Soluble Oligomers of the Pore-forming Toxin Cytolysin A from Escherichia coli Are Off-pathway Products of Pore Assembly*

The α-pore-forming toxin Cytolysin A (ClyA) is responsible for the hemolytic activity of various Escherichia coli and Salmonella enterica strains. Soluble ClyA monomers spontaneously assemble into annular dodecameric pore complexes upon contact with membranes or detergent. At ClyA monomer concentrations above ∼100 nm, the rate-limiting step in detergent- or membrane- induced pore assembly is the unimolecular reaction from the monomer to the assembly-competent protomer, which then oligomerizes rapidly to active pore complexes. In the absence of detergent, ClyA slowly forms soluble oligomers. Here we show that soluble ClyA oligomers cannot form dodecameric pore complexes after the addition of detergent and are hemolytically inactive. In addition, we demonstrate that the natural cysteine pair Cys-87/Cys-285 of ClyA forms a disulfide bond under oxidizing conditions and that both the oxidized and reduced ClyA monomers assemble to active pores via the same pathway in the presence of detergent, in which an unstructured, monomeric intermediate is transiently populated. The results show that the oxidized ClyA monomer assembles to pore complexes about one order of magnitude faster than the reduced monomer because the unstructured intermediate of oxidized ClyA is less stable and dissolves more rapidly than the reduced intermediate. Moreover, we show that oxidized ClyA forms soluble, inactive oligomers in the absence of detergent much faster than the reduced monomer, providing an explanation for several contradictory reports in which oxidized ClyA had been described as inactive.

Here we examined this alternative pore formation mechanism of ClyA further. We found no evidence for pore formation or efficient target cell lysis when these soluble oligomers were mixed with detergent or horse erythrocytes, respectively. In fact, the soluble ClyA oligomers showed only 1-2% hemolytic activity relative to ClyA monomers. In addition, no pores could be observed by electron microscopy after incubation of soluble oligomers with detergent (n-dodecyl-␤-D-maltopyranoside (DDM)) under conditions where the majority of ClyA monomers assembled into dodecameric pore complexes. We therefore interpreted the soluble oligomers as off-pathway products of ClyA pore formation. This conclusion is supported by the finding that the kinetics of soluble oligomer formation coincided with loss in hemolytic activity.
In addition, we compared oligomer formation and hemolytic activity of the reduced, dithiol form of ClyA (ClyA red ) with pore formation of the oxidized, disulfide form (ClyA ox ) to address contradictory reports in the literature, which had described ClyA ox either as assembly-incompetent and inactive (13,17,20) or as assembly competent (14). In the present study, we show that ClyA ox monomers have essentially the same specific hemolytic activity compared with monomeric ClyA red but form inactive, soluble oligomers 13-14 times faster than ClyA red upon incubation in the absence of detergent or membranes. The results provide a plausible explanation for previous, contradictory reports on the assembly competence of ClyA ox .
Finally we also investigated the mechanism of the DDM-induced monomer-to-protomer formation of ClyA ox with singlemolecule Förster resonance energy transfer (FRET) measurements, demonstrating that ClyA ox follows the same reaction pathway as ClyA red , in which an unstructured off-pathway intermediate (I ox ) is formed (21). Due to the lower stability of I ox compared with I red , assembly-competent protomers are even formed one order of magnitude faster from monomeric ClyA ox compared with monomeric ClyA red .

Experimental Procedures
Materials-Chemicals of the highest available purity were purchased from Merck KGaA (Darmstadt, Germany) or Sigma-Aldrich. Dithiothreithol (DTT), EDTA and ␤-mercaptoethanol were obtained from AppliChem. DDM was purchased from Anatrace. Brain total lipid extract was purchased from Avanti Polar Lipids, and horse erythrocytes were obtained from Oxoid AG.
Production and Purification of Reduced ClyA-To exclude any influence of an N-or C-terminal polyhistidine purification tag on the activity and assembly of ClyA, we introduced a TEV (tobacco etch virus) protease cleavage site between the N-terminal His 6 tag and the natural ClyA sequence into the previously described ClyA expression plasmid derived from pET11a (14). ClyA production in E. coli was carried out at 20°C for 15 h, as described (16). ClyA was purified from the soluble fraction of the cell extract by Ni 2ϩ -NTA affinity chromatography followed by chromatography on hydroxyapatite as described (14). The N-terminal His 6 tag of ClyA was then cleaved by recombinant His 6 -tagged TEV protease (22), and the resulting wild-type ClyA (residues Thr2-Val 303) was obtained in the flow-through of a second Ni 2ϩ -NTA affinity chromatography as described (16). All purification steps and the proteolytic cleavage were performed under reducing conditions with buffers containing 2 mM ␤-mercaptoethanol (when Ni 2ϩ -NTA columns were used) or 2 mM DTT. The yield of reduced wild-type ClyA (ClyA red ) was 16 mg of ClyA/liter of bacterial culture. The concentration of ClyA was determined via its specific absorbance at 280 nm (30370 M Ϫ1 cm Ϫ1 for ClyA red ). The ClyA red monomer was stored in PBS buffer (20 mM potassium phosphate, pH 7.3, 150 mM NaCl, 0.1 mM EDTA) with 2 mM DTT at 4°C and showed no oligomerization within 7 days.
Preparation and Reduction of Oxidized ClyA-To generate the disulfide-bonded oxidized form of ClyA, ClyA red (40 M) was incubated in PBS buffer with 0.5 mM CuCl 2 (i.e. 0.4 mM free CuCl 2 ) as a catalyst of air oxidation for 3-4 h at 22°C. These conditions guaranteed complete oxidation of the Cys-87/Cys-285 pair of ClyA (see below). Oxidized ClyA (ClyA ox ) was subsequently dialyzed at 4°C against PBS containing 2 mM EDTA and subjected to gel filtration on a Superdex 200 column (GE Healthcare Life Sciences) equilibrated with PBS to separate oxidized monomers from oxidized oligomers. The absence of free thiols in the purified oxidized monomers was confirmed by Ellman's assay (23) under denaturing conditions (4.0 M guanidinium chloride, pH 8.0) and by analytical reversed phase HPLC (RP-HPLC) at 30°C on a Zorbax SB300 C8 column (Agilent) using a water-acetonitrile gradient from 50 to 80% (v/v) acetonitrile in 0.1% trifluoroacetic acid (see Fig. 2A). The ClyA ox monomer was stored in PBS buffer without DTT at 4°C and showed no oligomerization within 1.5 days.
The reduction of ClyA ox (monomer, oligomer, or assembled pore complex in 0.1% DDM) was performed in PBS with 100 mM DTT, pH 7.3, at 37°C and quantified by RP-HPLC as described previously (16). The reduction of unfolded ClyA ox was carried out in 50 mM MOPS-NaOH, 4.0 M guanidinium chloride, pH 7.3, with 20 mM DTT at 37°C. Reactions were quenched after different time intervals by the addition of formic acid (12% final concentration), and samples were analyzed via RP-HPLC. The peaks corresponding to oxidized and reduced ClyA were integrated, and the data were evaluated according to pseudo first-order kinetics. The reaction between monomeric ClyA red (4 M) and oxidized DsbA ox (86 M, corresponding to the periplasmic DsbA concentration in E. coli (24)) at pH 7.3 and 37°C, was analyzed in the same way.
Purification and Oxidation of Förster Resonance Energy Transfer (FRET) Donor/Acceptor-labeled ClyA for Single-molecule Measurements-A ClyA variant with cysteine residues introduced at positions 56 and 252 (ClyACys) was expressed, purified, and labeled at Cys-56 with Alexa Fluor 488 and at Cys-252 with Alexa Fluor 594 as described previously (21). Air oxidation of the natural cysteine pair Cys-87/Cys-285 in the labeled ClyACys variant (4 M) was performed by incubation in 50 mM KH 2 PO 4 /K 2 HPO 4 pH 7.4, 150 mM NaCl, 10% (v/v) glycerol, and 0.5 mM CuCl 2 for 5 h at 22°C. Oligomers formed during this reaction were removed from the monomers by gel filtration on a Superdex 75 10/300 column (GE Healthcare Life Sciences) in PBS. The quantitative formation of the disulfide bonded Cys-87-Cys-285 was confirmed by reversed phase HPLC on an XTerra RP8 (30 ϫ 4.6 mm) column (Waters) in 0.1% (v/v) trifluoroacetic acid using a gradient from 30 to 60% acetonitrile.
Oligomerization of ClyA-Oligomers of reduced or oxidized ClyA were formed in the absence of detergents or lipids by incubation at 37°C (17). For preparative purposes, ClyA red (40 M) was incubated overnight at 37°C in PBS, pH 7.3, and subsequently subjected to gel filtration (Superdex 200) to separate the oligomer from the remaining monomer. In the case of ClyA ox , a 4-h incubation at 22°C (conditions of oxidation of ClyA) resulted in a sufficient amount of oligomeric ClyA ( Fig.  2A). The kinetics of oligomerization of ClyA ox and ClyA red in PBS at 4°C or 37°C were determined by quantification of the respective amount of monomeric or oligomeric ClyA via gel filtration on a ProSEC 300S column (Agilent) after different incubation times, starting with ClyA monomer at 5 M concentration. The decrease in the monomer over time was fitted mono-exponentially, assuming a unimolecular rate-limiting step, which is consistent with previous findings that oligomer formation is concentration-independent at micromolar monomer concentrations (17).
Negative Stain Transmission Electron Microscopy-To trigger pore formation from monomers, ClyA (2 M in PBS) was incubated in 0.1% DDM for 1 h at 22°C. Samples were adsorbed on glow-discharged 300-mesh carbon-coated copper grids (Quantifoil) and negatively stained with 2 mM uranyl acetate. Images were recorded by a KeenView CCD camera using a FEI Morgagni electron microscope operating at an acceleration voltage of 100 kV.
Hemolysis Kinetics-The kinetics of ClyA-dependent lysis of horse erythrocytes were measured as described previously (25) by following the decrease in optical density at 650 nm using a Varian Cary 100 spectrophotometer (Agilent). Horse erythrocytes at a density of 2 ϫ 10 6 cells/ml in PBS, pH 7.3, were lysed at 37°C by the addition of ClyA (final monomer concentrations of 2-100 nM). Reactions with ClyA red additionally contained 2 mM DTT. Hemolysis kinetics were evaluated by linearly fitting the data points in the middle of the lysis reaction between 35 and 75% of the initial optical density. The slope of the linear decrease in optical density was defined as the maximum lysis velocity. The linear dependence of the maximum lysis velocity on the concentration of ClyA was defined as the specific hemolytic activity of the different ClyA species (16).
Circular Dichroism (CD) Spectroscopy-The monomerto-protomer transition and pore assembly of ClyA ox and ClyA red (9 M) in PBS at 22°C was initiated by the addition of DDM (final concentration, 0.1% (w/v)) and followed via the CD signal change at 225 nm using a temperature-controlled J715 CD spectrometer (Jasco). Samples of reduced ClyA additionally contained 2 mM DTT. Before and after assembly, the CD spectra of ClyA ox and ClyA red were recorded. The rate-limiting step of the monomer-to-protomer conversion was fitted according to a first-order reaction.
Single-molecule Measurements of ClyA ox -Kinetic singlemolecule FRET measurements were performed and evaluated essentially as described for the reduced ClyACys variant (21). The measurements were recorded on a modified MicroTime 200 instrument (PicoQuant) using a setup for pulsed interleaved excitation (PIE (26)) with two excitation pulses for the donor dye alternating with a single acceptor pulse. The initial phase of protomer formation was measured in a microfluidic mixer (27), and the later phase as well as the pore formation kinetics was constructed from repeated manual mixing measurements (Table 1). All measurements were done at a nominal concentration of labeled ClyACys of 300 pM in PBS with 0.001% (w/v) Tween 20 and 0.1% (w/v) DDM at 22°C. For recording pore formation kinetics with FRET, excess unlabeled oxidized WT ClyA was added at 10 or 100 nM to achieve the concentrations required to observe the oligomerization of protomers.
The model fitting of the protomer formation was performed as described for the reduced variant (21). The three populations in the transfer efficiency histograms were fitted with Gaussian distributions, once with free amplitudes and once with the amplitudes determined by the kinetic model. Due to the extensive overlap of the peaks, the position and width of the monomer peak function were fixed to the values from the first histogram measured in the microfluidic chip and to those of a reference measurement for the microfluidic mixing and manual mixing data, respectively. As the intermediate is not sufficiently populated in the manual mixing experiments, the position and width of the peak function of the intermediate were fixed to those of the fit of the microfluidic mixing data. The uncertainties of the fitted rate coefficients were estimated with a bootstrap analysis on the level of the individual photon bursts. New sets of photon bursts were synthesized by random sampling with replacement from the original data. One thousand synthetic data sets were created, and the resulting histogram time series were fitted as described above. The resulting distri- butions of the fit parameters are well described by normal distributions.

Results
Soluble ClyA Oligomers Are Formed Faster by ClyA ox Than by ClyA red and Are Assembly-incompetent-WT ClyA contains a single cysteine pair at positions 87 and 285 that is able to form a disulfide bond in both the monomer and the protomer (14), consistent with the small C ␣ -C ␣ distances between the cysteines of 5.2 and 6.8 Å in the structures of the monomer and the protomer, respectively (15) (Fig. 1). Both ClyA redox forms have been described in vivo. Although ClyA lacks an N-terminal signal sequence (28), it is secreted to the periplasm where it accumulates in its oxidized form (20,29). In contrast, the assembled ClyA pore complexes that are exported by the bacteria to the extracellular medium in outer membrane vesicles (OMVs) are composed of reduced ClyA protomers (20). The mechanisms that underlie the secretion of ClyA into the periplasm and its assembly in OMVs are still unknown.
The DDM-induced assembly of ClyA had been characterized in detail for the disulfide-free form of ClyA (ClyA red ) (21). That study has revealed that the unimolecular formation of assembly-competent protomers (P red ) from monomers (M red ) is ratelimiting for the formation of dodecameric pore complexes at concentrations above ϳ100 nM and is accompanied by the reversible formation of a monomeric off-pathway intermediate (I red ): 12 I red 7 12 M red 7 12 P red 33 (fast) 33 dodecameric pore complex.
To study the assembly properties of ClyA ox , we first expressed and purified the authentic, untagged ClyA monomer under reducing conditions (2 mM ␤-mercaptoethanol or DTT) via chromatography of a ClyA variant with an N-terminal His 6 tag on Ni 2ϩ NTA-agarose and chromatography on hydroxyapatite followed by specific cleavage of the N-terminal His 6 tag by TEV protease. With this procedure, we obtained pure ClyA red (Thr-2-Val-303), for which Ellman's assay under denaturing conditions confirmed the presence of two thiol groups per polypeptide chain. ClyA ox was obtained from ClyA red after incubation for 4 h with 0.4 mM free CuCl 2 as a catalyst of air oxidation in PBS buffer, pH 7.3, at 22°C in the absence of detergent. Ellman's assay revealed the quantitative formation of the Cys-87-Cys-285 disulfide bond. This was confirmed by a single peak in reversed phase HPLC runs, which allows the quantitative separation of ClyA ox from ClyA red (Fig. 2B).
The preparation of ClyA ox was immediately applied to gel filtration, which revealed three distinct species: the oxidized monomer (M ox ) and a small (SO ox ) and large (LO ox ) oligomer of ClyA ox with apparent molecular masses of 580 and 1180 kDa, respectively ( Fig. 2A). All three ClyA ox species were isolated and again analyzed for quantitative disulfide bond formation by reversed phase HPLC (Fig. 2B). The analysis confirmed the presence of the disulfide bond in M ox , SO ox , and LO ox and also showed that purified ClyA red stayed completely reduced, proving to be resistant to air oxidation in the absence of Cu 2ϩ (Fig.  2B).  showing that ClyA red remained resistant to air oxidation in the absence of CuCl 2 , whereas it was oxidized quantitatively in the presence of CuCl 2 . Samples were separated on an analytical C8 column in 0.1% trifluoroacetic acid and eluted with a water/acetonitrile gradient. Due to the denaturing conditions, oligomeric ClyA dissociated so that only a single peak was detected in each run. The HPLC retention times corresponding to ClyA red and ClyA ox are indicated. C, negative stain electron micrographs of samples of the purified ClyA ox species LO ox , SO ox , and M ox (see A) and the isolated ClyA red species LO red and M red (see D) (2 M total monomer concentration in each sample) after incubation with 0.1% DDM for 1 h at 22°C. The results show that only M ox and M red were able to form intact pore complexes. Scale bar: 100 nm. D, comparison of the spontaneous oligomerization propensity of M ox (black lines) and M red (red lines) (5 M each) at 37°C in PBS buffer, pH 7.3. Purified M ox or M red was incubated for 0, 1, or 5 h at 37°C, and the reaction products were separated at 22°C in PBS on an Agilent 300S gel filtration column. The results show that M ox has a higher oligomerization tendency than M red and that both redox species can form large and small oligomers. E, kinetics of spontaneous oligomerization of M ox and M red (5 M each) at 37°C in PBS buffer, pH 7.3, analyzed via the decrease in the monomer peak. Data were normalized to the initial peak area of the monomer and were fitted to a first-order decay (solid lines), yielding apparent rate constants of 0.69 Ϯ 0.02 h Ϫ1 for ClyA ox and 0.052 Ϯ 0.002 h Ϫ1 for ClyA red , respectively.
To test the ability of M ox , SO ox , and LO ox to form dodecameric pore complexes, all species were isolated by gel filtration ( Fig. 2A) and incubated with 0.1% DDM to trigger pore formation under identical conditions (total monomer concentration of 2 M, 1-h incubation time, 22°C, and pH 7.3). Electron microscopy showed that only M ox readily assembled into pores complexes, whereas LO ox and SO ox showed no comparable pore formation activity (Fig. 2C, top panels).
Next, we found that ClyA red also formed small and large oligomers (SO red and LO red , respectively) in the absence of detergent with the same apparent molecular masses as SO ox and LO ox but with slower assembly kinetics. Fig. 2D shows the gel filtration profiles of the kinetics of spontaneous oligomerization of ClyA ox and ClyA red at pH 7.3 and identical initial monomer concentrations of 5 M, initiated by a temperature shift from 4 to 37°C. The kinetics of oligomer formation, recorded via the decrease in the peak areas of the monomers, revealed oligomerization half-lives of 60 and 800 min at 37°C for ClyA ox and ClyA red , respectively ( Fig. 2E and Table 2). Even at 4°C, slow ClyA oligomerization occurred under these conditions with half-lives of 103 and 1410 h for ClyA ox and ClyA red , respectively (Table 2). Together, these results show that the spontaneous formation of the species SO and LO is comparably slow and an intrinsic property of both ClyA redox forms ClyA ox oligomerizes about 14 times faster than ClyA red at both 4 and 37°C. Oligomerization is strongly favored at the physiological temperature of 37°C. In addition, Fig. 2C demonstrates that M ox , in contrast to previous reports describing the inability of ClyA ox to form pores (20) or its lack of cytotoxicity (13,17), readily assembles into annular pore complexes. Fig. 2D shows that the small oligomeric form SO red was much less populated than SO ox during spontaneous oligomerization of ClyA in the absence of detergent, so that we could not separate SO red from the much larger LO ox peak. However, we could isolate LO red by preparative gel filtration. Like LO ox , LO red proved to be assembly-incompetent upon the addition of 0.1% DDM, whereas M red efficiently assembled into pore complexes as shown previously (14,16) (Fig. 2C, lower panels). Thus, all oligomeric forms of ClyA that formed in the absence of detergent and could be isolated proved to be assembly-incompetent in vitro in the presence of DDM in a time frame in which monomeric ClyA was quantitatively incorporated into pores.
The Hemolytic Activity of Oligomeric Forms of Oxidized and Reduced ClyA Is Decreased 100-Fold Relative to the Respective Monomeric Forms-We next tested the ability of the purified oligomeric forms LO ox , SO ox , and LO red to form active pores in target cells. To this end, we compared their lytic activity toward horse erythrocytes with the hemolytic activity of the monomers M ox and M red . Fig. 3A shows the hemolysis kinetics at pH 7.3 and 37°C initiated by mixing erythrocyte suspensions (2 ϫ 10 6 cells/ml) with LO ox , SO ox , LO red , M ox , or M red at identical total monomer concentrations of 10 nM. The reaction was recorded via the decrease in optical density at 650 nm as a measure of erythrocyte lysis. Whereas M ox and M red proved to be highly active and, after a lag time of about 100 s, caused complete lysis within 200 s, the oligomers LO ox , SO ox , and LO red showed a strongly reduced activity, with lysis half-lives above 2000 s (Fig.  3A). For quantification of the specific hemolytic activities of all purified ClyA species, we used the recently established linear dependence of the maximum lysis velocity (see "Experimental Procedures" for the definition) for ClyA concentration in the range of 1 to 100 nM ClyA monomer (16). Fig. 3B and Table 3 show that freshly prepared M ox and M red had high specific hemolytic activity, with M ox even 1.2 times more active than M red (1.86 Ϯ 0.14 mOD s Ϫ1 nM Ϫ1 and 1.51 Ϯ 0.07 mOD s Ϫ1 nM Ϫ1 , respectively). In contrast, the specific activities of the oligomers of ClyA red and ClyA ox were dramatically reduced to 0.9% (LO ox and LO red ) and 1.4% (SO ox ) of the activity of M red ( Fig. 3B and Table 3). These results demonstrate that the spontaneous formation of ClyA oligomers in the absence of detergent or membranes inhibits the formation of active pores in target cells. This finding provides strong evidence that the species LO and SO of both ClyA redox forms are off-pathway products of pore formation, reminiscent of oligomeric off-pathway species in the ␣-PFT equinatoxin II from the sea anemone Actinia equina (30) that neither back-react rapidly to active monomers nor integrate into membranes and then become functional pores. The minute hemolytic activity of the oligomers detected in Fig. 3B may result from a very slow dissociation of inactive oligomers to active monomers upon dilution to the low ClyA concentration ranges (1-100 nM) used in the hemolysis assays.
The high hemolytic activity of M ox contradicts the previously reported assembly incompetence of oxidized ClyA (13,17,20). To test whether the contradictory data in the literature might be the result of the comparably rapid decrease in active monomers in preparations of ClyA ox (Fig. 2E), we tested whether the decrease in the concentration of assembly-competent monomers upon oligomerization of ClyA ox and ClyA red coincided with a loss of hemolytic activity. For this purpose, ClyA ox and ClyA red were again incubated under the conditions described in Fig. 2E, which favor oligomerization (initial monomer concen-   Fig.5D). MARCH 11, 2016 • VOLUME 291 • NUMBER 11 tration 5 M, 37°C, and pH 7.3). Samples were taken after different time intervals and analyzed for hemolytic activity as described above. Fig. 4 shows the recorded hemolysis profiles and the deduced kinetics of loss of hemolytic activity of ClyA ox and ClyA red , which declined within a factor of 1.6 with the same half-lives as the corresponding concentrations of M ox and M red (see Fig. 2E and Table 2). At 4°C, the half-lives of hemolytic activity loss of ClyA ox and ClyA red were about 50-fold longer compared with those recorded at 37°C (2.6 and 16.9 days, respectively ( Table 2)), showing that storage of ClyA ox at 4°C for 2 weeks leads to practically complete loss of activity. These results thus provide a plausible explanation for the fact that previous experiments had been interpreted such that ClyA ox is assembly-incompetent (13,17,20). ClyA ox Not Only Forms Annular Pore Complexes but Even Assembles Faster Than ClyA red -To identify potential differences in the assembly mechanisms of ClyA ox and ClyA red , we first compared the kinetics of the monomer-to-protomer transition and the kinetics of pore formation of ClyA ox and ClyA red in the presence of 0.1% DDM by far-UV CD spectroscopy. In the case of ClyA red , DDM triggered a rapid CD signal intensity increase, corresponding to the population of the off-pathway intermediate I red , followed by formation of the protomer P red with a more negative CD signal at 225 nm compared with M red , whereas the oligomerization of P red to pore complexes remained spectroscopically silent (14). We observed similar biphasic CD kinetics for ClyA ox at 225 nm (Fig. 5, A and B). However, the initial increase in the CD signal was less pro-nounced for ClyA ox (hinting at a lower transient population of the off-pathway intermediate), and the second, rate-limiting step of protomer formation proceeded 7.1 times faster (Fig. 5B). The CD spectra and CD kinetics (Fig. 5, A and B), together with the fact that oxidized and reduced monomers assembled to intact pore complexes in 0.1% DDM (Fig. 2C), thus provided a first hint that ClyA ox and ClyA red formed assembly-competent protomers via the same three-state reaction mechanism (I % M % P) (21) and that ClyA ox formed protomers faster than ClyA red .

Soluble ClyA Oligomers Are Pore Assembly Off-pathway Products
To get more quantitative information on the conformational states of ClyA ox populated during DDM-induced protomer formation, we next investigated the kinetics of P ox from M ox by single-molecule FRET experiments as described previously for ClyA red (21). To this end, we used a ClyA ox variant labeled with the FRET pair Alexa Fluor 488 and Alexa Fluor 594 at cysteine residues introduced at position 56 and 252, respectively, showing distinct conformation-specific transfer efficiencies (͗E͘) for M ox , I ox , and P ox (21) (Fig. 1). We used subnanomolar concentrations of ClyA that prevented protomer assembly, so that only the four unimolecular reactions in the scheme (I ox % M ox % P ox ) were observed. The time course of the transfer efficiency (͗E͘) histograms was fully consistent with the expected conversion of M ox (͗E͘ ϭ 0.42) to P ox (͗E͘ ϭ 0.67) with the transient population of the off-pathway intermediate I ox (͗E͘ ϭ 0.20) (Fig.  5C). Compared with ClyA red (21) (Fig. 5D, dotted lines), the formation of P ox was however about one magnitude faster than P red and virtually completed within 200 s (Fig. 5D), and the intermediate I ox was less populated. Compared with I red , shown to have molten globule-like characteristics (16,21), I ox showed an increased transfer efficiency (͗E͘ ϭ 0.20 versus 0.12) (21) (Fig. 5, C and E), indicating a more compact conformation of I ox in comparison with I red , likely due to the covalent linkage of helices B and G by the Cys-87-Cys-285 disulfide bond. Table 2 compares the microscopic rate constants obtained for the three-state mechanism of protomer formation of ClyA ox and ClyA red . The results are that the intermediate I ox is 0.5 kJ/mol less stable than M ox , whereas I red proved to be 4.4 kJ/mol more stable than M red (21). However, P ox was only 4.3 kJ/mol more  stable than M ox , whereas P red is 8.9 kJ/mol more stable than M red (21) (Fig. 5F). Overall, ClyA ox formed the protomer 7-8 times faster than ClyA red upon the addition of DDM (Fig. 5, B  and D), mainly because of a much lower population of the offpathway intermediate, I (maximum, 30% I ox after 10 s (Fig. 5D)) and, to a smaller extent, because of about a two times faster formation of P from M (Table 2). Finally, we also compared the kinetics of protomer association and pore assembly of ClyA ox and ClyA red by single-molecule FRET experiments at higher protein concentrations (10 nM), which allowed oligomerization to intact pores. This was achieved by adding an excess of the corresponding unlabeled ClyA redox form to donor/acceptor-labeled ClyA as described previously (21) (see also "Experimental Procedures" and the legend to Fig. 5E for the details). As observed for ClyA red (21), P ox was converted to a state with lower transfer efficiency (͗E͘ ϭ 0.53 versus 0.67 for P ox ) representing a mixture (O ox ) of annular pores and emerging, incomplete pores that cannot be distinguished spectroscopically (Fig. 5E) (21). These results clearly confirmed the assembly competence of P ox as demonstrated by electron microscopy (see Fig. 2C) and hemolysis experiments (Fig. 3). Moreover, the results showed that the assembly of P ox and P red occurred within comparable time frames (Fig. 5E). In summary, the single-molecule FRET measurements showed that the DDM-induced monomer-to-protomer transition of ClyA ox follows the same kinetic model as observed for ClyA red .
TheDisulfideofClyA ox IsNotSusceptibletoReductionIndependently of the Oligomerization State-The in vivo role of periplasmic ClyA ox has remained enigmatic, in particular because several groups have reported that ClyA ox has a lower intrinsic hemolytic activity than ClyA red (13,17,20), a finding that agrees with the observation that strains deficient in the periplasmic dithiol oxidase DsbA, which introduces disulfide bonds into periplasmic proteins, show ClyA-dependent hemolytic activity, whereas respective wild-type strains are hemolytically inactive (20). To investigate the possibility of a regulated change in the ClyA redox state in vivo, we tested the accessibility of the disulfide bond of ClyA ox for reduction by DTT. Fig. 6A shows that M ox , LO ox , and oxidized pore complexes were reduced only extremely slowly and remained more than 90% oxidized after 3 h even in 100 mM DTT. In contrast, unfolded M ox was completely reduced within 5 min by 20 mM DTT under the same conditions. The deduced rate constants of reduction (see legend to Fig. 6A) revealed that the tertiary structure of M ox , LO ox , and P ox in pore complexes protected the Cys-87-Cys-285 disulfide bond 10 5 -10 6 -fold against reduction relative to unfolded M ox . The reduction of ClyA ox by disulfide exchange at physiologically relevant rates would thus require unfolding, which is also consistent with the inaccessibility to solvent of the Cys-87/Cys-285 cysteine pair in the structures of the monomer and protomer (Fig. 6C). Finally, we also recorded the kinetics of oxidation of M red by DsbA. This reaction also proved to be extremely slow at physiological concentrations of DsbA ox (Fig.  6B). The results indicate that the formation of ClyA ox from folded ClyA red in the periplasm via disulfide exchange with DsbA would require the unfolding of M red . M for the indicated time intervals an diluted 500-fold with horse erythrocytes (final concentrations, 2 ϫ 10 6 cells/ml; 10 nM total monomer concentration), and the decrease in cell density was followed via the decrease in the optical density at 650 nm. C, kinetics of the decrease in hemolytic activity of ClyA ox and ClyA red at 37°C and pH 7.3. The maximum lysis velocity as a measure of hemolytic activity (see Fig. 3B) was plotted against incubation time. The decay in hemolytic activity yielded apparent rate constants of 0.53 Ϯ 0.03 h Ϫ1 for ClyA ox and 0.085 Ϯ 0.007 h Ϫ1 for ClyA red , respectively (solid lines).

Discussion
The Hemolytic Activity and Pore Formation Competence of ClyA ox -In the present study, we addressed two controversies in the literature on the assembly mechanism of ClyA. The first controversy is related to contradictory reports on the ability of ClyA ox to form pore complexes. Several reports have suggested that formation of the Cys-87-Cys-285 disulfide bond in ClyA prevents pore complex formation (13,17,20). A loss of assembly competence of ClyA ox was proposed based on the findings that ClyA in OMVs secreted by cytotoxic E. coli strains is present in the reduced form and that deletion of the periplasmic dithiol oxidase DsbA, which oxidizes ClyA in the periplasm, restores hemolytic activity (20). In addition, treatment of ClyA with excess Cu(phenanthroline) 2 , an established reagent for the oxidation of cysteine pairs in vitro, led to a loss of hemolytic activity that could be partially recovered by treatment with DTT, and a ClyA variant in which both cysteines were replaced by serine residues stayed active independently of the redox conditions (17). Based on these results, it was speculated that formation of the Cys-87-Cys-285 disulfide bond prevents the conformational transition to the assembly-competent protomers (17). All of these data were however in contrast to previous findings that ClyA ox forms active pores upon the addition of DDM in vitro and that both ClyA redox forms show comparable hemolytic activity (14).
We were able to resolve this controversy with our present study, in which we compared the mechanism of DDM-induced assembly of ClyA ox and ClyA red in detail in a quantitative manner. The key proved to be the important recent discovery by Fahie et al. (17) that ClyA shows a tendency toward spontaneous formation of larger oligomers in the absence of detergents, in particular at elevated temperatures. We thus compared ClyA ox and ClyA red for their ability to form oligomers sponta-neously. For this purpose, we first established the quantitative absence and presence of the Cys-87-Cys-285 disulfide bond in our ClyA preparations (Fig. 2B), and we then prepared ClyA ox FIGURE 5. Both oxidized and reduced ClyA monomers assemble via the same mechanism into functional pore complexes after the addition of the detergent DDM. A, far-UV CD spectra of ClyA red and ClyA ox monomers (solid lines) and after assembly to pores in 0.1% DDM at pH 7.3 (dashed lines), demonstrating the increased ␣-helix content of the protomers in the assembled pore complexes. The total monomer concentration in each sample was 9 M. B, reaction of M ox (gray symbols) and M red (red symbols) to assembly-competent protomers, recorded via the change in the far-UV CD signal at 225 nm. Reactions were initiated by the addition of DDM (manual mixing). The gray and red squares represent the CD signals of monomeric ClyA ox and ClyA red , respectively, in the absence of DDM. The rate-limiting step of monomer-to-protomer transition was fitted monoexponentially, resulting in apparent rates of protomer formation of 9.64 Ϯ 0.02 ϫ 10 Ϫ3 s Ϫ1 and 1.35 Ϯ 0.01 ϫ 10 Ϫ3 s Ϫ1 for ClyA ox and ClyA red , respectively. C and D, single-molecule FRET measurements of the kinetics of the monomer-to-protomer transition of ClyA ox labeled with Alexa Fluor 488 and 594. Protomer formation was initiated by the addition of 0.1% DDM, and the reactions were performed at low ClyA concentrations (300 pM), where oligomerization to pore complexes does not occur (21). CD signal intensities are shown as mean residue molar ellipticity (MRW). C, transfer efficiency histograms for the monomer-to-protomer transition of ClyA ox in DDM measured by microfluidics and manual mixing. Each line represents one histogram normalized to a total area of 1.  (21). E, dependence on ClyA subunit concentration of the kinetics of DDM-induced pore complex formation of ClyA red and ClyA ox recorded by single-molecule FRET experiments. Conditions were identical to those described in C and D, but reactions were performed at 10 nM ClyA subunit concentration allowing oligomerization of assembly-competent protomers and pore complex formation. The higher protein concentrations were achieved by mixing 300 pM FRET-labeled ClyA with excess unlabeled WT ClyA with the same redox state. Emerging and final pore complexes of ClyA ox (left panel) show identical transfer efficiencies of ϳ0.53 and appear as a single peak, (O ox ) n (orange) as observed for ClyA red (right panel). The transfer efficiencies of M ox , I ox , and P ox (0.42, 0.20, and 0.67, respectively) and of M red , I red , and P red (21) are also indicated. Transfer efficiency histograms were recorded after different reaction times, with color codes indicated in the bars at the top right of each panel. F, schemes of free energies (in kJ/mol) of M, I, and P of ClyA red and ClyA ox in 0.1% DDM at pH 7.3 and 22°C. The free energies of M ox and M red were assumed to be identical. I ox is 0.5 kJ/mol less stable than M ox , whereas I red is 4.4 kJ/mol more stable than M red . P ox is only 4.3 kJ/mol less stable than M ox , whereas P red is 8.9 kJ/mol more stable than M red . under mild oxidative conditions by using catalytic concentrations of Cu 2ϩ ions for Cu 2ϩ -catalyzed air oxidation. We fully confirmed the formation of two main oligomeric species for ClyA red reported previously (17). In addition, however, we showed that ClyA ox forms soluble oligomer 13-14 times faster than ClyA red , that oligomer formation is favored at elevated temperatures ( Table 2), and that all soluble oligomeric ClyA forms are hemolytically inactive (see below and Table 3). Thus, the formation of assembly-incompetent oligomers in preparations of ClyA ox provides a plausible explanation for the interpretations of previous studies that ClyA ox is hemolytically inactive. Specifically, we have shown that storage of ClyA ox at 4°C for 3 days already causes a loss of 50% of its hemolytic activity. Therefore, monomers of ClyA ox , freshly isolated from gel filtration chromatography, should be used immediately for assembly studies or kept frozen. We also suggest that ClyA red should be prepared and stored in the presence of reducing agents such as DTT to prevent air oxidation.
We analyzed the kinetic mechanism of DDM-induced pore formation for monomeric ClyA ox and ClyA red in detail, which clearly revealed that ClyA ox is not only fully assembly-competent and forms pores with even 20% higher specific hemolytic activity than pores of ClyA red (Fig. 3B), but also follows the same reaction mechanism in which the formation of assemblycompetent protomers is the rate-limiting step of pore formation at total monomer concentrations above ϳ100 nM. Protomer formation also is accompanied by the formation of a molten globule-like off-pathway intermediate. Notably, our single-molecule FRET experiments showed that the intermediate I ox is less stable than I red and therefore much less populated during protomer formation, so that assembly-competent protomers P ox even form about one order of magnitude faster than protomers of ClyA red (P red ) (Fig. 5). The formation of the Cys-87-Cys-285 disulfide bridge between helixes B and G thus accelerates protomer formation by decreasing the population of off-pathway intermediates during DDM-triggered pore formation.
ClyA in the periplasm of E. coli strains containing DsbA has been reported to be oxidized and to exhibit no cytotoxic activity (13,20). It remains to be established whether ClyA ox also assembles to inactive oligomers in the periplasm, analogous to its spontaneous assembly to oligomers in vitro. A recent study reported the co-localization of DsbA and ClyA in OMVs (31), where the reduced form is the main ClyA redox species. Our experiments show that the Cys-87/Cys-285 pair in folded ClyA red monomers cannot be oxidized efficiently by DsbA (Fig.  6B). This makes it unlikely that ClyA ox in OMVs needs to be activated by an unknown reducing agent, as proposed previously (20). An alternative explanation for the presence of ClyA red in OMVs could be that ClyA red is the predominant ClyA species in OMVs because it folds rapidly so that DsbA cannot access the Cys-87/Cys-285 pair and the ClyA molecules that stay in the periplasm are oxidized by molecular oxygen under aerobic conditions. Unraveling the translocation mechanism by which ClyA enters the periplasm in a Sec-independent manner appears to be essential for answering this open question.
Are Soluble ClyA Oligomers Really Intermediates of an Alternative Pore Formation Pathway?-Our previous work (21), together with the present data on the kinetic mechanism of DDM-induced pore formation of ClyA, clearly establishes a reaction pathway in which the unimolecular formation of protomers from monomers, with the transient population of a molten globule-like off-pathway intermediate, is the rate-limiting step of pore assembly at ClyA concentrations above ϳ100 nM. Here, we also examined the alternative ClyA assembly mechanism proposed by Fahie et al. (17), in which soluble ClyA oligomers, formed in the absence of detergent or membranes, represent prepores that can integrate into membranes and form hemolytically active pore complexes. We purified soluble oligomers of ClyA ox and ClyA red by gel filtration, but could detect neither a significant population of annular complexes by electron microscopy after addition of detergent (Fig. 2C) nor hemolytic activity above 1.4% relative to that of the reduced monomer after mixing the oligomers with erythrocytes (Fig.  3B). The interpretation of ClyA oligomers as prepore complexes by Fahie et al. (17) is based on findings that the oligomers showed hemolytic activity in end point hemolysis assays (see below) and single-channel conductance activity after reconstitution into planar lipid bilayers. So why did the work of Fahie et al. (17) and our present data lead to opposite conclusions on the role of soluble ClyA monomomers?
We believe that the main reason for the interpretation by Fahie et al. (17) is the fact that the dependence of hemolytic activity of ClyA oligomers on ClyA concentration was not investigated. Erythrocytes were incubated for 15 min at a total ClyA concentration of 0.8 M, and the amount of hemoglobin released after 15 min by ClyA oligomers and monomers was compared and used as a measure of hemolytic activity (17). Our Fig. 3, however, shows that the amount of lysed cells at a single time point is not a quantitative measure of hemolytic activity. Instead, the hemolytic activity of ClyA can be quantified accurately only via the dependence of maximum lysis velocity (or the inverse lag time of hemolysis (16)) on ClyA concentration (Fig. 2B). Therefore, determination of the specific hemolytic activity of ClyA requires the recording of hemolysis kinetics at different ClyA concentrations. Fig. 3A shows that the oligomers LO red , LO ox , and SO ox , already at concentrations of 10 nM, cause significant hemolysis after an incubation time of 1 h, for example, although their specific hemolytic activity relative to the monomer is only about 1%. In fact, ClyA monomers are highly active toxins, and the incubation of erythrocytes with only 10 nM ClyA monomers leads to complete lysis within 200 s (Fig. 3A). The experiments by Fahie et al. (17) were however performed at 80-fold higher ClyA concentrations, and ClyA activity was deduced from the amount of lysed cells after a single time point (15 min). Fig. 3 shows that we would have detected similarly high hemolytic activity in our preparation of purified ClyA oligomers if we had performed our experiments under the same conditions, i.e. at ClyA concentrations of 800 nM.
So what is the reason underlying the low apparent hemolytic activity of ClyA oligomers detected by us and Fahie et al.? A rough estimation based on the concentration dependence of ClyA activity (Fig. 3B) shows that even a minute fraction of about 0.1% of monomers spontaneously dissociating from the oligomers and reassembling into active pores via the classical assembly pathway can readily explain the hemolytic and singlechannel conductance activity of oligomers detected at the high ClyA oligomer concentrations used by Fahie et al. (17). Nevertheless, we cannot completely exclude the possibility that ClyA oligomers react directly to active pores after insertion into erythrocyte membranes, but we believe that this reaction is far too slow to be physiologically relevant.
Author Contributions-R. G., D. R., and B. S. designed the research. S. B. and D. R. prepared protein samples and performed the experiments. D. R., S. B., B. S., and R. G. analyzed and interpreted the data. D. R. and R. G. wrote the manuscript with the help of the other authors.