Identification of Cell Surface Molecules Involved in Dystroglycan-Independent Lassa Virus Cell Entry

Abstract

Although O-mannosylated dystroglycan is a receptor for Lassa virus, a causative agent of Lassa fever, recent findings suggest the existence of an alternative receptor(s). Here we identified four molecules as receptors for Lassa virus: Axl and Tyro3, from the TAM family, and dendritic cell-specific intercellular adhesion molecule 3-grabbing nonintegrin (DC-SIGN) and liver and lymph node sinusoidal endothelial calcium-dependent lectin (LSECtin), from the C-type lectin family. These molecules enhanced the binding of Lassa virus to cells and mediated infection independently of dystroglycan. Axl- or Tyro3-mediated infection required intracellular signaling via the tyrosine kinase activity of Axl or Tyro3, whereas DC-SIGN- or LSECtin-mediated infection and binding were dependent on a specific carbohydrate and on ions. The identification of these four molecules as Lassa virus receptors advances our understanding of Lassa virus cell entry.

INTRODUCTION

Lassa virus, a member of the Arenaviridae family of viruses, causes Lassa fever in humans (9). With more than 200,000 infections and several thousand deaths per year, Lassa fever poses a huge public health threat, especially in West Africa (43). In addition, more than 20 cases of imported Lassa fever in Japan, Europe, and North America have been reported, and the case fatality rate for imported cases is higher than that for nonimported cases (24). No vaccine against Lassa fever has been approved for human use. Ribavirin, a nucleoside analogue, is the sole drug to have shown at least partial efficacy in the treatment of Lassa fever (44).

The natural reservoir of Lassa virus is Mastomys natalensis, a species of rodent that prefers to occupy human habitats (47). Virus-contaminated urine is thought to be a major source of reservoir-to-human transmission, whereas airborne infection is rare (32). Close contact with the body fluids of infected patients is the highest risk factor for human-to-human transmission (43). Because poor inflammatory and immune responses are observed in fatal infection (43) and because in vitro infection of dendritic cells, macrophages, and endothelial cells downregulates the production of inflammatory mediators (3, 39–41), these cells appear to be early targets for Lassa virus infection in humans. Postmortem examinations have found mild histological lesions in the liver, adrenal gland, and kidney, and high viral burdens in the liver, lung, spleen, kidney, and heart have also been reported (43, 45, 65).

The interaction between a virus and its cellular receptor(s) is important for the determination of viral tissue and host tropisms. Arenaviruses express four viral proteins from two ambisense RNA genomes, one of which is a glycoprotein (glycoprotein precursor [GPC]) that mediates viral binding to and entry into cells (9). By using a virus overlay protein blot assay and the peptide sequence of the GPC of lymphocytic choriomeningitis virus (LCMV), Cao et al. (10) identified α-dystroglycan (α-DG) as a binding receptor for LCMV and also showed that Lassa virus and several other Arenavirus members use this molecule as a receptor. α-DG and β-DG constitute a DG complex; α-DG binds components of the extracellular matrix, such as laminin, while β-DG spans the cellular membrane and binds the intracellular cytoskeleton (29). DG is widely distributed, but its expression levels and glycosylation levels differ depending on the tissue (5, 28, 29, 52). Almost half of the O-linked glycosylation of α-DG is with O-mannosyl carbohydrates, which are rare among mammals (12, 57), and several glycosyltransferases for this O-mannosylation have been identified (52, 71). Defects in the glycosyltransferases reduce the level of O-mannosylation of DG and impair its ligand binding, with devastating effects on muscle fiber integrity and neural migration (42, 50).

Recently, O-mannosylation was reported to be necessary for DG to function as a receptor for Lassa virus (34). Expression of wild-type DG, but not expression of a mutant lacking O-mannosylation, conferred Lassa virus GPC-mediated infection of DG-null cells (35). Soluble α-DG mutants lacking O-mannosylation failed to bind Lassa virus particles, whereas enhanced glycosylation resulted in greater Lassa virus binding (34). Similar correlations among DG O-mannosylation, virus binding, and virus infection have been reported based on analyses with LCMV (30, 34, 35), suggesting a common property of GPC between Lassa virus and LCMV. However, although laminin is a ligand for DG and blocks the binding of Lassa virus GPC to DG (35), it cannot block Lassa virus GPC-mediated infection of Vero cells (34). The level of LCMV replication in mice that lack the gene for acetylglucosaminyltransferase-like protein (LARGE) or the gene for protein O-linked mannose β-1,2-N-acetylglucosaminyltransferase 1 (POMGnT1) in immune cells, both of which are responsible for the O-mannosylation of DG (52, 71), was comparable to that in wild-type mice (31). These findings suggest that DG is not the sole receptor for Lassa virus or LCMV.

Here we screened cellular cDNA libraries to identify molecules that conferred Lassa virus GPC-mediated infection. Four cellular transmembrane proteins were identified as binding receptors. When expressed in cells, each of these proteins allowed efficient Lassa virus GPC-mediated infection that occurred independently of DG modification and of DG itself.

MATERIALS AND METHODS

Cells.

Vero E6, HT1080, HeLa, 293T, Jurkat (clone E6-1), and Plat-GP (murine leukemia virus [MLV]-based packaging) cells were cultured as described previously (48, 61). The DG-knockout embryonic stem (ES) cell clone 354.B11 (25) was kindly provided by K. P. Campbell (Howard Hughes Medical Institute, University of Iowa) and was cultured by use of the EmbryoMax ES cell culture system (Millipore, Bedford, MA).

cDNA library.

The preparation of a cDNA library from Vero E6 cells in the pMX plasmid has been reported elsewhere (61). A human liver cDNA library in the pFB plasmid was purchased from Stratagene (La Jolla, CA). MLV retroviral vectors carrying these libraries were prepared as described previously (61) and were then used to infect Jurkat cells at a multiplicity of infection (MOI) of 0.2.

Pseudotype viruses.

Envelope proteins of Lassa virus (strain Josiah) (GPC), vesicular stomatitis virus (VSV; strain Indiana) (VSV G), severe acute respiratory syndrome coronavirus (SARS-CoV; strain Tor2) (Spike), and MLV (MLV Env, ecotropic) were expressed from the pCAGGS/MCS expression plasmid (33, 51). These proteins were used to pseudotype a human immunodeficiency virus (HIV)-based lentiviral vector or VSV as described previously (60, 63). Titration of pseudotype viruses carrying the green fluorescent protein (GFP) or Venus (a variant of yellow fluorescent protein [YFP]) gene was performed by inoculating cells with serially diluted virus stocks and counting reporter-positive cells under a fluorescence microscope (TE300; Nikon) 24 h (VSV pseudotype) or 48 h (HIV pseudotype) later.

Library screen.

Library-transduced Jurkat cells were incubated with an HIV lentiviral vector carrying Lassa virus GPC as an envelope protein and the feline CD2 gene as part of the genome [referred to as HIV-LaGPC(fCD2)] at a 1:5 dilution of the viral stock. Two days later, the cells were transferred to an anti-fCD2 antibody-coated dish to capture fCD2-positive cells (61). After 5 days of culture, colony-forming cells were further incubated with HIV-LaGPC(Venus) for 2 days. The cDNAs inserted into the Venus-positive colonies were amplified and sequenced as described previously (61).

Experiments with live Lassa virus.

All work with live Lassa virus was performed in a biosafety level 4 (BSL-4) laboratory at the National Microbiology Laboratory of the Public Health Agency of Canada. Jurkat cells were infected with the Josiah strain of Lassa virus at an MOI of 0.2 (a ratio of PFU titrated in Vero E6 cells to Jurkat cell numbers). Culture supernatants were harvested on the days indicated in Fig. 1C, and stocks were prepared and stored at −80°C until use. Infectious titers were measured by means of a conventional plaque-forming assay in Vero E6 cells.

Fig 1.

Lassa virus GPC-mediated infection. (A and B) Titers of pseudotyped viruses carrying Lassa virus GPC as an envelope protein in Jurkat cells expressing each of the following separately: the control (fCD2), Axl, Tyro3, DC-SIGN, or LSECtin. The backbones of the pseudotyped viruses are HIV (A) and VSV (B). Serially diluted pseudotyped viruses were inoculated onto cells, and 48 h (A) or 24 h (B) later, reporter (Venus or GFP)-expressing cells were counted under a fluorescence microscope. Data are means ± standard deviations (n = 3). IU, infectious units. (C) Replication of authentic Lassa virus in Jurkat cells. Cells were inoculated with authentic Lassa virus at an MOI of 0.2. Culture supernatants were harvested immediately after inoculation (day 0) or on the indicated days and were titrated in Vero E6 cells. Experiments were performed in duplicate, and means are shown. nd, not detectable.

Antibodies, proteins, and reagents.

Antibodies to the TAM family (clone 108724, clone 96201, goat polyclonal antibody AF154) and to dendritic cell-specific intercellular adhesion molecule 3-grabbing nonintegrin (DC-SIGN)/DC-SIGN-related (DC-SIGNR) (DC28) and recombinant human chimeric proteins (Axl/Fc and Dtk/Fc [also referred to as Tyro3/Fc]) were purchased from R&D Systems Inc. (Minneapolis, MN). An antibody against the hemagglutinin (HA) tag (clone HA-7) and anti-β-actin (AC-74) were purchased from Sigma (St. Louis, MO); anti-α-DG (IIH6C4) and anti-β-DG (43DAG1/8D5) were purchased from Upstate (Lake Placid, NY) and Abcam (Cambridge, MA), respectively. Polyclonal antibodies to Lassa virus GPC were obtained by immunizing rabbits with the synthetic peptide WTLSDSEGKDTPGGYC, which lies within the GP2 region of Lassa virus GPC. Soluble α-DGs were prepared as reported elsewhere (36) with some modifications. Two recombinant proteins, DGFc5, containing amino acids (aa) 30 to 653 of α-DG (for full-length α-DG), and DGFc1, containing aa 30 to181 of α-DG (for truncated α-DG), fused to the murine immunoglobulin κ chain signal sequence and the human immunoglobulin G1 Fc region at the N and C termini, respectively, were expressed from pCAGGS/MCS plasmids in 293T cells. Proteins were purified from culture supernatants of 293T cells by using the MAb Trap Kit (GE Healthcare, Uppsala, Sweden), and buffers containing the purified proteins were changed to phosphate-buffered saline (PBS) by use of a PD-10 column (GE Healthcare). Protein solutions were then concentrated by using a Microcon YM-30 instrument (Millipore). A series of recombinant Tyro3/Fc proteins (see Fig. 3D) was prepared by using the same strategy; the intact signal sequence of Tyro3, but not that of the murine immunoglobulin, was used. The tyrosine kinase inhibitor genistein was purchased from Sigma and was used at the concentrations indicated in Fig. 4.

Fig 3.

Schematic diagrams of Axl mutants (A and B), Tyro3 mutants (C), and soluble chimeric Tyro3 (D). The leftmost schematics of panels A and C represent intact Axl and Tyro3, respectively. N, amino terminus; Ig, immunoglobulin domain; FN, fibronectin type III domain; TM, transmembrane domain; TK, tyrosine kinase domain; C, carboxyl terminus. Solid squares, HA tags; solid circles, point mutations. Fc, Fc portion of immunoglobulin.

Fig 4.

Genistein inhibits infection of Jurkat cells expressing Axl (A) or Tyro3 (B) by a pseudotype virus carrying Lassa virus GPC (♢) or VSV G (□). Cells were infected with pseudotype viruses in the presence of genistein at the indicated concentrations; 48 h later, reporter (Venus)-expressing cells were counted. Infectivity was calculated as (titer in the presence of genistein)/(titer in the presence of the vehicle) × 100. Data are means ± standard deviations (n = 3).

Virus binding assays.

For the cell-based assay: cells were incubated with HIV-LaGPC(Venus) on ice with occasional mixing. One hour later, cells were washed three times with cold culture medium and were then suspended in 450 μl of culture medium. HIV antigens in the suspension were measured by using an HIV-1 p24 antigen enzyme-linked immunosorbent assay (ELISA) (ZeptoMetrix, Franklin, MA). For the Fc chimera-based assay, chimeric proteins fused with the Fc portion of human IgG were incubated with HIV-LaGPC(Venus) on ice. Thirty minutes later, protein G-Sepharose (GE Healthcare) was added, and the mixture was incubated, with occasional mixing, for 1 h at 4°C. The Sepharose was then washed three times with culture medium, and the HIV antigen was measured as described for the cell-based assay. When virus-like particles (VLPs) were used instead of pseudotype virus, washed cells or Sepharose was suspended in Tris-glycine sodium dodecyl sulfate (SDS) sample buffer (Invitrogen, Carlsbad, CA) and was analyzed by Western blotting with an antibody to Lassa virus GP2. Band densities were used to compare VLP quantities.

Flow cytometry.

Antibody staining and analyses were performed as described previously (61).

Knockdown of DG RNA.

To reduce DG expression by using RNA interference, a short hairpin RNA (shRNA) system based on the MLV retroviral vector was used. Two plasmids for this system, pmU6 and pSSSP, were kindly provided by Hideo Iba (University of Tokyo). The target sequences used were GGGTACAGTTCAACAGCAACA (for DG) and GGGACGATGATATCCGTCTGA (for LacZ, the control). MLV retroviral vectors were produced as described above and were used to infect Jurkat cells. Jurkat cells into which shRNA was introduced were selected by culturing in the presence of puromycin.

RESULTS

cDNA library screens identified members of the TAM family and the C-type lectin family as facilitators of Lassa virus cell entry.

Lassa virus replicates well in Vero E6 cells and causes extensive lesions in the liver (45, 65, 67, 68). Therefore, we selected this cell line and this tissue as sources for our cDNA libraries. Because Garbutt et al. (21) and Rojek et al. (54) reported high resistance of Jurkat cells to Lassa virus cell entry, we introduced the cDNA libraries from Vero E6 cells and liver into Jurkat cells by using an MLV vector to identify the gene(s) that conferred susceptibility to Lassa virus entry into cells. Cells were then incubated with HIV-LaGPC(fCD2), an HIV-based lentiviral vector carrying Lassa virus GPC as an envelope protein and the fCD2 gene as part of the genome. fCD2-positive cells were captured by incubation in an anti-fCD2 antibody-coated dish (61). To avoid false-positive selection caused by nonspecific infection, colony-forming cells in the dish were further incubated with HIV-LaGPC(Venus). From the genomes of the Venus-positive colonies, inserted cDNAs were amplified and were sequenced as described previously (61). Sequence analysis revealed that nine colonies harvested from cells carrying the Vero E6 cDNA library contained Axl cDNA and four colonies from cells carrying the liver cDNA library contained liver and lymph node sinusoidal endothelial calcium-dependent lectin (LSECtin) cDNA. Axl belongs to the TAM receptor-type tyrosine kinase family, which also includes Tyro3 (also called Dtk) and Mer (14, 37). LSECtin belongs to the calcium-dependent (C-type) lectin family, which includes dendritic cell-specific intercellular adhesion molecule-3-grabbing nonintegrin (DC-SIGN) and several other lectins. To examine the effects of these two protein families on Lassa virus entry into cells, we expressed human Axl, Tyro3, Mer, LSECtin, DC-SIGN, DC-SIGN-related (DC-SIGNR), macrophage galactose lectin (MGL), and fCD2 (a negative control) in Jurkat cells by using a lentiviral vector and evaluated the susceptibility of cells expressing these molecules by using the HIV pseudotype virus HIV-LaGPC(Venus) (with an HIV backbone, Lassa virus GPC as the envelope, and a genome encoding the reporter Venus). Expression of Axl, Tyro3, DC-SIGN, or LSECtin resulted in titers of HIV-LaGPC(Venus) 100- to 1,000-fold higher than those induced following fCD2 expression (Fig. 1A). No significant changes in titers were observed when Mer, DC-SIGNR, or MGL was expressed (data not shown). The susceptibility-enhancing effects of Axl, Tyro3, DC-SIGN, and LSECtin were also observed when another, VSV-based pseudotype virus, VSV-LaGPC(GFP) (with a VSV backbone, an envelope consisting of Lassa virus GPC, and a genome including the GFP gene), was used (Fig. 1B), indicating that the effects of Axl, Tyro3, DC-SIGN, and LSECtin are dependent not on the backbones of the pseudotype viruses but on Lassa virus GPC.

Ebola virus infection is enhanced by the expression of the TAM family and the C-type lectin family (2, 8, 23, 27, 61). We have reported that Ebola virus glycoprotein (GP)-mediated infection is enhanced by the TAM family members Axl and Tyro3 (61). Brindley et al. (8) and Hunt et al. (27) showed that endosomal uptake of Ebola virus particles is facilitated by Axl. The members of the C-type lectin family DC-SIGN and LSECtin bind Ebola virus particles and mediate viral entry (2, 23). Since we found that these four molecules—Axl, Tyro3, DC-SIGN, and LSECtin—also enhance Lassa virus GPC-mediated infection (Fig. 1A and B), we tested their specificity with respect to enhancing virus entry. We selected three viral glycoproteins—VSV G, SARS-CoV Spike (SARSSpike), and MLV Env—and examined the effects of Axl, Tyro3, DC-SIGN, and LSECtin on virus infection mediated by these viral glycoproteins. All three viral glycoproteins can pseudotype HIV and VSV, like Lassa virus GPC and Ebola virus GP; infection by these pseudotype viruses requires a low pH in the endosome, like infection by authentic Lassa virus and Ebola virus (26, 66, 70). Jurkat cells were highly susceptible to HIV-VSVG(Venus) (3.0 × 10⁸ infectious units [IU]/ml), and expression of the four molecules did not enhance their susceptibility to this virus (data not shown). Titers of HIV-SARSSpike(Venus) were not detectable in control-, Axl-, or Tyro3-expressing Jurkat cells and were low in DC-SIGN- or LSECtin-expressing cells (1.3 × 10¹ IU/ml in both; the detection limit of the assay is 6.7 × 10⁰ IU/ml). Similarly, we did not detect HIV-MLVEnv(Venus) infection in these cells. To ensure that this lack of infectivity of HIV-SARSSpike(Venus) and HIV-MLVEnv(Venus) originated from a lack of receptors for these envelope proteins, Jurkat cell lines expressing either the SARS-CoV receptor ACE2 or the MLV receptor mCAT were produced and tested for their susceptibilities to HIV-SARSSpike(Venus) and HIV-MLVEnv(Venus). The titers of HIV-SARSSpike(Venus) and HIV-MLVEnv(Venus) were 1.5 × 10⁵ IU/ml in ACE2-expressing cells and 1.6 × 10⁶ IU/ml in mCAT-expressing cells, respectively. These results indicate that the four molecules Axl, Tyro3, DC-SIGN, and LSECtin enhance the susceptibility of cells to HIV and VSV pseudotyped with Lassa virus GPC via specific interactions between Lassa virus GPC and these molecules.

Members of the TAM family and the C-type lectin family enhance authentic Lassa virus replication.

We next tested the importance of Axl, Tyro3, DC-SIGN, and LSECtin in authentic Lassa virus infection. As shown in Fig. 1C, Lassa virus replicated more rapidly in Jurkat cells expressing any of these four molecules than in control Jurkat cells; more than a 1-log difference was observed at days 2 and 3 for cells expressing Axl or Tyro and at days 1, 2, and 3 for cells expressing DC-SIGN or LSECtin. Similar results were obtained when real-time reverse transcription-PCR targeting the Lassa virus NP gene was performed (data not shown). These data suggest that Axl, Tyro3, DC-SIGN, and LSECtin may be receptors/coreceptors for Lassa virus.

Members of the TAM family and the C-type lectin family bind Lassa virus particles.

To examine the roles of Axl, Tyro3, DC-SIGN, and LSECtin in the entry of Lassa virus into cells, we performed a cell-based virus particle binding assay (see Materials and Methods). Cells were first incubated with virus particles and then washed to remove unbound particles, and virus antigens remaining on the cell surface were measured. We used an HIV-based pseudotype virus, HIV-LaGPC(Venus), because sensitive HIV p24 antigen detection ELISA kits are commercially available. Expression of Axl, DC-SIGN, or LSECtin, but not Tyro3, enhanced the binding of HIV-LaGPC(Venus) to Jurkat cells (Fig. 2A, left). The expression of LARGE, which is a putative N-acetylglucosaminyltransferase (52) and causes hyperglycosylation of DG (4), resulted in increased virus binding (Fig. 2A, center). Expression of Axl, Tyro3, DC-SIGN, or LSECtin did not increase the binding of HIV-VSVG(Venus), whose envelope contains VSV G protein (Fig. 2A, right). The interaction between Tyro3 and the virus particles was also assessed by using an Fc chimera-based assay. Proteins (e.g., Tyro3) fused with the Fc portion of immunoglobulin were mixed with virus particles; the mixture was then incubated with protein G-Sepharose; and virus antigens bound to the Sepharose were measured. Tyro3/Fc bound to HIV-LaGPC(Venus) particles, whereas Axl/Fc failed to bind to these virus particles (Fig. 2B, left), even though Lassa GPC binding was detected in the cell-based assay (Fig. 2A). The Fc chimera-based assay was, however, validated; DGFc5, a soluble form of O-mannosylated DG fused with Fc, bound to HIV-LaGPC(Venus) particles, but DGFc1, a truncated form of DG (36), did not (Fig. 2B, right).

Fig 2.

Virus-binding assays. (A and B) Binding of HIV-based pseudotype particles to Jurkat cells expressing the indicated molecules (cell-based assay) (A) or to soluble chimeric proteins (Fc chimera-based assay) (B) was measured by using an HIV-1 p24 antigen ELISA. (C and D) The binding of Lassa VLPs to Jurkat cells expressing the indicated molecules (C) or to soluble proteins (D) was examined by Western blotting with an anti-Lassa virus GP2 antibody. Band intensities were used to compare the amounts of GP2; β-actin densities were used for normalization. The experiments were performed twice, and representative results are shown. (E) Comparison of expression levels of Axl and Dtk. HA-tagged Axl and Tyro3 (Axl HA and Tyro3 HA, respectively) (Fig. 3) were expressed in Jurkat cells. Untagged, intact Axl and Tyro3 served as controls. Cells were analyzed by using flow cytometry with an anti-HA tag antibody. (F) The binding of HIV-based pseudotype particles to soluble Tyro3 proteins was measured by using an HIV-1 p24 antigen ELISA. Schematic diagrams of the soluble proteins used are shown in Fig. 3D.

Coexpression of the major structural proteins of Lassa virus—GPC, Z, and NP—in cells results in the production of Lassa virus-like particles (VLPs) in culture supernatants (56). We performed the cell-based and Fc chimera-based binding assays using the Lassa VLPs to find out whether results similar to those observed with the pseudotypes would be obtained. To compare the amounts of bound viral proteins, we performed Western blotting with an anti-GP2 antibody. In the cell-based assay, Lassa VLPs bound to cells expressing Axl, Tyro3, DC-SIGN, or LSECtin to a greater extent than to control cells (Fig. 2C). In the Fc chimera-based assays, Tyro3/Fc, but not Axl/Fc, showed considerable binding with Lassa VLPs (Fig. 2D), as was found with HIV pseudotypes (Fig. 2B, left). These results suggest that Axl, Tyro3, DC-SIGN, and LSECtin function as binding receptors for the virus.

To compare the expression levels of Axl and Tyro3 in Jurkat cells, an HA tag was inserted into a region proximal to the transmembrane domains of both molecules (Axl HA and Tyro3 HA [Fig. 3A and C]). Jurkat cells expressing Axl HA or Tyro3 HA were analyzed by flow cytometry. Untagged Axl and Tyro3 served as controls. An anti-Axl antibody reacted with Axl and Axl HA equally, while the anti-Tyro3 antibody reacted similarly with Tyro3 and Tyro3 HA (data not shown). The anti-HA antibody reacted more strongly with Axl than with Tyro3 (Fig. 2E), indicating that the expression levels of Axl were higher than those of Tyro3 and that the nonexistent or low binding of Lassa virus pseudotypes or VLPs to Tyro3-expressing cells (Fig. 2A and C) was due to the low expression level of Tyro3.

Roles of the extracellular and intracellular regions of Axl and Tyro3.

To determine which regions of Axl are important for Lassa virus infection, we constructed a series of Axl mutants (Fig. 3A) that had various truncations in the extracellular region and an HA tag in a region proximal to the transmembrane domain. The HA tag was used in flow cytometry to compare the mean fluorescence intensities (MFI) so as to determine the expression levels of the Axl mutants. To measure the binding of the pseudotype virus to the Axl mutants, we performed cell-based virus binding assays. We also examined the abilities of the Axl mutants to render Jurkat cells susceptible to Lassa virus infection by titrating the HIV-LaGPC(Venus) pseudotype in Axl mutant-expressing cells. Table 1 shows the results of these three experiments. Truncation of the first immunoglobulin domain (Axl del1D) (Fig. 3A) resulted in slightly better Axl detection but dramatically lower virus binding and HIV-LaGPC(Venus) titers than those for Axl HA (that is, Axl possessing the HA tag). Axl 1D, which lacked the second immunoglobulin domain and two fibronectin type III domains (leaving only the first immunoglobulin domain as the extracellular component of Axl), was detected at the cell surface at levels 4-fold higher than those of HA-tagged Axl. However, Axl 1D did not bind efficiently to LaGPC and did not efficiently confer susceptibility to Jurkat cells. Axl 1D2D, which lacked the two fibronectin type III domains, was similar to HA-tagged Axl in terms of cell surface detection levels, virus binding, and ability to confer HIV-LaGPC(Venus) susceptibility. These data suggest that the two immunoglobulin domains of Axl are necessary for Lassa virus infection.

Table 1.

Axl mutants with mutations in extracellular regions and HIV-LaGPC(Venus) binding and titers

Axl varianta	Level of detectionb	Virus bindingc	Titer ratiod
Axl HA	100	100	100
Axl del1D	135	3	0
Axl 1D	419	38	46
Axl 1D2D	104	92	101

^aSchematic structures of Axl mutants are shown in Fig. 3A. Axl HA, Axl possessing an HA tag in a region proximal to the transmembrane domain.

^bCalculated as (MFI of mutant)/(MFI of Axl HA) × 100. MFI of Axl mutants were examined by flow cytometry with an anti-HA antibody.

^cThe binding of HIV-LaGPC(Venus) particles to Jurkat cells expressing Axl mutants was measured by using an HIV p24 antigen ELISA and was calculated as (p24 value in mutant-expressing cells)/(p24 value in Axl HA-expressing cells) × 100.

^dTiters of HIV-LaGPC(Venus) were measured in Axl mutant-expressing Jurkat cells. Ratios were calculated as (titer in mutant-expressing cells)/(titer in Axl HA-expressing cells) × 100.

To determine the contributions of the intracellular regions of Axl, we produced three mutants (Fig. 3B) whose intracellular regions were truncated or contained point mutations. The detectable levels of the Axl mutants, the levels of binding of the pseudotype virus HIV-LaGPC(Venus) to mutant-expressing Jurkat cells, and HIV-LaGPC(Venus) titers in those cells are shown in Table 2. Axl delCT, which lacked the entire intracellular region, was detected at higher levels on the cell surface and bound to HIV-LaGPC(Venus) more efficiently than intact Axl, but it did not confer HIV-LaGPC(Venus) susceptibility on cells. The point mutation K567M (i.e., a lysine-to-methionine substitution at position 567), which is known to abolish the ATP-binding activity and thus the tyrosine kinase activity of Axl (7), had no effect on Axl cell surface detection levels and reduced Axl binding to the virus to only a limited extent. However, this mutation abolished the ability of Axl to confer HIV-LaGPC(Venus) susceptibility on cells. A tyrosine residue at position 821 in Axl is phosphorylated by the Axl tyrosine kinase upon signaling (7). A Y821F point mutation (Axl Y821F) did not affect Axl cell surface detection levels or virus binding but decreased the ability of Axl to confer HIV-LaGPC(Venus) susceptibility on cells to 31%. These data suggest that Axl-mediated infection requires intracellular signaling mediated by the tyrosine kinase activity of Axl.

Table 2.

Axl mutants with mutations in intracellular regions and HIV-LaGPC(Venus) binding and titers

Axla	Level of detectionb	Virus bindingc	Titer ratiod
Intact Axl	100	100	100
Axl delCT	305	186	4
Axl K567M	103	73	2
Axl Y821F	92	106	31

^aSchematic structures of Axl mutants are shown in Fig. 3B.

^bCalculated as (MFI of mutant)/(MFI of intact Axl) × 100. MFI of Axl mutants were examined by using flow cytometry with an anti-Axl antibody.

^cThe binding of HIV-LaGPC(Venus) particles to Axl mutant-expressing Jurkat cells was measured by using an HIV p24 antigen ELISA and was calculated as (p24 value in mutant-expressing cells)/(p24 value in intact Axl-expressing cells) × 100.

^dTiters of HIV-LaGPC(Venus) were measured in Axl mutant-expressing Jurkat cells, and ratios were calculated as (titer in mutant-expressing cells)/(titer in intact Axl-expressing cells) × 100.

Similar experiments were performed with Tyro3 (Fig. 3C) in order to understand its role in Lassa virus cell entry. Truncation of the first immunoglobulin domain (Tyro3 del1D) resulted in the detection of increased levels of Tyro3 at the cell surface but dramatically reduced the susceptibility of cells to HIV-LaGPC(Venus) (Table 3). Truncation of the second immunoglobulin domain and two fibronectin type III domains (Tyro3 1D) yielded results similar to those obtained with Tyro3 del1D. Deletion of the two fibronectin type III domains (Tyro3 1D2D) did not affect cell surface detection levels or the susceptibility of cells to HIV-LaGPC(Venus). To assess the binding between the Tyro3 extracellular regions and HIV-LaGPC(Venus) particles, soluble forms of Tyro3 mutants, which were fused to immunoglobulin Fc (Fig. 3D), were used in the Fc chimera-based assay. HIV-LaGPC(Venus) binding to soluble Tyro3 was detected only when both of the two immunoglobulin domains were present (Fig. 2F), indicating the importance of these domains for Tyro3-mediated Lassa virus infection.

Table 3.

Tyro3 mutants and HIV-LaGPC(Venus) titers

Tyro3 forma	Level of detectionb	Titer ratioc
Tyro3 HA	100	100
Tyro3 del1D	605	0
Tyro3 1D	178	0
Tyro3 1D2D	107	95

^aSchematic structures of Tyro3 mutants are shown in Fig. 3C.

^bCalculated as (MFI of mutant)/(MFI of Tyro3 HA) × 100. MFI of Tyro3 mutants were examined by flow cytometry with an anti-HA antibody.

^cTiters of HIV-LaGPC(Venus) were measured in Tyro3 mutant-expressing Jurkat cells, and ratios were calculated as (titer in mutant-expressing cells)/(titer in Tyro3 HA-expressing cells) × 100.

Because the intracellular amino acids of Tyro3 that are important for its activation/signaling remain unknown, the whole intracellular region was truncated (Tyro3 delCT) (Fig. 3C), and this mutant was compared with nontruncated Tyro3 in order to determine the contribution of this region. This truncation resulted in a 6-fold increase in the level of detection of Tyro3 at the cell surface but decreased the susceptibility of cells to HIV-LaGPC(Venus) (Table 4). Thus, as with Axl, Lassa virus likely requires intracellular signals mediated by the tyrosine kinase activity of Tyro3.

Table 4.

Truncation of the whole intracellular region of Tyro3 and HIV-LaGPC(Venus) titers

Tyro3a	Level of detectionb	Titer ratioc
Intact Tyro3	100	100
Tyro3 delCT	614	8

^aSchematic structures are shown in Fig. 3C.

^bCalculated as (MFI of mutant)/(MFI of intact Tyro3) × 100. MFI were examined by flow cytometry with an anti-Tyro3 antibody.

^cTiters of HIV-LaGPC(Venus) were measured in Tyro3 mutant-expressing Jurkat cells, and ratios were calculated as (titer in mutant-expressing cells)/(titer in intact Tyro3-expressing cells) × 100.

To assess the importance of tyrosine kinase activities in Lassa virus infection, Axl- and Tyro3-expressing Jurkat cells were infected with HIV-LaGPC(Venus) in the presence of a tyrosine kinase inhibitor, genistein. HIV-VSVG(Venus) served as a control virus. Genistein reduced the infectivity of HIV-LaGPC(Venus) more efficiently than that of HIV-VSVG(Venus) in both Axl- and Tyro3-expressing cells (Fig. 4), suggesting that tyrosine kinases contribute to Axl- and Tyro3-mediated Lassa virus infection.

Effects of carbohydrate and calcium on C-type lectin-mediated binding and infection.

DC-SIGN and LSECtin are members of the C-type lectin family and bind specific carbohydrates of glycoconjugates in a calcium ion-dependent manner. DC-SIGN mainly recognizes high-mannose-type glycans, and its binding can be blocked by mannan, a polymer of mannose. LSECtin has broad specificity; GlcNAcβ1-2Man is one of its high-affinity ligands (53). We therefore examined whether HIV-LaGPC(Venus) infection of lectin-expressing cells (Fig. 1A) was inhibited by mannan and GlcNAcβ1-2Man. HIV-LaGPC(Venus) infection of DC-SIGN- or LSECtin-expressing cells was dose-dependently inhibited by mannan or GlcNAcβ1-2Man, respectively (Fig. 5A). Next, we performed the cell-based virus binding assay in the presence of these carbohydrates (500 μg/ml). Interestingly, HIV-LaGPC(Venus) binding to DC-SIGN-expressing cells was inhibited by mannan but was not inhibited efficiently by GlcNAcβ1-2Man, whereas its binding to LSECtin-expressing cells was inhibited by the latter carbohydrate but not by the former (Fig. 5B). Furthermore, when the culture medium that was used for incubation and washing in the assay was replaced with PBS containing EDTA, the level of virus binding to DC-SIGN- or LSECtin-expressing cells was lower than that with PBS containing CaCl₂ (Fig. 5C). These results show that binding between Lassa virus GPC and DC-SIGN or LSECtin was both carbohydrate specific and calcium dependent, indicating that both lectins recognize Lassa virus GPC in a manner characteristic of C-type lectins. Thus, Lassa virus may infect cells that express DC-SIGN or LSECtin as a result of the recognition of the carbohydrate on its GPC by these lectins, leading to endocytosis mediated by the internalization signals of these C-type lectins.

Fig 5.

Carbohydrate and ion effects on the interaction between Lassa virus GPC and C-type lectins. (A) Effect of mannan (♢) or GlcNAcβ1-2Man (□) on HIV-LaGPC(Venus) infection of Jurkat cells expressing DC-SIGN (left) or LSECtin (right). The experiment was performed as described for genistein (Fig. 4). (B) Effects of carbohydrates (500 μg/ml) on the binding of HIV-LaGPC(Venus) to Jurkat cells expressing DC-SIGN (left) or LSECtin (right). (C) In the cell-based binding assay, the culture medium, PBS–Ca²⁺, or PBS–EDTA was used to wash Jurkat cells expressing DC-SIGN (left) or LSECtin (right). The amounts of cell-bound pseudotype viruses were measured by use of an HIV p24 antigen ELISA. All data are means ± standard deviations (n = 3).

DG is not involved in the ability of Axl, Tyro3, DC-SIGN, and LSECtin to enhance the efficiency of Lassa virus GPC-mediated virus infection.

We then examined the functional relationship in Lassa virus infection between Axl, Tyro3, DC-SIGN, or LSECtin and DG, a known Lassa virus receptor (10). DG has a unique modification, O-mannosylation, in its extracellular region, and this modification is necessary for DG to mediate Lassa virus infection (34). The anti-DG monoclonal antibody (MAb) IIH6C4, whose reactivity is dependent on O-mannosylation (16, 17, 46), failed to show a positive signal for Jurkat cells by flow cytometry or Western blotting, whereas MAb 43DAG1/8D5, which is specific for the cytoplasmic tail of DG, showed a positive band for Jurkat cells in Western blot analysis (data not shown), consistent with the findings of Rojek et al. (54) that the low infectivity of Lassa GPC pseudotype viruses in Jurkat cells is probably due to the carbohydrate modification status of DG. To test the possibility that the expression of the four molecules identified affects DG O-mannosylation in Jurkat cells, we used flow cytometry to examine the reactivities of MAb IIH6C4 with Jurkat cells upon expression of the molecules. None of the four molecules affected the reactivity of the MAb (Fig. 6A), indicating that these molecules did not affect the glycosylation status of DG. LARGE, a putative N-acetylglucosaminyltransferase whose expression compensates for any lack of O-mannosylation of DG (4), served as a control. LARGE expression in Jurkat cells increased IIH6C4 reactivity (Fig. 6A) and increased the HIV-LaGPC(Venus) titer (approximately 50-fold). Therefore, the increase in the infectivity of pseudotype viruses and authentic Lassa virus effected by Axl, Tyro3, DC-SIGN, and LSECtin (Fig. 1) was not due to enhanced glycosylation of DG in Jurkat cells.

Fig 6.

The effects of Axl, Tyro3, DC-SIGN, and LSECtin are independent of DG. (A) Reactivities of IIH6C4, an O-mannosylation-dependent anti-DG MAb, with Jurkat cells expressing the indicated molecules. An isotype-matched antibody (IgM) (thin lines) was also used. (B) Effects of knockdown (KD) by shRNA for LacZ (shlacZ) (control) or shDG on the infection-enhancing effects of the four molecules indicated. (C and D) Pseudotype virus infection of DG knockout (KO) cells. B11 ES cells (25) were transfected with plasmids to express the indicated proteins and were then used to titrate the pseudotype virus VSV-LaGPC(GFP) (C) or VSV-VSVG(GFP) (D). fCD2 served as a control.

To examine whether the susceptibility-enhancing effects of the four molecules depended on DG, we attempted to knock down DG mRNA by using RNA interference in Jurkat cells and tested the susceptibility of those cells to HIV-LaGPC(Venus). Jurkat cells were infected first with MLV shRNA retrovirus vectors for DG or LacZ (control) and then with a lentivirus vector expressing either a control molecule (fCD2), Axl, Tyro3, DC-SIGN, or LSECtin. Western blot analysis showed that DG protein levels in fCD2-, Axl-, Tyro3-, DC-SIGN-, or LSECtin-expressing cells in which DG RNA had been knocked down were 17%, 20%, 14%, 15%, and 15%, respectively, of the levels in cells in which LacZ had been knocked down. When these cell lines were used to titrate HIV-LaGPC(Venus), results similar to those shown in Fig. 1A were obtained (Fig. 6B), suggesting that the susceptibility-enhancing effects of Axl, Tyro3, DC-SIGN, and LSECtin in Jurkat cells are independent of DG.

To further examine the DG independence of the effects of these four molecules, we used DG knockout ES cells (clone B11) (25). Because our lentivirus expression system did not work efficiently in ES cells (data not shown), the pCAGGS/MCS expression plasmid was used to express the molecules of interest. The plasmid transfection efficiency was reasonable with B11 ES cells; when these cells were transfected with pCAGGS/MCS for expression of the fluorescence protein Venus, they showed 68% positivity in flow cytometry (data not shown). VSV pseudotyped with Lassa virus GPC was used in this experiment. As shown in Fig. 6C, B11 ES cells expressing any of the four molecules showed 2- to 4-fold higher titers of VSV-LaGPC(GFP) than control B11 ES cells. In this experiment, DG expression resulted in titers approximately 10-fold higher than those in control cells (Fig. 6C). Titers of VSV-VSVG(GFP) were similar for these cells (Fig. 6D).

Thus, the four molecules identified in this study participate in Lassa virus entry independently of the known virus receptor DG.

Axl and DG blocking experiments.

We performed blocking experiments with two cell lines, HT1080 (Axl positive, O-mannosylated DG negative) (Fig. 7A, left) and HeLa (Axl positive, O-mannosylated DG positive) (Fig. 7B, left). In HT1080 cells, an anti-Axl polyclonal antibody (10 μg/ml) partially but significantly inhibited HIV-LaGPC(Venus) infection (Fig. 7A, right). In contrast, this antibody did not inhibit HIV-LaGPC(Venus) infection in HeLa cells (Fig. 7B, right). However, DGFc5 (10 μg/ml), a soluble form of O-mannosylated α-DG, partially reduced HIV-LaGPC(Venus) infection in this cell line, and in the presence of DGFc5, an anti-Axl antibody was inhibitory (Fig. 7B, right, compare Control Ab + DGFc5 with Anti-Axl Ab + DGFc5). These findings indicate that both Axl and O-mannosylated DG are Lassa virus receptors but that O-mannosylated DG is more efficient as a receptor than Axl.

Fig 7.

Relationship between Axl and O-mannosylated DG with respect to Lassa virus GPC-mediated infection of the HT1080 (A) and HeLa (B) cell lines. (Left) The expression of Axl or DG was examined by means of flow cytometry with a MAb to Axl or with the O-mannosylation-dependent MAb IIH6C4 against DG (bold lines). Thin lines represent isotype-matched negative controls. (Right) Cells were infected with HIV-LaGPC(Venus) in the presence of the antibodies and/or soluble DG. DGFc5 is full-length α-DG, and DGFc1 is truncated DG. (A) Infectivity was calculated as (titer in the presence of the anti-Axl antibody)/(titer in the presence of the control antibody) × 100. (B) Infectivity was calculated as (titer in the presence of the indicated antibody plus the indicated DG)/(titer in the presence of the control antibody plus DGFc1) × 100. Data are means ± standard deviations (n = 3).

DISCUSSION

Here, we have identified four molecules, Axl, Tyro3, DC-SIGN, and LSECtin, as receptors for Lassa virus. These four molecules function as receptors independently of the known Lassa virus receptor DG (10), although none of them seems to function as efficiently as DG (Fig. 6C and 7B). Thus, if a single cell coexpresses O-mannosylated DG and one of these molecules, infection likely occurs via DG. In cells that do not express DG or that express DG with reduced O-mannosylation, Lassa virus infection may occur via one of these four molecules or via as yet unidentified molecules.

Binding of Lassa virus particles to Axl was detectable in cells expressing Axl (Fig. 2A and C) but not with soluble Axl/Fc (Fig. 2B and D), suggesting that an unidentified molecule(s) on the cell surface may be required for high-affinity binding of Lassa virus particles to Axl. Nonetheless, the cytoplasmic tyrosine kinase activity of Axl, in addition to its two extracellular immunoglobulin domains, is necessary for Lassa virus infection (Table 2). Because Lassa virus enters cells via endocytosis and Axl causes cell membrane ruffling by its natural ligand Gas6 (1), the high-affinity binding of Lassa virus particles to cells may cause intracellular signaling via the tyrosine kinase activity of Axl to induce the endocytotic engulfing of Lassa virus particles. Inhibition by genistein, a tyrosine kinase inhibitor, of Lassa virus GPC-mediated infection in Vero cells, which express Axl (61), is consistent with this notion (64). However, because genistein has multiple activities, including topoisomerase II inhibition, a G₂/M block of the cell cycle, and binding affinities to estrogen receptors, as well as tyrosine kinase inhibition (55), it is not clear whether the inhibitory effect of genistein on Lassa virus GPC-mediated infection in Vero cells (64) occurs solely via inhibition of the tyrosine kinase activity of Axl. Furthermore, the possibility exists that genistein treatment simply results in the downregulation of Axl and/or Tyro3. Studies with more-specific inhibitors will clarify the significance of tyrosine kinase activity in Lassa virus infection.

In contrast to Axl, a soluble chimeric Tyro3/Fc protein showed binding to Lassa virus pseudotype particles (Fig. 2B and D), indicating that Tyro3 binds viral particles directly. The lack or low level of viral binding to the cell surface (Fig. 2A and C) seemed to be due to a low level of Tyro3 expression (Fig. 2E), although this low expression was sufficient with regard to viral infection (Fig. 1). The domains important for Lassa virus GPC-mediated binding and infection were similar for Axl and Tyro3; the two extracellular immunoglobulin domains were necessary for both viral binding and infection (Fig. 2F; Table 3), and the intracellular domain was necessary for infection following binding (Table 4). Although the two extracellular immunoglobulin domains of Axl and Tyro3 are also necessary for binding to their natural ligand, Gas6 (58), we found no homologous sequences or motifs between Lassa virus GPC and the Gas6 proteins (data not shown).

Recently, Morizono et al. (49) found that Gas6 and protein S, another ligand in the TAM family, enhance viral infection in cell culture. These authors investigated the molecular mechanism of this enhancement in detail, in particular that for Gas6, and concluded that Gas6 bridges a virion envelope component, phosphatidylserine, to Axl on the cell surface. In Gas6 and protein S, the amino acid residues important for Axl binding are highly conserved (59). Because a point mutation at amino acid position 59 of Axl, Glu to Arg, dramatically reduces ligand binding (58), we expressed the mutant (Axl E59R; already reported in reference 60) in Jurkat cells and used it to titrate HIV-LaGPC(Venus). No differences in virus titers were observed between Axl HA- and Axl E59R-expressing Jurkat cells (data not shown), suggesting that the effects of Axl and Tyro3 on Lassa virus infection reported here are not likely to be mediated via bridging by the Gas6 or protein S ligand, both of which are present in fetal calf serum.

While Lassa virus GPC-mediated infection was enhanced by any of the four molecules identified, the C-type lectin family members DC-SIGN and LSECtin were more effective at enhancing virus infection than were the TAM family members Axl and Tyro3 (Fig. 1 and 6C). With authentic Lassa virus, a similar tendency was observed; virus titers in cells expressing the C-type lectin family members were higher than those in cells expressing the TAM family members (Fig. 1C, day 1 postinfection). Differences in the binding affinities of the molecules (Fig. 2) may be responsible for this difference, with the following descending order of efficiency for Lassa virus receptors: O-mannosylated DG, DC-SIGN and LSECtin, Axl and Dtk.

Laminin is a DG ligand that blocks the interaction between Lassa virus GPC and DG (35); however, it does not block Lassa virus GPC-mediated infection in Vero cells (34). These findings strongly suggest the existence of an alternative receptor(s) for Lassa virus infection in Vero cells. Vero cells express Axl (61), and we observed a small but significant inhibitory effect of the anti-Axl antibody on Lassa virus GPC-mediated infection in this cell line (∼30% inhibition at 10 μg/ml; P, <0.01 by the Student t test). Therefore, Lassa virus infection in Vero cells may be mediated, at least in part, by Axl. Immunohistochemical analysis with the anti-α-DG antibody IIH6, the reactivity of which depends on O-mannosylation (16, 17, 46), showed no positive signals in human liver samples (69), and another study with the same antibody also failed to detect DG in human hepatocytes (6). These results indicate that human hepatocyte DG does not have efficient O-mannosylated modifications and that therefore, Lassa virus cannot use DG to infect human hepatocytes. However, large amounts of Lassa virus have been detected in the livers of infected individuals, and tissue damage by Lassa virus infection is most prominent in liver parenchyma (15, 45, 65, 67, 68). Thus, Lassa virus likely uses a receptor(s) other than DG to infect hepatocytes. Axl may be one such alternative, since our flow cytometry analysis with an anti-Axl MAb showed Axl expression in human primary hepatocytes, although its signal was weak (data not shown). LSECtin is expressed on liver sinusoidal epithelial cells (38); therefore, it may also be involved in Lassa virus replication in the liver. The glycosylation status of hepatocyte DG in cell culture may be different from that in vivo; Rojek et al. (54) showed high expression of O-mannosylated DG in the hepatocyte line Huh7 by using Western blotting with a IIH6 antibody, and we observed a positive signal in another hepatocyte cell line (HepG2) by use of flow cytometry with the antibody (data not shown). Therefore, caution must be used in interpreting in vitro experiments designed to improve understanding of the molecular basis of Lassa virus tropism.

Patients with severe Lassa fever show markedly reduced platelet function (19). Platelets express members of the TAM family on their surfaces and are activated by Gas6 (22). Since Lassa virus particles bind Axl and Tyro3, two members of the TAM family, Lassa virus infection could alter the physiological function of platelets by interfering with the pathway for platelet activation that involves Axl and Tyro3, resulting in platelet dysfunction.

The four molecules we identified as Lassa virus receptors, Axl, Tyro3, DC-SIGN, and LSECtin, are known to be involved in Ebola virus entry into cells (2, 8, 23, 27, 60, 61). This is not surprising, considering their tropisms; both viruses preferentially target dendritic cells, macrophages, endothelial cells, and the liver (18). Furthermore, because Axl and Tyro3 are involved in endocytosis, and DC-SIGN and LSECtin are C-type lectins that recognize carbohydrate chains, it is possible for these four molecules to be involved in the cell entry of other viruses that possess glycosylated envelope proteins and use endosomes for infection. However, the cell entry mechanisms of Lassa and Ebola viruses are not identical; DC-SIGNR and MGL, which are used by Ebola virus for cell entry (2, 62), are not involved in Lassa virus GPC-mediated infection (present study; also data not shown), and to our knowledge, DG is not involved in Ebola virus infection. Moreover, the cathepsin cellular proteases and the endosomal membrane protein Niemann-Pick C1, which are needed for Ebola virus infection, are not required for Lassa virus infection (11, 13). Interestingly, Gallaher et al. (20) noted the structural similarity between Lassa virus GPC and Ebola virus glycoprotein and suggested a common ancestor for the two glycoproteins. Future studies comparing cell entry mechanisms are required for better understanding of both viruses.