Spontaneous Mobility of GABAA Receptor M2 Extracellular Half Relative to Noncompetitive Antagonist Action*

The γ-aminobutyric acid type A receptor β3 homopentamer is spontaneously open and highly sensitive to many noncompetitive antagonists(NCAs) and Zn2+.OurearlierstudyoftheM2cytoplasmic half (-1′ to 10′) established a model in which NCAs bind at porelining residues Ala2′, Thr6′, and Leu9′. To further define transmembrane 2 (M2) structure relative to NCA action, we extended the Cys scanning to the extra cellular half of the β3 homopentamer (11′ to 20′). Spontaneous disulfides formed with T13′C, L18′C, and E20′C from M2/M2 cross-linking and with I14′C (weak), H17′C, and R19′Con bridging M2/M3 intersubunits, based on single (M2 Cys only) and dual (M2 Cys plus M3 C289S) mutations. Induced disulfides also formed with T16′C, but there were few or none with M11′C, T12′C, and N15′C. These findings show conformational flexibility/mobility in the M2 extracellular half 17′ to 20′ region interpreted as a deformed β-like conformation in the open channel. The NCA radioligands used were [3H]1-(4-ethynylphenyl)-4-n-propyl-2,6,7-trioxabicyclo[2.2.2]octane ([3H]EBOB) and [3H]3,3-bis-trifluoromethylbicyclo[2.2.1]heptane-2,2-dicarbonitrile with essentially the same results. NCA binding was disrupted by individual Cys substitutions at 13′,14′,16′,17′, and 19′. The inactivity of T13′C/T13′S may have been due to disturbance of the channel gate; I14′S and T16′S showed much better binding activity than their Cys counterparts, and the low activities of H17′C and R19′C were reversed by dithiothreitol. Zn2+ potency for inhibition of [3H]EBOB binding was lowered 346-fold by the mutation H17′A. We propose that NCAs enter their binding site both directly, through the channel pore, and indirectly, through the water cavity of adjacent subunits.

The GABA A receptor (GABA A R) 2 of the ligand-gated ion channel (LGIC) superfamily (1,2) is blocked by noncompetitive antagonists (NCAs) of widely diverse structures (3) including some of the most important insecticides (4) and, at another site, by Zn 2ϩ (5). Complete mutational scanning of the 15 amino acids in the cytoplasmic half and adjacent regions of the M2 segments of the ␤ 3 homopentamer (Ala Ϫ4 Ј to Thr 10 Ј) 3 establishes that eight are directly or indirectly involved in NCA binding with Ala 2 Ј, Thr 6 Ј, and Leu 9 Ј of greatest importance (3). The extracellular half of the M2 segments (Met 11 Ј to Glu 20 Ј) has a proposed NCA interaction site at S17ЈC in an ␣ 1 subunit (6) and a Zn 2ϩ site at His 17 Ј and Glu 20 Ј in the ␤ 3 subunit (5).
The ␤ 3 homopentamer is ideal for defining the open state channel structure relative to NCA action because it is self-assembling, spontaneously open, and symmetrical with high sensitivities to NCAs and Zn 2ϩ (3,11,12). To further understand M2 structure and NCA action, here we have examined the role of the extracellular half of the M2 segments in NCA and Zn 2ϩ binding involving site-directed mutagenesis with Cys, Ser, and Ala replacements in M2 and M3 using two NCA radioligands. The structural aspects were studied with SCAM (substituted Cys accessibility method) and disulfide trapping (3,13). This investigation establishes the conformational flexibility of the extracellular half of M2 segments in the ␤ 3 homopentamer and suggests that NCAs access their binding sites both directly, through the channel pore, and indirectly, through the interface of adjacent subunits. The relationships considered lead to a model of the ␤ 3 homopentameric GABA A receptor M2 segments relative to NCA action (shown in Fig. 1). ). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 (1,3,7,14,15). Single Cys mutations were introduced for the entire 10 residues Met 11 Ј to Glu 20 Ј of the ␤ 3 M2 extracellular half. Four Ser mutations, i.e. T13ЈS, T14ЈS, N15ЈS, and T16ЈS, were generated for comparison with the corresponding Cys mutations. In addition, two Ala mutations, H17ЈA and E20ЈA, allowed further study of the sites of Zn 2ϩ interaction. It was possible that the engineered M2 Cys might interact with the single endogenous Cys (Cys 289 ) in M3. To avoid this possible complication, some dual mutations (13ЈC, 14ЈC, and 16ЈC to 20ЈC) were generated with Cys replaced by Ser at position 289 to give Cys-less M3. The human GABA A receptor ␤ 3 homopentamer and a series of mutants thereof were expressed in insect Sf9 cells. The methods for site-directed mutagenesis of M2 Ϫ4Ј to 10Ј, transfection, cell culture, and protein expression were discussed earlier (3,12). Similar protocols were followed here for single mutations to Cys, Ser, or Ala of M2 11Ј to 20Ј. Preparation of dual mutations involved a two-step PCR-based mutagenesis method, i.e. first M2 Cys was introduced as a template, and then M3 Cys was mutated to Ser. For example, the first mutational primer introduced M2 H17ЈC (CAACCATCAACACCTGC-CTTCGGGAGACCTTG), and the second primer generated the dual mutations M2 H17ЈC/M3 C289S (GACATGTACCT-TATGGGCTCCTTCGTCTTTGTGTTCCTGG) (mutated genetic code underlined). All mutations were confirmed by double-stranded DNA sequencing.

Mutagenesis, Transfection, and Expression-Alignment
Membrane Preparation, Western Blotting, and Radioligand Binding-The procedures for membrane preparation (500 ϫ g supernatant; 100,000 ϫ g pellet) and Western blotting (mouse anti-GABA A receptor ␤ 2,3 chain monoclonal antibody; Chemicon International, Temecula, CA) were described previously (3,12 Flow Cytometry Analysis-Flow cytometric analysis was used to determine receptor surface expression measured as membrane immunofluorescence (18). Sf9 cells were transfected with recombinant baculovirus for 48 h. Cells were harvested and washed with phosphate-buffered saline (PBS), and then 10 6 cells in 100 l of PBS were incubated for 30 min on ice with anti-GABA A R antibody as described above or, as a negative control, with PBS only. After washing twice with PBS, Alexa Fluor 488 goat anti-mouse IgG (1:200) (Invitrogen Molecular Probe) was added to each sample on ice for 30 min. Following three washes with PBS, the fluorescence density was measured using a Coulter Elite instrument (Beckman-Coulter) and analyzed with WinMDI 2.8 software provided by Duke University The pore-facing residues are considered to be Ϫ1Ј, 2Ј, 6Ј, 9Ј, 13Ј, 16Ј, 18Ј, and 20Ј. The ␣-helical structure of Ϫ1Ј to 16Ј (red) becomes a more ␤-like conformation for 17Ј to 20Ј (blue). Dotted lines represent cross-linking of M2 14Ј, 17Ј, and 19Ј with M3 Cys 289 . EBOB binds primarily at 2Ј and 6Ј, with enhancement by 9Ј, largely below the narrower 9Ј to 14Ј gating region. Zn 2ϩ binds at His 17 Ј, and anesthetics and alcohol interact with Asn 15 Ј. LGICs. The species considered are human for GABA A R and rat for glycine receptor (GlyR) ␣ 1, serotonin type 3 A receptor (5-HT3 A R), and nAChR ␣ 1 . The subunit-specific number is given at the left for the first residue of each aligned sequence (7). Index numbers (see Footnote 3) for positioning in M2 are shown at the top. Pore-facing residues of GABA A R ␣ 1 and nAChR ␣ 1 subunits are shown in bold type (14,15). In GABA A R ␤ 3 the lightly shaded residues are proposed to face the channel lumen and the dark-shaded residues to form M2/M3 disulfide bonds (Ref. 3 and this study). In the nAChR, the boxed area (9Ј to 14Ј) is the narrowest region of the channel pore identified by cryo-electron microscopy (1). In the GABA A R M3 region, the conserved endogenous Cys, indicated by an asterisk, is boxed for the ␤ 3 Cys 289 position, which was mutated to Ser in the dual mutants.
(Durham, NC). The expression of ␤ 3 homomer was evaluated as the mean fluorescence value.
Disulfide Cross-linking-Disulfide cross-linking profiles were determined for the single and dual mutants comparing spontaneous, oxidative, and reductive conditions (3,7,9). CuSO 4 was prepared as a 100 mM stock solution in water and o-phenanthroline (Sigma) as a 200 mM stock solution in ethanol. CuSO 4 and o-phenanthroline were freshly mixed to obtain the Cu:phen solution, which was added to samples to give a final concentration of 100:200 M. For oxidation, membrane protein (100 g) was incubated with Cu:phen for 5 min. Then the reaction was terminated by adding 10 mM N-ethylmaleimide and 1 mM EDTA. Following this treatment sequence, sample buffer (3) was added with or without 10 mM dithiothreitol (DTT). Finally, the cross-linked subunits were detected by immunoblotting. The results were compared with control samples without any treatment.
DTT Reduction, Cu:phen Oxidation, and NCA Binding-Membranes as described above were subjected to reduction and oxidation to determine the effect of the disulfide on receptor NCA binding activity. A similar procedure was used by others for cells with electrophysiology assay (7,9). Samples were incubated with 10 mM DTT for 15 min or 100:200 M Cu:phen for 5 min at room temperature. Samples with Cu:phen were also treated with 1 mM EDTA following oxidation to chelate the Cu 2ϩ in solution. Controls were samples without any treatment. Finally, each sample was subjected to the [ 3 H]EBOB binding assay. Activity values were expressed as percent of wild type (WT) binding level.
Sulfhydryl Modification and NCA Binding-2-Aminoethylmethanethiosulfonate hydrochloride (MTSEA ϩ ) (Toronto Research Chemical, North York, Canada) was used under previously described conditions (19) with analysis for its effect on [ 3 H]EBOB binding. Briefly, selected Cys mutants (more than 30% WT binding level) and WT were labeled for 10 min at room temperature with 2.5 mM MTSEA ϩ . Following incubation, each sample was centrifuged at 20,000 ϫ g for 15 min and the supernatant removed completely. The pellet was resuspended in PBS and subjected to a binding assay.
Molecular Modeling-Disulfide cross-linking interactions were examined with the ␤ 3 homopentamer model based on the homologous nAChR with ␣-helical structure throughout the transmembrane segments (3,20). All modeling was done with Maestro 6.5 (Schrödinger LLC). The M2 residues from 11Ј to 20Ј were mutated to Cys and the distances measured between sulfur centers.

RESULTS
Mutagenesis, Transfection, and Expression-PCR analysis indicated a recombination efficiency of nearly 100% for all mutants (Fig. 3A) with their identities confirmed by sequencing. Western blotting established similar expression levels for each mutant (Fig. 3B). A faint dimer from T13ЈC was resistant to reduction. Similar transfection efficiencies and protein expression levels were also observed for dual M2 and M3 mutations, i.e. T13ЈC, I14ЈC, T16ЈC, H17ЈC, L18ЈC, R19ЈC, and E20ЈC with C289S (data not shown). Flow cytometry was used to further characterize protein localization and quantity with special respect to mutant T13ЈS (with little activity for NCA binding, as considered later) using non-fixed, transfected Sf9 cells (Fig. 3C). Mean fluorescence values were similar for the WT and mutant T13ЈS, i.e. values of 16.1 and 16.4, respectively. This result shows strong ␤ 3 homopentamer surface expression, i.e. the mutation did not disturb the subcellular location of the homomer.
Disulfide Cross-linking Profiles -The 10 single Cys mutants (11Ј to 20Ј) were exposed to ambient oxygen in the presence and absence of Cu:phen catalyst (Fig. 4). Mutants forming disulfide bonds were evident by conversion of some or most of the band at ϳ55 to ϳ130 kDa based on Western blot analysis. The WT (3) and the M11ЈC mutant did not form any disulfide bonds. There were no dimers formed at positions 12Ј and 15Ј without Cu:phen, but extremely weak dimers were induced by Cu:phen. T16ЈC formed a strong dimer with the oxidant. Five positions (T13ЈC, H17ЈC, R18ЈC, L19ЈC, and E20ЈC) formed   DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50 disulfides spontaneously in the absence of Cu:phen, and there was some evidence of low spontaneous disulfide formation with I14ЈC. Cu:phen strongly increased the amount of dimer at these six positions, especially from 17Ј to 20Ј, where only a small amount of monomer remained. The dramatic increase in dimer formation at 17Ј to 20Ј indicated that intersubunit disulfide bonds dominated compared with possible intrasubunit M2/M3 disulfides, which would make the free Cys less available. The effect of DTT reduction on dimer yield varied with the mutants. All of the dimers at about 130 kDa were almost completely reduced by DTT except for T13ЈC and E20ЈC with faint resistant bands. A similar observation has been noticed in the 6Ј position (3,9).

GABA A R M2 Mobility and NCA Action
To determine M2/M2 versus possible M2/M3 intersubunit disulfide cross-bridging, dual mutants were generated by combining M3 C289S with M2 Cys mutants 13Ј, 14Ј, and 16Ј to 20Ј (Fig. 4). The selection of mutants was based on the spontaneous or induced disulfide formation and spatiality to M3 Cys 289 based on the GABA A R model (3). Dual M2 mutants 13Ј, 16Ј, 18Ј, and 20Ј showed strong spontaneous or oxidant-induced disulfide bond formation similar to that of the corresponding single Cys mutants establishing M2/M2 cross-linking. Spontaneous disulfide bond formation was greatly reduced in the dual Cys mutations 17Ј and 19Ј, and the slight spontaneous dimer at single Cys mutant I14ЈC was also diminished indicating that spontaneous disulfide bond cross-linking by these M2 single Cys mutants largely involves M3 Cys 289 . However, for unknown reasons, oxidation also led to extensive protein loss at these three positions.
Effect of Site-specific Mutations on NCA Binding-The sitespecific mutations had very different effects on NCA binding. whereas Ser at 14Ј and 16Ј greatly enhanced activity relative to Cys. The seven M2/M3 dual mutations showed different patterns compared with their single M2 Cys counterparts. The binding activity of T13ЈC/C289S was the same as the single mutants T13ЈC and T13ЈS. Dual mutations I14ЈC/C289S and T16ЈC/C289S did not give the high activity of the corresponding Ser mutations I14ЈS and T16ЈS. On the other hand, H17ЈC/ C289S and R19ЈC/C289S showed much better binding activity (60 -70%) than the corresponding single Cys mutants (20 -35%).
Effect of DTT Reduction and Cu:phen Oxidation on NCA Binding-Some single Cys mutants undergo partial spontaneous disulfide formation, which could affect [ 3 H]EBOB binding and might be reversed with DTT or increased with Cu:phen. Any of these structural changes might either have no effect or alter [ 3 H]EBOB binding. These relationships were further examined with the five single Cys mutants that undergo high spontaneous disulfide bond formation, i.e. 13Ј, 17Ј, 18Ј, 19Ј, and 20Ј (Fig. 6); all other Cys mutants showed no significant increase in binding activity with treatment (data not shown). DTT at 10 mM had essentially no effect or slightly decreased the binding level of [ 3 H]EBOB with WT, L18ЈC, and E20ЈC. T13ЈC with DTT showed an increased but still low level of binding. In contrast, mutants H17ЈC and R19ЈC showed greatly enhanced binding activity on the addition of DTT, with recovery to near the WT level. To check the effect of the oxidant, single Cys mutants 13Ј and 17Ј to 20Ј were treated with Cu:phen and EDTA. Before the addition of EDTA, Cu:phen inhibited the binding activity of WT, H17ЈC, L18ЈC, and E20ЈC about 15-40% but only slightly affected the binding of T13ЈC and R19ЈC. After the addition of EDTA to chelate the Cu 2ϩ and withdraw the oxidative effect of Cu:phen, all of them returned almost to the binding levels without treatment. The stronger oxidative environment with Cu:phen did not reduce the binding level of these Cys mutants. DTT did not recover any significant activity for the seven dual mutants (data not shown).
Effect of Sulfhydryl Modification on NCA Binding of Single Cys Mutants-Membrane-impermeable MTSEA ϩ was used to explore the accessibility of M2 single Cys mutants to sulfhydryl modification. Mutants T13ЈC, I14ЈC, T16ЈC, and R19ЈC with less than 30% of WT [ 3 H]EBOB binding activity could not be  assessed accurately, but all others were examined (Fig. 7). MTSEA ϩ inhibited [ 3 H]EBOB binding activity for the WT by 20%, but for mutants T12ЈC, N15ЈC, and L18ЈC the inhibition was more than 70%. The low inhibition (10 -30%) for H17ЈC and E20ЈC may come from the strong, spontaneous, disulfide bond formation, but L18ЈC still could be modified by MTSEA ϩ , indicating the greater availability of its free thiol moiety.
Effect of Zn 2ϩ on NCA Binding-[ 3 H]EBOB binding in the WT is very sensitive to Zn 2ϩ (IC 50 1.3 M) (Fig. 8)

DISCUSSION
Proposed M2 Structure of Spontaneously Open ␤ 3 Homopentameric Channel with High NCA Sensitivity-Findings are summarized in Table 1 and illustrated in Fig. 1 as a model for the proposed M2 structure and interaction sites. A detailed electrophysiological characterization of the functional ␤ 3 homopentameric GABA A R has shown that this subunit forms a spontaneously open channel, which can be inhibited by picrotoxin (PTX) and Zn 2ϩ and is insensitive to GABA (11,21). The cytoplasmic end of the channel may act as a fixed fulcrum to open the narrow domain of the pore between 9Ј and 14Ј (1,8,22) with little change in conformation during gating (23). Low mobility and limited accessibility from 2Ј to 6Ј are evident for the ␤ 1/3 subunit (3,13). Disulfide cross-linking at the 9Ј position supports its role as an important component of the channel gate (3). The current investigation defines how NCAs access their binding site at the cytoplasmic end by studying the conformational flexibility of the M2 extracellular half.
Important Role of 11Ј to 16Ј Region Relative to the Channel Gate -The ease of disulfide bridging depends on the average distances of Cys residues, the orientation of thiol moieties, and the mobility/flexibility of the protein region (9,24). Positions M11ЈC, T12ЈC, and N15ЈC, with little or no oxidant-induced disulfide formation, are probably not exposed in the channel lumen, although T12ЈC and N15ЈC can be accessed by sulfhydryl modification reagents (13). T13ЈC is the only position in the Ϫ1Ј to 16Ј region that spontaneously forms a strong disulfide bond, implying its flexibility in rapid opening and closing of the gate and/or the proximity of its thiol substituents in the narrowest region of the pore. The sulfurs in a disulfide bond are    separated, center-to-center, by 2 Å (24), and those for T13ЈC on adjacent subunits are only 6.8 Å apart (Fig. 9) consistent with their spontaneous formation of a strong disulfide cross-link. Disulfide bonds also form spontaneously between ␥ 2 13ЈC and ␣ 1 13ЈC (8) but not at ␤ subunits (9). I14ЈC of the ␤ 3 homopentamer gives very weak spontaneous disulfides associated with little M2 I14ЈC/M3 C289 cross-linking because of the considerably greater distance and unaligned orientation from M3 Cys 289 . Unaligned disulfides are also observed between ␥ 2 14ЈC and both ␣ 1 15ЈC and ␣ 1 16ЈC (8). Asn 15 Ј is not in the pore lumen and is important in alcohol or general anesthetic action (25). The induced but not spontaneous disulfide bond formation observed here with the ␤ 3 homopentamer places Thr 16 Ј in the pore lumen consistent with a previous SCAM study (14), suggesting its relatively low mobility and longer sulfur-sulfur distance on adjacent M2, measured as 10.0 Å (Fig. 9).
The M2 and M3 cross-links apparently yield only dimers but not trimers, quatermers, or pentamers. This finding is consistent with the observation that endogenous Cys 289 , both in WT (3) and M2 mutant M11ЈC (Fig. 4), does not form a spontaneous or Cu:phen-induced dimer or multimer, indicating that M3 Cys 289 will not cross-link with itself. The major product from the mutated Cys in M2 and endogenous M3 Cys 289 cross-linking is a dimer that probably "locks" the channel in an unfavorable conformation preventing further cross-linking to form multimers.
Conformational Mobility of 17Ј to 20Ј Region-The nAChRbased GABA A R model with an ␣-helical M2 structure places 17Ј and 20Ј in pore-facing positions (14,20) and gives average M2/M2 sulfur-sulfur distances for 17Ј, 18Ј, 19Ј, and 20Ј of 6.7, 13.6, 16.3, and 9.3 Å, respectively (Fig. 9). The pore-facing residues are spatially preferred to form disulfide bonds with relatively small movement asymmetrically (9), but this movement would rotate the non-pore-facing residues like 18Ј and 19Ј away from each other. However, all single Cys mutants in this region of the ␤ 3 homopentamer form disulfide bonds spontaneously, to a similar extent. The strong and newly observed spontaneous disulfide formations at L18ЈC and R19ЈC are therefore not consistent with the helical conformation. The ␤ 1 E20ЈC even forms a nonadjacent disulfide bond in the ␣ 1 ␤ 1 ␥ 2 receptor, suggesting a possible structural difference from the nAChR in this region (10). The M2 H17ЈC/M3 Cys 289 and M2 R19ЈC/M3 Cys 289 intersubunit sulfur-sulfur distances are quite long, ranging from 18 to 27 Å (Fig. 10). The observed M2/M3 intersubunit bridging by H17ЈC and R19ЈC further indicates that they probably trap in different directions from L18ЈC and E20ЈC and face out of the channel lumen (Fig. 1). Their cross-linking would be greatly facilitated by a ␤-like conformation, allowing closer proximity to the endogeneous M3 Cys.
We propose that in the ␤ 3 homopentamer open channel, the 17Ј to 20Ј region assumes a ␤-like conformation (Fig. 1) based on the disulfide formation profile. The extended deconformation structure, directly connected with the channel activation domain M2/M3 loop (22,26) (Fig. 10), may help to open or widen the channel gate. GABA activation alters the access of a sulfhydryl modification reagent in the 17Ј to 20Ј region (13) implying a conformational change that relies on channel functional states. The nAChR ␦ subunit extracellular part of M2 undergoes a pronounced change between closed and open conformations (27). Significant backbone deformation occurs at positions 13Ј, 16Ј, and 19Ј in nAChR ␣ subunit M2 during channel gating (28). If LGICs in the closed state assume a well ordered ␣-helix structure (1), then the ␤ 3 homopentamer may have undergone transition to a ␤-like conformation in this region on channel opening. The proposed GABA A R ␤-like conformation refers to the ␤ subunit only, because it is distinctive in flexibility and channel function relative to the ␣ or ␥ subunit (10,29,30).
NCA Interactions with the Extracellular Half-The nonpore-facing M11ЈC, T12ЈC, and N15ЈC had little effect on NCA binding and were not considered further here. Mutants T13ЈC and T13ЈS in the ␤ 3 homopentamer showed dramatically FIGURE 9. Sulfur-sulfur distances of Cys-mutated ␤ 3 homopentameric GABA A R based on an ␣-helical frame. M2/M2 average distances are shown for 13Ј to 20Ј with the native side chains mutated to Cys using Maestro 6.5. reduced NCA binding. ␤ 1 T13ЈA in an ␣ 1 ␤ 1 receptor does not respond to GABA and loses muscimol binding, suggesting its essential role in receptor function (29), possibly by disturbing the channel gate. L9Ј mutations change both the channel gating and NCA sensitivity (31), further indicating that NCA action is closely related with the gate domain. Cys mutations at Ile 14 Ј and Thr 16 Ј cause more damage than Ser on NCA binding, and DTT does not reverse the low binding activity; possibly the increased size of the Cys sulfur versus the Ser oxygen impedes NCA access to its binding site. With an ␣ 1 ␤ 1 ␥ 2 receptor, a thiol-reactive probe related to the NCA fipronil irreversibly inhibits GABAinduced chloride current at a Cys-mutated 17Ј position of the ␣ 1 subunit (6) . However, ␤ 3 H17ЈA in the present study did not change the binding activity. Zn 2ϩ binding affinity was sharply reduced by H17ЈA, consistent with earlier studies (5,21,32). The inhibition of [ 3 H]EBOB binding by Zn 2ϩ confirms M2 as the NCA target, but it is not clear whether the block is a direct or an allosteric effect (33).
The 17Ј to 20Ј region is particularly interesting relative to NCA binding. With the ␤ 3 homopentamer, DTT reverses the low activity of single Cys mutants H17ЈC and R19ЈC but does not affect L18ЈC and E20ЈC, which have original high binding levels ( Fig. 6), although these four single Cys mutants have similar levels of disulfide bond formation (Fig. 4). In contrast, disulfide formation at H17ЈC in the ␤ 1 subunit of the ␣ 1 ␤ 1 receptor is much less damaging on Cl Ϫ conductance than that at E20ЈC (9). The blocking effect from M2/M3 intersubunit disulfide cross-linking by 17ЈC/Cys 289 and 19ЈC/Cys 289 indicates that the NCAs may prefer the exposed water cavity of 17Ј and 19Ј as an access pathway.
NCA Access to the Binding Site-The NCA PTX has been proposed to access its binding site either through open channels or with different efficacies through closed channels (34,35). The present study suggests two pathways for NCA access to its ␤ 3 homopentamer binding site at the cytoplasmic end of M2 (Fig. 10). One is directly through the pore and another one possibly through the water cavities between adjacent subunits, similar to the suggested general anesthetics binding pocket in the water crevice behind M2 (25). The M2 segments are well shielded from the lipid bilayer by an outer ring of the helices M1, M3, and M4 to form channel pore and water cavities around M2 (1) (Fig. 10). There is precedent for these proposals because access to the ligand binding site of G-protein-coupled receptors is via a water-filled crevice (36 -39). The more cytoplasmic residues in M3 become available for modification by sulfhydryl reagents on GABA activation (40,41). The challenge is to define how the NCA can reach its binding site at the cytoplasmic end of M2 through this novel pathway. As an alternative interpretation, M2/M3 disulfide formation from 17ЈC/ Cys289 and 19ЈC/Cys289 may lock the channel pore in an unfavorable conformation for NCA binding. Similarly, disulfide bond formation at M2/M2 intersubunits or M2/M3 intrasubunit disrupts the channel conformation to alter the Cl Ϫ current (7-10). Although we cannot rule out this possibility, our new pathway suggestion has important implications for NCA access.
GABA A receptor noncompetitive antagonism is a complex phenomenon (34,35,42,43). PTX can reach its site through both hydrophilic and hydrophobic pathways (34). It is suggested to bind at an allosteric site to stabilize a closed or desensitized state of the channel (35) or to bind to both use-dependent and use-independent sites in GABA A receptors (43). The overall inference is that channel opening is not an absolute requirement for PTX to reach its site. The complexity of NCA antagonism probably is related in part to the conformational mobility of the M2 extracellular half and multiple access pathways to the binding site.