Discovery of an intrasubunit nicotinic acetylcholine receptor–binding site for the positive allosteric modulator Br-PBTC

Nicotinic acetylcholine receptor (nAChR) ligands that lack agonist activity but enhance activation in the presence of an agonist are called positive allosteric modulators (PAMs). nAChR PAMs have therapeutic potential for the treatment of nicotine addiction and several neuropsychiatric disorders. PAMs need to be selectively targeted toward certain nAChR subtypes to tap this potential. We previously discovered a novel PAM, (R)-7-bromo-N-(piperidin-3-yl)benzo[b]thiophene-2-carboxamide (Br-PBTC), which selectively potentiates the opening of α4β2*, α2β2*, α2β4*, and (α4β4)2α4 nAChRs and reactivates some of these subtypes when desensitized (* indicates the presence of other subunits). We located the Br-PBTC–binding site through mutagenesis and docking in α4. The amino acids Glu-282 and Phe-286 near the extracellular domain on the third transmembrane helix were found to be crucial for Br-PBTC's PAM effect. E282Q abolishes Br-PBTC potentiation. Using (α4E282Qβ2)2α5 nAChRs, we discovered that the trifluoromethylated derivatives of Br-PBTC can potentiate channel opening of α5-containing nAChRs. Mutating Tyr-430 in the α5 M4 domain changed α5-selectivity among Br-PBTC derivatives. There are two kinds of α4 subunits in α4β2 nAChRs. Primary α4 forms an agonist-binding site with another β2 subunit. Accessory α4 forms an agonist-binding site with another α4 subunit. The pharmacological effect of Br-PBTC depends both on its own and agonists' occupancy of primary and accessory α4 subunits. Br-PBTC reactivates desensitized (α4β2)2α4 nAChRs. Its full efficacy requires intact Br-PBTC sites in at least one accessory and one primary α4 subunit. PAM potency increases with higher occupancy of the agonist sites. Br-PBTC and its derivatives should prove useful as α subunit–selective nAChR PAMs.

Understanding the structural features of Br-PBTC's interactions with nAChRs will be crucial for developing subtype-selective nAChR PAMs. There are several methods to study ligandprotein interactions, including co-crystallization, cryo-electron microscopy (cryo-EM), photoaffinity labeling, mutagenesis, and docking (16,17). Co-crystallography and cryo-EM are the most direct methods for finding the Br-PBTC-binding site, but existing structures of ␣4␤2 nAChRs may not represent functional nAChRs because they were solved without lipids and with truncated cytoplasmic domains (18 -21). The lack of native lipid in structure determination complicates interaction of Br-PBTC with nAChRs. Lipid-nAChR interactions greatly influence conformational changes at the extracellular end of the transmembrane domain where Br-PBTC is likely to act (15). In addition, crystallization tends to resolve nAChRs in the most stable desensitized conformation (19,22), which may not carry a high-affinity Br-PBTC-binding site. Although type II PAMs can reactivate desensitized nAChRs, their effect on long-term desensitized nAChRs is limited (3,15,23,24). Therefore, they may not bind to the desensitized nAChRs obtained by crystallography. Photoaffinity labeling requires a photoreactive ligand. Modifying Br-PBTC to include photoreactive groups may reduce its binding affinity and produce false positives from nonspecific binding. To locate the Br-PBTC-binding site, we used mutagenesis of ␣4 subunits combined with a consensus docking approach. Consensus docking involves searching for binding sites identified by two different suites of docking software (25). Docking with one software can produce results that are biased by the idiosyncrasies of that docking software. Consensus docking can mitigate the biases of each individual software.
Our docking templates were homology models of ␣4 derived from corrected crystal structures of the Torpedo marmorata muscle-type nAChR structure in an open state (25). Torpedo remains the only Cys-loop receptor structure solved in its native membrane rather than after detergent solubilization (26,27). The registry of transmembrane domains was corrected by remapping the electron potential map, producing the template structure (25) that we use to create the ␣4 homology model in this study.
We located a Br-PBTC-binding site in an intrasubunit cavity between the extracellular ends of the four ␣4 transmembrane domains using consensus docking, which successfully predicted ␣4 mutations that reduce Br-PBTC activity: E282Q and F286S. We also discovered derivatives of Br-PBTC that can act on ␣5 subunits. In (␣4␤2) 2 ␣4, agonist occupancy of the accessory site at the ␣4/␣4 subunit interface increases Br-PBTC potency more than 9-fold. We find that Br-PBTC acts synergistically between the accessory and primary ␣4 subunits within (␣4␤2) 2 ␣4 to reactivate desensitized nAChRs.

Pharmacological comparison of Br-PBTC and dFBr
Recent studies of various nAChR PAMs suggest that Br-PBTC has a similar potentiation profile to those of 17␤estradiol and dFBr (15,28). Bermudez and co-workers (28) have shown that 17␤-estradiol and dFBr communicate with the Cysloop through Ile-601 in ␣4. When the last four ␣4 C-tail residues were mutated from AGMI to AAC, potentiation of ␣4␤2 activity by Br-PBTC decreased by 80% (15). These data suggest that Br-PBTC, dFBr, and 17␤-estradiol are all sensitive to the ␣4 C-tail (28).
We next investigated the effects of two mutants, I601F and I601W, on Br-PBTC potentiation ( Table 2). We chose to mutate isoleucine to aromatic amino acids because these mutants greatly reduced PAM activity of dFBr (28). Because PAM activity is evaluated in the presence of agonists, we studied whether these mutants affected activation by ACh. We chose the (␣4␤2) 2 ␤2 stoichiometry for initial screening of the mutants' effects to simplify the study because this stoichiometry only has one kind of agonist site, i.e. the high-sensitivity site at the ␣4/␤2 subunit interface (29 -31). Potencies of ACh were similar for I601F, I601W, and the WT. Maximum current was altered in some mutants, which could be due to changes in expression level. Unlike for dFBr (28), mutants I601F and

Binding site for the PAM Br-PBTC
I601W had minimal effect on Br-PBTC potentiation ( Table 2). I601F did not change the efficacy of Br-PBTC (Table 2), although it reduced the efficacy of dFBr 5-fold (28).
These data suggest that the binding site or mode of allosteric potentiation of Br-PBTC is slightly differently from that of dFBr and 17␤-estradiol. Experimental conditions in studies from Bermudez and co-workers are different from ours (28). They used a mixture of two stoichiometries at EC 10 concentrations of ACh to evaluate mutant effects on dFBr. We biased expression to the (␣4␤2) 2 ␤2 stoichiometry and used an EC 90 concentration of ACh. Our WT experiments using both PAMs on the two ␣4␤2 stoichiometries indicate that the two PAMs act similarly, but not identically.  (15). Br-PBTC is more potent and efficacious than dFBr at potentiating activation of (␣4␤2) 2 ␣4 nAChRs. As reported, Br-PBTC and dFBr showed decreased potentiation or even inhibition at 3 M and higher concentrations (15,60). These concentrations are not shown for comparison of the potentiation effect of the two compounds. C, reactivation effects on desensitized nAChRs by various concentrations of PAMs. Nicotine (0.5 M) was preincubated with cultured cells for 6 h before addition of PAMs. Br-PBTC greatly reactivated (␣4␤2) 2 ␣4 but had a very small effect on (␣4␤2) 2 ␤2 nAChRs. dFBr is less potent than Br-PBTC, but it partially reactivates both (␣4␤2) 2 ␤2 and (␣4␤2) 2 ␣4. Results are presented as mean Ϯ S.E., sample size n ϭ 3. Table 1 Potencies and efficacies of Br-PBTC and dFBr on activation and desensitization of (␣4␤2) 2 ␣4 and (␣4␤2) 2 ␤2 nAChRs As reported, Br-PBTC and dFBr showed decreased potentiation or even inhibition at 3 M and higher concentrations (15,61). These concentrations are not included in the analysis for comparison of the potentiation effect of the two compounds. When a curve failed to fit due to inhibition at higher concentrations, we defined the maximum efficacy (I max ) as the highest signal obtained, and we noted the I max value with # in the table and reported the confidence interval (CI) as not available (NA). Sample size is n ϭ 3.

Binding site for the PAM Br-PBTC
besides Glu-282, Phe-286, and the C-tail in ␣4 are also important for potentiation. These results suggest that Br-PBTC acts at a cavity in the transmembrane domain of ␣4 bordering the extracellular domain.

Consensus docking between SwissDock and Autodock Vina
We used a consensus docking approach (25,32,33) to locate the Br-PBTC-binding site within the ␣4 subunit. In consensus docking, we searched for the lowest root mean square difference (RMSD) between pairs of docked conformations from two different docking programs: SwissDock and Autodock Vina. Each pair contains one pose from the lowest energy cluster revealed by SwissDock and one pose from the top 20 poses reported by Autodock Vina. Both programs are free and widely used docking programs (34). Autodock Vina uses iterated local search and allows side chains to be flexible to predict binding poses of ligands within a customer-defined region of the protein. SwissDock uses exhaustive ranking and clustering of tentative binding modes in the vicinity of all target cavities. Lowenergy docked conformations and binding pockets identified by two docking programs are likely more accurate than the results of a single docking program (25,32,33). In addition, we hypothesize that dFBr is likely to bind in the same basic site as Br-PBTC because dFBr is sensitive to a Y283F mutation in ␣4 that is adjacent to E282Q and near a F286S mutation that inhibits Br-PBTC (28).
We first attempted to dock dFBr and Br-PBTC to the ␣4 subunit of one crystal and two cryo-EM structures of desensitized (␣4␤2) 2 ␤2 and (␣4␤2) 2 ␣4 (19, 22). We did not find a consensus pose using SwissDock and Autodock Vina within 10 Å of Glu-282 and Phe-286. Estimated binding energy was generally lower, and consensus RMSD was generally higher for those in desensitized ␣4␤2 structures compared with the ␣4 homology model based on the open-state Torpedo ␣1 structure (Table 3). Although the RMSD for Br-PBTC poses in the cryo-EM structure of (␣4␤2) 2 ␤2 is only 1.897 Å, Br-PBTC is too far from Glu-282 and Phe-286 in these poses to form any meaningful interactions. The charged nitrogen of Br-PBTC is 18.53 Å away from the side-chain oxygen in Glu-282 in Autodock Vina and 17.44 Å in SwissDock. The benzothiophene ring of Br-PBTC is more than 14.45 Å away from the phenyl ring of Phe-286 in Autodock Vina and 14.98 Å in SwissDock. The transmembrane domains in the desensitized structure were too closely packed near Glu-282 and Phe-286, preventing Br-PBTC from accessing these critical residues. This suggests the desensitized structure likely does not possess a Br-PBTC-binding site (Fig. 4). It Table 2 Effect of ␣4 mutations in (␣4␤2) 2 ␤2 on the efficacy and potency of Br-PBTC The (␣4␤2) 2 ␤2 nAChRs were expressed in Xenopus oocytes using ␣4 (wildtype or mutant) and ␤2 (wildtype) free subunits. Varying concentrations of Br-PBTC were co-applied with 500 M ACh. Results are presented as mean and 95% CI. ND represents values that cannot be determined because there is no observed PAM effect. Mutations E282Q and F286S significantly changed Br-PBTC potentiation as determined by two-way analysis of variance. F E282Q (1,8) ϭ 26 and P E282Q ϭ 0.0010; F F286S (1,8) ϭ 17 and P F286S ϭ 0.0035. Other mutations' effects were not statistically significant (p Ͼ 0.05). Efficacy of Br-PBTC is the maximum increase in peak response evoked by ACh in the presence of Br-PBTC compared with ACh alone. n represents number of oocytes tested.

␣4 mutation
ACh, EC 50 Figure 3. Location of ␣4 point mutations on the amino acid sequence and an ␣4 homology model. Mutated residues in the ␣4 subunit are shown as spheres with appropriate van der Waal radii. The ␣4 model shown here is derived from a revised Torpedo nAChR structure (25). Pink spheres represent carbon of side chains. The AGMI C-tail sequence is highlighted in red.

Binding site for the PAM Br-PBTC
is also known that Br-PBTC does not reactivate long-term desensitized (␣4␤2) 2 ␤2 (15), which suggests Br-PBTC may not bind to desensitized (␣4␤2) 2 ␤2. In addition, these crystal and cryo-EM structures could represent an unnatural "uncoupled" state due to detergent solubilization used to prepare the purified nAChR (18,22,25). Br-PBTC does reactivate desensitized (␣4␤2) 2 ␣4, but the effect is transient (15). There are multiple desensitized states (3,4), and Br-PBTC may not act on the conformation of desensitized (␣4␤2) 2 ␣4 resolved by cryo-EM. We next built an ␣4 homology model using the corrected cryo-EM structure of T. marmorata ␣ subunit in the open conformation, the only known nAChR structure solved in its native membrane (25). There is a registration mistake in the transmembrane domains in Unwin's structure, which is revised and remodeled in the corrected ␣1 structure (25). When we docked Br-PBTC and dFBr to our open-state ␣4 model, we saw an excellent consensus between SwissDock and Autodock Vina (Table 3). Both SwissDock and Autodock Vina placed dFBr and Br-PBTC derivatives within the same binding pocket that is formed by a triad of transmembrane domains: M1, M3, and M4 (Figs. 5A and 6A).
Both Autodock Vina and SwissDock position Br-PBTC and dFBr close to Glu-282 and Phe-286. Br-PBTC and dFBr bound with their aryl planes (benzothiophene or indole) oriented roughly parallel to the transmembrane helices. The positively charged piperidine moiety of Br-PBTC was closer to the extracellular domain, whereas the aromatic ends of Br-PBTC dock deeper into the transmembrane domain. dFBr locates its secondary amine at a similar level as its aromatic end. The docking site of dFBr is consistent with previous studies, which also observe dFBr docking in between the helix of Glu-282 and Phe-286 (28). The carboxylate group of Glu-282 is 7.20 Å away from Br-PBTC and 3.03 Å away from dFBr in Autodock Vina. In SwissDock, this distance is 4.21 Å for Br-PBTC and 2.83Å for dFBr. These suggest a possible coulombic interaction for binding of Br-PBTC and dFBr. Despite the relatively long distance, this salt bridge is plausible because it is shielded from solvent at the center of the binding pocket (35). The flanking hydrophobic residues from M1 and M4 could block solvent from screening and disrupting the electrostatic interaction between the charged nitrogen and Glu-282. This salt bridge could help explain the total loss of Br-PBTC PAM activity with the mutation E282Q. In the consensus poses for Br-PBTC and dFBr, the ligands' aromatic ring systems were nearly parallel with the aromatic face of Phe-286 (Figs. 5 and 6). This is consistent with a -stacking interaction (36) and explains the deleterious effect of F286S on Br-PBTC and dFBr potentiation (28). Consensus poses for both ligands are too far from Leu-275 and Thr-140 to form any significant interaction. This is consistent with mutagenesis data showing no effect of mutations L275D and T140R on Br-PBTC potentiation (Table 1).
Our consensus poses place Br-PBTC and dFBr too far from the AGMI C-tail sequence to form direct contact. Docked Br-PBTC and dFBr instead form direct contacts with M4 right before the C-tail begins. Our published results show that the C-tail sequence is critical for Br-PBTC function (15). Mutating the C-tail ␣4 from AGMI to AAC reduced potentiation by 80% (15). Instead of forming direct contacts with Br-PBTC, we believe C-tail mutations affect Br-PBTC potentiation by disturbing contacts between the C-tail and Cys-loop of ␣4, a possibility suggested in another study of dFBr (28).
Docking has limitations in finding precise bound conformations. Docking is heavily dependent on choice of receptor, treatment of ligand and side-chain rotations, and receptor backbone movements. Our homology modeling approach cannot accurately account for subtle differences in peptide backbone and tertiary structure between the template ␣1 subunit and ␣4 model. Docking is not a substitute for structural determination experiments, but by relying on consensus between multiple programs, we account for some idiosyncratic inaccuracies of each docking program and provide a reasonable hypothesis for Br-PBTC binding in the ␣4 subunit consistent with our mutagenesis data on Br-PBTC activity.

Br-PBTC derivatives acting on the ␣5 subunit
Promoting activation of ␣5 nAChR subtypes holds promise for the treatment of nicotine dependence (8, 10). Currently, there are no known ␣5-selective nAChR agonists or PAMs.

Binding site for the PAM Br-PBTC
Because Br-PBTC can act from a single ␣ subunit, and each nAChR naturally contains only one ␣5, Br-PBTC could be a good lead molecule for developing ␣5-selective PAMs. Therefore, we investigated whether Br-PBTC or its derivatives can act from the ␣5 subunit.
Based on our ␣4 model and sequence homology between ␣4 and ␣5, we hypothesized that a nonconserved tyrosine, Tyr-430, in the ␣5 M4/C-tail region (Fig. 3) is important for ␣5-selectivity for Br-PBTC derivatives. The ␣5 Y430A mutant was potentiated more by SR14270, SR14271, SR19678, and SR14273 than in WT ␣5, but not by SR13521 (Fig. 7E). Because not all ligands were affected by the Y430A mutant, the mutation's effects are likely specific for Br-PBTC derivatives potentiating ␣5 subunits rather than a nonspecific global effect on all PAMs. These data support the existence of a PAM site for Br-PBTC derivatives in the ␣5 subunit.
The position of Tyr-430 in the C-tail of ␣5 suggests that it is too far from Glu-282 and Phe-286 to directly contact the PAM. Because Tyr-430 is at the base of the ␣5 C-tail, Tyr-430 could be involved in transmitting the PAM effect to the rest of the subunit. This is similar to the role of the ␣4 C-tail that did not contact Br-PBTC in docking experiments. Despite the long distances from Br-PBTC in ␣4, mutations at Ile-601 and the AGMI C-tail sequence still affected Br-PBTC's PAM function. Tyr-430 in ␣5 may function similarly, ensuring the PAM effect is only transmitted for certain ␣5-selective PAMs. Y430A, like other M4/C-tail mutations, probably loosens selectivity between ␣5 and ␣4 subunits by subtly modifying the structural linkage that communicates PAM binding to the ion channel gating.

Binding site for the PAM Br-PBTC
Unique Br-PBTC pharmacology in the accessory ␣4 of (␣4␤2) 2 ␣4 It is known that accessory ␣ subunits in heteromeric nAChRs confer unique pharmacological properties to the pentamer (29). In ␣4␤2 nAChRs, there are two primary ␣4 subunits, each of which forms a high-affinity agonist-binding site with an adjacent ␤2 subunit. The fifth accessory subunit position forms a low-affinity agonist-binding site when it is occupied by an ␣4 or  Table 1 are colored olive green. All ligands dock with similar orientations, with the piperidine moiety placed next to Glu-282 and the aromatic ring buried further toward the cytoplasm. The lowest-energy conformational clusters of Br-PBTC (C) and dFBr (D) all fit in the same binding pocket within ␣4 between transmembrane domains M1, M3, and M4. SwissDock identifies hydrogen bonding (shown in green) between the carboxylate of Glu-282 and either the basic piperidine amine of Br-PBTC (C) or secondary amine of dFBr (D). Olive green denotes residues found to be critical for Br-PBTC potentiation: Glu-282 and Phe-286. The ␣-helix distortion near the intracellular end of M4 is due to an additional minimization step performed by SwissDock prior to docking. RMSD values between poses in Figs. 5 and 6 can be found in Table 2.  Table 2.

Binding site for the PAM Br-PBTC
no agonist site when it is occupied by a ␤2 subunit (19, 29 -31). Previously, we found that Br-PBTC potentiates activation of both stoichiometries of ␣4␤2. Now we find that activation of the (␣4␤2) 2 ␣4 stoichiometry selectively increases the potency of Br-PBTC.
To confirm involvement of the accessory ACh-binding site in increasing Br-PBTC's potency, we characterized potency of Br-PBTC in the presence of EC 90 ACh (500 M) or sazetidine-A (10 nM) using (␣4␤2) 2 ␣4 nAChRs expressed in oocytes. Sazetidine-A only binds primary agonist-binding sites formed by ␣4␤2 pairs, whereas ACh acts from both primary and accessory agonist sites (14). We found that the Br-PBTC EC 50 was 9-fold lower with ACh than with sazetidine-A (Fig. 9A). Even when the accessory PAM site was abolished by mutant E282Q, the EC 50 of Br-PBTC was still 17-fold lower in (␣4␤2) 2 ␣4 E282Q with ACh compared with sazetidine-A (Fig. 9B). This suggests that agonist binding to the accessory ACh-binding site causes the entire pentamer to enter a higher-affinity state for Br-PBTC. Single-channel studies support the existence of intermediate states between resting and open-state (␣4␤2) 2 ␣4 attributed to agonist binding at the accessory site (38,39). These intermediate states could explain how agonist binding at the accessory site leads to a high-affinity conformation of (␣4␤2) 2 ␣4 for Br-PBTC.
Another important effect of the accessory ␣4 is that it allows desensitized (␣4␤2) 2 ␣4 to be efficiently reactivated by Br-PBTC (15). The presence of an accessory ␣4 is critical for reactivation because very small reactivation by Br-PBTC was observed in (␣4␤2) 2 ␤2, which lacks an accessory ␣4 (15). We selectively deleted Br-PBTC sites on several ␣4 subunits in (␣4␤2) 2 ␣4 by inserting the E282Q mutation. This mutation did not greatly change the potency of agonist to activate nAChRs. Figure 9. Br-PBTC acts more than 10-fold more potently against (␣4␤2) 2 ␣4 and (␣4␤2) 2 ␣4 E282Q when co-applied with ACh rather than sazetidine-A (Saz-A). Activation by ACh increased sensitivity to potentiation by Br-PBTC, but sazetidine did not. ACh binds to primary and accessory ACh sites, but sazetidine binds only to primary ACh sites. Thus, activation of the accessory site is critical for agonist-induced sensitivity to Br-PBTC. Activation of the accessory ACh site increases sensitivity to potentiation by Br-PBTC even when the PAM-binding site on the accessory ␣4 subunit is blocked by the E282Q mutation. Xenopus oocytes were injected with 1:2 mass ratio of ␤2-␣4 ϩ ␣4 or ␤2-␣4 ϩ ␣4 E282Q mRNA, which yields the nAChR constructs A (␣4␤2) 2 ␣4 and B (␣4␤2) 2 ␣4 E282Q . Br-PBTC was co-applied with either saturating ACh (500 M) or saturating sazetidine-A (0.01 M). The percent increase in peak current was used to assess Br-PBTC potentiation. Pentamer diagrams are presented on the right of the Br-PBTC concentration response curve for each nAChR construct. ACh-, sazetidine-A-, and Br-PBTC-binding sites are noted in the graph. Results are presented as mean Ϯ S.E. Sample size is n ϭ 4.
The Br-PBTC-binding site we located overlapped well with that of other known nAChR PAMs that have been docked within the transmembrane domains (31,40). Some PAMs bind at similar intrasubunit cavities as Br-PBTC and dFBr (40), and some bind at intersubunit cavities in the transmembrane domains (25,40,41). A common transmembrane-binding region suggests that perturbation of the transmembrane domains is a critical structural feature needed to potentiate nAChR opening. It is known that nAChR function is very sensitive to conformational changes in the transmembrane domain M4, which directly connects to the C-tail (42,43). Subtle alterations in the tertiary structure of M4 could easily alter the conformation of the C-tail, thereby affecting nAChR function. It was shown that a single C418W mutation in M4 of ␣1 increased sensitivity to ACh by 16-fold (42). M4 is also believed to act as a "lipid sensor" in ␣1 (43). LY2087101, an ␣4and ␣7-selective PAM, is also sensitive to mutations at Glu-282 and Phe-286 in ␣4 like Br-PBTC (40). dFBr, an ␣4-selective PAM, is affected by mutations at Phe-286 and Ile-601 (28). 17␤-Estradiol, an ␣4*-selective PAM, has also been shown to depend on the ␣4 C-tail for its function (28). Swapping the C-tail of ␤2 with that of ␣4 allowed 17␤-estradiol to potentiate the chimeric ␤2 (44). This demonstrates the importance of conformational changes in ␣4 transmembrane domains for potentiating nAChR responses.

Binding site for the PAM Br-PBTC
Although there are ␣5-selective antagonists (46), SR14271 and SR14273 are the first PAMs that promote activation of ␣5-containing nAChRs. They need to be further optimized for ␣5-selectivity by eliminating their potentiation on ␣4-containing nAChRs.
We propose that Br-PBTC and similar nAChR PAMs that bind in the transmembrane domains potentiate nAChR opening by slowing down the closing of a desensitization gate in M2 and reactivate desensitized nAChRs by reopening this desensitization gate. In a published structure of desensitized (␣4␤2) 2 ␤2 (22), the five-leucine gate associated with activation near the middle of M2 creates an opening wide enough for a hydrated sodium cation to fit through, but another ring of five glutamates near the cytoplasmic end of M2 forms a ring too small to pass a hydrated cation. The presence of a desensitization gate was also observed in nAChRs and other Cys-loop receptors such as glycine and GABA A receptors (47,48). In (␣4␤2) 2 ␤2, the activation gate is shut in the resting state, and both the activation and desensitization gates are open in the conductive state; but in the desensitized state, the desensitization gate remains shut while the activation gate is held open. Because Br-PBTC can reactivate desensitized (␣4␤2) 2 ␣4, it must be opening the desensitization gate while keeping the activation gate open. Reactivation of long-term desensitized (␣4␤2) 2 ␣4 by Br-PBTC is transient (15). After reactivation, nAChRs again assume a desensitized state. The desensitization gate theory is consistent with the crystal structures of various Cys-loop receptors solved in open and desensitized states (47,48).
Although the (␣4␤2) 2 ␤2 and (␣4␤2) 2 ␣4 structures provided valuable structural information about ion channel gating (19,22), they were not optimal for docking our ligands. Br-PBTC does not reactivate long-term desensitized (␣4␤2) 2 ␤2, nor does it have a long-lasting effect on desensitized (␣4␤2) 2 ␣4 (15), and the structures are assumed to be in a desensitized state (19,22). Therefore, we are not surprised that Br-PBTC did not dock well to these structures. Newcombe et al. (25) found that an ␣7 model derived from the same crystal structure of (␣4␤2) 2 ␤2 bound to nicotine did not allow them to produce consensus docking results for their ␣7 PAMs. Similarly, we found that Br-PBTC could not dock to ␣4 subunits from the (␣4␤2) 2 ␤2 structure. They attributed this docking failure to conformational disruptions in (␣4␤2) 2 ␤2 crystals resulting from detergent solubilization. Detergent solubilization of muscle type nAChRs is believed to produce an "uncoupled" state (18). The extracellular domain and ion channel of the (␣4␤2) 2 ␤2 and (␣4␤2) 2 ␣4 structures may remain intact in detergent due to their water solubility, but the transmembrane domains are in intimate contact with lipids or detergent. This makes transmembrane domains especially sensitive to the chemical identity of membrane constituents. This suggests the (␣4␤2) 2 ␤2 and (␣4␤2) 2 ␣4 structure transmembrane domains may carry significant structural distortions. The transmembrane domains of an ␣4 homology model derived from a native-membrane muscle-type nAChR structure carry a sizable binding pocket, whereas the desensitized (␣4␤2) 2 ␤2 M1, M2, and M3 transmembrane domains are tightly compacted with the M4 domain extending away the nAChR body. The significant structural dif-ferences between the structures solved in native membrane and detergent suggest that detergent solubilization has significant effects on the transmembrane domain structure.
We used a single ␣4 subunit for docking because the position of Glu-282 in our ␣4 model and its homolog in the corrected Torpedo structure point toward an intrasubunit cavity and could not have reached into the intersubunit junction between ␣4 and ␤2 subunits. Br-PBTC's strong dependence on Glu-282 bounded our search area so that Br-PBTC could form a direct contact with Glu-282.
Based on our experiments with reactivation from desensitized states and lowering Br-PBTC's EC 50 in (␣4␤2)  In summary, we combined functional studies and mutagenesis with computational techniques to find a novel PAM-binding site in the transmembrane domains. A common dependence on Glu-282, Phe-286, and Ile-601 for Br-PBTC, dFBr, and LY2087101 potentiation combined with good overlap in docked Br-PBTC derivatives and dFBr suggest the binding site is likely to be accurate. Further refinement of our knowledge of the Br-PBTC-binding site will be important for designing new subtype-selective nAChR PAMs. Such PAMs may be useful for treating several neurological diseases.

Reagents
Synthesis of Br-PBTC and its derivatives were described previously (15,37). A85380 was purchased from Tocris Bioscience. Other reagents were purchased from Sigma or Thermo Fisher Scientific unless stated otherwise.

Binding site for the PAM Br-PBTC Mutagenesis
All mutations were made using the QuikChange TM hot-start PCR technique (52) with PfuUltra high-fidelity polymerase. An E282Q mutation was introduced into the ␣4 subunits of the ␤2-␣4 dimer and the ␤2-␣4-␤2 trimer with the same oligonucleotide used to generate E282Q in a single ␣4 subunit. All mutations were confirmed by sequencing.

Xenopus oocyte extraction
Oocytes were removed surgically from Xenopus laevis and defolliculated using published procedures (29).

Two-electrode voltage clamp on Xenopus oocytes
Currents were measured using the OpusXpress 6000A (Molecular Devices, Union City, CA), an automated two-electrode voltage-clamp amplifier that can record up to eight oocytes simultaneously (29). All oocytes were clamped to a holding potential of Ϫ50 mV. Fresh dilutions of drugs were made daily in recording buffer (ND-96: 96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 1.0 mM MgCl 2 , 5 mM HEPES, pH 7.6) with 0.5 M atropine. Agonists and PAMs were applied over 4 s at 3 ml/min to oocytes via the sidewall of each recording chamber. Between drug applications, oocytes received 253 s of buffer washing at a rate of 3 ml/min.
To compensate for variable nAChR expression in oocytes, peak current amplitudes of experimental responses were normalized to peak current amplitudes from ACh or nicotine. The extent of potentiation or Br-PBTC enhancement of nAChR opening was defined as the percent increase in peak current from co-application of Br-PBTC and agonist compared with application of the same concentration of agonist alone as shown in Equation 1. % potentiation ϭ 100% ϫ peak current with BrPBTC peak current agonist alone Ϫ 100% (Eq. 1) Dilutions of Br-PBTC and its derivatives in ND-96 were freshly prepared daily from 10 mM stock in DMSO. Mean and standard error were calculated for all concentrations of agonist and Br-PBTC. At least three oocytes were used per experimental group. When testing for Br-PBTC enhancement of nAChR opening, Br-PBTC was co-applied with an agonist, ACh, or nicotine. When testing for Br-PBTC's efficiency in reactivating desensitized nAChR, oocytes were perfused with 0.2 or 100 M nicotine in ND-96 with 0.5 M atropine at a rate of 2.0 ml/min. Concentration-response curves were fitted to the Hill equation to determine the EC 50 for any compound tested (15).

FLEXstation assay
Functional testing of nAChR was carried out in HEK cells permanently transfected with nAChR cDNA using a Flex-Station (Molecular Devices, Sunnyvale, CA) bench-top plate reader as described (53). All cells were maintained as described previously (51). Creation of HEK cell lines expressing defined stoichiometries of ␣4␤2 were described previously (15).
To enhance HEK cell expression of (␣4␤2) 2 ␣4 or (␣4␤2) 2 ␤2 nAChR, 96-well plates of cells were incubated at 29°C for 16 -24 h prior to testing to increase nAChR expression and thus enhance the fluorescent signal. Blue membrane potential sensitive fluorescent dye from Molecular Devices was used to measure membrane potential changes upon addition of agonists according to the manufacturer's instructions.

Homology modeling
The ␣4 subunit homology model was made from a corrected ␣ ␥ structure of Torpedo nAChR at open state (27) using SWISS-MODEL (54 -56). The sequence alignment for homology modeling was obtained from Clustal Omega (57). ␣ ␥ refers to the ␣ subunit that forms an ACh-binding site in conjunction with the ␥ subunit in Torpedo nAChR.

Docking
The ␣4 subunit in the desensitized nAChR crystal structure (PDB code 5KXI) and cryo-EM structures (PDB codes 6CNJ and 6CNK) were used directly in docking (22). Before docking, atomic coordinates of ligands were prepared using the "build structure" function in UCSF Chimera (58) and subjected to 5,000 steepest descent and 5,000 conjugated gradient minimization steps with a steepest descent and conjugate gradient step size of 0.02 Å.
In SwissDock, the ␣4 homology model and a minimized mol2 file of the desired ligand were directly input into the SwissDock server (59). Over a hundred docked ligand poses were sorted and clustered by location and conformational similarity. Each cluster was ranked by its average energy.
We used flexible docking function in Autodock Vina (version 1.1.2) (16) via a graphic interface MGLTools (Scripps Molecular Graphics Laboratory, San Francisco) (60). A 21 ϫ 24 ϫ 21 Å box encompassing the C-tail, Glu-282, Phe-286, and parts of each ␣4 transmembrane domain was set as the search space. Ten amino acid side chains in the search space were allowed to rotate freely: Tyr-220, Leu-224, Ile-266, Leu-275, Leu-279, Glu-282, Tyr-283, Phe-286, Leu-593, and Leu-597. Rotatable residues were distributed throughout the search space rather than concentrated in one region of the search space. This protects from biased docking toward the region containing the most rotatable side chains. Top 20 lowest energy poses were outputted for each ligand. To find a pair of consensus poses for a given ligand, we searched for the pose from Autodock Vina that had the lowest RMSD from the lowest energy SwissDock cluster.