Allosteric Modulation of Ca2+ flux in Ligand-gated Cation Channel (P2X4) by Actions on Lateral Portals*

Background: Ca2+ currents of ligand-gated ion channels are essential to cell signaling. Results: We show that the Ca2+ currents of P2X4 channels are subject to allosteric modulation. Conclusion: The fixed negative charge of a single amino acid is required for the allosteric effects of ivermectin on permeability, flux, and current deactivation. Significance: Allosteric modulators may provide therapeutic relief from symptoms of diseases such as peripheral neuropathy and hypertension. Human P2X receptors are a family of seven ATP-gated ion channels that transport Na+, K+, and Ca2+ across cell surface membranes. The P2X4 receptor is unique among family members in its sensitivity to the macrocyclic lactone, ivermectin, which allosterically modulates both ion conduction and channel gating. In this paper we show that removing the fixed negative charge of a single acidic amino acid (Glu51) in the lateral entrance to the transmembrane pore markedly attenuates the effect of ivermectin on Ca2+ current and channel gating. Ca2+ entry through P2X4 receptors is known to trigger downstream signaling pathways in microglia. Our experiments show that the lateral portals could present a novel target for drugs in the treatment of microglia-associated disease including neuropathic pain.

Human P2X receptors are a family of seven ATP-gated ion channels that transport Na ؉ , K ؉ , and Ca 2؉ across cell surface membranes. The P2X4 receptor is unique among family members in its sensitivity to the macrocyclic lactone, ivermectin, which allosterically modulates both ion conduction and channel gating. In this paper we show that removing the fixed negative charge of a single acidic amino acid (Glu 51 ) in the lateral entrance to the transmembrane pore markedly attenuates the effect of ivermectin on Ca 2؉ current and channel gating. Ca 2؉ entry through P2X4 receptors is known to trigger downstream signaling pathways in microglia. Our experiments show that the lateral portals could present a novel target for drugs in the treatment of microglia-associated disease including neuropathic pain.
The Ca 2ϩ currents of ligand-gated ion channels (LGICs) 3 play essential roles in cell signaling by regulating transmitter release, muscle contraction, and gene transcription (1,2). Most cells are exquisitely sensitive to [Ca 2ϩ ] i , and thus small changes in the amplitude of ligand-gated Ca 2ϩ currents can lead to dramatic effects on the regulation of downstream Ca 2ϩ -dependent signaling processes (3).
Recent works suggest that the ability to transport Ca 2ϩ is not necessarily a fixed channel property of LGICs. This was first demonstrated in hippocampal neurons where PKA-dependent phosphorylation enhances NMDA-gated Ca 2ϩ influx in dendritic spines by altering relative Ca 2ϩ permeability, leading to facilitation of long term potentiation (4). Other notable recent examples include PKC-dependent modulation of the Ca 2ϩ permeability of polymodal TRPV1 receptors (5) and agonist-dependent modulation of the Ca 2ϩ currents of TRPV1 (6) and TRPA1 receptors (7) receptors.
Despite these examples, a central unanswered question is whether the Ca 2ϩ currents of LGICs are susceptible to allosteric tuning by drugs (8). If this were so, then what is currently an interesting physiological phenomenon could be exploited pharmacologically. Thus, it might be possible to design drugs that allow channels to gate in response to their natural agonists but that alter the Ca 2ϩ flux of the conducted currents in a way that influences downstream cellular processes.
Here we report that the macrocyclic lactone, ivermectin (IVM), reduces fractional Ca 2ϩ current (Pf%) and relative Ca 2ϩ permeability (P Ca /P Cs ) through native and recombinant P2X4 receptor-channels, thus demonstrating for the first time allosteric modulation of the Ca 2ϩ current of a LGIC by an exogenously applied drug. This effect is absent in mutant human P2X4 receptors (hP2X4Rs) that lack the fixed negative charge of a specific acidic amino acid (Glu 51 ) in the lateral entrance to the transmembrane pore (9,10) and thereby identify a unique domain that might serve as a potential target for novel therapeutic agents.

EXPERIMENTAL PROCEDURES
Cell Culture of Mouse Cerebellar Microglia-Immortalized microglial C8-B4 cells were cultured in 35-mm dishes using Dulbecco's modified Eagle's medium (DMEM; Life Technologies, Grand Island, NY) supplemented with 10% fetal bovine serum (American Type Culture Collection, Manassas, VA), 2 mM glutamine, and antibiotics as previously described (11). The night before an experiment we added 1 g/ml lipopolysaccharide (LPS) to the culture medium and incubated the cells for an additional 12 h. On the morning of the experiment the cells were dispersed using a trypsin/EDTA (1ϫ) Hanks' buffered saline solution (Sigma), washed with DMEM, and replated onto polylysine-coated glass coverslips (Gold Seal; BD Biosciences) and bathed in a solution of DMEM and LPS. Electrophysiologi-cal recordings began 2-6 h later. We saw no ATP-gated currents in C8-B4 cells that were not exposed to LPS (11).
Mutagenesis-We studied recombinant hP2X4Rs that were made and expressed using conventional methods. Point mutations were engineered using the QuikChange Lightening Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA) and verified by automated DNA sequencing (Retrogen, San Diego, CA). Plasmid cDNA was delivered to cultured HEK293 cells using Lipofectamine LTX (Life Technologies).
Data Acquisition and Drug Application-Whole-cell currents were recorded with low resistance (1-3 megaohms), borosilicate glass pipettes (World Precision Instruments, Sarasota, FL), and AxoPatch 200B amplifiers (Molecular Devices, Sunnyvale, CA). Fura-2 fluorescence (510 nm excitation; 380 nm emission) was captured by a photomultiplier detection system (Photon Technology International, South Brunswick, NJ). Data were digitized at 10 kHz with Instrutech ITC-16 acquisition hardware (HEKA Instruments, Bellmore, NY) and AxoGraphX software (AxoGraph Scientific, Sydney, Australia). The stored data were analyzed off-line using AxoGraphX and Igor Pro software (WaveMetrics, Lake Oswego, OR). Fast solutions changes were achieved using a SF-77B Perfusion Fast- Step system (Warner Instruments, Hamden, CT). The concentration of IVM was 3 M for amplitude and deactivation experiments and 10 M for Pf% experiments except where otherwise noted.
Patch Clamp Photometry-The fraction of the total agonistgated current carried by Ca 2ϩ (i.e. Pf%) was measured using the dye-overload method of Neher (12) and Dani and co-workers (13). Our technique is described elsewhere in detail (14,15). In short, HEK293 cells transiently expressing P2X receptors were grown in 35-mm culture dishes and then re-plated at low density onto poly-L-lysine-coated glass coverslips (Gold Seal) 2-3 h before the start of the experiment. Whole-cell current and fluorescence were recorded from adherent cells using a recording pipette containing 140 mM CsCl, 10 mM tetraethylammonium chloride, 3 mM CsOH, 10 mM HEPES, and 2 mM K 5 -fura-2 at pH 7.3 (CsOH). The extracellular buffer contained 140 mM NaCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, and 10 mM HEPES at pH 7.4 (NaOH). Pf% was determined from Equation 1, Q Ca (in nC) equals ⌬F 380 /F max , ⌬F 380 is the change in fura-2 fluorescence caused by Ca 2ϩ entry (measured in "bead units"), and F max is a proportionality constant determined in separate sets of experiments (equal to 23.68 bead units/nC in most experiments). One bead units equals the average fluorescence of five Fluoresbrite 4.5 M microspheres (Polysciences, Warrington, PA). Q T is the total integrated agonist-gated current recorded using patch clamp electrophysiology (in nanocoulomb).
Bi-ionic Reversal Potential Measurements-We used round cells that were detached from cultures dishes by mechanical dispersion to minimize space-clamp errors. Whole-cell membrane current was recorded using indifferent electrodes suspended in 3 M KCl agar bridges in contact with the bath solution and the broken patch configuration of the whole-cell voltage clamp technique. The solution in the recording pipette was 150 mM CsCl, 10 mM EGTA, 10 mM HEPES, and pH 7.3 (CsOH). Relative Ca 2ϩ permeability (P Ca /P Cs ) was determined from the measured shift in reversal potential observed upon switching from the CsCl-based control solution to one composed of 110 mM CaCl 2 , 10 mM glucose, 10 mM HEPES, 2 mM CaOH 2 (pH 7.4). We changed the membrane voltage of cells bathed in each solution from Ϫ80 to 60 mV at a constant rate (1.4 V/s) before and during agonist application, and we measured the zero current level (E Rev ) from the leak-subtracted currents. Then, P Ca / P Cs was determined as where ⌬E rev equals E rev,Ca minus E rev,Cs (16). The ␥ Cs and ␥ Ca represent the activity coefficients for Cs ϩ (0.72 for 154 mM) and Ca 2ϩ (0.26 for 112 mM), respectively. Homology Model-We used the hP2X4R model built by Sébastien Dutertre (University of Queensland, St. Lucia, Australia) and based on the crystal structure of the zebrafish P2X4.1 receptor (zfP2X4.1R) (10) as previously described (9).
Statistical Tests and Reports-All data are presented as the mean Ϯ S.E. of at least six experiments. Drug treatments were analyzed by comparing data obtained before and during drug treatment using the Student's t test function. Groups of data were analyzed by one-way analysis of variance with significance determined from the Tukey's protected multiple comparison test using Instat (GraphPad Software, La Jolla, CA). A p Յ 0.01 was considered significant.

RESULTS
IVM Reduces the Contribution of Ca 2ϩ to ATP-gated Current of Mouse Cerebellar Microglia-The phenotypes of native homomeric P2X receptors are difficult to characterize because most tissues express more than one homologue and heteromeric receptors are common. We took advantage of the recent discovery of a pure population of mouse P2X4 receptors (mP2X4Rs) in LPS-stimulated C8-B4 microglia to study regulation of the Ca 2ϩ current of a native ATP-gated ion channel (11). IVM is a positive allosteric modulator of rat (17) and human (18) P2X4 receptors. In recombinant hP2X4Rs, low concentrations (EC 50 Ϸ 0.3 M) of IVM increase the maximum current evoked by saturating concentrations of ATP, and at higher concentrations (EC 50 Ϸ 2 M), IVM also slows deactivation (18). Both of these effects are readily apparent in the native mP2X4R of activated C8-B4 cells (11). To determine the effect of IVM on Pf%, we applied ATP (100 M, 3 s) while simultaneously measuring whole-cell current and fura-2 fluorescence. We found that the ATP-gated inward current (Fig. 1A) was accompanied by a decrease in the fluorescence intensity emitted at 510 nm by fura-2 when excited by 380 nm light, indicative of rise in the intracellular [Ca 2ϩ ] i (Fig. 1B). From these data, we calculated the Pf% of the mP2X4R to be 15.4 Ϯ 1.2% (Table 1), as expected from previous results (11). Next, we bathed the cells in 10 M IVM for 5 min and then reapplied ATP. The ATP-gated current was larger in the presence of IVM, and deactivation was slower (Fig. 1D). We again calculated the Pf% from the change in fluorescence (Fig. 1E) and found that the Pf% was significantly smaller (7.0 Ϯ 0.7%) than control ( Table 1). The decrease in Pf% was not due to dye saturation because the calibrated change in fluorescence (i.e. the Q Ca ) was a linear function of the total charge transfer across the surface membrane (i.e. the Q T ) (Fig. 1, C and F). Moreover, these linear functions show that the rise in free [Ca 2ϩ ] i results solely from Ca 2ϩ entry through the mP2X4R pore (19). To the best of our knowledge, this is the first report of allosteric modulation of the Ca 2ϩ current of a native LGIC.
IVM Also Decreases Pf% of Recombinant hP2X4Rs-Next, we used the wild-type recombinant hP2X4R to determine whether the inhibition of Pf% we saw in mouse cells was independent of tissue and conserved across species. Using recombinant receptors also allowed us to exploit site-directed mutagenesis to study the molecular basis of the IVM effect. In keeping with published results (9,14,15), we found that the Pf% of the hP2X4R equaled 14.0 Ϯ 0.7% in control conditions and was not significantly different from that measured from mP2X4Rs (Table 1). We also found that IVM had predictable effects (18); the ATP-gated currents of the hP2X4R evoked in the presence of 10 M IVM were larger than control and deactivated slower ( Fig. 2A). More importantly, we recapitulated the effect of IVM on the Pf% of native mP2X4Rs in the recombinant hP2X4Rs. That is, the Pf% of the ATP-gated current of the hP2X4R fell to 8.1 Ϯ 0.4% in the presence of IVM (Fig. 2B). IVM also reduced FIGURE 1. IVM reduces the Pf% of C8-B4 microglia. ATP-gated whole-cell current was recorded from mP2X4Rs of C8-B4 cells using patch pipettes containing the Ca 2ϩ -sensitive dye, fura-2. ATP (100 M) evoked an inward current (A) and a change in F 380 (B, blue trace). Q T (B, black trace) was determined by integrating the ATP-gated whole-cell current. The contribution of Q Ca to Q T is marked by the gray-shaded area of panel B. Panel C shows that Q Ca is a linear function of Q T . Panels D, E, and F show similar data obtained after a 5-min application of IVM (10 M). IVM increased peak current amplitude and prolonged deactivation (D) and at the same time decreased Pf% (E). The decrease in Pf% was not caused by saturation of fura-2 because Q Ca remained a linear function of Q T (F).

TABLE 1 Effect of IVM on Pf%
Asterisks denote a significant difference between Pf% measured in the absence (i.e. control) and presence of 10 M IVM. The number of trials for each experiment is indicated in parentheses.

Protein
Control IVM the P Ca /P Cs of the hP2X4R, measured from the change in reversal potential of ATP-gated currents obtained in extracellular solutions containing predominately Na ϩ or Ca 2ϩ (Fig. 2C). P Ca /P Cs decreased from a control value of 4.3 Ϯ 0.6 (n ϭ 8) to a new value of 2.1 Ϯ 0.1 (n ϭ 3) in the presence of IVM. Finally, we saw no effect of IVM on the Pf% of ATP-gated currents measured from HEK293 cells expressing either human P2X1 (hP2X1R) or rat P2X2 (rP2X2R) receptors ( Fig. 2D; see also Table 1). These experiments support the claim that the effects of IVM are limited to P2X4Rs (20,21).
We attempted to rigorously determine the IC 50 of IVM on the Pf% and P Ca /P Cs of the hP2X4R but were hindered by the low aqueous solubility of the lactone in water at concentrations of Ͼ10 M. Nevertheless, the effect of IVM was concentrationdependent because 3 M IVM had a significantly smaller effect on Pf% (11.1 Ϯ 0.6%; n ϭ 9) than that measured in the presence of 10 M IVM (supplemental Fig. S1A). Lower concentrations of IVM (0.1, 1 M) also appeared to have concentration-dependent effects (supplemental Fig. S1B), although the significance of these inhibitions was difficult to determine because of the limited range of attainable values of Pf% (ϳ11-15%) measured in the presence of less than 3 M IVM. Nevertheless, the data clearly show that a low (M) affinity IVM binding site mediates the reduction in Pf% (18).
In the remaining experiments we used site-directed mutagenesis of the recombinant wild-type (wt) hP2X4Rs and measurements of deactivation and Pf% to identify residues involved in the actions of IVM. We focused our study on sites previously shown to affect either channel gating (18,22,23) or Pf% (15,24).
Site-directed Mutagenesis of TM1 Has No Effect on Ability of IVM to Decrease Pf%-P2X receptors have two transmembrane-spanning domains, designated TM1 and TM2. TM2 lines the pore (25-28) and regulates both the permeability and conductance of the ATP-gated current (14, 29 -32). In contrast, there are sparse data to support such a role for TM1 in cation permeability or conduction (24,33,34). One exception is the finding that the Pf% of the rP2X2R is significantly reduced when either of two sites in TM1 (Tyr 43 and Phe 44 ) is mutated to a hydrophobic alanine (24); similar results were recently reported for rP2X3 receptors (35). Mutagenesis of the homologous residues (Tyr 42 and Val 43 ; Fig. 3A) of the hP2X4R to tryptophan reduces IVM sensitivity (22,36), and so we hypothesized that these residues might be important for the effect of IVM on Pf%. Therefore, we compared the three effects of IVM (potentiation of current, prolongation of deactivation, and reduction in Pf%) on the ATP-gated currents of the wt receptor and the tryptophan mutants (Fig. 3, B-D).
First, we found that tryptophan mutagenesis blunted the effect of IVM on the size of the hP2X4R, as expected from similar work using the rat P2X4 ortholog (36). The peak current amplitude of the wt hP2X4R measured in the presence of IVM (3 M) was 4.2 Ϯ 0.9-fold larger than control (Table 2). In contrast, IVM caused a smaller potentiation of the hP2X4-Y42W current (1.2 Ϯ 0.1-fold change) and inhibited the hP2X4-V43W current (0.6 Ϯ 0.1) (Fig. 3B; Table 2). Although these differences are significant, a firm conclusion cannot be drawn for the following reason. IVM increases the current amplitude of the wt hP2X4R primarily by increasing the open time/probability  Table 1).
of the channel (18). If so, then two reasons could explain the smaller effects of IVM on the TM1 mutants. First, the mutations could limit the ability of IVM to increase open time. Second, the open time of the tryptophan-substituted mutants could be significantly greater than that of the wt receptor, which would limit the ability of IVM to further increase P o . In-depth kinetic studies of the single channel currents of wt and mutant receptors are needed to identify the underlying cause, and such studies are problematic because of the extensive rundown in channel activity that is an innate property of hP2X4Rs (9, 18). We did not pursue these single channel experiments here because the focus of the present study is the effect of IVM on Pf%.
Next, we quantified deactivation by measuring the length of time it took for the agonist-gated current to fall from 90 to 10% of its peak current amplitude after washout of ATP (called the t 90 -10% time). We found that the -fold changes in the rates of deactivation of the mutant receptors were less pronounced in the presence of 3 M IVM (Fig. 3B), largely because tryptophan mutagenesis itself prolonged deactivation in the absence of IVM (Table 2). However, the absolute rates of deactivation of the wt and mutant receptors measured in the presence of IVM were not significantly different (see Table 2).
Finally, we measured the Pf% of these mutants before and after 10 M IVM (Fig. 3, C and D). We found that replacing Tyr 42 or Val 43 of the hP2X4R with tryptophan had no significant effect on the Pf% in the absence of IVM despite the fact that alanine mutagenesis at the homologous sites of the rP2X2R (Tyr 43 and Phe 44 ) significantly reduced Pf% (24). We also found that these mutations had no effect on the ability of IVM to attenuate the Pf% of the tryptophan-substituted receptors ( Table 1). We draw two conclusions from these data. First, the fact that mutagenesis of Tyr 42 and Val 43 has no effect on Pf% supports the hypothesis that TM1 makes a smaller contribution to permeation than gating (24,35,37,38 (14), in part because two acidic amino acids (Glu 51 and Asp 331 of the hP2X4R) provide an electrostatic environment that interacts with Ca 2ϩ (15). In our homology model of the hP2X4R, Glu 51 and Asp 331 lie just extracellular to the transmembrane domains and form part of the lateral por-tals that are the extracellular entrance to the pore (9) (Fig. 4A). IVM affects P2X4 currents by intercalating with the transmembrane domains (36), an interaction that affects the accessibility of engineered cysteines of the lateral portals to water-soluble thiol-reactive compounds (9). Thus, we hypothesized that IVM may reduce Pf% by changing the topology of the lateral portals in a way that lowers the capacity of Glu 51 and Asp 331 to facilitate Ca 2ϩ flux.
To test this hypothesis we investigated the IVM sensitivity of the hP2X4R mutants in which one or both of these acidic residues was replaced by the neutral amino acids that occupy the homologous positions (Gln 52 , Ser 326 ) in the IVM-insensitive rP2X2R. In the first set of experiments, we measured the Pf% of the mutant hP2X4Rs in the absence of IVM. Consistent with published results (15), removing the single charge of either Glu 51 (E51Q) or Asp 331 (D331S) had no effect on Pf% (Table 1), whereas removing both charges (E51Q/D331S) significantly reduced Pf% to 7.6 Ϯ 0.9%. After that, we measured the effect of 10 M IVM on the Pf% of the three mutant receptors and found that the D331S mutant resembled the wild-type receptor because both showed equivalent reductions in Pf% in the presence of IVM (Fig. 4, B and C). In contrast, IVM failed to reduce the Pf% of either the E51Q (Fig. 4D) or the E51Q/D331S mutants ( Fig. 4C; Table 1), which shows that the ability of IVM to regulate Pf% is critically dependent on the presence of the fixed negative charge of Glu 51 in hP2X4R.
Removing the Fixed Charge of Glu 51 Also Attenuates Effect of IVM on Current Deactivation-We then looked to see if removing the charge of Glu 51 and/or Asp 331 changed the ability of IVM to prolong deactivation after washout of 3 M ATP. We found that the t 90 -10% of the E51Q/D331S mutant increased from a control value of 0.13 Ϯ 0.01 s to new value of 1.7 Ϯ 0.4 s in the presence of 3 M IVM. Although the prolongation is significant, it is ϳ10-fold shorter than the ϳ125-fold change observed for the wt receptor ( Table 2). The effect of IVM on the t 90 -10% of the ATP-gated current mediated by the E51Q mutant was also attenuated 10-fold (Fig. 4E), increasing from a control value of 0.32 Ϯ 0.03 s in the absence of IVM to 3.9 Ϯ 1.1 s in the presence of IVM. In contrast, IVM caused an ϳ80-fold increase in the time course of deactivation of currents mediated by the D331S mutant as the t 90 -10% increased from 0.36 Ϯ 0.05 s to 27.6 Ϯ 6.4 s (Fig. 4E; Table 2).
Removing the fixed charge of Glu 51 also reduced the effect of 3 M IVM on current amplitude (Table 2). Again, the reduced ability to potentiate current could reflect either a change in the

Effect of IVM on peak current and deactivation
Deactivation rates are the t 90 -10% times measured from the falling phase of the ligand-gated current that occured upon washout of ATP. ATP was applied at a concentration of 100 M for zebrafish experiments and at a concentration of 3 M in all other experiments. The concentration of IVM was 3 M. The number of trials for each experiment is indicated in parentheses. behavior of IVM or an innate property of the mutant channels as discussed above; we did not further investigate this effect. Taken together, our data show that Glu 51 is necessary for the effect of IVM on Pf% and current deactivation. The strong correlation of the effect of IVM on Pf% and deactivation suggests that a common final pathway may underlie both effects.

Protein -Fold change in peak current by IVM Deactivation rate control Deactivation rate IVM -Fold change in deactivation
Substitution of Glu 51 into zfP2X4.1R Imparts Limited Sensitivity to IVM-The zfP2X4.1R lacks the fixed negative charge of Glu 51 and Asp 331 and has a Pf% that is smaller than that of the hP2X4R (14,15). Its sensitivity to IVM is unknown. We mutated the zfP2X4.1R to place acidic amino acids at sites equivalent to Glu 51 and Asp 331 of the hP2X4R and studied the effect of IVM on the wt and the double mutant (N34E/N334D) receptors. Our experiments produced four noteworthy results. First, in keeping with results obtained using other P2X4 orthologs (21), we found that 3 M IVM caused a ϳ9-fold change in the peak current amplitude of the wt zfP2X4.1R caused by applying 100 M ATP (Fig. 5A). To the best of our knowledge this is the first report of an effect of IVM on the zebrafish ortholog. Second, unlike its effects on the P2X4Rs of other species, IVM had no significant effect on the time course of current deactivation of the wt zfP2X4.1R (Fig. 5A). However, IVM substantially prolonged deactivation of the N54E/N334D double charge mutant (Fig. 5B), as the t 90 -10% changed from a control value of 0.4 Ϯ 0.1 s in control conditions to 14 Ϯ 2.5 s in the presence of 3 M IVM (Table 2). Again, in concordance with our results with the hP2X4R (see Fig. 4E), these data suggest that the effect of IVM on current deactivation requires the presence of a fixed charge in the lateral portal of P2X receptors. Third, inserting the fixed negative charges of glutamate and aspartate into the zfP2X4.1R had no effect on Pf% measured in the absence of IVM (Fig. 5C). This outcome was unexpected because the Pf% of the rP2X2R, which also lacks the two essential acidic amino acids, doubles after insertion of glutamate and aspartate at the equivalent sites (Fig. 5C (15)). Fourth, 10 M IVM had no effect on the Pf% of currents evoked in cells expressing either the wt zfP2X4.1R or the zfP2X4.1-N54E/ N334D double mutant (Fig. 5C). Why insertion of fixed charge failed to increase Pf% or impart IVM sensitivity to the Pf% of the zfP2X4.1R is presently unknown.
A Double Charge Mutant of rP2X2R Is IVM-insensitive-In a final set of experiments, we looked to see if we could impart full or limited sensitivity to IVM by inserting Glu 51 into an IVMinsensitive, non-P2X4 receptor. The hP2X1R has an acidic charge at the homologous site and a relatively high Pf% but is insensitive to the effects of IVM on Pf% (Fig. 2D), potentiation of current, and current deactivation. These data suggest that Glu 51 is not the sole determinant of IVM sensitivity. To test this hypothesis, we measured the IVM sensitivity of a double rP2X2R mutant (Q52E/S326D) that contains fixed charge at sites equivalent to Glu 51 and Asp 331 . We found that current amplitude and the rate of deactivation of the double mutant rP2X2R-Q52E/S326D were not affected by IVM (Fig. 5D). The Pf% of rP2X2R-Q52E/S326D was similarly unaffected (Fig. 5C).
The following picture, therefore, emerges when all of the data are considered together. We conclude that the fixed negative charge of Glu 51 is required for the slowed deactivation and reduced Pf% caused by the action of IVM on hP2X4Rs. However, Glu 51 does not bestow IVM sensitivity by itself, as the effects of IVM sensitivity are restricted to P2X4Rs. The later finding suggests that additional structural loci unique to P2X4Rs are required for the multiple effects of IVM.

DISCUSSION
IVM has two well described effects on P2X4R current. First, it potentiates whole-cell current by binding to a high affinity (nM) site, and second, it prolongs deactivation by binding a low affinity (M) site (17,18,39). In this report we show for the first time that IVM also reduces the contribution of Ca 2ϩ to the ATP-gated current of native and recombinant P2X4Rs. Furthermore, the effects of IVM on deactivation and Pf% are mark-edly attenuated by removing the fixed negative charge of a single acidic amino acid (Glu 51 ) in the extracellular entrance to the transmembrane pore. Our finding that drugs modulate the ATP-gated Ca 2ϩ current through actions on the lateral portals provide further support for the idea that these domains are flexible and move as the channel opens (9,18). Drugs that alter these movements may, therefore, provide therapeutic relief from symptoms of diseases such as peripheral neuropathy (40) and hypertension (41).
Rather than point to a third and distinct action of IVM on P2X4R current, our data suggest that a common transduction pathway may underlie the effects of IVM on Ca 2ϩ flux and current deactivation for the following two reasons. First, although the limited solubility of IVM in water prevented construction of a complete concentration-response curve, the available data suggest that IVM binds to a low affinity (M) site to mediate its effect on Pf% (see supplemental Fig. S1). The micromolar potency of IVM at this site is similar to the estimated EC 50 for the low affinity effect of IVM on deactivation. Second, removing the fixed negative charge of Glu 51 substantially attenuates the effects of IVM on both Pf% and current deactivation, thus demonstrating a common structural locus in the lateral portals for both effects. Again, the lateral portals are flexible and move when the channel opens (9,42). Our new data suggest that this movement may play a role in regulating the nature of cationic flux through the channel.
We speculate that the mechanism by which IVM attenuates Pf% and deactivation in P2X4Rs involves pushing the channel into a non-native conductance state with a lower P Ca /P Na . This hypothesis is partially supported by the single channel data of Priel and Silberberg (18) who showed that micromolar concentrations of IVM increase mean open time, open probability, and single channel conductance of the hP2X4R channel. The increase in conductance suggests that IVM decreases P Ca /P Na and Pf% by increasing Na ϩ permeability and flux. Although this may be so, it cannot fully explain the magnitude of the effects described here. That is, a ϳ100% increase in the size of the Na ϩ current is required to fully explain the observed 50% reduction in Pf% by IVM, which is far greater than the ϳ20% increase in single channel conductance reported by Priel and Silberberg (18). A concomitant decrease in Ca 2ϩ permeability and flux is, therefore, required too.
Our data strongly suggest that Glu 51 lies in a transduction pathway that mediates the low affinity effects of IVM. However, the data also imply that additional domains are involved in the selective effects of IVM on P2X4Rs. For example, the double mutant P2X2R (Q52E/S326D) has a glutamate at a position analogous to the Glu 51 of P2X4R, but neither the Pf% nor the deactivation was altered by IVM. Similarly, wt hP2X1Rs have the equivalent charges but are unaffected by IVM. In contrast, the ability to prolong deactivation by IVM was successfully conferred to the zfP2X4R by introducing fixed charge at the position analogous to Glu 51 of the hP2X4R. Interestingly, our data suggests that the wild-type zfP2X4R is still sensitive to the high affinity effects of IVM on current amplitude even though its deactivation lacks sensitivity to IVM. Further experiments comparing the zebrafish and mammalian P2X4R channels may help to tease apart the structural loci underpinning the separate effects of IVM on channel function and explain why the effects are specific to P2X4R-like receptors.
Finally, we previously showed that the amplitude of the Ca 2ϩ current of rat and human P2X1 and P2X4 receptors is highly dependent on the two relevant fixed charges of the lateral portals discussed here (15). The zfP2X4.1R lacks these acidic residues and has a lower Pf% than its mammalian counterparts. However, unlike our published results with rP2X2Rs (15), we did not record an increase in Pf% when carboxylates were added to the relevant positions in the zfP2X4.1R. We cannot fully explain the discrepancy between the double mutant rP2X2R and zfP2X4.1R results. Perhaps other undiscovered domains contribute to ion selection, and these domains are missing in the zfP2X4.1R. In addition, due to the low current densities achieved with maximal stimulation of these receptors in HEK293 cells (in the absence of IVM), we used long exposures to ATP (5-10 s) to obtain a Ca 2ϩ flux that was large enough to accurately measure Pf%. These long applications may push the receptor into the I 2 permeability state (43), which may have different Pf% properties.