Modulation of α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic Acid Receptor Desensitization by Extracellular Protons

Experimental Procedures

Cell Culture.

Primary cultures of hippocampal neurons were prepared according to established methods (Patneau et al., 1993). Briefly, the hippocampi were dissected from the brains of newborn (P0–1) Sprague-Dawley rats, incubated in papain (Worthington Biochemical Corp., Lakewood, NJ) and dissociated by trituration with a glass pipette. The cells were suspended in 90% Dulbecco's modified Eagle's medium/10% FBS supplemented with 2 mM Glutamax (Life Technologies, Rockville, MD) and 1% penicillin/streptomycin, plated onto glass coverslips precoated with collagen and poly-l-lysine, and maintained at 37°C in a humidified 10% CO₂ incubator. Under these conditions glia proliferate and few neurons survive. After the glial monolayer became confluent, 33 μg/ml uridine and 13 μg/ml 5′-fluorodeoxyuridine was added to the culture medium to inhibit mitosis. Neurons from a second dissociation were plated onto the glial monolayer and maintained in 94% Dulbecco's modified Eagle's medium/5% horse serum/1% FBS, supplemented with 2 mM Glutamax, 5′-fluorodeoxyuridine, and a neuronal supplement. Experiments were performed at room temperature (23–25°C) using neurons after 2 to 15 days in culture.

For expression of recombinant AMPA receptors, human embryonic kidney (HEK) 293 cells (American Type Culture Collection, Manassas, VA) were grown at 37°C in a humidified 5% CO₂ incubator in 90% minimal essential medium plus Earle's salts/10% FBS, supplemented with 2 mM l-glutamine. Plasmids containing cDNA for the glutamate receptor (GluR) subunits GluR-A, -B, -C and -D_, in both flip (i) andflop (o) isoforms, were gifts from Dr. P. Seeburg (Heidelberg, Germany). The plasmids were grown in Escherichia coli and purified by 2× CsCl density centrifugation or using a plasmid kit (Qiagen Inc., Valencia, CA). Cells were transfected with cDNA (10 μg/60 mm dish) using a modified CaPO₄precipitation method (Chen and Okayama, 1987). After 8 to 10 h, the medium containing precipitate was removed and the transfected HEK 293 cells were replated onto collagen/poly-l-lysine-treated coverslips in growth medium that contained 0.5 mM kynurenate to minimize excitotoxicity. Electrophysiological recordings were made 18 to 48 h later.

Recording Solutions.

Extracellular saline for neuronal experiments contained 166 mM NaCl, 2.5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 10 mM BES (or HEPES), 10 mM glucose, and 0.01 mg/ml phenol red; the osmolarity was adjusted to 325 mOsM, and the pH titrated with NaOH to the desired value. When varying pH, the buffer species was selected to maintain optimal buffering capacity over the pH range examined and was held constant in each experiment because significant effects of varying buffer were observed in preliminary experiments (data not shown). Thus HEPES (pK _a = 7.48) buffered the solutions for experiments examining the effects of alkalinization, whereas BES (pK _a = 7.09) buffered solutions for all other experiments. Tetrodotoxin (400 nM; Calbiochem, La Jolla, CA), bicuculline (5 μM), and (5R,10S)-(+)-5-methyl-10,11-dihydro-5H-dibenzo[a,d]cyclohepten-5,10-imine hydrogen maleate (MK-801; 500 nM) were added to block voltage-gated sodium channels, GABA receptors, and NMDA receptors, respectively. MK-801 was also added to neuronal cultures in the incubator to allow spontaneous synaptic activity to effect the essentially irreversible activity-dependent block of NMDA receptors before recording. The standard intracellular solution for neurons contained 125 mM CsMeSO₃ (Aldrich Chemical Co., Milwaukee, WI), 15 mM CsCl, 10 mM HEPES, 5 mM Cs₄BAPTA (Molecular Probes, Eugene, OR), 0.5 mM CaCl₂, 3 mM MgCl₂ and 2 mM Na₂ATP; for recording from nucleated macropatches, 10 mM CsF replaced 10 mM CsMeSO₃. The osmolarity was adjusted to 305 to 310 mOsM, and the pH was titrated to 7.2 with CsOH. In some experiments, 35 mM CsMeSO₃ was replaced with 20 mM creatine phosphate (di-Tris salt) and an additional 2 mM Na₂ATP and 50 U/ml creatine phosphokinase were included (to regenerate ATP).

Extracellular saline for recording from HEK 293 cells contained 145 mM NaCl, 5.4 mM KCl, 1 mM MgCl₂, 1.8 mM CaCl₂, 5 mM BES, and 0.01 mg/ml phenol red. The osmolarity was adjusted to 295 mOsm and the pH to 6.5 or 7.3 with NaOH. In some experiments, the [Ca²⁺] was lowered to 0.2 mM to reduce rundown of responses. Patch pipettes were filled with intracellular solution containing 135 mM CsCl, 10 mM CsF, 3 mM MgCl₂, 0.5 mM CaCl₂, 5 mM Cs₄BAPTA, 10 mM HEPES and 2 mM Na₂ATP. Osmolarity was adjusted to 295 mOsM and pH to 7.2 with CsOH. In some experiments, the intracellular solution was ATP-regenerating (see above).

Electrophysiology.

The recording chamber was continuously perfused with control extracellular saline at 0.2 to 0.5 ml/min. Recording electrodes were pulled from borosilicate glass, coated with Sylgard, and fire-polished; typical electrode resistance was 2 to 4 MΩ when filled with intracellular solution. An isolated neuron or HEK 293 cell was voltage-clamped in the whole-cell configuration at −60 mV using an Axopatch 200A amplifier (Axon Instruments, Foster City, CA) and standard techniques. In experiments to examine the kinetics of desensitization, nucleated outside-out patches (macropatches) were used (Sather et al., 1992; Patneau et al., 1993). Resistance in series with the cell or macropatch was typically 4 to 8 MΩ and was compensated by 60 to 90%. Data for on-line recording were filtered at 1 to 5 kHz and sampled at 2 to 20 kHz using pClamp (Axon Instruments).

Perfusion Techniques.

For whole-cell recording, rapid agonist application was achieved using a glass flowpipe array (6–8 parallel barrels, each 400 μm in diameter) mounted on a ‘bimorph’ ceramic wafer. The flowpipe array was placed near a voltage-clamped cell. Solutions were driven by a peristaltic pump (Minipuls3; Gilson, Middleton, WI) through three-way Isolatch valves (Parker Hannifin, Fairfield, NJ). Control solution (at the desired pH) from one barrel of the flowpipe continuously bathed the cell and its processes until agonist application was initiated; at this time the bimorph was charged, which rapidly moved an adjacent barrel containing agonist solution in front of the cell. Then, simultaneously, the valve regulating control solution flow was closed and the valve for agonist flow was opened. This system achieves solution exchange around a small neuron and its processes within approximately 10 ms. Other cells in the recording chamber were partially protected from pre-exposure to agonists by the inclusion of 0.5 mM kynurenate in the bath solution.

For fast perfusion of macropatches, solutions were gravity-fed through four-barreled square glass tubing pulled to a width of approximately 100 μM per barrel. Recording salines at pH 7.3 were in one horizontal pair of barrels (control and agonist), salines at pH 6.5 in the second pair. The tubing was mounted on a piezo translator driven by a 100 V power supply (PZ 100; Burleigh Instruments Inc., Fishers, NY). A macropatch was positioned in the control solution stream, near the interface between control and agonist-containing solutions. Upon charging or discharging of the piezo element, the interface between solutions was rapidly moved across the macropatch. This system achieves solution exchange on a nucleated macropatch in ≤ 300 μs, as determined from the 10 to 90% rise- and decay-time for a sodium concentration change in the presence of kainate. Vertical displacement of the recording pipette moved the macropatch between solutions at pH 7.3 and pH 6.5. At the end of each recording, to verify the correct positioning of the macropatch for optimal solution exchange, the macropatch was disrupted and junction potentials at the pipette tip were recorded ([Na⁺] was 6 mM lower in the control solutions; see Fig. 3 for examples). Data were excluded if the junction potentials indicated the macropatch was not optimally positioned.

Data Analysis.

Peak and steady-state amplitude and kinetic measurements were based on single or the average of a few (3–5) agonist-evoked responses. The majority of experiments examined the effects of protons on steady-state responses. Peak response amplitudes and kinetic measurements were only analyzed for macropatches, because the exchange time for whole-cell perfusion was too slow (see above) to make peak measurements reliable. Control responses preceded and followed an experimental manipulation. To correct for rundown, the experimental observation was normalized relative to the average of the two control responses. Numbers and statistical analyses in the text are based on these corrected responses. However, both control responses are illustrated in some bar graphs to demonstrate recovery (see Fig.1). Because substantial rundown can indicate a compromised recording, data exhibiting more than a 15% (neurons) or 25% (macropatches and HEK 293 cells) decline in control responses were excluded from analysis.

The kinetics of desensitization and deactivation were determined from exponential fits to responses from macropatches using a Chebyshev (pClamp) or Simplex (Kaleidagraph; Synergy Software, Reading, PA) fitting algorithm. Significance of pH-mediated effects was evaluated with paired-differences or Student's t tests. All data in the text are reported as mean ± S.D. unless noted otherwise. Bar graphs show mean ± S.E.M.

Materials.

Stock solutions of most drugs were prepared in the appropriate intracellular or extracellular solution, stored frozen, and added to experimental solutions immediately before use. When a very high final concentration (≥ 3 mM) of drug was required, the stock solution was made in NaCl-reduced saline to keep the final Na⁺ concentration constant across experimental solutions. Cyclothiazide was dissolved in dimethyl sulfoxide at 20 mM before dilution with extracellular solution, and an equivalent final concentration of dimethyl sulfoxide (0.5%) was added to all experimental and control solutions. The pK_avalues of the carboxylic acid group of glutamate, quisqualate, homocysteate, kainate, and AMPA were determined by standard pH titration methods, with pK_a values of 4.41, 4.30, 2.79, 4.32, and 5.28, respectively. Thus, the concentrations of the unprotonated form of the first four of these agonists in solution at pH 6.5 and 7.3 were negligibly different and did not require adjustment; AMPA was not used because it would be partially protonated at pH 6.5. AMPA, bicuculline, kynurenate, MK-801, and quisqualate were purchased from Tocris Cookson (Ballwin, MO). Other reagents were obtained from Sigma Chemical (St. Louis, MO) unless noted otherwise.

Results

Modulation of AMPA Receptors by Protons Is Agonist-Dependent.

Initial experiments focused on proton-mediated modulation of kainate- or glutamate-evoked responses in hippocampal neurons within a physiologically relevant pH range (pH 6.5–7.8). Alkaline pH (7.8) did not affect the amplitude of AMPA receptor-mediated responses to kainate or glutamate (data not shown). Because pathophysiological processes rarely alkalinize the interstitial milieu of the brain (Chesler, 1990) pH values greater that 7.8 were not tested. However, acidic pH (pH 6.8 or 6.5) did significantly affect responses to kainate and glutamate. This modulation was concentration-dependent, with larger effects observed at pH 6.5 than pH 6.8. The amplitudes of glutamate-evoked responses at pH 6.8 and pH 6.5 were 90 ± 3% and 72 ± 4% (P < .01, n = 6) of those at pH 7.3, respectively. Kainate-evoked responses at pH 6.8 and pH 6.5 were significantly reduced to 95 ± 2% and 86 ± 4% (P < .01, n = 10) of those at pH 7.3, respectively. To maximize the effect of protons while reproducing pathophysiologic conditions associated with ischemia, all further experiments compared agonist-evoked responses recorded at pH 7.3 (control) and pH 6.5.

The difference in the magnitude of inhibition of glutamate- and kainate-evoked responses in the initial experiments suggested that modulation by protons might be agonist-dependent. To test this hypothesis, the effects of pH 6.5 on glutamate- and kainate-evoked responses from single cells were compared (Fig. 1). Responses to glutamate were inhibited by 32 ± 6%, which was significantly greater than the 18 ± 4% inhibition of responses to kainate (paired differences t test, P < .001,n = 14). Similar results were obtained in experiments comparing proton-mediated inhibition of responses evoked by quisqualate and kainate. Responses evoked by kainate were inhibited by 19 ± 3% at pH 6.5 relative to pH 7.3 (P < .001), whereas responses evoked by quisqualate were inhibited by 35 ± 11% (P < .01), a significant difference (P< .05, n = 5). Thus, the differential effects of protons could be ascribed to agonist properties, with responses evoked by the strongly desensitizing agonists glutamate and quisqualate, more susceptible to inhibition than those evoked by the weakly desensitizing agonist, kainate.

Modulators of AMPA Receptor Desensitization Block Proton-Mediated Inhibition.

Because kainate and glutamate differ in their desensitization properties (Patneau et al., 1993), the agonist-dependence of proton modulation suggested that protons may affect desensitization. One way in which protons could modulate desensitization is selective interaction with the desensitized state of the receptor. To determine whether modulation of AMPA receptors by protons was state-dependent, extracellular pH was varied as glutamate was applied to neurons in the presence of trichlormethiazide or cyclothiazide, benzothiadiazine modulators that block desensitization of AMPA receptors by glutamate (Patneau et al., 1993; Yamada and Tang, 1993).

When glutamate was applied to hippocampal neurons in the presence of trichlormethiazide, no effect of protons was observed (Fig.2A). Responses to 1 mM glutamate at pH 6.5 averaged 99 ± 6% (n = 6) of those evoked at pH 7.3 (Fig. 2B). Similar results were obtained for glutamate-evoked currents in the presence of 100 μM cyclothiazide; responses at pH 6.5 were 97 ± 2% (n = 8) of those at pH 7.3. Thus, modulators that block AMPA receptor transitions to the desensitized state prevent inhibition of glutamate-evoked responses by protons.

Protons Do Not Affect AMPA Receptor Desensitization or Deactivation Kinetics.

The results of experiments described above suggested that protons modulate AMPA receptor desensitization. The effects of protons on the kinetics of desensitization were examined using macropatches pulled from hippocampal neurons and a piezo-based perfusion system to achieve submillisecond solution exchange. Responses to applications of three agonists, l-glutamate (3 mM),l-homocysteate (10 mM), and kainate (3 mM), were analyzed. Homocysteate is a strongly desensitizing full agonist with desensitization onset kinetics similar to those for glutamate, but with much lower affinity for the receptor (steady-state EC₅₀ value of 447 μM versus 19 μM for glutamate; Patneau and Mayer, 1990, 1991). Kainate is a weakly desensitizing partial agonist with desensitization kinetics that are significantly faster than glutamate and an intermediate affinity for the receptor (142 μM; Patneau et al., 1993).

To determine whether protons affect AMPA receptor desensitization, responses to 40 ms (kainate) or 100 ms (glutamate and homocysteate) applications were recorded at pH 7.3 and 6.5 (Fig.3A). The onset of desensitization, as reflected in the decay of the response to its steady-state amplitude in the presence of agonist, was then fitted with one (kainate) or the sum of two (glutamate and homocysteate) exponentials. There was no significant effect of pH on the kinetics of desensitization onset for any of the agonists, nor in the relative weights of the fast and slow components for responses to glutamate and homocysteate (Fig. 3C).

Receptor deactivation kinetics, as exhibited by the decay of the response after cessation of agonist application, were also examined to determine whether protons affect the rate of channel closing. For the strongly desensitizing agonists glutamate and homocysteate, a 1-ms pulse isolates deactivation from receptor desensitization (Fig. 3B). Because kainate-evoked responses are incompletely desensitized, sufficient steady-state current remained during a 40-ms pulse to determine deactivation kinetics from the decay of the response after removal of kainate. For all three agonists, decay of the current after removal of agonist was well-fit with a single exponential function at both pH 7.3 and pH 6.5, with time constants as illustrated in Fig. 3D. Deactivation time constants for glutamate at pH 7.3 (1.7 ± 0.4 ms, n = 9) were three times slower than for the low affinity agonist homocysteate (0.56 ± 0.12 ms, n= 7), indicating that the deactivation kinetics for glutamate primarily reflect agonist dissociation rather than channel closing. The mean deactivation time constant for kainate (1.3 ± 0.6 ms,n = 11) was consistent with its intermediate affinity for AMPA receptors. There was no significant difference between the deactivation time constants at pH 7.3 and pH 6.5 for any agonist, suggesting that neither channel closing nor agonist dissociation rates are significantly affected by protons.

Although protons did not alter the kinetics of desensitization onset, they could affect receptor desensitization by changing the rate of recovery from desensitization. This possibility was tested with a paired-pulse protocol, in which a pulse of 3 mM glutamate was followed by a second pulse at intervals of 2 to 800 ms. To estimate the potential effects of protons on synaptic responses, 1-ms pulses that mimic the time course of glutamate in the synaptic cleft were used in the initial analysis. The ratio of the peak amplitude of the second response to that of the first response was plotted against interpulse-interval, and the data were fitted with a single exponential function. The peak amplitudes of the first pulses were significantly inhibited, by 10 ± 3%, at pH 6.5 relative to pH 7.3 (n = 7, P < .001). But recovery from desensitization was not affected by pH, with time constants of 68 ± 11 ms and 64 ± 7 at pH 7.3 and 6.5, respectively. Because it is possible that occupancy of different conformational states may result from longer agonist applications that fully desensitize the receptor, we also examined recovery after 50 ms applications of glutamate, which produce almost complete receptor desensitization (see inset in Fig. 3E). Although the mean time constants for recovery from desensitization were slower than after 1-ms pulses (95 ± 12 ms at pH 7.3 and 90 ± 12 ms at pH 6.5, n = 3), there was no significant effect of pH (Fig. 3E).

Protons Increase Steady-State Desensitization of AMPA Receptors.

The rapid solution exchange possible on macropatches permitted comparison of the effects of protons on peak and steady-state responses. This analysis included all macropatches exposed to kainate. However, only macropatches with control steady-state responses to glutamate and homocysteate ≥ 10 pA were included because changes in the amplitude of smaller responses could not be distinguished from background noise. The peak amplitudes of responses to all three agonists were significantly inhibited at pH 6.5 relative to pH 7.3 (Fig. 3F). Inhibition of steady-state responses to glutamate (31 ± 13%) and kainate (22 ± 5%) in macropatches was comparable with that observed in whole-cell recording. Peak responses to glutamate and homocysteate were inhibited by approximately 10%, significantly less than steady-state responses (Fig. 3D); this indicates that the conformational states of the receptor occupied during instantaneous (peak) and equilibrium (steady-state) responses to glutamate are differentially sensitive to protons. The difference between inhibition of peak and steady-state responses to kainate at pH 6.5 was not significant.

However, examination of the data for the 11 macropatches exposed to kainate revealed a strong correlation (r = 0.89) between steady-state desensitization and proton-mediated inhibition. As shown in Fig. 4A, there was substantial variability in the amount of desensitization in kainate-evoked responses, which varied from 20 to 50%, and in proton-mediated inhibition of those responses, which varied from 17 to 32%. Kainate-evoked responses exhibiting the most desensitization were also most strongly inhibited at pH 6.5 (Fig. 4B). Because steady-state desensitization in kainate-evoked responses is affected by membrane voltage (Patneau et al., 1993), the relationship between desensitization and modulation by protons could also be examined within a single macropatch (Fig. 4C). Steady-state desensitization was strongly voltage-dependent at both pH values, with desensitization at +80 mV approximately 20% less than at −80 mV. But at all voltages, more extensive desensitization of kainate-evoked responses was observed at pH 6.5 than at pH 7.3 (Fig. 4D). Thus, protons seem to increase equilibrium desensitization, but by a mechanism that does not affect the macroscopic kinetics of desensitization.

Proton-Mediated Modulation Varies with AMPA Receptor Composition.

Molecular composition has previously been shown to affect proton modulation of NMDA and GABA receptors (Traynelis et al., 1995; Krishek et al., 1996). The experiments described above suggest that AMPA receptor desensitization and proton modulation interact. Subunit (GluR-A to -D or 1–4; Hollmann and Heinemann, 1994) and isoform (flip or flop) composition of recombinant AMPA receptors determine desensitization time constants and the magnitude of steady-state desensitization of glutamate-evoked responses (Mosbacher et al., 1994; Partin et al., 1994). The flip(i) isoforms exhibit slower rates of desensitization and less extensive steady-state desensitization, whereas theflop (o) isoforms, and particularly Do, desensitize very rapidly (τ<1 ms) to a steady-state response that is ≪1% of the peak. We therefore tested the hypothesis that proton-mediated modulation would vary with the desensitization properties of recombinant AMPA receptors. HEK 293 cells were transiently transfected with either flip or flopisoforms of two subunits, and agonist-evoked responses were recorded at pH 7.3 and pH 6.5.

Glutamate-evoked responses for many flop-containing receptors exhibited steady-state responses that were too small to be distinguished from background noise, but responses to kainate (1 mM) had amplitudes large enough to reliably measure proton-mediated effects. The magnitude of inhibition of kainate-evoked responses at pH 6.5 varied with both subunit and isoform composition, ranging from 10% to > 40%. Inhibition in receptors composed of flopisoforms (on average 30 ± 8%) was greater than in receptors composed of flip isoforms (on average 14 ± 5%). The specific compositions tested, and the percentage inhibition for each, were: AiBi, 11 ± 4 (n = 7); AiDi, 14 ± 6, (n = 10); BiCi, 18 ± 4 (n = 4); AoBo, 26 ± 8 (n = 8); AoDo, 41 ± 2 (n = 7); AoCo, 30 ± 3 (n = 8); and BoCo, 24 ± 1 (n = 4). Figure5 illustrates the effect of pH 6.5 on kainate-evoked responses for flip and flopisoforms in cells expressing heterodimeric receptors composed of GluR-A and -B or GluR-A and -D subunits. Note that receptors containing the subunit exhibiting the fastest desensitization, Do(Mosbacher et al., 1994), were most sensitive to proton-mediated inhibition.

Discussion

These experiments have produced three novel and significant findings. First, proton-mediated inhibition of AMPA receptors is dependent on the type of agonist. Second, protons seem to specifically interact with receptor desensitization but do not affect the macroscopic kinetics of desensitization. Finally, molecular receptor composition determines the magnitude of proton-mediated inhibition.

Previous research on proton-mediated modulation of non-NMDA receptors concluded that protons increase the closed time of the channel (Christensen and Hida, 1990; Traynelis and Cull-Candy, 1991). However, these studies did not determine whether protons increased closed time by directly affecting the transitions between open and closed states or, more indirectly, by affecting receptor desensitization. The experiments described above indicate that protons affect receptor desensitization. However, because protons do not affect desensitization kinetics or the affinity of the desensitized state, our results are inconsistent with a simple model of AMPA receptor gating that includes a single open state accessible from a nondesensitized state (e.g.,Patneau and Mayer, 1991). One way in which protons could affect steady-state receptor desensitization without affecting the macroscopic kinetics of desensitization is through modulation of a second open state accessible from a desensitized state. Inclusion of such an open state has been proposed previously by Raman and Trussell (1992) in their model for AMPA receptor gating in chick cochlear nucleus.

This cyclic model for AMPA receptor gating and desensitization is based on one originally proposed by Patneau and Mayer (1991), but includes a second open state that is accessible from the desensitized state. A represents the agonist; R and R_D represent the receptor in its unoccupied resting and desensitized states, respectively; A₂R is the agonist-bound, active (closed) state; A₂R_D is the agonist-bound, desensitized state; d₁ and d₋₁ represent the rate constants for entry into and recovery from the desensitized state, respectively; β and α designate rate constants of channel opening and closing for the first open state (Open), whereas β^′ and α^′ are the equivalent rate constants for the second open state (Open^′).
Support for protons specifically modulating a “desensitized” open state comes from several findings. First, the smaller effect of protons on responses to the weakly desensitized agonist kainate, compared with the strongly desensitizing agonist glutamate, suggest that the receptor is more susceptible to proton-mediated modulation when it spends more time in the desensitized state. Second, benzothiadiazine drugs that prevent AMPA receptor desensitization also block proton-mediated modulation. Because protons do not affect the rate of entry into or recovery from desensitization, this suggests that protons affect another open state that is not available when desensitization is blocked. Third, inhibition of kainate-evoked responses in macropatches was correlated with the magnitude of desensitization, and protons increased equilibrium desensitization without affecting the macroscopic kinetics of desensitization. Fourth, in transfected cells, receptor compositions previously shown to exhibit the fastest and most extensive desensitization were most sensitive to modulation. Finally, the model is supported by the findings that protons did not affect the kinetics of receptor deactivation for any of the three agonists examined and that protons inhibited peak responses to glutamate and homocysteate significantly less than steady-state responses. If there were only a single open state, and protons increased its closed time, receptor deactivation should have been affected. Although the kinetics of deactivation after removal of glutamate may reflect multiple openings with agonist still bound (see Fig. 3B), the very rapid deactivation observed after removal of the low-affinity agonist homocysteate would seem to accurately reflect the true channel-closing rate, which was not altered by protons. This result therefore suggests that the conformational states of the receptor underlying the instantaneous (peak) and equilibrium (steady-state) responses to glutamate are differentially sensitive to protons.

Our experiments indicate that modulation of AMPA receptors by protons in vivo will depend on both receptor composition and the magnitude of the pH change, whether the acidification is global and persistent, as in ischemia, or local and transient, as may occur in the cleft during synaptic transmission. Although local fluctuations in pH at the synapse have not been directly quantified, the processes that may contribute are known. Glutamatergic vesicles of hippocampal neurons are more acidic than the interstitial milieu (pH 5.67; Miesenbock et al., 1998). Thus, a rapid acidic transient is predicted to occur simultaneously with the release of neurotransmitter (Krishtal et al., 1987) and could be further amplified by deprotonation of glutamate upon release into the less acidic cleft. In addition, proton ATPases from synaptic vesicles may be incorporated in the presynaptic membrane in vesicle fusion and could transport protons from the presynaptic terminal into the cleft (Krishtal et al., 1987). The receptors directly under the site of release should see the largest acidic transient.

The temporal and spatial dispersion of acidic transients will be determined by the kinetics and capacity of the endogenous CO₂/HCO₃ ⁻buffering system. Physiological levels of bicarbonate in the brain confer a high buffer capacity, but substantial experimental evidence suggests that synaptic activation produces transient changes in interstitial pH that are not rapidly buffered (Chesler, 1990). Thus, rapid acidic transients in the synaptic cleft sufficient to significantly affect the function of postsynaptic AMPA receptors are likely to occur, in particular with high frequency synaptic activity or at synapses with multivesicular release.

In animal models of ischemia, significant protection from excitotoxic cell death is conferred by AMPA receptor antagonists (Sheardown et al., 1990; Gill, 1994). The global acidification of the brain that occurs during ischemia is neuroprotective for NMDA receptor-mediated cell death, and it was assumed that it would also reduce AMPA receptor-mediated cell death (Christensen and Hida, 1990; Giffard et al., 1990). In contrast, McDonald et al. (1998) found that acidification similar to that occurring during ischemia (pH 6.6) significantly potentiated AMPA receptor-mediated neurotoxicity in neuronal cultures.

Although our findings do not provide an explanation for the results ofMcDonald et al. (1998), they suggest a mechanism for the selective vulnerability of principal neurons to ischemia (Cervos-Navarro and Diemer, 1991). Brorson et al. (1995) found that AMPA receptor-mediated responses of Purkinje cells, which are among those cell types most vulnerable to ischemia, exhibit less complete desensitization than those of other cerebellar neurons. This difference in desensitization properties has recently been related to higher levels of expression offlip isoforms of AMPA receptor subunits (Tomiyama et al., 1999). Similarly, cortical pyramidal cells exhibit slower and less extensive desensitization than interneurons (Hestrin, 1993). The desensitization properties of pyramidal neurons are also highly correlated with expression of AMPA receptor flip isoforms (Lambolez et al., 1996). Thus, AMPA receptors that exhibit less steady-state desensitization (e.g., flip-containing receptors) will disproportionately contribute to the neuronal depolarization resulting from tonic glutamate release and will also be less sensitive to the potentially inhibitory effects of interstitial protons. Neurons expressing these AMPA receptors are predicted to be more vulnerable to glutamate-mediated excitotoxicity in ischemia.