The carboxyl-terminal end of protective antigen is required for receptor binding and anthrax toxin activity.

Anthrax toxin consists of three separate proteins produced by Bacillus anthracis: protective antigen (PA), lethal factor (LF), and edema factor (EF). Previous work showed that the process by which these proteins damage eukaryotic cells begins with binding of PA (83 kDa) to cell surface receptors. PA is then cleaved by a cell surface protease so as to expose a high-affinity binding site for LF or EF on the COOH-terminal, receptor-bound, 63-kilodalton fragment. In this report we more closely define a region of PA involved in receptor binding. The gene encoding PA was mutagenized so as to delete 3, 5, 7, 12, or 14 amino acids from the carboxyl terminus of the protein, and the truncated PA variants were purified from Bacillus subtilis or Escherichia coli. Deletion of 3, 5, or 7 amino acids reduced the binding of PA to cells and the subsequent toxicity of the PA.LF complex to J774A.1 cells and also the ability to cause EF binding to cells. Deletion of 12 or 14 amino acids completely eliminated all these activities. These results show that the carboxy terminus comprises or is part of the receptor-binding domain of PA.

Bacillus anthracis secretes three proteins that are collectively designated anthrax toxin. These proteins are protective antigen (PA),' lethal factor (LF), and edema factor (EF) (1)(2)(3). Each protein lacks toxic activity when administered alone; instead, the proteins act in binary combinations to produce two toxic activities. Thus, PA with EF (designated "edema toxin") produces edema in skin and inhibits the function of phagocytic cells (4-6). These effects are undoubtedly the result of elevated CAMP concentrations, because EF is an adenylyl cyclase that enters cells and acts on cytosolic ATP (7). PA in combination with LF ("lethal toxin") causes death of experimental animals (8) and lysis of mouse and rat mac-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. rophages (9) and is considered to be responsible for the death of infected animals (2, 10). The intracellular mechanism for LF action is currently unknown.
The genes for each of the anthrax toxin components have been cloned and sequenced (11)(12)(13)(14)(15). The PA gene encodes a polypeptide of 735 amino acids. We have reported previously that PA binds to a specific cell surface receptor and is then cleaved by a cell-associated protease which releases an NH2terminal 19-kDa fragment (16)(17)(18). Removal of this fragment creates or exposes a high-affinity binding site for LF and EF. PA can also be specifically cleaved in vitro at or near this site with trypsin, producing a 63-kDa fragment that retains bioactivity (16) and can insert in artificial lipid membranes (19). Thus PA appears to have at least three different functions: (i) receptor binding, (ii) LF or EF binding, and (iii) translocation of the toxic complex to the cytosol. The region involved in receptor binding has been studied in this paper by construction, expression, and characterization of PA proteins lacking 3-14 amino acids at the carboxyl terminus.

EXPERIMENTAL PROCEDURES
Reagents and General Procedures-Restriction endonucleases and DNA modification enzymes were purchased from Betbesda Research Labs, New England Biolabs, or Pharmacia LKB Biotechnology Inc. Low melting point agarose (Sea Plaque) was from FMC Corp. Oligonucleotides were synthesized on a PCR Mate (Applied Biosystems) and purified on oligonucleotide purification cartridges (Applied Biosystems). The polymerase chain reaction (PCR) was done using a DNA amplification reagent kit (GeneAmp) from Perkin-Elmer Cetus Instruments and a Perkin-Elmer Cetus Thermal Cycler. The amplification involved denaturation at 95 "C for 1 min, annealing at 37 'C for 2 min, and extension at 72 "C for 2 min, for a total of 30 cycles. A final extension was run a t 72 "C for 7 min. DNA sequencing reactions were done using the T7 sequencing kit from Pharmacia, and DNA sequencing gels were made with Hydrolink gel solution from A T Biocbem (Malvern, PA). [''ClAdenine (45 Ci/mmol) and '2sII-labeled Bolton Hunter reagent (2000 Ci/mmol) were purchased from Amersham Corp. J774A.1 cells were obtained from the American Type Culture Collection. Chinese hamster ovary cells (CHO) were obtained from the laboratory of Dr. Michael Gottesman (National Cancer Institute, National Institutes of Health).
Plasmid Construction-The protective antigen gene (pag) was mutated by oligonucleotide-directed mutagenesis of two previously described plasmids, pYS2 and pYS5 (17). pYS2 contains a T7 promoter driving expression of pug without a signal peptide sequence. pYS5 is an Escherichia coli-Bacillus subtilis shuttle vector containing the PA gene and additional 5' B. anthracis DNA that includes the pag promoter and signal peptide sequence.
Oligonucleotide-directed Mutagenesis-PA genes coding for proteins truncated at the carboxy terminus were generated by PCR amplification of the native pug gene from a pYS5 template. A nonmutagenic oligonucleotide primer corresponding exactly to nucleotides 2842-2864 of pag (numbering system of Welkos et al. (20)), a region containing a PstI recognition site unique to pug and pYS5, was used for all amplification reactions. Mutagenic oligonucleotide primers were designed to be complementary to the intended new 3' 15493 end of the PA coding sequence (near nucleotides 4054-4086) and to introduce two in-frame stop codons and a unique BamHI site. For example, the primer for deleting nucleotides 4087-4095 was 5'-CAC

C T A G A A T T A C C T G G A T C C T A T T A A T A G C C T T T T T T A
GAA AAG-3', where the underlined nucleotides are the reverse complement of nucleotides 4067-4086. Amplified PCR product was digested with restriction endonucleases PstI and BamHI and gelpurified after electrophoresis in 1.2% low-melting point agarose gels. Plasmids pYS2 and pYS5 were digested with PstI and BamHI, purified by the same method, and ligated to the PCR generated mutant gene fragments. Mutant constructions were identified by restriction analysis and the mutations confirmed by dideoxy sequencing of at least 200 nucleotides spanning the mutated region. PA deletion proteins were designated PA-732, PA-730, PA-728, PA-723, and PA-721, according to the number of residues remaining of the original 735. Proteins PA-732, PA-730, and PA-728 were expressed from pYS5-based shuttle vectors pYD15, pYD16, and pYD17 in B. subtilis DB104 (21), which lacks two major extracellular proteases. The longer deletions, proteins PA-723 and PA-721, were expressed in E. coli BL-21(XDE3) (22) from the pYS2-derived plasmids pYS18 and pYSl0 (Table I).
Expression and Purification of PA-B. subtilis DB104 containing the pYD plasmids were grown at 37 "C in 500 ml of FA medium (17) supplemented with 5% heat-inactivated fetal bovine serum, 2% glycerol, and 50 pg/ml kanamycin sulfate. When Asw reached 2.0, cultures were centrifuged at 10,000 X g to remove the cells. Supernatants were supplemented with NaCl to 0.5 M, EDTA to 10 mM, and phenylmethylsulfonyl fluoride to 1 mM and then sterilized by filtration through a 0.45-pm membrane. The culture filtrate was then adsorbed on a 1.5-ml column of anti-PA monoclonal antibody 10G4 (23), and the column was washed with 50 ml of phosphate-buffered saline, and eluted with 2 M sodium thiocyanate containing 20 mM HEPES, pH 7.4. Sodium thiocyanate was removed by passing the sample through a 10-ml column of Sephadex G-25 (Pharmacia PD-10) equilibrated with 10 mM HEPES, pH 7.4. Approximately 1 mg of purified protein from the PA mutants lacking 3, 5, or 7 amino acids were obtained from 500-ml cultures.
E. coli strains carrying the pYSl0 or pYS18 plasmids were grown in LB broth with 100 pg/ml ampicillin shaking at 250 rpm at 37 "C. When A, , reached 1.0, isopropyl-1-thio-P-D-galactopyranoside was added to a final concentration of 1 mM. Cells were induced for 120 min and then harvested by centrifugation at 4000 X g for 15 min at 4 "C. The cells from 1 liter were suspended in 50 ml of buffer (0.1 M HEPES, pH 7.4, 10 mM EDTA, 5% serum, 2% glycerol, and 1 mM 1,lO-phenanthroline) and sonicated in 5-ml batches for 2 min (30 s a t a time). After centrifugation at 10,000 X g the supernatants were diluted to 400 ml in 50 mM HEPES, pH 7.4,lO mM EDTA, 5% serum, 1 mM phenylmethylsulfonyl fluoride, and 200 mM NaC1, filtered through a 0.45-pm membrane, and purified on anti-PA monoclonal antibody columns as described above. The proteins were analyzed by polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulfate (SDS-PAGE). Gels were either stained with Coomassie Brilliant Blue, or the proteins were electrotransferred to nitrocellulose paper and probed with monoclonal antibody 3Dll(23) or a polyclonal rabbit anti-PA serum. The fraction of total protein that was full size PA was determined by scanning the Coomassie-stained gel with a laser densitometer (Pharmacia LKB Ultrascan XL).
Cell Culture Techniques and Cytotoxicity Measurements-The medium and growth conditions for J774A.1 cells have been described previously (24). The potencies of PA preparations were measured in vitro by the release of [I4C]adenine from J774A.1 cells (25). Nucleotide pools in cells were labeled by incubation with 0.1 pCi/ml [I4C]adenine in Dulbecco's modified Eagle's medium (DMEM) for 1 h at 37 "C. The cells were washed three times in situ to remove unincorporated adenine and then incubated at 37 "C for 3 h with various concentrations of truncated P A (0.01-10 pg/ml) in combination with 0.5 pg/ml LF. The radioactivity released into the medium was taken as a measure of cytotoxicity; release in the absence of toxin was typically 5% of the total incorporated.
Receptor-binding Assays-J774A.1 cells cultured in 12-well tissue culture plates were cooled by incubating in cold EMEM for 10 min and then placing the plates on ice. The medium was replaced with cold Williams E tissue culture medium supplemented with 1% bovine serum albumin and 25 mM HEPES (binding medium). Radiolabeled P A (0.1 pg/ml lZ5I-PA, 2.2 X lo6 cpm/pg) was added for 12 h at 4 "C.
Cells were washed four times with cold Hanks' balanced salt solution, solubilized in 0.5 ml of 0.1 M NaOH, and counted in a y counter (Beckman Gamma 9000). For measuring the binding activities of truncated PA proteins, the same assay was done except that varying concentrations of the nonradioactive proteins were included.
Assay of EF Binding-CHO cells were maintained as monolayer cultures in EMEM supplemented with nonessential amino acids and 10% heat-inactivated fetal bovine serum (binding medium). PA and truncated PA proteins at 1.0 pg/ml were added with 0.5 pg/ml EF to CHO cells which had been slowly cooled to 0 "C in binding medium. After 12-h incubation on wet ice, cell monolayers were washed three times with ice cold binding medium, and cell-bound EF was determined after lysing the cells in 0.2 ml of 0.1% Triton X-100 and assaying adenylyl cyclase activity as described previously (3).

RESULTS AND DISCUSSION
Structure-function studies of fragments produced by trypsin and chymotrypsin (3,17,26) have established that the COOH-terminal 63-kDa fragment of PA contains the receptor-binding domain, as well as the structural requirements for binding and internalization of LF and EF. T o locate more precisely the receptor-binding regions, we prepared and analyzed PA proteins truncated to varying extents at the carboxy terminus. These were generated by mutagenesis of two recombinant plasmids previously developed for expression of PA in E. coli and B. subtilis (17). Initial experiments showed that deletion of 65 amino acids at the carboxy terminus inactivated PA (data not shown). Therefore shorter deletions were made and the proteins purified. Proteins truncated by 3, 5, or 7 amino acids (PA-732, PA-730, and PA-728, respectively) were purified from supernatants of B. subtilis DB104 strains by immunoadsorbant chromatography with monoclonal antibody 10G4, as described previously (17). However, we obtained only about 2 pg of the purified proteins/ml of culture, compared with 20 pg/ml obtained from pYS5 making native PA. The low yield of these proteins may be due to their increased sensitivity to B. subtilis proteases. Electrophoresis (SDS-PAGE) of the purified proteins showed these proteins were at least 90% pure (Fig. 1).
Attempts to purify the proteins lacking 12 or 14 amino acids from B. subtilis supernatants were unsuccessful because of extensive degradation. Therefore these two proteins were purified from E. coli cell lysates, again using immunoadsorbant chromatography with monoclonal antibody 10G4. SDS-PAGE and Western blot analysis indicated that these products were at least 90% pure, although a large fraction of the material was proteolytically nicked or cleaved, so that the material migrating as intact 83-kDa protein constituted only 30-35% of the preparation (Fig. 1). In all subsequent assays, the concentrations of these deleted proteins were corrected to reflect the content of intact protein.
T o analyze the relative potencies of the purified proteins, dose-response analyses were performed with J774A.1 cells, using release of preloaded ['*C]adenine as a measure of toxicity (Table I). In the presence of a fixed concentration of LF (0.5 pg/ml), native PA at 0.1 pg/ml caused release of 80% of the adenine after 3 h (Fig. 2). The purified proteins lacking 3, 5 , or 7 amino acids were also toxic to J774A.1 cells when added with LF, but the cytotoxic activity was decreased 10to 20-fold compared with full-length PA. PA proteins with 12 or 14 amino acids deleted at the carboxyl terminus were not toxic to J774A.1 cells even at a 100-fold higher concentration (Fig. 2).
Because other data had suggested that the COOH-terminal region is required for receptor recognition, we considered it probable that the lower activity of the truncated PA proteins resulted from decreased receptor binding. To measure differences in the binding of truncated PA to cell surface receptors an assay was developed to measure competition between intact lZ5I-PA and the truncated PA proteins. Mutant PA proteins truncated by 3, 5 , or 7 amino acids had a 2-to 10-fold reduction in cell binding activity (Fig. 3), whereas binding and edema toxin, it was expected that the truncated PA activity was completely lost upon deletion of 12 or 14 amino proteins would also fail to promote EF binding. To test this acids. These results support the interpretation that the de-hypothesis, CHO cells were exposed for 12 h a t 0 "C to EF in creased toxicity of these truncated proteins is due to their the presence of native or truncated PA proteins. The CHO failure to bind to receptors. cells were washed extensively with binding medium, lysed, Because PA is a required component of both lethal toxin and the adenylyl cyclase activity ofthe cell lysate was assayed (3 surface receptor (as confirmed in the controls shown), the amount of adenylyl cyclase activity is a direct measure of the ability of PA to promote binding of EF. The data show that PA proteins deleted by 3,5, or 7 amino acids promoted binding (Table 11).
The data presented above show that the integrity of the carboxyl terminus is required for binding of PA to its cell receptor. However, the data do not allow us to unequivocally state that the amino acids near the *?minus-interact directly with the receptor, because the effects observed would also be expected if the terminal residues were instead (or in addition) needed for maintenance of the conformation of a spatially Although the evidence available cannot rule out either of these interpretations, an analysis of the properties of the smaller deletions (3, 5, and 7 amino acids) does lead us to favor the view that the terminal 7 residues are directly involved in receptor interactions. The two larger deletions (12 and 14 amino acids) differ qualitatively, being totally inactive (Fig. 2), and these will be discussed separately.
All of the decrease in activity of the smaller deletions can be attributed to their lowered affinity for receptor. Thus, the 2-to 10-fold decrease in receptor-binding affinity implied by the competitive binding data (Fig. 3) is sufficient to explain the similar losses in toxicity (Fig. 2). The smaller decreases (1-to 2-fold) observed in the assay for EF binding (Table 11) can be attributed to the use of the mutated PA proteins a t a single high concentration, 1.0 pg/ml. Thus, although the 3-, 5-, and 7-amino acid deletions have 10-fold lower affinity for receptor, their use a t 1.0 pg/ml will lead to saturation of the receptor. The fact that EF binding under these conditions nearly equals that with native PA shows that the mutant PA proteins are still capable of being nicked and of then binding EF with nearlv normal efficiencv. Further evidence that these J774A.1 cells were cultured in 24-well tissue culture plates. lZ5I-PA added in the wells and incubated for 12 h at 4 "C. At the end of (0.1 gg/ml) mixed with several concentrations of truncated PA was experiment cells were washed four times with cold medium containing 1% bovine serum albumin to remove nonspecifically bound PA. The cells were solubilized in 0.1 M NaOH and counted in a y counter. Controls showed that more than 85% of binding was specific (i.e. was competed by a 100-fold excess of nonradioactive PA).
The properties of the larger deletions (12 and 14 amino acids) cannot be so clearly explained. These proteins would obviously share with the shorter deletions any decreases in receptor binding due to the loss of the terminal seven amino acids. However it appears that these longer deletions also have additional defects in function. These proteins are much more susceptible to proteolysis in bacterial extracts and supernatants, implying that they have a less compact structure. The total inability of these proteins to bind to receptor or to promote toxicity could then be due either to the loss of an extended conformational structure involved in receptor binding or to the absence of the particular amino acids that constituted, with the terminal 7 residues, a sequence-determined receptor ligand.
Some additional insights into the structure of the receptorbinding structures in PA may be inferred from the behavior of certain anti-PA monoclonal antibodies that neutralize anthrax toxin. Antibodies 3B6 and 14B7 (23) cannot bind to PA that is already bound to its cell receptor and neutralize PA PA-723 PA-721 CHO cells were cultured in 12-well plates in EMEM containing nonessential amino acids and 10% fetal bovine serum. EF and the truncated PA proteins were added at the indicated final concentrations, and the cells were incubated for 12 h at 0 "C. Cells were washed to remove unbound proteins, solubilized in 0.1% Triton X-100, and the lysates assayed for adenylyl cyclase activity. Each value is the average & S.E. of three wells.
only if incubated with it prior to addition to cells. Therefore the antibodies are considered to bind at or very close to the site on PA which binds to the cell receptor. Because these antibodies fail to react with PA on immunoblots from SDS gels,' it appears that this site on PA consists of a conformationally determined structure rather than a sequence of contiguous amino acids.
Several precedents exist for the participation of the COOHterminal regions of protein toxins in receptor binding. For example, early evidence that the receptor-binding domain of diphtheria toxin was near the carboxyl terminus (27) was recently refined by showing that the last 54 amino acids of the toxin, released as a peptide by cleavage with hydroxylamine, can block binding of the toxin to receptor (28). Similarly, the receptor-binding domain of the Clostridium perfringens enterotoxin was localized to the COOH-terminal half of the protein by analysis of a recombinant fragment (29) and subsequently was shown to reside in the last 31 amino acids by expression and testing of this small region (32).
The carboxyl termini of some protein toxins may have roles other than or in addition to receptor binding. For instance, it was noted that Pseudomonas exotoxin A and cholera toxin have COOH-terminal sequences that resemble the sequence Lys-Asp-Glu-Leu (22), shown to be required for retention of newly synthesized eukaryotic proteins in the endoplasmic reticulum (30,31). In the case of exotoxin A, the Arg-Glu-Asp-Leu-Lys sequence is not involved in binding to the cell surface receptor, but is essential for correct intracellular trafficking. Although it is also possible that the COOH terminus of PA may play a role in determining the intracellular trafficking of the LF and EF toxin components, the data presented here show that one important function of the carboxyl terminus is to form part of the region required for recognition of the target cell surface receptor.
A complete understanding of the role of the COOH-terminal amino acids in PA will come only after an extensive mutational analysis of this region is completed. This should include use of substitutions as well as deletions. Also, if the interpretation presented above is correct, then it would be expected that synthetic peptides corresponding to the COOH terminus of PA might compete for receptor. However we have been discouraged from beginning such work by preliminary evidence that a purified fragment corresponding to residues Y. Singh and S. H. Leppla, unpublished data. 624-735 does not compete with "'1-PA for binding to receptor."