A novel mechanism for LSECtin binding to Ebola virus surface glycoprotein through truncated glycans

Abstract

LSECtin is a member of the C-type lectin family of glycan-binding receptors that is expressed on sinusoidal endothelial cells of the liver and lymph nodes. In order to compare the sugar- and pathogen-binding properties of LSECtin with those of related but more extensively characterized receptors, such as DC-SIGN, a soluble fragment of LSECtin consisting of the C-terminal carbohydrate-recognition domain has been expressed in bacteria. A biotin-tagged version of the protein was also generated and complexed with streptavidin to create tetramers. These forms of the carbohydrate-recognition domain were used to probe a glycan array and to characterize binding to oligosaccharide and glycoprotein ligands. LSECtin binds with high selectivity to glycoproteins terminating in GlcNAcβ1-2Man. The inhibition constant for this disaccharide is 3.5 μM, making it one of the best low molecular weight ligands known for any C-type lectin. As a result of the selective binding of this disaccharide unit, the receptor recognizes glycoproteins with truncated complex and hybrid N-linked glycans on glycoproteins. Glycan analysis of the surface glycoprotein of Ebola virus reveals the presence of such truncated glycans, explaining the ability of LSECtin to facilitate infection by Ebola virus. High mannose glycans are also present on the viral glycoprotein, which explains why DC-SIGN also binds to this virus. Thus, multiple receptors interact with surface glycoproteins of enveloped viruses that bear different types of relatively poorly processed glycans.

Mammalian glycan-binding receptors interact with endogenous glycans on cells and secreted glycoproteins and with sugar structures on the surfaces of pathogenic micro-organisms (1,2). Endocytosis mediated by sugar-specific receptors results in clearance of glycoproteins from circulation, so these receptors can participate in normal turnover of serum glycoproteins and in the scavenging of glycoproteins released as a result of damage to tissues (3,4). Glycan-binding receptors on cells of the immune system bind to glycans on the outer surfaces of viruses, bacteria, fungi or parasites and direct uptake of the pathogens, leading to their destruction and facilitating presentation of pathogen fragments to the immune system (5-7). Glycan-binding receptors are particularly common on the surfaces of cells in the immune system, such as macrophages and dendritic cells, and in the liver (8). In both cases, the distinct but overlapping roles of the different receptors remain poorly defined.

LSECtin is a recently described member of the C-type family of receptors that is found predominantly on sinusoidal endothelial cells of the liver and lymph nodes, suggesting that it plays a specific role in these cells (9). LSECtin has also been reported to be expressed on peripheral blood and thymic dendritic cells (10). It is most closely related in sequence to the receptors DC-SIGN¹ and DC-SIGNR (also known as L-SIGN) and shares with these receptors the ability to bind mannose, N-acetylglucosamine and related sugars. Both scavenging and pathogen-recognition functions for LSECtin can be envisioned, but its ligand-binding properties and its interactions with pathogens have not been extensively studied.

DC-SIGN and DC-SIGNR have been implicated in enhanced infection by enveloped viruses. For example, DC-SIGN on dendritic cells has been shown to present human immunodeficiency virus to T lymphocytes, dramatically increasing the efficiency of infection (11,12). In an analogous manner, DC-SIGN and DC-SIGNR on liver sinusoidal endothelial cells enhance infection of hepatocytes by hepatitis C virus in culture (13). LSECtin shares with these receptors the ability to bind to surface glycoproteins of enveloped viruses. Interaction of LSECtin with the surface glycoproteins of severe acute respiratory syndrome (SARS) coronavirus and Ebola virus has recently been described and LSECtin-mediated infection of cells Ebola virus has been demonstrated (14). DC-SIGNR, which shows a similar cellular distribution to LSECtin, also binds to Ebola virus, so there is potential overlap in their functions (15-17). However, there are also clear differences in the way that these receptors interact with viral glycoproteins, because LSECtin does not interact with lentiviral particles harboring the surface proteins of human immunodeficiency virus or hepatitis C virus, while DC-SIGN and DC-SIGNR do (10).

In order to clarify the differential interactions of LSECtin and the other, better characterized receptors, the binding specificity of LSECtin has been investigated by probing a glycan array and by employing other assays for sugar-binding activity. It is demonstrated that the most effective glycan ligands for LSECtin contain the disaccharide sequence GlcNAcβ1-2Man. Analysis of the Ebola virus surface glycoprotein shows the presence of both this target structure and high mannose oligosaccharides, which would account for the ability of LSECtin as well as DC-SIGN and DC-SIGNR to bind to the viral glycoprotein.

Experimental Procedures

Expression of the carbohydrate-recognition domain from LSECtin

A cDNA for LSECtin was isolated by polymerase chain reaction amplification from a human liver cDNA library from BD Biosciences (Oxford, UK). Synthetic oligonucleotides from Invitrogen (Paisley, UK) were combined with restriction fragments from the cDNA to insert the region encoding the carbohydrate-recognition domain (CRD) into a modified version of the vector ompA2 (18). The codons for the bacterial ompA signal sequence were followed immediately by the cDNA sequence beginning at residue 160 of LSECtin, which encodes the sequence AsnSerCys at the N-terminal end of the CRD. For expression of biotin-tagged protein, the C-terminal end of the cDNA was similarly modified by insertion of synthetic oligonucleotides so that the expression vector codes for a protein in which the C-terminal cysteine residue of LSECtin is followed by the sequence GlyLeuAsnAspIlePheGluAlaGlnLysIleGluTrpHisGlu (19).

The CRD of LSECtin was expressed in Escherichia coli strain JA221 grown in Luria-Bertani broth containing 50 μg/ml ampicillin. For addition of biotin, the cells were also transformed with plasmid birA, which encodes the gene for biotin ligase (20), and chloramphenicol at 20 μg/ml was included in the medium. For protein production, a 200 ml starter culture was diluted to 6 l and grown at 25 °C to OD₅₅₀ of 0.8, at which point additions were made to achieve final concentrations of 10 μg/l isopropyl-β-D-thiogalactoside, 12.5 μg/ml biotin, and 100 mM CaCl₂. After further growth overnight at 25 °C, bacteria were harvested by centrifugation for 15 min at 4,000 ×g. The pelleted cells were suspended in 150 ml of loading buffer (150 mM NaCl, 25 mM CaCl₂, 25 mM Tris-Cl, pH 7.8) and sonicated at full power with a Branson 250 sonicator five times for two minutes, with 30 minute periods of cooling on ice between sonication steps. Following centrifugation for 10 min at 10,000 ×g and 30 min at 150,000 ×g, the supernatant was filtered through glass wool and applied to a 10-ml column of mannose- or fucose-Sepharose (21). The column was washed with 16 ml of loading buffer and eluted in 2-ml fractions with elution buffer (150 mM NaCl, 2.5 mM EDTA, 25 mM Tris-Cl, pH 7.8).

Formation of streptavidin complexes

Streptavidin obtained from Sigma Chemical Co. (Poole, Dorset, UK) was iodinated by the chloramine T method (22) using Na¹²⁵I from Perkin-Elmer. Alexa 488-labeled streptavidin was purchased from the Molecular Probes division of Invitrogen. Approximately 250 μg of biotin-tagged CRD in 6 ml of eluting buffer was adjusted to 25 mM CaCl₂ and incubated overnight at 4 °C with 150 μg of labeled or unlabelled streptavidin. The complex was re-applied to a 1-ml column of fucose-Sepharose, which was rinsed with 5 ml of loading buffer and eluted with 0.5 ml aliquots of eluting buffer.

Generation of polyclonal antibody

A sample of 1 mg of purified CRD from LSECtin, without a biotin tag, was provided to Eurogentec (Seraing, Belgium) for immunization of rabbits following their standard protocol.

Expression of LSECtin in fibroblasts

The full length cDNA for LSECtin was inserted in to a retroviral expression vector that was used to generate pseudovirus that were in turn used to infect Rat-6 fibroblasts following exactly the protocols previously described for other glycan-binding receptors (23,24). Cells were harvested from two 225-cm² flasks by scraping in phosphate-buffered saline and were pelleted by spinning at 2000 ×g for 2 min. The pellet was suspended in 5 ml of eluting buffer containing 0.5 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride, 150 nM aprotinin; 1 μM E-64 protease inhibitor and 1 μM leupeptin from CalBiochem (Nottingham, UK) and sonicated for 5 sec at low power. Additions were made to bring the suspension to final concentrations of 1% Triton X-100 and 25 mM CaCl₂, followed by a further 5 sec sonication and incubation for 15 min on ice. The sample was centrifuged at 150,000 ×g for 20 min and the supernatant was applied to a 1-ml column of fucose-Sepharose, which was rinsed with 4 ml of loading buffer containing 0.1% Triton X-100 and eluted with 0.5 ml aliquots of eluting buffer containing 0.1% Triton X-100. Fractions were precipitated by addition of half a volume of 30% trichloroacetic acid and incubation for 10 min on ice. Precipitated protein was collected by centrifugation for 5 min at 18,000 ×g and the pellet was washed twice with 0.5 ml of ethanol:ether (1:1) and dried for 5 min under vacuum. The samples were dissolved directly in sample buffer and run on SDS-polyacrylamide gels that were blotted onto nitrocellulose (25). Blots were blocked for 30 min with 5% bovine serum albumin (BSA) in Tris-buffered saline, incubated for 90 min with primary antiserum at 1:500 dilution in the same buffer, washed with Tris-buffered saline, incubated with 20 μg/ml alkaline phosphatase-conjugated protein A from CalBiochem in Tris-buffered saline containing 5% BSA and finally washed again with Tris-buffered saline before incubation with 5-bromo-4-chloro-3-indolyl-phosphate/nitroblue tetrazolium phosphatase substrate from CalBiochem.

Screening for glycan ligands

The glycan array of the Consortium for Functional Glycomics was screened in buffer containing 150 mM NaCl, 2 mM CaCl₂, 2 mM MgCl₂, 20 mM Tris-Cl pH 7.4, 0.05% Tween 20 and 1% BSA. Glycoprotein blotting was performed as previously described (26). For competition binding assays (27) and pH-dependence assays (28), polystyrene wells were coated with CRD from LSECtin at a concentration of 50 μg/ml. Oligosaccharides were obtained from Dextra Laboratories (Reading, UK).

Other analytical procedures

SDS-polyacrylamide gels were performed by the method of Laemmli (29). Protein concentrations were determined by the method of Bradford (30).

Modeling of LSECtin binding site

The structure of rat mannose-binding protein A in complex with a mannose-containing ligand, Protein Database entry 2msb, was modified by changing side chains to their LSECtin equivalents using Insight II software from Accelrys (Cambridge, UK). The GlcNAcβ1-2Man disaccharide was abstracted from chain A in entry 1fc1 and the mannose residue was superimposed on the mannose in the primary binding site of mannose-binding protein. The model figure was created using Molscript (31).

Analysis of LSECtin binding to cells expressing Ebola glycoprotein

To assess binding of soluble LSECtin to Ebola glycoprotein on the cell surface, 293T cells were transiently transfected with the expression plasmid pcDNA3.1 encoding full length Ebola glycoprotein (32) or control vector without a cDNA insert. At 48 h post transfection, the cells were harvested, washed, resuspended in ice-cold phosphate-buffered saline containing 3% fetal calf serum and 0.1% NaN₃ and incubated with the indicated amounts of soluble, tetrameric Alexa 488-labelled LSECtin in the presence or absence of EDTA. Thereafter, the cells were again washed and analyzed using a FACSCalibur flow cytometer (Becton Dickinson).

Production of soluble Ebola virus glycoprotein

Soluble Ebola virus glycoprotein was produced by transient transfection of 293T cells with plasmid pAB61-ZEBOV-GP-Fc, which encodes the GP1 subunit of Ebola glycoprotein fused to the Fc portion of human IgG, as described previously (13). The supernatant of the transfected cells was collected and passed through a HiTrap protein A column from GE Healthcare (Munich, Germany). Bound protein was eluted with 0.1 M glycine, pH 3.0, mixed with 1 M Tris-Cl, pH 9.0, and stored at −80°C. The purity of the eluted samples was analyzed by SDS-gel electrophoresis and Coomassie blue staining.

Solid phase assay of LSECtin binding to Ebola glycoprotein

Streptavidin-coated wells from Perbio Sciences (Tattenhall, Cheshire, UK) were washed with loading buffer and incubated overnight with 100 μg/ml of biotin-tagged CRD in loading buffer containing 0.1% BSA. Wells were washed twice with loading buffer and incubated 6 h at room temperature with Ebola glycoprotein-Fc fusion protein at various dilutions. Following three washes with loading buffer, wells were incubated for 30 min at room temperature with 0.1 μg/ml alkaline phosphatase-conjugated protein A in loading buffer containing 0.1% BSA. After five further washes with loading buffer, wells were incubated for 15 min at room temperature with 100 μl of a 1 mg/ml solution of p-nitrophenylphosphate in 0.1 M glycine, pH 10.4, containing 1 mM MgCl₂ and 1 mM ZnCl₂. The reaction was stopped by addition of 50 μl of 1.5 M NaOH and the absorbance at 405 nm was read on a Wallac 1420 plate reader.

Analysis of glycans on Ebola glycoprotein

Approximately 100 μg of Ebola glycoprotein-Fc fusion protein was precipitated with trichloroacetic acid as described above. The pellet was rehydrated in 50 mM ammonium bicarbonate, pH 8.4 and digested at 37 °C overnight by the addition of tosylphenylchlormethyl ketone-trypsin at 1 mg/ml. N- and O-linked glycans were released and purified as described previously (33). Permethylation and sample repurification were performed using the sodium hydroxide protocol. Samples were redissolved in 10 μl of methanol:ultrapure water (4:1) and mixed in equal ratio with matrix (10 mg/ml 2,5-dihydroxybenzoic acid). Both MALDI mass spectrometry and tandem mass spectrometry were performed on an Applied Biosystems 4800 MALDI TOF/TOF mass spectrometer.

Results and Discussion

Molecular structure of LSECtin

The sequence of the extracellular domain of LSECtin suggests that this portion of the molecule is divided into an extended neck region and a C-terminal carbohydrate-recognition domain (CRD) (Fig. 1). The importance of the C-terminal domain in binding to carbohydrate was confirmed by expression of this portion of the protein in a bacterial system. Based on the arrangement of disulfide bonds in other C-type CRDs, the N-terminal boundary of the expressed fragment was placed just before the first cysteine common to these domains (www.imperial.ac.uk/research/animallectins). In order to enhance folding and disulfide bond formation, the polypeptide was fused to a bacterial signal sequence so that it would be directed to the periplasm. Following expression in the presence of Ca²⁺, the protein was released from the bacteria by sonication. Based on previous studies indicating that the best monosaccharide ligand for LSECtin is mannose (9), the CRD was purified by affinity chromatography on mannose-Sepharose (Fig. 2). In addition to protein that was retained on the column in the presence of Ca²⁺ and released with EDTA, there was evidence that some of the protein was retarded on the affinity resin but leached off during the wash. This behavior, reflecting relatively weak binding, is characteristic of other C-type CRDs expressed in isolation (23). Similar results were obtained using a fucose-Sepharose affinity column (Fig. 2), consistent with the previous demonstration that fucose binds to LSECtin with near the same affinity as mannose (9). The need for an extended column with very high densities of monosaccharide ligands to retain LSECtin probably explains why binding to mannose-containing columns was not detected in some recent experiments (10).

Fig. 1. Domain organization of LSECtin.

The diagram at the left indicates the type II transmembrane orientation, the locations of N-linked glycans, and the cysteine residues in the neck region that are potentially involved in interchain disulfide bond formation. The C-terminal end of the CRD, to which the biotin tag sequence has been appended, is also labeled. The diagram at the right shows the arrangement of the domains along the polypeptide, with the residue numbers corresponding to the domain boundaries labeled. The overall polypeptide molecular weight is 32,600 kDa. The heptad repeat pattern of hydrophobic amino acids in the neck region is also indicated.

Fig. 2. SDS-polyacrylamide gel electrophoresis of fractions from the affinity purification of biotin-tagged CRD from LSECtin expressed in bacteria.

A and B, elution of 10-ml columns of mannose- and fucose-Sepharose with buffer containing EDTA. Fractions of 2 ml were collected. C, the complex of biotin-tagged CRD with streptavidin was repurified by affinity chromatography on a 1-ml column. Elution fractions of 0.5 ml were collected. In all cases, aliquots of 20 μl were run on 17.5% gels that were stained with Coomassie blue. The expected molecular weight of the expressed CRD is 15,000 kDa.

In order to enhance the affinity of the protein for binding studies, a multivalent form of the CRD was created by appending a biotinylation sequence at the C-terminal end of the CRD. Expression of the CRD in bacteria also expressing the birA gene, which encodes biotin ligase, results in efficient biotinylation of the single lysine residue in the C-terminal extension (20). Complexes containing four CRDs can then be formed by allowing the tagged protein to interact with streptavidin. This approach has successfully been used to enhance the affinity of other C-type CRDs (34). The biotinylated protein was purified in the same way as the untagged CRD. Following incubation with streptavidin, the complex was efficiently repurified on a smaller affinity column, because of the enhanced affinity resulting from tetramerization (Fig. 2). The biotinylated monomeric and tetrameric forms of the protein were used in subsequent studies of the interaction of LSECtin with target ligands.

Previous studies have suggested that the extracellular domain of LSECtin assembles into forms that are larger than monomers, but a unique oligomeric structure has not been defined. The sequence of the neck region conforms very closely to the heptad repeat pattern of aliphatic hydrophobic amino acid side chains expected for an amphipathic α-helix that would be able to form coiled-coil structures (Fig. 1) (35). This domain also contains three cysteine residues, two of which are positioned at the hydrophobic positions that would be expected to face inwards in such a coiled coil. Although previous studies have suggested that the receptor does exist as disulfide-linked oligomers, these experiments were not performed on material selected for activity and thus the results may have been confounded by the presence of misfolded biosynthetic intermediates with heterogeneous disulfide-bond arrangements (9). In order to examine the possibility that the cysteine residues in the neck link polypeptides in the native receptor, the intact protein was expressed in rat fibroblasts and functional receptor was purified by affinity chromatography. Based on the results with the monomeric and tetrameric forms of the CRDs, it was expected that an oligomeric receptor would bind efficiently to fucose-Sepharose, and blotting with an antibody to the CRD revealed that the protein is retained on the column and can be eluted with EDTA (Fig. 3). This active material was run on gels in the presence and absence of reducing agent, revealing that it exists as a disulfide-linked dimer. Thus, binding results obtained with the streptavidin-complexed CRD can be taken as an indication of the type of affinity enhancement that would be expected for the native oligomeric protein.

Fig. 3. Gel electrophoresis of LSECtin isolated from transfected rat fibroblasts.

A Triton X-100 extract of cells was run on a 1-ml column of fucose-Sepharose and elution fractions of 0.5 ml were collected. Aliquots of 200 μl from the peak fraction (fraction 3) were precipitated with trichloroacetic acid and run on a 10% SDS-polyacrylamide gel that was blotted onto nitrocellulose and probed with antibody against the CRD of LSECtin followed by alkaline phosphatase-conjugated protein A. In addition to the major band, a smaller fragment that probably results either from proteolysis or underglycosylation is observed. A small amount of apparently dimeric material was observed in reduced samples and in unreduced samples a small amount of potentially tetrameric material was present. These bands were not seen in samples that were not concentrated by precipitation with trichloroacetic acid (data not shown).

Binding of LSECtin to truncated complex glycans

The binding specificity of LSECtin was initially investigated using the glycan array developed by the Consortium for Functional Glycomics (36). Fluorescently labeled streptavidin-CRD complex was created by incubating the biotin-tagged CRD with streptavidin that was pre-labeled with Alex-488. The purified complex was then used to probe the array at multiple concentrations (Fig. 4). At low concentrations, highly selective binding to a limited set of glycans was obtained. The common feature of the glycans with the highest signals is the presence of one or more terminal GlcNAcβ1-2Man sequences. However, some binding was also observed for a bi-antennary complex glycan in which both branches bear galactose residues linked β1-4 to the terminal GlcNAc residue and to a core trimannose structure.

Fig. 4. Binding of fluorescently-labeled LSECtin-streptavidin complex to a glycan array.

The ligands giving the highest signals are identified. Blue indicates glycans displaying terminal GlcNAcβ1-2Man, red denotes glycans that bear terminal Fucα1-2Gal and the mannose-terminated structure is shown in green. The complete list of glycans on the array is provided in Supplementary Table 1.

At higher concentrations of labeled probe, binding to an additional set of glycans with terminal α1-2 linked fucose residues was observed. There is some specificity to the interaction, because binding to glycans containing α1-3 and α1-4 linked fucose was not detected. Glycans with fucose linked α1-6 to the core are not present on the array. Although these results suggest that the receptor can bind to a second type of glycan, the binding occurs only at high concentrations of CRD presented in the form of a tetramer, indicating that the interaction would be weak. These results demonstrate the value of probing the glycan array at multiple concentrations of CRD and emphasize that preferential binding to potentially physiologically relevant ligands is revealed at lower concentrations of the labeled probe, while high concentrations may result in binding to low affinity ligands.

The glycan array results were confirmed using a competition assay to compare the binding of oligosaccharides to immobilized CRD from LSECtin. In this assay, which employs ¹²⁵I-Man-BSA as a reporter ligand at concentrations well below the K_D, the observed inhibition constants provide a good guide to the relative affinities of competing low molecular weight ligands. As predicted from the array results, the disaccharide GlcNAcβ1-2Man is the highest affinity ligand (Fig. 5). The branched trimannose oligosaccharide, which forms the core of the complex glycans and thus is also common to many of the ligands identified on the glycan array, binds only 2-fold better than mannose, reflecting the presence of two terminal mannose residues. It is unclear why this trisaccharide, when linked to a chitobiose core, is a ligand in the glycan arrays screened at low receptor concentrations. The presence of two GlcNAcβ1-2Man branches on a single glycan also does not enhance the affinity per terminal residue. The Fucα1-2Gal disaccharide shows only very slightly enhanced affinity compared to a simple monosaccharide, consistent with the conclusion that the affinity for this type of ligand is much lower than for the preferred GlcNAcβ1-2Man disaccharide as suggested by the glycan array results.

Fig. 5. Solid phase binding competition assays.

Immobilized CRD from LSECtin was probed with ¹²⁵I-Man-BSA in the presence of competing ligands. The K_I was calculated as the concentration of inhibitor needed for 50% inhibition of binding of the reporter ligand, based on nonlinear least squares fitting with a simple first order competition equation. A. Comparative results for a sample assay. B. Summary of results for multiple assays. Results for at least three experiments are reported for each ligand, except the two largest glycans for which the assay was performed twice. The mean ± standard deviation is given in each case.

The specificity of the binding of LSECtin is unusual for C-type CRDs. Many receptors that contain C-type CRDs bind to broad categories of glycans that share simple features, such as terminal galactose or mannose residues, although a few receptors, such as the selectins and the scavenger receptor C-type lectin, bind much more restricted sets of glycans. In the latter cases, the best ligands typically contain branched terminal structures, such as the sialylated or unsialylated versions of the Lewis^x and Lewis^a trisaccharides and binding is mediated by primary interactions with one terminal sugar residue and secondary interactions with the other terminal sugar residue. In contrast, LSECtin binds a simple linear disaccharide with micromolar affinity.

The affinity for GlcNAcβ1-2Man is also unusually high for such a simple ligand binding to a C-type CRD (37), so it is interesting to consider how this affinity may arise. The binding characteristics of multiple C-type CRDs have been studied in molecular detail and a common feature of binding is coordination of vicinal 3- and 4-hydroxyl groups to a conserved Ca²⁺ in the primary binding site (5,37). Based on this feature, and the fact that free mannose and GlcNAc have similar affinities for LSECtin (9) there are two ways in which LSECtin might interact with the GlcNAcβ1-2Man ligand, in which the 3- and 4-hydroxyl groups of either the GlcNAc or the mannose residue are coordinated to Ca²⁺ the primary binding site. The fact that ligands in which the GlcNAc residue is substituted with β1-4 linked galactose can still bind, albeit with reduced affinity, strongly suggests that binding is through the mannose residue, because the presence of galactose would mean that there would be no free 4-hydroxyl groups on the substituted GlcNAc residues available for coordination. With mannose in the primary binding site, the enhanced affinity could result from secondary contacts with the GlcNAc residue.

In order to explore the possible secondary interactions of a disaccharide ligand with LSECtin, some of the features of the binding site of LSECtin were modeled. Amongst C-type CRDs for which structure in complex with sugar ligands have been determined, the CRD of LSECtin is most closely related to the CRD of DC-SIGN in overall sequence. However, in the area around the Ca²⁺- and sugar-binding site, the sequence is more like serum mannose-binding protein, so this latter protein was used as a basis for the modeling studies (Fig. 6). When the disaccharide is modeled into the binding site with mannose ligated to Ca²⁺, potential secondary contacts between the GlcNAc residue and the side chain of residue Trp²⁵⁹ and the backbone of residue Gly²⁵³ are observed. The proposed packing of the indole ring of tryptophan against the methyl group of the N-acetyl substituent of GlcNAc is analogous to interactions of tryptophan with N-acetyl substituents in sugar-binding sites in a variety of animal, plant, and viral sugar-binding proteins (37,38). In other C-type CRDs, interactions of aromatic residues with an N-acetyl substituent at the 2 position on a sugar ligated in the primary binding site through hydroxyl groups at the 3 and 4 positions accounts for increased selectivity of this site (39,40), while in LSECtin the interaction would form a component of the secondary binding site. When the GlcNAc residue was placed in the primary binding site, the mannose residue did not appear to make secondary interactions that would account for the enhanced affinity of the disaccharide.

Fig. 6. Modeling of the LSECtin binding site based on the structure of mannose-binding protein (MBP).

The aligned sequences of regions of mannose-binding protein, DC-SIGN and LSECtin containing the critical Ca²⁺-binding site are shown at the bottom. The GlcNAcβ1-2Man disaccharide was positioned based on the arrangement of mannose in the binding site of MBP. For the mannose residue (yellow) in the primary binding site, hydrogens bonds are shown as solid lines and coordination bonds to Ca²⁺ are shown as dotted lines. The Ca²⁺ is shown as a black sphere. Potential interactions of the GlcNAc residue (orange) with backbone and side chain atoms are highlighted with green shading.

Taken together, the results from the glycan array, competition and modeling studies are consistent with identification of terminal GlcNAcβ1-2Man disaccharides as the preferred ligands for LSECtin. The selective binding of LSECtin to such truncated glycans was further confirmed by using biotinylated CRD from LSECtin complexed with ¹²⁵I-labelled streptavidin to probe blots of glycoproteins with different glycan structures (Fig. 7). The results confirm that glycoproteins with exposed terminal GlcNAcβ1-2Man disaccharides are the best ligands. Examples include partially degraded glycoproteins, such as agalacto-orosomucoid, and glycoproteins that naturally bear significant amounts of truncated hybrid structures, such as chicken ovalbumin. Weaker binding to asialoorosomucoid is consistent with the observed weaker binding to the galactose-terminated glycans. Thus, these studies identify glycoproteins with truncated complex N-linked glycans as potential target ligands for LSECtin. Such ligands could arise either from incomplete processing in the biosynthetic pathway or through limited degradation by exoglycosidases.

Fig. 7. Identification of glycoprotein ligands for LSECtin by probing of a glycoprotein blot.

Parallel gels of glycoproteins bearing different terminal glycans were stained with Coomassie blue (A) or blotted onto nitrocellulose and probed with ¹²⁵I-labelled streptavidin-LSECtin complex (B).

Effect of pH on binding of LSECtin to glycan ligands

There is evidence that LSECtin can participate in the uptake of ligands, such as viruses (14). Other receptors on liver sinusoidal endothelial cells, including the mannose receptor, mediate clearance of glycoproteins, particularly those released during tissue damage (4). Although GlcNAcβ1-4Man is not a common terminal sequence on secreted glycoproteins, glycoproteins with truncated glycans might be released from an intracellular compartment following cell lysis and rapid clearance of GlcNAc-terminated glycoproteins from human circulation has been described, although the site of clearance was not identified (41). In the case of related receptors containing C-type CRDs, such as DC-SIGN, the ability to mediate degradation of ligands is correlated with their ability to release ligands at endosomal pH (42). With the availability of the purified ligand-binding domain of the receptor, it was possible to test the effect of pH on ligand binding using the solid phase binding assay (Fig. 8). Surprisingly, binding of the ¹²⁵I-Man-BSA reporter ligand increases rather than decreases below neutral pH. In this respect, it much more closely resembles DC-SIGNR than DC-SIGN (42). It is interesting that both LSECtin and DC-SIGNR are expressed in endothelial cells and it is possible that they participate in specialized forms of endocytosis in these cells rather than the general clathrin-mediated uptake and recycling through acid endosomal compartments.

Fig. 8. pH-Dependence of ligand binding to LSECtin.

Immobilized CRD from LSECtin was probed with ¹²⁵I-Man-BSA at different pHs.

LSECtin can bind to Ebola virus glycoprotein bearing unusual terminal glycans

Based on the finding that the binding specificity of LSECtin is very different to the specificity of DC-SIGN and DC-SIGNR, it was of interest to examine the molecular details of the interaction of LSECtin with target glycoproteins on Ebola virus. There is a single surface glycoprotein on this virus, which is the product of an edited mRNA (43). Ebola glycoprotein is synthesized as a type 1 transmembrane protein of approximately 150 kDa that bears both N- and O-linked glycans and forms trimers on the surfaces of infected cells and virions (44). It is cleaved into two disulfide linked polypeptides, GP1 and GP2, by proteases in the secretory pathway.

The fluorescently labeled complex of streptavidin with the CRD from LSECtin was used to probe cells transfected with Ebola glycoprotein from the Zaire strain of the virus. Fluorescence-activated cell sorting at varying concentrations of labeled receptor revealed concentration-dependent binding (Fig. 9A). Controls in which either non-transfected cells were probed or transfected cells were probed in the presence of EDTA confirmed that the binding is specific (Fig. 9B). Further direct evidence for the interaction of LSECtin with the Ebola glycoprotein was obtained in a solid phase binding assay. The biotin-tagged CRD from LSECtin was immobilized on streptavidin-coated plates and binding of a fusion protein of the GP1 portion of Ebola glycoprotein and the Fc region of human IgG was detected using a protein A-alkaline phosphatase conjugate (Fig. 9C). The binding data could be fitted to a simple saturation binding equation. For two experiments each performed in duplicate, the calculated dissociation constant (K_D) was 3.7 ± 0.3 nM. Parallel assays with human IgG yielded negative results, indicating that the binding observed reflects interaction of LSECtin with glycans on the Ebola glycoprotein portion of the fusion protein rather than with the single glycan attached to the Fc portion of human IgG (45).

Fig. 9. Characterization of LSECtin binding to Ebola virus glycoprotein.

A, Fluorescence-activated cell sorter analysis of cells expressing Ebola glycoprotein probed with Alexa-LSECtin. 293T cells were transiently transfected with plasmid encoding Ebola glycoprotein or control plasmid and incubated with the indicated amounts of Alexa 488-labelled streptavidin-LSECtin complex prior to analysis. The results of a representative experiment are shown. Similar results were obtained in two separate experiments. B, Demonstration of specificity of soluble LSECtin binding to Ebola glycoprotein. Assays were performed as described for panel A, but LSECtin (12.5 μg/ml) was pre-incubated with 25 mM EDTA and the binding reaction carried out in the presence of 25 mM EDTA. Comparable results were obtained upon EGTA treatment (data not shown). SSC, side scattering. C, Solid phase assay of LSECtin binding to Ebola glycoprotein. Biotin-tagged CRD from LSECtin was immobilized on streptavidin-coated plates and probed with Ebola glycoprotein-Fc fusion protein followed by alkaline phosphatase-conjugated protein A. Enzyme activity measured as hydrolysis of p-nitrophenylphosphate was fitted to a simple binding equation. Experimental data are shown as dots and the fitted curve is shown as a solid line.

The binding of LSECtin to Ebola glycoprotein, combined with the accompanying analysis of LSECtin ligand-binding specificity, suggests that the complement of glycans on the viral protein might include glycans with terminal GlcNAcβ1-2Man structures that would form high affinity ligands for the receptor. Although little is known about the glycans attached to Ebola glycoprotein, the amino acid sequence of the protein core suggests that the GP1 portion would be modified with up to 15 N-linked glycans and possibly as many as 80 O-linked glycans (46). Release of N- and O-linked glycans from the Ebola glycoprotein-Fc fusion protein, followed by mass spectrometry, confirmed the presence of both types of glycans. Masses of many of the N-linked glycans suggest that they would be expected to have exposed terminal GlcNAc residues (Fig. 10). Tandem mass spectrometry confirmed that although several of these glycans could contain bisecting GlcNAc residues in β1-4 linkage to mannose, the glycans with m/z 1836, 2040, 2285 and 2735 must contain terminal GlcNAcβ1-2Man structures. Although it is not possible to examine directly the glycans on replication-competent Ebola virus, the results with the isolated surface glycoprotein suggest that GlcNAc-terminated glycans on the surface of the virus may be responsible for the demonstrated ability of LSECtin to mediate infection with the virus (14).

Fig. 10. Mass spectrometry of glycans from Ebola glycoprotein.

Glycans released with protein N-glycanase were permethylated and subjected to MALDI-TOF mass spectrometry. Structures of labeled peaks, except the high mannose oligosaccharides, were further characterized by MALDI-TOF-TOF tandem mass spectrometry. The structures indicated are consistent with the observed masses and fragmentation patterns as well as the known biosynthetic pathway for N-linked glycans. The shapes and colors of symbols for monosaccharides are described in Fig. 4.

The masses of the O-linked glycans released from Ebola glycoprotein are consistent with the presence of typical mucin-type core 1 and core 2 structures (Galβ1-3GalNAcα1-Ser/Thr and Galβ1-3(GlcNAcβ1-6)GalNAcα1-Ser/Thr), but only a small proportion contain sialic acid or fucose residues or have N-acetyllactosamine extensions (data not shown). None of the O-linked oligosaccharides that were detected correspond to high affinity ligands for LSECtin, indicating that the primary interaction between Ebola glycoprotein and LSECtin is likely to be through GlcNAc-terminated N-linked glycans.

The N-glycan analysis also revealed the presence of high mannose oligosaccharides, including Man₈ and Man₉ structures. Man₈ and Man₉ oligosaccharides are bound with high affinity by DC-SIGN and DC-SIGNR (42), so the glycan analysis is consistent with the ability of Ebola glycoprotein to interact with these receptors as well as LSECtin (14,16). Analysis of glycans on the spike glycoprotein of SARS coronavirus revealed that, as in the case of Ebola virus, both high mannose glycans and complex glycans with terminal GlcNAc residues are present (47). This analysis, combined with the demonstration of LSECtin specificity presented here, explains the ability of the SARS spike protein to interact with LSECtin as well as DC-SIGN and DC-SIGNR (14,48,49).

The presence of glycans with 8 to 9 mannose residues is a common feature of surface glycoproteins of enveloped viruses (17,50,51). These glycans represent relatively unprocessed forms of N-linked oligosaccharides, because they have not been subjected to the trimming reactions that are required to generate smaller oligomannose and complex glycans. The data presented here indicate that a substantial proportion of the N-linked glycans of Ebola virus have been processed to hybrid and complex forms, but fully processed N-linked glycans with terminal sialic acid residue are relatively uncommon. Thus, although they have reached a later stage in the processing pathway, these glycans are still truncated. In a similar way, only a small proportion of the O-linked glycans released from Ebola glycoprotein bear sialic acid residues. Therefore, the O-linked glycans also show reduced processing at the terminal stages of biosynthesis. Limited processing might reflect poor accessibility of glycans to processing glycosidases in the endoplasmic reticulum and Golgi apparatus, due perhaps to their dense packing on the glycoproteins or it might be caused by the high level of viral glycoprotein biosynthesis in infected cells, which could result in overloading of the processing machinery. Alternatively, the replication of virus might result in diminished expression of the galactosyltransferases and sialyltransferases that perform the terminal glycosylation steps in glycoprotein biosynthesis or the trafficking of the virus may bypass compartments in which terminal glycosylation normally occurs. In any case, the presence of various types of incompletely processed glycans may be a general feature of the surfaces of enveloped viruses.

Conclusions

LSECtin is encoded in the same gene cluster as DC-SIGN and DC-SIGNR and the CRDs of these three receptors are closely related in overall sequences and share the common features of mannose- and GlcNAc-binding receptors. However the data presented here demonstrate that they bind very distinct sets of glycan ligands. Thus, although LSECtin and DC-SIGNR are both expressed in sinusoidal endothelial cells, they would be expected to bind different sets of physiological and pathological ligands. The specificity data, combined with the analysis of glycans on the Ebola virus glycoprotein, explain the fact that human immunodeficiency virus and hepatitis C virus bind exclusively to DC-SIGN and DC-SIGNR, through high mannose oligosaccharides, while Ebola virus and SARS coronavirus bind to both these receptors and to LSECtin, because of the presence of both high mannose and truncated complex glycans. The results emphasize that the presence of various types of incompletely matured glycans may facilitate the interaction of enveloped viruses with multiple cell surface receptors.