A photoreactive analog of allopregnanolone enables identification of steroid-binding sites in a nicotinic acetylcholine receptor

Many neuroactive steroids potently and allosterically modulate pentameric ligand-gated ion channels, including GABAA receptors (GABAAR) and nicotinic acetylcholine receptors (nAChRs). Allopregnanolone and its synthetic analog alphaxalone are GABAAR-positive allosteric modulators (PAMs), whereas alphaxalone and most neuroactive steroids are nAChR inhibitors. In this report, we used 11β-(p-azidotetrafluorobenzoyloxy)allopregnanolone (F4N3Bzoxy-AP), a general anesthetic and photoreactive allopregnanolone analog that is a potent GABAAR PAM, to characterize steroid-binding sites in the Torpedo α2βγδ nAChR in its native membrane environment. We found that F4N3Bzoxy-AP (IC50 = 31 μm) is 7-fold more potent than alphaxalone in inhibiting binding of the channel blocker [3H]tenocyclidine to nAChRs in the desensitized state. At 300 μm, neither steroid inhibited binding of [3H]tetracaine, a closed-state selective channel blocker, or of [3H]acetylcholine. Photolabeling identified three distinct [3H]F4N3Bzoxy-AP–binding sites in the nAChR transmembrane domain: 1) in the ion channel, identified by photolabeling in the M2 helices of βVal-261 and δVal-269 (position M2–13′); 2) at the interface between the αM1 and αM4 helices, identified by photolabeling in αM1 (αCys-222/αLeu-223); and 3) at the lipid–protein interface involving γTrp-453 (M4), a residue photolabeled by small lipophilic probes and promegestone, a steroid nAChR antagonist. Photolabeling in the ion channel and αM1 was higher in the nAChR-desensitized state than in the resting state and inhibitable by promegestone. These results directly indicate a steroid-binding site in the nAChR ion channel and identify additional steroid-binding sites also occupied by other lipophilic nAChR antagonists.

tors (PAMs) 2 of synaptic and extrasynaptic ␥-aminobutyric acid type A receptors (GABA A Rs), anion-selective pentameric ligand-gated channels (pLGICs) (3)(4)(5), and negative allosteric modulators of muscle and neuronal nicotinic acetylcholine receptors (nAChRs), cation-selective pLGICs (6 -8). The complex effects of steroid PAMs on GABA A R gating and the locations of residues identified by mutational analyses as steroid sensitivity determinants suggest the existence of multiple steroid-binding sites in the transmembrane domain (TMD) of heteromeric GABA A Rs (9,10). Recent crystal structures of chimeric homopentameric receptors containing GABA A R ␣ subunit TMDs identify a binding site for steroid PAMs in the TMD at subunit interfaces (11,12), which is distinct from the intersubunit sites in the extracellular third of the TMD for etomidate, propofol, and barbiturates identified in heteromeric GABA A Rs by photolabeling and mutational analyses (13,14). A distinct intrasubunit site for pregnenolone sulfate, an inhibitory steroid, was also identified in a chimeric homomeric receptor TMD (11).
For muscle-type nAChRs, single-channel analyses of the acute effects of natural and synthetic steroids indicate a reduction of open channel lifetime (15)(16)(17), but it is not known whether this results from steroid binding within the ion channel or indirectly as a consequence of interactions at the lipid interface. As cholesterol is present in high concentrations in synaptic membranes and is important for facilitating agonistinduced conformational transitions, inhibition by steroids may result from perturbation of nAChR-cholesterol interactions (18 -20). Mutations of some amino acids in the M4 helices that are exposed at the lipid interface have as large an effect on channel gating as the substitutions at channel-lining residues (21,22). Also, substitutions at a position in ␣M4 exposed at the lipid interface reduced hydrocortisone potency (23). A muta-tion within the M2 ion channel domain that increased channel lifetime also reduces hydrocortisone potency, but there was no evidence for hydrocortisone competition with QX-222, an open channel blocker (16).
nAChR-rich membranes that can be isolated from the Torpedo electric organ provide a unique preparation to use photoaffinity-labeling techniques to identify steroid-binding sites in a muscle-type nAChR in its native membrane environment. Radiolabeled, photoreactive analogs of hydrophobic general anesthetics, including propofol, mephobarbital, and etomidate, have been shown to bind to sites in the TMD within the ion channel and to inter-and intra-subunit sites (24,25). Promegestone, a progestin steroid with intrinsic photoreactivity ( Fig. 1), is a potent inhibitor of Torpedo nAChRs expressed in Xenopus oocytes and of binding of [ 3 H]phencyclidine (PCP), a channel blocker, to nAChR-rich membranes (26). Whereas [ 3 H]promegestone photolabeled amino acids at the lipid interface in the M4 helices, no evidence was found of photolabeling ion channel residues. However, promegestone may have bound in the ion channel without efficient photolabeling.
To further characterize steroid-binding sites in the Torpedo nAChR, we now use 11␤-(p-azidotetrafluorobenzoyloxy)allopregnanolone (F 4 N 3 Bzoxy-AP) (Fig. 1), a photoreactive allopregnanolone derivative that is a potent general anesthetic and GABA A R-positive allosteric modulator (27). In contrast to promegestone, which upon UV irradiation forms a reactive ketyl radical at the 3-position of the steroid A-ring, F 4 N 3 Bzoxy-AP reacts by formation of a stabilized nitrene at the 11-position in the steroid C-ring (28,29). Because a nAChR agonist containing the same photoreactive group photolabeled aliphatic and aromatic amino acid side chains within the ACh-binding sites in the nAChR extracellular domain (30), F 4 N 3 Bzoxy-AP should have the capacity to photoincorporate into many of the amino acid side chains in the nAChR TMD. We found that, similar to promegestone, F 4 N 3 Bzoxy-AP inhibited binding of a channel blocker in the nAChR-desensitized state. In contrast to promegestone, [ 3 H]F 4 N 3 Bzoxy-AP photolabeled residues in the nAChR ion channel in the desensitized state (positions ␤M2-13Ј and ␦M2-13Ј, numbered from the conserved positive charges at the N termini of the M2 helices), which provides a first identification of a steroid-binding site in the nAChR ion channel. [ 3 H]F 4 N 3 Bzoxy-AP also identified sites accessible from the lipid, one within the ␣ subunit, identified by photolabeling of ␣Cys-222/␣Leu-223 at the interface between the M1 and M4 helices, and a site near the cytoplasmic surface of the TMD, identified by photolabeling of ␥Trp-453 in ␥M4, a residue photolabeled by [ 3 H]promegestone.

Radioligand binding assays
We compared the effects of F 4 N 3 Bzoxy-AP and alphaxalone on the equilibrium binding of [

Steroid-binding sites in an ␣␤␥␦ nAChR [ 3 H]F 4 N 3 Bzoxy-AP photolabeling of Torpedo nAChR-rich membranes
We compared patterns of nAChR subunit photolabeling after irradiation of nAChR-rich membranes at 365 or 254 nm in the absence or presence of an agonist (carbamylcholine (Carb)), PCP, alphaxalone, or the steroid noncompetitive antagonists alphaxalone or promegestone. After photolabeling with [ 3 H] F 4 N 3 Bzoxy-AP (3 M) and fractionation of membrane polypeptides by SDS-PAGE, the 3 H incorporation into the nAChR subunits and other membrane polypeptides was assessed by fluorography (Fig. 3A) for qualitative characterization and by liquid scintillation counting of excised subunit gel bands to quantify photoincorporation (Fig. 3, B-D). As seen by fluorography, after irradiation at 365 nm, the nAChR subunits photolabeled most prominently were the ␣ and ␥ subunits, with the most prominently photolabeled gel bands those of 34 and 32 kDa previously identified as the voltage-dependent anion channel and ADP/ATP carrier from contaminating mitochondrial fragments (35). Irradiation at 254 nm resulted in prominent, pharmacologically-specific photolabeling in the nAChR ␣ subunit, with Carb enhancing photolabeling compared with control and PCP inhibiting the Carb-enhanced photolabeling. Quantification of photolabeling by liquid scintillation counting (Fig. 3, B and C) established that for irradiation at 254 nm, Carb enhanced ␣ subunit photolabeling by ϳ50%, with the Carbenhanced photolabeling inhibited by PCP but not by 300 M alphaxalone or R-mTFD-MPAB, a barbiturate allosteric inhib-itor that binds to sites in the ion channel and at the ␥-␣ subunit interface (36). To test the effect of a second steroid antagonist on [ 3 H]F 4 N 3 Bzoxy-AP photolabeling at 254 nm, we used promegestone, which inhibits with an IC 50 of 10 M the ACh responses of Torpedo nAChRs expressed in Xenopus oocytes and the binding of the channel blocker [ 3 H]PCP to nAChR-rich membranes in the presence of Carb (Fig. 3D) (26). Carb enhanced nAChR ␣ subunit photolabeling by ϳ40%, and promegestone at 300 M reduced photolabeling to the level seen in the control condition or in the presence of Carb and PCP. Similarly for the ␤ and ␦ subunits, promegestone inhibited the small enhancement of photolabeling (ϳ20%) seen in the presence of Carb.

[ 3 H]F 4 N 3 Bzoxy-AP photolabeling in the nAChR ion channel
Pharmacologically-specific photolabeling was most prominent in the ␣ subunit after photolabeling at 254 nm, but we first characterizedphotolabelingwithintheM2channel-forminghelices after irradiation at 365 nm, a wavelength that minimizes nonspecific UV-induced protein degradation. Photolabeling was characterized in ␤M2 and ␦M2, because fragments beginning at their N termini (␤Met-249 and ␦Met-257) can be isolated readily (37)(38)(39) at ϳ10-fold higher mass levels than the ␣M2 fragment that requires a more complex purification procedure (40). Trypsin digests of ␤ subunits isolated from nAChRs photolabeled with [ 3 H]F 4 N 3 Bzoxy-AP in the absence or presence of Carb were fractionated by Tricine gel SDS-PAGE  3 H incorporation into nAChR subunit gel bands after irradiation at 365 nm (B) or 254 nm (C) from the same experiment as the fluorogram and from an independent experiment at 254 nm (D). The average cpm Ϯ S.D. are plotted for samples from two gels. Included in C and D are the p values, where statistically significant (p Ͻ 0.05, one-way ANOVA and Tukey's multiple comparison test for pairs of labeling conditions (GraphPad Prism 7)). A, left, electrophoretic mobilities are indicated of the nAChR ␣, ␤, ␥, and ␦ subunits, rapsyn (Rsn), the Na ϩ /K ϩ -ATPase ␣ subunit (␣ Na/K ), and the mitochondrial voltage-dependent anion channel (34 kDa) and ADP/ATP carrier (32 kDa).

Steroid-binding sites in an ␣␤␥␦ nAChR
and then rpHPLC (Fig. 4A). When the fragment beginning at ␤Met-249 was sequenced from the major peak of 3 H (Fig. 4B), the peak of 3 H release in cycle 13 identified photolabeling of ␤Val-261 (a channel-lining residue, position M2-13Ј) in the presence of Carb, whereas labeling in the absence of agonist, if it occurred, was at Ͻ10% that level. Similarly, after fractionation of an EndoLys-C digest of photolabeled ␦ subunit by Tricine-gel SDS-PAGE and then rpHPLC (Fig. 4C), sequence analysis of the major peak of 3 H established the presence of the fragment beginning at ␦Met-257 (Fig. 4D). The peak of 3 H release in cycle 13 in the presence of Carb indicated agonist-dependent photolabeling of ␦Val-269, the position in ␦M2 equivalent to ␤Val-261. In photolabelings with [ 3 H]F 4 N 3 Bzoxy-AP at 254 nm, we also determined that ␤Val-261 and ␦Val-269 were photolabeled in an agonist-dependent manner and at similar efficiency as for photolabeling at 365 nm (Table 1).

[ 3 H]F 4 N 3 Bzoxy-AP photolabels residues in ␣M1
Based upon the photoincorporation seen at the subunit level ( Fig. 3D), irradiation at 254 nm resulted in pharmacologicallyspecific photolabeling in the nAChR ␣ subunit. To identify photolabeling in the ␣ subunit, we took advantage of the fact that in gel digestion of the ␣ subunit with V8 protease results in the formation of three large, nonoverlapping subunit fragments that are readily resolved by SDS-PAGE (41,42), including one of 20 kDa (␣V8 -20) that begins at ␣Ser-173 and extends through the M1-M2-M3 helices and another of 10 kDa (␣V8 -10) that begins at ␣Asn-339 in the cytoplasmic domain and  (Table 1).

Table 1 Efficiency of [ 3 H]F 4 N 3 Bzoxy-AP photolabeling of amino acids in the Torpedo nAChR transmembrane domain
The photolabeling efficiency (cpm/pmol of the PTH-derivative) for each residue was calculated from the observed 3 H release, the initial peptide mass (I 0 ), and repetitive yield (R) as described under "Experimental procedures." Tabulated values are the mean (ϮS.D.), with the number of sequencing runs indicated in parentheses. With the exception of ␥Trp-453, the photolabeling efficiencies were determined from two (365 nm; ␣Cys-222 at 254 nm) or three (254 nm; M2-13Ј) independent preparative photolabelings. ND means not determined.

Labeled residues
Photolabeling efficiency 365 nm control/؉Carb 254 nm control/؉Carb a Because ␣Leu-223 immediately follows ␣Cys-222, which is photolabeled more efficiently, the efficiency of photolabeling of ␣Leu-223 cannot be calculated reliably.

Steroid-binding sites in an ␣␤␥␦ nAChR
extends through the M4 helix. Furthermore, digestion of ␣V8 -20 with EndoLys-C generates fragments readily separated by rpHPLC, one beginning at ␣His-186 that contains ␣M1 and a second beginning at ␣Met-243, the N terminus of ␣M2, that extends through ␣M3 (40). When an EndoLys-C digest of ␣V8 -20 isolated from nAChRs photolabeled at 254 nm in the presence of Carb was fractionated by rpHPLC ( Fig.  6A), the major peak of 3 H eluted at ϳ70% organic solvent B where the fragment containing ␣M1 is known to elute (40, 43), with little 3 H eluting in the more hydrophobic fractions where the fragment containing ␣M2 elutes. Sequence analysis of fractions containing the peak of 3 H (Fig. 6B) established the presence of the ␣His-186 fragment (I 0 ϭ 4 pmol), with no release of 3 H above background in 15 cycles of Edman degradation. Similarly, the fragment beginning at ␣Met-243 (I 0 ϭ 1 pmol), which eluted at ϳ85% organic solvent B, was sequenced without any 3 H release above background in 15 cycles of Edman degradation that extended to ␣M2-15Ј (Fig. 6C). These results indicated that the 3 H incorporation in ␣V8 -20 was likely to be within ␣M1.
To identify amino acids photolabeled in ␣M1, ␣ subunits were isolated from nAChRs photolabeled at 254 nm in the absence or presence of Carb and in the presence of Carb and promegestone. During sequencing of the fragment beginning at ␣Ile-210 (I 0 ϭ 60 pmol, each condition), isolated by rpHPLC from tryptic digests of nAChRs (36, 43), there was a peak of 3 H release in cycles 13/14 consistent with photolabeling of ␣Cys-222/␣Leu-223 (Fig. 6D). These residues were photolabeled at higher efficiency in the presence of Carb than in the absence (Table 1), and promegestone reduced the agonist-enhanced labeling by ϳ50%. To confirm photolabeling of ␣Cys-222/ ␣Leu-223, a fragment beginning at ␣Gln-208 was generated for sequencing by isolating the labeled ␣His-186 fragments from ␣ subunit EndoLys-C digests and treating them with cyanogen bromide to cleave at the C terminus of ␣Met-207 (43,44). When the ␣Gln-208 fragments were sequenced (Fig. 6E), the peak of Carb-enhanced 3 H release in cycle 15 confirmed labeling of ␣Cys-222. Sequence analysis of fragments beginning at ␣Ile-210 from an independent labeling experiment (ϩCarb and ϩCarb ϩ PMG) provided additional evidence that ␣Leu-223 was photolabeled as well as ␣Cys-222, because in this case the peak of 3 H release was in cycle 14 rather than 13 (Fig. 6F).

[ 3 H]F 4 N 3 Bzoxy-AP photolabeling in M4 and M3 helices
The nAChR M4 helices are most broadly exposed to lipid, with amino acids from the M1 and M3 helices also exposed at the lipid interface (42,45). Photolabeling studies with [ 3 H]promegestone, which forms a reactive ketyl radical upon irradiation at 312 nm, identified photolabeling of residues in ␣M4 (␣Cys-412 and ␣Cys-418), ␤M4 (␤Tyr-441 and ␤Cys-447), and ␥M4 (␥Cys-451 and ␥Trp-453) that are exposed at the nAChR-lipid interface and also photolabeled by small hydrophobic probes (26). Any [ 3 H]promegestone photolabeling of residues in ␤M2 or ␦M2, if it occurred, was at Ͻ10% the efficiency of any of the photolabeled cysteines. To identify photolabeling in ␣M4, we isolated by rpHPLC the fragment beginning at ␣Tyr-401, the ␣M4 N terminus, from trypsin digests of ␣V8 -10 isolated from nAChRs photolabeled at 254 and at 365 nm. No peaks of 3 H release above background were detected during sequence analyses of these fragments through 20 cycles of Edman degradation (Fig. 7, A and B), which indicated that labeling, if it occurred, was at Ͻ0.3 cpm/pmol, i.e. at Ͻ5% the efficiency of labeling of ␣Cys-222 but possibly at the labeling efficiency of ␦Val-269 (␦M2-13Ј).
Photolabeling in ␥M4 was characterized by a strategy similar to that used for ␣M4. The ␥-subunits were digested in gel with V8-protease to produce fragments of ϳ14 kDa (␥V8 -14) beginning at ␥Leu-373 and ␥Ile-413, and EndoLys-C digests of those fragments were fractionated by rpHPLC. For nAChRs photolabeled at 365 nm, sequencing of the fragment beginning at ␥Ala-450, the N terminus of ␥M4, revealed a peak of 3 H release in cycle 4, corresponding to photolabeling of ␥Trp-453 in the absence or presence of Carb at a photolabeling efficiency 4-fold higher than that of ␤Val-261 (␤M2-13Ј) (Fig. 7C and Table 1). In contrast, for nAChRs irradiated at 254 nm, photolabeling of ␥Trp-453, if it occurred, was at Ͻ10% the efficiency of ␤Val-261. This preferential photolabeling at 365 nm compared with 254 nm of ␥Trp-453, the only Trp in the nAChR TMD, may be the source of the increased ␥ subunit labeling seen after irradiation at 365 nm compared with 254 nm (Fig. 3).
Multiple lipid-exposed residues in the nAChR M3 helices have been photolabeled by small hydrophobic photoprobes (39,42), whereas other residues in ␥M3 contributing to a binding pocket at the ␥-␣ subunit interface have been photolabeled by

Steroid-binding sites in an ␣␤␥␦ nAChR
photoreactive etomidate and barbiturate analogs (36,46,47). For nAChRs photolabeled at 365 or 254 nm in the presence of Carb, we characterized [ 3 H]F 4 N 3 Bzoxy-AP photolabeling in ␥M3 and ␦M3 by sequencing for 30 cycles of Edman degradation fragments that begin at ␥Thr-276 and ␦Thr-281 and extend through the M3 helices. The protocol used (47) allowed efficient recovery of the fragments (I 0 ϭ 30 -90 pmol), but at either labeling wavelength, there were no peaks of 3 H release above background, which indicated that any labeling, if it occurred, was at Ͻ0.3 cpm/pmol.

Discussion
In this report, we use F 4 N 3 Bzoxy-AP, a photoreactive analog of alphaxalone and allopreganolone that is a potent general anesthetic and GABA A R PAM (27), to identify the locations of steroid-binding sites in the Torpedo nAChR in its native membrane environment. Many endogenous (glucocorticoids and progesterone) and synthetic (alphaxalone and promegestone) steroids act as noncompetitive antagonists of Torpedo and vertebrate skeletal muscle and neuronal nAChRs, but it is uncertain whether they act from the lipid interface or from sites within the nAChR (6,8). Promegestone, which potently inhibited Torpedo nAChR responses and the binding of a desensitized state selective channel blocker, photolabeled amino acids  3 H (E, control; •, Carb; ࡗ, Carb/promegestone) and PTH-derivatives (‚, control; OE, Carb; छ, Carb/promegestone) released while sequencing fragments begin at ␣Ile-210, the N terminus of ␣M1 (D and F), or at ␣Gln-208 (E). The fragment beginning at ␣Ile-210 was isolated by rpHPLC from trypsin digests, with the sequencing filters treated at cycle 2 with o-phthalaldehyde to prevent further sequencing of peptides not containing a proline at that cycle. The fragment beginning at ␣Gln-208 was isolated for sequencing by first isolating by rpHPLC from an EndoLys-C digest at the ␣His-186 fragment, which was sequenced for 18 cycles. The sequencing filter was then treated with cyanogen bromide to cleave at the carboxyl side ␣Met-207. The peaks of 3 H release in cycles 13 and 14 (D and F) or 15 and 16 (E) indicate photolabeling of ␣Cys-222 and ␣Leu-223 that was enhanced in the presence of Carb compared with control, and promegestone reduced the Carb-enhanced labeling.

Steroid-binding sites in an ␣␤␥␦ nAChR
at the nAChR-lipid interface without detectable photolabeling of amino acids in the ion channel (26). F 4 N 3 Bzoxy-AP also inhibits the binding of a channel blocker in the nAChR-desensitized state but not in the resting, closed channel state. In contrast to the photolabeling results for promegestone, we find that [ 3 H]F 4 N 3 Bzoxy-AP photolabels residues in the ion channel as well as at the lipid interface, a result providing direct evidence of an ion channel-binding site for steroids. The locations of the photolabeled amino acids in a homology model of the Torpedo nAChR structure (Fig. 8) establish that [ 3 H]F 4 N 3 Bzoxy-AP binds to three distinct sites in the nAChR TMD: (i) in the ion channel, photolabeling ␤Val-261 and ␦Val-269 in an agonistdependent manner; (ii) an intra-subunit site accessible from the lipid, identified by agonist-dependent photolabeling of ␣Cys-232/␣Leu-223, near the middle of the TMD; and (iii) at the lipid interface, photolabeling ␥Trp-453 near the cytoplasmic surface of the TMD. Based upon computational docking (Fig. 8), F 4 N 3 Bzoxy-AP is predicted to bind in the ion channel between positions M2-13Ј and M2-20Ј and between the ␣ subunit M1 and M4 helices, extending from ␣Cys-222 to the extracellular end of TMD.

Steroid binding in the ion channel
The agonist-enhanced photolabeling of ␤Val-261 and ␦Val-269 provides direct evidence that a steroid can bind in proximity to position M2-13Ј in the ion channel. Based upon the [ 3 H]TCP-and [ 3 H]tetracaine-binding assays, F 4 N 3 Bzoxy-AP binds in the ion channel in the desensitized state (IC 50 ϭ 30 M) with Ͼ10-fold higher affinity than in the closed channel state. As the structural changes in the ion channel associated with channel gating involve only small twists or tilts of the M2 helices, the residues that contribute to the lumen of the channel are the same in the closed, open, and desensitized states (50). The hydrophobic side chains of M2-13Ј, along with those at M2-9Ј, contribute to a hydrophobic plug preventing ion permeation in the closed channel state (37,51). Small charged (tetracaine) and uncharged (TID, benzophenone, and AziPm) drugs bind at that level preferentially in the absence of agonist (33,37,39,49). Similar to F 4 N 3 Bzoxy-AP, bulkier photoreactive etomidate and barbiturate analogs bind near M2-13Ј in the channel with higher affinity in the desensitized state (36,43,47).
Because more limited quantities of fragments containing ␣M2 or ␥M2 could be isolated for sequencing than ␤M2 or ␦M2, it remains possible that there is also unidentified labeling in those fragments at the same efficiency as ␦Val-269. The large quantities of ␤M2 and ␦M2 fragments allowed identification of agonist-dependent photolabeling in ␤M2 and ␦M2, even though it comprised a smaller fraction of subunit labeling than that seen for the ␣ subunit (Fig. 3). There was, however, agonist-enhanced PCP-inhibitable ␣ subunit labeling in nAChRs photolabeled at 254 nm, with the agonist-enhanced labeling inhibitable by 300 M promegestone. At the amino acid level, promegestone at 100 M reduced photolabeling of ␤Val-261 and ␦Val-269 by ϳ50%. Because promegestone is more potent than F 4 N 3 Bzoxy-AP as an inhibitor of [ 3 H]TCP binding (26), ϳ90% inhibition would be predicted if the two steroids bound in a mutually exclusive fashion. However, even PCP and chlorpromazine, two positively charged drugs, can bind simultaneously at different levels in the ion channel (52). Further studies are necessary to determine whether the inhibition of photolabeling is allosteric rather than competitive.

F 4 N 3 Bzoxy-AP binding at the nAChR-lipid interface
The photolabeling of ␥Trp-453 in ␥M4 provides evidence that F 4 N 3 Bzoxy-AP interacts with the same region of ␥M4 near the cytoplasmic end of the TMD as promegestone and a tricyclic aromatic probe, diazofluorene, which photolabel the same residue (26,53), and TID, which photolabels ␥Cys-451 (42). However, it was surprising not to find labeling of any positions in ␣M4, which is photolabeled at ␣Cys-412 and other residues by promegestone, diazofluorene, TID, and many general anesthetics (Fig. 8C). The lack of labeling of ␣Cys-412 cannot result from an inability of F 4 N 3 Bzoxy-AP to form a stable adduct with cysteines, because ␣Cys-222 in ␣M1 was one of the residues labeled most efficiently. One explanation for this unexpected result is that when F 4 N 3 Bzoxy-AP binds in proximity to ␣M4,  (Table 1). C, the peak of 3 H release in cycle 4 indicated photolabeling of ␥Trp-453 at 4-and 20-fold higher efficiency than the photolabeling in the ion channel of the M2-13Ј residues ␤Val-261 and ␦Val-269, respectively (Table 1).

Steroid-binding sites in an ␣␤␥␦ nAChR
the photoreactive azide incorporated at the steroid 11 position is oriented toward the lipid rather than the nAChR.
In the nAChR structure, the photolabeled residues in ␣M1 (␣Cys-222/␣Leu-223) are predicted to be accessible from the lipid, and computational docking predicts that F 4 N 3 Bzoxy-AP will intercalate between ␣M1 and ␣M4 in the outer half of the TMD (Fig. 8C). This is the location where cholesterol is predicted to be enriched in the outer leaflet of the lipid bilayer, based upon cryo-EM analyses of lipid distribution in the tubular vesicles formed from Torpedo nAChR-rich membranes (20). As interactions between the M4 and M1/M3 helices are important for channel function (19), our results indicate that binding of F 4 N 3 Bzoxy-AP or other steroids at this site can perturb cholesterol-nAChR interactions important for nAChR conformational transitions. Photolabeling of ␣Cys-222/␣Leu-223 was state-dependent, as evidenced by enhanced labeling in the presence of agonist, a result also seen for a convulsant barbiturate that binds in the same region, photolabeling ␣Val-218 (43). Consistent with this, mutations of ␣Val-218, ␣Pro-221, and ␣Cys-222 indicate that structural changes in this region are important determinants of nAChR gating and desensitization (54,55). The location of this intrasubunit-binding site for steroids between ␣M1 and ␣M4 identified by F 4 N 3 Bzoxy-AP is similar to a binding site for steroids identified by photolabeling in a prokaryotic pLGIC with a cation-selective ion channel, the homomeric Gloeobacter ligand-gated ion channel (GLIC) (56). In contrast, in a homomeric chimeric receptor containing a GABA A R ␣ subunit TMD, a binding site for pregnenolone sulfate, an inhibitory steroid, was identified between the M4 and M3 helices near the cytoplasmic end of the TMD (11).

Experimental procedures
Materials nAChR-rich membranes, containing 0.8 -1.1 nmol of [ 3 H] ACh-binding sites per mg of protein, were isolated from Torpedo californica electric organs (Aquatic Research Consultants, San Pedro, CA) as described (57) and stored in 38% (w/v) sucrose at Ϫ80°C until use. F 4 N 3 Bzoxy-AP and [ 3 H]F 4 N 3 Bzoxy-AP (45 Ci/mmol) were synthesized previously (27).  H]tetracaine was determined in the presence of 100 M Carb, PCP, or tetracaine, respectively. Stock solutions of F 4 N 3 Bzoxy-AP, alphaxalone, and promegestone were prepared at 60 mM in methanol, and all samples contained methanol at a final concentration of 0.5% (v/v). For each experiment, f x , the specifically bound 3 H (cpm total Ϫ cpm nonspecific ), was normalized to f 0 , the specifically bound 3 H in the absence of competitor. Individual experiments were carried out with duplicate samples, and data from independent experiments were combined for analysis. The concentration-dependent inhibition of [ 3 H]TCP binding was fit to the equation, f x /f 0 ϭ 1/(1 ϩ x/IC 50 ), where IC 50 is the total ligand concentration producing the half-maximal inhibition of binding. The numbers of independent experiments are noted in the figure legends.

[ 3 H]F 4 N 3 Bzoxy-AP photolabeling and gel electrophoresis
Conditions for photolabeling were identified by characterizing the absorption and photolysis characteristics of F 4 N 3 Bzoxy-AP in methanol. At the absorption maximum of 264 nm, the F 4 N 3 Bzoxy-AP extinction coefficient was 17,550 Ϯ 450 M Ϫ1 cm Ϫ1 . Photolysis with a 254-nm lamp (Spectronics EF160C) caused a progressive shift of the maximum absorption from 264 to 285 nm with a t1 ⁄ 2 of 20 s, associated with the formation of a broad secondary peak with an absorption maximum of ϳ340 nm. Photolysis with a 365-nm lamp (Spectronics EN-280L) decreased the absorption maximum with a t1 ⁄ 2 of 17 min. Based upon these results, nAChR-rich membranes equilibrated with [ 3 H]F 4 N 3 Bzoxy-AP were irradiated for 2 and 30 min with the 254-and 365-nm lamps, respectively. Before irradiation, Torpedo nAChR-rich membranes (2 M ACh sites; 2.5 mg of protein/ml in Torpedo saline buffer supplemented with 1 mM oxidized GSH as an aqueous scavenger) were incubated at 4°C with [ 3 H]F 4 N 3 Bzoxy-AP for 30 min and then for 30 min in the absence or presence of other ligands. nAChRs were photolabeled on analytical or preparative scales using 0.1 or 10 mg of protein per condition, respectively. After photolabeling, membrane polypeptides were resolved by Tris-glycine SDS-PAGE on gels composed of 8% polyacrylamide, 0.33% bisacrylamide, and membrane polypeptides were visualized with GelCode TM Blue Safe Protein Stain (Thermo Fisher Scientific). Invitrogen SeeBlue Plus2 pre-stained standards were used as molecular mass markers.
rpHPLC and sequence analysis rpHPLC was performed as described (47) using an Agilent 1100 binary rpHPLC system, a Brownlee Aquapore BU-300 column, and a mobile phase containing the aqueous solvent A (0.08% TFA in distilled water) and organic solvent B (3:2 acetonitrile/isopropyl alcohol supplemented with 0.05% TFA). Proteolytic digests were eluted at a flow rate of 0.2 ml/min using a nonlinear gradient, with organic solvent B increasing from 25 to 100% over 90 min with fractions collected every 2.5 min, and 3 H elution was determined by liquid scintillation counting of 10% of each fraction.
With the exception of rpHPLC fractions containing ␣M4, which were loaded onto polyvinylidene difluoride membrane filters at room temperature using Applied Biosystems ProSorb TM sample preparation cartridges, all rpHPLC fractions containing 3 H-labeled peptides were drop-loaded at 45°C onto Applied Biosystems Micro TFA glass fiber filters. After loading on filters, samples were treated with Biobrene Plus to stabilize the peptides and then subjected to Edman degradation sequencing on an Applied Biosystems Procise 492 protein sequencer. For some samples, as detailed in the figure legends, sequencing was interrupted at a predetermined cycle, and the filter was treated with o-phthalaldehyde to prevent further sequencing of any peptides not containing a proline at that cycle (57,59). As one method to characterize [ 3 H]F 4 N 3 Bzoxy-AP photolabeling in ␣M1, samples containing the fragment beginning at ␣His-186 were sequenced for 18 cycles, and filters were then treated with cyanogen bromide as described (44,60) to cleave at the C-terminal side of ␣Met-207 before ␣M1.
During N-terminal sequencing, either 1/6 or 2/3 of the material from each cycle of Edman degradation was used for phenylthiohydantoin (PTH)-derivative determination, and 5/6 (or 1/3) was collected to measure 3 H release by scintillation counting. The amount of PTH-derivative released in a given sequencing cycle (in picomoles) was determined by comparing the peak height of the derivative in the chromatogram to the height of its standard peak. I 0 , the initial amount of the peptide in the sample (in picomoles), was determined from the amounts of PTH-derivative in each cycle by using SigmaPlot 11 to the equation I x ϭ I 0 R x , where I x is the background-subtracted mass of the amino acid at cycle x, and R is the average repetitive yield of the Edman degradation. For samples containing multiple fragments, only PTH-derivatives unique to the fragment of interest were used in the fit, and amino acid derivatives whose amounts cannot be estimated accurately (His, Trp, Ser, Arg, and Cys) were omitted from the fits. When 2/3 of the material was used for PTHderivative determination, E(x), the photolabeling efficiency (cpm/pmol) of an amino acid residue in cycle x, was calculated according to the equation (2(cpm x Ϫ cpm (x Ϫ 1) ))/ (I 0 R x ), where cpm x is the 3 H release in cpm at cycle x. When 1/6 of the material was used for PTH-derivative determination, E(x) ϭ ((cpm x Ϫ cpm (x Ϫ 1) ))/(5 I 0 R x ).

Computational docking
A T. californica nAChR homology model was constructed based on the cryo-EM-derived structure of Torpedo marmorata nAChR (PDB code 2BG9) (45) using the Create Homology Model tool in Discovery Studio 2017 (Accelrys Inc., San Diego), with modifications introduced to correct for the previously identified error in assignment of amino acids in the M2 and M3 helices (61,62). To preserve the side-chain orientations of residues conserved between species, the nonconserved residues in each subunit (3, 17, and 11 in ␣, ␤, and ␦, respectively; in ␥, an added N-terminal Glu and eight residues) were mutated individually using the Build and Edit Protein Tool. To correct the placement of the M2 and M3 residues, the M1-M2 loops were shortened by four residues and the Create Homology Model tool was used to align ␣Thr-237-␣Ser-302 of the sequence with ␣Glu-241-␣His-306 from the structure, along with the equivalent alignments for the other subunits. The full structural model was minimized using CHARMm with the Generalized Born Implicit Membrane solvent model for 12 cycles (Smart Minimized method) to detect inappropriate residue placements and to reduce high-energy interactions (final energy, Ϫ66,779 kcal/mol). F 4 N 3 Bzoxy-AP was docked using 12-Å radius binding-site spheres centered as follows: 1) in the ion channel at the level of ␤Val-261 and ␦Val-269; and 2) at the lipid-␣ subunit interface at the level of ␣Cys-222 and ␣Leu-223. Each sphere was seeded with 12 distributed replicas of F 4 N 3 Bzoxy-AP with the CDOCKER module used to generate 40 molecular dynamicsinduced alterations for each replica, and then each altered structure was configured into 30 different starting orientations. The docking results for binding in the ion channel and at the lipid-protein interface between the ␣M1 and ␣M4 helices are shown as Connolly surface representations defined by a 1.4-Å diameter probe of the ensemble of 12 solutions with the lowest CDOCKER interaction energies. Similar docking results were obtained when F 4 N 3 Bzoxy-AP was docked in these sites of a T. californica homology nAChR model (43) based upon the crystal structure of an (␣4) 2 (␤2) 3 nAChR (Protein Data Bank entry 5KXI (63).