Evidence that translocation of anthrax toxin's lethal factor is initiated by entry of its N terminus into the protective antigen channel

Abstract

Entry of the enzymatic components of anthrax toxin [lethal factor (LF) and edema factor] into the cytosol of mammalian cells depends on the ability of the activated protective antigen (PA₆₃) component to form a channel (pore) in the membrane of an acidic intracellular compartment. To investigate the mechanism of translocation, we characterized N-terminally truncated forms of the PA₆₃-binding domain of LF (LF_N). Deleting 27 or 36 residues strongly inhibited acid-triggered translocation of LF_N across the plasma membrane of CHO-K1 cells and ablated the protein's ability to block PA₆₃ channels in planar lipid bilayers at a small positive voltage (+20 mV). Fusing a H₆-tag to the N terminus of the truncated proteins restored both translocation and channel-blocking activities. At +20 mV, N-terminal H₆ and biotin tags were accessible to Ni²⁺ and streptavidin, respectively, added to the trans compartment of a planar bilayer. On the basis of these findings, we propose that the N terminus of PA₆₃-bound LF or edema factor enters the PA₆₃-channel under the influence of acidic pH and a positive transmembrane potential and initiates translocation in an N- to C-terminal direction.

Many toxins produced by pathogenic bacteria are specialized enzymes that have evolved ways to gain access to the cytosolic compartment of mammalian cells. There, they catalyze reactions detrimental to the host. All such toxins must penetrate a cytosol-contiguous membrane, and some members of this class contain a domain, or a separate polypeptide, that forms an aqueous transmembrane channel (pore) to mediate translocation of the enzymatic part(s). How such channels are formed and how they mediate translocation of a catalytic moiety across the membrane are topics of current interest.

Anthrax toxin is an ensemble of three proteins: Two are intracellularly acting enzymes, and the third transports both of the enzymatic moieties to the cytosol (1, 2). The enzymatic components are an adenylate cyclase termed “edema factor” (EF, 89 kDa) and a Zn²⁺-protease termed “lethal factor” (LF, 90 kDa). The transport component is a receptor-binding, poreforming protein termed “protective antigen” (PA, 83 kDa). PA, EF, and LF, which are secreted from the anthrax bacillus as unassociated nontoxic monomers, combine at the surface of receptor-bearing cells to form a series of toxic cell-bound complexes (3). Assembly begins when PA binds to a cellular receptor and is activated by a member of the furin family of proteases (4). This action removes a 20-kDa fragment (PA₂₀) from the N terminus, leaving a 63-kDa fragment (PA₆₃) bound to the receptor. Once freed from steric constraints imposed by PA₂₀, PA₆₃ spontaneously oligomerizes, generating a ring-shaped heptamer, (PA₆₃)₇ (also called the prepore) (5). EF and LF bind to the heptamer, competitively and with nanomolar affinities, by means of a homologous N-terminal domain (EF_N or LF_N, respectively) (6). The footprint of this domain is such that a maximum of three molecules of EF or LF, or combinations thereof, can bind simultaneously per (PA₆₃)₇ (7). The assembled complexes are endocytosed and trafficked to an acidic compartment, believed to be the endosome. There, the acidic pH induces a conformational rearrangement in (PA₆₃)₇, allowing it to form an aqueous transmembrane channel that mediates translocation of EF and LF to the cytosol (8).

Electrophysiological data and crystallographic evidence support a model of channel formation in which the 2β₂–2β₃ loops of the seven monomers of (PA₆₃)₇ combine to form a 14-strand transmembrane β-barrel (9, 10). The limiting diameter of the PA₆₃ channel has been estimated from studies with tetraalkyl ammonium ions to be ≈12 Å, implying that a protein would need to unfold partially or fully to pass through (11, 12). Consistent with this concept, translocation of LF_N-DTA, a fusion protein containing LF_N linked to the N terminus of the catalytic A-chain of diphtheria toxin (DTA), was blocked either by introducing an artificial disulfide into the DTA moiety or by liganding this moiety with adenine or NAD⁺ (13). If unfolding of EF and LF is necessary for them to cross the membrane, they may thread through the channel vectorally, with the N or C terminus first. Here, we report results suggesting that a flexible 27-residue stretch of the N terminus of LF_N enters the (PA₆₃)₇ channel and initiates PA-dependent translocation across the endosomal membrane. These and ancillary findings suggest a model in which the N terminus of LF or EF initiates translocation in an N- to C-terminal direction. Further, our results suggest a role for transmembrane potential in the translocation process.

Materials and Methods

Preparation of LF_N Constructs. The plasmid pET15b-LF_N containing the N-terminal 1–263 residues of LF inserted between NdeI and BamHI sites was used for all cloning steps. LF_N constructs were made either with the N-terminal H₆-tag, in which the protein was inserted between the NdeI and BamHI sites, or without the N-terminal H₆-tag, in which it was inserted between the NcoI and BamHI sites of pET15b. We used 5′-CTAGGATCCTTACCCGTTGATCTTTAAGTTCTTCC-3′ to introduce a BamHI site and act as the reverse primer. The forward primers, designed to permit appropriate truncations and also introduce either an NcoI or NdeI site, were used in the amplification of non-H₆-tagged LF_N constructs or the H₆-tagged H₆-LF_N constructs. Constructs of H₆-LF_N, Δ27, Δ36, and Δ39 were kindly provided by D. Borden Lacy (Harvard Medical School). LF_N-DTA containing the A1C mutation was kindly provided by Steve Juris (Harvard Medical School).

Oligonucleotide 5′-CAT GGA TCC TTA GCC GTG ATG ATG ATG ATG ATG CGC CCG TTG ATC-3′ was used as a reverse primer to fuse a H₆-tag to the C terminus of LF_N and to introduce a BamHI site. Primers for LF_N and LF_N Δ27 were used as forward primers in the construction of LF_N-H₆ and LF_N Δ27-H₆.

Sequences were amplified from the pET15b-H₆-LF_N template by PCR. The PCR products were purified and double digested with either NcoI/BamHI or NdeI/BamHI. The digested fragments and pET15b vector were gel-purified, ligated, and transformed into Escherichia coli XL1-Blue by electroporation. Transformants were screened by digestion and verified by sequencing.

Constructs of H₆-LF_N Δ22 and H₆-LF_N Δ23 were made by using the QuikChange method of site-directed mutagenesis (Stratagene) using H₆-LF_N Δ21 and Δ24, respectively, as template. The PCR products were transformed into E. coli XL1-Blue and sequenced.

Protein Preparation. Purification of wild-type PA from BL21(DE3) and formation of PA₆₃ heptamer were performed as described in ref. 14. LF_N constructs that contained a H₆-tag at either their N or C terminus were expressed and purified according to the pET system manual from Novagen. Their N-terminal H₆-tag, which was removed by thrombin cleavage, consisted of the following 21 residues: Met–Gly–Ser–Ser–His–His–His–His–His–His–Ser–Ser–Gly–Leu–Val–Pro–Arg–Gly–Ser–His–Met. For conductance experiments, we used purified constructs containing either this 21-residue sequence or the product obtained by removal of the first 17 of these residues by thrombin treatment (hence leaving the additional Gly–Ser–His–Met sequence added to the N terminus). For cell-based assays, the sequences were as shown in Fig. 1, but with an additional Met residue at the N terminus. Protein concentrations were determined by using the Bio-Rad protein assay reagent.

Fig. 1.

N-terminal sequence of LF_N and deletion mutants. Shown are the sequences of the first 40 residues of wild-type (WT) LF_N and that of each of the deletion mutants tested. Acidic residues are underlined; basic residues are in boldface type. Additional sequences present at the N terminus of various constructs, including the 21-residue H₆-containing tag, are described in Materials and Methods.

Cell-Surface Binding and Translocation Assay. PA binding to the cell surface and translocation of ³⁵S-LF_N were assayed as described in refs. 13 and 15. Briefly, CHO–K1 cells in a 24-well plate were chilled on ice, washed, and incubated on ice for 2 h with trypsin-activated PA (saturating concentrations). The cells then were washed and incubated for an additional 2 h on ice with wild-type or mutant LF_N that had been labeled with [³⁵S]methionine in an in vitro transcription/translation system (Promega). The cells then were washed with ice-cold PBS at pH 5.0 or 8.0, incubated at 37°C for 1 min, and either treated with Pronase to digest residual untranslocated ³⁵S at the cell surface or left untreated as controls. The cells then were lysed, and ³⁵S liberated into the lysis buffer was assayed. The percent translocation was defined as dpm protected from Pronase/dpm bound to cells × 100.

Biotinylation of LF_N and LF_N-DTA. pET15b-H₆-LF_N was mutated to introduce a Cys residue either at its N terminus (A1C) or close to the C terminus (N242C) by using the QuikChange mutagenesis system (Stratagene). The N-terminal H₆-tag was cleaved with thrombin before biotinylation. Biotinylation was performed basically according to Jakes et al. (16), and the extent of biotinylation was assessed by an SDS/PAGE streptavidin-binding assay as described in Qiu et al. (17). To ensure that all of the effect observed on planar lipid bilayers came from biotinylated LF_N, the biotinylation products were purified through a monomeric avidin column (UltraLink, Pierce). Proteins purified on this column showed no evidence of residual nonbiotinylated LF_N in the gel assay.

Planar Lipid Bilayer Experiments. Bilayers were formed at room temperature by using the brush technique of Mueller et al. (18) across a 200-μm-diameter hole in a Delrin cup separating 1-ml cis and trans compartments. Symmetric solutions of 100 mM KCl, 25 mM succinic acid, and 1 mM EDTA (pH 5.5) were used. EDTA was omitted in experiments in which Ni²⁺ was added. Bridges containing 3 M KCl in 3% agar connected solutions to Ag/AgCl electrodes in 3 M KCl baths. The membrane-forming solution was 30 mg/ml diphytanoylphosphatidyl choline in n-decane. PA₆₃ heptamer (0.5–5 ng) was added to the cis compartment, which was held at +20 mV with respect to the trans compartment. The current was amplified through a BC-525C integrating bilayer clamp amplifier (Warner Instruments, Hamden, CT), filtered at a frequency of 0.1 kHz by a low-pass eight-pole Bessel filter, and computer-displayed through an analog/digital converter (Instrutech, Mineola, NY) with axograph 4.0 (Axon Instruments). LF_N was added to the cis compartment after (PA₆₃)₇ channel formation stabilized. Ni²⁺ or streptavidin was added to either the cis or the trans compartment at a concentration of 500 μM or 6 μg/ml, respectively, at indicated time points. Each compartment was stirred continuously throughout the experiment with a small stir bar.

Results

Cell-Surface Binding and Translocation of Truncated Forms of LF_N. To explore the manner in which EF and LF undergo PA₆₃-dependent translocation, we deleted various numbers of residues from the N terminus of LF_N and measured the PA₆₃-dependent binding and translocation activities of the resulting constructs. The constructs were labeled with [³⁵S]methionine in an in vitro transcription/translation system, and protection of cell-bound label against digestion with Pronase after an acid pulse was used as a measure of translocation. Earlier studies showed that the full-length LF_N is protected under the conditions of the assay (13). As shown in Fig. 2A, deletion of the first 27, 36, or 39 residues strongly inhibited acid-triggered translocation of the protein across the plasma membrane of CHO-K1 cells. Only the 39-residue deletion (Δ39) inhibited binding of LF_N to PA₆₃ on cells (Fig. 2B) or, slightly, in solution (Fig. 2C). The solution studies were based on the ability of LF_N to promote conversion of trypsin-activated PA to heptameric PA₆₃. Together, these experiments showed that N-terminal truncations of 27 or 36 residues strongly impaired translocation without significantly affecting binding to PA₆₃. These results suggested a pivotal role for this region of the protein in translocation.

Fig. 2.

The effects of various N-terminal deletions on the PA₆₃-binding and translocation activities of LF_N. H₆-tagged constructs (dark gray bars) and non-H₆-tagged constructs (light gray bars) were tested basically as described in ref. 13. (A) Translocation across the plasma membrane in response to low pH. (B) PA₆₃-dependent binding of LF_N to CHO-K1 cells. CHO-K1 cells were incubated with trypsin-nicked PA (2.4 × 10^–8 M) for 2 h on ice, and the cells then were washed and incubated with various ³⁵S-LF_N truncations for another 2 h. The cells were treated with buffer (pH 5.0) at 37°C for 1 min to trigger translocation and then with Pronase to digest ³⁵S-LF_N remaining at the cell surface. Cell-associated ³⁵S-LF_N was determined by scintillation counting as a measure of translocation. PA₆₃-dependent binding was measured in identical samples in which Pronase was omitted. The fraction translocated was defined as the amount of protected ³⁵S-LF_N divided by the amount bound. (Error bars, SEM.) The results shown in A and B are the average of three independent experiments. (C) LF_N-induced oligomerization of PA₆₃ in solution. Trypsin-activated PA (3 μg) was incubated with 1 μg of each LF_N truncation mutant at room temperature for 1 h, and the samples then were electrophoresed on a 4–12% polyacrylamide gel. Y236A-LF_N was included as a mutant form known to be deficient in PA binding.

Additional support for this hypothesis came from the finding that a H₆-tag fused to the N terminus of the Δ27, Δ36, and Δ39 constructs restored translocation efficiency to wild-type levels, or higher (Fig. 2 A and B). The H₆-tag dramatically enhanced translocation while slightly increasing the PA₆₃-dependent binding of truncated or wild-type forms of LF_N to CHO-K1 cells. These results suggested that the positively charged H₆-tag interacts with the cation-selective channel, strongly enhancing translocation efficiency.

Inhibition of PA₆₃-Mediated Ion Conductance in Planar Lipid Bilayers. Addition of EF, LF, or EF_N to the cis compartment of a planar lipid bilayer system (the compartment to which PA heptamer is added) causes a voltage-dependent reduction in ion conductance of the cation-selective channels formed by PA₆₃ (12). Also, LF or LF_N inhibits PA₆₃-dependent efflux of ⁸⁶Rb⁺ from preloaded CHO-K1 cells (19). Consistent with these results, we found that with the voltage held at +20 mV, adding full-length LF_N to the cis compartment caused a rapid 20- to 50-fold decrease of PA₆₃-mediated conductance (Fig. 3A).

Fig. 3.

Effects of N-terminal deletions on the ability of LF_N to inhibit PA₆₃-dependent conductance in planar bilayers. Purified (PA₆₃)₇ (2 ng) was added to the cis compartment, containing 1 ml of buffer (100 mM KCl/25 mM succinate/1 mM EDTA, pH 5.5). (A and B) The cis compartment was held at +20 mV, and after channel formation stabilized, we added full-length LF_N (A) or a truncated form, LF_N Δ27 (B), to the cis compartment to give a final concentration of 20 nM. The traces in A and B are representative of 12 or 11 different trials, respectively. (C) Various concentrations of full-length or truncated forms of LF_N were added to the cis compartment, and the percent current decrease was determined after 1 min. DTA, used as a negative control, did not interact with the channel. The curves in C are the average of three independent experiments, except for DTA, which was tested twice.

The effects of N-terminal truncations and H₆-tags on the inhibition of ion conductance correlated well with their effects on translocation. The Δ27 construct caused little or no inhibition of ion conductance, whereas Δ13 and Δ15 were as potent inhibitors of conductance as full-length LF_N (Fig. 3 B and C). Inhibitory activity tended to decrease as increasing numbers of residues were deleted, and constructs with deletions of 23 residues or more showed no inhibitory activity. A H₆-tag fused to the N terminus of Δ27 restored the ability of this mutated form of LF_N to block PA₆₃ channels (Fig. 4A). In fact, fusing a H₆-tag to the N terminus of any of the LF_N truncation constructs, including Δ18, Δ24, Δ27, Δ36, and Δ39, restored channel-blocking activity (data not shown). In contrast, a H₆-tag appended to the C terminus of either full-length LF_N or the Δ27 construct had no effect (Fig. 4).

Fig. 4.

Effect of H₆-tag on the channel-blocking activity of LF_N truncations. Conditions were as in Fig. 3. (A) After channel formation stabilized, LF_N Δ27 with a H₆-tag at either its N or C terminus was added to the cis compartment as indicated. The trace is representative of three independent experiments, all with identical results. (B) The percentage decrease in current 1 min after LF_N addition was plotted as a function of final LF_N concentration. The curves are averaged from three independent experiments.

Interaction of Modified Forms of LF_N with the PA₆₃ Channel. The results described above suggest that the N-terminal region of PA₆₃-bound LF_N can enter the channel, blocking ion conductance in planar bilayers and initiating translocation in vivo (Fig. 5). In support of this model, N-terminally biotinylated LF_N (A1C-biotin) blocked the PA₆₃ channel as effectively as wild-type LF_N, but preincubating it with streptavidin ablated blocking activity (Fig. 6A). Streptavidin did not inhibit binding of A1C-biotin to PA₆₃, as shown by the PA₆₃ oligomerization assay in solution (data not shown). Reduction of the disulfide bridge linking biotin to A1C released the streptavidin, allowing LF_N to regain blocking activity (Fig. 6A). LF_N biotinylated near the C terminus (N242C-biotin) retained channel-blocking activity in the presence or absence of streptavidin (Fig. 6B). A higher concentration of the derivatized LF_N was needed to block channels in the presence of streptavidin, however, suggesting steric interference by streptavidin of binding of LF_N to the PA₆₃ channel. Consistent with this hypothesis, the N242C-biotin:streptavidin complex bound PA₆₃ heptamer with a lower efficiency in solution (data not shown).

Fig. 5.

Illustration of the movement of N terminus of LF_N into the PA₆₃ channel under the influence of a transmembrane voltage of +20 mV.

Fig. 6.

Evidence that the N terminus of LF_N enters the channel formed by (PA₆₃)₇.(A and B) LF_N (0.6 μg) containing a biotin group disulfide-linked either to the N (A) or the C(B) terminus was incubated with 6 μg of streptavidin at room temperature for 10 min before being added to the cis compartment. The membrane-impermeant disulfide reducing agent Tris-carboxyethylphosphine (TCEP) (5 mM) was added as indicated. The voltage was held at +20 mV throughout the experiment. (C) Trans addition of Ni²⁺ slowed unblocking of N-terminal H₆-tagged LF_N. EDTA was omitted from the buffer. We added 500 μMNi²⁺ and 1 mM EDTA to the trans compartment at indicated time points. Although some of the channels became immediately unblocked at –20 mV, others took longer to unblock. This delay may be a consequence of the distribution of LF_N penetration of channels at a given voltage. (D) Use of biotin-LF_N-DTA to demonstrate penetration of the N terminus to the trans compartment. The buffer used was the same as in Fig. 3, except that 1 mM adenine was present. Eleven minutes before starting the recording, 2 ng of (PA₆₃)₇ was added to the cis compartment held at +20 mV. Then, biotin-LF_N-DTA (4 nM) was added to the same compartment. At the break, the cis compartment was perfused with 10 ml of buffer, and 10 μg of streptavidin was added to the trans compartment, as indicated. The voltage was shifted to –20 mV, and TCEP (5 mM) was added to the trans compartment, as indicated. The small response at –20 mV before TCEP addition may reflect a small amount of protein not binding streptavidin. The traces shown in A, B, and C were representative of three, four, and four independent trials, respectively, all of which were essentially identical. The trace in D is representative of six trials, all but one of which were successful.

The hypothesis that the N terminus of LF_N enters the PA₆₃ channel received additional support from an experiment showing that an N-terminal H₆ was accessible to Ni²⁺ added to the trans compartment. As shown in Fig. 6C, changing the voltage from +20 to –20 mV caused an immediate unblocking of channels. Because Ni²⁺ binds to the H₆-tag, trans Ni²⁺ should affect the rate at which the N terminus of LF_N returns to the cis side at negative voltages; that is, it should affect the rate of channel unblocking. This effect is precisely what we observed (Fig. 6C). When the voltage was stepped from +20 to –20 mV, the unblocking rate was slowed by Ni²⁺ added to the trans solution and was restored by the subsequent addition of EDTA to that solution. In contrast, trans Ni²⁺ had no effect on the unblocking rate if the H₆-tag was attached to the C terminus of LF_N instead of the N terminus (data not shown).

More definitive evidence that LF_N enters the PA₆₃ channel N terminus first came from a demonstration that the N terminus emerges into the trans solution while the C terminus is held in the cis compartment. For this experiment, we used LF_N-DTA containing biotin disulfide-linked to A1C, and we added adenine to the cis solution. Adenine binds to DTA at its NAD site, stabilizing its native conformation and blocking translocation in the cell-surface translocation assay (13, 20). Thus, with adenine in the cis compartment, the C-terminal DTA moiety of biotin-LF_N-DTA should be held on the cis side of the membrane. Under these circumstances, if the biotinylated N terminus enters channels and becomes accessible to streptavidin added to the trans side, unblocking of the channels at negative voltages should be prevented, and this is what we saw. After the addition of streptavidin to the trans solution, there was little unblocking at –20 mV (Fig. 6D). Subsequent addition of Tris-carboxyethylphosphine to the trans solution reduced the disulfide linking the biotin (with its attached streptavidin) to LF_N, and this allowed the N terminus to return to the cis side at –20 mV, thereby unblocking the channels (Fig. 6D).

Discussion

We used LF_N, the minimal PA₆₃-binding translocation-competent domain of LF, to probe the way in which LF and EF are translocated. The crystal structure of LF shows that LF_N contains a disordered stretch of 27 residues at the N terminus and that this stretch is connected to the compactly folded remainder of the domain (residues 40–263) by an outwardly projecting α-helix (helix 1α1; residues 28–39) (21). Our results suggest that this flexible N-terminal stretch fosters the initiation of translocation in an N- to C-terminal direction. Deleting the entire stretch strongly inhibited acid-induced translocation of LF_N across the plasma membrane, whereas deleting only 13 residues had no significant effect on this activity. Consistent with this result, the flexible N-terminal stretch in EF is 10 residues shorter than in LF, but the two proteins translocate with similar efficiencies (13).

The effects of N-terminal truncations on translocation correlate with their effects on the ability of LF_N to block the ion conductance of PA₆₃ channels. Deleting more than ≈23 residues caused loss of channel-blocking activity, whereas deleting 15 or fewer had no effect. Intermediate-length deletions had variable effects that may depend in part on the charge of residues near the N terminus. Thus, the Δ18 mutation had less effect than the Δ21, perhaps because two acidic residues are exposed at the N terminus of the former (Fig. 1). The loss of activity seen with the longer deletions apparently does not result from alterations in the ability of LF_N to bind to PA₆₃, because truncations as long as 36 residues caused no significant loss of affinity. The hypothesis that channel blocking is a consequence of actual entry of the N terminus of LF_N into the channel, as opposed to, for example, allosteric induction of channel closure, is supported strongly by the effect of trans streptavidin on biotin-tagged LF_N-DTA. Thus, while the C terminus was held on the cis side of the membrane by adenine-liganded DTA, the N-terminal biotin tag became accessible to streptavidin in the trans compartment at +20 mV. This result indicates that the LF_N moiety can span the entire breadth of the membrane under these conditions, with the C terminus held in the cis compartment and the N terminus extending into the trans compartment. We reported elsewhere (22) that when adenine is removed from the cis solution, translocation to the trans compartment can then occur (under higher positive membrane potential). Also, we found in the current study that the N-terminal H₆-tag of H₆-LF_N is accessible to Ni²⁺ added to the trans compartment, indicating that the N terminus is either in the channel or in the trans compartment.

Which properties of the N terminus of LF_N underlie its functionality in translocation and channel blocking? One may be its flexibility, presumably allowing movement into the channel. A second may be location. The structure of LF- or EF-liganded PA₆₃ is not yet known, but LF and EF bind to sites on domain 1′ of PA₆₃ at the mouth of the channel (14). A third feature is suggested by the striking effect of H₆ fused to the N terminus of the truncation mutants. The H₆-tag actually raised translocation and channel-blocking activities to levels above those of wild-type parents. The N-terminal H₆-tag had a major effect even when the 21-residue sequence containing it was tethered directly to the ordered globular portion of the LF_N structure (helix 1α1 in construct Δ27) (21). In contrast, a H₆-tag at the C terminus had no effect. The fact that the translocation and channel-blocking activities occur in a pH range in which the side chain of His is expected to be protonated suggests a role for electrostatic potential. The lumen of the prepore is negatively charged, and presumably so is the cation-selective channel (8, 9). A positively charged N terminus might be attracted into the channel by electrostatics, even in the absence of an applied voltage, and indeed we have shown this (A.F., unpublished results). Applying a positive transmembrane voltage would be expected to foster this entry, and, as we saw, small positive voltage (+20 mV) caused channel blocking in the planar bilayer system. A negative voltage caused immediate unblocking, implying withdrawal of the N terminus into the cis compartment, and reapplying the positive voltage caused immediate reblocking. The latter result implies that, consistent with the high affinity of LF and EF for (PA₆₃)₇, LF_N remains bound at the channel mouth even in the presence of a negative voltage.

There have been few studies of the transmembrane potential of the endosomal compartment, where translocation is believed to occur. There is consensus that the interior of the compartment is positive with respect to the cytosol: the same polarity that promotes channel blocking. Estimates of the magnitude of the potential vary widely, however, from +10 to +300 mV (23–25). Whereas the pore might discharge the transmembrane potential, the N terminus of LF or EF inserted into the prepore might block the pore before it opens. Recent evidence suggests that the pore may form within multivesicular endosomes, such that LF and EF initially are deposited into vesicles of cytosol contained within these endosomes (26). This reaction would be followed later by back fusion with the endosomal membrane, liberating LF and EF into the bulk cytosol. No estimates of the transmembrane potential of the internal vesicles have been made, to our knowledge.

The charge characteristics of the N-terminal regions of LF and EF are consistent with electrostatic potential drawing the N terminus of LF or EF into the PA₆₃ channel (Fig. 1). There is a high density of ionizable residues, including a majority of cationic ones (at pH 5.5), extending through the first two-thirds of the sequence. Such reasoning also may explain the surprising finding that fusing short tracts of Lys, Arg, or His residues to the N terminus of DTA enabled this protein to enter cells in the presence of PA and inhibit protein synthesis (27). Entry promoted by polycationic tracts was inefficient relative to LF_N-mediated entry; the most efficiently translocated of the constructs, Lys₈-DTA, entered cells with ≈1/10th the efficiency of LF_NDTA. Free LF_N did not block entry of these constructs, suggesting that they interacted with sites on (PA₆₃)₇ distinct from the EF/LF sites. It was suggested that these constructs might bind weakly to anionic sites at the cell surface [perhaps on (PA₆₃)₇] and be codelivered with (PA₆₃)₇ to the endosome (27). Regardless of the route, once within the same compartment as (PA₆₃)₇, the polycationic N terminus could enter the anionic (PA₆₃)₇ channel and initiate translocation of the molten globular DTA by the same mechanism as EF and LF.

Two reports (28, 29) have described findings that might appear to challenge our model of translocation. These studies showed that LF_N can mediate PA₆₃-dependent entry of a fusion protein with DTA, regardless of the order of the two domains within the construct. One study (28) reported that DTA-LF_N, in which the DTA moiety was N-terminal, inhibited protein synthesis in cells at ≈10% the efficiency of LF_N-DTA, where it was C-terminal. The other study (29) reported that DTA-LF_N with an N-terminal H₆-tag inhibited protein synthesis with about the same efficiency as LF_N-DTA. Comparison of the two studies suggests an increase in translocation efficiency by DTA-LF_N, from 10% to equivalence with that of LF_N-DTA, when an N-terminal H₆-tag is present. This result is consistent with our model.

We are left, then, with the question of how the undecorated N terminus of DTA-LF_N initiates translocation. First, we note that in our experiments LF_N retained a significant level of translocation activity after we deleted 27, 36, or 39 residues from the N terminus (Fig. 2 A), indicating that the flexible 27-residue stretch at the N terminus is not absolutely required for translocation. Recent data show that acidic pH promotes unfolding of LF_N in solution (30), and if (PA₆₃)₇-bound LF_N is similarly perturbed, this might allow the truncated N terminus of LF_N to enter the channel at low efficiency. Similar reasoning applies to DTA-LF_N. There is evidence that the DTA chain must unfold at acidic pH to be translocated by the B-chain of this toxin (31), and it is believed to adopt a molten globule state under acidic conditions (32). If the DTA moiety of DTA-LFN assumes a similar state while the fusion protein is tethered by means of LF_N to (PA₆₃)₇, this may provide sufficient flexibility for its N terminus to enter the channel and initiate translocation at the low efficiency observed.

In summary, the evidence presented suggests that the N terminus of bound LF_N enters the (PA₆₃)₇ channel under the influence of a small positive voltage and initiates threading of unfolded (or progressively unfolding) forms of this protein through to the opposite face of the membrane. Elsewhere, we reported evidence that at higher positive voltages (≥40 mV), the entire LF_N molecule translocates to the trans compartment, relieving the channel of blockage (22). The combined evidence from the two studies indicates that PA contains all of the molecular machinery needed to translocate LF_N, and, hence, presumably full-length LF and EF, across a phospholipid bilayer. Thus, there is no absolute requirement of a receptor or any other proteinaceous factor for this process to occur. However, interaction of PA with one of its two known receptors, CMG2, regulates the pH at which (PA₆₃)₇ forms a channel (33, 34), and other proteins may influence the anthrax toxin translocation process in cells, as shown for diphtheria toxin (35).