C-terminal Amino Acid Residues Are Required for the Folding and Cholesterol Binding Property of Perfringolysin O, a Pore-forming Cytolysin*

Perfringolysin O (θ-toxin) is a pore-forming cytolysin whose activity is triggered by binding to cholesterol in the plasma membrane. The cholesterol binding activity is predominantly localized in the β-sheet-rich C-terminal half. In order to determine the roles of the C-terminal amino acids in θ-toxin conformation and activity, mutants were constructed by truncation of the C terminus. While the mutant with a two-amino acid C-terminal truncation retains full activity and has similar structural features to native θ-toxin, truncation of three amino acids causes a 40% decrease in hemolytic activity due to the reduction in cholesterol binding activity with a slight change in its higher order structure. Furthermore, both mutants were found to be poor at in vitro refolding after denaturation in 6 m guanidine hydrochloride, resulting in a dramatic reduction in cholesterol binding and hemolytic activities. These activity losses were accompanied by a slight decrease in β-sheet content. A mutant toxin with a five-amino acid truncation expressed in Escherichia coli is recovered as a further truncated form lacking the C-terminal 21 amino residues. The product retains neither cholesterol binding nor hemolytic activities and shows a highly disordered structure as detected by alterations in the circular dichroism and tryptophan fluorescence spectra. These results show that the C-terminal region of θ-toxin has two distinct roles; the last 21 amino acids are involved to maintain an ordered overall structure, and in addition, the last two amino acids at the C-terminal end are needed for protein folding in vitro, in order to produce the necessary conformation for optimal cholesterol binding and hemolytic activities.

Thiol-activated cytolysins (1) comprise a family of bacterial protein toxins that are produced by Gram-positive bacteria. They share a high degree of homology in their amino acid sequences (40 -70%) (2-7) and have common biological characteristics, cholesterol binding and the formation of oligomeric pores on plasma membranes. Perfringolysin O (472 amino acids), known as -toxin, is such a toxin produced by Clostridium perfringens type A. Its cytolytic mechanism is thought to comprise at least four steps: binding to cholesterol in membranes, insertion into the membrane, oligomerization, and pore formation. -Toxin binds specifically to cholesterol on plasma membranes with high affinity (K d ϳ 10 Ϫ9 M) (8). By forming oligomeric pores on plasma membranes (9), -toxin causes cell disruption.
After several attempts to crystallize -toxin (10,11), its three-dimensional structure was recently revealed by x-ray diffraction (12). This analysis showed -toxin to be an elongated rod-shaped molecule rich in ␤-sheets and to consist of four discontinuous domains. Domain 4 ( Fig. 1b) (residues 363-472), the C-terminal domain, is an autonomous structure comprising a continuous amino acid chain. Six of the seven total tryptophan residues reside in domain 4, and three are located in the sequence of ECTGLAWEWWR (residues 430 -440), the longest conserved sequence among thiol-activated cytolysins (2, 3). From many efforts to achieve mutagenesis of this toxin family (13)(14)(15)(16)(17), it was shown that all mutations that inhibit cell binding activity reside in domain 4, suggesting that some region in domain 4 binds to membrane cholesterol upon binding to cells. This is consistent with our previous findings that a C-terminal tryptic fragment that contains predominantly domain 4 binds to cholesterol and to cholesterol-containing membrane (18). Our findings that the toxin binding to cholesterol in liposomal membrane triggers a conformational change around tryptophan residues in domain 4 also support this view (19,20). Recently, possible roles of the C-terminal region in cell binding were suggested by a report that a monoclonal antibody thought to bind near the C terminus specifically blocks cell binding, although the exact epitope was not identified (21). Despite this finding, it is not known whether the C-terminal region plays a role in cholesterol binding or membrane insertion activity, inasmuch as either one could affect toxin binding to cells. Recent x-ray crystallographic analysis showed that there are two ␤-strands in antiparallel orientation in the C-terminal end and that one of them is composed of 7 amino acids in the C-terminal end (12).
Here, we constructed and analyzed toxin mutants truncated in the C terminus to define the role of the C-terminal region on cholesterol binding activity. Using an ELISA 1 assay for quantitative analysis of cholesterol binding activity, we show that the C-terminal end is essential for folding of -toxin into the native conformation, thus ensuring activities of cholesterol binding and hemolysis.
Site-directed Mutagenesis-Plasmid pNSP10 containing the perfringolysin O gene (pfoA) (13) was used to construct six pfoA derivatives encoding truncated -toxins. Stop codons were introduced at appropriate sites in the pfoA gene by a site directed mutagenesis kit (CLONTECH) based on the unique site elimination method (22). The 5Ј-deoxyoligonucleotide, dGT-GACTGGTGAGGCCTCAACCAAGTC, was used to make a unique restriction site for the selection of all mutations. Stop codon insertion was performed using the following mutagenic primers for: ⌬471, 5Ј-deoxynucleotide dCAGTTTTTACTTTAGTaTAcTatTAAGTAATACTAG; ⌬470, dCTTTA-GTTTAATTtTAtcaAATACTgGATCCAGGGT; ⌬468, dGTTTAATTGTAA-GTttatCagGATCCAGGGT. Lowercase letters represent bases changed for mutagenesis. The DNA sequences in the resulting plasmids were confirmed by means of the dideoxynucleotide chain-termination method (23). Predicted amino acid sequences in the C-terminal ends of the mutant toxins are shown in Fig. 1.
Protein Production and Purification-Protein production and purification were performed as described previously (13) with slight modifications. Escherichia coli strains BL21(DE3) and BL21(DE3) harboring pLysS (24) (Novagen, Madison, WI) were used for the overexpression of wild type -toxin and mutant toxins. Wild type -toxin and mutant toxins were purified from the periplasmic fraction by a series of DEAE-Sephacel chromatographies. In the case of mutant toxins having no hemolytic activity, the fractions eluted from the first DEAE-Sephacel column were analyzed by immunostaining with anti--toxin antibody after SDS-PAGE. Then, the toxin fractions were loaded onto a second DEAE-Sephacel column equilibrated with 20 mM BisTris, pH 6.5, and eluted with the same buffer containing 40 mM NaCl. For further purification, the toxin fractions were applied to a hydroxylapatite column equilibrated with 20 mM sodium phosphate buffer, pH 7.5, and the toxins were eluted with 100 mM sodium phosphate. Then, the toxins were loaded onto a butyl-agarose column equilibrated with 20 mM Tris-HCl, pH 7.5, containing 1.7 M (NH 4 ) 2 SO 4 and eluted with 0.5 M (NH 4 ) 2 SO 4. The purity of the toxins was checked by SDS-PAGE (25).
Determination of the Hemolytic Activity of Toxins-Hemolytic activity was determined as described previously (26). The amount of toxin required for 50% hemolysis of 1 ml of 0.5% sheep erythrocytes in 30 min at 37°C (HD 50 ) was determined using the von Krogh equation (27). The hemolytic activity of each toxin was obtained as a 1/HD 50 value and expressed relative to the wild type toxin.
Binding of Wild Type and Mutant Toxins to Sheep Erythrocytes-After activation with 10 mM dithiothreitol for 30 min at 10°C, each toxin (0.3 g) was incubated with 0.5% hematocrit sheep erythrocytes in phosphate-buffered saline, pH 7.0, containing 1 mg/ml bovine serum albumin for 20 min at 20°C. The mixture was centrifuged at 250,000 ϫ g for 20 min at 4°C, and both the pellet and supernatant fractions were analyzed by Western blotting.
Binding to Cholesterol on TLC Plates-The cholesterol binding activity of each toxin was examined by using TLC plates as described previously (13,18).
ELISA Assay-The cholesterol binding activity of each toxin was determined by ELISA using the microtiter plate (Immulon 1, Dynatech Laboratories, Alexandria, VA). The wells were coated with various concentrations of cholesterol (12 -10,000 pmol) and treated with 10 mg/ml fatty-acid-free bovine serum albumin in Tris-buffered saline for 1 h for blocking. Toxins (each 1 ng) were then added to the wells and the mixtures were incubated for 1 h. After washing with Tris-buffered saline, the mixtures in the wells were incubated with anti-(-toxin) antibody for 1 h, followed by incubation with peroxidase-conjugated anti-rabbit IgG for 1 h. Toxins bound to the cholesterol on the microtiter plates were detected by measuring the intensity at 410 nm of the color development with 2,2Ј-azino-di[3-ethyl-benzthiazoline sulfonate(6)] (Kirkegaard & Perry Laboratories Inc., Gaithersburg, MD) as a peroxidase substrate.
Susceptibility of Toxins to a Protease-Purified toxins (600 ng each) were treated at 22°C with subtilisin Carlsberg at an enzyme to substrate ratio of 1:60 (28). At appropriate times, the digestion was stopped by the addition of 1 mM phenylmethanesulfonyl fluoride. The resultant fractions were analyzed by SDS-PAGE and Western blotting by using anti-(-toxin) antibody.
Measurment of Circular Dichroism Spectra-Circular dichroism (CD) spectra were recorded using a JASCO J-720 spectropolarimeter at room temperature with 1-or 5-mm pathlength cells. Purified proteins were diluted in 10 mM phosphate buffer with or without 150 mM NaCl. Scans from 250 to 190 nm were recorded with 1-mm cells in the absence of NaCl to minimize buffer noise. Molecular ellipticity ([]) was calculated based on the mean residue weight and extinction coefficient (E 280 0.1% ) estimated as 110.8 and 1.6, respectively. The CONTIN program for secondary structure estimation was kindly provided by Dr. S. W. Provencher.
Measurement of Trp Fluorescence-Fluorescent studies were performed with a Shimadzu spectrofluorophotometer RF-5000. Emission spectra were measured in the range of 300 -400 nm with an excitation wavelength of 280 or 295 nm. Purified toxins were diluted with Hepesbuffered saline, pH 7.0, to a protein concentration of 10 g/ml.
Unfolding and Refolding of Wild type Toxin and Mutant Toxins-Wild type and mutant toxins were unfolded by treatment with 6 M guanidine hydrochloride (GdnHCl) in 20 mM Tris-HCl, pH 8.0, 150 mM NaCl at room temperature for 10 -30 min. Denaturation was confirmed by a red shift in the fluorescence emission wavelength to 350 nm at an excitation wavelength of 280 nm. Refolding was carried out by dialyzing the samples against 20 mM Tris-HCl, pH 8.0, 150 mM NaCl at 4°C for 20 h.
N-terminal Sequence-N-Terminal sequences of wild type and mutant toxins were analyzed with a Biosystems protein sequencer 476A.
Mass Spectrometry-The molecular masses of the toxins were measured by electrospray ionization mass spectrometry (ESI-MS) using a Fourier transform ion cyclotron resonance mass spectrometer BioApex47E (Bruker Instruments) equipped with an external ESI source (Analytica of Branford). Before being injected into the source by a syringe pump operated at 30 l/h, the samples were desalted on a reverse-phase high pressure liquid chromatography column (Senshu Pak C8 -1251-N) eluted with a 10 -60% gradient of acetonitrile, 0.1% trifluoroacetic acid. In some cases, the samples were prepared from the SDS-PAGE gel by the method of Nakayama et al. (29). After SDS-PAGE, the gel was washed with distilled water and stained with 0.3 M CuCl 2 for 3 min. The toxin spot was excised from the SDS-PAGE gel and destained successively in 25 mM Tris-HCl, pH 8.3, and 12.5 mM Tris-HCl, pH 8.3. Then, the toxin was extracted from the gel in 50 mM Tris-HCl, pH 8.8, containing 50 mM EDTA and 0.1% SDS. To remove SDS and other impurities, the extracted toxin was applied to a Phenyl-5PW RP column (Tosoh, Tokyo, Japan) and recovered with 80% acetonitrile in 0.1% trifluoroacetic acid. When this method was used, the molecular mass of -toxin was determined by subtracting the mass of the adduct of copper, 63.4, from the observed mass.

RESULTS
Characterization of Truncated -Toxin--Toxin mutants truncated at the C terminus were constructed and expressed in E. coli as described under "Experimental Procedures" (Figs. 1 and 2). -Toxin has an intrinsic signal sequence at its N terminus and is secreted into the periplasm when expressed in E. coli ( Fig. 2 and Ref. 13). A similar expression profile was observed for ⌬471 and amounts comparable to wild type -toxin were recovered from the periplasmic fraction (Fig. 2). A slightly smaller amount was recovered in the case of ⌬470. Upon expression of the DNA construct for ⌬468 and mutants with larger truncations, the amounts of proteins with molecular sizes close to that of intact -toxin decreased with concomitant increases in the amounts of degradation products with sizes around 38 kDa ( Fig. 2 and data not shown). The results suggest that truncations at the C terminus affect the biosynthesis of -toxin and/or its secretion into the periplasm.
To further characterize the mutant toxins, the expressed proteins were purified from the periplasmic fraction by DEAE-Sephacel column chromatographies and their molecular masses were determined by ESI-MS (Table I). The observed molecular masses of wild type, ⌬471, and ⌬470 are within the range of the predicted molecular masses. In contrast, the observed moleculer mass of the protein recovered from the cells harboring the constructed plasmid for ⌬468 is smaller than the predicted mass for ⌬468 (Table I). This indicates that the production of the protein with a five-amino acid C-terminal truncation brings about a further truncated form. N-terminal sequence analysis revealed that the product has the same N-terminal sequence as the wild type toxin. From the results of N-terminal and molecular mass analyses, we conclude that the product comprises residues 1-451 (predicted M r , 50,375.6), with a 21-amino acid truncation at the C terminus. We designate the product as ⌬452 hereafter. The elution profile of ⌬452 is different from those of wild type -toxin and two mutants, ⌬471 and ⌬470; the former eluted from DEAE-Sephacel column at 110 mM NaCl, whereas the latter two mutants eluted at 60 mM.
The relative hemolytic activities of purified wild type and three mutant toxins were determined (Fig. 3, upper part). No differences in hemolytic activity were detected between wild type -toxin and ⌬471, indicating that the deletion of two amino acids from the C terminus of -toxin does not affect hemolytic activity. In contrast, ⌬470 showed a lower hemolytic activity, 40% that of wild type, while ⌬452 showed no hemolytic activity. These results indicate that truncation of the C terminus by 21 amino acids causes a loss of hemolytic activity.
Hemolysis by -toxin involves two important steps, binding and insertion into membranes, prior to pore formation. The binding activity to cells was measured and compared among the wild type and mutant toxins (Fig. 3, lower part). ⌬471 showed high-affinity binding to sheep erythrocytes similar to the wild type, but ⌬470 showed only very weak binding. ⌬452, which has no hemolytic activity, never bound to the cells. These results show a good correlation between cell binding activity and relative hemolytic activity.
Cholesterol on plasma membranes serves as a receptor for -toxin. Fig. 4a shows the cholesterol binding activity of mutant toxins on TLC plates as detected by immunostaining with   anti--toxin antibody. ⌬471 and ⌬470 were found to bind to cholesterol on TLC plates and to specifically recognize free cholesterol but not phosphatidylcholine or esterified cholesterol. Their manner of binding was the same as that of wild type toxin, although ⌬470 shows weaker spots. On the other hand, ⌬452 did not bind to cholesterol at all. Fig. 4b shows the quantitative analysis of the cholesterol binding activity of the toxins by ELISA. ⌬471 shows an activity comparable to the wild type toxin. The activity of ⌬470 is about 40% of the wild type, while no activity could be detected for ⌬452. The results show that the cholesterol binding activity of the toxins correlates well with their cell binding and hemolytic activities. We previously reported that mutants with Trp to Phe substitutions within the tryptophan-rich consensus sequence show decreased binding affinity for erythrocytes (13). We examined the cholesterol binding activity of two such mutants, W438F and W439F, by ELISA and compared the results with the cholesterol binding activity of the C-terminal truncation mutants (Fig. 4b). Mutants with Trp to Phe substitutions show cholesterol binding activity similar to that of the wild type toxin (Fig. 4b), showing that mutations of Trp in the consensus sequence has little effect on the cholesterol binding activity. The decrease in cell binding activity of the mutants should be attributable to step(s) other than cholesterol binding. This makes a district difference from the results for the mutants with C-terminal truncations.
Effect of C-terminal Truncation on the Structure of -Toxin-The defects in the cholesterol binding activities of ⌬470 and ⌬452 can be attributed to either the deletion of cholesterolbinding sites or conformational changes around the binding sites. To assess these possibilities, we first examined the susceptibility of mutant toxins to a protease (Fig. 5). Digestion of wild type -toxin and ⌬471 by subtilisin Carlsberg produced a distinctive 39-kDa fragment assigned as the C-terminal fragment (28); a smaller amount of this fragment was detected when ⌬470 was digested. In contrast, ⌬452 was digested over time into undetectable pieces showing no distinctive bands. Trypsin digestion also produced proteolytic fragments of 28 and 25 kDa (18) from ⌬471 and ⌬470, but not from ⌬452 (data not shown). The results indicate that the secondary or tertiary structures of ⌬452 has been changed by C-terminal truncation of 21 amino acid residues.
Because six out of the seven tryptophan residues in -toxin are located in the C-terminal region (see Fig. 1), it is reasonable to measure tryptophan fluorescence in order to monitor the conformational alterations of -toxin induced by truncation. When the toxins were excited at 295 nm, no differences in the peak emission wavelength at 338 nm were detected between wild type and two mutants, ⌬471 and ⌬470 (Table II), showing that the environmental changes around the Trp residues are not significant in those two mutants. However, environmental changes around some fluorophores other than tryptophan appear to have occurred, since the mutants exhibited a red shift in the maximal emission wavelength when excited at 280 nm (Table II). On the other hand, a distinctive red shift of the maximal emission wavelength was observed for ⌬452 as compared with the wild type toxin (Table II), indicating that the environment of the tryptophan residues in these mutants is more hydrophilic than in the wild type. Simultaneously, the intensity of the tryptophan fluorescence in ⌬452 excited at 295 nm was enhanced to 3.2 times that of the wild type -toxin. The results suggest that the inactive mutant ⌬452 has significant alteration in its tertiary structure around tryptophan residues, and that this leads to the loss in hemolytic activity.
When -toxin interacts with cholesterol on dioleoylphosphatidyl choline/cholesterol liposomes, there is an increase in the intensity of the tryptophan fluorescence (19). The two truncated toxins, ⌬471 and ⌬470, also showed increases in the intensity when incubated with dioleoylphosphatidyl choline/ cholesterol liposomes. In contrast, no enhancement of fluorescence intensity was detected for ⌬452 (data not shown). There- fore, this mutant lacks an appropriate structure for interaction with cholesterol in membranes.
In order to determine whether the deletion of C-terminal amino acids affects the secondary structure of the toxin, far ultraviolet CD spectra were measured. As shown in Fig. 6a, wild type -toxin has a ␤-sheet-rich structure and the spectra of ⌬471 and ⌬470 closely resemble that of the wild type toxin. On the other hand, drastic difference was detected in the spectrum of ⌬452 as compared with the wild type toxin. A significant increase in negative ellipticity was observed at 208 nm and shorter wavelengths. The CD difference spectrum obtained by subtracting the wild type spectrum from the ⌬452 spectrum exhibited a deep minimum at 200 nm or a shorter wavelength and a shoulder at around 225 nm. This difference spectrum resembles that usually taken to indicate an unfolded conformation (30). This observation suggests a large disorder in the secondary structures of ⌬452.
Effect of C-terminal Truncation on in Vitro Refolding-The structural analysis suggests that several amino acids at the C terminus play essential roles in in vivo protein folding and/or the maintenance of protein conformation. We carried out in vitro refolding experiments on the truncated mutants to define the function of C-terminal amino acids during folding. Wild type -toxin and mutant ⌬471 (truncated by two amino acids) were denatured in 6 M GdnHCl, renatured by dialysis, and their hemolytic activities were measured. As shown in Fig. 7a  (upper part), wild type -toxin recovered 81% of full hemolytic activity while ⌬471 displayed only 13% recovery, even though ⌬471 has an activity comparable to the wild type before denaturation. The refolded ⌬471 hardly bound to sheep erythrocytes as shown in Fig. 7a (lower part), showing a good correlation with relative hemolytic activity. To investigate whether the refolded ⌬471 recognizes cholesterol, the cholesterol binding activity was measured by ELISA. The refolded ⌬471 shows much less cholesterol binding activity than native ⌬471, while the activity of the wild type toxin is not changed by the denaturation refolding treatment (Fig. 7b). Although we could not judge whether all the refolded ⌬471 molecules have lower binding affinities than in the native state or whether a small population of ⌬471 refolds to the native form with full activity, it is clear that ⌬471 easily loses its ability to bind cholesterol during the denaturation-renaturation process. A decrease in the cholesterol binding activity was also observed after denaturationrefolding of ⌬470 (data not shown).
To rule out the possibility that there might be a minor contaminating protease that cleaves the ⌬471 protein during the refolding treatment and causes it to lose activity, the relative molecular masses of native and refolded ⌬471 were determined by ESI-MS (Table I). The relative molecular masses determined for the refolded ⌬471 and native ⌬471 are the same and within the range of the predicted one (Table I), indicating that no proteolytic cleavage occurs during the refolding process. Wild type toxin also maintains its intact size during refolding treatment, as shown by the relative molecular masses before and after treatment (Table I). These results indicate that the loss of ⌬471 activity after refolding is not caused by the FIG. 6. Far-ultraviolet CD spectra of wild type and truncated mutants. a, spectra for the wild type and three truncated toxins were measured in a 5-mm pathlength cuvette at room temperature. Samples were prepared at a toxin concentration of 30 g/ml in 20 mM phosphate buffer, pH 7.0, containing 150 mM NaCl. 1, wild type (solid line); 2, ⌬471 (dotted line); 3, ⌬470 (long dashed line); 4, ⌬452 (dot-dashed line). b, comparison of the far-ultraviolet CD spectra of ⌬471 before and after denaturation-refolding. Measurements of the far-ultraviolet CD spectra were performed in a 1-mm pathlength cuvette at room temperature. Native ⌬471 (1, dot-dashed line) and ⌬471 after denaturation-refolding (2, long dashed line) were prepared at a toxin concentration of 150 g/ml in 10 mM phosphate buffer, pH 7.0.

TABLE II
Maximal emission wavelengths and the relative intensity of wild type and truncated toxins The fluorescence measurements of wild type and truncated toxins were carried out at excitation wavelengths ( ex ) of 280 and 295 nm. The data represent mean Ϯ S.E. for three independent experiments. Maximal emission wavelengths are displayed in nanometers (nm), and the maximal intensity of each mutant is shown relative to the intensity of the wild type toxin. action of protease. The above results show that even just two amino acid residues at the C terminus are involved in the correct folding of -toxin. To assess whether the inactivation of ⌬471 by denaturation-refolding is accompanied by a conformational alteration, the structural properties of wild type and ⌬471 after denaturation-refolding treatment were studied by CD and fluorescence analyses. Compared with native ⌬471, the far-ultraviolet CD spectrum of refolded ⌬471 shows a slight alteration in the secondary structure (Fig. 6b), a 3% decrease in ␤-sheet content and a concomitant increase in random coil in the refolded ⌬471 as estimated by the algorism program, CONTIN (31). In the case of the wild type -toxin, no differences were observed in the CD spectra before and after refolding (data not shown). In the maximal emission wavelength of the fluorescence spectrum, no significant changes were observed in both wild type and ⌬471. These results clearly show that the secondary and tertiary structural changes in the refolded ⌬471 are small compared with those observed for ⌬452 (Fig. 6a and Table II), suggesting that the changes in refolded ⌬471 occur in a limited region of the toxin molecule.

DISCUSSION
The crystallographic study of -toxin showed that domain 4, the C-terminal domain supposed to contain the cholesterolbinding region, has nine ␤ strands folded into a compact ␤-sheet sandwich (12). Two ␤ strands in antiparallel orientation are located within the C-terminal 20 amino residues and form a part of one ␤ sheet (Fig. 1). In this study, focusing on the two C-terminal ␤ strands, we constructed several C-terminal truncated -toxin mutants to investigate how C-terminal amino acids contribute to the folding of the protein and its toxic action.
We first demonstrated that amino acids in the C-terminal ␤ strand play an important role in correct folding of the toxin. When ⌬471 was refolded after denaturation in 6 M GdnHCl, it lost its membrane binding and hemolytic activities with the reduction in cholesterol binding activity (Fig. 7), indicating the importance of the two C-terminal amino acids for correct folding in vitro into the conformation required for cholesterol binding. However, ⌬471 showed essentially the same hemolytic activity and secondary structure as wild type -toxin. This indicates that the mutant folds into the native conformation in vivo. Taking the difference between in vivo and in vitro folding into consideration, probably chaperone-like molecules help to achieve correct folding in vivo (32,33). As shown in ⌬470, a threeamino acid truncation affects folding both in vivo and in vitro.
The truncation of five amino acids from the C terminus leads to a further truncation of the protein in host E. coli, indicating that the C-terminal ␤ strand protects the protein against proteolytic cleavage in host cells. For toxin production, we used E. coli B strain, BL21(DE3), as a host, because it lacks both lon and ompT proteases. Some other minor protease(s) in E. coli may contribute to cleaving the product during synthesis or secretion into the periplasm (34). The product, ⌬452, lacks the two ␤ strands in the C terminus and completely loses its cell binding and hemolytic activities due to its inability to recognize cholesterol (Fig. 4). It is likely that the molecular structure required for the specific binding of cholesterol molecules is absent or not correctly organized in ⌬452. Spectroscopic data indicate its partially unfolded secondary structure and an environmental alteration around tryptophan residues to a more hydrophilic and unrestricted state ( Fig. 6a and Table II). Since the elimination of the two ␤ strands from the C-terminal end causes this remarkable disorder in structure, the C-terminal two ␤ strands are suggested to play key roles in constructing the overall structure of the toxin.
There are two distinct steps in cell binding by -toxin, cholesterol recognition and membrane insertion. In this study we demonstrated that truncation of the C terminus abolishes cholesterol-recognition ability, resulting in the loss of cell binding activity. This is in contrast to mutants with Trp to Phe substitutions within the tryptophan-rich consensus sequence (residues 430 -440), which show a loss in cell binding activity despite their ability to bind cholesterol ( Fig. 4b and Ref. 13). We previously suggested that the tryptophan-substituted mutants have some deficiency in membrane insertion activity (13,20) that could cause them to lose cell binding activity. The tryptophan-rich consensus sequence locates in close proximity to one of the two ␤ sheets in domain 4, and distant from the other ␤ sheet to which the C-terminal ␤ strand belongs (Fig. 1b). The crystallographic data indicate that there are some amino residues in the proximal ␤ sheet that are possible quenchers of the fluorescence of Trp-436 and Trp-439, Trps within the consensus sequence. Thus, a change in tryptophan fluorescence intensity could be a sensitive marker for a change in three-dimensional arrangement among these Trp residues and quenchers. Since either ⌬470 or refolded ⌬471 shows no significant change FIG. 7. The effect of denaturation-refolding treatment on wild type -toxin and ⌬471. a, hemolytic and binding activities to sheep erythocytes determined before and after denaturation-refolding. Hemolytic activities (shown as bars) of the refolded wild type and mutant toxins are expressed relative to the corresponding native forms. Binding of native and refolded toxins to sheep erythrocytes was examined as described under "Experimental Procedures." After centrifugation, the resultant supernatant (S), pellet (P), and total (T) fractions were analyzed by SDS-PAGE and immunoblotting. SRBC, sheep red blood cells. b, cholesterol binding activities of native wild type (open circles), refolded wild type (closed circles), native ⌬471 (open triangles), and refolded ⌬471 (closed triangles). Each activity was determined by ELISA and is expressed relative to the activity of the native wild type toxin at 10,000 pmol of cholesterol.
in tryptophan fluorescence compared with the wild type toxin, the microenvironment around the Trp residues remains intact in these mutants. This strongly suggests that the C-terminal truncation does not affect the structural features around the tryptophan-rich consensus sequence despite its significant effect on cholesterol binding activity. Probably the site of cholesterol binding is in different region in domain 4 from that of membrane insertion.
Since the molecular mass of ⌬471 remains unchanged by the unfolding-refolding treatment (Table I), the activity loss in cholesterol binding should be ascribed to a conformational change. It is likely that the molecular structure required for the specific binding of cholesterol molecules is not correctly organized after treatment. An approximately 3% decrease in ␤sheet content in ⌬471 was detected after treatment. No change occurs in tryptophan fluorescence as discussed above. This implies that the change occurs within a limited region upon refolding of ⌬471 and that this local change in structure directly affects the cholesterol-binding site. Since the C-terminal ␤ strand interacts directly with the next ␤ strand (residues 453-460) by hydrogen bonding to form part of an antiparallel ␤ sheet ( Fig. 1 and Ref. 12), it is likely that the decrease in ␤-sheet content produced by treatment occurs near these strands. It has been reported that C-terminal truncation of pneumolysin, another cholesterol-binding cytolysin, causes a loss of cell binding activity (17). It would be interesting to know whether this loss in the cell binding activity of pneumolysin is due to the loss of cholesterol binding activity, although conformational studies and molecular mass determination of the truncated species of pneumolysin would be required to draw a conclusion.
There have been several reports showing that C-terminal residues are important for the correct folding and maintenance of the native protein conformations (35,36). Among them, -toxin is a distinct example since the deletion of only two amino acids from the C terminus out of a total of 472 residues seriously affects its folding and the maintenance of the functional conformation. We have reported chemically modified -toxin as a new probe for cholesterol (37). If the relationships between binding activity and the conformation of the C-terminal region are clarified and the minimum binding unit is identified, further design of useful probes can be realized.