Oligomerization of Vibrio cholerae Cytolysin Yields a Pentameric Pore and Has a Dual Specificity for Cholesterol and Sphingolipids in the Target Membrane*

Vibrio cholerae cytolysin permeabilizes animal cell membranes. Upon binding to the target lipid bilayer, the protein assembles into homo-oligomeric pores of an as yet unknown stoichiometry. Pore formation has been observed with model liposomes consisting of phosphatidylcholine and cholesterol, but the latter were much less susceptible to the cytolysin than were erythrocytes or intestinal epithelial cells. We here show that liposome permeabilization is strongly promoted if cholesterol is combined with sphingolipids, whereby the most pronounced effects are observed with monohexosylceramides and free ceramide. These two lipid species are prevalent in mammalian intestinal brush border membranes. We therefore propose that, on its natural target membranes, the cytolysin has a dual specificity for both cholesterol and ceramides. To assess the stoichiometry of the pore, we generated hybrid oligomers of two naturally occurring variants of the toxin that differ in molecular weight. On SDS-polyacrylamide gel electrophoresis, the mixed oligomers formed a pattern of six distinct bands. Ordered by decreasing electrophoretic mobility, the six oligomer species must comprise 0 to 5 subunits of the larger form; the pore thus is a pentamer. Due to both lipid specificity and pore stoichiometry, V. choleraecytolysin represents a novel prototype in the class of bacterial pore-forming toxins.

Vibrio cholerae O1 El Tor is the causative agent of the current seventh cholera pandemic. The severe purging of cholera is due to the effects of cholera enterotoxin upon the small intestine epithelial cells (1). Apart from enterotoxin, many strains of V. cholerae El Tor secrete a hemolytic toxin, V. cholerae cytolysin (VCC). 1 The latter is also produced by most V. cholerae non-O1 strains, which may cause intestinal and extraintestinal infections (2).
In vitro, V. cholerae cytolysin acts upon a variety of target cells, among these are enterocytes and immune cells (3)(4)(5). A pathogenetic role for VCC in enteritis is suggested by the observation that the purified protein elicits fluid accumulation in rabbit ileal loops (6,7). Homologous cytolysins are secreted by various Aeromonas and Vibrio species that may elicit diarrheal infections and by Vibrio vulnificus, a related organism that causes wound and septicemic infections (8).
During secretion of VCC, a signal peptide of 25 amino acids is cleaved (9,10). The extracellular procytolysin (79 kDa) requires further proteolytic activation (3,11), which can be accomplished by a variety of proteases (12) and occurs by removal of an N-terminal fragment of 14 kDa; the latter is important for proper folding of the molecule (13). The mature VCC of 65 kDa may undergo an additional proteolytic cleavage close to the C terminus to yield a second active species of about 50 kDa (4).
Recently, V. cholerae cytolysin has been identified as an oligomerizing, pore-forming protein (4,14,15). The process of pore formation involves two separate steps. The water-soluble monomer first binds reversibly to the target membrane (15). Subsequently, an unknown number of membrane-associated monomers assemble into an oligomer that inserts into the membrane to surround a water-filled pore of 1.5 nm diameter (4,16). Binding is essentially unsaturable and of low specificity, since the cytolysin readily associates with phosphatidylcholine liposomes (15). However, in this synthetic system, membrane permeabilization has only been observed when phosphatidylcholine (PC) was supplemented with cholesterol (14). Even then, the liposomes thus obtained lag far behind animal cell membranes in sensitivity; while rabbit erythrocytes and human enterocytes are lysed at cytolysin concentrations of 170 pM (5,15), a 1000 times higher amount had to be employed for permeabilization of PC/cholesterol liposomes (14). Since V. cholerae cytolysin does not detectably interact with rabbit erythrocyte membrane proteins (15), we reasoned that the superior susceptibility of natural membranes should be accounted for by lipid constituents other than cholesterol. Characterization of these incremental lipid species should both contribute to understanding the molecular properties of the cytolysin and allow for improved liposome models useful in further studies on this interesting toxin.
As an initial confirmation of our hypothesis, we found that the susceptibility of liposomes was strongly enhanced when PC was replaced by a crude lipid extract of bovine brain. When the constituents of this lipid mixture were tested singly, liposome sensitization was apparent with sphingolipids, particularly galactosylceramide, whereas only minor effects were observed with glycerophospholipids. Comparison of various sphingolipids indicates that the element essential for interaction with the cytolysin consists in the common ceramide moiety.
In the second part of this study, the brain lipid liposome model was utilized to examine the stoichiometry of the cytolysin oligomer. To this end, we employed two naturally occurring forms of the cytolysin that differ in molecular weight but are equally endowed with the capability to form oligomeric pores. When these two forms were simultaneously applied to liposomes, a mixture of oligomers was formed that could be resolved into six distinct species by SDS-PAGE; from this pattern, a pentameric stoichiometry of the oligomeric pore can be inferred. The results of this study are relevant to both molecular and pathogenetic aspects of V. cholerae cytolysin.

MATERIALS AND METHODS
Bacterial Strains-V. cholerae O1 El Tor strain 8731, which was used for isolation of the VCC 65-kDa form, was a generous gift of Dr. R. Hall, Washington D. C. The 50-kDa form was isolated from V. cholerae O1 El Tor strain KM 169 (4).
Purification of V. cholerae Cytolysin-Both forms of VCC were isolated according to published procedures. Briefly, the 65-kDa form was precipitated from culture supernatants of strain 8731 with ethanol and purified by sequential isoelectric focusing in sucrose density gradients and hydroxyapatite chromatography (15). For the isolation of the 50-kDa form, V. cholerae strain KM 169 was cultured in liquid minimal medium. The cytolysin was purified by ammonium sulfate fractionation and sequential chromatography on DE52 cellulose (Whatman), Ultrogel AcA-44 (Amersham Pharmacia Biotech), and a Mono Q column (Amersham Pharmacia Biotech) (7).
Preparation of Liposomes-Cholesterol, phosphatidylethanolamine (PE), and PC from egg yolk, ceramide, D-sphingosine, ganglioside G M1 , and bovine brain extract (composed of phosphatidylserine (PS) 10 -15%, PC 15-20%, PE 20 -25%, sphingomyelin (SPM) 10 -15%, glycoceramides 30 -40%, cholesterol 1-2%) were purchased from Fluka AG, Buchs, Switzerland. The enumerated constituents of the brain extract were also purchased singly in purified form (purity Ն98%) from either Sigma or Fluka; glucosylceramide from human spleen was obtained from Sigma. The lipids were dissolved in chloroform with or without 33% methanol and admixed at the desired molar ratios (see "Results") in a 250-ml round bottomed flask. The solvent was evaporated under a stream of nitrogen, and the lipid film was dried for 30 min under vacuum. Following resuspension of the lipids in Hepes/NaCl to 5 mg/ml, large unilamellar vesicles (LUV) were formed by repeatedly extruding the suspension through polycarbonate membranes (Nuclepore, CA; 100-nm pore size) (17), whereby the extrusion apparatus (Lipex Biomembranes, Vancouver, Canada) was thermostatted at 40 or 50°C if required. The lipid concentration in the final sample was determined using a commercial enzymatic cholesterol assay (Boehringer Mannheim). If the liposomes did not contain cholesterol, phospholipids were quantitated by phosphorus analysis (18).
Calcein Release Assay-Large unilamellar vesicles (LUV) were produced as above, whereby the lipids after drying were resuspended in Hepes/NaCl containing 50 mM calcein (2Ј,7Ј-bis-[N,N-bis-(carboxymethyl)aminomethyl]fluorescein). Following extrusion, the liposome suspension was passed over a column of Sephadex G-50 (Amersham Pharmacia Biotech) equilibrated with Hepes/NaCl to remove nonentrapped calcein. The void volume fractions were pooled, and the lipid concentration was determined. Small unilamellar vesicles (SUV) were prepared in an analogous manner, except that sonication (Branson probe sonifier 250) for 6 min was substituted for membrane extrusion. Liposomes containing 15 g of total lipid were incubated for 10 min at 37°C in a volume of 100 l with the amounts of cytolysin indicated under "Results." Subsequently, each sample was diluted into 3 ml of Hepes/ NaCl (pH 7.5) and immediately assayed for calcein fluorescence ( ex , 488 nm, em , 520 nm) in a SPEX Fluoromax fluorimeter. The fraction of calcein released was calculated from the increase of fluorescence over that of a control sample incubated without cytolysin, whereas the fluorescence maximum corresponding to 100% release was determined on a sample solubilized with sodium deoxycholate (final concentration, 6 mM).
Assay of VCC Binding to Liposomes-Liposomes varying in composition were incubated with 1 g of the cytolysin under the same conditions as applied in the calcein release assay. The sample was then supplemented with sucrose to 6% and spun in an air-driven ultracentrifuge (Beckman Airfuge) for 15 min at 100,000 ϫ g in order to float the liposomes. Unbound cytolysin was sampled from below the liposome film by aspiration and quantitated by hemolytic titration, whereas each sample was analyzed in triplicate. 2-Fold dilution series in phosphatebuffered saline, 0.05% bovine serum albumin were prepared in a microtiter plate, and rabbit erythrocytes were added to 1.25% final concentration. After incubation at 37°C for 2 h, unlysed cells were pelleted by centrifugation. The supernatants were diluted 6-fold with phosphate-buffered saline, and the amount of hemoglobin released was determined photometrically at 440 nm. The hemolytic titer was read as the highest dilution sufficient for Ն60% hemolysis.
Polyacrylamide Gel Electrophoresis (PAGE)-SDS-PAGE was performed according to the Laemmli method. For native PAGE, a continuous buffer system was used containing 50 mM Tris, 50 mM glycine, and 4 mM sodium deoxycholate. Gels were prepared with 1,4-bis(acryloyl)piperazine (Fluka) as the cross-linker (molar ratio cross-linker/acrylamide 0.015; total concentration of acrylamide, 5%). Electrophoresis was carried out at 5 V/cm for 1.5 h. Protein bands were visualized by silver staining (19) or with Coomassie Blue R-250.

Cytolysin Susceptibility of Phosphatidylcholine/Cholesterol
Large Unilamellar Vesicles-The release of calcein from liposomes composed of PC and cholesterol has been demonstrated previously (14). In those experiments, the cytolysin was employed at a concentration of 170 nM, which is about 1000 times more than the amount required for lysis of rabbit erythrocytes (15). The liposomes had been obtained by sonication, which usually results in the formation of small unilamellar vesicles (SUV). Since the low susceptibility of the SUV might conceivably have been related to their small size, we compared them to large unilamellar vesicles (LUV) that were prepared by extrusion through polycarbonate membranes (17). The liposomes were incubated with various concentrations of the cytolysin (10 min, 37°C), and the fraction of calcein released was quantitated fluorimetrically. As shown in Fig. 1A, the LUV proved even less susceptible toward the cytolysin than SUV. Since LUV are generally considered a more appropriate model of natural membranes than SUV, this finding even accentuates the gap in susceptibility between these artificial target membranes and natural ones.
Brain Lipids Strongly Enhance Liposome Permeabilization by VCC-For an initial test of the hypothesis that lipids other than cholesterol are important in the interaction of V. cholerae cytolysin with target membranes, we employed a crude mixture of phospholipids and glycolipids extracted from bovine brain. These lipids were supplemented with cholesterol to the same molar content as above (30%) and used for the preparation of LUV. Fig. 1B shows that calcein was released from the cholesterol-enriched brain lipid liposomes at very much lower cytolysin dosages than those containing PC (cf. Fig. 1A). Obviously, the brain extract does contain one or more lipid species that greatly augment VCC pore formation.
We next examined the contribution of cholesterol to the sensitivity of the brain lipid liposomes. Since the crude brain extract has a residual cholesterol content of 1-2%, we employed a blend of purified bovine brain lipids instead of the crude extract for the preparation of LUV without cholesterol. PC, PE, PS, galactosylceramide, and SPM were admixed at their respective molar fractions also prevalent in the crude extract (see "Materials and Methods"). As seen in Fig. 1B, the cholesterolfree liposomes required a fairly high dosage of cytolysin to yield any detectable calcein release at all, and only very few cytolysin oligomers were detected on the liposome membranes by SDS-PAGE. This observation reinforces the previously established important role of cholesterol (14). On the other hand, it may be stated that the blended brain lipid liposomes are still similarly sensitive to VCC as are those composed of egg yolk PC and cholesterol (cf. Fig. 1A). Thus, it appears that both cholesterol and the incremental brain lipid species impart a low level sensitivity to membranes when present alone, but they have a strong cooperative effect when employed in combination.
In membrane permeabilization by VCC, monomer binding can be distinguished from oligomerization and pore formation, and the sensitizing effect of particular membrane constituents might be related to either of these steps. It has previously been shown that VCC efficiently binds to membranes consisting of egg yolk PC alone (15). PC was a major constituent of all liposome species that were employed here. Accordingly, with all of these liposome preparations, the extent of toxin binding ranged from 50 to 90% as assayed by hemolytic titration. Therefore, the binding step could only account for a 2-fold variation in membrane susceptibility, which means that the much larger differences that were observed experimentally must mainly be due to the oligomerization step.
Galactosylceramide Is the Major Constituent Responsible for the Sensitivity of Brain Lipid Liposomes to VCC-We then sought to identify the individual lipid constituents of the crude brain extract involved in promoting the oligomerization of VCC. To this end, the major lipid species were obtained in pure form and incorporated into liposomes at their respective molar fraction in the crude extract (see "Materials and Methods"); the residual lipids were cholesterol (molar fraction, 30%) and egg yolk PC. As an exception, bovine brain PC was employed at 70% with no egg yolk PC. The liposomes were added to serial dilutions of VCC. Fig. 2 displays the fraction of calcein released from the various liposome species. The most pronounced sensitization was evidently induced by galactosylceramide (Gal-Cer), which was employed at 40% by mole and approached the crude brain lipid mixture in efficacy. A slight enhancement of susceptibility was also observed with SPM (molar content in the membrane, 10%). In contrast, very little sensitization was evident with any of the glycerophospholipids, with the sole exception of bovine brain PE (membrane content, 20%) which was similar in efficacy to SPM and clearly superior to bovine brain PC. A similar relationship was observed with PE and PC derived from egg yolk (not shown). The main difference between the two lipid species should consist in their respective choline and the ethanolamine head groups; the choline head group therefore appears to have an inhibitory effect upon VCC pore formation.
The Ceramide Moiety Is Essential in Enhancement of VCC Oligomerization by Sphingolipids-In the above experiments, the sphingolipids SPM and GalCer had performed quite differently in sensitizing the respective liposomes to VCC. However, GalCer had been employed to a four times higher amount, which might conceivably account for its superior effect. For a more precise comparison of their interaction with the cytolysin, liposomes were produced with matched contents of SPM and GalCer. In this series of experiments, further lipid species were also included to learn more about the importance of various structural features of the sphingolipid molecule to the oligomerization of VCC. Fig. 3 displays the cytolysin dosages required for calcein release of Ն50% from liposomes containing 10, 20, or 40% of the respective sphingolipids (missing values for 20 or 40% indicate that homogeneous and stable liposome preparations could not be obtained under our experimental conditions.) The essential findings can be stated as follows: SPM is clearly inferior to GalCer also if both are employed at equivalent amounts. The ganglioside G M1 (which possesses a tetrasaccharide head group attached to ceramide) is similar in efficacy to SPM. Both glucosylceramide and free ceramide closely match GalCer in their degree of liposome sensitization, which indicates that ceramide does not require any head group to be attached. Sphingosine, however, is only weakly effective, indicating that removal of the amide-linked fatty acid from the ceramide molecule impairs its ability to support VCC oligomerization. We conclude that the ceramide moiety is crucial in the interaction of sphingolipids with VCC.
The 50-kDa Form of V. cholerae Cytolysin Lacks the C Terminus of the 65-kDa Form-In a previous report, we showed that a naturally occurring 50-kDa form of VCC shares the N terminus of the 65-kDa form, implying that it must have been FIG. 1. VCC-mediated release of calcein from liposomes differing in size or composition. Liposomes with entrapped calcein were added to a dilution series of V. cholerae cytolysin (final concentration of lipid, 150 g/ml). After 10 min at 37°C, the calcein released was quantitated fluorimetrically; 100% release corresponds to the fluorescence of a detergent-lysed sample. A, SUV and LUV consisting of phosphatidylcholine (molar content, 70%) and cholesterol (30%). B, LUV prepared from crude bovine brain lipids, supplemented with cholesterol to a molar content of 30% or from a corresponding mixture of purified brain lipids lacking cholesterol. Inset, SDS-PAGE of VCC incubated with liposomes. The upper bands correspond to the oligomer. Left, brain lipid LUV with cholesterol; right, brain lipid LUV without cholesterol.
FIG. 2. The effect of individual brain lipid species upon membrane permeabilization by VCC. The major components of bovine brain lipids were incorporated to the molar ratios indicated into liposomes, whereby the residual lipids were made up by cholesterol (30%) and egg yolk phosphatidylcholine. Membrane permeabilization was assayed by release of calcein as detailed in the legend to Fig. 1. GalCer, SPM, PE, and PS were employed at their respective molar ratios also prevalent in the crude brain extract. Brain PC (which amounts to 15% in the extract) was used at 70% without any egg yolk PC.
proteolytically cleaved close to its C terminus (4). In those experiments, the 50-kDa form was only characterized by SDS-PAGE, so the possibility was not ruled out that the proteolytic fragments remain associated under non-denaturing conditions. With the homologous V. vulnificus cytolysin, a proteolytically nicked form has been described, the fragments of which are held together by disulfide bonds (20). Fig. 4A shows that this is not the case with the 50-kDa form of VCC, since identical migration was observed under reducing and non-reducing conditions. If electrophoresis was performed in the presence of the non-denaturing detergent deoxycholate, the 50-kDa form again migrated ahead of the 65-kDa form (Fig. 4B). This confirms that the C terminus is indeed lost upon proteolytic cleavage and is not required for cytolytic activity.
The V. cholerae Cytolysin Pore Is a Pentamer-The 65-kDa form assembles into SDS-resistant oligomeric pores on suitable membranes (14). This also holds for the 50-kDa form, whereby the two oligomer species differ in electrophoretic mobility in good correlation with the difference in molecular mass of the respective monomers (Fig. 5, lanes 1 and 5). We reasoned that a mixture of the two cytolysin forms should yield hybrid oligomers. The composition of these hybrids should be randomly distributed, and it should be reflected by their respective electrophoretic mobility. Fig. 5 (lanes 2-4) shows that this is indeed the case. The mixed samples formed patterns of six evenly spaced bands, which included those representing the homogeneous oligomers. The oligomers migrating in the lowest of these six bands thus consisted of 50-kDa subunits only. With each of the five subsequent bands, the number of 65-kDa subunits increased by one, reaching a maximum of five with the topmost band. The total number of subunits thus is always five; therefore the VCC pore is a pentamer. When the oligomers were dissociated by boiling in SDS, their subunits were recovered with unaltered electrophoretic mobility, which ruled out any artifacts due to changes in covalent structure (e.g. by protease contamination; Fig. 5B).

DISCUSSION
The cytolysin of V. cholerae belongs to a homologous family of toxins that are widespread among Vibrio and Aeromonas species (21). Several of these are capable of eliciting diarrheal disease, including enterotoxin-negative strains of V. cholerae O1 El Tor (22, 23) and non-O1 serotypes (24,25). In the latter instance, VCC most probably is pathogenetically significant, since experimentally the purified protein imparts marked damage to the small intestine epithelia of rabbits and mice (6,7,26).
The susceptibility of the epithelial cells to these cytolysins probably reflects the adaptation of V. cholerae and related species to the intestinal environment, but it has not yet been explained in molecular terms. The mammalian intestinal brush border membrane is distinguished by its high content of glycolipids, which may contribute to the tubular shape of the microvilli (27,28) and, like phospholipids and cholesterol, amount to one-third of the total lipids. Monohexosylceramides or ceramide represent the major glycolipids in rats (29, 30) and humans (31). This composition of lipids very much resembles the most susceptible synthetic liposomes characterized in the present study. We therefore propose that its high content of glycolipids, in conjunction with cholesterol, accounts for the sensitivity of the brush border membrane toward V. cholerae cytolysin.
One of the most striking findings about the effect of lipids upon the oligomerization of VCC consists in the pronounced cooperativity between sphingolipids and cholesterol. Various sphingolipids have been reported to associate with cholesterol in mixed membranes, which raises the possibility that the oligomerization of VCC is mediated by sphingolipid-cholesterol complexes rather than by the individual lipid molecules. The interaction with cholesterol has been most thoroughly studied with sphingomyelin. Model monolayers of sphingomyelin were significantly condensed upon addition of cholesterol, and the sterol was also more firmly retained by these monolayers than by ones consisting of PC, respectively (32,33). Similar findings have been reported for dihexosylceramides, whereas monohexosylceramides did not appreciably associate with cholesterol (34). To our knowledge, evidence of stable complexes consisting of cholesterol and free ceramide is lacking as well. Since monohexosylceramides and free ceramide are clearly superior to SPM with respect to the enhancement of VCC oligomerization, it appears that association of sphingolipids and cholesterol is not essential for their interaction with the cytolysin.
Another interesting example of a protein simultaneously requiring ceramides and cholesterol in its target lipid membrane is provided by the E1 envelope protein of Semliki Forest virus (35,36), which triggers fusion of the viral envelope to the endosome membrane. Fusion, like pore formation, requires physical separation of laterally interacting lipid molecules, which might constitute the common rationale behind the unusual combined specificity for two lipid molecules, both of which are largely buried within the apolar core of the bilayer. In line with this interpretation, both ceramide and cholesterol are not required in binding of the cytolysin monomer but essentially contribute in the subsequent event of oligomerization and pore formation.
Apart from the apolar moieties of the membrane lipids, their polar head groups also appear to play a role in the oligomerization of VCC. Of note, a choline head group is present in both PC and sphingomyelin, and the efficiency of oligomerization in the presence of either molecule was clearly inferior to that observed with homologous lipid species (PE and ceramide, respectively). The inhibitory effect of the choline head group apparent from these results may be shared by the complex polar tetrasaccharide moiety of the ganglioside G M1 . In this context, it should be noted that a reduction in sensitivity of rabbit erythrocytes to VCC has been obtained by neuraminidase treatment (37). Removal of sialic acid from the membrane glycolipids exposed terminal galactose moieties, and susceptibility of the cells could be restored by treating the cells with galactose oxidase. It thus appears that VCC may interact with terminal galactosyl residues (which also occur in the ganglioside G M1 ) in a non-functional manner. On the other hand, galactosylceramide was very similar in its capability to sensitize membranes to VCC as were glucosylceramide and free ceramide. Possibly, the distance separating galactose from the ceramide moiety determines whether or not oligomerization of VCC proceeds following its binding to the sugar residue.
Sensitive model liposomes containing both cholesterol and glycosphingolipids were utilized to assess the stoichiometry of the V. cholerae cytolysin pore. The electrophoretic analysis of heteromers applied here to elucidate the oligomer stoichiometry was inspired by previous work on the heptamer of Staphylococcus aureus ␣-hemolysin (38). In the work cited, the two toxin variants required were obtained by chemical modification of single cysteine mutants. Where chemical modification or limited proteolysis are inappropriate, extending or truncating the termini of a protein molecule at the DNA level might be used to produce the variant species for heteromer analysis. The latter method should thus be more generally useful to determine the stoichiometries of toxin pores and, if combined with non-denaturing analytical separation, other homo-oligomeric proteins.
Among the bacterial pore-formers, ␣-hemolysin provides the only example of an oligomer structure determined at high resolution (39). Its transmembrane portion consists of a ␤-barrel with 7-fold rotational symmetry. Heptameric stoichiometry and ␤-barrel structure has also been confirmed with the protective antigen component of anthrax toxin (40,41). As a pentamer, VCC so far is unique within this particular class of toxins, but there are several previous examples of pentameric transmembrane channels in general. One of those is provided by the B subunit oligomer of V. cholerae enteroxin. The hollow center of the latter is lined by five ␣-helices (42). Helical structure has also been suggested for the intramembranous portion of the nicotinic acetylcholine receptor (43) and for the cardiac calcium channel phospholamban (44). It would be most interesting to determine which one of the two above structural paradigms applies to the transmembrane part of the VCC pore.
In sum, the present study shows that V. cholerae cytolysin has a dual specificity for both cholesterol and ceramides, which reflects the composition of its natural target membranes and that the pore is a pentamer. Both these properties qualify VCC as a novel prototype within the class of bacterial pore-forming toxins, and they probably apply to a series of homologous toxins of other Vibrionaceae as well.