|Home | About | Journals | Submit | Contact Us | Français|
Protective antigen (PA) is the cell surface recognition unit of the binary anthrax toxin system and the primary immunogenic component in both the current and proposed “next-generation” anthrax vaccines. Several studies utilizing animal models have indicated that PA-specific antibodies, acquired by either active or passive immunization, are sufficient to protect against infection with Bacillus anthracis. To investigate the human antibody response to anthrax immunization, we have established a large panel of human PA-specific monoclonal antibodies derived from multiple individuals vaccinated with the currently approved anthrax vaccine BioThrax. We have determined that although these antibodies bind PA in standard binding assays such as enzyme-linked immunosorbent assay, Western blotting, capture assays, and dot blots, less than 25% are capable of neutralizing lethal toxin (LT) in vitro. Nonneutralizing antibodies also fail to neutralize toxin when present in combination with other nonneutralizing paratopes. Although neutralizing antibodies recognize determinants throughout the PA monomer, they are significantly less common among those paratopes that bind to the immunodominant amino-terminal portion of the molecule. These findings demonstrate that PA binding alone is not sufficient to neutralize LT and suggest that for an antibody to effectively block PA-mediated toxicity, it must bind to PA such that one of the requisite toxin functions is disrupted. A vaccine design strategy that directed a higher percentage of the antibody response toward neutralizing epitopes may result in a more efficacious vaccine for the prevention of anthrax infection.
The Bacillus anthracis binary toxin system contributes directly to anthrax pathogenicity in the host (3, 14). The cell surface recognition element of this toxin system is an 83-kDa protein known as protective antigen (PA83). Antibodies that bind PA protect against infection (8, 12), and PA is the primary immunogenic component in the anthrax vaccine currently licensed for use in the United States (BioThrax, or anthrax vaccine adsorbed [AVA]; Emergent Biosystems). Ongoing attempts to develop a “next-generation” anthrax vaccine are relying on a recombinant form of PA as the sole immunogenic component. PA's role as an important vaccine target has driven a significant amount of research into both the biology of this protein toxin and the immunobiology of its interaction with the immune system of the vaccinated or infected host.
PA83 binds to the cell surface receptors tumor endothelial marker 8 and the capillary morphogenesis gene 2 product (4, 20). Bound PA is cleaved by cell surface-associated furin proteases to release the 20-kDa amino-terminal portion of the molecule (PA20), which has no further role in intoxication. Following proteolytic cleavage, cell-bound PA63 self-assembles to form a heptameric prepore structure that can bind several molecules of the catalytic toxin components lethal factor (LF) and/or edema factor (EF). Receptor-mediated endocytosis results in the internalization of the complex, which inserts into the membrane of the endocytic vacuole. LF and/or EF is then actively translocated into the cytoplasm of the cell. The structure of PA, both as a monomer and as a heptamer, has been determined (15, 19), and the regions of the molecule (domains) involved in the various functions described above have been identified (1, 6, 15, 18, 19).
The immunobiology of the immune response to PA in vaccinated humans has only recently been explored at the molecular level. PA elicits a polyclonal antibody response in vaccinated humans that utilizes a wide variety of immunoglobulin variable (V)-region genes. Preliminary studies have indicated that after vaccination, antibodies undergo the somatic hypermutation and class switch normally associated with affinity maturation (21). We have previously demonstrated the human antibody response to PA to be significantly biased toward epitopes associated with the amino-terminal domain of the PA protein (PA20) and have postulated that these antibodies may be deficient in their ability to neutralize toxin (16).
In this study, we determined the toxin neutralization potentials of a large panel of individual monoclonal antibodies isolated from seven individuals vaccinated with AVA vaccine, using a cell-based assay of LT-mediated cytotoxicity. We found that only 24% of the component antibodies that comprise the overall response are capable of neutralizing PA-mediated cytotoxicity in vitro. We found no direct correlation between the relative PA binding ability of the individual antibodies and their ability to neutralize anthrax toxin. We also determined that toxin-neutralizing paratopes occur less frequently among those antibodies that recognize the immunodominant epitopes associated with the amino-terminal domain of the PA monomer. These findings suggest that the efficacy of future PA-based vaccines might be improved by modifying the immunogen such that a greater proportion of the antibody response is directed toward those epitopes that lead to toxin neutralization.
The donors analyzed in this report were recruited from individuals taking part in a larger study of the response to AVA vaccine being conducted at Baylor College of Medicine. Human subject protocols were reviewed and approved by the Institutional Review Boards at both Children's Hospital Oakland and Baylor College of Medicine.
Fab expression libraries were constructed from mononuclear cells enriched for PA-specific B cells in a manner similar to that previously described for PA- and polysaccharide-specific antibody expression libraries (16, 17, 21; J. Zhou and D. C. Reason, unpublished data). PA83 was purchased from List Biological Laboratories, Campbell, CA. PA-specific Fabs were identified using a sensitive 125I-labeled PA capture assay and lysates of individual Escherichia coli expression cultures. Positive isolates were recloned, heavy (H)- and light (L)-chain gene sequence determined, and PA-specific binding confirmed by enzyme-linked immunosorbent assay (ELISA). Initial sequence analysis utilized the NCBI IgBlast server (http://www.ncbi.nlm.nih.gov/igblast/) to identify candidate germ line gene (2). Subsequent analysis, alignments, and translations were performed using MacVector (Accelrys Inc., Princeton, NJ). H- and L-chain V-region gene nomenclature is as described in the IMGT database (11). Complementarity-determining regions are as defined previously (9).
In vitro expression of full-chain immunoglobulin G1 (IgG1) antibodies utilized the Flp-In Chinese hamster ovary (CHO) cell system from Invitrogen (Carlsbad, CA). H- and L-chain V-region gene segments were isolated from PA-specific Fabs and inserted into the FLP recombination target vector as a bicistronic eukaryotic expression cassette utilizing an internal ribosomal entry segment sequence. Flp-In CHO cells were plated at 3.5 × 105 cells per well (in 2 ml Flp-In medium) in Nunclon Delta six-well plates and then incubated at 37°C with 5% CO2 overnight. Once cells reached 80% confluence, they were transfected with pOG44 and the FLP recombination target vector (9:1 ratio) using the TransFast transfection reagent (Promega). Forty-eight hours after transfection, the cells were trypsinized and placed in a fresh six-well plate under drug selection with 600 μg/ml hygromycin. Antibody was concentrated from the cell culture supernatant for use in binding and toxin neutralization assays.
PA and LF were purchased from List Biological Laboratories, Campbell, CA. RAW 264.7 cells were plated at 4 × 104 cells per well (in 65 μl assay medium) in Nunclon Delta 96-well plates and incubated at 37°C with 5% CO2 for approximately 4 to 5 h to ensure proper settling and attachment. Antibodies and PA/LF were preincubated in Greiner Bio-One 96-well plates at 37°C with 5% CO2 for 1 h and then were transferred to the 96-well plate containing RAW 264.7 cells. The plate was then placed back into the incubator at 37°C with 5% CO2 for overnight incubation. The following morning, 20 μl Cell-Titer Blue reagent (Promega) was added to each well of the assay plate. Optical density (at 570/595 nm) was determined for each well 4 hours later using a microtiter plate reader.
Antibody concentration was determined by a capture ELISA in which goat anti-human Fc (The Binding Site, Birmingham, United Kingdom) immobilized on a microtiter plate captures IgG, which is then detected by alkaline phosphatase-labeled goat anti-human L chain (Biosource International, Camarillo, CA). This assay is standardized using a purchased IgG1 protein standard (Sigma). PA binding in ELISA was determined for IgG1 antibodies on 96-well plates coated with 5 μg/ml PA83 and developed with alkaline phosphatase-conjugated goat antibody specific for human kappa or lambda light chains.
Neutralizing antibodies were preincubated with modified forms of PA to verify the specificity of the antibody-PA binding in the toxin neutralization assay. PA20 (residues 1 to 191) was expressed fused to green fluorescent protein (GFP) as previously described (16). A nonfunctional mutant of PA (PArb−) was constructed using QuikChange mutagenesis (Stratagene) by mutating two residues in the domain 4 region of wild-type PA (N682A and D683A) to remove its ability to bind to the cell surface receptor. This modification was necessary to prevent PA added to the assay as an inhibitor from participating in LF-mediated toxicity. Neutralizing antibodies were incubated with the modified PA constructs overnight at concentrations sufficient for 50% inhibition of LT-mediated cytotoxicity. These were then added to the neutralization assay mixture as described above, and the degree to which the preincubation blocked the ability of the antibodies to neutralize toxin was calculated.
Differences between groups were analyzed using the Mann-Whitney U test and the two-sided Fisher's exact test as implemented within the Prism graphic software package.
We have isolated and characterized a large panel of PA-specific monoclonal antibodies from AVA vaccine-immunized donors (16, 21). Sequence analysis of H-chain joining-region usage and V H-chain complementarity determining region 3 sequence and length allowed us to assign the over 120 somatically mutated and differentiated antibodies isolated from these donors to 64 families that arose from unique and independent B-cell rearrangement events (16). To determine the toxin neutralization potentials of the 64 individual members of the panel, the Ig V-region-coding genes of each isolate were cloned into a bicistronic, eukaryotic expression vector and transfected into CHO cells to facilitate the production of each antibody paratope as a bivalent IgG1 molecule. CHO cells were maintained in culture, and expressed human monoclonal antibody was concentrated from the culture supernatant and quantitated prior to analysis. Six of the 64 antibodies transferred into the CHO expression system failed to express a sufficient quantity of antibody for testing. To determine the LT neutralization potentials of the 58 expressed antibodies, we employed a well-established in vitro RAW 264.7 cell-based assay of LT-mediated cytotoxicity. Titrated quantities of antibody were incubated with PA and LF to allow antibodies to bind PA. Antibody-toxin mixtures were then added to previously seeded RAW 264.7 cells in 96-well plates, and cell viability was assessed using CellTiter-Blue (Promega) following overnight culture. The optical density in each antibody-containing well was measured and calculated as a percentage of that in wells containing no toxin. Data were expressed as percent viability versus antibody concentration (Fig. (Fig.11).
The majority of the individual paratopes failed to neutralize LT in culture at any concentration tested. The maximum concentration of antibody tested varied for each clone based on the level of expression that we were able to achieve in the CHO cell culture for that particular clone. All were tested at molar concentrations in excess of the 3.6 nM PA in the assay and in excess of that required to demonstrate neutralization by the poorest neutralizing clone in the panel (see below). Several nonneutralizing clones, when tested in combination of up to four antibodies, were also unable to neutralize toxin (data not shown). To demonstrate the specificity of the antibody-mediated neutralization, antibodies were preincubated with either a mutated form of PA (contains a 2-amino-acid substitution in the receptor binding region of the molecule and is nonfunctional) or the amino-terminal PA20 region of PA fused to GFP prior to their addition to the RAW assay. This preincubation abrogated the neutralizing potential of the antibodies in an epitope-specific manner (Fig. (Fig.2)2) for most antibodies, demonstrating that the toxin neutralization we observed arose from the PA-specific antibody binding to the PA in the assay. Monoclonal antibody 4A12, which is specific for an epitope associated with the domain 4 region of the monomer (16), was not inhibited by preincubation with PArb− in this assay. The two mutations introduced into PA to render it nonfunctional for this assay were in the domain 4 region and likely contribute to 4A12 binding. Monoclonal antibody 25G9 also does not inhibit, as might be expected in this assay. This antibody did not capture PA20 from solution in the capture assay, and specificity was assigned based on its ability to bind PA20 in a solid-phase ELISA. It is possible that the relevant PA20-associated epitope is not retained or efficiently displayed by this construct in the soluble phase required for this assay. Overall, approximately 24% of the monoclonal antibodies in the panel were capable of neutralizing LT in the in vitro assay. The toxin-neutralizing potential of each monoclonal antibody in the panel is shown in Table Table11 and summarized in Table Table22.
In addition to specificity, the affinity of the individual antibody combining sites for the target antigen and the region of the antigen bound (epitope) may determine the ability of an individual toxin-specific antibody to block toxin function. To determine if toxin-neutralizing and nonneutralizing antibodies differed in their relative abilities to bind PA, the titers of monoclonal antibodies were determined in a PA-specific ELISA and the antibody concentration resulting in 50% binding (BC50) ascertained. BC50 values for nonneutralizing antibodies varied from 12 to 42,500 ng/ml (median = 35 ng/ml), with the majority of values falling between 16 and 3,200 ng/ml. Neutralizing antibodies also varied widely in their BC50 values (median = 60 ng/ml). When the relative avidities for the neutralizing antibodies were compared to the overall antibody pool (Fig. (Fig.3),3), BC50 distributions for the two groups overlapped, and no significant differences were observed between antibodies capable of neutralizing toxin and those lacking efficacy in the cytotoxicity assay (Mann-Whitney U test, P = 0.72).
We have previously demonstrated that the antibody response to PA is highly biased toward epitopes present in the amino-terminal portion of the PA monomer (PA20), with about 63% of the antibodies binding to this region of the molecule (which represents only 25% of the PA monomer's mass) (16). PA20 is cleaved from the remainder of the PA molecule rapidly following cell surface binding and has no further role in the intoxication process. It is possible that antibodies specific for epitopes located on this region of the molecule might be deficient in their ability to effect toxin neutralization. Compared as a function of PA domain specificity (Table (Table1),1), toxin neutralization was more frequently observed within the subset of antibodies specific for epitopes associated with the PA63 region of the PA monomer. Seven of the 16 antibodies (44%) specific for PA63-associated determinants (including those specific for domain 4) were capable of neutralizing toxin. Seven of the 38 antibodies (18%) specific for PA20-associated determinants neutralized toxin in the in vitro RAW cell assay. PA63-specific antibodies were therefore over twice as likely to neutralize LT as those specific for PA20-related epitopes (P = 0.08 by the two-sided Fisher's exact test).
In addition to determining the relative frequency of toxin-neutralizing paratopes within the two epitope-specific populations, we also sought to determine if the two antibody populations varied in their relative efficacy in toxin neutralization. The concentration of antibody resulting in 50% toxin neutralization (NC50) in the RAW cell assay was determined for each antibody clone. NC50 values for the two groups were compared to determine if this amount varied as a function of domain specificity. On average, PA63-specific clones required about 3.8-fold less antibody to achieve 50% neutralization compared to clones specific for PA20-associated epitopes (Fig. (Fig.4).4). Statistical comparison of the two groups indicated that PA63-specific antibodies were significantly more efficient in neutralizing LT (P = 0.01 by the Mann-Whitney U test) than PA20-specific antibodies.
The human antibody response to a large protein antigen such as PA is complex. Exposure to such a molecule activates a large collection of B cells, each of which produces an antibody specific for an individual antigenic epitope associated with the immunogen. Most, if not all, of these antibodies undergo further somatic hypermutation, giving rise to a diverse assemblage of antibody paratopes, each characterized by the three-dimensional structure on the target molecule it recognizes (its epitope) and the strength with which it binds that structure (its affinity). For antibodies that bind determinants physically associated with a pathogen (such as a bacterial cell wall), the functionality of the antibody is often determined by the constant region of the antibody molecule. Complement fixation and opsonization, for example, are facilitated by the constant region and are the primary means by which antibody mediates the destruction of many pathogens. The mechanism by which antibodies neutralize soluble agents such as toxins is not well understood. The findings we present here demonstrate that avid antibody binding alone is insufficient to block PA's role in intoxication and suggest that the antigenic epitope recognized is the primary determining factor in antibody function.
Given the complexity of the role that PA plays in anthrax intoxication, the finding that the majority of PA-specific antibodies in vaccinated individuals do not neutralize LT is unexpected. For PA-mediated cell death to occur in the in vitro assay employed in this study (and presumably in vivo as well), PA must bind to the cell surface receptor and be cleaved by furin to yield cell-associated PA63. Cell-bound PA63 must then form homoheptamers, the PA63 heptamers must bind LF, and the PA/LF complex must be internalized and released into the cytosol. Our data demonstrate that this complicated chain of events can proceed unimpeded in the presence of antibody bound to each of the participating PA monomers. The predominance of PA20-specific paratopes in the response may offer a partial explanation of the phenomenon in serum. Antibodies binding the PA20 region of the molecule would be detached from cell-bound PA83 (along with the PA20 fragment) following furin cleavage. Additionally, as free PA20 accumulates in the culture supernatant (or in the serum ), this proteolytic fragment could compete for antibody binding with those PA20-associated epitopes still associated with intact PA83. Nine of the 16 PA63-specific antibodies that we assayed were also nonneutralizing. It remains unclear how PA retains its functionality when complexed with these antibodies.
While we believe the collection of antibodies we isolated to be representative of the overall response, it is not complete, and further investigations of these same donors would undoubtedly uncover additional PA-specific paratopes. Serum from donor 3, for example, exhibited very low, but nevertheless above-background, toxin neutralization activity when tested at low dilutions. None of the antibodies that we isolated from this individual neutralized toxin. This could result either from insufficient sampling or from the fact that the respective B cells were not circulating at the time blood was collected. As a group, the antibody panel represents a wide range of relative avidities, recognizes epitopes distributed throughout the PA monomer, and has both neutralizing and nonneutralizing members. These factors suggest no obvious sampling bias and that the paratope distribution we observe reflects the paratope distribution present during the ongoing immune response in a vaccinated individual.
Establishing the relationship between antibody paratope, antigenic epitope, and antibody function is crucial to the understanding of how toxin-based vaccines give rise to efficacious antibody responses. If binding alone is insufficient, it is a reasonable assumption that antibody binding must disrupt an essential toxin function in order for that antibody to be effective. It has been postulated, for example, that the primary mechanism of action of anti-PA antibodies would be the blockade of binding to the cell surface receptor, and vaccine formulations based solely on the primary receptor binding domain (domain 4) of PA have been proposed (5, 7). Our findings suggest that these assumptions are premature. Only one of the neutralizing antibodies we isolated (4A12) reacts with the domain 4 region of the molecule, and it is unlikely that any of the PA20-specific antibodies interfere with receptor binding. Residues associated with heptamer formation, LF/EF binding, or furin cleavage could also give rise to antigenic epitopes, and antibodies recognizing these epitopes might also be effective in neutralizing toxin. We have initiated studies to determine the mechanism by which each of the antibodies we isolated neutralizes PA-mediated toxicity in order to determine if they function through any of these modalities.
Although a subset of PA20-specifc antibodies are capable of neutralizing PA-mediated toxicity, their dominance in the response following vaccination may nevertheless have negative implications for the protective efficacy of PA-based anthrax vaccines. Only 18% of these antibodies neutralize the toxin, and they are less efficient, requiring a higher concentration to achieve neutralization. In addition, in vivo, these antibodies may be effectively blocked by free PA20. In comparison, the population of PA63-specific paratopes contains a higher ratio of neutralizing specificities, and cell-free PA83 or PA63 encountered in serum would not diminish their effectiveness.
Although other anthrax-derived antigens have been shown to elicit protective immune responses and have been proposed for inclusion in new vaccine formulations, the second-generation anthrax vaccines currently under development are based solely on PA (10). The findings presented here, when considered along with our previous demonstration of a profound domain bias in the PA-specific response toward PA20-specific epitopes in vaccinated humans (16), suggest that factors intrinsic to the immunobiology of PA itself may diminish its effectiveness in inducing toxin-neutralizing antibodies. The mechanisms underlying the domain bias of the antibody response to PA remain unknown. We have postulated (16) that differential antigen processing of free PA20 and cell-bound PA63 may give rise to a preponderance of PA20-specific antibodies following vaccination, and we have suggested that sequence alterations in the furin recognition sequence that render PA protease resistant might produce a more immunogenic form of the PA monomer. Such a design strategy aimed at shifting epitope dominance toward neutralizing determinants might result in a more efficacious vaccine for the prevention of anthrax infection.
We gratefully acknowledge Nanette Bond for assistance with sample collection and Betty M. Ho for critically reading the manuscript.
This work was supported by Public Health Service grants AI57932 and AI066508 from the National Institute of Allergy and Infectious Diseases. This research was conducted in a facility constructed with support from Research Facilities Improvement Program grant C06 RR-16226 from the National Center for Research Resources, NIH.
Editor: A. Camilli
Published ahead of print on 17 February 2009.