An intermolecular electrostatic interaction controls the prepore-to-pore transition in a cholesterol-dependent cytolysin

Significance

Bacterial pathogens produce pore-forming toxins that damage eukaryotic membranes, whereas the pore-forming immune defense proteins produced by vertebrates can damage bacterial membranes. Despite the opposite functions of these proteins in pathogenesis or protection, many use a common pore-forming mechanism whereby membrane-bound monomers oligomerize into a circular structure, termed the prepore, which then assembles a β-barrel structure that punches a hole in the membrane. Here we show that once the prepore is assembled, an intermolecular electrostatic interaction is established that drives the formation of the pore. This mechanism is likely to be used by toxins and other pore-forming proteins that span the biological domains of life.

Abstract

β-Barrel pore-forming toxins (βPFTs) form an obligatory oligomeric prepore intermediate before the formation of the β-barrel pore. The molecular components that control the critical prepore-to-pore transition remain unknown for βPFTs. Using the archetype βPFT perfringolysin O, we show that E183 of each monomer within the prepore complex forms an intermolecular electrostatic interaction with K336 of the adjacent monomer on completion of the prepore complex. The signal generated throughout the prepore complex by this interaction irrevocably commits it to the formation of the membrane-inserted giant β-barrel pore. This interaction supplies the free energy to overcome the energy barrier (determined here to be ∼19 kcal/mol) to the prepore-to-pore transition by the coordinated disruption of a critical interface within each monomer. These studies provide the first insight to our knowledge into the molecular mechanism that controls the prepore-to-pore transition for a βPFT.

The cholesterol-dependent cytolysins (CDCs) make up the largest class of bacterial β-barrel pore-forming toxins (βPFTs) and are present in nearly 50 Gram-positive opportunistic pathogens. A central paradigm of βPFTs is the formation of an obligatory intermediate termed the prepore (1–6). The prepore is a membrane-bound oligomerized ring-shaped complex in which the membrane-spanning β-barrel pore has not formed. Pore conversion, characterized by β-barrel insertion, only occurs after prepore completion. For all βPFTs, the molecular mechanism or mechanisms that control the transition from the prepore to the pore remains unknown.

The archetype CDC, perfringolysin O (PFO), coordinates the assembly of about 37 monomers into a prepore and then assembles and inserts into the membrane a β-barrel pore composed of 74 amphipathic hairpins (7–12). On prepore completion, a signal is generated within the oligomeric complex that triggers this transition. The nexus of the prepore-to-pore transition is the disruption of the interface that domain 3 (D3) forms with domains 1 and 2 (D1,2) (10) (Fig. 1). This interface is formed between D1,2 and one of the two D3 α-helical bundles. The D3 α-helical structures ultimately unfurl and refold into the two transmembrane β-hairpins (TMH1 and TMH2), which contribute to the formation of the giant β-barrel pore (Fig. 1) (11–13). Pore formation also requires that the prepore undergo a 40-Å vertical collapse that allows the TMHs to span the bilayer (7, 8, 14).

Fig. 1.

PFO structure and pore complex assembly. (A) The PFO ribbon structure (19) with magnified D3 (Right) with pertinent residues and structures identified. The D3 core β-sheet (β1–β4) is shown in cyan, and the β5α1 loop is shown in magenta. The α-helical bundles that refold into TMH1 and TMH2 are shown in yellow and purple, respectively (11, 12). The waters surrounding N197 are shown as red spheres. The membrane-binding interface (34–37) is circled. Structures were created and rendered using UCSF Chimera (38). (B) Schematic overview of the PFO prepore assembly and its conversion to the pore complex (39). The D3 counterparts (dashed circle) are colored the same as in A. Bound monomers oligomerize into the prepore structure, and then the D3 interface with domains 1 and 2 (D1,2) is disrupted and the α-helical bundles unfurl and refold into the TMHs to form the β-barrel pore.

Here we show that the prepore-to-pore transition is controlled by an intermolecular electrostatic interaction between E183 and K336 in PFO. Loss of either charged residue traps PFO in a stable prepore state, whereas pore formation can be restored by weakening the D3–D1,2 interface by a single mutation or by increasing the temperature. Molecular measurements with cross-linkers and molecular dynamics simulations suggest these two residues are positioned to form a strong electrostatic interaction or salt bridge. Our studies suggest the interactions that stabilize the D3–D1,2 interface are the primary barrier to the prepore-to-pore transition and further show that the E–K interaction provides the necessary free energy to overcome this transition state barrier.

Results

Conserved E–K pair Is Essential for PFO and Streptolysin O Activity.

A Cys scan of the PFO primary structure revealed that K336 and E183 mutation reduced pore-forming activity >10-fold. E183 is located in D3 at the end of β1, whereas K336 is located on the opposite side of D3 at the tip of a loop formed by β-strand 5 (β5) and α-helix 1 (α1) (Fig. 1A). An essentially identical arrangement is observed in the streptolysin O (SLO) crystal structure (15). Primary structure alignments of known CDCs show conservation of these two residues in about 50% of the CDCs, with all but one of these CDCs clustering into the same large phylogenetic clade (Fig. S1).

The contribution of these charged residues to PFO pore-forming activity was examined in several purified mutants (Table 1). In essence, these results show that uncharged substitutions for either residue eliminate or reduce pore-forming activity >99%, whereas conservative substitutions (Glu→Asp or Lys→Arg) or swapping the E–K pair positions were not well tolerated.

Table 1.

Features of the E and K mutants of PFO and SLO

Toxin	Percentage activity	SDS-stable oligomer
PFO	100	Yes
E183A	ND	Yes†
E183C	ND*	Yes
E183D	ND*	Yes
K336A	ND*	Yes
K336C	ND*	Yes
K336G	ND	Yes
K336R	19 ± 4	Yes
E183A/K336A	ND*	Yes
E183C/K336C	ND	Yes
E183K	ND	Yes
K336E	ND	Yes
E183K/K336E	ND*	Yes
SLO	100	Yes
E245A	ND	No
K407A	ND	No

The relative percentage pore-forming activity (and SD) for each mutant relative to PFO or SLO and formation of an SDS stable oligomer are shown. ND, pore formation not detected in the range of toxin concentration used. ND*, the EC₅₀ could not be determined, as 100% activity was not achieved at the highest toxin concentration used.

^†Partially resistant to SDS dissociation.

Partial activity of the electrostatic Cys mutants could be restored in either PFO^E183C or PFO^K336C by chemical modification with the positively and negatively charged sulfhydryl-specific reagents MTS-EA(+) (2-aminoethyl methanethiosulfonate) or MTS-ET(+) [2-(trimethylammonium)ethyl MTS] and MTS-ES(−) (2-sulfonatoethyl MTS) (Table S1), but only when each Cys mutant was modified with the cognate charge. More than 90% of the sulfhydryl groups were modified with these reagents, yet only ∼20% activity was restored, suggesting the correct positioning of the charges is necessary for efficient pore formation.

E183 and K336 Mutants Trap PFO in a Prepore State.

The late-stage prepore of PFO is characterized by its resistance to dissociation by SDS (without heating to 95 °C for 5 min) when separated by SDS-agarose gel electrophoresis (AGE) (5, 10), whereas the pore complex resists dissociation by SDS and heat. Single or double mutants of E183 and K336 were able to form SDS-resistant or, in the case of PFO^E183A, partially resistant prepore complexes (Table 1) but were not resistant to SDS and heat.

Chemical modification of PFO^E183C and PFO^K336C with the appropriately charged MTS reagents (described in Table S2) was equally effective at restoring partial activity, whether they were added before or after prepore formation (Fig. 2). Hence, neither E183 nor K336 are necessary for prepore formation, but they are necessary for the prepore-to-pore transition.

Fig. 2.

Chemical modification of PFO^E183C and PFO^K336C with sulfhydryl-specific charged reagents trigger pore conversion. PFO^E183C and PFO^K336C (1 μM) were modified with charged MTS-ES(−) or MTS-ET(+), respectively, before (solid lines) or after (dashed lines) prepore formation on CF liposomes. CF release was continuously monitored over time. Arrows indicate addition of the MTS reagents after prepore formation. Graphs are representative of three independent experiments.

E183 and K336 Undergo a Polar-to-Nonpolar Transition on PFO Prepore Formation.

If E183 and K336 formed an intermolecular electrostatic interaction at the interface between the monomers in the prepore, then the side-chains of both residues would be predicted to transition from a polar to a less polar environment at this interface. A sulfhydryl-specific environmentally sensitive probe N,N′-dimethyl-N-(iodoacetly)-N′-(7-nitrobenz-2-oxa-1,3-diazol-4-yl) ethylenediamine [IANBD (12)], the emission of which is quenched by water, was coupled to the Cys sulfhydryl of PFO^E183C and PFO^K336C, and the changes in its fluorescence emission were monitored as these proteins move from the soluble monomer to the prepore complex (12). The probe entered a less polar environment in both cases during prepore assembly, as evidenced by a blue-shifted and increased fluorescence emission of the probe that likely results from the exclusion of water at the interface (Fig. S2).

Distance Measurements Suggest the Charged Centroids of E183 and K336 Are Juxtaposed in the PFO Prepore Complex.

The double Cys mutant PFO^E183C:K336C was subjected to cross-linking analysis with various length homobifunctional cross-linkers to estimate the distance between the α-carbons of the Cys residues. This measurement was then compared with the estimated distance separating the α-carbons of an E–K pair that forms a strong electrostatic interaction or salt bridge (Fig. 3B). The PFO^E183C:K336C prepore is dissociated by the combination of SDS and heat; hence, the Cys sulfhydryls alone are not sufficiently close to form disulfides and stabilize its prepore [Fig. 3A, the lane without added bismaleimidoethane (BMOE) cross-linker].

Fig. 3.

Distance separating the α-carbons of E183 and K336 before and after β5 disengages from β4. Prepore complexes of PFO^E183C:K336C were treated with homobifunctional, sulfhydryl-specific cross-linkers (40). (A) SDS-AGE panel 1: lanes 1–3, PFO or its pore complex treated as shown (SDS+Heat, 95 °C for 5 min). Panels 2–4: PFO^E183C:K336C prepore treated with a 0–5 M excess of BMOE, BMB, and DBB, and then heated (95 °C for 5 min) in SDS sample buffer before SDS-AGE. (B, Upper and Lower, respectively) Approximate distances separating the ɑ-carbons of Cys residues cross-linked with BMOE or of an E–K pair that form an electrostatic interaction or salt bridge. (C) PFO^{E183C:K336C:Y181A} was incubated with BMOE and BMB as in A. Images are representative of three independent experiments. L, liposomes; M, monomer; O, oligomer; X-L, molar excess of cross-linker to protein.

Highly efficient (85–98%) stabilization of the PFO^E183C:K336C prepore complex was achieved at a 1:1 and 5:1, but not a 1:4, molar ratio of the BMOE cross-linker to protein (Fig. 3A). The longer [bismaleimidobutane (BMB)] and the shorter [dibromobimane (DBB)] cross-linkers were less efficient and did not stabilize the prepore (Fig. 3A), although several intermediate-sized oligomers were evident. These results suggest that the Cys α-carbons are ∼14 Å apart in the prepore, similar to the estimated distance separating the α-carbons of a E–K pair that position their centroids of charge sufficiently close (≤4 Å) to form a strong electrostatic interaction or salt bridge (16, 17), as schematically shown in Fig. 3B.

Disengagement of β5 from β4 Is Necessary to Appose E183 and K336.

Restricting the movement of β5 away from β4 with an engineered disulfide results in an SDS-sensitive early-stage prepore (18). Mutation of Y181 to Ala also results in an SDS-sensitive prepore complex, suggesting the disengagement of β5 from β4 is also restricted in this mutant, which we indeed determined to be the case (Fig. S3). Thus, this mutation was introduced to the PFO^E183C:K336C background (rather than introducing additional cysteines for a disulfide bridge) to determine whether β5 must disengage from β4 to be repositioned sufficiently near E183 to be cross-linked. Neither BMOE nor the longer cross-linker BMB stabilized the prepore of PFO^E183C:K336C with Y181A mutation (Fig. 3C). Hence, restricting the disengagement of β5 from β4 prevents K336 from being positioned near E183 in the prepore complex.

Altering the D3–D1,2 Interface Stability Restores Pore Formation in PFO^E183C and PFO^K336C.

In the CDCs that maintain the E–K pair, an Asn is conserved at the D3–D1,2 interface (Fig. S1). PFO N197 occupies a hydrophilic pocket at this interface where its side chain forms hydrogen bonds with Ser-93 (water mediated) and Asp-91 and is surrounded by a network of hydrogen-bonded water molecules, which also contribute to the interface stability (19) (Fig. 1A). Therefore, could the pore-forming activity be restored in PFO mutants that lacked either E183 or K336 by replacing N197 with a residue that would disrupt these stabilizing interactions?

Tryptophan was substituted for N197, as modeling of this residue in the PFO structure suggested that its hydrophobic side chain would exhibit difficulty packing into the interface, be incapable of forming the N197 interactions, and disrupt the water-mediated interactions. This mutation restored pore-forming activity in both PFO^E183C and PFO^K336C to near PFO levels and increased the rate of pore formation over that of PFO (Fig. 4A). When Cys was substituted for N197, it did not restore activity in the cognate alanine mutants for E183 and K336. However, chemical modification of the Cys sulfhydryl with N-ethylmaleimide (NEM) after the assembly of the prepore triggered rapid and quantitative conversion to the pore (Fig. 4B).

Fig. 4.

Mutations for N197 restore pore-forming activity in PFO^E183C and PFO^K336C. (A) Trp was substituted for N197 in PFO^E183C and PFO^K336C, and then the rate and extent of pore formation on CF liposomes over time was compared with PFO. (B) PFO^E183A:N197C and PFO^K336A:N197C were incubated with CF-liposomes at RT to assemble the prepore complex while the fluorescence emission was continually recorded (20 min). NEM (20 molar excess over protein) was then injected (arrow) into the stirred cuvette, and CF release was monitored an additional 20 min. Results are representative of three independent experiments.

N197 Mutations Decrease the Melting Temperatures of PFO and Its Derivatives.

The introduction of Trp likely destabilized the D3–D1,2 interface, which allowed the prepore-to-pore transition. Consistent with this premise, the introduction of Trp at the D3–D1,2 interface decreased the melting temperatures (T_m) by 3.5 °C and 4.1 °C for PFO^E183C and PFO^K336C, respectively (Table 2). The N197C mutation also decreased the T_m of PFO^E183A and PFO^K336A by 4.3 °C and 2.8 °C, respectively, but it did not restore activity (Fig. 4). Modification with NEM decreased the T_m further and also restored activity. The stability of the D3–D1,2 interface is apparently the primary determinant of the T_m, as an engineered disulfide between D3 and D2 (PFO^PP), which prevents the α-helical bundle of D3 from disengaging from this interface and traps PFO in a prepore (10), increased the T_m of oxidized PFO^PP >4 °C over PFO, but decreased about 2 °C less than PFO when this disulfide was reduced (Table 2).

Table 2.

T_m of PFO and derivatives

Toxin	Tm (°C)	∆Tm (°C)
PFO	49.0 ± 0.2	—
PFO + NEM	48.7 ± 0.5	−0.3
E183C	47.6 ± 0.3	−1.4
E183C:N197W	44.1 ± 0.5	−4.9 (−3.5)
E183A:N197C	44.7 ± 0.5	−4.3
E183A:N197CNEM	43.9 ± 0.4	−5.1 (−0.8)
K336C	47.6 ± 0.3	−1.4
K336C:N197W	43.5 ± 0.3	−5.5 (−4.1)
K336A:N197C	46.2 ± 0.2	−2.8
K336A:N197CNEM	44.7 ± 0.3	−4.3 (−1.5)
PFOPP (Ox)	53.2 ± 0.4	+4.2 (+6.0)*
PFOPP (Red)	47.2 ± 0.2	−1.8

The T_m values for PFO, PFO^E183C, and PFO^K336C with N197W or NEM-labeled N197C, and PFO^PP under oxidizing (Ox) or reducing (Red) conditions. Experiments were performed in triplicate. ∆T_m, difference in T_m between PFO and mutant. Parenthesis indicates ∆T_m between indicated derivative and the cognate nonsubstituted or unlabeled counterpart.

^*∆T_m between the Ox and Red forms of PFO^PP itself.

The Arrhenius Activation Energy for the Prepore-to-Pore Transition.

The T_m results showed that the D3–D1,2 interface stability is the primary determinant of the PFO T_m and suggested that elevating the temperature might drive prepore-to-pore conversion when the electrostatic interaction is absent. The PFO^E183C pore-forming activity progressively increased as the temperature was increased (Fig. 5A), and it achieved similar activity to that exhibited by the PFO^E183C:N197W mutant and PFO (Fig. 4A) when the temperature exceeded its T_m.

Fig. 5.

Arrhenius E_a of the prepore-to-pore transition. (A) Pore-forming activity of PFO^E183C on CF-liposomes with increasing temperature. (B) The initial velocities of the CF release from A were used to calculate the E_a of pore formation by PFO^E183C. (C) The E_a of PFO^E183C pore formation was determined as in B, but only after it had first formed the prepore at RT for 20 min on CF liposomes, which eliminates the contribution of binding and oligomerization to the E_a. (D) The E_a for PFO was determined as in B, with PFO added directly to liposomes held at the indicated temperatures. The data are representative of two experiments.

These results allowed us to determine the activation energy (E_a) of the prepore-to-pore transition, as the E_a for binding and oligomerization (20) could be eliminated by first allowing PFO^E183C to assemble into the prepore at room temperature. Using the initial rates (k₀) of pore formation at 30–50 °C, the E_a was calculated to be ∼40 kcal/mol for pore formation when the PFO^E183C prepore was not first formed on the liposomes, which reflects the total E_a required to bind, oligomerize, and form the pore (Fig. 5B). When the PFO^E183C prepore was first formed at room temperature before determining the k₀ at each temperature, the E_a for the prepore-to-pore transition was calculated to be ∼19 kcal/mol (Fig. 5C). Hence, the E_a for binding and oligomerization is ∼21 kcal/mol (i.e., 40–19 kcal/mol), which is similar to the E_a of about 19 kcal/mol determined previously for the binding and oligomerization of PFO on human erythrocytes (20).

The E_a for PFO was determined to be 24 kcal/mol (Fig. 5D). This value is consistent with the E_a of 22–23 kcal/mol for pore formation on erythrocytes by PFO (20, 21). Because the E–K pair remains intact in PFO, it provides the necessary free energy to disrupt the D3–D1,2 interface, so this component is largely absent from the E_a for PFO.

Molecular Dynamic Simulations of the E183 and K336 Interaction.

Molecular dynamic (MD) simulations of the PFO pore support the formation of an intermolecular salt bridge between the E183 and K336 residues of adjacent monomers. CDC monomers assume an upright position perpendicular to the membrane (Fig. 1) and then undergo a 40-Å collapse to insert the β-barrel pore (Fig. 1B) (7, 8, 14). These experimental data are consistent with a substantial tilt of the monomers with respect to the membrane surface (22). At the initiation of the MD simulation, when the prepore intermediate is not yet tilted, the charge centroids of E183 and K336 in adjacent monomers are separated by ∼16 Å, and the α-carbons by ∼19 Å (Fig. 6A). As the MD simulation proceeds, monomers within the prepore oligomer undergo a 30° tilt, K336 moves closer to E183 (Fig. 6B), and finally, the charge centroids come within ∼2 Å of each other and spontaneously form an intermolecular salt bridge (Fig. 6C). In addition, as the two residues move closer together, they concomitantly draw β-strands 1 and 4 closer together so they can form backbone hydrogen bonds.

Fig. 6.

MD simulations of the E183 and K336 interaction in the prepore complex. (A–C) Three images of two adjacent monomers are shown from simulations of the PFO prepore-to-pore transition depicting the distance between the E–K pair (shown as sticks inside the circles) as the monomers tilt within the prepore complex before spontaneously forming a salt bridge. (C) Includes a magnification of the E–K pair forming a salt bridge.

Discussion

The displacement of the β5α1 loop from its interaction with β4 (Fig. 1) is necessary for stable prepore formation (18, 23). We now show that after prepore assembly, the β5α1 loop plays a second critical role: it is repositioned so that K336 is placed sufficiently near E183 to form a strong intermolecular electrostatic interaction or salt bridge. This interaction then drives the disruption of the D3–D1,2 interface, thereby committing the prepore to undergo the transition to the pore complex. The loss of pore-forming activity with conservative substitutions or by swapping the charged pair positions suggests correct positioning of the charge centroids is critical for this interaction. It appears that this control system is also present in SLO, as knocking out either residue of its homologous E–K pair eliminates its pore-forming activity (Table 1). These studies reveal the first molecular mechanism to our knowledge by which a βPFT controls the critical prepore-to-pore transition.

Measurements using homobifunctional sulfhydryl-specific cross-linkers estimated the distance between the charged centroids for the E–K pair to be 2–4 Å (corresponding with 12–14 Å between their α-carbons) in the prepore complex, consistent with the formation of a strong intermolecular electrostatic interaction or a salt bridge (17, 24). The distance between these two residues increased when the movement of the β5α1 loop, which contains K336 at its tip (Fig. 1A), was restricted (22). BMOE cross-linking did not restore pore formation, which suggests the distance between the charged centroids of E183 and K397 is necessarily reduced to less than 4 Å (perhaps forming a salt bridge, as suggested by the MD simulation) to disrupt the D3–D1,2 interface and trigger the prepore-to-pore transition.

The loss of pore-forming activity was suppressed by mutation of the N197 to Trp, NEM modification of Cys-substituted N197, or increasing the temperature. The decreased T_m for the various mutants were consistent with this scenario and with the fact that we could restore pore-forming activity in the prepore trapped mutants by increasing the temperature to approximately the T_m of PFO. Interestingly, Cys substitution for N197 in the Ala-substituted mutants for E183 and K336 did not restore activity, even though the T_m of PFO^E183A:N197C was decreased about 4 °C. In both cases, the chemical modification of its sulfhydryl group with NEM triggered rapid and quantitative conversion to the pore. The small difference (<1 °C) in T_m of the unmodified (inactive) and NEM-modified PFO^E183A/N197C (active) suggests the stability of this interface is finely balanced between the prepore and pore states.

The soluble PFO monomer, the prepore, and the pore complex are intermediate structures that represent local minima in the energy landscape of pore formation. The first major transition state barrier for PFO-mediated pore formation is the assembly of the oligomeric prepore; the second major transition state barrier is the prepore-to-pore conversion. The E_a for the entire assembly mechanism (binding to pore formation) is about 40 kcal/mol. This E_a is consistent with the sum of the independently derived E_a of ∼19 kcal/mol for binding and oligomerization (20) and the E_a of ∼19 kcal determined here for the prepore-to-pore transition. The free energy contributed by the formation of a strong electrostatic interaction or salt bridge typically exceeds −1 kcal/mol (16); hence, the formation of one such interaction per monomer of the prepore oligomer (7) would easily exceed the E_a of 19 kcal/mol required to disrupt the D3–D1,2 interface and drive the prepore-to-pore transition. It also explains why E_a for pore formation by PFO is 24 kcal/mol, as the free energy provided by the intact E–K pair would eliminate the prepore-to-pore E_a.

We propose that once the prepore is established, the monomers tilt in a concerted fashion, a process that may be facilitated by the π-stacking of aromatic residues Y181 and F318 (18). This action positions the K336 side chain near the monomer–monomer interface, where E183 resides in the ring complex. The electrostatic interaction between these two residues likely progresses from a weak to strong interaction, which further decreases the distance between them and draws β1 closer to β4. Because β1 is contiguous with the α-helical bundle that resides at the D3–D1,2 interface, this interaction disrupts the remaining stabilizing interactions at the D3–D1,2 interface, which allows the α-helical bundle to unfold and refold into TMH1.

CDCs that exhibit a conserved E–K pair and the cognate Asn at the D3–D1,2 interface cluster in a single large clade when their primary structures are aligned (Fig. S1). Most other CDCs that lack the cognate E–K–N trio instead contain an E–G–G trio. For unknown reasons, these CDCs have apparently evolved a different approach to controlling the prepore-to-pore transition, but as with the electrostatic controlled mechanism, it is likely to be directed at disrupting the D3–D1,2 interface.

Compelling evidence suggests that the unrelated βPFT, Staphylococcus aureus α-hemolysin, may also use an electrostatic mechanism to control the prepore-to-pore transition. α-Hemolysin forms a heptameric pore complex; each monomer contributes a single TMH to the β-barrel (25). Bayley and colleagues (26, 27) showed that the mutation of K110 or D152, which are located at the beginning and end of its folded TMH structure in the monomer, respectively, trapped α-hemolysin in a stable prepore. Subsequently, the crystal structures for the α-hemolysin monomer (28) and pore complex (25) showed that K110 is displaced about 16 Å from its position in the monomer to form a salt bridge with D152 in the pore complex (Fig. S4). Hence, an electrostatically controlled prepore-to-pore transition, similar to that observed here for PFO, may take place in α-hemolysin. A similar arrangement may exist in the octameric hetero-oligomer of S. aureus γ-hemolysin (29) but remains to be shown experimentally.

In summary, the formation of a single intermolecular electrostatic interaction is a vital element that controls the prepore-to-pore transition by supplying the free energy necessary to disrupt a critical interface in the prepore complex, which commits it to inserting the β-barrel pore. These studies will form the basis for the investigation into the mechanisms that control prepore-to-pore conversion by other members of the large βPFT family of toxins (not just the CDCs), as well as the membrane attack complex/perforin family of pore-forming proteins, which are structurally and mechanistically related to the CDCs (30, 31).

Materials and Methods

Plasmids, Bacterial Strains, and Toxin Production and Purification.

The amino acid sequence for the PFO Cys-less derivative (PFO^C459A) was codon optimized for E. coli expression and cloned into pET-15b (Novagen) by GenScript. Protein expressed from this plasmid (pRT30) is referred to as PFO here. QuikChange mutagenesis (Stratagene) was used to introduce mutations into PFO, using pRT30 as template. The DNA of each mutant was sequenced and then analyzed using Sequencher.

Hexahistidine-tagged, signal peptide-deficient PFO and derivatives were transformed into E. coli Tuner cells for expression (Novagen). Cells were cultured and the recombinant protein purified, assayed for concentration, and stored under conditions described previously (12), with the exception that cysteine-containing proteins were stored with 50 µM [Tris(2-carboxyethyl)phosphine] at −80 °C.

Liposome Preparation.

Liposomes containing a 45:55 mol % ratio of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (Avanti Polar Lipids) to cholesterol (Steraloids) with and without encapsulated 5 (6)-carboxyfluorescein (CF) (Sigma) were made as previously described (12, 32).

PFO Modification with IANBD and Charged MTS Reagents.

PFO derivatives with a Cys at various positions were labeled at the Cys sulfhydryl with the environmentally sensitive fluorescent dye N,N′-dimethyl-N-(iodoacetly)-N′-(7-nitrobenz-2-oxa-1,3-diazol-4-yl) ethylenediamine (IANBD; Invitrogen) and the free dye removed by previously described methods (18).

Modification of PFO derivatives with the MTS reagents (Affymetrix) was carried out with ∼30,000 molar excess of MTS reagents to protein (per the manufacturers instructions). Modification of proteins with NEM (Invitrogen) was carried out using a 20 molar excess of NEM to protein.

Steady-State and Kinetic Fluorescence Spectroscopy.

All steady-state and kinetic fluorescence-based experiments were performed on a Fluorolog-3 Spectrofluorometer with the FluorEssence software. The pore-forming activity (i.e., EC₅₀) of PFO and its derivatives on CF liposomes was determined as described (32). CF release kinetics before and after modification of the various PFO derivatives with NEM or the MTS reagents were examined by setting the excitation wavelength to 470 nm and following the changes in emission at 515 nm (1 nm bandpass) over time. Data were collected using an integration time of 1 s.

Cross-Linking and SDS-AGE.

Analysis of oligomer formation on liposomes was determined by SDS-AGE, as described previously (5). To ensure the sulfhydryl groups were reduced, PFO^E183C:K336C was pretreated with 10 mM DTT for 30 min at room temperature (RT) before the DTT was removed with Sephadex G-50 (PBS with 50 mM EDTA). Toxin (3.5 μM, 7 μM Cys residues) was then incubated with liposomes (20 min at RT) to assemble the prepore before cross-linkers BMOE, BMB, and DBB (Santa Cruz Biotechnology) were added at 0.25, 1, or 5 times molar excess of the cysteines in the PFO derivatives and incubated for 60 min. The reactions were quenched with 10 mM DTT to prevent further cross-linking. SDS-AGE sample buffer (6 μL) and 6 μL of a 25% stock of SDS (in water) were added to each sample, which were then heated at 95 °C for 5 min and separated by SDS-AGE. Cross-linkers were suspended in DMSO or ethanol (DBB) and diluted (PBS-EDTA) to a working concentration of 0.1 mM.

Thermal T_m Measurements.

The T_ms of the various PFO mutants (final concentration, 0.5 mg/mL in PBS) was determined using a 7500 Fast Real-Time PCR System using the Protein Thermal Shift Dye Kit (Applied Biosystems) according to the manufacturer’s protocol. The effect of NEM labeling on the T_m for PFO, PFO^E183A/N197C, and PFO^K336A/N197C was determined by incubating the proteins with NEM (20 molar excess) for 1 h at RT before T_m analysis. The PFO^PP was incubated with 1 mM DTT (1 h on ice) to reduce the disulfide before T_m analysis (1 mM DTT had no effect on the T_m of PFO).

Arrhenius E_a Measurements.

PFO and PFO^E183C (88 pmol/50 µL) were incubated with CF-liposomes (10 μL) in a final volume of 100 μL for 1 min on ice to minimize prepore formation. This mixture was then injected into 2.45 mL Hepes-buffered saline (HBS) in the fluorometer cuvette held at a specific temperature and CF emission measured over time. The same experiment was repeated with the exception that PFO^E183C was first allowed to form the prepore on the CF liposomes at RT for 20 min before injection into HBS. The initial velocity (k₀) of pore formation (i.e., the linear portion of the CF release curve) was determined from 20 °C to 50 °C. The E_a was then calculated from the equation E_a = −R*m, where m is the slope of the line derived by plotting ln(k₀) versus 1/T(K) and R is the gas constant (8.314 J/molK).

MD Simulation.

A 36-monomer (1,384,727 total atoms) molecular model of the PFO prepore structure was assembled on the basis of the PFO monomer crystal structure (Protein Data Bank ID code 1PFO). The monomers were evenly rotated about a central axis and were solvated and ionized to be electrically neutral with 0.15 M NaCl. Cylindrical boundary conditions were used to restrict the simulation to a radius of 170 Å and a height of 140 Å. MD equilibrium simulations were performed at 310 K, using NAMD 2.9.51 (33) on an IBM BlueGene/Q for 5 ns. An additional model was built (36 monomers) tilted ∼30° vertically in a counterclockwise direction. To induce the tilting movement, weak harmonic constraints were applied (0.1 kcal/mol/Ų) to the monomer backbone in D1–D3 (residues 29–380) to the new template positions, an effective method for applying a bulk movement to large complexes without greatly interfering with side-chain conformations. This motion was repeated six times to sample conformational pathways and side-chain interactions, with each simulation on the order of 10 ns in length (with constraints applied). These were followed by 5–10 ns of simulations with constraints removed.