Transmembrane AMPA Receptor Regulatory Proteins and Cornichon-2 Allosterically Regulate AMPA Receptor Antagonists and Potentiators*

AMPA receptors mediate fast excitatory transmission in the brain. Neuronal AMPA receptors comprise GluA pore-forming principal subunits and can associate with multiple modulatory components, including transmembrane AMPA receptor regulatory proteins (TARPs) and CNIHs (cornichons). AMPA receptor potentiators and non-competitive antagonists represent potential targets for a variety of neuropsychiatric disorders. Previous studies showed that the AMPA receptor antagonist GYKI-53655 displaces binding of a potentiator from brain receptors but not from recombinant GluA subunits. Here, we asked whether AMPA receptor modulatory subunits might resolve this discrepancy. We find that the cerebellar TARP, stargazin (γ-2), enhances the binding affinity of the AMPA receptor potentiator [3H]-LY450295 and confers sensitivity to displacement by non-competitive antagonists. In cerebellar membranes from stargazer mice, [3H]-LY450295 binding is reduced and relatively resistant to displacement by non-competitive antagonists. Coexpression of AMPA receptors with CNIH-2, which is expressed in the hippocampus and at low levels in the cerebellar Purkinje neurons, confers partial sensitivity of [3H]-LY450295 potentiator binding to displacement by non-competitive antagonists. Autoradiography of [3H]-LY450295 binding to stargazer and γ-8-deficient mouse brain sections, demonstrates that TARPs regulate the pharmacology of allosteric AMPA potentiators and antagonists in the cerebellum and hippocampus, respectively. These studies demonstrate that accessory proteins define AMPA receptor pharmacology by functionally linking allosteric AMPA receptor potentiator and antagonist sites.

AMPA receptors play fundamental roles in controlling behavior, learning, and memory; dysfunction of these receptors likely underlies a variety of neuropsychiatric disorders (15)(16)(17)(18)(19)(20)(21). Accordingly, numerous pharmacological efforts have sought either to promote AMPA receptor function with potentiators or to block channel function with antagonists. AMPA receptor potentiators enhance channel permeation by either slowing deactivation (channel closure following glutamate removal) or blunting desensitization (channel closure in the continued presence of glutamate) (22). AMPA potentiators can enhance synaptic transmission and thereby promote growth factor release and neurogenesis (23)(24)(25). Preclinically, AMPA potentiators have shown promising activity in models of depression (26), cognitive impairment (27), and Parkinson disease (28). AMPA receptor antagonists that blunt synaptic transmission have been pursued as therapeutics for epilepsy, neurodegeneration, and neuropathic pain. Multiple classes of AMPA receptor antagonists have been identified. Antagonists, such as 6-cyano-7-nitroquinoxaline-2,3-dione, compete with glutamate at the agonist binding site and occlude channel opening (29). Noncompetitive antagonists, such as GYKI-53655, reportedly interact at an alternative site to allosteric potentiators and inhibit channel gating downstream of glutamate binding (30,31).
TARPs have complex interactions with AMPA receptor modulator drugs. TARPs convert certain competitive antagonists, including 6-cyano-7-nitroquinoxaline-2,3-dione, into partial agonists (32). By increasing glutamate affinity, TARPs also blunt effects of competitive antagonists (33). By contrast, TARP ␥-2 increases the affinity of certain non-competitive antagonists, such as GYKI-53655 (33). The molecular mechanism for channel block by non-competitive antagonists remains unclear (31), and resolving this may provide insights for mechanisms that underlie channel gating.
Interactions between AMPA receptors potentiators and antagonists have also yielded valuable clues regarding their mechanisms of action. The IC 50 for GKYI-53655 is shifted 10-fold to right by cyclothiazide (34). Although initially interpreted as evidence for direct interaction between these sites, subsequent studies showed that cyclothiazide cannot displace a radiotracer from the GYKI-53655 binding site (9). Additional insight was provided by binding studies with [ 3 H]-LY395153, which potently labels the potentiator site (35). Interestingly, GYKI-53655 potently blocks binding of [ 3 H]-LY395153 to AMPA receptors in brain membranes but does not affect binding to recombinant receptors (35). Here, we addressed this discrepancy and discovered an unexpected role for AMPA receptor accessory proteins in functionally linking AMPA receptor potentiator and non-competitive antagonist binding sites. These results provide valuable insights for understanding and refining the neuropharmacology of AMPA receptors.

EXPERIMENTAL PROCEDURES
Buffers, Reagents, Plasmids, and Cell Culture-All buffers and reagents were purchased from either Sigma-Aldrich or Thermo Fisher Scientific (Pittsburgh, PA). All cDNAs were human, except for rat iGluA2R, and were cloned into pcDNA 3.1 mammalian expression plasmids (Invitrogen). Compounds used in binding assays were synthesized at Lilly Research Laboratories (Indianapolis, IN). For electrophysiology experiments, HEK293T cells were maintained at 37°C in 5% CO 2 high glucose DMEM medium supplemented with 10% fetal calf serum and 1% penicillin-streptomycin and split bi-or triweekly. HEK293T cells were plated onto 25-mm coverslips and were transiently transfected using FuGENE 6 according to the manufacturer's protocols (11814443001; Roche Applied Science). Experiments were conducted 48 -72 h post-transfection.
Tissue and Recombinant Cell Membrane Preparation-Stargazer and wild-type mice were euthanized with CO 2 and decapitated. Brains were dissected rapidly and homogenized using a Polytron in 10 volumes of ice-cold 50 mM Tris-HCl buffer (pH 7.4). Cells were pelleted and homogenized using a Polytron in 10 volumes of ice-cold 50 mM Tris-HCl buffer (pH 7.4). Homogenates were centrifuged at 1000 g to remove nuclei and unbroken cells. Both tissue and cell homogenates were centrifuged again at 4°C at 38,000 ϫ g for 20 min. To remove endogenous glutamate, pellets were resuspended, washed with buffer, and centrifuged for 20 min. This process was repeated a total of four times. After the final wash, pellets were frozen on solid CO 2 and stored at Ϫ80°C.
Radioligand Binding-Membranes were incubated with 50 nM [ 3 H]-LY450295 (ViTrax Radiochemicals, Placentia, CA) and other pharmaceutical agents as indicated for 2 h at 4°C. Assay buffer comprised 50 mM Tris-HCl (pH 7.4) and 500 M L-glutamate (Tocris Bioscience, Ellisville, MO). Nonspecific binding was determined by including 10 M LY450108, a related AMPA receptor potentiator (36). All binding was terminated by rapid filtration using a TOMTEC 96-well cell harvester (Hamden, CT) through GF/A filters presoaked with 0.3% polyethyleneimine. The filters were washed with 5 ml of icecold 50 mM Tris buffer (pH 7.4) and air-dried overnight. The dried filters were placed on PerkinElmer Life Sciences MeltiLex A melt-on scintillator sheets, and the radioactivity was counted using a PerkinElmer Life Sciences Wallac 1205 Betaplate counter (Perkin Elmer Life Sciences). For binding studies, homomeric GluA transfections were used to ensure a uniform receptor composition. GluA2 was selected for binding studies due to its inclusion in most hippocampal (GluA1/GluA2 heteromeric) and cerebellar neuronal (GluA2/GluA3 and GluA2/GluA4 heteromeric) AMPA receptors (37). In some experiments, experimental variability caused binding to exceed 100% of control.
Transfected HEK293T cells were lifted and perfused with a 16-barrel glass capillary pipette placed 100 -200 m from the cell (VitroCom, Mountain Lakes, NJ). Solutions were switched by sliding the pipette array with a solution exchange rate of Ͻ20 ms. Glutamate (1 mM), LY450295 (10 M), and GYKI-53655 (10 M) were applied where indicated. Because of the low conductance from homomeric GluA2 receptors used in binding studies (39,40), we transfected HEK293T cells with GluA1 ϩ GluA2 in the presence or absence of ␥-2. GluA2 incorporation into heteromeric complexes with GluA1 was confirmed via a linear I-V curve from Ϫ80 to ϩ 80 mV. Preincubation of the potentiator, empirically determined to maximize intercell potentiation reliability, was for a period of 1 min followed by 30-s pulses of agonist in the presence of compounds.
Autoradiography-Sagittal brain sections were cut at 12 m, thaw mounted onto gelatin-coated slides, and stored at Ϫ80°C. Sections were incubated for 2 h in 50 mM Tris-HCl containing 50 nM [ 3 H]-LY450295, 500 M L-glutamate, and other agents as indicated. Sections were rinsed with 50 mM ice-cold Tris-HCl for 10 min, dried, and exposed to a Fujifilm Imaging Plate for 15 days.
Data Analysis and Statistics-Radioligand binding studies were analyzed using a Microsoft Excel TM workbook and were graphed using GraphPad Prism software (La Jolla, CA). The electrophysiology data are represented as mean Ϯ S.E. and were the result of at least three independent experiments. Analyses involving three or more data sets were performed with a oneway analysis of variance with a Tukey Kramer post hoc analysis using GraphPad Prism software. Analyses involving two data sets were performed with an uncorrected Student's t test or with a Student's t test with a Welsh correction, only if the variances were statistically different. Percent inhibition was calculated as, where I Glu-SS is the glutamate-evoked steady state current in LY450295 and GYKI-53655 or just LY450295. Significance was set as a p value of Ͻ0.05. B 0 refers to the specific potentiator binding, which occurs in the absence of any added antagonist.

Non-competitive Antagonists Displace [ 3 H]-LY450295 Binding from Brain but Not from Recombinant AMPA Receptors-
Previous studies characterized binding of AMPA receptor potentiator [ 3 H]-LY395153 both to brain membranes and to recombinantly expressed GluA subunits (35). Curiously, the non-competitive antagonist, GYKI-53655 potently displaced [ 3 H]-LY395153 binding to cerebrocortical membranes but did not displace binding from transfected GluA4 membranes (35). We first asked whether this discrepancy was particular to a specific brain region or GluA subunit. Using GluA2 because of its predominant incorporation into cerebellar neuronal AMPA receptors (37), we observed that GYKI-53655 readily displaced the AMPA potentiator, [ 3 H]-LY450295, from cerebellar membranes but did not displace [ 3 H]-LY450295 binding from transfected GluA2 membranes (Fig. 1A). Furthermore, a structurally distinct non-competitive AMPA receptor antagonist, CP-465,022, showed a similar selectivity for displacement of potentiator only from native tissues (Fig. 1B). Because GluA2 is incorporated into AMPA receptors across multiple brain regions, we focused on this GluA subunit (37).
Autoradiography Demonstrates TARP-mediated Sensitivity of [ 3 H]-LY450295 Binding-␥-2 is the predominant TARP subunit in the cerebellum, whereas ␥-8 predominates in hip-  pocampus (48). We used autoradiography to visualize [ 3 H]-LY450295 binding throughout the brain. Sagittal sections were incubated with ϳ50 nM [ 3 H]-LY450295, and co-application of unlabeled LY450108 served as a measure of non-specific binding (Fig. 7, A and D). In wild-type mice, [ 3 H]-LY450295 binding was present in diverse brain regions with the highest levels occurring in the hippocampus and cerebellum (Fig. 7A). The non-competitive antagonist CP-465,022 reduced [ 3 H]-LY450295 binding in both brain regions (Fig. 7, A and E).
In stargazer, [ 3 H]-LY450295 binding is reduced substantially in the cerebellum and reduced modestly in the hip-pocampus (Fig. 7, B and D). In ␥-8 knock-out, [ 3 H]-LY450295 binding was reduced significantly in hippocampus but not in cerebellum (Fig. 7, C and D). Importantly, the [ 3 H]-LY450295 binding that remained in the cerebellum of stargazer mice was not displaced by CP-465,022 (Fig. 7E). In the ␥-8 knock-out hippocampus residual [ 3 H]-LY450295 was partially resistant to displacement by the non-competitive antagonist (Fig. 7E). Together, these findings establish that TARPs and accessory proteins such as CNIH-2 allosterically regulate AMPA receptor pharmacology in specific brain regions.

DISCUSSION
The principal finding of this study is that TARP subunits functionally link potentiator and antagonist binding sites within AMPA receptors. As published previously (35), we found that GYKI-53655 displaces AMPA potentiators from native AMPA receptors but not from recombinant receptors containing only a GluA subunit. We resolved this paradox by observing that TARPs ␥-2/␥-8, and to a lesser extent CNIH-2, confer sensitivity of AMPA receptor potentiator binding to displacement by GYKI-53655 and another non-competitive AMPA receptor antagonist. Although GYKI-53655 did not dis-  1960 versus ϳ 1470 fmol/mg protein, respectively) and reduced binding affinity (K d ϳ 32 nM versus ϳ 60 nM, respectively) relative to wild-type (wild-type CB, filled circles). Non-competitive antagonists GYKI-53655 (B) and CP-465,022 (C) more weakly displace [ 3 H]-LY450295 binding from stargazer cerebellar membranes as compared with wild-type cerebellar membranes (GYKI, ϳ46% versus ϳ77% displaced, respectively; CP, ϳ42% versus ϳ72% displaced, respectively). Data are presented as mean Ϯ S.E. place [ 3 H]-LY450295 in recombinant ␥-2-lacking AMPA receptors, it nonetheless inhibited LY450295 potentiated currents from such receptors. Furthermore, ␥-2 increases the affinity of both AMPA potentiators and antagonists (7-10, 33). Using receptor autoradiography and transgenic stargazer or ␥-8 knock-out mice, we visualized the essential roles for these TARPs in controlling AMPA receptor pharmacology across the brain.
Structural studies provide insight regarding regulation of AMPA receptors by pharmacological agents and auxiliary subunits. GluA subunits are three-pass transmembrane proteins and the ligand binding domain (LBD) comprises amino acid residues from the extracellular N terminus and the loop between the second and third transmembrane domains. Structural studies of the LBD (49) and more recently of the GluA2 tetramer (50) have provided insight regarding molecular mechanisms for AMPA receptor gating. The tetrameric complex is assembled in a dimer-of-dimer conformation. Agonist binding induces closure of the clam-shaped LBD, which leads to sepa-ration of the attached transmembrane helices and then to channel opening (51). This is followed rapidly by rearrangement of the LBD dimer interface and channel desensitization. AMPA receptor potentiators bind at the interface of LBD dimers and prevent the conformational changes that cause desensitization (52).
To test how TARPs exert their actions on AMPA potentiator binding, we measured the potentiator affinity. Our results showed that ␥-2 increased potency of AMPA potentiator binding, indicating allosteric modulation. AMPA potentiators blunt receptor desensitization by binding to and stabilizing the interface between dimer pairs of the LBD (52). TARPs also blunt channel desensitization, and this is thought to occur through interactions of the first extracellular loop of TARP with the LBD (58). Our data support this model and suggest that TARPs facilitate dimerization of LBDs and therefore enhance the binding affinity of AMPA potentiators.
AMPA receptor antagonists also have been valuable tools to understand receptor gating mechanisms. Pharmacological studies have identified non-competitive antagonists including the 2,3-benzodiazepines, such as GYKI-53655 (59), and quinazolinones, such as CP-465,022 (60). Radioligand binding indicates that the 2,3-benzodiazepines and the quinazolinones interact with overlapping sites on the receptor; however, their binding site remains uncertain. Blocking receptor desensitization by cyclothiazide (30) or by mutating GluA1 leucine 497 to a tyrosine (31) decreases the affinity of GYKI-53655. Also, GYKI-53655 dissociates more rapidly from activated than from non-activated receptors suggesting that GYKI-53655 binds preferentially to a closed state of the AMPA receptor (31).
A domain-swapping and site-directed mutagenesis strategy showed that residues between LBD and the channel transmembrane domains are essential for antagonism by GYKI-53655. This region therefore was interpreted to be the binding site for non-competitive antagonists (31). This finding suggested a model whereby conformational changes in the regions that link the LBD to the transmembrane helices during channel opening distort the binding site for GYKI-53655 and reduce its affinity. This model was questioned by studies showing that co-transfection with ␥-2 restored GYKI-53655 inhibition of the GYKI-53655-"insensitive" GluA1 variant (33). Indeed, ␥-2 increases inhibitory potency of GYKI-53655 on non-mutated GluA1 (33). As TARPs stabilize the open non-desensitized state, the ␥-2mediated increase in GYKI-53655 affinity would not have been predicted by the previous model.
Our studies showed that displacement of AMPA potentiator [ 3 H]-LY450295 binding by either GYKI-53655 or CP-465,022 requires a TARP or, to a lesser extent, a CNIH-2 subunit within the receptor complex. These data resolved a paradox from previous studies showing that AMPA potentiator binding to native AMPA receptors but not recombinant GluA4 alone is sensitive to inhibition by GYKI-53655 (35). Our data also showed that TARPs enhance affinity of [ 3 H]-LY450295 binding. As predicted by these results, we found a reduction in [ 3 H]-LY450295 binding to cerebellar membranes from stargazer mice. The residual binding was insensitive to displacement by GYKI-53655 with the residual displacement likely resulting from other TARPs. We further characterized [ 3 H]-LY450295 phar-macology across mouse brain by autoradiography. In ␥-8 knock-out mice, we found a dramatic loss of [ 3 H]-LY450295 binding in the hippocampus, which fits with the regional distribution for ␥-8 (61,62). We also found that the residual hippocampal binding of [ 3 H]-LY450295 in ␥-8 knock-outs exhibited somewhat reduced displacement by a non-competitive antagonist. Robust hippocampal CNIH-2 expression and/or alternative TARP likely accounts for the residual antagonist activity.
Our studies provide key insights regarding the functional interactions of TARPs with AMPA receptors. It had been shown previously that the first extracellular loop of ␥-2 controls AMPA receptor gating, whereas the C terminus is important for surface and synaptic trafficking (8,63). The first extracellular domain of ␥-2 also mediates the increase in GYKI-53655 affinity (33). Furthermore, ␥-2 can change the conformation of the linker between the ligand binding core and the transmembrane domain of AMPA receptors to restore GYKI-53655 antagonism for a receptor with mutations in this linker region (33). Our findings support the hypothesis that TARPs enable non-competitive antagonists to block potentiator binding, suggesting that TARPs propagate perturbation of the juxtamembrane linker region to the LBD. Promoting conformational communication between these regions of AMPA receptors may explain how TARPs regulate AMPA receptor pharmacology to enhance channel gating. Future structural studies of AMPA receptors in complex with TARP and other auxiliary subunits will clarify the nature of these molecular interactions.
Either augmenting or antagonizing AMPA receptor function represents intriguing pharmacological approaches for a variety of neuropsychiatric disorders. However, the ubiquity of neuronal AMPA receptors suggests that subtype selectivity may be desired for modulating transmission in specific pathways. Understanding how auxiliary subunits modify AMPA receptor pharmacology may help design more selective agents for targeting AMPA receptors.