Monomer-Monomer Interactions Propagate Structural Transitions Necessary for Pore Formation by the Cholesterol-dependent Cytolysins*

Background: The cholesterol-dependent cytolysins (CDCs) undergo a complex set of structural transitions to form the homo-oligomeric pore complex. Results: Structural transitions are propagated between monomers of the oligomeric complex. Conclusion: Specific structural changes establish the geometry of the oligomeric pore complex and promote the completion of existing oligomers. Significance: CDCs use membrane binding and ordered intermolecular interactions to drive assembly of their β-barrel pore. The assembly of the cholesterol-dependent cytolysin (CDC) oligomeric pore complex requires a complex choreography of secondary and tertiary structural changes in domain 3 (D3) of the CDC monomer structure. A point mutation was identified in the archetype CDC, perfringolysin O, that blocks detectable D3 structural changes and traps the membrane-bound monomers in an early and reversible stage of oligomer assembly. Using this and other mutants we show that specific D3 structural changes are propagated from one membrane-bound monomer to another. Propagation of these structural changes results in the exposure of a β-strand in D3 that allows it to pair and form edge-on interactions with a second β-strand of a free membrane-bound monomer. Pairing of these strands establishes the final geometry of the pore complex and is necessary to drive the formation of the β-barrel pore. These studies provide new insights into how structural information is propagated between membrane-bound monomers of a self-assembling system and the interactions that establish the geometry of the final pore complex.

The assembly of the cholesterol-dependent cytolysin (CDC) oligomeric pore complex requires a complex choreography of secondary and tertiary structural changes in domain 3 (D3) of the CDC monomer structure. A point mutation was identified in the archetype CDC, perfringolysin O, that blocks detectable D3 structural changes and traps the membrane-bound monomers in an early and reversible stage of oligomer assembly. Using this and other mutants we show that specific D3 structural changes are propagated from one membrane-bound monomer to another. Propagation of these structural changes results in the exposure of a ␤-strand in D3 that allows it to pair and form edge-on interactions with a second ␤-strand of a free membrane-bound monomer. Pairing of these strands establishes the final geometry of the pore complex and is necessary to drive the formation of the ␤-barrel pore. These studies provide new insights into how structural information is propagated between membrane-bound monomers of a self-assembling system and the interactions that establish the geometry of the final pore complex.
Cholesterol-dependent cytolysins (CDCs) 3 are secreted by a large number of Gram-positive bacterial pathogens and con-tribute to pathogenesis in a variety of ways. The soluble (CDC) monomers bind to their receptor (cholesterol or CD59) (for review, see Ref. 1) and assemble on cell membranes into large homo-oligomeric doughnut-shaped pore complexes. The CDC from Clostridium perfringens, perfringolysin O (PFO), has been used extensively as a model system to study the conformational changes that occur during the transition of the CDCs from soluble monomers to membrane-embedded pore complexes. The interaction of PFO monomers with cholesterol-rich membranes initiates a series of secondary and tertiary structural changes that result in the formation of a large ring-shaped ␤-barrel pore (for review, see Ref. 2) composed of 34 -36 monomers (3). Prior to the formation of the ␤-barrel pore, the only structures that contact the membrane are three short loops and the residues of the conserved undecapeptide located at the tip of domain 4 (D4) (4 -6) (Fig. 1). This interaction initiates the D3 structural changes that lead to oligomerization and pore formation.
After binding, membrane-bound monomers establish intermolecular contacts through the formation of backbone hydrogen bonds and -stacking interactions (Tyr-181 and Phe-318 in PFO) (7) between ␤-strands 1 (␤1) and 4 (␤4) of the fourstranded core ␤-sheet of D3 ( Fig. 1) (7). These interactions require that ␤5 rotates away from ␤4, thus freeing up the edge of ␤4 of one membrane-bound monomer to form edge-on interactions with ␤1 of another membrane-bound monomer. Also, two D3 ␣-helical bundles extend from the core ␤-sheet of D3 ( Fig. 1) and unravel to form the twin extended transmembrane ␤-hairpins (TMHs), which then contribute to the formation of the large ␤-barrel pore (8,9). To unfold and extend these ␣-helices into ␤-hairpins the interface between domains 3 and 2 is disrupted, the twist in the core ␤-sheet is relieved, and intermolecular hydrogen bonds and the -stacking interaction are formed between the remaining residues of ␤-strands 1 and 4 to create the transmembrane ␤-barrel (Fig. 1B).
These structural changes and the oligomerization of PFO monomers into the pore complex do not occur in the absence of membrane binding: oligomeric structures of PFO that resemble the oligomeric membrane complexes are not detected in solution, even at the high protein concentrations required for crystallization (10,11). Hence, membrane binding via D3 is necessary to initiate conformational changes in D3 that are required for oligomerization and pore formation. This mechanism is further supported by the observation that the membrane insertion of the undecapeptide tryptophan residues in D3 is conformationally coupled to the membrane insertion of the D3 TMHs (4). We have also observed a coupled rotation of D4 and D1-3 around the D2-D4 interface that disrupts some D2-D3 interac-tions in different crystal forms of PFO (12), which suggests a pathway for allosterically coupling membrane binding to the activation of the D3 structural transitions necessary to the formation of the oligomeric pore complex.
Although membrane binding is clearly the initiating event for pore formation, we have shown that monomer-monomer cooperativity is important in driving the assembly of the ␤-barrel pore (13). Here, we show that a point mutant in a conserved tryptophan residue traps PFO in an early stage of prepore formation that blocks the detectable D3 structural transitions and restricts the oligomer geometry to a linear structure rather than the typical circular structure. The study of this mutant suggests that prior to any detectable D3 structural changes the membrane-bound PFO monomers initially form weak, but specific intermolecular interactions. Our studies also suggest that these initial interactions synergistically propagate further D3 structural changes, which strengthen the interactions between monomers and drive the major structural changes in D3. These structural changes irrevocably commit the monomers to oligomerization and dictate the pore geometry, which leads to the formation of the oligomeric pore complex.

MATERIALS AND METHODS
Preparation of PFO Derivatives-The gene for PFO A459C , the cysteine-less PFO derivative in which the cysteine at residue 459 was replaced with alanine, was cloned into pTrcHisA (Invitrogen) as previously described (9). This plasmid (pRT20) was used as the template for all other amino acid residue substitutions. Various amino acid substitutions in PFO were generated by PCR QuikChange mutagenesis (Stratagene) and each derivative sequenced (Oklahoma Medical Research Foundation Core DNA sequencing facility). The expression and purification of recombinant His-tagged PFO and its derivatives from Escherichia coli were carried out as described (6). The toxin was stored in 5 mM dithiothreitol (DTT) and 10% (v/v) sterile glycerol at Ϫ80°C until used.
Hemolytic Activity-The hemolytic activity of each labeled or unlabeled PFO derivative was determined on human red blood cells (hRBCs) as described previously (14). The EC 50 is defined as the effective concentration of toxin required to lyse 50% of the hRBCs. The hemolytic activity of each mutant was compared with that of recombinant wild-type PFO.
SDS-Agarose Gel Electrophoresis (SDS-AGE)-SDS-AGE was carried out as described previously (9,14,15). Briefly, PFO (171 nM) was incubated in the presence or absence of POPC:cholesterol liposomes for 30 min at 37°C. Samples were solubilized with SDS sample buffer at 37°C for 2 min, and then the monomeric and oligomeric complexes were resolved on a 1.5% SDSagarose gel.  (11). D3 and the structure containing Trp-165 in domain 1 are shown in detail in the right panel. The D3 core ␤-sheet comprises the four ␤-strands that are contiguous with the twin ␣-helical bundles that ultimately unravel to form the twin TMHs that contribute to the formation the ␤-barrel pore. B, synopsis of the PFO pore-forming mechanism is shown. The SDS-sensitive prepore stage represents several different substates where different mutations trap PFO in an SDS-sensitive prepore structure (7,13), including the mutation of Trp-165 reported herein. SDS resistance of the prepore structure is achieved by the formation of the intermolecular backbone hydrogen bonds between ␤-strands 1 and 4 and the -stacking of interaction between Tyr-181 and Phe-318 of the core ␤-sheets of two monomers (7). As a point of reference, the locations of residues Asn-197 and Val-322 are also shown in the mechanism model in B when the two D3 ␣-helical bundles are unraveled to form the two membrane-spanning ␤-hairpins (8,9). D1-D4, domains 1-4. The PFO structure was generated using Visual Molecular Dynamics (29). The structural models of the prepore and pore monomers in B were derived from the cryo-electron microscopy derived structures of the pneumolysin prepore and pore fitted with the alpha-carbon backbone of the PFO crystal structure (30).

Modification of PFO with Fluorescent and Nonfluorescent
Probes-The labeling of PFO cysteine-containing mutants at the cysteine sulfhydryl with the environmentally sensitive dye N,NЈ-dimethyl-NЈ-(7-nitrobenz-2-oxa-1,3-diazol-4-yl) ethylenediamine (IANBD, Invitrogen) was carried out as described previously (7). The labeled protein was separated from free dye by gel filtration (Sephadex G-50; inner diameter 1.5 ϫ 25 cm) in HBS (50 mM HEPES, pH 7.5, 100 mM NaCl). For pyrene labeling, N-(1-pyrene)maleimide (Invitrogen) in dimethyl sulfoxide was added to a 4-fold molar excess over the PFO in HBS. After 2 h at room temperature DTT was added to 5 mM to stop the reaction. The mixture was passed through a Sephadex G-50 column equilibrated with HBS buffer to separate pyrene-labeled PFO from the free dye. All labeled PFO derivatives were made 10% (v/v) in glycerol and stored at Ϫ80°C. Prior to use the sample was spun at 21,000 ϫ g for 10 min to remove any precipitated protein and the protein concentration determined using the Bio-Rad protein assay according to the manufacturer's instructions.
Förster Resonance Energy Transfer (FRET) Analysis-FRET measurements were carried out as described previously (13) with the following changes. PFO W165T was labeled with sulfhydryl-specific maleimide derivatives of donor (D) (Alexa Fluor 488) or acceptor (A) (Alexa Fluor 568) fluorescent dyes at cysteine-substituted Asp-30. Asp-30 is located at the N terminus of mature PFO, and substitution of cysteine for this residue does not affect the activity of PFO (data not shown).
An equimolar mixture of donor (D)-and acceptor (A)-labeled PFO W165T (22 nM each; the DA sample) was stirred in HBS at 37°C. To correct for light scattering and direct excitation of the acceptor, a sample was prepared in parallel in which unlabeled PFO W165T (U) replaced the donor-labeled PFO to create the UA sample, therefore net DA ϭ DA Ϫ UA. Net D fluorescence in the absence of A (DU) was determined by DU-UU. Emission was determined between 500 and 600 nm (excitation at 470 nm; bandpass at 4 nm). Liposomes were then added to each sample, and the contents were mixed for 30 min at 37°C to allow PFO W165T to bind and oligomerize on the liposomes prior to reading the emission. To determine whether the association of the D-and A-labeled PFO W165T could be disrupted, a 4 molar excess of ␤4-␤5 disulfide-locked PFO (PFO ␤4␤5 ) (7) or prepore-locked PFO (PFO PPL ) (16) were added to the oligomerized PFO W165T in both the DA and DU samples, and after an additional 30 min the net D fluorescence was determined.
Electron Microscopy of PFO Oligomeric Complexes-To form the oligomeric complexes, PFO protein solutions (ϳ0.1 mg/ml) in buffer (20 mM HEPES, pH 7.0) were pipetted into Teflon wells as 13-l droplets and coated with 1 l of a 0.5 mg/ml lipid containing a 1:1 molar mixture of cholesterol and 1,2 dioleoylsn-glycero-3-phosphocholine. After incubation at room tem-perature for 30 min, the oligomeric PFO (or its derivatives) structures were transferred to carbon support films on electron microscopy (EM) grids and negatively stained with 1% (w/v) uranyl acetate and observed with a CM200 FEG microscope operating at 120 kV, under low dose conditions. Images were taken with a 2K ϫ 2K Teitz CCD camera.

RESULTS
Cytolytic Activity of Trp-165 Mutants-A cysteine scan of PFO revealed that substitution of Trp-165 resulted in a cytolytically inactive PFO derivative. Trp-165 is conserved in all known CDCs, suggesting that it is located at a structurally sensitive site in the CDC structure and likely requires an aromatic residue at this location. This was confirmed by a series of mutants containing either aromatic or non-aromatic side chains. As shown in Table 1 substitutions with aromatic residues retained between 19 and 100% of the cytolytic activity of wild-type PFO, whereas mutants that contained non-aromatic residues did not exhibit detectable cytolytic activity.
Membrane Oligomer Formation by the Trp-165 Mutants-Membrane-bound monomers of PFO oligomerize to form characteristic prepore and pore structures that are visualized by SDS-AGE (17). SDS-AGE analysis of the Trp-165 mutants showed that aromatic substitutions formed SDS-resistant oligomers similar to wild-type PFO. The non-aromatic substitution mutants, however, did not form detectable SDS-resistant oligomers as exemplified by the Cys and Thr substitutions (Fig. 2).
We have previously characterized oligomer-forming mutants, which are sensitive to dissociation with SDS but still form the typical doughnut-shaped rings of the CDC oligomers as visualized by EM (13,16). The mutants containing an aromatic residue for the native tryptophan exhibited the typical ring-shaped pore complexes (Fig. 3). The oligomers of both aromatic substitutions were slightly larger than native PFO by 1-2 monomers based on the size of the monomer crystal structure and the ring size demonstrated by EM. The Thr and Cys mutants, however, formed linear oligomers (Fig. 3) that were representative of other non-aromatic substitution mutants of Trp-165 (data not shown).
Membrane Binding by PFO W165T -The linear geometry and SDS-sensitive phenotype of the non-aromatic-substituted mutants suggest that they are defective in one or more stages of oligomer assembly, which prevents the formation of the typical circular complexes (17). To elucidate the structural basis of this defect each stage of assembly of the oligomeric pore complex in the PFO W165T mutant was evaluated. We first evaluated mem-

Relative hemolytic activity of Trp-165 mutants
The EC 50 (effective concentration for 50% lysis) was determined for wild-type PFO and the PFO T165 substitution mutants. Shown is the relative hemolytic activity of each mutant compared with WT toxin (% activity ϭ (WT EC 50 /mutant EC 50 ) ϫ 100). ND, hemolytic activity was not detectable in the range of concentrations using in the assay. All assays were performed in triplicate, and in all cases the S.D. was Ͻ5%.

PFO mutant
Hemolytic activity 19 W165A, C, E, I, K, L, M, N, S, T, V ND brane binding by monitoring the change in the intrinsic fluorescence of the native tryptophans of the undecapeptide at the base of D3 as they enter the nonpolar bilayer upon binding (4,5,18). Upon the addition of cholesterol-rich liposomes to either PFO or PFO W165T a similar increase in the intrinsic tryptophan fluorescence intensity was observed for both proteins (Fig. 4A), showing that native PFO and PFO W165T bound and inserted their undecapeptide tryptophans similarly into the membrane. Therefore, the W165T mutation did not affect the interaction of D4 with the membrane. Disengagement of PFO W165T Domains 2 and 3-Two major D3 structural transitions occur that lead to the formation of the ␤-barrel pore. One of these transitions is the disruption of the interface between D2 and D3 that is necessary to extend and transform the two D3 ␣-helical bundles into the twin TMHs, which then contribute to the formation of the ␤-barrel pore (8,9). Asn-197 is located in TMH1 and is buried in the nonpolar D2-D3 interface. Upon pore formation Asn-197 makes the transition to the polar environment of the pore lumen as the ␣-helical bundle unravels to form TMH1, which, along with TMH2, extends across the bilayer (9). Therefore the disruption of the D2-D3 interface is monitored by placement of the environmentally sensitive probe NBD on a cysteine substituted for Asn-197 and then measuring the change in emission intensity as it makes this transition. As expected, we observed a decrease in NBD emission intensity for PFO N197C-NBD as it moves from the nonpolar D2-D3 interface to the pore lumen (Fig. 4B). In contrast, little change in the environment of the NBD probe was seen for PFO W165T/N197C-NBD (Fig. 4B), indicating that Asn-197 remains buried in the D2-D3 interface.
Disengagement of ␤-Strand 5 from ␤-Strand 4-The other major D3 structural transition that occurs along with the disruption of the D2-D3 interface is the disengagement of the D3 ␤-strand 5 (␤5) from ␤-strand 4 (␤4) (7,13). This transition is necessary to free up the edge of ␤4 so that it is available to form the intermolecular backbone hydrogen bonds with ␤1 of another membrane-bound monomer. This interaction contributes to the formation of SDS-resistant oligomers. Val-322 is buried beneath ␤5 in the PFO monomer but is exposed to the aqueous environment when ␤5 rotates away from ␤4 (7). Therefore, an NBD probe positioned at this location also undergoes a transition from a nonpolar to polar environment, which is reflected by a decrease in its fluorescence intensity. The expected decrease in fluorescence intensity of the NBD probe in the functional PFO V322C-NBD was observed as ␤5 rotates away from ␤4, whereas only a small change in the environment of the probe was observed for PFO W165T/V322C-NBD (Fig. 4C). Therefore, ␤4 still remains buried under ␤5 in PFO W165T thereby preventing ␤4 from pairing with ␤1 of another membrane-bound monomer.
To confirm that the intermolecular pairing of ␤1-␤4 between monomers of PFO W165T was not taking place we measured excimer formation resulting from the -stacking of pyrene probes placed on cysteine-substituted residues Thr-179 in ␤1 of one monomer and Val-322 on ␤4 of a second monomer. If ␤1 and ␤4 are paired then these pyrenes will be oriented in such a way that will allow the stacking interaction to take place, which results in the formation of the excimer emission (7). As expected, the pyrenes in the membrane-bound PFO monomers formed a -stacking interaction, as evidenced by the appearance of a broad excimer emission centered at about 470 nm (Fig. 4D). No excimer emission was observed for the PFO W165T mutant, demonstrating that the pyrenes in ␤1 and ␤4 were unable to stack and hence that ␤1 and ␤4 in adjacent monomers were not properly juxtaposed in the oligomers formed by PFO W165T .
The combined data shown in Fig. 4 demonstrate that substitution of Trp-165 with threonine does not affect membrane binding of the monomer D4, but prevents both the detectable structural transitions within D3 and the intermolecular pairing of ␤1 and ␤4 between membrane-bound monomers.
Monomer-Monomer Interactions of PFO W165T Are Reversible-Native PFO monomers form membrane oligomers that are resistant to dissociation by SDS and cannot freely dissociate once they are incorporated into the native oligomer (16,17). Because the oligomers formed by PFO W165T are SDS-sensitive we determined whether they could freely exchange with monomers added after the PFO W165T linear oligomers were allowed to form. This was accomplished by determining whether FRET between the monomers of donor (D) and acceptor (A) fluorophore-labeled PFO W165T , which are incorporated into preformed linear oligomers, could be disrupted by the subsequent addition of various unlabeled PFO derivatives. If PFO W165T monomers were in equilibrium between the monomer and olig-   JULY 13, 2012 • VOLUME 287 • NUMBER 29 omer state on the membrane then the addition of unlabeled monomers would exchange with the D-and A-labeled monomers in the oligomer and disrupt the FRET-dependent A quenching of the D fluorescence.

Intermolecular Interactions Drive Assembly of a Pore Complex
As expected, we observed FRET-dependent donor quenching as D-and A-labeled PFO W165T interacted to form the linear oligomers (Fig. 5, compare DA with DU; U ϭ unlabeled PFO W165T ). We then determined whether PFO W165T monomers incorporated into the preformed oligomers could be exchanged by the addition of unlabeled mutants of PFO, which are trapped at various stages of pore assembly. These included unlabeled versions of PFO ␤4␤5 , which contains an engineered disulfide between ␤4 and ␤5 that prevents the interaction of ␤1 and ␤4 of adjacent monomers and is trapped at an early SDSsensitive stage of the prepore, (see the last section under "Results") (7); PFO PPL , which can oligomerize into an SDS-resistant prepore oligomer but cannot insert the ␤-barrel due to an engineered disulfide (16), and PFO W165T itself.  (4). B, disruption of the domain D2-D3 interface was detected by monitoring the aqueous exposure (and resulting decrease in the emission intensity) of the buried NBD, located at cysteine-substituted Asn-197, as D3 swings away from D2 (9). C, disengagement of ␤5 from ␤4 was detected by monitoring the decrease in the emission of NBD positioned at cysteine-substituted Val-322, which is buried beneath ␤5. As ␤4 and ␤5 separate, NBD moves from a nonpolar to polar environment, and its emission intensity decreases (7). D, intermolecular association of ␤1 and ␤4 from adjacent membrane-bound monomers is shown by the -stacking of pyrene dyes attached to Cys-substituted Thr-179 (located in ␤1) and Val-322 (located in ␤4) and the resultant broad excimer emission near 470 nm (7). Pyrene emission spectra are shown for an equimolar mixture of pyrene-derivatized PFO T179C and PFO V322C in wild-type (WT) and PFO W165T backgrounds in the absence (solid line) and presence (dashed line) of liposomes.
As shown in Fig. 5, A-C, the addition of unlabeled PFO ␤4␤5 , PFO PPL , or PFO W165T disrupted FRET between the D-and A-labeled PFO W165T . A 4-fold molar excess of each PFO derivative nearly eliminated FRET, indicating that the monomers in the PFO W165T oligomer were exchangeable after it formed the linear oligomers. The fact that a 4-fold molar excess of each unlabeled PFO derivative could nearly completely abolish FRET between D-and A-labeled PFO W165T indicates that the exchange with the unlabeled variants completely randomized their distribution within the linear oligomers. On average, inserting four monomers between each pair of D-and A-labeled PFO W165T would increase their separation by more than 100 Å, which would decrease the FRET efficiency to near zero. PFO ␤4␤5 and PFO PPL acted similarly to PFO W165T to disrupt FRET between the PFO W165T DA pair, suggesting that they formed interactions with PFO W165T similar to those it forms with itself. PFO PPL has been shown to form SDS-resistant oligomers; and when the same experiment was performed with the donor-and acceptor-labeled PFO PPL , the FRET could not be disrupted by the addition of unlabeled PFO PPL after the oligomers had formed (Fig. 5D). Therefore, unlike PFO W165T the monomers of the PFO PPL oligomeric complex cannot be exchanged once they interact to form the oligomer. As described below in the last section under "Results," the same is true for the interaction of the PFO ␤4␤5 monomers, which cannot form the intermolecular ␤1-␤4 interaction.
We also showed that this interaction was not the result of a weak, nonspecific interaction of the membrane-bound monomers, but that it was a specific interaction of the PFO monomers. The addition of a 4-fold molar excess of pneumolysin (PLY ␤4␤5 ), a CDC from Streptococcus pneumoniae that exhibits about 40% identity with PFO, did not disrupt of the FRET between the PFO W165T monomers (Fig. 5E).
These combined data show three important features of the PFO W165T monomer-monomer interaction: (i) PFO W165T appears to be trapped at an early stage of oligomerization that is reversible; (ii) the complementary interfaces mediating this interaction are intact in the various disulfide-locked mutants of PFO; and (iii) the interaction is specific to PFO and its derivatives.
Functional PFO Drives D3 Structural Transitions in PFO W165T -The above studies showed that disulfide-locked PFO derivatives could interact with PFO W165T . We therefore asked whether the interaction of PFO PPL with PFO W165T could drive the PFO W165T monomers to undergo the D3 structural transitions that were blocked by this mutation. When the disulfide of PFO PPL is in the oxidized state (PFO PPLox ) it still can form ␤1-␤4 intermolecular interactions but cannot insert its TMHs into the membrane until the disulfide is reduced (16). Therefore, we determined whether the interaction of PFO PPLox with PFO W165T could drive further the disengagement of ␤5 from ␤4 in PFO W165T (16). The disengagement of ␤5 from ␤4 was measured by monitoring a NBD probe located at V322C in ␤4 in PFO W165T oligomers as described in Fig. 4C. The rotation of ␤5 away from ␤4 exposures the NBD probe to the polar environment, and the emission of the probe is quenched by water.
The fluorescence emission of the NBD probe was determined for the PFO W165T/V322C-NBD monomers before and after the addition of liposomes to allow the formation of the linear oligomers (Fig. 6A, compare dotted and dashed lines). Following formation of the PFO W165T oligomers a 4-fold molar excess of unlabeled PFO PPLox was introduced and allowed to intercalate into the PFO W165T linear oligomers. The fluorescence emission of the NBD probe was again measured (Fig. 6A, solid  gray line). The disulfide in PFO PPLox was then reduced (PFO W165Tred ), and an emission scan was again performed (Fig.  6A, solid black line). Oxidized and reduced PFO PPL further drove the disengagement of ␤5 from ␤4 in PFO W165T as evidenced by the increased quenching of the NBD emission, signifying its transition to an increasingly more polar environment as ␤5 rotates away to open up the edge of ␤4 (Fig. 6A). The magnitude of the decrease in the NBD fluorescence emission after the reduction of the disulfide of PFO PPL was similar to that observed in PFO (Fig. 4C). A parallel experiment was performed in which additional unlabeled PFO W165T was added in place of the PFO PPL (Fig. 6B). No change in the fluorescence emission of the NBD occurs after the initial 30-min incubation period or after the addition of reducing agent. Hence, the increased polarity of the environment of the NBD in PFO W165T/V322C-NBD after the addition of the PFO PPL in its oxidized and reduced states resulted from the propagation of structural changes from PFO PPL to D3 of PFO W165T .
We then determined whether reduction of the PFO PPL disulfide also drove the membrane insertion of PFO W165T TMHs. To detect membrane insertion of PFO W165T TMHs we positioned the environmentally sensitive NBD probe at cysteinesubstituted Ala-215 in TMH1 of PFO W165T . Ala-215 is a membrane-facing residue in TMH1 that we have used extensively to monitor membrane insertion of the PFO TMHs (8,9,13,16,17,19). The reduction of the PFO PPLox disulfide allows it to complete the transition to the membrane pore by inserting its ␤-barrel pore into the membrane. In the mixed oligomers of PFO PPL and PFO W165T/A215C-NBD where the disulfide of PFO PPL was reduced it drove the membrane insertion of the PFO W165T TMHs, as shown by the increased emission intensity of the NBD probe in PFO W165T/A215C-NBD as it entered the membrane (Fig. 6C).
These results show that the oxidized and reduced forms of PFO PPL could drive the necessary D3 structural transitions and membrane insertion of the TMHs of PFO W165T . The interactions between functional PFO molecules and adjacent PFO W165T molecules therefore provided the energy and direction necessary to drive the D3 structural transitions required for pore formation. Because PFO W165T can be transformed from its inactive trapped conformation to the active state by the cooperative effect of adjacent functional PFO molecules, its inactive conformation appears to be an intermediate in the pathway of PFO structural changes that create a pore.
PFO ␤4␤5 Monomers Interact on the Membrane Surface-The rotation of ␤5 away from ␤4 is a critical step in the formation of SDS-resistant oligomers and is blocked in both the PFO W165T mutant and in the disulfide-locked PFO ␤4␤5 mutant (7). Previous studies of the PFO ␤4␤5 mutant suggested that it either remained in the monomer state or formed an SDS-sensitive oligomer (7). We observed that it could disrupt FRET between monomers of PFO W165T (Fig. 5A), which suggested that it also was forming an oligomer. Analysis of the PFO ␤4␤5ox -treated liposomal membranes by EM revealed the presence of short curvilinear arcs (Fig. 7A) that were converted to the typical ring-and arc-shaped structures when the engineered disulfide between ␤-strands 4 and 5 was reduced prior to their addition to the liposomes (Fig. 7B).
We then used FRET to determine whether monomers of PFO ␤4␤5 in these oligomeric membrane complexes were also in equilibrium between monomer and oligomer states on the membrane, similar to that observed for PFO W165T . As expected from the EM analysis, FRET-dependent quenching of the D emission is seen when the D-and A-labeled PFO ␤4␤5 are added to the membrane. However, no change in the quenching of the donor emission was observed when a 4 molar excess of unlabeled PFO PPL was subsequently added to the preformed PFO ␤4␤5 oligomers (Fig. 7C, dashed line). Unlike the interaction of PFO W165T monomers, the interaction of membranebound PFO ␤4␤5 monomers is not reversible, which indicates that they have formed additional contacts that prevent the reversal of the interaction of monomers within the SDS-sensitive PFO ␤4␤5 oligomer.

DISCUSSION
The coordination among membrane binding, monomermonomer interaction, and the D3 structural transitions in PFO that results in assembly of the oligomeric pore complex is a critical, though complex, process. The studies described here provide key insights into this process by showing that monomer-monomer interactions are obligatorily required to drive the D3 conformational changes necessary for the formation of the oligomer and ␤-barrel pore. Our studies suggest that prior to D3 structural changes the membrane-bound monomers initially form weak transient, but specific, interactions. Through normal thermal fluctuations a fraction of these interactions form stable interactions via the intermolecular pairing of ␤-strands 1 and 4, which are propagated to other monomers that interact with the growing oligomer. These interactions set the intermolecular geometry of monomer-monomer interaction that establishes the pore size and irreversibly commits the monomers to the assembly of the oligomeric prepore and pore complexes. Hence, the intermolecular interaction of membrane-bound monomers drives changes in the molecular structure of the CDCs that determine the pore size and leads to the completion of the oligomeric pore complex.
The intermolecular pairing of the D3 ␤-strands 1 and 4 between two membrane-bound monomers is a critical interaction that leads to the formation of the ring-shaped oligomeric complex of PFO. This interaction was lost in PFO W165T and is also blocked in the disulfide-locked mutant PFO ␤4␤5 , which contains an engineered disulfide that restricts rotation of ␤5 away from ␤4 unless the disulfide is reduced. Both PFO W165T and PFO ␤4␤5 form oligomers that are sensitive to SDS and exhibit an increased curvature of radius. In contrast, PFO PPL forms the intermolecular ␤1-␤4 interaction, and its oligomers exhibit architecture similar to that of the native pore complex (3,16). When the disulfide of PFO ␤4␤5 was reduced prior to addition to the membranes to allow the formation of the ␤1-␤4 intermolecular interaction, its oligomers exhibited an architecture that was indistinguishable from that of native PFO. Hence, the intermolecular ␤1-␤4 interaction locks membrane-bound monomers into their final intermolecular geometry and establishes the final size of the pore.
We also showed that in mixed oligomers that PFO PPLox could further drive the disruption of the ␤4 and ␤5 interaction in PFO W165T thereby showing that it propagated this change to the latter. The reduction of the engineered disulfide in PFO PPL further drove the disruption of the ␤4-␤5 interaction and the membrane insertion of the PFO W165T TMHs. These results show that D3 structural changes can be propagated from one monomer to another. Regulating the D3 structural transitions by monomer-monomer contact may be important to establish the proper environment for unfolding and assembling the twin TMHs from the D3 ␣-helical bundles (8,9). We have shown that if the D2-D3 interface is disrupted in soluble CDC monomers it causes the D3 ␣-helical bundles to unfurl prematurely and results in the rapid and irreversible loss of cytolytic activity and aggregation of the soluble monomers (20). Therefore, if membrane binding alone were sufficient to initiate the unfurling of the D3 ␣-helical bundles the membrane surface would be populated by individual monomers with their ␣-helical bundles unfolded to different extents, thereby making them susceptible to nonspecific, off-pathway interactions with other monomers and proteins on the membrane surface. Our studies suggest that monomer-monomer contact drives the D3 structural transitions that lead to transition of the ␣-helical bundles to the TMHs, which would minimize the residence time of the TMHs in the disordered state and therefore minimize possible offpathway interactions.
Our studies also suggest an explanation for the nonstochastic assembly of the PFO oligomer observed by Shepard et al. (17). They observed that at the earliest stages of assembly of the PFO oligomer the only detectable species were the monomer and the terminal prepore oligomer, which suggests that cooperative completion of oligomers was favored over the initiation of new oligomers. Based on the studies herein, we propose that the energy barrier to the D3 structural changes that leads to stabilizing intermolecular interactions (i.e. intermolecular pairing of ␤1 and ␤4) is met in only a fraction of the encounters between membrane-bound monomers (shown schematically in Fig. 8). Transient higher energy states of D3 due to normal thermal fluctuations would result in the partial disruption of the ␤4-␤5 hydrogen bonds. Although these states are transient they would be sampled many times by the interactions between membrane-bound monomers typified by those observed for PFO W165T . In a fraction of these encounters where the free edge of ␤4 is exposed by these thermal fluctuations it could pair with ␤1 of another monomer. Because ␤1 can form more potential backbone hydrogen bonds with ␤4 (eight) than ␤4 forms with ␤5 (four) the reversal of this interaction would be energetically unfavorable, thus stabilizing the dimer (see B in Fig. 8). The formation of this stable interface would induce structural FIGURE 7. PFO ␤4␤5 forms SDS-sensitive oligomers on membranes. A and B, EMs of membrane-bound PFO ␤4␤5 in its oxidized state (A) and reduced (pore-forming) state (B). Scale bars, 50 Å. The intermolecular interaction of monomers was determined by FRET between donor-and acceptor-labeled PFO ␤4␤5 . C, FRET-dependent donor quenching measured in its oxidized membrane-bound form. Compare the DA and DU samples. After PFO ␤4␤5 had been allowed to interact and donor quenching was at its maximum we added unlabeled PFO PPL (dashed line) to determine whether the interaction of the PFO ␤4␤5 was reversible, like that observed for PFO W165T (Fig.  5B). changes in the second monomer that would disrupt the ␤4-␤5 interaction at the free interface of the dimer, similar to what we observed when oxidized and reduced forms of PFO PPL interacted with PFO W165T . This would free up ␤4 on the open interface of the dimer (see C in Fig. 8), or, at least, increase the probability that ␤4 was in an open state and available to pair with ␤1 of another monomer (see D in Fig. 8). Therefore, unlike the formation of the stable dimer pair, the addition of monomers to the growing end of the oligomer would not be subject to the same energetic constraints facing dimer formation because the ␤4 of the dimer, or higher order oligomer, would exhibit a higher probability of being in the open conformation, which could then pair with ␤1 of a free membrane-bound monomer. Hence, the D3 structural changes that are initiated in the stable dimer would be propagated cooperatively to each monomer as they contacted the growing end of the oligomer, a process that would favor the extension of oligomers over the initiation of new oligomers.
It is important to note that the intermediate stages of oligomer assembly exhibited by PFO W165T and PFO ␤4␤5 would not normally be observed in native PFO. These two mutants trap PFO at what would otherwise be transient stages of the monomer-monomer interaction: in native PFO once two monomers establish a stable dimer by the intermolecular ␤1-␤4 pairing the intermolecular geometry between the monomers is also established. Hence, neither the long linear oligomers of PFO W165T nor the short curvilinear oligomers of PFO ␤4␤5 will be observed in native PFO due to their transient nature. We should also note that although we do not detect the major structural changes in D3 of PFO W165T that lead to the formation of the prepore and pore complexes, it is likely that membrane binding leads to some small structural changes in PFO W165T that facilitate the interaction of the monomers, interactions that are not observed in the soluble monomers.
The membrane attack complex/perforin (MACPF) protein family has recently been shown to exhibit a D3-like structure, and it has been proposed that they may exhibit features of the CDC pore-forming mechanism (21-26). They also contain a ␤4 analog, and some exhibit a ␤5 analog suggesting that some MACPF proteins may employ a mechanism similar to the CDCs to regulate assembly of their pore. Aspects of the assembly of the CDC oligomeric pore complex are also reminiscent of the mechanism suggested for the initiation of fibril formation in protein folding diseases such as Alzheimer disease. Protein folding diseases are typified by the conversion of soluble functional globular proteins into ␤-strand containing fibrillar aggregates (27,28). Chiti and Dobson (28) have suggested that the amyloidogenic state is accessible through the normal thermal fluctuations of the native proteins, which results in the formation of intermolecular edge-on backbone hydrogen bond interactions between ␤-strands and subsequent formation of ␤-sheet-rich fibrils. This mechanistic model is not unlike the mechanism of assembly for the CDCs suggested by the studies herein.
In summary, mutation of the conserved Trp-165 blocks all detectable D3 structural transitions and reveals that its membrane-bound monomers interact in a low affinity and reversible mode that does not allow adjacent monomers to adopt a functional conformation. The blockage was overcome when functional PFO derivatives in the hybrid oligomers induced PFO W165T to undergo the structural transitions that culminated in the rotation of ␤5 away from ␤4 and the formation and membrane insertion of the TMHs. Thus, early monomer-monomer interactions promote conformational changes that most likely align specific structures between monomers that increase the number of intermolecular interactions and thereby strengthening the association even more. These stabilizing structural changes are then propagated to each monomer as it encounters the growing end of the oligomer. This interaction triggers the conformational changes in D3 that establish the geometry of the ring complex and irreversibly commit the membrane-bound monomers to the formation of the oligomeric pore structure.