CADD-based discovery of novel oligomeric modulators of PKM2 with antitumor activity in aggressive human glioblastoma models

Abstract

Pyruvate kinase isoform M2 (PKM2) is a multifunctional enzyme capable of transitioning between monomeric, dimeric, and tetrameric states, with its oligomeric equilibrium playing a pivotal role in tumour progression and survival. The unique exon ten at the dimer-dimer interface represents an attractive target for isoform-specific modulation, offering opportunities for disrupting this equilibrium and altering tumour cell dynamics. This study identifies a novel druggable pocket at the PKM2 dimer interface through conformational analysis. This pocket was exploited in a virtual screening of a large small-molecule library, identifying two promising candidates, C599 and C998. Both compounds exhibited dose-dependent antiproliferative effects in glioblastoma cell lines and induced apoptosis, as evidenced by caspase 3/7 activation. These effects were directly linked to their inhibition of PKM2 enzymatic activity, validating the proposed mechanism of action in their rational design. ADMET studies further highlighted their strong potential as lead PKM2 inhibitors for GBM treatment. Molecular dynamics (MD) simulations and post-MD analyses, including Dynamic Cross-Correlation Maps (DCCM), Probability Density Function (PDF), and Free Energy Landscape (FEL), confirmed the stability of the protein-ligand interactions and highlighted critical residues at the dimer-dimer interface. The Steered MD simulations demonstrated the high affinity of the compounds for PKM2, as evidenced by the requirement of high rupture forces to induce an unbinding event. These results highlight the potential of the compounds as oligomeric modulators of PKM2. These findings position C599 and C998 as promising lead compounds for antitumor applications. Future studies will focus on optimising these candidates and assessing their efficacy in vivo glioblastoma models, reassuring the thoroughness of our research and the potential for further advancements.

Keywords: Docking based virtual screening; Molecular dynamics; PKM2; Pharmacological inhibitors.

Fig. 1

(A). PKM2 Tetrameric conformation. The tetramer is the association of a pair of dimers along the C terminal domain of each monomer. (B). The PKM2-specific 22 amino acids encoded by exon 10 are within the two α helices, Cα1 and Cα2 (light green).

Fig. 2

Metabolic reprogramming in glioblastoma cells driven by PKM2. Glioblastoma cells exhibit increased energy production and biosynthetic demands, leading to anaerobic glycolysis (the Warburg effect) even in the presence of oxygen [70]. This metabolic reprogramming is facilitated by several key mechanisms involving both canonical and non-canonical activities of PKM2: a) upregulation of glucose transporters (GLUTs) and PKM2 facilitates increased glucose conversion to pyruvate, leading to elevated lactate production [71,72]; b) excess lactate suppresses antitumor immune responses, promotes extracellular acidosis, and contributes to tumour growth, progression, and therapy resistance; c) glycolytic intermediates are redirected to the pentose phosphate pathway (PPP) to provide amino acids and nucleotides necessary for cancer cell proliferation; d) dimeric PKM2 migrates to the nucleus, where it activates oncogenic transcription factors that regulate proliferation, angiogenesis, tumorigenesis, and epigenetic modifications and PKM2 expression [73,74]; e) mitochondrial translocation of PKM2 protects GBM cells from oxidative stress-induced apoptosis through Bcl-2 phosphorylation [75]. Finally, PKM2 conformational dynamics is tightly regulated by numerous mechanisms including ligand-induced allosteric regulation (fructose bisphosphate (FBP), amino acids, Succinyl amino imidazole carboxyamide ribose-5-phosphate (SAICAR), several post-translational modifications (PMTs) and reactive species oxygen (ROS).

Fig. 3

Interaction radar plot graphic and Interface physicochemical parameters. The radar area exceeding the 50 % borderline at most points indicates that the interface is not a crystal packing artefact.

Fig. 4

Identification of the druggable pocket within the dimer-dimer interface. (A). Graphical representation of the Z-score calculated for the region encompassing residues 300–506 of PKM2. The shaded grey area includes residues from the Cα1 and Cα2 helices and most residues within the identified druggable pocket. (B). The image illustrates the druggable pocket at the dimer-dimer interface of PKM2, with a ribbon representation of the protein and the pocket volume depicted as a light green surface. (C). The pocket is shown with the protein surface in grey, highlighting the druggable pocket with light green spheres. The labelled residues within the pocket are encoded by exon 10 of PKM2.

Fig. 5

Chemical diversity of the compounds. The matrices show the similarity among the top 10 compounds based on the Tanimoto Coefficient of Similarity (1 indicates identity) using Morgan and MACCS 2D fingerprints.

Fig. 6

(A). Apoptosis as determined with a caspase 3/7 activity assay after treatment with C599 and C998 compounds in U87 cell line (B). Pharmacological inhibition of PKM2 enzymatic activity. PK activity in the presence of different concentrations of C599 and C998 with U87 cell line (n = 3). Data were normalised to DMSO-treated PK activity. Mean ± SD, ∗p ≤ 0.05; ∗∗p ≤ 0.01; ∗∗∗p ≤ 0.001; ∗∗∗∗p ≤ 0.0001.

Fig. 7

(A). Root-mean-square deviation (RMSD) (B). Free PKM2 (black) and PKM2 in complex with C599 (turquoise) and C998 (orange) as a function of time. (C) Root-mean-square fluctuation (RMSF) of PKM2 in complex with C599 (turquoise) and C998 (orange).

Fig. 8

(A). Dynamic cross correlation matrix (DCCM) and (B) Probability density function (PDF) analysis for free PKM2 and PKM2 in complex with compounds C599 and C998.

Fig. 9

(A) The number of Hydrogen bonds formed in the PKM2-C4 complex concerning simulation time throughout the trajectory. (B) Non-bonded interactions within the protein-ligand complex throughout the trajectory.

Fig. 10

(A). 3D Gibbs free energy landscapes (FEL) of PKM2-C599 and PKM2-C998 complexes as function of projections of the MD trajectory onto PC1 and PC2 eigenvectors, (B). 2D Diagram with close contact interactions of C599 and C998 with the binding pocket. (C). 3D Surface representation of PKM2-C599 and PKM2-C998 complexes.

Fig. 11

(A). Representative plots of the pulling force between each ligand and PKM2 throughout the SMD simulation. (B). Representative plots of the centre-of-mass (distance).