Polar effect of staphylococcal α-toxin on epithelial cells
To test whether MDCK and CaCoII cells were sensitive to α-toxin, we measured the effect of the toxin on the transepithelial resistance (TER) of cell monolayers. The toxin was added to filter grown cells either only to the apical or only to the basolateral compartment. To our surprise, both polarized cell types were sensitive to the toxin only when exposed on the basolateral side (Fig. 1).
To get a more direct estimate of the pore-forming activity of α-toxin on these polarized cells, we measured the change in intracellular potassium. If α-toxin forms a channel in the apical or basolateral plasma membrane, this should lead to potassium efflux across this membrane. Once more, α-toxin was added either apically or basolaterally to filter grown cells. As can be seen in figure 2 (upper panel), potassium efflux was significant only when α-toxin was added to the basolateral membrane of either MDCK or CacoII cells. The kinetics of channel formation were fast, since cells were essentially depleted of potassium within 1 h of exposure to a 500 ng/ml α-toxin concentration as shown for MDCK cells.
The basolateral membrane of a polarized epithelial cell is often considered as equivalent to the plasma membrane of a non-polarized cell, the apical membrane being considered more specialized. We therefore expected non-polarized MDCK cells to be sensitive to α-toxin. Cells were platted at high dilution (1/10) in order to prevent cell contacts, and potassium efflux measurements were performed. To our surprise, non-polar MDCK cells were insensitive to α-toxin (not shown because equal to controls). To test whether basolateral sensitivity required the full establishment of polarity, cells were platted on filters at a 1:1 dilution and the sensitivity to basolateral α-toxin was measured every day for 4 days. As shown in figure 2 (lower panel), 1 day after platting, MDCK cells were already sensitive to basolateral α-toxin. These observations suggest that cell contact and formation of a tight monolayer, but not full polarization which requires approx. 4 days for MDCK cells, is required for MDCK cells to exhibit basolateral sensitivity to α-toxin.
Differential binding of α-toxin to the apical and basolateral membranes of MDCK cells
We next monitored whether binding occurred preferentially on the basolateral side. As shown in figure 3, binding to the apical side is far lower than binding to the basolateral side. Binding to the basolateral side was specific since it could be competed with a 50-fold excess of unlabeled toxin (Fig. 3B). The fact that a 50-fold excess was sufficient to inhibit binding by more than 80% also indicates that there were a limited number of binding sites.
Channel formation by α-toxin is inhibited upon cholesterol removal
It has previously been observed by us and by others that certain bacterial pore-forming toxin such as aerolysin and Clostridum α-toxin as well as related toxins such as Helicobacter pylori VacA or anthrax toxin protective antigen, preferentially associate with raft-like domains. To investigate whether the mode of action of Staphylococcal α-toxin also involves cholesterol-rich domains, we analyzed the effect of cholesterol depletion using β-methyl-cyclodextrin (β-MCD). Removal of cholesterol from the basolateral side of the MDCK monolayer led to a drastic inhibition of the α-toxin-induced potassium efflux (Fig. 4A; since β-MCD itself led to a decrease in intracellular potassium, results were expressed as a percentage of potassium at time 0). We then investigated what step in the mode of action was affected and found that binding of α-toxin to β-MCD-treated cells was essentially abolished (Fig. 4C). Since β-MCD could also affect cell polarity and thereby explain the loss of sensitivity, we tested the effect of cholesterol depletion by β-MCD on a non-polarized α-toxin-sensitive cell line, namely HeLa. As for MDCK cells, we found that β-MCD inhibited the toxin-induced potassium efflux (Fig. 4B) and toxin binding (Fig.4C).
The effect observed upon β-MCD treatment was due to the removal of cholesterol and not to the concomitant release of another important factor since reloading of cholesterol restored α-toxin sensitivity (Fig. 5A, B). These observations raise the possibility that as for certain cholesterol-dependent toxins (CDTs), cholesterol could be the receptors. This seems unlikely since preincubation of α-toxin with cholesterol did not affect the activity in agreement with previous observations on liposomes. Pretreatment with cholesterol has been shown to inhibit pore formation by certain cholesterol-binding CDTs.
The cholesterol dependence of α-toxin binding raises the possibility that it binds to raft-like domains. As a rough estimate of this, we probed for the association of α-toxin with detergent-resistant membranes. Polarized MDCK cells were treated with α-toxin from the basolateral side, solubilized in Triton X-100 at 4°C and detergent-soluble material was separated by detergent-insoluble material by high-speed centrifugation. As shown in figure 6, α-toxin was mainly associated with the detergent-soluble fraction in contrast to aerolysin, which was here used as a control, suggesting that unlike aerolysin, alpha toxin preferentially binds to detergent-soluble membranes.
Inhibition of glycosphingolipid synthesis leads to an increase in α- toxin sensitivity
To further explore the possibility that α-toxin binds to lipid rafts, we analyzed the effect of depleting cells of another major component, namely glycosphingolipids. For this we made use of a mouse melanoma cell line MEB4 and its glycolipid-deficient mutant GM95. GM95 has a defect in ceramide glucosyltransferase I (CerGlcTI) that catalyzes the first step of glycosphingolipid synthesis. As shown in figure 7A, MEB4 cells were insensitive to α-toxins, but surprisingly GM95 cells showed sensitivity, which was lost upon recomplementation of the cells with the CerGlcTI gene (CG1 cells). Differences between these different cell lines could be partly due to charges in lipid composition, in addition to the lack of glycolipids. This behavior is similar to that observed for the earthworm pore-forming toxin lysenin. Interestingly lysenin also shows a polarized effect on epithelial cells, with only the basolateral membrane being sensitive. α-toxin and lysenin, however, do not bind to the same sites at the target cell surface since we found no effect of cholesterol depletion on pore formation by lysenin (Fig. 7B, right panel).
Lysenin is well known to bind to sphingomyelin, but to only special arrangements of sphingomyelin that are perturbed by the presence of glycosphingolipids (which would explain why these sphingomyelin arrangements are preferentially found on the basolateral membrane of polarized cells). It could be that α-toxin also binds to arrangements of sphingomyelin, yet different from those recognized by lysenin. α-toxin was indeed found not to bind to sphingomyelin vesicles unless cholesterol was present. We therefore tested the effect of sphingomyelinase treatment on the α-toxin-induced potassium efflux. While sphingomyelinase treatment abolished the sensitivity of cells to lysenin, it only had a mild effect on the sensitivity to α-toxin (Fig. 7B, left panel), indicating that α-toxin does not bind to sphingomyelin, or to a pool of sphingomyelin that is not removed by the treatment.
Studies in the late 1980’s pointed towards a requirement of phosphocholine-containing lipid head groups for α-toxin binding and activity using artificial liposomes. The crystal structure of the α-toxin heptamer revealed the presence of a phosphocholine binding pocket between the rim and stem domains of the pore-forming complex. To test the importance of phosphocholine-binding in pore-forming activity of target cells, we analyzed the effects of phosphocholine chloride on the toxin-induced efflux from HeLa cells, using choline chloride as a control. We found that increasing concentrations of phosphocholine led to a gradual inhibition of toxin activity (Fig. 8).