Article Information

• PDF (563 kb)
• Full Text with Large Figures
• Cited by in Scopus (4)

PubMed

• Articles by Nicholas A. Mitchell
• Articles by Mark W. Fleck

Related Articles

• TARP Subtypes Differentially and Dose-Dependently Control Sy...

  TARP Subtypes Differentially and Dose-Dependently Control Synaptic AMPA Receptor Gating Neuron, Volume 55, Issue 6, 20 September 2007, Pages 905-918 Aaron D. Milstein, Wei Zhou, Siavash Karimzadegan, David S. Bredt and Roger A. Nicoll Summary A family of transmembrane AMPA receptor regulatory proteins (TARPs) profoundly affects the trafficking and gating of AMPA receptors (AMPARs). Although TARP subtypes are differentially expressed throughout the CNS, it is unclear whether this imparts functional diversity to AMPARs in distinct neuronal populations. Here, we examine the effects of each TARP subtype on the kinetics of AMPAR gating in heterologous cells and in neurons. We report a striking heterogeneity in the effects of TARP subtypes on AMPAR deactivation and desensitization, which we demonstrate controls the time course of synaptic transmission. In addition, we find that some TARP subtypes dramatically slow AMPAR activation kinetics. Synaptic AMPAR kinetics also depend on TARP expression level, suggesting a variable TARP/AMPAR stoichiometry. Analysis of quantal synaptic transmission in a TARP γ-4 knockout (KO) mouse corroborates our expression data and demonstrates that TARP subtype-specific gating of AMPARs contributes to the kinetics of native AMPARs at central synapses. Summary | Full Text | PDF (1714 kb)

• Arc/Arg3.1 Mediates Homeostatic Synaptic Scaling of AMPA Rec...

  Arc/Arg3.1 Mediates Homeostatic Synaptic Scaling of AMPA Receptors Neuron, Volume 52, Issue 3, 9 November 2006, Pages 475-484 Jason D. Shepherd, Gavin Rumbaugh, Jing Wu, Shoaib Chowdhury, Niels Plath, Dietmar Kuhl, Richard L. Huganir and Paul F. Worley Summary Homeostatic plasticity may compensate for Hebbian forms of synaptic plasticity, such as long-term potentiation (LTP) and depression (LTD), by scaling neuronal output without changing the relative strength of individual synapses. This delicate balance between neuronal output and distributed synaptic weight may be necessary for maintaining efficient encoding of information across neuronal networks. Here, we demonstrate that Arc/Arg3.1, an immediate-early gene (IEG) that is rapidly induced by neuronal activity associated with information encoding in the brain, mediates homeostatic synaptic scaling of AMPA type glutamate receptors (AMPARs) via its ability to activate a novel and selective AMPAR endocytic pathway. High levels of Arc/Arg3.1 block the homeostatic increases in AMPAR function induced by chronic neuronal inactivity. Conversely, loss of Arc/Arg3.1 results in increased AMPAR function and abolishes homeostatic scaling of AMPARs. These observations, together with evidence that Arc/Arg3.1 is required for memory consolidation, reveal the importance of Arc/Arg3.1's dynamic expression as it exerts continuous and precise control over synaptic strength and cellular excitability. Summary | Full Text | PDF (2146 kb)

• Postsynaptically Silent Synapses in Single Neuron Cultures

  Postsynaptically Silent Synapses in Single Neuron Cultures Neuron, Volume 21, Issue 6, 1 December 1998, Pages 1443-1451 Stephen N Gomperts, Anuradha Rao, Ann Marie Craig, Robert C Malenka and Roger A Nicoll Summary We have used the synapses that isolated hippocampal cells in culture form onto themselves (autapses) to determine if some synapses lack functional AMPA receptors (AMPARs). A comparison of the synaptic variability of the AMPAR- and NMDAR-mediated evoked responses, as well as of miniature synaptic responses, indicates that a population of events exists that only contains an NMDAR component. Spillover of glutamate from adjacent synapses cannot explain these results because in single cell cultures all synaptic events mediated by AMPARs should be detected. Immunocytochemical analysis of these cultures clearly reveals a population of synapses with puncta for NR1 (NMDAR) but not for GluR1 (AMPAR). These results provide strong anatomical and physiological evidence for the existence of postsynaptically silent synapses. Summary | Full Text | PDF (315 kb)

• …more

Copyright © 2007 The Biophysical Society. All rights reserved.
Biophysical Journal, Volume 92, Issue 7, 2392-2402, 1 April 2007

doi:10.1529/biophysj.106.095091

Channels, Receptors, and Electrical Signaling

Targeting AMPA Receptor Gating Processes with Allosteric Modulators and Mutations

Nicholas A. Mitchell and Mark W. Fleck^,  

Center for Neuropharmacology & Neuroscience, Albany Medical College, Albany, New York 12208

Address reprint requests to Mark W. Fleck, Ph.D., Center for Neuropharmacology & Neuroscience, Albany Medical College, MC-136, 47 New Scotland Avenue, Albany, NY 12208. Tel.: 518-262-6534; Fax: 518-262-6534.

AMPA-type glutamate receptors (AMPARs) are ligand-gated ion channels that represent the principal mediators of fast excitatory synaptic transmission in the mammalian CNS. The dysfunction of AMPARs has been linked to Alzheimer’s disease, depression, epilepsy, stroke, and neurodegenerative disease 1,2. Their implication in these various neuropathologies makes them attractive targets for the rational design and development of novel therapeutics.

AMPAR-mediated synaptic currents are exceedingly brief (∼1–2ms), reflecting the transient elevation of glutamate in the synaptic cleft 3,4 and rapid inactivation of AMPARs by the combined processes of deactivation and desensitization 5,6,7. Deactivation is measured by the decay of current after the removal of agonist, which follows an exponential time course having a time constant (τ_deact) of 0.5–2ms. Desensitization is measured by the decay of current in the continuous presence of agonist and has a somewhat slower time constant (τ_des) on the order of 1–10ms 5,6,8. Both desensitization and deactivation are empirical measures of current decay from the open state, and so both measures necessarily involve the rate of channel closing (α). Beyond that, it remains unclear to what extent the underlying rates of agonist dissociation (k_off) and isomerization to the desensitized conformation (k_des) are truly independent of one another. Nonetheless, because these rates collectively govern the magnitude and time course of synaptic transmission, an objective in the development of novel AMPAR modulators has been to target these processes selectively.

The molecular mechanisms of AMPAR gating and desensitization are the subject of intense investigation. Structural studies suggest that glutamate binds within an agonist-binding clamshell 9. The binding domains are linked to one another in a back-to-back dimer configuration 10,11 and held together in part by a salt-bridge hydrogen bond network at the dimer interface 12. AMPAR activation involves closure of the upper and lower lobes of the binding pocket around the agonist, which pulls against the rigid dimer interface to cause channel opening by rotational rearrangement of the transmembrane domains 13. Deactivation is thought to represent simply the reverse of this process, whereas desensitization is related to instability of the dimer interface contacts that are necessary to link the agonist-binding and transmembrane domains 11,12,13,14.

Allosteric modulators bind remotely but influence the agonist binding site or agonist-induced conformational rearrangements associated with channel gating 7,15,16,17,18,19,20,21. Interest in AMPAR allosteric modulators began in earnest with the discovery that cyclothiazide (CTZ) potentiates AMPAR currents in hippocampal neurons by slowing desensitization 21,22. CTZ binds within and stabilizes the glutamate receptor (GluR) dimer interface 11 and owes its preference for AMPAR-flip isoforms 7,22,23 to a single serine residue (S750) in the flip/flop domain 7. In addition to slowing desensitization, CTZ also slows AMPAR deactivation 7 and increases the apparent affinity for activation by agonist 24. These secondary actions of CTZ might be an indirect consequence of slowing desensitization or a direct effect of CTZ on agonist dissociation. The question is complicated to the extent that desensitization contributes to deactivation and limits agonist potency by preventing channel activation at low agonist concentrations 25,26.

To answer this question we used a combinatorial approach that employed a “nondesensitizing” AMPAR. Similar to CTZ, mutation of a leucine to a tyrosine at position 497 of GluR1 27 (GluR2-L483Y 28) greatly reduces AMPAR desensitization. The GluR1-L497Y receptor allowed us to discriminate between drug effects that were a consequence of blocking desensitization and those effects that were unique and separable from desensitization. The actions of a related benzothiadiazide, trichlormethiazide (TCM), and an unrelated Ampakine, 2H,3H,6aH-pyrrolidino[2′,1′-3′,2′]1,3-oxazino[6′,5′-5,4]benzo[e]1,4dioxan-10-one (CX614), were also examined. GluR1-wt and GluR1-L497Y receptors were expressed in HEK 293 cells, and drug effects on measures of deactivation, desensitization, and agonist EC50 were determined using ultrafast solution exchange. An additional mutation, GluR1-S750Q 7, was used to confirm benzothiadiazide binding at the dimer interface and resulted in identification of S750 as a pivotal linkage that connects the L497Y mutation to the agonist-binding pocket. Results indicate that desensitization and deactivation can be selectively targeted, at least by site-directed mutagenesis, which suggests that future modulators may also be capable of such selectivity. Actions of modulators alone, and interactions between modulators and mutations on channel gating, are characterized in an effort to offer potential strategies for targeting specific AMPAR gating transitions.

Ultrapure salts and chemicals were purchased from Sigma (St. Louis, MO).

Human Embryonic Kidney 293 fibroblasts (HEK293, CRL, ATCC, Manassas, VA) were cultured in minimal essential medium (MEM) supplemented with 10% fetal bovine serum and 2mM Glutamax (Life Technologies, Rockville, MD). Cells were incubated at 37°C in a 5% CO₂ environment. Cells were plated into 25cm² Falcon flasks and passed twice weekly to fresh flasks without reaching confluence.

Cells intended for transfection and subsequent recording were plated to poly-d-lysine–coated 35-mm NUNC dishes at a density of 80,000cells/ml and transfected the following day using Lipofectamine 2000 reagents (Life Technologies). Cells were cotransfected with EGFP at a 9:1 ratio. cDNA plasmids containing GluR1, GluR1-L497Y, GluR1-S750Q, or GluR1-L497Y/S750Q were combined with Lipofectamine reagents for a final concentration of 1μg/μl cDNA per 35-mm NUNC dish. Transfected cells were incubated for 36–72h before use for electrophysiological recordings.

Cells were continuously superfused with a standard extracellular solution containing (in mM): 150 NaCl, 3 KCl, 5 HEPES, 1 MgCl₂, 1.8 CaCl₂, 10 glucose, and 0.1 mg/ml phenol red, with an adjusted pH of 7.3. Transfected cells were identified by fluorescent expression of EGFP. Recording microelectrodes were fabricated from thin-walled borosilicate glass capillary tubes (TW150, World Precision Instruments, Sarasota, FL). Electrode open-tip resistance was typically 2–4MΩ when filled with an intracellular solution comprised of (in mM): 135 CsCl, 10 CsF, 10 HEPES, 5 EGTA, 1 MgCl₂, 0.5 CaCl₂, pH 7.2, and 295mOsm. Outside-out patch recordings were performed in voltage-clamp mode at a holding potential of −70mV using an Axopatch 200B amplifier (Axon Instruments, Foster City, CA). Current signals were filtered at 2–5kHz with an eight-pole Bessel filter (Cygnus Technologies, Delaware Watergap, PA), digitized at 20kHz, and stored on a Macintosh PowerPC-G3 computer using an ITC-16 interface (Instrutech, Port Washington, NY) under the control of the data acquisition and analysis program Synapse (Synergy Research, Silver Spring, MD). All recordings were performed at room temperature (20–22°C).

All recordings were conducted in the outside-out patch configuration. Ultrafast solution exchange was achieved by using an LSS-3100 piezotranslator (Burleigh Instruments, Fisher, NY). Control and agonist solutions were driven simultaneously at a rate of 0.25ml/min through two parallel barrels of a θ-tube having a tip diameter ∼200μm. Allosteric modulators were preapplied through the control barrel. Membrane patches were positioned in the control stream near the solution interface, and a piezotranslator was used to rapidly move the θ-tube ∼100μm such that the solution interface passed over the patch. The timing and rate of solution exchange were determined by open-tip junction currents at the conclusion of all recordings and were typically <200μs (10–90% rise time); these were based on current deflections produced by a 5% change in NaCl concentration. Representative figure traces as well as those used for kinetic analyses were averages of 5–10 consecutive responses. Time constants for deactivation (τ_deact) and desensitization (τ_des) were derived from one or two exponential fits as required using a least-squares fitting algorithm. Current decays were fit from 60% to 95% of peak to steady state. Glutamate was prepared as 1M stock solution at neutral pH and added to extracellular solutions by serial dilutions as required. CTZ, TCM, and CX614 were prepared in DMSO and diluted in external solution to final concentrations of 100μM, 500μM, and 100μM, respectively (final DMSO concentration 0.5%). DMSO controls up to 10 times the concentrations present in drug solutions did not alter deactivation or desensitization kinetics in the absence of drug.

Mean EC50 and kinetic measures of deactivation and desensitization were compared using a one-factor ANOVA. Post-hoc analyses employed two-tailed Student’s t-tests with Bonferroni correction (where applicable). Paired t-tests were employed in some cases to compare pre- and postdrug measures from the same patches. Statistical significance levels are indicated in the figure legends. EC50 values were determined from best-fits of the logistic equation I=I_max/(1+(EC50/(GLU))^nH), where I is the current at a given agonist concentration, I_max is the maximal current at saturation, EC50 is the glutamate concentration giving half-maximal current, and nH is the Hill coefficient. A least-squares algorithm from KaleidaGraph software was used to generate curve fits (Synergy Software, Reading, PA). Best-fitting nH values were generally near unity (1.2–1.4). The number of independent observations from different patches is indicated (n) within figures.

Computational models were constructed, and simulations were tested, using AxoGraph software (Axon Instruments). Several kinetic schemes have been used to model AMPAR behavior 7,25,29, and the underlying assumptions of modeling AMPAR behavior have been well described 29,30. We adopted the 4-site scheme of Robert and Howe 29, which was simplified by excluding secondary desensitized states. Briefly, the model assumes that 1), AMPARs are tetrameric channels 2,31,32 that possess four agonist binding sites 9,10 and exhibit three intermediate conductance levels 31,33, 2), agonist binding initiates either channel activation or desensitization, and 3), the decay during prolonged agonist pulses represents receptor entry into desensitized states. The parameters of the models were optimized to simultaneously reproduce the values of activation (τ_act), deactivation (τ_deact), desensitization (τ_des), and agonist potency (EC50) that were observed experimentally for wild-type, modulator-bound, and mutant receptors; the latter required iterative changes to one or more rate constants: k_on/k_off, β/α, or k_des/k_res. Initially, rate constants were modified individually, and the effects of these changes on channel behavior were recorded. Subsequent modifications to multiple rate constants were then combined in an iterative manner to arrive at an optimal solution (see Table 2). Simpler two-site kinetic schemes required similar solutions except that the changes were quantitatively different.

Many effects of CTZ and the L497Y mutation on GluR function are documented, but it is difficult to compare these various results from different preparations, methods, and laboratories 7,22,23,27,34,35. Specifically, we sought to compare (quantitatively) the effects of CTZ and L497Y mutation on desensitization, deactivation, and agonist potency (EC50) in GluR1 AMPARs. Our initial experiments therefore replicated findings from other sources not only to confirm the findings but also to make them quantitatively comparable for subsequent studies (Fig. 1). Consistent with previous reports, homomeric wild-type GluR1-flip channels in outside-out membrane patches deactivated in <1ms (Figure 1C) and desensitized to <1% of peak within 2–3ms (Figure 1A). Preapplication of 100μM CTZ or insertion of a leucine-to-tyrosine (L/Y) mutation at position 497 of the mature protein prevented desensitization during 50-ms pulses of 10mM glutamate. Prolonged exposure to glutamate for up to 2s typically produced <25% decay to steady state in GluR1+CTZ and L/Y mutant receptors (Figure 1B, Table 1). Although desensitization is slowed extensively by CTZ and the L497Y mutation, as indicated in Fig. 1, receptors did still desensitize (see Table 1 for values). Differences in the time course of deactivation were obvious from the relaxation kinetics after glutamate removal; decay of currents for GluR1+CTZ appeared much faster than GluR1-L497Y. To characterize deactivation kinetics more precisely, and for comparison to wild-type GluR1, brief 1-ms exposure and rapid removal of 10mM glutamate were used to minimize wild-type desensitization and to focus on the faster processes of channel closing (α) and agonist dissociation (k_off). These effects are depicted in Figure 1C and summarized in Table 1. Deactivation time constants (τ_deact) were significantly slower than wild type in both cases (P<0.001 unpaired Student’s t-test), but more so for the L/Y mutant than for GluR1+CTZ. Preincubation of GluR1 receptors with 100μM CTZ slowed receptor deactivation by 2.4-fold, whereas the L497Y mutation slowed GluR1 deactivation by more than 13-fold compared to wild type. These values are similar to those reported elsewhere 7,34. Activation rates were evaluated by comparing the current onset time constants (τ_act) of receptors at glutamate concentrations ranging from 0.3 to 3mM. As expected, activation rates were concentration dependent; the relation between concentration and activation was not different between groups, indicating that rates of agonist binding (k_on) and channel opening (β) were not modified by these manipulations.

Display large version of this figure

Figure 1

Effects of CTZ and L497Y mutation on AMPAR function. (A) Fast desensitization of GluR1 receptors was measured during 50-ms pulses of 10mM glutamate (arrow indicates peak of GluR1 wild-type current, WT). Desensitization was absent after preexposure (>10s) to 100μM CTZ or introduction of the L497Y mutation at the dimer interface. The GluR1-L497Y mutant trace was normalized to allow for direct comparison of decay kinetics after the removal of glutamate. Traces here and in subsequent figures are averages of 5–10 consecutive responses. (B) Although CTZ and the L497Y mutation effectively alleviated desensitization during 50-ms pulses, partial, slow desensitization was apparent during longer 2-s agonist pulses. (C) Deactivation after 1-ms pulses of 10mM glutamate. Fast deactivation of wild-type GluR1 was moderately slower after preincubation with CTZ. Introduction of the L497Y mutation slowed channel deactivation more extensively. The L497Y mutant currents were normalized for comparison of kinetics. Identical results were obtained if CTZ was preapplied but absent during the agonist pulse, as in the traces shown, or by coapplication of CTZ after preexposure. (D) Concentration-response curves for wild-type GluR1 in the absence and presence of 100μM CTZ and for GluR1-L497Y. Data are fit by the logistic equation (see Methods). Error bars represent SEM for individual data points.

Immediately after the application of glutamate, receptors begin to desensitize at a rate of ∼450s⁻¹ (τ_des ∼2.2ms). This limits their contribution to the peak response and theoretically should do so most profoundly at lower agonist concentrations where agonist-binding rates (k_on) are equivalent to or slower than the rate of desensitization. Thus, blocking entry into the desensitized states should facilitate receptor activation and increase the apparent affinity for glutamate. To confirm this relation, we compared the effects of CTZ and the L/Y mutation on the glutamate concentration-response relation. The curves were derived using a 10mM maximum concentration of glutamate and sequential application of six lower agonist concentrations before returning to the maximum concentration to assess rundown; data from patches that exhibited >20% rundown of the maximal currents were excluded. Peak response data were normalized to the maximal concentration, plotted as a function of glutamate concentration, and fit by a logistic equation (see Methods). Best-fitting EC50 values are given in Table 1. Notably, even though both conditions achieved near-complete block of desensitization, CTZ produced only a 9.4-fold increase in agonist potency, whereas the L497Y mutation provided nearly a 40-fold increase in agonist potency relative to wild-type controls (Figure 1DF and Figure 2DF and Table 1). This more pronounced effect of the L497Y mutation on agonist potency was paralleled by slower deactivation rates (Figure 1CE and Figure 2CE). Although both CTZ and the L/Y mutation produced equivalent block of desensitization, their additional effects on deactivation and EC50 were evidently very different.

The quantitatively different effects of CTZ and the L/Y mutation on deactivation and agonist EC50 could be explained in one of two ways. 1), Both CTZ and the L/Y mutation have independent actions on desensitization, deactivation, and agonist binding. Their varied effects are best explained by a difference in efficacy for deactivation and agonist affinity. 2), Slowing deactivation and lowering the EC50 is a consequence of blocking desensitization for both GluR1+CTZ and L/Y receptors, but only CTZ has an additional negative allosteric effect that increases receptor deactivation and decreases the apparent affinity for agonist.

To determine which of these explanations was valid, we examined the effects of CTZ on L/Y mutant receptors. L/Y mutant receptors showed little desensitization during 50-ms agonist pulses in the absence or presence of CTZ. Longer agonist pulses (2s) allowed L/Y receptors to desensitize to ∼75% of their peak currents (Figure 2B, see Table 1 for values). This residual desensitization was blocked by preapplication of 100μM CTZ (Figure 2B) and was accompanied by an unexpected threefold faster deactivation of L497Y mutant receptors (Figure 2C–E). CTZ also produced a rightward shift in the glutamate concentration-response curve (Figure 2D–F); EC50 values were sixfold higher after preincubation with CTZ. All CTZ effects were reversed by prolonged washout of the drug. In summary, CTZ further slowed GluR1-L497Y receptor desensitization but paradoxically accelerated receptor deactivation and decreased agonist potency. This result indicates that CTZ and the L/Y mutation effects are additive with respect to desensitization but competitive with respect to deactivation and EC50. Such competition, albeit only for the latter effects, implies convergence at a common site or process that could result from CTZ binding in proximity to the L/Y residue 11 or at another unknown site. We therefore questioned whether the noncompetitive and competitive effects might be mediated by CTZ binding to multiple sites.

Display large version of this figure

Figure 2

Inverse actions of CTZ at L497Y mutant receptors. (A) Current traces showing desensitization kinetics of GluR1-L497Y in the absence and presence of 100μM CTZ during 50-ms pulses of 10mM glutamate. CTZ appeared to have little effect on L497Y currents during 50-ms pulses. (B) GluR1-L497Y mutant receptor currents were nondesensitizing during 50-ms pulses, but longer pulses (2s) disclosed a partial and slowly desensitizing response to 10mM glutamate. The residual L497Y receptor desensitization was alleviated after preincubation with 100μM CTZ. (C) Deactivation kinetics of GluR1-L497Y (1-ms pulse, 10mM glutamate) in the absence and presence of CTZ. Deactivation was about three times faster after preincubation with 100μM CTZ. (D) Concentration-response curves for the GluR1-L497Y mutant in the absence and presence of CTZ. Data are fit by the logistic equation (see Methods); error bars represent SEM. (E, F) Summary bar plots comparing deactivation time constants (E) and agonist potency (F) for GluR1-wt and the GluR1-L497Y mutant receptors in the absence and presence of CTZ. Error bars represent SEM. Numbers of observations are given in parentheses. ***P<0.001 versus R1-wt controls; ^†††P<0.001 comparing GluR1+CTZ versus GluR1-L497Y; ^‡‡‡P<0.001 comparing GluR1-L497Y in the absence and presence of CTZ. Unpaired t-tests were used with Bonferroni correction for multiple comparisons.

To determine if these separate and opposing actions of CTZ required multiple binding sites, we introduced the S750Q mutation to both GluR1-wt and GluR1-L497Y constructs to prevent CTZ from binding to its known binding site at the dimer interface. Previous studies have shown that CTZ binding critically involves the SNQ residue at position 750 (in GluR1), with serine (S) being the most active and glutamine (Q) being inactive 7,11. The GluR1-S750Q single mutant served as a control for these experiments. The GluR1-S750Q mutant exhibited desensitization and deactivation kinetics that were slightly faster than those of wild-type GluR1 (Figure 3A–C). GluR1-S750Q receptor rates of deactivation and desensitization were consistent with those previously reported 7. We also determined that the GluR1-S750Q receptor EC50 for glutamate was essentially unchanged from GluR1-wt (see Table 1). Addition of the S750Q mutation to the L/Y mutant resulted in several unexpected and significant changes in receptor phenotype, the most important of which was elimination of both the positive and negative allosteric effects of CTZ (Figure 3B–D). As expected, all CTZ effects on GluR1-wt and L497Y mutant channels are made possible by the S750 residue and therefore a single binding site.

Display large version of this figure

Figure 3

Negative allosteric actions of CTZ require the S750 residue. (A) Wild-type GluR1 and GluR1-S750Q mutant responses to 50-ms application of 10mM glutamate are overlaid to compare desensitization kinetics. The GluR1-S750Q mutant was unaltered after treatment with 100μM CTZ. (B) The GluR1-L497Y mutant and GluR1-L497Y/S750Q double mutant responses to 50-ms application of 10mM glutamate are overlaid to compare desensitization kinetics. Note that the GluR1-L497Y/S750Q double mutant exhibits some desensitization (average ∼27%), which is absent or slower in the L/Y parent. GluR1-L497Y/S750Q mutant kinetics were unaltered after treatment with 100μM CTZ. (C) Deactivation kinetics of the GluR1-S750Q mutant receptor are similar in the absence and presence of 100μM CTZ. Traces are pre- and postdrug responses to 1-ms applications of 10mM glutamate. The GluR1-wt trace is provided for reference. (D) Deactivation kinetics of the GluR1-L497Y/S750Q double mutant are unaltered by 100μM CTZ. The overlay is provided for direct comparison of pre- and postdrug kinetics and the GluR1-wt response. (E) Summary bar plot comparing the desensitization time constants of GluR1 wild-type and mutant receptors. Error bars represent SEM. Numbers of observations are given in parentheses. The GluR1-L497Y/S750Q double mutant receptor displayed a partial block of desensitization (73±3% nondesensitizing); τ_des of the remaining decay from peak to steady state was 16.4±0.4ms. ND indicates the majority of current was nondesensitizing during 50-ms agonist pulses. (F) Summary bar plot comparing deactivation time constants for GluR1-S750Q and GluR1-L497Y/S750Q double mutant receptors in the absence and presence of CTZ. Error bars represent SEM. Numbers of observations are given in parentheses. Pre- and postdrug measures were not significantly different for either receptor by paired t-tests. A significant difference was disclosed only between GluR1-S750Q and GluR1-L497Y/S750Q (**P<0.01, unpaired t-test with Bonferroni correction).

The GluR1-L497Y/S750Q double mutant may represent the keystone to understanding how CTZ and L/Y mutation effects are both related to the S750 residue. Comparable to the GluR2-L483Y/S754D mutant described by Sun et al. 11, the GluR1-L497Y/S750Q mutant desensitized partially during 50-ms glutamate pulses. The extent of desensitization was far from complete, and its onset was sevenfold slower than GluR1-wt but still fourfold faster than the L/Y parent (see Table 1). Equilibrium (steady-state) currents during 50-ms glutamate pulses were 73±3% of peak (Figure 3B). These measures were notably similar to the equilibrium currents recorded for WT+CTZ and L497Y mutant receptors during the 2-s agonist pulses (Figure 1B). Closer examination of the GluR1-L497Y/S750Q double mutant indicated that deactivation of this receptor was 10-fold faster than that in the L/Y parent, and the EC50 for glutamate was ∼40-fold higher. Both of these values are remarkably comparable to GluR1-wt and the GluR1-S750Q mutant (see Table 1 and Figure 1CCC and Figure 2CCC and Figure 3CCC for comparison). These observations indicate that effects of the L/Y mutation on deactivation and agonist potency are not a consequence of slow desensitization but occur independently. Moreover, this mutation demonstrates that the S750 residue is required not only for all the effects of CTZ but also for the effects of the L497Y mutation on deactivation and agonist potency (Figure 1CDD and Figure 3CDD). Faster deactivation and reduced agonist potency of the CTZ bound L/Y receptor implies that CTZ disrupts a direct or indirect interaction between Y497 and S750 to promote faster channel closing or agonist dissociation.

The appearance of negative allosteric effects of CTZ (i.e., faster deactivation and reduced agonist potency) in addition to its positive allosteric effect (i.e., slower desensitization) at L497Y receptors could not have been predicted. We questioned whether these counteractive secondary effects were unique to CTZ. We first examined the actions of TCM, a congener of CTZ, on L/Y mutant receptors. TCM is also a positive allosteric modulator of AMPA-type channels, but it is not as potent as CTZ 21,36. Little is known about the binding of TCM, but the fact that TCM and CTZ are benzothiadiazides having similar structures and activities suggests they bind to a common site at the GluR1 dimer interface. In agreement with previous studies 21,36, TCM (500μM) slowed GluR1 wild-type desensitization by sevenfold and increased the steady-state currents to nearly 90% of peak (Figure 4A). This concentration of TCM did not alter the nondesensitizing phenotype of GluR1-L497Y mutant receptors but did significantly increase the rate of GluR1-L497Y deactivation and produced a rightward shift in the concentration-response curve (Fig. 4 and Table 1). These actions of TCM were qualitatively similar in all respects to CTZ, and the results therefore support the hypothesis that the two drugs share a common binding site and mechanism of action. Notably, TCM was ineffective at GluR1-S750Q mutant receptors, and its effects were generally less robust than CTZ (Table 1). These differences may correspond to the lower affinity of the congener or to differences in the relative efficacies of the modulators on the various processes examined.

Display large version of this figure

Figure 4

Effects of other allosteric modulators on GluR1-L497Y receptors. (A) Responses of GluR1 wild-type to 1- and 50-ms pulses of 10mM glutamate in the absence and presence of other positive allosteric modulators. Traces are averages of 5-10 consecutive responses measured under control conditions and after preincubation with 100μM CTZ, 500μM TCM, or 100μM CX614. (B) Responses of GluR1-L497Y receptors to 1- and 50-ms pulses of 10mM glutamate under control conditions and after preincubation with 500μM TCM or 100μM CX614. (C) Traces are superimposed to compare deactivation kinetics of GluR1-L497Y (1-ms pulse of 10mM glutamate) under control conditions and after preincubation with TCM or CX614. (D) Summary bar plot of deactivation time constants. Error bars represent SEM. Numbers of observations are given in parentheses. **P<0.01, ***P<0.001 versus L497Y receptors in the same patches before drug treatment (paired t-tests). Note that deactivation after exposure to CX614 was biexponential; a weighted τ of 21.9±1.4ms is plotted. (E) Concentration-response curves for GluR1-L497Y receptors in the absence and presence of modulators. Data are fit by the logistic equation (see Methods); error bars represent SEM for individual data points. Note the moderate rightward shift in the presence of TCM, which was not observed in the presence of CX614. The logistic fit to TCM data had an unusually shallow slope with a Hill coefficient of 0.7, as compared to values of 1.2 to 1.4 in most other cases. (F) Summary bar plot of agonist potency under control and drug conditions. Error bars represent SEM. Numbers of observations are given in parentheses. Best-fitting EC50 values were determined for individual patches. ***P<0.001 versus untreated L497Y receptors (unpaired t-test with Bonferroni correction).

To confirm that the triad of effects displayed by CTZ on GluR1 receptors (slower desensitization, and counteractive effects on deactivation and EC50) was unique to benzothiadiazides, we investigated another family of positive allosteric modulators. CX614 is a benzamide that belongs to a family of modulators known collectively as Ampakines 37,38. Ampakines also potentiate AMPAR currents but are believed to do so predominantly by slowing the rate of channel closing, although to a lesser extent they also inhibit desensitization 7,28,39. At a concentration of 100μM, CX614 slowed desensitization of wild-type GluR1 by twofold, increased the nondesensitizing steady-state currents (from <1% before drug treatment to 39±4% after treatment), and slowed deactivation by 2.5-fold (n=5) (Figure 4A). CX614 also slowed deactivation of L497Y mutant receptors by nearly twofold (Figure 4CD). Similar results were obtained with 3mM aniracetam, a first-generation Ampakine, at L497Y mutant receptors (τ_deact 21±2ms). Despite the marked slowing of deactivation in the presence of CX614, no appreciable change was observed in the L497Y glutamate concentration-response curve (Figure 4EF). Such profound slowing of deactivation without a concomitant change in EC50 is not consistent with changes to agonist dissociation rates (k_off) but is consistent with slower channel-closing rates (α). These data corroborate conclusions that have been made regarding the actions of aniracetam and CX614 on GluR2 flip and GluR1–3 flop splice variants 7,28,38.

To further inform conclusions about the effects of modulators and mutations on AMPAR gating, we simulated their actions by computational modeling. We began with a simplified version of the Robert and Howe model 29, as shown in Fig. 5, which includes four binding sites linked to three subconductance states displayed by AMPARs 31,34. The assumptions of the model have been described 29, and the kinetic rate constants that were derived for GluR1-wt (Table 2) produced an accurate simulation of the GluR1-wt phenotype (τ_des 2.1ms, τ_deact 0.8ms, and 597μM EC50) (Figure 5BC).

Display large version of this figure

Figure 5

Computational simulations of modulatory effects. (A) Kinetic scheme of the four-site model used for simulations. The form of the model and starting rate constants were adapted from Robert and Howe 29, where R represents receptors in the closed, activatable state, D represents the desensitized receptor state, O represents the open-channel state, and subscripted numbers indicate the number of agonist molecules bound in the various states. O₂, O₃, and O₄ conductance levels were 8, 15, and 23 pS, respectively. Agonist binding (k⁺) and unbinding (k⁻) reactions are indicated by right and left arrows. Channel-opening (β), -closing (α), desensitization (δ), and resensitization (γ) reactions are indicated by up or down arrows. Optimized reaction rates are given in Table 2. (B) Simulated current traces comparing the kinetics of deactivation (1ms agonist pulses, left) and desensitization (50ms agonist pulses, right) of the models after optimization of model parameters to match experimental results. (C) Simulated concentration-response curves from the optimized models. Peak currents were measured from 50-ms simulations, normalized, and plotted relative to their individual maxima. Data are fit with the logistic equation (see Methods). Best-fitting EC50 values are given in the figure.

We first determined the minimum changes necessary to reduce the rate and extent of desensitization to simulate GluR1+CTZ and GluR1-L497Y. A “nondesensitizing” phenotype was replicated in several ways, including slowing entry into the desensitized states (k_des), accelerating exit from the desensitized states (k_res), or simply eliminating the desensitized states. Changing k_des and k_res had complementary effects except that k_res primarily impacts the magnitude of the steady-state current, whereas k_des governs the rate of decay from peak to steady state and has greater impact on the peak current amplitudes. Changing these rates together gave the best approximation of CTZ effects (40-fold slower k_des, 4-fold faster k_res), whereas slowing k_des alone by 150-fold was sufficient to model the L/Y mutant currents. These modifications gave the best approximation of CTZ and L/Y mutant currents having an initial peak and slow decay to a steady-state level at ∼80% of peak.

Regardless of how the “nondesensitizing” phenotype was modeled, reducing or eliminating desensitization from the model by itself predicted very little change in either τ_deact (0.8 to 1.0ms) or EC50 (570 to 416μM). In addition to alterations in k_des/k_res, more realistic models of CTZ and L/Y mutation effects on these measures required additional changes to agonist dissociation rates (k_off). They could not be reproduced by slowing channel-closing rates (α), which slowed deactivation but did not lower the agonist EC50. CTZ effects on GluR1 were best simulated having 3.6-fold slower k_off than wild type (τ_deact 2.3ms, EC50 125μM), and L/Y mutation effects required a 20-fold slower k_off (τ_deact 13.9ms, EC50 25μM). All other modulator and mutation effects, except for CX614, were readily simulated by altering these same rates (Figure 5BC, and Table 1). The additive effects of CTZ on L/Y mutant desensitization were reproduced in the L/Y model having fourfold faster k_res, as in the CTZ model, whereas the acceleration of deactivation and reduction of agonist potency by CTZ was readily simulated by increasing k_off in the L/Y model by threefold (τ_deact 3.84ms, EC50 71μM). The best-fitting model describing the behavior of the L/Y-S750Q double mutant receptor required 20-fold slower k_des, sixfold faster k_res, and 30% faster k_off as compared to GluR1-wt. Notably, the small change in k_off made here was also sufficient to produce the slight increase in EC50 and faster τ_deact and τ_des seen for the GluR1-S750Q mutant 7 (Table 1).

In contrast to these, the preferential effects of CX614 on deactivation with lesser effect on agonist EC50 were reproduced by slowing channel closing rates threefold in the wild-type model. This change alone slowed τ_des and τ_deact 2.6-fold because both are measured as decay of current from the open states and are therefore limited by channel closing. Truly accurate simulations also required fivefold faster k_res to reproduce the larger equilibrium currents associated with CX614 without further slowing of desensitization 28 (Table 1).

Our experimental results can therefore be reasonably well described by modifications to existing models of GluR function. Within the context of these models, the multiple effects of modulators and even the single-point mutations cannot be simulated by changes to desensitization alone (k_des/k_res) but require additional changes to agonist dissociation (k_off) or channel closing (α).

Emerging interest in the development of AMPAR allosteric modulators as potential therapeutics makes it increasingly important to understand both the effects of lead compounds and the nature of the gating processes they modulate. AMPAR gating is an allosteric process by which glutamate binding promotes conformational transitions from closed to open and desensitized channel states. Although the molecular details of channel opening and desensitization are unclear, it has been proposed that closure of the ligand-binding domain around ligand pulls against a rigid dimer interface to cause opening of the channel pore. Desensitization is thought to result from disintegration of the dimer interface, which uncouples ligand binding from channel opening. Functional and atomic-level structural analyses support this model of AMPAR gating 11,12,13,14,28,29,40,41. Mutations that promote stability of the dimer interface slow desensitization, as in the case of L497Y, whereas mutations that destabilize the dimer interface accelerate desensitization 12,14,40. CTZ binds to specific residues within the dimer interface to promote dimer stability and slow desensitization 7,11.

The structural model does not explain why the effects of modulators and mutations that stabilize the dimer interface are not limited to desensitization. Rather, such manipulations tend to produce a constellation of effects, simultaneously altering deactivation, desensitization, and agonist potency. These secondary effects on deactivation and agonist potency might result in part or entirely from the slowing of desensitization itself, to the extent that desensitization limits channel activation at low agonist concentrations and provides a route of inactivation from the ligand-bound and possibly open states. Otherwise, they might reflect secondary actions of drugs and mutation on channel-closing (α) or agonist dissociation (k_off) rates that are entirely independent of effects on desensitization. If the former were true, we hypothesized that block desensitization by any means should have quantitatively similar effects on deactivation and agonist EC50. Moreover, these effects of modulators should be occluded in cases where desensitization is already blocked by mutation.

To test these predictions, we compared the effects of CTZ and TCM on GluR1-wt and GluR1-L497Y mutant receptors. Consistent with previous reports, CTZ, TCM, and L/Y mutation profoundly slowed desensitization of GluR1 during prolonged exposure to agonist 7,27,34,36. Yet, despite the near absence of desensitization during test pulses, these treatments had quantitatively very different effects on deactivation and glutamate potency (see Table 1). Furthermore, although CTZ and TCM further slowed desensitization in the L/Y mutant, they paradoxically accelerated deactivation and reduced agonist potency, opposite their normal effects on GluR1-wt. Results therefore indicate that the effects on deactivation and agonist potency are not a consequence of slower desensitization. Predictions of computational models support this conclusion and further suggest that CTZ, TCM, and L/Y mutation alter desensitization (k_des/k_res) and agonist binding (k_off) independently. The effects on k_des/k_res most likely involve greater dimer stability, whereas the modulation of k_off appears to involve interactions with the S750 residue specifically. Greatest support for this conclusion comes from the L497Y-S750Q double mutant where S/Q mutation restores L/Y mutant deactivation and EC50 to wild-type levels.

Aniracetam and CX614 are unique from CTZ and TCM. These modulators bind between the lower lobes of the ligand-binding domain within the dimer interface, where they inhibit the relaxation from the closed- to open-cleft configuration 28. These and related benzamide modulators generally slow receptor deactivation with relatively less effect on desensitization or agonist potency. Their actions persist in the L497Y and S750Q mutants and are best modeled by assuming faster recovery from desensitization and slower channel closing 7,28,37,38,42.

There are several possible explanations for the conflicting actions of these modulators on deactivation and agonist potency in the L/Y mutant. One interpretation is that modulator-receptor interactions supersede interactions made by the mutant tyrosine residue so that L/Y mutant and wild-type receptors behave similarly when bound to modulators. Indeed, τ_deact and EC50 values of WT+CTZ (or TCM) and L/Y+CTZ (or TCM) are more similar than those of L/Y receptors in the absence and presence of modulators (Table 1). However, several observations contradict this explanation. First, both modulators further slowed desensitization of L/Y mutant receptors. The additive nature of effects suggests that both the modulator and L/Y interactions across the dimer interface remain intact. Second, τ_deact and EC50 values are significantly different between WT+CTZ (TCM) and L/Y+CTZ (TCM), in some cases by two- to threefold. Unfortunately, the L/Y+CTZ crystal structure has not been solved, which might further refute this explanation. More limited insight may be gained by comparison of related structures, including GluR2+AMPA (PDB: 1FTM), GluR2+GLU+CTZ (PDB: 1LBC), and GluR2-L483Y+AMPA (PDB: 1LB8) 10,11. Superimposition of these structures indicates that the L/Y residue is 8 to 8.7Å from the nearest point of CTZ, suggesting there is unlikely to be any direct interaction between the L/Y residue and modulators. Examination of all residues within 4Å of CTZ, the L/Y residue, or the bound agonist in the superimposed structures reveals only minuscule differences in the position or orientation of these residues that cannot account for the profound effects of CTZ or TCM on L/Y deactivation and agonist potency. We could find no indication that CTZ supersedes any direct interactions made by the L/Y residue.

A more plausible explanation for the seemingly paradoxical actions of modulators on L/Y mutant receptors is that CTZ and TCM have two opposing actions. The first is to block desensitization, which by itself only modestly slows deactivation and increases agonist potency. The second is to destabilize agonist binding, which accelerates deactivation and decreases agonist potency. Quantitative differences among CTZ, TCM, and L/Y mutation effects and their paradoxical interactions can thus be explained by how effectively these manipulations engage the two processes. Some previous studies have alluded to a second inhibitory action of CTZ that occurs independently of desensitization. Patneau et al. 43 noted a modest inhibition of kainate-evoked currents in native GluRs from hippocampal neurons. Others have reported reduced affinities for radioligands binding to native or recombinant AMPARs treated with CTZ 44,45 and an approximately twofold increase in the agonist EC50 at some nondesensitizing GluR chimeras 27. These data are most consistent with effects on agonist dissociation (k_off). All of the effects on desensitization, deactivation, and EC50 can be attributed to a single CTZ binding site because they are abolished by the S750Q mutation 7,11. Moreover, the profound slowing of desensitization in the L/Y-S750Q double mutant without effects on τ_deact or EC50 argue that the S750 residue is directly involved in modulation of k_off in addition to its being required for modulator binding.

S750Q mutation in wild-type GluR1 has only a modest impact on receptor function 7,11,23. Such a profound effect on L/Y mutant receptor function is therefore quite remarkable. Sun et al. 11 reported a similar, slowly desensitizing phenotype for the GluR2-L483/S754D double mutant. These mutations target the same sites we examined in GluR1 except that the S/D mutation alone accelerates desensitization more than S/Q 7,11; τ_deact and EC50 values were not available for comparison. The authors concluded that these two residues act independently to regulate the transitions into and out of the desensitized states. Our results support their conclusion with respect to the rate and extent of desensitization, which we agree is most readily explained by the effects of these mutations on dimer stability. On the other hand, these residues clearly interact with respect to deactivation and agonist potency. In fact, our data suggest that the S750 residue is essential for modulation of k_off by L/Y mutation. Unfortunately, because this residue is required for CTZ binding, its downstream involvement in benzothiadiazide modulation cannot be ascertained. Nonetheless, the L/Y-S750Q double mutant clearly demonstrates that desensitization can be modified independently of deactivation, at least by mutagenesis targeting the dimer interface.

Insights from comparison of these modulators and mutations may offer strategies to more selectively target desensitization or other functional measures. CTZ and TCM differ by an R3 substitution, which differentiates their size, hydrophobicity, and charge. It may be significant that that the R3 moiety occupies a hydrophobic pocket near the L/Y residue 11. Although CTZ and TCM have substantially different affinities for AMPARs, both produce similar block of desensitization. More importantly, their modulatory effects on τ_deact and EC50 are significantly different, suggesting that other R3 substitutions may provide a means to selectively target these parameters. Numerous benzothiadiazides have been described, including additional R3 substitutions, some of which slow AMPAR desensitization 21. Aside from CTZ and the present characterization of TCM, none of these compounds has been thoroughly examined for effects on AMPAR deactivation or agonist potency. Additional characterization of this family of compounds could be useful in the generation of more functionally selective modulators that would have both experimental utility in dissecting GluR function and therapeutic value in correcting GluR dysfunction in disease.