Introduction

Altered metabolism has recently been recognized as a nearly universal trait of cancer cells1. The best-known phenotype of upregulated metabolic profiles in cancer is aerobic glycolysis, known as the Warburg effect, which describes the preference of glycolysis and increased lactate production, regardless of oxygen availability2,3. PGAM1, an important enzyme in glycolysis, is markedly upregulated in a number of human cancers, such as hepatocellular carcinoma4, colorectal cancer5 and urothelial bladder cancer6. Overexpression of PGAM1 is related to the loss of TP53 in cancer cells due to the negative regulation of PGAM gene expression by TP53. Hitosugi and co-workers7 have reported that increased PGAM1 decreases the level of 3-phosphoglycerate (3-PG), thus leading to the disinhibition of 6-phosphogluconate dehydrogenase in the oxidative pentose phosphate pathway (PPP), whereas the increase in 2-phosphoglycerate (2-PG) activates 3-phosphoglycerate dehydrogenase, thereby steering 3-PG in glycolysis for serine synthesis. Inhibition of PGAM1 expression by shRNA or small molecules attenuates cancer cell proliferation and tumor growth. Hitosugi and co-workers have demonstrated the mechanism by which the abnormal expression of PGAM1 benefits growing tumors. Together, these studies suggest that PGAM1 is a promising metabolic target for human cancer treatments.

PGAM1 catalyzes isomerization between 3-PG and 2-PG, which occurs at the branch point during the glycolytic/gluconeogenic pathway and serine biosynthesis. This enzyme is cofactor-dependent and requires 2,3-BPG as a phosphate group donor to initialize the reaction in the first phosphorylation stage during catalysis. 2,3-BPG is also an intermediate of the reaction process. This enzyme and its homologues have been extensively investigated by using basic enzyme experiments and crystallographic studies and determining disease associations. A ping-pong reaction mechanism of PGAM has been proposed several decades ago8,9,10,11,12. In brief, a phosphorus group is covalently linked to histidine 11 (H11) of PGAM1 from 2,3-BPG or other phosphorus group donors. The phosphorus group of phospho-H11 is then transferred to C-2 of 3-PG, thus promoting the formation of 2,3-BPG after 3-PG localizes at the catalytic site of PGAM1. When 2,3-BPG is created in the active site, it may change its orientation and facilitate the re-phosphorylation of H11 with the C-3 phosphate group by 2,3-BPG. Finally, PGAM1 reverts to its activated state and releases the product 2-PG. However, the molecular mechanisms of 2,3-BPG reorientation within the catalytic pocket and how H11 phosphorylation makes PGAM1 more activated remain elusive.

The structural details of PGAM1 are a prerequisite for understanding its catalytic mechanism and may guide suitable chemical entity design for cancer therapy. The available PGAM1 structures reveal a dimerization mode and contain a conserved folding type among other members of the PGAM superfamily13,14,15,16. However, the residues beyond 234 of the C-terminal domain are disordered, as evidenced by weak electron density. Among the PGAM family members, only two structures, Escherichia coli (PDB code: 1e58)14,17 and human bisphosphoglycerate mutase (BPGM, PDB code: 2hhj)18,19, contain a relatively well-ordered C-terminal region. Because the conformations of the C-terminal portion of these two structures are clearly different, they represent two achievable states corresponding to the closed and semi-closed forms of PGAM1. Because of the different conformations and insufficient electron densities in previous structures, the C-terminal region of PGAM1 is believed to be conformationally variable and may play a crucial role in substrate entrance and product release during the catalytic cycle. Lysine residues of the C-terminal domain are conserved and may interact with substrates or residues in the catalytic cavity. The mutase activity of PGAM from Saccharomyces cerevisiae is lost after removal of the last seven residues of the C-terminal region by limited proteolysis, although the phosphatase activity is not affected20. The in vivo evidence from Kowalski et al21 indicates that the C-terminal region of PGAM is regulated by the concentration of exogenous lactate, thus modulating the effectiveness of the glycolytic enzyme complex. Accordingly, the C-terminal region is likely to play an important role during the catalytic cycle of PGAM1, although the detailed mechanism remains to be further investigated.

In this article, we elucidated the role of the C-terminal domain in PGAM1 by performing microsecond (μs)-scale MD simulations and demonstrate the flexible nature of the C-terminal segment. We characterized the structural transition of PGAM1 from the closed state to the open state, which is induced by large motions of the C-terminal region. The electrostatic potentials of the open and closed states were calculated and revealed the positively charged catalytic pocket of PGAM1. Concurrently, the C-terminal segment was determined to interact with residues located in the substrate/cofactor binding pocket and to affect 2,3-BPG binding. The C-terminal segment helps to maintain the closed conformation of PGAM1 and encloses the catalytic pocket, thereby facilitating 2,3-BPG binding and further catalysis. As a case study on a member of the PGAM family, the results presented here underscore the flexible nature of the C-terminal tail and demonstrate the structural transition of PGAM1 from the closed state to the open state. The electrostatic calculations should help guide the design of novel PGAM1 inhibitors. The proposed swing mechanism of the dynamic C-terminal region improves understanding of the PGAM catalytic cycle.

Materials and methods

Model construction and preparation

Owing to the absence of a C-terminal 10-residue tail in the crystal structures of human PGAM1, we chose two structures (PDB entry: 1e58, 2hhj) with a well-ordered C-terminal region as templates for homology modeling. Both structures show reasonable sequence identity (no less than 50%) to the human PGAM1 sequence. Crystal coordinates of 1e58 and 2hhj were obtained from the Protein Data Bank (PDB). A homology modeling approach was used to generate three-dimensional (3D) models of human PGAM1 by using Modeller v1422. Two apo-monomer structures of PGAM1 were generated on the basis of the abovementioned crystal structures. The sequence (Entry Number: P18669) of human PGAM1 containing 254 amino acids was retrieved from UniProtKB23 (http://www.uniprot.org/). During the model building process, 60 comparative models were initially generated and then ranked by using discrete optimized protein energy (DOPE) assessment scores and probability density functions. The model with the best performance was selected for further evaluation to determine the compatibility of various structural parameters with Procheck24. Subsequently, four simulation systems (two apo models and two holo models with cofactor 2,3-BPG) were designed on the basis of the model structures of PGAM1 and the BPGM complex structures. Before simulations, the protonated states of the titratable residues from the models were set according to predicted pKa values with Propka25,26 at a neutral bulk pH. The crystal structure of 2,3-BPG was obtained from PDB entry 2hhj. The parameters for 2,3-BPG were produced by antechamber on the basis of its crystal coordinates. The protonated state of 2,3-BPG was predicted with Epik in the Schrodinger suite27. The RESP charges of 2,3-BPG were derived with antechamber28 and Gaussian 09 software (Gaussian, Inc)29 at the HF/6-31 (d, p) level. Tleap in AmberTools 13 was used to construct all simulation systems, and the amber0330 force field and general Amber force field (GAFF)31 were used to generate the parameters for simulations. In all simulation systems, the protein was submerged in explicit TIP3P water in cubic boxes with an extra 10 Å extension along each axis of the protein. The net charge of the system was neutralized by adding a suitable number of counterions.

Classical molecule dynamics (CMD) simulations

Acpype32 was used to convert the amber topology into Gromacs format. All CMD simulations were performed with the Gromacs molecular dynamics package version 4.5.733,34 using the amber03 force field30. The particle mesh Ewald35 (PME) method was used to account for the long-range electrostatics. A 10 Å cutoff was applied for van der Waals forces and short-range electrostatic interactions. The periodic boundary condition (PBC) was used at 300 K and 1 atm. During the CMD runs, the integration time step was set to 2 fs. The LINCS algorithm36 was applied to constrain the length of the hydrogen bonds. To avoid unfavorable steric clashes, the starting model structure of PGAM1 was energy-minimized by using the steepest-descents algorithm. First, the protein/protein complex in the system was constrained to equilibrate the solvent. Second, the protein was released to minimize the whole simulation system. The 2,3-BPG in holo systems remained fixed during this procedure. Then, the protein complex was maintained without any constraints, and the whole holo system was minimized. Third, each system was gradually heated to 300 K in the canonical (NVT) ensemble, and this was followed by a 10 ns equilibration of the whole system under isothermal isobaric (NPT) conditions. The equilibrated systems for PGAM1 models were subjected to the production period of simulation. During the production period, the Lennard-Jones potential cutoff was set at 0.9 nm. The solvent, ions and protein/protein complex were separately coupled in a thermal bath by using a modified Berendsen thermostat method37 with a coupling time of 0.1 ps. The isotropic pressure coupling approach with a coupling time of 1 ps was used for the equilibration and production runs. Coordinates were saved every 10 ps throughout the production runs.

Trajectory analysis

The root-mean-square deviation (RMSD) is a critical indicator of trajectory stability. Owing to large fluctuations of the C-terminal segment, the Cα of other domains in PGAM1 were selected for RMSD calculations by using the starting structure as a reference. To demonstrate the highly flexible portions, the root-mean-square fluctuations (RMSFs) were calculated per residue during the simulations. Other common structural analyses, such as radius of gyration (Rg) and hydrogen bonds determined. The secondary structure elements of each frame in the trajectory were identified by using the DSSP program38 as previously described39. All desired analyses were performed using the Gromacs package. The molecular viewers VMD and Pymol were used to analyze the protein structure.

Principal component analysis (PCA)

PCA was performed to detect the primary motions of flexible regions in PGAM1 by using the positional covariance matrix of the atomic coordinates and its eigenvectors. The principles of PCA used for MD simulations have been described in detail in previous publications40,41. The PCA applied in this study was performed by using the protocol42 in the Gromacs molecular dynamics package tools. The motions of the protein were analyzed by projecting the top three eigenvectors, including the maximum motions. PCA results were visualized using the Normal Mode Wizard (NMWiz) plugin43 of VMD.

Free energy landscape (FEL) analysis

The FEL of a protein allows us to explore the conformational transition near the native state structures. Here, we use the sampled conformations in apo systems to show the states during the transition pathway. FEL analysis was carried out using a protocol described in previous publications40,41.

Results and discussion

Published crystal structures and the intrinsic sequence of PGAM suggest a flexible C-terminal region

A representative set of available 3D structures of human PGAM1 and its homologous proteins were aligned for comparison (Figure 1A). We found further evidence that the C-terminal tail (residues 234-254) of human PGAM1 structures (PDB code: 1yfk, 1yjx, 4gpi and 4gpz) lacks interpretable electron density. Several homologous structures have a similar lack of density, except for structures of phosphoglycerate mutase from E. coli (PDB code: 1e58, 1e59) and human BPGM (PDB code: 2a9j, 2f90, 2h4x, 2h4z, 2h52, 2hhj and 3nfy). A protein sequence alignment of the C-terminal region was performed (Figure 1B). The C-terminal region is considered a comparatively conserved domain, thus indicating the conservation of function among a set of orthologous proteins. These results directly suggest the highly dynamic nature of the C-terminal region. We hypothesized that the C-terminal segment is of great importance by virtue of its substantial function-related conformational change. Walter et al20 have highlighted the crucial role of the C-terminal region in S cerevisiae through limited proteolysis and site-directed mutagenesis assays. Their results, as well as reported structural information, indicate that the C-terminal tail should have a tendency to be intrinsically disordered. In the next step, we determined whether the C-terminal tail is natively unstructured because of its specific amino acid sequence. The complete amino acid sequence of human PGAM1 was scored by a protein disorder predictor, PONDR-FIT (http://www.disprot.org/pondr-fit.php)44,45. Residues with a VSl2B score44 larger than 0.5 were recognized as disordered. The results clearly showed that the amino acids of the C-terminal segment were highly disordered (Figure 1C). Additionally, to confirm the sequence-dependent conformational variations of the isolated C-terminus, MC simulations were carried out in Campari software46. The MC simulations showed that the C-terminal region undergoes extensive conformational changes covering previously reported configurations, such as that from 2hhj and 1e58 (Figure S1). These preliminary analyses add to current understanding of the conformational flexibility of the C-terminal domain in human PGAM1.

Figure 1
figure 1

Superimposed structures of PGAM and sequence-based prediction of intrinsically disordered regions. (A) A cartoon representation showing the 3D structure alignment of the available crystal structures of PGAMs; structures of human PGAM1: magenta (PDB code: 4gpi), light marine (PDB code: 1yfk); the structure of human BPGM: gray (PDB code: 2hhj); the structure of Escherichi coli PGAM: green (PDB code: 1e58). The image was created with Pymol. (B) C-terminal sequence alignment of PGAMs from human, mouse, Escherichia coli, and Saccharomyces cerevisiae and human BPGM. Motif QGK is the most highly conserved motif within the C-terminal region. Identical residues are shown in red. Similar residues are highlighted in red and boxed in blue. (C) PGAM1 with four disordered regions at residues 31-44, residues 99-117, residues 121-151 and residues 235-254. Residues with a prediction score above 0.5 are predicted as disordered.

PowerPoint slide

Full size image

Intrinsic flexibility of the C-terminus from MD simulations

The sequence of the PGAM1 C-terminal region reflects its function-related dynamic behavior, as indicated by the PONDR-FIT prediction results and MC simulations. Despite the ergodicity of conformational space in the MC simulations, reliable interactive configurations cannot be obtained from the isolated C-terminus without other PGAM1 domains. MD simulations are routinely used computational tools for studying the structural flexibility of macromolecules, as shown previously40,47,48,49,50,51,52,53. To determine the actual dynamic behavior of the C-terminal segment within its surrounding environment, we first performed MD simulations on the apo-PGAM1 models in aqueous solution. Two available homologous structures (PDB code: 1e58 and 2hhj) were selected to construct PGAM1 models, because these two homologous crystal structures share high sequence similarity with PGAM1 and contain a relatively complete C-terminal region that adopts distinguished conformations. Thus, we built two systems to solve the two apo model structures in water, namely, 1E58-apo and 2HHJ-apo (Table S1). The model building procedures and system preparations are described in the Methods section. For each system, a 1.5-μs CMD simulation was carried out. The structural features of PGAM1 were analyzed using the Gromacs build-in tools as previously described. During the simulations, the 1E58-apo system reached a stable condition up to 300 ns according to the RMSD diagram of the C-α backbone of the protein (the C-terminal region was not included) over time (Figure S2). However, the RMSD of the 2HHJ-apo model without the C-terminal region reached a relatively stable state after 1200 ns. The RMSF values of the two models demonstrated that the C-terminus undergoes large fluctuations during simulations (Figure S3A). Moreover, the central kink and back helix regions (residues 133–137) were detected as fluctuating regions, as were the helix region (residues 12-23), which have previously been reported to be flexible16. These conformational changes may be involved in the interaction networks induced by motions of the C-terminal tail. The secondary structure of the C-terminal segment was inspected for a detailed analysis of its organizational diversity during the simulations. The results of both systems indicated the flexible nature of the C-terminal tail in PGAM1. There was no fixed secondary structure in either system for residues 246 to 254. However, in contrast with the 1E58-apo system, the 2HHJ-apo system contained more helical regions (residues 243–245) than did the 1E58-apo system until 1200 ns. During a simulation time from 1200 ns to 1500 ns, both C-terminal regions reached the same state of equilibrium with a similar percentage of helical elements, in agreement with the crystal structure of human PGAM1 (PDB code: 1yfk, 1yjx, 4gpi and 4gpz), in which the C-terminal segment is presented as a helical element between residues 236-243 and lacks electron density beyond residue methionine 243. These simulations further suggest the flexible nature of the C-terminal region in PGAM1. Thus, we concluded that the C-terminal region of PGAM1 is indeed a flexible part that exhibits diverse conformational changes.

Interconversions between the open and closed state

Significant motions of the C-terminal region should reflect its related biological functions. To extract the large-scale motion modes and explore their functional relevance, we applied PCA of the MD trajectories across the two apo systems. We selected the top three principal components (Figure 2A–C), which separately accounted for 44.3%, 33.8% and 4.1% of the overall motions in the 2HHJ-apo system. The first component (PC1) indicated that the C-terminal region gradually extends into the bulk solvents and moves away from the central kink region. Simultaneously, the central kink region moves opposite from the motion direction of the C-terminus. Nevertheless, PGAM1 displays a common closed conformation in the presence of 2,3-BPG. According to the second component (PC2), the C-terminal region shows a strong tendency for the system toward the open state, which includes a more disordered C-terminal region, as compared with the closed state of PGAM1. In the third component (PC3), the central kink region attempts to restore its orientation to the closed state. These motions reveal new conformational space of the structural transitions. The FEL was calculated according to the plot selecting PC1 and PC2 as the reaction coordinates (Figure 2D, 2E). The FELs of the two apo models provide a direct view of the intrinsically disordered C-terminal region and a process of state conformational exchange between the open (I in Figure 2D, 2E) and closed (III in Figure 2D and IV in Figure 2E) state. Our MD simulations of the PGAM1 systems probed the open forms of PGAM1, which have not previously been identified. These observations also address the extensive conformational transition from the catalytically active (closed) state of PGAM1 to an open state. The contour map of free energy estimation indicates that the open state is the most frequently distributed state among the conformational states from apo-PGAM1 simulations (II and III in Figure 2D; IV in Figure 2E). It also shows that the open state should be the native conformation when PGAM1 is in an unbound state, thereby explaining the disordered tail in the crystal structures of PGAM1. As described in Figure 2D, the initial structure of the 1E58-apo model has a less closed conformation than that in the 2HHJ-apo model. Notably, the C-terminal tail (residues 243-254) reshapes its coordinates and orientation (II in Figure 2D). The conformational changes in the central kink region are coupled with fluctuations of the C-terminal tail. Consequently, the motions of these portions uncover the substrate-binding pocket, thus serving to recruit substrates or release products. Moreover, the information provided by the free energy landscape also reflects the ability of PGAM1 to sustain the open/closed conformation. Replica exchange molecular dynamics (REMD) simulations were performed to enhance the conformational sampling of PGAM1. The results reproduced the large-magnitude functional motions from CMD simulations and emphasized the elastic characteristics of the C-terminal region in PGAM1 (Figure S4).

Figure 2
figure 2

The C-terminal region is involved the primary motions and transition pathway of PGAM1 during the MD simulations. (A-C) Motions corresponding to PC1-PC3 of PGAM1 in the 2HHJ-apo system. PC1-PC3 directions are shown by yellow, blue and magenta arrows. (D, E) Bottom, energy landscape for the conformational transition pathway along projections onto the PC1 and PC2 of systems 1E58-apo and 2HHJ-apo; top, representative snapshots of the local energy minima during simulations of the two apo systems.

PowerPoint slide

Full size image

We also calculated the electrostatics of the open/closed state by using APBS (see Figure S5). When the C-terminal region covered the catalytic pocket and PGAM1 was in the closed conformation, the majority of the basic residues located in the catalytic pocket were wrapped by the C-terminal domain. Nevertheless, the basic residues were exposed to solvents in the open conformation (Figure S5B). The open pocket with high positive electrostatics would quickly attract the negatively charged substrate 3-PG, which contains a phosphate group. After the substrate binds to basic residues, such as K100, R116 or R117, the C-terminal region rescues its conformation in the closed state and maintains a necessary biochemical environment for enzyme catalysis. Several inhibitors have been reported to target PGAM17,54,55,56. MJE3 covalently inactivates PGAM1 at K100. However, it has been speculated that PGMI-004A binds exclusively to an allosteric pocket on PGAM1. Owing to the positive potential of the catalytic pocket and knowledge of the flexible C-terminus, more attention should be paid to determine the regular function of PGAM1. Only an electronegative molecule could occupy the substrate/cofactor binding pocket, thereby blocking access of the substrate to the catalytic pocket. Some non-specific polyanionic compounds have been described as mimics of the substrate that collapse the phospho-enzyme intermediate56. Therefore, analogs of the substrate or cofactor may act as effective inhibitors. Moreover, the design of a novel allosteric or covalent inhibitor is a promising strategy to mediate the biological activity of PGAM1.

Key residues involved in the dynamic mechanisms of PGAM1

To explore the mechanisms underlying the dynamic nature of PGAM1, we carefully examined all snapshots of trajectories from MD simulations. We found that the coordinates of K100 were associated with the conformational change in the central kink region. We then calculated the distance between K100 and H11. H11 is in the center of the catalytic pocket; therefore, the position of H11 was chosen as the reference coordinate set. We observed that the distance from the snapshots over the simulation time was larger than that from the initial model structures (Figure 3A). At the end of the simulations, the semi-open form of the central kink region had an increased kink angle compared with the initial model structures. The movement of this part further exposes the substrate-binding cavity composed of basic amino acid residues and may be the key process underlying product release. R116 and R117 are two flexible residues that are involved in the binding of cofactor 2,3-BPG and substrate 3-PG17,19. Owing to their critical roles in the catalytic cycle, detailed hydrogen bond analyses were performed to determine whether they are affected by other domains. R116 forms hydrogen bonds with its adjacent residues, such as K113 and V112. Moreover, K100 was also found to contact R116 through hydrogen bonds. We observed that R117 interacts with E203 and E207. In the MD trajectory of the 1E58-apo system, these hydrogen bonds are fleetingly present throughout most of the simulation time but especially before 400 ns. However, during the simulations of the 2HHJ-apo system, hydrogen bonds between R117-E203/E207 are present only before 800 ns (Figure 3B). When R117 is captured by E203 or E207, it cannot facilitate recognition and binding to the cofactor or substrate. These interactions inhibit a substrate from entering the enzyme's active site as well as product release. Residue K254 forms hydrogen bonds with R117 within a short period, thus indicating the interplay between the C-terminal region and the substrate-binding pocket.

Figure 3
figure 3

Dynamic features of key residues during MD simulations. (A) Left, distance between K100 and H11 in the 1E58-apo (orange) and 2HHJ-apo (cyan) systems versus simulation time. Time-smoothed values of distance are shown on the light background (raw data), 100 raw data points were averaged at a time; right, distance between K100 (cyan) and H11 (orange). (B) Left, time evolution of the hydrogen bond existence map of R117 in the 1E58-apo (top) and 2HHJ-apo (bottom) systems. Rows correspond to various hydrogen bonds formed. A hydrogen bond is defined when the acceptor atom lies within 3.5 Å of the donor atom in a cone of 120°. Right, R117, and its related residues are shown as sticks with hydrogen bonds depicted as green dashed lines.

PowerPoint slide

Full size image

Holo system analysis

The apo systems provided with abundant, intrinsically dynamic information on PGAM1 in the unbound state. It is unclear whether more dynamic features are induced when the cofactor binds in the catalytic pocket of PGAM1. Specific questions regarding the underlying mechanism of critical residues for cofactor binding and catalytic activity remain unknown. To understand the conformational changes and the precise molecular mechanisms involved, we built two holo systems, 1E58-holo and 2HHJ-holo (Table S1), to investigate the effects introduced by the intermediate product, 2,3-BPG. We speculated that such a strategy (including the intermediate) may be necessary to obtain new information with respect to the interactions between the C-terminal region and the catalytic pocket of PGAM1. Details of model construction can be found in the Methods section. To determine the stability of PGAM1 in the four simulated systems, we compared their gyration radius changes over simulation time. The compactness of protein structures from the 2HHJ-holo system was homogeneous. The C-terminal region underwent large motions in the two 1E58-based systems. In the 2HHJ-apo system, the gyration radius was similar to that in the 2HHJ-holo system, which underwent large-amplitude fluctuations during the simulation and reached the same level of the two 1E58-based systems. We then clustered the trajectories of all holo systems and found three major conformations of the complex structures (Figure 4B–D). As mentioned above, a large conformational transition was observed in the 1E58-holo system. The representative conformation observed during the early simulation time is described in Figure 4B. The C-terminal tail generated the open state conformation, as demonstrated in Figure 4C, even in the presence of 2,3-BPG. This configuration reveals that the C-terminal tail is not necessary for the phosphatase activity of PGAM1, as corroborated by a previous study20. The structure of the cluster center in the compact 2HHJ-holo system is shown in Figure 4D. This structure shows few differences with the complex structure of human BPGM.

Figure 4
figure 4

Structural characterization of the C-terminal region and its influence on 2,3-BPG binding in two holo systems. (A) Time evolution of Rg in 1E58-apo (orange), 2HHJ-apo (cyan), 1E58-holo (purple) and 2HHJ-holo (gray) systems. (B-D) Typical conformations of PGAM1 with various C-terminal tails in the holo systems. 2,3-BPG is indicated by sticks. The central kink region is colored cyan, and the C-terminal tail is colored orange. (E-F) Hydrogen bond occupancy between 2,3-BPG and key residues for substrate/cofactor binding in 1E58-holo (E) and 2HHJ-holo (F) systems. A hydrogen bond is defined when the acceptor atom lies within 3.5 Å of the donor atom in a cone of 120°.

PowerPoint slide

Full size image

Furthermore, the occupancy of hydrogen bonds present in the 2,3-BPG binding pocket was analyzed during the simulation period. Key residues that contact 2,3-BPG were identified (Figure 4E-F). On the basis of the hydrogen bond occupancy, we determined that 2,3-BPG has dissimilar binding modes in the two holo systems. R116, R117 and K100 play dominant roles in the interaction with 2,3-BPG in both systems, with an occupancy rate as high as 35%. Their positive electrical potential stabilizes the negatively charged carbonyl oxygen and phosphate oxygen in 2,3-BPG, which is the direct reason for their role during substrate/cofactor binding. Hydrogen bonds between 2,3-BPG and arginine 10 (R10) appeared frequently in the trajectory of the 1E58-holo system; however, these interactions were not often observed in the 2HHJ-holo system. This phenomenon indicated that 2,3-BPG slightly shifts outward when PGAM1 is in the open state. In the case of the 2HHJ-holo system, we found that glutamine 249 (Q249) is the critical anchor for the interaction between 2,3-BPG and the C-terminal region. Q249 may interact directly with 2,3-BPG via the formation of a hydrogen bond, thus prompting partial portion adjustments of the catalytic pocket. Q249 can also form hydrogen bonds with R116. Additionally, Q249 is a perfectly conserved site in the homologous protein (Figure 1B), thus suggesting that Q249 may be crucial for enzyme function. In summary, the C-terminal region of PGAM1 exhibits remarkable conformational freedom. We confirmed that the flexible C-terminal region serves as a modulator affecting the binding mode of the substrate or intermediate product.

Swing model of the C-terminal region during the catalytic cycle

On the basis of the evidence in this study and previous reports20, we propose the following swing model to describe the functional role of the C-terminal region within the catalytic pathway (Figure 5). The C-terminal region reorganizes its secondary structure and exhibits large-scale conformational transitions (swing) between the open and closed configurations in the apo state. The dynamics of other domains (the kink region and its adjacent helix region) are accompanied by movement in the C-terminus. In the 1E58-holo system, however, 2,3-BPG can localizes in the binding cavity without large motions, even when the C-terminal region is in the open state, thus indicating the minimal role of the C-terminal domain in phosphatase activity. Additionally, the 2HHJ-holo simulation system indicates that the C-terminal region is primarily invoked as a means to constrain structural flexibility to stabilize the binding mode of the intermediate so that R116, R117 and other basic residues interact favorably with 2,3-BPG. Overall, we observed the dynamic motions of the C-terminal region of PGAM1 and described its role in different stages of catalysis by using computational methods.

Figure 5
figure 5

Swing model of the C-terminal region during the catalytic cycle. The surface area of the C-terminal region is colored purple, and its secondary scaffold is shown in light green. The kink region is colored light blue, and its adjacent helix region (residues 12–23) is filled with yellow. Other parts of PGAM1 are shown in gray. 3-PG, phosphohistidine 11 and 2,3-BPG are represented as sticks.

PowerPoint slide

Full size image

Conclusions

Crystal structures of the C-terminal region of PGAM1 do not show a fixed 3D structure, and residues within the C-terminal tail are strongly predicted to be natively disordered, according to PONDR-FIT. Here, different conformations of the C-terminal region were explored by using MC simulations. The results are in agreement with the available structure of the C-terminal region. We also validated the flexible nature of the C-terminal region and revealed the transition pathway of PGAM1 from the closed state to the open state, by using CMD simulations. The open or closed state of the C-terminal region may be involved in events associated with substrate binding or product release during catalysis. On the basis of the MC and CMD simulations, which were operated with different force fields, we found that the C-terminal region is inherently plastic, in a manner dependent on its amino acid sequence. Thus, the dynamic features of the C-terminus are not the result of force field deficiencies.

Considering the relationship between the C-terminal region and the catalytic core, we performed CMD simulations of complex systems. In the 1E58-holo system, the coil of the C-terminal tail underwent large-amplitude fluctuations and ultimately generated the open state conformation, even in the presence of 2,3-BPG. In contrast, most conformations in the 2HHJ-holo system were structurally stable, in accordance with the crystal structure of the BPGM complex in the closed state, thus demonstrating that the C-terminal region may play a different role in the presence of cofactor 2,3-BPG. Moreover, some residues of the C-terminal region were determined to interfere with substrate/cofactor binding, in agreement with previous reports.

Furthermore, it is difficult to identify small-drug molecules that target the binding cavity, owing to its physicochemical properties. Hence, an allosteric inhibitor or a covalent modulator may provide a new direction for drug design against PGAM1. In summary, this simulation study sheds new light on the spontaneous interactions of the flexible C-terminus and active pocket of PGAM1 and its homologous proteins. Future efforts will focus on the process of 2,3-BPG repositioning (i.e., transferring the C3 phosphate to H11) by using MD simulations.

Author contribution

Cheng LUO and Lu ZHOU designed the study; Shi-en LIU and Cheng LUO performed the simulations, analyzed the data and wrote the paper; Jun-chi HU, Hao ZHANG, Wei WAN, Pan XU, Ming-yue ZHENG, Kun-qian YU, Hong DING and Hua-liang JIANG discussed and edited the manuscript. All authors contributed to writing the manuscript and have approved the final version of the manuscript.