Functional Domains of a Pore-forming Cardiotoxic Protein, Volvatoxin A2*

Volvatoxin A2 (VVA2), a novel pore-forming cardiotoxic protein was isolated from the mushroom Volvariella volvacea . We identified an N-terminal fragment (NTF) (1–127 residues) of VVA2 as a domain for oligomerization by limited tryptic digestion. On preincubation of NTF with VVA2, NTF was found to inhibit VVA2 hemolytic activity by inducing VVA2 oligomerization in the solution in the same manner as liposomes. By site-di-rected mutagenesis, the amphipathic (cid:1) -helix B of NTF or VVA2 was shown to be indispensable for its biological functions. Interestingly, at a molar ratio of recombinant NTF (reNTF)/VVA2 as low as 0.01, reNTF was able to inhibit VVA2 hemolytic activity and induce VVA2 oligomerization. This indicates that reNTF can trigger VVA2 oligomerization by a seeding effect. Furthermore, the recombinant C-terminal fragment (128–199 residues) was found to be a functional domain that mediates the membrane binding of VVA2. We found a fragment localized at the C-terminal half of VVA2 containing (cid:2) 6, -7, and -8, which is protected from protease digestion because of its insertion of a membrane. We also identified a putative heparin binding site (HBS) located in the VVA2 C terminus (166–194 residues), which Model of cytolysis induced by VVA2: membrane binding, oligomerization, and pore formation. A , VVA2 appears to expose its amphipathic (cid:1) -helix B responsible for the VVA2 oligomerization upon interacting with NTF, a truncated form of VVA2 lacking the CTF shielding of amphipathic (cid:1) -helix B. This change in VVA2 conformation leads to its oligomerization. These observations suggest that the amphiphilicity of (cid:1) -helix B provides the seed that is responsible for an accelerated aggregation of the soluble VVA2. monomers of VVA2 first bind to the cell membrane by its CTF, and then form oligomers by its amphipathic (cid:1) -helix B of NTF, followed by (cid:2) -barrel insertion of CTF and pore formation.

protein that causes cardiac arrest by activation of Ca 2ϩ -dependent ATPase activity and inhibition of Ca 2ϩ accumulation in the microsomal fraction of the sarcoplasmic recticulum of the ventricular muscle (3). VVA is composed of VVA1 and VVA2, of which only VVA2 is endowed with hemolytic and cytotoxic activity (1). Our previous studies showed that VVA2 consists of 199 amino acid residues without cysteine residues, and that it forms pores in response to its oligomer formation in human erythrocyte membranes and liposomes. This toxic protein also permeabilized the cell membrane and allowed the passage of 70-kDa dextran rhodamine B into cultured Xenopus myocytes. Furthermore, conformational changes occurred when VVA2 interacted with liposomes (2).
Naturally occurring hemolytic toxic proteins can be classified into three groups based on their mechanisms of lysis of red blood cells: (a) pore forming, (b) enzymatic degradation of membrane lipids, and (c) solubilizing cell membrane by a detergentlike action (4). Several hemolytic toxic proteins isolated from microorganisms are pore-forming proteins that require oligomerization and specific membrane targets for pore formation such as cholesterol, integrin, and glycophosphatidylinositol (5)(6)(7). They can be divided into ␣ pore-forming toxins or ␤ poreforming toxins according to the insertion structure of ␣-helices or a ␤-barrel (8).
Two models of pore formation by the toxic protein have been proposed: (a) Palmer et al. (9) proposed an assembly model for streptolysin O, a member of the family of cholesterol-dependent cytolysins, in which the pore formation is initiated by insertion of a small oligomeric complex of streptolysin O into the membrane and then enlarged in a progressive manner by the addition of soluble toxin monomers; (b) Hotze et al. (10) proposed a pre-pore model for perfringolysin O in which the formation of oligomeric pre-pore complex precedes its membrane insertion.
The relationships between structure and function of poreforming toxins have been studied extensively. The ␣-hemolysin produced by Staphylococcus aureus is divided into three domains: the rim, cap, and stem responsible for its membrane binding, oligomerization, and insertion, respectively (11). Perfringolysin O, another member of the cholesterol-dependent cytolysins family, the function of membrane binding is assigned to domain 4, oligomerization to domain 1, and insertion to domain 3 (5,12,13).
This report describes the structure and function of a poreforming protein VVA2. We demonstrated that VVA2 contains a site involved in heparin binding on the cell membrane (HBS, FIG . 1. VVA2 sequence. A, nucleotide and deduced amino acid sequence of VVA2 cDNA. The open reading frame consists of 199 amino acids and a putative signal peptide of 18 amino acids. The first amino acid and nucleotide sequence of VVA2 are boxed. A stop codon that maps at nucleotide position 652 is denoted by an asterisk. The underlining in the nucleotide sequence indicates a potential polyadenylation site. Nucleotides and amino acid residues are numbered on the left and right, respectively. The primers used in rapid amplification of cDNA ends-PCR are marked by arrows. The amino acid sequence deduced from cDNA is shown in one-letter code. B, the amino acid sequence of VVA2 was aligned with that of CytB isolated from B. thuringiensis. The secondary structural elements of VVA2 predicted from the PROFsec program are illustrated at the top of the sequence (pink) and those of CytB (green) from the crystal structure are shown below; the arrows represent ␤-strands, and the rods represent ␣-helices. The secondary structural elements of VVA2 with PROF scores below 5 are shown in light pink.
166 -194 residues), which is conserved among cardiotoxins isolated from snake venoms (14,15). The amphipathic ␣-helix B (78 -86 residues) of the N-terminal fragment (NTF, 1-127 res-idues), either as an isolated fragment, NTF, or in the intact VVA2 molecule, was shown to be crucial for VVA2 oligomerization. A C-terminal fragment (CTF, 128 -199 residues) with a pI FIG. 2. Heparin binding site of VVA2. A, alignment of the amino acid sequence of VVA2-HBS with the heparin binding domains of 10 snake cardiotoxins (CTXs). CTX Tg represents toxin g from Naja Naja nigricollis venom, whereas CTX A and CTX M are from Naja Naja atra and Naja mossambica mossambica venom, respectively. Conserved amino acid residues among those toxins are surrounded by boxes. Asterisks indicate the residues might be involved in the heparin binding of VVA2. B, affinity column chromatography of VVA2. VVA2 (100 g) was applied to a heparin affinity column (0.7 ϫ 2.5 cm) that was pre-equilibrated with 10 mM sodium acetate buffer, pH 4. The column was washed with the same buffer and then eluted with the same buffer containing a linear gradient of 0 to 0.5 M NaCl. The absorbance of elution fractions was measured at 280 nm. Each 1-ml fraction was collected. The dashed line in the panel represents the NaCl gradient from 0 to 0.5 M. C, heparin inhibits the oligomer formation of VVA2 by liposomes. VVA2 (5 g) was preincubated with heparin or dextran at a molar ratio of 1:1, and then liposomes were added to the mixture. The reaction was carried out at 37°C for 30 min, and then terminated with SDS sample buffer. The reaction products were analyzed by SDS-PAGE, and the proteins were visualized by silver staining. value of 9.6 was shown to mediate its membrane binding. Moreover, the insertion domain of VVA2 was identified by its transmembrane ␤-barrel in CTF. Altogether, our results provide insight supportive of the pre-pore model as the cytolytic mechanism of VVA2.

EXPERIMENTAL PROCEDURES
Materials Taq-DNA polymerase and pGEM-T vector were purchased from Promega (Madison, WI). Restriction endonucleases, T4 DNA ligase, pTYB2 vector, and chitin beads were from New England Biolabs (Beverly, MA). The Marathon cDNA amplification kit was from Clontech (Palo Alto, CA). 1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine and cholesterol were purchased from Avanti Polar Lipids (Alabaster, AL). Heparin (porcine intestinal mucosal), dermatan sulfate (porcine intestinal mucosal), chondroitin sulfate (shark cartilage), dextran, and glutathioneagarose beads were from Sigma. HiTrap heparin column, CNBr-activated Sepharose 4B, CM5 or L1 sensor chips, the amine coupling kit, and pGEX-2T vector were from Amersham Biosciences. The Rapid Translation System RTS 500 Escherichia coli circular template kit and complete protease inhibitor mixture were from Roche Applied Science. VVA2 was isolated from the fruit body of Volvariella volvacea, grass mushrooms, as described previously (1). All other chemicals were of analytical grade.
Preparation of Lipid Vesicles-To obtain multilamellar vesicles for membrane-bound proteolysis analysis, cholesterol was mixed with 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (1:2, mol/mol) in chloroform, dried under a stream of nitrogen, and evaporated in a rotary evaporator. The dry lipid film was hydrated and dispersed by Hepesbuffered saline (10 mM Hepes, 150 mM NaCl), pH 7.4, to give a final lipid concentration of 10 mol/ml. Small unilamellar vesicles for BIAcore assay were extruded 15 times through a 50-nm polycarbonate filter, sonicated by a sonifier, centrifuged at 108,000 ϫ g for 25 min at 15°C, and further spun at 60,000 ϫ g for 3 h to remove large unilamellar vesicles. Liposome was quantitated by the method of Bartlett (16).
Proteolysis of VVA2 and Liposome-bound VVA2-Limited protease digestion was carried out by treating VVA2 (30 g) with N-tosyl-Lphenylalanine chloromethylketone (TPCK)-trypsin or N ␣ -p-tosyl-L-lysine chloromethylketone chymotrypsin (Sigma) in phosphate-buffered saline (PBS) at room temperature for 1 h, using an enzyme/substrate ratio (w/w) of 0.1. The reaction was terminated by the addition of SDS sample buffer, and then boiling at 95°C for 5 min. Finally, the reaction products were analyzed by 15% SDS-PAGE and visualized by Coomassie Blue staining. The proteolytic fragment was isolated from the gel with a gel elutor (model 422 Electro-Elutor, Bio-Rad) and its molecular weight was determined by electrospray ionization mass spectrometry (ESI-MS) (Finnigan MAT LCQ).
Proteolysis of membrane-bound VVA2 was performed as described by Du et al. (17) with minor modifications. Briefly, VVA2 (50 g) was preincubated with multilamellar vesicles in phosphate-buffered saline (PBS) at 37°C for 1 h, and then the VVA2-liposome complex was washed three times with PBS to remove unbound VVA2. For protease K, Pronase, and subtilisin digestion, liposome-bound VVA2 was treated with the proteases at a ratio of 1:5 (w/w, liposome bound VVA2:protease) at 37°C for 1 h. After washing the digested VVA2-liposome complex three times with PBS, the membrane-bound fragments were analyzed by 16.5% Tricine SDS-PAGE (18). The peptides were transferred to polyvinylidene difluoride membranes and the amino acid sequence of the peptide was determined with an ABI 476 A sequencer.
cDNA Cloning of VVA2 and Construction of Expression Vectors for the Various Mutants of VVA2-The poly(A) mRNA was isolated from the total cellular RNAs on an oligo(dT)-cellulose column (19), and poly(A)-rich mRNA was reverse transcribed with the Marathon cDNA amplification kit. The cDNAs were ligated to Marathon cDNA adaptors for 5Ј and 3Ј rapid amplification of cDNA ends, and used as the template for the subsequent PCR amplification. For the first PCR, VVA2 cDNA was amplified with sense adaptor primer-1 (AP-1) and antisense degenerate primer A, corresponding to amino acids 137-143. VVA2 cDNA was then amplified by PCR with sense degenerate primer B corresponding to amino acids 123-129 and antisense primer adaptor primer-1. The DNA sequence of the products obtained from the first and second PCR steps was used to design gene-specific primers 1 and 2, which correspond to the first eight N-terminal and last eight C-terminal amino acids, respectively. The VVA2 cDNA was obtained by PCR amplification with sense gene-specific primer 1 and antisense gene-specific primer 2 (Fig. 1A).
NTF and CTF constructs were obtained by PCR using VVA2 cDNA as a template and the 5Ј primers were designed in-frame with the NdeI site. NTF ␣-helix B mutant I82E/L86K (NTF I82E/L86K), and VVA2 I82E/L86K constructs were generated by PCR site-directed mutagenesis (20). The double cleaved fragments obtained by digestion with NdeI and SmaI were ligated to a pTYB2 vector for expression of NTF or NTF I82E/L86K in E. coli or a pIVEX 2.3 MCS vector for expression of VVA2 I82E/L86K or CTF in the E. coli cell-free system. The VVA2 heparin binding site fragment (HBSF) (165-199 residues) expression vector was constructed as follows. HBSF was first amplified by PCR with the 5Ј primer behind a BamHI and the 3Ј primer with a stop codon following an EcoRI, and the PCR product was inserted into the BamHI-EcoRI site of the pGEX-2T vector for expression.
Expression of Various Mutants of reVVA2-For expression of recombinant NTF (reNTF) and reNTF I82E/L86K, the expression vectors were transformed into E. coli strain ER2566, and grown aerobically in Luria-Bertani medium at 37°C. When the density of culture reached A 600 nm ϭ 0.7, isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 0.3 mM, and the growth was continued with shaking at 15°C for 16 h. The cells were lysed by ultrasonication with lysis buffer containing 20 mM Tris, pH 8, 0.5 M NaCl, 1 mM EDTA, 0.1% Tween 20, and complete protease inhibitor mixture. After centrifugation at 39,191 ϫ g at 4°C for 30 min, the supernatant was incubated with chitin beads, and the on-column bound proteins were cleaved with 20 mM Tris, pH 8.0, buffer containing 0.5 M NaCl, 1 mM EDTA, and 50 mM dithiothreitol at 4°C for 16 h (21). The protein fractions were further purified on a Superdex 75 column (1 ϫ 30 cm) equilibrated with PBS using a FPLC system (Amersham Biosciences).
To avoid cytotoxicity or formation of inclusion bodies of expression proteins, the expression of reVVA2 I82E/L86K or reCTF in vitro was performed with a Rapid Translation System RTS 500, which utilizes an enhanced E. coli lysate for an in vitro transcription/translation reaction (22). Reactions were carried out at 28°C for 24 h and then the reaction mixtures were centrifuged at 20,000 ϫ g for 30 min. The supernatant was adjusted to 300 mM NaCl and 1 mM imidazole, and then 0.5 ml of a 50% slurry of nickel-nitriloacetic acid-agarose (Qiagen, Germany) was added to the mixtures, and incubated by rotating at 4°C for 6 h. The reaction products were loaded onto the column (10 ϫ 5 mm), washed with 10 column volumes of wash buffer containing 15 mM imidazole, 300 mM NaCl, and 50 mM sodium phosphate, pH 7.5, and then eluted with elution buffer consisting of 50 mM sodium phosphate, 150 mM imidazole, and 300 mM NaCl.
To prepare reHBSF, the expression plasmid was transformed into E. coli strain TG1. After growth to the density A 600 nm ϭ 0.5, the cells were induced with 0.01 mM isopropyl-1-thio-␤-D-galactopyranoside, and the culture was further incubated for 3 h. The glutathione S-transferase-HBSF fusion protein was then subjected to affinity purification with glutathione-agarose beads, and the purified fusion protein was digested with thrombin at room temperature for 16 h. The reaction products were purified with a glutathione-Sepharose 4B column (10 ϫ 5 mm). The homogeneity of HBSF was analyzed by SDS-PAGE and amino acid sequence analysis. All protein concentrations were measured by the bicinchoninic acid method (23).
Hemolysis Assay-Human red blood cells were diluted to a concentration of 2 ϫ 10 8 cells/ml in PBS, and 0.2 ml of reaction mixtures containing 0.1 ml of washed red blood cells and 0.1 ml of various concentrations of VVA2, reVVA2 I82E/L86K, reNTF, reNTF I82E/ L86K, or reCTF were incubated at 37°C for 30 min. The supernatants were collected by centrifugation at 13,000 ϫ g for 5 min, and hemolytic activity was quantitated using 96-well tissue culture plates and measuring the absorbance of the supernatant at 540 nm (24). 100% hemolysis was defined as the same volume of human red blood cell solution in the presence of 0.1% Triton X-100.
GAGs Affinity Chromatography-Affinity column chromatography was performed on a column (0.7 ϫ 2.5 cm) containing 1 ml of various sulfated glycoaminoglycan (GAG) gels at 4°C. VVA2 (100 g) or reHBSF (200 g) was applied onto the column pre-equilibrated with 10  (43), and proteins from a HiTrap heparin columns, CS or DS column, were eluted with NaCl. b VVA2 or reHBSF was equilibrated with 10 mM sodium acetate, pH 4 or 10 mM sodium phosphate, pH 7.5, respectively. mM sodium acetate, pH 4, or 10 mM sodium phosphate, pH 7.5, respectively. The column was washed with the equilibration buffer described above until absorbance at 280 or 212 nm of the eluate returned to baseline, and the column was then eluted with a linear gradient of 0 -0.5 M NaCl at a flow rate of 1 ml/min using an Ä KTA purifier system (Amersham Biosciences).
Binding Analysis by Biosensor-Real time protein-protein or proteinlipid interactions were examined using a BIAcore X instrument (BIA-coreAB, Uppsala, Sweden). For protein-protein interactions, immobilization of VVA2 on the surface of the CM5 sensor chip was performed following the standard amine coupling procedure with minor modifications. Briefly, the chip surface was first activated by injection of 75 The lipid bilayers were formed in a hydrophobic chip L1 that is suitable for the study of hydrophobic interactions (25). 0.5 mM Cholesterol/1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (1:2, mol/mol) liposomes (small unilamellar vesicles) in 10 mM Hepes-buffered saline, pH 7.4, was injected at a flow rate of 2 l/min for 10 min and the lipid layer was washed with a brief injection of 20 mM NaOH at an increased flow rate to 100 l/min for 1 min to wash off the unbound phospholipids. Then, 0.1 mg/ml bovine serum albumin was used to block nonspecific binding at a flow rate of 50 l/min for 1 min. Various concentrations of reVVA2 I82E/L86K, reNTF, or reCTF over a range of 3 to 0.125 M in 10 mM Hepes-buffered saline, pH 7.4, were injected for 5 min at a flow rate of 20 l/min.
The sensor chips were regenerated after each cycle of measurement by injecting 50 l of 20 mM NaOH. Sensorgrams were then subtracted from control sensorgrams obtained with reference surface to yield true binding responses. All sensorgrams were analyzed by using BIAevaluation software, version 3.0. Kinetic constants were obtained from the association and dissociation curves generated from various concentrations of analytes using the 1:1 binding model (A ϩ B ϭ AB) available in the software.  4. Requirement of amphipathic ␣-helix B for VVA2 oligomerization. A, amphipathic ␣-helix B of NTF is required because of its inhibitory effects on VVA2 hemolytic activity. VVA2 (0.2 g) and reNTF or reNTF I82E/L86K were pre-mixed at various molar ratios as indicated and the percentage of hemolysis was calculated as described under "Experimental Procedures." Each value represents the mean Ϯ S.D. of three independent experiments. B, amphipathic ␣-helix B of NTF is required because of triggering of VVA2 oligomerization. The formation of various oligomers of VVA2 triggered by various molar ratios of VVA2/reNTF, as indicated, but not by VVA2/reNTF I82E/L86K at a 1:1 molar ratio was analyzed by 12% SDS-PAGE and visualized by silver staining. 5 g of VVA2 was used in these experiments. C, amphipathic ␣-helix B is required for interactions between NTF and VVA2, which were studied by SPR analysis. Overlay of sensorgrams for the interactions of immobilized VVA2 with each reNTF or reNTF I82E/L86K at the same concentration of 375 nM is shown. a, reNTF; b, reNTF I82E/L86K. Every injection at a flow rate of 20 l/min was started at 100 s, stopped at 400 s, and followed by dissociation in HBS-EP buffer for 300 s. The amounts of interacting reNTF or reNTF I82E/L86K as a relative response, in surface plasmon resonance units (RU), are shown on the y axis; the x axis is the flow time in seconds (s). The results for the non-immobilizing control were subtracted from each sensorgram. D, amphipathic ␣-helix B of full-length VVA2 is required for VVA2 oligomerization by liposomes. VVA2 (5 g) or reVVA2 I82E/L86K (2.5 g) in the absence or presence of liposomes at 1:10 molar ratio as indicated were analyzed by SDS-PAGE and visualized by silver staining.

Molecular Cloning and Structural Characteristics of VVA2-
The cloned cDNA contained the coding region of VVA2 with 651 nucleotides, and the amino acid sequence deduced from the cDNA nucleotide sequence was identical to that previously determined by protein sequencing (Fig. 1A) (2). A BLAST search of the protein data bank revealed that partial homology (48.7%) exists between VVA2 and CytB isolated from Bacillus thuringiensis (26). CytB consists of an outer layer of helix hairpins flanked by several ␤-strands, which were suggested for membrane insertion (17,26,27). The results of alignment of the predicted secondary structure elements of VVA2 by the PROFsec program in the PredictProtein server network systems 2 (28,29) with those of CytB from the crystal structure are shown in Fig. 1B.

FIG. 5. VVA2 binds to cell membrane by its CTF.
A, binding of VVA2 or its various mutants on human red blood cells was carried out as described previously (42). 5 g of VVA2 or its various mutants were incubated with human red blood cells at 37°C for 30 min. The reaction mixtures were then centrifuged at 20,000 ϫ g, and the supernatant and pellets were subjected to Western blot analysis. reVVA2 I82E/L86K or reCTF were visualized with an anti-His antibody and VVA2 or reNTF with anti-VVA2 or anti-NTF polyclonal antibodies, respectively. S represents the supernatant fraction and P the membrane fraction. Control corresponds to red blood cells without adding VVA2 or its mutants.  a The association and dissociation constants were derived from the raw data of different sensorgrams using a 1:1 binding model (A ϩ B ϭ AB) included with BIAevaluation software, version 3.0.
Gly(X 6 )-Lys-(X 12 )-Val-Lys-Lys-Gly 194 -, which is similar to that of 10 kinds of cardiotoxin, -Lys 12 -Thr-(X 3 )-Gly-(X 5 )-Lys-(X 10 )-Val-Lys-Arg-Gly 37 -, in that eight amino acid residues were conserved among them (14,30) (Fig. 2A). The binding of GAGs to VVA2 or reHBSF was assessed by affinity column chromatography, and the results indicated that both VVA2 and reHBSF bound to the HiTrap heparin column. VVA2 was eluted at 0.27 M NaCl and reHBSF at 0.12 M NaCl (Fig. 2B, Table I). Similar results were obtained when chondroitin sulfate and dermatan sulfate-Sepharose columns were used (Table I). In addition, preincubation of VVA2 with heparin abolished VVA2 oligomerization by liposomes, whereas dextran did not (Fig. 2C). Because cardiac myocytes contain a higher amount of GAGs on the cell membrane, it suggests that the HBS located at ␤-strands 7, 8, and ␣-helix D of VVA2 plays an important role in its cardiotoxicity (15).
Domain for VVA2 Oligomerization-To study the structurefunction relationships of VVA2, the toxic protein was digested with TPCK-trypsin. A fragment (Fig. 3A, lane 2) with a molecular mass of 14,236 Da analyzed by ESI-MS was obtained (data not shown), and the cleavage site was found to be at Lys 127 , which was designated as NTF, whereas the fragment 128 -199, was CTF (Fig. 3B). A fragment was also obtained by TLCKchymotrypsin digestion (Fig. 3A, lane 4) and the cleavage site was found to be at Trp 132 (data not shown). These protease accessible sites are located at the loop between ␤-5 and -6, and ␤-6 strand, respectively (Fig. 1B), corresponding to their location at the surface of CytB (26).
Interestingly, we found that one molecule of reNTF completely inhibits the hemolytic activity of 100 molecules of VVA2 (Fig. 4A). In addition, SDS-PAGE showed that reNTF can induce VVA2 oligomerization at the same level of concentration required to inhibit VVA2 hemolytic activity (Fig. 4, A and B).
These results indicate that reNTF triggers VVA2 oligomerization by a seeding mechanism and prevents it from causing hemolysis.
Amphipathic ␣-Helix B Required for VVA2 Oligomerization-To determine which motif in NTF is required for VVA2 oligomerization, the secondary structure of NTF was analyzed. It was revealed that there is an amphipathic ␣-helix B with a hydrophobic moment calculated to be 0.7 according to the method of Eisenberg et al. (31). We used site-directed mutagenesis to generate the mutant, NTF I82E/L86K, with a hydrophobic moment of 0.2. The results showed that at the molar ratio of reNTF I82E/L86K/VVA2, which was 100-fold higher than that of wild type reNTF, the reNTF I82E/L86K neither inhibited VVA2 hemolytic activity nor induced its oligomerization (Fig. 4,  A and B, lane 10, respectively). This suggests that the amphipathic ␣-helix B of NTF is a functional motif required for the induction of VVA2 oligomerization and subsequent inhibition of VVA2 hemolytic activity.
To obtain insight into whether direct interaction between NTF and VVA2 is required for facilitation of VVA2 oligomerization, real-time detection of protein-protein interaction by SPR analysis was performed. To compare the interaction of reNTF and reNTF I82E/L86K with VVA2, each fragment was injected at the concentration of 375 nM over immobilized VVA2, and binding was measured over time. Under these conditions, reNTF yielded a much stronger binding than reNTF I82E/ L86K (Fig. 4C). To further characterize the binding of reNTF or reNTF I82E/L86K, a series of concentrations of each of these proteins was injected over immobilized VVA2 to obtain the kinetic parameters. Table II shows that reNTF had a faster association rate (k a ϭ 56.7) than reNTF I82E/L86K (k a ϭ 5.67), and the binding of reNTF to VVA2 (K D ϭ 102 nM) appeared to be tighter than that of reNTF I82E/L86K (K D ϭ 595 nM). The binding strength was consistent with the capacity to induce oligomerization, leading to different degrees of inhibition on VVA2 hemolytic activity between these two proteins (Fig. 4, A  and B). These results indicate that reNTF interacts directly with VVA2 through its amphipathic ␣-helix B, and hence induces VVA2 oligomerization.
Study of the effects of amphipathic ␣-helix B in a full-length VVA2 on its functions showed that reVVA2 I82E/L86K, unlike VVA2, lacked both hemolytic activity (Fig. 4A) and oligomerization activity in the presence of liposomes (Fig. 4D, lane 2 and  4). This suggests that the amphipathic ␣-helix B in a fulllength VVA2 is indispensable for its cytolytic activity, which is dependent on its oligomerization on cell membrane.
To assess whether mutations in amphipathic ␣-helix B affect their biological function by misfolding of proteins, circular dichroism (CD) measurement was performed. The results showed that the CD spectrum of reNTF is similar to that of reNTF I82E/L86K, and that of VVA2 is also similar to reVVA2 I82E/ L86K (data not shown). These findings indicate that the expressed mutants were properly folded as wild type proteins.  The association and dissociation constants were derived from the raw data of different sensorgrams using a 1:1 binding model (A ϩ B ϭ AB) included with BIAevaluation software, version 3.0. b ND, not determined due to loss of binding.
Domains of VVA2 for Membrane Binding and Insertion-To examine the membrane binding domain of VVA2, the binding of VVA2, reVVA2 I82E/L86K, reCTF, or reNTF to human red blood cells was analyzed by Western blotting. Fig. 5A shows that reVVA2 I82E/L86K and reCTF bound to cell membrane, whereas reNTF did not. The results indicate that CTF is the region responsible for membrane binding and that it has the same membrane binding activity as that of full-length VVA2 or VVA2 I82E/L86K. To further examine the kinetic constants of membrane binding of reNTF, reCTF, and reVVA2 I82E/L86K, BIAcore surface plasmon resonance study was performed. Typical sensorgrams of various concentrations of reVVA2 I82E/ L86K and reCTF binding to lipid bilayer are shown in panels a and b of Fig. 5B, respectively. The signal intensity in surface plasmon resonance units increased as a function of reVVA2 I82E/L86K or reCTF concentrations, suggesting that the amounts of protein bound to the lipid bilayer were proportional to protein concentrations. However, reNTF did not bind to the lipid bilayer (panel c in Fig. 5B). Detailed binding parameters for reVVA2 I82E/L86K or reCTF are summarized in Table III. The results indicate that CTF, but not NTF, appears to be necessary for membrane binding.
To identify the membrane insertion domain of VVA2, proteolysis of liposome-bound VVA2 was carried out with protease K, Pronase, or subtilisin. A major fragment with a molecular weight of about 10,000 was obtained (Fig. 6). Amino acid sequencing revealed that the major cleavage site was at Glu 111 for subtilisin digestion and Gln 120 for protease K or Pronase digestion (data not shown). These results indicate that these fragments contain three ␤-strands (␤6, -7, and -8) and that the N terminus Glu 111 and Gln 120 of VVA2 are membrane-embedded and thus are resistant to proteolytic digestion (Fig. 1B). DISCUSSION VVA2 is a novel pore-forming cardiotoxic protein. We identified two domains of VVA2 responsible for its membrane binding, oligomerization, and membrane insertion. In the cholesterol-dependent cytolysins family, it has been shown that the C-terminal domain is responsible for membrane attachment and contains the longest conserved tryptophan-rich sequence motif, a situation similar to equinatoxin II, a well known eukaryotic pore-forming toxin (12,32). Here, we used Western blot and SPR analyses to show that VVA2 binds to cell membranes by its CTF. Consistently, CTF also contains two tryptophan residues at 169 and 170, suggesting their role in membrane attachment (Fig. 1B). In addition, the pI value of CTF was found to be 9.6, whereas that of NTF was 4.0, indicating that CTF has a higher affinity to the anionic surface of general cell membranes (33). Furthermore, the fact that CTF or VVA2 I82E/L86K carries functional sites for VVA2 membrane binding but not for its oligomerization suggests that binding of toxin to membrane is the first step that precedes its oligomerization.
GAGs have been shown to be a potential targets for cobra cardiotoxin with high affinity and specificity through a cationic belt at the concave surface of the polypeptide (14,15). We demonstrated that the binding site of VVA2 with GAGs was located at ␤-strands 7, 8, and ␣-helix D by affinity column chromatography, and preincubation of VVA2 with heparin prevented VVA2 oligomerization by liposomes. This provides evidence of interaction of the VVA2-GAGs complex where the cationic belt of the conserved residue in HBS initiates ionic interaction with GAGs followed by binding of VVA2 on the cell membranes of cardiac myocytes. These results suggest that heparin may constitute a potential target for VVA2 and help FIG. 7. Model of cytolysis induced by VVA2: membrane binding, oligomerization, and pore formation. A, VVA2 appears to expose its amphipathic ␣-helix B responsible for the VVA2 oligomerization upon interacting with NTF, a truncated form of VVA2 lacking the CTF shielding of amphipathic ␣-helix B. This change in VVA2 conformation leads to its oligomerization. These observations suggest that the amphiphilicity of ␣-helix B provides the seed that is responsible for an accelerated aggregation of the soluble VVA2. B, monomers of VVA2 first bind to the cell membrane by its CTF, and then form oligomers by its amphipathic ␣-helix B of NTF, followed by ␤-barrel insertion of CTF and pore formation. the toxin exert its cardiotoxic effect.
Studies on CytB by x-ray crystallography (26), protease digestion in model membrane (17), and fluorescence technique (27) support that the long amphiphilic ␤-strands (␤5, ␤6, and ␤7) are responsible for membrane insertion in the pore formation of CytB. In this study, the results of protease digestion of membrane-bound VVA2 showed that the proteolysis-resistant region of VVA2 encompasses the amphipathic ␤-strands, being corresponding to those of CytB. Thus, the ␤-barrel might be involved in membrane insertion, and VVA2 could be classified as a ␤-pore-forming toxin based on its ␤-barrel insertion structure (8,34).
Membrane association, protein-protein interaction, or oligomerization through amphipathic ␣-helices has been established in previous studies (35)(36)(37). The N-terminal amphipathic ␣-helix of equinatoxin II participates in membrane binding and pore formation (32), whereas the amphipathic ␣-helix B of NTF or VVA2 is crucial for VVA2 oligomerization. In this study, we found that amphipathic ␣-helix B plays an important role in self-interacting and the triggering of VVA2 oligomerization. However, the observation that VVA2 is a monomer in solution, as demonstrated by size-exclusion chromatography or the results obtained by SPR analysis in which no interaction was observed between VVA2 or reVVA2 I82E/L86K and immobilized VVA2 (data not shown), indicates that amphipathic ␣-helix B may be inaccessible in the absence of NTF or liposomes. NTF, like liposomes, can induce a conformational change of VVA2 as measured by intrinsic emission fluorescence spectroscopy (data not shown). This suggests that VVA2 exposes the amphipathic ␣-helix B in the presence NTF or liposomes. Thus, this change in VVA2 conformation may represent the first of a series of events leading to VVA2 oligomerization and subsequent insertion into membrane. Based on the very low ratio of reNTF/VVA2 available to facilitate VVA2 oligomerization, the amphipathic ␣-helix B may provide the seeds that are responsible for an accelerated aggregation of soluble monomeric VVA2. This interesting phenomenon could occur in nature. The truncated proteins produced by deletion mutation might induce the oligomerization of wild type proteins, leading to dysfunction of affected cells. The induction of oligomerization in many proteins has been reported to play important roles in exerting their individual biological functions or producing infectious material, such as ␣-synuclein (38), proapoptotic protein, Bid (39), and amyloid ␤-protein (40). Because VVA2 oligomerization occurred at 4 or 37°C in the presence of liposomes, while VVA2 lysed red blood cells at 37°C but not 4°C (data not shown), it is likely that VVA2 oligomerization precedes membrane insertion, a step that is temperature-dependent.
Based on the findings of this study, we propose that VVA2 comprises two functional domains. The first, CTF, is thought to bind to the cell membrane. The second, NTF, is responsible for oligomerization upon the exposure of its amphipathic ␣-helix B in response to conformational changes of VVA2 molecule. This leads to the penetration into cell membrane by ␤-strands 6, 7, and 8 located in CTF.
In summary, based on the findings of this study we propose the model cytolysis induced by VVA2 illustrated in Fig. 7B. VVA2 is a monomer in solution in the absence of membrane. VVA2 first binds on the cell membrane by its CTF and then forms oligomers by its amphipathic ␣-helix B. This results in permeabilization of the cell membrane by insertion of ␤-barrels of CTF and induces a non-colloid-osmotic cytolysis (41). Because the amphipathic ␣-helix B of reNTF appears to be exposed to trigger VVA2 oligomerization, it functions as the seed for the formation of the large pre-pore complex (Fig. 7A) prior to the insertion of the transmembrane ␤-barrel, favoring the pre-pore model of VVA2 function.