Transcriptional Profiling of Bacillus anthracis Sterne (34F2) during Iron Starvation

Lack of available iron is one of many environmental challenges that a bacterium encounters during infection and adaptation to iron starvation is important for the pathogen to efficiently replicate within the host. Here we define the transcriptional response of B. anthracis Sterne (34F2) to iron depleted conditions. Genome-wide transcript analysis showed that B. anthracis undergoes considerable changes in gene expression during growth in iron-depleted media, including the regulation of known and candidate virulence factors. Two genes encoding putative internalin proteins were chosen for further study. Deletion of either gene (GBAA0552 or GBAA1340) resulted in attenuation in a murine model of infection. This attenuation was amplified in a double mutant strain. These data define the transcriptional changes induced during growth in low iron conditions and illustrate the potential of this dataset in the identification of putative virulence determinants for future study.


Introduction
Bacillus anthracis, the causative agent of anthrax, is a highly virulent pathogen that has been used in recent history as a biological weapon [1]. In the environment, B. anthracis exists primarily as an infectious spore, which can remain dormant for many years [2]. Upon infection of a mammalian host, the spores are believed to associate with regional phagocytes, rapidly germinate and begin to divide [3,4]. Following these early events, vegetative B. anthracis escape the phagocytes and replicate in the lymphatic system prior to entering the bloodstream [5]. Titers can reach as high as 10 7 or 10 8 bacteria per milliliter of blood. This high bacterial load can lead to septicemia, toxemia, and eventually the death of the host [5]. The success of B. anthracis as a pathogen is at least partly due to the development of mechanisms allowing the rapid transition from environmental dormancy to active replication within the host. A better understanding of the mechanisms involved in this adaptation could aid in the development of improved treatment options and vaccines.
Pathogens encounter a variety of signals during infection of a mammalian host. Our laboratory has previously shown that the macrophage environment in which B. anthracis exists during early infection induces significant transcriptional changes in the bacterium, presumably altering the protein profiles of the organism for optimal growth in this environment [6]. Earlier studies have also examined the response of this pathogen to oxidative stress, a specific signal that the bacterium is likely to encounter during its time within the macrophage cytosol [7]. Also, another host-related signal, increased carbon dioxide levels is known to alter expression of the B. anthracis toxin genes as well as others [8,9,10].
Though there is a substantial body of work examining the response of B. anthracis to several signals encountered in the host, the response of the pathogen to iron starvation remains unstudied on a global scale. In the human host, the concentration of free iron available to the bacterium is limiting for growth [11]. Effective strategies for adaptation to this altered environmental condition and, subsequently, the acquisition of iron, are vital to the survival of most bacterial pathogens. Many pathogens undergo significant changes in their gene and protein expression to adapt to growth in iron limiting conditions, including Francisella tularensis, Mycobacterium tuberculosis, and Streptococcus pneumoniae [12,13,14].
Here we report the transcriptional response of B. anthracis grown in iron limiting conditions. When compared to bacteria that were grown in iron replete conditions, significant changes in gene expression were observed as early as 2 hours post-inoculation. Clear patterns of gene regulation were observed with significant changes in genes involved in a wide range of processes being seen. One set of differentially regulated genes, annotated in genome sequence databases as internalins, were chosen for further study. These data fill a significant gap in our current knowledge of B. anthracis pathogenesis and provide the basis for examination of specific mechanisms of iron acquisition and general pathogenesis of this organism.

Response of B. anthracis Sterne to iron starvation
In the study reported here, we sought to define the response of B. anthracis to low iron concentrations, a signal that mimics conditions encountered by the bacterium within a mammalian host. Growth characteristics of B. anthracis in iron depleted media (IDM) has been previously reported [15,16]. We first wanted to examine the growth kinetics of this pathogen in media containing high and low concentrations of iron to identify optimal timepoints for RNA isolation. Vegetative bacilli were inoculated into either IDM or iron replete media (IRM) at an initial OD 600 of 0.05 and growth was monitored hourly (Figure 1). Based on the growth characteristics of the bacteria in these media, hourly intervals from 2 to 4 hours were chosen as time points for RNA isolations (Figure 1-arrows). The growth rate of B. anthracis in IDM slowed noticeably beyond the four hour timepoint, with entrance into stationary phase apparent by five hours. Because of the growth phase difference between the two conditions (late log for IRM and early stationary for IDM) the five hour timepoint was not included in the microarray comparison.
In order to gain a thorough understanding of the response of B. anthracis to iron starvation, we utilized DNA microarrays to examine differences in transcriptional profiles following growth in either IDM or IRM. RNA was isolated from B. anthracis following 2, 3, or 4 hours of growth in either IDM or IRM. Each of the RNA samples (four replicates were performed for each timepoint) was then subjected to microarray analysis using a custom B. anthracis Affymetrix GeneChip [6,7,17]. Statistically significant differences in gene expression between B. anthracis grown in high (IRM) or low (IDM) iron concentrations were defined as a J5 score greater than two as well as a fold change greater than two (IDM vs. IRM). The J5 statistical test was chosen for its stringency when looking at a data set with a low n, limiting the chance of picking up false positives [18]. Statistical comparisons were performed at each of the three timepoints independently. At two hours the differences observed were minimal, with only 41 genes exhibiting statistically significant changes based on the above criteria. Beyond two hours, more significant changes were observed, presumably as the trace amounts of iron remaining in the IDM were consumed by the bacteria. At the three and four hour timepoints, significant changes were observed in the expression levels of 124 and 282 genes, respectively. The majority of the genes that exhibited significant changes in expression level at two hours were also observed to be significant at both three and four hours (87.8%). In addition, 86.3% of the genes identified as significant at three hours were significantly altered at the four hour timepoint as well.
Those genes exhibiting significantly altered expression following four hours of growth in IDM were put into Gene Pattern [19] for hierarchical clustering (Figure 2). Figure 2A shows a clustering of log 2 transformed normalized intensity values of all gene found to be significantly changed at four hours. Data for the entire timecourse in IDM are shown for these genes to illustrate the observed changes over time in this medium. The clustered data demonstrate a clear pattern of transcriptional regulation over time. Cluster A represents a small subset of genes that appear to peak at three hours, although their expression remains significantly higher at four hours (compared to IRM). Cluster B represent genes that peak at three hours and remain highly expressed through the end of the experiment while cluster C represents genes with a peak in expression at four hours. Finally, cluster D consists of those genes exhibiting a decrease in expression during iron starvation.
We next sought to compare the gene expression profiles between IDM and IRM at this single timepoint ( Figure 2B). Here the data cluster into two distinct groups, with 136 genes upregulated and 145 genes down-regulated. Since the majority of the genes differentially regulated at earlier timepoints were also identified at four hours, all subsequent analyses will focus on the genes identified at the final timepoint. The top 50 genes that were up-regulated after four hours in IDM are listed in Table 1. The complete list of significant genes identified at four hours, along with those for the other two timepoints, can be found in the supplementary information (Table S1-up-regulated genes Table  S2 -down-regulated genes).

Validation of microarray results using quantitative real time PCR
In order to confirm the microarray results, ten genes were chosen (five up-regulated and five down-regulated) for analysis using quantitative real time PCR (Q-PCR). Q-PCR experiments were carried out on RNA isolated following four hours of growth in IDM or IRM. Among the up-regulated genes tested were GBAA4786, GBAA2368, GBAA1393, GBAA3328, and GBAA1086. Down-regulated genes used to confirm the microarray analysis included GBAA2145, GBAA4027, GBAA5481, GBAA0410, and GBAA2609. When measured by Q-PCR, each of these genes exhibited a pattern of expression nearly equal to what was observed on the microarrays ( Figure 3A white bars = microarray data; black bars = Q-PCR). A correlation plot comparing microarray data to those generated using Q-PCR demonstrated a strong positive association between these two data sets ( Figure 2B). Taken together, these data validate the microarray results, solidifying the findings reported above.

Analysis of genes based on COG classification
In order to more thoroughly examine these results, the uniquely regulated genes were grouped based on the annotated COGs (Clusters of Orthologous Groups). Related COGs were grouped together into the categories shown in Figure 4. The pie charts show the categorical designation of the unique genes that are induced or repressed after four hours in IDM ( Figure 4). Each chart represents 100% of the COG designations for the genes that were induced or repressed as discussed above. Each slice within a given chart represents the percentage of genes identified from a given category. By examining these general relationships between genes regulated by iron starvation, many interesting observations can be made.
Immediately, it was clear that the most striking changes, up-and down-regulated, were in genes involved in transport and metabo- anthracis vegetative cells were used to inoculate either IRM (black lines) or IDM (gray lines) at an initial OD 600 = 0.05. Growth was monitored over time by measuring change in OD 600 . RNA was harvested at two, three, and four hours after inoculation (arrows). doi:10.1371/journal.pone.0006988.g001 lism, with this category comprising nearly 75% of each pie ( Figure 4, dark blue slice). Because each of these slices was identical, it was difficult to determine what changes were occurring in this group. To examine this more closely, we looked at each of the COGs that was used to make up the 'transport and metabolism' group in Figure 4. Figure 5 shows the pie charts with this more specific categorical representation of the COG designations for the metabolic genes either induced (76) or repressed (96) in IDM. Here we can see clear differences in the genes involved in these transport and metabolic pathways following growth under iron depeleted conditions. The COG 'inorganic ion transport and metabolism' (purple slice) represents 42% of the total COG designations for genes upregulated in IDM. The genes up-regulated in this category encode orthologs of proteins with annotated roles in iron acquisition, while those being repressed during iron starvation are genes involved in transport of other ions including phosphate and nitrate. In fact, among the genes most up-regulated in IDM was the B. anthracis isd locus (GBAA4782-4789), which is known to be regulated by environmental iron concentration through the presence of known ferric regulatory regions (FUR boxes) [20]. This locus encodes proteins involved in iron binding, uptake, and release [20]. Work with the virulent Vollum strain of B. anthracis has shown that three genes from this locus (GBAA4787-4789) are not required for virulence in the guinea pig model of infection [21]. However, the fact that other members of this operon have been shown to play a role in iron acquisition from heme, indicates that some details about the role of the isd locus in B. anthracis virulence may remain to be elucidated [22].
Components of another other iron uptake system of specific interest, encoded by GBAA5328-5330, were induced by iron starvation. It has recently been shown that a B. cereus homolog of the protein encoded by GBAA5330 is able to bind to the siderophore petrobactin [23]. Similar binding properties were observed for another iron binding protein, GBAA4766, the expression of which was also up-regulated during growth in IDM [23]. Synthesis of petrobactin is vital to B. anthracis pathogenesis and mutants unable to produce this molecule are significantly impaired for growth in IDM and attenuated in a mouse model of infection [15,24]. This is in sharp contrast to mutants unable to synthesize bacillibactin, the other siderophore produced by B. anthracis, which grows well in IDM and shows no significant attenuation in a mouse model of infection [15]. Unexpectedly, the genes that comprise the petrobactin biosynthesis pathway were not significantly up-regulated during growth in IDM. In light of these data, it is not surprising that expression of the uptake systems for this siderophore would be induced, potentially enabling the bacterium to acquire more ironcontaining petrobactin. In fact, we have shown that one of these genes is important in the pathogenesis of B. anthracis and is required for growth in IDM, as well as for full murine virulence (Carlson et. al. manuscript submitted).
Another group of metabolic genes that were induced during iron starvation were the flavodoxins (GBAA1394 and GBAA3596). These genes, which are part of the 'energy production and conversion' category ( Figure 5, red slices), appear to be co-regulated with adjacent hypothetical proteins. Flavodox-  ins are involved in a variety of reactions in bacteria and are essential in many bacterial species [25]. These proteins have previously been shown to replace ferrodoxins in NADP + and N 2 reduction reactions during growth in low iron conditions [25]. Not surprisingly, our microarray data show that B. anthracis increases the expression of flavodoxins and decreases expression of ferrodoxins during growth in iron depleted conditions (Tables S1  and S2). Interestingly, flavodoxins have previously been associated with B. anthracis infection, being up-regulated following growth in macrophages [6]. Our data, along with those previously published, provide increasing evidence of a role for flavodoxins during B. anthracis infection of mammalian hosts. The specific role of these proteins is currently under investigation. Another category of specific interest, which may be related to the metabolic changes observed, includes genes involved in the control of transcription and translation (Figure 4, red slices). Altered regulation of a single transcription factor in response to an environmental signal could lead to significant changes in the transcriptome of the organism. Although only few members of this category exhibit altered regulation during iron starvation (8 upregulated genes and 5 down-regulated), the downstream effects of these proteins could be significant. Among the transcription factors differentially regulated during iron starvation are members of the AraC, LacI, and GntR families of transcriptional regulators. Proteins from each of these families play a role in carbon catabolite repression, a process known to be regulated by the presence or absence of environmental carbon sources [26,27,28]. Although the presence of genes thought to play a role in CCR is interesting, the regulation of these genes in this particular system (iron starvation) seems unusual. For the experiments presented here, the carbon composition of the two media types compared here, IDM and IRM, is identical. The differential regulation of these genes may indicate either a novel mechanism of metabolic control or, more likely, a function for these particular transcription factors unrelated to CCR.
Taken together, the data on uptake systems, transcription factors, and general metabolism indicate a significant change in the global metabolism of B. anthracis growing under iron limiting conditions. Further evaluation of the metabolic and transport pathways induced in B. anthracis during iron starvation, and other host specific cues, may provide insight into the specific metabolic changes the bacterium requires for adaptation to life within a mammalian host. Further characterization of these changes may lead to the development of treatments that target specific aspects of B. anthracis metabolism within the host, thereby preventing the organism from thriving in the host environment. A significant portion of the genes differentially regulated in iron depleted conditions do not have a known function (Figure 4, gray slice). Approximately 40% of the B. anthracis Sterne genome does not have an annotated function, so it is expected that we would see a significant number of these genes being differentially regulated under most conditions. Searching through this list, however, should provide some interesting candidates for future study. Many of the hypothetical proteins encoded by the B. anthracis genome are conserved across bacterial species; however there are genes which are specific to members of the Bacillus genus. Many genes induced by low iron levels, including GBAA1395, GBAA1458, and GBAA2286 encode proteins that do not share significant homology with proteins outside of the pathogenic Bacillus species. Genes such as GBAA1092 and the plasmid encoded GBAA_pXO1_0120 are of particular interest as they have no significant homology outside of B. anthracis. Further evaluation of these genes, specifically those up-regulated in conditions that mimic a host environment (e.g. iron starvation), may facilitate the identification of novel findings regarding mechanisms of B. anthracis pathogenesis.

Attenuation of internalin deficient B. anthracis
Two other genes of unknown function that were induced during iron starvation, GBAA0552 and GBAA1346. These genes have been annotated as putative internalins based on the presence of leucine rich repeat (LRR) domains within the predicted protein sequence [29]. These LRR domains are found in all Listeria internalins and are known to facilitate the interaction with their respective host-cell receptors [29]. The protein encoded by GBAA0552 also contains a NEAT domain, which has been associated with heme binding properties in other B. anthracis proteins [22]. The protein encoded by GBAA1346 lacks this domain, but contains three SLH domains, commonly found in protein that are anchored to the S-layer of the cell wall [30]. Among Bacillus species, these genes are only found within the pathogenic members, though their potential role in virulence remains to be elucidated [31]. To examine the potential role of these proteins in B. anthracis growth during iron starvation, we created non-polar deletion mutants lacking either one or both of these putative internalins. Curiously, none of these mutants exhibited a reduced ability to grow in IDM (data not shown).
Homologs of these putative internalin genes were identified as candidate B. cereus virulence factors due to their expression during infection of insects [32]. Although the function of these proteins in B. anthracis is currently unknown, internalins play an important role in the virulence of other pathogens, though their mechanisms of action remain unclear [29,31,33,34]. In Listeria monocytogenes, some internalins are known to promote invasion into epithelial cells through the binding of E-cadherin expressed on the host cell surface [33,34]. Interestingly, increased expression of these genes in B. anthracis also was observed during oxidative stress, which is another condition the bacterium would encounter during infection of a mammalian host [7].
Although no differences in IDM growth were observed for these mutants, the combination of B. anthracis gene expression data and the fact that internalin-like proteins have been clearly linked to virulence in other pathogens led us to examine the role of these two genes in B. anthracis virulence. We sought to determine if B. anthracis strains lacking these putative internalins were attenuated for virulence in a mouse model of inhalational anthrax. Wild-type B. anthracis, D0552, D1346, and D0552/1346 strains were inoculated intratracheally into DBA/2J mice at various doses to determine the LD 50 of each strain. The single deletion strains both exhibited attenuation in this model, with the loss of either GBAA0552 or GBAA1346 resulting in an increase in LD 50 of approximately 110 and 60 times that of wild-type spores, respectively ( Table 2). The double mutant exhibited the highest degree of attenuation, with an LD 50 160 times that of wild-type B. anthracis. A representative survival curve for the double mutant is shown in Figure 6. This level of attenuation is comparable with what has been observed for internalin mutants in L. monocytogenes [35,36,37]. Further study will be required to fully elucidate the role of these proteins in the pathogenesis of B. anthracis.
Collectively, the data presented here show significant changes in the transcriptional profile of B. anthracis during growth in iron limiting conditions. Significant increases in iron transport mechanisms and genes involved in various aspects of metabolism indicate adaptation to growth in these conditions. These data fill a gap in our current knowledge about the pathogenesis of this organism and provide the basis for future work examining the role of these genes during B. anthracis infection, as well as the potential of these genes to be utilized as the target of novel therapeutics.

Bacterial cultivation
All of the work described in this manuscript was performed using the Sterne 34F 2 (pXO1 + , pXO2 2 ) strain of B. anthracis. Initial spore stocks were prepared as follows. A single B. anthracis culture was  grown overnight at 37uC, and then 1 ml was used to inoculate 500 ml of G medium as described previously [35]. For all iron depletion studies, iron depleted media (IDM) was used [15]. Growth curve and RNA isolation experiments were performed as follows. Spores were germinated in brain heart infusion (BHI) broth and incubated overnight at room temperature. The following day, vegetative bacilli were diluted 1:100 in fresh BHI and grown 1-2 hours. Early log-phase bacilli were then washed 3x with PBS and 3x with IDM to ensure removal of nutrients and potential iron sources carried over from the BHI. Washed bacteria were then used to inoculate cultures of IDM or iron replete media (IRM) [IDM+20 mM ferrous sulfate] at an OD 600 = 0.05.

RNA isolation and analysis
RNA isolations were performed as previously described [17]. Briefly, at indicated timepoints, cultures were passed over a 0.2 mM filter. Bacteria were resuspended from the filter with ice cold nuclease free water. Lysis buffer (2% SDS, 10 mM EDTA, 200 mM NaCl) was immediately added to the resuspended bacteria and the mixture was placed in a boiling water bath for three minutes. Nucleic acid was subsequently isolated through two hot phenol extractions (65uC) followed by phenol:chloroform and chloroform extractions (22uC). The isolated nucleic acids were then precipitated overnight at 220uC following the addition of 0.1 vol. ammonium acetate and 2.5 vol. ethanol. Resulting RNA was treated with DNase according to manufacturer's protocol (Turbo DNA-free, Ambion). DNA-free RNA was then precipitated using ammonium acetate/ethanol. RNA quantity was measured by spectrophotometry and quality was confirmed at each step using an Agilent Bioanalyzer according to manufacturer's protocol.

Microarray data analysis
Analysis of the microarray data was carried out using Gene Expression Data Analyzer (GEDA; http://bioinformatics.upmc. edu/GE2/GEDA.html) [18]. Briefly, this online software package was used for statistical comparisons of individual conditions at each timepoint. Data were grouped into categories, IDM or IRM, and subjected to the J5 statistical analysis, which is designed for data sets with limited numbers of replicates and reduces the chances of false positives [18]. Genes were considered statistically significant if they had a J5 score greater than 2 and exhibited a fold change greater than 2. Supplementary tables S1 and S2 report both the J5 value and fold change for genes that changed significantly at each timepoint. Genes meeting these requirements were then entered into GenePattern [19] and data are presented as a hierarchical clustering (Pearson correlation) of log 2 transformed normalized intensity values for these genes. Our datasets have been deposited to the ArrayExpress database under the accession number E-MEXP-2272.

Quantitative PCR
cDNA synthesis was performed using Superscript III (Invitrogen) and 1 mg of total RNA. Real time reactions were performed with a 1:500 final dilution of template cDNA. Primer sets were designed for genes indicated using Primer 3 [36] and reactions were carried out on an BioRad ICycler real time machine using SYBR Green labeling according to manufacturer'r protocol (BioRad). The acoB gene (GBAA2775) was used as the internal reference, as it was observed to have no change in expression on microarray data (data not shown). The Q-PCR data are presented as log 2 transformed fold change values (iron depleted/iron rich). Data are presented as the mean6SEM for 4 independent experiments.

Creation of internalin mutants
Mutants were generated by allelic exchange, using the method of Janes and Stibiwitz [37]. Each mutant allele was designed to replace codons 10 through 14 with the following sequence: TAA TAG TGA GGA TCC. This introduced three in-frame stop codons followed by the recognition site for the restriction endonuclease BamHI (to facilitate screening of the mutant allele). All other codons of the gene remain wild-type, therefore the mutant is predicted to be non-polar. The double mutant was generated by the introduction of the GBAA1346 mutation into the D0552 mutant by the same method.

Mouse infections
Intratracheal infections of DBA/2J mice (Jackson Laboratories) were performed as previously described [38]. Groups of eight mice were infected with either wild-type or mutant spores at a variety of doses ranging from 1.5610 5 through 1.5610 8 spores per mouse. Mice were monitored over a period of fourteen days. LD 50 values were calculated using the Moving Average Interpolation program available at (http://falkow.stanford.edu/whatwedo/software/software.html) [39]. All mouse infections performed in compliance with the National Institutes of Health Guidelines on the Care and Use of Animals and an animal study protocol approved by the University of Michigan's Institutional Animal Care and Use Committee.