PH-TRIGGERED CONFORMATIONAL SWITCHING OF THE DIPHTHERIA TOXIN T-DOMAIN: THE ROLES OF N-TERMINAL HISTIDINES

Abstract

pH-Induced conformational switching is essential for functioning of diphtheria toxin, which undergoes a membrane insertion/translocation transition triggered by endosomal acidification as a key step of cellular entry. In order to establish the sequence of molecular rearrangements and side chain protonation accompanying the formation of the membrane-competent state of the toxin’s translocation (T) domain, we have developed and applied an integrated approach that combines multiple techniques of computational chemistry (e.g., long, µsec-range, all-atom molecular dynamics simulations; continuum electrostatics calculations; and thermodynamic integration) with several experimental techniques of fluorescence spectroscopy. Thermodynamic integration calculations indicate that protonation of H257 causes the greatest destabilization of the native structure (6.9 kcal/mole), which is consistent with our early mutagenesis results. Extensive equilibrium MD simulations with a combined length of over eight µsec demonstrate that histidine protonation, while not accompanied by the loss of structural compactness of the T-domain, nevertheless results in substantial molecular rearrangements characterized by the partial loss of secondary structure due to unfolding of helices TH1 and TH2, and the loss of close contact between the C- and N-terminal segments. The structural changes accompanying the formation of the membrane-competent state ensure an easier exposure of the internal hydrophobic hairpin formed by helices TH8 and TH9, in preparation for its subsequent transmembrane insertion.

Conformational switching, broadly defined as an alteration in the spatial organization of a macromolecule in response to environmental changes, binding of a ligand or proteolytic cleavage, is at the core of any cell function. The design of specific switches leading to changes between folded and disordered states or between distinct folds, as well as between different oligomeric states, is also an important topic in protein engineering ¹. pH-triggered switching, which utilizes the ability of the cell to regulate proton concentrations in different compartments, has been implicated in the entry of multiple bacterial toxins (e.g., diphtheria, botulinum, tetanus, and anthrax ^{2; 3; 4; 5; 6; 7}) and viruses ^{8; 9} sharing the endosomal pathway as well as in the regulation of apoptosis ^{10; 11; 12}. Here we utilize experimental approaches, an array of molecular modeling methods, and novel computational capabilities of ANTON, a specialized super-computer, designed to generate extra-long (µsec-range) molecular dynamics (MD) simulations ^{13; 14}, to elucidate the molecular mechanism of conformational switching by protonation in diphtheria toxin translocation (T) domain.

The function of the T-domain is to translocate the catalytic domain across the lipid bilayer in response to acidification of the endosome ¹⁵, which requires a substantial refolding of the globular structure (see Fig. 1A for crystallographic structure ¹⁶) and insertion into the membrane 17; 18; 19; 20 Previously we have demonstrated that the first step of this complex conformational transition of the T-domain, namely the formation of the membrane-competent state, occurs in solution at mildly acidic pH ²⁰ and that histidine protonation is likely involved in various stages of conformational switching ^{21; 22; 23}. It remains unknown, however, what specific structural changes occurring in response to histidine protonation prepare the hydrophobic core of the T-domain for subsequent membrane insertion. To characterize molecular rearrangements and to explore the role of histidine protonation in conformational changes, we carried out multi-µsec molecular dynamics (MD) simulations of the T-domain in explicit solvent. Two MD simulations were performed, starting with the known structure of the T-domain in solution (Fig. 1A): one with all six native histidine residues in a standard (unprotonated) state (this corresponds to a membrane-incompetent state of the T-domain populated at neutral and high pH); and another with all histidines protonated (this corresponds to a membrane-competent state of a protein formed upon acidification). The simulations revealed a specific and subtle conformational transition, which prepares the protein for further interactions with the lipid bilayer. Several predictions based on MD simulations were validated experimentally using fluorescence spectroscopy. The role of individual histidines in T-domain conformational switching has been evaluated by all-atom Thermodynamic Integration (TI) simulation and by continuum electrostatics calculations, based on solving the Poisson-Boltzmann (PB) equation as pioneered by Honig and co-workers ^{24; 25}, and implemented in the HARLEM molecular modeling package 26; 27

Figure 1.

(A) Crystallographic structure of the diphtheria toxin T-domain¹⁶ is represented as a backbone ribbon with the hydrophobic membrane insertion hairpin TH8–9 highlighted in brown. Protonation of histidine residues (shown as red sticks) is implicated in modulating pH-dependent refolding of the T-domain, which is studied here by various computational methods (see Fig. 5). Extended MD simulations indicate unfolding/refolding of the long N-terminal helix TH1 and helix TH2 (shown in green) is an important step in the formation of membrane-competent state of the T-domain (see Fig. 2 and Supplemental Information movies, SI_movie_l_all.avi, SI_movie_THl_TH2.avi). Changes in proximity of residues W206 and Q369 (shown as space filling structures) are used to verify the predictions of the MD simulations using methods of fluorescence spectroscopy (see Fig. 4). Grey space filling structure corresponds to W281. (B and C) pH-Dependent conversion of T-domain from soluble W-state into membrane-competent W⁺-state. (B) Identification of W-to-W⁺ transition through membrane binding at lipid saturation. FCS-based mobility measurements (diamonds) and measurements of FRET between donor-labeled T-domain and acceptor-labeled vesicles (circles) were performed as described in with raw data presented in SI Fig. SI A. Triangles represent increase in fluorescence of environment-sensitive probe bimane attached to a single cysteine in position 369. Solid line represents a fit of combined data with Eq. 4 of the Supplemental Experimental Methods. (C) Identification of W-to-W⁺ transition through spectroscopic measurements in the absence of lipid. Relative changes in Trp intensity (triangles), ellipticity at 222 nm (circles) are the averages of triplicate measurements. Squares represent relative changes in the amplitude of the unquenched lifetime component of fluorescence decay component of bimane dye attached to residue 369 (more on quenching mechanism in text and in Fig. 4). The error bars for bimane amplitude represent results of the support-plane analysis that accounts for the cross-correlation of the fluorescence decay parameters. All data are presented as fractional changes in the corresponding signals between values measured at pH 8.0 and 5.5 (representative raw data are shown in Fig. 4C and SI Figs. SIB and SIC) and overlaid with the fit to binding data from panel B (solid line) for visual reference.

Results

Experimental evidence of pH-dependent formation of membrane-competent conformation

Formation of the membrane-competent form (W⁺-state) of the T-domain is the first step along a complex pathway, leading from a soluble conformation with a known crystallographic structure (W-state) ultimately to membrane-inserted states, for which no high-resolution structural information is available. The W-to-W⁺ conversion is the first pH-dependent conformational transition along this pathway and is normally identified in a membrane binding assay conducted under conditions of saturation with lipid vesicles ²⁰. We illustrate in Fig. 1B the application of three independent membrane binding methods to the T-domain’s interaction with vesicles (raw data are shown in Supplementary Information Figs. S1A-C) based on: diffusion measurements by FCS (diamonds), association of donor-labeled T-domain and acceptor-labeled vesicles by FRET (circles) and changes in polarity of a specifically attached environment-sensitive bimane probe by increase in fluorescence intensity (triangles). All datasets are adequately described by a single fitting curve with a pKa of 6.2 and Hill coefficient of about 2. Is it possible to correlate this pH-dependent binding with conformational changes in solution? This is not trivial, because in the absence of membranes under acidic conditions the T-domain aggregates avidly ^{28; 29}, leading to poorly reproducible results, especially at pH<5.5. Nevertheless, when the tryptophan fluorescence intensity and ellipticity at 222 nm measured in solution are normalized in the pH range of 5.5÷8 (Fig. 1C), their relative changes correlate well with an accurately measured binding curve (solid line). This confirms that the W⁺-state is not just a virtual thermodynamic state (e.g., similar to unfolded melittin on membrane interface ³⁰), or an unstable intermediate on the kinetic path from the W-state to a membrane-bound protein, but a relatively stable experimentally observable state of the protein with conformation different from that of the W-state. Next, we examine structural rearrangements along the W-to-W⁺ transition via MD simulations employing protonation of all six histidine residues as a pH-dependent trigger.

Extended MD simulations of conformational change triggered by protonation of histidines

T-domain is composed of nine alpha helices (further termed TH1-TH9) ¹⁶ and has a globular structure in aqueous solution at neutral pH. Helices TH8 and TH9 constitute the hydrophobic core of the protein partially protected from water by the rest of the structure (Fig. 1A). At low pH the T-domain has been demonstrated to insert into lipid bilayer in a transmembrane conformation ^{17; 18; 19; 20}. However, little is known about specific structural changes initiated by His residue protonation. In order to shed light on how structural changes are initiated in solution upon protonation of the His residues, we performed molecular dynamics simulations of the T-domain in explicit aqueous solution mimicking both low-pH and high-pH conditions. To simulate the globular membrane incompetent W-state of the protein that occurs at neutral or basic pH, it was assumed that all protonatable residues of the protein were in their standard protonation states, including six histidines in neutral form. To simulate formation of a membrane-competent W⁺-state, populated at mildly acidic pH ²⁰, all six His residues of the T-domain were protonated. Protonation of all histidines provides the largest possible driving force for structure destabilization associated with lowered pH. Thus, we design a simulation with enhanced possibility of observing a structural transition on the MD simulation time-scale. Currently, there are no direct measurements of the protonation of the histidines in the T-domain, while indirect measurements give conflicting answers to the question whether histidines in the T-domain are protonated sequentially or cooperatively in solution or near the membrane interface ^{21;22; 23} . Therefore in our analysis we considered two separate questions: (1) what is the impact of His protonation on the protein structure and (2) what is the role of individual His residues in initiating and driving structural transition toward the W⁺-state?. The former question is addressed in this section, while the latter one is deferred to the next (see the next section for a detailed analysis and discussion of individual pKas and their role in driving structural transition of the protein).

In the neutral pH simulation of approximately 2 µsec, the protein retained its globular structure in a natively folded state. The absence of significant structural changes during the entire simulation is evident from low values (about 1.9 Å) of the average root mean squared deviation (RMSD) of the Cα atoms in TH1–9 from their crystallographic positions (blue trace in Fig. 2A).

Figure 2. Conformational changes in T-domain triggered by histidine protonation established by MD simulations.

(A) Comparison of RMSD traces obtained from MD runs with unprotonated (blue trace) and protonated histidines (yellow trace). MD-generated structures were rotated and translated to minimize the RMSD values relative to crystallographic coordinates of Ca atoms in helices TH1 to TH9. The conformational transition apparent from the abrupt change in the latter trace at ~0.5 µs affects the secondary (B) and tertiary structure (C) of the T-domain in the W⁺-state. (B) Changes in helical content obtained by averaging the probability of helical conformation during four different stages of the conformational transition triggered by histidine protonation. Colored lines represent (i) initial membrane-incompetent W-state (blue, 0–0.5 µs); (ii) main conformational transition (red, 0.5–1.5 µs); (iii) subsequent refolding (green, 1.5–4.0 µs) and final conformation of membrane-competent W⁺-state (yellow, 4.0–6.7 )µs). These data indicate the loss of helical structure corresponding to crystallographic helices TH2 and TH7 as well as refolding of a long helix TH1 into two short helical segments. (C) MD snapshots illustrating the changes in fold of the TH1–2 helical region. The original structure formed by two well-defined helices (198.42 ns snapshot, blue) is lost during conformational change caused by histidine protonation (800.58 ns, red) and is transformed into two short helical regions folded against each other and a substantial unfolded segment (2878.92 ns, green). This refolding also substantially increases the distance between H223 and H257 (SI_movie_3_H223_H257.avi). As a result, the most hydrophobic helix TH8 (grey) becomes more accessible for subsequent interaction with the lipid bilayer.

In stark contrast to the neutral pH simulation, large structural changes were observed in the protein during the 6 µs simulation at low pH. In Fig. 2A, the Cα RMSD plot shows an abrupt increase at 532 ns (dark yellow trace). This is primarily due to formation of a kink in helix TH1, initiated by rotation of backbone dihedral angles of K216 (see also Supplemental Information movie SI_movie_1_all.avi). This results in a decrease in α-helical content on the C-terminus of TH1 followed by unfolding of its N-terminal. The subsequent partial refolding of TH1 is initiated in its N-terminus at 1037 ns. The resulting structure is stabilized by packing against TH9. Helix TH2 unwinds at 1414 ns followed by the C-terminal refolding of TH1 at 1482 ns (see also the description and the movie in Supplemental Information, SI_movie_2_TH1_TH2.avi). Thereafter, the Cα RMSD remains at a stable value, around 3.8 Å, over the last 2 µsec of the simulation. Note that the low value of RMSD prior to ca. 500 ns is similar to that for the neutral pH simulation. Therefore a conventional MD simulation (typically tens of nanoseconds in length for a system of this size) would have failed to detect the structural implications of histidine protonation. The dynamics of partial unfolding, subsequent refolding and further reorganization of the entire protein secondary structure is clearly seen in the plot of a fraction of α-helical content shown in Fig. 2B. Note that the observed conformational changes in the global structure also resulted in repositioning of the two histidines H223 and H257 with respect to each other. These residues, located in TH1 and TH5, respectively, are in close proximity to each other in the initial undisturbed structure, but move apart in the partially unfolded and refolded structures (see Supplemental Information, SI_movie_3_H223_H257.avi). This conformational transition also caused H257 to become exposed to the solvent, which is likely to provide the driving force for the observed refolding process (see section Thermodynamic integration and continuum electrostatics analysis of histidine protonation) As a consequence of the observed transition, the hydrophobic helices TH8 and TH9 become somewhat exposed to the solution, forming a membrane-competent state, a precursor state for the association with the membrane in the insertion process. The solvent accessible surface area (SASA) of hydrophobic side chains in helices TH8–9 increases along the low-pH T domain trajectory (red traces in Figs. 3A and 3B) relative to the average SASA of the same helices along the neutral-pH trajectory (shown as blue dashed lines in Figs. 3A and 3B).

Figure 3. Changes in solvent accessible surface area (SASA) of hydrophobic side chains in helices TH8 (A) and TH9 (B).

The light blue and yellow traces correspond to the MD simulations with neutral and protonated histidines respectively. The dark blue and red lines represent running averages over a 60 ns window. Blue dotted lines are drawn for visual representation of the average SASA values in the W-state for TH8 (SASA=39), and TH9 (SASA=116).

Experimental validation of MD simulation results

We have used several spectroscopic techniques to validate various aspects of the MD-based conformational predictions. First, we examined the overall changes in secondary structure by measuring CD spectra (Fig. 1C and Supplemental Results Fig. S1C). At mildly acidic pH, the T-domain undergoes partial unfolding, and the resulting loss of ellipticity is consistent with that expected from unwinding of the TH1–2 segment (SI_movie_2_THl_TH2.avi). Second, we measured the intrinsic fluorescence from the two tryptophan residues (W206 and W281). While formation of the membrane-competent state is accompanied by a small decrease in intensity, the position of the maximum of emission remains constant in this pH range (from 8 to 5.5) and does not shift until a much lower pH is reached (Supplemental Results Fig. S1B). This is consistent with a lack of exposure of tryptophan residues to the aqueous phase predicted by the MD simulations. Third, the overall retention of the compact conformation during the formation of the membrane-competent state (SI_movie_l_all.avi) is confirmed by FRET measurements between native tryptophans and AEDANS dye selectively attached at a single cysteine in a Q369C mutant (Fig. 4D).

Figure 4.

Experimental validation of the MD-predicted structural changes between membrane-incompeten W-state (illustrated by 459.54 ns snapshot structure in blue) and membrane-competent W⁺-state (illustrated by 4865.22 ns snapshot structure in yellow). (A) The W-to-W⁺ conformational rearrangement triggered by histidine protonation and illustrated in Fig. 2 results in a change in proximity of residues Q369 (C-terminal helix TH9) and W206 (N-terminal helix TH1), which can be verified by two specifically designed fluorescence experiments. (B) Distance between alpha-carbons of W206 and Q369 from MD simulation of nonprotonated (blue) and protonated T-domain (yellow). (C) Validation of proximity changes in panel A by short-range aromatic quenching³¹ of bimane probe attached to single cysteine mutant at position 369: fluorescence decay following the pulsed excitation is strongly quenched in W-state (pH 8, black line), but not in W⁺-state (pH 5.5, red line). (D) Long-range FRET measurements between two donors, W206 and W281, and AEDANS acceptor probe attached to single cysteine mutant at position 369: excitation spectra measured at 500 nm at pH 8.0 (black line) and 5.5 (red line) showed no variation in FRET level, while unfolding with guanidine HC1 (blue line) reduced FRET, as seen by the loss of donor contribution at 280–290 nm. Both spectroscopic experiments confirm MD results, suggesting local rearrangements but not a global refolding for W-to-W transition.

To further verify specific conformational changes predicted by the MD simulations, we have designed a spectroscopic experiment based on the following observation. As predicted by MD simulations and illustrated in Fig. 4, conformational reorganization of the protein upon protonation results in an increase in the distance between Q369 in the C-terminal of TH9 and W206 in the N-terminal of TH1 (see also Supplemental Information, SI_movie_4_W206_Q369.avi). Based on this prediction, we designed a spectroscopic experiment sensitive to changes on this distance scale by replacing Q369 with a cysteine and labeling it with bimane dye. Fluorescence of bimane has been demonstrated to be strongly quenched by tryptophan in a short-distance range (less than 10 A) ³¹. This is different from the relatively long-range FRET measurements described above, which are not sensitive to distance changes in this range (the Förster radius for Trp-AEDANS donor-acceptor pair is on the order of 25–30 Å). Indeed, our fluorescence lifetime measurements of bimane-labeled T-domain at pH 8 show highly quenched kinetics with a pronounced sub-nanosecond component (Fig. 4C, black curve). This quenching is relieved at pH 5.5 (red), consistent with the expectation from MD (Fig. 4A, red arrow). At intermediate pH values the kinetics can be represented by the mixture of quenched and unquenched fluorescence decays. Once the conformational change revealed by MD is validated by experiments, the MD TI procedure is used to calculate the free energy of protonation of His residues.

Thermodynamic integration and continuum electrostatics analysis of histidine protonation

We have utilized an MD-based Thermodynamic Integration (TI) approach to determine the thermodynamic consequences of protonation of individual histidines ³². The values of free energy of protonation ΔG, calculated for all histidines in the context of the native folded state, range from somewhat favorable (e.g., for H223) to highly unfavorable (Fig. 5A). Protonation of H257 has the highest free energy penalty, and hence the most destabilizing effect on the native structure, which is consistent with our mutagenesis results²² The free energy of protonation of H257 increases even more when calculation is repeated assuming that the neighboring H223 is already protonated (this is a reasonable assumption given the negative value of ΔG for H223).

Figure 5. Thermodynamics of histidine protonation in the T-domain.

(A) TI calculations of the free energy of protonation ΔG of all of the histidines (see text for details). Time traces of pKa’s for residues H223 (B), H251 (C) and H257 (D) calculated using PB approach from the MD simulations. Light blue and yellow traces correspond to the calculation for MD simulations with nonprotonated and protonated histidines, respectively (Fig. 2A), while dark blue and red represent corresponding pKa traces averaged over a 60 ns-window (pKa values were calculated from MD-generated structures every 600 ps). Blue dotted lines are drawn for visual representation of the average pKa values in the W-state for H223 (average pKa=6.3), H251 (average pKa=5.3) and H257 (average pKa=4.7). As further discussed in the text, both TI and PB calculations indicate different roles for H257 (difficult to protonate in the W-state; protonation provides the major driving force of refolding into the W⁺-state) and H223 (easy to protonate in the W-state, prevents premature protonation of H257).

In order to examine how the environments of individual histidine residues change during formation of the membrane-competent state, we have also calculated pKa values along the MD trajectory using a simpler continuum electrostatics approximation via solving a Poisson-Boltzmann (PB) equation numerically, as described in Computational Methods section and ^{26; 27}. In Fig. 5 we illustrate such pKa traces for three N-terminal histidine residues: H223, H251 and H257. (Note that we have recently demonstrated that the three C-terminal histidines do not affect folding in solution, but are important for the later steps of the membrane insertion/translocation pathway of the T-domain ²³. Therefore the PB calculations for H322, H323 and H372, shown in Fig. S2, will be discussed elsewhere.) PB calculations for the unprotonated W-state (light and dark blue, for 100 point average), plotted on a negative timescale for convenience of comparison, correspond well to more rigorous, yet computationally costly TI calculations. For all residues, the pKa values fluctuate substantially, yet the average values for H257 (4.7) and H251 (5.3) are much lower than that for H223 (6.3). The pKa traces, however, calculated along the conformational transition toward the membrane-competent state (dark yellow and red, for 100 point average), demonstrate a more complex behavior. First, the initial equilibration leads to an immediate increase in pKa, with an approximate value of 7, which is achieved via local rearrangements of atoms around either histidine residue. The subsequent dynamics, however, are quite different for the three residues, with H223 and H251 experiencing little change throughout the entire simulation time. The environment for H257, on the other hand, becomes more accommodating for the positive charge, which is reflected in the rise of its average pKa to the level of 9.5. This additional stabilization, starting at ca. 4 µsec, is derived from the hydrogen bonding with backbone oxygens of residues A254 and K227 (not shown).

Discussion

In this study we have used an integrated approach, which combines experimental and computational methods to characterize pH-dependent conformational switching between membrane-incompetent and -competent states of diphtheria toxin T-domain (Fig. 1). We have taken advantage of novel capabilities of the ANTON supercomputer to generate long MD runs, which allowed us to observe conformational change triggered by protonation of six native histidines (Fig. 2). We verified our predictions using specifically designed spectroscopic experiments (Fig.4)which gave us confidence in applying thermodynamic analysis to histidine protonation (Fig. 5).

Our results reveal that the protonated W⁺-state, originally thought of as a molten globulelike state, has a defined conformation, which nonetheless differs from that of the W-state. The main difference is that helical conformation of the TH2 segment is completely lost, while the long TH1 helix is split into two short fragments that fold on each other (Fig. 4). Interestingly, in order to get to this final conformation, TH1 is first almost completely unfolded. Similarly, the TH4 segment completely loses its helicity during early stages of transition, yet regains structure after 1.5 µsec of transition (Fig. 4B). As a result of this conformational change, the most hydrophobic helical hairpin, TH8–9, which upon interaction with membranes adopts a trans- bilayer conformation²⁰, becomes more exposed. In addition, conformational change, triggered by His protonation, leads to disruption of several salt bridges between different helices (e.g., K212 of TH1 and E327 of TH8). While the overall direct contribution of these relatively exposed salt bridges to protein stability is unlikely to be very high^33;34, their disruption frees positively charged residues to interact with negatively charged lipids, which were demonstrated to play a critical role during the next stages of refolding, on the lipid membrane interface²⁰ Thus, despite its folded structure, the W⁺-state is poised to interact with the membrane and convert to the next insertion intermediate, the interfacially refolded insertion-competent state²⁰.

Thermodynamic integration results indicate diverse roles for different histidines in the pH-triggered conformational switching of the T-domain. For example, H223 is a likely candidate for early protonation, owing to the favorable negative free energy estimate for protonation in the context of the original structure (Fig. 5A). In contrast, H257 has the highest ΔG of protonation, thus exerting the largest perturbation on the native structure. (This is consistent with our recent mutagenesis study demonstrating partial unfolding of the H257R mutant, which is physiologically active and gains structure upon interaction with its ligand, the lipid bilayer²².) The highest ΔG of protonation computed for H257 correlates well with the largest upward change of the pKa value of this residue computed along the protonated T-domain MD trajectory. Recently we demonstrated that the C-terminal histidines H322, H323 and H372 are important at later stages of the inertion pathway ²³, although the positive ΔG values observed for H323 and H372 (as well as that for H251) suggest that their protonation is likely to contribute to the cooperativity of the initial refolding transition.

Remarkably, the free energy of protonation of H257 is raised even higher (from 6.9 to 10.2 kcal/mole) when easily protonatable H223 is charged (Fig. 5A). This suggests that H223 plays an interesting role in modulating the protonation of H257. At neutral pH, H223 gets protonated before other histidines but does not cause much of the protein unfolding (negative ΔG). Its positive charge, however, influences the protonation of H257 by increasing its free energy of protonation and hence shifting the pKa of H257 to even lower values. We further illustrate these findings in Figure 6, which presents the results of Poisson-Boltzmann calculation of pKa values for the two histidines of the diphtheria toxin T-domain, H223 and H257. In the W-state (upper panel) H223 exhibits a unimodal pKa distribution with a maximum at pKa 6.5 (blue histogram). This suggests that it is likely to be protonated even before endosomal encapsulation (pH range associated with the acidic environment of the endosome is shown as a grey band ^35,36,37.) On the contrary, H257 has much lower average pKa of 4.7, distributed at two peaks centered at pH of 4 and 6.3 (green histogram). (This does not mean, however, that one needs to reach a very low pH to protonate H257 in the cell, because the T-domain undergoes a structural refolding coupled with protonation. Note that the pKa distributions for the membrane-competent W⁺-state (Fig. 6B) are well above the value of pH 5.5 (dotted line), at which the transition from W-to-W⁺ is completed (Fig. 1).) The distribution for H257 is shifted even further to acidic pH when calculation is repeated assuming H223 is already protonated (red histogram). These differences are consistent with Thermodynamic Integration calculations (Fig. 4A), which suggest that protonation of the H257 in the context of the W-state is energetically very costly, especially when H223 is protonated. As a result the titration curve for H257 is completely removed from the range of neutral pH, and the probability of its protonation under these conditions is minimal. We hypothesize that this may be important physiologically, because otherwise the protonation of H257 would have caused substantial unfolding before the endosomal apartment is reached and would trigger a non-productive interaction with the membrane at an early stage of the insertion pathway. Thus, H223 can be compared to a safety device, which reduces protonation of the crucial H257 by further shifting its pKa, and holding it in a state resembling a loaded spring until the protein is poised for translocation in the endosomal compartment. Once acidification of an endosome lowers pH sufficiently for the protonation of H257 to occur, the safety latch can no longer hold and the spring is released, causing the conformational change that results in formation of the membrane-competent state, membrane insertion and translocation. We intend to test this safety latch hypothesis in future studies, using mutagenesis and NMR measurements of protonation of individual histidines of the T-domain.

Figure 6.

pKa distributions of key histidine residues, H223 and H257, calculated in PB approximation from MD traces for (A) membrane-incompetent W-state and (B) membrane-competent W⁺-state (region of 4.0–6.7 µsec in protonated MD run from Fig. 5 B, D) (see text for details).

Perspectives

MD computer simulation is a technique that can, in principle, provide a detailed, “atom-by-atom” picture of the structural reorganization of a molecule during a relatively fast relaxation process. In recent years MD methodology has been significantly refined and is becoming a widely used technique for modeling a variety of structural and dynamic properties of proteins (such as structure refinements, NMR relaxation and free energy computations). Yet its main limitation has been a relatively short timescale – typically, within tens of nanoseconds – which is certainly insufficient to analyze processes such as structural reorganization and folding/unfolding of proteins. This limitation, due to a shortage of computational power, is quickly becoming less severe, especially with the introduction of massively parallel special purpose super-computers, such as the ANTON system of Shaw Corp. ^{13; 14} used in this study. Here, we present simulations of the diphtheria toxin T-domain in explicit atomistic aqueous solution of a cumulative length of 8 µsee, a timescale sufficient to directly observe the conformational reconfiguration that the T-domain undergoes upon protonation of histidine residues. Various structural predictions of the simulation are verified experimentally, and thermodynamic predictions made using two alternative approaches (TI and PB calculations) are consistent with earlier mutagenesis studies.

Our results on the variation of the free energy of protonation for different histidines of the T-domain (Fig. 5A), as well as changes of the pKa of H257 along the W-to-W⁺ transition (Fig. 5D), clearly illustrate that a large variation in pKa values (over 4 pH units) can be achieved in a relatively compact state and do not require substantial unfolding. This potentially enables both subtle and large-scale conformational switching by changes in pH, similar to that occurring in proteins involved in viral fusion ^{8; 9}, assembly of multimeric proteins ³⁸ and possibly also in intrinsically disordered (ID) proteins. The analysis of amino acid compositions of ID and folded proteins indicates that folding is promoted by higher hydrophobicity and lower mean charge Thus, protonation of histidines is expected to favor unfolding propensity, as it would introduce a charge and lower the hydrophobicity. The possibility that histidine protonation may play a role in regulation of the folding transition in ID proteins is also supported by the observation that when placed on Vihinen's flexibility scale ⁴⁰ with other residues, histidines occupy a borderline between residues more frequently found in folded and unfolded protein segments ³⁹. We expect that the integrated approach developed in this study, which relies on extended MD simulations combined with experimental verification, not only unveils the role of histidine protonation in conformational switching on later stages of the T-domain membrane-insertion pathway but will also be instrumental in the broader task of deciphering its role in regulating folding transitions in other proteins.

MATERIALS AND METHODS

Experimental methods are described in Supplemental Experimental Methods

MD simulations setup and analysis

The structural model of the T-domain (residues 201–380) was obtained from a high resolution structure of the entire toxin [Protein data bank ID code 1F0L ⁴¹]. Hydrogen atoms and a rectangular box of water solvent molecules have been added to the protein model using AMBER 9 tLeap utility. The simulation box contained 4587 TIP3P⁴² water molecules and had initial dimensions of 59 × 59 × 63 Å. The following two molecular mechanics models of the T-domain were constructed: (a) a model of the T-domain with histidine residues in unprotonated epsilon tautomeric form (the equivalent of membrane-incompetent W-state, populated at neutral pH) and (b) a model in with all six histidine residues are in the protonated form (the equivalent of membrane-competent W⁺-state, populated at mildly acidic pH as illustrated in Fig. 1). Other titratable residues were kept in their standard protonation states in both models. Sodium ions were added to neutralize the charge of the system. AMBER ff99SB force field parameters were used in all simulations ⁴³. Periodic boundary conditions were applied and Particle Mesh Ewald (PME) method ⁴⁴ with cutoff radius of 8 Å was used to compute electrostatic interactions. Energy minimization of the systems performed by steepest descent and conjugate gradient methods were followed by a 20 ns and 16 ns equilibration MD runs for the neutral pH and acidic pH models, respectively. Equilibration MD simulations were executed using the AMBER 9 PMEMD ⁴⁵ program with all covalent bonds involving hydrogen atoms constrained with SHAKE ⁴⁶ and a simulation time step of 2 fs. Pressure was kept at 1 atm using Berendsen barostat ⁴⁷ and temperature was kept at 300 K using the Langevin thermostat ⁴⁸ with a collision frequency of 2 ps⁻¹. The equilibrated structures were used as initial configurations for production of MD simulations on ANTON supercomputer ^{13; 14}. The parameters of production MD simulations were the same as for the equilibration run, except that the Gaussian Split Ewalds method ⁴⁹ with grid size of 64 × 64 × 64 and cutoff radius 11.56 Å was used to compute electrostatic interaction. Berendsen thermostat was used to keep the temperature of the systems at 310 K with time constant of 1.2 ps. Atom coordinates were saved every 60 ps. MD simulations were carried out for 2060 ns and 6723 ns for the neutral- and the low-pH T domain models, respectively. Trajectory post-processing was performed with VMD 1.8.7 ⁵⁰ and the ptraj program available in AmberTools 1.4 suite ⁵¹. Secondary structure identification was carried out using the DSSP program ⁵². Molecular figures were generated with VMD 1.8.7.

Thermodynamic Integration calculations of ΔG values of histidine protonation

Thermodynamic Integration(TI) calculations for the protonation of individual histidine residues in the T-domain in the W-state were performed using a molecular mechanics model similar to that described in Molecular Dynamics section. The computational model contained the same residues 201–380 as the T domain solvated in a periodical box with 4587 TIP3P water molecules. The AG values were computed from free energy differences of quasi-chemical transformations of histidine residues from unprotonated (δ) state to the protonated state in water $\mathrm{\Delta }{G}_{\mathit{\text{unprot}}\to \mathit{\text{prot}}}^{\mathit{\text{water}}}$ and in the solvated protein fragment. Free energies of quasi-chemical transformation of histidine residues from the unprotonated to the protonated form in water $\mathrm{\Delta }{G}_{\mathit{\text{unprot}}\to \mathit{\text{prot}}}^{\mathit{\text{water}}}$ and in protein $\mathrm{\Delta }{G}_{\mathit{\text{unprot}}\to \mathit{\text{prot}}}^{\mathit{\text{protein}}}$ computed from 〈H_prot−H_unprot〉_λ averages over MD trajectories executed with intermediate state Hamiltonian H_λ = λ*H_unprot + (1-λ)*H_prot at 3 λ values (0.11271, 0.5, 0.88729) and Gaussian quadrature formula. The length of the MD runs for each λ value was between 14 and 21 ns. All MD simulations with H_λ Hamiltonians have been started from geometry obtained after initial energy minimization of water-solvated protein fragment obtained from crystallographic structure followed by 10 ps equilibration MD run.

Continuum electrostatic (Poisson-Boltzmann) calculations

Continuum electrostatic calculations were based on finite difference solutions of the Poisson-Boltzmann equation provided by DELPHI program ²⁴. In this approach proteins were treated as a body of low dielectric (ε = 4) surrounded by a solvent modeled by a continuum of high dielectric constant (ε = 80). The protein and high dielectric boundary conditions were mapped in a cubic grid 121 × 121 × 121. The protein atoms van der Waals radii and charge parameters were set to PARSE radii³⁴ and to AMBER 94 ⁵³ charge parameters, respectively. The radius of water molecules was set at 1.4 Å and the ionic strength was set at 0.15 M. Intrinsic pK_a values were calculated for only histidine residues and the N-terminus followed by calculation of the apparent pK_a values by Metropolis Monte-Carlo sampling ⁵⁴ as implemented in HARLEM program ^26;27. The number of MC steps was 10⁶, which was determined to provide converged average values. These calculations were repeated every 4 ps and 600 ps over the equilibration and production MD trajectories, respectively.

Supplementary Material

01. movie_1_all.avi.

A fragment of the MD trajectory with all HIS protonated (low pH trajectory) shows unfolding of helix TH1 at 532 ns, which corresponds to the transition between the blue and red states shown in Fig. 2B. Also note that the side chain of H223 becomes more exposed to the solvent following unfolding of helix TH1 and increases its distance to H257 (both shown as red sticks, the top one is H257 and the lower one is H223). Notice that W206 (green space filling representation) moves away from Q369 (ochre space filling representation). Helices TH1-4 (residues 206 to 267), TH5-7 (residues 268 to 318), and TH8-9 (residues 319 to 375) are shown in green, translucent grey and ochre new cartoon representation respectively. W281 is shown in translucent grey space filling representation.

02. movie_2_TH1_TH2.avi.

A fragment of the MD trajectory with all HIS protonated shows partial unfolding and subsequent refolding of helix TH1 and unfolding of helix TH2. This fragment shows the trajectory from 15 ns to 1815 ns.TH1-2 (residues 201 to 239) is shown in a cartoon representation and W206 (sticks) is highlighted to indicate the N-terminal of TH1. The TH1 backbone rearrangement occurs at 532 ns resulting in significant loss of the α-helical content. The N-terminal of TH1 recovers the α-helical conformation at 1037 ns and helix TH2 unfolds at 1414 ns. The C-terminal of TH1 recovers the α-helical conformation at 1482ns, which corresponds to the transition from red to green state shown in Fig. 2B. AAn α-helical structure is represented by magenta color, a 3₁₀helix by blue color, a π helix by red color, a turn is shown in cyan, and a coil is shown in white color. The secondary structure elements were identified via the STRIDE program available in VMD 1.8.7.

03. movie_3_H223_H257.avi.

A fragment of the MD trajectory with all HIS protonatedshows separation ofH223 (green sticks) andH257 (red sticks) relative to each other. This is associated with the conformational changes in helix TH1 at 532 ns. This fragment shows the trajectory from 255 ns to 1755 ns. Notice that helix TH4 undergoes secondary structure fluctuations over this MD segment before TH1 unfolds at 532 ns. The side chains of H223 (green sticks) and H257 (red sticks) become mainly solvent exposed after TH1 unfolding. Refolding of the TH1 N-terminal occurs at 1037 ns, and unfolding of TH2 occurs at 1414 ns. Furthermore, the TH1 C-terminal refolds into an α-helix at 1482 ns. In this movie, helices TH1-2 (residues 201 to 239) and TH3-4 (residues 240 to 267) are shown in cartoon representation. An α-helical structure is represented by magenta color, a 3₁₀helix by blue color, a π helix by red color, a turn is shown in cyan, and a coil is shown in white color. The secondary structure elements were identified via the STRIDE program available in VMD 1.8.7.

04. movie_4_W206_Q369.avi.

A fragment of the MD trajectory with all HIS protonated shows W206 (green sticks) moving away from Q369 (blue sticks) after unfolding of helix TH1 at 532 ns. This fragment shows the trajectory from 243 ns to 1143 ns. While a partial refolding of TH1 occurs at 1037 ns W206 and Q369 remain at an extended distance from each other. Notice that the C-terminal of helix TH9 (370-375) adopts a turn or 3₁₀conformation over this MD segment. In this movie, helices TH1-2 (residues 206 to 239) and TH8-9 (residues 326 to 375) are shown in cartoon representation. An α-helical structure is represented by magenta color, a 3₁₀helix by blue color, a π helix by red color, a turn is shown in cyan, and a coil is shown in white color. The secondary structure elements were identified via the STRIDE program available in VMD 1.8.7.

• Conformational switching is investigated by computational and experimental methods

• Microsecond MD simulations reveal conformational change upon histidine protonation

• Thermodynamic integration is used to calculate free energy of histidine protonation

• Protonation of H257 ensures conformational changes

• Protonation of H223 protects from premature protonation of H257

ABBREVIATIONS

T-domain

diphtheria toxin T-domain

MD

Molecular Dynamics

TI

thermodynamic Integration

PB

Poisson-Boltzmann

FCS

Fluorescence Correlation Spectroscopy

FRET

Förster Resonance Energy Transfer