From soil to gut: Bacillus cereus and its food poisoning toxins


  • Editor: Fergus Priest

Correspondence: Per Einar Granum, Department of Food Safety and Infection Biology, Section for Food Safety, Norwegian School of Veterinary Science, PO Box 8146 Dep., N-0033 Oslo, Norway. Tel: +47 22 96 48 45; fax: +47 22 96 48 50; e-mail:


Bacillus cereus is widespread in nature and frequently isolated from soil and growing plants, but it is also well adapted for growth in the intestinal tract of insects and mammals. From these habitats it is easily spread to foods, where it may cause an emetic or a diarrhoeal type of food-associated illness that is becoming increasingly important in the industrialized world. The emetic disease is a food intoxication caused by cereulide, a small ring-formed dodecadepsipeptide. Similar to the virulence determinants that distinguish Bacillus thuringiensis and Bacillus anthracis from B. cereus, the genetic determinants of cereulide are plasmid-borne. The diarrhoeal syndrome of B. cereus is an infection caused by vegetative cells, ingested as viable cells or spores, thought to produce protein enterotoxins in the small intestine. Three pore-forming cytotoxins have been associated with diarrhoeal disease: haemolysin BL (Hbl), nonhaemolytic enterotoxin (Nhe) and cytotoxin K. Hbl and Nhe are homologous three-component toxins, which appear to be related to the monooligomeric toxin cytolysin A found in Escherichia coli. This review will focus on the toxins associated with foodborne diseases frequently caused by B. cereus. The disease characteristics are described, and recent findings regarding the associated toxins are discussed, as well as the present knowledge on virulence regulation.


Who is Bacillus cereus? It is a quiet soil dweller that thrives in a diversity of habitats or a part of the intestinal flora of different animals. It has the ability to withstand time and harsh environments because it can form endospores that are resistant to heat, dehydration and other physical stresses. When allowed access to mammalian tissues it is an opportunistic pathogen that may cause severe local or systemic infections such as endophthalmitis and septicaemia (reviewed in Drobniewski, 1993; Kotiranta et al., 2000), and its close relative Bacillus anthracis is infamous for its potential to cause the severe disease anthrax (Mock & Fouet, 2001). Bacillus cereus is commonly present in food production environments by virtue of its highly adhesive endospores, spreading to all kinds of foods. It produces a range of virulence factors that may cause unpleasant disease in humans when present in food or the gastrointestinal tract and it is one of the major foodborne pathogenic bacteria, although in most cases disease is mild and of short duration. Interestingly, the spectrum of potential B. cereus toxicity ranges from strains used as probiotics for humans (Hong et al., 2005) to highly toxic strains reported to be responsible for food-related fatalities (Mahler et al., 1997; Lund et al., 2000; Dierick et al., 2005). The bacterium causes two types of gastrointestinal disease, the diarrhoeal and the emetic syndromes, which are caused by very different types of toxins. The emetic toxin, causing vomiting, has been characterized and is a small ring-formed peptide (Ehling-Schulz et al., 2004b), while the diarrhoeal disease is caused by one or more protein enterotoxins, thought to elicit diarrhoea by disrupting the integrity of the plasma membrane of epithelial cells in the small intestine. The three toxins that have been implicated as aetiological agents of the diarrhoeal disease are the pore-forming cytotoxins haemolysin BL (Hbl), nonhaemolytic enterotoxin (Nhe) and cytotoxin K (CytK) (Beecher & MacMillan, 1991; Lund & Granum, 1996; Lund et al., 2000). These cytotoxins are part of a virulence regulon that is activated by the transcriptional regulator PlcR (Lereclus et al., 1996; Gohar et al., 2002); however, it is becoming increasingly evident that other regulatory factors are involved, playing a role in determining the pathogenic potential of individual strains.

The organism: characteristics and identification

The ‘B. cereus group’, also known as B. cereus sensu lato, is an informal but widely used term describing a genetically highly homogeneous subdivision of the genus Bacillus, comprising six recognized species: B. cereus sensu stricto, B. anthracis, Bacillus thuringiensis, Bacillus mycoides, Bacillus pseudomycoides and Bacillus weihenstephanensis. The type strain of B. cereus sensu stricto is American Type Culture Collection (ATCC) 14579, which was isolated from air in a cow shed more than one hundred years ago (Frankland & Frankland, 1887). Bacillus thuringiensis is distinguished from B. cereus by the production of insecticidal δ-endotoxins during sporulation, and is commercially used for biological control of insects in crop protection (Aronson & Shai, 2001). Bacillus anthracis causes the fatal animal and human disease anthrax, and has in recent years become known for its use as a biological weapon (Mock & Fouet, 2001; Jernigan et al., 2002). The species B. mycoides and B. pseudomycoides are phenotypically differentiated from B. cereus by rhizoidal colony shape and fatty acid composition (Flugge, 1886; Nakamura, 1998). Bacillus cereus was originally described as a mesophilic organism, growing between 10 and 50 °C and with an optimum temperature of 35 and 40 °C (Johnson, 1984; Claus & Berkeley, 1986). During the last few decades, increasing numbers of psychrotolerant B. cereus strains were described, which led to the description of a new psychrotolerant species within the B. cereus group, named B. weihenstephanensis. This species is characterized by the ability to grow below 7 °C but not at 43 °C, and specific signature sequences in 16S rRNA and cold-shock protein genes (Lechner et al., 1998). Bacillus cereus sensu stricto comprises all strains of the B. cereus group that do not belong to any of the other species due to the absence of distinctive traits.

The word bacillus means small rod, and cereus can be translated from Latin to mean wax-like. The name reflects the easily recognizable morphology of B. cereus when viewed in the microscope or on blood agar plates. Bacillus cereus is a large (1.0–1.2 μm by 3.0–5.0 μm) Gram-positive rod-shaped bacterium which grows on common agar media to large colonies (3–8 mm diameter) with a rather flat, greyish and ‘ground-glass’ appearance, often with irregular borders. On blood agar, the colonies are surrounded by zones of β-haemolysis (Kramer & Gilbert, 1989), the size of which is often large, but can vary depending on culturing conditions.

Most strains will form endospores within a few days on commonly used agar media. Bacillus cereus spores are ellipsoidal, centrally or paracentrally placed, and do not distend the cell (Gilbert & Kramer, 1986). Employing phase contrast microscopy or spore staining techniques, the placement and morphology of the spores are much used criteria to distinguish the species of the genus Bacillus (Fritze, 2002). Other commonly used features for identification are motility, haemolysis, carbohydrate fermentation (B. cereus does not ferment mannitol) and the very active lecithinase (phospholipase) production (Johnson, 1984). Various plating media are used for the isolation, detection and enumeration of B. cereus from foods, including MYP (mannitol-egg yolk-phenol red-polymyxin-agar) and PEMBA (polymyxin-pyruvate-egg yolk-mannitol-bromthymol blue-agar) (Holbrook & Anderson, 1980; Mossel et al., 1967). In addition to selective compounds like polymyxin, these media utilize the bacterium's lecithinase production (egg-yolk reaction giving precipitate zones) and lack of mannitol fermentation. A thorough description of these media is found in Kramer & Gilbert (1989). More recently, chromogenic media have been developed for several food pathogens, including B. cereus (for instance Cereus–Ident-Agar from heipha Dr Müller GmbH, and chromogenic B. cereus Agar from Oxoid Ltd). These new media have been evaluated together with standard plating media by Fricker et al. (2008).

The dilemma in B. cereus group taxonomy

Although B. anthracis, B. cereus and B. thuringiensis are differentiated by phenotypic characteristics and pathological properties, genome sequencing data have shown that they are closely related, both in gene content and synteny (Helgason et al., 2000; Rasko et al., 2004), and their 16S rRNA gene sequences share greater than 99% similarity (Ash et al., 1991). Phylogenetic studies based on chromosomal markers show that there is no taxonomic basis for B. cereus and B. thuringiensis having separate species status (Carlson et al., 1994; Helgason et al., 2004; Hill et al., 2004; Ko et al., 2004; Priest et al., 2004; Guinebretière et al., in press), while B. anthracis can basically be considered a clone of B. cereus (Keim et al., 2000; Hill et al., 2004; Muzzi et al., 2007). The distinguishing features between the species are encoded by genes located on plasmids, which are well-recognized as highly mobile genetic elements, also within the species of the B. cereus group (Thomas et al., 2000; Van der Auwera et al., 2007). Bacillus thuringiensis is defined by the presence of plasmids carrying cry genes encoding δ-endotoxins, while B. anthracis carries two large plasmids encoding the two main virulence factors of this species; pXO1 encoding the anthrax toxin complex and pXO2 encoding the poly-γ-d-glutamic acid capsule, as well as the positive regulator of the virulence factors AtxA, located on pXO1 (Leppla, 2006). The importance of plasmids as virulence determinants within the B. cereus group is also demonstrated by the recent discovery that the genetic determinants of the B. cereus emetic toxin, the ces genes, are present on a large plasmid (Hoton et al., 2005; Ehling-Schulz et al., 2006a). This plasmid was observed to be almost exclusively present in a single monomorphic cluster of B. cereus sensu stricto strains (Ehling-Schulz et al., 2005a), although cereulide-producing strains have been described which differed from the highly homogeneous cluster in genotypic and phenotypic properties (Apetroaie et al., 2005). Supporting this observation, a recent study employing multilocus sequence typing (MLST) identified cereulide-producing strains belonging to a phylogenetic cluster different from the main monomorphic emetic cluster (Vassileva et al., 2007). Additionally, two B. weihenstephanensis isolates were demonstrated to produce cereulide and contain the cesB gene, even though detection of the plasmid was not reported (Thorsen et al., 2006). The emetic type of B. cereus has been suggested to differ from nonemetic strains in properties such as starch hydrolysis, haemolysis, lecithinase reaction and temperature limits for growth (Andersson et al., 2004; Ehling-Schulz et al., 2004b; Carlin et al., 2006).

The dilemma in definition of species within the B. cereus group is therefore that the principal virulence factors that distinguish B. thuringiensis and B. anthracis from B. cereus do not correlate with phylogeny studies based on chromosomal markers, as illustrated by the phylogenetic tree prepared using MLST shown in Fig. 1. An interesting exception is a newly discovered cluster comprising only three known strains, including the B. cereus strain NVH 391/98 responsible for three deaths due to diarrhoeal disease (Lund et al., 2000). MLST analysis and genomic sequencing have indicated that this group is sufficiently far from the main B. cereus group cluster to warrant novel species status (Fagerlund et al., 2007; Lapidus et al., 2007;Fig. 1), and the name ‘Bacillus cytotoxicus’ has informally been proposed for these strains (Lapidus et al., 2007). These three strains can not be distinguished from the other B. cereus group strains based on virulence factors, but they are able to grow at temperatures 6–8 °C higher than the mesophilic B. cereus strains, making them thermotolerant representatives of the B. cereus group (Sorokin et al., 2007; Guinebretière et al., in press). In comparison, the species B. weihenstephanensis was described to distinguish psychrotolerant B. cereus strains from mesophilic strains. Several typing methods have suggested that B. weihenstephanensis strains group in a separate clade within the B. cereus group along with B. mycoides strains (Cherif et al., 2003a, b; Priest et al., 2004; Sorokin et al., 2006; Guinebretière et al., in press). However, psychrotolerant B. cereus group strains do not always conform to the B. weihenstephanensis species criteria (Stenfors & Granum, 2001; Stenfors Arnesen et al., 2007), and a genetic group composed of psycrotolerant B. cereus and B. thuringiensis strains has been identified, which is phylogentically distant to the B. weihenstephanensis clade. Interestingly, the temperature tolerance limits for strains within the B. cereus group appear to correlate with different phylogenetic clusters (Guinebretière et al., in press).

Figure 1.

 Neighbor-joining phylogenetic tree prepared from the concatenated housekeeping gene sequences of different strains of the Bacillus cereus group. The tree was based on the MLST scheme described at the University of Oslo's B. cereus group MLST website ( Genetic distances were estimated using the Kimura model and bootstrap confidence values were generated using 1000 permutations. Bootstrap values (in %) are shown next to the appropriate nodes. Bc, B. cereus; Bt, Bacillus thuringiensis; Bw, Bacillus weihenstephanensis; Bm, Bacillus mycoides.

The discussion of classification of B. cereus group strains is not only of academic and taxonomic interest, but also relates to issues concerning public health. For example, B. cereus isolates harbouring B. anthracis virulence factors have been detected in cases of severe anthrax-like illness (Hoffmaster et al., 2004). Furthermore, while B. cereus is widely recognized as a food poisoning organism, B. thuringiensis is used as a biological insecticide for crop protection. However, because genes encoding the cytotoxins associated with diarrhoeal disease and other opportunistic B. cereus infections are generally chromosomally encoded, they are present in all species of the B. cereus group, although they are silent in B. anthracis (Mignot et al., 2001). In particular, B. thuringiensis has a similar distribution and expression level of genes encoding extracellular virulence factors as B. cereus (Damgaard, 1995; Rivera et al., 2000; Swiecicka et al., 2006), and has caused human infections similar to those caused by B. cereus (Samples & Buettner, 1983; Jackson et al., 1995; Damgaard et al., 1997; Hernandez et al., 1998; Ghelardi et al., 2007b). Food poisoning caused by B. thuringiensis is probably under-reported, as methods for identification of B. cereus group strains in food and clinical settings do not distinguish between B. cereus and B. thuringiensis (Granum, 2002). Therefore, unless otherwise stated, in the remainder of this review the discussion related to B. cereus also applies to B. thuringiensis and B. weihenstephanensis strains.

It has been proposed that B. cereus, B. thuringiensis and B. anthracis should be considered one species based on genetic evidence (Helgason et al., 2000), but no consensus on this matter has been reached. This ambiguous taxonomic state of the B. cereus group illustrates the difficulties encountered with species definition within bacterial systematics, in particular in the genomic era. Traditional phylogenetic analysis of B. cereus group taxonomy is furthermore complicated by extensive horizontal gene transfer between strains (Cardazzo et al., 2008). However, whereas genetically the B. cereus group could be considered one species, a good argument for retaining the current nomenclature is the principle that ‘medical organisms with defined clinical symptoms may continue to bear names that may not necessarily agree with their genomic relatedness so as to avoid unnecessary confusion among microbiologists and nonmicrobiologists’ (Stackebrandt et al., 2002), according to rule 56a(5) in the Bacteriological Code (Lapage et al., 1992).

Reservoirs and lifestyles

Bacillus cereus is described as being of ubiquitous presence in nature and can be found in many types of soils, sediments, dust and plants (Gilbert & Kramer, 1986; Kramer & Gilbert, 1989; von Stetten et al., 1999; Kotiranta et al., 2000; Schoeni & Wong, 2005). Spores may be passively spread and thus found also outside natural habitats. It is believed that B. cereus sensu lato exists in soil as spores, and germinates and grows when brought in contact with organic matter or an insect or animal host. Interest in the ecology of this bacterium spurred a study showing that B. cereus could germinate, grow and sporulate in soil, thus demonstrating a saprophytic life cycle (Vilain et al., 2006). Furthermore, a multicellular phenotype with a filamentous mode of growth was observed and suggested to be a means of translocation through soil (Vilain et al., 2006). A multicellular, filamentous mode of growth has also been observed in the gut of insects. The intestines of insects were suggested as a habitat for B. cereus when sporeforming bacteria, later identified as B. cereus, were isolated from guts of different soil-dwelling arthropod species, where the bacteria appear to exist in symbiosis with their invertebrate host (Margulis et al., 1998). The role of the insect gut microbial communities as a natural niche for part of the B. cereus life cycle is further discussed by Jensen et al. (2003), and it is also suggested that the existence of different morphological modes used by B. cereus, such as the filamentous mode, may be adaptations to different life cycles like the ‘normal’ cycle of life as a symbiont or the more infrequent pathogenic life cycle with rapid growth.

Bacullus cereus has been reported to be present in stools of healthy humans at varying levels (Johnson, 1984; Kramer & Gilbert, 1989; Yea et al., 1994; Jensen et al., 2003). Its ubiquitous low level presence in environments, feed and foods would ensure B. cereus a transient presence in the mammalian gut (Kramer & Gilbert, 1989). However, genomic data from the B. cereus type strain ATCC 14579 and from B. anthracis suggested that their metabolic capacity is more adapted to the use of proteins as a nutrient source than carbohydrates, and furthermore that genes for establishment within a host were conserved (Ivanova et al., 2003; Read et al., 2003). Adding another nuance to the picture, a recent genomic and phenotypic comparison between B. cereus strains ATCC 14579 and ATCC 10987 revealed that ATCC 14579 actually has the capacity to metabolize a larger number of carbohydrates than what was initially believed based on genomic analysis alone (Mols et al., 2007). These data suggest that in addition to a full life cycle in soil, where it is richly present, B. cereus is also adapted to a lifestyle in a host, as a pathogen or perhaps as a part of intestinal flora, as well as to growth in foods. The possible adaptation of B. cereus to the environment of the animal gut could be the basis of their proposed probiotic effect. Such use can not uncritically be considered safe for humans because all B. cereus strains are able to produce at least one of the toxins associated with diarrhoeal disease (Duc le et al., 2004; Hong et al., 2005). However, certain strains producing negligible amounts of toxin at 37 °C have been approved for probiotic use by the European Food Safety Authority (EFSA).

Being present in so many environments, it is expected that B. cereus should also be found in water; however, there are not many data on the presence of B. cereus in water sources, and standard methods for the detection from water are not available. Norwegian surface waters were investigated for presence of B. cereus spores, and cytotoxic strains were isolated from several rivers (Østensvik et al., 2004). This suggests the possibility that the water supply may be a means by which B. cereus enters the food-processing chain.

Transfer from soil to food

Bacillus cereus can be isolated from a remarkable range of different foods and food ingredients, including rice, dairy products, spices, dried foods and vegetables (Kramer & Gilbert, 1989). Cross-contamination can distribute spores or cells to other foods, such as meat products (Johnson, 1984; Gilbert & Kramer, 1986; Granum, 2007). At harvest, B. cereus cells or spores may accompany plant material into food production areas and establish on food-processing equipment. Bacillus cereus is a common contaminant of milk (Johnson, 1984; Kramer & Gilbert, 1989; Andersson et al., 1995; Te Giffel et al., 1997; Lin et al., 1998), and it can cause a defect known as sweet curdling in dairy products. Spores or cells of B. cereus can contaminate udders of cows during grazing (Andersson et al., 1995), or enter the dairy farm through bedding material or feed (Kramer & Gilbert, 1989). In a recent study, high counts of B. cereus were found in the upper layers of dairy farm bedding (Magnusson et al., 2007).

Bacillus cereus spores represent a huge advantage for the organism, allowing attachment, as well as survival of heat treatment or other procedures which remove species of vegetative bacteria which could otherwise outgrow B. cereus. Strain differences in spore characteristics, such as hydrophobicity, exosporium and appendages, have been shown to significantly affect the ability of the spore to adhere to surfaces such as food processing lines (Wiencek et al., 1990; Tauveron et al., 2006; Faille et al., 2007). Bacillus cereus spores are not necessarily removed by regular cleaning of surfaces (Andersson et al., 1995; Faille et al., 2002). The ability of B. cereus to enter yet another lifestyle when forming biofilms (Wijman et al., 2007) is most likely of importance for its persistence in food industry equipment, such as dairy pipelines. The biofilm protects spores and vegetative cells against inactivation by sanitizers (Ryu & Beuchat, 2005).

Modern large-scale food production technology, with extended use of refrigeration as a means of conservation, has created a cold niche well suited for bacteria that are not very competitive, but that can survive heat treatment and also grow at low temperatures. For instance, B. weihenstephanensis as well as B. cereus and other Bacillus species are frequently isolated from dairy products and environments which extensively use cooling as a means of controlling the growth of microorganisms (Wong et al., 1988; Te Giffel et al., 1997; Larsen & Jørgensen, 1999). In addition to dairy products, lightly heat-treated foods with extended refrigerated storage also represent a new and favourable environment for B. cereus group species.

Considering the ubiquitous presence of B. cereus, its resilient spores, and the nonfastidious nature of this microorganism, no type of food with pH>4.8 (Gilbert & Kramer, 1986) can be excluded as a possible vehicle or as representing a risk of food spoilage or foodborne disease. Failure by consumers to follow basic food preparation rules, i.e. slow or inadequate cooling, storage at ambient temperature or prolonged heat-keeping at <60 °C, may allow growth of B. cereus and is commonly part of the story in cases of foodborne disease.

Characteristics of foodborne disease

Two distinct foodborne disease types, emetic and diarrhoeal, are associated with B. cereus. Both are generally mild and self-limiting, although more serious and even lethal cases have occurred (Granum, 1994b; Mahler et al., 1997; Lund et al., 2000; Dierick et al., 2005). Bacillus cereus was established as an organism of foodborne disease in the 1950s, with the first described outbreaks of the diarrhoeal type of disease in hospitals in Norway in 1947–1949 (Hauge, 1950, 1955). Earlier descriptions of disease which could probably be attributed to B. cereus lack the nomenclature and epidemiological framework that would allow this attribution, however there is little doubt that B. cereus has been implicated in foodborne disease historically (Kramer & Gilbert, 1989).

The emetic syndrome was first identified after several outbreaks caused by eating cooked rice in the United Kingdom in the early 1970s (Mortimer & McCann, 1974). This disease is an intoxication caused by the B. cereus emetic toxin, named cereulide, produced in foods before ingestion. The course of the disease is characteristic, with nausea and emesis occurring only a few hours after the meal. The incubation time was originally described as 1–5 h (Kramer & Gilbert, 1989), but more recently to be as short as 0.5 h, and up to 6 h (Ehling-Schulz et al., 2004b; Table 1). The duration of the emetic disease is normally 6–24 h (Ehling-Schulz et al., 2004b). The most important differential diagnosis is intoxication with Staphylococcus aureus enterotoxins, which causes similar symptoms; however, in this disease emesis is commonly accompied by diarrhoea (Seo & Bohach, 2007). Several severe and even lethal cases of emetic foodborne B. cereus disease have been reported (Mahler et al., 1997; Jääskeläinen et al., 2003; Ehling-Schulz et al., 2004b; Dierick et al., 2005; Fricker et al., 2007).

Table 1.   Characteristics of the two types of Bacillus cereus foodborne disease. Adapted from Granum (2007)
CharacteristicsDiarrhoeal diseaseEmetic disease
Type of toxinProtein; enterotoxin(s): Hbl, Nhe, CytK implicatedCyclic peptide; emetic toxin (cereulide)
Location of toxin productionIn the small intestine of the hostPreformed in foods
Infective dose105–108 cfu (total)
The total number required is lower for spores compared to vegetative cells*
105–108 cells g−1 is often found in implicated foods, but live cells are not required for intoxication
Cereulide: 8–10 μg kg−1 body weight (animal models)
Incubation time8–16 h (occasionally >24 h)0.5–6 h
Duration of illness12–24 h (occasionally several days)6–24 h
SymptomsAbdominal pain, watery diarrhoea and occasionally nausea Lethality has occurredNausea, vomiting and malaise. A few lethal cases (possibly due to liver damage)§
Foods most frequently implicatedProteinaceous foods; meat products, soups, vegetables, puddings, sauces, milk and milk productsStarch-rich foods; Fried and cooked rice, pasta, pastry and noodles

The diarrhoeal syndrome is thought to be a toxicoinfection caused by vegetative cells, ingested as viable cells or spores, producing protein enterotoxins in the small intestine (Granum et al., 1993; Andersson et al., 1998a; Clavel et al., 2004). It is easily confused with the foodborne disease caused by another sporeforming bacterium, Clostridium perfringens (Granum, 1990), and typically presents with abdominal pain, watery diarrhoea and occasionally nausea and emesis. The incubation time is over 6 h, normally in the range of 8–16 h, and on average 12 h, but in rare cases longer incubation times have been observed. The duration of the disease is normally 12–24 h but cases lasting several days have been reported (Kramer & Gilbert, 1989; Table 1).

Infective doses

For both types of B. cereus foodborne disease, a relatively high number of cells has generally been found in foods implicated in disease. For the diarrhoeal type, 105–108 cells or spores have been indicated as the infective dose, although lower as well as much higher counts have been found in implicated foods. However, doses as low as 103 B. cereus CFU g−1 of food have been found in foods causing disease (Gilbert & Kramer, 1986). The lowest count of B. cereus found in a confirmed foodborne outbreak in Norway was 200 CFU g−1 food (Granum, 1994a), but further investigations showed that the actual number was closer to 104 CFU g−1 food, and that the underestimate was due to the bacilli being present as aggregated spores (T. Stalheim and P. E. Granum, unpublished data). Lower numbers of spores compared with vegetative cells can probably cause diarrhoeal disease, as the spores are better equipped to survive the passage through the gastric acid (Clavel et al., 2004).

The number of B. cereus cells required to produce sufficient emetic toxin to cause disease has not been determined, but in foods incriminated in cases of emetic disease, levels of 103–1010 CFU g−1 food have been found, in most cases at least 105 CFU g−1 food (Gilbert & Kramer, 1986). In animal experiments, a minimal emesis-causing dose of cereulide of 8–10 μg kg−1 body weight was reported (Agata et al., 1994, 1995; Shinagawa et al., 1995). This was supported by investigations on the cereulide content of a food dish that caused a serious outbreak of emetic disease, suggesting a dose of ≤8 μg kg−1 body weight (Jääskeläinen et al., 2003).

No specific population groups are described as being of special risk for B. cereus foodborne disease. However, individuals with lowered stomach acidity, for example elderly people or people suffering from achlorydria, may be more susceptible to B. cereus diarrhoeal disease, because a larger number of cells are expected to survive gastric transit (Clavel et al., 2004).

Outbreaks of B. cereus foodborne disease

Bacillus cereus is an important cause of foodborne disease worldwide (Clavel et al., 2007; Granum, 2007), although it is probably highly under-reported in official lists of foodborne disease causes. In the European Union, Bacillus species (including non-cereus) were reported to be responsible for 1.4% of foodborne outbreaks in 2005 (Anonymous, 2006). In the years 1992–2006, 45 outbreaks of gastroenteritis attributed to Bacillus spp. in England and Wales were reported to the Health Protection Agency Centre for Infections ( Between 1993 and 1998 in the Netherlands, B. cereus accounted for 12% of foodborne disease outbreaks where a causative agent was identified (Schmidt, 2001). Several factors contribute to the number of foodborne B. cereus disease being largely under-reported. It is a consequence of the generally short and mild course of disease, which does not motivate the patient to seek medical attention. Furthermore, when diagnosed, the disease is not reportable. In addition, cases and/or outbreaks may not always be attributed to B. cereus, because the symptoms of the emetic disease are not easily distinguished from those caused by S. aureus intoxication, and the B. cereus diarrhoeal disease shows the same symptoms as C. perfringens type A food poisoning. The number of cases of B. cereus foodborne disease is reportedly increasing in industrialized countries (Gilbert & Kramer, 1986; Kotiranta et al., 2000). However, as the surveillance systems for foodborne disease differ between countries, it is difficult to compare data and obtain true incidence estimates. Examples of cases and outbreaks are well described in several publications (see e.g. Johnson, 1984; Gilbert & Kramer, 1986; Kramer & Gilbert, 1989; Kotiranta et al., 2000; Granum, 2007).

Many kinds of food have been associated with B. cereus foodborne disease, including spices, meats, poultry, sprouts, rice and pasta (Johnson, 1984; Gilbert & Kramer, 1986; Kramer & Gilbert, 1989; Kotiranta et al., 2000). Different types of food are more commonly associated with either of the two types of foodborne disease: the emetic type of disease has often been connected with consumption of fried and cooked rice (Gilbert & Kramer, 1986), pasta, pastry and noodles (Schoeni & Wong, 2005; Granum, 2007). The diarrhoeal type is commonly associated with proteinaceous foods, sauces and vegetables (Kramer & Gilbert, 1989), meat products, soups, puddings and milk products (Gilbert & Kramer, 1986; Kotiranta et al., 2000; Granum, 2007). Paradoxically, the emetic type of B. cereus strains are thus more often found in starch-rich foods, although their metabolic capacities are not necessarily well adapted to the nutrient composition of these foods, as strains of this type are generally not able to hydrolyse starch (Ehling-Schulz et al., 2004b). Perhaps a lack of readily available nutrients is a factor which can trigger the production of virulence factors?

A somewhat different distribution between countries is observed for the emetic and diarrhoeal diseases, which could partly be a reflection of the association of the two types of disease with different food vehicles: in Japan and the UK, the emetic disease dominates (Gilbert & Kramer, 1986; Shinagawa et al., 1995), while in Northern Europe and North America, the diarrhoeal disease seems more prevalent (Kotiranta et al., 2000). At least part of the difference in disease pattern is probably due to different eating habits, but it is difficult to document whether the distribution is truly different and not a result of reporting differences.

Cereulide, the emetic toxin

The rapid onset of the emetic disease caused by B. cereus, generally from 0.5 to 6 h after consumption of the meal, indicates that this is an intoxication by toxin preformed in the food. Cereulide, the emetic toxin, is a cyclic dodecadepsipeptide with molecular mass 1.2 kDa and the structure: [d-O-Leu-d-Ala-d-O-Val-d-Val]3 (Agata et al., 1994; Ehling-Schulz et al., 2004b). Cereulide is produced by a nonribosomal peptide synthetase, encoded by the 24-kb cereulide synthetase (ces) gene cluster (Ehling-Schulz et al., 2005b), which is located on a megaplasmid related to pXO1 (Ehling-Schulz et al., 2006a). The plasmid was originally named pBCE4810, but has also been referred to as pCER270 (Rasko et al., 2007). Because cereulide is resistant towards acid conditions, proteolysis and heat, it will not be destroyed by gastric acid, the proteolytic enzymes of the intestinal tract or by reheating foods that have been stored at room temperature after a first heating (Johnson, 1984; Agata et al., 1994; Shinagawa et al., 1996).

The mechanism by which cereulide causes emesis in humans has not been definitely determined, although animal feeding experiments have shown a receptor-mediated mechanism. Following release from the stomach into the duodenum, cereulide binds to the 5-HT3 receptor, and stimulation of the vagus afferent causes vomiting in Suncus murinus, an animal model (Agata et al., 1995). Several biological effects of cereulide have been described. The toxin acts as a cation ionopore, like valinomycin, and is therefore able to inhibit mitochondrial activity by inhibition of fatty acid oxidation (Mikkola et al., 1999). This effect of cereulide was the reason for the liver failure in two lethal cases of emetic foodpoisoning where a 17-year-old Swiss boy and a 7-year-old Belgian girl died (Mahler et al., 1997; Dierick et al., 2005). In an experiment where mice were injected intraperitoneally with high doses of synthetic cereulide, massive degeneration of hepatocytes occurred. The serum values of hepatic enzymes were highest on days 2–3 after the inoculation of cereulide, and rapidly decreased thereafter. General recovery from the pathological changes, and regeneration of hepatocytes, were observed after 4 weeks (Yokoyama et al., 1999). Cereulide has also been shown to cause cellular damage (Shinagawa et al., 1996) and inhibit human natural killer cells of the immune system (Paananen et al., 2002).

Cereulide production commences at the end of logarithmic phase during vegetative growth of B. cereus, with the highest level of production at early stationary phase of growth, and the production is not associated with sporulation. Cereulide synthesis takes place at temperatures ranging from c. 12 to 37 °C, although maximal production of emetic toxin appears to occur between 12 and 22 °C (Finlay et al., 2000; Häggblom et al., 2002). However, two isolates belonging to the psychrotolerant species B. weihenstephanensis were recently shown to produce cereulide at 8 °C (Thorsen et al., 2006).

Different foods have varying ability to sustain cereulide production. In infant formulas, levels from 0.02 to 2 μg cereulide mL−1 food were reached after 24 h incubation at room temperature. Cereulide production was influenced by the composition of the formula, with a combination of dairy and cereal ingredients giving higher levels of cereulide production than rice and nondairy ingredients (Shaheen et al., 2006). In another study, cereulide production was quantified in various types of food. In egg and meat products as well as in liquid foods such as milk and soy milk, only low cereulide levels were detected. In contrast, boiled rice and farinaceous foods could sustain production of high levels of cereulide (Agata et al., 2002). In a study by Rajkovic et al. (2006), two B. cereus strains were used to inoculate potato puré, pasta and boiled rice. At static incubation at 28 °C, lower cereulide levels were detected in boiled rice compared with the other two foods, while the counts of B. cereus were equally high (108 CFU g−1) in all three foods.

Differences in levels of cereulide production between strains have also been observed (Häggblom et al., 2002; Rajkovic et al., 2006) and are possibly due to differences in regulation, because the ces genes themselves show only a low level of heterogeneity (Ehling-Schulz et al., 2005b). Furthermore, environmental factors such as oxygen, pH, temperature and the presence of specific amino acids have been described to influence the production of cereulide (Agata et al., 1999; Finlay et al., 2000; Ehling-Schulz et al., 2004b; Jääskeläinen et al., 2004), and it has been shown that stationary incubation of food supports a higher level of cereulide production compared with aerated incubation (Rajkovic et al., 2006; Shaheen et al., 2006). However, the mechanisms regulating cereulide synthesis are still largely unknown.

Detection of cereulide

A range of different methods have been employed for cereulide detection. In early days, monkey feeding tests were used due to lack of other suitable detection methods (Melling et al., 1976). Different variants of cell culture assays using HEp-2 cells take advantage of the ability of cereulide to cause vacuolization of this cell line (Hughes et al., 1988; Sakurai et al., 1994; Finlay et al., 1999). A boar sperm biological assay was developed based on inhibition of boar sperm motility due to the mitochondria-damaging activity of cereulide (Andersson et al., 1998b, 2004; Hoornstra et al., 2003). Rat liver mitochondria were utilized to establish a quantitative detection method based on the ability of cereulide to uncouple mitochondrial respiratory activity (Kawamura-Sato et al., 2005). Neither of the abovementioned tests specifically detects cereulide, and currently this can only be done conclusively by rather laborious and costly HPLC-MS analysis (Häggblom et al., 2002). Nonetheless, in contrast to the use of live animals, or the somewhat labour-demanding cell culture assays, the boar sperm bioassay is easily performed and is well suited for screening B. cereus isolates (Andersson et al., 2004; unpublished data from the Norwegian national reference laboratory). Furthermore, it was shown to correlate well with HPLC-MS analysis within a range of cereulide concentrations (Häggblom et al., 2002) and with LC-MS analysis (Shaheen et al., 2006).

The genes encoding the biosynthetic apparatus for production of cereulide appear to be restricted to emetic toxin-producing strains (Ehling-Schulz et al., 2005b, 2006a), rendering PCR techniques highly relevant for identifying potentially harmful strains. The first PCR assay for detection of emetic strains was published in 2004 (Ehling-Schulz et al., 2004a), and after identification of the nonribosomal peptide synthetase responsible for cereulide synthesis, a PCR assay specific for the ces genes was developed (Ehling-Schulz et al., 2005b). More recently, a real-time PCR method for use in food, targeting the ces genes, was developed (Fricker et al., 2007).

Cytotoxins associated with B. cereus foodborne disease

The diarrhoeal disease was early on attributed to an enterotoxin because culture filtrates of B. cereus caused fluid accumulation in rabbit ileal loops (Spira & Goepfert, 1972; Glatz et al., 1974). This assay has traditionally been considered a decisive test of enterotoxic activity (Bergdoll, 1988). Because the toxins are presumed to elicit diarrhoea by disrupting the integrity of the plasma membrane of epithelial cells in the small intestine, cell culture assays measuring the cytotoxic activity of cell-free culture supernatants is now more commonly used to detect the presence of B. cereus diarrhoeal toxins, and these give a good indication of the cytotoxic potential of B. cereus strains. However, as B. cereus produces a large number of secreted cytotoxins and enzymes that may contribute to diarrhoeal disease, the identity of the enterotoxin(s) is still a controversial topic. The three cytotoxins Hbl, Nhe and CytK are currently considered the aetiological agents of B. cereus diarrhoeal foodborne disease (Beecher & MacMillan, 1991; Lund & Granum, 1996; Lund et al., 2000). Hbl and Nhe are related three-component toxins, while the single-component CytK belongs to the family of β-barrel pore-forming toxins. In addition, several other protein cytotoxins, haemolysins and degradative enzymes have been described that may potentially contribute to the pathogenicity of B. cereus diarrhoeal disease. These include cereolysin O (Kreft et al., 1983), haemolysin II (Baida et al., 1999), haemolysin III (Baida & Kuzmin, 1995), InhA2 (Fedhila et al., 2003) and three phospholipases C (Kuppe et al., 1989).

Before the discovery of Nhe and CytK, Hbl was suggested to be the primary virulence factor in diarrhoea caused by B. cereus. However, food poisoning outbreaks have been caused by strains lacking Hbl (Granum et al., 1996), for example the hbl- and cytK-negative strain B. cereus NVH 0075/95, in which Nhe was first identified (Lund & Granum, 1996), and B. cereus NVH 391/98, from which CytK was originally isolated (Lund et al., 2000). Strain NVH 391/98 belongs to a phylogenetic group distantly related to the main cluster of B. cereus group strains, and was initially thought to be negative for both hbl and nhe. CytK was therefore acknowledged to be the enterotoxin responsible for the outbreak of foodborne disease caused by this strain, which presented as necrotic enteritis (Lund et al., 2000; Dietrich et al., 2005). However, Nhe was later detected in this strain (Fagerlund et al., 2007; Lapidus et al., 2007), and may therefore have contributed to its pathogenicity. Genes encoding Nhe are now thought to be present in all known B. cereus group strains. In contrast, hbl and cytK are present in less than 50% of randomly sampled strains (Ehling-Schulz et al., 2005a, 2006b; Moravek et al., 2006), although higher frequencies of cytK and hbl are observed in studies of clinical and food-associated isolates (Guinebretière et al., 2002; Swiecicka et al., 2006).

Several lines of evidence implicate Nhe as the most dominant diarrhoeal toxin, for example the strong correlation of cytotoxicity with the concentration of Nhe in culture supernatants in 100 B. cereus strains (Moravek et al., 2006) and neutralization of cytotoxicity using an Nhe-specific antibody in 20 of 20 strains tested (Dietrich et al., 2005). Furthermore, the cytotoxic activity of supernatant from the hbl- and cytK-negative B. cereus NVH 0075/95 was abolished by an nhe mutation (Fagerlund et al., 2008), while no change in cytotoxicity was observed upon deletion of hbl or cytK in B. thuringiensis 407 Cry (Ramarao & Lereclus, 2006). However, because the disruption of the hbl operon in the laboratory strain B. cereus ATCC 14579 caused a major reduction in cytotoxic activity towards Vero cells (Lindbäck et al., 1999), the most important toxin may vary between strains. Most likely, multiple toxins may act together to cause gastroenteritis (Callegan et al., 2003; Fedhila et al., 2003). Possibly, toxins may also act synergistically in the gastrointestinal tract, similar to that observed with erythrocytes where phospholipases C have been shown to enhance the lytic activity of Hbl (Beecher & Wong, 2000a).

Bacillus cereus tripartite cytotoxin family

Hbl and Nhe are both three-component toxin complexes, which are restricted to the B. cereus group (From et al., 2005). Hbl consists of the three proteins L2, L1 and B (Beecher & MacMillan, 1991), encoded by the genes hblC, hblD and hblA, respectively, and are cotranscribed from one operon (Heinrichs et al., 1993; Ryan et al., 1997; Lindbäck et al., 1999). Nhe is composed of the proteins NheA, NheB and NheC, encoded by the nheABC operon (Granum et al., 1999). The proteins of Nhe and Hbl show homology, both between the three components of each complex and between the proteins of Nhe and Hbl, with amino acid identities ranging from 18% to 44%. The proteins show no significant sequence similarity towards any other known proteins. The observed similarities between the six nhe and hbl genes suggest that they have originated from a common gene, and that Hbl and Nhe constitute a family of tripartite toxins. Of all six Nhe and Hbl proteins, NheB and NheC show the highest sequence identity (44%), and of the three Hbl proteins both NheB and NheC are most similar to component L1 (Fig. 2). This indicates that the latest gene duplication event in the nhe/hbl gene family was the generation of nheB and nheC. However, when it comes to similarities between hydrophobic regions of the proteins, NheA and Hbl L2 contain no hydrophobic segments, NheB and Hbl L1 each contain a hydrophobic segment of 54 and 60 amino acids, respectively, while NheC and Hbl B each contain one shorter stretch of 25 and 17 hydrophobic amino acids (Table 2). These hydrophobic regions are located in corresponding positions in the proteins, and were previously predicted to be one or two transmembrane helices (Granum et al., 1999; Schoeni & Wong, 2005). Despite the similarities between the two toxins, co-operation between the components of Hbl and Nhe appears to be limited (Lund & Granum, 1997).

Figure 2.

 Neighbour-joining tree showing the molecular relatedness of the Hbl and Nhe proteins from Bacillus cereus ATCC 14579. The tree was generated as described in Fig. 1 but using the proteins sequences of Hbl and Nhe from B. cereus ATCC 14579. Scale bar indicates 10% divergence.

Table 2.   Selected characteristics of the Hbl and Nhe toxin components
  • *

    Hydrophobic sequences are from B. cereus ATCC 14579.

NheADetected by the TECRA-BDE kit from Tecra
Does not contain hydrophobic segments
NheBThe only Nhe component that bound directly to the Vero cell surface
Contains a hydrophobic segment of total length 54 amino acids (aa 234–287): AIIIGSSVATALGPIAIIGGAVVIATGAGTPLGVALIAGGAAAVGGGTAGIVLA*
NheCExcess concentration inhibits cytotoxicity
Produced in lower amounts than NheA and NheB
Contains a hydrophobic segment 25 amino acids in length (aa 227–251), which contains a pair of cysteine residues: MVIAGGVLCVALITCLAGGPMIAVA*
Hbl L2Detected by the BCET-RPLA kit from Oxoid
Does not contain hydrophobic segments
Hbl L1Excess concentration inhibits haemolysis in blood agar
Possibly produced in lower amounts than L2 and B under certain conditions
Contains a hydrophobic segment of total length 60 amino acids (aa 234–293): VLAWSIGGGLGAAILVIAAIGGAVVIVVTGGTATPAVVGGLSALGAAGIGLGTAAGVTAS*
Hbl BExcess concentration inhibits haemolysis in blood agar
Reaction with erythrocytes is the rate-limiting step of haemolysis by Hbl
Contains a hydrophobic segment 17 amino acids in length (aa 237–253):
The crystal structure has been determined (Protein Data Bank entry 2nrj)

For both Nhe and Hbl, all three components are necessary for maximal biological activity (Beecher & MacMillan, 1991; Beecher et al., 1995b; Lindbäck et al., 2004). Nevertheless, using proteins purified from culture supernatants, 10–15% haemolysis has been observed for Hbl in the absence of either L1 or L2 (Beecher & MacMillan, 1991; Beecher & Wong, 1994c), and limited toxic activity has been observed for NheA and NheB in the absence of NheC (Lindbäck et al., 2004). However, because the limited lysis in the blood agar diffusion assay observed using only B and L1 was abolished when recombinant B component was used (Heinrichs et al., 1993), and recombinant clones containing NheA and NheB alone were not cytotoxic (Lindbäck et al., 2004), it is likely that these observations were due to the presence of minute amounts of copurified Hbl L1, L2 or NheC, respectively. However, it has not been established whether limited lytic activity can occur for example at elevated concentrations of two of the three components of each toxin complex.

Biological activity and mechanism of Hbl

Hbl was originally purified from B. cereus strain F837/76, isolated from a postoperative wound (Turnbull et al., 1979). It was initially thought to be a binary toxin composed of B, the ‘binding’ component, and L, the ‘lytic’ component (Beecher & MacMillan, 1990), but upon further examination, the three components L2, L1 and B were identified (Beecher & MacMillan, 1991). Molecular properties of the Hbl proteins are summarized in Table 3. Hbl has been determined to cause fluid accumulation in rabbit ileal loops (Beecher et al., 1995b), show dermonecrotic activity, vascular permeability (Beecher & Wong, 1994b), cytotoxic activity towards Vero cells and retinal tissue (Beecher et al., 1995a; Lund & Granum, 1997), and haemolytic activity towards erythrocytes from several species (Beecher & MacMillan, 1990; Beecher & Wong, 2000a). A model for the action of Hbl has been proposed from studies of Hbl activity towards erythrocytes, including osmotic protection experiments, suggesting that the three components independently bind to erythrocytes, and then assemble into a membrane-attacking complex which lyses erythrocytes by a colloid osmotic lysis mechanism by forming a transmembrane pore (Beecher & Wong, 1997).

Table 3.   Molecular properties of Hbl proteins
Hbl componentSignal peptide*
(amino acids)
Mature protein
(amino acids)
Predicted MW of
mature protein (kDa)
MW from
Estimated pI
F837/76ATCC 4579KBAB4 (Hbla)F837/76ATCC14579KBAB4 (Hbla)F837/76ATCC 4579KBAB4 (Hbla) F837/76ATCC 4579KBAB4 (Hbla)

The stoichiometry of the three components of the Hbl complex forming the proposed transmembrane pore has not been determined, although a positive response in the rabbit ileal loop assay was obtained when the three Hbl components were injected in equimolar amounts (Beecher et al., 1995b). However, maximal haemolytic activity appeared to occur also when the concentration of either Hbl L1 or L2 was lower than the concentration of Hbl B (Beecher & Wong, 1994c). Hbl produces a distinct ring-formed (discontinuous) haemolysis pattern when it diffuses from a well in blood agar, where haemolysis begins in a ring away from the well containing Hbl (Beecher & MacMillan, 1991; Beecher & Wong, 1994b). This pattern was suggested to be the result of a mutually inhibitory effect of B and L1 and the slow reaction between the B component and the erythrocyte membrane, which was the rate-limiting step of haemolysis. Thus, in blood agar initiation of lysis begins at a distance away from the source of Hbl, where the priming reaction by B can occur before the B and L1 components accumulate by diffusion to inhibitory concentrations (Beecher & Wong, 1997). Indeed, when the concentration of L1 was reduced compared with that of Hbl B and L2, lysis occurred more quickly near the well containing the Hbl components (Beecher & Wong, 2000b). The molecular basis for the inhibition of Hbl by excess Hbl L1 and B remains elusive, but could indicate that optimal activity is obtained when one or two components is present in reduced amounts relative to the other components, although inhibition of Hbl haemolysis was also seen in suspension assays when the concentration of all three components was increased at a constant ratio (Beecher & Wong, 1994c). Although all three Hbl components have readily been isolated from culture supernatants and often appear to be present in approximately equal amounts (Beecher & MacMillan, 1991; Beecher & Wong, 1994c; Dietrich et al., 1999; Gohar et al., 2005), Hbl component L1 was not always detected on two-dimensional gels (Gohar et al., 2002; Gilois et al., 2007), indicating that, at least during certain phases of growth, it may be produced in lower amounts than L2 and B.

Biological activity and mechanism of Nhe

Nhe was first characterized after a large food poisoning outbreak in Norway in 1995 caused by the hbl-negative B. cereus strain NVH 0075/95 (Granum et al., 1995; Lund & Granum, 1996). Initially, Nhe was thought to be a cytotoxin complex composed of NheA, NheB and a 105-kDa protein (Lund & Granum, 1996, 1997), but the 105-kDa protein was later shown to be a collagenase (Lund & Granum, 1999), not part of the Nhe complex. Sequencing of the nhe operon identified the gene encoding NheC (Granum et al., 1999), which was subsequently confirmed to be a component of Nhe (Lindbäck et al., 2004). Molecular properties of the Nhe proteins are summarized in Table 4. Both NheA and NheB appear to be present in culture supernatants in two forms with slightly differing sizes, where the smallest form represents a further processed variant of the largest form. The smallest forms of NheA and NheB lack 11 and 12 N-terminal amino acids, respectively, in addition to the 26 and 30 residues of their signal peptides (Beecher & Wong, 1994a; Lund & Granum, 1996, 1997). Trypsin digestion of the largest form of NheA yielded a fragment with mobility identical to the smaller one (Lund & Granum, 1997). Both variants of NheA and NheB show similar biological activity (Lund & Granum, 1996, 1997).

Table 4.   Molecular properties of Nhe
Nhe componentSignal peptide* (amino acids)Mature protein (amino acids)Predicted MW of mature protein (kDa)MW from SDS-PAGEEstimated pI
NVH 0075/95KBAB4 (plasmid)NVH 391/98NVH 0075/95KBAB4 (plasmid)NVH 391/98NVH 0075/95KBAB4 (plasmid)NVH 391/98NVH 0075/95KBAB4 (plasmid)NVH 391/98
NheA262624360/34936336341.0/39.841.341.545§/40 and 415.07/5.044.895.12

The maximal cytotoxic activity towards Vero cells was obtained when the molar ratio between NheA, NheB and NheC was c. 10 : 10 : 1. Furthermore, addition of excess NheC inhibited the cytotoxic activity of Nhe against Vero cells, both in B. cereus culture supernatants and using purified proteins (Lindbäck et al., 2004). Presumably, the initial lack of identification of NheC as part of the Nhe toxin was a result of NheC being produced by the bacterium in much lower concentration than NheA and NheB, in order to obtain a toxin complex with optimal ratio of components. This probably also explains why NheC was not detected among the secreted proteins of B. cereus by two-dimensional electrophoresis analysis (Gohar et al., 2005), and why the NheC antigen titre in B. cereus culture supernatants were considerably lower than the titre of NheA and NheB, although the latter could be related to differing affinities of the antibodies used (Dietrich et al., 2005; Moravek et al., 2006). An inverted repeat located between nheB and nheC has been suggested to mediate translational repression of nheC resulting in lower expression of nheC compared with that of nheA and nheB (Granum et al., 1999; Lindbäck et al., 2004).

Recently, the nature of the cytotoxic activity of Nhe towards epithelial cells was further examined, showing rapid disruption of the plasma membrane following exposure to Nhe, and formation of pores in planar lipid bilayers (Fagerlund et al., 2008). These results were consistent with the insertion of transmembrane pores rather than activation of endogenous channels. Osmotic protection experiments and measures of increases in cell size upon Nhe exposure further indicated that Nhe causes cell death through colloid osmotic lysis by forming transmembrane pores. Nhe was also shown to have haemolytic activity towards erythrocytes from several mammalian species in suspension assays (Fagerlund et al., 2008). However, the relative levels of haemolytic activity of Nhe and Hbl have not been examined, and it is possible that Nhe was previously found to be nonhaemolytic on bovine blood agar plates (Lund & Granum, 1996) due to lower haemolytic activity compared with that of Hbl.

Several independent attempts to prepare deletion mutants of the complete nhe operon in B. cereus and B. thuringiensis strains have failed (M. Ehling-Schulz, pers. commun.; Ramarao & Lereclus, 2006; Fagerlund et al., 2008), although a B. cereus nheBC mutant has been obtained (Fagerlund et al., 2008) and a B. cereus strain with a frameshift mutation in the 5′-end of nheC has been identified (E. Märtlbauer, pers. commun.). In B. anthracis, which produces a low level of NheA protein despite a nonfunctional PlcR protein, an nheA insertional mutant has successfully been prepared; however, this mutant renders two truncated proteins, one of which lacks only the first 77 amino acids (Mendelson et al., 2004). As the nhe operon is found in every B. cereus group strain examined to date, these observations may suggest the intriguing possibility that nheA could be an essential gene in B. cereus. As cytotoxic activity is not likely to be crucial for cell viability, it could be possible that NheA may have a dual role, with a function essential to cell viability in addition to its role as a secreted toxin component.

The Hbl/Nhe family and ClyA constitute a toxin superfamily

The Hbl and Nhe proteins do not show significant sequence homology towards any other known protein family. However, the crystal structure of Hbl component B determined by a structural genomics consortium (Protein Data Bank entry 2nrj; Fig. 3a) showed remarkable tertiary structure resemblance with the pore-forming toxin cytolysin A (ClyA) (Fig. 3b; Protein Data Bank entry 1qoy; Wallace et al., 2000; Fagerlund et al., 2008). ClyA, also known as HlyE or SheA, is a haemolytic and cytotoxic monooligomeric protein toxin of 34 kDa expressed during anaerobic growth in Escherichia coli, Shigella flexneri and Salmonella enterica serovars Typhi and Paratypi A (Oscarsson et al., 1996, 2002; Ludwig et al., 1999; Wallace et al., 2000).

Figure 3.

 Comparison of the structures of Hbl component B and ClyA determined by X-ray crystallography. (a) Hbl component B, (b) E. coli ClyA. Protein structures are shown in ribbon format, with the β-hairpins in blue. (c) Structural alignment visualized as a 3D superimposition of Hbl B (blue) and ClyA (grey), viewed as a Cα-trace. Figures from Fagerlund et al. (2008).

The crystal structures of ClyA and Hbl B consist of long, four/five α-helix bundles that wrap around each other in left-handed supercoils, and a unique subdomain containing a hydrophobic β-hairpin flanked by two short α-helices. The main structural difference, the orientation of the subdomain (Fig. 3c), may possibly represent two different conformational states that both molecules may adopt, with the subdomain and the main helix bundle being connected by a hinge region. This is supported by the observation that the two structures represent different crystallization states, as Hbl B was crystallized as a monomer while ClyA was a dimer in a head-to-tail conformation, where the subdomain containing the β-hairpin was buried against a second hydrophobic surface patch on the opposite end of the protein structure (Wallace et al., 2000). Hinge movements within the subdomain containing the β-hairpin in ClyA were also suggested by results from electron microscopy showing that the oligomeric pores formed by ClyA were of significantly greater length than the water-soluble monomeric protein structure, indicating significant structural changes upon pore formation (Wallace et al., 2000; Eifler et al., 2006; Tzokov et al., 2006).

NheB and NheC show sufficient sequence identity towards Hbl B for generation of 3D homology models based on the Hbl B structural template. Interestingly, as observed for Hbl B and ClyA, the hydrophobic segments of NheB and NheC correlate with the predicted β-hairpin in the homology models. Despite limited sequence identities, the strong structural and functional similarities suggested that the Hbl/Nhe family and the ClyA family of toxins constitute a new superfamily of toxins (Fagerlund et al., 2008).

Membrane binding and pore formation

To date, no host cell receptor for Nhe or Hbl has been identified, and the nature of a putative receptor is an open question. For Hbl, all three components have been determined to bind individually to the erythrocyte surface in experiments where the addition of one component to erythrocytes, followed by washing and subsequent addition of the two remaining components, resulted in haemolysis regardless of which component was the initial component added (Beecher & Wong, 1997). For Nhe, NheB was the only component for which binding directly to the Vero cell surface could be demonstrated, and this association was inhibited by the presence of excess NheC (Lindbäck et al., 2004). The ability of Nhe to rapidly form pores in synthetic lipid bilayer membranes using low toxin concentration shows that Nhe has innate pore-forming ability in phospholipid membranes (Fagerlund et al., 2008), and it is thus possible that Nhe is not dependent on a protein or carbohydrate receptor for toxin binding and activity. For the structurally related ClyA, the hydrophobic β-hairpin was suggested to be the part of the toxin responsible for membrane interaction (Wallace et al., 2000; Eifler et al., 2006; Tzokov et al., 2006). As the β-hairpins in the Hbl B crystal structure and in the NheB homology model also comprise hydrophobic residues, membrane binding by these proteins could occur by a similar mechanism, whereby reorientation of hinge regions in the subdomain correlates with association of the β-hairpin with the membrane. However, the hydrophobic sections of Hbl B and L1 have also been suggested to serve as mediators of oligomerization (Schoeni & Wong, 2005).

Although both Hbl and Nhe appear to be pore-forming toxins, it is not known how the three components in each complex interact or to what extent they oligomerize in the process of forming a transmembrane pore. Because NheB appears to be the sole component of Nhe that binds to the cell membrane (Lindbäck et al., 2004), it is also possible that Nhe, instead of acting as a classical heterooligomeric pore-forming toxin, may act in a manner reminiscent of the A-B type toxins, in which the catalytic and receptor binding functions reside on separate polypeptides or protein domains. The size of the functional pores formed by Hbl has been estimated by osmotic protection experiments using carbohydrates of increasing size to have an approximate diameter of ≤1.2 nm (Beecher & Wong, 1997), while lipid bilayer experiments and osmotic protection assays have indicated that Nhe and ClyA forms pores of comparable size (Fagerlund et al., 2008), which in the case of ClyA were estimated to be 2.5–3.0 nm in diameter (Ludwig et al., 1995; Ludwig et al., 1999; Oscarsson et al., 1999). The pores formed by Nhe and ClyA were also similar in that they were moderately cation-selective in lipid bilayers (Ludwig et al., 1999; Fagerlund et al., 2008). Given that Hbl and Nhe each require three separate proteins for maximum cytotoxic action it is difficult to predict the pore structures formed by Hbl and Nhe based on those formed by the homooligomeric ClyA. However, assuming that Hbl and Nhe, like ClyA, do not encounter major changes in secondary structure upon pore-formation, it is likely that putative oligomeric pores will be built mainly from α-helices, because β-barrels formed from the β-hairpins would be too short to span the predicted thickness of the target membrane (Eifler et al., 2006).

Heterogeneity and genetic organization of hbl and nhe

It has been reported that Hbl proteins from different strains show a high degree of hetereogeneity. Two homologous sets of all three Hbl components, with distinct physical properties, were isolated from B. cereus strain MGBC145 (Beecher & Wong, 2000b), indicating that two distinct homologues of hbl genes may exist in a single strain. Western blot analysis of several strains has furthermore identified proteins of several sizes for each of the three Hbl proteins, and in one strain three Hbl B protein bands were detected (Schoeni & Wong, 1999). Although it is possible that some of the detected Hbl proteins represent forms processed after secretion, similar to those observed for NheA and NheB (Beecher & Wong, 1994a; Lund & Granum, 1996, 1997), the current availability of genomic sequences from several B. cereus group strains has made it possible to establish that two different types of hbl operons indeed exist.

In what appears to be the most common variant of the hbl operon, the ORF hblB is located immediately downstream of the hblCDA genes. The hblB gene was originally identified when the hblA gene was sequenced in B. cereus F837/76 (Heinrichs et al., 1993), but is probably a pseudogene, as it is not transcribed at a detectable level and the hblCDA mRNA transcript appears to terminate within hblB (Agaisse et al., 1999; Lindbäck et al., 1999). Based on sequence analysis, hblB appears to have been generated by duplication of the first 1092 bp of the 1128 bp long hblA gene and a fusion with an ORF in the 3′ end (Økstad et al., 1999). The hblCDAB operon is highly conserved, with DNA sequence identities of 97–99% between strains. Nevertheless, frameshift mutations in the hblC and hblD genes in strain F837/76 have resulted in the C-terminal ends of the L2 and L1 proteins being eight amino acids longer and 23 amino acids shorter, respectively, in this strain compared with in the strains for which genomic sequences are currently available (Table 2).

Sequence similarity searches revealed a second variant of the hbl operon in B. cereus 03BB108 (GenBank entry ABDM00000000) and in B. weihenstephanensis KBAB4 (GenBank entry CP000903; Lapidus et al., 2007). These two hbl operons consist of only three genes, organized as hblCDA operons lacking the hblB pseudogene. Bacillus cereus 03BB108 also harbours the hblCDAB operon and hence contains two hbl operons. Interestingly, the determined N-terminal sequences of the two sets of Hbl proteins isolated from B. cereus MGBC145 (Beecher & Wong, 2000b) indicated that one set of Hbl proteins originated from an hblCDAB operon, while the second set, denoted Hbla and consisting of proteins L2a, L1a and Ba, appeared to be more similar to the Hbl proteins encoded by the three-gene hbl operons from strains 03BB108 and KBAB4, which correspondingly are referred to as hbla. The hbla genes show 75–82% identity towards the corresponding genes of the hblCDAB operons. There is a greater divergence between the sequences of the two identified hbla operons, showing only 86% sequence identity, than between the more homogeneous hblCDAB operons which show 97–99% sequence identity between strains. Differences in molecular properties between Hbl proteins from B. cereus strains F837/76 and ATCC 14579 and the Hbla proteins from B. weihenstephanensis KBAB4 are listed in Table 2.

The hblCDAB operon is chromosomally encoded and the genomic location is conserved between strains. It is part of a 17.7-kb 11-gene insertion flanked on one side by a degenerate ISRso11 transposase fragment, and on both sides by a direct repeat covering the 3′ end of an uvrC-like gene, suggesting that these hbl genes have been acquired as a mobile genetic element (Han et al., 2006). Other genes in this inserted cluster include gerIABC encoding spore germination proteins, and trrA encoding a transcriptional regulator. In contrast, the genomic location of the hbla operons differed between strains 03BB108 and KBAB4. The KBAB4 hbla operon is chromosomally encoded, and flanked upstream by a β-lactamase gene and downstream by a gene encoding a predicted 99-kDa S-layer domain protein. In contrast, the 03BB108 hbla operon is located in a region containing genes with similarities towards genes present on the pXO1 plasmid, potentially indicating that it could be plasmid-borne. However, as the 03BB108 genome sequence is currently not assembled it is not known whether this hbla operon is chromosomally encoded or located on a plasmid.

Heterogeneity for Nhe proteins has been reported for the nhe operons present in the group of strains represented by B. cereus NVH 391/98 (Fagerlund et al., 2007; Lapidus et al., 2007). The nhe operons in these strains show only about 77% identity towards the nhe operons in the main B. cereus group cluster, in correspondence with the remote phylogenetic relationship that these strains show towards other B. cereus group strains (Fagerlund et al., 2007; Fig. 1). In contrast, the identities between nhe operons from other strains are on average about 90%, but approach 100% between strains belonging to the B. anthracis and emetic B. cereus clonal clusters. The greatest differences between nhe operons from different strains reside in the intergenic regions, in particular in the region between nheB and nheC, which also varies in length between strains. All currently available genome sequences of B. cereus group strains contain a single nhe operon, except B. weihenstephanensis KBAB4, which contains two distinct nhe operons. One of these operons is similar to the nhe operons found in the other B. cereus group strains, and likewise chromosomally encoded. The second operon, however, is located on a 400-kb megaplasmid named pBWB401 (GenBank entry CP000904; Lapidus et al., 2007). The sequence identity of this operon towards the other known B. cereus group operons is only around 58%. The Nhe proteins encoded by the plasmid-borne nhe operon are thus more divergent from the known chromosomally encoded Nhe proteins than the Hbl and Hbla groups of proteins are divergent from each other, as illustrated in the tree representation in Fig. 4. Differences in molecular properties between the Nhe proteins from B. cereus strains NVH 0075/95, NVH 391/98, and those encoded by the B. weihenstephanensis KBAB4 megaplasmid pBWB401 are listed in Table 3.

Figure 4.

 Neighbor-joining tree showing the molecular relatedness of different variants of Hbl and Nhe proteins from selected Bacillus cereus sensu lato strains. The tree was generated as described in Figs 1 and 2. B′ indicates the deduced protein sequence of hblB. NheAp, NheBp and NheCp denotes the Nhe proteins encoded on the KBAB4 megaplasmid pBWB401. Scale bar indicates 10% divergence.

For Hbl, functional differences have been determined between the two Hbl toxins isolated from B. cereus MGBC145, because the Hbla variant did not produce the ring-formed haemolysis pattern in blood agar diffusion assays characteristic for Hbl, although both toxins showed haemolytic activity (Beecher & Wong, 2000b). The biological activity of the Nhe toxins encoded by the recently described nhe operons in strain NVH 391/98 and on the KBAB4 megaplasmid remains to be determined, as these proteins have not been isolated and characterized. Interestingly, the only B. weihenstephanensis strain sequenced to date, KBAB4, contains both variants of the nhe operon and the rare hbl variant, hbla. This may reflect the higher level of genetic exchange that is observed among B. weihenstephanensis strains compared with other species of the B. cereus group (Sorokin et al., 2006). In any case, it is clear that a greater diversity than what was initially apparent exists within the family of B. cereus tripartite cytotoxins. Furthermore, the implications for genetic transfer suggested by the observation that the hblCDAB operon may have been acquired as a transposable element and the presence of a new variant of the nhe operon on a plasmid are intriguing.

The B. cereusβ-barrel pore-forming toxins

Bacillus cereus produces two single-component protein toxins that are members of the family of β-barrel pore-forming toxins, namely CytK and HlyII (Baida et al., 1999; Lund et al., 2000). This toxin family includes β-toxin of C. perfringens (Steinthorsdottir et al., 2000) and α-haemolysin of S. aureus (Gouaux, 1998), the best characterized member. These toxins are secreted as water-soluble monomers that associate into oligomeric prepores at the target cell surface, which subsequently insert their pore-forming regions into the cell membrane forming a transmembrane pore (Bhakdi & Tranum-Jensen, 1991). The crystal structure of S. aureusα-haemolysin shows a mushroom-shaped heptamer of about 10 nm in height and diameter, with an amphiphatic β-barrel membrane-spanning pore ranging in size from 1.4 to 4.6 nm (Song et al., 1996).

CytK is a 34-kDa protein with dermonecrotic, cytotoxic and haemolytic activities, and shows similar cytotoxic potency towards cell cultures as Hbl and Nhe (Lund et al., 2000). It is identical to the toxin referred to as haemolysin IV (HlyIV), partially characterized by Beecher et al. (2000). CytK was originally isolated from B. cereus strain NVH 391/98, which was responsible for a severe foodborne outbreak of diarrhoeal disease in a French nursing home in 1998, in which several people developed bloody diarrhoea and three elderly people died (Lund et al., 2000). Based on the necrotic activity of CytK and the apparent lack of both Nhe and Hbl in the outbreak-associated strain, CytK was implicated as the toxin responsible for the severe symptoms and uncharacteristic bloody diarrhoea presenting in this outbreak (Lund et al., 2000; Dietrich et al., 2005). However, because genes encoding nhe were later identified in this strain (Fagerlund et al., 2007; Lapidus et al., 2007), contribution by Nhe to the pathogenicity of B. cereus NVH 391/98 cannot be excluded.

In accordance with the remote phylogenetic relationship of B. cereus NVH 391/98 towards the majority of other B. cereus group strains, the CytK protein from these two groups of strains show only 89% sequence identity (Fagerlund et al., 2004, 2007). Characterization of the two variants of CytK showed that the CytK protein from NVH 391/98 had five-fold greater cytotoxic activity towards Caco-2 and Vero cells than the most common CytK variant, represented by CytK from B. cereus NVH 1230/88, which was initially named CytK-2. The differences in cytotoxicity correlated with the most common CytK variant forming a greater number of small-sized pores in synthetic lipid bilayers (Fagerlund et al., 2004). The greater cytotoxic activity of the CytK protein from NVH 391/98 offered a plausible explanation for the severe food poisoning outbreak caused by this strain. However, while NVH 391/98 was shown to have an exceptionally high level of cytK expression (Brillard & Lereclus, 2004), B. cereus NVH 883/00 harbouring the same cytK variant as NVH 391/98 was noncytotoxic in a cell culture assay due to low amounts of toxins produced under the tested conditions (Fagerlund et al., 2007). This implies that the level of virulence gene expression is probably more important than the presence of specific gene variants for determining the level of cytotoxicity of a particular strain.

HlyII, the second β-barrel pore-forming toxin of B. cereus, is haemolytic and cytotoxic towards human cell lines (Andreeva et al., 2006), but has never been implicated as the toxin responsible for diarrhoea caused by B. cereus. It has been suggested that this may be due to a trypsin digestion site in the β-loop constituting the transmembrane domain of the toxin, resulting in inactivation by trypsin in the small intestine (Lund et al., 2000), as is observed for β-toxin of C. perfringens (Granum, 1990), but this remains to be tested experimentally. The toxin, with a deduced mass of 42.3 kDa, was originally characterized by Sinev et al. (1993) and Baida et al. (1999). Compared with other known members of the family of β-barrel pore-forming toxins family, HlyII has a 94-amino acid, C-terminal extension not required for pore formation or haemolytic activity (Baida et al., 1999; Miles et al., 2002). Interestingly, HlyII expression appears to be independent of PlcR, the central transcriptional regulator for virulence genes in B. cereus, which is required for transcription of hbl, nhe and cytK. This may suggest that HlyII may have a different role than Hbl, Nhe and CytK in B. cereus.

Both CytK and HlyII have been shown to form anion-selective channels with functional diameters of c. 7 Å in planar lipid bilayers (Hardy et al., 2001; Miles et al., 2002; Andreeva et al., 2007). Because they readily form pores in synthetic lipid bilayers, both toxins have innate pore-forming ability in phospholipid membranes. It is thus unlikely that protein or carbohydrate receptors are absolute requirements for binding and lysis by CytK and HlyII, and it has correspondingly been reported that HlyII has no specific receptor on erythrocytes (Andreeva et al., 2006). However, for the related S. aureusα-haemolysin, phosphocholine appears to be the host cell binding receptor. A low affinity binding site for a phosphocholine head group allows concentration of α-haemolysin toxin monomers in microdomains enriched in cholesterol and sphingolipids (lipid rafts). This results in high local concentrations allowing toxin oligomerization and thus stable membrane-anchored binding to target host cells, giving the appearance that certain cell types have high-affinity toxin binding sites (Valeva et al., 2006). β-toxin from C. perfringens has likewise been shown to concentrate in lipid rafts (Nagahama et al., 2003). Based on the functional similarities within the family of β-barrel pore-forming toxins, it is tempting to speculate that membrane binding of CytK and HlyII may occur by a similar mechanism.

Secretion of cytotoxins from the bacterial cell

CytK and all six components of Hbl and Nhe all contain secretory signal peptides indicating that they are secreted by the general secretory (Sec) pathway, which is considered to be the main translocation system on which bacterial protein secretion relies (van Wely et al., 2001). Nevertheless, it has been suggested that the three Hbl components are secreted from the bacterial cell using the flagellar export apparatus (Ghelardi et al., 2002). This conclusion was based on the absence of all three Hbl components in culture supernatant from a nonflagellated flhA mutant. Observations of reduced Hbl protein levels in culture supernatants have recently also been observed in two additional nonflagellated strains (Ghelardi et al., 2007a) and in an flhF mutant strain also showing reduced numbers of flagella and altered motility behaviour (Salvetti et al., 2007). However, these studies did not address the question of whether the reduced levels of Hbl proteins in culture supernatant were mediated at the transcriptional, translational or post-translational level, or due to a secretion defect. Interestingly, a study by (Bouillaut et al., 2005) demonstrated that the nonflagellated flhA mutant showed a 50% reduction in hbl transcription.

Recent evidence has, however, suggested that all three Hbl proteins, in addition to the Nhe components and CytK, are indeed secreted by the Sec pathway, as inhibition by azide of SecA, an essential component of the Sec translocase, resulted in reduced secretion and intracellular accumulation of the toxin components. In contrast, the nonflagellated flhA mutant showed reduced secretion of Hbl, Nhe and CytK, but the lack of intracellular accumulation of toxin proteins in this strain suggested that the absence of secreted toxin proteins was due to reduced toxin production and not a secretion defect (A. Fagerlund & P. E. Granum, unpublished results).

Detection of B. cereus cytotoxins

Owing to the widespread presence and hardy nature of B. cereus, it must be expected to be present in different foods and raw materials, and thus detection of the bacterium is not always the main issue for food safety purposes. Instead, ability to detect the possibly harmful strains, or their toxic products, is the highly desired goal. As cereulide and the three cytotoxins Hbl, Nhe and CytK are the main known virulence factors in B. cereus foodborne disease, focus has been on their detection.

Antibodies have been produced for the three-component toxins Nhe and Hbl (Dietrich et al., 1999, 2005), and two antibody-based detection kits targeting these toxins are commercially available (Buchanan & Schultz, 1994; Day et al., 1994). The BCET-RPLA kit (Oxoid Ltd., UK) is a semi-quantitative assay detecting, by reversed antibody agglutination, the L2 component of Hbl in foods and in cultures of B. cereus (Beecher & Wong, 1994a). The sensitivity of the test is reported to be 2 ng mL−1 test extract. The TECRA-BDE kit (Tecra International Pty Ltd., Australia) detects the NheA component of the Nhe toxin by an enzyme-linked immunosorbent assay (ELISA) sandwich test (Beecher & Wong, 1994a). The sensitivity reported by the manufacturer is >1 ng mL−1 prepared sample, and the kit is intended for use on foods and environmental samples. Neither of the kits will confirm the presence of biologically active toxin, because only one of each of the three-component toxins is detected. For the third and more recently described toxin CytK, there is at present no commercially available detection kit.

For nonspecific detection and characterization of B. cereus enterotoxins, different laboratory animal and tissue culture assays have been employed. Among the tests involving live animals are the rabbit ileal loop (RIL) test, performed by injection of B. cereus cultures or extracts into ligated rabbit intestinal loops followed by observation of fluid accumulation, the guinea pig skin reaction, and the vascular permeability assay (Kramer & Gilbert, 1989). The use of tissue culture assays for detecting B. cereus enterotoxins has been shown to correlate well with results from traditional methods, and represent a convenient alternative for screening purposes (Gilbert & Kramer, 1984; Thompson et al., 1984; Shinagawa et al., 1991; Jackson, 1993; Fermanian et al., 1996). The cell culture lines used include CHO cells (Buchanan & Schultz, 1994; Beattie & Williams, 1999; Hsieh et al., 1999), McCoy cells (Jackson, 1993; Fletcher & Logan, 1999), Caco-2 cells (Hardy et al., 2001; Rowan et al., 2001) and Vero cells (Lund & Granum, 1996; Dietrich et al., 1999; Prüss et al., 1999; From et al., 2005).

For specific detection of the genes encoding the B. cereus toxins Hbl, Nhe and CytK, several PCR schemes, including multiplex PCR, have been developed (see for instance Mäntynen & Lindström, 1998; Hansen & Hendriksen, 2001; Guinebretière et al., 2002; Yang et al., 2005). Considering the wide distribution of cytotoxin genes among B. cereus strains (Rusul & Yaacob, 1995; Mäntynen & Lindström, 1998; Prüss et al., 1999; Rivera et al., 2000; Guinebretière et al., 2002; Ehling-Schulz et al., 2005a), the use of PCR techniques to identify diarrhoeal strains is of little use for practical food safety purposes, because detection of a toxin gene does not reveal the level of toxin production and thus can not predict the potential pathogenicity of a particular B. cereus strain.

Regulation of cytotoxin expression

In general, production of bacterial toxins is tightly regulated. Bacteria sense their environment and respond by producing virulence factors when they are needed, for example in the host environment or upon encountering nutrient limitation, and by turning off toxin production when it would be a selective disadvantage, as synthesis and transport of toxins requires a considerable amount of energy. Because virulence factors of bacteria are often co-ordinately regulated, toxin genes are frequently members of regulons that include additional genes encoding virulence determinants (Finlay & Falkow, 1997). The majority of the B. cereus protein cytotoxins are members of a regulon controlled by the transcriptional activator PlcR, but it is becoming increasingly clear that additional regulatory mechanisms must be involved.

The PlcR quorum sensing system

PlcR is the major virulence regulator of B. cereus (Lereclus et al., 1996). It is part of a ‘quorum sensing’ system that allows B. cereus to regulate virulence genes in a cell density-dependent manner, and it activates expression of a regulon comprising several extracellular virulence factors, including Hbl, Nhe, CytK, degradative enzymes (phospholipases, proteases) and surface proteins (Lereclus et al., 1996; Agaisse et al., 1999; Gohar et al., 2002). Activation of the 34-kDa PlcR protein is dependent on PapR, a 48 amino acid peptide encoded downstream of plcR, which is thought to be secreted by the Sec pathway and extracellularly processed. The PapR autoinducer peptide is then reimported via an oligopeptide permease (Opp), apparently as a heptamer (Agaisse et al., 1999; Gominet et al., 2001; Slamti & Lereclus, 2002; Declerck et al., 2007). When high bacterial densities are reached, the concentration of PapR inside the cells increases, and PapR then interacts with PlcR facilitating binding of PlcR to a conserved palindromic motif known as the PlcR box (TATGNAN4TNCATA) upstream of target genes to activate their transcription (Agaisse et al., 1999; Slamti & Lereclus, 2002).

PlcR has been crystallized as an asymmetric dimer in complex with PapR, and contains an N-terminal helix-turn-helix DNA binding domain and a C-terminal regulatory domain composed of 11 helices with which PapR interacts. Structure modelling based on small angle X-ray scattering analysis has further suggested that binding of PapR triggers oligomerization of PlcR dimers into a supramolecular structure forming a right-handed spiral that may associate with DNA (Declerck et al., 2007). PlcR is phylogenetically related to all quorum sensors that bind directly to their autoinducer peptide inside the cell, which form a superfamily referred to as the RNPP family, restricted to the Gram-positive class Firmibacteria and the orders Bacillales and Clostridiales (Declerck et al., 2007).

As PlcR is positively autoregulated, activation of PlcR by PapR causes a positive feedback loop, presumably responsible for the sharp initiation of PlcR activation. The initiation of PlcR expression at the transition between exponential and stationary phase in liquid culture indicated that PlcR was also regulated by transition state regulators (Lereclus et al., 1996). The plcR promoter contains two Spo0A binding sites on either side of the PlcR box, and plcR transcription was strongly upregulated in a spo0A deletion mutant, but abolished in sporulation specific medium, indicating that PlcR was repressed by the transcriptional regulator Spo0A∼P (Lereclus et al., 2000), which is responsible for initiation of sporulation in the stationary phase of growth (Phillips & Strauch, 2002). Activation of PlcR thus most likely requires at least two conditions to be fulfilled: (1) that the cell density is high enough for quorum sensing to occur, and (2) that the nutritional state of the cell is such that Spo0A∼P is at a sufficiently low concentration to allow plcR transcription.

The significance of PlcR was demonstrated by showing that a deletion of plcR resulted in a 50% decrease in the amount of proteins secreted at the onset of stationary phase compared with the wild-type B. cereus strain, at which time the majority of secreted proteins were putative virulence factors (Gohar et al., 2002). Furthermore, PlcR was shown to directly influence B. cereus and B. thuringiensis pathogenicity as disruption of plcR caused a strong reduction in virulence against both insect larvae, mice and rabbit eyes (Salamitou et al., 2000; Callegan et al., 2003). However, although virulence was reduced it was not abolished, indicating that additional factors not regulated by PlcR contributed to virulence, and that activation of PlcR was not sufficient to account for the pathogenicity of B. cereus. In B. anthracis, the PlcR regulon is silent due to a nonsense mutation in plcR, resulting in a truncated inactive PlcR protein (Agaisse et al., 1999; Gohar et al., 2005). The PlcR regulon may have been counterselected in B. anthracis due to incompatibility between the AtxA and PlcR regulons, as their simultaneous expression resulted in a sporulation defect (Mignot et al., 2001). Interestingly, the upstream promoter regions of both the hbla operon and the plasmid-borne nhe operon of B. weihenstephanensis KBAB4 contain consensus PlcR-boxes, indicating that these operons are regulated by PlcR. A PlcR-box was, however, not identified upstream of hbla in B. cereus 03BB108.

Regulation of HlyII

HlyII is one of the few secreted virulence factors of B. cereus that does not appear to be regulated by PlcR (Budarina et al., 2004; Gohar et al., 2005). It has instead been shown to be negatively regulated by the transcriptional regulator HlyIIR, encoded immediately downstream of hlyII (Budarina et al., 2004). HlyII expression is additionally predicted to be regulated by the ferric uptake regulator (Fur), as the hlyII promoter contains a Fur binding site overlapping the transcriptional start site (Harvie et al., 2005). Fur regulates iron metabolism and represses genes involved in iron uptake when sufficient iron is present, but upon sensing iron limitation the repression of genes involved in iron uptake and transport is lifted, enabling the bacterium to obtain sufficient iron for growth. Deletion of fur in B. cereus resulted in reduced virulence in an insect infection model, demonstrating a link between virulence and iron metabolism (Harvie et al., 2005). The observations that HlyII has neither been implicated as the enterotoxin responsible for an outbreak of foodborne disease, nor demonstrated to be important in nongastrointestinal infections, in addition to the predicted coregulation with iron metabolism genes, could suggest that the haemolytic action of HlyII is a mechanism by which the bacterium gains access to iron.

Other regulatory mechanisms

It has long been observed that the level of toxic activity produced by a B. cereus culture is dependent on environmental factors such as pH, temperature, glucose concentration and oxygen tension (Glatz & Goepfert, 1976; Sutherland & Limond, 1993). More recently, the regulation of Hbl and Nhe expression has been linked to the metabolic state of the cell, as B. cereus produced more Hbl during fermentative (anaerobic) growth than during respiratory (aerobic) growth (Duport et al., 2004), and a low oxidoreduction potential (ORP) during anaerobiosis strongly favoured Hbl and Nhe production (Zigha et al., 2006). These results seem to reconcile well with the anaerobic, highly reducing fermentative conditions present in the small intestine (Moriarty-Craige & Jones, 2004), where B. cereus must produce toxins in order to induce diarrhoeal disease (Granum et al., 1993). The two-component system ResDE and the transcriptional regulator Fnr, whose primary roles are to modulate the metabolism of the cell in response to oxygen availability and redox conditions, also mediates the regulation of Hbl and Nhe expression in response to these factors (Duport et al., 2006; Zigha et al., 2007). Production of Hbl and Nhe was essentially abolished under all conditions tested in a fnr mutant and in a resE (sensor kinase) mutant with intact resD (response regulator) (Duport et al., 2006; Zigha et al., 2007). ResDE is also a positive regulator of the B. anthracis toxin complexes (Vetter & Schlievert, 2007), and interestingly, Fnr is a positive regulator of E. coli ClyA (Green & Baldwin, 1997), belonging to the same superfamily of toxins as Hbl and Nhe. The control of Hbl and Nhe production by ResDE and Fnr is not mediated through PlcR, and putative ResD binding sites have been identified in the promoter regions of hbl and nhe (Duport et al., 2006).

Hbl and Nhe have also been suggested to be subject to catabolite repression, at least during anaerobiosis, because transcription of hbl was repressed by increasing concentrations of glucose (Duport et al., 2004), and growth on sucrose gave higher levels of Hbl and Nhe production compared with growth on glucose (Ouhib et al., 2006). Genes regulated by catabolite repression harbour catabolite responsive element(s) (cre sites), for which the consensus sequence in Bacillus subtilis has been determined to be TGWNANCGNTNWCA (Hueck & Hillen, 1995) or WWTGNAARCGNWWWCAWW (Miwa et al., 2000). A search of the hbl, nhe and cytK regulatory regions for the presence of this sequence revealed two potential cre sites in the nhe regulatory region (Fig. 5). The concept of B. cereus toxin regulation by catabolite repression is perhaps not unexpected, as from a bacterial point of view, deploying virulence factors to liberate required nutrients does not appear necessary when easily metabolized carbohydrates are available.

Figure 5.

 The nhe operon with promoter and regulatory sites. The indicated regulatory sequences are from Bacillus cereus ATCC 14579. The consensus PlcR-box (Agaisse et al., 1999) and the putative PlcR-box with one mismatch in strain ATCC 14579 (underlined) and six bases between the palindromic flanks of the recognition sequence instead of four as in the established consensus (Granum et al., 1999), and the two predicted cre sites (with mismatches towards the B. subtilis consensus underlined) are shown as boxes. The inverted repeat between nheB and nheC (Granum et al., 1999) is indicated as a stem loop structure. The bent arrows indicate the positions of transcriptional start sites, preceded by putative −10 and −35 regions. The transcriptional start site closest to the nheA gene was identified using RNA isolated from strains NVH 0075/95 and NVH 1230/88 (Lindbäck et al., 2004), while the one further upstream was identified using a plasmid carrying the nhe promoter from Bacillus thuringiensis strain 407 (Agaisse et al., 1999). The scale in basepairs is shown in the lower part of the figure.

Observations also point towards a regulatory link between expression of motility and virulence factor genes in B. cereus. For example, in a plcR mutant, flagellin expression and motility was reduced (Gohar et al., 2002; Callegan et al., 2003), and inactivation of flhA, encoding a component of the flagellar export apparatus, has been shown to affect flagellation, sporulation, secretion of Hbl, transcription of hbl and plcA, and production of Hbl, Nhe and CytK (Ghelardi et al., 2002; Bouillaut et al., 2005; A. Fagerlund & P. E. Granum, unpublished results). Also, Hbl production increases during swarming migration (Ghelardi et al., 2007a), which is a differentiated state where elongated and hyperflagellate swarm cells collectively move across solid surfaces (Henrichsen, 1972; Harshey, 1994). However, the molecular mechanisms that putatively couple the expression of virulence factors to motility have not been elucidated, and it is not known whether motility plays a role in B. cereus infection.

The complexity of regulation of toxin production by B. cereus is also illustrated by the apparent differential regulation of Hbl, Nhe and CytK synthesis, illustrated by the variation in time-course of production of these toxins (Gilois et al., 2007). The highest specific production of Nhe was determined to occur early during exponential growth, while Hbl was produced later, early in the stationary phase of growth (Zigha et al., 2006). Furthermore, the nhe operon appears to contain two transcriptional start sites (Agaisse et al., 1999; Lindbäck et al., 2004;Fig. 5), indicating the presence of two promoters which may potentially be differently regulated. Furthermore, it appears that the temperature of growth affects toxin production in B. cereus group species in a strain-dependent manner (Christiansson et al., 1989; Stenfors Arnesen et al., 2007), although the mechanism of such control is not known. However, the most notable observation regarding B. cereus virulence regulation is the substantial variation in the level of toxin production between individual strains. While certain B. cereus strains have been used as probiotics (Hong et al., 2005), others are the cause of lethal foodborne disease (Mahler et al., 1997; Lund et al., 2000; Dierick et al., 2005). Because the presence of cytotoxin genes or gene variants does not seem to be sufficient to explain the level of virulence of a particular strain with respect to the diarrhoeal type of disease, it is likely that the reason why only some strains of B. cereus appear to be pathogenic lies in strain-dependent differences in regulation of toxin expression. Virulence gene regulation therefore appears be essential for understanding B. cereus pathogenesis.

Concluding remarks

Bacillus cereus shows a wide range of variation in phenotypes and virulence types. The diverse nature of the bacterium is all the more fascinating when the strong elements of likeness within the B. cereus group members are considered. There is a huge contrast between the strongly clonal strains with plasmid-borne virulence factors, such as B. anthracis and the emetic type of B. cereus, and the more diverse strains of B. cereus and B. thuringiensis. As described in this review, the level of virulence between strains is highly variable, ranging from harmless to lethal strains. While the role of cereulide in causing the emetic syndrome of B. cereus is well established, the role of the protein cytotoxins as aetiological agents of diarrhoeal disease appears more complex. Strong evidence points towards the Hbl, Nhe and CytK cytotoxins being the main virulence factors in B. cereus foodborne diarrhoeal disease. However, final proof of their role has not been obtained, and the disease appears to be more complex than for example that observed for C. perfringens type A food poisoning, where a single enterotoxin (CPE) alone accounts for all symptoms. Currently, evidence points towards Nhe being the major cytotoxic membrane-damaging factor secreted by most B. cereus strains (Dietrich et al., 2005; Moravek et al., 2006). However, the difficulties in establishing a single factor as the aetiological agent of gastroenteritis due to B. cereus probably reflects that the basis for the disease is multifactorial, where a number of virulence factors may contribute to the overall cellular damage, possibly in a strain-dependent manner.

The most intriguing toxins of B. cereus are the related three-component pore-forming toxins Hbl and Nhe. These unique toxins appear to be distantly related to the homooligomeric toxin ClyA found in certain species of Enterobacteriaceae, a relationship that was discovered due to the structural similarities between the recently determined crystal structure of Hbl component B and ClyA (Fagerlund et al., 2008). This superfamily of toxins appears to represent the only known pore-forming toxins with mammalian targets that are mainly α-helical in structure. However, while ClyA assembles into a transmembrane pore of identical subunits (Wallace et al., 2000), Hbl and Nhe appear to have evolved into tripartite toxin complexes through gene duplication. Further structural and functional studies will hopefully reveal more about the role of the three proteins in each complex, including the molecular basis of the unusual inhibitory effect on cell lysis by excess concentration of individual toxin components. At least two genetically different operons encoding both Hbl and Nhe have been revealed through genomic sequencing of B. cereus group strains, suggesting that the family of B. cereus tripartite toxins may harbour much greater diversity than originally conceived. Further study remains in order to determine the significance of these newly discovered toxin complex variants in relation to foodborne disease.

Increasing evidence points towards gene regulation being the key to understanding the ecology and pathogenesis of B. cereus. Because B. cereus is not a strict pathogen, it may have developed its regulation of pathogenesis using ‘established’ regulatory systems. Apparently, several systems are involved in B. cereus virulence regulation, in a cross-talk between metabolism and toxin production. As an organism inhabiting a diversity of niches, B. cereus employs a complex network of gene regulation for optimal use of resources in all situations, a complexity which is also reflected in its regulation of foodborne virulence. Expression of B. cereus virulence factors implicated in foodborne disease is temporally controlled in response to cell density, environmental conditions, nutrient availability and the metabolic state of the cell, in addition to being co-ordinately regulated with motility genes. The level of toxin gene expression appears to play a major role in determining the pathogenic potential of a particular strain, while in comparison, the presence of specific toxin genes or gene variants appear to be of less importance. A better understanding of the multiple regulatory mechanisms involved in B. cereus toxin production will help to understand the adaptation of B. cereus to its pathogenic lifestyle, and may prove to be the key for identification of potentially harmful strains.

Like its relative B. subtilis, widely used as a model organism, B. cereus is a species receiving considerable attention where we can learn even more because it exists as both virulent and avirulent types. This may reflect the diverse nature of B. cereus, primarily existing as a soil saprophyte, with physiology well adapted for the intestinal tract, also acting as an opportunistic pathogen involved in local and systemic infections.