Structural basis of the multiple ligand binding mechanisms of the P2X1 receptor

Abstract

As important modulators of human purinergic signaling, P2X1 receptors form homotrimers to transport calcium, regulating multiple physiological processes, and are long regarded as promising therapeutic targets for male contraception and inflammation. However, the development of drugs that target the P2X1 receptor, such as the antagonist NF449, is greatly hindered by the unclear molecular mechanism of ligand binding modes and receptor activation. Here, we report the structures of the P2X1 receptor in complex with the endogenous agonist ATP or the competitive antagonist NF449. The P2X1 receptor displays distinct conformational features when bound to different types of compounds. Despite coupling to the agonist ATP, the receptor adopts a desensitized conformation that arrests the ions in the transmembrane (TM) domain, aligning with the nature of the high desensitization rates of the P2X1 receptor within the P2X family. Interestingly, the antagonist NF449 not only occupies the orthosteric pocket of ATP but also interacts with the dorsal fin, left flipper, and head domains, suggesting a unique binding mode to perform both orthosteric and allosteric mechanisms of NF449 inhibition. Intriguingly, a novel lipid binding site adjacent to the TM helices and lower body of P2X1, which is critical for receptor activation, is identified. Further functional assay results and structural alignments reveal the high conservation of this lipid binding site in P2X receptors, indicating important modulatory roles upon lipid binding. Taken together, these findings greatly increase our understanding of the ligand binding modes and multiple modulatory mechanisms of the P2X1 receptor and shed light on the further development of P2X1-selective antagonists.

Keywords: Structural biology; Ligand binding mode; Ion channel; Purinergic P2X1 receptor

Introduction

Purinergic signaling is an important signal transduction system in mammals that is mediated by P2Y and P2X purinergic receptors [1]. In contrast to P2Y receptors, which are G protein-coupled receptors [2], ligand-gated P2X receptors are nonselective cation channels pivotal for many physiological processes and include seven subtypes in humans (P2X1–7) [3, 4]. P2X receptors are activated by adenosine 5’-triphosphate (ATP) and form a nonselective pore to transit cations, playing diverse and important roles in mediating specific cellular events, including muscle contraction [5], inflammation [6], and vasoconstriction [7]. Therefore, dysfunction of P2X receptors is implicated in multiple diseases [3, 8, 9].

P2X receptors are formed by three individual subunits. Each subunit of P2X receptors has intracellular NH₂- and COOH-terminal domains, two transmembrane helices, and an extracellular domain where the ATP-binding site is located. P2X receptors vary in their desensitization kinetics and binding affinity for different types of ligands, indicating the complexity of their purinergic signaling and the potential of their pharmacology [10]. The crystal structures of the P2X4 receptor in apo-closed or ATP-bound states revealed the dolphin-like shapes of P2X receptors for the first time, depicting the ATP binding mode and activation mechanism of P2X4 [11, 12]. In recent years, more structural and functional findings have shown that P2X receptors have multiple binding sites for ligands and lipids, highlighting the plasticity of P2X physiological regulation [13–17]. P2X1 is a distinct member because of its rapid desensitization rate, multiple binding sites for ligands, and ability to form different types of heterotrimers [6].

Among P2X receptors, the P2X1 receptor is located on platelets, neutrophils, and the smooth muscle of the vas deferens and is the most potent P2X receptor [6, 18]. Upon ATP activation, P2X1 is involved in inflammation, bladder dysfunction, and multiple pathological processes [19, 20]. Therefore, P2X1 antagonists have long been regarded as promising drug targets for male contraception, thrombolysis, and anti-inflammatory treatment [18, 21, 22]. To date, several compounds have been identified that specifically target the P2X1 receptor, including antagonists such as NF449 [23], MRS2159 [24], aurintricarboxylic acid [25], and PSB-2001 [26], as well as positive allosteric modulators such as MRS2219 [27]. However, most compounds encounter challenges in further drug development [6]. The most potent P2X1-selective antagonist, NF449, which has subnanomolar potency, has severe off-target effects and poor bioavailability [28]. To date, although the structures of other types of P2X receptors have been solved [11, 12, 14], no structures of P2X1 have been reported, which greatly hinders the understanding of the ligand recognition and functional regulatory mechanisms of P2X1. Without more precise structural information, the identification of the molecular determinants that underpin the selectivity and pharmacological efficacy of P2X1 compounds remains insufficient.

To obtain a more thorough understanding of the molecular basis for ligand binding to the P2X1 receptor, we determined the structures of the P2X1 receptor in complex with the agonist ATP or the competitive antagonist NF449 (Fig. 1a, b). By combining functional studies and molecular dynamics (MD) simulation experiments, these new data provide a key structural basis for the interaction patterns between P2X1 and its compounds and reveal potential multiple ligand binding sites on the P2X1 receptor. This insight will facilitate the further discovery of drugs that target more diverse types of P2X1 compounds.

Fig. 1. Cryo-EM structures of ATP-P2X1 and NF449-P2X1 complexes.

a, b Cryo-EM maps of the ATP-P2X1 and NF449-P2X1 complexes colored according to their chains. c, d Structures of the ATP-P2X1 and NF449-P2X1 complexes. The structures are displayed in membrane and extracellular views. e, f Comparison of the conformations of the ATP- and NF449-P2X1 structures with previously solved P2X structures bound to ligands. The structural alignments are shown in an overall view, and the extracellular view focuses on the TM domain. e Comparison between the ATP-P2X1 and ATP-P2X7 (PDB ID: 6U9W) structures. f Comparison between the ATP-P2X1 and apo-P2X7 (PDB ID: 6U9V) structures.

Methods

Construct cloning and protein expression

The human P2X1 (wild-type) gene (399 residues; UniProt P51575) was synthesized with codon optimization by Sangon Biotech. The hP2X1 gene was subsequently inserted into a pFastBac1 vector under the regulation of the GP64 promoter, with the gene positioned upstream of a C-terminal PreScission protease cleavage site, a 10×-His tag, and a Flag tag. To improve the biochemical stability and expression level of hP2X1, we truncated 40 amino acids from the C-terminus and 25 amino acids from the N-terminus (amino acid residues 26–359).

Modified hP2X1 was expressed in Spodoptera frugiperda (sf9) insect cells using the Bac-to-Bac baculovirus expression system. The cells were grown to a density of 2 × 10⁶ cells/ml. The hP2X1 DNA was mixed with Fugene HD and transfected into sf9. P0 virus was produced after 96 h of culture at 27 °C and 300 rpm. After infection with P0 virus at a ratio of sf9 cells to P0 virus of 1:100, P1 virus was generated through culture at 27 °C and 125 rpm for 48 h. Subsequently, the precipitate was collected by centrifugation at 2000 rpm for 15 min and stored at −80 °C.

Receptor purification

The P2X1 protein was purified at 4 °C. The cell precipitate was initially rinsed with PBS and then alternately washed with low-salt media (10 mM HEPES pH 7.5, 4 °C, 10 mM MgCl₂, and 20 mM KCl) containing protein inhibitors (1 mg/ml 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 0.005 mg/ml E64, 0.005 mg/ml aprotinin, and 0.005 mg/ml leupeptin) and high-salt media (low-salt media plus 1 M NaCl) using a Dounce homogenizer, and the membrane components were separated by ultracentrifugation at 40,000 rpm for 30 min. Before dissolution, the membrane was incubated with 2 mg/ml iodoacetamide for 1 h and then dissolved in 2% (w/v) DDM, 0.4% (w/v) CHS, 100 mM Tris-HCl (pH 8.0), 150 mM NaCl, and protein inhibitors for 3 h. The precipitate was eliminated by centrifugation at 40,000 rpm for 30 min. The supernatant, supplemented with 10 mM imidazole and Talon resin, was incubated overnight at 4 °C.

The mixture was centrifuged at 800 × g for 5 min, all the supernatant was removed, and 5 column volumes of exchange buffer (25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.12% (w/v) GDN, and 5% glycerol) were added and rotated at 4 °C for 2 h. The resin was washed with 20 column volumes of washing buffer containing 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.01% (w/v) GDN, 5% glycerol, and 30 mM imidazole. The target protein was eluted with 5 column volumes of an elution buffer containing 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.01% (w/v) GDN, 5% glycerol, 250 mM imidazole, and protein inhibitors. The protein solution was subsequently concentrated to 500 μl via a 100 kDa molecular weight cutoff concentrator and further purified on a Superdex 200 Increase 10/300 GL equilibrated with 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 0.01% (w/v) GDN. The peak fractions of the protein samples were concentrated to ~15 mg/ml, flash-frozen in liquid nitrogen, and stored at –80 °C until use.

For the NF449 and ATP binding structures, the purification method was conducted as described above. Only 100 μM NF449 and 1 mM ATP were added to the utilized buffers before membrane dissolution and during the purification procedure.

Cryo-EM data collection

Three microliters of the P2X1 complex protein sample was applied to a cryo-EM grid (ANTcryo R1.2/1.3, Au 300 mesh). To improve the particle distribution, we conducted 45 s of glow discharge and then vitrified it into liquid ethane using a Mark IV Vitrobot, with a blot force of 1 and a blot time of 3 at 4 °C and 100% humidity. This sample preparation protocol ensures the standardization and consistency of each sample, which benefits cryo-EM sample optimization and data collection. Data were collected on a 300 kV Titan Krios G3 electron microscope equipped with a Gatan K3 summit direct detection camera and a GIF-Quantum energy filter at a magnification of 81,000. With a bin2 pixel size of 1.071 Å, the superresolution counting mode of SerialEM was used to collect movie stacks. The defocus values of the cryo-EM movies varied from −0.8 to −1.5 μm, and the dose rate was 1.75 electrons per Ų per frame, with a 3 s exposure time in a total of 40 frames.

Data processing and P2X1 model construction

In total, 8669 movies for the ATP-P2X1 complex and 11,736 movies for the NF449-P2X1 complex were collected. To estimate and adjust the particle movements caused by electron beam damage, we subjected the movies to beam-induced motion correction using MotionCor2. The CTFFIND4 software in CryoSPARC was used to estimate the contrast transfer function (CTF) parameters of each image. To create the initial particle-picking templates without introducing any preset information, the blob picker was used to pick particles from 500 random micrographs. After several rounds of two-dimensional (2D) classification, 145,850 particles for the ATP-P2X1 complex or 129,640 particles for the NF449-P2X1 complex were selected to generate three-dimensional (3D) templates. Guided by the templates generated above, the template picker module in CryoSPARC was used to select targeted particles more precisely, and 5,949,390 or 5,783,231 particles were picked from the ATP-P2X1 or NF449-P2X1 image datasets, respectively, via the template picker program in CryoSPARC. Following two rounds of 2D classification for particle polishing, 1,272,255 particles of the ATP-P2X1 complex and 1,801,896 particles of the NF449-P2X1 complex were subjected to ab initio reconstruction, generating 5 classes of initial models for both datasets without any preset information. The particles in the best-looking classes of each dataset were selected to perform further rounds of heterogeneous refinement and ab initio reconstruction. After discarding the classes of blurry particles, 346,074 particles and 391,273 particles were subjected to nonuniform refinements of the ATP-P2X1 and NF449-P2X1 complexes, yielding initial density maps of 4.2 Å and 3.2 Å, respectively. To further optimize the maps, the datasets of the two complexes were subjected to Bayesian polishing using RELION-4. The polished particles were subsequently converted into CryoSPARC to perform 3D classification without alignments by setting the number of classes to ten. The particles of the seven best-looking classes were chosen to perform the final round of nonuniform refinement and local refinement, resulting in two final maps with global resolutions of 3.6 Å and 3.1 Å, respectively. The gold-standard Fourier shell correlation was used to evaluate the resolutions of the two maps with the 0.143 criterion. Local resolution estimation was performed via ResMap.

The models of ATP-P2X1 and NF449-P2X1 complexes were built using the predicted P2X1 receptor model generated by AlphaFold2. The initial models of the two complexes were docked into the cryo-EM density maps using ChimeraX, and either ATP or NF449 was introduced into the P2X1 models according to the maps. After model fitting among several types of lipid compounds, the cholesterol modules were confidentially fitted into the adjacent regions of the TM helices of the three subunits in both P2X1 maps. The models were manually adjusted in COOT and refined in Phenix using real_space refinement. The final models were validated using MolProbity. The structural analysis and figure preparations were completed using Chimera and PyMOL (https://pymol.org/2/).

Whole-cell patch-clamp electrophysiological recording

hP2X1 and its mutants were subcloned and inserted into the pcDNA3.1 vector and tagged with EGFP to facilitate subsequent electrophysiological recordings. Whole-cell patch-clamp recordings were conducted on HEK-293 cells at room temperature (25 ± 2 °C) 24–48 h posttransfection using Lipofectamine 3000 (Thermo Fisher, USA). The recordings were performed with a Multiclamp 700B amplifier (Axon Instruments, USA) and a Digidata 1440 A digitizer (Axon Instruments, USA). Current data were sampled at 10 kHz and filtered at 2 kHz, with analysis carried out using pCLAMP 10 software (Molecular Devices, USA). Patch electrodes pulled from borosilicate glass capillaries (World Precision Instruments) had resistances ranging from 1.5 to 4 MΩ when filled with an intracellular solution composed of 145 mM NaCl, 10 mM EGTA, 10 mM glucose, and 10 mM HEPES (adjusted to pH 7.3 with NaOH). The extracellular solution contained 147 mM NaCl, 2 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 10 mM HEPES, and 13 mM glucose (adjusted to pH 7.3 with NaOH). Extracellular solutions containing ATP or varying concentrations of compounds were applied using an RSC-200 Rapid Solution Changer (BioLogic, France), which is equipped with a 9-tube head for continuous perfusion. The delivery system included pipettes with a tip diameter of 500 µm positioned ~300 µm from the cell, allowing for solution switching within 10 ms. All electrophysiological recordings were analyzed using Clampfit 10.6 (Molecular Devices, USA).

Molecular dynamics simulations

The coordinates of the P2X1 complexes were extracted from the ATP-P2X1 complex structure, with or without the ATP module. Using the CHARMM-GUI membrane builder, the protonation states of all titratable residues and the disulfide bonds in P2X1 receptors were evaluated at pH 7.4 with or without the ATP module, and the P2X1 and ATP-P2X1 complex structures were embedded in a membrane-mimic bilayer composed of the lipids 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine and 1,2-dioleoyl-sn-glycero-3-phosphocholine at a 50:50 ratio. The simulation systems were solvated in a water box containing 150 mM NaCl and 150 mM CaCl₂, with a minimum water height of 22.5 Å at both the top and bottom of the systems, resulting in a total of 144,296 atoms for the ATP-P2X1 system and 144,392 atoms for the P2X1 system. The CHARMM36m force field was used to model the receptor and membrane atoms.

The simulations were then performed using the CPU implementation of the parallelized program in Gromacs. Each of the two constructed systems was energy minimized and heated from 0 to 310 K using Langevin dynamics with a constant box volume. The system was subsequently equilibrated in the constant pressure and temperature ensemble (NPT). Further two-step equilibration was carried out at 310 K for 1 ns with harmonic restraints applied to the protein only, starting with a force constant of 5 kcal/mol/Ų on the protein backbone and ending with 0.1 kcal/mol/Ų on the protein Cα atoms. After that, a 1 ns simulation without any restraints was performed at 310 K and 1 bar in the NPT ensemble. After the equilibration of the systems, classical molecular dynamics (cMD) simulations were performed on the equilibrated systems using the GPU version of Gromacs. To achieve this, a 10-ns short cMD simulation was used to collect potential statistics for further MD parameter values, and finally, a total of three independent MD production trajectories were run with a 600 ns simulation time. The coordinates were saved every 1000 steps for data analysis. After the simulations were completed, the trajectories of each system were analyzed with VMD software. The trajectories from 200 to 600 ns were averaged and analyzed as the final output structures of the MD simulations.

Results

Structure determination of ATP/NF449-bound P2X1 complexes

To facilitate receptor expression and protein yield, the optimized P2X1 construct yielded high-quality protein samples after careful screening of receptor N/C-terminus lengths (Supplementary Fig. 3a–d). Specifically, the 25 residues in the N-terminal region of human P2X1 were truncated, and the 40 residues in the C-terminus were replaced with polyhistidine and flag tags for protein expression detection and purification (Supplementary Fig. 3e, f). Compared with that of the wild-type (WT) receptor, the functional assay showed that the signaling of the P2X1 construct used for structure solving was severely impaired, which may reflect the important roles of the P2X1 N/C-terminus in receptor signaling transduction [29] (Supplementary Fig. 3g). The optimized P2X1 construct was cloned and inserted into the pFastBac1 vector and expressed with the Bac-to-Bac expression system in Sf9 insect cells. NF449 or ATP was subsequently introduced into the P2X1 receptor purification process to obtain stable ligand-bound P2X1 complexes (Supplementary Fig. 3e, f).

Images of the complex samples were collected via cryogenic electron microscopy (cryo-EM) and then subjected to single-particle analysis. The results of further data processing yielded two maps of the P2X1 receptor in complex with NF449 or ATP at resolutions of 3.1 Å and 3.6 Å, respectively (Fig. 1a, b, Supplementary Fig. 1). The well-defined density of maps enabled unambiguous model building for most of the residues in the P2X1 receptor, whether in complex with NF449 or ATP. However, no continuous density was observed for the region of the transmembrane domain of NF449-bound P2X1, indicating highly dynamic features of these regions (Supplementary Fig. 2b). Overall, the clear density of maps allowed the good modeling of NF449- or ATP-bound P2X1 structures, which is sufficient for further analysis and structural findings.

Conformational changes in the P2X1 receptor in different states

Despite differences in protein modifications (truncation, mutation, etc.) and structure-solving technologies, as with other previously determined P2X receptor structures [12, 13, 15], each subunit of the P2X1 receptor structures we solved has a dolphin-like shape, which consists of a head, upper and lower bodies, right and left flippers, dorsal fin, and transmembrane (TM) helices (Fig. 1a). The three subunits of the P2X1 receptor are closely intertwined from the extracellular domain to the TM domain, forming a threefold axis of symmetry centered on the receptor. This arrangement aligns perpendicularly to the membrane plane, reflecting the essential structural organization for receptor physiological processes (Fig. 1c, d). In the extracellular region, the receptor adopts a similar conformation in complex with either NF449 or ATP, with a Cα root mean square deviation (Cα r.m.s.d.) of 1.443 Å in the extracellular domain (S58–D327). In both structures, the extra density is shown around the sidechain of the asparagine amino acid at positions 153, 184, 242, and 300, which are potentially N-glycosylated according to a previous study [30] (Supplementary Fig. 2c). Although the other three residues are located aloof from the orthosteric ligand pocket of the P2X1 receptor, the N184 residue sits beside the phosphate group of NF449 in the structure, indicating the potential impact of N-glycosylation on P2X1 ligand binding (Supplementary Fig. 2d).

In contrast to the similarity of the receptor extracellular domain, the conformation of the TMD varies between the two structures, with an r.m.s.d. of 3.789 Å in the structural deviation of the two TM helices. TM helices I and II shift away at ~9 and 6 Å (measured at the end of the helix) in NF449-bound P2X1 relative to those in the ATP-bound structure. As an endogenous agonist and a competitive antagonist, ATP and NF449 control the distinct modulatory processes of P2X1 activation [26] by exerting completely opposite effects. Accommodated by the physiological roles of the two compounds, the ATP-bound P2X1 structure forms larger ion channel pores than the NF449-bound P2X1 structure does (Fig. 1c, d). Compared with the P2X7 receptor in the open and closed states previously determined by cryo-EM, NF449-bound P2X1 forms a closed ion pore similar to P2X7 in the closed state (PDB ID: 6U9V). In terms of structure, three helix II of the P2X1 subunits intersect proximally, forming a narrow pore via residues I339 and G340 (Fig. 1f). Like the P2X3 or P2X7 structures in the closed state, helix I of P2X1 stands beside helix II to further strengthen the shrinkage of the ion pores (Fig. 1f, Supplementary Fig. 3i). Comparing the sizes of the ion channels with those of P2X7 in the closed state, NF449-bound P2X1 shows similar structural features in the TM domain, forming a very narrow tunnel with a minimum radius of 0.5 Å (Fig. 2b).

Fig. 2. Channel characteristics of P2X1 in different states.

a, b Pore-lining surface analysis and plots of the pore radius along the channel axis of P2X1 receptors in different states. The data were calculated by the program HOLE2 (https://www.holeprogram.org/). A radius <1.15 Å is colored red, a radius between 1.15 and 2.30 Å is colored green, and a radius >2.30 Å is colored blue. The channel volumes of P2X1 receptors are quantified by HOLE2 calculations, and the corresponding receptor domains of the radius curves are illustrated by dashed lines. a, Comparison of the pore-lining surfaces of ATP-P2X1 and ATP-P2X7. b Comparison of the pore-lining surfaces of NF449-P2X1 and apo-P2X7. c, d Cryo-EM maps and structures of ATP-P2X1 at checkpoints 1 and 2, shown in the extracellular views, focused on D97 or S337 residues. e Structural alignment of ATP-P2X1 and apo-P2X1 (using the structure of ATP-P2X1 with the ATP module removed) after 600 ns of MD simulation. The structures are produced by averaging the results of 400–600 ns of simulation. f Illustration of the different conformational changes of P2X1 bound to ATP or NF449.

However, unlike the common features of open-gated pores exhibited in P2X7, in the structure of ATP-bound P2X1, the TM helixes of each subunit move closer to the central symmetry axis of the structure. Moreover, helix I constrains the outward movement of helix II by surrounding it more uprightly, forming an angle of ~5° perpendicular to the cellular membrane (Fig. 1e). As a result, the ion tunnel of P2X1 bound to ATP has a similar shape to that of P2X7 in the open state, although it has smaller channel sizes at checkpoints 1 and 2 in the extracellular and TM domains. The TM region of ATP-bound P2X1 forms a smaller pore than P2X7 in the open state, with a measured radius of 0.8 Å at residues G343 and V344, and such a small ion pore cannot enable the transportation of cations (Fig. 2a). Interestingly, two ion-like densities were found to be surrounded by residues D97 or S337, right at the adjacent areas of the checkpoints mentioned above. The sidechains of these residues all point to ion-like density, forming bottlenecks at checkpoints 1 or 2 (Fig. 2c, d). The P2X1 receptor has long been regarded as having the highest ATP affinity and the fastest desensitization rate among all P2X receptors [31]; thus, our structure is consistent with the conformation of fast desensitization after activation. In the desensitization state presented by our structure, residues D97 and S337 form two ionic interaction networks, arresting calcium to block the transportation process.

To further understand the gating cycle of the P2X1 receptor, we performed three independent MD simulations of the ATP-bound P2X1 structure, with or without the ATP molecule. After 600 ns of simulation, the data revealed that P2X1 was stable in the desensitized state, with or without ATP coupling (Supplementary Figs. 3h, j). By averaging the structure from 400 ns to 600 ns, the conformations of P2X1 in the desensitized state are barely influenced by ATP coupling (Fig. 2e). In the areas of the two checkpoints, the formation of ion networks still persisted during the MD simulation process, indicating the stability of this desensitizing conformation. Taken together, P2X1 undergoes conformational changes stimulated by ATP, leading to the expansion of ion channel pores with the outward movement of the TMD. P2X1 then desensitizes rapidly to block cation transport, as reflected in the structural features discussed above, including the constraint of TM helices, the formation of ion checkpoints, and the shrinkage of ion channels (Fig. 2a, f). In addition, the binding of NF449, an antagonist, further triggers the closing of ionic pores to inhibit P2X1 signaling, allowing the receptor to remain in a closed state (Fig. 2b, f).

Orthosteric pocket of P2X1

As an ATP-gated cation channel, the P2X1 receptor couples ATP to transport ions such as calcium. To date, several ATP-bound P2X receptor structures have been solved, shedding light on the ligand binding mechanism of this family. ATP couples to receptors in a “U”-shaped structure, interacting with a cluster of basic and polar residues that are highly conserved in the P2X family [12]. NF449, a competitive antagonist of P2X1, is known for its high potency and selectivity among the P2X receptor family, and its inhibitory potency was demonstrated via a whole-cell patch clamp electrophysiology assay. However, owing to the large differences in molecular weight and skeleton, the molecular mechanism of NF449 selectivity and inhibitory potency is ambiguous without structural information, which greatly hinders the development and optimization of P2X1-selective antagonists.

Strikingly, in our NF449-bound P2X1 structure, NF449 not only occupies the binding pocket of ATP but also extensively interacts with other regions of the ECD (Fig. 3a). Upon binding to ATP, the residues K68, K70, K190, R292, and K309 of the P2X1 subunits form salt bridge interactions with the phosphate group of ATP, generating a conserved binding site similar to other P2X receptors reported [12, 32]. Furthermore, the head of the γ-phosphate group of ATP couples to the magnesium ion, strengthens the interaction with D170, and forms a hydrogen bond with Q142 via a water connection, which contributes to the highly polar environment for ATP binding (Fig. 3b). In NF449-bound P2X1, one benzene ring of NF449 occupies a binding site similar to that of ATP in P2X1. Specifically, two phosphate heads of the NF449 benzene ring are surrounded by the residues K68, K70, K190, R292, and K309, forming a similar interaction pattern as the phosphate group of ATP (Fig. 3c). To verify our findings, single mutations of those key residues to alanine resulted in a notable decrease in both ATP activation and the NF449 inhibitory effect, indicating that these polar residues are determinants of the competitive inhibitory effect of NF449 on ATP (Fig. 3e). Consistent with the finding that D170A interacts with ATP alone, the mutation D170A has no obvious inhibitory effect on NF449, further confirming the differences in NF449 and ATP binding in the orthosteric pockets (Fig. 3b, c, e).

Fig. 3. Binding modes of ATP and NF449.

a Interaction interface between P2X1 and ATP or NF449. The interaction regions are highlighted by red dashed circles. The interface is formed by two P2X1 subunits, colored yellow and purple blue. b–d ATP or NF449 interactions at the interfaces. The key residues involved in the interactions are shown as sticks and colored according to their chains. The backbone of ATP is colored forest green, and the backbone of NF449 is colored magenta. e NF449-mediated inhibition of ATP-evoked currents in hP2X1 and its mutants measured via a whole-cell patch-clamp electrophysiological recording assay. The inhibition ratios of the constructs are represented by bars. Supplementary Table 1 provides detailed independent experiment numbers (n) and statistical evaluation. *P < 0.05, **P < 0.001, ***P < 0.0001 by one-way analysis of variance followed by Dunnett’s post hoc test compared with the response of the wild type. f Conformational changes induced by ATP or NF449 in the P2X1 receptor; the movements of different regions are indicated with red arrows. g Model of P2X1 conformational changes coupled to different ligands.

Despite the similar coupling features of the two ligands in the region of the phosphate group, the other parts exhibit different or even opposite effects on NF449 and ATP. The adenine base of ATP is located at the bottom of the orthosteric pocket and forms hydrogen bonds with the sidechains of T186 and K140 and the main chain oxygen of K70, which are conserved among P2X receptors [12]. Additionally, the residues F188, V229, M214, and L205 form a hydrophobic core to accommodate the interactions with the ATP adenine group, indicating the utmost importance for ATP activation by dragging the dorsal fin regions of P2X1 to trigger the downstream activation process (Fig. 3b). In NF449-bound P2X1, however, the other three benzene rings of NF449 behave like a “scissor” and stabilize the conformations of the left flipper, head, and dorsal fin domains. One benzene ring of NF449 penetrates into the cleft formed by the head and left body, interacting with R139, K136, F185, and T186 and preventing the downward shift of the head domain. By forming hydrogen bonds with residue K215 and extensive hydrophobic interactions with residues L218, V229, and F188, the other two benzene rings of NF449 prevent further inward movement of the dorsal fin domain, which results in the inhibitory effect of P2X1 receptor activation (Fig. 3d). The mutants K136A, R139A, K215A, and F188A in P2X1 strongly impaired the inhibitory potency of NF449 by 5%–40% (Fig. 3e). These data demonstrate the distinct binding modes and inhibition mechanisms of NF449 and identify the key factors involved in P2X1 agonism and antagonism.

As mentioned above, the ATP- or NF449-bound structures are similar in the extracellular domain, whereas the lower body domain of the NF449-bound structure shifts downward by 3.6 Å compared with that of ATP-bound P2X1 (measured at D264). This deviation is caused by the steric hindrance of NF449, which prevents P2X1 activation (Fig. 3f). Owing to its strong interactions with the dorsal fin and head domains, NF449 functions as a versatile molecule, not only competitively occupying the orthosteric pocket of ATP but also constraining the downward movement of the dorsal fin and head domains. This, in turn, pushes the lower body downward, resulting in the narrowing of the ion channel pore (Fig. 3g). Our findings reveal the molecular binding mechanisms of ATP and NF449, elucidate the structural basis of the inhibitory effect of NF449, and may shed light on the further development of P2X1 antagonists.

Novel lipid binding site in P2X1

The physiological functions of P2X receptors are regulated by different types of lipids [17, 33]. Some previous functional and structural studies have demonstrated the critical binding motifs of lipids at the N/C-terminus of P2X receptors, which strongly influence the activation of P2X receptors. In our structures, strikingly, a novel lipid binding site was observed, which plays an important role in P2X1 activation.

In both the density maps of ATP- and NF449-bound P2X1, clear densities were found beside the upper side of the TM regions of each P2X1 subunit, attached to the loops of the lower body domain and the beginning region of TM1. According to the shape of the lipid density, cholesterol was confidently modeled after the different types of membrane lipids were screened (Fig. 4a). Further molecular docking experiments confirmed the justification of cholesterol modeling, with a coupling energy of -6.8 kJ. Specifically, residues F49, Y55, and F326 form hydrophobic interactions with the steroid group of cholesterol, and residue W265 of the lower body region forms a hydrogen bond with the alcohol head of cholesterol, resulting in a tight connection between the lower body and TM regions (Fig. 4b). A single alanine mutation was introduced into those key residues to reveal the impact of cholesterol on P2X1 activation. The mutants W265A and F326A strongly reduced P2X1 signaling at a 20%–40% ratio, and F49A and Y55A almost completely abolished P2X1 activation (Fig. 4e). This finding suggests that cholesterol functions as a mediator to transit the movement of the lower body to the TM domain, triggering TM1 outward movement and TM2 rotation, which are critical for P2X1 activation. In particular, the mutation of residues F326, W265, and F49 to alanine also notably decreased the inhibitory effect of NF449 in a functional assay at a ratio of 40%–60% (Fig. 4e). These findings may indicate that cholesterol is needed in the transition from the open or desensitization state to the closed state, transiting the downward movement of the lower body triggered by NF449 and pushing the TM1 helix to shift inwardly, leading TM2 helixes of three subunits to form a narrow pore to stabilize the receptor in the closed state. Therefore, the binding of cholesterol may be the key modulating factor facilitating conformational changes in the gate cycle of the P2X1 receptor, which highlights the importance of this novel binding site for the further development of allosteric drugs.

Fig. 4. Cholesterol binding mode in P2X1 structures.

a The binding sites of cholesterol in the P2X1 receptor bound to ATP. The backbone of cholesterol is colored silver. b Details of the interactions between cholesterol and the P2X1 receptor. c, d Structural comparisons of different P2X receptors. The key interacting residues are shown as sticks and colored according to their chains. e Ratios of NF449 inhibition or ATP-evoked current peak density for the P2X1 receptor and its mutants. The data are plotted from at least three independent experiments. *P < 0.05, **P < 0.001, ***P < 0.0001 by one-way analysis of variance followed by Dunnett’s post hoctest compared with the response of the wild type. f Model of P2X1 conformational changes with the assistance of cholesterol at lipid binding sites.

To further explore this binding site among the P2X receptor family, sequence alignment was performed, which revealed high conservation of the lipid binding site in all P2X receptor family members. For example, residues Y55 and F326 of P2X1 are conserved across all P2X receptors, and the regions near residues W265 and F49 also show high conservation of amino acid properties (Supplementary Fig. 4). In structural alignments with P2X3, P2X4, and P2X7, identical or similar residues form the same lipid binding site as observed in our P2X1 structure (Fig. 4c, d). Moreover, in recently solved structures of P2X receptors, densities were observed in areas similar to the lipid binding site, although no lipids were modeled (Supplementary Fig. 5a, b, c; EMD-20703; EMD-36668).

These findings strongly indicate that P2X receptors may also share these conserved lipid binding sites to mediate receptor conformational changes in several states. A previous study also revealed the important roles of cholesterol in P2X receptor activation and sensitization [17, 34, 35]. Consistent with our findings, P2X7 activation rates were increased with the depletion of cholesterol, revealing the potential negative modulator role of cholesterol [34]. However, all previous studies on the lipid regulation of P2X receptors focused on the N/C-terminal regions of these receptors, in which cholesterol might modulate pore formation and trigger downstream receptor activation. Our findings revealed that cholesterol is a key mediator of changes in P2X receptor conformation by interacting with the ECD and TMD regions of the P2X1 receptor directly, transiting the movements from upward to downward. Similar to the results of a previous study, cholesterol not only acts as a remote effector but also participates in domain shifting and conformational changes, whether induced by an agonist or an antagonist (Fig. 4F). Taken together, these data expand the potential of this novel binding site to all P2X receptors and may pave a new way to understand the allosteric modulating mechanism of P2X receptors.

Agonism and antagonism of P2X receptors

Among the published P2X structures bound to different compounds, AF-219 in P2X3, BX430 and BAY-1797 in P2X4, and A804598 in P2X7 reveal several allosteric binding sites of P2X receptors, which are located from the head and upper body to the lower body domains [14, 36, 37]. These negative allosteric modulators play important roles in different sites of P2X receptor activation chains, stabilizing the conformations in inactive states or inhibiting further activation processes. However, those allosteric binding sites are all located in the extracellular domains of P2X receptors, where they either inhibit structural changes in the upper body domain (BX430 and BAY-1797 in P2X4) [37], block the relative motions between the left flipper and lower body-dorsal fin domains (AF-219) [36], or stabilize the turret formed by the lower body and left flipper domains (A804598 in P2X7) [14]. The cholesterol in the lipid binding site of our structures facilitates the transition of activating motions from the extracellular to TM domains by connecting the loops of the lower body and TM domains and coordinating the upper motion of the lower body with the outward movements of TM1 and TM2.

By combining all the data, a model is proposed to illustrate the agonism and antagonism of P2X receptors. In apo-state P2X receptors, each domain of the receptor rests in a relatively flexible state, and the lower body and left flipper domains maintain closed conformations without the influence of ATP binding (Fig. 5a). As a result, the TM helices of the subunits interact with each other, forming narrow pores insufficient for ionic transportation. While the agonist ATP binds to P2X receptors, the phosphate and adenine base groups of the ATP agonist interact with the head and dorsal fin domains, pushing out the left flipper from the ATP binding site and triggering the upper movement of the lower body. Transmitted through the cholesterol we identified, such movements cause outward shifts and rotations of the TM domains, resulting in the opening of the ion channel [12] (Fig. 5b). Despite the consistency of agonism in the P2X family, the extent of ion channel opening varies across subtypes of P2X receptors, with different ATP affinities and desensitization rates. Like P2X1, a receptor that is rapidly desensitized on a subsecond timescale [31], the TM region is constrained in the closed state, potentially mediated by the cytoplasmic region that is not observed in our structures [38]. In this state, the bottleneck of the ion channel in the TM domain is formed by residue S337 in three subunits, trapping cations in the helix bundles and blocking signaling (Fig. 5c).

Fig. 5. Model of the P2X receptor activation cycle.

a–d Cartoon model of P2X in the activation cycle, combining the modulatory effects of ATP, NF449, and cholesterol mentioned above.

In the antagonism of P2X receptors, all types of inhibitors work by blocking the ATP activation process or stabilizing the closing conformation of P2X receptors. The orthosteric antagonists occupy the ATP binding pocket to prevent the initiation of ATP activation, resulting in the closed states of P2X receptors (Fig. 5d). However, the other allosteric modulators constrain the conformational changes in the activation process mentioned above and impair the potency of ATP activation without total inhibition. Together, the findings of agonism and antagonism may provide insights into the modulatory mechanism of P2X receptors, deepen the understanding of connections at different modulating sites, and facilitate structure-based drug design and the study of drug combination therapy.

Discussion

As an important drug target, the P2X1 receptor has attracted much attention because of its modulation mechanism and antagonist development [6]. In this work, we report the cryo-EM structures of the human P2X1 receptor in complex with ATP or NF449, revealing the interaction modes of agonists and antagonists. The reported ligand-bound P2X structures have provided some insight into ATP activation, including the conserved motifs and clusters in the P2X orthosteric pocket. Combined with the results of the functional assay, we revealed the molecular basis of the competitive binding mechanism between the two ligands. Moreover, while the phosphate groups of the two ligands have similar binding effects in the orthosteric pocket, the other parts exhibit different structural features in terms of binding modes and activation modulation. The other three benzene rings of NF449 form extensive interactions with the head, left flipper, and lower body domains, preventing further movements of these domains and inhibiting the P2X1 activation process. In contrast, several antagonists of other P2X receptors only occupy orthosteric sites, of which TNP-ATP, PPADS, and PPNDS interact with the conserved basic clusters of the P2X7 ATP binding site [15, 16], revealing the molecular mechanism of competitive inhibition effects. Therefore, the unique structural features of NF449, an antagonist with a large molecular weight, may contribute greatly to its high affinity and selectivity, and our findings may provide the structural basis for the further optimization and development of P2X1-selective antagonists. During the preparation and submission of our manuscript, several P2X1 cryo-EM structures, including apo-, ATP-, or NF449-bound states, have been published. These structures shared very similar structural features with our findings, such as the same P2X1 desensitization state during ATP coupling, the dual-binding mode of NF449, and the similar cholesterol binding site. In addition to these similarities, we discovered several unique structural features that further clarify the conformational changes in ion channels, such as two checkpoints in the ECD and TMD. Taken together, these findings strongly support our conclusions and help us gain a more thorough understanding of P2X1 activation and ligand modulation.

Cholesterol is an allosteric modulator of many membrane protein receptors. In our structures, a novel lipid binding site is identified at the connecting region of the lower body and TM domains. In conjunction with functional assays, cholesterol has been demonstrated to be the main contributor to the transit of conformational movements during P2X1 activation, potentially revealing the allosteric modulating effects of this lipid binding site in P2X receptors. In recent years, several compounds that occupy orthosteric or allosteric pockets have been identified to promote drug discovery in the P2X family. Because of the high sequence homology among P2X receptors, most antagonists share similar binding modes with different P2X receptors, resulting in poor selectivity and side effects for those antagonists in clinical trials. In contrast with traditional drug pockets, this new lipid-binding site of the P2X receptor may provide novel insight into the development of drugs for the whole P2X receptor family. The new drug candidate designed on the basis of this lipid binding site may not occupy the traditional site of the P2X receptor to interrupt normal ATP binding but may constrain the receptor to a specific conformational state by coadministration with other specific antagonists, resulting in increased selectivity and potency in disease treatment. Our current findings are insufficient to validate the feasibility of this drug development strategy, which requires more structural information and relevant functional assays. In the future, we expect more detailed structural and functional findings of P2X receptors related to lipid function to reveal the important regulatory roles and therapeutic potential of those key mediators.

In addition to the importance of these binding sites in P2X homotrimer activation, our findings may provide some insight into the mechanism by which P2X heterotrimers are modulated. The majority of P2X receptors are capable of forming heterotrimers, which exhibit different physiological properties [39]. The P2X1 receptor can form heterotrimers with P2X2, P2X3, P2X5, and P2X6, resulting in distinct features in ATP affinity, desensitization, and pH modulation [4, 40]. However, the molecular basis of P2X1-related heterotrimer formation and modulation is still ambiguous. In our study, the lipid binding site was confirmed to be conserved among P2X receptors, as were the highly conserved ATP binding pockets, which may establish similar agonism to that of the P2X homotrimers discussed above. The ATP molecule couples to similar binding pockets of P2X1 heterotrimers, triggering the activating motions of the P2X1 subunit, which then passes through the TM region through cholesterol, resulting in the expansion of the ion pore. Despite the similarity of their agonism, the different selective compounds of P2X receptors may vary in affinity or modulating effects owing to the different binding pockets formed by heterotrimers compared with those formed by homotrimers. In addition, the different types of P2X cytoplasmic regions may have different modulatory effects on P2X1 heterotrimer desensitization rates and internalization intentions. Overall, the different modulatory mechanisms of P2X compounds and the molecular details of our study highlight the diversity and complexity of P2X heterotrimers, which need to be further studied.

Collectively, our work reveals the structural basis for the coupling of ATP and NF449, the agonist and antagonist, to P2X1 receptors. By combining MD simulations and functional assays, the distinct lipid binding sites strengthen our understanding of P2X1 agonism and antagonism. These findings shed light on the basis of P2X1 signaling modulation and improve the development of P2X1 drug discovery.

Author contributions

YTQ established and optimized the protein purification procedures, prepared protein samples for cryo-EM data collection, and helped with manuscript preparation. PPW performed the functional assays and helped with the data analysis. XL and YTC helped with plasmid cloning and protein preparation. LP expressed the proteins. LKZ helped with the protein purification and data analysis. ZBG planned and analyzed the experiments and helped with manuscript preparation. KC collected the cryo-EM data, performed the cryo-EM data processing and analysis, as well as model building and structure refinement, wrote the first version of the manuscript, and helped with the data analysis. QZ initiated the project, planned and analyzed the experiments, supervised the research, and finalized the manuscript with input from all coauthors.

Data availability

The coordinates and cryo-EM density maps for the structures of the ATP-P2X1 and NF449-P2X1 complexes have been deposited in the PDB database with the identification codes 9LX5 and 9LXC, respectively, and in the Electron Microscopy Data Bank under accession codes EMD-63466 and EMD-63473, respectively.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41401-025-01512-y.