A Novel Pharmatope Tag Inserted into the β4 Subunit Confers Allosteric Modulation to Neuronal Nicotinic Receptors*

α-Bungarotoxin, the classic nicotinic antagonist, has high specificity for muscle type α1 subunits in nicotinic acetylcholine receptors. In this study, we show that an 11-amino-acid pharmatope sequence, containing residues important for α-bungarotoxin binding to α1, confers functional α-bungarotoxin sensitivity when strategically placed into a neuronal non-α subunit, normally insensitive to this toxin. Remarkably, the mechanism of toxin inhibition is allosteric, not competitive as with neuromuscular nicotinic receptors. Our findings argue that α-bungarotoxin binding to the pharmatope, inserted at a subunit-subunit interface diametrically distinct from the agonist binding site, interferes with subunit interface movements critical for receptor activation. Our results, taken together with the structural similarities between nicotinic and GABAA receptors, suggest that this allosteric mechanism is conserved in the Cys-loop ion channel family. Furthermore, as a general strategy, the engineering of allosteric inhibitory sites through pharmatope tagging offers a powerful new tool for the study of membrane proteins.

␣-Bungarotoxin, the classic nicotinic antagonist, has high specificity for muscle type ␣1 subunits in nicotinic acetylcholine receptors. In this study, we show that an 11-amino-acid pharmatope sequence, containing residues important for ␣-bungarotoxin binding to ␣1, confers functional ␣-bungarotoxin sensitivity when strategically placed into a neuronal non-␣ subunit, normally insensitive to this toxin. Remarkably, the mechanism of toxin inhibition is allosteric, not competitive as with neuromuscular nicotinic receptors. Our findings argue that ␣-bungarotoxin binding to the pharmatope, inserted at a subunit-subunit interface diametrically distinct from the agonist binding site, interferes with subunit interface movements critical for receptor activation. Our results, taken together with the structural similarities between nicotinic and GABA A receptors, suggest that this allosteric mechanism is conserved in the Cys-loop ion channel family. Furthermore, as a general strategy, the engineering of allosteric inhibitory sites through pharmatope tagging offers a powerful new tool for the study of membrane proteins.
Nicotinic acetylcholine receptors (nAChRs) 1 are prototypical members of the superfamily of ligand-gated pentameric ion channels. They are important for peripheral and central nervous system function and have long been known to mediate synaptic transmission at the vertebrate neuromuscular junction (1). The biochemical and molecular characterization of muscle nAChRs, the first representatives of the superfamily to be characterized in molecular terms, has been greatly facilitated by the availability of highly specific snake venom-derived ␣-neurotoxins, such as ␣-bungarotoxin (Bgtx). These small proteins are high affinity competitive antagonists that interact with a subset of agonist binding determinants to disrupt agonist binding (2). Furthermore, the commercial availability of a large number of diversely labeled conjugates of Bgtx has led to major advances in the understanding of the biogenesis and developmental regulation of muscle nAChRs (3). Unfortunately, neuronal heteromeric nAChRs are not recognized by the ␣-neurotoxins, and comparably effective tools for their study are lacking.
Biochemical and structural studies show that the major determinants responsible for Bgtx binding to muscle nAChRs are located in Loop C of the ␣-subunit. The structure of the molluscan acetylcholine-binding protein (AChBP), a generally accepted model for the extracellular domain of homopentameric nAChRs, indicates that Loop C residues form a hairpin turn at the end of two conserved ␤ strand motifs (␤9 and ␤10). These ␤ strands lie on the outer perimeter of the ␣ subunit surface with the hairpin region in close proximity to the surface of the adjoining subunit. In heteromeric nicotinic receptors, the adjacent subunit is always a non-␣ (4 -6). There is considerable evidence that Bgtx binding to the Loop C region prevents the association of acetylcholine (ACh) to its binding site, also in close proximity to Loop C, and as a consequence, the conformational changes leading to channel gating are precluded (7)(8)(9)(10).
In contrast to muscle nAChRs, vertebrate heteromeric neuronal nAChRs are insensitive to Bgtx inhibition and lack high affinity Bgtx binding sites (11). We have shown that as few as 5 amino acids from Loop C of the muscle type ␣1 subunit, when substituted into the homologous region of the Bgtx-insensitive neuronal ␣3 subunit, are sufficient to impart functional sensitivity to Bgtx (12). These studies, together with observations that Bgtx binds with high affinity to isolated short peptide fragments derived from ␣1 Loop C, led us to speculate that short sequences derived from ␣1 Loop C might be effective in transferring Bgtx sensitivity to a more distantly related target, such as the nicotinic ␤4 subunit, one of the widely expressed neuronal non-␣ subunits. If successful, we reasoned that such a sequence, which we term a pharmatope in analogy to epitope tagging, could provide a general tool for introducing pharmacological sensitivity into membrane proteins (2).
In the study reported here, we wanted to determine whether insertion of an ␣1-derived pharmatope sequence into the neuronal ␤4 subunit could produce a high affinity Bgtx binding site. We tested this in Xenopus oocytes using functional neuronal heteromeric nAChRs produced by heterologous expression of the chimeric ␤4 subunit along with the Bgtx-insensitive ␣3 subunit. Two experimental ␤4 constructs were prepared: one with the pharmatope inserted en bloc into the L1 loop (4), near the N terminus, and the other with the pharmatope replacing endogenous residues in the homologous Loop C region of ␤4. We asked whether Bgtx binding to these ectopic sites, distant from the subunit interfaces responsible for ACh binding, could inhibit ACh-evoked activity. We also wanted to determine by independent means whether a reporter Bgtx conjugate could be bound to surface receptors containing either of the two pharmatope-tagged ␤4 constructs. The observed pharmacological sensitivity introduced through one of the two pharmatope insertions provides novel insights into the conformational changes that couple activation of neuronal nAChRs to channel opening, and we emphasize for the first time an important role for the subunit interfaces distinct from those forming the agonist binding site.

MATERIALS AND METHODS
Chimeric Subunit Constructs-The ␣3 cDNA construct used in this study was in pcDNA3.1/Zeo, whereas ␤2 and ␤4 were in the pGEMHE vector provided by Charles Luetje. After linearization by restriction enzyme digestion, DNA was purified by ethanol precipitation, and in vitro RNA transcription was performed with the mMESSAGE mMA-CHINE kit from Invitrogen. cRNA was purified by lithium chloride precipitation and stored at Ϫ80°C. Chimeras were generated by QuikChange PCR mutagenesis (Stratagene, La Jolla, CA), and the resulting plasmids were transformed into XL-1 Blue Supercompetent cells. Sequences were determined by PCR cycle sequencing.
Oocyte Preparation and Injection-Oocytes were collected from mature Xenopus laevis by survival surgery and were prepared for injection essentially as described previously (12). Oocytes were injected with 46 nL of cRNA and maintained in antibiotic-supplemented buffer for 2-6 days at 15°C before recording. Equal parts of ␣ to ␤ subunit cRNA (0.125 g/l unless otherwise noted) were diluted into OR2 buffer (115 mM NaCl, 2.5 mM KCl, 1.8 mM CaCl 2 , 10 mM HEPES, 0.003 mM atropine sulfate, pH 7.4).
Electrophysiological Recordings-ACh-evoked currents were measured using the two-electrode voltage clamp method with a Warner OC-725C. Electrodes of resistance 0.5-2.0 megaohms were filled with 3 M KCl, and recordings were performed in a Warner RC-3Z chamber with an OC-725 bath clamp headstage attached. The flow of various drug solutions in OR2 into the chamber was regulated by solenoid valves driven by a Warner BPS-8 controller. The flow rate of the perfusion fluid was adjusted to 5 ml/min by gravity feed. As noted by others, neuronal nAChRs can display rundown with repeated agonist application (35). For each subunit combination, we perform rundown control determinations by measuring currents for the same dose (ϳEC 50 ) of ACh given 6 -10 times over a period similar to that used for the experimental data acquisitions. In cases in which rundown is observed, values for ACh-evoked inward currents are corrected by linear interpolation. Co-application experiments were used to measure the onset of toxin block. After collecting initial responses evoked by a control dose of ACh, oocytes were incubated in a solution containing the desired Bgtx concentration for 15 min. To measure recovery from toxin block, ACh responses were collected from each oocyte prior to and after exposing the oocyte to Bgtx. Perfusion to wash out the Bgtx was commenced, and cell responsiveness to 100 M ACh was determined at various times.
Membranes (derived from 10 oocytes, at 10 -50 mg of protein/ml) were incubated with 125 I-Bgtx (5 nM) for 5 h at room temperature in incubation buffer (140 mM NaCl, 20 mM sodium phosphate, 1 mM EDTA, 1 mM EGTA, 0.5 mM phenylmethylsulfonyl fluoride, 200 units of aprotinin) at a final volume of 400 l (4 -20 mg of protein). Nonspecific binding was determined in the presence of 1 M Bgtx. For the determination of competition by ACh, the acetylcholine esterase inhibitor diisopropyl fluorophosphate (200 M) together with ACh (100 M) were added to the membranes prior to the addition of 125 I-Bgtx. At the end of the incubation period, the suspension was diluted with 5 ml of ice-cold wash buffer (10 mM NaCl, 10 mM sodium phosphate, pH 7.4) and filtered through a double layer of DE81 filters soaked for 4 h in 20% bovine serum albumin in ND50 buffer. Radioactivity was determined on triplicate samples using a gamma counter.
Binding Assays with a Fluorescent Conjugate of Bgtx-Intact oocytes expressing chimeric nAChRs were standardly incubated overnight in ND96 buffer (12) with 100 nM Alexa Fluor 647®-Bgtx, although significant binding could be observed after 1 h of incubation. The solution containing unbound Bgtx conjugate was removed, and the oocytes were rinsed five times before being examined by confocal scanning fluorescence microscopy using a filter set optimized for fluorescence emission at 633 nm. Nonspecific binding was determined by the addition of 1 M Bgtx prior to the addition of the Bgtx conjugate. No fluorescence was observed with oocytes expressing ␣3 and ␤4 subunits or with uninjected oocytes.

Two Sites in the Extracellular Domain of ␤4 Targeted for
Pharmatope Insertion-Initially, we compared the Loop C amino acid sequence from an ␣-bungarotoxin-sensitive ␣1 subunit found in a skeletal muscle type of nAChR (i.e. from Torpedo californica electric organ) with comparable regions from the toxin-insensitive rat ␣3 and ␤4 subunits. As shown in Fig.  1a, the putative Loop C region of ␤4 is highly variant but anchored by flanking invariant residues. In addition, it contains 4 fewer amino acid residues than Torpedo ␣1 and 3 less than rat ␣3. Also, ␤4 Loop C lacks the adjacent Cys residues (Cys-192-Cys-193; Torpedo ␣1 numbering) characteristic of all nicotinic ␣ subunits (1). We replaced 7 contiguous residues from the presumed turn 7 of Loop C in ␤4 (i.e. VNPQDPS) with a sequence that we expected, based on NMR studies of receptor peptide fragments, to be sufficient to generate a high affinity Bgtx binding site (8,13,14). Our candidate pharmatope sequence (WVYYTCCPDTP) was derived from the tip of Loop C in the Torpedo ␣1 subunit. The chimeric ␤4/␣1 [11]Loop-C subunit (Fig. 1a) contains a net addition of 4 amino acids in the Loop C region and is identical in length to the Loop C in ␣1. We produced a second chimeric construct, ␤4/␣1 [11]Helix, by inserting the same sequence, WVYYTCCPDTP, into a site distant from Loop C. We chose the L1 region very near the N terminus of the extracellular domain of ␤4. In the structure of the AChBP, the L1 loop follows a region (residues 1-13) that is ␣-helical (Fig. 1b). In nAChRs, the homologous L1 loop is most likely situated near the cusp of the receptor vestibule (4). The structures of related ligand-gated receptor subunits, from glycine receptors and GABA C receptors, accommodate natural sequence insertions in this region (4).
The Pharmatope Sequence Is Well Tolerated-Heterologous expression in Xenopus oocytes was used to prepare functional heteromeric nAChRs containing the ␤4 pharmatope constructs. As observed previously, the ␤4 subunit alone is incapable of forming functional ACh-gated ion channels; co-expression of an ␣ subunit such as ␣3 or ␣4 is required (11). Therefore, cRNA corresponding to each ␤4 chimera was injected into Xenopus oocytes along with cRNA encoding the wild-type, Bgtx-insensitive rat ␣3 subunit. Neuronal nAChR (␣3␤4)-mediated currents were measured in the oocytes using the two-electrode voltage clamp recording technique as described previously (12). The ␤4/␣1 [11]Loop-C chimera, when co-expressed with ␣3, generated ACh-evoked currents comparable in size to those of wild- type ␤4 (Table I). ␤4/␣1 [11]Loop-C alone or when co-expressed with wild-type ␤4 did not generate functional nAChRs in oocytes (data not shown). The insertion of ␣1 Loop C residues does not convert the ␤4 subunit into an ␣ subunit. As with wild-type ␤4, the ␤4/␣1 [11]Loop-C chimera must co-assemble with an agonist binding ␣-type subunit to generate functional nAChRs (15). Furthermore, the nAChRs comprised of the ␤4/ ␣1 [11]Loop-C chimera and ␣3 were activated by ACh as in wild-type channels. We obtained an apparent EC 50 for ACh of 121 M, a value comparable with the EC 50 of ϳ100 M reported for wild-type ␣3␤4 nAChRs (12). Our results indicate that insertion of the pharmatope sequence, WVYYTCCPDTP, within the Loop C region of ␤4 does not disrupt subunit synthesis or co-assembly with the ␣3 subunit, critical steps required for generation of functional surface receptors. Pharmatope insertion into the L1 region (␤4/␣1 [11]Helix construct) did not affect cell surface expression of functional receptors upon co-expression with ␣3 but did decrease the apparent EC 50 for ACh ϳ4-fold (Table I). Further investigations would be required to address the mechanisms by which this alteration in the L1 loop, which is distant from the ACh binding pocket, brings about the observed change in the EC 50 .
The Pharmatope Inserted within Loop C of ␤4 Confers Bgtx Sensitivity-After confirming that wild-type ␣3␤4 receptors are insensitive to Bgtx (11,12) (Table I), we tested the toxin sensitivity of ACh-evoked currents in oocytes expressing ␣3 and the chimeric ␤4/␣1 [11]Loop-C subunit. Bgtx blocked AChevoked responses in oocytes expressing ␣3 and ␤4/␣1 [11]Loop-C subunits with remarkably high affinity. The IC 50 for Bgtx block was 34 nM (Table I; Fig. 2a). In addition, we calculated the apparent Hill coefficient for Bgtx block of the ␤4/ ␣1 [11]Loop-C chimera to be ϳ0.7, consistent with occupancy of a single Bgtx binding site per functional receptor being sufficient for functional inhibition.
Next, we tested the reversibility of Bgtx inhibition of ␤4/ ␣1 [11]Loop-C following removal of the toxin from the oocyte perfusion. Bgtx block was long-lived, with an apparent t1 ⁄2 for dissociation of ϳ140 min (Fig. 2b). In comparison, Bgtx block of native muscle type nAChRs expressed in oocytes is essentially irreversible over the same time interval (16).
Bgtx Inhibition of nAChRs Containing ␤4/␣1 [11]Loop-C Does Not Involve the ACh Binding Site-Bgtx inhibits native muscle nAChRs by competitive antagonism at the agonist binding site; numerous studies have shown that agonists such as ACh inhibit radiolabeled ␣-neurotoxin binding to muscle-type nAChRs competitively (17). To investigate the mechanism of toxin block in oocytes expressing ␣3 and chimeric ␤4/ ␣1 [11]Loop-C, we performed competition binding assays on isolated oocyte membranes using 125 I-Bgtx (18). In control experiments, we confirmed that 100 M ACh decreases Bgtx binding to native ␣1 muscle-type receptors by ϳ90% (Fig. 3). ACh binding to the agonist binding site prevents association of Bgtx with binding determinants that are common to the two ligands (2, 7-9). In contrast, 100 M ACh has no significant effect on 125 I-Bgtx binding to oocyte membranes containing nAChRs composed of ␣3 and ␤4/␣1 [11]Loop-C subunits. These results show that the functional Bgtx binding site in ␤4/ ␣1 [11]Loop-C-bearing receptors is structurally separate from the agonist binding site. Bgtx must therefore inhibit AChevoked responses in ␤4/␣1 [11]Loop-C-bearing receptors by an allosteric mechanism, not by steric occlusion of the agonist binding site.
Bgtx Binds to Intact Oocytes Expressing nAChRs Containing the ␤4/␣1 [11] Helix Subunit-In contrast to ␤4/␣1 [11]Loop-C, which renders nAChRs sensitive to Bgtx inhibition, the chi-  [11]Loop-C by Bgtx and recovery following removal of Bgtx. a, oocytes expressing ␣3 and ␤4/ ␣1 [11]Loop-C subunits were incubated with the indicated concentration of Bgtx for 15 min prior to the application of three test doses of 100 M ACh. Control ACh-evoked responses were obtained prior to Bgtx addition and were used to normalize the data. The data shown represent the mean from five replicate determinations. b, oocytes expressing ␣3 and ␤4/␣1 [11]Loop-C subunits were incubated with 200 nM Bgtx for 15 min, after which the Bgtx was removed by bath perfusion. At the times indicated, responses to test applications of 100 M ACh were recorded. The data represent the mean of triplicate determinations and are fit to a linear function given that the experimental time points were in the linear range of the expected exponential decay. The error bars indicate the standard error of the mean. meric ␤4/␣1 [11]Helix subunit does not (Table I). We next determined whether Bgtx binds to ␤4/␣1 [11]Helix-bearing nAChRs but without inhibiting the channel function. We incubated intact oocytes with a fluorescent derivative of Bgtx, Alexa Fluor 647®-Bgtx. Specific binding of the fluorescent Bgtx was evident in oocytes expressing either the ␤4/␣1 [11]Helix subunit or the ␤4/␣1 [11]Loop-C subunit along with wild-type ␣3 (Fig.  4). In both cases, fluorescent Bgtx binding could be blocked by an excess of unlabeled Bgtx, consistent with specific binding. Specific binding was not observed with uninjected oocytes, with oocytes expressing wild-type ␣3 and ␤4, nor with oocytes expressing ␤4/␣1 [11]Loop-C alone. In addition, the rates of dissociation of fluorescent Bgtx from oocytes expressing either of the two pharmatope-tagged ␤4 subunits were similar (data not shown). These results indicate that Bgtx binds to the pharmatope sequence introduced into either the N-terminal L1 region or the Loop C region of ␤4, but only in the latter position does it inhibit channel activation. DISCUSSION We have shown that 11 contiguous amino acids from Loop C of the muscle-type ␣1 subunit have the capability to generate high affinity binding sites for Bgtx when introduced ectopically into the extracellular region of the non-␣ nAChR subunit, ␤4. Although insertion of the pharmatope sequence into the Loop C region produces Bgtx binding and functional block, the same pharmatope produces high affinity Bgtx binding but no detectable pharmacological activity when inserted into the N-terminal L1 loop.
The ability of Bgtx to block ACh-evoked responses in nAChRs formed from co-expressed ␣3 and ␤4/␣1 [11]Loop-C subunits ( Fig. 2; Table I) is unexpected based on the conventional view of Bgtx as a competitive inhibitor of ACh. Our results suggest that we have created an allosteric inhibitory site in ␤4. Furthermore, our findings are entirely consistent with structural considerations based on the crystal structure of the ACh-binding protein (Protein Data Bank accession code 1I9B), a homo-pentamer homologous to the extracellular domain of nAChRs (4). Bgtx, the prototype of the ␣-neurotoxins, is a competitive inhibitor at the agonist binding site, which is situated at the subunit-subunit interface of the ␣ subunit.(1, 2, 7-10, 13) As revealed in the crystal structure of the AChbinding protein, the agonist binding site consists of three loops, A, B, and C, from the principal (ϩ) face of the ␣ subunit and three loops, D, E, and F, contributed from the complementary (Ϫ) face of the adjacent subunit (e.g. ␥, ⑀, or ␦ in the case of the muscle nAChR or the ␤4 subunit, as in this study). The introduction of a Bgtx-binding pharmatope into Loop C of the ␤4 subunit would be expected to direct Bgtx binding to the (ϩ) face of the ␤4 subunit and diametrically away from the agonist binding site located at least one subunit interface away at the (Ϫ) face of the ␤4 subunit (Fig. 5). The distance from one subunit interface to the next, along the circumference of the extracellular domain at the level of Loop C, is predicted to be ϳ40 -50 Å. In comparison, Bgtx extends ϳ40 Å in its longest dimension and, therefore, is incapable of interacting simultaneously with the pharmatope-tagged ␤4 Loop C and with Loop C of the ␣3 subunit. Consequently, no direct competitive interaction between ACh and Bgtx is expected based on the current structural model. The physical separation of the Bgtx binding site from the agonist binding site in ␤4/␣1 [11]Loop-C-bearing nAChRs offers the advantage that the full range of amino acid mutations can now be used to study the contribution of individual amino acids in Loop C to Bgtx binding. This has been technically difficult due to the overlap in determinants for Bgtx and agonist binding (19). In addition, mutational strategies can be used to improve further the affinity of Bgtx for the ectopic binding site.
The proposed structural relationship of the Bgtx binding site relative to the agonist binding site in ␤4/␣1 [11]Loop-C-bearing nAChRs (Fig. 5) is strikingly reminiscent of the spatial relationship between the allosteric regulatory site in ligand-gated GABA A receptors and the agonist (GABA) binding site. Nicotinic and GABA A receptors belong to the same superfamily of Cys-loop ligand gated ion channels and share many structural features (1,4,20). In GABA A receptors, so-called inverse agonists of the ␤-carboline drug class bind at a subunit interface that is also responsible for the binding of the classic allosteric regulator, benzodiazepine (4, 20 -24). Binding of benzodiazepine to this site potentiates the action of the neurotransmitter GABA, whereas binding of inverse agonists to the same site diminishes responsiveness to GABA, but not in a competitive manner.
Current GABA A receptor models suggest that the inverse agonist binding site is at a subunit interface distinct from the GABA binding site and is formed by the Loop C region at the (ϩ) face of the GABA A ␣ subunit (e.g. ␣1) and the (Ϫ) face of the adjoining ␥ subunit (e.g. ␥2 in the case of ␣1␤2␥2, the most abundant GABA A receptor subtype in rat brain) (20,21). Meanwhile, the (Ϫ) face of the ␣ subunit contributes to the adjoining GABA binding site in a manner homologous to the predicted contribution of the (Ϫ) face of the nicotinic ␤4 subunit to the ACh binding site in neuronal nAChRs (4, 20 -24). The exact mechanism by which binding of inverse agonists to the benzodiazepine site in GABA A receptors produces inhibition of GABA-evoked currents is unknown, but a decrease in the frequency of GABA-evoked channel openings has been observed (25). The structural similarities between nicotinic and GABA A receptors suggest a mechanism common between the action of inverse agonists on GABA A receptors and the inhibition produced by Bgtx on ␤4/␣1 [11]Loop-C-bearing nAChRs. Furthermore, although several nicotinic potentiating ligands with allosteric characteristics have been described, their sites of action remain to be fully elucidated (26,27). Our findings, together with analogies with the benzodiazepine site, would suggest that nicotinic non-␣ subunit interfaces should receive more attention as potential targets for drug discovery.
What is the molecular mechanism responsible for the allosteric inhibition (Figs. 2 and 3) observed with chimeric ␤4/␣1 [11]Loop-C-bearing nAChRs? Based on the types of structural consider- ations described above, we propose that Bgtx interferes with rigid body-like subunit rotations that are normally essential for gating in neuronal nAChRs (Fig. 6). Such rotations along the subunit axis of quasi-symmetry have been suggested for the homomeric nAChRs, but to date, there is no direct evidence for this from the AChBP crystal structures (6,28). Based on electron microscopic studies of membranes enriched in muscle-type nAChR, it has been proposed that ACh binding evokes ϳ15°axial rotations of the extracellular domain in the two ␣ subunits of the receptor (29 -32). However, no similar rotations in the non-␣ subunits, ␤1, ␥, and ␦, were observed, leading to models that emphasize the functional importance of ␣1 subunit rotations. Whether neuronal ␤ subunits undergo ACh-evoked subunit rotations is currently unknown, but it is notable that the extracellular domains of neuronal ␤ subunits have several features that make them more similar to ␣ subunits than to the non-␣ subunits found in muscletype nAChRs.
Neuronal nAChR ␤ subunits, unlike the muscle-type non-␣ subunits, contain the conserved Trp-149 (Torpedo numbering) and Tyr-93 (although ␤3, like ␣5, has a conserved aromatic Phe at this position) found in all ␣ subunits, and their Loop F regions, located at the (Ϫ) face of the subunit interface, are identical in length to that of ␣ subunits. This suggests that neuronal ␤ subunits may be more likely to undergo subunit rotations akin to those seen in muscle-type ␣ subunits. If such rotations are essential for gating, then a wrench-in-the-workslike interference due to a stabilization of the resting state subunit interface by bound Bgtx could be the mechanism underlying the functional blockade mediated by the ␤4/␣1 [11]-Loop-C subunit (Fig. 6).
Alternatively, it is possible that Bgtx may prevent rotation of the ␣3 subunit by stabilizing the resting state ␤(ϩ)␣(Ϫ) interface of chimeric ␤4/␣1 [11]Loop-C subunit-containing nAChRs. In either case, the functional blockade of ␤4/␣1 [11]Loop-C subunit-containing nAChRs by Bgtx supports the view that rigid body-like subunit movements at subunit interfaces play an important role in receptor gating. Whether such movements FIG. 4. Binding of Alexa Fluor 647®-Bgtx to intact oocytes expressing chimeric nAChRs. Intact oocytes expressing ␣3 and ␤4/ ␣1 [11]Loop-C subunits (A) or ␣3 and ␤4/␣1 [11]Helix subunits (B and C) were incubated overnight with Alexa Fluor 647®-Bgtx. Unbound Bgtx was removed by five buffer exchanges, and the oocytes were immediately examined by confocal scanning fluorescence microscopy. Nonspecific binding (C) was determined by the addition of 1 M Bgtx prior to the addition of the fluorescent-Bgtx conjugate. Similar nonspecific binding was observed with oocytes expressing ␣3 and ␤4/␣1 [11]Loop-C subunits and was comparable with the total binding observed with uninjected oocytes.  (4), and the position of bound Bgtx is based on NMR studies using a Bgtx-binding peptide fragment of the rat ␣7 homomeric receptor (8). The extracellular domain of the ␤4/ ␣1 [11]Loop-C subunit is shown on the left, contributing its (ϩ) face to the Bgtx binding site, where Loop C, with Bgtx in close proximity below, is shown extending toward the ␣3 subunit on the right. The vertical stripes in Loop C indicate the approximate position of the introduced pharmatope sequence. The rotating arrows at the top of the figure depict the clockwise direction of subunit rotation that has been observed with Torpedo ␣1 subunits upon agonist activation (30). The exposed loop, L1, immediately following the N-terminal helical region is highlighted in the ␤4/␣1 [11]Loop-C subunit.
are restricted to ␣ subunits as suggested previously for muscletype nAChRs or whether neuronal ␤ subunits also undergo conformational movements of their extracellular domains remains to be fully elucidated.
Previous studies have shown that neuronal nAChRs are pentameric with an expected stoichiometry of two ␣ subunits and three ␤ subunits (33,34). This would suggest that our ␤4 chimera generate three potential Bgtx binding sites per receptor: two at the ␤(ϩ)␣(Ϫ) interfaces and one at the ␤(ϩ)␤(Ϫ) interface (Fig. 5). The Hill coefficient of ϳ0.7, estimated from our dose-response analysis, argues that Bgtx needs to bind to only a single site on the receptor complex to produce functional inhibition. We do not presently know whether all three potential Bgtx binding sites per ␤4/␣1 [11]Loop-C-bearing receptor are capable of binding toxin and producing blockade. We have observed, however, that ␤4/␣1 [11]Loop-C-bearing receptors with ␣4 subunits substituting for ␣3 are somewhat less sensitive to Bgtx inhibition, suggesting that the (Ϫ) face of the ␣ subunit may directly contribute to the introduced Bgtx binding site. 2 This would suggest that it is binding to the ␤(ϩ)␣(Ϫ) interface that is responsible for the observed functional block.
The successful introduction of a Bgtx-binding pharmatope into the extracellular domain of the ␤4 subunit suggests that similar modifications can be made to the other neuronal nAChR ␤ subunits, and in preliminary studies, we find that insertion of a pharmatope sequence into Loop C of the rat ␤2 subunit also imparts functional Bgtx sensitivity. 2 In addition, in those cases in which high affinity Bgtx binding is conferred in the absence of pharmacological sensitivity, such constructs would be of considerable utility for studies of receptor localization and metabolism given the many commercially available reporter group derivatives of Bgtx. The relatively small size of Bgtx (ϳ8 kDa) as compared with antibodies may offer a further advantage in tissue penetrance and accessibility. Finally, our findings offer the possibility that similar pharmatope sequences can be similarly inserted into the extracellular domains of other ligand-gated ion channels or, more broadly, into other membrane proteins for which appropriate probes are lacking.