Detection of Anthrax Toxin in the Serum of Animals Infected with Bacillus anthracis by Using Engineered Immunoassays

Sandwich ELISA for PA detection.The M18 scFv construct had previously been cloned into pMoPac16 (20), expressed as a scAb, and purified by immobilized metal affinity chromatography (IMAC) via a C-terminal polyhistidine tag and size-exclusion chromatography (30). ANTXR2 was a gift from Robert Liddington's group (The Burnham Institute, La Jolla, Calif.) and was prepared as previously mentioned (41). Rabbit anti-PA polyclonal, native PA, and Y688A PA were produced as previously described (39). Human serum was obtained from a donor at the University of Texas Health Center, Austin. For overnight coating, 50 μl of M18 scAb (4 μg/ml) or ANTXR2 (4 to 30 μg/ml) was incubated overnight in a 96-well ELISA plate (no. 3590; Corning Inc., Corning, NY) at 4°C. The plate was then brought to room temperature and incubated with 2% dried milk (Carnation, Nestle)-phosphate-buffered saline (PBS) or 2% bovine serum albumin (BSA; Sigma)/PBS for 2 h. The plate was washed three times with 1× PBS-0.5% Tween 20 and then once with 1× PBS. For initial assays, native or Y688A PA was diluted into 2% milk-PBS or human serum at 1 μg/ml and serially diluted onto the ELISA plate. For assays detecting PA in blood of infected animals, serum was added to the plate initially diluted 1:1 in 2% milk-PBS and then serially diluted across the plate. Preexposed serum spiked with quantified PA served as the positive control and was serially diluted as well. After a 1-h incubation, the plate was washed as described above. Rabbit anti-PA polyclonal serum was diluted 1:1,000 in 2% milk-PBS and added to the plate for a 1-h incubation. The plate was then washed again, followed by the addition of rabbit anti-goat immunoglobulin G (IgG)-horseradish peroxidase (HRP) conjugate (Bio-Rad) diluted in 2% milk-PBS for 1 h. ELISA reactions were developed with o-phenylenediamine (OPD) tablets (Sigma) and quenched by the addition of 50 μl of 4.5 M H₂SO₄. All ELISAs were run in duplicate, averaging each data point with standard deviation. Experimental assay data determining the detection of PA were applied to Prizm 3.03 software (GraphPad Software Inc., San Diego, CA) to calculate the statistical significance for differences between detection methods of each model. In addition, data were curve fitted to a nonlinear regression model to determine “goodness of fit.”

Sandwich ELISAs for LF detection.Recombinant PA₈₃ and PA₆₃ were purchased from List Laboratories (New Jersey). PA₆₃ is a cleavage product that is capable of binding LF (45). For the sandwich ELISA, 50 μl of PA at 63 kDa and 83 kDa (6 μg/ml) was applied to a 96-well plate and blocked with 2% milk-PBS as described above. For initial assays, LF was diluted in PBS or human serum at 5 μg/ml. For assays detecting LF in infected animals, serum was added to the plate initially diluted 1:1 in 2% milk-PBS and then serially diluted across the plate in duplicate. After a 1-h incubation, the plate was washed as described above. Goat anti-LF polyclonal serum (List Labs) was diluted 1:1,000 in 2% milk-PBS and added to the plate for a 1-h incubation in duplicate. The plate was then washed, followed by the addition of goat anti-rabbit IgG-HRP conjugate (Bio-Rad) diluted in 2% milk-PBS for 1 h. ELISA reactions were developed with OPD tablets (Sigma) and quenched by the addition of 50 μl of 4.5 M H₂SO₄. Observed concentrations were calculated from positive control. Data were applied to Prizm 3.03 (GraphPad Software Inc., San Diego, CA) software for statistical analysis.

Animal spore exposure and histology.Female Hartley guinea pigs (225 to 305 g) (Charles River Laboratories, Massachusetts) were housed individually in a One Cage 2100 AllerZone Interchangeable Micro-Isolator high-density housing system (Lab Products Inc., Seaford, DE), and New Zealand rabbits (2.0 to 2.5 kg) (Charles River Laboratories, Wilmington, MA) were housed individually in ventilated rabbit cage racks (model no. RBV243116US6-2X; Allentown Caging Equipment Company, Inc., Allentown, NJ). Animals were housed in a biosafety level 4 facility at The Southwest Foundation for Biomedical Research (San Antonio, TX) for spore challenge. Anthrax spore inocula, both the Ames and Vollum strains, were prepared on the day of challenge and diluted to the desired concentration in PBS. Both guinea pigs and rabbits were sedated with ketamine (80 mg/kg of body weight) and xylazine (10 mg/kg) during all bleeds, injections, and inhalation instillations, according to the approved IACUC protocol. Guinea pigs were exposed to B. anthracis Vollum 1B (3) spore inoculum of 500× LD₅₀ (2.0 × 10⁷) (100-μl total volume), and rabbits were exposed to a 20× LD₅₀ (2.1 × 10⁶) (48) of B. anthracis Ames spores via intranasal inoculation in both cases. Blood samples were taken from the femoral artery at various times. The guinea pigs were monitored for signs of toxemia and euthanized by cardiac injection of sodium pentobarbital when considered moribund. Symptoms of toxemia include labored breathing, curling of digits, edema, abnormal color in nose/feet/tail, and immobility. For the rabbit studies, euthanasia was not used and blood samples were taken only at the time of death. Serum was obtained by centrifugation and separation using StatSampler columns (StatSpin, Norwood, MA).

For histology, lungs and spleens were removed from the guinea pigs postmortem. A 100-mg section was excised and homogenized with 0.4 ml of sterile PBS in a sterile tissue grinder. The homogenates were serially diluted in sterile PBS and then plated in duplicate on sheep blood agar (SBA) plates. After 24 h of incubation at 37°C, the colonies were enumerated to determine the number of CFU/100 mg of tissue.

PA detection ELISAs.A highly sensitive sandwich ELISA for PA capture was developed based on the engineered M18 antibody. Briefly, the M18 murine scFv (19) was isolated from an error-prone library derived from the anti-PA 14B7 murine monoclonal antibody (29) using a new bacterial display technology referred to as anchored periplasmic expression (APEx) (19). The M18 scFv was cloned into the pMoPac 16 vector (20) to produce a fusion between a human kappa chain and the heavy-chain variable domain of the scFv. The resulting construct is referred to as a scAb (21). The purified M18 scAb fragment binds PA with an equilibrium dissociation constant (K_D) of 35 pM (19).

Figure 1A shows a schematic of the anti-PA detection ELISA. The well bottoms of a 96-well plate were coated with the M18 scAb to serve as the capture agent. PA was serially diluted in 2% milk-PBS and human serum and then incubated for 1 h on the plate. After washing, rabbit anti-PA polyclonal serum was diluted in 2% milk-PBS and added to the plate. After washing, goat anti-rabbit polyclonal IgG-HRP conjugate was added for detection of the bound rabbit IgG. Note that having two polyclonal detection antibodies adds potential signal amplification steps. The results from these serial dilution assays revealed that PA can be detected in 2% milk-PBS and human serum with an observed lower limit of detection of ≈1 ng/ml (Fig. 1B) based upon the concentration of PA at which a threefold-higher signal over background or negative control was seen. The significance between the curves was calculated as a two-way analysis of variance (ANOVA) model using Prism 3.03 software (GraphPad Software, San Diego, CA), resulting in an overall P value of <0.0001. “Goodness of fit” for each curve was determined via a nonlinear regression model using the same software to reveal R² values of 0.990 (serum) and 0.994 (milk) for PA detection.

• Open in new tab
• Download powerpoint

FIG. 1.

PA detection immunoassay. (A) Schematic for PA detection assay with M18 scAb. (B) For experimental detection, PA was detected at 1 μg/ml in milk-PBS (purple squares) or human serum (blue diamonds) and serially diluted in the plate assay. After probing with the primary and secondary antibodies as described, results showed that PA is detectable in milk-PBS and serum at very similar levels, with an overall lower detection limit of ≈1 ng/ml.

Engineered ELISA for PA variant.The 14B7 antibody was originally identified as a neutralizing antibody and shown to act by blocking binding of PA to cells (29). It has been shown to bind to the PA domain 4 small loop region involved in receptor binding (39, 41). Domain IV is responsible for the majority of binding to the ANTXR2 (CMG2) (42) and ANTXR1 (ATR/TEM8) (7) receptors, although a contribution from PA domain II was recently identified by X-ray crystallographic analysis of the complex (41).

The ability to genetically modify a potential biowarfare agent to evade current and future therapeutics must be taken into consideration. Rosovitz and colleagues showed that a PA mutant having the Y688A substitution in domain IV retains full toxicity (39), even though binding by some anti-PA antibodies was disrupted. We sought to develop an assay capable of detecting anthrax strains containing a genetically modified PA component that retains receptor binding, and thus virulence, but might evade any single capture ELISA antibody. The ANTXR2 receptor was chosen as a capture agent due its reported high affinity to PA (7). The soluble component of the ANTXR2 receptor was substituted into the capture ELISA described above in place of the M18 scAb construct. The ELISA was then performed corresponding to the schematic shown in Fig. 2A. Plates containing either the ANTXR2 soluble receptor or the M18 scAb were analyzed for detection of serially diluted native PA and PA Y688A. Results from the assays are shown in Fig. 2B. Assays using ANTXR2 as the capture agent revealed very similar curves for both native PA (R² = 0.9935) and PA Y688A (R² = 0.9979), with a reported P value of <0.0001 using the two-way ANOVA model. On the other hand, the M18-based assay showed a marked difference in sensitivity between the two PA samples, with the assay for PA Y688A (R² = 0.9851) being close to an order of magnitude less sensitive than M18 binding native PA (R² = 0.9986). Interestingly, despite a relatively high affinity reported for this receptor binding to PA, the ANTXR2-based assay was considerably less sensitive than the ELISA based on M18, even in the case of PA Y688A detection. The discrepancy could involve the way in which the ANTXR2 receptor binds to the plate, possibly leading to lower overall assay sensitivity by precluding unobstructed access to the PA toxin binding site.

• Open in new tab
• Download powerpoint

FIG. 2.

Sandwich detection ELISA for native PA and mutant PA Y688A. (A) The soluble anthrax receptor, ANTXR2, and M18 scAb were capture ligands for detection of native PA and the mutant PA Y688A in domain IV. (B) Experimental results show that ANTXR2 bound to both native and mutant PAs with similar affinities. There was an order-of-magnitude decrease in sensitivity for M18 to the mutant PA compared to native toxin; however, overall the M18 assay was still more sensitive in both cases.

LF detection ELISA.The PA 63-kDa fragment was purchased from List Labs (New Jersey) and applied as the capture ligand for LF detection in a sandwich ELISA (Fig. 3A). Full-length PA₈₃ was used as a negative control since this fragment cannot bind LF, whereas PA₆₃ was presumed to be in heptameric form and possess binding sites for LF (33) due to solution conditions from the supplier. As for our PA ELISA, detection of captured LF was accomplished by primary polyclonal anti-LF followed by HRP-conjugated secondary antibodies. As expected, using PA₆₃, but not PA₈₃, as the capture agent produced an ELISA signal. “Goodness of fit” resulted in R² values of 0.9904 for serum and 0.9899 for milk PA detection. Two-way ANOVA calculated a P value of <0.0001 for comparison between the two detection methods. A lower detection limit of ≈20 ng/ml LF (Fig. 3B) was determined for the PA₆₃-based assay in serial dilution experiments, which may correspond to the binding affinity (low nM) of LF to PA (14).

• Open in new tab
• Download powerpoint

FIG. 3.

Sandwich detection ELISA for LF. (A) Schematic shows the PA 63-kDa fragment as the capture ligand in the ELISA for LF. (B) For experimental detection, the PA 63-kDa or 83-kDa (control) fragment was applied as the capture ligand for LF detection. LF was detected at 4 μg/ml in milk-PBS (blue diamonds) or human serum (yellow triangles). The results showed that LF is detectable in milk-PBS and serum at very similar levels, with an overall lower detection limit of ≈20 ng/ml when the PA 63-kDa fragment was used as the capture ligand. There was no detection of LF when the PA 83-kDa fragment was the capture ligand. The lower detection limit of the assay was calculated as a signal threefold higher than the PA 83-kDa negative control (purple squares and turquoise circles).

Diagnostic detection of PA from exposed guinea pigs.To examine the ability of the assay to detect PA produced during infection, female Hartley guinea pigs (225 to 300 g) were exposed to a 500× LD₅₀ (2 × 10⁷) of Vollum anthrax spores (3) and then bled twice daily until euthanasia or death. PA was detected in blood in systemic circulation using the M18 capture ELISA in 4 of the 5 animals before death (Table 1). As mandated by the approved IACUC protocol, animals displaying any observed signs of toxemia were immediately euthanized. However, two animals expired from anthrax in spite of constant monitoring for signs, and their serum samples were taken postmortem. For the three animals euthanized, two showed detectable levels of circulating PA at the time of euthanasia. Both of the animals that died from anthrax possessed detectable serum levels of PA prior to onset of toxemic signs as well as after death. One animal (25539) was found to have a relatively high serum PA level (24 ± 5 μg/ml) at the time of death, compared to the level of 0.4 ± 0.06 μg/ml detected in the same animal 9 h earlier.

Taken together, our results demonstrate that PA can be detected in the late stages of guinea pig infection, especially within 12 h of death. Tissue samples from all five animals were taken for histology tests and revealed that animals considered moribund, and thus euthanized, possessed equivalent CFU (≈10⁷ to 10⁹) of bacteria compared to those that expired due to infection (data not shown).

Toxin detection from rabbits exposed to Ames spores.As a preliminary investigation into the generality of the approach to PA detection and to evaluate the effectiveness of the LF assay, another animal model, New Zealand White rabbits, was investigated using the other fully virulent laboratory strain of anthrax, the Ames strain. Two rabbits were exposed to a 20× LD₅₀ of Ames spores via intranasal inoculation. Upon death (mean, ≈48 h), serum samples were taken for detection of PA and LF components of the toxin. Table 2 summarizes the ELISA results. The two animals expired with a mean time to death (MTTD) of ≈48 h, an MTTD very similar to those of previous reports of exposure to a 150× LD₅₀ (34, 48). Upon death, both animals possessed detectable levels of PA and LF in the systemic circulation. PA was found to be present at concentrations of 80 ± 7 μg/ml and 100 ± 13 μg/ml in the two animals, using the M18 antibody capture ELISA. The elevated PA levels compared to those in the guinea pigs might indicate that toxin secretions differ among animal models and/or between different strains (Ames versus Vollum 1b). The two animals expired with an MTTD of ≈48 h, an MTTD very similar to those of previous reports of exposure to a 150× LD₅₀ (34, 48). At time of death, blood samples were taken and PA values of 80 to 100 μg/ml were detected. The elevated PA levels compared to those of the guinea pigs might indicate that toxin secretions differ among animal models and/or between different strains (Ames versus Vollum 1b).