The lethal and edema factors of anthrax toxin bind only to oligomeric forms of the protective antigen

Abstract

The three proteins that comprise anthrax toxin, edema factor (EF), lethal factor (LF), and protective antigen (PA), assemble at the mammalian cell surface into toxic complexes. After binding to its receptor, PA is proteolytically activated, yielding a carboxyl-terminal 63-kDa fragment (PA₆₃) that coordinates assembly of the complexes, promotes their endocytosis, and translocates EF and LF to the cytosol. PA₆₃ spontaneously oligomerizes to form symmetric ring-shaped heptamers that are capable of binding three molecules of EF and/or LF as competing ligands. To determine whether binding of these ligands depends on oligomerization of PA₆₃, we prepared two oligomerization-deficient forms of this protein, each mutated on a different PA₆₃–PA₆₃ contact face. In solution or when bound to receptors on Chinese hamster ovary K1 cells, neither mutant alone bound ligand, but a mixture of them did. After the two mutants were proteolytically activated and mixed with ligand in solution, a ternary complex was isolated containing one molecule of each protein. Thus EF and LF bind stably only to PA₆₃ dimers or higher order oligomers. These findings are relevant to the kinetics and pathways of assembly of anthrax toxin complexes.

Anthrax toxin comprises three nontoxic monomeric proteins that assemble at the mammalian cell surface to form toxic complexes. Edema factor (EF) and lethal factor (LF) are enzymes that are delivered to the cytosol of mammalian cells by the third protein, protective antigen (PA). EF is an adenylyl cyclase that elicits edema when coinjected with PA into animals and likely serves to impair the immune response to infection (1–3). LF is a zinc protease that cleaves certain mitogen-activated protein kinase kinases (4–6). In combination with PA, LF causes death of animals by a still poorly understood sequence of events (7).

Assembly of the toxic complexes begins when PA binds to a cellular receptor, a widely expressed type 1 membrane protein termed anthrax toxin receptor (8). Receptor-bound PA is cleaved by a member of the furin class of proteases, causing removal of an amino-terminal 20-kDa segment of the protein (PA₂₀), and leaving the carboxyl-terminal 63-kDa fragment (PA₆₃) bound to anthrax toxin receptor (9). Unlike native PA, PA₆₃ oligomerizes to form a ring-shaped heptamer (10). The heptamer binds up to three copies of EF and/or LF competitively and with high affinity (K_d ≈ 1 nM; refs. 11 and 12). Whereas native PA persists on the cell surface, the heptamer is endocytosed (13), presumably because oligomerization aggregates anthrax toxin receptor. The endocytosed toxic complexes are trafficked to endosomal compartments where the low pH causes the PA₆₃ heptamer to insert into the membrane and form a water-filled channel (14, 15). Delivery of EF and LF to the cytosol occurs in concert with channel formation and may involve passage of these proteins through the channel (16).

Comparison of the crystal structures of PA and the PA₆₃ heptamer showed that removal of the PA₂₀ moiety relieves a steric constraint to oligomerization (17). Association of PA₆₃ monomers occurs spontaneously and is mediated by interactions of domains 1′ and 2 of one monomer with domains 3 and 2, respectively, of an adjacent monomer. The hypothesis that EF and LF bind to the surface of domain 1′ exposed by removal of PA₂₀ is supported by identification of mutations within domain 1′ that affect ligand binding (18). No crystallographic structure of complexes of PA₆₃ with EF and/or LF has been solved, however, and the specific contacts involved in this interaction have remained undefined.

In an earlier study we discovered that an oligomerization-deficient mutant of PA (PA–D512A) did not bind as much ligand as wild-type PA, suggesting that ligand binding might depend on PA₆₃ oligomerization (19). To investigate this question we prepared two complementary PA mutants with lesions on different PA₆₃–PA₆₃ contact faces. The presence of a mutated oligomerization face would be expected to block oligomerization of either mutant protein alone. When the proteins are mixed, however, their complementary oligomerization-competent faces should allow mixed dimers to form. Using these mutant proteins singly and in combination, we obtained strong evidence that oligomerization of PA₆₃ is required for ligand binding.

Materials and Methods

Plasmid Construction.

The plasmid pET22b-PA contains the entire PA gene except for the portion that encodes the signal sequence (20). QuikChange mutagenesis was performed according to manufacturer instructions (Stratagene) to introduce site-directed mutations. Oligonucleotides were obtained from Integrated DNA Technologies (Coralville, IA). The oligonucleotides GGTTGATGTCAAAAATGAAAGAACTTTTCTTTCACC, CAATTTTGAAAATGGAGCAGTGGACGTGGATACAGGCTCG, GGTTAATCCTAGTAAACCATTAGAAACG, and their complements were used to introduce the mutations K199E, R468A/R470D, and D512K, respectively.

Preparation of Proteins.

PA, PA₆₃, and LF_N, the amino-terminal domain of LF, were purified from Escherichia coli as described (16, 21). Protein concentrations were determined by using Bradford protein assay reagent (Bio-Rad).

Cell-Surface Binding Assay.

Binding assays were performed essentially as described (16). Briefly, Chinese hamster ovary K1 cells were incubated on ice with 2 × 10⁻⁸ M trypsin-nicked PA for 2 h. The cells were washed twice with PBS and then incubated on ice with ³⁵S-labeled LF_N for 2 h. The cells were washed twice with PBS, and radioactive content was determined by scintillation counting.

Native Gel Electrophoresis.

Trypsin-nicked PA (2 μg) and LF_N (3 μg) were mixed, as indicated, in 20 mM Tris⋅HCl (pH 8.0)/150 mM NaCl (in a volume of 8 μl) and incubated for ≈10 min at room temperature. Two microliters of loading dye (bromophenol blue in 50% glycerol) were added to the mixtures, which then were loaded onto 4–20% acrylamide, Tris-glycine gels (BioWhittaker). The running buffer was 25 mM Tris-base/192 mM glycine. The gels were stained with 0.05% Coomassie blue R-250/50% methanol/10% acetic acid and then destained with 10% acetic acid/10% methanol.

Multiangle Laser Light Scattering.

Light scattering was performed as described (11). Trypsin-nicked PA (300 μg from each oligomerization-deficient mutant) was incubated for 30 min at room temperature in the absence or presence of LF_N (1 mg). Samples (140 μl) were loaded into a 100-μl loop and injected onto a Shodex KW-803 column equilibrated with 20 mM Tris⋅HCl (pH 8.2)/200 mM NaCl at a flow rate of 0.5 ml/min. The column was connected to a DAWN EOS 18-angle light-scattering detector (Wyatt Technology, Santa Barbara, CA) and an OPTILAB DSP interferometric refractometer (Wyatt Technology). Molecular mass calculations were performed by ASTRA software on normalized voltages from detectors 8–18 by using a dn/dc (refractive index increment) value of 0.185 ml/g. Normalization of the detectors was performed with the isotropic scatterer, BSA, to correct for slight differences in electronic gain among the detectors.

Results

Two Complementary Oligomerization-Deficient PA Mutants.

In a prior study we identified amino acid 512 as being important for oligomerization of PA₆₃ (19). Mutating this residue from aspartic acid to alanine diminished oligomerization of PA₆₃, and unexpectedly also impaired the protein's ability to bind LF_N, as a sample ligand. To assess whether ligand binding depends on oligomerization of PA₆₃, we made two other oligomerization-deficient PA mutants. For the first, we replaced D512 with lysine to generate a more severe oligomerization defect on the basis of the fact that D512 is on a loop of domain 3 that inserts into a positively charged cleft in domain 1′ of a neighboring subunit of the heptamer. For the second mutant, we placed mutations on the complementary oligomerization face of PA. We found empirically that multiple mutations on this face were needed to create a strong oligomerization defect and ultimately chose the triple mutant PA–K199E/R468A/R470D for further study.

Ligand Binds Only to Mixtures of the Two Mutants.

PA–D512K and PA–K199E/R468A/R470D were activated with trypsin and incubated singly or together with Chinese hamster ovary K1 cells at 4°C, a temperature at which endocytosis does not occur. After unbound PA was washed away, the cells were exposed to [³⁵S]LF_N, the amino-terminal PA₆₃-binding domain of LF, and bound LF_N was quantified by scintillation counting. Cells incubated with either PA–K199E/R468A/R470D or PA–D512K alone did not bind LF_N (Fig. 1). In contrast, cells incubated with an equimolar mixture of PA–D512K and PA–K199E/R468A/R470D bound almost as much LF_N as cells incubated with wild-type PA. These results support the notion that a high-affinity ligand site was generated when the two mutant forms of PA₆₃ interacted (Fig. 1).

Figure 1.

Neither of two oligomerization-deficient forms of PA₆₃ on cells stably binds LF_N, but a mixture of the two does. Trypsin-nicked PA (2 × 10⁻⁸ M) wild type (WT), D512K, K199E/R468A/R470D, or a mixture of the two mutant forms (1 × 10⁻⁸ M of each) was incubated with Chinese hamster ovary K1 cells at 4°C for 2 h. The cells were washed twice with PBS and incubated with [³⁵S]LF_N at 4°C for 2 h. The cells then were washed with PBS, and radioactive content was measured by scintillation counting. The units on the ordinate are percentages of control (wild-type PA.) Nonspecific binding of LF_N to cells (less than 10%) was subtracted from the experimental measurements to yield specific binding. The error bars represent SE of the mean.

A Ternary Complex Formed in Solution.

The oligomerization state of PA₆₃ can be assessed by gel electrophoresis under nondenaturing conditions. Trypsin cleaves at the furin site, and in trypsin-nicked PA, the PA₂₀ and PA₆₃ fragments tend to remain associated, thereby blocking oligomerization of PA₆₃. The nicked protein migrates on native gels with high mobility relative to that of liganded PA₆₃ heptamer, formed by incubating nicked PA with LF_N (Fig. 2, compare lanes 2 and 3). The electrophoretic mobilities of the nicked PA mutants differ from each other and from that of nicked wild-type PA because of differences in charge. Consistent with their oligomerization defects, neither of the mutants produced a low-mobility species when incubated with LF_N (lanes 4–7). However, when a mixture of the two mutants was combined with LF_N, a new band was observed with a slightly lower mobility than that of the nicked PA–D512K mutant (compare lanes 8 and 9). We inferred that this band represented a ternary complex generated by interaction of LF_N with a PA₆₃ dimer formed from the two mutants.

Figure 2.

Native gel electrophoresis of oligomerization-deficient PA mutants in the presence and absence of ligand. Trypsin-nicked PA (nPA) of wild type (WT) or mutant (2 μg) was incubated in the presence or absence of LF_N (3 μg), and the mixture was electrophoresed on a 4–20% acrylamide Tris-glycine gel. The gels were stained with Coomassie blue. A band in lane 9 corresponding to a putative heterotrimer containing a molecule of LF_N and a molecule of PA₆₃ from each of the two mutant PAs is indicated by the arrow.

To confirm the identity of this putative ternary complex, we nicked each mutant PA, mixed the two, divided the mixture in half, and added excess LF_N to one half (8:1 molar ratio of LF_N/PA₆₃ monomer). Each sample then was chromatographed on a Shodex KW-803 size-exclusion column, and the effluent was passed sequentially through a multiangle laser light-scattering instrument and an interferometric refractometer. The latter served as a continuous monitor of protein concentration. Combining the data from these two instruments permitted calculation of the particle masses of protein peaks in the column effluent. As shown in Fig. 3, the sample containing LF_N gave two protein peaks with measured molecular masses of 163 ± 1.9 and 36.1 ± 0.4 kDa. The first value corresponds closely to the predicted mass (160 kDa) of a PA₆₃ dimer complexed with a single molecule of LF_N, and the second corresponds well with the predicted mass of the unbound LF_N (34 kDa). The presence of a small amount of PA₂₀ at the trailing edge of the LF_N peak probably accounts for the negative slope of the molecular mass values across this peak. These results support our interpretation of the additional band seen in lane 9 of Fig. 2 as representing a ternary complex as described.

Figure 3.

Use of multiangle laser light scattering to identify a ternary complex of LF_N and the two oligomerization-deficient forms of PA₆₃. A mixture of LF_N (1 mg) with trypsin-nicked PA–D512K (300 μg) and PA–K199E/R468A/R470D (300 μg) was chromatographed over a Shodex KW-803 gel filtration column connected to a light-scattering instrument (Lower) and an interferometric refractometer (Upper). The values of molecular mass determined in volume increments across each peak are shown (arrows). The light-scattering peak at ≈7 ml elution volume represents a very small amount of contaminating high molecular mass material (note the absence of interferometric refractometer signal). The negative refractive index values at 14–15 ml elution volume represent differences between the composition of the sample and elution buffers.

When the two oligomerization-deficient mutants were mixed in the absence of LF_N and chromatographed, only a single peak was seen, with a measured mass of 83.4 ± 2.1 kDa, corresponding to PA₆₃ complexed with PA₂₀ (data not shown), which correlates with the fact that we did not observe a band by PAGE, corresponding to a heterodimer of the two mutant forms of PA₆₃ (Fig. 2). This result suggests that PA₆₃ dimers are inherently unstable and form only transiently during the course of heptamerization. It is noteworthy in this regard that no subheptameric intermediates are seen in chromatographic fractionation of PA₆₃. Thus only the fully formed heptamer of PA₆₃, in which every subunit is bound to two neighboring subunits, is apparently stable in the absence of ligand.

Discussion

Our results indicate that the enzymatic moieties of anthrax toxin bind stably only to dimeric or oligomeric forms of PA₆₃ (Fig. 4). Dependence of ligand binding on oligomerization would be expected if the ligand site contained contributions from two adjacent monomers or if subunit–subunit interactions stabilized a binding-competent conformation. In an accompanying paper, evidence is presented that the binding site of LF_N spans the interface between PA₆₃ subunits of the dimer (18), which substantiates the first of these possibilities.

Figure 4.

Complex formation by wild type and oligomerization-deficient mutants of PA₆₃. Only one potential pathway of formation of the ternary complex is illustrated.

That LF_N does not associate stably with monomeric PA₆₃ is at variance with the conclusion of an earlier study reporting that LF bound to PA that had been adsorbed to a plastic surface and then nicked with trypsin (22). The adsorbed PA may have oligomerized in this experiment by diffusion on the plastic surface or via other unrecognized pathways. Alternatively, the observed binding may have reflected combined interactions of LF with both monomeric PA₆₃ and the plastic surface.

Our results have implications regarding the pathways by which anthrax toxin complexes assemble from their component parts. Earlier results have demonstrated that PA₆₃ can assemble to the homoheptamer in the absence of ligand, and that the heptamer can bind up to three molecules of ligand (10, 11). Our results imply that ligand can bind to and stabilize subheptameric oligomers of PA₆₃, creating stable heterooligomeric intermediates. The stability of such intermediates would be expected to accelerate formation of toxic complexes containing heptameric PA₆₃.

Fig. 4 illustrates one pathway by which liganded ternary complex containing dimeric PA₆₃ could form. However, two such pathways are possible: (i) ligand could bind to a transient intermediate of two PA₆₃ molecules, or (ii) ligand could bind to monomeric PA₆₃ followed by interaction of this complex with a second molecule of PA₆₃. The kinetics of these pathways depend on the relative stabilities of the PA₆₃–PA₆₃ and PA₆₃–ligand intermediates, which are unknown. Values of K_d and the on- and off-rate constants of the interactions of the enzymatic moieties for oligomeric PA₆₃ have been measured (12), but other parameters affecting the assembly process have not. Regardless of the pathway by which it formed, the liganded dimer of the mixed oligomerization-deficient mutant PAs is sufficiently stable in solution to be identified as a peak from a size-exclusion column and a band on PAGE.

Abbreviations

EF

edema factor

LF

lethal factor

PA

protective antigen

PA_n

n-kDa fragment of protective antigen

LF_N

amino-terminal domain of LF