Ionophoretic properties and mitochondrial effects of cereulide

The emetic toxin of B. cereus


M. S. Salkinoja-Salonen, Department of Applied Chemistry and Microbiology, POB 56, FIN-00014 University of Helsinki, Finland. Fax: + 358 9 70859322, Tel.: + 358 9 70859300, E-mail:


The emetic toxin of Bacillus cereus, found to cause immobilization of spermatozoa and swelling of their mitochondria, was purified and its structure found to be identical to the earlier known toxin cereulide. It increased the conductance in black-lipid membranes in KCl solutions in an ionophore-like manner. It formed adducts with K+, Na+, and NH4+ but the conductance was highly selective for K+ in relation to Na+ and H+ (three orders of magnitude). The increase in the kinetics of conductance indicated a stoichiometric ratio between the cereulide and K+. Its ionophoretic properties are thus similar to those of valinomycin. In addition, its effects on rat liver mitochondria were similar: it stimulated swelling and respiration in respiring mitochondria in the presence but not in the absence of K+, it reduced the transmembrane potential under these conditions. In nonrespiring mitochondria, swelling was seen in KNO3- but not in NaNO3-containing media, less in acetate. In NaNO3 media addition of the cereulide caused a transient diffusion potential which was reduced by adding K+. It is concluded that the toxic effects of cereulide are due to it being a K+ ionophore.


mitochondrial inner membrane transmembrane potential


black-lipid membrane


matrix-assisted laser desorption ionization time-of-flight MS


rat liver mitochondria.

The emetic toxin of strains of Bacillus cereus, that may cause serious food poisoning, has been isolated and found to be structurally related to valinomycin [1]. It was reported to cause vacuole formation and the vacuoles were then found to be swollen mitochondria in mouse liver and HeLa cells [2]. The emetic toxin, also called cereulide [1], stimulated respiration like the uncoupling agent 2,4-dinitrophenol. Similar observations were made in liver failure following a fatal food poisoning due to the toxin which caused lowering of the respiratory control ratio in isolated rat liver mitochondria [3]. These findings are analogous to the early findings on the mechanism of action of valinomycin, found to cause a mitochondrial energy-dependent uptake of K+, ejection of H+, stimulation of respiration in the presence of K+ and swelling [4,5]. Several other bacterial toxins have ionophoretic or channel-forming properties: examples are diphtheria toxin, the enterohemorrhagic haemolysin of Escherichia coli[6] and an insecticidal toxin from B. thuringiensis[7].

We have earlier reported swelling of mitochondria in boar spermatozoa induced by extracts of water-damaged indoor building materials heavily contaminated by microbes [8]. The cessation of spermatozoic motility was used as a sensitive test for cereulide and related toxins [9]. In this study we report the purification from B. cereus of the active agent, its structural features and ionophoretic properties in black-lipid membranes (BLM) and its effect on mitochondrial swelling and function. It is shown that the agent is structurally identical to cereulide and that it acts like valinomycin, i.e. it is a specific K+ ionophore.

Materials and methods

Electron microscopy

Electron microscopy was performed as described previously [8].

Purification and analysis of the sperm toxin from B. cereus

The methanol extract of B. cereus 4810/72 [9] was diluted to 90% (v/v) with water and injected to the Sep-pak C18 cartridge (Waters Co, Milford, MA, USA), washed with methanol/water (9 : 1) and finally eluted with methanol. The eluate was analysed by reversed phase HPLC (Smart, Pharmacia Biotech, Sweden). The column used was Sephasil C8 SC 2.1 mm in diameter × 100 mm, 5 µm. Elution solutions were A, 0.1% trifluoroacetic acid in water; and B, 0.075% trifluoroacetic acid in acetonitrile. A gradient from 10% A and 90% B to 100% B in 5 min was used with a flow rate of 100 µl·min−1. For detection A215 was used. Before injections the methanol eluate was evaporated to dryness and dissolved in 90 : 10 acetonitrile/water containing 0.075% trifluoroacetic acid. Valinomycin and gramicidin D (Sigma) were used as external reference compounds in the assay of cereulide.

MS analysis

Matrix-assisted laser desorption ionization time-of-flight MS (MALDI-TOF MS) was performed in the delayed-extraction mode with a BIFLEX™ mass spectrometer (Bruker-Franzen Analytik, Bremen, Denmark) using a 337-nm nitrogen laser. A thin-layer matrix preparation was used according to Vorm et al. [10].

Preparation of rat liver mitochondria

Rat liver mitochondria (RLM) was performed by a conventional procedure as described previously [11].

Conductance measurements in BLM

BLM was formed from phosphatidylcholine from soy beans (Sigma) in a hole diameter of 0.5 mm between two compartments according to Mueller et al. [12]. The lipid was dissolved in n-decane at a concentration of 20 mg·mL−1, applied around the hole, dried, 5 mL electrolyte solution added in the cis and trans chambers and a small drop added with a capillary pipette followed by bubbling leading to the formation of a BLM. The conductances at various transmembrane voltages were measured at 25 °C as described previously [13]. The electrolyte concentration was 100 mm, pH 6.0, if not otherwise stated.


Mitochondrial protein was measured with fluorescamine using excitation wavelength 398 nm and emission wavelength 480 nm with BSA as standard [14]. Respiration was followed polarographically with the aid of a Clark electrode. The membrane potential (Δψ) was measured with rhodamine 123 as a fluorescent probe using the excitation at 503 nm and emission at 527 nm [15] at room temperature with the Hitachi F4000 fluorometer. Mitochondrial swelling was followed by measuring changes in absorbance at 520 nm in the Shimadzu UV3000 photometer. Respiration was measured polarographically. Octanol/water partition coefficients were estimated with the aid of the logkow program for Microsoft Windows 3.1, version 1.57 (Syracuse Research Corporation, Environmental Science Center, North Syracuse, NY, USA).


To study the properties of the toxin causing immobilization of spermatozoa, the toxin was purified and its biological effect compared to that of purified cereulide using the spermatozoan mobility test to verify that the active component was purified. Its ion-conductance in BLM could then be characterized and its effects on RLM studied.

Swelling of spermatozoan mitochondria

Figure 1 shows the effect of the emetic toxin on spermatozoan mitochondria. Compared with control mitochondria (Fig. 1, top) toxin-treated spermatozoa exhibited some highly swollen, vacuole-like mitochondria with the outer membrane partially removed (Fig. 1, bottom). Cyclosporin A, 160 ng·mL−1 did not prevent the mitochondrial swelling (data not shown).

Figure 1.

Figure 1.

Thin sections of the mitochondrial sheath of boar spermatozoa exposed to cell extracts ofB. cereusstrains. Extended boar semen containing 40 × 106 spermatozoa·mL−1 were exposed for 7 days to extracts of 2 mg B. cereus cells·mL−1 (w/w). The upper panel shows the mitochondrial sheath of a spermatozoon exposed to cell extract from the nontoxic B. cereus type strain ATCC 14579: the mitochondria are of normal size and shape. The lower panel shows mitochondrial damage of a spermatozoon exposed to cell extract prepared from a cereulide-producing B. cereus strain F5881: 320 ng of cereulide·mL−1 was added to diluted boar semen. The picture shows mitochondria increased about 10 times in size, with disrupted outer membranes. Bar, 200 nm.

Purification, analysis and quantitation of cereulide

The single peak obtained by HPLC of the methanol extract of B. cereus, was evaporated by a stream of nitrogen gas and the residue was dissolved in methanol. This methanol solution was analysed by MALDI-TOF MS. Three peaks were obtained (Fig. 2) with m/z 1170.46, 1175.54 and 1191.54. The first represents the ammonium, the second the sodium and the last the potassium adduct of a single molecular species. The MALDI-TOF MS data and results published elsewhere [9] confirm that the sperm toxin is a cyclic dodecadepsipeptide the structure of which is identical to that reported for cereulide. The concentration of Sep-pak C18 cartridge purified cereulide was 126 µg·mL−1.

Figure 2.

Figure 2.

MALDI-TOF MS spectrum of the purified sperm toxin fromB. cereus4810/72. The peaks show ammonium [M + NH4]+m/z 1170.46, sodium [M + Na]+m/z 1175.54, and potassium [M + K]+m/z 1191.54 adducts of cereulide. For details of purification and analysis, see Material and methods.

Conductance and ion selectivity of the cereulide in BLM

The cereulide, 0.5 ng·mL−1, was applied on the cis side, and a negative potential of 50 mV was applied from the same side. The cereulide caused an increase in the conductance of BLM using KCl in both compartments. In contrast to gramicidin D cereulide showed no sudden changes in conductance indicative of channel activities (data not shown). In Fig. 3 the conductance is plotted as a function of the cereulide concentration: there was a linear relationship indicating a stoichiometric ratio between cereulide and the carried cation. Valinomycin gave a corresponding relationship (data not shown).

Figure 3.

Figure 3.

Dependence of the conductance on the concentration of cereulide. Both compartments contained 100 mm KCl. The transmembrane potential was −50 mV applied on the cis side. The diameter of the membrane hole was 0.5 mm.

Table 1 shows the ion selectivities using various salts. It is seen that the cereulide is highly selective for K+, whereas it was three orders of magnitude less for Na+ and H+. The conductance for Ca2+ was also tested and found to be two orders of magnitude lower.

Table 1. Ion selectivity of the lipid bilayer in the presence of cereulide. Conditions correspond to those in Fig. 4 with 100 mm concentrations of monovalent cation salts in both compartments. In the case of Ca2+, the concentration of the cation was 10 mm. In the experiment on H+, HCl was added to 50 mm NaCl present in both compartments. The cereulide concentration was 0.5 ng·mL−1 on the cis side.
Na+< 1.4 × 10−3
H+< 2.8 × 10−3
Ca2< 3.0 × 10−2

Figure 4A shows the zero current potential as a function of a KCl gradient. The experimental points fit almost ideally the theoretical curve (dashed line) of Nernst′ potential for an ideal K+ selectivity. Fig. 4B shows the influence of NaCl on the zero potentials. It is seen that no potentials developed (within the experimental error ± 1 mV) when NaCl was added to the 100 mm KCl solution (dashed line) whereas large potentials were formed when KCl was added to NaCl solutions (solid line). These data confirm the data in Table 1 that the ratio of Na+ permeability to K+ permeability of the bilayer modified by the cereulide is less than 1.4 × 10−3. The n-Octanol/water partition coefficient, logKow, for cereulide was calculated to be 7.46 and 5.99 for valinomycin, the former thus being more hydrophobic.

Figure 4.

Figure 4.

Zero-current potentials as function of the KCl and NaCl gradients. The zero-current potential was recorded as the potential applied to cause the current to become zero. (A) the KCl gradient was varied: the initial [KCl] in both compartments was 100 mm, The diameter of the membrane hole was 0.5 mm and the concentration of KCl in the trans compartment was kept constant, while [KCl] in the cis compartment was varied. r, Ratio of [KCl] in the cis to trans compartments. (B) The effect of NaCl was studied. Solid line: KCl was added to the cis compartment with 100 mm NaCl being present initially in both compartments. Broken line: NaCl was added to the cis compartment with 100 mm KCl present in both compartments.

Effects of cereulide on RLM

Swelling experiments were carried out in media containing 100 mm of various salts buffered with 10 mm Hepes, pH 7.3. Under nonenergized conditions both cereulide and valinomycin stimulated the swelling in the presence of potassium salts in nitrate-containing media but at most only slightly in the presence of acetate (Fig. 5). The swelling commenced after a lag in the case of valinomycin. No swelling was induced by valinomycin or cereulide when sodium or choline salts were used (data not shown).

Figure 5.

Figure 5.

The effect of cereulide and valinomycin on the swelling of nonrespiring RLM. The medium contained 100 mm KNO3 or acetate (trace Ac), 10 mm Hepes, pH 7.3, 1 µg·mL−1 antimycin A; and RLM were added to give a concentration of 0.275 mg of protein·mL−1. Additions: 1 µg cereulide·mL−1 (traces Ac and C); 0.8 µg valinomycin·mL−1 (trace V; arrows). Downward deflection indicates decrease in A520, i.e. swelling.

The energization of RLM induced by addition of the respiratory substrate succinate led to the formation of a Δψ in RLM, seen as a quenching of the rhodamine 123 fluorescence, Fig. 6A. In a medium containing 21 mm K+ both cereulide and valinomycin reduced the Δψ that was not further changed by the addition of a protonophorous uncoupler. In a medium in which Na+ had replaced K+, addition of the cereulide had no effect. When nonrespiring mitochondria were used in this medium, addition of the cereulide or valinomycin caused the formation of Δψ that was slowly decaying, Fig. 6B. Addition of 1 mm KCl at this point decreased Δψ.

Figure 6.

Figure 6.

The effect of cereulide and valinomycin on theΔψin RLM. RLM (1.1 mg·mL−1), were suspended in a medium containing 200 mm sucrose, 20 mm KCl, 10 mm Hepes, pH 7.3, 1 mm KH2PO4, 0.2 mm MgCl2, 2 µm rotenone and 0.1 µm rhodamine 123 (SKH) or in a medium in which Na+ salts were used instead of K+ (SNaH). The fluorescence was followed at room temperature with the excitation wavelength of 507 nm and emission wavelength of 527 nm; downward deflection indicates formation of Δψ. (A) Succinic acid (2 mm) neutralized with Tris base was added (S). (B) No respiratory substrates were present. Additions: 0.24 µg cereulide·mL−1 (C); 0.2 µg valinomycin·mL−1 (V); 10 µm carbonyl cyanide p-trifluoromethoxyphenylhydrazone, 1 mm KCl.

Using a medium without K+, addition of 0.36 µg cereulide· mL−1 (or valinomycin, data not shown) inhibited succinate respiration slightly (Fig. 7, right trace). Addition of 1 mm KCl then stimulated respiration clearly. Addition of a further aliquot of K+ increased the stimulation. In a medium containing 20 mm KCl, addition of cereulide stimulated respiration, a further addition had no effect (Fig. 7, left trace). Valinomycin had similar effects (data not shown).

Figure 7.

Figure 7.

The effect of cereulide and valinomycin on the respiration of RLM. RLM (0.60 mg·mL−1) were suspended in a medium similar to that in Fig. 6, but containing 1 mm Mg2+, 0.65 mm EGTA, 2 mm Tris/succinate and 2 µm rotenone; the fluorescent indicator was omitted. Additions: 0.36 µg of cereulide·mL−1 (C). The oxygen consumption was followed polarographically with the aid of an oxygen electrode at 25 °C. In the right trace a medium was used in which Na+ salts were used instead of K+, and 1 mm KCl added (arrows).


The structure of the cereulide is analogous to that of valinomycin

Cereulide (-OLeu-Ala-OVal-Val-)3[1] was found to be a cyclic dodecadepsipeptide resembling valinomycin (-Val-OAla-Val-OVal-)3[1].

The cereulide is a K+-specific ionophore

Figure 3 shows an increase in conductance induced by the cereulide indicating that cereulide has formed a complex with K+. The cereulide is cation-specific. The data in Fig. 4 and in Table 1 show a high selectivity for K+ over Na+, Ca2+ and H+. The cereulide thus has a cation selectivity like that of the known K+ carrier valinomycin [5,16].

Mitochondrial effects of the cereulide

The swelling seen in the mitochondria of boar spermatozoa mitochondria (Fig. 1) could be due to several mechanisms, i.e. activation of phospholipase A2[17], the permeability transition that is stimulated by Ca2+ and prooxidants [18], or ionophoretic uptake of K+ driven by the transmembrane potential [19]. The properties of cereulide indicating that it is a K+-specific ionophore like valinomycin makes the latter mechanism the most probable. This is supported by the experiments carried out on RLM. The mechanism of uptake of a cation can be studied by using suitable anions. Swelling in the presence of an anion that is capable of penetrating the membrane, such as one with a net negative charge, indicates a uniport mechanism, whereas swelling in the presence of acetate, that can enter the matrix as acetic acid, indicates an antiport mechanism involving protons [20]. Figure 5 shows that cereulide, like valinomycin, mediates uptake of K+ by a uniport mechanism in the presence of nitrate. No swelling was seen with Na+ as the monovalent cation.

The Δψ measurements in RLM also support the mechanism of cereulide acting like valinomycin. Figure 6A shows the formation of Δψ in succinate respiration, and the collapse of Δψ caused by cereulide or valinomycin in the presence of K+ but not in the Na+-containing medium. The absence of an effect in the latter shows that cereulide has no uncoupling effect, i.e. it does not increase the proton conductance. This may explain the observation that valinomycin and cereulide did not deplete spermatozoa of ATP in spite of the fact that both cause mitochondrial swelling [9]. The collapse of Δψ in RLM occurred at a very low concentration of cereulide, 0.5 ng·mL−1. At this same concentration of cereulide boar spermatozoan motility became inhibited [9] indicating that the motility of the sperm cell was regulated by Δψ in mitochondria. It is known that motilities of boar and human spermatozoa, but not bovine spermatozoa, are exclusively dependent on the functioning of mitochondria [9]. Figure 6B shows that in the absence of K+ in the medium, addition of cereulide or valinomycin caused the formation of a diffusion potential of K+ that is present in the matrix. The valinomycin-induced formation of a diffusion potential according to the Nernst equation has been widely used to estimate Δψ in mitochondria [19,21].

The conclusion of cereulide being a K+-specific ionophore similar to valinomycin was also borne out by the respiration experiments. Addition of the substances to a mitochondrial suspension in a K+-free medium had little effect on the respiratory rate or even inhibited it slightly (Fig. 7). It was found earlier that the emetic toxin acted like a protonophorous uncoupler in stimulating respiration [3] but that must have been due to the presence of a few mm K+ in the medium since care was not taken to exclude this cation. Addition of a small amount of K+ was sufficient to stimulate respiration in the presence of cereulide (Fig. 7). In the medium which contained 20 mm KCl the addition of cereulide stimulated respiration.

Toxicological aspects

Cereulide is the causative agent of the most serious, frequently fatal, type of food poisoning caused by B. cereus[1,3]. In this study we have shown it to be a potent K+ ionophore, as was already anticipated by Agata et al. [1], who first noticed its structural analogy to valinomycin, but did not report its ionophoretic properties. Valinomycin is a product of Streptomyces sp. [22] and has no known involvement in food poisoning.

Recent research on several bacterial protein toxins indicates that their toxicity mechanisms are probably due to their capacity to induce formation of K+ channels. The haemolytic toxin of enterohemorrhagic E. coli O157:H7 forms a cation channel in asolectin membranes, with high selectivity for K+ over Na+ or other monovalent cations [6]. B. thuringiensis CrylA(a) is an insecticidal toxin and kills by forming K+ channels [7] in the larval gut [23]. In mammalian cells valinomycin is known to interfere with vital cellular signal pathways by attenuating protein tyrosine phosphorylation [24] and to induce apoptosis by collapsing the mitochondrial Δψ[25]. We have shown here that cereulide is at least as potent a K+ ionophore as valinomycin, suggesting that this property may be responsible for the toxic effects. To our knowlegde, cereulide is the first nonproteinaceous ionophore shown to be involved in food poisoning.


The authors thank J. Helin for sharing his expertise in MS and M. Ritama for technical assistance. The study was supported by grants from the Helsinki University Fund for Centers of Excellence, the European Union contract number ENV4-CT97–0379 and Academy of Finland to M. S. S.-S. and to N.-E. L. S. and M. S. S.-S. from the Academy of Finland for cooperation with Russian research groups, and from the Finnish Society of Sciences and Letters.