Role of the α clamp in the protein translocation mechanism of anthrax toxin

Abstract

Membrane-embedded molecular machines are utilized to move water-soluble proteins across these barriers. Anthrax toxin forms one such machine through the self-assembly of its three component proteins—protective antigen (PA), lethal factor (LF), and edema factor (EF). Upon endocytosis into host cells, acidification of the endosome induces PA to form a membrane-inserted channel, which unfolds LF and EF and translocates them into the host cytosol. Translocation is driven by the proton motive force, comprised of the chemical potential, the proton-gradient (ΔpH), and the membrane potential (ΔΨ). A crystal structure of the lethal toxin core complex revealed an “α clamp” structure that binds to substrate helices nonspecifically. Here we test the hypothesis that through the recognition of unfolding helical structure the α clamp can accelerate the rate of translocation. We produced a synthetic PA mutant in which an α helix was crosslinked into the α clamp to block its function. This synthetic construct impairs translocation by raising a yet uncharacterized translocation barrier shown to be much less force dependent than the known unfolding barrier. We also report that the α clamp more stably binds substrates that can form helices than those, such as polyproline, that cannot. Hence the α clamp recognizes substrates by a general shape-complementarity mechanism. Substrates that are incapable of forming compact secondary structure (due to the introduction of a polyproline track) are severely deficient for translocation. Therefore, the α clamp and its recognition of helical structure in the translocating substrate play key roles in the molecular mechanism of protein translocation.

Introduction

The secretion [1] and degradation [2] of proteins are essential for a variety of processes including protein trafficking [1], membrane and organelle biogenesis [3], microbial toxin secretion [4] and subsequent entrance into host cells [5, 6], antigen presentation [6], and destruction of damaged proteins [6]. These cellular processes require large and complex molecular machines that are typically composed of multiprotein complexes, though this is not always the case [7]. These multiprotein complexes, called translocases, form aqueous pores in lipid bilayers thereby allowing other substrate proteins and peptides to be translocated across the membrane. Of course, translocases require substantial energy inputs, usually in the form of ATP binding and hydrolysis [2, 6] or the proton motive force (PMF) [8].

Anthrax toxin [9], one of two key virulence factors produced by Bacillus anthracis, the causative agent of anthrax disease, is an ideal model for studying transmembrane translocation [10–21]. It consists of three separately nontoxic proteins that associate to make toxic complexes. The three proteins include an 83 kDa channel-forming protein, protective antigen (PA), and two ~90 kDa enzyme effectors, lethal factor (LF) and edema factor (EF). First, PA is proteolytically activated to the 63 kDa monomer (PA₆₃), which assembles into heptameric or octameric prechannels that bind three or four moieties of LF or EF, respectively [15, 16]. These toxic complexes are endocytosed, and acidification of the compartment drives the PA prechannel oligomer to convert into a channel. Finally, the PMF that develops then drives LF/EF substrate unfolding [14, 22] and translocation into the host cytosol [13, 17].

A high-resolution cryo-electron microscopy (cryoEM) structure of the PA channel has recently been elucidated [23]. The channel is large and ~180 Å long with a narrow constriction point at its phenylalanine clamp (ϕ clamp) of about 6 Å. The remaining lumen diameter is, however, larger. The lumen of the ~110 Å long β barrel below the ϕ clamp is ~15 Å in diameter, and the lumen in the cap above the ϕ clamp is ~20 Å in diameter. Thus while the narrowest point at the ϕ clamp cannot accommodate an α helix, these wider diameter lumens can [22].

An emerging consensus on the dynamics of LF/EF translocation is that the proton gradient (ΔpH) component of the PMF created by the endosomal acidification drives substrate translocation via a charge-state Brownian ratchet [13, 17–21]. In brief, acidic residues in the substrate are more frequently protonated on the low pH endosomal side of the membrane, resulting in a net positive stretch of peptide that is allowed to move bidirectionally, according to Brownian motion, through the cation-selective channel. When the translocating chain reaches the higher pH cytosol, the acidic residues are more frequently deprotonated, yielding an anionic region no longer capable of retrotranslocation. In this way, movement is resolved in one direction, and productive translocation occurs until the entire substrate reaches the cytosol. However, a Brownian ratchet would only produce modest forces on the folded substrate. Hence, with this translocation model, it is not clear how larger forces may develop from the PMF that are sufficient to drive unfolding.

Our understanding of how the substrate unfolds is complicated by our lack of knowledge about the degree of folding. Is some helical secondary structure maintained in the translocating chain? Recently, the crystal structure of the octameric prechannel bound to LF’s PA-binding domain (LF_N) was reported [16]. It revealed that the α1 helix and β1 strand of LF_N unfold and bind a hydrophobic grove near the top of the PA channel, termed the α clamp. Mutational analysis of the substrate revealed that the α clamp will bind non-specifically to helices with a wide variety of side chain size, charge, and polarity [16]. Additionally, in the cryoEM structure, the clamp is positioned in such a way that extending the substrate’s α helix would orient it toward the channel lumen [23]. On one hand, the α clamp may function as a helix nucleation site, in accordance with the Zimm-Bragg formalism [24]. On the other hand, the clamp may be an allosteric helix sensor and binds helices after they nucleate inside the PA pore. This aspect of the mechanism of the α clamp remains to be resolved.

Here, we show how occlusion of the α clamp disrupts the translocation step of the mechanism. Using substrates incapable of forming more-compact α-helical states, we show that they have defective translocation rates and efficiency—results consistent with obligate helical translocation. We present a model for how helical translocation could be a necessary step in driving tertiary unfolding.

Results

A mechanism-based α clamp inhibitor construct disrupts substrate binding

Previously, we have shown that interactions between the α clamp and LF’s α1/β1 region provide 2.5 to 4 kcal mol⁻¹ of stabilization for the bound state [16]. Interestingly, replacements of LF’s α1/β1 region with other peptides from LF and EF do not readily alter the interaction with the α clamp, revealing its large degree of nonspecificity. The consequence of this property of the α clamp is that it is difficult to completely eliminate binding to the site with point mutations, and ultimately its role in the translocation mechanism is unknown. In point of fact, the clamp cannot be disrupted by mutating the structural twin-calcium ion binding sites, as they are too integral to the overall structure and stability of PA oligomer. To develop an approach that probes α clamp function yet maintains structural integrity of the oligomer, we attached a sequence corresponding to LF’s α1/β1 sequence (residues 26-49) to PA₆₃’s amino terminus, i.e. directly after PA₈₃’s cleavage site and preceding the first residue of post-cleavage PA₆₃ (Fig. 1A). The four Lys residues in the α1/β1 sequence were mutated to Ala to prevent unwanted tryptic cleavage when PA₈₃ is treated with trypsin immediately prior to oligomerization. These Lys residues are not critical for binding the α clamp [16]. Furthermore, PA’s P173 residue, one of six unstructured amino acids at PA₆₃’s amino terminus was mutated to a Gly to increase the region’s flexibility and allow the appended sequence to occupy the α clamp and act as a “plug”. As such, we termed the construct “PA_α-plug.” Additionally, we made several constructs in which this interaction was stabilized by cysteine crosslinking: ^236-40PA_α-plug, ^464-32PA_α-plug, and ^465-30PA_α-plug, where the first superscripted number is the position in PA mutated to Cys and the second is the equivalent LF residue on the appended helix also mutated to Cys.

Figure 1. α clamp occlusion inhibits substrate binding.

(a) (top) Sequences of WT PA and PA_α-plug. PA₂₀, which is cleaved prior to oligomerization, is highlighted (gray). PA₆₃ is shown colored (green), with the P173G mutation, added for linker flexibility (underlined). The point of proteolytic cleavage is indicated with a dashed line. The appended sequence, corresponding to LF residues 29-46, is colored (red). (bottom) A structural model of PA_α-plug, based on the crystal structure of LF_N bound to the octameric PA prechannel (3KWV) [16]. Adjacent PA₆₃ subunits are colored (green and blue) with the appended LF-plug sequence (red). Six unstructured residues of PA’s N terminus not seen in the crystal structure are drawn in as the flexible linker connecting PA to the appended LF helix. (b) Fraction of PA bound to substrate, inferred from the ratio of current after and before substrate addition (I/I_o), as a function of WT LF_N ligand concentration, [L], fit to a single-state binding model: I/I_o = 1 − a/(1 + K_d/[L]). The parameter a is a baseline parameter that estimates the value of 1 − I/I_o under saturating concentrations of substrate. The equilibrium dissociation constant, K_d, is shifted from 120 (±30) pM for WT PA (black) to 530 (±60) pM for PA_α-plug (green), 860 (±150) pM for ^236-40PA_α-plug (magenta), 3.3 (±0.6) nM for ^464-32PA_α-plug (blue), and 7.4 (±1.7) nM for ^465-30PA_α-plug (red). The a baseline for conductance block at saturating concentrations of LF_N is 93 (±4)% for WT PA, 88 (±2)% for PA_α-plug, 83 (±3)% for ^236-40PA_α-plug, 72 (±2)% for ^464-32PA_α-plug, and 72 (±3)% for ^465-30PA_α-plug. Each dataset is representative of experiments performed on multiple membranes.

To measure the substrate LF_N-binding thermodynamics, we used a planar lipid bilayer electrophysiology binding assay [16]. Here, two aqueous chambers, cis and trans, are separated by a planar lipid bilayer. The PA prechannel oligomer is added to the cis side of the membrane under an asymmetric KCl gradient ([KCl]_cis = 100 mM, [KCl]_trans = 0 mM, pH_cis = 6.5, pH_trans = 7.4), thereby allowing for detection of PA channel insertion and current increase at a ΔΨ of 0 mV (ΔΨ ≡ ΔΨ_cis − ΔΨ_trans; ΔΨ_trans ≡ 0 mV). Once the current stabilizes, excess prechannel is removed by perfusion with pH_cis 7.40 buffer to maintain a KCl gradient simultaneously while removing the pH gradient. Substrate LF_N is added at a variety of concentrations. Conductance blockade is observed in response to each addition of LF_N. The fraction of unblocked channels as a function of substrate concentration is well fit using a single-site binding model.

The equilibrium dissociation constant, K_d, for wild type (WT) LF_N with WT PA channels is 120 (±30) pM (Fig. 1B). With the non-crosslinked PA_α-plug construct, that value increases modestly to 530 (±60) pM. The ^236-40PA_α-plug channel slightly increases the extent of binding disruption, bringing the K_d to 860 (±150) pM. However, reactions with Ellman’s reagent suggest poor crosslinking in this construct. The other constructs, ^464-32PA_α-plug, and ^465-30PA_α-plug, had more substantial effects, increasing the K_d to 3.3 (±0.6) nM and 7.4 (±1.7) nM, respectively. For the most severe mutant, ^465-30PA_α-plug, this corresponds to a loss of 2.4 (±0.2) kcal mol⁻¹ of stabilization for the bound state, which was the previously reported value for stability imparted by the α clamp as determined by truncation of LF’s first helix and strand, α1/β1 [16].

Furthermore, the baseline indicating the maximum amount of current blocked under saturating concentrations of WT LF_N is shifted when the α clamp is rendered inaccessible (Fig. 1B). For WT PA, this baseline is 93 (±4)%. The non-crosslinked PA_α-plug and unsuccessfully crosslinked ^236-40PA_α-plug only shift this value to 88 (±2)% and 83 (±3)%, respectively. Once again, the largest effects come from ^464-32PA_α-plug and ^465-30PA_α-plug, where the maximum block baselines are 72 (±2)% and 72 (±3)%, respectively. Hence for the PA_α-plug mutations, while substrate binding to the α clamp was disrupted by up to 2.5 kcal mol⁻¹, the ability of substrate to initiate and dock into the channel’s conductance-blocking site, the ϕ clamp was perturbed but not fully disrupted.

PA_α-plug mutant possesses defective translocase activity

Because the substrate could still initiate into the channel, we then measured the translocation kinetics in the α-plug mutant background. As it was the more severe mutant, we specifically focused on ^465-30PA_α-plug. Planar lipid bilayer electrophysiology was again used to make translocation kinetics measurements [11–21]. Here the planar bilayer separated symmetrical aqueous chambers of 100 mM KCl, pH 5.6. PA prechannel oligomer was added to the cis side of the membrane under a ΔΨ of 20 mV. An increase in current followed by stabilization indicated successful channel formation, and excess prechannel was removed by perfusion. Next, a saturating concentration of substrate was added and allowed to fully block the channel, as inferred by the decrease in ion flow, before a second round of perfusion. Translocation was then initiated by increasing the ΔΨ. The observed translocation kinetic records are complex and multi-exponential. To estimate the rate of translocation, the time for half of the substrate to translocate (t_½) is measured. This approximation allows the rate-limiting step of the translocation reaction to be monitored according to the established theory of transit times used in enzymology. With this latter parameter, we can estimate the activation energy of translocation (ΔG^‡) by ΔG^‡ = RT ln (t_½/c), where R is the gas constant, T is the temperature, and c is a constant of 1 s.

As a function of the driving force, ΔG^‡ does not vary linearly with ΔΨ as expected for a simple system with only one major driving force-dependent rate-limiting barrier. Rather, in a ΔG^‡ versus ΔΨ plot, there is a steep slope at lower driving forces and a much shallower slope at higher driving forces. This phenomenon is consistent whether the driving force is a ΔΨ, ΔpH, or a combination thereof [13, 14, 17]. The two slopes correspond with the two major sets of barriers in the translocation mechanism. One set of barriers is highly driving force-dependent, and they dominate at lower driving forces. The other set of barriers is largely driving force-independent, and they are rate limiting at higher driving forces. Earlier work has extensively characterized the driving force-dependent barrier and has shown it to be limited by substrate unfolding, specifically the unfolding of LF_N’s β-sheet subdomain region [14]. The second barrier has yet to be identified, but it is likely a post-unfolding step involving translocation of the unfolded chain. Interestingly, the ^465-30PA_α-plug mutant has a different barrier profile than WT PA (Fig 2A). The major trend is that the driving force-independent barrier is raised, but the high driving force-dependent barrier is less affected.

Figure 2. Blocking the α clamp inhibits WT LF_N translocation more severely at greater driving forces.

(a) Activation energy versus ΔΨ results for the translocation of WT LF_N with either WT PA (black) or PA_α-plug (red). These translocations were conducted at a symmetrical pH of 5.6. Data are fit to a two-barrier model [14]. The error bars are the means ± standard deviation (s.d.) (n = 2–4). (b) Values of ΔΔG^‡ = ΔG^‡_α-plug - ΔG^‡_WT for the driving force levels displayed in panel a.

PA_α-plug mutant disrupts the more force-independent translocation step

The upward shift in activation energies shows that the ^465-30PA_α-plug mutant is defective in translocating WT LF_N (Fig. 2A). However, this translocation defect is more pronounced with higher driving forces. At 40 mV, our lowest driving force assayed, the change in activation energy for translocation (ΔΔG^‡ = ΔG^‡_α-Plug − ΔG^‡_WT) was increased less than 0.5 kcal mol⁻¹ (Fig. 2B). As the magnitude of the driving force is increased, ΔΔG^‡ rises as well, until it eventually levels off slightly above 1 kcal mol⁻¹ at the highest ΔΨ values we measured.

Our earlier model suggested that the α clamp played a role in unfolding, so we investigated the ability of the PA_α-plug mutant to translocate a construct with a mutation previously demonstrated to destabilize it, LF_N L145A [14]. Once again, translocation was slowed at all voltages, but in this case the extent of the defect did not vary with the magnitude of the driving force (Fig. 3A). The ΔΔG^‡ values for LF_N L145A remained just below 1 kcal mol⁻¹ across the same range of driving forces under which the ΔΔG^‡ doubled with WT LF_N (Fig. 3B).

Figure 3. Blocking the α clamp inhibits translocation of a destabilized mutant LF_N regardless of the driving force magnitude.

(a) Activation energy versus ΔΨ results for the translocation of LF_N L145A with either WT PA (black) or PA_α-plug (red). These translocations were conducted at a symmetrical pH 5.6. Data are fit to a two-barrier model [14]. The error bars are the means ± s.d. (n = 2–3). (b) Values of ΔΔG^‡ = ΔG^‡_α-plug − ΔG^‡_WT for the driving force levels displayed in panel a.

Finally, it should be noted that ^465-30PA_α-plug cannot be fully blocked by either WT or mutant substrate. Even with our low pH and small ΔΨ pre-translocation conditions, there remained some unblocked current. We are reporting the kinetics only for the substrates that are bound. If some channels bind no substrate, it stands to reason that a portion of the blocked channels have only a single substrate bound, whereas WT PA heptamers can bind up to three. It is possible that clearance of WT PA channels requires three translocation events while ^465-30PA_α-plug only requires one. Thus, we may be in fact be underestimating the extent to which the ^465-30PA_α-plug is inhibiting translocation.

Altering the shape of LF_N’s α1/β1 sequence disrupts channel binding and translocation

A complementary approach to obstructing the α clamp is to alter the α1/β1 sequence (residues 30-47) that typically binds there. We previously showed that the α clamp is a nonspecific site, and it stably binds a variety of polypeptide sequences [16]. The current hypothesis is that the α clamp may recognize substrates via a general steric shape-complementarity mechanism. To test this idea, we replaced LF_N α1/β1 (positions 30-47) with varying densities of Pro, which would result in a drastic change in their backbone configuration. Pro residues are highly disruptive to α-helix formation due to steric interference and the lack of amide hydrogens for hydrogen binding. Consecutive prolines do form a helical structure, the left-handed polyproline II helix, which in steric terms is narrower and consequently longer than a typical α helix [25] (Fig. 4A).

Figure 4. Polyproline helix disrupts binding and translocation.

(a) (top) Comparison of an α helix (based on LF’s α1 helix) and a polyproline II helix (middle) with an equal number of residues. (bottom) The sequences of the WT LF_N substrate and mutants with part or all of the α1/β1 region replaced with Pro residues. (b) Fraction of WT PA bound to substrate, inferred from the ratio of current after and before substrate addition (I/I_o), as a function of [L] fit to a single-state binding model: I/I_o = 1 − a/(1 + K_d/[L]). The equilibrium dissociation constant, K_d, is shifted from 120 (±30) pM for WT LF_N (black) to 7.8 (±0.2) nM for LF_N Pro_30-47 (red), 1.2 (±0.3) nM for LF_N Pro_36-47 (green), and 310 (±50) pM for LF_N Pro_43-47 (blue). The baseline parameter a for conductance block at saturating concentrations of LF_N is 93 (±4)% for WT PA, 78 (±3)% for LF_N Pro_30-47, 73 (±3)% for LF_N Pro_36-47, and 90 (±2)% for LF_N Pro_43-47. (c) Translocations of WT LF_N (black) and LF_N Pro_30-47 (red), and LF_N 6Pro_30-47 (cyan) at symmetrical pH 5.6 and with a ΔΨ of 50 mV. Each trace is representative of experiments performed on multiple membranes.

In our initial constructs, we replaced either the entire α1/β1 region (LF_N Pro_30-47), a portion of the α1 helix and the β1 strand (LF_N Pro_36-47) or just the β1 region (LF_N Pro_43-47) with consecutive Pro residues. If the proline substitution causes the α1/β1 region to form a shape that the α clamp cannot recognize, this substrate should be deficient in its ability to block the channel. Indeed, LF_N Pro_30-47 bound WT PA ~100 times weaker than WT LF_N, with the K_d increasing to 2.1 (±0.5) nM for the mutant from 120 (±30) pM for the WT substrate (Fig. 4B). This is a loss of 1.7 (±0.2) kcal mol⁻¹ of stabilization for the bound state. While substantial, this value is somewhat lower than ~2.5 kcal mol⁻¹ observed by occlusion of the α clamp (Fig. 1B) or full truncation of the α1/β1 region [16]. Furthermore, this mutant can only achieve a maximum block of 78 (±3)% compared to WT LF_N’s 93 (±4)%, again a defect similar to though slightly smaller in magnitude compared with what was observed when the α clamp was made inaccessible. Blocking only a portion of the α1 helix and β1 sheet with LF_N Pro_36-47 yielded similar results with a K_d of 1.2 (±0.3) nM and a maximum block parameter of 73 (±3)%. However, LF_N Pro_43-47 behaved much more like WT LF_N, with K_d of 310 (±50) pM and maximum block of 90 (±2)%. This indicates that disrupting the β1 sheet is tolerated; however, the α1 helix cannot be replaced with a polyproline sequence.

Proline-substituted LF_N substrates are deficient in translocation

While the polyproline sequence imparts binding deficiencies, the LF_N Pro_30-47 construct can block PA channels under our normal pre-translocation conditions (pH_cis = 5.6; ΔΨ of 20 mV). This allows us to explore whether the α clamp plays a role in translocation. If the α clamp plays no role post-initiation, there should be no difference in translocation between WT LF_N and LF_N Pro_30-47 under these conditions. Differences would only appear if interaction with the α clamp and/or the shape of the translocating chain matter in downstream translocation steps.

Indeed, there appear to be post-initiation translocation differences between WT LF_N and LF_N Pro_30-47. LF_N Pro_30-47 was unable to translocate with a 50 mV membrane potential, conditions under which WT LF_N can readily translocate (Fig. 4C). We next looked to see if we could sufficiently disrupt the shape the α1/β1 region using six helix-disrupting proline substitutions spaced throughout this sequence (Fig. 4A). This construct, LF_N 6Pro_30-47, would be predicted to disrupt α helix formation but would not form the unique polyproline II helix. Surprisingly, LF_N 6Pro_30-47 was not deficient in translocation under the same driving forces (Fig. 4B). If these mutants are indeed interacting with the α clamp despite the helix-disrupting presence of prolines, the site’s specificity may be even broader than we initially expected.

Discussion

The α clamp functions via a general shape complementarity mechanism

Previous work demonstrated the importance for the α clamp in substrate binding but did not definitively establish its role in the mechanism of translocation [16]. By occluding the α clamp with the PA_α-plug mutations (Fig. 1B) and disrupting LF_N’s α1/β1 region with LF_N Pro_30-47’s polyproline II helix (Fig. 4B) we have found further support for α clamp’s role in substrate recognition and binding. However, surprisingly LF_N Pro_30-47 did not disrupt binding as extensively as other approaches. This result suggested some weakened interaction may occur between the α clamp and a polyproline II helix. Hence, the expected activity of the α clamp may be to recognize polypeptide by a shape-complementarity mechanism. According to this shape-complementarity mechanism α helix is preferred by ~1.7 kcal mol⁻¹ relative to the narrower polyproline helix.

While the polyproline mutant LF_N Pro_30-47 translocated poorly relative to wild type LF_N, the mutant LF_N 6Pro_30-47 translocated similarly to wild type LF_N. This mutant had 6 proline substitutions. The discrepancy suggested the α clamp remains functional with less dense proline substitution. It should not be surprising that prolines are tolerated by the PA translocase, since they occur naturally in LF and EF. A peptide with fewer proline substitutions, however, could make a bulkier structure within the clamp site than pure polyproline. Hence, in accordance with a general shape complementarity mechanism, we propose the clamp may recognize thicker non-helical polymers with some proline substitutions; however, uniform polyproline are not well tolerated and these sequences are not present in LF and EF.

The other caveat of the polyproline mutant LF_N Pro_30-47 is that it substituted 4 lysine residues, which may have impacted its charge-dependent translocation. Based on prior work, it was found that positive charges in the 40s cassette (residues 40-49) were critical for voltage-dependent translocation [18]. Hence it is possible some of the measured defect for this mutant comes from the loss of positive charge. This caveat does not affect this interpretation of the binding result, since these charges have been shown to be superfluous to binding in a prior study [16].

Although earlier work determined the binding stability imparted by the α clamp based on α1/β1 truncations, those experiments could not address the α clamp’s role in channel blockade. With the PA_α-plug and LF_N Pro_30-47 mutants, we can disrupt the α clamp-α1/β1 interaction while still maintaining a WT N terminus in LF_N and a WT phenylalanine clamp (ϕ clamp) site in PA. At our highest substrate concentrations, LF_N will dock with the channel’s top surface even in the absence of the α clamp interaction, because the α clamp only comprises about half the total binding site. Neither PA_α-plug nor LF_N Pro_30-47 have mutations that can disrupt that interaction. That full blockade cannot be achieved under saturating conditions, however, indicates that bound substrates cannot fully block conductance, suggesting a role for the α clamp in guiding the substrate’s amino terminus into the channel. Future work will be required to understand the partial blockade phenomenon we observe for polyproline-containing sequences.

The α clamp’s role in the post-unfolding step(s) of translocation

Our main interest in these investigations of the α-clamp site is its role in substrate translocation. Its nonspecific helix-binding nature led us to hypothesize a role in nucleating helix formation. This led to two main predictions: (i) blocking access to the α clamp could inhibit substrate initiation at the ϕ clamp; and (ii) disrupting the substrate’s ability to bind the α clamp and form an α helix may generally impede the largely force-independent translocation step. Indeed, our data provides some insight on these models. Based on prior work, substrate initiation is defined as conductance blockade at the ϕ-clamp site, which is the major conductance-blocking site in PA [12]. We found that occlusion of the α-clamp site with the PA_α-plug mutations and introduction of polyproline in substrate sequences did not block initiation. However, we did find that substitution of a polyproline sequence in place of LF_N’s α1/β1 sequence greatly inhibited translocation (Fig. 4). Occlusion of the α clamp revealed an increase in the force-independent translocation barrier.

If plugging the α clamp in effect raises the force-independent translocation barrier, then we would predict a different set of results if we translocated a highly destabilized substrate (e.g., LF_N L145A). Such a substrate would have a reduced unfolding barrier. Therefore, the force-independent translocation barrier would dominate at all driving forces. Raising this latter barrier should affect translocation equally regardless of driving-force strength and to an extent comparable to the maximum change observed with the WT substrate (Fig. 5). Indeed, this is the case (Fig. 3), and the α clamp thus raises the force-independent translocation barrier by ~1 kcal mol⁻¹.

Figure 5. Occluding PA’s α clamp raises the driving force-independent translocation barrier.

The two major barriers for anthrax toxin translocation are depicted: the “unfolding” barrier and the “translocation” barrier. The unfolding barrier dominates at low driving forces and is strongly force dependent; and the translocation barrier dominates at high driving forces and is largely force independent [13, 14, 17]. The solid-line energy diagrams for WT PA with WT LF_N (top diagrams) and WT PA with LF_N L145A (destabilized LF_N mutant, bottom diagrams) are shown. Low (left) and high (right) driving force barrier diagrams are shown for each substrate and channel combination. The energy diagrams shown here explain the changes in translocation activation energy (ΔΔG^‡ = ΔG^‡_MUT - ΔG^‡_WT) that would be predicted with a mutated PA_α-plug channel (MUT) that increases the latter barrier (dashed line). With a WT substrate (top), the ΔΔG^‡ would be predicted to increase as the driving force increases. However, with a destabilized substrate, the second barrier would always be rate limiting and the ΔΔG^‡ should be consistent across a wider range of driving forces. These energetic depictions are consistent with the data collected in Figs. 2 and 3.

This more force-independent step is likely post-unfolding, because a prior study showed that destabilized point mutants failed to significantly shift the force-independent barrier [14]. What might a force-independent, post-unfolding translocation barrier look like structurally? One hypothesis is that this step involves the translocation of the unfolded chain, where the unfolded chain may contain helical structure. In support of this hypothesis, the translocation of unfolded chain will have lower force requirements than unfolding. One other aspect that is worth mentioning is the shape of the kinetics. Because the translocation kinetics remained S-shaped even for the PA_α-plug mutations tested, they imply that this force-independent barrier is really a series of similar height barriers. Hence post-unfolding translocation is likely the crossing of a series of similar height barriers. This model agrees in principle with prior studies examining single-channel translocation, which showed similar S-shaped kinetics at larger membrane potentials [26]. Thus the α-clamp site may operate as guide for translocating helical chain. This model suggests that this chain may not be fully unstructured, since the α-clamp site recognizes helix and not unstructured chain or thinner polymer structures such as polyproline. Therefore, we now hypothesize that the force-independent barrier corresponds to helix formation in the substrate, and the α-clamp site effectively lowers the activation energy for helix formation by serving as an upstream guide for helical structure into the channel.

Current translocation models

Given the results described here and prior work described elsewhere (12, 14, 16–18), we can compare and contrast two hypothesized models for translocation. One model is called the “extended-chain Brownian-ratchet model” [27]; the second model is the “allosteric helix-compression model” [5]. There are common initial steps in either model. A ΔpH is established such that the inside of the endosome is acidic (pH 5–6) relative to the outside cytosol (pH 7.5). The substrate LF/EF binds first to the top C-terminal docking site, and then the substrate’s α1/β1 N-terminal region docks into the α-clamp site. The models diverge in subsequent steps of the mechanism.

For the extended-chain Brownian-ratchet model, the peptide that initiates into the channel diffuses within the channel as extended-chain with no helical structure [27]. Negatively charged residues (Asp and Glu) are protonated and neutralized in charge. This step allows the neutralized peptide chain to diffuse past the charge clamp selectivity filter in the channel. The charge clamp selectivity filter excludes anions, and so the neutralized peptide could pass the filter at that point. Then the peptide deprotonates down gradient causing the peptide to become anionic. This deprotonation thus prevents retrotranslocation due to the charge clamp’s exclusion of anionic charge. The peptide can productively protonate again for a new round of extended-chain translocation until the entire peptide is translocated. This model suffers from its inability to explain how the protein is unfolded, because Brownian ratchets typically produce low forces. The model also does not include a significant role for the α-clamp site in the mechanism.

By contrast, the allosteric helix-compression model (Fig. 6) allows the extended chain to initially enter into the channel, but upon protonation, the chain compresses into a shortened α helix. (Note that extended chain is about 3.5 Å per residue, but α helix is about 1.5 A per residue.) This proton-driven compression into helix could then apply an unfolding force on a folded domain in LF or EF. The α clamp may then act as an allosteric trigger when helix is bound to the site. This triggers a conformational change in the channel, forcing the helical structure to populate extended chain. This allosteric change is a putative feature of this model but consistent with the recent high-resolution cryoEM structure of the PA channel, which shows the central ϕ clamp is too narrow to accept α helix, but it could accept extended chain [23]. When allosterically triggered, the peptide will be pushed past the charge clamp selectivity filter. Then the peptide can deprotonate down gradient, allowing it to become electrostatically captured by the charge clamp. A new round of translocation can then take place, where helix reforms inside the channel. The latter model is currently favored because it incorporates a role for the α clamp, helical structure in the substrate, and may allow for larger unfolding forces (power strokes) to be developed, explaining in part how folded domains can be unfolding during translocation.

Figure 6. Helix-compression mechanism of translocation.

Schematic of the PA channel with a substrate “Load” leading its amino terminus into the lumen. In the helix-compression mechanism negatively charged residues are protonated and the amino terminus converts from extended chain in an α helix. This conversion contracts the end-to-end distance of the terminus, causing a force to be applied on the load. Also in order to accommodate helical structure, the central ϕ clamp site dilates. Helix binding in the α clamp causes an allosteric conversion of the ϕ clamp site to the more closed state, which favors extended chain. The extended chain state extends peptide past the charge clamp, where it deprotonates to the high pH side of the membrane. The cycle can then repeat on the next segment of peptide in the substrate.

Materials & Methods

Proteins

Recombinant PA, LF, and mutants were expressed and purified as described [14, 15]. Assembly PCR and QuikChange were used to construct PA_α-plug and LF_N polyproline mutants described in the Results [16, 17]. The six-histidine amino-terminal tags were removed from substrates using bovine α thrombin [14]. PA₇ prechannel oligomers were assembled as described [15].

Electrophysiology

Planar lipid bilayers were formed as described [14] by painting [28] a membrane-forming solution (3% 1,2-diphytanoyl-sn-glycerol-3-phosphocholine in n-decane; Avanti Polar Lipids) across a 100 μm aperture in a 1-mL, white-Delrin or polysulfone cup [14–18]. A capacitance test confirmed the quality of the membrane. The membrane separates the cis and trans chambers, which each contained 1 mL of universal bilayer buffer (UBB; 10 mM oxalic acid, 10 mM MES, 10 mM phosphoric acid, 1mM EDTA) or a supplemented UBB (SUBB; 6 mM oxalic acid, 6 mM MES, 6 mM phosphoric acid, 6 mM TAPS, 6 mM boric acid, 1mM EDTA). Generally, an equivalent of 100 mM KCl was added to each buffer, except where indicated. Ag/AgCl electrodes bathed in saturated 3 M KCl were linked to the chambers via 3 M KCl-agar salt bridges. PA currents were obtained via an Axopatch 200B amplifier and recorded using AXOCLAMP10.

Translocation assays

Bilayers were bathed in symmetrical UBB or SUBB. ΔΨ ≡ Ψ_cis − Ψ_trans (Ψ_trans ≡ 0). PA₇ prechannel was added to the cis chamber (held at 20 mV), and conductance was blocked by the addition of substrate (WT LF_N or LF_N L145A) to the cis side (still held at 20 mV in symmetric pH 5.6 experiments). The substrate blockade was >95% of the original current, except where indicated. Excess substrate was perfused by a custom hand-cranked, push-pull perfusion system. Substrate translocation was driven by increasing the ΔΨ. An approximate translocation activation energy (ΔG^‡) was determined: ΔG^‡ = RT ln t_½/c [14]. The t_½ value is the time for half the substrate to translocate; c is a 1-sec reference.

Binding assays

Bilayers were bathed in asymmetric KCl solutions buffered in 10 mM potassium phosphate ([added KCl]_cis = 100 mM, [added KCl]_trans = 0 mM, pH_cis = 6.5, pH_trans = 7.40). Once PA channel insertion was complete the cis buffer was perfused and exchanged to pH 7.40, 100 mM KCl. (The pH of the cis and trans buffers were matched to 0.01 units.) LF_N was then added to the cis side of the membrane at small concentration increments, allowing for binding equilibrium to be maintained. Final current (I) levels were recorded, and the equilibrium current-block versus ligand concentration, [L], curves were fit to a simple single-binding site model, I/I_o = 1 − a/(1 + K_d/[L]), to obtain the equilibrium dissociation constant, K_d, where I_o is the current amplitude with no substrate and the baseline a estimates the value of 1 − I/I_o under saturating concentrations of substrate.

Highlights.

• Anthrax toxin protective antigen is a model transmembrane translocase

• Protective antigen oligomers contain an α clamp structure that binds to α helices

• An inhibitor (called α plug) was made by fusing a helix into the α clamp cleft

• Occlusion of the α clamp disrupts the translocation step of the mechanism

• Helix-compression mechanism is now favored over pure Brownian ratchet

Abbreviations

EF

edema factor

LF

lethal factor

LF_N

lethal factor’s amino-terminal domain

PA

protective antigen

PA₆₃

63 kDa PA fragment

PA_α-plug

inhibited PA

PMF

proton motive force

SUBB

supplemented universal bilayer buffer

UBB

universal bilayer buffer

WT

wild type

α clamp

cleft on PA that binds α helices

ΔpH

proton gradient

ΔG^‡

activation free energy

ΔΨ

membrane potential

ϕ clamp

phenylalanine clamp