Mechanism of partial agonism in AMPA-type glutamate receptors

Abstract

Neurotransmitters trigger synaptic currents by activating ligand-gated ion channel receptors. Whereas most neurotransmitters are efficacious agonists, molecules that activate receptors more weakly-partial agonists-also exist. Whether these partial agonists have weak activity because they stabilize less active forms, sustain active states for a lesser fraction of the time or both, remains an open question. Here we describe the crystal structure of an α-amino-3-hydroxy-5-methyl-4-isoxazolepropionate receptor (AMPAR) ligand binding domain (LBD) tetramer in complex with the partial agonist 5-fluorowillardiine (FW). We validate this structure, and others of different geometry, using engineered intersubunit bridges. We establish an inverse relation between the efficacy of an agonist and its promiscuity to drive the LBD layer into different conformations. These results suggest that partial agonists of the AMPAR are weak activators of the receptor because they stabilize multiple non-conducting conformations, indicating that agonism is a function of both the space and time domains.

Figure 1. Structure of the GluA2 LBD tetramer bound by FW.

(a) Side view of the GluA2-toxin-(R,R)-2b-FW complex (PDB: 4u5c), with the subunits coloured as follows: A (green), B (red), C (blue) and D (yellow). The Con-Ikot-Ikot toxin is shown in orange and the agonist fluorowillardiine (FW) and the modulator (R,R)-2b are represented in black and cyan spheres, respectively. (b) The GluA2_NW structure (PDB: 4u4f) represented as in a. The partial agonist nitrowillardiine is coloured in black spheres. (c) Crystal structure of the soluble LBD in complex with FW (sLBD_FW) at 1.23 Å resolution. Domain 1 (light green) and domain 2 (green) harbour the agonist binding site. The inset shows the positive difference electron density omit map around FW contoured at 2 sigma. (d) The upper panel shows the tetrameric sLBD_FW structure oriented as in the full-length structures presented in A and B. FW is shown as black spheres. The relative dimer orientation of 120° is measured by the angle between the centre of mass between Ala665 and Leu748 of subunits A and C. The lower panel shows the top view (turned 90° along the x-axis) onto the crystal structure of sLBD_FW, which forms a tetramer with its symmetry mates. (e) In the left panel a cartoon showing the localization of the mutant L2, which was used to validate the sLBD_FW structure is presented. According to the histidine mutation model based on our sLBD_FW structure (right panel) the L2 mutant should trap between the subunits B (red) and D (yellow). To probe this interface, a binding site for Zn²⁺ was introduced by using the native histidine at position 435 and replacing residue K434 for a histidine. The fictive Zn²⁺ ion is shown as a green sphere and was manually placed between the four histidines.

Figure 2. L2 traps selectively in presence of partial agonists but not in the Apo or the fully active states.

(a,b) Left panels show the L2 mutant models based on the apo (PDB, 4u2p) and the fully active (PDB, 4yu0) structures. The trapping histidines in these models are too distant (distances are measured between the imidazole N atoms of residues His434 and His435 in subunits B & D) to allow trapping. The right panels show patch clamp experiments at low concentrations of glutamate of 10 μM, testing the Apo state and 10 mM, testing the full activated state after the application of CuPhen. As predicted from the modelled histidines we do not observe changes in the active fraction. (c) Dose–response curves in Glutamate and 100 μM CTZ for WT GluA2 (EC₅₀=450±70 μM) (open circles), the peak current, measured at the beginning of the application of Zn²⁺ for the mutant L2 (closed circles) (EC₅₀=350±30 μM). Zinc did not modify currents and the active fraction following zinc application (yellow circles) was fit by a linear function (c=0.99; n=4). (d) In the left panel, the L2 mutation modelled into the sLBD_FW structure could allow coordination of a fictive Zn²⁺ ion (green sphere) by K434H and H435 from subunits B and D with minimal conformational change. Dashed lines indicate distances between fictive Zn²⁺ and the imidazole N atoms. Trapping after the application of 100 μM FW and 10 μM Zn²⁺ (left panel) and 1 mM KA and 10 μM Zn²⁺ (right panel). Arrows indicate the reduction of the current after trapping in presence of Zn²⁺. Open circles indicate double exponential fits of the recovery after trapping.

Figure 3. Partial agonists trap the A–C interface.

(a) Top view of the LBD layer of the CA structure²³. Subunits A (green) and C (blue) are covalently linked by a disulfide bond formed between the introduced Cys665 on subunits A and C. The lower panel cartoon shows the localization of the position A665C. (b) Patch clamp experiments showing test pulses for cysteine mutations on positions 664-666. CTZ (100 μM) was present throughout the whole experiment. The recovery of the current in 10mM glutamate and DTT following trapping in the presence of CuPhen (10 μM) with 5-fluorowillardiine (FW) (right panel) or Kainate (KA) (left panel) was recorded. Arrows indicate a reduction of the current after trapping, which was observed for I664C, A665C and V666C but not for wild type (WT, bottom panel). Open circles indicate double exponential fits to the recovery after trapping. The time constants are summarized in Fig. 5. The difference between the active fraction after trapping in the presence of FW and KA for I664C, A665C and V666C compared with WT was significant for FW (P<0.005) and KA (P<0.05).

Figure 4. Failure to trap the D1 lateral interface is particular to saturating FW.

(a) Cartoon showing the localization of the mutants in the LBD layer. The red cross indicates that the HHH mutant is expected not to trap with FW, whereas the green ticks indicate proximity consistent with trapping. (b) Wild-type receptors did not show a decrease in the active fraction after the application of Zn²⁺ and FW. (c) The mutated residues are depicted in the sLBD_FW structure. Expected distances between the histidine imidazole N-atoms and fictive Zn²⁺ ions (green spheres) are denoted by dotted lines and expressed in Ångstroms. (d) Patch-clamp experiment showing the response of the HHH mutant, which was not modified in saturating FW. The difference between the active fraction after trapping in the presence of FW versus WT was not significant (P=0.5436). The difference between the active fractions after the trapping of T1 and HH was induced by FW compared with WT was significant (P<0.009). Arrows indicate the reduction of the current after trapping. The recovery of current in 10mM glutamate and EDTA following trapping (fit with double exponential function, circles) showed a common fast component, corresponding to the exchange of solution, and a slow component.

Figure 5. Summary of current recovery after trapping.

(a) The reduction of the active fraction and the time constants obtained from the double exponential fitting to the current during the recuperation after trapping (τ_fast and τ_slow) for cysteine mutants (for their locations, see Fig. 3). The fast time constant corresponds to glutamate activation, and the slow component is assumed to correspond to the breaking of the trapping bridge. (b) Equivalent data for histidine mutants (for their locations, see Figs 1 and 4). The graphs at the base of each column show the active fraction from all of the mutants versus the different partial agonist. The row end graphs show the individual mutants versus the different partial agonists. All errors are s.e.m., and for all data columns, n = 5 patches.

Figure 6. HHH traps at intermediate concentrations of FW.

(a) Trapping at different concentrations of FW. Three records from different patches showing trapping after the application of Zn²⁺ (10 μM) for 100 nM (dark blue), 3 μM (orange) and 100 μM FW (light blue). The inset shows the current after trapping displaying a robust reduction occurring at 3 μM. The fits (dots) are double exponentials fitted to the recovery after trapping the arrows and symbols indicate measurements plotted in c. (b) Currents from WT GluA2 were not modified at any concentration of FW. (c) Dose–response curves for FW in Zn²⁺ and 100 μM CTZ for WT GluA2 (EC₅₀=1.7±0.2 μM) (closed blue circles) and the mutant HHH (closed yellow circles) (EC₅₀=2.7±0.3 μM). The active fraction for WT GluA2 is flat (blue open dots). The active fraction for the HHH mutant (open yellow circles) is fit by a log-normal function with a maximum inhibition at 3 μM FW. All error bars are s.e.m. (n=5).

Figure 7. The promiscuity of cross-linking is inversely correlated with agonist efficacy.

(a) The table presents the cartoons of the closed angle, Loose and Tight conformations, showing the localization of the engineered bridges that either crosslink the upper D1 or the lower D2 lobe of the individual subunits (A in green, B in red, C in blue and D in yellow). The entries in the table indicate observed crosslinks for the different trapping mutations that were tested in the presence of KA, FW and glutamate. Full circles indicate trapping at saturating concentrations, whereas, half-filled circles indicate trapping at intermediate concentrations of the respective agonist. (b) Trapping profiles for all the mutants tested (different colours) in different partial agonists. Each plot compares the active fraction versus the time constant of recovery from trapping (τ_slow). All error bars are s.e.m. (n=5). (c) Each bridge has a different propensity to trap a range of distinct conformations (left panel). The crosslinking promiscuity (right panel), assessed over all agonist concentrations (filled symbols) or only for fully-bound LBD tetramers (open symbols) is plotted in relation to efficacy. A crosslink in saturating ligand was scored 1 unit, and in partially occupied LBD layers, a crosslink was scored as 0.25 units, because of the greater conformational variation in the latter case. (d) Cartoon illustrating the proposed mechanistic relation between conformational ensembles, efficacy and the number of different states trapped for the three classes of agonist.