Allosteric interaction between zinc and glutamate binding domains on NR2A causes desensitization of NMDA receptors

Methods

Transfection of HEK cells

Cells from the human embryonic kidney cell line HEK 293 (CRL 1573; ATCC, Rockville, MD, USA) were maintained at 37°C and 5% CO₂ and plated on 12 mm diameter glass coverslips as previously described (Zheng et al. 1998). Low-confluency cells were transiently transfected by the calcium phosphate method with cDNA encoding NR1-1a (GenBank U11418, U08261; pCIneo vector; hereafter NR1), NR2A (D13211; pCIneo) and GFP at a ratio of 1: 2: 0.2 (0.25 μg ml⁻¹ NR1) for 4–12 h. After transfection, NMDA antagonists (200 μm 2-amino-5-phosphonovaleric acid, 2 mm Mg²⁺) were added to the culture medium.

Whole-cell patch-clamp recording from HEK 293 cells

Patch-clamp recording in the whole-cell configuration was made with an Axopatch 200B amplifier (Axon Instruments, Union City, CA, USA). Recording electrodes (5–12 MΩ) were filled with (mm): 140 caesium gluconate, 5 Hepes, 4 NaCl, 2 MgCl₂, 0.5 CaCl₂, 1 ATP, 0.3 GTP and 5 BAPTA (pH 7.4, 23°C). The recording chamber was continually perfused with recording solution composed of (mm): 150 NaCl, 10 Hepes, 1 CaCl₂, 3 KCl and 10–20 mannitol (pH 7.4 unless otherwise noted). Glutamate (100 μm) was applied using a theta glass pipette driven by a piezo-electric translator (Burleigh Instruments, Fishers, NY, USA). We have previously measured the solution exchange time of this system by stepping a cell held at 0 mV in the absence of agonist into a solution in which 150 mm NaCl was replaced by KCl (Mott et al. 2001). The time course of the change in the holding current had a 10–90% rise time of 16 ms and could be fitted with a single exponential component with a time constant of 7.3 ms.

Glycine (50 μm) was present in all solutions except where noted. Data used for analysis were collected within 5–15 min of obtaining the whole-cell patch configuration to minimize time-dependent changes in glycine-independent desensitization. In some experiments, series resistance was corrected off-line (Traynelis, 1998). Correction of series resistance did not alter the time constant for zinc-induced current relaxation.

Curve fitting and statistical analysis

The time course of desensitization was fitted with one or two exponential components with custom software NPM (S. F. Traynelis, Emory University) using the equation where τ is the decay time constant. The zinc concentration dependence for the rate of desensitization was used to calculate the zinc binding and unbinding rates using the equation: where τ_on is the time constant for fast desensitization and k_on and k_off are the zinc binding and unbinding rates, respectively. Concentration–response relationships for zinc inhibition were fitted with the equation where I₀ is the whole cell current in the absence of extracellular Zn²⁺, n is the Hill slope, and I_Zn is the residual current at saturating extracellular Zn²⁺.

An allosteric model was fitted simultaneously to the averaged NR1/NR2A response waveforms obtained under six different recording conditions (ChanneLab, http://www.synaptosoft.com). Proton concentrations were corrected with an activity coefficient of 0.8. Averaged waveforms were normalized to the largest amplitude, and fitted by comparing the normalized simulated currents and normalized response waveforms at each step of a simplex algorithm using least squares criteria. Simulations were generated using a 5th order Runge-Kutta numerical integration algorithm. All pooled data are expressed as mean ±s.e.m. Student's unpaired t test was used unless stated otherwise.

Results

Zinc and glutamate binding sites are allosterically coupled

Currents from HEK 293 cells expressing NR1/NR2A NMDA receptors were elicited by rapid application of 100 μm glutamate in the continuous presence of 50 μm glycine (Fig. 1A, holding potential V_hold=+50 mV). In the absence of free Zn²⁺ in the external solution (ambient zinc chelated by 10 μm EDTA), the current exhibits only a slow form of glycine-independent desensitization. When free Zn²⁺ is in the nanomolar range (free concentration buffered by tricine), a fast component of desensitization is observed. We have previously proposed that fast desensitization of NR1/NR2A receptors is caused by an allosteric interaction between the high-affinity nanomolar Zn²⁺ binding site on the amino terminal domain and the glutamate binding site (Zheng et al. 2001; see Monod et al. 1965). A key feature of this hypothesis is the proposed intra-subunit allosteric regulation between the NR2A amino terminal domain and the NR2A S1S2 glutamate binding domain. Figure 1B shows the difference in the IC₅₀ values for peak and steady-state currents, which approximate the apparent Zn²⁺ affinity for glutamate-unbound and glutamate-bound receptors, respectively. Figure 1C illustrates the dependence of the desensitization rate on extracellular Zn²⁺ concentration. The linear relationship of the rate of desensitization on the free zinc concentration is consistent with a single zinc binding event being the rate-limiting step in the desensitization process. Based on linear regression of data presented in Fig. 1C, the binding and unbinding rates for zinc are 3.40 × 10⁷m⁻¹ s⁻¹ and 1.38 s⁻¹, respectively (see Methods). From these microscopic rate constants, the K_d for zinc is calculated to be 40 nm, which is in agreement with the zinc IC₅₀ determined for the steady-state currents (26 nm; Fig. 1B).

Because glutamate binding causes both channel opening and the hypothesized switch in zinc affinity, it remains possible that it is channel opening rather than glutamate binding that causes the switch in zinc affinity. To test this hypothesis, a set of experiments parallel to those in Fig. 1 was performed in which glutamate was present continuously and the co-agonist glycine was rapidly applied (Fig. 2). If the intra-subunit allosteric hypothesis is correct, then pre-equilibration with glutamate will shift receptors into the high-affinity state for zinc binding even before glycine application. Glycine application would then lead to channel opening, but no apparent desensitization. On the other hand, if rapid application of glycine causes the same desensitization as rapid application of glutamate, it would suggest that zinc affinity is controlled by the conductance state of the channel pore.

Figure 2 shows that rapid application of glycine causes no desensitization and that the dose–response relationship for zinc inhibition of currents in response to glycine application exhibits high apparent affinity for zinc inhibition. Figure 2A plots the current from one representative cell using the same glutamate application protocol as in Fig. 1. Figure 2B shows the current from an independent representative cell using a protocol of rapid application of 100 μm glycine in the continuous presence of 50 μm glutamate. Note that there is little desensitization of the current, consistent with the receptors already being in a high-affinity state for zinc binding due to pre-equilibration with glutamate before activation of the current by glycine. Figure 2C illustrates the intra-subunit hypothesis that zinc affinity is controlled by the occupancy of the glutamate binding site and not by glycine binding nor channel opening. Consistent with this hypothesis, Fig. 2D shows that the IC₅₀ for zinc inhibition of the peak currents in response to glycine application is 29 nm, almost identical to the IC₅₀ for zinc inhibition of steady-state currents in response to glutamate application (26 nm) and a lower IC₅₀ than peak currents in response to glutamate application (123 nm, dashed line). The results are consistent with the zinc affinity being controlled by the occupancy of the glutamate binding site, rather than by the glycine binding site or opening of the channel pore. Thus, intra-domain interactions within the NR2A subunit appear to fully mediate the allosteric interactions between Zn²⁺ and glutamate.

An allosteric model accounts for the zinc and pH dependence of desensitization

In order to assess the zinc-induced desensitization in the context of the putative connection between zinc inhibition and regulation of NMDA receptors by protons, we recorded glutamate-induced whole cell currents over a range of Zn²⁺ concentrations and proton concentrations (pH). Figure 3 illustrates the pH dependence of zinc-dependent desensitization, with more desensitization observed when the proton concentration is elevated (lower pH). Multiple lines of evidence have thus far suggested that Zn²⁺ inhibition is mediated through an increase in sensitivity to tonic proton inhibition at physiological pH (Choi & Lipton, 1999; Low et al. 2000; Zheng et al. 2001). We therefore hypothesize that a positive allosteric interaction also occurs between Zn²⁺ binding and proton binding, although the compact nature of proton binding sites may create special properties for such an interaction (see Discussion). Figure 4 shows the simplest representation of this dual allosteric regulation between Zn²⁺–glutamate, and Zn²⁺–protons within a single glutamate-binding subunit, in which thermodynamic balance has been maintained. In this model, we have simplified all potential pre-gating steps (Popescu et al. 2004; Erreger et al. 2005) to a single closed–open transition (Lester & Jahr, 1992). This allows us to focus attention within the complex model on the dual intra-subunit allosteric interactions while still adequately describing macroscopic current and open probability.

To quantitatively test this model, the time course of currents over a range of zinc concentrations and pH values were recorded, and composite average traces obtained across multiple patches. These data were simultaneously fitted by this model (see Methods), which embodied the concepts of dual positive allosteric interactions between Zn²⁺ and glutamate as well as Zn²⁺ and protons (Fig. 4). Our goal in evaluating this model was to test whether we could predict the time course and amplitude of responses over a wide range of conditions by incorporating two ideas into the model: (1) glutamate binding increases the affinity for zinc, and (2) zinc alters protonation rates resulting in greater proton affinity. This model assumes that Zn²⁺-bound receptors can open, consistent with the significant plateau currents observed for Zn²⁺ inhibition curves at saturating Zn²⁺ concentrations (e.g. Williams, 1996; Paoletti et al. 1997; Low et al. 2000). By contrast, we assume protonated channels do not open, as suggested by Banke et al. (2005). Intuitively, this is a reasonable assumption since the concentration–response relationship for proton inhibition is complete, with no current detectable at saturating proton concentrations (Low et al. 2000). Moreover, known structural determinants for proton inhibition lie in the same region of the receptor hypothesized to be critical for channel gating, so it seems likely that protonation directly impacts the gating process (Low et al. 2003; Banke et al. 2005; Hu & Zheng, 2005). The model contains a slow desensitized state to account for the gradual decrement of the current even in the absence of Zn²⁺.

Before evaluating the ability of this model to predict the response time course that we recorded, we placed several simplifying constraints on the model. First, we assumed a proton association rate for all steps of 1.4 × 10⁹m⁻¹ s⁻¹ (Banke et al. 2005). We set the closing rate to 269 s⁻¹, which is the reciprocal of the mean open time previously reported in outside-out patches (3.72 ms from Erreger et al. 2005). We then set the opening rate to a value (560 s⁻¹) that gave an open probability of 0.5 at pH 7.3 (Erreger et al. 2005). The glutamate dissociation rate (80 s⁻¹) was selected to give a relaxation time course after a brief application of agonist with a time constant of 32 ms (Erreger et al. 2005). The glutamate binding rate was set to a value (8.4 × 10⁶m⁻¹ s⁻¹) that gave a glutamate EC₅₀ value of 4.4 μm at steady state (Zheng et al. 2001).

We subsequently fitted current responses recorded in the presence of EDTA at pH 6.7, 7.3 and 7.8 with the model varying only desensitization rates and proton dissociation rates. Once these rate constants were obtained, the slow desensitization rates were fixed in subsequent evaluation of the model in the presence of Zn²⁺. By implementing these constraints, assumptions and thermodynamic balance in loops, we reduced the full model to only five free parameters – the ‘on’ and ‘off’ rates for Zn²⁺ binding to receptors that lack glutamate, the ‘on’ and ‘off’ rates for Zn²⁺ binding to receptors that have bound glutamate, and the ‘off’ rate for protons from Zn²⁺-bound receptors.

We subsequently fitted a set of six composite averaged current response waveforms with this model, varying these five parameters using a least-squares criteria and a simplex algorithm (Fig. 5A). Figure 5A–C shows how this straightforward model with only five free parameters can predict the amplitude and time course of Zn²⁺ inhibition over a wide range of pH values and Zn²⁺ concentrations. Table 1 summarizes the rate constants for this fit. Despite several simplifying assumptions and constraints, the model successfully predicts the relative amplitudes and time courses for conditions across the full range of parameters for both Zn²⁺ and proton concentration. In addition, the model also predicts the effect of Zn²⁺ on the glutamate EC₅₀. The glutamate dissociation rate was set to give an EC₅₀ for steady-state currents of 4.4 μm (Zheng et al. 2001). The fitted model predicted an EC₅₀ for Zn²⁺-bound receptors of 1.8 μm, similar to the value (2.3 μm) previously observed in the presence of saturating Zn²⁺ (Zheng et al. 2001). In addition, the IC₅₀ value for Zn²⁺ inhibition of peak and steady-state currents was 141 and 64 nm, respectively, similar to previously determined experimental values (123 and 26 nm, see Fig. 1). Thus, to a first approximation, we can account for the main features of this data set by the dual positive allosteric interactions between Zn²⁺ and both proton and glutamate binding. We also refitted the data with a series of models in which the Zn²⁺-bound and Zn²⁺-unbound open states were uncoupled or had independent closing and/or opening rates (data not shown). In all cases, similar fits to the data were obtained, confirming that this allosteric representation of the interactions between Zn²⁺ and glutamate is relatively insensitive to the nature of the representation of the gating steps in the model.

Figure 5D illustrates one of the main hypotheses encapsulated in the model: the regulation of proton affinity by Zn²⁺ binding. The model was used to simulate the fraction of receptors in the protonated state at steady state in the presence of 1 mm glutamate as a function of Zn²⁺ concentration. The pK_a is the pH at which half of the proton binding sites are occupied. Zn²⁺ increases sensitivity to protons, which is manifested as a leftward shift in the functional proton concentration–response curve (Fig. 5E). Therefore the model accurately describes the established Zn²⁺-induced shift in the proton concentration–response relationship (Choi & Lipton, 1999; Low et al. 2000).

Sensitivity of simulated synaptic currents to Zn²⁺ inhibition

The complex nature of the effects of extracellular Zn²⁺ on NMDA receptor currents (as well as uncertainty about the magnitude and time course of Zn²⁺ concentrations at synapses) make prediction of the effects of Zn²⁺ on synaptic currents difficult. However, our elucidation of the rate constants describing Zn²⁺ binding, glutamate binding, as well as the allosteric interaction between Zn²⁺ and glutamate binding provides a tool that can be used to explore the conditions under which Zn²⁺ might have meaningful effects on synaptic NMDA receptor function. We initially estimated the sensitivity of synaptic responses (excitatory postsynaptic currents, EPSCs) to low steady-state concentrations of Zn²⁺ by driving the allosteric model described above with a synaptic glutamate concentration waveform similar to that described for hippocampal neurones, with a peak concentration of 1.1 mm and an exponential decay time of 1.2 ms (Clements et al. 1992). We evaluated the sensitivity of a single pulse of glutamate to different steady-state levels of Zn²⁺ at pH 7.3. Figure 6A confirms that NMDA receptors activated by brief synaptic-like stimuli can be inhibited by extracellular Zn²⁺ in a similar fashion to the inhibition observed for currents in response to a prolonged application of glutamate (Fig. 1). Note that the fast desensitization apparent during prolonged glutamate application (Fig. 1) does not occur under this protocol of a single brief glutamate exposure. Moreover, the occupancy of the zinc binding site remains almost constant as the ∼1 ms synaptic pulse of glutamate is insufficient in duration to significantly influence zinc binding. Nonetheless, exposure to steady-state zinc in the nanomolar range inhibits the amplitude of the EPSC. Figure 6B demonstrates the concentration-dependent nature of this inhibition.

We next investigated how Zn²⁺ would influence EPSC amplitude if it were released simultaneously with glutamate at a synapse. To accomplish this, we simulated coincident release of glutamate (1.1 mm, τ1.2 ms) and a constant amount of Zn²⁺ such that there is a sustained, step elevation of Zn²⁺ concentration. We chose to evaluate a step concentration change in Zn²⁺ since it would set an upper limit on the possible effects of Zn²⁺ co-released with glutamate, and because the time course of Zn²⁺ concentration changes in the synaptic cleft is unknown. Figure 6C demonstrates that 300 nm Zn²⁺ co-released with glutamate would have no impact on the amplitude of the synaptic current because the occupancy of the Zn²⁺ binding site does not become significant until well after the peak of the current. Figure 6D illustrates the relative insensitivity of an individual synaptic current to co-release of Zn²⁺ over a wide concentration range. Zn²⁺ dose–response data for peak current, total charge transfer, and decay time constant are plotted in Fig. 7. The IC₅₀ value for Zn²⁺ inhibition of a simulated synaptic NR1/NR2A current (160 nm, Fig. 7A) is similar to the low affinity inhibition observed for peak whole-cell currents in response to rapid application of glutamate (123 nm, Fig. 1). The sensitivity of the charge transfer (estimated under voltage clamp) is somewhat shallower than that of the peak response because the Zn²⁺ binding increases glutamate affinity and modestly slows the deactivation time course. Figure 7B demonstrates that zinc co-released with glutamate does not appreciably alter EPSC properties. These simulation results suggest that steady-state zinc has a much stronger effect on a single synaptic NMDA current than zinc co-released with glutamate. While all simulations in Figs 6 and 7 were performed at pH 7.3, a similar insensitivity to co-released zinc is predicted by simulations at lower pH. For example, at pH 6.8, 1 μm Zn²⁺ co-released with glutamate reduces the current by only 5.1%.

In order to evaluate the frequency dependence of zinc inhibition, we subsequently drove the same allosteric model presented above with a variable length train of synaptic glutamate pulses over a frequency range of 0.5–100 Hz (Fig. 8A). Figure 8B summarizes the effects of extracellular Zn²⁺ on total charge transfer associated with responses at different frequencies. As expected, these simulations show that increasing the synaptic stimulus frequency can cause more inhibition of charge transfer by Zn²⁺ due to the Zn²⁺-induced desensitization. This reflects, as described above, increasing occupancy of the glutamate binding site, the consequent allosteric coupling of glutamate binding to Zn²⁺ binding, and the resulting relaxation to a new equilibrium as Zn²⁺ binds to its site at a higher affinity. Interestingly, the IC₅₀ value for the responses at the end of a 1 s, 100 Hz train (136 nm, Fig. 8D steady state) is slightly lower than that associated with inhibition of a single pulse (160 nm, Fig. 7A), but quite similar to that associated with inhibition of a whole-cell current recorded in response to continuous application of glutamate (123 nm, Fig. 1A and B). These results suggest that any condition (such as high frequency stimulation) that promotes increased occupancy of the glutamate binding site will enhance Zn²⁺ inhibition by engaging the allosteric mechanism described here. These data suggest that the concentration–effect relationship for Zn²⁺ inhibition of synaptically evoked NMDA receptor responses resembles that for the peak current in our concentration-jump experiments. Thus, the conclusions drawn from this study of recombinant currents appear directly applicable to the synaptic situation of a brief glutamate time course.

Discussion

The NR2A subunit is expressed widely in the central nervous system. Recombinant NR1/NR2A receptors have been used extensively as a model system to investigate the underlying mechanisms of desensitization of native NMDA receptors (Krupp et al. 2002; Ren et al. 2003; Chen et al. 2004). The main finding of this study is that fast glycine-independent desensitization of NR1/NR2A is mediated by an allosteric interaction between the glutamate and zinc binding site, which are both known to reside within the NR2A subunit. Activation of NMDA receptor currents by rapid application of glycine in the continued presence of glutamate does not result in this form of desensitization, demonstrating that the zinc affinity is specifically regulated by glutamate binding, not by glycine binding and not by opening of the ion channel pore.

A hypothetical model for allosteric regulation of NR2A-containing NMDA receptors

Our working hypothesis is that the semi-autonomous amino terminal domain (ATD) and the glutamate-binding domain of NR2A (S1S2 domain) interact to produce use-dependent regulation of NMDA receptor function (Fig. 4). This hypothesis is supported by our demonstration that a simple allosteric model of glutamate and Zn²⁺ binding can quantitatively account for the magnitude and time course of zinc-induced desensitization. Moreover, incorporation into this model of the idea that Zn²⁺ binding changes the pK_a of a residue(s) that controls gating can further account for the pH dependence of Zn²⁺-induced desensitization (Zheng et al. 2001). The model represents the following sequence of events. First, glutamate binds to the S1S2 domain of NR2A. Second, the glutamate-binding S1S2 domain and the zinc-binding ATD interact in a manner that shifts the zinc binding site from a low to a high-affinity state following glutamate binding. Third, as the system relaxes into a new equilibrium, the occupancy of the zinc binding site increases in a time-dependent manner. Fourth, zinc binding to the ATD of NR2A causes conformational changes within the receptor that enhance the binding of protons to the pH-sensitive gating elements (Low et al. 2003), thereby reducing channel open probability (Traynelis & Cull-Candy, 1991; Banke et al. 2005). The enhancement of proton binding is sufficient to quantitatively account for the degree of inhibition by zinc (Fig. 5; Low et al. 2000; Zheng et al. 2001). In this model, the zinc-induced desensitization results from re-equilibration of zinc binding to the NMDA receptors and subsequent enhancement of tonic proton inhibition. The time course of zinc-induced desensitization represents the rate-limiting step, which is zinc binding, since glutamate binding and protonation both occur more rapidly than the time course measured for the fast component of desensitization (e.g. Lester & Jahr, 1991; Erreger et al. 2005; Banke et al. 2005).

The hypothetical model proposed here has a number of strengths and weaknesses. In terms of strengths, the model can reproduce the main features of the data by only incorporating a positive allosteric interaction between semi-autonomous domains that bind glutamate (S1S2) and Zn²⁺ (amino terminal domain), a positive allosteric interaction between Zn²⁺ binding and protonation rates, and thermodynamic balance. Moreover, the model can account for the data over a wide range of conditions (10-fold changes in Zn²⁺ and proton concentration). This quantitative representation of the working hypothesis strengthens the conclusions by demonstrating the compatibility of the hypothesis with data. The model also contains several weaknesses. For example, the Zn²⁺ affinity in simulations, which was previously reported to be enhanced at alkaline pH (Low et al. 2000), is modestly decreased in the model at alkaline pH. This may reflect a separate pH sensitivity of the histidine residues that chelate Zn²⁺ (Low et al. 2000), which is not incorporated into the model. In addition, the principle of thermodynamic balance as implemented in the model may not fully apply to protonation of an individual residue for the following reason. Binding of Zn²⁺ may alter the interaction of the ATD with a proton-sensing region of the receptor through a large number of van der Waals contacts, hydrogen bonds, and ionic or electrostatic interactions. It is likely that only a subset of these interactions is necessary to shift the pK_a of a key ionizable residue that is intimately involved in gating. This enhancement of proton inhibition probably involves alteration in the pK_a of the proton sensor through alteration in either the partial charge or hydrogen-bonded network around an ionizable residue. Therefore, it seems an oversimplification to expect protonation of a single residue to mediate the full complement of reverse effects on the full set of protein–protein contacts. The complexity of this issue will only be resolved as gating elements are determined at the atomic level, and the structural basis for pH sensitivity of gating becomes understood. Other shortcomings of the model are the simplification of the binding scheme to a single subunit, when it is clear that there are probably two NR2A subunits with unknown functional coupling. We have neglected any potential cooperativity between Zn²⁺-bound ATDs (Hatton & Paoletti, 2005). Finally, we have hypothesized that Zn²⁺-bound receptors can open on the basis that saturating Zn²⁺ concentrations can leave a pH-sensitive plateau current (Williams, 1996; Paoletti et al. 1997; Low et al. 2000). Yet we know little about the properties of the Zn²⁺-bound receptor, and thus the gating representation is an oversimplification. However, sensitivity analysis of the gating steps in the model show that the model can still reproduce all features of Zn²⁺ regulation of the receptor if one were to alter the closing rate of the Zn²⁺-bound open state, alter the opening rate of the Zn²⁺-bound closed state, or alter the Zn²⁺ sensitivity of this state (data not shown). That is, the model is insensitive to the kinetics of Zn²⁺-bound channel opening and does not require Zn²⁺ binding to the open state. One notable caveat to the interpretation of the simulations generated using the model is that we have assumed that pH is constant, although it is known that protons are released at synapses and can lower the pH of the cleft under some conditions (Krishtal et al. 1987; DeVries, 2001).

Allosteric regulation of other glutamate receptors

Whereas our data were obtained from a pure NR1/NR2A subunit population, there is strong evidence for incorporation of more than one NR2 subunit in an individual native NMDA receptor complex (Chazot et al. 1994: Sheng et al. 1994). A recent study with heterotrimeric NR1/NR2A/NR2B receptors has suggested that an individual NR2A ATD can bind Zn²⁺ with high affinity but will only induce a partial inhibition compared to having Zn²⁺ bind to both NR2 ATDs in the receptor–channel complex (Hatton & Paoletti, 2005). We hypothesize that this incomplete inhibition reflects a smaller Zn²⁺-induced shift in proton binding rates (and thus proton IC₅₀) compared to receptors that have heterodimeric NR1/NR2A subunit composition.

It is possible that the allosteric interaction described here might exist for other members of the glutamate receptor family and may be involved in the regulation of glutamate receptor function by endogenous modulators and compounds used therapeutically. A similar cooperativity has been suggested between the ifenprodil and glutamate binding site for receptors comprising NR1/NR2B subunits (Kew & Kemp, 1998; Zheng et al. 2001). The binding of phenolethanolamines (typified by ifenprodil) to the ATD of NR2B has also been demonstrated to enhance proton sensitivity in much the same way as zinc binding to the ATD of NR2A is hypothesized to enhance proton inhibition (Pahk & Williams, 1997; Mott et al. 1998; Zheng et al. 2001; Perin-Dureau et al. 2002). Similarly, a low affinity Zn²⁺ binding site has been proposed to also exist within the ATD of the NR2B subunit (Rachline et al. 2005). The amino terminal domain of other ionotropic glutamate receptor subunits may contain binding sites for other extracellular regulators, and such regulatory sites could be allosterically coupled to the agonist-binding site in the S1S2 domain. The allosteric interaction described here within the NR2A subunit may be a fundamental principle of the regulation of glutamate receptor function that could represent a target for the development of subunit-specific modulators.

Effects of Zn²⁺ on synaptically activated NMDA receptors

Our simulations clearly show that steady-state levels of Zn²⁺ can alter the response to synaptic stimuli by occupying the Zn²⁺ binding site in its lower affinity state before agonist binding. The model presented in Fig. 4 and Table 1 also allows us to evaluate the effect of Zn²⁺ released from synaptic vesicles, which has been a topic under intense investigation (Kay, 2003; Qian & Noebels, 2005; reviewed in detail by Frederickson et al. 2005). Although a consensus has yet to emerge on the concentration range and time course of synaptically released Zn²⁺ that may exist within the cleft at various synapses, our direct estimation of Zn²⁺–NR2A association rates allows the evaluation of whether synaptically released Zn²⁺ is capable of altering the synaptic NMDA channel response. To do this we used a step change in Zn²⁺ concentration coincident with simulated glutamate release since there is no information about the rate of clearance of Zn²⁺ from the synaptic cleft. Although such an event is unlikely to occur at synapses, it provides an opportunity to explore the limits at which Zn²⁺ could potentially alter the NMDA receptor response time course. The simulations demonstrate a relative insensitivity of individual synaptic currents to co-release of Zn²⁺. These results make intuitive sense given that the activation (10–90% rise time ∼7 ms) and relaxation of NR2A-mediated currents (τ∼30 ms) occur more rapidly than Zn²⁺ binding (Fig. 6). At the higher range of concentrations simulated for synaptic release of zinc (> 10 μm) binding proceeds rapidly. However, in this situation the subsequent re-equilibration of protonation becomes rate-limiting due to the Zn²⁺-induced slowing of proton unbinding, and thus activation still lags behind the rise time of the EPSC. This effect minimizes Zn²⁺-induced inhibition at the peak of the response. Although the peak amplitude of a single synaptic event appears insensitive to co-released Zn²⁺, it remains likely that co-released Zn²⁺ would have substantially greater effects on the response to high frequency stimulation, with the magnitude of the effect approaching that observed with steady-state levels of Zn²⁺ (Fig. 8). One caveat associated with the simulations of co-released Zn²⁺ is that low-affinity voltage-dependent channel block by Zn²⁺ becomes a significant factor at concentrations > 10 μm (Williams, 1996; Paoletti et al. 1997). This inhibition is not represented in our model, and may result in greater inhibition of EPSCs by Zn²⁺ than predicted by our modelling of high-affinity voltage-independent inhibition.

Our data suggest that NR1/NR2A receptors exhibit sensitivity to Zn²⁺ in the nanomolar range but only under conditions of steady-state Zn²⁺ or high frequency activity. Low frequency co-release of Zn²⁺ with glutamate is predicted to have a minimal direct impact on NR1/NR2A NMDA receptor synaptic currents, although synaptically released Zn²⁺ could be important in setting the steady-state levels. Whereas it is often suggested that NR2A-containing receptors may be tonically inhibited by Zn²⁺ due to the high-affinity binding site, we propose that not only the concentration but also the time course of both Zn²⁺ and glutamate release are critical in determining the effects of Zn²⁺ on synaptic NMDA receptors. While the IC₅₀ value for high-affinity inhibition of NR1/NR2A NMDA channels in sustained glutamate is ∼30 nm, our simulations suggest that the IC₅₀ for steady-state inhibition of synaptic NR2A-containing NMDA receptors is 160 nm Zn²⁺, implying that synaptic NMDA currents may not necessarily be under tonic inhibition by Zn²⁺. While the magnitude of both basal and activity-dependent Zn²⁺ concentrations at the synapse remains the subject of active investigation and debate, the current study places limits on the concentration and time course of Zn²⁺ required for inhibition of NR1/NR2A NMDA receptors and provides a framework in which to interpret new data on the magnitude and time course of Zn²⁺ at the synaptic cleft.