Increased Stability upon Heptamerization of the Pore-forming Toxin Aerolysin*

Aerolysin is a bacterial pore-forming toxin that is secreted as an inactive precursor, which is then processed at its COOH terminus and finally forms a circular heptameric ring which inserts into membranes to form a pore. We have analyzed the stability of the precursor proaerolysin and the heptameric complex. Equilibrium unfolding induced by urea and guanidinium hydrochloride was monitored by measuring the intrinsic tryptophan fluorescence of the protein. Proaerolysin was found to unfold in two steps corresponding to the unfolding of the large COOH-terminal lobe followed by the unfolding of the small NH2-terminal domain. We show that proaerolysin contains two disulfide bridges which strongly contribute to the stability of the toxin and protect it from proteolytic attack. The stability of aerolysin was greatly enhanced by polymerization into a heptamer. Two regions of the protein, corresponding to amino acids 180–307 and 401–427, were identified, by limited proteolysis, NH2-terminal sequencing and matrix-assisted laser desorption ionization-time of flight, as being responsible for stability and maintenance of the heptamer. These regions are presumably involved in monomer/monomer interactions in the heptameric protein and are exclusively composed of β structure. The stability of the aerolysin heptamer is reminiscent of that of pathogenic, fimbrial protein aggregates found in a variety of neurodegenerative diseases.

The pore-forming toxin aerolysin is secreted by the human pathogen Aeromonas hydrophila as an inactive precursor, called proaerolysin (for review, see Refs. 1 and 2). According to the crystal structure of proaerolysin, each monomer is formed of two domains, a small globular NH 2 -terminal region (domain 1, amino acids 1-82) and a long elongated lobe that can be divided into 3 domains that are not continuous in the linear sequence ( Fig. 1) (3). In the crystal structure proaerolysin is organized as a dimer in which the two monomers are arranged in an antiparallel fashion with the two NH 2 -terminal domains clasping each other in a crossover fashion (3). Dimeric aerolysin was also observed in solution at high toxin concentrations (4,5). However, at concentrations below 1 mg/ml, proaerolysin starts separating into monomers as recently suggested by gel filtration experiments. 1 In order to become active, the precursor proaerolysin must be proteolytically processed at its COOH terminus (7)(8)(9)(10). The mature aerolysin is then able to oligomerize by circular assembly of seven monomers (11,12) which leads to the formation of a central channel in a manner similar to that which has been observed for other toxins as well as for GroEL, Trp RNAbinding attenuation protein, and the proteasome (for review, see Ref. 13). The aerolysin heptamer exposes hydrophobic surfaces (14) which enable the complex to spontaneously insert into a lipid bilayer and form a channel (15).
Here we have analyzed the stability of both proaerolysin and the aerolysin heptamer. We show that the unfolding of proaerolysin occurs in at least two steps and is reversible under nonreducing conditions. The two disulfide bridges in the proaerolysin molecule greatly contribute to the overall stability of the protein. These studies moreover show that the NH 2 -terminal domain 1 of the toxin constitutes a separate folding domain and is the most stable part of the toxin. The second part of this work concerns the analysis of the effects of heptamerization on the stability of aerolysin. Urea (8 M) had no effect on the structure of the complex. Partial unfolding, without disassembly of the complex, was, however, observed in GdnHCl 2 (6 M). Again the disulfide bonds were found to significantly contribute to the overall stability. In order to identify regions of the protein involved in maintaining the oligomer assembled, limited proteolysis studies were performed. Results from NH 2 -terminal sequencing of the fragments indicated that domains 1 and 2 could be entirely removed and that domains 3 and 4 are crucial for the maintenance and the stability of the heptameric complex.

EXPERIMENTAL PROCEDURES
Protein Purification-Wild type and mutant proaerolysins were purified according to a published procedure (16). Generation of C159S was described previously (17,18). Concentrations were determined by measuring the optical density at 280 nm, considering that a 1 mg/ml sample has an OD of 2.5 (9). Heptameric aerolysin was prepared as described previously (10,12). Hemolytic activities were measured as described previously (19). When oligomerization had to be prevented activation was performed in 150 mM NaCl, 50 HEPPS, pH 8.5 (20).
Cyanogen Bromide Fragmentation-Cyanogen bromide fragments * This work was supported by a grant of the Swiss National Science Foundation (to G. v. d. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
were obtained by treating proaerolysin for 24 h with a 500-fold molar excess of cyanogen bromide in 75% formic acid at room temperature. The sample was then frozen and evaporated by vacuum centrifugation. The sample was then rinsed with water, frozen, and evaporated again. This procedure was performed twice.
Unfolding with Chaotropic Agents-Equilibrium unfolding experiments were performed as described by Pace et al. (21). Urea and GuHCl stock solution were prepared as described by Pace et al. (21) in either 150 mM NaCl, 20 mM HEPES, pH 7.4, or 150 mM NaCl, 20 mM HEPPS, pH 8.4. The denaturant concentrations of each stock solution were determined by weight and refractive index (21). These solutions were used only if the difference between the two values was less than 1%. Buffer, denaturant, and protein (20 g/ml) from concentrated stocks were mixed and samples were incubated in a water bath at 25°C for at least 24 h, unless specified otherwise. All experiments were performed at least 3 times. To test for reversibility, a solution of proaerolysin in the post-transitional region (4 -8 M urea) of the denaturation curve was submitted to a 2-fold serial dilution with buffer in a 96-well plate. The samples were then incubated for 2 h at room temperature to allow refolding of the protein. Trypsin (1/50 protease/protein ration, w:w, 10 min incubation) was then added to each well to convert proaerolysin into aerolysin and the hemolytic activities were then determined as described previously (19).
Tryptophan Fluorescence-Fluorescence measurements were made with a PTI spectrofluorimeter equipped with a thermostated cell holder. The excitation wavelength was either 278 or 295 nm and slit widths were 5 and 2.5 nm for excitation and emission, respectively. For each recorded spectrum, the Raman scatter contribution was removed by subtraction of a buffer blank. Very similar unfolding curves were obtained whether the samples were excited at 278 or 295 nm. In order to measure the effect of denaturants on the fluorescence intensity, samples were excited at 278 nm (allowing excitation of both tryptophans and tyrosines). The ratio between the intensity at 345 nm (which varied significantly) and the intensity at 315 nm (which varied little) was calculated for each condition. Experiments were performed at least three times and the average ratio calculated. The protein concentration was 20 g/ml.
Limited Proteolysis with Trypsin and NH 2 -terminal Sequencing-Heptameric aerolysin was prepared as described previously (10,12). Limited proteolysis of the heptamer with trypsin was either performed in solution or in a SDS gel. For cleavage in solution, the heptamer was first incubated with 0.1% SDS for 3 h. Trypsin was then added in a 1:1 (w:w) ratio and the sample was incubated for 24 h at 37°C. The same amount of trypsin was added every 24 h. After 72 h incubation, 10-fold trypsin inhibitor was added the sample was analyzed by SDS-PAGE. For in-gel proteolysis, the heptamer was run on a 7.5% SDS gel. The oligomeric band was revealed by Coomassie Blue staining, excised, and incubated in 100 mM ammonium bicarbonate, pH 8, in the presence of trypsin for 24 h at 37°C. The trypsin to protein ratios between 0.1:1 and 12:1 were tested and led to the same results. The samples were concentrated 2-fold using a Speed-Vac. The samples were then again analyzed by SDS-PAGE using 7.5% gels and transferred onto a PVDF Porablot membrane (0.2 m, Macherey-Nagel, Dü ren, Germany) according to a procedure adapted from Laurière (22). After electrophoretic migration, the gel was sealed in a plastic pocket containing water and frozen at Ϫ80°C for 15 min between two glass plates. The gel was then incubated in a buffer containing 0.1% (w/v) SDS, 15 mM lactic acid, 25 mM Tris, pH 8.4, for 15 min. The membrane was activated in methanol for 1 min. Proteins were transferred under semi-dry conditions on the Bio-Rad semi-dry transfer cell (Trans-blot SD). The following buffers were used. Buffer I contained 1.2% (w/v) SDS, 60 mM lactic acid, 100 mM Tris, pH 8.4; buffer II, 0.1% (w/v) SDS, 15 mM lactic acid, 25 mM Tris, pH 8.4; buffer III, 20% methanol, 60 mM Lactic acid, 20 mM Tris, pH 3.8; and buffer IV, 20% methanol, 100 mM Tris, pH 10.4. The gel and blotting membrane were assembled with Whatmann No. 3 paper and sponges in a blotting sandwich as follows, going from the cathode to the anode: 1) sponge in buffer I; 2) 3 Whatmann papers in buffer II; 3) gel; 4) plastic frame; 5) PVDF membrane; 6) 3 Whatmann papers in buffer III; and 7) sponge in buffer IV. Samples were transferred for 120 min at 150 mA. The membrane was then washed three times for 10 min in water, dried between two Whatman papers, and stained twice for 5 min with Coomassie Blue prepared in methanol. The membrane was destained in methanol, dried under vacuum for 30 min, and kept at 4°C.
Limited Proteolysis with Boilysin-Heptameric aerolysin (0.4 mg/ml, in 30 mM NaCl, 20 mM HEPES, pH 7.4) was incubated in the presence of 5 mM CaCl 2 and 0.1% SDS for 10 min at 70°C. Then boilysin, which was kindly provided by Dr. van den Burg (23) was added at a protein to protease ratio of 200 and the sample was incubated for an additional 10 min at 70°C. Proteolysis was performed in the presence of 0.1% SDS in order to disperse aggregated heptamers. Similar results were obtained in the absence of SDS but lead to bands that were not as sharp (not shown). The boilysin-treated sample was then precipitated with chloroform/methanol. A fraction of the sample was analyzed by SDS-PAGE. The remainder of the sample was treated with 70% formic acid for 30 min at room temperature to disassemble the heptameric complexes. The obtained peptide sample was split in two. Part was analyzed by matrix-assisted laser desorption ionization-time of flight mass spectrometry and the rest was analyzed on Tricine-SDS gels (24). The peptide gel was transferred onto a PVDF Porablot membrane overnight at 40 V using a transfer buffer containing 12.5 mM Tris, 92 mM glycine, and 10% methanol.
Other Methods-In order to disrupt the aerolysin heptamer, the complex was treated with 70% formic acid for 30 min at room temperature. The sample was then frozen and evaporated using a Speed-Vac. The sample resuspended in water, frozen, and evaporated again. Unless specified protein transfer onto PVDF membranes was performed using classical procedures. Amino-terminal amino acid sequencing was performed using an ABI 477A or 473A protein sequencer and the standard Edman chemistry provided by the manufacturer (Applied Biosystems and ABI, Weiterstadt, Germany).

RESULTS
Identification of Disulfide Bridges-Proaerolysin contains 4 cysteine residues ( Fig. 1), which according to the crystal structure are involved in two disulfide bridges: Cys 19 -Cys 75 and Cys 159 -Cys 164 (3). For our studies, it was important to confirm the existence of these disulfide bridges experimentally.
To reveal the existence of Cys 19 -Cys 75 disulfide bridge, we made use of the fact that methionine residues, which constitute cyanogen bromide (CNBr) cleavage sites, separate these two cysteines. CnBr fragmentation of proaerolysin is predicted to give rise to a variety of fragments among which two fragments of 9 and 4.5 kDa, respectively, containing Cys 75 and Cys 19 . If these two cysteines are bridged, a 13.5-kDa fragment should be observed under nonreducing conditions but not under reducing conditions. As shown in Fig. 2A, this is indeed the case.
To confirm the existence of a bridge between Cys 159 and Cys 164 , we used another approach. It has been previously shown that treatment of proaerolysin with trypsin only leads to the removal of a COOH-terminal peptide and that the mature aerolysin thus obtained is very resistant to further proteolysis (7). However, as shown Fig. 2B, when wild type proaerolysin was treated with trypsin in the presence of 10 mM DTT, two additional bands appeared with apparent molecular masses of 16 and 30 kDa indicating that an additional cleavage site became accessible upon reduction of disulfide bridges. The same cleavage pattern was observed when cleaving a proaerolysin mutant in which Cys 159 was mutated to serine thereby preventing the formation of a disulfide bond. Amino-terminal sequencing revealed that fragment 1 in Fig. 2B contained two fragments starting at Arg 163 and Cys 164 , respectively, indicating that cleavage had occurred between the two cysteines, and band 2 had as expected the same NH 2 terminus as the fulllength toxin. These observations suggest that in the native toxin, Cys 159 and Cys 164 are indeed bridged. In agreement with the fact that all cysteines are involved in disulfide bridges, we could not detect any free cysteines by 5,5Ј-dithionitrobenzoic acid labeling. In contrast, one free sulfhydryl group was found by this method in the C159S mutant (data not shown).
These observations unambiguously demonstrate that proaerolysin contains two disulfide bridges: Cys 19 -Cys 75 and Cys 159 -Cys 164 (Fig. 1). The proteolysis experiments moreover indicate that the Cys 159 -Cys 164 bridge, which is present in a surface exposed loop (3), protects this loop from proteolytic attack.
Urea Unfolding of Proaerolysin-We next studied the stability of proaerolysin by measuring its unfolding in urea. Preliminary experiments showed that after 24 h incubation in up to 7 M urea (in 150 mM NaCl, 20 mM Tris, pH 7.4) unfolding was reversible since 75-100% of the initial hemolytic activity could be recovered upon subsequent dilution in urea free buffer (not shown).
The extent of unfolding of proaerolysin in urea was followed by measuring the intrinsic fluorescence of the molecule. Proaerolysin contains 18 tryptophans and 21 tyrosines. These residues are scattered throughout the structure, although a higher concentration is found in domain 2, which contains 11 out of the 18 tryptophans (Fig. 1). The fluorescence emission spectra of proaerolysin in 0, 5.4, and 8.1 M urea are shown in Fig. 3A. Whereas the fluorescence intensity of most proteins decreases upon unfolding, the fluorescence of proaerolysin significantly increased. Concomitantly the maximal emission wavelength was shifted from 335 to 345 nm and finally to 348 nm for 0, 5.4, and 8.1 M urea, respectively. The ratio between the intensity at 345 nm (which varied significantly) and the intensity at 315 nm (which varied little) was plotted as a function of the urea concentration (Fig. 3B). The unfolding curve then obtained was clearly biphasic. Similar curves were obtained when plotting the maximal emission wavelength as a function of urea (not shown). The first midpoint transition [urea]1 ⁄2 was at 2.9 M urea and the second at 7.1 M urea (Table  I). Unfolding curves for proaerolysin were independent of toxin concentrations within the 7-100 g/ml range we studied. This observation suggested that at the proaerolysin concentration used in these experiments (20 g/ml), the toxin is mainly monomeric in agreement with our gel filtration experiments. 1 The biphasic nature of the urea unfolding curve (Fig. 3B) suggests that domains of the protein unfold sequentially. The possibility that the two steps correspond to the unfolding of two independent folding units is further strengthened by the observation that both temperature and pH selectively affected only the first step (not shown).
Effects of Disulfide Bonds on the Stability of Proaerolysin in FIG. 2. Proaerolysin contains two disulfide bridges. A, proaerolysin was subjected to CNBr fragmentation and analyzed on a Tricine-SDS gel under reducing (with ␤-mercaptoethanol, ϩ) and nonreducing (without ␤-mercaptoethanol, Ϫ) conditions. As predicted, in the absence of ␤-mercaptoethanol, the three largest peptides had molecular masses of 27.8, 13.6, and 8.8 kDa. Upon reduction the 13.6-kDa fragment led to two fragments of 9.0 and 4.5 kDa. The difference in migration pattern indicate that Cys 19 and Cys 75 , that are separated by 3 methionine residues, formed a disulfide bridge in the native protein. B, wild type proaerolysin was incubated in the presence of 10 mM DTT for 15 min at room temperature and then treated with trypsin (1/50, mol:mol) for 20 min. SDS-PAGE analysis of the sample revealed the presence of two bands in addition to the aerolysin band. The same migration pattern was observed after treatment of the C159S proaerolysin mutant with trypsin in the absence of DTT. NH 2 -terminal sequencing revealed that the 30-kDa band contained two fragments with the following NH 2 termini, XGDKTAI and XXXKTAIKV, and the 16-kDa band contained one fragment with the same NH 2 terminus as the full-length toxin. When the wild type toxin is treated with trypsin in the absence of DTT, the precursor is fully converted to aerolysin without any evidence for further breakdown. Urea-We next studied the contribution of the two disulfide bridges to the stability of proaerolysin. Urea denaturation experiments were first performed using the wild type toxin in the presence of DTT. Under these conditions unfolding was irreversible and therefore thermodynamic parameters could not be calculated except for the midpoint transition (Table I). As shown in Fig. 4, DTT had a dramatic effect on both unfolding steps leading to an apparently monophasic curve. These observations indicate that at least one of the two bridges strongly contributes to the stability of the protein.
To specifically investigate the role of the Cys 159 -Cys 164 bridge, we analyzed the urea denaturation of the C159S proaerolysin mutant. Interestingly breaking of the Cys 159 -Cys 164 bridge by this mutation only affected the first unfolding transition, whereas the second was similar to the one observed for the wild type protein (Fig. 4). Thus the Cys 159 -Cys 164 bridge appears to be important for stability despite the fact the two cysteines are only 5 amino acids apart. The fact that DTT did have an effect on the stability of the C159S mutant indicates that breaking the Cys 19 -Cys 75 bridge had a strong destabilizing effect. Therefore both disulfide bridges contribute to the stability of proaerolysin.
These results also indicate that the biphasic unfolding of proaerolysin corresponds to the unfolding of the large lobe of the protein (containing the Cys 159 -Cys 164 bridge), followed at higher urea concentrations by the unfolding of domain 1 (containing the Cys 19 -Cys 75 bridge). The two disulfide bridges strongly contribute to the stability of the protein domain in which they are located.
Unfolding of the Aerolysin Heptamer-We next studied the stability of the aerolysin heptamer, which is its channel forming configuration by analyzing the effect of urea on the tryptophan fluorescence of the heptamer. Quite remarkably, no change in fluorescence intensity or in maximal emission wavelength could be observed after 24 h incubation of the complex with 8 M urea (Fig. 5A). Unfolding of the heptamer required high concentrations of a more potent chaotropic agent, guanidinium hydrochloride (GdnHCl). Unfolding appeared very cooperative and occurred in a single step with a midpoint transition at 3.8 M (Fig. 5B, Table I). For comparison, unfolding of proaerolysin in GdnHCl was also analyzed. As observed in urea, unfolding followed a two-step process (Fig. 5B). This first midpoint transition was at 0.9 M GdnHCl and the second at 2.9 M. We found that DTT affected unfolding of the heptamer indicating that the disulfide bridges contributed to its stability.
The above experiments illustrate the uncommon stability of the noncovalent aerolysin heptamer. In order to identify the regions of the protein that are involved in maintaining the heptamer, we have performed limited proteolysis under conditions that would not lead to full heptamer disassembly.
Partial Proteolysis of the Aerolysin Heptamer-We were able to generate a partially degraded aerolysin heptamer after prolonged incubation (7 days) in the presence of 0.1% SDS and of residual trypsin remaining after activation (Fig. 6A, lane 2). This truncated complex ran as a double band slightly below the 200-kDa molecular mass marker on acrylamide gradient gels. To determine were cleavage had occurred, the truncated complex was dissociated into monomers by formic acid treatment (25) (Fig. 6A, lane 3). In addition to the aerolysin band (presumably resulting from the nondegraded full-length heptamer), a major polypeptide with an apparent molecular mass of 37 kDa and a less abundant band (corresponding to a doublet) with a apparent molecular masses of 22 kDa were detected. Amino-terminal sequencing revealed that the 37-kDa fragment and one of the fragments running at 22 kDa had the same NH 2 terminus, i.e. Asp 139 , whereas the second 22-kDa fragment started at Trp 149 . The calculated molecular mass of a fragment going from Asp 139 to the trypsin activation site at Lys 427 is 32.4 kDa. A heptamer formed by such fragments would have a mass of 227 kDa, which is in close agreement with the apparent molecular mass of the digested complex (Fig.  6A, lane 2). It is therefore likely that all monomers in the truncated heptamer lacked the first 138 amino acids and that some also underwent COOH-terminal cleavage.
We next performed limited proteolysis of the heptamer with trypsin at an enzyme to protomer ratio of two to one for 3 days. Shorter incubation times could be used when the heptamers was treated with trypsin within the SDS gel itself (see "Experimental Procedures"). As shown Fig. 6B (lane 2), trypsin treatment led to the accumulation of three oligomeric species of approximately 200, 160, and 75 kDa apparent molecular mass. The fact that all species ran higher than the aerolysin monomer indicates that the heptameric organization was at least  partially maintained. This was confirmed by the fact that these bands could no longer be seen when the trypsin-treated sample was incubated with formic acid prior to SDS-PAGE (not shown). Truncated oligomers were transfer to PVDF membranes and NH 2 -terminal sequencing was carried out to identify the fragments. Transfer of the 200-kDa complex was so inefficient that we were unable to obtained sequence information. The 160-kDa complex was found to contain a single NH 2 terminus, starting at Gln 195 (Table II, Fig. 6). However, considering that the limit of sensitivity is about 1 pmol, we cannot exclude that other less abundant NH 2 termini were present. The 75-kDa complex contained 5 detectable NH 2 termini. The two most abundant NH 2 termini, each corresponding to Ϸ27% of the total, were His 186 and Gly 402 (Table II, Fig. 6B). The three less abundant NH 2 termini found in this complex were Gln 195 (Ϸ18% of total), Ser 192 (Ϸ14% of total), and Ala 398 (Ϸ14% of total). The detectable NH 2 termini present the 75-kDa complex indicate that cleavage occurred in the 185-194 region (in domain 3) and in the 397-401 region (at the domain 2-domain 3 boundary). The simplest interpretation of these results is that trypsin had released at least 185 residues for the NH 2 terminus in all seven monomers and that all contained the COOH-terminal region (amino acids 398 to the end, i.e. Lys 427 ).
Proteolysis of the Aerolysin Heptamer Using the Thermostable Enzyme Boilysin-The heptamer was also treated using an unusually stable mutant of thermolysin named boilysin which is active at high temperatures (23). Cleavage was performed at 70°C, a temperature at which the heptamer is slightly unfolded (a 2-nm red shift in maximum emission wavelength was observed). The use of boilysin at 70°C allowed proteolysis for much shorter times (10 min) than with trypsin . This treatment led to 2 fragments as well as material at the migration front. Samples were analyzed by SDS-PAGE using the Laemmli buffer system (6) and a linear gradient of 3-15% acrylamide. B, the aerolysin heptamer was treated with trypsin as described under "Experimental Procedures." The trypsin-treated sample was then analyzed by SDS-PAGE using an acrylamide gradient gel. Three truncated complexes could be identified. Lane M, molecular weight marker.

TABLE II NH 2 -terminal sequencing of the aerolysin heptamer proteolytic fragments
The NH 2 -terminal sequence of the proteolytic fragments shown in Figs. 6 and 7 are summarized. The calculated full-length molecular weight (M r ) corresponds to the calculated weight of the peptide started at the determined N terminus and ending at Lys-427, the trypsin activation site. The NH 2 -terminal labeled with three asterisks (***) each accounted for 27% of all NH 2 -terminal in the 75-kDa complex, the NH 2 -terminus labeled with two asterisks (**) for 18%, and the NH 2terminal labeled with one asterisk (*) for 14% each. (3 days) and at a 400 -2000 lower enzyme to protein ratio. Boilysin cleavage led to a similar cleavage pattern as trypsin with two major bands with apparent masses of 160 and 75 kDa (Fig. 7A, lane 3). Upon dissociation of the complex by formic acid treatment, three bands of similar molecular weights could be detected on Tris glycine peptide gels (Fig. 7B, indicated by  lines in lane 3). The three bands were sequenced separately by cutting the blot in very thin strips and not sequencing as a mixture. The top major peptide started at Leu 179 as determined by NH 2 -terminal sequencing (Table II). The second minor peptide started at Phe 184 , and finally the smallest minor peptide started at Val 189 (Table II). The cross-contamination levels between the three bands were low. Matrix-assisted laser desorption ionization-time of flight mass spectrometry analysis indicated that these three peptides had a mass of 14.38, 13.82, and 13.24 kDa, respectively. When taking into account the NH 2 termini and the masses, it appears that all three ended at Glu 307 . The above proteolysis studies show that amino acids 1-178 can be removed without triggering full disassembly of the heptamer. Since the boilysin fragments ended at Glu 307 and no other peptides could be detected, it appears that domains 1 and 2 had been completely removed from the heptameric complex. Although peptides containing residues 401 to 427 (corresponding to the trypsin activation site) were not detected in the boilysin-treated samples it seems unlikely that they had been removed by the enzyme. They indeed are part of a 4-stranded ␤-sheet in domain 3 (Fig. 1). More likely this small peptide of 2.7 kDa was not detected on our peptide gels. This hypothesis is supported by the presence of NH 2 termini starting in the 397-401 regions in the truncated complexes obtained by trypsin treatment (Fig. 6B, Table II). DISCUSSION In the present work we have analyzed the stability of the pore-forming toxin aerolysin both in its precursor form and in its heptameric complex. Unfolding of the proaerolysin in urea was biphasic. The first phase could be attributed to the unfolding of the COOH-terminal elongated lobe (domains 2 to 4) of the proaerolysin molecule and the second to the unfolding of domain 1 (NH 2 -terminal domain) which was by far the most stable part of the protein. Both the large COOH-terminal domain and the NH 2 -terminal domain 1 were found to owe their stability in part to the presence of a disulfide bridge (Cys 159 -Cys 164 and Cys 19 -Cys 75 , respectively). It was somewhat surprising to find that the Cys 159 -Cys 164 (Fig. 1) had a role in stability since this bridge links two cysteines that are only 5 amino acids apart in the primary sequence. This bridge was also found to protect a surface exposed loop at the top of domain 2 from proteolysis.
The stability of the toxin was dramatically enhanced upon heptamerization of aerolysin. Incubation for 24 h with 8 M urea failed to disassemble the subunits and did not even affect the fluorescence of tryptophan residues indicating that they were still in a rigid environment. Also, incubation of the heptamer with 1% SDS for 3 h did not affect the tertiary structure as witnessed by near ultraviolet circular dichroism (not shown). Loosening of the structure, without separating the subunits, however, required a 24-h incubation with 4 M of the more potent chaotropic reagent GdnHCl. Such incredible stability is reminiscent of fibrillar protein aggregates which form upon self-assembly of prion protein or Alzheimer ␤-amyloid peptide (26 -28). Extremely high stability is also found in enzymes produced by extremophilic bacteria (29). In many of these proteins, increased stability is reached by association of subunits (30).
To identify the regions of the protein that are responsible for maintaining the heptamer so tightly assembled, we established proteolysis conditions that would cleave the protein but not fully disassemble the complex. Not surprisingly, considering our observations on the very high stability of the complex, the heptamer was very insensitive to proteolysis. Nicking required either very high trypsin concentrations and several days of incubation or treatment at 70°C with the engineered thermostable enzyme boilysin (23). It is attractive to speculate that the resistance of the heptamer to proteolysis under physiological conditions would be an explanation for the inability of mammalian cells to recover after aerolysin treatment. We have indeed never seen recovery of aerolysin-treated cells in contrast to the situation reported for Staphylococcal ␣-toxin (31) and Vibrio cholerae El Tor cytolysin (32), two toxins that have modes of action very similar to that of aerolysin. In contrast to the aerolysin heptamer, the Staphylococcal ␣-toxin heptamer is heat labile and sensitive to proteases (33,34) and might therefore be degraded by cellular proteases after internalization of plasma membrane patches that contain the pore. This type of repair mechanism has indeed been proposed but remains to be demonstrated.
The analysis of the fragments obtained after harsh proteolytic treatment showed that the truncated complexes contained residues Ϸ180 to 307 and 401 to 427 of aerolysin, which constitute domains 3 and 4 (Fig. 8, in dark gray). These results suggest that domain 1 and domain 2 could be excised from the heptamer without provoking complete disassembly of the complex. This indicates that domain 1 is not involved in maintaining the heptamer assembled, even though it is the most stable part of proaerolysin and appears to be crucial for dimerization at high toxin concentrations. Heptamerization, in contrast to dimerization, might therefore not occur through a domainswapping interaction. Alternatively it could occur by a swapping mechanism involving different domains. The fact that domain 2, which is the largest domain in the protein (Figs. 1 and 8), could also be excised from the complex in addition to domain 1 is in agreement with the model of the aerolysin channel proposed by Parker et al. (3). According to this model, domain 1 of one monomer interacts with domain 2 of the next monomer in the circular assembly. Therefore, upon removal of domain 1, domain 2 would no longer be involved in monomermonomer contacts and could become accessible to proteases.
Thus domains 3 and 4, and more specifically residues 180 to 307 and 401-427, are sufficient to maintain the complex assembled and to confer SDS resistance since the truncated heptamers could be detected after SDS-PAGE. Domains 3 and 4 are exclusively composed of ␤-strands and random structure indicating that, as for amyloids, stable self-assembly of aerolysin involves ␤-sheet domains (26,30,35). Self-association of ␤-sheet domains resulting in the formation of pathogenic protein aggregates is a characteristic feature of various medical disorders including Huntington's, Alzheimer, and Creutzfeld-Jacob's diseases (36). Further studies on the dissociation and reassociation mechanisms of aerolysin, which forms a smaller complex that is more amenable to biochemical and spectroscopical analyzes, are likely to contribute to understanding the aggregation processes of these pathogenic protein complexes.