Neurotransmitter Funneling Optimizes Glutamate Receptor Kinetics

Abstract

Ionotropic glutamate receptors (iGluRs) mediate neurotransmission at the majority of excitatory synapses in the brain. Little is known, however, about how glutamate reaches the recessed binding pocket in iGluR ligand-binding domains (LBDs). Here we report the process of glutamate binding to a prototypical iGluR, GluA2, in atomistic detail using unbiased molecular simulations. Charged residues on the LBD surface form pathways that facilitate glutamate binding by effectively reducing a three-dimensional diffusion process to a spatially constrained, two-dimensional one. Free energy calculations identify residues that metastably bind glutamate and help guide it into the binding pocket. These simulations also reveal that glutamate can bind in an inverted conformation and also reorient while in its pocket. Electrophysiological recordings demonstrate that eliminating these transient binding sites slows activation and deactivation, consistent with slower glutamate binding and unbinding. These results suggest that binding pathways have evolved to optimize rapid responses of AMPA-type iGluRs at synapses.

Keywords: electrophysiology; free energy; glutamate receptors; ligand binding; ligand-gated ion channels; molecular dynamics simulations.

Figure 1

Dynamics of Glutamate Binding Time points (lower left of each panel) are relative to the start of Movie S1. This trajectory corresponds to the LBD dimer system T_dim1 (see Table S1); only the LBD that binds glutamate is shown for clarity. (A) Prior to ligand entry to the binding pocket, the LBD is open; (ξ₁, ξ₂) = (12.3, 12.2 Å). The ligand’s γ-carboxylate contacts R453 on Lobe 1. (B) Close-up view of (A). (C) Glutamate slips into the binding cleft. (D) The ligand contacts R661 on Lobe 2. (E) A metastable interaction forms across Lobes 1 and 2. The ligand’s γ-carboxylate contacts E657, R660, and R661 on helix F, and the ligand’s α-carboxylate contacts R485. R485 flickers out of the binding pocket to interact with the ligand. (F) R485 relaxes toward the binding pocket. (G) The metastable interaction at the ligand’s α- and γ-carboxylate between Lobes 1 and 2 persists. (H) The ligand's α-carboxylate and amine move to interact with binding pocket residues in Lobe 1. (I) The ligand shifts into the binding pocket, with its α-carboxylate contacting R485. Lobe 2 interactions with helix F are broken. In the pocket, the ligand’s amine is coordinated by P478 and L480. Cleft closure is initiated once helix F undergoes a backward tilt to form a pocket for the ligand’s γ-carboxylate. (J) Glutamate adopts the crystallographic conformation. (K) As the cleft closes to secure the ligand, the ligand’s amine contacts E705 on Lobe 2, and the ligand’s γ-carboxylate contacts the backbone amine of S654. (L) Expanded view of (K). The LBD closes around the ligand in the crystallographic conformation: (ξ₁, ξ₂) = (11.8, 10.8 Å). See also Figures S1 and S8, Movie S1, and Table S1.

Figure 2

Glutamate Binding Pathways and Metastable Binding Sites (A) The PMF calculated from the ligand density using a hard-sphere van der Waals approximation on a grid spacing of 0.5 Å along the x, y, and z axes contoured at 1.89 kcal mol^–1. Data are from the monomeric LBD system with 10 ligands (T_mon2,3). The primary binding pathways for glutamate are depicted by arrows. (B) The PMF, contoured at 1.16 kcal mol^–1, defines the regions of metastable interaction, sites 0–3. Site 0 is the site of stable binding. Site 1 is shared by Pathways 1 and 2, site 2 is encountered in Pathway 2, and site 3 is encountered in Pathway 3. Site 1 spans the two lobes, involving R453 in Lobe 1 and E657, R660, and R661 in Lobe 2. Site 2 involves E657, R660, and R661 in Lobe 2. Site 3 involves R675 and R684 in Lobe 2. The residues involved in site 0 are shown in Figure 3A. Error analysis is provided in Figure S2. (C) The PMF, contoured at 0.32 kcal mol^–1, shows the global free energy minimum. This minimum overlaps well with the ligand density derived from crystal structures of the glutamate-bound complex (e.g., PDB: 1FTJ). (D) The one-dimensional representation of the WT GluA2 ligand-binding PMF was obtained by first computing the three-dimensional PMF (A–C). Values of the three-dimensional PMF were indexed increasing in the z, y, then x directions, from (x_min, y_min, z_min), to produce the one-dimensional representation. The positions of sites 0–3 are indicated. Site 0 is the global free energy minimum and is set to 0 kcal/mol; sites 1–3 form local minima with free energies of 0.29, 0.83, and 0.75 kcal/mol, respectively. See also Figures S2 and S8.

Figure 3

Conformations of Bound Glutamate and the LBD (A) The bound ligand conformation, taken from PDB: 1FTJ. The ligand’s α-carboxylate contacts R485, and its γ-carboxylate contacts the backbone amine of S654. The ligand’s amine is coordinated to the side chains of T480 and E705 and to the backbone of P478. (B) The inverted conformation of the ligand. The ligand’s α-carboxylate contacts the backbone amine of S654, and its γ-carboxylate contacts R485. The ligand’s amine is coordinated to E705. P478 has moved upward to accommodate a water molecule that also contacts the ligand’s amine. (C) The two-dimensional order parameter (ξ₁, ξ₂) used to characterize large-scale conformational transitions in the GluA2 LBD. ξ₁ and ξ₂ each indicate the distance between the centers of mass of the clusters of atoms shown in blue and green, respectively. (D) (ξ₁, ξ₂) measures the degree of cleft closure for the LBD in apo (blue) and ligand-bound conformations. The ligand occupies either the crystallographic (orange) or the inverted (green) poses. Each point represents a snapshot taken every 120 ps from simulations of the monomer system at 3.9 mM glutamate concentration (T_mon1,4-8). The marginal histograms indicate distribution densities. The “X” indicates the degree of cleft closure in the crystal structure (PDB: 1FTJ). See also Figures S3 and S4 and Movies S2 and S3.

Figure 4

Activation and Deactivation of Receptors with Mutations in Pathway 1 (A) Sites of mutations that were tested functionally. The green-colored side chains correspond to the WT residues. The tan-colored residues contact bound glutamate directly but were not mutated in the functional tests. (B) The blue oval indicates LBD residues proximal to metastable site 1 (Figure 2B) in Lobe 1. R453 interacts with the ligand in the simulations. Mutants tested include single-charge swaps R453D and K458D, and the triple mutant R453A-D456A-K458A (RDK-AAA). (C) Activation of Lobe 1 mutants (R453D, blue; K458D, green; RDK-AAA, red) by a long pulse of 10 mM glutamate. Solution exchange measured after the experiment is shown as the upper black trace. A typical WT GluA2 response is plotted with a dotted line. The individual 10%–90% rise times of the currents (t_rise) are shown in the bar chart in the right panel, with the WT mean value as a dashed gray line. Asterisk indicates p < 0.005, Student’s t test. (D) Left panel shows deactivation of Lobe 1 mutants in response to ∼1 ms pulse of 10 mM glutamate. Color coding is as in (C), with monoexponential fits indicated by open circles. Right panel shows bar chart of individual deactivation decay values. Asterisk indicates p < 0.005, Student’s t test. See also Figures S1, S5, S6, and S8 and Table S3.

Figure 5

Activation and Deactivation of Receptors with Mutations in Pathways 1 and 2 (A) Residues in Lobe 2 that interact with the ligand at metastable sites 1 and 2 (Figure 2B) include R660, R661, and E657 (blue oval). The following mutations were made to test these sites: R660E, R661E, and E657A-R660A-R661A (ERR-AAA). (B) Left panel shows activation of Lobe 2 mutants (R660E, green; R661, blue; ERR-AAA, red) in response to a long pulse of 10 mM glutamate. The individual 10%–90% rise times of the currents (t_rise) are shown in the bar chart in the right panel, with the WT mean value as a dashed gray line, with asterisk indicating p < 0.05 versus WT from t test. (C) Left panel shows deactivation of Lobe 2 mutants following a 1 ms pulse of 10 mM glutamate, with color coding as in (B). Individual rise times are plotted in the right panel. (D) The affinity for glutamate is unchanged for the ERR-AAA mutant relative to WT. Dose-response curves in glutamate, measured at the peak current response, for WT GluA2 (EC₅₀ = 330 ± 90 μM; black circles), and for the mutant ERR-AAA (EC₅₀ = 410 ± 30 μM, red circles). By comparing the fits to responses from individual cells, the glutamate EC₅₀ for ERR-AAA was indistinguishable from that of WT GluA2 (p = 0.5; t test, n = 3). (E) The PMF for the ERR-AAA mutant, contoured at 2.62 kcal/mol. Data are from the monomeric LBD with 10 ligands (T_mut1-4). The PMF shows a loss of ligand density along Pathways 1 and 2, proximal to helix F. The ERR-AAA ligand-binding pathway, indicated by the red arrow, resembles Pathway 3 of the WT LBD. Interactions between R684 and R675 on Lobe 2 at site 4 are preserved in both the mutant and WT protein. These residues metastably interact with the ligand prior to binding in T_mut1. See also Figure S6, Tables S3 and S4, and Movie S4.

Figure 6

Activation and Deactivation of Receptors with Mutations in Pathway 3 and Off-Pathway Mutants (A) The blue ovals indicate residues of the LBD that interact with the ligand. K449 participates in transferring the ligand from metastable site 3 to site 0 (Figure 2B). D447 and K449 were mutated to alanine (DK-AA). R684 participates in site 3, and two residues, R684 and E688, were separately mutated to alanine (RE-AA). (B) The activation of receptors in response to a long pulse of 10 mM glutamate was slower than WT GluA2 (dashed line) for both DK-AA (blue trace) and RE-AA (red trace). The upper black trace shows solution exchange. Individual 10%–90% rise times are plotted in the right panel with the mean value for WT GluA2 indicated by a dashed line and standard error shaded in light gray. Asterisks indicate p < 0.01 versus WT GluA2, t test, n = 3–6. (C) Deactivation of DK-AA and RE-AA mutants in response to a 1 ms pulse of 10 mM glutamate with color coding as in (B). Monoexponential fits are represented by open circles. Individual deactivation decay constants are plotted in the bar graph (right). The asterisk indicates p < 0.05 versus WT GluA2, t test, n = 3–6. (D) Blue ovals indicate positions in the LBD of two triple-alanine mutants, located away from the binding pathways (Figure 2A). One set of off-pathway mutants is located in Lobe 1 (K409A-K410A-E422A, KKE-AAA) and one set in Lobe 2 (R715A-K716A-D769A, RKD-AAA) (Figure 4A). (E) Left panel shows activation of receptors by long pulses of 10 mM glutamate. Responses for the KKE-AAA (blue trace) and RKD-AAA (green trace) mutants are overlaid with a typical WT GluA2 response as in (C), with individual rise times plotted in the right panel. Asterisk indicates p < 0.05. (F) Left panel shows deactivation of off-pathway mutants following a 1 ms pulse of 10 mM glutamate. Color coding as in (E). Individual rise times are plotted in the right panel. Asterisk indicates p < 0.01. See also Figure S7 and Table S3.

Figure 7

Kinetic Modeling of AMPA Receptor Currents Activated by Glutamate Recapitulates the Effect of Slowed Binding Reactions on Receptor Activation and Deactivation (A) Kinetic model constructed according to principles outlined in previous studies of GluA2 kinetics (Robert et al., 2005). Four glutamate-binding sites, two open states (green, ^∗), and five desensitized states (red) are included. Low conductance open states and connections between desensitized states were omitted for simplicity. For each simulation, the association and dissociation rates were multiplied by a common factor, f, to represent the effects of mutants to slow binding rates through disruption of glutamate binding pathways. The rate constants were as follows: beta = 5,000 s^–1, alpha = 3,000 s^–1, k+ = 5 × 10⁶ M^–1 s^–1, k– = 10,000 s^–1, d+ = 250 s^–1, d– = 60 s^–1, d₀+ = 1 s^–1, d₀– = 9 s^–1. The conductance of the open state A4R^∗ was set at twice that of A3R^∗. (B) Example simulated currents (normalized to the maximum possible response) for a realistic concentration jump of 800 μs with rise time of 300 μs, using the model in (A). Six color-coded current profiles generated using the RCJ scripts (see STAR Methods) are shown, with the binding rate factor ranging from 2 to 0.05. Note that deactivation is more strongly affected than activation (because efficacy for channel opening is >1). Also, for fast binding reactions, the rise time is faster than the solution exchange. (C) Kinetic measurements from GluA2 mutants and WT (black circles, WT GluA2 marked) were well described by a linear fit with slope of ∼8 with intercept close to the origin. 95% confidence intervals for the line are shown as gray dashed curves. Uncertainties in both abscissa and ordinate were used for the fit (ODR = 2 in Igor 7). The reduced chi-square value for this fit is ∼6, meaning that the linear description of the data is, at best, approximate. Off-pathway control mutants (red circles) lie away from this curve. Data from simulations with different rise times for the glutamate pulse are plotted as open symbols, with the relevant f value for each group of simulations indicated with a dotted line. The colors of these symbols relates to the simulated 10%–90% solution exchange time seen by receptors. All solution exchange rates predict the similar steep, approximately linear relations between activation time (10%–90%) and decay time constant. The simulated kinetics span a similar range to the electrophysiological recordings of pathway-disruption mutants. The best agreement between simulation and experiment comes from solution exchange times at an intact patch in the physically plausible range of 200–400 μs.