Site-Specific Labeling and ¹⁹F NMR Provide Direct Evidence for Dynamic Behavior of the Anthrax Toxin Pore ϕ-Clamp Structure

Abstract

The anthrax toxin protective antigen (PA), the membrane binding and pore-forming component of the anthrax toxin, was studied using ¹⁹F NMR. We site-specifically labeled PA with p-fluorophenylalanine (pF-Phe) at Phe427, a critically important residue that comprises the ϕ-clamp that is required for translocation of edema factor (EF) and lethal factor (LF) into the host cell cytosol. We utilized ¹⁹F NMR to follow low-pH-induced structural changes in the prepore, alone and bound to the N-terminal PA binding domain of LF, LF_N. Our studies indicate that pF-Phe427 is dynamic in the prepore state and then becomes more dynamic in the transition to the pore. An increase in dynamic behavior at the ϕ-clamp may provide the necessary room for movement needed in translocating EF and LF into the cell cytosol.

The anthrax toxin protective antigen (PA) is secreted by Bacillus anthracis and binds to host immune cells.² Upon binding, proteolytic cleavage⁴ of PA leads to the formation of either a heptameric⁵ or octameric prepore,⁶ providing a surface that allows binding of two enzymatic components of the toxin: edema factor (EF) and/or lethal factor (LF). Following binding, the toxin is then endocytosed into the cell⁷ and at low pH forms a membrane-spanning pore that allows entry of EF and LF into the cell cytosol, where they target functions that severely disrupt the immune response.⁸ Several toxins utilize a pore mechanism to allow entry of toxin components into the cell, including the Clostridioides difficile toxin⁹ and Clostridium perfringens iota toxin.¹⁰ The mechanism by which these toxins transport their cargo into cells remains unclear in part due to the very narrow channel comprised of phenylalanines, termed the ϕ-clamp,¹¹ that the polypeptides must traverse to gain entry into the cell cytosol. Early studies showed that Phe427 in PA played a key role in translocation of EF and LF into cells. Mutations at this site have been shown to prevent translocation of LF_N-DTA, a fusion of the N-terminal PA binding domain of LF and the catalytic domain of diphtheria toxin, into CHO-KI cells.^12,13 X-ray structural studies of the heptameric prepore by Lacy and co-workers showed that residues near Phe427 (Phe427 was not observed in the crystal structure) resided within the lumen of the prepore.¹⁴ Finally, seminal studies by Krantz and Melnyk showed that the low-pH-induced transition to the pore state caused a constriction of the Phe427 residues to form a narrow iris, the ϕ-clamp.¹¹ Mutations at this site also revealed that translocation was dependent on the integrity of the ϕ-clamp and that while LF_N could bind and engage the ϕ-clamp and prevent ion conductance, certain mutations (Phe427 to Ala, Gly, Ileu, and Asp for instance) would not translocate LF_N across the bilayer membrane.

The recent cryo-electron microscopy (cryo-EM) structure of the pore state at high resolution (2.9 Å) confirmed that the ϕ-clamp forms a narrow 6 Å iris, where the Phe427 rings lie in a nearly face-to-face arrangement (Figure 1), and suggested that only completely unfolded polypeptides lacking secondary structure can pass through.¹ However, even in a completely unfolded polypeptide chain, the Cα to side chain end atoms of arginine, lysine, or tryptophan are on the order of 6 Å, suggesting that to translocate EF or LF in an unfolded state, the ϕ-clamp must be able to expand and contract to allow these side chains to pass through. We modeled the first 10 amino acids from the N-terminal end of LF (AGGHG-DVGMH), using the single poly-L-proline II helix from the collagen α2(I) chain as a template,¹⁵ and placed this through the ϕ-clamp iris. The modeled structure shows obvious steric clashes among the His, Met, and Asp residues, highlighting the need for expansion to allow translocation (Figure 1).

Figure 1.

Anthrax toxin pore and regions encompassing the α-clamp and the ϕ-clamp. Panels A and B show top-down and side views, respectively, of the cryo-EM structure of the PA pore (Protein Data Bank entry 3J9C).¹ We have highlighted the α-clamp region in cyan and yellow on the surface of the pore, and the ϕ-clamp ring of Phe427 residues is also highlighted. In panel C, we modeled LF (residues 1–10) (cyan) into the center of the ϕ-clamp iris. One can see that there are clear steric clashes between the Phe427 residues and the side chains of LF, including Met, His, and Asp. This figure was generated using ChimeraX.³

Although the 6 Å diameter seen by cryo-EM suggests that the passage is too narrow to allow unfolded polypeptides to pass through, early ion conductance studies of the anthrax pore in planar lipid bilayers using tetraalkylammonium ions of differing sizes placed an upper diameter of the pore at 12 Å.^16,17 Similarly, studies using different sized polyethylene glycols (PEGs) suggested an upper diameter of 20 Å,¹⁸ which even if these are overestimates suggests that despite the narrow ϕ-clamp structure observed by cryo-EM, larger diameters can be achieved. We now present direct evidence, using ¹⁹F NMR, that the ϕ-clamp is in fact highly dynamic.

To do this, we utilized the Fürter system¹⁹ that allowed us to site-specifically incorporate p-fluorophenylalanine (pF-Phe) into the 83 kDa form of PA (pF-PhePA₈₃) at Phe427. The purified pF-PhePA₈₃ was ~73% labeled as determined by LC-MS/MS (Supplementary Data S1). Partial digestion with trypsin allows cleavage of the first 157 amino acids of PA within domain 1, yielding a 63 kDa fragment that oligomerizes into a heptameric prepore (pF-PhePA₆₃)₇.⁵

Figure 2A shows the ¹⁹F NMR spectrum of 3 μM prepore (pF-Phe427-PA₆₃)₇ at pH 8.5, collected at 600 MHz using a fluorine cryoprobe, and shows that there are two overlapping resonances in the prepore state (−36.97p and −37.21p). This was quite striking, as the MW of the heptameric prepore is 63 × 7 = 441 kDa, and we attribute the ability to see resonances in this state to the increased dynamic behavior of pF-Phe427. As mentioned, in the crystal structure of the heptameric prepore state, electron density is missing for Phe427, suggestive of increased dynamic motion.¹⁴ The middle figure is after addition of the detergent Fos-14 (1:1000, pF-PhePA₆₃:Fos-14) and shows that addition of Fos-14 resolves a smaller upfield peak (~37.34p), suggesting that the pF-Phe427 in the prepore structure occupies at least two different conformational states. It also indicates that the prepore remains intact in this detergent; Fos-14 was deemed the best detergent in a myriad of detergents tested that could largely prevent aggregation of the pore at low pH while maintaining structural integrity.²⁰ Finally, the top figure presents the spectrum after the addition of sufficient 1 M phosphoric acid to induce a transition to pH 5.5, a pH that allows pore formation to occur.²¹ The spectrum shows that the −37.34p resonance is now sharper (Lorentzian line fitting gives an approximate line width change from 316 to 190 Hz) with a higher intensity, suggesting that there is a major population shift to the upfield resonance.

Figure 2.

(A) ¹⁹F NMR spectra of the heptameric (pFPhe427PA₆₃)₇ prepore and pore states. The bottom (blue) represents 3.3 μM heptameric prepore in 20 mM Tris (pH 8.5), 0.4 M NaCl, and 1 mM CaCl₂ buffer, with the addition of 50 μL of D₂O and 12.5 μL of 2 mM 5-fluorotryptophan as an internal reference (−46.03 ppm). The middle spectrum (red) is after the addition of 20 μL of 1 M Fos-14 detergent in water, and the top spectrum (black) after the addition of 10 μL of 1 M phosphoric acid. (B) ¹⁹F NMR spectra of the heptameric (pFPhe427PA₆₃)₇ prepore and effects of the addition of LF_N. Spectra and sample conditions are the same as in panel A, except the concentration of PA was higher [7 μM heptameric prepore (blue)], with the addition of LF_N (red) to give a final ratio of LF_N to heptameric prepore of 3:1. The green spectrum is after the addition of Fos-14, and the black spectrum after the addition of phosphoric acid. Note the new resonance in the spectrum at pH 5.5 (arrow). Asterisks denote small molecule impurities in the samples. Spectra in panels A and B represent 40000 transients collected at 20 °C on a Bruker 600 MHz instrument equipped with a ¹⁹F cryoprobe and were processed with 10 Hz of line broadening.

LF and EF bind to the prepore, and the N-terminal α-helical regions of LF and EF have been shown in recent cryo-EM studies to engage the α-clamp (Figure 1), which positions each LF or EF for translocation through the pore.^22,23 Engagement of the α-clamp has also been shown to favor an allosteric change at the ϕ-clamp, which in single-channel conductance measurements could exist in the pore state in either a more conductive open state (unclamped empty) or less conductive open state (clamped empty).²⁴ Engagement favors a transition to the less conductive clamped state, and in planar lipid bilayer experiments, LF_N (the N-terminal PA binding domain of LF) is able to engage the ϕ-clamp and block channel conductance.²⁵ To determine if the pF-Phe resonance can sense an environmental change when LF_N is bound, we repeated the experiments described above in the same manner but in the presence of LF_N. In Figure 2B, after the addition of LF_N in a 1:3 (pF-Phe427-PA₆₃)₇:LF_N ratio, the resonance appears to broaden, and after the addition of Fos-14, at least three resonances become apparent. The sharp resonance (asterisk) in the spectra is likely a small molecule contaminant from the purification process. After the addition of phosphoric acid, the resonance we attribute to the pore state is now sharper, the line width decreasing to ~90 Hz, but is asymmetric, suggesting that other states are likely present in smaller populations. We see another resonance that becomes more apparent at −36.6 ppm (arrow). Because this resonance is not present in the prepore in the absence of LF_N, we take this to indicate that LF_N is likely causing an environmental change at the ϕ-clamp.

The narrowed line width and greater intensity observed in our spectrum of the pore state (either with or without LF_N) are surprising. As mentioned above, the cryo-EM structure of the pore state showed that the Phe427 side chains form a narrow iris that is only 6 Å wide, and each of the Phe427 side chains is in a face–face orientation.¹ In studies of crystals of pF-Phe, the aromatic rings are oriented in a face-to-face arrangement and showed a relatively high barrier to rotation about the Cβ–aromatic bond (~20 kcal/mol), in contrast to crystals of Phe (~5 kcal/mol).²⁶ The higher barrier was attributed to the closer packing of the pF-Phe rings; that if rotation does occur, other planar rings within van der Waals contact distance must also rotate in a cooperative manner. Given the proximity of the Phe rings in the pore state, rotation around the Cβ–aromatic bond should also be sterically hindered from rotation and assume a rotational correlation time (τ_i) that matches that of the entire pore structure.

In elegant studies of fluorotyrosine-labeled alkaline phosphatase, the theoretical basis for relaxation processes that give rise to fluorine line widths was described.^27,28 It was shown that relaxation via chemical shift anisotropy (CSA) contributes substantially to the observed line widths, increases as the rotational correlation time (τ_c) of the molecule increases, increases as the field strength increases, and increases as the degree of motion around the Cβ–aromatic bond decreases (τ_i approaches isotropic motion).²⁸ Assuming a spherical shape for the prepore, a minimum estimate of τ_c is ~185 ns and is expected to increase in the transition to the more elongated pore.¹ Thus, our expectation was that the transition to the pore state would lead to extreme broadening of the pF-Phe427 resonance into the baseline. However, the line width of the pore resonance is narrower than the prepore and narrows even further when LF_N (which is expected to increase τ_c even further) is bound.

Due to sample limitations, we could not account for relaxation mechanisms (T1 and T2) or CSA contributions to the observed line widths. However, our data suggest that the narrower resonance observed in the pore state is due to increased dynamics at the pF-Phe427 site. In early planar lipid bilayer experiments, current fluctuations between open and closed conductance states of the pore could be observed from single channels that occur on a subsecond time scale.²⁹ Further single-channel conductance studies also indicated a fast (~20 ms) flickering between open and closed conductive states.³⁰ The fast flickering between open and closed states of the PA pore has also been observed by others,^31,32 and closures are not due to contaminating molecules present in solution.³³ The narrowed line width of the pore state may represent this fast-flickering motion between the open and closed states. How then can we explain the even narrower line shape of the pore resonance in the presence of LF_N?

One explanation is that the presence of LF_N may prevent solvent exposure at the ϕ-clamp, narrowing the line width by decreasing the level of solvent dipolar relaxation. However, as mentioned above in the recent cryo-EM structure of the pore with bound EF and/or LF, N-terminal portions of LF and EF bind to the α-clamp.²² Using peptides representing the N-terminal end of LF (residues 1–50), binding of peptides to the α-clamp induces an allosteric change at the ϕ-clamp.²⁴ In these studies, peptide binding shifts the equilibrium between two separate open subconductance states, favoring a more closed, clamped state.²⁴ If these states were present in our studies, binding of LF_N would shift the equilibrium to the more closed clamped state, limiting the fast flickering to a more narrow range of open conformational states. This could give rise to the narrower line width that we observe. In addition, single-channel recordings of the PA pore in the presence of helix-favoring peptides revealed at least two intermediate conductance states, aside from the open and closed states, again indicating that the ϕ-clamp is a dynamic structure and not static.³⁴ To allow for translocation of EF and LF through the narrow ϕ-clamp, there must be some dynamic flexibility in this structure, and our ¹⁹F NMR data would support a more dynamic ϕ-clamp that is likely able to expand and contract to sterically accommodate all 20 amino acids and combinations of sequences translocating through the channel.

ABBREVIATIONS

NMR

nuclear magnetic resonance

PA

protective antigen

EF

edema factor

LF

lethal factor

T1

longitudinal relaxation time

T2

transverse relaxation time

CSA

chemical shift anisotropy