Ability of Bacillus cereus Group Strains To Cause Food Poisoning Varies According to Phylogenetic Affiliation (Groups I to VII) Rather than Species Affiliation

The B acillus cereus group (B. cereus sensu lato) can induce diarrhea by producing heat-labile enterotoxins during growth in the small intestine (8) and emesis by producing a single heat-stable peptide toxin called cereulide (1). A suitable test for the detection of emetic toxin producers is available (7). For diarrheal syndrome, three enterotoxins have been described: the cytotoxin CytK (17), the nonhemolytic enterotoxin Nhe (9), and the hemolysin HBL (2). Also, other factors may have a role in the enterotoxicity of certain isolates. A way of evaluating the enterotoxicity produced by bacterial strains is to test the cytotoxicity of culture filtrates on cell lines, such as Vero cells (18) or Caco2 cells (4), which encompass all soluble factors.

Assessing the capabilities to cause food poisoning of B. cereus sensu stricto and other species of the group (B. thuringiensis, B. mycoides, B. pseudomycoides, B. weihenstephanensis, and B. anthracis) may not be feasible because these are not separate genomic species, some being distributed between a few distinct phylogenetic groups (10) and others both belonging to the same phylogenetic group (Table 1). In contrast, the seven phylogenetic groups described in Table 1 cover the entire B. cereus group (sensu lato), are based on genetic delimitation (10), and represent a more homogeneous basis to assess the food poisoning capability. The aim of the present study was to determine food poisoning toxins and global cytotoxicity associated with each of the seven phylogenetic groups to evaluate their food poisoning potential.

A total of 391 independent strains (10) belonging to the B. cereus group were used to determine the presence of known toxin genes. In the first step, total DNA was extracted for each strain as previously described (13). Detection of hbl genes and nhe genes was performed by PCR and Southern blotting (to check false-negative responses) as previously described (11). New primer pairs were also selected to detect nhe genes in the two most distant groups (I and VII), taking into account the sequence polymorphisms observed between the sequence of genome NVH 391-98 (group VII) and those of other sequenced genomes (ATCC 14579, E33L, Ames, ATCC 10987, and KBAB4): 5′-GTAAATGCTGCVGATAGYCAAAC-3′/5′-GGCATVATRTTYCCTGCTGC-3′ and 5′-GGTTCRAAYGCTTTAGTAATGG-3′/5′-ATTCCWGCRTCTTGACTAGC-3′, targeting nheAB genes and the nheB gene, respectively. Strains carrying the cytK gene and its two polymorphic forms, cytK-1 and cytK-2, were detected using a duplex PCR (validated to be without false-positive or false-negative responses) as described previously (12). Cereulide producers were detected using a ces gene-specific PCR (7).

A subset of 97 strains, belonging to the different phylogenetic groups (II to VII), was characterized for cytotoxic activity of culture filtrates. Cytotoxic activity was measured on Caco2 cells as described previously (4). The Caco2 cell viability (V) was measured based on the optical density at 620 nm (OD₆₂₀), as a percentage of total viability (100% viability was based on the OD₆₂₀ obtained for Caco2 cells treated with noninoculated brain heart infusion medium instead of culture filtrate), for serial 2-fold-diluted culture filtrates (dilution [D], 0 to 1/32). The nonlinear regression [V = f(D)] was obtained from duplicate values of the six tested dilutions. This curve and the D value for which V was 50% (50% cytotoxicity index [IC₅₀]) were calculated using the GraphPad Prism software (San Diego, CA). When no cytotoxic activity was recorded, the IC₅₀ was approximated as 1 to allow calculations. The cytotoxic activity (CA) of culture filtrates was expressed as follows: CA = 1 − IC₅₀. All the strains were tested in the course of two independent experiments.

Fisher's test was used to determine whether bacterial populations were significantly distant for the tested character by using the XLSTAT 2008 software package.

As two emetic strains (20) have been recently associated with groups other than the two usually known lineages (6, 21), we examined the distribution of the cereulide-producing strains over all seven phylogenetic groups. In our study, cereulide producers (7) were found only in group III, in subgroup III-2 containing the emetic strain F4810/72, and subgroup III-3 (Table 1), which correspond to the known emetic lineages. None of the 137 tested strains in group VI (B. thuringiensis VI, B. weihenstephanensis, and B. mycoides) carried the ces gene, which suggests that emetic B. weihenstephanensis strains, first observed by Thorsen et al. (20), are probably extremely rare.

n he genes were carried by 100% of the tested strains in each group and subgroup (Table 1), in accordance with data from available sequenced genomes affiliated with the B. cereus group (82 sequenced genomes). We can therefore now definitively assert that nhe genes are a constant part of the B. cereus group strains. In contrast, the hbl gene frequency varied from 40% to 97% between phylogenetic groups I, II, IV, V, and VI, and this operon was seldom carried by strains of phylogenetic group III, particularly in subgroups III-2 and III-3 (emetic strains) and III-4 (containing the B. anthracis strains) (Table 1). Thus, the rarity of hbl genes is not a specificity of the “emetic” lineages as previously published (6) but also concerns other subgroups in phylogenetic group III (III-4). As for hbl genes, the distribution of the cytK gene was disrupted. However, it showed some interesting specificities (Table 1): the c ytK-1 form was specific for the most distant group VII “B. cytotoxicus” (versus the 387 strains tested from the other groups); the cytK-2 form was particularly frequent in mesophilic groups III and IV (except for the 12 tested B. anthracis strains that were negative in the III-4 subgroup), whereas it was rare or absent in the psychrotolerant or moderately psychrotolerant groups (VI, II, and V). The c ytK gene was also absent in group I strains.

The disrupted distribution of hbl and cytK genes is in accordance with what is observed from available sequenced genomes. These findings open the way to new hypothetical mechanisms for the spread history of enterotoxins, such as horizontal gene transfer (HGT) in the B. cereus group. This is coherent with traces of HGT previously observed on the B. cereus genome around hbl genes (14) or with the existence of plasmids carrying enterotoxin-like genes (15, 16). On this assumption, some phylogenetic groups might have failed to acquire cytK or hbl (for example, group VI for the cytK gene or subgroup III-2 for hbl), because they occupy very specific thermal niches unfavorable to cohabitation and exchange of genetic material with other phylogenetic groups.

Low or no cytotoxic activity of culture filtrates was recorded for all the strains of group VI and subgroup III-3 (Fig. 1). In contrast, higher cytotoxic activity (P < 0.001) was recorded for strains of the other groups (Fig. 1a), reaching very high values for a great number of strains in group III. Cytotoxic activity was heterogeneous in group III (Fig. 1b), with high cytotoxic activity in groups III-1 and III-2 and very low cytotoxic activity in subgroup III-3 (P < 0.0001). In addition, the cytotoxicity recorded for strains of subgroup IV-2 (Fig. 1b) was higher than for subgroups IV-1 and IV-3 (P < 0.05) in group IV. Finally, strain 391-98, representing group VII (B. cytotoxicus), exhibited a high level of cytotoxic activity (mean, 0.75).

The enterotoxic potential conferred by each phylogenetic group, independent of species affiliation, has implications for health risk. Based on the frequency of highly enterotoxic strains (generally associated with the diarrheal syndrome) recorded in each group, food poisoning risk should be highest for group III (containing B. cereus III, B. thuringiensis III, and B. anthracis-like strains), particularly subgroups III-1, III-2, and III-4. Subgroups III-2 and III-3 may be also dangerous for their emetic potential. This risk remains high for groups VII, IV, and II, decreases with group V, and seems very low for group VI. Strains belonging to this last group should be the safest with regard to diarrheal syndrome. This is consistent with the absence of strains isolated from food-borne diseases in this group (10). Concerning emetic syndrome, the possible production of emetic toxin by some group VI strains, although presumably very rare, needs to be clarified. Because B. weihenstephanensis and B. mycoides strains belong to group VI, we can also consider these two species as representing a low level of risk, on the condition that they are reliably identified. For B. cereus sensu stricto and B. thuringiensis species, the level of risk is impossible to determine without taking into account phylogenetic affiliation of strains (I to VII), as they can belong to phylogenetic groups II to VI and thus reflect several different levels of food poisoning risk.

In conclusion, affiliation to a phylogenetic group provides a more accurate indication of the risk than affiliation to a current species of the B. cereus group. Associating the phylogenetic number (I to VII) with the species name when identifying new isolates offers a first useful indicator of risk. The challenge is thus to propose reliable methods for identifying the seven phylogenetic groups.

As shown in this study, the validated duplex PCR method (12) permits identification of cytK-1-carrying strains specific to the rare and hazardous group VII (B. cytotoxicus). A rapid method to assign bacterial isolates to groups I to VII is to compare the panC gene sequence from new isolates with those of reference strains published by Guinebretiere et al. (10). For this purpose, we propose the online tool at the following URL: https://www.tools.symprevius.org/Bcereus/english.php . As the panC gene sequence was previously shown to reliably delimit the phylogenetic groups and subgroups (10), this tool allows a rapid and reliable identification (100% exact responses) at the phylogenetic group level (I to VII). It compares a query sequence with those in a database (85 panC sequences for which the phylogenetic affiliations are known). Protocols and a homology search algorithm are available from the website.

This phylogenetic approach takes a first step toward evaluating the potential risk associated with the B. cereus group strains. It could in the future be extended to include the presence or production of pathogenic markers. To date, the ces gene is the most relevant marker for the emetic syndrome. The toxins Nhe and HlyII have been directly or indirectly linked to the diarrheal syndrome (3, 19), but they do not yet explain the pathogenicity of all strains. Some new associated virulence factors (5) have not been evaluated yet as pathogenic makers, and others are doubtlessly still unknown. Further knowledge of virulence-associated factors secreted by the B. cereus group strains is thus currently needed for a more accurate detection of food poisoning strains.