Immunogenicity and Protective Efficacy of Bacillus anthracis Poly-γ-d-glutamic Acid Capsule Covalently Coupled to a Protein Carrier Using a Novel Triazine-based Conjugation Strategy*

The capsular polypeptide of Bacillus anthracis is composed of a unique polyglutamic acid polymer in which d-glutamate monomers are joined by γ-peptidyl bonds. The capsule is poorly immunogenic, and efforts at exploiting the polymer for vaccine development have focused on increasing its inherent immunogenicity through chemical coupling to immune-stimulating protein carriers. The usual strategy has employed carbodiimide-based condensing reagents for activation of free α-carboxyl groups, despite reports that this chemistry may lead to chain scission. We have purified the high molecular mass capsule to >95% homogeneity and have demonstrated that the polymer contains >99% poly-γ-d-glutamic acid. The predominant structure of the polymer as assessed by circular dichroism and multiangle laser light scattering was unordered at near-neutral pH. We investigated the effects of various activation chemistries, and we demonstrated that carbodiimide treatment under aqueous conditions results in significant cleavage of the γ-peptidyl bond, whereas scission is significantly reduced in nonaqueous polar solvents, although undesired side chain modification was still observed. An activation chemistry was developed using the triazine-based reagent 4-(4,6-dimethoxy (1,3,5)triazin-2-yl)-4-methylmorpholinium chloride, which allowed for controlled and reproducible derivatization of α-carbonyls. In a two-pot reaction scheme, activated capsule was derivatized with a sulfhydryl-reactive heterobifunctional moiety and was subsequently coupled to thiolated carrier protein. This conjugate elicited very high capsule-specific immune titers in mice. More importantly, mice immunized with conjugated capsule exhibited good protection against lethal challenge from a virulent B. anthracis strain in two models of infection. We also showed, for the first time, that treatment of capsule with carbodiimide significantly reduced recognition by capsule-specific antisera concurrent with the reagent-induced reduction of polymer mass. The data suggested that for vaccine development, maintenance of the high mass of the polymer may be important.

can be lethal and typically occurs through one of three routes: cutaneous, gastrointestinal, or inhalation. Cutaneous anthrax is typically contracted through open skin wounds in contact with infected animal materials such as hides, wool, etc. It is the most common and least severe form of the disease, accounting for 95% of reported cases. Gastrointestinal infection is usually the result of ingestion of undercooked infected meat, and respiratory infection involves the inhalation of airborne spores. High mortality rates are common with the latter two forms. Early symptoms of inhalational anthrax are nonspecific, but death can follow rapidly with little opportunity for therapeutic intervention (2).
Anthrax has long been considered one of the most likely bio-warfare weapons. Its potential use as an agent of bio-terrorism was dramatically highlighted in 2001 by a series of postal mailing attacks in the United States that resulted in 11 cases of cutaneous and 11 cases of inhalational anthrax, including 5 fatalities as a result of the latter (3). The most effective defensive strategy in the face of such an event is the availability of a highly efficacious and well tolerated vaccine.
Two major plasmid-derived bacterial virulence factors have been identified in B. anthracis. The first is the tripartite toxin encoded by plasmid pXO1 that is composed of lethal factor, edema factor, and their associated receptor-binding component, protective antigen (PA) 2 (4). The second is a surface capsular polypeptide composed of poly-␥-Dglutamic acid (PGGA), the product of plasmid pXO2 (5). Historical vaccines such as the Pasteur and Sterne formulations were based on live attenuated strains, but the currently licensed vaccine, Anthrax Vaccine Adsorbed, is a sterile cell-free filtrate of bacterial culture medium adsorbed to an aluminum-based adjuvant. This vaccine has been shown to confer good protection in humans (6), but it is a fairly undefined formulation and requires a 6-injection dosing regimen. Furthermore, there has been an ongoing debate about its safety, although large trials of the vaccine have indicated an acceptable safety level (7,8). As a result of these considerations, several approaches to a second generation vaccine are in development, the majority of which target PA. Such vaccines would provide indirect protection because they would act to neutralize the toxin following the establishment of infection rather than offering sterilizing immunity. A second bacterial component of interest is the capsule. This polypeptide is an attractive vaccine candidate for several reasons. First, acapsular strains of B. anthracis have been shown to be highly attenuated in a variety of animal models (9,10). Second, the ␥-peptidyl linkage and D-amino acid composition of the polymer are unique against the background protein repertoire of the host and would be expected to generate a highly specific immunological response.
Finally, recent studies have demonstrated that antibodies against the capsule are opsonophagocytic, suggesting that vaccination with the capsule may produce a response that would target the bacillus directly and prevent establishment of initial infection, generating sterilizing immunity (11). This hypothesis is further supported by the demonstration of good passive protection in mice administered a capsule-specific monoclonal antibody and challenged in a lethal pulmonary model of infection (12).
Early studies indicated that PGGA isolated from B. anthracis is poorly immunogenic and confers little or no protective efficacy in animal models of infection (13). More recent work has confirmed that the capsule functions as a T-independent antigen (14), although the report by Chabot et al. indicated some protection in mice from a subcutaneous spore challenge with a pXO1 Ϫ , pXO2 ϩ derivative of the Ames strain and enhanced protection against fully virulent pXO1 ϩ , pXO2 ϩ Ames strain when given with PA (15). An immune-enhancing strategy, which has proved successful for related T-independent bacterial polysaccharide antigens, involves covalently coupling the biopolymer to a highly immunogenic protein carrier. These immunoconjugates often produce specific serum antibody titers several orders of magnitude higher than the antigen alone, and immune responses can often be manipulated depending on the choice of carrier or adjuvant (16). Recent literature reports have described this approach for PGGA where the carriers used were bovine serum albumin (BSA) (15), Pseudomonas aeruginosa exotoxin A (17), keyhole limpet hemocyanin (11), and PA (18). Mice immunized with these conjugates produced anti-capsule-specific sera that induced killing of toxin-minus, capsule-positive bacilli in an in vitro opsonophagocytic assay (17); furthermore, mice immunized with PGGA-PA conjugates were protected against lethal challenge by coadministered protective antigen and lethal factor toxin components (18).
Here we describe preparation of highly purified and well characterized conjugates of PGGA covalently coupled to the outer membrane protein complex (OMPC) of Neisseria meningitidis serotype B (19). We chose this carrier based on the fact that it is a component of a currently licensed pediatric vaccine against Haemophilus influenzae type B (20) and has been used as a carrier in investigational vaccines for other bacterial capsular polysaccharides, including several serotypes of Streptococcus pneumoniae (21). Furthermore, OMPC presents certain unique immune-enhancing effects that render it superior to similar proteinbased carriers (22). It was our desire to produce a highly purified PGGA as starting material and to avoid excessive antigen processing during purification and conjugation in the event that conformational epitopes present on high molecular weight material might be important neutralizing determinants. The purification and conjugation protocols described herein met these goals. We developed a coupling strategy that allowed for precise and controllable derivatization of PGGA and OMPC separately, an important consideration in being able to exercise fine control over modification of the antigen and achievable loading levels of polypeptide on carrier. Finally, we tested our conjugates in two mouse challenge models, utilizing vegetative and spore forms of B. anthracis, and we evaluated the passive protective efficacy of the polyclonal animal serum.

MATERIALS AND METHODS
Bacterial Strain, Inoculation, and Culture Methods-B. anthracis strain Ames (United States Army Medical Research Institute of Infectious Diseases) was streaked for isolation onto Miller LB (Difco; 10 g of pancreatic casein digest, 5 g of yeast, 10 g of sodium chloride, 15 g of agar) agar plates from a glycerol freeze-down and grown overnight at 37°C. A single bacterial colony was used to inoculate 5 ml of LB broth in a 50-ml polyethylene disposable centrifuge tube. The culture was incubated overnight in an air shaker at 37°C. NBY/bicarbonate plates were prepared from 8 g of nutrient broth powder (Difco), 12 g of agar, and 3 g of yeast extract in 900 ml of water. The medium was autoclaved, and 100 ml of an 8% 0.45-m filtered sodium bicarbonate solution was then added prior to plates being poured. A total of 20 NBY/bicarbonate plates were streaked using an inoculating loop with the overnight Ames culture, and the plates were incubated at 37°C with 20% CO 2 for 24 -48 h.
Isolation of Crude PGGA-Bacteria were removed from confluent plates using a disposable spreader and resuspended in water. The suspension was autoclaved at 121°C for 1 h and clarified by centrifugation at 11,000 ϫ g for 10 min at 4°C. The supernatant liquid was recovered, and sterility was verified. The extract was buffered by addition of 0.1 volume of 0.1 M Tris, pH 8.0, 0.1 M sodium chloride, 0.01 M magnesium chloride, 0.2% (w/v) sodium azide. Nuclease digestion was performed by addition of 100 units of DNase I (Invitrogen) and 10 units of RNase ONE (Promega, Madison, WI) overnight at 4°C. Trichloroacetic acid (50% w/v) was added to a final concentration of 5%, and the mixture was incubated on ice for 30 min. The preparation was clarified by centrifugation at 12,000 ϫ g for 15 min at 4°C, and the supernatant was recovered. The pH was adjusted to 7 by slow addition of 6 N NaOH, and the material was exhaustively dialyzed against water for 3 days at 4°C. The crude product was lyophilized to dryness, and a second sterility confirmation was performed.
Polishing of Crude PGGA-Crude PGGA was dissolved in water at 2 mg/ml, and the solution was mixed with an equal volume of hydroxyapatite (HPT) chromatography Buffer A (0.004 M sodium phosphate, pH 7.0, 1.0 M NaCl). Sample was applied to a column (25 mm inner diameter ϫ 200 mm) of HPT type II (Bio-Rad) at a flow rate of 8 ml/min on a Waters Delta 600 (Milford, MA) chromatography system. Elution was monitored by absorbance at 215 nm. Nonbound material was washed out with Buffer A, and PGGA was eluted with a linear gradient from 0 to 100% Buffer B (0.4 M sodium phosphate, pH 7.0, 1.0 M NaCl) at 5 ml/min. Fractions (8 ml) were collected during elution, and those containing PGGA by analytical size exclusion chromatography were pooled. The HPT product was concentrated and diafiltered against water using a 5-kDa PES Pellicon XL50 tangential-flow ultrafiltration membrane (Millipore, Billerica, MA) to give the final polished product. Purified PGGA was shell-frozen, lyophilized to dryness, and stored desiccated at Ϫ70°C.
NMR Analysis-All analyses were performed on a 600-MHz Varian NMR instrument. Crude or purified PGGA powder was weighed and dissolved in a fixed volume of D 2 O (99.999%) containing 0.01% Me 2 SO of known concentration for the quantification of polypeptide concentration and purity. The spectral chemical shift was internally referenced with 0.02% d 6 -2,2-dimethyl-2-silapentanesulfonic acid. The acquisition was carried out in 5-mm tubes at a probe temperature of 25°C. NMR diffusion spectra were generated with the pulse sequence, Bi-polar Longitudinal Eddy Current Delay (23), at a pulsed field gradient of 0.18 torr/m and 2-ms duration. Two-dimensional HMBC NMR of PGGA was recorded at 25°C. The spectral width of the proton and carbon dimensions were 4001.6 and 15080.1 Hz, respectively. Because of the limited spectral width, the carbon chemical shift of carbonyl carbon was folded into the spectrum. The size of the acquired spectral matrix was 2416 ϫ 128, and the 90°pulse of proton and carbon was 7 and 17 s, respectively. The delay for the development of one and multiple bound proton-carbon correlation was optimized for 145 and 10 Hz, respectively. The raw data were multiplied with 90°shifted sine wave function along both dimensions. The final spectrum matrix after Fourier transformation was 2048 ϫ 2048.
Determination of Absolute Configuration-Purified PGGA (10 l) was dried under vacuum in a sealed reaction vessel for ϳ20 min using a Pico-Tag Work station (Waters). The residue was solubilized in 400 l of 6 N constant boiling HCl and incubated under nitrogen at 110°C for 20 h. Following hydrolysis, 200 l of HPLC-grade water was added; the reaction tube was vortexed, and the mixture was transferred to a clean vial. The wash step was repeated four times, and the diluted hydrolysate was dried in a speed-vac concentrator. The residue was taken up in 200 l of (R)-2-butanoic HCl (5:1 v/v mixture of R)-2-butanol (Sigma) and acetyl chloride) and incubated at 120°C for 20 min. Residual (R)-butanoic HCl was evaporated under nitrogen. Heptafluorobutyric anhydride, 100 l, was added, and the mixture was incubated at 120°C for 10 min. The solvent was removed under nitrogen, and the N-heptafluorobutyryl (R)-2-butyl diester derivative of glutamic acid was recovered in a mixture of 100 l of ethyl acetate and 100 l of acetic anhydride before being transferred to a gas chromatography (GC) autosampler vial. As controls, D-and L-glutamic acid (Sigma) were prepared at 0.7 g/ml in methanol, and a 25-l aliquot of each was derivatized as for PGGA. GC analysis was performed on a Hewlett-Packard HP6890 with flame ionization detector (FID). The column was an HP-5MS (30 ϫ 0.32 mm) coated with 5% phenyl, 95% dimethylpolysiloxane (Agilent, Palo Alto, CA). Temperatures of injection port and FID were maintained at 250 and 300°C, respectively. The temperature of the GC oven was programmed starting from 80°C, holding for 5 min, ramping at a rate of 8°C/min to 275°C, and holding for 2 min. Helium gas, at a flow rate of 2.2 ml/min, was utilized as the mobile phase. Sample injection volume was 2 l.
Sizing Analyses-Molecular mass determination of PGGA was performed by high performance size exclusion chromatography (HPSEC) coupled with multiangle laser light scattering (MALLS) detection on an 1100 series liquid chromatography system (Agilent) and Dawn TM EOS 18 angle light scattering detector with quasi-elastic light scattering and Optilab TM DSP options (Wyatt, Santa Barbara, CA). Chromatography was performed at a flow rate of 0.5 ml/min using two Ultrahydrogel TM Linear 30-cm columns in series behind an Ultrahydrogel TM guard column (Waters). Running buffer was 50 mM sodium phosphate, pH 7.2, 0.15 M sodium chloride containing 8 ppm Proclin 150 TM as a preservative. Columns and detectors were maintained at 35°C. The refractive index increment (dn/dc) of purified PGGA was determined empirically. Briefly, five PGGA solutions of approximate concentration 0.2-0.8 mg/ml were prepared in phosphate-buffered saline and dialyzed exhaustively against the same buffer. The refractive index of each sample (0.2 m filtered) was measured using a Optilab DSP interferometric refractometer (Wyatt), and concentration was determined from the absorption at 215 nm using an empirically determined extinction coefficient of 7628. The concentration was input into the Dn/Dc software (Wyatt Technology DNDC version 5.90.03) to obtain a value of 0.150 Ϯ 0.002, which was used for all subsequent mass determinations. PGGA sonication studies were performed in 50 mM phosphate, 0.15 M NaCl, pH 7.5, using a Branson Ultrasonics Sonifier (St. Louis MO) with either 2-inch or high intensity cup/horn probes. Sonication was carried out at an output energy of ϳ180 watts.
Amino Acid Analysis-Quantitative amino acid analysis was performed using the AccQTag TM system (Waters). PGGA samples were appropriately diluted in water and hydrolyzed in constant boiling 6 N HCl containing 2% phenol at 110°C for either 20 or 70 h, dried, and reconstituted in 20 mM HCl. Hydrolysates were labeled with AccQFluor TM reagent following the manufacturer's protocol and heated at 55°C for 10 min prior to analysis. Chromatography was performed on an 1100 series HPLC (Agilent) using an AccQTag TM Amino Acid Analysis Column at a flow rate of 1.0 ml/min. Fluorescence was monitored at excitation and emission wavelengths of 250 and 395 nm, respectively.
CD Analysis-CD measurements were performed on a J-810 spectropolarimeter equipped with a Peltier temperature controller set at 25°C (Jasco, Easton, MD). Spectra were collected from 185 to 300 nm at a scan speed of 100 nm/min for 10 accumulations/sample with a data interval of 0.1. The polymer concentration of samples was accurately quantified by amino acid analysis, and the CD signal was converted to molar ellipticity, [⍜], as shown in Equation 1, where MRW is the mean residue weight, which in this case was the polymer repeat unit molecular mass (128.1Da); l is path length in cm, and c is concentration in mg/ml.
Nucleic Acid Analysis-PGGA samples were dissolved in water at a concentration of 1 mg/ml (w/v) and analyzed for total nucleic acid content using the PicoGreen TM fluorescent dye-binding assay (Molecular Probes, Eugene, OR) as per the manufacturer's recommendations. Quantitation was performed by using either calf thymus DNA or total yeast RNA as nucleic acid standards.
Aqueous Carbodiimide Activation of PGGA-Aqueous activation studies were performed by dissolving PGGA to a final concentration of 4 mg/ml in either 0.25 M sodium phosphate, pH 7.3, or 0.25 M MES, pH 4.5. Both buffers contained 0.15 M NaCl and 0.75 mg/ml N-(⑀-maleimidocaproic acid)hydrazide (EMCH, Pierce). 1-Ethyl-3(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC, Pierce) was added to final concentrations of 5 or 0.5 mg/ml. Reactions were allowed to proceed at room temperature for 1-3 h in the dark. At each time point, the reactions were stopped by three consecutive rounds of buffer exchange over a 5,000-Da centrifugal concentrator Samples were analyzed by HPSEC/ MALLS, and aliquots were removed for ELISA testing, snap-frozen in liquid nitrogen, and stored at Ϫ70°C. Development of Nonaqueous Activation Chemistry-In order to conserve purified PGGA, initial studies were performed using poly-␣-Dglutamic acid (PAGA) (Sigma) and subsequently confirmed with PGGA. In order to render the polymers soluble in nonaqueous solvents, they were first converted from the sodium to tetrabutylammonium (TBA ϩ ) salt by chromatography on AG50W-X8 resin (Bio-Rad). Resin (200 -400 mesh, H ϩ form) was washed in distilled deionized water (DI water) and used to pack a 10-ml column that was washed successively with 10 volumes water, 1 M tetrabutylammonium hydroxide, and water. PAGA or PGGA dissolved in DI water (3-5 mg/ml) was applied at 1 ml/min and eluted with DI water. The eluent was dried and product yield determined by weight. The TBA ϩ adduct was reconstituted in anhydrous dimethylformamide (DMF) at a nominal concentration of 10 mg/ml and placed on ice. EMCH was added at a equimolar ratio to total carboxyl content. For comparison of condensing reagents, either N,NЈdiisopropylcarbodiimide (DIPC) or 4-(4,6-dimethoxy(1,3,5)triazin-2yl)-4-methylmorpholinium chloride (DMTMM; Acros Organics, Pittsburgh, PA) dissolved in DMF was added at an equimolar ratio to carboxyl. For the DIPC reaction, 1-hydroxybenzotriazole was added at a 0.1 molar eq to carboxyl. Reactions proceeded in dark under N 2 for 1 h on ice followed by ambient temperature overnight. Activated PAGA was desalted by gel permeation chromatography in DMF using Sephadex LH-20 resin (Amersham Biosciences), and the product was diluted with 4 volumes DI water. The product was dialyzed successively versus 1 M NaCl and DI water and dried under vacuum. For comparison of maleimidation reagents, either EMCH or 5-aminopentylmaleimide (APM) was added to PGGA dissolved in DMF at varying molar ratios (0.2 to 1) to carboxyl. DMTMM (0.2-1 eq) was added, and reactions were carried out under N 2 in dark on ice for 1 h followed by ambient temperature for 3 h. Product was counter-ion exchanged by dialysis as described above, omitting the gel permeation step, and dried under vacuum. For evaluation of DMTMM under aqueous conditions, PGGA was dissolved in 0.1 M HEPES, pH 7.3, 0.5 M NaCl, 2 mM EDTA buffer at 3 mg/ml, with addition of EMCH or APM and subsequent reaction conditions as described above.
Preparation of PGGA-OMPC Conjugate-Purified outer membrane protein complex (OMPC) of N. meningitidis serotype B was activated by thiolation of lysine residues as described previously (24). Total sulfhydryl content of the activated carrier was determined using 5,5Ј-dithionitrobenzoic acid (25). PGGA maleimidation was optimized based on activation studies. Briefly, PGGA (TBA ϩ ) was dissolved in DMF at 5 mg/ml, and EMCH was added at an equimolar equivalent to carboxyl. After 10 min, DMTMM was added at a 0.2 molar eq, and the reaction was purged with N 2 and incubated on ice in the dark for 1 h followed by 2.5 h at ambient temperature. The reaction mixture was diluted 4-fold with DI water and dialyzed as described above in the dark at 4°C, with continuous N 2 purging of the dialysis buffer. The maleimide-adducted product was 0.22-m filtered and buffered to a final concentration of 20 mM HEPES, pH 7.3, 0.5 M NaCl, 2 mM EDTA. Thiolated OMPC was added at an equimolar ratio of sulfhydryl to maleimide, and the reaction was allowed to proceed for 28 h at ambient temperature in the dark. An OMPC-only control was carried forward in parallel. Residual unreacted thiol and maleimide groups were successively quenched using a 5-fold molar excess of iodoacetamide and N-acetylcysteamine, respectively, overnight at ambient temperature in the dark. To remove excess reagents and residual free PGGA, products were pelleted by centrifugation at 289,000 ϫ g for 60 min at 4°C and resuspended in 20 mM HEPES, pH 7.3, 0.15 M NaCl buffer. This step was repeated twice, and the final resuspension was in 0.15 M NaCl. Product and control were centrifuged at 1,000 ϫ g for 10 min to remove any aggregated material. Protein was determined by modified Lowry assay (26). Conjugate and control were adsorbed onto aluminum hydroxyphosphate adjuvant (alum, Merck) and diluted to appropriate PGGA concentrations in sterile saline.
Animal Immunizations-BALB/c mice were immunized by intraperitoneal injection of 0.2 ml of alum-formulated conjugate vaccines.
PGGA ELISA-Serum IgG antibody titers against PGGA were measured by ELISA as follows. Costar high binding plates were coated with 2 g/ml (50 l/well) purified PGGA in phosphate-buffered saline and incubated overnight at 4°C. Plates were blocked overnight at 4°C with 5% fetal bovine serum. Serum samples were tested at 1:5 serial dilutions on the ELISA plates after they were prediluted at 1:10 or 1:100. The plates were then incubated overnight at 4°C. After washing, alkaline phosphatase-labeled goat anti-mouse IgG (Southern Biotech, Birmingham, AL) at 1:2,000 dilution was used to detect bound IgG antibody. The plates were developed using p-nitrophenyl phosphate substrate (Sigma), and absorbance was measured at 405 nm. Titers were calculated as described previously (27).
Animal Challenge Studies-BALB/c mice, immunized as described above, were challenged with either B. anthracis vegetative or spore preparations via the intraperitoneal route. B. anthracis Ames strain spores were maintained at Ϫ70°C. To prepare the vegetative bacillus inoculum, the frozen spore stock material was thawed, and a loop of material was streaked on a sheep red blood agar plate and incubated for 18 h at 35°C. Immediately prior to challenge, the 18-h vegetative bacillus colonies were suspended in saline, vortexed vigorously, and diluted to achieve an inoculum shown previously to provide a specific number of colony-forming units (cfu) per 0.2 ml. To prepare the spore inoculum, the frozen spore stock material was thawed and diluted in sterile water to achieve an inoculum shown previously to provide a specific number of spores per 0.2 ml. Mice were observed multiple times each day for 25 days after the final challenge. In preparation for passive serum transfer studies, mice were immunized with 0.1 g of PGGA-OMPC. Serum was collected from these mice before (preimmune) and 14 days after the immunization. For the final experiment, 0.2-ml aliquots of neat sera, either preimmune or 14-day, was then mixed with 0.1 ml of saline solution aliquots containing 1.5 ϫ 10 4 B. anthracis Ames strain cfu for 30 min at 37°C. Groups of 10 mice were inoculated intraperitoneally with 0.3 ml of preimmune or day-14 sera incubated with B. anthracis. A control group of 18 mice was each inoculated with 0.2 ml of saline that had been preincubated with 0.1 ml of the B. anthracis vegetative preparation. Finally, a final group of 10 mice received the same inoculum of B. anthracis incubated with 1 mg of an IgG1 anti-PGGA mAb. Mice were observed multiple times each day for a 12-day period. BALB/c mice were housed in Microisolator boxes and given water and food ad libitum. All animal studies and housing were approved and conducted following recommendations from the Institutional Animal Care and Use Committee at Merck. Within each challenge experiment, differences in survival distributions among the alternative treatment groups were assessed by the log-rank test using the LIFETEST procedure in SAS (SAS Institute Inc., Cary, NC).

RESULTS
Purification of PGGA-PGGA was isolated from plate-grown cultures of B. anthracis by heat extraction followed by nuclease digestion and an acid precipitation step. The yield of crude product from two independent isolations was ϳ12 mg/plate. Trichloroacetic acid treatment proved to be a highly effective step for removing exogenous protein contaminants as the glutamic acid polypeptide remained soluble in the supernatant. The one-dimensional 1 H NMR analysis of representative crude and purified products (lot 2, see Table 1) is presented in Fig. 1. The ratio of specific PGGA concentration to gravimetric powder yielded a purity of 60 -70% for crude PGGA. Unidentified contaminant bands evident in Fig. 1A were somewhat broad, suggesting that these might be due to residual protein or nucleic acid species. Ultraviolet spectroscopy of the product revealed a broad but minor absorbance centered at ϳ263 nm along with the expected strong absorbance at ϳ220 nm (data not shown), again suggesting the possible presence of nucleic acid. HPSEC analysis suggested that the contaminant peak was of low molecular mass and also possessed 215 nm absorbance. Hydroxyapatite chromatography was investigated as a means for removing this material. PGGA was observed to bind to the resin at high ionic strength and could be eluted with a linear gradient of increasing phosphate. The purity of fractions was assessed by analytical HPSEC, and it was found that the contaminants were quantitatively removed in the unbound flow-through fraction. Tangential flow ultrafiltration was used to remove salts and other residual low mass contaminants, and the product was stored as a freeze-dried powder. The 1 H NMR analysis of polished material (Fig. 1B) yielded powder purity of 96%. Purified PGGA was soluble in water at 3-5 mg/ml, showing a concentration-dependent viscosity that could be significantly decreased by raising the ionic strength of the medium as described previously for polyglutamate polymers (28,29). The complete analytical profile for two representative preparations is summarized in Table 1. Most surprisingly, nucleic acid levels as measured by the Pico-Green fluorescent dye binding were very low for both crude and purified PGGA. We showed using PAGA that the presence of high levels of polypeptide did not reduce the response of either DNA or RNA in the assay. Although Pico-Green is typically used for DNA quantification, we found that RNA was detectable as well. The assay showed that purified PGGA preparations did not contain significant amounts of either nucleic acid. Amino acid analysis suggested that the unidentified contaminants were most likely protein or peptide fragments that co-purified through the trichloroacetic acid and dialysis steps. Although these were of low mass, it is conceivable that they were ionically bound to the polymer and removed during hydroxyapatite processing as a result of the high ionic strength of the loading buffer. The high level of purity determined by NMR was confirmed by amino acid analysis, which showed glutamic acid as the only detected residue (data not shown).
PGGA Structural Analysis-The unambiguous determination of polypeptide linkage was confirmed by two-dimensional HMBC NMR (30) at 25°C. Fig. 2A shows the spectral results with peak assignments labeled. At the chemical shift of an amide carbonyl carbon (ϳ178 ppm, spectral folding due to limited spectral width of the indirect carbon dimension), its correlation with H-␥ can be observed ( Fig. 2A, shaded  box). The cross-peak on HMBC confirms the backbone linkage through the ␥-carboxylic group. Determination of the absolute configuration of glutamic acid in PGGA (Fig. 2B) was based on the analysis of diastereoisomers by GC-FID (31) and was accomplished through polypeptide hydrolysis, derivatization, and GC-FID analysis. Both D-and L-glutamic acid, purchased commercially, were used as reference for the identification. The small impurity peaks in the chromatography arose from the impurity, (S)-2-butanol, in (R)-2-butanol. For D-Glu, the main peak comprised 89.8% of the total area. The purified PGGA sample, which co-eluted with this peak, accounted for 89.2% of the total area, yielding a calculated value of 99.2% D-Glu present in the polymer. CD analysis of purified PGGA (Fig. 2C) shows the overall conformation for both lots to be similar. These spectra are dominated by a large positive ellipticity at ϳ197 nm and a single negative minima at 217 nm. For typical L-amino acid homopolymers and reference proteins, this pattern corresponds to a class A spectrum and is characteristic of a ␤-sheet conformation (32). However, polymers containing all D-amino acids exhibit a CD spectra, which is the exact mirror image of the corresponding L-form (33, 34). As  such, the spectra for the PGGA lots correspond to a class U, or unordered conformation.
Sizing Characterization of PGGA-The results of HPSEC/MALLS analysis on crude and purified PGGA are given in Table 2. The average mass of the product is high, suggesting that the isolation conditions do not have a negative impact on the native polymer in contrast to previous reports where purified PGGA had a molecular mass of less than 100,000 Da (13). Although the distribution showed some degree of heterogeneity, the symmetry of the chromatographic peaks was generally good, and 75% of the mass was found to be between 100,000 and 1,000,000 Da. The behavior of lot 2 following purification is anomalous as the polymer appears to increase in mass. Analysis of molar mass as a function of elution time indicated that the polymer mass initially decreased across the peak, as expected, and then showed an abruptly sharp increase on the tailing end (data not shown). This type of behavior is indicative of nonspecific interactions between the highly charged polymer and the chromatographic resin that can affect the average calculated mass. The sample concentrations determined from the dn/dc agree well with the independent NMR determinations. Plots of the root mean square radius versus molecular mass are indicative of the structure of the molecule in solution (35). The slopes of these plots fall in the range of 0.62 to 0.67 and suggest an extended random coil conformation (data not shown). The relationship of the root mean square radius to the hydrodynamic radius in each case also suggests that the molecules are present in a random coil configuration. This finding is consistent with our CD analysis as well as with various literature reports of polyglutamic acids at neutral pH (36,37).
PGGA Activation Studies-Historic approaches for conjugation of polyglutamic acids utilized carboxyl activation mediated by water-soluble carbodiimides such as EDC (38). However, the resultant mass of capsule in these studies was often quite low. To investigate the effect of carbodiimide exposure on mass and immunogenicity, we treated PGGA with EDC under aqueous conditions at both pH 7.2 and 4.5, varying temperature, reaction time, and reagent concentration. For mass determination, aliquots were removed at given time intervals, and the reaction products were analyzed by HPSEC/MALLS. Fig. 3A shows a representative analysis examining the effect of room temperature treatment with 0.5 or 5 mg/ml EDC at pH 4.5. The average molecular mass relative to untreated control was reduced within 1 h at both concentrations of EDC, although the size decrease was more pronounced at the lower concentration. To determine whether mass reduction related to EDC treatment might detrimentally affect the ability of the capsule to function as an effective immunogen, we analyzed the reaction products for their ability to bind a mouse monoclonal capsule-specific hybridoma supernatant using a competitive ELISA. This antibody had been raised against a capsule-conjugate immunogen. The results from these experiments are shown in Fig. 3B. Carbodiimide treatment of PGGA resulted in a decreased ability of the polymer to bind to the antibody, which correlated with both the decrease in mass and length of EDC treatment. Similar results were obtained for EDC treatments performed at pH 7.2, room temperature; pH 4.5, 4°C; and pH 7.2, 4°C (data not shown). In all cases, a time-dependent reduction in PGGA mass was observed, with the most dramatic decrease occurring at room temperature, pH 7.2. Most interestingly, although mass reduction appeared greater at pH 7.2 using 5 mg/ml EDC, at pH 4.5 the size decrease was more pronounced at 0.5 mg/ml EDC at either temperature. In all cases, mass reduction increased with longer reaction times. Similarly, loss of binding to the PGGA-specific monoclonal antibody was seen under all treatment protocols (data not shown). In most cases an increase of the IC 50 by ϳ10fold was observed; however, the samples prepared at pH 4.5 invariably showed much more dramatic effects on antibody binding. In several cases the relative IC 50 was increased by greater than 100-fold. This is particularly striking, because the relative mass reduction of these samples was not significantly different from their pH 7.2 counterparts. To confirm further that polymer mass reduction correlated with decreased immunogenicity, we subjected PGGA to controlled sonication to generate a range of mass species and tested these in the competitive ELISA. We again observed a similar size dependence, with a 10-fold decrease in mass resulting in Ͼ50% decrease in antibody binding (data not shown). Stock solutions of PGGA in water (1 mg/ml) were adjusted to 10 mM sodium phosphate, pH 7.2, and spectra were acquired at 25°C. The polymer concentration was calculated by amino acid analysis, and the spectrum was converted to molar ellipticity as described under "Materials and Methods." To avoid the aqueous carbodiimide-induced mass reduction, we evaluated the efficiency of activation under nonaqueous conditions. A variety of solvents including methanol, DMF, and 1-methyl-2-pyrrolidinone was evaluated for direct solubilization of either PGGA or PAGA, but none proved useful. Counter ion-exchange of sodium ions for more hydrophobic H ϩ or TBA ϩ ions followed by lyophilization was found to render both polymers soluble in DMF at concentrations of 5 mg/ml. HPSEC/MALLS showed that the mass of PGGA was best maintained after TBA ϩ treatment, so this was used for subsequent experiments. In order to conserve purified PGGA, we used PAGA as a model for activation studies, and we evaluated DIPC as a representative carbodiimidebased condensing agent and the novel triazine-based compound DMTMM. In both cases, the heterobifunctional compound coupled to the polyglutamic acid was EMCH, which contained a carbonyl-reactive hydrazide moiety and a thiol-reactive maleimide functionality separated by a six-carbon atom spacer. Following coupling, excess reagents were removed by gel permeation chromatography in DMF, and the desalted product was diluted into water and dialyzed successively against 1 M NaCl followed by water prior to drying. This step was found to be necessary to exchange the activated polymer back to the Na ϩ form and render it water-soluble for subsequent conjugation. Fig. 4 shows the 1 H NMR profiles of DIPC and DMTMM-activated PAGA samples. The derivatization level or side chain loading (SCL) percentage was defined as the percentage of derivatized ␥-glutamate per polypeptide repeating units and was determined from the peak integral ratio of maleimidederivatized side chain to H-␣. The covalent linkage of EMCH to polypeptide was confirmed by the use of diffusion edited spectra, as shown in the lower trace of Fig. 4, A and B. Under Brownian motion, small molecules such as HDO and Me 2 SO diffuse much faster than polymer, disappearing from the diffusion edited proton NMR spectrum, whereas covalently linked derivatived groups remain. The identification of maleimide in the pre-and post-diffusion edited spectra confirm the covalent nature of the EMCH addition.
At a given loading ratio and reaction time, both DIPC and DMTMM afforded ϳ3% SCL as determined by quantification of the maleimide peak, but carboxyl derivatization by isopropyl groups was measured at 5.4% for the DIPC-mediated reaction (Fig. 4A), suggesting that nonproductive adduction of reagent was occurring, possibly through formation of the stable N-acylisourea (39). In contrast, the spectrum for DMTMM-mediated coupling was much cleaner, showing only maleimide and proton resonances derived from the caproic acid spacer (Fig.  4B). HPSEC/MALLS analysis of the reaction products showed no cleavage of derivatized PAGA under these conditions (data not shown).
Next, a series of small scale PGGA activations were carried out under nonaqueous conditions using DMTMM and either EMCH or APM as the heterobifunctional activator. For this and subsequent work, the gel permeation step was eliminated, and both reagent removal and counter ion exchange were accomplished by dialysis. The results are summa-   FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 rized in Table 3. It was observed that the degree of activation could be adjusted by varying either the molar ratio of DMTMM relative to carboxyl or the overall reaction time. Because DMTMM was reported to be useful for amide synthesis under aqueous conditions, we also investi-gated this route. PGGA was solubilized in 20 mM HEPES, pH 7.3, 0.15 M NaCl buffer and reacted with either EMCH or APM in the presence of 0.5 molar eq of DMTMM. The maleimide incorporation was less than 0.5% in each case, although the amount of carboxyl derivatization was   . PAGA activation under nonaqueous conditions. One-dimensional 1 H NMR was used to assess EMCH derivatization of PAGA mediated by DIPC (A) or DMTMM (B). For reaction, PAGA was converted to the TBA ϩ ion form and dried. The powder was solubilized in DMF, and EMCH was added at a 1 molar eq to carboxyl groups. For the carbodiimide reaction only, 1-hydroxybenzotriazole was added at a 0.1 molar eq to carboxyl. DIPC or DMTMM was added at a 1 molar eq to carboxyl, and the reaction was allowed to proceed under N 2 for 1 h on ice and then overnight at ambient temperature. Reaction products were desalted by gel permeation chromatography in DMF and dialyzed successively against 1 M NaCl and water, followed by drying. The top trace in each panel shows the regular proton spectrum, and the bottom trace is the diffusion edited spectrum. Peak assignments for PAGA protons are labeled along with the maleimide reporter peak derived from EMCH. Other nonassigned EMCH-derived shifts are labeled with asterisks. The residual isopropyl groups of the adducted DIPC are noted in A.

Triazine-based Polyglutamic Acid Conjugates
comparable with equivalent nonaqueous conditions. However, sizing analysis showed that the polymer mass had been reduced by 40-fold. Reducing the reaction time by half lowered the derivatization, but the maleimide incorporation was still low. Polymer scission also appeared to be time-dependent, because the relative mass reduction was 20 -25fold at 3 h.
PGGA-OMPC Conjugates-The use of a two-step conjugation protocol in which PGGA and carrier were independently and differentially activated prior to mixing enabled the subsequent coupling to be directed and more highly controlled than would be possible in a one-pot preparation. The heterobifunctional chemistry enabled by use of EMCH was preferable to previous reports where homobifunctional reagents such as adipic acid dihydrazide were used for activation (17). The disadvantage to this latter method is that the potential for inappropriate cross-links is significantly increased. OMPC was rendered maleimide-reactive by thiolation of surface-accessible lysine residues (24), a process that was highly reproducible and typically yielded 0.2-0.3 mol of thiol/mg of protein with total protein recoveries of 60 -70%. Activation of PGGA in DMF using a 1 mol eq of EMCH and 0.2 mol eq of DMTMM, relative to total PGGA carboxyl, provided 10.2% maleimide SCL with 14.0% total carbonyl modification for one lot and 15.7% SCL with 21.7% derivatization for a second lot. A slight mass reduction of ϳ2-fold was observed during the activation process, but the final polymer mass was Ͼ350,000 Da for both lots. Determination of conjugation efficiency was performed by comparison of the glutamate content between conjugate and OMPC-only control. To determine whether noncovalent association of PGGA and OMPC was a factor, a portion of the OMPC-only control was mixed with nonmaleimidated PGGA and incubated 1 h prior to the reagent removal steps. Amino acid analysis showed no increased glutamate content in the physical mix relative to the OMPC-only control. The percent incorporation by mass of PGGA in the conjugate was 8.5 and 10.1%, respectively, for two independent lots.
Mouse Immunization and Protection-A dose-ranging study (Fig. 5) was performed by immunizing groups of 10 BALB/c mice with increas-ing amounts of PGGA-OMPC conjugates at 2-week intervals using a 3-dose vaccination regimen. Controls received either aluminum adjuvant-adsorbed unconjugated PGGA (10 g) or OMPC alone. The OMPC dose was matched with that in the highest dose of conjugate vaccine (ϳ50 g). Mice responded rapidly to the first dose of PGGA-OMPC conjugate vaccines. There was little difference in response between the 10-and 1.0-g dose levels, whereas the response to 0.1-g dose was ϳ10-fold lower than that elicited by the 10-g dose level. Antibody titers rose ϳ10-fold after the second and third doses. By contrast, mice injected with unconjugated PGGA (formulated on aluminum adjuvant) generated a modest response to the first injection that declined after subsequent doses. The results indicated that conjugation of PGGA to the potent OMPC carrier resulted in a markedly enhanced immune response.
In order to demonstrate that the conjugation method was reproducible, a second conjugate was prepared and compared with the first for immunogenicity. In this experiment (data not shown), the dose ranging was extended to 0.001 g/mouse. Mice responded well to a single dose of either vaccine lot. The 0.001-g dose of conjugate induced a response equivalent to that generated with a 1.0-g dose of unconjugated PGGA, and the geometric mean titer induced with the 0.1-g dose of PGGA-OMPC conjugate vaccine was 100-fold higher than that elicited with unconjugated PGGA. Overall, there was little difference in potency between the two conjugate preparations.
B. anthracis Challenge Studies-In a series of experiments to explore vaccine efficacy in the mouse model, mice were immunized and then challenged with either vegetative bacilli or spores. Groups of 9 -10 mice from the experiment shown in Fig. 5 were challenged 2 weeks after the third dose of vaccine with 10 3 cfu (day 0) and 5 ϫ 10 3 cfu (day 12) of vegetative bacilli. The number of surviving mice was assessed daily until day 25. As shown in Fig. 6, all mice immunized with 0.1-10 g of the PGGA-OMPC conjugate vaccine survived the 10 3 cfu challenge, and 100% of mice immunized with the highest dose (10 g) of conjugate survived after the 5 ϫ 10 3 cfu challenge dose. There was no statistical evidence of a difference in survival distribution among the three PGGA-OMPC conjugate vaccine groups (p ϭ 0.60). By contrast, 30% of mice immunized with unconjugated PGGA survived the first challenge dose, and only 20% remained alive after the second dose. Although 40% of mice immunized with OMPC survived the first challenge, none survived the second. Although the survival distributions of the unconjugated PGGA-and OMPC-only groups were not statistically different from one another (p ϭ 0.87), they each were highly statistically different  from each of the PGGA-OMPC conjugate vaccine groups (p Յ 0.002). These results indicated that mice immunized with even the lowest dose of PGGA-OMPC conjugate vaccine were well protected from a lethal challenge with vegetative bacilli.
Additional experiments were conducted to evaluate protection against B. anthracis spore challenge. BALB/c mice were immunized on day 0 and day 14 with 0.01-1.0 g of PGGA-OMPC conjugate vaccine, and serum was collected on days 14 and 28 for ELISA. As shown in Table  4, there was a dose-dependent antibody response to the conjugate vaccine with titers ranging from 62,602/ml to the lowest (0.01 g) vaccine dose to 2,070,645/ml to the highest (1.0 g) vaccine dose. Approximately 8 weeks after the second vaccine dose, all four groups of mice were challenged with a subcutaneous injection of 2 ϫ 10 4 B. anthracis spores. Survival was monitored for 2 weeks. Vaccination with even the lowest dose (0.01) g of the vaccine resulted in good (80%) protection from challenge with spores from a virulent strain of B. anthracis.
Passive Protection with Immune Serum-Serum collected from BALB/c mice before (preimmune) and 14 days after a single immunization with 0.1 g of PGGA-OMPC conjugate vaccine was pooled and tested for passive protection of naive BALB/c mice from B. anthracis vegetative challenge. The ELISA titer of preimmune serum was Յ5 per ml, whereas the titer of the pooled post vaccination immune serum was 12,161 per ml. For this experiment, 0.2 ml of neat serum was mixed with 0.1 ml of vegetative bacilli (0.1 ml ϭ 1.5 ϫ 10 4 cfu) and incubated for 30 min at 37°C. Groups of 10 mice were then inoculated with 0.3 ml of the mixture and observed for 12 days. A separate group of 10 mice received an inoculum of 1 mg each of an IgG1 anti-PGGA mAb (ELISA titer ϳ10 8 per mg) that had been preincubated for 30 min with 1.5 ϫ 10 4 vegetative bacilli. A group of 18 unimmunized control mice was also injected with 1.5 ϫ 10 4 vegetative B. anthracis bacteria that had been preincubated with saline for 30 min prior to challenge. As shown in Fig.  7, 40% of mice challenged with B. anthracis that had been preincubated with immune serum or with the anti-PGGA mAb survived for the length of the experiment (12 days). There was no statistical evidence of a difference in survival distribution between the immune serum and anti-PGGA mAb groups (p ϭ 0.89). In contrast, no mice in the preimmune serum treatment group survived beyond day 4, and only 11% of control mice survived to day 12. The survival distributions of the preimmune serum and nonimmunized control groups were not statistically different from one another (p ϭ 0.08), whereas the immune serum and anti-PGGA mAb groups each were highly statistically and significantly different from the nonimmunized control group (p Յ 0.005). The results suggested that protection induced by vaccination with the PGGA-OMPC conjugate vaccine was antibody-mediated.

DISCUSSION
This study demonstrates for the first time that a covalently conjugated form of the native B. anthracis polypeptide capsule is able to elicit an immune response capable of protecting experimentally vaccinated animals from a lethal bacterial challenge. It has been reported that the isolated capsule is only weakly immunogenic (14,17,40), behaving in that respect similarly to the polysaccharide capsules of organisms such as pneumococci and meningococci (41). Recent studies employing conjugated forms of either PGGA derived from B. lichenformis (18) or synthetic poly-␥-D-glutamic acid peptides (17) demonstrated the enhanced immunogenicity of these formulations and showed in the latter study the resultant sera capable of opsonophagocytic killing in vitro. However, neither of these demonstrated protective efficacy of the vaccine candidates.
The ability to maintain PGGA capsular antigen in a form that closely mimics the native structure on the surface of the bacillus may well be a critical consideration for its effectiveness at inducing a protective response. It is well documented that polyglutamic acids can exist in defined structural conformations (36,37,42,43) and that these conformations are sensitive to the effects of pH and ionic strength (28,29,36). It is likely that such extended and regularly repeating structures could be disrupted by random cleavage of polymer chains, particularly when the molecular mass is reduced below a certain critical degree of polymerization. It has been postulated that some bacterial polysaccharide capsular antigens may contain conformational epitopes that are mass-dependent, and the correlation of an effective immunologic response with  high mass has been demonstrated for meningococcal type B polysialic acid (44) and for other capsular antigens (45)(46)(47). In order to minimize perturbation to the native capsule, we felt it was important to develop a purification scheme that retained as closely as possible the native size in the event that mass-dependent conformational effects were important for induction of neutralizing antibody. Our process yielded a highly purified product of Ͼ400,000 Da from two independent isolations, in contrasted with previous reports in which PGGA was recovered with a mass of Ͻ100,000 (13). The chemical identity and homogeneity of the PGGA preparation were confirmed by NMR and amino acid analysis, and the unambiguous structural determination verified the ␥-carboxyl isopeptide linkage. In agreement with previous reports (13,36), GC-FID analysis showed that Ͼ99% of the polymer was composed of D-glutamic acid. This isomeric homogeneity appears to be a unique characteristic of B. anthracis and has further implications for vaccine antigen development in that others have proposed use of polyglutamic acids derived from bacilli including Bacillus subtilis and Bacillus lichenformis, despite the fact that their capsules are often a mixture of D-and L-isomers (28,36). The use of HPSEC/MALLS allowed for accurate mass determinations independent of column migration behavior, and the detailed mass studies of crude and purified product were instrumental in both furthering the characterization of this antigen and in demonstrating the effect of size reduction on its immunological profile. To our knowledge, this is the first reported measurement of hydrodynamic radius for PGGA based on light scattering data. The findings are in agreement with the CD spectra and show the molecular conformation to be that of an extended random coil at neutral pH and low ionic strength, consistent with historic studies employing both CD and optical rotary dispersion analyses (36,37,42). Most interestingly, the hydrodynamic radius did not show any appreciable change at pH 4.5 for the control in the EDC study. Balasubramanian et al. (36) showed the ellipticity increase at 210 nm closely followed the proton titration curve of the carboxylic acid. At pH 4.5, that dissociation factor was still Ͼ0.5, indicating that, overall, the molecule was still in an ionized state. The incomplete coil-to-helix transition may account for the similar radius values. It should be noted that the optical measurement techniques yield an average conformation over the molecule and do not preclude the possible existence of shorter segments that contain a more defined structure. The study of Kimura et al. (37) demonstrated that helical transition in low mass L-polyglutamic acids (Ͻ25 kDa) followed a stepwise route, with small stretches of localized helix eventually combining to form long extended rods. It is possible that in the high mass native PGGA local environmental effects can favor and stabilize regions of secondary structure.
We have now shown definitively that aqueous carbodiimide treatment significantly reduces both molecular mass and antigenicity of the PGGA polymer. A study by Paterson and Leach (38) had demonstrated previously carbodiimide-induced cleavage of poly-L-glutamic acid at pH 5.0 in the presence or absence of ethanolamine. In discussing putative scission mechanisms, they suggested that a likely route was attack by free carboxyl groups on the O-acylisourea intermediate, basing this on the observation that cleavage was decreased in the presence of ethanolamine, which would suggest that as amide formation proceeded, fewer reactive free carboxyls were available for backbone cleavage. As their native polyglutamic acid mass was ϳ25,000 Da, it can be reasoned that in a polymer such as PGGA, with a mass 10-fold higher, the proportion of reactive groups is significantly increased. Furthermore, at high concentrations of the carbodiimide, the potential for water-mediated conversion of the O-acylisourea back to the reactive carboxyl is greatly increased, consistent with the observed increase in scission at the higher EDC concentrations of our study. Although we do not propose a mechanism for the activation-induced cleavage of PGGA, the reaction appears to proceed readily under aqueous conditions. Carboxyl activation with EMCH in DMF did not result in chain scission using either DIPC or DMTMM, although removal of the intermediate was less efficient with the carbodiimide reagent. This lends convincing support to the postulated role for water playing a direct role in the cleavage reaction, as outlined above. Alternatively, a study of ␥-glutamyl peptide bonds in collagen suggested side reactions leading to internal formation of glutarimide or acyl diketopiperazine ring structures in the presence of water-soluble carbodiimide. Such structures may be more susceptible to cleavage with subsequent chain scission (48). The observation that at high concentrations of EDC the mass reduction was less pronounced at pH 4.5 argues that increased protonation of the carboxyl groups has a protective effect. This would be contradictory to the putative mechanism where the protonated carboxylic acid is the reactive species. However, at the 0.5 mg/ml EDC concentration, mass reduction was more pronounced at pH 4.5. Furthermore, the degree of antibody binding was more significantly reduced for PGGA treated at pH 4.5, regardless of reagent concentration or temperature, suggesting that side chain modification may proceed without resultant polymer scission.
The observed lability of PGGA in the presence of carbodiimides has considerable implications for several recent studies in which this method of activation was employed to prepare carrier-capsule conjugates. Recently, Chabot et al. (15) reported that although native PGGA showed some efficacy against a toxin Ϫ , capsule ϩ B. anthracis strain, the protective efficacy was completely abrogated following conjugation of the capsule to BSA using EDC at pH 4.7. This was despite the fact that the measured PGGA-specific IgG titers were increased by nearly 20-fold. This provides compelling evidence that a conformational component to the immune response may indeed be critical for protection and is consistent with our current observations that EDC-mediated reduction in capsule mass and treatment at low pH results in altered immunogenicity.
PGGA covalently coupled to OMPC by the DMTMM method described here was a highly effective immunogen, inducing polymerspecific titers in mice that were at least 3 logs higher than PGGA alone at a 100-fold lower dose. The immune response to the conjugate was observed to increase with subsequent boosting, although no increase was seen for PGGA alone post-dose 2 or 3 in the initial dose-ranging study. Most surprisingly, even a single injection of 0.01 g of conjugated polymer provided a very robust response compared with 1.0 g of unconjugated antigen. It is significant to note that the PGGA-OMPC conjugate raised high levels of IgG and displayed greater than 80% efficacy in both vegetative bacillus and spore challenge models, whereas unconjugated PGGA provided only 20% protection. A previous report comparing nonconjugated with BSA-conjugated PGGA found that IgM-specific titers were lower in animals vaccinated with conjugate, although the IgG titers were also significantly higher (15). However, because in this case the conjugate-immunized animals showed a complete lack of protection, the authors postulate that IgM may be the protective entity and that protective epitopes may be destroyed by the conjugation procedure. On the contrary, the positive correlation between IgG and protection demonstrated in our study clearly shows that IgG antibodies are protective. The discrepancy between the two studies can be explained by the differences in the conjugation procedures. In the previous study, the nonprotective BSA-PGGA conjugate was prepared using EDC-mediated coupling, which we have shown to significantly affect antigen mass and integrity resulting in decreased reactivity with antiserum to native PGGA. It should be recalled that in FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 the previous study the nonconjugated PGGA was protective against a pXO1 Ϫ , pXO2 ϩ strain but not wild type, whereas our conjugate vaccine showed good efficacy against the fully virulent wild type strain (pXO1 ϩ , pXO2 ϩ ). Recent reports using conjugated PGGA (11,17,18) demonstrate a robust IgG response induced by the peptide-carrier conjugate. An antibody-mediated basis of protection is strongly supported by our demonstration of passive protection in mouse transfer experiments, as well as by previous studies that showed that conjugate-induced anti-PGGA immune sera contains opsonophagocytic activity against pXO1 Ϫ , pXO2 ϩ strains of B. anthracis (17).

Triazine-based Polyglutamic Acid Conjugates
The ability to convert T-independent antigens such as polysaccharides and D-polypeptides to T-dependent ones capable of inducing mature IgG responses is strongly related to the choice of protein carrier (49). In this respect, OMPC offers certain advantages. First, OMPC is a component of a licensed pediatric vaccine and has a proven safety record (50). Second, in a comparison study, Donnelly et al. (51) demonstrated that in addition to the usual carrier-mediated immunoglobulin class switching, OMPC also demonstrated potent adjuvant capabilities, likely related to B-cell activation by its lipopolysaccharide component. It is postulated that this is the reason why OMPC-polysaccharide conjugates induce high serum antibody titers following one-dose immunizations, whereas other carriers typically require a three-dose regimen. The data from the present study extend that finding to D-polypeptides. Third, recent studies indicate the involvement of Toll-like receptors 2 and 4 in the immunopotentiating effects of Neisseria outer membrane protein preparations, including OMPC (52,53). Such effects are mediated by the major membrane porins or associated lipopolysaccharide, or both, and lead to the activation of B-cells through a MyD88-dependent mechanism (52). Finally, the activation and conjugation scheme described here allowed for the covalent coupling of high mass PGGA, in contrast to earlier studies where carbodiimide-mediated coupling would inadvertently reduce the mass (15,17) or where deliberate mass reduction was employed to allow effective coupling (18).
In conclusion, we have demonstrated that both active and passive protection can be induced with a B. anthracis capsular polypeptide-OMPC conjugate vaccine and that mice could be protected from challenge with either vegetative bacteria or spores from the fully virulent Ames strain. The use of covalent OMPC conjugates is a highly effective means of generating a robust IgG-based immune response to the otherwise poorly immunogenic native polypeptide. The sizing and activation studies described here clearly demonstrate that excessive perturbation of the native polymer structure can be detrimental to the elicitation of a protective antibody-mediated response. This provides indirect evidence for the existence of conformational epitopes within the repeat structure of the polymer, analogous to those proposed for certain bacterial capsular polysaccharides. Future NMR studies on highly purified PGGA could offer insight into the nature of such structures. In addition, carefully controlled derivatization could be helpful in eliciting what types of modifications are detrimental to these protective epitopes.