Dengue virus (DV) is a mosquito‐borne flavivirus that causes haemorrhagic fever in humans. DV primarily targets immature dendritic cells (DCs) after a bite by an infected mosquito vector. Here, we analysed the interactions between DV and human‐monocyte‐derived DCs at the level of virus entry. We show that the DC‐specific ICAM3‐grabbing non‐integrin (DC‐SIGN) molecule, a cell‐surface, mannose‐specific, C‐type lectin, binds mosquito‐cell‐derived DVs and allows viral replication. Conclusive evidence for the involvement of DC‐SIGN in DV infection was obtained by the inhibition of viral infection by anti‐DC‐SIGN antibodies and by the soluble tetrameric ectodomain of DC‐SIGN. Our data show that DC‐SIGN functions as a DV‐binding lectin by interacting with the DV envelope glycoprotein. Mosquito‐cell‐derived DVs may have differential infectivity for DC‐SIGN‐expressing cells. We suggest that the differential use of DC‐SIGN by viral envelope glycoproteins may account for the immunopathogenesis of DVs.
Dengue virus (DV) is an arthropod‐borne flavivirus that belongs to the Flaviviridae family (Rice, 1996). Four antigenically distinct serotypes, DV types 1–4, are transmitted to humans through the mosquito vector, Aedes aegypti (Guzman & Kouri, 2002). DVs are lipid‐enveloped viruses with a single‐stranded, positive‐sense RNA genome that encodes the structural proteins C (capsid), M (membrane) and E (envelope), and eight non‐structural proteins, NS1 to NS5 (Rice, 1996). The E‐glycoprotein, which is exposed on the surface of the viral membrane (Kuhn et al., 2002), mediates viral attachment to cells (Hung et al., 1999). The DV E‐glycoprotein has two potential sites for N‐linked glycosylation at positions Asn 67 and Asn 153, which are differentially used by the four DV serotypes (Johnson et al., 1994).
DV infection can cause self‐limiting fever or severe haemorrhagic fever and shock‐syndrome (Halstead, 1989; Rothman & Ennis, 1999; Lei et al., 2001). The pathogenesis of DV is still poorly understood. Immature dendritic cells (DCs), in contrast with macrophages and monocytes, are permissive for DV infection (Palucka, 2000; Wu et al., 2000; Ho et al., 2001; Marovich et al., 2001). The skin‐epidermal Langerhans cells, which are considered to be immature DCs, have been proposed to be the primary target cells after the initial bite by an infected Aedes mosquito (Wu et al., 2000). Early interactions between mosquito‐cell‐derived DV and DCs may be crucial for transporting viral antigens to secondary lymphoid organs and for developing anti‐viral immunity (Kelsa et al., 2002). Consistent with this, DV infection of immature DCs is considered to be a crucial step in the establishment of viral infection. However, DV interactions with DCs are poorly understood at the molecular level (Wu et al., 2000). The attachment sites for viral entry into leukocytes might contribute to the severity of the disease (Morens et al., 1991). Here, we show the importance of the lectin, DC‐specific ICAM‐3‐grabbing nonintegrin (DC‐SIGN; also known as CD209), for the productive infection of monocyte‐derived DCs (MDDCs) by mosquito‐cell‐derived DVs.
Results and Discussion
The ability of immature human MDDCs that lack the monocytic CD14 marker and express DC‐SIGN (Fig. 1A) to support DV type‐1 (DV‐1) virus infection was investigated. MDDCs were infected with the virulent DV‐1 virus, FGA/NA d1d, which was grown in Aedes AP61 cells (Desprès et al., 1998; Duarte dos Santos et al., 2000). Inoculation with five AP61 focus‐forming units (FFU) per cell was needed to infect 50% of MDDCs at 40 h post‐infection, as determined by immunofluorescent staining of DV‐1 antigens (Fig. 1B; using the anti‐DV antibody) and of the DV‐1 NS1 protein, a non‐structural protein that indicates active replication (Fig. 1B; using the anti‐NS1 antibody).
It has been suggested that carbohydrates that are present on the DV virion glycoprotein contribute to both binding and to penetration of the virus into host cells (Hung et al., 1999). The FGA/NA d1d E‐glycoprotein has two N‐linked oligosaccharides (Courageot et al., 2000), and attached carbohydrates for mosquito‐cell‐derived DV are of the high‐mannose type (Johnson et al., 1994). Therefore, we reasoned that DC‐SIGN, which is a type‐II integral tetrameric protein (Geijtenbeek et al., 2000b; Steiman, 2000; Feinberg et al., 2001; Figdor et al., 2002), might interact with the DV‐1 E‐glycoprotein. Indeed, the DC‐SIGN carbohydrate‐recognition domain (CRD) binds mannose residues within high‐mannose oligosaccharides in a calcium‐dependent manner (Feinberg et al., 2001). High‐mannose oligosaccharides are present on envelope glycoproteins of human immunodeficiency virus (HIV; Lin et al., 2003), cytomegalovirus (CMV; Halary et al., 2002), Ebola virus (Alvarez et al., 2002; Lin et al., 2003) and hepatitis C virus (Lozach et al., 2003; Pöhlmann et al., 2003). Incubation with the DC‐SIGN CRD‐specific monoclonal antibodies 1B10 or 8A5 (Halary et al., 2002), or with the soluble, tetrameric ectodomain of DC‐SIGN (sDC‐SIGN; Halary et al., 2002), reduced infection by more than 90%, as assessed by immunodetection of DV antigens (Fig. 1C) or by titration of viral progeny (Fig. 1D). As a positive control, the neutralizing anti‐DV E monoclonal antibody, 9D12, which recognizes an accessible epitope in the DV‐1 virion (Desprès et al., 1993; Duarte dos Santos et al., 2000), reduced infection by 70% (Fig. 1B).
To further evaluate the ability of DC‐SIGN alone to render cells susceptible to DV infection, we compared the ability of parental THP‐1 cells and DC‐SIGN‐expressing THP‐1 cells (THP/DC‐SIGN; Kwon et al., 2002) to support DV‐1 virus replication. THP/DC‐SIGN cells internalize HIV and CMV in a DC‐SIGN‐dependent manner (Geijtenbeek et al., 2000a; Kwon et al., 2002; Halary et al., 2002). THP‐1 and THP/DC‐SIGN cells were infected with various multiplicities of infection (m.o.i.) of FGA/NA d1d virus, and the efficiencies of infection were monitored by immunostaining for viral antigens at 40 h post‐infection (Fig. 2A). Parental THP‐1 cells were almost completely refractory to FGA/NA d1d virus infection (0.5% infected cells were infected at a m.o.i. of 10), whereas THP/DC‐SIGN cells were infected in a dose‐dependent manner, reaching 60% infection at an m.o.i. of 10 (Fig. 2A). DV‐1 that was generated from infected THP/DC‐SIGN cells (at a m.o.i. of 5) was ∼1 × 105 AP61 FFU ml−1 at 48 h post‐infection, and cell death was detected after 72 h (data not shown). Thus, the infectivity shown by mosquito‐cell‐derived DV‐1 was similar in MDDCs and DC‐SIGN‐expressing THP‐1 cells. The ability of DC‐SIGN that was expressed in otherwise refractory Jurkat cells (a lymphoblastoid T‐cell line) to mediate DV‐1 infection (30% cells were positive for DV antigens at a m.o.i. of 5; data not shown) further substantiates the important function of the C‐lectin in the DV life cycle.
As is the case for MDDCs, the use of the anti‐DC‐SIGN monoclonal antibody 1B10, EDTA or sDC‐SIGN abolished or markedly restricted FGA/NA d1d virus infection in THP/DC‐SIGN cells (Fig. 2B). Furthermore, pre‐incubation of FGA/NA d1d virus with concavalin A (ConA), which binds to N‐linked high‐mannose, or treatment of THP/DC‐SIGN cells with yeast mannan, reduced virus infectivity by ∼75% (Fig. 2B). The inhibition by ConA supports the theory that α‐mannose carbohydrate residues participate in the attachment of DV to DC‐SIGN. Exposure of the FGA/NA d1d virus to enzymatic attack by N‐glycosidase F (PNGase F) reduced virus infectivity by 50%, thus proving that carbohydrate moieties of the DV‐1 E‐glycoprotein are involved in DV/DC‐SIGN interactions (Fig. 2B). These results suggest that DC‐SIGN functions as a DV‐binding molecule that is required for the productive infection of MDDCs.
We investigated whether the extent of E‐glycoprotein glycosylation correlates with the effective use of DC‐SIGN for viral entry. DV‐1 and DV‐3 E‐glycoproteins are glycosylated at Asn 67 and Asn 153, whereas DV‐2 and DV‐4 E‐glycoproteins are glycosylated only at Asn 67 (Johnson et al., 1994). The potential glycosylation site at Asn 67 is unique among flaviviruses. THP‐1 cells are refractory to infection by the virulent, wild‐type strain, Jam, of DV‐2 virus that was grown in AP61 cells (Johnson et al., 1994; and data not shown). Immunofluorescent staining of infected cells showed marked differences in infectivity between DV‐1 and DV‐2 viruses at 40 h post‐infection (Fig. 3A). At an m.o.i. of 1, 50% and 5% of THP/DC‐SIGN cells were infected by DV‐1 and DV‐2, respectively. Similar results were seen with mosquito‐cell‐derived DV‐3 and DV‐4, with respect to the predicted number of carbohydrate moieties attached (Fig. 3A). Inoculation with 50 AP61 FFU per cell was required to infect ∼10–20% of THP/DC‐SIGN cells with DV‐4 or DV‐2 (data not shown). Because DVs show a variation in their ability to infect DC‐SIGN‐expressing cells, immature MDDCs were exposed to highly purified DV‐1 FGA/NA d1d and DV‐2 Jam viruses that were grown in mosquito cells (Fig. 3B). On average, ∼25% of MDDCs were infected by DV‐2 and ∼70% by DV‐1 at the highest m.o.i. tested (p < 0.001, from t‐tests carried out in accordance with the method of Fisher and Yates). Thus, mosquito‐cell‐derived DVs may have differential infectivities for MDDCs. Notably, monoclonal antibody 1B10 can block DV‐2 infection of MDDCs, as is the case for DV‐1 virus (Fig. 3B, +1B10). These findings support the theory that the high‐mannose N‐oligosaccharide at Asn 153 might account for the higher affinity of the E‐glycoprotein binding to the oligomeric DC‐SIGN CRD (Feinberg et al., 2001) and, ultimately, for the efficient infection of MDDCs.
We examined whether the wild‐type strain IS‐98‐ST1 of West Nile (WN) virus and the 17D vaccine strain of yellow fever (YF) virus use DC‐SIGN for the infection of monocytic cells. Asn 153 in WN IS‐98‐ST1 E‐glycoprotein (Mashimo et al., 2002) carries a carbohydrate (P.D., unpublished data) and the YF 17D E‐protein has an N‐glycosylation site that is not used (Desprès et al., 1991). Neither THP‐1 nor THP/DC‐SIGN cells were infected by YF virus at an m.o.i. of 50 vero‐FFU per cell (Fig. 3C). From these results, it is reasonable to predict that carbohydrate moieties attached to the flavivirus E‐glycoprotein are crucial for efficient and productive infection, which requires host‐cell expression of DC‐SIGN. Interestingly, the replication of WN virus in THP‐1 cells was not dependent on DC‐SIGN (Fig. 3C). This suggests that alternative receptor molecules that allow the infection of monocytic cells by flaviviruses, such as WN virus, must exist.
Heparan sulphate (HS) glycosaminoglycans have been found to function in the cell binding of DVs and have been suggested to be putative receptors for the infection of susceptible cells by DV (Chen et al., 1997; Hung et al., 1999). Consistent with a recent report by Germi et al. (2002), we found that heparin‐lyase I, which specifically removes HS, partially inhibited YF 17D virus infection in VERO cells (data not shown). Treatment of THP/DC‐SIGN cells with heparin lyase I or pre‐incubation of DV‐1 with heparin 6,000 (an HS analogue) had no effect on virus infection (data not shown), indicating that HS did not influence the interactions between DV and DC‐SIGN. Flavivirus uptake involves the endocytic pathway, in which acidification and endosomal vesicles are required (Heinz & Allison, 2000). Treatment with the pH‐interfering drugs bafilomycin A1 (Fig. 4A) and chloroquine (Fig. 4B) caused a dose‐dependent reduction of DV‐1 infectivity in THP/DC‐SIGN cells. This suggests that DC‐SIGN‐mediated DV entry is a pH‐dependent process that probably requires the endocytosis of incoming virions.
Our data suggest that DC‐SIGN is crucial for DV binding to MDDCs, the putative primary host cells in humans. However, they do not show whether DC‐SIGN functions in both attachment and penetration of DV or constitutes part of a putative viral receptor complex. The dependence of the viral entry of prototype strain New Guinea C of DV on HS proteoglycan interactions has been reported in cells that lack DC‐SIGN, such as epithelial cells (Chen et al., 1997; Germi et al., 2002). Although the physiological relevance of this observation needs to be elucidated, it underlines the existence of molecules other than DC‐SIGN that show viral attachment properties for DV (Bielefeldt‐Ohmann et al., 2001). The blocking effect of pH‐interfering drugs suggests that the endocytic pathway contributes to the penetration of DV. Two sorting signals for the endocytic pathway are located in the cytoplasmic tail of DC‐SIGN, and they could account for the DV entry process (Soilleux et al., 2000). However, we cannot formally rule out the possibility that an unidentified protein that shares the ability to promote DV uptake with DC‐SIGN is the target of the blocking effect of bafilomycin A1.
Our findings identify DC‐SIGN as an attachment factor for DV, which enters the blood after an initial bite by an infected mosquito. Although the relevance of these observations remains to be ascertained in vivo, we propose that the immunopathogenesis of DV infection might depend on both the expression of DC‐SIGN in human DC subpopulations and the ability of viral envelope glycoproteins to interact with DC‐SIGN.
MDDCs were generated from human blood monocytes using granulocyte macrophage colony stimulating factor and interleukin‐4, as described in Halary et al. (2002). The human monocytic cell line, THP‐1, was grown in RPMI 1640 containing 10% FCS, 2 mM L‐glutamine and antibiotics. THP/DC‐SIGN cells were provided by D.R. Littman (Kwon et al., 2002).
Antibodies and soluble DC‐SIGN.
Anti‐DC‐SIGN monoclonal antibodies 1B10 and 8A5 have been described previously (Halary et al., 2002). Anti‐DV E monoclonal antibody 9D12 was a gift from M.K. Gentry and R. Putnak. Anti‐LCMV (lymphocytic choriomeningitis virus) monoclonal antibody BD12.5 (IgG γ2a) was used as a negative control. sDC‐SIGN has been described previously (Halary et al., 2002).
The production of DV‐1 strain FGA/NA d1d (GenBank accession number AF226686), DV‐2 strain Jam (M20558), DV‐3 strain H‐87 (NC001475), DV‐4 strain H‐241 (NC002640), and WN virus strain IS‐98‐ST1 (AF481864) from Aedes pseudoscutellaris AP61 cell monolayers and virus titration on AP61 cells by focus immunodetection assays (FIAs) were performed as described previously (Desprès et al., 1993). DV‐1, DV‐2 and WN viruses were highly purified on sucrose gradients, as described previously (Desprès et al., 1993). Infectivity titres were expressed as FFUs in AP61 cells. Vaccine strain 17D‐204 of YF virus (STAMARIL; Aventis Pasteur Vaccins; GenBank accession number X15062) was propagated twice in African green monkey kidney VERO cell monolayers and purified using sucrose gradients. Infectivity titres were expressed as FFU in VERO cells.
Cells were adhered to glass Lab‐tek chambers (Nalge Nunc International) coated with poly‐L‐lysine (Sigma; 5 × 104 cells cm−2). Adherent cells were washed once with RPMI 1640, infected with flavivirus in RPMI 1640 supplemented with 0.2% BSA, pH 7.5, for 2 h at 37 °C, and incubated with RPMI, 2% FCS for 40 h at 37 °C.
Cells were with fixed with 3.2% paraformaldehyde (PFA) in PBS for 20 min, treated with 50 mM NH4 Cl in PBS for 10 min, and permeabilized with 0.1% Triton X‐100 for 5 min. Intracellular viral antigens were stained with anti‐DV‐specific hyperimmune mouse ascites fluids (HMAF), anti‐YF‐virus‐specific HMAF or anti‐WN‐virus‐specific HMAF, and the secondary antibody used was a FITC‐conjugated goat‐anti‐mouse IgG (Sigma). Cells were examined using an AXIOPLAN 2 fluorescence microscope (Zeiss). Images were processed using RS Image 1.07, Adobe Photoshop and Powerpoint software.
Deglycosylation of dengue virus virions.
Highly purified FGA/NA d1d virus (1 × 108 AP61 FFU) was incubated with 1 unit of PNGase F (Roche Applied Science) in 20 mM sodium phosphate (pH 7.6) for 7 h at 37 °C. PNGase‐F‐treated virus and mock‐treated virus were used to infect THP/DC‐SIGN cells for 48 h. Infected cells were then fixed with 3.2% PFA, washed, and incubated sequentially with anti‐DV‐1 HMAF and phycoerythrin‐conjugated anti‐mouse IgG antibody (Sigma). Cells were analysed using a FACScan machine (Becton‐Dickinson) and data were processed using CellQuest 3.3 software.
pH‐interfering drug treatments.
Bafilomycin A1 and chloroquine were obtained from Sigma. Cells were infected with DV as described above. Infections were performed in the presence of the bafilomycin A1 or chloroquine, followed by washing to remove unbound virus, and cells were further incubated for 3 h with the same drugs. Cells were then washed, and were incubated at 37 °C until 40 h after infection.
The authors thank M. Flamand for providing the anti‐DV NS1 monoclonal antibody. We acknowledge the assistance provided by P.‐E. Lozach, M.‐T. Drouet, I. Staropoli and C. Houlès. This work was supported by grants from Direction de la Valorisation et des Partenariats Industriels (Pasteur Institute) and the French National AIDS Research agency (ANRS). E.N.‐S. is funded by scholarship funds from the SFERE‐CONACYT.
- Copyright © 2003 European Molecular Biology Organisation