Periplasmic Proteins HdeA and YmgD by ... - ACS Publications

Oct 27, 2016 - activity.8 For HdeA, the strength of the dimer interaction is reduced by a factor of ... version 1.5 (The PyMOL Molecular Graphics Syst...
0 downloads 0 Views 6MB Size
Article pubs.acs.org/JPCB

Probing the Structure of the Escherichia coli Periplasmic Proteins HdeA and YmgD by Molecular Dynamics Simulations Eileen Socher and Heinrich Sticht* Division of Bioinformatics, Institute of Biochemistry, Friedrich-Alexander-University Erlangen-Nürnberg (FAU), Fahrstraße 17, 91054 Erlangen, Germany S Supporting Information *

ABSTRACT: HdeA and YmgD are structurally homologous proteins in the periplasm of Escherichia coli. HdeA has been shown to represent an acid-activated chaperone, whereas the function of YmgD has not yet been characterized. We performed pH-titrating molecular dynamics simulations (pHtMD) to investigate the structural changes of both proteins and to assess whether YmgD may also exhibit an unfolding behavior similar to that of HdeA. The unfolding pathway of HdeA includes partially unfolded dimer structures, which represent a prerequisite for subsequent dissociation. In contrast to the coupled unfolding and dissociation of HdeA, YmgD displays dissociation of the folded subunits, and the subunits do not undergo significant unfolding even at low pH values. The differences in subunit stability between HdeA and YmgD may be explained by the structural features of helix D, which represents the starting point of unfolding in HdeA. In summary, the present study suggests that YmgD either is not an acid-activated chaperone or, at least, does not require unfolding for activation.



INTRODUCTION Enteropathogenic bacteria, which are for instance swallowed with food or water, need to survive the acid conditions in the host stomach before they can infect the intestine. For fulfilling this challenging task, different strategies have emerged: One mechanism for acid resistance is the amino acid decarboxylase system, which can be found in many enteric bacteria. These enzymes decarboxylate glutamate, arginine, or lysine, thereby consuming cytoplasmic protons,1 which in turn raises the pH value in the cytoplasm. Furthermore, bacteria can slow the influx of protons into the cytoplasm by reversing their innermembrane potential.2 These mechanisms allow the cytoplasmic pH to be maintained at tolerable pH values (pH ≈ 4.5) because the inner membrane has only a relatively low permeability for protons. In contrast to the cytoplasm, the bacterial periplasm is enclosed by an outer membrane with a high permeability for protons, so that the pH of the periplasm decreases rapidly to the same value as the environmental pH (pH ≈ 2). To prevent the aggregation of periplasmic proteins, some of the enteric bacteria, for instance, Escherichia coli, Shigella flexneri, and Brucella abortus, express periplasmic proteins HdeA and HdeB. These proteins are expressed under normal physiological conditions, and they can be upregulated in response to moderately low pH values. Experimental data demonstrated that the survival and growth of E. coli, S. flexneri, and B. abortus was compromised after the disruption of the hdeAB operon, which is part of the genomic acid fitness island.3 A clear inverse correlation can also be observed between possessing HdeA/ © XXXX American Chemical Society

HdeB and the infectious dose. Salmonella enterica serovar Typhimurium (S. typhimurium) and Vibrio cholerae, which do not express HdeA and HdeB, need 10−109 higher dosages to cause infections.4 The structures of both HdeA (Protein Data Bank (PDB) codes: 1BG8,5 1DJ86) and HdeB (PDB code: 2XUV7) are known from experimental data. Both proteins are dimeric and exhibit a similar tertiary structure of the subunits. However, they use different interfaces for dimer formation.7 Functional in vitro studies indicated that HdeA is more efficient than HdeB at pH 2, whereas HdeB has a higher pH of 4−5 for optimal activity.8 For HdeA, the strength of the dimer interaction is reduced by a factor of 200 when pH decreases from 7 to 2, suggesting that HdeA is likely to be monomeric when it functions to suppress aggregation of denatured proteins under acid conditions.6 Acid-induced dissociation of HdeA is rather slow and occurs at a rate of 8 s−1,9 impeding a study of the underlying processes by conventional simulation methods. To enhance sampling, replica exchange methods were previously applied in both temperature space and pH space.10,11 To further enhance the sampling of the transition state region, a two-dimensional (2D) window-exchange umbrella sampling approach has recently been devised, which allowed a more detailed characterization of HdeA folding intermediates.12 Received: June 16, 2016 Revised: October 20, 2016 Published: October 27, 2016 A

DOI: 10.1021/acs.jpcb.6b06091 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Simulations in Implicit Solvent. All simulations in implicit solvent started with two rounds of minimization, each comprising 2500 steps of steepest descent, followed by 2500 steps of conjugate gradient minimization. In the first round, the hydrogen atoms were minimized while all heavy atoms were restrained with a constant force of 10 kcal mol−1 Å−2 to their initial positions. In the second round, no restrains were used to allow minimization of the entire system. Thereafter, a two-step equilibration was performed for the systems. In the first part, the temperature was raised from 10 to 310 K within 0.1 ns, and the protein was restrained with a constant force of 5 kcal mol−1 Å−2. In the second step (0.4 ns length), only the Cα atoms of the protein were restrained with a constant force of 5 kcal mol−1 Å−2. Subsequent pHtMD simulations followed a previously established protocol13 and consisted of a series of short CpHMDs (1 ns length each). After each nanosecond at constant pH, the pH value of the solvent was slightly lowered or elevated, and a further nanosecond was simulated using the final coordinates and velocities from the previous simulation round as an input. Using this procedure, a large number (>100) of consecutive 1 ns MD simulations were performed, slightly changing the solvent pH value in a defined direction, thus resembling a classical wet-lab titration experiment. pHtMD simulations of HdeA started at pH 7, at which the maximal stability of the dimer was observed.6 To study progressive destabilization, the pH was decreased to pH 2 with a pH changing rate of −0.02 pH units/ns. For one of the simulations, a reduced changing rate of −0.01 pH units/ns was used. For HdeA, there was already one simulation available from a previous work,13 which was used to supplement the five simulations performed in the present work. Several simulations were extended by short CpHMD simulations (50 ns at pH 2), which were performed in an identical manner as the pHtMD simulations but with the pH changing rate set to zero. As it was not known whether the YmgG dimer stability also is pH-dependent, pHtMD simulations of YmgD started at pH 5, at which the experimental dimer structure was determined. All YmgD calculations were conducted with a pH changing rate of −0.02 pH units/ns. The main set of five simulations was performed for decreasing pH from 5 to 2 to assess whether acid-induced structural changes are observed. In addition, a set of three control simulations was performed in which the pH was increased from 5 to 7 to verify whether there are differences in protein stability over this pH range. All MD simulations were carried out with Amber 12 or Amber 14 16 using the ff99SB 17 force field. The salt concentration was set to 0.1 M (on the basis of DebyeHückel), bonds involving hydrogen atoms were constrained with the SHAKE algorithm, and a time step of 2 fs was used. In addition, the Langevin dynamics with a collision frequency of 2 ps−1 was used. In our simulation setup, the protonation check was performed at each time step of the MD simulation. For all parts of the simulation, the generalized Born implicit solvent model igb = 218,19 and a cutoff of 30.0 Å were used for nonbonded interactions. MD Simulations Performed in Explicit Solvent. MD simulations with explicit solvent were performed using version 14 of the AMBER MD software package16 (ambermd.org) and the ff14SB force field.20 In addition to the preparatory steps described above, the dimeric YmgD system was electrically neutralized with 8 Na+ ions and solvated with TIP3P21 water molecules using the AMBER tool LEaP. Otherwise, system

The aim of the present study was to investigate whether additional HdeA homologs that might also potentially act as pH-dependent chaperones exist. For this purpose, we conducted a search with HdeA (PDB code: 1BG8) for structurally similar proteins against all deposited structures in the PDB. In addition to HdeB, for which the structural homology was already described,7 the functionally uncharacterized protein YmgD (PDB code: 2LRM) was found. The reverse DALI search with YmgD as the input structure revealed that HdeA is the most structurally similar protein to YmgD known today. On the basis of this finding, we investigated whether YmgD exhibits an unfolding behavior similar to that of HdeA. To take into account that unfolding of HdeA is generally slow but is accelerated at low pH, we used our recently developed pHtitrating molecular dynamics (pHtMD) protocol13 for the simulations. pHtMD relies on the overall concept of constant pH molecular dynamics (CpHMD) in implicit solvent but performs a consecutive series of molecular dynamics (MD) simulations with small pH changes, which allows to change the solvent pH over simulation time.13,14 Thus, pHtMD allows for enhanced sampling by using implicit solvent, and it additionally facilitates HdeA unfolding by lowering the pH to acidic values, at which faster HdeA unfolding is observed. To provide a sufficient structural basis for analysis of unfolding, we performed multiple independent pHtMD simulations of HdeA and YmgD. A comparison of these simulations revealed that the structural changes of HdeA and YmgD are markedly different. HdeA displays a coupled unfolding and dissociation mechanism, whereas YmgD exhibits a weak dimer interface resulting in monomer−dimer equilibrium without unfolding of the subunits. In summary, the computational investigations above suggest that YmgD either is not an acid-activated chaperone or, at the very least, does not require unfolding for activation. On the basis of this observation, alternative molecular functions of YmgD will be discussed.



COMPUTATIONAL METHODS Structure Preparation. Simulations of HdeA and YmgD were based on PDB entries 1BG85 (chains A and B) and 2LRM (model 15), respectively. Both crystal structures available for HdeA (PDB codes: 1BG8, 1DJ8) were determined at a pH of ∼4 and contain multiple copies of HdeA per asymmetric unit. Extra chains, water molecules, and ions were deleted from the PDB file. If the terminal residues did not represent the natural termini after the cleavage of the signal peptide, an acetyl group (ACE) was added to the N-terminus and/or an N-methylamine group (NME) was added to the C-terminus with PyMOL version 1.5 (The PyMOL Molecular Graphics System; Schrödinger, LLC, http://www.pymol.org/). The cysteine residues involved in disulfide bonds were renamed from CYS to CYX. In addition, titrating ASP and GLU residues were renamed as AS4 and GL4, respectively, to allow proper treatment within the pHtMD simulations. As none of the proteins investigated contained histidines, this residue type was not considered in the setup of the titratable groups. After these preparation steps of the PDB file, missing hydrogens were added, disulfide bonds were defined, and the topology and the coordinate files were created with tleap. Finally, the cpin file was created with the cpinutil.py program from the Amber 12 MD package.15 In the cpin file, all aspartate, glutamate, and lysine residues were defined as titrating. B

DOI: 10.1021/acs.jpcb.6b06091 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Figure 1. Structure and sequence comparison of YmgD and HdeA. (a) The overlay of the monomers shows that HdeA and YmgD are structural homologs. (b) Superimposition of YmgD (green/yellow-green) with HdeA (blue/light blue) shows that the dimer interface of YmgD has a high similarity to the dimer interface of HdeA. (c) Structure-based sequence alignment of HdeA and YmgD. The sequence numbering refers to that of the mature proteins after the N-terminal signal sequences were cleaved. Identical amino acids are colored in the sequence alignment. The conserved disulfide bond is marked with a bracket. The lower part of the sequence alignment shows the secondary structure assignments by DSSP (H: helix, L: coil). Identical assignments are colored.

nanosecond of simulation. The last frame was taken to allow the structure to adapt its conformation to the respective pH. The Dali server27 (http://ekhidna.biocenter.helsinki.fi/dali_ server/start) was used to compare the protein structures against all other deposited structures in the PDB (performed on 201604-22). Structural alignments were done with the DaliLite pairwise option28 (version 3.1) (http://ekhidna.biocenter. helsinki.fi/dali_lite/start). The PDBe PISA v.1.51 tool (http://www.ebi.ac.uk/msd-srv/prot_int/cgi-bin/piserver) was used to analyze the interface area of the dimers. All visualizations of the molecules were done with VMD 1.9.2,29 and to follow the changes in the structural elements, the secondary structure assignment was updated after each frame with the sscache.tcl script by Andrew Dalke (http://www.ks. uiuc.edu/Research/vmd/script_library/scripts/sscache/). Plots were created with gnuplot 4.6 (http://www.gnuplot.info/).

equilibration was performed as described above for the simulations in implicit solvent. The subsequent production phases (500 ns length) were performed on GPUs22−24 at 310 K and without any restraints. Furthermore, the constant pressure periodic boundary conditions were applied with an average pressure of 1 bar by using the Berendsen barostat with isotropic position scaling (NTP = 1). A Berendsen thermostat was used to maintain the temperature of the system at 310 K. A time step of 2 fs and the SHAKE algorithm for bonds involving hydrogen were applied during the equilibration and the production phases. Structure Analysis and Visualization. The pKa values were calculated for the aspartate and glutamate residues as described in Socher and Sticht.13 For each titratable group, the deprotonated fraction (fdeprot) was used to calculate its pKa value according to eq 1 fdeprot =



1 1 + 10

n(pK a − pH)

(1)

RESULTS AND DISCUSSION

HdeA and YmgD Are Structurally Homologous Proteins in the Periplasm of E. coli. HdeA and YmgD are two periplasmic E. coli (strain K12) proteins of similar size. In addition, the monomeric forms of these proteins are structurally homologous, as proven by a pairwise structure comparison with DALI (root-mean-square deviation (RMSD) between the two monomers: 3.5−3.6 Å, Z-score: 7.7−7.9; Figure 1a). This RMSD is slightly higher, but still in a similar range, than the value obtained between the two acid-activated chaperones HdeB and HdeA (RMSD: 2.2−3.0 Å, Z-score: 8.7− 8.9). HdeA and YmgD both form homodimers by using the same interface (Figure 1b), whereas the homologous HdeB uses a different interface for dimerization. Notably, the

Fits according to this equation allow the estimation of pKa values at the midpoint of titration and of the Hill coefficient, n, which describes the cooperativity of various titrating groups.25 Structural properties such as the number of intermolecular contacts, the radius of gyration of each chain, and the hydrophobic surface were measured with cpptraj.26 For the hydrophobic surface, the solvent-accessible surfaces of the hydrophobic amino acids alanine, valine, phenylalanine, proline, methionine, isoleucine, and leucine were added. The total charge of the HdeA and YmgD monomers was computed from the protonation state of the ionizable groups after every C

DOI: 10.1021/acs.jpcb.6b06091 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B sequence identity between HdeA and YmgD is very low; only 11 residues are identical in the sequence alignment derived from the structural alignment (Figure 1c). Among these conserved residues are also two aspartates (Asp25/Asp69 in HdeA and Asp27/Asp71 in YmgD) as residues with titratable side chains. In addition, the two cysteines are conserved, and the experimentally determined protein structures demonstrate that they form an intramolecular disulfide bridge. The analysis of secondary structure elements with DSSP showed that the secondary structure is substantially more conserved than the primary structure, which is also evident from the 3D superimposition (Figure 1a). Owing to this significant similarity of YmgD to HdeA, which was described to act as an acidactivated chaperone, we hypothesized that the yet uncharacterized protein YmgD might also represent an acid-activated chaperone. To test our hypothesis, we performed pHtMD simulations of HdeA and checked whether the properties found for HdeA can also be observed in pHtMD simulations of YmgD. HdeA Forms Loosened Dimer Structures and Requires Unfolding of Both Subunits before Dissociation. The pHtMD simulations of the dimeric HdeA in implicit solvent started at pH 7, and the solution pH was then lowered over the simulation time to pH 2. The unfolding pathway for one of the simulations is shown in Figure 2, and the structural changes of all six runs are analyzed in more detail in Figures 3 and S1. A more detailed analysis of run 1 is shown in Figure 3 (top panel) revealing that unfolding of the subunits is accompanied by a strong increase of the radius of gyration to values >18 Å. After both subunits are unfolded, the subunits start to dissociate as indicated by the loss of intermolecular contacts. Unfolding and dissociation both result in an increase of hydrophobic surface (Figure S1), which is likely a prerequisite for chaperone activity.30,31 Moreover, the total charge of the subunits was calculated at the end of each nanosecond by analyzing the protonation states of all titratable side chains. As exemplarily shown for chain A in Figure S1, the total charge progressed from negative to large positive values. This evolution was also discussed in the work of Zhang et al.32 as a potential explanation for why the unfolding and dissociation of the dimers is facilitated at low pH values. To validate the observed unfolding pathway observed for run 1, the remaining five pHtMD simulations were analyzed in the same manner. Although unfolding of the subunits starts earlier in the simulation, run 6 (Figure 3) is qualitatively rather similar to run 1 with respect to the sequence of the unfolding events, and unfolding of both subunits occurs before dissociation. In addition, unfolding of the chains starts again at the C-terminal helix. Run 4 (Figure 3) represents a simulation in which dissociation is rather delayed with respect to the unfolding of the chains. In this run, the dissociation is not complete at the end of the pHtMD experiment, as is evident from the remaining intermolecular contacts but can be observed within an additional 50 ns CpHMD simulation at pH 2 (data not shown). For the remaining three runs (2, 3, and 5) unfolding was only detected in one of the subunits (Figure 3). Interestingly, none of these three simulations displayed a dissociation of the subunits. We also extended runs 2 and 3 by 50 ns of CpHMD simulation in the same manner as done for run 4; however, no dissociation was observed. This finding suggests that an unfolding of both subunits is a prerequisite for dimer

Figure 2. Snapshots from the pHtMD simulation (run 1) of HdeA. The secondary structure elements are color-coded: β-sheet (yellow), α-helix (pink), 310 helix (blue), turn (cyan), coil (white). In the snapshots taken at 150 and 200 ns, it can be seen that the partial unfolding starts at the C-terminus, and the two C-terminal salt bridges Asp69−Lys77 and Asp76−Lys79 are not stable over the simulation. At 250 ns, the HdeA dimer is dissociated into mainly disordered monomers.

dissociation. Interestingly, the structural properties observed in these simulations are similar to those of the I2 intermediate described in a recent study of HdeA coupled folding and binding by 2D window-exchange umbrella sampling.12 This observation suggests that the structures in our simulation runs 2, 3, and 5 were trapped in a rather stable intermediate state. As already noted in previous theoretical studies, HdeA undergoes a coupled unfolding and unbinding process with high-energy transition barriers between the different conformations requiring the application of enhanced sampling techniques to monitor conformational transitions.11,12 In our present approach, enhanced sampling is achieved by using an implicit solvent model. In addition, the pH is lowered to acidic values, at which faster HdeA unfolding is observed. Sampling may be further enhanced by increasing the overall simulation time. This is suggested from the results of run 6, in which the pH changing rate was decreased from 0.02 to 0.01 pH units/ns. This simulation shows unfolding and subsequent D

DOI: 10.1021/acs.jpcb.6b06091 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Figure 3. pHtMD simulation of the periplasmic protein HdeA. Radius of gyration for chain A (blue) and chain B (light blue), and number of intermolecular contacts (gray) between chain A and chain B over the simulation. E

DOI: 10.1021/acs.jpcb.6b06091 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Figure 4. pHtMD simulation of the periplasmic protein YmgD. Radius of gyration of chain A (green) and chain B (yellow-green), and intermolecular contacts (gray) between both subunits of the YmgD dimer.

structural property was also observed for the I2 intermediate described in the replica exchange study of Ahlstrom et al.11 and for the I4 intermediate detected in the umbrella sampling simulation of Dickson et al.12 In addition, the dimeric I2 intermediate, which still contains one completely folded

dissociation of both subunits (Figure 3), and the structure is not trapped in the intermediate state observed above in runs 2, 3, and 5. In the present study, unfolding of the subunits generally starts at the C-terminal helix (helix D; residues 74−83). This F

DOI: 10.1021/acs.jpcb.6b06091 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

of the radius of gyration. In some of the simulations, transient unfolding of one or both chains was observed, but in all cases YmgD again adopted more compact conformations during the further course of the simulations. Consequently, YmgD did not show a continuous increase of hydrophobic surface over simulation time as previously observed for HdeA (Figure 3). Similar to HdeA, YmgD is negatively charged at pH 5 and becomes positively charged at low pH. However, the overall positive charge of YmgD at pH 2 is smaller compared with that of HdeA (Figures S1 and S3). The rapid dissociation upon lowering the pH prompted us to investigate whether the same effect is also observed when the pH is increased. For this purpose, three pHtMD simulations were performed in which the pH was increased from 5 to 7. Again, a rapid dissociation was observed (data not shown), suggesting that dissociation is not dependent on pH but rather indicative of a weak dimer interface. To investigate this point in more detail and to exclude that the effects observed critically depend on the implicit solvent model used, additional simulations in explicit solvent were performed. As CpHMD simulations in explicit solvent are computationally rather expensive, we performed conventional constant protonation MD simulations. To select a proper protonation state for the titratable groups, we first calculated the respective pKa values on the basis of the pHtMD experiments according to the strategy described in Socher and Sticht.13 The data in Table 1

monomer, described in the latter work is structurally similar to the conformations detected in our runs 2, 3, and 5. Thus, the present study and the work by Dickson et al.12 provide evidence for the existence of rather stable loosened dimer states that have previously been proposed on the basis of NMR studies.33 This further supports the idea that dimeric in addition to monomeric species may also interact with HdeA substrates.10 We also investigated, whether a decrease of pH during simulation is a prerequisite for observing HdeA unfolding. For that purpose, constant protonation and CpHMD simulations at neutral pH were performed in implicit solvent. Under both conditions, unfolding is observed in some of the simulations following the same unfolding pathway as the pHtMD simulations (Figure S2). This indicates that the enhanced sampling caused by implicit solvent alone is sufficient to detect HdeA unfolding, which prompted us to inspect the experimentally determined dissociation constants for various pH values: The Kd values at pH 7, 4, and 2 are 50, 1, and 0.25 μM, respectively.6 These experimental data indicate that a monomer−dimer equilibrium already exists at higher pH and that HdeA generally exhibits a certain degree of conformational lability, which is enhanced under acidic conditions. The observed high rate of unfolding in our simulations might be attributed to two facts: (1) the implicit solvent generally allows an enhanced sampling compared with simulations in water, (2) dissociation may be further facilitated by the fact that simulations in implicit solvent would correspond to a very low protein concentration due to the lack of periodic boundary conditions thereby favoring dissociation. As a consequence, HdeA exhibits a high conformational lability under these simulation conditions, which leads to fast unfolding and dissociation regardless of pH changes. Therefore, it is not possible to clearly distinguish between a pHindependent unfolding, which is most likely favored by the use of the implicit solvent, and pH-induced unfolding. Although the present data from the pH 7 simulations indicate that use of the pHtMD is no strict requirement to observe HdeA unfolding, we still consider pHtMD as a valid approach because the unfolding pathway detected is in good agreement with previous theoretical studies applying alternative simulation methods (see detailed comparison above).11,12 Thus, we conclude that our approach is at least capable of qualitatively monitoring the sequence of unfolding events in HdeA and is therefore also applicable to probe unfolding of YmgD. More detailed information about the HdeA unfolding pathway and the structure of the intermediates may be gained from CpHMD simulations in explicit solvent.25,34−38 However, one should keep in mind that simulations in explicit solvent are computationally more expensive and exhibit a reduced conformational sampling compared with CpHMD in implicit solvent. Thus, much longer simulation times may be required in explicit solvent, limiting the applicability of such approaches to processes like HdeA unfolding. YmgD Exhibits a Weak Dimer Interface and a High Conformational Stability of the Subunits. For the investigation of the yet uncharacterized protein YmgD, five independent pHtMD simulation runs were performed in an analogous manner as described above for HdeA. YmgD behaved very similarly in all simulations and always showed a rapid dissociation as evidenced by complete loss of all intermolecular contacts (Figures 4 and S3). As a second difference to HdeA, YmgD did not show an irreversible increase

Table 1. Predicted pKa Values of YmgDa

a

residue

pKa from pHtMD predictiona

Glu12 Glu19 Glu22 Asp27 Glu33 Asp47 Glu51 Glu55 Glu62 Asp71 Asp81 Asp82

4.64 4.78 4.21 3.68 4.41 3.72 4.21 3.86 3.97 4.26 3.06 3.28

Averaged over both protein chains and all YmgD simulations.

show that none of the acidic residues has a pKa value >5, indicating that all these residues are predominantly negatively charged in the pH range from 5 to 7. Therefore, these residues were modeled as negatively charged for the MD simulation in explicit solvent. YmgD does not contain histidines; therefore, no particular considerations of the protonation state of this residue type are required. Figure 5 shows the RMSD values and number of interface contacts over the simulation time for the two independent simulations performed. In run 1 (Figure 5a), the RMSD of the subunits is rather constant at ∼3−4 Å, indicating that the subunits remain folded, but the RMSD of the dimer increases to values >20 Å during the simulation. Analysis of the intersubunit interactions reveals that this high RMSD is due to a complete dissociation of the subunits after 220 ns (Figure 5b). After 270 ns a reassociation of the subunits is observed, resulting, however, in a different interface for interaction. This reassociation may be favored by the high protein concenG

DOI: 10.1021/acs.jpcb.6b06091 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Figure 5. Constant protonation MD simulation of YmgD in explicit solvent. (a, c) RMSD values calculated for the dimer (black) and the subunits A (green) and B (yellow-green). Please note the different scale of the y axis in panel c. (b, d) Total number of intermolecular contacts (gray) and native intermolecular contacts (black) between both subunits of the YmgD dimer.

Figure 6. Gly80 in helix D of HdeA generates kinked helices. (a) The HdeA starting structure and (b) the structure within the first nanosecond of pHtMD simulation run 1. The Gly80 residues are depicted as spheres scaled to their van der Waals radii (arrows). Helix representation was done by using Bendix39 and the RGB color scheme. Thereby, red indicates the helix areas with the highest axis angle, followed by green for intermediate angles and blue for very small angles.

trations in the periodic boundary water box, which hinders a further separation of the subunits after dissociation. The conformational instability of the dimer interface also becomes evident from a second control simulation (Figure 5c, d). In this simulation, the increase of the dimer RMSD up to 6−7 Å is less pronounced than that in the first run, indicating that dissociation is not accomplished within 500 ns of simulation time. However, the loss of native intermolecular contacts (Figure 5d) indicates a considerable rearrangement of the subunits. Taken together, both simulations confirm the results of the pHtMD simulations that YmgD exhibits a rather unstable dimer interface, whereas the individual subunits exhibit a rather high conformational stability. All simulations above thus support the idea of a weak dimer interface, resulting in monomer−dimer equilibrium, in which the dimer is predominantly formed at high protein concentrations. One possible explanation for the different interface stabilities of HdeA and YmgD comes from the size of the interface, which is approximately 30% smaller in YmgD compared with that in

HdeA. Consequently, fewer residues contribute to the interface in YmgD than those in HdeA. In our simulation, the larger interface of HdeA was disrupted only after unfolding of both subunits, confirming a coupled unfolding and dissociation mechanism. In contrast, these two processes are not coupled for YmgD, for which dissociation of the subunits occurs more readily in the simulations without loss of the tertiary structure in the subunits (Figure 5). Structure of Helix D Offers an Explanation for the Differences in Subunit Stabilities of HdeA and YmgD. We also investigated potential structural reasons for this higher conformational stability of the YmgD monomers. Firstly, we compared the pKa values of the titratable groups (Table 1) to those of HdeA (Table S1) that were calculated from the pHtMD simulations (Figure S4). For both proteins, the pKa values of the acidic residues are in the same range and only moderately shifted compared with the textbook values. Thus, inspection of the pKa values alone does not provide a clue about the structural origin of the differences in conformational H

DOI: 10.1021/acs.jpcb.6b06091 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B stability. Next, we investigated whether the spatial location and interactions of the titratable residues (instead of the pKa value itself) might cause the differences observed. We focused on helix D because this helix represents the starting point of the unfolding in the HdeA simulations. Helix D of HdeA contains a glycine (Gly80), which is generally considered a helix breaker. In the pHtMD simulations, a kink emerges at Gly80 within the first few nanoseconds of simulation time (Figure 6), suggesting that this site represents a weak point of helix stability. Counteracting features that contribute to helix stability are two salt bridges; one is formed within the helix (Asp76−Lys79) and the other in the sequentially adjacent turn preceding helix D (Asp69− Lys77). Taken together, the data above indicate that helix D of HdeA is intrinsically rather unstable but is stabilized by two local salt bridges (Asp76−Lys79, Asp69−Lys77). Protonation of both aspartates upon lowering pH values will decrease the stability of the helix, thus offering a structural explanation for the pH-dependent unfolding observed in the pHtMD simulations. In contrast, YmgD has no glycine in helix D, suggesting that the intrinsic stability of helix D in YmgD is higher compared with that in HdeA. In addition, the pattern of charged residues and also the interactions formed by these residues are markedly different in YmgD. The positions of the basic residues in the Cterminal region are not conserved (Figure 1c), and YmgD consequently also lacks the two salt bridges found in HdeA. However, YmgD contains an alternative salt bridge (Asp81− Arg46) that anchors helix D to the globular core of the molecule. Notably, Asp81 has the lowest predicted pKa value of all acidic residues in YmgD (Table 1). This indicates that Asp81 remains charged over a wide pH range, thus allowing for a strong ionic interaction with Arg46. Potential Functions of YmgD. The large differences in domain stability observed between HdeA and YmgD clearly argue against the idea that YmgD is an acid-activated chaperone that acts by an HdeA-like mechanism, that is, uses large disordered hydrophobic patches for substrate binding. This raises the question about the molecular function of YmgD, for which at least three different scenarios are feasible: (1) YmgD is an acid-activated chaperone that acts by an HdeB-like mechanism: Like HdeA, HdeB is also an acidactivated chaperone but has a higher pH of 4−5 for optimal activity.8 More importantly, a recent NMR study has shown that HdeB remains a well-folded dimer in the respective pH range but undergoes significant μs to ms timescale conformational exchanges.8 Although the exact mechanism for HdeB chaperone activity still remains elusive, the study above suggests that HdeB activation is coupled to its intrinsic dynamics instead of structural changes. If YmgD were to function by a mechanism similar to that of HdeB, the present simulations performed would probably be too short to detect the functionally relevant conformational exchanges (μs−ms timescale). Thus, longer simulations with enhanced sampling techniques will be needed in the future to study their respective dynamics. (2) YmgD is activated in the basic pH range: On the genomic level, there is one key difference between HdeA/HdeB and YmgD. The YmgD gene is not on the same acid fitness island in the E. coli K12 genome. Whereas the expression of HdeA is upregulated through

acid conditions, the expression of YmgD was described to be strongly induced by a base.40 (3) YmgD has an alternative molecular function: Apart from its structural similarity to HdeA, YmgD shows structural similarity to the immunity protein Rap1a from Serratia marcescens (RMSD: 3.0−3.1 Å, Z-score: 6.5−6.7), which acts as a periplasmic protease inhibitor.41 Like E. coli, S. marcescens is also a Gram-negative bacterium in the Enterobacteriaceae family. The sequence identity of YmgD to Rap1a is slightly lower (11%) compared with that of HdeA (15%). In addition, the position of the disulfide bridge is not conserved between YmgD and Rap1a, which argues against a high functional similarity between YmgD and Rap1a. Rap1a specifically binds the small secreted protease Ssp1.42 Owing to the low sequence identity between Rap1a and YmgD, YmgD is unlikely to bind an Ssp1 ortholog in E. coli. However, YmgD might still interact with other proteins of the E. coli periplasm and control their activity. None of the three hypotheses above can be readily verified or discarded solely by computer analysis and will require further experimental testing. Hypotheses (1) and (2) imply investigating chaperone activity at different pH values, whereas hypothesis (3) may be verified by a search for candidate YmgD interaction partners in the periplasm. On the basis of the outcome of these experiments, simulation methods can then be applied to further unravel the details of the respective molecular processes in a manner similar to that done for the structural homologue HdeA.



CONCLUSIONS YmgD is a structural homologue of HdeA, which is an acidactivated chaperone in the periplasm of E. coli. Like HdeA, YmgD is described as a periplasmic protein of E. coli; however, the function of YmgD is not yet characterized. We have investigated HdeA and YmgD with pHtMD simulations to compare the unfolding behavior of both proteins and to assess whether YmgD also represents an acid-activated chaperone similar to HdeA. Upon decreasing pH, HdeA forms loosened dimer structures, which lead to subunit dissociation after unfolding of both subunits. This is consistent with previous experimental and computational work and indicates that HdeA requires large disordered hydrophobic patches for substrate binding. In contrast to HdeA, which shows coupled unfolding and dissociation, YmgD displays an entirely different behavior. YmgD exhibits a relatively weak dimer interface leading to monomer−dimer equilibrium even at physiological pH values, whereas the individual subunits are stable and do not undergo unfolding. This difference compared to HdeA may be explained by structural alterations within helix D, which represents the starting point of unfolding in HdeA. Helix D of HdeA has a rather low conformational stability due to a glycine within the helix, which is counterbalanced by the presence of two local salt bridges (Asp76−Lys79, Asp69−Lys77). These aspartates become uncharged at low pH resulting in the loss of the salt bridges, thus favoring acid-induced unfolding. In summary, the computational investigations above suggest that YmgD either is not an acid-activated chaperone or does not require an HdeA-like mechanism for activation. On the basis of this observation, three putative molecular functions are I

DOI: 10.1021/acs.jpcb.6b06091 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Acid-Activated Chaperone Allows Promiscuous Substrate Binding. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 5557−5562. (10) Ahlstrom, L. S.; Law, S. M.; Dickson, A.; Brooks, C. L., III. Multiscale Modeling of a Conditionally Disordered pH-Sensing Chaperone. J. Mol. Biol. 2015, 427, 1670−1680. (11) Ahlstrom, L. S.; Dickson, A.; Brooks, C. L., III. Binding and Folding of the Small Bacterial Chaperone HdeA. J. Phys. Chem. B 2013, 117, 13219−13225. (12) Dickson, A.; Ahlstrom, L. S.; Brooks, C. L., III. Coupled Folding and Binding with 2D Window-Exchange Umbrella Sampling. J. Comput. Chem. 2016, 37, 587−594. (13) Socher, E.; Sticht, H. Mimicking Titration Experiments with MD Simulations: A Protocol for the Investigation of pH-Dependent Effects on Proteins. Sci. Rep. 2016, 6, No. 22523. (14) Oecal, S.; Socher, E.; Uthoff, M.; Ernst, C.; Zaucke, F.; Sticht, H.; Baumann, U.; Gebauer, J. M. The pH-Dependent Client Release from the Collagen-Specific Chaperone HSP47 Is Triggered by a Tandem Histidine Pair. J. Biol. Chem. 2016, 291, 12612. (15) Case, D. A.; Darden, T. A.; Cheatham, T. E., III; Simmerling, C. L.; Wang, J.; Duke, R. E.; Luo, R.; Walker, R. C.; Zhang, W.; Merz, K. M.; et al. AMBER 12; University of California: San Francisco, 2012. (16) Case, D. A.; Babin, V.; Berryman, J. T.; Betz, R. M.; Cai, Q.; Cerutti, D. S.; Cheatham, T. E., III; Darden, T. A.; Duke, R. E.; Gohlke, H.; et al. AMBER 14; University of California: San Francisco, 2014. (17) Hornak, V.; Abel, R.; Okur, A.; Strockbine, B.; Roitberg, A.; Simmerling, C. Comparison of Multiple Amber Force Fields and Development of Improved Protein Backbone Parameters. Proteins: Struct., Funct., Bioinf. 2006, 65, 712−725. (18) Onufriev, A.; Bashford, D.; Case, D. A. Exploring Protein Native States and Large-Scale Conformational Changes With a Modified Generalized Born Model. Proteins: Struct., Funct., Bioinf. 2004, 55, 383−394. (19) Onufriev, A.; Bashford, D.; Case, D. A. Modification of the Generalized Born Model Suitable for Macromolecules. J. Phys. Chem. B 2000, 104, 3712−3720. (20) Maier, J. A.; Martinez, C.; Kasavajhala, K.; Wickstrom, L.; Hauser, K. E.; Simmerling, C. ff14SB: Improving the Accuracy of Protein Side Chain and Backbone Parameters from ff99SB. J. Chem. Theory Comput. 2015, 11, 3696−3713. (21) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. Comparison of Simple Potential Functions for Simulating Liquid Water. J. Chem. Phys. 1983, 79, 926. (22) Le Grand, S.; Götz, A. W.; Walker, R. C. SPFP: Speed Without CompromiseA Mixed Precision Model for GPU Accelerated Molecular Dynamics Simulations. Comput. Phys. Commun. 2013, 184, 374−380. (23) Götz, A. W.; Williamson, M. J.; Xu, D.; Poole, D.; Le Grand, S.; Walker, R. C. Routine Microsecond Molecular Dynamics Simulations with AMBER on GPUs. 1. Generalized Born. J. Chem. Theory Comput. 2012, 8, 1542−1555. (24) Salomon-Ferrer, R.; Götz, A. W.; Poole, D.; Le Grand, S.; Walker, R. C. Routine Microsecond Molecular Dynamics Simulations with AMBER on GPUs. 2. Explicit Solvent Particle Mesh Ewald. J. Chem. Theory Comput. 2013, 9, 3878−3888. (25) Swails, J. M.; York, D. M.; Roitberg, A. E. Constant pH Replica Exchange Molecular Dynamics in Explicit Solvent Using Discrete Protonation States: Implementation, Testing, and Validation. J. Chem. Theory Comput. 2014, 10, 1341−1352. (26) Roe, D. R.; Cheatham, T. E., III. PTRAJ and CPPTRAJ: Software for Processing and Analysis of Molecular Dynamics Trajectory Data. J. Chem. Theory Comput. 2013, 9, 3084−3095. (27) Holm, L.; Rosenström, P. Dali Server: Conservation Mapping in 3D. Nucleic Acids Res. 2010, 38, W545−W549. (28) Holm, L.; Park, J. DaliLite Workbench for Protein Structure Comparison. Bioinformatics 2000, 16, 566−567. (29) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual Molecular Dynamics. J. Mol. Graphics 1996, 14, 33−38.

postulated for YmgD: (1) as an acid-activated chaperone that, in analogy to HdeB, does not require unfolding for function, (2) as a base-activated chaperone, or (3) as a protease inhibitor such as the structural homologue Rap1a.



ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcb.6b06091. Hydrophobic surface as well as overall charge from pHtMD simulations of HdeA (Figure S1) and YmgD (Figure S3); radius of gyration and number of intermolecular contacts between both subunits of HdeA in constant pH and constant protonation MD simulations (Figure S2); titration curves of Glu26 in HdeA (Figure S4); experimental and predicted pKa values of HdeA (Table S1) (PDF)



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel: +49-9131-85 24614. Fax: +49-9131-85 22485. ORCID

Heinrich Sticht: 0000-0001-5644-045X Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors gratefully acknowledge the HPC group of the Regional Computing Center Erlangen (RRZE) for computational resources and the Deutsche Forschungsgemeinschaft (SFB796, project A2) for funding. Moreover, the authors would like to thank Victoria Jackiw (Language Center, Univ. ErlangenNürnberg) for reading the manuscript.



REFERENCES

(1) Foster, J. W. Escherichia coli Acid Resistance: Tales of an Amateur Acidophile. Nat. Rev. Microbiol. 2004, 2, 898−907. (2) Richard, H.; Foster, J. W. Escherichia coli Glutamate- and Arginine-Dependent Acid Resistance Systems Increase Internal pH and Reverse Transmembrane Potential. J. Bacteriol. 2004, 186, 6032− 6041. (3) Mates, A. K.; Sayed, A. K.; Foster, J. W. Products of the Escherichia coli Acid Fitness Island Attenuate Metabolite Stress at Extremely Low pH and Mediate a Cell Density-Dependent Acid Resistance. J. Bacteriol. 2007, 189, 2759−2768. (4) Hong, W.; Wu, Y. E.; Fu, X.; Chang, Z. Chaperone-Dependent Mechanisms for Acid Resistance in Enteric Bacteria. Trends Microbiol. 2012, 20, 328−335. (5) Yang, F.; Gustafson, K. R.; Boyd, M. R.; Wlodawer, A. Crystal Structure of Escherichia coli HdeA. Nat. Struct. Biol. 1998, 5, 763−764. (6) Gajiwala, K. S.; Burley, S. K. HDEA, a Periplasmic Protein that Supports Acid Resistance in Pathogenic Enteric Bacteria. J. Mol. Biol. 2000, 295, 605−612. (7) Wang, W.; Rasmussen, T.; Harding, A. J.; Booth, N. A.; Booth, I. R.; Naismith, J. H. Salt Bridges Regulate Both Dimer Formation and Monomeric Flexibility in HdeB and May Have a Role in Periplasmic Chaperone Function. J. Mol. Biol. 2012, 415, 538−546. (8) Ding, J.; Yang, C.; Niu, X.; Hu, Y.; Jin, C. HdeB Chaperone Activity Is Coupled to Its Intrinsic Dynamic Properties. Sci. Rep. 2015, 5, No. 16856. (9) Tapley, T. L.; Korner, J. L.; Barge, M. T.; Hupfeld, J.; Schauerte, J. A.; Gafni, A.; Jakob, U.; Bardwell, J. C. A. Structural Plasticity of an J

DOI: 10.1021/acs.jpcb.6b06091 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B (30) Wu, Y. E.; Hong, W.; Liu, C.; Zhang, L.; Chang, Z. Conserved Amphiphilic Feature Is Essential for Periplasmic Chaperone HdeA to Support Acid Resistance in Enteric Bacteria. Biochem. J. 2008, 412, 389−397. (31) Hong, W.; Jiao, W.; Hu, J.; Zhang, J.; Liu, C.; Fu, X.; Shen, D.; Xia, B.; Chang, Z. Periplasmic Protein HdeA Exhibits Chaperone-like Activity Exclusively within Stomach pH Range by Transforming into Disordered Conformation. J. Biol. Chem. 2005, 280, 27029−27034. (32) Zhang, B. W.; Brunetti, L.; Brooks, C. L., III. Probing pHDependent Dissociation of HdeA Dimers. J. Am. Chem. Soc. 2011, 133, 19393−19398. (33) Garrison, M. A.; Crowhurst, K. A. NMR-Monitored Titration of Acid-Stress Bacterial Chaperone HdeA Reveals That Asp and Glu Charge Neutralization Produces a Loosened Dimer Structure in Preparation for Protein Unfolding and Chaperone Activation. Protein Sci. 2014, 23, 167−178. (34) Baptista, A. M.; Teixeira, V. H.; Soares, C. M. Constant-pH Molecular Dynamics Using Stochastic Titration. J. Chem. Phys. 2002, 117, 4184. (35) Donnini, S.; Tegeler, F.; Groenhof, G.; Grubmüller, H. Constant pH Molecular Dynamics in Explicit Solvent with λDynamics. J. Chem. Theory Comput. 2011, 7, 1962−1978. (36) Wallace, J. A.; Shen, J. K. Charge-leveling and Proper Treatment of Long-range Electrostatics in All-atom Molecular Dynamics at Constant pH. J. Chem. Phys. 2012, 137, No. 184105. (37) Goh, G. B.; Knight, J. L.; Brooks, C. L., III. Constant pH Molecular Dynamics Simulations of Nucleic Acids in Explicit Solvent. J. Chem. Theory Comput. 2012, 8, 36−46. (38) Lee, J.; Miller, B. T.; Damjanović, A.; Brooks, B. R. Enhancing Constant-pH Simulation in Explicit Solvent with a Two-Dimensional Replica Exchange Method. J. Chem. Theory Comput. 2015, 11, 2560− 2574. (39) Dahl, A. C. E.; Chavent, M.; Sansom, M. S. P. Bendix: Intuitive Helix Geometry Analysis and Abstraction. Bioinformatics 2012, 28, 2193−2194. (40) Maurer, L. M.; Yohannes, E.; Bondurant, S. S.; Radmacher, M.; Slonczewski, J. L. pH Regulates Genes for Flagellar Motility, Catabolism, and Oxidative Stress in Escherichia coli K-12. J. Bacteriol. 2005, 187, 304−319. (41) Srikannathasan, V.; English, G.; Bui, N. K.; Trunk, K.; O’Rourke, P. E. F.; Rao, V. A.; Vollmer, W.; Coulthurst, S. J.; Hunter, W. N. Structural Basis for Type VI Secreted Peptidoglycan DL-Endopeptidase Function, Specificity and Neutralization in Serratia marcescens. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2013, 69, 2468− 2482. (42) English, G.; Trunk, K.; Rao, V. A.; Srikannathasan, V.; Hunter, W. N.; Coulthurst, S. J. New Secreted Toxins and Immunity Proteins Encoded within the Type VI Secretion System Gene Cluster of Serratia marcescens. Mol. Microbiol. 2012, 86, 921−936.

K

DOI: 10.1021/acs.jpcb.6b06091 J. Phys. Chem. B XXXX, XXX, XXX−XXX