Orientation of d-Tubocurarine in the Muscle Nicotinic Acetylcholine Receptor-binding Site*

Ligand modification and receptor site-directed mutagenesis were used to examine binding of the competitive antagonist,d-tubocurarine (dTC), to the muscle-type nicotinic acetylcholine receptor (AChR). By using various dTC analogs, we measured the interactions of specific dTC functional groups with amino acid positions in the AChR γ-subunit. Because data for mutations at residue γTyr117 were the most consistent with direct interaction with dTC, we focused on that residue. Double mutant thermodynamic cycle analysis showed apparent interactions of γTyr117 with both the 2-N and the 13′-positions of dTC. Examination of a dTC analog with a negative charge at the 13′-position failed to reveal electrostatic interaction with charged side-chain substitutions at γ117, but the effects of side-chain substitutions remained consistent with proximity of Tyr117 to the cationic 2-N of dTC. The apparent interaction of γTyr117with the 13′-position of dTC was likely mediated by allosteric changes in either dTC or the receptor. The data also show that cation-π electron stabilization of the 2-N position is not required for high affinity binding. Molecular modeling of dTC within the binding pocket of the acetylcholine-binding protein places the 2-N in proximity to the residue homologous to γTyr117. This model provides a plausible structural basis for binding of dTC within the acetylcholine-binding site of the AChR family that appears consistent with findings from photoaffinity labeling studies and with site-directed mutagenesis studies of the AChR.

The muscle nicotinic acetylcholine receptor (AChR) 1 is a member of the ligand-gated ion channel superfamily. It is a pseudo-symmetric pentamer with a subunit stoichiometry of ␣ 2 ␤␥␦. The agonist-binding sites, which are responsible for activating channel opening, lie at the interfaces between the ␣and ␥-subunits and between the ␣and ␦-subunits (1). The two sites are similar and share conserved features contributed by their respective ␣-subunits as well as the amino acids that remain constant among the homologous ␥and ␦-subunits. The ␥and ␦-subunits, however, also determine affinity differences of the two sites for various ligands. Affinity differences between the sites extend to both agonists, such as epibatidine (2) and carbamylcholine (3), and to competitive antagonists, such as d-tubocurarine (4).
Identification of binding site residues has been carried out by affinity and photoaffinity labeling, by cross-linking studies, by analysis of expressed chimeric receptors, and by site-directed mutagenesis (5). These approaches have found that the binding sites consist of residues from two subunits from several regions (also referred to as loops) that are distant in the linear subunit sequences. Loop c in the ␣-subunit includes residues ␣184 -␣198, which constitute a significant part of the binding determinant for ␣-bungarotoxin (6). It also includes an unusual, highly conserved vicinal disulfide bond between ␣Cys 192 and ␣Cys 193 (7) and several conserved tyrosine residues (␣Tyr 190 and ␣Tyr 198 ). The two other regions within the ␣-subunit include the residues ␣Tyr 93 (8,9) and ␣Trp 149 (10), respectively. The latter residue appears particularly important for interaction with agonist ammonium moieties (11).
Studies addressing the binding site contributions from the ␥and ␦-subunits have likewise found contributions from sequence-separated regions of amino acids. Amino acid ␥Trp 55 (12) lies within one region. A second region includes residues ␥Leu 109 -␥Tyr 117 . The alternating pattern of residues contributing to binding in this region led to the proposal that the sequence makes a hairpin turn in their vicinity of the binding site (13). Several other ␥-subunit residues dispersed in the sequence contribute to ligand binding, either directly or through allosteric effects. They include ␥Lys 34 , ␥Ser 161 , ␥Phe 172 , and ␥Asp 174 (14 -16).
The detailed interactions between ligands and this sizable set of amino acid residues are largely unknown. However, the prevalence of aromatic amino acids as well as detailed studies using unnatural amino acid substitution support the hypothesis that cation -electron interaction are critical for stabilizing the ammoniums of agonists (11,17). Such studies also yielded observations on the possible interactions of several other aromatic residues; Nowak et al. (17) showed that ␣Tyr 93 is likely to act as a hydrogen bond donor, whereas the aromatic ring of ␣Tyr 198 appears to interact with the quater-nary ammonium of acetylcholine.
Antagonists such as dTC have been utilized to study binding site interactions, partly because of the significant affinity difference between the two binding sites. dTC has ϳ100 -500-fold higher affinity for the ␣␥ site than the ␣␦ site (4). Nonetheless, ␣-subunit residues also affect binding of dTC. ␣Tyr 198 has a strong impact on dTC affinity; when mutated to ␣Y198F, the AChR displayed ϳ10-fold greater affinity for dTC, but the mutation had little effect on acetylcholine affinity or efficacy (18). This suggests a unique interaction between the ␣Tyr 198 hydroxyl and dTC. Because mutations at ␣Tyr 198 and ␥Tyr 117 had similar effects on metocurine (sometimes referred to as dimethyl-tubocurarine) affinity, it was proposed that these amino acids each stabilize one of the two quaternary ammoniums on metocurine through cationelectron interactions (19).
The recent atomic resolution structure of the acetylcholinebinding protein from Lymnaea stagnalis (AChBP), a protein homologous to the N-terminal, ligand-binding domain of the AChR, shows that many of the residues implicated by the studies listed above are in proximity to a binding pocket where they can contribute to stabilization of binding of agonists and antagonists (20). The structure included a solvent Hepes molecule that indicates the likely binding locus. The structure is in accord with many of the prior observations regarding ligandbinding site structure and substantially refined the current thinking about ligand-receptor interactions. However, the details of the interactions with cholinergic ligands remain to be fully elucidated, as do the conformational changes that correlate with channel opening and desensitization.
Our previous studies (21,22) have taken advantage of dTC analogs to examine the importance of various functional groups to binding interactions and conformational transitions. The structures of the analogs are shown in Fig. 1. For both mouse and Torpedo AChR, we demonstrated that the stereochemistry at the 1 carbon was important for high affinity binding but that the ammoniums at the 2-and 2Ј-positions need not be quaternary for high affinity binding (21). Furthermore, dTC ring D, which includes the 12Ј-and 13Ј-positions, interacts in a siteselective manner, consistent with a possible interaction with the ␥and ␦-subunits. 13Ј-Modification further altered the propensity of dTC to desensitize the Torpedo AChR, an observation that may serve as a clue for the structural changes that occur upon conformational shifts of the AChR. In order to pinpoint such interactions between a ligand functional group and a particular amino acid residue, ligand binding energies must be examined in concert with receptor mutagenesis.
In this study, we utilize a double mutant thermodynamic cycle analysis to analyze the interaction between our library of dTC analogs (21) and residues in the AChR-binding site, with particular emphasis on the interaction with residue ␥Tyr 117 . We demonstrate that both the 2-N and the 13Ј-positions of dTC appear to interact with this amino acid but with a proximal interaction only to the 2-N. Based on this conclusion and using the structure of the AChBP, we present a plausible structural model for the orientation and position of dTC within the binding pocket.

EXPERIMENTAL PROCEDURES
Materials-The ␣-subunit cDNA clone of the mouse nicotinic AChR was a gift from Dr. Mike White (MCP Hahnemann University); the ␤-, ␥-, and ␦-subunit cDNAs were gifts from Dr. James Patrick (Baylor College of Medicine). Restriction enzymes were purchased from Invitrogen or from New England Biolabs (Beverly, MA). The QuickChange mutagenesis kit was obtained from Stratagene (La Jolla, CA), and Endo-free Mega and Giga kits were obtained from Qiagen (Valencia, CA). HEK293 cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA). PEI was obtained either from Miles or from Aldrich. 125 I-␣-Bungarotoxin was obtained from Amersham Biosciences. Whatman GF/C filters were obtained through VWR Scientific (Houston, TX). All other chemicals were obtained from Sigma or from standard sources.
Synthesis of d-Tubocurarine Analogs-Synthesis of most of the dtubocurarine analogs was described previously (21). Synthesis of several new analogs is described as follows. For 13Ј-sulfo-d-tubocurarine (sulfo-dTC, 7Ј,12Ј-dihydroxy-6,6Ј-dimethoxy-13Ј-sulfonate-2,2Ј,2Ј-trimethyltubocuraranium trifluoroacetate), 1 g of dTC was treated with 5 ml of concentrated H 2 SO 4 on ice for 3 h. The product was precipitated by dilution with 500 ml of cold ether and collected by centrifugation. The product was purified by cation exchange chromatography over a CM-25 Sephadex column (Amersham Biosciences) and eluted with a gradient of 50 -400 mM ammonium acetate, pH 6.8. Pure fractions were collected and lyophilized with a net yield of 58%. The material was desalted on a reversed phase Beckman C-18 Ultrasphere preparative HPLC column (22 ϫ 150 mm). The column was eluted with a gradient of 10 -40% of 0.9% trifluoroacetic acid/CH 3 CN. The final product was lyophilized to a white powder. The identity was verified by mass spectroscopy (m/z ϭ 689 and 345, for the di-cation). For diacetyl-d-tubocurarine (7Ј,12Јdiacetoxy-6,6Ј-dimethoxy-2,2Ј,2Ј-trimethyltubocuraranium dinitrate), 1.07 g of dTC was treated with 2.5 ml of acetic anhydride in 8 ml of glacial acetic acid and 874 mg of sodium acetate, essentially according to Dutcher (23). The product was isolated by crystallization of the nitrate salt with a yield of 66%. This product was 95% pure as judged by reversed phase HPLC; the remaining contaminant corresponded to a monoacetylated compound. For iodo-chondocurarine (7Ј,12Ј-dihydroxy-6,6Ј-dimethoxy-13Ј-iodo-2,2,2Ј,2Ј-tetramethyltubocuraranium ditrifluoroacetate), ICl (0.5 mmol) in MeOH was slowly added to 0.2 mmol of chondocurarine (21) dissolved in 0.1 M acetate buffer (pH 5.6, 100 ml) on ice. After 30 min of reaction, 0.5 mmol of neat 2-mercaptoethanol was added to quench the excess ICl. The mixture was diluted to 250 ml, adjusted to pH 8.0 using NH 4 OH, and subjected to a CM-25 Sephadex cation exchange column chromatography. The column was eluted with a linear gradient from 50 to 300 mM NH 4 HCO 3 , pH 8.0. The fractions containing pure product were identified by reversed phase HPLC, pooled, and lyophilized. The dry salts were dissolved in 0.1% trifluoroacetic acid/H 2 O and applied to a 21.2 ϫ 150-mm Ultraprep C 18 reversed phase HPLC column (Beckman) in four batches and then eluted with a gradient of 20 -50% CH 3 CN over 60 min. The pure fractions were pooled and lyophilized to a white powder (47.4 mg, 49 mol, 25%). This product was judged pure by HPLC, UV-visible spectroscopy, 1 H NMR spectroscopy (two aromatic protons were shifted to ␦7.62 and ␦7.12 compared with ␦7.11 and ␦7.06 of chondocurarine), and mass spectrometry (m/z ϭ 749.2, m/z ϭ 375.1 for the double ion). For di-demethyl-d-tubocurarine (6,6Ј,7Ј,12Ј-tetrahydroxy-2,2Ј,2Ј-trimethyltubocuraranium ditrifluoroacetate), dTC (10 mg) was heated to 160°C in 300 l of phosphoric acid. The reaction mixture was diluted with 0.1% trifluoroacetic acid/H 2 O and applied to a preparative reversed phase HPLC column. A major and a minor product were isolated, both essentially pure, and lyophilized. The major product corresponded to 6,6Ј-didemethyl-d-tubocurarine by mass spectroscopy (m/z ϭ 581.2 and m/z ϭ 291.2 for the double ion). The minor product was singly demethylated.
Recombinant Plasmids-For the expression studies, the cDNA of each subunit of the mouse muscle AChR (␣, ␤, ␥, and ␦) was released from the original plasmid by using appropriate combinations of restriction enzymes and purified by agarose gel electrophoresis. The cDNA for each of the four subunits was subsequently cloned into the eukaryotic expression vector pCDNA3 (Invitrogen) to create the expression plasmids pCDNA3.NA-␣, pCDNA3.NA-␤, pCDNA3.NA-␦, and pCDNA3.NA-␥, respectively.
Site-directed Mutagenesis-The pCDNA3.NA-␥ plasmid was digested with the restriction enzymes EcoRV and XhoI, and the resulting two DNA fragments were purified by agarose gel electrophoresis. The small DNA fragment (NA-␥.S) contained the sites for introducing the desired mutations, and the large DNA fragment (pCDNA.NA-␥.L) contained the remainder of the ␥-subunit cDNA attached to the pCDNA vector; it was saved for later use. The NA-␥.S DNA fragment was subsequently cloned into the pBlueScript vector (Stratagene), which was pre-digested with the restriction enzymes EcoRV and XhoI. Point mutations were created by site-directed mutagenesis using appropriate pairs of mutation-specific oligonucleotide primers and the QuickChange mutagenesis kit, following the manufacturer's protocol. Subsequently, the plasmid DNA was transformed into the Escherichia coli XL-1 blue. Overnight cultures of three to four colonies from each transformation plate were grown, and plasmids were purified from these cultures using the Wizard plasmid purification kit (Promega, Madison, WI). The sequence of the entire DNA insert of each putative mutant plasmid was analyzed to confirm the desired mutation and to verify the sequence fidelity of the rest of the DNA. The NA-␥.S DNA fragment containing a given mutation was then released from the plasmid using the restriction enzymes EcoRV and XhoI, purified by preparative agarose gel electrophoresis, and ligated with the previously purified pCDNA.NA-␥.L DNA fragment (see above). After verification by DNA sequencing, a recombinant plasmid containing the mutated full-length NA-␥ cDNA was selected for large scale purification. Large scale purifications were carried out using either the Endofree Plasmid Mega (for 500 ml of culture) or the Endofree Plasmid Giga (2,500 ml of culture) kit. This purification was necessary for optimal expression of AChRs.
Cell Culture and Transfection-HEK293 cells were grown to 70 -90% confluence in 100-mm culture dishes in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum, 0.1 unit/ml penicillin, and 0.1 g/ml streptomycin. Generally, cells passaged less than 15 times in culture were used for transient expression of the wild type and the mutant AChRs. For transient expression, the pCDNA3-based AChR subunit expression plasmids were mixed in a ratio of 2:1:1:1 of ␣/␤/␥/␦ (21 g of total DNA/100-mm dish). A stock solution of PEI (25kD, branched form) was prepared at a concentration of 4.3 mg/ml (0.1 M in nitrogen) in phosphate-buffered saline (PBS) (24). For each dish, 6.3 l of PEI stock was diluted with 1.5 ml of sterile PBS, and the DNA was then added and mixed. The PBS/PEI/DNA mixture was incubated for 20 -45 min at ambient temperature before addition to 9 ml of culture medium. The old culture medium was then removed from the dish, and the PBS/PEI/DNA mixture was added. Cells were returned to the incubator and harvested 48 -72 h after transfection.
Ligand Binding-[ 3 H]Acetylcholine binding and binding of 125 I-␣-BgTx to Torpedo AChR was carried out as described previously (21). For determination of binding constants to mutant mouse AChR expressed in tissue culture cells, the following protocol was observed. Cells were scraped into PBS, centrifuged, and resuspended in high potassium/ Ringer's containing 140 mM KCl, 5.4 mM NaCl, 1.8 mM CaCl 2 , 1.7 mM MgCl 2 , 25 mM Hepes, pH 7.0, with 30 g/ml bovine serum albumin (16). Samples were incubated with the indicated concentration of competing ligand for 30 min at ambient temperature before the addition of 0.1-0.3 nM 125 I-␣-BgTx (2000 Ci/mmol). A first order binding rate was observed for more than 2 h. Binding was stopped at 2 h with excess, cold, non-radioactive BgTx, and the samples were stored on ice. Samples were filtered and washed with PBS over GF/C filters that had been soaked cold overnight with 4% Carnation Instant Nonfat Dry Milk in potassium/Ringer's buffer. Filters were counted in a gamma counter (Beckman). Data were plotted and analyzed using SigmaPlot (Jandel, SPSS, Chicago, IL). To block binding to the ␣␦ site, 30 -80 nM CTX was added to each assay for most experiments; experiments using wild type receptors confirmed that this concentration of conotoxin only blocks ligand binding to the ␣␦ site and does not change the affinity of the ␣␥ site (25).
We observed that the mouse AChR expressed in HEK293 cells displayed higher affinity for tubocurine and for iodo-dTC than the affinities observed previously for binding to the mouse AChR expressed by BC3H1 cells (22). We found that the lower affinities observed earlier had been due to the higher pH conditions of the former assays (ϳ7.5-8.0) and also to unidentified components in the former tissue culture media that specifically affected the binding of iodo-dTC. Nonetheless, the affinity changes of tubocurine and iodo-dTC described below were consistent with our earlier conclusions that the 2Ј-N likely interacted with the ␣-subunits or with conserved aspects of the heterologous subunits (21,22) and that the 12Ј, 13Ј region interacted with the heterologous subunits.
Data Analysis-For each binding curve, the K I was determined by fitting to the equation for single site binding f ϭ A(K I /(K I ϩ L)) ϩ background, where A represents maximum fmol bound, and L is the ligand concentration. Binding curves obtained in the absence of CTX were fitted to models for two-site binding with either equal or unequal site stoichiometries as described previously (22). To determine the degree of interaction between a site on the ligand and a site on the receptor, we used a double mutant thermodynamic cycle analysis (26,27). Four different binding curves were obtained in a single experiment; two ligands differing in a discrete position were measured with both wild type and mutant AChR. The interaction factor ⍀ was then calculated from the K I values using the equation ⍀ ϭ (K W,L1 /K W,L2 )/(K m,L1 / K M,L2 ), where the subscripts indicate the following: W for wild type AChR, M for mutant receptor, and L1 and L2 for the two ligands being compared. An ⍀ value significantly different from 1 indicates an interaction between the functional group on the ligand and the amino acid of the receptor.
Molecular Modeling-Energy minimizations of dTC and the AChBP (Ref. 20, Protein Data Bank code 1I9B) were conducted with the soft-ware package HyperChem (versus 5.1, Hypercube Inc., Gainesville, FL) using the MMϩ force field. The initial structure of dTC was taken from the coordinates of a crystal form (28) and was energy-minimized, which produced only minor changes in the structure. Initial simulations were sped up by using a reduced structure of the AChBP. Three of the five subunits were deleted; of the remaining two subunits, only the amino acids in immediate ligand contact plus one layer of surrounding amino acids were considered during energy minimization. For the initial manual alignment of dTC in the AChBP-binding site, dTC was placed in the binding site with the 2Ј-ammonium at the site of the ethanesulfonic acid-piperazine nitrogen of Hepes, a ligand found in the original structure of the binding site (20). dTC was energy-minimized in several rounds without allowing changes in protein structure. To resolve steric conflicts or trapping in local energy minima, occasional manual changes in dTC location were made. When a preliminary structure was obtained for bound dTC, it was then merged with two complete subunits of the acetylcholine-binding protein. Subsequent dTC energy minimization included minimization of all contact residues in the context of two complete AChBP subunits.
To mimic the structure of the mouse AChR ␣␥-binding site, the aligned, homologous mouse residues were substituted for the side chains of contact residues. The substituted side chains are shown in Fig.  8C. After each substitution, the side chain was minimized to conform to the surrounding protein environment. dTC was then energy-minimized in the context of the new structure. Two separate rounds of minimizations were conducted. First, dTC was minimized while the protein was kept rigid. Second, after the approximate dTC conformation was found, dTC and 31 nearby residues were then all minimized collectively to yield the final structure.

RESULTS
In order to examine the proximal relationship of AChR amino acid residues to functional groups on dTC, we assessed the degree of interaction by double mutant thermodynamic cycle analysis. In general, the affinities of mutant and wild type AChRs were measured for a pair of dTC analogs. The four K I values thus obtained were used to determine the value of the interaction coefficient, ⍀ (see "Experimental Procedures"). An initial goal was to identify amino acids that had specific interactions with dTC. Our previous work (21,22) had shown that modification of dTC to metocurine or to iodo-dTC shifted the ligand affinities and that the magnitude of the shifts differed at the two binding sites. Comparison of dTC with these analogs would test for interactions at the 2-N ammonium and at the 7Ј-, 12Ј-, and 13Ј-C positions. Therefore, our preliminary experiments utilized these two analogs, although not exclusively.
To determine amino acids likely to interact at these dTC sites, we focused on those ␥-subunit residues that had been identified to contribute to site-selective binding. Because dTC binds with higher affinity to the ␣␥ site than the ␣␦ site on both Torpedo AChR and mouse AChR, ␥-subunit amino acids were initially mutated to those amino acids found in the homologous, aligned positions of ␦or ⑀-subunits of either Torpedo or mouse muscle AChR. We first examined those residues identified by chimeric constructs to contribute to site-selective dTC binding: ␥Ile 116 , ␥Tyr 117 , and ␥Ser 161 (16). Conservative mutations of ␥I116M and ␥I116V displayed small, 2-4-fold affinity increases for metocurine over the wild type AChR and small affinity decreases for iodo-dTC. Mutations ␥Y117T and ␥Y117S displayed significant (10-fold) affinity changes for metocurine as well as for chondocurarine, a ligand that differs from dTC by a single 2-N methyl group (see Fig. 1), whereas smaller affinity changes were observed for dTC itself. Mutations ␥S161A, ␥S161M, and ␥S161K had more varied effects as follows: ␥S161A caused only small changes in iodo-dTC affinity; ␥S161M caused ϳ5-fold lower affinities in several dTC derivatives; and ␥S161K affected neither dTC nor chondocurarine affinity but caused a 5-fold reduction of iodo-dTC affinity. Although there were clearly observed changes in affinity, further experiments with ␥Ser 161 mutants failed to reveal interaction with any particular locus on dTC. Although the mutations we examined at residues ␥Ile 116 , ␥Tyr 117 , and ␥Ser 161 had the expected effects on metocurine affinity, according to the observations of Sine (16), only mutations at ␥Tyr 117 correlated with specific dTC interactions.
Next, we examined two residues found to be photoaffinitylabeled by dTC or by 4-benzoylbenzoylcholine (13,29), ␥Leu 109 and ␥Ser 111 . Neither conservative nor nonconservative mutations of ␥Leu 109 produced substantial changes in the affinities for dTC or of several analogs. ␥S111Y caused small changes in affinity, whereas ␥S111R resulted in consistent 3-10-fold changes on all analogs tested. We further tested a residue implicated in the site-selective binding of ␣-conotoxin, ␥Phe 172 (15). Mutation of ␥F172D produced little change in dTC affinity. ␥F172H caused only small changes in affinity for several analogs.
Of all the residues tested, mutations at ␥Tyr 117 yielded ⍀ values that most consistently correlated with structural changes in dTC. Therefore, subsequent experiments focused on this residue in more detail. This residue had been further examined by Fu and Sine (19) who proposed that stabilization of curare-type antagonist binding occurred through -electron interactions to one of the two curare ammoniums. Our preliminary experiments were consistent with ␥Tyr 117 interactions at the 2-N ammonium but did not exclude other possible interaction loci.
␥Tyr 117 Interactions with dTC-Two ligands, dTC and chondocurarine, were compared to test 2-N interaction with ␥Tyr 117 . We initially examined two mutants, ␥Y117F and ␥Y117A, that represent a conservative change and a large size change in side-chain structures, respectively. The mutation ␥Y117F resulted in higher affinity for both dTC and chondocurarine ( Fig. 2B) with similar 10-fold changes in both ligand affinities. The calculated interaction coefficient ⍀, therefore, was near 1. In contrast, the ␥Y117A mutation resulted in higher affinity for dTC than for chondocurarine, yielding an average ⍀ value of 10 ( Fig. 2A). This result suggested that there is an interaction between the 2-N and ␥Tyr 117 for large changes in side-chain structure, whereas the more conservative change from ␥Y117F yielded higher affinity for both ligands but no apparent interaction. These two mutations, ␥Y117A and ␥Y117F, yielded similar affinities for dTC, despite the disparity in side-chain structure. The difference in size suggests that simple steric effects alone do not account for the interaction. Because alanine is a non-aromatic residue, cation--electron stabilization also appears not to be required for high affinity binding.
To assess whether the 2-N position was truly the predominant interaction locus and to assess the nature of the interaction, we examined the interaction between a broader selection of ligand pairs and ␥Tyr 117 substitutions. The ␥Y117A substitution had such a strong effect on dTC affinity that we also examined ␥Y117V as representative of a large, aliphatic amino acid. ␥Y117D and ␥Y117K were included to examine charge effects on binding. We chose ligand pairs to examine specific functional groups on dTC in matched experiments. The results of this series of experiments are presented in Fig. 3, and a summary of corresponding K I values obtained for wild type and mutant receptors is found in Table I. Comparison of tubocurine and dTC reflects a change of a methyl group at the 2Ј-N (Fig. 3). This change yielded uniformly low ⍀ values, indicating no interaction between the 2Ј-N functional group and ␥Tyr 117 . Comparison of dTC with chondocurarine reflects a change of a methyl group at the opposite, 2-N ammonium. This change displayed an ⍀ value ranging from 1 to 20 that depended on the mutation that was analyzed. Examination of the 7Ј-hydroxyl, 12Ј-hydroxyl, the 7Ј,12Ј-hydroxyls (as a pair), and 6,6Ј-methoxy groups (as a pair), generally yielded low interaction values. Comparison of dTC with its 13Ј-substituted analog, iodo-dTC, revealed interactions of similar magnitude to those observed for the 2-N substituted analog. Overall, these data suggest interactions of ␥Tyr 117 with the 2-N of dTC and with the 13Ј-position of dTC with the degree of interaction highly dependent on the particular side-chain substitution.
Electrostatic Interactions at the 13Ј-Position-The Reynolds and Palmer (28) crystal structure of dTC shows the distance between the 2-N and the 13Ј-position to be 7.4 Å, a distance that is near the longest dimensions of the amino acid side chains examined. Therefore, it seemed unlikely that there would be van der Waals contact between ␥Tyr 117 and both the 2-N and the 13Ј-positions. To determine whether one functional group represented a physically closer interaction than the other, we examined the binding of another 13Ј-derivative, sulfo-dTC. This compound bears a negatively charged sulfonate at the 13Ј-position (see Fig. 1). With this ligand, we could gauge proximity through charge attraction or repulsion to ␥117 when that residue was mutated to charged side chains.
We measured affinities of three ligands, dTC, sulfo-dTC, and chondocurarine (Fig. 4, A-C, respectively) for the AChR with mutations to the ionic side chains ␥Y117D and ␥Y117K. For comparison, we also measured the affinities for their respective closest neutral congeners, ␥Y117N and ␥Y117M. Comparison of dTC affinities with sulfo-dTC affinities (Fig. 4, A and B) shows similar patterns of affinity shifts upon side-chain substitution. Sulfo-dTC had ϳ100-fold lower affinity for each side chain substitution than did dTC, indicating that the 13Ј-sulfonate moiety does decrease affinity. However, the affinity shifts were similar, regardless of the charge change on the ␥117 side chain. This result indicated that 13Ј-sulfonate was not likely in proximity to ␥117 side chains. To quantify the degree of interaction, we determined the corresponding ⍀ values for both changes, from neutral to cationic and from neutral to anionic side chains. Both ⍀ values were near 1 (Table II), values consistent with no electrostatic interaction.
For further comparison, measurement of iodo-dTC affinities for these mutated AChRs were intermediate between those of dTC and sulfo-dTC but retained a similar pattern of affinity shifts (Table II). This observation was also consistent with a lack of proximal interaction between the 13Ј-position and ␥117.  It remains unclear whether the consistently lower affinities observed for sulfo-dTC were due to the volume of the sulfonate ion, to its negative charge, or to both. When the neutral to charged amino acid changes were examined with chondocurarine, the dTC analog with an added methyl group on the 2-N, the pattern of side-chain effects on affinity was clearly distinct from that seen for dTC and sulfo-dTC (Fig. 4C). This was consistent with interaction between ␥117 and the 2-N. Because we do not have an analog that neutralizes or reverses charge at the 2-N position, it was not possible to test electrostatic interactions by ⍀ values. Nonetheless, affinity shifts upon charge changes at ␥117 were clearly consistent with cationic interactions; mutation of ␥Y117K lowered affinity for all the ligands tested, whereas mutation of ␥Y117D increased affinity (Table I  and Table II), albeit to a lesser extent than the decrease caused by ␥Y117K. These data suggest a proximal interaction between ␥Tyr 117 and the 2-N ammonium of dTC.
Interactions of Iodo-chondocurarine-An explanation for the apparent interaction of ␥Tyr 117 with both the 2-N position and the 13Ј-position, as seen in Fig. 3, is that the 13Ј-modifications shift the position or the conformation of dTC within the binding site such that interactions with the 2-N position are affected. Such a linkage in binding between the two dTC functional groups should be apparent in a thermodynamic cycle analysis of dTC analogs. To complete such an analysis, we synthesized a ligand that had been derivatized at both the 13Ј and the 2-N positions, iodochondocurarine (Fig. 1). FIG , iodo-dTC (ƒ), and iodo-chondocurarine (Ⅺ) were determined by inhibition of the initial rate of 125 I-␣-BgTx binding to Torpedo AChR-rich vesicles as described under "Experimental Procedures." Data points are averages of duplicate determinations; the solid curves represent the best fits of the data to two-site binding models with a variable site ratio. B, K I values for iodo-chondocurarine binding to mouse AChR in the absence (q) and presence (E) of 40 nM CTX were determined by inhibition of the initial rate of 125 I-␣-BgTx as described under "Experimental Procedures." Data points are averages of triplicate determinations; the solid curves are the best fits to two-site and to one-site binding models, respectively. C, K I values for binding of dTC (E, q) and iodo-chondocurarine (Ⅺ, f) to ␥Y117S (q, f) and ␥Y117T (E, Ⅺ) were determined in the presence of 40 nM CTX. The solid curves represent the best fits to a single site binding model.

TABLE II
Binding of 13Ј-modified d-tubocurarine analogs to charged and neutral ␥Tyr 117 mutants Sets of paired experiments to directly compare K I values for inhibition of 125 I-␣-Bgtx binding to wild type (wt) and mutant mouse receptors were determined as described under "Experimental Procedures." The errors listed are standard deviations from n independent determinations, given in the parentheses. ⍀ was computed as described under "Experimental Procedures" using the K I values from dTC and sulfo-dTC measured on pairs of charged and neutral side chain mutations: ␥Y117D versus ␥Y117N and ␥Y117K versus ␥Y117M. Initial characterization of this compound by Torpedo AChR binding revealed substantially higher site selectivity between the ␣␥ and ␣␦ site affinities, when compared with dTC, chondocurarine, and iodo-dTC (Fig. 5A). Affinities for the mouse AChR sites were similarly affected with slightly enhanced affinity for the ␣␥ site and 3-4-fold decreased affinity for the ␣␦ site ( Fig. 5B and Table III). Binding in the absence and presence of 40 nM ␣-conotoxin MI to block binding to the ␣␦ site confirmed that iodo-chondocurarine bound with higher affinity to the ␣␥ site of the AChR. The iodo-chondocurarine affinity for the ␣␥ site was not significantly different (Table III) from the affinities of dTC, chondocurarine, and iodo-dTC (Table I), indicating no interaction between the 13Ј and 2-N functional groups in the context of the native ␥Tyr 117 residue.
The affinity changes at the ␣␦ site, however, were significant. Because the natural amino acid in the ␦-subunits of both mouse and Torpedo AChR in the position homologous to ␥Tyr 117 is Thr and because the ␥Y117T mutation caused large changes in affinity of chondocurarine (Table I), we examined the interactions of ␥Y117T mutation with iodo-chondocurarine. For comparison, we also examined the closest congener, ␥Y117S, the difference in side-chain structure being the loss of a methyl group. Binding measurements of dTC and iodo-chondocurarine show that dTC affinity was affected strongly by this change in side-chain structure, whereas the affinities for iodochondocurarine were insensitive to the change in side-chain structure but were substantially lower (Fig. 5C).
We constructed a set of linked thermodynamic cycles (Fig. 6) using the K I values in Tables I and III. The outer edges of the diagram (in black) constitute a cycle that represents stepwise ligand structural changes for the ␥Y117T mutant (⍀ ϭ 3.7); the inner cycle (in gray) has the corresponding changes for the ␥Y117S mutant (⍀ ϭ 1.4). The ⍀ values indicate that the interaction between 13Ј-and 2-N positions is stronger in the context of the mutation ␥Y117T than ␥Y117S. The corresponding ⍀ value for the native ␥Tyr 117 is 1.4, which is similar to that for ␥Y117S. The four trapezoidally shaped cycles, which connect the inner and outer cycles, represent the interactions between the amino acids and stepwise changes in ligand functionality. Addition to dTC of either the 2-N-methyl alone or the 13Ј-iodine alone yielded low ⍀ values of 1-2. However, further addition of either second substituents produced higher interaction values (4.3 and 10); thus, interaction was only observed upon the second ligand modification. Examination of the affinities for the ␥Y117T mutant shows that either single dTC modification resulted in decreased ligand affinity, but the second modification did not further affect affinity. In contrast, for ␥Y117S, each modification decreased affinity independently.
Whether interactions occur appears to be highly sensitive to the amino acid residue context at position ␥117.
␥Tyr 117 and the 2-N Interactions-To try to assess the nature of the interaction between ␥Tyr 117 and the 2-N of dTC, we grouped dTC and chondocurarine affinity data by amino acid side-chain functionality (Fig. 7). In this figure, the two K I values are connected with a line for each amino acid. This allows a simple visualization of thermodynamic cycle interactions, parallel lines indicate no interaction and diverging lines indicate interaction. Most side-chain substitutions yield parallel lines, and most differ from the flat, wild type ␥Tyr 117 side chain. This explains the high number of amino acids indicative of interactions in Fig. 3 (column 2N) for these two ligands where ⍀ values were calculated against the wild type ␥Tyr 117 . Of all the amino acids, only ␥Y117L and ␥Y117M display negative slopes, and the charged amino acids have the steepest slopes. Examination of the plot for trends reveals that neither change in polarity nor charge is an indicator of interaction. A plot of side-chain volume versus K I also showed no clear trend (not shown). Nor did simple inspection of affinity changes reveal clear patterns; high affinity was achieved with three structurally dissimilar amino acids, ␥Y117F, ␥Y117D, and ␥Y117A.
In one case, correlation of side-chain structure with interaction was suggestive but nonetheless inconclusive; comparison of ␥Y117I and ␥Y117L was consistent with an interaction due to a side-chain change involving the displacement of a methyl group from ␤-carbon to the ␥-carbon. This observation might indicate involvement of the ␤-carbon branched methyl group of ␥Y117I (and perhaps of ␥Y117V). However, other side-  chain substitutions, without ␤-carbon branched groups, have slopes similar to those that do (e.g. ␥Y117A, ␥Y117N, and ␥Y117D), indicating no formal interaction.
Modeling of dTC into the Structure of the AChBP-We computer-modeled dTC into the binding pocket of the AChBP to determine whether our observations on the interaction of dTC were consistent with the structure of this binding site. To mimic the conditions of the mutation experiments, we further changed the side chains of the residues in contact with dTC to those of the mouse AChR. We used the aligned ␣-subunit residues for those corresponding to the A-chain of the AChBP and the ␥-subunit residues for those corresponding to the B-chain. Energy minimization yielded a satisfactory structure of dTC in the binding pocket without bad contacts or unrealistic bond lengths and angles. The binding site residues moved only minimally in order to accommodate dTC (Fig. 8). Movement of A-Trp 143 (homologous to AChR ␣Trp 149 ) was critical for accommodating the 6Ј-hydroxymethyl moiety of dTC (compare Fig. 8,  C and D). The displacement was primarily rotation about the tryptophan ␣-␤ carbon bond (48°) and the ␤-␥ carbon bond (29°). Without this movement, rings A and B of dTC became severely distorted during minimization. There was also movement of A-Tyr 89 (␣Tyr 93 ) about its ␤-␥ carbon axis to accommodate the movement of A-Trp 143 , lesser movement of A-Tyr 185 (␣Tyr 190 ) to accommodate the bulk of the 2Ј-quaternary ammonium, and of A-Tyr 192 (␣Tyr 198 ) due to steric interaction between its 4-hydroxyl and dTC ring D.
This model is appealing because the dTC 2-N is clearly apposed to ␥Tyr 117 , as suggested by our data, and the opposite, 2Ј-ammonium, is deep in the conserved binding pocket, which is also consistent with our data. dTC maintains close protein contact in the binding pocket without gross disturbance of the backbone configuration. Although it is possible that dTC can have distinct binding modes, the simplest alternatives appear unlikely. Rotation of dTC about its long axis places the 2Ј-N quaternary ammonium nearer ␣Trp 149 and ␥Trp 55 ; however, this part of the binding site is sterically too crowded to accommodate the 2Ј-ammonium without significant distortion of the A and B rings of dTC or substantial movement of the loop c region. Rotation of dTC about the axis normal to the structure in Fig. 8A yielded a plausible, energy-minimized structure with the 2Ј-N near ␥Tyr 117 and the 2-N deep in the conserved pocket. Although this orientation may constitute an alternative binding mode, it is inconsistent with the observed interactions between ␥Tyr 117 and the 2-N.
The conformation of bound dTC is distinct from the conformations observed by atomic resolution structures of curare compounds (28, 30 -32) and from any of the low energy conformations of dTC found by Zhorov and Brovtsyna (33) using molecular modeling. The region of the 2-N and 1-C of dTC is relatively strained, and if dTC in its bound conformation is removed from the context of the protein and permitted to relax by energy minimization, this region changes structure. The bound conformation is, therefore, not a local minimum energy state of dTC but is stabilized by close interactions with the surrounding protein. The close contact with the binding site in the current model suggests that binding of distinct dTC conformers would require substantial movement of the protein to bind. Energy minimizations with dTC placed in the binding pocket in the same general orientation but with various other conformations often displaced ␣-subunit loop c (residues ϳ188 -199; data not shown) with concomitant disruption of contacts between loop c and the ␥-subunit. If such dislocations of the protein occur naturally, this might permit alternative modes of dTC binding beyond those considered here. DISCUSSION Our goal was to determine the nature of interaction of dTC with the AChR. In addition, we tested whether a double-mutant thermodynamic cycle analysis could be applied to analysis of ligand-protein binding in a system where the ligand was substantially smaller than the peptide toxins typically employed (e.g. Refs. 34 and 35). During the course of the studies, predominantly weak interaction energies were observed; therefore, determination of proximal interactions required examination of the patterns from many mutations. ␥Tyr 117 apparently interacts with dTC at both the 13Ј and the 2-N positions; however, the results from charge changes were inconsistent with proximity to the 13Ј-position. Therefore, the 13Ј-interactions are likely allosteric: changes in structure of either dTC or FIG. 7. The affinities of dTC and chondocurarine are affected by ␥117 side-chain structure. K I values for inhibition of 125 I-␣-Bgtx binding to wild type and mutant mouse receptors were determined as described under "Experimental Procedures" and are plotted for all amino acids tested. For numerical data, see Table I. For each mutant, the two K I values for chondocurarine and for dTC are connected by a line. the protein. The data consistently supported ␥Tyr 117 interaction with the dTC 2-N position; we used this fact to construct a model for binding of dTC to the AChR. As discussed below, this model is largely consistent with the body of data describing the interaction of dTC or its analogs with various AChRs. The model is testable and may provide insight into the nature of the conformational changes that occur in the binding site upon ligand binding and subsequent channel opening and desensitization.
Structure-Activity of d-Tubocurarine-Of particular interest to this study was mapping the AChR sites of interaction of the 12Ј-and 13Ј-dTC positions because we had shown these sites to affect site-selective binding and desensitization (21,22). Whereas we observed interactions at the 13Ј-position to ␥Tyr 117 , these were inconsistent with proximal interactions. This observation suggested that 13Ј-modifications invoke conformational changes of dTC, the protein, or both that produced the interaction energy or that dTC has an alternative mode of binding where the interactions are distinct, such as an orientation rotated to swap the 2-N and 2Ј-N positions. Protein allosteric effects might be consistent with the increased desensitization caused by 13Ј-modified dTC analogs and suggest a possible role for ␥Tyr 117 in desensitization of the AChR.
These observations also suggested an allosteric interaction between the 13Ј and the 2-N position of dTC, mediated either by changes in protein structure or dTC structure. We tested for that interaction by comparing binding of iodo-chondocurarine to dTC, iodo-dTC, and chondocurarine. Interaction between the functional groups was strongly dependent on the ␥117 sidechain structure; for the native ␥Tyr 117 and for ␥Y117S, there were no apparent interactions, whereas interaction was observed with ␥Y117T. The mixed results may indicate compensatory structural changes in the ligand or protein upon binding. They also show that some side-chain structures can greatly reduce affinity of 13Ј-modified ligands. Clearly, curare binding is sensitive to such changes in structure, but full understanding of either direct or indirect interactions will require further experimentation.
The moderate interaction energies observed and the apparent allosteric interactions limited the scope of our conclusions. The largest affinity changes and interaction energies were usually observed upon charge changes, as seen in other examples utilizing thermodynamic cycle analysis (36). The modest changes in interaction energies seen otherwise may reflect the intrinsic flexibility of the ligand or the protein to compensate for structural changes. In this respect, dTC is known from crystal structures (28,30,31) and molecular modeling (33) to have several distinct, stable conformations. These concerns may limit the general applicability of this method to understanding binding of flexible ligands to allosteric proteins.
Molecular Modeling of dTC-We constructed a model for the binding of dTC to the AChR using the crystal structure of the AChBP (20). This model is consistent with much of the photoaffinity labeling data that indicated proximity between dTC and Torpedo residues ␣Tyr 190 , ␣Tyr 198 , ␣Cys 192 (37), ␦Trp 57 (the primary site of labeling in the heterologous subunits and homologous to ␥Trp 55 ), as well as with minor labeling of ␥Tyr 117 (13). dTC photoaffinity labeling of ␥Tyr 111 was minor, relative to other residues (13), and this residue does not contact protein shown. C, dTC within structure of the putative binding site showing the 31 residues that were included in the minimization. Labels for several important residues are according to the mouse AChR sequence. D, the corresponding starting structure of the AChBP prior to amino acid substitution and minimization, from the protein data bank. Amino acid side chains corresponding to those in C are labeled according to the AChBP sequence.
FIG. 8. Conformational changes of dTC upon binding the AChBP. The atomic resolution structures of dTC and the AChBP served as the starting points for generating a model for the binding of dTC in the putative acetylcholine-binding site as described in "Experimental Procedures." The side chains surrounding the binding pocket were changed to the aligned residues in the mouse ␣and ␥-subunits. Stereoscopic images of the following structures are shown. A, the starting structure of dTC based on the dTC crystal structure (28). B, the structure of dTC after energy minimization in the AChBP, with no dTC in the model shown in Fig. 8. Mutagenesis experiments have shown that ␣Y198F enhances dTC affinity about 10-fold (18). This observation is consistent with the close contact of the 4-OH group of ␣Tyr 198 with ring D of dTC ( Figs. 1 and 8); removal of this OH group would clearly relieve steric strain. The type of interaction seen here is consistent with the aromatic-aromatic interaction suggested previously (17,18). Mutagenesis of ␣Tyr 93 , ␥Trp 55 , or ␦Trp 57 has little effect on dTC affinity, whereas ␣Y190F reduces dTC affinity 10-fold (8). The rationale for these observations was not clear from inspection of the model.
Of the three residues identified by analysis of chimeras (16) as being important for site-selective dTC binding, ␥Ile 116 , ␥Tyr 117 , and ␥Ser 161 , only ␥Tyr 117 appears to be in direct contact with dTC in our model. ␥Ile 116 points away from the binding site into the interior wall of the protein. ␥Ser 161 faces into the binding pocket but is not in van der Waals contact with dTC. It seems feasible, however, that mutation to a cationic or large residue at this position could affect dTC binding, as observed, by virtue of its proximity.
It was also important to assess whether the model satisfied our observations on the affinities of the dTC analogs. The model agreed with our principal observation, the interaction between ␥Tyr 117 and the 2-N of dTC. We observed that 7Јmodification increased affinity of dTC (Table I), consistent with proximity of this functional group to a cavity in the binding site model that may accommodate the added bulk of a substituent. The 12Ј-position appears in close contact with the protein, suggesting that added steric bulk should interfere with binding, consistent with the general decrease in affinity we observed for modification at this position. In contrast, we could not rationalize the effects of 13Ј-modification, a position that also appears in close contact with the protein, particularly with ␥Leu 109 . Our preliminary experiments with mutations at ␥Leu 109 did not show interactions. Although the environment of the 13Ј-position is sterically crowded, it is not obvious why the affinities of the 13Ј-modified analogs tend to be sensitive to the ␥117 side-chain structure nor why they induce increased desensitization of Torpedo AChR. We observe little change in affinity upon demethylation at the 6-and 6Ј-oxygens. The model predicts the 6Ј-hydroxymethyl of dTC to be in close contact with the protein, where loss of a methyl group might be expected to relieve steric strain and increase affinity. Although our model is consistent with many observations, it cannot rationalize all observations and therefore represents only a first step toward understanding dTC-AChR interactions.
We also sought to understand the nature of the interaction between ␥Tyr 117 and the dTC 2-N. The interaction between ␥Tyr 117 and the dTC analog metocurine had been proposed to occur through cation-electron interactions (19). The proximity of the 2-N of dTC and ␥Tyr 117 in our model appears reasonably consistent with such a mechanism. However, we observe increased affinity of the mutation ␥Y117F, which should yield cation-interactions of similar strength (38). There is no obvious interaction of the tyrosine hydroxyl that explains this affinity change. Other factors clearly may influence the affinity as well; small side chains such as ␥Y117A or ␥Y117D yielded affinities as high as that of ␥Y117F. It is possible that small side chains reduce steric hindrance and permit a lower energy conformation of dTC, thereby improving affinity.
A better understanding of dTC binding will appear with a direct structural determination and knowledge of the structural differences between the resting and desensitized conformations of the AChR. Our model clearly does not permit simple entry of dTC into the binding site without breathing or flexing in the binding pocket. Kinetic measurements of dTC associa-tion and dissociation rates may provide a test of our model and determine which protein conformation is better represented by the crystal structure of the AChBP. During energy minimization of various dTC conformations, we sometimes observed displacement of the ␣-subunit loop c. Such movement may be representative of the conformational changes that take place upon desensitization (or opening) of a native channel. Although proposing such a conformational movement is speculative, it may also explain some of the inconsistencies in our model. Structural changes in dTC may be accommodated by complementary structural changes in the binding site, affecting the overall conformational equilibrium, and yielding modest net changes in affinity.