A Hybrid Approach Combining Shape-Based and Docking Methods to Identify Novel Potential P2X7 Antagonists from Natural Product Databases

Abstract

P2X7 is an ATP-activated purinergic receptor implicated in pro-inflammatory responses. It is associated with the development of several diseases, including inflammatory and neurodegenerative conditions. Although several P2X7 receptor antagonists have recently been reported in the literature, none of them is approved for clinical use. However, the structure of the known antagonists can serve as a scaffold for discovering effective compounds in clinical therapy. This study aimed to propose an improved virtual screening methodology for the identification of novel potential P2X7 receptor antagonists from natural products through the combination of shape-based and docking approaches. First, a shape-based screening was performed based on the structure of JNJ-47965567, a P2X7 antagonist, using two natural product compound databases, MEGx (~5.8 × 10³ compounds) and NATx (~32 × 10³ compounds). Then, the compounds selected by the proposed shape-based model, with Shape-Tanimoto score values ranging between 0.624 and 0.799, were filtered for drug-like properties. Finally, the compounds that met the drug-like filter criteria were docked into the P2X7 allosteric binding site, using the docking programs GOLD and DockThor. The docking poses with the best score values were submitted to careful visual inspection of the P2X7 allosteric binding site. Based on our established visual inspection criteria, four compounds from the MEGx database and four from the NATx database were finally selected as potential P2X7 receptor antagonists. The selected compounds are structurally different from known P2X7 antagonists, have drug-like properties, and are predicted to interact with key P2X7 allosteric binding pocket residues, including F88, F92, F95, F103, M105, F108, Y295, Y298, and I310. Therefore, the combination of shape-based screening and docking approaches proposed in our study has proven useful in selecting potential novel P2X7 antagonist candidates from natural-product-derived compounds databases. This approach could also be useful for selecting potential inhibitors/antagonists of other receptors and/or biological targets.

Keywords: P2X7 receptor; antagonists; natural products; shape-based model; virtual screening.

Figure 1

Schematic representation of the hierarchical sequence of selection filters applied to the MEGx and NATx databases aiming at the selection of potential P2X7 antagonists.

Figure 2

Two-dimensional representation of the structure of the JNJ-47965567, a selective P2X7 antagonist.

Figure 3

Shape-based model building. (A) Schematic representation of the AmP2X7-JNJ-47965567 complex (PDB ID: 5U1X; resolution: 3.20 Å). (B) Close-up view of JNJ-47965567 bound to the allosteric AmP2X7 binding site. (C) The shape-based model was generated using ROCS. The molecular shape surface is represented in gray. JNJ-47965567 is represented as sticks, with carbon, oxygen, nitrogen, and sulfur atoms colored in green, red, blue, and yellow, respectively. The figure was prepared using PyMOL v.2.5.0 and ROCS v.3.5.0.2.

Figure 4

Representation of the molecular interaction fields (MIFs) calculated using the GRID program for the allosteric binding site of AmP2X7 used as the reference pocket for the docking calculations in our VS protocol. White, blue, and red surfaces represent the hydrophobic, hydrogen-bond donor, and hydrogen-bond acceptor MIFs obtained using the DRY probe (energy cutoff value of −1.0 kcal.mol⁻¹), N1 probe (energy cutoff value of −7.0 kcal.mol⁻¹), and O probe (energy cutoff value of −7.0 kcal.mol⁻¹), respectively. The main binding site residues are represented as sticks, and the protein 3D structure is shown as a cartoon representation. Residues from subunits A and B are labeled in black and blue, respectively. The figure was prepared using PyMOL v.2.5.0.

Figure 5

Schematic representation of the predicted binding modes (docking solutions with the best score) of two representative compounds selected by our VS protocol as potential P2X7 antagonists. (A) Superposition between the best scoring poses from the GOLD (carbon atoms in yellow) and DockThor (carbon atoms in pink) programs for compound NP-016468 from the MEGx database. (B) Superposition between the best scoring poses from the GOLD (carbon atoms in yellow) and DockThor (carbon atoms in pink) programs for compound NAT13-340161 from the NATx database. The AmP2X7 structure is represented as a green cartoon. Representative compound structures and binding site residues that make hydrophobic and/or hydrogen bond interactions with the representative compounds are shown as sticks. Oxygen, nitrogen, and fluorine atoms are colored red, blue, and light blue, respectively. Hydrogen bonds are represented by dashed lines (in cyan for GOLD docking poses and in magenta for DockThor docking poses). Residues from different subunits are labeled in black and in blue. The figure was prepared using the PyMOL v.2.5.0 software.

Figure 6

Variation in the RMSD of the compounds complexed with the P2X7 receptor over the simulation time. (A) NP-016468. (B) NP-025357. (C) NP-025047.

Figure 6

Variation in the RMSD of the compounds complexed with the P2X7 receptor over the simulation time. (A) NP-016468. (B) NP-025357. (C) NP-025047.

Figure 7

Interaction map between protein residues and compound NP-016468 using the most representative structures obtained over 50 ns, clustered using the gromos method (cut-off 0.14 Å). Residues are identified by their names, position in the primary structure, and chain. Each of the most representative structures is associated with an average frame of the computational simulation. The maps were generated using the free version of the LigPlot+ v.2.2 program.

Figure 8

Interaction map between protein residues and compound NP-025357 using the most representative structures obtained over 50 ns, clustered using the gromos method (cut-off 0.14 Å). Residues are identified by their names, position in the primary structure, and chain. Each of the most representative structures is associated with an average frame of the computational simulation. The maps were generated using the free version of the LigPlot+ v.2.2 program.

Figure 9

Interaction map between protein residues and compound NP-025047 using the most representative structures obtained over 50 ns, clustered using the gromos method (cut-off 0.14 Å). Residues are identified by their names, position in the primary structure, and chain. Each of the most representative structures is associated with an average frame of the computational simulation. The maps were generated using the free version of the LigPlot+ v.2.2 program.