Structural rearrangements of NR1/NR2A NMDA receptors during allosteric inhibition

Disruption of the hydrophobic contacts at the ABD dimer interface increases zinc sensitivity

The crystal structure of the isolated NR1/NR2A ABD heterodimer shows that both subunits are arranged in a “back-to-back” fashion that is similar to the arrangement of subunits in the non-desensitized state of the AMPA GluR2 ABD homodimer (Furukawa et al., 2005; Fig. 1). In particular, the NR1/NR2A ABD interface comprises numerous contacts made by residues found at -or in the vicinity of- helices αJ and αD of both subunits, resulting in two subsites of dimerization: site A, formed by helix J of NR1 and helix D of NR2A; and site B, related to site A by a pseudo two-fold axis of symmetry, formed by the helix D of NR1 and helix J of NR2A. Both sites contain a cluster of highly conserved hydrophobic residues making numerous inter-subunit Van der Waals contacts that are expected to be important in dimer stabilization. We identified NR1-L777 and the homologous residue NR2A-L780 as residues likely to be essential for subunits dimerization since both leucines are buried at the core of the hydrophobic cluster with their side chain methyl groups involved in extensive hydrophobic contacts with the neighboring hydrophobic residues including NR2A-I514 and NR2A-S519 for NR1-L777, and NR1-A524 and NR1-I519 for NR2A-L780 (Fig. 2A). In addition, these two leucines are homologous to residue GluR2-L751 (Fig. 1B), mutation of which strongly promotes desensitization in GluR2 receptors (Horning and Mayer, 2004).

To explore the roles of NR1-L777 and NR2A-L780 in NMDAR zinc inhibition, we mutated each of these two leucine residues separately in alanine, whose shorter lateral chain should strongly disrupt the hydrophobic interface cluster. Compared to wild-type (wt) NR1/NR2A receptors, both NR1-L777A/NR2Awt and NR1wt/NR2A-L780A receptors display greatly enhanced zinc sensitivity (Fig. 2B–C). At 5 nM zinc, the lowest zinc concentration tested during initial mutation screening, currents were already strongly reduced (Fig. 2B). Whereas at wt NR1/NR2A receptors, zinc inhibits agonist-induced currents with an IC₅₀ around 10 nM, in the mutated receptors, zinc IC₅₀ was decreased by ~10-fold (1.1 and 1.2 nM for NR1-L777A/NR2Awt and NR1wt/NR2A-L780A receptors, respectively; Fig. 2C and Table 1). Another striking modification concerns the maximal level of zinc inhibition. Whereas, as previously reported, at wt NR1/NR2A receptors, saturating the high-affinity zinc binding site results in partial inhibition (maximal inhibition of 60–80%; Paoletti et al., 1997; Choi and Lipton, 1999; Low et al., 2000), at NR1-L777A/NR2Awt and NR1wt/NR2A-L780A receptors, residual currents in saturating zinc concentrations were dramatically diminished (maximal inhibition of 99% and 97%, respectively; Fig. 2C and Table 1). To make sure that the observed inhibition was driven by zinc binding to the NR2A NTD and not to other potential sites of lower affinity located outside the NTD (see Rachline et al., 2005), we checked that high-affinity zinc inhibition was absent in receptors combining the NR1-L777A mutation with the NR2A-H128S mutation, a mutation that selectively abolishes zinc binding to the NR2A NTD (Fayyazuddin et al., 2000; Low et al., 2000; Paoletti et al., 2000). Indeed, 100 nM zinc, a concentration that induces an almost complete inhibition of NR1-L777A/NR2Awt receptors barely affected the doubly mutated NR1-L777A/NR2A-H128S receptors (relative current 93 ± 3% [n=3]).

To gain further insight into the importance of leucines NR1-L777 and NR2A-L780 in zinc inhibition of NMDARs, we introduced additional mutations at these positions. As expected from the introduction of polar residues in an hydrophobic cluster, substitutions by a serine or a cysteine induced a leftward shift in zinc sensitivity, similar to that observed with the alanine substitutions, with both an increase in zinc apparent affinity (IC₅₀) and “efficacy” (maximal level of inhibition) (Table 1). Interestingly, introducing a phenylalanine, a bulky hydrophobic residue, also strongly increased zinc sensitivity (Table 1). Of note, currents carried by NR1wt/NR2A-L780F were so small (< 40 nA, due to strong proton inhibition at pH 7.3; see below) that full zinc concentration-response curves could not be reliably obtained although it was clear that at these receptors zinc was very potent (>50% inhibition at 5 nM). Thus, the series of NR1-L777 and NR2A-L780 mutations reveal that perturbing the hydrophobic packing between the two adjacent ABDs, either by side chain truncation or change in hydrophobicity (A, C and S mutations), or through a likely steric clash (F mutation), markedly increases zinc sensitivity of NR1/NR2A receptors. These results also indicate that the two pseudo-symmetrical dimerization sites A and B contribute to a comparable extent to dimer stability, an observation that is not surprising given the strong sequence homology between the two protomers in regions constituting the dimer interface.

We next mutated the residues that, in the crystal structure, partner with leucines NR1-L777 and NR2A-L780 through intersubunit contacts. For each leucine, three interacting residues were identified: NR2A-I514, NR2A-E516 and NR2A-S519 for NR1-L777 and the homologous residues NR1-I519, NR1-N521 and NR1-A524 for NR2A-L780 (Fig. 2A). Each of these residues participates to the dimer interface stability by making an interdomain Van der Waals contact with the distal methyl group of the adjacent leucine. The influence of the interactions involving NR2A-E516 and NR1-N521 is difficult to evaluate by mutagenesis because these residues contact the corresponding leucine through their backbone Cα. A similar situation applies to NR2A-S519 with the involvement of its Cβ methylene group that would only be eliminated with a substitution by glycine. Glycine substitutions, however, are known to disrupt secondary structures (O’Neil and DeGrado, 1990) and were therefore avoided. Substituting isoleucine R1-I519 either by shorter aliphatic residues (valine or alanine) or by a large aromatic residue (phenylalanine) produced very modest effects on zinc sensitivity (Table 1). Similarly, receptors containing an alanine mutation at the homologous position in NR2A (NR2A-I514A) displayed unchanged zinc sensitivity compared to wt receptors (despite strong effects on agonist sensitivity; Table 1 and see Experimental Procedures). That valine can substitute for the isoleucine is not unexpected given that NR1-I519 contacts NR2A-L780 through its Cβ branch methyl group, an interaction which could be preserved in the valine mutant. In contrast, the results obtained with the alanine mutants are surprising since the one methyl side chain of the alanine is expected to be too short to form favourable Van der Waals contacts with the key leucines. Moreover, the homologous mutation GluR2-I481A has a strong effect on desensitization of AMPA receptors indicative of a much weakened dimer interface (Horning and Mayer, 2004). However, because GluR2 ABDs form homodimers with a two-fold symmetric dimerization, each individual mutation is present twice at the dimer interface while at heteromeric NMDARs, there is only one copy of the NR1 or NR2A mutation per heterodimer. Thus, to match more closely the AMPA receptor situation, we co-injected NR1-I519A with NR2A-I514A and found that doubly-mutated receptors displayed significantly enhanced maximal level of zinc inhibition compared to singly-mutated receptors in agreement with a significant decrease in stability of dimer contacts (Table 1). Introducing a charged aspartate at these positions also resulted in markedly enhanced zinc sensitivity, most particularly in the case of the NR1-I519D mutation, as expected for mutations that strongly disrupt the hydrophobic cluster (Table 1). Similarly, introducing charged residues at position NR1-A524 (D, E or K mutation) induced marked increase in zinc sensitivity (Table 1). Taken together, these results indicate that isoleucines NR2A-I514 and NR1-I519 as well as alanine NR1-A524 (and potentially NR2A-S519) are likely to contribute to dimer stability although some structural flexibility at the isoleucine positions can be supported without apparent destabilization of the interface.

The role of the hydrophobic interactions at the dimer interface in setting zinc sensitivity of NR1/NR2A receptors was further investigated by mutating additional residues of sites A and B in the vicinity of NR1-L777 and NR2A-L780. Residue NR1-L774 (site A) and the homologous residue NR2A-L777 (site B) on helices αJ form the exterior side of the hydrophobic cluster and contact the methylene groups of residues from the adjacent subunit NR2A-E516 and NR2A-S519 (site A), and NR1-N521 and NR1-A524 (site B) (Fig. 2A). When NR1-L774 and NR2A-L777 were individually mutated to alanine, zinc sensitivity increased. In addition, when NR1-L774A was co-injected with NR2A-L777A, the corresponding doubly-mutated receptors displayed enhanced zinc sensitivity compared to the singly-mutated receptors (Table 1), reinforcing the importance of the ABD intermolecular Van der Waals contacts in the control of zinc sensitivity. The shifts in sensitivity, however, were clearly not as marked as with NR1-L777 or NR2A-L780 mutated receptors. This is consistent with the fact that NR1-L774 and NR2A-L777 have their side chain partly solvent accessible and thus are expected to contribute significantly less than the other two leucines to dimer stability. Finally, we tested mutations at NR2A-V526. This residue, part of site B hydrophobic cluster, does not make direct contacts with the adjacent NR1 subunit but instead forms an intrasubunit contact with the key leucine 780 of NR2A (Fig. 2A). This latter interaction, however, is expected to be important for the ABD dimer assembly because of its likely involvement in the positioning of NR2A-L780. Consistent with this hypothesis, the alanine truncation mutant NR2A-V526A led to a small, but significant, increase in zinc sensitivity, while introduction of a less conservative mutation, NR2A-V526K strongly increased zinc sensitivity (Table 1).

Disruption of hydrogen bonds at the ABD dimer interface increases zinc sensitivity

Inspection of the NR1/NR2A ABD crystal structure indicates that, beside hydrophobic contacts, hydrogen bonds also participate to the stabilization of the ABD dimer interface. This is the case of residue NR1-E781 at the apex of helix J in site A. NR1-E781 is in a fully extended conformation and makes a bidentate contact with the base of helix D of the adjacent NR2A subunit very much like GluR2-E755 in AMPA receptors (Horning and Mayer, 2004). Residue NR1-E781 makes a direct hydrogen bond with the backbone amide nitrogen of NR2A-E516 and a hydrogen bond with the backbone amide nitrogen of NR2A-E517 via a water molecule (Fig. 3A). The favourable electrostatic interaction between the negatively charged glutamate and the dipole of the αD helix of NR2A, may also contribute to the dimer interface cohesion. To remove these interactions, NR1-E781 was first substituted into alanine. This resulted in a marked enhancement of zinc sensitivity with an effect on both zinc IC₅₀ and the maximal level of zinc inhibition (Zn IC₅₀ of 2.4 nM and maximal inhibition of 92% [n=9], Table 1). When the glutamate side chain was replaced by a more conservative aspartate, a shorter negatively charged residue that should weaken both hydrogen bond network and charge-dipole interaction, zinc sensitivity was increased but to a lesser extent (Table 1). Surprisingly, replacing NR1-E781 by a glutamine, a neutral residue of similar length which should permit formation of one hydrogen bond with the neighbouring D helix but not the charge-dipole interaction, strongly increased zinc sensitivity (Zn IC₅₀ of 1.3 nM and maximal inhibition of 97% [n=8], Fig. 3B). This result suggests that the charge-dipole interaction between NR1-E781 and the NR2A αD helix may contribute to the overall dimer stability. The fact, however, that the Q mutation has a “stronger” phenotype than the A mutation is difficult to explain assuming that the sole effect of the mutation is to weaken dimer stability and likely reflects additional effects of the Q mutation on other conformational state(s) of the dimer (possibly through stabilization of a “desensitized” state; see Armstrong et al., 2006). In conclusion, the hydrogen bond network at the ABD dimer interface appears to be an important structural determinant involved in the control of the zinc sensitivity of NR1/NR2A receptors most likely by contributing to the stability of the ABD dimer interface.

Receptors with cysteine cross-linked ABDs display redox-state dependent zinc sensitivity

A prediction based on the preceding results showing that disrupting the intersubunit ABD dimer contacts enhances zinc sensitivity is that strengthening dimer stability should also affect zinc sensitivity. We first tried to increase dimer stability by introducing either a) polar residues to mimic the intersubunit salt links present in AMPA and kainate ABDs (Horning and Mayer, 2004; Priel et al., 2006; Plested and Mayer, 2007), b) tyrosines to mimic the non-desensitizing GluR2-L483Y receptor situation (Sun et al., 2002; Furukawa et al., 2005), or c) mutations to consolidate the hydrogen bond network between the two partner subunits. Disappointingly, introducing polar residues or tyrosines produced either no significant change in zinc sensitivity (NR1-S773N/NR2A-S519E, NR1-N521Y/NR2A-E516Y) or increased zinc sensitivity (NR1-A524E/NR2A-V526K-D776N) (Table 1). In contrast, introducing H-bond acceptor groups at position NR2A-G784 (homologous to NR1-E781) did result in a small, but significant, decrease in zinc sensitivity (Table 1 and Fig. S1) suggesting that enhanced dimer stability may render zinc inhibition less potent.

To test the influence of dimer stabilization on zinc sensitivity more directly, we used an approach based on engineered disulfide bridges. In their study of the NR1/NR2A ABD structure, Furukawa et al. (2005) showed that it is possible, in full-length receptors, to cross-link adjacent NR1 and N2A subunits by introducing two pairs of cysteines at appropriate positions in the ABD NR1/NR2A heterodimer interface (NR1-N521C-L777C and NR2A-E516C-L780C, hereafter named NR1-CC and NR2A-CC, respectively; see Fig. 4A). By performing Western blot experiments on whole oocyte extracts, we first verified that in our hands and only under non-reducing conditions, such doubly-cysteine mutated NR1 and NR2A subunits do form heterodimers in an intact receptor (Fig. S2). We then assessed zinc sensitivity of NR1CC/NR2A-CC receptors under different redox conditions. In control conditions (i.e. before any redox treatments), NR1-CC/NR2A-CC receptors exhibited a maximal zinc inhibition significantly lower than that observed on NR1wt/NR2Awt receptors (64% vs 81%; Fig. 4B–C) as would be expected if a significant proportion of surface NMDA receptors has a reduced zinc sensitivity because of restrained conformational mobility of covalently-attached ABDs. We obtained further evidence that this is likely to be the case by treating the oocytes with DTNB, an oxidizing agent that promotes disulfide bond formation. As shown in Figure 4B–C, DTNB treatment strongly reduced the sensitivity to zinc, as judged by the much lowered maximal level of inhibition (37%). The fact that zinc sensitivity of DTNB-treated receptors was not completely abolished may be due either to a limited access of DTNB to the dimer interface resulting in a mixed population of receptors with reduced and intact cross-links or to insufficient time of treatment. Another possibility could be that even in the cross-linked receptors, some movements at the ABD interface are still possible following zinc binding to the NTD. Whatever the explanation, we predicted that breaking the engineered disulfide bonds should restore zinc sensitivity. Consistent with this idea, DTE treatment (5 mM, 15 min) dramatically increased zinc sensitivity (100% maximal inhibition; Fig. 4B–C). The sensitivity of DTE-treated R1-CC/2A-CC receptors to zinc was in fact similar to that of receptors harbouring a single cysteine mutation at the corresponding αJ helix leucine positions (R1-L777C/2Awt or R1wt/2A-L780C receptors; see Table 1) as expected if most of the disulfide bonds have been cleaved by the DTE treatment. To verify that the observed effects resulted from a selective modification of the two disulfide bonds introduced at the subunit interface, we performed similar redox treatments on oocytes expressing either wt NR1/NR2A receptors or receptors containing only one set of double cysteine mutations, which do not give rise to disulphide-linked NR1/NR2A heterodimers (R1-CC/2Awt and R1wt/2A-CC receptors; Fig. S2). As shown in Fig. 4D, none of the redox treatments affected the maximal level of zinc inhibition of these receptors. Finally, we also found that it was possible to reversibly switch NR1-CC/NR2A-CC receptors from a low- to a high-zinc sensitivity state by sequential applications of DTNB and DTE (Figs 4B and D). These results reveal that cross-linking ABD protomers leads to receptors with greatly diminished zinc sensitivity and support a model in which conformational rearrangements of the agonist binding cores are required to couple zinc binding to the NTD to subsequent channel closure.

Disruption of the heterodimer interface between ABDs increases proton sensitivity

A distinctive property of the high-affinity zinc inhibition of NR1/NR2A receptors is that it saturates at a maximal level of 60–80% (Paoletti et al., 1997; Choi and Lipton, 1999; Low et al., 2000). Based on the observation that the extent of maximal zinc inhibition depends on extracellular pH (inhibition by a saturating zinc concentration increases as proton concentration increases), it has been proposed that zinc exerts its inhibitory action on NR1/NR2A receptors through an enhancement of tonic proton inhibition (Choi and Lipton, 1999; Low et al., 2000). According to this model, the presence of protons is required for zinc inhibition to occur. The residual current observed at saturating zinc concentrations and at physiological pH reflects the fact that saturating the high-affinity zinc binding site does not lead to saturation of the inhibitory proton binding site at pH 7.3. Given the strong effects of the ABD dimer interface mutations on zinc sensitivity, and in particular on the maximal level of zinc inhibition, we decided to examine the pH sensitivity of our mutant receptors. We found that mutation of the central leucines of the ABD interface hydrophobic cores, which greatly enhanced maximal zinc inhibition (see Fig. 2), dramatically increased proton sensitivity. Thus, as illustrated in Figure 5A, switching the agonist solution from pH 7.3 to 8.3, which results in a modest 1.15-fold increase of wt currents, induced a 4- and 2.6-fold increase of currents carried by NR1-L777A/NR2Awt and NR1wt/NR2A-L780A receptors, respectively. Currents from mutant receptors were still profoundly inhibited by H⁺ at pH 8.3 since further potentiation could be induced by increasing pH to 9.3 (Fig. 5A). Using the series of mutations that we had produced at these positions, we found that the potentiation was largest for NR1/NR2A-L780F receptors with a remarkable 30-fold increase in current amplitude when jumping from pH 7.3 to 9.3 (Fig. 5A). Performing full concentration-response curves revealed that pH sensitivity of these mutant receptors was shifted by more than two orders of magnitude in comparison to wt receptors (pH of IC₅₀ of 9.13 [n=4–7] for NR1-L777A/NR2Awt receptors, 9.16 [n=4–8] for NR1wt/NR2A-L780A receptors and 9.08 [n=4] for NR1/NR2A-L780F receptors, vs 6.92 [n=14] for wt receptor; Fig. 5B).

To obtain a more complete view of the relationship between zinc and proton sensitivities at NR1/NR2A receptors, we measured pH sensitivity of many dimer interface mutants covering a large spectrum of zinc sensitivity (Table 1). Mutant receptors that displayed enhanced zinc sensitivity also displayed enhanced pH sensitivity. Two mutant receptors deviated from this general trend: NR1-I519D/NR2Awt which showed unusual biphasic pH sensitivity (with a potentiating and an inhibitory component; data not shown) and NR1-A524E/NR2Awt (and to a lesser extent NR1-A524D/NR2Awt) which showed very little shift in pH sensitivity despite a marked increase in zinc sensitivity. We believe that for these mutants, the “abnormal” pH sensitivity reflects the presence of an additional pH effect involving highly perturbed pKa’s of the introduced carboxylic groups (probably because of the clustering of multiple charged residues in a relatively low polar environment; see Fersht, 1999). Notwithstanding these latter mutants, we found that the maximal level of zinc inhibition and proton sensitivity were strongly correlated (Fig. 6). This result agrees well with a model in which NTD-controlled zinc inhibition results from an enhancement of proton inhibition and indicates that common mechanisms are likely to operate during zinc and proton inhibition.

We gained additional evidence for a functional interaction between the ABD dimer interface and the NTDs by measuring pH sensitivity of receptors combining a NR1 ABD interface mutation (NR1-E781A) with a NTD-deleted NR2A subunit (NR2A-delNTD; see Rachline et al., 2005). Although the interface mutation or the NTD deletion separately caused a modest enhancement of proton sensitivity (pH of IC₅₀ of 7.3 and 7.2, respectively; Table 1 and Fig. S3), the combination of both modifications resulted in a strong synergistic increase in proton sensitivity (pH of IC₅₀ of 9.1; Fig. S3).

Increased accessibility of the ABD dimer interface upon zinc inhibition

Mutations expected to weaken (or strengthen) ABD dimer stability selectively enhance (or decrease) zinc and proton sensitivities of NR1/NR2A receptors. This observation strongly suggests that disruption of the ABD dimer interface, with one protomer moving away from the other, constitutes a central conformational rearrangement occurring during zinc and proton inhibitions, very much like the movement occurring during AMPA and kainate receptor desensitization (Armstrong et al., 2006; Weston et al., 2006; Priel et al., 2006). To probe for this putative conformational change, we introduced a cysteine residue at the ABD dimer interface and compared the rate of modification of the mutant receptors by the thiol modifying agent sodium (2-sulfonatoethyl)methane thiosulfonate (MTSES) with and without zinc. We hypothesized that attachment of the bulky (2-sulfonatoethyl)methane moiety at the ABD interface of the mutated receptors would increase their sensitivity to zinc and pH, thereby decreasing current sizes. According to a model in which the ABD interface becomes more accessible upon zinc inhibition, the rate of MTSES modification should be faster in zinc than without. We chose to introduce the cysteine substitution at position 2A-V526 because of the following criteria: first, this residue, as judged by the NR1/NR2A ABD dimer crystal structure (Furukawa et al., 2005), is poorly accessible to the bulk solvent in the “resting” state of the ABD dimer; second, its mutation to cysteine has only a very minor effect on zinc sensitivity (Zn IC₅₀ = 10.9 nM and maximal inhibition of 81%, [n = 3]) such that residual currents in saturating zinc concentrations are of large enough amplitude to be accurately measured; third, its replacement by a large charged residue such as a lysine induced a marked increase in zinc sensitivity (see Table 1) presumably through disruption of the dimer interface; fourth, in GluR2 AMPA receptors, the homologous residue (K493) becomes more exposed to the solvent during receptor desensitization (Armstrong et al., 2006).

We observed that the reduction of currents carried by NR1wt/NR2A-V526C receptors induced by MTSES application was both faster and of greater amplitude in the presence of zinc (applied at 500 nM, a saturating concentration) than in its absence (Fig. 7A–B). The enhanced magnitude of inhibition (61 ± 4% [n = 4] inhibition in zinc vs 39 ± 4% [n = 5] in DTPA) is expected if in both situations, the decrease in currents induced by MTSES reflects a similar enhancement of proton inhibition but starting from receptors under higher tonic proton inhibition in the zinc conditions (see Low et al., 2000). From a kinetic point of view, in both cases, application of MTSES was best described by a two-component time course. Since the fast component, but not the slow one, was also observed during treatment of NR1wt/NR2Awt receptors with MTSES (most likely due to modification of an endogenous cysteine; data not shown), we concentrated our analysis on the slow component. On average, this latter component was 1.9-fold faster in the presence of zinc than in its absence (Fig. 7C–D). Thus, the thiol moiety of the interface residue 2A-V526C becomes more accessible upon NTD-driven zinc inhibition.

Molecular Biology

The pcDNA3-based expression plasmids for NR1-1a (named NR1 herein) and NR2A (all rat clones), the mutagenesis strategy and sequencing have been described previously (Rachline et al., 2005).

Electrophysiology

Recombinant NMDA receptors were expressed in Xenopus laevi oocytes after co-injection of cDNAs (at 10 ng/μl; nuclear injection) coding for the various rat NR1 and NR2A subunits (ratio 1:1). Oocytes were prepared, injected, voltage-clamped and superfused as described previously (Paoletti et al., 1995, 1997). For all experiments, except for those aimed at measuring the pH sensitivity (see below), the standard external solution contained (in mM): 100 NaCl, 0.3 BaCl₂, 5 HEPES. The pH was adjusted to 7.3 with KOH.

For free zinc concentrations in the 1 nM-1 μM range, tricine (10 mM) was used to buffer zinc (Paoletti et al., 1997) and the following relationship was used to calculate the free zinc concentrations: [Zn]_free=[Zn]_added/200 (Fayyazuddin et al., 2000). In the control “0” added zinc solution, the heavy metal chelator DTPA (10 μM) was present in addition to tricine. For the zinc concentration-response curves of the mutant receptors displaying a very high zinc sensitivity (Figs 2 and 3), zinc was not buffered with tricine but with ADA (1mM) and the following equation was used to calculate free zinc concentrations ([Zn]_free=[Zn]_added/17000; Fayyazuddin et al., 2000). For the pH sensitivity experiments, an enriched HEPES external solution was used to insure proper pH buffering. This solution contained (in mM): 60 NaCl, 2.5 KCl, 40 HEPES. pH was first adjusted to 10.3 with NaOH. More acidic solutions were then obtained by decreasing pH with concentrated HCl. This procedure was chosen because it maintains Na⁺ concentration at a constant value irrespective of the pH. To chelate trace amounts of contaminant zinc, DTPA (10 μM) was added in all the solutions used for pH experiments.

NMDA currents were induced by simultaneous application of saturating concentrations of L-glutamate and glycine (agonists). These concentrations were set at 100 μM each for wt receptors. However, since, at NR1/NR2A receptors, both Zn IC₅₀ and maximal inhibition depend on the concentration of glutamate used to activate the receptors (with stronger inhibition with increased glutamate concentrations; Zheng et al., 2001; Erreger et al., 2005), special care was taken to estimate, for each mutant receptor, the appropriate glutamate concentration required to maximally activate the receptors. For NR1-L777A/NR2Awt and NR1wt/NR2A-L780A receptors, which display the strongest shifts in zinc sensitivity, full agonist dose-response curves were performed, revealing agonist sensitivity close to that of wt receptors (Fig. S4). Accordingly, 100 μM glutamate (together with 100 μM glycine) was used with these receptors. For all other mutants, currents elicited by 100 μM and 1 mM agonists were initially compared. If these currents differed by less than 15% (as observed with wt receptors), glutamate and glycine were used at 100 μM each. This was the case for most of the mutants described in the present study. For mutant receptors displaying a larger difference (NR1wt/NR2A-I514A, NR1wt/NR2A-I514D, NR1-CC/NR2A-CC), full agonist concentration-response curves were performed (Figs S5 and S6) and, in the following zinc/pH experiments, glutamate and glycine concentrations were adjusted to provide maximal or nearly maximal activation of the receptors (1 mM glutamate and glycine for NR1-CC/NR2A-CC, 100 μM glycine + 1 mM glutamate for NR1wt/NR2A-I514A, 100 μM glycine + 10 mM glutamate for NR1wt/NR2A-I514D).

When performing full pH concentration-response curves, at very alkaline pH values, agonist concentrations were adjusted to compensate for the loss of protonation of the α-amine moiety (pKa ~9.7 for both L-glutamate and glycine). Thus, at pH 9.3, 9.8 and 10.3, glutamate and glycine concentrations were increased by 1.4 fold, 2.1 fold and 5 fold, respectively. At all other pH values, agonist concentrations were kept identical to those used at pH 7.3.

Currents were measured at a holding potential of −60 mV and experiments were done at room temperature. Error bars represent the standard deviations of the mean relative currents.

Redox treatments of receptors with engineered disulphide bonds

Since NR1-CC/NR2A-CC receptors exhibit lower agonist affinities compared to wt receptors (Furukawa et al., 2005 and Fig. S6), agonist concentrations had to be increased from 100 μM (wt receptors) to 1 mM when studying these receptors in order to be at saturating concentrations. This was also the case for NR1wt/NR2A-CC receptors. To promote disulfide bridge cleavage, oocytes were incubated with the reducing agent DTE (5 mM) in a Barth solution containing (in mM): 88 NaCl, 1 KCl, 0.33 Ca(NO₃)₂, 0.41 CaCl₂, 0.82 MgSO₄, 2.4 NaHCO₃, 10 HEPES supplemented with gentamycin (50 μg/mL) and the NMDAR antagonist D-APV (50 μM) to prevent NMDAR-induced cell toxicity. pH was adjusted to 8.0. To induce the formation of disulfide bonds, oocytes were incubated with the oxidizing agent DTNB (0.5 mM) in the same Barth solution supplemented with gentamycin, D-APV and the heavy-metal chelator DTPA (10 μM).

Immunoblotting

Xenopus oocytes were prepared and injected as for the electrophysiological experiments. After 2–5 days, expression level for NMDA receptors was checked using two-electrode voltage clamp and only oocytes that showed NMDA currents >300 nA were subsequently used. Oocytes were first treated with 0.5 mM DTNB for 5 min. They were then homogenized in a buffer (10 μl/cell) containing 20 mM Tris-HCl (pH 8.0), 50 mM NaCl, 1/100 volume of mammalian protease inhibitor cocktail (Sigma) and 1 mM N-ethylmaleimide. After centrifugation (12,200 g for 10 min at 4°C), the pellet was then solubilized in the same buffer complemented with 1% β-D-dodecyl-maltoside (Anatrace). Sample amounts equivalent to two oocytes were loaded on 8% gels (SDS-PAGE) either in presence or absence of 100 mM DTE. Gels were incubated 10 min in 1 mM DTE before transfer on PVDF membranes. PVDF membranes were blocked with fat-free milk (5% in TBS-Tween buffer) and incubated overnight (4°C) with 1/2,000 dilution of mouse anti-NR1 antibody (amino acids 660–811, Chemicon) followed by 1/5,000 HRP-conjugated goat anti-mouse secondary antibody and revealed with SuperSignal West Pico Kit (Pierce).

Data Analysis

Data were collected and analyzed using Clampex 7.0 and Clampfit 9.2 (Axon Instruments, Foster City, CA). Data were fitted using KaleidaGraph 4.0 (Synergy Software, Reading, PA). For zinc concentration-response curves, the experimental data points were fitted with the following Hill equation: I_Zn/I_Control = 1 − Inhib_max/(1+(IC₅₀/[Zn])^nH), where I_Zn/I_Control is the mean relative current, [Zn] is the concentration of free zinc, nH is the Hill coefficient, Inhib_max is the maximal zinc inhibition, and IC₅₀ is the concentration of zinc producing 50% of the maximal inhibition. IC₅₀, Inhib_max, and nH were fitted as free parameters.

When performing pH experiments, we observed that changes in pH of the extracellular solution leads to small (a few mV) but significant changes in the potential of the bath reference electrode, an effect which results in small shifts of the reversal potential of NMDA responses. To correct for these effects, the measured NMDA current values were multiplied by the following factor: 0.97 at pH 6.3, 0.98 at pH 6.8, 1.03 at pH 7.8, 1.05 at pH 8.3, 1.07 at pH 8.8, 1.08 at pH 9.3, 1.09 at pH 9.8 and 10.3. To generate pH concentration-response curves, a two step procedure was used. For each cell, the experimental data points were first fitted using the following Hill equation: I_pH/I_Control = b/(1+10^(n_H (pH_IC50−pH))), where I_pH/I_Control is the relative current (control current is obtained at pH 7.3), b is the maximum potentiation obtained at alkaline pH, pH_IC50 is the pH producing 50% of inhibition, and n_H is the Hill coefficient. pH_IC50, n_H, and b were fitted as free parameters. For each cell, all data points were then divided by b to have a maximum relative current normalized to 1. Resulting values obtained for each individual cell and for each pH were then averaged and the mean data points were then fitted with the same Hill equation as above but with b fixed to 1.0.

To calculate the ΔG of maximal zinc inhibition (ΔG°_Zn), we assumed that, at the plateau of the zinc-concentration response curve (i.e. at saturating zinc concentrations), the agonist-bound receptors are in equilibrium between two states, a zinc-bound inactive state (RZn) and a zinc-bound active state (RZn*). Thus, the equilibrium constant K_Zn between these two states is given by K_Zn = [RZn*]/[RZn] = (1 − Inhib_max)/Inhib_max (with Inhib_max being the maximal inhibition produced by zinc; see above) and the corresponding Gibbs free energy (ΔG°_Zn) is defined as: ΔG°_Zn = − RT Ln(K_Zn) = −RT Ln(10) log((1 − Inhib_max)/Inhib_max), with R being the gas constant and T the absolute temperature. As an approximation, we used ΔG°_Zn (kcal.mol⁻¹) = − 1.4 log((1 − Inhib_max)/Inhib_max) at room temperature. For each mutant receptor (mut), we then calculated the difference ΔΔG°_Zn(mut) = ΔG°_Zn(mut) − Delta;G°_Zn(wt), where ΔG°_Zn(wt) is the ΔG of maximal zinc inhibition for wt receptors expressed in the same batch of oocytes as the studied mutant. We used this procedure to circumvent some variability seen in the maximal level of zinc inhibition at wt NR1/NR2A receptors (max. inhibition 0.77–0.83 range). To quantify proton inhibition, we used the equilibrium constant K_H between the protonated (inactive) and deprotonated (active) state of the agonist-bound receptor to calculate the corresponding free energy: ΔG°_pH = − RT Ln(K_H) = RT Ln(10) pH_IC50. At room temperature and at pH 7.3, we used ΔG°_pH (kcal.mol⁻¹) = 1.4 pH_IC50. For consistency with the analysis of the maximal zinc inhibition, the difference ΔΔG°_pH(mut) = ΔG°_pH(mut) − ΔG°_pH(wt) was then calculated for each mutant receptor. Of note, K_H accounts for a binding equilibrium, whereas K_Zn accounts for an equilibrium between two zinc-bound states and thus does not involve any binding step.

Chemicals

HEPES, L-glutamate, glycine, DTPA (diethylenetriamine-pentaacetic acid), Tricine (N-tris(hydroxymethyl)methylglycine), ADA (N-[2-Acetamido]-2-iminodiacetic acid), DTE (dithioerythritol) and DTNB (5,5′-dithio-bis(2-nitrobenzoïc acid)) and NEM (N-ethylmaleimide) were obtained from Sigma (Saint-Louis, MO, USA), D-APV (D-(−)-2-Amino-5-phosphonopentanoic acid) from Ascent Scientific (Weston-Super-Mare, UK), gentamycin from Gibco (Invitrogen, Rockville, MD, USA) and sodium (2-sulfonatoethyl)methane thiosulfonate (MTSES) from Toronto Research Chemicals (North York, On., Canada). L-glutamate and glycine stock solutions (100 μM – 1 M) were prepared in bi-distilled water. DTNB was prepared as a stock solution (50 mM) in an HEPES solution (500 mM, pH 8.0) whereas DTE was freshly prepared directly from powder. MTSES was prepared as 200 mM stock solutions in bi-distilled water, aliquoted in small volumes (200 μL) and immediately frozen at −20°C; aliquots were thawed just before use and replaced every 30 min. Zinc was added as chloride salts (ZnCl₂, ACS reagent quality, Sigma) by dilution from 100 mM stock solutions prepared in 10⁻² N HCl.

3D structure illustrations

Figures from the crystal structure of the NR1/NR2A ABD heterodimer (pdb coordinates 2A5T; Furukawa et al., 2005) were prepared with PyMOL (DeLano, 2002).