Distinct Roles for Sialoside and Protein Receptors in Coronavirus Infection.

Coronaviruses (CoVs) are common human and animal pathogens that can transmit zoonotically and cause severe respiratory disease syndromes. CoV infection requires spike proteins, which bind viruses to host cell receptors and catalyze virus-cell membrane fusion. Several CoV strains have spike proteins with two receptor-binding domains, an S1A that engages host sialic acids and an S1B that recognizes host transmembrane proteins. As this bivalent binding may enable broad zoonotic CoV infection, we aimed to identify roles for each receptor in distinct infection stages. Focusing on two betacoronaviruses, murine JHM-CoV and human Middle East respiratory syndrome coronavirus (MERS-CoV), we found that virus particle binding to cells was mediated by sialic acids; however, the transmembrane protein receptors were required for a subsequent virus infection. These results favored a two-step process in which viruses first adhere to sialic acids and then require subsequent engagement with protein receptors during infectious cell entry. However, sialic acids sufficiently facilitated the later stages of virus spread through cell-cell membrane fusion, without requiring protein receptors. This virus spread in the absence of the prototype protein receptors was increased by adaptive S1A mutations. Overall, these findings reveal roles for sialic acids in virus-cell binding, viral spike protein-directed cell-cell fusion, and resultant spread of CoV infections.IMPORTANCE CoVs can transmit from animals to humans to cause serious disease. This zoonotic transmission uses spike proteins, which bind CoVs to cells with two receptor-binding domains. Here, we identified the roles for the two binding processes in the CoV infection process. Binding to sialic acids promoted infection and also supported the intercellular expansion of CoV infections through syncytial development. Adaptive mutations in the sialic acid-binding spike domains increased the intercellular expansion process. These findings raise the possibility that the lectin-like properties of many CoVs contribute to facile zoonotic transmission and intercellular spread within infected organisms.

could be attributed to S proteins. Intriguingly, JHM HEϪ -CoV virus particles agglutinated human erythrocytes, while the related A59-CoV strain did not (Fig. 1A), even at particle-associated S-protein input levels slightly exceeding the corresponding JHM HEϪ -CoV (Fig. S1B). These findings suggested that JHM uniquely engages host sialic acids.
To validate the presumed JHM virus spike-host cell sialic acid interaction, we developed an MHV-based virus-like particle (VLP) system, in which VLPs contained reporter proteins that can be tracked during virus-cell binding and entry (Fig. 1B). In this system, VLPs were produced from 293T cells cotransfected with MHV structural genes S, E, M, and N (38)(39)(40). The N genes were modified by in-frame fusion with dual-split protein (DSP) encoding sequences, the two of which complement to form Renilla luciferase (Rluc) and green fluorescent protein (GFP) reporters. This allowed for the release of VLPs that were intrinsically Rluc and GFP positive.
We noted that MHV VLPs incorporated intact JHM spike proteins at levels that were similar to authentic viruses. This was in contrast to JHM spikes that were incorporated into vesicular stomatitis virus (VSV)-based pseudoparticles, most of which were degraded (Fig. 1C). These findings indicated that the VLPs structurally resemble authentic virus structures, at least with respect to spike stability and particle incorporation. As spike stability and multivalent presentation are frequently required to detect sialate interactions (21,41), we considered the VLPs to be well-suited for receptor binding studies. Additionally, VLPs are equally produced either with or without incorporated S proteins (42), making it possible to use VLPs in quantifying S-protein-dependent virus-cell binding processes.
VLPs, either with or without JHM or A59 S proteins, were applied to mCEACAMnegative HeLa cells at equivalent (Rluc) multiplicities. Under these conditions, the S-protein levels on the A59 VLPs exceeded those of JHM VLPs (Fig. 1D). Yet, in agreement with the hemagglutinating effects of the viruses, the JHM S proteins increased VLP-HeLa cell binding above background levels, while the A59 S proteins did not (Fig. 1E). Notably, the JHM VLPs did not bind to HeLa cells that were pretreated with neuraminidase ( Fig. 1F, left), suggesting that the VLP binding was dependent on cellular sialic acids.
Surprisingly, VLP binding to cells was not increased by expression of the MHV protein receptor mCEACAM, and furthermore, VLPs did not bind to mCEACAMexpressing cells that had been pretreated with neuraminidase ( Fig. 1F, right). Together, these data suggest that, during MHV-JHM infection, the initial viral binding process is mediated by S-proteinϪsialic acid interactions, not by the protein receptor mCEACAM.
Virus binding to sialic acids facilitates protein receptor-mediated entry. We next aimed to identify the biological importance of the JHM spike-sialate interaction. We reasoned that this interaction could be sufficient for viral entry or may instead be an early step preceding viral engagement with protein (mCEACAM) receptors. To assess virus entry, we utilized VLPs containing N proteins fused to only one of the DSP fragments (DSP [8][9][10][11] ), such that VLP entry into target cells containing DSP 1-7 fragments could be measured by reporter complementation into active Rluc or GFP moieties ( Fig. 2A). We compared VLP entries in HeLa and HeLa-mCEACAM cells and found that Rluc entry signals required mCEACAM presence (Fig. 2B). The entry signals were elevated by the presence of TMPRSS2 (Fig. 2B), a known viral fusion-triggering protease (11,12,(43)(44)(45)(46)(47)(48). This facilitating effect of TMPRSS2 supported the contention that the VLP-based entry assay faithfully reflected authentic CoV entry processes. In conjunction with VLP binding data, these findings suggest that JHM-CoV cell entry is a multistep process, with initial virus-sialate binding and subsequent mCEACAM receptor-driven entry.
We next determined whether virus binding to sialic acids elevates protein receptordependent virus entry. Here, authentic JHM HEϪ -CoV particles were bound for 2 h to DBT (murine), HeLa-mCEACAM (human), or BHK (hamster) cells, with or without prior cell exposure to neuraminidase. Infection was then measured by quantifying viral firefly luciferase (Fluc) reporter expression. The neuraminidase pretreatments decreased (E) HeLa cells were formalin fixed, rinsed, and incubated for 2 h at 4°C with VLPs bearing A59 or JHM S. Rluc input multiplicities were equal for all VLP inoculations. Media were removed, cells were lysed, and cell-associated Rluc activities were quantified. Data are presented after subtracting background ("No S") Rluc-positive (Rluc ϩ ) VLP levels, with each data point representing averages from independent experiments (N ϭ 8 [A59] and 7 [JHM]; n ϭ 4 technical replicates per experiment). Error bars present standard errors (SE) from the mean. Statistically significant deviations from background ("No S") binding were assessed by unpaired Student's t test and are indicated as follows: ns, not significant; *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ****, P Ͻ 0.0001. (F) HeLa or HeLa-mCEACAM cells were fixed and treated with vehicle (PBS) or neuraminidase (NA) for 3 h at 37°C. JHM-CoV VLPs were then added for 2 h at 4°C, cell-associated Rluc activities were quantified, and data are presented after subtracting background ("No S") Rluc ϩ VLP levels. Error bars present standard errors (SE) from the mean. Statistically significant deviations were assessed by unpaired Student's t test and are indicated as follows: ns, not significant; *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ****, P Ͻ 0.0001.

Qing et al.
® JHM HEϪ -CoV entry by 38% (DBT), 38% (HeLa-mCEACAM), and 54% (BHK) (Fig. 2C). After neuraminidase pretreatments, the cells were similarly resistant to IAV pp transduction, but not VSV G protein (VSV G)-containing pseudoparticle (VSVG pp) transduction (Fig. 2C). Partial resistance to IAV pp, a known sialic acid-requiring virus (49)(50)(51), is best explained by incomplete removal of sialic acids by neuraminidase. We infer that the partial neuraminidase resistance of CoV VLPs is similarly explained by incomplete sialic acid removal. These results support a model in which initial virus binding to sialic acids facilitates protein receptor-mediated entry.   Sialic acid receptors promote JHM viral spread without requiring mCEACAM protein receptors. CoV infections spread through the canonical process of progeny virus release and reinfection, and also through fusion, i.e., syncytial spread, with uninfected cells. JHM S proteins can uniquely mediate syncytia in several conditions, even when mCEACAM protein receptors are absent (33,35). To determine whether sialic acids operate independently of mCEACAM receptors in JHM cell-cell membrane fusion and syncytial development, we cultured JHM CoV-infected mouse (DBT) cells with mCEACAM-negative human (HeLa) cells that separately harbored either DSP 1-7 or DSP [8][9][10][11] , such that HeLa cell-cell fusions would permit DSP complementation into active Rluc (Fig. 3A). Using live-cell Rluc substrate, the development of active Rluc was measured over time, and in the presence of increasing neuraminidase doses. The results (Fig. 3B) revealed neuraminidase dose-dependent reductions in JHM CoVinduced syncytia, up to 7.4-fold at the highest neuraminidase dose (2 U/ml). Similar, but less pronounced reductions were observed in parallel cultures of CEACAM-positive HeLa cells (Fig. 3C). Of note, the neuraminidase treatments significantly reduced cell-surface sialic acids, as revealed by adherence of fluorescent wheat germ agglutinin (WGA) to treated cells (Fig. S2). From these findings, we conclude that sialic acids can enable a CoV S-mediated cell-cell fusion process without requiring a prototypic proteinaceous receptor.
A JHM-CoV spike mutation increases mCEACAM-independent cell binding and membrane fusion. To date, analyses of the multidomain CoV S proteins have revealed that the S1A domains interact with sialic acids (18,19,21,41) and that most of the S1B domains bind to protein receptors (22,52,53). The exceptions are the MHV beta-CoVs, which have S1A domains that bind to protein (CEACAM) receptors (25). From these findings, we inferred that the JHM-CoV S1A domains contain a novel dual-receptor binding capability, able to bind both sialic acid and CEACAM receptors. Sialic acidbinding sites on beta-CoV S proteins were originally inferred from mutagenesis studies (19), and most recently, from structural resolution of S proteins in complex with sialosides (18,54). Mutations in the inferred site did decrease sialoside binding (19,41), even though they are distal from the structurally resolved sialoside binding grooves (18,54). Here, we noted a mutation in the original inferred site. Among a stretch of four residues important for sialate-binding activity of bovine CoV (B-CoV) S1A, JHM differed only at residue 176, where there is a Gly on a divergent loop in place of the orthologous B-CoV Glu170 (Fig. 4A). To evaluate the importance of this divergence to sialic acid binding, we constructed a G176E mutant JHM-CoV S protein. G176E S-protein expression, proteolytic processing, and VLP incorporation were all comparable to the wild type (WT) (Fig. 4B). Notably, VLPs with G176E spikes bound more tightly to HeLa cells, and G176E spike-containing VLP binding was suppressed by neuraminidase pretreatment of the cells (Fig. 4C). To correlate this increased cell binding with membrane fusion, we compared the syncytial potency of JHM wild-type and G176E mutant S proteins. Cell-cell fusions were measured by DSP 1-7 ϪDSP 8 -11 complementation (Fig. 4D), visualized as GFP-positive syncytia (Fig. 4E), and quantified by measuring Rluc signals (Fig. 4F). Of note, when mCEACAM receptors were present, synyctia induced by wild type and mutant S proteins developed comparably, but in the absence of mCEACAM, the G176E spikes were significantly more robust in their syncytiuminducing activity (Fig. 4F). We concluded that a single point mutation in the JHM-CoV S1A domain increases the cell binding and membrane fusion activities of S proteins, independent of the prototype JHM protein receptor.

MERS-CoV spikes mediate cell fusion without requiring hDPP4 protein receptors.
MERS-CoV S1B domains bind to host protein (hDPP4) receptors (22)(23)(24), and MERS-CoV S1A domains bind to host sialic acids (21). This raised the question of whether MERS-CoV spikes can catalyze cell fusions after the binding of their S1A domains to cells, and independently of high-affinity protein receptors, similar to the JHM-CoV spikes. To address this question, MERS-CoV S-mediated cell-cell fusions were measured by DSP 1-7 ϪDSP 8 -11 complementation (Fig. 5A) and quantified by measuring Rluc signals (Fig. 5B). Of note, syncytial development mediated by MERS-CoV S proteins did not require hDPP4 but did require the S-protein-activating protease human TM-PRSS2 (hTMPRSS2) (Fig. 5B). The catalytically inactive TMPRSS2(S441A) did not facilitate cell-cell fusion (Fig. 5B). The MERS S-mediated cell fusions contrasted with those generated by HCoV-229E S proteins, which do not bind sialic acids (21). 229E S-mediated fusions absolutely required target-cell human aminopeptidase N (hAPN) receptors (Fig. 5B). Thus, with sufficient cell-surface protease activities, the primary hDPP4 receptor was dispensable, conceivably because the secondary sialate-binding S1A RBDs can operate at a stage in the process.
To determine whether the MERS S1A domains operate in the cell fusion process, S1A-Fc proteins (Fig. S5) were added during the assays. Exogenous S1A proteins interfered with hDPP4-independent MERS S-mediated fusion but had no effect on fusion when hDPP4 was present (Fig. 5C). To determine whether sialic acids operate in the cell fusion process, neuraminidase was added during the assays. Exogenous neuraminidase reduced hDPP4-independent MERS-S-mediated fusion but had no effect on the cell fusion when hDPP4 was present (Fig. S3). These results further suggest that either of the two S domains, S1A or S1B, can tether S proteins to target cells for subsequent cell-cell fusions, with distinct host receptors for each domain.
An analogous MERS-CoV spike mutation similarly increases hDPP4-independent cell binding and membrane fusion. We next determined whether analogous changes in S1A of the related MERS-CoV also modified viral attachment. The S1A interaction with sialic acids has biological significance, as antibodies against S1A protect mice from lethal MERS-CoV infections (55), and S1A acquires adaptive mutations in humans (56) and in mouse models of human MERS-CoV lung infection (57,58). In mice, this selectivity was at Asn222 (57, 58), a known glycan addition site (17,59,60), and a documented hypervariable residue, evident in a group of MERS-CoV-related HKU4 bat viruses but absent in the HKU5 bat virus groups (Fig. 6A). Remarkably, in the context of CoV S1A structure, the Asn222 position overlaps closely with B-CoV residue Glu170 and MHV-CoV residue Gly176 (Fig. 6B). We constructed MERS VLPs to determine whether this Asn222 change impacted cell binding. In initial tests, we constructed MERS-CoV VLPs and evaluated their sialate binding properties. Human red blood cells (RBCs) were agglutinated by MERS VLPs, while neuraminidase-treated RBCs were not (Fig. S4), indicating that VLPs do bind sialosides. We then constructed wild-type and ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) N222D VLPs containing internal Rluc and compared their Rluc and S-protein levels relative to VLPs lacking S proteins (Fig. 6C). VLPs were applied to human lungderived Calu3 cells, which express the human DPP4 receptor for MERS-CoV, and to mouse lung-derived LET-1 and mouse brain-derived DBT cells, neither of which express human DPP4 and thus do not bind MERS-CoV via a protein receptor. VLP binding was assessed by quantifying cell-associated Rluc. Asn222 change did not change VLP binding to human Calu3 cells but clearly enhanced VLP binding to murine DBT and LET-1 cells (Fig. 6D). We next considered whether this adaptation for increased cell binding had consequences in MERS S-mediated cell-cell membrane fusion, similar to the JHM-CoV G176E mutant. We found that the N222D mutant S proteins did indeed exhibit enhanced cell fusion, relative to wild-type S proteins (Fig. 6E), but this was only observed by eliminating hDPP4 (Fig. 6E), which removes the dominant S1B-protein receptor interaction and demands S1A utilization. To further probe the relationships between S1A-cell binding and cell fusion, we determined whether N222D S1A-Fc proteins would suppress cell fusion more effectively than WT S1A-Fc. Indeed, we found that exogenously added N222D S1A-Fc proteins suppressed cell fusions more potently than the corresponding wild-type S1A-Fc did (Fig. 6F and Fig. S5), further reinforcing the direct relationship between S1A binding to cells and resultant cell-cell fusion. Altogether, these findings support the hypothesis that MERS-CoV and JHM-CoV can acquire increased cell binding through S1A mutation, which consequently allows for more robust cell-cell fusion capability. Both of these CoVs can procure these cell binding and cell fusion properties through similarly localized S1A mutations.

DISCUSSION
Viruses frequently begin infection by attaching to cellular sialic acids (61). Such viruses include several CoVs, which attach to sialates via spike S1A domains (11,12), at low affinity (18), but relatively high multivalent avidity (21,41). Atomic resolution structures of 9-O-acetylated sialic acid in complex with the human OC43-CoV S1A domain (18) as well as several sialosides in complex with the MERS-CoV S1A (54) have revealed architectures of CoV-sialate binding sites. Yet even with this detailed understanding, it is not clear whether sialic acids confer susceptibility to CoV infection on their own or whether proteinaceous CoV receptors are also required. In considering this question, we first focused on the murine MHV JHM-CoV strain. This strain can infect cells and mice that lack the proteinaceous MHV receptor, mCEACAM1a (27, 30, 32-35, 62, 63). This strain is also unusually sensitive to proteolytic activation of spike-mediated membrane fusion (46), and therefore, we hypothesized that a low-affinity cell binding event, conceivably to cellular sialoglycans, might be sufficient for subsequent JHM-CoV protease-triggered fusion activation and cell entry.
We found that JHM-CoV binding to cells depended on host sialic acids. This binding accounted for initial virus attachment to cells, which facilitated subsequent virus engagement with proteinaceous CEACAM receptors. Later in the infection cycle, JHM-CoV spikes facilitated cell fusions and further virus dissemination in a sialic acid-dependent manner. MERS-CoV spike proteins could similarly fuse cells together without requiring prototype protein (hDPP4) receptors, presumably by utilizing sialoside receptors to engage neighboring cells. These findings, summarized in Fig. 7, suggest that CoV-cell entry and intercellular spread involve the lectin-like activities of spike proteins during both virus-cell attachment and infected-cell expansion into syncytia. green), A59 S1A (PDB accession no. 3JCL, magenta), and MERS S1A (PDB accession no. 5X4R, yellow) using PyMol. MERS S1A residue N222 (stick representation) is located proximally to B-CoV S1A residue E170. (C) MERS VLP binding characterization. MERS-VLPs with equivalent Rluc activities (right) were processed for Western blotting to image S-protein abundance and integrity (left). (D) VLPs were added at equivalent Rluc input multiplicities to human Calu3 or murine LET-1 cells. After 2 h at 4°C, cellassociated Rluc activities were quantified, and data were presented after subtracting background ("No S") Rluc ϩ VLP levels. (E) Effector cells expressing the indicated S proteins were cocultured with target cells expressing hTMPRSS2 and/or hDPP4. (F) MERS S-expressing effector cells were mixed with hTMPRSS2expressing target cells, in the presence of the indicated Fc proteins. All Fc proteins were at 10 M concentration throughout the coculture period. Experimental procedures, data acquisition, and processing were performed as described in the Fig. 1 legend.
Sialic Acids Facilitate Coronavirus Dissemination ® While sialic acids affected virus-cell binding, infection by JHM-CoV and cell entry by MERS-CoV VLPs required the proteinaceous receptors mCEACAM and hDPP4, respectively. Therefore, one possibility is that multivalent CoV-sialate interactions are not sufficient to hold viruses throughout a virus particle-cell membrane fusion process and that sialic acids may instead only tether viruses, transiently, onto host cell surfaces, as they diffuse two-dimensionally and then bind more stably to high-affinity protein receptors. Yet, while insufficient for the initial infection, sialic acid binding may advance the spike-mediated cell-cell membrane fusions that are observed in several CoV infection processes. In cell-cell fusion, spikes on infectedcell surfaces are steadily directed at uninfected (target) cells. They are not liable to diffuse away as cell-bound viruses would, yet they still may require attachment to target cell sialates in order to effect membrane fusion. In support of this contention, we found correlations between sialic acid abundance and membrane fusion activity. This was observed for both JHM-and MERS-CoV spikes, and notably, both the JHMand MERS-CoV spike-directed cell fusions took place independently of any proteinaceous receptors. These correlations suggest that sialic acids on uninfected cells can arrange spikes such that they will then operate cooperatively to pull opposing plasma membranes into coalescence.
This differential utilization of sialoside and protein receptors raises questions about CoV evolution. The CoV S1A domains fold into a galectin-like structure (14, 15, 17-19, 25, 64), and several CoV S1A domains demonstrably bind carbohydrate ligands (18,21,41,54), but prior to this report, the MHV-CoV S1A domains were considered to bind only protein (mCEACAM) receptors (25). However, Peng et al. (19) proposed that ancestral MHV-CoVs bound carbohydrates, with adaptive evolution generating the present-day CEACAM-binding sites. This proposal infers the existence, past or present, of evolutionary intermediates capable of binding both mCEACAM and sialic acid. The MHV JHM-CoV strain meets the criteria for such an intermediate-a virus in which a single domain retains two "receptor" binding activities, one for carbohydrate and the other for protein. Conceivably, these two receptor sites interact, such that increased affinity for one ligand reduces affinity for the other. Mutants with relatively high affinity for sialic acid may bind poorly to mCEACAM, while strains lacking sialic acid binding may demonstrate strong mCEACAM affinity. An inverse relationship would be consistent with the study of Peng et al., who proposed that acquisition of a CEACAM-binding site concomitantly destroyed a sialate site (19). Yet perhaps more intriguing relationships are between CoV S and hemagglutinin-esterase (HE) proteins, both of which bind  (65,66). Esterase activities within HE proteins deacetylate sialosides, and it has been documented, in HKU1-CoV infections, that HE activity destroys an S-specific sialoside receptor, keeping viruses detached from cells and allowing for virus dissemination (67). Several MHV strains, including JHM, express a viral HE, which binds and deacetylates 4-O-acetyl sialiate (68). Similar to HKU-1-CoV HE, MHV HE could destroy MHV sialoside receptors and facilitate dissemination. MHV strain A59 did not bind sialic acids, and it does not retain an intact HE open reading frame (36,37), while strain JHM did bind sialic acids and does express HE, suggesting that HE is necessary for those CoVs with relatively high S-protein affinity for sialosides. MERS-CoV binds sialic acids, albeit weakly (21), and increases in its affinity for sialates may be limited by the absence of a MERS-CoV HE gene.
This study revealed S-protein mutations that may endow CoVs with expanded tropisms, beyond that determined by prototype proteinaceous CoV receptors. The JHM-CoV mutation G176E, engineered to reflect sialate-utilizing B-CoV, increased viral S-protein binding to cells, as well as S-protein-mediated membrane fusion, independently of mCEACAM. The MERS-CoV mutation N222D, an adaptation for virus growth in mouse lungs (58), operated remarkably similar to the JHM-CoV G176E change, with mutant S proteins showing increased hDPP4-independent cell binding and cell fusion. Our findings fit with the hypothesis that viruses with these mutations bind relatively tightly to sialoside receptors. However, it is clear that the changes are not present in currently identified sialoside-binding sites (18,54), which raises alternative hypotheses that include mutation-induced allosteric restructuring of sialoside-binding sites. Alternatively, the mutations may restructure distinct receptor-or coreceptor-binding sites, such as presumed binding sites for orthologous CEACAMs on MHV S proteins (69), or sites for CEACAM5 (70) or GRP78 (71) on MERS-CoV S proteins.
Finally, this report illuminates understanding of CoV pathogenesis. Among the CoVs, the JHM-CoV strain is known for causing lethal brain infection, even in mice that lack the principal MHV receptor, CEACAM1a (30). JHM spike was identified to be the major contributor to this phenotype (30), and JHM-CoV, but not the related A59-CoV, spread interneuronally both in vivo and within in vitro cultures of central nervous system (CNS)-derived cells (89,90). Notably, neural cell membranes are known for their abundant sialic acid content (72). These findings, combined with evidence that cell-tocell syncytial spread correlates with pathogenesis in several infection models (30,62,63,(73)(74)(75), prompts a hypothesis that JHM-CoV sialic acid binding potential accounts for an interneuronal syncytial spread that is rapidly lethal. A prediction is that variants of JHM-CoV exhibiting enhanced sialic acid affinity will have unusually high neurovirulence. Similarly, the MERS-CoV strain causes lethal pneumonia, and here it is significant that antibodies specific for the MERS-CoV S1A domains both neutralize the virus and reduce infection and pathogenesis in a mouse MERS-CoV model system (55,59,76). Conceivably, these antibodies interfere with sialic acid binding, reducing expansion of MERS-CoV that may take place via cell-cell fusion. Variants of MERS-CoV with enhanced cell binding may be useful in assessing the in vivo significance of the findings presented in this report.
Expression plasmids (1 g) were complexed with polyethylenimine (PEI) (Polysciences) (6 g) in 0.2 ml of Opti-MEM (Life Technologies) for 15 min at room temperature and then added dropwise to 5 ϫ 10 5 HEK293T cells in 1 ml of DMEMϪ10% FBS. For spikeless VLPs, S expression plasmids were replaced with empty vector plasmid DNAs. Six hours later, the cells were replenished with fresh DMEMϪ10% FBS. Media were collected at 48 h posttransfection.
Hemagglutination assay. Serial twofold dilutions of 100ϫ concentrated virus and pseudovirus particle stocks were prepared using FBS-free DMEM as diluent. Diluted samples were placed into V-bottom 96-well plates and incubated for 2 h at 4°C. Washed adult human erythrocytes (0.5%) (in phosphate-buffered saline [PBS]) were then added, and scoring was performed after 2-to 12-h incubation at 4°C. For live-cell assays, adherent, confluent cells in 24-or 96-well plates were pretreated with vehicle or neuraminidase (1 U/ml) (catalog no. N2876; Sigma) in FBS-free DMEM for 2 h at 37°C. The cells were rinsed three times with PBS and then used in virus transduction assays.
VLPs were incubated for 2 h at 4°C to adherent, confluent cells, either with or without prior neuraminidase treatments. Incubations were in FBS-free DMEM and at known Rluc VLP/cell ratios. After incubation, unbound VLPs were removed, and cells were rinsed variably with PBS. To measure cell-bound VLPs, adherent cells were dissolved into PLB, and Rluc activity was quantified.
VLP entry assay. VLPs used for entry assays were prepared by cotransfection with plasmids encoding CoV S, E (envelope), M (matrix), and DSP 8-11 -N. The resulting VLPs contained DSP fragments in their interiors. To quantify VLP-associated DSP levels, VLPs were mixed with excess DSP 1-7 , in passive lysis buffer, for 30 min at 37°C, to allow for DSP complementation. The excess DSP 1-7 was obtained from HEK293 cells overexpressing these fragments. After postlysis complementation, samples were introduced into opaque microplate wells and luminescence readings were used to infer DSP [8][9][10][11] levels.
Virus and pseudovirus particle entry assays. Virus particles were incubated with cells, either with or without prior neuraminidase treatments, for 2 h at 4°C. Cells were then rinsed three times with PBS and replenished with FBS-containing DMEM/MEM. After 16 h at 37°C, cells were dissolved in lysis buffer (25 mM  Cell-cell fusion assay. Effector and target cells were prepared by introducing expression plasmids, using the PEI transfection method. Effector cells were cotransfected with pDSP 1-7 and the indicated S-expressing plasmids. S-expressing plasmids included pcDNA3.1-229E S C9 (GenBank accession no. AB691763.1, obtained from Fang Li, University of Minnesota). Control effector cells received pDSP 1-7 and empty vector plasmids. Target cells were cotransfected with pDSP [8][9][10][11] and the indicated S receptorexpressing plasmids. These plasmids included pCMV6-Entry-hDPP4 FLAG (GenBank accession no. NM_001935; purchased from OriGene) and pcDNA3.1-hAPN (GenBank accession no. M22324; obtained from Fang Li). Indicated experiments included additional cotransfection of target cells with pCAGGS-TMPRSS2 FLAG (44).
At 6 h posttransfection, effector and target cells were rinsed with PBS, lifted with 0.05% trypsin, and mixed at 1:1 ratios. The cocultures were incubated for 2 to 18 h at 37°C. Fused cells were visualized microscopically as green fluorescent protein-positive (GFP ϩ ) cells, and extents of cell-cell fusion were quantified as Rluc activities present in PLB cell lysates. Samples were chilled and maintained at 4°C during Rluc quantifications to eliminate DSP postlysis complementation. In time course experiments, EnduRen substrate (Promega E6482) was used in place of coelenterazine.
HEK293T cells were transfected using the PEI method and then incubated in FBS-free DMEM containing 2% (wt/vol) Cell Boost 5 (HyClone). Supernatants were collected on days 4 and 7 and clarified by sequential centrifugation (300 ϫ g, 4°C, 10 min; 4,500 ϫ g, 4°C, 10 min). The Fc-tagged proteins were then purified using HiTrap protein A high-performance columns (GE Healthcare) according to the manufacturer's instruction. The resulting purified proteins were quantified spectrophotometrically and stored at -20°C until use.
Microscopy and image acquisition. Live-cell images were captured in a z series on an electronmultiplied charge-coupled-device digital camera (EMCCD Cascade 2; Photometrics) and deconvolved using SoftWoRx. Identical conditions were applied to all acquisitions. Deconvolved images were analyzed by an identical algorithm in Imaris 8.3.1 (Bitplane).
Statistical analysis. Unless stated otherwise, all experiments were repeated independently at least three times. Each data point graphed represents the mean of an independent repeat (N), generated from three or four technical repetitions (n ϭ 3 or 4). Statistical comparisons were made by unpaired Student's t test. Error bars indicate the standard errors of the data. P values less than 0.05 were considered statistically significant.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
This research was supported by the National Institutes of Health (NIH) under award P01 AI060699.
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We thank Edward Campbell (Loyola University Chicago) for assistance with fluorescent confocal microscopy. We thank Arlene Barlan and Gautami Galpalli for expert technical support.