Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Comparative Transcriptional Profiling of Bacillus cereus Sensu Lato Strains during Growth in CO2-Bicarbonate and Aerobic Atmospheres

  • Karla D. Passalacqua,

    Affiliation School of Biology, Georgia Institute of Technology, Atlanta, Georgia, United States of America

  • Anjana Varadarajan,

    Affiliation School of Biology, Georgia Institute of Technology, Atlanta, Georgia, United States of America

  • Benjamin Byrd,

    Affiliation School of Biology, Georgia Institute of Technology, Atlanta, Georgia, United States of America

  • Nicholas H. Bergman

    Affiliations School of Biology, Georgia Institute of Technology, Atlanta, Georgia, United States of America, Electro-Optical Systems Laboratory, Georgia Tech Research Institute, Atlanta, Georgia, United States of America

Comparative Transcriptional Profiling of Bacillus cereus Sensu Lato Strains during Growth in CO2-Bicarbonate and Aerobic Atmospheres

  • Karla D. Passalacqua, 
  • Anjana Varadarajan, 
  • Benjamin Byrd, 
  • Nicholas H. Bergman



Bacillus species are spore-forming bacteria that are ubiquitous in the environment and display a range of virulent and avirulent phenotypes. This range is particularly evident in the Bacillus cereus sensu lato group; where closely related strains cause anthrax, food-borne illnesses, and pneumonia, but can also be non-pathogenic. Although much of this phenotypic range can be attributed to the presence or absence of a few key virulence factors, there are other virulence-associated loci that are conserved throughout the B. cereus group, and we hypothesized that these genes may be regulated differently in pathogenic and non-pathogenic strains.

Methodology/Principal Findings

Here we report transcriptional profiles of three closely related but phenotypically unique members of the Bacillus cereus group—a pneumonia-causing B. cereus strain (G9241), an attenuated strain of B. anthracis (Sterne 34F2), and an avirulent B. cereus strain (10987)—during exponential growth in two distinct atmospheric environments: 14% CO2/bicarbonate and ambient air. We show that the disease-causing Bacillus strains undergo more distinctive transcriptional changes between the two environments, and that the expression of plasmid-encoded virulence genes was increased exclusively in the CO2 environment. We observed a core of conserved metabolic genes that were differentially expressed in all three strains in both conditions. Additionally, the expression profiles of putative virulence genes in G9241 suggest that this strain, unlike Bacillus anthracis, may regulate gene expression with both PlcR and AtxA transcriptional regulators, each acting in a different environment.


We have shown that homologous and even identical genes within the genomes of three closely related members of the B. cereus sensu lato group are in some instances regulated very differently, and that these differences can have important implications for virulence. This study provides insights into the evolution of the B. cereus group, and highlights the importance of looking beyond differences in gene content in comparative genomics studies.


Bacillus species are a highly diverse, spore-forming group of bacteria that can be found ubiquitously in the environment. These microbes range from being beneficial or benign to causing diseases such as anthrax, pneumonia, periodontal disease, and food-borne illness [1][3]. Much effort has been put into elucidating the pathogenic potential of one of the more notorious Bacillus species, B. anthracis, the causative agent of anthrax. However, in 1994, a B. cereus strain (G9241) was isolated from a welder suffering from severe pneumonia that was similar in many ways to inhalational anthrax, and other similar cases of anthrax-like disease caused by B. cereus have since been observed [4], [5]. Significantly, although the genomes of B. cereus G9241 and B. anthracis share many similarities [6], [7], the two species carry different plasmids, and they are phenotypically quite different (e.g., B. cereus G9241 is motile and hemolytic, and B. anthracis is neither). Given all of this, the relationship between B. anthracis, B. cereus G9241 and other related strains of B. cereus provides a unique perspective from which to study both the evolutionary origins and the genetic basis for phenotypic diversity in the B. cereus group.

Large-scale genomic variation amongst species is a primary source of bacterial identity, but phenotypic differences are also manifested by smaller differences that drive selective expression of gene repertoires. These subtle differences between strains of B. cereus and B. anthracis have not yet been explored. The global transcriptional behavior of B. anthracis has recently been investigated under a variety of conditions [8][12], and several studies have demonstrated that the presence of CO2/bicarbonate, either in the host or in vitro, induces the expression of genes encoding the anthrax toxin components [13], [14]. CO2/bicarbonate is the main pH buffering system in the body, and the relative levels of these molecules affect many cellular, biochemical and physiological processes [15]. Bicarbonate is a small and labile biomolecule that converts to CO2 readily in various conditions, both spontaneously and when catalyzed by carbonic anyhydrase enzymes, and pH levels are influenced as interconversion occurs [15]. It seems intuitive, therefore, that bacteria interacting with mammalian hosts (particularly pathogens) might have developed ways of responding to and thriving within a bicarbonate-rich environment that is most likely different from typical conditions encountered during growth in rich media in vitro. Indeed, both CO2 and bicarbonate have been shown to be important in the interaction between the host and a variety of bacteria, including B. anthracis [16], [17]. The full genome of the pathogenic B. cereus strain G9241 has been sequenced [6], but the transcriptional behavior of this microbe has not been elucidated.

Given the connection between CO2/bicarbonate and pathogenesis in B. anthracis, and the fact that the CO2/O2 balance is known to affect B. anthracis gene expression in a profound way, we felt that a global study of gene expression in varying atmospheric growth conditions would provide meaningful insights into the functional and transcriptional differences between strains of B. cereus and B. anthracis. With this in mind, in this study we characterized the relative transcriptional profiles of three closely related but phenotypically unique Bacillus strains between CO2/bicarbonate-rich and highly oxygenated atmospheres in a semi-defined growth medium. We assayed the growth characteristics of two pathogenic Bacilli (B. cereus G9241 and the attenuated B. anthracis strain Sterne 34F2) [6], [18], [19] as well as an avirulent laboratory strain, (B. cereus 10987) that is closely related to B. anthracis [5], [6], [20] in each condition, then generated relative mRNA profiles using Nimblegen gene expression microarrays. The data show that the two pathogens are, in terms of transcriptional regulation, more sensitive to atmospheric conditions than avirulent strain 10987, and that their plasmid gene expression is almost universally induced in the CO2/bicarbonate background. We identified 27 genes sharing >90% protein identity in all 3 species that were similarly regulated in the two growth conditions in all 3 strains, revealing a core repertoire of transcriptional changes for Bacilli between CO2 and O2. We also observed that groups of genes putatively involved in cellular characteristics such as motility, cell morphology, and pathogenesis were transcribed differentially between the strains. Of special note, we observed that many genes putatively regulated by the transcriptional regulators PlcR and AtxA were differentially expressed in strain G9241 in O2 and CO2, respectively. Therefore, unlike B. anthracis, strain G9241 may utilize both PlcR and AtxA transcriptional regulators, but in exclusive environments. This study refines our knowledge of Bacillus genome architecture and regulation and has highlighted specific functional aspects of two pathogenic species that should aid in focusing future research into the physiological aspects of Bacillus behavior.

Results and Discussion

Growth in Modified G Medium: CO2 versus O2

For characterization of the global transcriptional patterns of B. cereus G9241, B. anthracis Sterne 34F2, and B. cereus 10987 (hereafter referred to as G9241, B. anthracis, and 10987, respectively), we first determined the growth characteristics of these three strains in a semi-defined medium under both high aeration and 14% CO2+0.8% bicarbonate conditions (hereafter referred to as O2 and CO2). The medium used in this study, Modified G Medium (MGM), is typically used to produce high levels of B. anthracis spores (see Methods). This medium was used in earlier transcriptional profiling studies of B. anthracis, and its use here provides for continuity with those studies [8], [9]. Figure 1 illustrates the growth characteristics of the three Bacilli in MGM in O2 and CO2 (1 experiment representative of 5). Overall, the three strains grew at a slightly slower rate in CO2 than in O2, which is unsurprising, since energy production is generally more efficient in aerobic atmospheres. Depending on the time points used to calculate doubling times, rates of doubling in O2 ranged from 24 to 44 minutes, whereas doubling times in CO2 ranged from 31–55 minutes. However, within each growth condition, the strains displayed similar growth kinetics (Figure 1A), and they all reached stationary phase within 6 hours.

Figure 1. Growth curves of strains B. cereus G9241, B. anthracis Sterne 34F2, and B. cereus 10987 in MGM with normal aeration and with 14% CO2+0.8% bicarbonate.

(A) The three Bacillus strains plotted against each other in each condition, exhibiting similar growth rates. (B) The three Bacillus species plotted against themselves in each condition, showing slightly slower growth rates for G9241 and B. anthracis Sterne in the CO2 environment. Curves are representative examples of 5 experiments with similar trends.

After observing that growth for each strain in each condition was robust, we sought to: (i) ascertain the major differences in the transcriptional profiles of each individual strain between growth in two very different atmospheres; and then (ii) make an inter-strain comparison of these overall transcriptional trends to ascertain differences and similarities between three closely-related but phenotypically unique Bacilli. Hence, because the strains grow at a slower rate in the CO2 atmosphere, we harvested RNA for transcriptome comparisons at equivalent population densities as measured by OD600 (∼0.4), which falls within the range of exponential growth for all strains. Microscopic examination of the strains in each condition did not indicate any overt morphological differences in either condition.

Overview of gene expression in CO2 v O2 - chromosomes & plasmids

Microarrays are powerful tools for global gene expression studies, but they are limited by the fact that they require sequenced genomes and can only measure what is contained on the chip. Although the genome sequence for strain G9241 has been reported [6], it has not yet been entirely assembled or annotated, and as a consequence, the custom microarray made for us by Nimblegen has several features to be noted here. First, G9241 contains several extrachromosomal features; a cryptic phage (pBC_Clin29) and two large plasmids (pBCXO1 and pBC218). At the time of our microarray design, only a portion of the pBCXO1 plasmid sequence was available to us for probe selection; so of the 177 putative pBCXO1 genes, 111 (62%) are included on our array. Also note that plasmid pBC218 was originally reported in Hoffmaster et al. [6] as being ∼218 kb, but a subsequent study of B. cereus plasmids revealed that it is more likely ∼210 kb ([21], and Dave Rasko, personal communication). Hence, this genetic element has been reported as both pBC218 and pBC210 in the literature. Here, we refer to it by its original name, pBC218, since locus tag numbers in the public databases still refer to it as such.

Figure 2 and Table 1 outline the general gene expression patterns in terms of numbers of genes differentially expressed between growth in CO2 versus O2 for each of the three strains. In this primary analysis, we included all genes that were differentially expressed as assessed by SAM (Significance Analysis of Microarrays - see Methods) with differences in expression level >2-fold (5 microarrays each strain in each condition: 4 biological replicates plus 1 technical replicate - see Methods). Note that SAM analyses result in fold-difference comparisons that are reported relative to the sample that is used as the initial query. Hence, in our comparisons, genes with decreased expression in CO2 were viewed as having increased expression in O2, and are referred to as such for clarity.

Figure 2. Global differences in transcription between CO2 and O2 growth conditions for three Bacillus strains.

(A–C) Histograms show numbers of genes more highly expressed in CO2 and O2 conditions (per SAM analysis), displaying total number of genes, number of genes differentially expressed by ranges of Fold Difference, and numbers of genes from various genetic elements (chromosomes and plasmids). (D) Sample Density Matrices (SDM) for microarray samples comparing levels of similarity between CO2 and O2 conditions (black representing high similarity and red low similarity). Each quadrant is representative of 5 microarray experiments in each growth condition. See text for details on SDM.

Table 1. Numbers of genes differentially expressed for 3 Bacillus strains during log-phase growth in MGM in CO2+bicarbonate versus O2.

Strain G9241 showed the greatest overall difference in expression profile between the two growth atmospheres (535 genes up in CO2 v 640 up in O2), followed by B. anthracis (265 up in CO2 v 238 up in O2), with 10987 showing the smallest difference in gene expression between the two conditions (182 up in CO2 v 109 up in O2) (Table 1). Given that G9241 is a recent clinical isolate, and has not been passaged for years under laboratory conditions as have the commonly used B. anthracis Sterne and B. cereus 10987 strains, the more robust differences in transcriptome profiles are perhaps not surprising. Looking more closely at these data, we noted that G9241 also induced genes at significantly higher levels (6–19 fold and >20 fold) in each condition than either B. anthracis or 10987 (Fig. 2A–C, Table 1). More interestingly, we observed that plasmid genes for G9241 and B. anthracis are almost exclusively expressed more highly in CO2 (Fig. 2A and B, Table 1). For G9241, 39 genes were expressed >20 fold higher in CO2, and 34 of them (∼87%) were plasmid-encoded. No genes from pBC218 were expressed higher in O2 at any level. For B. anthracis, 23 genes on pXO1 were more highly expressed in CO2, but only 9 of these were at levels >6 fold. (Fig. 2B, Table 1).

Global differences between transcriptional profiles in CO2 and O2 environments can be more clearly seen in sample distance matrices (SDM), which allowed us to directly visualize the overall similarity/dissimilarity between global expression profiles. Figure 2D shows SDMs for each chromosome and plasmid carried by the three strains, with the 5 separate microarray datasets arranged from top to bottom and left to right in the same order. The diagonal represents self-self comparisons and is left blank, while all other squares are colored based on the similarity between the two datasets that are represented by the row and column that intersect there (with black indicating high similarity and red indicating low similarity). Thus, the upper left and lower right quadrants show comparisons of the 5 replicate datasets collected in the same atmospheric condition, (i.e., CO2 v CO2 and O2 v O2); whereas the upper right and lower left quadrants show comparisons between the gene expression profiles observed under different atmospheric conditions (i.e., CO2 v O2). The SDM show that all three G9241 genetic elements had a striking change in global expression profile between CO2 and O2, as was suggested in the numbers of genes differentially expressed (Fig. 2, Table 1). In other words, in terms of overall expression, G9241 shows very different expression patterns in each of the two conditions. B. anthracis also showed a significantly different expression profile between the two conditions for both chromosomal and plasmid-encoded genes, but to a slightly lesser degree than G9241 (Fig. 2D). Strain 10987 showed more subtle differences in chromosomal gene expression between CO2 and O2, but expression from the pBC10987 plasmid appears to be quite similar in all samples. This plasmid is derived from the pXO1/pBCXO1 group [20], [21], but at least in terms of gene expression in CO2, it appears to have diverged considerably from its pathogenic relatives.

Whereas 88 genes on G9241's pBC218 plasmid showed increased expression in CO2 (out of 188 putative ORFs, ∼47%), and pBCXO1 had 24 genes upregulated (24 out of 111, ∼22%), we see that only a select region of the B. anthracis pXO1 plasmid showed an increase in expression (23 out of 204 ORFs, ∼11%) in CO2. Table 2 lists all plasmid genes for pBC218 (G92410), and Table 3 lists all genes for pBCXO1 (G9241) and pXO1 (B. anthracis) plasmids that had increased expression in CO2 >6-fold, and we noted several salient observations. First, genetic loci 0059–0073 from G9241's pBC218 represent a putative gene cluster for capsule biosynthesis [6], [22], and each of these showed substantially increased expression in CO2 (9–77 fold) (Table 2 footnote b). Sue et al. [22] observed (but did not quantify) capsule production in G9241 in both the presence and absence of CO2, in a variety of growth media, and our transcriptional data extend that finding by suggesting that CO2 may cause an increase in capsule production. More interestingly, we noted an extraordinary increase in expression in G9241 of three genes on the pBCXO1 plasmid (loci pBCXO1_0108-0110; up 123–226 fold) that belong to a family of transferases with undefined function (Table 3). Considering the dynamic range of measurements made by microarrays, fold-differences this high are rarely observed and indicate extreme differences in relative transcript abundance. Interestingly, one of these genes - galU (pBCXO1_0109) - shares 100% protein identity with the B. anthracis pXO1 galU gene (pXO1_0129), which was the only B. anthracis plasmid gene that showed higher expression in O2 (2.8 fold). Although the G9241 pBCXO1 and B. anthracis pXO1 plasmids differ slightly in terms of sequence [6], they appear to have diverged much more in terms of transcriptional regulation, at least as can be seen between O2 and CO2 growth (Table 3). This may be linked to the acquisition by B. anthracis of the pXO2 virulence plasmid, which contains genes that encode the enzymes necessary for the synthesis of a very different capsule [23] and its regulators [24]. It is tempting to speculate that these highly expressed pBCXO1 genes in G9241 might possibly be involved in augmenting pBC218 encoded capsule production for this strain, [5], and it may be interesting to explore the possibility of a synergistic relationship between pBCXO1 and pBC218 in capsule formation. Some genes located on G9241's pBC218 plasmid also displayed what can be considered extreme differences in transcript abundance in CO2 (>50 fold) (Table 2), including several hypothetical proteins, a non-hemolytic enterotoxin component (see section “Pathogenesis genes”), and an arsR transcriptional regulator (locus 0045). Overall, these data show that the pathogenic Bacillus strains have evolved ways to adapt gene expression in a more profound and extensive way in different atmospheric conditions. Avirulent strain 10987, on the other hand, stands in stark counterpoint to the pathogens, where relative transcriptional activity between CO2 and O2 appears to be only slightly regulated.

Table 2. Genes with increased expression (≥6-fold) for the B. cereus G9241 pBC218 plasmid in CO2+0.8% bicarbonate.

Table 3. Genes with increased expression in CO2+0.8% bicarbonate (≥6-fold) for the B. cereus G9241 pBCXO1 plasmid and the B. anthracis pXO1 plasmid.

Core Bacillus gene expression

Comparing transcriptomes of phenotypically unique but closely related bacteria [25] is not a common approach for microarray studies, but here it provides a perspective on core expression programs that are used by related strains in adapting to environmental stimuli. Therefore, with expression data for three unique microbes in hand, we asked what core similarities exist amongst these three related but unique strains that could indicate a highly conserved set of genes important for adaptation to varying atmospheric conditions for the Bacillus cereus group. First, we performed genome-wide BLASTp analysis and generated 2 lists: (i) all genes that share >90% protein identity between the two pathogens G9241 and B. anthracis (n = 3,076 genes); and (ii) a list of genes that share >90% protein identity in all 3 strains (n = 2,624) (see Methods). We then compared these lists to each other and identified the genes that were both conserved and differentially expressed in each strain in CO2 and O2. Finally, we compared the conserved and expressed gene lists to each other to identify common genes that were differentially expressed in CO2 and O2 in all 3 strains. Figure 3 shows four Venn diagrams outlining these results, and Tables 4, S1 and S2 list specific genes with fold-differences.

Figure 3. Venn diagrams representing numbers of genes differentially expressed in CO2 and O2 that share >90% protein identity in the Bacillus strains indicated.

(A) Comparisons of the two pathogenic strains, B. cereus G9241 and B. anthracis (Sterne 34F2) from a list of 3,076 putatively common genes. (B) Comparisons of all three Bacillus strains (including B. cereus 10987) from a list of 2,624 putatively common genes.

Table 4. List of conserved genes (>90% protein identity) that are differentially expressed in CO2 and O2 environments in all 3 Bacillus strains. (>2-fold).

In comparing the two pathogenic strains, G9241 and B. anthracis (Fig. 3A), 68 common genes were more induced in CO2, and 60 common genes were more induced in O2 (Tables S1 and S2). Of note are several genes that were substantially (>20 fold) induced in the two conditions. Two ATP-binding transporters (cydC and cydD) were substantially up-regulated in CO2 in these strains (29–65 fold), suggesting that as in B. subtilis, the cytochrome bd complex may be active under CO2 conditions [26]. Also, our data suggest that L-lysine catabolism is important for growth in CO2 in MGM for G9241 and B. anthracis, since the kamA transcript (L-lysine 2,3-aminomutase) is >23-fold more abundant in this atmosphere [27] (Table S1). In the O2 environment, the most conspicuous genes conserved and differentially expressed in the two pathogens were two genes putatively involved in trehalose transport and metabolism (treB and treC) (Table S2), revealing another way in which the two pathogens have adopted modes of gene regulation not observed in strain 10987.

For genes that are conserved in all three strains, 20 were commonly induced in CO2, while 7 were common in O2 (Fig. 3B, Table 4). These data suggest that the various cytochrome c oxidases, reductases, CoA synthetase and dehydrogenase genes are metabolic proteins most important to Bacilli in a CO2/bicarbonate atmosphere. Two specific genes jump out as perhaps being of utmost importance for growth in each conditions: (i) an anaerobic ribonucleoside triphosphate reductase in CO2, and (ii) an oxalate∶formate antiporter in O2.

The biochemical properties of anaerobic ribonucleoside triphosphate reductase have been studied extensively in E. coli [28][31], and this enzyme is needed for the making of deoxyribonucleoside triphosphates in anaerobic conditions for this microbe. Considering the substantial upregulation of this gene in CO2 (8–11 fold difference) in all three strains, it is likely of critical importance for Bacillus replication in a CO2/bicarbonate environment. Previously, the ribonucleotide reductase gene NrdF of B. anthracis was characterized and suggested as a possible antimicrobial target [32]. However, because the CO2/bicarbonate environment is considered to be somewhat analogous to conditions within the host [16], this other nucleic acid gene/enzyme may be another candidate as a possible target for antimicrobial reagents for all pathogenic Bacilli. Regarding the putative oxalate∶formate antiporter more highly expressed in O2 in all three strains, special note should be taken of the label “putative”, since the COG term for this gene is listed as “nitrate/nitrite antiporter”, and the InterPro and Pfam descriptions are simply listed as “Major Facilitator Superfamily”. Hence, it is difficult to say whether or not this transporter specifically moves oxalate/formate, but the conserved expression in an O2 environment suggests that it is an important transporter in an aerobic environment. However, because this type of growth is not considered to be analogous to the host, the conserved expression of this gene is mainly illustrative of the general metabolic capabilities and gene regulation in the Bacilli.

Relative transcriptional profiles in CO2 versus O2 (≥6-Fold Difference)

The previous global analyses outlined differential gene expression profiles for the three Bacillus strains in a broad and global view, and included all genes with fold differences >2 between CO2 and O2. We list in Tables 5, 6, 7, 8, 9 those genes with >6-fold difference between the two growth states (CO2 and O2) by putative biological & cellular function for each strain. They are organized first by broad putative functional family (gene families include Energy & Metabolism, Spore Function, Cellular & Enzymatic Activity, Amino Acids, Nucleic Acids, Transport, Regulation, Motility & Chemotaxis, and Pathogenesis & Toxins), then by Bacillus strain, and then by more specific COG, GO or Pfam categories (Genes of “Unknown Function” and “Hypothetical Proteins” expressed >6-fold are listed in Tables S3 and S4). The >6-fold cutoff was chosen with the goal of highlighting those genes to be of the most biological significance, and the complete list of differentially expressed genes >2-fold are listed in Supplemental Tables S5, S6, S7, S8, S9, S10 by locus tag number. Several trends of note, including cellular attributes such as motility, cell structure, and genes putatively involved in pathogenic potential, are discussed.

Table 5. Chromosomal genes with increased expression (≥6-fold) for three Bacillus strains in MGM+CO2+0.8% bicarbonate for functional families: Energy & Metabolism; Spore Function; and Cellular & Enzymatic Activity.

Table 6. Chromosomal genes with increased expression (≥6-fold) for three Bacillus strains in MGM+CO2+0.8% bicarbonate for functional families: Amino Acids; Nucleic Acids; and Transport.

Table 7. Chromosomal genes with increased expression (≥6-fold) for three Bacillus strains in MGM with normal aeration (O2) for functional families: Energy & Metabolism and Cellular & Enzymatic Activity.

Table 8. Chromosomal genes with increased expression (≥6-fold) for three Bacillus strains in MGM with normal aeration (O2) for functional families: Amino Acids; Nucleic Acids; and Transport.

Table 9. Chromosomal genes with increased expression (≥6-fold) for three Bacillus strains in MGM with normal aeration (O2) for functional families: Regulation; Motility & Chemotaxis; and Pathogenesis & Toxins.

Because our array data showed that the two pathogens, B. anthracis and G9241, exhibited such pronounced differential expression patterns between the two conditions (often >9–200 fold difference), we performed qRT-PCR on a select set of genes for these two strains to verify that the trends we had observed were reproducible by a complementary approach. The genes assayed were from both chromosomal and plasmid loci, and represent both up- and down-regulated genes between the two growth conditions (see Tables S11 and S12 for gene names and primer sequences). These results are shown in Figure 4, and clearly display good correlation between the two assays (Pearson r = −0.7735 for B. anthracis and −0.7757 for G9241, p<0.0004 for both).

Figure 4. Spearman correlations comparing microarray and SYBR-Green qRT-PCR gene expression results for select genes in two conditions - CO2 and O2.

(A) B. cereus G9241; expression for 12 genes in 2 conditions (5 chromosomal, 4 from pBC218, and 3 from pBCXO1); Spearman r = −0.7797; p<0.0001; and (B) B. anthracis Sterne 34F2; 8 genes in two conditions (5 chromosomal, 3 from pXO1); Spearman r = −0.7735; p = 0.0004. Genes and primer sequences listed in Tables S11 and S12.

Motility genes

B. anthracis is non-motile, but retains a set of motility genes, some of which contain disrupting mutations [18], [33]. B. cereus strains exhibit a range of motility behavior [3], and both G9241 and 10987 displayed observable motility by microscopy when grown in rich broth. G9241 showed conspicuously increased expression of motility genes during growth in O2 (Table 9), including those encoding flagellar components and chemotaxis sensory genes (18 genes >6-fold; 19 genes between 2–5 fold (Table S13 summarizes differentially expressed motility genes for G9241 and B. anthracis)). This substantial trend suggested that motility may be reduced in MGM in the CO2 environment for G9241, and we confirmed this trait via microscopy of cells from liquid cultures. Because microscopic examination is qualitative, we also used B. subtilis as a baseline model for robust motility. Both B. cereus 10987 and B. subtilis 168 maintained motility in the CO2/bicarbonate environment as observed by microscopy, but were qualitatively more motile in the O2 non-bicarbonate environment. G9241 did not appear motile in CO2, but was quite motile in MGM in full aeration, as well as in rich LB medium.

For B. anthracis, 7 methyl-accepting chemotaxis protein (MCP) genes and 1 putative flagellar motor hook gene were increased in expression in O2 between 2–5 fold (Table S13). Of these 8 genes, 5 are included in the list of genes conserved and differentially expressed commonly with G9241 (Table S2). Therefore, even though B. anthracis has not been reported to be motile in any previous study (and was negative for observable motility in all media and atmospheres tested in this study) we see that various genes implicated in motility have been maintained and are expressed in a similar fashion to a fully motile relative, perhaps suggesting that these chemo-sensors play roles in traits other than motility. One of these MCP genes that is present in both B. anthracis and G9241 (GBAA5317 homologous to G9241_5185) may be expressed as part of a two-gene operon, since their homologous upstream neighbors (GBAA5318 and G9241_5186: putative endonuclease/exonuclease/phosphatase family protein) were also more highly expressed in O2 for both strains, and there is a very small (25 nt) intergenic region. Transcriptome sequencing data from a different project support this observation (Passalacqua et al., unpublished), where these two genes for B. anthracis are apparently transcribed on a contiguous mRNA unit. BLASTp analysis of this putative endonuclease in the JGI IMG ( shows homologs present in the pathogenic Bacillus cereus sensu lato group >90% (B. anthracis, B. cereus, B. thuringiensis), but identities quickly drop off for microbes in the rest of this database, implicating a specific role for these genes in this group of bacteria.

S-Layer genes

S-layers are protective cell coats that are involved in a variety of bacterial characteristics, such as cell adhesion, permeability layers, and protein scaffolds, and multiple genes in many prokaryotic genomes are implicated in S-layer structure and affiliation (see Review [34]). In B. anthracis, the chromosomal genes for the S-layer proteins EA1 and Sap have been previously characterized and were shown to be the main components of the B. anthracis S-layer [35]. Also, a pXO1-encoded S-layer protein (BslA) was recently shown to play a role in the ability of B. anthracis vegetative cells to adhere to host cells [36]. We observed that the B. anthracis chromosomal EA1 (GBAA0887) gene and the pXO1 encoded gene BslA (pXO1_0124) were both significantly more highly expressed in CO2 (21 and 35 fold respectively, Tables 5 and 3) (all putative S-layer genes differentially expressed in CO2 and O2 for all three Bacillus strains are listed together in Table S14). However, no transcriptional differences were observed in B. anthracis for Sap, and both microarray fluorescence intensities and transcriptome sequencing data suggest that this gene is constitutively expressed at a substantial level in both conditions (Passalacqua et al., unpublished). Of 11 other putative S-layer proteins on the B. anthracis chromosome, one had higher expression in CO2 (GBAA1130, 6-fold) and three were more highly expressed in O2 (GBAA0981, 1926, and 2315, up 3, 5 and 2.4-fold in O2 respectively; Table S14). These data suggest that in addition to the important Sap component of the B. anthracis S-layer, this microbe expresses other S-layer components differentially, likely due to growth condition-specific needs.

The differential expression of the G9241 putative S-layer genes showed both similarities to and differences from the pattern in B. anthracis. First, the G9241 homolog of BslA, located on the pBCXO1 plasmid (pBCXO1_0105), was, as in B. anthracis, increased substantially in CO2 (12 fold) (Tables 3 and S14). However, for the two G9241 chromosomal S-layer homologs of EA1 and Sap, the G9241 EA1 gene showed no difference in transcript abundance in CO2 and Sap was ∼3 fold more highly expressed in O2. The G9241 S-layer gene locus 1115 (homolog to B. anthracis locus 1130) was increased in CO2 as in B. anthracis (∼2.5 fold each - 96% protein identity) (Table S14). Also, the chromosomal homolog to one of the B. anthracis genes more highly expressed in O2 showed increased expression in G9241, but to a substantially larger degree (Tables 7 and S14) (G9241 locus 0996 up 28 fold versus 3 fold for B. anthracis locus 0981 (95% protein identity)). One last chromosomal S-layer gene for G9241 showed higher expression in O2 (locus 4903 up 9.5 fold) (Table 7), and this gene shares ∼96% protein identity with the B. anthracis locus 5054 and the 10987 locus 4952. The 10987 homolog was the only putative S-layer protein to be differentially expressed in this strain (up ∼2 fold in O2), whereas the B. anthracis version did not show differential expression.

Two putative S-layer genes on the pBC218 plasmid of G9241 were highly upregulated in CO2, by 28 and 41 fold (pBC218_0012 and 0043, respectively) (Tables 2 and S14). The former gene is a homolog to the B. anthracis pXO2-encoded S-layer protein, which would not be detected in these experiments for B. anthracis due to the lack of pXO2 in the Sterne strain. The latter gene does not have a closely related homolog in either B. anthracis or in 10987, and may represent a unique S-layer component in G9241, since a protein BLAST analysis of this gene with a threshold of >90% homology against all microbial genomes in the JGI IMG database resulted in no hits, not even in the numerous other sequenced Bacillus genomes. The differential expression patterns of S-layer genes between CO2 and O2 conditions suggest that B. anthracis evolved a regulatory pattern for the EA1 S-layer component unique from both B. cereus strains and that the pathogenic Bacillus strains may adapt unique S-layer structures differentially in response to various growth conditions, contributing to the phenotypic uniqueness of each strain.

Pathogenesis genes

Anthrax toxin has been an active topic of research, and control of toxin gene expression in B. anthracis in CO2/bicarbonate by the important regulator AtxA has been well-studied [11], [24], [37][39]. The anthrax toxin genes and the AtxA regulator are encoded on the B. anthracis pXO1 virulence plasmid, but this regulator also controls the expression of various chromosomal genes [11]. Regarding toxin expression profiles for B. anthracis in this study, we note here mainly that our results match previous work in showing increased expression of the key genes involved in toxin production in CO2/bicarbonate growth (lef, pagA, cyaA, and atxA; pXO1_0172, 0164, 0142, and 0146 respectively) (Table 3 and Table S7). Recently, it has been shown that a particular bicarbonate transporter (B. anthracis Sterne loci BAS2712-14, orthologous to Ames Ancestor GBAA2920-22) is important for B. anthracis virulence [16]. However, these transporter genes were not differentially expressed between CO2 and O2; perhaps they are constitutive, which seems consistent with a critical role in rapid assimilation to a bicarbonate-rich environment.

Outside of the well-characterized determinants in B. anthracis (i.e., toxins), what is considered a virulence factor in B. anthracis is a matter of interpretation. We simply note that the scope of this study for B. anthracis was to attain a global view of transcriptional activity between CO2/bicarbonate and O2, and so we direct the reader to the various tables for more gene-specific details. Regarding G9241, however, a primary goal of this study was to elucidate this unusual pathogen's transcriptional behavior in both a global and specific sense, and in terms of putative virulence determinants, this microorganism displays a high level of CO2 versus O2 specific transcriptional regulation.

B. anthracis and G9241 harbor two operons putatively encoding siderophore biosynthetic genes for iron acquisition: the petrobactin operon (B. anthracis loci GBAA1981-86; G9241 loci BCE_G9241_1997-2002), and the dhb operon (B. anthracis loci GBAA2368-72; G9241 2333-37). Strain 10987 only harbors the dhb operon, which is highly conserved in Bacilli (BCE_2398-2402) [40]. The petrobactin operon has been implicated in the virulence of B. anthracis [41], and we have observed that expression of this operon for B. anthracis during exponential growth is constitutive (KDP, unpublished), whereas the dhb operon of B. anthracis is tightly regulated by iron (KDP, unpublished), akin to that seen in B. subtilis [42]. For G9241, several genes from both operons were more highly transcribed in the CO2 environment (Table S5): dhb genes up ∼3–11 fold; petrobactin genes up ∼3–5 fold). Regarding the petrobactin operon, it does not appear that iron itself is the key regulatory factor for this operon in B. anthracis, and thus, we only note that the 100 bp upstream intergenic sequences where putative regulatory sequences most likely are located are slightly divergent between G9241 and B. anthracis (4 differences of 1–2 nt each), and may be contributing to the small differential expression for G9241. More interestingly, the 100 bp upstream nucleotide sequences of the dhb operon for B. anthracis and 10987 differ by only 2 single nucleotides, with exact matching Fur iron regulatory protein consensus sequences located 65 base pairs upstream of the translational start sites [43], whereas the canonical Fur binding sequence in G9241 is located 75 bp upstream of the translational start site. Because changes in regulatory regions are a major force in the evolution of genetic regulation [44], further analysis of these important sequences should shed much light on the evolution of gene regulation in bacteria.

B. cereus strains are very diverse in the specific toxin genes that they harbor, and these toxins have been implicated in different types of food-borne illnesses [45][48]. Here, G9241 displayed very strong induction of multiple chromosomal enteroxin, hemolysin, and phospholipase genes in O2 (33–200 fold differences), the homologs of which have been shown previously to be regulated by the transcriptional regulator PlcR in B. cereus strain 14579 during stationary phase [49] (19 out of 27 G9241 genes sharing >85% protein identity with strain 14589 plcR regulated genes in [49]) (Table 9 - Pathogenesis & Toxins, PlcR regulated homologs marked with **). Hence, in our study, it seems that putative PlcR-regulated virulence determinants were strongly correlated with growth in an aerobic environment. This is in agreement with previous observations that cereulide toxin production in B. cereus is influenced by the presence of atmospheric O2 [50]. The G9241 plcR gene itself (locus 5525, ∼72% protein identity to B. cereus 14579 PlcR) is ∼2.4 fold more highly expressed in O2 as well (Table S6). Homologs of many of these toxin genes exist in the genomes of B. anthracis and 10987, but do not show differential expression in O2 for these strains. Interestingly, three additional, putative enterotoxin genes are found on G9241's pBC218 plasmid (pBC218_0039, 0040 and 0042) (Table 2) and, unlike the chromosomally encoded toxins, they are substantially more highly expressed in CO2 (42, 75 and 24 fold difference, respectively). These genes do not share protein identity >41% with any bacteria in the JGI IMG database, so it appears that G9241 has evolved various regulatory strategies for unique toxin expression in each growth condition. The differential expression of various enterotoxin genes has been observed in different B. cereus under a variety of conditions, including anaerobic growth and availability of specific nutrients [51], [52]. Given, however, that G9241 has only been seen to cause a pneumonia-like disease [5], and has not been demonstrated to cause food-borne illness like many of the other B. cereus strains, it is not clear what role enterotoxins or their differential regulation may play in the pathogenesis of this strain.

In B. anthracis, the plcR gene contains a nonsense mutation that prevents the protein from being fully translated [53]. However, as mentioned, B. anthracis has the AtxA regulator, whose gene is located on the pXO1 plasmid, and it controls both pXO1 virulence determinants and chromosomal genes [11]. We observed that the atxA gene and the toxin genes it regulates in B. anthracis were more highly expressed in CO2 for this bacterium (Table 3). It has been proposed that a certain incompatibility may exist between PlcR and AtxA regulons [54], particularly in regard to sporulation. In G9241, however, the situation appears to be somewhat complex, where two homologs to the B. anthracis atxA gene exist and an active PlcR regulator appears to function (as inferred from the upregulation of various toxin genes; see previous). In looking at this situation in more detail, we noted that gene locus pBCXO1_0126 in G9241 is currently annotated as a “conserved hypothetical protein”. However, the nucleotide sequence of this gene (1428 bp) is 100% identical to the B. anthracis pXO1_0146 atxA gene, indicating an error in the pBCXO1 annotation. As in B. anthracis, this G9241 gene that we now consider to be the atxA gene showed higher expression in CO2 by ∼5.6 fold (Table 3). A less conserved gene on the pBC218 plasmid (locus 0050) is annotated as a “trans-acting activator”, and it shares ∼79% protein identity with B. anthracis atxA, but it was not differentially expressed between CO2 and O2. Lastly a small gene on G9241's pBC218 plasmid (locus 0049) is annotated as a “trans-acting positive regulator AtxA”; however, it is only 420 bp compared to 1428 bp full length B. anthracis atxA gene, and it actually shares ∼81% protein identity with the B. anthracis pXO1_0148 hypothetical protein; therefore, it is unclear what relation it bears to the B. anthracis AtxA regulator, if any. The G9241 pBCXO1 loci 0105 and 0106 are increased in expression in CO2 (∼13 and 10-fold, respectively), and these genes are homologous to the B. anthracis pXO1 loci 0124 and 0125 which have been shown to be strongly induced by AtxA [11] (Table 3). Therefore, upon closer inspection, we see that G9241 likely has both an active PlcR-like regulator and an AtxA-like regulator, which are themselves differentially expressed between O2 and CO2. This suggests that any incompatibility between the two regulators may be avoided in G9241 by this microbe's specific utilization of each regulator during unique growth environments.

Conclusions and future directions

Global transcriptional analyses via microarrays provide wonderful genome-scale views of gene expression, and inevitably supply multiple avenues for more detailed, gene-specific studies. Here, we chose to explore global gene expression patterns in three closely related but phenotypically unique Bacillus strains in two very different atmospheric growth conditions (CO2/bicarbonate versus O2). The data presented highlight both conservation and divergence amongst the three Bacilli, and in particular, revealed a high level of atmosphere-dependent genetic regulation for the two pathogenic strains (B. cereus G9241 and B. anthracis). A set of genes involved in general metabolic processes showed a conserved pattern of differential expression in all three strains between the two growth conditions, and mutagenesis experiments might reveal how vital these genes are for the genus Bacillus as a whole. Conversely, genes involved in motility, S-layer structures, and pathogenesis appeared to be expressed in a regulated manner only in the pathogens, suggesting that these microbes are able to modify their metabolic and structural features in a more profound way in response to the environment. In particular, transcription of genes located on virulence plasmids for G9241 and B. anthracis only showed increased expression in the CO2 environment, while the one plasmid of B. cereus 10987, despite being ancestrally derived from the pXO1 plasmid family, did not show any regulatory, transcriptional changes between the two conditions. Interestingly, G9241 did not display motility when grown planktonically in CO2/bicarbonate, but rather appeared to be upregulating capsule production in this putatively host-like environment. The most unusual finding was that, unlike B. anthracis, G9241 may utilize both PlcR and AtxA regulators, but each in a different growth condition. Gene-specific mutagenesis and biochemical strategies will surely continue to elucidate the complex regulatory repertoires of these interesting bacteria.

Materials and Methods

Bacterial strains and culture conditions for RNA isolation

Strains used in this study are Bacillus anthracis Sterne (34F2); Bacillus cereus G9241 (acquired from the Centers for Disease Control (CDC), Atlanta, GA); and Bacillus cereus 10987 (acquired from the American Type Culture Collection (ATCC)). Overnight cultures were started by picking a fresh colony from a TSA blood agar plate (B. anthracis was negative for hemolysis, and B. cereus strains showed positive hemolysis) and inoculating 5 ml of Luria-Bertani (LB) medium and grown at 37°C with shaking at 150 rpm. Overnight cultures were back-diluted in the morning to an optical density (600 nm) (OD600) of 0.1 in 5 ml of LB medium and allowed to recover to OD600 0.3. These cultures were then used to inoculate 60 ml of Modified G Medium (MGM) (recipe available at, or 60 ml of MGM+0.8% sodium bicarbonate (NaHCO3), to an OD600 of 0.01, in 500 ml volume erlenmeyer flasks. Cells were grown to mid-exponential phase (OD600 0.4–0.5) at 37°C with shaking at 250 rpm. Cells were grown in either ambient air (MGM only) or in the presence of 14–15% CO2 (MGM+0.8% NaHCO3). All flasks were sealed with “Bugstopper” vented plugs (Fisher Scientific) to allow gas exchange. All experiments were performed such that 4 biological replicates were collected for each strain in each condition.

RNA Isolation and cDNA Synthesis

RNA collection was done essentially as described previously [9], [55]. Briefly, when cells reached the target OD600, they were filtered in sterile 150 ml Nalgene 0.2 µm filter flasks, immediately resuspended in 13 ml RNase/DNase-free sterile water and moved to a sterile 50 ml RNase/DNase-free centrifuge tubes (Corning 4558). For B. cereus G9241 grown in the presence of CO2, cells could not be filtered due to capsule expression (filters were clogged quickly), so they were spun by centrifugation at top speed for 3 minutes, after which supernatants were poured off. Boiling lysis buffer (6.5 ml total: 0.2% SDS, 16 mM EDTA, 200 mM NaCl) was added and the tubes were incubated at 100°C for 5 minutes. The time between cell filtering, centrifugation and suspension in boiling lysis buffer was <5 minutes. Note that in previous studies we directly compared mRNA collected by centrifugation and by filtration and found no differences in the two methods (N. Bergman and E. Anderson, unpublished). We also found no differences in cell morphology, CFU, or RNA yield in any of the strains or conditions examined in this study.

For RNA extraction, samples were extracted three times with 20 ml of 65°C phenol (Sigma P4682), 20 ml of Phenol∶Chloroform∶Isoamyl Alcohol (25∶24∶1) (Sigma P3803), and finally, 20 ml of Chloroform∶Isoamyl Alcohol (Sigma C0549), after which the final aqueous layer was removed to a fresh centrifuge tube and precipitated with 2.5 volumes of 100% ethanol (Sigma E7023). After washing the pellet with 25 ml 70% ethanol, the RNA was resuspended in 200 µl sterile RNase/DNase-free water and stored at −20°C. For small RNA, tRNA, and genomic DNA removal, the Qiagen RNEasy Mini-Kit RNA Cleanup protocol was used per the manufacturer's instructions for 100 µg of each raw RNA extraction with on-column DNase digestion. All samples were assayed for RNA integrity on the Bio-Rad Experion Automated Electrophoresis Station with Prokaryotic RNA StdSens chips.

After RNA cleanup, 10 µg of each RNA sample were used to make cDNA as follows: all RNA used had a 260/280 ratio≥1.7, and a 260/230 ratio≥1.5. (Following protocol based on Nimblegen cDNA synthesis procedures). The Invitrogen SuperScript II Double Stranded cDNA synthesis kit (Cat. no. 11917) was used per standard protocols (full detailed protocol available at, with 200 ng Invitrogen Random hexamer primers for first-strand synthesis. All RNA and cDNA concentrations were checked on a Nanodrop 1000 spectrophotometer.

Growth Curves

Strains B. anthracis Sterne 34F2, B. cereus 10987, and B. cereus G9241 were grown on TSA blood agar plates. Colonies were picked to inoculate 3 ml of LB medium and grown at 37°C to OD600∼0.4–0.5. These cultures were used to inoculate 30 ml of MGM or MGM+0.8% Sodium Bicarbonate to an OD600 of ∼0.01. Cultures were shaken at 250 rpm at 37°C for 6 hours (bicarbonate flasks incubated in 14–15% CO2 conditions) and OD600 was recorded every hour. Growth was plotted in Prism 5 for Mac OSX. Doubling times were calculated per the method of Moat, Foster & Spector [56] using OD600.

Microarray Expression Analysis

cDNA prepared as stated above was sent to Roche Nimblegen to be run on Gene Expression Microarrays: B. anthracis Ames Ancestor chip design TI261594 (Catalog design for Bacillus anthracis str. ‘Ames Ancestor’ covering NC_007322, NC_007323, NC_007530. Probes selected for 5617/5617 sequences. Median number of probes/sequence is 17 with an average of 17.00.); B. cereus 10987 chip design TI222523 (Catalog design for Bacillus cereus ATCC 10987 covering NC_003909, NC_005707. Probes selected for 5844/5844 sequences. Median number of probes/sequence is 13 with an average of 13.00.); and a custom B. cereus G9241 chip designed by Nimblegen using files provided by Timothy Read, containing 6147 sequences, resulting in an array with 14 probes per gene and 5 copies of each probe. In total, 30 expression arrays were run as follows: for each of the 3 Bacillus species, 5 arrays were run for each condition (ambient air versus CO2 for 10 arrays/species), representing 4 biological replicates and one additional replicate representing cDNA made from a mix of RNA from each of 4 biological replicates. Raw expression data were normalized using RMA [57]. Fluorescence intensities for each of the probe replicates were averaged, log2-transformed, and used for further analysis. Expression analyses were done on Mev 4.0 (TIGR software) for Mac OSX. Significance Analysis of Microarrays (SAM) analyses were done with a false discovery rate (FDR)<0.001 and a fold-change cutoff of 2.0. Sample Density Matrices were also generated in Mev 4.0. Functional family analyses were done using the EASE algorithm [58] and TIGRFAM and GO tables compiled from the TIGR Comprehensive Microbial Resource ( Note that for B. cereus G9241, a functional family list was created to correspond to its closest hits in the B. anthracis genome as follows. BLAST was performed for all G9241 genes against a formatted database consisting of genes from the B. anthracis Ames Ancestor strain. The e-value cutoff was 1e-20. In cases of genes for which more than one homolog were found, only the best BLAST hit was considered and the others were ignored. The genes found in G9241 were mapped to the families of their corresponding homologs found in the B. anthracis Ames Ancestor list. The genes from G9241 that had no homologs in B. anthracis were classified as “unique” genes and were stored in a separate file.

Microarray Data

Microarray data are available in the EBI ArrayExpress database under accession number E-MEXP-2036.

SYBR-Green quantitative RT-PCR of select genes

SYBR Green quantitative RT-PCR (qRT-PCR) was used to confirm differential transcript expression between samples. For B. anthracis 34F2, 8 genes were chosen for analysis: 5 located on the chromosome and 3 located on the pXO1 plasmid. For B. cereus G9241, 12 genes were chosen for analysis: 5 located on the chromosome; 4 located on the pBC218 plasmid, and 3 putatively located on the pBCXO1 plasmid (see Supplemental Tables S11 and S12 for gene names and primer sequences). The fusA gene (B. anthracis GBAA 0107), which has been previously shown to be expressed constitutively in a wide variety of growth conditions [9], [55] and its closest homolog in B. cereus G9241 (BCE_G9241_0105) were used for relative quantitation of transcripts. Primers were designed to amplify sequences within the open reading frames and to result in PCR products between 150–200 base pairs. All primer pairs were tested with genomic DNA to confirm that only one amplification product was produced at the optimal annealing temperature. Experiments were performed using Applied Biosystems (ABI) Power RNA-to-Ct SYBR Mix 1-Step (ABI 4389986) with 3 experimental replicates, 2 no RT controls, and 1 no RNA control (all negative). 20 µl reactions were mixed in 96-well ABI Optical Reaction Plates as follows: 10 µl 2× Reaction Mix, 2 µl forward primer and 2 µl reverse primer (100 nM final), 0.16 µl RT mix, 1 µl RNA (70 ng total) and 4.84 µl H2O. Each primer pair was used for a separate assay using RNA from CO2 and from O2 samples. Plates were analyzed on an ABI Prism 7000 with the following protocol: 1× 48°C 30 min, 1× 95°C 10 min, 40× 95°C 15 s, 60°C 60 s. Data were analyzed using the ABI Prism 7000 SDS Software with the Auto Ct as threshold. Microarray fluorescence intensities (the average of 5 biological replicates - log2) and qRT-PCR Ct values (averages of 3 technical replicates) were plotted in Prism 5 for Mac OSX and analyzed by a Spearman Rank Correlation test.

Motility Assays

Planktonic cells of strains B. cereus G9241, B. cereus 10987, B. anthracis Sterne 34F2, and B. subtilis 168 (obtained from ATCC) were mounted on glass cover slips and hanging drop slides and observed under 40× magnification via phase contrast microscopy for motility. Bacteria were grown as follows: fresh colonies from a TSA blood-agar plate were picked to inoculate 15 ml of LB medium, MGM, and MGM+0.8% sodium bicarbonate, and grown at 37°C with normal aeration or with 10% CO2 for bicarbonate cultures, with shaking at 200 rpm to mid log phase (OD600∼0.4–0.5). Flasks were sealed with “Bug-Stopper” vented plugs for gas exchange.

Bioinformatic analyses for Venn Diagram

FASTA protein sequences from Bacillus anthracis Ames Ancestor (AA), Bacillus cereus G9241, and Bacillus cereus 10987 genomes were compared to each other using a local installation of BLAST as follows: AA versus G9241, G9241 versus 10987, and 10987 versus AA. Then, 2 lists were compiled as follows. List one comprised genes between AA and G9241 that shared >90% protein identity at the amino acid level (total = 3,076). List two comprised genes that shared >90% identity at the amino acid level across all three genomes (total = 2,624). If more than one gene shared identity at >90%, the higher value was kept. For Venn diagrams, all genes that were differentially expressed in each species per SAM analysis were compared to the aforementioned 90% lists, and those contained on the list(s) were separated into a new “conserved & differentially expressed” list. Then, these conserved expression lists were compared to each other for Venn diagram counts.

Nucleotide sequence alignments for intergenic sequences

Nucleotide sequence alignments for the 150 base pair regions for B. anthracis, B. cereus G9241 and B. cereus 10987 directly upstream of the petrobactin and dhb biosynthetic operons were done on CLC DNA Workbench version 4.1.2 for Mac OSX, on the slow (very accurate) setting with Gap Open and Gap Extension Costs of 10 (default).

Supporting Information

Table S1.

Genes more highly expressed in CO2+bicarbonate for B. cereus G9241 and B. anthracis Sterne 34F2 that share >90% protein identity

(0.13 MB PDF)

Table S2.

Genes more highly expressed in O2 for B. cereus G9241 and B. anthracis Sterne 34F2 that share >90% protein identity

(0.12 MB PDF)

Table S3.

Hypothetical and Unknown Function Chromosomal genes with increased expression (≥6-fold) in CO2+0.8% bicarbonate

(0.11 MB PDF)

Table S4.

Hypothetical and Unknown Function Chromosomal genes with increased expression (≥6-fold) in O2

(0.10 MB PDF)

Table S5.

Genes with increased expression in B. cereus G9241 in CO2 (MGM+0.8% bicarbonate)

(0.27 MB PDF)

Table S6.

Genes with increased expression in B. cereus G9241 in MGM in O2

(0.30 MB PDF)

Table S7.

Genes with increased expression in B. anthracis Sterne (34F2) in CO2 (MGM+0.8% bicarbonate)

(0.17 MB PDF)

Table S8.

Genes with increased expression in B. anthracis Sterne (34F2) in MGM in O2

(0.16 MB PDF)

Table S9.

Genes with increased expression in B. cereus 10987 in CO2 (MGM+0.8% bicarbonate)

(0.12 MB PDF)

Table S10.

Genes with increased expression in B. cereus 10987 in MGM in O2

(0.10 MB PDF)

Table S11.

Primers for SYBR-Green qRT-PCR for B. anthracis Sterne 34F2

(0.08 MB PDF)

Table S12.

Primers for SYBR-Green qRT-PCR for B. cereus G9241

(0.08 MB PDF)

Table S13.

Putative motility genes more highly expressed in O2 for B. cereus G9241 and B. anthracis Sterne (34F2)

(0.08 MB PDF)

Table S14.

Putative S-Layer genes differentially expressed in CO2 and O2 in three Bacilli.

(0.08 MB PDF)


We thank Dr. Alex Hoffmaster for supplying B. cereus G9241 for this study. We thank Timothy Read and David Rasko for helpful information regarding genome annotations.

Author Contributions

Conceived and designed the experiments: KDP NB. Performed the experiments: KDP AV BB. Analyzed the data: KDP AV NB. Wrote the paper: KDP NB.


  1. 1. Passalacqua KD, Bergman NH (2006) Bacillus anthracis: interactions with the host and establishment of inhalational anthrax. Future Microbiol 1: 397–415.
  2. 2. Rasko DA, Altherr MR, Han CS, Ravel J (2005) Genomics of the Bacillus cereus group of organisms. FEMS Microbiol Rev 29: 303–329.
  3. 3. Jensen GB, Hansen BM, Eilenberg J, Mahillon J (2003) The hidden lifestyles of Bacillus cereus and relatives. Environ Microbiol 5: 631–640.
  4. 4. Miller JM, Hair JG, Hebert M, Hebert L, Roberts FJ Jr, et al. (1997) Fulminating bacteremia and pneumonia due to Bacillus cereus. J Clin Microbiol 35: 504–507.
  5. 5. Hoffmaster AR, Hill KK, Gee JE, Marston CK, De BK, et al. (2006) Characterization of Bacillus cereus isolates associated with fatal pneumonias: strains are closely related to Bacillus anthracis and harbor B. anthracis virulence genes. J Clin Microbiol 44: 3352–3360.
  6. 6. Hoffmaster AR, Ravel J, Rasko DA, Chapman GD, Chute MD, et al. (2004) Identification of anthrax toxin genes in a Bacillus cereus associated with an illness resembling inhalation anthrax. Proc Natl Acad Sci U S A 101: 8449–8454.
  7. 7. Zwick ME, Kiley MP, Stewart AC, Mateczun A, Read TD (2008) Genotyping of Bacillus cereus strains by microarray-based resequencing. PLoS ONE 3: e2513.
  8. 8. Bergman NH, Anderson EC, Swenson EE, Janes BK, Fisher N, et al. (2007) Transcriptional profiling of Bacillus anthracis during infection of host macrophages. Infect Immun 75: 3434–3444.
  9. 9. Bergman NH, Anderson EC, Swenson EE, Niemeyer MM, Miyoshi AD, et al. (2006) Transcriptional profiling of the Bacillus anthracis life cycle in vitro and an implied model for regulation of spore formation. J Bacteriol 188: 6092–6100.
  10. 10. Passalacqua KD, Bergman NH, Lee JY, Sherman DH, Hanna PC (2007) The global transcriptional responses of Bacillus anthracis Sterne (34F2) and a Delta sodA1 mutant to paraquat reveal metal ion homeostasis imbalances during endogenous superoxide stress. J Bacteriol 189: 3996–4013.
  11. 11. Bourgogne A, Drysdale M, Hilsenbeck SG, Peterson SN, Koehler TM (2003) Global effects of virulence gene regulators in a Bacillus anthracis strain with both virulence plasmids. Infect Immun 71: 2736–2743.
  12. 12. Sirard JC, Mock M, Fouet A (1995) Molecular tools for the study of transcriptional regulation in Bacillus anthracis. Res Microbiol 146: 729–737.
  13. 13. Sirard JC, Mock M, Fouet A (1994) The three Bacillus anthracis toxin genes are coordinately regulated by bicarbonate and temperature. J Bacteriol 176: 5188–5192.
  14. 14. Koehler TM, Dai Z, Kaufman-Yarbray M (1994) Regulation of the Bacillus anthracis protective antigen gene: CO2 and a trans-acting element activate transcription from one of two promoters. J Bacteriol 176: 586–595.
  15. 15. Casey JR (2006) Why bicarbonate? Biochem Cell Biol 84: 930–939.
  16. 16. Wilson AC, Soyer M, Hoch JA, Perego M (2008) The bicarbonate transporter is essential for Bacillus anthracis lethality. PLoS Pathog 4: e1000210.
  17. 17. Yang J, Hart E, Tauschek M, Price GD, Hartland EL, et al. (2008) Bicarbonate-mediated transcriptional activation of divergent operons by the virulence regulatory protein, RegA, from Citrobacter rodentium. Mol Microbiol 68: 314–327.
  18. 18. Read TD, Peterson SN, Tourasse N, Baillie LW, Paulsen IT, et al. (2003) The genome sequence of Bacillus anthracis Ames and comparison to closely related bacteria. Nature 423: 81–86.
  19. 19. Ravel J, Jiang L, Stanley ST, Wilson MR, Decker RS, et al. (2008) The complete genome sequence of Bacillus anthracis Ames “Ancestor”. J Bacteriol.
  20. 20. Rasko DA, Ravel J, Okstad OA, Helgason E, Cer RZ, et al. (2004) The genome sequence of Bacillus cereus ATCC 10987 reveals metabolic adaptations and a large plasmid related to Bacillus anthracis pXO1. Nucleic Acids Res 32: 977–988.
  21. 21. Rasko DA, Rosovitz MJ, Okstad OA, Fouts DE, Jiang L, et al. (2007) Complete sequence analysis of novel plasmids from emetic and periodontal Bacillus cereus isolates reveals a common evolutionary history among the B. cereus-group plasmids, including Bacillus anthracis pXO1. J Bacteriol 189: 52–64.
  22. 22. Sue D, Hoffmaster AR, Popovic T, Wilkins PP (2006) Capsule production in Bacillus cereus strains associated with severe pneumonia. J Clin Microbiol 44: 3426–3428.
  23. 23. Green BD, Battisti L, Koehler TM, Thorne CB, Ivins BE (1985) Demonstration of a capsule plasmid in Bacillus anthracis. Infect Immun 49: 291–297.
  24. 24. Drysdale M, Bourgogne A, Hilsenbeck SG, Koehler TM (2004) atxA controls Bacillus anthracis capsule synthesis via acpA and a newly discovered regulator, acpB. J Bacteriol 186: 307–315.
  25. 25. Rehren G, Walters S, Fontan P, Smith I, Zarraga AM (2007) Differential gene expression between Mycobacterium bovis and Mycobacterium tuberculosis. Tuberculosis (Edinb) 87: 347–359.
  26. 26. Winstedt L, Yoshida K, Fujita Y, von Wachenfeldt C (1998) Cytochrome bd biosynthesis in Bacillus subtilis: characterization of the cydABCD operon. J Bacteriol 180: 6571–6580.
  27. 27. Ruzicka FJ, Lieder KW, Frey PA (2000) Lysine 2,3-aminomutase from Clostridium subterminale SB4: mass spectral characterization of cyanogen bromide-treated peptides and cloning, sequencing, and expression of the gene kamA in Escherichia coli. J Bacteriol 182: 469–476.
  28. 28. Eliasson R, Reichard P, Mulliez E, Ollagnier S, Fontecave M, et al. (1995) The mechanism of the anaerobic Escherichia coli ribonucleotide reductase investigated with nuclear magnetic resonance spectroscopy. Biochem Biophys Res Commun 214: 28–35.
  29. 29. Ollagnier S, Mulliez E, Gaillard J, Eliasson R, Fontecave M, et al. (1996) The anaerobic Escherichia coli ribonucleotide reductase. Subunit structure and iron sulfur center. J Biol Chem 271: 9410–9416.
  30. 30. Ollagnier S, Mulliez E, Schmidt PP, Eliasson R, Gaillard J, et al. (1997) Activation of the anaerobic ribonucleotide reductase from Escherichia coli. The essential role of the iron-sulfur center for S-adenosylmethionine reduction. J Biol Chem 272: 24216–24223.
  31. 31. Sun X, Ollagnier S, Schmidt PP, Atta M, Mulliez E, et al. (1996) The free radical of the anaerobic ribonucleotide reductase from Escherichia coli is at glycine 681. J Biol Chem 271: 6827–6831.
  32. 32. Torrents E, Sahlin M, Biglino D, Graslund A, Sjoberg BM (2005) Efficient growth inhibition of Bacillus anthracis by knocking out the ribonucleotide reductase tyrosyl radical. Proc Natl Acad Sci U S A 102: 17946–17951.
  33. 33. Sterne M, Proom H (1957) Induction of motility and capsulation in Bacillus anthracis. J Bacteriol 74: 541–542.
  34. 34. Sleytr UB, Beveridge TJ (1999) Bacterial S-layers. Trends Microbiol 7: 253–260.
  35. 35. Mesnage S, Tosi-Couture E, Mock M, Gounon P, Fouet A (1997) Molecular characterization of the Bacillus anthracis main S-layer component: evidence that it is the major cell-associated antigen. Mol Microbiol 23: 1147–1155.
  36. 36. Kern JW, Schneewind O (2008) BslA, a pXO1-encoded adhesin of Bacillus anthracis. Mol Microbiol 68: 504–515.
  37. 37. Hoffmaster AR, Koehler TM (1997) The anthrax toxin activator gene atxA is associated with CO2-enhanced non-toxin gene expression in Bacillus anthracis. Infect Immun 65: 3091–3099.
  38. 38. Hoffmaster AR, Koehler TM (1999) Control of virulence gene expression in Bacillus anthracis. J Appl Microbiol 87: 279–281.
  39. 39. Mignot T, Couture-Tosi E, Mesnage S, Mock M, Fouet A (2004) In vivo Bacillus anthracis gene expression requires PagR as an intermediate effector of the AtxA signalling cascade. Int J Med Microbiol 293: 619–624.
  40. 40. Rowland BM, Grossman TH, Osburne MS, Taber HW (1996) Sequence and genetic organization of a Bacillus subtilis operon encoding 2,3-dihydroxybenzoate biosynthetic enzymes. Gene 178: 119–123.
  41. 41. Cendrowski S, MacArthur W, Hanna P (2004) Bacillus anthracis requires siderophore biosynthesis for growth in macrophages and mouse virulence. Mol Microbiol 51: 407–417.
  42. 42. Bsat N, Helmann JD (1999) Interaction of Bacillus subtilis Fur (ferric uptake repressor) with the dhb operator in vitro and in vivo. J Bacteriol 181: 4299–4307.
  43. 43. Baichoo N, Helmann JD (2002) Recognition of DNA by Fur: a reinterpretation of the Fur box consensus sequence. J Bacteriol 184: 5826–5832.
  44. 44. Wilson MD, Barbosa-Morais NL, Schmidt D, Conboy CM, Vanes L, et al. (2008) Species-specific transcription in mice carrying human chromosome 21. Science 322: 434–438.
  45. 45. Ehling-Schulz M, Guinebretiere MH, Monthan A, Berge O, Fricker M, et al. (2006) Toxin gene profiling of enterotoxic and emetic Bacillus cereus. FEMS Microbiol Lett 260: 232–240.
  46. 46. Gaviria Rivera A, Granum P, Priest F (2000) Common occurrence of enterotoxin genes and enterotoxicity in Bacillus thuringiensis. FEMS Microbiology Letters 190: 151–155.
  47. 47. Ngamwongsatit P, Buasri W, Pianariyanon P, Pulsrikarn C, Ohba M, et al. (2008) Broad distribution of enterotoxin genes (hblCDA, nheABC, cytK, and entFM) among Bacillus thuringiensis and Bacilus cereus as shown by novel primers. International Journal of Food Microbiology 121: 352–356.
  48. 48. Pruss B, Dietrich R, Nibler B, Martblauer E, Scherer S (1999) The hemolytic enterotoxin HBL is broadly distributed among species of the Bacillus cereus group. Applied and Environmental Microbiology 65: 5436–5442.
  49. 49. Gohar M, Faegri K, Perchat S, Ravnum S, Okstad OA, et al. (2008) The PlcR virulence regulon of Bacillus cereus. PLoS ONE 3: e2793.
  50. 50. Jaaskelainen EL, Haggblom MM, Andersson MA, Salkinoja-Salonen MS (2004) Atmospheric oxygen and other conditions affecting the production of cereulide by Bacillus cereus in food. Int J Food Microbiol 96: 75–83.
  51. 51. Duport C, Zigha A, Rosenfeld E, Schmitt P (2006) Control of enterotoxin gene expression in Bacillus cereus F4430/73 involves the redox-sensitive ResDE signal transduction system. J Bacteriol 188: 6640–6651.
  52. 52. Ouhib O, Clavel T, Schmitt P (2006) The production of Bacillus cereus enterotoxins is influenced by carbohydrate and growth rate. Curren Microbiology 53: 222–226.
  53. 53. Agaisse H, Gominet M, Okstad OA, Kolsto AB, Lereclus D (1999) PlcR is a pleiotropic regulator of extracellular virulence factor gene expression in Bacillus thuringiensis. Mol Microbiol 32: 1043–1053.
  54. 54. Mignot T, Mock M, Robichon D, Landier A, Lereclus D, et al. (2001) The incompatibility between the PlcR- and AtxA-controlled regulons may have selected a nonsense mutation in Bacillus anthracis. Mol Microbiol 42: 1189–1198.
  55. 55. Liu H, Bergman NH, Thomason B, Shallom S, Hazen A, et al. (2004) Formation and composition of the Bacillus anthracis endospore. J Bacteriol 186: 164–178.
  56. 56. Moat AG, Foster JW, Spector MP (2002) Microbial Physiology, 4th edition. New York: Wiley-Liss.
  57. 57. Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, et al. (2003) Summaries of Affymetrix GeneChip probe level data. Nucleic Acids Res 31: e15.
  58. 58. Hosack DA, Dennis G Jr, Sherman BT, Lane HC, Lempicki RA (2003) Identifying biological themes within lists of genes with EASE. Genome Biol 4: R70.