Zinc Effects on NMDA Receptor Gating Kinetics

Abstract

Zinc accumulates in the synaptic vesicles of certain glutamatergic forebrain neurons and modulates neuronal excitability and synaptic plasticity by multiple poorly understood mechanisms. Zinc directly inhibits NMDA-sensitive glutamate-gated channels by two separate mechanisms: high-affinity binding to N-terminal domains of GluN2A subunits reduces channel open probability, and low-affinity voltage-dependent binding to pore-lining residues blocks the channel. Insight into the high-affinity allosteric effect has been hampered by the receptor's complex gating; multiple, sometimes coupled, modulatory mechanisms; and practical difficulties in avoiding transient block by residual Mg²⁺. To sidestep these challenges, we examined how nanomolar zinc concentrations changed the gating kinetics of individual block-resistant receptors. We found that block-insensitive channels had lower intrinsic open probabilities but retained high sensitivity to zinc inhibition. Binding of zinc to these receptors resulted in longer closures and shorter openings within bursts of activity but had no effect on interburst intervals. Based on kinetic modeling of these data, we conclude that zinc-bound receptors have higher energy barriers to opening and less stable open states. We tested this model for its ability to predict zinc-dependent changes in macroscopic responses and to infer the impact of nanomolar zinc concentrations on synaptic currents mediated by 2A-type NMDA receptors.

Introduction

Zinc is an essential micronutrient that accumulates in brain and is required for normal development and function (1). Both deficiency and excess of zinc alter behavior and can cause brain abnormalities and neuropathies, of which epilepsy, ischemia, and Alzheimer's degeneration have been the most studied (2,3). Aside from catalytic and structural functions in many proteins, ionic zinc (Zn²⁺) may play important roles in neurotransmission (4). Free Zn²⁺ accumulates in the synaptic vesicles of a specific subset of glutamatergic neurons and is coreleased with glutamate in an activity-dependent manner (5–7). Upon release, free Zn²⁺ may modulate neurotransmitter receptors and transporters, activate zinc-sensing metabotropic receptors, and/or gain cellular access through Ca²⁺-permeable channels (8–10). At certain glutamatergic synapses, a primary role for vesicular zinc is to reduce N-methyl-D-aspartate (NMDA) receptor currents (11). A wide range of extracellular Zn²⁺ concentrations directly and specifically inhibit NMDA receptor responses (12–16), and in the hippocampus, a region highly enriched in vesicular zinc, zinc-positive glutamatergic synapses are also enriched in NMDA receptors (17). The inhibitory effects of Zn²⁺ on NMDA receptors have received considerable attention due in part to the pivotal role played by these receptors in synaptic transmission and plasticity. Still, the mechanism by which the inhibition occurs is incompletely understood.

NMDA receptors are glutamate-gated channels responsible for the slow component of excitatory postsynaptic currents at central synapses. They assemble as tetramers of two glycine-binding GluN1 subunits and two glutamate-binding GluN2 subunits (18–20). Optional splicing of three exons in GluN1 subunits and differential expression of GluN2 subunits from four separate genes (A–D) result in substantial molecular diversity of NMDA receptors during development, across brain regions and during physiologic and pathologic conditions (21–24). The subunit composition of NMDA receptors determines their biophysical and pharmacological properties, including sensitivity to Zn²⁺ (25–28). The extracellular portion of each NMDA receptor subunit consists of an N-terminal domain (NTD) that binds allosterically active ligands including Zn²⁺, and a contiguous ligand-binding domain (LBD), which binds the obligatory coagonists glutamate and glycine. The LBD also connects with the transmembrane domain (TMD), which forms the ion-conducting pore and presumably the gate (Fig. 1 A) (29,30). The recently elucidated 3-D structure of a tetrameric receptor reveals that across subunits homologous modules organize in layers and identifies important intermodule interfaces that may mediate allosteric communication (31). Two types of Zn²⁺-binding sites are located on separate modules and inhibit NMDA receptor currents through distinct mechanisms.

Figure 1.

N/G responses have reduced sensitivity to block by extracellular Mg²⁺. (A, upper) Diagram of a 1/2A dimer highlights the low-affinity Zn²⁺-binding site within the TMD and the high-affinity Zn²⁺-binding site within the NTD (red). (A, lower) Topology of GluN subunits with expanded pore sequence illustrates the asparagine residues (boxed) involved in cation coordination and the pore sequence of the block-resistant construct. (B, upper) Whole-cell currents recorded in the presence and absence of EDTA (10 μM) reveal tonic inhibition of WT but not N/G receptors. (B, lower) Traces recorded in several external concentrations of Mg²⁺ (mM) illustrate the resistance of N/G receptors to block by external Mg²⁺. (C) Dose response of N/G (gray; 7.3 nM and 1.0 μM) and WT (black; 5.8 nM and 2.5 μM) receptors show similar sensitivities to nanomolar concentrations of Zn²⁺.

Several divalent cations including Mg²⁺, Ca²⁺, and Zn²⁺ bind with voltage-dependent affinities to residues in the TMD and transiently block current flow through open channels. Asparagine residues in the second transmembrane helix, N616 in GluN1-1a and N615 in GluN2A, contribute to the narrowest portion of the permeation pathway and are critical for voltage-dependent block (32–36). Glycine substitutions at these critical positions, GluN1-1a(N0G) or GluN2A(N+1G), increase pore diameter and reduce channel sensitivity to voltage-dependent block by orders of magnitude (32,35,37) (Fig. 1, A and B). In addition, zinc ions also produce voltage-independent inhibition of NMDA receptor currents. This effect is mediated by NTD of GluN2 subunits and has a purely allosteric mechanism: Zn²⁺ binding changes the receptor's gating kinetics but not its conductance (28,38–40). The NMDA receptor sensitivity to this type of inhibition is remarkably isoform-specific, with 2A receptors having high sensitivity (IC₅₀ = 5–30 nM) and 2B receptors having low sensitivity (∼1 μM) (25,27,28,41). The mechanism, although incompletely understood, is believed to be similar for both isoforms, with the large difference in sensitivity reflecting a much slower Zn²⁺ dissociation rate from 2A receptors (28).

Within the NTD of GluN2 subunits, Zn²⁺ binds to polar residues that line the cleft between two hinged lobes and stabilizes closed-cleft conformations (42–46). Glutamate binding to residues in the adjacent LBD facilitates this change in NTD structure, which in turn lowers the channel's open probability (P_o) (47,48). Because glutamate binding is required for channel opening and at the same time increases the NTD's affinity for Zn²⁺, when Zn²⁺ concentrations are subsaturating, the initial glutamate-elicited current declines relatively quickly as more receptors bind Zn²⁺, a phenomenon known as Zn²⁺-dependent desensitization (48,49). However, in saturating concentrations of Zn²⁺, when NTDs of resting receptors are fully occupied, this phenomenon is absent and Zn²⁺ inhibition reflects solely the altered gating kinetics of Zn²⁺-bound receptors (39). Why Zn²⁺-bound receptors gate with lower P_o than Zn²⁺-free receptors is unknown.

To focus specifically on this question, we examined the gating kinetics of block-resistant Zn²⁺-bound NMDA receptors over the entire range of experimentally resolvable gating transitions. We found that when compared to Zn²⁺-free receptors, Zn²⁺-bound receptors had higher energy barriers to opening and populated open states that were less stable. These changes caused Zn²⁺-bound receptors to dwell longer in pre-open states, to remain open for shorter durations, and to accumulate in desensitized states, although the desensitized states themselves were not altered. This redistribution of state occupancies provided a quantitative account for the observed lower P_o of Zn²⁺-bound NMDA receptors.

Materials and Methods

Rat GluN1-1a (NR1-1a, U08261) and GluN2A (NR2A, M91561) or GluN2A (N615G) were expressed together with green fluorescent protein in HEK293 cells. Single-channel currents were recorded with the cell-attached patch-clamp technique in extracellular solution containing (in mM) 150 NaCl, 2.5 KCl, 10 HEPBS (pK_a, 8.3), 1 Glu, and 0.1 Gly, adjusted to pH 8. EDTA or tricine were added as specified. Free Zn²⁺ concentrations in 10 mM tricine-buffered solutions were calculated using Maxchelator software (www.stanford.edu/∼cpatton/maxc.html) using K_d = 10⁻⁵ M, as previously reported (28). For processing and analyses, we selected only one-channel records that required minimal processing, as described in detail in a previous work (50). Idealization was done with the segmental-k-means (SKM) algorithm in QuB on 12 kHz filtered data, with 0.15 ms dead time. Modeling was done with the maximum-interval-likelihood (MIL) algorithm. Whole-cell currents (clamped at −70 mV) and excised patch currents (clamped at −100 mV) were recorded with intracellular solution consisting of (in mM) 135 CsF, 33 CsOH, 2 MgCl₂, 1 CaCl₂, 10 HEPES, and 11 EGTA, pH 7.4. Clamped cells were perfused with extracellular solutions containing (in mM) 150 NaCl, 2.5 KCl, 0.5 CaCl₂, 0.1 Gly, and 10 HEPBS, pH 8. Glutamate, chelators and tricine-buffered Zn²⁺ were included as specified. Differences were evaluated with two-tailed Student's t-tests assuming equal variance and were considered significant for P-values <0.05. (See Supporting Material.)

Results

To determine how Zn²⁺ binding to the NTD affects the gating sequence of GluN1-1a/GluN2A receptors (1/2A) it is necessary to evaluate quantitatively and compare all the kinetic species populated by Zn²⁺-free and Zn²⁺-bound receptors. This is experimentally challenging for several reasons: NMDA receptors gate with multistate kinetic mechanisms; several soluble ligands modulate channel kinetics and also have reciprocal effects on each other's potency (39,48); and divalent cations such as Mg²⁺, Ca²⁺, and Zn²⁺, which are experimentally difficult to control simultaneously, block the channel in a voltage-dependent manner, further complicating the single-channel signal.

We isolated allosteric effects from voltage-dependent effects by examining block-insensitive receptors, 1/2A(N615G), with stationary single-channel recordings from cell-attached one-channel patches clamped at low membrane potentials (+40 mV in the recording pipette) (35,37). To minimize the effects of the various ligands on each other's affinity, we used maximally effective concentrations of glutamate (Glu, 1 mM), glycine (Gly, 0.1 mM), and free Zn²⁺ (67 nM, 10 mM tricine with 200 μM added Zn²⁺), and minimally effective proton concentrations (10 nM, pH 8).

N/G receptors are resistant to Mg²⁺-block but retain high-affinity Zn²⁺ inhibition

First, to test whether the N-to-G substitution in the pore-lining region of GluN2A subunits eliminates block by trace divalent cations, we examined whole-cell responses from HEK293 cells expressing wild-type 1/2A (WT) or 1/2A(N615G) receptors (N/G). Adding a strong metal chelator (EDTA, 0.01 mM) to the extracellular solution potentiated WT responses but not N/G responses, indicating that, as expected, N/G receptors were rendered resistant to trace divalent cations (Fig. 1 B). Consistent with this interpretation, WT responses but not N/G responses, were reduced substantially by adding Mg²⁺ (0.07 or 0.7 mM) to the extracellular solution (Fig. 1 B). Importantly for our purpose, N/G receptors maintained high-affinity Zn²⁺ inhibition. Successive applications of increasing concentrations of nanomolar-range tricine-buffered Zn²⁺ reduced to the same extent the equilibrium WT and N/G whole-cell responses (IC₅₀, 5.8 nM and 7.3 nM, respectively) (Fig. 1 C). Based on these results, we conclude that N/G receptors are well suited for investigations into the mechanism of high-affinity Zn²⁺ inhibition.

We did not anticipate that the N-to-G pore substitution would affect receptor gating. However, the faster and deeper desensitization of N/G whole-cell responses was a strong indication of intrinsic changes in the activation of mutant receptors, which forced us to examine the kinetic mechanism of N/G receptors (Fig. 1 B). We recorded stationary currents from one-channel patches of cells transfected with WT or N/G receptors in the presence of saturating concentrations of agonists, with or without EDTA (1 mM), and +100 mV in the recording pipette. Consistent with the macroscopic measurements, WT receptors in the absence of EDTA displayed low single-channel activity (P_o = 0.10 ± 0.02, n = 8) and EDTA increased the observed activity (0.65 ± 0.04, n = 12) (Fig. 2, A and B, and Table S1 in the Supporting Material). This large increase in P_o reflected two major differences in the single-channel record, both attributable to relief from low-affinity voltage-dependent block. First, in the absence of EDTA, WT openings were short (mean open time (MOT) = 1.3 ± 0.1 ms) and of uniform duration (only one time component). The addition of EDTA revealed ∼10-fold longer openings (11.7 ± 0.6 ms) and complex open interval distributions. Second, when EDTA was omitted, bursts included frequent brief gaps responsible for the additional, sixth component in the closed-event distributions (τ_block = 1.2 ± 0.1 ms, a_block = 35 ± 1%). This single-channel structure indicated, most likely, that in the absence of EDTA, currents through long-lived open receptors were interrupted often by trace Mg²⁺-binding to pore residues. The apparent binding (1/MOT = ∼800 s⁻¹) and dissociation rates (1/τ_block = ∼1000 s⁻¹) were within the range reported previously for this receptor isoform, assuming ∼1 μM contaminant Mg²⁺ (51). As expected, the frequent 1-ms gaps observed for WT receptors in the absence of EDTA, which most likely represented Mg²⁺-bound open receptors, were absent in N/G records, whether EDTA was present or not (Fig. 2, C and D, and Table S1).

Figure 2.

N/G receptors are resistant to block by contaminant Mg²⁺ and have altered kinetics. Cell-attached traces (open is down) displayed at two time resolutions. Event histograms below each trace represent interval distributions for one record obtained with (A) WT with no added EDTA (104,352 events), (B) WT with 1 mM EDTA (509,043 events), (C) N/G with no added EDTA (188,412 events), or (D) N/G with 1 mM EDTA (416,360 events) (+100 mV in cell-attached pipette, pH 8).

Despite lacking Mg²⁺ block, N/G channels displayed lower activity than WT, indicative of mutation-induced changes in gating. EDTA slightly but significantly increased single-channel P_o of N/G channels, from 0.22 ± 0.05 (n = 7) to 0.36 ± 0.04 (n = 10) (P < 0.05), an indication that EDTA-chelatable allosteric inhibitors, most probably Zn²⁺, may also contaminate our recording solutions (49). However, comparing N/G and WT records obtained in the presence of EDTA, we were able to attribute the observed 45% reduction in P_o specifically to the N-to-G pore substitution. It is important to point out that although the N/G mutation results in measurable differences in pore diameter, as reported previously (35), it had no effect on the channel conductance for monovalent cations. In our recordings, where Na⁺ was the principal charge carrier, both the WT and N/G receptors had statistically indistinguishable single-channel amplitudes: 8.5 ± 0.4 and 9.6 ± 0.7 pA, respectively (P > 0.5).

Like WT, N/G records had five closed components (E₁–E₅) and two to four open components (τ_f and at least one of τ_L, τ_M, and τ_H), an indication that the mutation did not change the fundamental gating mechanism of NMDA receptors (49,52–55). The lower P_o measured in N/G records was fully accounted for by 2.5-fold longer closures and 30% shorter openings (Fig. 3 A). Except for the principal desensitized component τ_E5, which was unaffected by the mutation, all other detected closures were around twofold longer (Fig. 3 B and Table S2). The slight decrease in MOT reflected a fivefold decrease in the area of the longest open component (a_H), with no change in the individual open-component durations (Table S3). Based on this observation and the knowledge that within the NMDA receptor gating reaction open states are coupled, we opted to interpret the data with a simplified kinetic model, 5C1O, in which all open states are represented as a single aggregate state. Fitting this model to individual one-channel records within the two data sets, showed that the kinetic differences between N/G and WT receptors can be explained by slower rate constants for the three activation transitions: C₃→C₂, C₂→C₁, and C₁→O (Fig. 3 C). Based on these results, we suggest that by removing the asparagine side chain, which is hypothesized to face the channel lumen of GluN2A subunits, the N/G substitution increases the receptor's activation energy barriers (Fig. 3 D).

Figure 3.

Reaction mechanism of N/G receptors. (A) Single-channel kinetic parameters for N/G (white; n = 10, 8 × 10⁵ events) relative to WT (black; n = 12, 3.3 × 10⁶ events). (B) Closed-component time constants (τ_En) for N/G (white) relative to those for WT (black). (C) Reaction mechanisms estimated from fits of 5C1O models to individual one-channel records obtained from WT (upper) and N/G (lower). Rounded means of rate constants for each condition are in s⁻¹. (D) Relative free-energy profiles calculated from the mechanisms illustrated in C displayed relative to the initial liganded state C₃; off-pathway transitions into the long-lived C₄ and C₅ states are omitted. (E) Predicted fractional occupancies for WT and N/G receptors. Bars and whiskers represent means and standard errors; ^∗P < 0.05 in a Student's t-test.

Notably, these results demonstrate that the pore mutation had no observable effect on the kinetics of microscopic desensitization (Fig. 3 C), although the N/G macroscopic traces we measured desensitized around twofold faster (τ_D, 0.5 ± 0.1 s vs. 1.2 ± 0.1 s) and ∼1.3-fold deeper (I_ss/I_pk, 0.47 ± 0.05 vs. 0.60 ± 0.02) (Fig. 1 B). To investigate this possible discrepancy, we simulated macroscopic traces with the models in Fig. 3 C assuming equal glutamate association rates for N/G and WT receptors (1.7 × 10⁶ M⁻¹ s⁻¹) (56). Despite the differences in experimental conditions for the single-channel (cell-attached) and macroscopic (whole-cell) experiments, the models reproduced well the measured change in time course, predicting that N/G receptors will generate currents that desensitize 1.9-fold faster (1.2 s vs. 2.1 s) and 1.3-fold deeper (59% vs. 79%) (Fig. 1 B). Based on this result, we conclude that the measured increase in macroscopic desensitization produced by the N/G substitution is not indicative of changes in microscopic desensitization. Rather, it illustrates that delayed access into open conformations due to higher activation barriers causes the accumulation of N/G receptors in desensitized states (Fig. 3, D and E).

Zn²⁺ effects on the gating kinetics of N/G receptors

Having established that N/G receptors retain the fundamental gating mechanism of WT receptors and their high sensitivity to Zn²⁺ inhibition, and are resistant to Mg²⁺ block, we proceeded to examine how nanomolar concentrations of zinc change this receptor's gating reaction. We recorded, analyzed, and compared cell-attached one-channel traces obtained from Zn²⁺-free (10 mM tricine, n = 6) and Zn²⁺-bound N/G receptors (67 nM free Zn²⁺, n = 9) in the presence of high concentrations of glutamate (1 mM) and glycine (0.1 mM) and in low proton concentrations (pH 8) (Fig. 4, A and B). As an added precaution against voltage-dependent block, these recordings were done at lower membrane potentials, by applying only +40 mV in the recording pipette as compared to the +100 mV applied in the measurements described above. Zn²⁺ binding had no significant effect on single-channel current amplitudes elicited from N/G receptors (5.0 ± 0.4 pA vs. 5.5 ± 0.4 pA, P > 0.3), but Zn²⁺-bound receptors gated with threefold-lower P_o (0.10 ± 0.02 vs. 0.31 ± 0.07, P < 0.003) and had shorter openings (4.9 ± 0.4 ms vs. 7.8 ± 0.6 ms, P < 0.002) (Fig. 4 C and Table S4). These results are in agreement with conclusions reached previously based on single-channel measurements done at physiologic pH, which showed that: the high-affinity Zn²⁺ inhibition of 1/2A receptors occurs through an entirely allosteric mechanism, is incomplete, and results, at least in part, from shorter channel openings (39,57,58). In addition, because all our recordings originated from one-channel patches, we were able to determine that Zn²⁺-bound channels also had longer mean closed durations (63 ± 12 ms vs. 21 ± 4 ms, P < 0.02) (Fig. 4 C and Table S4).

Figure 4.

Allosteric effects of extracellular Zn²⁺ on individual N/G receptors. (A) Cell-attached recordings of N/G currents in the absence (left, 10 mM tricine) and presence of Zn²⁺ (right, 67 nM free Zn²⁺). Top traces illustrate a 25-s period that is expanded in the five traces beneath it, with the underlined segment shown at higher resolution at the bottom. (B) Histograms for the records illustrated in A (24 min, 122,562 events; and 27 min, 39,961 events). (Insets) Component time constants (τ) and areas (a). (C) Summary of kinetic parameters estimated for added zinc (gray, N/G with 67 nM free Zn²⁺, n = 9) relative to control (black, N/G with zero zinc, n = 6). Control open components (in ms): τ_f, 0.44; τ_L, 4.3; τ_M, 12; τ_H, 41. Control closed components (in ms): τ_E1, 0.41; τ_E2, 3.5; τ_E3, 9.8; τ_E4, 53; τ_E5, 3112. Bars and whiskers represent means and standard errors; ^∗P < 0.05 in a Student's t-test.

N/G receptors maintained the complex structure of open interval distributions characteristic of NMDA receptors even in the absence of strong metal chelators, an indication of their immunity to contaminant Mg²⁺. In some records, we could clearly separate four open components, a clear indication that Zn²⁺-bound receptors retain modal gating (56). Further, we were able to determine that the overall shorter openings reflected shorter openings in each mode (Fig. 4 C) and that zinc did not significantly change the relative areas of open components (P > 0.2). In contrast, the cause of longer closures could be traced specifically to longer intraburst closures, specifically τ_E2 (5.4 ± 0.5 ms vs. 3.5 ± 0.4 ms, P < 0.02) and τ_E3 (20 ± 3 vs. 10 ± 1, P < 0.02) (Fig. 4 C). It is important to note that we were able to determine that Zn²⁺ did not change the durations of desensitized intervals τ_E4 and τ_E5 but increased their relative areas from 3 ± 1% to 10 ± 2% (P < 0.01) and from 0.5 ± 1% to 1.3 ± 0.3% (P < 0.02), respectively (Fig. 4 C).

The observation that Zn²⁺ did not affect the number of components in closed- and open-interval distributions represents a strong argument that Zn²⁺-bound receptors activate with a basic mechanism similar to that of Zn²⁺-free receptors. For this reason, we chose to fit these two data sets by the same 5C1O model used to describe WT and N/G activity (Fig. 3 C). This analysis showed that Zn²⁺ binding further decreased the activation rate constants C₃→C₂, C₂→C₁, and C₁→O that were already slowed by the N/G mutation (Fig. 5 A). In addition, the closing rate constant O→C₁ was significantly faster, an indication that the open states of Zn²⁺-bound receptors were overall less stable. Notably, the rates at which receptors entered the desensitized states C₄ and C₅ were not significantly changed by Zn²⁺. Instead, the observed increase in the areas of long closed components was explained solely on the basis of longer dwells in states from which desensitized states can be accessed. Thus, the model postulates that although Zn²⁺-bound receptors have intact microscopic desensitization rates, at equilibrium, more receptors will accumulate in desensitized states, because by dwelling longer in preopen conformations they have more opportunities to transition into desensitized states. These analyses explain the allosteric effect of Zn²⁺ at NMDA receptors as an increase in the activation energy barriers to opening and a destabilization of open states (Fig. 5 B). They also predict a substantially increased occupancy of desensitized states (∼35–55%) at the expense of open states (approximately fourfold lower occupancy) (Fig. 5 C).

Figure 5.

Kinetic mechanism of NMDA receptor inhibition by Zn²⁺. (A) Reaction mechanisms estimated for N/G receptors with 10 mM tricine (n = 6, 5.7 × 10⁵ events) (upper) or with tricine-buffered zinc (free Zn²⁺, 67 nM) (n = 9, 7.5 × 10⁵ events) (lower). Rounded mean rate constants for each condition are in s⁻¹; ^∗P < 0.05 in a Student's t-test. (B) Free-energy profiles calculated from the kinetic models in A. The desensitized states C₄ and C₅ are omitted. (C) Changes in state equilibrium occupancies calculated from the models in A (bar graph) and predicted fractional occupancies for Zn²⁺-free and Zn²⁺-bound receptors (pie chart).

These conclusions necessarily rely on the assumption that the 5C1O model we used to evaluate single-channel activity is accurate, at least with regard to the salient features of Zn²⁺-induced changes in receptor gating. As a way of testing this assumption, we asked whether the mechanism we derived from single-channel data can match the known effects of Zn²⁺ on macroscopic currents, such as dose response of inhibition and kinetics of onset and recovery. We were successful in reproducing the dose response relationship illustrated in Fig. 1 C with the tiered model illustrated in Fig. 6 A, which has the following features. First, the upper arm represents the activation reaction of zinc-free N/G receptors: the reaction is initiated by the sequential binding of two glutamate molecules to resting (C_U) and monoliganded (C_M) receptor states, respectively (k₊ = 1.7 × 10⁷ M⁻¹ s⁻¹, k₋ = 60 s⁻¹, K_d = 3 μM) (59). Further, fully liganded receptors transition through: several preopen states (C₃-C₂-C₁), a collection of open states (C₁-O), and two separate desensitized states (C₃-C₅ and C₂-C₄) (50). Second, the lower arm of the tiered model represents the activation for Zn²⁺-bound receptors and postulates identical transitions that occur along a slightly altered energy landscape, as determined in this study (Fig. 5 B). Third, the model incorporates explicit Zn²⁺-binding and dissociation steps for glutamate-bound, but not for resting, receptors. This feature represents the higher affinity for Zn²⁺ of glutamate-bound receptors, as shown previously by elegant experiments from the Traynelis group (48).

Figure 6.

Zn²⁺ effects on nonstationary NMDA receptor responses. (A) Tiered model represents glutamate binding and the gating reaction of Zn²⁺-free receptors (upper), the binding of Zn²⁺ (vertical arrows, blue), and the gating reaction of Zn²⁺-bound receptors (lower, gray-shaded). Except for binding reactions, which are in M⁻¹ s⁻¹, all other rate constants are in s⁻¹. [Glu] was set to 1 mM (G) and [Zn²⁺] to 67 nM (Zn). (B) Simulated responses with the N/G model in A during a 5-s pulse of 1 mM Glu in the absence (black), coapplication (blue), or ambient presence (red) of 67 nM Zn²⁺. (C) Responses from WT receptors measured in excised patches after fast application of 1 mM Glu for 5 s (left) or 1 ms (right). Conditions are as in B.

We used this model to simulate the effects of several Zn²⁺ concentrations (0–300 nM) on Zn²⁺-free resting receptors. All 100 receptors invoked initially occupied the C_U state of the upper tier arm, were stimulated with a prolonged pulse of 1 mM Glu, and had intermittent pulses of Zn²⁺ superimposed on the Glu pulse. We obtained the best match between simulated traces and measured whole-cell responses by assuming Zn²⁺ rate constants of: k₊ = 7 × 10⁷ M⁻¹ s⁻¹ and k₋ = 0.5 s⁻¹. These corresponded to our experimental value of K_d = 7 nM (N/G receptors, pH 8) and were very close to values reported previously for WT receptors at pH 7.4 by two groups: k₊ = 10⁸ M⁻¹ s⁻¹; k₋ = 0.6 s⁻¹ (28) and k₊ = 3.4 × 10⁷ M⁻¹ s⁻¹; k₋ = 1.4 s⁻¹ (49). These results illustrate that the relatively simple operational model we derived from single channels residing in cell-attached patches reproduced well salient behaviors of macroscopic whole-cell responses.

Next, we asked how Zn²⁺ may affect NMDA receptors in physiological situations that at present are technically off-limits to direct observation, namely, how synaptic NMDA receptors may be affected by nanomolar concentrations of free Zn²⁺ when 1), the modulator is coreleased with the neurotransmitter glutamate, or 2), the modulator predates synaptic release, as, for example, during high-frequency trains. We set out to mimic these situations with two stimulation paradigms: 1), brief (1 ms) pulses of glutamate (1 mM) (60,61), and 2), seconds-long glutamate applications. For each, the following conditions were compared: 1), 1 mM Glu in zero Zn²⁺ (using only the upper arm of the model); 2), 1 mM Glu and 67 nM coapplied Zn²⁺ (tiered model with all receptors starting in the Zn²⁺-free resting state); and 3), 1 mM Glu applied onto receptors preequilibrated with Zn²⁺ (using only the lower arm of the model). Results showed that for brief and long pulses alike, the peak N/G response was not affected by coapplied Zn²⁺, but was dramatically reduced (>50%) when Zn²⁺ predated the Glu pulse (Fig. 6, B and C). In addition, as observed after long pulses (5 s), coapplied Zn²⁺ increased dramatically the rate and degree of macroscopic desensitization. This result represents additional validation for the model, because it demonstrates the model's ability to reproduce a prominent and well-known feature of Zn²⁺ effects on macroscopic NMDA receptor responses: the fast zinc-dependent component of macroscopic desensitization (25,28,48).

Zn²⁺ effects on synaptic-like NMDA receptor responses

Finally, to test whether the predictions made based on experimental results obtained from block-resistant N/G receptors also held true for WT receptors, we measured currents elicited by 5-s or 1-ms fast applications of 1 mM Glu onto excised patches of HEK293 cells expressing WT GluN1/GluN2A receptors. Responses were recorded successively from the same patch in the absence of Zn²⁺ (10 mM tricine); in the ambient presence of 67 nM free Zn²⁺ (preequilibrated); and when 67 nM free Zn²⁺ was coapplied with Glu (Fig. 6 C). Results showed that coapplied Zn²⁺ had no effect on the peak or time course of NMDA receptor responses to brief stimulation, as may be the case during normal transmission. However, when Zn²⁺ predated the Glu pulse, responses had drastically lower peaks, desensitized faster, and equilibrated to lower activity levels. Traces elicited with 5-s coapplications of Zn²⁺ and Glu revealed the same effects on WT currents as predicted from the N/G model: faster and deeper desensitization. The model explains this result by postulating a combined effect of nanomolar concentrations of co-applied Zn²⁺ on NMDA receptor responses to fast applications of Glu: Glu-elicited increase in Zn²⁺-affinity and altered activation kinetics of Zn²⁺-bound receptors.

Discussion

In this article, we examined how zinc binding to the N-terminal domains of NMDA receptors affected the receptors' gating kinetics. We compared current traces obtained from individual block-resistant receptors in the absence and presence of nanomolar concentrations of extracellular zinc and found that zinc shortened all openings, increased the duration of specific closures within bursts, and had no effect on intervals residing outside bursts. Further, based on these data, we derived a gating model for zinc-bound receptors that explained the inhibitory effect of zinc as a combination of increased energy barriers to opening and less stable open states. The deduced mechanism predicted correctly the known effects of zinc on macroscopic responses, including dose response and macroscopic kinetics. Last, we used the model to ask how nanomolar zinc concentrations may influence NMDA receptor responses to physiologically relevant stimulation protocols. Responses recorded from WT receptors residing in excised patches exposed to fast pulses of glutamate confirmed that zinc had no effect on peak currents when coapplied with glutamate, but it strongly decreased receptor-mediated charge transfer when zinc predated the glutamate pulse.

Together, these results make a strong case for the mechanism of allosteric zinc inhibition proposed here. However, a good fit to experimental data was possible only after making a number of experimental simplifications and theoretical assumptions. Keeping these in mind helps to place results in perspective and identifies current obstacles toward the goal of understanding in quantitative detail the mechanism of synaptic transmission at glutamatergic synapses. Among the experimental limitations faced, we note that NMDA receptor kinetics vary considerably, though not fundamentally, between preparations. Perhaps recordings from cell-attached patches of intact cells are closest to native responses, given the observation that recordings from whole cells and excised patches both illustrate changes in kinetics related to the time of membrane rupture (62–65). Thus, without knowledge about the mechanisms responsible for these changes in kinetics, quantitative fits cannot be expected. Despite these limitations, the operational model developed in this study reproduced with useful accuracy the behaviors of individual receptor molecules and allowed important, testable predictions about Zn²⁺ effects on NMDA receptors.

In general, kinetic models developed so far for NMDA receptors have been useful in uncovering new aspects of the receptor mechanism precisely because they offer a general agreement between single-channel behaviors and macroscopic recordings (52,54–56,65,66). This advancement is yet to be accomplished for any other ionotropic glutamate receptors; however, even NMDA receptor models offer only coarse simplifications of the receptors' conformational landscapes and are far from representing all functional states known to be present in solution. To enumerate only the most obvious simplifications, the model ignores any influences from glycine and proton binding equilibria. This is an acceptable simplification only when using concentrations outside the physiologic range (in this study, 0.1 mM glycine, pH 8). However, in situ and under nonstationary conditions, these and other modulators will exert changes on the NMDA receptor response that the model does not anticipate. In particular, the mechanism by which zinc and protons collaborate to modify receptor gating and how glutamate dissociation from Zn²⁺-free receptors modulates the deactivation time course require further investigation.

In summary, we developed a kinetic model for zinc inhibition of NMDA receptor kinetics by observing the effects of zinc on NMDA receptor single-channel and macroscopic responses. The model explains zinc inhibition by postulating a free-energy landscape of zinc-bound receptors that has higher energy barriers to opening and higher-energy open sates. Our results have two important implications. First, they show that zinc-binding to the N-terminal domain has no effect on the receptor's microscopic desensitization kinetics, thus endorsing the view that the molecular determinants of receptor desensitization are separate and independent from conformations controlled by high-affinity zinc-binding. Second, our results suggest that synaptic zinc will strongly inhibit NMDA-receptor-mediated charge transfer only when zinc is present in the cleft before stimulation, perhaps due to slow clearance of zinc released by preceding stimuli. This may represent a mechanism for synaptic short-term depression of the NMDA receptor response and may explain some of the unique features of zinc-containing glutamatergic synapses.