Sialic acid-containing glycolipids mediate binding and viral entry of SARS-CoV-2

Abstract

Emerging evidence suggests that host glycans influence severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. Here, we reveal that the receptor-binding domain (RBD) of the spike (S) protein on SARS-CoV-2 recognizes oligosaccharides containing sialic acid (Sia), with preference for monosialylated gangliosides. Gangliosides embedded within an artificial membrane also bind to the RBD. The monomeric affinities (K_d = 100-200 μM) of gangliosides for the RBD are similar to another negatively charged glycan ligand of the RBD proposed as a viral co-receptor, heparan sulfate (HS) dp2-dp6 oligosaccharides. RBD binding and infection of SARS-CoV-2 pseudotyped lentivirus to angiotensin-converting enzyme 2 (ACE2)-expressing cells is decreased following depletion of cell surface Sia levels using three approaches: sialyltransferase (ST) inhibition, genetic knockout of Sia biosynthesis, or neuraminidase treatment. These effects on RBD binding and both pseudotyped and authentic SARS-CoV-2 viral entry are recapitulated with pharmacological or genetic disruption of glycolipid biosynthesis. Together, these results suggest that sialylated glycans, specifically glycolipids, facilitate viral entry of SARS-CoV-2.

Extended Data Fig. 1 ∣. Structures of 139 defined glycans used in CaR-ESI-MS screening against RBD and S protein.

These glycans comprise 9 groups: Lewis antigens, blood group ABH antigens, antigen-related glycans, globo-series glycans, ganglioside oligosaccharides, human milk oligosaccharides, Neu5Acα2-6 linked glycans, rhamnose-containing glycans (Library P) and others.

Extended Data Fig. 2 ∣. CaR-ESI-MS screening against defined glycan libraries at 25 °C.

Normalized abundances of released glycans from the SARS-CoV-2 RBD by CaR-ESI-MS at 25 °C. Summary of the charge-normalized (relative to 71) abundances of released ligands measured by CaR-ESI-MS screening of Library A – O against SARS-CoV-2 RBD. Measurements were performed in negative ion mode with a UHMR Orbitrap mass spectrometer at an HCD energy of 50 V. Aqueous ammonium acetate (100 mM, pH 6.9) solutions of SARS-CoV-2 RBD (10 μM) and glycan library (containing 50 nM of each glycan) at were used for screening. The different classes of oligosaccharides are distinguished by colour: red - Lewis antigens (1–14); yellow - blood group A antigens (15–20); orange - blood group B antigens (21–26); green - blood group H antigens (27–32); blue - sulfated compounds (34–38) and HS oligosaccharides (39–41), light blue - antigen-related glycans (42–50), purple - globo (51–60), pink - ganglioside oligosaccharides (61–76), white – HMOs and other glycans (77–128), light grey - Neu5Acα2-6-linked oligosaccharides (129–132), and dark grey – rhamnose-containing compounds (133–139). NA, not applicable. Asterisk indicates that the relative abundances (in CaR-ESI-MS) were estimated from their relative (to 69) K_d. Black-outlined white-filled circles represent data points. Data represent means ± s.d., n = 3 independent experiments.

Extended Data Fig. 3 ∣. Zero-charge mass spectrum for an aqueous ammonium acetate solution of SARS-CoV-2 RBD.

a, Zero-charge mass spectrum of RBD acquired in positive ion mode with a UHMR Orbitrap mass spectrometer for an aqueous ammonium acetate (100 mM, pH 6.9) solution of SARS-CoV-2 RBD (10 μM). Deconvolution was performed using the Thermo BioPharma Finder Software^™. The glycan composition of each species is indicated (N ≡ HexNAc, H ≡ Hex, F ≡ Fuc, S ≡ Neu5Ac). b, Heatmap showing the distribution of the number of Neu5Ac and Fuc residues in the RBD glycoforms.

Extended Data Fig. 4 ∣. Comparison of glycan affinities for RBD measured by ESI-MS and their relative abundances measured by CaR-ESI-MS screening.

Comparison of glycan (15, 21, 27, 35, 36, 69, 71, 76 and 129) affinities for RBD measured by ESI-MS and their relative abundances measured by CaR-ESI-MS screening (performed at 25 °C). Black bars represent the ratio of the abundance of released glycans normalized to that of 71 under same conditions. White bars represent the ratios of dissociation affinities of RBD against 69 to dissociation affinities of RBD against glycan (L). ND, not detected. For relative abundances by CaR-ESI-MS, data represent ± s.d., n = 3 independent experiments. For K_d values, means ± s.d., n = 3 independent experiments for each glycan concentration.

Extended Data Fig. 5 ∣. CaR-ESI-MS screening ganglioside-containing nanodiscs against RBD.

a, ESI mass spectrum acquired in negative ion mode for an aqueous ammonium acetate solution (100 mM, 25 °C and pH 7.4) of RBD (10 μM) and 5 μM ganglioside-containing ND (1% GM1, GM2, GM3, GD1a, GD2, and GT1b, each at approximately 10 μM). b, CID mass spectrum acquired for ions centered at m/z 3,540, with a window width of 100 m/z units (the region highlighted in light blue in (a)) using a Trap collision energy of 50 V. c, CID mass spectrum acquired in negative ion mode for ions centered at m/z 3,540, with a window width of 100 m/z units, produced from an aqueous ammonium acetate solution (100 mM, 25 °C and pH 7.4) of and 10 μM ganglioside-containing ND; the Trap collision energy was 50 V.

Extended Data Fig. 6 ∣. ACE2 expression, RBD binding, and pseudoviral infection in HEK293 cells with and without ACE2.

a,b, Expression of human ACE2 (a) and trimeric fluorescent RBD binding (b) to WT (left) and ACE2⁺ (right) HEK293 cells as detected by flow cytometry. c, Flow cytometry gating of GFP-encoding pseudotyped SARS-CoV-2 lentivirus infection of WT (top) and ACE2⁺ (bottom) HEK293 cells.

Extended Data Fig. 7 ∣. Representative flow cytometry gating strategy and histogram for Fig. 4.

a-c, Side scatter (SSC) versus forward scatter (FSC) was used to discriminate against dead cells (SSC-A) and doublets (SSC-W). The singlets were then plotted in a histogram and the mean fluorescence intensity (MFI) was determined. Gating panels corresponding to Fig. 4a (a), 4b (b) and 4c (c). d, Representative gating strategy for Fig. 4d to determine % GFP⁺ cells in the pseudoviral infection assays when treated with 3FNeu5Ac. e, Representative flow cytometry gating and histogram for ACE2 binding in 3FNeu5Ac treated cells. This strategy corresponds to Fig. 4e.

Extended Data Fig. 8 ∣. Representative flow cytometry gating strategy and histogram for Fig. 5.

a, b, d, f, Side scatter (SSC) versus forward scatter (FSC) was used to discriminate against dead cells (SSC-A) and doublets (SSC-W). The singlets were then plotted in a histogram and the mean fluorescence intensity (MFI) was determined. Gating strategy of Genz-123346 treated cells (a, b) and UGCG⁺ or UGCG^−/− HEK293 ACE2 cells (d, e, f) used to plot Fig. 5b (a), 5c (b), 5f (d), 5g (e), and 5h (f). c,g, Representative gating strategy for Fig. 5d (c) and 5i (g) to determine % GFP⁺ cells in the pseudoviral infection of HEK293 ACE2 UGCG⁺ and UGCG^−/− cells.

Fig. 1 ∣. Glycan library screening and glycan affinities for RBD and S protein.

a,b, Normalized abundances of released glycans from the SARS-CoV-2 RBD and S protein by CaR-ESI-MS at 37 °C. Summary of the charge-normalized (relative to 69) abundances (Ab_rel) of released ligands (L) measured by CaR-ESI-MS screening of libraries A–O against SARS-CoV-2 RBD (a) and S protein (b). Measurements were performed in negative ion mode with an ultrahigh mass range (UHMR) Orbitrap mass spectrometer on aqueous ammonium acetate (100 mM, pH 6.9) solutions of RBD (10 μM) and glycan library (A–O) and library P (50 nM of each glycan) or S protein (280 μg ml⁻¹) and glycan library (A–O) and library P (500 nM of each glycan). A high-energy C-trap dissociation (HCD) energy of 50 V (using nitrogen as collision gas) with an isolation window of m/z 3,000–4,000 was used for screening against RBD; an HCD energy of 300 V (using argon) with an isolation window of m/z 8,000–15,000 was used for screening against S protein. The different classes of oligosaccharides are distinguished by color: red, Lewis antigens (1–14); yellow, blood group A antigens (15–20); orange, blood group B antigens (21–26); green, blood group H antigens (27–32); blue, sulfated compounds (34–38) and HS oligosaccharides (39–41); light blue, antigen-related glycans (42–50); purple, globo (51–60); pink, ganglioside oligosaccharides (61–76); white, HMOs and other glycans (77–128); light gray, Neu5Acα2-6-linked oligosaccharides (129–132); dark gray, rhamnose-containing compounds (133–139); NA, not applicable. An asterisk indicates that the relative abundances (in CaR-ESI-MS) were estimated from their relative (to 69) K_d. Data represent mean ± s.d.; n = 3 independent experiments. c, Affinities of glycans (15, 21, 27, 35, 36, 39–41, 69, 71, 76, 129 and 140) for RBD. K_d (μM) values were measured by ESI-MS using aqueous ammonium acetate (100 mM, pH 6.9) solutions containing RBD (5 μM) and each glycan (three different concentrations ranging from 30 to 150 μM) at 25 °C (white bars) and 37 °C (black bars); ND, not detected. Data represent mean ± s.d.; n = 3 independent experiments for each glycan concentration.

Fig. 2 ∣. Screening of natural N-glycan libraries against RBD.

a,c, CaR-ESI-MS screening results obtained for aqueous ammonium acetate solutions (100 mM, pH 6.9, 37 °C) of RBD (10 μM), P dimer of the Saga strain (P_ref, 4 μM) and the N-glycan library from lung (200 μg ml⁻¹) (a) and intestinal (125 μg ml⁻¹) (c) tissues. Ions with an m/z of 3,000–4,000 were subjected to HCD using a collision energy of 50 V. b,d, Chromatograms of 2-aminobenzamide (2-AB)-labeled N-glycans released from lung (b) and intestinal (d) tissues acquired using HILIC with fluorescence detection. The relative abundances of the Neu5Ac content of the N-glycans are indicated graphically. e, Heat maps of relative affinities of ligands identified by CaR-ESI-MS screening at 37 °C; ND, not detected.

Fig. 3 ∣. Inhibition of glycan binding to RBD by MeαNeu5Ac and HS oligosaccharides.

a, Scatter plot of the relative change in normalized abundances (to 71) of released neutral and acidic glycans in the presence and absence of MeαNeu5Ac. Measurements were performed in aqueous ammonium acetate solutions (100 mM, pH 6.9, 25 °C) of RBD (10 μM) and each of the defined libraries (50 nM of each glycan) with 0, 200, and 500 μM MeαNeu5Ac. Circles and diamonds represent neutral and acidic glycans, respectively, and solid lines and dashed lines represent the median values and first and third quartiles, respectively. Statistical significance was calculated based on an unpaired Mann–Whitney test because the distribution of data failed numerous normality tests (D’Agostino–Pearson omnibus, Shapiro–Wilk and Kolmogorov–Smirnov); n = 3 independent experiments for each glycan. b, Change in normalized abundance of released glycan 69 following the addition of glycan 41 or MeαNeu5Ac to the solution. Measurements were performed in aqueous ammonium acetate solutions (100 mM, pH 6.9, 25 °C) of RBD (5 μM) and 69 (50 nM) in the absence and presence of 41 (triangle) or MeαNeu5Ac (circle) at concentrations ranging from 0.1 to 100 μM (41) or 500 μM (MeαNeu5Ac). Ions in the range of m/z 3,000–4,000 were subjected to HCD using a collision energy of 50 V. Data represent mean ± s.d.; n = 3 independent experiments. c, Zero-charge mass spectra of RBD (5 μM) alone and in the presence of glycan 69 (25 μM) or glycan 41 (50 μM) or a mixture of 69 (25 μM) and 41 (50 μM). An asterisk indicates a sodium adduct. Experiments were performed in aqueous ammonium acetate solutions (100 mM, pH 6.9, 25 °C). Deconvolution was performed using the Thermo BioPharma Finder Software.

Fig. 4 ∣. Decreasing Sia on ACE2⁺ cells decreases RBD binding and SARS-CoV-2 pseudotyped lentiviral infection.

a,b, Changes in Sia levels from 3FNeu5Ac treatment of HEK293 ACE2⁺ cells determined by SNA (a) and PNA (b) lectin staining by flow cytometry. These n = 3 biological replicates are represented as mean fluorescence intensity (MFI) values with 95% confidence interval (95% CI) error. c, 3FNeu5Ac treatment decreases RBD binding. The MFI values for n = 3 biological replicates within a single experiment (left) and the average of n = 3 independent experiments (right) are shown. The average RBD binding to control cells was set to 100%. d, 3FNeu5Ac treatment decreases SARS-CoV-2 pseudovirus infection of HEK293 ACE2⁺ cells. The percentage of GFP⁺ cells for n = 3 biological replicates within a single experiment (left) and the averages of n = 3 independent experiments (right) are shown. The error bars associated with the single experiment are 95% CI. e, ACE2 expression level does not change following 3FNeu5Ac treatment. The MFI with 95% CI for n = 3 biological replicates is shown. f–j, Flow cytometry results for changes in Sia levels within CMAS^−/− ACE2⁺ HEK293 cells determined by SNA (f) and PNA (g) lectin staining, RBD binding (h), SARS-CoV-2 pseudovirus infection (i) and ACE2 expression (j) on CMAS⁺ versus CMAS^−/− ACE2⁺ HEK293 cells. Each point represents the average of n = 3 biological replicates for each individual clone with 95% CI error bars. k, SARS-CoV-2 pseudovirus infection of CMAS⁺ versus CMAS^−/− ACE2⁺ HEK293 cells treated with or without 3FNeu5Ac. The percentage of GFP⁺ cells for n = 3 biological replicates with 95% CI error bars is shown. l,m, Pretreatment of Vero-E6 cells (l) and ferret lung tissue sections (m) with neuraminidase decreases RBD binding. Scale bars represent 100 μm (l) and 200 μm (m), and results represent data from one representative sample from n = 5 experiments. n, Quantification of fluorescence intensity for RBD binding and lectin staining of Vero-E6 cells and ferret lung tissue sections. Error bars represent ±s.d. of n = 3 replicates. Statistical significance was calculated based on either two-tailed unpaired Student’s t-test (a–k) or analysis of variance (ANOVA) with multiple comparisons (l,m). Flow cytometry data were analyzed on FlowJo version 9.9.6, and graphs were plotted on GraphPad Prism 8; NS, not significant.

Fig. 5 ∣. Pharmacological and genetic abrogation of glycolipids decrease RBD binding and SARS-CoV-2 pseudoviral infection.

a,b, Changes in CTX binding on ACE2⁺ HEK293 cells after GENZ-123346 treatment determined by flow cytometry. Representative histogram results (a) and quantification of the MFI (b) are shown. Error bars represent 95% CI of n = 3 biological replicates. c,d, RBD binding to ACE2⁺ HEK293 cells (c) and SARS-CoV-2 pseudovirus infection in ACE2⁺ HEK293 cells (d) following GENZ-123346 treatment; n = 3 biological replicates are shown with 95% CI error bars. e, SARS-CoV-2 pseudovirus infection in CMAS⁺ and CMAS^−/− ACE2⁺ HEK293 cells following GENZ-123346 treatment. The percentage of GFP⁺ cells for n = 3 biological replicates and 95% CI error bars are shown. f–j, CTX staining (f), ACE2 expression levels (g), RBD binding (h) and percentage of GFP⁺ cells (i) of UGCG⁺ and UGCG^−/− ACE2⁺ HEK293 cell clones. Error bars represent 95% CI of n = 3 biological replicates. j, Pseudoviral infection of ACE2⁺ UGCG^−/− HEK293 cells with supplementation of glycolipids isolated from wild-type HEK293 cells. The percentage of GFP⁺ cells for n = 4 biological replicates within a single experiment (left) and the averages of n = 3 independent experiments (right) with error bars representing 95% CI are shown. Statistical significance was calculated based on two-tailed unpaired Student’s t-test (b–j). Flow cytometry data were analyzed on FlowJo version 9.9.6, and graphs were plotted on GraphPad Prism 8; NS, not significant.

Fig. 6 ∣. Pharmacological abrogation of glycolipids decreases authentic SARS-CoV-2 viral infection.

a,b, SARS-CoV-2 infection by two authentic SARS-CoV-2 viral strains, VIDO and 72b Calg, in ACE2⁺ HEK293 cells after GENZ-123346 treatment (a) or in UGCG⁺ or UGCG^−/− ACE2⁺ HEK293 cells (b). Shown is the viral titer of n = 3 biological replicates with 95% CI error bars. c, Infection of GENZ-123346-treated human primary nasal epithelial cells by 72b Calg virus. Open and solid symbols represent n = 2 biological replicates each containing n = 2 technical replicates at 0.1 and 0.5 multiplicity of infection (MOI). Viral titer was quantified by a plaque-forming unit (PFU) assay. Error bars represent 95% CI. Statistical significance was calculated using two-tailed unpaired Student’s t-test (a–c). Graphs were plotted using GraphPad Prism 8; NS, not significant.