Loop 2 Structure in Glycine and GABAA Receptors Plays a Key Role in Determining Ethanol Sensitivity*

The present study tests the hypothesis that the structure of extracellular domain Loop 2 can markedly affect ethanol sensitivity in glycine receptors (GlyRs) and γ-aminobutyric acid type A receptors (GABAARs). To test this, we mutated Loop 2 in the α1 subunit of GlyRs and in the γ subunit of α1β2γ2GABAARs and measured the sensitivity of wild type and mutant receptors expressed in Xenopus oocytes to agonist, ethanol, and other agents using two-electrode voltage clamp. Replacing Loop 2 of α1GlyR subunits with Loop 2 from the δGABAAR (δL2), but not the γGABAAR subunit, reduced ethanol threshold and increased the degree of ethanol potentiation without altering general receptor function. Similarly, replacing Loop 2 of the γ subunit of GABAARs with δL2 shifted the ethanol threshold from 50 mm in WT to 1 mm in the GABAA γ-δL2 mutant. These findings indicate that the structure of Loop 2 can profoundly affect ethanol sensitivity in GlyRs and GABAARs. The δL2 mutations did not affect GlyR or GABAAR sensitivity, respectively, to Zn2+ or diazepam, which suggests that these δL2-induced changes in ethanol sensitivity do not extend to all allosteric modulators and may be specific for ethanol or ethanol-like agents. To explore molecular mechanisms underlying these results, we threaded the WT and δL2 GlyR sequences onto the x-ray structure of the bacterial Gloeobacter violaceus pentameric ligand-gated ion channel homologue (GLIC). In addition to being the first GlyR model threaded on GLIC, the juxtaposition of the two structures led to a possible mechanistic explanation for the effects of ethanol on GlyR-based on changes in Loop 2 structure.

Alcohol abuse and dependence are significant problems in our society, with ϳ14 million people in the United States being affected (1,2). Alcohol causes over 100,000 deaths in the United States, and alcohol-related issues are estimated to cost nearly 200 billion dollars annually (2). To address this, considerable attention has focused on the development of medications to prevent and treat alcohol-related problems (3)(4)(5). The development of such medications would be aided by a clear understanding of the molecular structures on which ethanol acts and how these structures influence receptor sensitivity to ethanol.
A series of studies that employed chimeric and mutagenic strategies combined with sulfhydryl-specific labeling identified key regions within Cys-loop receptors that appear to be initial targets for ethanol action that also can determine the sensitivity of the receptors to ethanol (7-12, 18, 19, 26 -30). This work provides several lines of evidence that position 267 and possibly other sites in the transmembrane (TM) domain of GlyRs and homologous sites in GABA A Rs are targets for ethanol action and that mutations at these sites can influence ethanol sensitivity (8,9,26,31).
Growing evidence from GlyRs indicates that ethanol also acts on the extracellular domain. The initial findings came from studies demonstrating that ␣1GlyRs are more sensitive to ethanol than are ␣2GlyRs despite the high (ϳ78%) sequence homology between ␣1GlyRs and ␣2GlyRs (32). Further work found that an alanine to serine exchange at position 52 (A52S) in Loop 2 can eliminate the difference in ethanol sensitivity between ␣1GlyRs and ␣2GlyRs (18,20,33). These studies also demonstrated that mutations at position 52 in ␣1GlyRS and the homologous position 59 in ␣2GlyRs controlled the sensitivity of these receptors to a novel mechanistic ethanol antagonist (20). Collectively, these studies suggest that there are multiple sites of ethanol action in ␣1GlyRs, with one site located in the TM domain (e.g. position 267) and another in the extracellular domain (e.g. position 52).
Subsequent studies revealed that the polarity of the residue at position 52 plays a key role in determining the sensitivity of GlyRs to ethanol (20). The findings with polarity in the extracellular domain contrast with the findings at position 267 in the TM domain, where molecular volume, but not polarity, significantly affected ethanol sensitivity (9). Taken together, these findings indicate that the physical-chemical parameters of residues at positions in the extracellular and TM domains that modulate ethanol effects and/or initiate ethanol action in GlyRs are not uniform. Thus, knowledge regarding the physicalchemical properties that control agonist and ethanol sensitivity is key for understanding the relationship between the structure and the actions of ethanol in LGICs (19, 31, 34 -40).
GlyRs and GABA A Rs, which differ significantly in their sensitivities to ethanol, offer a potential method for identifying the structures that control ethanol sensitivity. For example, ␣1GlyRs do not reliably respond to ethanol concentrations less than 10 mM (32,33,41). Similarly, ␥ subunit-containing GABA A Rs (e.g. ␣1␤2␥2), the most predominantly expressed GABA A Rs in the central nervous system, are insensitive to ethanol concentrations less than 50 mM (42,43). In contrast, ␦ subunit-containing GABA A Rs (e.g. ␣4␤3␦) have been shown to be sensitive to ethanol concentrations as low as 1-3 mM (44 -51). Sequence alignment of ␣1GlyR, ␥GABA A R, and ␦GABA A R revealed differences between the Loop 2 regions of these receptor subunits. Since prior studies found that mutations of Loop 2 residues can affect ethanol sensitivity (19,20,39), the non-conserved residues in Loop 2 of GlyR and GABA A R subunits could provide the physical-chemical and structural bases underlying the differences in ethanol sensitivity between these receptors.
The present study tested the hypothesis that the structure of Loop 2 can markedly affect the ethanol sensitivity of GlyRs and GABA A Rs. To accomplish this, we performed multiple mutations that replaced the Loop 2 region of the ␣1 subunit in ␣1GlyRs and the Loop 2 region of the ␥ subunit of ␣1␤2␥2 GABA A Rs with corresponding non-conserved residues from the ␦ subunit of GABA A R and tested the sensitivity of these receptors to ethanol. As predicted, replacing Loop 2 of WT ␣1GlyRs with the homologous residues from the ␦GABA A R subunit (␦L2), but not the ␥GABA A R subunit (␥L2), markedly increased the sensitivity of the receptor to ethanol. Similarly, replacing the non-conserved residues of the ␥ subunit of ␣1␤2␥2 GABA A Rs with ␦L2 also markedly increased ethanol sensitivity of GABA A Rs. These findings support the hypothesis and suggest that Loop 2 may play a role in controlling ethanol sensitivity across the Cys-loop superfamily of receptors. The findings also provide the basis for suggesting structurefunction relationships in a new molecular model of the GlyR based on the bacterial Gloeobacter violaceus pentameric LGIC homologue (GLIC).

Materials
Adult female Xenopus laevis frogs were purchased from Nasco (Fort Atkinson, WI). Gentamicin, 3-aminobenzoic acid ethyl ester, glycine, GABA, ethanol, zinc chloride, strychnine, picrotoxin, diazepam, and collagenase were purchased from Sigma. All other chemicals used were of reagent grade. Glycine, GABA, and strychnine stock solutions were prepared from powder. Stock solutions of picrotoxin and diazepam were prepared in DMSO and then diluted to an appropriate concentration with the extracellular solution just before use. To avoid adverse effects from DMSO exposure, the final concentration (v/v) of DMSO was not higher than 0.5%. Picrotoxin stocks and solutions were wrapped in foil to avoid UV exposure.

Expression in Oocytes
The amino acid sequences for ␣1GlyR and ␦and ␥GABA A R subunits were aligned, and the Loop 2 regions were compared (Table 1). Individual point mutations in the ␣1GlyR or ␥GABA A R subunit cDNA were created so that the resulting Loop 2 region matched that of the ␦GABA A R or the ␥GABA A R subunits. Xenopus oocytes were isolated and injected with human GlyR cDNAs (1 ng/32 nl) or GABA A R cDNAs (1:1:10 ratio for a total volume of 1 ng of ␣1␤2␥2) cloned into the mammalian expression vector pCIS2 or pBKCMV, as described previously (33), and verified by partial sequencing (DNA Core Facility, University of Southern California). After injection, oocytes were stored in incubation medium (modified Barth's saline supplemented with 2 mM sodium pyruvate, 0.5 mM theophylline, and 50 mg/liter gentamycin) in Petri dishes (VWR, San Dimas, CA). All solutions were sterilized by passage through 0.22-m filters. Oocytes, stored at 18°C, usually expressed GlyRs the day after injection and GABA A Rs 3-4 days after injection. Oocytes were used in experiments for up to 7 days after injection.
Native ␦-containing GABA A Rs (␣4␤2/3␦ and ␣6␤2/3␦) have been shown to be sensitive to low ethanol concentrations (1-3 mM) in a variety of preparations (44 -51) However, these receptors are difficult to express in oocytes. This topic has been the subject of several reviews (52)(53)(54). The goal of the present study was to test the hypothesis that the structure of Loop 2 can markedly affect the ethanol sensitivity of GlyRs and GABA A Rs. We used the ␦ Loop 2 as a vehicle for testing this hypothesis. In this context, and given the difficulties described above, we did not include WT ␦-containing GABA A Rs in the current paper.

Whole Cell Two-electrode Voltage Clamp Recordings
Two-electrode voltage clamp recording was performed using techniques similar to those previously reported (33). Briefly, electrodes pulled (P-30; Sutter Instruments, Novato, CA) from borosilicate glass (1.2-mm thick walled filamented glass capillaries (WPI, Sarasota, FL)) were back-filled with 3 M KCl to yield resistances of 0.5-3 megaohms. All electrophysiological recordings were conducted within a chamber that contains a vibration-resistant platform that supports an oocyte bath, two micro positioners (WPI (Sarasota, FL) or Narishige International USA, Inc. (East Meadow, NY)), and bath clamp (33). Oocytes were perfused in a 100-l oocyte bath with modified Barth's saline with or without drugs via a custom high pressure drug delivery system (Alcott Chromatography, Norcross, GA) at 2 ml/min using 1 ⁄ 16 OD high pressure PEEK tubing (Upchurch Scientific, Oak Harbor, WA). Oocytes were voltageclamped at a membrane potential of Ϫ70 mV using a Warner Instruments model OC-725C (Hamden, CT) oocyte clamp. A chart recorder (Barnstead/Thermolyne, Dubuque, IA) continuously plotted the clamped currents. The peak currents were measured and used in data analysis. All experiments were performed at room temperature (20 -23°C).

Application of Agonist
For agonist concentration response experiments, WT or mutant GlyRs or GABA A Rs were exposed to 1 M to 3 mM glycine or 1 M to 10 mM GABA for 60 s, using 5-15-min washouts between applications to ensure complete receptor resensitization.
Application of Ethanol-We used a concentration of glycine or GABA producing 10 Ϯ 2% of the maximal effect (EC 10 ). Agonist EC 10 was applied as a control pre-and post-ethanol treatment. When testing ethanol potentiation, the oocytes were preincubated with ethanol for 60 s prior to co-application of ethanol and agonist for 60 s (18). Washout periods (5-15 min) were allowed between agonist and drug applications to ensure complete resensitization of receptors. WT and mutant ␣1GlyR responses were measured across an ethanol concentration range of 1-30 mM. GABA A R responses were measured across an ethanol concentration range of 1-50 mM. Ethanol, in the absence of glycine or GABA, did not significantly affect the holding currents of the GlyRs and GABA A Rs tested.

Application of Antagonists and Modulators
Zinc Chloride-Oocytes expressing WT, ␦L2, and ␥L2 GlyRs were tested for response to low (10 M) and high (100 M) concentrations of zinc chloride (ZnCl 2 ), a bimodal allosteric modulator of the GlyR. Glycine EC 10 was applied for 60 s. Oocytes were preincubated with ZnCl 2 for 60 s, followed by co-application with glycine EC 10 for 60 s. Wash-out periods (5-15 min) were allowed between drug applications to ensure complete resensitization of receptors.
Strychnine and Picrotoxin-Oocytes expressing WT, ␦L2, and ␥L2 GlyRs were tested for response to the competitive GlyR antagonist strychnine or the noncompetitive GlyR antagonist picrotoxin. Glycine EC 10 was applied for 60 s. Oocytes were preincubated with strychnine (50 nM) or picrotoxin (100 M) for 60 s, followed by co-application with glycine EC 10 for 60 s. Washout periods (5-15 min) were allowed between drug applications to ensure complete resensitization of receptors.
Diazepam-Oocytes expressing WT and ␦L2 GABA A Rs were tested for response to the benzodiazepine agonist diazepam. GABA EC 10 was applied for 60 s. Oocytes were preincubated with diazepam (1 M) for 60 s, followed by co-application with GABA EC 10 for 60 s. Washout periods (5-15 min) were allowed between drug applications to ensure complete resensitization of receptors.

Cell Surface Biotinylation and Immunoblotting
Biotinylation of surface-expressed proteins was modified from a previous protocol published by Chen et al. (55). Four days after cDNA injections, oocytes (15 oocytes/group) were incubated with 1.5 mg/ml membrane-impermeable sulfosuccinimidyl 2-(biotinamido)-ethyl-1,3-dithiopropionate (Pierce) for 30 min at room temperature. After washing once with 25 mM Tris (pH 8.0) and twice with phosphate-buffered saline, oocytes were homogenized in 500 l of lysis buffer (40 mM Tris (pH 7.5), 110 mM NaCl, 4 mM EDTA, 0.08% Triton X-100, 1% protease inhibitor mixture (Vector Laboratories, Burlingame, CA)). The yolk and cellular debris were removed by centrifugation at 3600 ϫ g for 10 min. Aliquots of the supernatant were mixed with 2ϫ SDS loading buffer and stored at Ϫ20°C to assess total receptor fraction. The remaining supernatant was incubated with streptavidin beads (Pierce) overnight at 4°C. Beads were washed three times with lysis buffer, and the biotinylated proteins were eluted by heating at 95°C for 10 min in SDS loading buffer. The surface and total proteins were separated using SDS-PAGE and transferred to polyvinylidene fluoride membranes. The membranes were incubated overnight with rabbit anti-GlyR antibody (1:500 dilution; Chemicon International, Temecula, CA), followed by incubation with the appropriate horseradish peroxidase-conjugated secondary antibody. Protein bands were visualized using enhanced chemiluminescence (Pierce). The blots were then scanned and analyzed to obtain images.

Molecular Modeling
Models of the WT and ␦L2 mutant GlyRs were built using Discovery Studio 2.1 (Accelrys, San Diego, CA). The GlyR and the mutant sequence with the ␦GABA Loop 2 were aligned with the "Align multiple sequences" module, a derivative of ClustalW. To ensure compatibility with the literature, a two-step procedure was used to test the alignments. First, we used the alignment of ␣1GlyR with ␣1nAChR suggested by Sixma and co-workers (56). Second, we used the alignment of ␣1nAChR with GLIC suggested by Changeux and co-workers (57,58). The resulting alignment of GlyR with GLIC proved to be correct (Table 1). We then submitted the two alignments to the "Modeler" module with the restriction that the Cys-loop cysteine disulfide bond (Cys 138 -Cys 152 ) should be preserved. For each alignment, 10 initial models were produced, and then each of these was subjected to loop refinement to yield a total of 50 models for WT and mutant receptors. The "best" model for each alignment was selected based on total force field PDF energy (a calculated value called the probability density function, which is derived from spatial restraints when building the initial models and can be used to identify high energy regions of the structure). Then each model was further refined with the "Loop refinement" module. At this point, a harmonic restraint of 10 kcal/(mol A 2 ) was applied to all backbone atoms of the homopentamers, and this restraint was maintained for all of the following steps. Both models were optimized to a gradient of 0.0001 kcal/Å in Discovery Studio with a conjugate gradient algorithm using the Accelrys version of the CHARMm force field and the default spherical nonbond cut-off of 14 Å. Then the models were relaxed with 50,000 1-fs steps of molecular dynamics at 300 K. Finally, the models were optimized again as described above. The WT and mutated models had final potential energies of Ϫ88,054 and Ϫ88,487, respectively. These values indicate that the models are stable. However, due to the extensive changes in amino acids, the energies of the models cannot be compared to determine which is more stable. In both models, intersubunit and intrasubunit interactions of residues in Loop 2 were detected with two methods. First, the hydrogen bond detection module was enabled. Second, all residues within 5 Å of any atom in Loop 2 were selected and manually examined.

Data Analysis
Data for each experiment were obtained from 4 -8 oocytes from at least two different frogs. n refers to the number of oocytes tested. Results are expressed as mean Ϯ S.E. Where no error bars are shown, they are smaller than the symbols. We used Prism (GraphPAD Software, San Diego, CA) to perform curve fitting and statistical analyses. Agonist concentration response data were analyzed using non-linear regression anal- , where I is the peak current recorded following application of a range of agonist concentrations, [A]; I max is the estimated maximum current; EC 50 is the glycine concentration required for a half-maximal response, and n H is the Hill slope). Data were subjected to Student's t tests, one-or two-way analysis of variance (ANOVA) with Dunnett's multiple comparison or Bonferroni post-tests when warranted. To determine the threshold concentration at which a significant effect of ethanol was first detected in WT and mutant receptors, we compared the absolute values of agonistinduced chloride currents in the presence and absence of ethanol across ethanol concentrations using two-way ANOVA, followed by Bonferroni post-tests. Statistical significance was defined as p Ͻ 0.05.

Agonist Concentration Response
GlyRs-Glycine produced inward Cl Ϫ currents in WT and mutant GlyRs in a concentration-dependent manner (Fig. 1). There were no significant differences between WT and mutant GlyRs in glycine I max or Hill slope ( Table 2). The ␦L2 mutation in ␣1GlyRs caused a significant reduction in EC 50 in these receptors compared with WT ␣1GlyRs. In contrast, the ␥L2 GlyRs did not differ significantly from WT in terms of EC 50 . Cell surface biotinylation followed by immunoblotting analysis did not show a significant difference between cell surface biotinylated fraction or total expression of GlyR protein between WT and any of the mutant GlyRs tested (Fig. 2). This suggests that the differences in EC 50 of WT versus ␦L2 GlyRs do not reflect differences in surface expression levels due to receptor internalization.
GABA A Rs-GABA produced inward Cl Ϫ currents in WT and mutant GABA A Rs in a concentration-dependent manner (Fig. 3). The ␣1␤2␥2(␦L2) GABA A R mutation caused a nonsignificant left shift in EC 50 . There were no significant differ-ences in I max or Hill slope between WT and mutant GABA A Rs (Table 3).

Ethanol Concentration Response
GlyRs-We predicted that mutating the Loop 2 region in ␣1GlyRs to the homologous residues from the ␦GABA A R subunit would increase ethanol sensitivity of ␣1GlyRs. As predicted, the Loop 2 substitution in WT ␣1GlyRs reduced the threshold for ethanol sensitivity from 30 mM in WT GlyRs to 1 mM in the ␦L2 mutant and increased the degree of ethanol potentiation at all concentrations tested (Fig. 4). On the other hand, mutating the Loop 2 region in ␣1GlyRs to the homologous residues from ␥GABA A R did not significantly affect ethanol sensitivity compared with WT GlyRs. Therefore, changes in ethanol sensitivity caused by mutating Loop 2 of the ␣1GlyR to the Loop 2 sequence found in ␦and ␥GABA A R subunits,  Table 2. Each data point represents the mean Ϯ S.E.

FIGURE 2. Western blot analysis of total and cell surface protein from
Xenopus oocytes expressing WT, ␦L2, and ␥L2 ␣1GlyR subunits. Western blot analysis revealed no differences between WT and mutant GlyRs with respect to total cell lysates and cell surface biotinylated fractions. Results shown are for 1 ng of WT or mutant GlyR cDNA injected into each oocyte. Immunoprecipitates were run on SDS-polyacrylamide gel and then transferred to polyvinylidene difluoride membrane. Blots were then probed with rabbit antibody against the ␣1 subunit of the human GlyR. respectively, parallel the relative ethanol sensitivities of the GABA A Rs from which the Loop 2 sequence was taken. Overall, these findings support the notion that the structure of Loop 2 plays a key role in determining ethanol sensitivity in GlyRs. GABA A Rs-If Loop 2 plays a key role in the ethanol sensitivity of GABA A Rs, then mutating Loop 2 of the ␥ subunit of ␣1␤2␥2 GABA A Rs to the homologous sequence in the ␦GABA A R subunit should increase ethanol sensitivity of ␣1␤2␥2 GABA A Rs. As predicted, the ␦L2 mutation in the ␥ subunit of GABA A Rs shifted the threshold for ethanol sensitivity from 50 mM in WT, to 1 mM in the GABA A ␥-␦L2 mutant receptor and markedly increased the magnitude of the ethanol response compared with WT GABA A Rs (Fig. 5). Overall, the results support the notion that the structure of Loop 2 plays a key role in determining the ethanol sensitivity of GABA A Rs.

Additional Tests of Receptor Function
Zinc Chloride-Zinc is an allosteric modulator of the GlyR that modulates the receptor in a bimodal manner. Submicromolar to micromolar concentrations of ZnCl 2 enhance GlyR function by acting at a high affinity Zn 2ϩ binding site, whereas micromolar concentrations of ZnCl 2 Ն100 M cause inhibition of GlyR function at a low affinity Zn 2ϩ binding site (59,60). In agreement with previous work, low concentrations of ZnCl 2 (10 M) enhanced EC 10 glycine-activated currents, whereas higher concentrations of ZnCl 2 (100  Table 3. Each data point represents the mean Ϯ S.E.   . The ␦L2 GlyR mutation did not affect biphasic modulation by Zn 2؉ in GlyRs. Zn 2ϩ allosterically modulated WT, ␦L2, and ␥L2 GlyRs in a bimodal manner. 10 M ZnCl 2 caused enhancement of glycine-activated currents, whereas 100 M ZnCl 2 caused inhibition in both WT and mutant GlyRs. One-way ANOVA followed by Dunnett's post-tests revealed no significant differences between WT and ␦L2 GlyRs with respect to modulation by Zn 2ϩ at either concentration. The response to 10 M ZnCl 2 of ␥L2 GlyRs did not differ significantly from WT, but the response to 100 M ZnCl 2 was significantly reduced in these receptors. Data are shown as mean Ϯ S.E. percentage of control (where the EC 10 control response is 100%).

TABLE 3 Summary of non-linear regression analysis results for GABA concentration responses in WT and mutant GABA A Rs
GABA EC 50 , Hill slope (n H ), and maximal current amplitude (I max ) are presented as mean Ϯ S.E. Student's t test revealed no significant differences between WT and mutant GABA A Rs in I max , EC 50 , or Hill slope. M) inhibited glycine-activated currents in WT GlyRs (Fig.  6). The ␦L2 mutation did not significantly alter the effects of ZnCl 2 at either concentration tested. 100 M ZnCl 2 caused a significantly greater inhibition of glycine-activated currents in the ␥L2 mutant receptor (Fig. 6). Strychnine-Strychnine is a competitive antagonist of the glycine binding site in ␣1GlyRs (61). In order to test if Loop 2 mutations interfered with strychnine binding, oocytes expressing WT and Loop 2 mutant GlyRs were tested for response to 50 nM strychnine. In agreement with previous work (33), strychnine inhibited glycine-activated currents in WT ␣1GlyRs (Fig.  7). The ␦L2 mutation did not alter the effects of strychnine on these mutant ␣1GlyRs. There was a significant increase in strychnine inhibition of glycine-activated currents in the ␥L2 mutant receptor.

Receptor
Picrotoxin-Picrotoxin is a plant alkaloid convulsant that inhibits homomeric ␣1GlyRs with a high potency by blocking the channel pore (62). In order to test if Loop 2 mutations interfered with the effects of picrotoxin, oocytes expressing WT and Loop 2 mutant GlyRs were tested for response to 100 M picrotoxin. Picrotoxin inhibited glycine-activated currents in WT ␣1GlyRs (Fig. 8). The ␦L2 and ␥L2 mutations did not alter the effects of picrotoxin in ␣1GlyRs.
Diazepam-Diazepam is the prototypical benzodiazepine agonist and potentiates the GABA responses through binding to an allosteric modulatory site on the receptor (63). In order to test if mutations to the ␥ subunit of the GABA A R interfered with the effects of diazepam, oocytes expressing WT and ␦L2 mutant GABA A Rs were tested for response to 1 M diazepam. Diazepam potentiated GABA-activated currents in WT ␣1␤2␥2 GABA A Rs (Fig. 9). The ␦L2 mutation did not significantly alter the effects of diazepam on the receptor.
Collectively, these findings in GlyR and GABA A R suggest that replacement of non-conserved residues in Loop 2 with those of ␦GABA increase ethanol sensitivity and that these changes in ethanol sensitivity cannot be explained by changes in the basic receptor function. Interestingly, the ␦L2 mutations did not affect allosteric modulation by Zn 2ϩ in GlyRs or by diazepam in GABA A Rs, which indicates that the changes in ethanol sensitivity produced by this mutation do not extend to all allosteric modulators.

Molecular Modeling of WT Versus ␦L2 GlyR
The model of the ␣1GlyR based on the template of the prokaryotic LGIC GLIC (Protein Data Bank code 3EAM) showed that Loop 2 is a tight ␤ turn (Fig 10A). This is an important point in that the previous best x-ray structure we used as a template for the ligand-binding domain had a more rounded structure for Loop 2. This template was the acetylcholine-binding protein (Protein Data Bank code 1I9B), and the differences in Loop structure are probably due to the acetylcholine-binding protein being a water-soluble protein with Loop 2 facing the aqueous environment, whereas Loop 2 in GlyR is at the relatively hydrophobic interface of two domains. Another notable feature of this GlyR model is that Lys 276 extends out from the TM2-TM3 FIGURE 7. The ␦L2 GlyR mutation did not affect inhibition by strychnine in GlyRs. 50 nM strychnine inhibited WT, ␦L2, and ␥L2 ␣1GlyRs. Oneway ANOVA followed by Dunnett's post-tests showed no significant difference in the degree of strychnine inhibition between WT and ␦L2 mutant GlyRs. In contrast, strychnine inhibited ␥L2 GlyRs significantly more than WT GlyRs. Data are shown as mean Ϯ S.E. percentage of control (where the EC 10 control response is 100%).  linker and makes a salt bridge with Glu 53 in Loop 2 of the adjacent subunit. It is noteworthy that this salt bridge now extends directly across the intersubunit cavity.
Three GlyR subunits are shown (Fig. 10A) in order to emphasize the intersubunit interactions that are possible, whereas specific interactions within Loop 2 are shown in an expanded view of the WT ␣1GlyR (Fig. 10B). Here we consider interactions of GlyR residues 50 -59 with other residues within Loop 2 and in the ␤ strands surrounding them. Ser 50 interacts directly across the top of Loop 2 and forms a hydrogen bond with Arg 59 . Ile 51 points toward the ␤ sheet below, whereas Ala 52 points more toward the ion pore. In addition, Ala 52 is approximately in the "i" position of a ␤ turn (39) and Glu 53 points away from the center of the turn and forms a salt bridge with Arg 218 in the Pre-TM1 segment of its subunit and with Lys 276 of the neighboring subunit. Thr 54 forms a hydrogen bond with Ser 273 in the TM2-TM3 linker and interacts with Phe 187 . Thr 55 interacts across Loop 2, whereas Met 56 points away from Loop 2. Asp 57 forms a salt Although only 4 of the 10 residues in Loop 2 are conserved in the mutated construct that we made, the global structure of the backbone of Loop 2 is essentially identical in the GlyR WT (Fig.  10B) and the ␦L2 mutant construct (Fig. 10C). This is remarkable, because each of the two sequences was independently used by the Modeler module of Discovery Studio to build the models. The best of 50 models was selected based on potential energy in the CHARMm force field, and then side chain positions were adjusted with the autorotomer module, a short molecular dynamics run was made, and then the two final structures were reoptimized. The positions of other residues that interact with those in Loop 2 were also conserved, especially Lys 104 and Leu 136 . Lys 276 still projects away from the TM2-TM3 linker and forms a salt bridge with Glu 53 in the adjacent subunit. The most notable changes are how Arg 218 interacts with Glu 53 with a much different form of salt bridge. This change resulted in a small distortion of the pre-TM1 segment compared with the WT GlyR. As expected, the substitution of Asp 57 with glutamate resulted in a shift of the salt bridge with Lys 104 to compensate for the increased length of the glutamate side chain.

DISCUSSION
The present study tests the hypothesis that the structure of extracellular domain Loop 2 can markedly affect ethanol sensitivity in both GlyRs and GABA A Rs. We found that replacing Loop 2 of the ␣1GlyR subunit with that of the ␦GABA A R subunit, but not the ␥GABA A R subunit, reduced the threshold for ethanol sensitivity and increased the degree of ethanol potentiation without altering the general function of the receptor. Similarly, replacing the Loop 2 region of the ␥ subunit of GABA A Rs with the Loop 2 region of ␦GABA A R shifted the threshold for ethanol sensitivity from 50 mM in WT to 1 mM in the GABA A R ␥-␦L2 mutant. These results indicate that manipulations of Loop 2 structure can have profound effects on ethanol sensitivity of these receptors. Given the relatively high structural homology between the Cys-loop superfamily of receptors (36,38,56,64), these findings in GlyR and GABA A R could extend to nAChRs and 5-hydroxytryptamine 3 receptors.
As presented, the ␦L2 mutations increased ethanol sensitivity without altering sensitivity of GlyR and GABA A R, respectively, to allosteric modulation by Zn 2ϩ and diazepam. Further work is necessary to test other allosteric modulators of GlyRs and GABA A Rs, particularly other general anesthetics like isoflurane, halothane, and propofol (65)(66)(67). Nonetheless, the lack of change in sensitivity of ␦L2 mutant GlyRs and GABA A Rs to the allosteric modulators tested suggests that the changes in ethanol sensitivity by this mutation do not extend to other allosteric modulators and may be specific for ethanol or ethanollike agents.
The mechanism by which mutation of Loop 2 alters ethanol sensitivity in GlyRs and GABA A Rs is unknown. However, the current and previous studies provide some insights. With one exception, a left shift in glycine EC 50 in the ␦L2 GlyR, Loop 2 mutations that increased ethanol sensitivity did not alter recep-tor EC 50 , I max , or Hill slope. Similarly, the ␦L2 GABA A R mutation resulted in increased ethanol sensitivity without a significant change in GABA sensitivity. Prior studies also found that mutation of position 52 in Loop 2 could alter ethanol sensitivity in GlyRs without changing EC 50 (19,20). Moreover, the ␦L2 mutation in GlyRs did not significantly affect the response of the receptors to strychnine or picrotoxin. Together, these findings indicate that the increase in ethanol sensitivity in ␦L2 mutants cannot be explained by changes in receptor conformation that alter basic receptor function.
Interestingly, prior studies indicate that ethanol sensitivity in recombinant ␣1␤2␦ GABA A Rs expressed in Xenopus oocytes is not increased. Rather, the ethanol sensitivity of this subunit combination is similar to that seen in WT ␣1␤2␥2 GABA A Rs (51). Further studies are necessary to ensure incorporation of the ␦ subunit in this work. Nonetheless, these findings suggest that there is an important interaction between ␣ and ␦ subunits that is involved in making WT ␣4␤2/3␦ and ␣6␤2/3␦ GABA A Rs highly sensitive to ethanol. Taken in conjunction with the present results, these findings in ␣1␤2␦ GABA A Rs also support the conclusion that the structure of Loop 2 plays a critical role in producing high ethanol sensitivity in the ␦L2 mutant GABA A Rs and probably also the ␦L2 mutant GlyRs, tested in the present study.
Mutations of Loop 2 structure could alter ethanol sensitivity by changing the physical-chemical characteristics of the amino acids at key locations and their interactions within Loop 2 and/or with the TM domain. This notion is supported by several lines of evidence and by the models described below. Prior studies provide evidence that position 52 in Loop 2 of the extracellular domain and position 267 in the TM domain of ␣1 GlyRs are sites of ethanol action (8, 18 -20, 31, 32) and that ethanol causes qualitatively different (position-specific) effects when acting on these targets (19). Further studies used cysteine mutations at these positions in combination with propyl methanethiosulfonate to suggest that these sites were part of the same ethanol pocket (19). Given that this pocket contains multiple sites that are capable of producing ethanol effects, we describe the pocket as an ethanol "action pocket" to distinguish it from classical high affinity binding sites. Molecular modeling revealed a cavity that extends ϳ28 Å from the C␣ atoms of Ala 52 to Ser 267 that could function as this alcohol action pocket (19). As proposed by these authors, this pocket would be large enough to hold several ethanol molecules. The estimated 28-Å distance between positions 52 and 267 precludes action by one ethanol molecule on both sites simultaneously. Hence, the probability that ethanol molecule(s) will be acting on one or more of these sites at a given moment increases as the ethanol concentration increases. The net response to ethanol on receptor function will represent the summation of the actions of ethanol on these potentially independent targets.
Interestingly, further study found that the polarity of the residue at position 52 plays a key role in determining the sensitivity of GlyRs to ethanol (20). The findings with polarity contrast with the findings at position 267 in the TM domain, where others found that molecular volume, but not polarity, significantly affects ethanol sensitivity (9). Taken together, these findings indicate that the physical-chemical parameters at positions in the extracellular and TM domains that modulate ethanol effects and/or initiate ethanol action in GlyRs are not uniform and may respond to different concentrations of ethanol. GABA A Rs have not been investigated extensively in this respect, but parallel studies that implicate the homologous positions in GABA A Rs as targets for ethanol action and modulation, combined with the structural homology between GlyRs and GABA A Rs (8,38,56,64), suggest that the same factors may apply for GABA A Rs. Knowledge regarding the physical-chemical properties that control ethanol sensitivity is key for understanding the relationship between structure and the actions of ethanol in receptors and for building molecular models of the ethanol sites of action.
Several molecular models of LGICs have been developed that have begun to describe possible pairwise ionic interactions between critical residues in the extracellular and TM domains that may contribute to agonist action (36,38,39,64,68,69). These studies employed techniques such as charge reversal and cysteine cross-linking to identify conformational changes in receptor proteins, including GlyRs and GABA A Rs that may be involved in agonist activation or transduction. Molecular models have been developed that identify putative sites of ethanol action in GlyRs (19,26,40). However, these models have not addressed possible molecular mechanisms that initiate, transduce, or modulate the actions of ethanol.
Here, we present a molecular model of the GlyR threaded on the x-ray structure of GLIC. In addition to being the first GlyR model threaded on GLIC, it is the first model that offers a mechanistic explanation for the effects of ethanol on the GlyR based on changes in Loop 2 structure. The latter are revealed by juxtaposing the models derived from threading the WT versus the ␦L2 GlyR sequences onto GLIC. The change in conformation as a result of the ␦L2 substitution in mutant GlyRs changes the manner in which Arg 218 (pre-TM1) interacts with Glu 53 (Loop 2) with a much different form of salt bridge. The delocalized charge of the three partially positive nitrogen atoms (N-H ϩ groups) at the guanidinium end of the arginine side chain allows it to form a salt bridge with the glutamate carboxyl group either straight-on (the longest net distance) or at either side of the arginine side chain (shorter net distance and not linear). The result of the ␦ Loop 2 mutation is to form the more distorted side-on salt bridge in our modeling. This change causes a small distortion of the pre-TM1 segment compared with the WT GlyR. Moreover, the ␦L2 mutant GlyR has an aspartic acid residue at position 57 in place of the glutamic acid found in WT. As expected, the substitution of Asp 57 with glutamate results in a shift of the salt bridge with Lys 104 to compensate for the increased length of the glutamate side chain. However, it is unlikely that these are just static changes. Rather, they would change the ensemble of conformations that may occur during gating and may be affected by the presence of alcohol molecules, which could alter ethanol sensitivity. If valid, this suggests that these dynamic movements are involved in causing and/or transducing the action of ethanol in Loop 2.
Despite the low homology between Loop 2 residues in ␣1GlyR and ␦GABA A R, the global structure of the ␤ turn is conserved in the chimera, illustrating the importance of structural homology across the Cys-loop superfamily. This suggests that insights provided by the current model may generalize to GABA A Rs and other members of the superfamily. Two notable differences in the model in Fig. 10 stand out. First, the side chain of Lys 276 extends out from the TM2-3 linker to make contact with the conserved Glu 53 in Loop 2, forming an intersubunit salt bridge. This intersubunit salt bridge has not been observed in previous x-ray or cryoelectron microscopy structures and is not present in the GLIC template used for modeling. It is possible that the solvation/ desolvation of this salt bridge is important for the structural rearrangements that accompany the gating transition (70). Second, the salt bridge between Arg 218 and Glu 53 has a different conformation in the ␦L2 mutant GlyR. The altered length of this salt bridge may contribute to the differences in sensitivity to glycine and ethanol. In addition, it should be noted that the partial negative charges on Glu 53 , at the tip of the ␤ turn in Loop 2, are shared between Arg 218 and Lys 276 . These complicated electrostatic and steric interactions might be especially sensitive to the presence of ethanol molecules in the adjacent cavity. These findings exploring the role of Loop 2 and the ␦ GABA sequence exemplify how increasing our knowledge regarding the structures that can modulate ethanol sensitivity can increase our understanding of the targets for ethanol and structure-function relationships.
GlyRs and GABA A Rs are widely held to represent initial targets for ethanol action that underlie a broad spectrum of ethanol-induced acute and chronic behavioral effects. Behavioral effects in humans can be detected at blood ethanol concentrations beginning at ϳ0.03% (w/v) (7 mM) (71). The legal limits for alcohol consumption while driving are 0.05% (w/v) (11 mM) in most European Union countries and 0.08% (w/v) (17 mM) in the United States (72). A blood alcohol concentration of 0.40% (w/v) (88 mM) is lethal in 50% of the population (73). Therefore, the present studies in recombinant receptors, which identify Loop 2 as a structure that can modulate ethanol sensitivity across this broad range of behaviorally and toxicologically relevant concentrations, could provide insight into the structural basis for individual differences in ethanol sensitivity.
The findings also suggest the exciting possibility that structural modifications of Loop 2 in GlyR and GABA A R might be used to markedly increase the ethanol sensitivity in target receptor populations (e.g. specific receptor subtypes or brain regional populations) in transgenic animals. This approach could result in new tools for measuring the effects of ethanol on sensitized receptors in which overexpression of high ethanol sensitivity mutant receptors in neurons will enable us see the effects of ethanol on these receptors at very low concentrations (ϳ1 mM) that should not elicit responses from endogenous receptors. Hence, we should be able to detect this effect of ethanol on the neuron without interference from its action on endogenous receptors. If valid, this would provide an alternative strategy that could be used to map the specific behavioral effects of ethanol caused by its actions on respective receptor populations. Increased knowledge regarding the initial sites for ethanol action and the structures that affect sensitivity to ethanol also could provide new targets for the development of therapeutic agents to prevent or help treat alcohol-related disorders.