Multiple-Locus Sequence Typing Analysis of Bacillus cereus and Bacillus thuringiensis Reveals Separate Clustering and a Distinct Population Structure of Psychrotrophic Strains

Strains and growth conditions.All the strains used in this study are shown in Fig. 1. The strains for which experimental sequence data were obtained in this study were given designations beginning with the letters KB. The 115 strains labeled KBAnn, KBBnn, and KBCnn are from the Versailles Collection, which has been described elsewhere (33). They were isolated from the soil of a forest at La Minière near Versailles, France (Ile-de-France region) on 6 December 2000. B. thuringiensis strains were distinguished from B. cereus strains by light microscopic detection of parasporal crystals (33). We also applied MLST analysis to five psychrotrophic B. weihenstephanensis strains—WSBC10204, WSBC10206, WSBC10297, WSBC10311, and WSBC10315 (20, 27) (Fig. 1, KB008 to KB012, respectively)—obtained from S. Scherer (Germany); five strains of B. cereus—41, F4810/72, F0837/76, F2038/78, and ATCC 6464—reported to have smaller than normal chromosomes (4) (KBBI1 to KBBI5, respectively), obtained from A.-B. Kolsto (Norway); a pathogenic B. cereus isolate, NVH 391-98 (21) (KB007), from vegetable puree, obtained from D. Lereclus (France); and B. cereus strain BGSC 6A1(1) (KB002) and four strains of B. thuringiensis, BGSC 4H2, BGSC 4D1, ATCC 35646, and BGSC 4Q7 (5, 8, 9, 30) (KB003 to KB006, respectively), from public collections. We also included the sequences of the loci considered from the entirely sequenced strains B. cereus ATCC 14579 (17) (KB001), B. cereus ATCC 10987(28), and B. anthracis Ames (29) in our analysis.

• Open in new tab
• Download powerpoint

• Open in new tab
• Download powerpoint

FIG. 1.

UPGMA-based dendrogram and BURST-based clonal complexes of the allelic profiles of the 134 strains used in this study. Each strain name is followed by the strain origin marker (consisting of the letters Bt, Bc, Ba, or Bw [phenotypically defined B. thuringiensis, B. cereus, B. anthracis, or B. weihenstephanensis, respectively], or BtF or BtS [strains from the Versailles Collection originating from deep in the forest or the forest edge], and a number indicating the soil sample [33]), the sequence-based strain cluster designation, the cold-growth phenotype (M, P, I, or N [see Materials and Methods for an explanation]), the ST, and the corresponding allelic profile. The allelic profile, in parentheses, contains arbitrary allelic numbers for the clpC, dinB, gdpD, panC, purF, and yhfL loci, respectively. Most of the strains fall into clonal complexes (or groups, designated G1, G2, etc.) of closely related STs, defined as groups of STs in which every ST shares at least four of six identical alleles with at least one other ST in the group. Singletons are STs not belonging to the clonal complexes. Groups marked with asterisks have ancestral allelic profiles identified by START. Relationships between single-locus variants (red), double-locus variants (blue), and the ancestor are indicated by concentric circles or connecting lines of the corresponding color, generated by START (18). (A, B, and C) Allelic profile trees for the T, W, and C clusters, respectively. Green vertical bars indicate clonal complexes of psychrotrophic strains.

All strains were cultured overnight on Luria-Bertani agar plates at 30°C, as previously described (16), and stored at room temperature until MLST analysis. The cells (approximately 10⁷) were picked by toothpick into 300 μl of water and resuspended by vortexing, and 3 μl of the resulting suspension was used for PCR. The ability of strains to grow in the cold was tested by streaking bacteria from plates incubated overnight at 30°C onto fresh agar plates. These plates were then incubated for 14 days at 6°C. Growth was assessed by eye, with the density of bacterial growth compared to that at 30°C and expressed as a percentage of this maximal growth. This procedure was repeated three times, and we estimated the statistical error of this method of growth estimation at 30%. The cold-growth status of strains was designated as follows: P (for psychrotrophic) if growth exceeded 70% of the maximum, M (mesophilic) if less than 30% of the maximum level of growth was observed, I for intermediate growth, and N for strains not tested (see Fig. 1).

PCR amplification and sequencing.We carried out PCR with the Expand Long Template PCR system (Roche Diagnostics) to obtain templates for sequencing. The cycling program was as follows: 94°C for 5 min; 12 cycles of 94°C for 10 s, 55°C for 10 s, and 65°C for 12 min; 24 cycles of 94°C for 10 s, 55°C for 10 s, and 65°C for 12 min in the first cycle, increasing by 15 s with each subsequent cycle. The mixture was subjected to a final extension step for 10 min at 72°C. PCR products were treated for 1 h at 37°C with exonuclease I and shrimp alkaline phosphatase (USB Corporation). Sequencing was performed with an ABI PRISM sequencing kit (Applied Biosystems). Sequencing products were precipitated in ethanol and analyzed with an ABI 3700 sequencer (Applied Biosystems).

MLST.We used multiple-locus sequence typing to compare strains. Alleles at six unlinked loci were identified by sequencing of 350- to 700-bp internal fragments of the genes, and the sequence type (ST) was defined by the string of allele numbers at the six loci. The six genes selected for characterization of the diversity of strains by MLST (22) and the corresponding primers are listed in Table 1. They are evenly distributed over the chromosome of the entirely sequenced strain B. cereus ATCC 14579 (17), referred to here as KB001. The alternative primers used for amplification and sequencing of the strains of the W cluster were based on the sequences of flanking regions of the selected loci for strains KB008, KB009, and KBAB4, considered representative of this cluster. These sequences are available from NCBI under accession numbers DQ301422 to DQ301433.

The sequences from each strain were assembled and consensus sequences generated with xbap software (7). The consensus sequences were compared using the multiple alignment program CLUSTALX 1.8 (31), and phylogenetic trees were generated with NJplot software (24). START software was used for the analysis of allelic data in the MLST context (18). The program decoupe_msf, used for the elimination of identical bases from closely related aligned sequences, was kindly provided by J.-M. Batto (Génétique Microbienne, INRA). The sequences were deposited in the NCBI database and can also be used for BLAST analysis at spock.jouy.inra.fr (click on“ Links” and “BaCerId” in the site menu).

Linkage analysis and prediction of clonal complexes.We assessed the correlation between alleles in the population and used this information to determine clonal status (23) by calculating the index of association (I_A) with START software (18). Clonal complexes or groups of strains, defined as strains with no more than two different alleles, and the ancestral strains were predicted using the BURST algorithm implemented in START.

Nucleotide sequence accession numbers.The sequences determined for the 115 strains studied were deposited in the NCBI database under accession numbers DQ296198 to DQ296463 and DQ300927 to DQ301422.

The phylogeny based on sequence alignment reveals three major strain clusters.Six genes—clpC, dinB, gdpD, panC, purF, and yhfL—were systematically sequenced for 115 strains of the Versailles Collection, including 61 and 54 strains that had previously been identified as B. cereus and B. thuringiensis, respectively (33). These genes were also sequenced for 19 reference strains, as described in Materials and Methods. Phylogenetic analysis based on the sequences of individual loci (not shown) or concatenated sequences and the neighbor-joining method implemented in CLUSTALX (25) indicated that most strains could be assigned to one of three clusters: C, T, and W (Fig. 2). The set of oligonucleotides selected on the basis of available sequences for KB001 and B. anthracis Ames (17, 29) gave no PCR amplification for many cluster W strains (not shown). We therefore amplified entire regions containing the selected genes for strains KB008, KB009, and KBAB4, which are representative of this cluster, using primers corresponding to the coding regions of the nearest conserved genes. We obtained PCR fragments of 1.5 to 5 kb. Sequencing of these fragments by primer walking provided the necessary information for the selection of an alternative set of oligonucleotides, making it possible to amplify the loci of cluster W strains in a reproducible manner. Table 2 shows the statistical characteristics of the entire sequencing data set and a comparison of the characteristics of the three clusters. The assignment of individual strains to these clusters is shown in Fig. 1. Four strains from the forest soil collection could not be assigned to any of the three clusters; these strains were assigned the cluster name“ X.” Strain KB007 was unusual in that it had highly divergent alleles for three genes (clpC, gdpD, and yhfL) but could clearly be assigned to cluster W based on analysis of the other three genes (dinB, panC, and purF).

• Open in new tab
• Download powerpoint

FIG. 2.

Concatenated sequence-based phylogenetic tree. Three clusters containing most of the strains are designated by the letters C, T, and W. Each strain name is followed by the ST number and the number of identical strains corresponding to this ST. Approximate locations of clonal clusters (or groups) are indicated, including G1, G2, and G8, corresponding to the lineages Tolworthi, Sotto, and Kurstaki identified by Priest et al. (26). The small cluster X, containing strains KBAE4, KBBC4, KBCF5, and KBCC8, is not labeled. Bc10987, B. cereus 10987; BAames, B. anthracis Ames.

The six genes sequenced therefore gave consistent separation of the strains into three major, nonoverlapping clusters. We carried out allele profile analysis for each of these clusters, both separately and together. This separation into clusters was also used as the basis for a separate calculation of indices of association.

STs and clonal complexes.Each unique sequence of a gene was regarded as an allele and was assigned an arbitrary allele number. The set of allele numbers for a given strain represents an allelic profile or sequence type, also arbitrarily numbered. Figure 1 shows dendrograms constructed using the unweighted-pair group method using average linkages (UPGMA), with allelic profiles or STs. The strains were resolved into 93 STs, 80 of which were identified only once. The most common, ST-93, was identified 12 times, whereas two others, ST-54 and ST-57, were detected in 11 and 8 isolates, respectively, and 10 STs corresponded to 2 or 3 strains. Most of the 115 isolates of the Versailles Collection and the 19 additional strains analyzed in this work were grouped into 10 clonal complexes, with the isolates of a given complex identical at four, five, or six loci (Fig. 1). The clonal complexes for which the ancestral ST was successfully identified with START software are represented in the form of concentric circles. The ancestral ST is shown in the central circle. STs differing by only one (single-locus variants) or two (double-locus variants) loci are shown separated from the ancestor by one or two circles, respectively. This formalism of the presentation of very close phylogenetic relationships between bacterial strains, and the concept of clonal complexes, which is very useful for the interpretation of MLST data, were suggested by Feil et al. (11). START also uses color coding to show more complex relationships between STs (see the legend to Fig. 1).

The three clusters—C, T, and W—consisted of 15, 55, and 41 strains, respectively, isolated from forest soil. The significance of this formal cluster assignment is made clear by the inclusion in our data set of sequences from independently isolated strains with established phenotypic characteristics. Several B. thuringiensis strains—Bacillusthuringiensis serovar israelensis ATCC 35646 (KB005), B. thuringiensis serovar israelensis BGSC 4Q7 (KB006), Bacillus thuringiensis serovar kurstaki BGSC 4D1 (KB004), and Bacillusthuringiensis serovar canadensis BGSC 4H2 (KB003)—and two strains of B. cereus—ATCC 14579 (KB001) and BGSC 6A1 (KB002)—appeared to belong to cluster T. Moreover, almost all the strains of our forest collection found to synthesize parasporal crystal toxin protein (33) were also found in this cluster (Fig. 1A). These strains are therefore genetically different from the strains in clusters C and W even in the absence of genes encoding entomopathogenic toxins, as in the case of strains KB001 and KB002. In our forest strain collection, only one group (or clonal complex), designated G8, contained a mixture of strains assigned to B. cereus and B. thuringiensis based on phenotype. Cluster X (Fig. 1C) was also a mixed group. It contained both phenotypic classes and was not identified as a clonal complex, because allelic profiles differed at more than two loci. In total, we detected five clonal complexes (G1, G2, G6, G7, and G8) for cluster T, four of which had an identifiable ancestral strain (Fig. 1A). These clonal complexes displayed strain relationships different from those for cluster W (Fig. 1B). Most of the cluster T strains had identical STs or differed from the ancestral strain by only one allele. One possible reason for this is that strains may be clones, and such findings may result from an artifact of strain isolation. All precautions were taken to avoid this (33), and at least two other lines of evidence suggest that this is not the case: (i) only five strains with the same phenotype were isolated from the same soil sample, and (ii) if there were clones, we would expect similar population structures for the three clusters (T, C, and W), whereas these clusters were actually found to have different population structures.

The strains that did not synthesize parasporal crystals detectable by light microscopy belonged to two major clusters: W and C. Most of the cluster W strains could be combined into two clonal complexes (G3 and G5), demonstrating their close genetic relationship (Fig. 1B). Only one genetically different group, designated G10, was identified in cluster W, although six strains, including the pathogenic strain KB007 and the independently isolated psychrotrophic strain KB010 (27), would clearly form another clonal complex if more strains were analyzed (Fig. 1B).

Cluster C contained about one-third of all the tested strains of the Versailles Collection identified as B. cereus on the basis of phenotype (Fig. 1C). This cluster also contained pathogenic B. anthracis Ames, B. cereus ATCC 10987, and emetic strain KBBI2. Two psychrotrophic strains, KB011 and KB012, isolated independently (27), also belonged to this cluster, although testing of these and all other strains of this cluster (see below) showed a tendency to grow less well in the cold. Indeed, our solid-medium test at 6°C characterized strains KB010, KB011, and KB012 as mesophilic, whereas they were classified as psychrotrophic based on culture in liquid medium at 8°C (see below). Cluster C contained fewer strains than cluster W, accounting for the lack of isolation of rare strains closely related to KB011 and KB012. We identified two clonal complexes in this cluster, G4 and G9, but were unable to identify the ancestral STs.

Relationship between strain clusters and psychrotrophy.The ability to grow in the cold is not an intrinsic property of all B. cereus or B. thuringiensis strains. When the Versailles Collection was established, psychrotrophy was not suspected in many strains and was not systematically tested (33). However, the close genetic relationship between strains from clonal complex G3 of cluster W and the psychrotrophic type strains KB008 and KB009 (20) suggests that many strains of this clonal complex may be able to grow at low temperatures. We therefore systematically tested all strains that did not synthesize detectable parasporal crystals (clusters W and C) and most of the strains that did produce such crysals (cluster T) for the ability to grow at 6°C on solid medium. These results are shown on Fig. 1. In this strict test (ability to grow at 6°C), most of the cluster W strains were found to be psychrotrophic. Almost all these strains were assigned to two clonal complexes—G3 and G5—containing 22 and 7 strains, respectively. These findings indicate that the strains that are able to grow in the cold are closely related genetically. Group G10 of cluster W, which was genetically different, contained two mesophilic strains. Genetic characterization of five independently isolated psychrotrophic strains (20, 27), designated KB008 to KB012, revealed that KB008 belonged to the main psychrotrophic clonal complex of the Versailles Collection and that KB009 was closely related. However, three other independently isolated psychrotrophic strains, KB010, KB011, and KB012, did not seem to be closely related to the strains in this collection. KB011 shares two identical alleles with several strains of cluster C (group G9), but several forest strains from this cluster (e.g., KBAC5, KBCA4, and KBCB1) grew poorly in the cold in our test, and others (e.g., KBCA5, KBCE4, and KBCG5) did not grow in the cold at all. In this strict test of growth in the cold (6°C), strains KB010, KB011, and KB012 were actually classified as mesophilic (Fig. 1). However, these strains grew to a high density (optical density at 600 nm, 3) in liquid medium at 8°C (not shown). The 14 strains of the W cluster, phylogenetically identified as singletons not belonging to any clonal cluster, differed in their ability to grow in the cold. Nine of these strains grew reproducibly at 6°C, three did not grow at this temperature, and growth was not reproducible for the other two (Fig. 1). We therefore conclude that, in the collection tested, only the genetically closely related G3 and G5 clonal complexes contain multiple strains with a high potential for growth in the cold.

Statistical recombination tests and the population structure of the B. weihenstephanensis and B. thuringiensis forest soil community.The large number of strains from the T and W clusters made it possible to compare the population structures of these clusters. Visual comparison of the topologies of the phylogenetic trees constructed for the T and W clusters (Fig. 1A and B) indicated differences in the population structures of the strains in these two clusters, with presumably clonal population structures for cluster T and effectively sexual (or panmictic) population structures for cluster W. Several clonal complexes of cluster T (G1, G2, and G7) contained a number of strains with the same STs (Fig. 1A). In contrast, the topologies of trees for clonal complexes G3, G5, and G10 differed (Fig. 1B), since only a few strains had the same allelic profiles. The topology of clonal complex G9 in cluster C also seems to resemble those of the clonal complexes of the W cluster more closely than those of other clonal complexes of the C cluster.

A quantitative parameter for distinguishing between different bacterial population structures, the index of association (I_A), was proposed by Maynard Smith et al. (23). Differences in this parameter, calculated for the whole population and its parts, in particular those containing STs taken as units, can be used to distinguish between completely sexual (I_A = 0), epidemic, sexual at fine scale, and completely clonal (I_A ≫ 0 for all parts of the population) populations. For the forest strains used here, the index of association was higher than zero (I_A, 2.917 [n= 115 strains]) and similar to that calculated for all strains (I_A, 2.815 [n = 134]). A similar value was obtained if the strains from deep in the forest or from the forest edge were considered separately (I_A, 2.807 [n = 81] and 3.861 [n = 34], respectively), presumably indicating a high correlation of sequenced alleles and the absence of intensive recombination in the population as a whole. However, the index was much lower if only strains from the W cluster were considered (I_A, 0.593 [n= 41]) but remained high for the T and C clusters (I_A, 4.257 [n = 55] and 2.705 [n =15], respectively). For these two clusters, the index for all clonal complexes, identified by START, exceeded 2.5. In contrast, groups G3 and G5 had lower indices (I_A, 0.151 [n =20] and 0.243 [n = 7], respectively), indicating effective sexual exchange of alleles, decreasing the correlation between them. Inclusion of the 19 independently isolated strains in the analysis had no significant effect on these findings.

Recombination-to-point mutation ratio in B. weihenstephanensis and B. thuringiensis.As reported for other bacterial populations (12), we estimated the ratio between allelic variations due to point mutations and those due to recombination events. We used the approach described by Feil et al. (11). Allelic profiles from clonal complexes in clusters W and T and sequence differences between alleles were analyzed to determine whether the alleles differed from their ancestors by a single nucleotide change, indicating that they were generated by mutation or recombination, or displayed multiple changes, indicating that they were generated by recombination alone (11). The corresponding data are presented in Table 3. For cluster W, the ratio of multiple- to single-change alleles was 13 to 7 (1.9:1), and for cluster T it was 5 to 9 (0.6:1). The corresponding estimates for the probability of changes per site were 15:1 and 3.5:1, respectively, for the W and T clusters. These figures are smaller than those reported for Neisseria meningitidis (100:1), Streptococcus pneumoniae (61:1), and Staphylococcus aureus (24:1) (12), especially for the T cluster population. The higher frequency of recombinational allele evolution in B. weihenstephanensis (W cluster) is consistent with the effectively sexual nature of this community, contrasting with the clonal nature of B. thuringiensis (cluster T).