Constant-pH MD simulations of the protonation-triggered conformational switching in diphtheria toxin translocation domain

Abstract

Protonation of key residues in the diphtheria toxin translocation (T)-domain triggered by endosomal acidification is critical for inducing a series of conformational transitions critical for the cellular entry of the toxin. Previous experiments revealed the importance of histidine residues in modulating pH-dependent transitions. They suggested the presence of a “safety latch” preventing premature refolding of the T-domain by a yet poorly understood mechanism. Here, we used constant-pH molecular dynamics simulations to systematically investigate the protonation sequence in the wild-type T-domain and the following mutants: H223Q, H257Q, E259Q, and H223Q/H257Q. Comparison of these computational results with previous experimental data on T-domain stability and activity with the H-to-Q replacements confirms the role of H223 ($\mathrm{p}{K}_{\mathrm{a}}$ = 6.5) in delaying the protonation of the main trigger, H257 ($\mathrm{p}{K}_{\mathrm{a}}$ = 2.2 in the WT and $\mathrm{p}{K}_{\mathrm{a}}$ = 4.9 in H223Q). Our calculations also reveal a very low $\mathrm{p}{K}_{\mathrm{a}}$ for a neighboring acidic residue E259, which does not get protonated even during simulations at pH 3. This residue also contributes to the formation of the safety latch, with the $\mathrm{p}{K}_{\mathrm{a}}$ of H257 increasing from 2.2 to 5.1 upon E259Q replacement. In contrast, the latter replacement has virtually no effect on the protonation of the H223. Thus, we conclude that the interplay of the protonation in the H223/H257/E259 triad has evolved to prevent triggering the accidental refolding of the T-domain by a fluctuation in the protonation of the main trigger at neutral pH, before the incorporation of the toxin inside the endosome. Subsequent acidification of the endosome overcomes the safety latch and triggers conformational switching via repulsion of H223⁺ and H257⁺. This protonation/conformation relationship corroborates experimental findings and offers a detailed stepwise molecular description of the transition mechanism, which can be instrumental in optimizing the potential applications of the T-domain for targeted delivery of therapies to tumors and other diseased acidic tissues.

Graphical abstract

Significance

Changes in the protonation states of amino acid residues are often involved in triggering physiologically important conformational transitions in proteins. A representative example of such conformational switching involves acid-induced refolding of the translocation domain of the diphtheria toxin, which is responsible for cellular infection. Here, we have applied constant-pH molecular dynamics simulations to decipher the complex molecular mechanisms of protonation-structure coupling in the wild-type translocation domain and several of its mutants exhibiting variable pH-dependent activity. Our results 1) provide a proof-of-concept for the application of such calculations to conformational switching and 2) are instrumental for the future development of molecular platforms to selectively target tumors and other diseased acidic tissues.

Introduction

The protonation of amino acid side chains in proteins is an important process that couples their conformational space with changes in the local environment pH and subsequently modulates several cellular functions (1,2,3,4,5,6). Such proton-triggered conformational switching is often utilized by bacterial toxins and viruses entering the cell via the endosomal pathway (1,2,7,8). Following the receptor-mediated endocytosis, the translocation of active parts of toxins and viruses into the cytosol occurs as a direct result of acidification of the endosomal environment triggering conformational changes in the translocation protein machinery. One of the paramount examples of the process is the pH-triggered refolding of the translocation domain of the diphtheria toxin (T-domain), leading to a multistep conformational change from a water-soluble to membrane-competent and membrane-inserted states, delivering the catalytic domain of the toxin into the cell (4,5,6,9,10,11,12). Being a natural delivery mechanism at acidic pH, the interest in this protein has pushed research toward tailoring its application to drug delivery in acidic environments, such as the microenvironment of late and aggressive tumors (13).

Understanding the mechanism of action of the diphtheria toxin T-domain is of great importance, as it provides an important reference point to cellular entrees of other bacterial toxins and colicins (2,14,15), to the general principles of translocation across the bilayers (16), and the action of other self-inserting proteins, such as Bcl-2 apoptotic regulators (17,18). Consequently, the various aspects of the T-domain functioning and folding/insertion were subject to numerous studies that are reviewed in (8,12). Although the complex process of translocation is not fully understood on a mechanistic level (19), several studies have used various mutagenesis strategies to address the nature of triggering the process via protonation (11,20,21). Two distinct residue groups are usually implicated in pH-dependent transitions inside endosomes: 1) acidic residues, which will lose a negative charge upon acidification of the endosome milieu, and 2) histidines, which will gain a positive charge. The six histidine residues of the T-domain can be further subdivided into two groups: 1) N-terminal H223, H251, and H257, responsible for the early stages of the conformational switching occurring in solution (6,22), and 2) C-terminal H322, H323, and H372 involved on the later stages of the translocation process (23). Specifically, a safety-latch hypothesis has been suggested, which states that the protonation of H257 acts as a primary trigger for the initial transition in solution, while easily protonatable H223 acts as a safety latch, preventing premature unfolding in solution (22,24). This hypothesis was further explored in our previous study, which applied various experimental techniques, such as x-ray crystallography and fluorescence, CD, and NMR spectroscopy, alongside the computational technique of the constant-pH molecular dynamics (CpHMD) simulations (6). These initial CpHMD computations suggested an intriguing possibility that another residue, acidic E259 with an unusually low $\mathrm{p}{K}_{\mathrm{a}}$ may play an important role alongside H223 and H257. In this work, we have systematically explored the molecular details of the protonation of all titratable residues in both wild-type (WT) T-domain and several of its mutants. Our CpHMD results reveal that the protonation/conformation interplay within the H223-H257-E259 triad is pivotal for modulating the early stages of the conformational transition in diphtheria toxin T-domain triggered by the acidification.

Materials and methods

System setup

The T-domain system used in this work was based on our previous work (6), with the x-ray crystallography structure (PDB: 1F0L) inserted in a dodecaedric box and solvated with 8404 SPC (25) water molecules. In this work, we have produced four mutants, changing key residues to glutamines using PyMOL (26), resulting in H223Q, H257Q, E259Q, and the double mutant H223Q/H257Q. All five systems (WT + four mutants) were simulated with GROMACS 5.1.5 (27) and the GROMOS 54A7 force field (28). All systems were energy minimized using the steepest descent algorithm and initialized in three steps: first 100 ps MD to generate velocities using the v-rescale thermostat (29) to set the temperature of the system to 310 K. Position restraints on the protein were applied with a force of 1000 kJ mol⁻¹ nm⁻². A second step of 200 ps initialized the pressure of the system with the Parrinello-Rahman barostat (30) at 1 bar and compressibility of 4.5 10⁻⁵ bar⁻¹ with the same position restraints. A final initialization step of 200 ps was performed reducing the position restraints to 10 kJ mol⁻¹ nm⁻² to allow the protein structure to relax.

CpHMD settings

The stochastic titration CpHMD method (31,32,33) was used to produce three replicate simulations of all systems at four different pH values: 3.0, 4.5, 5.5, and 7.0, ranging from highly acidic to neutral conditions. For the WT system, we used as a starting point the 150 ns simulated in our previous work (6) and extended them to 500 ns. For the four mutant systems, we ran the simulations for 200 ns. Both N/C-termini and all aspartic, glutamic, and histidine residues were allowed to titrate during the simulation time. The molecular mechanics (MM)/molecular dynamics (MD) runs of the CpHMD method used the same temperature and pressure settings as the final initialization step and with a time step of 2 fs. The particle mesh Ewald (34) was used to treat long-range electrostatic interactions with a Verlet scheme and a single cutoff of 1.4 nm for both Coulomb and van der Waals interactions (35). As a consequence of using particle mesh Ewald, our simulation box is required to be near neutrality (36). Hence, 20 ns preruns were performed at each pH value to estimate the charge of all systems and adjust the number of ions to ensure that the charge fluctuations of every simulation box are around zero (37). In this work and due to limitations in the calibration of the CpHMD framework, we used unscreened charges for charged residues and ions, instead of scaling them by a dielectric constant (e.g., 2), reflecting the electronic polarization effects that are missing in the G54A7 fixed-charge scheme (38,39). The protein bonds were constrained with p-LINCS (40), and the water molecules were constrained using the SETTLE algorithm (41).

The CpHMD method estimates the protonation of each residue using a combination of Poisson-Boltzmann (PB) calculations and Monte Carlo (MC) simulations. The PB/MC calculations occur every 20 ps of MD simulations. Although this frequency of protonation attempts is high for most (higher) pH values, we have previously shown that the protonation sampling distribution is not affected (31,32). PB calculations were performed using the Delphi v.5.1 software (42,43) where the atomic charges are obtained directly from the used force field and the atom radii were derived from Lennard-Jones parameters of each atom type at 2 RT (44). To perform this calculation, the dielectric constant was set to 2 and 80, for solute and solvent, respectively. The ionic strength of the system was set to 0.1 M with an ion exclusion layer of 2 Å and the molecular surface of the protein was obtained with a probe of 1.4 Å. The calculation is done in two steps, first PB is calculated on a coarse grid of 1 Å, followed by a focusing second step using a small grid with 0.25 Å. The electrostatic potential was calculated in all three dimensions with relaxation factor values of 0.2 and 0.75 for the linear and nonlinear iterations, respectively.

MC simulations were used to sample the final protonation states from the PB-derived free energies. A total of 10⁵ MC cycles were performed with PETIT with the inclusion of proton tautomerism in the titratable residues (45,46). Protonation changes were attempted following the Metropolis criterion (47) for individual or pairs of sites with interactions larger than 2 $\mathrm{p}K$ units.

System analysis

System equilibration was followed using tools from the GROMACS software package to extract root mean square deviation (RMSD), radius of gyration, and secondary structure content. The $\mathrm{p}{K}_{\mathrm{a}}$ values were calculated for all titratable sites from data fitting to Hill curves, and the $\mathrm{p}{K}_{\mathrm{a}}$ errors were estimated with the standard error of the mean from the three replicates. The distance landscapes were calculated with the GROMACS “mindist” tool and the density scatterplots were calculated approximating our sampling to a Gaussian distribution using a mesh size of 0.015 nm, the probability density values were then converted into conditional free energies with:

$$E\left(\mathrm{i}\right)=-RT\ln \frac{P\left(\mathrm{i}\right)}{P_{\max }}$$ (Equation 1)

where $P_{\max }$ is the maximum of the probability density function $P_{\mathrm{i}}$ (48). This maximum probability density was arbitrarily assigned with zero energy.

Results and discussion

Using advanced computational techniques, such as CpHMD simulations, we obtained a detailed molecular description of the T-domain system at different pH values and aimed to capture the pH-dependent conformational transition mechanism of this protein. To assess the equilibration and convergence of our simulations, we followed several common structural properties such as RMSD, radius of gyration, and secondary structure (Figs. S1–S4), including the protonation states and distances between several key residues (Figs. S5–S9). All simulations equilibrate relatively fast and, to accommodate those that take slightly longer, we have discarded the initial 50 ns of all simulations.

In our previous work (6), we had already studied the protonation profiles of the large array of titrable amino acids present in the WT T-domain of the diphtheria toxin (Fig. 1, A and B) at pH 7, 4.5, and 3.

Figure 1.

Protonation study of the diphtheria toxin T-domain WT using CpHMD simulations. The crystallographic structure of the T-domain (PDB: 1F0L) is shown as a backbone ribbon with histidines (A) and acidic residues (B) highlighted in sticks. The function of the T-domain is to deliver the catalytic part of the toxin across the membrane in response to the acidification of the endosome’s local environment. In our previous paper, we performed a series of 150 ns CpHMD simulations at pH 7, 4.5, and 3 to gain structural and thermodynamic insights into the early stages of acid-induced conformational switching (6). To these, we have now added a new pH value (5.5) and extended the simulation time to 500 ns. These simulations allowed us to estimate the $\mathrm{p}{K}_{\mathrm{a}}$ values of all titratable residues as illustrated in (C) for all six histidines, and in (D) for a selected group of acidic residues (all data are color coded with the labels on the graphs). The full set of $\mathrm{p}{K}_{\mathrm{a}}$ values is presented in Table S1. All points have error bars estimated through the standard error of the mean, although some are too small to be noticeable. The wide range of $\mathrm{p}{K}_{\mathrm{a}}$ values revealed by the CpHMD simulations supports the experiment-based suggestion of the complex and overlapping roles of individual residues in initiating the pH-dependent transition in the T-domain (22), including the “safety latch” hypothesis (24).

In the present work, we have complemented the previous simulations with an extra pH value (5.5) to improve our $\mathrm{p}{K}_{\mathrm{a}}$ predictions (Fig. 1, C and D), and extended them until 500 ns, attempting to observe any conformational transition on this system. From our $\mathrm{p}{K}_{\mathrm{a}}$ analysis, we captured a stepwise protonation process of the histidines near the T-domain’s N-terminus, starting from H223, then H251, and finally H257. In this mechanism, H257 is the final trigger that promotes the conformational transition to the membrane-competent state. The $\mathrm{p}{K}_{\mathrm{a}}$ curves also indicate the presence of a safety latch involving residues H223 and E259, which was proposed in previous works (6,24) to be responsible for the shifted $\mathrm{p}{K}_{\mathrm{a}}$ of H257.

The complex interplay between H223, H257, and E259 provides an acidity-induced trigger for the T-domain to transition into the desired membrane-competent conformations where the safety latch protects H257 from protonation until the medium acidifies to a critical pH value. Below this pH threshold, H257 finally protonates triggering the conformational transition. To test this hypothesis, we performed CpHMD simulations on several mutated combinations of these residues, some that should abolish the safety latch, namely H223Q and E259Q, and others that targeted directly H257, namely H257Q and the H223Q/H257Q double mutant.

An initial analysis of the $\mathrm{p}{K}_{\mathrm{a}}$ values calculated from our simulations showed that no significant shifts were induced on residues away from the N-terminus region described above (Table S1). This agrees with the model where residues located in the C-termini (H322, H323, and H372), which are relatively well solvated, do not contribute to the initial steps of the conformational transition.

To analyze the mutations’ impact, we focused on the N-terminus pocket where the key residue triad (H223, H257, and E259) are located (Fig. 2 A).

Figure 2.

Examining the molecular mechanism of the “safety latch” hypothesis (24) using CpHMD simulations of the T-domain WT and several of its mutants. (A) Structural representation of the T-domain with the three residues of the putative safety latch triad highlighted: H223 (red), H257 (blue), and E259 (olive). The pH-dependent protonation of each residue of the triad is plotted separately for H223 (B), H257 (C), and E259 (D). The protonation was examined in the context of T-domain WT (black in all three panels) and the following mutants: H223Q (red in C and D), H257Q (blue in B and D), E259Q (olive in B and C), and H223Q/H257Q double mutant (olive in D). For clarity, the water molecules were omitted from the figure in (A). All points in (B)–(D) have error bars estimated through the standard error of the mean, although some are too small to be noticeable. The demonstrated interplay of the protonation of the triad, uncoupled by mutations, advances our understanding of the modulation of the pH-triggered conformational switching (details in text).

As expected, the $\mathrm{p}{K}_{\mathrm{a}}$ values are only significantly affected when mutations involve charge changes in that pH range. This is illustrated by the fact that the H223 $\mathrm{p}{K}_{\mathrm{a}}$ value is unaffected by the H257Q mutation (Fig. 2 B), while the H257 $\mathrm{p}{K}_{\mathrm{a}}$ value is very sensitive to either the H223Q or the E259Q mutation (Fig. 2 C). The interaction between charged H223 and E259 is the essence of the safety latch and mutations that target and abolish it (H223Q and E259Q) lead to the recovery of the H257 $\mathrm{p}{K}_{\mathrm{a}}$ value to the unperturbed range. As a direct consequence of this lack of protection in H257 protonation, we expect the conformational transition to happen more freely. This has been observed in previous experimental works where the H223Q mutation was produced and membrane penetration for this mutant was achieved at a higher pH value when compared with WT (24). When considering the effect of mutations on E259 (Fig. 2 D), we noted that only by removing its safety latch partner (H223), both on H223Q and the double mutant H223Q/H257Q, we see a recovery of its $\mathrm{p}{K}_{\mathrm{a}}$ value. Without the ionic interaction with this positive histidine residue, the glutamate turns to the solvent and its $\mathrm{p}{K}_{\mathrm{a}}$ value increases.

After presenting the data illustrating the key role of H223 and E259 in forming the safety latch in the T-domain, we noted that the effects of replacing H223 and E259 with an uncharged glutamine residue produced a similar effect on the $\mathrm{p}{K}_{\mathrm{a}}$ of H257. This implies a similarity in the safety latch perturbation of H223Q and E259Q mutants, which needs to be further examined experimentally.

In several of our simulations, either promoted by the low pH or by the absence of the safety latch, we have the conditions to observe the initial steps of the T-domain conformational transition mechanism (Fig. 3).

Figure 3.

Structural changes observed in the T-domain WT and its mutants with the disrupted pH trigger and safety latch. The overall helicity (A) undergoes only small changes, mostly localized in the TH2 helix (B). The average helicity values for each variant and pH are shown as boxes and the individual replicate average values are shown as small points. The large deviation between replicates of TH2 helicity in pH 3.0 of WT and the double mutant is explained by the loss of helicity in only one replicate (Fig. S4). To illustrate the structural differences in the unfolding of TH2, structures are shown at 200 ns simulation of WT, H223Q/H257Q, and E259Q mutants (from left to right) at pH 7 (C–E) and pH 3 (F–H). For clarity, the water molecules were omitted from the figures. These structures illustrate that the TH2 helix (yellow) is disrupted at low pH for the WT (F) and E259Q mutant (H), but not for the mutants with histidine replacements (e.g., G).

Therefore, we focused our analyses on the secondary structure content present in our simulations. Overall, the total helicity of the T-domain remained relatively stable throughout all pH values (Fig. 3 A), which is unsurprising taking into consideration the limited timescale sampled by our CpHMD simulations (200–500 ns). Nevertheless, two exceptions stand out at low pH for the WT and E259Q mutant where one replicate undergoes some loss of helicity in the TH2 helix (Fig. 3 B). Previous works proposed that TH2 helicity loss is one of the initial conformational changes promoting the switch from soluble to membrane-competent state (6). This TH2 unwinding seems to be induced by H257 protonation (Fig. S10), which is regulated by the repulsion between H223 and H257 (Fig. 3, F and H). This is particularly evident in E259Q due to the absence of the safety latch (Fig. 3 H). Despite the qualitative nature of these observations, due to the lack of more robust sampling, we nonetheless propose that these local destabilizations are the initial steps that trigger the larger conformational transition into the membrane-competent state. We also searched for specific interactions between H257 and the residues of TH2. When in safety-latch conditions, the most stable interaction is between the H257 and I226 main chain (Fig. S11). When the safety latch is broken, by H223 deprotonation, H257 protonation, or by mutation (Fig. S12), we observe a destabilization of this specific interaction (H257-I226), including when TH2 unwinds.

We quantified the two effects that contribute the most to the propensity to undergo the conformational transition. We calculated the distances between the side chains of residues 223 and 259 to evaluate the quality of the safety latch and between residues 223 and 257 as a measure of the putative electrostatic repulsion that needs to be overcome before H257 protonation and the final conformational trigger. The values from both these distances were used to produce energy density maps that allow us to evaluate the structural details of each mutation upon pH acidification (Fig. 4).

Figure 4.

Effects of the pH (columns) and mutations (rows) on the correlation of the distances between the residues of the triad. The energy landscape (color coded on the right axis) of the distances between the residues in the safety latch (H223 and E259) and the repelling histidine residues (H223 and H257) for the WT T-domain and all four tested mutants. Four quadrants on this landscape are shown, with their respective abundance percentage, representing distinct structural features. The dotted lines were positioned at 0.5 nm to delimit a typical H-bond distance threshold (see text for details).

The maps have been divided into four distinct quadrants with different structural characters. The top-left quadrant represents the typical physiological structure of the T-domain, having H223 and H257 deprotonated, allowing their proximity, while the H223-E259 latch is not formed due to the lack of charge in H223. In the bottom-left quadrant, H223 is protonated and establishes a strong safety latch with E259, locking the protein until H257 is also protonated. In the bottom-right quadrant, we have unusual conformations where the two histidines H223 and H257 are repelling each other, but the safety latch is still formed. Finally, the interesting top-right quadrant corresponds to conformations ready to transition into the membrane-competent state, where the safety latch is broken and the two histidine residues are already repelling each other.

In the WT T-domain, we observe a clear formation of the safety latch as soon as pH is decreased, moving our ensemble from the top-left quadrant to the bottom-left. Only at a very low pH (3.0), do we see a small conformational shift toward the rightmost quadrants, highlighting the histidine repulsion. Unsurprisingly, the E259Q and H223Q mutations are the most prone to transition their conformation due to the absence of the safety latch. The larger population of the top-right quadrant in E259Q results from adding the effect of histidine repulsion, which is not present in H223Q. It should be noted that the glutamine residue used in the H223 mutation also interacts strongly with E259 by hydrogen bonding, limiting our sampling of the open conformations, typical of a system with no latch. These timescale limitations in our sampling may also contribute to the differences observed between the H223Q and E223Q. The H257Q and the double mutant exhibited almost no open conformations, even when the safety latch was removed (double mutant). These results highlight the key role of H257 protonation in transition into the membrane-competent state and suggest that, by mutating this residue, we should lose all ability of the T-domain to respond to pH changes.

Conclusions and perspectives

Deciphering the mechanisms of protonation-induced conformational switching in proteins is critical for understanding the physiological functioning of many cellular systems. Here, we have used an important example of the diphtheria toxin’s translocation domain, which responds to endosomal acidification to undergo a series of protonation-driven conformational changes to infect the target cell. We have applied our computational approach of CpHMD simulations to probe the initial transitions in the T-domain occurring in solution, and specifically to test the involvement of the H223/H257/E259 triad implicated in forming a safety latch that prevents the premature triggering of refolding (6,24).

Our computational approach has provided solid evidence of the tight involvement of the safety latch and histidine repulsion in triggering the conformational transition in the T-domain of the diphtheria toxin. Our $\mathrm{p}{K}_{\mathrm{a}}$ analysis highlighted H257 as the final residue to protonate, triggering the initial conformational changes compatible with the membrane-compatible state. The H257 $\mathrm{p}{K}_{\mathrm{a}}$ value is modulated mainly by the safety latch formed by neighboring residues H223 and E259. This protonation/conformation interplay is consistent with already published experimental results obtained with several T-domain mutants and provides further guidelines for designing other mutants (e.g., E259Q) with altered active pH ranges. The latter will have potentially important biomedical applications for specific targeting of acidic diseased tissues.

This study can be also considered a proof of concept for applying CpHMD calculations to study other systems of physiological importance, such as Bcl-2 apoptotic regulators that share some structural features of the T-domain (17,18). Obviously, the potential targets for this approach will also include other toxins entering the cell via the endosomal pathway and experiencing pH-dependent conformational switching, such as tetanus, botulinus, and anthrax. To successfully apply our approach to these and other systems several methodological advances should be further developed, namely expanding the range of such simulations into a microsecond range and the inclusion of the lipid bilayer (the latter is likely to go beyond the simple environmental effects and changes in proton concentration, and is expected to affect the $\mathrm{p}{K}_{\mathrm{a}}$ values of key residues in a lipid-dependent manner, as implied by our precious experiments (18,49)). None of these limitations appear to be fundamental and our future studies will be focused on using this approach to decipher the mechanisms of the T-domain action in the later stages of the membrane insertion pathway.

Acknowledgments

We thank Diogo Vila Viçosa for fruitful discussions. We acknowledge financial support from Fundação para a Ciência e a Tecnologia through grants 2021.06409.BD and CEECIND/02300/2017, and projects UIDB/04046/2020 (https://doi.org/10.54499/UIDB/04046/2020) and UIDP/04046/2020 (https://doi.org/10.54499/UIDP/04046/2020). A.S.L. was supported by the NIH grant R01GM145626. This study was also supported by the European Union (TWIN2PIPSA - Twinning for Excellence in Biophysics of Protein Interactions and Self-Assembly, GA 101079147). Views and opinions expressed are those of the author(s) only and do not necessarily reflect those of the European Union or European Research Executive Agency (REA). Neither the European Union nor the granting authority can be held responsible for them.

Author contributions

A.S.L. and M.M. designed the research. N.F.B.O. carried out all simulations and analyzed the data. All authors interpreted the results and wrote the article.

Declaration of interests

The authors declare no competing interests.

Editor: Ana-Nicoleta Bondar.