Certhrax Is an Antivirulence Factor for the Anthrax-Like Organism Bacillus cereus Strain G9241

ABSTRACT

Bacillus cereus G9241 caused a life-threatening anthrax-like lung infection in a previously healthy human. This strain harbors two large virulence plasmids, pBCXO1 and pBC210, that are absent from typical B. cereus isolates. The pBCXO1 plasmid is nearly identical to pXO1 from Bacillus anthracis and carries genes (pagA1, lef, and cya) for anthrax toxin components (protective antigen [called PA1 in G9241], lethal factor [LF], and edema factor [EF], respectively). The plasmid also has an intact hyaluronic acid capsule locus. The pBC210 plasmid has a tetrasaccharide capsule locus, a gene for a PA1 homolog called PA2 (pagA2), and a gene (cer) for Certhrax, an ADP-ribosyltransferase toxin that inactivates vinculin. LF, EF, and Certhrax require PA for entry into cells. In this study, we asked what role PA1, PA2, LF, and Certhrax play in the pathogenicity of G9241. To answer this, we generated isogenic deletion mutations in the targeted toxin gene components and then assessed the strains for virulence in highly G9241-susceptible (A/J) and moderately G9241-sensitive (C57BL/6) mice. We found that full virulence of G9241 required PA1 and LF, while PA2 contributed minimally to pathogenesis of G9241 but could not functionally replace PA1 as a toxin-binding subunit in vivo. Surprisingly, we discovered that Certhrax attenuated the virulence of G9241; i.e., a Δcer Δlef mutant strain was more virulent than a Δlef mutant strain following subcutaneous inoculation of A/J mice. Moreover, the enzymatic activity of Certhrax contributed to this phenotype. We concluded that Certhrax acts as an antivirulence factor in the anthrax-like organism B. cereus G9241.

INTRODUCTION

Bacillus cereus is a spore-forming soil organism that is considered an opportunistic pathogen because it rarely causes severe infections except in immunocompromised individuals (1). Some strains of B. cereus that carry certain enterotoxin genes can cause foodborne gastrointestinal illnesses that are often acute in onset but of short duration. Thus, it was a surprise when in 1994 a B. cereus strain (subsequently named G9241) was isolated from the sputum and blood of a patient who sought care at a Louisiana hospital 2 days after onset of an illness that resembled inhalational anthrax (2). He was a welder by occupation and had no underlying health issues. The similarity of the patient's illness to inhalational anthrax was subsequently explained when Hoffmaster et al. reported that G9241 harbors two large plasmids, called pBCXO1 and pBC210 (originally called pCB218); pBCXO1 is nearly identical to the large virulence plasmid, pXO1, of Bacillus anthracis. The pBCXO1 plasmid encodes the anthrax toxin components lethal factor (LF), edema factor (EF), and protective antigen (PA), as well as an operon involved in the synthesis of a hyaluronic acid capsule. Such an operon is also present on the pXO1 plasmid of B. anthracis, but the capsule is not made due to the presence of a stop codon in the first gene of the operon (3). The pBC210 plasmid of G9241 has a second capsule synthesis locus for a tetrasaccharide capsule that is distinct from the poly-γ-d-glutamic acid (PDGA) capsule encoded on the B. anthracis plasmid pXO2 (2, 4). Moreover, the pBC210 plasmid carries two other loci for unique products of potential pathogenic significance for G9241. These are PA2, a homolog of PA (called PA1 in G9241) that shares 60% amino acid identity (2), and Certhrax, a novel ADP-ribosyltransferase toxin (ADPRT) that posttranslationally modifies and inactivates human vinculin (5).

Other B. cereus strains that caused anthrax-like disease in immunocompetent people have subsequently been identified. Five of the cases were in metal workers from the southern regions of the United States who ultimately succumbed to their illnesses (6–8). The B. cereus strains isolated from these individuals also express a capsule and/or genes for the anthrax toxin components. One of the strains appears to be indistinguishable (based on PCRs) from G9241 and carries pBCXO1- and pBC210-like plasmids (7). Another isolate of B. cereus, strain BcFL2013, caused a cutaneous anthrax-like illness in Florida and carries an incomplete pBCXO1-like plasmid and a partial version of pBC210 (9). Finally, B. cereus isolates from great apes that died of anthrax-like illness in several African countries are now called biovar anthracis. Indeed, these African strains were recently placed on the Health and Human Services select agent list because they express the PDGA capsule of B. anthracis and the anthrax toxins (10). Since B. cereus is often considered a contaminant when isolated in hospital samples, it is important to be aware of the emergence of such strains with increased pathogenic potential.

The virulence of B. anthracis primarily reflects production of lethal toxin (LT), edema toxin (ET), and the antiphagocytic PDGA capsule (11). The toxins are comprised of an enzymatically active component (LF or EF) and a cell-binding component, PA. Neither component is toxic alone (12). LF is a zinc metalloprotease that cleaves mitogen-activated protein kinase kinases (MAPKKs) and disrupts numerous signaling pathways. EF is a calmodulin-dependent adenylyl cyclase that catalyzes the production of cyclic AMP (cAMP) and leads to edema (13). PA is an 83-kDa protein that binds to the anthrax toxin receptors tumor endothelial marker 8 (TEM8; also called ATXR1) and capillary morphogenesis protein 2 (CMG2; also called ATXR2) on the surface of eukaryotic cells (11). After PA₈₃ binds to an anthrax receptor, it gets cleaved by the protease furin into a 63-kDa monomer (PA₆₃) that then forms a heptamer or an octamer (14, 15). This oligomerization step exposes sites on PA₆₃ that permit up to three EF or LF molecules to bind at a time (16). The complex is then taken up by the cell through clathrin-mediated endocytosis in which, in the presence of an acidic environment, the PA oligomer undergoes a conformation change to form a pore in the membrane of the endosome that subsequently allows for unfolded LF and EF to translocate to the cytosol of the cell (17).

Although B. cereus strain G9241 caused a severe illness in a person and can produce anthrax toxins, it does not behave as does the fully virulent B. anthracis Ames strain in certain animal models (18). This apparent reduced virulence of G9241 could be attributed to the lack of the PDGA capsule. However, with the production of anthrax toxins, two distinct capsules, and PA2 and Certhrax, the pathogen does have an extensive repertoire of putative virulence factors, some of which we already know to contribute to pathogenicity in mice. For example, in single intraperitoneal (i.p.) challenge dose studies with C57BL/6 mice, the hyaluronic acid capsule and PA1 were required for full virulence of G9241, while PA2 did not appear to contribute to pathogenicity at an inoculum of 10⁵ spores (19, 20). In this study, we sought to better characterize the role of toxins in pathogenesis of G9241. In particular, we asked whether Certhrax and PA2 contribute to the virulence of the bacterium. A transcriptome analyses study previously showed that both genes were transcribed to higher levels in the presence of growth conditions that induce expression of known virulence factors in B. anthracis (21). Furthermore, structural analyses showed that similar to LF and EF, Certhrax has an N-terminal PA-binding domain and requires PA for translocation into host cells to cause toxicity (22). Therefore, we constructed deletion strains of these toxin components, as well as PA1 and LF, and characterized their pathogenicity in two mouse models. We found that PA1 and LF were essential for full virulence of G9241 after infection of both sensitive (A/J) and moderately resistant (C57BL/6) mice and that PA2 had a slight impact on pathogenicity of G9241 but could not replace PA1 as a major toxin binding subunit. We also showed that the genes for PA2 (pagA2) and Certhrax (cer) are in an operon but that each also has its own promoter. Lastly and most surprisingly, we discovered that Certhrax attenuated virulence of G9241 in A/J mice after subcutaneous inoculation and that this effect may be attributed to the enzymatic activity of the toxin. To our knowledge, this is the first report of an ADPRT behaving as an antivirulence factor.

RESULTS

PA2 minimally contributes to virulence of B. cereus G9241.

To assess the role of PA2 in pathogenesis of G9241, an isogenic deletion mutant strain here referred to as the ΔpagA2 mutant, was generated through allelic exchange and the absence of the expressed protein was confirmed by Western blot analyses with an anti-PA2 antibody (see Fig. S1A in the supplemental material). Six-week-old female A/J and C57BL/6 mice were then inoculated with the ΔpagA2 spores through the subcutaneous (s.c.) and intranasal (i.n.) routes, and the 50% lethal dose (LD₅₀) was calculated. As shown in Table 1, there was no significant difference in lethality in A/J mice inoculated through either route with the ΔpagA2 spores and wild-type (WT) G9241. The LD₅₀ values were calculated to be 20 and 1 × 10⁴ spores for the s.c. and i.n. experiments, respectively. There was also no difference in the median times to death for mice that succumbed to infection following challenge with the ΔpagA2 spores and mice challenged with WT G9241 (data not shown). However, the calculated percent survival in A/J mice inoculated s.c. with the lowest (10¹) dose of ΔpagA2 spores was 60%, which is statistically higher than the 13.3% survival rate in mice inoculated with the WT strain (P = 0.0280 via Fisher's exact test). In the C57BL/6 mouse model, there was no difference in lethality in mice inoculated via the i.n. route, where the LD₅₀ was 6 × 10⁵ spores for the ΔpagA2 strain and 4 × 10⁵ spores for the WT. However, there was a statistically significant attenuation in virulence in mice following s.c. challenge with the ΔpagA2 spores (LD₅₀ of 200 spores for the ΔpagA2 mutant, compared to 40 spores for the WT [Table 1]). Moreover, similar to what was seen in A/J mice, there was a statistical difference in the survival rates at the lowest inoculum dose, where 80% of C57BL/6 mice survived inoculation with 10¹ of ΔpagA2 spores and only 27% survived WT G9241 at the same dose (P = 0.0154). For the mice that did succumb to the infection, there was no difference in the actual times to death between ΔpagA2 and WT G9241 challenge (data not shown). In vivo production of PA2 was subsequently confirmed by the presence of anti-PA2 antibodies in mouse sera following G9241 infection (Fig. S1B). These results suggest that PA2 does contribute to virulence of B. cereus G9241 but in a limited capacity, as the effect can be seen only at the lowest spore doses inoculated. The data also implicate PA1 as the more vital toxin binding subunit in G9241.

TABLE 1.

LD₅₀ spore values of WT G9241 and mutant strains following inoculation of A/J or C57BL/6 mice by the subcutaneous or intranasal route

Genotype	LD50
	A/J		C57BL/6
	s.c.	i.n.	s.c.	i.n.
WT	10a	3 × 103a	40a (16–100)	4 × 105a (2 × 105–9 × 105)
ΔpagA2	20a	1 × 104a	200a,c (70–500)	6 × 105a (2 × 105–3 × 106)
ΔpagA1	2 × 104b	2 × 105a,c (6 × 104–5 × 105)	>3 × 107	2 × 106b
ΔpagA1 ΔpagA2	6 × 104b	3 × 105a,c (1 × 105–7 × 105)	NA	NA
pagA1::pagA2 ΔpagA1	3 × 105b	NAe	NA	NA
Δcer	<20	5 × 102a (4 × 101–4 × 103)	70a (20–200)	3 × 105a (1 × 105–8 × 105)
Δlef	7 × 105a,c (2 × 105–3 × 106)	2 × 105a,c	>2 × 107	5 × 106b
Δcer Δlef	1 × 103a,c,d (2 × 102–5 × 103)	2 × 105a,c (5 × 104–5 × 105)	>2 × 107	4 × 106b
cer.rest Δlef	5 × 106b	NA	NA	NA
inact.cer Δlef	2 × 104a,c,d (4 × 103–105)	NA	NA	NA

^aLD₅₀ values calculated via probit analyses, with 95% confidence intervals provided in parentheses.

^bLD₅₀ values calculated by the Reed and Muench method when the criteria for probit analysis were not met.

^cValues are significantly different from those for WT G9241 as determined by probit analysis (P < 0.05).

^dValue is significantly different from that for the Δlef G9241 strain as determined by probit analysis (P < 0.05).

^eNA, not applicable (the experiment was not conducted).

PA1 is essential for virulence of B. cereus G9241 during s.c. and i.n challenge of mice.

Since PA1 appears to be the required binding subunit involved in anthrax toxin translocation in G9241, we sought to determine whether expression of PA1 masked more substantive in vivo effects of PA2 in the two mouse models used in this study. To do so, an isogenic deletion mutant strain of pagA1 (ΔpagA1 mutant) and a double pagA1 and pagA2 deletion mutant strain (ΔpagA1 ΔpagA2 mutant) were generated. The ΔpagA1 ΔpagA2 mutant strain was constructed by deleting pagA2 in the ΔpagA1 strain background.

We first assessed the contribution of PA1 to virulence of G9241 during s.c. and i.n. infection of A/J and C57BL/6 mice through inoculation with ΔpagA1 mutant spores (Table 1). The mutant strain was avirulent in the s.c. inoculation model of C57BL/6 mice, and just a few mice died with the 10⁷ dose following i.n. inoculation. There was a significant attenuation in virulence of the ΔpagA1 mutant in the A/J mice as well compared to WT G9241. Approximately a 2-log attenuation was observed in A/J mice inoculated via the i.n. route. The LD₅₀ value for the s.c. route was calculated to be 2 × 10⁴ spores via the Reed and Muench method (23) since the data did not satisfy the criteria for probit analysis (i.e., there was not a single inoculum group in which 100% of the mice died). The results were consistent in two different experiments with 5 mice per group inoculated per dose.

Since the ΔpagA1 mutant strain was almost avirulent by both inoculation routes in the C57BL/6 mice, only the A/J mice were inoculated with the ΔpagA1 ΔpagA2 mutant spores (Table 1). The double knockout was as attenuated as the ΔpagA1 mutant strain in mice inoculated via the i.n. route, with a calculated LD₅₀ of 3 × 10⁵ spores. The double knockout was highly attenuated when given to A/J mice s.c., with an LD₅₀ value calculated to be 6 × 10⁴ spores. Thus, there was no apparent difference in virulence between the ΔpagA1 and ΔpagA1 ΔpagA2 mutant strains. Therefore, the data further support the dominant role of PA1 in the pathogenesis of G9241 and the minimal involvement of PA2.

PA2 cannot functionally replace PA1 as the major toxin-translocating component in B. cereus G9241.

Beyond the amino acid differences between PA1 and PA2, there are also differences in the upstream promoter regions of the two genes. Indeed, these regions are not homologous. Additionally, pagA1 is part of an operon with the autoregulator pagR, but there is no pagR homolog downstream of pagA2. Therefore, we sought to investigate whether PA2 could functionally replace PA1 if all other expression conditions were the same between the two. Therefore, we generated a pagA2 knock-in mutant strain, the pagA1::pagA2 ΔpagA1 strain, in which pagA2 was cloned into the pagA1 locus within the ΔpagA1 ΔpagA2 mutant strain background, as described in Materials and Methods. The in vitro expression of PA2 was confirmed via Western blot analyses (Fig. S1A).

Since the PA2 effect was observed only in the s.c. inoculation model, A/J mice were inoculated with a range of doses of pagA1::pagA2 ΔpagA1 spores and monitored for 14 days. The lethality pattern of the knock-in mutant strain was similar to that of the ΔpagA1 and ΔpagA1 ΔpagA2 G9241 strains, with the LD₅₀ value calculated to be 3 × 10⁵ spores (Table 1). These data suggest that PA2 cannot functionally replace PA1 to effectively translocate anthrax toxins to host cells.

cer and pagA2 are cotranscribed when B. cereus G9241 is expressed under toxin-inducing conditions, but each gene has its own promoter.

Passalacqua et al. demonstrated by transcriptome analyses that cer is expressed by B. cereus G9241 (21). However, the growth conditions used in that study are different from the toxin-inducing conditions used in this study. Therefore, we sought to establish that cer was indeed expressed when G9241 was grown in heart infusion (HI) broth supplemented with 8% sodium bicarbonate and 5% CO₂ at 37°C. Reverse transcriptase PCRs (RT-PCRs) were set up with WT G9241 cDNA and cer-specific primers to test for the presence of cer transcript (primer pair 1 in Fig. 1A). Additionally, pagA2-specific primers (primer pair 2 in Fig. 1A) were used as a positive control to confirm that the growth conditions were conducive to toxin component expression. As shown in Fig. 1B, both cer and pagA2 were expressed in all three samples of WT cDNA. No product was observed in any of the “no RT” samples, an observation that indicated that there was no DNA contamination in the isolated RNA. Additionally, with the use of primers that amplified the intergenic region between cer and pagA2 (primer pair 3 in Fig. 1A), we observed a transcript (Fig. 1B). Thus, we concluded that the two genes are, in fact, cotranscribed as an operon.

FIG 1.

The cer and pagA2 genes are cotranscribed by B. cereus G9241 during growth under toxin-inducing conditions. RNA was isolated from WT G9241 grown under toxin-inducing conditions. One microgram of RNA was used to make cDNA (+) and in a “no RT” control (−), where no reverse transcriptase was added. For the PCR, cer-specific primers (1), pagA2-specific primers (2), and primers that amplified the cer-pagA2 intergenic region (3) were used as depicted in the schematic (A). The resultant PCR products are shown in panel B. WT DNA and water were used as positive and negative controls, respectively. Data are representative of results from three biologically independent WT G9241 cultures.

To identify where the promoter for cer is located and to determine whether pagA2 has its own promoter despite being the downstream gene in the operon, we constructed several promoter fusions for each gene. Three different-size regions upstream of each gene were amplified and then cloned into the pHT304-18z vector (24), such that the fragments were inserted in front of a promoterless lacZ gene (Fig. S2A) with the goal of detecting β-galactosidase (β-Gal) activity. Each construct, as well as an empty vector control, was electroporated into wild-type G9241. For the cer constructs, the expression of lacZ was observed in the pHT304-18z::cerup2 and pHT304-18z::cerup3 G9241, but not in pHT304-18z::cerup1 (Fig. S2B). No β-Gal activity was observed in any of the empty vector controls. These results suggest that the cer promoter may be located within 267 bases upstream of the cer ATG and that the first 109 bp may be insufficient to promote β-Gal expression. For the pagA2 constructs, β-Gal activity was noted for all pagA2 promoter constructs but not in the empty vector control (Fig. S2C). Therefore, we surmised that pagA2 does in fact have its own promoter that may lie within 155 bp upstream of the pagA2 ATG.

Certhrax is a cell-associated protein in B. cereus G9241.

Beyond the above-mentioned transcription analyses, experiments were carried out to show expression of Certhrax protein in vitro. For that purpose, cultures were grown under toxin- and capsule-inducing conditions. Supernatants and cell lysates of the cultures were subjected to Western blot analyses and probed with purified anti-ADPr antibody. Based on the primary amino acid composition, the predicted size for Certhrax is 55 kDa. However, no immune-reactive protein was detected in G9241 culture supernatants or cell lysates (data not shown). Since the inability to detect Certhrax in the sample could be due to protein concentrations falling below the detection limits of the assay, cer was placed under the control of an isopropyl-β-d-1-thiogalactopyranoside (IPTG)-inducible promoter within the vector pUTE657 (25). The gene was cloned into pUTE657 with and without an N-terminal His tag so that an anti-His antibody could be used to detect the protein in conjunction with the anti-ADPr. The two constructs, pCer and pHisCer, were transformed into Escherichia coli Mach1 competent cells and were also electroporated into a cer deletion mutant strain of G9241 (Δcer mutant). In E. coli cell lysates, Certhrax and His-Certhrax were expressed at approximately 55 kDa and were detected in induced and uninduced cell lysates (Fig. 2A), although a smaller band of approximately 50 kDa was also detected in the induced cultures. The anti-His antibody seemed to detect only the higher-molecular-mass band and not the lower-molecular-mass protein (Fig. 2B). To detect expression in G9241, the Δcer pCer strain, WT G9241, the Δcer mutant, and the Δcer mutant carrying the empty pUTE657 vector (pEmpty) were grown under toxin-inducing conditions with IPTG induction, and cell lysates and supernatants were collected. As shown in Fig. 2C, there was a detectable band in the Δcer pCer cell lysates at approximately 55 kDa and a stronger band at approximately 50 kDa. No product was observed in the WT or Δcer cell lysates or in the empty vector control. Additionally, no protein was detected in the culture supernatants for any of the strains when probed with anti-ADPr (data not shown). Collectively, these data suggest that Certhrax is a cell-associated protein of approximately 55 kDa that may undergo processing in B. cereus G9241 to a protein of approximately 50 kDa. The processing of the protein may be N terminal, as suggested by the absence of reactivity of the anti-His antibody with Δcer pHisCer cell lysates (data not shown). Additionally, the protein may be expressed at low levels in G9241 in vitro and can therefore be detected only when overexpressed.

FIG 2.

Certhrax expression in E. coli and B. cereus G9241 cell lysates from IPTG-inducible constructs. A promotorless cer and his-cer, used to express an N-terminally tagged His-Certhrax, were cloned into vector pUTE657 to generate constructs pCer and pHisCer, respectively, where cer expression was placed under the control of an IPTG-inducible promoter. The constructs were transformed into E. coli Mach1 and electroporated into B. cereus G9241 Δcer for induction with IPTG, as described in Materials and Methods. Cell lysates of induced and uninduced E. coli transformed with pCer or pHisCer were subjected to SDS-PAGE and the proteins were transferred to nitrocellulose membranes for Western blot analyses with anti-ADPr (A) or anti-His antibodies (B). WT G9241, the Δcer mutant strain, and the Δcer pCer and Δcer pEmpty transformant strains, which carried either the pCer vector or the empty pUTE657 vector, respectively, were grown under toxin-inducing conditions with IPTG induction. The cell lysates were subjected to SDS-PAGE and the proteins were transferred to a membrane for Western blot analyses with the anti-ADPr antibody (C). The lines in panel C represent the well that each sample was loaded into; the larger line for Δcer pCer indicates sample spillover into the neighboring left well.

While lethal factor contributes to virulence in B. cereus G9241, Certhrax attenuates virulence of the pathogen.

To determine whether Certhrax plays a role in pathogenicity of G9241, the virulence of a markerless deletion mutant strain, the Δcer strain, was assessed in A/J and C57BL/6 mice inoculated through the s.c. and i.n. routes. The LD₅₀ value of the mutant strain for the s.c. route in A/J mice was calculated to be <20 spores, while in the C57BL/6 mice it was 70 spores. For the i.n. route, the LD₅₀ values were 5 × 10² and 3 × 10⁵ in A/J and C57BL/6 mice, respectively. The mutant strain, therefore, appeared as virulent as WT G9241 in all of the in vivo inoculation models (Table 1). Furthermore, there was no statistical difference in the median times to death during s.c. inoculation of either mouse strain (data not shown), even at the lowest inoculum dose (10¹). These data suggest that Certhrax does not play a prominent role as a virulence-enhancing factor in G9241, as originally hypothesized. Conversely, it was observed that A/J mice inoculated with Δcer spores through the s.c. route appeared to form a larger abscess at the site of inoculation (a large bump at the right flank of the mouse) than did WT G9241 (data not shown). Furthermore, we saw 100% lethality even at the 2 × 10¹-spore inoculum dose for the Δcer strain following s.c. inoculation of A/J mice, compared to 13% survival in the WT G9241 group challenged with the same dose. These data suggest that perhaps the Δcer mutant induced more pathology than WT G9241 in the s.c. challenge of A/J mice, but a more accurate LD₅₀ for the mutant strain was below the limit of detection in our assay.

Since lethal toxin is an important virulence factor in B. anthracis, studies were carried out to characterize the role of LT in pathogenicity of G9241 and to determine if its in vivo activity could mask any observable effects of Certhrax. For these experiments, we generated a markerless deletion mutant strain of lef (Δlef mutant) as well as a double cer and lef knockout strain (Δcer Δlef mutant). The Δcer Δlef mutant strain was constructed via deletion of lef in the Δcer mutant strain background. A/J and C57BL/6 mice were inoculated s.c. and i.n. with each mutant strain. As shown in Table 1, the deletion of lef significantly attenuated virulence in both mouse strains. In fact, the Δlef mutant was avirulent during s.c. inoculation of C57BL/6 mice and just a few mice died following intranasal inoculation with the highest (10⁷) dose of mutant spores. There was also a significant attenuation in virulence of the Δlef strain compared to that of WT G9241 in A/J mice. The calculated LD₅₀ values were 7 × 10⁵ and 2 × 10⁵ spores for the s.c. and i.n. routes, respectively (Table 1). These data show that LT is essential for full virulence of B. cereus G9241 in these mouse models.

To determine whether expression of LF could mask any in vivo effects of Certhrax, mice were inoculated with Δcer Δlef G9241 spores. The Δcer Δlef mutant strain had LD₅₀ values comparable to those of the Δlef mutant following i.n. inoculation of A/J mice and for both administration routes in C57BL/6 mice (Table 1). In each case, the double-knockout strain was significantly attenuated compared to WT G9241, but it was not different from the Δlef strain. However, an exception was noted following the s.c. inoculation of A/J mice. The calculated LD₅₀ value for the Δcer Δlef strain in the s.c. model was 1 × 10³ spores, a value that is statistically lower than what was observed with the Δlef mutant strain. These results were consistent in three separate infection experiments and suggest that Certhrax behaves as an antivirulence factor after subcutaneous inoculation of A/J mice.

The enzymatic activity of Certhrax is responsible for its antivirulence activity.

To determine whether the increased-virulence phenotype of the Δcer Δlef strain compared to that of the Δlef strain was due to its enzymatic activity, two cer restorants were generated within the Δcer Δlef mutant strain background. One restorant, the cer.rest Δlef strain, was created by cloning wild-type cer into its native locus in Δcer Δlef G9241. The other restorant, the inact.cer Δlef strain, was made with an enzymatically inactive cer as described in Materials and Methods. Both strains expressed cer transcript at levels comparable to that of WT G9241 following growth of cultures under toxin-inducing conditions in vitro (data not shown). The virulence phenotype of each restorant strain was assessed in the s.c. inoculation model of A/J mice. As shown in Table 1, the cer.rest Δlef mutant strain of G9241 was highly attenuated compared to WT G9241, with a calculated LD₅₀ of 5 × 10⁶ spores. Furthermore, this restorant was significantly more attenuated than the parent Δcer Δlef strain, with a phenotype similar to that of the Δlef strain. The inact.cer Δlef mutant strain, on the other hand, had an LD₅₀ value similar to that of the Δcer Δlef mutant strain of G9241. In fact, the LD₅₀ values of the two strains were not statistically different. These results show that the enzymatically inactive cer did not restore the virulence of the Δcer Δlef mutant strain to Δlef strain levels, while the enzymatically active cer did so. Therefore, these data suggest that the enzymatic activity of Certhrax may be responsible for the observed virulence difference between the Δlef and Δcer Δlef strains.

We next sought to assess whether there was a difference in in vivo dissemination between the two restorant strains of G9241. To do so, 20 A/J mice were s.c. inoculated with a dose of 10⁶ spores of the inact.cer Δlef restorant, while another group of 20 mice were inoculated with the same dose of cer.rest Δlef spores. The intent was to measure differences in spore and bacterial burdens over time in the abscess formed at the site of inoculation and in the spleens of the mice inoculated with the two strains. The weights of the collected abscesses and spleens are depicted in Fig. 3A and B, respectively. There was no difference in spleen weights between mice inoculated with the cer.rest Δlef and inact.cer Δlef spores at each time point, but there was a gradual increase in the average spleen weights over time (Fig. 3B). However, a significant difference in abscess weights was noted at 24 h postinoculation between the two inoculum strains (P < 0.0001 as calculated via a two-way analysis of variance [ANOVA] with a Bonferroni adjustment for multiple comparisons) (Fig. 3A). Mice inoculated with the inact.cer Δlef restorant formed significantly larger abscesses at the site of inoculation than mice inoculated with the other mutant strain. The abscess weights were comparable at all other time points between the two strains. Furthermore, despite the difference noted at 24 h, there was no difference in the overall bacterial burdens in the collected abscesses between the two mutant strains at any time point (Fig. 3C). However, the average spore count (per gram of tissue) was higher, albeit not statistically significantly different, in the abscesses of mice inoculated with the inact.cer Δlef restorant at the 6-h time point (Fig. 3D). The numbers of spores per gram were comparable at the later time points. Conversely, while also not statistically significant with just 5 mice per group, there was a larger overall bacterial burden in the spleens of mice inoculated with the cer.rest Δlef mutant strain of G9241 at 6 and 24 h postinoculation. At 72 h the CFU per gram were comparable between the two strains. We also noted that mice that died of the infection (marked with “×” in Fig. 3), or were moribund, had a bacterial burden greater than 10⁷ in the spleen. Mice that were inoculated with the cer.rest Δlef restorant and did not die of the infection had a lower bacterial count in the spleen compared to the mouse that succumbed to the illness in the same group (Fig. 3E). The mice inoculated with the inact.cer Δlef restorant that succumbed to the infection all had a high bacterial burden in the spleen (P < 0.0001 compared to the value for the cer.rest Δlef restorant). Additionally, with the exception of a few mice, most of the bacteria in the spleen were vegetative cells and not spores (Fig. 3F). Collectively, the data suggest that the inact.cer Δlef mutant strain persists locally at the site of inoculation longer than the cer.rest Δlef mutant strain, which starts to disseminate within the first 6 h postinoculation. Despite the earlier dissemination, however, mice inoculated with cer.rest Δlef G9241 appeared to control the infection better than inact.cer Δlef strain-infected mice; bacterial counts remained lower in the spleens of the cer.rest Δlef strain-infected mice. Therefore, Certhrax may play a role as a virulence-attenuating factor of B. cereus G9241 early in infection.

FIG 3.

The cer.rest Δlef and inact.cer Δlef mutant strains exhibit differences in dissemination during s.c. challenge of A/J mice. Mice were inoculated with a dose of 10⁶ cer.rest Δlef (blue) or inact.cer Δlef (red) mutant spores. At 6, 24, and 72 h postinoculation, five mice per group were euthanized and their spleens and abscesses, formed at the site of inoculation, were collected. The last group of five mice (referred to as “Natural”) for each strain succumbed to the infection and the tissues were collected upon death. The abscesses (A) and spleens (B) were weighed when collected. The horizontal lines in panels A and B represent average weights for each group, and the error bars represent the SEMs. The abscesses were subsequently mashed in PBS, while the spleens were homogenized. Half of each sample was untreated (C and E), while the other half was heat treated to kill vegetative cells (D and F). All samples were serially diluted and plated for enumeration of CFU. The calculated abscess (C and D) and spleen (E and F) CFU per gram are represented for each mouse. The lines in panels C to F represent the geometric means for each group, and the error bars represent the 95% confidence interval. The “×” symbols represent mice that were dead upon tissue isolation. The dashed black line represents the limit of detection. Statistical differences were assessed with a two-way ANOVA; a P value of <0.05 is considered significant.

DISCUSSION

The overarching goal of the work described here was to determine the role that toxins play in pathogenesis of the anthrax-like B. cereus strain G9241. The approach taken in this study was to generate a series of single- and double-toxin-component deletion strains and, subsequently, to characterize their virulence phenotype in vivo following subcutaneous and intranasal challenge of mice. The two mouse strains used in the study, A/J and C57BL/6, are commonly employed to evaluate B. anthracis virulence. A/J mice are often so tested because of their susceptibility to anthrax infection, even with the attenuated unencapsulated B. anthracis strain Sterne (26). The increased sensitivity of this mouse strain is due to its deficiency in the expression of a C5 component of complement, which leads to a weakened host immune response to infection. C57BL/6 mice, which are C5⁺, are considered to be of intermediate susceptibility to infection with B. anthracis wild-type strains (27). The combined use of these two mouse strains occasionally reveals roles for B. anthracis virulence factors that may be otherwise masked in anthrax infections of other inbred murine strains.

We showed that the anthrax toxin components PA and LF were essential for full virulence of B. cereus G9241 whether A/J or C57BL/6 mice were infected by the s.c. or i.n. route. These findings were not unexpected based on the role of lethal toxin in the pathogenicity of B. anthracis in mice. Nevertheless, this was the first study to explore the contribution of LF to virulence of the G9241 strain. A ΔpagA1 mutant strain of G9241 was previously assessed for lethality after intraperitoneal challenge of C57BL/6 mice with 10⁵ spores; the ΔpagA1 mutant caused fewer deaths at that dose than WT G9241 (19, 20).

While anthrax toxin involvement in virulence of G9241 was one focus of the work described here, the major focus of the investigation was on the two less characterized toxin components, PA2 and Certhrax. PA2, as a homolog of protective antigen, was hypothesized to have a role similar to that of PA1 in toxin translocation into host cells. Certhrax, on the other hand, is a novel ADP-ribosyltransferase that was identified through in silico analyses by Fieldhouse et al. (28). While the protein shares structural homology with other known ADPRTs, it appears to be unique to G9241 and, perhaps, B. cereus 03BB87, a strain that is genetically indistinguishable from G9241. Certhrax was shown to inactivate vinculin in vitro, a target that prior to the study by Simon and Barbieri had not been described for ADPRTs (5). These investigators theorized that Certhrax could contribute to bacterial spread within the host following infection with G9241 because the toxin, when transfected into cells, disrupts actin polymerization and subsequently changes cell morphology. We speculated that since cer and pagA2 are located in close proximity on the pBC210 plasmid, PA2 may be the preferred binding component for Certhrax translocation into the cell.

The role of PA2 in G9241 pathogenicity was previously assessed in an i.p. inoculation model of C57BL/6 mice (20), and the ΔpagA2 mutant strain was as virulent as WT G9241 at a single spore dose of 10⁵. Tissue recovered from moribund or dead mice had G9241 loads similar to those described for mice exposed to WT spores. Oh et al. therefore concluded that PA2 is not involved in pathogenesis (20). In the s.c. and i.n. inoculation models described here, the ΔpagA2 mutant strain was also as virulent as WT G9241 at the higher spore inoculum doses. However, we observed a slight but significant increase in LD₅₀ values in C57BL/6 mice injected subcutaneously with the ΔpagA2 mutant spores compared to WT G9241. Furthermore, there was a decrease in the percentage of deaths when both A/J and C57BL/6 mice were inoculated with the lowest dose of the mutant strain compared to that for the mice exposed to WT G9241 spores. These data suggest that PA2 does contribute to the virulence of B. cereus G9241 but to a lesser extent than PA1. Furthermore, the ΔpagA1ΔpagA2 strain had LD₅₀ values comparable to those of the ΔpagA1 mutant strain, which adds to the idea that PA1 is the essential anthrax toxin-translocating subunit.

We subsequently sought to determine whether PA2 could functionally replace PA1 if it was expressed under the same conditions as the latter protein. Besides the amino acid differences between the two proteins, and the fact that they are expressed from two different virulence plasmids, there are other variables that could be at play. First, the promoter regions between the two genes are not homologous. Second, pagA1 is the first gene in an operon with the autorepressor pagR; there is no pagR-like gene immediately downstream of pagA2 (29). Therefore, we generated a pagA1::pagA2 ΔpagA1 knock-in mutant strain in the ΔpagA1 ΔpagA2 background, where pagA2 was inserted into the pagA1 locus. Surprisingly, the pagA1::pagA2 ΔpagA1 knock-in strain was as attenuated as the ΔpagA1 and ΔpagA1 ΔpagA2 mutant strains of G9241 following s.c. challenge of A/J mice. Therefore, PA2 cannot functionally replace PA1 in anthrax toxin translocation in B. cereus G9241.

There are several different possibilities that could explain why PA2 cannot substitute for PA1 as an anthrax toxin binding component. First, PA2 may not bind to either CMG2 or TEM8 receptors that PA of B. anthracis can use; there is just 45% amino acid identity between the receptor-binding domains of PA2 and PA of B. anthracis (20). Second, PA2 may not oligomerize or may not form a stable oligomer. Third, if PA2 can in fact oligomerize the way PA does, it is conceivable that the formed oligomer may not be capable of binding LF and EF. According to Cunningham et al. (16), there are seven residues in domain 1 of PA₆₃ that comprise the binding site for LF and EF in the oligomer. If any of the residues is mutated, there is essentially a complete loss of ligand binding capacity. Based on an amino acid alignment of PA2 to PA of B. anthracis, it appears that all of the residues are conserved in PA2 except for R178. Instead of an arginine, a positively charged residue, PA2 contains glutamic acid, a negatively charged amino acid, at that location. R178 is located in a basic interface between two bound PA subunits of the oligomer and, if mutated, can lead to 90% loss of substrate binding (16). Therefore, it is likely that the glutamic acid substitution in PA2 could affect LF and EF binding to the oligomer.

A further possible explanation for why PA2 cannot substitute for PA is that PA2 may not be involved in the translocation of the anthrax toxin components in G9241 at all. Instead, PA2 may in fact be the preferred binding subunit for Certhrax translocation into host cells. This hypothesis is particularly intriguing because we found that cer and pagA2 are cotranscribed when B. cereus G9241 is grown under toxin-inducing conditions. However, each gene also appears to also be expressed alone, as pagA2 and cer seem to have their own strong promoters.

As for the role that Certhrax plays in virulence, the data yielded some unexpected results. When cer was deleted in B. cereus G9241, the resulting mutant strain was as virulent as WT G9241 in A/J and C57BL/6 mice following s.c. and i.n. inoculation. Therefore, like PA2, Certhrax is not a major contributor to enhancement of virulence in G9241. However, an observation was made that A/J mice challenged with Δcer spores formed larger abscesses at the site of inoculation than did mice inoculated with the WT strain. This finding suggested that there may be more pathology at the site of inoculation with the mutant spores and perhaps the mutant strain was, paradoxically, more virulent than WT G9241. The construction of the Δcer Δlef double-deletion mutant strain and its subsequent evaluation in mice lent further credence to the idea that Certhrax may be an antivirulence factor of G9241 (Table 1). Indeed, the double mutant strain had a lower LD₅₀ in A/J mice following s.c. challenge than the strain with only the lef deletion. This phenotype was observed only in A/J mice, while in the C57BL/6 mouse model the Δcer Δlef mutant was avirulent. As the host immune response accounts for the biggest difference between the two mouse strains, we speculated that Certhrax may be involved in immune modulation. This possibility would be in keeping with some of the effects observed from the anthrax toxins as well (13).

As Certhrax has an N-terminal PA-binding domain and a C-terminal ADP-ribosyltransferase domain, we developed two hypotheses to explain the increased-virulence phenotype of the Δcer Δlef mutant strain, although they may not be mutually exclusive. The first theory was that Certhrax competes with EF and LF for binding to the PA oligomer. Based on amino acid alignments to EF and LF, Certhrax appears to have the conserved amino acids necessary for LF and EF to bind to the PA oligomer (30). Furthermore, Visschedyk et al. determined that Certhrax requires PA for entry into host cells to cause cytotoxicity (22). Thus, if Certhrax competes for binding with EF and LF, more of the anthrax toxin components could bind PA in the Δcer mutant strain and more anthrax toxin would be translocated into the cell to cause pathology.

The second hypothesis was that the loss of the ADP-ribosylating activity of Certhrax was responsible for the virulence phenotype of the Δcer Δlef mutant strain of G9241. Vinculin has been identified as the target for Certhrax in HeLa cells (5). Vinculin is a critical protein in cell-cell and cell-matrix adhesion complexes, as it binds actin to promote its polymerization and recruits remodeling proteins to the complexes. When vinculin is absent, actin polymerization is compromised, which, in turn, causes a detrimental effect on cell morphology (31). Vinculin inactivation also results in defects in cell migration (32). Therefore, the inactivation of vinculin in G9241-infected cells perhaps disrupts their migration and prevents G9241 from rapidly spreading within the host. Consequently, when cer is deleted and vinculin is unaffected, G9241 spreads more rapidly.

To test this hypothesis directly, we generated enzymatically active and enzymatically inactive cer restorants in the Δcer Δlef mutant strain background. Certhrax was enzymatically inactivated by mutating two catalytic residues in the ADPr domain, as has been previously described in the literature (22, 33). These mutant strains were examined for virulence in the s.c. challenge model of A/J mice, as this model appears to be more sensitive for studying toxin involvement in pathogenesis of G9241. The cer.rest Δlef mutant strain showed an attenuated virulence phenotype compared to that of the parent Δcer Δlef strain, and the LD₅₀ was comparable to that of the Δlef mutant strain of G9241. The enzymatically inactive mutant strain, on the other hand, remained as virulent as the Δcer Δlef mutant strain. Therefore, these data suggest that the ADP-ribosylating activity of Certhrax contributes to the observed virulence phenotype between the Δcer Δlef and Δlef strains. However, the possibility that Certhrax competes for binding with LF and EF has not been ruled out by these experiments.

We further sought to determine whether the inact.cer Δlef mutant strain was more pathogenic because it was able to disseminate more rapidly within the host than the cer.rest Δlef strain. There were several differences observed between the two restorant strains during the experiment, in which we followed spore and bacterial counts over time (Fig. 3). Many of the differences were apparent at the first time point evaluated. These kinetics suggest that Certhrax elicits its effect early in infection. At 6 h postinoculation, no bacteria were isolated from spleens of most mice inoculated with the inact.cer Δlef strain, while spleens collected from mice inoculated with the cer.rest Δlef strain had an average of 10³ CFU/ml of bacteria. This observation may indicate that early in infection the inact.cer Δlef strain does not disseminate from the site of inoculation as rapidly as does the cer.rest Δlef mutant strain. This finding is the opposite outcome from what was initially hypothesized. At that same time point, abscesses collected from mice inoculated with the inact.cer Δlef restorant had more spores in the tissue than the mice inoculated with the cer.rest Δlef strain. This difference in counts was also true for the vegetative bacteria at that time, but the difference was not as pronounced. These results imply that the enzymatically inactive restorant remains at the site of infection longer but appears to disseminate more effectively and at larger numbers later in infection, which may explain its increased-virulence phenotype.

The most striking difference between the two restorant strains, however, was the difference in abscess size at 24 h postinoculation. The abscesses collected from mice challenged with the inact.cer Δlef spores were much larger at that time point. Furthermore, there was no difference in bacterial counts in the abscesses at the 24-h time point, an observation that raises the question, what is the actual composition of the abscesses? Later in infection, however, the abscess sizes were comparable between the two mouse groups. This once again suggests that Certhrax appears to act early in infection, and the effect may be more localized to the site of inoculation. Additionally, as mentioned above, since this effect is observed only in A/J mice, Certhrax may have a greater effect on mice whose immune response is compromised.

These results paint a complex picture for Certhrax involvement in pathogenesis, as some of the data are counterintuitive. As depicted in Fig. 4, perhaps when Certhrax is expressed by B. cereus G9241, as is the case in WT G9241 and the Δlef and cer.rest Δlef mutant strains, Certhrax is produced by germinated bacilli inside phagocytes and is able to target and ADP-ribosylate vinculin in those phagocytic cells (Fig. 4A). Inactivation of vinculin subsequently leads to actin rearrangement and loss of cell integrity, and these events allow the bacteria to escape the cell to disseminate in the blood. This dissemination occurs soon after inoculation, and, therefore, low levels of bacteria are detected in the spleen at the 6-h time point. Additionally, if vinculin is inactivated there may be a disruption in cell migration of the immune cells. Therefore, as the infection progresses, the phagocytes cannot be used effectively by G9241 as a vehicle to disseminate throughout the host.

FIG 4.

Proposed model of Certhrax involvement in pathogenesis of B. cereus G9241. The bacterial spore is represented by the black circle, a phagocytic cell is the yellow star-like shape, vegetative bacterial cells are represented by dark blue rods, and the active Certhrax ADPRT is represented by the red star. (A) Spores of B. cereus G9241 strains that can express Certhrax (WT, Δlef, and cer.rest Δlef) get engulfed by phagocytic cells at the site of the inoculation. The spores germinate into vegetative bacteria in the phagocytic cell, where some bacteria are killed by the phagocyte, while others survive and begin to replicate. The vegetative cells secrete various virulence factors and Certhrax. Certhrax ADP-ribosylates vinculin, which, in turn, leads to actin depolymerization and loss of cell integrity of the phagocytic cells that lyse, allowing for released bacteria to disseminate into the blood. Furthermore, vinculin inactivation disrupts cell migration so that the phagocytic cells do not efficiently support bacterial dissemination. Thus, low numbers of bacteria reach the spleen early in infection. (B) Spores of B. cereus G9241 strains that do not express an enzymatically active Certhrax (Δcer, Δcer Δlef, and inact.cer Δlef) are engulfed by phagocytic cells, where the spores begin to germinate. The vegetative cells that survive killing replicate but do not express Certhrax. The phagocytic cell remains intact until the bacteria replicate to high enough levels to lyse the cell with a subsequent delay in dissemination. Since vinculin remains unaffected, some phagocytic cells with intracellular bacteria are able to migrate throughout the host and promote bacterial spread. Upon lysis of the phagocyte, high numbers of bacteria are released into the bloodstream and seed the spleen in larger quantities than in panel A.

Conversely, when Certhrax is enzymatically inactive or the gene is deleted, G9241 spores are still taken up by the phagocytic cells. However, vinculin is not targeted by the enzymatically inactive Certhrax and the phagocytic cells remain intact (Fig. 4B). Phagocytes are able to kill some of the germinated bacteria but not all. The surviving organisms replicate within the immune cells until the sheer number of bacteria lyses the cell. This delay in bacterial dissemination at the early time points amplifies the infection and allows for more bacteria to disseminate to the spleen later in infection. Additionally, since vinculin is active, the immune cells are able to migrate from the site of infection more efficiently to distribute G9241 within the host.

Another possibility is that enzymatically active vinculin actually enhances the B. cereus G9241 infection. In fact, in the case of S. aureus, researchers found that vinculin binds Rab5 and cooperatively enhances S. aureus uptake in nonphagocytic cells (34). Rab5 is a membrane-trafficking protein located in membranes of early endosomes and the plasma membrane that plays a role in the internalization of many molecules, including bacteria. Overexpression of vinculin enhances S. aureus phagocytosis, while knockdown of vinculin expression by small interfering RNA (siRNA) technology results in a reduced infection (34). Therefore, it is possible that vinculin also enhances B. cereus G9241 phagocytosis by nonphagocytic or phagocytic cells at the site of inoculation.

Despite our findings, there are limitations to our study. The most obvious of these is that demonstration of the antivirulence effect of Certhrax relied upon simultaneous deletion of lethal factor in the parent G9241 strain. Given that the WT G9241 strain LD₅₀ was 10 spores in the s.c. A/J mouse model, we were not able to demonstrate that the Δcer strain was more virulent than the wild type in this model; it is too technically challenging to accurately administer a lower inoculum dose than the lethal 10 spores. Simply stated, the WT strain killed the mice too efficiently to be able to see enhanced virulence once Certhrax was deleted. Despite this limitation, two different observations strongly supported the conclusion that the Δcer strain was more virulent than WT: (i) the formation of larger bumps at the site of inoculation of the Δcer spores (an observation that was also made for the enzymatically inactive restorant strain) and (ii) the deaths of 100% of the A/J mice following inoculation even with the lowest dose of Δcer spores, which was a higher rate of killing than seen with the WT. Clearly, more data are needed to fully understand the involvement of Certhrax in pathogenesis of B. cereus G9241. First, it could be important to analyze the composition of the abscesses through microscopy of the isolated abscess material to determine what immune cells are predominantly present. The presence of pus suggests that neutrophils may be involved. Additionally, it is possible that Certhrax targets a protein other than vinculin. Vinculin was identified as the target in HeLa cells (5, 33), but it is conceivable that there could be a different target in phagocytic cells.

In conclusion, our findings suggest that PA2 does contribute to G9241 pathogenesis, albeit in a more limited capacity than PA and LF. Our studies also indicate that there are differences in the functional capabilities between PA and PA2. Furthermore, and quite unexpectedly, we showed that Certhrax, a novel ADP-ribosyltransferase, behaves as an antivirulence factor following subcutaneous challenge of A/J mice with B. cereus G9241 spores.

MATERIALS AND METHODS

Bacterial growth conditions and spore preparation.

The bacterial strains and plasmids used in this study are listed in Table S1 in the supplemental material. The Bacillus cereus G9241 and Escherichia coli strains were routinely grown in Luria-Bertani (LB) broth (BD Biosciences, East Rutherford, NJ) overnight at 37°C with shaking at 225 rpm. Antibiotics were added as needed at the following concentrations: kanamycin (Kan) at 100 μg/ml, spectinomycin (Sp) at 200 μg/ml, and erythromycin (Erm) at 5 μg/ml for B. cereus and ampicillin (Amp) at 100 μg/ml and Kan at 50 μg/ml for E. coli. To promote toxin expression, an overnight culture of G9241 was grown in heart infusion (HI) broth (BD Biosciences) and diluted 1:1,000 in HI broth supplemented with 0.8% sodium bicarbonate. That culture was then incubated for 4 h at 37°C in 5% CO₂ with shaking (called toxin-inducing conditions).

G9241 spores were prepared as described by Scarff et al. (35). Briefly, G9241 was grown overnight in 50 ml of brain heart infusion (BHI) broth (BD Biosciences) at 37°C with shaking at 225 rpm. The whole culture was added to 1 liter of Difco sporulation medium (DSM) consisting of 0.8% nutrient broth, 0.012% MgSO₄, 0.1% KCl, 1 mM Ca(NO₃)₂, 6 mM NaOH, 1 μM FeSO₄, and 0.01 mM MnCl₂. The culture was grown for 5 days at 30°C with shaking at 225 rpm. The bacteria were pelleted by centrifugation, washed three times with sterile water, resuspended in sterile water, and heat treated at 65°C for 1 h to kill vegetative cells. The heat-resistant spores were stored at 4°C.

All work with B. cereus G9241 vegetative cells and spores was done in a biosafety cabinet, in accordance with biosafety level 2+ (BSL2+) safety standards. The personal protective equipment (PPE) for researchers who worked with spores included an N95 respirator, extended cuff disposable gloves, safety goggles, and a disposable lab coat.

Construction of B. cereus mutant strains and recombinant plasmids.

B. cereus deletion mutations were generated through homologous recombination with a temperature-sensitive plasmid as described by Scarff et al. (35). Plasmid pJMS1 was derived from pLM4 (19). For each mutant strain, 1-kb regions flanking the gene of interest were amplified via PCR with gene-specific primers listed in Table S2 in the supplemental material. The PCR products were first ligated into a pGEM-T vector (Promega, Madison, WI), such that the 3′ region immediately followed the 5′ flanking region. The 2-kb fragment was then subcloned into pJMS, and the recombinant plasmid was transformed into One Shot Mach1-T1R chemically competent E. coli cells (Invitrogen by Life Technologies). The plasmid was isolated from the Mach1 cells and subsequently transformed into dam- and dcm-deficient E. coli (New England BioLabs, Ipswich, MA) to obtain unmethylated DNA. One microgram of the purified unmethylated plasmid DNA was electroporated into B. cereus G9241 as described by Vergis et al. (36). Transformants were recovered on LB agar plates supplemented with Kan at 30°C. Subsequently, transformant B. cereus G9241 colonies were subjected to several passaging steps. First, to force plasmid integration, overnight cultures of the transformants were passaged at a 1:100 dilution 2 or 3 times at 37°C in the presence of Kan. Second, the integrants were passaged 8 or 9 times at 30°C in the absence of antibiotic to promote secondary recombination and loss of the pJMS1 vector. Colonies were screened for kanamycin sensitivity through patching on plain LB agar plates and on plates supplemented with Kan. A complete description of the cloning strategy for each mutant strain can be found in the supplemental material.

Each mutant strain was confirmed through Southern blot analysis (see Fig. S3 and S4 in the supplemental material) with digoxigenin (DIG)-labeled gene-specific probes as per the manufacturer's guidelines (Roche, Pleasanton, CA). Furthermore, each strain was characterized as follows (data not shown): expression of PA, LF, EF, and PA2 was confirmed by Western blot analyses with specific antibodies, growth rates in comparison to that of wild-type G9241 were measured in LB broth, expression of both capsules was visualized through Maneval staining with and without hyaluronidase treatment (35), and in vitro germination of spores was assessed in the presence of l-alanine and inosine (18). Only mutant strains that showed no defects in any of these additional phenotypic assays were utilized in this study.

Vector pUTE657 (25) was used to generate two distinct IPTG-inducible cer expression constructs through application of In-FusionHD cloning technology (Clontech, Mountain View, CA). Primers P35 and P37 were used to amplify cer (84 to 1,530 bp) to construct plasmid pCer, whereas primers P36 and P37 were used to amplify cer with an N-terminal histidine tag (sequence CATCATCACCATCACCAC) to generate plasmid pHisCer. The primers are listed in the supplemental material Table S2. The constructs were transformed into E. coli Mach1 competent cells (Amp^r). The transformants were confirmed through PCR and sequencing. The pCer and pHisCer vectors, as well as the pUTE657 empty vector, were transformed into dam- and dcm-deficient cells and subsequently electroporated into Δcer G9241.

Purification of an anti-ADPr antibody from immunized rabbit serum.

A glutathione S-transferase (GST)-tagged ADPr region of Certhrax was used for affinity column purification of an anti-ADPr antibody from rabbit sera. GST-ADPr was first expressed from a pET-42b(+) expression vector (Novagen by Millipore Sigma) that was generated via PCR amplification of the cer ADPr region with primers P46/P47. The amplified product and pET-42b(+) were digested with PshAI and XhoHI endonucleases and then ligated. The construct was transformed into BL21(DE3) pLysS E. coli cells for expression following induction with 100 mM IPTG. Four hours postinduction, cells were pelleted, suspended in binding buffer (10 mM imidazole, 50 mM sodium phosphate, 200 mM NaCl [pH 8]), and sonicated to lyse the bacteria. Insoluble material was removed by centrifugation at 12,000 rpm for 20 min. Cell lysate supernatants were collected and filtered through a 0.4-μm membrane. The supernatant that contained the GST-ADPr was then bound to a glutathione affinity column (GE Healthcare) for purification as described in the manufacturer's manual. The purified GST-ADPr was covalently bound to an N-hydroxysuccinimide (NHS)-activated fast protein liquid chromatography (FPLC) column from GE Healthcare as described in the manufacturer's user manual.

Serum was obtained from a rabbit that was immunized with a purified His-tagged ADPr domain of Certhrax (Pocono Rabbit Farm & Laboratory Inc., Canadensis, PA). The His-ADPr was generated through a collaboration with Joseph Barbieri and Nathan Simon as described by Simon et al. (33). The rabbit ADPr antisera was subsequently concentrated via affinity purification with the GST-ADPr column. The rabbit antiserum sample was loaded onto the column and was eluted with 0.1 M glycine, pH 2.8. The eluent was desalted into phosphate-buffered saline (PBS) and then incubated with plasmid-cured G9241 lysates to remove nonspecific antibodies to Bacillus. The nonspecific antibody-bacterium mixture was pelleted by centrifugation, and the supernatant was filtered through 0.22-μm membrane.

In vitro expression of Certhrax.

The IPTG-inducible constructs pCer and pHisCer, transformed into E. coli Mach1 cells or electroporated into the Δcer mutant strain of B. cereus G9241, were used to detect Certhrax expression via Western blot analyses with the anti-ADPr antibody.

The E. coli strains that contained the pCer and pHisCer constructs were grown overnight in LB broth supplemented with Amp. The overnight cultures were subcultured at a 1:50 dilution into two tubes that contained fresh LB-Amp and grown to an optical density at 600 nm (OD₆₀₀) of 0.4 to 0.8. One of the cultures was induced with 100 μM IPTG and further grown overnight at 37°C with shaking, while the other culture was left uninduced and grown overnight as a control. The samples were centrifuged and the cell pellets were resuspended in 200 μl of 4× SDS-PAGE buffer.

WT, Δcer, Δcer pCer, Δcer pHisCer, and Δcer pEmpty (empty vector control) G9241 strains were grown under toxin-inducing conditions, except that the cultures were grown for 24 h without shaking instead of 4 h with shaking. During the 24-h incubation period, the cultures were induced with 100 μM IPTG at 0 and 22 h. Following incubation, the cultures were centrifuged to collect the cell lysates. G9241 cell lysates were prepared based on the protocol described by Hammerstrom et al. (37). Briefly, the cell pellets were washed once with 1 ml KTE-PIC buffer (10 mM Tris-HCl [pH 8], 100 mM KCl, 10% ethylene glycol, and 1× complete EDTA-free proteinase inhibitor [Roche by Millipore Sigma]). The pellet was resuspended in 550 μl of KTE-PIC and transferred to a tube that contained approximately 400 μl of 0.1-mm zirconia/silica beads (BioSpec Products, Bartlesville, OK). The pellet was beaten in the presence of the beads for 2 min with a Turbo Mixer vortex attachment, and the sample was then centrifuged at 10,000 × g at 4°C. The supernatant was collected, and 4× SDS-PAGE buffer was added to the sample.

The E. coli and B. cereus G9241 cell lysates and supernatants were subjected to SDS-PAGE. The protein was then transferred to a nitrocellulose membrane with the iBLOT dry blotting system apparatus (Life Technologies) for Western blot analyses, as described by Scarff et al. (35). Certhrax was detected with an anti-ADPr antibody, and His-tagged Certhrax was detected with an anti-His antibody (Qiagen, Germantown, MD; 34660), followed by horseradish peroxidase (HRP)-conjugated secondary antibodies (Bio-Rad, Hercules, CA; 170-6516/170-6515). The blots were visualized with the ImageQuant LAS4000 (GE Life Sciences, Piscataway, NJ).

RNA extraction and reverse transcriptase PCR.

Bacillus RNA was extracted by protocols modified from those of Lessard et al. (38) and Gilbreath et al. (39). Bacillus cultures were grown under toxin-inducing conditions to late exponential phase for 4 h. The bacteria in the culture (12.5 ml) were pelleted via centrifugation at 10,000 × g for 15 min at 4°C. The pellet was resuspended in 1 ml of lysozyme (20 mg/ml in Tris-EDTA [TE] buffer) (Millipore Sigma) and incubated for 2 h at 37°C, with shaking at 250 rpm. Two hundred microliters of 20% SDS (Millipore Sigma) was added to lyse the cells, followed by 10 μl of 20-mg/ml proteinase K (Millipore Sigma). The samples were then incubated for 1 h at 37°C with shaking. Three milliliters of TRIzol (Thermo Fisher Scientific, Waltham, MA) was added, and the tubes were shaken in a vortex 3 times for 5 s each and then allowed to sit at room temperature for 3 to 5 min. Chloroform was added at 0.2 ml per 1 ml of TRIzol, and the samples were vortexed for 15 s. The samples were centrifuged at 3,000 × g for 15 min at 4°C, and the top aqueous phase was transferred to a fresh 15-ml conical tube (Corning). An equal volume of 100% isopropanol was added, and the samples were shaken in a vortex at low speed. The RNA was subsequently isolated via a vacuum manifold procedure (39) in which the samples were loaded onto RNeasy columns (Qiagen) and placed onto the vacuum manifold. The RNA was eluted with RNase-free H₂O two times and was subjected to gel electrophoresis on a 1% agarose and 0.5× Tris-borate-EDTA (TBE) gel to assess the quality of the RNA.

The RNA was converted to cDNA with the Quantitect reverse transcription kit (Qiagen). Reverse transcriptase PCR was done to assess whether cer and pagA2 were cotranscribed. The PCR was carried out with the following primer pairs (Table S2): primers P29/P30 to detect cer transcript, primers P51/P52 for the pagA2 transcript, and primers P53/P54 for the cotranscript.

Additionally, quantitative RT-PCR (qRT-PCR) was used to confirm that cer.rest Δlef and inact.cer Δlef strains express wild-type levels of cer transcript. The reactions were carried out using primers P55/P56 and SYBR green master mix (Thermo Fisher Scientific).

Mouse infection and dissemination studies.

All mouse infections were done under animal biosafety level 2+ conditions and in accordance with the guidelines established by the Institutional Animal Care and Use Committee of the Uniformed Services University. Six-week-old female C57BL/6 and A/J mice were obtained from Jackson Laboratories (Bar Harbor, ME) and were housed in filter top cages with standard food and water. Five mice per group were inoculated with a range of B. cereus G9241 spore doses via the s.c. or i.n. route as described by Wilson et al. (18). For s.c. inoculation, 100 μl of spores was injected into the right flank of the mouse. For the i.n. inoculation, mice were lightly anesthetized via isoflurane inhalation, and 50 μl of spore inoculum was pipetted into each naris to be inhaled. A/J and C57BL/6 mice were inoculated via both routes with the following strains: the Δcer, Δlef, Δcer Δlef, ΔpagA2, ΔpagA1, and ΔpagA1 ΔpagA2 mutants. Additionally, A/J mice were challenged s.c. only with the cer.rest Δlef, inact.cer Δlef, and pagA1::pagA2 ΔpagA1 G9241 mutant strains. At least two independent experiments were done with each strain and for each inoculation route. Following inoculation, mice were observed for morbidity and mortality, and moribund animals were humanely euthanized. The combined data were collected and the LD₅₀ values calculated via probit analyses with SPSS software (IBM, Armonk, NY). The Reed-Muench method (23) was used to calculate the LD₅₀ when the requirements for the probit analysis were not met. The median times to death were calculated for all challenge groups and compared statistically with the Mann-Whitney rank sum test. The percent survival of mice inoculated s.c. with the ΔpagA2 or Δcer mutant strains was compared to that of mice inoculated with the WT G9241 strain value by Fisher's exact test.

Forty A/J mice were inoculated via the s.c. route in a dissemination experiment. Twenty mice were inoculated with 10⁶ of cer.rest Δlef spores and the rest with 10⁶ inact.cer Δlef spores. For each group, 5 mice per cage were humanely euthanized at 6 h, 24 h, or 72 h postinoculation, while mortality in one cage of mice was followed through day 5 of infection. Any surviving mice in that cage were humanely euthanized at day 5 postinoculation. For each time point, mouse abscesses and spleens were collected for enumeration of CFU. The mouse abscesses were excised from the site of injection and placed into 1 ml of PBS in 2-ml microcentrifuge tubes (USA Scientific). The abscesses were weighed and then mashed with a plunger from a 10-ml syringe (BD, Franklin Lakes, NJ). The solid tissue was removed and discarded. The spleens were collected in 15-ml tubes with 1 ml of PBS. The organs were weighed and homogenized. For both tissues, half of each sample was heat killed at 65°C for 30 min, and then both halves were serially diluted and plated on LB agar plates for enumeration. The data were plotted and analyzed with GraphPad Prism software. Data were compared with a two-way ANOVA and the Bonferroni adjustment for multiple comparisons.