Siglec-6 mediates the uptake of extracellular vesicles through a noncanonical glycolipid binding pocket

Abstract

Immunomodulatory Siglecs are controlled by their glycoprotein and glycolipid ligands. Siglec-glycolipid interactions are often studied outside the context of a lipid bilayer, missing the complex behaviors of glycolipids in a membrane. Through optimizing a liposomal formulation to dissect Siglec-glycolipid interactions, it is shown that Siglec-6 can recognize glycolipids independent of its canonical binding pocket, suggesting that Siglec-6 possesses a secondary binding pocket tailored for recognizing glycolipids in a bilayer. A panel of synthetic neoglycolipids is used to probe the specificity of this glycolipid binding pocket on Siglec-6, leading to the development of a neoglycolipid with higher avidity for Siglec-6 compared to natural glycolipids. This neoglycolipid facilitates the delivery of liposomes to Siglec-6 on human mast cells, memory B-cells and placental syncytiotrophoblasts. A physiological relevance for glycolipid recognition by Siglec-6 is revealed for the binding and internalization of extracellular vesicles. These results demonstrate a unique and physiologically relevant ability of Siglec-6 to recognize glycolipids in a membrane.

Fig. 1. Optimizing a liposome formulation for probing Siglec–ganglioside interactions.

a Schematic of the cell assay where binding between Siglec-1-expressing CHO cells and fluorescently labeled GLLs are quantified via flow cytometry. b Representative flow cytometry histograms and quantification of 3 mol% GM1 liposomes with varying mol% of PEG₄₅–DSPE to Siglec-1 CHO cells (n = 5 technical replicates). c Binding of liposomes formulated with increasing ganglioside (GM1, GM2, GM3, GD1a) content against Siglec-1 CHO cells (n = 4 technical replicates). d Mass spectrometry-derived dissociation constant for the interaction between soluble Siglec-1 and ganglioside (GM1, GM2, GM3, GD1a) oligosaccharides (n = 4 technical replicates). e Ganglioside (GM1, GM2, GM3, GD1a) oligosaccharide density titration using a liquid glycan array (LiGA) against Siglec-1 CHO cells (5 ≥ n ≥ 4 technical replicates). f Liposomes formulated with GM1 and GA1 binding to Siglec-1-expressing CHO cells in the cell assay (n = 3 technical replicates). Glycan structures are represented using Symbol Nomenclature for Glycans (SNFG; blue circle, glucose; yellow circle, galactose; yellow square, GalNAc; purple diamond, Neu5Ac). Data are represented as the mean ± one standard deviation of at least three technical replicates. For panels b and f a Brown–Forsythe and Welch one-way ANOVA was used for statistical analysis. Not Significant (NS), P > 0.5.

Fig. 2. Investigating Siglec–ganglioside interactions using optimized liposomes reveals interactions.

a Heatmap summarizing the binding interactions of each human Siglec (4 ≥ n ≥ 3 technical replicates). Color is representative of the log₁₀(MFI_AF647) of each GLL subtracted from the log₁₀(MFI_AF647) of the same GLL against UT CHO cells. † Denotes binding was measured after treatment of cells with neuraminidase. b Schematic of the enzyme-linked immunosorbent assay (ELISA) used to investigate Siglec–ganglioside interactions outside the context of a lipid bilayer. c ELISA results of Siglec-1, -6, and -7 binding to nine gangliosides (n = 4 technical replicates). d Schematic of the liposome over lectin assay (LOLA) where binding is read out on a fluorescence plate reader. e Results of the LOLA against a select number of GLLs (n = 5 technical replicates). f Schematic of the bead assay where binding is read out by flow cytometry. g Results of the bead assay against a select number of GLLs (n = 4 technical replicates). All results are represented as the mean ± one standard deviation of at least three technical replicates. The dotted line on the ELISA, LOLA, and bead assay results represents two standard deviations above the blank well background for the ELISA and a naked liposome for the LOLA and bead assay. For panels c, e, and g, a Brown–Forsythe and Welch one-way ANOVA was used. Not Significant (NS), P > 0.5; *0.05 > P ≥ 0.01; ** 0.01 > P ≥ 0.001; ***0.001 > P ≥ 0.0001; ****P < 0.0001.

Fig. 3. Exploring the glycolipid binding specificity of Siglec-6 using nGLLs.

a Schematic of a glycolipid structure broken down into three components. b Structures of nGLs 3 and 4 presenting the oligosaccharide of GM1 and GM3, respectively, through an amide-linkage to a 1,3-di-O-hexadecyl glycerol scaffold. c Binding of liposomes formulated with 3 mol% 3 and 4 to WT Siglec-6 CHO cells in the cell assay, relative to liposomes formulated without a ligand (n = 4 technical replicates). d, Structures of nGLs 5, 6, and 7 presenting the oligosaccharide of GM3 triazole-linked to 1,3-di-O-hexadecyl glycerol, phosphatidyl sphingomyelin, and distearoylphosphatidylcholine scaffold, respectively. e Binding of liposomes formulated with nGLs 5, 6, and 7 to WT Siglec-6 CHO cells in the cell assay relative to naked liposomes (n = 4 technical replicates). f Structures of nGLs 8, 9, and 10 presenting an α-(2 → 3)- or α-(2 → 6)-linked sialoside on an underlying lactose or LacNAc core, triazole-linked to 1,3-di-O-hexadecyl glycerol scaffold. g, Binding of liposomes formulated with 5 and 8−10 to WT and R122A Siglec−6 CHO cells in the cell assay relative to liposomes formulated with 3 mol% nGL 5 (n = 4 technical replicates). Data is representative of the mean ± one standard deviation of four technical replicates. For panel c, a two-tailed Student’s t-test was used for statistical analysis. For panels e and g, a Brown–Forsythe and Welch one-way ANOVA was used for statistical analysis.

Fig. 4. Identifying the location of the noncanonical glycolipid binding pocket in Siglec-6.

a AlphaFold-generated molecular model of Siglec-6 extracellular domains. b Binding of 5 nGLLs to CHO cells expressing each Siglec-6/8 chimera. The dotted line represents the average binding of 5 nGLLs to UT CHO cells (n = 4 technical replicates). Relative Siglec staining is defined by the amount of anti-Siglec staining of the chimeric cell line over the anti-Siglec staining of the WT cell line. c Expansion of the molecular model of Siglec-6 at the interface of the V-set domain and the first C2 domain, with residues of interest highlighted as sticks (green, canonical arginine residue; red, noncanonical arginine residue; orange, cysteine residues involved in the interdomain disulfide bridge, non-arginine residues; yellow). d and e nGLLs formulated with 5 binding to CHO cells expressing Siglec-6 mutants (n = 4 technical replicates). Liposome binding data is representative of the mean ± one standard deviation of four technical replicates. For panels b and d, a Brown–Forsythe and Welch one-way ANOVA was used for statistical analysis. For panels d and e, a statistical comparison was between each mutant and WT Siglec-6. Not Significant (NS) P > 0.5.

Fig. 5. Targeting Siglec-6 on human cells and tissues.

a Siglec-6 expression levels on primary mast cells isolated from healthy human spleen donors and LAD2 cells. For the primary mast cells, each datum represents a biological replicant (n = 5 biological replicates). b Binding of 5 nGLLs to LAD2 after treatment with an anti-Siglec-6 antibody quantified by flow cytometry (n = 3 technical replicates). c and d 5 liposome binding to human naïve and memory B-cells isolated from the blood of four healthy biological replicants (represented with Roman numerals) without and with pretreatment of Siglec-6 blocking antibody, respectively (4 ≥ n ≥ 3 technical replicates). e Representative confocal microscopy images of human syncytiotrophoblasts after treatment with naked and 5 nGLLs. Quantification of the total number of naked liposomes and 5 nGLLs binding to human placental explants per µm³ (f n = 4 biological replicates) and the number of 5 nGLL colocalized with Siglec-6 per µm³ (g n = 3 biological replicates). h Binding of 5 nGLLs to human syncytiotrophoblasts after treatment with an anti-Siglec-antibody (n = 3 biological replicates). For panels f, g, and h each datum is representative of a different donor, which introduces variability between panels. All results are represented by the mean ± one standard deviation of at least three technical replicates. For panels a, c, d, f, g, and h, a two-tailed Student’s t-test was used for statistical analysis. For panel b, a Brown–Forsythe and Welch one-way ANOVA was used for statistical analysis. Not Significant (NS) P > 0.5.

Fig. 6. Siglec-6 binds and internalizes extracellular vesicles through glycolipids independent of its conserved arginine residue.

a Binding of EVs to Siglec-6 expressing CHO cells pre-treated with an anti-Siglec-6 antibody (n = 3 technical replicates). b Binding of EVs to UT CHO cells and CHO cells expressing C46A, C172A, WT, R122A, and R92K Siglec-6 (4 ≥ n ≥ 3 technical replicates). c, Binding of EVs to WT and R92K Siglec-6 in the bead assay (n = 4 technical replicates). d Blocking of EV binding to Siglec-6 with 5 mol% 5 nGLLs in the bead assay (n = 4 technical replicates). e Binding of EVs isolated from two different donors to LAD2 cells pretreated with an anti-Siglec-6 antibody (n = 4 technical replicates). f Binding of neuraminidase A and neuraminidase S treated EVs to Siglec-6 in the bead assay (n = 3 technical replicates). g Binding of EVs isolated from WT and β1-4GalNT1^−/− N2a cells to WT Siglec−6 in the bead assay (n = 4 technical replicates). h Time-dependent fluorescence of pHrodo labeled EVs incubated with Daudi cells transduced with WT, R122A Siglec-6, and an empty vector (n = 3 technical replicates). i Representative imaging flow cytometry images of empty vector and WT Siglec-6 virally transduced Daudi cells incubated with AF488 labeled EVs at 4 or 37 °C with the EV fluorescence overlaid over the brightfield image. Scale bars represent 7 µm. j Quantification of internalization of EVs at 4 or 37 °C by Daudi cells transduced with WT Siglec-6 and an empty vector (n = 4 technical replicates). Data is represented by the mean ± one standard deviation of at least three technical replicants. For panel a (WT, C46A, R122A, C172A), b, d, f, g, h, and j a Brown–Forsythe and Welch one-way ANOVA was used for statistical analysis. For panels a (WT vs. R92K) c and e a two-tailed Student’s t-test was used for statistical analysis. Not Significant (NS), P > 0.5.