Requirement of N-glycan on GPI-anchored proteins for efficient binding of aerolysin but not Clostridium septicum alpha-toxin

Abstract

Aerolysin of the Gram-negative bacterium Aeromonas hydrophila consists of small (SL) and large (LL) lobes. The alpha-toxin of Gram-positive Clostridium septicum has a single lobe homologous to LL. These toxins bind to glycosylphosphatidylinositol (GPI)-anchored proteins and generate pores in the cell's plasma membrane. We isolated CHO cells resistant to aerolysin, with the aim of obtaining GPI biosynthesis mutants. One mutant unexpectedly expressed GPI-anchored proteins, but nevertheless bound aerolysin poorly and was 10-fold less sensitive than wild-type cells. A cDNA of N-acetylglucosamine transferase I (GnTI) restored the binding of aerolysin to this mutant. Therefore, N-glycan is involved in the binding. Removal of mannoses by alpha-mannosidase II was important for the binding of aerolysin. In contrast, the alpha-toxin killed GnTI-deficient and wild-type CHO cells equally, indicating that its binding to GPI-anchored proteins is independent of N-glycan. Because SL bound to wild-type but not to GnTI-deficient cells, and because a hybrid toxin consisting of SL and the alpha-toxin killed wild-type cells 10-fold more efficiently than GnTI- deficient cells, SL with its binding site for N-glycan contributes to the high binding affinity of aerolysin.

Fig. 1. Different expressions of GPI-anchored proteins on three aerolysin-resistant mutant cells. Cells were stained for CD59 and DAF. (A) Control, wild-type CHO cells stained with isotype-matched non-relevant antibodies. (B) CHO(wt), wild-type CHO cells. (C) GPI(–).O, GPI-anchor-deficient CHO cells. (D) GPI(–).U, CHO cells defective in GPI transamidase. (E) GPI(+), GPI-anchor-sufficient, aerolysin-resistant CHO cells.

Fig. 2. Aerolysin-resistant GPI(+) cells are not defective in GPI biosynthesis. Lipids extracted from cells metabolically labeled with [³H]mannose were analyzed by TLC in a solvent consisting of CHCl₃:MeOH:H₂O = 10:10:3. Mannolipids termed according to Hirose et al. (1992) are indicated. Lane 1, mutant GPI(–).O; lane 2, mutant GPI(–).U; lane 3, mutant GPI(+); lane 4, wild-type CHO(wt) cells.

Fig. 3. The aerolysin resistance of GPI(+) cells was due to the inefficient binding of the toxin. (A) Viabilities of mutant CHO cells after treatments with increasing concentrations of aerolysin. Percent cell viability measured by MTT assay is shown as a function of the proaerolysin concentration. (B) Binding of fluorescent-tagged proaerolysin (FLAER) to mutant cells. Cells incubated with FLAER at various concentrations (shown on the right) were analyzed by flow cytometry. (a) CHO(wt); (b) GPI(+); (c) GPI(–).U cells.

Fig. 4. GPI(+) cells are defective in GnTI. (A) GnTI cDNA restored binding of FLAER to GPI(+) cells. GPI(+) cells transfected with rat GnTI cDNA (solid line) or a mock vector (dotted line) were stained with 5 nM FLAER and analyzed by FACS. (B) Lec1 CHO cells, an authentic GnTI mutant, are defective in aerolysin binding, similar to GPI(+) mutant. Lec1 cells transiently transfected with GnTI cDNA (a and c) or a mock vector (b and d) were incubated with 10 µg/ml FITC–PHA-P (a and b) or 5 nM FLAER (c and d). Thin lines, untransfected Lec1 cells; bold lines, transfectants.

Fig. 5. GPI(+) cells are defective in maturation of N-glycan. (A) Expression of a smaller CD59 by GPI(+) mutant cells. CD59 immunoprecipitates from NP-40 extracts of wild-type CHO(wt) (lane 1), GPI(+) mutant (lane 2) and GPI(–).U mutant (lane 3) cells were analyzed by western blotting with anti-CD59 mAb. Size makers are shown on the left. (B) The smaller sized CD59 in GPI(+) cells was due to an abnormal N-glycan not GPI anchor. Left panel: wild-type CHO cells (lanes 1 and 3) and GPI(+) mutant CHO cells (lanes 2 and 4) were transfected with a FLAG-tagged GPI-anchored form of CD59 (lanes 1 and 2) or its non-N-glycosylation mutant (N43A) (lanes 3 and 4). Two days later, proteins were immunoprecipitated by anti-CD59 mAb, followed by western blotting against anti-FLAG mAb. Right panel: a FLAG-tagged transmembrane form of CD59 was transfected instead of the FLAG-tagged GPI-anchored form. FLAG-tagged transmembrane CD59 was transfected into wild-type (lane 5) and GPI(+) mutant (lane 6) CHO cells.

Fig. 6. Inefficient binding of proaerolysin to cells treated with swainsonine. To inhibit α-mannosidase II, wild-type CHO cells were treated with 10 µg/ml swainsonine or PBS for 3 days and then incubated with 10 µg/ml FITC-conjugated PHA-P (A) and 5 nM FLAER (B). Solid lines, swainsonine-treated cells; dotted lines, buffer-treated cells.

Fig. 7. Deficiency of GnTII does not affect binding of proaerolysin. GPI-deficient B-lymphoblastoid cells (JY5), the wild-type counterpart (JY25) and B-lymphoblastoid cells from a patient with GnTII deficiency [GnTII(–)] were incubated with PHA-P (A–C) and FLAER (D–F) at various concentrations (shown on the right).

Fig. 8. Normal binding of aerolysin to Lec2 and Lec8 cells. (A and B) Binding of fluorescent-tagged PHA-P (10 µg/ml) to Lec2 (A) and Lec8 (B) cells. Bold lines, PHA-P; thin lines, PBS. (C and D) FLAER binding to Lec2 and Lec8 cells. Cells were incubated with 5 nM FLAER. Bold lines, FLAER; thin lines, PBS.

Fig. 9. N-glycan of GPA is the binding determinant for aerolysin. (A) GPA and asialo GPA inhibited binding of FLAER to CHO cells. FLAER (5 nM) was pre-incubated with GPA, asialo-GPA (0.5 mg/ml) or buffer for 10 min, and then incubated with CHO cells. Thin lines, buffer-treated FLAER; bold lines, GPA-treated (a) and asialo GPA-treated (b) FLAER. (B) Treatment with PNGase F abolished the inhibitory activity of asialo GPA. Samples of asialo-GPA (0.3 mg/ml) treated with PNGase F or buffer only were incubated with FLAER (5 nM) for 10 min. The mixtures were then incubated with CHO cells. Thin lines, binding of non-treated FLAER; bold lines, binding of FLAER in the presence of PNGase F-treated asialo GPA (a) or buffer-treated asialo GPA (b).

Fig. 10. Binding of SL to N-glycan of GPI-anchored proteins. (A) Fluorescent-tagged SL bound to wild-type CHO cells (thin line) but not to GnTI-deficient GPI(+) cells (bold line) and GPI(–).U cells (dotted line). (B) SL recognized the same binding determinant as intact aerolysin. CHO(wt) cells were incubated with biotinylated proaerolysin (10 nM) (bold line) or PBS (dotted line) plus streptavidin–PE in the first step (a). In the second step, the cells were incubated with 10 µM fluorescent-tagged SL (b; bold line, cells pre-treated with biotinylated aerolysin; dotted line, cells pre-treated with PBS).

Fig. 11. No requirement of N-glycan for the binding of C.septicum α-toxin. (A) GPI(+) cells were as sensitive to α-toxin as CHO(wt) cells. Percent viability is plotted as a function of the α-toxin concentration. (B) Efficient binding of α-toxin to GPI(+) cells. CHO(wt) cells (a), GPI(+) mutant (b) and GPI(–).U mutant (c) cells were incubated with various concentrations of fluorescent-tagged α-toxin.

Fig. 12. CHO(wt) cells were 10-fold more sensitive to hybrid toxin consisting of SL and α-toxin than GPI(+) cells. CHO(wt), GPI(+) mutant and GPI(–).U mutant cells were incubated with the hybrid toxin. Percent cell viability determined by MTT assay is plotted as a function of the toxin concentration.