Deprotonation states of the two active site water molecules regulate the binding of protein phosphatase 5 with its substrate: A molecular dynamics study

Abstract

Protein phosphatase 5 (PP5), mainly localized in human brain, can dephosphorylate tau protein whose high level of phosphorylation is related to Alzheimer's disease. Similar to other protein phosphatases, PP5 has a conserved motif in the catalytic domain that contains two binding sites for manganese (Mn²⁺) ions. Structural data indicate that two active site water molecules, one bridging the two Mn²⁺ ions and the other terminally coordinated with one of the Mn²⁺ ions (Mn1), are involved in catalysis. Recently, a density functional theory study revealed that the two water molecules can be both deprotonated to keep a neutral active site for catalysis. The theoretical study gives us an insight into the catalytic mechanism of PP5, but the knowledge of how the deprotonation states of the two water molecules affect the binding of PP5 with its substrate is still lacking. To approach this problem, molecular dynamics simulations were performed to model the four possible deprotonation states. Through structural, dynamical and energetic analyses, the results demonstrate that the deprotonation states of the two water molecules affect the structure of the active site including the distance between the two Mn²⁺ ions and their coordination, impact the interaction energy of residues R275, R400 and H304 which directly interact with the substrate phosphoserine, and mediate the dynamics of helix αJ which is involved in regulation of the enzyme's activity. Furthermore, the deprotonation state that is preferable for PP5 binding of its substrate has been identified. These findings could provide new design strategy for PP5 inhibitor.

Introduction

Phosphorylation and dephosphorylation are two fundamental regulation mechanisms in cells. These reversible processes are regulated by two kinds of enzymes: the protein kinases and the protein phosphatases (PPs).1 As a large family of PPs, protein serine/threonine phosphatases (PSPs) include PP1, PP2A, PP2B (also known as calcineurin), PP4, PP5, PP6, and PP7.2, 3 PP5, as one of the important members of PSPs, is widely expressed in mammalian tissues.4 Specifically, it is highly expressed in the human brains where it regulates the phosphorylation of tau protein, whose high level is related to Alzheimer's disease.5, 6 PP5 is involved in cell cycle progression, including cell proliferation, differentiation, survival and death, and the abnormal expression of PP5 could cause diseases such as breast and liver cancers.7, 8, 9, 10

The PSPs members share a conserved sequence (DXH(X)_nGDXXD(X)_mGNHD/E, (X means any amino acid)) in their catalytic domain.11 This conserved sequence forms a βαβαβ secondary structure which contains a dinuclear metal ion binding site. Crystal structures of PP5 have shown that two manganese (Mn²⁺) ions are located in the binding sites [PDB: 1S95,4 5HPE,12 3H60/3H61/3H62/3H63,13 1WAO,14 4ZVZ/4ZX215], which is the same as those of PP1 [PDB: 1FJM,16 1JKT,17 1IT6,18 3E7A19] and PP2A [PDB: 2IE4,20 2NPP,21 3DW822]. The catalytic domain of PP5 shares 35∼45% sequence identity with other family members of PSPs.4 And the superposition of PP5 with PP1, PP2A and PP2B shows that the root mean square deviation (RMSD) for the catalytic domain is small (less than 2Å). Furthermore, PP5 can bind the inhibitor of PP1 and PP2A such as okadaic acid and microcystin.23, 24 These data indicate that PP5 has a similar active site structure to other PSPs, which makes it a general model to investigate the substrate binding.

Recently, a crystal structure of PP5 in complex with a Cdc37 phosphomimetic peptide has been reported,12 but there is still no structure available for PSPs binding with a phospoprotein substrate. However, the crystal structure of PP5 bound with a phosphate ion (PDB: 1S95)4 provides a structural basis for studying the binding of PP5 with its phosphorylated substrate. In the crystal structure (PDB: 1S95), each Mn²⁺ ion is coordinated by six residue ligands forming a slightly distorted octahedral geometry (Fig. 1). The ligands to Mn²⁺ in Mn1 site are D242, H244, D271 (a conserved aspartate coordinating both Mn1 and Mn2), a water‐binding site (N1 site) bridging Mn1 and Mn2, a nonbridging water‐binding site (N2 site) coordinated only to Mn1, and an oxygen atom (OP1) of phosphate ion. The ligands to Mn²⁺ in Mn2 site are D271, N303, H352, H427, N1 water‐binding site, and an oxygen atom (OP2) of phosphate ion. The structure has shown that the phosphate ion coordinates the two Mn²⁺ ions in a bidentate mode, which may illustrate the binding mode of phospoprotein substrate.

Figure 1.

Initial structure of PP5 with its substrate phosphoserine (pSer). The detailed structure of the active site showing the coordination of Mn²⁺ ions (in Mn1 and Mn2 sites) is represented in the enlarged figure. Four possible deprotonation states of the two water molecules (in the N1 and N2 sites) were simulated: (I) OH/OH, a hydroxide ion (OH) in both the N1 and N2 sites; (II) OH/W, a hydroxide ion in the N1 site and a water molecule (W) in the N2 site; (III) W/OH, a water molecule in the N1 site and a hydroxide ion in the N2 site; and (IV) W/W, a water molecule in both the N1 and N2 sites

To date, how PSPs bind and catalyze the phospoprotein substrate is still not clearly known, but several catalytic mechanisms have been proposed.4, 25, 26 The most possible catalytic mechanism of PP5, like other PSPs, counts on a metal‐ligated water for the nucleophilic attack of the phosphoryl group.4 In the active site of PP5, there are two possible candidates as the nucleophile: a bridging water in the N1 site and a nonbridging water in the N2 site (Fig. 1). Swingle et al.4 have indicated that a hydroxide ion (deprotonated water) in the N1 site is the most possible nucleophile. But since hydrogen atoms data are not available in crystal structures, it is hard to identify whether the nucleophile is a water molecule or a hydroxide ion. Recently, a density functional theory (DFT) study indicates that both of the water molecules in the active sites of PP5 can be deprotonated to keep the active site neutral for catalysis,27 which suggests the nucleophile could be a hydroxide ion.

To investigate the effects of deprotonation states of the two water molecules on the binding of PP5 with its substrate, a phosphoserine was first modeled based on the structure of phosphate ion, and then independent molecular dynamics (MD) simulations were performed for the four possible deprotonation states. Our results indicate that it is necessary for the bridging water to be deprotonated to maintain the stable structure of the active site. Compared to the state of both water molecules being deprotonated, the state with a deprotonated bridging water and not deprotonated nondridging water is energetically preferable for PP5 binding of its substrate. In addition, different dynamics of helix αJ in the two deprotonation states were identified, which may reveal the regulation of PP5 activity by the effect of deprotonation states. The findings of this work may be useful for designing novel PP5 inhibitors.

Results and Discussion

To indentify whether the deprotonation states of the two active site water molecules can impact the binding of PP5 with its substrate phosphoserine (pSer) [Supporting Information Fig. S1(a)], four systems were set up to mimic all the possible deprotonation states (Fig. 1): (I) OH/OH, a hydroxide ion (OH, deprotonated water) in both the N1 and N2 sites; (II) OH/W, a hydroxide ion in the N1 site and a water molecule (W) in the N2 site; (III) W/OH, a water molecule in the N1 site and a hydroxide ion in the N2 site; and (IV) W/W, a water molecule in both the N1 and N2 sites.

Starting from different initial velocity, four simulation trajectories of 100 ns were performed for systems I and II, and three trajectories were carried out for systems III and IV. Before data analyses, the equilibrations of the simulations were determined by checking the root mean square deviation (RMSD) of PP5. The time‐dependent RMSD relative to the initial structure was represented in Supporting Information Figure S2. It showed that the RMSD of PP5 reached stable state before 60 ns in all the simulations, thus the last 40 ns trajectories were used for data analyses.

A deprotonated bridging water is required for a stable active site

The active site of PP5 contains two Mn²⁺ ions‐ (Mn1 and Mn2 in Fig. 1) and two water‐binding sites (N1 and N2 sites in Fig. 1) which play essential roles in substrate binding and catalysis. To investigate the structural change of the active site, the distance between the two Mn²⁺ ions was computed and averaged for each system [Fig. 2(a)]. For comparison, we also presented the averaged distance between Mn²⁺ ions calculated from the PDB structures of PP5 (PDB: 1S95,4 5HPE,12 3H60/3P61/3H62/3H63,13 1AWO,14 and 4ZVZ/4ZX215). In addition, the distance of the two Mn²⁺ ions in the reactant complex of the DFT study27 was averaged and shown as a dotted line in Figure 2(a). Results showed that the averaged distance was 3.45 ± 0.08 Å for system I, 3.42 ± 0.09 Å for system II, 3.66 ± 0.11 Å for system III, 3.93 ± 0.11 Å for system IV, and 3.28 ± 0.23 Å for the PDB structures. Compared to that in PDB structures, the distance showed no dramatic change in systems I and II, while it significantly increased in systems III and IV. Since both crystal structures and computational data4, 12, 27 reveal that close distance between the two Mn²⁺ ions is maintained during PP5 binding and catalyzing its substrate, the active site of systems I and II is structurally preferred over that of systems III and IV.

Figure 2.

Structural change of the active site. (a) Comparison of the distance (Å) between the two Mn²⁺ ions in simulation systems with that in PDB structures. * indicates that the difference is statistically significant, while NS indicates that the difference is not significant. The dotted line shows the distance between the two Mn²⁺ ions averaged from the reactant complex in the DFT study (27). (b) Snapshots at 100 ns for the active site of systems I (I‐1), II (II‐1 and II‐2), III (III‐1, III‐2, and III‐3) and IV (IV‐1, IV‐2, and IV‐3). The similar snapshots for systems I (I‐2, I‐3, and I‐4) and system II (II‐3 and II‐4) are shown in Figure S3. A dotted line is drawn between Mn²⁺ ion and the coordinating ligand. The final states of the two water molecules in the N1 and N2 sites are labeled in orange. ‐ indicates that there is no water in the N1 or N2 site. The water molecule in N2 site of system II is not stable and moves away from the binding site in system II″ (containing II‐2 and II‐4). The octahedral geometry for the coordination of Mn²⁺ ion is impaired in systems III and IV due to the increased distance between the two Mn²⁺ ions and the loss of coordination of residue H427 and water molecule (in the N1 and/or N2 site)

In the crystal structures of PP5 bound with its substrates (PDB: 1S95 and 5HPE), each Mn²⁺ ion is coordinated by six ligands which form a slightly distorted octahedral geometry and keep the active site structurally stable. To check whether the deprotonation states can affect the ligand coordinations for the Mn²⁺ ions, the detailed structures of the active site at 100 ns of the simulations were represented in Figure 2(b) and Supporting Information Figure S3. In the figures, a dotted line was drawn if a Mn²⁺ ion is coordinated with a ligand. For system I, all the ligands including the two hydroxide ions coordinated with the two Mn²⁺ ions and maintained an octahedral geometry in all the simulations [I‐1 in Fig. 2(b), and I‐2, I‐3, and I‐4 in Supporting Information Fig. S3]. For system II, two statuses for the coordination of Mn²⁺ ions were found: in II‐1 [in Fig. 2(b)] and II‐3 (in Supporting Information Fig. S3), all the coordinating ligands for Mn²⁺ ions were kept in an octahedral geometry, which is like the coordination of system I; however, in II‐2 [in Fig. 2(b)] and II‐4 (in Supporting Information Fig. S3), the ligands were coordinated with Mn²⁺ ions except for the water molecule in the N2 site since it had moved away from the binding site. Under these two statues, the distance between the two Mn²⁺ ions was not varied. The different coordination of Mn²⁺ ion among the simulations of system II indicates that the binding of water molecule in N2 site is not stable.

The change of coordination for Mn²⁺ ions was observed in the active site of systems III and IV (Fig. 2). For system III, the water molecule initially in the N1 site moved away in all the simulations. Accompanied by the escape of the water molecule, residue H427 lost its coordination to Mn2 in III‐1 and III‐3. It is noted that in III‐3 the hydroxide ion initially binding in the N2 site moved to the N1 site (labeled by yellow arrow), indicating that a hydroxide ion rather than a water molecule is favorable to bind in the N1 site. For system IV, the water molecule initially in the N1 site escaped from the site in all the simulations. And the water molecule initially in the N2 site also moved away in IV‐2 and IV‐3. Similar to III‐1 and III‐3, the loss of coordination of H427 to Mn2 was also observed in IV‐1. Thus, the loss of water molecule in the N1 and/or N2 site of the two systems directly impaired the octahedral geometry for the coordination of Mn²⁺ ions.

Due to the increased distance between Mn²⁺ ions and the loss of coordination of water molecules and/or residue H427, systems III and IV are structurally not suitable for PP5 binding and catalyzing its substrate. Therefore, the deprotonation states of water molecules in these two systems are not considered for further analyses. In addition, to distinguish the distinct binding mode of water molecule in the N2 site, system II is divided into two groups: system II′ (containing II‐1 and II‐3, in which water molecule is bound in the N2 site) and system II″ (containing II‐2 and II‐4, in which water molecule is not bound in the N2 site). For systems I, II′ and II″, there is a hydroxide ion in the N1 site [as shown in the first row of Fig. 2(b)]. The major difference between these systems is the binding status of water in the N2 site: There is a hydroxide ion in system I, a water molecule in system II′, and no water binding in system II″.

The state with a deprotonated bridging water and not deprotonated nonbridging water favors the binding of PP5 with its substrate

To determine which deprotonation state of the two active site water molecules is preferable for the binding of PP5 with its substrate pSer, the averaged interaction energy for systems I, II′ and II″ was calculated and then decomposed based on each residue of PP5 [Fig. 3(a)]. For the interaction energy, a negative or positive value means that the residue favors or disfavors the binding. Results revealed that the interaction energies for R275, R400, N303, and H304 were low in systems I, II′ and II″, indicating that these four residues play important roles in binding. Among these residues, the interaction energy for N303 is similar between systems (–4.32 kcal/mol in system I, −4.40 kcal/mol in system II′ and −3.97 kcal/mol in system II″), thus the interaction difference between the systems mainly came from residue R275, R400 and H304.

Figure 3.

Interaction between PP5 and its substrate pSer in the representative structure of systems I (left), II′ (middle) and II″ (right). (a) Interaction energy (kcal/mol) decomposed based on each residue. Four key residues (R275, R400, N303, and H304) play an important role in binding. The final deprotonation states of the two water molecules in the N1 and N2 sites are labeled in orange. ‐ indicates that there is no water in the N1 or N2 site. (b) Interaction between the four key residues and the phosphate group of pSer. Interactions (with a distance cutoff of 3.5 Å) are represented as dotted lines. (c) Surface electrostatic potential of PP5. Positive potential is shown in blue, while negative potential is shown in red. The binding region for substrate is labeled in a dotted circle. The deprotonation state of the two water molecules in system II′ is energetically favorable for PP5 binding of its substrate

To give a clear view of the interaction between the four key residues and pSer, the representative structures for systems I, II′ and II″ were shown in Figure 3(b). Since the difference between these systems lies in the N2 site, we focused on how the N2 site affected the binding of PP5 with the substrate. We found that N2 site affected the binding through its interaction with residue R275 which plays a dominant role in binding. In system I, the hydroxide ion in N2 site formed a hydrogen bond (H‐bond) with residue R275 [Fig. 3(b)] as the distance between the hydroxide ion and residue R275 was 3.14 ± 0.24 Å [Supporting Information Fig. S4(a)]; whereas in system II′, no H‐bond was formed between the water molecule in the N2 site and residue R275 since their distance was increased to 4.71 ± 0.50 Å [Supporting Information Fig. S4(a)]. Compared to the water molecule in N2 site of system II′, the negatively charged hydroxide ion of system I dragged the positive residue R275 about 0.7 Å away from pSer [the distance between R275 and pSer was 3.61 ± 0.29 Å in system I and 2.90 ± 0.22 Å in system II′ as shown in Supporting Information Fig. S4(b)]. This resulted in a less favorable interaction energy for R275 [–10.88 kcal/mol in system I vs. −16.11 kcal/mol in system II′ in Fig. 3(a)]. For system II″, since there was no water in N2 site, residue R275 was flexible which also caused a less favorable interaction energy (–9.79 kcal/mol) than system II′.

In system I, the four key residues R275, R400, N303, and H304 respectively interacted with an oxygen atom of pSer [left in Fig. 3(b)]. The two hydroxide ions together with D271, D242 and pSer (containing −2 charges) neutralized the active site with the positive residues including R275, R400 and the two Mn²⁺ ions. Compared to system I, a water molecule but not a hydroxide ion is in the N2 site of system II′. Since the water molecule only partially neutralized the active site, the side chain of H304 underwent a conformational change and moved away from pSer [middle in Fig. 3(b)], leading to a less favorable interaction energy for H304 (–0.23 kcal/mol in system II′ vs. −2.50 kcal/mol in system I). It is noted that H304 acts as a general acid and binds the leaving phosphate product in the catalysis.4 The conformational change of H304 in system II′ might make it less suitable for catalysis. In system II″, the water molecule escaped from the N2 site, which subsequently induced an extra positive charge in the active site. To keep the neutral state of the active site, residue R400 moved away from the active site and lost its interaction with pSer [right in Fig. 3(b)], resulting in a less favorable interaction energy for residue R400 (–4.47 kcal/mol in system II″ in contrast to −6.77 kcal/mol in system I and −12.04 kcal/mol in system II′). Residue R400 in system II′ had greater interaction energy than it in system I. This may result from the conformational change of H304 in system II′. Accompanied with the side chain of H304 away from the substrate, R400 moved closer to the oxygen atom of pSer and enhanced its interaction energy with the substrate of system II′. By summing the interaction energies of R275, R400, N303 and H304 (Supporting Information Fig. S5), we found that these four residues in system II′ has greater interaction energy than them in the other two systems (–32.78 kcal/mol for system II′, −24.47 kcal/mol for system I, and −20.96 kcal/mol for system II″).

To further check the interaction between PP5 and pSer, the surface electrostatic potential for PP5 was calculated and shown in Figure 3(c). In the figure, the blue color means positive potential while red color means the negative potential. The electrostatic potential for the pSer‐binding site of PP5 is labeled as a dotted circle in Figure 3(c). It showed that the surface potential for the binding site is almost neutral (white) in system I, while it is a little bit positive (light blue) in system II′, and it is negative (red) in system II″. The positive surface of the binding site in system II′ would favor the binding of PP5 with the negatively charged pSer. Thus, the results suggest that the deprotonation state of active site water molecules in system II′, i.e., a hydroxide ion in the N1 site and a water molecule in the N2 site, is energetically preferable for PP5 binding with its substrate.

For system II, the escape of the water molecule from the N2 site is observed in two of the four simulations. This means the binding of water molecule in the N2 site of system II is less stable than that of hydroxide ion in the N2 site of system I. As this simulation work focuses on the binding of PP5 with its substrate, we do not further investigate the binding free energy of the water molecule. Based on the work of Ribeiro et al.27 and our simulation data, we propose a possible scenario for the binding of substrate and water molecules before the catalysis procedure: PP5 first utilizes the deprotonation state of system II′ (i.e., a deprotonated bridging water and a nonbridging water molecule) to bind the phosphoserine substrate, then the water molecule in N2 site is depotonated to a hydroxide ion or is exchanged with a hydroxide ion in the intracellular solvent, and after that PP5 adopts the deprotonation state of system I (i.e., both of the water molecules deprotonated) to catalyze its substrate. If it is the case, the water molecule in the N2 site needs to be deprotonated before enzyme reaction. The procedure of deprotonation can not be investigated by MD simulation since the bond between the oxygen and hydrogen atoms of the water molecule would be broken. To check whether this happens before PP5 catalyzing its substrate, further study such as using the combined quantum mechanics/molecular mechanics (QM/MM) method is needed.

The dynamics of PP5 can be mediated by the deprotonation states of the two active site water molecules

The flexibility of PP5 was tested by calculating the root mean square fluctuation (RMSF) of the Cα atoms that measures the fluctuation amplitude of each residue over the equilibrated trajectories (Fig. 4). To give a view of how RMSF was changed for the secondary structure of PP5, secondary structure analysis was performed (Supporting Information Fig. S6) and the secondary structure elements indicated in the PDB structure (PDB: 1S95) were shown in Figure 4 and Supporting Information Figure S6, α helices A‐J are shaded in gray and β strands 1–14 are shaded in cyan. Compared to the loop regions, the RMSF values for α helices and β strands were small and kept the same between systems I, II′, and II″, except for helices αD and αJ. These two helices in systems I and II″ had larger fluctuation than them in system II′ (Fig. 4). The secondary structure data also showed that the occupancy for helices αD and αJ varied between systems. The occupancy for helix αD was 42.96% in system I, 44.89% in system II′ and 32.34% in system II″; while that for helix αJ was 72.26% in system I, 36.28% in system II′ and 61.72% in system II″ (Supporting Information Fig. S6). It is noted that the substrate pSer contacted with residues M309 and Y313 of helix αD, and Y313 in turn has hydrophobic interactions with L494 of helix αJ [Supporting Information Fig. S1(b)]. Thus, the deprotonation states of two active site water molecules may alter the flexibility and secondary structure of helices αD and αJ through the interaction between the substrate pSer and helix αD.

Figure 4.

Root mean square fluctuation (Å) for each Cα atom of PP5 calculated from the equilibrated trajectories. The secondary structure elements indicated in the PDB structure (PDB: 1S95) are shown in the shaded regions: α helices A‐J are shaded in gray and β strands 1–14 are shaded in cyan. Helices αD and αJ in system II′ show less fluctuation than them in systems I and II″

To assess the correlated dynamics between residues of PP5, the dynamical cross‐correlation maps (DCCM) was calculated and presented in Figure 5(a). In DCCM, red color means there is correlated motion between residues, while blue color means there is anticorrelated motion between residues. It showed that the most significant dynamical difference between systems was localized in helix αJ [red box in Fig. 5(a)]. Compared to system II″, helix αJ had stronger anticorrelated motions (dark blue) in system I, whereas it had weaker anticorrelated motions (light blue) in system II′. To further investigate the dynamical change of helix αJ, principal component analysis (PCA) was performed [Fig. 5(b)]. PCA is often used to extract protein's collective motion which may reveal the physiological character of long trend dynamics.28 PCA results showed that helix αJ had a leftward motion in systems I and II″, while it had upward motion in system II′. It has been reported that helix αJ plays a role in regulation of PP5 activity.14 The different PCA motion of helix αJ may reveal how the enzyme's activity is mediated by helix αJ under distinct deprotonation states of the active site water molecules.

Figure 5.

Comparison of the dynamical change between simulation systems. (a) Dynamical cross‐correlation maps showing the degree of correlated motion for system I (bottom left in the two figures), system II′ (top right in left figure) and system II″ (top right in right figure). The red color shows the correlation between residues, whereas the blue color shows the anticorrelation between residues. The region of helix αJ is shown in the red boxes. (b) Principal component analysis showing the collective motion of helix αJ in system I (left), II′ (middle) and II″ (right). Different dynamics of helix αJ may reveal its role in regulation of PP5 activity

Conclusions

Four possible deprotonation states of the two water molecules in the active site were simulated to investigate their effect on the binding of PP5 with the substrate phosphoserine. The structural analysis of the active site demonstrated that systems III and IV containing bridging water molecules (not deprotonated) were not suitable for PP5 binding of its substrate since the octahedral geometry for the coordination of Mn²⁺ ions was impaired by the increased Mn‐Mn distance and the loss of coordinating ligands including H427 and/or the two water molecules. Of the two states containing the deprotonated bridging water in the N1 site (i.e., systems I and II), the binding of water molecule in the N2 site (system II) is less stable than that of hydroxide ion (system I), as the water molecules moved away in two of the four simulations (referred as system II″). However, when the water molecule is maintained in the N2 site (i.e., system II′), the state has a favorable interaction energy for the four key residues (R275, R400, N303, and H304) and a positive electrostatic potential for the binding site to attract the negative charged substrate, indicating it is a energetically preferred state for PP5 binding of its substrate. The flexibility, correlated dynamics and collective motion of helix αJ were varied in systems I and II′. This may be caused by the effect of deprotonation states of the two water molecules through the interaction between substrate and helix αD which in turn contact with helix αJ. The molecular insight into the effect of the active site water molecules on the structural, energetic and dynamical change of PP5 may provide new drug design strategy for PSPs inhibitor.

Materials and Methods

Simulation systems

The crystal structure of PP5 bound with a phosphate ion (PDB: 1S95)4 was used as the starting structure. Besides water molecules in the PDB structure, the two catalytic Mn²⁺ ions and the phosphate ion in the active site were retained. Since phosphoamino acid analysis and proteomic analysis have indicated that phosphoserine (pSer) is more abundant than phosphothreonine and phosphotyrosine,29, 30 pSer is used to model the substrate in this work. The phosphate group of pSer is modeled by using the structure of the phosphate ion, and the backbone atoms of pSer were formed automatically using the LEAP module of AMBER10.31 Then pSer is capped with ACE (acetylation of the N‐terminus) and NME (amidation of the C‐terminus) groups to mimic the phosphoprotein substrate [Supporting Information Fig. S1(a)]. In the modeled structure, the phosphate group of pSer is located in a pocket formed by R275, R400, N303, and H304. And the main chain of pSer, ACE and NME are located in a hydrophobic environment formed by M309, Y313 and aliphatic part of R400 [Supporting Information Fig. S1(b)].

Based on the starting model, four systems containing different deprotonation states of the two active site water molecules were set up (Fig. 1): (I) OH/OH, a hydroxide ion (OH, deprotonated water) in both the N1 and N2 sites; (II) OH/W, a hydroxide ion in the N1 site and a water molecule (W) in the N2 site; (III) W/OH, a water molecule in the N1 site and a hydroxide ion in the N2 site; and (IV) W/W, a water molecule in both the N1 and N2 sites. The hydroxide ion was changed manually by deleting a hydrogen atom from the corresponding water molecule.

Each simulation model was placed in a box filled with 11,365 TIP3P water molecules.32 The size of the water sphere was chosen so that the distance between every atom in the protein and the boundary of the water was at least 10 Å. Physiological salt concentration was made by adding 150 mM NaCl in the systems, and counter ions were also added to keep the systems neutral.

Molecular dynamics simulations

All the MD simulations were carried out by AMBER10 programs package31 and the AMBER03 force field33 was used to describe the molecular interactions. The force field parameters for Mn²⁺ ions34 as well as pSer carrying an unprotonated (–2) charge35 were obtained from the AMBER parameter database. The force field parameter for hydroxide ion was the same to the work of Lee et al.36

The simulation procedure is the same as our previous works.37, 38, 39, 40, 41, 42 To remove bad contacts from the initial structures, each system was minimized using 2,500 steps of steepest descent method followed by 2,500 steps of the conjugate gradient method in which the protein was constrained by a force constant of 500.0 kcal/mol‐Ų. Then, another 5,000 steps energy minimization was carried out using the same process but with the constraints switched off. After energy minimization, the whole system was gradually heated from 0K to 300 K for 300 ps under NVT (constant temperature and volume) condition. Subsequently, 100 ns MD simulations were performed in NTP (constant temperature and pressure) ensemble for each system. In the MD simulations, the Berendsen thermostat43 was used to keep the temperature (300 K) and the pressure (1 atm) constant. Particle‐mesh Ewald (PME)44 method is used to treat long‐range electrostatic interactions and the cutoff for the nonbonded interactions is 10 Å. The SHAKE algorithm45 was employed to constrain all bonds involving hydrogen atoms. The time step of the simulations was 2 fs and the trajectories were collected every 1 ps.

Structural and dynamical analyses

All the structural and dynamical analyses for PP5 were performed using the PTRAJ program of AMBER10.31 RMSD for Cα atoms of PP5 were calculated to determine the system's equilibration tendencies and its convergence. To evaluate secondary structural changes of PP5, DSSP method46 was used to determine whether an amino acid residue belonged to an α‐helix or a β‐strand. The occupancy of each residue in an α‐helix or a β‐strand was determined on the basis of the percentage of time that the residue existed in the α‐helix or β‐strand over the simulation. The helix percentage for αD and αJ was calculated by summing the overall helicity of the residues in the helix and then dividing by the residue number of the helix. Clustering analysis was performed to get the representative structure of PP5 using the MMTSB toolset.47 The K‐means algorithm48 based on the RMSD similarity of the structures was used in the clustering analysis. To compare the distance of the two Mn²⁺ ions and the interaction energy between systems, the average values and standard deviation for the variables were calculated. Significant differences for the studied variables were determined using the Student's t test49 with 95% confidence.

To study the flexibility of PP5, root mean square fluctuation (RMSF) was calculated on a residue‐by‐residue basis and averaged over the equilibrated simulations. Dynamical cross‐correlation maps (DCCM)50 between residues were calculated to provide correlated motion between residues and the degree of the correlation. MATLAB51 was used to generate the cross‐correlation plots. Principal component analysis (PCA) was performed to characterize the collective motion of PP5 residues. In MD simulation, both local dynamical fluctuations and collective motions of residues occur simultaneously. PCA is a multivariate statistical method to separate the collective motions from the local dynamics by reducing high‐dimensional motional data sets into a small subset composed of principal components that describe the collective motion.52 In this study, only the first principal component that dominates the collective motion was shown. The electrostatic potential for PP5 was computed by APBS53 and was mapped onto the molecular surface of PP5 using VMD.54 In addition, VMD was also used in the visualization of the configuration of PP5.

Interaction energy calculations

To tell which residue is essential for the binding of the substrate, the interaction energy was calculated and decomposed using the molecular mechanics/generalized Born surface area (MM/GBSA) approach.55 The procedure of interaction energy calculation is similar to our previous work.41 In brief, the MM/GBSA energy was computed by summing the molecular mechanics energy and solvation free energy between PP5 and its substrate. Then, the MM/GBSA energy was decomposed based on each residue of PP5. In the energy decomposition, the equilibrated simulation trajectories were used with all counterions and water molecules stripped.

Supporting Information

Figure S1 shows the initial model of the substrate phosphoserine (pSer). Figure S2 shows the time‐dependent root mean square deviation (RMSD) for PP5. Figure S3 shows the snapshots at 100 ns of simulation for systems I (I‐2, I‐3, and I‐4) and II (II‐3 and II‐4). Figure S4 shows the distance between R275 and water in the N2 site and that between R275 and pSer. Figure S5 shows the sum of interaction energy from residues R275, R400, N303 and H304. Figure S6 shows the secondary structure of PP5.

Supporting information

Supporting Information.