Loose Protein Packing around the Extracellular Half of the GABA A Receptor (cid:1) 1 Subunit M2 Channel-lining Segment*

GABA A receptors are ligand-gated ion channels formed by the pseudosymmetrical assembly of five homologous subunits around the central channel axis. The five M2 membrane-spanning segments largely line the channel. In the present work we probed the water surface accessibility of the (cid:1) 1 subunit M2 segment us- ing the substituted cysteine accessibility method. We assayed the reaction of the negatively charged sulfhy-dryl-specific reagent, p- chloromercuribenzenesulfonate (pCMBS (cid:2) ), by its effect on subsequent currents elicited by EC 50 and saturating GABA concentrations. pCMBS (cid:2) , applied with GABA, reacted with 14 of the 19 residues tested. At the M2 cytoplasmic end from 2 (cid:1) to 6 (cid:1) only (cid:1) 1 A252C (2 (cid:1) ) and (cid:1) 1 T256C (6 (cid:1) ) were pCMBS (cid:2) -reactive in the presence of GABA. We infer that the M2 segments are tightly packed in this region. Toward the extracellular half of M2 all residues from (cid:1) 1 T262C (12 (cid:1) ) through (cid:1) 1 E270C (20 (cid:1) ) reacted with pCMBS (cid:2) applied with GABA. We infer that this region is highly mobile and loosely packed against the rest of the protein. Based on differences in pCMBS (cid:2) reaction rates two domains can be distinguished on the putative channel-lining side

tors the most common subunit stoichiometry is 2␣, 2␤, and 1␥ subunit (6 -9). Each subunit has an ϳ200-amino acid, extracellular, N-terminal domain and a similar sized C-terminal domain with four ␣ helical, membrane-spanning segments (M1, M2, M3, M4) (10,11). GABA binding at two sites in the extracellular domain at the ␤-␣ subunit interfaces (12)(13)(14) induces the transient opening of a transmembrane, anion-selective channel that is an integral part of the receptor protein. The transmembrane ion channel is largely lined by the M2 segments of each subunit (11,15). Recent x-ray crystallographic and cryoelectron microscopic studies have provided new insights into the structure of these receptors. The x-ray crystal structure of a snail ACh-binding protein that is homologous to the extracellular domain provides a foundation for structural studies of the residues forming the agonist binding sites (14, 16 -19). The 4 Å resolution structure of the homologous Torpedo ACh receptor membrane-spanning domain solved by cryoelectron microscopy of two-dimensional crystals provides a view of the closed state of the transmembrane, channel-forming domain (11). This latter study showed that the five M2 segments form an inner ring of helices that line the closed channel. The M2 segments appear to be loosely packed against the outer ring of helices formed by the M1, M3, and M4 segments.
To probe the protein packing around the M2 segment we assayed the accessibility of cysteines (Cys) substituted for the GABA A receptor ␤ 1 subunit M2 segment residues to react with the sulfhydryl-specific reagent, p-chloromercuribenzenesulfonate (pCMBS Ϫ ). pCMBS Ϫ is a rigid, negatively charged molecule that will fit into a right cylinder that is ϳ6 Å in diameter and ϳ10 Å in length. pCMBS Ϫ reacts with Cys and covalently couples HgC 6 H 4 SO 3 Ϫ onto the sulfhydryl. pCMBS Ϫ is membrane-impermeant (20,21) and reacts 1,000 times faster with ionized thiolates (-S Ϫ ) than with thiols (-SH) (22); thus, a reaction is much more likely with water-accessible Cys that can ionize. To facilitate comparisons with the residues in the M2 segments of other superfamily members an index numbering system is used (23) where the absolutely conserved positively charged residue at the cytoplasmic end of M2, ␤ 1 Arg 251 , is 0Ј, and the residue aligned with the extracellular ring of charge (24), ␤ 1 Glu 270 , is 20Ј (Fig. 1). In the substituted-cysteine-accessibility method (SCAM) reaction of sulfhydryl reagents with engineered Cys is usually assayed by their effects on subsequent functional responses of the channel (25). Previous substituted cysteine accessibility studies of the GABA A receptor ␣ 1 and ACh receptor ␣ subunit were performed using saturating concentrations of agonist for the test pulses before and after application of the sulfhydryl reagents (15,26). It has become apparent that modification at some positions alters currents elicited by EC 50 concentrations of agonist but not saturating concentrations (27)(28)(29)(30). This presumably arises because covalent modification at these positions alters channel gating but does not alter single channel conductance or maximal open probability. Alternatively, it may arise from counteracting/ compensating effects on gating/desensitization and conductance. Thus, screening with only saturating agonist test pulses results in failure to identify all of the reactive Cys.
In the present study we screened the reactivity of pCMBS Ϫ with Cys replacing the GABA A receptor ␤ 1 subunit M2 segment residues from ␤ 1 Ala 252 (2Ј) to ␤ 1 Glu 270 (20Ј) using both EC 50 and saturating GABA test pulses. We found that pCMBS Ϫ reacted with Cys substituted for all residues between the 12Ј and 20Ј levels. This implies that all of the residues are at least transiently water-accessible. Toward the cytoplasmic end of the channel reactivity is much more limited, suggesting tighter protein packing around the cytoplasmic end of the channel.

EXPERIMENTAL PROCEDURES
Mutagenesis and Oocyte Expression-PCR mutagenesis of the rat ␤ 1 subunit in the pGEMHE vector, in vitro mRNA transcription, and preparation and injection of Xenopus oocytes were performed as described previously (27). All mutants were sequenced to confirm mutation and rule out secondary mutations. For certain mutants, cDNA was prepared from mRNA by reverse transcription-PCR (One-Step RT-PCR Kit, Qiagen, Valencia, CA) and sequenced separately. The ␤ 1 Cys mutants were coexpressed with wild type ␣ 1 and ␥ 2S subunits by injecting oocytes with 50 nl of RNA (200 pg/nl) mixed in a ratio of 2:1:2 ␣ 1 :␤ 1 :␥ 2 to discourage the formation of ␤ 1 homopentamers. Oocytes were incubated at 16 -17°C for 2-3 days before use (31).
Electrophysiology-Currents were recorded under two-electrode voltage clamp from oocytes continuously perfused at 5 ml/min with Ca 2ϩfree Frog Ringer buffer (115 mM NaCl, 2.5 mM KCl, 1.8 mM MgCl 2 , 10 mM HEPES pH 7.5 with NaOH) using equipment and procedures described previously (27). The perfusion chamber volume was ϳ200 l. Agarose cushion electrodes (32) were filled with 3 M KCl and had a resistance of less than 2 megohms. The ground electrode was connected to the bath by a 3 M KCl/agar bridge. The holding potential was maintained at Ϫ45 mV unless otherwise specified.
To determine EC 50 and maximal GABA concentrations, GABA concentration-response relationships were fitted to the Hill equation, where I max is the maximal peak current, [A] is the agonist concentration, and n is the Hill coefficient, using Prism3 software (GraphPad, San Diego) as we have done previously (33).
To determine the irreversible effects of pCMBS Ϫ on the GABAinduced currents, test pulses of GABA were applied at EC 50 and saturating concentrations before and after application of pCMBS Ϫ . The following series of GABA applications preceded and followed the pCMBS Ϫ application: 10 ϫ EC 50 GABA, 20 s; 10 ϫ EC 50 GABA, 20 s; EC 50 GABA, 20 s; EC 50 GABA, 20 s. To ensure stability of the responses, if paired current responses either before or after pCMBS Ϫ were not within 10% of each other the oocyte was not used. The GABA applications were separated by 3-5-min washes with Ca 2ϩ -free Frog Ringer buffer to allow complete recovery from desensitization. The percent effect was determined by %effect ϭ ͑͑I GABA, after /I GABA, before ͒ Ϫ 1͒*100 (Eq. 2) where I GABA, after is the average peak current of the GABA test pulses after pCMBS Ϫ application, and I GABA, before is the average peak current of the initial GABA applications. Data are given as means Ϯ S.E. For screening, pCMBS Ϫ was applied for 1 min at 0.5 mM either in the absence or in the presence of a saturating concentration of GABA. This combination of time and concentration was chosen because they were the maximal concentration and duration that caused no significant increase in the leak conductance of uninjected Xenopus oocytes. This limits our ability to detect reactive residues. As discussed below, given the variability of responses, application of a reagent must cause a net change in current greater than ϳ30% to be significantly different from wild type by a one-way analysis of variance (ANOVA) (for n between 3 and 6). With this threshold and the pCMBS Ϫ reaction conditions (0.5 mM applied for 1 min), if complete reaction caused 100% inhibition of the GABA-induced current we would detect reactive positions with a second order reaction rate Ͼ12 liters/mol/s. Measurement of Reaction Rates-pCMBS Ϫ reaction rates were determined by the effect of sequential brief applications of pCMBS Ϫ as we have done previously (29,30). A test pulse of GABA was applied to measure the GABA-induced current. pCMBS Ϫ (2 M-0.5 mM) Ϯ GABA was applied for 15-60 s. Following washout a GABA test pulse was applied, and the induced current was measured. The effect of five to eight brief, sequential applications of pCMBS Ϫ Ϯ GABA were determined. The magnitudes of the GABA test currents were normalized relative to the initial test current. The normalized current was plotted as a function of the cumulative duration of pCMBS Ϫ application and fitted with a single exponential function using Prism3 software (Graph-Pad). The second order rate constant was calculated by dividing the pseudo-first order rate constant obtained from the exponential fit by the pCMBS Ϫ concentration. For a given mutant, second order rate constants were independent of the pCMBS Ϫ concentration used. For two Cys mutants, S265C and R269C, reaction rates were calculated by fitting an exponential function to the pCMBS Ϫ -activated inward current induced in these mutants.
Statistics-Data are expressed as the percent change of current after modification Ϯ S.E. The significance of differences between each mutant and wild type was determined by one way ANOVA using the Dunnett's post hoc test (Prism3). Reaction rates are given as the mean Ϯ S.E.

RESULTS
Characterization of the Mutants-All of the ␤ 1 subunit Cys mutants were expressed with wild type ␣ 1 and ␥ 2S subunits in Xenopus oocytes. GABA-induced currents were recorded from oocytes expressing each of the Cys mutants. For wild type receptors the maximum GABA-induced currents were 1,163 Ϯ 181 nA (n ϭ 6). For the mutants the maximum currents ranged from 118 Ϯ 27 nA (n ϭ 6) for ␤ 1 V258C (8Ј) to 1,454 Ϯ 202 nA (n ϭ 6) for ␤ 1 G254C (4Ј). Oocytes expressing ␣ 1 ␤ 1 L259C␥ 2S (9Ј) had a large holding current in the absence of GABA. The ␤ 1 L259C channels have a high spontaneous open probability and significant decrease in agonist EC 50 as has been reported for 9Ј mutations in this and other gene superfamily members (34 -41). For wild type ␣ 1 ␤ 1 ␥ 2S receptors the GABA EC 50 was 38 Ϯ 7 M (n ϭ 4) ( Table I). The GABA EC 50 for the mutants ranged from 1. (n ϭ 3) for ␤ 1 T260C (10Ј). The GABA EC 50 was significantly reduced in a number of the Cys mutants (Table I).
Effect of pCMBS Ϫ Applied in the Presence of GABA-In the presence of saturating GABA concentrations receptors undergo fluctuations between the open and desensitized states (42)(43)(44). Thus, during pCMBS Ϫ application in the presence of GABA the channels occupy multiple conformational states, and we cannot distinguish whether the reaction is occurring in one or more of these states.
A 1-min application of 0.5 mM pCMBS Ϫ in the presence of 100 M GABA irreversibly altered the subsequent GABA-induced test currents in 14 of the 19 mutants tested but had no effect on wild type ( Fig. 2). At eight of the positions (T256C, V258C, L259C, T260C, T262C, I264C, S265C, and E270C) the reaction significantly altered both EC 50 and saturating test pulses. At three positions (A252C, T266C, and L268C) only currents elicited by saturating GABA test pulses were significantly altered, and at three other positions (T257C, H267C, and R269C) only currents elicited by EC 50 GABA test pulses were significantly altered. At four positions (T257C, H267C, R269C, and E270C) a reaction with the negatively charged pCMBS Ϫ caused potentiation of the subsequent GABA-induced currents. In contrast, at the other 10 positions covalent modification by pCMBS Ϫ resulted in inhibition of the subsequent GABA-induced currents.
In addition to the effects on subsequent GABA currents, pCMBS Ϫ application significantly increased the subsequent holding current for T260C. This presumably occurred because of an increase in spontaneous opening probability following covalent modification of the Cys.
At the other five positions (L253C, G254C, I255C, M261C, and T263C) pCMBS Ϫ application did not significantly alter the subsequent GABA-induced currents or the holding current. We cannot distinguish whether these positions are not reactive with pCMBS Ϫ or whether reaction at these positions had no functional effect on the macroscopic currents elicited by EC 50 and saturating GABA. We were surprised that pCMBS Ϫ application to T263C had no functional effect. In all other Cys-loop receptor subunits tested the 13Ј position reacted with sulfhydryl reagents (15,26,28,30,45). To ensure that the T263C mutant was correct we have sequenced both the plasmid DNA and the product of a reverse transcription-PCR on the in vitro transcribed mRNA. Both confirm that the Cys mutant is correct (data not shown). Application of a larger sulfhydryl-reac-tive methanethiosulfonate (MTS) derivative, MTS-benzophenone, inhibited subsequent currents in oocytes expressing the T263C mutant (data not shown). Thus, we infer that T263C is a sulfhydryl-reactive position but that reaction with pCMBS Ϫ had no functional effect.
pCMBS Ϫ Reaction Rates in the Presence of GABA-The reaction rates of pCMBS Ϫ applied in the presence of GABA with the engineered Cys mutants were measured as illustrated in Fig. 3. The reaction rates are summarized in Table II. The reaction rates ranged between 40 Ϯ 10 M Ϫ1 s Ϫ1 with ␤ 1 T257C (7Ј) and 31,978 Ϯ 12,251 M Ϫ1 s Ϫ1 with ␤ 1 T262C (12Ј). To investigate whether the negative charge at the 20Ј position (␤ 1 Glu 270 ) affected the pCMBS Ϫ reaction rate with the Cys substituted at the adjacent 17Ј position, ␤ 1 H267C (Table II), we constructed a double mutant ␤ 1 H267C/E270N. Removal of the negative charge at the 20Ј position did not significantly alter the pCMBS Ϫ reaction rate with the 17Ј Cys (data not shown).
We were unable to measure the pCMBS Ϫ reaction rate with ␤ 1 L268C because it produced biphasic effects at low concentrations (data not shown). Initially the GABA test responses were potentiated, and after further pCMBS Ϫ application they were inhibited. Because there are two ␤ subunits in each functional receptor we infer that for the L268C mutant the reaction rates with the two Cys were different leading to the biphasic effects.
Effect of pCMBS Ϫ Applied in the Absence of GABA-Ideally, by determining whether pCMBS Ϫ reacted in the absence of GABA we would be determining accessibility in the closed state of the receptor. In the absence of GABA, the wild type receptors are largely in the closed state because of their very low spontaneous open probability. The spontaneous open probability of the Cys mutants, however, is not known. For all of the mutants except ␤ 1 L259C the initial holding current was not significantly greater than the current from oocytes expressing wild type receptors.
A 1-min application of 0.5 mM pCMBS Ϫ in the absence of GABA irreversibly altered the subsequent GABA currents of eight of the 18 Cys mutants (Fig. 4). At four positions (V258C, T262C, I264C, and L268C) both EC 50 and saturating test pulses were altered. At two positions (T266C and R269C) only the subsequent saturating GABA currents were altered, and at two other positions (T256C, S265C) only the subsequent EC 50 GABA currents were significantly altered. The pCMBS Ϫ reaction at all of these positions caused inhibition of the subsequent GABA-induced currents. At R269C there was a significant increase in the holding current after pCMBS Ϫ application in the absence of GABA. This was presumably because of an increase in the spontaneous open probability following covalent modification. This was not observed when R269C was modified by pCMBS Ϫ in the presence of GABA.

DISCUSSION
In the presence of GABA, pCMBS Ϫ irreversibly altered the subsequent GABA-induced currents of 14 of the 19 engineered ␤ 1 M2 Cys mutants (Fig. 5). We infer that the Cys at these positions were covalently modified by pCMBS Ϫ . The ability of pCMBS Ϫ to react with these Cys implies that they are, at least transiently, on the water-accessible protein surface. Furthermore, the access pathway from bulk solution to each of these Cys must be sufficiently large for pCMBS Ϫ , with dimensions ϳ6 Å in diameter and ϳ10 Å in length, to pass through.
On casual inspection, the pattern of accessibility in ␤ 1 M2 is quite different from the pattern observed with the GABA A receptor ␣ 1 M2 segment (15,26). It should be noted that the ␣ 1 M2 segment was screened using only saturating concentrations of GABA test pulses. Had we only used saturating GABA concentrations in the current experiments we would have identified only 11 sulfhydryl-reactive positions, two more than the nine sulfhydryl-reactive positions in the ␣ 1 M2 segment (15). Thus, screening with both saturating and EC 50 agonist concentrations provides the highest probability that all reactive positions will be identified.
The recently solved 4 Å resolution structure of the homologous Torpedo ACh receptor demonstrates conclusively that the M2 segments are ␣ helical as had previously been inferred (11,15,26,46,47). Thus, the fact that the water-accessible surface of M2 extends around the entire circumference of the helix over most of the length of M2 implies either that the M2 segments are loosely packed against the outer ring of helices and/or that the M2 segments are highly mobile and rotating so that residues on most of M2 are transiently water-exposed. We believe that the explanation for the extent of the water-accessible surface of M2 is a combination of these two. The former explanation, loose packing, is consistent with the loose packing of M2 observed in the ACh receptor structure (11). Because of the loose packing, water can presumably intercalate into the space between the M2 segments, in the inner ring of helices, and the M1, M3, and M4 segments that form the outer ring of helices. In addition, our previous studies using disulfide trapping indicated that the 17Ј-20Ј region of M2 is highly mobile in the resting state (31). The ability to form disulfide bonds between engineered channel-lining Cys residues implies that this is a region of high thermal protein mobility which allows both rotational and translational movement to bring aligned, channellining positions into close proximity. This high degree of thermal mobility is also consistent with loose protein packing.
In other Cys-loop receptor subunits studied by SCAM similar patterns of circumferential accessibility have been observed in the 12Ј-20Ј region when they were screened with both EC 50 and saturating agonist test pulses (Fig. 6A) (15,26,28,30,45). This implies that there is significant mobility/flexibility in the M2 segments of these members of the gene superfamily as well.
Near the cytoplasmic end of the channel from 2Ј to 6Ј the water-accessible surface of M2 appears to be more limited (Fig.  6B). In this region pCMBS Ϫ only reacted with ␤ 1 A252C (2Ј) and ␤ 1 T256C (6Ј). There was no effect of pCMBS Ϫ on the intervening Cys mutants. Similar patterns of reactivity have been found at the aligned positions in SCAM studies of other superfamily members (Fig. 6A) (15,26,30,48). The narrow accessible face suggests that in this region the protein is more tightly packed, and thus pCMBS Ϫ only has significant access to channel-lining residues. Of note, the reaction rate at 7Ј was more than 50-fold slower than the rate at 6Ј, suggesting restricted access to the 7Ј position or limited residence time in the channel lumen. ␤ 1 L253C (3Ј) and ␤ 1 I255C (5Ј), two residues that flank the channel-lining face in this region, did not react with pCMBS Ϫ (Fig. 6B). Cys substituted at the aligned positions in the GABA A ␣ 1 , ACh ␣ and ␤, and 5-HT 3A subunits do not react with sulfhydryl reagents in the open state (15,26,28,30,45).
Access to these residues may be limited because of close packing interactions with the neighboring M2 segments. The limited accessible face and inferred tight protein packing are consistent with previous work that suggested that in the open state the narrow region of the Cys-loop receptor channels extends from Ϫ1Ј to 2Ј, forming the size-and charge-selectivity filters (15, 49 -54). In this region of the GABA A receptor we inferred that the open channel diameter was at least 9 Å because picrotoxin binds at the 2Ј level (50), and picrotoxin is a rigid, roughly spherical molecule that is ϳ9 Å in diameter.
It should be noted that Cys substituted at the 11Ј position has never been shown to react with charged sulfhydryl reagents (Fig. 6A). In the ACh receptor closed state structure this position is noted to be in close proximity with the M1 segment (11). Our results suggest that access to the 11Ј position is restricted in the open/desensitized states as well. Interestingly, mutations at the ACh receptor ␦ subunit 11Ј position had virtually no effect on the gating equilibrium constant, implying that the relationship between this residue and its environment changes little during gating (55).
pCMBS Ϫ modification of the Cys substituted for ␤ 1 Arg 269 (19Ј) had opposite effects on subsequent GABA-induced currents depending on whether it was modified in the absence or in the presence of GABA (Figs. 2 and 4). Modification in the absence of GABA caused inhibition of subsequent GABA-induced currents but an increase in the spontaneous open probability. In contrast, modification in the presence of GABA resulted in potentiation of subsequent currents with no apparent change in the spontaneous open probability. This implies that the 19Ј Cys was trapped in two different positions depending on the state in which it was modified. The two modified states were not interconvertible. This implies that this region undergoes a conformational change during gating. A similar effect was observed following modification of the ACh receptor ␣ 12Ј Cys mutant ␣S252C (26).
The 13Ј position, ␤ 1 T263C, is predicted to lie on the channellining face, and in the closed state ACh receptor structure it forms the upper part of the channel gate (11,15,26,56). Strikingly, at ␤ 1 T263C, pCMBS Ϫ application either in the absence or in presence of GABA had no effect on the subsequent GABA-induced currents (Figs. 2 and 4). As noted above, application of MTS-benzophenone to ␤ 1 T263C did irreversibly inhibit subsequent GABA currents. This implies that the engineered Cys is reactive. We do not understand why reaction with pCMBS Ϫ is functionally silent. In all other Cys-loop receptor M2 segments tested, the GABA A ␣ 1 , ACh ␣ and ␤, and the 5-HT 3a receptor subunits, reaction of charged sulfhydryl reagents with Cys substituted at the 13Ј position had detectable functional effects (Fig. 6A) (15,26,28,30,45). In the ACh receptor ␤ subunit, single channel recordings demonstrated that modification of the 13Ј Cys only affected open probability but not single channel conductance (28).
The pCMBS Ϫ reaction rate with an engineered Cys depends on two major factors: 1) accessibility of the Cys to bulk solution, and 2) reactivity of the Cys with the sulfhydryl reagents. Accessibility depends on steric and electrostatic factors in the access pathway from bulk solution to the site of the Cys. Furthermore, the local electrostatic potential will alter the local concentration or time-averaged concentration of the reagent in proximity to the Cys. Reactivity depends on the local Cys environment where factors that affect reaction rate include steric constraints and Cys ionization state, which is influenced by fractional time on the water-accessible surface and local electrostatic potential (25). To a certain extent it is difficult to compare reaction rates with Cys at different positions because it is difficult to determine which of the many factors are influ-  encing the reaction rate. Nevertheless, there are two aspects of the pCMBS Ϫ reaction rates with the ␤ 1 M2 Cys mutants which should be noted. First, the reaction rates with Cys substituted in the 17Ј-20Ј region are all less than 1,000 M Ϫ1 s Ϫ1 . In contrast, aside from T257C (7Ј), the rates with the other pCMBS Ϫreactive positions were all greater than 1,000 M Ϫ1 s Ϫ1 (Table  II). Based on the ACh receptor structure, at least one of the residues in the 17Ј-20Ј region should be facing into the channel, most likely 17Ј and 20Ј and they should be readily accessible to pCMBS Ϫ . Thus, the slow rates in this region might be the result of limited residence time of pCMBS Ϫ in this region of the channel perhaps because of the local electrostatic potential. Reaction with R269C (19Ј), predicted to lie on the protein-facing side of M2 (11,15), may be limited by steric factors, although the rates at 12Ј and 8Ј, which flank the 19Ј position on the ␣ helix (Fig. 5), are 3-40 times faster, despite being deeper in the membrane. The second aspect of the rates to consider are the rates between the 2Ј and 16Ј positions on the presumptive channellining helix face that extends around from T262C (12Ј) to I264C (14Ј), the left side of the helix in Fig. 5. This face can be divided into two domains, one with rates Ͼ 7,000 M Ϫ1 s Ϫ1 extending from T262C (12Ј) to T263C (13Ј) and the other with rates Ͻ4,000 M Ϫ1 s Ϫ1 extending from T256C (6Ј) to I264C (14Ј) (Fig.  5). The rates are 2-25-fold slower with the residues in the second domain. Interestingly, if one takes the MTSEA ϩ reac-  (20). pCMBS Ϫreactive residues are indicated by black boxes. Numbers next to each reactive residue indicate the mean second order reaction rate in M Ϫ1 s Ϫ1 . Rates are from Table  II. The open square on L263C indicates that it reacted with MTS-benzophenone but that there was no functional response to pCMBS Ϫ application. As discussed under "Results," for technical reasons reaction rates with ␤ 1 L268C were not measured.
tion rate results of Pascual and Karlin (57) with the ACh receptor ␣ subunit one finds that they too can be divided into two similar domains on an ␣ helical wheel plot. At the 2Ј, 9Ј, 13Ј, and 16Ј positions the MTSEA ϩ reaction rates in the presence of ACh were all greater than 100 M Ϫ1 s Ϫ1 , and the rates in the absence of ACh were significantly slower. These positions align with the fast reacting domain of the GABA A ␤ 1 M2 seg-ment. In contrast, the MTSEA ϩ reaction rates with 6Ј and 10Ј positions were Ͻ 10 M Ϫ1 s Ϫ1 and were not significantly different in the absence and presence of ACh. This corresponds to the slower reacting face of the GABA A ␤ 1 M2 segment. Based on disulfide trapping experiments (31), we suggested previously that opening of the GABA A receptor channel involved a rotation of the M2 segments, and similar suggestions have been made for the ACh receptor based on cryoelectron microscopic images (11,58). We suggest that because of thermal motion and loose protein packing the M2 segments continuously rotate back and forth independently of each other, between a closed conformation and an open conformation; agonist stabilizes the open conformation. This would require a rotation of the M2 segment by 80 -100 o . The slow and fast reacting faces of M2 correspond to the channel-lining faces in the closed and open conformations. When all of the subunit M2 segments rotate into the open conformation the channel opens and conducts ions, but when fewer than five subunits are in the open conformation the channel is partially lined by "closed" state residues, and although it is nonconducting some of the channellining residues are accessible to sulfhydryl reactive reagents at a very slow rate.
In the closed state ACh receptor structure the narrowest region in the channel that presumably forms the closed channel gate is in the 9Ј-14Ј region (11). Yet Cys substituted for residues in this region and at more cytoplasmic positions reacted with pCMBS Ϫ added extracellularly (Fig. 4). The reaction rate at T256C in the absence of GABA was 77 Ϯ 12 M Ϫ1 s Ϫ1 (n ϭ 3), greater than 10-fold slower than the reaction rate in the presence of GABA. Similar results were reported previously at the 6Ј position in the GABA A and glycine receptors (48). There are several potential problems with using SCAM to identify the location of the closed state gate. The reaction could be occurring during spontaneous channel openings, or the sulfhydryl reagent itself may act as a weak partial agonist. At the ␤ 1 9Ј position Cys substitution caused an increase in the spontaneous open probability. Similarly, other mutations in M2 have been shown to increase spontaneous open probability at a variety of positions (59 -61). We have not measured the spontaneous open probability of the Cys mutants reactive in the absence of GABA. A second potential problem is that pCMBS Ϫ might be a weak partial agonist in the mutant channels; pCMBS Ϫ is not that structurally different from piperidine-4sulfonic acid (P4S) a known GABA A receptor partial agonist. Thus, the reagent itself might act both as a partial agonist to induce a low level of channel opening and to modify the Cys made accessible by the channel opening. We cannot discriminate between these possibilities, and given the potential problems noted above we will not use our data to infer the position of the gate. It is notable that in the 5-HT 3A receptor, in the absence of 5-HT, MTS reagents did not react with any Cys substituted at a position more cytoplasmic than the 14Ј level, suggesting that the gate was below the 14Ј level (45). In the ACh receptor, based on SCAM experiments, the gate was inferred to be near the cytoplasmic end of the channel (57,62).
In summary, our current experiments indicate that the cytoplasmic end of the channel-lining M2 segments between 2Ј and 6Ј is tightly packed with the rest of the protein thereby limiting access to sulfhydryl reagents and presumably to water as well. Above this level the M2 segments are loosely packed and have a high degree of thermal protein motion. This motion includes both a rotation of the M2 segments around the helix axis of perhaps 80 -100 o and translational movements perpendicular to the channel axis. This motion is likely to be involved in the channel gating process. Further studies will be necessary to elucidate the extent of the thermal motion and its relation to channel gating. Data are from this work and Refs. 15, 26, 30, and 45. Residues are indicated by the index numbering system. B, schematic of a crosssection of the channel in the region between the 2Ј and 6Ј residues. Each circle represents an M2 segment. The channel lumen lies in the center. An ␣ helical wheel overlays the ␤ subunit on the right showing the positions of the pCMBS Ϫ -reactive and nonreactive residues. pCMBS Ϫreactive residues are indicated by black squares, and the mutant name is in bold type. The arc between 2Ј and 6Ј is indicated by a thick line to indicate the putative channel-lining face.