The NMDA receptor opens in response to binding of NMDA and glycine. However, it remains unclear where and how gating of the NMDA receptor pore is accomplished. We show that different point mutations between S645 and I655 (thus including the highly conserved SYTANLAAF motif) of M3c in NR2B lead to constitutively open channels. The current through these constitutively open channels are readily blocked by external Mg2+ and MK-801 [(+)-5-methyl-10,11-dihydro-5H-dibenzo [a,d] cyclohepten-5,10-imine maleate]. Also, the open-channel blocker MK-801 can no longer be trapped in these channels when NMDA and glycine are washed off. Moreover, M3c residues at or below A651(NR2B, A7 in SYTANLAAF) react with external methanethiosulfonate (MTS) reagents ∼500 to 1000-fold faster in the presence than in the absence of agonists NMDA and glycine. In fact, the MTS modification rate shows exactly the same NMDA concentration dependence as channel activation. In contrast, those residues external to A651 are always modified with similar kinetics whether NMDA and glycine are present or not. Interestingly, MTS modification of A651C(NR2B) holds the channel constitutively open. Mutations of A651(NR2B) into arginine, tryptophan, or phenylalanine, and similar mutations of the corresponding A652 in NR1 also lead to constitutively open channels. Double-mutant cycle analysis further shows that the effects of A652(NR1) and A651(NR2B) mutations are evidently non-additive (i.e., cooperative) if mutated into residues with large side chains or with compensatory charges [e.g., A652E(NR1)+A651R(NR2B)]. The side chain of A7 thus plays a determinant role in the intersubunit distance at this level, which is directly responsible for the activation gate and activation–deactivation gating of the NMDA receptor.
- activation gate
- constitutive opening
- gating mechanism
- intersubunit distance
- mutant cycle analysis
- NMDA receptor
The NMDA receptor opens and closes in response to binding of agonists glutamate (or NMDA) and glycine. Although the agonist-binding domains of the NMDA receptor have been characterized (Kuryatov et al., 1994; Laube et al., 1997; Anson et al., 1998; Furukawa and Gouaux, 2003; Chen et al., 2005; Furukawa et al., 2005), it remains unclear how opening and closing of the pore is accomplished. Each NMDA receptor is composed of four (two NR1 and two NR2) subunits. Each subunit has three transmembrane segments (M1, M3, and M4), a reentrant loop (M2), an extracellular N terminus [the N-terminal domain (ATD)], and an intracellular C terminus (for review, see Dingledine et al., 1999; Mayer and Armstrong, 2004). The NMDA channel pore is formed internally by the M2 loop and externally by pre-M1, M3c (the C-terminal part of M3) and M4n (the N-terminal part of M4) from all four subunits (Kuner et al., 1996; Beck et al., 1999; Kashiwagi et al., 2002).
The pore region of the NMDA receptor shares evolutionary and structural kinship with an “invertedly” positioned K+ channel (Chen et al., 1999; Kuner et al., 1996, 2003; Panchenko et al., 2001). Crystal structures of the KcsA (closed) and MthK (open) K+ channels reveal that the inner helices form a gate-like constriction (the bundle crossing) at the intracellular end of the pore (Doyle et al., 1998; Jiang et al., 2002b). Similar localization of the activation gate at the intracellular end of the pore has also been inferred for Na+ channels (Kuo and Liao, 2000; Sunami et al., 2004) and Ca2+ channels (Xie et al., 2005). However, the access of Ag+ or methanethiosulfonate (MTS) regents to the pore is not equivalently limited by the bundle crossing in cyclic nucleotide-gated and small-conductance Ca2+-activated K+ channels (Flynn and Zagotta, 2001; Bruening-Wright et al., 2002), suggesting that the bundle crossing may not form a physical gate in all channels with evolutionary kinship. It is thus interesting to see whether the activation gate of the NMDA receptor has similar location and structural features to Kv channels.
The M3c domain of all glutamate receptor (GluR) subunits contains a highly conserved nine-amino acid motif (SYTANLAAF). The A-to-T mutation at position 8 of this motif (A8T mutation) in the δ2 glutamate receptor results in constitutive opening of the channel and causes neurodegeneration in the lurcher mutant mouse (Zuo et al., 1997). The T648A (T3A) and A649C (A4C) mutations in NR1 also made constitutively open channels (Kashiwagi et al., 2002; Sobolevsky et al., 2007). The M3c domain also shows significant gating conformational changes and thus may play an essential role in activation–deactivation gating of the receptor (Sobolevsky et al., 1999; Kohda et al., 2000; Jones et al., 2002; Kashiwagi et al., 2002; Low et al., 2003; Yuan et al., 2005) (but see Beck et al., 1999; Sobolevsky et al., 2002a,b, 2007). Based on findings that the M3c residues below but not above A651(NR2B, A7 of SYTANLAAF) show state-dependent external MTS modification kinetics tightly correlated with channel activation and that the effects of A652(NR1) and A651(NR2B) double mutations are evidently cooperative in holding the channel constitutively open, we conclude that A652(NR1)–A651(NR2B) play a determinant role in the intersubunit configuration at this level, which is directly responsible for the activation–deactivation gating of the NMDA receptor.
Materials and Methods
Site-directed mutagenesis and expression of NMDA receptors.
The rat cDNA clones of NMDA receptors used in this study are the NR1a variant and the NR2B clone. The sequence of amino acid residues in the NR1 and NR2B subunits is numbered from the initiator methionine as in the original papers of NR1 (Moriyoshi et al., 1991) and NR2B (Monyer et al., 1992), respectively. Site-directed mutagenesis was performed using the QuikChange kit (Stratagene, La Jolla, CA). For the preparation of double mutations in the same subunit, cDNA templates that already contained one mutation were used for a second mutation. Mutations were verified by DNA sequencing, and two independent clones for each mutant were tested to preclude any inadvertent mutations. The full-length capped cRNA transcripts were then synthesized from each of NR1 and NR2B using T7 and T3 mMESSAGE mMACHINE transcription kits (Ambion, Austin, TX), respectively. cRNA was stored in aliquots at −80°C. To minimize the probability of formation of homomeric NR1 receptors, a mixture of NR1 and NR2 cRNAs in a ratio of 1:5 (i.e., 0.1–4 ng of NR1 and 0.5–20 ng of NR2) (Kashiwagi et al., 2002) was injected into Xenopus oocytes (stages V–VI). Oocytes were maintained in the culture medium (96 mm NaCl, 2 mm KCl, 1.8 mm MgCl2, 1.8 mm CaCl2, 5 mm HEPES, and 50 μg/ml gentamycin, pH 7.6) at 18°C for 2–3 d before electrophysiological recordings. The culture medium was replaced daily.
Oocytes were placed in a small working-volume (<20 μl) perfusion chamber (OPC-1; AutoMate Scientific, San Francisco, CA) that is optimized for fast solution exchange. The NMDA currents were recorded at a holding potential of −70 mV and at room temperature (∼25°C) with two-electrode voltage clamp using an OC-725C amplifier (Warner Instruments, Hamden, CT). The microelectrodes, pulled from borosilicate glass and filled with 1 m KCl, had resistances of 0.5–4 MΩ. Oocytes were continuously perfused with Mg2+-free ND96 solution (96 mm NaCl, 2 mm KCl, 0.3 mm BaCl2, and 5 mm HEPES, pH 7.6). Solution exchanges were obtained using a pressure-regulated, computer-controlled perfusion system (ValveLink 8; AutoMate Scientific). The solution flow rate was ∼6 ml/min, and the kinetics of solution exchange were quantified as the following. The NMDA current (elicited by application of 100 μm NMDA plus 10 μm glycine) was recorded at a holding potential of −70 mV in Mg2+-free ND96 solution. When the current amplitude reached a steady state, Na+ ions were replaced with impermeable N-methyl-d-gluconate (NMG) ions for an instantaneous change in the driving force. The NMDA current then decayed with a two-exponential time course. The fast component had a relative amplitude of 94% and a time constant of ∼0.43 s. The remaining (6%) component decayed with a time constant of ∼1.4 s. This indicates that ∼95% of solution exchange is completed within 2 s (assuming that approximately four time constants are enough to “complete” the reaction). NMDA and glycine (Sigma, St. Louis, MO) were dissolved in water, and MK-801 [(+)-5-methyl-10,11-dihydro-5H-dibenzo [a,d] cyclohepten-5,10-imine maleate] (Tocris Biosciences, Bristol, UK) was dissolved in ethanol to make 100, 10, and 10 mm stock solutions, respectively. The stock solutions were then diluted into Mg2+-free ND96 solution to make 0.1 μm to 1 mm NMDA, 10 μm glycine, and 50 nm or 10 μm MK-801 immediate before the experiment. The final concentration of ethanol (≤0.1%) had no detectable effect on NMDA currents.
Constitutive open index.
To investigate macroscopic currents through the constitutively open channel, we obtained currents in the absence and presence of extracellular Na+. As shown in Figure 1A, oocytes were voltage clamped at −70 mV and perfused with Na+-free ND96 solution that contained the impermeable NMG ion (“NMG–ND96”: 96 mm NMG, 2 mm KCl, 0.3 mm BaCl2, and 5 mm HEPES, pH 7.6). The perfusion was then changed to one containing 96 mm NaCl (“Na+–ND96”: 96 mm NaCl, 2 mm KCl, 0.3 mm BaCl2, and 5 mm HEPES, pH 7.6) to elicit the current (“Na+-elicited current”). Subsequently, NMDA channel pore blockers Mg2+ (2 mm) or MK-801 (10 μm) were applied with Na+–ND96 solution to the oocytes (to block most, if not all, of the currents). The Mg2+-sensitive Na+-elicited current (the Na+-elicited current that could be blocked by Mg2+ and MK-801 in the absence of NMDA and glycine) is referred to as the NMDA current through the constitutively open NMDA channel. Conversely, oocytes were initially perfused with NMG–ND96 solution and then changed to Na+-ND96 solution containing 100 μm NMDA and 10 μm glycine to elicit the agonist-elicited NMDA current, which is also readily blocked by 2 mm Mg2+. In the wild-type and mutant channels, the amplitude of Mg2+-sensitive, Na+-elicited current is divided by that of Mg2+-sensitive “agonist-elicited current” to give the constitutive open index (percentage).
MTS reagents (Toronto Research Chemicals, Toronto, Ontario, Canada) were made as stocks in water everyday under ice-cold condition and mixed into the recording solution within a few minutes before use. Modification rates were determined by either the rate of change in current amplitude during continuous application of MTS reagents (see Fig. 3A) or coapplication of MTS reagents and d-2-amino-5-phosphonovaleric acid (APV) (100 μm) for different periods of time and documentation of the change in current amplitude (see Fig. 3B). Application of APV with MTS in this case would help to eliminate the possibility of contamination of the experimental results by any residual glutamate and glycine in the bath solution. The MTS concentration was raised to as high as 3 mm to measure the slowest rates and lowered to 20 μm to measure the fastest rates. High MTS concentrations could also produce a rapid, reversible reduction (10–20%) in current size that was also seen in the wild-type channel. This rapid phase was ignored in determining the specific modification rate. The change in current size was fitted with a single-exponential function to obtain the time constant (τ). The apparent second-order rate constant for MTS modification (k) was given by k = 1/(τ[MTS]), in which [MTS] is the concentration of the MTS reagent. The rate of solution exchange should be much faster than that of MTS modification itself for an accurate measurement of the modification rate. With our fast perfusion system, MTS concentrations were selected so that τ was ∼10–50 and ∼80–500 s for the continuous and periodic protocols, respectively. Current rundown was examined for all mutations in the absence of MTS reagents and was always <10%.
Data were acquired using the Digidata-1322A analog/digital interface with pClamp software (Molecular Devices, Palo Alto, CA), and the sampling rates were 0.5–1 kHz. All data are expressed as mean ± SEM. The Student's t test was used for comparison between experimental groups, and p < 0.05 was considered as statistically significant.
Mutations in M3c of the NR2B or the NR1 subunits lead to constitutively open channels
We made a series of single alanine substitutions in and around ATD, pre-M1, M2, and M3c of the NR2B subunit (i.e., alanine mutagenesis scan). For ATD, the critical residues for the high-affinity binding site of ifenprodil (a potent gating modifier of the channel) were selected (Perin-Dureau et al., 2002). For M2, residues N615 and N616 were selected. These residues presumably form the narrowest part of the channel pore and are directly responsible for the binding of pore blockers such as Mg2+ and MK-801. For pre-M1 and M3c, most of the residues were selected because these two domains presumably constitute the wall of the external pore mouth and are most likely related to channel gating (see Introduction). If the selected residue itself is alanine, it is substituted into nucleophilic cysteine (e.g., A644C, A648C, A651C, and A652C). All mutations were screened by the experimental protocol described in Figure 1A (see also Materials and Methods). The mutant receptors that have a significantly different (p < 0.05) response from the wild-type receptor are considered to contain mutations at the “key” positions. These key mutations could be approximately correlated with those mutants that had a constitutive open index >5% (vs 2 ± 0.4% for the wild-type receptor) (Fig. 1B). We found six key positions in the 46 NR2B mutations tested (all of which generated functional receptors, except F554A, W559A, N649A, and F653A). Interestingly, these six key mutations are clustered in two regions, namely M3c (S645A, T647A, A648C, and I655A) and pre-M1 (S555A and D557A), both being regions likely to line the pore. These results are consistent with the idea that M3c and pre-M1 are critically related to channel gating. Because the two positions in pre-M1 (S555 and D557) of NR2A were not substantially modified by external MTS reagents regardless of whether the agonists are present or not (Thomas et al., 2006), we would focus on the key positions in M3c. NMDA receptors are heterotetrameric receptors composed of the NR1 and NR2 subunits, both contributing to the formation of the channel pore. We therefore further examined the constitutive open index in NR1 mutant receptors. Very much analogous to the observation in the homologous residues in NR2B, T648A, A649C, V656A, and L657A mutations in NR1 also lead to constitutively open channels (Fig. 1B).
MK-801 can no longer be trapped in the constitutively open channels
It has been shown that externally applied MK-801, an open channel blocker of the NMDA receptor, can be “trapped” behind the closed activation gate after dissociation of agonists from the receptor (Huettner and Bean, 1988). In other words, external MK-801 can enter the open NMDA channel and bind to the “blocking site” located deep in the pore (probably very close to the selectivity filter) (Kashiwagi et al., 2002). Occupancy of this site, however, does not prevent channel closure, and the blocking molecule can be kept behind the closed activation gate for a long time. Conversely, subsequent openings of the activation gate would greatly facilitate the trapped blocker to leave the channel (Antonov et al., 1998; Sobolevsky et al., 1999; Qian and Johnson, 2002). More recently, it was also proposed that MK-801 and other trapping channel blockers probably bind close to and interact with the activation–deactivation gating machinery of the NMDA receptor (Yuan et al., 2005; Dravid et al., 2007). MK-801 therefore could be a very useful probe to detect whether the activation gate of the NMDA receptor is closed or open. We examined recovery from block by MK-801 in the wild-type and selected NR2B mutant receptors. Figure 2 shows that recovery from MK-801 block of the wild-type receptor in the absence of agonists is very slow, with a time constant of ∼20 min, consistent with the idea of trapped MK-801 in the closed channel (Huettner and Bean, 1988). In contrast, T647A mutation speeds the recovery course with shortening of the time constant to ∼60 s (Fig. 2C). This result indicates that MK-801 cannot be trapped in T647A mutant receptor after dissociation of agonists from the receptor. Moreover, similar findings can be obtained in the other constitutively open mutants (e.g., S645A, A648C, and I655A) but not in the nonconstitutively open mutants (e.g., A651C and A652C) (Fig. 2B). These results support that these constitutively open channels indeed have an open activation gate.
The inner but not the outer part of M3c (NR2B) shows state-dependent accessibility to external MTS reagents
Figure 3A shows agonist-elicited NMDA currents recorded in Xenopus oocytes expressing cysteine-substituted NR2B mutant (V640–Q656) receptors. Application of external 2-aminoethyl-methanethiosulfonate (MTSEA) typically reduces current amplitude [e.g., T647C and M654C (Fig. 3A); V640C, L643C, A648C, I655C and Q656C (data not shown)]. However, MTSEA increases rather than decreases the currents in two instances, namely A651C and A652C (Fig. 3A). For all cysteine-substituted mutant receptors, changes in the current amplitude with application of MTSEA could be well fitted with single-exponential functions. The apparent second-order rate constants (i.e., k value in the equation in Materials and Methods) for MTS modification in the presence of agonists (i.e., when the receptor is predominantly distributed into the “activated” states, which presumably include both the conducting open and the nonconducting desensitized conformations) are derived from the time constants from the fits and are 889, 379, 769, and 833 m−1s−1 for T647C, A651C, A652C, and M654C mutant receptors, respectively (Fig. 3A). We further measured the modification rate of cysteine-substituted NR2B mutant receptors in the absence of agonists (i.e., when the receptor is predominantly distributed into the closed state). Four typical examples of the apparent second-order rate constants derived from the time constants from the single-exponential fits to the courses of modification are 1.86, <0.83, 49, and 204 m−1s−1 for T647C, A651C, A652C, and M654C mutant receptors, respectively (Fig. 3B). Figure 3, C and D, summarizes the pattern of state-dependent reactivity of single cysteine substitution in M3c and M2 based on experiments similar to the preceding ones. The substituting cysteines at positions N615 and N616 in M2, and V640, L643, T647 and A651 in M3c are modified almost exclusively in the activated state, with at least an ∼500- to 1000-fold faster rate of modification in the activated than in the closed state. These results are consistent with the previous reports that N616C, V644C, T648C, and A652C of the NR1 subunit (homologous to N615, L643, T647, and A651 of NR2B, respectively) show significant state-dependent reaction kinetics with external MTSEA (Jones et al., 2002; Sobolevsky et al., 2002a; Yuan et al., 2005). Conversely, residues above T647–A651 (i.e., positions A652–Q656) show much smaller reactivity changes associated with channel gating (at most an approximately threefold to fivefold preference for the activated state). This overall pattern leads us to hypothesize that access of MTS reagents to the deeper region (N615–N616 in M2 and V640–T647 in M3c) of the pore is controlled by the activation gate of the channel, very much similar to the way how the access of traveling ions (e.g., Na+) to the pore is regulated. At positions M654–Q656 in NR2B, the access does not seem to be controlled by the activation gate. Instead, there is probably a gating-associated movement that results in small (approximately threefold to fivefold) changes in side-chain accessibility. Small changes in cysteine reactivity associated with gating have also been observed in the lower part of S6 of the Shaker K+ channel (Liu et al., 1997).
MTS modification rate shows exactly the same NMDA concentration dependence as channel activation
To confirm that the marked state dependence of modification rates was indeed ascribable to channel gating, we characterized the NMDA concentration dependence of 2-trimethylammonioethyl-methanethiosulfonate (MTSET) modification in detail in the L643C(NR2B) mutant receptor (i.e., L643 is presumably a pore-lining residue located below T647–A651). Figure 4 shows that the second-order rate constants obtained at different NMDA concentrations are well superimposable on the NMDA channel activation (opening probability) curve. It is clear that the modification rate rises steeply with the increasing NMDA concentration in parallel with channel gating (the NMDA channel activation curve) and reaches a plateau value at the NMDA concentration giving the maximum current. The perfect superposition of the two curves is reasonable considering the similar diameters of the smallest hydrated Na+ ion (∼7 Å) and MTSEA, MTSET, and MK-801 (∼3.6, ∼5.8, and ∼7.2 Å, respectively). The tight correlation between the gated access of external MTSET to the substituted cysteine and the gated Na+ permeation through the pore strongly argues that both processes are controlled by the same gate of the channel.
Modification of A651C(NR2B) with external MTS reagents holds the channel constitutively open
We have demonstrated that A651C(NR2B) mutant receptor shows remarkably state-dependent reaction kinetics with external MTS reagents (Fig. 3C,D). The effect of MTS modification on A651C(NR2B) mutant receptor, however, is quite different from the other tested residues. Modification of A651C(NR2B) with external MTS reagents happens essentially only in the presence of agonists and then would tend to hold the channel open. MTS modification therefore enhances the NMDA current. It is already evident in Figure 3A that, after MTESA modification, the channel allows a much larger inward current and remains open even after removal of agonists. Subsequent application of Mg2+ (an open channel blocker) (Fig. 3A) but not APV (a competitive NMDA site antagonist; data not shown) blocks this constantly open channel. Moreover, modification of homologous A652C(NR1) by external MTSET shows essentially the same findings (data not shown) (Jones et al., 2002; Yuan et al., 2005). Because of the structural similarity between MTS-modified cysteine and arginine (Fig. 5A), we examined arginine substitutions in A651(NR2B) and homologous A652(NR1) to see whether they also make constitutively open channels. Figure 5B shows that there are very similar current amplitude and activation kinetics between the Na+-elicited current and agonist-elicited current (or more precisely “agonist- and Na+-elicited current” because, in this case, the external solution was switched from agonist-free NMG–ND96 to agonist-containing Na+–ND96) in A651R(NR2B) mutation. The constitutive open index is 96 ± 2.2 and 71 ± 1.7% for A651R(NR2B) and A652R(NR1) mutations, respectively (both n = 10) (Table 1). The fact that the presence of agonists (agonist-elicited current) or not (Na+-elicited current) does not cause a discernible difference in the elicitation of the NMDA currents is consistent with the foregoing view that A651C(NR2B)+MTS already holds the channel constitutively open. Similar to the wild-type channel, the Na+-elicited currents of these two constitutively open channels are readily blocked by Mg2+ and MK-801. Most interestingly, N615C(NR2B), a residue presumably located deep in the pore (i.e., at the tip of M2 loop) and responsible for Mg2+ as well as Ca2+ binding, can be modified by external MTSET in the A651R+N615C(NR2B) double mutation in the complete absence of agonists (Fig. 5C). Moreover, the modification kinetics of MTSET are very similar in single-mutant N615C(NR2B) in the presence of agonists (the N615C mutant receptor is not constitutively open and cannot be modified by MTSET in the absence of agonists) and double-mutant A651R+N615C(NR2B) in the absence of any agonists (k = 298 ± 21 and 267 ± 18 m−1s−1, respectively; p > 0.05; n = 8) (Fig. 5C). These results strongly suggest that A651R(NR2B) mutation leads to a constitutively open channel with an activation gate conformation very similar to that of the open wild-type channel (and also with essentially unaltered functional properties of the pore concerning ion permeation or Mg2+ and MK-801 block).
Double-mutant cycle analysis shows cooperative interaction between homologous residues A652(NR1) and A651(NR2B)
Because A652(NR1) and A651(NR2B) seem to play a unique role in NMDA channel gating, we further investigated the possible interactions between these two homologous residues to elucidate the molecular mechanisms underlying the control of the activation gate. If the difference in free energy (ΔG) between closed and activated NMDA receptors is simplistically calculated according to the Boltzmann equation as follows: ΔG = RT ln(A/C), where R is the gas constant, T is the absolute temperature, and A/C is the ratio between the steady-state occupancy of the activated states (open plus desensitized states) and the closed state of the NMDA receptor. Assuming a constant ratio between the occupancies of the open and desensitized states and nearly complete occupancy of the receptor in the activated states (i.e., scarcely any receptor in the resting state) when 100 μm NMDA/10 μm glycine is present, we may readily obtain A/C from the ratio between constitutive open index and (1 − constitutive open index) for each different receptor. The change in ΔG caused by each mutation (ΔΔG) could be calculated as follows: ΔΔG = ΔGmut − ΔGwt. To differentiate whether two mutations interact independently or cooperatively in channel gating, we used double-mutant cycle analysis (Hidalgo and MacKinnon, 1995) to calculate the coupling energy (ΔΔGint) for each pair of mutations. If the two residues are coupled in energy terms, there will be a coupling energy ΔΔGint given by ΔΔGint = ΔΔGA12 − (ΔΔGA1 + ΔΔGA2), where A1 and A2 are single mutations, and A12 is the double mutation A1 + A2. Table 1 shows that different charged single mutations (e.g., A to R, E, K, or D) of these two homologous residues, A652(NR1) and A651(NR2B), lead to channels with constitutive open indexes of ∼15–96%. Most interestingly, A652R(NR1)+A651E(NR2B) and A652E(NR1)+A651R(NR2B) have a much smaller constitutive open index than single mutations A652R(NR1) and A651R(NR2B), suggesting a cooperative (non-additive) action between these two homologous residues. Moreover, the significant negative coupling energies (ΔΔGint = −2.36 to −3.37 kcal/mol) (Fig. 6A) further indicate that there is a significant direct interaction between these two homologous compensatory-charge mutant residues in NR1 and NR2B. These findings strongly argue that A652(NR1) and A651(NR2B) may play a critical role in setting the configuration of the activation gate. We therefore also explored the steric or voluminal effect of the residues at this critical position. Single and double mutations by substituting A652(NR1) and A651(NR2B) with residues of “medium” side chains (e.g., cysteine or threonine, in contrast with the “small”-size alanine in the wild-type channel) cannot make constitutively open channels (Fig. 6B, Table 1). Although single mutations to “large” side chains (e.g., leucine or phenylalanine) in A652(NR1) and A651(NR2B) also do not lead to constitutively open channels, double mutations [A652L(NR1)+A651L(NR2B) and A652F(NR1)+A651F(NR2B)] significantly increase the constitutive open index to ∼46–56%. Moreover, single mutations by substituting tryptophan (which has the “largest” side chain) to A652(NR1) and A651(NR2B) result in a constitutive open index of 10 and 30%, respectively, whereas the homologous double mutation [A652W(NR1)+A651W(NR2B)] leads to a nearly fully constitutively open channel with an index up to ∼97% (Fig. 6B). The significant positive coupling energies (ΔΔGint = 1.57–3.34 kcal/mol) (Table 1) of these double mutations also indicate a strong cooperative interaction between A652(NR1) and A651(NR2B), well consistent with the proximity between A652(NR1) and A651(NR2B) demonstrated by the evident cooperative effect in the compensatory-charge double mutations. These findings altogether suggest an essential role of intersubunit configuration at the level of A652(NR1)–A651(NR2B) in NMDA receptor activation–deactivation (see Discussion).
The activation gate of the NMDA receptor is constituted by M3c and is located in the external vestibule of the pore
We show that different single point mutations between S645 and I655 of M3c in NR2B (and also homologous residues in NR1) lead to constitutive opening of the NMDA receptor (Fig. 1). The currents through these “constitutively open channels” are readily blocked by external Mg2+ and MK-801. Also, MK-801 can no longer be trapped in these constitutively open channels (Fig. 2), consistent with the ideas that the foregoing residues are located in the external vestibule of the pore and are closely related to the activation gating control of the channel (see Introduction). In this regard, it is interesting to note that felbamate, the only approved anticonvulsant with significant gating-modification effects on NMDA receptors (Kuo et al., 2004), also binds to this region and acts as an “opportunistic” pore blocker with the “entrance” of its binding site undergoing a large gating conformational change (Chang and Kuo, 2007a,b, 2008). Moreover, MTSEA and MTSET modification of M3c (NR2B) is ∼500- to 1000-fold faster in the activated than in the closed channels for the positions below A648–A651 (Fig. 3C,D) (Jones et al., 2002; Sobolevsky et al., 2002a,b, 2007) (but see Beck et al., 1999). In contrast, the positions above A648–A651 (e.g., M654 and I655) are readily modified in both activated and closed channels. Most importantly, Figure 4 demonstrates that the NMDA concentration dependence of MTSET modification rate at L643C, a residue below A648–A651, is exactly superimposable on the NMDA concentration dependence of channel activation (opening probability of the channel). These findings indicate that the access of external MTS reagents and the permeating Na+ to the channel pore is controlled by the same activation gate.
A651(NR2B) and A652(NR1) of M3c constitute the critical part of the activation gate and ensure the essential intersubunit configuration for normal gating
The significant negative coupling energies (ΔΔGint = −2.36 to −3.37 kcal/mol) (Fig. 6A) of the compensatory-charge double mutations of A652(NR1) and A651(NR2B) by thermodynamic cycle analysis (Hidalgo and MacKinnon, 1995) indicate significant interactions between these two homologous residues in NR1 and NR2B. We have demonstrated that residues L643 and T647 of M3c in NR2B (and homologous V644 and T648 in NR1) constitute the binding ligands for felbamate (Chang and Kuo, 2008), suggesting that L643 and T647 approximately face the same side on the α-helical M3c domain (Beck et al., 1999; Kashiwagi et al., 2002) and are both responsible for pore lining. The significant interactions between A652(NR1) and A651(NR2B) in compensatory-charge double mutations would then strongly argue that these two residues still line the pore and thus should be located close to the interphase of two different subunits if L643 and T647 in the same α-helix are also in the pore. Consistently, single and double mutations at A652(NR1) and A651 (NR2B) demonstrate evident steric or voluminal effect on constitutive activation of the channel (Fig. 6B). Although single mutations to large side chains in A652(NR1) and A651 (NR2B) (both being A7 in the SYTANLAAF motif), such as A7L or A7F, do not lead to constitutively open channels, double mutations of A7 [A652L(NR1)+A651L(NR2B) and A652F(NR1)+A651F(NR2B)] result in a substantial constitutive open index. Moreover, modification of A7 by external MTS reagents can happen only in the activated but not the resting channel and then markedly enhances NMDA currents as well as holds the channel constitutively open (Fig. 3A). These findings indicate that proper interactions between A7 and its local environment are critical for the activation–deactivation gating and that the distance between each two subunits at the A7 level is a critical determinant for the status of the NMDA receptor activation gate. The presumably very limited space to accommodate these critical residues in the resting channel, for example, may explain why alanines are selected and remain so highly conserved for this position (Fig. 7A). An approximately similar pattern of state-dependent modification of the SYTANLAAF motif has also been demonstrated for NR1 and NR2C (Sobolevsky et al., 2002a, 2007), and it seems that MTS modification of A7C in any subunit always lead to markedly enhanced constitutive opening of the channel (Beck et al., 1999; Jones et al., 2002; Yuan et al., 2005; Sobolevsky et al., 2007). However, the A7C mutations in NR1 and NR2A behave differently in response to partial glutamate site agonists (Jones et al., 2002). Also, there is probably a staggering alignment of M3c by approximately four residues between subunits NR1 and NR2C. Moreover, the NMDA currents are markedly enhanced by MTS modification of only A7C and A8C in NR1, but a similar effect is observed for A3C, N4C, A7C, and A8C in NR2C, with more M3c residues in NR1 facing the central pore than in NR2C (Sobolevsky et al., 2002a,b, 2007). Thus, there may well be a different alignment of M3c between NR1 and NR2A or NR2C, reminiscent of the findings that the tip of M2 is asymmetrically positioned between NR1 and NR2C (Kuner et al., 1996), although the M3c in NR2B seems to be well aligned to its homologous residues in NR1 in our case (Fig. 6) (Chang and Kuo, 2008). At any rate, it seems that the intersubunit distance at the level of A7 is well preserved as a common determinant of channel activation for all different subunits. It would be interesting to further characterize the differences in the local placement of M3c and related functional consequences for different subunits.
A652(NR1) and A651(NR2B) may mark the bundle-crossing point of the four M3c helices
The pore region of the NMDA receptor is probably a minimal K+ channel, which is “inserted” into a bacterial periplasmic protein in an “inverted” way (Kuner et al., 1996, 2003; Panchenko et al., 2001; Mayer and Armstrong, 2004). Identification of a prokaryotic ionotropic glutamate receptor with a K+-selective pore (GluR0) (Chen et al., 1999) confirms this idea and strengthens the view that ionotropic glutamate receptors are structurally related to K+ channels. Sequence alignment of the M3 segment of the NMDA or GluR0 receptors with the M2 (inner) segment of KcsA or the S6 segment of Shaker K+ channels (Fig. 7A) shows that A7 (A652 in NR1 and A651 in NR2B) in the NMDA receptor may correspond to A111 in KcsA and V478 in Shaker, the “bundle-crossing” points of those channels (Doyle et al., 1998; Hackos et al., 2002; Kitaguchi et al., 2004). This is consistent with our previous analysis that the intersubunit distance at the A7 level is a common determinant of NMDA channel activation. A7 thus probably marks the narrowest (or the bundle-crossing) region of the external vestibule in the closed channel (Fig. 7B,C). The other M3c mutations leading to constitutively open channels in the alanine scan (Fig. 1) presumably directly or allosterically change the intersubunit distance at this A7 level. This is reminiscent of the constitutive activation of the Shaker K+ channel caused by mutations within the S6 activation gate region (Sukhareva et al., 2003). The ligand-binding core of the NMDA receptor consists of two domains formed by the N-terminal extracellular domain (S1) and the extracellular loop between M3 and M4 (S2) (Kuryatov et al., 1994; Laube et al., 1997; Anson et al., 1998; Furukawa and Gouaux, 2003; Chen et al., 2005; Furukawa et al., 2005). Because the conserved SYTANLAAF motif of the NMDA receptor is located at the end of M3c, which is directly connected to the S2 domain [the Helix F (Furukawa and Gouaux, 2003)] by a brief (∼30 amino acids) post-M3 segment, via which the en bloc movement caused by ligand binding may be readily transduced. The diameters at the levels V644(NR1)–L643(NR2B) and T648(NR1)–T647(NR2B) of M3c in the open channel are estimated to be ∼18 and ∼20 Å, respectively (from α-carbon to α-carbon in the main chain) (Chang and Kuo, 2008). With an α-helical secondary structure, the diameter at the level of A652(NR1)–A651(NR2B) then should be ∼22 Å (α-carbon to α-carbon) or ∼17 Å (side chain tip to tip) in the open channel. If the diameter at this level should be <7 Å (the smallest diameter of an hydrated Na+ ion) in the closed channel, there would be a >10 Å gating conformational change at the activation gate of the NMDA receptor, a figure slightly larger than that in K+ channels (e.g., KcsA, MthK, and Kv1.2), in which diameters of ∼4–5 and ∼12 Å at the intracellular end were reported for the closed and open channels, respectively (Doyle et al., 1998; Jiang et al., 2002a,b; Long et al., 2005a,b).
Effective opening of the activation gate requires movement of all four M3c helices
The opening probability of NMDA channel is greatly increased only when all of the four agonist sites are occupied by appropriate agonists (e.g., two glutamates and two glycines), indicating that effective opening of the activation gate requires all of four M3c helices. Consistently, we have seen that single mutations of A7 into residues with large hydrophobic side chains (leucine or phenylalanine) lead to just negligible, and even single mutations into tryptophan (which has the largest side chain) lead to only small, constitutive channel opening (Fig. 7D). In contrast, when all four M3c helices contain bulky substituting side chains at A7, there is a disproportionately enhanced constitutive opening in these double-mutant NMDA receptors. The arrangement and motion of the four M3c helices at the A7 activation gate must be so designed that movement of only two subunits would have a disproportionately smaller effect than movement of all four subunits, and only when all M3c helices tend to decline a close apposition with their neighbors at the A7 level could the channel fully open, regardless of whether the movement is achieved by binding of glutamate and glycine or by steric hindrance and/or other repulsive forces between different M3c helices.
This work was supported by Grant NSC 96-2320-B-002-009 from the National Science Council and Grant NHRI-EX96-9606NI from the National Health Research Institutes, Taiwan. H.-R.C. was a recipient of the MD–PhD Predoctoral Fellowship from the National Health Research Institutes, Taiwan. We thank Drs. S. Nakanishi and K. Williams for sharing cDNAs encoding NR1a and NR2B subunits.
- Correspondence should be addressed to Chung-Chin Kuo, Department of Physiology, National Taiwan University College of Medicine, 1, Jen-Ai Road, 1st Section, Taipei 100, Taiwan.