Computer‐aided identification of a novel pyruvate kinase M2 activator compound

Abstract

Objectives

The aim of this study was to obtain antitumour molecules targeting to activate PKM2 through adequate computational methods combined with biological activity experiments.

Methods

The structure‐based virtual screening was utilized to screen effective activator targeting PKM2 from ZINC database. Molecular dynamics simulations were performed to evaluate the stability of the small molecule‐binding PKM2 complex systems. Then, cell survival experiments, glutaraldehyde crosslinking reaction, western blot, and qPCR experiments were used to detect the effects of top hits on various cancer cells and the targeting specificity of PKM2.

Results

Two small molecules in 1,5‐2H‐pyrrole‐dione were obtained after virtual screening. In vitro experiments demonstrated that ZINC08383544 specifically activated PKM2 and affected the expression of upstream and downstream genes of PKM2 during glycolysis, leading to the inhibition of tumour cell growth. These results indicate that ZINC08383544 conforms to the characteristics of PKM2 activator and is potential to be a novel PKM2 activator as antitumour drug.

Discussion

This work proves that ZINC08383544 promotes the formation of PKM2 tetramer, effectively blocks PKM2 nuclear translocation, and inhibits the growth of tumour, and ZINC08383544 may be a novel activator of PKM2. This work may provide a good choice of drug or molecular fragments for the antitumour strategy targeting PKM2. Screening of targeted drugs by combination of virtual screening and bioactivity experiments is a rapid method for drug discovery.

1. INTRODUCTION

Energy of tumour cells is mainly from glycolysis, even under the condition of sufficient oxygen, the way to access energy is called aerobic glycolysis, also known as the Warburg effect.1, 2, 3 Pyruvate kinase (PK) regulates the final rate‐limiting step of glycolysis.4, 5 PK has four kinds of isozymes PKM1, PKM2, PKL, PKR.6 PKM1 and M2 are derived from the variable splicing of PKM pre‐mRNA. PKM2 is mainly found in the proliferation of tissues,3, 6 such as normal proliferation, embryonic stem cells, and tumour cells. PKL, PKR, and PKM1 are highly active tetramer in the cells. PKM2 has a low‐active dimeric state and a highly active tetrameric state, and there is a dynamic equilibrium between them. Compared with normal tissues, dimeric PKM2 is more common in tumour cells. Recent studies have shown PKM2 as a new tumour marker,7, 8, 9, 10 and PKM2 mediates an important link between aerobic glycolysis and tumour development. Therefore, it is known that PKM2 is an important target for tumour treatment. At present, antitumour drug research strategies targeting PKM2 are mainly achieved by downregulating PKM2 expression, inhibiting enzyme activity,11 and activating the dimer activity.12, 13 Although several PKM2 inhibitors have been identified, there are always some concerns in targeted PKM2 inhibition. On the one hand, PKM2 is expressed in some normal tissues as well. On the other hand, knockout of PKM2 gene cannot completely inhibit the proliferation of cancer cells. Previous studies have shown that compared with PKM1, PKM2 can be converted from tetramer to monomers to facilitate nucleus entry, and thus promote aerobic glycolysis and cell cycle progression.14, 15 PKM2 could directly interact with Histone H3 and phosphorylate Histone H3‐T11 at the activation of epidermal growth factor receptor,14, 16, 17 and nuclear PKM2 could phosphorylate signal transducer and activator of transcription 3 (STAT3) at Y705 to activate mitogen/extracellular signal‐regulated kinase transcription 5 (MEK5), thereby contributing to cell proliferation.18 Early studies15, 19 have pointed out the role of PKM2 in the nucleus, PKM2 coactivates β‐catenin to induce c‐Myc expression, resulting in upregulation of glucose transporter 1 (GLUT1) and lactate dehydrogenase A (LDHA). PKM2 activator can induce the formation of tetrameric PKM2 and prevent PKM2 from entering nucleus from cytoplasm, thus inhibiting the function of PKM2 in nucleus. Moreover, it was also found that PKM2 activator could enhance the allosteric activation of fructose‐1,6‐diphosphate (FBP) to PKM2.20 In summary, the treatment strategy of activating PKM2 is supposed to be a good choice to suppress tumour. Based on the above‐mentioned ideas, researchers have designed and synthesised a variety of PKM2 activators,21, 22, 23, 24 such as dual sulphonamides, sulphonyl indoles, and thieno pyrrolopyrazine ketones. Generally, the binding site of PKM2 activator was obtained through structural studies and high throughput screening could be used to get lead compounds, and structure activity relationship analysis could facilitate the design and synthesis of new small molecule compounds that have promoting effect on PKM2.22, 23, 24

In recent decades, the virtual screening of drugs has become a new way of drug discovery and has been widely applied given its high throughput, fast speed, and low cost.25 The therapeutic strategy of activating PKM2 may rectify tumour cells to normal cellular metabolism level and restore the state characteristics of normal differentiated cells. The state of PKM2 high active tetramer structure can inhibit the survival of tumour cells. In this work, the structure‐based virtual screening combined with molecular dynamic simulation was performed to identify potential activator compounds targeting dimeric PKM2. Screening known compounds from the compound library is faster and saving, while enriching effective compounds. Then, the enzyme activity was detected after candidate activators treated with PKM2 and the cell proliferation inhibition effect of candidate activators was evaluated in several cell lines. The effects of candidate activators on PKM2 nuclear function and binding specificity were tested (Figure 1). We find that ZINC08383544 can specifically activate PKM2, effectively block nuclear translocation of PKM2 and inhibit tumour growth.

Figure 1.

Experimental flow chat

2. MATERIALS AND METHODS

2.1. Computer‐aided virtual experiment

2.1.1. Initial structure preparation

The X‐ray crystal structure of PKM2 (PDB code: 3ME3)21 was obtained from the RCSB protein databank ( www.rcsb.org). We chose A and B subunits as a receptor molecule. (Figure S1) The 3D structure data set of candidate compounds was downloaded from specs database subset in ZINC database ( http://zinc.docking.org/catalogs/specs, updated on 2014‐09; pH: 6~8), which contained more than 30 000 compounds that could be directly used for molecular docking.

2.1.2. Molecular docking

Molecular docking was performed by UCSF DOCK (version 6.5, http://dock.compbio.ucsf.edu/).26 The Dock Prep tool of the Chimera program (http://www.cgl.ucsf.edu/chimera)27 was used to prepare the receptor files. The docking pose was ranked based on grid score, and Hawkins GB/SA Score model28 was used to evaluate the docking conformation. The Hawkins GB/SA score is an implementation of the Molecular Mechanics Generalized Born Surface Area (MM/GBSA) method.28, 29, 30 The embedded ligand 3SZ in the crystal structure was set as a reference. After sorting the scores, we selected compounds with higher scores than 3SZ for further analysis.

2.1.3. Docking model verification

In our study, we also considered the credibility of molecular docking. In order to verify the reliability of this screening method, we performed a molecular docking simulation and analysed the receiver operating characteristics (ROC).31, 32 We established a test data set containing positive activator and decoys. The composition of the positive ligand set was derived from literatures21, 23, 24, 33, 34 (Table S1), and the mol2 (3D) format of all compounds was converted to SMILES string by Open Babel35 toolbox. DecoyFinder1.136 was used to construct a set of 390 negative ligands. The docking simulation test was carried out using the same method and parameters as previous docking screening. The ROC curves and areas under the ROC curves (AUC) were calculated by the pROC37, 38 package in the R software39 environment.

2.1.4. Molecular dynamics simulations

In order to further evaluate the binding activity and interaction of these potential candidate compounds, 50 ns molecular dynamics (MD) simulation was performed. Root mean square deviations (RMSDs) of complexes backbones in MD simulation can reflect the state of conformation dynamics of the complexes. In view of the natural allosteric activator FBP of PKM2, we performed the MD simulation using PKM2‐FBP and PKM2 without FBP, respectively. All the MD simulations took advantage of the GROMACS 4.5.5 program40 adopting the Amber ff99SB force field41 and the TIP3P water mode. The topological parameters of small molecules were generated using antechamber42 and tleap tools in AmberTools12,43 and the PYthon Parser interfacE (ACPYPE) tool44 was used to convert the AnteChamber to a file suitable for Gromacs.

Then, we carried out energy minimisation for all atoms in MD.45 During the simulation, all bond lengths were constrained adopting the LINCS algorithm46 and long‐range electrostatics were calculated using the particle mesh E _wald (PME)47 method. Energy and trajectory information was collected once per 2 ps for further analysis. The trajectory information was analysed using GROMACS utilities, UCSF Chimera and VMD (www.ks.uiuc.edu/Research/vmd).48 Root mean square fluctuations (RMSF) of each residue and RMSD of the backbone were calculated by g_rmsf software and g_rms software (version 4.5),40 respectively.

2.1.5. Binding‐free energy calculation and energy decomposition

In this work, the single trajectory approach was applied to estimate the energy. The binding‐free energy (ΔG _bind) of each PKM2‐ligand complex was calculated by the g_mmpbsa package (http://rashmikumari.github.io/g_mmpbsa/).49 g_mmpbsa is developed using two widely used open source software, namely, GROMACS45 and Adaptive Poisson‐Boltzmann Solver (APBS) (http://www.poissonboltzmann.org/).50 The tool calculates components of binding energy using molecular mechanics Poisson Boltzmann surface area (MM‐PBSA) method except the entropic term and energetic contribution of each residue to the binding using energy decomposition scheme. 2 ns of the equalised trajectory (Table S4) in MD were chose to calculate the MM/PBSA binding‐free energy. The binding‐free energy includes the vacuum potential energy plus the salvation free energy and subtracts the product of the absolute temperature and entropy change. Binding‐free energy can be expressed in the following formulas.

$$\mathrm{\Delta }{G}_{\mathrm{bind}}=G_{\mathrm{complex}}-G_{\mathrm{protein}}-G_{\mathrm{ligand}}=E_{\mathrm{MM}}+G_{\mathit{sol}}-T\mathrm{\Delta }S,$$ (1)

there

$$E_{\mathrm{MM}}=E_{\mathrm{bonded}}+E_{\mathrm{nonbonded}}=E_{\mathrm{bonded}}+(E_{\mathrm{vdW}}+E_{\mathrm{elec}})$$ (2)

$$G_{\mathrm{sol}}=G_{\mathrm{polar}}+G_{\mathrm{nonpolar}}$$ (3)

Since there are a lot of residues in the receptor protein PKM2, the contribution of entropy needs a lot of counting time and the accuracy in computational is not high; herein, we excluded the entropy.31, 49

2.2. Experimental test and verification

2.2.1. The effect of small molecular compounds on the activity of human‐PKM2

In order to verify the effect of candidate compounds on PKM2 activity and anticancer effect, we carried out relevant biochemical experiments. Small molecules were purchased from the SPECS database ( http://www.specs.net/). Firstly, we used enzyme‐linked immunosorbent assay (ELISA) to detect the activation of PKM2 by candidate compounds as activators. The ELISA experimental protocol is shown in the supplementary document.

2.2.2. The effect of small molecular compounds on the proliferation of cancer cells

We adopted common cancer cell lines HeLa, HCT116, SKBR3, HepG2, and H1299 as the experimental groups, and the normal healthy liver cell LO2 as normal control for testing the effect of candidate compounds on cells. Cell viability was measured in 96‐well plate using the Cell Counting Kit (CCK‐8 Kit) (KGA317, Nanjing keygen).

2.2.3. The effect of small molecular compounds on PKM2 nuclear translocation

Although the role of nuclear PKM2 is not dependent on the enzyme activity,18 the activated PKM2 is a tetrameric conformation, which affects the function of nucleus PKM2. Therefore, in this work we measured the changes of PKM2 content in the nucleus and cytoplasm of HeLa and HepG2 cells by western blot. We used the homologous crosslinking agent glutaraldehyde51 to cross link the human‐PKM2 after treated with candidate molecules and carried out Sodium dodecyl sulphate‐polyacrylamide gel electrophoresis to detect the molecular weight of protein after crosslinking. In addition, phosphorylation of histone H3‐T11 and STAT3‐Y705 that might be affected by nuclear PKM2 level was examined, and the expression of GLUT1 and LDHA was also measured by quantitative real‐time polymerase chain reaction (qPCR). Antibody sources and the information of primers are displayed in the supplementary document.

2.2.4. PKM2 mediates the inhibition of cancer by small molecule compounds

In order to further explore whether candidate compounds can specifically target PKM2, we used siRNA to knockdown PKM2 in HeLa and HepG2 cells. The negative control group and the experimental group were set up. The information of siRNAs are displayed in the supplementary document. After these siRNA were transfected into cells 24 hours, cells were treated with candidate compounds, and then, cell viabilities were detected by CCK‐8 test 24 hours later.

2.3. Statistical analysis

All the presented data and results were confirmed in at least three independent experiments. These data were expressed as the mean ± SD. Statistical comparisons were made by analysed with one‐way analysis of variance, correlation analysis, and multiple comparisons. P < 0.05 was considered statistically significant.

3. RESULTS

3.1. Two kinds of small molecules were screened out as PKM2 activators from docking

After two steps scored that using the same parameter setting with the docking process, we calculated the AUC of each docking score method. Generally, the closer AUC is to 1, the better performance of the classification method.31 As shown in Figure 2, the AUC values of grid score and Hawkins GB/SA score were 0.8147 and 0.9169, respectively. These two values had shown that our model and method for screening in this work have a good ability to distinguish between true and false.

Figure 2.

ROC curves of the simulated docking test. Docking test set includes 15 positive ligands from the literature (Table S1) and 390 decoys constructed from DecoyFinder1.1

According to the above docking method, we screened out 65 small molecules (Table S2) with higher affinity after removing duplication and arranged them in order of their toxicity. In addition to two molecules without LD₅₀ information, there were five compounds ZINC08383507, ZINC08383544, ZINC08695186, ZINC00850070, and ZINC06149696 with higher LD₅₀. By comparing the structural characteristics of these five compounds (Table S3), we found that ZINC08383507, ZINC08383544, and ZINC08695186 were 1,5‐dihydro‐2H‐pyrrol‐2‐ones and the others were 3,7‐dihydro‐1H‐purine‐2,6‐dions (xanthines) (Figure 3A). These compounds were significantly enriched in anticancer compounds targeting PKM2 from docking.

Figure 3.

2D structures of five small molecules (A) and the backbone RMSD of six small molecule‐PKM2 complexes (B) in 50 ns

3.2. MD analysis identified two small molecules as candidate compounds of PKM2 activator

Root mean square deviations is the main reference standard to evaluate the stability of protein‐ligand system in the 50 ns dynamic simulation. We compared the fluctuations of RMSD values of five small molecule compounds, respectively (Figure 3). As the whole, these RMSD values were less than 0.4 nm, suggesting that the dynamics of receptor‐small molecule complexes were reasonable, which can provide a reasonable basis for the following studies. In Figure 3, the backbone RMSD of PKM2‐3SZ complex has been larger and continued to rise. RMSD fluctuations of PKM2‐ZINC00850070 and PKM2‐ZINC06149696 complex are relatively explicit during the entire MD process and other three small molecules with PKM2 all had a relatively stable equilibrium period. Therefore, we chose three small molecules ZINC08383507, ZINC08383544, ZINC08695186 for the following analysis.

Calculation of the RMSF of each dynamic simulation system was based on a period of equilibrium time (Table S4) of the trajectory file. RMSFs of the three compound‐protein complexes residues were essentially in agreement (Figure 4A). Three complex systems in the vicinity of the substrate‐binding site and the ATP‐binding site showed different advantages (Figure S2). As for allosteric site (Trp482 and Arg489, Figure S3), three complexes loaded with or without FBP showed lower RMSF values than 3SZ.

Figure 4.

The contribution of key residues in protein stability and ligand‐binding affinity. A, RMSF of PKM2 residues during 2 ns equilibrium of MD simulations. B, Ligand‐residue interaction spectra for 3SZ‐ PKM2 and candidate compound‐ PKM2 complexes. C, 2D diagrams of the interactions of PKM2 with small molecules, plotted using LIGPLOT+. Hydrophobic interactions are represented by red arcs. Hydrogen bonds are indicated by broken green lines. Ligands are represented in purple. C, N, and O atoms are shown in black, blue, and red, respectively. The red circles and ellipses highlight the key residues identified in binding‐free energy decomposition

The predicted binding‐free energy and energy contribution were calculated in Table 1. The binding‐free energies of small molecules ZINC08383507, ZINC08383544, and ZINC08695186 were −55.929 kJ/mol, −84.285 kJ/mol, and −43.438 kJ/mol, showing ZINC08383507 and ZINC08383544 had better binding ability than others. As shown in Figure 4B, the energy contribution was mainly concentrated on several special amino acids complexes, while the contribution of energy in PKM2‐ZINC08695186 complex is obviously different from that of other compounds. In Figure 4B, Phe26‐His29, Phe308‐Lys311, Ala350‐Asp354, and Tyr390‐Glu397 both in the chain A and B are significantly involved in the combination of small molecules and PKM2 and participate in the formation of binding‐free energy compared to other amino acids. The energy changes near binding sites of PKM2‐ZINC08383507, ZINC08383544 complexes were consistent with the changing trend of PKM2‐3SZ complexes, indicating that the energy contribution of these complexes had good performances in terms of system stability maintenance.

Table 1.

The predicted binding‐free energy and energy contribution (KJ/mol)

Small molecular	ΔE vdW	ΔE ele	ΔE solv	ΔG SASA	ΔG bind
3SZ	−260.809 ± 8.634	−53.759 ± 7.876	241.207 ± 15.435	−22.460 ± 0.701	−95.822 ± 14.721
ZINC08383507	−296.694 ± 10.193	−95.320 ± 14.950	363.390 ± 23.775	−27.305 ± 0.813	−55.929 ± 28.033
ZINC08383544	−317.205 ± 13.682	−49.523 ± 6.121	309.585 ± 12.808	−27.142 ± 0.648	−84.285 ± 14.431
ZINC08695186	−266.686 ± 5.994	22.452 ± 27.971	228.452 ± 16.808	−27.657 ± 1.199	−43.438 ± 23.967
ZINC06149696	−261.940 ± 5.554	−13.603 ± 8.077	310.815 ± 27.686	−25.084 ± 0.850	10.188 ± 32.368
ZINC00850070	−307.959 ± 11.978	−41.718 ± 22.210	347.361 ± 13.193	−25.805 ± 0.570	−28.120 ± 21.794

Energy and trajectory information was collected every 8 ps and 12 times in total. (n = 12, E = mean ± SD).

ΔE _vdW: van der Waals energy contribution from MM; ΔE _ele: electrostatic energy in the gas phase as calculated by the MM force field; ΔE _solv: sum of the non‐polar and polar contributions to solvation; ΔG _SASA: non‐polar contribution to the solvation‐free energy calculated by solvent‐accessible surface area (SASA); ΔG _bind: final estimated binding‐free energy calculated from the terms mentioned above.

The important interactions identified in the PKM2 binding system of the key amino acids in the MD process were analysed and presented by the LIGPLOT⁺ software (version 1.4.5).52 As shown in Figure 4C, residues that interact with the small molecules are shown visually. Compared with the original ligand 3SZ, those in the red circle are important amino acids involved in the interaction of activation sites, which is basically the same as the amino acids that have a greater contribution to the energy in Figure 4B. Although more key amino acids appear in the PKM2‐ZINC08695186 complex, the binding energy is too high, so the other two were selected as candidate verified compounds.

3.3. Two candidate compounds can activate PKM2 and inhibit the proliferation of cancer cells

According to the enzyme activity test, the activity of 15.6 μg/mL human‐PKM2 was 6.5 U/mL. The results of small molecules on PKM2 activity are shown in Table 2, and two candidate compounds can enhance the activity of PKM2. In the cytotoxic test of small molecules on cancer cells, these two candidate compounds could inhibit the growth of several cancer cells, but had less inhibition effect on normal liver cells LO2. Moreover, the two kinds of small molecules had a dose‐dependent effect on the growth inhibition of cancer cells, in which the IC₅₀ value of ZINC08383544 on HeLa and HCT116 cells was below 100 nmol/L (Figure 5A).

Table 2.

The effect of small molecule compounds on activity of PKM2

Molecule ID	PKM2 activity (U/mL)
	0	50 nmol/L	100 nmol/L	150 nmol/L	200 nmol/L	250 nmol/L	300 nmol/L
ZINC08383507	6.5 ± 0.053	6.53 ± 0.096	6.87 ± 0.061*	7.13 ± 0.056**	7.27 ± 0.122*	7.72 ± 0.132**	8.45 ± 0.168*
ZINC08383544	6.5 ± 0.053	6.82 ± 0.081	7.54 ± 0.121**	8.18 ± 0.062**	8.25 ± 0.072**	8.67 ± 0.108**	9.63 ± 0.044**

Each sample was detected three times. U = mean (n = 3) ± SD. Statistical analyses were performed using one‐way ANOVA. *P < 0.05, **P < 0.01.

Figure 5.

ZINC08383544 can inhibit the growth of cancer cells and effectively block the nuclear translocation of PKM2. A, Inhibitory rates of various concentrations of candidate compounds on multiple cancer cells. B, The content of PKM2 in the nucleus and cytoplasm. C, Glutaraldehyde crosslinking. The concentration of ZINC08383507 and ZINC08383544 is 250 nmol/L. D, Changes in phosphorylation of Histone H3‐T11 and STAT 3‐Y705 in HepG2 and HeLa cells. E, Real‐time quantitative PCR was used to detect the expression of GLUT1 and LDHA in HepG2 and HeLa cells treated with candidate compounds. There are three repetitions per concentration. These data were expressed as the mean ± SD. Statistical comparisons were made by using one‐way ANOVA. *P < 0.05, **P < 0.01. (In B and D, the concentrations of ZINC08383507 and ZINC08383544 were 400 and 200 nmol/L, respectively. The same volume of DMSO was added to the control group.)

3.4. ZINC08383544 can effectively block the nuclear translocation of PKM2

In view of the influence of candidate molecules on PKM2 activity, it would be lead to conformational changes and further affect the function of nuclear PKM2. Thereby reducing the amount of PKM2 entering nucleus, and the content of PKM2 in cytoplasm would rise correspondingly. The result of western blot (Figure 5B) had shown that ZINC08383544 treatment was consistent with this prediction. Moreover, the content of PKM2 changed more obviously after ZINC08383544 treatment. At the same time, the glutaraldehyde crosslinking reaction showed that the content of PKM2 tetramer after candidate compounds treatment increased significantly (Figure 5C). The phosphorylation of histone H3‐T11 phosphorylation and STAT 3‐Y705 decreased in the nucleus (Figure 5D). This result indicated that candidate compounds, especially ZINC08383544, could block the nuclear translocation of PKM2.

The results of GLUT1 and LDHA expression detection (Figure 5E) showed no matter which candidate compounds treated cells, the expression of LDHA decreased continuously with the increase of concentration. The treatment effect of ZINC08383544 was obviously better than ZINC08383507. After ZINC08383544 treatment, the expression of LDHA in HeLa cells changed more significantly. After ZINC08383507 treatment, the expression of GLUT1 decreased at first and then increased, indicating that ZINC08383507 is able to affect the PKM2 signalling pathway, and ZINC08383544 could reduce the expression of GLUT1.

3.5. Anticancer activity of ZINC08383544 is dependent on PKM2 expression

After siRNA transfected into cells 24 hours, the efficiency of transfection was observed using fluorescence microscopy (Figure 6A) and the qPCR test was used to detect the knockdown conditions of PKM2. The results (Figure 6B) showed that after siRNA transfection, the expression of PKM2 was less than 15% to the normal level. The CCK‐8 test (Figure 6C) showed that the inhibitory effect of ZINC08383544 on the growth of cancer cells was relieved and distinguished from that of ZINC08383507. Therefore, these results suggest that antitumour effect of ZINC08383544 depends on the expression of PKM2.

Figure 6.

Anticancer activity of ZINC08383544 is dependent on PKM2 expression. A, Transfection of siRNA in cells. 24 h later the transfection efficiency was observed by fluorescence microscope. The scale shown in the figure is 100 μm. B, Detection of PKM2 gene knockdown conditions by real‐time quantitative PCR. Statistical comparisons were made by using one‐way ANOVA. *P < 0.05, **P < 0.01. C, After PKM2 knockdown, the effect of candidate compouds on survival of HepG2 and HeLa

4. DISCUSSION

As evidence53, 54 suggested a key role played by the low‐activity nuclear PKM2 in tumour progression, it emerges as an attractive target in cancer therapy and the intervention of active conformation of PKM2 as a new way for cancer therapy.55 However, most of the drug discovery was mainly based on empirical screening,24, 56 known as random screening. Virtual screening is an alternative technique to solve the problem of low hit rate in drug discovery at present and offers a practical route to discovering new reagents and leads for pharmaceutical research. Combined with biological detection, the rate of molecular hit is obviously improved.

In the previous virtual screening targeting PKM2,24, 57 most of the receptor structures were chosen to concentrate on the monomer PKM2 instead of the dimer, which may be due to the various binding sites and complex structures of the PKM2 dimer. In this work, PKM2 dimer structure was selected as the receptor in virtual experimental, which is closer to the physiological state in tumour cells. The docking model used in this study was verified by AUC, which avoided subjectivity in the process of threshold selection. After dock scored, the AUC value reached 0.9169, indicating that our dock model had a good screening effect for the compound as a candidate activator. In order to evaluate the stability of the PKM2‐small molecule complexes, we analysed the RMSD change of complex backbones in the whole simulation process, and selected 2 ns equilibrium time to analyse the binding‐free energy. Integrated various aspects, two small molecular compounds were screened out from the ZINC database. Enzymatic reaction directly reflected the activation effect of screened small molecules on PKM2. For the firmness of study, several cell lines were selected for confirmation. In vitro biological tests showed that ZINC08383544 can activate PKM2 specifically, stabilise the conformation of PKM2 tetramer, reduce the amount of nuclear PKM2, and inhibit the aerobic glycolysis and cell proliferation. Furthermore, it was found that the inhibitory effect of ZINC08383544 on the proliferation of cancer cells was weakened when PKM2 knockdown by siRNA. This suggests that ZINC08383544 is more likely to depend on the expression of PKM2 to exert its effect of inhibiting the proliferation of cancer cells. Taken together, our data suggest that ZINC08383544 is a novel and potent PKM2 activator that might have therapeutic implications for cancers. It is valid to consider that ZINC08383544 can be used for further antitumour experimental research or as a lead compound targeting PKM2 anticancer drugs for drug design and development. Compared with other studies,16, 19 this work involves virtual screening and in vitro experiments. In order to confirm the role of small molecules and drug resistance more affirmatively, further animal experiments and drug metabolism experiments are needed in the near future.

COMPETING INTERESTS

The authors declare that they have no competing interests.

AUTHOR CONTRIBUTION

The contributions of each author to the contents of this manuscript are as follows. Study design: JB, XL, CW; experiments guidance: CW, XL; experiments operation: YL, MB, CY, JC; data collection and analysis: YL, MB, CY, RS, SZ; manuscript write: YL, XL.

Supporting information

 

 

 

 

Li Y, Bao M, Yang C, et al. Computer‐aided identification of a novel pyruvate kinase M2 activator compound. Cell Prolif. 2018;51:e12509 10.1111/cpr.12509