Bacillus anthracis, Bacillus thuringiensis and Bacillus cereus are members of the Bacillus cereus group. These bacteria express virulence in diverse ways in mammals and insects. The pathogenic properties of B. cereus and B. thuringiensis in mammals results largely from the secretion of non-specific toxins, including haemolysins, the production of which depends upon a pleiotropic activator PlcR. In B. anthracis, PlcR is inactive because of a nonsense mutation in the plcR gene. This suggests that the phenotypic differences between B. anthracis on the one hand and B. thuringiensis and B. cereus on the other could result at least partly from loss of the PlcR regulon. We expressed a functional PlcR in B. anthracis. This resulted in the transcriptional activation of genes weakly expressed in the absence of PlcR. The transcriptional activation correlated with the induction of enzymatic activities and toxins including haemolysins. The toxicity of a B. anthracis PlcR+ strain was assayed in the mouse subcutaneous and nasal models of infection. It was no greater than that of the parental strain, suggesting that the PlcR regulon has no influence on B. anthracis virulence. The PlcR regulon had dramatic effects on the sporulation of a B. anthracis strain containing the virulence plasmid pXO1. This resulted from incompatible interactions with the major AtxA-controlled virulence regulon. We propose that the PlcR-controlled regulon in B. anthracis has been counterselected on account of its disadvantageous effects.
Bacillus anthracis, Bacillus thuringiensis and Bacillus cereus are spore-forming soil bacteria that belong to the Bacillus cereus group. These bacteria are of particular interest because they are pathogens. B. anthracis causes anthrax in mammals, B. thuringiensis is highly pathogenic to the insect (for a review, see Schnepf et al., 1998), and B. cereus is involved in various pathologies such as periodontitis, food poisoning and fulminant endophthalmitis (Drobniewski, 1993; Kotiranta et al., 1998; Beecher et al., 2000). The virulence of B. anthracis and B. thuringiensis is dependent upon the production of specific toxins, namely the oedema and the lethal toxins for B. anthracis and the Cry proteins for B. thuringiensis. B. cereus virulence is thought to be a consequence of its secretion of diverse non-specific virulence factors such as phospholipases and enterotoxins (Lund and Granum, 1996; Beecher et al., 2000). Most B. thuringiensis strains also secrete these factors and are opportunistic mammalian pathogens (Hernandez et al., 1999; Salamitou et al., 2000).
Genetic evidence suggests that B. anthracis, B. thuringiensis and B. cereus are members of the same species (Helgason et al., 2000). B. thuringiensis and B. cereus are highly polymorphic (Helgason et al., 1998), whereas B. anthracis is one of the most molecularly monomorphic bacteria known (Keim et al., 1997). B. cereus may be the ancestral species (Helgason et al., 2000) from which B. thuringiensis and B. anthracis emerged by acquisition of extrachromosomal elements such as the toxin-encoding plasmids Cry (B. thuringiensis) and pXO1 (B. anthracis). This would indicate that the pathogenic phenotypes of these bacteria are mainly conferred by the plasmids. However, this is only partially true, as there are major phenotypic differences between B. anthracis on the one hand and B. thuringiensis and B. cereus on the other: B. anthracis does not secrete significant amounts of degradative enzymes, and it is non-motile. These characteristics have even been used to discriminate B. anthracis from the other members of the group (Drobniewski, 1993). Such differences could result from the absence, in B. anthracis, of structural genes that are present in the other group members. Alternatively, the regulation circuits leading to the expression of genes encoding these functions may be different in B. anthracis. Understanding the molecular basis of these differences may reveal how B. anthracis emerged in the B. cereus group.
In B. thuringiensis and B. cereus, the transcription of a large regulon involving genes encoding several non-specific virulence factors is under the control of a pleiotropic transcriptional activator, PlcR (Lereclus et al., 1996; Agaisse et al., 1999; Økstad et al., 1999; Lund et al., 2000). The PlcR-controlled regulon governs the synthesis of toxins including haemolysin BL, which is involved in the food poisoning and endophthalmitis syndromes caused by several strains of B. cereus. This regulon may therefore be responsible for the opportunistic properties of B. thuringiensis and B. cereus (Agaisse et al., 1999; Lindbäck et al., 1999; Beecher et al., 2000) and, indeed, ΔplcR strains of B. thuringiensis and B. cereus have attenuated virulence in a mouse nasal instillation model of infection (Salamitou et al., 2000). Interestingly, the virulence of the B. cereus strain was not completely abolished, implicating PlcR-independent factors and underlining the differences between the group members with regard to pathogenicity (Salamitou et al., 2000).
Remarkably, B. anthracis carries a plcR gene, which has a nonsense mutation, thus encoding a truncated inactive activator (Agaisse et al., 1999). Thus, in this bacterium, a complete regulon of non-specific virulence factors may be inactive because of a single mutation in a transcriptional regulator. Presumably, these functions were lost during recent evolution, because they are inessential or, possibly, they were counterselected because incompatible with the pathogenesis of B. anthracis. The defect in the PlcR regulon in B. anthracis could explain some of its phenotypic specificities (i.e. the lack of haemolysins and degradative enzymes).
We investigated the effect of the expression of a functional PlcR on B. anthracis and, in particular, on the synthesis of non-specific toxins, virulence and sporulation. We discuss the consequences of the results in terms of pathogenesis and evolution.
The PlcR-controlled regulon involves a wide panel of genes in B. anthracis
In B. thuringiensis and B. cereus, genes have been assigned to the PlcR-controlled regulon genetically and as a result of the presence of a highly conserved palindromic motif TATGNAN4TNCATA, a probable PlcR recognition target, in their promoters (Agaisse et al., 1999; Økstad et al., 1999). The location of this motif, relative to the transcription start sites of the target genes, was variable and could be found as far as position −284 (Agaisse et al., 1999). The B. anthracis chromosome sequence was screened for known PlcR target genes and the PlcR recognition site (Agaisse et al., 1999; Økstad et al., 1999): 52 open reading frames (ORFs) were identified. The characteristics of 19 ORFs encoding predictable functions are summarized in Table 1.
Table 1. Predicted PlcR-regulated functions as indicated by genomic analysis of B.anthracisa
Twenty-nine of these ORFs encode proteins without signal peptides. The functions of only six of these could be predicted from similarity with characterized sequences (Table 1). The only cytoplasmic protein in B. thuringiensis and B. cereus known to be under PlcR control is PlcR itself. Our analysis suggests that numerous cytoplasmic functions are controlled by PlcR. They include a eukaryotic-like Ser/Thr kinase. An enzyme of this type has been described in the B. cereus group, but only in virulent B. thuringiensis and B. cereus strains (Guttmann and Ellar, 2000).
Twenty-three ORFs were predicted to encode secreted proteins, consistent with the proposal that PlcR controls the synthesis of many exported proteins; 13 have predictable functions (Table 1). They fall into the three broad classes described previously (Agaisse et al., 1999): degradative enzymes, surface proteins and toxins. Most of the genes described as being PlcR regulated in B. thuringiensis were found in B. anthracis, i.e. the operon encoding phosphatidylcholine-specific phospholipase (PlcB) and sphingomyelinase (CerB) and the genes encoding the non-haemolytic enterotoxin (NHE). Interestingly, the genes encoding the components of the haemolysin BL were absent.
This genomic analysis (Table 1) indicates that the phenotypic properties of B. anthracis are not caused by the absence of structural genes, but may result from the silencing of a regulon.
Expression of PlcR in B. anthracis leads to the activation of a weakly expressed regulon
In B. thuringiensis and B. cereus, plcA encoding phosphatidylinositol-specific phospholipase C, sfp encoding a serine-protease of the subtilase family and a gene encoding a putative surface SLH-containing protein are members of the PlcR-controlled regulon (Agaisse et al., 1999; Økstad et al., 1999). Orthologues of plcA and sfp, and ORF02613, encoding a putative SLH-containing protein, are present on the B. anthracis chromosome. To check the PlcR dependence of their regulation, we assayed the expression of these genes in the presence and absence of an active PlcR from B. thuringiensis. The B. thuringiensis plcR coding sequence and its promoter region were cloned in a replicative multicopy vector, giving rise to pATCR, and transferred into the plasmidless B. anthracis strain 9131. The expression of plcA, sfp, ORF02613 and eag, a PlcR-independent control, was assayed by reverse transcription–polymerase chain reaction (RT–PCR) (Fig. 1). RT–PCR with RNA extracted from the control strain 9131(pAT28) (Table 2) only generated faint signals with plcA-, sfp- and ORF02613-specific primers, suggesting that these genes are expressed weakly (Fig. 1, lanes 1–3). In contrast, RT–PCR performed on RNA extracted from strain 9131(pATCR) generated strong signals of the expected size in all cases (Fig. 1, lanes 5–7). eag expression was similar in both backgrounds (Fig. 1, lanes 4 and 8), confirming that there was no substantial difference in the yield and quality of RNA extractions. Hence, the transcription of predicted PlcR target genes in B. anthracis is low, but was significantly enhanced by a functional PlcR. Therefore, the nonsense mutation of the plcR gene in B. anthracis silences a complete regulon.
Table 2. Oligonucleotides used for RT–PCR experiments.
PlcR induces degradative enzyme activities in B. anthracis
We tested whether the transcription of the sfp and plcA orthologues was associated with the induction of corresponding enzymatic activities. Strains 9131(pAT28) and 9131(pATCR) were grown on appropriate plates. Strong protease and lecithinase activities were secreted by strain 9131(pATCR) (Fig. 2A, plates 1 and 2 respectively). These activities were also secreted by strain 9131(pAT28), but very weakly. These results confirm the findings of the RT–PCR experiment, indicating that protease and lecithinase activities are residual in the wild-type B. anthracis and can be activated further by PlcR.
Protease zymogram analysis was performed to estimate the number and size of proteolytic enzymes induced by PlcR (Fig. 2B, plate 1). The proteolytic profiles of strains 9131(pATCR) and 9131(pAT28) were different: two bands corresponding to proteins of apparent molecular weights of about 40 and 45 kDa were detected in strain 9131(pATCR) (Fig. 2B, plate 1), whereas strain 9131(pAT28) gave only weak bands (Fig. 2B, plate 1). This suggests that probably two distinct proteases are induced by PlcR in B. anthracis. Lecithinase zymogram analysis indicated a single band corresponding to an apparent molecular weight of 30 kDa in strain 9131(pATCR) (Fig. 2B, plate 2). Thus, at least one protein displaying lecithinase activity was induced by PlcR in B. anthracis.
A human erythrocyte-specific sulphydryl-activated cytolysin is induced by PlcR in B. anthracis
In B. thuringiensis and B. cereus, PlcR activates the synthesis of haemolysins and other enterotoxins. The haemolytic activities of strains 9131(pAT28) and 9131(pATCR) were tested against sheep and human erythrocytes (RBCs). Strain 9131(pATCR) displayed a low but significant haemolytic activity against sheep RBCs (80 units ml−1), whereas 9131(pAT28) did not. Only strain 9131(pATCR) generated lytic plaques on sheep blood agar plates (Fig. 3A). The haemolytic activity of strain 9131(pATCR) was 30 times higher on human RBCs (2560 units ml−1; Fig. 3B) than on sheep RBCs. On human RBCs, strain 9131(pAT28) also displayed a weak but significant haemolytic activity (40 units ml−1; Fig. 3B). Hence, in B. anthracis, the major PlcR-controlled haemolytic activity is active on human RBCs.
Bacillus thuringiensis and B. cereus produce cholesterol-binding cytolysins (CBCs), named thuringiolysin O and cereolysin O respectively (Alouf, 2000). We assayed the susceptibility of the haemolytic activity of B. anthracis against human RBCs to cholesterol, a potent inhibitor of CBCs (Alouf, 2000). The haemolytic activities of strains 9131(pAT28) and 9131(pATCR) against human RBCs were completely abolished by cholesterol in a dose-dependent manner (Fig. 3C). Cholesterol partially inhibited the haemolytic activity of strain 9131(pATCR) on sheep erythrocytes; full inhibition could only be obtained by the addition of both cholesterol and EDTA (data not shown). These results suggest that the haemolytic activity against human RBCs only involves CBC(s), whereas that against sheep RBCs results from the combination of CBC(s) and metalloenzyme(s).
As indicated by genomic analysis, a gene (clo) encoding an orthologue of cereolysin O is located downstream from a PlcR motif (Table 1). The translation product of this gene contains the classical ECTGLAWEWWR C-terminal signature of the CBCs. RT–PCR analysis showed that transcription of the clo orthologue is significantly higher in strain 9131(pATCR) (Fig. 3D, lane 2) than in strain 9131(pAT28) (Fig. 3D, lane 1). Thus, B. anthracis is haemolytic towards human RBCs as a result of the activity of a PlcR-activated orthologue of cereolysin O. Furthermore, this activity is already present at a low level in the absence of PlcR.
The PlcR regulon does not influence B. anthracis virulence in two mouse models of infection
PlcR significantly increases the synthesis of non-specific functional toxins in B. anthracis. Thus, full expression of the regulon may have effects on B. anthracis virulence. However, the apparent counterselection of the regulon by evolution suggests that it is inessential for the expression of full B. anthracis virulence. We assayed the virulence of a B. anthracis strain expressing a functional PlcR. To avoid any effects resulting from a non-physiological expression of PlcR, we constructed a strain expressing PlcR from a single-copy insertion into the chromosome: the plcR gene and its regulatory region were inserted into the eag gene encoding the S-layer protein EA1, a locus that does not affect B. anthracis virulence (the LD50 of the eag– derivative of the 7702 strain is 2.5 × 105 spores per mouse), on the chromosome of strain 7702, harbouring the lethal and oedema toxin-encoding plasmid pXO1. Strain 7702CR (which is pXO1+plcR+) was tested for the PlcR-induced phenotypes. As for strain 9131(pATCR), 7702CR displayed PlcR-induced protease, lecithinase and haemolytic activities (data not shown; Fig. 3A), but with a 24 h delay. Such a difference may be explained by the fact that a high gene dosage (pAT28 is a high-copy-number plasmid) probably leads to the overexpression of PlcR. The lethality of strain 7702CR was assayed in the mouse subcutaneous model of infection. The LD50 of 7702CR was ≈ 5 × 105 spores per mouse, which is the same as that of strain 7702 eag– and the parental strain 7702. The toxicity of strains 7702 and 7702CR was also investigated by nasal instillation, a model used to assay B. thuringiensis lethality in mice (Salamitou et al., 2000). Both strains harboured low and equivalent toxicities (LD50 > 108). We conclude that expression of the PlcR regulon does not affect the virulence of B. anthracis overall, at least in our models of infection.
Co-expression of the PlcR- and the AtxA-controlled regulons hampers efficient sporulation in B. anthracis
In B. thuringiensis, the expression of the plcR gene is repressed by the phosphorylated form of the sporulation transition state regulator Spo0A (Lereclus et al., 2000). This suggests a strict temporal control of the PlcR-activated regulon; possibly, its silencing is required for sporulation to begin. The sporulation efficiency of strain 7702CR was only 20% of that of strain 7702 [(0.5 ± 0.1) × 108 spores per NBY plate for 7702CR versus (2.5 ± 0.5) × 108 spores per NBY plate for 7702]. We then investigated the effect of plcR expression from a multicopy plasmid on sporulation efficiency. All the bacilli of a 7702 strain harbouring pAT28 sporulated within 24 h of inoculation (Table 3). In contrast, strain 7702 harbouring pATCR failed to sporulate (Table 3). Most of the 7702(pATCR) cells remained viable (Table 3) and had the rod shape of viable stationary phase cells throughout the experiment (Fig. 4). Therefore, this phenotype is not the result of PlcR-induced loss of viability. In a pXO1– background, the expression of PlcR did not influence sporulation: strains 9131(pAT28) and 9131(pATCR) sporulated with equivalent efficiencies (2.2 × 108 and 2.7 × 108 spores per NBY plate respectively). Expression of PlcR from a multicopy plasmid did not affect the sporulation of B. thuringiensis. These results show that the sporulation defect of strain 7702(pATCR) results from the incompatibility of the PlcR regulon and a pXO1-encoded function.
Table 3. PlcR hampers the sporulation of an atxA+B. anthracis strain.
Many pXO1-encoded functions including the anthrax toxin components are positively regulated by AtxA and comprise the AtxA-controlled regulon (Hoffmaster and Koehler, 1997). The sporulation kinetics of an atxA mutant expressing PlcR [strain BATX1(pATCR)] were assayed as described above (Table 3). Strain BATX1(pATCR) sporulated with the same kinetics as strain 7702(pAT28) with all the bacteria becoming spores within 24 h (Table 3, Fig. 4). This demonstrates that the simultaneous expression of the PlcR- and AtxA-controlled regulons alters the initiation of sporulation in B. anthracis, an overexpression of PlcR leading to a complete block.
In B. anthracis, the plcR gene harbours a single nonsense mutation inactivating a pleiotropic regulator of extracellullar virulence factors (Agaisse et al., 1999). Expression of a functional PlcR in B. anthracis leads to the activation of a large regulon encoding enzymatic and toxin activities including protease, phospholipase and haemolysis. These activities can be attributed as follows. (i) The transcription of sfp encoding a putative serine-protease may account for some of the proteolytic activity. (ii) In B. cereus, the phospholipase activity results from the combined activities of PI-PLC (PlcA) and PC-PLC (PlcB). The plcA and plcB genes are present on the B. anthracis chromosome, and we show that plcA transcription is activated by PlcR. PlcB is a lecithinase. Thus, in B. anthracis, PlcR probably induces the synthesis of both PlcA and PlcB. (iii) Haemolysis of human RBCs is attributable to a CBC that is probably the product of the clo gene orthologue (CLO). The haemolysis of sheep RBCs results from the activities of both CLO and metalloenzyme(s). As PlcB contains zinc in its active site (Martin et al., 1996) and displays haemolytic activity (Gilmore et al., 1989), it may be involved in sheep erythrocyte haemolysis. Our data also suggest that CLO has a higher specificity for human RBCs than for sheep RBCs. Combined with genomic data, these results show that most B. cereus virulence factors are present and functional in B. anthracis. The non-haemolytic phenotype of this bacterium is thus clearly the result of downregulation as a consequence of the plcR mutation.
The PlcR target genes we tested encode activities that are functional. This suggests two possibilities that are not mutually exclusive. (i) The point mutation in the plcR sequence is too recent for deleterious mutations to have accumulated within the target genes. (ii) Selection pressure upon these genes has prevented the accumulation of mutations. PlcR-induced activities were detectable, albeit weakly, in the absence of a functional PlcR. This suggests that these activities are relevant to B. anthracis. Indeed, they may be regulated by other signals. Consequently, it may be interesting to study the contribution of genes such as plcA, plcB and clo to B. anthracis virulence: new virulence factors may thus be identified.
In experimental models, B. anthracis 7702 lethality against mice results from the synthesis of the synergistic lethal and oedema toxins (Brossier et al., 2000). We found that the PlcR regulon did not affect B. anthracis toxicity after subcutaneous injection. Thus, the synthesis of non-specific toxins has negligible effects on B. anthracis toxicity. B. thuringiensis and B. cereus are toxic to mice when administered at high doses (≥ 108) by intranasal instillation (Salamitou et al., 2000). Intranasal B. anthracis is poorly toxic (Salamitou et al., 2000). Any potentiation of B. anthracis virulence by PlcR would therefore be expected to be detected in this model. That this was not the case indicates that the mechanisms of toxicity of B. anthracis plcR+ are not the same as those of B. thuringiensis and B. cereus. This may result from differences in the levels of toxin production. Alternatively, B. anthracis may lack key toxins, one example being haemolysin BL (Lindbäck et al., 1999). The absence of any effect of the PlcR-controlled regulon on B. anthracis virulence is consistent with its evolutionary counterselection: the regulon is inessential and does not contribute to virulence. Such examples of reductive evolution are common, the most extreme case being Mycobacterium leprae, an organism in which major metabolic pathways are not functional (Cole et al., 2001).
However, this does not appear to be the sole explanation: the expression of the PlcR-controlled regulon interferes with that of the AtxA-controlled regulon and inhibits sporulation. Our results suggest that initiation of sporulation is prevented at a step before asymmetric cell division. In B. anthracis, the anthrax toxin and the capsule are major virulence factors, and the absence of either of these elements leads to a 104 times reduction in virulence. Thus, AtxA is a key regulator, as it activates the synthesis of the lethal and oedema toxins in the presence of a CO2 environmental signal and potentiates capsule synthesis. Indeed, an atxA mutant strain is avirulent (Dai et al., 1995). In anthrax pathogenesis, AtxA is considered a central mediator of CO2 signalling and virulence factor activation. Our experiments were conducted in air and, therefore, provide indirect evidence that AtxA is also involved in a non-CO2-dependent manner in the control of target genes. Our results, and the evidence that AtxA also controls pXO1-encoded non-toxin genes (Hoffmaster and Koehler, 1997), highlight the importance of the AtxA-controlled regulon in B. anthracis physiology.
Sporulation is a key process in the B. anthracis infectious cycle, as the spore is the infecting form. Therefore, the PlcR regulon may have been counterselected because it gave B. anthracis a clear selective disadvantage: bacilli that do not sporulate cannot compete with those that do. It is probable that the mutation within plcR occurred after the acquisition of plasmid pXO1 and thus allowed the emergence of a bacterium fully able to persist and propagate.
plcR Bt was inserted into the eag gene as follows: the 1.7 kb SalI–SmaI fragment of pHT304plcRBt was blunt ended and ligated to pSAL322 (Mesnage et al., 1997) previously digested by BamHI and blunt ended.
Construction of recombinant strains
Recombinant plasmids were transferred from E. coli to B. anthracis by heterogramic conjugation (Trieu-Cuot et al., 1987). 7702CR was obtained by allelic exchange between the mutated eag::plcR gene and the chromosomal locus and selected as described previously (Pezard et al., 1991).
The characteristics of the oligonucleotides used for RT–PCR are listed in Table 1. RT–PCRs were carried out on 1 µg of total RNA with the Superscript One-Step RT–PCR kit (Gibco BRL) according to the manufacturer's recommendations. The RNA dependence of the amplifications was checked by conducting PCR without the reverse transcription step. For this, the reaction mix was preheated for 5 min at 95°C to inactivate the RT, and the PCR was then carried out in standard conditions.
Enzyme activity assays
Protease, lecithinase and haemolytic activities were assayed on BHI agar plates containing 5% milk, 5% egg yolk (Difco) and 5% sterile sheep's blood (Bio-Rad) respectively. For zymogram analysis, casein-containing gels or egg yolk-containing gels (5%) were loaded with proteins from culture supernatants (5 µg). After SDS–PAGE, the gels were treated for renaturation as described by Mesnage et al. (2000). Casein gels were stained with Coomassie brilliant blue.
Haemolytic activities against sheep erythrocytes or human erythrocytes were assayed as described previously (Sirard et al., 1997) with supernatants from B. anthracis cultures grown to stationary phase on 0.5% sheep/human erythrocytes. One haemolytic unit is defined as the largest dilution of 1 ml of supernatant that results in complete lysis of the erythrocytes. Cholesterol and EDTA inhibition assays were carried out as follows: solutions containing 26 µM cholesterol (Sigma) or 26 µM cholesterol and 2 mM EDTA were diluted serially twofold and added to approximately 1 haemolytic unit in the presence of 0.5% human/sheep erythrocytes.
In vivo experiments
The toxicities of the B. anthracis strains were assayed on 7-week-old female OF/1 outbred mice (Iffa Credo) by subcutaneous injection and by nasal instillation as described previously (Mesnage et al., 2000; Salamitou et al., 2000).
Sporulation efficiencies were determined as follows: 6–8 × 105 exponential phase B. anthracis cells were used to inoculate NBY medium plates and incubated for 7 days at 30°C. Spores were collected in 1 ml of sterile H2O, heated for 30 min at 65°C to eliminate vegetative forms and counted. The values in the text are the mean values of five independent assays. For sporulation kinetics assays, NBY plates were collected, and suspensions were counted before and after treatment at 65°C.
Protein sequence analysis
Translated ORFs from TIGR were used for blast searches on the non-redundant protein database set at NCBI, USA (http://www.ncbi.ntm.nih.gov/). Peptide signals were predicted using the sigcleave program available at Pasteur, France (http://www.pasteur.fr).
The authors wish to thank T. Read and TIGR for making available the sequences of the PlcR target genes. M. Lévy, F. Brossier and M. A. Lopez-Vernaza were of great help with the animal experiments. M. Haustant gave excellent technical assistance. We wish to thank A. Ullmann for her highly critical reading of the manuscript, as well as P. Goossens for helpful discussions. The work at TIGR is funded by ONR/DOE/NIH/ DERA. This work was supported by DGA 99 34 032/DSP/STTC. T.M. had a fellowship from MENRT.