Modulation of triheteromeric NMDA receptors by N-terminal domain ligands

Read Also

By using a novel approach based on a combination of point mutations, we have been able to functionally isolate recombinant NMDA receptors containing NR1 and two different NR2 subunits (triheteromeric receptors). This allowed us to establish the functional stoichiometry of high-affinity Zn inhibition at NR1/2A receptors and to define the sensitivity of triheteromeric NMDA receptors to allosteric modulators binding to the NR2-NTDs. We have unmasked a general rule for NTD function, which is that triheteromeric assemblies including only one ligand-binding NR2-NTD suffices to confer high-sensitivity to its ligand, be it Zn or ifenprodil. Such assemblies, however, show a greatly reduced degree of inhibition compared to their diheteromeric counterparts containing two copies of the same NR2-NTD. High-efficacy inhibition, as seen on NR1/2A and NR1/2B receptors, requires occupancy of both available NR2-NTDs by their respective ligands, a principle applicable even to receptors containing two different NR2 subunits.

One potential problem of our approach is that it was necessary to introduce a strong glutamate binding mutation into the NR2A subunit to isolate the triheteromeric receptor assemblies. Because little is known about the nature of interactions between the NTDs and the agonist binding domains, the possibility exists that NTD ligands could act differently on our mutated triheteromeric receptors than on native triheteromeric NMDARs. We believe that this is not the case for the following reasons. Firstly, on NR1/2A_N164KT690I/2B receptors, the super-additivity experiments demonstrate that both NR2-NTDs are functional despite the presence of the glutamate mutation. Secondly, either Zn or ifenprodil can inhibit these receptors with similar plateaus although the Zn binds to the glutamate mutated subunit, whereas ifenprodil binds to a wt subunit. Thirdly, and most importantly, we have obtained direct experimental evidence that NR2-NTD function is not affected by the presence of the glutamate binding mutation by recording from cells expressing NR1, NR2A and NR2A_H128SN614K subunits under conditions of minimal contamination by all-mutant receptors (see supplementary Fig. 2). Zn concentration-inhibition curves in these cells are almost identical to those recorded from NR1/2A/2A_{H128SN614KT690I} receptors. Together, these findings indicate that the T690I-mutation does not affect the ability of NTD-ligands to inhibit NMDARs, suggesting that our current findings are directly applicable to native triheteromeric assemblies.

The discovery that triheteromeric NR1/2A/2B receptors are inhibited by ifenprodil with high-affinity but only to a limited degree, might explain the apparent controversy surrounding the ifenprodil sensitivity of NR1/2A/2B receptors. Kew et al. (1998) suggested that ifenprodil sensitive NR1/2A/2B receptors were present in cultured cortical neurons, but Tovar and Westbrook (1999) argued against this because they saw only limited ifenprodil inhibition in HEK-293 cells co-expressing NR1, NR2A and NR2B subunits. Interpretation of the findings in either case is complicated by the fact that they are based on results from cells expressing mixtures of di- and tri-heteromeric receptors. Using a direct biochemical approach, Hawkins et al. (1999) found that in HEK-293 cells transfected with the NR1, NR2A and NR2B subunits, receptors purified with an anti-NR2A antibody showed significant high-affinity binding to the ifenprodil derivative [³H]Ro 25-6981, implying that NR1/2A/2B receptors bind ifenprodil and its derivatives with high-affinity, as suggested by Kew et al. (1998), but apparently at odds with the observation of Tovar and Westbrook (1999). Our present observation that NR1/2A_N614KT690I/2B receptors have a high-affinity for ifenprodil but are inhibited to only a limited extent ties together these apparently divergent observations. Further, our finding that macroscopic on- and off-rates of ifenprodil inhibition on these receptors differ from those on NR1/2B wt receptors, may now serve as a useful tool for detecting the presence of triheteromeric NR1/2A/2B receptors in vivo.

The NR2-NTDs are structurally related to the bacterial protein LIVBP (leucine/isoleucine/valine binding protein) and to other LIVBP-like domains found in other membrane receptors, in particular the agonist-binding domains (ABDs) of metabotropic glutamate and GABA receptors (Paoletti et al., 2000; Pin et al., 2003). Functional and crystallographic studies of a number of LIVBP-like domains have revealed that these domains fold as two lobes separated by a central cleft and that they oscillate between an “open” and “closed” cleft conformations, the later being stabilized upon ligand binding within the cleft (“Venus-flytrap” mechanism; see Pin et al., 2003). Both Zn and ifenprodil act via binding sites located in the cleft of their respective NR2-NTDs and have been proposed to promote cleft closure (Paoletti et al., 2000; Perin-Dureau et al., 2002). In other words, Zn and ifenprodil can be considered as agonists of their respective NTD-binding sites, the activation of which leads to inhibition of the receptor. Exactly how activation (closure) of the NTD-domains leads to inhibition of the receptor is still poorly understood. One suggestion is that binding of Zn or ifenprodil to the NR2-NTDs induces a shift of the proton sensitivity rendering the receptors more sensitive to inhibition by ambient protons (Mott et al., 1998; Low et al., 2000). In Fig. 8, we show two possible NTD activation mechanisms which could account for NR2-NTD inhibition of NMDARs. They are based on our knowledge of the stoichiometry of NMDARs (two NR1 and two NR2 subunits) and on our present results. We know that there are two potential NR2-NTD binding sites per receptor, and that both singly- and doubly-NTD-liganded states can lead to inhibition, but that the degree of inhibition of the doubly-NTD-liganded receptors is greater (higher efficacy) than that of singly-NTD-liganded receptors. What we do not know is whether, the NR2-NTDs move in a concerted fashion (scheme 1) or independently (scheme 2).

            In the first case, we assume that the NR2-NTDs can only change conformation in a concerted manner, being either both open or both closed. The difference in efficacy of inhibition between singly- and doubly-NTD-liganded states results from the fact that the stability of the doubly-NTD-liganded closed state is greater than the singly-NTD-liganded closed state (efficacy E2 = b2/a2 > E1 = b1/a1). In the second case, where the NR2-NTDs can move independently from one another, inhibition can arise from receptors with either one or both NR2-NTDs closed. In this scenario, receptors with one open NR2-NTD and one closed NR2-NTD have only partial access to the downstream inhibitory machinery, and only receptors with both NR2-NTDs closed have full access (potentially by partial or full unmasking of a proton sensor). Importantly, in this scheme, singly-NTD-liganded receptors give access only to the state with partial access to the downstream inhibitory machinery whereas doubly-NTD-liganded receptors give access to both singly- and doubly-NTD-shut states, with partial or full access to the inhibitory machinery.  

Using parameters chosen to give the correct IC₅₀ for the wt receptors (18 nM), and correct relative inhibitions for wt and triheteromeric receptors, the schemes shown in Fig. 8A result in the predicted concentration-inhibition curves and Hill slope-concentration curves shown in Fig. 8B/C. Both models predict Zn IC₅₀s for the triheteromeric receptors in the low nanomolar range as experimentally observed. Scheme 1 predicts the correct tendency with regards to the relative IC₅₀s of wt and triheteromeric receptors, with an IC₅₀ of 25 nM for triheteromeric receptors with only one active NR2-NTD (red curves), very close to the experimentally determined value for NR1/2A/2A_{H128SN614KT690I} receptors (29 nM). Scheme 2 in contrast, predicts an IC₅₀ of 10 nM (red curves), a shift in the wrong direction. However, scheme 2 gives a much better prediction than scheme 1 of the shallow Hill-slopes, characteristic of high-affinity Zn and ifenprodil inhibitions of wt NMDARs (Paoletti et al., 1997; Chen et al., 1997; Kew et al., 1998; Masuko et al., 1999; Perin-Dureau et al., 2002). Scheme 2 predicts a Hill slope of 1.07, very close to the experimentally determined value (1.06), whilst scheme 1 predicts a steeper slope of 1.33 (see Fig. 4 and green curves, Fig. 8). The shallow Hill-slopes can at first seems at odds with the presence of two potential binding sites per receptor (two NR2 subunits). Our findings now provide an explanation for this, as both our models predict a shallowing of the concentration-inhibition curve when singly-liganded inhibition is taken into account (green vs blue lines). Interestingly, a precedent in favour of scheme 2 comes from dimeric metabotropic glutamate receptors (mGluRs) which possess ABDs structurally related to the NTDs of NMDARs. Indeed, Kniazeff et al. (2004) found that in mGluRs, binding of agonist to both ABDs is required for full activity, but that partial activity can be observed when one ABD is occupied by glutamate while the other is blocked open by an antagonist. For NMDARs, further investigation will be needed to differentiate between the two models.

One interesting observation made during the course of this study was that, whilst the IC₅₀ for Zn inhibition of the NR1/2A/2A_{H128SN614KT690I} receptors was very similar to that of NR1/2A wt receptors, the glutamate EC₅₀ of these triheteromeric receptors was, by contrast, very different from that of either NR1/2A wt or all-mutant NR1/2A_{H128SN614KT690I} receptors. This suggests that the nature of the interactions between the glutamate binding sites of the NR2A subunits is very different from the interactions of their NTDs, despite the fact that the two domains share similar bi-lobate folds and Venus-flytrap mechanism (Mayer and Armstrong, 2004). At the level of the agonist binding sites, it is known that NMDA receptor gating has an absolute requirement for the binding of two molecules of glutamate and two molecules of glycine (Benveniste and Mayer, 1991; Clements and Westbrook, 1991). Again, this is fundamentally different from the effects of inhibitory agonists of the NR2-NTD where one agonist molecule is sufficient to produce partial inhibition (see above). Since all the agonist binding sites need to be occupied for the channel to open, introducing one subunit with a mutation which causes a very large shift in glutamate sensitivity, such as T690I, might prevent channel activation. That T690I-containing NR2A subunits did not act in a negative dominant manner (see Fig. 3) provides new information about the nature of the interactions between glutamate binding sites. It suggests a strong cooperativity between the two glutamate binding sites, such that the site of the T690I containing subunit increases in affinity when the site of the wt subunit is occupied. We cannot say how this cooperativity arises, but a likely explanation for this phenomenon is a direct physical interaction from one glutamate binding domain to the other. This seems a particularly attractive hypothesis, since isolated glutamate binding domains of AMPA GluR2 subunits form dimers (Sun et al., 2002) and since NR1 glycine binding domains have also been suggested to form homodimers (Neugebauer et al., 2003; but see also Furukawa and Gouaux, 2003). Thus, despite the fact that NTDs fold similarly to the ABDs and also dimerize (Kuusinen et al., 1999; Ayalon and Stern-Bach, 2001; Meddows et al., 2001), NTDs and the ABDs appear to function differently. This could reflect differences in domain-domain interaction, as the interface seen in dimeric GluR2 ABDs (Sun et al., 2002) differs from that seen in dimeric mGluR ABDs (Kunishima et al., 2000) which are structurally related to the NTD of ionotropic glutamate receptors. Alternatively, this could indicate differences in quarternary arrangement, with homodimers, NR1-NR1, NR2-NR2, formed at the level of the ABDs but heterodimers, NR1-NR2 formed at the level of the NTDs.

The presence of NR1/2A/2B receptors in adult forebrain neurons was demonstrated as early as 1994 (Sheng et al., 1994; see also Didier et al., 1995; Chazot and Stephenson, 1997; Luo et al., 1997), yet little is known of their functional properties. Similarly, there is increasing evidence that triheteromeric NR1/2A/2C receptors are present in the adult cerebellum (Wafford et al., 1993, Chazot et al., 1994; Cathala et al., 2000). There is also accumulating evidence that Zn²⁺ ions, which are stored and released at many glutamatergic synapses in the brain, are important endogenous regulators of NMDA receptor activity (Smart et al., 2004). Basal Zn levels are thought to provide tonic inhibition of NMDARs by binding at high-affinity (nanomolar) Zn binding sites, while synaptically released Zn could provide phasic inhibition of NMDARs by binding at sites of lower (micromolar) affinity (see Rachline et al., 2005). We now demonstrate that both NR1/2A/2B and NR1/2A/2C receptors retain NR1/2A-like nanomolar Zn sensitivity. This suggests that NR2A-containing triheteromeric NMDARs, similarly to NR1/2A receptors, may also be subject to tonic inhibition at physiological concentrations of ambient Zn. However, because the degree of inhibition is reduced compared to diheteromeric NR1/2A receptors, the regulation of NR2A subunit expression relative to that of NR2B or NR2C subunits might provide a way by which cells can fine tune background NMDA receptor activity as a function of ambient Zn concentration. Another striking feature of NR1/2A/2B receptors, is their graded inhibition over a range of Zn concentrations, spanning almost three orders of magnitude, from the nanomolar to micromolar range. So, unlike NR1/2A receptors, which are strongly inhibited by tonic (nanomolar) Zn levels, and NR1/2B receptors, which only sense phasic (micromolar) Zn (Rachline et al., 2005), triheteromeric NR1/2A/2B receptors could sense both tonic and phasic Zn. Finally, we find that triheteromeric NR1/2A/2B receptors retain high-affinity ifenprodil sensitivity but that the degree of inhibition strongly depends on the occupancy of the neighbouring NR2A NTD high-affinity Zn binding site and thus on ambient Zn levels. Since ifenprodil is the prototype compound of a family of NR2B-specific antagonists with promising neuroprotective properties (Kemp and McKernan, 2002), this finding could have important implications for the effects of ifenprodil-like compounds in vivo and their therapeutic potentials.