A Single Amino Acid in the A2 Linker of Cholera Toxin Affects the Efficiency of Toxin Disassembly and Resulting Toxicity

Abstract

Cholera toxin (CT) from Vibrio cholerae and heat-labile enterotoxin (LT) from enterotoxigenic Escherichia coli are highly related AB₅-type protein toxins that contain a catalytic A1 subunit, an A2 linker, and a cell-binding homopentamer of B subunits. The two proteins bind the same GM1 receptor and have similar levels of enzymatic activity toward their shared Gsα target, yet CT is more potent than LT. This difference has been mapped to a sequence of 11 amino acids (residues 226–236) near the C-terminus of the A2 linker. We now identify a single amino acid, aspartic acid at position 229, as a key contributor to the function of CT: substituting it for glutamic acid 229 from LT reduced the toxicity of CT to the level of LT. Toxin disassembly by protein disulfide isomerase (PDI), which is an essential event in the intoxication process, was less efficient for CT_D229E and LT than for wild-type CT. This was further confirmed by a CT variant, in which four residues were converted into the LT sequence (D229E, I230V, T232I, and H233Y). To test our previous hypothesis that the positioning of the A1 subunit over the B pentamer critically determines toxin disassembly and thus potency, we determined crystal structures of both toxin variants to 1.6 Å resolution and calculated the angle at which the A2 linker exits the pore. The conformations of the two toxin chimeras lie between CT and LT, with the A2 C-terminus of the CT quadruple variant adopting essentially the same fold as LT. The A subunits of the toxin variants as well as that of LT showed a higher conformational freedom than CT as determined by molecular dynamics simulations, suggesting that a more rigid positioning of the A subunit drives toxin disassembly and potency. Our collective data uncover the critical role of D229 in toxin folding and structural dynamics, which in turn impacts toxin potency through effects on the efficiency of PDI-driven toxin disassembly.

Introduction

Vibrio cholerae and its cholera toxin (CT) induce the profuse, life-threatening diarrhea of cholera. , Enterotoxigenic Escherichia coli, which produces heat-labile enterotoxin (LT), causes a milder diarrheal disease that still causes significant morbidity and pediatric mortality in developing countries. , CT and LT are AB₅ toxins with a catalytic A1 subunit, an A2 linker, and a cell-binding B homopentamer (Figure A,B). The A1 subunit (residues 1–192 for CT and 1–194 for LT) sits above the ring-like B pentamer and is connected to the rest of the toxin by a single disulfide bond between the C-terminus of the A1 subunit and the N-terminus of the A2 subunit. The A2 subunit (residues 193–240 for CT and 195–240 for LT) also maintains numerous non-covalent contacts with both the A1 subunit and the B pentamer, thus tethering the A1 subunit to the rest of the toxin. The intact holotoxin is transported by vesicle carriers from the plasma membrane to the endoplasmic reticulum (ER) of a target cell. The A1/A2 disulfide bond can be reduced at the resident redox state of the ER, but this alone does not lead to spontaneous toxin disassembly. , Protein disulfide isomerase (PDI) is specifically required to displace the reduced A1 subunit from the A2/B₅ complex. − The released A1 subunit then crosses the ER membrane and enters the cytosol to elicit a cytopathic effect through ADP-ribosylation of the α subunit of the heterotrimeric G protein Gs (Gsα).

1.

CT and LT structural differences. CT (A; Protein Data Bank (PDB) entry 1S5E ) and LT (B; PDB entry 1LTS ) are each composed of a catalytic A1 subunit (red), an A2 linker (yellow), and a cell-binding B homopentamer (gray). For activation, the A subunit needs to be cleaved (indicated by the scissors) and reduced (S–S; orange sticks), resulting in A1 and A2 subunits. The A2 subunit extends into the central pore of the ring-like B pentamer (ribbon representation), which is transparent to highlight the locations of the A2 C-terminal residues. The greater cellular potency of CT in comparison to LT has been mapped to four amino acid differences near the C-terminus of the A2 subunit: residues 229 (red), 230 (green), 232 (orange), and 233 (blue). The cyan sphere in the A1 subunit of CT represents a conserved sodium ion. (C) CT variants were generated by changing the CTA2 residues to the corresponding LTA2 residues at either one (229; CTv1) or all four differing residues (229, 230, 232, 233; CTv4) between CT and LT. An additional amino acid difference between CTA2 and LTA2 at position 237 in the ER retention motif was previously shown to have no impact on relative toxin potency. The residue at position 225 also differs between the two toxins (G in CT and D in LT) but has not been probed in our current analysis nor in previous works.

CT and LT share ∼80% sequence identity. Both toxins exhibit high-affinity interactions with their shared GM1 ganglioside surface receptor and display similar in vitro ADP-ribosylation activities against their shared Gsα target. However, CT is much more potent than LT in cell culture: it elicits a greater cytopathic effect than LT and does so more quickly than LT. − The difference in toxicity has been mapped to 11 amino acid residues (226–236) near the C-terminus of the non-catalytic A2 subunit (Figure C). This region of the A2 linker is located within the central pore of the B pentamer (Figure A,B). Replacing the C-terminal region of LTA2 with the corresponding sequence from CTA2 produced a chimeric LT that was more potent than wild-type LT in cell culture. Likewise, a chimeric CT containing the C-terminal region of LTA2 exhibited cellular toxicity that was less potent than wild-type CT.

We recently demonstrated that the PDI-driven disassembly of LT is less efficient than that of CT. This would limit LTA1 access to the cytosol and the extent of LT intoxication. We hypothesized that the interaction between PDI and LT differs from the interaction between PDI and CT because the A1 chains for the two toxins are positioned at different angles above their respective B pentamers, with a 49° angle for CTA1 and a 40° angle for LTA1. These observations suggested that the C-terminal region of the A2 subunit affects toxin potency by positioning the A1 subunit within the holotoxin, thereby influencing the efficiency of PDI-driven toxin disassembly and, ultimately, the cellular activity of the released A1 subunit.

Here, we show that amino acid residue 229 in the CTA2 subunit plays a key role in establishing the efficiency of PDI-driven toxin disassembly and toxin potency. This residue represents one of the four differences between CTA2 and LTA2 in the critical C-terminal peptide of the A2 subunit (Figure C). We focused our studies on this specific amino acid residue because it represents the site where CTA2 bends in the B pentamer pore, which could influence the tilt angle of the A2 α-helix. As documented by enzyme-linked immunosorbent assay (ELISA) and surface plasmon resonance (SPR) spectroscopy, the PDI-driven disassembly of CT_D229E (henceforth referred to as CTv1) was less efficient than wild-type CT. CTv1 was accordingly less active against cultured CHO cells than wild-type CT, exhibiting a level of potency comparable to wild-type LT. A second CT variant (CTv4) with four amino acid substitutions (D229E, I230V, T232I, and H233Y), which completely converted the C-terminus of the CTA2 subunit to the LTA2_226‑236 sequence, behaved similarly to CTv1. Crystal structures of CTv1 and CTv4 were both determined to 1.6 Å resolution. We observed that, within the pore where the substitutions reside, the toxin variants differ considerably from CT. The C-termini of CTv1 and CTv4 exhibit an extended conformation that, for CTv4, ends in a turn that anchors the A2 subunit in the pore, as in LT. Molecular dynamics (MD) simulations of the different toxins show higher flexibility of the A1/A2 subunits of LT and the CT variants compared to wild-type CT, which may affect the efficiency of PDI-driven disassembly.

Materials and Methods

Materials

Ganglioside GM1 was purchased from Sigma-Aldrich (catalog no. 345724). Trypsin, soybean trypsin inhibitor, and TMB were purchased from Fisher Thermo Scientific. The rabbit anti-CTA1 antibody was purchased from Advanced Targeting Systems (catalog no. AB-43), and the HRP-conjugated goat anti-rabbit IgG antibody was purchased from Jackson ImmunoResearch (catalog no. 111-035-003).

Toxin Constructs

Dr. Randall K. Holmes kindly provided the pARCT5 vector encoding the ctxAB operon under the control of an l-arabinose-inducible promoter and containing the signal sequence from E. coli LT-IIb. Site-directed mutagenesis (Q5 Site-Directed Mutagenesis Kit, NEB), following the standard manufacturer’s protocol, was performed to obtain two chimeric constructs of CT with either a single substitution (D229E, CTv1) or four amino acid substitutions (D229E, I230V, T232I and H233Y, CTv4) to the LT sequence. DNA oligo nucleotides (Eurofins Genomics) are provided in Table S1.

For the expression of LT from ETEC strains infecting humans (hLT) or pigs (pLT), we designed the vectors pARhLT5 and pARpLT5. The corresponding DNA was synthesized by Genscript into pUC57 vectors with LT-IIb signal sequences and subcloned into a mutated pARCT5 vector containing only two (instead of three) NcoI restriction sites. These sites were used to replace the DNA encoding CTA and CTB with the sequences encoding LTA and LTB.

Production and Purification of CT, CTv1, CTv4, and pLT

Production of CT wild-type and variants was performed as described recently. Briefly, V_max X2 cells transformed with the corresponding plasmids were grown overnight at 30 °C and 120 rpm in LB-v2 salts medium containing 25 μg/mL chloramphenicol (CAM). The pre-cultures were used to inoculate 500 mL of LB-v2 salt medium (12.5 μg/mL CAM) and grown until the optical density at 600 nm (OD₆₀₀) reached ≈0.8. Expression was induced with 0.2% l-arabinose (Sigma), and cultures were incubated at 30 °C, 120 rpm for 20–22 h.

Both CT wild-type and variants are secreted to the growth medium and were purified following the same protocol. After removing the cells by centrifugation, the growth medium was filtered and loaded on a 5 mL d-galactose affinity chromatography (d-galactose agarose resin from Thermo Scientific), previously equilibrated with the loading buffer (50 mM sodium phosphate pH 7.4, 200 mM NaCl). The column was washed with 15 column volumes (CV) of loading buffer, and proteins were eluted with 10 CV of elution buffer (50 mM sodium phosphate pH 7.4, 200 mM NaCl, 500 mM d-galactose). The samples were concentrated by ultrafiltration (4 °C, 4000g) using an Amicon Ultra Centrifugal Filter Units with a 30 kDa MWCO (Merck) and further purified by size-exclusion chromatography (SEC). Proteins were applied to a Superdex 200 increase 10/300 GL column (Cytiva) equilibrated with SEC buffer (50 mM Tris–HCl pH 7.5, 200 mM NaCl), and fractions containing pure protein were used directly to set up crystallization experiments.

pLT was produced and purified by using the equivalent procedure.

Production and Purification of hLT

hLT was produced in OverExpress C43 (DE3) cells (Sigma) and recovered from the bacterial periplasmic space following a protocol described earlier. Briefly, cells were grown at 37 °C in Terrific Broth (TB) medium containing 25 μg/mL CAM until OD₆₀₀ reached 2, before induction with 0.2% l-arabinose (Sigma). After 3 h, cells were harvested by centrifugation. The resulting pellet was then resuspended in 1/40^th volume of Talon A buffer (50 mM sodium phosphate pH 8.0, 300 mM NaCl), supplemented with cOmplete protease inhibitor cocktail (Roche), 1 mg/mL polymyxin B sulfate (Sigma) and benzonase (Sigma). This solution was incubated at 37 °C for 15 min with mild shaking and subsequently centrifuged (8000g, 20 min, 4 °C). The supernatant (containing the periplasmic extract) was filtered and loaded to a HiTrap TALON crude 5 mL column (Cytiva) previously equilibrated with Talon A buffer. After the washing step, hLT was eluted with 10 CV Talon B (50 mM sodium phosphate pH 8.0, 300 mM NaCl, 50 mM imidazole). The protein was concentrated and loaded on a HiLoad 16/60 Superdex 200 prep grade column (Cytiva) equilibrated with phosphate-buffered saline (PBS). As two overlapping peaks corresponding to the holotoxin and the B pentamer were obtained, only fractions from the first half of the holotoxin peak were pooled in order to reduce holotoxin contamination by B pentamer.

Holotoxin Cleavage by Trypsin Treatment

Before toxicity and disassembly experiments, the A subunits of the recombinant holotoxins were converted to reduced A1/A2 heterodimers through a 30 min incubation at 4 °C in 10 mM MES (pH 7.4) containing 10 μg/mL of trypsin, 1 mM DTT, and 150 mM NaCl in a 1 mL reaction volume. A trypsin inhibitor (20 μg/mL) was then added to the toxins, which were at 10 μg/mL concentrations.

Purification of Recombinant PDI with an N-Terminal Hexahistidine Tag

E. coli strain BL21­(DE3)­pLysS transformed with the pOLR130 plasmid was grown in 1 L Luria broth with 100 μg/mL of ampicillin to an OD₆₀₀ of 0.6. It was then induced with 1 mM IPTG for overnight incubation at 25 °C. The culture was centrifuged at 8500g for 20 min at 4 °C, and the resulting cell pellet was resuspended in 37 °C lysis buffer (50 mM Tris–HCl, 150 mM NaCl, 1% Triton X-100, 1% deoxycholate, and 200 μg/mL of lysozyme, pH 7.4) before sonication. The supernatant collected from a 30 min 4 °C spin (12,000g) of the cell lysate was added to the TALON resin (5 mL slurry, Clontech) that had been washed three times with extraction buffer (20 mM sodium phosphate, 300 mM NaCl, pH 7.4). After overnight mixing at 4 °C, the PDI-bound resin was washed three times with extraction buffer containing 5 mM imidazole. The resin was then transferred to a gravity column, and PDI was eluted using 2 mL volumes of extraction buffer containing 10, 20, 40, 60, 100, and 300 mM imidazole. SDS-PAGE was used to identify fractions with pure PDI, which were then pooled and dialyzed using a 20 kDa MWCO cassette (Thermo Fisher Scientific) with 1 h and overnight exchanges in 1 L of 10 mM MES buffer (pH 7.4). Following the determination of protein concentration by the BCA protein assay (Thermo Fisher Scientific), 100 or 500 μg protein samples were aliquoted into microcentrifuge tubes, frozen at −80 °C, and after overnight storage lyophilized for long-term storage at −80 °C.

Toxicity Assay

CHO cells grown overnight to ∼80% confluency in 24-well plates were incubated in serum-free medium with various concentrations of CT, CTv1, CTv4, or LT. The cells were then lysed by a 20 min 4 °C incubation in ice-cold acidic ethanol (1:100 HCl/EtOH) and air dried overnight. cAMP levels in buffer-reconstituted cell extracts were quantified with the Direct BioTrack EIA, an ELISA-based kit from GE Healthcare. The basal quantity of cAMP from unintoxicated cells was background-subtracted from all experimental samples, and the resulting values were then expressed as percentages of the maximum cAMP response in the experiment (i.e., challenge with 100 ng/mL of CT).

Toxin Disassembly Monitored by ELISA

Ganglioside GM1 (6 μg/mL in ethanol) was added in 50 μL volumes to the wells of a 96-well plate and allowed to air dry overnight. The dried plate was washed once with 150 μL of phosphate-buffered saline (pH 7.4) containing 0.05% Tween20 (PBST). The toxin (1 μg/mL) was then added to the wells in 100 μL PBST with 2.5% bovine serum albumin (BSA). After a 1 h incubation at 4 °C, the wells were washed four times with PBST and incubated for 1 h at 25 °C with 100 μL of PDI (20 μg/mL) that had been pre-reduced by a 30 min 25 °C incubation in PBST containing 1 mM DTT. The wells were washed four times with PBST after exposure to PDI and incubated for 1 h at 4 °C with a rabbit primary anti-CTA1 antibody (100 μL at 1:1000 dilution in PBST/2.5% BSA). This was followed by four washes with PBST, a 30 min 4 °C incubation with an HRP-conjugated goat anti-rabbit IgG antibody (100 μL at 1:2500 dilution in PBST/2.5% BSA), six washes with PBST, and addition of 100 μL TMB substrate. Stop solution was added 5-10 min after the addition of TMB, and absorbance was measured at 450 nm with a Synergy 2 plate reader (BioTek). Values from GM1-coated wells lacking any toxin were background-subtracted from all experimental samples. The percentage of toxin disassembly was then calculated as [1.00 – (PDI-treated A1 signal/untreated A1 signal)] × 100.

Toxin Disassembly Monitored by SPR

Experiments were performed using an OpenSPR instrument and hydrophobic sensors, both from Nicoya. Injection volumes of 200 μL were used at a flow rate of 20 μL/min for all steps unless otherwise noted. After sensor surface activation with an injection of 40 mM OGP at a flow rate of 150 μL/min, GM1 ganglioside diluted to 150 μg/mL in perfusion buffer (10 mM MES at pH 7.4 with 150 mM NaCl) was perfused over the sensor. The toxin (10 μg/mL) was perfused over the GM1-coated sensor, followed by a blocking injection of 2.5% BSA in PBS. The baseline signal corresponding to the mass of the captured holotoxin was set at 0 resonance units (RU). PDI that had been incubated in perfusion buffer containing 1 mM DTT for 30 min at room temperature was then added to the sensor for 150 s at a flow rate of 40 μL/min for a total of 500 s. A resulting loss of mass from the sensor, indicative of A1 subunit release from the holotoxin, would drop the initial RU signal from zero to a negative value. Subsequent injection of an anti-CTA1 polyclonal antibody at 1:1000 dilution was used to determine whether the A1 subunit was completely or partially removed from the rest of the holotoxin by PDI. A final probe using an anti-CTB polyclonal antibody at a 1:1000 dilution ensured that the toxin B subunit remained on the sensor at the end of the experiment.

Crystallization

CTv1

CTv1 was crystallized by sitting-drop vapor-diffusion in Swissci 96-well 2-drop PS plates (Molecular Dimensions) immediately after purification by SEC, using the fraction containing the highest concentration of pure protein. 0.2 μL of protein at 13 mg/mL was mixed with an equal amount of reservoir solution of the PACT premier screen (Molecular Dimensions) using a crystallization robot (Oryx4, Douglas Instruments) and incubated at 20 °C. Crystals were obtained in condition D9 (200 mM LiCl, 100 mM Tris–HCl pH 8, 20% w/v PEG 6000) over the course of 2 weeks. They were harvested, mounted in loops, cryoprotected in mother liquor supplemented with 20% glycerol, and flash-cooled in liquid nitrogen.

CTv4

SEC fractions containing pure CTv4 were pooled and concentrated by ultrafiltration (4 °C, 4000g) using Amicon Ultra Centrifugal Filter Units 30 kDa MWCO (Merck) and thereafter diluted to a concentration of 6.5 mg/mL by addition of d-galactose-containing buffer (50 mM Tris–HCl pH 7.5, 200 mM NaCl, 300 mM d-galactose) to a final concentration of 50 mM d-galactose. Sitting-drop vapor-diffusion experiments were set up in Swissci 96-well 2-drop PS plates (Molecular Dimensions) using a crystallization robot (Oryx4, Douglas Instruments), mixing 0.3 μL of protein solution with an equal amount of reservoir solution and incubated at 20 °C. Crystals were obtained from a PACT premier screen (Molecular Dimensions) condition C8 (200 mM NH₄Cl, 100 mM HEPES pH 7.0, 20% w/v PEG 6000) over the course of 3 weeks. After cryoprotection in mother liquor supplemented with 20% glycerol, crystals were flash-cooled in liquid nitrogen and stored for X-ray data collection.

X-ray Data Collection and Refinement

Synchrotron data collection was performed at beamlines ID30-B (CTv1) and ID23-2 (CTv4), at the European Synchrotron Radiation Facility (ESRF) in Grenoble, France, at 100 K. Diffraction data for CTv1 were processed with the EDNA_proc autprocessing pipeline, whereas for CTv4, data were processed with XDSAPP. Both structures were determined by molecular replacement with Phaser from the CCP4 suite using the CT wild-type structure (PDB: 8QRE and PDB ID: 1S5E, for CTv1 and CTv4, respectively) as a search model. Refinement was carried out by alternating cycles of manual building in Coot , and refinement with REFMAC5. After improving the protein main chain, water molecules and ions were added in a step-wise fashion. The CTv4 structure contains several d-galactose molecules bound to its B pentamer, while CTv1 binds glycerol or PEG-fragments. All ions and ligands were modeled and refined. Finally, alternative conformations were added, their relative occupancies were refined with phenix.refine, and the structures were deposited in the PDB with PDB IDs 8Q6I (CTv1) and 8OXS (CTv4). Data collection and refinement statistics are summarized in Table .

1. Data Collection and Refinement Statistics.

	CTv1	CTv4
Data Collection
Beamline	ESRF ID30-B	ESRF ID23-2
Wavelength (Å)	0.8731	0.8731
Space group	P21	P21
Molecules in asymmetric unit	1	2
Unit cell parameters
a, b, c (Å)	60.3, 91.1, 72.3	60.3, 89.6, 141.1
β (°)	106.4	98.2
Resolution (Å)	48.8–1.6 (1.66–1.60)	48.9–1.6 (1.62–1.60)
CC 1/2 (%)	99.9 (43.6)	99.2 (16.7)
Completeness (%)	97.1 (92.1)	99.6 (92.8)
Multiplicity	7.1 (7.4)	6.7 (5.9)
No. of unique reflections	95,507 (8864)	195,717 (9026)
(I)/σ(I)	10.8 (0.7)	6.5 (0.6)
Wilson B-factor (Å2)	26.6	22.5
Refinement
R/R free (%)	18.5/22.2	20.7/23.1
No. of atoms
protein	6383	13,700
ligand/ion	83/2	266/2
water molecules	408	764
B-Factors (Å2)
protein	30.4	25.4
ligand	46.4	25.9
water	33.8	29.3
r.m.s.d. bonds (Å)	0.010	0.010
r.m.s.d. angles (°)	1.61	1.45
Ramachandran Plot
favored (%)	98.3	98.0
allowed (%)	1.5	1.8
outliers (%)	0.2	0.2
PDB ID	8Q6I	8OXS

^aValues for the highest-resolution shell are shown in parentheses.

Analysis of Different Crystal Forms and Selection of Toxin References for Comparison

All CT and LT structures published in the PDB are listed in Table S2, with space group and unit cell parameters specified. For wild-type CT, seven structures have been deposited, five with space group P2₁ (PDB ID: 1XTC, 1S5E, 1S5B, 1S5C, and 8QRE ) and two with space group P2₁2₁2₁ (PDB ID: 1S5F and 1S5D ). Of these structures, those with amino acid substitutions (1S5B, 1S5C and 1S5D ) were disregarded as main references for comparison with our models. Of the remaining CT structures, PDB entry 1S5E had the highest resolution (1.9 Å) and excellent geometry. It was therefore selected as main CT reference for structural comparisons.

For LT, all structures considered in the PDB belong to the same space group, P2₁2₁2₁, most of them (except 1LTA ) with similar unit cell parameters (Table S2). The highest resolution LT structures are 1LTS (1.95 Å), 1LT3 (2.0 Å), and 1LT4 (2.0 Å). We excluded entries 1LT4, 1LTG, and 1LTI as main references for structural comparisons since these structures either harbor a substitution or are part of a complex, which may influence the properties of the protein and therefore our conclusions regarding CTv4 and CTv1. We selected PDB entry 1LTS as the main LT reference, which has very good geometry.

Superposition of Structures and Visualization

To identify differences in the positioning of the A subunit, structures were superimposed on their B pentamers. Alignments were performed with the cealign command in PyMOL version 2.1 (Schrödinger, LLC), and root-mean-square difference (r.m.s.d.) values for specific protein regions were calculated with the rms_cur command. Hydrophobicity and electrostatic potentials of the toxin B pentamers were calculated and visualized with UCSF ChimeraX. All other figures were generated using PyMOL (Schrödinger, LLC).

Analysis of A2 Tilt Angles in the Crystal Structures

Angles between the A2 α-helix and the B pentamer plane were calculated using the Helix Orientation (HELO) algorithm developed by Tatulian. This algorithm uses the Cα coordinates to determine the angle of a helical axis in the protein coordinate system. In the current analysis, we determined the orientation (angle δ) of the A2 helix with respect to the normal of the B pentamer plane defined by the Cα atoms of Leu72 of all five B subunits, as described in more detail by Serrano et al. (2022). The procedure was completed for 22 overlapping quadruplets starting at amino acid residue number 200, i.e., residues 200, 201, 202, 203, then 201, 202, 203, 204, ... and finally 221, 222, 223, 224. For each angle δ, the angle θ between the respective helical segment and the B5 plane was calculated as θ = 90° – δ and plotted vs the residue number (θ corresponding to residue i represents the angle for the i to i + 3 stretch, i.e., approximately one helical turn). All 22 values of θ were averaged to find the overall orientation of the A2 helix relative to the plane of the B pentamer. It should be noted that the planes between the A2 quadruplet axes and the B pentamer normal may or may not be coplanar; this cannot be distinguished by HELO and requires 3D examination of the crystal structures.

MD Simulations

MD simulations were carried out on our CTv1 and CTv4 crystal structures, as well as for wild-type CT (PDB: 1S5E) and LT (PDB: 1LTS), using GROMACS version 2021.5 or 2023.5. , Residues that were not modeled in the crystal structures were added manually using weak electron density maps as a reference in Coot. , In addition, we exchanged the 3-methylhistidine modeled at position 131 in the CTv1 structure for a regular histidine residue, as the methylated residue was not recognized as a standard amino acid by GROMACS. Additional buffer components found in the crystal structures like ions, glycerol, and galactose were not considered in the simulations, with the exception of the structural sodium ion present in the A1 subunit of CT and the variants. The models were completed by addition of missing hydrogen atoms and choosing standard protonation at neutral pH for all the ionizable groups. Initial simulations were performed with inactivated toxins, where the A1/A2 subunits were not cleaved (part of the same polypeptide chain) and connected through a disulfide bridge between Cys187–Cys199. To mimic the host ER conditions, simulations were also performed on activated toxins, harboring two separate A1/A2 subunits (cleaved between residues 192–193) and a reduced disulfide bridge.

Each system was solvated in a periodic cubic box (12 × 12 × 12 nm³) filled with water molecules (including crystallographic waters) and neutralized. The salt concentration was adjusted to 150 mM NaCl. The Amber99SB-ILDN was used to parameterize the protein and ions, while water was represented with the TIP3P model. All bonds that included hydrogen atoms were constrained using the LINCS algorithm. Long-range electrostatic interactions were computed with the Particle Mesh Ewald (PME) method, while van der Waals interactions were calculated using the Lennard–Jones function with a cut-off value of 10 Å. Each system was relaxed through a two-step energy minimization process using the steepest descent algorithm (5000 steps with the heavy atoms fixed to minimize the position of the hydrogen atoms only, followed by 5000 additional steps without constraints to minimize the positions of all the atoms). After that, the systems were equilibrated in the isothermal-isobaric (NPT) ensemble using the Berendsen barostat, with a coupling constant of 2 ps and target pressure 1 bar. Temperature equilibration was obtained by randomly setting the velocities of all atoms using the Maxwell–Boltzmann distribution at 10 K and then performing simulated annealing to reach a final temperature of 300 K within 4 ns. After equilibration, all the systems were subjected to a 1 μs production run in the NPT ensemble with a Parrinello–Rahman barostat, using the coordinates and velocities of the last step of the equilibration run.

We ran three independent replicates of each system starting from the same energy-minimized configuration and choosing different random seeds to generate the velocities during the equilibration step. In all MD simulations, the equations of motion were solved by using the leapfrog scheme with a time step of 2 fs. For LT, we ran a total of five replicates since one of the original three replicates showed a significantly different tilt angle compared with the two other replicates.

All simulations were analyzed with GROMACS , and the python package MDAnalysis. Data is available for download at the NIRD repository through the link: https://doi.org/10.11582/2026.aeg2t121.

Analysis of A2 Helix Tilt-Angle and Geometry in MD Simulations

Calculation of the angle between the A2 helix and the B pentamer was performed using MDAnalysis throughout our 1 μs simulations. Unlike the HELO algorithm, which computes the angle between local helix axes (formed by four contiguous residues in the helix) and the B pentamer plane, we performed the angle calculations in our simulations using the global helix axis. This global axis was determined using the module HELANAL , from MDAnalysis. The angle (α) between the global helix vector and the normal vector to the B pentamer plane (defined by the coordinates of Cα atoms of Leu 72 from all five B-subunits) was calculated using their dot product. The tilt angle (θ) was defined as 90° – α.

For the initial angle calculations, we used the Cα coordinates of A2 residues 200–221 to define the global helix axis. However, due to partial unfolding of the helix between residues 200–206 observed in several of our simulations, we repeated the calculations using only residues 207–221, as this region provides a more accurate calculation of the helix vector. In addition, we analyzed the degree of bending of the helix using the function all_bends in HELANAL, which compares the angle between pairs of local helix axes.

Calculation of Water Occupancy in the B Pentamer Pore

To evaluate water occupancy near the top of the B pentamer pore, two probes were defined in this region as described previously. The probes were placed at the midpoint of the triangle formed by the Cα carbon of residue 225 from the A2 subunit with the Cα carbons of I74 from chains F and G (probe 1), or I74 from chains G and H (probe 2). The number of waters within a radius of 2.8 Å from these probes was counted throughout the simulation trajectory using MDAnalysis.

Results

Impact of Amino Acid 229 on Toxin Activity and PDI-Driven Toxin Disassembly

Site-directed mutagenesis was used to generate two variants of CT: CTv1 contains the single substitution D229E, and CTv4 contains four amino acid substitutions that fully convert its A2 subunit to the LTA2 sequence between the critical residues 226–236 (Figure C). Both chimeric toxins as well as wild-type CT and wild-type LT were purified by affinity chromatography. Trypsin treatment converted the A chains of all four toxins to separate A1 and A2 subunits (Figure ). Neither of the CT variants exhibited resistance to trypsin, which is a phenotype that can result from alterations in the A subunit.

2.

Substitutions in the A2 linker do not affect proteolytic nicking of the A subunit into distinct A1 and A2 subunits. CT, CTv1, CTv4, and LT were incubated with trypsin protease (indicated by +) for 30 min at 4 °C before visualization by SDS-PAGE with Coomassie stain. The unnicked 26 kDa A subunit is composed of CTA1 (21 kDa) and CTA2 (5 kDa) subunits. The B pentamer is composed of five identical 11.6 kDa protomers. The mobility of trypsin, a 24 kDa protein, is denoted by an asterisk.

Our recombinant toxins were used in a cell-based toxicity assay that monitored the elevated levels of cAMP resulting from Gsα activation (Figure A). Each toxin elicited a dose-dependent response from cultured CHO cells. A more robust cAMP response was recorded for wild-type CT (circles) than for either CTv1 (closed diamonds), CTv4 (open diamonds), or LT (squares). The latter three toxins all produced comparable effects. Substituting one or four of the CTA2 residues with the corresponding amino acid(s) from LTA2 thus reduced the potency of CT to the level of wild-type LT.

3.

Substitutions in the A2 linker alter the cellular activity and PDI-driven disassembly of CT. (A) CHO cells were incubated with the stated concentrations of CT (circles), CTv1 (closed diamonds), CTv4 (open diamonds), or LT (squares) for 2 h before cAMP levels were quantified. Data were expressed as percentages of the maximal cAMP response for the assay and are presented as the mean ± standard error of six independent experiments with triplicate samples. (B) CT, CTv1, CTv4, and LT were appended to the wells of a GM1-coated ELISA plate before incubation with reduced PDI for 1 h at 25 °C. After extensive washing, sequential incubations with primary and HRP-conjugated secondary antibodies were used to detect the A1 subunit. The percentage of toxin disassembly was calculated using the maximum A1 signal obtained from the corresponding untreated holotoxin. Data from six independent experiments with six replicate wells per condition are presented in the box-and-whisker format. Asterisks indicate statistically significant differences (**p < 0.01; ****p < 0.0001, one-way ANOVA).

A substitution in the non-catalytic A2 subunit could reduce toxin potency by limiting the efficiency of PDI-driven toxin disassembly. We examined this possibility with an established ELISA that monitors the release of the A1 subunit from its plate-anchored A2/B₅ complex (Figure B). Previous work showed that PDI only recognizes the A1 subunit, thus, a substitution in the A2 subunit would not directly affect PDI binding to the toxin. Yet, as shown in Figure B, the D229E substitution in CTv1 reduced the efficiency of its PDI-driven disassembly in comparison to native CT. Disassembly of CTv4 was less efficient than that of CTv1 and comparable to LT.

The results of our ELISA experiments were confirmed with a second assay using SPR spectroscopy (Figure ). In this assay, each toxin was immobilized on a GM1-coated sensor and exposed to PDI. The signal from an SPR sensor is proportional to the mass on the chip. The release of the A1 subunit from the sensor-bound A2/B₅ complex will therefore produce a negative resonance unit (RU) signal when compared to the initial baseline signal representing the mass of the holotoxin. Addition of PDI to the sensor-bound holotoxin resulted in a clear loss of signal for CT, a minor loss of signal for CTv1, and no appreciable loss of signal for CTv4 or LT (Figure A–D). The complete disassembly of wild-type CT was confirmed with sequential injections of CTA1 and CTB antibodies at the end of the experiment: the absence of an A1 antibody signal and presence of a B antibody signal demonstrated that CTA1 had been lost from the remaining sensor-anchored B pentamer. In contrast, an A1 antibody signal could be detected for CTv1, CTv4, and LT. This indicated that at least some intact holotoxin remained bound to the chip after exposure to 1 μM PDI. We have previously shown that a higher PDI concentration of 10 μM will facilitate the complete disassembly of LT, which was also observed here for CTv1 (Figure E) and CTv4 (Figure F). PDI can thus displace the A1 subunit from all four toxins, but its disassembly of CT is much more efficient than the disassembly of CTv1, CTv4, or LT.

4.

Substitutions in the A2 linker reduce the efficiency of toxin disassembly by PDI. Following a baseline measurement corresponding to the mass of the sensor-bound holotoxin (0 RU), PDI (1 μM unless otherwise noted) was added to the perfusion buffer at 0 s. This was followed by sequential injections of antibodies against the toxin A1 and B subunits, as indicated by the arrowheads. Asterisks denote the removal of PDI or antibody from the perfusion buffer. For comparative purposes, traces from separate experiments are overlaid on the same graph. (A) CT vs. LT. (B) LT from panel A vs. two independent runs with CTv1. (C) LT from panel A vs. two independent runs with CTv4. (D) CT from panel A with color-matched traces of CTv1 from panel B and CTv4 from panel C. (E) CT from panel A vs. two independent runs with CTv1 sensors exposed to 10 μM PDI. (F) CT from panel A vs. two independent runs with CTv4 sensors exposed to 10 μM PDI.

Crystal Structures of CTv1 and CTv4

The C-terminal region of the A2 subunit might affect toxin disassembly and potency by influencing the A1 tilt angle within the holotoxin. We accordingly predicted that the A1 tilt angles in CTv1 and CTv4 would be closer to the 40° angle in LT than the 49° angle in CT. To test this prediction, we crystallized and determined the X-ray structures of CTv1 and CTv4 (Table ). Both toxin variants crystallized in space group P2₁, with CTv1 featuring one molecule in the asymmetric unit, while CTv4 has two. As with other crystal structures of the CT holotoxin (e.g., PDB ID: 1S5E and 1S5F ), the toxins were not proteolytically nicked and feature intact A subunits. The structures are of high quality, as judged from the high resolution (1.6 Å in both cases), low R-factors, and good geometry (Table ). They are characterized by well-defined electron density for most of the holotoxin. Only the A subunits have two regions of weak electron density: a flexible loop around the trypsin cleavage site (residues 190–194) and the region close to the C-terminus of the toxins (Figure ). This region showed more flexibility or disorder in CTv1, where no residues could be modeled beyond D231, while for CTv4, clear density was observed until residue R235 or I236 (Figure , insets). The KDEL sequence (residues 237–240) responsible for retaining CT in the ER is not visible in the electron density maps (Figure , insets), which is a common phenomenon for published CT and LT toxin structures.

5.

Structures of CTv1 and CTv4. (A) Overall structure of CTv1. The active A1 subunit is colored teal, the A2-linker light teal, and the B pentamer gray. The purple sphere represents a conserved sodium ion. A2 C-terminal residues are shown in the inset, with their respective 2mF _o –DF _c electron density maps (blue mesh at 1.0 σ). (B) Overall structure of CTv4 with the B pentamer (gray), the active A1 subunit (beige), and the A2-linker (light beige). The A2 C-terminal residues are shown in the inset. To aid visualization, the central helices of only three neighboring B-subunits are displayed. The density of CTv1 extends only to D231 (compared to I236 for CTv4), reflecting disorder of the C-terminal residues. Note that although A1 and A2 subunits are shown in different colors, crystals were obtained from intact holotoxins with uncleaved A subunits.

Comparison of Holotoxin Structures

The overall structures of CTv1 and CTv4 are highly similar to wild-type CT (represented by PDB ID: 1S5E ), with r.m.s.d. values for both CT variants of 1.2 Å (Figure A). The structures are also similar to each other (r.m.s.d. = 0.4 Å) and more distinct from LT (represented by PDB ID: 1LTS ), with respective r.m.s.d. values of 1.8 Å and 2.0 Å (Figure B). The main differences are observed in the orientation of the A subunit “head” (A1 and A2 helixes; Figure A–C), with r.m.s.d values of 2.8 Å and 2.9 Å for CTv1 and CTv4 compared to CT and 4.5 Å and 4.4 Å compared to LT, respectively. The A2 tilt angle is very similar for both CT variants and lies between CT and LT (Figure C), whereas the A2 linker region inside the B pentamer pore closely resembles LT (Figure B inset and Figure ).

6.

Superimposition of CTv1 and CTv4 structures with wild-type CT and LT. (A) Wild-type CT (dark green, PDB: 1S5E ) superimposed with CTv1 (teal, PDB ID: 8Q6I; this work) and CTv4 (beige, PDB ID: 8OXS; this work). CTv1 and CTv4 superimpose very well (r.m.s.d. = 0.4 Å), but their C-terminal regions are distinct from CT. The C-terminal region of the A2 linker, which threads through the B pentamer pore, is enlarged in the inset and is shown in more detail in Figure . The four residues probed by mutagenesis are shown as sticks. The conserved sodium ion present in CT and the variants is displayed as a purple sphere. (B) Wild-type LT (brown, PDB: 1LTS ) superimposed onto CTv1 (teal, PDB ID: 8Q6I) and CTv4 (beige, PDB ID: 8OXS). The A2 C-terminus is enlarged in the inset. (C) Orientation of the A2-helix of CT, CTv1, CTv4, and LT relative to their respective B pentamers, with that of the variants lying in between CT and LT. Left panel: front view displaying the B pentamer of CT in surface representation. Right panel: top view in cartoon representation.

7.

Comparison of A2 C-terminal residues inside the B-pentamer pore. Superimposition of LT (PDB: 1LTS, brown) with (A) CT (PDB: 1S5E, dark green), (B) CTv1 (this work, light teal), and (C) CTv4 (this work, beige). Residues that differ between each pair of toxins are shown in bold. The four residues probed by mutagenesis in this work are underlined (E229 in CTv1; E229, V230, I232, and Y233 in CTv4). In CTv4, the conformation of the A2 peptide in the B pentamer pore is virtually identical for both molecules in the asymmetric unit. The only minor difference arises from the fact that we observe dual side-chain conformations for the A2 peptide in molecule 1 between F223 until S228 (Figure S1). This is also observed in CTv1 (B), but only for residues located close to the pore entry (I222, S224, and Y226). From residue 232, electron density is very weak (Figure A, inset), indicating disorder. In addition to the residues probed by mutagenesis, residue 225 differs between CT and LT (G vs D). G225 falls within the favored and allowed regions of the Ramachandran plot for both CT and the variants.

Detailed Description of Interactions inside the B Pentamer Pore

CT, LT, and the toxin variants exhibit the same amino acids lining the B pentamer pore (Figure S2). Therefore, the interaction with the A2 subunit is exclusively determined by the different A2 residues of the toxins. These interactions are shown in Figure . At the pore entry, the A2 residues of the wild-type toxins and the variants have very similar conformations (Figure ). Interactions in this region are mainly between hydrophobic residues, and no water molecules are observed (Figure B). Differences start to become apparent from residue 225, which corresponds to a glycine in CTv1 and CTv4 (like wild-type CT) and an aspartate in LT. In the variants, G225 is positioned close to I74 (chain G) and T78 (chain G and F), while D225 in LT extends upwards (Figure ), forming a salt bridge with R146 from the A1 subunit. Despite this difference, all subsequent residues of CTv1 and CTv4 in the pore superimpose very well with LT (Figure , r.m.s.d. for Cα atoms are 1.8 Å and 1.6 Å for CTv1 and CTv4, respectively). In all three structures, the polypeptide chain is stretched after Q227, resulting in a register shift of 1.5 helical turns that positions S228 and all subsequent residues deeper in the pore (Figure ). As a result, the A2 residues in this region are found in a very different environment in CTv1 and CTv4 compared to CT (Figure C). For instance, the substituted residue E229 (D229 in CT) is positioned within H-bonding distance to R73 in the B pentamer (chain F) and to Q227 (A2), whereas D229 in CT is located further up in the pore (distant from the membrane in a biological context), where it H-bonds with two water molecules and is surrounded by I74 from chains D, H, and G. At the pore exit, the orientation of the residues is similar for CTv4 and wild-type toxins, although they are located in slightly different environments (Figure D). The ER-targeting KDEL sequence is not defined by electron density in any of the toxins.

8.

Interactions inside the B pentamer pore of CT, CTv1, CTv4, and LT. (A) Structures of CT (dark green, PDB: 1S5E ), CTv1 (teal, PDB ID: 8Q6I; this work), CTv4 (beige, PDB ID: 8OXS; this work) and LT (brown, PDB: 1LTS ), highlighting A2 residues at the pore entry, middle pore section, and pore exit. (B) Interactions at the pore entry. The B pentamer is represented in gray, and residues from different B chains are identified with a superscript except for chain D: chain E = *, chain F = **, chain H = #. Chain G is not shown. In all toxins, F223 is placed in a hydrophobic pocket formed by the side chain of T78 from the B pentamer (chain D) and I222 from the A2 peptide. S224 is stacked between the hydroxymethyl groups of T78 from B pentamer chains E and F and is exposed to the side chain of L77 (chain F). In addition, it H-bonds to the backbone carbonyl of R220. In CTv4, residue 224 has two alternative conformations. Residue 225 in CT and the variants corresponds to a glycine, whereas LT features an aspartate. G225 is positioned within 4 Å of I74 (chain G, not shown) and not far from T78 (chains G and F), while D225 in LT extends upwards, forming a hydrogen bond with R146 from the A1 subunit. In all toxins, Y226 is stacked between the side chains of T78 from two neighboring subunits (D and H). It additionally interacts with F223 and L77 (chain D). (C) Interactions in the central section of the pore. In contrast to the hydrophobic pore entry (Figure S2B), this region is lined mostly by polar residues and harbors multiple water molecules. The conformation of the A2 residues differs significantly between the toxins, as the polypeptide chain in LT and CT variants is stretched, positioning Q227 and all subsequent residues deeper in the pore. The orientation of the A2 residues of CTv1, CTv4, and LT is virtually identical. Q227 adopts two alternative conformations in CTv4 (Figure S1). In its major conformation (like CTv1 and LT), the Q227 side chain points downwards, forming H-bonds with the substituted A2 residue E229 as well as with R73 from chains E and F, one of which is water-mediated. S228 is stacked between the side chains of R73 and D70 (chains H and D). In contrast, Q227 and S228 in CT are located closer to the pore entry, in a hydrophobic environment formed by chains D, E, and F. The substituted residue E229 in CTv1 and CTv4 has the same orientation as in LT. It is positioned within H-bonding distance to R73 in the B pentamer (chain F) and to Q227 (A2). Conversely, D229 in CT points toward chains D and H (opposite to E229 in LT and CT variants), where it H-bonds to two water molecules. Residue 230, corresponding to an isoleucine in CTv1, is exposed to a solvent pocket formed by D70 (chains D and E) and R67 (chain D). In CTv4, this residue is exchanged to valine and is also oriented towards a solvent pocket, which is instead formed by D70 from chains D and H and R235 (as in LT). (D) Interactions at the lower region of the pore. D231 is the last residue visible in CTv1, which we modeled pointing down the pore, close to K63 (chains D and E). In CTv4, D231 stretches towards K63 (chain F), forming a salt bridge, and is also positioned closely to N234 (A2). Residue 232 (Thr in CT vs Ile in CTv4 and LT) H-bonds with a water molecule in CT, whereas I232 in CTv4 and LT is stabilized by the hydrocarbon side chains of K63 (E) and E66 (E and F). H233 (CT) faces chain F, while Y233 (CTv4 and LT) is stacked between K63 from two different B subunits. N234 emerges from the bottom of the pore into the solvent, while R235 folds back into the B pentamer, forming a salt bridge to D70 (chain D). I236 is visible only in CTv4, where it was modeled pointing out of the pore, connected via its main-chain NH back to the I232 carbonyl group.

Interestingly, a network of 10 water molecules with a near 5-fold symmetry is observed above and below E229 in CTv1, CTv4, and LT (Figure A–C). These water molecules form two pentagons: one surrounding R73 from the B pentamer and a second one next to D70. In CT, only this second water pentagon is complete (Figure D), whereas the first one is missing the water molecule adjacent to R73 from chain G (red arrow in Figure A); another water molecule lies outside of the “pentagon plane” (marked by the pink arrow in Figure D). The disrupted water symmetry in CT results from the different orientations of residues Q227 and S228 inside the pore for LT and the toxin variants compared to CT (Figures C and ). In CT, Q227 is positioned higher in the pore, dragging the nearby water molecule out of alignment, while the backbone carbonyl of S228 is positioned less than 2 Å from the water molecule close to chain G in the other structures (red arrow in Figure A) and would potentially create a clash. The missing water molecule is consistently absent in all published CT structures (Figure S3), , while the complete water network is conserved among different LT structures. ,, This network is also observed in crystal structures of the B pentamer alone (from both CT and LT), ,− indicating that the conformation of the A2 linker in CT partially disrupts the 5-fold symmetry of the water molecules inside the pore, whereas the extended conformation of the A2 residues observed in LT and the CT variants does not affect this water arrangement. Water molecules are also found farther down the pore in the holotoxin structures but do not form symmetric pentagon structures as in the middle pore section.

9.

Distribution of water molecules around the central region of the B pentamer pore. Water molecules (green spheres) are symmetrically distributed above and below residue E229 in (A) CTv1, (B) CTv4, and (C) LT, forming two pentagons. (D) In contrast, the CT structure has only nine water molecules in this region, which are less symmetrical. The pink arrow in (D) points at a water molecule outside the pentagon plane, while the red arrow in (A) points at the water molecule H-bonding to R73 from chain G (not shown), which is missing in the CT structure. Here, R73 H-bonds instead to the backbone carbonyl of S228 (not shown). Residues from different B-subunits are identified with a superscript except for chain D: chain E = *, chain F = **, chain H = #. Chain G is not shown.

Geometric HELO Analysis of A2 Tilt Angle

We have shown earlier that in LT, the α-helix of the A2 linker is more steeply inclined toward the B5 pentamer surface than in CT by ∼10°. This results in different positionings of the A1 subunit over the B pentamer in the two toxins, which might explain their differences in PDI-assisted disassembly. To probe our hypothesis, we applied the same geometric analysis to the CTv1 and CTv4 structures (PDB IDs: 8Q6I and 8OXS, respectively) (Figure S4). Average tilt angles of the chimeric toxins (θ = 47° and 48° for CTv1 and CTv4, respectively) lie between the values of the parent toxins (θ = 49° and 40° for CT and LT, respectively), as seen in the structural superimposition shown in Figure C. Because of the small difference between the θ angles of wild-type CT and the two CT variants, the low efficiency of CTv1/CTv4 disassembly by PDI cannot be attributed to the average A2 tilt angle as calculated using the HELO algorithm. We therefore examined the structures of CTv1 and CTv4 for other LT-like properties that could potentially explain the low efficiency of their disassembly.

Analysis of CTv1/CTv4 Dynamic Properties

In addition to their atomic coordinates and molecular interactions, the dynamic properties of biomolecules often crucially determine their functions. We investigated protein dynamics by B-factor analysis (Figure S5) and MD simulations. Specifically, we performed 1 μs MD simulations of CT, LT, and the CTv1 and CTv4 variants using their X-ray crystal structures as starting points, repeating the simulations over multiple independent runs to ensure meaningful statistical analysis. Initially, the simulations were set up with inactivated toxins (i.e., uncleaved A1/A2 subunits), but we did not observe significant differences between the toxins when compared to the crystal structures in terms of their helix tilt angle and interactions inside the B pentamer pore. We also performed principal component analysis (PCA) to filter the collective motions of the proteins from their local (fast) movements, but no consistent trends were observed. Therefore, we initiated a second set of simulations with cleaved and reduced toxins, which is the physiologically relevant state of the toxins upon interaction with PDI in the ER. Our results regarding the analysis of cleaved toxins are described below.

Local Fluctuations and Interactions inside the B Pentamer Pore

To identify protein regions with greater local flexibility, the root mean square fluctuation (RMSF) of Cα atoms was calculated. The A1/A2 subunits in all toxins showed similar patterns (Figure S6), with larger fluctuations on three functionally important loops: the activation loop (residues 25–40), the active-site loop (residues 47–56), and the cleavage site region (residues 185–200). The C-terminus also exhibited higher RMSF values, which is not surprising given that this region contains the flexible ER-retention motif (not visible in any of the crystal structures). The main differences were observed in residues 200–207 of the A2-helix, which displayed higher fluctuations in some replicas of CTv1, CTv4, and LT, whereas this region appeared relatively more rigid in CT (Figure S6). In fact, the A2 helix partially unfolded during these simulations (see Videos S1–S4 and Figure S7), but in CT this was less pronounced and only affected residues 200–204.

The A2 linker inside the B-pentamer pore is another region where LT and the CT variants exhibit higher flexibility than wild-type CT. In LT, the main variations occur in the orientation of S228, E229, V230, and D231, which point in different directions inside the B pentamer throughout the simulation time (see representative Ramachandran plots in Figure S8). In CTv1 and CTv4, these residues not only sample different rotamers but also change their backbone positions during the simulations, adopting a more extended conformation in some of the replicas (Figure S9). As a result, S228 and downstream residues sit deeper in the pore and engage in different interactions with B-pentamer residues than those observed in the crystal structures. In contrast, the positions of the backbone and side chains of most A2 residues in CT remain very similar to the crystal structure and do not vary significantly throughout the simulations, with the exception of N234 and I236 (which show different rotamers) and the C-terminal residues containing the flexible KDEL motif.

Variation of the Helix Tilt Angle over the Simulation Time

In the crystal structures, the helix tilt angle is very similar for CT and the variants (47°–48°), which challenged our previous hypothesis that C-terminal substitutions in the A2 linker affect the positioning of the helix with respect to the B pentamer. Crystal structures, however, only provide a snapshot of the proteins in a single state and do not represent the protein’s dynamic behavior. Therefore, we decided to track the evolution of the A2 tilt angle in MD simulations. Although the angles fluctuate over time (Figure S10), CT and the variants showed average values of approximately 48°, similar to the angles calculated from the crystal structures (Figure S11). In contrast, the angles calculated for LT varied across different replicas and stayed consistently below 45° (Figure S11). While the difference in the tilt angle between CT and LT is only 5°lower than the 9° difference observed in the crystal structuresthe larger variation among replicas in LT indicates that there are differences in the intrinsic flexibility of the toxins, which may provide a better explanation for the differential toxicity than the tilt angle.

Principal Component Analysis

To reduce the dimensionality of the MD data and identify essential motions in the toxins, we performed principal component analysis (PCA). This technique returns a set of vectors called principal components (PCs) that describe the directions of collective atomic motions. The first principal component (PC1) accounts for the largest movement in the trajectory, and each subsequent component captures progressively less movement. In our simulations, the first five principal components (PC1–PC5) were associated with the greatest movements of the toxins (Figure S12A). These PCs represent two main motions: twisting (clockwise or anticlockwise; Videos S5–S8) and lateral movement of the A1/A2 subunits relative to the B pentamer (Figure and Videos S9–S12). The twisting movement was more pronounced in LT and decreased progressively through CTv4, CTv1, and CT. Twisting was observed in all the replicas of LT and the CT variants, but only in two of the three wild-type CT replicas. While the lateral movement of the A subunit in LT was represented by the first two PCs in three replicas (Figure S12C), in CT and the variants these first components were mainly associated with random loop movements. For these systems, twisting and lateral motions were predominantly observed in PC3 and PC4 instead. Both the larger magnitude of the movements in LT and their higher frequency indicate that this toxin has a more flexible “A” subunit head compared to CT and the variants.

10.

Principal component analysis (PCA) of dominant motions in LT, CTv4, CTv1, and CT over the last 500 ns of the simulations. Twisting (clockwise and anticlockwise) and lateral (sideways) movements of the A1/A2 subunit were the main modes contributing to the overall motion of the toxins. The magenta arrows in the porcupine plots indicate the direction of the movement, while their length represents the magnitude. These movements were more pronounced in LT (larger magenta arrows) and decreased progressively in CTv4, CTv1, and CT. In LT, the lateral movement of the A1/A2 “head” was represented by the first eigenvector (PC1) in 2 out of 5 replicas (Figure S12).

Water Permeability inside the B Pentamer Pore

We also analyzed the water occupancy at the top of the B pentamer, which has been identified as an important difference between CT and LT in previous MD simulations of the toxins. Like in the earlier work by Craft et al., we observed that CT excludes water at the top of the pore, whereas LT, CTv1, and CTv4 allow multiple water molecules to enter this region (Figure S13). This is due to the orientation of S228: in CT, this residue is positioned closer to the pore edge, blocking the passage of water, whereas the same residue in LT is found to be deeper in the pore and closer to the central pore axis. Interestingly, the C-terminal A2 substitutions introduced in CTv1 and CTv4 changed S228 into the conformation observed in LT, increasing water permeability at the top of the B pentamer.

Discussion

In this work, we found that the PDI-driven disassembly of CTv1, harboring the single substitution D229E, was less efficient than that of wild-type CT. CTv1 was accordingly less active against cultured CHO cells than was wild-type CT. For both of these events, CTv1 mimicked the general properties of the wild-type LT. Thus, a single substitution in the non-catalytic CTA2 subunit had a substantial impact on toxin disassembly and resulting toxicity. The disassembly of CTv4 was slightly less efficient than the disassembly of CTv1 and was equivalent to the disassembly of LT. Substituting the whole C-terminal region of the CTA2 subunit with the LT sequence (D229E, I230V, T232I, and H233Y) in the critical 11 amino acids spanning residues 226–236 fully converted CT to the LT phenotype.

To understand the structural underpinnings of the modulated activities, we determined the crystal structures of CTv1 and CTv4 (PDB IDs: 8Q6I and 8OXS, respectively). With a resolution of 1.6 Å, the crystal structures of both toxin variants represent the highest resolution holotoxin structures in the Protein Data Bank (PDB). Both structures also have excellent geometry, as judged from the Ramachandran plot and low deviations from ideal geometry (Table ).

Mechanism of Toxin Disassembly by PDI

After delivery to the ER, CT and LT are disassembled in a process that requires the action of PDI. , Only the A1 subunit is recognized by PDI, although the site of contact between the two proteins remains unknown. Structural studies using Fourier-transform infrared spectroscopy have documented the partial unfolding of PDI upon binding to the A1 subunit. We suggested that the expanded hydrodynamic size of unfolded PDI allows it to act as a wedge for toxin disassembly by pushing between the A1 subunit and A2/B₅ complex. In support of this model, conditions that prevent the substrate-induced unfolding of PDI also block its ability to disassemble CT. , PDI also exhibits a remarkable conformational flexibility that allows its a and a′ domains to undergo multiple rounds of unfolding and refolding to the native state, which is consistent with its proposed mechanism of toxin disassembly. This model served as the foundation for our structural analysis of the CTv1 and CTv4 toxins.

Intersubunit Geometry

According to our initial hypothesis, the different C-terminal sequences of the CTA2 and LTA2 subunits induce different tilt angles of the A2-helix relative to the B pentamers. This, in turn, affects their mode of interaction with PDI, efficiency of PDI-driven toxin disassembly and, consequently, cellular toxicity. In preliminary small-angle X-ray scattering (SAXS) experiments (unpublished), we could show that small but significant differences between the conformation of CT and LT also exist in solution. When superimposing the structures of toxin variants determined in this work, we found that the tilt angles of CTv1 and CTv4 lie in between CT and LT (Figure C). We also calculated the A2 helix angles using the HELO algorithm, as previously done for CT and LT, which resulted in average tilt angles of CTv1 and CTv4 of 47°/48°, which is very similar to the 49° angle in wild-type CT. It is worth noting, however, that the HELO algorithm was originally developed for determining the angle between two helices (thus in two dimensions) and therefore has limitations when applied to our 3D system. It should also be noted that crystal structures are influenced by crystal contacts, and care needs to be taken when analyzing these structures to avoid conclusions tainted by crystal artifacts.

When comparing the different published crystal structures of CT and LT, we observed that the orientation of the A2 helix is almost identical in all CT structures (Figure S14), despite crystallizing in different space groups (P2₁ and P2₁2₁2₁; see Table S2). This suggests that the steeper helix is a characteristic of CT, rather than the result of crystal packing artifacts. In contrast, all the LT structures have crystallized in space group P2₁2₁2₁, but their A1 subunits vary slightly in orientation. This may be due to the flexibility of the LTA “head” observed in our MD simulations, allowing the toxin to adapt to different crystal environments. The simulations also showed greater variation in the average helix tilt angle of LT, whereas the angle in CT and variants remained consistent among different replicas.

In summary, we observe a correlation among conformations adopted by the different toxin variants, PDI disassembly, and toxicity from CT over CTv1 and CTv4 to LT. However, the overall structural differences are smaller than those of the phenotypic changes.

Consideration of Potential Factors Affecting Toxin Structure/Function

Before examining the structures of CTv1 and CTv4 for additional LT-like characteristics that could impede their disassembly by PDI, we considered whether other factors could influence our structural comparisons. We considered, for example, the possibility that some CT and LT crystal structures could be nicked between the A1 and A2 subunits. Except for LT structure 1LTB, there is no evidence of this (Table S2). Our unnicked CTv1 and CTv4 toxins thus provide a direct comparison to the published crystal structures of unnicked CT (1S5E ) and LT (1LTS ).

We further noticed that all published LT structures in the PDB represent LT infecting pigs (pLT), whereas our toxicity and disassembly assays were done with LT from human ETEC isolates (hLT). Both LT variants have highly similar sequences (98% identity), with the A subunits only exhibiting three differences in A1 residue 4 (K in hLT, R in pLT), A2 residue 213 (K in hLT, E in pLT), and A2 residue 238 in the C-terminal ER-targeting sequence (RNEL for hLT and RDEL for pLT). Previous work demonstrated that the K4R substitution in pLT greatly reduces its cellular activity through an effect on the catalytic site, but we confirmed that PDI disassembles both pLT and hLT with similar kinetics (Figure S15). It should therefore be valid to draw conclusions based on comparisons between crystallographic studies of pLT and disassembly or toxicity assays using hLT.

CT and LT exploit the retrograde trafficking of their GM1 receptors to reach the ER. The toxin is then retained in the ER through an interaction between the host KDEL receptor and the C-terminal tetrapeptide of the A2 subunit. The C-termini of pLT (RDEL), hLT (RNEL), and CT (KDEL) each contain a different yet functional tetrapeptide motif for the KDEL receptor. Previous works concluded that the C-terminal motif does not contribute to the differential toxicity of CT versus LT, comparing KDEL with RDEL. Toxin potency is reduced when the motif is removed from CT and hLT, yet CT remains more toxic than hLT under these conditions. CT, CTv1, and CTv4 share the same KDEL motif but differ in disassembly and toxicity. From these collective observations, we conclude that variations in the C-terminal tetrapeptide sequence do not impact PDI-driven toxin disassembly or toxin potency.

It should also be noted that the residues lining the B pentamer pores of CT and LT are fully conserved (Figure S2), along with an overall sequence identity of approximately 80%. This means that differences in the structures of CT, CTv1, CTv4, and LT cannot be caused by variations of the pore but most likely result from the A2 amino acid substitution(s) in CTv1 and CTv4.

Comparison of Holotoxin Structures

In CTv1 and CTv4, the A2 peptide in the B pentamer pore exhibits essentially the same conformation as LT (Figure B,C), despite the two additional amino acid differences at positions 225 (glycine in CT and the variants versus aspartate in LT) and 237 (ER retention motif), which were not probed in our work. Until S224, all structures superimpose quite well (r.m.s.d. = 0.7 Å between CT, CTv1, and CTv4; 2.1 Å between LT and CT for residues 222-224), but further down, wild-type CT diverges considerably (r.m.s.d = 3.9 Å for residues 225-236 between CT and LT; see Figure ).

Interestingly, one single substitution (D229E) was sufficient to induce a register shift of 1.5 helical turns that converts the A2 linker into the extended conformation observed in LT (Figure ). To what extent the register shift is triggered by the methylene extension in the longer E229 vs D229 is difficult to judge. D229 keeps a significant distance from the residues lining the B-pentamer pore. There is enough space to accommodate a longer residue in this position, but instead, the A2 linker adopts an extended conformation. Another aspect to consider is that the longer side chain of glutamate makes it more flexible than that of aspartate. This may enable glutamate to adopt different conformations within the pentamer and increase the flexibility in the A2 linker. Although we did not observe alternative conformations for E229 in the CT variants, several A2 residues adopt dual conformations in both CTv1 (I222, S224, and Y226) and CTv4 (F223-228). We also observe higher B factors in this region (Figure S5), indicating that the substitution could indeed increase the flexibility of the A2 linker. This was supported by our MD simulations, where the A2 residues inside the pore of the variants showed more variable conformations, especially after Q227 (Figures S6 and S9), whereas these residues in CT showed less conformational flexibility (except for the terminal KDEL motif, which is flexible in all of the toxins).

Interactions of A1 Subunit

While the focus of this article is on the A2 subunit, we should not forget that the subunit to be displaced for toxin activation is the A1 subunit. The interactions between A1 and the B pentamer are described in Figure S16. In addition, the A1 subunit interacts with the A2 helix along its entire length in both CT and LT, as described in detail for LT by Sixma et al. However, it is difficult to judge how the interactions between the A1/A2 subunits and the B pentamer in the crystal structures affect toxin disassembly, as these are flexible proteins, and dynamics can play a role. Moreover, to induce changes in toxicity, it was sufficient to substitute residues in the A2 linker inside the B-pentamer pore without changing interactions with A1, as observed by us and others. The interactions with the A1 subunit remained essentially unchanged (Figure S16). Therefore, the main focus of our analysis concerns the structure and induced dynamics of A2.

Structural Dynamics

An analysis of the dynamic properties of the holotoxins, by comparing the normalized B-factors, showed that CTv1 and CTv4 have three residues (Y226, Q227, and S228) with very high B-factors at the entrance of the pore, close to the membrane interface (Figure S5). In one molecule of the CTv4 structure, residues 223–228 even adopt dual conformations. Because Y226 and Q227 are located just before the change of conformation of the C-terminus (S228), this could indicate that both toxin variants are forced into a new structure by the shared D229E substitution, increasing the mobility of the two amino acids N-terminal to this residue to be able to adopt the new linker fold. The trigger for this structural adaptation may be the first substitution within the A2 linker, D229E, which due to its longer side chain leads to steric constraints. In addition, the mobility differences in this structural region are likely influenced by the one-residue difference between CT (and the variants) and LT at position 225 (G to D). In LT, D225 is fixed in a bidentate salt bridge to R146 of the A1 subunit (a residue conserved between CT and LT), just where the A2 helix enters the pore (Figure S16), contributing to stabilization. We did not specifically alter residue 225 in our analysis because the Hirst lab previously reported that the difference in CT vs LT toxicity could be mapped to amino acid residues 226–236. Our work corroborates the Hirst study and further demonstrates that a single D229E substitution in CTA2 converts CT to an LT-like phenotype in terms of its toxin structural dynamics, disassembly, and potency. Future studies could examine whether residue 225 further influences the properties of CT and LT.

Our MD simulations showed that the flexibility of the A subunit is higher for LT and decreases progressively from CTv4 over CTv1 to CT. In a previous study, PDI has been suggested to act as a wedge to dislodge the A1 subunit from the holotoxin. If the A subunit is flexible and can bend/tilt, it is likely more difficult for PDI to bind to the toxin than if binding occurred to a more rigid structure. As the “A” head in LT and the CT variants is more flexible, PDI might also struggle to achieve sufficient leverage, because the A1 subunit might be moving away from PDI as it approaches the toxin. This may explain why the efficiency in toxin disassembly is lower for LT, CTv1 and CTv4 than for CT (Figure ).

11.

Proposed model of differential toxicity. (A) A flexible A1 subunit reduces the efficiency of toxin disassembly, as PDI may struggle to bind and/or achieve sufficient leverage to dislodge the toxin, resulting in lower toxicity. (B) In contrast, a rigid A1 subunit may improve PDI binding and provide a better handle for toxin disassembly, thereby increasing toxicity.

Previous MD simulations of CT and LT found that CT excludes water from the top of the B pentamer, whereas LT allows multiple water molecules to enter. This difference in water permeability was proposed to increase the stability of CT and result in toxicity. The MD simulations performed in our current work confirmed the marked difference in water occupancy between the CT and LT. Interestingly, the toxin variants also allowed the entrance of water molecules at the top of the pore, although the occupancy was lower than that for LT. This is not surprising given that the A2 linker inside the B pentamer pore of the variants exhibits the same extended conformation observed in LT (Figure ). In these toxins, S228 is positioned deeper in the pore and closer to the central pore axis, whereas in CT, S228 sticks towards the pore edge and is found higher up, blocking water access at the pore entry (Figure ). It is possible that the higher water occupancy disrupts interactions between the A2-linker and the B pentamer, increasing the flexibility of the A1/A2 subunits.

These observations strengthen the evidence that mobility changes induced by substitutions in the A2 pore peptide are the main cause for the differences in PDI-mediated disassemblyfor instance, by allowing (or blocking) access of PDI to the toxin surface or by providing a better (or worse) binding target and/or handle for toxin disassembly.

Conclusion

A single amino acid substitution, D229E, is sufficient to convert CT into a toxin that resembles LT with respect to disassembly and toxin activation. Remarkably, this residue resides in the middle of the B pentamer pore, 15 Å from CTA1. Substituting Asp (D) with the longer Glu (E) positions the C-terminal residues further down in the pore, converting the conformation of A2 to that of LT. While the effect on the A2-helix tilt angle is small, CT to LT substitution results in more water molecules entering the pore at the A1/B pentamer interface and concomitant increased flexibility. The crystal structure analysis is corroborated by MD simulations. Flexibility is found to be especially pronounced at the A1/A2 cleavage site. The increased flexibility of the A1/A2 subunits in LT and the CT variants may make it harder for PDI to bind the toxins and achieve sufficient leverage to displace the A1 subunit, compared to CT. This can explain why CT is more toxic than LT.

Glossary

Abbreviations

ADP

adenosine diphosphate

BCA

bicinchoninic acid

CAM

chloramphenicol

cAMP

cyclic adenosine monophosphate

CT

cholera toxin

CTA1

cholera toxin subunit A1

CTA2

cholera toxin subunit A2

CTB

cholera toxin B pentamer

CV

column volumes

ELISA

enzyme-linked immunosorbent assay

ER

endoplasmic reticulum

GM1

monosialotetrahexosylganglioside

Gsα

alpha subunit of the heterotrimeric G protein

IPTG

isopropyl β-d-1-thiogalactopyranoside

LT

heat-labile toxin

LTA1

heat-labile toxin subunit A1

LTA2

heat-labile toxin subunit A2

LTB

heat-labile toxin B pentamer

MES

2-(N-morpholino)­ethanesulfonic acid

MD

molecular dynamics

OD

optical density

PCA

principal component analysis

PDB

Protein Data Bank

PDI

protein disulfide isomerase

PEG

polyethylene glycol

r.m.s.d.

root mean square difference

RMSF

root mean square fluctuation

SAXS

small-angle X-ray scattering

SEC

size-exclusion chromatography

SPR

surface plasmon resonance

TB

terrific broth

Tris

tris­(hydroxymethyl)­aminomethane

All data will be made available upon reasonable request. MD simulation data are freely available for download at the NIRD repository. https://doi.org/10.11582/2026.aeg2t121.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.6c07022.

• Superimposition of C-terminal A2 linker residues of the two molecules in the asymmetric unit of CTv4 (Figure S1); Conservation of residues lining the B pentamer pore (Figure S2); Conservation of pore water molecules in pentagon arrangement (Figure S3); Analytical geometric analysis of CT/LT chimera compared to wild-type toxins (Figure S4); Probing protein mobility of toxin A subunits by normalized B-factor analysis (Figure S5); Effect of toxin cleavage on MD simulations (Figure S6); Unfolding of the A2 helix near the cleavage site (Figure S7); Ramachandran plots of residues 228–231 over 1 μs MD trajectories (Figure S8); Snapshots of A2 residues inside the B pentamer pore during MD simulations (Figure S9); Variation of the helix tilt angle over 1 μs MD simulations (Figure S10); Probability distribution plots of the tilt angle throughout the 1 μs MD simulations (Figure S11); Principal component analysis (PCA) of cleaved toxins (Figure S12); Water occupancy near the top of the B pentamer pore (Figure S13); Orientation of the A2 helix in published CT structures (Figure S14); Inefficient disassembly of both hLT and pLT by PDI (Figure S15); Residues at the interface of A1 subunit and B pentamer (Figure S16); DNA oligonucleotides (Table S1); Analyzed CT and LT crystal structures from PDB (Table S2)­(PDF)

• Video S1. 1 μs molecular dynamics (MD) simulation trajectory of CT (MP4)

• Video S2. 1 μs molecular dynamics (MD) simulation trajectory of CTv1 (MOV)

• Video S3. 1 μs molecular dynamics (MD) simulation trajectory of CTv4 (MOV)

• Video S4. 1 μs molecular dynamics (MD) simulation trajectory of LT (MOV)

• Video S5. Principal component 3 (PC3) for LT: twisting of the A1/A2 subunits (MP4)

• Video S6. Principal component 4 (PC4) for CTv4: twisting of the A1/A2 subunits (MP4)

• Video S7. Principal component 3 (PC3) for CTv1: twisting of the A1/A2 subunits (MP4)

• Video S8. Principal component 3 (PC3) for CT: twisting of the A1/A2 subunits (MOV)

• Video S9. Principal component 1 (PC1) for LT: lateral movement of the A1/A2 subunits (MOV)

• Video 10. Principal component 5 (PC5) for CTv4: lateral movement of the A1/A2 subunits (MOV)

• Video S11. Principal component 2 (PC2) for CTv1: lateral movement of the A1/A2 subunits (MOV)

• Video S12. Principal component 4 (PC4) for CT: lateral movement of the A1/A2 subunits (MOV)

#.

Viral Pathogenesis Section, Laboratory of Viral Diseases, National Institutes of Health, Bethesda, MD, United States

△.

F.K. and N.M. contributed equally to the work. K.T. conceived the study, and J.B.H., N.M., and U.K. additionally contributed to designing the experiments. A.S. and J.L.G. performed the toxicity assays and SPR analysis, supervised by K.T. F.K., J.B.H. and N.M. performed the experiments related to gene expression and protein purification to X-ray structure determination and refinement, supervised by G.C. and U.K., who also validated the crystal structures (F.K. and J.B.H. additionally contributed to crystallization and data collection). A.S. performed analytic geometry computations, supervised by S.A.T., and N.M. performed additional calculations. N.M. ran and analyzed all MD simulations, supervised by MC, with additional contributions by H.M.C. F.K., U.K. and K.T. wrote the original draft of the manuscript; A.S. and S.A.T. contributed the description of the HELO analysis. After a major revision by N.M. and U.K., the manuscript was finalized by all authors.

The authors declare no competing financial interest.