Engineering a high-affinity allosteric binding site for divalent cations in kainate receptors

Abstract

Kainate receptors are allosterically regulated by sodium ions. Removal of Na+ from the extracellular solution, or replacement of Na+ by larger monovalent cations, inhibits kainate receptor activity. Sodium binds at a negatively charged cavity in the extracellular neurotransmitter binding domain that is capped by a small hydrophobic residue. Prior work revealed that introduction of aspartic acid at this site strongly reduces GluK2 sensitivity to monovalent cations of different size. We found that the GluK2 M739D mutant is also insensitive to substitution of Na+ by the large organic cations Tris and NMDG. Because these are excluded from the Na+ binding site by steric hindrance, we investigated the possibility that divalent cations can substitute for Na+. We show that in Na+ free solutions with low concentrations of Ca2+ and Mg2+ the GluK2 M739D mutant is inhibited by EDTA; that divalent ions in the micromolar concentration range act as positive allosteric modulators; and that the chemistry of the mutant cation binding site is typical of Ca2+ and Mg2+ binding sites found in protein crystal structures. Hence, the apparent insensitivity of the M739D mutant to monovalent cations is due to the adventitious allosteric effects of divalent ions at physiological concentrations and below.

Figure 1. GluR6 M739D mutant is relatively insensitive to NMDG and Tris

(A) In normal saline (Na), outward currents at + 60 mV due to wild-type GluR6 receptors activate rapidly in response to a 100 ms pulse of 10 mM glutamate and then desensitize (open circles represent a monoexponential fit, k_des = 110 s⁻¹). In the same patch, when external sodium is replaced by either Tris or N-methyl-D-glucamine (NMDG), outward currents are strongly inhibited. The lower trace shows the solution exchange into glutamate, measured after the experiment. (B) Expanded view of the current response in NMDG and Tris at +60 mV. The response in Na is shown as a dashed line. Monoexponential fits are shown with open circles. The rate of desensitization is ~ 15-fold faster (in Tris, 1600 ± 400 s⁻¹ and in NMDG, 1600 ± 100 s⁻¹ n = 5) than in sodium-containing solutions. (C) The desensitization and peak current of the GluR6 M739D mutant is much less sensitive to substitution of sodium with NMDG or Tris. Monoexponential fits are shown as open circles (in this patch, Na k_des = 99 s⁻¹; NMDG k_des = 110 s⁻¹; Tris k_des = 124 s⁻¹). The lower trace is the solution exchange measured after the experiment. (D) Bar plot indicating inhibition of the slope conductance at positive potentials for wild-type and M739D receptors. The limited inhibition of the M739D mutant by NMDG and Tris probably derives in part from channel block. Data are mean ± SEM for five patches.

Figure 2. Organic cations and monovalent and divalent metal cations used in this study

Alkali earth metal ions that bind in the kainate receptor cation binding site, Magnesium and Calcium are shown as spheres with radii according to Shannon (Li, 0.59Å; Na, 1.02 Å; Cs, 1.67 Å; Mg, 0.72 Å; Ca, 1.06 Å). Tris and N-methyl-D-glucamine are drawn in CPK representation (see Methods).

Figure 3. The GluR6 M739D mutant is modulated by divalent cations

(A) Glutamate activated outward currents recorded in the same patch, in control NaCl solution, and when Na⁺ was replaced by Tris. In contrast to the near complete loss of response for wild type GluR6, the rate of onset of desensitization increased only 1.6-fold, and the peak amplitude decreased by only 30 ± 5 %. Both solutions contained 100 μM Ca²⁺ and 100 μM Mg²⁺. (B) In the same patch as in (A), when Ca²⁺ and Mg²⁺ were omitted and trace divalent cations chelated with 100 μM EDTA, the removal of divalent ions had little effect on the outward current in 150 mM NaCl, but resulted in near-total abolition of responses recorded with Tris as the external cation (6 ± 3% of control, n = 5). (C) Bar plot showing the effect of 100 μM Ca²⁺ and 100 μM Mg²⁺ (+) and EDTA (-) on slope conductance and %desensitization for responses recorded with either Na or Tris. Steady-state desensitization in Tris in the absence of divalent cations was not determined (nd) because the residual current was too small to measure. Data shown from six patches; error bars indicate SEM. (D) Consistent with channel block by divalent cations, desensitization of the M739D mutant was slower at positive potentials for responses with 100 μM Ca²⁺ and 100 μM Mg²⁺ present (Na⁺). The rates of desensitization were the same at −40mV and at +60 mV when divalent cations were chelated with EDTA (Na−). We did not observe appreciable inward currents in external Tris, and the outward current in Tris without divalent cations was too small to reliably determine the desensitization rate. Bars represent the mean ± SEM for five to eleven patches.

Figure 4. The GluR6 M739D mutant is preferentially binds calcium

(A) Normalized outward currents for the M739D mutant recorded from the same patch with 100 μM each Ca²⁺ and Mg²⁺ (k_des = 154 s⁻¹); 100 μM Ca²⁺ (k_des = 212 s⁻¹); or 10 μM Ca²⁺ (k_des = 430 s⁻¹) with 150 mM Tris present in all solutions. EDTA was included to chelate other trace divalent ions, and buffer free calcium for the responses measured in 10 and 100 μM Ca²⁺. Dotted lines represent single exponential fits. (B) Bar plots summarizing changes in peak amplitude and desensitization rate for responses recorded with either Ca²⁺ or Mg²⁺ at concentrations of 10 and 100 μM indicate a weak preference for Ca²⁺. Bars show the means of 5 observations, error bars represent SEM.

Figure 5. Model divalent ion binding sites

(A) Structural model of the GluR6 wild-type cation binding site based on the Na-GluR5 complex (3c32), with sodium as a gold sphere. Transparent spheres represent the Shannon radius. The Methionine residue that caps the binding site is shown in CPK representation. (B) Structural model of the GluR6 M739D mutant with calcium bound (violet sphere). The model is derived from the Na-GluR5 complex (3c32) with the ‘capping’ isoleucine replaced by an aspartic acid. The bound sodium ion was replaced by a calcium ion (violet sphere) placed equidistant between the proximal carboxyl groups of E509 and M739D. (C) Ca (violet sphere) bound to copper amine oxidase (1OAC) with five protein ligands. The Ca ion is buried but the crown of three aspartic acid side chains that coordinate it is partly solvent exposed. Two backbone carbonyl oxygen atoms and a water molecule (W1) complete the six-fold coordination. (D) Mg (yellow sphere) bound in the active site of inositol monophosphatase in a substrate-free crystal form (2BJI). The coordination sphere has three carboxylates and two waters (red spheres, W1 and W2). A site is available for third water molecule to give octahedral coordination, although it was not modeled.