding to the biosafety level (BSL), the pseudotyped virus was categorized to BSL2.

Production of Virus-like particles (VLPs).
VLPs were produced as reported previously with some modification 19,20 . Briefly, HEK293T cells (4 × 10 6 ) were seeded on a 10 cm culture dish one day before transfection. These cells were transfected with pCAGGS-GP and pCAGGS-VP40 by using Trans-IT LT1. DNA and transfection reagent (40 μL of Trans-IT LT1 with 10 μg of each plasmid) were mixed in 0.5 mL of Opti-MEM (Gibco) and added to the cells. The mixture was incubated at 37 °C for 48 hours. The culture medium was harvested 48 hours later and centrifuged at 6,000 rpm for 20 min to remove cellular debris. The supernatant was then layered on a 25% sucrose cushion and ultra-centrifuged at 27,000 rpm for 1.75 hours at 4 °C using a himac CP 70 MX (Hitachi). The supernatant was discarded, and the pellet was resuspended in PBS. The suspension was regarded as VLPs and stored at −20 °C until further use. The protein amount of VLP solution was measured by BCA protein assay kit (Pierce).
Cell culture. HEK293T cells and K562 human chronic myelogenous leukemia cells were grown in DMEM-HG or RPMI 1640 medium, respectively, supplemented with 10% FCS. Peripheral blood mononuclear cells (PBMCs) were isolated from venous blood drawn from healthy volunteers at the Tokyo Metropolitan Red Cross Blood Center (Tokyo, Japan) by centrifugation on a Percoll density gradient. Monocytes were purified from PBMCs by incubation with anti-CD14 mAb-coated microbeads and separation by a magnetic cell separation system (MACS, Miltenyi Biotec, Bergisch Gladbach, Germany) with column type LS. Immature DCs were prepared by culturing monocytes for 7 days in RPMI containing 7.5% FCS with 500 U/ml GM-CSF (Kyowa-Kirin) and 100 ng/mL interleukin-4 (GT #2204). To examine the purity of immature DCs, flow cytometric analysis with fluorescein isothiocyanate (FITC)-labeled anti-human CD1a (Biosource), CD14 (Dako Cytomation), CD83 (eBioscience), and HLA-DR (BD Biosciences Pharmingen) was performed.
Expression of soluble recombinant MGL/CD301. A plasmid encoding the putative extracellular domain of MGL/CD301 was prepared as described previously 32 . E. coli with expressed proteins were recovered, first washed with TBS containing 0.5% TritonX-100 (Sigma) and 10 mM EDTA, and then washed with H 2 O. Washed pellets were solubilized with 2 M NH 4 OH, and soluble recombinant MGL/CD301 was refolded in a mixture of 2 mM reduced glutathione and 0.2 mM oxidized glutathione. The refolded protein was dialyzed against 20 mM MOPS buffer (pH 7.0) containing 20 mM CaCl 2 , 0.5 M NaCl, and 0.02% NaN 3 at 4 °C. Soluble rMGL/ CD301 was purified by affinity chromatography on a galactose-Sepharose 4B column. by using an anti-MGL/CD301 monoclonal antibody (mAb) MLD-1. Cloning of transfectants was performed by the limiting dilution method. Flow cytometric analysis with mAb MLD-1 (1:400-diluted ascites fluid) and FITC-labeled anti-mouse immunoglobulin G (Zymed, 5 μg/mL) was performed to determine the levels of MGL/ CD301 expression on the obtained cell clones.
Titration of VSVΔG*-pseudotyped virus. Vero E6 cells (5 × 10 4 cells/well) were seeded on 96-well plate one day before the infection. Serially diluted viral solutions were made after treatment with anti-VSVG antibody at room temperature for 30 min to neutralize VSV pseudotyped with VSVG in the viral solution, and diluted viral solution was added 50 μL/well to seeded cells. Cells were incubated at 37 °C for 16 h, then the number of infected cells were analyzed by flow cytometory. In our infection experiment, virus titer was defined as below. SDS-PAGE and immunoblotting analysis. VSV pseudotyped with GPs or VLPs bearing GP were solubilized. Viral GPs were separated by SDS-PAGE and blotted onto PVDF membrane. The membrane was blocked with 3% BSA in PBS at 4 °C overnight, and incubated with anti-EBOV GP1 mAb 42/3.7 (1:1000 dilution), followed by HRP goat anti-mouse IgG(H + L) (ZyMed) to confirm equal loading. Biotinylated lectins (2.5 μg/ml in TBS containing Ca 2+ ) were reacted for 2 hours at room temperature, and then ZyMax TM streptavidin alkaline phosphatase conjugate (Zymed, 1:1000 dilution) was used as the second reagent for 1 hour at room temperature. After washing, the membranes were developed using ECL TM Western blotting detection reagent (Amersham) for HRP or alkaline phosphatase substrate Kit IV (Vector) for alkaline phosphatase. Band intensities of the MGL/CD301 binding and the anti-EBOV GP1 mAb 42/3.7 binding to the GPs was quantified with Image J or ImageQuant TL (GE Healthcare) software. In the measurement of MGL/CD301 binding to GPs, the amounts of GPs were normalized as stated below. First, the loading amounts of viral lysates or VLPs were normalized to the total protein amount measured using BCA protein assays. Western blotting analysis was performed using mAb 42/3.7. Bands intensities of GPs were quantified by ImageJ or ImageQuant TL software, and loading volume of viral lysate or VLPs were re-normalized to load the estimated equal amount of GPs.
Structural estimation of N-glycans. Immunoprecipitated GPs from solubilized VLPs with GPs were separated on an 8% SDS-PAGE gel. The gel was then fixed and stained with CBB (BioRad). The stained portions of GPs in the gel were excised and subjected to in-gel tryptic digestion. Digestion was carried out for 16 h with MS-grade trypsin (Wako). Peptides were extracted from the gel with 0.1% TFA (containing 50% acetonitrile (ACN)). Peptides were dried by speed vac and dissolved in H 2 O. Dissolved peptide samples were treated with PNGase F (Roche) to release N-glycans. Released N-glycans were purified and labeled using the Glycoblotting method (Sumitomo Bakelite). Glycoblotting was performed according to a previously described procedure 25 . Labeled N-glycans were analyzed by MALDI-TOF MS and the TOF/TOF method using UltraFLEX II (Bruker).
Glycosidase treatment. VLPs bearing ZGP or RGP were treated with 100 mU of Neuraminidase [Clostridium perfringens] (Sigma) at 37 °C for 2 hours. Then, 3 mU of β-galactosidase [Streptococcus] (Seikagaku Co.) and MnCl 2 (final concentration: 10 mM) were added and incubated at 37 °C for 24 hours. Glycosidase-treated VLPs were subjected to immunoblotting to confirm the removal of NeuAc and Gal as well as the changes of MGL/ CD301 binding before and after enzyme treatment.
Statistical analysis. Student's paired t-test was used to determine the statistical significance of the differences. A p-value of less than 0.05 was considered statistically significant.