A shared mechanism for TNP-ATP recognition by members of the P2X receptor family

P2X receptors (P2X1–7) are non-selective cation channels involved in many physiological activities such as synaptic transmission, immunological modulation, and cardiovascular function. These receptors share a conserved mechanism to sense extracellular ATP. TNP-ATP is an ATP derivative acting as a nonselective competitive P2X antagonist. Understanding how it occupies the orthosteric site in the absence of agonism may help reveal the key allostery during P2X gating. However, TNP-ATP/P2X complexes (TNP-ATP/human P2X3 (hP2X3) and TNP-ATP/chicken P2X7 (ckP2X7)) with distinct conformations and different mechanisms of action have been proposed. Whether these represent species and subtype variations or experimental differences remains unclear. Here, we show that a common mechanism of TNP-ATP recognition exists for the P2X family members by combining enhanced conformation sampling, engineered disulfide bond analysis, and covalent occupancy. In this model, the polar triphosphate moiety of TNP-ATP interacts with the orthosteric site, while its TNP-moiety is deeply embedded in the head and dorsal fin (DF) interface, creating a restrictive allostery in these two domains that results in a partly enlarged yet ion-impermeable pore. Similar results were obtained from multiple P2X subtypes of different species, including ckP2X7, hP2X3, rat P2X2 (rP2X2), and human P2X1 (hP2X1). Thus, TNP-ATP uses a common mechanism for P2X recognition and modulation by restricting the movements of the head and DF domains which are essential for P2X activation. This knowledge is applicable to the development of new P2X inhibitors.

ATP-induced activation has been studied intensively by means of structural and functional biology for P2X receptors of various species and subtypes [28][29][30][31][32][33][34].These studies reveal very conserved ATP recognition and activation mechanisms for the P2X receptor family members (ATP act mainly in the cavity formed by two adjacent subunits, and interacted with the head, dorsal fin (DF) and left flipper (LF) domains), e.g., the close similarity between Gulf coast tick P2X receptor (Amblyomma maculatum P2X, AmP2X) [31] and human P2X3 (hP2X3) [28], not to mention the highly conserved ATP recognition and gating mechanism among mammalian panda P2X7 (pdP2X7) [30], rat P2X7 (rP2X7) [32] and hP2X3 [28].However, some of the allosteric mechanisms are not so obvious when comparing the open and closed structures, such as the upper body domain and the upper vestibule, which are rigid architectures [30,35]; completely different binding conformations were obtained for different P2X receptors bound by the same inhibitor, e.g., TNP-ATP [28,36].
Currently, two TNP-ATP-bound structures of P2X receptors have been determined, namely hP2X3 (PDB ID: 5SVQ) and chicken P2X7 (ckP2X7, PDB ID: 5XW6) [28,36].In contrast to the highly conserved mechanism of P2X receptor activation by ATP, these two structures show completely different overall conformations and TNP-ATP recognition mechanisms.The TNP-ATP-bound ckP2X7 exhibits an incompletely activated pore conformation (the transmembrane (TM) region is extended outward, but the channel pore is still impermeable to ions).Moreover, its extracellular domain shows an extended shape somewhat similar to the open state of the P2X receptor [28,29,31,36], which resembles the TNP-ATP-bound zebra fish P2X4 (zfP2X4) by NMR analysis [37].In addition, TNP-ATP in this structure exhibits a "U-shaped" occupancy pose resembling ATP binding [28][29][30][31], with its 2,4,6 trinitrophenyl (TNP) moiety interacting with amino acid residues in the head and DF domains of the receptor.This binding mode is somewhat consistent with a previous result on the P2X1 receptor by voltage clamp fluorometry analysis [34].In contrast, the structure of the hP2X3 receptor bound by TNP-ATP resembles the apo or closed conformation of P2X [28,38].In this structure, the TNP-ATP molecule shows a "Y-shaped" occupancy and its TNP moiety is rotated ~180 • toward the lower body (LB) domain of P2X3 and is encapsulated by the LF region [28].These two different TNP-ATP binding modes might arise from subtype or species differences, which would be inconsistent with the highly conserved ATP-binding mode across different species and subtypes of P2X receptors [28][29][30][31][32][33][34].Alternatively, the final binding conformation of TNP-ATP to the P2X family members may be very similar, and either or both of determined structures could represent intermediate states.Moreover, it cannot be ruled out that the differences might simply arise from the different experimental approaches used.
Here, we evaluated the interaction of TNP-ATP with P2X receptors based on the above two structures by combining enhanced conformational sampling, covalent occupation, and engineered disulfide bond analysis.Our results reveal a general mechanism for the interaction of TNP-ATP with the P2X receptors, in which the polar triphosphate group of TNP-ATP interacts with the orthosteric site, while its TNP moiety is deeply embedded in the interface between the head and DF domains, which restricts the conformational changes associated with channel opening in these two domains.In addition, the TM region, which is indirectly coupled to the DF domain via the body domain, has an outward motion but no ion permeation in the final state.Our model generally agrees with the binding mode revealed by the ckP2X7/TNP-ATP structure complex, but has an important refinement, in which V130 and H131 of ckP2X7, instead of T202 and T112 as revealed by structural biology, are key sites for TNP-ATP recognition.We confirmed this binding mode in hP2X3, rP2X2 and hP2X1, demonstrating that it represents a common recognition mechanism of the P2X receptors to this class of non-selective competitive inhibitors, which should be instructive to future designs of P2X receptor drugs.

Chemicals and mutagenesis
Unless otherwise stated, all compounds were purchased from Sigma-Aldrich (USA).TNP-ATP triethylammonium salt was purchased from Tocris bioscience (Catalog No.: 2464), with a purity of more than 98%.hP2X3 plasmid was purchased from Open Biosystems; ckP2X7 cDNA was a generous gift of Dr. Osamu Nureki and was subcloned into pcDNA3.1 vector; pcDNA3-rP2X2, pcDNA3-rP2X3 plasmids were generous gifts of Drs.Alan North and Linhua Jiang.All mutants were created using the KOD-Plus-mutagenesis kit (Toyobo, SMK-101) and confirmed by DNA sequencing [39].

Cell culture and electrophysiology
As we described previously [40][41][42], human embryonic kidney (HEK 293) cells were purchased from Shanghai Institutes for Biological Sciences and cultured in Dulbecco's Modified Eagle Medium supplemented with 10% fetal bovine serum (FBS), 1% penicillin-streptomycin and 1% GlutaMAX™ at 37 • C, 5% CO 2 and 95% air in a humidified environment.Plasmids were transfected into cells using transfection agents containing two solutions (Solution A: 250 mM CaCl 2 in pure water; Solution B (in mM): 1.5 Na 2 HPO 4 , 140 NaCl and 50 HEPES, pH adjusted to 6.96).After mixing 2.5-3 μg of the plasmid and solution A, the mixture was added dropwise to an equal volume of solution B, followed by stirring with a pipette tip.The mixture was left to stabilize at room temperature for 3-5 min and then added to the cell culture dish.After 6 h of transfection, the medium in the culture dish was replaced with fresh medium.

Engineered disulfide bond cross-linking and gel analysis
HEK293 cells were transfected with rP2X3-WT or mutant plasmids, washed three times with phosphate buffered saline (PBS; pH 7.4), and then incubated with 2 mL sulfo-NHS-LC-biotin (Pierce, Germany).The cells were then placed in a refrigerator at 4 • C for 30 min with shaking and agitation every 10 min to label membrane proteins on the cell surface.Subsequently, glycine was added to terminate the reaction.Cells were washed three times with PBS, and then RIPA lysis buffer (200 μL) was added.Cell lysates were then collected from the bottom of the culture dish with a cell spatula and then centrifuged at 12,000 rpm for 30 min at 4 • C.Then, 20% (v/v) of the supernatant was added to SDS loading buffer to determine total protein content, with (for reducing gel analysis) and without (for non-reducing gel analysis) β-mercaptoethanol (β-Me), followed by a 5-minute metal bath.Anti-EGFP (1:3000; Sigma, United States) antibodies were incubated overnight at 4 • C.After washing, the blot was incubated with secondary Goat anti-mouse IgG (H+L) HRP (Sungene Biotech, China) for 2 h at room temperature.Finally, the blot was visualized by exposure with automated chemiluminescence-fluorescence image analysis systems (Tanon 5200, Multi) for 1-3 min in the ECL solution (Thermo).Analysis of protein expression was repeated by at least three independent experiments.

Conventional molecular dynamics (CMD) simulations and metadynamics (MetaD) simulations
The energy-minimized structures of hP2X3/TNP-ATP (PDB ID:5SVQ), ckP2X7/TNP-ATP (PDB ID:5XW6) were used as the initial structures for CMD simulations.A large 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC, 300 K) bilayer, available in System Builder of DESMOND [42,44], was built to generate a suitable membrane system based on the OPM database [45].The hP2X3/TNP-ATP/POPC and ckP2X7/TNP-ATP/POPC systems were dissolved in simple point charge (SPC) water molecules.The DESMOND default relaxation protocol was applied to each system prior to the simulation run. 1) 100 ps simulations in the NVT (constant number of particles (N), volume (V), and temperature (T)) ensemble with Brownian kinetics using a temperature of 10 K with solute heavy atoms constrained; 2) 12 ps simulations in the NVT ensemble using a Berendsen thermostat with a temperature of 10 K and small-time steps with solute heavy atoms constrained; 3) 12 ps simulations in the NPT (constant number of particles (N), pressure (P), and temperature (T)) ensemble using a Berendsen thermostat and barostat for 12 ps simulations at 10 K and 1 atm, with solute heavy atoms constrained; 4) 12 ps simulations in the NPT ensemble using a Berendsen thermostat and barostat at 300 K and 1 atm with solute heavy atoms constrained; 5) 24 ps simulations in the NPT ensemble using a Berendsen thermostat and barostat at 300 K and 1 atm without constraint.After equilibration, MD simulations were performed for ~0.5-1.0 µs.Long-range electrostatic interactions were calculated using the smooth particle grid Ewald method.Trajectory recording intervals were set to 100-200 ps and for other settings default parameters of DESMOND were used in the CMD simulation runs.All simulations used the all-atom OPLS_2005 force field [46,47], which is used for proteins, ions, lipids and SPC waters.The Simulation Interaction Diagram (SID) module in DESMOND was used for exploring the interaction analysis between TNP-ATP and hP2X3/ckP2X7.MetaD simulations were performed by DESMOND under NPT and periodic boundary conditions using the default parameters at constant temperature (330 K) and pressure (1 bar).The parameters for height, width of the Gaussian, and the interval were set to 0.12 kcal/mol, 0.05 Å and 0.09 ps, respectively.The DESMOND default relaxation protocol was applied to each system prior to the MetaD simulation run (same steps as for CMD simulations, see above).All MetaD simulations were lasted for 120 ns until they showed free diffusion along the defined CV.The sum of the Gaussians and the free energy surface (FES) were generated by METADYNAMICS ANALYSIS tools of DESMOMD.The bias V (s, t) is typically constructed in the form of periodically added repulsive Gaussians, where s is the chosen CV which could be multidimensional.Therefore, the free energy surface can be constructed in the space spanned by those CVs.The bias potential V (s, t) at time t can be written as: where ω is the Gaussian height controlled by the deposition stride, Si is one of d CVs, and σ i is the Gaussian width.This method pushes the system to escape the local minima to find the nearest saddle point on the FES.When the transient happens, the bias provides the free-energy estimate as: where C is an arbitrary additive constant, and F(S) is free energy.Since the absolute free energy is normally not important, this constant can be readily eliminated for calculating the free-energy difference.All simulations were performed on DELL T7920 with NVIDIA TESLA K40C or CAOWEI 4028GR with NVIDIA TESLA K80.The simulation system was prepared, trajectory analyzed and visualized on a DELL T7500 graphic workstation with 12 CPUs.

Data analysis
All electrophysiological recordings were analyzed using Clampfit 10.6 (Molecular Devices).Pooled data were expressed as mean and standard error of the mean (s.e.m.).Comparisons between multiple independent groups were performed using one-way ANOVA followed by Dunnett's tests.Comparisons between two groups were made using paired (before and after engineered disulfide bond cross-linking) or unpaired Student's t tests, as appropriate.The dose-response curves were fitted using the Hill 1 equation: where I is the normalized current at a given concentration of ATP, I max is the maximum normalized current, EC 50 is the concentration of ATP yielding one half of maximal currents, and n is the Hill coefficient.The inhibitory concentration-effect curves were fitted using the Hill 1 equation: I/I max = 1/[1+(IC 50 /[TNP-ATP]) n ], where I is the normalized current at a given antagonist concentration, I max is the maximum normalized current induced by the agonist, IC 50 is the concentration of the antagonist exhibiting the half-maximum effect, [TNP-ATP] is the concentration of TNP-ATP, and n is the Hill 1 coefficient.

Comparison of two modes of TNP-ATP recognition by P2X receptors and their stability during conventional molecular dynamics (CMD) simulations
The P2X receptor structure has a grail-shaped trimeric structure, and each subunit consists of a large hydrophilic extracellular domain, two TM helices, and intracellular termini [28][29][30][31][32]37,38].ATP and its derivative TNP-ATP (Fig. 1A) act mainly in the cavity formed by two adjacent subunits.Despite the differences in subtypes and species, the ATP molecule adopts almost the same binding mode in the zfP2X4, hP2X3 and pdP2X7 receptors, with an overall "U-shaped" insertion (Fig. 1B, C).For TNP-ATP, two different mechanisms of action have been proposed, with differences in the recognition mode and the overall conformational changes induced by TNP-ATP (Fig. 1C-G).
For recognition, in the ckP2X7/TNP-ATP structure complex (PDB ID: 5XW6), TNP-ATP has a "U-shaped" orientation, like the ATP molecule shown in Fig. 1B, C. The inhibitory TNP moiety interacts with the head, LB and DF domains of the ckP2X7 receptor, forming strong hydrogen bonds (H-bonds) with the amino acids ck7 K66, ck7 T112 and ck7 T202 (the upper corner mark ck7 indicates the sequence number of ckP2X7, hereinafter), and it affects the opening of the channel by preventing the head and DF domains from approaching each other (Fig. 1C, D).
In contrast, in the structure of TNP-ATP-bound hP2X3 (PDB ID: 5SVQ), TNP-ATP exhibits a "Y-shaped" orientation (Fig. 1C, E), with the TNP group rotated approximately 180 • toward the LB and LF domains, preventing the downward movement of the LF domain, which is required for channel opening.Interestingly, the TNP group here does not form any strong H-bonds with the hP2X3 receptor, but only interacts with h3 F174 in a hydrophobic manner (Fig. 1E), despite the fact that several polar residues in the LF domain are sufficiently close to TNP-ATP.
The changes in the overall structure of the P2X receptor induced by TNP-ATP binding are also inconsistent.The overall structure of the TNP-ATP-bound hP2X3 resembles the apo/closed state of hP2X3, which is particularly evident in the TM region (Fig. 1F), i.e., no change at all; whereas that of TNP-ATP-bound ckP2X7 exhibits a conformation that is somewhat suggestive of an incomplete open state as it exhibits features of both the apo/closed state and ATP-bound open state (Fig. 1G).
These results suggest that the two conformations determined by structural biology have some degree of stability, which could represent either stable or semi-stable intermediate states.At least during CMD simulations at a time scale of μs, these two conformations did not undergo significant conformational changes or state transitions.However, since some of the structures were obtained after purification of the proteins by soaking with the ligand (commonly used to obtain structures of complexes [28][29][30]), such as hP2X3 [28], it cannot be ruled out that the obtained conformation only existed under conditions of the structural determination.

The pocket formed by the LF and LB domains is not essential for TNP-ATP recognition of hP2X3
To validate the recognition mechanism of TNP-ATP in the hP2X3/ TNP-ATP complex, we introduced a pair of cysteine residues ( h3 K201C/V274C) between the LF and DF domains of the P2X3 receptor, which can form inter-subunit disulfide bonds and is theoretically able to block the insertion of the TNP group into the gap between the LF and DF domains (Fig. 3A), thereby reducing the inhibitory effect of TNP-ATP.Since the expression efficiency and ATP-evoked current of the h3 K201C/V274C double mutant was markedly decreased, we chose rat P2X3 (rP2X3) as a replacement (Fig. 3B, C), whose orthosteric site and sequences of the LF and DF domain are almost identical to hP2X3 (Fig. S1).
Western blots obtained from non-reducing SDS-PAGE gels showed that r3 K201C/V274C exhibited more trimers with a molecular weight of ~268 kDa as compared to r3 wild type ( r3 WT) and single cysteine Fig. 2. Conventional molecular dynamics simulations (CMD) of hP2X3/TNP-ATP (PDB ID: 5SVQ) and ckP2X7/TNP-ATP (PDB ID: 5XW6) complexes.(A) 0.4-μs CMD simulations of hP2X3/TNP-ATP complex showing interactions between key residues and TNP-ATP.The interactions are classified into four types: hydrogen bonds (green), hydrophobic (light purple), ionic (pink) and water bridges (blue).The histogram of the stacking is normalized over the course of the trajectory.(B) CMD simulations showing the total interactions of hP2X3 and TNP-ATP.The darker orange color indicates a higher number of residues interacting with the ligand, as some amino acids have multiple specific contracts with the ligand.(C) 0.4-μs CMD simulations of the ckP2X7/TNP-ATP complex showing amino acid and TNP-ATP interactions.(D) CMD simulations showing the full interaction of ckP2X7 with TNP-ATP.(E, F) Conformational evolution of each rotatable bond in the ligand during the simulation trajectory of hP2X3/TNP-ATP (E) and ckP2X7/TNP-ATP (F).Two-dimensional schematics of TNP-ATP are shown as color-coded rotatable bonds.The radial plots represent the conformation of the torsion bodies.The center of the radial plot represents the beginning of the simulation, with the temporal evolution plotted in the radial direction outwardly.The histogram summarizes the data of the corresponding radial plot, which represents the probability density of the torsion.mutants, which disassembled into monomers of ~90 kDa after breaking the disulfide bond with β-mercaptoethanol (β-Me) (Fig. 3D).This indicated that disulfide bonds could be formed between subunits.Compared with r3 WT, the affinity of r3 K201C/V274C to ATP was reduced by about 15-fold (EC 50 for ATP: r3 WT, 0.832 ± 0.140 μM; r3 K201C/V274C, 12.9 ± 2.7 μM ; Hill 1 function fit; Fig. 3E), which is consistent with the conclusion that S275 of the LF domain is involved in the recognition of ATP [28,48].Because TNP-ATP is a competitive inhibitor, to avoid false positives/false negatives associated with agonist nonsaturation, we chose a saturating concentration of 100 μM ATP to examine the inhibition of r3 K201C/V274C by TNP-ATP, and compared it with r3 WT stimulated with a saturating concentration of 10 μM ATP.Even with a 10-fold increase in ATP concentration for the drug to compete, the apparent inhibitory efficiency of TNP-ATP in r3 K201C/V274C was only mildly decreased compared to r3 WT (IC 50 : r3 WT, 0.117 ± 0.092 nM; r3 K201C/V274C, 0.362 ± 0.071 nM; Hill 1 function fit; Fig. 3F, G).
In addition, we evaluated the contribution of h3 F714, the only residue showing interaction with the TNP group in the hP2X3/TNP-ATP structure.Since it has been reported that the F 174 A mutant of hP2X3 does not alter the inhibition by TNP-ATP [49], we designed a tryptophan mutant ( h3 F174W) to create a bulkier side chain.However, the inhibition by TNP-ATP remained unchanged (inhibition ratio = 0.810 ± 0.044 for h3 F174W, p > 0.05 vs h3 WT 0.764 ± 0.033, n = 8-12, unpaired t-test; Fig. 3H, I), indicating that h3 F174 indeed has no significant contribution to the inhibition by TNP-ATP.These results suggest that the site formed by the LF and LB domains might not contribute to TNP-ATP recognition by hP2X3.
To further verify the role of the head and DF domains of hP2X3 in TNP-ATP recognition, we introduced a pair of disulfide bonds, h3 L127C/ T202C (Fig. 4F), in these two regions.If these two regions are involved in TNP-ATP recognition, the disulfide bond would strongly weaken the inhibitory effect of TNP-ATP.Indeed, the h3 L127C/T202C mutant displayed a significantly reduced inhibition by TNP-ATP (inhibition ratio = 0.324 ± 0.069 for H 2 O 2 + (disulfide formed) vs. H 2 O 2 -0.806 ± 0.032, n = 5, p < 0.001, unpaired t-test; Fig. 4G, H ). Thus, the head and DF domains of hP2X3, rather than the LB and LF regions shown by the structure of the hP2X3/TNP-ATP complex, actually play an important role in TNP-ATP recognition.

Ehanced conformational sampling by metadynamics (MetaD) simulations demonstrates low relative free energy with the TNP moiety of TNP-ATP bound at the head and DF domains of hP2X3
The above results implied a different TNP-ATP recognition mode than that revealed by the structure of the hP2X3/TNP-ATP complex.
Since the CMD simulations on µs time scale (Fig. 2) could not detect a large conformational change or allow transition to another state, we employed MetaD simulations for enhanced conformational sampling [50].This method allowed not only the transition from the TNP moiety being accommodated by the LF and LB regions to that by the head and DF domains, but also an evaluation of the relative binding free energies between them.Since the distance will alter significantly if TNP-ATP flips, we chose the distance between the O atom on the h3 K113 backbone of the hP2X3 and the N atom on the 4' position nitroxide of TNP as the collective variables 1 (CV1)；the dihedral angle consisting of four consecutive atoms on the backbone of the triphosphate group as CV2 (bright yellow, Fig. 5A) for the accelerated sampling of MetaD simulations.Combinatorial alterations to CV1 and CV2 can affect the relative free energies of the two TNP-group conformations further into the ATP-binding pocket and at the head-DF domain interface.Based on the 120 ns MeatD simulations and the three-dimensional (3D) reconstruction of the free energy surface, the initial conformation of TNP-ATP (POSE 0) corresponded to a ΔG of + 14.79 kcal/mol, and the two conformations with the lowest free energy, POSE I and POSE II corresponded to ΔG of + 1.59 kcal/mol and 0 kcal/mol, respectively (Fig. 5D).In both conformations with the lowest energy, the TNP moiety was flipped and inserted between the head and DF domains, which is similar to the pattern shown in the ckP2X7/TNP-ATP complex, Fig. 5B,  C).We superimposed the lowest energy POSE II conformation with the ckP2X7/TNP-ATP complex (PDB ID: 5XW6) and found that the binding modes of TNP-ATP in both were approximately the same, differing only slightly in the depth of the insertion of the TNP moiety to the gap between the head and DF domains (Fig. 5E).
More importantly, the amino acid residues that interact with the TNP moiety of TNP-ATP in these two conformations with the lowest relative free energy, i.e., h3 K113 and h3 L127 in the head domain and h3 R204 in the DF domain, are the same as those we verified above to be important for the TNP-ATP action by analyzing mutants (Fig. 4B, E).This implies that the structurally determined hP2X3/TNP-ATP interaction mechanism (Fig. 1E) likely represents an intermediate state during the interaction of TNP-ATP with hP2X3, rather than a final natural state.

Enhanced conformational sampling by MetaD simulations reveals low relative free energy of the ckP2X7/TNP-ATP complex with the TNP moiety interacting with the head and DF domains
For further inverse validation, we performed MetaD-enhanced sampling on ckP2X7, choosing the distance between the N atom on the nitro group at the 4' position of TNP and the C atom of the ck7 K181 backbone in the LF domain as the distance variable CV1, and the four consecutive atoms of the triphosphate group backbone as the dihedral angle variable CV2 (bright yellow, Fig. 5G).And the resulted showed that the ΔG of the initial conformation (POSE 0) was + 5.71 kcal/mol, and that of the two conformations with the free energy minimums were 0 kcal/mol for POSE I and + 0.03 kcal/mol for POSE II (Fig. 5 F).In both conformations, the TNP group of TNP-ATP remains position between the head and DF domains (Fig. 5 H, I).However, the ΔG of POSE III, which resembled the recognition mode of the hP2X3/TNP-ATP complex (i.e., TNP interacts with the LB and LF domains, PDB ID: 5SVQ), is + 14.94 kcal/ mol, much higher than that of POSE 0, POSE I and POSE II (Fig. 5F, J).By superimposing the initial (dark blue) and final (gray) poses after 120 ns of MetaD simulations, we were able to observe that the TNPgroup had moved from the ATP-binding pocket to the interface between the head and DF domains (Fig. 5K).Thus, the enhanced conformational sampling of hP2X3/TNP-ATP and ckP2X7/TNP-ATP by MetaD simulations suggests the recognition of TNP-ATP by both P2X receptors to be more in line with the mode revealed by the ckP2X7/TNP-ATP structure, i.e., the interaction of the TNP moiety with the head and DF domains of the P2X receptor.

Residues ck7 V130 and ck7 H131 in the head domain and mild outward expansion of the TM region contribute to TNP-ATP recognition by ckP2X7
To validate the information obtained from ckP2X7 interaction with TNP-ATP in MetaD-enhanced conformational sampling, as well as the structure of the ckP2X7/TNP-ATP complex (PDB ID：5XW6), we analyzed mutants with amino acid substitutions in the head and DF domains of ckP2X7.A ck7 V130W/H131W double tryptophan mutant was made to restrict the rotation of the TNP moiety, which increased the efficiency of TNP-ATP inhibition of ATP-evoked current by more than 20-fold compared to ck7 WT (IC50 : ck7 WT, 4.28 ± 0.47 μM; ck7V130W/ H131W, 0.199 ± 0.036 μM ; Hill 1 function fit; Fig. 6A -C).Its apparent affinity to ATP was comparable to ck7 WT (EC 50 : ck7 WT, 3.96 ± 0.78 μM; ck7 V130W/H131W, 5.36 ± 0.49 μM; Hill 1 function fit; Fig. 6D).This indicates that ck7 V130 and ck7 H131 in the head domain are critical for the recognition of TNP-ATP.The ckP2X7/TNP-ATP and hP2X3/TNP-ATP models differ in the TM region in addition to the TNP-ATP recognition site (Fig. 1F, G): in hP2X3, the TNP-ATP-bound state is similar to the closed state, while in ckP2X7, the TM region of the TNP-ATP-bound state is outwardly expanded, implying that alterations in this region may partially contribute to the binding free energy (including the entropic and enthalpy changes from additional interactions, which are, of course, partially offset by the increased protein strain energy).We reasoned that if the TM region is restricted, it would also affect the inhibitory effect of TNP-ATP on ckP2X7.To test this, we introduced an additional pair of disulfide bond mutants, ck7 L45C/L320C (Fig. 6E), in the TM region of ckP2X7.After forming the disulfide bond with the treatment of H 2 O 2 , the inhibition ratio of TNP-ATP was 0.153 ± 0.035, significantly lower than after the disulfide bond was broken with a subsequent treatment of DTT, which had an inhibition ratio of 0.384 ± 0.027 (p < 0.01, paired ttest, n = 3; Fig. 6F).As a control, these treatments did not affect the TNP-ATP inhibition effect on ck7 WT (0.234 ± 0.029 vs. 0.244 ± 0.067, DTT treatment vs. H 2 O 2 treatment, respectively, p > 0.05, paired t-test, n = 3, Fig. 6F), indicating that the TM region of ckP2X7/TNP-ATP indeed contributes significantly to TNP-ATP inhibition.

Further refinement and confirmation of the interactions revealed by the structure of the ckP2X7/TNP-ATP complex
Next, we explored the contributions of amino acids that form Hbonds with the TNP group as revealed by the ckP2X7/TNP-ATP complex.In the structure, the TNP moiety of TNP-ATP forms H-bonds with ck7 T112 (head domain) and ck7 T202 (DF domain) (Fig. 1D), which may play a key role in its recognition.To verify this, we first introduced tryptophan at residues ck7 T112 and ck7 A113 in the head domain to explore the effects of increasing the side chain size on TNP-ATP inhibition.
Statistics on the evolution of TNP-ATP at individual rotatable bonds (RB) interacting with ckP2X7 during CMD simulations revealed that the two nitro groups of TNP-ATP facing the head and DF domains approached a 360 • rotation (Fig. 2F), implying that this position is not restricted by ckP2X7.The ck7 V130W/H131W double tryptophan mutant may limit this fluctuation, resulting in a 20-fold increase in TNP-ATP inhibition efficiency (Fig. 6B, C).In contrast, the ck7 V130T/H131T mutant with smaller side chains did not cause a significant change in inhibition caused by 10 μM TNP-ATP (inhibition ratio = 0.847 ± 0.069 for ck7 V130T/H131T, vs. ck7 WT 0.844 ± 0.022, n = 4-21, p > 0.05, unpaired t-test, Fig. S3A, B).On the other hand, the dynamic recognition of TNP group may also be one of the reasons for the non-significant effect of the ck7 T112C mutant in the head domain and ck7 T202C mutant in the DF domain on TNP-ATP inhibition.
In addition, we tried to explore the role of conserved residues in the upper body domain, such as ck7 K181 and h3 K176 on TNP-ATP inhibition.These residues were shown to interact with TNP-ATP in both the ckP2X7 and hP2X3 crystal structures (Fig. 2).However, mutant made at neither of these sites was functional (Fig. S3C), occluding further functional analysis.
Thus, the structure of the ckP2X7/TNP-ATP complex yielded relatively precise receptor-ligand interactions, showing that the head, DF and TM regions played roles in the inhibition of TNP-ATP.However, the key recognition residues may not be ck7 T112 and ck7 T202 as shown in the ckP2X7/TNP-ATP complex structure.Functional studies suggest that ck7 V130 and ck7 H131, located in the head domain, are probably more important.

The lumen between the head and DF domains is equally important for TNP-ATP recognition of other P2X subtypes
Although a similar mechanism exists for TNP-ATP action on hP2X3 and ckP2X7, whether it is also applicable to the recognition of TNP-ATP by other P2X subtypes needed to be further examined.We have verified this being a common mechanism for rat P2X2 (rP2X2, Fig. 8A-E) and human P2X1 (hP2X1, Fig. 8F-G).

Discussion
Here, we evaluated two distinct modes of TNP-ATP binding to P2X receptors (hP2X3 and ckP2X7) suggested by recent structural studies and propose an improved mechanism of P2X/TNP-ATP action.Our results suggest that TNP-ATP interacts with P2X receptors in a similar pattern to that revealed by the structure of the ckP2X7/TNP-ATP complex, i.e., with TNP-ATP embedded in the P2X receptor adapting a "U" shape, similar to ATP, but not a "Y" shape as revealed by the structure of the hP2X3/TNP-ATP complex.TNP-ATP acts by interacting with amino acids in the head and DF domains, where it blocks the opening of the channel by restricting the movement of these two domains.When bound by TNP-ATP, the overall structure of the P2X receptor resembles semi-activation, but without ion permeation.
In the structure of the hP2X3/TNP-ATP complex, the TNP moiety of TNP-ATP is inserted between the LF and LB domains, which could limit channel opening by interfering with the movement of these two regions (Fig. 1E).However, although the introduction of the r3 K201C/V274C disulfide bond mutant indeed limited channel opening (Fig. 3C), the affinity to TNP-ATP was only slightly reduced in this mutant (Fig. 3F, G), suggesting that the effect most likely resulted from a weakened binding of the triphosphate group, which is common in both ATP and TNP-ATP, to rP2X3 (Fig. 3E).On the other hand, mutating F174, the only residue suggested by the resolved structure to interact with the TNP moiety of TNP-ATP, to an amino acid with a bulkier side chain also did not affect the inhibitory effect of TNP-ATP (Fig. 3H, I).These results imply that the resolved structure of the hP2X3/TNP-ATP complex may not represent a final state.This is supported by the results of our MetaD simulations, in which the structure of hP2X3/TNP-ATP flipped to a more stable ckP2X7/TNP-ATP-like conformation (POSE I and POSE II) during accelerated conformational sampling, and the TNP moiety that exerts the inhibitory effect interacted with residues in the head and DF domains (Figs. 4A-H and 5A-D).These results suggest that the resolved hP2X3/TNP-ATP structure may differ from the natural ligand recognition.This discrepancy between structural biology and other experiments might be due to the use of ligand soaking (the ckP2X7/TNP-ATP structure was determined by co-crystallization method) or some other conditions during sample preparation; for example, it is known that high salt concentrations and low temperatures affect the ligand-receptor interaction and stabilize it in some unnatural states.We also found high free energy barriers for the ckP2X7/TNP-ATP complex (POSE III, Fig. 5J) when it adapts a structure similar to the resolved hP2X3/TNP-ATP complex during MetaD-enhanced conformational sampling.
The structure of the ckP2X7/TNP-ATP complex showed that the cavity between the head and DF domains of the P2X receptor is the region where the TNP moiety of TNP-ATP is recognized, but mutations at ck7 T112 of the head domain and ck7 T202 of the DF domain, which show H-bonds with TNP in the crystal structure of the ckP2X7/TNP-ATP complex, did not affect the recognition of TNP-ATP (Fig. 7A-H).Instead, two other amino acids, ck7 V130 and ck7 H131, which do not show explicit interactions with the TNP moiety in the crystal structure, were found to have an important role in the recognition of TNP-ATP by ckP2X7 (Fig. 6B-D).This difference could result from the following reasons.First, the crystallization conditions might lead to some deviations such that the MetaD stimulation reveals more stable conformations (POSE I and POSE II; Fig. 5H, I) with ck7 V130 and ck7 H131 as the more critical residues that coordinate the interaction of TNP-ATP with ckP2X7.Second, as shown by CMD simulations, the TNP moiety of TNP-ATP has an important role in the recognition of ckP2X7, which could create multiple transient states during the ligand-receptor interaction (Fig. 2E, F).This means that the TNP-interacting amino acids in the head and DF domains of ckP2X7 may not be fixed and the resolved structure of the ckP2X7/TNP-ATP complex represents just one of the trapped conformations.Third, it cannot be ruled out that at a resolution of 3.1 Å, it is quite difficult to accurately define the side chain orientation of threonine in the ckP2X7/TNP-ATP structure.Therefore, whether the OG1 (oxygen) atom of either of the threonine side chains is actually oriented toward the TNP moiety of TNP-ATP to form a hydrogen bond is still questionable.Nonetheless, despite the details of the amino acids involved in binding to TNP, the contribution of the TM region to TNP-ATP inhibition, which was predicted by the ckP2X7/TNP-ATP complex structure but not the hP2X3/TNP-ATP structure, was confirmed in our study (Fig. 6E, F), demonstrating the relative soundness of the ckP2X7/TNP-ATP interaction mode.
Finally, we showed that equivalent amino acid residues in the head and DF domains of other subtypes of P2X receptors from other species, like rP2X2 and hP2X1, also play important roles in TNP-ATP recognition (Fig. 8A-H), indicating that the mechanism underlying the action of TNP-ATP in P2X receptors is a shared one, i.e., by inserting the TNP moiety in between the head and DF domains of the P2X receptor, blocking the conformational change of these two regions, thus preventing the opening of the channel to produce an inhibitory effect.Interestingly, the amino acids identified in the head and DF domains of hP2X3, ckP2X7, rP2X2 and hP2X1 that are critical for TNP-ATP recognition, i.e., h3 R126/L127, h3 K113, h3 M200, h3 K201, h3 R204, ck7 V130, ck7 H131, r2 R118, r2 V119, r2 H120, r2 K212, h1 G123 and h1 K215, are not conserved, which may stem from species and subtype differences.Particularly, the P2X7 subtype possesses a fragment between β2,3 with a long loop that is absent in all other subtypes [32].However, residues contributing to TNP-ATP recognition are more conserved in other subtypes with relatively similar structures (hP2X3, rP2X2, and hP2X1).This implies both conserved and specific mechanisms of action of the nonselective inhibitor TNP-ATP on different subtypes of P2X, which may be exploited for the development of subtype-specific competitive inhibitors.

Conclusion
Taken together, our study reveals a common mechanism of action of TNP-ATP on P2X receptors, opening new avenues for understanding the gating mechanism of P2X receptors and designing new modulators targeting P2X receptors.

Fig. 5 .
Fig. 5. Enhanced sampling by metadynamics (MetaD) shows relatively low binding free energy when TNP moiety is accommodated by the head and DF domains of the hP2X3 and ckP2X7 receptors.(A) Definition of two collective variables (CVs), CV1 (yellow) and CV2 (yellow), in MetaD simulations of hP2X3/ TNP-ATP (PDB ID: 5SVQ).The values of distance (CV1) and dihedral angle (CV2) are derived from the structure of hP2X3/TNP-ATP.(B, C) Two binding poses of TNP-ATP with lower relative free energy in the MetaD trajectory of hP2X3/TNP-ATP.(D) Three-dimensional (3D) reconstruction of the free energy surface (FES) based on MetaD simulations.POSE 0, POSE I and POSE II correspond to ΔG of + 14.79, + 1.59 and 0 kcal/mol, respectively.(E) The superimposition of the sampled conformation of hP2X3 (POSE II, lowest free energy, blue) and the structure of the ckP2X7/TNP-ATP complex (PDB ID: 5XW6, pink), showing the similarity of the TNP-molecule orientation after MetaD simulations of hP2X3 and ckP2X7.(F) 3D reconstruction of the FES.ΔG values are + 5.71, 0, + 0.03 and + 14.94 kcal/mol for POSE 0, POSE I, POSE II, and POSE III, respectively.(G) Definition of two collective variables (CVs), CV1 (yellow) and CV2 (yellow), during MetaD simulations of the ckP2X7/TNP-ATP complex (PDB ID: 5XW6).The values of distance (CV1) and dihedral angle (CV2) are derived from the structure of ckP2X7/TNP-ATP.(H, I) Two binding poses with lower relative free energy in the MetaD trajectory of ckP2X7/TNP-ATP.(J) POSE III derived from the MetaD trajectory of ckP2X7/TNP-ATP showing a similar.TNP-molecule orientation to hP2X3/TNP-ATP, but with higher free energy.(K) TNP-group moving from the ATP-binding pocket to the interface between the head and DF domains, as revealed by superimposing the initial (dark blue) and final (gray) poses after 120 ns of MetaD simulations.