Pore-Forming Activity of Alpha-Toxin Is Essential for Clostridium septicum-Mediated Myonecrosis

Abstract

Clostridium septicum alpha-toxin is a β-barrel pore-forming cytolysin that is functionally similar to aerolysin. Residues important in receptor binding, oligomerization, and pore formation have been identified; however, little is known about the activity of the toxin in an infection, although it is essential for disease. We have now shown that deletion of a small portion of the transmembrane domain, so that the toxin is no longer able to form pores, completely abrogates its ability to contribute to disease, as does replacement of the sole cysteine residue with leucine. However, although previous biochemical and cytotoxicity assays clearly indicated that mutations in residues important in oligomerization, binding, and prepore conversion greatly reduced activity or rendered the toxin inactive, once the mutated toxins were overexpressed by the natural host in the context of an infection it was found they were able to cause disease in a mouse model of myonecrosis. These results highlight the importance of testing the activity of virulence determinants in the normal host background and in an infectious disease context and provide unequivocal evidence that it is the ability of alpha-toxin to form a pore that confers its toxicity in vivo.

Clostridium septicum is a gram-positive anaerobe and is the primary causative agent of nontraumatic or spontaneous gas gangrene in humans (1, 45) and a leading cause of malignant edema in farm animals (23). Disease in humans is primarily associated with colorectal cancer and hematological diseases (14), which presumably provide the predisposing conditions that allow the vegetative cells or spores to disseminate from the intestine and establish an infection (46). Disease progresses rapidly and is frequently fatal, with aggressive debridement of affected tissue in conjunction with antibiotic therapy being the only effective treatment option (27, 28).

The C. septicum alpha-toxin is a lethal and necrotizing small, pore-forming toxin that belongs to the same family as aerolysin from Aeromonas hydrophila (9) and ɛ-toxin from Clostridium perfringens types B and D (15). The crystal structure of aerolysin reveals it to be predominantly made up of β-sheets and to consist of four domains arranged as a large and a small lobe (37). Aerolysin-based homology modeling of alpha-toxin suggests that it only has a single large lobe made up of three domains (16, 30). C. septicum alpha-toxin is encoded by the csa gene and is secreted as an inactive 46-kDa monomer (10, 25), which is cleaved to its 43-kDa active form by host cell-associated proteases following binding of the toxin to glycosylphosphatidylinositol (GPI)-anchored proteins on the cell surface (19, 24). The propeptide is known to be a potent inhibitor of oligomerization and acts to prevent the toxin from forming inactive aggregates before it binds to the cell surface (42). Binding of the monomers on the cell surface effectively concentrates the toxin and enhances monomer oligomerization (47). It was recently demonstrated that this process is further facilitated by binding of the toxin to GPI-anchored proteins within detergent-resistant microdomains (21). It has been shown that alpha-toxin binds to different types of GPI-anchored proteins in different species, affording broad host specificity (20, 48). Upon oligomerization, alpha-toxin undergoes a conformational change such that the β-hairpin transmembrane domain (TMD) within domain 2 (Fig. 1) is exposed and the oligomer forms a heptameric prepore β-barrel structure (30), which is then inserted into the host cell membrane to form a 1.6-nm pore (41).

FIG. 1.

Alpha-toxin model with residues investigated in this study identified. Residues R133 and N296 (blue) were mutated to alanine, while C86 (purple) was mutated to leucine to give the three receptor-binding mutants. The oligomerization mutant was generated by mutation of S178 (blue) to a cysteine in combination with the mutation of C86 to an alanine. Amino acids 212 to 222, which were deleted from the TMD, are shown in red. Amino acid deletions and substitutions are as previously described (30, 31). The model was generated with Swiss-Model (http://swissmodel.expasy.org/) by using aerolysin as a template (ExPDB template code 1preB). For clarity, after the model was constructed, the protoxin domain (domain 3) and residues G88 to F98 (domain 1) were hidden in the image without altering the structure.

Recent studies (30, 31) provided greater insights into the role of the three domains of alpha-toxin. Alanine scanning mutagenesis of the entire protein revealed that residues important in receptor binding were restricted to loops 1 and 3 and to an α-helix in domain 1. Residues in loop 2 of domain 1 were found to be important in prepore-to-pore conversion since replacement of these residues with alanine abolished cytotoxicity, even though the toxin was still able to bind and oligomerize on SupT1 cell membranes (31). Cysteine scanning in an alpha-toxin C86A background revealed that several residues throughout the protein are required for effective oligomerization; however, the greatest concentration of residues was found to be in domain 3 (31).

Previous work showed that alpha-toxin was essential for the virulence of C. septicum (26). Deletion of the csa gene rendered C. septicum avirulent in the mouse myonecrosis model, while complementation of this mutation with the wild-type csa gene restored the virulent phenotype (26). Histopathological examination of the affected tissue revealed alpha-toxin to be the major myonecrotic toxin, which also acted to reduce the levels of polymorphonuclear leukocytes (PMNLs) at the site of infection (26).

In this study, we have assessed the importance of individual alpha-toxin residues in the context of an infection. Building on our establishment of methods for the genetic manipulation of C. septicum (26), mutated copies of the csa gene were transferred into an alpha-toxin null mutant by conjugation. The resultant strains, which produced variant alpha-toxin derivatives deficient in receptor binding, oligomerization, prepore-to-pore conversion, or pore formation, were assessed for virulence with the mouse model (6, 26). The results indicate that alpha-toxin did not confer virulence in the host in the absence of receptor binding and that pore formation was essential for the pathogenesis of disease.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

The C. septicum strains used in this study are outlined in Table 1. Strains were cultured anaerobically as previously described (26). Where required, antibiotics were used at the following concentrations: rifampin, 20 μg ml⁻¹; erythromycin, 50 μg ml⁻¹; tetracycline, 10 μg ml⁻¹; thiamphenicol, 10 μg ml⁻¹. For cloning work, Escherichia coli strain DH5α (Gibco BRL-Life Technologies) was used, while E. coli strain S17-1 (43) was used as the donor for conjugations. E. coli strains were cultured aerobically with 2 × YT medium supplemented with erythromycin (150 μg ml⁻¹) or chloramphenicol (30 μg ml⁻¹).

TABLE 1.

C. septicum strains used in this study and characteristics of the expressed toxin

Strain	Designation	Genotype	Functional deficitb	Cytotoxicityb,c
JIR6086	Wild typea	csa+	NAe	NA
JIR6111	JIR6111a	csa::erm(B)	NA	NA
JIR6146	JIR6146WTa	csa::erm(B) (csa+)	NA	NA
JIR6220	JIR6220VECTOR	csa::erm(B) (vector)	NA	NA
JIR6221	JIR6221R133A	csa::erm(B) (csaR133A)	Prepore-to-pore conversion	0
JIR6222	JIR6222N296A	csa::erm(B) (csaN296A)	Receptor binding (L)d	0
JIR6256	JIR6256C86L	csa::erm(B) (csaC86L)	Receptor binding (H)f	0
JIR6225	JIR6225TMD	csa::erm(B) (csaTMDΔ212-222)	Pore formation	0
JIR6226	JIR6226OLIGO	csa::erm(B) (csaS178C:C86A)	Oligomerization	0

^aConstructed and analyzed as previously described (26).

^bAs previously determined with purified toxin (31).

^cPercentage of wild-type toxin LD₅₀ on SupT1 cells.

^dL, receptor-binding affinity reduced 483-fold.

^eNA, strain does not carry altered toxin.

^fH, receptor-binding affinity reduced 7 × 10⁶-fold.

Molecular techniques.

Previously published methods were used for the isolation of small-scale plasmid DNA (35) and genomic DNA (34) from C. septicum and for E. coli-C. septicum conjugations (29). Small-scale plasmid DNA was isolated from E. coli with QIAprep Spin Miniprep kits (Qiagen). All other manipulations were undertaken as previously described (38) or by following the manufacturers’ instructions.

Cloning manipulations.

The modified csa genes (30, 31) were PCR amplified from expression vectors with the high-fidelity DNA polymerase Pwo (Roche) and primers JRP1075 (5′-ATCCCGCGAAATTAATAC-3′) and JRP1923 (5′-CGTTAATTAATATCAATTTTTTTATCA-3′), where JRP1923 incorporates a unique PacI site and the additional 3′ adenosine native to JIR6086 and other C. septicum alpha-toxin sequences (4). The product was then cleaved with PacI and BbsI and cloned into csa suicide plasmid pJIR2230, which carries an erm(B) gene flanked by 5′ and 3′ csa fragments (26), so that the erm(B) gene was replaced with an internal csa fragment that included the required mutation. Plasmid pJIR2230 includes 343 bp of the sequence upstream of csa; therefore, the mutated csa gene will be under the control of its native promoter. The clostridial rep region and erm(B) gene were then blunt end cloned with EcoRV/HpaI-digested DNA from pJIR410 (44) into the StuI site upstream of the csa gene. A vector control plasmid was constructed in a similar fashion by replacing the csa gene (removed by using flanking PvuII sites) with the same rep-erm(B) fragment. The orientation of this fragment was shown to be the same in all of the resultant plasmids by PCR and restriction analysis. Sequence analysis confirmed that each of the recombinant csa plasmids had the predicted sequence in the csa gene region. These plasmids were transferred to C. septicum csa mutant JIR6111 by conjugation from E. coli S17-1; the resultant genotypes are listed in Table 1.

Hemolysin assay.

The hemolytic activity of culture supernatants was assessed as described previously (26). Briefly, 10×-concentrated supernatants were serially diluted twofold in Dulbecco's phosphate-buffered saline (PBS) and 30-μl aliquots were added to 0.5-cm wells punched into 15-ml 5% horse blood agar (HBA) plates. Plates were incubated aerobically at 37°C for 24 h, and the titer was taken as the log₂ of the last dilution to show a distinct zone of hemolysis.

Western blotting.

Cultures were grown to a turbidity of 2.0 to 2.5 at 600 nm, and the cells were removed by centrifugation at 3,500 × g for 15 min at 4°C. With Millipore concentrator columns (molecular mass cutoff, 10 kDa), supernatants were buffer exchanged with Dulbecco's PBS and concentrated 10-fold. Protein concentration was determined with a Bradford assay kit (Bio-Rad), and equal concentrations of protein were separated by 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) before being transferred to a Hybond N+ nitrocellulose membrane. Polyclonal affinity-purified anti-alpha-toxin antibodies (41) were used to probe the membrane, and horseradish peroxidase-conjugated anti-rabbit immunoglobulin G antibodies (Chemicon) were used for chemiluminescence detection.

Binding and oligomerization assays.

Toxins were assessed for binding and oligomerization on cell membranes as described before (30), with modifications. Briefly, cells were lysed by sonication and membranes were collected by centrifugation at 37,000 × g. Concentrated culture supernatants were treated with trypsin at 37°C for 30 min to activate the toxin, and the reaction was stopped with Complete protease inhibitor cocktail tablets (Roche); the activated supernatants were then added to the cell membrane preparation. Following incubation at 37°C for 2 h, membranes were again collected by centrifugation, separated by 4 to 15% SDS-PAGE, and probed for alpha-toxin by Western blotting.

Cell culture and cytotoxicity assays.

Cell cultures used for cytotoxicity assays were maintained as follows. C2C12 mouse myoblast and rat smooth aortic muscle (RASMC) cells were cultured in Dulbecco modified Eagle medium (Gibco) containing l-glutamine and HEPES and supplemented with 10% fetal calf serum (DF10). SupT1 human T cells were maintained in RPMI 1640 medium (Gibco) containing l-glutamine and HEPES and supplemented with 10% fetal calf serum (RF10). For cytotoxicity assays, C2C12 and RASMC cells were seeded at 1 × 10⁴/well and SupT1 cells were seeded at 5 × 10⁵/well in Nunc clear, flat-bottom 96-well plates. At all times, cells were maintained at 37°C in an atmosphere of 5% CO₂. Supernatants were prepared from three biological replicates of each culture, grown, and concentrated as for Western blotting. The relative levels of alpha-toxin were assessed with a Western blot assay with alpha-toxin-specific antibodies. Concentrated supernatants were serially diluted twofold in 50-μl volumes in microtiter plates and added to the cells for 4 h for SupT1 cells, as described before (30, 31), and 16 h for C2C12 and RASMC cells as these time frames were found to give the best range of cytotoxicities across the dilution series. Cytotoxicity was determined with the CellTitre Blue assay reagent (Promega) in accordance with the manufacturer's instructions, and the titer at which 50% cell death was obtained (LD₅₀) was determined.

Assay for pore formation.

A modification of the ⁸⁶Rb release assay was used to assess pore formation (11). C2C12 cells were cultured to mid-log phase, harvested, and resuspended at a concentration of 1.5 × 10⁶/100 μl of medium. The cell suspension was treated with 100 μCi ⁵¹Cr (specific activity, 9.2 × 10⁴ Ci/g; MP Biomedicals) and incubated at 37°C for 90 min. Cells were then washed to remove extracellular ⁵¹Cr, diluted to a concentration of 1 × 10⁵/ml, and dispensed in 100-μl aliquots into a microtiter plate. Serially diluted supernatants were then added to the cell suspensions to a final volume of 200 μl. Following incubation for 4 h at 37°C, 100 μl of supernatant was collected and analyzed for radioactivity with a Wallac 1470 Wizard gamma counter. Release of ⁵¹Cr from cells due to pore formation was expressed as a percentage of the total release (cells lysed with 100 μl of 10% SDS) minus the background (medium alone).

Mouse myonecrosis model.

Virulence in the murine model of myonecrosis was determined as previously described (6, 26), with slight modifications. Bacterial starter cultures were grown for 6 h prior to subculture into 90 ml of supplemented brain heart infusion broth to a final turbidity of 0.15 at 600 nm. These cultures were then maintained at 37°C for a further 4.5 to 5 h until they reached a turbidity of 1.7 to 1.9 at 600 nm, which ensured that the cultures were in mid-log phase and that the cells were predominantly short rods. The cells were washed once and resuspended to three times the packed cell volume in PBS (pH 7.5). Groups of five 6- to 8-week-old female BALB/c mice were injected intramuscularly in the right thigh with 50 μl of the washed cell suspension, which equates to a viable count of approximately 5 × 10⁸ cells. Mice were monitored hourly for up to 24 h for signs of disease, and the level of pathology was scored as either 0 (no disease or very slight disease), 0.5 (intermediate disease), or 1 (severe disease) for the parameters of limping, swelling of the thigh and footpad, blackening of the thigh and footpad, and malaise. When symptoms of disease became severe, usually when the mice were assigned a score of 1 in three or more parameters, mice were humanely killed via CO₂ inhalation. The combined scores for all mice per group were graphed for each parameter, and the scores assigned at time of killing are included in the combined score for the remainder of the trial. For histopathological examination, muscle tissue was dissected and fixed in 10% formalin for hematoxylin and eosin staining. All observations during the virulence trial and all histopathological examinations were performed blind. All animal experiments were conducted in accordance with Victorian State Government regulations and were approved and monitored by the AMREP Animal Ethics Committee.

Statistical analysis.

Data were analyzed with GraphPad Prism software. One-way analysis of variance and Tukey's posttest were used to analyze graphs of disease progression; Kaplan-Meier survival curves were analyzed by log rank tests.

RESULTS

Construction of complemented derivatives.

To study the importance of the various steps in the conversion of inactive alpha-toxin monomers to heptameric pore complexes, in the context of the infection process, we selected a series of alpha-toxin derivatives with substitutions for residues involved in receptor binding (N296A and C86L), oligomerization (S178C and C86L), or prepore-to-pore conversion (R133A) or a 10-amino-acid deletion within the TMD (TMD_212-222); these derivatives were constructed and assayed for in vitro alpha-toxin activity in earlier studies (30, 31). A series of isogenic C. septicum strains were constructed by complementing the previously reported csa mutant JIR6111 (26) with an isogenic plasmid (Table 1) that encodes a copy of the csa gene mutated such that the resultant toxin contained one of the required amino acid substitutions. Several attempts were made to clone the wild-type csa gene into the same vector backbone as the mutated genes; however, the presumed toxicity of the wild-type alpha-toxin for E. coli DH5α prevented it from being recloned in this manner; therefore, previous complementation strain JIR6146 (here designated JIR6146_WT; Table 1) (26) was used as a positive control.

Complemented C. septicum strains produce the altered toxins but are not hemolytic.

The hemolytic activity of the C. septicum strains was initially assessed by culturing them on 5% HBA. In agreement with previous results that showed purified preparations of the toxins not to be cytotoxic (30, 31), all strains complemented with a variant toxin were nonhemolytic; with the exception of JIR6226_OLIGO, which showed weak hemolysis, while the wild-type and JIR6146_WT strains were strongly hemolytic, as expected (Table 2). To confirm these observations, 10-fold-concentrated culture supernatants of the mutant strains were assayed quantitatively for hemolytic activity. No detectable hemolysis was observed for any of the strains, with the exception of the original wild-type strain and JIR6146_WT, which had titers of 5.0 ± 0.63 and 8.17 ± 0.41, respectively. Western blot assays showed that the altered toxins were being successfully expressed at levels comparable to that found for the complemented wild-type strain, which corresponded to approximately 1 μg of toxin per 10 μl of concentrated supernatant (Fig. 2). With the exception of the vector control strain, the level of secreted toxin was considerably higher than that of the wild-type strain—a consequence of the toxin gene being present on a multicopy plasmid—this level was actually closer to the amount of toxin produced by a separate clinical isolate of C. septicum, JIR6144.

TABLE 2.

Cytotoxicity of concentrated supernatants compared to their hemolytic activity and virulence

Strain	Hlya	Virulenceb	Mean LD50c ± SEM
			C2C12	RASMC	SupT1
Wild type	++	++	17.1 ± 0.53	14.8 ± 0.71	14.8 ± 0.06
JIR6146WT	+++	++	20.6 ± 1.0	15.6 ± 0.30	16.1 ± 0.44
JIR6220VECTOR	−	−	<1	<1	<1
JIR6221R133A	−	+	1.1 ± 0.09	<1	<1
JIR6222N296A	−	+	5.2 ± 0.56	<1	<1
JIR6256C86L	−	−	<1	<1	<1
JIR6225TMD	−	−	<1	<1	<1
JIR6226OLIGO	−/+	+	6.2 ± 1.6	4.5 ± 1.4	7.8 ± 0.78

^aRelative size of hemolytic zone when cultured on HBA. This measure of hemolysis was used as all strains except the wild type and JIR6146_WT had a titer of <1 in the hemolysin assay.

^b++, wild-type virulence; +, nearly wild-type virulence; −, avirulent.

^cLD₅₀, titer of supernatant that resulted in 50% loss of viability of tissue culture cells.

FIG. 2.

Western blot assay of concentrated culture supernatants. Cell-free culture supernatants were concentrated and separated by SDS-PAGE, and alpha-toxin was identified by Western blot assay with alpha-toxin-specific antibodies. As controls, 1 μg of purified wild-type toxin (1 μg WT) and supernatant from the C. septicum (Cs) clinical isolate JIR6144 were included. All supernatant samples were concentrated 10-fold, with the exception of that of the wild type, which was concentrated 20-fold. Note the reduced size of the TMD deletion toxin expressed by JIR6225_TMD. The value on the left is a molecular size in kilodaltons.

Virulence testing of the complemented strains reveals several to be pathogenic.

To characterize the in vivo activities of the substituted alpha-toxin derivatives, we used a mouse myonecrosis infection model to measure the progression of disease (26). To our surprise, the results showed that strains JIR6221_R133A and JIR6222_N296A, which were nonhemolytic and produced alpha-toxins that were previously shown (31) to be deficient in prepore-to-pore conversion and receptor binding (483-fold-reduced affinity), respectively, were virulent, as was the slightly hemolytic strain JIR6226_OLIGO (Fig. 3). The progression of disease in the mice infected with these strains closely mirrored the progression observed in the wild-type and JIR6146_WT strains, with a rapid increase in the observed swelling, blackening, limping, and malaise, suggesting that, under these conditions, the altered toxins were as effective as the wild-type toxin. Despite this similarity in disease progression, analysis of the survival curve (Fig. 3) revealed that the three mutant strains JIR6221_R133A, JIR6222_N296A, and JIR6226_OLIGO were less virulent than the wild-type and JIR6146_WT strains (JIR6221_R133A, P < 0.05; JIR6222_N296A, P = 0.0002; JIR6226_OLIGO, P = 0.002). By contrast, strains JIR6225_TMD and JIR6256_C86L, which produced alpha-toxins with a deleted TMD and 7 × 10⁶-fold reduced receptor-binding affinity, respectively, were avirulent. In these infections, swelling of the thigh and malaise were seen toward the end of the experiment, as observed in the mice infected with control strain JIR6220_VECTOR. JIR6225_TMD and JIR6256_C86L were significantly different from the wild-type strain (P < 0.0001) and similar to JIR6220_VECTOR (P > 0.05) in most of the parameters measured. To assess the possibility that strains JIR6221_R133A, JIR6222_N296A, and JIR6226_OLIGO were virulent because of a recombination event that restored a wild-type copy of the gene, all strains were recovered from the infected tissue and cultured on 5% HBA to assess their hemolytic phenotype. All strains were found to be phenotypically unaltered by their passage through the animals.

FIG. 3.

Cumulative gross pathology of mice infected with isogenic strains of C. septicum. Three groups of five mice were inoculated with a C. septicum strain, and the progression of disease was monitored for 24 h; the score is expressed as a percentage of the highest possible score. Symbols: wild type, □; JIR6146_WT, ▵; JIR6220_VECTOR, ▿; JIR6221_R133A, ○; JIR6222_N296A, *; JIR6226_OLIGO, ⋄; JIR6225_TMD, •; JIR6256_C86L, ▪. Note that the survival plot for JIR6221_R133A was moved 0.1 U along the x axis for clarity.

Histological evaluation revealed that the tissue of mice infected with the virulent strains was characterized not only by the expected necrosis of muscle tissue but also by considerable interstitial hemorrhage and edema, which correlates with the rapid blackening of the thigh and footpad that was observed (Fig. 4A to D). Such extensive hemorrhage was not observed in previous work, where bacterial strains were cultured differently (26). The histology of the muscle tissue of mice infected with strain JIR6220_VECTOR, JIR6225_TMD, or JIR6256_C86L also revealed the activity of an additional potential myonecrotic toxin. While gross pathology indicated no significant disease in these mice, aside from swelling of the thigh and associated restricted movement of the limb, histological examination revealed some necrosis in the absence of hemorrhage at 5 h after inoculation (Fig. 4E) and extensive necrosis associated with the influx of leukocytes at 24 h postinoculation (Fig. 4F). These results provide evidence that alpha-toxin mediates major interstitial hemorrhage, as well as tissue necrosis, and also suggest that C. septicum may secrete a second myonecrotic toxin.

FIG. 4.

Histopathological examination of infected muscle tissue. Extensive necrosis and interstitial hemorrhage are evident in tissue from mice where blackening was noted, with little to no PMNL infiltrate (A, B, C, and D). Despite minimal gross pathological changes, significant necrosis is evident in tissue from mice infected with avirulent strains (E and F) although interstitial hemorrhage is absent, and PMNL infiltrate is evident at 24 h (F). Panels: wild type, A; JIR6146_WT, B and C; JIR6226_OLIGO (as a representative of virulent mutants), D; JIR6256_C86L (as a representative of avirulent mutants), E; JIR6220_VECTOR at 24 h, F. Necrosis (white arrowhead), hemorrhage (H), thrombosis (Th), leukocytes (L), and bacterial cells (*) are indicated. Scale bars: black 10 μm; white, 5 μm. Samples were from mice killed at 5 to 6 h, except for panel F. Cs, C. septicum.

Nonhemolytic strains are cytotoxic against nucleated cells.

To determine why these nonhemolytic strains were virulent in mice, we undertook quantitative cell toxicity assays with a range of cell lines including mouse muscle (C2C12), RASMC, and human T cells (SupT1), the latter because SupT1 cells were used in the original studies of the altered toxin derivatives (30, 31). The supernatants of several nonhemolytic strains were cytotoxic against these cell lines, although the amount of toxin required to kill 50% of the cells varied greatly, depending on the cell type (Table 2). Notably, the supernatants from JIR6226_OLIGO, which showed slight hemolysis and was still virulent, showed a higher level of activity on all of the cell types tested, although much less than those from the wild-type strains. C2C12 cells were the only cell type found to be susceptible to all of the mutants that were virulent in mice, including strains JIR6221_R133A and JIR6222_N296A, which were not cytotoxic against SupT1 cells, as previously indicated (31). Therefore, C2C12 cell cytotoxicity was the best in vitro correlate of virulence.

The alpha-toxin mutant with the S178C and C86A substitutions is still able to oligomerize and form functional pores.

To further characterize the alpha-toxin activity of JIR6226_OLIGO, we determined whether the expressed toxin could bind to and oligomerize on C2C12 cell membranes. Membrane preparations of C2C12 cells were incubated with supernatants from strains JIR6146_WT, JIR6226_OLIGO, and JIR6225_TMD. Binding and oligomerization of the alpha-toxin derivatives were assessed by using Western blot assays to identify monomeric and oligomeric forms of the toxin. The results indicated that the toxin secreted by JIR6226_OLIGO was able to bind to the cell membrane and form a small amount of oligomeric complex, although to a lesser degree than the wild-type toxin and the TMD deletion protein (Fig. 5A). To show that oligomerization of the toxin resulted in pore formation, not just inactive aggregation, we used a ⁵¹Cr release assay in which pore formation would result in the release of intracellular ⁵¹Cr. The supernatants of JIR6226_OLIGO caused 50% release of the chromium at a dilution factor of 1.2 ± 0.35, whereas supernatants of JIR6146_WT caused 50% release at a significantly higher dilution factor of 4.5 ± 0.15 (P < 0.01) (Fig. 5B). As expected, treatment of cells with supernatants from JIR6220_VECTOR and JIR6225_TMD, controls for binding and oligomerization without pore formation, caused no significant release of ⁵¹Cr, even though the toxin produced by JIR6225_TMD was able to bind and form an oligomeric complex (Fig. 5A). These data indicate that amino acid substitutions S178C and C86A did not completely abolish oligomerization when the toxin was at high concentrations and that the resulting oligomeric complexes were able to form functional transmembrane pores.

FIG. 5.

Functional assays for alpha-toxin oligomerization and cell lysis. (A) Concentrated supernatants of JIR6146_WT, JIR6226_OLIGO, and JIR6225_TMD were mixed with equal volumes of C2C12 cell membrane preparations and separated on a 4 to 15% gel, and blots were probed with anti-alpha-toxin antibodies that recognized the monomeric (M) and oligomeric (O) forms. Note the difference in the amount of oligomeric form and the decreased size of the toxin expressed by JIR6225_TMD. (B) ⁵¹Cr release from C2C12 cells in response to alpha-toxin. Supernatant was added to cells in 10-fold dilutions, and the percent release of ⁵¹Cr was measured compared to that of SDS-lysed cells (± standard deviation). Symbols: JIR6146_WT, ▴; JIR6220_VECTOR, ○; JIR6225_TMD, □; JIR6226_OLIGO, ♦.

DISCUSSION

This study provides further evidence that alpha-toxin, a pore-forming cytolysin, is an essential virulence factor of C. septicum and shows that its TMD is essential for activity in vivo, since the deletion of a 10-amino-acid segment within this region abrogated virulence. This deletion still allowed the toxin to be cleaved, bind to its receptor, and form a heptameric oligomer (Fig. 5A) (30). However, the fact that strain JIR6225_TMD, which secreted this toxin variant, was avirulent, indicates that pore formation by alpha-toxin is essential for the virulence of C. septicum. As commented upon in a recent review (5), our previous study represents one of the few (26) in which a pore-forming toxin has been shown to be required for virulence. To the best of our knowledge, our present study represents the first in vivo demonstration of the specific requirement for pore formation for the effect of a small, pore-forming toxin on virulence.

It was surprising that for the JIR6226_OLIGO, JIR6221_R133A, and JIR6222_N296A derivatives there was no correlation between in vitro hemolytic activity and virulence. In a similar study, site-directed mutagenesis of the protective antigen, lethal factor, and edema factor of Bacillus anthracis demonstrated a close correlation between the in vitro and in vivo activities of the toxins (12). Another study on listeriolysin O reported a good correlation between loss of hemolytic activity in vitro and a reduction in virulence (32), despite the fact that the primary role of listeriolysin is to enable Listeria monocytogenes to escape from the phagocytic vacuole (40).

In our experiments, there was a slightly better correlation between virulence and cytotoxicity against nucleated cells rather than red blood cells, especially with the mouse-derived myoblast cell line C2C12. C2C12 cells, and mouse cell lines generally, have been previously shown to be particularly sensitive to alpha-toxin, and C2C12 cells have several GPI-anchored proteins that bind alpha-toxin, proteins that are not present on other cell types (20, 21). It is also possible that C2C12 cells may have other receptors to which the receptor-binding mutants were still able to bind but which were not present on the human-derived SupT1 T cells used to determine the original binding activity of the mutants (31). These alternative receptors may also be available on the surfaces of muscle cells in mice. Unfortunately, attempts to demonstrate hemolytic activity by using mouse erythrocytes were confounded by the presence in the supernatants of an additional hemolysin (data not shown), which was inactive against horse erythrocytes and may be C. septicum δ-toxin (33). The greater toxicity of alpha-toxin against nucleated cells compared to erythrocytes suggests that, in addition to pore formation, alpha-toxin may elicit an additional cell response that is distinct from the osmotic lysis measured in hemolysin assays. There is a growing body of evidence that describes additional activities of pore-forming toxins (5, 18, 36), and we are currently investigating other intracellular effects of C. septicum alpha-toxin.

We propose that the capacity of the toxins that had reduced in vitro cytotoxicity to complement an alpha-toxin null mutant to virulence was due to the concentration of the toxin present in our focal-infection model coupled with the use of an in trans complementation approach. With the culture method described here for the generation of the virulence trial inoculum, wild-type complemented strain JIR6146_WT was no more virulent than the wild-type strain, despite producing approximately 16-fold more toxin, according to hemolysin assays. In such an acute model, where a large number of bacteria are confined to a small focus of infection, a threshold of toxic activity may be reached. In this situation, once enough toxin is present to induce the symptoms of myonecrosis and hemorrhage, the addition of extra toxin may only act to form more pores per cell in an already saturated system. More simply, when relating the in vivo and in vitro results, the concentration of toxin in the infection model may be best compared to the concentration of toxin achieved at concentrations higher than the LD₅₀ in the cytotoxicity and ⁵¹Cr release assays. This situation is an unavoidable limitation of the existing animal model, in which to reproducibly establish an infection with anaerobic bacteria in viable aerobic tissue, a large number of bacterial cells must be inoculated. Unfortunately, the severe effects of this disease are such that the determination of an LD₅₀ is no longer ethically feasible. Nonetheless, this study clearly showed that alpha-toxin does not have any additional effects that arise from its presence in the extracellular milieu, or from binding and oligomerization in the absence of pore formation, as revealed by the avirulent nature of strains JIR6256_C86L and JIR6225_TMD, respectively.

Although the receptor-binding mutant JIR6256_C86L was also inactive in vivo, this result does not suggest that it is this single cysteine residue in alpha-toxin that is essential for activity, since the active oligomerization mutant JIR6226_OLIGO (S178C and C86A) was still virulent in vivo. Rather, it suggests that the slightly larger nonpolar leucine residue, which was substituted for the cysteine, occluded binding or altered the structure so that the toxin was inactive. In fact, JIR6226_OLIGO showed the least reduction in activity of all of the mutants and even caused slight hemolysis on HBA. Previous work has shown that alpha-toxin oligomerizes in solution in a concentration-dependent manner (8) and would be more likely to oligomerize when bound to membranes because of the effective concentration of the toxin on the membrane compared to the extracellular milieu (2). Our results indicate that the oligomerization mutant was still active at high concentrations, as shown by the cytotoxicity and ⁵¹Cr release assays (Table 2; Fig. 5B). While the S178C and C86A substitutions appreciably reduced oligomerization of the toxin in biochemical assays, the large number of bacteria injected, expressing high levels of toxin, would presumably lead to a high concentration of toxin at the site of infection. Since these amino acid changes were not sufficient to completely hinder oligomerization, this concentration would encourage the formation of heptameric oligomers and the toxin could be considered biologically active, albeit at a proportionally lower level than the wild-type toxin.

Analysis of gross pathology showed that active alpha-toxin resulted in rapid swelling and blackening of the thigh and footpad, which corresponded to necrosis of the muscle tissue, interstitial hemorrhage, and edema. A similar pathology has also been attributed to the activity of ɛ-toxin of C. perfringens types B and D, which are structurally related to C. septicum alpha-toxin (3, 15, 39). In C. septicum infections, hemorrhage is likely to be due to destruction of the microvasculature, resulting in reduced blood flow to the site of infection and ultimately leading to ischemia, which would favor the survival of C. septicum in the absence of external trauma. It is highly likely that alpha-toxin acts directly on the endothelial cells of the microvasculature since it is able to bind to the ubiquitous GPI-anchored proteins, allowing it to target a broad range of cell types (13, 21). Consistent with this hypothesis, we recently showed, by using intravital microscopy of microvasculature perfused with C. septicum culture supernatants, that alpha-toxin activity profoundly compromises blood flow (22).

In addition, histopathological examination of tissue from mice infected with “avirulent” strains that caused no gross pathology prompts us to suggest that C. septicum produces a second myonecrotic toxin that is distinct from the activity of alpha-toxin. Similar results were obtained from the virulence testing of a phospholipase C mutant of C. perfringens (6, 17). Deletion of the gene that encodes phospholipase C, which is the primary virulence factor in C. perfringens-mediated gas gangrene, rendered C. perfringens avirulent in the mouse myonecrosis model, as measured by examination of gross pathology (6). However, later histopathological examination revealed that some tissue necrosis was occurring in the absence of phospholipase C and that this was due to the activity of an additional C. perfringens toxin, perfringolysin O (7, 17). Necrosis of C. septicum-infected tissue was apparent even at relatively early stages of infection with the avirulent strains, but in mice infected with strains with active alpha-toxin, the necrosis was obscured by the tissue and coagulative necrosis associated with alpha-toxin activity and ischemia. This additional necrosis could have been due to the δ-toxin or the activity of other virulence factors secreted by C. septicum.

In conclusion, this study has provided important evidence that pore formation is essential for the biological activity of alpha-toxin and the virulence of C. septicum. We have shown that the cytotoxicity of alpha-toxin for nucleated cells, rather than its ability to cause hemolysis of erythrocytes, is a better in vitro measure of its effect on virulence. Most importantly, this study has highlighted the importance of the use of in vivo models to obtain an accurate assessment of the activity of bacterial virulence factors in disease.