Subtype-specific control of P2X receptor channel signaling by ATP and Mg²⁺

Significance

ATP is an important extracellular signal that activates P2X receptor channels. Although a large fraction of ATP is bound to divalent cations in vivo, the forms of ATP that activate P2X receptors are unknown. Here we show how the activity of homomeric P2X receptors is tuned by Mg²⁺ in some subtypes by preventing activation by free ATP, and in others by binding to a distinct regulatory site. We also find that both regulatory mechanisms are disengaged in heteromeric channels to form a sensitive ATP signaling pathway. These fundamental properties of P2X receptors will be valuable for investigating their physiological functions.

Abstract

The identity and forms of activating ligands for ion channels are fundamental to their physiological roles in rapid electrical signaling. P2X receptor channels are ATP-activated cation channels that serve important roles in sensory signaling and inflammation, yet the active forms of the nucleotide are unknown. In physiological solutions, ATP is ionized and primarily found in complex with Mg²⁺. Here we investigated the active forms of ATP and found that the action of MgATP²⁻ and ATP⁴⁻ differs between subtypes of P2X receptors. The slowly desensitizing P2X2 receptor can be activated by free ATP, but MgATP²⁻ promotes opening with very low efficacy. In contrast, both free ATP and MgATP²⁻ robustly open the rapidly desensitizing P2X3 subtype. A further distinction between these two subtypes is the ability of Mg²⁺ to regulate P2X3 through a distinct allosteric mechanism. Importantly, heteromeric P2X2/3 channels present in sensory neurons exhibit a hybrid phenotype, characterized by robust activation by MgATP²⁻ and weak regulation by Mg²⁺. These results reveal the existence of two classes of homomeric P2X receptors with differential sensitivity to MgATP²⁻ and regulation by Mg²⁺, and demonstrate that both restraining mechanisms can be disengaged in heteromeric channels to form fast and sensitive ATP signaling pathways in sensory neurons.

Seven subtypes of P2X receptors have been identified in mammals that can form either homomeric (P2X1, P2X2, P2X3, P2X4, P2X5, P2X7) or heteromeric (P2X1/2, P2X1/4, P2X1/5, P2X2/3, P2X2/5, P2X2/6, P2X4/6, and possibly, P2X4/7) channels (1–8). These subtypes of P2X receptors have distinct gating properties, pharmacology, and cellular distributions. P2X1 and P2X3 receptors desensitize within a few hundred milliseconds when opened by ATP, and their distributions are restricted to either smooth muscle cells and platelets (P2X1) or a subset of sensory neurons (P2X3) (1, 9–14). P2X2 and P2X4 receptors exhibit slow desensitization during prolonged ATP application, and these receptors are the most abundant subtypes in the central nervous system (15). P2X2 subunits also express in a subset of sensory neurons; however, in these cells they only form heteromeric channels with P2X3 subunits (12, 16, 17). In sensory neurons, P2X3 homomeric channels together with P2X2/3 heteromeric channels play important roles in mediating the primary sensory effects of ATP, and knock-out animals with either P2X3 deletion or P2X2 and P2X3 double-deletions have revealed critical roles in taste, pain, oxygen sensing, and bladder filling (17–20).

A long-standing conundrum in P2X receptor-mediated signaling concerns the forms of ATP that activate these channels. In neutral solutions, ATP is ionized and exists mostly as free ATP (ATP⁴⁻), an efficient chelator of divalent cations such as Mg²⁺, and to a lesser extent Ca²⁺ (21). In extracellular biological compartments, such as the synaptic cleft, Ca²⁺ and Mg²⁺ are present in the millimolar range, and therefore only a relatively small fraction of ATP released from vesicles is present in the free form. Although a range of important studies have explored the regulatory effects of Ca²⁺ and Mg²⁺ on P2X receptor channels (22–30), the essential question of which forms of ATP serve as agonists remains unresolved. Several previous studies have reported that P2X2, P2X7, and the native P2X receptors in cilia are activated by ATP in solutions containing low concentrations of divalent cations, and that the addition of divalent cations shifts the concentration dependence for activation of the channels to higher ATP concentrations, suggesting that either ATP⁴⁻ is the most active form of ATP or that divalent cations regulate those subtypes through allosteric mechanisms (27, 30–36). In the present study, we investigated the form(s) of ATP that serve as agonists for a range of subtypes of P2X receptor channels. Our primary focus was to determine whether ATP⁴⁻ or MgATP²⁻ are the principal agonists and to explore whether Mg²⁺ might serve specific regulatory roles. Our results demonstrate that the action of MgATP²⁻ and ATP⁴⁻ differ between subtypes of P2X receptors, and reveal that heteromeric channels can have unique hybrid phenotypes, findings that will be crucial for understanding the physiological functions of these channels in both the peripheral and central nervous systems.

Results

ATP⁴⁻ Activates All Subtypes of Homomeric P2X Receptor Channels.

We began by establishing whether ATP⁴⁻ can activate each subtype of P2X receptor that can be heterologously expressed as a homomeric channel. To test for activation by ATP⁴⁻, we used the patch-clamp recording technique to record ATP-activated currents when applying the nucleotide in the presence of 10 mM EDTA to remove contaminating divalent ions that are present in aqueous solutions. Because divalent ions stabilize seals between electrode glass and cell membranes that are required for patch-clamp recordings (37, 38), we first established GΩ seals with 0.5 mM Mg²⁺ and 2 mM Ca²⁺ in the extracellular solution, and only studied cells further if the seal was exceptionally tight (>10 GΩ resistance) and stable for 5 min after switching to a 10-mM EDTA solution. Using this approach to study HEK cells transfected with P2X receptor cDNA, we observed that 300 μM ATP evokes currents with kinetic properties resembling those that have been reported in solutions with divalent cations (1) (Fig. 1A). Both P2X1 and P2X3 receptors rapidly desensitized in response to ATP⁴⁻, whereas the others exhibited a more slowly desensitizing phenotype. For P2X1, P2X2, P2X3, and P2X4 receptors, the EC₅₀ concentrations for ATP activation in the divalent-free solution were between 1 and 2 μM (Fig. 1B). In agreement with previous studies, the ATP-activated P2X5 currents are typically very small, and P2X7 receptors activate slowly and with low affinity for ATP, limitations that precluded further experiments with these subtypes (34, 39). We also did not study P2X6 receptors because they are difficult to express as homomeric channels (40, 41).

Fig. 1.

Divalent-free ATP can open P2X receptor channels. (A) ATP (300 μM) activates P2X receptors expressed in HEK cells when the agonist is applied in a divalent-free solution containing 10 mM EDTA. Macroscopic ATP-activated currents were measured using a holding voltage of −60 mV. (B) Concentration–response relations for ATP activation of P2X1, P2X2, P2X3, and P2X4 receptors in divalent-free solution. Each point is the average ± SEM of four measurements. The Hill equation was used to fit the data, and the EC₅₀ and (n) are 1.2 ± 0.2 μM and (1.5 ± 0.2) for P2X1; 1.4 ± 0.1 μM and (1.3 ± 0.1) for P2X2; 2.0 ± 0.7 μM and (1.1 ± 0.3) for P2X3; and 1.3 ± 0.1 μM and (1.3 ± 0.1) for P2X4.

Differential Responses to MgATP²⁻ for Homomeric P2X Receptor Channels.

The above results demonstrate that ATP can activate P2X receptor channels in the absence of divalent cations, indicating that ATP in its free anionic form is an agonist. Our next objective was to test whether MgATP²⁻ can also activate P2X receptors. We reasoned that if Mg²⁺-bound ATP is also an agonist of P2X receptors, a solution containing a substantial amount of MgATP²⁻ and a minimal amount of ATP⁴⁻ would effectively activate these channels. To test this possibility, we activated P2X receptors with a saturating concentration of ATP (10 μM) in divalent-free solution, and compared the extent of activation to that observed in the presence of 5 mM MgCl₂, a solution that contains ∼0.2 μM ATP⁴⁻ and ∼9.8 μM MgATP²⁻ (Fig. 2; Materials and Methods). According to the concentration dependence for activation of P2X1–4 receptors by ATP in divalent-free solution (Fig. 1B), the currents activated by a solution containing ∼0.2 μM ATP⁴⁻ and ∼9.8 μM MgATP²⁻ should be between 5% and 10% of the current activated by 10 μM ATP⁴⁻ if MgATP²⁻ does not activate the channel.

Fig. 2.

MgATP²⁻ activates P2X1 and P2X3, but not P2X2 and P2X4 receptors. Macroscopic currents activated by 10 µM ATP in the absence and presence of Mg²⁺ were recorded at −60 mV from HEK cells expressing P2X1 (A), P2X3 (B), P2X2 (C), or P2X4 (D). (A–D) The solution of 10 μM ATP plus 5 mM MgCl₂ contains ∼0.2 μM ATP⁴⁻ and ∼9.8 μM MgATP²⁻. Between ATP pulses, cells were perfused with divalent-free solution. To estimate rundown, 10 μM ATP was applied twice at 90-s intervals in the divalent-free solution, and all subsequent ATP applications were given at 90-s intervals. Similar experiments were repeated on 3–4 different cells. (Right) Summary of normalized currents in response to each ATP pulse. Numbering indicates the sequence of ATP applications from first (1) to last (4). P2X1–P2X3 subtypes were studied with conventional whole-cell recording, and P2X4 was studied with the perforated-patch configuration. I_R observed in P2X2 and P2X4 are indicated with an asterisk. (E and F) Enlarged views of I_R for P2X2 (E) or P2X4 (F). Scale bars on the left show the amplitude of currents normalized to that activated by 10 μM ATP⁴⁻.

In the case of P2X1 and P2X3 receptor channels, >90% of maximum activation was observed in the presence of 5 mM MgCl₂ (Fig. 2 A and B), suggesting that MgATP²⁻ is an agonist for these two subtypes. However, in the case of P2X2, this Mg-ATP solution activated very little steady-state current (Fig. 2C), and for P2X4, the fractional activation was ∼10% (Fig. 2D). Notably, upon exchanging the test extracellular solution containing both Mg²⁺ and ATP with the control solution containing EDTA, we detected a rapid increase in the inward current for cells expressing either P2X2 or P2X4, which we termed a resurgent current (Fig. 2 C and D, asterisk, and Fig. 2 E and F, expanded views). This resurgent current (I_R) required solution exchange times of <50 ms to be consistently detected, and it decays with similar kinetics for deactivation of these channels following removal of ATP⁴⁻. We therefore hypothesized that MgATP²⁻ can bind to P2X2 and P2X4 receptor channels, but that MgATP²⁻ cannot open P2X receptor channels as efficiently as ATP⁴⁻. During external application of the Mg-ATP solution, both ATP⁴⁻ and MgATP²⁻ bind to the channel (with occupancies determined by their relative affinities and concentrations), but the steady-state current is dominated by the relatively small fraction of channels with ATP⁴⁻ bound. In this scenario, upon removal of Mg²⁺ and ATP, I_R is observed because Mg²⁺ unbinds before ATP⁴⁻, enabling a larger fraction of channels to open than would be possible if only ATP⁴⁻ bound to the receptor when applied in the presence of Mg²⁺. In the following sections, we further explore this hypothesis, focusing on the P2X2 subtype, a channel that displays the least desensitization and most stable ATP-activated currents.

MgATP²⁻ Binds to P2X2 Receptors.

If I_R results from Mg²⁺ unbinding from channels that already have MgATP²⁻ bound, then the amplitude of that current should vary with the concentration of MgATP²⁻ in the external test solution. To test whether this is the case, we varied the concentrations of MgATP²⁻ in the external solution from ∼0.9 to 29.9 μM by increasing the concentration of both Mg²⁺ and total ATP while maintaining ATP⁴⁻ at 0.1 μM, a concentration that by itself would cause minimal activation of P2X2 receptor channels (Fig. 1B). As the concentration of Mg²⁺ and ATP increase, we observed steady-state currents that were consistently small when the test solution was applied; however, the amplitude of I_R increased along with the concentration of MgATP²⁻ (Fig. 3 A and B), in line with the idea that MgATP²⁻ can bind to P2X2 receptors and that this complex is not effective at opening the channel.

Fig. 3.

MgATP²⁻ can bind to P2X2 receptors. (A) P2X2 currents activated by 10 μM free ATP or 0.1 μM ATP⁴⁻ and varying concentrations of MgATP²⁻. EDTA solution was applied to the cell between each ATP application. Resurgent currents (I_R) are marked with an asterisk. The four test MgATP²⁻ solutions were made using the following concentrations of ATP and MgCl₂: 0.1 μM ATP⁴⁻/0.9 μM MgATP²⁻ (1 μM ATP plus 1 mM MgCl₂); 0.1 μM ATP⁴⁻/2.9 μM MgATP²⁻ (3 μM ATP plus 3 mM MgCl₂); 0.1 μM ATP⁴⁻/9.9 μM MgATP²⁻ (10 μM ATP plus 10 mM MgCl₂); and 0.1 μM ATP⁴⁻/29.9 μM MgATP²⁻ (30 μM ATP plus 30 mM MgCl₂). (B) Normalized resurgent current (I_R/I_Max) plotted against MgATP²⁻ concentration (n = 4). I_max is the current activated by saturating ATP⁴⁻. (C) Comparison of currents activated with an EC₂₀ concentration of ATP⁴⁻ with either low (0.6 µM) or high (10 µM) MgATP²⁻. The concentrations of ATP⁴⁻, MgATP²⁻, and Mg²⁺ are 0.4 μM, 0.6 μM, and 199.4 μM in 1 μM ATP plus 200 μM MgCl₂ solution, and 0.4 μM, 9.6 μM, and 2.99 mM in 10 μM ATP plus 3 mM MgCl₂ solution. (D) Superimposed activation of P2X2 receptor channel currents shown in C (second trace in green and third trace in orange). (E) Kinetics of MgATP²⁻ binding to P2X2 receptors. A 10-μM ATP plus 5-mM MgCl₂ solution was applied to cells expressing P2X2 receptors for 20 ms, 100 ms, 500 ms, or 5 s, and then switched to EDTA solution. I_R was normalized to the maximal I_R value obtained after 5 s ATP plus MgCl₂ application (n = 3). (F) Kinetics of resurgent current activation. P2X2 receptor channels were activated by 10 μM ATP plus 5 mM MgCl₂ before returning to an external EDTA solution to evoke I_R. The red trace shows currents measured when an open tip is switched between two solutions with different ionic composition. Similar results were obtained from three cells. (G) Enlargement of the part of the current in purple box shown in F. A piezoelectric perfusion system was used for all experiments shown in C–G.

To confirm these results, we performed additional experiments using a piezoelectric-driven solution exchange system that can routinely achieve solution exchange within 4 ms. We designed solutions to contain approximately an EC₂₀ concentration of ATP⁴⁻ (0.4 µM) together with either low (0.6 μM) or high (10 μM) MgATP²⁻ (Fig. 3C). According to the titration curve shown in Fig. 3B, binding of MgATP²⁻ to P2X2 receptors should be minimal at 0.6 μM, but substantial at 10 μM. The results show that the solution with 0.4 μM ATP⁴⁻ and 0.6 μM MgATP²⁻ elicited only steady-state current (without a measurable I_R) that is 13 ± 2% (n = 6) of maximum current activated by saturating concentration of ATP⁴⁻ (Fig. 3C). In contrast, the solution containing 0.4 μM ATP⁴⁻ and 10 μM MgATP²⁻ activates a smaller steady-state current, which is only 6 ± 1% (n = 6) of the maximum current, and when Mg²⁺ and ATP are removed I_R is now observed (Fig. 3C). All of these observations are consistent with the idea that ATP⁴⁻ binds with higher affinity compared with MgATP²⁻, and that the former promotes channel opening much more effectively. The activation kinetics for the steady-state current activated by 0.4 μM ATP⁴⁻ and 0.6 μM MgATP²⁻ are slower than that activated by 0.4 μM ATP⁴⁻ and 10 μM MgATP²⁻ (Fig. 3D), which is consistent with the notion that both ATP⁴⁻ and MgATP²⁻ bind to the receptor. We also investigated the amplitude of I_R as a function of how long MgATP²⁻ is applied to the external solution, and determined that 100 ms is sufficient for I_R to reach its maximum (Fig. 3E), confirming that binding of MgATP²⁻ is rapid. After the MgATP²⁻ solution is removed, I_R develops within a few milliseconds (Fig. 3 F and G), suggesting that Mg²⁺ unbinds very rapidly when it is bound together with ATP.

The Site of Mg²⁺ Action on P2X2 Receptors.

Our results thus far for P2X2 receptor channels are consistent with the idea that Mg²⁺ binds to ATP (in solution and on the receptor), but that MgATP²⁻ binds to the channel with much lower affinity than ATP⁴⁻ (compare Fig. 1B with Fig. 3 A and B) and promotes opening less effectively than ATP⁴⁻ (Fig. 3 A and B); in essence, this is an inhibitory mechanism by which Mg²⁺ would appear to behave as a simple competitive antagonist. To investigate this possibility, we obtained concentration-response relations for ATP activation of P2X2 over a range of Mg²⁺ concentrations and used Schild analysis to determine whether the effects of Mg²⁺ obey a simple competitive mechanism (42). Because P2X receptors might be blocked by Mg²⁺, we determined the range of MgCl₂ concentrations over which block was not significant. We did not observe any obvious voltage dependence under conditions where Mg²⁺ produced robust inhibition of the P2X2 receptor channel (Fig. 4A), arguing against pore block as the primary mechanism of inhibition. In experiments in which the channels are activated with 1 mM ATP to ensure that ATP⁴⁻ is saturating in the presence of all tested Mg²⁺ concentrations, increasing MgCl₂ from 0.1 mM to 12.5 mM has only modest effects on P2X2 currents (Fig. 4B). Next, we constructed concentration–response relations for ATP activation in solutions with four different concentrations of MgCl₂ (0.1, 0.5, 2.5, and 12.5 mM), and observed that although the relations are similar in shape, they shift to higher ATP concentrations as the concentration of Mg²⁺ increases (Fig. 4C). The slope of the Schild plot for the EC₅₀ concentrations obtained in different concentrations of Mg²⁺ is 1.1 ± 0.1 (Fig. 4D), suggesting that Mg²⁺ behaves as a simple competitive inhibitor of P2X2 receptors. It is notable that the log(K_d) obtained from the Schild plot for Mg²⁺ shifting the concentration–response relation for ATP activation of the channel is −4.2, which is indistinguishable from the stability constant of 4.1 for formation of MgATP²⁻ in aqueous solution (where the stability constant is logK_a) (21, 43). Taken together, these results demonstrate that MgATP is not an effective agonist for P2X2 receptor channels, and that ATP⁴⁻ is the principal agonist that activates this channel subtype.

Fig. 4.

The site of Mg²⁺ action on P2X2 receptors. (A) Magnesium inhibition is not voltage dependent. (Left) Macroscopic P2X2 receptor channel currents recorded with 1.5-s voltage ramps from −100 mV to +50 mV. The solution of 10 μM ATP plus 5 mM MgCl₂ contains ∼0.2 μM ATP⁴⁻, 4.9 mM Mg²⁺, and ∼9.8 μM MgATP²⁻. (Right) Current recorded in 10 μM ATP plus 5 mM MgCl₂ was normalized to that recorded in 0.2 µM ATP plus EDTA. (B) Effects of magnesium on P2X2 currents activated by 1 mM ATP. The lowest ATP⁴⁻ concentration in these solutions is 9.6 μM (12.5 mM MgCl₂ plus 1 mM ATP). All currents were recorded from the same cell. The amplitudes of ATP-activated currents for the second through sixth ATP applications were 99 ± 1%, 94 ± 2%, 95 ± 2%, 95 ± 4%, and 88 ± 5% of the first application (n = 4). (C) Dose–response curves for ATP activation of P2X2 receptors obtained in EDTA solutions and in solutions with 2.5 mM MgCl₂. Each point is the average ± SEM of four measurements. The Hill equation was fit to the data, and the EC₅₀ and (n) are 1.4 ± 0.1 μM and (1.3 ± 0.1) for that in EDTA solution and 17.9 ± 0.9 μM and (1.9 ± 0.2) for that in 2.5 mM MgCl₂. (D) Schild plot for Mg antagonism of P2X2 activation with ATP. Dose ratio (DR) was calculated by dividing the EC₅₀ in the presence of MgCl₂ with the EC₅₀ in the absence of MgCl₂. The slope of the Schild plot is 1.1 ± 0.1.

Mg²⁺ Inhibits P2X3 Receptors by Binding to the Extracellular Central Cavity.

The results above for P2X1 and P2X3 receptors show that MgATP²⁻ is an effective agonist for these subtypes (Fig. 2). However, previous studies have shown that Mg²⁺ can also inhibit the P2X3 subtype when the divalent is given sufficient time to equilibrate with the receptor (26). The duration of Mg²⁺ application in our experiments thus far was relatively brief because we applied Mg²⁺ together with ATP; we therefore investigated the actions of Mg²⁺ when it was preequilibrated with the channel for 90 s before application of ATP. For these experiments, we tested the effects of 1 mM MgCl₂ on P2X3 currents activated with 300 μM ATP, a solution predicted to contain ∼40 μM ATP⁴⁻ (a saturating concentration for activating P2X1–4 receptors according to Fig. 1B), ∼260 μM MgATP²⁻, and ∼740 μM Mg²⁺. Consistent with previous results, we observed that Mg²⁺ inhibited P2X3 under these conditions (Fig. 5A). Even more pronounced inhibition by Mg²⁺ was observed for P2X1 receptors, whereas relatively modest inhibition was evident for P2X2 or P2X4 receptors (Fig. S1). Next we constructed concentration–response relations for ATP activation of P2X3 receptors in the continuous presence of 0.1, 0.5, or 2.5 mM MgCl₂ (Fig. 5B), and observed that currents activated by saturating concentrations of ATP decrease as the concentration of Mg²⁺ increases, and that near maximal inhibition is obtained at submillimolar Mg²⁺ concentrations. In contrast to the inhibitory mechanism discussed above for P2X2 receptors, Mg²⁺ inhibition of P2X3 cannot be overcome by increasing the concentration of ATP, suggesting a distinct regulatory mechanism. Interestingly, the concentration–response relationship is detectably shifted toward lower ATP concentrations in the presence of 0.1 mM MgCl₂, suggesting that MgATP²⁻ has either a higher binding affinity or efficacy for activating P2X3 receptors compared with ATP⁴⁻.

Fig. 5.

Mg²⁺ inhibition of P2X3 receptor channels. (A) Effects of preincubation with 1 mM MgCl₂ on P2X3 receptors activated with 300 μM ATP. Macroscopic ATP-activated currents were measured using a holding voltage of −60 mV. To estimate rundown, 300 μM ATP was applied twice in EDTA solution with a 90-s interval between repeated ATP applications. Five seconds after the second ATP pulse, the external solution was switched to one containing 1 mM MgCl₂, and ATP was applied again 90 s later to estimate the effects of MgCl₂. In control experiments in divalent-free solution, the rundown to repeated ATP applications was estimated to be 7 ± 4%. Compared with currents in divalent-free solutions, the currents in solutions with 1 mM MgCl₂ are 66 ± 7% smaller (n = 4). (B) Dose–response curves for ATP activation of P2X3 WT receptors in the presence of EDTA or different concentrations of MgCl₂. Each point is the average ± SEM of 3–4 measurements. The Hill equation was fit to the data, and the EC₅₀ and (n) are 2.0 ± 0.7 μM and (1.1 ± 0.3) for that in EDTA solution, 0.6 ± 0.1 μM and (1.1 ± 0.2) for that in 0.1 mM MgCl₂, 0.7 ± 0.1 μM and (1.1 ± 0.2) for that in 0.5 mM MgCl₂, and 0.6 ± 0.1 μM and (1.2 ± 0.2) for that in 2.5 mM MgCl₂. (C) Sequence alignment of two extracellular segments of P2X receptors. (D) Effects of 1 mM MgCl₂ on P2X3 T87A mutant receptors activated with 300 μM ATP. Experimental protocol is the same as that shown in A. In control experiments in divalent-free solution, the rundown to repeated ATP applications was estimated to be 5 ± 3%. Compared with currents in divalent-free solutions, the currents in solutions with 1 mM MgCl₂ are 11 ± 5% smaller (n = 3). (E) P2X3 T87A mutant channel dose–response curves for ATP activation in the presence of EDTA or different concentrations of MgCl₂. Each point is the average ± SEM of 3–4 measurements. The Hill equation was used to fit the data, and the EC₅₀ and (n) are 1.1 ± 0.2 μM and (1.3 ± 0.2) for that in EDTA solution, 0.7 ± 0.1 μM and (1.1 ± 0.1) for that in 0.5 mM MgCl₂, and 0.9 ± 0.3 and (1.1 ± 0.3) for that in 2.5 mM MgCl₂. (F) Summary of the effect of 1 mM MgCl₂ on P2X3 WT and mutant channels. The experiments were carried out in the same way as that shown in A and E. Currents after 1 mM MgCl₂ application were normalized to the currents activated by 300 μM ATP⁴⁻. (G) Homology models of hP2X3 receptors constructed with SWISS-MODEL using the apo and ATP-bound zfP2X4 crystal structures PDB ID codes (3H9V and 4DW1). Residues E167 and N170 are colored green, and T87 is colored red. The apo P2X3 model (Upper) is viewed from the side, whereas the apo (closed) and ATP-bound (open) models (Lower) are viewed from the intracellular side.

To identify the binding site for Mg²⁺ in P2X3 receptors, we examined the Gd³⁺ sites identified crystallographically in zebrafish P2X4 (zfP2X4) (44). Gd³⁺ is commonly used to localize putative divalent cation binding sites in X-ray structures of proteins, and two types of Gd³⁺ binding sites have been identified in the closed state zfP2X4 crystal structure—a singular site at the threefold axis within the central chamber of the extracellular domain, and three identical peripheral sites within that domain (Fig. 5 C and G) (44). As shown in that structure, E98 from each of the three zfP2X4 subunits form a single binding site in the central chamber, and D184 and N187 from each subunit form the three sites at the periphery of the trimeric receptor (Fig. S2). We mutated E167 and N170 in P2X3, positions equivalent to D184 and N187 in the peripheral Gd³⁺ sites of zfP2X4, and T87 in P2X3, a residue located within the central chamber nearby E98 in zfP2X4, and investigated whether preincubation with 1 mM MgCl₂ would affect the channel currents activated by saturating concentrations of ATP (Fig. 5 C–G). Although Mg²⁺ inhibition was unaffected in the E167A and N170A mutants in the peripheral sites (Fig. 5 C, F, and G), inhibition was greatly attenuated in the case of the T87A mutant in the central chamber (Fig. 5 C–F). We also constructed concentration–responses relations for ATP activation in the presence of either 0.5 or 2.5 mM MgCl₂ (Fig. 5E) for the T87A mutant. No inhibition was observed at 0.5 mM MgCl₂, compared with nearly maximal inhibition at that concentration for WT P2X3. Indeed, at this concentration of MgCl₂, we observed a modest shift of the concentration–response relationship for activation by ATP to lower ATP concentrations, consistent with MgATP²⁻ having a higher affinity or efficacy for the P2X3 receptor. At 2.5 mM, MgCl₂ inhibited the maximum activation by 30 ± 1%, which is significantly reduced compare with the 73 ± 1% inhibition for the WT, suggesting that this mutant displays a diminished affinity for Mg²⁺.

To rule out the possibility that Mg²⁺ also binds to the central chamber in the P2X2 receptor channel or to the peripheral sites to produce inhibition, we looked for residues in P2X2 that might be related to those forming the two types of Gd³⁺ sites in zfP2X4. The Glu in the central cavity and the Asp in the periphery are not conserved in rP2X2, but we identified nearby residues that might play similar roles. We mutated E91 and D315 (rP2X2) in the central cavity, and M179 plus N182 at the peripheral site to Ala, and tested if Mg-ATP solution could activate these mutant channels (Fig. S3). Although the concentration dependence for activation by ATP⁴⁻ was altered to varying degrees in these mutant channels (in particular for D315A mutant), all were weakly activated in the Mg-ATP solution and display I_R when removing Mg²⁺ and ATP (Fig. S3).

Regulatory Constraints of Mg²⁺ Are Disengaged in Heteromeric P2X2/3 Receptor Channels.

The results presented thus far demonstrate that physiological concentrations of Mg²⁺ inhibit the function of both P2X2 and P2X3 receptors. Although the inhibitory mechanisms for P2X2 and P2X3 are distinct, in both cases the binding sites for Mg²⁺ are likely located at subunit interfaces. For P2X2 receptors, inhibition results because MgATP²⁻ binds to the ligand binding site with lower affinity than ATP⁴⁻ and promotes opening less efficiently than ATP⁴⁻. In the X-ray structure of the zfP2X4 receptor, the ATP binding site is formed by a pocket at the interface between adjacent subunits (45). For P2X3 receptors, inhibition appears to result from Mg²⁺ binding to the central chamber formed at the interface between all three subunits; this raises an intriguing possibility that both inhibitory mechanisms might be altered in P2X2/3 heterotrimeric receptors, a specific combination of subunits involved in forming P2X receptor channels in sensory neurons. Heteromeric P2X2/3 receptor channels exhibit slow desensitization similar to P2X2, yet can be activated by αβ-methylene ATP (αβ-MeATP), an ATP analog that activates P2X3 with much higher affinity and efficacy compared with P2X2, together providing the means for distinguishing P2X2/3 heteromeric channels from P2X2 or P2X3 homomeric channels (12, 46). When P2X2 and P2X3 receptors are coexpressed in HEK cells, we observed that 10 μM ATP in divalent-free solution only activates slow-desensitizing currents that are similar to those activated by αβ-MeATP (Fig. 6 A and B), suggesting that under these expression conditions, heteromeric P2X2/3 channels are more abundant than homomeric P2X2 or P2X3 channels. Application of 10 μM ATP with 5 mM MgCl₂ (a solution that contains ∼0.2 μM ATP⁴⁻ and ∼9.8 μM MgATP²⁻) produces activation of the slowly desensitizing current that is much larger than that activated by 0.2 μM ATP⁴⁻ (Fig. 6A). This current could not arise from P2X3 receptors because it desensitizes slowly, and it could not arise from P2X2 receptors because that subtype is not efficiently activated by MgATP²⁻ (Fig. 2). These results demonstrate that heterologously expressed P2X2/3 heteromeric channels can be activated by both ATP⁴⁻ and MgATP²⁻. To investigate whether Mg²⁺ inhibits P2X2/3 channels through the central chamber site identified in P2X3 receptors, we tested whether currents activated by saturating concentrations of ATP are inhibited by 1 mM MgCl₂ when preequilibrated for 90 s (Fig. 6B), the same protocol we used with homomeric P2X3 channels. The results show that P2X2/3 channel currents activated by 100 μM ATP are not detectably inhibited by Mg²⁺, suggesting that central chamber Mg²⁺ site is not robustly engaged in these heteromeric channels.

Fig. 6.

Characterization of P2X2/3 heteromeric channels in HEK cells and DRG neurons. (A) Testing the sensitivity of P2X2/3 heteromeric channels to ATP⁴⁻, MgATP²⁻, and αβ-MeATP. Macroscopic current traces recorded from a HEK cell transfected with P2X2 and P2X3 with a cDNA ratio of 5:1 using a holding voltage of −60 mV. The solution of 10 μM ATP plus 5 mM MgCl₂ contains ∼0.2 μM ATP⁴⁻ and ∼9.8 μM MgATP²⁻. In all experiments, αβ-MeATP was applied in the standard extracellular solution with 2 mM CaCl₂, 0.5 mM MgCl₂, and 10 mM glucose. (Right) Summary of currents in response to each ATP pulse from four cells normalized to the first ATP application. Numbering indicates the sequence of ATP applications from first (1) to last (4). (B) P2X2/3 heteromeric channels in HEK cells are not inhibited after 90 s preincubation with 1 mM MgCl₂ (n = 3). (C) Three types of ATP-activated currents observed in small- and medium-sized DRG neurons. (Left) The fast desensitizing current is likely the P2X3 channel current and was seen in four neurons. (Center) This current trace has two components corresponding to a mixture of P2X3 and P2X2/3 channel populations and was observed in three neurons. (Right) The slow activating and desensitizing current is likely the P2X2/3 channel current and was seen in 15 neurons. The slower activation and deactivation kinetics observed in this population of cells may be the consequence of their larger size, which slows solution exchange compared with the smaller P2X3-expressing neurons and HEK cells. (D) The properties of P2X2/3 heteromeric channels in a medium-sized DRG neuron are similar to P2X2/3 receptors expressed in HEK cells (n = 6).

To examine whether the properties of P2X2/3 heteromeric channels expressed in native dorsal root ganglion (DRG) neurons are similar to those in HEK cells, we initially applied ATP in a physiological solution containing divalent cations and observed detectable currents in 22 of 38 neurons tested. The responsive neurons exhibited three types of currents, resembling those of P2X3 only (Fig. 6C, Left; four cells), P2X3 together with P2X2/3 (Fig. 6C, Center; three cells), or P2X2/3 only (Fig. 6C, Right; 15 cells), consistent with previous studies (16, 47). The neurons that were nonresponsive to ATP in solution containing divalent cations did not respond to ATP in divalent-free solution, suggesting that homomeric P2X2 receptors are not expressed in these cells. Also consistent with previous results, the slowly desensitizing P2X receptors in DRG neurons could be activated by αβ-MeATP, further suggesting that they are P2X2/3 heteromeric channels (Fig. 6D). In this latter class of cells, 0.2 μM ATP⁴⁻ only activated small currents that are <10% of those activated by 10 μM ATP⁴⁻, and a solution containing 0.2 μM ATP⁴⁻ and 9.8 μM MgATP²⁻ activated currents that are comparable to those activated by 10 μM ATP⁴⁻, indicating that P2X2/3 in DRG neurons can be activated by MgATP²⁻ (Fig. 6D).

Discussion

The objective of the present study was to determine which forms of ATP can serve as agonists for different subtypes of P2X receptor channels and to explore whether Mg²⁺ serves any regulatory roles. Collectively, our results demonstrate that ATP in its free ionic form is an active ligand for all subtypes of homomeric P2X receptors, and that the activity of P2X1–P2X4 receptors are inhibited by the presence of physiological concentrations of Mg²⁺ (Table S1). For P2X2 receptors, Mg²⁺ constrains receptor activation because MgATP²⁻ can bind to the agonist binding site, but its affinity and efficacy are lower than ATP⁴⁻ (Figs. 2 and 3). In contrast to what was observed for P2X2 and P2X4, both P2X1 and P2X3 were strongly activated by MgATP²⁻, demonstrating that there are two classes of P2X receptors that differ in the forms of ATP that can serve as agonists (Table S1). Because Ca²⁺ can also chelate ATP⁴⁻, and is abundant in physiological solutions, it is interesting to consider whether it would behave similarly to Mg²⁺. Although quantitative comparison with Mg²⁺ is complicated by the fact that Ca²⁺ permeates P2X receptor channels to different degrees, and thus can change the unitary conductance, we obtained qualitatively comparable results with Ca²⁺ that are consistent with the general idea that divalent-bound ATP can serve as an agonist for P2X3, but not P2X2 receptor channels (Fig. S4).

The structural differences between the two classes of ATP binding sites in P2X receptor channels are likely subtle; however, it is interesting that their sensitivities to MgATP²⁻ are paralleled by other pharmacological distinctions. P2X1 and P2X3 receptor channels can be activated by αβ-MeATP with an EC₅₀ concentration about 1 μM, but P2X2 and P2X4 receptor channels can only partially be activated by 100 μM αβ-MeATP. In addition, P2X1 and P2X3 receptors are inhibited by nanomolar concentrations of trinitrophenyl-ATP, whereas P2X2 and P2X4 are much less sensitive (8, 46, 48). It appears that those receptors that can trigger opening with Mg²⁺ bound to ATP in the agonist binding pockets also have room for the larger agonists and antagonists. These pharmacological distinctions imply that the agonist binding site of P2X4 is most similar to P2X2, raising the possibility that the mechanism of Mg²⁺ inhibition is similar for these two subtypes. We observed resurgent currents for P2X4 that resemble those of P2X2 (Fig. 2), and the apparent affinities of ATP for P2X2 and P2X4 receptors are similar in physiological solutions containing divalent ions, and lower than observed for P2X1 and P2X3 receptors (1, 2). In addition, Mg²⁺ produces only modest inhibition of P2X4 in the presence of high concentrations of ATP (Fig. S2), demonstrating that much of the Mg²⁺ inhibition can be overcome by increasing the concentration of ATP. Thus, the available evidence is most consistent with the idea that both P2X2 and P2X4 are most efficiently activated by ATP⁴⁻. One important implication of this selective activation of P2X2 and P2X4 receptors by the physiologically scarce form of ATP is that these subtypes will require substantially higher concentration of released ATP for efficient signaling to occur when they are expressed as homomeric receptors. Perhaps in these instances, P2X2 and P2X4 receptors primarily serve to detect high concentrations of ATP, as might be expected with repetitive ATP release or with inflammation where cells are dying and releasing cytoplasmic ATP.

In contrast, P2X1 and P2X3 receptor channels can be activated by lower concentrations of ATP in the presence of Mg²⁺ because MgATP²⁻ is an effective agonist for these channels (Fig. 2 A and B). In the case of P2X3 receptors, we observed that the concentration dependence for activation by ATP was shifted to lower ATP concentrations in the presence of low Mg²⁺ concentrations, suggesting that MgATP²⁻ has either higher affinity or efficacy compared with ATP⁴⁻. However, both P2X1 and P2X3 receptors are partially inhibited by Mg²⁺ through a distinct mechanism that likely involves binding of the metal to the central chamber in the extracellular domain (Fig. 5). In the X-ray structure of the apo zfP2X4 receptor, Gd³⁺ binds to the central chamber and is coordinated intimately by E98 (Fig. S2A). Interestingly, in both apo and ATP-bound structures solved in the absence of Gd³⁺, the side chain of E98 swings away from the central axis (Fig. S2 B and C), demonstrating that Gd³⁺ binding induces a reorientation of the E98 side chain. In a P2X3 homology model made from apo zfP2X4, the critical threonine (Thr) we identified (T87) is located within the central chamber and its side chain projects toward the threefold axis of symmetry (Fig. 5G). The hydroxyl moieties in this homology model are positioned within 7 Å of each other, compatible with a role in coordinating Mg²⁺ (Fig. 5G). In contrast, the expansion of the central chamber that occurs in the open-state zfP2X4 structure (and open P2X3 homology model) would be less compatible with stable coordination of Mg²⁺ (Fig. 5G and Fig. S2) (45, 49). Thus, we speculate that the Mg²⁺ inhibits both P2X1 and P2X3 receptors by stabilizing the extracellular domain in a nonconductive conformation. One possibility is that Mg²⁺ inhibits P2X3 receptors by stabilizing a desensitized state of the receptor, as previously proposed (26).

One intriguing common feature of the mechanisms for Mg²⁺ inhibition of homomeric P2X2 and P2X3 receptors is that the sites of Mg²⁺ action are localized to subunit interfaces. In the case of P2X2 receptors, Mg²⁺ binds to ATP in the agonist binding pocket formed by adjacent subunits, and in the case of P2X3, Mg²⁺ likely binds to the central chamber at the threefold axis. P2X2 and P2X3 receptors have been demonstrated to form heteromeric channels in sensory neurons, and in this instance the heteromeric channels contain one P2X2 subunit and two P2X3 subunits (50, 51). Thus, the Mg²⁺ regulatory site in the central chamber would be expected to be weakened because it is missing one of three Thr residues. Similarly, none of the three agonist binding sites would be formed purely by P2X2 subunits; two would be formed at the interface between P2X2 and P2X3 subunits, and one would be constructed at a purely P2X3/P2X3 interface. In particular for heteromeric P2X2/3 receptors, it is not surprising that these heteromeric channels would be robustly activated by MgATP²⁻ given that all three agonist binding sites have contributions from P2X3 subunits.

The key discoveries described here are that (i) the active forms of ATP are not the same for different subtypes of P2X receptors, with P2X2 and likely P2X4 requiring free ATP, whereas P2X1 and P2X3 are also sensitive to MgATP²⁻; (ii) the MgATP²⁻-sensitive P2X1 and P2X3 receptors are partially inhibited by Mg²⁺ binding to the central chamber; and (iii) both of these mechanisms that constrain activation of the receptor can be disengaged when heteromeric receptors form. These findings likely have a range of important implications for understanding the physiological functions of P2X receptor channels in both the central and peripheral nervous systems. In particular, the slowly desensitizing homomeric P2X2 and P2X4 receptor channels would be well-suited for signaling roles when the concentration of extracellular ATP is sufficiently high, whereas the rapidly desensitizing homomeric P2X1 and P2X3 receptors would be more sensitive to lower concentrations of ATP. However, heteromerization of these two classes of P2X receptors channels allows the creation of hybrid channels that are sensitive to MgATP²⁻, slowly desensitizing and relatively insensitive to Mg²⁺ inhibition involving the central chamber. It will be fascinating to explore whether the present studies with heteromeric P2X2/3 receptor channels can be generalized to other complexes to provide a common logic for heteromerization. Our collective results would predict similar hybrid behavior for previously documented heteromeric P2X1/2 and P2X1/4 receptor channels, and it will be exciting to investigate the role of the P2X6 subunits when it forms heteromeric channels with P2X2 and P2X4, all three of which are widely expressed within the central nervous system (15, 26, 52–55).

Materials and Methods

Channel Constructs.

Human P2X1 (hP2X1) in pcDNA3 was generously provided by Richard Evans (University of Leicester, Leicester, United Kingdom). rP2X2 in pcDNA1 was generously provided by David Julius (University of California, San Francisco, San Francisco). hP2X2 and hP2X3 in pCMV6-XL4 were purchased from OriGENE. rP2X4 and rP2X7 in pcDNA3 were generously provided by Florentina Soto (Washington University in St. Louis, St. Louis). rP2X5 was generously provided by Alan North (University of Manchester, Manchester, United Kingdom) and was cloned into pcDNA1. rP2X2 was used in most experiments; however, the key results shown in Figs. 2C and 6A were also repeated with hP2X2 and similar results were obtained. rP2X3 and hP2X3 share 94% sequence identity, and none of the minor differences localize to the ATP binding sites observed in zfP2X4, suggesting that these two species will share the properties described here for hP2X3.

Cell Culture and Transfection.

HEK cells were cultured in DMEM supplemented with 10% (vol/vol) FBS and 10 mg/L gentamicin. All cell culture reagents were obtained from GIBCO. Trypsin-treated HEK293 cells were seeded onto glass coverslips in six-well plates 1 d before transfection and placed in a 37 °C incubator with 95% air and 5% CO₂. Transfections were performed using FuGENE6 Transfection Reagent (Roche Applied Science). P2X receptors were cotransfected with a GFP cDNA construct in pGreen-Lantern (Invitrogen) at a ratio of 2:1. Cells were used for whole-cell recording 18–30 h after transfection.

DRG Preparation.

Isolated DRG neurons were prepared from 7- to 14-d-old rats using procedures similar to what has been previously described (56). After anesthetizing rats with isoflurane and decapitation, segments of the spinal cord (thoracic and lumbar segments) were removed and placed in a cold Ca²⁺, Mg²⁺-free HBSS (Gibco) containing (in mM) 137 NaCl, 5.3 KCl, 0.33 Na₂HPO₄, 0.44 KH₂PO₄, 4.1 NaHCO₃, 5 Hepes, and 5.5 glucose (pH 7.4) with NaOH. The bone surrounding the spinal cord was cut away to expose DRG, which were removed with forceps. Ganglia were cut in half and incubated in a Ca²⁺, Mg²⁺-free HBSS containing 20 U/mL papain (Worthington Biochemical Corporation) and 5 mM cysteine for 20 min at 37 °C. After incubation with papain, ganglia were washed and transferred into a Ca²⁺, Mg²⁺-free HBSS containing 0.1 U/mL collagenase and 0.8 U/mL Dispase (Roche Applied Science), and incubated at 37 °C for 20 min. After two successive incubations in enzymes, ganglia were transferred into Leibovitz’s L-15 medium (Invitrogen) supplemented with 10% FCS, 5 mM Hepes, and 50 ng/mL NGF (Invitrogen). Individual cells were dispersed by mechanical trituration using a fire-polished Pasteur pipette and plated on poly-d-lysine/laminin-coated glass coverslips (BD Biosciences). Cells were incubated in a 37 °C incubator with 95% air and 5% CO₂ for 3 h, after which cells were stored at room temperature and used for patch-clamping within 24 h.

Electrophysiology.

Standard whole-cell patch-clamp recording was used to record most P2X receptor channel currents from transiently transfected HEK293 cells and isolated DRG neurons. Perforated patch-clamp recording was used to record P2X4 receptor channel currents to minimize the rundown in response to repeated ATP applications (57). Membrane currents were recorded under voltage clamp using an Axopatch 200B patch-clamp amplifier (Axon Instruments, Inc.) and digitized online using a Digidata 1322A interface board and pCLAMP 9.2 software (Axon Instruments, Inc.). Currents were filtered at 2 kHz using eight-pole Bessel filters, and digitized at 5 or 10 kHz. The pipette solution for standard whole-cell recording on HEK cells contained (in mM) 140 NaCl, 10 EGTA, and 10 Hepes, adjusted to pH 7.0 with NaOH. The pipette solution for standard whole-cell recording on DRG neurons contained (in mM) 65 CsCl, 65 CsF, 10 NaCl, 2 MgCl₂, 10 EGTA, 10 Hepes, 14 Tris⋅creatine PO₄, 4 Mg-ATP, and 0.3 Na-GTP (pH 7.2) with CsOH. The pipette solution for perforated patch-clamp recording on HEK cells contained (in mM) 65 K₂SO₄, 10 NaCl, 1 CaCl₂, 1 MgCl₂, 10 Hepes (pH 7.2) with KOH. The divalent-free extracellular solution contained (in mM) 140 NaCl, 10 EDTA, and 10 Hepes, adjusted to pH 7.3 with NaOH. Assuming that divalent ion contaminations from reagents and water is less than 1 μM, any possible free divalent cations in the 10 mM EDTA solution would be less than 0.5 nM. The extracellular solutions with different concentrations of Mg²⁺ contained (in mM) 140 NaCl, 10 Hepes, and desired concentration of MgCl₂, adjusted to pH 7.3. K₂ATP (referred to as ATP) was used in all of the experiments. The stability constants of ATP binding to cations present in our solutions are Mg²⁺ 10^4.10, Ca²⁺ 10^3.76, Na⁺ 10^0.83, and K⁺ 10^1.17 (National Institute of Standards and Technology Critically Selected Stability Constants of Metal Complexes). Max Chelator (http://maxchelator.stanford.edu/) was used to calculate concentrations of Mg²⁺, MgATP²⁻, and ATP⁴⁻ in solutions with different concentrations of MgCl₂ and ATP. Recording solutions were adjusted to the correct pH after adding MgCl₂ and ATP. Solution exchange was achieved using either the Rapid Solution Changer RSC-200 (BioLogic), which has the capacity of switching between nine solutions, or a Piezo-driven system (AutoMate Scientific), which can change two solutions within 1 to 4 ms, depending on the size of the cell.

Data Analysis.

To generate the dose–response relationship, a reference concentration of ATP was given every time before a test concentration was applied, as previously described (31). The Hill equation was fit to the data according to

where I is the normalized current at a given concentration of ATP, I_max is the maximum normalized current, EC₅₀ is the concentration of ATP ([ATP]) producing half-maximal currents, and n is the Hill coefficient.