Preferential use of unobstructed lateral portals as the access route to the pore of human ATP-gated ion channels (P2X receptors)

Abstract

P2X receptors are trimeric cation channels with widespread roles in health and disease. The recent crystal structure of a P2X4 receptor provides a 3D view of their topology and architecture. A key unresolved issue is how ions gain access to the pore, because the structure reveals two different pathways within the extracellular domain. One of these is the central pathway spanning the entire length of the extracellular domain and covering a distance of ≈70 Å. The second consists of three lateral portals, adjacent to the membrane and connected to the transmembrane pore by short tunnels. Here, we demonstrate the preferential use of the lateral portals. Owing to their favorable diameters and equivalent spacing, the lateral portals split the task of ion supply threefold and minimize an ion's diffusive path before it succumbs to transmembrane electrochemical gradients.

P2X receptors are a family of ligand-gated ion channels that open when bound to extracellular ATP, providing a regulated path for depolarizing cation current across cell surface membranes. In mammals, the resulting entry of Ca²⁺ and Na⁺ triggers muscle contraction, release of neurotransmitters, hormones, and inflammatory mediators, as well as activation of downstream signaling cascades (1, 2). In combination with a ubiquitous distribution, the widespread actions of ATP suggest that P2X receptors play essential roles in the body. Indeed, advances with mouse genetics and pharmacology show that P2X receptors are involved in a variety of important physiological responses and disease states (3), making them attractive targets for drug discovery. A key goal is to understand how P2X receptors work at a molecular level.

The P2X receptor family is made of seven gene products, designated P2X1–P2X7, that can form homomeric and/or heteromeric assemblies (4). One family member, the PX4 receptor, is particularly prevalent in the central nervous system (5). This receptor is involved in microglia–neuron interactions, and mounting evidence suggests it plays a critical role in neuropathic pain, a persistent and prevalent form of pain whose molecular and cellular neurobiology is poorly understood and for which there is no effective treatment (6).

Each P2X4 receptor is made of three parts: an extracellular domain containing the ATP binding site and the extracellular vestibule of the ion channel, a transmembrane pore containing the channel gate and selectivity filter, and an intracellular vestibule in contact with the cytoplasm (7). The possibility that the permeation pathway extends outward from the extracellular vestibule was not considered until recently, when the newly solved crystal structure of a closed state of the zebrafish P2X4.1 receptor (zfP2X4.1R) revealed, but failed to discriminate between, two possible paths in the extracellular domain for ion passage to the pore (8). The first is the central pathway that spans the entire length of the extracellular domain (Fig. 1A). The central pathway is lined with negative charge and appears well suited to regulate cationic current (8, 9). The second pathway is cast from three lateral portals formed at the interfaces of adjoining subunits just above the outer leaflet of the membrane (Fig. 1B). The lateral portals are less charged than the central pathway, and less obstructed. Dual pathways are also observed in acid sensing channels (10), but in neither case is it known which pathway is functionally used as a cation pore.

Fig. 1.

Alternative ion access routes to the pore of the hP2X4R. Illustrations of the hP2X4R viewed parallel to the membrane plane. The cytoplasmic N- and C-termini are missing because these domains are absent in the crystal structure of the truncated zfP2X4.1 (8). The dotted lines show the limits of the membrane. (A) A sagittal section (4 Å) of the central pathway shows the pore-lining inner surface (gray mesh grid), highlighting the vestibules and constrictions defined by Kawate et al. (8). The orange spheres are spaced in 1-Å steps along the long axis of the pore and suggest a likely path for cations to the cytoplasm if they entered through the central pathway (calculated by using PoreWalker 1.0). (B) A space-filling model of the lateral portals, extracellular vestibule, and transmembrane pore (orange) as calculated by using CAVER. Each portal (P1 to P3) has a wide (8 Å) and direct opening to the extracellular vestibule of the transmembrane pore.

Early work showed that replacing Ile³²⁸ of the rat P2X2 receptor (rP2X2R) with cysteine results in a mutant receptor whose ATP-gated currents are sensitive to water-soluble, thiol-reactive reagents (11, 12), and a recent report used the P2X2R-I328C mutant to conclude that the lateral portals form an access route to the transmembrane pore (13). However, Ile³²⁸ sits at the top of the second transmembrane domain of the rP2X2R and is accessible to thiol-reactive reagents that may travel through either the central pathway or the lateral portals and, thus, this mutant does not firmly establish the primary route of ion flow. In the present report, we used whole-cell as well as single-channel recordings, and applied cationic and anionic methanethiosulfonates (MTS) and Cd²⁺ in the presence and absence of ATP, to measure effects on gating and conduction of a wide range of cysteine-substituted mutants of the human P2X4 receptor (hP2X4R). Taken together, our experiments provide convincing evidence that the lateral portals form the primary access route to the hP2X4R and show that some residues in the lateral portals are subject to allosteric modulation.

Results

We used cysteine-scanning mutagenesis (14) and measurements of fractional Ca²⁺ current (Pf%) (15, 16) to investigate the contributions of the two pathways to ionic current of the hP2X4R. We studied the hP2X4R because of its high sequence identity to the zfP2X4.1 form (Fig. S1) and, because, unlike the zfP2X4.1R (17), it readily expresses in HEK293 cells and gives reproducible responses to repeated applications of micromolar concentrations of ATP (EC₅₀ = 4.4 ± 0.9 μM; n = 6; Fig. 2A). Using a homology model of the hP2X4R closed state, we identified residues that line the central pathway and the lateral portals in their narrowest regions and, thus, might be expected to interact with ions in the two putative permeation pathways. We then mutated the residues to cysteines, expressed the mutants in HEK293 cells, and measured the accessibility of the engineered thiol side chains to water-soluble methanethiosulfonates (MTS) by means of electrophysiology (SI Methods). In control experiments, we found no effect of 1 mM [2-(trimethylammonium)ethyl] MTS⁺ (MTSET⁺) on the ATP-gated (100 μM) current of the wild-type (wt) hP2X4R (Fig. 2 B and C).

Fig. 2.

MTSET⁺ has no effect on wt hP2X4R and upper vestibule mutants. (A) Concentration–response curve for the hP2X4R generated by using perforated-patch electrophysiology (SI Methods). In each cell, 30 μM ATP was applied first, and then the effect of subsequent doses were normalized to the current caused by 30 μM ATP (I/I_30μM, y axis). (B) Sequential overlaid traces of ATP-gated current, separated by 180s, are shown before (black) and during (red) an application of MTSET⁺. MTSET⁺ (1 mM; red bar) has no effect on membrane current of the wt hP2X4R when coapplied with ATP (100 μM; black bar). (C) MTSET⁺ (1 mM) has no effect on the wt hP2X4R when applied in the absence of ATP. C Upper shows the ATP-gated currents caused by repeated applications (each 1-s application is separated by 180s) before and after a 60 s application of 1 mM MTSET⁺ (pink shaded rectangle). A subsequent application of DTT (1 mM, 60 s; blue shaded rectangle) also had no effect on the ATP-gated currents. C Lower shows the averaged data of individual measurements from five or more cells. Peak amplitudes of ATP-gated currents were normalized to the peak current of the ATP-gated current that immediately preceded the application of MTSET⁺ (trial_-1). (D) Coapplied MTSET⁺ has no effect on cysteine-substituted mutants of the central pathway. The illustration shows one subunit of the trimeric hP2X4R. The positions of six of the nine substituted cysteines in the central pathway are shown as spheres, and the traces at Right show the lack of effect of MTSET⁺ (red bars) on currents evoked by ATP (black bars). (E) MTSET⁺ (1 mM) and DTT (1 mM) have no effect on mutants of the narrow constriction of the central pathway when applied in the absence of ATP. Each data point is the average of individual determinations from seven or more cells measured using perforated-patch electrophysiology.

To demonstrate feasibility, we used the published reports of accessibility of cysteine-substituted rP2X2Rs (11, 12, 18, 19) to suggest a residue in the transmembrane pore of hP2X4R (Ser³⁴¹, equivalent to rP2X2R-T336C) that should be accessible to MTSET⁺ (Fig. S2A). As expected, we found that 1 mM MTSET⁺ irreversible reduced (99 ± 1%, n = 4) the agonist-gated current of the hP2X4R-S341C when coapplied with ATP (Fig. S2 B and E), and that the rate of modification (K_on = 4,387 ± 1,472 M⁻¹⋅s⁻¹) was comparable to that measured when MTSET⁺ and ATP are coapplied to cysteine-substituted rP2X2R (18). Because the large (6 × 6 × 10 Å) MTSET⁺ must first traverse either the central pathway or the lateral portals to reach S341C, this experiment serves as a positive control to demonstrate that the relevant permeation pathway is fully permeable to the thiol-modifying reagent.

Upper Vestibule.

We examined the central pathway by placing cysteines at sites in the upper vestibule where the diameter of the channel is narrow (Fig. 2D). The first site represents the outermost entrance to the central pathway; it is obstructed in the closed state model of hP2X4R, in part because the side chain of Leu³⁰³ caps the opening. The entrance must open for current to flow, and we hypothesized that modification by MTSET⁺ of L303C would plug the open channel and reduce ATP-gated current if the central pathway formed a conducting pore. However, we saw no effect of coapplying ATP and MTSET⁺ on L303C currents (Fig. 2D), and no effect of MTSET⁺ on the neighboring D302C and A304C residues (<5% change; n = 5). We then moved inward, and placed cysteines near the top (A297C; pore diameter ≈ 10 Å) and bottom (P92C and N93C; pore diameter ≈ 5 Å; Fig. S3) of the upper vestibule, and again saw no effect (<5% change; n = 5) of MTSET⁺ (for examples, see Fig. 2D). The bulkier (≈8 × 16 × 16 Å) Texas Red-2-sulfonamidoethyl MTS (MTSTR) was also without effect at all of the aforementioned residues, despite the fact that it irreversibly blocks the current of the S341C transmembrane mutant (Fig. S2 C and E).

Next, we probed the narrow constriction (≈2.3 Å) between the upper and central vestibules of the zfP2X4.1R formed in part by three residues (Gln⁹⁷, Glu⁹⁸, Asp⁹⁹), including Glu⁹⁸ that coordinates Gd³⁺ just above the entrance to the central vestibule (8). We placed cysteines at the equivalent sites of the hP2 X 4R (Gln⁹⁴, Glu⁹⁵, Glu⁹⁶; Fig. S3) and failed to measure an effect of MTSET⁺ (Fig. 2 D and E) or MTSTR. Although it is possible that MTSET⁺ and MTSTR modify residues in the constriction without changing current amplitude (14), the comparative dimensions of the MTS head groups and the diameter of the pathway (8) render this highly unlikely. Another possibility is that residues in the constriction are inaccessible to MTSET⁺ and MTSTR, either because the central pathway does not form a part of the ion channel pore, or because large MTS reagents are impermeable. Although the latter option seems unlikely because of the demonstrated accessibility of the downstream S341C mutant, we nevertheless tested this possibility by using Cd²⁺, a small metal element that binds to thiol side chains exposed to water, as a permeable probe (19). We expected to measure a change in the amplitude of the ATP-gated current if the coordinated ion occludes the channel or preferentially locks it in the open or closed state, but we found that Cd²⁺ (20 μM) failed to modify (<5% change; n = 6–7) the ATP-evoked currents of any of these three hP2X4R mutants. When considered in concert with the A93C/A297C/L303C experiments, our results argue against the suggestion that the central pathway forms a major conduit for ion conduction in the hP2X4R (8, 9).

Lateral Portals.

We then tested the alternative hypothesis that cations enter the transmembrane pore through the lateral portals. We started by studying two acidic residues in the lateral portals of the hP2X4R, Glu⁵⁶ and Asp⁵⁸ (Fig. 3A), that are common to most (P2X2-P2X7) but not all (P2X1) human P2X receptors, and that are thought to concentrate cations in the extracellular vestibule of the zfP2X4.1R pore (8). Fig. 3 shows the effect of coapplying MTSET⁺ and ATP. The results were qualitatively similar in both mutants: MTSET⁺ caused irreversible reductions in the amplitudes of the ATP-gated currents. In cells expressing the E56C mutant, MTSET⁺ reduced current by 58 ± 4% (n = 9), with an apparent modification rate (K_on) of 7,057 ± 421 M⁻¹⋅s⁻¹ (Fig. 3B). By comparison, MTSET⁺ caused a slower (K_on = 3,545 ± 577 M⁻¹⋅s⁻¹) but more complete block (95 ± 1%, n = 8) of the D58C mutant (Fig. 3C). The faster block of E56C may reflect the fact that the substituted cysteine resides on the outer surface of the portal where it is more accessible to MTSET⁺ than D58C (Fig. S3).

Fig. 3.

MTSET⁺ modifies residues of the lateral portals. (A) Illustrations of hP2X4R showing the position of Glu⁵⁶ (blue spheres) and Asp⁵⁸ (red spheres) in the lateral portals. (B) MTSET⁺ (1 mM; red bars) inhibited current through hP2X4R-E56C when coapplied with ATP (100 μM; black bars). (C) MTSET⁺ (1 mM) caused a near complete inhibition of current through hP2X4R-D58C when coapplied with ATP (100 μM). (D and E) MTSET⁺ (1 mM for 60 s; pink shade rectangles) also inhibited current when applied in the absence of ATP. The inhibitions by MTSET⁺ were reversed by 1 mM DTT (blue shaded rectangles). (F) Averaged data for experiments like those of D and E. Each data point is the average of individual measurements from 11 to 16 cells obtained by using perforated-patch electrophysiology.

Although Glu⁵⁶ and Asp⁵⁸ are positioned in the lateral portals/extracellular vestibule, they also lie at the lowest point of the central pathway. Therefore, it remained possible that the MTSET⁺-sensitive currents move through the central pathway before reacting with the engineered thiolates. We addressed this possibility.

First, we measured MTSET⁺ reactivity in the absence of ATP. The central pathway contains several narrow constrictions that must open to allow MTSET⁺ to access Glu⁵⁶ and Asp⁵⁸ (8). In contrast, there is no barrier to current flow in the portals that would prevent modification of the closed states of E56C and D58C mutants. We established baseline responses to ATP (100 μM), and then applied MTSET⁺ for 2–60 s in the absence of ATP. In all cases, MTSET⁺ caused significant and persistent inhibitions of currents elicited by subsequent applications of ATP (Fig. 3 D and E). In some cases, the ATP-gated currents showed a partial recovery in the first 3 min after washout of MTSET⁺. A similar effect is reported for MTSET⁺ inhibition of cysteine-substituted rat P2X4Rs, which is thought to reflect recycling of unmodified receptors to the plasma membrane (20) (see also ref. 7). However, in no case did the amplitude of the ATP currents return to their control values, and a significant inhibition remained even after a 15-min wash (Fig. 3 D–F). Further, the persistent inhibition of ATP-gated current by MTSET⁺ was immediately and fully reversed by the reducing agent, DTT (1 mM; 60 s) (Fig. 3 D–F). The same protocol had no effect on the wt hP2X4R (Fig. 2C), or on mutant receptors containing cysteine substitutions in the narrowest constriction of the central pathway (Fig. 2E), suggesting that the majority of the plasmalemmal E56C and D58C receptors were covalently modified by MTSET⁺.

Second, we measured the ability of an anionic MTS reagent to modify E56C and D58C. The profoundly acidic surface of the central vestibule of the zfP2X4.1R provides a long-range negative electrostatic environment that favors the passage of cations over anions (8, 13). The hP2X4R shows a modest Cl⁻ permeability (21), and we find that the anionic thiol-reactive drug, (2-sulfonatoethyl) MTS⁻ (MTSES⁻), causes a fast (8,348 ± 867 M^-1⋅s⁻¹) and near complete block of the S341C mutant (Fig. S2 D and E), as previously shown for the homologous residue (Thr³³⁶) of the rP2X2R (11). Like MTSET⁺, MTSES⁻ must pass through either the central pathway or the lateral portals before reaching S341C. Therefore, we coapplied MTSES⁻ (1 mM) and ATP (100 μM) to measure accessibility and found that it modified both E56C and D58C mutant receptors. MTSES⁻ caused a small (34%) but significant (P = 0.0003) potentiation of current through hP2X4R-E56C and a near complete block of ATP-gated current through hP2X4-D58C (Fig. S4 A–C). In cysteine substituted accessibility experiments, potentiation and block are both equally insightful and interpreted to indicate side chain accessibility (14). Surprisingly, the apparent modification rate at position D58C was faster for MTSES⁻ (K_on = 10,681 ± 1,123 M^-1⋅s⁻¹, n = 7) than for MTSET⁺ (≈3,500 M⁻¹⋅s⁻¹). We do not know why the anionic drug reacts quicker. Nonetheless, the fast reaction rate suggests that MTSES⁻ has an unimpeded route of access to the thiol side chain of D58C, a situation that favors the short and less charged environment of the lateral portals over the longer and highly electronegative central pathway.

Finally, we measured the effect of MTSET⁺ and MTSES⁻ on T57C because it sits between the two acidic amino acids of the lateral portals. Coapplication of MTSET⁺ caused a moderate (27 ± 5%; n = 5) but significant (P = 0.0004) inhibition of the ATP-gated current of hP2X4R-T57C (Fig. S4D). In contrast, MTSES⁻ had no affect. However, an initial application of MTSES⁻ prevented the inhibition of ATP-gated current by a subsequent application of MTSET⁺ (Fig. S4E), showing that the anionic reagent modifies the cysteine at position Tyr⁵⁷.

Studies of Single Channel Current.

The block of the ATP-gated currents of hP2X4R-E56C and hP2X4-D58C by MTSET⁺ reflects either a change in conductance or a change in gating (14). Although both effects signal accessibility, a change in conductance is the more direct demonstration that the modified side-chain lines the permeation pathway (14, 22, 23). To measure a change in conductance, we recorded ATP-gated single channel currents from the membrane patches of HEK293 cells voltage-clamped at −120 mV. The hP2X4R shows a fast and progressive loss of single channel activity in excised patches that severely limits before and after comparisons of drug effects (24). Rundown also occurs in cell-attached patches, but at a slower rate that allows repeated measurements of single channel current amplitude (Fig. S5A; ref. 24). Thus, we used the cell-attached mode to compare channel amplitudes from populations of cells bathed in a physiological salt solution with or without 1 mM MTSET⁺ before seal formation (SI Methods).

We recorded bursts of single-channel current from wt (Fig. S5B) and mutant hP2X4Rs (Fig. 4A) in the presence of ATP (0.3–1 μM) that were never seen in the absence of agonist or in untransfected cells. We found no difference in the conductances of the wt hP2X4R and the E56C single channel currents that measured 17.8 ± 0.8 pS (n = 7) and 16.2 ± 0.4 pS (n = 23), respectively. In contrast, the conductance of the D58C mutant (21.7 ± 0.6, n = 26) was significantly larger than that of the other two (P < 0.01), providing the first clue that D58C lines the permeation pathway. We then compared the average current amplitudes of the two cysteine-substituted receptors expressed in HEK293 cells bathed for 3 min in 1 mM MTSET⁺ and saw significant (P < 0.0001) reductions (≈15% for E56C and ≈40% for D58C) in the size of the single channel currents (Fig. 4B). The fact that cysteine modification decreases conductance provides robust support for the model that E56C and D58C line the permeation pathway and, thus, validates the hypothesis that the lateral portals are a primary entrance to the transmembrane pore.

Fig. 4.

Thiolation alters in single channel current. (A) Representative single-channel currents of unmodified E56C and D58C receptors are shown at two different time scales (Left). Center shows the reduction of current through E56C and D58C in cells bathed in 1 mM MTSET⁺ before patch formation. Right shows the increase in current amplitude in a cell expressing the E56C mutant and exposed to 1 mM MTSES⁻. The dashed red lines indicate the average single-channel current amplitude of the relevant unmodified mutant receptors. (B) The histogram presents the average of all of the data shown in Fig. S5C. Asterisks denote significant differences (P < 0.01) in current amplitudes for the E56C and D58C mutants measured before and after thiolation.

The ATP-gated whole-cell current of the hP2X4R-E56C is potentiated by coapplication of MTSES⁻ (Fig. S4A). To determine whether this potentiation reflects a change in conduction or gating, we measured single channel current after anionic thiolation and found a significant (P < 0.0001) increase (≈24%) in the size of the single channel current compared with control (Fig. 4 A and B). The observation that MTSET⁺ and MTSES⁻ have opposite effects suggests a role for electrostatics in the effect of thiolation on ATP-gated current at this position in the lateral portals.

No Effect of Removing the Fixed Negative Charge of Glu⁵⁶, Asp⁵⁸, and Glu⁹⁵ on Ca²⁺ Current.

In light of the apparent role of electrostatics described above, we next tested the hypothesis put forth by Kawate et al. that the fixed negative charge of Glu⁵⁶ and Asp⁵⁸ creates a long-range electrostatic potential that concentrates cations and repels anions near the entrance to the transmembrane pore (8). Ca²⁺ carries twice the charge of Na⁺ and is more sensitive to local changes in electrostatic fields. Thus, we measured the fraction of the total ATP-gated current carried by Ca²⁺, called the Pf%, by using patch clamp photometry as described (21, 25). We constructed single (E56Q, D58N) and double (E56Q/D58N) hP2X4R mutants with reduced charge and found that the Pf% was unchanged (Fig. S6). We therefore conclude that neither Glu⁵⁶ nor Asp⁵⁸ make a significant contribution to cation or anion selection in the hP2X4R, despite the fact that they lie in the permeation pathway. This result is consistent with the fact that the negative charges at these positions are not absolutely conserved among cationic P2X receptors. In a similar vein, we measured no effect of removing the carboxyl side chain of Glu⁹⁵ of the central pathway on the Pf% of the hP2X4R-E95C mutant (Fig. S6 D and E), despite the fact that this site forms a cation-binding site in the zfP2X4.1R (8).

Our finding that removing the fixed negative charge of Glu⁵⁶ has no effect on Pf% is interesting in light of the opposite effects of MTSET⁺ and MTSES⁻ on the single channel conductance of the E56C mutant (Fig. 4). Glu⁵⁶ is on the outer rim of the lateral portal where the negative charge of the native carboxylate may be poorly positioned to alter current (Fig. S3). In contrast, the addition of the bulky (≈6 Å) anionic headgroup of MTSES⁻ to the cysteine mutant may place fixed negative charge closer to the path of ion flow and, thus, increase single channel current by attracting cations to the lateral portals.

Central Vestibule.

We then extended our cysteine scan to determine the accessibility of four sites (Ser⁵⁹, Val⁶¹, Ser⁶², Asn⁹⁷) in the central vestibule to a range of MTS reagents (Fig. S3). Our rationale for doing this experiment was as follows: The central vestibule sits just above the lateral portals, and although most cations that exit the portals probably move passively down the transmembrane electrochemical gradient and into the cell, some may be attracted to the strong electronegative pull of acidic amino acids in the central vestibule (8, 9). In keeping with this hypothesis, we found that MTSET⁺ significantly potentiated the ATP-gated currents of S59C, S62C, and N97C mutants (Fig. S7). In contrast, MTSES⁻ potentiated the current of the receptors carrying the S59C mutation that sits just above the lateral portals, but neither altered ATP-gated current nor masked the effects of MTSET⁺ on S62C and N97C. Neither MTSET⁺ nor MTSES⁻ affected V61C. Thus, our results support the hypothesis of Kawate et al. that the negative electrostatic environment of the central vestibule attracts cations and repels cations (8). The head groups of MTSET⁺ and MTSES⁻ are approximately the same size (14), which suggests that the lack of effect of MTSES⁻ on S62C and N97C cannot be explained by bulk exclusion. Rather, the data imply that the strong electronegative environment of the central cavity impedes the flow of anions through the middle third of the central pathway, despite the fact that MTSES⁻ readily modifies downstream engineered cysteines in the lateral portals (E56C, D58C) and transmembrane pore (S341C).

Effect of Ivermectin on Thiolation.

Ivermectin (IVM) increases single channel conductance and lengthens mean open time of P2X4 receptors (7, 24, 26), at least in part through intercalation with the transmembrane helices (27). To determine whether the lateral portals lie in the allosteric pathway, we assessed the effects of MTSET⁺ on the E56C and D58C mutants before and after 5-min preincubations in 1–10 μM IVM (Fig. 5A). IVM changed the effect of MTSET⁺ in two ways. First, the rates of reaction were significantly slower (K_on values after 10 μM IVM were 1,780 ± 387 (n = 5) and 1,084 ± 200 (n = 8) for E56C and D58C, respectively). Second, the inhibitions by MTSET⁺ were significantly smaller (Fig. 5B). These findings are supported by experiments using the smaller Cd²⁺ as modifier. We found that Cd²⁺ caused a pronounced potentiation and inhibition of the ATP-gated currents of the E56C and D58C mutants, respectively, and that both effects of Cd²⁺ were absent after incubation with IVM (Fig. 5 C–E). We interpret the data with MTSET⁺ and Cd²⁺ to indicate that IVM changed the shape of the portals through an allosteric mechanism and, hence, the nature of the thiol modification.

Fig. 5.

IVM reduces MTSET⁺ and Cd²⁺ modification of the lateral portals. (A) The figure shows the ATP-gated current before (black traces) and during (red traces) coapplication of 1 mM MTSET⁺ in two cells preincubated in 10 μM IVM for 5 min, and then maintained in IVM for the duration of the experiment (≈10 min). The traces are normalized to their peak current amplitudes. MTSET⁺ still blocks the ATP-gated current in cells expressing E56C and D58C and soaked in IVM, but the effect is smaller and slower. (B) Summary data for the block of MTSET⁺ in the absence (black) and presence (red striped) of IVM. **, significantly different from wt hP2X4; *, significantly different from the wt and the −IVM condition (P < 0.001). (C) Normalized traces show control (black; 100 μM ATP alone) and test traces (green; 100 μM ATP followed by 20 μM Cd²⁺ applied for the duration indicated by the green bar) for wt hP2X4 and the indicated cysteine mutants. Arrows point to profound potentiation and block of E56C and D58C mutants by Cd²⁺. (D) As in C, but for cells that had been bathed in 1 μM IVM for 2–5 min before the start of the recording and then also for the duration of the recording. (E) Summary data for experiments such as those in C and D for wt hP2X4R and E56C and D58C mutants. Relative to the wt hP2X4R, Cd²⁺ applications to E56C and D568C mutants significantly potentiated and reduced ATP-evoked responses; further these effects were abolished in the presence of IVM. The data are shown as average ± SEM (n = 5–16). **, significantly different from wt hP2X4R; *, significantly different from the −IVM condition.

Discussion

The main findings of this study are that cations gain access to the ion selective transmembrane pore via the lateral portals of P2X receptors and that ion flow through the portals is allosterically regulated.

The coordination of Gd³⁺ by Glu⁹⁸ in the closed state structure of the zfP2X4.1 was interpreted to suggest that the central pathway might be a conduit for cation flow (8). However, Glu⁹⁸ is not well conserved, and the surmountable nature of the block of ATP-evoked currents by Gd³⁺ is inconsistent with pore block (8). In relation to these facts, we found no evidence using substituted cysteine mutagenesis or measurements of Pf% to suggest that the central pathway is a conductive entrance to the pore.

However, our experiments provide strong evidence to suggest that ions preferentially flow to the pore via the symmetrically placed lateral portals. Thus, coapplication of ATP and either MTSET⁺ or MTSES⁻ causes a near complete block of current through the D58C mutant (Fig. 3C and Fig. S4B), showing that, at least under these circumstances, all or almost all of the current that enters the transmembrane pore passes first through the lateral portals. Further, MTSES⁻ is excluded from parts of the central pathway (Fig. S7) but is still capable of modifying S341C in the transmembrane pore (Fig. S2D). In some ways the use of lateral portals makes intuitive sense, because some ions that traverse the central pathway might exit the pore through the three wide lateral portals rather than proceed into the single transmembrane pore. Further, the portals are short and unobstructed. The use of lateral portals therefore appears to be an efficient mechanism to facilitate unimpeded ion flow as it splits the task threefold between portals, exploits the favorable electrostatic environment near the plasma membrane outer leaflet, and reduces the diffusive path for cations.

Our finding that MTSET⁺ reduces both whole-cell and single-channel current of the D58C mutant suggests this modification exerts a “bulk” effect in the portals that impedes the flow of ions. Cationic thiolation of D58C results in a significant (40%) reduction in single channel current that is smaller than the block of the whole-cell current (95%). The inability of thiolation to completely block single channel current by adding bulk to portal-lining residues is not surprising for two reasons. First, the lateral portals are unusually wide by comparison with the pore-lining transmembrane domains that are typical targets of thiolation studies of ion permeation (8). Second, each portal presents a single D58C side chain for modification, unlike transmembrane pores that show multiple axes of symmetry. Thus, the addition of the headgroup of MTSET⁺ to a single cysteinyl side chain in a wide lateral portal may not fully plug the opening. If so, then the almost complete reduction in whole-cell current by MTSET⁺ must reflect changes in both conduction and gating. A change in gating after thiolation is more often the rule than the exception (14). Nevertheless, our finding that MTSET⁺ (and MTSES⁻) change the amplitudes of the single channel currents of E56C and D58C provides unusually strong support for the hypothesis that the lateral portals form a primary entrance to the transmembrane ion conduction pathway.

A dual effect on conduction and gating may also explain the smaller effect of MTSET⁺ on homologous residues of the rP2X2R. The effect of MTSET⁺ on the ATP-gated current of rP2X2R-D57C and rP2X2R-E59C is smaller (28) than the effect on the homologous residues (E56C, D58C) of hP2X4R, which could be explained if some of the effect of thiolation results from a change in gating, as discussed above. In fact, the gating behavior of rP2X2R and hP2X4R differ in sensitivity to agonists, antagonists, and allosteric modulators like pH and IVM (29). Perhaps then we should not be surprised that the effect of thiolation also differs between P2X receptors. Nevertheless, the route of access is the main consideration, and our data clearly show that MTS reagents alter ATP-gated current of the hP2X4R by entering the lateral portals.

Some of the accessible residues (Ser⁵⁹, Ser⁶², Asn⁹⁷) that are potentiated by thiolation map to a region between the proposed ATP binding site and the outer end of TM1, composed of “connecting rods” that link ATP binding to movement of the transmembrane domains (30). Residues in this part of the protein are thought to be involved in gating (30, 31). Our results suggest that the connecting rods form a surface of the portals and vestibules that is accessible to application of water-soluble thiol-reactive reagents that travel through the lateral portals even when the transmembrane channel gate is closed. Further, our data with IVM, a known allosteric regulator of P2X4 receptors, provides good evidence that altered gating can affect the nature of accessibility in the portals. This observation, along with the connecting rods model, implies that ion flow through the lateral portals and receptor gating may be linked. These findings raise the possibility that ion access to the hP2X4R portals could be targeted with designer drugs, which may find applications in therapy of P2X4-associated disorders. In future work it will be useful to explore this possibility when a 3D structure of the open channel bound to ATP becomes available.

In conclusion, our experiments show a pattern of hits in cysteine scanning accessibility experiments (Fig. S8) that strongly suggest that ions access the pore of a P2X receptor by making preferential use of three radially symmetric and unobstructed lateral portals. Although we found no evidence to support current flow through the central pathway, it is still a possibility that this route is used under conditions that differ from those used in our experiments, for example, when the concentration of extracellular divalents is lowered to promote pore dilation (1). Finally, our findings are directly relevant to trimeric acid sensing channels (9, 10) and to structural and mechanistic studies of other membrane proteins containing portals (32, 33), which may also be regulated and used by the transported species.

Methods

Methods are fully described in SI Methods. In short, whole-cell and single-channel currents were recorded from cultured HEK293 cells that transiently expressed either wt or cysteine-substituted hP2X4Rs. Broken-patch voltage-clamp was used to record whole-cell current when one or two applications of ATP were sufficient to compete the goals the experiment. To prevent rundown with repeated drug applications, the perforated-patch voltage clamp was used to record whole-cell current when more than two applications were required. Thiol modification was determined from changes in ATP-gated (100 μM) current measured before, during, and after applications of MTS (1 mM) or Cd²⁺ (20 μM). Single-channel current was recorded from cell-attached patches of cells whose membrane potential was zeroed by a bath solution containing 154 mM K⁺. The transmembrane potential was set to −120 mV by applying current to the recording electrode. Recording electrodes were filled with a solution containing 140 mM NaF and 0.1–1.0 μM ATP. Pf% was measured as described (20, 24).