Channel gating in response to extracellular ATP is a fundamental process for the physiological functions of P2X receptors. Here we identify coordinated allosteric changes in the left flipper (LF) and dorsal fin (DF) domains that couple ATP-binding to channel gating. Engineered disulphide crosslinking or zinc bridges between the LF and DF domains that constrain their relative motions significantly influence channel gating of P2X4 receptors, confirming the essential role of these allosteric changes. ATP-binding-induced alterations in interdomain hydrophobic interactions among I208, L217, V291 and the aliphatic chain of K193 correlate well with these coordinated relative movements. Mutations on those four residues lead to impaired or fully abolished channel activations of P2X4 receptors. Our data reveal that ATP-binding-induced altered interdomain hydrophobic interactions and the concomitant coordinated motions of LF and DF domains are allosteric events essential for the channel gating of P2X4 receptors.
P2X receptors are ATP-gated ion channels, and the opening of P2X receptors in response to extracellular (EC) ATP is a fundamental process in signal transduction in both excited and non-excited cells1. The delineation of ATP-binding, allosteric conformational changes in EC domains, together with pore dynamics correlated to channel gating, will promote the development of new pharmacological agents for P2X receptors2,3. Recently, the solution of crystal structures of the zebra fish P2X4 (zfP2X4, PDB entry codes: 3H9V and 4DW1) receptor in the presence and absence of ATP provided an insight into the architecture of the ATP-binding site and the ion channel pore, together with the conformational changes associated with channel gating4,5. Those structures confirm the predicted topology of P2X receptors that contains a large, glycosylated and disulphide-rich EC domain, two transmembrane (TM) domains and N- and C-intracellular termini3,4,5. The architecture of P2X receptors resembles the shape of a dolphin4,5, containing a rigid central EC body domain, which is attached by the structurally different head, left flipper (LF), right flipper (RF) and dorsal fin (DF) domains.
Numerous studies have demonstrated that a downward motion of the cysteine-rich head domain, tightening of the ATP-binding site jaw and upper vestibule expansion are essential for activation, desensitization and sustained activation of P2X receptors5,6,7,8,9,10, with data obtained from the application of multiple approaches, including electron microscopy10, voltage-clamp fluorometry9, engineering EC metal bridges8,11, normal mode analysis (NMA)7, molecular dynamics (MD) simulations7 and fast-scanning atomic force microscopy6. Those studies provide new insights into the dynamics of EC domain as well as the movements of the whole receptor during channel gating. Notably, a more pronounced deformation of the EC domain was observed in biophysical, biochemical and computational analysis of the P2X receptor than that observed in ATP-bound open structure, especially for the head domain and the upper vestibule6,10. These apparently less pronounced allosteric conformational changes may be attributable to different data from distinct subunits of P2X receptors. In addition, the possibility that the crystallization conditions have an effect12 on the native conformation of zfP2X4 cannot be completely excluded. Making further explorations on the allosteric changes of EC domain will contribute to our understanding of the nature and extent of ATP-binding-induced conformational changes on channel activation of this distinct family of trimeric ligand-gated ion channels.
Besides the head domain, the open structure of zfP2X4 also suggests that LF and DF domains undergo conformational changes during channel activation5. However, whether those allosteric changes are essential for channel activation or not and how ATP-binding contributes to allosteric changes of those two domains remain unclear. Taking advantage of the known crystal structures of closed and open states, and with the use of a multidisciplinary approaches including molecular modelling, site-directed mutagenesis, electrophysiology, protein expression analysis, engineered disulphide crosslinking and engineered zinc bridge, here we uncover key residues involved in ATP-binding-induced coordinated movements of LF and DF domains, and demonstrate the essential role of those relative motions in channel gating of P2X4 receptors.
Structural dynamics of LF and DF domains during simulations
The closed structure4 of zfP2X4 outlined the static architecture of this ion channel (Supplementary Fig. 1a). To further illustrate the inherent structural dynamics at the atomic level as well as the correlation between the dynamics and channel activation, a 300-ns MD simulation was carried out under conditions in which the effects of temperature (300 K), water and membrane on channel dynamics were adequately considered. Root mean square deviation calculations indicated that it took ~40–50 ns for this structure to attain the equilibrium state (Supplementary Fig. 1b), a period that is significantly longer than that of one of structurally similar trimeric channels13, that is, the ASIC3 channel14. This implied that the zfP2X4 receptor underwent significant structural rearrangements under normal temperature and pH conditions during the timescale of the MD simulations (Supplementary Movie 1). A superposition and comparison of the initial and the equilibrated representative conformation after MD simulations confirmed the structural dynamics of zfP2X4 at the room temperature in the absence of ATP, especially for β12-13, head, LF and DF domains (Fig. 1a and Supplementary Movie 1). Surprisingly, the integrated structural rearrangements of these domains resulted in a spontaneous closure of the ATP-binding site jaw (Fig. 1b and Supplementary Movie 1), and an expansion of the upper vestibule (Fig. 1c, Supplementary Fig. 1c and Supplementary Movie 1). Those structural changes are widely accepted to associate with channel opening6,7,8,10. Nevertheless, a comparison of pore size of the initial and representative conformation after MD simulations (Fig. 1c), and continuous pore-size measurements (Supplementary Fig. 1d) revealed that there was no obvious dilation of the TM pore observed even though these open-related structural events had taken place. These observations confirmed the essential role of inherent dynamics in favouring the tightening of ATP-binding site jaw as well as the expansion of the upper vestibule. Meanwhile, these results also implied that other open-related allosteric changes may be curbed because of the absence of ATP during MD simulations.
As revealed by a superposition of the initial and the equilibrated representative conformations (Fig. 1a) as well as a continuous measurement of Cα–Cα distances between zfP125 (head; the superscript ‘zf’ denotes zfP2X4 numbering, similarly hereinafter), zfV291 (LF) and zfH219 (DF) in MD simulations (Supplementary Fig. 2a,b,e,f), both the LF and DF domains underwent upward movements and came closer to the head domain when bound ATP is absent in MD simulations. Principal component analysis (PCA) of snapshots from MD simulations (Fig. 1d) or a low-frequency dynamics mode derived from NMA of the closed structure (Fig. 1e and Supplementary Movie 2) confirmed the tendency of upward motions of both the LF and DF domains in the absence of ATP. Interestingly, those movements are distinct from an upward motion of the DF domain and a downward motion of the LF domain observed in the ATP-bound open crystal structure of zfP2X4 (Fig. 2a). In addition, the relative motions with a downward motion of the LF domain and an upward motion of the DF domain were also observed in another set of MD simulation on the closed zfP2X4 structure complexes with pre-docked ATP (see below). These data indicated that the bound ATP might break the inherent dynamics of LF and DF domains, which has an essential role in channel gating of P2X4 receptors.
Disulphide crosslinking between LF and DF domains
To test this idea, we constrained the relative motions between LF and DF domains by introducing cysteine residues15 that can form interdomain disulphide bridge (Fig. 2b). Owing to the low potency of ATP on the zfP2X4 receptor expressed in mammalian HEK-293 cells5, all mutagenesis, electrophysiological recording and protein expressions measurements were carried out on identical residues of rat P2X4 (rP2X4), although the molecular modelling was based on the crystal structures of zfP2X4.
Cysteine-scanning mutagenesis in the LF and DF domains of the rP2X4 receptor has identified two positions, rV288 and rT211 (corresponding to zfV291 and zfS214 in zfP2X4, respectively; the superscript ‘r’ denotes rP2X4 numbering, similarly hereinafter; Fig. 2a), and when both are substituted by cysteine could form an interdomain/intersubunit disulphide bond (Fig. 2b). The wild-type rP2X4, and single mutants rV288C and rT211C, migrated on SDS-PAGE gels predominantly at position expected for the monomer subunit (~57 kDa; Fig. 2b). In contrast, no obvious monomeric form was observed for the subunit containing cysteine at both positions (rV288C/T211C; Fig. 2b). The observed higher molecular mass protein presumably represents a disulphide bond trimer, because β-mercaptoethanol (β-ME, 1%) reduced it to a monomeric size. β-ME also caused a modest shift of wild-type subunit or subunits with single cysteine substitution (Fig. 2b)owing to its effect on the large number of native cysteine residues within EC domain, which is also observed in another subtype of P2X receptors16. Nevertheless, it can be reasonably inferred that an intersubunit/interdomain disulphide is formed between the LF and DF domains in the double cysteine mutant rV288C/T211C.
The Cβ–Cβ distance between identical residues of zfP2X4, zfV291 and zfS214, was slightly reduced in open state (5.1 Å) compared with that of resting state (5.9 Å) (Fig. 2c). The Cα atom of zfV291 moved downwards ~5.2 Å and the Cα atom of zfS214 moved upwards ~4.9 Å along with relative motions between LF and DF domains (Fig. 2a; red arrows). Those relative movements led to a final ‘translocation’ between zfV291 and zfS214 (Fig. 2a,c). Based on closed crystal structure of zfP2X4, the predicted Cβ–Cβ distance of two cysteine residues of zfP2X4C291/C214 was reduced to ~3.71 Å (Fig. 2d). The predicted dihedral angle χ between the Cβ–Sγ–Sγ–Cβ atoms of zfC291 and zfC214 based on the closed crystal structure of zfP2X4) is 91.9° (Fig. 2d), a value very close to that of crystal structures (χCβ-Sγ-Sγ-Cβ=±90o)17. During MD simulations, the predicted dihedral angle χ and Cβ–Cβ distance based on the closed structure were stabilized at ~90° and ~3.8 Å (Supplementary Fig. 3b). In contrast, the predicted dihedral angle χ and Cβ–Cβ distance of zfV291C/S214C based on the open crystal structure of zfP2X4 markedly drifted from the original values during MD simulations (83.5° and 4.09 Å, Supplementary Fig. 3a), resulting in more obvious spontaneous structural rearrangements in the LF and DF domains (Supplementary Fig. 3f) than wild type (Supplementary Fig. 3e). Thus, the disulphide bridge between the LF and DF domains would be more compatible with the closed state rather than the open state, and will stabilize the conformation of those two domains in the closed sate. This means that the dihedral angle χ between Cβ–Sγ–Sγ–Cβ atoms of zfC291 and zfC214 would help to block the ‘translocation’ of zfC291 and zfC214, as well as relative motions between LF and DF domains during channel gating.
As a result, after expression in HEK-293 cells, the mutated channel of containing cysteine at both positions (rV288C/T211C) produced a smaller response to ATP (147±47 pA, n=11), when compared with channel carrying single substitutions rT211C (3833±731 pA, n=4) or rV288C (427±73 pA, n=8). The current of rV288C/T211C increased ~2–5 folds after dithiothreitol (DTT) application (Fig. 3a,c; n=9); the current declined again after treatment with H2O2; the second application of DTT could increase the current of rV288C/T211C again, suggesting that an interdomain disulphide formation between rV288C and rT211C impaired channel gating of rP2X4. Since DTT slightly decreased rather than increased ATP currents in HEK-293 cells expressing wild-type rP2X4 receptor (Fig. 3b,c; n=12), the breaking of disulphide bond between rV288C and rT211C may be responsible for the DTT-application-evoked increases in current amplitude of rV288C/T211C. The disulphide crosslinking between C288 and C211 decreased the apparent ATP affinity of rV288C/T211C (EC50, the concentration of ATP yielding a current that is half of the maximum, 31.4±13.7 μM, nH=1.74±0.3, and 6.44±2.5 μM, nH=1.78±0.16, for before and after DTT applications, respectively; mean±s.e.m. of EC50 values derived from 4–6 independent measurements; Fig. 3d,e), indicating that the relative motions between LF and DF domains were correlated with ATP-sensing of P2X4 receptors. However, the DTT application is also capable of potentiating the saturated ATP (0.3–1 mM)-induced current of rP2X4 receptors (Fig. 3d,e), suggesting that the relative motions between LF and DF domains are also essential allosteric changes that couple ATP-binding to channel activation. Those data suggested that the free relative motions between the LF and DF domains are pivotal for ATP-evoked activation of P2X4 receptors.
Engineered zinc bridge between LF and DF domains
To further confirm the crucial role of the relative motion of the LF and DF domains in channel activations, we introduced an engineered zinc bridge into the mutated P2X4 receptor containing histidine residues at those two positions. The zinc-bridge model of zfP2X4V291H/T214H based on the closed crystal structure of zfP2X4 was constructed as previously described8 (see Methods section). The distance (2.17 Å) and angle (107.7°) between the Zn2+ and coordinating atom NE2 of histidine residues (Fig. 4a) were very close to those obtained from the analysis of 111 crystal structures18. The zinc-bridge model of zfP2X4V291H/T214H based on the open crystal structure also produced reasonable values for the bond angle (110.2°) and bond lengths (2.14 Å; Supplementary Fig. 3c). However, the predicted model of the metal bridge between the LF and DF domains based on the open crystal structure was less stable during MD simulations when compared with the model based on the closed crystal structure (Supplementary Fig. 3d). During 50-ns MD simulations, the Cβ–Cβ distance and dihedral angles χ1 (along Cα–Cβ) of zfV291H and zfS214H were significantly altered, indicating that the zinc bridge may be not compatible with the open structure of zfP2X4. Similarly, the marked spontaneous structural rearrangements of LF and DF domains after such timescale of MD simulations were observed in zfV291H/S214H (Supplementary Fig. 3f) but not in the open structure of wild-type zfP2X4 (Supplementary Fig. 3e). Therefore, the zinc bridge between zfH291 and zfH214 would mimic the role of the disulphide crosslinking by constraining the relative motions between LF and DF domains through coordination bonds NE2–Zn–NE2 and the increased Cβ–Cβ distance (7.28 Å) between zfH291 and zfH214 (Fig. 4a).
Indeed, Zn2+ dose-dependently inhibited 100 μM ATP (saturating)-induced channel activation of rP2X4V288H/T211H (IC50, the concentration of Zn2+ yielding half of the inhibition=25.6±5.7 μM; nH=−0.97±0.1; Fig. 4b,c). As control, Zn2+ (200 μM, saturating) slightly potentiated rather than inhibited ATP currents of wild-type rP2X4 receptors as well as the single histidine mutant rT211H (Fig. 4b,c). Unexpectedly, Zn2+ inhibited ATP currents of another single histidine mutant rV288H (Fig. 4c), indicating that the native histidine residues or backbone atoms could form a metal site together with rV288H. Indeed, the double-mutant rV288H/H286A abolished Zn2+-mediated inhibition. However, Zn2+ inhibited ATP currents of triple mutants rV288H/T211H/H286A and rV288H/T211H/H140A, and the quadruple mutant rV288H/T211H/H140A/H286A (Fig. 4c), suggesting that the Zn2+-mediated blockage in rV288H/T211H was not relying on nearby native histidine residues (H140 and H286).
Taking advantage of lacking native histidine residue at the identical position of rH286 and rH140 in zfP2X4 (Supplementary Fig. 4c), we further tested the effect of Zn2+ on ATP currents of zfP2X4V291H/S214H. Zn2+ inhibited ATP currents of zfV291H/S214H but not of single mutants zfV291H and zfV214H, as well as the wild-type zfP2X4 (Supplementary Fig. 4d,e), confirming the idea that an interdomain metal bridge between LF and DF is sufficient to inhibit ATP currents of P2X4 receptors. Therefore, an engineered zinc bridge between LF and DF domains is capable of inhibiting the channel activation of P2X4 receptors by preventing the relative motion of LF and DF domains.
Altered hydrophobic interactions between LF and DF domains
To further address the mechanism underlying ATP-binding-induced coordinated motions of LF and DF domains, the ATP-binding mode at the very beginning of ATP contacting with resting zfP2X4 was established by in silico docking (Supplementary Fig. 5a). A superposition and comparison of the resting and open structures of zfP2X4 (Supplementary Fig. 5c), both with bound ATP, showed that the side chains of key residues K70, K72, R143 and T189 directly interacting with ATP underwent conformational changes during channel gating (Supplementary Fig. 5c). In contrast, the conformation of other residues R298, K316, N296, I232 and L191 had no significant changes during channel gating (Supplementary Fig. 5c). Notably, some partially conserved hydrophobic residues (Supplementary Fig. 4a) within the P2X receptor family, such as zfL217, zfI208, zfV291 and aliphatic side chain of zfK193 interacting with each other via hydrophobic interactions, underwent conformational changes passing from resting to open state (Fig. 5a). The dihedral angles of χ2 (along Cβ–Cγ) of zfK193 and zfL217 experienced mild changes during transformation from resting to open states (χ2zfK193=−178o versus −171o; χ2−zfL217=−71.7o versus −60.9o, resting state versus open state), whereas the χ1 angle (along Cα–Cβ) of zfV291 and χ2 angle of zfL208 underwent marked alterations (χ1−zfV291=−75° versus −170°; χ2-zfI208=173.2° versus −85.8°, resting state versus open state). The free-energy profiles for those interaction pairs were reconstructed by metadynamics, ‘a powerful algorithm that can be used for both reconstructing the free energy and accelerating rare events in systems described by complex Hamiltonians’19, where the dihedral angles of those four residues mentioned above were defined as the collective variables (CVs). The lowest free-energy paths (red arrows) plotted onto free-energy profiles passing from resting to open states revealed many ‘transition’ states (blue arrows) corresponding to higher potential energy along transition process, during which the bound ATP may help to overcome these energy barriers. Most importantly, those four residues were located at the interface between LF and DF domains (Fig. 5a), implying their possible roles in ATP-binding-induced coordinated movements of those two domains.
To examine the involvement of those hydrophobic residues in the movements of LF and DF domains, another set of normal MD simulation was used, in which the ATP-binding mode at the very beginning of ATP contacting with the resting zfP2X4 was set as the initial structure to replay the process of ATP-binding-induced coordinated movements of LF and DF domains. ATP made no contacts with side chains and main chains of zfK70, zfK72 and zfT189 at the initial ATP-bound zfP2X4 resting structure. However, as revealed by increased total number of H-bonds from 4 to 6–8 (Supplementary Fig. 6a), these contacts were quickly established or strengthened after 10 ns MD simulations. Furthermore, the N–O distance between zfK193 and the α-phosphate of ATP was shortened from 7.5 to 4.8 Å after 20 ns simulations, achieved by a long-distance electrostatic attraction (Supplementary Fig. 6b). In contrast, the zfK193–zfI208 distance made no significant changes (Supplementary Fig. 6c), indicating that the hydrophobic interaction between zfK193–zfI208 remained despite the change of the location of zfK193. Meanwhile, the distance of zfL217 (DF)-ATP was shortened from ~8 to 4.5–5.8 Å via hydrophobic interactions among the adenine base of ATP, DF and head domains after 20-ns simulations (Supplementary Fig. 6d), so were the distances of zfV291 (LF)–zfI208 and zfL217 (DF)–zfI208 (Supplementary Fig. 6e,f). Along with those structural rearrangements, zfV291, located on LF domain, was repelled by the ribose of ATP (Supplementary Fig. 6g). Most importantly, along with those allosteric changes, the relative motions between LF and DF domains comparable to the open crystal structure were observed (Supplementary Fig. 6h), especially for the LF domain, indicating that alterations in hydrophobic interactions of zfV291, zfI208, zfL217 and zfK193 were correlated well with ATP-binding-induced coordinated motions of those two domains.
Results of the lowest free-energy paths passing from resting to open states (Fig. 5b–e) and normal MD simulations also disclosed other details for bound-ATP-induced structural rearrangements of those four residues during channel gating. The transition paths revealed that the transition of the dihedral angle χ1 of zfV291 from −75°to −170° has priority over changes in dihedral angles of zfK193, zfI208 and zfL217 (Fig. 5b–d) in the process of passing from resting to open state, especially for zfI208 (Fig. 5d). Indeed, at the very beginning (3 ns) of normal MD simulations on closed zfP2X4 complexes with pre-docked ATP, we observed bound-ATP-induced changes of χ1 angle of zfV291 from −75° to −170°, whereas only mild changes occurred in the χ2 angles of zfK193, zfI208 and zfL217 (Supplementary Fig. 7a). The pre-docked ATP in closed zfP2X4 was also able to stabilize the χ1 angle of zfV291 at −170° during normal MD simulations (Supplementary Fig. 7b). Those data indicated that the bound-ATP-induced changes of χ1 angle of zfV291 and the following repelling action on zfV291 caused conformational changes of zfK193, zfI208 and zfL217. Those allosteric changes led to final relative motions between LF and DF domains.
Role of hydrophobic interactions between LF and DF domains
As illustrated in Fig. 6a, ATP-binding-induced coordinated movements of LF and DF domains may be achieved by the repelling of zfV291 (LF) and the closing movements of zfK193, zfL217 (DF) and zfI208 (DF) to ATP. In this process, the changed interactions of zfV291, zfI208, zfL217 and the aliphatic chain of zfK193 are essential for the relative motions between LF and DF domains. Indeed, mutations on rK190 of rP2X4 (corresponding to zfK193 of zfP2X4), rK190A, rK190L and rK190F, which lacked the basic group (rK190L), or had either smaller (rK190A) or bulkier (rK190F) size of the aliphatic side chain, strongly reduced the saturated ATP-induced maximal current of rP2X4 (Fig. 6b,c). These reduced activities may result from blocking the rK190-ATP interaction and the altered entire architecture of the hydrophobic interactions between LF and DF domains. Meanwhile, mutations in other three residues rI205, rL214 and rV288 of rP2X4 (corresponding to zfI208, zfL217 and zfV291 of zfP2X4, respectively) involved in hydrophobic interactions contributing to DF and LF domains coordinated movements, rI205G, rI205C, rI205D, rI205W, rL214A, rL214C, rL214W, rV288G, rV288A, rV288C and rV288F, with smaller or bulkier size of side chains, or with an introduction of hydrophilic side chain to break the hydrophobic interaction, also significantly diminished or fully abolished the maximal currents of rP2X4 (Fig. 6b,c). In contrast, consistent with the observation that zfL191, zfI232, zfL209 and zfI212, close to zfI208, zfL217, zfV291 and zfK193, displayed no significant conformational changes during the coordinated movements of LF and DF domains (Fig. 6a), mutations on the corresponding residues of rP2X4, rL188F, rL206A, rI209G, rI229A and rI229F only partially reduced the maximal current of rP2X4 (Fig. 6b,c).
In addition, the surface expression levels of both the functional or loss-of-function mutants of rP2X4 exhibited no pronounced changes when compared with that of the wild-type rP2X4 receptor, except for rI229F (Fig. 6d), indicating their critical role in channel gating rather than surface expression. Furthermore, alanine substitutions on the rI205, rL214 and rV288 led to significant reduction of maximal currents of rP2X4 (Fig. 6c), but only slightly decreased their apparent affinities of ATP (EC50=13.8±4.1, 22.5±6.3, 12.9±1.3 μM for rI205A, rL214A and rV288A, n=3–4, respectively) when compared with that of wild-type rP2X4 (EC50=7.34±2.8 μM, n=4) (Fig. 6e). Meanwhile, alanine replacement on rK190 led to a significantly increased EC50 value of rK190A (233±69 μM, n=4, ~30-fold decreased the ATP-sensitivity of rP2X4) possibly because of its direct electrostatic contact with ATP, indicating its important role in ATP-sensing. Notwithstanding, the contribution of rK190 in ATP-sensing is much weaker than other key residues involved in directly contacting with ATP20. Our data revealed that the residues involved in ATP-binding-induced altered interdomain hydrophobic interactions are essential for channel activation of rP2X4 receptor rather than the surface expression and the ATP-sensing, at least for rI205, rL214 and rV288.
Here we identify the indispensable role of relative motions between the LF and DF domains in channel gating of P2X4 receptors using a combination of computational and experimental studies. All the molecular modelling was done on the crystal structure of zfP2X4, whereas the mutagenesis was done with the rP2X4. This is because ATP is not very potent at activating the zebra fish receptor expressed in HEK cells. Despite the difference in P2X sequences among species, the number of residues that form the LF, DF and head domains are the same between rP2X4 and zfP2X4 receptors. The key hydrophobic residues located at the interface of the LF and DF domains are also highly conserved between rP2X4 and zfP2X4 (Supplementary Fig. 4a). As a result, mutations of those residues in zfP2X4 significantly reduced ATP (1 mM)-induced currents (Supplementary Fig. 4f,g). A metal bridge between the LF (zfV291H) and DF (zfV214H) domains also inhibited ATP currents of zfP2X4 (Supplementary Fig. 4d,e). All this evidence suggests that the same mechanism is responsible for coupling LF and DF’s dynamics with channel gating in rP2X4 and zfP2X4.
The molecular modelling together with mutagenesis in our work revealed that ATP-binding-induced alterations in interdomain hydrophobic interactions and the concomitant coordinated motions of the LF and DF domains are allosteric changes essential for the channel activation of P2X4 receptors. Those conformational changes are distinct from previously identified allosteric events essential for channel activations, such as the downward motion of cysteine-rich head domain, tightening of the ATP-binding site jaw and the upper vestibule expansions5,6,8,9. As revealed by the spontaneous closure of the ATP-binding site jaw (Fig. 1b) as well as the upper vestibule expansion in the absence of ATP-binding during MD simulation (Fig. 1c), inherent dynamics has a dominant role in determining those allosteric events. However, MD simulations together with NMA analysis revealed that the bound ATP may break the inherent dynamics of LF and DF domains, suggesting that the relative motions between the LF and DF domains evoked by bound ATP may act as a conformational switch to mediate the channel activation of P2X4 receptors. This assumption was supported by the findings that the engineered disulphide crosslinking and Zn2+ bridge between the LF and DF domains significantly influenced the activation of P2X4 receptors (Figs 3 and 4).
Although ATP-binding-induced relative motions between LF and DF domains are allosteric events indispensable for the channel activation of P2X4 receptors, the mechanism underlying this process still remains unclear. This may be because of the following possible reasons: first, the hydrophobic interactions of the LF and DF domains at the resting state may develop an energy barrier to prevent channel activations. The downward motion of LF domain induced by ATP-binding is essential for overcoming of this energy barrier and favours downstream allosteric changes coupling to channel activations of P2X4 receptor. Second, given the rigidity of the body domain, like a ‘lever’, the ATP-bound crystal structure reveals a ‘lever’ mechanism that link binding site closure to iris-like pore opening5,21. The DF domain is structurally coupled to this ‘lever’. The upward movement of the DF domain induced by hydrophobic interactions among DF and head domains may drive the expanding motion of the lower body domain (‘lever arm’). Third, LF domain is located at the intersubunit interface of two lower body domains. One side of the LF domain is structurally coupled to the body domain of one subunit, and the other side contacts with the body domain and DF domain of another subunit. Bound ATP-evoked altered hydrophobic interactions between the LF and DF domains, and the downward motion of the LF domain might favour outward flexing of those two lower body domains and facilitate the final opening of P2X4 receptors. Indeed, during MD simulations on the closed zfP2X4 complexes with pre-docked ATP, we have observed a mild pore dilation in the TM regions, which became more obvious along with the relative motions of the LF and DF domains (Supplementary Fig. 8a,c) when compared with the simulations on receptors in the absence of pre-docked ATP (Supplementary Fig. 8c). The obvious expansion of central vestibule (formed by lower body domains of three subunits) and upward movements of linker regions (Supplementary Figs 1a and 8a–c) resulted in outward movements of TM1 and TM2 (Supplementary Fig. 8a). Those allosteric changes led to a final pore dilation of zfP2X4, consistent with the observation in the open crystal structure (Supplementary Fig. 8d). In contrast, during MD simulations on the closed structure without pre-docked ATP, only a mild spontaneous expansion of central vestibule and the upward movements of linker regions were detected (Supplementary Fig. 8e). Therefore, the rigid lower body domains and linker regions may enable the coupling of the relative motions between LF and DF domains and the outward motions of the TM helixes.
Finally, taking into consideration the crucial role of relative motions between LF and DF in channel gating of P2X4 receptors, the cleft existing among LF, DF and head domains provides an opportunity to develop new subtype-specific small molecules interacting with this site (Supplementary Fig. 4b). We proposed that a bound small molecule at this site may potentially block the activation of P2X receptors. Here are some hints on this hypothesis: first:, the bound small molecules at this site may stabilize the interactions between LF and DF domains, and curb their relative motions via contacts with the hydrophobic residues in the interface of these two domains, such as rV288, rI205, rL214 and their nearby residues. Those interactions will increase the energy barrier for the channel gating. Second, as revealed by the crystal structures of zfP2X4 (refs 4, 5), the bound ATP makes contacts with residues located at the interface of the LF, DF and head domain via weak hydrophobic interactions (Supplementary Fig. 4b). The bound small molecules occupying this cleft/cavity may compete with the ATP. Finally, the residues exposed to this cleft/cavity are not conserved throughout the P2X receptor family (Supplementary Fig. 4b,c), providing a possibility to develop subunit-specific blockers of P2X receptors. Thus, the identification of open-related movements of LF and DF domains together with identified allosteric changes of head domain will help to develop subunit-specific blockers of P2X receptors.
In summary, ATP-binding-evoked P2X channel gating is a very precise and complex process, including allosteric conformational changes in head, LF, DF, body and TM domains. ATP-binding has roles in tightening the binding site and coordinating the motion of whole receptors, which leads to the reduction of energy barrier and the subsequent channel activation. The identification of open-related movements of those domains would enable the evaluation of new strategies aiming at developing new subtype-specific blockers of P2X receptors.
ATP, ZnCl2 and most of other drugs were purchased from Sigma (St Louis, MO, USA). The plasmid pcDNA3-rP2X4 is the gift from Drs Alan North and Lin-Hua Jiang, and the cDNA of zfP2X4.1 obtained from Dr Eric Gouaux was subcloned into the pcDNA3 vector (all the primers that have been used see Supplementary Table 1). Each mutant was generated with a QuikChange mutagenesis kit. Individual mutations were verified by DNA sequencing. All constructs were expressed in HEK-293 cells. HEK-293 cells were cultured in DMEM medium at 37 °C in a humidified atmosphere of 5% CO2 and 95% air. Transfections of plasmids were performed using Hilymax (Dojindo Laboratories, Kumamoto, Japan) following the manufacturer’s recommendation. Electrophysiological measurements were performed 24–48 h after transfection as we described previously22.
The electrophysiological recordings were performed using the conventional whole-cell configuration under the voltage clamp at room temperature (23±2 °C) as our previous description14,22. In brief, patch pipettes were pulled from glass capillaries on a two-stage puller (PC-10, Narishige). The resistance between the recording electrode was filled with pipette solution and the reference electrode was 3–5 MΩ. Membrane currents were measured using an Axon patch 200B patch clamp amplifier (Molecular Devices) and were filtered at 2 kHz using low-pass Bessel filter. All currents were sampled and analysed using a Digidata 1440 interface and a computer running the Clampex and Clamp-fit 10.0 software (Molecular Devices). The bath solution contained 150 mM NaCl, 5 mM KCl, 10 mM glucose, 2 mM CaCl2.2H2O, 10 mM HEPES and 1 mM MgCl2.6H2O; the pH was adjusted to 7.35. Patch electrodes were filled with solution containing 30 mM NaCl, 120 mM KCl, 1 mM MgCl2.6H2O, 0.1 mM CaCl2.2H2O and 5 mM EGTA; the pH was adjusted to 7.35. In most experiments, 70–90% of the series resistance was compensated. The membrane potential was held at −60 mV throughout the experiment under voltage-clamp conditions. ATP solutions were prepared in batch buffer for 2 h at most and applied using a fast pressure-driven microperfusion system (OctaFlow08P, computer controlled, ALA Scientific Instruments). Responses were normalized to cell membrane capacitance. Dose–response data were collected from recording of a range of ATP concentrations buffer, and the corresponding currents were normalized to the highest current amplitude; ATP-gated currents were recorded after regular 3 s ATP application every 2 min. For higher ATP concentrations (>10 μM), the pulses were spaced up to 5–15 min to avoid receptor desensitization. To abolish the rundown of currents in multiple-dose applications of ATP, the amphotericin-perforated patch clamp23 was also used for current recordings of dose-dependent activations of rP2X4V288C/T211C in initial absence and the subsequent presence of reducing agents, as well as the dose-dependent inhibition of Zn2+ on ATP currents of rP2X4V288H/T211H. During this procedure, ATP-gated currents were recorded after regular 15–20 s ATP application every 8 min to avoid receptor desensitization.
Cell surface biotinylation and western blotting analysis
Cell surface biotinylation and western blotting were performed as we previously described24. In brief, the HEK-293 cells transfected with rP2X4 or its mutants were washed in chilled PBS+/+, and then incubated with sulfo-NHS-LC-biotin. The reaction was terminated by further incubating the cells with glycine in PBS. The cells were then collected and lysed with RIPA buffer. Biotinylated proteins were separated from the intracellular protein fraction using agarose resin linked to NeutrAvidin by incubation overnight and subsequent centrifugation. The resins were then washed and bound proteins were eluted with the boiling SDS sample buffer, while 10% of the volume of the supernatant was diluted and used as the total protein fraction. Protein samples from biotinylation assay were analysed by western blotting. The samples were separated by SDS-PAGE and transferred to a polyvinylidene difluoride (PVDF) membrane. The membrane was incubated overnight at 4 °C with anti-EE-tag (1:1,000; Abcam, catalogue number ab40767) or anti-GAPDH (1:1,000; Sungene Biotech, catalogue number KM9002) antibodies, followed by appropriate HRP-conjugated secondary antibodies for EE-tag (25 °C, 1 h, 1:1,000; Goat-Rabbit IgG(HtL)-HRP; Sungene Biotech, catalogue number: LK2001) or GAPDH (25 °C, 1 h, 1:3,000, Goat-Mouse IgG(HtL)-HRP; Sungene Biotech, catalogue number LK2003) and finally visualized using an ECL solution (Thermo) and exposure with ImageQuant RT ECL system (GE Healthcare, USA) for 1–3 min. All western blots and gels are accompanied by the locations of molecular weight markers (Thermo PageRuler Prestained Protein Ladder 10–170 kDa, catalogue number 26617) (Supplementary Fig. 9).
Protein dynamics simulations
All MD simulations were performed by using the programme DESMOND 3.0 (ref. 25) with constant number of particles, pressure and temperature (NPT) and periodic boundary conditions (10 Å buffer distance to each dimension of the simulation box), which uses a particular ‘neutral territory’ method called the midpoint method to efficiently exploit a high degree of computational parallelism. A default OPLS_2005 force field, following the functional form of the OPLS-AA family of force fields with additional stretch, bend and torsional parameters for better coverage of ligand functional groups26, was used for the protein and ligand molecules. The energy-minimized structures of zfP2X4 by DESMOND 3.0 at the resting (PDB ID code 3H9V) or the open state (PDB ID code 4DW1) were used as the starting structures for MD simulations. The ATP-binding mode at the resting state of zfP2X4 was established by Autodock 4.2 (ref. 27); the pose with similar features as the open crystal structure of zfP2X4 was minimized by DESMOND 3.0 and was selected as the initial structure in the ATP-resting zfP2X4 complexes simulation system. A large dimyristoylphosphatidylcholine bilayer consisting of 253 membrane molecules, was constructed to generate a suitable membrane system in which the TM domain of the zfP2X4 could be embedded. The protein/dimyristoylphosphatidylcholine system was then solvated in a bath of simple point charge water molecules. Counter ions were subsequently added to compensate for the net negative charge of the system. NaCl (150 mM) was added into the simulation box that represents background salt at physiological conditions. To maintain the system at a constant temperature of 300 K, the Berendsen algorithm was applied to couple protein and other molecules separately with a coupling time of 0.1 ps. All of the bond lengths including hydrogen atoms were constrained by the Linear Constraint Solver algorithm. Electrostatic interactions between charged groups at a distance <9 Å were calculated explicitly; long range electrostatic interactions were calculated using the smoothed particle mesh Ewald method. All of the MD simulations were run on the DAWNING TC2600 (with 200 AMD Opteron 8374HE CPUs). Preparation, analysis and visualization were performed on a 12-CPU CORE DELL T7500 graphic working station. The pore radius was measured by using the programme HOLE28. PCA and NMA analysis were conducted by using the programme ProDy 1.1 (ref. 29). PCA modes of zfP2X4 dynamics were obtained by essential dynamics analysis of snapshots of MD simulations of zfP2X4. Low-frequency NMA modes of apo zfP2X4 were predicted by Anisotropic Network Model) Module of Prody 1.1 (ref. 29). These two measurements were performed using the default Prody parameters. Ten modes were generated in both NMA and PCA analysis by Prody. The modes with a tendency of closing the ATP-binding site jaw were selected.
Models of zfP2X4V291C/214C and zfP2X4V291H/S214H
The structures of the closed (PDB ID code 3H9V) and open (PDB ID code 4DW1) zfP2X4 receptors were taken as the templates to construct the zfP2X4V291C/214C and zfP2X4V291H/S214H models with the programme MODELLER30. For the disulphide bridge construction, the disulphide bond between two cysteine residues C291 and C214 was auto-connected by MODELLER30. Zinc-bridge model was constructed according to previous procedure8 using MODELLER. In brief, for zinc-binding site reconstructions, a distance constrain (2.0–2.2 Å) was added between zinc and the NE2 atom of histidines and then was minimized by DESMOND 3.0 with OPLS_2005 force field31. The resulting two models of zfP2X4V291C/214C and zfP2X4V291H/S214H were further optimized by 1.2-ns MD simulations using programme DESMOND 3.0 with OPLS_2005 force field31. After such timescale of MD simulations, various parameters of zinc and disulphide bridges, including distances between atoms, bond angles and dihedral angles were very closing to those obtained by analysis of crystals.
Metadynamics19,32,33 is a technique in which the potential for one or more chosen variables is modified by periodically adding a repulsive potential of Gaussian shape at the location given by particular values of the variables. In the present study, all metadynamics analysis were conducted by programme DESMOND 3.0 (ref. 25) under NPT and periodic boundary conditions using the default parameters at constant temperature (300 K) and pressure (1 bar) by using the Berendsen coupling scheme with one temperature group. All simulations used the OPLS_2005 force field for proteins and ions and the simple point charge model for water. The dihedral angles of side chain of key residues were set up as CVs. To set up CVs, four carbon atoms (Cα, Cβ, Cγ and Cδ for zfK193, zfI208 and zfL217, and Cα, Cβ, Cγ and one backbone carbon atom for zfV291) were selected to define the dihedral angles. The parameters for height, width of the Gaussian, and the interval were set to 0.03 kcal mol−1, 5° and 0.09 ps, respectively. All metadynamics runs were protracted for ~20,000 ps until they showed a free diffusivity along the defined CVs. This criteria was chosen according to the guidelines of assessing the accuracy of metadynamics runs33. The sum of the Gaussians and the free-energy surface were generated by METADYNAMICS ANALYSIS tools of DESMOMD 3.0.
The results are expressed as the means±s.e.m. Statistical comparisons were made using the t-test, where P<0.05 (*) or P<0.01 (**) was considered significant. Concentration–response relationships for ATP activation of wild-type or mutated channels were obtained by measuring currents in response to different concentrations of ATP, and all of the results used to generate a concentration–response relationship were from the same group. The data were fit to the Hill equation: I/Imax=1/[1+(EC50/[ATP])n], where I is the normalized current at a given concentration of ATP, Imax is the maximum normalized current, EC50 is the concentration of ATP yielding a current that is half of the maximum and n is the Hill coefficient.
How to cite this article: Zhao, W.-S. et al. Relative motions between left flipper and dorsal fin domains favour P2X4 receptor activation. Nat. Commun. 5:4189 doi: 10.1038/ncomms5189 (2014).
We thank Drs Lin-Hua Jiang, Alan North and Eric Gouaux for their kind gifts of plasmids pCDNA3-rP2X4 and pFatBac OP-1-zfP2X4.1. We also thank Dr Tian-Le Xu for his permission to use DAWNING TC2600, and thank Drs Michael Xi Zhu, Ying Xue, Shang-Wei Hou and Ming Zhou for their helpful comments on this manuscript. This study was supported by grants from the National Excellent Young Scientist Foundation of China (no. 31222018), National Program on Key Basic Research Project of China (2014CB910300, 2014CB910302 and 2010CB529800), National Natural Science Foundation of China (no. 31170787), Key National S&T Program ‘Major New Drug Development’ Grant (no. 2012ZX09504001-003), Natural Science Foundation of Shanghai (12ZR1415600) and Natural Science Foundation of Jiangsu Province (SBK201241903), ‘Shanghai Jiao Tong University-SMC Mutual Funds’ for Excellent Young Scholar and ‘Opening project’ of Chinese Academy of Sciences, Shanghai Institute of Organic Chemistry.
Spontaneous conformational changes of the closed zfP2X4 structure during MD simulations. Significant structural alterations in head, LF, DF and β12-13 domains, the spontaneous closure of the ATP-binding site jaw, and the expansion of upper vestibule were observed during 300-ns MD simulations.
The downward movement of the head domain, upward movements of the LF and DF domains in low frequency NMA mode of the closed structure of zfP2X4 receptors.