Significant role of host sialylated glycans in the infection and spread of severe acute respiratory syndrome coronavirus 2

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has been transmitted across all over the world, in contrast to the limited epidemic of genetically- and virologically-related SARS-CoV. However, the molecular basis explaining the difference in the virological characteristics among SARS-CoV-2 and SARS-CoV has been poorly defined. Here we identified that host sialoglycans play a significant role in the efficient spread of SARS-CoV-2 infection, while this was not the case with SARS-CoV. SARS-CoV-2 infection was significantly inhibited by α2-6-linked sialic acid-containing compounds, but not by α2–3 analog, in VeroE6/TMPRSS2 cells. The α2-6-linked compound bound to SARS-CoV-2 spike S1 subunit to competitively inhibit SARS-CoV-2 attachment to cells. Enzymatic removal of cell surface sialic acids impaired the interaction between SARS-CoV-2 spike and angiotensin-converting enzyme 2 (ACE2), and suppressed the efficient spread of SARS-CoV-2 infection over time, in contrast to its least effect on SARS-CoV spread. Our study provides a novel molecular basis of SARS-CoV-2 infection which illustrates the distinctive characteristics from SARS-CoV.

SARS-CoV-2 infection [24]. These findings highlight the need for further investigation on sialic acid's role in SARS-CoV-2 infection.
In this study, we screened a series of compounds using SARS-CoV-2 cell culture-based assay and identified α2-6-linked sialic acid-containing compounds to inhibit viral infection. SARS-CoV-2 infection was inhibited by α2-6SLN-lipo-PGA compound in VeroE6/TMPRSS2 and Calu-3 cells. α2-6-linked sialylated glycans targeted SARS-CoV-2 spike S1 subunit to inhibit viral attachment to the host cells. Furthermore, enzymatic removal of endogenous sialic acids on the cell surface by neuraminidase suppressed the spread of SARS-CoV-2 infection, but this was not the case with SARS-CoV. Our study revealed that SARS-CoV-2 utilizes cell surface sialic acid which can support the virus binding to its receptor ACE2. The finding provides novel insight into the molecular basis for the unique characteristics of SARS-CoV-2 infection and spread.

α2-6SLN-lipo-PGA inhibited SARS-CoV-2 infection
By evaluating a series of compounds using SARS-CoV-2 infection cell culture system, we found that α2-6-linked sialic acid-containing compounds inhibited SARS-CoV-2 infection. In this screening we investigated the antiviral activity of compounds using α2-3SLN-PGA In contrast, α2-6SLN-lipo-PGA reduced the SARS-CoV infection marginally to around 10% at most (Fig 2C, blue). An immunofluorescence detection of SARS-CoV N protein also showed only a little or no effect of α2-6SLN-lipo-PGA compound (Fig 2B, panel n), which almost completely eliminated the SARS-CoV-2 N protein in the same infection assay (Fig 2B, panel g). We then investigated the antiviral activity of the component of α2-6SLN-lipo-PGA. Independent treatment of N-acetylneuraminic acid (Neu5Ac, see Fig 1 and S1 for the structure), which is the sialic acid in α2-6SLN-lipo-PGA, exhibited antiviral activity (Fig 2D, left), while this was not the case with SARS-CoV infection (Fig 2D, right). We further confirmed that methyl α-Neu5Ac and α2-6 linked triose (α2-6 sialyllactose) inhibited SARS-CoV-2 infection (S2D and S2E Fig). On the contrary, lactosylsphingosine-PGA [βGal-(1-4)-βGlc-1-sphingosine-poly-α-L-glutamic acid] (LS-PGA, see Figs 1 and S1 for the structure), which is composed of similar structure to α2-6SLN-lipo-PGA without having sialic acid, did not reduce the viral RNA ( Fig 2E). The data indicates that the sialic acid, and not the SLN-lipo-PGA backbone, is responsible for the antiviral activity of α2-6SLN-lipo-PGA. We also confirmed that α2-6SLN-lipo-PGA showed a log-scale reduction in SARS-CoV-2 infection in a humanderived lung epithelial cell line Calu-3 cells (Fig 2F).

α2-6SLN-lipo-PGA inhibited SARS-CoV-2 attachment
To investigate which step in the viral life cycle was blocked by α2-6SLN-lipo-PGA, we performed a time of addition assay ( Fig 3A). The antiviral activity of compounds was measured under three different conditions: (a, whole) compounds present throughout the experiment, (b, entry) compounds were treated for 1 h during viral inoculation and additional 2 h and then removed, and (c, post-entry) compounds were added at 2 h after washing free viruses and remained throughout the experiment. As shown in Fig 3B, SARS-CoV-2 replication inhibitor remdesivir treated at (a) whole and (c) post-entry showed significant viral RNA reduction

PLOS PATHOGENS
Role of host sialylated glycans in SARS-CoV-2 infection and spread without showing significant effect on viral entry (b). Chloroquine, a known SARS-CoV-2 entry inhibitor, was confirmed to show antiviral activity in the early stages of infection (Fig 3B, Chloroquine, b). Since this assay allows multiple rounds of re-infection, entry inhibitors can show antiviral effects when added at (c) post-entry (Fig 3B, Chloroquine, c). Since α2-6SLNlipo-PGA treatment inhibited viral infection at all three conditions (Fig 3B, α2-6SLN-lipo-PGA, a-c), α2-6SLN-lipo-PGA was suggested to inhibit the viral entry process. We further confirmed the result using SARS-CoV-2 pseudovirus; which consists of vesicular stomatitis virus-based pseudovirus carrying SARS-CoV-2 spike, to evaluate SARS-CoV-2 spike-dependent cell entry. As Fig 3C shows, SARS-CoV-2 pseudovirus infection was significantly inhibited by the treatment with α2-6SLN-lipo-PGA, as well as heparin, which inhibits cell attachment through heparan sulfate proteoglycan in multiple viruses including SARS-CoV-2 [27,28]. We then examined the effect of the compound on viral attachment to the host cell surface, by incubating the cells with SARS-CoV-2 at 4˚C to allow virus-cell attachment but not the following process, and quantifying viral RNA attached on the cell surface ( Fig 3D). Heparin significantly inhibited SARS-CoV-2 attachment to cells, while chloroquine, which targets

PLOS PATHOGENS
Role of host sialylated glycans in SARS-CoV-2 infection and spread

α2-6SLN-lipo-PGA targets SARS-CoV-2 particles to inhibit viral attachment
We then examined whether α2-6SLN-lipo-PGA targeted SARS-CoV-2 particles or host cells to inhibit viral infection. At first, to determine whether α2-6SLN-lipo-PGA targeted viral particles, we prepared "compound-treated SARS-CoV-2 particles" by incubating SARS-CoV-2 with or without compounds for 1 h and then removing unbound compounds by ultrafiltration. The capacity of the prepared virus for cell attachment was examined by treating VeroE6/ TMPRSS2 cells for 30 min at 4˚C and quantifying the attached virus. The viral attachment was greatly reduced by the treatment with α2-6SLN-lipo-PGA, as well as heparin that targets viral particle, but not with anti-ACE2 antibody ( Fig 4A). To determine whether α2-6SLN-lipo-PGA targets host cells, we prepared "compound-treated cells" by incubating VeroE6/TMPRSS2 cells with compounds for 30 min. After washing out free compounds, the cells were inoculated with virus for 30 min at 4˚C in the absence of compounds and attached viral particles were detected. Cell susceptibility to SARS-CoV-2 attachment was significantly reduced when treated with anti-ACE2 antibody, but not with α2-6SLN-lipo-PGA ( Fig 4B). These results suggest that α2-6SLN-lipo-PGA targets SARS-CoV-2 particles to inhibit viral infection.

α2-6-linked sialosides interacted with the S1 subunit of SARS-CoV-2 spike protein
We further analyzed the direct interaction between SARS-CoV-2 spike and sialic acids in vitro. Spike protein is divided into S1 and S2 subunits that are responsible for viral attachment to cell and virus-cell membrane fusion, respectively. Since α2-6-linked sialosides inhibited viral attachment to host cells (Fig 3D), we hypothesized that α2-6 sialylated glycans might target SARS-CoV-2 S1 subunit; we tested this hypothesis by glycan array. ACE2 (positive control), asialo-SGP (negative control), α2-3-SGP, and α2-6-SGP (see Fig 1 for the structure) were immobilized on a glass slides and were incubated with recombinant SARS-CoV-2 S1 protein (aa 16-685) ( Fig 4C). As Fig 4D shows, incubated recombinant S1 interacted with ACE2 and yielded the fluorescence signal, confirming that the assay is working properly, whereas asialo-SGP, which does not contain sialosides, showed much less response. α2-6-SGP yielded significant responses in a dose-dependent manner, while α2-3-SGP exhibited as much response as the negative control asialo-SGP, regardless of the concentration. This result indicates that α2-6-linked sialylated glycans directly interact with the S1 subunit of SARS-CoV-2 spike.

PLOS PATHOGENS
Role of host sialylated glycans in SARS-CoV-2 infection and spread neuraminidase from either Arthrobacter ureafaciens or Vibrio cholerae, SARS-CoV-2 pseudovirus infection significantly reduced (Fig 5A, left). The removal of sialic acid from VeroE6/ TMPRSS2 cells by treatment with neuraminidase was confirmed by lectin staining (S3 Fig). In contrast, infection of SARS-CoV pseudovirus was not affected by neuraminidase treatment (Fig 5A, right). Consistently, removal of sialic acid reduced SARS-CoV-2 attachment to the cell surface, as monitored by detecting viral RNA and spike protein derived from attached viral particles, although the cell expression of ACE2 was not decreased upon neuraminidase treatment ( Fig 5B). We then examined the role of sialic acid in virus-ACE2 binding by incubating neuraminidase-treated or untreated cells with SARS-CoV-2 at 4˚C to examine spike-ACE2 binding by coimmunoprecipitation assay: SARS-CoV-2 spike was copurified with endogenous ACE2 in intact cells ( Fig 5C, lane 1), whereas this ACE2-coprecipitated SARS-CoV-2 spike was impaired by depletion of sialic acids upon neuraminidase treatment ( Fig 5C,  lane 2). Moreover, we found that the interaction of ACE2 and SARS-CoV-2 spike S1 was impaired when mammalian cell-derived recombinant ACE2 was pretreated with neuraminidase ( Fig 5D, lanes 2, 3). The removal of sialic acid on the recombinant ACE2 by neuraminidase was confirmed by mobility shift of the band in gel by immunoblotting ( Fig 5D, lanes 4,  5). Thus, host sialic acid on the cell surface facilitates the binding of SARS-CoV-2 S1 and ACE2. Fig 5E indicates that dual-inhibition of ACE2 and sialic acid had more inhibitory effect than mono-treatment of anti-ACE2 antibody in SARS-CoV-2 pseudovirus infection, which was not the case with SARS-CoV pseudovirus infection. Finally, time-dependent spread of SARS-CoV-2 and SARS-CoV infection was examined. The spread of SARS-CoV-2 infection detected by SARS-CoV-2 N protein over time post infection was more moderate upon neuraminidase treatment (Fig 5F-a, upper), which was quantitatively demonstrated by detecting virus infected area (Fig 5F-b, upper) or viral RNA in culture supernatant (Fig 5F-c). In contrast, the spread of SARS-CoV infection was readily observed irrespective of the removal of sialic acid (Fig 5F-a, b, c, lower). These results suggest that SARS-CoV-2 utilizes sialic acid to facilitate cell attachment and binding to ACE2, which leads to an efficient spread of the viral infection.

Discussion
In this study, using cell culture-based infection assay, we identified that α2-6-linked sialic acid-containing compounds inhibited SARS-CoV-2 infection, but not SARS-CoV. It was revealed that the compound targeted virus particles to inhibit viral attachment to cells. In vitro analysis using recombinant SARS-CoV-2 spike protein and sialylglycopeptides consistently showed that α2-6-linked sialylated glycan directly interacted with the S1 subunit of SARS- The attachment of compound-pretreated virus to cells. SARS-CoV-2 particles were pretreated with the compounds at 37˚C for 60 min, then the free compounds were removed by ultrafiltration. The prepared viruses were treated to VeroE6/TMPRSS2 cells in the absence of compounds at 4˚C for 30 min to examine virus-cell attachment. (B) The viral attachment to compound-pretreated cells. The indicated compounds were treated to VeroE6/TMPRSS2 cells at 37˚C for 30 min and were then washed out extensively. The prepared cells were incubated with SARS-CoV-2 in the absence of compounds at 4˚C for 30 min to examine virus-cell attachment. (A, B) Anti-ACE2 antibody, 100 μg/ml; heparin, 10 U/ml; α2-6SLN-lipo-PGA, 10 μM. (C) Schematic representation of glycan array. Glycan tip immobilized with sialylglycopeptides on a slide glass was incubated with recombinant SARS-CoV-2 S1 at room temperature for 1 h, and then the fluorescence signal was detected. Anti-SARS-CoV-2 S1 antibody was used as primary antibody, and Cy3-conjugated anti-rabbit IgG was used as secondary antibody. (D) The fluorescent response of the glycan array due to the interaction between recombinant SARS-CoV-2 S1 (aa 16-685, Sino Biological) and indicated protein or glycans. The amount of glycan on the neoglycoproteins were 10.8 mol (α2-6-SGP), 10.5 mol (α2-3-SGP) and 10.8 mol (asialo-SGP) of glycan per 1 mol of BSA. The dashed line indicates the nonspecific response level derived from negative control asialo-SGP. Asialo-SGP, 0.7 mg/ml; α2-3-SGP, 0.1 and 1 mg/ml; α2-6-SGP, 0.1 and 1 mg/ml; recombinant ACE2, 0.1 mg/ml.

PLOS PATHOGENS
Role of host sialylated glycans in SARS-CoV-2 infection and spread The sialic acid-binding capacity is one of the many determinants of biological diversity in the family of coronavirus. Several coronaviruses have been reported to interact with sialic acids, including HCoV-OC43, HCoV-HKU1, bovine CoV (BCoV), mouse hepatitis virus, MERS-CoV, transmissible gastroenteritis virus, and infectious bronchitis virus [16][17][18]29,30]. For HCoV-OC43 and HCoV-HKU1, 9-O-acetylated-Sia (Neu5,9Ac 2 ) is the sole receptor, in which the binding critically determines the cell entry of the virus. In contrast, the sialic acid use of SARS-CoV-2 is more similar to that of MERS-CoV, which involves binding to the cell surface of sialylated glycans contributing to the efficiency of binding to their entry receptor DPP4. Fantini et al. [31] proposed a dual receptor/attachment model for SARS-CoV-2 with the spike protein involved in ACE2 receptor recognition and finding sialylated sphingoglycolipid (ganglioside)-rich landing area (lipid rafts) at the cell surface. Considering our result that depletion of cell surface sialic acids reduces the capacity of cells for the binding of SARS-CoV-2 particles with the impairment of ACE2 interaction to the viral spike protein, the interaction of SARS-CoV-2 with sialic acids may be responsible for the early contact with host cells, which can serve in concentrating the viruses on the surface of target cells, facilitating their binding to high-affinity receptor ACE2.
The genome sequence homology of SARS-CoV-2 exhibits 79.6% similarity with SARS-CoV, and they share the same receptor ACE2 for their entry into host cells [12][13][14]. Despite the similarities, our data demonstrated one of their differences that SARS-CoV-2 has the ability to use host cell sialosides while SARS-CoV does not. The N-terminal domain (NTD) of spike protein is responsible for glycan binding in other coronaviruses such as BCoV, HCoV-OC43, and MERS-CoV. The NTD sequence of SARS-CoV-2 and SARS-CoV show only 53.5% of homology, while full-length spike proteins share almost 76% identities [19]. Seyran et al. [20] proposed the presence of the distinct flat and non-sunken sialic acid-binding domain in SARS-CoV-2 S1-NTD which enables faster viral surfing over the epithelium. This promotes faster viral movement to ACE2 receptor compared to other human coronaviruses such as SARS-CoV. Moreover, the structure-based sequence comparison of the spike protein of SARS-CoV, SARS-CoV-2, and MERS-CoV identified that SARS-CoV-2 S1-NTD contains structural regions reminiscent of MERS-CoV sialoside binding pockets [21,22]. The sialic acid binding of recombinant SARS-CoV-2 spike was experimentally investigated by Baker et al. [23] using detected. The attached viral spike protein and endogenous protein expression of ACE2 and actin in VeroE6/TMPRSS2 cells was determined by immunoblotting. The number under the blot shows the relative intensity (fold) of the band measured using ImageJ program. (C) The interaction between ACE2 and SARS-CoV-2 spike under pretreatment with neuraminidase. VeroE6/TMPRSS2 cells were treated with or without neuraminidase for 2 h at 37˚C, then were inoculated with SARS-CoV-2 for 1 h at 4˚C. After washing unbound viruses, the cells were harvested to perform immunoprecipitation with anti-ACE2 antibody (IP: ACE2) or normal mouse IgG as a negative control (IP: IgG). Viral spike protein and ACE2 in the precipitates were detected by immunoblotting. (D) The effect of sialic acid on ACE2 for the interaction with SARS-CoV-2 spike S1. The fluorescent response due to the interaction between recombinant SARS-CoV-2 spike S1 and recombinant ACE2 protein was detected. Recombinant ACE2 proteins were incubated with or without 2 U/ml neuraminidase (Arthrobacter ureafaciens) for 2 h at 37˚C, and each ACE2 sample were immobilized on a glass slide to analyze the interaction with SARS-CoV-2 spike S1 protein. The dashed line indicates the nonspecific response level of protein-free glass slide. The schematic representation of the assay is shown under the data. The removal of sialic acid on ACE2 by treatment with neuraminidase was confirmed by immunoblotting.

PLOS PATHOGENS
Role of host sialylated glycans in SARS-CoV-2 infection and spread gold nanoparticles bearing sialic acid, which demonstrated that αNeu5Ac is a ligand for spike glycoprotein, although they did not examine the function of αNeu5Ac in virus infection. Nguyen et al. [24] presented a new idea that receptor binding domain (RBD) of SARS-CoV-2 spike is involved in the recognition of sialylated glycans. They performed in vitro glycan library screening against SARS-CoV-2 RBD and S protein, which revealed that several sialic acid-containing structures are recognized. The usage of glycolipids by SARS-CoV-2 was confirmed by pharmacological abrogation in viral infection. These findings are consistent with our result that SARS-CoV-2 S1 interact with sialic acid-containing molecules, which supports viral entry into host cells.
As the viral entry process is essential for initiation, spread, and maintenance of infection, this step is an attractive target for the development of antiviral agents. Our finding, that SARS-CoV-2 utilizes sialic acids for its entry, provides novel insight into understanding the molecular mechanisms of SARS-CoV-2 infection and the role of sialylated glycans in viral entry as well as the developing antiviral strategies.

SARS-CoV-2 and SARS-CoV infection assay
VeroE6/TMPRSS2 cells were seeded in 96 well plates a day before virus infection. The cells were inoculated with SARS-CoV-2 Wk-521 strain, a clinical strain isolated from a COVID-19 patient [47], or SARS-CoV Frankfurt strain at a multiplicity of infection (MOI) of 0.001 or 0.003 for 1 h and the unbound virus was removed by washing. Cells were cultured for 24 h prior to measuring the extracellular viral RNA or detecting viral encoded N protein. Compounds were added during virus inoculation (1 h) and replenished after washing (24 h). Infection assay with Calu-3 cells was performed by incubation with virus at a MOI of 0.1 for 3 h. Extracellular viral RNA was detected at 72 h post-inoculation. For the calculation of MOI, viral titers (50% tissue culture infective dose: TCID 50 ) were determined by Behrens-Karber method [51].

Cell viability
At 24 h post infection, the same time point when culture supernatant was collected to measure extracellular viral RNA, cells were washed with PBS and fixed with 4% paraformaldehyde. Then the cells were stained with DAPI. Cell viability was determined by the quantification of survived cell numbers by counting the number of detected nucleus with three cell culture wells (one microscopic field in each well) in a 96 well plate using a high content imaging analyzer ImageXpress Micro Confocal (Molecular Device) [53,54].

Indirect immunofluorescence analysis
The cells were fixed with 4% paraformaldehyde and permeabilized with 0.3% Triton X-100. To detect viral protein expression, the cells were treated with a rabbit anti-SARS-CoV/SARS-CoV-2 N antibody [55] as a primary antibody, and then incubated with Alexa594-conjugated secondary antibody and DAPI. N-positive cells (Fig 5F-b) were quantified by calculating fluorescent stained area of three cell culture wells (one microscopic field in each well) in a 96 well plate using hybrid cell count system in BZ-X Analyzer (BZ-X710, KEYENCE).

Time of addition analysis
VeroE6/TMPRSS2 cells were inoculated with virus at an MOI of 0.001 for 1h and the unbound virus was removed by washing. Cells were cultured for 24 h and the amount of viral RNA in the supernatant was measured as an indicator of released virus particles. Compounds were added with three different timing: (a, whole) compounds present throughout the experiment, (b, entry) compounds were treated for 1 h during viral inoculation and an additional 2 h and then removed, and (c, post-entry) compounds were added at 2 h after washing free viruses and remained throughout the experiment [56][57][58]. Viral RNA in culture supernatant was measured as described above.

SARS-CoV-2 and SARS-CoV pseudovirus infection assay
SARS-CoV-2 and SARS-CoV pseudoviruses were produced using vesicular stomatitis virus (VSV)pseudovirus system as described previously [59,60] using expression plasmids encoding SARS-CoV-2 or SARS-CoV spike protein and the G-deficient VSV which carries the luciferase gene instead of the VSV-G gene. This pseudoviruses infect the target cells in a spike-dependent manner and produce luciferase activity in infected cells. VeroE6/TMPRSS2 cells were inoculated with SARS-CoV-2 or SARS-CoV pseudovirus in the presence or absence of compounds and intracellular luciferase activity was measured at 24 h post-inoculation using Luciferase Assay System (Promega).

Virus-cell attachment assay
VeroE6/TMPRSS2 cells were incubated with viruses in the presence or absence of compounds for 5 min (Fig 3D) or 60 min (Fig 5B)

Compound treatment to SARS-CoV-2 particles or host cells
To evaluate the effect of compounds on SARS-CoV-2 particles, virus was treated with or without compounds for 60 min at 37˚C, and unbound compounds were subsequently removed by ultrafiltration using centrifugal filter devices (Pall Corpotation). The viral sample ultrafiltrated without compound was used as a control. The attachment capacity of compound-treated virus was evaluated by treating VeroE6/TMPRSS2 cells in the absence of compounds for 30 min at 4˚C, and then extracting the attached viral RNA with RNeasy Mini Kit (QIAGEN) for quantification by real time RT-PCR. To evaluate the effect of compounds on host cells, VeroE6/TMPRSS2 cells were pretreated with compounds for 30 min at 37˚C before being washed. The capacity of compound-treated cells to virus attachment was measured by inoculating the cells with virus for 30 min at 4˚C in the absence of compounds and detecting the attached viral RNA.

The glycan array
Glycopeptide conjugated BSA was prepared by glycoconjugation of BSA using N-(3-maleimidobenzoyloxy) succinimide (Nacalai), with each glycopeptide by catalytic reaction. Preparation of the glycopeptide conjugated BSA microarray and binding assay of glycan microarray were carried out as reported previously [61,62]. The amount of glycans on the neoglycoprotein was determined by liberation of glycans, fluorescence labeling and quantified by HPLC (α2-6-SGP and asialo-SGP were 10.8 mol glycan/ 1 mol BSA and α2-3-SGP was 10.5 mol glycan/1 mol of BSA). Anti-SARS-CoV-2 S1 antibody (prepared in house) was used as a primary antibody, and Cy3-conjugated anti-rabbit IgG was used as secondary antibody. Following MIRAGE guidelines, information on the glycan array is shown.

Neuraminidase treatment
To examine the role of sialic acid in virus infection, VeroE6/TMPRSS2 cells were incubated with neuraminidase derived from either Vibrio cholerae (Roche) or Arthrobacter ureafaciens (Nacalai) for 2 h at 37˚C. After washing cells to remove neuraminidase, SARS-CoV-2 and

PLOS PATHOGENS
Role of host sialylated glycans in SARS-CoV-2 infection and spread SARS-CoV pseudovirus infection assay or SARS-CoV-2 attachment assay was performed as described above.

Lectin staining
For detection of sialic acids, VeroE6/TMPRSS2 cells were fixed with 4% paraformaldehyde. Biotinylated elderberry bark lectin and biotinylated Maackia amurensis lectin II (Vector Labs) were used to stain α2-6 and α2-3 sialic acid, respectively. The lectins were added into fixed cells for 1 h at room temperature, then streptavidin-conjugated Alexa Fluor 594 (Thermo Fisher Scientifics) and DAPI were added for 1 h at room temperature. The fluorescence images were captured using confocal microscopy TCS SP8 (Leica).

Co-immunoprecipitation analysis
The cells were lysed with the buffer [50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM EDTA, 1% NP-40 and complete protease inhibitor], followed by an immunoprecipitation with anti-ACE2 antibody (R&D Systems) or mouse normal IgG as a negative control as described previously [63].

Immunoblot analysis
Immunoblot analysis was performed as described previously [64] using anti-ACE2 (R&D Systems), anti-SARS-CoV-2 spike (Sino Biological) or anti-actin (Sigma) antibodies as primary antibodies. The intensity of the bands in Fig 5B was measured using ImageJ software.

Statistics
All statistical data were obtained by triplicates of three cell culture wells in an assay. All experiments were repeated at least two times in each assay and the representative data are shown in figures. Immunoblot analysis, indirect immunofluorescence analysis and lectin staining (Figs 2B, 5B-5D, 5F-a and S3) show one representative data, however, these experiments were repeated at least two times in each assay. Statistical significance was estimated using the twotailed Student's t test ( � p < 0.05; �� p < 0.01; N.S., not significant).