Abstract
Ion channels alternate stochastically between two functional states, open and closed. This gating behavior is controlled by membrane potential or by the binding of neurotransmitters in voltage- and ligand-gated channels, respectively. Although much progress has been made in defining the structure and function of the ligand-binding cores and the voltage sensors, how these domains couple to channel opening remains poorly understood. Here we show that the M3 transmembrane segments of the NMDA receptor allosterically interact with both the ligand-binding cores and the channel gate. It is proposed that M3 functions as a transduction element whose conformational change couples ligand binding with channel opening. Furthermore, amino acid homology between glutamate receptor M3 segments and the equivalent S6 or TM2 segments in K+ channels suggests that ion channel activation and gating are both structurally and functionally conserved.
- ion channel gating
- affinity
- efficacy
- receptor structure
- ligand binding
- neurotransmitters
- activation mechanism
- NMDA receptor
Cation channels consist of multiple domains or subunits arranged around a central ion-conducting pore. The pore-forming regions of K+ channels and glutamate receptors (GluRs) display significant amino acid sequence conservation (Wo and Oswald, 1995; Wood et al., 1995), suggesting that they have evolved from a common ancestor (Wood et al., 1995). Identification of GluR0, a prokaryotic ionotropic glutamate receptor with a K+-selective pore (Chen et al., 1999), confirmed this notion and further strengthened the idea that K+ channels and glutamate receptors are structurally related (Fig.1A). The amino acid sequence of the “missing link” GluR0 allowed us to extend this homology to the transmembrane segment that follows the P region (Fig.1B). In glutamate receptors, this M3 segment contains a nine amino acid sequence (SYTANLAAF) that is highly conserved throughout all members of the family. The alanine at position 8 of this motif is mutated to threonine in the δ2glutamate receptor of the lurcher mutant mouse (Zhou et al., 1997). Whereas wild-type δ2 receptors are nonfunctional in heterologous expression systems, the lurcher mutation (A654T) results in constitutive activation of the channel and causes neurodegeneration. Introducing the same mutation in the GluR1 AMPA receptor increases agonist potency and converts a competitive antagonist into an agonist (Taverna et al., 2000). Introducing thelurcher mutation into AMPA, kainate, and NMDA receptors reduced receptor desensitization and reduced the rate of receptor deactivation (Kohda et al., 2000). Because the M3 segment is not part of the ligand-binding core, these data suggest a role for M3 in channel activation.
The ligand-binding site in glutamate receptor subunits is formed by two domains, S1 in the N terminal and S2 in the M3–M4 linker (Stern-Bach et al., 1994) (Fig. 1A). The S1S2 ligand-binding core of GluR2 has been crystallized, with a partial agonist molecule bound to the structure (Armstrong et al., 1998). More recently, additional structures have been obtained for the GluR2 ligand-binding core, including an agonist-free (apo) conformation, conformations with full and partial agonists, and an antagonist-bound conformation (Armstrong and Gouaux, 2000). The S1–S2 structure reveals two lobes connected by a hinge, forming a clamshell-like structure, similar to the bacterial periplasmic binding proteins (Oh et al., 1993). These novel GluR2 structures confirm and extend what was known for the bacterial periplasmic binding proteins, namely that ligand binding induces a conformational change in which the lobes rotate on the hinge and collapse around the ligand. These crystallographic data suggest a model for agonist activation of glutamate receptors (Armstrong and Gouaux, 2000), in which the binding of ligand stabilizes the closed form of the clamshell (Fig. 1C–E).
Because glutamate receptors are tetramers (Laube et al., 1998;Rosenmund et al., 1998), functional receptors contain four ligand-binding sites. NMDA receptors are a subtype of glutamate receptor, the activation of which requires binding of the co-agonist glycine in addition to glutamate (Johnson and Ascher, 1987). NMDA receptors (NRs) are heteromeric; they are composed of glycine-binding NR1 subunits (Kuryatov et al., 1994; Hirai et al., 1996) and glutamate-binding NR2 subunits (Anson et al., 1998). An analysis of agonist- and antagonist-binding rates suggested that each receptor contains two glycine- and two l-glutamate-binding sites (Benveniste and Mayer, 1991; Clements and Westbrook, 1991). Despite a rather detailed understanding of ligand binding in glutamate receptors and other ligand-gated ion channels, it is not clearly understood how closure of the ligand-binding lobes leads to opening of the channel gate (Colquhoun, 1998). The data presented here demonstrate that the M3 transmembrane segments of the NMDA receptor couple ligand binding to channel gating.
MATERIALS AND METHODS
Mutations were generated using the megaprimer PCR method, as described previously (Wood et al., 1995), and were confirmed by DNA sequencing. cRNA for wild-type and mutant NMDA receptor subunits were prepared by in vitro transcription, as described previously (Wood et al., 1995).
Oocyte preparation and injection. Adult female Xenopus laevis were anesthetized, and a few ovarian lobes containing stage V or VI oocytes were removed. The incision was sutured and the frog was allowed to recover. Oocytes were dispersed by manually disrupting the egg sac and were digested for 2 hr with collagenase type I (2 mg/ml) (Invitrogen, Gaithersburg, MD) to further remove the follicular layer. Injection of cRNA (75 nl of a 10–100 ng/μl solution) was performed under a dissecting microscope with a micrometer-driven micropipetter. Oocytes were then transferred to SOS buffer (in mm: 100 NaCl, 2 KCl, 1.8 CaCl2, 1 MgCl2, 5 HEPES, pH 7.6) supplemented with gentamicin (100 μg/ml) and fungizone (100 μg/ml) and maintained at 19°C.
Two-electrode voltage clamp electrophysiology. Functional expression was assessed 2–5 d after injection using a two-electrode voltage-clamp amplifier (OC-725; Warner Instrument, Hamden, CT) by application of agonists during continuous perfusion of a buffer solution containing (in mm): 100 NaCl, 5 KCl, 0.5 BaCl2, 10 HEPES, 10 μm EDTA, pH 7.3, at 25°C. Barium, rather than calcium, was used as the divalent cation to minimize secondary activation of calcium-activated Cl−currents (Leonard and Kelso, 1990). EDTA was used to chelate trace amounts of the soft-metal divalent cations Cd2+ and Zn2+, which have been reported to contaminate buffer solutions (Paoletti et al., 1997) and inhibit the NMDA receptor by binding to a high-affinity site (Low et al., 2000;Paoletti et al., 2000). EDTA also removes a zinc-dependent component of desensitization (Zheng et al., 2001). Oocytes were placed in a perfusion chamber (Warner Instrument) that was optimized for laminar flow. Solution changes were accomplished using a gravity-fed, computer-controlled perfusion system, which uses BIO-SIL silicone rubber tubing (3.8 mm outer diameter; 1.5 mm inner diameter; Sil-Med, Taunton, MA) and solenoid valves (General Valve, Fairfield, NJ). During solution exchanges, the solution flow rate was ∼15 ml/min. Solution exchange kinetics was quantified as follows. Inward NMDA currents were recorded at a holding potential of −60 mV in a buffer containing (in mm) 100 NaCl and 0.5 BaCl2, after application of 10 μm glycine plus 100 μml-glutamate. When steady state was reached, all of the Na+ ions were replaced with impermeableN-methyl-d-gluconate ions, resulting in an instantaneous change in the driving force. The inward NMDA current decayed with a biexponential time course. The fast component had a time constant of 180 msec and a relative amplitude of 89%. The remaining 11% decayed with a time constant of 2.3 sec. This implies that ∼90% of the solution exchange is complete within 200 msec, but the remaining 10% takes several seconds to complete. Low-resistance glass microelectrodes (0.5–2.0 MΩ) were filled with 3m KCl and 10 mm HEPES, pH 7.2, and used to impale the oocyte. Current traces were recorded from a holding potential of −60 mV. Data acquisition and voltage control were accomplished with pClamp hardware and software (Axon Instruments, Burlingame, CA).
Curve fitting. The time courses of NMDA current deactivation in Figure 3 were fitted with exponential models by minimizing the residual sum of squares (RSS), using the Solver function in Microsoft Excel (Microsoft, Redmond, WA). One- and two-exponential models were compared using the asymptotic information criterion (AIC) (Akaike, 1981; DiStephano and Landaw, 1984): AIC = Nlog(RSS) + 2P, where N is the number of data points and P is the number of parameters in the model. The best model is the one that minimizes AIC. The remaining time courses of deactivation and 2-(aminoethyl)methanethiosulfonate hydrobromide (MTSEA) modification were fitted with exponential models using the Clampfit module in pClamp 6.0 (Axon Instruments).
RESULTS
The role of the M3 segment in the activation pathway of glutamate receptors was examined because (1) the lurcher phenotype suggested that an M3 residue (A654) had a role in receptor gating and (2) the strict amino acid conservation in the C-terminal half of this transmembrane domain among all of the members of the glutamate receptor family suggested an important functional role. Amino acids in the SYTANLAAF region of NMDA receptor subunits were individually substituted with cysteines. Because functional NMDA receptors are heteromeric assemblies of NR1 and NR2 subunits, mutations were introduced into each subunit separately and coexpressed inXenopus oocytes with the complementary wild-type subunit. NMDA receptors containing mutant subunits were then exposed to the thiol-modifying reagent MTSEA (Akabas et al., 1992) to explore the function of these residues in activation of the receptor. Because both the NR1 and NR2A subunits contain several extracellular cysteines, the effect of MTSEA treatment was first characterized in the wild-type NMDA NR1 plus NR2A receptor.
Effects of MTSEA on wild-type NMDA receptor function
MTSEA treatment of the wild-type NMDA receptors in the absence of agonist had no significant effect on current amplitude (Fig.2A). Simultaneous application of MTSEA and agonists did cause a significant inhibition of NMDA current, but this effect was reversed after washout (Fig. 2B). Given its size and charge, it is likely that MTSEA can act as an open channel blocker in NMDA receptors, as has been shown for other channels (Liu and Siegelbaum, 2000). To test this idea, MTSEA was applied at positive holding potentials, where it exerted no effect on current amplitude (data not shown). Therefore, the reversible inhibition by MTSEA in the presence of agonist likely represents open channel block and not covalent modification of a thiol group. Because MTSEA application had no lasting effect on the current amplitude evoked from wild-type NMDA receptors, we conclude that either the endogenous cysteines of the receptor were not available for modification or their modification caused no discernible change in current amplitude.
The SYTANLAAF region
Amino acids in the SYTANLAAF region of NMDA receptor subunits were individually substituted with cysteines. Eight of the nine cysteine-substituted NR1 constructs were functional. However, NMDA receptors containing the NR1-S646C mutation did not yield current. The remaining receptors were exposed to the thiol-modifying reagent MTSEA, both in the absence and in the presence of agonists. Four of the cysteine-substituted NR1 receptors exhibited state-dependent responses to MTSEA application. Simultaneous exposure to MTSEA and agonist reduced the current amplitudes of receptors containing the NR1 mutants T649C, N651C, or L652C by 75 ± 8%, 64 ± 9%, and 40 ± 8%, respectively (data not shown). However, exposure to MTSEA in the absence of agonists did not reduce the NMDA currents in these mutants (data not shown). MTSEA modification of NR1-A653C revealed a critical role in receptor activation. This mutation targets the alanine at position 7 of the SYTANLAAF region and will be referred to as A7C. It should be noted that the A7C mutation immediately precedes thelurcher mutation A8T. The remainder of this article characterizes the unique properties conferred by MTSEA modification of NMDA receptors containing the A7C mutation in either the NR1 or NR2 subunit.
Access to residue A7C requires receptor activation
The amplitudes of currents evoked from NMDA receptors containing the A7C mutation in the SYTANLAAF region of either NR1 or NR2A were comparable with that of the wild-type receptor. The glutamate and glycine concentration–response curves of the A7C mutants did not indicate substantial changes in the apparent affinity (EC50) of either of the co-agonists (Table1). This suggested that the overall structure of the channel was not grossly disrupted by the mutation. NMDA current evoked from receptors containing the A7C mutation in either the NR1 or NR2A subunit was not affected by exposure to MTSEA in the absence of agonist (Fig. 2C,E), although simultaneous application of MTSEA and agonists potentiated the current amplitude and markedly decreased the deactivation rate (Fig.2D,F). This indicates a pronounced activation-state dependence for MTSEA modification of the A7C-containing receptors. Similar results were obtained when MTSEA was substituted with the membrane-impermeable analog [2-(trimethylammonium)ethyl] methane thiosulfanate (data not shown), indicating that access to residue A7C occurs from the extracellular space. Taken together, these data suggest that MTSEA gains access to the A7C residue only when the receptor is activated.
MTSEA modification slows deactivation of A7C NMDA receptors
The current evoked during activation of ligand-gated ion channels is rapidly abolished after removal of the agonist. This deactivation occurs as the agonist dissociates from its binding site and the ion channel returns to its resting state. Deactivation of wild-type NR1/NR2A NMDA receptors proceeded with a biexponential time course (Fig. 3). This time course may in part be determined by the relatively slow perfusion system (see Materials and Methods). The deactivation kinetics of A7C-containing NMDA receptors was also biexponential, and the time constants obtained were statistically indistinguishable from the wild-type receptor (Fig. 3). MTSEA application in the absence of agonists did not affect the deactivation kinetics of the wild-type or A7C-containing NMDA receptors. Concurrent application of MTSEA and agonist had no effect on the wild-type NMDA receptor, but it dramatically slowed deactivation of NR1-A7C and NR2A-A7C receptors. Figure 3 summarizes the effects of MTSEA modification on the deactivation kinetics of A7C-containing NMDA receptors, as well as the requirement for the presence of agonists. MTSEA modification also potentiated the amplitude of currents evoked from the A7C NMDA receptors. Interestingly, the potentiation of NMDA current occurred despite activation of the A7C NMDA receptors with supramaximal concentrations of agonists (Fig. 2D). The effect that MTSEA exerts on the deactivation kinetics and current amplitudes of the A7C NMDA receptors was examined more thoroughly.
Modification of A7C requires closure of both glutamate- and glycine-binding cores
The experiments described above suggested that activation of the NMDA receptor is required for MTSEA to gain access to the A7C residue. Agonist binding induces receptor activation by a series of conformational changes that culminate in channel opening. This process is initiated by closure of the bilobate binding cores (Armstrong and Gouaux, 2000). In the NMDA receptor, activation requires binding of both glycine and glutamate at the NR1 and NR2A subunits, respectively. Therefore, we investigated whether the application of either co-agonist alone was sufficient to allow MTSEA access to the A7C residue. Coapplication of MTSEA and l-glutamate failed to evoke current from either NR1-A7C- or NR2A-A7C-containing receptors and did not result in their covalent modification (data not shown). However, coapplication of MTSEA and glycine did produce a very slowly developing current in A7C-containing NMDA receptors that required several minutes to reach steady state (Fig. 4). Interestingly, glycine by itself was found to partially activate the NMDA receptor to a level between 1 and 5% of the maximal activation obtained with both co-agonists (Fig. 4). These data suggest that our glycine solutions may have been contaminated by trace amounts ofl-glutamate, or alternatively, that glycine may act as a partial agonist at the glutamate-binding site. In both scenarios, the glutamate-binding site is partially activated by an agonist. Therefore, we investigated whether the glutamate-binding site plays a critical role in the glycine-only NMDA current, which supports the modification by MTSEA. Coapplication of the competitive l-glutamate antagonist APV reduced the glycine-only currents evoked from the A7C NMDA receptors in a dose-dependent manner (Fig. 4). Concomitant with the reduction in receptor activation, APV also reduced the rate and extent of MTSEA modification observed during glycine application to the A7C NMDA receptors (Fig. 4). Although these results do not fully resolve the nature of the glycine-only currents evoked from A7C-containing NMDA receptors, they indicate that MTSEA modification of A7C requires that both the glycine- and glutamate-binding sites be occupied by effective agonists. The absence of glycine from its binding site or the presence of an antagonist in the glutamate site is sufficient to prevent MTSEA modification.
From the data described thus far, it could be concluded that lobe closure of both the glutamate- and glycine-binding sites is required to allow the conformational change in M3 that makes residue A7C accessible to MTSEA. However, lobe closure at the ligand-binding sites induces a cascade of conformational rearrangements that result in channel opening, Therefore, it is not clear which of the several steps involved directly causes MTSEA accessibility to residue A7C. The conformational change in M3 could result directly from ligand binding, or alternatively, it could be a consequence of channel gating. This point was further elaborated by investigating the correlation between the rate of MTSEA modification with ligand binding and gating parameters. Because thiol modification by MTSEA is extremely rapid, up to 105m−1sec−1 (Stauffer and Karlin, 1994), the chemical reaction is not the rate-limiting step of modification. Rather, the rate of MTSEA modification depends critically on the relative amount of time that the cysteine at position A7 is exposed to the aqueous environment of the receptor.
MTSEA modification rate depends on agonist concentration
The relationship between MTSEA modification rate and agonist concentration was investigated to correlate the activation state of the A7C NMDA receptors with MTSEA accessibility to residue A7C. When lower concentrations of agonists were used, MTSEA was found to potentiate the NMDA currents substantially (Fig.5A), suggesting that modified receptors are much more sensitive to agonists. The rate of MTSEA modification was much slower when lower concentrations of agonists were used (Fig. 5A). The time course of current potentiation was best fit by a single exponential function, and the time constant was dependent on the concentration of bothl-glutamate and glycine. However, the MTSEA-induced potentiation of current evoked from the NR2A-A7C receptor became biexponential at the highest agonist concentrations. In all of the cases, the NMDA current potentiation became faster with increasing agonist concentration and the kinetics saturated at higher concentrations. Modification rate constants were plotted as a function of the agonist concentration and fitted with the Hill equation. The resulting half-maximal concentrations (EC50) and Hill coefficients are summarized in Table 1. Normalization of the rate constants allowed for direct comparison between the modification kinetics and the activation state of the NMDA receptor, as reflected by the concentration–response curves (Fig. 5B,C). The kinetics of current amplitude potentiation paralleled the concentration-dependent activation of the receptors. The two curves, steady-state activation and MTSEA modification rate, overlapped when plotted as a function of glycine concentration for the NR1-A7C receptors (Fig. 5B). However, the MTSEA modification rate curve of the NR2A-A7C receptors was significantly left-shifted from the activation curve when plotted as a function of glutamate concentration (Fig. 5C). This discrepancy suggests that A7C accessibility is not directly determined by the activation state (the open probability) of the receptor, as reflected by the normalized concentration–response curve. Instead, the conformational changes in M3 display a differential sensitivity to occupancy at the glycine- and glutamate-binding sites that cannot be derived from channel gating alone.
MTSEA modification of residue A7C depends on agonist efficacy
The recent crystallographic structures of the GluR2 ligand-binding core suggest a model in which interaction of ligands with their binding sites is characterized by two distinct processes, agonist binding and domain closure (Fig. 1C–E). These molecular reactions can be described by either a linear three-state model or a more general four-state allosteric model (Colquhoun, 1998). In the linear model, the agonist interacts with the open conformation of the ligand-binding core in an initial binding reaction. Once bound, the agonist stabilizes a closed conformation of the ligand-binding clamshell (Fig.1D). This closed conformation of the ligand-binding core promotes receptor activation and ultimately results in channel opening. In the allosteric model, the ligand-binding core is able to visit the closed conformation spontaneously, even in the absence of bound agonist, although the probability is very low (Fig.1E). However, agonist occupancy facilitates domain closure by stabilizing the closed conformation.
The concentration dependence of the MTSEA modification rate (Fig.5B,C) correlates A7C accessibility with agonist occupancy. However, there are two distinct conformations in which the agonist occupies the ligand-binding core: the agonist-bound–open (A·R) state and the agonist-bound–closed (A·R*) state (Abele et al., 2000). Therefore, we wanted to determine which of these conformations support the accessibility change of residue A7C. This important issue was investigated by activating A7C NMDA receptors with partial agonists before exposure to MTSEA. The models describing the interaction between the agonist and its binding site (Fig. 1C,D) are reduced to a two-state scheme at saturating agonist concentrations, in which the probability of observing the unoccupied receptor state(s) becomes vanishingly small. Under these conditions, the ligand-binding cores alternate stochastically between the closed (A·R*) and open (A·R) conformations. The degree of domain closure induced by agonists can vary substantially (Armstrong and Gouaux, 2000) and is an important determinant of the efficacy of the agonist (E in Fig. 1D,E). Experimentally, agonist efficacy is estimated by its intrinsic activity, α (Ariens, 1954).
By activating A7C NMDA receptors with saturating concentrations of agonists with different intrinsic activities, it is possible to correlate the accessibility of residue A7C with the position of the equilibrium between the A·R and A·R* conformations of the ligand-binding core. MTSEA modification of A7C NMDA receptors proceeded more slowly when partial agonists were substituted forl-glutamate or glycine (Fig.6A). Therefore, accessibility of residue A7C, as measured by the rate of MTSEA modification, depends critically on the stability of the agonist-bound–closed conformation (A·R*) of the ligand-binding core. The MTSEA modification rate of the NR1-A7C receptor correlated linearly with the intrinsic activity of the glycine-site agonists, and the modification rate of the NR2A-A7C receptor correlated linearly with the intrinsic activity of the glutamate-site agonists (Fig.6B,C). This could be expected because closure of the NR1 ligand-binding core is induced by glycine and glycine-like compounds, and closure of the NR2 ligand-binding domains are induced byl-glutamate and glutamate-like compounds.
Surprisingly, MTSEA modification of NR2A-A7C receptors also correlated linearly with the intrinsic activity of the agonist bound to the NR1 subunit (Fig. 6C). However, modification of the NR1-A7C receptor did not correlate linearly with the intrinsic activity of NR2-bound agonists (Fig. 6B). For example, substituting l-glutamate with agonists that have intermediate intrinsic activities did not reduce the rate of MTSEA modification, although substitution with (±)-cis-piperidine-2,3-dicarboxylic acid (PDA), a ligand of particularly low efficacy (α = 0.07), did result in a significantly slower modification. These data suggest that the access of MTSEA to residue A7C is promoted when the binding sites are in the agonist-bound–closed (A·R*) state. Furthermore, the data revealed a distinct asymmetry in the intersubunit communication between the NR1 and NR2A NMDA receptor subunits, in which the agonist bound at the glycine site exerts a dominant influence over the M3 conformation in its own NR1 subunit as well as the neighboring NR2 subunits.
MTSEA modification increases the efficacy of partial agonists
MTSEA modification of the A7C NMDA receptors increased the current amplitude to levels that were equal to or greater than the responses evoked by full agonists. This occurred in 9 of 10 cases (Fig.6D,E). The current amplitude evoked by the lowest-efficacy glycine agonist, (±)-3-amino-1-hydroxy-2-pyrrolidone (HA-966), was potentiated 70-fold by MTSEA modification. Whereas before modification the relative current amplitudes evoked by partial agonists varied widely, reflecting large differences in intrinsic activity, MTSEA modification potentiated current amplitudes to levels that are much more similar (Fig. 6D,E). Therefore, MTSEA modification of A7C NMDA receptors appears to increase the intrinsic activity of partial agonists, converting the majority to full agonists.
DISCUSSION
The data shown here implicate the M3 segment as a critical element in the activation pathway of NMDA receptors. Cysteine substitution of the alanine at position 7 of the SYTANLAAF motif (A7C) revealed agonist-induced accessibility changes. A previous report that studied the same cysteine substitutions (Beck et al., 1999) concluded that none of the positions displayed agonist-dependent accessibility to MTSEA. This apparent conflict can be reconciled by considering the different experimental approaches used. In the present experiments, glutamate and glycine were always coapplied and the buffer used between agonist applications was agonist-free. In the studies performed by Beck et al. (1999), glycine was present throughout the experiment and receptors were activated by the addition of l-glutamate. Consequently, when A7C was treated with MTSEA, it was found that the baseline current slowly increased, which made additional testing difficult. As we have shown, glycine by itself will minimally activate NMDA receptors and permit MTSEA modification, which very slowly increases the size of the current. This modification is prevented by occupying the glutamate-binding site with an antagonist, suggesting that agonist occupancy of both binding sites is required to allow MTSEA access to A7C. Alternatively, it is possible that occupancy of the glycine-binding site is sufficient to activate the receptor partially. In this scenario, the inhibitory effect of APV suggests that glycine binding at the NR1 subunit allows the glutamate-binding core to close in the absence of ligand.
The activation-dependent accessibility of residue A7C does not by itself place M3 in the chain of events that links ligand binding to channel opening. M3 could undergo a conformational change concomitant with channel opening without having a causative role. However, the additional data shown here strongly support the idea that M3 is a transduction element critically involved in coupling ligand binding to channel opening (Fig. 7).
M3 is functionally distinct from the gate
The NMDA receptor is unique among receptors in its dual agonist requirement. Binding reactions at all four subunits allosterically project to a single gate, which “integrates” the information and generates a single output, the open probability. If the accessibility of A7C is completely determined by open probability, then MTSEA modification rate should depend only on the state of activation. However, the modification rate shows a concentration dependence that is distinct (in both its Hill coefficient and its midpoint) from the normalized response curves (Fig. 5, Table 1). Furthermore, the MTSEA modification rate of NR1-A7C is the same for the partial agonist quinolinic acid (QA) and the full agonist l-glutamate, despite that fact that QA activates the NMDA receptor by only 50% (Fig. 6B). Therefore, A7C accessibility is not determined strictly by the open probability of the channel. It appears that M3 receives information on events taking place at the glycine- and glutamate-binding sites that cannot be extracted from channel gating alone.
M3 interacts allosterically with the glutamate- and glycine-binding sites
The MTSEA modification rate of A7C-containing NMDA receptors was determined by both agonist concentration and efficacy. Supramaximal concentrations of the full agonists glycine and glutamate, which promote efficient closure of the ligand-binding lobes, permitted rapid modification. Likewise, low concentrations of full agonists or high concentrations of partial agonists, both of which induce infrequent or incomplete domain closure, significantly slowed the rate of modification. These results imply that the activation state of the ligand-binding cores modulates conformational changes in M3 (Fig. 7,arrow a). Conversely, M3 also affects the behavior of the ligand-binding lobes, as MTSEA modification of A7C increased the efficacy of the partial agonists. This suggests that MTSEA modification of position A7 in the SYTANLAAF region of the M3 domain alters the interaction between the ligand-binding core and its agonist (Fig. 7,arrow b). Introducing a bulky side chain at A7C allosterically shifts the equilibrium between the open and closed cleft conformations. From this it is concluded that the M3 segments and the ligand-binding cores of NMDA receptors interact through a reciprocal, allosteric mechanism that determines agonist efficacy.
M3 interacts allosterically with the gate
The conformation of M3 affects the behavior of the gate, which is the pore structure directly responsible for opening the channel. MTSEA modification of A7C dramatically affects NMDA receptor gating by decreasing the rate at which the receptor deactivates after agonist removal (Fig. 3C). MTSEA modification also results in activation of the NMDA receptor to a level not obtainable by agonists alone (Figs. 2F, 6E). Taken together, these results suggest that activation of the NMDA receptor evokes conformational changes in the M3 domain, which are then allosterically transduced to affect the behavior of the gate (Fig. 7,arrow c). An allosteric interaction between M3 and the gate must necessarily be bidirectional, such that alterations in gating behavior (caused by a mutation or drug) should affect conformational changes in M3 (Fig. 7, arrow d). We have not presented data that support this idea.
Cooperativity and asymmetry
The model in Figure 7 schematically illustrates how the M3 segment allosterically mediates the coupling between the ligand-binding core and channel opening. However, the model is incomplete because it shows the ligand-binding site and M3 segment for a single subunit, whereas functional NMDA receptors are most likely tetrameric. Moreover, our results have uncovered a high degree of cooperativity between the movement of the M3 segments in NMDA receptor subunits that needs to be taken into account. Closure of just the glutamate- or glycine-binding lobes is not sufficient to induce the conformational changes in M3 that make residue A7C accessible to MTSEA. However, when both types of binding lobes are closed, MTSEA gains access to residue A7C. This was required whether the cysteine substitution was introduced in the NR1 or NR2A subunit. The rate of modification of a NR2A-A7C-containing NMDA receptor depended critically on the activation status of both its “own” glutamate-binding core as well as the glycine-binding core localized to the neighboring NR1 subunit. The MTSEA modification of the NR1-A7C receptor was similarly dependent on the conformation of the ligand-binding cores, although the relationship between the modification rate and efficacy for partial glutamate agonists was distinctly nonlinear. This suggests that domain closure of the individual glycine and glutamate bilobate binding cores does not provide enough free energy to induce a conformational change in the M3 segments. Closure of the ligand-binding lobes at both the NR1 and NR2A subunits is required to produce the change in accessibility.
This high degree of cooperativity in the behavior of the M3 segments can be explained by their interaction with a common, single gate (Colquhoun, 1998). Alternatively, the subunits may interact with each other, potentially through the M3 or M1 segments, to produce the cooperativity that was observed. This could also make it easier to explain the asymmetry that has been observed in the interaction between the glutamate- and glycine-binding sites (Benveniste et al., 1990;Lester et al., 1993) (Fig. 6B,C).
A conserved transduction element in glutamate receptors and K+ channels
The M3 domains allosterically interact with both the gate and the ligand-binding sites in the NMDA receptor and mediate their coupling. Ligand binding alters the conformation of M3, which in turn changes the probability that the channel will open. What is the nature of this conformational change that occurs in M3? Data obtained recently for K channels may provide a clue. Pore-forming regions are structurally conserved between ionotropic glutamate receptors and K+ channels (Fig. 1A,B). The glutamate receptor M3 segment corresponds to TM2 in the bacterial KcsA K channel, which has been crystallized recently (Doyle et al., 1998). Electron paramagnetic resonance spectroscopy measurements in KcsA have shown that the α-helical TM2 segment displays rotational and translational movements after activation of the channel (Perozo et al., 1998, 1999). The data presented here are consistent with a similar rotation/translation of the M3 segment underlying NMDA receptor activation.
Furthermore, introduction of a cysteine in the C-terminal half of S6, the equivalent transmembrane segment of the voltage-gated Shaker K+ channel, results in a phenotype very similar to that of A7C in the NMDA receptor. Like A7C-containing NMDA receptors, modification of the S6 cysteine requires channel activation. In addition, deactivation of the Shaker K+ channel was slowed by the presence of soft-metal divalent cations, which can form intersubunit metal bridges (Holmgren and Yellen, 1998). Therefore, we propose that the M3, TM2, and S6 segments are structurally and functionally conserved. They represent universal transduction elements whose translocation couples ligand binding or voltage-sensor movement to channel opening. It therefore appears that glutamate receptors and K channels not only contain structurally highly conserved pore-forming regions but also share a common gating mechanism.
Footnotes
This work was supported by National Institutes of Health Grants NS31557 and MH61506 to A.M.J.V. and by a Minority Predoctoral Fellowship in Neuroscience from the American Psychological Association to K.S.J. We thank Dr. John York for his critical reading of the manuscript.
Correspondence should be addressed to Antonius M. J. VanDongen, Department of Pharmacology and Cancer Biology, Duke University Medical Center, P.O. Box 3813, Durham, NC 27710. E-mail: vando005{at}mc.duke.edu.