(cid:1) -Conotoxin BuIA, a Novel Peptide from Conus bullatus , Distinguishes among Neuronal Nicotinic Acetylcholine Receptors*

Nicotinic acetylcholine receptors (nAChRs) are pentameric ligand-gated ion channels. (cid:1) Subunits, to-gether with (cid:2) 2 and/or (cid:2) 4 subunits, form ligand-binding sites at (cid:1) / (cid:2) subunit interfaces. Predatory marine snails of the genus Conus are a rich source of nAChR-tar-geted peptides. Using conserved features of the (cid:1) -cono-toxin signal sequence and 3 (cid:1) -untranslated sequence region, we have cloned a novel gene from the fish-eating snail, Conus bullatus ; the gene codes for a previously unreported (cid:1) -conotoxin with unusual 4/4 spacing of amino acids in the two disulfide loops. Chemical synthesis of the predicted mature toxin was performed. The resulting peptide, (cid:1) -conotoxin BuIA, was tested on cloned nAChRs expressed in Xenopus oocytes. The peptide potently blocks numerous rat nAChR subtypes, with highest potency for (cid:1) 3- and chimeric (cid:1) 6-containing nAChRs; BuIA blocks (cid:1) 6/ (cid:1) 3 (cid:2) 2 two-step oxidation the peptides the disulfide bridge between Cys 2 and Cys by the peptide into an equal of m M ferricyanide, 0.1 M 7.5. The allowed to react for 30 min, and the monocyclic peptide was purified by reverse- phase HPLC. Simultaneous removal of the S -acetamidomethyl groups closure of the disulfide bridge between Cys 3 and Cys 13 was carried out by iodine oxidation. The monocyclic peptide and HPLC eluent was

Acetylcholine acts on nicotinic acetylcholine receptors (nAChRs) 1 to mediate fast excitatory neurotransmission or to modulate neurotransmitter release. nAChRs appear to be involved in pain sensation, attention, memory, learning, and development (1). Medications that affect nicotinic transmission may be useful for the treatment of pain, memory disorders, Parkinson's disease, schizophrenia, and nicotine addiction. To maximize the ratio of therapeutic benefit to side effects, medications must discriminate between a plethora of subtypes of nAChRs.
Neuronal nAChRs are ligand-gated cationic channels composed of ␣ and, in many cases, ␤ subunits. These pentameric proteins have at least two ligand binding sites located at the interface of two subunits. For neuronal nAChRs, the ligand binding ␤ subunit appears to be either ␤2 or ␤4. Both of these subunits are widely expressed, often within the same areas of the nervous system (reviewed in Refs. 2 and 3). Probes to distinguish among different ␣ and ␤ subunit-containing nAChRs are needed to identify which subtypes underlie the particular effects of nicotine.
Predatory marine snails of the genus Conus utilize nAChR antagonists to immobilize and capture their prey. The ϳ500 species of cone snails prey upon a broad diversity of organisms (five different phyla). Each Conus species appears to have a unique complement of nAChR antagonists, making Conus a rich source of novel ligands, often with unique specificities. In this report, we describe the cloning of a gene that encodes a novel ␣-conotoxin that distinguishes among ␣ subunit-containing nAChRs and kinetically discriminates between ␤2and ␤4-containing nAChRs.

EXPERIMENTAL PROCEDURES
Identification and Sequencing of a cDNA Clone Encoding ␣-Conotoxin BuIA-cDNA was prepared by a reverse transcription of RNA isolated from the Conus bullatus venom duct as described previously (4). The resulting cDNA served as a template for PCR using oligonucleotides corresponding to the conserved signal sequence and the 3Ј-UTR sequence of ␣-conotoxin prepropeptides. The resulting PCR product was purified using the HIGH-PURE PCR product purification kit (Roche Applied Science) following the suggested protocol of the manufacturer. The eluted DNA fragment was annealed to plasmid pAMP1 vector, and the resulting product was transformed into competent DH5␣ cells, with the clone AMP pAMP System for Rapid Cloning of Amplification Products (Invitrogen) following the suggested protocols of the manufacturer. The resulting product was transferred into competent DH5␣ cells as described (4). The nucleic acid sequences of the resulting clones were determined according to the standard protocol for Sequenase Version 2.0 DNA Sequencing kit as described (5).
Chemical Synthesis-The peptide was synthesized, 0.45 mmol/g, on a Fmoc amide resin using Fmoc chemistry and standard side protection except on cysteine residues. Cys residues were protected in pairs with either S-trityl on Cys 2 and Cys 8 , or S-acetamidomethyl on Cys 3 and Cys 13 . The peptide was removed from the resin and precipitated. A two-step oxidation protocol was used to selectively fold the peptides as described previously (6). Briefly, the disulfide bridge between Cys 2 and Cys 8 was closed by dripping the peptide into an equal volume of 20 mM potassium ferricyanide, 0.1 M Tris, pH 7.5. The solution was allowed to react for 30 min, and the monocyclic peptide was purified by reversephase HPLC. Simultaneous removal of the S-acetamidomethyl groups and closure of the disulfide bridge between Cys 3 and Cys 13 was carried out by iodine oxidation. The monocyclic peptide and HPLC eluent was dripped into an equal volume of iodine (10 mM) in H 2 0/trifluoroacetic acid/acetonitrile (78:2:20 by volume) and allowed to react for 10 min. The reaction was terminated by the addition of ascorbic acid, diluted 20-fold with 0.1% trifluoroacetic acid, and the bicyclic peptide was purified by HPLC on a reverse-phase C 18 Vydac column using a linear gradient of 0.1% trifluoroacetic acid 0.092% trifluoroacetic acid, 60% acetonitrile, remainder H 2 O.
Mass Spectrometry-Measurements were performed at the Salk Institute under the direction of Jean Rivier. MALDI-TOF/MS (matrixassisted laser desorption ionization time-of-flight mass spectrometry) was utilized.
Voltage Clamp Recording-Oocytes were harvested and injected with cRNA encoding nAChR subunits as described previously (7). The oocyte recording chamber was fabricated from Sylgard and was 300 l in volume. Oocytes were gravity-perfused with ND96 (96.0 mM NaCl, 2.0 mM KCl, 1.8 mM CaCl 2 , 1.0 mM MgCl 2 , 5 mM HEPES, pH 7.1-7.5) containing 1 M atropine (ND96A) with or without toxin at a rate of ϳ5 ml/min. All solutions also contained 0.1 mg/ml bovine serum albumin to reduce nonspecific adsorption of peptide. The perfusion medium could be switched to one containing peptide or ACh by use of a distributor valve (SmartValve, Cavro Scientific Instruments, Sunnyvale, CA) and a series of three-way solenoid valves (model 161T031, Neptune Research, Northboro, MA). All recordings were made at room temperature (ϳ22°C). ACh-gated currents were obtained with a twoelectrode voltage clamp amplifier (model OC-725B, Warner Instrument, Hamden, CT). Glass microelectrodes, pulled from fiber-filled borosilicate capillaries (1-mm outer diameter Åϳ 0.75-mm inner diameter) (WPI, Sarasota, FL) and filled with 3 M KCl, served as voltage and current electrodes. Resistances for voltage and current electrodes were 0.5-3.0 M⍀ and 0.5-2.0 M⍀, respectively. The membrane potential was clamped at Ϫ70 mV, and the current signal, recorded through virtual ground, was low pass-filtered (5 Hz cut-off) and digitized at a sampling frequency of 20 Hz. Data acquisition, measurement of peak responses, and control of the distributor and solenoid valves were automated by a homemade virtual instrument constructed with the graphical programming language LabVIEW (National Instruments).
To apply a pulse of ACh to the oocyte, the perfusion fluid was switched to one containing ACh for 1 s. This was automatically done at intervals of 1-5 min. The shortest time interval was chosen such that reproducible control responses were obtained with no observable desensitization. This time interval depended on the nAChR subtype being tested. The concentration of ACh was 100 M. The ACh was diluted in ND96A for tests of all nAChR subtypes except ␣7, in which case the diluent was ND96. For control responses, the ACh pulse was preceded by perfusion with ND96 (for ␣7) or ND96A (all others). No atropine was used with oocytes expressing ␣7, because it has been demonstrated to be an antagonist of these receptors (8). For responses in toxin (test responses), the perfusion solution was switched to one containing toxin while maintaining the same interval of ACh pulses. Toxin was continuously perfused until responses reached a steady state. All ACh pulses contained no toxin, for it was assumed that little, if any, bound toxin would have washed away in the brief time (Ͻ2 s) it takes for the responses to peak. The average peak amplitude of three control responses just preceding exposure to toxin were used to normalize the amplitude of each test response to obtain "% response" or "% block." Each data point of a dose-response curve represents the FIG. 1. Prepropeptide and encoded toxin of ␣-BuIA. Post-translational processing sites are indicated by the arrows, and the mature toxin is indicated. The arrow following the K is a putative proteolytic processing site. The glycine following the C-terminal cysteine in the mature toxin is assumed to be processed to a C-terminal amide.  Table I. Cloning of Mouse ␣3, ␤2, and ␤4 cDNAs-The mouse ␤2 cDNA was cloned as described previously (9). For the cloning of the ␣3 and ␤4 cDNAs, first strand cDNAs were generated from 2 g each, DBA/2Ibg adrenal gland total RNA and P19 teratocarcinoma cell total RNA using AMV reverse transcriptase (Promega, Madison, WI), 1ϫ reverse transcriptase buffer (Promega) 200 M each, dATP, dCTP, dGTP, and dTTP, 5 mM MgCl 2 and 2.5 M random hexamers (Promega). cDNA synthesis was performed for 1 h at 42°C. Following synthesis, the first strand cDNAs were purified using the Qiagen (Valencia, CA) PCR purification protocol according to the manufacturer's instructions. The ␣3 (1612 bp) and ␤4 (1578 bp) cDNAs subsequently were amplified from the DBA/ 2Ibg adrenal gland first strand cDNA and P19 teratocarcinoma cell first strand cDNA, respectively, using Pfu Turbo Polymerase (Stratagene, La Jolla, CA), 1ϫ Pfu buffer, 200 M each, dATP, dCTP, dGTP, dTTP, and 0.4 M each, ␣3and ␤4-specific forward and reverse primers. ␣-BuIA was perfusion-applied to oocytes expressing the indicated rat nAChRs as described under "Experimental Procedures." After block was complete, toxin was washed out and response to ACh monitored. Note that nAChRs with the same ␣ subunit but different ␤ subunit have very different recovery rates; in all cases, the respective ␤4-containing receptor has the longest recovery.
Construction of Rat ␣6/␣3 Chimera-cDNA clones encoding rat nAChR subunits were provided by S. Heinemann (Salk Institute, San Diego, CA) with plasmid constructs as described (10). The rat ␣6 subunit does not express with the rat ␤2 subunit (11). We therefore used a chimera that contains the N-terminal extracellular ␣6 subunit sequence linked to the remaining portion of the ␣3 subunit protein as a model of activity at ␣6 subunit-containing nAChRs. The ␣6/␣3 chimera was provided by James Garrett (Cognetix Inc., Salt Lake City, UT). The chimeric nAChR consists of amino acids 1-237 of the rat ␣6 subunit protein linked to amino acids 233-499 of the rat ␣3 subunit protein.
The chimeric junction is located at the paired arginine residues immediately preceding the M1 transmembrane segment of the ␣3 subunit. The resulting chimeric nAChR represents the extracellular ligandbinding domain of the ␣6 subunit linked to the membrane-embedded region of the ␣3 subunit. The ␣6/␣3 cDNA was constructed by introduction of BspEI sites at the chimeric junction into the ␣6 and ␣3 cDNA sequences using mutagenic primers to introduce the restriction sites through silent codon changes. The ␣6 and ␣3 segments were generated by PCR of rat brain cDNA clones using primers in the 5Ј-and 3Јuntranslated regions of the corresponding cDNAs along with the internal mutagenic primers. The PCR products were digested with BspEI and ligated to generate the chimeric construct. The final chimeric construct was cloned and completely sequenced to confirm the correct cDNA sequence. The protein sequence was an exact match to the rat ␣6 and ␣3 sequences in GenBank TM , except for a valine to alanine change at amino acid 278 in the chimeric construct. To further improve expression levels, most of the 5Ј-and 3Ј-untranslated regions of the nAChR cDNA were deleted, leaving only 12 bp of 5Ј-UTR and 34 bp of 3Ј-UTR sequence. The chimeric construct was cloned into the Xenopus expression vector pT7TS, placing Xenopus globin 5Ј-and 3Ј-UTR regions around the nAChR cDNA. The expression construct, pT7TS/r␣6␣3, was transcribed with T7 RNA polymerase to generate sense-strand RNA for oocyte expression. To improve expression levels of the ␣6/␣3 chimera, it was co-injected with ␤2 and ␤4 subunits (provided by Charles Luetje) that were engineered into the pGEMHE high expression vector as described (12).

Cloning of ␣-Conotoxin
BuIA-In common with other known families of conotoxins, the ␣-conotoxins are proteolytically processed from a larger precursor protein. In the case of the ␣-conotoxins, this prepropeptide is ϳ40 amino acids long, with the mature ␣-conotoxin moiety of ϳ13-18 amino acids located at the C terminus of the precursor. A processing site, usually consisting of a basic amino acid, immediately precedes the mature toxin in the precursor sequence. An unusual feature of the conotoxins is that while the mature toxin peptides are highly variable in sequence, the precursor proteins are highly conserved. The signal sequence region is practically invariant among the different ␣-conotoxin precursors, and this remains true even for phylogenetically distant Conus species (13,14). Also, the 3Ј-untranslated region of the ␣-conotoxin mRNA is highly conserved. We utilized the conserved features of the ␣-conotoxin gene structure to design oligonucleotide primers for polymerase chain reaction amplification of the ␣-conotoxincoding region. The resulting cDNA clone from C. bullatus is shown in Fig. 1.
Chemical Synthesis of ␣-BuIA-Solid phase chemical synthesis of the predicted mature toxin was undertaken. The glycine at the C terminus was assumed to be post-translationally modified to a C-terminal amide. It was also assumed that the disulfide bridging of ␣-conotoxin BuIA was analogous to all previously characterized ␣-conotoxins, that is, Cys 2 to Cys 8 and

␣-BuIA Distinguishes among Neuronal nAChRs
Cys 3 to Cys 13 ; Cys groups were orthogonally protected in pairs to direct disulfide bond formation in this configuration. Acidlabile S-trityl was removed simultaneously with peptide cleavage from the resin and closure of the disulfide bridge between these Cys residues was accomplished with FeCN. The monocyclic peptide was purified by HPLC, and the acid-stable acetomidomethyl groups were removed, and the disulfide bridges formed by iodine oxidation. The folded peptide was subsequently analyzed with matrix-assisted laser desorption mass spectrometry. The mass of the synthetic peptide was consistent with the amidated sequence (monoisotopic MH ϩ : calculated, 1311.5; observed, 1311.4).
Peptide Effect on nAChRs-␣-Conotoxin BuIA was tested on various subunit combinations of neuronal nAChRs heterologously expressed in Xenopus oocytes. Concentration response analysis indicated that, unlike some ␣-conotoxins, ␣-BuIA was active against a broad spectrum of nAChR subtypes. The ␣ nAChR subunit had a profound influence on the effect of BuIA. The peptide was most potent on nAChRs containing the extracellular portion of the ␣6 subunit and next most potent on the closely related ␣3 subunit-containing nAChRs. The ligand had little effect on ␣4␤2 nAChRs (IC 50 Ͼ10 M). Results are shown in Fig. 2 and Table I.
Effects of BuIA on ␤2versus ␤4-containing nAChRs-The rate of unblock by toxin was monitored subsequent to washout of ligand. The results are shown in Fig. 3. Recovery from toxin block was markedly slower for ␤4 versus ␤2 subunit-containing nAChRs. This effect is particularly noticeable when nAChRs containing the same ␣ subunit are compared. We further investigated this effect on mouse ␣3␤2 versus mouse ␣3␤4 nAChRs, and human ␣3␤2 and human ␣3␤4 nAChRs. As was the case for rat nAChRs, the rate of recovery from toxin block was significantly slower for ␤4versus ␤2-containing nAChRs (Fig. 4).
Although the off-rate for ␤4-containing nAChRs was much slower, surprisingly in several instances the corresponding IC 50 values of ␣-BuIA for ␤2-containing nAChRs was lower than that of the ␤4-containing receptors. Since affinity is a ratio of off-rate to on-rate, this implies that the on-rate for ␤2-containing nAChRs is faster than that of the corresponding ␤4-containing nAChRs. Time to steady state for toxin block of rat ␣3␤2 and ␣3␤4 nAChRs was examined by perfusing the toxin over the oocyte and then assessing a response to ACh (Fig. 5). The results are consistent with a markedly faster on-rate for the ␤2-containing receptor. For ␣x␤2 nAChRs the recovery t1 ⁄2 was longest for ␣6/␣3␤2 nAChRs. The K i of functional block of ␣6/␣3␤2 nAChRs, calculated from k off /k on (Fig. 6A) is consistent with the IC 50 determined from concentration response analysis (0.69 nM versus 0.26 nM respectively). Kinetics of block and unblock were difficult to accurately quantitate for ␣x␤4 receptors due to the very long on-and off-times combined with limitations of oocyte life span and a tendency for the ACh response to drift over extended periods of time. However, kinetic constants for the ␣2␤4 nAChR were determined (Fig. 6B) and the calculated K i of functional block by BuIA (62.9 nM) is consistent with the IC 50 value (121 nM) determined by concentration response experiments. We further assessed kinetic constants for all ␣x␤4 nAChRs by determining k obs of block by three different toxin concentrations (Fig. 6C). The k on and k off for ␣2␤4 were 1.99 ϫ 10 5 min Ϫ1 M Ϫ1 and 0.0125 min Ϫ1 , respectively, and this compares well to the value determined from the toxin wash in and washout experiments (1.84 ϫ 10 5 min Ϫ1 M Ϫ1 and 0.0106 min Ϫ1 , Fig. 6B). The K i values determined by this method were also consistent with IC 50 values determined by concentration response analysis (Fig. 6C and Table II). However, whereas the on-rates determined by this method had reasonable 95% confidence intervals, the offrates did not. The off-rate is determined by the y-intercept and for these toxins the y-intercept is near the origin (Fig.  6C). When the confidence interval includes the origin, the off-rate range becomes infinite. Therefore, as an additional check, we estimated off-rates by using the k on value determined by linear regression analysis of k obs values (Fig. 6C) and by assuming that the K i is approximately equal to the IC 50 . k off values determined by this method were consistent with those obtained by the analysis of k obs values (Table II). For ␣2␤2, ␣3␤2, and ␣4␤2 nAChRs the toxin off-rate was rapid (t1 ⁄2 Ͻ 1 min, see Fig. 2); we therefore were unable to further quantitate k off for these receptor subtypes due to receptor desensitization that occurs from ACh exposure more frequent than once per minute. DISCUSSION C. bullatus is found from Mozambique and Zanzibar to Marquesas and Hawaii. It lives in muddy sand, coral rubble and gravel, often beneath dead coral rocks outside and inside the reef. It is known to feed on fish and molluscs, and is preyed upon by skates and stingrays as well as mollusceating cone snails (15). In this report, we describe the first FIG. 5. Observed on-rate for ␣-BuIA. A, 10 nM ␣-BuIA was perfusion-applied to oocytes expressing rat ␣3␤2 nAChRs as described under "Experimental Procedures." Time to equilibrium was examined by perfusing toxin over the oocyte at a toxin concentration 2-4ϫ greater than the IC 50 , and then assessing response to a 1-s pulse of ACh every 1-2 min. B, 100 nM ␣-BuIA was perfusion-applied to Xenopus oocytesexpressing rat ␣3␤4 nAChRs. The time course was monitored as described in A. Data are from three oocytes. Error bars are the S.E. ␣-BuIA Distinguishes among Neuronal nAChRs toxin to be isolated from this species. The toxin-encoding gene is homologous to peptides that are members of the A-superfamily of Conus toxins. This superfamily consists of peptides that act on nAChRs, potassium channels and sodium channels (14,16,17).
␣-Conotoxin BuIA is a 13-amino acid peptide with homology to ␣-conotoxins isolated from other cone species. It is unusual however in the spacing between Cys residues. Previously isolated ␣-conotoxins fall into three broad categories. There are those that are referred to in the literature as having ␣3/5 spacing, indicating that there are three and five amino acids, respectively, between Cys residues in the two loops of the toxin; these are paralytic toxins isolated from Indo-Pacific cone snails that hunt fish, and potently inhibit the muscle nicotinic receptor. A second group is made up of those toxins having ␣4/7 spacing; these conotoxins predominately target neuronal subtypes of nicotinic receptors. The third group are the ␣4/3 peptides, isolated thus far only from Conus imperialis (18,19) (Table III). ␣-Conotoxin BuIA has unusually broad specificity for different subtypes of neuronal nAChRs compared with previously characterized ␣-conotoxins. It is unknown whether the unique ␣4/4 spacing influences specificity.
Total chemical synthesis of the new peptide was carried out assuming the disulfide bond configuration of previously characterized ␣-conotoxins. We note that native ␣-conotoxin BuIA has never been isolated from venom. Therefore, it is possible that there are post-translational modifications present in the native peptide that could influence the properties reported for the synthetic peptide described in this report.
Another striking feature of ␣-conotoxin BuIA is its ability to discriminate, on the basis of off-rate kinetics, between ␤2and ␤4-containing nAChRs. Off-rates are very slow for ␤4-containing nAChRs, in contrast to the relatively rapid off-rates for ␤2-containing receptors. The ␤2 and ␤4 subunits are found within both the central and peripheral nervous systems. Within the central nervous system, the transcripts for the ␤2 subunit have widespread expression, whereas the ␤4 subunit expression is more restricted. The areas within the brain where these two subunits are co-localized include the habenula-interpeduncular pathway, the locus ceruleus, and a few structures within the sensory and motor areas of the brainstem (25,26). Within the peripheral nervous system, both ␤2 and ␤4 subunits are found in the autonomic ganglia (27,28) and contribute to autonomic ganglionic neurotransmission (29 -31). Knockout studies in mice suggest a more prominent role for the ␤4 subunit in regulation of certain visceral functions, such as cardiac and intestinal autonomic regulation and bladder contractility (29,31,32). However, mice that lack either the ␤4 or the ␤2 subunit grow to adulthood with no visible phenotypical abnormalities, thus indicating a degree of redundancy between 0.00212 min Ϫ1 . C, toxin was perfusion applied to oocytes at three different concentrations for each nAChR subtype. The response to a 1-s pulse of ACh was monitored once per minute. Since k obs ϭ k on ϫ F ϩ k off , where F ϭ the free toxin concentration, the graph of k obs versus F has a slope of k on and a y-intercept of k off .
FIG. 6. Kinetics of block and unblock by ␣-BuIA. A, ␣-BuIA at 0.3 nM was perfusion-applied (triangle) to oocytes expressing rat ␣6/␣3␤2 nAChRs and subsequently toxin was washed out (square). The response to a 1-s pulse of ACh every 1 min was assessed. An individual experiment is shown. Data averaged from 3 oocytes gave a k on of (2.77 Ϯ 0.59) ϫ 10 8 min Ϫ1 M Ϫ1 . The k off was 0.168 Ϯ 0.022 min Ϫ1 . B, ␣-BuIA at 150 nM was perfusion applied to rat ␣2␤4 nAChRs and response to a 1-s pulse of ACh was monitored every 2 min. Results from a single experiment are shown (triangle). For three experiments the average k on ϭ 1.84 ϫ 10 5 min Ϫ1 M Ϫ1 . ␣-BuIA was bath applied for 5 min at 1 m and then washed out. Recovery from block was monitored by recording the response to a 1-s pulse of ACh every 2 min. A single experiment is shown (square). Results from three oocytes yielded a k off of 0.0106 Ϯ ␣-BuIA Distinguishes among Neuronal nAChRs the ␤2 and ␤4 subunits (28,29,31,32). Only the deletion of both the ␤2 and ␤4 subunits is lethal (29), and results in death soon after birth, therefore suggesting important contributions of both subunits to autonomic function. However, since few ligands can distinguish between the two subunits, it has been difficult to determine the exact contribution of each subunit to nAChRs that are present in autonomic ganglia in genetically normal animals.
␣-BuIA represents a novel probe for discriminating, by differences in off-rate kinetic, between ␤2and ␤4-containing receptors. We note that in certain instances, ␤2 and ␤4 subunits may form part of the same receptor complex (33). The kinetics of block by ␣-BuIA for such a receptor subtype are unknown. Heteromeric nAChRs have two ligand binding sites for acetylcholine formed between the interface of ␣ and ␤ subunits. Two molecules of acetylcholine are generally believed to be required for receptor activation. If an ␣-conotoxin occupies one of the two acetylcholine binding sites, ACh is unable to activate the receptor. If ␣-conotoxin BuIA blocks by binding to the subunit interface, we feel it likely that the toxin would have slow off-rate kinetics for receptors containing a ␤4 subunit at the ligand binding interface. The binding of other ligands is known to interact with nAChR ␣ and ␤ subunits, and nAChR subunit chimeras have been used to localize the residues that contribute to subunit-dependent ligand selectivity (34). The agonist cytisine selectively activates rat ␣x␤4 versus ␣x␤2 nAChRs expressed in Xenopus oocytes (35). k-Bungarotoxin blocks rat ␣3␤2 nAChRs with slow off-rate kinetics compared with ␣3␤4 nAChRs. Subunit residue number 59 (threonine in ␤2 and lysine in ␤4) appears to be the major determinant of k-bungarotoxin binding differences between these subunits (36). Cocaine, in addition to blocking the dopamine transporter, preferentially blocks ␤4-containing versus ␤2-containing nAChRs (37). Likewise, substance P, a small peptide that acts at neurokinin type 1 receptors, also noncompetitively blocks nAChRs with a 20-to 30-fold higher affinity for ␤4versus ␤2 containing rat nAChRs (38). Species differences in nAChR subunits may have substantial effects on toxin binding. For example, dihydro-␤-erythroidine blocks rat ␣3␤2, ␣4␤2, and ␣4␤4 nAChRs expressed in Xenopus oocytes with nearly equal IC 50 values (in M, 0.41, 0.37, and 0.19, respectively) (39). In contrast, dihydro-␤-erythroidine blocks human ␣3␤2, ␣4␤2 and ␣4␤4 nAChRs expressed in Xenopus oocytes with ϳ10 -100-fold differences in IC 50 values (in M, 1.62, 0.11, and 0.01, respectively) (40). For this reason, we examined the off-rate kinetics of block by ␣-conotoxin BuIA in three species. We demonstrate in this report that the ability of ␣-conotoxin BuIA to kinetically discriminate between ␣3␤4 and ␣3␤2 nAChRs is true for rat, mouse, and human subtypes as expressed in Xenopus oocytes. For each species, the t1 ⁄2 is less than 1 or 2 min for the ␣3␤2 subtype, and greater than 30 min for the ␣3␤4 subtype. Previously analyzed ␣-conotoxins have been shown to bind to determinants on both the ␣ and ␤ subunits of the nAChR. Receptor mutation analysis indicates that ␣-conotoxin MII interacts with Lys 185 and Ile 188 of the ␣3 subunit, and Thr 59 of the ␤2 subunit (41). Analysis of ␣-conotoxin PnIA indicates that it interacts with overlapping but distinct residues on the ␣3 subunit, including Pro 182 , Ile 188 , and Gln 198 (42). It should be possible using a similar approach to identify residues on the ␤4 subunit that confer slow off-rate kinetics for ␣-conotoxin BuIA.  a Calculated by k obs ϭ k on (F) ϩ k off (see Fig. 6C).
c Calculated by k off ϭ k on ϫ k i , assuming that K i ϭ IC 50 . d Calculated from K i ϭ k off /k on , with k off and k on determined as in Fig. 6C. e See Fig. 2. f Numbers in parentheses are 95% confidence intervals.