Anthrax Toxin Entry into Polarized Epithelial Cells

Abstract

We examined the entry of anthrax edema toxin (EdTx) into polarized human T84 epithelial cells using cyclic AMP-regulated Cl⁻ secretion as an index of toxin entry. EdTx is a binary A/B toxin which self assembles at the cell surface from anthrax edema factor and protective antigen (PA). PA binds to cell surface receptors and delivers EF, an adenylate cyclase, to the cytosol. EdTx elicited a strong Cl⁻ secretory response when it was applied to the basolateral surface of T84 cells but no response when it was applied to the apical surface. PA alone had no effect when it was applied to either surface. T84 cells exposed basolaterally bound at least 30-fold-more PA than did T84 cells exposed apically, indicating that the PA receptor is largely or completely restricted to the basolateral membrane of these cells. The PA receptor did not fractionate with detergent-insoluble caveola-like membranes as cholera toxin receptors do. These findings have implications regarding the nature of the PA receptor and confirm the view that EdTx and CT coopt fundamentally different subcellular systems to enter the cell and cause disease.

Although a large number of bacterial toxins are known to act within the cytosol of mammalian cells, there is still no toxin for which we fully understand the mechanism of entry. The entry process is known to be complex, at least for most intracellularly acting toxins, and a broad variety of experimental approaches will be needed to elucidate its many facets. In the course of our studies on toxins that elevate intracellular cyclic AMP (cAMP) levels, we have made use of the human polarized epithelial cell line, T84, which exhibits cAMP-regulated Cl⁻ secretion. Cl⁻ secretion from T84 cells can be measured electrically with a high degree of sensitivity and temporal resolution (7, 18, 19). Using this system, we have shown that the action of cholera toxin (CT) requires that the toxin be trafficked to the cis Golgi or endoplasmic reticulum (17, 18) and that its entry into acidic endosomes is not sufficient to elevate cAMP levels (21).

In the present studies, we used T84 cells to extend our knowledge of the action of anthrax edema toxin (EdTx). Like CT, EdTx elevates cAMP levels within cells, but it does so through a different mechanism. Whereas CT activates the host cell’s adenylate cyclase by ADP ribosylation of a subunit of the regulatory trimeric G protein (32), EdTx contains a subunit (edema factor [EF]; size, 89 kDa) that is a calmodulin-dependent adenylate cyclase, which directly catalyzes formation of cAMP in the cytosol (22, 23).

EdTx is a binary toxin which is assembled at the surface of receptor-containing mammalian cells from its component parts, EF and anthrax protective antigen (PA; size, 83 kDa). PA serves as the toxin’s B moiety, mediating receptor binding, self assembly, and translocation of EF to the cytosol. (PA also mediates delivery of an alternative enzymic moiety, lethal factor, to the cytosol.) PA binds to an as-yet-unidentified cell surface receptor which is saturable and at least partly proteinaceous (10). The receptor appears to be ubiquitous, as all cell types thus far tested respond to EdTx. PA is then cleaved by furin or a related protease, generating a small N-terminal 20-kDa fragment, which is released into the medium, and a C-terminal 63-kDa fragment (PA63), which remains bound to the receptor (12, 23, 38). PA63 spontaneously oligomerizes to form a heptameric, ring-shaped oligomer (4, 31) which binds EF. The stoichiometry and order of these assembly steps are not known. The EF-PA63-receptor complex enters the cell by endocytosis and is trafficked to an acidic compartment where low pH triggers insertion of the PA63 heptamer into the membrane and translocation of EF to the cytosol (4, 35).

As detailed below, T84 cells are sensitive to EdTx, and the overall characteristics of EdTx entry closely resemble those found on nonpolarized cells. However, our studies have revealed an unexpected asymmetric distribution of receptors on these cells. The results of these studies give clues to the identity of the anthrax toxin receptor and serve as the basis for further application of this system to study toxin action.

MATERIALS AND METHODS

Materials.

CT was purchased from Calbiochem, La Jolla, Calif. Polyclonal anti-PA antibody was raised in rabbits. All other reagents, including the bafilomycin A₁ and the horseradish peroxidase-labeled goat anti-rabbit antibody, were obtained from Sigma Chemical Co. (St. Louis, Mo.) unless otherwise stated. Hank’s balanced salt solution (HBSS) was used for all intact cell assays and contained 0.185 (per liter) g of CaCl₂, 0.098 g of MgSO₄, 0.4 g of KCl, 0.06 g of KH₂PO₄, 8 g of NaCl, 0.048 g of Na₂HPO₄, and 1 g of glucose, to which was added 10 mM HEPES (pH 7.4).

Toxin preparation and purification.

PA and EF were purified from Bacillus anthracis (Sterne strain), as previously described (23), or from Escherichia coli BL21(DE3). Toxins from both sources gave similar results. The PA purification from E. coli was performed as described previously (4), with modifications as follows: periplasmic proteins were extracted first with a 10-min incubation in 0.4 volume of 30 mM Tris-HCl (pH 8.0)–20% sucrose–1 mM EDTA. This mixture was centrifuged, and the resulting pellet was resuspended in 5 mM MgSO₄ and incubated on ice for 10 min. The mixture was again centrifuged, and the resulting supernatant was concentrated, and the buffer was exchanged into 20 mM Tris (pH 8.0) by using a minitan tangential flow concentrator (Amicon, Beverly, Mass.). The concentrated supernatant was subjected to anion-exchange chromatography as described previously, but the buffer A lacked dithiothreitol and EDTA. Proteins were stored in buffer A at −80°C until use. The mutant forms of PA, PASSSR, and PA--D, were purified from E. coli as described above. The construction and characterization of the mutants were done as described previously (2, 12, 39).

For EF purification from E. coli, the wild-type gene for EF was amplified from the B. anthracis toxin plasmid pXO1 with the following primers: 5′-GATCGATCCATATGAATGAACATTACACTGAGAG-3′ and 5′-GATCGATCGGATCCTCATTATTTTTCATCAATAATTTTTTGG-3′. The PCR product was digested with NdeI and BamHI and ligated into the E. coli expression vector pET15b (Novagen, Milwaukee, Wis.). Ligation products were transformed into E. coli XL1-Blue (Stratagene, La Jolla, Calif.). Following sequencing to confirm proper cloning, the plasmid was put into E. coli BL21(DE3) for expression. EF was purified on a nickel column via its histidine tag following the manufacturer’s protocol (Novagen). After overnight dialysis against 1 liter of 20 mM Tris (pH 8.0), the eluate was subjected to anion-exchange chromatography (Mono Q; Pharmacia, Piscataway, N.J.) in 20 mM Tris-HCl (pH 8.0). The preparation gave a single band by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and Coomassie staining.

Cell culture.

T84 cells were obtained from the American Type Culture Collection (Manassas, Va.) and cultured as previously described (7) in equal parts of Dulbecco’s modified Eagle medium plus 1 g of glucose per liter and Ham’s F-12 medium supplemented with 5% newborn calf serum, 15 mM HEPES, 14 mM NaHCO₃, 40 mg of penicillin per liter, 0.9 mg of streptomycin per liter, and 8 mg of ampicillin per liter. Cells were seeded at confluent density onto 5-cm² or 0.33-cm² Transwell inserts (Costar, Cambridge, Mass.) coated with a dilute collagen solution (25). They were then grown for 7 to 15 days, during which time they formed confluent monolayers of polarized columnar epithelium with tight intercellular junctions, high transepithelial resistances (>1,000 Ω/cm²), and a regulated Cl⁻ secretory pathway analogous to that found in intact intestine (9).

Electrophysiology.

Confluent T84 monolayers on Transwell inserts (5 cm² or 0.33 cm²) were moved to HBSS for measurements of short-circuit current (I_sc) and transepithelial resistance, as previously described (7, 19, 25). Five percent agar bridges made with Ringer’s buffer (114 mM NaCl, 5 mM KCl, 1.65 mM Na₂HPO₄, 0.3 mM NaH₂PO₄, 25 mM NaHCO₃, 1.1 mM MgSO₄, 1.25 mM CaCl₂) were used to interface serosal and mucosal reservoirs with calomel and Ag-AgCl electrodes. Transepithelial potentials in the absence or presence of applied 25- or 50-μA currents were measured with a dual voltage/current clamp device (University of Iowa, Iowa City, Ia.) I_sc was calculated by Ohm’s law. Monolayers that did not maintain a resistance of >500 Ω/cm² were excluded from the study.

Where indicated, bafilomycin A₁ was applied basolaterally to a final concentration of 0.5 μM, diluted from a stock made with dimethyl sulfoxide (DMSO). Dimethyl sulfoxide alone in these concentrations had no effect. The Cl⁻-free buffer contained 140 mM sodium gluconate, 5 mM potassium gluconate, 1.25 mM calcium acetate, 1 mM magnesium acetate, 5 mM NaH₂PO₄/Na₂HPO₄, 10 mM HEPES, and 5 mM glucose (pH 7.4).

Selective cell surface binding.

T84 monolayers (5 cm²) were cooled to 4°C for 15 min, rinsed in HBSS, and transferred to a clean 6-well plate containing HBSS at 4°C. PA was added to either apical or basolateral reservoirs at 3 μg/ml (3.6 × 10⁻⁸ M), final concentration. The tissue culture dish was covered in ice. After 1 h at 4°C, the monolayers were washed extensively, with careful attention to keep the apical and basolateral reservoirs from mixing. The filters were cut from the Transwell inserts, placed in 1 ml of differential extraction buffer (150 mM NaCl, 10 mM Tris, 1% Triton X-100, 350 μM phenylmethylsulfonyl fluoride, 20 μg of chymostatin per ml [pH 7.5]), vortexed extensively, and tumbled end over end for 30 min at 4°C. After being vortexed again, the extracts were clarified by centrifugation at 15,000 × g for 15 min at 4°C. The postnuclear supernatant was moved to a fresh centrifuge tube. Proteins were precipitated from 200 μl of the postnuclear supernatant by incubation for at least 1 h with 1 ml of acetone at −20°C. The protein pellet was harvested by centrifugation for 15 min at 15,000 × g (4°C) and resuspended in 50 μl of 6× reducing protein sample buffer (1). Samples were boiled for 2 min at 100°C, resolved by SDS-PAGE (7.5% gel), and transferred overnight onto nitrocellulose (Transblot transfer medium; Bio-Rad, Hercules, Calif.). PA was detected by Western blotting, with rabbit antiserum raised against PA (1:2,500 dilution) and horseradish peroxidase-labeled goat anti-rabbit secondary antibody (Sigma). The Western blot was developed by chemiluminescence (Supersignal substrate Western kit; Pierce, Rockford, Ill.) and exposed on Kodak Biomax MR-1 film.

Sucrose equilibrium density centrifugation.

One or two confluent monolayers of T84 (45 cm² each) were used for isolation of detergent-insoluble membranes. All steps were performed at 4°C. Cells were scraped into 2 ml of ice-cold differential extraction buffer (see above) and homogenized with five strokes in a tight-fitting Dounce homogenizer on ice. The homogenate was adjusted to 40% sucrose by the addition of 2 ml of 80% sucrose in differential extraction buffer, layered under a linear 5 to 30% sucrose gradient, and centrifuged at 39,000 rpm for 16 to 20 h in an SW41 rotor (Beckman Instruments, Palo Alto, Calif.). The presence of a floating membrane fraction was noted visually. Sequential 0.5- or 1-ml fractions were collected from the top of the gradient. From each fraction 20 μl was analyzed by SDS-PAGE and Coomassie staining or Western blotting. Sucrose density was monitored by refractometry.

RESULTS

Anthrax EdTx (10 μg of PA per ml and 0.1 μg of EF per ml) was found to elicit a strong transepithelial current (I_sc) from human intestinal T84 cells when it was applied to the basolateral surface of these cells but not when it was applied to the apical surface (Fig. 1A). The viability of cells treated apically with EdTx was confirmed by demonstrating that a cAMP agonist, vasoactive intestinal peptide (VIP), elicited a robust I_sc. The I_sc elicited by basolaterally added EdTx was dependent on the presence of both EF and PA. Two PA mutants known to be defective in mediating EF entry into CHO-K1 and RAW264.7 cells (12, 39) were unable to elicit EdTx-dependent Cl⁻ secretion (Table 1). One of the mutants (PASSR) lacked the furin cleavage site necessary for PA activation (12), and the second (PA--D) had a deletion in the chymotrypsin-sensitive loop that mediates pore formation and translocation (35, 39). Bafilomycin A₁, which inhibits the vacuolar H⁺ ATPase and collapses the pH gradient of endocytic vesicles, completely inhibited the EdTx-induced Cl⁻ secretory response (Table 1), consistent with prior evidence that EdTx action depends on its entry into acidic endosomes (11, 29, 30).

FIG. 1.

EdTx elicits an increase in I_sc from T84 cells. (A) Time courses of CT- and EdTx-induced I_sc. All toxins were applied at time zero. EdTx (10 μg of PA per ml and 0.1 μg of EF per ml) was added to either the apical or basolateral chamber. CT (20 nM) was added to the apical chamber. VIP was added to the control monolayers at the end of the experiment to confirm viability. VIP is a cAMP agonist that is used to show that the cells can respond to increases in cAMP. (B) Dose dependency of EdTx action. The indicated concentrations of PA were added to the basolateral chamber of T84 monolayers in the presence of 0.1 μg of EF per ml. Peak I_sc values (means ± standard deviations at steady state ≈97 min after EdTx application, n = 2) are plotted and fit to Michaelis-Menten kinetics. Apparent ED₅₀, 3 μg/ml.

TABLE 1.

Response of T84 cells to anthrax EdTx^a

Toxin	Cell surface applied and treatment	Lag (min)	dIsc/dt (μA/cm2/min)	Peak Isc (μA/cm2)
WT PA
n = 28	basolateral	17 ± 7	0.7 ± 0.5	28 ± 17
n = 8	apical	>80	ND	1.2 ± 0.2
WT PA (trypsin treated)
n = 7	basolateral	20 ± 4	0.3 ± 0.1	17 ± 6b
n = 9	apical	>80	ND	1.4 ± 0.4
PASSSR, n = 8	basolateral	>80	ND	1.5 ± 0.5
PA--D	basolateral	>80	ND	1.9 ± 0.6
WT PA, n = 8	basolateral;  Baf A1	>80	ND	1.7 ± 0.7

^aValues are reported as means ± standard deviations. ND, not detectable. 

^bTrypsin-treated EdTx induced a slightly lower I_sc than did whole EdTx (30% reduction), as has been observed previously (33). 

The effects of basolaterally added EdTx were dose dependent. When EF was held constant at 0.1 μg/ml, maximal currents of approximately 45 μA/cm² were seen at 30 μg of PA per ml and the 50% effective dose was 3 μg of PA per ml (Fig. 1B). Concentrations of 10 μg of PA per ml and 0.1 μg of EF per ml were used for all subsequent studies. Under these conditions, we consistently observed a lag of ≈17 min between the addition of EdTx and increases in I_sc (Fig. 1A; Table 1).

By several criteria, the I_sc induced by EdTx was dependent on a cAMP-induced Cl⁻ current. Substitution of membrane impermeant gluconate for Cl⁻ in both the apical and basolateral buffers abrogated the I_sc induced by EdTx, and the I_sc response could be restored by replenishing Cl⁻ (Fig. 2). Bumetanide (10 μM), a specific inhibitor of the Na⁺ K⁺ 2Cl⁻ uptake pathway (3, 8) reduced EdTx-induced I_sc by ∼70% (n = 3; data not shown). Ba²⁺ (3 mM) and charybdotoxin (100 nM) were used to distinguish between cAMP- and Ca²⁺-induced secretory responses (15, 24, 36). Pretreatment of cells with Ba²⁺, which inhibits cAMP-dependent K⁺ channels, inhibited Cl⁻ secretion induced by EdTx, whereas charybdotoxin, which inhibits Ca²⁺-dependent K⁺ channels, had no effect (Fig. 3). Finally, we also showed that EdTx and the [Ca²⁺]-dependent agonist carbachol (data not shown) (5, 6, 27) acted synergistically, consistent with these agonists activating different pathways.

FIG. 2.

The source of the EdTx induced current is Cl⁻ transport. PA (10 μg/ml) and EF (0.1 μg/ml) were added to the basolateral chamber of T84 cells in either HBSS or a gluconate buffer which lacked Cl⁻ (see Materials and Methods). For this and the rest of the time courses, time zero was the time at which toxin was added to the cells. Control monolayers not exposed to EdTx were incubated in gluconate buffer. Cl⁻ was added back to the monolayers in gluconate buffer at 48 min (I_sc) (means ± standard deviations n = 3). VIP was added to control monolayers at the end of the experiment to confirm the viability of the monolayers.

FIG. 3.

EdTx-induced I_sc depends on cyclic nucleotide, but not Ca²⁺, as a second messenger. Time courses of I_sc induced by 10 μg of PA per ml and 0.1 μg of EF per ml added basolaterally to T84 monolayers pretreated for 30 min in either 3 mM barium or 100 nM charybdotoxin or not pretreated with an inhibitor. Control monolayers were incubated in HBSS alone.

To probe the basis of the polarity of EdTx action on T84 cells, we examined the hypothesis that these cells may not be able to affect proteolytic activation of PA applied to the apical surface. PA that had been activated with trypsin in vitro (0.1 μg of trypsin per ml for 5 min on ice) was inactive when it was applied apically, although it remained active when it was applied basolaterally (Table 1). This result implies that the inactivity of apically applied EdTx cannot be attributed solely to the absence of an activating protease.

The possibility that receptors for EdTx might be absent on the apical membranes of T84 cells was assessed by measuring binding of PA. PA (3 μg/ml) was applied to T84 cells apically or basolaterally. After incubation for 60 min at 4°C, cells were washed to remove unbound protein, and the bound proteins were extracted. Bound PA was quantified by SDS-PAGE and Western blot analysis (Fig. 4A). The quantity of PA bound to apical surfaces was always at least 30-fold less than that bound to basolateral surfaces (Fig. 4B), and in some experiments no PA could be detected on the apical surface. Also, no cell-associated PA was detectable after incubation with the apically applied protein at 37°C for periods up to 60 min, indicating that amounts of the protein taken up by fluid-phase endocytosis are below the detection limits of our assay (Fig. 5).

FIG. 4.

PA binds to the basolateral but not the apical membrane of T84 cells. PA (10 μg/ml) was bound to T84 monolayers at 4°C for 1 h, and unbound material was washed away. Cell extracts were analyzed for PA by Western blotting. The anti-PA antibody recognized a nonspecific cellular protein with a size of approximately 55 to 60 kDa (indicated by *), which proved useful as an internal control for protein loading. The molecular size markers are indicated on the outside of each blot and represent molecular sizes of 104, 80, 47, and 33.5 kDa. (A) Western blot of extracts from T84 cells that had been treated with PA apically or basolaterally at 4°C or without treatment. PA recovered from cells treated basolaterally was cleaved to its PA63 form, while the PA standard (not exposed to T84 cell surfaces) was not. (B) Dilutions of protein extracts from T84 cells exposed to PA. Extracts from cells treated basolaterally with PA were diluted in protein sample buffer, as indicated above each lane and Western blotted for PA. Undiluted extracts from cells treated apically with PA were run on the same blot. At least 30-fold-less PA was recovered from cells treated apically with PA than from those treated basolaterally.

FIG. 5.

Assembly of PA in T84 cells. PA was bound to T84 cells at 4°C and then incubated at 37°C for the indicated times. A Western blot of cell extracts is shown. Three forms of PA could be detected in the proteins recovered from the basolateral membrane of cells incubated at 37°C: full-length PA (PA83), nicked PA (PA63), and a high-molecular-weight band representing the PA oligomer (indicated by an arrow). No toxin was recovered from the apical membrane.

PA63, as well as whole PA (83 kDa), was recovered from cells treated basolaterally with PA, indicating that, as in CHO-K1 and L6 cells (13, 16), proteases endogenous to T84 cells can activate PA at the cell surface. We also observed the time-dependent formation of a high-molecular-weight, SDS-resistant PA63 oligomer (Fig. 5, arrow) after incubation of PA (at 37°C) with basolateral but not apical cell surfaces. The oligomer was not observed after incubations at 4°C. These results are consistent with the notion that formation of the SDS-resistant PA63 oligomer requires endocytosis and correlates with toxin activity (31).

Figure 6 shows that, whereas complexes of CT and its receptor ganglioside GM1 fractionate in sucrose gradients with detergent insoluble caveola-like membrane domains (fractions 11 and 12), the PA-receptor complex fractionates with all other detergent-soluble proteins at the bottom of the gradient. These results provide evidence that the receptor for PA differs fundamentally from that for CT, although we cannot formally rule out the alternative possibility that PA may dissociate from its receptor by treatment with 1% Triton X-100 at 4°C.

FIG. 6.

Association of the CT-GM1 and PA-receptor complexes with caveola-like membrane domains. Sucrose gradient of extracts from T84 cells were exposed apically to CT B subunits (CTB, 20 nM) or basolaterally to PA (PA83, 20 nM) and analyzed for CT B subunits or PA by Western blotting. Only fractions 10 to 20, which correspond to a linear sucrose gradient from 15 to 32% sucrose (top to bottom), are shown. Less than 1% of the total cellular protein floated into the gradient (fractions 11 and 12, representing 18.2 to 22.8% sucrose). This fraction contains caveolin 1 (40).

DISCUSSION

The results of these studies define the human intestinal T84 cell as a model to study the cell biology of anthrax EdTx action on eukaryotic cells. This model system complements other methods for studying EdTx action; the sensitivity and temporal resolution are high, allowing the kinetics of toxin action to be measured precisely. Additionally, the same monolayer of cells can be monitored repeatedly for changes in response. As the SDS-resistant form of the PA heptamer can be monitored and correlated temporally with EdTx action, T84 cells represent a sensitive model to examine the structure and function of anthrax toxin activation, assembly, and trafficking.

In T84 cells, EdTx elicits a response only when it is applied to basolateral cell surfaces, and our data imply that this is a result of the polarized localization of the toxin receptor. As formal binding isotherms have not yet been performed with T84 cells, it is possible that this association is nonspecific. Nonetheless, we find evidence for receptor-mediated endocytosis restricted to the basolateral membrane, and available evidence from other cell systems suggests that the PA receptor is ubiquitous.

The protein and lipid components of the apical membrane are often unique to a specific cell phenotype. The basolateral membrane, on the other hand, exhibits many (if not all) housekeeping or structural proteins required for cell viability (28, 37). Thus, while clearly an oversimplification (as basolateral membranes can also harbor proteins and lipids of specialized functions), the fact that the receptor for PA is ubiquitously expressed and sorted strictly to basolateral membranes of polarized cells suggests that the receptor serves a basic function common to all cell types.

The mechanism of cell entry for EdTx differs from that for CT, as evidenced both by the kinetics of toxin action and by the characteristics of PA receptors, which do not fractionate with caveola-like membranes in T84 cells. To begin, EdTx binds a basolateral receptor, whereas CT works from the apical membrane. In T84 and Caco2 cells, CT partitions into caveola-like membrane domains for endocytosis and trafficking within the cell (34, 41). As EdTx does not fractionate with caveolae, it must enter T84 cells via clathrin-coated pits or via another non-clathrin-dependent transport system. The faster kinetics of EdTx action (Fig. 1) support a more direct trafficking mechanism for EdTx than for CT. The long lag period for CT activity likely represents its requirement for trafficking through the endoplasmic reticulum (20, 26), while EdTx is thought to exhibit a simple trafficking pathway into endosomes. Our results are consistent with results obtained in CHO-K1 cells, where a lag of approximately 10 min was measured before EdTx activity was detectable (14). Taken together, these studies show that EdTx and CT co-opt different subcellular systems for entry.

The physical characteristics displayed by EdTx receptors in T84 cells, which include basolateral polarity and exclusion from caveolae, may prove useful in receptor purification and identification. These results imply that the receptor for PA may be a protein that performs a generalized cellular function.