ABSTRACT
The virulence of Bacillus anthracis, the causative agent of anthrax, stems from its antiphagocytic capsule, encoded by pXO2, and the tripartite toxins encoded by pXO1. The accepted paradigm states that anthrax is both an invasive and toxinogenic disease and that the toxins play major roles in pathogenicity. We tested this assumption by a systematic study of mutants with combined deletions of the pag, lef, and cya genes, encoding protective antigen (PA), lethal factor (LF), and edema factor (EF), respectively. The resulting seven mutants (single, double, and triple) were evaluated following subcutaneous (s.c.) and intranasal (i.n.) inoculation in rabbits and guinea pigs. In the rabbit model, virulence is completely dependent on the presence of PA. Any mutant bearing a pag deletion behaved like a pXO1-cured mutant, exhibiting complete loss of virulence with attenuation indices of over 2,500,000 or 1,250 in the s.c. or i.n. route of infection, respectively. In marked contrast, in guinea pigs, deletion of pag or even of all three toxin components resulted in relatively moderate attenuation, whereas the pXO1-cured bacteria showed complete attenuation. The results indicate that a pXO1-encoded factor(s), other than the toxins, has a major contribution to the virulence mechanism of B. anthracis in the guinea pig model. These unexpected toxin-dependent and toxin-independent manifestations of pathogenicity in different animal models emphasize the importance and need for a comprehensive evaluation of B. anthracis virulence in general and in particular for the design of relevant next-generation anthrax vaccines.
INTRODUCTION
Bacillus anthracis, the causative agent of anthrax, is a Gram-positive aerobic spore-forming bacillus. The classical virulence factors commonly associated with anthrax pathogenesis are encoded by the pXO1 and pXO2 virulence plasmids. The plasmid pXO2 encodes functions required for biosynthesis of the antiphagocytic capsule, composed of poly-d-glutamic acid, while the toxin component genes, i.e., pag coding for protective antigen (PA), lef coding for lethal factor (LF), and cya coding for edema factor (EF), are carried on pXO1 (40). PA represents the cell-binding component required for the entry into target cells of the enzymatic moieties, either LF or EF (10, 29). LF, a zinc metalloprotease, cleaves most isoforms of mitogen-activated protein kinase kinases, disrupting normal cellular signaling pathways in immune cells (39). EF, a calmodulin-dependent adenylate cyclase, elevates the intracellular cyclic AMP (cAMP) concentration, interfering with cellular signaling and membrane permeability regulation (38). B. anthracis binary toxins are thus termed lethal toxin (LT) (LF plus PA) and edema toxin (ET) (EF plus PA). B. anthracis strains cured of pXO1 exhibit complete loss of virulence, indicating the possible essential role of these toxins in pathogenicity (reference 20 and this study), although such strains are reported to retain virulence in mice (41), in which the capsule is of major importance.
The role of the toxins has been studied in a variety of animal models, including mice (15, 28), guinea pigs (37), rabbits, and nonhuman primates (14; for a review, see reference 18). Although nonhuman primates are probably considered the most suitable model for the human anthrax disease, their use is limited because of high cost and the inherent difficulty in performing large experiments necessary for statistically significant results. Therefore, for large-scale studies, guinea pigs and rabbits, which exhibit susceptibility to toxemia and bacteremia, are referred to as the models of choice for anthrax research. The use of the popular murine model is limited by the high sensitivity of mice to the B. anthracis capsule. Therefore, most mouse model research utilizes the attenuated noncapsulated Sterne strain (18).
A genetic approach has been undertaken to determine the contribution of each of the three B. anthracis toxin components of the toxin to virulence in the mouse model by Pezard et al. (32, 33). In that study, a deletion of either the pag, lef, or cya gene in the Sterne noncapsulated (ΔpXO2) toxinogenic strain was generated. The virulence of each of the three different single mutants was determined by subcutaneous (s.c.) infection of mice and compared to that of parental Sterne strain. The results demonstrated that in mice the Δpag strain lost its virulence and was as attenuated as the strain cured of the entire pXO1 plasmid. Similar results were observed for the lef null mutant, whereas the cya mutant was only 10-fold less virulent than the parental strain. The conclusion from these studies was that PA and LF, unlike EF, are major contributors to the pathogenicity of the noncapsulated B. anthracis strain in mice. In contrast, several recent studies with a clinical isolate and the Sterne strain as well as the purified toxin demonstrated the importance of edema toxin in B. anthracis pathogenicity in the murine model (11, 12, 15, 19).
In studies with the fully virulent Vollum strains, it was shown recently that the deletion of either the lef or the cya gene had little or no effect on the virulence of the mutants in guinea pigs and rabbits (25, 26). These findings indicate that a single toxin, either LT or ET, may be sufficient for host colonization and mortality, yet some differences with respect to the roles of the toxins in the pathogenicity of B. anthracis in guinea pigs and rabbits were noted. Furthermore, the roles of the toxins not only may vary in the different animal models but also may depend on the infection route, which activates different colonization processes. For example, in studies addressing the role of additional B. anthracis virulence factors it was shown that following airway infection of guinea pigs the mntA mutant is as virulent as the parental wild-type Vollum strain but that it is at least 10,000-fold less virulent following s.c. infection (17; A. Shafferman unpublished results).
In an attempt to elucidate the role of each of the three components of the toxins in the background of the fully virulent B. anthracis ATCC 14875 Vollum strain, we carried out a systematic genetic study deleting the pag, lef, and cya genes and combinations thereof, followed by evaluation of the effects of the mutations on virulence in rabbits and guinea pigs by two routes of infection: subcutaneous injection and intranasal (i.n.) instillation. We demonstrate that the accepted paradigm of an essential role for the toxins in B. anthracis pathogenicity appears to be confirmed in the rabbit model yet that in the guinea pig model the contributions of the various components of the toxins to virulence are highly complex and other factors may be even more critical for effective pathogenesis. The implications of these results for our understanding of the virulence mechanism in different models and for vaccine development are discussed.
MATERIALS AND METHODS
Bacterial strains, media, and growth conditions.The bacterial strains used in this study are listed in Table 1. In this study, we used the wild-type and fully virulent ATCC 14578 strain (which should not be confused with the attenuated Vollum 1B strain [27]). We can state with sufficient confidence, on the basis of numerous studies performed and published, that in every parameter that we are aware of, the wild-type Vollum ATCC 14578 strain is as virulent as the Ames strain, at least on the basis of 50% lethal doses (LD50s) for different routes of infection tested in various animal models as well as in mean time to death (MTTD). The LD50 for the Ames strain was reported to be in the range of 1 × 104 to 1 × 105 CFU for aerosol infection and about 100 CFU for i.n. or s.c. infections in different models, which are comparable to the data presented in Tables 2 and 3 for our Vollum strain. B. anthracis and Escherichia coli strains were cultivated in Terrific broth (35) at 37°C with vigorous shaking (250 rpm). For the induction of toxins and capsule production, NBY-bicarbonate broth (42) was used. Sporulation was carried out using G broth, as previously described (23). E. coli strains were used for the facilitation of plasmid construction. Antibiotic concentrations used for selection in Mueller-Hinton (MH) agar (Difco) or Terrific broth were as follows: for E. coli strains, ampicillin at 100 μg/ml; for B. anthracis strains, kanamycin at 10 μg/ml and erythromycin at 5 μg/ml.
Bacterial strains, plasmids, and oligonucleotide primers used in this study
Plasmid and strain construction.The plasmids and oligonucleotide primers used in this study are summarized in Table 1. The oligonucleotide primers were designed according to the genomic sequence of the B. anthracis Ames ancestor strain (accession no. NC_007530; NCBI). Genomic DNA (containing the chromosomal DNA and the native plasmids, pXO1 and pXO2) for cloning the target gene fragments was extracted from the Vollum strain as previously described (24). PCR amplifications were performed using the AccuTaq LA (Sigma) systems.
Prior to transformation into the Vollum strain, all plasmids were first propagated in the methylation-deficient E. coli strain GM2929 (Table 1). B. anthracis cells were electrotransformed as described previously (9).
To disrupt the target genes by homologous recombination, a previously described method was used (21, 26). In general, gene deletion was accomplished by a markerless allelic exchange technique that replaced the complete coding region with the SpeI restriction site. At the end of the procedure, the resulting mutants did not contain any foreign sequences, and the only modification was the null mutation of the target gene. The method was used, sequentially, to generate the double and triple mutants. The different mutants were verified by PCR for (i) the deletion of the target gene, (ii) the presence of the other, nonmutated toxin-related genes, and (iii) verification that no major rearrangements occurred in the vicinity of the deletion. All the mutants were tested for their ability to produce the capsule and the relevant toxin component, following incubation in NBY-CO2 growth medium. The capsule was visualized by negative staining with India ink. The presence or absence of toxins was determined by Western blotting using anti-toxin-specific monoclonal antibodies to PA, LF, and EF. The single deletion mutations in the lef or cya genes were previously shown (25, 26) to have little or no effect on the pathogenicity in guinea pigs or rabbits. Deletion of the pag gene was previously described (32) and, as described in this paper, has a relatively small effect on pathogenicity in guinea pigs. Therefore, to confirm the absence of any other unintended mutation/genetic alteration that may affect the phenotypes (minor loss of function), two independent mutants were tested (see Fig. 1).
The stability of the mutations was validated by specific and relevant PCR tests performed on blood cultures collected from each animal that died as a result of infection with a given mutant.
DNA preparation and PCR.DNA was purified from bacterial cultures as described previously (24, 26). For colony screening, each colony was resuspended in 50 μl of sterile double-distilled water (DDW) in a 0.2-ml PCR tube. The tube was then placed in a thermocycler for two cycles of 95°C for 10 min. The tubes were then centrifuged in a minicentrifuge, at maximal speed, for 1 min at room temperature. The clear supernatant was transferred to a clean tube, and 5 μl was used for the PCRs.
All PCRs (25-μl mixtures) were performed in PCR buffer (3.5 mM MgCl2, deoxynucleoside triphosphates [dNTPs] [0.2 mM each], and 0.04 U/μl of TaKaRa Taq DNA polymerase [all from TaKaRa Bio Inc. [R001A]) with ∼2 ng of DNA. The general thermocycling program for the PCR was 94°C for 30 s followed by 40 cycles of 94°C for 1 min, 55°C for 30 s, and 72°C for 1 min and then one cycle of 72°C for 5 min. The PCR products were separated on a 1.1% agarose gel using Tris-borate-EDTA (TBE) as running buffer (35).
Infection of guinea pigs and rabbits.Female Hartley guinea pigs (Charles River Laboratories), weighing 220 to 250 g, or New Zealand White rabbits, weighing 2.2 to 2.5 kg, were used. The animals were infected with spore preparations of either the mutant strains or the parental Vollum strain. Prior to infection of the animals, spore preparations were heat shocked (70°C, 20 min) and serially diluted in saline to produce spore suspensions within the range of 102 to 109 per milliliter. A spore dose of 0.1 ml (guinea pigs) or 1 ml (rabbits) was administered subcutaneously (s.c.) or intranasally (i.n.) to each animal. The remaining spore dose suspensions were plated for total viable counts (CFU/ml). The animals were monitored daily for 21 days or for the indicated period. Upon death, blood samples were plated and DNA was extracted, followed by PCR analysis in order to determine the identity of the strain responsible for the animal's death.
This study was carried out in strict accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals of the National Research Council. The protocols were approved by the Committee on the Ethics of Animal Experiments of the IIBR.
Statistical analysis.The spore lethal dose required to kill 50% of the animals (LD50) was calculated by the method of Reed and Muench (34) (wild type and Δlef or Δcya mutant [25, 26]). For cases where a mutant caused only partial mortality in a given group, the LD50 was estimated from the dose that crossed the 50% lethality line in plots of lethality versus dose (dashed lines in Fig. 2 and 3) and was referred to as “estimated LD50.” The significance of the difference between the estimated LD50s or calculated LD50s was determined using the Fisher test in GraphPad Prism version 5.00 for Windows (GraphPad Software, San Diego, CA).The mean times to death (MTTDs) were calculated as the sum of the days of death of all the animals that succumbed divided by the total number of dead animals. In this type of calculation, no score is given to animals that survived the infection. Analysis of differences between various MTTDs was determined using Mann-Whitney and t tests in GraphPad Prism version 5.00 for Windows.
RESULTS
Generation and validation of toxin deletion mutants.The different mutants of the fully virulent ATCC 14578 Vollum strain (pXO1+ pXO2+) were generated by the method of markerless mutagenesis (see Materials and Methods) (26). Two independent single-deletion toxin gene mutants were isolated, characterized, and then used to generate the various double- and triple-deletion independent mutant progenitors, as described in Fig. 1A. Multiple genotypic and phenotypic analyses were used to verify the correct mutation in each isolate. It should be noted that for the multiple mutants we used different genealogical routes to generate the same mutant type. For example (Fig. 1A), in order to generate the first type of triple mutant we started with the Δcya, then generated Δcya Δlef, and finally into the latter introduced the Δpag mutation (two isolates were picked up each time for analysis). The same type of triple mutant was generated by a different genealogical route whereby the starting mutant was the Δpag mutant, into which we introduced Δlef to generate the Δpag Δlef mutant, which was then used to introduce the Δcya mutation (again in every step, two independent clones were isolated); thus, for the triple mutant we had two pairs of isolates generated by different genealogical routes.
Generation and analyses of the various pag, cya, and lef mutants. (A) Construction strategy for the different single, double, and triple mutants. For each gene, two independent single mutants were generated from the fully virulent Vollum strain (wild type [wt]). Two independent isolates of the Δcya mutants and a single isolate of the Δpag mutants were used to generate the two independent Δcya Δlef double mutants and the Δcya Δpag and Δpag Δlef double mutants. The Δcya Δlef and Δpag Δlef mutants were used to generate the two independent isolates of the Δcya Δlef Δpag and Δpag Δlef Δcya triple mutant. All independently isolated pairs of a given mutant type exhibited identical genotype and phenotype (see the text). Note the different genealogy of the single, triple, and some of the double mutants (see the text). (B) Example of PCR validation analysis: the case of the Δpag mutant. Depicted is an analysis of 3 isolated colonies (designated 1, 2, and 3) which were picked for evaluation from a specific mutagenesis experiment. Primers 1 and 6c represent primers from outside the integration fragments (outside the pag gene in this example). Primers s3 and s4c represent primers inside the coding region of each of the 3 different toxin-related genes coding for PA, LF, and EF. The primer target genes are marked PA-pag, LF-lef, and EF-cya. (Not shown are the results of PCR using primers 1 and 6c of the lef and the cya genes that were used to verify the presence of these genes in the pag mutant isolates 1, 2, and 3.) (C) Phenotypic validation of the different mutants by Western blot analysis of the different toxins secreted into medium following growth under induction conditions, using specific anti-PA, anti-LF, and anti-EF monoclonal antibodies. Data for only one of the two isolates of each mutant are presented, although all isolates of the same genotype manifested the same Western blot profile.
The genotypes were confirmed by PCR analyses of DNA extracted from each of the different mutants. For every mutant generated, PCR analyses verified the absence of any coding sequences of the deleted genes as well as the intactness of neighboring sequences. An example of this type of PCR analysis is depicted in Fig. 1B for construction of the Δpag mutant. In Fig. 1B we demonstrate that for 3 independently isolated mutants, Δpag has the precise deletion (Pa 1-6c and Pa s3-4c) and that the lef and cya genes are intact (LF s3-4c and EF s3-4c, respectively). In addition to the genetic analysis, each mutant was grown under conditions that induce toxin production, and the expression of the relevant toxin component was verified. To this end, we used Western blot analysis of secreted proteins to verify that the phenotypes of the mutants are indeed in line with their expected PCR-genetic profiles (Fig. 1C). This type of analysis was performed on each of the isolates of the various single, double, or triple mutants. Thus, each Δpag mutant expressed the LF and EF, and the independent Δpag Δlef, Δpag Δcya, and Δcya Δlef double mutants expressed EF, LF, and PA, respectively. In in vitro studies we demonstrated that the growth and sporulation profiles as well as capsule production of all mutants were indistinguishable from those observed for the parental wild-type Vollum strain (data not shown). For in vivo studies, in both guinea pigs and rabbits, we tested the two independently isolated mutants from each of the different mutants generated. All these genotypically identical pairs of bacterial isolates showed identical behavior. To ensure the in vivo stability of the mutants and to demonstrate that the inocula of the mutants were indeed pure and that no reversions occurred in the animals, blood cultures were grown from each animal that died and subjected to PCR analysis. All tests of samples taken from these animals validated the genotypes and the stabilities of the input B. anthracis strains used for infection. Altogether, the seven Vollum toxin mutants types were generated: Δpag, Δcya, Δlef, Δpag Δcya, Δpag Δlef, Δcya Δlef, and Δpag Δcya Δlef. These mutant strains were tested side by side with the fully virulent parental wild-type Vollum strain and its fully attenuated pXO1-cured derivative strain, designated Vollum ΔpXO1, as described below.
Role of the toxins in subcutaneous and airway infections in the guinea pig model. (i) Subcutaneous inoculation.We have previously demonstrated (25, 26) that the deletion of either LF or EF had minor effects on the LD50s of the mutants in s.c. infection of guinea pigs (Fig. 2; Table 2). These findings indicate that a single toxin, either LT or ET, may be sufficient for host colonization and mortality; however, they raise additional questions about the role of the toxins in the pathogenicity of B. anthracis. The survival curve for guinea pigs infected with the pag null mutant as a function of infection dose is presented in Fig. 2, and the “estimated LD50” (the dose that intersects the 50% survival line [see Materials and Methods]) and MTTD are presented in Table 2. The estimated LD50 of the pag mutant was about two orders of magnitude higher than that of the wild-type strain. The MTTD was about 1 day longer, but mortality was not complete even at doses 100- to 1,000-fold higher (106 to 107 CFU) than the estimated LD50. The pag mutant is at least 7,000-fold more virulent than Vollum ΔpXO1 (Fig. 2A; Table 2) (P < 0.001 compared to the Δpag mutant). Therefore, it was important to determine if EF and LF alone can contribute to virulence in the absence of PA, the polypeptide which is considered essential for the entry of EF and LF into the cell, as well as to determine whether or not pXO1 contribute to the virulence in the absence of pag. Results with the Vollum Δpag Δlef double mutant, which produces only EF, or with the Δpag Δcya mutant, which expresses only LF, were quite similar to those obtained with Vollum Δpag, and the estimated attenuation indices for all of these three mutants range between 150 and 350 (Table 2) (P = 0.17 compared to the Δpag mutant). Furthermore, as was the case for the Δpag mutant, for these double mutants the calculated MTTDs also were extended relative to that of the wild type by 1 to 2 days. However, most remarkably, for all these mutants, we consistently observe that some animals survived the infection at doses which are much higher than the “estimated LD50.” For the sake of completeness and to clarify that neither PA in itself nor the virulence factors (EF and LF) alone can be responsible for the observed relatively minor attenuation, we also generated and tested the virulence of the double Δcya Δlef and the triple Δpag Δcya Δlef mutants. Following s.c. infection, these multiple mutants exhibited estimated LD50 and MTTD values similar to those determined for the pag null mutant (P = 0.17) and to those of the other tested double mutants (Fig. 2; Table 2). Essentially, in all animals that were infected and tested for the presence of bacteria in the circulation, we did detect the B. anthracis bacteria, which were genetically verified as the infecting mutant, indicating that the bacteria had disseminated to body organs. In addition, in most of the animals infected with the different mutants, pleural fluid accumulation was detected, as is usually detected in animals infected with the wild-type strain. For some representative animals that survived the infection, blood was collected. Specific seroconversion was detected in all of these tested animals, indicating both that the mutants had the ability to disseminate in the host and that they generated in the host the expected immune response. For example, anti-PA antibodies were detected in the sera of guinea pigs that survived s.c. (or i.n. [see below]) infection with the Δlef Δcya mutant. These results demonstrate that (i) both of the toxins are necessary for optimal virulence in s.c. infection of guinea pigs, (ii) a single functional toxin is sufficient for effective pathogenicity, since elimination of ET or LT leads at most to a 5- to 10-fold increase in attenuation index (Table 2), and, most strikingly, (iii) in the absence of all three factors PA, LF, and EF, the attenuation index is only 250, far from the attenuation index of >2,500,000 for the strain cured of pXO1, the plasmid which codes for these factors. These results clearly suggest that in s.c. infection of guinea pigs, other important mechanisms of pathogenesis exist which are directly or indirectly dependent on pXO1-encoded factors.
Effects of mutations in the pag, lef, and cya genes on the virulence in the guinea pig model. (A) Subcutaneous infection. Groups of guinea pigs were infected subcutaneously with different spore doses of the indicated mutants (5 or 10 animals per mutant type), along with the wild-type strain or its ΔpXO1 derivative. The percentage of survivors is presented as a function of infection dose (CFU). The data are the results of one or the mean value from two experiments with one or two independent mutants as follows: Δpag, single experiment (n = 5; 5 × 107, 106, 104, and 103 CFU); Δpag Δcya Δlef, single experiment (n = 5; 105, 104, 103, and 102 CFU) and two independent mutant isolates (n = 5 × 2; 5 × 107 and 106 CFU). (B) Intranasal infection. Guinea pigs were infected intranasally with different spore doses of the mutant (5 or 10 animals per mutant type) or the wild-type strain. The percentage of survivors is presented as a function of infection dose (CFU). The data presents the results of one or mean value from two experiments with one or two independent mutants as follows: Δpag, single experiment (n = 5; 5 × 107 CFU) and two independent mutants (n = 5 × 2; 107, 106, and 105 CFU); Δpag Δcya Δlef, single experiment (n = 5; 107 and 106 CFU) and two independent mutants (n = 5 × 2; 5 × 107 CFU).
Estimated LD50s, MTTDs, and attenuation indices of mutants in subcutaneously and intranasally infected guinea pigs
(ii) Intranasal instillation.A single mutation in the pag, cya, or lef gene had either a minimal or no effect on the LD50 of the airway infection in guinea pigs (Fig. 2B; Table 2). In the absence of PA, a very significant effect on MTTD (Table 2) was observed (MTTD increased by 4 days). Furthermore, as was found in the s.c. infection, some animals could survive doses of infection that were almost 100 times higher than the LD50. The effects of the various double mutations and of the triple mutation on virulence in the i.n. airway infection were essentially similar to those observed with this animal model following s.c. infection, with some notable differences. Most significantly, the Vollum Δpag Δlef mutant, which expresses only EF, exhibits an attenuation index of only 10 (Fig. 3B) (P = 0.9 compared to the Δpag mutant and P = 0.001 compared to the triple mutant), yet in these experiments some animals survived i.n. instillation at doses of as high as 107 CFU. These results appear to suggest that somehow EF alone can contribute in a PA-independent manner to the virulence mechanism, at least in intranasal infection of guinea pigs. Consistent with this hypothetical role of EF, we find that a deletion of EF in the Δpag Δcya and Δcya Δlef double mutants led to increases of 130- and 1,000-fold in attenuation, respectively, relative to the wild-type strain (Table 2) (P = 0.02 and 0.017, respectively, compared to the Δpag mutant and P = 0.023 and 1, respectively, compared to the triple mutant).
Effects of mutations in the pag, lef, and cya genes on virulence in the rabbit model following subcutaneous and intranasal infection. Groups of rabbits were infected subcutaneously (A) or intranasally (B) with different spore doses of the mutant (2 or 4 animals per group). The percentage of survivors is presented as a function of infection dose (CFU). Results for the wild type are based on data from references 25 and 26) and a repeat for s.c. infection (n = 2; 103 CFU) and i.n. infection (n = 2; 106 CFU). For s.c. infection with Δpag, n = 2 (105 and 106 CFU) and n = 4 (5 × 107 CFU). For the Δlef Δcya mutant and for the Δpag Δcya Δlef triple mutant, only the highest dose of 5 × 107 CFU was tested with 4 animals for each mutant in both s.c. and i.n. infections. ΔpXO1 data represent experiments with 4 animals at a dose of 5 × 107 CFU in both s.c. and i.n. infection.
Following i.n. infection, the triple mutant (Δpag Δcya Δlef) exhibited an attenuation index of 1,600 (based on an estimated LD50 of 5 × 107 CFU, the highest concentration used), concomitant with significant effects on the MTTD (Fig. 2B; Table 2). Unlike in the case of the triple mutant (Δpag Δcya Δlef), all the Vollum ΔpXO1-infected animals survived the dose of 5 × 107 CFU. It should be noted that for the fully virulent Vollum strain, the i.n. LD50 is about 30,000 CFU, and since the highest dose used in the study did not exceed 5 × 107 CFU, only an upper limit to attenuation due the curing of pXO1 can be estimated. In other words, the ceiling for the measured attenuation index in i.n. infection is ca. 1,600 while in s.c. infection it can be as high as 2,500,000 because the LD50 for the wild-type strain in s.c. infection is about 20 CFU. As indicated above for the s.c. infection, all the animals were bacteremic at death, suggesting that all the mutants disseminated to the internal organs. In most of the animals infected with the different mutants, pleural fluid accumulation was detected, and seroconversion was noted in surviving animals, as was reported for the subcutaneous infection. Additional studies using aerosol models should be carried out to confirm and validate these findings and conclusions.
Contributions of toxins and their components to the pathogenicity of B. anthracis in rabbits.In view of the surprising results in guinea pigs with some of the deletion mutants, we decided that it would be important to determine the effect of these mutations in the background of another animal model, the rabbit model.
The picture that emerges from the rabbit model is quite simple and in marked contrast to the one described above for the guinea pig model. Deletion of PA in itself was sufficient for a complete attenuation (Fig. 3; Table 3), and any Vollum strain bearing a Δpag mutation behaved exactly like B. anthracis cured of pXO1. This behavior was independent of the route of infection. Thus, all Δpag mutants exhibited attenuation indices of higher than 2,500,000 and 1,250 in the subcutaneous and intranasal infection routes, respectively. For verification purposes we determined that as expected, Western blots obtained with sera collected from rabbits infected with the Δpag mutant did not cross-react with PA but reacted with both EF and LF. Conversely, sera collected from rabbits infected with the Δcya Δlef mutant (which was also completely attenuated), cross-reacted only with PA (data not shown). As observed earlier (25, 26), the deletion of either EF or LF had a very small effect on virulence following s.c. infection, while in the i.n. route of infection the mutation effect was more pronounced, leading to a reduction of 25- to 75-fold in virulence (Table 3).
Estimated LD50s, MTTDs and attenuation indices of mutants tested in rabbits
DISCUSSION
The accepted paradigm states that anthrax is an invasive and toxinogenic disease and that the lethal toxin (LT) and edema toxin (ET) play major roles in the pathogenicity of B. anthracis. The role of the toxin was studied in a variety of animal models and cell lines from different origins. Mice are sensitive to the capsule, which is the major virulence factor in this species, as encapsulated strains devoid of pXO1 (41) or pag (3) are not significantly attenuated. Therefore, most of the research to determine the roles of toxins in the murine model is based on the attenuated, noncapsulated Sterne strain. A genetic study (32) using this model determined that LF but not EF has a major role in pathogenicity. Mutation in the pag or lef gene resulted in full attenuation of the Sterne strain in mice, whereas removal of the cya gene resulted in an LD50 increase of one order of magnitude. Basically, these findings, together with LT toxicity exhibited on murine macrophage lines, established the major role of this toxin in B. anthracis pathogenicity. Recently we initiated a study (25, 26) aimed to test these interpretations and paradigm using strains with null mutations in lef or cya genes in the background of the fully virulent Vollum strain in the guinea pig and rabbit models. We found that unlike the results for the Sterne challenge in a murine model, the absence of LF or EF had a relatively minor effect on virulence, and the effect was dependent on the route of infection and the animal model. Here we describe an extension of these studies with seven mutants carrying single, double, and triple deletions in one or a combination of the pag, cya, and lef genes (Fig. 1A). These mutants of the Vollum strain were used to infect guinea pigs and rabbits subcutaneously and intranasally (Fig. 2 and 3).
The results obtained in the rabbit model (Fig. 3 and Table 3) are in complete accordance with the accepted paradigm, demonstrating the critical role of the toxins in the pathogenicity of B. anthracis. All rabbits, independent of the route of infection, survived infection with as much as 5 × 107 CFU of the Vollum strain harboring the Δpag mutation alone or in combination with the other toxin component-related mutations, such as Δcya and/or Δlef (Fig. 3). Thus, by eliminating the ability of the B. anthracis Vollum strain to produce PA, the LD50 increases by at least 6 or 3 orders of magnitude in the subcutaneous or the airway infection route, respectively (Table 3). In the rabbits, the effect of the deletion of pag was equivalent to that observed after curing the virulence plasmid pXO1 from Vollum. However, bacteria expressing at least a single toxin, ET or LT, retain most of their pathogenic potential in subcutaneous infection but exhibit a decrease in virulence of one to two orders of magnitude in airway infection. We may therefore conclude that in rabbits, the anthrax disease is absolutely dependent on expression of at least one of the classical B. anthracis toxins, ET or LT, and that the virulence of B. anthracis and the lethality of the disease can be completely abolished by damaging the ability of the bacteria to produce functional PA molecules.
Unlike the clear-cut results seen in rabbits, the virulence and pathogenicity of B. anthracis in guinea pigs were much more complex. Most significantly, the virulence of the Vollum strain was not abolished as a result of the deletion of either pag or even all three of the toxin-related genes, pag, cya, and lef (Fig. 2). This was observed in both the intranasal and subcutaneous infections, but its full extent could be better appreciated and quantified in the s.c. infection mode, mainly because the LD50 of the wild-type virulent Vollum strain is relatively low (20 CFU in s.c. infection versus 30,000 CFU in i.n. infection). Thus, while the pag-containing mutants exhibit attenuation indices which are in the range of 150 to 350, the attenuation index determined for the ΔpXO1 derivative of Vollum is significantly higher, exceeding 2,500,000 (Table 2). This means that B. anthracis strains which are unable to produce both toxins LT and ET or just PA still retain, in the guinea pig model, a very substantial level of pathogenicity which is dependent on the presence of pXO1. The virulence plasmid pXO1 carries, besides pag, cya and lef, genes encoding 28 other polypeptides within its pathogenicity island (most with undefined function [1, 30]), and at least 40 additional hypothetical or unknown open reading frames (ORFs). One or more of these ORFs could contribute to or be directly responsible for the observed toxin-independent virulence in the guinea pigs. In addition, there are numerous chromosomal genes that are regulated by pXO1 (2, 6, 31). This cross talk between pXO1 and the chromosome involves modulation of expression of several genes, which could also contribute to the unique pathogenicity of B. anthracis in guinea pigs. It is worth noting in this context that in recent years several groups identified chromosomal genes which could lead to at least a 10,000- to over a 1,000,000-fold decrease in virulence in guinea pigs (e.g., mntA [17], purH [22], and htrA [5]). Most notably, in most cases it was demonstrated that these chromosomal mutations did not affect the ability of the infecting B. anthracis strain to express and produce the classical virulence factors (e.g., LT, ET, or the capsule) in the infected animals.
Another intriguing observation, exhibited mainly in guinea pigs, which remains unexplained is that infection via the s.c. route with most of the mutants did not lead to the death of all the infected animals even when the infecting dose was several orders of magnitude above the estimated LD50. This was a consistent and a reproducible observation, found in the identical pairs of independently isolated mutants (Fig. 1) and in replicate experiments (Fig. 2). Actually, in the guinea pigs, LD50 plots that followed the classical behavior (namely, all animals succumbed to death at doses much higher than the LD50) were observed only for the wild-type Vollum strain and its two single-mutant derivatives expressing functional LT or ET. Finally, we also note that in guinea pigs infected intranasally, EF alone could apparently contribute to the pathogenicity (see the results for the Δpag Δlef and Δcya Δlef double mutants in Fig. 2B and Table 2). Studies are now in progress to better understand this EF-dependent virulence mechanism. It appears from all the above that guinea pigs, being less susceptible to the toxins, allow the manifestation of other virulence mechanisms that are usually masked by the toxin effect in other animals. These conclusions are supported by previous studies (14, 17) and recent studies on HtrA (5) and the purH mutant of B. anthracis (22). In the latter, purH mutants of the Ames strain were found to be as virulent as the parental strain in either mice or even rabbits, but in guinea pigs, the purH mutant was at least 10,000-fold more attenuated than the parental Ames strain.
The fact that a critical component of virulence of B. anthracis in guinea pigs is pXO1 dependent but toxin independent obviously needs to be further explored. However, this finding in itself may already have important practical implications with respect to future directions in the development and evaluation of next-generation anthrax vaccines. Guinea pigs are used for anthrax vaccine production and potency testing before batch release (U.S. Pharmacopeia USP 34/NF29-2011; European Pharmacopeia 6.0, 01/2008:2188). On the other hand, according to the “two-animal model rule,” the FDA proposes the use of rabbits and monkeys as relevant animal models for evaluation of anthrax vaccines (13). Thus, if as observed here for rabbits it will be shown also for monkeys that the virulence mechanism is completely PA dependent, then it will suffice for any efficacious anthrax vaccine to include PA formulated with some appropriate adjuvant. However, if the more complex virulence and pathogenicity mechanism observed in guinea pigs is somehow relevant to the disease in humans, then it may be prudent to include in future vaccines additional B. anthracis factors besides PA. This could require extension of studies already initiated in recent years (16, 36) such as “reverse vaccinology” along with more extensive proteomic studies (7, 8) and even new approaches toward development of novel live attenuated vaccines (9) that include pXO1 genetic elements (for a review, see reference 4). Concomitantly, it may be prudent to invest more efforts in order to determine the specific parameters in the various animal models which could be considered most relevant to the human anthrax disease and thus most suitable for evaluation of next-generation anthrax vaccines.
ACKNOWLEDGMENT
We thank Nili Rothschild for her excellent technical assistance.
FOOTNOTES
- Received 8 March 2012.
- Accepted 7 May 2012.
- Accepted manuscript posted online 14 May 2012.
- Copyright © 2012, American Society for Microbiology. All Rights Reserved.
