Gating-enhanced Accessibility of Hydrophobic Sites within the Transmembrane Region of the Nicotinic Acetylcholine Receptor's δ-Subunit

General anesthetics often interact more strongly with sites on open than on closed states of ligand-gated ion channels. To seek such sites, Torpedo membranes enriched in nicotinic acetylcholine receptors (nAChRs) were preincubated with the hydrophobic probe 3-(trifluoromethyl)-3-(m-iodophenyl) diazirine ([125I]TID) and exposed to agonist for either 0 ms (closed state), 1.5 and 10 ms (activated states), 1 s (fast desensitized state), or ≥1 h (equilibrium or slow desensitized state) and then rapidly frozen (<1 ms) and photolabeled. Within 1.5 ms, the fractional change in photoincorporation relative to the closed state decreased to 0.7 in the β- and γ-subunits, whereas in the α-subunit, it changed little. The most dramatic change occurred in the δ-subunit, where it increased to 1.6 within 10 ms but fell to 0.7 during fast desensitization. Four residues in the δ-subunit's transmembrane domain accounted for the enhanced photoincorporation induced by a 10-ms agonist exposure both when TID was added simultaneously with agonist and when it was preincubated with membranes. In the published closed state structure, two residues (δThr274 and δLeu278) are situated toward the extracellular end of helix M2, both contralateral to the ion channel and adjacent to the third residue (δPhe232) on M1. The fourth labeled residue (δIle288) is toward the end of the M2-M3 loop. Contact with these residues occurs on the time scale of a rapid phase of TID inhibition in Torpedo nAChRs, suggesting the formation of a transient hydrophobic pocket between M1, M2, and M3 in the δ-subunit during gating.

Among the most sensitive of these targets are the ligand-gated ion channel superfamily of receptors that include anion channels gated by ␥-aminobutyric acid (GABA) 1 and glycine and cation channels gated by acetylcholine (ACh) and serotonin (1,2). In general, the action on anion channels is to shift the agonist concentration-response curve to the left (enhancement) (3), probably as a result of stabilizing the open state (4). In cation channels, only a few of the smallest general anesthetics (e.g. urethane and ethanol) act similarly, whereas most are noncompetitive open channel inhibitors (5). General anesthetics interact with these receptors in a conformationally sensitive manner, often having their highest affinity for those transient conformational states that occur immediately after the agonist binds to the closed state (5). Structural information is key to a complete understanding at the molecular level of anesthetic interactions with the open state, but, given its transitory nature, other techniques will be required for the foreseeable future.
Currently, the structure of the transmembrane domain of the Torpedo nicotinic acetylcholine receptor (nAChR) is known to 4.0 Å in the closed state (6). The nAChR consists of four subunits, ␣, ␤, ␦, and ␥ with a stoichiometry 2:1:1:1, each having a bundle of four transmembrane helices (M1-M4, 27-34 residues in length). The five M2 helices are arranged about a central axis orthogonal to the membrane forming the channel lumen. A 9-Å structure of the open state (7) is of too low a resolution to resolve secondary structure.
Electrophysiological studies combined with site-directed mutagenesis provide kinetic evidence that general anesthetics interact with an enhancing site in anion channels that is located within the four-helix bundle of a given subunit (8) and with an inhibitory site in cation channels located in the channel lumen (9). Whereas these models of anesthetic action are selfconsistent, much work is necessary to rule out alternative explanations. For example, the mutations might increase the affinity for a distant general anesthetic site by stabilizing a conformation that has high affinity for the anesthetic (5). Thus, a complementary approach that can provide structural information such as ligand-protein contact points is desirable.
Photolabeling is an approach that has provided abundant information about agonist and antagonist sites on the nAChR, a member of the superfamily that is abundantly available from the electric tissue of Torpedo (10,11), and some, more difficult to achieve, information on GABA A receptors (GABA A Rs) from the brain (12). With some exceptions (13,14), these studies have been on the closed and desensitized states, conformations that exist at equilibrium, but an improved method of timeresolved photolabeling has now been introduced that has allowed nanomoles of nAChR to be efficiently photolabeled following exposure to agonists and other ligands for times as short as 1 ms (15).
The interaction of the lipophilic photoactivatable probe 3-(trifluoromethyl)-3-(m-iodophenyl) diazirine (TID) with every subunit of the nAChR in the closed and desensitized states has been thoroughly characterized (16 -19), as has the ability of general anesthetics, such as barbiturates, to modulate its photoincorporation (20). Preincubation of nAChRs with TID causes closed state inhibition, which develops on the same 100-ms timescale (21) as the increase in photolabeling of the channel lumen, whereas equilibration with the lipid-protein interface is complete within a few milliseconds (22). In addition to the above closed state inhibition, rapid perfusion studies of mouse muscle nAChRs reveal a phase of inhibition that occurs on the 10-ms time scale immediately following agonist-induced activation (23).
In the present work, we find that during channel gating, TID interacts uniquely with the nAChR ␦-subunit, where, within 10 ms after the addition of agonist, photoincorporation has increased about one and a half times but then decreases during fast desensitization. With reference to the nAChR closed state structure (6), three of the additional activation-dependent amino acid residues photolabeled by TID are clustered together on M1 and M2, and the fourth is on the M2-M3 loop close to the start of M3, suggesting the formation of a transient hydrophobic pocket in that region during activation.
Preparation of nAChR-rich Membranes-nAChR-rich membranes, prepared from the electric organs of Torpedo californica (Aquatic Research Consultants, San Pedro, CA) as described (24), were stored at Ϫ80°C in 38% sucrose, 0.02% NaN 3 . Protein concentrations were determined by the micro-BCA assay method (Pierce), and the ACh binding site concentration was determined by an [ 3 H]ACh (PerkinElmer Life Sciences) binding assay.
Time-resolved Photolabeling of nAChR-rich Membranes-The method was detailed previously (15,22). Briefly, the rapid mixing device was completely filled with buffer from the drive syringes to the two six-way sample valves, each of which contained a 0.5-ml sample loop loaded with appropriate reagents. The pneumatic ram delivered sufficient buffer both to force the reagents in the sample loops through the mixer in Ͻ1 ms and to expel only the mixed sample from the aging tube (Ն1.3 ms) onto a rotating stainless steel disk, precooled in liquid nitrogen, where it was frozen in a thin film within 1 ms (freeze quenching). Incubation times were varied by changing the velocity of the ram and the length of the aging tube. Typically, one sample loop contained 0.5 ml of membranes (4 mg/ml protein), sometimes premixed with [ 125 I]TID and/or carbamylcholine (Carb), and the other contained 0.5 ml of Torpedo physiological saline, Carb, and/or [ 125 I]TID. After mixing, the concentrations were 3.5 M [ 125 I]TID and 10 mM Carb. Stainless steel tubing was used throughout. Freeze-quenched membranes were irradiated (Blak-Ray UV lamp model UVL-56) on the slowly rotating disk for 30 min at 366 nm at a distance of ϳ3 cm. To obtain sufficient material for identification of labeled amino acids (preparative photolabeling), six samples per condition were photolabeled in the frozen state, and the material was subsequently combined for analysis. Twelve samples can be handled on a typical day.
SDS-PAGE-Aliquots of the thawed [ 125 I]TID-labeled nAChR-rich membranes were analyzed for protein concentration. The rest of the frozen membranes were thawed in an appropriate amount of 4ϫ sample buffer such that the final mixture contained 10% glycerol, 2% lithium dodecyl sulfate, 140 mM Tris-base, 106 mM Tris-HCl, 0.5 mM EDTA, 0.22 mM Serva Blue G250, 0.175 mM phenol red, 0.23 M ␤-mercaptoethanol, pH 8.5. The protein concentrations of the resuspended membranes were used to calculate the appropriate volumes to load for mass-balanced analytical SDS-polyacrylamide gels. The nAChR-rich membrane polypeptides were resolved by SDS-PAGE using the Laemmli buffer system (25) with 8% acrylamide separating gels (1.5 mm thick) containing 0.32% bisacrylamide. Polypeptides were visualized by Coomassie Blue stain, and the bands corresponding to the nAChR subunits as well as bands at 43 and 90 kDa, containing the non-nAChR peptides rapsyn and the Na ϩ /K ϩ ATPase ␣-subunit, respectively, were excised from the gel for ␥-counting (Micromedic 4/800 plus, Titertek Instruments, Huntsville, AL) and/or further processing. EndoLys-C digests were separated on 1.5-mm-thick 16.5% T/6% C Tricine SDS-polyacrylamide gels (19). The labeled bands of interest were detected within the wet gel by phosphorimaging (1-8-h exposure at 25°C) using a Storm Phospho-rImager (Amersham Biosciences), and the resulting image was used as a template for cutting up the gel. Prestained molecular mass standards (M-4038; Sigma) were used to determine the migration distances of the labeled fragments. Subunit or fragment bands of interest were eluted passively (3 days) in 12 ml of elution buffer (100 mM NH 4 HCO 3 , 0.1% SDS, 2.5 mM dithiothreitol, pH 8.4), filtered, concentrated in Vivaspin 15-ml concentrators (Vivascience, Hanover, Germany), and either precipitated in 75% acetone (Ϫ20°C, Ͼ3 h) or separated directly by reversed-phase HPLC.
Enzymatic Digestion of [ 125 I]TID-labeled nAChR Subunits-En-doLys-C digestions were performed in 25 mM Tris, 0.5 mM EDTA, 0.1% SDS, pH 8.6, for 2 weeks at 25°C. For V8 protease digestion of material after HPLC, the fractions (ϳ350 l) were first neutralized by the addition of 200 l of 25 mM Tris, pH 8.6, containing 0.1% SDS and 0.5 mM EDTA. The acid-neutralized pool was then rotary-evaporated to remove most of the organic solvent, and V8 protease (100 g) was added for 2 days at 25°C.
Reversed-phase HPLC Purification of [ 125 I]TID-labeled Fragments-Purification was performed on an Agilent 1100 HPLC with an inline degasser, column heater, and external absorbance detector. Separations were achieved at 40°C using a Brownlee Aquapore C-4 column (100 ϫ 2.1 mm) with a C-2 guard column. All solvents were HPLC grade. The aqueous phase (solvent A) was 0.08% trifluoroacetic acid, and the organic phase (solvent B) was 60% acetonitrile, 40% 2-propanol, 0.05% trifluoroacetic acid. The flow rates were 0.2 ml/min, and fractions of 0.5 ml were collected.
Amino-terminal Sequence Analysis of [ 125 I]TID-labeled Fragments-Samples were sequenced on a Procise 492 protein sequencer (Applied Biosystems) modified such that one-sixth of each cycle was used for amino acid identification and quantification, and five-sixths were collected to measure 125 I. HPLC fractions for sequencing were pooled and drop-loaded onto Biobrene-treated glass fiber filters (Applied Biosystems catalog no. 401111) at 45°C. When samples in detergent were loaded onto the filters, the detergent was removed by a prewash consisting of 5 min of gas trifluoroacetic acid, followed by 5-min washes with ethyl acetate and then n-butylchloride. The cpm detected in each cycle of Edman degradation were corrected by subtraction of the background, which varied from 10 to 25 cpm, in each of the channels of a Packard Cobra B5005 ␥-counter. All samples were sequenced between 60 and 100 days after the labeling and were back-corrected for 125 I decay to the date of labeling. Phenylthiohydantoin-derivatized amino acids were quantified from chromatographic peak heights, and the corrected 125 I cpm and pmol released are reported. Initial pmol (I 0 ) and repetitive yields (R) were calculated from a nonlinear least squares fit (Sigma Plot, Jandel Scientific) of the equation, f(x) ϭ I 0 ⅐R x , where f(x) is the pmol of the amino acid in cycle x. Arginines, serines, histidines, tryptophans, and cysteines were excluded from the fit because of known problems with their recovery. To quantify [ 125 I]TID photoincorporation into specific residues, the increase in 125 I cpm released at that cycle (i.e. cpm n Ϫ cpm (n -1) ) was divided by 5 times the pmol of the amino acid calculated from the values of I 0 and R. For some samples, the sequencing run was interrupted, and the material on the filter was treated with o-phthalaldehyde (OPA) as described (26). OPA reacts with primary amines preferentially over secondary amines (i.e. proline), and it may be used at any sequencing cycle to block Edman degradation of peptides not containing an N-terminal proline (27).
Patch Clamp Electrophysiology-Methods have been described previously (23). All experiments were performed at 21-23°C. Internal pipette and external buffers were K-100 (97 mM KCl, 1 mM MgCl 2 , 0.2 mM EGTA, and 5 mM Hepes, pH 7.5). Oocyte membrane patches were excised in the outside-out configuration on borosilicate pipettes (1.2-3 megaohms) and voltage-clamped at Ϫ50 mV. Patches were positioned in the outflow of a custom-built 2 ϫ 2 quad-barrel superfusion pipette, coupled to two orthogonal piezo electric elements. Application of RCdamped high DC voltages to the piezo elements moved the superfusion pipette and resulted in switching between adjacent superfusion solutions in Ͻ1 ms (10 -90% rise time of open pipette junction potential). In some experiments, two superfusion barrels were used: one containing K-100 buffer and an adjacent barrel containing ACh or ACh plus TID. If TID preincubation was required, three barrels were used: K-100, TID, and ACh plus TID. Patch currents stimulated by ACh were monitored and filtered (8-pole bessel, 1 kHz) with an Axopatch 200A amplifier (Axon Instruments, Foster City, CA). Recording of digitized data (1-2 kHz) and control of the piezo-driven superfusion device were achieved using a Digidata 1200 series interface and pClamp 7.0 software (both from Axon Instruments, Foster City, CA).
Analysis of Electrophysiological Data-Current traces were analyzed offline. Base-line leak currents were subtracted digitally. Current decay rates were determined by fitting exponential decay functions to data points from the peak current to the point at which ACh superfusion ceased. Nonlinear least squares fitting was performed using Clampfit 7.0 (Axon Instruments, Foster City, CA).

RESULTS
Overall Experimental Design Considerations-Our experimental design was based on knowledge of Torpedo nAChR kinetics. Agonist-induced channel activation occurs in ϳ100 s (28), within the dead time of our apparatus, and peak Carbinduced cation flux in Torpedo vesicles is linear over at least 10 ms, being terminated by fast desensitization with a time constant of ϳ200 ms (29,30). In addition, in the absence of agonist, ϳ10% of receptors are in the desensitized rather than the closed state (31). Therefore, we chose agonist incubation times of 0 (resting states), 1.5 and 10 ms (activated states), 1 s (fast desensitized state), and ϳ1 h (slow desensitized state). Initially, the membranes were preincubated with TID before being rapidly mixed with saturating concentrations of agonist so that nAChR-TID interaction kinetics with the closed state (see Introduction) (15,22) would not be superimposed upon agonistinduced changes. Subsequently, TID was added simultaneously with agonist to obtain a sense of the relative accessibility of the TID interaction sites.
Time Dependence of Agonist Action on TID Photoincorporation into nAChR Subunits- Fig. 1A shows a typical phosphor image of an SDS-polyacrylamide gel of membranes preincubated with TID and exposed for 10 ms either to 10 mM Carb or to buffer before freeze quenching and photolabeling. In this experiment, Carb increased photoincorporation of [ 125 I]TID into the ␦-subunit markedly, whereas that into the ␥-subunit decreased, and that into all other bands changed little.
The combined results from several separate experiments are shown in Fig. 1B, where the data for each experiment have been normalized to the zero time (no Carb) control (see legend). The overall percentage S.D. (coefficient of variation) for the data set of ratios was 12%. The ␦-subunit showed the most complex changes with exposure time to agonist. Photoincorpo-ration increased 1.6-fold 10 ms after adding Carb, followed by an equally dramatic decrease first to just below control values at 1 s and then to much lower values similar to the ␤and ␥-subunits at equilibrium. The ␣-subunit did not decrease at early times but showed a slight decline at later times. The ␤and ␥-subunits were the only subunits to decline significantly at 1.5 ms. They declined further by 1 s but did not change thereafter. Modest photoincorporation into the non-nAChR polypeptides was always observed but did not change upon exposure to agonist.
[ 125 I]TID Photoincorporation into the Transmembrane Domain of the ␦-Subunit-The ␦-subunit was chosen for more detailed study because of its unique kinetics (Fig. 1). nAChRrich membranes preincubated with [ 125 I]TID were exposed to 10 mM Carb or to buffer (six replicates each) for 10 ms, freezequenched, and photolabeled. The membrane polypeptides were separated by SDS-PAGE, and the ␦-subunits (visualized by Coomassie stain) were excised, eluted, concentrated, and acetone-precipitated. The resuspended ␦-subunits were digested in solution with EndoLys-C, which produces subunit fragments of ϳ21 kDa (␦EKC-21, beginning at ␦His 20 /␦His 26 and containing most of the extracellular domain), ϳ10 kDa (beginning at ␦Met 257 , the beginning of ␦M2), and ϳ12 kDa (beginning at ␦Phe 206 and containing a site of N-linked glycosylation (␦Asn 208 ) and ␦M1) (22,32).
Identification of [ 125 I]TID-labeled Amino Acids in ␦M2-␦M3-For each condition, HPLC fractionation of the ␦EKC-10 band produced a single peak of 125 I (Fig. 2B) eluting at 70% organic solvent that contained ϳ75% of the recovered radioactivity. Sequence analysis (Fig. 3A) of aliquots of the combined fractions 26 and 27 demonstrated the presence of a single fragment beginning at the N terminus of ␦M2, ␦Met 257 (ϪCarb, I 0 ϭ 9 pmol; ϩCarb, I 0 ϭ 8 pmol), with no other peptide sequences detected (Ͻ0.5 pmol). For the ϪCarb sample, there was an increase in 125 I release in cycle 9 (cpm n Ϫ cpm (n Ϫ 1) ) of 4,265 cpm with lower level increases in cycles 13 (670 cpm) and 16 (350 cpm), corresponding to the labeling of ␦Leu 265 , ␦Val 269 , and ␦Leu 272 , residues in the lumen of the closed channel that were labeled by [ 125 I]TID in previous freeze quench studies (22). After exposure to Carb for 10 ms, peaks of 125 I release were retained in cycles 9 (2,570 cpm), 13 (360 cpm), and 16 (220 cpm), but there was also 125 I release in cycles 18 (290 cpm), 22 (170 cpm), and 32 (340 cpm) that indicated labeling of ␦Thr 274 , ␦Leu 278 , and ␦Ile 288 .
Although in the ϩCarb sample the increased release of 340 cpm in cycle 32 was only ϳ10% of the 2,570 cpm in cycle 9, ␦Ile 288 was potentially labeled as efficiently as ␦Leu 265 , because the repetitive yield of Edman degradation was ϳ95%/ cycle. Two additional experiments were carried out to confirm that the cycle 32 release did result from labeling of ␦Ile 288 . First, we used OPA, which prevents subsequent sequencing of all peptides except those with N-terminal prolines, to determine whether the 125 I release in cycle 32 was two cycles after a proline, as predicted by the presence of ␦Pro 286 . An aliquot of the ϩCarb sample from fractions 26 and 27 was sequenced, but after 29 cycles of Edman degradation, the sequencer was stopped, the filter was treated with OPA, and sequencing was then continued for another 11 cycles (Fig. 3B). Enhanced release of 125 I in cycle 32 (340 cpm) was preserved after the OPA treatment, consistent with the presence of ␦Pro 286 in cycle 30 and 125 I incorporation in ␦Ile 288 . Second, we sequenced a sample after digestion with V8 protease, which cleaved the ␦Met 257 fragment after ␦Glu 280 , exposing ␦Pro 286 after five cycles of Edman degradation. In this sample, the filter was treated with OPA after five cycles, and sequencing was continued for 25 additional cycles. Although the identification of the ␦-subunit fragments was prevented by the presence of V8 protease, there was a single peak of 125 I release in cycle 8 (2,670 cpm; Fig. 3C). 2 Thus, the labeling of ␦Ile 288 was confirmed relative to ␦Pro 286 in these two separate experiments (Fig. 3, B and C).  Table  I. Inset, a replot of the data on an expanded cpm scale to highlight the agonist-dependent release in cycles 18, 22, and 32. B, another aliquot of the ϩCarb sample was sequenced, with sequencing interrupted after cycle 29, when ␦Pro 286 was the N-terminal amino acid, and the filter was treated with OPA (2) before resumption of sequencing (39,000 cpm loaded, 16,130 cpm remaining). The sequence beginning at ␦Met 257 (Ⅺ, I 0 ϭ 10 pmol, R ϭ 92%) was identified for all 40 cycles of Edman degradation, and after OPA treatment, 125 I release in cycle 32 was preserved. C, sequence analysis of a third aliquot of the ϩCarb sample after further digestion with V8 protease (see "Experimental Procedures"). Sequencing was interrupted after cycle 5 (2) for treatment with OPA (35,190 cpm loaded, 5,425 cpm remaining after 30 cycles).
For the ϪCarb and ϩCarb samples, the efficiencies of photoincorporation at each of the labeled amino acids (cpm/pmol), calculated from the observed increases in 125 I and phenylthiohydantoin-derivative releases, are tabulated in Table I. Exposure to Carb for 10 ms reduced by ϳ50% the efficiency of labeling of each of the labeled amino acids within the channel lumen (␦M2-9, -13, and -16). 3 For the amino acids labeled only after brief exposure to agonist, ␦Ile 288 (␦M2-32) was labeled at 5-9-fold higher efficiency than ␦Thr 274 (␦M2-18) or ␦Leu 278 (␦M2-22) and at similar levels to the most highly labeled amino acid in the channel lumen, ␦Leu 265 (␦M2-9).
State-dependent Labeling in ␦M2-␦M3-The above studies with membranes preincubated with [ 125 I]TID revealed two classes of labeled amino acids: those that preexist in the nAChR closed state and those that appear only after a brief exposure to an activating concentration of agonist. To obtain information about their relative accessibility in the closed and open states, experiments were performed in which [ 125 I]TID, with or without Carb (six replicates each), was added to nAChR-rich membranes for 10 ms before freeze quenching. Subunit labeling was determined by phosphorimaging after SDS-PAGE. As observed in a representative experiment (Fig.  4A), agonist-enhanced photoincorporation occurred only in the ␦-subunit. The nAChR subunits as well as non-nAChR polypeptides were excised from such gels, and 125 I incorporation was quantified. In each experiment, the mean of the ϩCarb samples was normalized to the mean of the no agonist samples, and these values were averaged with propagation of errors. For each nAChR subunit at 10 ms, the ratio of ϩCarb to ϪCarb photoincorporation was as follows: ␣, 1.0 Ϯ 0.3; ␤, 1.3 Ϯ 0.15; ␥, 1.3 Ϯ 0.1; ␦, 3.1 Ϯ 0.6 (mean Ϯ S.D. of three different experiments). The addition of Carb resulted in no significant changes in labeling of the non-nAChR polypeptides. Thus, the unusual behavior of the ␦-subunit relative to the others was maintained, but the magnitude of this agonist-enhanced subunit 3 To aid in locating amino acid residues on the structure of the nAChR, two conventions are adopted. First, residues are numbered from the first residue after the charged residue at the N-terminal end of M2; thus, ␦M2-9 is ␦Leu 265 . Second, residues on M1 are located relative to the central conserved Pro; thus, ␦M1-(P-3) is the ␦Phe 232 three residues before the Pro on ␦M1.  Fig. 2C) revealed the presence of the fragment beginning at ␦Phe 206 at a high level (I 0 ϭ 14 pmol, R ϭ 90%, data not shown), which indicated that the labeled and unlabeled fragments were separated by reversed-phase HPLC. B, for the nAChRs exposed for 10 ms to [ 125 I]TID with or without Carb before freezing and photolabeling, the only sequence detected began at ␦Phe 206 (Ⅺ, ϩCarb, I 0 ϭ 13 pmol, R ϭ 90%; छ, ϪCarb, I 0 ϭ 15 pmol, R ϭ 95%). For the ϩCarb sample, the sequencing filter was treated with OPA after cycle 19 (2), with 22,600 cpm loaded and 9,570 cpm remaining on the filter after 40 cycles of Edman degradation. For the ϪCarb sample, 3,430 cpm were loaded and 1,950 cpm remained. The 125 I release in cycle 20 was not seen in the sequence analysis of similar samples and probably was associated with the interruption of sequencing for treatment with OPA. photolabeling relative to the closed state was doubled compared with labeling for nAChRs preincubated with [ 125 I]TID (compare Fig. 1B). This was anticipated, because in the closed state TID equilibrates with its site in the ion channel with a time constant of about 75 ms (15).
For sequencing, two sets of six aliquots of membranes were freeze-quenched after 10-ms exposure to [ 125 I]TID with or without 10 mM Carb and photolabeled. The ␦-subunit was then isolated from the membranes by SDS-PAGE and digested with EndoLys-C. After these digests were separated by Tricine SDS-PAGE, phosphorimaging (Fig. 4B) revealed a major band of 125 I at ϳ13 kDa (␦EKC-13) and a secondary band at ϳ10 kDa. When the material eluted from ␦EKC-13 was further purified by reversed-phase HPLC (Fig. 4C), 125 I was distributed in two peaks, fraction 23 (55% organic, 30% of recovered 125 I) and fractions 26 and 27 (70% organic, 55% recovered 125 I), in each of which 125 I in the ϪCarb sample was Ͻ20% of that in the ϩCarb sample. The more hydrophobic peak (fractions 26 and 27) contained the ␦M2-␦M3 fragment, whereas the more hydrophilic peak (fraction 23) contained ␦M1 (sequenced in Figs. 5 and 6B, respectively; for details, see below).
Sequence analysis (Fig. 5) of pooled fractions 26 and 27 revealed the presence of a single fragment beginning at ␦Met 257 (ϪCarb or ϩCarb, I 0 ϭ 23 pmol). As expected, for the ϪCarb sample, the major peak of 125 I release was in cycle 9 (900 cpm), with lower level release in cycles 13 (170 cpm) and 16 (60 cpm), consistent with labeling within the lumen of the closed ion channel, and there was also a 60-cpm release in cycle 22. For the ϩCarb sample, in addition to increased release in cycles 9 (1,240 cpm), 13 (305 cpm), and 16 (60 cpm), there was release at cycles 18 (810 cpm), 22 (225 cpm), and 32 (380 cpm, after treatment of the sequencing filter with OPA after the 29th cycle of degradation). Table I shows that the efficiency of [ 125 I]TID incorporation was higher in ␦Leu 265 (␦M2-9) than in ␦Val 269 (␦M2-13) and ␦Leu 272 (␦M2-16), independent of the presence of Carb. The most efficiently labeled residues, ␦Thr 274 (␦M2-18) and ␦IIe 288 , were labeled at least 30-fold more efficiently in the presence than in the absence of Carb, whereas for ␦Leu 278 (␦M2-22), which was labeled at a lower level, Carb enhanced the efficiency 4-fold. For membranes exposed to [ 125 I]TID for only 10 ms (when TID binding within the lumen of the ion channel is at only 10 -20% of equilibrium (22)), the agonist dependent labeling at cycles 18 and 32 (28 and 40 cpm/pmol) was actually at higher efficiency than the labeling in cycle 9 (22 cpm/pmol).
We also compared the photolabeling efficiency of amino acids in ␦M2 for nAChRs preincubated with TID and exposed to Carb for as brief a time as possible, 1.5 ms, versus those desensitized by equilibration with 10 mM Carb. As before, the fragments beginning at ␦Met 257 were isolated from EndoLys-C digests of labeled ␦-subunit, but in this case they were sequenced only for 25 cycles of Edman degradation. For samples exposed to Carb for 1.5 ms, ␦M2-9, -13, -16, -18, and -22 were each labeled, with ␦M2-9 labeled at highest efficiency and ␦M2-18 and ␦M2-22 at lowest efficiency (Table I). For the nAChRs preequilibrated with Carb, and therefore in the desensitized state, the efficiency of labeling at ␦M2-9, -16, and -13 was 20 -30 times less than that in the sample exposed to agonist for 1.5 ms, and any labeling of ␦M2-18 or -22, if it occurred, was at Ͻ10% the efficiency of that labeling.
State-dependent Labeling in ␦M1-Based upon our previous characterization of the sites of photolabeling of [ 125 I]TID and [ 14 C]halothane in the ␦-subunit transmembrane domain (22,33), we expected that the hydrophilic peak of 125 I in the HPLC purifications of ␦EKC-13 (Figs. 2C and 4C) would contain a glycosylated fragment beginning at ␦Phe 206 and extending through ␦M1. Fraction 23 from Fig. 2C had been isolated from nAChRs preincubated with [ 125 I]TID and then exposed to buffer or to Carb for 10 ms. When the ϪCarb sample was sequenced (Fig. 6A), no release of 125 I above background was detected during 40 cycles of Edman degradation. In contrast, in the ϩCarb sample, there was a sharp increase of 125 I release in cycle 27 (490 cpm) with possible additional release in cycles 28 4 and 31. In these samples, the expected ␦-subunit fragment was present, but at very low levels (ϩCarb, I 0 ϭ 0.2 pmol; a For each data column, the 125 I release was determined from a single sequencing run. To allow comparison of labeling efficiencies of amino acids within fragments that were sequenced at times varying from 1 to 2 months after the initial photolabeling, the experimentally determined 125 I release (cpm n Ϫ cpm (n Ϫ 1) ) for each sequence run was corrected for the 125 I decay since the date of photolabeling.
b An upper limit of labeling efficiency was estimated from the local variability in cpm from cycle to cycle during Edman degradation, which in the absence of specific labeling was characterized by an S.D. of ‫%01ف‬ of the mean cpm over 5-10 cycles. The indicated upper limit was estimated as 20% of the mean cpm. c NS, not sequenced. d Estimated by combining the separate sequencing results (see Fig. 6A and "Results") for HPLC fractions 22 and 23 in Fig. 2C.
ϪCarb, I 0 ϭ 0.4 pmol). Although the mass levels were low, the 125 I release seen in cycle 27 was associated with labeling of ␦Phe 232 , since this release was seen after the sequencing filter had been treated with OPA before the second cycle of Edman degradation (␦Pro 207 ). The low mass level detected during the sequencing of fraction 23 suggested that the labeled and unlabeled fragments may have been separated during HPLC, and, consistent with this, HPLC fraction 22 was sequenced and found to contain the fragment beginning at ␦Phe 206 at a much higher level (ϩCarb, I 0 ϭ 14 pmol). Consequently, for the material isolated from nAChRs exposed to [ 125 I]TID for 10 ms in the absence or presence of Carb (Fig. 4C), fractions 22 and 23 were pooled before being sequenced (Fig. 6B), with OPA treatment after 19 cycles of Edman degradation corresponding to the expected location of ␦Pro 225 . The only detected sequence began at ␦Phe 206 (ϩCarb, I 0 ϭ 13 pmol; ϪCarb, I 0 ϭ 15 pmol). For the ϩCarb sample, once again there was a clear peak of 125 I release in cycle 27 (360 cpm), corresponding to ␦Phe 232 , with a shoulder of release in cycle 31 (␦Cys 236 ). In the absence of Carb, 125 I release in cycle 27 was minor (30 cpm). When compared on a cpm/pmol basis (Table I), it was only ϳ2% of that in the ϩCarb sample.
Electrophysiological Studies of Inhibition by TID-Electrophysiological recordings from outside-out oocyte membrane patches expressing Torpedo nAChRs revealed evidence for multiple steps leading to inhibition by TID, as previously reported for mouse muscle nAChRs (23). Simultaneous exposure to 1 mM ACh and a high concentration of TID (5 M) resulted in an ϳ10% reduction of peak patch currents (Fig. 7, top and  middle), and there was an increase in the rate of the current decay following the peak from 16 to 42 s Ϫ1 . Preexposure of patches to 5 M TID resulted in complete inhibition of AChinduced current (not shown). TID preincubation at concentrations of 0.16 M (not shown) and 0.5 M (Fig. 7, bottom) prior to activation with ACh and TID resulted in a diminished peak current response and a biphasic decay of the current. We observed both a TID-dependent rapid decay phase and a slow phase comparable with desensitization in the control. Systematic studies (not shown) revealed that the peak response decreased with TID preincubation time at an average rate of 5 Ϯ 1.8 s Ϫ1 (n ϭ 6, mean Ϯ S.D.). This rate was independent of TID concentration. The fast current decay accelerated with TID preincubation time, plateauing after 300 ms at 50 Ϯ 13 s Ϫ1 at 0.16 M TID and 150 Ϯ 78 s Ϫ1 at 0.5 M TID. Finally, the steady-state current after the fast phase declined more rapidly than peak current (not shown), just as it did in earlier observations with mouse nAChRs (23), indicating that the mechanism of TID inhibition is similar in both species. DISCUSSION The salient new finding in this study is that within 10 ms of the addition of agonist, when the fraction of nAChRs in the open state is maximum, [ 125 I]TID photoincorporates into a novel set of amino acids (activation-dependent residues) that are either not or are only inefficiently photolabeled in the closed or desensitized states and that are located only on the ␦-subunit within the M1 and M2 transmembrane helices and on the M2-M3 loop. During brief exposure to agonist (1.5-10 ms), only the ␦-subunit experienced an increase in photoincorporation. In contrast, the other subunits showed no such increase; instead, they experienced a steady decline with increasing time.
Another set of amino acids (activation-independent residues) on the channel face of the ␦M2 helix was photolabeled both in the closed state and following exposure to agonist for 10 ms. In earlier studies at room temperature, 80% of TID that photoincorporated into the closed state of the nAChR did so on the channel face of M2, and that labeling was ϳ5-fold less in the equilibrium desensitized state (16,34). The current time-resolved frozen state data show that, with the exception of the ␦-subunit, this decrease is apparent within milliseconds and is maximal within 1 s, suggesting that the reduction is associated with fast desensitization. However, the ␦-subunit again differed, experiencing a further decrease in photoincorporation when the conformation changed from the fast to the slow desensitized state. Although no agonist-induced increase was observed in the ␣-, ␤-, or ␥-subunits, we cannot rule out the possibility that small increases in some locations were masked by larger decreases in other locations.
A special role for the ␦-subunit was noted previously during linear free energy analysis of electrophysiological data. It is the slowest subunit to respond during the conformational wave following agonist binding, and it makes a distinct contribution, influencing both agonist binding and channel gating (35,36).
Although there are several intriguing questions arising from the subunit level data, this study focused on the ␦-subunit during gating. At 10 ms, the enhanced photoincorporation into the activation-dependent residues (␦Thr 274 , ␦Leu 278 , ␦Phe 232 , ␦Ile 288 ) is only partially offset by the decrease in the activationindependent residues (␦Leu 265 , ␦Val 269 , and ␦Leu 272 ). The net FIG. 7. Inhibition of Torpedo nAChR requires preexposure to TID. Top, three current traces from the same oocyte patch are displayed: pre-and post-control traces elicited with 1 mM ACh alone and a current elicited with 1 mM ACh plus 5 M TID. Precontrol (dashed line) peak current ϭ Ϫ220 pA; decay (desensitization) rate ϭ 16 s Ϫ1 ; ACh plus TID (solid line) peak current ϭ Ϫ195 pA, decay rate ϭ 42 s Ϫ1 ; postcontrol (dotted line) peak current ϭ Ϫ210 pA, decay rate ϭ 17 s Ϫ1 . Middle, peak current for this patch evolved in about 2 ms, and TID causes about 10% inhibition of the peak. Bottom, exposure to 0.5 M TID for 600 ms before activation with 1 mM ACh plus 0.5 M TID resulted in reduced peak response (solid trace; Ϫ72 pA) and biphasic current decay. The fast decay rate proceeds at a rate of 155 s Ϫ1 , and the slow decay rate is 16 s Ϫ1 . The slow decay rate is close to the desensitization rate of 12 s Ϫ1 seen in the control trace for this patch after activation with 1 mM ACh alone (dotted line). The peak control current for this patch, Ϫ210 pA, was scaled to the peak after TID exposure, in order to illustrate the different decay phases. increase in Table I accounts for the observed increase in the ␦-subunit in Fig. 1, suggesting that there are no other undetected contributions to agonist-induced photolabeling.
When TID was rapidly mixed with nAChRs to reveal relative access rates of open and closed states, the labeling of the activation-dependent residues was greatly enhanced, whereas the labeling of the activation-independent residues was similar in the absence or presence of agonist (Table I). The former result also suggests that preequilibration with TID does not profoundly influence subsequent conformation changes, consistent with previous stopped flow spectroscopic observations (21).
Inhibition by TID-Little functional inhibition is observed when nAChRs from either Torpedo or mouse muscle are exposed simultaneously to agonist plus TID, whereas variable length preincubation with TID results in inhibition of initial currents that develops at a rate of ϳ10 s Ϫ1 (this work) (21,23). In the closed state, access of TID to the activation-independent site also occurs at a rate of ϳ10 s Ϫ1 , which is much slower than access to the lipid bilayer or to the lipid-protein interface (15,22), and, therefore, prior occupation of this site on the closed state is required if TID is to act as an inhibitor during rapid agonist-induced activation.
In addition to the above pathway of inhibition, continuous current monitoring in rapidly perfused, excised outside-out oocyte membrane patches revealed an additional rapid phase of inhibition that only occurred after preincubation with TID. The rate of this phase was faster at ϳ10 2 s Ϫ1 and depended on TID concentration, suggesting a bimolecular action (Fig. 7, lower panel) (23). In the freeze quench experiments, the aqueous concentration of TID has been estimated to fall from a few micromolar to ϳ10 2 nM (in Ͻ1 ms in the case of rapid mixing) upon the addition to membranes (15). Such free aqueous concentrations are comparable with those experienced by oocyte membranes during continuous perfusion experiments. Because this rapid phase of inhibition and the activation-dependent photolabeling of the ␦-subunit both occur at similar TID concentrations and within 10 ms of addition of agonist, it seems likely that they are two manifestations of the same underlying process.
The Location of the TID Sites-The activation-independent residues in the ␦-subunit are all located in the ion channel's lumen in the closed state structure of the nAChR (6) (Fig. 8, dark blue residues). Of the four activation-dependent residues (Fig. 8, cyan residues), two, ␦Thr 274 and ␦Leu 278 (␦M2-18 and ␦M2-22), are situated on the opposite face of the M2 helix from the activation-independent residues. Another, ␦Phe 232 , is on the M1 helix, three residues above the conserved proline (␦M1-(P-3)), and adjacent to the ␦M2-18 and ␦M2-22 residues noted above. Examination of the Connolly surface in this region (Fig.  8C) reveals that ␦Thr 274 , ␦Leu 278 , and ␦Phe 232 are on the surface of an accessible pocket within the M1-M3 helix bundle that measures ϳ9 ϫ 5 ϫ 5 Å, which, consistent with the lack of prominent labeling in this region in the closed state, is too small to easily accommodate TID (ϳ11 ϫ 7 ϫ 7 Å). Because TID fits in this pocket in the activated state, it follows that conformation changes during activation must be associated with an expansion of this pocket sufficient to accommodate TID. However, the observed labeling pattern raises two questions about such a simple proposal. First, why are other residues that line this pocket, particularly those on ␦M3, not photolabeled? Second, why is the other activation-dependent residue (␦Ile 288 , M2-M3L) ϳ15 Å from the center of this pocket? A simple explanation for these discrepancies is that in the open state ␦Ile 288 is closer to the other labeled residues than in the closed state model and itself forms part of the TID-binding pocket.
Although uncertainty surrounds the structural changes occurring during it (37), one current concept of gating supposes that upon activation all five M2 helices rotate in a clockwise direction some 15°, which would move the activation-dependent ␦M2-18 and ␦M2-22 residues away from M3 and toward the ␣-␦ subunit interface (6,38). Such a rotation is supported by time-resolved fluorescence studies of rhodamine-labeled ␤M2-19 (39). The rotation of M2 is driven by agonist-promoted motions in the extracellular agonist-binding domain that are closely coupled to the M2-M3 loop (6,40). In rotating M2 clockwise, the M2-M3 loop probably moves toward the ␣-␦ interface. If, consequently, during gating M3 tilts toward the space between M2 and M3, it would both force M1 and M2 apart, admitting TID, and decrease the distance between ␦Ile 288 and the other three activation-dependent residues, perhaps allowing ␦Ile 288 to form the roof of a cleft whose floor is the cluster of three ␦M1-M2 residues. While speculative, such a model has the advantage of efficiently rationalizing our timeresolved photolabeling observations. Additional evidence for agonist-induced changes in M1 exposure in nAChRs comes  (6)) looking down the channel from the synaptic side. The cylinders approximate the individual transmembrane helices from the five subunits. B, the same model rotated 90°(viewed along the red arrow in A) and limited to the ␦and ␤-subunits as well as one ␣M2 helix. Colored blue are side chains in the channel that are labeled by TID in the closed state, including the activation-independent residues on the ␦-subunit (this work) and previously reported residues on other subunits (16,34). Colored cyan are the side chains in the ␦-subunit labeled by TID in an activation-dependent manner (after 10-ms agonist exposure). Also included are the side chains labeled in the desensitized state either by the anesthetic analogs azioctanol and/or azietomidate (green) or by halothane (yellow) (32,33,43). A Connolly surface representation of TID is included to scale. C, a stereo view of the top (extracellular) part of the ␦M1, ␦M2, and ␦M3 helices as viewed along the black arrow in B. The ␦M4 helix was removed to allow a better view of the interior Connolly surface of the helical bundle and of a solvent-accessible pocket, which extends down to ␦Cys 236 and is lined by all of the activation-dependent residues except ␦Ile 288 . Also included to the same scale is a Connolly surface representation of TID. from cysteine accessibility studies that found that positions on ␣M1 and ␤M1 equivalent to ␦Phe 232 were modified only in the presence of agonist (41,42).
It is also worth noting that at the current resolution of the closed state structure (6), it is difficult to unambiguously identify individual amino acids, so that some uncertainty exists in the assignments of the M2-M3 loop and the M3 ␣-helix. As published, three hydrophobic amino acids at the N terminus of M3 (in ␦-subunit, Tyr 291 , Leu 292 , and Met 293 ) have been positioned one helical turn above the plane of the lipid bilayer. An ϳ3-amino acid register shift would move ␦Ile 288 into the first turn of the ␦M3 helix at the level of the other activation-dependent residues in M1 and M2, thus completing the pocket. Activation would still be required to enlarge the pocket enough to accommodate TID.
Relevance to General Anesthetic Action on nAChRs-We have used the well characterized hydrophobic photolabel, TID, to probe for hydrophobic cavities that are unique to conformations associated with channel gating. Such cavities, apart from what they reveal about gating, are of pharmacological interest, because general anesthetics tend to selectively target activated states. How does the cavity we have detected relate to what is known about anesthetic action on this superfamily? A number of equilibrium studies with general anesthetic photolabels on the nAChR have been reported. The clinical general anesthetic halothane, a small photolabel with an exclusive preference for aromatic residues, photolabels in an agonist-sensitive manner ␦Tyr 228 (␦M1-(P-7); Fig. 8, yellow), one turn N-terminal to ␦Phe 232 (33). In the desensitized state, azietomidate, a close analog of a clinical intravenous agent, and 3-azioctanol both photolabeled ␣Glu 262 (␦M2-20; Fig. 8, green) (32,43). In the closed state structure, the ␣-carbon of this residue is separated by only 10 -12 Å from the ␣-carbons of our three activation-dependent residues on ␦M1 and ␦M2, distances that probably decrease during the hypothesized rotation of M2 during gating discussed above. It will be interesting to see whether analogs of these photolabels with the diazirine group in different positions, such as 7-azioctanol (44), provide evidence in favor of a general anesthetic site that spans the interface between the subunits. Such an interfacial site would be analogous to that of the classic allosteric effector of hemoglobin, 2,3-diphosphoglycerate (45).
Most site-directed mutagenesis studies of general anesthetic action in the nAChR have focused on the channel lumen (9,46), generally in the regions where TID photolabels the activationindependent residues. However, for neuronal nAChRs containing ␤ 4 and either ␣ 2 or ␣ 4 subunits, substitutions at M2-15 cause changes in alcohol sensitivity, and, based on the reactivity of Cys-substituted residues, ␣M2-15 is in an aqueous environment (47). In addition, at a position in the M2-M3 loop homologous with Torpedo ␦Ile 288 , an isoleucine to methionine substitution in homomeric chimeras of ␣ 3 ␣ 7 nAChRs decreased the sensitivity to halothane inhibition and shifted the ACh concentration-response curve to the left (48). These authors suggested that the mutations acted indirectly by changing gating rather than by ablating anesthetic binding, but our results increase the possibility that binding is part of the explanation.
Relevance to General Anesthetic Action on GABA A Rs-In the anionic channels of the ligand-gated ion channel superfamily, the hypothesis that there is a site for general anesthetics, contralateral to the channel and bounded by M1, M2, and M3, that is responsible for shifting the agonist concentration-response curve to the left has received detailed attention. Different classes of anesthetic often interact with different subunits of the GABA A R, but in general a strong circumstantial case has been built up using site-directed mutagenesis and cysteine accessibility to support the hypothesis that there are residues at M1-(P-1), M2-15, and at the extracellular end of M3 that modulate the action of general anesthetics (reviewed in Refs. 1 and 49). However, the general anesthetic propofol has been found to compete with cysteine reagents at the extracellular end of M3 but not with the M2-15 residue, suggesting that the later residue is not in the main part of the propofol binding pocket but that its role may be either to indirectly enlarge the site and/or to alter gating so as to enhance anesthetic sensitivity (50). Cysteine accessibility studies have established that amino acids in the outer part of M3 contribute to a solventexposed cleft whose accessibility is modulated by agonist (51) and by alcohols (52).
Our finding that on a homologous member of the ligandgated ion channel superfamily, there is TID photoincorporation in the same general region only during gating adds substantial support to the hypothesis that mutagenesis in this region of the GABA A and glycine receptors is modulating anesthetic binding directly, although such experiments are complicated by concomitant effects on gating. Two other aspects of our work have bearing on the anion channel studies. First, anesthetics stabilize the GABA A R open state (4,5), suggesting that the pocket becomes enlarged during gating just as it appears to do on the nAChR ␦-subunit. Second, the subunit dependence of TID interaction with the nAChR echoes that of general anesthetic action in the GABA A R mentioned above.
Perspective-Our identification of a novel transient TID site on the nAChR establishes the value of using the rapid mixing, freeze quench technique. In the future, it will be important to use the method with recently developed photoreactive general anesthetics (32,43) to map transient anesthetic-binding sites on nAChRs and GABA A Rs.