Monomer-Monomer Interactions Drive the Prepore to Pore Conversion of a β-Barrel-forming Cholesterol-dependent Cytolysin*

Perfringolysin O (PFO), a cholesterol-dependent cytolysin, forms large oligomeric pore complexes comprised of up to 50 PFO molecules. In the present studies a mutant of PFO (PFOY181A) has been identified that traps PFO in a multimeric prepore complex that cannot insert its transmembrane β-hairpins and therefore cannot form a pore. Remarkably, PFOY181A can be induced to insert its transmembrane β-hairpins if functional PFO is incorporated into the PFOY181A oligomeric prepore complex. Furthermore, the transition from prepore to pore appears to be an “all or none” process; partial insertion of the transmembrane β-barrel does not occur. Therefore, cooperative interactions between the monomers of the prepore drive the prepore to pore conversion that results in the formation of the transmembrane β-barrel.

Perfringolysin O (PFO), a cholesterol-dependent cytolysin, forms large oligomeric pore complexes comprised of up to 50 PFO molecules. In the present studies a mutant of PFO (PFO Y181A ) has been identified that traps PFO in a multimeric prepore complex that cannot insert its transmembrane ␤-hairpins and therefore cannot form a pore. Remarkably, PFO Y181A can be induced to insert its transmembrane ␤-hairpins if functional PFO is incorporated into the PFO Y181A oligomeric prepore complex. Furthermore, the transition from prepore to pore appears to be an "all or none" process; partial insertion of the transmembrane ␤-barrel does not occur. Therefore, cooperative interactions between the monomers of the prepore drive the prepore to pore conversion that results in the formation of the transmembrane ␤-barrel.
Many pore-forming bacterial toxins, outer membrane porins, and autotransporter proteins of Gram-negative bacteria, as well as the outer membrane porins of mitochondria and chloroplasts, utilize amphipathic ␤-sheets to span the lipid bilayer. Although the structure of the membrane-spanning ␤-sheet has been elucidated for a wide variety of membrane proteins, the mechanism of its assembly into the membrane remains largely unexplored. Amphipathic ␤-sheets are used to penetrate and form pores in the membrane by several toxins that have been designated ␤-PFTs or "␤-barrel pore-forming toxins" (1). Members of this family currently include Staphylococcus aureus ␣-hemolysin (2,3), Bacillus anthracis protective antigen (4,5), and Clostridium perfringens perfringolysin O (PFO) 1 (6,7). In contrast to the porins that form a membrane-spanning ␤-barrel from a single protein, these toxins assemble a membranespanning ␤-sheet by the oligomerization of individual toxin molecules. Each monomer contributes one or two amphipathic ␤-hairpins (dependent on the specific toxin) to the formation of the transmembrane ␤-sheet. Hence, the ␤-barrel of these toxins is assembled at the membrane from many protein monomers rather than from a single protein, as are the porins. The mechanism of pore formation by the ␤-PFTs encompasses the following three basic steps: targeting to the membrane surface, oligomerization into a prepore complex, and the conversion of the prepore to the inserted pore complex by the insertion of the transmembrane ␤-sheet. However, it is likely that additional intermediate states exist for this transition, and these states will be important in defining the mechanism by which the ␤-PFTs form a pore in the membrane.
The size of the oligomeric complex, and therefore the size of the pore, varies dramatically for the various types of poreforming toxins. Toxins such as aerolysin, ␣-hemolysin, and anthrax-protective antigen form heptameric oligomers and pores of 1-2 nm in diameter, whereas the cholesterol-dependent toxins (a family of ␤-PFTs whose cytolytic activity exhibits an absolute requirement for cholesterol), such as PFO, form oligomers of 40 -50 monomers and pores of 20 -30 nm (reviewed in Refs. 1 and 8). The formation of the prepore intermediate has been demonstrated for ␣-hemolysin (9, 10), PFO (11)(12)(13), and anthrax-protective antigen (14), although the function of the prepore intermediate has not been determined. We have recently shown (11)(12)(13) that the insertion of the transmembrane ␤-sheet of PFO, a cholesterol-dependent cytolysin (CDC), was significantly more rapid if the toxin monomers were first allowed to form the prepore complex. These observations suggested that formation of the prepore complex was a rate-limiting step and that the prepore complex is an obligatory intermediate in the insertion of the large transmembrane ␤-barrel formed by PFO.
In the present report we describe a novel mutant of PFO, PFO Y181A , 2 that is trapped in the prepore complex. Surprisingly, if PFO Y181A monomers form a mixed oligomer with functional PFO monomers, the insertion of the PFO Y181A ␤-hairpins occurs. Furthermore, these studies also reveal that insertion of the prepore appears to be an "all or none" process such that partial insertion of the prepore does not occur. These studies provide compelling evidence that monomers do not insert their transmembrane hairpins individually. Instead, cooperation between toxin monomers within the prepore complex is required for and drives the prepore to pore conversion, presumably in one concerted movement. * This work was supported by National Institutes of Health Grants AI37657 (to R. K. T.) and RRO7720 (to Z. S.) and the Robert A. Welch Foundation (to A. E. J.). 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.

EXPERIMENTAL PROCEDURES
Generation and Purification of PFO and Its Derivatives-All amino acid substitutions were generated via PCR overlap mutagenesis using the gene for the cysteine-less derivative of PFO, PFO C459A , 3 in which Cys-459 had been changed to an alanine (6). Expression and purification of PFO C459A and its derivatives used in this study and the determination of their hemolytic activity were performed as described previously (6). In all cases the mutants of PFO used in this study were generated from the gene for the cysteine-less, fully active, mutant PFO C459A . The functional PFO used in the various assays described below is PFO C459A .
Labeling Cysteine-substituted PFO with a Fluorescent Dye-In a typical labeling reaction, 2 mg of a single cysteine-substituted PFO derivative (1-2 mg/ml) was passed over a G-50 Sephadex column equilibrated in 50 mM HEPES (pH 8.0), 100 mM NaCl to remove excess dithiothreitol. The fractions containing toxin were combined, and solid guanidine hydrochloride was added, if necessary, to 3 M for labeling with 5-iodoacetamidofluorescein (IAF), N,NЈ-dimethyl-N-(iodoacetyl)-NЈ-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)ethylenediamine, or tetramethylrhodamine-6-maleimide (TMR). A 20-fold molar excess of fluorescent reagent was added to the toxin and allowed to incubate at 22°C for 2 h. After incubation, the mixture was passed over a second G-50 column in 50 mM HEPES (pH 7.5), 100 mM NaCl (buffer A) to separate the labeled protein from the free dye. The labeled fractions were pooled, made 10% (v/v) in glycerol, and quick-frozen in small aliquots in liquid nitrogen before storage at Ϫ80°C. All fluorescent probes were obtained from Molecular Probes (Eugene, OR). The extent of labeling was determined spectroscopically using the published extinction coefficient of the fluorescent probes (Molecular Probes) and PFO (6). In all cases labeling efficiency was 80 -100%, and the activity of the unlabeled and labeled toxin derivatives was not significantly different.
Liposome Production-Phospholipids and cholesterol were obtained from Avanti Polar Lipids (Alabaster, AL) and Steraloids (Wilton, NH), respectively. Liposomes were prepared by extrusion in a Liposofast extruder using 100 nm pore membranes (Avestin, Ottawa, Ontario, Canada) with a mixture of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) and cholesterol at a ratio of 45:55 mol % as described previously (6). For lipophilic quenching studies, the ratio of cholesterol to phospholipid was maintained at 55:45 mol %, except that a fraction of the POPC was replaced by 1-palmitoyl-2-stearoyl-(7-doxyl)-sn-glycero-3-phosphocholine (7-NO-PC) to yield cholesterol/POPC/7-NO-PC vesicles at 55:25:20 mol %. In all of the experiments requiring liposomes, the liposomes had been titrated with PFO as described previously (6) to ensure that an excess of liposomes was used to quantitatively bind and insert the available PFO.
Fluorescence Intensity Measurements-All fluorescence intensity measurements were performed using an SLM-8100 photon-counting spectrofluorimeter (SLM Instruments) as described in Shepard et al. (6). The excitation wavelength was 470 nm with a bandpass of 4 nm for all NBD measurements, and the emission intensity was scanned between 500 and 600 nm with an integration time of 1 s. Emission spectra were obtained at 37°C in 2 ml of buffer A containing 88 nM NBDlabeled toxin. In a typical experiment, excess liposomes were added to the stirred solution of toxin and allowed to incubate for 30 min at 37°C to ensure complete oligomerization and insertion of PFO into the membrane before the scans were repeated on the samples. Fluorescence quenching experiments were performed in the same manner using the 7-NO-PC-containing liposomes, in which 20 mol % of the PC was replaced with 7-NO-PC. In all cases the emission of a control sample in which the NBD-labeled toxin was replaced with unlabeled toxin was subtracted from the experimental data.
Fluorescence Resonance Energy Transfer (FRET) Measurements-To label specifically the various PFO derivatives with donor (fluorescein) or acceptor (rhodamine) fluorescent dyes, single-cysteine mutations were generated in PFO Y181A and PFO C459A at aspartate 30. Asp-30 is located at the amino terminus of mature PFO, and the substitution of cysteine for this residue does not affect the activity of PFO (data not shown). When irradiated at its excitation wavelength, the donor dye (D) can nonradiatively transfer its excited state energy to a second chromophore or dye (the acceptor, A). The efficiency of this transfer depends on, among other things, the extent of overlap of the emission of D and the absorption spectra of A, the relative orientation of D and A, and the distance between D and A. The transfer of excitation energy from D to A results in a decrease in D emission intensity.
The D30C derivatives of PFO Y181A and PFO C459A were labeled with either IAF or TMR. An equimolar mixture of donor-and acceptorlabeled PFO molecules (22 nM each; the DA sample) was stirred in buffer A containing 1 mM dithiothreitol at 37°C. To correct for light scattering and direct excitation of the acceptor, a sample was prepared in parallel in which unlabeled PFO(U)-replaced donor-labeled PFO to create the UA sample. Net donor emission was determined between 500 and 600 nm (excitation at 470 nm; bandpass at 4 nm) and subtracting the UA sample signal from the DA signal. Liposomes were then added, and the contents were mixed for 30 min at 37°C prior to reading the emission to allow the toxin to bind to and oligomerize on the liposomes. The net donor emission in the presence of liposomes was calculated as above and corrected for dilution before being compared with donor emission in the absence of liposomes (Fig. 4A). The same protocol was followed for an equimolar mixture of donor-and acceptor-labeled PFO Y181A and for a mixture of 22 nM donor-labeled PFO Y181A and 22 nM acceptor-labeled PFO C459A .
Fluorescence Lifetime () Measurements-Time-resolved fluorescence measurements were made using an ISS (Urbana, IL) K2-002 multifrequency cross-correlation phase and modulation spectrofluorimeter as described in Shepard et al. (6). Samples of 100 -200 nM NBD-labeled PFO Y181A derivatives containing 0 -400 nM functional PFO (PFO C459A ) were incubated with an excess of liposomes (the molar ratio of protein to total lipid was 1:1000) for 40 min at 37°C before the measurements were made. Phase and modulation data were analyzed using GLOBALS UNLIMITED, obtained from the University of Illinois (Urbana, IL), as before (6). In samples containing liposomes, the lowest 2 value was obtained with a three-component fit consisting of a single exponential with a of 0.001 ns to correct for scatter and two discrete exponentials for the NBD emission lifetimes. The fractional contribution of the scattering determined by this analysis corresponded closely to the fraction of total signal intensity that resulted from Raman and Rayleigh scattering. In the absence of liposomes, the scatter component was omitted during data analysis.
Pore Formation Assay-␤-Amylase from sweet potato and glutathione (Sigma) were labeled with fluorescein isothiocyanate (Molecular Probes, Eugene, OR) using the same procedures described in Heuck et al. (12) for carbonic anhydrase. Liposomes containing fluorescein-labeled ␤-amylase or glutathione were prepared as before for carbonic anhydrase (12). These liposomes (100 M total lipids) were suspended in buffer A containing an excess of anti-fluorescein antiserum that quenches fluorescein emission intensity by 80% upon binding to the dye. After thermal stabilization at 25°C for 5 min, the initial fluorescence was determined (F 0 ). The toxin (or toxin mixture) was then added at the indicated concentration to a final volume of 1.6 ml, and data collection was begun 15 s later, as described in Heuck et al. (12). Blank measurements were made using an otherwise identical sample that lacked the toxin, and the blank signal was subtracted from the corresponding sample signal.
Gel Electrophoresis-Denaturing agarose gel electrophoresis (SDS-AGE) was performed as described previously (11). Briefly, in all samples in which PFO was incubated with liposomes, the liposomes (cholesterol/POPC at 55:45 mol %) were incubated with PFO (8 g of toxin for Coomassie-stained and 1.5 g of TMR-labeled toxin for fluorescent gels) for 30 min at 37°C. Oligomeric complexes were solubilized with sample buffer containing 2.2% (w/v) SDS and then separated on 1.5% (w/v) SeaPlaque-agarose (FMC, Rockland, ME) in SDS gel reservoir buffer (15). The gel was typically run at 100 V for 70 -100 min. Gels were fixed (10% (v/v) acetic acid, 30% (v/v) methanol) overnight and then dried in a Hoefer gel dryer (San Francisco, CA). The dried gel was stained with Coomassie Brilliant Blue R and then destained to visualize the protein bands. TMR-labeled PFO or PFO Y181A were visualized on an UV transilluminator immediately after electrophoresis to detect the dye-labeled protein.
Atomic Force Microscopy-Mutant PFO was added to supported lipid bilayers composed of egg phosphatidylcholine/cholesterol (Avanti Polar Lipids, Alabaster, AL) at a 1:1 molar ratio, following a procedure described previously (16). The protein was injected to a final concentration of ϳ15 g/ml into a buffer of 5 mM dithiothreitol, 10 mM sodium phosphate (pH 7) covering the supported bilayer. After incubating for 45 min, the sample was extensively washed and then imaged with AFM under the same buffer. Imaging was performed in the contact mode with a Nanoscope II AFM (Digital Instruments, Santa Barbara, CA) using oxidesharpened "twin tip" Si 3 N 4 cantilevers with a spring constant of 0.06 N/m. The typical scan rate was 7 Hz, and the applied force was minimized to 0.1 nN. The outer diameter of the rings was determined from the center-tocenter distance between nearest neighbor complexes.

RESULTS
Oligomer Formation by PFO Y181A -Tyrosine 181 of PFO C459A was substituted with alanine to yield the mutant PFO Y181A . This derivative was found to exhibit less than 1% of the hemolytic activity of PFO C459A on human erythrocytes (data not shown). PFO C459A lacks the only native cysteine of PFO but exhibits the same activity as native PFO and is hereafter considered functional PFO (6). Tyrosine 181 is located immediately upstream of the first transmembrane ␤-hairpin of PFO ( Fig. 1 and see Ref. 6). The basis for the loss of activity was initially unknown, although a previous study with the highly homologous streptolysin O suggested that mutations in the analogous region of streptolysin O abrogated the ability of this toxin to form membrane-bound oligomers (17). Therefore, we initially examined the ability of PFO Y181A to form oligomers on liposomal membranes.
PFO forms large SDS-resistant oligomeric complexes on cholesterol-containing membranes that can be visualized by SDS-AGE and electron microscopy (11,13). The formation of oligomers by PFO Y181A on cholesterol-containing liposomes was investigated by SDS-AGE. As shown in Fig. 2A, PFO Y181A efficiently generated oligomers ( Fig. 2A, lane 8). However, in contrast to PFO C459A , these oligomeric complexes were not stable to SDS unless they were chemically cross-linked with glutaraldehyde prior to the addition of SDS (compare lanes 6 and 8). Because the PFO Y181A cross-linked oligomers migrated more slowly than both uncross-linked ( Fig. 2A, lane 2) and cross-linked PFO C459A oligomers ( Fig. 2A, lane 4), the PFO Y181A prepore oligomer is apparently either more structurally relaxed (i.e. less compact) than the inserted oligomer of PFO C459A and/or contains a few more monomers than the oligomer of the functional toxin. No oligomer was detected when monomers of PFO or PFO Y181A were incubated with an excess of cross-linker for 30 min in the absence of membranes ( Fig. 2A,  lanes 3 and 7, respectively). Therefore, as with functional PFO (i.e. PFO C459A ), the formation of PFO Y181A oligomers was dependent on the presence of membranes.
Visualization of the PFO Y181A oligomers on cholesterol-containing membranes by atomic force microscopy (AFM) (Fig. 2B) demonstrated that the oligomers formed by this mutant have an internal diameter of ϳ25 nm, similar to that of the oligomers formed by functional toxin (11,13). Therefore, PFO Y181A can bind and oligomerize on the membrane surface, but its inability to lyse erythrocytes suggests that the mutant gets trapped in a prepore state.
PFO Y181A Cannot Insert Its TMHs into the Bilayer-Conver-sion of the prepore complex to an inserted pore complex involves the insertion of two transmembrane ␤-hairpins (TMHs) per PFO molecule into the membrane (6,7,13). The interaction of these hairpins within the oligomeric complex of PFO generates a transmembrane ␤-barrel that is composed of up to 100 amphipathic ␤-hairpins (18).
To determine whether the transmembrane hairpins of PFO Y181A were inserted into the membrane, we individually analyzed the membrane exposure of residues in each PFO Y181A TMH. The residues responsible for the formation of the transmembrane ␤-sheet have been rigorously mapped for PFO by the use of multiple independent fluorescence techniques (6, 7). Therefore, selected residues that face the membrane in TMH1 (Ser-194 and Ala-215) and TMH2 (Lys-288 and Ile-303) were substituted with cysteine and then modified via their sulfhydryl groups with the NBD fluorescent dye. Because the fluorescence emission of NBD is strongly quenched by water, its emission intensity and lifetime increase as it moves into a nonpolar environment. NBD is therefore an excellent indicator for membrane insertion of the transmembrane ␤-sheet of PFO (6,7,(11)(12)(13).
The changes in the emission spectra of NBD dyes positioned at the above four locations are shown in Fig. 3 (left column) as functional PFO (PFO C459A ) moves from its water-soluble monomer state to a membrane-inserted pore complex. By comparison, the emission spectra (Fig. 3, right column) and lifetimes (Table I) of NBD located at these same positions within PFO Y181A did not change significantly as soluble PFO Y181A monomers were allowed to oligomerize on liposomes. Thus, the Y181A mutation allows the prepore oligomer to form, but it prevents the prepore complex from inserting its hairpins into the membrane and thereby prevents the formation of the pore.
Y181A Can Form Mixed Oligomers with Functional PFO-Because the PFO Y181A oligomer did not make the transition from a prepore to pore complex, it presented an excellent opportunity to determine directly if neighboring monomers within the prepore complex cooperate to help drive the conversion of the prepore complex to a fully inserted pore complex. If such monomer-monomer interactions are important in enabling this event, then mixing PFO C459A with the PFO Y181A mutant may induce PFO Y181A to insert its TMHs and lead to pore formation. Therefore, we initially determined whether PFO Y181A could form a mixed oligomer with PFO C459A using fluorescent resonance energy transfer (FRET). FRET is a nondestructive approach to demonstrate a close proximity between proteins and has been used previously by us to demonstrate the association of PFO monomers in the oligomeric complex (13,19,20).
As discussed in detail under "Experimental Procedures," the association of two proteins can be detected by labeling one with a fluorescein dye (the donor dye) that can transfer its excited state energy to a rhodamine dye (the acceptor dye) attached to the second protein. This transfer of energy, and therefore the loss of donor emission intensity, only occurs when the donor and acceptor dyes are very close (Ͻ100 Å), so no FRET would be observed at the concentration of PFO used in our experiments unless the proteins had associated with each other. As shown in the top panel of Fig. 4, the addition of cholesterol-containing liposomes to an equimolar mixture of donor-and acceptorlabeled PFO molecules results in a large decrease in donor emission. As we have shown previously (13,19), this reduction in donor intensity is due to FRET that occurs as the PFO monomers associate to form the oligomeric pore complex. A similar reduction in donor emission intensity occurs upon adding liposomes to donor-and acceptor-labeled PFO Y181A (Fig. 4,  middle panel). In addition to the cross-linked oligomer analysis and the AFM analysis (Fig. 2), the FRET provides spectral evidence that the PFO Y181A molecules are associating into oligomeric complexes on the membrane surface. Furthermore, PFO C459A and PFO Y181A form mixed oligomers when mixed in the presence of membranes (Fig. 4, bottom panel). The FRET observed in the lower panel shows that an equimolar mixture of PFO C459A and PFO Y181A forms complexes on the membrane that are spectroscopically indistinguishable from the complexes formed by homogeneous mixtures of PFO C459A or PFO Y181A .
Mixed Oligomers of Functional PFO and PFO Y181A Are Stable to Dissociation with SDS-The FRET analysis showed clearly that PFO C459A and PFO Y181A are capable of associating on the membrane to form mixed oligomers. However, the oligomers of PFO Y181A are completely dissociated by SDS unless covalently linked by a cross-linker, whereas functional PFO (PFO C459A ) forms SDS-resistant oligomers (Fig. 2). We therefore considered the consequence of forming the mixed oligomer on the stability of the oligomeric complex: could functional PFO C459A induce structural changes in PFO Y181A that might result in a SDS-stable oligomer or would the oligomer remain sensitive to SDS? To visualize the incorporation of PFO Y181A into mixed oligomers by SDS-AGE, PFO Y181A was labeled with TMR via the sulfhydryl group of a cysteine substituted for aspartate 30 in PFO Y181A (Fig. 5). As expected, in the presence of cholesterol-containing liposomes (11,13), functional PFO C459A (also labeled at D30C) with TMR (Fig. 5, lane 3) formed SDS-resistant oligomers. Also, as expected from the results in Fig. 2, homogeneous oligomers of labeled PFO Y181A were not resistant to SDS (Fig. 5, lane 4); only monomer and incomplete oligomers were observed. It should be noted that the partially dissociated oligomers of PFO Y181A are in comparatively small quantities as they are not observed for PFO Y181A in the Coomassie-stained gels in Fig. 2A. The intermediatesized oligomers for uncross-linked PFO Y181A are only observed when a sensitive detection system, as in the case of the fluorescently tagged proteins, is employed to visualize the oligomers.
When unlabeled functional PFO (PFO C459A ) was mixed with an equimolar amount of rhodamine-labeled PFO Y181A (Fig. 5,  lane 5), we observed that about 50% of the TMR-labeled PFO Y181A was incorporated into SDS-resistant oligomers. Increasing the ratio of PFO C459A to a 4-fold molar excess over PFO Y181A converted nearly all of the TMR-labeled PFO Y181A into an SDS-stable oligomer (Fig. 5, lane 6). Therefore, the functional PFO was clearly affecting the nature of the mixed oligomer such that this pore complex exhibited the characteristic resistance to dissociation by SDS that is observed for a homogeneous oligomer of PFO C459A .
The Insertion of the Transmembrane ␤-Hairpins of PFO Y181A in Mixed Oligomers-The results in Fig. 5 showed that as the proportion of functional PFO (PFO C459A ) in a mixed oligomer of PFO C459A and PFO Y181A increases, the fraction of the oligomer that exhibits resistance to dissociation by SDS also increases. Therefore, we suspected that the functional PFO in the mixed oligomers was forcing PFO Y181A to undergo structural transitions that allowed it to form a more stable oligomer and possibly to insert its transmembrane hairpins. To test the latter possibility directly, we examined the membrane insertion of the TMHs of PFO Y181A in the presence of various molar ratios of functional PFO.
Membrane-bound oligomers were formed by mixing various ratios of unlabeled PFO C459A with each of four derivatives of PFO Y181A that were labeled with NBD at positions S194C, A215C, K288C, or I303C and then incubating the toxin mixtures with excess liposomes. PFO Y181A TMH insertion was then determined by the magnitude of the increase in the emission intensity of the NBD probe at each of these locations. As shown in Fig. 6A, the emission intensity of the NBD increased significantly in all cases as the ratio of functional PFO (PFO C459A ) to PFO Y181A increased. These results demonstrated that the TMHs of PFO Y181A were inserting into the membrane in the presence of functional PFO, whereas, as shown in Fig. 3, PFO Y181A alone cannot insert its TMHs. Hence, as the concentration of functional PFO increased in the mixed oligomers, it forced the insertion of the TMHs of PFO Y181A .
Although we have previously demonstrated that these four residues, among many other residues in these transmembrane ␤-hairpins, face the membrane in functional PFO (6, 7), we confirmed the membrane location of these residues for PFO Y181A in the mixed oligomers by determining the accessibility of the NBD probe to collisional quenching by a nitroxide moiety attached to a phospholipid acyl chain (7-NO-PC) that is restricted to the hydrophobic core of the membrane (6,7,13). Mixed oligomers were again formed with each of the four NBDlabeled PFO Y181A derivatives using a 4-fold molar excess of unlabeled functional PFO (PFO C459A ) to stimulate maximal insertion of the PFO Y181A hairpins. The emission spectra were then taken before and after incubation of these mixtures with cholesterol/POPC (55:45 mol %) liposomes or with cholesterol/ POPC/7-NO-PC (55:25:20 mol %) liposomes (Fig. 6B). In all cases, the fluorescence intensity of the NBD probe is quenched more than 75% by the membrane-restricted nitroxide, a result that is only possible if the NBD is positioned within the nonpolar interior of the membrane.
The extent of PFO Y181A TMH insertion into the membrane was quantified by fluorescence lifetime analysis of the NBD emission when the probes were located at either position 215 in The distribution of the NBD probe lifetimes in samples with an NBD located at either position 215 (TMH1) or 303 (TMH2) of PFO Y181A was determined as described under "Experimental Procedures." In all cases, less than 4% of the signal was due to scattering. When liposomes were present in the sample, the lifetime data were best fit (i.e. the analyses have the lowest 2 ) to two discrete lifetimes for NBD, L and S , that represent NBD probes in a non-aqueous milieu with a long lifetime or an aqueous milieu with a short lifetime, respectively, and one scatter component ( fixed at 0.001 ns) (6,7). The "mol % buried" column shows estimates of the fraction of dyes in each sample that have lifetimes Ն9.0 ns and hence are buried in the nonpolar interior of the membrane. A corresponding to a residue buried in the membrane interior ( L Ͼ6 ns) was not observed in the absence of functional PFO (PFO C459A ). In those samples, more than 82% of the NBD probes are in an aqueous environment. The minor component of slightly longer lifetimes (Ͻ4 ns) represents probes exposed to less aqueous environments within the protein (7) TMH1 or position 303 in TMH2. In Table I we show that whenever the mol % of functional PFO reached or exceeded 50%, more than 40% of the NBD probes, and hence the TMHs of PFO Y181A , in the sample were inserted into the membrane, as indicated by their long lifetimes (Ͼ9 ns). At a 4:1 ratio of PFO C459A to PFO Y181A , more than 60% of the NBD probes in the samples make the transition to a long lifetime. These experiments demonstrate that significant fractions of both TMH1 and TMH2 of PFO Y181A are inserting into the membrane in the mixed oligomers. Therefore, functional PFO can evidently drive structural transitions in PFO Y181A that result in the insertion of its TMHs. Because the extent of conversion of these mixed oligomers to an inserted pore complex is dependent on the amount of functional PFO in the mixed oligomers, it suggests that only those oligomers with sufficient functional PFO can overcome the energetic barrier posed by the presence of the Y181A mutation and undergo the prepore to pore transition. Based on the lifetime data, ϳ40% of the prepore complexes do not have enough functional PFO to insert into the membrane under the conditions of the experiments. Does the Pore Size Change in the Mixed Oligomers?-Because the conversion of the mixed prepore oligomers to an inserted pore complex is dependent on the concentration of functional PFO in the mixture, we examined the nature of the pore formed by these mixed oligomers. Previous studies (11,13) have suggested that the prepore complex must attain an insertion-competent size to form a pore in the membrane. To determine whether the pores formed by mixed oligomers containing the PFO Y181A mutant were significantly reduced in size, we used liposomes that had encapsulated either fluorescein-labeled glutathione or fluorescein-labeled ␤-amylase. Initially we examined the release of these markers by a homogeneous mixture of functional PFO (PFO C459A ). As seen in Fig. 7A, little difference was observed in the release of the two markers demonstrating that the functional PFO formed a large pore that stimulated the release of both the large and small markers at similar rates. As expected, a homogeneous mixture of PFO Y181A did not induce the release of GSH (Fig. 7A) or ␤-amylase (not shown). We then compared the release kinetics for fluorescein-glutathione (ϳ600 Da) and fluorescein-␤-amylase (ϳ200,000 Da) from liposomes treated with mixtures of functional PFO and PFO Y181A containing 50 (Fig. 7B), 66 (Fig. 7C), or 80 mol % (Fig. 7D) of PFO Y181A . Very little difference in the release kinetics was observed for glutathione and ␤-amylase, even though the hydrodynamic radii of these two molecules differ greatly. Thus, each pore created by the toxin mixtures was large enough to pass glutathione and ␤-amylase with approximately equal efficiency. These results strongly indicate that the reduction in the rate of fluorescence-detected pore formation observed with mixed oligomers (Fig. 7) occurs be-cause of a reduced rate in the prepore to pore transition but not to a reduction in the average pore size. Therefore, as the ratio of PFO Y181A to functional PFO increases fewer complexes can make the transition to a functional pore suggesting that only those complexes that reach an insertion-competent size, and that contain sufficient functional PFO to overcome the inhibition posed by the presence of the PFO Y181A monomers, can make the transition. DISCUSSION The functional consequence of the Y181A mutation in PFO is that PFO Y181A oligomers cannot make the transition from the prepore to the pore complex. The Y181A mutation therefore raises the activation energy required for PFO insertion into the membrane. Yet the PFO Y181A monomers within the oligomer are able to make this transition if the prepore complex contains a significant number of functional PFO molecules. Thus, functional PFO molecules are able to induce the proper conformational changes in the nonfunctional PFO Y181A molecules such that they can insert their transmembrane ␤-hairpins. By monitoring emission intensities, fluorescence lifetimes, and accessibilities to membrane-restricted collisional quenchers of NBD probes positioned within the TMHs of PFO Y181A , we were able to demonstrate unequivocally that as the fraction of functional PFO increased in the mixed oligomers, an increasing fraction of the PFO Y181A TMHs moved into the membrane. These data show, for the first time, that interactions between the monomers of a CDC prepore complex are required to drive the conversion of the prepore complex to the inserted pore complex.
FRET, AFM, and SDS-AGE oligomer analyses showed that PFO Y181A is fully capable of forming homo-oligomers and that it forms oligomeric complexes similar in size to those of functional PFO. Furthermore, we have shown that PFO Y181A is capable of oligomerizing with functional PFO to form mixed oligomers. Whereas the oligomer of PFO Y181A was clearly dissociated by SDS, mixed oligomers with sufficient functional PFO are almost completely stable to SDS. However, it was necessary to use a 4-fold molar excess of functional PFO to induce the conversion of all of the mixed oligomers to an SDSstable form, suggesting that the barrier to stable oligomer formation posed by the Y181A mutation is significant. Because the PFO Y181A proteins form less stable prepore complexes, the monomer-monomer affinity in the mutant prepore complex is reduced relative to that in the wild-type PFO oligomers. The molecular basis of the reduced affinity is not yet known, but the fact that this tyrosine is conserved in all CDCs suggests that the aromatic ring may be important in stabilizing monomermonomer interactions and effecting the conformational changes necessary to accomplish the prepore to pore conversion. Consistent with this conclusion, PFO mutants with Tyr-181 replaced with Phe are significantly more active in pore formation than mutants with Tyr-181 replaced with Ile, Ala, Cys, and Thr (data not shown). 4 Because the prepore complex formed by PFO Y181A is unable to insert its TMHs into the membrane, PFO Y181A cannot undergo the proper conformational changes that are required for the alignment and insertion of the TMHs. One possible explanation for the observed behavior of PFO Y181A is that it is unable to align its TMHs with those of adjacent subunits in the prepore complex. The absence of hydrogen bonds between adjacent 4 E. M. Hotze and R. K. Tweten, unpublished data. TMHs would prevent a concerted movement of the TMHs into the bilayer and the formation of the pore. However, functional toxin monomers in an oligomer with the PFO Y181A induce the proper conformational changes in this mutant so that its transmembrane ␤-strands align and form intra-and inter-strand hydrogen bonds, and thereby allow the concerted insertion of the ␤-sheet to proceed.
In the homo-oligomer of PFO Y181A , the TMHs did not detectably insert into the membrane (Fig. 3; Table I), but successful insertion of the PFO Y181A hairpins occurred with increasing efficiency as the fraction of functional PFO in the mixed oligomers was increased. Thus, the insertion of the PFO Y181A TMHs could only result from the effects of the neighboring functional PFO monomers in the mixed prepore oligomer. This result further showed that PFO Y181A insertion could not be initiated with only a few functional monomers within the prepore oligomer (cf. a domino effect) but instead required a substantial fraction of functional PFO (Ն50%) to induce the required conformational changes in PFO Y181A necessary to overcome the misalignment and transition energy barrier imposed by the Y181A mutation ( Fig. 6; Table I). It is also important to note that the liposome release assays showed that the average size of the pore did not decrease when fewer functional PFO molecules were present in the mixed oligomers (Fig. 7). As the PFO:PFO Y181A ratio decreased fewer of the mixed oligomers contained a sufficient level of functional PFO to induce the prepore to pore transition. Thus, the mixed prepore complexes did not appear to partially insert into the membrane and form small pores that would have been reflected by a difference in the rate of release of the small (GSH) and large (␤-amylase) markers. Instead, the insertion of the prepore appeared to be an all or none process, i.e. the prepore to pore transition occurred only if the prepore was of a sufficient size and a sufficient number of functional molecules were present to overcome the barrier posed by the presence of PFO Y181A in the mixed oligomers.
The experiments reported here demonstrate directly that TMH insertion into the bilayer does not occur independently for PFO. Instead, TMH insertion is coupled with the insertion of its neighbors to create the ␤-barrel in the bilayer. Monomermonomer interactions therefore not only promote insertion but cooperative interactions between PFO monomers appear to be required to drive TMH insertion and ␤-barrel formation. The model of ␤-barrel biogenesis derived from this study of PFO may be a paradigm for the creation of other ␤-barrel pores in membranes. For example, the formation of the TOM complex in the outer mitochondrial membrane may involve a cooperative and concerted insertion of more than one Tom40p molecule (21,22). Experiments can now be done to ascertain the generality of cooperative TMH insertion in the biogenesis of ␤-barrel poreforming proteins.