Structural insights on the efficient catalysis of hydroperoxide reduction by Ohr: Crystallographic and molecular dynamics approaches

Organic hydroperoxide resistance (Ohr) enzymes are highly efficient Cys-based peroxidases that play central roles in bacterial response to fatty acid hydroperoxides and peroxynitrite, two oxidants that are generated during host-pathogen interactions. In the active site of Ohr proteins, the conserved Arg (Arg19 in Ohr from Xylella fastidiosa) and Glu (Glu51 in Ohr from Xylella fastidiosa) residues, among other factors, are involved in the extremely high reactivity of the peroxidatic Cys (Cp) toward hydroperoxides. In the closed state, the thiolate of Cp is in close proximity to the guanidinium group of Arg19. Ohr enzymes can also assume an open state, where the loop containing the catalytic Arg is far away from Cp and Glu51. Here, we aimed to gain insights into the putative structural switches of the Ohr catalytic cycle. First, we describe the crystal structure of Ohr from Xylella fastidiosa (XfOhr) in the open state that, together with the previously described XfOhr structure in the closed state, may represent two snapshots along the coordinate of the enzyme-catalyzed reaction. These two structures were used for the experimental validation of molecular dynamics (MD) simulations. MD simulations employing distinct protonation states and in silico mutagenesis indicated that the polar interactions of Arg19 with Glu51 and Cp contributed to the stabilization of XfOhr in the closed state. Indeed, Cp oxidation to the disulfide state facilitated the switching of the Arg19 loop from the closed to the open state. In addition to the Arg19 loop, other portions of XfOhr displayed high mobility, such as a loop rich in Gly residues. In summary, we obtained a high correlation between crystallographic data, MD simulations and biochemical/enzymatic assays. The dynamics of the Ohr enzymes are unique among the Cys-based peroxidases, in which the active site Arg undergoes structural switches throughout the catalytic cycle, while Cp remains relatively static.


Introduction
Oxidants such as fatty acid hydroperoxides are signaling molecules involved in host-pathogen interactions, and therefore, their levels are strictly controlled by peroxidases and other mechanisms [1][2][3][4]. Ohr (Organic hydroperoxide resistance) proteins are Cys-based, dithiol-dependent peroxidases that display unique biochemical and structural properties [5,6]. Ohr enzymes play central roles in the bacterial response to peroxynitrite and fatty acid hydroperoxides, two oxidants involved in host-pathogen interactions [1]. These enzymes are found in bacteria and fungi, and they are absent in their hosts (plants and animals) [7], making them promising targets for drug discovery. Some examples of pathogenic bacteria that express Ohr proteins are Pseudomonas aeruginosa, Vibrio cholerae and Xyllela fastidiosa [7]. Xylella fastidiosa is a plant pathogen with agronomic interest, causing disease in citrus, grapes and olives [8].
Ohr protein was first identified in Xanthomonas campestris pv. phaseoli due to its involvement in the bacterial response to organic hydroperoxides, but it is not involved in the H 2 O 2 response [9]. This unusual organic hydroperoxide resistance phenotype is related to the ability of Ohr enzymes to reduce organic hydroperoxides with higher efficiency than H 2 O 2 [5,6,10]. Ohr, Prxs (peroxiredoxins) and Gpx (GSH peroxidases) are all Cys-based, thiol-dependent peroxidases; however, Ohr and Prx/Gpx enzymes belong to distinct families, as their biochemical/enzymatic properties and structures are distinct [6,11]. Instead, Ohr proteins share structural and amino acid sequence similarities with OsmC proteins, which were initially related to the bacterial response to osmotic stress [12]. Later, it was demonstrated that OsmC enzymes are also endowed with thiol peroxidase activity [13,14]. Therefore, Ohr/OsmC is a family of Cys-based proteins that also comprise proteins (such as YhfA from Escherichia coli) whose biochemical activity is still unknown [7,12,14].
Proteins belonging to the Ohr/OsmC family display a barrel-like structure formed by a tightly folded homodimer, in which two six-stranded β-sheets wrap around two central α-helices [6,11,15]. The two active sites are located at the dimer interface on opposite sides of the protein, and the reactive Cys, also called the peroxidatic Cys (C p , Cys61 in Ohr from Xylella fastidiosa-XfOhr), is located in one of the central α-helices. C p and two other residues (Arg19 and Glu51 in XfOhr) constitute the catalytic triad. The involvement of catalytic Arg in the ability of Ohr enzymes to reduce hydroperoxides was directly assessed by site-directed mutagenesis in Ohr from Pseudomonas aeruginosa (PaOhr) [6]. The carboxylic group of catalytic Glu orients the guanidinium group of Arg toward C p in a configuration that appears to be optimal for the reduction of organic hydroperoxides [6,11]. Recently, we showed that fatty acid hydroperoxides are biological substrates of Ohr enzymes [1], displaying properties expected for ligands of these enzymes, such as an elongated shape and hydrophobicity. Peroxynitrite is also one of the biological oxidants of Ohr enzymes, but other features are associated with this catalysis [1]. In spite of all these advances, several aspects related to the extremely high efficiency of Ohr enzymes to reduce hydroperoxides remains elusive, such as the possible occurrence of structural movements along the catalytic cycle.
The reaction of hydroperoxides with C p generates a sulfenic acid (C p -SOH), which undergoes condensation with the resolving Cys (C r , which is Cys125 in XfOhr), generating an intramolecular disulfide bond [6,11]. Moreover, the loop that contains the catalytic Arg (herein named the Arg19 loop) was observed far away from C p and the catalytic Glu in the crystal structure of Ohr from Deinococcus radiodurans (DrOhr) [15]. In this case, the two Cys residues form a disulfide bond [15]. Therefore, we previously hypothesized that Ohr enzymes in the socalled "closed state" [6,11] would present catalytic Arg in an orientation able to activate C p for hydroperoxide reduction, whereas Ohr enzymes in the so-called "open configuration" [15] would be more prone to recycling by the reducing substrate. We have since shown that the reducing substrates of XfOhr are lipoylated proteins [10], in contrast to the Prx/Gpx counterparts that are mainly reduced by thioredoxin or GSH [16]. Here, for the first time, we present crystal structures for the same Ohr protein in the open and closed states, allowing for the validation of the in silico simulations. Additionally, to better understand the structural changes during the catalytic cycle, molecular dynamics (MD) simulations were applied to the XfOhr structure in its closed and open states, in distinct protonation and oxidation states, and after in silico mutagenesis. The same mutagenesis was also performed in the recombinant Ohr protein to evaluate its biochemical properties. Among other findings, our results indicate that polar interactions among the C p , Arg19 and Glu51 residues are important to stabilize XfOhr in the closed state, and they are also required to activate the thiolate for hydroperoxide reduction. The disruption of any of these polar interactions releases some of the constraints on the Arg19 loop movement.

Crystallization trials, data collection and processing
The procedures concerning XfOhr expression and purification have been previously reported [5]. XfOhr (10 mg/ml) was treated with 1.2 mM lipoamide at 310 K for 1 h and crystallized using the hanging-drop vapor diffusion method. The optimal crystallization condition was obtained using reservoir solution pH 6.0 (0.1 M sodium cacodylate and 0.4 M sodium citrate). The XfOhr crystal, cryoprotected by the mother liquor solution supplemented with 20% glycerol, was cooled to 100 K in a nitrogen gas stream, and X-ray diffraction data were collected at protein crystallography beam line D03B-MX1 at the Brazilian Synchrotron Light Laboratory, LNLS. The data set was processed using the programs MOSFLM [17] and SCALA [18,19] from the CCP4i package [20].

Structure determination, model building and refinement
The Matthews coefficient (2.18) revealed three Ohr chains per asymmetric unit, and the monomer structure of the XfOhr (1ZB8) was used as a search model in molecular replacement protocols using the program Phaser [21]. The model was constructed by consecutive cycles of manual modelling, using the program Coot [22], and refinement using Refmac [23]. The stereochemical parameters of the final model were evaluated using the programs PROCHECK [24] and WHATCHECK [25]. Cα superposition was performed using Coot [22], and molecular graphical representations were generated using PyMOL [26].

Site-directed mutagenesis
The pET15b/XfOhr plasmid was used as a template to generate the individual Ohr mutants carrying mutations of Arg19 to Ala (R19A) and Glu51 to Ala (E51A). The mutagenesis protocols were performed according to the manufacturer's instructions (Quick Change II Kit; Stratagene) with the following primers: XfOhrR19A_F (5' CAACTGGTGGCGCCGATGGCAGC 3'), XfOhrR19A_R (5' GCTGCCATCGGCGCCACCAGTTG 3'), XfOhrE51A_F (5' GGT ACCAATCCAGCGCAACTGTTTG 3') XfOhrE51A_R (5' CAAACAGTTGCGCTGGATTGGT ACC 3'). The reaction products were treated with Dpn I to remove the parental methylated plasmids, and the E. coli XL1-Blue strain was used as the host and transformed by electroporation. Single colonies were selected and their plasmids were extracted and sequenced with the BigDye Terminator v3.1 Cycle Sequencing Kit using an automatic sequencer, the ABI 3730 DNA Analyzer (Thermo Scientific), to confirm the codon substitutions. The plasmids harboring the mutations were transformed into the E. coli BL21 (DE3) strain by electroporation. The procedures concerning XfOhr mutants expression and purification were the same as for the wild-type.

Lipoamide-lipoamide dehydrogenase peroxidase-coupled assay
The lipoyl peroxidase activity levels of wild-type XfOhr and its mutants (R19A and E51A) were determined as previously described [10]. The reactions were followed by the decay of absorbance at 340 nm (e = 6,290 M −1 Ácm −1 ) due to NADH oxidation. pK a determination by monobromobimane alkylation assay Wild-type XfOhr and its mutants (R19A and E51A) were reduced with 100 mM DTT (dithiothreitol) for 2 hours at room temperature. The DTT excess was then removed by gel filtration (PD-10 desalting column-GE), and the Ohr proteins (10 μM) were incubated with monobromobimane (2 μM) in buffers (50 mM) at different pH values (3.0 to 7.0) for 20 minutes at room temperature. The rates of alkylation by monobromobimane were determined by extrapolation of the maximum inclination of the curves [27]. Subsequently, the pK a values were determined by the Henderson-Hasselbach equation in GraphPad 1 Prism4.

Circular dichroism
All measurements were carried out in Tris buffer (10 mM) pH 7.4, and wild-type XfOhr and its mutants (R19A and E51A) were used at 15 μM. CD spectra were recorded from 180 to 320 nm using a JASCO spectropolarimeter, model J720 at the Central Analítica of IQUSP, SP.

Morph conformations
Morph conformations were generated using UCSF Chimera [28]. For this purpose, we applied the corkscrew interpolation method with 40 interpolation steps and used two crystal structures of XfOhr in its closed (1ZB8) and open states (4XX2) to generate the first set of morph conformations. Subsequently, the first, the average and the last snapshots of the XfOhr-SS trajectory (see below) were used to generate the second set of morph conformations.

MD simulations
The XfOhr structure in the closed conformation (PDB entry 1ZB8, 2.4 Å resolution) was subjected to MD simulation studies in two conditions: (1) in the reduced form (C p as thiolate, Cys61) with 12 crystal water molecules, having B-factors < 25 Å 2 and (2) with an artificial intramolecular disulfide between C p and Cys125, which was built with SYBYL [29]. After the disulfide bond formation, the neighboring residues had their geometry optimized using Tripos force field and the Powell method [30,31]. The XfOhr trajectories in the reduced and oxidized forms were named XfOhr-Sand XfOhr-SS, respectively. Furthermore, an XfOhr trajectory with C p as a protonated thiol (named XfOhr-SH) was similarly built and subjected to MD simulation. To evaluate the roles of the Arg19 and Glu51 residues in the conformational change of XfOhr, the R19A and E51A XfOhr mutants were built in silico using SYBYL-X [32]. The wildtype residues were replaced by alanine, and their neighboring residues were minimized, as described above.
Finally, the open conformation of XfOhr described here (PDB entry 4XX2) was also used as a starting point for MD simulation, having, however, its C p reduced to thiolate by breaking the disulfide bond and deprotonated using SYBYL-X [32]. Subsequently, the geometry of the C p neighboring residues was optimized as described for XfOhr-SS, and the minimized structure was named Open-S -. Arginine and lysine were protonated, whereas aspartic and glutamic acids were deprotonated. Histidine was protonated at its ε-nitrogen atoms.
All MD simulations were performed using GROMACS 4.6.3 [33,34] and G54a7 force field [35]. Force field parameters for cysteine as thiolate were taken from those available for Cys without adding a hydrogen atom to the Sγ atom. Partial charges for Sγ and Cβ atoms were assigned as -0.7 and -0.3, respectively, which correspond to the mean values calculated using the Gasteiger Marsili, Hückel, Pullman, MMFF94, Gasteiger Hückel methods available in SYBYL-X [32]. The starting structure was initially minimized in vacuum, using the steepest descent method and the conjugated gradient algorithm (2000 steps each). The minimized structure was placed in a 100 Å cubic box, solvated with simple point-charge (SPC) water [36] and neutralized by adding sodium ions. Periodic boundary conditions were applied, and all covalent bonds containing hydrogen were fixed at equilibrium lengths using the LINCS algorithm [37]. The particle-mesh Ewald method [38,39] was used and a 9 Å cutoff value was applied for van der Waals interactions. The system energy was further minimized using the steepest descent method and the conjugate gradient method (2000 steps each). Subsequently, a position restraint dynamics simulation was performed for 2.5 ps at 200 K, keeping rigid all protein atom positions. The whole system was heated from 100 K to 300 K over 37 ps, followed by a period of 100 ps of equilibration. The temperature and pressure were kept at 300 K and 1 atm, respectively, by the V-rescale [40] and Berendsen [41] approaches. Subsequently, MD simulations were carried out for 50/150 ns at 300 K. A 2 fs integration time step was used, and configurations were collected every 2 ps.
VMD [42] was used to align all trajectories to their corresponding starting structures. The root-mean-square deviation (RMSD) values of all backbone atoms with respect to the initial conformation were calculated by VMD [42], and their average values were used to determine the overall backbone dynamics. The snapshot closest to the average structures was used as a representative of each simulation. The root-mean-square fluctuation (RMSF) of all protein residues with respect to their average position was calculated with VMD [42] and used to analyze protein residue flexibility. The conformational change of XfOhr in the simulation was followed by measuring the distance between the Arg19-Cα and C p -Cα/Glu51-Cα atoms throughout the simulation time using VMD [42]. For the residues Arg19, Glu51 and C p , the stability of the hydrogen bond interactions was measured by hydrogen bond (Hbond) occupancy throughout the entire trajectory using the default parameters of the VMD hydrogen bond tool (donoracceptor distance and angle values of 3.0 Å and 20˚, respectively). The stability of the saltbridge interactions between these residues was measured considering the distance between all N-O/S pairs throughout the simulation using VMD [42]. These distances were analyzed by Tukey box-plots generated by R [43], and only residues having at least one N-O/S pair whose median distance value was lower than 4 Å [44,45] were considered to be stable. PyMOL [26] and VMD [42] were used for visualization of both the trajectories and the representative structures. MD simulation movies were generated using UCSF Chimera [28].

Crystal structure of XfOhr as a disulfide in its open form
Ohr proteins contain a distinct α/β fold, and there are currently only six structures deposited in the Protein Data Bank. Therefore, it is relevant to make new Ohr crystal structures available for comparative studies. Two of these structures are of XfOhr, and both are in the closed state [11]. A new structure of XfOhr in the open state is described here.
XfOhr was crystallized by the hanging drop vapor diffusion method, and the corresponding crystal belongs to space group C2 (Table 1). A complete data set was collected up to 2.15 Å resolution. The molecular replacement solution contained three monomers in the asymmetric unit. As expected from the previous characterizations [6,11], the overall structure of XfOhr in the disulfide state is an elliptically shaped homodimer. The superposition of the two XfOhr structures (PDB entries 4XX2 and 1ZB8/1ZB9) resulted in an RMSD = 1.21 Å. The XfOhr open-state structure was obtained in the oxidized state (disulfide bond) despite the presence of a reducing agent (dihydrolipoamide) in the solution. It is possible that the growth of the crystal started after the oxidation of dihydrolipoamide. Nevertheless, the same phenomenon occurred with DrOhr, but in this case, DTT instead of dihydrolipoamide was used as the reducing agent [15]. It is well documented that the efficacy of thiols as reductants decreases over time [46].
In the XfOhr open-state structure, the two fully conserved Cys residues are linked by a disulfide bond (Cys61-S-S-Cys125, S1 Fig), and the Arg19 loop is displaced far away from C p (Cys61) in an open configuration (Fig 1A), in contrast to the reported XfOhr closed-state structure ( Fig 1B) [11]. Other differences between the two XfOhr states are: (i) the α-helix that contains C p is slightly bent in the open form and (ii) a Gly-rich loop containing residues 35 to Structural dynamics of Ohr 46 ( Fig 1B), which is referred herein as the Gly-rich loop. In spite of these differences, the overall fold of XfOhr is quite similar to that of other Ohr structures ( Fig 1C).
The main chains of the Arg19 loops of the XfOhr open state and of DrOhr overlapped well (Fig 1C). In the case of DrOhr, it was not possible to assign electronic densities to the side chains of the Arg19 loop [15], whereas the corresponding assignment for XfOhr was possible (S1  In the other structure, XfOhr is in the closed state and Arg19 makes polar interactions with Glu51 and with C p (Fig 1D). The opening of the loop would probably then be facilitated by the loss of some of the polar interactions that occurred when C p was oxidized to Cys-SOH, subsequently forming an intramolecular disulfide bond with the C r (Fig 1A).
To gain insights into the conformational changes between the open and closed states, we made a morph conformations movie using 1ZB8 (open) and 4XX2 (closed) as the reference XfOhr crystal structures (see Fig 2A and S1 Video). This morph conformations movie suggested a concerted movement involving the Arg19 and Gly-rich loops. While the Arg19 loop moved far away from the XfOhr active site, the Gly-rich loop occupied this space, which is near the active site.

Molecular dynamics of XfOhr: From closed to open states
To evaluate the role of polar interactions in the structural movements of the Arg19 loop, MD simulations were performed after "in silico" reconstruction of the disulfide bond between its C p and C r in the XfOhr closed state structure. With this disulfide bond, no polar interactions between C p and Arg19 can occur, and therefore, some of the constraints on the dynamics of the Arg19 loop were relieved (S3 Fig). As a control, the XfOhr closed state in the reduced form (Cys61 as thiolate) was also subjected to MD simulation. Both of these simulations were performed for 150 ns. These two XfOhr simulations in the oxidized and reduced forms of the closed state structure are referred to herein as XfOhr-SS and XfOhr-S -, respectively. The XfOhr-Strajectory displayed an average RMSD = 2.14 ± 0.31 Å, indicating that its overall fold was stable [47] throughout the entire simulation ( Table 2). In contrast, the average RMSD value for XfOhr-SS was 3.24 ± 0.50 Å, which was consistent with some conformational changes taking place. Notably, the RMSD values of the XfOhr-SS structure increased rapidly in the first 30 ns, reaching values up to 4 Å (S4 Fig). Next, we calculated the per-residue Cα RMSF to measure more localized fluctuations along the simulations. For the Arg19 residue, the values were approximately 1.0 and 1.8 Å during the XfOhr-Sand XfOhr-SS trajectories, respectively (Fig 3A). The higher values observed for the XfOhr-SS trajectory than those of the XfOhr-Strajectory indicated that the Arg19 loop underwent conformational movements in the first case (Fig 3A-3C). Notably, the X-ray diffraction data (B-factors) also indicated that the Arg19 loop displayed a higher mobility in the open state than in the closed state (Fig 3D). Although the correlation of the intensities between the MD simulations and the X-ray data is not perfect, the overall profile of peaks and valleys displayed high correspondence (Fig 3A and 3D).
The Arg19 loop was not the region that presented the highest RMSF values. Instead, the residues from positions 33 to 48 (KLSVPQGLGGPGGSGT) in both simulations displayed the highest RMSF values, which is consistent with the fact that this loop (the Gly-rich loop) is mainly composed of short side chain residues (Fig 3A-3C). Residues 71 to 82 and 88 to 98 also displayed a higher mobility than the Arg19 loop (Fig 3A-3C). Moreover, the high flexibility of these three regions was also experimentally observed (c.f., B-factors of Fig 3D).
For further analysis of the dynamics of these loops along the XFOhr-SS trajectory, we made a morph conformations movie using the starting, the average and final conformation observed during this trajectory (Fig 2B and 2C and S2 Video). From this analysis, we again observed a concerned movement between Arg19 and the Gly-rich loops. According to the previous results (Fig 3), the Gly-rich loop movement was more pronounced than the Arg19 loop movement.
Due to its importance for catalysis, the movement of the Arg19 loop was further analyzed by measuring the distances of the Arg19-Cα atom to the C p -Cα and Glu51-Cα atoms throughout the simulations (Fig 4). For comparison, the average distance of Arg19-Cα to Glu51-Cα was 10 Å in the closed-state crystal structure and 18 Å in the open-state crystal structure. Likewise, the average distance between Arg19-Cα and C p -Cα was 6 Å and 15 Å in the crystals structures in the closed and open states, respectively. The XfOhr-Strajectory appears to be more stable in a conformation more similar to the closed-state crystal structure of XfOhr, whereas the XfOhr-SS trajectory displayed more freedom with intermediate distances between  Structural dynamics of Ohr the two crystal structures (Fig 4). These findings further suggested that the Arg19 loop was less constrained in the XfOhr-SS state, being able to move away from the active site. In contrast, the Arg19 loop kept its position close to C p throughout the entire XfOhr-Ssimulation. The polar interactions among C p -Arg19-Glu51 residues were further investigated by analyzing the distances involving atoms of the side chains (Fig 5A and 5B). As expected, the Arg19-Glu51 and Arg19-Cys61 salt-bridge interactions were stable in the XfOhr-Strajectory (median values <4 Å for nearly all N-O/S pairs, c.f., Material and Methods). Likewise, the Arg19-Glu51 and Arg19-Cys61 hydrogen bond (Hbond) interactions were observed throughout the XfOhr-Strajectory, with occupancy values equal to 65 and 37%, respectively (Fig 5C). In contrast, these interactions were unstable or even absent during the XfOhr-SS trajectory (median values > 4 Å for all N-O/S pairs and hydrogen bond occupancy values = 0%,   Fig 5C). Indeed, the corresponding median values of the Arg19-Glu51 and Arg19-Cys61 distances for XfOhr-SS trajectories were high, reaching values of approximately 10 Å (Fig 5A  and 5B).

Fig 4. Distance values between Arg19-Cα and Glu51-Cα atoms (A and B) and between Arg19-Cα and Cp-Cα atoms (C and D) for the XfOhr-S -(black) and XfOhr-SS (red) trajectories (150 ns each
The snapshot closest to the average structure throughout each MD simulation was used to represent the entire XfOhr-SS and XfOhr-Strajectories (Fig 6). No significant difference in the overall structure was observed between the two average structures (Fig 6A, with RMSD = 1.93 Å). Despite these similarities, their Arg19 loops adopted two different conformations. In the XfOhr-Srepresentative structure, the Arg19 loop adopted a closed orientation in both chains, which were very similar to the closed crystal structure (Fig 6B and S3 Video). Furthermore, Arg19 kept stable hydrogen bond and salt-bridge interactions with Glu51 and C p (both identified as stable during the entire XfOhr-Ssimulation). In contrast, the Arg19 loop of the XfOhr-SS representative structure underwent a conformational change, moving away from the active site, which was similar to the conformation observed in the open crystal structure (Fig 6C and S4 Video). Both chains adopted this open state; however, their Arg19 loop orientations were somewhat different (S5 Fig). Another difference observed between the XfOhr-SS and XfOhr-Srepresentative structures is the Glu51 side chain orientation. In the XfOhr-Srepresentative structure, the Glu51 side chain is oriented toward the guanidinium group of Arg19, establishing polar contacts (Fig 6B). On the other hand, in the XfOhr-SS representative structure, the Glu51 side chain is oriented toward the Gly-rich loop (Fig 7C). From a visual inspection of the Glu51 side chain movement along the XfOhr-SS trajectory, we observed that Glu51 initially formed polar contacts with Arg19. However, as Arg19 moved away from the XfOhr active site, the Arg19-Glu51 interactions were lost. As a result, the Glu51 side chain was less constrained, being able to establish hydrogen bond interactions with other residues, in particular with those of the Gly-rich loop. Probably, these changes observed in the Glu51 conformations are part of the concerted movement between the Arg19 and Gly-rich loop described previously (S2 and S4 Videos).
Next, we investigated whether the polar interactions between the Arg19 and C p residues could restrict the movement of the corresponding loop. Therefore, another MD simulation (50 ns) was performed, again with XfOhr in the closed state, but now having both thiols artificially protonated. This simulation is referred to herein as XfOhr-SH. Since the Arg19 loop movement started immediately after the initial 30 ns in the XfOhr-SS trajectory (S4 Fig), we assumed that 50 ns of simulation would be enough to observe similar movement in the XfOhr-SH trajectory. Indeed, the Arg19 loop of both chains moved away from the active site, leaving the Arg19 side chain highly exposed to the solvent (Figs 5D, 5F, 5I and 7 and S5 Video). Furthermore, the Arg19 -Glu51 salt-bridges and hydrogen bond interactions were unstable during the XfOhr-SH trajectory (Fig 5D and 5F, respectively). Moreover, the distances between Arg19-Cα and Glu51-Cα were comparable with those observed for the XfOhr-SS simulation, with values up to 19 Å (S6 Fig). These findings further support the notion that the salt-bridge interaction between Arg19 and the negatively charged C p (Cys61-S -) plays a relevant role in stabilizing the Arg19 loop near the active site. Thus, our simulations indicated that the Arg19 loop movement is constrained by polar interactions between C p -Sand Arg19. Considering the overall XfOhr-SH simulation, the average RMSD values were in between those observed for the XfOhr-Sand XfOhr -SS trajectories ( Table 2). (50 ns each). Distance values (yellow dashed lines) measured for the representative structures: (F) XfOhr-S -(green, values = 2.2-3.7 Å), (G) XfOhr-SS (blue marine, values = 11.9 and 13.1 Å), (H) XfOhr-SH (cyan = 8.0-9.0 Å) and (I) E51A XfOhr-S -(orange, value = 19.0 Å). Each representative structure corresponds to the MD simulation snapshot closest to the average structure calculated for its trajectory (50 or 150 ns). The protein backbone atoms are show in cartoon representation, and the 33-48 loop was removed for clarity. Arg19, Glu51, C p and C r side chain atoms are shown in stick representation, including their polar hydrogen atoms. https://doi.org/10.1371/journal.pone.0196918.g005

Structural dynamics of Ohr
Subsequently, we addressed whether Glu51 could also contribute to the stabilization of the XfOhr structure in the closed state. Thus, the E51A XfOhr-Smutant was artificially built and subjected to MD simulations (50 ns). Again, the Arg19 loop moved away from the active site (Figs 5E, 5F, 5J and 7 and S6 Video). In this case, the movement occurred in an even shorter interval than those observed for the XfOhr-SS and XfOhr-SH trajectories. The highest Arg19-Cα to Ala51-Cα distance observed during the E51A XfOhr-Strajectory was 27 Å (S7 Fig). Furthermore, the Arg19 -C p salt-bridge and hydrogen bond interactions were also unstable during this trajectory (Fig 5E and 5F). Therefore, Glu51 is also crucial for stabilizing the Arg19 loop close to the active site, in this case independently of the C p oxidative state.
Finally, we artificially built a R19A substitution in the XfOhr-Sstructure in the closed state. According to our hypothesis, the mutated "Ala19" loop would move away from the active site. The median distance values for Ala19-Cα -Glu51-Cα and Ala19-Cα -C p -Cα distance values were equal to 16 and 12 Å, respectively (S8 Fig and S7 Video). Interestingly, these values were slightly shorter than those obtained for the E51A XfOhr-Smutant. The more restricted behavior of the R19A mutant could be related to the aliphatic side chain that confers hydrophobic properties to the Ala residue. As a consequence, polar interactions between the Ala19 side chain and solvent water molecules are not favorable, destabilizing the open-state conformation (S9 Fig).

Molecular dynamics of XfOhr: From the open to closed states
Since our MD simulations indicated that upon C p oxidation, XfOhr could move from the closed to the open state, we decided to verify whether the opposite process could occur, i.e., if the reduction of the Cys61-Cys125 bond would lead to the reverse movement (from the open to the closed state). In this case, the XfOhr open state was used as a starting point for MD simulation (150 ns), being named the Open XfOhr-Strajectory. Thus, the disulfide bond was artificially reduced, considering C p (Cys61) and C r (Cys125) as a thiolate (RS -) and a thiol (RSH), respectively.
Unexpectedly, the Arg19 loop did not undergo major movements but remained in its open state throughout the entire trajectory (S8 Video). Indeed, the observed distances of the Open XfOhr-Sstructure were similar to those measured in the XfOhr open-state crystal structure (S10 Fig). One hypothesis for this is that the crystal packing contacts in the crystal structure of the XfOhr open state (4XX2) artificially kept the Arg19 loop orientation more exposed to the solvent than in its native condition. Therefore, the starting structure used for the Open XfOhr-Strajectory would have the Arg19 loop orientation more distant from the active site than the biological one, preventing the closure of the Arg19 loop. Alternatively, entropic factors related to the dehydration of the Arg19 loop may have also prevented XfOhr from assuming the closed state. Indeed, the Ohr active site is surrounded by hydrophobic residues [11]. Furthermore, the Gly-rich loop (comprising residues 33-48) might have impaired the closure of the Arg19 loop. Our MD simulations are consistent with this possibility, as the Gly-rich loop appears to prevent the movement of the Arg19 loop back to the active site by steric hindrance effects (S8 Video), at least during the simulation time employed (150 ns). Further studies are required to understand the possible roles of the Gly-rich loop in catalysis.

Biochemical analysis of XfOhr mutants (R19A and E51A)
It is well accepted that the presence of Arg and Glu in the active site of Ohr in close proximity to C p is an important factors for the high reactivity of this peroxidase toward hydroperoxides [1,6,11]. Furthermore, our MD simulation data presented here indicate that Glu51 and C p are required to stabilize the Arg19 loop in the closed state. Therefore, two mutations (R19A and E51A) were generated by site-directed mutagenesis into the XfOhr recombinant protein to experimentally validate these in silico findings. As expected, compared with the wild-type protein, XfOhr R19A and XfOhr E51A presented only residual dihydrolipoamide-dependent peroxidase activity (Fig 8A). Moreover, these mutations resulted in significant changes in the pK a of the catalytic Cys thiolate 8.09 (± 0.11) for R19A and 7.20 (± 0.11) for E51A) compared that observed in the wild-type protein (5.92 ± 0.11; Fig 8B to 8D). Previously, we also observed that wild-type XfOhr displays an acidic pK a [48]. Here, we show for the first time that mutation of the catalytic Glu impairs the enzymatic activity of Ohr. Previously, the relevance of the catalytic Arg was analyzed by mutation in PaOhr [6]. Therefore, these results are consistent with the proposed roles of Arg19 and Glu51 in catalysis, as well as with our MD simulations. Indeed, the thiolate pK a values for the R19A and E51A mutants are more similar to that of free cysteine [49] than the pK a values corresponding to the wild-type protein. Circular dichroism spectra of the wild-type and mutant proteins in the reduced and oxidized states were very similar, excluding the hypothesis that the R19A and E51A mutations might provoke major problems to the overall structure of XfOhr (S11 Fig).

Discussion
A working hypothesis for the catalytic mechanism of Ohr enzymes is presented in Fig 9. Most likely, the catalytic cycle of Ohr enzymes is more complex, and additional steps occur between the closed (Fig 9i) and open (Fig 9iv) states.
There are six crystal structures of Ohr enzymes deposited in the RSCB Protein Data Bank, and all of them are either in the open or in the closed state that corresponds either to snapshot (i) or (iv) in Fig 9, respectively. The closed state appears to be an optimal conformation for hydroperoxide reduction, as the catalytic Arg (Arg19 in XfOhr) is near to C p , which probably results in increased C p nucleophilicity and ROOH electrophilicity. Indeed, in peroxiredoxins (another type of Cys-based peroxidase), a catalytic Arg plays a similar role [50]. In the open state, the entrance of the active site is wider [10], which might better accommodate lipoylated proteins that are the reducing agent of XfOhr [15]. Prior to this work, no information on intermediate states was available.
Initially, the opening of the Arg19 loop was investigated by MD simulations. The overall fold was stable throughout the XfOhr-SS simulation, but the Arg19 loop underwent an opening movement, among other conformational changes. In contrast, the Arg19 loop was stabilized in the closed form when C p was reduced and unprotonated (Figs 4-6 and S3 Video), which is consistent with a pK a for the thiolate group of C p equivalent to 5.92 (Fig 9B). In contrast, when the thiol group of C p was oxidized or protonated (Figs 4-8 and S6 Fig), Arg19 displayed greater freedom, moving away from the active site (Figs 6 and 7, S4 and S5 Videos). Therefore, according to our hypothesis, the stability of the Arg19 loop depends on the oxidative state of C p .
However, other factors also contribute to the stability of the Arg19 loop in the closed state, and a major one is the polar interaction of Arg19 with Glu51. Indeed, the mutation of Glu51 to Ala resulted in increased mobility of the Arg19 loop (Figs 5 and 7; S7 Fig and S6 Video), even when C p was in the thiolate form. Therefore, the disruption of either the Arg19-C p or the Arg19-Glu51 polar interaction facilitated the opening of the Arg19 loop.
We also performed MD simulations starting from the structure with the Arg19 loop in the open form in an attempt to investigate the closing of this loop. Contrary to our expectations, the Arg19 loop did not return to the closed state in any of the conditions and intervals analyzed (S10 Fig and S8 Video). Possibly, the removal of water molecules solvated to the Arg19 loop is required prior to the approximation of this region toward C p . One hypothesis is that the Structural dynamics of Ohr reducing agent (lipoylated proteins) might assist the closing process of the Arg19 loop. Another possibility is that this movement would require much longer simulation time (> 150 ns) or would be better observed using a different in silico technique, such as Normal Mode analysis [51]. Investigations in these directions are underway. This is the first report that describes a biochemical feature associated with the Gly-rich loop (comprising residues from position 35 to 48) that was part of the XfOhr, which exhibited the greatest flexibility (Fig 3A). Hydrophobicity is another feature of the Gly-rich loop that is highly conserved among Ohr family members [7,12,14]. Remarkably, some of these residues interact with polyethylene glycol by hydrophobic interactions [11]. Indeed, docking studies [1] have indicated that the hydrophobic interactions are major factors for lipid hydroperoxide binding within the XfOhr active site. Interestingly, the position of the catalytic Arg considerably differs between Ohr and OsmC proteins that comprise two of the major sub-families in the Ohr/OsmC superfamily [14]. The catalytic Arg is in the Gly-rich loop in OsmC proteins. Therefore, our studies open new perspectives in the understanding of enzyme-substrate interactions in proteins belonging to the Ohr/OsmC superfamily, which may foster investigations aiming to identify inhibitors of these enzymes.
Furthermore, this study contributes information to help distinguish Ohr from other Cysbased peroxidases, such as Prx. Indeed, we have previously reported that Ohr and Prx display distinct biochemical and structural properties [1]. For instance, Ohr and Prx are not homologous proteins, as they do not share amino acid sequence or structural similarities [7,12,52]. For After peroxide reduction, C p is oxidized to sulfenic acid (SOH), which is then attacked by the sulfhydryl group of the resolving Cys (Cys125 of XfOhr), forming an intra-molecular disulfide. Our working hypothesis is that this condensation reaction releases constraints for Arg19 loop movements. (iv) The last step involves the reduction of the disulfide by a lipoylated protein and a rearrangement of the loop to the close state (taken from [11]). Steps (ii) and (iii) are hypothetical, as substrate and product (respectively) were inserted based on the co-crystallization of PEG with XfOhr [11]. https://doi.org/10.1371/journal.pone.0196918.g009 Structural dynamics of Ohr most Prx enzymes, the reductant is Trx, whereas for Ohr, the reductants are probably lipoylated proteins [10]. In this report, we present other features that distinguish Ohr from Prx enzymes. For instance, it is well known that Prx enzymes switch back and forth between the so-called fully folded and locally unfolded states when catalytic Cys residues undergo large movements to allow disulfide formation, as these two residues are far apart in the reduced state [53]. In contrast, the catalytic Arg remains relatively static throughout the catalytic cycle of Prx. In the case of XfOhr, the two catalytic Cys residues remain relatively static throughout the catalytic cycle, whereas the catalytic Arg19 undergoes movement between the closed and open states. Therefore, distinct mechanisms were selected throughout evolution that allowed for the development of two different systems, operating with extraordinary efficiency in hydroperoxide reduction and attaining rates in the 10 7 −10 8 M -1 s -1 range [1,53,54].
Finally, it is important to emphasize that the high correspondence between the crystallographic and simulation data and the biochemical characterizations indicate the robustness of our analysis. Understanding catalytic cycle dynamics might be relevant for the development of Ohr inhibitors. Since Ohr enzymes are present in pathogenic bacteria and fungi [7] but are absent in their hosts, such as plants and animals, these enzymes might be promising targets for drug design.

Accession codes
The crystal structure in this paper has been deposited in the Protein Data Bank as 4XX2.