Enantiomeric barbiturates bind distinct inter- and intrasubunit binding sites in a nicotinic acetylcholine receptor (nAChR)

Nicotinic acetylcholine receptors (nAChRs) and γ-aminobutyric acid type A receptors (GABAARs) are members of the pentameric ligand-gated ion channel superfamily. Drugs acting as positive allosteric modulators of muscle-type α2βγδ nAChRs, of use in treatment of neuromuscular disorders, have been hard to identify. However, identification of nAChR allosteric modulator binding sites has been facilitated by using drugs developed as photoreactive GABAAR modulators. Recently, R-1-methyl-5-allyl-5-(m-trifluoromethyl-diazirinylphenyl) barbituric acid (R-mTFD-MPAB), an anesthetic and GABAAR potentiator, has been shown to inhibit Torpedo α2βγδ nAChRs, binding in the ion channel and to a γ+–α− subunit interface site similar to its GABAAR intersubunit binding site. In contrast, S-1-methyl-5-propyl-5-(m-trifluoromethyl-diazirinylphenyl) barbituric acid (S-mTFD-MPPB) acts as a convulsant and GABAAR inhibitor. Photolabeling studies established that S-mTFD-MPPB binds to the same GABAAR intersubunit binding site as R-mTFD-MPAB, but with negative rather than positive energetic coupling to GABA binding. We now show that S-mTFD-MPPB binds with the same state (agonist) dependence as R-mTFD-MPAB within the nAChR ion channel, but it does not bind to the intersubunit binding site. Rather, S-mTFD-MPPB binds to intrasubunit sites within the α and δ subunits, photolabeling αVal-218 (αM1), δPhe-232 (δM1), δThr-274 (δM2), and δIle-288 (δM3). Propofol, a general anesthetic that binds to GABAAR intersubunit sites, inhibited [3H]S-mTFD-MPPB photolabeling of these nAChR intrasubunit binding sites. These results demonstrate that in an nAChR, the subtle difference in structure between S-mTFD-MPPB and R-mTFD-MPAB (chirality; 5-propyl versus 5-allyl) determines selectivity for intra- versus intersubunit sites, in contrast to GABAARs, where this difference affects state dependence of binding to a common site.

Excitatory nicotinic acetylcholine receptors (nAChRs) 2 and inhibitory ␥-aminobutyric acid type A receptors (GABA A Rs) are members of the pentameric ligand-gated ion channel superfamily that also contains receptors for the neurotransmitters serotonin and glycine in vertebrates and for glutamate and biogenic amines in invertebrates (1)(2)(3)(4)(5). Drugs that modulate nAChR function have many potential therapeutic uses, as nAChR subtypes are widely expressed in neuronal and nonneuronal tissues and are involved in a wide variety of physiological and pathological processes, including neuromuscular disorders, cognition, nicotine addiction, and Alzheimer's and Parkinson's diseases (6 -8). Neuronal nAChR positive allosteric modulators (PAMs), some with subtype selectivity, have been identified through screens of natural products and chemical libraries (9,10). Muscle-type ␣ 2 ␤␥␦ nAChR PAMs, potentially of use in treatment of ALS, myasthenia gravis, and other neuromuscular disorders, have been harder to identify, and neuronal nAChR PAMs often act as channel blockers of muscle-type nAChRs (11,12).
Identification of binding sites for allosteric modulators is facilitated by the models of receptor structure that can be derived from crystal structures of soluble, homomeric acetylcholine binding proteins (13,14) and from X-ray or cryo-electron microscopy structures of distantly related prokaryotic channels and homomeric invertebrate glutamate-gated chloride channels and vertebrate glycine, GABA A , and serotonin receptors (15), as well as a structure of a heteromeric ␣4␤2 neuronal nAChR (16). In an ␣ 2 ␤␥␦ nAChR, agonist-binding sites are in the extracellular domain at the ␣-␥ and ␣-␦ subunit interfaces, and allosteric modulator sites have been identified in proximity to the agonist-binding sites and at corresponding positions at other interfaces (11,17). The transmembrane domain (TMD) of each receptor subunit consists of a bundle of four transmembrane ␣ helices (M1-M4), with M2 helices from each subunit associating at the central axis of the receptor to form the ion channel and the M1, M3, and M4 helices forming an outer ring in contact with membrane lipids. In addition to drug-binding sites in the TMD within the ion channel, potential intrasubunit drug binding sites are present in pockets within each subunit helix bundle and intersubunit binding sites in pockets at each subunit interface.
Studies using photoreactive analogs of general anesthetics, including etomidate, propofol, and mephobarbital that act as GABA A R PAMs but as nAChR inhibitors, have identified two homologous classes of intersubunit binding sites in the TMD of ␣1␤3␥2 GABA A Rs ( Fig. 1A; reviewed in Refs. 18 and 19). In ␣ 2 ␤␥␦ nAChRs, these drugs each bind to sites in the ion channel and to intrasubunit sites within the nAChR ␦ subunit helix bundle and/or intersubunit sites at the ␥ ϩ -␣ Ϫ subunit interface, where a photoreactive etomidate analog that acts as a low efficacy PAM also binds (20,21).
R-and S-mTFD-MPAB act as potent inhibitors of the Torpedo ␣ 2 ␤␥␦ muscle-type nAChR, each binding with high affinity to a site in the ion channel in the desensitized state and with R-mTFD-MPAB also binding to a site at the ␥ ϩ -␣ Ϫ interface (26). In this report, we identify the Torpedo nAChR amino acids photolabeled by [ 3 H]S-mTFD-MPPB to determine whether or not it binds to the same sites as R-mTFD-MPAB. We find that both barbiturates bind to the same site in the ion channel. In addition, S-mTFD-MPPB binds in the transmembrane domain to intrasubunit sites within the ␣ and ␦ subunits, but not to the intersubunit site that binds R-mTFD-MPAB. These results provide the first demonstration of the subtle difference in structure that is sufficient to determine drug selectivity for inter-or intrasubunit sites in a heteromeric nAChR.

Radioligand binding assays
The effect of mTFD-MPPB on the equilibrium binding of [ 3 H]ACh was determined at [ 3 H]ACh concentrations sufficient to occupy ϳ20% of nAChR-binding sites to allow determination of enhancement or reduction of binding. Similar to R-mTFD-MPAB (26), S-and R-mTFD-MPPB each increased [ 3 H]ACh binding with EC 50 values of 6.4 Ϯ 2.3 and 7.6 Ϯ 2.8 M, respectively ( Fig. 2A and Table 1). The ϳ20% maximal enhancement of binding indicated lower efficacy as a desensitizing agent than for proadifen, a prototypic desensitizing non-competitive antagonist (27) that in parallel experiments increased binding by 30% (EC 50 ϭ 0.95 Ϯ 0.40 M).
We also characterized the effects of S-and R-mTFD-MPPB on the binding of cationic channel blockers that bind preferentially to the nAChR ion channel in the desensitized state stabi-

[ 3 H]S-mTFD-MPPB photolabeling of Torpedo nAChR-rich membranes
After irradiation, membrane suspensions were fractionated by SDS-PAGE, and the covalent incorporation of [ 3 H]S-mTFD-MPPB was characterized by fluorography (Fig. 3A) and by liquid scintillation counting of bands excised from the stained gels (Fig. 3B). In the absence of other drugs (control conditions), the nAChR ␣ subunit was labeled most prominently. Photoincorporation into each nAChR subunit was enhanced in the presence of agonist (Carb) compared with control. Tetracaine did not reduce subunit photolabeling in the absence of agonist, but the enhanced nAChR subunit photolabeling in the presence of Carb was reduced in the presence of PCP or R-mTFD-MPPB. For the nAChR ␣, ␤, and ␥ subunits, PCP or R-mTFD-MPPB reduced photolabeling to levels close to that in the control condition, whereas for the ␦ subunit, the inhibition was partial. These results suggest that the Carb-enhanced nAChR subunit photolabeling results from [ 3 H]S-mTFD-MPPB photolabeling in the ion channel in the nAChR-desensitized state, with an additional PCPinsensitive binding site within the nAChR ␦ subunit.
To provide an initial characterization of the location of photolabeled residues within a nAChR subunit, we used in-gel digestion of labeled ␣ subunits to generate four large non-overlapping fragments of 4 kDa (␣V8-4, predominately starting from ␣Ser-1), 18 kDa (␣V8-18, beginning at ␣Thr-52 and con-

H]ACh (A) and [ 3 H]TCP (؉ Carb) and [ 3 H]tetracaine (؉ ␣-bungarotoxin) (B).
Binding assays were performed at 4°C by centrifugation. Each independent experiment was performed in duplicate, and the data were normalized to the specific binding in the absence of competitor. Pooled data (average Ϯ S.D.) are plotted. See Table 1   , and aliquots in duplicate were fractionated by SDS-PAGE. After staining the gel with GelCode TM Blue Safe Protein Stain, one set was prepared for fluorography (A), and gel bands were excised from the second for 3 H determination (B). The electrophoretic mobilities of the nAChR ␣, ␤, ␥, and ␦ subunits, rapsyn (Rsn), and the Na ϩ /K ϩ -ATPase ␣ subunit (␣ Na/K ) are indicated on the left of A. C, 3 H incorporation in the large nAChR ␣ subunit fragments generated by in-gel digestion of ␣ subunits with V8 protease. Gel bands containing ␣ subunits were isolated by SDS-PAGE from nAChR-rich membranes photolabeled with 1.5 M [ 3 H]S-mTFD-MPPB in the absence (control) or presence of 1 mM Carb, without or with 100 M PCP (Carb and Carb/PCP, respectively). The gel bands containing ␣V8-20, ␣V8-18, ␣V8-10, and ␣V8-4 were excised from the stained mapping gels, and 3 H incorporation was determined by liquid scintillation counting.

Table 1 The potencies of S-mTFD-MPPB, R-mTFD-MPPB, and proadifen as modulators of [ 3 H]ACh, [ 3 H]TCP (؉Carb), and [ 3 H]tetracaine (؉␣-bungarotoxin) equilibrium binding to Torpedo nAChRs
For each independent equilibrium binding assay, binding at each modulator concentration was determined in duplicate, and the specific binding at each concentration was normalized to the total specific binding in the absence of modulator. n, number of independent experiments; ND, not determined.

[ 3 H]S-mTFD-MPPB photolabels residues in ␣M2 and ␣M1
To identify photolabeled amino acids within ␣V8-20, this fragment was isolated from nAChRs photolabeled on a preparative scale in three conditions (control, Carb, and Carb/PCP). When EndoLys-C digests of ␣V8-20 were fractionated by rpHPLC (Fig. 4A), there were peaks of 3 H eluting at ϳ60 and ϳ80% organic solvent, where fragments beginning at ␣His-186 and extending though ␣M1 and at ␣Met-243, the N terminus of ␣M2, respectively, are known to elute (33). 3 H within both peaks was increased in the presence of agonist, but PCP strongly reduced labeling only in the more hydrophobic peak.
For nAChRs labeled in the presence of Carb, sequence analysis of the fragment beginning at ␣Met-243 revealed a major peak of 3 H release in cycles 5 and 6 with additional peaks in cycles 9 and 13, consistent with photolabeling of ␣Ile-247, ␣Ser-248, ␣Leu-251, and ␣Val-255 at positions M2-5Ј, M2-6Ј, M2-9Ј, and M2-13Ј that line the lumen of the ion channel (Fig.  4B). The efficiencies of photolabeling (cpm/pmol) at M2-5Ј and -6Ј were 10-fold higher in the presence of Carb than in its absence, and PCP inhibited that photolabeling by Ͼ90%, whereas both the state dependence and PCP sensitivity were reduced at M2-9Ј (Table 2).
Sequence analysis of the fragment beginning at ␣His-186 from nAChRs photolabeled in the absence of agonist revealed no peaks of 3 H release during 15 cycles of Edman degradation, which included the core aromatics ␣Tyr-190 and ␣Tyr-198 of ACh binding site segment C (top panel in Fig. 4C). To identify labeling within ␣M1, the filter was then treated with CNBr to cleave at ␣Met-207 before M1. Sequencing through ␣M1 then identified a single peak of 3 H release in cycle 11, consistent with photolabeling of ␣Val-218, but only for the sample from nAChRs photolabeled in the presence of Carb and PCP (bottom panel in Fig. 4C and Table 3).
To confirm [ 3 H]S-mTFD-MPPB photolabeling of ␣Val-218 in ␣M1, fragments beginning at ␣Ile-210 were isolated for sequence analysis by rpHPLC from trypsin digests of ␣ subunit (26) from an independent photolabeling experiment in the presence of Carb or Carb plus PCP (Fig. 4D). The peak of 3 H release in cycle 9 confirmed photolabeling of ␣Val-218 at ϳ3-fold higher efficiency in the presence of Carb and PCP than

[ 3 H]S-mTFD-MPPB photolabels ␦ subunit residues in the ion channel and in the helix bundle pocket
That PCP produced only a partial inhibition of the Carbenhanced labeling in the ␦ subunit ( Fig. 3B) indicated that [ 3 H]S-mTFD-MPPB may photolabel nAChR ␦ subunit residues in addition to those in the ion channel. The ␦ subunit helix bundle pocket is a likely site, because the photoreactive propofol analog [ 3 H]AziPm binds there in an agonist-dependent manner, photolabeling residues in ␦M1 (␦Phe-232) and ␦M2 (␦Thr-274 and ␦M2-18Ј), with photolabeling inhibited by propofol but not by PCP (35).

Table 2 [ 3 H]S-mTFD-MPPB photoincorporation efficiencies at amino acids within Torpedo nAChR M2 helices that line the ion channel (cpm/pmol of PTH-derivative)
The photolabeling efficiency (cpm/pmol of PTH derivative) for each residue was calculated from the observed 3 H release, the initial peptide mass, and repetitive yield as described under "Experimental procedures." n, the number of samples sequenced. For n ϭ 2, the average efficiencies Ϯ S.D. are tabulated.

Table 3 Pharmacological specificity of [ 3 H]S-mTFD-MPPB photolabeling of Torpedo nAChR amino acids within intrasubunit binding pockets in the ␣ and ␦ subunit helix bundle pockets (cpm/pmol of PTH-derivative)
The photolabeling efficiency (cpm/pmol of PTH derivative) for each residue was calculated from the observed 3 H release, the initial peptide mass, and repetitive yield as described under "Experimental procedures." For Experiment 1, single samples were sequenced to determine photolabeling efficiency in the absence of Carb (E Control ), and two independent experiments were carried out in the presence of Carb to determine photolabeling efficiencies in the absence (E Carb ) and presence of PCP (E Carb/PCP ). To take into account the differences in E Carb between experiments, the effect of PCP was quantified as the ratio E Carb/PCP /E Carb for each paired experiment. For Experiment 2, two ␣M1 and ␦M2 samples and single samples for the ␦ subunit fragments were sequenced. The effect of propofol was quantified as the ratio E Carb/PCP/propofol /E Carb/PCP for each sample. Averages Ϯ S.D. were tabulated when two samples were sequenced.
Inter-and intrasubunit nAChR anesthetic binding sites elutes, and a minor peak of 3 H eluted at ϳ60% organic where the fragment beginning at ␦Phe-206 before ␦M1 elutes. Sequence analysis of the fragment beginning at ␦Met-257 ( Fig.  5B) showed that for nAChRs labeled in the presence of Carb, there was a major peak of 3 H release in cycle 9 with smaller peaks of release in cycles 6, 13, 17, and 18, indicating primary photolabeling at ␦M2-9Ј in the ion channel with lower level photolabeling of channel lining residues at ␦M2-6Ј, 13Ј, and 17Ј as well as labeling of ␦M2-18Ј (␦Thr-274) that contributes to the ␦ subunit helix bundle. Photolabeling of M2-6Ј and M2-9Ј was at 5-10-fold higher efficiency in the desensitized state (Carb) than in the absence of agonist, whereas PCP in the presence of Carb inhibited photolabeling at M2-6Ј and M2-9Ј by ϳ90 and ϳ70%, respectively, with little, if any, inhibition of photolabeling at M2-13Ј and M2-18Ј (Tables 2 and 3). Sequence analysis of the fragment beginning at ␦Phe-206 (Fig. 5C) revealed a single major peak of 3 H release at cycle 27, consistent with photolabeling of ␦Phe-232 in ␦M1, for the sample from nAChRs labeled in the presence of agonist. That residue was photolabeled at Ͼ10-fold higher efficiency in the presence of Carb than in its absence, and PCP in the presence of Carb slightly enhanced rather than inhibited photolabeling (Table 3).

Propofol inhibits intrasubunit binding site photolabeling
Propofol, a widely used intravenous anesthetic and GABA A R PAM, binds to intersubunit binding sites in GABA A Rs (37). In Torpedo nAChRs, propofol acts as a desensitizing negative allosteric modulator and, based upon inhibition of [ 3 H]AziPm photolabeling, it binds in the ␦ subunit helix bundle pocket and also within the ion channel (35). To determine whether propofol also bound in the ␣ subunit helix bundle pocket, we examined the effects of propofol on [ 3 H]S-mTFD-MPPB photolabeling in the presence of Carb with PCP included to enhance [ 3 H]S-mTFD-MPPB photolabeling ␣M1 and minimize photolabeling in the ion channel. As shown in Fig. 6A and Table 3, sequence analysis through ␣M1 established that propofol inhibited photolabeling of ␣Val-218 in the presence of Carb and PCP. Similarly, sequence analysis of photolabeling in ␦M1 and ␦M2 (Fig. 6, B and C) established that propofol also inhibited photolabeling in the ␦ helix bundle pocket of ␦Phe-232 and ␦Thr-274 (␦M2-18Ј) ( Table 3) as well as in the ion channel (␦M2-9Ј, -13Ј, and -17Ј). Consistent with the results of Fig. 5B, no photolabeling of ␦M2-6Ј was seen in the presence of PCP. These results indicate that in Torpedo nAChRs, S-mTFD-MPPB binds to sites within the ␣ and ␦ subunit helix bundle pockets, and propofol inhibits binding at both sites.

[ 3 H]S-mTFD-MPPB photolabeling in ␤M2 and ␥M2
To extend the characterization of photolabeling in the M2 ion channel domain, we also sequenced fragments beginning at the N termini of ␤M2 and ␥M2, fragments beginning at ␤Met-249 that can be isolated from subunit trypsin digests by SDS-PAGE and rpHPLC (38) and at ␥Cys-252 that can be isolated by rpHPLC from EndoLys-C digests of ϳ24or 14-kDa ␥ subunit fragments produced by in-gel digestion with V8 protease fragment (26,39). Sequencing through ␤M2 from nAChRs photo-labeled in the presence of agonist (Fig. 7A) revealed major peaks of 3 H release in cycles 6 and 9, with additional peaks in cycles 13 and 17, consistent with photolabeling ion channel residues M2-6Ј, -9Ј, -13Ј, and -17Ј Sequencing through ␥M2 (Fig. 7B) revealed a major peak of 3 H release in cycle 6 with an additional peak in cycle 9. As seen for photolabeling in ␣M2 and ␦M2, labeling efficiency at M2-6Ј was increased by ϳ7-fold in the presence of agonist compared with the absence, and PCP strongly inhibited photolabeling at M2-6Ј and -9Ј, but not at M2-13Ј or M2-17Ј (Table 2).

Photolabeling in the M3 helices
Inspection of nAChR structural models allows identification of residues in M3 helices that are predicted to be exposed to lipid, to intersubunit interfaces, or to the intrasubunit helix bundles pocket. [ 3 H]R-mTFD-MPAB photolabeled residues in ␥M3 (␥Met-299) and ␣M1 (␣Leu-231) that contribute to a binding pocket at the ␥ ϩ -␣ Ϫ interface (26). Photolabeling of

Inter-and intrasubunit nAChR anesthetic binding sites
␥Met-299 was state-independent and insensitive to PCP, but it was inhibited in a concentration-dependent manner by R-mTFD-MPAB. In contrast, the hydrophobic probe [ 125 I]TID, which also contains the trifluoromethylphenyl diazirine reactive group, photolabeled ␥Phe-292, ␥Leu-296, and ␥Asn-300, residues exposed at the lipid interface, and the corresponding residues in ␤M3 and ␦M3 (40). To determine whether [ 3 H]S-mTFD-MPPB photolabeled ␥Met-299 or other residues in ␥M3, we isolated and sequenced a fragment beginning at ␥Thr-276 from nAChRs labeled in the absence or presence of Carb or in the presence of Carb and PCP (Fig. 8A). We found no evidence of photolabeling (peaks of 3 H release above background) in 30 cycles of Edman degradation. Similarly, there was no evidence of photolabeling in ␤M3 (Fig. 8B). [ 3 H]S-mTFD-MPPB incorporation in ␥M3 and ␤M3, if it occurred, was at Ͻ10% the photolabeling efficiency of ion channel residues (Carb). In contrast, [ 3 H]R-mTFD-MPAB photolabeled ␥Met-299 at the same efficiency as the most prominently labeled residues in the ion channel.
Sequence analysis of the corresponding fragment from the ␦ subunit, which begins at ␦Thr-281, revealed a single major peak of 3 H release in cycle 8 (␦Ile-288) for nAChRs labeled in the presence of agonist but not in the absence (Fig. 8C and Table 3). Similar to ␦Phe-232 in ␦M1 and ␦Thr-274 in ␦M2, ␦Ile-288, which is in the M3 helix, contributes to the binding pocket near the extracellular end of the ␦ subunit helix bun-dle. All three residues were also photolabeled in an agonistdependent manner by [ 125 I]TID (41,42) and by the photoreactive propofol analog [ 3 H]AziPm (35), but not by [ 3 H]R-mTFD-MPAB (26).

[ 3 H]S-mTFD-MPPB labeling in ␣M4
To further examine photolabeling at the lipid interface, we also sequenced the fragment beginning at ␣Tyr-401, which contains ␣M4 (Fig. 9). Within ␣M4, the single major peak of 3    , Carb (f), and Carb/ PCP (ƒ)) released during sequencing are shown for fragments isolated by rpHPLC from V8 protease digests of nAChR ␤, ␥, and ␦ subunits from the photolabeling experiment of Fig. 5. The major peaks of 3 H from the rpHPLC fractionations of the subunit digests were sequenced with OPA treatment at cycle 6 of Edman degradation (indicated by the arrows), which prevents further sequencing of peptides not containing a proline at this cycle and chemically isolates the fragments beginning at ␥Thr-276, ␤Thr-273, and ␦Thr-281. A, after OPA treatment, sequencing continued for the fragment beginning at ␥Thr-276 (I 0 ϭ 38 (Ⅺ) and 55 (f, ƒ) pmol). B, after OPA treatment, sequencing continued of the fragment beginning at ␤Thr-273 (I 0 ϭ 110 (Ⅺ, ƒ) and 170 (f) pmol). No evidence was seen for labeling in ␥M3 or ␤M3, based upon the absence of any peaks of 3 H release Ͼ25% above the background level of release. C, after OPA treatment, sequencing continued for the fragment beginning at ␦Thr-281 (I 0 ϭ 140 (Ⅺ, ƒ) and 220 (f) pmol). The peak of 3 H release at cycle 8 in C indicated photolabeling of ␦Ile-288 at Ͻ0.2 cpm/pmol (control), 2.0 cpm/pmol (Carb), and 1.5 cpm/pmol (Carb/PCP). The progressive increase in background 3 H release in cycles 13-32 of Edman degradation results from random cleavages of other fragments in the sequenced sample that contain residues labeled in ␦M2 and ␦M1. Although present, sequencing of those fragments was prevented by treatment of the sequencing filters with OPA in cycle 6, which blocks further sequencing of peptides not containing a proline at that cycle (44,54).

Discussion
S-mTFD-MPPB, a convulsant in vivo, acts as an inhibitor of ␣␤␥ GABA A Rs, whereas R-mTFD-MPAB, which differs only by chirality and the presence of a 5-allyl rather than 5-propyl substituent, acts as an anesthetic in vivo and as an ␣␤␥ GABA A R PAM (22,24). In ␣1␤3␥2 GABA A Rs, photoaffinity labeling studies established that S-mTFD-MPPB and R-mTFD-MPAB bind to the same binding site in the TMD at the ␥ ϩ -␤ Ϫ subunit interface, but with the opposite state dependence and in different orientations (23,25). S-mTFD-MPPB binds preferentially in the presence of bicuculline, an inverse agonist, whereas R-mTFD-MPAB binds preferentially in the presence of GABA. In a muscle-type nAChR, R-mTFD-MPAB acts as an inhibitor, binding to sites in the TMD in the ion channel and at the ␥ ϩ -␣ Ϫ subunit interface (26).
To determine whether S-mTFD-MPPB and R-mTFD-MPAB also bind to the same binding sites in a nAChR, in this report, we used radioligand binding assays and photoaffinity labeling to identify binding sites in the Torpedo nAChR for S-mTFD-MPPB. We found that S-mTFD-MPPB binds to the same site in the nAChR ion channel in the desensitized state as R-mTFD-MPAB and with similar high affinity. However, our results establish that S-mTFD-MPPB does not bind to the intersubunit site that binds R-mTFD-MPAB. Rather, S-mTFD-MPPB binds to homologous intrasubunit sites in the ␣ and ␦ subunits in pockets formed by each subunit's bundle of transmembrane helices. Furthermore, propofol, but not the positively charged channel blocker PCP, inhibits binding of S-mTFD-MPPB to those intrasubunit sites. Whereas anesthetics, including halothane, propofol, and the photoreactive propofol analog AziPm, have been shown previously to bind in a state-dependent manner within the ␦ subunit helix bundle pocket (35,43), this is the first time, to our knowledge, that anesthetics or other drugs have been found to bind within the ␣ subunit intrasubunit site.
A comparison of S-mTFD-MPPB and R-mTFD-MPAB actions and binding sites in Torpedo nAChR and ␣1␤3␥2 GABA A R is shown in Table 4. The locations of the amino acids photolabeled by [ 3 H]S-mTFD-MPPB that define the ion channel and intrasubunit binding sites are shown in Fig. 10 in a Torpedo californica nAChR homology model based upon the recently determined structure of an expressed human (␣4) 2 (␤2) 3 nAChR (16). Also shown in Fig. 10 (B-E) are the most energetically favorable binding poses predicted by computational docking for S-mTFD-MPPB in each of the binding sites.

H]S-mTFD-MPPB, [ 125 I]TID, and [ 3 H]AziPm, but not by [ 3 H]R-mTFD-MPAB
The efficiency of [ 3 H]S-mTFD-MPPB photolabeling of ␣Val-218 was 3-fold higher in the presence of Carb and PCP than in the presence of Carb alone, whereas PCP inhibited [ 3 H]S-mTFD-MPPB photolabeling of the ion channel residues M2-6Ј by ϳ90% and M2-9Ј by 50 -80%. The simplest interpretation of the enhanced ␣Val-218 photolabeling is that when PCP binds in the ion channel at the level of M2-2Ј and M2-6Ј, there is a change in structure of the ␣ subunit helix bundle that increases S-mTFD-MPPB binding affinity. Interestingly, for TDBzl-etomidate, which binds at the ␥ ϩ -␣ Ϫ interface and photolabels ␣M2-10Ј, PCP increased by ϳ2-fold ␣M2-10Ј photolabeling in the presence of Carb (44). These results indicate that PCP binding in the ion channel stabilizes an nAChR structure that differs from that of the desensitized state stabilized by agonist alone.

S-mTFD-MPPB and propofol bind within the ␦ subunit helix bundle pocket
As seen for other nAChR negative allosteric modulators, including [ 14 C]halothane (43), [ 125 I]TID (36,41), and the photoreactive anesthetics ([ 3 H]TFD-etomidate and [ 3 H]AziPm (35,45)), [ 3 H]S-mTFD-MPPB photolabeled in an agonist-dependent manner residues contributing to the ␦ subunit helix bundle pocket (Fig. 10E). [ 125 I]TID photolabeling was greatly enhanced in the nAChR open channel and transient desensitized states compared with the equilibrium desensitized state (36,42), and further studies using rapid-mixing and freezequench techniques will be necessary to determine whether [ 3 H]S-mTFD-MPPB has a similar state dependence. Propofol inhibition of [ 3 H]S-mTFD-MPPB photolabeling of these residues is consistent with its previously reported inhibition of [ 3 H]AziPm photolabeling (35). In contrast to the enhanced [ 3 H]S-mTFD-MPPB photolabeling of ␣Val-218 in the presence of PCP, little, if any, enhancement was seen for photolabeling in the ␦ subunit helix bundle. That PCP did not enhance ␦ intrasubunit photolabeling serves as a control that the enhanced photolabeling seen at ␣Val-218 results from positive allosteric coupling between PCP binding in the channel and the ␣ intrasubunit site and is not simply due to an increase in the free [ 3 H]S-mTFD-MPPB concentration resulting from its displacement by PCP from the ion channel.

S-mTFD-MPPB binding in the nAChR ion channel
Barbiturates of diverse structures act as state-dependent inhibitors of Torpedo nAChR (47), and they probably bind to sites in the ion channel because they fully inhibit binding of channel blockers (48,49). Our photolabeling results establish that S-mTFD-MPPB binds to the same region in the ion channel as R-mTFD-MPAB (26) and with the same Ͼ10-fold selectivity for the desensitized state compared with the resting, closed channel state. Both barbiturates photolabel residues at M2-6Ј and M2-9Ј most efficiently and also photolabel M2-13Ј and M2-17Ј. Consistent with the location in the agonist-stabilized desensitized state of the high affinity PCP binding site near the cytoplasmic end of the ion channel (50) and the capacity of PCP and uncharged anesthetics to bind simultaneously in the channel, PCP fully inhibited photolabeling at the level of M2-6Ј but did not inhibit labeling at the level of M2-13Ј or M2-17Ј. Whereas S-mTFD-MPPB and R-mTFD-MPAB, which are N-methylated, bind in the ion channel preferentially in the desensitized state, many barbiturates lacking the N-methyl group bind preferentially in the resting, closed channel state (49). In recently solved crystal structures of the cationic prokaryotic nAChR homolog GLIC in a locally closed state, the barbiturate binding site in the ion channel has also been localized to the level of M2-2Ј to M2-9Ј (51).
S-mTFD-MPPB and R-mTFD-MPAB each bind with higher affinity to a site in the ion channel than to intra-or intersubunit sites, and it is binding to the ion channel site that most likely produces nAChR desensitization and inhibition for nAChRs equilibrated with either drug. However, further studies defining the kinetics of inhibition and kinetics of binding to these different classes of sites will be necessary to determine whether either or both barbiturates act primarily as an open channel blocker upon transient exposure to drug and agonist. Electrophysiological and time-resolved photolabeling studies have shown that TID binding to the ␦ subunit intrasubunit site contributes to inhibition upon initial exposure and that binding in the ion channel occurs more slowly (36).

Computational docking calculations
Based upon calculated CDOCKER interaction energies, S-mTFD-MPPB is predicted to bind to the ␣ (Ϫ49 kcal/mol) and ␦ (Ϫ40 kcal/mol) subunit intrasubunit sites and with lower affinity in the ion channel (Ϫ37 kcal/mol at the cytoplasmic end centered near M2-2Ј, Ϫ28 kcal/mol at the level of M2-6Ј and -9Ј). However, the results of these calculations also predict that R-mTFD-MPAB binds with similar affinity as S-mTFD-MPPB to the intrasubunit sites that it does not photolabel and that S-mTFD-MPPB binds with similar affinity as R-mTFD-MPAB (Ϫ30 kcal/mol) to the ␥ ϩ -␣ Ϫ site that it does not photolabel. Potentially, the use of improved lipid-embedded nAChR structural models and docking algorithms in conjunction with molecular dynamics simulations may facilitate computational predictions consistent with the experimental evidence that S-mTFD-MPPB binds to intrasubunit sites, whereas R-mTFD-MPAB binds to an intersubunit site.

Functional significance of intrasubunit and intersubunit binding sites in muscle-type nAChR
There is great interest in developing PAMs for muscle-type nAChRs that could be of use in the treatment of ALS, myasthenia gravis, and other neuromuscular disorders. However, this has proven challenging, as most general anesthetics that act as GABA A R PAMs (21) and many neuronal nAChR PAMs act as potent ␣ 2 ␤␥␦ nAChR channel blockers (11,12). For Torpedo

Inter-and intrasubunit nAChR anesthetic binding sites
nAChRs, studies with photoreactive anesthetics led to the identification of drugs that bind with only low affinity to the ion channel, including TDBzl-etomidate, a low efficacy PAM that binds to the ␥ ϩ -␣ Ϫ intersubunit site (44) and TFD-etomidate, a potent inhibitor that binds to that intersubunit site and to the ␦ subunit helix bundle pocket (45). The identification of S-mTFD-MPPB as a drug binding to the ␣ subunit intrasubunit pocket will facilitate the identification of other drugs binding potentially with higher affinity and selectivity to that site.

Equilibrium binding of [ 3 H]ACh, [ 3 H]TCP, or [ 3 H]tetracaine
to Torpedo nAChR-rich membranes in Torpedo physiological saline buffer (250 mM NaCl, 5 mM KCl, 3 mM CaCl 2 , 2 mM MgCl 2 , and 5 mM sodium phosphate, pH 7.0) was determined by centrifugation as described (45). In brief, membrane suspensions were pre-equilibrated with radioligand for 30 min on ice and then incubated with various concentrations of non-radioactive drugs for 1 h at 4°C before centrifugation at 18,000 ϫ g for 45 min. After removal of the supernatants, membrane pellets were resuspended in 200 l of 10% SDS overnight, with pellet and supernatant 3  For each radioligand, f x , the specifically bound 3 H (cpm total Ϫ cpm nonspecific ) in the presence of competitor at concentration x, was normalized to f 0 , the specifically bound 3

[ 3 H]S-mTFD-MPPB photolabeling and gel electrophoresis
[ 3 H]S-mTFD-MPPB photolabeling of Torpedo nAChR-rich membranes (0.9 nmol of ACh-binding sites/mg of protein; 2 mg of protein/ml in Torpedo physiological saline buffer supplemented with 1 mM oxidized glutathione as an aqueous scavenger) was performed at 4°C on analytical or preparative scales using 0.1 or 10 mg of protein per condition, respectively. After incubation with [ 3 H]S-mTFD-MPPB (0.4 -0.9 M) for 30 min and an additional 30-min incubation in the absence or presence of other ligands, the membranes on ice were irradiated for 30 min using a 365-nm UV lamp (model EN-280L, Spectronics Corp., Westbury, NJ) at a distance of Ͻ2 cm. After photolabeling, membrane polypeptides were resolved by Tris-glycine SDS-PAGE on gels composed of 8% polyacrylamide, 0.33% bisacrylamide and visualized with GelCode TM Blue Safe Protein Stain (ThermoFisher). For analytical photolabelings, duplicate samples were separated by SDS-PAGE, with stained subunit bands excised from one set for 3 H quantification by liquid scintillation counting and the other set analyzed by fluorography using Amplify (GE Healthcare). For preparative photolabelings, bands containing the nAChR ␣, ␤, ␥, and ␦ subunits were excised from the stained gels, and material was eluted passively for 3 days at room temperature in elution buffer (100 mM NH 4 HCO 3 , 2.5 mM DL-dithiothreitol, 0.1% SDS, pH 8.4). Eluted samples were filtered, concentrated to a final volume of Ͻ400 l by centrifugal filtration using Vivaspin 15 M r 5000 concentrators (Vivascience, Stonehouse, UK), precipitated by 75% acetone overnight at Ϫ20°C, and resuspended in digestion buffer (15 mM Tris, 0.5 mM EDTA, 0.1% SDS, pH 8.1). For most preparative photolabelings, only 25% of the ␣ and ␥ subunit gel bands were eluted, with 75% of those gel bands used for in-gel digestion with V8 protease (100 g) on 15% polyacrylamide, 0.76% bisacrylamide mapping gels (32,40). The resultant subunit fragments (␣V8-20, ␣V8-10, ␥V8-24, and ␥V8-14) were recovered from gel bands by passive elution, concentrated, and resuspended in digestion buffer. In addition, ␣ subunits from Torpedo nAChR-rich membranes (0.5 mg of protein) photolabeled on an analytical scale with 1.5 M [ 3 H]S-mTFD-MPPB were digested in gel by V8 protease (5 g), with 3 H distribution in the fragments determined by liquid scintillation counting.

Inter-and intrasubunit nAChR anesthetic binding sites
Genapol to reduce the SDS concentration to 0.02% and then digested with 200 g of TPCK-treated trypsin in the presence of 0.4 mM CaCl 2 overnight (␤ subunit) or for 2-3 days (␣ subunit and ␣V8-10). ␣V8-20, ␥V8-24, ␥V8-14, and 60% of ␦ subunit in digestion buffer were digested with 0.5 units of EndoLys-C for 2 weeks. Aliquots of ␤ (50%), ␥ (25%), and ␦ (40%) subunits were digested with 200 g of V8 protease for 2-3 days in digestion buffer. Small pore Tricine SDS-PAGE gels (16.5% T, 6% C) (36,53) were used to fractionate the ␤ and ␦ subunits after digestion in solution with trypsin and EndoLys-C, respectively. The resultant fragments were recovered from Tricine gel bands by passive elution and concentrated by centrifugation for the further purification by rpHPLC. rpHPLC and sequence analyses rpHPLC was performed as described (45) using an Agilent 1100 binary rpHPLC system, a Brownlee Aquapore BU-300 column, and a mobile phase consisting of aqueous solvent A (0.08% trifluoroacetic acid) and organic solvent B (3:2 acetonitrile/isopropyl alcohol and 0.05% trifluoroacetic acid). Material was eluted at a flow rate of 0.2 ml/min using a non-linear gradient with solvent B increasing from 25 to 100% over 90 min. Fractions were collected every 2.5 min, and 3 H was deterimined by counting 10%.
rpHPLC fractions containing ␣M1, ␣M4, or ␦M1 were loaded onto PVDF membrane filters using Applied Biosystems ProSorb TM sample preparation cartridges. All other rpHPLC fractions containing 3 H-labeled peptides were drop-loaded onto the Applied Biosystems Micro TFA glass fiber filters at 45°C. Samples were treated with Biobrene Plus after loading to stabilize the peptides on the filters and then sequenced on an Applied Biosystems Procise 492 protein sequencer. For certain samples, sequencing was interrupted at predetermined cycles to treat the filter with OPA to prevent further sequencing of other peptides not containing a proline at that cycle (52,54). To facilitate identification of [ 3 H]S-mTFD-MPPB photolabeling in ␣M1, samples containing the fragment beginning at ␣His-186 were sequenced for 15 cycles, and filters were then treated with CNBr as described (17,55) to cleave at the C-terminal side of ␣Met-207 before ␣M1.
During N-terminal sequencing, two-thirds of the material was injected into an rpHPLC system for quantifying the amino acid at each cycle, and one-third of the sample was collected for determination of 3 H release. The masses of released phenylthiohydantoin (PTH)-amino acid derivatives were fit versus cycles of Edman degradation (SigmaPlot version 11) according to the equation, f x ϭ I 0 ϫ R x , where f x is the mass of amino acid on cycle x, I 0 is the initial mass of the peptide, and R is the repetitive yield of Edman degradation. The labeling efficiencies (cpm/pmol) of residues photolabeled by [ 3 H]S-m TFD-MPPB were calculated based on the expression, (2 ϫ (cpm x Ϫ cpm (x Ϫ 1) ))/(I 0 ϫ R x ).

Computational docking
A homology model of the T. californica nAChR was constructed based on the recently solved (3.9 Å) crystal structure of a neuronal (␣4) 2 (␤2) 3 nAChR (Protein Data Bank entry 5KXI (16)) using the Create Homology Model tool in Discovery Stu-dio 2017 (Biovia). Torpedo ␣ sequences were substituted for the two ␣4 subunits, and the three ␤2 subunits were replaced with Torpedo ␤, ␥, and ␦ subunits in their known positions relative to the ␣ subunits (i.e. clockwise ␣␤␦␣␥ when viewed from the extracellular side). The ␣4 to ␣ substitution required a singleresidue insert in loop C of the agonist binding site (Torpedo ␣Thr-196). To align the Torpedo ␤, ␥, and ␦ subunits with ␤2, the following adjustments were made. (i) For ␤ and ␥ subunits, ␤2Pro-14 and ␤2Ser-15 were deleted. (ii) Insertions were made between ␤2Glu-165 and ␤2Val-166 of 10 (␤), 8 (␥), or 12 (␦) residues in the structurally undefined agonist site loop F region. Because the structures of these large inserts are unknown, these inserts were removed from the model before docking. (iii) Four residue insertions were made in ␥ and ␦ subunit loop C regions between ␤2Asp-192 and ␤2Asp-193. (iv) A single residue was inserted for ␥ in the M1-M2 loop between ␤2Cys-237 and ␤2Gly-238. Nicotine molecules bound to the two agonist sites in the ␣4␤2 structure were retained in the homology model. The model was placed in a membrane force field, and the entire model was minimized to Ϫ85,106 kcal/mol. Initial attempts to dock S-mTFD-MPPB to a pocket between the M1, M2, and M3 helices at the extracellular end of the ␣ or ␦ transmembrane helix bundles were unsuccessful. To create a pocket in this region, S-mTFD-MPPB was placed within the helix bundle between M1, M2, and M3 adjacent to the labeled residues (for both ␣ and ␦ subunits), and the model was minimized to Ϫ107,840 and Ϫ84,853 kcal/mol for ␣ and ␦ subunits, respectively. A binding site sphere of 12-Å radius was placed around the minimized S-mTFD-MPPB, and the sphere was seeded with 12 copies each of S-mTFD-MPPB and R-mTFD-MPAB. Each seeded molecule was subjected to 40 molecular dynamics-induced alterations, and each altered structure was rotated/translated into 40 different starting orientations. Each sampling was minimized with the residues within the sphere, the final interaction energy was determined, and the lowest-energy solutions were collected along with the predicted orientations of the bound molecules. S-mTFD-MPPB and R-mTFD-MPAB were also docked using 12-Å binding site spheres in the pocket at the ␥-␣ interface and at three levels in the ion channel using binding site spheres centered at M2-2Ј, M2-6Ј/9Ј, and M2-13Ј/17Ј. For individual binding sites, docking results are displayed as Connolly surface representations defined by a 1.4-Å diameter probe of the 10 solutions with the lowest CDOCKER interaction energies.
Author contributions-Z. Y. and J. B. C. designed and analyzed the experiments that were performed by Z. Y. Both authors contributed to the writing of the manuscript and approved the final version of the manuscript.