Cell Surface Tumor Endothelium Marker 8 Cytoplasmic Tail-independent Anthrax Toxin Binding, Proteolytic Processing, Oligomer Formation, and Internalization*

The interaction of anthrax toxin protective antigen (PA) and target cells was assessed, and the importance of the cytosolic domain of tumor endothelium marker 8 (TEM8) in its function as a cellular receptor for PA was evaluated. PA binding and proteolytic processing on the Chinese hamster ovary cell surface occurred rapidly, with both processes nearly reaching steady state in 5 min. Remarkably, the resulting PA63 fragment was present on the cell surface only as an oligomer, and furthermore, the oligomer was the only PA species internalized, suggesting that oligomerization of PA63 triggers receptor-mediated endocytosis. Following internalization, the PA63 oligomer was rapidly and irreversibly transformed to an SDS/heat-resistant form, in a process requiring an acidic compartment. This conformational change was functionally correlated with membrane insertion, channel formation, and translocation of lethal factor into the cytosol. To explore the role of the TEM8 cytosolic tail, a series of truncated TEM8 mutants was transfected into a PA receptor-deficient Chinese hamster ovary cell line. Interestingly, all of the cytosolic tail truncated TEM8 mutants functioned as PA receptors, as determined by PA binding, processing, oligomer formation, and translocation of an lethal factor fusion toxin into the cytosol. Moreover, cells transfected with a TEM8 construct truncated before the predicted transmembrane domain failed to bind PA, demonstrating that residues 321–343 are needed for cell surface anchoring. Further evidence that the cytosolic domain plays no essential role in anthrax toxin action was obtained by showing that TEM8 anchored by a glycosylphosphatidylinositol tail also functioned as a PA receptor.


Shihui Liu and Stephen H. Leppla ‡
From the Microbial Pathogenesis Section, NIAID, National Institutes of Health, Bethesda, Maryland 20892 The interaction of anthrax toxin protective antigen (PA) and target cells was assessed, and the importance of the cytosolic domain of tumor endothelium marker 8 (TEM8) in its function as a cellular receptor for PA was evaluated. PA binding and proteolytic processing on the Chinese hamster ovary cell surface occurred rapidly, with both processes nearly reaching steady state in 5 min. Remarkably, the resulting PA63 fragment was present on the cell surface only as an oligomer, and furthermore, the oligomer was the only PA species internalized, suggesting that oligomerization of PA63 triggers receptor-mediated endocytosis. Following internalization, the PA63 oligomer was rapidly and irreversibly transformed to an SDS/heat-resistant form, in a process requiring an acidic compartment. This conformational change was functionally correlated with membrane insertion, channel formation, and translocation of lethal factor into the cytosol. To explore the role of the TEM8 cytosolic tail, a series of truncated TEM8 mutants was transfected into a PA receptor-deficient Chinese hamster ovary cell line. Interestingly, all of the cytosolic tail truncated TEM8 mutants functioned as PA receptors, as determined by PA binding, processing, oligomer formation, and translocation of an lethal factor fusion toxin into the cytosol. Moreover, cells transfected with a TEM8 construct truncated before the predicted transmembrane domain failed to bind PA, demonstrating that residues 321-343 are needed for cell surface anchoring. Further evidence that the cytosolic domain plays no essential role in anthrax toxin action was obtained by showing that TEM8 anchored by a glycosylphosphatidylinositol tail also functioned as a PA receptor.
Anthrax toxin, the major virulence factor of Bacillus anthracis, consists of three polypeptides: protective antigen (PA, 1 83 kDa), lethal factor (LF, 90 kDa), and edema factor (EF, 89 kDa) (1,2). These three proteins are individually non-toxic. To intoxicate mammalian cells, PA binds to a ubiquitously ex-pressed, recently identified cellular receptor, tumor endothelium marker 8 (TEM8) variant 2 (3), and is cleaved at the sequence RKKR 167 on the cell surface by furin or furin-like proteases (4,5). Proteolysis yields the amino-terminal 20-kDa fragment (PA20), which is released into the medium, and the carboxyl-terminal 63-kDa fragment (PA63), which remains bound to the receptor and self-associates to form a ring-shaped heptamer (6,7). The heptamer binds up to 3 molecules of LF or EF (8,9). The resulting oligomeric complex is then internalized into endosomes, where the decreased pH causes the PA63 heptamer to insert into the endosomal membrane and produce a channel through which LF and EF translocate to the cytosol (10). Therefore, PA is the central part of anthrax toxin, serving as the delivery vehicle for binding and translocation of LF and EF into the cytosol of the cells. The combination of PA plus LF kills animals (11,12) and certain cells, including mouse macrophages (13,14). LF is a zinc-dependent metalloprotease that cleaves several mitogen-activated protein kinase kinases (MAPKK) in their amino-terminal regions (15,16). How this cleavage triggers the lethal effects of the toxin and whether there are additional cellular substrates remains unclear. EF is a calmodulin-dependent adenylate cyclase that elevates intracellular cAMP concentrations (17), thereby causing diverse effects in cells including the impairment of phagocytosis (18).
Previous studies on the interaction of PA with host cells have often used cytotoxicity assays to infer internalization mechanisms, or have used radiolabeled or chemically labeled PA that may behave differently due to modification. In the present studies, we directly assessed PA binding, proteolytic processing, and internalization by target cells using a highly sensitive and specific rabbit antiserum to PA. We found that following binding and processing by cell surface furin, the cleaved PA immediately forms the PA63 oligomer and that this oligomer is the only species of PA that is internalized. In addition, following internalization, the oligomer is quickly transformed into an SDS/heat-resistant form, a process coincident with insertion and channel formation in endosomal membranes.
In related studies we extended the understanding of toxin internalization obtained in the recent breakthrough that identified TEM8 variant 2 as a PA receptor (3). Currently, there are three reported cDNAs that result from splicing variations in TEM8 (GenBank TM accession number NM_032208, NM_053034, and NM_18153). The physiological functions of these have not been studied. Beyond the fact that TEM8 variant 2 functions as a PA receptor, the only information available is that implicit in the initial identification that TEM8 expression is up-regulated in tumor endothelium (19,20). Thus, it remains unknown whether other TEM8 variants can also function as PA receptors, and whether TEM8 has functions beyond binding PA in anthrax toxin action. To answer these questions, in the present work, we constructed a series of TEM8 truncated mutants, transfected them into a PA receptor-deficient Chinese hamster ovary (CHO) cell mutant, and found that all constructs having a membrane anchor functioned as PA receptors.
Cell Lines and Culture Media-CHO cell clone 6 (CHO CL6) is a line recloned in this laboratory from CHO 10001, a subclone of CHO-S (26), which was obtained from Dr. Michael Gottesman (National Institutes of Health, Bethesda). CHO FD11, a furin-deficient derivative of CHO CL6, was developed in our laboratory by chemical mutagenesis (27). CHO PR230 is a spontaneous PA receptor-deficient mutant derived from CHO WTP4, which is derived from the thioguanine-and ouabainresistant cell WTB111 (28), which was derived from CHO-K1. All CHO cells were grown in ␣-minimal essential medium supplemented with 5% fetal calf serum, 2 mM glutamine, 50 g/ml gentamycin, and 25 mM HEPES.
PA Binding and Internalization by CHO Cells-PA binding was assessed at both 37 and 4°C. Cells were grown in 24-well plates to confluence. Cells were incubated with 1 g/ml PA for different lengths of time and then washed five times with Hanks' balanced salt solution (HBSS) (Biofluids, Rockville, MD). The cells were lysed in 100 l of modified RIPA lysis buffer (50 mM Tris-HCl, pH 7.4, 1% Nonidet P-40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml each of aprotinin, leupeptin, and pepstatin). In measurements of PA internalization, the cells were first treated with 0.5 ml of 0.5 mg/ml trypsin in HBSS per well at 37°C for 5 min to remove proteolytically the cell surface-bound PA, then washed, and lysed. The cell lysates were subjected to SDS-PAGE or native-PAGE using 4 -20% Tris-glycine gradient gels (NOVEX, San Diego). Prior to loading, the cell lysates were boiled for 10 min in 1ϫ SDS sample buffer (50 mM Tris-HCl, pH 6.8, 2% SDS, 100 mM dithiothreitol, 0.01% bromphenol blue, 6% glycerol) for SDS-PAGE or were incubated for 10 min at room temperature in 1ϫ native buffer (NOVEX) for native-PAGE. The proteins were then transferred to nitrocellulose membranes, followed by Western blotting as described (29). PA was visualized by chemiluminescence using the West Pico Kit (Pierce). For the two-dimensional analysis, cell lysate was first separated on native-PAGE, and the gel strip was then sequentially equilibrated for 15 min each in Buffer I (125 mM Tris-HCl, pH 6.8, 1% SDS, 8.7% glycerol, 5 mM dithiothreitol) and Buffer II (125 mM Tris-HCl, pH 6.8, 1% SDS, 8.7% glycerol, 2% iodoacetamide) and subjected to SDS-PAGE, followed by Western blotting as described above.
LF Translocation and MAPKK1 Cleavage Assay-In measurements of LF translocation, CHO CL6 cells grown in 24-well plates were incubated with 1 g/ml LF along with 1 g/ml PA or PA-⌬ FF for 1 h at 37°C, washed once with HBSS, and treated with 0.5 ml 0.5 mg/ml trypsin in HBSS per well at 37°C for 5 min to remove proteolytically the cell surface-bound toxin. The cells were then washed and permeabilized by saponin to allow efflux of cytosol as described (30). Briefly, the cells were resuspended and incubated in 100 l of phosphatebuffered saline containing 50 g/ml saponin, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml each of aprotinin, leupeptin, and pepstatin for 30 min at 4°C. The soluble fraction was separated from the particulate fraction by centrifugation at 15,000 ϫ g for 5 min at 4°C. The pellet was washed in HBSS and solubilized in RIPA lysis buffer. The samples from the soluble and pellet fractions were analyzed by native-PAGE followed by Western blotting using LF antiserum (number 5309). In measurements of MAPKK1 cleavage by LF, CHO CL6 cells grown in 24-well plate were incubated with 1 g/ml LF along with 1 g/ml PA or PA-⌬ FF for 1 h at 37°C, washed, lysed, and analyzed by SDS-PAGE followed by Western blotting using an antibody against the amino-terminal sequence of MAPKK1 (MEK1-NT).
Effect of Vacuolar pH Elevation on the Formation of the SDS/Heatresistant PA63 Oligomer-To assess the effect of vacuolar pH elevation on de novo formation of SDS/heat-resistant PA63 oligomer, CHO CL6 cells were incubated at 4°C with 1 g/ml PA for 3 h and then washed five times with ice-cold HBSS. The cells were then incubated in fresh medium in the presence or absence of 0.2 M bafilomycin A1 at 37°C. Incubations at 37°C varied from 5 to 180 min. To assess the effect of vacuolar pH elevation on the preformed SDS/heat-resistant PA63 oligomer, cells were incubated with 1 g/ml PA at 37°C for 1 h and washed five times. One set of cells was then further incubated in fresh medium containing 0.2 M bafilomycin A1 and the other in medium without the drug. The incubations at 37°C varied from 10 to 180 min. The cells were lysed, and the lysates were processed by SDS-PAGE and Western blotting as described above.
Construction and Transfection of Human TEM8 Variants into CHO PR230 Cells-Human TEM8 variant cDNA fragments were isolated by reverse transcriptase-PCR from human fetal brain mRNA (catalog number 11438-017) purchased from Invitrogen. First-strand cDNA was synthesized by using the SuperScript First-strand Synthesis System (catalog number 11904-018) purchased from Invitrogen. We used 5Ј primer PR5 (AAGTGTACAATGGCCACGGCGGAGCGGAGAGCCCT-CGGCATCGGCT, the start codon ATG is underlined and the BsrGI site for cloning is in boldface) in combination with various 3Ј primers to amplify different carboxyl-terminal truncated TEM8 variants, as diagramed in Fig. 5A). Thus, primer 115 aa (CCCACAAGGCATCGAGTT-TTCCCTT, stop codon provided by the expression vector) was used to obtain the TEM8 variant TEM8-115 aa, having a 115-residue cytosolic tail. Similarly, primer 26 aa (CGGGATCCTAAGCGTAATCTGGAAC-ATCGTATGGGTAACCATCATCATCTTCTTCCTCACTCTCCTCGG-CA, the antisense of stop codon is underlined, the BamHI site is in boldface, and the sequence encoding for an influenza virus hemagglutinin (HA) tag is in italic) was used to obtain TEM8 -26 aa. Primer 16 aa (CGGGATCCTAAGCGTAATCTGGAACATCGTATGGGTAGGCA-GGGGGTGGAGGGACCTCCTTGATAAT, underlining, etc., as above) was used to obtain TEM8 -16 aa. Primer 0 aa (CGGGATCCTACCAG-AACCACCAGAGGAGAGCCAGGGCTA, underlining, etc., as above, and having no HA tag) was used to obtain TEM8 -0 aa. Primer ED (CGGGATCCTAACCGTCAGAACAGTGTGTGGTGGTGATGATGA-CA, underlining, etc., as above) was used to obtain TEM8-ed, the variant having only the extracellular domain, residues 1-320. Finally, primer v3 (CTATTCCATGCAAGCAGCTGTTGTGGGGCCTGATGCA-ATTTTGTGGAGGCTACAGTGTGTGGTGGTGATGATGACAGAACT-GGA, the antisense of stop codon is underlined) was used to obtain TEM8 variant 3. We found it is difficult to amplify full-length cDNA for TEM8 variant 1 due to its exceptionally high GC content, and instead we synthesized the 3Ј cDNA region of variant 1 chemically and ligated it into TEM8-115 aa, resulting in full-length TEM8 variant 1. The TEM8 variant cDNA fragments were digested by BsrGI alone or BsrGI and BamHI and then cloned between the BsrGI and EcoRV or BsrGI and BamHI sites of pIRESHgy2B (catalog number 6939-1, Clontech Laboratories, Inc., Palo Alto, CA). This bicistronic mammalian expression vector contains an attenuated version of the internal ribosome entry site of the encephalomyocarditis virus, which allows both the gene of interest and the hygromycin B selection marker to be translated from a single mRNA. We also constructed a glycosylphosphatidylinositol (GPI)-anchored TEM8 by fusion of the TEM8 extracellular region to the GPI anchoring sequence of urokinase plasminogen activator receptor (uPAR) (31). To do so, the GPI sequence of human uPAR was amplified by using primers U5 (TATCGTACGTTGTAACCACCCAGACCTG-GATGTCCAGT, the BsiWI cloning site is in boldface) and U3 (AATTC-CAGCACACTGGTTAGGTCCAGAGGAGAGTGCCT, the antisense of the stop codon is underlined, and the BstXI site is in boldface), with a template of uPAR plasmid phuPAR (kind gift from Dr. Thomas H. Bugge, National Institutes of Health, Bethesda). The PCR product was digested by BsiWI and BstXI and cloned between the BsiWI and BstXI sites of the plasmid encoding TEM8-ed, resulting in an expression plasmid encoding the TEM8 extracellular part (residues 1-317) and the GPI anchoring sequence of uPAR (residues 293-335) linked by short tripeptide IVR. All the expression plasmids were confirmed by DNA sequencing and were transfected into CHO PR230 cells using Lipo-fectAMINE Plus Reagent (Invitrogen), and stably transfected cells were selected by growth in hygromycin B (500 g/ml) for 2 weeks. Hygromycin-resistant colonies were either isolated individually or pooled for further analysis.
Cytotoxicity Assay with MTT-Cells were grown in 96-well plates to ϳ50% confluence. Serial dilutions of PA (0 -1000 ng/ml) combined with FP59 (100 ng/ml) were added to the cells to give a total volume of 200 l/well and were incubated for 48 h. Cell viability was then assayed by adding 50 l of 2.5 mg/ml 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide (MTT) in ␣-minimal essential medium. The cells were incubated with MTT for 45 min at 37°C; the medium was removed, and the blue pigment produced by viable cells was solubilized with 100 l/well of 0.5% (w/v) SDS, 25 mM HCl, in 90% (v/v) isopropyl alcohol. The plates were vortexed, and the oxidized MTT was measured as A 570 using a microplate reader.
RNA Isolation and Northern Hybridization-Total RNA was isolated from exponentially growing CHO cells by using TRIzol Reagent (Invitrogen), separated on 1.0% agarose, 6.66% formaldehyde gels and then transferred onto nylon membranes (Immobilon-N, Millipore). Membranes were hybridized with a 32 P-labeled 1.0-kb CHO TEM8 cDNA fragment isolated by reverse transcriptase-PCR by using 5Ј primer TTCTGCCAGGAGGAGACACTTACATGC, and 3Ј primer CCCACAAGGCATCGAGTTTTCCCTT. DNA sequencing analysis showed that this fragment was 90 and 89% identical to the corresponding mouse and human TEM8 sequences, respectively. Hybridization was performed in QuikHyb hybridization solution (catalog number 201220, Stratagene, La Jolla, CA) containing denatured salmon sperm DNA (0.5 mg/ml) at 60°C overnight. Membranes were then washed in 2ϫ SSC (catalog number 750020, Research Genetics, Huntsville, AL), 0.1% SDS at room temperature for 5 min, and then twice with 0.1ϫ SSC, 0.1% SDS at 60°C for 25 min.

PA63 Produced from PA by Cell Surface Proteolytic Cleavage
Rapidly Forms Oligomers-We used a high titered polyclonal anti-PA serum (number 5308) to assess PA binding and its subsequent processing and internalization by CHO cells. Analysis of unmodified PA eliminated concerns that radiolabeled or chemically labeled PA might behave differently from native PA due to the modification. The results showed that PA bound to CHO CL6 cells and was rapidly cleaved to PA63. More than half the PA bound was cleaved to PA63 within 5 min (Fig. 1A, CL6 lanes). Because furin is the major cell surface protease that cleaves PA (4, 5), we compared processing of PA by CHO FD11, a derivative of CHO CL6 cells that lacks furin (27). The FD11 cells bound PA efficiently, but cleavage to PA63 was very slow (Fig. 1A, FD11 lanes). In addition to intact PA and PA63, another PA species revealed on SDS-PAGE was an SDS/heatresistant PA63 oligomer that migrated very slowly (Fig. 1A). Because the formation of the PA63 oligomer requires proteolysis, the PA63 oligomer was hardly detected in cell lysates from FD11 (Fig. 1A). The cell lysates were further analyzed by native-PAGE. The FD11 cell lysates contained mainly the intact PA (Fig. 1B, FD11 lanes), as expected, whereas the lysates from CHO CL6 cells contained intact PA as well as two higher order PA63 oligomers, but no monomeric PA63 (Fig. 1B, CL6  lanes). When the binding assay was performed by using trypsin-nicked PA in which the furin site was pre-cleaved by limited trypsin digestion (22), as expected the major PA species detected were the oligomers, and just a negligible amount of monomer PA was shown (Fig. 1C). The nature of these two distinct PA63 oligomers is unclear. The faster migrating species, termed oligomer A in this study (Fig. 1, B and C), is probably free PA63 heptamer, whereas the more slowly migrating species, termed oligomer B (Fig. 1, B and C), may be a complex of the PA63 heptamer with cellular components such as the PA receptor or detergent-resistant membrane structures. When these oligomeric species (in Fig. 1C, lane 1h) were subjected to second dimension SDS-PAGE, interestingly, oligomer A dissociated to PA63 monomer, whereas oligomer B turned out to be the mixture of both SDS-sensitive and -resistant oligomers (Fig. 1D). Together these results showed not only that bound PA is rapidly cleaved by furin but also that the resulting PA63 monomer very rapidly oligomerizes. Thus, the cell surface-associated PA63 mimics the behavior of PA63 produced in vitro, which forms the heptamer in neutral aqueous solutions (21). These heptamers have been visualized previously by electron microscopy (6), x-ray diffraction (7), and elec-trophoresis (22,32). Absence of PA63 monomer further indicated that oligomerization of PA63 is effectively irreversible. PA63 detected by SDS-PAGE (Fig. 1A) evidently resulted from the resolution of PA63 oligomer by boiling in SDS loading buffer.
PA63 Oligomer Is the Only PA Species That Is Internalized-The data above indicated that cells exposed to PA contain intact PA and PA63 oligomers on their surface (Fig. 1B). To explore whether these PA species are equally internalized, we performed a PA trypsin protection assay. After incubation with PA at 4 or 37°C, cells were treated with trypsin to remove the cell surface-bound PA, allowing identification of those materials internalized by endocytosis. Remarkably, PA63 oligomer constituted the major protected PA species at 37°C (Fig. 1, A  and B, CL6 lanes), indicating that the PA63 oligomer was the only form of PA to be internalized. Also present were small amounts of a PA fragment, probably the carboxyl-terminal 47-kDa receptor-binding portion remaining bound to receptor after incomplete cleavage by trypsin (33). Endocytosis is tem- FIG. 1. Binding, proteolytic processing, and internalization of PA by CHO cells. A and B, PA63 oligomer is the only PA species internalized. CHO CL6 and FD11 cells were incubated with PA (1 g/ml) for different lengths of time at 37°C, washed to remove unbound PA, and lysed. Cell lysates prepared directly or after the cells were incubated with trypsin to remove cell surface-bound PA species were separated by SDS-PAGE (A) or native-PAGE (B) followed by Western blotting using PA antiserum (number 5308). C and D, two-dimensional electrophoretic analysis of PA oligomers. CHO CL6 cells were incubated with the trypsin-nicked PA (nPA, 1 g/ml) for 0 or 1 h, and cell lysates were separated by native-PAGE (C). A gel slice like the 1-h sample containing oligomers A and B was subjected to a second dimension SDS-PAGE analysis (D). E, cell surface-associated PA can be efficiently removed by trypsin. CHO CL6 cells were incubated with PA at 4°C and treated as in A. F, uncleavable PA cannot be internalized. CHO CL6 cells were incubated with PA-U7 at 37°C and treated as in A.
perature-dependent (10), and therefore all surface-bound PA should be removed by trypsin from cells incubated at 4°C. Thus, in a control for the previous experiment, we showed that trypsin removed all the cell-associated PA (Fig. 1E), with the exception of the 47-kDa fragment mentioned above. Further evidence that intact, monomeric PA is not internalized into cells was obtained using PA-U7, an uncleavable variant of PA that can bind but cannot be proteolytically activated by cellular furin (24). This PA mutant was not internalized to a trypsinresistant site even when incubated with cells at 37°C for 4 h (Fig. 1F). These results demonstrated that the proteolytic cleavage of receptor-bound PA is an absolute prerequisite not only for the biochemical property of self-assembly but also for its subsequent biological activity of undergoing endocytosis.
Oligomeric PA63 Endocytosed to Acidic Compartments Is Rapidly Transformed to an SDS/Heat-resistant Form-Previous studies showed that in solution, purified PA63 can form two types of oligomers, an SDS-sensitive type, which forms at neutral pH and can be resolved into PA63 monomer by SDS, and the SDS-resistant type, which forms at acidic pH and persists in the presence of SDS (34). This suggested that the SDS/heat-resistant oligomer shown above (Fig. 1A) may be the counterpart of this SDS-resistant oligomer formed in acidic solution, and therefore may be produced following endocytosis and delivery to acidic compartments. In fact, when CHO CL6 cells were incubated with PA, the SDS/heat-resistant oligomer was formed at 37°C (Fig. 1A, CL6 lanes) but not at 4°C, a temperature at which endocytosis does not occur (Fig. 1E, 1h  lane). Moreover, native-PAGE analysis revealed that the PA63 oligomer A formed at both 4 and 37°C ( Fig. 2A). Based on these observations we hypothesized that the PA63 oligomer formed on the cell surface encounters a progressively more acidic environment along the endocytic pathway and undergoes conformational changes and membrane insertion at acidic pH that renders it resistant to SDS/heat. To verify this hypothesis, we incubated CHO CL6 cells with PA at 4°C, washed, and then shifted to 37°C for various lengths of time in the absence or presence of bafilomycin A1, a potent and specific inhibitor of the vacuolar (H ϩ )-ATPase proton pumps that maintain the pH gradients of acidic compartments (35,36). We showed that following 3 h of incubation at 4°C, little, if any, SDS/heatresistant oligomer formed (Fig. 2B, ϪBafilomycin, 0 lane), although a significant amount of oligomer was present when lysates were analyzed by native gels (data not shown). However, the resistant oligomer formed efficiently in a time-dependent manner when the cells were shifted to 37°C following binding at 4°C (Fig. 2B, Ϫ Bafilomycin lanes). Moreover, the amount of resistant oligomer quickly declined after peaking at 30 -40 min of incubation at 37°C, and less than half remained after 60 min (Fig. 2B, Ϫ Bafilomycin lanes). In contrast, formation of the SDS/heat-resistant PA63 oligomer was potently inhibited when bafilomycin A1 was included at the time when the cells were shifted to 37°C (Fig. 2B, ϩ Bafilomycin lanes), indicating that the SDS/heat-resistant PA63 oligomer is produced within acidic compartments. Because the resistant oligomer began to appear as early as 5 min after shifting to 37°C, it likely formed in early endosomes, the first compartments along the endocytic pathway that have a significantly acidic pH. Because only trace amounts of monomeric PA63 were detected in the trypsinized samples shown above (Fig. 1A, trypsin ϩ lanes), we conclude that the majority of internalized PA63 oligomer is rapidly transformed to the SDS/heat-resistant form in endosomes following internalization.
The chymotrypsin-sensitive loop in PA domain 2 is involved in membrane insertion (7,22,37). We used the PA protein altered in this loop to show that formation of the SDS/heatresistant oligomer requires membrane insertion. The mutated PA protein, PA-⌬ FF, with 313 FF 314 at the tip of the chymotrypsin-sensitive loop deleted (22), was bound, processed, and internalized normally (Fig. 3A), but its transformation into the SDS/heat-resistant form in acidic compartments was greatly decreased (Fig. 3B). This in turn made the mutated PA unable to translocate LF into the cytosol of CHO CL6 cells, observed both biochemically (Fig. 3C) and through the failure to cause MAPKK1 cleavage (Fig. 3D). These data show that membrane insertion and channel formation by the chymotrypsin-sensitive loop is the basis of the conformational changes that render the oligomer resistant to boiling in SDS sample buffer.
In the experiment using bafilomycin, we noted that the small amount of resistant oligomer that appeared late in the presence of bafilomycin A1 persisted for at least 120 min (Fig. 2B, ϩ Bafilomycin lanes), suggesting both its formation and degradation were arrested by bafilomycin A1. We therefore assessed the oligomer degradation more systematically. The acidic pH of early endosomes is important for sorting endocytosed material, influencing whether materials are recycled to the plasma membrane, or transported to lysosomes for degradation. The SDS/ heat-resistant oligomer is likely transported to lysosomes for degradation. To examine oligomer stability, cells were incubated with PA for 1 h at 37°C and then washed to remove PA. At this point, all the compartments in the endocytic pathway, including cell surface, endosomes, and lysosomes, should be loaded with the corresponding PA species. During a subsequent incubation at 37°C in the absence of bafilomycin A1, it was found that the SDS/heat-resistant PA63 oligomer remained at the same level for 30 min and then decreased with time, completely disappearing after 180 min (Fig. 4, Ϫ Bafilomycin  lanes). In contrast, oligomer degradation was completely inhibited in the presence of bafilomycin A1 (Fig. 4, ϩ Bafilomycin lanes). Bafilomycin A1 probably arrested the degradation of the SDS/heat-resistant PA63 oligomer by inhibiting its transport to lysosomes and also by decreasing the activity of lysosomal FIG. 2. SDS/heat-resistant PA63 oligomer forms in acidic endocytic compartments. A, PA63 oligomer forms at both 37 and 4°C. CHO CL6 cells were incubated with 1 g/ml PA at 4 or 37°C for 1 h, washed, and lysed. The cell lysates were subjected to native-PAGE, followed by Western blotting using PA antiserum (number 5308). B, SDS/heatresistant PA63 oligomer formation requires internalization and endosome acidification. CHO CL6 cells were incubated with PA (1 g/ml) at 4°C for 3 h, washed, and changed to fresh medium. The cells were shifted to 37°C in the absence or presence of bafilomycin A1 (0.2 M) for various lengths of time and lysed. The cell lysates were analyzed as in Fig. 1A. proteases that prefer an acidic environment (38). The stability of the SDS/heat-resistant PA63 oligomer in cells treated with bafilomycin A1 also suggests that the acid-induced conformational change is effectively irreversible, because elevating the pH of endocytic compartments does not produce monomeric PA63.
TEM8-mediated PA Binding, Proteolytic Processing, and Endocytosis Are Independent of Its Cytoplasmic Sequence-An important advance in understanding anthrax toxin action was the recent demonstration that TEM8 variant 2 (368 residues, designated ATR) acts as a PA receptor (3). However, no information is available regarding the ability of other TEM8 gene products to serve as PA receptors. TEM8-related proteins produced through alternative splicing also include TEM8 variant 1 (564 residues) and variant 3 (333 residues) (Fig. 5A). TEM8 variants 1 and 2 contain a putative 23-residue transmembrane region (residues 321-343), whereas TEM8 variant 3 lacks this sequence and instead has a unique 15-residue carboxyl-terminal sequence whose properties suggest that the protein will be released as a soluble form that would not function as a receptor. TEM8 variant 2 (ATR) has a very short putative cytoplasmic tail (25 residues) with a 360 EESEE 364 acidic cluster that was suggested to cause co-localization of ATR with the furin protease whose action is required for PA activation (3). The TEM8 variant 1 cytoplasmic region (221 residues) includes a more extended acidic cluster 360 EESEEEDDD 368 , a prolinerich region (34 prolines within the last 57 residues), and 14 potential phosphorylation sites.
To explore the roles of the various TEM8 variants as PA receptors and experimentally evaluate the roles of the motifs in the cytoplasmic region of TEM8 in PA binding, proteolytic processing, and internalization, we constructed eight variants of human TEM8 (Fig. 5A). The corresponding cDNA fragments were cloned into the bicistronic mammalian expression vector pIRESHyg2b. The resulting expression plasmids were transfected into PR230, a spontaneous CHO cell PA receptor-deficient mutant. In contrast to its parental CHO cell line WTP4, PR230 is specifically defective in PA binding and thus is resistant to PA plus FP59 but sensitive to diphtheria toxin and Pseudomonas exotoxin A (data not shown). Stably transfected cells were established by hygromycin B selection for 2 weeks, and the hygromycin-resistant colonies were either isolated individually or pooled. We found that Ͼ80% of the hygromycinresistant clones expressed the transfected genes (data not shown). Therefore, in addition to analyzing representative isolated clones from selected TEM8 variant constructs (Fig. 5, B   FIG. 3. PA-⌬ FF has decreased ability to form SDS/heat-resistant oligomer and to translocate LF into the cell cytosol. A and B, PA-⌬ FF is impaired in SDS/heat-resistant oligomer formation. CHO CL6 cells were incubated with 1 g/ml PA or PA-⌬ FF for 1 h at 37°C and treated as in Fig. 1, and cell lysates were separated by SDS-PAGE (A) or native-PAGE (B) for Western blotting using PA antiserum (number 5308). C and D, PA-⌬ FF fails to translocate LF into the cytosol (C) or to produce MAPKK1 cleavage (D). CHO CL6 cells were incubated with 1 g/ml PA or PA-⌬ FF plus 1 g/ml LF for 1 h at 37°C; the cells were then trypsinized, permeabilized by saponin, and separated into cell pellet and supernatant cytosol fractions as described under "Experimental Procedures." The cell pellet lysates (p) and the cytosol (c) were separated by native-PAGE followed by Western blotting using LF antiserum (number 5309) (C). CHO CL6 cells were incubated with 1 g/ml PA or PA-⌬ FF plus 1 g/ml LF for 1 h at 37°C; the cell lysates were then separated by SDS-PAGE followed by Western blotting using a antibody that detects the amino-terminal sequence of intact but not cleaved MAPKK1 (MEK1-NT) (D). and C), we fully analyzed pools of transfectants to obtain results characteristic of average expression levels (Fig. 5, D and  E). Remarkably, all the TEM8 variants or truncated mutants containing the putative transmembrane domain and extracellular region, including TEM8 variant 1, TEM8 -115 aa, TEM8 -26 aa, TEM8 -16 aa, and TEM8 -0 aa, were functional PA receptors (Fig. 5). These transfected PR230 cells regained the ability to bind and proteolytically process PA, leading to the internalization and conversion of the PA63 oligomer into the SDS/heat-resistant form to an extent matching that of the parental WTP4 cells (Fig. 5, B and D). Moreover, these transfected cells became sensitive to killing by PA plus FP59 (Fig. 5,  C and E). In contrast, cells transfected with the TEM8-ed construct that is truncated before residues 321-343, the putative transmembrane domain, failed to bind PA and were resistant to PA plus FP59 (Fig. 5, D and E). Similarly, cells transfected with TEM8 variant 3 or mouse vacuolar protein sorting protein 11 (Vps11) (39), an irrelevant gene, also were unable to bind PA and remained resistant to PA plus FP59 (Fig. 5, D and  E). These data support the prediction that residues 321-343 constitute a transmembrane anchor and imply that TEM8 variant 3 is a naturally occurring soluble form of TEM8. Taken together, the results demonstrate that the TEM8 extracellular region and transmembrane domain together constitute the minimum PA receptor structure. As an alternative to being anchored by transmembrane domains, some cell surface pro-teins are retained at the surface by post-translational linkage to GPI. To explore further what properties are required in a TEM8 protein for it to act as a PA receptor, we constructed an expression vector that expresses a fusion of the TEM8 extracellular domain (residues 1-317) with the GPI-anchoring sequence of uPAR (residues 294 -335). Interestingly, this GPIanchored TEM8, designated TEM8-GPI, could also function as a PA receptor, supporting the binding and processing of PA into PA63, formation of the SDS/heat-resistant oligomer, and making cells sensitive to PA plus FP59 (Fig. 5). The GPI-anchored nature of TEM8-GPI was confirmed by showing that PI-PLC treatment to release GPI-anchored proteins (40) significantly decreased PA binding to TEM8-GPI-transfected cells without affecting binding to parental WTP4 cells or cells transfected with TEM8 -0 aa (Fig. 6). These results clearly demonstrate that anthrax toxin receptor-mediated internalization can be mediated by TEM8 regardless of how it is anchored to the cell surface.
Like CHO CL6 cells (Figs. 1-4), WTP4 cells can only internalize the oligomeric form of PA63 (Fig. 7), as demonstrated by trypsin treatment method. We found that constructs with no cytosolic tail (TEM8 -0 aa, Fig. 7, middle panel) or with a GPI anchor (TEM8-GPI, Fig. 7, right panel) had this same ability to internalize the oligomeric PA but not the monomer PA (Fig. 7).
Northern blot analysis revealed that CHO WTP4 cells express one major TEM8 transcript, one having a size slightly FIG. 5. TEM8-mediated PA binding, proteolytic processing, and endocytosis are independent of its cytoplasmic sequence and the type of cell surface anchoring. A, schematic representation and the sequences alignment of various TEM8s constructed and transfected into CHO PR230 cells. TM, putative transmembrane region shown in green; AC, acidic cluster shown in red; PC, proline cluster shown in blue. Residues 1-300 are not shown in the sequence alignment; HA tag is not capitalized when present; GPI sequence is underlined; the additional five residues DRTLT at the carboxyl terminus of TEM8 -115 aa was produced due to DNA manipulation. Interactions of PA with the various TEM8transfected PR230 cells either isolated (B) or pooled (D). The stably transfected cells were incubated with 1 g/ml of PA for 1 h at 37°C; the cells were then analyzed by SDS-PAGE and Western blotting as in Fig. 1A. Cytotoxicity of PA plus FP59 to the various TEM8 transfected PR230 cells either isolated (C) or pooled (E). Cells were incubated with various concentrations of PA plus 100 ng/ml of FP59 for 48 h, and the cell viability was assessed as described under "Experimental Procedures." larger than 4.4 kb (Fig. 8). This is likely to be TEM8 variant 1, which has a full-length cDNA of 5.2 kb in mouse and 5.4 kb in human. This transcript was also expressed by PR230 cells but at a significantly decreased level. This suggests that one allele in PR230 does not produce a stable TEM8 transcript. This first defective allele could either pre-exist in the CHO WTP4 parent or result from the mutation in PR230, whereas the second defective allele in PR230 must consist of a more subtle change that still allows production of the transcript observed in Fig. 8.
The CHO PR230 mutant cell line is one of a number of PA receptor-deficient CHO mutants obtained in this laboratory by either spontaneous mutation, chemical mutagenesis, or retrovirus insertional mutagenesis. 2 All the PA receptor-deficient CHO mutants analyzed so far by cell fusion belong to a single genetic complementation group, and their phenotype is restored to that of parental cells by transfection with TEM8 -26 aa (data not shown). These data strongly suggest that the naturally occurring TEM8 variants, all produced from a single gene, are the only cell surface molecules that act as functional PA receptors. DISCUSSION PA is a central part of anthrax toxin, serving as a cellular binding moiety and a delivery vehicle for translocation of LF and EF into the cytosol of cells. Therefore, the interaction between PA and the target cell has been a central issue in studies of anthrax toxin. Here, we took advantage of a highly sensitive and specific anti-PA antiserum, instead of using radiolabeled or biotin-labeled PA that may behave differently due to modification, to assess directly the binding, processing, and internalization of PA by host cells. These analyses revealed that PA not only rapidly binds but also is rapidly cleaved by furin protease on the CHO cell surface, with both processes nearing completion in 5 min. Remarkably, the resulting PA63 was very rapidly converted to the oligomer, so that only monomeric PA (PA83) but not monomeric PA63 could be detected. Surprisingly, we did not find any monomeric form of PA inside the cells, and the PA63 oligomer was the only form of PA internalized. These results are consistent with the previous report (41) that the proteolytic activation of PA on macrophages promotes its internalization. Taken together, these results indicate that oligomerization of PA63 on the cell surface triggers receptor-mediated endocytosis, explaining that only PA63 oligomer but not the monomeric form of PA was internalized. Therefore, proteolytic cleavage of receptor bound PA is a fundamental prerequisite not only for PA self-assembly and LF and EF binding but also for the subsequent step of endocytic internalization.
We further assessed the role that the acidic pH of intracellular compartments plays in the formation and degradation of the PA oligomers. PA could be processed to form an SDS/heatsensitive PA63 oligomer on the cell surface at 4°C, a temperature at which endocytosis does not occur. However, when endocytosis was allowed by shifting the temperature from 4 to 37°C, the PA63 oligomer began to transform to the SDS/heatresistant form, in a process that was efficiently blocked by bafilomycin A1, a specific inhibitor of the vacuolar ATPases that generates the pH gradient of acidic compartments. These results clearly demonstrated that transformation of PA63 oligomer to the SDS/heat-resistant form was induced by acidic pH in endocytic compartments. Furthermore, this conformational change was functionally correlated with membrane insertion, channel formation, and translocation of LF into cytosol, because a mutated PA protein defective in membrane insertion, PA-⌬ FF, underwent significantly decreased SDS/heat-resistant oligomer formation and failed to translocate LF into cytosol as assessed by MAPKK1 cleavage. It was also noted that the SDS/heat-resistant PA63 oligomer formation was irreversible, because bafilomycin A1 could not affect the resistant oligomer already formed inside the cells. Another finding in this work is that bafilomycin A1 could completely inhibit the PA63 oligomer degradation which otherwise is complete within 3 h, suggesting that the degradation takes place in a low pH compartment, probably the lysosome.
Because the PA63 band formed by dissociation of the SDS/ heat-sensitive oligomer was barely detectable inside cells (Fig.  1A, trypsin ϩ lanes), we concluded that the majority of the internalized PA63 oligomer rapidly transformed to the SDS/ heat-resistant form in the acidic environment in endosomes following the endocytosis. Therefore, the PA63 SDS/heat-sensitive oligomer, the source of the PA63 band in the SDS gels, mainly exists on the cell surface. Thus, the progressive decrease of PA63 on the SDS gel (Fig. 4 reflects the endocytic process. The clearance of PA63 (Fig. 4, ϪBafilomycin lanes) required about 2 h, indicating that the endocytic process is a slow and rate-limiting step in toxin internalization.
Taken together, the data show that the anthrax toxin system has evolved to deliver efficiently LF and EF to the cytosol. PA is retained on cell surface receptors until it is cleaved and converted to the oligomer, the first species able to bind LF and EF (9). Endocytosis, initiated after oligomerization, is relatively slow compared with furin cleavage, so that the PA oligomer remains on the surface for a sufficient time to allow LF (or FP59) and EF to bind. This mechanism makes it probable that the internalized PA oligomers are loaded with at least one molecule of the enzymatic moieties before entry. This may explain why CHO cells are very sensitive to PA, having an EC 50 only 1-2 ng/ml of PA in the presence of FP59.
The other objectives of this work were to determine whether TEM8 variants 1 and 3 can also function as PA receptors, to explore the roles of the TEM8 cytosolic tail in PA binding, proteolytic processing, oligomerization, and internalization, and to experimentally define the transmembrane region. The three reported TEM8 variants share the same amino-terminal extracellular part but differ in length and sequence in their putative cytosolic regions. TEM8 variant 1 is the longest and possesses a more complicated cytoplasmic region. Surprisingly, all the TEM8 mutants we constructed that retained a membrane anchor functioned as PA receptors, as determined by PA binding, processing, oligomer formation, and translocation of FP59 into cells. The shortest mutant of TEM8 that acted as a PA receptor is TEM8 -0 aa, which was truncated at the carboxyl terminus of the putative transmembrane domain. When the predicted transmembrane domain (residues 321-343) was further deleted, the resulting mutated TEM8-ed transfected cells failed to bind PA, demonstrating that residues 321-343 are responsible for TEM8 cell surface anchoring. Cells transfected with either the naturally occurring TEM8 variant 3 or the TEM8-ed construct could not bind PA, suggesting that the resulting proteins are soluble forms of TEM8. Secretion of a soluble receptor by some cells may have an impact on the normal function of TEM8 and any endogenous ligands that normally bind to it.
The evidence that a simple transmembrane anchor was sufficient for TEM8 to function as a PA receptor suggested that other forms of membrane anchor might also be sufficient. Surprisingly, when we linked the TEM8 extracellular domain to the GPI-anchoring sequence of uPAR, the chimeric protein still functioned as a PA receptor, clearly demonstrating that the ability of TEM8 to act as a receptor is independent of the type of membrane anchor. The physiological function of TEM8 is still unknown. A portion of the extracellular part of TEM8 shares high sequence homology with the von Willebrand factor type A domain. Because von Willebrand factor type A domains are often found in extracellular domains of integrins where they constitute ligand-binding sites, TEM8 may be involved in the interaction of cells with the surrounding extracellular matrix. Although the complex TEM8 cytosolic domain is not required for PA receptor function, it may play a crucial role in its unidentified normal physiological function. Therefore, it will be important to identify any natural TEM8 ligands and to determine the consequences of ligand binding, in part because of the role TEM8 plays in anthrax toxin action. In this respect, a mutated PA defective in TEM8 binding may be considered in the development of new anthrax vaccines so as to avoid any potential effects of PA that follow from its binding to TEM8.