A polycyclic scaffold identified by structure-based drug design effectively inhibits the human P2X7 receptor

Abstract

The P2X7 receptor is an ATP-gated ion channel that activates inflammatory pathways involved in diseases such as cancer, atherosclerosis, and neurodegeneration. However, despite the potential benefits of blocking overactive signaling, no P2X7 receptor antagonists have been approved for clinical use. Understanding species-specific pharmacological effects of existing antagonists has been challenging, in part due to the dearth of molecular information on receptor orthologs. Here, to identify distinct molecular features in the human receptor, we determine high-resolution cryo-EM structures of the full-length wild-type human P2X7 receptor in apo closed and ATP-bound open state conformations and draw comparisons with structures of other orthologs. We also report a cryo-EM structure of the human receptor in complex with an adamantane-based inhibitor, which we leverage, in conjunction with functional data and molecular dynamics simulations, to design a potent and selective antagonist with a unique polycyclic scaffold. Functional and structural analysis reveal how this optimized ligand, termed UB-MBX-46, interacts with the classical allosteric pocket of the human P2X7 receptor with subnanomolar potency and high selectivity, revealing its significant therapeutic potential.

Fig. 1. Structures of human, mouse, and rat P2X7Rs in the apo closed state conformation reveal distinct allosteric ligand-binding sites.

A Ribbon representation of the apo closed state structures of human (shades of blue and gray), mouse (shades of purple and gray), and rat (shades of green and gray, PDB code: 8TR5) P2X7Rs, aligned with ChimeraX. The classical allosteric ligand-binding site is boxed in red. Two CHS molecules per protomer (inner and outer; both tan), located at the interface between TM1 and TM2 on the extracellular side of the membrane, are found only in the hP2X7R. GDP (tan) and Zn²⁺ ions (slate gray) are labeled and shown within the cytoplasmic ballast (Supplementary Fig. 2 A). B Magnified and 120° rotated view of A, highlighting cryo-EM density for two molecules of CHS bound to one protomer of the hP2X7R. C Same view as B, highlighting the hydrophobic and hydrogen bonding interactions between the inner CHS molecule and the hP2X7R. D 50° rotated view from C highlighting the hydrophobic and hydrogen bonding interactions between the outer CHS molecule and the hP2X7R. E-G Magnified view of the classical allosteric ligand-binding site from A, highlighting the differences between human (E, light blue and gray), mouse (F, light purple and gray), and rat (G, light green and gray) P2X7R orthologs. Residues 95, 108, and 312 (red labels) are the key residue differences in the classical allosteric pockets between the three orthologs. The larger V312 in hP2X7Rs (A312 in mP2X7Rs and rP2X7Rs) forces the neighboring residue Y295 (orange labels) to adopt an alternative rotameric conformation that condenses the classical allosteric pocket in the human ortholog only (Supplementary Fig. 7).

Fig. 2. ATP binds to the orthosteric pocket of the hP2X7R.

A Dose response curves from TEVC experiments measuring the activation of full-length wild-type human (blue), mouse (purple), and rat (green) P2X7Rs by ATP (EC₅₀ = 89 ± 8.3 µM, 70 ± 17 µM, and 34 ± 8.4 μM, respectively). Data points and error bars represent the mean ± SD of normalized current, respectively, across triplicate experiments. B Representative BLI sensorgram for a dilution series of ATP (shades of blue) binding to biotinylated hP2X7R immobilized on streptavidin (SA) biosensors. Kinetic data were globally fit using a Langmuir 1:1 model to determine the equilibrium dissociation constant (K_D) of ATP to the hP2X7R as K_D = 650 ± 120 nM, representing the mean ± SD across triplicate experiments. For kinetic analysis, a 90-s association time and a 60-s dissociation time were used for analysis. C Ribbon representation of the hP2X7R in the ATP-bound open state conformation at 3.0 Å colored by protomer (blue, light-blue, and gray), highlighting the orthosteric ATP-binding site. D, E Magnified view of the orthosteric ATP-binding site from A, highlighting the cryo-EM density for ATP (D) and residue interactions that coordinate ATP (E). ATP is coordinated by seven residues conserved across all P2XR subtypes (K64, K66, T189, K193, N292, R294, and K311) and four P2X7R subtype-specific residues (Q143, L191, I214, and Y288).

Fig. 3. Cryo-EM analysis and molecular dynamics of UB-ALT-P30 bound to the hP2X7R show that larger caged alkyls fit the classical allosteric pocket in the human ortholog.

A 2D chemical structure of UB-ALT-P30. B Concentration-dependent inhibition of agonist-induced calcium influx (IC₅₀ curves) for UB-ALT-P30 on human, rat, and mouse P2X7Rs. Receptors were recombinantly expressed in 1321N1 astrocytoma cells and activated by an EC₈₀ of the agonist BzATP. C Calcium influx assays were used to measure the potency of UB-ALT-P30 at hP2X1Rs, hP2X2Rs, hP2X3Rs, and hP2X4Rs stably expressed in 1321N1 astrocytoma cells. Receptors were activated by an EC₈₀ of ATP at the respective subtype (hP2X1R, 100 nM; hP2X2R, 1000 nM; hP2X3R, 100 nM; hP2X4R, 300 nM). *An IC₅₀ value of 707 ± 185 nM was estimated for UB-ALT-P30 at the hP2X3R. B, C Data represent mean ± SD of at least three independent experiments performed in duplicates. For each trace, the data were normalized to the calcium signal induced by the respective EC₈₀ concentration of agonist (when no antagonist was added). D Ribbon representation of the classical allosteric ligand-binding site in hP2X7R, located at the interface of two protomers (gray and light blue) with one molecule of UB-ALT-P30 (tan) shown with corresponding cryo-EM density (blue mesh) at 2.8 Å. E Residues in the classical allosteric ligand-binding site of the hP2X7R that interact with UB-ALT-P30. The polycyclic group of UB-ALT-P30 forms hydrophobic interactions with the receptor. F Computational values that describe the binding properties of UB-ALT-P30 within the classical allosteric pocket of the hP2X7R. The hydrophobic volume reported is the hydrophobic surface of the adamantyl group calculated via Maestro; Schrödinger. Unoccupied space surrounding the adamantyl was calculated via POVME3.0. LogP values calculated via Maestro; Schrödinger. G Same view as E showing the accessible volume of the classical allosteric pocket in the hP2X7R as calculated by Fpocket using the UB-ALT-P30-bound inhibited state structure of the hP2X7R with the ligand removed. Several voids flanking UB-ALT-P30, as well as vacant space above and below the ligand, could be targeted areas for further ligand development.

Fig. 4. UB-MBX-46 is a potent and selective allosteric antagonist for the hP2X7R.

A 2D chemical structure of UB-MBX-46. B Concentration-dependent inhibition of agonist-induced calcium influx (IC₅₀ curves) for UB-MBX-46 on human, rat, and mouse P2X7Rs. Receptors were recombinantly expressed in 1321N1 astrocytoma cells and activated by an EC₈₀ of the agonist BzATP. C Calcium influx assays were used to measure the potency of UB-MBX-46 at hP2X1Rs, hP2X2Rs, hP2X3Rs, and hP2X4Rs stably expressed in 1321N1 astrocytoma cells. Receptors were activated by an EC₈₀ of ATP at the respective subtype (hP2X1R, 100 nM; hP2X2R, 1000 nM; hP2X3R, 100 nM; hP2X4R, 300 nM). B, C Data represent mean ± SD of at least three independent experiments performed in duplicates. For each trace, the data were normalized to the calcium signal induced by the respective EC₈₀ concentration of agonist (when no antagonist was added). D Ribbon representation of the classical allosteric ligand-binding site in hP2X7R, located at the interface of two protomers (gray and light blue) with one molecule of UB-MBX-46 (tan) shown with its corresponding cryo-EM density (blue mesh) at 2.5 Å. E Residues in the classical allosteric ligand-binding site of the hP2X7R that interact with UB-MBX-46. The polycyclic group of UB-MBX-46 forms hydrophobic interactions with the receptor and the backbone carbonyl from D92, as well as the side chain hydroxyl of Y298 form hydrogen bonding interactions with the hydrazide linker. F Computational values that describe the binding properties of UB-MBX-46 within the classical allosteric pocket of the hP2X7R. The hydrophobic volume reported is the hydrophobic surface of the polycyclic group calculated via Maestro, Schrödinger. Unoccupied space surrounding the polycyclic core was calculated via POVME3.0. LogP values calculated via Maestro, Schrödinger. G Same view as E showing the accessible volume of the classical allosteric pocket in the hP2X7R calculated by Fpocket using the UB-MBX-46-bound inhibited state structure of the hP2X7R with the ligand removed. UB-MBX-46 tightly fits the classical allosteric pocket of the hP2X7R, although unoccupied voids remain above and below the ligand. H Determination of binding kinetics by TEVC recordings. Normalized current responses to 5s-pulses of 300 µM ATP over time during continuous superfusion with the indicated concentrations of UB-MBX-46. All data are represented as mean ± SD with n = 3, 4, 4, and 4 for 3 nM, 10 nM, 30 nM, and 100 nM respectively. I (left) Time course of antagonist dissociation. Normalized responses to 300 µM ATP after a 90–100% block by the antagonists are shown, highlighting virtually irreversible binding of UB-MBX-46 to the hP2X7R. (right) Experimentally determined on-rates (k_obs) were plotted against antagonist concentrations F to obtain an estimate for the off-rate constant k_off (y-intercept) according to the formula k_obs = k_on* F + k_off. All data are represented as mean ± SD with n = 3 for UB-MBX-46 and n = 5 for UB-ALT-P30 for the dissociation (left) and n = 4 for UB-MBX-46 and n = 7 for UB-ALT-P30 for the on-rates (right). Since there was minimal dissociation of UB-MBX-46 after 10 min, its graphically determined off-rate is clearly overestimated, and better values for K_off were obtained based on the measured or extrapolated IC₅₀ values (Supplementary Fig. 10).

Fig. 5. A potent and selective scaffold for the allosteric modulation of the hP2X7R.

A–D Schematic representation of the unoccupied classical allosteric pocket in human, mouse, and rat P2X7R compared to the classical allosteric pocket for the hP2X7R occupied by UB-MBX-46. Surface rendering for one P2X7 protomer is shown to highlight the size and identity of ortholog-specific residues comprising the classical allosteric pocket. The classical allosteric pockets of the hP2X7R (A), the mP2X7R (B), and the rP2X7R (C) in the apo closed state conformation. The larger residues in the allosteric site of the hP2X7R appear to condense the pocket compared to the mP2X7R or the rP2X7R. Specifically, V312, F95, and the sterically rotated Y295 in the hP2X7R occupy more volume deep within the pocket. D The classical allosteric pocket of the hP2X7R is occupied by the potent and selective antagonist UB-MBX-46. The ligand is well tailored to the size and shape of the classical allosteric pocket in the human ortholog.