The Roles of AtxA Orthologs in Virulence of Anthrax-like Bacillus cereus G9241

Abstract

AtxA is a critical transcriptional regulator of plasmid-encoded virulence genes in Bacillus anthracis. Bacillus cereus G9241, which caused an anthrax-like infection, has two virulence plasmids, pBCXO1 and pBC210, that each harbor toxin genes and a capsule locus. G9241 also produces two orthologs of AtxA: AtxA1, encoded on pBCXO1, and AtxA2, encoded on pBC210. The amino acid sequence of AtxA1 is identical to that of AtxA from B. anthracis, while the sequences of AtxA1 and AtxA2 are 79% identical and 91% similar to one another. We found by qRT-PCR that AtxA1 and AtxA2 function as positive regulators of toxin (AtxA1) and capsule operon (both) transcription in G9241 and that a ΔatxA1 mutant produced lower levels of the anthrax toxins and no hyaluronic acid capsule. Deletion of atxA1 or atxA2 decreased the virulence of spores administered intranasally or subcutaneously to C57BL/6 mice but not to A/J mice, and deletion of both genes rendered spores avirulent in A/J mice. In addition, unlike AtxA1, AtxA2 did not form stable homomultimers in vitro, although AtxA1 and AtxA2 formed heterodimers. Our data show that AtxA1 is the primary regulator of G9241 virulence factor expression and that AtxA1 and AtxA2 are both required for full virulence.

Introduction

Bacillus cereus and Bacillus anthracis are members of the B. cereus sensu lato group. B. cereus is a causative agent of food poisoning and opportunistic infections, while B. anthracis is the etiologic agent of anthrax. Strains of B. anthracis harbor two virulence plasmids, pXO1 and pXO2, that play critical roles in anthrax disease. The plasmid pXO1 contains the genes pagA, lef, and cya that encode the anthrax toxin components protective antigen (PA), lethal factor (LF), and edema factor (EF), respectively (Vodkin & Leppla, 1983, Robertson & Leppla, 1986, Tippetts & Robertson, 1988, Okinaka et al., 1999). The plasmid pXO2 contains the genes necessary for production of a poly-γ-D-glutamic acid capsule (Makino et al., 1988). AtxA, encoded by atxA on pXO1, is a positive regulator of pagA, lef, and cya, and the capsule biosynthesis operon capBCADE in B. anthracis (Uchida et al., 1993, Dai et al., 1995). AtxA-mediated capsule gene regulation occurs via control of the paralogous regulators, AcpA and AcpB, which are encoded by genes on pXO2 (Drysdale et al., 2003, Drysdale et al., 2005). AtxA functions as a dimer, and dimerization is optimal when B. anthracis is cultivated in the presence of CO₂/bicarbonate (Hammerstrom et al., 2011). AtxA has two amino-terminal helix-turn-helix (HTH) DNA-binding domains, two phosphoenolpyruvate-dependent phosphotransferase regulatory domains, PRD1 and PRD2, and a carboxy-terminal EIIB-like domain that is required for dimerization (Tsvetanova et al., 2007, Hammerstrom et al., 2011, Hammerstrom et al., 2015). Phosphorylation of H199 in PRD1 enhances AtxA activity, whereas phosphorylation of H379 disrupts dimer formation and ablates AtxA activity (Tsvetanova et al., 2007, Hammerstrom et al., 2015).

B. cereus strains that harbor pXO1 and pXO2-like plasmids, named B. cereus Biovar anthracis, have been isolated as the causative agents of anthrax-like infections in primates in Cameroon and Côte d’Ivoire (Klee et al., 2006, Brezillon et al., 2015). Similar B. cereus strains that produce the anthrax toxins were identified as the etiological agents of anthrax-like respiratory infections in metalworkers in Texas and Louisiana and an anthrax-like cutaneous infection in Florida (Miller et al., 1997, Hoffmaster et al., 2004, Avashia et al., 2007, Wright et al., 2011, Gee et al., 2014). The respiratory infections in metalworkers involved many overlapping signs and symptoms with those exhibited by the victims of the 2001 inhalational anthrax attacks: fever/chills, labored breathing, cough, nausea/vomiting and/or diarrhea, bacteremia, and pleural effusions or infiltrates. However, there were differences in disease presentation between B. anthracis and B. cereus infection. For example, a widened mediastinum was observed in patients infected with B. anthracis but not B. cereus. Furthermore, the B. cereus, but not the B. anthracis, infections were also associated with hemoptysis (Avashia et al., 2007). Lastly, autopsy findings showed that inhalational anthrax patients had hemorrhage in multiple organs, including the lungs, while hemorrhage was predominantly found only in the lungs of patients with B. cereus infection (Abramova et al., 1993, Grinberg et al., 2001, Guarner et al., 2003, Wright et al., 2011).

B. cereus strain G9241 was isolated from a welder in Louisiana who suffered from a severe anthrax-like respiratory illness (Hoffmaster et al., 2004). G9241 contains two virulence plasmids, pBCXO1 and pBC210 (identified by Hoffmaster et al. as pBC218), as well as pBClin29, a linear plasmid that harbors cryptic prophage genes. The plasmid pBCXO1 has high similarity to pXO1 and contains the toxin genes pagA, lef, and cya. The amino acid sequences of PA, LF, and EF are 99.7%, 99%, 96% identical, respectively, to their counterparts in B. anthracis (Hoffmaster et al., 2004). The genes that encode a PA paralog, PA2, and the novel ADP-ribosyltransferase, Certhrax, are closely linked to one another on pBC210 (Hoffmaster et al., 2004, Visschedyk et al., 2012, Simon et al., 2013, Simon et al., 2014). Each plasmid contains genes required for synthesis of a unique polysaccharide capsule; the enzymes needed for hyaluronic acid (HA) capsule production are encoded by the hasACB operon on pBCXO1, and the proteins required for elaboration of a putative tetrasaccharide (TS) capsule are encoded by the bps locus on pBC210 (Hoffmaster et al., 2004, Oh et al., 2011). Each virulence plasmid also carries a distinct atxA allele. The amino acid sequence of AtxA1, produced from atxA1 on pBCXO1, is 100% identical to that of the AtxA encoded by B. anthracis pXO1, and the sequence of AtxA2, from atxA2 on pBC210, is 79% identical and 91% conserved to that of B. anthracis AtxA/G9241 AtxA1 (Hoffmaster et al., 2004). While B. cereus G9241 and B. anthracis Ames are similar because they elaborate capsule(s) and the anthrax toxins, the virulence of G9241 in rabbits was found to be more similar to that of B. anthracis Sterne strain 34F2 (pXO2⁻) that only produces the toxins (Wilson et al., 2011).

In this study, we investigated the roles of AtxA1 and AtxA2 in the regulation of toxin and capsule expression in G9241. We constructed isogenic atxA1 and atxA2 mutants in G9241, and demonstrated that each gene product exerted a regulatory role on capsule production in vitro and that AtxA1 was the predominant regulator of toxin production in vitro. We also demonstrated that deletion of atxA1 and/or atxA2 affected the virulence of G9241 in both A/J and C57BL/6 mice. Finally, we determined that AtxA2 has reduced activity compared to AtxA1, and that AtxA1 and AtxA2 have the capacity to form heteromultimers. Our study provides mechanistic insights into the complex regulatory network used by G9241 to control the production of toxins and capsules that are essential for full virulence of the organism.

Results

AtxA1 and AtxA2 control of capsule production

To investigate the roles of AtxA1 and AtxA2 in regulation of capsule and toxin production in G9241, we generated isogenic strains with atxA1 and/or atxA2 deleted (Table 1). We confirmed that pBCXO1 and pBC210 were retained in the mutants by PCR amplification with 10-12 primer pairs that span the circumference of each plasmid (Fig. S1). The growth rates in rich medium (Fig. S2) and germination rates in the presence of alanine and inosine (data not shown) of the atxA mutants were similar to those of G9241.

Table 1.

Strains and plasmids used in this study

Name	Description	Reference
Strains
B. cereus
G9241	Wild type strain with pBCXO1 and pBC210	(Hoffmaster et al., 2004)
Δ atxA1	G9241 ΔatxA1::Ω-kan; Kanr	This study
Δ atxA2	G9241 ΔatxA2	This study
Δ atxA1 ΔatxA2	G9241 ΔatxA1::Ω-kan, ΔatxA2::Ω-spec; Kanr Spr	This study
Δ atxA1 ΔatxA2::atxA1	G9241 ΔatxA1 ΔatxA2 with pHT304-atxA1; Kanr Spr Ermr	This study
Δ atxA1 ΔatxA2::atxA2	G9241 ΔatxA1 ΔatxA2 with pHT304-atxA2; Kanr Spr Ermr	This study
pBCXO1-pBC210-	G9241 cured of pBCXO1 and pBC210	(Wilson et al., 2011)
B. anthracis
UT376	Sterne ANR1 lef promoter - lacZ fusion (Plef- lacZ) at native lef locus, atxA-null	(Hammerstrom et al., 2011)
Plasmids
pUTE657	Expression vector derived from pDR111 and pBC16 with IPTG-inducible Phyper-spank; Spr Apr	(Pflughoeft et al., 2011)
pUTE991	pUTE657::atxA1-6xHis (His6 tag on the C- terminus of AtxA1); the atxA1 ribosome binding site and coding region controlled by Phyper-spank	(Hammerstrom et al., 2011)
pUTE992	pUTE657::atxA1-FLAG (FLAG tag on the C- terminus of AtxA1); the atxA1 ribosome binding site and coding region controlled by Phyper-spank	(Hammerstrom et al., 2011)
pUTE1096	pUTE657::atxA2-6xHis (His6 tag on the C- terminus of AtxA2); the atxA2 ribosome binding site and coding region controlled by Phyper-spank	This study
pUTE1122	pUTE657::atxA2-FLAG (FLAG tag on the C- terminus of AtxA2); the atxA2 ribosome binding site and coding region controlled by Phyper-spank	This study
pUTE1013	pUTE657::gfp-FLAG (FLAG tag on the C- terminus of GFP); the gfpmut3a ribosome binding site, coding region	(Hammerstrom et al., 2011)
pHT304	Low copy number expression plasmid for Bacillus, Apr in E. coli, Ermr in Bacillus	(Arantes & Lereclus, 1991)
pHT304-atxA1	pHT304::atxA1 and upstream promoter region	This study
pHT304-atxA2	pHT304::atxA2 and upstream promoter region	This study

Kan^r, kanamycin resistant; Sp^r, spectinomycin resistant; Ap^r, ampicillin resistant;; Erm^r, erythromycin resistant, Cm^r, chloramphenicol resistant

To determine whether AtxA1 and/or AtxA2 mediated control of capsule expression in B. cereus G9241, we assessed the relative levels of capsule gene transcripts in G9241 and the isogenic atxA mutants. We isolated RNA from bacteria grown to late-exponential phase under capsule-inducing conditions (14% CO₂, 0.8% sodium bicarbonate, 50% horse serum), and determined the expression levels of hasA, the first gene in the hyaluronic acid (HA) capsule locus on pBCXO1, and bpsA, the second gene in the putative tetrasaccharide (TS) capsule locus on pBC210, by quantitative real-time polymerase chain reaction (qRT-PCR) (Fig. 1). Compared to G9241, ΔatxA1 exhibited a 150-fold reduction in hasA transcript level, while ΔatxA2 was unaltered for hasA transcript production (Fig. 1A). In agreement with atxA1-specific control of hasA, the hasA transcript level in ΔatxA1ΔatxA2 was comparable to that of atxA1. The hasA transcript level in ΔatxA1ΔatxA2 was restored to levels comparable to G9241 by the introduction of atxA1 in trans (ΔatxA1ΔatxA2::atxA1), but no change in hasA expression was observed when atxA2 was introduced in trans (ΔatxA1ΔatxA2::atxA2). For the pBC210 TS capsule locus, bpsA transcript levels in ΔatxA1 and ΔatxA2 were unchanged compared to G9241. However, bpsA transcripts were reduced approximately 10-fold in ΔatxA1ΔatxA2. The bpsA transcript was restored to levels comparable to G9241 in ΔatxA1ΔatxA2::atxA1 and ΔatxA1ΔatxA2::atxA2, which suggests functional similarity of the two proteins with regard to positive control of bpsA. To determine if deletion of atxA1 or atxA2 altered the transcription of the other allele in capsule-inducing conditions, we assessed atxA1 and atxA2 transcript levels in the atxA mutants and the complemented mutants (Fig 1B). Transcript levels of atxA1 were comparable among G9241, ΔatxA2, and ΔatxA1ΔatxA2::atxA1. In contrast, atxA2 transcription was increased 20-fold in ΔatxA1 relative to G9241. Additionally, ΔatxA1ΔatxA2::atxA2 overexpressed atxA2 by a factor of 28 compared to G9241. These data indicate that AtxA1 is necessary for HA capsule production, but that either AtxA1 or AtxA2 is sufficient for production of the TS capsule.

Fig. 1.

qRT-PCR analyses of capsule gene and atxA transcripts in G9241 and atxA mutants. RNA was isolated from bacteria grown to mid-exponential phase in capsule-inducing conditions. (A) Relative expression of hasA and bpsA from HA and TS capsule loci, respectively, in G9241 and isogenic atxA mutants. (B) Relative expression of atxA1 and atxA2 in atxA mutants. Bars represent the mean, and error bars represent the standard deviation. # Significantly different from G9241 (P < 0.01); ^ significantly different from all other strains (P < 0.01); @ significantly different from ΔatxA1ΔatxA2 (P < 0.0001) as determined by one-way ANOVA with Tukey’s multiple comparisons post-test. n=5 for G9241, ΔatxA1, ΔatxA2, ΔatxA1ΔatxA2; n=3 for ΔatxA1ΔatxA2::atxA1 and ΔatxA1ΔatxA2::atxA2.

To assess the effects of the atxA1 and atxA2 deletions on capsule production, we grew G9241 and the atxA mutant strains in capsule-inducing conditions, treated the cells with or without hyaluronidase, and visualized the capsule(s) with the Maneval staining procedure (Maneval, 1941). Bacilli were first prepared on a slide with the negative stain, Congo red, followed by administration of the Maneval stain. The fuchsin in the Maneval stain makes the bacteria appear pink, while the acidity of the stain changes the Congo red to a blue/purple color. The capsule is apparent as a white halo around the pink bacteria. As was previously reported (Oh et al., 2011), G9241 produced both HA and TS capsules (Fig. 2A, 2H). The capsules were visible as large, round halos around untreated G9241 cells (Fig. 2A). After hyaluronidase treatment removed the larger HA capsule, the more compact TS capsule remained visible on the surface of G9241 (Fig. 2H). The ΔatxA1 mutant produced only the TS capsule, as evidenced by the unchanged appearance of the cells with and without hyaluronidase treatment (Fig. 2B, 2I). Similarly to G9241, ΔatxA2 produced both capsules; the larger halo indicative of HA capsule was visible without hyaluronidase treatment and the compact TS capsule remained apparent after hyaluronidase treatment (Fig. 2C, 2J). No capsule was detected on the surface of ΔatxA1ΔatxA2 cells (Fig. 2D) or on the negative control pBCXO1-/pBC210- strain that lacks both virulence plasmids (Fig. 2E). In ΔatxA1ΔatxA2::atxA1, the expression of both capsules was restored (Fig. 2F, 2K), but only the TS capsule was produced by ΔatxA1ΔatxA2::atxA2 (Fig. 2G, 2L). The capsule visualization results are in agreement with the qRT-PCR data. HA capsule was produced only in those strains with a functional AtxA1, and these strains had the highest amount of hasA transcript. Similarly, the only strain that did not produce TS capsule, ΔatxA1ΔatxA2, was the only strain with a significant reduction in bpsA transcript. Taken together, these data indicate that AtxA1 regulates production of the HA capsule, while either AtxA1 or AtxA2 is sufficient for production of the TS capsule in G9241.

Fig. 2.

Capsule production in atxA mutant strains. Capsule production by G9241 (A, H), ΔatxA1 (B, I), ΔatxA2 (C, J), ΔatxA1ΔatxA2 (D), pBCXO1-pBC210- (E) ΔatxA1ΔatxA2::atxA1 (F, K), ΔatxA1ΔatxA2::atxA2 (G, L). Each strain was grown in capsule inducing conditions for 24 hrs. Cells were treated with water (A-G) or 200U hyaluronidase in water (H-L) at 37°C for 2hrs. Capsules were visualized with Maneval stain under 1,000X magnification.

AtxA1 and AtxA2 control of toxin expression

To evaluate the roles of AtxA1 and AtxA2 in anthrax toxin expression, we grew G9241 and the atxA mutants to late-exponential growth phase in toxin-inducing conditions (5% CO₂, 0.8% sodium bicarbonate) and determined transcript levels with qRT-PCR. We elected to analyze transcript levels of pagA, atxA1, and atxA2 only among strains grown under toxin-inducing conditions because the abundance of the lef, cya, and pagA2 transcripts in G9241 was over 10-fold lower than the transcript level for pagA and varied among experiments. Expression of pagA was reduced 19-fold and 58-fold in ΔatxA1 and ΔatxA1ΔatxA2, respectively, compared to G9241, while deletion of atxA2 did not affect pagA transcription (Fig. 3A). Complementation of ΔatxA1ΔatxA2 with either atxA1 or atxA2 restored the amount of pagA transcript to an amount that was similar to that of G9241 (Fig. 3A). To determine whether deletion of atxA1 or atxA2 altered the transcription of the other allele in toxin-inducing conditions, we assessed atxA1 and atxA2 transcript levels in the atxA mutants (Fig 3A). The levels of atxA1 transcript in G9241, ΔatxA2, and ΔatxA1ΔatxA2::atxA1 were similar, while the amount of atxA2 transcript was similar between G9241 and ΔatxA1 and increased 8-fold in ΔatxA1ΔatxA2::atxA2 compared to G9241 (Fig. 3A).

Fig. 3.

Analyses of toxin component transcripts and protein production in G9241 and atxA mutants. (A) RNA was extracted from cells grown to late-exponential phase in toxin-inducing conditions. Relative expression of pagA, atxA1, and atxA2 [n=5 for G9241, ΔatxA1, ΔatxA2, ΔatxA1ΔatxA2; n=4 for ΔatxA1ΔatxA2::atxA1 and ΔatxA1ΔatxA2::atxA2]. (B) Supernatants from cells grown under toxin-inducing conditions for 4h were collected, normalized for cell density, and concentrated. Representative Western blots to detect PA, LF, and EF, and PA2. (C) Integrated density of Western blots [n=4 for PA, LF and EF; n=3 for PA2]. Bars represent the mean, and error bars show the standard deviation. # Significantly different from G9241 (P < 0.05); ^ significantly different from all other strains (P < 0.05); @ significantly different from ΔatxA1ΔatxA2 (P < 0.0001) as determined by one-way ANOVA with Tukey’s multiple comparisons post-test.

We next evaluated the toxin component protein levels in supernatants from cultures grown in toxin-inducing conditions. G9241 and ΔatxA2 produced similar amounts of PA, while ΔatxA1 and ΔatxA1ΔatxA2 produced less PA than did G9241 (Fig. 3B). Complementation of ΔatxA1ΔatxA2 with atxA1 or atxA2 in trans restored PA production. LF, EF, and PA2 followed the same pattern of expression as that of PA among the strains (Fig. 3B). PA2 was detected with an antibody raised against a peptide that is unique to PA2, and the antibodies used to detect PA and PA2 did not cross react with the other protein. We previously showed that the anti-PA antibody did not detect PA2 in supernatants from a pBCXO1- strain (Wilson et al., 2011), and we confirmed that the anti-PA2 peptide antibody did not react with supernatants from a G9241 derivative that was cured of pBC210 (data not shown). Densitometry analysis of the Western blots showed that ΔatxA1ΔatxA2::atxA1 produced more of each toxin component than did all other strains (P < 0.05; Fig. 3C). Additionally, ΔatxA1, ΔatxA1ΔatxA2, and ΔatxA1ΔatxA2::atxA2 produced significantly less PA than did G9241 (P < 0.05). G9241 produced significantly less EF than ΔatxA1ΔatxA2::atxA1 (P < 0.0001) and significantly more EF than the other strains (P < 0.05; Fig 3C). Of note, complementation of ΔatxA1ΔatxA2 with atxA2 resulted in an intermediate toxin production phenotype with levels between those observed for G9241 and the ΔatxA1ΔatxA2 mutant. Taken together, these data indicate that toxin production is primarily regulated by AtxA1.

Virulence of atxA deletion mutants in mice

To determine whether the altered capsule and toxin phenotypes of the atxA mutants affected G9241 virulence, we inoculated mice via either subcutaneous (s.c.) or intranasal (i.n.) routes with G9241, ΔatxA1, ΔatxA2, or ΔatxA1ΔatxA2 spores. Mice were monitored for morbidity and mortality, and the 50% lethal dose (LD₅₀) was determined for each strain (Table 2). We first utilized A/J mice, which are C5 deficient and have reduced macrophage and neutrophil recruitment to the sites of infection. These mice are also susceptible to low doses of toxigenic but nonencapsulated B. anthracis strains (Cinader et al., 1964, Welkos & Friedlander, 1988, Welkos et al., 1989, Wetsel et al., 1990). In A/J mice, the LD₅₀s for s.c. challenge with G9241 and ΔatxA1 were 7 and 24 spores, respectively, while the LD₅₀ for mice infected with ΔatxA2 was less than 20 spores (100% of mice inoculated with 20 spores succumbed to infection). For i.n. infection of A/J mice, the LD₅₀s for G9241 and ΔatxA2 were both 3 × 10³ spores, and the LD₅₀ for ΔatxA1 was 8 × 10³ spores. The ΔatxA1ΔatxA2 mutant was avirulent when administered s.c. or i.n. to A/J mice at doses up to 10⁷ spores. The median time-to-death for mice challenged s.c. with 100 spores of ΔatxA1 was about 2 days later than that for mice given the same dose of G9241 (Fig. 4A). We observed no differences in survival time among mice that were inoculated s.c. with 10² G9241 or ΔatxA2 spores, or mice that were inoculated i.n. with 10⁴ G9241, ΔatxA1, or ΔatxA2 spores (Fig. 4A).

Table 2.

LD₅₀ values for mouse infection with G9241 or isogenic atxA mutants

	A/J mice		C57BL/6 mice
Strain	s.c. LD50	i.n. LD50	s.c. LD50	i.n. LD50
G9241	7 a	3 × 103 a	44 a (16 – 100)	4 × 105 a (2 × 105 – 9 × 105)
Δ atxA1	24 a (7 – 36)	8 × 103 a	800 a, c (200 – 3 × 103)	8 × 106 b
Δ atxA2	<20	3 × 103 a (1 × 103 – 6 × 103)	300 a, c (100-900)	2 × 106 a (7 × 105 – 9 × 107)
Δ atxA1 Δ atxA2	>107	>107	-	-

^aLD₅₀ calculated by probit analysis with 95% confidence interval, if given by SPSS software, in parentheses

^bLD₅₀ calculated with Reed and Muench method as criteria for probit analysis were not met (Reed & Muench, 1938)

^cSignificantly different from G9241 LD₅₀ as determined by comparison of relative median potencies (P < 0.05)

Fig. 4.

Median survival of mice infected with G9241 (squares), ΔatxA1 (circles), or ΔatxA2 (triangles). (A) Time until death of A/J mice after infection with 10² spores s.c. or 10⁴ spores i.n. (B) Time until death of C57BL/6 mice after a dose of 10² spores s.c. or 10⁷ spores i.n. Doses are equivalent to 10-times the LD₅₀ of G9241. Bars represent the median and mice that survived the challenge were assigned a time to death of 14 days (termination of the experiment). * Significantly different from G9241 (P < 0.05) as determined by Kruskal-Wallis with Dunn’s multiple comparisons post-test.

We also assessed the virulence of G9241, ΔatxA1, and ΔatxA2 spores in the more anthrax-resistant C57BL/6 mice. These immunocompetent mice are 100-fold less susceptible to toxigenic B. anthracis strains that do not produce capsule, but are still susceptible to low doses (<20 spores) of fully virulent B. anthracis strains that express both toxin and capsule (Welkos et al., 1986). The LD₅₀s for ΔatxA1 and ΔatxA2 administered s.c. were 800 and 300 spores, respectively, and were significantly higher than the G9241 LD₅₀ of 44 spores (Table 2). The LD₅₀s after i.n. inoculation of C57BL/6 mice were 8 × 10⁶ spores for ΔatxA1 and 2 × 10⁶ spores for ΔatxA2, both of which were higher than the G9241 LD₅₀ of 4 × 10⁵ spores. Nevertheless, neither comparison reached statistical significance (Table 2). C57BL/6 mice that were administered 10⁷ ΔatxA1 spores i.n. had a significant delay in time until death compared to mice that received the same dose of G9241 spores (Fig. 4B). However, there was no difference in the survival time of C57BL/6 mice challenged with G9241 compared to those that were given ΔatxA1 spores s.c. or ΔatxA2 spores s.c. or i.n. (Fig. 4B). Taken together, our mouse virulence data indicate that at least one AtxA ortholog is required for G9241 to cause disease and that AtxA1 may play a larger role than AtxA2 in G9241 virulence.

Production of capsule by atxA mutants in vivo

To evaluate capsule production in vivo, we inoculated mice s.c. with spores and collected abscesses from the infection site of moribund mice that were humanely euthanized. Bacteria were isolated from the abscesses and treated with hyaluronidase. We observed a large halo that was indicative of HA capsule production prior to treatment (Fig. 5A, 5C) and the compact halo suggestive of the TS capsule after hyaluronidase treatment (Fig. 5E, 5G) on bacteria isolated from mice infected with either G9241 or ΔatxA2 spores. Bacteria from the abscesses of the ΔatxA1-infected mice produced only TS capsule, as evidenced by the consistent capsule size with and without hyaluronidase treatment (Fig.5B, 5F). The bacteria isolated from abscesses after infection with ΔatxA1ΔatxA2 made no capsule, as expected (Fig. 5D). These data are in agreement with our in vitro findings that AtxA1 is required for HA capsule production and either AtxA1 or AtxA2 is necessary for TS capsule expression (Fig. 2).

Fig. 5.

Capsule production by G9241 and atxA mutants in vivo. Capsule production by bacteria isolated from abscesses of mice infected s.c. with 10⁴ G9241 (A, E), ΔatxA1 (B, F), ΔatxA2 spores (C, G), or 10⁷ ΔatxA1ΔatxA2 spores (D). Abscesses were isolated when mice were moribund or 6 h post-inoculation for ΔatxA1ΔatxA2-infected mice. Bacteria were treated with water (A-D) or 1,000U hyaluronidase (E-G) for 24 h and capsule was visualized with Maneval stain under 1,000X magnification.

Activity of AtxA1 and AtxA2

We used a reporter assay to assess the capacity of AtxA2 to act as a transcriptional regulator, as was done previously for AtxA in B. anthracis (Hammerstrom et al., 2011). We measured the activity of AtxA1-His and AtxA2-His (expressed episomally) in cultures of UT376, a B. anthracis ΔatxA reporter strain that carries the AtxA-regulated promoter Plef fused to lacZ. Our previous investigations showed that addition of the 6xHis affinity tag to the carboxy-terminus of AtxA did not affect activity of AtxA (Hammerstrom et al., 2011). Expression of AtxA1 and AtxA2 was induced to similar levels with IPTG (30 μM and 100 μM, respectively) as shown by Western blot probed with an antibody against the 6xHis tag (Fig. 6B). We found that the β-galactosidase activity in the AtxA2-His expression strain was approximately five-fold lower than the activity in the AtxA1-His expression strain (Fig. 6A). These data indicate that AtxA1 enhances lef transcription to a significantly greater degree than does AtxA2. Moreover, since expression of AtxA2 at levels comparable to AtxA1 required three-fold more IPTG, we speculate that AtxA2 is less stable than AtxA1 in B. anthracis.

Fig. 6.

Activity of AtxA1 and AtxA2. The β-galactosidase activity of B. anthracis atxA-null strains that harbor the Plef::lacZ reporter and IPTG-inducible atxA alleles was measured after AtxA1-His and AtxA2-His were induced with 30 μM and 100 μM IPTG, respectively. (A) β-galactosidase activity (Miller Units) was measured in cell lysates collected at the transition from exponential to stationary phase of growth. (B) Activity was standardized to the amount of AtxA produced as determined by densitometry analysis of α-His Western blots. (C) The β subunit of RNA-Polymerase (RNAPβ) was used as an internal loading control. The graph shows the mean and standard deviation for three independent experiments. *Significantly different from AtxA1-His (P < 0.05) as determined by Student’s t-test.

Multimerization of AtxA proteins

Previous reports show that AtxA activity requires dimerization of the protein and that mutants that fail to dimerize are inactive (Hammerstrom et al., 2011). To test for homomultimer formation by AtxA2, we used co-affinity purification to detect protein-protein interactions. B. anthracis ANR-1 ΔatxA strains that expressed AtxA1-His, AtxA1-FLAG, AtxA2-His, AtxA2-FLAG, or GFP-FLAG were pooled and lysed in pairs: (1) AtxA2-His and AtxA2-FLAG, (2) AtxA1-His and AtxA1-FLAG, (3) AtxA2-His and GFP-FLAG, and (4) AtxA2-FLAG and GFP-FLAG, as indicated in Fig. 7A. AtxA dimerization has been demonstrated previously, so pair (2) served as a positive control (Hammerstrom et al., 2011). GFP-FLAG was used as a negative control. His-tagged proteins were captured with Ni²⁺-NTA resin, eluted with imidazole, and detected by Western blot. All of the appropriate tagged proteins were present in pooled lysates prior to incubation with the Ni²⁺-NTA resin (Fig. 7A, lanes1-4). Protein complexes that contained a 6xHis tag bound to the Ni²⁺-NTA resin and were eluted with imidazole (Fig. 7A, lanes 5-8). As expected, AtxA1-FLAG co-eluted with AtxA1-His (Fig 7A lane 6). AtxA2-FLAG co-eluted with AtxA2-His, as evidenced by a faint band in the anti-His Western blot (Fig. 7A, lane 5). The AtxA2-FLAG band was much less intense than that of the AtxA1-FLAG band that co-eluted with AtxA1-His (Fig. 7A, lane 5 versus lane 6). These data suggest that AtxA2 can form homomultimers, but that the AtxA2 homomeric interaction is weaker than that of AtxA1 homomultimers.

Fig. 7.

Dimerization of AtxA2. (A) Cultures of B. anthracis atxA-null strains with plasmids that encode IPTG-inducible AtxA1-His, AtxA2-His, AtxA1-FLAG, AtxA2-FLAG or GFP-FLAG were co-incubated as indicated, then co-affinity purified with Ni²⁺-NTA resin. (A) Proteins present in the mixed lysates prior to (Load, lanes 1-4) and after purification (Eluate, lanes 5-8) were subjected to SDS-PAGE and Western blot with α-His and α-FLAG antibodies as indicated. Arrows indicate the predicted size of AtxA1-His or AtxA2-His [AtxA1/2-His]; AtxA1-FLAG or AtxA2-FLAG [AtxA1/2-FLAG]; and GFP-FLAG. (B, C) His-tagged AtxA, AtxA-H379D, and AtxA2-His proteins were induced by IPTG in a B. anthracis atxA-null strain. Lysates were incubated without (B) and with (C) the crosslinking agent BMH. Proteins were detected with anti-His antibody. Arrows indicate the predicted size of monomers (B) and dimers (C) of AtxA proteins.

As a parallel approach to explore AtxA2 homomultimerization, we employed protein crosslinking with bis(maleimido)hexane (BMH). BMH reacts specifically with free cysteines within 13 Å to irreversibly link the residues. AtxA1 has cysteines at positions 96, 161, 202, 356, 370, and 402, and crosslinking by BMH at C402 results in homomeric protein dimers (Hammerstrom et al., 2011). AtxA2 has the same cysteine residues as AtxA1, except for C161, which is an alanine residue. Cell lysates from individual cultures that expressed 6xHis-tagged recombinant AtxA1, AtxA-H379D, and AtxA2 were treated with BMH, subjected to SDS-PAGE, and probed with an anti-His antibody via Western blot. In the absence of the crosslinker, AtxA migrated as a dense band near the 50-kDa marker in each of the lysates (Fig. 7B). The predicted molecular weights of the three AtxA variants is 55.6 kDa. The crosslinking of two AtxA1 proteins was indicated by a band of the size commensurate with an AtxA1 dimer (110 kDa) detected in the AtxA1 lysate treated with BMH (Fig. 7C). The H379D amino acid mutation in AtxA abolishes dimerization (Hammerstrom et al., 2015), and only monomer was detected in the BMH-treated lysate from the AtxA H379D mutant (Fig. 7C). Similarly, BMH treatment of the AtxA2 lysate did not produce a band suggestive of a dimer. Smears of apparent high molecular weight cross-reactive protein were present in the cross-linked AtxA1 and AtxA2 lysates, an observation that suggests that each AtxA protein has the capacity to form some higher-ordered protein-protein interactions (Fig. 7C).

The capacity for AtxA1 and AtxA2 to form heterodimers was also investigated with co-affinity purification. B. anthracis ANR-1 ΔatxA strains that expressed AtxA1-His, AtxA1-FLAG, AtxA2-His, or GFP-FLAG episomally were pooled and lysed in pairs as depicted in Fig. 8. Lysates were incubated with Ni²⁺-NTA resin and eluted with imidazole, then separated by SDS-PAGE and analyzed by Western blot. As expected, AtxA1 multimerization was demonstrated by detection of AtxA1-FLAG in AtxA1-His eluates (lane 6). GFP-FLAG was not detected in any eluates, which indicated that non-specific proteins were washed from the Ni²⁺-NTA resin and that the 6xHis and FLAG tags did not interact. The presence of AtxA1-FLAG in AtxA2-His eluates revealed that, in these conditions, AtxA1-FLAG formed a stable interaction with AtxA2-His. Taken together, the BMH crosslinking and co-affinity purification data suggest a model in which AtxA2 forms weak or unstable homomultimers relative to AtxA1 homomultimers, and AtxA2 forms a relatively stable heteromultimer with AtxA1.

Fig. 8.

Multimerization of AtxA1 and AtxA2. Cultures of B. anthracis atxA-null strains that contain plasmids that encode IPTG-inducible AtxA1-His, AtxA2-His, AtxA1-FLAG, or GFP-FLAG were mixed as indicated and co-affinity purified with Ni²⁺-NTA resin, then subjected to SDS-PAGE and Western blot. Proteins present in the mixed lysates prior to (Load, lanes 1-5) and after purification (Eluate, lanes 6-10) were detected by Western blots with α-His and α-FLAG antibodies as indicated. Arrows indicate the predicted size of AtxA1-His or AtxA2-His [AtxA1/2-His]; AtxA1-FLAG; and GFP-FLAG.

Discussion

In this study, we investigated the roles of AtxA1 and AtxA2 in gene regulation and virulence of B. cereus G9241. AtxA1 is identical to the B. anthracis virulence regulator AtxA that positively affects transcription of the anthrax toxin genes, the capsule biosynthesis operon, and multiple other genes on the B. anthracis plasmids and chromosome (Bourgogne et al., 2003, Drysdale et al., 2005). Here, we found that AtxA1 and the paralogous protein AtxA2 regulate expression of the toxin components and the capsules in B. cereus G9241, although they exhibit clear differences in activity (Fig. 9). These activity differences could be attributed to changes to the primary, and, consequently, tertiary structure of AtxA2 compared to that of AtxA1.

Fig. 9.

Model of regulation by AtxA1 and AtxA2 in G9241. A solid line denotes a positive regulation of that virulence factor by AtxA1 and/or AtxA2. A dotted line indicates that regulation of toxins by AtxA2 resulted from overexpression of AtxA2.

A linear representation of the five domains of AtxA and a comparison of the AtxA1 and AtxA2 amino acid sequences is shown in Fig. 10. There are 37 amino acids that are not conserved between AtxA1 and AtxA2, only four of which are in the amino-terminal DNA-binding domains of the protein and, as such, might affect target specificities. The remainder of the non-conserved amino acids are located in the PRDs and the EIIB-like multimerization domain. Notably, the phosphorylation sites of AtxA, H199 of PRD1 and H379 of PRD2, that are known to affect AtxA activity (Tsvetanova et al., 2007, Hammerstrom et al., 2015) are conserved in AtxA1 and AtxA2 (highlighted in yellow in Fig. 10). In the current model for AtxA structure and function, phosphorylation of H199 enhances AtxA activity, while phosphorylation at H379 prevents dimerization and abrogates AtxA function (Hammerstrom et al., 2015). The crystal structure of AtxA shows that several amino acids interact at the dimer interface, and, with the exception of the asparagine to serine change at position 389, these residues are conserved among the proteins (highlighted in pink in Fig. 10). Nevertheless, despite this conservation of residues, we were unable to detect robust homomultimerization of AtxA2 with either co-affinity purification or crosslinking. Amino acid differences in AtxA2 may result in conformational changes that affect accessibility of the histidine residues for phosphorylation and/or alter the positions of residues that interact at the dimer interface and lead to reduced dimerization and activity of AtxA2. It is also possible that a slight alteration in the AtxA2 structure changes the position of C402, a residue that is required for BMH-mediated crosslinking of two AtxA molecules (Hammerstrom et al., 2011). Our data also suggest that AtxA2 may be less stable than AtxA1. When atxA1 and atxA2 expression were controlled by the same IPTG-inducible promoter, three-fold more IPTG was required to induce AtxA2 to a level comparable to AtxA1. Finally, it is also possible that, for reasons not yet clear, dimerization is not essential for AtxA2 function.

Fig. 10.

AtxA domains and alignment of AtxA proteins. (A) Linear representation of AtxA proteins. AtxA has 2 helix-turn-helix (HTH) domains, 2 phosphoenolpyruvate-dependent phosphotransferase system regulator domains (PRD), and an EIIB-like domain. (B) Alignment of AtxA from B. anthracis (Ba AtxA) with AtxA1 and AtxA2 from B. cereus G9241. Domains are underlined, conserved H199 and H379 residues are highlighted in yellow, and residues that interact at the dimer interface are highlighted in pink. Non-conserved amino acid changes in AtxA2 are highlighted in gray.

AtxA1 is the predominant regulator of toxin component gene transcription in G9241, while both AtxA1 and AtxA2 can both positively influence capsule production (Fig. 9). In fact, AtxA2 only affected toxin component production when it was overexpressed in the absence of atxA1, and toxin component production in ΔatxA1ΔatxA2::atxA1 was greater than that of ΔatxA1ΔatxA2::atxA2 despite the higher level of atxA2 transcript relative to that of atxA1. These results are in agreement with our Plef:lacZ reporter data that showed that AtxA2 had approximately five-fold less activity than AtxA1. AtxA1 positively regulated transcription of hasACB to promote production of the HA capsule, while AtxA1 and AtxA2 both enhanced transcription of the bps capsule locus that is required for TS capsule synthesis. The amount of atxA2 transcript was higher in ΔatxA1 than in G9241 in capsule-inducing conditions but not in toxin-inducing conditions. Notably, we did not detect capsule on bacteria that were grown in the toxin-inducing cultures (i.e. without serum), and, while toxin is likely produced in the capsule-inducing cultures, the high concentration of serum in the cultures obscured detection of the toxin components in supernatants. These results indicate a possible compensatory mechanism for the loss of AtxA1 or suggest that in cultures with high concentrations of serum, AtxA1 has a negative effect on atxA2 transcription.

The toxin and capsule phenotypes of the G9241 atxA mutants are consistent with previous reports of B. cereus Serovar anthracis strains that were isolated from great apes in Cameroon. The great ape strains contain pBCXO1 from pathogenic B. cereus and a pXO2-like plasmid called pBCXO2; consequently, these strains produce both the HA capsule and the classical B. anthracis poly-D-glutamic acid (PDGA) capsule as well as the anthrax toxin components (Klee et al., 2006, Klee et al., 2010, Brezillon et al., 2015). A derivative of one of these strains with a spontaneous deletion mutation in atxA did not produce PA or LF. Additionally, transcription of both capsule operons was positively influenced by AtxA1, whether directly (hasACB) or indirectly via activation of the transcriptional regulators AcpA and AcpB (capBCADE) (Brezillon et al., 2015). G9241 lacks pXO2, and no proteins with homology to AcpA and AcpB have been identified in the G9241 genome. Whether the bps promoter(s) are directly regulated by AtxA1 and/or AtxA2 or are regulated by a secondary protein remains unknown. A consensus sequence for AtxA-mediated transcriptional control has not been identified, and specific DNA-binding has not been demonstrated for any AtxA-regulated gene. Therefore, we cannot discern meaningful differences in the promoter regions of the regulated genes.

We used two well-established murine models of B. anthracis virulence to evaluate the contribution of each AtxA to G9241 disease. A/J mice have reduced capacity for macrophage and neutrophil recruitment to the site of infection as a result of a complement component C5 deficiency (Cinader et al., 1964, Wetsel et al., 1990). C57BL/6 mice are used as an immunocompetent model of B. anthracis disease and are less susceptible to nonencapsulated strains of B. anthracis. We also utilized two challenge methods to represent two forms of human disease: s.c. challenge for cutaneous infection and i.n. challenge to represent inhalational/pneumonia infections with these unique B. cereus strains. We previously reported that the LD₅₀s for G9241 spores isolated after growth on solid modified germination (G) media and administered s.c. or i.n. to A/J and C57BL/6 mice were similar to those of B. anthracis Sterne strain 34F2 (pXO1+ pXO2-) (Wilson et al., 2011). We changed our method of spore preparation to a liquid Difco sporulation medium (DSM) growth protocol that is more similar to that used by others in the field (Stojkovic et al., 2008) and found that the LD₅₀s for G9241 spores administered s.c. or i.n. to A/J and C57BL/6 were more similar to those for toxigenic and encapsulated B. anthracis (pXO1+ pXO2+) than those for toxigenic but nonencapsulated B. anthracis Sterne 34F2 [Table 2, (Wilson et al., 2011)]. These data underscore the importance of standardized spore growth and isolation procedures within the field.

The role of toxin in infection of toxin-susceptible A/J mice has been investigated with capsule negative strains of B. anthracis (Sterne). When Sterne strain 7702 in which only 40% of the inoculum could produce LF was administered s.c. to the ears of A/J mice, there was no change in germination at the site of infection and dissemination in the host (Lowe et al., 2014). However, when the percentage of LF+ clones in the inoculum was reduced to 10% or 0.3%, defects in dissemination from the ear (10%) and colonization of the ear (0.3%) were observed (Lowe et al., 2014). That study showed not only that LF is necessary for colonization and dissemination by B. anthracis but also that there is a threshold of LF production that is required for disease progression (Lowe et al., 2014). Our observation of an increased time until death for A/J mice infected with ΔatxA1 spores supports the findings of Lowe et al. While ΔatxA1 produced less toxin than did G9241, the strain still produced the TS capsule; we hypothesize that the presence of capsule protects the vegetative bacteria from clearance until the strain produces sufficient toxin to disseminate and cause lethal disease.

Previous investigations about the importance of toxins and capsule(s), individually, on the virulence of B. anthracis and the anthrax-like B. cereus strains in more resistant mouse strains are contradictory. When the great ape strains that contain pBCXO1 and pBCXO2 were administered to outbred mice, deletion of the hasACB locus had no effect on virulence when the capBCADE operon was present. However, when the hasACB locus was deleted and pBCXO2 was cured, the toxigenic, but nonencapsulated, mutant was severely attenuated in mice after s.c. challenge with an LD₅₀ similar to that of Sterne strain 7702. Further, a spontaneous atxA mutant in the pBCXO1+/pBCXO2- strain produced neither the HA capsule nor the toxin components and was avirulent (Brezillon et al., 2015). Oh et al. showed that a G9241 ΔpagA1 strain administered via intraperitoneal injection to C57BL/6 mice was attenuated and that a nonencapsulated G9241 mutant was avirulent (Oh et al., 2011). In our studies, both ΔatxA1 and ΔatxA2 were attenuated in C57BL/6 mice compared to G9241 after s.c. or i.n. inoculation. Similarly, the strain that had reduced toxin production, ΔatxA1, was more attenuated than ΔatxA2, which produced wild-type levels of the toxin components.

Overall, we demonstrate that AtxA1 is the predominant regulator of toxin and HA capsule expression in G9241. The paralogous regulator AtxA2 has reduced activity when compared to AtxA1, but, like AtxA1, can positively control bps operon transcription. Our data suggest a model in which AtxA1/AtxA1 homomultimers have the highest activity and AtxA2/AtxA2 homomultimers have the lowest activity; we propose that AtxA1/AtxA2 heteromultimers have an intermediate activity that does not have a dramatic impact on AtxA-mediated transcriptional regulation due to low abundance and/or stability of the heteromultimer. These data expand our insight into the AtxA family of regulators and their roles in virulence gene expression. Additionally, our study demonstrates that the B. anthracis paradigm of AtxA regulation of virulence is applicable to anthrax-like B. cereus strains, but other factors likely contribute as well. With the continued emergence of B. anthracis-like B. cereus strains, it is important to investigate the virulence networks that are unique to these emerging pathogens.

Experimental Procedures

Bacterial strains, culture conditions, and spore preparation

The bacterial strains and plasmids used in this study are shown in Table 1. For routine growth, B. cereus, B. anthracis, and E. coli strains were cultivated in Luria-Bertani (LB) broth (BD Biosciences, East Rutherford, NJ) overnight at 30°C or 37°C and 225 rpm. To induce capsule expression, an overnight culture of bacteria grown in heart infusion (HI) broth (BD Biosciences) was diluted 1:1,000 in HI broth supplemented with 50% heat-inactivated horse serum (Life Technologies, Carlsbad, CA) and 0.8% sodium bicarbonate; cultures were incubated at 37°C in 14% CO₂ without aeration. To induce toxin expression, cultures grown overnight at 37°C in HI broth were diluted 1:1,000 into HI broth supplemented with 0.8% sodium bicarbonate and then incubated at 37°C and 5% CO₂ with aeration. The following antibiotics were added to bacterial cultures as needed: ampicillin (100 μg ml⁻¹), chloramphenicol (20 μg ml⁻¹), kanamycin (50 μg ml⁻¹), or spectinomycin (150 μg ml⁻¹) for E. coli, and kanamycin (100 μg ml⁻¹), spectinomycin (200 μg ml⁻¹), or erythromycin (5 μg ml⁻¹) for B. cereus.

For spore preparation, each strain was grown overnight in 25 ml brain heart infusion (BHI) media at 37°C with shaking. Either 500 ml or 1 L of Difco Sporulation Media [DSM, 0.8% nutrient broth, 0.1% KCl, 0.012% MgSO₄, 6 mM NaOH, 1 mM Ca(NO₃)₂, 0.01 mM MnCl₂, and 1 μM FeSO₄] was seeded with the 25 ml overnight culture, then cultivated at 30°C for 5 days with aeration (Schaeffer et al., 1965). When the spore density in the culture reached ~95%, the spores were isolated by centrifugation at 10,000 × g for 10 min, then washed three times with sterile water. Spores were resuspended in sterile water, heat-killed at 65°C for 1 h to remove any residual vegetative bacteria, and stored at 4°C until use.

Plasmids and generation of G9241 mutant strains

Deletion mutants were constructed by either allelic exchange mutagenesis in which the gene of interest was replaced with an antibiotic resistance cassette or by markerless deletion mutagenesis with a temperature-sensitive plasmid. A complete description of mutagenesis can be found in the Supporting Information. A list of strains and plasmids used for mutagenesis and primers used can be found in Tables S1 and S2, respectively. Briefly, to generate the vectors for insertional mutagenesis, an Ω-kanamycin or Ω-spectinomycin cassette was cloned between ~1 kb PCR products that flank the 5’ and 3’ ends of the genes of interest in pGEM-T, subcloned into pUTE583, and introduced into G9241 as described previously (Vergis et al., 2013). For markerless deletion, the ~1 kb PCR products that flank the 5’ and 3’ ends of the gene of interest were cloned into pGEM-T, subcloned into pJMS1, and introduced into G9241 as described in the Supporting Information. The genes for atxA1 and atxA2 with their respective native promoters were cloned into pCR-Blunt II-TOPO (Life Technologies) and then subcloned into pHT304 for complementation of ΔatxA1ΔatxA2.

All mutant and complemented strains were screened by PCR to ensure that the pBCXO1 and pBC210 plasmids remained intact, as described previously (Wilson et al., 2011). In addition to fabG, bclA (P14/P15) was also used as a chromosomal gene. Since introduction of pUTE583 caused the loss of atxA1 on pBCXO1 in some cases, primers specific for atxA1 (P16/P17) were added to the primer screening panel; only those mutant strains that retained otherwise intact virulence plasmids were used. Mutations were also confirmed by Southern blot with a DIG-labeled probe, according to the manufacturer’s guidelines (Roche, Pleasanton, CA).

RNA isolation and qRT-PCR

Bacteria grown under capsule-inducing or toxin-inducing conditions were harvested at late-exponential phase by centrifugation at 35,000 × g or 10,000 × g, respectively, for 15 min. Each cell pellet was resuspended in 1 ml RNAzol^® RT (Sigma-Aldrich, St. Louis, MO) and placed in a 1.5 ml screw cap tube that contained 0.2-0.3 ml of 0.1 mm RNase-free glass beads (Next Advance, Averill Park, NY). Cells were lysed with a Turbo Mixer vortex attachment for 4-6 min, with repeated 30 sec cycles of agitation and rest on ice. RNA was isolated with the total RNA isolation protocol for RNAZol RT (Molecular Research Center, Inc., Cincinnati, OH). The RNA was precipitated by addition of sodium acetate (pH 5.2) to a final concentration of 0.3 M and two volumes of 100% ethanol. RNA was incubated on dry ice for 15 min and pelleted by centrifugation at 16,000 × g for 30 min. The RNA pellet was washed with 70% ethanol and resuspended in TE buffer. DNA contamination in RNA samples was digested with the Turbo DNA-free kit (Life Technologies).

For qRT-PCR, RNA was converted to cDNA with the Quantitect Reverse Transcription kit (QIAGEN, Germantown, MD). The cDNA from capsule-inducing cultures was amplified by qRT-PCR with primers specific for 16S rRNA (P18/P19), hasA (P20/P21), bpsA (P22/P23), atxA1 (P16/P24), and atxA2(P25/P26) with the Rotor-Gene SYBR Green PCR kit in a Rotor-Gene Q (QIAGEN). The cDNA from toxin-inducing cultures was amplified by qRT-PCR with primers specific for 16S rRNA, pagA (P27/P28), atxA1, and atxA2. Primer sequences can be found in Table S2. The ΔCT values were calculated by subtraction of the CT value for 16S rRNA, the internal control, from the CT value for the gene of interest. The log₁₀ relative expression [2^(−ΔCT)] was calculated for each gene of interest. GraphPad Prism 6 (GraphPad Software, La Jolla, CA) was used to conduct one-way ANOVA with Tukey’s multiple comparisons post-test to identify significant differences among strains for each gene of interest.

Hyaluronidase treatment and capsule visualization

Strains were cultivated in capsule-inducing conditions for 24-26 h, fixed by the addition of formaldehyde to 10% (v/v), and stored at 4°C. Fixed capsule cultures were washed three times with PBS to remove the formaldehyde and resuspended in PBS to the initial volume. A 50 μl sample of each culture was treated with 200 U of hyaluronidase (Sigma-Aldrich) or water at 37°C for 2 h. After treatment, hyaluronidase was removed by centrifugation at 16,000 × g for 10 min. The pellet was resuspended in the original sample volume. The capsules were stained with the Maneval stain (Maneval, 1941). Briefly, 8 μl of the bacterial suspension was added to 4 μl of 1% Congo red on a slide and smeared. The slides were allowed to air dry, then Maneval stain (Carolina Biological Supply Co., Burlington, NC) was applied to the dried smear. After 5 min, the Maneval stain was rinsed from the slide with water and the slide was allowed to air dry. Slides were viewed on an Olympus BX60 microscope under 1,000X magnification, and images were captured with a SPOT Insight Gigabit Color Camera (SPOT Imaging Solutions, Sterling Heights, MI).

Western blot for toxin components

Supernatants were collected from toxin-inducing cultures grown to late-exponential phase. Each supernatant was carefully removed from the cell pellet and filter-sterilized with a 0.22 μm filter (Corning, Corning, NY). The supernatants were adjusted to the same OD₆₀₀, at least an equal volume of ice-cold ethanol was added, and the samples were placed at −20°C overnight. Precipitated protein was collected by centrifugation at 12,000 × g for 15 min at 4°C. The supernatant was discarded and the protein pellet was washed with ethanol, air dried, and resuspended in 400 – 500 μl 4x SDS-PAGE loading buffer [106 mM Tris HCl, 141 mM Tris Base, 2% SDS, 10% glycerol, 0.51 M EDTA].

Samples were denatured at 98°C for 10 min and separated in a 4-12% Bis-Tris gel (Life Technologies) in MOPS buffer (Teknova, Hollister, CA) at 125V. Total protein in gels was detected with Oriole stain (Bio-Rad, Hercules, CA). For Western blot analysis, the proteins were transferred to nitrocellulose membranes with the iBLOT Dry Blotting System (Life Technologies). The membrane was blocked with PBS supplemented with 0.1% Tween-20 and 5% skim milk (PBS-T-SM) for at least 1 h at room temperature. Primary antibodies were incubated overnight at 4°C in PBS-T-SM. PA and LF were detected with mouse monoclonal antibodies (GenWay, San Diego, CA), diluted 1:10,000; EF was detected with rabbit polyclonal sera diluted 1:1,000 (kindly provided by Dr. Stephen Leppla); and PA2 was detected with rabbit antibody that was generated against a peptide unique to PA2, diluted 1:1,000. The PA2 peptide antibody was produced by GenScript (Piscataway Township, NJ) against peptide RSANMNREIVDEDNC. Blots were washed three times with PBS supplemented with 0.1% Tween 20 (PBS-T). Secondary antibody, goat anti-mouse Ig or goat anti-rabbit Ig conjugated to horseradish peroxidase (Bio-Rad), was diluted 1:10,000 in PBS-T and applied to the membrane at room temperature for at least 1 h. The nitrocellulose membrane was washed once with PBS-T and twice with PBS. Protein was detected with ECL reagent (GE Healthcare) in an ImageQuant LAS4000 (GE Healthcare).

We used ImageQuant TL software to quantify total protein in the Oriole-stained gels and toxin levels in the Western blots of supernatants. A normalization value for each concentrated toxin supernatant was computed by calculation of the total intensity of the 20-60 kDa range in each lane of an Oriole-stained gel. Only the protein bands between 20-60 kDa were used for normalization so as to exclude the 60-100 kDa region that includes the toxins. The volume of each well over the same area was determined in ImageQuant and divided by the area to get relative intensity. A normalization value was determined from the relative intensity of the Oriole-stained wells to adjust for differences in protein loading. The relative intensity was calculated in the same manner for the bands detected in Western blots, and the same value from an empty well was subtracted from the relative intensity of each band to control for the background of each blot. The intensity was then multiplied by the normalization factor determined by the Oriole stain to calculate the integrated density for each reactive band. Significant differences among integrated densities were determined by one-way ANOVA with Tukey’s multiple comparisons post-test in GraphPad Prism 6.

Mouse infection and abscess collection

All mouse infections were completed under ABSL-2 conditions with approval from the Institutional Animal Care and Use Committee of the Uniformed Services University of the Health Sciences. Six-week-old female A/J or C57BL/6 mice (Jackson Laboratories, Bar Harbor, ME) were housed in filter-top cages with access to standard food and water ad libitum. Groups of 5-10 mice each received either an i.n. or s.c. inoculation of different doses of spores or water as described previously (Wilson et al., 2011). Briefly, for i.n. infection, mice were lightly anesthetized by isoflurane inhalation and a total of 50 μl of spores was introduced onto both nares. For s.c. infection, 100 μl of spores was injected on the right flank below the foreleg. Mice were closely monitored for morbidity and mortality for 14 days post-infection, and moribund animals were humanely euthanized. Both mouse strains were challenged with each bacterial strain via both routes in at least two independent experiments. The LD₅₀ values were calculated with probit analysis in SPSS software, version 22.0 (IBM, Armonk, NY), when applicable. If criteria for probit analysis were not met, the LD₅₀ values were calculated by the method of Reed and Muench (Reed & Muench, 1938). Significant differences among the LD₅₀ values were determined by comparison of the relative median potency values in SPSS software. If the 95% confidence interval of the relative median potency between two groups did not contain the value one, the LD₅₀s were statistically different. Significant differences in time until death of mice after infection with spores were determined by Kruskal-Wallis with Dunn’s multiple comparisons post-test.

Abscesses were collected from mice inoculated by s.c. injection with G9241, ΔatxA1, ΔatxA2, or ΔatxA1ΔatxA2 spores. Mice were euthanized by inhalational overdose of isoflurane followed by cervical dislocation 6 h after infection with ΔatxA1ΔatxA2 spores and when moribund for mice challenged with the other strains. Abscesses were apparent as thickened skin at the injection site. Each abscess was excised from the underlying skin and mashed with the plunger from a 1 ml syringe into 1 ml PBS to dislodge the bacteria. The abscess skin was removed and a cell pellet was collected by centrifugation at 16,000 × g for 10 min. The cell pellet was resuspended in PBS supplemented with 10% formaldehyde (v/v) to fix the sample and stored at 4°C. For capsule detection, 30 μl of abscess bacteria were washed with PBS, treated with 1,000 U hyaluronidase for 22-24 h, and visualized with Maneval stain as described above.

AtxA activity assay

Activity of AtxA and AtxA2 was quantified with a transcriptional reporter system as described previously (Hammerstrom et al., 2011). Briefly, UT376 is a B. anthracis Sterne ANR1 ΔatxA derivative in which the AtxA-regulated pXO1-encoded lethal factor gene, lef, was replaced with a promoterless lacZ. The atxA1 and atxA2 alleles were engineered to encode a 6xHis tag at the carboxy-terminus and were cloned under control of the IPTG-inducible hyper-spank promoter (Britton et al., 2002) to generate pUTE991 and pUTE1096, respectively. The plasmids were introduced individually into UT376. BHI broth supplemented with spectinomycin (100 μg ml⁻¹) was inoculated with spores from each expression strain (or an empty vector control) and incubated at 30°C with aeration for 8-9 h. Cultures were diluted into 25 ml casamino acid medium supplemented with 0.8% sodium bicarbonate (CA-CO₃) (Thorne & Belton, 1957, Hammerstrom et al., 2011) to an initial OD₆₀₀ of 0.08, then incubated with aeration at 37°C with 5% CO₂. Expression of the atxA1 and atxA2 alleles was induced with 30 (atxA1) or 100 (atxA2) μM IPTG at the early exponential growth phase and samples were collected at the transition to stationary phase. Relative levels of the 6xHis-tagged proteins were determined by Western blot with anti-His antibody (Genscript). Samples with comparable expression levels of AtxA and AtxA2 were assayed for β-galactosidase activity as described previously (Hadjifrangiskou & Koehler, 2008).

Co-affinity purification

AtxA variants were co-affinity purified from pooled lysates as described previously (Hammerstrom et al., 2011). B. anthracis UT376 strains with plasmids that encode AtxA1-His (pUTE991), AtxA2-His (pUTE1096), AtxA1-FLAG (pUTE992), AtxA2-FLAG (pUTE1122), and GFP-FLAG (pUTE1013) were cultivated individually in 25 ml of CA-CO₃ at 37°C with 5% CO₂. Protein expression was induced with IPTG (50-100 μM) at the early exponential growth phase and 20 ml of each culture were collected at the transition to stationary phase. The cultures were pooled as described in Figs. 9 and 10, pelleted by centrifugation, and washed with 10 ml binding buffer [5 mM imidazole pH 7.9, 0.5 M NaCl, 20 mM Tris, pH 7.15, 5 mM β-mercaptoethanol] with EDTA-free Complete Protease Inhibitor Cocktail (PIC; Roche). Cell pellets were flash frozen and stored at −80°C for future use or resuspended immediately in binding buffer-PIC. Cell suspensions were lysed mechanically in a Mini BeadBeater (BioSpec Products, Barlesville, OH) for 2 min at 4°C, and insoluble material was pelleted by centrifugation at 10,000 × g for 5 min at 4°C. The lysates were incubated for 20 min at 37°C with 5% CO₂, then mixed with Ni²⁺-NTA resin (QIAGEN). The resin was washed to remove non-specifically bound proteins. His-tagged proteins, as well as any associated proteins, were eluted with elution buffer [20 mM Tris, pH 7.15, 800 mM imidazole, pH 7.9, 100 mM NaCl, 5 mM β-mercaptoethanol] and analyzed by Western blot with anti-His and anti-FLAG antibodies (GenScript).

BMH crosslinking

B. anthracis strain UT376 with plasmids that encode AtxA-His (pUTE991), AtxA-H379-His (pUTE991), and AtxA2-His (pUTE1096) were cultured and induced with IPTG (50 μM, 50 μM, 100 μM respectively). After 2 h, 20 ml of each culture was collected by centrifugation and washed twice with 5 ml PBS with 10mM EDTA, pH 7.2 (PBS-E). Cells were resuspended in 1 ml PBS-E and lysed by mechanical disruption. Insoluble debris was removed from cell lysates by centrifugation at 10,000 × g for 5 min at 4°C. For each experiment, 250 μl of soluble lysate was mixed with 5 μl of 20 mM bis(maleimido)hexane (BMH, Thermo Scientific, prepared freshly in DMSO) and incubated at 4°C with end-over-end mixing for 2 h. Control reactions without BMH contained DMSO only. Reactions were quenched by the addition of cysteine to a final concentration of 40 mM, and samples were vortexed for 15 min at room temperature. Samples were boiled in SDS loading buffer [5% glycerol, 100 mM DTT, 2% SDS, 40 mM Tris-Cl, pH 6.8] and analyzed on 4-15% polyacrylamide SDS gels (Bio-Rad). AtxA1, AtxA-H379D, and AtxA2 were detected by Western blot with an anti-His antibody.