Ethanol inhibition of N-methyl-D-aspartate receptors is reduced by site-directed mutagenesis of a transmembrane domain phenylalanine residue.

N-Methyl-D-aspartate (NMDA) receptors (NRs) are ionotropic receptors activated by glutamate and the co-agonist glycine. Ethanol inhibits NMDA receptor function, although its site of action is undefined. We hypothesized that ethanol acts at specific amino acids contained within the transmembrane (TM) domains of the receptor. In this study, NR1 and NR2A subunits were altered by mutagenesis and tested for sensitivity to ethanol. Three NR1 mutants (W636A, F817A, and L819A) and one NR2A mutant (F637A) failed to generate functional receptors. Pre-TM1 (I546A, L551A, F554A, and F558A), TM1 (W563A), and TM2 (W611A) NR1 mutations did not affect ethanol sensitivity of heteromeric receptors. In contrast, altering a TM3 phenylalanine to alanine (F639A) reduced the ethanol inhibition of NMDA receptors expressed in oocytes and human embryonic kidney 293 cells. Mutation of the nearby methionine (M641) to alanine did not affect ethanol sensitivity, whereas changing Phe(639) to tryptophan slightly enhanced ethanol inhibition. NR1(F639A) did not alter the agonist potency of glutamate but did produce a leftward shift in the glycine concentration response for receptors containing NR2A and NR2B subunits. NR1(F639A) also reduced the potency of the competitive glycine antagonist 5,7-dichlorokynurenic acid and increased the efficacy of the glycine partial agonist 3-amino-1-hydroxy-2-pyrrolidinone ((+)-HA-966). These results suggest that ethanol may interact with amino acids contained in the TM3 domain of NMDA subunits that are involved in transducing agonist binding to channel opening.

N-Methyl-D-aspartate (NMDA) receptors (NRs) are ionotropic receptors activated by glutamate and the coagonist glycine. Ethanol inhibits NMDA receptor function, although its site of action is undefined. We hypothesized that ethanol acts at specific amino acids contained within the transmembrane (TM) domains of the receptor. In this study, NR1 and NR2A subunits were altered by mutagenesis and tested for sensitivity to ethanol. Three NR1 mutants (W636A, F817A, and L819A) and one NR2A mutant (F637A) failed to generate functional receptors. Pre-TM1 (I546A, L551A, F554A, and F558A), TM1 (W563A), and TM2 (W611A) NR1 mutations did not affect ethanol sensitivity of heteromeric receptors. In contrast, altering a TM3 phenylalanine to alanine (F639A) reduced the ethanol inhibition of NMDA receptors expressed in oocytes and human embryonic kidney 293 cells. Mutation of the nearby methionine (M641) to alanine did not affect ethanol sensitivity, whereas changing Phe 639 to tryptophan slightly enhanced ethanol inhibition. NR1(F639A) did not alter the agonist potency of glutamate but did produce a leftward shift in the glycine concentration response for receptors containing NR2A and NR2B subunits. NR1(F639A) also reduced the potency of the competitive glycine antagonist 5,7-dichlorokynurenic acid and increased the efficacy of the glycine partial agonist 3-amino-1-hydroxy-2pyrrolidinone ((؉)-HA-966). These results suggest that ethanol may interact with amino acids contained in the TM3 domain of NMDA subunits that are involved in transducing agonist binding to channel opening.
N-Methyl-D-aspartate (NMDA) 1 receptors are calcium-permeable ion channels expressed by neurons and require both glutamate and glycine for activation. Combinations of NMDA receptor 1 (NR1) and NR2 subunits yield receptors with different biophysical and pharmacological properties such as differences in desensitization and sensitivity to agonists and antagonists. NMDA receptors play an important role in neuronal development and are required for some forms of synaptic plas-ticity such as associative long-term potentiation that may underlie some forms of learning and memory (1). NMDA receptors are also involved in the excitotoxic effects of glutamate that accompany traumatic brain injury and stroke-induced ischemia.
Ethanol inhibits native NMDA receptor function in vitro and in vivo (2)(3)(4)(5)(6)(7). Chronic exposure of neurons to ethanol results in up-regulation of NMDA receptor function and enhanced glutamate-mediated excitotoxicity (8 -10). NMDA antagonists block the seizures associated with ethanol withdrawal (11,12), and human alcoholics report ethanol-like subjective effects after administration of ketamine, a nondissociative anesthetic that inhibits NMDA channel function (13). Despite the wealth of knowledge indicating that the NMDA receptor is an important target for ethanol in the brain, there is no consensus as to how ethanol inhibits receptor function. Ethanol behaves as a noncompetitive and voltage-independent antagonist of the receptor, and attempts to correlate its inhibitory actions with any of the known modulatory sites on the receptor have been largely negative (14 -16). In single-channel studies, the inhibitory effects of ethanol were best accounted for by decreases in the mean open time and frequency of channel opening, effects consistent with an allosteric reduction in agonist-induced channel gating (17).
Studies with recombinant NMDA receptors have shown that receptors containing NR1/2A or NR1/2B subunits are generally more sensitive to ethanol inhibition than NR1/2C or NR1/2D receptors (16,18,19). In addition, ethanol inhibition of NMDAinduced currents in oocytes expressing NR1, NR2A, and NR2C subunits was less than that observed with NR1 and NR2A receptors, suggesting that subunit composition significantly influences overall ethanol sensitivity (19). Recent studies from this laboratory have also shown that ethanol inhibition of NR1/2A receptors expressed in human embryonic kidney 293 (HEK293) cells is reduced by Fyn tyrosine kinase-mediated phosphorylation of the NR2A subunit as well as by conditions that block calcium-dependent inactivation of NR1/2A receptors (20,21). However, these manipulations only partially reduce ethanol inhibition of receptor function and C-terminal truncated NMDA subunits retain substantial sensitivity to inhibition by ethanol (21)(22)(23). Overall, these data suggest that although C-terminal modifications may influence the ethanol sensitivity of the NMDA receptor, it is unlikely that these intracellular domains represent the major site of action for ethanol.
Results from recent mutagenesis studies with alcohol-and anesthetic-sensitive ␥-aminobutyric acid A (GABA A ) and glycine receptors have shown that mutation of a serine residue in the second transmembrane (TM) domain or an alanine residue in the TM3 greatly affected the potentiation of GABA A and glycine channel function by ethanol and volatile anesthetics (24). The magnitude of this effect was correlated with the molecular volume of the substituted amino acid, with larger amino acids producing inhibition by ethanol and volatile anesthetics and smaller amino acids producing enhanced potentiation. These results suggested that specific amino acids in these subunits may define an alcohol-and anesthetic-sensitive site (25). Although the sequence identity and structural homology between GABA A and glycine receptors and glutamate receptors is extremely low, we hypothesized that NMDA receptors also possess an ethanol-sensitive site that is defined by specific amino acids contained in one or more of the TM domains of the receptor. We reasoned that these amino acids would be relatively large and conserved between various NMDA subunits and not face the pore of the ion channel.
In this study, a series of amino acids fitting these criteria were altered by site-directed mutagenesis, and the resulting mutant receptors were tested for their ethanol sensitivity. The results demonstrate that substitution of a single phenylalanine residue in the TM3 domain markedly reduces ethanol inhibition of NMDA receptor currents and that this effect is influenced by amino acid volume. A preliminary report of these findings has been presented in abstract form (26).

EXPERIMENTAL PROCEDURES
Molecular Biology and Mutagenesis-The rat NR1 and NR2 cDNA clones were kindly provided by Dr. S. Nakanishi, Dr. P. H. Seeburg, and Dr. D. Lynch. These cDNAs were subcloned into cytomegalovirus-containing vectors (pcDNA3, Invitrogen; or pGFP-N3, CLONTECH) as needed. The QuickChange site-directed mutagenesis kit by Stratagene (La Jolla, CA) was used to make point mutations, and the resulting cDNA plasmids were sequenced by automated dideoxy sequencing to verify the mutation. For mRNA synthesis, linearized cDNA was used as the template in an in vitro transcription reaction using the mMessage mMachine kit (Ambion, Inc., Austin, TX).
Xenopus laevis Oocyte Preparation and Injection-Adult oocyte-positive X. laevis frogs (Xenopus One, Ann Arbor, MI) were housed in distilled, dechlorinated water (18 -20°C) with a 12-h/12-h light/dark cycle and fed twice weekly. Frogs were anesthetized with 0.25% ethyl m-aminobenzoate 222 (MS 222, Sigma), and a portion of the oocytes was removed after surgery. Oocytes were incubated in 1 mg/ml collagenase type 1A (Sigma) for 60 -90 min. NMDA receptor mRNAs synthesized in vitro were injected into oocytes at a 1:1 ratio using a variable Nanoject injector (Drummond Scientific Co., Broomall, PA). Oocytes were incubated for 2-7 days in 0.5ϫ Leibovitz L-15 medium (Sigma) supplemented by antibiotics.
Cell Culture and Transfections-HEK293 cells (ATCC, Manassas, VA) were maintained as previously described and transfected with various NMDA receptor subunits using the LipofectAMINE (Life Technologies, Inc.) reagent (21). A red-shifted variant (S65T) of the green fluorescent protein (GFP-N3; CLONTECH) previously generated in this laboratory was used to allow for identification of transfected cells under epifluorescent conditions (27). After transfection, cells were maintained in minimal essential medium with 200 M ketamine for 24 -48 h to prevent excitotoxicity (28). Ketamine was removed by extensive washing before electrophysiological recordings.
Two-electrode Voltage Clamp Analysis-Oocytes were placed in a perfusion chamber (total volume, ϳ40 l) and perfused with bariumcontaining extracellular solution (115 mM NaCl, 2.5 mM KCl, 1.8 mM BaCl 2 , and 10 mM Hepes, pH 7.2). In most experiments, Na 2 EDTA (10 M) was also included in the extracellular solutions to eliminate residual zinc found in experimental solutions. Oocytes were impaled with two microelectrodes filled with 3 M KCl (0.5-3 m⍀), and the resting membrane potential was held at Ϫ80 mV using an Axon GeneClamp amplifier (Axon Instruments Inc., Foster City, CA). NMDA receptors were activated by switching to a solution containing 100 M NMDA (or glutamate as noted) and 10 M glycine. Oocytes were then exposed to a solution containing the desired concentration of ethanol (10 -200 mM) for ϳ10 -30 s followed by a solution containing ethanol, NMDA, and glycine. The ethanol was purchased as 100% from AAper Alcohol and Chemical Co. (Shelbyville, KY) and was diluted with the extracellular recording solution to the final desired concentration. After washout, a second control recording was obtained, and these control currents were averaged. In each oocyte tested, ethanol inhibition was calculated using the formula 100% Ϫ (I etoh /I control ), where I control was the average of currents obtained before and after the ethanol challenge. Unless otherwise noted, current amplitudes were measured at the end of the agonist pulse. Currents were filtered at 10 Hz, digitized with a 16-bit analog-to-digital interface (Instrutech, Port Washington, NY) and analyzed on a Macintosh computer using Pulse Control voltage clamp software (Richard Bookman, University of Miami) running under the IGOR-Pro graphics environment (Wavemetrics, Lake Oswego, OR). Agonist and antagonist concentration-response curves were determined by normalizing current amplitude at each concentration of agonist or antagonist to the amplitude of a maximal concentration of agonist for each oocyte. Similarly, the effects of the partial agonist (ϩ)-HA-966 (Sigma) on NMDA-stimulated currents were expressed as percent stimulation produced in each oocyte by 100 M NMDA and 10 M glycine.

TM3 Phenylalanine and Ethanol
binding site and to prevent glycine-dependent desensitization. Na 2 EDTA (0.01 mM) was added to all external solutions to eliminate residual zinc found in experimental solutions. HEK293 cells showing green fluorescent protein fluorescence were selected for recording and were held at Ϫ50 mV. Series resistance was routinely compensated by 70 -80%. A multibarrel fast perfusion system (SF77A; Warner Instruments, Hamden, CT) with a switching time of ϳ8 ms was used to perfuse the cells with extracellular control solution. NMDA receptormediated currents were activated by switching from the control solution to one containing glutamate (200 M) and glycine (50 M). Currents were filtered at 5 kHz, low pass-filtered at 0.2 kHz, digitized using an Instrutech analog-to-digital interface, and collected on a Macintosh computer running the Pulse Control voltage clamp software under the Igor Pro graphics platform (Wavemetrics).
Data Collection and Analysis-Data were analyzed for statistical significance (p Ͻ 0.05) using either Student's t test or analysis of variance followed by Tukey's posthoc test where appropriate. Regression analysis for calculating EC 50 values was performed using Graph-Pad Prism software. Results for the experiments are given as mean Ϯ S. E. unless indicated otherwise. Fig. 1 shows the sequences of the TM domains of the NR1 and NR2 NMDA receptor subunits. Asterisks indicate residues shown in the NR1 subunit that have been previously assigned as pore-facing by cysteine scanning mutagenesis (29,30). Residues in the NR1 subunit shown in boldface were mutated to alanine in the present study and tested for ethanol sensitivity. Most of these mutant receptor subtypes gave rise to large NMDA-mediated currents when expressed in Xenopus oocytes (Table I). However, some of the NR1 mutants (W636A, F817A, and L819A) and NR2A mutants (F637A) tested did not yield reproducible currents and were not further investigated. None of the mutants tested appeared to be tonically active in the absence of agonist based on analysis of holding currents at Ϫ80 mV. Ethanol inhibition of wild-type NMDA receptors was determined the same day as that of mutant receptors to control for any seasonal variability. These control data were pooled where appropriate to give an average value of ethanol inhibition and are consistent with previously published results using oocytes (19,22).

RESULTS
Mutations made at four different sites within the pre-TM1 domain of the NR1 subunit resulted in functional and ethanolsensitive NMDA receptors when co-expressed with the wildtype NR2A subunit. Receptors composed of NR1 I546A, L551A, F554A, or F558A plus the NR2A subunit were all inhibited to the same extent by 25-100 mM ethanol (Fig. 2A). The range of ethanol sensitivity among these mutants was not different from that determined for wild-type NR1/2A receptors. Similarly, selected mutations made in either the TM1 (W563A) or TM2 (W611A) domain of the NR1 receptor resulted in functional NMDA receptors that showed normal sensitivity to ethanol (Fig. 3, A and B).

TM3 Phenylalanine and Ethanol
In contrast, expression of the TM3 mutant NR1(F639A) with the NR2A subunit yielded receptors that were significantly less sensitive to ethanol than wild type (Fig. 4A). This effect was manifested as a rightward and downward shift in the ethanol dose-response curve and persisted over ethanol concentrations from 10 to 200 mM (Fig. 4B). Concentrations of ethanol Ͼ200 mM produced unstable responses in voltage-clamped oocytes and were not tested. Mutation of the nearby methionine (Met 641 ) to alanine did not alter the inhibitory effect of ethanol compared with wild-type receptors (Fig. 4C). In addition, mutation of NR1(F639A) to the larger tryptophan residue (F639W) produced receptors that were slightly more sensitive to ethanol than wild-type receptors (Fig. 4D).
To determine whether the effects of the F639A mutation on ethanol sensitivity were NR2 subunit-dependent, NR1(F639A) was co-expressed with either NR2B or NR2C subunits. The ethanol sensitivity of both NR1(F639A)/2B and NR1 (F639A)/2C receptors was also significantly less than that determined for their respective wild-type counterparts (Fig. 5, A  and B). Expression of NR1(F639A) with NR2A, NR2B, or NR2C subunits in HEK293 cells also significantly reduced the inhibitory effects of 100 mM ethanol compared with wild-type receptors (Fig. 5C).
Current-voltage experiments revealed no differences in the reversal potential or slope conductance between NR1 (F639A)/2A receptors and wild-type receptors (data not shown). In addition, expression of NR1(F639A) with the NR2A subunit did not significantly alter the ability of the physiological agonist glutamate to activate the receptor (Fig. 6A). However, the F639A substitution in the NR1 subunit shifted the concentration response for glycine to the left of that of the wild-type receptor (Fig. 6B). Calculated EC 50 values for the wild-type (NR1/2A) and mutant (NR1(F639A)/2A) receptors were 0.94 M (Hill slope, 1.58) and 0.38 M (Hill slope, 1.40), respectively. This effect of the F639A mutation on glycine potency was even more pronounced in receptors co-expressing NR2B subunits (Fig. 6C). The glycine EC 50 value for wild-type NR1/2B receptors was 0.18 M (Hill slope, 2.00). When NR1(F639A) was co-expressed with the NR2B subunit, significant receptor activation was observed even in the absence of added glycine. This effect prevented an accurate calculation of the EC 50 value for this subunit combination. The activation of NR1(F639A)/2B receptors in solutions lacking added glycine was blocked by the glycine site antagonist 5,7-dichlorokynurenic acid (5,7-DCK; data not shown). The glycine sensitivity of receptors expressing NR1(F639A) and NR2C subunits (EC 50 , 0.18 M; Hill slope, 1.44) was not significantly different from wild-type NR1/2C receptors (EC 50 , 0.31 M; Hill slope, 1.96; Fig. 6D).
To investigate possible mechanisms underlying the shift in apparent glycine sensitivity with the mutant receptor, the sensitivity of wild-type and mutant receptors to a competitive glycine antagonist and a glycine partial agonist were determined. At a fixed glycine concentration of 10 M, the competitive antagonist, 5,7-DCK dose-dependently inhibited NMDAstimulated currents from oocytes expressing wild-type NR1/2A receptors with an IC 50 value of 0.68 M (Fig. 7A). Expression of NR1(F639A)/2A receptors shifted the concentration-response curve for 5,7-DCK to the right and increased the IC 50 value to 2.37 M.
The effect of the F639A mutant on glycine efficacy was examined by using (ϩ)-HA-966, a high-affinity, low-efficacy agonist at the glycine site. Oocytes expressing either wild-type NR1/2A or NR1(F639A)/2A receptors were stimulated with NMDA (100 M) and increasing concentrations of (ϩ)-HA-966 in the absence of any added glycine. The currents obtained in the presence of each concentration of (ϩ)-HA-966 were normalized to the current produced in each oocyte by a maximum concentration of NMDA and glycine. In the absence of any added glycine, NMDA application resulted in currents from both wild-type and NR1(F639A)/2A receptors that were ϳ5-10% of the response obtained in the presence of saturating concentrations of NMDA and glycine (Fig. 7B). Addition of (ϩ)-HA-966 up to 300 M to NMDA-containing solutions lacking added glycine did not significantly increase the amplitude of currents from wild-type receptors. In contrast, (ϩ)-HA-966 dose-dependently increased the amplitude of NMDA-stimulated currents in oocytes expressing NR1(F639A)/2A subunits, reaching a maximum of ϳ30% at 300 M.

DISCUSSION
The major goal of this study was to test the hypothesis that the inhibition of NMDA receptor currents by ethanol is mediated via an interaction with one or more amino acids contained within transmembrane domains of the receptor. Because eth-anol inhibition of NMDA receptor currents does not resemble that of channel-blocking drugs such as MK801 or ketamine, we initially selected amino acids that were not thought to be porefacing (29). Subsequent cysteine scanning studies revealed that some of the pre-TM1 residues tested (Phe 554 and Phe 558 ) were accessible to sulfhydryl-modifying agents, suggesting that the pre-TMI domain contains amino acids that contribute to the outer vestibule of the channel (30). The pre-TM1 domain of NR2 subunits also contains amino acids involved in regulating part of the glycine-independent desensitization of the NR2Aand NR2B-containing receptors (31). In the present study, none of the NR1 subunit pre-TMI mutants tested modified the NMDA receptor sensitivity to ethanol inhibition. Although it is possible that changes in ethanol sensitivity may have been seen by modifying NR2 residues specifically involved in regulating glycine-independent desensitization, nondesensitizing receptor subtypes (e.g. NR1/2C) are still significantly inhibited by ethanol, suggesting that these amino acids are not likely sites of ethanol interaction.
Mutation of Phe 639 to alanine in the TM3 domain of the NR1 subunit significantly decreased the ethanol inhibition of NMDA receptors expressed in either oocytes or HEK293 cells. Substitution of the slightly larger tryptophan residue at Phe 639 resulted in receptors that were slightly more sensitive to ethanol inhibition than wild-type receptors, suggesting that some physical or chemical property of the amino acid substitution at this position may be an important determinant of ethanol sensitivity.
This has been more carefully studied in mutagenesis experiments on GABA A and glycine receptors. In those receptors, residues in TM2 and TM3 have been shown to influence the degree of potentiation of receptor function by ethanol and vol- atile anesthetics such as isoflurane (24,25). Replacement of the conserved TM2 serine with amino acids of different volumes resulted in receptors that were either potentiated (small amino acids), inhibited (large amino acids), or insensitive to these compounds (25,32). Whether these amino acids are involved in defining a binding site or pocket for ethanol or whether they indirectly affect ethanol sensitivity by altering channel gating is unknown. However, replacement of the TM2 serine in GABA A and glycine receptors with amino acids that reduce ethanol potentiation are often associated with a significant leftward shift in the agonist dose-response curves for these receptors. This mutation-induced leftward shift in agonist potency often reduces further potentiation by alcohols and inhaled volatile anesthetics.
In the present study, the F639A mutation caused a significant leftward shift in the potency of the receptor for glycine, with less effect on that for glutamate. This was NR2 subunitdependent and was especially marked for receptors containing the NR2B subunit. The reduction in glycine but not glutamate potency is consistent with previous studies suggesting that NR1 subunits contribute the glycine binding site of the receptor, whereas NR2 subunits provide the glutamate binding site. The leftward shift in the concentration-response curve for glycine may have resulted from a change in the affinity of the NR1 subunit for glycine or in an increase in the efficacy of glycine as a receptor agonist. Results obtained with the glycine site com-petitive antagonist 5,7-DCK were not conclusive, because the reduced sensitivity of F639A mutant receptors to this antagonist may have resulted from an increase in the binding affinity for glycine. However, for several reasons, it seems unlikely that the effect of the F639A mutation on glycine potency is attributable to a change in glycine affinity. First, amino acids that regulate glycine binding to the NMDA receptor are located in the extracellular S1 and S2 lobes of the NR1 subunit. Most of these map to homologous positions shown by x-ray crystallography to define the glutamate binding site of the GluR2 subunit (33). Second, in the present study, oocytes expressing NR1(F639A)/2A receptors showed significant currents when stimulated with NMDA and the high-affinity partial agonist (ϩ)HA-966. These results suggest that Phe 639 may be part of an activation domain that couples receptor binding to channel opening. It is likely that other domains are important in defining this site, because the magnitude of the effect of the F639A mutation on glycine potency was influenced by the NR2 subunit expressed. For example, F639A did not significantly alter the glycine potency for NR2C-containing receptors as it did for those composed of NR2A or NR2B subunits. Nonetheless, this mutation reduced the inhibitory effects of ethanol when combined with the NR2C subunit. These results suggest the possibility that F639A may be important in defining a physical site important for the interaction between ethanol and the NMDA receptor and that changes in glycine efficacy are secondary to this change.
The Phe 639 residue in the TM3 domain of the NR1 subunit is one of several that were found to be insensitive to cysteinemodifying agents, suggesting that it does not face the pore of the ion channel (30). In contrast, a long stretch of TM3 residues at the C-terminal end of TM3 was accessible to these modifying agents, and this domain (SYTANLAAF) is the most highly conserved among all ionotropic glutamate receptors. This has led to the suggestion that the TM3 domain may be located near the central axis of the NMDA receptor ion channel and that amino acids in the outer portion of this domain interact with the aqueous lumen of the channel, whereas residues deeper within the membrane face amino acids located on other transmembrane domains. A mutation in the N-terminal end of TM3 of the glutamate ␦2 subunit gives rise to the Lurcher mouse phenotype that is characterized by constitutive channel activation (34). Mutation of homologous sites in various ionotropic glutamate receptors, including the NMDA receptor, alters channel properties and kinetics, suggesting that the TM3 domain may be an important regulator of receptor gating (35).
Finally, although the F639A mutation reduced the ethanol sensitivity of all heteromeric NMDA receptors tested in this study, it did not fully eliminate ethanol inhibition. However, because NMDA receptors probably contain at least two copies of an NR1 and NR2 subunit, mutation of both NR1 and NR2 subunits at the Phe 639 site may be required to create an ethanol-insensitive NMDA receptor. Interestingly, all NMDA and non-NMDA subunits have a phenylalanine at the site homologous to Phe 639 of the NR1 subunit, suggesting a key role for this residue. In the present study, expression of wild-type NR1 or NR1(F639A) subunits with NR2A subunits carrying the phenylalanine to alanine substitution (F637A) did not yield functional receptors. Further analysis of NMDA receptors carrying other amino acid substitutions at this site may reveal additional determinants of ethanol sensitivity.