Lassa virus entry requires a trigger-induced receptor switch

See allHide authors and affiliations

Science  27 Jun 2014:
Vol. 344, Issue 6191, pp. 1506-1510
DOI: 10.1126/science.1252480

How Lassa virus breaks and enters

Lassa virus, which spreads from rodents to humans, infecting about half a million people every year, can lead to deadly hemorrhagic fever. Like many viruses, Lassa virus binds to cell surface receptors. Jae et al. now show that to enter a cell, the virus requires a second receptor, this one inside the infected cell. This requirement sheds light on the “enigmatic resistance” of bird cells to Lassa virus observed three decades ago. Although bird cells have the cell surface receptor, the intracellular receptor cannot bind the virus, stopping it in its tracks.

Science, this issue p. 1506


Lassa virus spreads from a rodent to humans and can lead to lethal hemorrhagic fever. Despite its broad tropism, chicken cells were reported 30 years ago to resist infection. We found that Lassa virus readily engaged its cell-surface receptor α-dystroglycan in avian cells, but virus entry in susceptible species involved a pH-dependent switch to an intracellular receptor, the lysosome-resident protein LAMP1. Iterative haploid screens revealed that the sialyltransferase ST3GAL4 was required for the interaction of the virus glycoprotein with LAMP1. A single glycosylated residue in LAMP1, present in susceptible species but absent in birds, was essential for interaction with the Lassa virus envelope protein and subsequent infection. The resistance of Lamp1-deficient mice to Lassa virus highlights the relevance of this receptor switch in vivo.

Lassa virus binds to glycosylated α-dystroglycan (α-DG) at the cell surface to enter cells (1, 2). Over 30 years ago, it was reported that Lassa virus infects a broad suite of cells from different species, with the exception of chicken (3). This was recapitulated by a recombinant vesicular stomatitis virus (VSV) that enters cells using the Lassa virus glycoprotein (rVSV-GP-LASV) (4). Because wild-type VSV was unaffected, this indicated a defect in Lassa glycoprotein (GP)–mediated entry (fig. S1A). Birds, however, generate glycosylated α-DG (5), and the Lassa envelope protein recognized avian α-DG (fig. S1, B and C).

To identify host factors affecting virus entry independent of α-DG binding, we carried out a haploid screen in receptor-deficient cells. For this, we made use of their incomplete resistance phenotype (fig. S2). This showed that neither α-DG nor factors glycosylating α-DG acted as host factors under these conditions (Fig. 1A; fig. S3, A and B; and tables S1 and S2) (6). Instead, we found genes involved in glycosylation, Golgi function, and heparan sulfate biosynthesis. The latter were not identified in wild-type cells (fig. S3C and tables S1 and S3) (4), suggesting that in the absence of α-DG, Lassa virus used heparan sulfate, a commonly used virus attachment factor (7). The lysosomal transmembrane protein LAMP1 and factors involved in N-glycosylation and sialylation, including the α-2,3-sialyltransferase ST3GAL4, stood out in both genotypes. Cells deficient for LAMP1 or ST3GAL4 were comparibly resistant to wild-type Lassa virus as those lacking α-DG (Fig. 1B and fig. S4, A and B). Expression of human but not chicken LAMP1 sensitized chicken fibroblasts to infection with rVSV-GP-LASV (Fig. 1C and fig. S4C) and imposed virus susceptibility in LAMP1-deficient human cells (Fig. 1D and fig. S5). This requirement for LAMP1 was specific for Lassa virus and not shared by the related lymphocytic choriomeningitis virus (fig. S6). Thus, LAMP1 and ST3GAL4 were important for Lassa virus infection independent of α-DG, and host factor function of human LAMP1 was not shared by its chicken ortholog.

Fig. 1 Human LAMP1 is an α-DG–independent host factor for Lassa virus and bypasses an infection barrier in avian cells.

(A) Haploid genetic screen for host factors required for infection with rVSV-GP-LASV in cells lacking α-DG. The y axis indicates the significance of enrichment of gene-trap insertions in particular genes as compared with nonselected control cells. Solid circles represent genes, and their size corresponds to the number of insertion sites identified in the virus-selected cell population. Hits were colored if they passed the statistical criteria described in (6). Significant hits were grouped by function horizontally, and data are displayed until –log(P) = 0.01. (B) HAP1 cell lines with nuclease-generated mutations in the corresponding genes were exposed to wild-type Lassa virus and stained with antibodies specific for viral antigens to measure infected cells. LAMP1-deficient cells were complemented with human LAMP1 cDNA. (C) Chicken fibroblasts were transduced with retroviruses expressing chicken (c) or human (h) LAMP1 and challenged with rVSV-GP-LASV. Average number (±SD) of infected cells per field (eGFP-positive) is indicated. (D) Wild-type or LAMP1-deficient HAP1 cells transduced with retroviruses expressing cLAMP1 or hLAMP1 (L1) were exposed to rVSV-G or rVSV-GP-LASV. Percentage (±SD) of infected cells (expressing eGFP) is indicated. Scale bars, 50 μm.

Because LAMP1 deficiency neither causes pronounced phenotypes in mice (8) nor detectably impaired the endo-lysosomal compartment in cultured cells (fig. S7), we asked whether the Lassa virus envelope protein could bind to LAMP1. As the majority of LAMP1 is localized in the acidic interior of lysosomes (9, 10), these experiments were carried out at neutral and acidic pH. Immobilized Flag-tagged Lassa-GP bound α-DG at neutral pH, but this interaction was lost at acidic pH, at which Lassa-GP instead strongly bound to LAMP1 (Fig. 2A and fig. S8, A and B). Lassa-GP molecules that had previously bound α-DG were capable of subsequent binding to LAMP1 (fig. S8C). Likewise, intact virions were captured by the luminal region of LAMP1 at acidic but not at neutral pH (fig. S9). Last, this interaction was observed for both human and mouse LAMP1 but not chicken LAMP1 or human LAMP2 (Fig. 2B and fig. S10).

Fig. 2 Lassa-GP undergoes a pH-induced switch to engage LAMP1.

(A) Flag-tagged Lassa-GP was immobilized on beads and incubated with cell lysates from human embryonic kidney (HEK) 293T cells at the indicated pH. Bound proteins were subjected to immunoblot analysis, and uncoated beads served as a control. IP, immunoprecipitation. (B) Flag-tagged Lassa-GP was immobilized on beads and incubated with lysates from human, mouse, and chicken cells at the indicated pH. Bound proteins were subjected to immunoblot analysis. (C) Electron micrographs of wild-type and LAMP1-deficient HEK-293T cells that were infected with rVSV-GP-LASV. LAMP1-deficient cells show an accumulation of the bullet-shaped viral particles (arrows) in intracellular vesicles. Scale bars, 100 nm. (D) Flag-tagged Lassa-GP was immobilized on beads and incubated with purified LAMP1-Fc at the indicated pH. Complexes (IP) were precipitated and subjected to immunoblot analysis. The supernatant (Sup) was analyzed for the release of Lassa-GP1. (E) LAMP1-deficient (top) or LAMP1d384-expressing HEK-293T cells (bottom) were transfected with expression vectors for Lassa-GP and GFP and exposed to pH 5.5. Cell boundaries were visualized with fluorescent wheat germ agglutinin (red). Large, homogenous green fluorescent area results from Lassa-GP–induced syncytia formation (yellow outline). Scale bars, 50 μm.

Virus particles containing enhanced green fluorescent protein (eGFP) fused to the VSV matrix protein (MeGFP, allowing direct visualization of incoming fluorescent virions) were internalized in cells lacking LAMP1 or ST3GAL4 (fig. S11) but accumulated in vesicles of LAMP1-deficient cells (Fig. 2C). In wild-type cells, fusion of viral and cellular membranes leads to release of MeGFP protein into the cytoplasm (11), but in LAMP1-deficient cells, MeGFP remained localized to vesicles (figs. S12 and 13), suggesting that the association of Lassa-GP with LAMP1 precedes membrane fusion. In agreement with this, LAMP1 interacted with Lassa-GP in a prefusion configuration when the GP1 subunit of the viral envelope protein is still part of the complex (12, 13) but not when GP1 was fully released from GP2 by low pH (Fig. 2D and fig. S14). To test whether Lassa-GP–mediated membrane fusion was affected by LAMP1, we carried out cell-cell fusion experiments in the presence of a LAMP1 mutant that localizes to the cell surface (LAMP1d384) (fig. S15A) (9). Expression of this mutant led to increased syncytia formation as a consequence of membrane fusion (Fig. 2E and fig. S15, B and C). This activity of LAMP1 was independent of α-DG (fig. S15D). Thus, Lassa virus likely engages α-DG at the cell surface, enters the endocytic pathway and binds to LAMP1 upon reaching the acidic interior of lysosomes, before membrane fusion.

To test this model, we engineered an artificial virus infection scenario. Cells in which α-DG was knocked out that expressed LAMP1d384 were largely resistant to rVSV-GP-LASV under normal conditions, but lowering the extracellular pH during virus exposure led to productive infection (fig. S16, A to E). Lassa virus entry normally depends on acidification of endosomes (14) and is sensitive to bafilomycin (15). The engineered entry route was, however, bafilomycin-insensitive (fig. S16, F and G). Thus, the requirement for α-DG could be bypassed by rerouting LAMP1 to the cell surface and triggering binding to Lassa-GP.

Besides LAMP1, the screens identified the α-2,3-sialyltransferase ST3GAL4 as an α-DG–independent host factor. Because targets modified by this enzyme could display genetic interactions, we searched for host factors depending on ST3GAL4. ST3GAL4-deficient cells were mutagenized and selected with rVSV-GP-LASV. Like experiments in wild-type cells, this screen identified DAG1 and its modifiers (4). As expected, the disrupted ST3GAL4 locus did not act as a host factor under these conditions, but neither did LAMP1 (Fig. 3A; figs. S3, C and D, and S17A, and tables S1 and S4). Therefore, we investigated a putative biochemical connection between them. LAMP1 is glycosylated (16) with both N- and O-glycans (17). LAMP1 derived from ST3GAL4-deficient cells showed reduced binding to lectins that preferentially capture α-2,3–linked sialic acid (fig. S17, B and C) (18) and lost its ability to bind to Lassa-GP (Fig. 3B). Thus, LAMP1 was only able to act as a host factor in the context of ST3GAL4 proficiency.

Fig. 3 Binding of Lassa-GP to LAMP1 depends on ST3GAL4, and LAMP1-Asn76 is critical for host factor function.

(A) Haploid genetic screen pointing out genetic interactions between ST3GAL4 and other Lassa entry factors. ST3GAL4-deficient cells were mutagenized and exposed to rVSV-GP-LASV. Gene-trap insertion sites were mapped in the resistant cell population, and data was analyzed as in Fig. 1A. (B) Flag-tagged Lassa-GP was immobilized on beads and incubated with cell lysates from wild-type and ST3GAL4-deficient HAP1 cells at pH 5.5. Bound proteins were subjected to immunoblot analysis. (C) Wild-type (WT) and LAMP1-deficient (ΔL1) HAP1 cells complemented with cDNAs expressing the distal domain of LAMP1 containing mutations at the indicated glycosylation sites were exposed to rVSV-GP-LASV. Percentage (±SD) of infected cells (eGFP-positive) is shown. (D) Comparison of LAMP1 polypeptides from indicated species highlights Asn76 as a marker of susceptibility to Lassa virus infection. (E) Flag-tagged Lassa-GP was immobilized on beads and incubated with lysates from LAMP1-deficient HEK-293T cells expressing human LAMP1 or “chickenized” LAMP1 carrying the Asn76Ser substitution at the indicated pH.

LAMP1 consists of three luminal domains: a membrane-proximal domain, an O-glycosylated hinge region, and a distal domain. The distal domain contains 11 N-glycosylation sites (UniProt, P11279) and was sufficient to support rVSV-GP-LASV infection by itself (fig. S18). Reconciling that genes for N-glycosylation and sialylation acted as host factors and that LAMP1 derived from ST3GAL4-mutant cells was not recognized by Lassa-GP, we speculated that one of these glycosylation sites was important for host factor function. Indeed, we found that only Asn76 was essential for VSV-GP-LASV infectivity (Fig. 3C). This residue is present in LAMP1 from species susceptible to Lassa virus but absent in birds (Fig. 3D) (16). Substitution of this amino acid in human LAMP1 for the respective avian residue (Asn76Ser) was sufficient to block infection (fig. S19) and binding to Lassa-GP (Fig. 3E). Reciprocally, insertion of a region surrounding human Asn76 into chicken LAMP1 converted the avian protein into a host factor (fig. S20). Thus, we identified the sialyltransferase ST3GAL4 as a critical enzyme required for LAMP1 to function as a host factor and mapped the interaction between sialylated LAMP1 and Lassa-GP to a single glycosylated amino acid present in sensitive species but absent in birds.

Because Lassa virus has a rodent reservoir, we examined whether LAMP1 is required for the propagation of wild-type Lassa virus in vivo. After intraperitoneal injection, virus was cleared in mice in which Lamp1 was knocked out, whereas infection was manifest in all organ samples taken from wild-type or heterozygous animals (Fig. 4A and fig. S21).

Fig. 4 Lamp1 knockout mice are resistant to wild-type Lassa virus, and the receptors require distinct glycosyltransferases.

(A) Lassa virus propagation in Lamp1+/+, Lamp1+/−, and Lamp−/− mice. Mice were injected intraperitoneally with wild-type Lassa virus, and viral titers (y axis, plague-forming units/mL) were determined after 6 days in the indicated tissues. The horizontal line marks the detection limit. (B) Flag-tagged Lassa-GP was immobilized on beads and incubated with cell lysates from wild-type, TMEM5-, or ST3GAL4-deficient cells at the indicated pH. The glycosyltransferase TMEM5 is needed to generate an epitope on α-DG that is recognized by Lassa-GP (4). Bound proteins were subjected to immunoblot analysis. Asterisk indicates nonspecific background band.

Here, we have shown that Lassa virus entry requires a pH-regulated engagement of α-DG and LAMP1, both of which need to be glycosylated. However, the glycan structures that are needed for host factor function are unrelated and constructed by distinct enzymes (Fig. 4B and fig. S22). Unlike in rodents (19), the human upper airway mainly contains α-2,6–linked sialic acid moieties rather than α-2,3–linked sugars (20) generated by enzymes such as ST3GAL4. It has been proposed that this is an adaptation to evade pathogens like avian influenza (21), but it may also limit human-to-human spread of Lassa virus (22). Lassa virus has been described as a “late-penetrating” virus (23) that requires low pH (24). Our findings rationalize these observations and emphasize the emergence of intracellular receptors for virus entry.

Supplementary Materials

Materials and Methods

Figs. S1 to S22

Tables S1 to S4

References (2544)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank T. Sixma, A. Perrakis, E. von Castelmur, D. Lefeber, and members of the Brummelkamp group for dicussion; M. Rusch for mouse breeding; S. Kunz for a plasmid encoding Lassa-GP; E. Ollmann-Saphire for an Fc-fusion vector; R. Schoepp for GP1 antibodies; and M. Verheije for DF1 cells. This work was supported by, Nederlandse Organisatie voor Wetenschappelijk Onderzoek Vidi grant 91711316, and European Research Council (ERC) Starting Grant (ERC-2012-StG 309634) to T.R.B.; Deutsche Forschungsgemeinschaft (DFG SPP1580 and GRK1459) to P.S.; and NIH grants AI081842 and AI109740 to S.P.W. J.M.D was supported by the Defense Threat Reduction Agency (CB3947). The HAP1 cells that were used are distributed under a materials transfer agreement. Opinions, interpretations, conclusions, and recommendations are those of the author and are not necessarily endorsed by the U.S. Army. T.R.B. is a cofounder and shareholder of Haplogen GmbH, a company involved in haploid genetics. Sequencing data are accessible at (accession SRP041566).
View Abstract

Navigate This Article