Crucial Role of Perfringolysin O D1 Domain in Orchestrating Structural Transitions Leading to Membrane-perforating Pores

Background: Pore formation by perfringolysin O consists of well defined stages, but the mechanism is not fully understood. Results: We mapped regions in toxin that undergo stabilization and destabilization upon oligomerization. Conclusion: Properly folded D1 orchestrates structural transitions in other domains necessary for pore formation. Significance: Study of structural changes in cytolysin domains upon binding to membrane is crucial for understanding pore formation mechanism. Perfringolysin O (PFO) is a toxic protein that binds to cholesterol-containing membranes, oligomerizes, and forms a β-barrel transmembrane pore, leading to cell lysis. Previous studies have uncovered the sequence of events in this multistage structural transition to a considerable detail, but the underlying molecular mechanisms are not yet fully understood. By measuring hydrogen-deuterium exchange patterns of peptide bond amide protons monitored by mass spectrometry (MS), we have mapped structural changes in PFO and its variant bearing a point mutation during incorporation to the lipid environment. We have defined all regions that undergo structural changes caused by the interaction with the lipid environment both in wild-type PFO, thus providing new experimental constraints for molecular modeling of the pore formation process, and in a point mutant, W165T, for which the pore formation process is known to be inefficient. We have demonstrated that point mutation W165T causes destabilization of protein solution structure, strongest for domain D1, which interrupts the pathway of structural transitions in other domains necessary for proper oligomerization in the membrane. In PFO, the strongest changes accompanying binding to the membrane focus in D1; the C-terminal part of D4; and strands β1, β4, and β5 of D3. These changes were much weaker for PFOW165Tlipo where substantial stabilization was observed only in D4 domain. In this study, the application of hydrogen-deuterium exchange analysis monitored by MS provided new insight into conformational changes of PFO associated with the membrane binding, oligomerization, and lytic pore formation.

(D1-D4) ( Fig. 1) that play different roles in the consecutive stages of pore formation (Fig. 2). The recognition of cholesterol-rich membranes and direct initial interaction with the membrane is mediated by the region located in the D4 domain (17). Data presented by Iwamoto and co-workers (18,19) identified the region responsible for anchoring PFO in cholesterol-rich membranes as a tryptophan-rich peptide composed of 11 amino acid residues (ECTGLAWEWWR), termed the undecapeptide, which is located near the C terminus of the D4 domain ( Fig. 1). Recently, two additional residues in loop L1 of the D4 domain, namely Thr-490 and Leu-491, were identified as indispensable for recognition of cholesterol in the lipid membrane (20). The D4 domain loops provide only initial contact and do not insert deeply into the membrane. Final penetration of PFO across the membrane, involving conversion of transmembrane helices (TMH1 and TMH2) of domain D3 into membrane ␤-hairpins, is provoked by structural changes induced by monomer-monomer interaction. TMH insertion into the bilayer does not occur independently for each monomer. Instead, TMH insertion is coupled with the insertion of TMHs of neighboring monomers to create the ␤-barrel in the bilayer (21). The initial monomer-monomer contacts are known to trigger a cascade of structural changes in domain D2 and in TMH-containing domain D3 that increase the number of intermolecular interactions, properly position monomers, and establish the geometry of the ring complex (21)(22)(23). Particularly, oligomerization is promoted by formation of hydrogen bonds between ␤-strand 1 (␤1) in D3 of one monomer and ␤4 from D3 of a second monomer. Association of both molecules is enabled by rotation of the ␤5 strand away from ␤4 in D3 domain, leading to the exposure of ␤4 to the polar environment (24). Mutational analysis showed that an interaction between Tyr-181 in ␤1 of one subunit and Phe-318 in ␤4 of a neighboring subunit forces a proper orientation of the two monomers, and this change in orientation defines pore geometry and size (24) as the oligomers grow by attachment of subsequent monomers. As a result of oligomeric stabilization, the D2 domain undergoes a vertical collapse with concomitant disruption of its solution state structure. This change destabilizes the D2-D3 contact surface and leads to profound rearrangement within the D3 domain, causing insertion of this domain into the lipid bilayer (25). Six short transmembrane helices (three from TMH1 and three from TMH2) in domain D3 are unfolded, and pairs of amphipathic transmembrane ␤-hairpins are generated (24,26,27). Insertion of the ␤-hairpins into a membrane is the final step of creation of a large, transmembrane ␤-barrel (28).
The scenario described above focuses on the involvement of domains D2, D3, and D4 in pore formation by PFO. Interestingly, an interference in the D1 domain by substitution of Trp-165 with non-aromatic amino acids resulted in loss of lytic activity of PFO (29). Further analysis of the W165T variant indicated that this mutation does not block D4 binding to lipid membrane but disturbs the formation of the ring-shaped oligomeric PFO complex by preventing ␤4 from pairing with ␤1 of another membrane-bound PFO monomer. Instead of ringshaped oligomers, linear oligomers were found in the W165T protein (29). The solution structure of PFO shows that D1 is in direct contact with D2 and D3 domains, which undergo major structural changes during incorporation into the membranes and formation of pores. Combined with the ability of a point mutation in D1 to disrupt the pore formation, these data strongly suggest that D1 might be crucial for transmitting conformational transitions between the other domains.
Structural insight into the properties of the PFO W165T mutant can bring new information helpful to elucidate the role of D1 in conformational changes of D2 and D3 domains during pore formation. Such analysis requires access to structural data for a protein-lipid membrane assembly. Recent advances in hydrogen-deuterium exchange analysis monitored by mass spectrometry (HDX-MS) provide a new tool that allows for insight not only into the structural properties of a protein in solution (30) but also into structural changes accompanying its  . Schematic illustration of a conformational change in perfringolysin O during its interaction with a cholesterol-containing membrane and transition from the prepore to the conducting pore state. Main domains D1, D2, D3, and D4 are indicated. Water-soluble monomer bound to the cholesterol-containing membrane (I) triggers the structural rearrangements required to initiate the formation of prepore complex (II) that ultimately turns into a transmembrane, oligomeric pore (III). Cholesterol is represented as green diamonds. The figure is based on Heuck et al. (22).
insertion into the lipid membrane as exemplified by the studies of the bacterial chemotaxis receptor (31). Typical HDX analysis yields only medium resolution data as it allows the measurement of HDX kinetics averaged over protein peptic peptides. Conversely, HDX can be measured under native conditions at relatively low micromolar protein concentrations with no molecular mass limit. In many instances, this methodology has provided useful data revealing structural constraints that have allowed the verification of molecular models of proteins and their structural transitions. Motivated by this, we have carried out a comparison of the patterns of exchange of amide protons (solution versus membrane structure) in PFO. As a basis of our study, we have used a PFO variant bearing a single point mutation, C459A, that was shown previously to have no effect on the pore formation process (26). We have also studied PFO bearing an additional mutation, W165T, that causes deranged pore formation. As a result, we have mapped the regions in which the HDX kinetics are changed upon the insertion of PFO and its PFO W165T mutant into the lipid bilayer. This information was helpful to define the molecular mechanism of the involvement of D1 domain in the process of pore formation.

EXPERIMENTAL PROCEDURES
Purification of Recombinant Proteins-A synthetic gene for perfringolysin O of C. perfringens was prepared by GenScript basing on cDNA sequence number CP000246.1 at NCBI with a single point mutation in which Cys-459 had been changed to an alanine (26). The sequence was optimized for expression in Escherichia coli. The synthesized gene was devoid of a leader sequence coding for 28 N-terminal amino acids to ensure intracellular accumulation of the expressed protein. The functional PFO used in the various assays described below is PFO C459A ⌬28.
To obtain PFO with a GST tag at the N terminus, the construct was cloned into pGEX4T vector using BamHI and EcoRI sites. pGEX4T-3 vectors containing the PFO sequence separated from the GST sequence by a motif encoding ENLYFQG recognized by tobacco etch virus protease were obtained as described by Kwiatkowska et al. (32). Substitution of W165T in the PFO sequence was performed by site-directed mutagenesis method with a pair of primers: 5Ј-GAT GAA CTG GTC AGC  AAA ACC AAC GAA AAA TAC AGC AGC AC-3Ј and  5Ј-GTG CTG CTG TAT TTT TCG TTG GTT TTG CTG ACC  AGT TCA TC-3Ј. Briefly, cells of E. coli BL21(DE3) strain transformed with the vectors were grown at 37°C in LB medium containing 100 g/ml ampicillin to an OD of 0.6 when 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside was added. The culture was continued at 18°C for 20 h, and harvested bacteria were lysed in the presence of 0.35 mg/ml lysozyme and protease inhibitor mixture (10 min; 4°C). The suspension was supplemented with 1% Triton X-100 and sonicated on ice (15 min; 0.3 cycle; amplitude, 33%) using a UP200S Hielscher sonifier (Germany). Obtained lysates were clarified by centrifugation at 18,000 ϫ g for 30 min at 4°C and loaded onto a glutathione-Sepharose 4B column (GE Healthcare). GST-PFO and GST-PFO W165T proteins were eluted from the column with 10 mM glutathione, 5 mM DTT, 50 mM Tris, pH 8.0. Fractions containing recombinant proteins were pooled, and the buffer was exchanged to buffer containing 50 mM Tris, 100 mM NaCl, 1 mM DTT, pH 7.4 using a PD-10 desalting column (GE Healthcare). To obtain PFO and PFO W165T without a GST tag, the proteins (1 mg) were bound to a glutathione-Sepharose 4B column and cleaved in the presence of 50 g of tobacco etch virus protease for 3 h at 30°C in 50 mM Tris, pH 8.0 buffer supplemented with 1 mM DTT. Proteins released from the column were collected and loaded onto a column with HisLink Protein Purification Resin (Promega) to remove tobacco etch virus protease. PFO and PFO W165T recovered from the column were loaded onto a PD-10 desalting column for buffer exchange into 100 mM NaCl, 1 mM DTT, 50 mM Tris, pH 7.4. The final samples were supplemented with 20% sucrose and frozen in liquid nitrogen. In some sets of experiments, recombinant proteins PFO and PFO W165T linked with GST tag were also applied.
Determination of Molecular Mass by Size Exclusion Chromatography-The molecular mass of PFO and PFO W165T was determined by gel filtration on an FPLC system (Amersham Biosciences). Approximately 100 g of protein in 100 l of 150 mM NaCl, 30 mM Tris-HCl, pH 7.5 was injected onto a Superdex 200 10/300 GL column (GE Healthcare) and analyzed at a 0.5 ml/min flow rate. Elution of proteins was followed at 280 nm. The void volume of the column was determined with blue dextran. The calibration standards (Bio-Rad) were: thyroglobulin (670,000 Da), ␥-globulin (158,000 Da), ovalbumin (44,000 Da), myoglobin (17,000 Da), and vitamin B 12 (1,350 Da).
Carboxyfluorescein Release from SUVs-Small unilamellar vesicles (SUVs) were prepared from 1,2-di-(9Z-octadecenoyl)sn-glycero-3-phosphocholine (DOPC; Avanti) and various cholest-5-en-3␤-ol (cholesterol; Sigma) concentrations (in a range of 10 -70%; the total lipid concentration was 2 mM) in chloroform:methanol (1:1, v:v) and dried under a nitrogen stream. The lipid film was resuspended in TBS (150 mM NaCl, 50 mM Tris, pH 7.5) supplemented with 50 M 6-carboxyfluorescein (Sigma) and vortexed for 5 min. Finally, SUVs were obtained by six cycles of freeze-thawing and centrifugation (10,000 ϫ g; 30 min; 4°C). The pellet of liposomes was resuspended in TBS and incubated with 1 M PFO or its PFO W165T mutant. After incubation for 30 min at 25°C, the suspension was centrifuged, and the level of released carboxyfluorescein in supernatants was measured in a Jasco FP 6500 spectrofluorometer at an excitation/emission of 492/517 nm. The results were expressed as the percentage of maximum dye release induced by addition of 0.1% Triton X-100 to SUVs.
Protein/Lipid Overlay Assay-To test the specificity of lipid binding by PFO and PFO W165T , 1 l of DOPC (Avanti), cholesterol, or N-(hexadecanoyl)-sphing-4-enine-1-phosphocholine (SM, Sigma-Aldrich) in chloroform:methanol:water mixture (1:1:0.3, v/v) containing various amounts of lipid (50 -10,000 pmol) was applied onto nitrocellulose according to the procedure described by Abdel Shakor et al. (33). The membranes were incubated overnight with 10 g/ml GST-tagged PFO or PFO W165T and exposed to goat anti-GST IgG conjugated with horseradish peroxidase (HRP) (Rockland). Interactive spots were visualized with ECL Western blotting detection reagents (GE Healthcare). Amounts of PFO proteins bound to cholesterol were quantified densitometrically with GelQuant.NET software provided by BiochemLab Solutions. For normalization, the densitometric data were expressed in relation to a signal generated by GST-PFO bound to 20 nmol of cholesterol and arbitrarily equalized to 1.
Binding of PFO and PFO W165T to Liposomes-SUVs composed of DOPC/cholesterol with different cholesterol/PFO molar ratios (from 20:1 to 1000:1) were prepared by drying lipids mixture, suspension in TBS, and sonication on ice for 30 min. After pelleting (100,000 ϫ g; 45 min; 4°C), SUVs were resuspended in 30 l of PBS. Liposome suspensions (1 mM final lipid concentration) were incubated with 1 M PFO (with GST) for 45 min at 25°C, centrifuged, and resuspended in 0.2% Triton X-100 in TBS. Samples were spotted onto nitrocellulose, and proteins bound to liposomes were detected by goat anti-GST IgG conjugated with HRP. Immunoreactive spots were visualized by chemiluminescence and quantified densitometrically using GelQuant.NET software. For normalization, the densitometric data were expressed in relation to the PFO level detected in liposomes at a cholesterol/PFO ratio of 1000:1 and arbitrarily equalized to 100.
Surface Plasmon Resonance Measurements-To analyze the interaction of PFO proteins with large unilamellar vesicles, a BIAcore 3000 apparatus (BIAcore, GE Healthcare) equipped with an L1 chip was used. Large unilamellar vesicles composed of DOPC/cholesterol and DOPC/SM at a 1:1 molar ratio were prepared as described previously, giving a 1 mM final lipid concentration (34). Measurements were conducted according to Kulma et al. (34) using a 1 M solution of PFO and PFO W165T at a flow rate 5 l/min for 300 s (association phase). Dissociation of deposited protein was followed for another 300 s (dissociation phase). Determinations of equilibrium binding constants were calculated using BIAevaluation software (GE Healthcare).
Electron Microscopy of Oligomeric Complexes-Ultrastructural studies of PFO oligomeric complexes were performed according to Hotze et al. (29) with some modifications. PFO proteins in 20 mM HEPES, pH 7.4 (0.1 or 0.2 mg/ml (1.8 or 3.6 M, respectively)) were spotted onto a Teflon surface as 13-l droplets and coated with 1 l of 1 mg/ml lipid mixture composed of cholesterol and DOPC (1:1 molar ratio; 2 mM total lipid concentration). After incubation at room temperature for 30 min, samples were transferred to carbon-coated grids and negatively stained with 2% uranyl acetate. In a series of experiments, 0.1-0.2 mg/ml protein was incubated with 1 mg/ml lipid mixture (cholesterol/DOPC at a 1:1 molar ratio) directly on the grids. This technique facilitated assembly of oligomers by PFO W165T . Samples were examined under a JEM-1200EX (JEOL) microscope.
Susceptibility of Toxins to a Protease-Liposomes composed of DOPC/cholesterol (molar ratio, 1:1; 2 mM total lipid concentration) were incubated with 1 M PFO protein for 30 min at 25°C in 30 mM Tris, 150 mM NaCl, pH 7.5 buffer. After centrifugation (20 min; 14,000 rpm), the pellet was resuspended in TBS or TBS supplemented with 0.05% Triton X-100 and incubated for 5 min at 25°C after which the pH of the sample was adjusted to pH 2.4, and the sample was incubated with pepsin (Promega) for 3 min at 4°C at an enzyme to substrate ratio of 1:100 (w/w). The reaction mixture was supplemented with SDS sample buffer, and products of digestion were analyzed by SDS-PAGE.
Sample Preparation for Hydrogen-Deuterium Exchange-For HDX experiments, lipid vesicles were prepared by addition of the stock solutions of DOPC and cholesterol to a chloroform: methanol mixture (1:1, v/v) and then evaporated under a stream of nitrogen. The samples were then suspended in PBS, vortexed for 3 min, and subjected to a freeze-thaw procedure between liquid nitrogen and a water bath; the freeze-thaw cycle was repeated eight times. In the following step, liposomes were centrifuged at 24,000 ϫ g for 20 min at 4°C, washed in PBS, and pelleted as above. Freshly made lipid vesicles (2 mM lipid concentration) were incubated with 1 M PFO in PBS at room temperature for 45 min. The molar ratio of PFO/cholesterol was 1:1000. Subsequently, the liposomes were pelleted and rinsed in PBS as described previously. The final pellet was resuspended in 30 mM Tris, 150 mM NaCl, pH 7.5 and used in HDX experiments.
Hydrogen-Deuterium Exchange Measurements-Hydrogendeuterium exchange measurements were performed as described previously by Sitkiewicz et al. (30) with several modifications. Experiments were started by creating a list of peptic PFO peptides using a non-deuterated protein sample. For this purpose, 5 l of PFO sample (at least 15 M) was diluted with 45 l of H 2 O-based Reaction buffer (30 mM Tris-HCl, 150 mM NaCl, pH 7.5) and immediately mixed with 10 l of H 2 O Stop buffer (2 M glycine, 150 mM NaCl, pH 2.4) and 0.2% Triton X-100. The sample was injected into the nanoACQUITY UPLC system (Waters, Milford, MA). After an on-line pepsin digestion of perfringolysin O, peptides were trapped on a C 18 column. Next the sample was passed through a second precolumn (Acquity BEH C 18 VanGuard precolumn, 1.7-m resin, Waters) that was directly connected to a reversed phase column (Acquity UPLC BEH C 18 column, 1.0 ϫ 100 mm, 1.7 m resin, Waters). The second precolumn was used to prevent the introduction of lipids to the analytical column (35,36). The SYNAPT G2 HDMS mass spectrometer (Waters) was operated in TOF mode. The spectrometer operating parameters were as follows: electrospray ionization positive mode; capillary voltage, 3 kV; sampling cone voltage, 35 V; extraction cone voltage, 3 V; source temperature, 80°C; desolvation temperature, 175°C; desolvation gas flow, 800 liters/h. The calibration of the spectrometer was performed using standard calibration solutions. Identification of peptides was achieved using ProteinLynx Global Server software (Waters). The list of identified peptides comprising peptide m/z, charge, and retention time was processed in the DynamX 2.0 hydrogen deuterium data analysis program (Waters).
In hydrogen-deuterium exchange reactions, two sets of samples were analyzed and compared: PFO in aqueous solution (PFO aq ) and PFO incorporated into liposomes (PFO lipo ). The Reaction buffer and the Stop buffer were prepared using D 2 O (99.8%; Cambridge Isotope Laboratories, Inc.) with pH adjusted using DCl and NaOD (Sigma). The hydrogen-deuterium exchange reaction was initiated by addition of 45 l of D 2 O Reaction buffer (30 mM Tris, 150 mM NaCl, pH 7.5) to 5 l of a sample. The sample was exposed to D 2 O Reaction buffer for varying periods of time (10 s, 20 min, 1 h, up to 24 h) at room temperature. After the specified time, the sample was acidified by mixing with 10 l of prechilled D 2 O Stop buffer (2 M glycine, 150 mM NaCl, pH 2.4), which quenched the hydrogen-deuterium exchange reaction. Subsequently, deuterated protein was

Structural Consequences of PFO Interaction with Liposomes
OCTOBER 10, 2014 • VOLUME 289 • NUMBER 41 released from liposomes by addition of Triton X-100 to a final concentration of 0.2% and shaken for 2 min at 900 rpm at 0°C (Thermomixer MKR 13, Ditabis). Immediately after stopping the reaction, the sample was manually injected into the nano-ACQUITY UPLC system. Pepsin digestion, LC, and MS analysis were conducted exactly as reported for the non-deuterated sample.
A back-exchange control experiment was conducted as described previously by Kupniewska-Kozak et al. (37). Briefly, to determine a maximum exchange (back-exchange control), 5 l of protein stock was diluted with 45 l of D 2 O Reaction buffer, incubated overnight, mixed with D 2 O Stop buffer, and analyzed as described above. The deuteration level in the backexchange experiment was calculated using DynamX 2.0 and described as 100% exchange (M ex 100 ). The average mass of peptides before exchange was defined as 0% exchange (M ex 0 ). The above scheme of experiments (including back-exchange control and the HDX experiment with the sample incubated with D 2 O buffer for varying time periods) allowed us to collect the data for a set of identical peptides in both the control experiment and the HDX experiment. For each incubation time, the HDX reaction was repeated at least three times. The data presented below correspond to a value of a mean and S.D. calculated for those experiments.
HDX Data Analysis-Peptides were identified using Protein-Lynx Global Server (PLGS) software. The level of deuterium uptake of each PFO peptide was calculated using DynamX 2.0 software based on the list of perfringolysin O peptic peptides obtained from ProteinLynx Global Server software. The DynamX 2.0 acceptance criteria were: minimum intensity threshold of 1000 and minimum products per amino acid at a level of 0.3. The values of deuterium uptake in the exchange reaction (M ex ), a non-deuterated sample (M ex 0 ), and a control back-exchange experiment (M ex 100 ) obtained from the DynamX analysis were next manually verified. Isotopic envelopes that were inconclusive, overlapping, or representing a carryover effect were deleted from further analysis. Data analysis was completed as described by Sitkiewicz et al. (30). The fraction of exchange (f) of a given peptide was calculated by taking into account the non-deuterated sample and the backexchange control values with the following equation.
Mean values of exchange and standard deviations were computed using data collected in at least three independent HDX experiments. The difference in exchange between protein in aqueous solution (f aq ) and protein incorporated into liposomes (f lipo ) was achieved by subtracting the values of exchanged fractions (f aq and f lipo ). The same calculation was performed for comparison of PFO and its mutant (PFO aq Ϫ PFO W165T aq ). The error bars for subtracted data were calculated as the square root of the sum of variances of subtracted numbers. Finally, all data generated by Origin 8.0 were presented in graphs and visualized on the PFO structure obtained from the Protein Data Bank using YASARA software.

Characterization of the Binding of PFO and PFO W165T to
Cholesterol-Before characterizing the structure of PFO and PFO W165T by HDX, we tested lipid binding, lytic activity, and oligomerization status of these proteins.
Because cholesterol is known to serve as a target for PFO, the binding of PFO to cholesterol-containing liposomes and the influence of the W165T mutation on recognition and binding to the lipid were tested using various approaches. A lipid overlay assay was used to reveal the specificity of binding of N-terminally GST-tagged PFO and PFO W165T to cholesterol. We demonstrated that both PFO and PFO W165T at 10 g/ml (125 nM) were specifically bound to cholesterol and not to sphingomyelin or DOPC. Densitometric analysis of dots revealed that binding of PFO W165T to cholesterol was reduced by about 40% (Fig. 3A).
The ability of GST-PFO and GST-PFO W165T to bind to cholesterol was also analyzed using SUVs containing various cholesterol concentrations yielding cholesterol/PFO molar ratios ranging from 20:1 to 1000:1 (Fig. 3B). As expected, we found that binding of both proteins to SUVs increased with increasing concentration of cholesterol with a marked improvement of the binding above a 400:1 cholesterol/PFO ratio. However, binding of PFO W165T to SUVs at a 1000:1 cholesterol/PFO ratio was diminished by about 45% in comparison with the binding of PFO. The decreased binding of the PFO mutant to cholesterolcontaining liposomes was confirmed by surface plasmon resonance. For this analysis, large unilamellar vesicles composed of cholesterol/DOPC (molar ratio, 1:1) and cholesterol/protein (ratio, 1000:1) were used. As shown in Fig. 3C, the binding of PFO to cholesterol/DOPC reached a plateau of about 3700 relative units after 300 s of dissociation time (Fig. 3C) and was ϳ1.8-fold stronger than PFO W165T (ϳ2000 relative units after dissociation). Analysis of surface plasmon resonance sensorgrams demonstrated that, in contrast to cholesterol/DOPC liposomes, unspecific binding of PFO and its W165T mutant to DOPC/SM large unilamellar vesicles was very weak and almost completely reversible (ϳ50 relative units after 300-s dissociation) (Fig. 3C). Obtained sensorgrams for cholesterol/DOPC liposomes allowed analysis of the data by fitting the sensorgrams to a 1:1 binding model. Table 1 shows that the equilibrium dissociation constant (K D ) of PFO was equal to 8.58 ϫ 10 Ϫ12 M. PFO W165T was bound much more weakly to cholesterol/DOPC liposomes with a dramatic increase in the K D value of ϳ15,000-fold.
As a consequence of the interaction with cholesterol-containing membranes, PFO oligomerizes, becomes incorporated into the membrane, forms pores, and evokes hemolysis of erythrocytes. To determine the relative lytic activities of wildtype PFO and its W165T mutant, a hemolytic assay and fluorometric measurements of carboxyfluorescein release from liposomes were performed. In agreement with previous studies (29), PFO evoked hemolysis of sheep erythrocytes in a dose-dependent manner at 37°C (data not shown). In contrast, the W165T point mutation led to loss of the hemolytic activity even at 100 nM PFO (data not shown). Previous results showing weak binding of PFO W165T to cholesterol (Fig. 3, A, B, and C) sug-gested that the higher concentrations of cholesterol in membranes may be necessary to detect lytic activity of this protein.
Indeed, in experiments with toxin-mediated release of carboxyfluorescein from liposomes containing 50% cholesterol in com-parison with erythrocytes containing 28% of the lipid (38), a low level lytic activity of PFO W165T was observed (data not shown). However, this effect was much weaker than for PFO and did not exceed 5% of carboxyfluorescein release at 40 nM PFO W165T .

JOURNAL OF BIOLOGICAL CHEMISTRY 28743
The level of carboxyfluorescein released from liposomes reached 100% for liposomes containing 40% cholesterol or more (Fig. 4A). Increasing the amount of cholesterol in liposomes had no effect on PFO W165T lytic activity, and release of fluorescent dye was not detected even for 70% cholesterol-containing liposomes (Fig. 4A).
The lytic activity of PFO correlates with its ability to form SDSresistant oligomers. Electron microscopy analysis confirmed the ability of PFO to form ring-shaped oligomers in the presence of cholesterol (Fig. 4C, panel I) in agreement with previous data (29). In contrast to PFO, mutant PFO W165T assembled into either a mixture of ring-shaped structures of relatively large perimeter and linear oligomers (Fig. 4C, panel II) or only linear oligomeric structures (Fig. 4C, panel III), the latter described earlier by Hotze et al. (29). Oligomers assembled by PFO W165T were easily disrupted by the mechanical force used during sample preparation. The data presented above indicated that PFO W165T cannot assemble into stable oligomers.
We also examined the protein in its preoligomerization state in solution using size exclusion chromatography. In agreement with previous studies (14), PFO in solution was mostly dimeric (Fig. 3D). PFO W165T was eluted from the gel filtration column in the same fractions as PFO, corresponding to a protein mass of 110 kDa, which is expected for PFO dimers. This result indicate that PFO W165T retains the ability to dimerize as found for PFO and that disturbances in oligomerization of PFO W165T upon binding to cholesterol-containing membranes do not originate from the differences in the preoligomeric status of the protein in solution.
Structural Changes in PFO in Solution as Monitored by HDX-The pattern of exchange of amide protons for deuterium was measured in PFO and PFO W165T in the aqueous buffer and after incorporation into liposomes. Pepsin digestion resulted in sequence coverage of 93.9 and 78.4% for PFO and PFO W165T , respectively, with overlapping peptides in multiple regions. Analysis of SDS-PAGE revealed that PFO bound to liposomes was partially protected from the proteolytic digestion (Fig. 4B). However, pretreatment of liposomes with Triton X-100 increased the availability of PFO for pepsin digestion (Fig. 4B). This result strongly indicated that permeabilization of liposomes with Triton X-100 before pepsin digestion allowed  peptides from PFO fragments located both on the surface and incorporated into the lipid bilayer to be obtained. To carry out hydrogen-deuterium exchange, samples were incubated with D 2 O for different periods, and each experiment was carried out in triplicate, allowing us to assess the reproducibility of the results. Data collected at shorter times of incubation revealed the occurrence of deuteration processes in flexible regions, whereas longer times of hydrogen-deuterium exchange provided insight into regions characterized by a more stable structure. For PFO aq and PFO lipo , the levels of exchange in all identified peptic peptides after 10 s (Fig. 5A) and 20 min (Fig. 5B) of exchange are shown aligned along the protein sequence. Also, the data presented in Fig. 5, A and B, are overlaid on domain partition (Fig. 5, E and F) and secondary structures (see Fig. 9A) based on the known crystallographic structure of PFO (Protein Data Bank code 1PFO) (16). In general, the dynamics of HDX in wild-type PFO aq is high, and the exchange levels smaller than 20% were found only for a few regions after 20 min (Fig. 5B). These regions include two peptides spanning positions 268 -278 (strand ␤3 at the interface between domains D1 and D3) and a few peptides in the region 389 -415 (domain D4 central ␤-strands S17 and S18 linked by loop L2) (Figs. 1, 5A, and 9A). These most stable peptides are localized within the internal core strands of domains D3 and D4 surrounded by other strands on each side. The ␤3 strand protrudes into the D1 domain and ␤-strand S17 of D4 is preceded by ␤-strand S16 of D2 broken by a short D4-D2 flexible linker. In addition to the most stable internal short regions at the D1-D3 and D2-D4 interfaces, only a few more short stretches remain relatively protected at longer incubation times, namely peptide 100 -110 in D1 and N-terminal peptides of TMH1, ␤3, and ␤4. In other regions, the exchange is substantial. Conversely, the number of peptides for which the exchange is very fast (Ͼ70% at 10 s) is also quite small with these regions including positions 54 -74 (D2 initial 20 amino acids), positions 133-142 (D1 region), positions 319 -340 (␤4-loop-␤5 in D3 including the Gly-Gly sequence), and positions 485-491 (interface between neighboring ␤-strand and loop L1 in D4) (Fig. 5, A and E). In D3, a strongly protected ␤3 strand is close in space in the structure to ␤4-loop-␤5, which is highly dynamic in solution. Other regions reveal intermediate levels of protection against exchange, but after 20 min, the majority of peptides are exchanged by more than 60% (Fig. 5, B and F), indicating an overall flexible structure of the protein, which is nevertheless precisely organized in space by a set of well defined stretches of relative stability.
Structural Changes of PFO in Liposomes-Upon liposomal binding, PFO undergoes profound structural changes allowing the formation of an ion channel, perforating the membrane. These changes alter the accessibility of amide protons to exchange. The pattern of exchange for PFO lipo along with data on PFO aq allowed us to observe the stabilization of the structure along the majority of the sequence, encompassing all four domains (Fig. 6, A and B). After 20 min, the number of regions that had undergone exchange of less than 20% increased with new stable regions at positions 100 -109 and 114 -122 (D1), 352-366 (D3/D1), 448 -462 (D4), and 482-496 (loop L1 in D4) (Figs. 5B and 6B). Loop L1 from D4 was among the least protected in PFO aq , whereas in PFO lipo , this loop was very strongly stabilized. Loop L1 is directly engaged in anchoring the PFO in the membrane, an event that precedes the channel formation. Besides loop L1 in D4, the more strongly stable regions in PFO lipo appeared in the D1 domain. No region in PFO lipo was exchanged by more than 70% at 10 s of exchange (Figs. 5A and 6A). Regions that were most strongly affected during transition to PFO lipo were visualized in the form of differential plots in which the fraction exchanged in PFO lipo was subtracted from the fraction exchanged in PFO aq for each peptide (Fig. 6, A and  B). When we analyzed differences in exchange between PFO aq and PFO lipo after 20 min of deuteration, we found that the most strongly stabilized fragments in PFO lipo included the majority of D1 peptides in regions 92-174 and 351-366, the C-terminal part of D4 covering L3, undecapeptide, and most strongly L1; also included were strands ␤1, ␤4, and ␤5 of D3 (Figs. 5F, 6B, and 9B). Peptides covering transdomain strands ␤2 and ␤3, some of them already considerably protected in PFO aq , became even more protected in PFO lipo (Fig. 7).
In D1, the most pronounced changes were observed for peptides 123-142, 133-142, and 143-160 (Figs. 5, A, B, E, and F, and 6, A and B). When these peptides were overlaid on the PFO structure, they were localized in a region consisting of ␤-strands S6 and S7 and part of helix 3 ( Figs. 1 and 9, A and B). Peptide 133-142 was one of the most exposed in PFO aq and became strongly stabilized in PFO lipo . Similarly, the region ␤4-loop-␤5 in D3 and loop L1 in D4, also among the most exposed in PFO aq , gained substantial stability in PFO lipo . According to a proposed model of interactions (29), stabilization in strands ␤1 and ␤4 of D3 reflects the formation of the monomer-monomer interface in oligomers. Interestingly, the N-terminal peptide of ␤4 (positions 315-320), C-terminally flanking TMH2, became strongly exposed upon transition to liposomes (Fig. 6, A and B). The N-terminal TMH2-flanking region (positions 279 -294) also became more flexible in PFO lipo (Fig. 6B). The TMH1-flanking regions were intermediate in stability (Fig. 6, A and B) and remained unchanged upon transition to liposomes. However, a peptide in TMH1 covering position Asn-197, which upon disruption of the D2-D3 interface was transferred from a nonpolar to a polar environment, also revealed increased exposure to exchange (Figs. 5, A and B,  and 6, A and B). Thus, the linker segments, localized between the transmembrane regions and the rest of the protein, retained flexibility in the PFO pores. The protection at both TMH regions did not change substantially; these regions may be equally well protected when buried inside the protein and inside the membrane. In transition to PFO lipo , large parts of D1, D3, and D4 became stabilized, whereas small regions flanking TMH2 in D3 became strongly exposed (Figs. 5, E and F; 6, A and  B; and 9, A and B).
Lytic activity remains equally high for higher cholesterol concentration. We also tested whether HDX patterns remain the same for higher cholesterol concentration. Fig. 8 compares level of HDX in PFO incorporated into liposomes containing 50 or 60% cholesterol. The differences in the exchange do not exceed 10% and are thus insignificant.
Structural Features of PFO W165T in Solution-HDX experiments were also conducted for the PFO W165T mutant that did not exhibit lytic activity (Fig. 4A) (29). The patterns of hydro-

Structural Consequences of PFO Interaction with Liposomes
OCTOBER 10, 2014 • VOLUME 289 • NUMBER 41  gen-deuterium exchange in the examined peptides of protein in aqueous solution (PFO W165T aq ) and those incorporated into liposomes (PFO W165T lipo ) were obtained for 10 s (Fig. 5, C and E) and 20 min of deuteration (Fig. 5, D and F). As before, peptic peptide positions were assigned to known PFO domains and to previously described regions of the protein.
An overall view of the PFO W165T aq exchange pattern indicated that the protein was even more dynamic than PFO aq . After 20 min of exchange, only a few peptides revealed lower than 40% deuteration, particularly at positions 225-230 (in ␤2 in D3), 315-320 (in ␤4 in D3), and 402-406 (L2 in D4) (Fig. 5D). As compared with wild-type PFO aq , the exchange level was generally higher in the mutant along nearly the entire sequence, indicating a more relaxed structure of the whole protein and not only at its selected domains.
Analysis of differential plots of the exchange pattern for PFO W165T and PFO in solution localized the most striking loss of stability to the entire D1 region 90 -174 (bearing the mutation site at position 165) and its flanking regions from D2 and D3 domains, extending the loss-of-stability region to positions 50 -270. The destabilization also propagated to the N-terminal part of D4 (Fig. 6E).
Structural Features of PFO W165T in Liposomes-Analysis of differential plots revealed that the anchoring of PFO W165T in the membrane led to much smaller changes in HDX patterns than observed for PFO (Fig. 6, C and D). Upon incorporation into liposomes, PFO W165T lipo showed some stabilization along the whole protein sequence, but the effects were much less pronounced than for PFO lipo itself. Very strong stabilization of PFO W165T in liposomes was observed only in D4, the undecapeptide region, and loop L1 (Figs. 5, C, D, E, and F, and 6 C and D). This stabilization reflects the anchoring event. Only a few regions destabilized in liposomes were observed in PFO W165T , namely 161-186 (␤1 in D1), 212-230 (C-terminal part of TMH1 and N-terminal part of ␤2 in D3), 315-320 (N-terminal part of ␤4 in D3), and 402-410 (L2 in D4) (Fig. 6, C and D). With the exception of the ␤4 fragment, these regions differed from those observed for wild-type PFO lipo (Figs. 6, A-D, and 9, B and D).
The observed anchoring of PFO W165T in the membrane leads to some structural changes, but the effects in comparison with  PFO are either much smaller as for the D1 domain or the ␤4 strand or opposite as for ␤1 and ␤2, reflecting the inability of the PFO W165T mutant to oligomerize in a proper way. Also, in PFO W165T , the peptide spanning position Asn-197 along with other TMH-flanking regions does not show exposure as observed for PFO (Fig. 5, B and F), indicating disruption of the D2-D3 interface (26). The single point mutation of W165T in D1 causes profound destabilization of D1 itself as detectable already in solution (Fig. 9C) that disrupted further structural transitions necessary for pore formation (Fig. 9D).

DISCUSSION
Previous studies on the structure of PFO combined with the analysis of a series of point mutants provided the basic sequence of events leading from a toxin in solution to a membrane-embedded pore (8,22). PFO attaches to the cholesterol molecules within the membrane only via loops at the tip of the D4 domain. Binding to the membrane leads to two concerted actions: protein oligomerization and profound structural changes in the monomers. The causative pathway linking the attachment signal with subsequent changes and the molecular mechanism by which the signal is transmitted from the D4 tip to the other domains are still not fully explained. The HDX-MS method applied in this work allowed us to gain structural insight not only for PFO in solution but also for its membrane-bound form and provided unique experimental constraints for further modeling of the process. Our results underscore the role of the D1 domain as it proved to be the most significantly affected region both upon transition to the membrane environment and upon W165T mutation.
The identity of the cholesterol recognition motif was until recently localized in the undecapeptide region, but more recent data have shown that the cholesterol recognition motif must also include a loop 1 conserved threonine-leucine pair (residues 490 and 491 of L1 loop in PFO) (20). In agreement with this finding, we demonstrated that during embedding of PFO into liposomes, the strongest stabilization was observed in L1 loop peptides in D4. Loops L3 and undecapeptide also become more stable as they are known to participate in anchoring the protein in the membrane (Figs. 5, E and F; 6, A and B; and 9B).
The first step of direct interaction of PFO with the membrane initiates the structural changes leading to prepore and finally pore formation. Changes include the following steps: oligomerization, disengagement of strand ␤5 and exposure of the ␤4 strand in D3, intermolecular stacking of ␤4 to expose ␤1 of another monomer, and transmembrane helix conversion to membrane-embedded ␤-strands. These events lead to structural collapse of PFO along its longitudinal axis connected to D2 collapse during the transition from the prepore state to pore formation (29). In the simplest scenario, the initial contact of the protein with the membrane is allosterically transmitted to D3, leading to breakage of the D2-D3 interface (40) while the disengaged ␤5 rotates around the glycine-glycine swivel, exposing one edge of ␤4 to the solvent. Our data demonstrated that the ␤4-␤5 region is among the least structured in PFO aq , indicating a small energy barrier required for this transition, and in PFO lipo , this region becomes stabilized by ␤1-␤4 intramolecular pairing, which was not observed for PFO W165T . ␤4 exposure was found previously to coincide with PFO binding to the membrane, leading to the conclusion that ␤4 exposure precedes oligomerization (24). Recently, however, it was shown that mutant W165T (D1 domain) becomes trapped before exposure of ␤4 but still retains the ability to form linear oligomers (29). We found similar linear structures formed by PFO W165T that in certain conditions were accompanied by ring-shaped oligomers (Fig. 4C). This observation implies the presence of monomer-monomer interactions preceding the ␤1-␤4-driven assembly and involvement of D1 in the propagation of structural changes. In agreement with this prediction, our results indicate D1 as a region characterized by the most striking changes accompanying pore formation. The stability in PFO lipo is increased along nearly the entire D1 sequence, excluding a few small fragments. The strongest gain in protection was observed for two regions: 110 -122 (corresponding to a loop linking strands S5 and S6 in D1) and 133-157 (strand S7-S8 loop, strand S8, and the N terminus of helix 3). Especially region 133-143, covering a loop linking strands S7 and S8, was highly exposed in solution and became significantly more protected in PFO lipo , indicating the most stable contact site (Figs. 1  and 9, A and B). This region exactly matches the epitope 136 -157 of an antibody specifically blocking oligomerization, not the cell binding step (41,42). Moreover, the majority of regions most strongly destabilized in PFO W165T in solution, as compared with PFO, belong to the D1 domain, indicating that the insufficient stability of D1 in W165T might be responsible for its inability to expose ␤4 (Figs. 5, E and F; 6E; and 9, A and C). Particularly, the Trp-165 residue (Fig. 1) directly contacts the loop linking strands 7 and 8, the above mentioned monomermonomer contact site. Appropriate rigidity of D1 may be crucial to trigger further changes. In agreement, oligomeric struc-tures formed by PFO W165T upon cholesterol binding were sensitive to mechanical stress and disintegrated during SDSagarose electrophoresis, which failed to reveal PFO W165T oligomers (Fig. 4C), as described earlier by Hotze et al. (29). Accordingly, PFO W165T lacks lytic activity (Fig. 4A) (29). The inability of PFO W165T to assemble into stable oligomers can be a result of decreasing affinity of the protein for cholesterol in comparison with PFO (Fig. 3, A, B, and C), suggesting that proper oligomerization of PFO strengthens its binding to the lipid (29).
D1 is highly interwoven with other domains, especially with D3 by means of ␤-strands penetrating the domain borders, with the longest one, 376 -398, linking the D2 and D4 domains and a ␤2-␤3 strand system linking domains D1 and D3. The interdomain communication may be realized by fine-tuning of the stiffness of a network of these strands. The ␤2-␤3 system is characterized by a strong curvature at the domain border, indicating a structural stress at the D1-D3 interface imposed by multiple D2-D3 contacts. As suggested, straightening the curved ␤-sheet can lead to rotation of D3 (29,43). D1 stability may be necessary to fine-tune the structural stress so that a trigger signal of sufficient strength relieves the stress and flexes D3. In other words, native D1 decreases the free energy barrier for release of the D3-D2 contacts, and this role cannot be fulfilled in PFO W165T mutant due to insufficient stability of this domain (Fig. 10). The trigger signal results from a combined structural change caused by membrane binding and monomermonomer interactions. It may be transmitted from D1 by the changes in stability of the ␤2-␤3 system. In agreement with this idea, we observed increased stability of ␤2-␤3 peptides in PFO lipo as compared with PFO aq (Fig. 7). It has been shown previously that the PFO W165T mutant forms pores if it is supplemented by wild-type PFO (29). The same effect was observed earlier for the Y181A mutant (21). Thus, D1, which is unstable in PFO W165T , may be stabilized from outside by intermolecular interactions with properly folded D1 of PFO, leading to the required structural changes.
Preformed oligomers are further stabilized by ␤1-␤4 pairing. The TMH helices unfold into ␤-hairpins stabilized by their edgewise association in multiple monomers. Cooperative stabilization of monomers within the prepore is coupled with the concerted TMH insertion into the membrane (21) made possible by the vertical collapse of the whole molecule. The TMH ␤-hairpins are extensions of antiparallel ␤-sheet ␤2-␤3 at the D1-D3 interface. The C-terminal part of ␤4, containing the Val-322 residue, is known to become exposed to solvent in the pore in contrast with its hydrophobic environment in monomers (43). In our study, the amide protons in the peptide covering position 322 are not protected in PFO aq and become stabilized in PFO lipo due to interaction with ␤1. Observed stabilization of the main chain does not exclude exposure of the side chain to solvent. In contrast, the amides of the N-terminal part of ␤4, which are stable in PFO aq , undergo a strong exposure in PFO lipo . This peptide contains Phe-318, which was shown to be required not only to stabilize the prepore complex but also to align the adjacent ␤1 and ␤4 strands in the proper register in the transmembrane ␤-barrel (24). Our results show that the amide protons in the N-terminal ␤4 segment, being located on the interface between membrane-embedded TMH2 and ␤4, undergo significant exposure. This does not exclude its stable side chain-side chain interactions. Similarly, slight destabilization in PFO lipo is observed at the ␤1-TMH1 and ␤3-TMH2 interfaces. In the pores, the H-bonded networks in the linker segments, which are localized at the interface between the membrane and the protein, become less stable than in solution.
The HDX-MS method has been broadly applied to study protein dynamics and structure (15,39). It allows the limitations of the classic methods, like the molecular mass limit and high protein concentration required for NMR or the necessity to obtain a protein crystal, to be overcome. This technique enables us to study the protein in its natural environment (broad pH range and aqueous or hydrophobic/lipid conditions) at relatively low, micromolar concentration in principle without mass limit. For PFO, it provided an effective tool to approach the structure of the protein both in solution and in the lipid membrane. The method, however, has its limitations. First, it reports on the exchange of amide protons, so it only provides information on the stability of main-chain hydrogen bonds or burial inside the structure without information on the side chains. Second, its resolution is limited as the exchange is measured at the peptic peptide level and not the level of single amides. For a particular protein, the effectiveness of the method depends on the effectiveness of the pepsin digestion and separation of peptides in LC. In our studies, PFO was digested by pepsin with very high efficiency (the sequence coverage was above 90%). Also, the analysis of protein-liposome complex might in some cases be challenging because not all detergents are recommended for HDX-MS (SDS is forbidden). Each HDX-MS experiment requires a protocol optimization, and the data analysis is timeconsuming (it involves manual validation of data). Nevertheless, this methodology allowed us to map the changes in conformational dynamics in PFO upon interaction with the membrane.
Our data underscore the role of the D1 domain in orchestrating the structural transitions leading to pore formation. In addition, we have established an analytical platform enabling further studies of different mutants that will eventually lead to elucidation of the molecular mechanisms covering the full pathway of events leading from monomers in solution to pores. The energy levels and energy barriers shown in the diagram are speculative and are arranged for illustrative purposes only to best account for major results obtained in this work. These results indicate the role of D1 domain in orchestrating the structural transition. In PFO, a stable D1 enforces strain upon neighboring D2 and D3 domains that is transmitted by transdomain ␤-strands. In solution, this strain is not sufficient to break their interactions. Membrane binding and concomitant oligomerization increase the stability of D1, triggering the structural transition and separation of D2 and D3 domains either by decreasing the free energy barrier, stabilizing the final state with separated D2 and D3 domains, or both. In the mutant, due to loss of stability of the D1 domain, the D2-D3 interaction is not strained, and oligomerization cannot affect the energy levels of different structural states, so the state with separated D2 and D3 domains is no longer favored.