SLO co-opts host cell glycosphingolipids to access cholesterol-rich lipid rafts for enhanced pore formation and cytotoxicity

Abstract

Streptolysin O (SLO) is a virulence determinant of group A Streptococcus (S. pyogenes), the agent of streptococcal sore throat and severe invasive infections. SLO is a member of a family of bacterial pore-forming toxins known as cholesterol-dependent cytolysins, which require cell membrane cholesterol for pore formation. While cholesterol is essential for cytolytic activity, accumulating data suggest that cell surface glycans may also participate in the binding of SLO and other cholesterol-dependent cytolysins to host cells. Here, we find that unbiased CRISPR screens for host susceptibility factors for SLO cytotoxicity identified genes encoding enzymes involved in the earliest steps of glycosphingolipid (GSL) biosynthesis. Targeted knockouts of these genes conferred relative resistance to SLO cytotoxicity in two independent human cell lines. Inactivation of ugcg, which codes for UDP-glucose ceramide glucosyltransferase, the enzyme catalyzing the first glycosylation step in GSL biosynthesis, reduced the clustering of SLO on the cell surface. This result suggests that binding to GSLs serves to cluster SLO molecules at lipid rafts where both GSLs and cholesterol are abundant. SLO clustering and susceptibility to SLO cytotoxicity were restored by reconstituting the GSL content of ugcg knockout cells with ganglioside GM1, but susceptibility to SLO cytotoxicity was not restored by a GM1 variant that lacks an oligosaccharide head group required for SLO binding, nor by a variant with a "kinked" acyl chain that prevents efficient packing of the ganglioside ceramide moiety with cholesterol. Thus, SLO appears to co-opt cell surface glycosphingolipids to gain access to lipid rafts for increased efficiency of pore formation and cytotoxicity.

Importance: Group A Streptococcus is a global public health concern as it causes streptococcal sore throat and less common but potentially life-threatening invasive infections. Invasive infections have been associated with bacterial strains that produce large amounts of a secreted toxin, streptolysin O (SLO), which belongs to a family of pore-forming toxins produced by a variety of bacterial species. This study reveals that SLO binds to a class of molecules known as glycosphingolipids on the surface of human cells and that this interaction promotes efficient binding of SLO to cholesterol in the cell membrane and enhances pore formation. Understanding how SLO damages human cells provides new insight into streptococcal infection and may inform new approaches to treatment and prevention.

Keywords: Streptococcus pyogenes; cholesterol-dependent cytolysin; glycan; glycosphingolipid; group A Streptococcus; lipid raft; receptor; streptolysin O; toxin.

Fig 1

An oligosaccharide that binds SLO in vitro inhibits SLO binding to human epidermal cells. (A) Representative flow cytometry profile for the binding of SLO to A431 cells in the absence or presence of oligosaccharide inhibitor lacto-N-neotetraose (LNnt), ganglioside asialo GM1 head group (A-GM1), or cellobiose. (B) Quantification of mean fluorescence intensity values for ≥3 independent replicates of the experiment illustrated in panel A. The mean value for SLO in the absence of inhibitor is set at 100%. ****P < 0.0001.

Fig 2

CRISPR-Cas9 screens identify glycosphingolipid biosynthetic enzymes as key host factors for SLO-mediated cytotoxicity. (A) Schematic of CRISPR screen workflow. (B–E) Results of four independent CRISPR screens in HAP1-Cas9 cells using native SLO (panels B and C) or SLO L562D (panels D and E). Each data point represents a single gene plotted as total sgRNA read counts on the x-axis versus number of unique sgRNA sequences on the y-axis. Selected genes involved in cholesterol homeostasis (blue) or glycosphingolipid biosynthesis (green) are indicated. See also Table 1. (F) Schematic of the earliest glycosylation steps in the biosynthesis of glucosyl glycosphingolipids. Enzymes encoded by genes identified in CRISPR-Cas9 screens for SLO susceptibility factors are labeled in red. UGCG, UDP-glucose ceramide glucosyltransferase; B4GALT5, β-1,4-galactosyltransferase 5; GALE, UDP-galactose-4-epimerase; GlcCer, glucosylceramide; and LacCer, lactosylceramide.

Fig 3

Targeted inactivation of glycosphingolipid biosynthetic genes confers resistance to SLO cytotoxicity. (A–C) Dose-response curves for SLO-mediated cytotoxicity of HAP1-Cas9 cell lines with targeted knockouts in individual genes encoding glycosphingolipid biosynthetic enzymes (red) compared to that for HAP1-Cas9 control cells (blue, repeated for clarity in each panel). (D) Dose-response curve for SLO-mediated cytotoxicity of a ugcg targeted knockout in A431 cells (red) compared to that for A431 control cells (blue). In panels A–D, dashed lines indicate results for cells treated with supplemental cholesterol prior to SLO exposure. (E) Resistance to SLO cytotoxicity of HAP1-Cas9 cell lines deficient in the expression of glycosphingolipid biosynthetic enzymes is recapitulated by challenge with live bacteria. Data represent mean ± SD percentage of surviving cells after exposure of cell monolayers to GAS strain 854. For all panels, data represent mean values ± SD for three independent experiments. *P < 0.05, **P < 0.01, and ****P < 0.0001 for comparison with control. Abbreviations are as in Fig. 2; KO, knockout.

Fig 4

Resistance of GSL mutants to SLO toxicity cannot be fully explained by gene deletion effects on membrane cholesterol or toxin binding. (A) Plasma membrane cholesterol is not significantly reduced in HAP1-Cas9 cell lines deficient in the expression of glycosphingolipid biosynthetic enzymes. Each data point represents mean fluorescence intensity of filipin staining of the cell surface from at least 25 cells per slide on ≥3 slides per cell sample without (red bars) or with (blue bars) the addition of supplemental cholesterol. (B) Representative flow cytometry profile of SLO binding to wild-type A431 cells or ugcg knockout A431 cells. (C) Quantification of mean ± SD fluorescence intensity values for three independent replicates of the experiment illustrated in panel B. (D) Representative flow cytometry profiles for the binding of SLO to HAP1-Cas9 control cells or HAP1-Cas9 cell lines with targeted knockouts in individual genes encoding glycosphingolipid biosynthetic enzyme. (E) Quantification of mean fluorescence intensity values for three independent replicates of the experiment illustrated in panel D. Abbreviations are as in Fig. 2 and 3.

Fig 5

Reconstitution of glycosphingolipid-deficient A431 cells with exogenous GM1 ganglioside increases SLO clustering on the cell surface. (A–D) Representative STORM images of SLO bound to the surface of wild-type A431 cells, the ugcg knockout, and the knockout supplemented with ganglioside GM1 C18:0 or with GM1 18:1 (A, B, C, and D, respectively). Cell monolayers were exposed to (non-hemolytic) SLO G395V G396V labeled with Alexa Fluor 647 and then imaged by STORM. Typical clusters of bound SLO are indicated by black arrows for each condition. Scale bar is 500 nm. (E–G) Clustering analysis of SLO binding to wild-type A431 cells or to ugcg knockout cells without or with supplemental ganglioside GM1. Representative curves from five independent replicates of the experiments illustrated in panels A–D are plotted according to Ripley’s K-function analysis of clustering of labeled SLO molecules into domains of radius r on the x-axis versus L(r) – r on the y-axis, in which for a given radius r, L(r) is the radius within which the number of points would be distributed if the distribution were completely random. Thus, the maximum of each curve corresponds to the cluster diameter containing the highest density of SLO molecules. (E) Vertical lines mark the maximum of curves corresponding to SLO binding to wild-type A431 cells (black) or ugcg knockout cells (red). (F and G) A vertical line marks the maximum of a curve corresponding to SLO binding to ugcg KO cells supplemented with ganglioside GM1 18:0 (panel F, blue) or GM1 18:1 (panel G, cyan). The curve from panel E for unsupplemented ugcg knockout cells is included for comparison (red). (H) Summary data for experiments illustrated in panels E–G. Statistically significant differences between groups are indicated. *P < 0.05 and ***P < 0.001. (I) Schematic of ganglioside GM1 with a fully saturated fatty acyl chain (C18:0) or with a single double bond at position 9 (C18:1). The GM1 oligosaccharide head group is represented by “R” (drawing modified from reference 33). An adjacent cholesterol molecule (shown in gray) packs efficiently with native GM1 C18:0 but not with GM1 C18:1 due to the bend in its acyl chain. (J) For SLO clustering experiments in panels F and G, glycosphingolipid-deficient cells were loaded with 5 µM GM1 C18:0 (blue) or 0.5 µM GM1 C18:1 (green), which resulted in similar levels of cell surface GM1 expression. Data represent mean fluorescence intensity as assessed by flow cytometry after staining cells with cholera toxin B-Alexa Fluor 488, which binds to ganglioside GM1.

Fig 6

Reconstitution of glycosphingolipid-deficient A431 cells with exogenous GM1 ganglioside restores susceptibility to SLO cytotoxicity. A431 ugcg knockout cell monolayers were treated with the indicated concentration of glycan prior to exposure to 0.6 nM SLO. Glycans are (A) ganglioside GM1 C18:0, (B) glucosylceramide, or (C) ganglioside GM1 C18:1. **P < 0.01 and ****P < 0.0001 for comparison to no added glycan.