CytK is a cytotoxin isolated from a strain of Bacillus cereus cultured from cases of necrotic enteritis and the amino acid sequence of the protein suggests that it may belong to the family of β-barrel pore-forming toxins. We show here in planar lipid bilayers the toxin is able to form pores which are weakly anion selective and exhibit an open channel probability close to one. The predicted minimum pore diameter is approximately 7 Å. We also show that cytK is a potent cytotoxin against human intestinal Caco-2 epithelia. CytK, like other β-barrel pore-forming toxins, spontaneously forms oligomers which are resistant to sodium dodecyl sulphate (SDS), but not to boiling. CytK represents a pore-forming toxin linked with human cases of necrotic enteritis.
We recently described a new cytotoxic protein, cytK, secreted by a strain of Bacillus cereus cultured from an outbreak of food poisoning in which three people died of necrotic enteritis . CytK was cytotoxic against Vero cells and was haemolytic. The deduced amino acid sequence had approximately 30% identity to a group of related β-barrel pore-forming toxins including Staphylococcus aureusα-toxin and Clostridium perfringens type C β-toxin [1,2]. The crystal structure of the α-toxin of S. aureus has been resolved to <2 Å resolution by crystallography and indicates that a heptameric β-barrel forms the transmembrane domain . C. perfringens type C β-toxin is implicated in the aetiology of necrotic enteritis in animals  and recently has been shown to form pores in planar phospholipid bilayers . We present evidence to support the concept that cytK belongs to the β-barrel pore-forming toxins and possesses potential enterotoxic activity.
2Materials and methods
2.1Strain, culture medium and culture conditions
B. cereus strain 391-98  was cultured in modified CGY medium  for production of cytK at 32°C for 5 h and EDTA (1 mM) was added at the time of harvest. Extracellular proteins were separated from the cells by centrifugation and the supernatant proteins were precipitated with 70% saturated ammonium sulphate solution.
2.2Purification of proteins
Proteins were purified as described in , using chromatography on a DEAE-Sephacel column (Pharmacia) with Bis-Tris buffer, pH 5.9 and chromatography on a hydroxyapatite column (Bio-Rad) with sodium phosphate buffer, pH 6.8. The last purification step employed chromatography on a Resource Q (ReQ) column (Pharmacia) with triethanolamine buffer, pH 8.1 with a linear 40 ml NaCl gradient from 0 to 0.2 M.
2.3Caco-2 cell assay
The human colon cancer line Caco-2  was cultivated in RPMI 1640 plus 10% foetal calf serum, gentamicin and l-glutamine in 24-well chambers until confluent monolayers were obtained (approximately 7 days). Toxicity was determined by measuring the inhibition of protein synthesis according to the method of Sandvig and Olsnes .
2.4Planar lipid bilayer recording
Planar lipid bilayer recordings were carried out using a system described previously . Briefly, bilayers were formed from a dispersion of 15 mg ml−1 palmitoyl-oleoylphosphatidylethanolamine and 15 mg ml−1 palmitoyl-oleoylphosphotiditylserine in n-decane, drawn across a 0.4 mm diameter hole. The cis chamber (to which the toxin and drugs were added) was held at ground (0 mV), and the trans chamber was clamped to a range of potentials using a GeneClamp 500 patch-clamp amplifier/CV-5B 100G headstage (Axon Instruments). The sign of the membrane potential refers to the trans chamber, and currents are defined as ‘positive’ when cations flow from trans to cis. Transmembrane currents were low-pass filtered at 500 Hz (8 pole Bessel) digitised at 5 kHz, and recorded directly to computer disk via a Digidata 1200 AD interface (Axon Instruments). Membrane capacitance was measured by differentiating a triangle wave input of 0.2 kHz. Only bilayers that had a conductance of less than 10 pS were used for incorporation of the toxin. Initial capacitance was at least 300 pF (corresponding to a specific capacitance of 0.25 pF cm−2). All recordings were made at room temperature (19–22°C), and 30–180 s recordings for each holding potential were analysed off-line using PAT v7.0 software (Strathclyde Electrophysiology Software). Maximum current amplitudes were determined from the peaks of Gaussian functions fitted to amplitude histograms.
CytK of strain 391-98 was purified as described previously  and no other bands were visible when the protein fraction was analysed by SDS–PAGE (Fig. 1, lane 1). Purified cytK consisted mainly of the monomeric form, but small amounts of polymers of approximately 200 kDa, which were resistant to SDS, but not to boiling, were also present (Fig. 1, lanes 2 and 3).
The toxicity of cytK against intestinal epithelia was examined using the Caco-2 cell line, a cell line with mixed large and small bowel phenotypes. The amount of protein necessary for 50% inhibition of protein synthesis in 5×104 cells was found to be 16 ng (mean value of duplicate measurements). This value is five times lower than that found for the monkey kidney Vero cell line epithelium.
We examined the ability of purified cytK to form pores in planar phospholipid bilayers. Addition of ∼40 ng cytK to 500 μl bathing solution resulted in the appearance of step current increases within 1–20 min, consistent with the insertion of single channels into the bilayer. The current steps almost never showed any closures but instead remained at the new level at both positive or negative voltages (between ±100 mV). Current voltage relationships were linear and yielded a slope conductance of 159±4 pS (mean±S.E.M.) in 250 mM NaCl (Table 1). Current amplitudes that fell outside ±2 standard deviations of the mean were rare (<1%) indicating that pores of fixed size were dominant, rather than yielding a broad range of conductance values as seen with other β-barrel toxins such as C. perfringens type C β-toxin and S. aureusα-toxin [5,10].
Table 1. Conductance of cytK pores in different salt concentrations
All solutions contained 5 mM HEPES/NaOH, pH 7.2. Values represent mean±standard errors.
To estimate the selectivity of the channels between cations and anions, reversal potential measurements were made in asymmetrical bi-ionic solutions (1:0.2 M and 1:0.5 M NaCl). The reversal potentials obtained were not significantly different from zero indicating little or no discrimination between charged monovalent ions. Due to the small number of channels measured in each bilayer (between one and five) an alternative method was used to determine the ion selectivity. Single channel conductances were measured in symmetrical salt solutions with the anion and cation replaced by gluconate− and choline+ respectively. Data are summarised in Fig. 2. Replacement of either cation or anion significantly reduced channel conductance (P<0.01 for both choline Cl− and Na+ gluconate, ANOVA, t-test) with anion replacement having greatest effect indicating a slight preference for anions over cations.
The conductance of the pores in sodium and potassium solutions followed the predicted mobility of the cations in water with the conductance higher in 0.1 M KCl (g=57.6±4.6 pS; n=3 bilayers) than in 0.1 M NaCl (g=48.1±2.6 pS; n=10 bilayers, comparison of slope conductances: P<0.01). The conductances of the pores in increasing NaCl solutions are given in Table 1. The relationship between pore conductance and the conductivity of the salt solution is shown in Fig. 3 and the slope of the line yielded a value of 1.16 when fitted by nonlinear regression as a double logarithmic plot (Fig. 3, inset). This indicates that the pore is filled with water (i.e. it does not show saturation with increasing ion concentration). Under such conditions the pore can be assumed to act as a water-filled cylinder and the diameter can be estimated using the formula:
where G=conductance, σ=conductivity of the solution, l=length of the pore. Assuming l=100Å (taken from S. aureusα-toxin ) the function G/σ at the intercept yields a value of 3.3×10−9 cm−1 which gives a pore diameter of 6.6 Å.
Unlike previous findings with S. aureusα-toxin and C. perfringens type C β-toxin [5,10], addition of 2 mM zinc ions to the cis chamber had no effect on ion channel conductance (167±3 pS in 250 mM NaCl, n=5 bilayers, P>0.05).
The single channel activity of cytK in pure phospholipid bilayers confirms the predicted pore-forming ability predicted from sequence homology. The pores exhibited characteristics comparable to other β-barrel toxins: water-filled pores with single channel conductances similar to that of S. aureusα-toxin, aerolysin, and C. perfringensβ-toxin (Table 2), and little selectivity between monovalent anions and cations. However, certain features of the pores are distinct: cytK pores yielded a linear I/V curve and were unaffected by addition of millimolar concentrations of zinc ions. C. perfringens type C β-toxin exhibited rectification at negative voltages  and along with S. aureusα-toxin  exhibited voltage-dependent closure after addition of zinc ions .
Table 2. Comparison of conductance values (pS) of selected pore-forming toxins in varying salt concentrations
The reversal potentials for cytK measured in bi-ionic solutions were close to zero, consistent with a non-selective water-filled pore. It would be expected that summation of the conductances in choline and gluconate would match that obtained in NaCl (159 pS). The reduced conductance in sodium gluconate (49 pS) might be accounted for by the difference in ionic activity coefficient at that concentration, although gluconate may exert channel blocking or allosteric events to modulate the bulk conductance.
The estimated channel diameter of cytK was ∼7 Å. This value assumes a channel length of 100 Å (taken from the crystal structures of S. aureusα-toxin and aerolysin). Using Eq. 1, Menestrina  estimated the diameter of the S. aureusα-toxin to be 11 Å, a value close to the narrowest point in the pore lumen (14 Å) estimated from crystallography . The same method predicted a channel lumen diameter of 9 Å for aerolysin , whereas subsequent estimates from crystallography of proaerolysin indicate a minimum of 17 Å. At 7 Å the cytK pore is smaller than the other β-barrel pore-formers, although Aeromonas sobria aerolysin was reported to have a pore diameter of 8 Å. The variation in predicted pore diameter remains a problematic issue and perhaps the report that staphylococcal α-toxin appears to form pores of varying diameter, depending on the concentration of toxin used , offers a partial explanation.
The identity between the previously identified β-barrel pore-forming toxins never exceeds 30% homology  and cytK aligns to a similar extent. While the channel properties of cytK and S. aureusα-toxin have certain features in common, the biological function of these pore-forming toxins may be very different. Staphylococcal α-toxin attacks erythrocytes, monocytes, lymphocytes and keratinocytes but intestinal epithelia (Int 407 cells) are relatively resistant . The difference in binding properties may be explained by differences in amino acid sequence between the two toxins. Residues Arg-66, Glu-70, Arg-200, Asp-254, Asp-255, and Asp-276 in α-toxin are probably involved in binding to rabbit erythrocytes . In the regions that align with these residues in α-toxin, cytK is very different from α-toxin, containing additional amino acids (see Fig. 3 in ) that may form a receptor-binding domain.
We show that cytK can spontaneously form oligomers that are resistant to SDS but not boiling, a feature of other β-barrel pore-forming toxins. The size of the cytK oligomer (approximately 200 kDa) may correspond to a pentamer, hexamer or heptamer. It will be of interest to determine whether cytK forms similar oligomers in membranes of target cells.
To support the hypothesis that cytK is an enterotoxin we find that cytK is cytotoxic to Caco-2 cells at a concentration five times lower than in Vero cells (derived from monkey kidney) measured in a similar manner . This differs from C. perfringensβ-toxin which is only weakly active against another intestinal cell line , requiring ∼0.4 μg (approximately 10 000-fold less potent than cytK). C. perfringensβ-toxin is proposed to be cytotoxic but not cytolytic. We assume that the activity of cytK in Caco-2 cells is directly related to its pore-forming ability. The toxicity may reside at least in part due to the fact that the pore is non-selective and remains in the open configuration between ±100 mV whereas C. perfringensβ-toxin was selective for cations, displayed inward rectification due to a reduced open probability at positive potentials. The features of cytK described here may account for the cytotoxicity of the protein despite a small predicted pore diameter. Initial studies indicate that cytK is not widely found in isolates of B. cereus from diarrhoeal illness . This isolate from a food poisoning outbreak in which three people died lacks the other putative enterotoxin genes (HBL, NHE) as detected by polymerase chain reaction and hence is atypical. Consequently, we propose a role for β-barrel pore-forming toxins in the aetiology of necrotic enteritis in humans as well as animals.
This work was supported by The Research Council of Norway, Grant 119303/112 to T.L., and TINE Norske Meierier. The authors thank C. Flamigni for assistance with bilayer experiments.