Subunit-specific Coupling between γ-Aminobutyric Acid Type A and P2X2 Receptor Channels*

ATP and γ-aminobutyric acid (GABA) are two fast neurotransmitters co-released at central synapses, where they co-activate excitatory P2X and inhibitory GABAA (GABA type A) receptors. We report here that co-activation of P2X2 and various GABAA receptors, co-expressed in Xenopus oocytes, leads to a functional cross-inhibition dependent on GABAA subunit composition. Sequential applications of GABA and ATP revealed that αβ- or αβγ-containing GABAA receptors inhibited P2X2 channels, whereas P2X2 channels failed to inhibit γ-containing GABAA receptors. This functional cross-talk is independent of membrane potential, changes in current direction, and calcium. Non-additive responses observed between cation-selective GABAA and P2X2 receptors further indicate the chloride independence of this process. Overexpression of minigenes encoding either the C-terminal fragment of P2X2 or the intracellular loop of the β3 subunit disrupted the functional cross-inhibition. We previously demonstrated functional and physical cross-talk between ρ1 and P2X2 receptors, which induced a retargeting of ρ1 channels to surface clusters when co-expressed in hippocampal neurons (Boué-Grabot, E., Emerit, M. B., Toulme, E., Seguela, P., and Garret, M. (2004) J. Biol. Chem. 279, 6967–6975). Co-expression of P2X2 and chimeric ρ1 receptors with the C-terminal sequences of α2, β3, or γ2 subunits indicated that only ρ1-β3 and P2X2 channels exhibit both functional cross-inhibition in Xenopus oocytes and co-clustering/retargeting in hippocampal neurons. Therefore, the C-terminal domain of P2X2 and the intracellular loop of β GABAA subunits are required for the functional interaction between ATP- and GABA-gated channels. This γ subunit-dependent cross-talk may contribute to the regulation of synaptic activity.

and ligand-gated channels reciprocally affect their functions by direct interaction between intracellular domains, as illustrated for D5 and GABA A , 1 or D1 and N-methyl-D-aspartic acid (2,3). Cross-talk between distinct ligand-gated channels has been described between ATP P2X receptors and either acetylcholine nicotinic receptors (4 -7), 5-hydroxytryptamine 3 receptors (8,9), or GABA A receptors in dorsal root ganglia neurons (10), as well as between GABA and glycine receptors in spinal cord neurons (11). Cross-talk between ligand-gated channels is characterized by current occlusion during simultaneous agonist application, although the mechanisms remain unclear. Intracellular phosphorylation pathways underlie asymmetric cross-inhibition between GABA A and glycine receptors (11). Cross-talk between GABA A and P2X receptors expressed in dorsal root ganglia neurons has been described as a chlorideand calcium-dependent interaction (10), whereas we demonstrated that a physical interaction involving the intracellular domains of each receptor led to the functional cross-talk between 5-HT 3 and P2X 2 receptors (9), as well as between P2X 2 and 1/GABA C (12). Because GABA and ATP are synaptically co-released in spinal cord and hypothalamic neurons (13,14), where they activate co-localized ATP-and GABA-gated channels, cross-talk between inhibitory GABA A and excitatory P2X receptors may represent an important, fast process for controlling the signal transmission phenomenon. Therefore, our aim was to investigate a putative cross-modulation between GABA A and P2X receptors and the mechanisms underlying this process.
GABA and P2X receptors have distinct structures. GABA A receptors are pentameric structures formed by differential assembly of multiple subunits (␣1-6, ␤1-3, ␥1-3, ␦, , and ⑀), and GABA C receptors are composed of 1-3 subunits (15). Each of the subunits contains four hydrophobic transmembrane domains and extracellular termini. GABA-gated chloride channels mediate fast inhibition and thus, play a fundamental role in the physiology and physiopathology of the nervous system (16). The diversity of their functional properties, pharmacology, and subcellular targeting depends to a large extent on subunit composition (16). P2X receptors are ATP-gated cation channels consisting of a family of seven subunits with two transmembrane domains, a large extracellular loop, and intracellular termini (17). P2X receptors are implicated in fast excitatory synaptic transmission (18). However, little is known about purinergic transmission in the brain, despite the wide distribution of P2X receptor subunits throughout the central nervous system (19).
Xenopus Oocyte Preparation and Injection-Oocytes were prepared as previously described (21). Stage V and VI oocytes were manually defolliculated before cytoplasmic or nuclear injection (Nanoject II, Drummond) of cRNA or cDNA, respectively. 0.2 ng of cRNA coding for P2X 2 or P2X 2 YFP, or 15 ng of RNA coding for P2X 2 TR, together with 10 -20 ng of each of the GABA subunits, were injected to reach similar expression levels for both channels. Then the oocytes were incubated in Barth's solution containing 1.8 mM CaCl 2 and gentamycin (10 g/ml, Sigma) at 19°C for 1-4 days prior to electrophysiological recordings. For competition experiments, RNAs coding for minigenes were injected (50 -60 ng each to reach a 2:1 minigene:receptor ratio) immediately after injection of receptor RNAs.
Electrophysiology-Two-electrode voltage clamp recordings were carried out at room temperature using glass pipettes (1-2 M⍀) filled with 3 M KCl solution to ensure a reliable holding potential. Oocytes were voltage clamped at Ϫ60 mV, and the membrane currents were recorded through an OC-725B amplifier (Warner Instruments) and digitized at 500 Hz. Oocytes were perfused at a flow rate of 10 -12 ml/min with Ringer's solution (Na ϩ -R), pH 7.4, containing 115 mM NaCl, 5 mM NaOH, 2.5 mM KCl, 1.8 mM CaCl 2 or BaCl 2 , and 10 mM HEPES. Cation permeability was addressed using an extracellular solution (NMDG-R) with a low sodium concentration containing 5 mM NaCl, 5 mM NaOH, 2.5 mM KCl, 1.8 mM CaCl 2 , 110 mM NMDG, and 10 mM HEPES at pH 7.4. All drugs (purchased from Sigma) were dissolved in the perfusion solution and applied using a computer-driven valve system (BPS8, Ala Scientific). To allow for the differences in GABA current and ATP current kinetics, we compared the peak of actual responses to the peak of predicted additive responses obtained using Axograph software (Axon Instruments) and not with the sum of the peaks of individual responses. Numerical values are expressed as mean Ϯ S.E. Statistical comparisons were assessed using Student's t test. The differences were considered significant if p Ͻ 0.05. For I/V experiments, currents were measured at the peak of the response obtained at several membrane potentials (Ϫ60 to ϩ30 mV). Reversal potentials (V rev ) were estimated from the I-V relationship, determined by linear regression test analysis. All data were analyzed using Prism 4.0 (GraphPad, San Diego, CA).
Cell Culture and Transfection-Neurons were cultured as previously described (12). Briefly, hippocampi of rat embryos were dissected on day 18. Dissociation was achieved after trypsinization, with a Pasteur pipette. Cells were counted and plated on poly-D-lysine-coated (50 g/ml), 15-mm diameter coverslips (Electron Microscopy Sciences), at a density of 50,000 -70,000 cells per 16-mm dish (250 -350 per mm 2 ), in complete Neurobasal medium supplemented with B27 (Invitrogen), containing 1 mM L-glutamine, and penicillin G (10 units/ml)/streptomycin (10 mg/ ml). Four hours after plating, the coverslips were transferred to a 90-mm dish containing conditioned medium obtained by incubating the complete medium described above on a glial culture (70 -80% confluency) for 24 h. The medium was partially changed every 3-4 days. Hippocampal neurons were transfected at 7-8 days in vitro as follows: for each coverslip, a total of 1.5 g of plasmid DNA was mixed with 50 l of Neurobasal medium without B27 supplement. For co-transfection experiments, the proportions of each plasmid were first adjusted to reach similar expression levels and the plasmids were thoroughly mixed prior to the addition of 1.5 l of LipofectAMINE 2000 (Invitrogen) in 50 l of Neurobasal medium. 150 l of complete Neurobasal medium containing B27 supplement was applied to the neuronal cul-ture and incubated for 3 h at 37°C. Expression was then conducted for 48 h in the original medium that was added back to the neurons. Under these conditions, the transfection rate reached 10 -25%, and 60 -70% of transfected neurons were co-transfected with two plasmids.
Immunocytochemistry and Confocal Imaging-Immunofluorescence was performed 48 h after transfection (9 -10 days after plating). Coverslips were washed twice with PBSϩ (phosphate-buffered saline containing 0.1 mM CaCl 2 and 0.1 mM MgCl 2 ) at 37°C, and fixed for 15 min with paraformaldehyde (4%) containing 4% sucrose, at 37°C in PBSϩ. After three 10-min washes in PBS, they were incubated for 30 min in antibody buffer (2% bovine serum albumin, 3% normal goat serum, 3% normal donkey serum, and 0.1% Triton X-100 in PBS). 1 chimeras were labeled with affinity purified anti-1 antibodies (1:200) directed against the extracellular N-terminal sequence (22) for 2 h at room temperature, and revealed with AlexaFluor 568 -conjugated donkey antirabbit IgG (Jackson ImmunoResearch, West Groove, PA) at 1:1000 dilution, for 1 h at room temperature. Immunofluorescence images were generated using a Leica TCS-400 laser scanning confocal microscope. Contrast and brightness were adjusted to ensure that all relevant pixels were within linear range. Images are the product of a 16-fold line average. For double labeling experiments, pictures were generated using Adobe Photoshop 5.0.
Quantitative Measurements of Fluorescence Profiles-Fluorescence profiles along labeled neurites were measured using Lucia 4.71 imaging software (Nikon). For each neuron, the contrast and brightness of confocal images were adjusted using identical parameters. The longest labeled neurite was then fitted with a polyline along which the fluorescence profile was plotted against the distance from the cell body. In all experiments, neurons overexpressing either receptor were excluded from the study. In co-transfection experiments, only neurons expressing similar levels of both receptors were included in the study. Variability between individual neurons was overcome by using the cumulated fluorescence profiles obtained with 50 neurons of each group as a relevant parameter for quantifying the radial distribution of 1 chimeras 48 h after transfection.

RESULTS
Reciprocal Cross-inhibition between ␣␤-containing GABA Receptors and P2X 2 Receptors-To assess whether GABA A and P2X receptors cross-modulated, we first co-expressed different combinations of GABA A subunits with P2X 2 channels in Xenopus oocytes and recorded responses to separate or combined application of ATP and GABA. In oocytes co-expressing ␣2␤3 and P2X 2 channels, the application of saturating concentrations of either GABA (100 M) or ATP (100 M) evoked inward currents characteristic of the activation of GABA A and P2X 2 receptors, respectively ( Fig. 1A and Table I summarizes data in Figs. 1 and 2). However, co-application of 100 M GABA plus 100 M ATP evoked inward responses (Fig. 1A, Actual) significantly smaller (p Ͻ 0.0005, n ϭ 15) than the sum of individual responses. The amplitude of I ATPϩGABA represented 69.40 Ϯ 3.70% of the predicted current, i.e. the arithmetic sum of I GABA and I ATP . A significant inhibition (p Ͻ 0.0005, n ϭ 8) was also recorded when GABA was applied during ATP application, or when ATP was applied during GABA application. I ATP then GABA and I GABA then ATP represented 76.90 Ϯ 3.94% and 71.69 Ϯ 2.6% of the predicted current, respectively (Fig. 1, A and B, and Table I). In oocytes co-expressing P2X 2 and GABA A receptors formed either by ␣1␤3, ␣3␤3, or ␣2h␤1 subunits, co-activation by both agonists elicited responses significantly smaller than the sum of the individual application of ATP and GABA (Fig. 1B). I ATPϩGABA represented 68.58 Ϯ 6.7% (n ϭ 5), 60.3 Ϯ 2.25% (n ϭ 5), and 78.04 Ϯ 5.95% (n ϭ 5) of the predicted current for ␣1␤3, ␣3␤3, and ␣2h␤1, respectively. Moreover, identical inhibition was observed during sequential application (ATP then GABA or GABA then ATP). As depicted in Fig. 1B and Table I, the inhibition level (20 -40%) was unaffected by receptor density (ATP plus GABA current ranged from 0.5 to 6 A) or ATP versus GABA current ratio. In oocytes expressing P2X 2 or GABA A receptors alone, ATP did not gate or modulate GABA-gated channels, and, conversely, GABA did not activate or modulate P2X 2 channels (data not shown).
These results showed that co-activation of P2X 2 and ␣␤-containing GABA A receptors induced an instantaneous, bidirectional current occlusion, demonstrating that GABA A and P2X 2 receptors are functionally non-independent.
Interestingly, when GABA was applied during ATP application, the current amplitude was not significantly different (p Ͼ 0.5, n ϭ 15) from the sum of I GABA and I ATP in oocytes coexpressing P2X 2 and ␣2␤3␥3 (I ATP then GABA ϭ 106 Ϯ 4.1% of the predicted current). P2X 2 and ␣2␤3␥2 currents were equally additive when GABA application started during continuous application of ATP (I ATP then GABA ϭ 97.5 Ϯ 2.5% of the predicted current, n ϭ 6, Fig. 2, B and C) showing that GABA decreased the ATP response of P2X 2 receptors, whereas ATP had no effect on the GABA current of ␥ subunit-containing receptors. As illustrated in Fig. 2D, the enhancement by flunitrazepam (1 M) of the current evoked by 10 M GABA (I GABAϩflunitrazepam ϭ 152 Ϯ 15% of the control, n ϭ 6) indicated that ␥ subunits were associated with ␣ and ␤ subunit complexes. Taken together, these results show that ATP-gated channels interact functionally with GABA A receptors irrespective of their subunit composition and, more importantly, that the direction of current occlusion depends on the molecular composition of the GABA A receptors. GABA A receptors closed substantial proportions of P2X 2 receptors during co-activation (and vice versa), whereas ATP-gated channels failed to close GABA-gated channels associated with a ␥ subunit. Holding membrane potential (V h ) ϭ Ϫ60 mV was monitored during all recordings.

TABLE I Cross-inhibition between P2X 2 and various combinations of GABA A subunits
Amplitude of currents recorded by applying a saturating concentration of ATP (100 M), GABA (100 M or 1 mM with ␣ 3␤ 3), or a mixture of both agonists at a holding potential of Ϫ60 mV. The average inhibition (%) indicates that similar current occlusion occurred between P2X 2 and various ␣ ␤ GABA A receptors during either simultaneous (ATPϩGABA) or sequential applications: ATP during application of GABA (GABA/ATP) and reciprocally (ATP/GABA). On the contrary, the ␥ -subunit containing GABA A receptors inhibited P2X 2 , but P2X 2 receptors did not inhibit GABA A . Voltage-, Calcium-, and Chloride-independent Current Occlusion-Sokolova and coworkers (10) described a current occlusion between GABA-and ATP-induced currents in dorsal root ganglia neurons and postulated that chloride ions were the main coupling agent. We therefore investigated a potential chloride or voltage dependence of the cross-talk by recording currents evoked by the activation of P2X 2 and GABA A receptors at different holding potentials. Currents induced by co-application of ATP plus GABA were significantly smaller (p Ͻ 0.0005, n ϭ 7) than the sum of the individual I ATP and I GABA , at holding potentials ranging from Ϫ60 to ϩ30 mV (I ATPϩGABA was between 64 and 76% of the predicted current, Fig. 3A). We also recorded crossinhibition in the absence of CaCl 2 in the extracellular solution (Fig. 3A). Cross-inhibition was observed with outward GABA currents, i.e. chloride entry, suggesting that chloride was not essential to the occlusion process. However, ATP-induced outward currents are difficult to record due to a marked inward rectification of P2X 2 channels (23). We therefore used cationic GABA receptors to determine whether chloride was implicated in the cross-talk. Mutation of amino acids in the M1-M2 loop of the ␤3 subunit to the corresponding amino acids of the ␣7 nicotinic acetylcholine subunit rendered the GABA A cation-selective, with no significant permeability to anions, on co-expression with wild type ␣ and ␥ subunits (20). We co-injected oocytes with cDNAs coding for P2X 2 , ␣2, and human mutated ␤3 subunits (h␤3SGE) and recorded responses to individual and combined application of agonists at a holding potential of Ϫ60 mV. Application of GABA (100 M) induced a slowly desensitizing inward current (I GABA ϭ Ϫ0.71 Ϯ 0.4 A, n ϭ 7), and ATP evoked non-desensitizing responses (I ATP ϭ Ϫ2.8 Ϯ 1.1 A, n ϭ 7). Co-activation of both receptors by simultaneous or sequential application evoked inward responses significantly smaller (p Ͻ 0.0005, n ϭ 7) than the predicted sum of I ATP and I GABA (Fig. 3C). I ATPϩGABA , I ATP then GABA , and I GABA then ATP represented 73.26 Ϯ 4.7%, 73 Ϯ 4.0%, and 65 Ϯ 5.0% of the predicted current, respectively. Current-voltage relationships for ␣2h␤3SGE receptors and wild-type ␣2h␤3 receptors were determined in normal Ringer solution Na ϩ -R and in low Na ϩ Ringer (NMDG-R). As depicted in Fig. 3D, the I-V curve for ␣2h␤3SGE revealed a shift in reversal potential of ϳ26 mV for NMDG-R (E rev ϭ Ϫ5.17 Ϯ 0.7 mV, n ϭ 5) compared with Na ϩ -R (E rev ϭ Ϫ31.33 Ϯ 2.0 mV, n ϭ 5). In contrast, the I-V curve for wild type ␣2h␤3 receptors exhibited a minor shift of ϳ4 mV for NMDG-R (E rev ϭ Ϫ 34.22 Ϯ 2 mV, n ϭ 4) compared with Na ϩ -R (E rev ϭ Ϫ 30.28 Ϯ 2 mV, n ϭ 4). These data are in agreement with the cation-selective permeability of the ␣2h␤3SGE receptor (20). Taken together, these experiments demonstrated that reciprocal cross-inhibition between GABA A and P2X 2 was a receptor-mediated, voltage-, chloride-, and calcium-independent phenomenon.
Interaction of Intracellular Domains of P2X 2 and GABA A Receptors Are Crucial for Cross-inhibition-Previous studies showed that intracellular domains of receptor channels were involved in cross-inhibition between P2X and 5-HT 3 or 1/ GABA C receptors (9,12). We investigated whether the C-terminal domain of P2X 2 and the intracellular domains of GABA A subunits participated in the current occlusion (Fig. 4A). We first co-expressed a C-terminal truncated form of P2X 2 (P2X 2 TR) with ␣2␤3 GABA A receptors. Co-activation of both receptors by co-application of ATP and GABA evoked inward currents (Ϫ2.06 Ϯ 0.92 A) not significantly different (p Ͼ 0.5, n ϭ 5) from the sum of the individual I ATP (Ϫ0.54 Ϯ 0.19 A) and I GABA (Ϫ1.5 Ϯ 0.73 A). I ATPϩGABA represented 100 Ϯ 1.4% of the predicted value (Fig. 4, B and F) showing that P2X 2 TR and GABA A are functionally independent. To further examine the role of the intracellular domains of each receptor in the cross-inhibition mechanism, we carried out competition experiments with minigenes encoding the N or C termini of P2X 2 receptors (X2-NT corresponding to amino acids 1-29 and X2-CT corresponding to amino acids 374 -472) and a minigene coding for the intracellular loop between hydrophobic domains M3 and M4 of the ␤3 GABA A subunit (␤3-IL2 corresponding to amino acids 327-450). Although overexpression of X2-NT had no effect (Fig. 4C), X2-CT (Fig. 4D) and ␤3-IL2 (Fig. 4E) significantly blocked the cross-inhibition between wild-type P2X 2 and ␣2␤3 receptors. Co-application of ATP plus GABA elicited responses not significantly different (p Ͼ 0.5) from the sum of the individual responses of ATP and GABA. I ATPϩGABA represented 103.1 Ϯ 7.8% (n ϭ 5) and 108.8 Ϯ 11.17% (n ϭ 5) of the predicted sum in presence of X2-CT and ␤3-IL2, respectively. On the contrary, I ATPϩGABA was significantly smaller (I ATPϩGABA ϭ 72.03 Ϯ 4.3% of the predicted current p Ͻ 0.0005, n ϭ 5) than the sum of separate applications in the presence of X2-NT. These data demonstrate that the C-terminal domain of P2X 2 or the intracellular loop of the ␤3 subunits are required for functional interaction between P2X and GABA A receptors.
FIG. 2. Asymmetric current occlusion between P2X 2 and ␥ subunit containing GABA-gated channels. A and B, in oocytes coexpressing P2X 2 plus ␣2␤3␥3 (A) or P2X 2 plus ␣2␤3␥2 (B), co-application of ATP plus GABA induced currents (Actual) significantly smaller than the sum (Predicted) of the individual ATP and GABA responses. ***, p Ͻ 0.0005, n ϭ 22 and 6, respectively. Additive responses occurred when GABA was applied during ATP application, n ϭ 15 and 6, respectively, whereas non-additive responses were recorded when ATP was applied during GABA application. Holding membrane potential (V h ) ϭ Ϫ60 mV. Currents from two oocytes are illustrated (panels a and b). C, mean current amplitudes of responses normalized to the predicted response from each cell. ns, not significant. D, representative currents evoked by a subsaturating concentration of GABA (10 M), either alone or in presence of 1 M flunitrazepam Differential Cross-inhibition between Chimeric 1/GABA A and P2X 2 Receptors-Given that homomeric 1/GABA C receptors interact functionally with P2X channels and that the intracellular loop between M3 and M4 of the 1 subunit is involved in the mechanism (12), we generated chimeric GABA C / GABA A subunits to determine which of the GABA A subunits was responsible for the cross-inhibition. The sequence downstream from the third transmembrane domain of the 1 subunit was swapped with the homologous domain of either ␣2, ␤3, or ␥2 subunits of GABA A receptors (Fig. 5A). In Xenopus oocytes expressing either 1-␣2, 1-␤3, or 1-␥2 chimeric subunits, the application of GABA (10 M) induced non-desensitizing inward currents similar to wild-type 1 channels, showing that the chimeric subunits formed functional homomeric GABA receptors. In addition, ATP (100 M) did not gate or modulate GABA-evoked responses (Fig. 5A).
In oocytes co-expressing P2X 2 and 1-␣2 channels, simultaneous application of GABA (10 M) and ATP (100 M) induced currents (Ϫ1.61 Ϯ 0.34 A, n ϭ 24) that were not significantly different (p Ͼ 0.5, n ϭ 24) from the sum of responses to separate applications of ATP (Ϫ1.25 Ϯ 0.34 A) and GABA (Ϫ0.65 Ϯ 0.13 A). I ATPϩGABA represented 95.47 Ϯ 3.40% of the predicted current (Fig. 5, B and E). When GABA was applied during ATP application or ATP during GABA responses, I ATP then GABA and I GABA then ATP represented 85.34 Ϯ 1.8% and 87.10 Ϯ 2.0% of the predicted current, respectively. These values are significantly different (p Ͻ 0.0005) from those previously recorded between wild-type 1 and P2X 2 channels (ϳ75% of the predicted current), suggesting that 1-␣2 channels exhibit a less marked functional interaction with P2X 2 . In oocytes co-expressing P2X 2 and 1-␤3 chimeric receptors, significant inhibition (p Ͻ 0.0005, n ϭ 10) was observed with both simultaneous and sequential application. I ATPϩGABA , I ATP then GABA , or I GABA then ATP represented 66.54 Ϯ 4.35%, 67. 34 Ϯ 1.80%, or 64.32 Ϯ 2.50% of the predicted current, respectively (Fig. 5, C and E), showing that 1-␤3 fully reconstituted the ability of 1 channels to interact functionally with P2X 2 channels. Finally, recordings of oocytes (n ϭ 18) expressing 1-␥2 and P2X 2 channels showed that simultaneous application evoked responses representing 95.8 Ϯ 6.4% of the predicted sum of individual ATP and GABA responses (Fig. 5, D  and E). Additive responses were also recorded when GABA was applied during ATP application and vice versa (I ATP then GABA ϭ 92.63 Ϯ 5.57% and I GABA then ATP ϭ 89.21 Ϯ 5.5% of the predicted current) demonstrating that 1-␥2 and P2X 2 functioned independently in oocytes. These data suggest that the intracellular domain of ␤ subunits is determinant in the functional cross-inhibition between GABA A and P2X 2 receptors.
Differential Molecular Interactions between 1 Chimeras and P2X 2 Receptors Expressed in Hippocampal Neurons-In addition to the cross-inhibition, we have previously shown a physical association of 1/GABA C and P2X 2 receptors by co-purification of both receptors. Because the physical link of the two receptor types was conveyed by the co-clustering of 1 and P2X 2 subunits and the retargeting of 1 induced by the presence of the P2X 2 subunit in transfected neurons (12), hippocampal neurons were transfected with the wild-type 1 subunit (1-wt) or the three chimeras (1-␣2, 1-␤3, or 1-␥2), together with the P2X 2 -YFP subunit. Fluorescence profiles of anti1/Alex-aFluor 568 and YFP revealing 1 and P2X 2 , respectively, over-  5). Non-additive ATP and GABA responses were observed between Ϫ60 to ϩ30 mV. I ATPϩGABA recorded at all tested potentials (black filled circles) and in calcium-free extracellular solution (black filled squares) was significantly smaller (p Ͻ 0.0005) than the predicted current (open circles). B and C, reciprocal current occlusion was also observed in oocytes co-expressing cationic ␣2h␤3SGE GABA A receptors and P2X 2 channels. ***, p Ͻ 0.0005, n ϭ 7. D, current-voltage relationships for GABA A receptors ␣2h␤3 and ␣2h␤3SGE obtained in normal extracellular ringer solution (black squares) and low sodium ringer solution (black filled circles). A negative shift in reversal potential was observed for the ␣2h␤3SGE receptor when extracellular sodium was replaced by NMDG, whereas no shift was observed for the wild-type ␣2h␤3 receptors. lapped in proximal dendrites in 40 -50% of co-transfected-neurons, demonstrating a clustering of both receptors (Fig. 6). Matching of the fluorescence profiles of the two labels was also detected in 1-␤3 and P2X 2 -YFP co-transfected neurons, revealing compartment clusters. On the contrary, no co-localization of P2X 2 -YFP clusters with 1-␣2 or 1-␥2 was observed (Fig. 6).
When expressed alone in hippocampal neurons, 1 exhibited a punctate distribution restricted to the close vicinity of cell bodies and were scarcely found distally (Fig. 7A). Conversely, P2X 2 -YFP was uniformly distributed throughout the neurons with a predominantly surface topology (Fig. 7A). These data were consistent with previous reports (24,25). A clear increase in the distance from cell bodies (radial topology) of 1-wt in the presence of P2X 2 -YFP subunits was revealed by cumulated fluorescence intensity profiles (Fig. 7, B and C), indicating that the presence of P2X 2 -YFP affected the routing of the 1-wt subunits. An increase in radial topology of 1-␤3 was also observed in the presence of P2X 2 -YFP, whereas there was no difference between the cumulated fluorescence profiles obtained with neurons transfected with 1-␣2 or 1-␥2, in the presence or absence of P2X 2 -YFP subunits (Fig. 7, B and C).
The specific rerouting of both wild-type 1 and 1-␤3 receptors triggered by P2X 2 receptors suggested that the intracellular domain of ␤ subunits was involved in molecular interaction between GABA A and P2X 2 receptors. However, significant differences were also noticed when we compared 1-␤3 to 1-wt. First, although 1-␤3 and P2X 2 -YFP were co-localized in cotransfected neurons, these areas were typically intracellular (Fig. 6), consistent with vesicular co-targeting. Second, the change in radial topology in the presence of P2X 2 -YFP was less pronounced when compared with 1-wt (Fig. 7C). This suggests that chimera modified the anchoring properties of the complex within neuron membranes. Data in Figs. 4 -7 demonstrate that the intracellular domain of the ␤ subunit is necessary for molecular and functional interaction between GABA A receptors and P2X channels. DISCUSSION In this study, we provide evidence in both Xenopus oocytes and transfected hippocampal neurons that molecular and functional interaction between P2X 2 and GABA A receptors is a receptor-mediated phenomenon, dependent on GABA A subunit composition.
Simultaneous application of ATP and GABA triggered an instantaneous current occlusion in oocytes co-expressing P2X 2 and GABA A receptors containing various ␣ and ␤ subunits, with or without ␥ subunits (see Table I). These results showed that P2X 2 channels interacted functionally with all major types of GABA A receptors. However, sequential applications of agonists revealed that the relative contribution of ATP or GABA receptors to current occlusion was dependent on the composition of the GABA receptors. Although ␣␤ or ␣␤␥ GABA A receptors inhibited P2X 2 channels, P2X 2 channel activation failed to inhibit GABA A receptors containing ␥2 or ␥3 subunits. Crossinhibition between P2X receptors and either nicotinic, 5-HT 3 , or 1/GABA C receptors was reciprocal (4 -7, 9, 12), whereas asymmetrical current occlusion was observed between GABA A and P2X or GABA and glycine receptors (10,11). We demonstrated that the occlusion direction was regulated by the subunit composition of the GABA A receptors, giving the first evidence of a specific cross-talk mechanism between two unrelated ligand-gated channels. The ␥2 subunit has been shown to modify the functional properties of GABA A receptors (26,27). The open state of GABA A receptors promoted by a ␥ subunit may decrease the ability of P2X receptors to occlude GABA-gated channels. It is noteworthy that the percentage of current occlusion when both agonists were applied together was maintained when the GABA A complex contained a ␥ subunit. These data suggest that the cross-inhibition between these channels is the result of an equilibrium between their open and closed states. Because ␥ subunits in GABA A also play a central role in receptor localization and clustering by interacting with associated proteins, such as GABARAP or GODZ (28,29), another possibility is that protein interactions promoted by this subunit stabilize the receptor complex or modify the coupling with P2X 2 and, consequently, the bidirectional cross-talk.
Current occlusion cross-talk observed between GABA A and P2X in dorsal root ganglia neurons was described as being chloride-and calcium-dependent (10). Our data showed that cross-talk between P2X 2 and GABA A could occur in the absence of extracellular calcium and at a range of holding potentials from Ϫ60 to ϩ30 mV, i.e. irrespective of the direction of chloride or calcium fluxes. Similar calcium and voltage independence was demonstrated for the cross-talk between P2X 2 and either nicotinic, 5-HT 3 , or GABA C receptors (4,7,9,12). Moreover, the use of cation-selective GABA A receptor-channels, generated by mutation of ␤3 subunits (20), did not prevent the functional cross-inhibition with P2X 2 receptors, clearly demonstrating that chloride is not involved in this coupling. However, cross-talk between other P2X subtypes and GABA A receptors has yet to be investigated.
Modulation of the state of receptor phosphorylation is responsible for the asymmetrical cross-talk between GABA A and glycine observed in spinal neurons (11). Although GABA A -and P2X-receptor functions are modulated by phosphorylation mechanisms (21,30), the unchanged current kinetics and im-FIG. 6. Differential co-localization of chimeric 1/GABA A receptors and P2X 2 channels in transfected hippocampal neurons. Hippocampal neurons were co-transfected with wild-type 1 (1-wt) or each chimera plus P2X 2 -YFP subunits. 1-wt and 1-␤3 co-localized with P2X 2 -YFP, whereas 1-␣2 and 1-␥2 did not. Anti-1/AlexaFluor 568 and YFP fluorescence profiles were measured on polylines drawn along the neurites. Note the overlap for 1-wt and 1-␤3 but not 1-␣2 or 1-␥2 with P2X 2 (bottom panels). Co-localization of 1-wt and P2X 2 -YFP occurred partially on surface clusters, whereas co-localization of 1-␤3 and P2X 2 -YFP was essentially intracellular. Representative images are shown. Scale bars: 10 m. mediate recovery of both currents after co-activation are incompatible with such mechanisms.
The functional independence between the C-terminal truncated form of P2X 2 receptors and ␣2␤3 GABA A receptors, as well as the suppression of cross-talk between wild-type P2X 2 and ␣2␤3 GABA A receptors in competition experiments with either the C-terminal domain (but not with the N-terminal) of P2X 2 or the intracellular loop (IL2) of ␤3 GABA A subunits demonstrated the involvement of intracellular domains in the functional interaction between these channels. Because 1/ GABA C receptors are homomeric complexes interacting functionally and physically with P2X 2 channels (12), we co-expressed P2X 2 and chimeric 1-GABA A channels with ␣2, ␤3, or ␥2 sequences downstream from TM3. This demonstrated that the cytoplasmic domain of the ␤3 subunit was indispensable for functional cross-talk, whereas that of ␣2 and ␥2 GABA A was unnecessary. Similarly, we observed a co-clustering of 1-␤3 chimeras (but not 1-␣2 or 1-␥2 chimeras) with P2X 2 in cotransfected hippocampal neurons and a modification of the 1-␤3 radial topology in the presence of P2X 2 subunits, indicating a physical association between the receptors. Thus, we propose that the physical and functional interactions between GABA A and P2X receptors specifically involve the large intracellular loop (IL2) of GABA A ␤ subunits and the C-terminal domain of P2X subunits.
These results and previous data suggest that a generic molecular mechanism underlies the functional coupling between ATP-gated channels and members of the nicotinic superfamily. Because the IL2 loop is not conserved among the nicotinic superfamily, the C-terminal domain of P2X subunits are of varying lengths and sequences, and the cytoplasmic domains of ligand-gated channels are involved in interactions with several binding partners, it is likely that specific protein complexes promote associations between these ligand-gated channels. Interacting complexes may propagate conformational changes induced by activation of one type of receptor to neighboring receptors of different types. This type of conformational spread has been proposed as a universal mechanism for signal integration (31). Molecular complexes implicated in the cell surface localization or anchoring of GABA A receptors remain unclear, even though the involvement of proteins such as gephyrin, tubulin, dystrophyn, and GABA receptor-associated protein has been demonstrated (28,(32)(33)(34). Recently, it has been suggested that a conserved cytoplasmic motif stabilizes surface FIG. 7. Modification of radial topology of homomeric 1 chimeras in the presence of P2X 2 -YFP receptors in transfected hippocampal neurons. A, determination of radial topology for P2X 2 -YFP and 1-wt subunits. Fluorescence profiles for YFP and anti-1/Alex-aFluor 568 were measured along polylines drawn on the longest neurites. Note that P2X 2 -YFP subunits were distributed distally to the cell body and found along the entire length of the neurites, whereas 1-wt subunits were usually found proximal to the cell body. B, determination of radial topology for 1 chimeras in the absence (top row) or presence (bottom row) of P2X 2 -YFP subunits. The presence of P2X 2 -YFP subunits changed the radial topology of 1-wt and 1-␤3, but not 1-␣2 and 1-␥2. C, cumulated fluorescence profiles measured as above, for 50 neurons in each group. From left to right: 1-wt, 1-␣2, 1-␤3, 1-␥2. Gray curves: without P2X 2 -YFP, black curves: with P2X 2 -YFP. The average radial trafficking was increased by the presence of P2X2-YFP for 1-wt and 1-␤3 but not for 1-␣2 and 1-␥2.
GABA and ATP are co-released in the spinal cord and brain, and their associated receptors are co-localized in the postsynaptic density (13,14), thus providing the physiological conditions for cross-talk between GABA A and P2X receptors. Furthermore, postsynaptic ATP/GABA transmission examined at a potential between the reversal potentials of both receptors revealed a lack of mixed excitatory and inhibitory currents in lateral hypothalamus synapses (14). Non-additive ATP and GABA currents observed in these neurons are consistent with negative cross-talk between GABA A and P2X receptors described in this work.
Synaptic GABA A receptors are composed of ␣␤␥ subunits, whereas neuronal ATP-gated channels are thought to be homoor heteromeric associations of P2X 2 , P2X 4 , and P2X 6 subunits (18). The finding that a ␥ subunit within GABA A receptors prevents their inhibition by the activation of P2X receptors indicates that cross-talk may have a stronger effect on the ATP component of mixed ATP/GABA synapses in the dorsal horn of the spinal cord, as well as in the lateral hypothalamus. Reciprocal cross-talk may also regulate the functioning of extrasynaptic P2X 2 and GABA A receptors thought to be composed of ␣␤ or ␣␤␦ subunits (36,37), which may be co-activated by ATP and GABA released from neighboring astrocytes (38 -40). Both GABA A and P2X receptors have also presynaptic roles, as shown in afferent terminals of dorsal root ganglia neurons (41). ATP facilitates glutamate release (13,42), whereas GABA produces a presynaptic inhibition (43). Cross-inhibition between ATP-and GABA-gated channels may modify the balance between excitation and inhibition and, consequently, contribute to sensory information processing. Taken together with other published data, our findings showed that cross-talk between separate ligand-gated channels is a widespread mechanism, regulated by their molecular complexity, that profoundly influences receptor functions.