Discovery of a proteinaceous cellular receptor for a norovirus

Main text

Noroviruses (NoV) are non-enveloped positive-sense RNA viruses (1, 2). Due to the strict species tropism of viruses in the NoV genus and the lack of robust replication of human noroviruses (HNoV) in animal models, murine norovirus (MNoV) has emerged as a model system to uncover basic mechanisms of NoV biology in vitro and in vivo (1, 3–8). MNoVs can establish persistent enteric infection enabling studies of the interplay between viral persistence, resident enteric microorganisms, and the host immune system (4–6). The capacity of HNoV and MNoV to bind cells, and the susceptibility of humans to HNoV, have been linked to expression of cell surface and secreted carbohydrates (1, 9–13), while another member of the Caliciviridae utilizes a proteinaceous receptor (14, 15). Host factors including receptor(s) required for NoV infection and pathogenesis have largely defied molecular identification; their discovery would aid in understanding mechanisms of NoV replication, vaccination, species tropism, and enteric viral persistence. To identify host molecules required for MNoV infection, we undertook an unbiased forward genetic approach (Fig. S1).

MNoV replicates and induces cell death in murine macrophage-like cells, including the microglial BV2 cell line, allowing identification of genes essential for MNoV replication using CRISPR-Cas9 technology. We introduced four independent genome-wide subpools of single guide RNAs (sgRNAs) of the murine Asiago library into BV2 cells stably expressing Cas9 and then infected the cells with MNoV strains that either cause acute systemic infection (MNoV^CW3) or persistent enteric infection (MNoV^CR6) in mice (16–18). sgRNA sequences from the surviving cells were sequenced and analyzed using the STARS gene-ranking algorithm (Fig. 1A) (17). sgRNAs targeting Cd300lf (CLM-1, CMRF35, MAIR-V, LMIR3), a cell surface immunoglobulin (Ig) domain-containing molecule within a family of proteins involved in binding lipids, were most significantly enriched for both MNoV strains (Fig. 1A and Table S1–3) (19). We generated two independent BV2ΔCD300lf clones; the growth of MNoV^CW3 and MNoV^CR6 was abolished in both clones while the replication of other viruses was unaffected (Fig 1B and S2). We also validated several additional molecules predicted to be important for MNoV-induced cell death (Fig. 1A and S3). Taken together, these data provide a systematic overview of the molecules required for NoV replication in these cells.

We selected CD300lf for further analysis because of its cell surface expression and the importance of viral receptors for conferring permissiveness for viral replication and species tropism. Transfection of MNoV^CW3 RNA into BV2ΔCD300lf cells was sufficient to rescue MNoV^CW3 production demonstrating, that CD300lf is essential for viral entry (Fig. 2A). Preincubation of cells with a polyclonal antibody targeting CD300lf (α-CD300lf) blocked MNoV^CW3-induced cytopathic effects in BV2 cells (Fig. 2B). Similarly, incubating MNoV^CW3 with recombinant CD300lf ectodomain (sCD300lf) neutralized MNoV^CW3-induced cytopathic effects while Annexin V or phosphoserine treatment had no effect (Fig. 2C and S4). Similar results were obtained for infection of the B-cell line M12, bone marrow-derived dendritic cells (BMDC), and bone marrow-derived macrophages (BMDM) indicating the essential role of CD300lf in multiple cell types (Fig. 2D). Taken together these data indicate that interactions between MNoV and CD300lf are essential for MNoV infection.

To directly test if CD300lf functions as a binding receptor for MNoV, we analyzed the attachment of MNoV^CW3 to BV2 cells on ice. BV2ΔCD300lf cells were impaired in MNoV^CW3 binding (Fig. 2E). Additionally, treating BV2 cells with α-CD300lf, or preincubating virus with sCD300lf, reduced MNoV binding (Fig. 2E). In BV2ΔCD300lf cells, binding of MNoV^CW3 was not further inhibited with α-CD300lf or by premixing virus with sCD300lf (Fig. 2E). Taken together, these data indicate that CD300lf mediates viral binding and is a functional receptor for MNoV.

Previous reports have suggested carbohydrates facilitate the binding of MNoV and HNoV to host cells and control the susceptibility of humans to HNoV infection (1, 9–11). Therefore, we assessed the relative contribution of carbohydrates to MNoV attachment and infection. Surprisingly, mice deficient in Fut2, which controls histo-blood group antigen (HBGA) secretor status, had similar MNoV loads compared to wild-type, littermate controls (Fig. S5A). Also, treating cells with the mannosidase I inhibitor kifunensine significantly reduced cell surface carbohydrates, but did not significantly alter MNoV^CW3 binding to cells (Fig. S5B and S5C). These data suggest that MNoV binding and infection is dependent upon Cd300lf but not on protein-associated glycans or secretor status within the sensitivity of these assays.

In addition to CD300lf, we discovered that efficient MNoV binding to cells requires a cofactor present in serum (Fig. 2F). The cofactor is present in delipidated serum and is resistant to proteinase K and heat denaturation (95°C; Fig. 2F). Size fractionation of serum indicates that the cofactor is present in fractions with an average molecular weight of less than 5,000 Da (Fig 2F). Thus, the serum cofactor is consistent with a small non-proteinaceious heat stable molecule that interacts with MNoV and/or CD300lf to facilitate cellular binding.

Next, we sought to assess the physiologic relevance of MNoV and CD300lf interactions. MNoV isolated from the spleens of MNoV^CW3-infected mice remained sensitive to sCD300lf inhibition, indicating that sCD300lf neutralizes MNoV regardless of the source of the virus (Fig. 3A). To test the role of CD300lf and MNoV interactions in vivo, we incubated MNoV^CW3 with either sCD300lf or a control protein prior to infection of Stat1^−/− mice, which succumb to MNoV^CW3 infection. In a dose-dependent manner, sCD300lf preincubation was able to protect from MNoV-induced lethality (Fig. 3B and 3C). Finally, we generated Cd300lf^−/− mice to test the in vivo role of CD300lf in MNoV infection. Cd300lf^−/− mice were resistant to MNoV^CR6 infection compared to littermate controls (Figure 3D). These data demonstrate that CD300lf is the primary physiological receptor for MNoV in vivo.

MNoV replicates in murine dendritic cells, macrophages, and B cells, but not in epithelial cells or human cells due to a restriction at viral entry (20, 21). Therefore, we tested if expression of murine CD300lf was sufficient to confer susceptibility of HeLa cells to MNoV. As expected, HeLa cells transfected with a control plasmid were unable to support MNoV replication (Fig. 4A). In contrast, HeLa cells expressing murine but not human CD300lf were susceptible to MNoV (Fig. 4A). These data indicate that CD300lf expression is sufficient for MNoV growth in human cells and suggest that differences between human and murine CD300lf could contribute to MNoV species restriction.

In mice there are eight CD300 family members (22). Therefore, we sought to determine whether other CD300 molecules are capable of functioning as an MNoV receptor. After transfection into HeLa cells only CD300lf and CD300ld conferred susceptibility to MNoV in HeLa cells while the expression of other CD300 family members or the unrelated phosphatidylserine binding protein Tim1 was unable to support MNoV infection (Fig. 4B and S6) (23). We further confirmed that expression of CD300lf and CD300ld but not CD300lh was sufficient for viral replication in human cells (Fig. 4A). However, in contrast to murine CD300lf the recombinant ectodomains of murine CD300ld, CD300lh, or human CD300lf failed to neutralize viral infection (Figure 4C). Additionally, BV2ΔCD300ld cells are susceptible to MNoV infection suggesting that CD300ld can act as an MNoV receptor when ectopically expressed, but it is not a universal requirement (Fig. S7). However, we cannot exclude the possibility that CD300ld may play a role in viral tropism in some circumstances in vivo. Experiments in Cd300lf^−/− mice indicate that this putative role is not essential for intestinal infection or shedding.

We next sought to define the mechanism for MNoV entry via CD300lf and to determine how MNoV discriminates between mouse and human CD300lf proteins. Importantly, the intracellular domain of CD300lf was not required to make HeLa cells susceptible, indicating that species tropism is determined by the ectodomain (CD300lfΔCT, Fig. 4D). We determined the crystal structure of the CD300lf ectodomain at 1.6Å resolution (Fig. 4E and Table S4). Interestingly, density corresponding to a bound HEPES molecule, present in the purification buffers, and a coordinated metal were visible in a surface cleft formed between the CDR3 loop and the β-hairpin turn that connects the C-C′ β-strands. CD300lf has been reported to mediate the phagocytosis of apoptotic cells through the calcium-dependent binding of lipids (24–26). HEPES has chemical resemblance to a phospholipid headgroup (Fig. S8). Our structure also reveals a metal coordinated primarily by CD300lf Asp98 and two CDR3 loop carbonyl oxygens. While mutation of murine CD300lf Asp98 has been shown to disrupt apoptotic cell surface binding, this mutant still allows MNoV^CW3 infection of HeLa cells, suggesting that viral entry can occur in the absence of bound metal (Fig. 4D and S8) (25, 27). The Ig-domains of murine and human CD300lf share 59% sequence identity, and structurally the largest variation occurs in CDR3 and the CC′-loop (Fig. S9 and S10) (27). Interestingly, individual substitutions of human CC′-loop and CDR3 into murine CD300lf diminished and abolished MNoV^CW3 infection of HeLa cells, respectively (Fig. 4D and 4F). Additionally, purified recombinant proteins harboring CC′-loop and CDR3 human sequences failed to neutralize MNoV infection (Fig. 4C and S11). However, the reciprocal CDR3 mutation (murine CDR3 into human CD300lf) was not sufficient for MNoV infection (Fig. 4D). Independently comparing murine CD300lf and CD300lh also indicated that the CD300lf CC′-loop and CDR3 sequences are necessary but not sufficient for receptor utilization by MNoV (Fig. S12). These data provide a framework for understanding how MNoV discriminates between CD300 family members.

Our work establishes that CD300lf is a functional MNoV receptor that mediates binding to the cell surface and is both necessary and sufficient for viral entry and replication in vitro and in vivo. Because MNoV serves as a model system for understanding how viruses persist and shape the immune system, the modulation of receptor availability either genetically or chemically may foster understanding of immunomodulation, persistence, and tropism of MNoV. This work also enables the future study of MNoV replication in human cells, which may uncover novel mechanisms of viral replication and pathogenesis and allow a direct identification and mechanistic dissection of the cellular factors required for NoV replication across species. Additionally, our work has implications for understanding HNoV infections. HNoV binds to HBGA and susceptibility to HNoV is correlated with host HBGA status (12, 13). These reports are the foundation for the hypothesis that glycans are HNoV receptors (9). However, HBGA alone cannot, to date, explain species tropism or the entry barrier for HNoV. In contrast, our data indicate that murine CD300lf is sufficient to explain tropism for MNoV and more broadly suggest the possibility that other NoVs utilize proteinacious receptors in addition to small molecule cofactors that are present in serum or other biological fluids. It is intriguing that CD300 molecules can bind a range of host lipids and, additionally, bacterial products (22, 28). It may therefore be that NoV cell and tissue tropism is determined in a combinatorial fashion by proteinaceous receptors interacting with permissiveness cofactors present at different sites.

Supplementary Material