Effects of Spontaneous Deamidation on the Cytotoxic Activity of the Bacillus anthracis Protective Antigen*

Protective antigen (PA) is a central virulence factor of Bacillus anthracis and a key component in anthrax vaccines. PA binds to target cell receptors, is cleaved by the furin protease, self-aggregates to heptamers, and finally internalizes as a complex with either lethal or edema factors. Under mild room temperature storage conditions, PA cytotoxicity decreased (t½ ≈ 7 days) concomitant with the generation of new acidic isoforms, probably through deamidation of Asn residues. Ranking all 68 Asn residues in PA based on their predicted deamidation rates revealed five residues with half-lives of <60 days, and these residues were further analyzed: Asn10 in the 20-kDa region, Asn162 at P6 vicinal to the furin cleavage site, Asn306 in the pro-pore translocation loop, and both Asn713 and Asn719 in the receptor-binding domain. We found that PA underwent spontaneous deamidation at Asn162 upon storage concomitant with decreased susceptibility to furin. A panel of model synthetic furin substrates was used to demonstrate that Asn162 deamidation led to a 20-fold decrease in the bimolecular rate constant (kcat/Km) of proteolysis due to the new negatively charged residue at P6 in the furin recognition sequence. Furthermore, reduced PA cytotoxicity correlated with a decrease in PA cell binding and also with deamidation of Asn713 and Asn719. On the other hand, neither deamidation of Asn10 or Asn306 nor impairment of heptamerization could be observed upon prolonged PA storage. We suggest that PA inactivation during storage is associated with susceptible deamidation sites, which are intimately involved in both mechanisms of PA cleavage by furin and PA-receptor binding.

Protective antigen (PA) is a central virulence factor of Bacillus anthracis and a key component in anthrax vaccines. PA binds to target cell receptors, is cleaved by the furin protease, self-aggregates to heptamers, and finally internalizes as a complex with either lethal or edema factors. Under mild room temperature storage conditions, PA cytotoxicity decreased (t1 ⁄ 2 ≈ 7 days) concomitant with the generation of new acidic isoforms, probably through deamidation of Asn residues. Ranking all 68 Asn residues in PA based on their predicted deamidation rates revealed five residues with half-lives of <60 days, and these residues were further analyzed: Asn 10 in the 20-kDa region, Asn 162 at P 6 vicinal to the furin cleavage site, Asn 306 in the pro-pore translocation loop, and both Asn 713 and Asn 719 in the receptor-binding domain. We found that PA underwent spontaneous deamidation at Asn 162 upon storage concomitant with decreased susceptibility to furin. A panel of model synthetic furin substrates was used to demonstrate that Asn 162 deamidation led to a 20-fold decrease in the bimolecular rate constant (k cat /K m ) of proteolysis due to the new negatively charged residue at P 6 in the furin recognition sequence. Furthermore, reduced PA cytotoxicity correlated with a decrease in PA cell binding and also with deamidation of Asn 713 and Asn 719 . On the other hand, neither deamidation of Asn 10 or Asn 306 nor impairment of heptamerization could be observed upon prolonged PA storage. We suggest that PA inactivation during storage is associated with susceptible deamidation sites, which are intimately involved in both mechanisms of PA cleavage by furin and PA-receptor binding.
The Gram-positive spore-forming bacterium Bacillus anthracis, the causative agent of anthrax, produces a bipartite A/B-type toxin. The B subunit is the 83-kDa protective antigen (PA) 3 receptor-binding moiety (named for its use as a vaccine), and the two catalytic A subunit moieties are edema factor (EF; 89 kDa) and lethal factor (LF; 90 kDa) (1). EF is a Ca 2ϩ -and calmodulin-dependent adenylate cyclase (2). LF is a Zn 2ϩ protease that cleaves and inactivates mitogen-activated protein kinase kinase-1 and -2 (3,4). Following PA binding to cell receptors (5)(6)(7)(8), it is cleaved by a furin family protease (9) to a 20-kDa N-terminal fragment with no known further function and to a 63-kDa fragment that forms ring-shaped heptamers on the cell surface (10). The heptamers then bind up to three molecules of EF or LF (11,12). Following binding, the heptamer-LF/EF complex undergoes endocytosis, and LF or EF is released to the cytosol (13)(14)(15). The crystal structure (16) and functional studies (11,12,(17)(18)(19)(20) have demonstrated that the PA polypeptide is folded into four distinct domains with well defined functions. Domain I (residues 1-258) prevents premature PA polymerization and harbors the furin cleavage site, which is located in an unstructured flexible loop; domain II (residues 259 -487) is involved in heptamerization and is in the membrane insertion loop; domain III (residues 488 -595) is also involved in heptamerization (18); and domain IV (residues 596 -735) is the C-terminal receptor-binding domain, with a small and a large subdomain loop. Mutagenesis studies implied the involvement of the small loop (residues 679 -693), but not the large one (residues 704 -722), in PA-receptor binding (21). Other studies have shown the involvement of the large loop in binding (20). Recent x-ray crystallography studies of the PA-receptor complex revealed two contact regions of PA with the receptor, residues 681-688 and 714 -716, suggesting the contribution of both the small and large loops, respectively, in PA binding (22).
The currently approved human vaccines against anthrax are prepared from crude bacterial supernatant enriched with PA in addition to other B. anthracis-related proteins such as LF and EF (23,24). Previous studies have shown that protective immunity to anthrax disease correlates with induction of neutralizing anti-PA antibodies (25)(26)(27). "Second generation" recombinant anthrax vaccines are based on the recombinant protein as the sole or major B. anthracis antigenic component (28,29) or on a live attenuated bacterial strain expressing recombinant PA (30,31).
PA has been reported previously to be a thermally unstable protein, losing its in vitro cytotoxic activity upon storage at 37°C within 48 h (32) or within few minutes above 40°C as a result of aggregation (33). Protein loss of function under mild storage conditions is generally attributed to a variety of nonenzymatic modifications such as deamidation, isomerization, oxidation, and alternative disulfide pairings (34,35). Spontaneous deamidation, which occurs mainly at Asn side chains and at a much slower rate at Gln residues (36), is a major and well documented degradation pathway in proteins. Deamidation rates depend on pH, temperature, primary sequence ("nearest neighbor" effect), and protein conformation (37)(38)(39)(40). Deamidation proceeds by nucleophilic attack on the side chain carbonyl carbon of Asn by the nitrogen of the adjacent peptide bond, resulting in the formation of an unstable five-member succinimide ring, which is hydrolyzed to produce mainly L-Asp and L-isoaspartic acid at a ratio of ϳ1:2 (41). Thus, deamidation can induce structural and functional perturbations in proteins through the introduction of a new negatively charged amino acid (Asp) and the insertion of an additional methyl residue in the polypeptide backbone (isoaspartic acid). As a result, generation of new acidic isoforms is considered a hallmark of deamidation (42,43). It has been hypothesized that in vivo deamidation functions as an internal clock, regulating the half-life of proteins (39,44), and may have significant physiological consequences, resulting in autoimmune diseases (45), Alzheimer disease (46), lens cataracts (47)(48)(49), and in other physiological systems (43,44,50). Furthermore, the deamidation process is of major concern in the production of pharmaceutical proteins mainly during their purification and under long-term storage, as reported for recombinant human growth hormone (51), growth hormone-releasing factor (52), recombinant plasminogen activator (53), and recombinant human interleukin-11 (54), and even in monoclonal antibody preparations following in vivo administration (55). In addition, it was demonstrated recently that deamidation of tetanus vaccine under long-term storage results in the impairment of antigen processing and presentation (56).
In this study, we aimed to clarify the molecular mechanism(s) by which PA undergoes loss of function. We present evidence supporting the notion that inactivation of PA upon storage is caused by spontaneous deamidation of at least three vulnerable Asn sites, perturbing crucial steps in its mechanism of action. Production and Purification of PA and LF-PA and LF were purified from B. anthracis strain V770-NP1-R (ATCC 14185) cultivated as described previously (31,55,57) and stored at Ϫ20°C until used.

EXPERIMENTAL PROCEDURES
PA Cytotoxicity Assay-Purified stocks of PA and LF were kept at Ϫ20°C until used. Cytotoxicity was determined in J774A.1 cells (American Type Culture Collection, Manassas, VA) as described previously (19,27). In brief, cells (6 ϫ 10 5 /ml) were plated in 96-well cell culture plates. The cytotoxicity assay was initiated by the addition of PA (in serial dilutions) and LF, followed by incubation for 5 h at 37°C under 5% CO 2 . Cell viability was monitored by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay (58). Cytotoxicity was determined by linear regression of each tested sample from a PA standard curve and is expressed as percent activity at time 0.
PA Storage Conditions and Peptide Deamidation-PA was stored at 25°C in 50 mM NH 4 HCO 3 buffer (pH 8.0) for up to 30 days in sealed tubes in a humid and sterilized (0.2-m filtration) environment. Storage conditions were selected on the basis of the PA purification methodology (31,57) and heat stability studies (33). Peptides were incubated for a few days in sealed tubes at 37°C essentially as described above in 50 mM NH 4 HCO 3 (pH 8.0). pH and sample volumes were stable throughout the incubation period.
Isoelectric Focusing and SDS-Polyacrylamide Gels-Isoelectric focusing was carried out at a pH range of 3-9, and SDS-pretreated samples were analyzed using precast gels (PhastSystem, Amersham Biosciences) and Bio-Rad mini-gels.
Gel Scanning-Gel scanning and densitometry were carried out using a GS-800 calibrated imaging densitometer (Bio-Rad) with Quantity One software.
Peptide Synthesis and Purification-Simultaneous Fmoc solid-phase peptide synthesis was performed manually in "T-bags" (59). Peptides were purified on a semipreparative reversed-phase HPLC column (Supelcosil LC-18-DB 300A, 5-m pore size, 250 ϫ 10 mm) using a Waters system equipped with 600E delivery pumps, a 996 photodiode array detector, and a 717 autosampler and controlled by Millenium software. About 2 mg of each peptide were purified (Ͼ95%), dried (Savant SpeedVac), and stored at Ϫ20°C until used.
MALDI-TOF/MS and Amino Acid Analysis-Reversed-phase HPLC-purified peptide samples were applied to a MALDI-TOF/MS target in 0.1% trifluoroacetic acid as described previously (60). Peptide quantitation was by amino acid analysis using vapor acid hydrolysis (110°C, 16 h) and phenylthiocarbamyl pre-column derivatization (phenylthiocarbamyl workstation, Waters Corp.).
In-gel Asp-N Digestion-Coomassie Blue-stained protein bands were excised from the gel and in gel-digested with 2 g/ml endoproteinase Asp-N in 50 mM NH 4 HCO 3 and 30% CH 3 CN. Following 2 h of incubation at 37°C, peptides were extracted in 20 l of 1% trifluoroacetic acid for 20 min at room temperature and analyzed by MALDI-TOF/MS.
Enzyme Assays and Michaelis-Menten Constant Determination-K m(app) and V max(app) were determined using GraphPad Prism software. The data obtained were fitted to the hyperbolic Michaelis-Menten rate equation. Peptide cleavage was initiated by the addition of 1 l of furin (ϳ4 ng) to 100 l of peptide solution containing 150 mM Tris-HCl (pH 7.4) and 1 mM CaCl 2 over a peptide concentration range of 0.05-20 M at 37°C, well within the affinity range of furin for the peptide substrates used in this work. At 1-2-min intervals (for up to 10 min), reactions were stopped by the addition of 10 l of 10% trifluoroacetic acid, followed by injection into a Supelcosil LC-318 analytical column (250 ϫ 4.6 mm). The intact peptide and its N-terminal cleaved product were identified by MALDI-TOF/MS. Cleavage rates were calculated as the ratios of product peak area versus the sum of the substrate and product peaks at a given reaction time. The theoretical isotopic distribution of identical amino acid sequences for each peptide was determined as 0% deamidation (using Micromass MassLinx software).
PA Heptamerization Assay-The PA heptamerization process was followed essentially as described previously (17,61) with the following modifications. PA samples (2 mg/ml) were cleaved by trypsin (0.25 g/ml) for 30 min at 37°C in 50 mM NH 4 HCO 3 (pH 8.0). The reaction was stopped by 1 mM phenylmethylsulfonyl fluoride (final concentration). LF was added to the PA samples at various molar ratios (25°C), and the mixtures were immediately separated by native gel (4 -15%) electrophoresis (PhastSystem) or by gel filtration on a Superdex 200 10/30 fast protein liquid chromatography column (Amersham Biosciences).
PA Biotinylation and Binding Assay-PA was labeled with biotin (Pierce kit 21430), at a 1:50 biotin/PA molar ratio according to the manufacturer's instructions. Biotin substitution was evaluated following PA dialysis at 4°C in 50 mM NH 4 HCO 3 for 16 h by the 2-hydroxyazobenzene-4Ј-carboxylic acid assay (Pierce kit 21430). PA binding studies were performed at 4°C using biotinylated PA (1.3 mol of biotin/ mol of PA). Chinese hamster ovary cells were grown in 24-well plates to 1.5 ϫ 10 5 cells/well in ␣-minimal essential medium and 10% fetal calf serum in an CO 2 atmosphere (62). The PA binding assay was performed at 4°C. Biotinylated PA was added at various concentrations (0.25-10 g/ml) in the absence or presence of a 100-fold molar excess of unlabeled PA (for nonspecific binding determination). Two h later, cells were washed with cold phosphate-buffered saline and lysed in 250 l of cold buffer solution containing 50 mM Tris-HCl (pH 7.4), 1% Nonidet P-40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM EDTA, 2 mM MgCl 2 , 10 units/ml DNase, and protease inhibitor mixture (Sigma). Extracted PA was applied to streptavidin-precoated enzyme-linked immunosorbent assay plates (Pierce). Biotinylated bound PA was quantitated using guinea pig anti-PA polyclonal antibodies.

Loss of PA Cytotoxic Activity Correlates with Generation of Acidic Isoforms
PA cytotoxicity in the presence of LF decreased gradually upon storage at 25°C and pH 8.0 with an apparent half-life of ϳ7 days (Fig. 1A). Loss of cytotoxic activity was accompanied by gradual appearance of acidic isoforms exhibiting pI values lower than the calculated value of 5.8 (Fig. 1B, right panel) with no apparent molecular breakdown (left panel), suggesting the involvement of a deamidation process (42,63). Note the presence of some acidic isoforms at time 0 ( Fig. 1B, right panel), suggesting that a certain level of deamidation had occurred already during PA isolation and purification (see Fig. 5). To determine putative deamidated residues, we first ranked all of the 68 Asn residues of PA (Protein Data Bank code 1ACC) according to their theoretical deamidation half-lives calculated on the basis of the algorithm of Robinson and Robinson (available at www.deamidation. org) (39), which takes into account empirical studies as well as primary, secondary, and three-dimensional structures of the protein, under standard experimental conditions. This ranking procedure resulted in predicted Asn deamidation half-lives ranging from 24 days (Asn 713 ) to Ͼ2.5 ϫ 10 6 days (Asn 602 ). Based on the PA storage time frame used in this study, a cutoff half-life of 60 days was selected. Only five Asn residues with calculated deamidation half-lives of Ͻ60 days fall in this category (Fig. 2B). These include Asn 10 , with no known role in the mechanism of PA cytotoxicity (1); Asn 162 at P 6 in the cleavage recognition sequence of furin (9,64); Asn 306 in the heptameric pre-pore loop (19); and finally, Asn 713 and Asn 719 in the receptor-binding large loop domain (20), in close proximity to the PA-receptor direct contact interface region (Gly 714 -Thr 716 ) (8,22). Their actual deamidation upon PA storage and the relationships between deamidation processes and PA inactivation were further evaluated.

Analysis of the Five Asn Residues with Predicted Susceptibility to Deamidation
The five Asn residues in PA with the shortest predicted deamidation half-lives (Fig. 2B) were subjected to analysis for possible deamidation by mass spectrometry. During deamidation, one expects a 1-Da mass increase (NH 2 versus OH, ⌬ ϭ 1 Da) of Asn-containing peptide fragments. This mass shift is used to evaluate the deamidation process at the selected Asn residues (65) as detailed below.
Asn 10 and Asn 162 -In an attempt to mass analyze the peptide fragments encompassing Asn 10 and Asn 162 in domain I, PA was cleaved by trypsin, followed by SDS-PAGE separation of the resulting 20-kDa polypeptide (cf. Fig. 6B). (Note that prior to this treatment, all stored PA preparations contained essentially intact 83-kDa polypeptide.) The 20-kDa band was subsequently in gel-digested with Asp-N. MALDI-TOF/MS analysis resulted in three major peptides (suitable for isotopic mass resolution) corresponding to the predicted Asp-N cleavage products: Glu 1 -Ser 25 DECEMBER 2, 2005 • VOLUME 280 • NUMBER 48 peptide has in fact two Asn residues at positions 150 and 162, which may represent two potential deamidation sites. Based on the algorithm of Robinson and Robinson (39), however, the predicted halflives of Asn 150 and Asn 162 are 21 ϫ 10 3 and 38 days, respectively. It is therefore likely that Asn 162 is fully responsible for the observed deamidation. This assumption was further confirmed experimentally using a model peptide of identical sequence, which was deamidated, cleaved by trypsin, and analyzed by mass spectrometry (data not shown).

Deamidation of B. anthracis Protective Antigen
Asn 306 -To test the possible deamidation of Asn 306 (predicted half-life of 53 days) (Fig. 2), located in heptamerization domain II (residues 259 -487) (19), PA samples were digested with endoproteinase Asp-N. Fingerprint MALDI-TOF/MS analysis was used to identify the Asp-N peptide product Asp 283 -Phe 314 at time 0 and following 14 days of PA storage (Fig.  4). The similarity between the spectra obtained and the theoretical isotopic model distribution suggests that Asn 306 is not subjected to any detectable deamidation during the PA storage conditions studied.
Asn 713 and Asn 719 -Asn 713 and Asn 719 , with predicted short deamidation half-lives of 24 and 49 days (Fig. 2), respectively, are located within the exposed "large loop" of 20 amino acids (positions 703-722), which was suggested to be involved in PA-receptor interaction (8,20,22), and in close proximity to one of the PA-receptor contact interfaces (amino acids 714 -716) (22). Tryptic mass fingerprints of native PA at time 0 and following 7 and 14 days of storage were analyzed and compared with the model isotope distribution of an identical peptide sequence. As shown in Fig. 5A, the theoretical monoisotopic ratio of the expected tryptic amino acid peptide sequence, 0.32 ( Fig. 5A and legend to Fig. 3C), was used as a reference point for 0% deamidation. Unexpectedly, the monoisotopic ratio of the tryptic fragment from the PA sample at time 0 was 0.14, indicating Ͼ50% deamidation. This suggests that a deamidation process had already occurred in this region during PA isolation and purification (see acidic isoform distribution in Fig. 1B at  time 0). It should be noted that this analyzed tryptic peptide, E 704 NTIINPSENGDTSTNGIKK 723 , has in fact four potential deamidation sites at Asn 705 , Asn 709 , Asn 713 , and Asn 719 , with theoretical halflives of 3007, 1273, 24, and 49 days, respectively. To confirm the predicted relative deamidation tendency of these residues, a model peptide of identical sequence was examined. Deamidation of this peptide fol- lowing incubation for 10 days at 37°C resulted in a 2-Da gain in mass (data not shown), indicating complete deamidation of two Asn residues. This deamidated peptide was further analyzed by Edman amino acid sequence analysis. As shown in Fig. 5B, the phenylthiohydantoin (PTH)-Asn signals at P 2 and P 6 (Asn 705 and Asn 709 ) were not affected upon deamidation compared with the control untreated peptide, whereas the PTH-Asn residues at cycles 10 and 16 (corresponding to Asn 713 and Asn 719 ) were completely replaced with PTH-Asp residues, clearly indicating a deamidation process at these specific Asn sites (Fig.   5B). In addition, peptides with the sequences ENTIINPSE (positions 704 -712) and NGDTSTNGIKK (positions 713-723) were stored for various time periods and analyzed for deamidation. The NGDTST-NGIKK peptide underwent rapid deamidation (half-life of 12 h at pH 8.0 and 37°C), whereas the ENTIINPSE peptide remained stable (following 5 days of storage). Therefore, we conclude that the observed deamidation of the Glu 704 -Lys 723 peptide in PA (Fig. 5A) proceeds through deamidation of both Asn 713 and Asn 719 , with no significant contribution of Asn 705 and Asn 709 . We found that three of the five selected Asn residues undergo deamidation, and most notably, these are the residues with the shortest predicted half-lives (Fig. 2).

Possible Relationship between Deamidation and PA Functional Impairment
The Furin Cleavage Step-Cleavage of PA by furin at RKKR 167 2 is a crucial step in the PA mechanism of action, allowing its heptamerization and co-internalization with either LF or EF into target cells (9,64). To examine whether PA inactivation during storage correlates with deamidation of Asn 162 in the P 6 recognition sequence (Fig. 6A), we first examined the susceptibility of PA to furin cleavage at different storage times. As shown in Fig. 6 (B and C, left panels), the native PA fraction could be readily cleaved by furin (time 0). However, stored fractions became progressively insensitive to cleavage. To ensure that reduced PA sensitivity to furin upon storage is not caused by hindering its furin cleavage site due to nonspecific protein denaturation and aggregation (33), we subjected the same stored samples to tryptic digestion. Trypsin cleavage efficiency was not significantly affected by PA storage (Fig. 6, B and C, right panels). The limited cleavage of PA by trypsin mainly at Arg 167 is also consistent with the conclusion that the overall conformation of PA is preserved.   DECEMBER 2, 2005 • VOLUME 280 • NUMBER 48

JOURNAL OF BIOLOGICAL CHEMISTRY 39901
Asn 162 Deamidation Reduces Furin Cleavage Efficiency-To examine whether the reduced sensitivity of stored PA fractions to furin, from 95% to only 50% cleavage upon 30 days of PA storage (Fig. 6C), is indeed due to the observed deamidation at P 6 in the furin recognition sequence, a model peptide (L 156 KQKSSNSRKKR2STSAG 172 ) encompassing the furin cleavage site in PA was synthesized as a substrate for furin. This peptide was further subjected to deamidation in vitro (see "Experimental Procedures"). Deamidation of this peptide is clearly manifested by a typical decrease in the monoisotopic mass of 1904.1 Da concomitant with increased intensities of successive isotopes (1907.2 and 1908.2 Da) (Fig. 7A). Both the native and resulting deamidated peptides were used as potential substrates for furin. As shown in Fig. 7B, although the native peptide could be readily cleaved by furin (as indicated by the reduced peak intensity of the substrate peak at 1904.1 Da and the appearance of the expected product of 1501.9 Da), the deamidated peptide appeared to be cleaved by furin at a much slower rate (Fig. 7B, right panel). For quantitative comparison, proteolytic reaction rates were evaluated by analytical reversed-phase HPLC (see "Experimental Procedures"). As depicted in Fig. 7C, incubation of the native peptide (10 M) with furin resulted in 60% peptide cleavage within 10 min compared with Ͻ8% cleavage of the deamidated peptide. Reaction rates as a function of substrate concentration were plotted for the native and deamidated peptides (Fig. 7C). Both curves fit the normal Michaelis-Menten equation, which was used to calculate the kinetic parameters for these reactions. As shown in TABLE ONE, the K m of furin with the native peptide substrate was 1.1 M, in good agreement with the previously reported K m of 2 M when PA protein was used as a substrate (9). Following peptide deamidation, the K m increased to 8 M, with a 2-fold decrease in V max (Fig. 7C and TABLE ONE, Peptide 2). Deamidation of Asn side chains results in the formation of mainly L-Asp and L-isoaspartic acid at a ratio of ϳ1:2 (41,66). Although the modification of Asn to Asp inserts one negative charge for each Asn residue in the polypeptide chain, the racemization process results in the insertion of a -CH 2 -group, in addition to the net negative charge. Using appropriate peptide analogs, the effects of substituting Asp or isoaspartic acid for Asn on the furin cleavage efficiency were evaluated and compared. As shown in TABLE ONE, substitution of Asn with Asp (Peptide 3) resulted in a 10-fold decrease in cleavage efficiency (k cat /K m ), mainly due to an increase in the K m , whereas isoaspartic acid substitution (Peptide 4) caused further reduced efficiency, due to both increased K m and decreased V max values. Reduced substrate efficiency was also observed with the Asn-to-Glu analog (Peptide 5). Finally, substitution of Asn 162 with Ala or ␤Ala had no significant effect (Peptide 6 versus Peptide 7), further emphasizing the dominant role of the negative charge versus the addition of a methyl group in reducing the furin cleavage efficiency. Taken together, these results indicate that the insertion of a negative charge at P 6 (Asn to Asp) is the main cause for the reduced furin cleavage efficiency, whereas the effect of inserting an additional -CH 2 -group into the peptide backbone (Peptide 2 versus Peptide 3) or into the side chain (as in the case with the Asn-to-Glu analog, Peptide 5) is marginal.
Heptamerization of PA Is Not Affected by Prolonged Storage-Heptamerization, involving domains II and III of the PA molecule (16,18,19), has a central role in PA cytotoxicity. Although no evidence for Asn 306 deamidation could be detected (Fig. 4), it may still be possible that other deamidation processes could be indirectly involved in PA heptamerization impairment, which may in turn contribute to the  . A deamidation t1 ⁄2 of 2.6 days was calculated based on first-order kinetics. B, the native and deamidated peptides (10 M each) were incubated in the presence of furin (40 ng/ml) for 60 min, and the reaction mixtures were analyzed by MALDI-TOF/MS. The arrows indicate the peptide products. C, the furin cleavage rates of the native and deamidated peptides was monitored by reversed-phase HPLC. The native and deamidated peptides (10 M each) were incubated in the presence of 40 ng/ml furin for 10 min. The arrowheads indicate the N-terminal fragment product Ac-LKQKSSNSRKKR peak (identified following collection and analysis by MALDI-TOF/MS; not shown). The indicated percent cleavage was calculated as the substrate peak area versus the sum of peak areas of the product and the remaining substrate peaks. The lower curve represents the initial furin (40 ng/ml) cleavage rates of the native and deamidated peptides versus peptide substrate concentrations. DECEMBER 2, 2005 • VOLUME 280 • NUMBER 48 observed inactivation of PA. The effects of storage on the PA heptamerization process were thus examined by inducing polymerization of trypsin-cleaved PA samples upon the addition of LF toxin at increasing LF/PA molar ratios (0 -1.5 mol/mol) (see "Experimental Procedures"). No apparent differences could be observed between the oligomerization patterns of the native versus 85% inactivated (14 days) stored samples as monitored by native gel analysis (18) and gel filtration experiments (data not shown). We thus conclude that the heptamerization process does not contribute to PA inactivation upon storage.

Deamidation of B. anthracis Protective Antigen
PA Binding Is Impaired during Storage-As shown above, the two susceptible sites (Asn 713 and Asn 719 ) were already partially deamidated at time 0 and underwent further deamidation upon PA storage. These residues are located in close proximity to one of the receptor interface segments (Gly 714 -Thr 716 ) (22). This proximity might affect the efficiency of binding of stored PA to cell receptors, impairing its cytotoxic activity. For binding assays, PA was first labeled with biotin at a biotin/PA molar ratio of 2, resulting in the substitution of 1.3 mol of biotin/ mol of PA, which had no significant effect on its cytotoxic activity (data not shown). Binding experiments using biotinylated PA samples after 0, 7, and 14 days of storage were then performed. As shown in Fig. 8, the maximum binding site value for native PA (stored at Ϫ20°C) was found to be 13,000 PA molecules/Chinese hamster ovary cell, in agreement with a previous study (62). A sharp decrease in PA binding of 47 and 85% was observed following 7 and 14 days of storage, respectively, concomitant with a similar loss of cytotoxic activity (Fig. 1).

DISCUSSION
PA is of great pharmacological importance as a vaccine against anthrax. In addition, PA is also an attractive model system for studying the impact of deamidation on protein-protein interaction and the mechanism of action. PA has a relatively high abundance of Asn residues (Asn/total amino acid ratio of 0.09 versus a normal statistical distribution of 0.05) (21,67). Furthermore, PA has a complex mechanism of action, involving multiple forms of protein-protein interactions: as a substrate for furin, as a ligand to cell receptors, as a self-forming homooligomer (heptamer), and as a complex with LF and EF (6,16,68). We, as well as others (33,69), have observed that, under various storage conditions, PA loses its cytotoxic activity in macrophage cell line assays. We have demonstrated here that, under mild conditions (25°C, pH 8.0), loss of activity is not due to proteolytic degradation or denaturation because the inactivated PA preparations preserve (a) their molecular integrity (Fig. 1B, left panel) and (b) their global conformation, as indicated by their limited trypsin cleavage pattern (Fig. 6, B and C, right  panels), confined mainly to the Arg 167 -Ser 168 furin cleavage bond. On the other hand, loss of biological activity upon storage appears concomitant with the formation of new acidic isoforms (Fig. 1B). This phenomenon of multiple acidic isoform formation has been demonstrated previously to be the hallmark of deamidation (40,42,43,54,63). When all PA Asn residues were ranked on the basis of their predicted deamidation half-lives (available at www.deamidation.org), only five of the 68 residues were found to have theoretical half-lives of up to 60 days (a cutoff that was selected based on the observed kinetics of PA inactivation upon storage). The five Asn residues (Fig. 2) are scattered in protein domains, which may be critical for PA biological function: two residues are located in domain I (Asn 10 , in a region with no known function, and Asn 162 , at the proximity of the furin cleavage site); one residue (Asn 306 ) is located in domain II within the heptamerization insertion loop; and two residues (Asn 713 and Asn 719 ) are located in domain IV in the PA receptor-binding region. Heptamerization, cell receptor binding efficiency, and cleavage susceptibility to furin were therefore monitored in an attempt to correlate functional impairment with Asn deamidation.
Heptamerization Process-According to the current model (70), furin cleavage enables the polymerization of PA as heptamers on the cell surface, which then interact with either LF or EF. Because this heptamerization involves the interaction of domains II and III (16,18), it was of interest to evaluate whether this complex process is impaired upon 14 days of PA storage concomitant with the deamidation of Asn 306 and cytotoxic inactivation. As shown in Fig. 4, Asn 306 did not appear to undergo deamidation. In addition, heptamerization proceeded similarly in the fully activated PA as well as inactivated stored PA preparations as judged by native gel analysis and size exclusion chromatography experiments (data not shown). We conclude that heptamerization is probably not related to the observed PA inactivation upon storage.
PA Binding to Cell Receptors-PA binding to cell receptors involves its C-terminal domain IV (8,16,20,22). Studies of the PA-receptor complex implied direct interaction of PA with the receptor at the contact region between positions 714 and 716 (8,22). We observed that both Asn 713 and Asn 719 , with predicted short half-lives of 24 and 49 days, respectively, underwent deamidation ( Fig. 5) with concomitant impairment of PA binding (Fig. 8) and reduced cytotoxicity (Fig. 1). It is thus tempting to speculate that the close vicinity of the two susceptible Asn residues to a PA-receptor contact region may impair PA binding upon deamidation, resulting in a decrease in LF-mediated PA cytotoxicity. Yet, in view of site-directed mutagenesis studies (21) in which Asn 713 and Asn 719 replacement with Ala had no substantial effect on PA toxicity, it is difficult to determine the contribution of deamidation at these positions to the overall decrease in PA cytotoxicity. PA Proteolytic Cleavage by Furin-Asn 162 , a predicted candidate residue for deamidation (Fig. 2), is located six amino acids (P 6 ) upstream of the site of serine cleavage by furin. Furin typically cleaves protein substrates at the consensus sequence Arg-Xaa-(Lys/Arg)-Arg (9,71). Although the requirements for Arg at P 1 and P 4 have been documented extensively in the majority of furin-processed proproteins, those at P 6 are less characterized. Yet, in furin natural protein substrates, cleavage requirements are restricted merely to alkaline residues at either P 4 or P 6 (72)(73)(74). Thus, the importance of the amino acid at P 6 is expressed mainly when a basic residue at P 4 is missing. We have demonstrated here that, upon storage, PA sensitivity to furin decreases concomitant with substantial deamidation of Asn 162 . To test the hypothesis of the causal relationship between deamidation at P 6 and PA inactivation, we compared the efficiency of furin cleavage of a synthetic peptide substrate in its native state and following deamidation. In addition, we generated a series of peptide substrate analogs to directly investigate the major determinants affecting furin cleavage efficiency upon deamidation. In quantitative kinetic studies, we have demonstrated that deamidation impairs furin cleavage efficiency due to the insertion of a negatively charged residue at P 6 in the furin recognition sequence (TABLE  ONE). These results are consistent with the notion that PA inactivation upon storage is at least partially due to PA deamidation at Asn 162 , resulting in its reduced sensitivity to furin (Figs. 6 and 7). Indeed, the important role of P 6 in the RKKR recognition sequence (in which P 4 is positively charged) has been addressed previously (75). In addition, a recent three-dimensional x-ray model of furin demonstrates a canyon-like crevice of the active site with a negative surface potential (Glu 230 and Asp 233 ) adjacent to P 6 of the substrate (16). Although deamidation of Asn 162 may reduce the efficiency of PA cleavage by furin and contributes to the overall reduced PA cytotoxicity upon storage, it cannot explain the extensive reduced PA cytotoxicity (75%) (Fig. 1) following 14 days of storage because, at this time, Asn 162 underwent only 28% deamidation (Fig. 3B), with an ϳ30% reduced cleavage by furin (Fig. 6C). It is thus conceivable that additional factors, such as binding impairment, are involved in the PA inactivation mechanism (76).
Physical Factors Affecting PA Deamidation Rates-In good agreement with the predicted rules (39), we have found that the three Asn residues with the predicted shortest half-lives indeed undergo deamidation. It should be noted that the "standard storage conditions" applied for theoretical predictions were 37°C at pH 7.4 (39), whereas we applied 25°C at pH 8.0. In addition to the nearest neighbor effect of the intrinsic protein and its three-dimensional structure, deamidation susceptibility is highly dependent on various incubation conditions, such as temperature and pH, which were shown to have the most profound effect on deamidation. For example, it has been shown recently that deamidation of glutamate dehydrogenase produced in Escherichia coli can be substantially reduced by temperature decreases during growth and during protein isolation (63). Additional studies have indicated the pH effect on the deamidation rate (77). Indeed, preliminary studies in our laboratory clearly indicate a decrease in the deamidation rate upon PA storage (as deduced from the reduction in the appearance of acidic isoforms) at pH 7.4 and at lower temperature concomitant with an improved biological stability of the protein.
In summary, five of the 68 Asn residues in B. anthracis PA were selected and analyzed based on their predicted deamidation half-lives for their possible involvement in protein inactivation upon storage. Indeed, three of these residues have been demonstrated here to undergo spontaneous deamidation. These deamidation events could be correlated with a decrease in PA cytotoxicity, caused by impairment of two distinct steps in PA cytotoxicity: cell receptor binding and furin cleavage efficiency. The possible effect (if any) of PA deamidation on the induction of protective immunity by PA-based vaccines remains to be determined.