Portal de Periódicos da CAPES

Curariform Antagonists Bind in Different Orientations to Acetylcholine-binding Protein

2003; Elsevier BV; Volume: 278; Issue: 25 Linguagem: Inglês

10.1074/jbc.m301151200

1083-351X

Fan Gao, Nina Bern, Alicia J. Little, Hailong Wang, Scott B. Hansen, Todd T. Talley, Palmer Taylor, Steven M. Sine,

Receptor Mechanisms and Signaling

Acetylcholine-binding protein (AChBP) recently emerged as a prototype for relating structure to function of the ligand binding domain of nicotinic acetylcholine receptors (AChRs). To understand interactions of competitive antagonists at the atomic structural level, we studied binding of the curare derivatives d-tubocurarine (d-TC) and metocurine to AChBP using computational methods, mutagenesis, and ligand binding measurements. To account for protein flexibility, we used a 2-ns molecular dynamics simulation of AChBP to generate multiple snapshots of the equilibrated dynamic structure to which optimal docking orientations were determined. Our results predict a predominant docking orientation for both d-TC and metocurine, but unexpectedly, the bound orientations differ fundamentally for each ligand. At one subunit interface of AChBP, the side chain of Tyr-89 closely approaches a positively charged nitrogen in d-TC but is farther away from the equivalent nitrogen in metocurine, whereas, at the opposing interface, side chains of Trp-53 and Gln-55 closely approach the metocurine scaffold but not that of d-TC. The different orientations correspond to ∼170° rotation and ∼30° degree tilt of the curare scaffold within the binding pocket. Mutagenesis of binding site residues in AChBP, combined with measurements of ligand binding, confirms the different docking orientations. Thus structurally similar ligands can adopt distinct orientations at receptor binding sites, posing challenges for interpreting structure-activity relationships for many drugs. Acetylcholine-binding protein (AChBP) recently emerged as a prototype for relating structure to function of the ligand binding domain of nicotinic acetylcholine receptors (AChRs). To understand interactions of competitive antagonists at the atomic structural level, we studied binding of the curare derivatives d-tubocurarine (d-TC) and metocurine to AChBP using computational methods, mutagenesis, and ligand binding measurements. To account for protein flexibility, we used a 2-ns molecular dynamics simulation of AChBP to generate multiple snapshots of the equilibrated dynamic structure to which optimal docking orientations were determined. Our results predict a predominant docking orientation for both d-TC and metocurine, but unexpectedly, the bound orientations differ fundamentally for each ligand. At one subunit interface of AChBP, the side chain of Tyr-89 closely approaches a positively charged nitrogen in d-TC but is farther away from the equivalent nitrogen in metocurine, whereas, at the opposing interface, side chains of Trp-53 and Gln-55 closely approach the metocurine scaffold but not that of d-TC. The different orientations correspond to ∼170° rotation and ∼30° degree tilt of the curare scaffold within the binding pocket. Mutagenesis of binding site residues in AChBP, combined with measurements of ligand binding, confirms the different docking orientations. Thus structurally similar ligands can adopt distinct orientations at receptor binding sites, posing challenges for interpreting structure-activity relationships for many drugs. The superfamily of pentameric ligand-gated ion channels activated by acetylcholine (ACh), 1The abbreviations used are: ACh, acetylcholine; AChBP, acetylcholine-binding protein; d-TC, d-tubocurarine; MD, molecular dynamics simulation; r.m.s.d., root mean square deviation. γ-aminobutyric acid, glycine, and serotonin mediate rapid synaptic transmission throughout the nervous system. Their strategic position in the pathway of information flow makes them strategic loci for disease processes as well as logical targets for drugs used clinically. The synaptic protrusion of these channels contains a ligand binding domain, which harbors structural components specialized for binding agonists and competitive antagonists. The ligand binding domain is formed at interfaces between subunits where conserved aromatic and hydrophobic residues are clustered (1Sine S.M. J. Neurobiol. 2002; 3: 431-446Crossref Scopus (162) Google Scholar, 2Corringer P.J. Le Novere N. Changeux J.P. Annu. Rev. Pharmacol. Toxicol. 2000; 40: 431-458Crossref PubMed Scopus (706) Google Scholar, 3Karlin A. Akabas M. Neuron. 1995; 15: 1231-1244Abstract Full Text PDF PubMed Scopus (564) Google Scholar). In the nicotinic receptor found at the motor endplate, the alpha subunit forms one face of the ligand binding domain, whereas a non-alpha subunit, gamma, delta or epsilon, forms the other face. Studies of subunit chimeras and site-directed mutations have pinpointed key residues critical for stabilizing bound agonists and antagonists. In particular, residues on both alpha and non-alpha subunits are critical for binding competitive antagonists of the curare family, suggesting that these antagonists prevent ACh binding by bridging the subunit interface (4Fu D-X. Sine S.M. J. Biol. Chem. 1994; 269: 26152-26157Abstract Full Text PDF PubMed Google Scholar). The potential for understanding ligand binding at the atomic structural level recently arose with the discovery of acetylcholine-binding protein (AChBP) (5Brejc K. van Dijk W.J. Klaassen R.V. Schuurmans M. van Der Oost J. Smit A.B. Sixma T.K. Nature. 2001; 411: 269-276Crossref PubMed Scopus (1580) Google Scholar), a soluble protein homologous to the ligand binding domains of pentameric ligand-gated ion channels. Moreover, the crystal structure of AChBP elegantly confirmed the subunit interface hypothesis of the ligand binding site and provided atomic coordinates of residues known to be essential for binding agonists and antagonists (6Smit A.B. Syed N.I. Schaap D. van Minnen J. Klumperman J. Kits K.S. Lodder H. van der Schors R.C. van Elk R. Sorgedrager B. Brejc K. Sixma T.K. Geraerts W.P. Nature. 2001; 411: 261-268Crossref PubMed Scopus (465) Google Scholar). AChBP therefore emerges as a prototype for gaining in-depth understanding of how small ligands bind to an acetylcholine binding site. A competitive antagonist from the curare family, d-tubocurarine (d-TC), was found to bind with high affinity to AChBP (5Brejc K. van Dijk W.J. Klaassen R.V. Schuurmans M. van Der Oost J. Smit A.B. Sixma T.K. Nature. 2001; 411: 269-276Crossref PubMed Scopus (1580) Google Scholar). We therefore sought to understand binding of d-TC and its methylated analog metocurine at the atomic structural level by applying computational methods in combination with site-directed mutagenesis and measurements of antagonist binding. Unexpectedly, our computational results predict that these two structurally similar ligands bind to AChBP in distinctly different orientations where each ligand contacts different side chains at the subunit interface. These predictions are confirmed by mutagenesis of key interface residues in AChBP followed by measurements of ligand binding. The results have profound implications for understanding ligand binding to pentameric ligand-gated ion channels. Materials—125I-α-Bungarotoxin was purchased from PerkinElmer Life Sciences, and d-tubocurarine (d-TC) chloride was purchased from ICN Pharmaceuticals, Inc. The fully methylated analog of d-tubocurarine, metocurine iodide, was a gift from the Eli Lilly Co. Human embryonic kidney cells (293 HEK) were from the American Type Culture Collection. Molecular Dynamics and Ligand Docking Simulation—We employed a docking strategy that allows flexibility in both the ligand and the protein to predict structures of the complexes formed between AChBP and either d-TC or metocurine. We first carried out molecular dynamics (MD) simulation for AChBP in the presence of explicit solvent molecules at room temperature using the AMBER 7 program (9Pearlman D.A. Case D.A. Caldwell J.W. Ross W.S. Cheatham T.E. Debolt S. Ferguson D. Seibel G. Kollman P. Comp. Phys. Commun. 1995; 91: 1-41Crossref Scopus (2670) Google Scholar) and collected multiple frames of the equilibrated dynamic AChBP structure. Then we ran ligand docking simulations to each frame of the AChBP structure using the AUTODOCK 3.0.3 program (10Morris G.M. Goodsell D.S. Halliday R.S. Huey R. Hart W.E. Belew R.K. Olson A.J. J. Comp. Chem. 1998; 19: 1639-1662Crossref Scopus (9209) Google Scholar), with rotation allowed around bonds to hydroxyl and methoxy groups of each curare analog. We then employed a cluster analysis to identify predominant modes of ligand docking and found a predominant docking orientation for each ligand. A further MD simulation was performed to refine the predominant receptor-ligand complex, and the time average structure was output as the final docked structure. Atomic coordinates of AChBP (5Brejc K. van Dijk W.J. Klaassen R.V. Schuurmans M. van Der Oost J. Smit A.B. Sixma T.K. Nature. 2001; 411: 269-276Crossref PubMed Scopus (1580) Google Scholar) were downloaded from the Protein Data Bank (PDB code: 1I9B). To prepare the protein structure for docking, water, ions, and HEPES were first removed from the crystal structure. AChBP was then solvated in a water box, using the TIP3P model for water molecules extended at least 10 Å in each direction from the solute (24,274 water molecules included), and 60 Na+ and 15 Cl– counter ions were added to neutralize the system by employing the LEAP module of the AMBER 7 program (parm99 force field included). The system was first energy-minimized then heated for 30 ps to 298 K, and the Particle Mesh Ewald method was employed to calculate long range electrostatic interactions. Following the heating step, the system was maintained at 298 K, and the MD simulation was computed at 1.0-fs intervals with frames collected every 1.0 ps. After 500 ps of simulation, the system reached equilibrium and trajectories from the subsequent 1500-ps MD simulation were collected for docking computation. During the simulation, the SHAKE algorithm was turned on to constrain bonds involving hydrogen atoms, and the non-bonded interaction cutoff was set to 8.0 Å. From each of the 1500 frames of the dynamic AChBP structure, atomic coordinates of one pair of the binding site-forming subunits (subunits A and B from the crystal structure) were collected for docking computation. Partial charges were then assigned to each atom of AChBP using the AMBER force field, which employed the restrained electrostatic potential charge model. Crystal structures of metocurine (11Sobell H.M. Sakore T.D. Tavale S.S. Canepa F.G. Pauling P. Petcher T.J. Proc. Natl. Acad. Sci. U. S. A. 1972; 69: 2212-2215Crossref PubMed Scopus (33) Google Scholar) and d-TC (12Codding P.W. James M.N.G. Acta Crystallogr. 1973; 29: 935-942Crossref Google Scholar) were used to prepare the ligands for docking (Fig. 1). Hydrogen atoms were added by employing QUANTA, followed by a 500-step energy minimization of all the hydrogen atoms using the CHARMM module in QUANTA. Partial charges were assigned to each atom of the ligand using the semi-empirical charge model-1 (CM1) in the AMSOL 6.7.2 program (13Winget P. Thompson J.D. Cramer C.J. Truhlar D.G. J. Phys. Chem. 2002; 106: 5160-5168Crossref Scopus (31) Google Scholar) with the following settings: the SM5.4PDA solvation mode, the AM1 semi-empirical method, water as solvent, and the “TRUSTE” optimizer and “CHARGE” set to +2 based on the net charge of both ligands. Docking computations were performed using the AUTODOCK 3.0.3 program. Non-polar hydrogens were removed from each ligand, and their partial atomic charges were united with the bonded carbon atoms. For each frame of the AChBP structure, the ligand was arbitrarily positioned at the subunit interface using QUANTA, then docking computations were performed 10 times using the Lamarkian genetic algorithm, with grid sizes 40 × 40 × 40 (grid spacing, 0.375 Å), yielding 10 docked conformations per frame. Parameters for the docking computation were assigned default values. The resulting 15,000 docked orientations for each ligand produced by AUTODOCK 3.0.3 were analyzed by grouping them into clusters with similar ligand orientations. We employed a cluster analysis that classified structures as similar by specifying translational and rotational limits of the docked ligands. We used 1 Å as the translation cutoff for the distance between mass centers of the different docking orientations and 10° as the rotation cutoff for the corresponding differences in rotation axis and rotation angles. For both d-TC and metocurine, one docking orientation predominated and was selected as the most probable structure of the AChBP-ligand complex. The resulting AChBP-ligand complexes were further refined by energy minimization with explicit water molecules and counter ions using the SANDER module of AMBER, followed by a 30-ps heating step and then a 500-ps equilibration MD simulation using the parm99 force field and Particle Mesh Ewald methods described above. CM1 partial atomic charges were assigned to each atom of the ligands to mimic their electrostatic properties in the MD simulation. Time-average structures of the AChBP-ligand complexes from the 500-ps equilibration were output as the final ligand-bound structures. 2Note: PDB files of the protein-ligand complexes are available upon request from the corresponding author. Plasmids and Mutagenesis—The previously described synthetic cDNA encoding AChBP (7Hansen S.B. Radic Z. Talley T.T. Molles B.E. Deernick T. Tsigelny I. Taylor P. J. Biol. Chem. 2002; 277: 41299-41302Abstract Full Text Full Text PDF PubMed Scopus (92) Google Scholar) was subcloned into the cytomegalovirusbased vector pRBG4 (8Lee B.S. Gunn R.B. Kopito R.R. J. Biol. Chem. 1991; 266: 11448-11456Abstract Full Text PDF PubMed Google Scholar), and an MF2 epitope tag was constructed in-frame with the C-terminal coding sequence. Starting with the human alpha7 cDNA subcloned into pRBG4 (28Quiram P.A. Sine S.M. J. Biol. Chem. 1998; 273: 11001-11006Abstract Full Text Full Text PDF PubMed Scopus (63) Google Scholar), we isolated a 3860-bp HindIII to SalI fragment and ligated that with a 575-bp HindIII to BspEI fragment, derived from the AChBP cDNA, and a synthetic double-stranded oligonucleotide that included the MF2 epitope with a stop codon at the end. The 5′-end of the resulting construct contained five unwanted codons from alpha7; these were removed by bridging BsgI and HindIII sites with a synthetic double-stranded oligonucleotide. The final coding region therefore begins with the human alpha7 signal peptide (MRCSPGGVWLGLAASLLHVSLQ), which is followed by the AChBP coding sequence and terminates with the MF2 epitope sequence (DYKDDDDK). Point mutations were constructed using the QuikChange kit from Stratagene. The final AChBP construct and all mutations were confirmed by automated dye terminator sequencing. Expression and Purification of Mutant Receptors and Ligand Binding Measurements—293 HEK cells were transfected with mutant or wild type subunit cDNAs using calcium phosphate precipitation (14Sine S.M. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 9436-9440Crossref PubMed Scopus (167) Google Scholar). Two to three days following transfection, wild type or mutant AChBP secreted into the culture medium was purified in one step using anti-MF2 coupled to an agarose gel (Sigma-Aldrich, St. Louis, MO). For a typical determination of competitive antagonist binding, 100 μl of culture medium was combined with 100 μl of anti-MF2-agarose suspended in Tris buffer (50 mm Tris, 150 mm NaCl, pH 7.5) containing protease inhibitors (5 mm EDTA, 0.2 mm phenylmethylsulfonyl fluoride, 0.002 mm aprotinin, 0.01 mm pepstatin A), and 200 μl of Tris buffer containing 0.1% bovine serum albumin was added. After mixing for 1 h at 4 °C, the samples were centrifuged, the supernatant removed, and the washing process was repeated two more times. The immobilized protein was suspended in Tris buffer containing protease inhibitors and stored on ice until use. Purity and size of the protein were determined by elution from the agarose gel, electrophoresis on an SDS-PAGE gel, and transfer to 0.2-μm nitrocellulose paper. Protein detection was accomplished with a chemiluminescence kit (Roche Applied Science, Manheim, Germany) using biotin-conjugated anti-MF2 to label the protein, and streptavidin-conjugated horseradish peroxidase, hydrogen peroxide, and luminol for detection. Ligand binding was measured by competition against the initial rate of 125I-α-bungarotoxin binding by modification of the previously described method (15Sine S.M. Taylor P. J. Biol. Chem. 1979; 254: 3315-3325Abstract Full Text PDF PubMed Google Scholar). Briefly, AChBP immobilized on agarose was suspended in Tris buffer at a concentration of ∼1 nm in α-bungarotoxin binding sites and incubated with competitive antagonist for 15 min. 125I-α-Bungarotoxin at a final concentration of 5 nm was added and allowed to bind for 4 min to allow occupancy of approximately half of the total binding sites. The binding reaction was terminated by dilution in Tris buffer, the samples were centrifuged, and the supernatant was removed. The washing process was repeated a second time, and radio-activity bound to the pellet was quantified using a γ-counter. The initial rate of toxin binding was calculated to yield the fractional occupancy, Y, and the competition measurements were analyzed according to the equation, 1−Y=1/(1+([ligand]/Kapp)n)(Eq. 1) where n is the Hill coefficient and Kapp is the apparent dissociation constant. To understand binding of competitive antagonists to AChBP at the atomic structural level, we studied binding of d-TC and metocurarine (Fig. 1) using computational methods combined with mutagenesis of key binding site residues and experimental determinations of ligand binding. Because proteins are intrinsically flexible, we employed a computational strategy that accounts for this key property. We therefore used molecular dynamics (MD) simulation to generate an ensemble of protein structures accessible to the ligand and computed docking of ligand to each snapshot of the dynamic structure (16Carlson H.A. McCammon J.A. Mol. Pharmacol. 2000; 57: 213-218Crossref PubMed Scopus (61) Google Scholar). In MD simulation, the structure reaches an energy minimum when a plateau is reached in the root mean-squared deviation (r.m.s.d.) of atomic distances as a function of time. For AChBP in the presence of explicit water molecules and counter ions, such a plateau is achieved after ∼500 ps of MD simulation, as measured by the r.m.s.d. of either the alpha carbons or all heavy atoms (Fig. 2). The subsequent 1.5-ns period therefore represents the AChBP structure in dynamic equilibrium; the corresponding 1500 frames were collected for computation of ligand docking. Application of AUTODOCK 3.0.3 to these frames generated 10 docked structures per frame to give a total of 15,000 docked structures. We then grouped these into structurally similar clusters by specifying translation and rotation limits of the center of mass of the ligand within the binding pocket (see “Experimental Procedures”). For both ligands, our cluster analysis reveals a predominant ligand orientation within the binding pocket (cluster 1), as well as several smaller clusters with different orientations (clusters 2–6; Fig. 3). For metocurine, comparison of clusters 1 and 2 reveals only a minor shift within the binding pocket, whereas for d-TC, clusters 1 and 2 represent fundamentally different orientations (Fig. 3, lower panel). To assess relative energies of stabilization of d-TC in clusters 1 and 2, we ran a 500-ps MD simulation of the two clusters, extracted 50 frames, and evaluated binding energy of each frame using the MM-GBSA (molecular mechanics-generalized Born surface area) method within AMBER 7 (29Srinivasan J. Cheatham III, T.E. Cieplak P. Kollman P.A. Case D.A. J. Am. Chem. Soc. 1998; 120: 9401-9409Crossref Scopus (1396) Google Scholar). By this method, the mean energy of cluster 1 was 23 kcal/mole more negative than that of cluster 2, indicating that cluster 1, which is most probable, is also the most stable. These findings agree with previous work showing that the predominant orientation produced by AUTODOCK agrees with bound ligand orientations observed in x-ray structures (30Osterberg F. Morris G.M. Sanner M.F. Olson A.J. Goodsell D.S. Proteins. 2002; 46: 34-40Crossref PubMed Scopus (413) Google Scholar). Our final predicted structure was obtained by energy minimization of the average structure of the predominant cluster followed by a 500-ps MD simulation of the AChBP-ligand complex. Unexpectedly, despite having identical molecular scaffolds and rigid cyclic structures, d-TC and metocurine are predicted to bind to AChBP in distinctly different orientations (Fig. 4). For d-TC, the tertiary nitrogen atom (N1 in Figs. 1 and 4) juxtaposes the aromatic side chain of Tyr-89 of the principal face of the binding site, whereas for metocurine, the equivalent nitrogen is remote from Tyr-89 and framed by Tyr-192 of the principle face and Trp-53 of the complementary face. This difference in positioning of N1 arises from a 170° rotation of the curare scaffold about the N1–N2 axis, plus a 30° tilt normal to the axis. Consequently, the cis-methoxy groups of metocurine are positioned for favorable hydrophobic interaction with Gln-55 at the complementary face of the binding site, whereas the equivalent methoxy and hydroxyl groups of d-TC penetrate deep into the subunit interface away from key binding site residues. For both ligands, common stabilizing residues include Trp-143 and Tyr-192 from the principal face of the binding site and Leu-112 and Met-114 from the complementary face. Table I lists computed distances between positively charged nitrogen atoms of the two ligands and key residues at the principal and complementary faces of the AChBP binding site. For each ligand the distances correspond to the orientation in the most probable cluster (cluster 1 in Fig. 3); for d-TC, distances are also listed for cluster 2. These inter-atomic distances indicate cation-π interactions that are common for both ligands as well as interactions that are unique for each one. A common interaction involves the indole ring of Trp-143 and the positively charged nitrogen N1 in both ligands; the separation of 4–5 Å is within the range expected for a stable cation-π interaction (27Dougherty D.A. Science. 1996; 271: 163-168Crossref PubMed Scopus (2357) Google Scholar). On the other hand, the ring center of Tyr-89 is more than 1 Å closer to N1 in d-TC than to the equivalent nitrogen in metocurine, whereas the ring center of Trp-53 is 1.5 Å closer to N1 in metocurine than to the counterpart in d-TC. The ring center of Tyr-192 is closer to N1 of metocurine compared with N1 in d-TC, but Tyr-192 also has a large area of contact with both ligands, suggesting similar stabilization of each. Other obvious candidates for stabilization include the two negatively charged residues Asp-108 and Glu-110 on the complementary face of the binding site, but these are far away from charged nitrogens in either ligand.Table IDistance measurement for long range electrostatic interaction between positive charge centers of ligands and AChBP residuesAtoms for measurementMetocurined-Tubocurarine (1)d-Tubocurarine (2)ÅN1(ligand)-6′ ring center (Tyr-89, P subunit)6.24.86.1N1(ligand)-6′ ring center (Trp-53, C subunit)5.26.78.8N1(ligand)-6′ ring center (Trp-143, P subunit)5.13.67.5N1(ligand)-6′ ring center (Tyr-192, P subunit)5.06.53.4N2(ligand)-CG(Asp-108, C subunit)10.815.715.8N2(ligand)-CD(Glu-110, C subunit)7.210.111.1 Open table in a new tab For d-TC the orientation in cluster 2 lacks important interactions present in the most probable cluster 1. In cluster 2 the ring center of Tyr-89 is more than 1 Å farther away from N1 than in cluster 1. Also in cluster 2, the ring center of Trp-143 is more than 2 Å farther away from N1 than in cluster 1. The lower frequency of occurrence, the lower energy of stabilization, and the following mutagenesis results indicate that cluster 2 is not the predominant orientation of bound d-TC. To test our computational results indicating different orientations of d-TC and metocurine bound to AChBP, we mutated residues from each of the seven loops that form the ligand binding site (1Sine S.M. J. Neurobiol. 2002; 3: 431-446Crossref Scopus (162) Google Scholar), transfected each mutant construct into 293 HEK cells, and purified soluble AChBP from the culture medium (see “Experimental Procedures”). Both wild type and mutant AChBP express in robust quantities and migrate as sharp bands on SDS-PAGE gels (Fig. 5). The apparent molecular mass of ∼24 kDa is in good agreement with the theoretical molecular mass of 23.8 kDa and suggests little or no glycosylation. A noteworthy exception is the mutation K34S in binding site loop D, which did not express and thus could not be examined. Also, despite robust expression of protein, the mutations Y185F and Y192T did not bind detectable amounts of α-bungarotoxin (not shown), preventing further study. Binding of each ligand was determined by competition against the initial rate of α-bungarotoxin binding. d-TC binds to native AChBP with a dissociation constant of 67 nm, similar to the previously reported value of 93 nm (6Smit A.B. Syed N.I. Schaap D. van Minnen J. Klumperman J. Kits K.S. Lodder H. van der Schors R.C. van Elk R. Sorgedrager B. Brejc K. Sixma T.K. Geraerts W.P. Nature. 2001; 411: 261-268Crossref PubMed Scopus (465) Google Scholar), whereas metocurine binds with a dissociation constant of 119 nm (Table II). Moreover, mutation of key binding site residues alters ligand affinity in accord with our computational predictions of the predominant orientation of each ligand. The mutations W53F and Q55L elicit large changes in metocurine affinity, whereas they do not affect d-TC affinity (Figs. 6 and 7). On the other hand, the mutation Y89T diminishes d-TC affinity but does not affect metocurine affinity. The mutagenesis results also confirm that a common set of residues stabilizes both ligands, including Trp-143, Tyr-192, Leu-112, and Met-114. Residues that do not contribute to ligand binding are also confirmed and include Thr-57, Asp-108, and Glu-110. Thus, our results from computation, mutagenesis, and ligand binding measurements demonstrate that d-TC and metocurine bind to AChBP in distinctly different orientations.Table IIBinding parameters for metocurine and d-tubocurarine to wild type AChBPAChBP (wt)Metocurined-TubocurarineMeanSDMeanSDKapp (nM)119 (15)aNumbers in parentheses indicate the number of independent measurements.3767 (22)20Hill coefficient0.960.200.760.14Mutagenesis of AChBP, binding affinity changes for metocurine and d-tubocurarineLog (Kmut/Kwt)MutationMetocurined-TubocurarineMeanS.D.MeanS.D.W53F0.88 (8)0.150.06 (6)0.18Q55L-0.54 (3)0.05-0.02 (5)0.01T57V0.13 (1)0.15 (5)0.12Y89T0.14 (2)0.050.80 (5)0.18D108N-0.05 (2)0.020.02 (3)0.06E110Q-0.14 (3)0.020.1 (4)0.13L112K1.13 (2)0.121.53 (4)0.23M114Y0.78 (2)0.170.75 (2)0.11M114T1.05 (1)1 (1)W143F0.75 (3)0.100.80 (2)0.06H145Y0.26 (3)0.120.44 (2)0.15Y192F-0.47 (5)0.23-0.32 (6)0.14D160A-0.03 (1)-0.03 (2)0.13D161A-0.3 (1)-0.52 (1)Y164T-0.8 (5)0.19-0.91 (3)0.25a Numbers in parentheses indicate the number of independent measurements. Open table in a new tab Fig. 7Changes in ligand binding affinity following mutation of AChBP. Changes of binding affinity for metocurine and d-tubocurarine are compared for the indicated mutations. See Table II for a summary of results for these and other mutations. Note that, although W53F, Q55L, and Y89T distinguish between the two ligands, L112K, M114T, W143F, and Y192F show similar contributions to each.View Large Image Figure ViewerDownload (PPT) Curare is the prototypical competitive antagonist of the nicotinic receptor found at the motor endplate and served as a key probe in establishing the subunit interface structure of the ligand binding sites (1Sine S.M. J. Neurobiol. 2002; 3: 431-446Crossref Scopus (162) Google Scholar). Demonstration that curariform ligands distinguish between the two binding sites of the heteropentameric AChR suggested that the different subunits neighboring the alpha subunits give rise to this distinction (17Sine S.M. Taylor P. J. Biol. Chem. 1981; 256: 6692-6699Abstract Full Text PDF PubMed Google Scholar, 18Neubig R.R. Cohen J.B. Biochemistry. 1979; 18: 5464-5475Crossref PubMed Scopus (181) Google Scholar). Subunit omission experiments, together with measurements of curare binding, confirmed that the non-alpha subunits are responsible for the ability of competitive antagonists to distinguish between the two binding sites (19Blount P. Merlie J.P. Neuron. 1989; 3: 349-357Abstract Full Text PDF PubMed Scopus (228) Google Scholar, 20Sine S.M. Claudio T. J. Biol. Chem. 1991; 266: 19369-19377Abstract Full Text PDF PubMed Google Scholar). Studies of chimeric subunits and site directed mutations showed that seven distinct regions, far apart on the protein chain, converge to form the ligand binding site; three regions originate from the alpha subunit and four from the non-alpha subunit (1Sine S.M. J. Neurobiol. 2002; 3: 431-446Crossref Scopus (162) Google Scholar, 2Corringer P.J. Le Novere N. Changeux J.P. Annu. Rev. Pharmacol. Toxicol. 2000; 40: 431-458Crossref PubMed Scopus (706) Google Scholar). The precise atomic interactions between curare and the binding site remained unknown, but promise for determining these arose with the discovery and structural determination of AChBP (5Brejc K. van Dijk W.J. Klaassen R.V. Schuurmans M. van Der Oost J. Smit A.B. Sixma T.K. Nature. 2001; 411: 269-276Crossref PubMed Scopus (1580) Google Scholar, 6Smit A.B. Syed N.I. Schaap D. van Minnen J. Klumperman J. Kits K.S. Lodder H. van der Schors R.C. van Elk R. Sorgedrager B. Brejc K. Sixma T.K. Geraerts W.P. Nature. 2001; 411: 261-268Crossref PubMed Scopus (465) Google Scholar), a prototype of the ligand binding site of nicotinic receptors. The present work uses a computational approach incorporating protein flexibility to define docking orientations of d-TC and metocurine and tests the predicted orientations by mutagenesis combined with measurements of ligand binding. Our results show that d-TC and metocurine bind to AChBP in distinct orientations in which the curare scaffold rotates 170° around the nitrogen-nitrogen axis and tilts 30° normal to this axis to contact different residues in the binding site. The standard method for computing ligand docking orientation is to use a high resolution x-ray crystal structure combined with a single, favored conformer of the ligand. Such rigid body docking provides tremendous advantages in the speed of the calculations but neglects the intrinsic flexibility of the protein. Recently, however, molecular dynamics simulations have been used to generate an ensemble of snapshots of the dynamic protein to which rigid body docking computation is applied (21Kua J. Zhang Y. McCammon J.A. J. Am. Chem. Soc. 2002; 124: 8260-8267Crossref PubMed Scopus (94) Google Scholar, 22Lin J.-H. Perryman A.L. Schames J.R. McCammon J.A. J. Am. Chem. Soc. 2002; 124: 5632-5633Crossref PubMed Scopus (362) Google Scholar). The combination of these two well-established procedures, molecular dynamics simulation and rigid body docking, accounts for flexibility inherent to proteins and allows for the theoretical possibility of conformational selection in the ligand binding process, which is an alternative pathway to the induced fit mechanism, but can yield the same final complex. Our approach includes partial flexibility of the ligand in which rotation about several bonds is allowed, but the cyclic scaffold of the ligand is maintained in the rigid conformation of the x-ray structure. The x-ray structure of d-TC was found to be similar to the structure determined by NMR (23Fraenkel Y. Gershoni J.M. Navon G. Biochemistry. 1994; 33: 644-650Crossref PubMed Scopus (18) Google Scholar), indicating that the conformation in the crystal approximates that in solution. Thus our computational analysis of curariform ligand binding to AChBP is made possible by the availability of x-ray structures of both AChBP and the curare derivatives, established MD and docking methods, and the high processing speed of present day computer clusters. What is the root cause of the different orientations of d-TC and metocurine bound to AChBP? Both ligands have the same hydrocarbon scaffold, well-defined hydrophobic concave and hydrophilic convex faces, and two electron-deficient nitrogen atoms. Also, the tetrahydroisoquinoline ring bearing nitrogen N2, and the connecting benzene and phenol rings have similar orientations in d-TC and metocurine (12Codding P.W. James M.N.G. Acta Crystallogr. 1973; 29: 935-942Crossref Google Scholar). However, in metocurine, methylation of N1 in the second tetrahydroisoquinoline ring causes it to rotate away from the center of the molecule, producing a more open structure and an increase of the N1–N2 distance from 8.97 to 10.7 Å (11Sobell H.M. Sakore T.D. Tavale S.S. Canepa F.G. Pauling P. Petcher T.J. Proc. Natl. Acad. Sci. U. S. A. 1972; 69: 2212-2215Crossref PubMed Scopus (33) Google Scholar, 12Codding P.W. James M.N.G. Acta Crystallogr. 1973; 29: 935-942Crossref Google Scholar). The docked orientations described here indicate that the conformations of free d-TC and metocurine are minimally altered when bound to AChBP. Thus the differences in bound orientations of the two ligands owe to the tertiary configuration of N1 in d-TC versus the quaternary configuration in metocurine, leading to a more compact structure of d-TC relative to metocurine and different distances separating N1 and N2. The present results illustrate the versatility of the binding site in AChBP to dock curariform ligands. Despite their very different bound orientations, d-TC and metocurine bind with affinities that differ by only 2-fold; the higher affinity of d-TC likely results from penetration of its more compact structure deeper into the binding pocket. However, both ligands are nearly completely enveloped by the binding site, maximizing van der Waals contacts, and the presence of two electron-deficient nitrogen atoms in each ligand ensures that at least one of these will be stabilized by the aromatic-rich binding pocket through cation-π interactions. These include stabilization of N1 by Trp-143 in both ligands of the principal face of the binding site and, differentially, for d-TC, by Tyr-89 of the principal face, and for metocurine, by Trp-53 of the complementary face. Tyr-192 of the principal face stabilizes both ligands but in different ways. For metocurine there is the expected π-cation interaction with N1, but for d-TC the distance of separation is too great; instead Tyr-192 makes a π-π interaction through contact with the benzene ring of d-TC. For both ligands, nitrogen N2 is located at the periphery of the docked structure and is not involved in electrostatic interactions. N2 is too far away from aromatic residues in the center of the binding pocket, and, although it is theoretically close enough to be stabilized by Asp-108 and Glu-110 at the periphery, our mutagenesis results indicate no contribution of either negatively charged side chain. A variety of hydrophobic interactions are also established with both ligands. For metocurine, the concave hydrophobic face fits into a bulge created by the disulfide-bonded cysteines 187 and 188 that form a hairpin between β-strands 9 and 10 of the principal face. For d-TC, on the other hand, the concave hydrophobic face rotates away from this hairpin to juxtapose with Met-114 from the complementary face. Similarly, Leu-112 of the complementary face makes hydrophobic contacts with the tetrahydroisoquinoline ring harboring N2 in both ligands, although opposite faces of the ring are contacted by Leu-112. The mutations M114T and L112K cause the greatest losses of affinity among the mutations studied, indicating that hydrophobic interactions are major sources of stabilization for each ligand. Conversely, Gln-55 of the complementary face closely opposes the benzene ring of metocurine, perhaps making a carbamino-π contact, whereas Gln-55 is far from d-TC, owing to its deeper penetration into the binding pocket; the mutation Q55L selectively enhances affinity of metocurine, presumably through enhanced hydrophobic contacts. Our overall results show that both d-TC and metocurine are stabilized by residues from both subunits at the binding site interface, in agreement with extensive studies of ligand binding to the nicotinic receptor at the motor endplate. However, contrary to expectations, the two ligands bind in fundamentally different orientations to AChBP, interacting with different residues in the binding pocket. These findings raise the possibility that, for the AChR at the motor endplate, different curariform ligands bind in different orientations, while still relying on contributions from both subunits at the binding site interface. Using homology models of pentameric ligand-gated ion channels (24Le Novere N. Corringer P.J. Changeau J.P. Proc. Natl. Acad. Sci. 2002; 99: 3210-3215Crossref PubMed Scopus (257) Google Scholar, 25Sine S.M. Wang H.-L. Bren N. J. Biol. Chem. 2002; 277: 29210-29223Abstract Full Text Full Text PDF PubMed Scopus (70) Google Scholar, 26Molles B.E. Tsigelny I. Nguyen P.D. Gao S.X. Sine S.M. Taylor P. Biochemistry. 2002; 41: 7895-7906Crossref PubMed Scopus (40) Google Scholar), the combination of the computational and experimental approaches described here should provide an understanding of competitive antagonist binding at the atomic structural level. We thank Dean Johnson for valuable contributions implementing computer hardware and software.