A Non-classical Assembly Pathway of Escherichia coli Pore-forming Toxin Cytolysin A

Background: Cytolysin A (ClyA) is an α-pore-forming toxin secreted from pathogenic E. coli. Results: ClyA monomer assembles to an oligomeric pre-pore structure independently of lipid membrane and detergent. Conclusion: Our results support a model that ClyA proteins may oligomerize to a prepore within outer membrane vesicles before arriving on the target cell membrane. Significance: The proposed model for ClyA represents a non-classical pathway to attack eukaryotic host cells. Cytolysin A (ClyA) is an α-pore forming toxin from pathogenic Escherichia coli (E. coli) and Salmonella enterica. Here, we report that E. coli ClyA assembles into an oligomeric structure in solution in the absence of either bilayer membranes or detergents at physiological temperature. These oligomers can rearrange to create transmembrane pores when in contact with detergents or biological membranes. Intrinsic fluorescence measurements revealed that oligomers adopted an intermediate state found during the transition between monomer and transmembrane pore. These results indicate that the water-soluble oligomer represents a prepore intermediate state. Furthermore, we show that ClyA does not form transmembrane pores on E. coli lipid membranes. Because ClyA is delivered to the target host cell in an oligomeric conformation within outer membrane vesicles (OMVs), our findings suggest ClyA forms a prepore oligomeric structure independently of the lipid membrane within the OMV. The proposed model for ClyA represents a non-classical pathway to attack eukaryotic host cells.

Pore-forming toxins (PFTs) 2 represent the largest family of bacterial protein toxins and constitute important bacterial virulence factors (1,2). Their cytolytic function operates by introducing a large, water-filled pore into target cell membranes. These pores either deliver toxic effector proteins to the target cell or lead to cell lysis through leakage (3). Most bacterial PFTs are secreted into the extracellular environment in a water-soluble form, where they subsequently diffuse and assemble on host cell membranes. Pore-forming toxins are classified as ␣-PFTs or ␤-PFTs depending on the structure of the transmembrane pore, i.e. ␣-PFTs contain ␣-helical transmembrane domains, and ␤-PFTs form a ␤-barrel (1,2).
Cytolysin A (ClyA), also known as silent hemolysin A (SheA) or hemolysin E (HlyE), is a cytolytic ␣-PFT that causes the hemolytic phenotype of several Escherichia coli (E. coli) strains (4 -9). Its homologs are also found in other pathogenic organisms, including Salmonella typhi and Shigella flexneri (10,11). The E. coli ClyA monomer is a 34-kDa soluble protein that has a rod shape formed by a core bundle of four long ␣-helices (ϳ90 Å long) (10). At the end of the bundle that contains the N-terminal region, an additional shorter (ϳ30 Å long) helix from the C-terminal region packs against the core bundle, forming a five-helix bundle for about one-third the length of the molecule. The structure of the transmembrane pore shows a hollow funnel consisting of 12 subunits (protomers) (12,13). Each protomer contributes one amphipathic ␣-helix that packs in an iris-like structure to form the transmembrane barrel. The narrowest opening of the channel at the transmembrane site has a diameter of 35 Å while the top of the funnel is 70 Å (12).
The general model for multimeric PFT attack action involves three steps (1): (i) docking of the soluble monomer to the target cell membrane; (ii) assembly of the monomer into a ring-like prepore structure that lacks the transmembrane domain structure at the membrane surface; (iii) penetration of the transmembrane domain across the target cell membrane. In the first step, cholesterol, carbohydrates, or membrane proteins on the host cell surface may serve as receptors for the association of the toxins with the membrane (14 -18). ClyA toxin is also believed to follow a similar strategy when attacking host cells (1). However, unlike the well-studied ␤-toxins ␣-hemolysin and protective antigen from anthrax toxin, which are secreted by Gram-positive bacteria into the extracellular environment as a soluble monomer (19,20), ClyA is secreted from E. coli via a vesicle-mediated pathway (21)(22)(23). Similar to the budding of yeast cells, the outer membrane of E. coli bubbles out and pinches off to form the outer membrane vesicles (OMVs) (24). During the formation of the OMV, many outer membrane proteins as well as periplasmic proteins are incorporated into the OMV. The vesicle-mediated pathway has been found to deliver several toxins, including heat-labile enterotoxin and Shiga toxin (25)(26)(27). So far, it remains unclear how these toxins are released from the OMVs to carry out their cytolytic function (25).
Of note, Wai et al. discovered that ClyA in the OMV adopted a ring-like oligomeric structure when viewed under an electron-microscope (21). This observation contradicts the typical phenylmethanesulfonylfluoride (Sigma). The cell lysate was incubated on ice for 30 min. To reduce the viscosity, 50 kU DNase I and 3 mM MgCl 2 were added to the mixture and incubated for 30 min at room temperature. The lysate was centrifuged at 20,000 ϫ g for 30 min. After passing through a 0.45 m filter, the supernatant sample was loaded onto the Ni-NTA affinity column that was equilibrated with 50 mM Tris⅐HCl, pH 8.0, 150 mM NaCl. His-tagged ClyA proteins were eluted with the same buffer containing 0.5 M imidazole. The eluted protein was further purified by Superdex 200 10/300 gel filtration column (GE Healthcare) that was equilibrated with 25 mM Tris⅐HCl, pH 8.0, 150 mM NaCl. The monomeric ClyA proteins were collected and stored at Ϫ80°C. The purity of the protein (Ͼ95%) was verified by 15% SDS-PAGE (supplemental Fig. S1).
Gel Filtration Analysis of Oligomerization-Gel filtration was performed at room temperature using a Superdex 200 10/300 gel filtration column (GE healthcare). For detergent-triggered oligomerization, samples were analyzed in buffer 50 mM Tris⅐HCl, pH 8.0, 150 mM NaCl, 0.01% (w/v) n-dodecyl ␤-Dmaltoside (DDM) at room temperature. For detergent-independent oligomerization, samples were incubated with buffer containing no DDM at the indicated temperature prior to GFC analysis. The buffer used for GFC was 50 mM Tris⅐HCl, pH 8.0, 150 mM NaCl.
Intrinsic Fluorescence Studies-ClyA proteins were analyzed at a concentration of ϳ3 M at 25°C. For intrinsic fluorescence measurements the excitation wavelength was 280 nm and emission spectra were collected at 290 -410 nm using a Fluorolog-3 spectrofluorimeter. The fluorescence emission spectra of corresponding buffers were subtracted from the emission spectrum of each ClyA sample. For DDM-induced oligomerization studies, 15 l of DDM stock solution (10% w/v) was added to 1.5 ml of ClyA monomer at a final concentration of 0.1% (w/v). Subsequently, the emission fluorescence spectra were recorded at 3-min intervals.
Liposome Preparations-All lipids were purchased from Avanti Polar Lipids (Alabaster, AL). E. coli total extract or brain total extract lipids in chloroform was dried at 20 -23°C under nitrogen and then kept under vacuum for at least 3 h. Lipids were re-hydrated in buffer 50 mM Tris⅐HCl, pH 8.0, 150 mM NaCl to a 25 mg/ml final concentration of total lipids, and incubated for 30 min at 20 -23°C with vortexing at 5 min intervals. The suspended lipid mixture was frozen in liquid nitrogen and thawed at 37°C for a total of three cycles to reduce the number of multi-lamellar liposomes. Hydrated lipids were extruded 21 times through a 0.4 m pore size polycarbonate filter (Whatman) using an Avanti Mini-Extruder (Alabaster, AL). The resultant liposomes were stored at 4°C and used within 2 weeks of production.
Analysis of Hemolytic Activity-The liquid hemolysis assay with sheep blood cells was used to measure the hemolytic activity of the ClyA proteins (28). Briefly, Remel sheep defibrinated blood (Thermo Scientific) was washed with HyClone Dulbecco's phosphate-buffered saline (DPBS) buffer (Thermo Scientific) and diluted 4-fold in DPBS buffer. ClyA proteins (7 g) were added to 250 l of blood cells and incubated at 37°C for 15 min. Samples were then centrifuged at 22,000 ϫ g at 4°C for 8 min. The absorbance of the supernatant at 540 nm was measured to determine the released hemoglobin. Total hemolysis (100%) was defined by incubation of red blood cells in MilliQ water, in place of buffer.
Analysis of Pore-forming Activity by Fluorescence Quenching-The pore-forming activity of ClyA proteins was also assessed by using a Tb(DPA) fluorescence quenching assay. Tb(DPA) 3 3Ϫ loaded liposomes (6 l of ϳ 2 mM giving a final concentration of 12.5 M total lipids) were added to 994 l of quenching buffer (100 mM NaCl, 50 mM HEPES, 5 mM EDTA, pH 7.5) containing 250 nM protein just before the measurement. Samples were excited at 278 nm and the net initial emission intensity (F 0 ) at 544 nm was determined after equilibration of the sample at 25°C for 2 min. The samples were then incubated for 1 h at 37°C. After re-equilibration at 25°C, the final net emission intensity (F f ) of the sample was determined and the fraction of quenched fluorophore was calculated using 1 Ϫ F f /F 0 . Measurements were repeated 4 times for each condition.
Oxidation of ClyA Proteins-The formation of intramolecular disulfide bonds were catalyzed using the oxidizing reagent Cu(phenanthroline) 2 following a previous protocol (29). ClyA proteins were incubated with 1.5 mM Cu(phenanthroline) 2 at room temperature for 30 min. EDTA (5 mM, final concentration) was then added to quench the reaction. All the chemicals were then removed by buffer exchange using a Centricon (Millipore) with a 3 kDa cut-off. To reduce the sample, oxidized proteins were incubated with 20 mM DTT for 60 min at room temperature. DTT was then removed by buffer exchange before the analysis of hemolytic activity.
Single Channel Study of ClyA-Planar lipid bilayer experiments were performed in an apparatus partitioned into two chambers with a 25 m-thick Teflon film. An aperture of ϳ100-m diameter had been made near the center of the film with an electric arc. Each chamber was filled with 25 mM Tris⅐HCl, pH 8.0, 1 M KCl. A Ag/AgCl electrode was immersed in each chamber with the cis chamber grounded. A positive potential indicates a higher potential in the trans chamber. 1,2diphytanoyl-sn-glycerol-3-phosphocholine (Avanti Polar Lipids) dissolved in pentane (10% w/v) was deposited on the surface of the buffer in both chambers and monolayers formed after the pentane evaporated. The lipid bilayer was formed by raising the liquid level up and down across the aperture, which had been pretreated with a hexadecane/pentane (1:10 v/v) solution. The current was amplified with an Axopatch 200B integrating patch clamp amplifier (Axon Instruments, Foster City, CA). Signals were filtered with a Bessel filter at 2 kHz (unless otherwise stated) and then acquired by a computer (sampling at 50 s) after digitization with a Digidata 1320A/D board (Axon Instruments). Data were analyzed with Clampex 10.0 software.

RESULTS
Assembly of ClyA in the Absence of Detergent-It is often observed that PFT monomers assemble to form the oligomeric transmembrane pores upon contacting with detergents or lipid vesicles (30 -32). Similar to previous findings, the retention volume of ClyA monomers shifted from 15 ml to 10 ml by gel filtration chromatography (GFC) after incubation with 0.1% (w/v) DDM overnight at 4°C (Fig. 1a) (12,33). This high molecular weight oligomer was previously shown to represent the transmembrane pore structure and is termed ClyA TM here (12, 33). During the purification, we noticed that the ClyA protein showed a tendency to form high molecular weight oligomers even in the absence of detergent or lipids. We therefore explored the oligomerization of ClyA in the absence of detergent or membranes using GFC. Two peaks with retention volumes of 8.5 ml and 9.5 ml appeared in chromatograms after incubation of the ClyA monomer at 37°C in buffer containing 50 mM Tris⅐HCl, pH 8.0, 150 mM NaCl for 2 h (Fig. 1b). Fractions corresponding to two peaks were named oligomer 8 (ClyA O8 ) and oligomer 9 (ClyA O9 ) based on their respective retention volumes. Notably, these two fractions migrated faster in the GFC column than ClyA TM (Fig. 1, a and b). Because the ClyA monomer contains two cysteines, it is possible that ClyA formed oligomers through disulfide bond formation during the incubation at 37°C. To assess this, the same experiment was repeated in buffers containing freshly prepared 10 mM DTT at 37°C. The presence of DTT did not modify the elution profile of ClyA oligomers, excluding the involvement of disulfide bridges (data not shown). The ClyA O8 and ClyA O9 were pooled separately and re-analyzed by GFC. The two proteins were still eluted at their original retention volumes of 8.5 ml and 9.5 ml, respectively (supplemental Fig. S2). No monomer was observed in the chromatogram, indicating that ClyA O8 and ClyA O9 did not dissociate. Thus, we conclude that ClyA was able to oligomerize in the absence of detergents/membranes and that the isolated oligomers were stable in solution.
Conversion of Oligomer 8 and Oligomer 9 to the Transmembrane Pore-To investigate whether ClyA O8 and ClyA O9 can convert to the transmembrane pore, we incubated them with 0.1% DDM overnight at 4°C. The ClyA O8 shifted to a peak eluted at 15 ml, corresponding to the monomer, and a peak at 10 ml, corresponding to the fully-assembled transmembrane pore (Fig. 1c). The ClyA O9 also shifted to peak at 10 ml but little monomer is observed during this process (Fig. 1d). These results indicate that ClyA O8 may contain loosely packed monomers that can be solubilized by DDM. By contrast, ClyA O9 had a stable structure that did not disassemble in detergent solution. Rather, DDM triggered a conformational change in ClyA O9 that transformed it into the transmembrane pore. Because a portion of ClyA O8 also appeared to have converted to ClyA TM , it may contain a mixture of ClyA O9 -like oligomers with a stable structure and loosely packed monomers.
Many assembled ␤-toxin oligomers have shown tolerance to sodium dodecyl sulfate (SDS) treatment and remain as oligomers on SDS-PAGE (31, 34 -36). We were interested to see if ClyA oligomers would have different stabilities in SDS and migrate differently in the SDS-PAGE. We found that all three oligomeric forms of ClyA dissociated to monomer on SDS-PAGE (supplemental Fig. S3). This result suggests that the interaction in the oligomeric interface of ClyA is not as strong as those in ␤-PFTs.
Hemolytic Activities of the ClyA Proteins-The cytolytic activity of the oligomeric forms, ClyA O8 and ClyA O9 was determined using a hemolytic assay. We also quantitatively compared its functionality to the ClyA monomer. Sheep blood cells were incubated with ClyA proteins of the same weight concentration at 37°C for 15 min. The release of the hemoglobin associated with cell lysis was then measured. As a negative control, a denatured ClyA sample, which were pre-heated at 95°C for 30min prior to the hemolytic assay, was tested. A sample of PBS buffer was included to define zero activity. Results show that thermally-denatured ClyA has no hemolytic activity while the monomer has the highest activity (Fig. 2). ClyA O9 and ClyA O8 exhibited 65 and 30% of the monomer activity, respectively. The transmembrane pore (collected as peak 10 from GFC) only exhibited a background level of hemoglobin release, supporting the notion that this population had lost its ability to lyse cells. This result is in agreement with previous observations that transmembrane ClyA pores formed in detergent octyl-glucoside solution showed no pore-forming activity on cell membranes (37). Because both ClyA O9 and ClyA O8 were still functional in forming pores on erythrocyte membranes, we concluded that they might be oligomeric states with an ordered structure instead of disordered aggregates.
Conductance of ClyA Pores-Because ClyA can form transmembrane pores by converting the monomer or oligomer ClyA O9 in the presence of detergents, we asked if the stoichiometry of the transmembrane pores formed from these two pathways are the same. Because the ion conductance of a membrane channel is proportional to the pore size, which, in turn, is determined by the number of protomers in the transmembrane pore (38), we could address this question by comparing the conductance of ClyA channels formed in two different pathways. Here, single channel insertion of ClyA proteins into planar lipid bilayers was monitored by bilayer current recording experiments (Fig. 3a). Similar to the results of the electrophysiological studies of E. coli ClyA and its homologs from Salmonella typhi (6,39,40), the ClyA channels formed by monomer show a broad distribution from 5-15 nS with a major peak at around 11 nS (Fig. 3b), suggesting that ClyA might create pores of variable size. The broad distribution of conductance observed by electrophysiology agrees with the cryo-EM data showing ClyA pores of variable size in detergent solution (12). The ClyA channels formed by ClyA O9 and ClyA O8 exhibited similar conductance distribution patterns with the majority of ClyA O8 pores exhibiting a conductance between 6 -10 nS and ClyA O9 pores a conductance between 9 -16 ns (Fig. 3, c and d). The population of ClyA O9 channels shifted to slightly larger pores than those formed directly from monomer while ClyA O8 pores appeared smaller. So far, it is unclear why the conductance of the ClyA O8 , ClyA O9 , and ClyA TM pores are different considering that they migrate at the same retention volume by GFC (Fig. 1). Further detailed structural studies will be carried out to address this issue in future.
Interestingly, when we measured the conductance of the ClyA monomer sample after 7 days incubation at 4°C in DDM, the histogram of conductance showed almost a single population of 11 nS pores (Fig. 3e). This suggests that the 11 nS pore is the most stable form, since it survives a long-time incubation in detergent. By contrast, many other forms of ClyA lose the ability to insert into the lipid bilayer, probably due to aggregation or degradation. The dodecameric structure of the transmembrane pore determined by x-ray crystallography was also obtained from a DDM sample which contained a large variety of ClyA pores of different sizes (12). We suggest that the ClyA transmembrane pore with dodecameric structure might correspond to the 11 nS pores observed in the electrophysiology recordings.
Temperature-and Concentration-dependent Oligomerization-We were interested in how the temperature and the protein concentration affect ClyA assembly in solution. ClyA monomer was incubated at temperatures ranging from 4 -42°C for 2 h. At a temperature lower than 37°C, little ClyA oligomerization was observed (Fig. 4). Unlike the assembly in the detergent micelles which proceeded to completion at 4°C for 30 min, the detergentindependent assembly was strictly temperature-dependent and was a much slower process relative to the former. To see if oligo-merization was also concentration dependent, we incubated 0.1 mg ClyA protein at concentrations ranging from 1 g/ml to 0.1 mg/ml at 42°C for 2 h. Samples were then analyzed by GFC. In order to load enough protein sample to the GFC column for detection, the two lower concentration samples, 10 g/ml and 1 g/ml, were concentrated to 1 ml using an Amicon centrifuge concentrator at 4°C. To eliminate the possibility that oligomerization might also be induced during the centrifugation concentrating procedure, we incubated 100 ml 1 g/ml at 4°C for 2 h and concentrated the sample to 1 ml and loaded it to the GFC column as a control. We found ClyA remained as a monomer and no protein was eluted at either 8 or 9 ml retention volume demonstrating the concentration step does not induce oligomer formation (data not shown). The percentage of each population (ClyA O8 , ClyA O9 , and monomer) was calculated from the area of the peaks eluted at 8.5 ml, 9.5 ml, and 15 ml in the chromatogram that were analyzed by the Unicorn software. The data are summarized in Fig. 5. Even at the  lowest concentration (1 g/ml, ϳ30 nM), the ratio of the oligomer to monomer was not reduced compared those of the higher concentrations, suggesting the oligomerization was independent of the protein concentration. Because of the detection limit of GFC, we were not able to investigate lower concentrations. This result demonstrates that ClyA has a strong tendency to oligomerize even at a low concentration at physiological temperature.
Probing the Structural Arrangement of Oligomers by Fluorescence-To understand the structural arrangement of ClyA O8 and ClyA O9, we measured their intrinsic fluorescence spectra. The ClyA monomer contains two tryptophan and 13 tyrosine residues. As shown in Fig. 6a, the monomer had a fluorescence maximum at 315 nm while the fully assembled pore ClyA TM had a maximum at 340 nm. The fluorescence emission spectra of the ClyA O8 and ClyA O9 were very similar with the fluorescence maxima also at 340 nm, close to that of the transmembrane pore. The spectra from 345-400 nm overlapped well with that of the transmembrane pore while there is a blue shift of 2 nm in the spectra from 300 -340 nm (Fig. 6a). Because the spectra of ClyA O8 and ClyA O9 appeared to be in between the monomer and transmembrane pore, we suspected they might be an intermediate state. To test this, we studied a series of intermediate states of ClyA proteins during the transition to transmembrane pores. Here, DDM was added to ClyA monomer to a final concentration of 0.1% (w/v) of DDM at 23°C to trigger the pore formation. The spectrum was taken every 3 min until the spectra stabilized. Fig. 6b shows that the fluorescence intensity rose sharply upon the addition of detergent and the wavelength of the peak maximum shifted from 315 to 340 within the first 3 min. After the initial fluorescence jump, the fluorescence intensity gradually decreased and became stable after 30 min. The normalized spectra of these intermediates revealed the peak maxima undergo subtle blue-shift during the transition (Fig. 6c). Interestingly, the spectrum taken 9 min after the addition of the DDM overlaid well with the spectrum of ClyA O8 and ClyA O9 (Fig. 6d) , indicating that the structure of ClyA O8 and ClyA O9 might mimic an intermediate state of the oligomerization process.
We also investigated how the composition of the lipid membrane affects the assembly of ClyA. To avoid the occurrence of oligomerization in solution that would contribute to fluorescence spectrum, the ClyA monomers were incubated with liposomes prepared from porcine brain lipids or E. coli lipids extract at room temperature in a 1:50 protein to lipid molar ratio. The spectrum of the brain lipid liposomes sample overlapped very well with the spectrum of ClyA TM (Fig. 6e), which demonstrated that ClyA proteins have transformed into transmembrane pores on the brain lipid liposomes. In contrast, the spectrum of the ClyA incubated with E. coli liposomes did not resemble the DDM treated sample (Fig. 6f). Instead, it was located in between the monomer and the spectrum of the ClyA O9, suggesting that ClyA does not form the transmembrane pore on E. coli lipid membranes.
Effect of Lipid Composition on Pore-forming Activity-To further examine the pore-forming activity of ClyA proteins on liposomes, we used a fluorescence quenching assay. The liposomes with Tb(DPA) 3 3Ϫ encapsulated inside were added to buffers containing the fluorescence quencher EDTA. EDTA can diffuse into the liposome through ClyA pores on the liposome membrane and quench the fluorescence. Fig. 7 shows that ClyA monomer caused a significant fluorescence decrease in the brain lipid liposomes. Consistent with our hemolytic assay, ClyA O9 and ClyA O8 also induced fluorescence quenching but with reduced activity. On the contrary, no fluorescence quenching was observed with any of the ClyA protein samples incubated with E. coli liposomes. This experiment confirms that ClyA proteins do not form transmembrane pores on E. coli lipid membranes.
Direct Conversion of ClyA O8 and ClyA O9 to Transmembrane Pores-The data above revealed that Cly O8 and Cly O9 can convert to the transmembrane pore when in contact with detergent micelles or lipid membranes. This could be achieved through two possible pathways: (i) ClyA O8 and ClyA O9 dissociate into monomers then re-assemble in the micelles/lipid membrane or, (ii) ClyA O8 and ClyA O9 directly convert to the transmembrane pore without undergoing the dissociation step. To distinguish between these pathways, we monitored the conversion of ClyA O8 and ClyA O9 in the presence of brain lipid membranes by intrinsic fluorescence. Liposomes were added to monomer and the fluorescence was recorded (Fig. 8a). After 5 min, the spectrum showed two peak maxima at 314 nm and 340 nm indicating there was still monomer remaining in the sample. This result suggests the association process of the ClyA monomers to oligomers on the lipid membrane occurs on the time scale of minutes, which is much slower than the ␣HL toxin (Ͻ5ms) (41). If ClyA O8 and ClyA O9 dissociate into monomers, we expect to observe the spectrum of the monomeric species due to the relatively slow association process of ClyA monomers. Adding liposomes to ClyA O8 and ClyA O9 induced a slight change at the 300 -340 nm range (Fig. 8, b and c), however we did not observe a significant shift of the spectrum toward the monomer peaks. Thus, this experiment suggests that ClyA O8 and ClyA O9 directly convert to the transmembrane pore on the lipid membrane without first dissociating to monomers. Effect of Redox Environment on the Activity of ClyA-ClyA monomer contains two cysteines (C87, C285). The structure of ClyA monomer revealed that these cysteines were located within 5.2 Å (C␣ to C␣), a distance that would allow a disulfide bond to form. In the transmembrane pore structure, the cysteines move away from each other to be 6.8 Å apart, which is the longest possible distance for a disulfide bond to form between to cysteine residues (42). This raised a possibility that the disulfide formed between the cysteines might act as a switch to con-trol ClyA activity. To test this, we induced the disulfide bond formation by incubating the protein with Cu(phenanthroline) 2 . GFC analysis of the sample showed ClyA proteins remained monomeric, indicating the protein did not form any inter-molecular disulfide bonds (supplemental Fig. S4). The oxidized ClyA was subjected to the hemolytic assay. Fig. 9 shows that the oxidized ClyA lost 90% of hemolytic activity. Incubation in 20 mM DTT for an hour could recover 60% the activity. The incomplete recovery of the activity is likely due to the close packing of the helix bundles that occludes the disulfide bond from the reducing chemicals. As a control, the double cysteine knock out mutant ClyA ⌬Cys exhibited no response to the oxidation/reduction procedure. These results indicated that formation of the intra-molecular disulfide bond could inhibit the pore-forming activity presumably by preventing the occurrence of conformational change necessary for the formation of oligomeric transmembrane pore.

DISCUSSION
Previous work on ClyA has also shown that ClyA oligomerizes in solution without detergent at 37°C (43). In this example, monomer, dimers and high molecule-weight oligomers (8-10mers as speculated by the authors) were observed by GFC (43). The 8 -10mer fraction lacked hemolytic activity. However, our data clearly demonstrated that both ClyA O8 and ClyA O9 lysed blood cells while the thermally denatured ClyA did not.  Because ClyA O8 and ClyA O9 retains both a stable structure and hemolytic activity, we expect that they are not a disordered aggregate but rather an active oligomeric form of ClyA proteins. In addition, the intrinsic fluorescence of these soluble oligomers matched that of an intermediate state. This suggests the detergent-independent oligomer resembles an intermediate state between the monomer and fully assembled transmembrane pore. Since ClyA O8 and ClyA O9 were formed in solution, it is unlikely that the ␣-helical transmembrane barrel has formed, as this would expose the hydrophobic outer surface of the barrel to aqueous solution. Therefore, we believe that ClyA O8 and ClyA O9 could be oligomeric forms without the hydrophobic transmembrane barrel domain, i.e. the prepore structure. Also, we have shown that Cly O8 and ClyA O9 can convert directly to transmembrane pore. In summary, we propose that ClyA O8 and ClyA O9 are intermediate states between the monomer and transmembrane ClyA channel. Since these two populations were eluted earlier than transmembrane pores in GFC, it might have a less compact structure than the transmembrane pore.
Although many PFTs require detergents or lipid bilayers for assembly to oligomeric prepore, the protective antigen of anthrax toxin also oligomerizes to a prepore structure in the absence of detergents or lipid membranes (35,44). In fact, the water-soluble prepore was formed by the oligomerization of protective antigen 63 which derives from the proteolytic removal of a 20 kDa fragment from the full length anthrax monomer. Similarly, Monalysin, a PFT from the Drosophila pathogen Pseudomonas entomophila, forms a multimeric structure in solution after proteoactivation (45). Therefore, not all PFTs require lipid membranes or similar environments to form prepore structures. Although in vitro the PA63 and Monalysin undergo oligomerization in solution after protease treatment, in nature the oligomerization of the PA63 and Monalysin is triggered by the proteases expressed on the host plasma membrane, where they subsequently assemble. The activation of ClyA may be regulated by the change of its redox status during the secretion of ClyA into OMVs (10,21,43). ClyA in the periplasm contains a C87-C285 disulfide bond (6,21,46), whereas ClyA within OMV is reduced (21). The disulfide bond may prevent assembly of functional pore complexes, as oxidized ClyA showed a decreased hemolytic activity compared with reduced ClyA, which is consistent with previous studies (10,21,43). This notion is further supported by the observation that ClyA expressed in an E. coli (dsbA Ϫ and dsbB Ϫ ) strain, which is deficient in periplasmic disulfide bond formation, displays a significant hemolytic activity on blood agar plates, an effect that is not detectable for ClyA  expressed in wild-type strain (21). Therefore, the secretion of ClyA to OMVs where DsbA and DsbB are absent may reduce the intracellular disulfide bonds and activate the toxin for oligomerization (21).
Some speculate that ClyA forms transmembrane pores upon interaction with the membrane of the OMV (21). However, this contradicts two observations: (i) the pre-formed ClyA transmembrane pore lacks hemolytic activity as demonstrated by our study and previous data (37). A transmembrane ClyA pore embedded in the OMV membrane loses functionality for further attack. (ii) the pore-forming activity of ClyA is strongly dependent on cholesterol in the membrane, which suggests that cholesterol may facilitate the transmembrane domain insertion (46). Since bacterial outer membranes contain no cholesterol, the transformation of ClyA into the transmembrane pore inside of the OMV should be very slow. Our intrinsic fluorescence analysis supports this notion: the spectra of ClyA with liposomes reveal transmembrane pores in brain lipids but ClyA cannot transform in the presence of E. coli liposomes. The Tb fluorescence quenching assay further confirmed that ClyA cannot form transmembrane pores on E. coli lipid membranes. Thus, it is reasonable to believe the oligomeric structure of ClyA in the OMV may not be the transmembrane pore.
The average expression level of a protein in E. coli is ϳ1000 copies, roughly 1 M in the cytosol (47). Our data demonstrated that ClyA forms oligomers at 37°C at concentrations as low as 30 nM. Under physiological conditions ClyA can associate into ClyA O9 -like prepore structures once activated in the OMV and this process can occur independently of lipid membrane. Taken together, these data strongly suggest that the secreted ClyA proteins oligomerize to a prepore structure instead of the transmembrane pore inside the OMV before they reach the target cell membrane.
We propose that ClyA adopts a non-classical assembly pathway including the following steps ( Fig. 10): (i) ClyA is expressed as a monomer in the cytosol and exported to the periplasmic space where it remains inactive in an oxidized form; (ii) ClyA is secreted in outer membrane vesicles in monomer form; (iii) Cleavage of the disulfide bond in the OMV activates ClyA monomers which subsequently oligomerize to the prepore structure. How the OMV then delivers the active ClyA prepore to the host cell remains unknown. One plausible mechanism involves internalization of the vesicle and subsequent release of the cargo proteins to the cytoplasm of the target cells (25). Further studies are underway to identify how ClyA is released.
In conclusion, we have shown that ClyA protein assembles into a functional oligomer in the absence of detergents and membranes. The oligomeric form represents a prepore intermediate state that may resemble the ClyA oligomer structure in the OMVs. These data provide insight regarding the oligomeric state of ClyA proteins in the OMV, which represents an important step toward resolving the overall mechanism of ClyA. High-resolution crystal structures of the prepore might be obtained in the near future, and they will be essential for determining how the prepore becomes the hemolytic pore. The classical PFT attacking mechanism involves the secreted monomer docking to the host membrane, followed by rapid assembly to the prepore structure. A conformational change triggers the formation of the transmembrane domain. By contrast, monomeric ClyA is secreted in OMVs where the redox environment cleaves intramolecular disulfide bonds and activates ClyA. ClyA then assembles to a prepore structure in the vesicle. Through an unknown mechanism, OMV release ClyA, which then form a transmembrane pore on the host cell.