Investigating the Genome Diversity of B. cereus and Evolutionary Aspects of B. anthracis Emergence

Abstract

Here we report the use of a multi-genome DNA microarray to investigate the genome diversity of Bacillus cereus group members and elucidate the events associated with the emergence of B. anthracis the causative agent of anthrax–a lethal zoonotic disease. We initially performed directed genome sequencing of seven diverse B. cereus strains to identify novel sequences encoded in those genomes. The novel genes identified, combined with those publicly available, allowed the design of a “species” DNA microarray. Comparative genomic hybridization analyses of 41 strains indicates that substantial heterogeneity exists with respect to the genes comprising functional role categories. While the acquisition of the plasmid-encoded pathogenicity island (pXO1) and capsule genes (pXO2) represent a crucial landmark dictating the emergence of B. anthracis, the evolution of this species and its close relatives was associated with an overall a shift in the fraction of genes devoted to energy metabolism, cellular processes, transport, as well as virulence.

INTRODUCTION

The B. cereus sensu lato group is comprised of multiple species including B. cereus (Bc), B. thuringiensis (Bt), B. mycoides, B. pseudomycoides, B. weihenstephanensis (Bw) and B. anthracis (Ba). Based upon 16s rDNA sequence, these species share 99% sequence identity and therefore, phylogenetically they belong to one group [1-4]. The naming of species within this group has placed a historical emphasis on the distinct biological phenotypes displayed by members of the B. cereus group, most notably, the mammalian pathogen B. anthracis [5]. The reconciliation of contradictory relationships exhibited by members of this phylogenetic group of species is still ongoing. There is a growing number of complete and partial genomic sequences available in public databases that has confirmed and expanded our appreciation of the diversity displayed by the Bc group members [6]. Genome size ranges from 5-6 Mb can be attributed to high degree of plasmid heterogeneity but also to variation in the chromosomally encoded genes [6-13]. One aspect of this diversity may be explained by the dynamic repertoire of plasmids found in B. cereus group isolates [8-11,13-23]. The number and size of plasmids found in these isolates suggest that the plasmids are a significant reservoir of gene novelty enabling species fitness in a wide array of environmental niche. The specific plasmid complements encode a substantial number of genetic determinants that influence B. anthracis virulence (pXO1, pXO2), or B. thuringiensis insecticidal/pathogenic character (pBT) and Bc emetic strains (pCER270), for example [7,16,17,24]. There is evidence that mobility of plasmid encoded sequences contribute to the apparently high rate of species diversification [25-28]. One genome sequencing project, reported a B. cereus isolate recovered from a metal worker presenting symptoms consistent with inhalation anthrax [13]. This report altered the previously held belief that the virulence plasmids, pXO1, encoding the primary virulence factors, Lef, Pag and Cya were found solely in Ba [13]. These observations have been subsequently extended to other Bc isolates that encode toxin genes that cause invasive disease [17]. It remains unclear to what extent plasmid inheritance resulting in such fundamental phenotypic alterations occurs within this group.

The life cycle of B. anthracis begins with the infection of the host by the spore [24,29]. The spore germinates and become vegetative and metabolically active cells. Upon shedding or host death, the vegetative cells are often returned to the soil, where the vegetative cells go through the process of sporulation to form highly resistant spores. B. anthracis spends the majority of its life cycle as an inert spore. This may imply that Ba may have substantially reduced opportunity for gene acquisition by horizontal transfer compared to Bc counterparts that more commonly exist in the environment as vegetative cells. Comparative analysis of B. anthracis genomes indicates that Ba belong to a monomorphic group with limited diversity. This is in contrast to other Bc group genomes that display greater degrees of diversity. There is evidence that members of the Bc group undergo genetic exchange with other members of this group [14,18,20,23]. Despite several illuminating studies conducted in the last decade with regard to the genome composition and population structure of Bc group (for review see [6,10]), the evolution and emergence of Ba remains unclear. This limitation can be attributed to our lack of knowledge regarding the ecology of the Bc group members and especially Ba [5,30-33]. An additional barrier pertains to our relative inability to identify the whole set of virulence factors and accessory genes required for niche adaptation and pathogenicity [9,12,32,34,35]. Large scale comparative genomic studies while useful must be complemented with continued functional characterization of genes and their function in natural environments in order to better elucidate aspects of the evolution of ecotypes within this heterogeneous group [36].

Comparative genomic analysis of the Bc group involving complete DNA sequence and SNPs have been employed [8,9,11-13,37-39]. Here we have used comparative genome hybridization (CGH) as a means of defining genomic differences within this group, specifically to address the emergence of B. anthracis. While powerful, CGH analysis has discrete limitations related to the use of arbitrary reference genomes. As such, one may only identify genes that have been lost in query genomes, relative to that reference genome. Likewise, the DNA hybridization assay using DNA microarrays has broad, but limited capacity to detect genes that have undergone substantial divergence resulting in a relatively high frequency of false negative gene presence measurements.

In the current study we used two complementary approaches to estimate the genomic diversity among the members of the Bc group, and elucidate evolutionary aspects regarding one of the recently emerged lineages containing B. anthracis and its close relatives. Initially we implemented a directed sequencing strategy we refer to as Gene Discovery (GD) that, like subtractive hybridization, enables the identification and rapid characterization of strain-specific sequences present in microbial genomes. In order to gain insights into the unique genomic features encoded in the B. anthracis genome in the context of the B. cereus group members we applied the GD strategy to identify those genomic segments that are uniquely present in the genomes of seven diverse B. cereus strains, relative to the genome of B. anthracis. We report the analysis of genomic sequences that appear characteristic among the seven query genomes. The number of unique sequences obtained from each strain ranged significantly. Based on the unique genome sequences identified in this study and other publically available sources we developed a more comprehensive species 70-mer oligonucleotide DNA microarray for comparative genomic analyses. We carried out comparative genome hybridization (CGH) on 35 diverse Bc and six Ba strains. Due to its inclusiveness, the multi-genome microarray is able to closely approximate the gene complements of query genomes. In so doing, we were able to identify genomic events associated with the emergence of B. anthracis as a distinct lineage within the B. cereus group. We also demonstrate that the evolution of B. anthracis as a hyper-virulent, invasive phenotype is coupled with reduced energy metabolism, biosynthesis of cofactors, and transport functions compared to its predecessors. At the same time acquisition of the plasmids harboring the genes for the three toxins, and the capsule, through horizontal gene transfer, seem to have been accompanied by the acquisition of a specific set of the chromosomal virulence associated factors.

MATERIAL AND METHODS

Bacterial Isolates

A total of 48 strains, 41 B. cereus (Bc), one B. weihenstephanensis (Bw), and six B. anthracis (Ba) strains, were investigated (Table 1). Many of these strains have been previously characterized by various population genetic analysis methods such as whole genome sequencing, multi-locus enzyme electrophoresis (MLEE), multi-locus sequence typing (MLST) as well as CGH [4,12,39]. The isolates correspond to all three major groups within the B. cereus sensu lato [38,39] and represent a wide phenotypic and genotypic diversity, and geographic coverage of the species. Seven strains, were selected initially for gene discovery to complement the complete genome sequences already available in the public databases. These seven strains were selected to represent major lineages with this taxon on the basis of previously reported population genetic studies using CGH [12] and MLST [39].

Table 1.

Strains used in the analysis.

Species	Strain	Alternative Designations	Year of isolation	Geographic Origin	Source	Type of infection (human isolates)	MLST sequence type (ST)&	Plasmid Profile	Main phylogenetic cluster$
B. cereus	AH259*	6A1 in BGSC					Not typed		II
B. cereus	AH607*			Norway	Dairy		17		I
B. cereus	AH535*		1994	Norway	Soil (strawberry field)		77		III
B. cereus	AH819*		1995	Norway	Human	Periodontitis	40		I
B. cereus	AH812*		1995	Norway	Human	Periodontitis	38		I
B. cereus	AH1123*			France	Human	Blood	45		I
B.weihenstephanensis	AH1143*			Germany	Dairy (milk)		68		III
B. cereus	ATCC 10987*		1930	Canada	Spoiled cheese		2		I
B. cereus	ATCC 14579*		1916	USA	Air, farmhouse		33		II
B. cereus	AH404			Finland	Dairy		Not typed		III
B. cereus	AH405			Norway	Dairy		Not typed		II
B. cereus	AH407			Finland	Dairy		9		III
B. cereus	AH533		1993	Norway	Soil (strawberry field)		Not typed		IV§
B. cereus	AH540		1993	Norway	Soil		Not typed		I
B. cereus	AH557		1993	Norway	Soil		Not typed		I
B. cereus	AH564		1993	Norway	Soil		Not typed		I
B. cereus	AH568		1993	Norway	Soil		Not typed		I
B. cereus	AH597			Norway	Dairy		Not typed		III
B. cereus	AH598			Norway	Dairy		Not typed		I
B. cereus	AH599			Norway	Dairy		Not typed		I
B. cereus	AH601			Norway	Dairy		Not typed		II
B. cereus	AH602			Norway	Dairy		Not typed		III
B. cereus	AH603			Norway	Dairy		Not typed		III
B. cereus	AH604			Norway	Dairy		Not typed		I
B. cereus	AH608			Norway	Dairy		Not typed		I
B. cereus	AH648		1994	Norway	Soil		Not typed		III
B. cereus	AH810		1994	Norway	Human	Periodontitis	36		I
B. cereus	AH811		1995	Brazil	Human	Periodontitis	37		II
B. cereus	AH813		1995	Brazil	Human	Periodontitis	39		I
B. cereus	AH814		1995	Norway	Human	Periodontitis	84		II
B. cereus	AH815		1995	Norway	Human	Periodontitis	85		II
B. cereus	AH816		1995	Norway	Human	Periodontitis	39	pPER272	I
B. cereus	AH817		1995	Norway	Human	Periodontitis	3	pPER272	I
B. cereus	AH818		1995	Brazil	Human	Periodontitis	39		I
B. cereus	AH820		1995	Norway	Human	Periodontitis	39		I
B. cereus	AH823		1996	Norway	Human	Periodontitis	40		I
B. cereus	AH825		1996	Norway	Human	Periodontitis	40		I
B. cereus	AH826		1996	Norway	Human	Periodontitis	40		I
B. cereus	AH827		1996	Norway	Human	Periodontitis	40		I
B. cereus	AH828		1996	Norway	Human	Periodontitis	40		I
B. cereus	AH830		1995	Norway	Human	Periodontitis	41		I
B. cereus	AH831		1996	Norway	Human	Periodontitis	42		I
B. anthracis	Ames		1981	USA	Cow		1	pXO+ /pXO2+	I
B. anthracis	Australia94	A0039	1994	Australia	Cattle		1	pXO+ /pXO2+	I
B. anthracis	Tsiankovskii-I			Former USSR	Livestock vaccine strain		1	pXO+ /pXO2+	I
B. anthracis	Vollum	A4088	1935	UK	Cow		1	pXO+ /pXO2+	I
B. anthracis	Sterne	34F2	1937	South Africa	Cow (animal vaccine strain)		1	pXO+ /pXO2-	I
B. anthracis	Pasteur						1	pXO- /pXO2+	I

^*Strains used for gene discovery

^&According to the optimized scheme of Tourasse, Helgason et al., (2006)

^$As determined by MLST, MLEE, and/or AFLP typing data (HyperCAT database, http://mlstoslo.uio.no)

Preparation of Genomic DNA

For each strain in Table 1, a single colony from a plate of LB growth was inoculated into 10 mL LB and grown overnight at 30°C. Fresh LB medium (250 mL) was inoculated from the overnight culture at a 1:50 dilution and grown for 4 hours at 30°C with shaking (250 rpm). The culture was harvested by centrifugation at 8500 rpm for 10 min at 20°C. Genomic DNA was extracted and purified from the pellets as previously described [39]. All genomic DNA (gDNA) samples were assessed for DNA integrity by gel electrophoresis before further analysis.

Gene Discovery (GD): Genomic Library Preparation, Screening and Sequencing

Gene Discovery strategy was applied as previously described [40]. Library generation and screening for GD are summarized in Figure 1s. Briefly, a partial Tsp509I digestion was performed with 3 μg each of the seven selected Bc strains to generate a mass peak centered at ~500 base pair fragments. Digested DNA was electrophoresed on 1% ultra pure agarose gel (Invitrogen, Carlsbad, CA), and fragments of an apparent MW ~300-800bp. were excised and recovered using the QIAquick gel extraction kit as per manufacturer’s instruction (QIAgen, Valencia, CA). These fragments were cloned into the EcoRI site of pUC19 [41] and transformed into ElectroMAX DH10B cells (Invitrogen, Calsbad, CA) by electroporation. High-throughput plasmid purification was performed at the J. Craig Venter Science Foundation, Joint Technology Center (Rockville, MD). To determine how many plasmids to screen from each genomic library we used Moore’s formula [42] assuming an average insert size of base pairs. For each strain, 37,632 plasmids were produced representing greater than 5X coverage of each genome. Slides for genomic library screening (GD) were generated by printing plasmid DNA templates at a concentration of ca. 200ng/μL. In order to identify sequences uniquely present in query strains, each library was screened through flip-dye competitive hybridizations using differentially labeled probes from the genomic DNA of the strain that the library was generated from, and B. anthracis Ames strain [12]. Plasmids were printed and screened using Bacillus spp./ and Ba genomic probes labeled with either Cy3 or Cy5. Replicate hybridizations were conducted with flip-dye replicates. Genomic DNA hybridizations were performed essentially as described at http://pfgrc.jcvi.org/index.php/microarray/protocols.html [40,43]. The slides were dried and then scanned using the GenePiX 4000B scanner (Axon Instruments, Union City, CA). The signal intensity ratios were used to indicate those plasmids that contained inserts that were divergent or uniquely present in the query Bacillus spp. genome relative to Ba. Hybridization signals displaying a query Bacillus spp./reference Ba. log-ratio of 5.0 or greater were identified and the corresponding plasmid DNAs were subjected to DNA sequencing in both directions using vector-based primers. All sequence reads were assembled using the Celera Assembler [44]. Automatic annotation was performed to identify entire or partial open reading frames (ORFs) as previously described [45,46]. Annotation of the contigs produced putative ORFs or “features.” These features represent both complete and partial genes and are referred to as “pORFs” hereafter.

Gene Discovery Sequence Analysis

Figure 2s summarizes the bioinformatic approach used to characterize the sequences from B. cereus strains subjected to gene discovery. Briefly, template sequences obtained from the GD process were compared to publicly available genome sequences from Bacillus as well as non-Bacillus species to identify unique sequences. Obtained sequences were initially filtered for uniqueness against B. anthracis, and B. cereus genomes, using the blastn algorithm [47]. Nucleotide matches of <100 nucleotides and e-values >10^-5 were preliminarily considered novel. Translated sequences were then compared against the non-redundant amino acid (NRAA) database using blastp [47]. ORFs showing BLAST hits with homology e-value <10^-5 were considered homologous to previously characterized proteins, while those with no match or with e-values >10^-5, were considered unique.

Species Microarray Design, Preparation, and Hybridization

ArrayOligoSelector [48] was used to design seventy base oligonucleotides (70-mer genomic markers) for the multi-genome species microarray. A total of 29,682 ORFs representing unique features from GD and from other partial and complete genome sequencing projects (NCBI release 148) were printed on the species microarray. Oligonuceotide preparation, printing and hybridizations were performed as previously described [12,40]. Ba Ames gDNA was used as reference in all CGH experiments.

DNA Microarray Data Analysis

Microarray data were analyzed as previously described [40,49]. Raw microarray images were analyzed using SpotFinder version 2.2.2 from the TM4 Microarray Software Suite. Fluorescence intensities were normalized using Histogram Mode Centering (HMC) algorithm with spots less than 20,000 relative fluorescence units filtered from subsequent analysis.

CGH Data Clustering

In order to minimize bias and noise due to the individual log-ratio values when investigating global gene presence/absence patterns among the query strains, the final data set were assigned one of three different values: 0 = absent, 0.5 = divergent and 1 = present. Hierarchical clustering of the query genomes based on CGH data was performed as previously described [40,49-53] using the Multiple Experiment Viewer (MeV) of TM4 Software as well as MrBayes.

Allelic Grouping and Gene Calling

Our oligonucleotide design approach allowed us to distinguish subtle sequence differences among variants of orthologous and paralogous genes. In order not to confound the final gene calls or the total gene estimates, we identified and clustered related gene variants into “allelic groups” as previously described [40]. Each group includes putative variants of the same coding DNA sequence (CDS). Clustering of the orthologous sequences into allelic groups was primarily based on oligo-to-gene, gene-to-gene, and protein-to-protein relationships. Allelic groups containing multiple paralogs were sub-divided when possible such that each paralogous variant had its own sub-group.

Statistical Analysis of the CGH Data

The Fisher exact test [54] was used to perform genomic marker association analyses (MAA) between the genomic attributes i.e. gene presence or absence, and clades or groups of strains displaying a particular phenotype. Markers, and consequently their corresponding ORFs, were considered “characteristic for” or “prevalent in” a particular group as inferred by MAA using p<0.05 as a cut-off value for statistical significance [40].

RESULTS

Genomic Library Screening

We have developed a directed sequencing strategy referred to as Gene Discovery (GD) that enables the identification and rapid characterization of strain-specific sequences present in microbial genomes (Figure 1s). We applied GD analysis to nine strains belonging to the B. cereus sensu lato group. They consisted of two fully sequenced strains (ATCC10987 and ATCC14579), and seven uncharacterized strains. Random genomic libraries from each strain consisting of nearly 38,000 recombinant pUC plasmids with inserts of 300-800 bp were printed onto glass slides. Each library was screened through flip-dye hybridization using two differentially labeled probes. The reference channel in each case was B. anthracis Ames [12] paired with the strain that was used for library construction. Based on hybridization intensity, plasmids that hybridized to the query strain (library) but not B. anthracis, were considered candidate strain-specific sequences. Fully sequenced strains ATCC10987 and ATCC14579 were included to evaluate the efficiency of the novel gene identification. None of the sequence reads obtained from either library – ATCC10987 and ATCC14579, aligned significantly with any identified Ba sequences. The fraction of strain-specific genes encoded in the ATCC10987 and ATCC14579 recovered in the screen was 100% for both control strains (e-value <10^-5 over >100 nucleotides, Table 1s, [9,11]). These findings suggest that both the experimental and bioinformatic thresholds were appropriate. Over 75% of the recovered sequences displayed significant similarity at the nucleotide and protein levels to sequences derived from previously sequenced Bacillus spp. or other gram-positive bacteria. The number of unique and orthologous reads varied for each query genome screened (7%-35%). Strains AH1123 and AH1143 had the largest numbers of unique reads identified (614 and 913 respectively). Strains AH812 and AH259 had the smallest number of novel sequences identified (46 and 48 respectively).

Analysis of Unique Genomic Features

A total of 4,630 contigs and 4,008 singletons were obtained from the seven query strains. Contig size ranged from 270 – 4,800 bp with an average length and sequence coverage of 740 bp and 1.9 reads per contig respectively. The amount of novel sequence information in the query B. cereus strains (the total length of contigs and singletons) relative to B. anthracis ranged from 104.4 Kb (AH812) to 1.5 Mbp (AH535) (Table 2). Genes encoding hypothetical proteins constituted the largest group of features (35-76%, Figures 2 and 3s). Among pORFs with a match in the NRAA database, the vast majority (87%) shared sequence identity to Gram-positive bacteria and bacteriophage. While 71 % of the novel sequence displayed the strongest similarities to other Bacillus spp., pORFs with identity to other non-Bacillus spp., low G+C Gram-positives such as Clostridium spp. and Listeria spp. were also prevalent (Figure 1).

Table 2.

Comparative sequence analysis summary of B. cereus novel sequences found from gene discovery.

B. cereus strains:	AH819		AH607		AH535		AH1123		AH812		AH1143		AH259
Total bases sequenced / strain:	1,076,892	bp	1,237,775	bp	1,500,190	bp	443,218	bp	104,276	bp	1,422,879	bp	177,983	bp
Fraction relative to the size of an average B. cereus genome:	20%		23%		28%		8%		2%		27%		3%
Total number of annotated features:	2,041		2,204		3,138		871		200		2,874		383
Species matches
B. anthracis	43	2% a	122	6% a	132	4% a	52	6% a	7	4% a	92	3% a	11	3% a
B. anthracis pXO1 & pXO2	4	0% a	0	0% a	0	0% a	0	0% a	1	1%	2	0% a	0	0% a
B. cereus 10987	503	25% a	510	23% a	431	14% a	85	10% a	34	17%	373	13% a	4	1% a
pBc10987	121	6% a	3	0% a	1	0% a	1	0% a	17	9%	0	0% a	0	0% a
B. cereus ATCC14579	143	7% a	183	8% a	427	14% a	64	7% a	15	8%	394	14% a	271	71% a
B. cereus (other strain genomes)	200	10% a	275	12% a	284	9% a	192	22% a	21	11%	227	8% a	12	3% a
B. cereus (other strain plasmids)	49	2% a	70	3% a	112	4% a	18	2% a	12	6%	73	3% a	0	0% a
B. thuringiensis	97	5% a	245	11% a	490	16% a	52	6% a	7	4%	474	16% a	40	10% a
B. thuringiensis plasmids	0	0% a	6	0% a	3	0% a	0	0% a	0	0%	0	0% a	0	0% a
Bacillus spp.	1	0.05% a	1	0.05% a	5	0.16% a	0	0.00% a	0	0.00%	6	0.21% a	0	0.00% a
Bacillus spp. plasmids	0	0% a	2	0% a	0	0% a	0	0% a	0	0%	1	0% a	0	0% a
Bacillus spp. phages	0	0% a	0	0% a	0	0%	0	0% a	0	0%	0	0% a	0	0% a
Overall Number of features similar to Bacillus species by BlastN:	1,161	57% a	1,417	64% a	1,885	60% a	464	53% a	114	57% a	1,642	57% a	338	88% a
Remaining relatively novel features (by BlastN):	880	43% a	787	36% a	1,253	40% a	407	47% a	86	43% a	1,232	43% a	45	12% a
Truly unique features from the relatively novel gene set c:	452	22% a	312	14% a	887	28% a	324	37% a	71	36% a	1,054	37% a	19	5% a

^aFraction from the total number of annotated features in the corresponding strain.

^bRemaining pORFs whose amino acid sequence shows no significant homology to proteins in databases by BlastP.

Figure 2.

Comparative summary of genes considered missing or novel among the seven query B. cereus strains relative to B. anthracis. Absent genes are identified from a previous CGH study [12] whereas the novel genes are those discovered through gene discovery (GD).

Figure 1.

Sequence analysis summary of the novel genomic content present in B. cereus strains based on blast [47]. Species hit summary results are based on feature (partial ORF) sequence analysis resulting from blastp.

We observed differences among the unique features obtained from each query genome with respect to predicted functions (Figure 3s, Table 2s). The majority of the novel features (67%) encoding proteins with predicted functions defined only seven major role categories including: cell envelope (14%), transport (14%), regulatory/signal transduction (13%), cellular processes (10%), energy metabolism (9%) and mobile elements (7%). Among transporters, proteins predicted to be responsible for the uptake of glycopeptides, amino acids and their derivatives, trace elements, calcium, sodium, phosphates as well as sulfates were commonly found. In addition to some S-layer proteins, the novel cell envelope fraction was dominated by proteins involved in various steps of peptidoglycan or capsule synthesis; most of them were sugar transferases, epimerases, amidases, transfereases, D-alanine-D-alanine ligases N-acetylglucosaminyltransferases, or murein hydrolases. Novel features related to regulatory functions included various members of two-component regulatory systems, transcriptional regulators belonging to MarR, TerRAcrC, LysR, LuxR, LytR, AraC/XylS families, σ⁵⁴-dependent transcriptional activator, BglG family anti-terminators, as well as kinases involved in sprorulation. Among the unique pORFs, we identified several that are predicted to perform important house-keeping functions within the cell such as protein synthesis and fate, and DNA metabolism. This group of features was enriched for variants of tRNA-synthetases. Among non-essential genes there appeared to be significant variation in genes encoding chaperones, proteases, and restriction modification enzymes (Figures 3 and 3s, and Table 2s).

Figure 3.

Simulation model for predicting the effect of additional strain sampling to estimate (A) the number of novel pORFs, and (B) the limits of unique genomic DNA expected among B. cereus group members.

Distribution of Virulence-associated and Anti-microbial Resistance Genes

We mined the novel sequence data to identify genes with similarity to well characterized penicillin binding proteins (pbp- or fmt-like) and other antibiotics, heavy metals, multi-drug efflux pump regulators and lantibiotic transporters (Table 2s and 3s). Virulence associated genes (VAGs) included orthologs of Listeria spp. internalin, enterotoxins, putative cereolysin O (a thiol-activated cytolysin variant from B. weihenstephanensis), collagen adhesion protein (B. cereus ATCC10987), mouse virulence factor mviN, and a perfringolysin O precursor; almost all occurrences of these genes were found in non-clinical (environmental) isolates. Interestingly, we found a gene derived from the strain Bw AH1143 annotated as “protective antigen.” This CDS (2,127 bp) displayed significant sequence identity (61 % nt, 47 % aa) to the B. anthracis protective antigen. Interestingly, the Bw AH1143 pag-ortholog had 97 % nucleotide identity over its length to a 1846 bp contig, encoding a partial CDS also annotated as protective antigen, in the recently sequenced genome of B. cereus BDRD-ST196 (NCBI Accession Number: ACMD01000248).

B. cereus Pan-genome Predictions

To evaluate the comprehensiveness of gene discovery as it relates to the complete gene pool encoded by B. cereus sensu lato, we performed statistical modeling based on the novel genomic content observed among the query genomes (Figure 3). The number of novel genes obtained from each query strain follow a strong regression pattern approximating the Boltzmann equation (R²=0.96). This analysis predicts that members of the Bc group may vary in genome size by as much as 1.5 Mb, nearly 30% of an average genomes coding capacity. It is interesting to note that this value is quite close to traditional threshold for species differentiation (≥ 30 % genome difference by DNA-DNA hybridization) [55,56]. Following the identification of novel genes from seven Bc strains, we estimated that an asymptote was reached predicting that the number of the new genes that may be discovered through additional screening of genomes may not exceed 250. The fact that the number does not drop to zero is consistent with the existing knowledge that B. cereus sensu lato is part of an open genome [45].

B. cereus/B. anthracis Multi-Genome DNA Microarray Design

Annotated genes derived from novel gene sequences discovered in this study and previous genome and plasmid sequencing projects were used to create a database of B. cereus group DNA sequence to enable 70-mer oligonucleotide design using a modified version of the Pick-70 tool [48]. Table 4s presents a summary of the 29,977 unique 70-mer oligonucleotides that together, represent 29,682 putative ORFs. We were able to assign these ORFs to a smaller number of 15,548 orthologous groups including sequence variants of orthologous coding DNA sequences (CDSs). It should be noted that many singleton reads annotated as unknowns, may represent unlinked sequence belonging to the same ORF or orthologous sequence. Therefore, the frequency and identity of feature redundancy for these ORFs is unknown. Given these uncertainties, the estimated number of oligonucleotides per allelic group is 1.93. There were 9,909 (59%) CDS represented by single oligonucleotides. Among the orthologous groups with two or more CDSs, 3,117 (48%), 1,823 (28%), 580 (9%), 344 (5%), and 225 (3%) are represented by two, three, four, five and six oligonucleotides respectively. Allelic grouping was then used as the basis for estimating the genomic content of each query strain based on hybridization results.

Genome Size and Conservation Estimates

We interrogated the B. cereus/B. anthracis species DNA microarray with a set of 35 diverse Bacillus isolates including soil, dairy and periodontal isolates and six B. anthracis strains (Table 1). CGH profiles of the strains are summarized in Table 5s. In order to assess the accuracy of using CGH data to estimate gene content and genome size, CGH-derived total gene estimates for four B. anthracis strains (Ames, Sterne 34F2, Vollum, Australia 94, Tsiankovskii-I) and three B. cereus 14579, 10987 and AH820) were compared to the annotated genes from their published genome sequences. The margin of error of the CGH-based method ranged from -7.9 to 13% (median 2%, average 4%). Using this procedure to estimate genome size, we observed that genome sizes are generally conserved within the members of this group (Figure 4s). For the majority of query strains, total gene counts ranged from 5,000 to 6,000. These results are comparable to genome size estimates determined by complete genome sequencing (average genome size among B. cereus group members varies between 5.3-5.6 Mb). Some notable exceptions to the overall genome size conservation included three strains (AH533, AH815 and AH830) that had gene counts below 4,000 and five others (AH404, AH597, AH601 AH604 and AH648) that had total gene estimates exceeding 6,000.

CGH Profiling of Mobile Elements

Plasmid profiles inferred from CGH patterns appeared complex (Table 6s, Figure 4). Therefore, we focused our analysis on the best-known and most clinically relevant plasmids such as pXO1, pXO2, and pBC218 [13,27]. Plasmid pXO1 has been shown to share a high degree of sequence similarity, with three completely sequenced plasmids i.e. pBC10987, pCER, and pPER272 [16]. Here we refer to the presence of a plasmid or a plasmid backbone if CGH results indicate that ≥ 20% of genes normally associated with a plasmid are present. In no case have we confirmed that the detected genes are indeed plasmid-based. Among the 35 Bc strains investigated, 19 (54 %) appeared to harbor pXO1-like plasmids. These findings confirm previous observations that pXO1-like plasmid variants are common among B. cereus group members [6,8,11-14,16].

Figure 4.

Summary of comparative clade analysis based on the marker association analysis. The dendrogram on the left represents strain clustering based on the global CGH patterns. Comparative clade analyses are color-coded with square brackets accordingly. Markers representing clade-characteristic CDSs are summarized in the form of predicted functional role categories of the genes they represent. Stars next to the role category sections on each bar indicate that there is a statistical difference with respect to their corresponding fractions between groups.

CGH profile analysis indicated that variants of capsule carrying plasmids such as pXO2 or pBC218, may also be harbored by some Bc strains but at a significantly reduced frequency. pXO2 variants were found in four Bc strains (AH813, AH815, AH816, and AH818). With one exception (AH813), all these strains also appeared to harbor pXO1-like plasmids. Traces of pBC218 plasmid (10-19% of its genes) were observed in eight Bc strains, four of which (AH817, AH598, AH599, AH601), also seemed to harbor pXO1-like plasmid variants. Despite the prevalence of the virulence plasmid variants among Bc strains investigated, the presence of B. anthracis toxin genes was very rare. Among the B. cereus strains examined, we observed a positive signal for the regulator of tripartite toxins gene expression, atxA (pXO1) only in strain AH568. Five other strains, i.e. AH601, AH597, AH648, AH404 and AH818, generated a positive signals for pXO2-borne capA gene. Similarly, strain AH608 generated a positive signal for capC. Only two additional B. cereus strains, AH815, AH811, generated positive signals for both capA and capC, however both were negative for capB as well as the regulatory genes controlling cap gene expression (Table 7s).

Phylogenomic Relationships of Bacillus Strains

We performed hierarchical clustering to reveal strain relatedness on a genomic scale. The dendrogram shown in Figure 4 represents a summary of phylogenomic relationships based on the data for all 29,977 oligonucleotides using the trinary designations (“absent”, “divergent” and “present”). Strain relationships based on CGH patterns (trinary values) of plasmid marker sets are summarized in Figure 5s. Overall, phylogenomic clustering resulted in similar strain relationships as those generated by other approaches such as MLEE, MLST, AFLP or sequencing [4,6,8,38,39,57,58]. The finding that four periodontal isolates, strains AH820, AH813, AH816 and AH817, clustered tightly with the B. anthracis clade (Clade 6) was of interest. Nearly all of these isolates harbor pXO1- and/or pXO2-like plasmids. These isolates were distinct from six other B. cereus strains of clinical origin i.e. AH823, AH825, AH826, AH827, AH828 and AH831 (Clade 2) that constitute a separate clade, somewhat distant from B. anthracis.

Phylogenomic cluster analysis further support previous observation that the B. anthracis clade appears to be closely related to a distinct group of B. cereus (Clade 3) [4,39]. Based on the data generated, seven Bc isolates, together with the B. anthracis strains, define a major clade (Clade 5). It is also of interest to note that the three strains in this clade are of environmental origin (Clade 5). The pathogenic potential of these isolates is unknown. Strains constituting Clade 2 were the closest relatives of Clade 4. Clades 2 and 4 appear to have originated from a common B. cereus ancestral lineage that correspond to the previously defined B. cereus group I [38,39]. Most of the strains within this group appear to harbor pXO1-like plasmid variants. The remaining major Clade 1 was a diverse group comprised almost exclusively of environmental or dairy isolates. Four of the seven members of this clade appeared to harbor pBC218-like plasmids.

Comparative Clade Analysis Reveals Genomic Flux and Alteration Events

Based on the identified phylogenomic relationships, we next conducted a comparative analysis of the gene complements of isolates within each distinct clade. The focus of our analyses was the identification of genomic patterns (gene content) associated with the emergence of B. anthracis using the Fisher’s Exact Test (Figure 4, Table 5s). We investigated the main role category profiles that provided an overview of the differences. We then conducted a more detailed examination of sub-role categories and KEGG pathways. We considered that the function(s) or pathways may be impaired if the number of missing genes classified under such functions or pathways was ≥ 2 [40].

We noticed several differences with respect to the main functional categories across the clades. For example, the fraction of genes involved in energy utilization, cellular processes, transport, and in some house-keeping functions such as amino acid biosynthesis, protein synthesis and protein fate, was relatively smaller among B. anthracis genomes and their neighbors compared to their distant relatives (e.g. Clade 8 vs. Clade 6 and/or Clade 4 vs. Clade 2), (Figure 4). This observation was also true when comparing clinical isolates with those of limited or no virulence potential. Furthermore, the fraction of transport, energy metabolism, biosynthesis of cofactors cellular processes, amino acids metabolism, and protein synthesis and fate functions as a whole in Clade 4 members is half of the corresponding group compared to the remainder of B. cereus (18% vs. 35%, Figure 4). The evolution of Clade 4, and especially B. anthracis, have been associated with the acquisition of mobile elements and most dominantly, genes of unknown function. The expansion of mobile elements in B. anthracis compared to other clades suggests that the primary mechanism for the numerous gene acquisition events occurring throughout the B. cereus group are mediated by mobile element movement.

We found a significantly smaller number of genes representing ORFs predicted to be involved in protein production i.e. amino acid biosynthesis, protein fate (post-translational modification) and synthesis among Clade 3 members compared to those belonging to Clade 1 (29 vs. 316). Likewise, fewer ORFs predicted to be involved in protein production were observed when comparing Clade 4 vs. Clade 2. (19 vs. 90) or Clade 4 vs. the remainder of all other Bc genomes (29 vs. 158). We also note a smaller number of genes devoted to the metabolism of n-acetylglucosamine, n-acetylgalactosamine or n-acetylmuramic acid derivatives (classified under cell envelope functions) among Clade 3 genomes compared to those belonging to Clade 1 (10 vs. 40) as well as among Clade 4 vs. the remainder of all other Bc genomes (11 vs. 17). Similar patterns were also observed when investigating genes involved in sugar metabolism, which is classified under energy metabolism. For example, genes involved in the uptake and metabolism of sugars in general, and mannose, fructose, ribose/ribulose or glucose in particular, were more prevalent among Clade 1 members compared to those in Clade 3 (48 vs. 11). Similarly, more genes involved in sugar metabolism were found when comparing Clade 2 members with to those belonging to Clade 4 (31 vs. 8). Clade 4 genomes also appeared to possess fewer genes involved in sugar metabolism when compared the remainder of all other Bc genomes (3 vs. 12).

Results of sub-role functions or KEGG pathways survey indicated that seven functions or pathways may be incomplete or impaired in Clade 3 members compared to those of Clade 1 affecting the metabolism of: arginine and proline, aspartate derivatives involved in amino acid biosynthesis, metabolism of butanoate, starch and sucrose, biosynthesis of secondary metabolites as well as metabolism in diverse environments. Likewise, when compared to the rest of B. cereus, Clade 4 members appeared to have incomplete or impaired functions or pathways affecting the biosynthesis of glutamate-, aspartate- and pyruvate derivatives involved in amino acid biosynthesis, degradation of proteins, peptides, and glycopeptides, degradation of amino acids and amines, biosynthesis of thiamine, adaptation to atypical conditions as well as DNA recombination and repair. On the other hand, Clade 4 members appeared to possess significantly large numbers of features representing mobile elements and enzymes of unknown functions. In addition, we discovered that Clade 4 members seemed to have four genes belonging to a segment of the inositol phosphate metabolism which enable the catabolism of myo-inositol to Acetyl –CoA or Glyceraldehyde-3P, which in turn may enter the glycolysis/glyconeogenesis pathways.

Transporters were another functional group of interest. We found that Clade 1, which contains more environmental isolates, has more transporters than Clade 3, which moslty contains strains of clinical origin. Similarly, both Clade 4 and Clade 8 members, compared to the rest of B. cereus genomes, seem to possess fewer transporters in general, especially sugar transporters. On the other hand, we found more urea/amide transporters among Clade 3 members compared to those of Clade1, and more iron transporters in B. anthracis (Clade 8) than in B. cereus genomes.

The evolution of B. anthracis (comparing Clade 3 to Clade 4) appear to have involved an overall increase in genome size and coding capacity. For example the average gene number estimates for Clade 1, and Clades 3 and Clade 4 (excluding B.a.) are 5,600, 5,200 and 5,000 respectively. By contrast, the average number of genes encoded by human-associated Clade 4 members is 460 genes larger than those of environmental origin. B. anthracis genomes also contain on average 480 more genes than their four nearest neighbors. A large fraction of the additional genes are the result of B. anthracis-specific pXO1 and pXO2 CDS and the four Ba chromosomal phage insertions. In addition, we conducted a more detailed comparative gene content analysis among Ba (Clade 8, excluding plasmid deficient mutant strains) and Clade 6 members. The results of this analysis have been summarized in Figure 5. Overall, we found 1,459 CDSs that were shared by at least two strains in Clade 7 beyond the conserved gene set. Among this group of 1,459 partially conserved genes, 475 were found in common between B. anthracis and at least another member of Clade 6. From this analysis, we also identified 645 ORFs unique to B. anthracis with respect to Clade 6 members. Besides of consisting of a significantly large number of genes, the Ba unique gene set also displayed a characteristic functional role profile. Mobile elements (34%) and genes encoding for proteins of unknown functions (45%) constituted 79% the B. anthracis unique gene set with respect to Clade 6 genomes. Interestingly, the remainder of the unique genes consisted of almost equal shares of ORFs coding for proteins involved in cell envelope (5%), cellular processes (5%), transport (4%), and regulation of gene expression (4%). It is worth noting here that of the 27 CDSs constituting the latter group of genes involved in regulatory functions only six were found within phage features. This finding suggests that, aside from several phage- and plasmid-based genes that could be just hitch-hikers within phages or plasmids, Ba seems to have acquired at least 21 single genes or operons that may contribute to its distinct pathobiology. While two of them – atxA and capR are well-characterized, the role of the other putative regulatory genes remains yet to be discovered. On the other hand, 45% (13/29) of the genes predicted to participate in cell envelope functions consisted of those encoding proteins involved in LPS biosynthesis, lipoproteins, and S-layer proteins. Finally over half (54%, 13/24) of the genes involved in transport functions consisted of efflux pumps and protein involved in transport of amino acids and trace elements.

Figure 5.

Comparison of gene content among B. anthracis (Clade 8) and four B. cereus genomes constituting Clade 6 (Figure 4, Table 5s). A. Unique and shared genes. We defined the conserved gene set shared by Clade 7 members as the group of ORFs that are found among ≥75% of both Clade 6 and Clade 8. The remaining genes from each genome were either unique or shared by at least one other genome in Clade 7. We used B. anthracis Ames as a representative of Clade 8. The number in the inner circle represent the gene set commonly found among members of both Clade 6 and Clade 8 (present in >= 75% of the genomes of Clade 7). Going outward, numbers in the second circle represent genes of a particular strain that are shared with at least one other member of the Clade 7. The remainder of the numbers represents the number of unique genes for each individual strain. B. Role category profiling of genes that are unique for B. anthracis Ames and Clade 6 B. cereus genomes.

Acquisition of Virulence Associated Genes in B. anthracis

We conducted a specific mining of the dataset to identify the characteristic genomic events associated with the evolution and the emergence of B. anthracis. Initially we identified a group of 1,512 genes that were characteristic for B. anthracis compared to the all B. cereus group genomes analyzed, (which appeared to have 302 genes in common). It should be noted that not all the genes that were characteristic for B. anthracis by marker association analysis were unique. Some of them, despite being present in all B. anthracis genomes, were occasionally present in other B. cereus genomes [6]. The picture that emerges from this analysis supports the idea that the evolution of the B. anthracis lineage from its predecessors is complex involving more than the acquisition of the two virulence plasmids. Therefore, we expanded our analysis to additional genes that may have contributed to the enhanced B. anthracis virulence, beyond those encoding tripartite toxins and the capsule.

Among the oligonucleotides represented on the DNA microarray, are 772 markers that were annotated to represent putative virulence associated genes (VAGs). This VAG set may be divided into two sub-groups. The first one is comprised primarily of toxins, capsular genes, cytadhesins, invasins, hemolysins, collagenases, and phospholipases. The second sub-group includes genes whose products are predicted to contain, or are in the involved acquisition and utilization of trace elements such as iron, cobalt, zinc, and copper. With the exception of zinc, which is a component of metallo-proteases, other elements function as co-factors primarily involved in redox pathways. While the proteins of the first group are considered primary virulence factors, those belonging to the second group may be considered as accessory genes that play a role in adaptation to the host environment, especially under stress conditions imposed by the human immune response [59-71].

Marker association analysis identified a small group of 31 VAGs (Ba-VAGs) that are characteristic for B. anthracis compared to the remainder of B. cereus genomes analyzed (Table 3). With the exception of toxin and capsule encoding genes, many Ba-VAGs are shared among other Clade 4 members.

Table 3.

Characteristic virulence associated genes for the clades of B. anthracis and its near neighbors.

Locus	Common Name	Characteristic for
		Clade 4 genomes vs. other B. cereus	B. anthracis vs. B. cereus
BA0552	internalin, putative	(+)a	+
BA1489	superoxide dismutase	+
BA1760	cobalamin synthesis protein, putative	+	+
BA1902	multicopper oxidase family protein	+
BA1925	cobalamin synthesis protein/P47K family protein	+	+
BA1981 o	siderophore biosynthesis protein, putative	+	+
BA1982 o	siderophore biosynthesis protein, putative	+	+
BA1983 o	AMP-binding protein	+	+
BA1984 o	hypothetical protein	+	+
BA1985 o	hypothetical protein	+	+
BA1986 o	conserved hypothetical protein	+	+
BA1992	phospholipase, putative	+
BA2222	alcohol dehydrogenase, iron-containing	+	+
BA2372 o	nonribosomal peptide synthetase DhbF	+	+
BA2588	alcohol dehydrogenase, zinc-containing	+	+
BA2642	cobalt transport protein	+	+
BA2730	neutral protease	+
BA3100	copper homeostasis protein CutC, putative	+
BA3131	alcohol dehydrogenase, zinc-containing		+
BA3189	manganese ABC transporter, substrate-binding protein/adhesin	(+)a	+
BA3269	iron-sulfur cluster-binding protein	+	+
BA3299	microbial collagenase, putative		+
BA3307 o	L-serine dehydratase, iron-sulfur-dependent, alpha subunit	+	+
BA3308 o	L-serine dehydratase, iron-sulfur-dependent, beta subunit	+	+
BA3435 o	alcohol dehydrogenase, zinc-containing		+
BA3442	neutral protease	+	+
BA3515	alcohol dehydrogenase, zinc-containing, authentic point mutation	+
BA3584	microbial collagenase, putative	+
BA3703	phospholipase/carboxylesterase family protein	+
BA3922	zinc protease, insulinase family	+
BA4766 o	iron compound ABC transporter, iron compound-binding protein	+	+
BA4767 o	iron compound ABC transporter, permease protein	+	+
BA4784 o	iron compound ABC transporter, ATP-binding protein	+
BA5701	channel protein, hemolysin III family	+
BXA0142	calmodulin-sensitive adenylate cyclase, Cya		+ b
BXA0146	transcriptional activator AtxA, (pXO1-119)		+ b
BXA0163	protective antigen-related protein, (pXO1-111)		+ b
BXA0164	protective antigen, PagA		+ b
BXA0172	lethal factor, Lef		+ b
BXA0197	zinc-binding lipoprotein AdcA domain protein, (pXO1-130)		+ b
BXB0060	capsule synthesis trans-acting positive regulator, putative, (pXO2-53)		+
BXB0062 o	hypothetical protein, CapE (pXO2-54)		+ b
BXB0063 o	gamma-glutamyltranspeptidase, capD (pXO2-55)		+ b
BXB0064 o	capsule biosynthesis protein CapA, (pXO2-56)		+
BXB0065 o	capsule biosynthesis protein CapC, (pXO2-57)		+ b
BXB0066 o	capsule biosynthesis protein CapB, (pXO2-58)		+ b
BXB0084	capsule synthesis trans-acting positive regulator, CapR (pXO2-64)		+ b

^aPresent in all B. anthracis strains and three B. cereus of Cl.4, Clade2, and Clade1.

^bPresent in all B. anthracis strain and three B. cereus one of which belong to Clade 4.

^oGene is part of an operon.

In our analysis of VAGs in B. cereus lineages we noted that clades comprised of human associated strains do not necessarily encode more VAGs than strains of environmental origin. For example, we identified 29 VAGs that were characteristic for Clade 3 compared to 103 that were characteristic for Clade 1 strains. Likewise, Clade 4 genomes appear to have fewer characteristic VAGs when compared to the other Bc strains — 37 vs. 47. We also conducted a global clade distribution analysis with respect to VAGs and found that that B. anthracis and its near neighbors have indeed fewer VAGs than their distant relatives (Figure 6). Taken together, these findings may indicate that the evolution of B. anthracis was associated with the acquisition and/or maintenance of a limited but specific set of chromosomally encoded VAGs in addition to those encoded on the virulence plasmids.

Figure 6.

Summary of VAG content distribution analysis among various groups of genomes.

DISCUSSION

The generation of diversity is of fundamental importance for most microbial populations. These diversification processes, together with environmental selection, drive cell adaptation [72-74]. In this study, we attempted to elucidate evolutionary aspects or trends associated with the emergence of one of the hyper-virulent variants of this taxon, B. anthracis. Genes identified through targeted genome sequencing of diverse isolates were complemented with sequences from publicly available sources to design a comprehensive oligonucleotide-based species microarray. This microarray was then used to screen a diverse group of strains for their genomic content by CGH.

Total gene estimates for the majority of the query strains varied from 5,000 to 6,000, which is consistent with previous findings regarding genome size of spore forming members of the B. cereus group (NCBI (http://www.ncbi.nlm.nih.gov ) and [4-6,10,11,13,39]). Some exceptions were noted as some strains displayed markedly reduced coding capacity and total gene estimates as low as 3,400 – 4,500, or as high as 6,500. These findings are consistent with a recent report on a highly cytotoxic B. cereus strain with genome size of 4.2 Mbp [75].

The profiling of putative role categories associated with both novel sequence identification and CGH data provided a global perspective with regard to the diversity of genome complements defining Bc group members. Based on all data generated, we conclude that 67% of the strain-variable CDS identified in the seven Bc strains analyzed correspond to seven role categories including genes of unknown function, mobile elements, transport, regulatory functions, cellular processes, energy metabolism, and cell envelope. Interestingly, the same role categories constitute the majority of functions that were absent in these seven query strains as compared to the reference genome. Hence, we infer that the heterogeneity of gene complements throughout this lineage has been driven by significant horizontal gene acquisition, and orthologous and non-orthologous gene displacements events.

Plasmids play an important role in the biology of Bacillus species. Together with other mobile elements such as bacteriophages, they enable horizontal gene transfer among members of this genus, a major force driving genetic diversity. Plasmids vary considerably in size and gene content among Bacillus spp. and contribute strongly to genome variability. pXO1-like plasmids are common among Clade 3 members, and together with pXO2 variants, pXO1-like plasmids are prevalent among isolates of human origin. In contrast, pBC218 variants are frequently found among non-clinical strains, or strains outside of the Clade 3 lineage. Our data do not allow us to define gene content as plasmid based or chromosomal. However, the presence of previously identified virulence factors and capsular genes among the gene complements of B. cereus isolates is of interest and potential importance. Two of the four Clade 6 genomes appear to harbor parts of both pXO1 and pXO2. The relevance these plasmid sequences in the pathobiology of the periodontal-associated strains constituting Clade 6 is not yet clear. Clade 6 strains in this study, together with isolates obtained from metal workers in Texas, US, [17] and wild great apes from Coté d’Ivoire, Cameroon, [15] causing anthrax-like disease are the only B. cereus strains to date that appear to harbor both pXO1- and pXO2-like plasmids. These periodontal strains, however, are missing both the B. anthracis pathogenicity island and the cap gene locus, further emphasizing the major role of these two loci in the emergence and pathogenicity of B. anthracis. Taken together, these observations suggest that the evolution of both B. anthracis plasmids appears to have been a complex process involving multiple gene acquisition events.

The phylogenomic relationships inferred by complete gene complement-based clustering do not in general, reproduce the phylogeny of a species in terms of vertical evolution, but rather depict the overall relatedness of genomes to one another. We compared the B. cereus var. anthracis CI genome [76] content and its relatedness with other genomes we investigated in this study by simulating CGH as previously reported [40]. From both marker association as well as phylogenomic analyses B. cereus var. anthracis CI appears to represent a distinct lineage within Clade 3 that is closely related to Clade 4. This finding further suggests that acquisition of both pXO1 and pXO2—as well as the full expression of the virulence genes they carry—by B. cereus var. anthracis CI may have taken place in a particular genomic (chromosomal) background.

A systematic comparative clade analysis indicates that gene acquisition and fixation appear to have driven early Clade 3 members and subsequently those constituting the Clade 4 lineage toward an altered interaction with the (mammalian) host, which eventually resulted in variants with a distinct pathobiology. The observed lineage-specific gene gain and loss events may reflect step-wise adaptation to a new selective host environment or may reflect fitness adaptation in response to the emergence of a lineage with a hyper-virulent phenotype. Although Clade 5 and Clade 6 represent near-neighbors of B. anthracis, neither the gene acquisition and gene loss events, nor the specific relevance of all these events to anthrax disease can be deduced with absolute clarity. However, it does appear that the emergence of B. anthracis is complex, involving many events, not limited simply to the acquisition of the pathogenicity island and cap genes on pXO1 and pXO2. The evolution of B. anthracis or its predecessors appears to be characterized by reductions in the number of genes involved in metabolism of arginine and proline, biosynthesis of glutamate-, aspartate- and pyruvate derivatives involved in amino acid biosynthesis, degradation of proteins, peptides, and glycopeptides, degradation of amino acids and amines, biosynthesis of thiamine, metabolism of butanoate, starch and sucrose, sugar transport, biosynthesis of secondary metabolites, metabolism in diverse environments, adaptation to atypical conditions as well as DNA recombination and repair. In addition, all these alterations appear to be associated with the acquisition of mobile elements and CDSs of unknown functions.

The finding that Clade 4 has a significantly reduced number of features involved in DNA recombination and repair is interesting and may provide support to the hypothesis of Didelot et al [77] that suggests that the pattern of gene acquisition may be indicative of a shift in recombination boundaries. A similar pattern has been previously observed between Campylobacter jejuni and C. coli resulting in changes in their ecology [78]. According to this hypothesis, different from other bacteria, the mismatch repair system in Bacillus may play a moderate role in the prevention of recombination resulting in higher promiscuity of gene acquisition. The CGH data we present in this report are congruent with those from the MLST study by Didelot et al. [77] indicating that there is a major gene acquisition shift that appears to have taken place within Clade 4 that strongly correlates with the emergence of a lineage—B. anthracis—that is highly invasive for mammals.

Comparative genome analyses enabled the identification of several mechanisms employed by pathogens to survive within the host. Observed differences between Clade 1 and Clade 3 with respect to urea/amide transporters are interesting, as some pathogens, especially those associated with enteric infections, are known to use urea metabolism as a means to neutralize acidic conditions [59,79,80]. The finding that B. anthracis possesses more transporters involved in iron acquisition is also significant, further underscoring the importance of iron in the virulence process [60,66]. Another mechanism pertained to myo-inositol degradation. Myo-Inositol is highly abundant within the host serving as a structural component of the eukaryotic cell membrane, which also acts as a signaling molecule. It appears that Clade 4 members including B. anthracis, employ scavenging mechanism for energy resources. This pattern is congruent with the sialic acid degradation pathway employed by pathogenic vibrios [81].

The acquisition of specific genes, e.g. VAGs, in strains with elevated pathogenicity, appears to have occurred in concert with an overall genome size reduction in B. anthracis, indicating niche speciation [82]. Gene acquisition events may lead to a shift in host environment or tissue tropism. This shift may have a fundamental impact on the genome since previously useful genes may no longer serve as such. In such cases, the rate of fixed mutations increases dramatically and involves pseudogene formation, IS element activity and movement, genome rearrangements and gene loss [74,83]. Adaptation to the host environment may be driving the loss of genes that are no longer essential in the host environment. The gene loss profile observed among the seven genomes comprising Clade 4, and especially B. anthracis, suggests that genome evolution is driving the microbial cell toward deeper dependency on the host for energy and metabolites. At the same time, this process has been accompanied by the acquisition of VAGs by early Clade 3 and especially Clade 4 members that may represent the initial steps towards niche speciation i.e. in a mammalian host [29,84]. Then, the acquisition of the pathogenicity islands coding for the three toxins and cap gene locus may have further enhanced the invasive capabilities of a clone, resulting in a genomic variant that is able to consistently cause anthrax. Therefore, acquisition of these two loci must have been a critical point in the evolution of the B. anthracis. The combined evidence from comparative clade analysis and plasmid profile suggests that members of this lineage have been undergoing rapid evolution towards a higher level of niche adaptation.

Analysis of strains by the marker set representing presumed virulence factors suggests that the evolution of highly virulent B. anthracis from its predecessors is associated with the fixation of a specific set of VAGs. In addition to being functionally characterized as virulence determinants, some of these proteins have been shown to have immunogenic properties and are possible therapeutic targets [85-87]. The identification of genes encoding proteins containing or involved in the acquisition of trace elements, besides those involved in cytadherence or invasion, is significant. Enzymes that bind trace elements are mainly involved in redox processes which are essential for microbial growth and survival, especially under stress conditions e.g. inflammation or phagocytosis [70,88-93]. Therefore it is no surprise that many pathogens, including B. anthracis, appear to possess a large repertoire of genes involved in trace element metabolism [59-71,94]. These genes may play an important role in establishing and maintaining infection, particularly within the phagolysosomes of macrophages or polymorpho-nuclear leukocytes [70,88-93].

Extensive research on B. cereus pathogenesis has resulted in the identification of many virulence factors and our knowledge about their mechanisms and interactions is still expanding. As an opportunistic pathogen, B. cereus is often associated with diarrheal and emetic food poisoning. Additionally, in rare cases involving individuals with impaired immune defenses or trauma B. cereus has been shown to cause severe systemic or local infection [17,31,34]. On the other hand, there is mounting evidence originating from genome research and eco-microbiological studies indicating that members of the B. cereus group may be normal inhabitants of arthropod intestines [6,95,96]. From the analysis of sequences obtained from gene discovery we found that almost all of the annotated VAGs were found among non-clinical isolates. CGH profiling also demonstrated that non-clinical strains, in general, had larger sets of putative virulence factors. It is known that the pleotropic regulatory gene plcR, is fully functional among the B. cereus but inactive in B. anthracis due to a frame-shift mutation [6,95,96]. In the case of Clade 4 genomes and B. anthracis in particular, evolution seem to entail reduction of VAGs for which regulatory factors such as PlcR are no longer necessary. It is worth noting here that almost all the chromosomal VAGs found to be characteristic for either B. anthracis or its near neighbors within Clade 4 were found only occasionally in genomes belonging to non-clinical B. cereus strains from distant lineages.

The finding of a PagA variant in the genome of an environmental isolate (B. weihenstephanensis AH1143) may provide some additional insights about the natural reservoir and extent of diversity of one of the B. anthracis toxin genes. Our finding, together with previously published evidence [13,97], further supports the hypothesis that variants of B. anthracis toxin genes such as pag and lef may exist among other members of B. cereus sensu lato group, although their role outside the mammalian host has yet to be determined.

Taken together, results of this study allowed us to better recognize the sources of genome diversity, understand the relationships among members of B. cereus group, and further elucidate aspects of their genome evolution, especially events associated with the evolution of B. anthracis lineage and its close relatives within this taxon. Furthermore, the type of analytical approach we presented here will help improve diagnostics and open the avenues for developing predictive cladistic models to improve our understanding of genome evolution associated with clone emergence. Such efforts will enable the identifications genomic markers that can be used for diagnostic purposes and finally assist both drug- and vaccine development.

Supplementary Material

07. Figure 1s.

Gene Discovery (GD). Random genomic libraries are generated for each B. cereus strain. Cloned genomic fragments are common (uncolored boxes) or unique (colored boxes). Plasmid DNAs are printed onto glass slide microarrays and subsequently probed with Cy-labeled query B. cereus and B. anthracis in a competitive hybridization. Those plasmids that do not hybridize with B. anthracis were considered to contain unique genomic fragments (novel genes) and were sequence characterized. Other plasmids that hybridized weakly were considered to contain divergent sequences were also sequenced.

08. Figure 2s.

Summary of the bioinformatic approach used to characterize the unique genomic fragments from B. cereus strains subjected gene discovery.

09. Figure 3s.

Relative frequencies of novel genes discovered among seven query B. cereus genomes based on predicted functional role categories.

10. Figure 4s.

Total gene content estimates for Bacillus cereus group members.

11. Figure 5s.

Comparative cluster analysis of strains based on CGH data. Left, clustering based on the global gene hybridization patterns (~29977 70-mer oligonucleotides) using the trinary designations “0”, “0.5” and “1” corresponding to “absent”, “divergent” and “present” CDSs respectively. Right, clustering of the 950 markers representing known plasmid sequences.