Molecular Mechanisms of Methoctramine Binding and Selectivity at Muscarinic Acetylcholine Receptors

Methoctramine (N,N'-bis[6-[[(2-methoxyphenyl)-methyl]hexyl]-1,8-octane] diamine) is an M2-selective competitive antagonist of muscarinic acetylcholine receptors and exhibits allosteric properties at high concentrations. To reveal the molecular mechanisms of methoctramine binding and selectivity we took advantage of reciprocal mutations of the M2 and M3 receptors in the second and third extracellular loops that are involved in the binding of allosteric ligands. To this end we performed measurements of kinetics of the radiolabeled antagonists N-methylscopolamine (NMS) in the presence of methoctramine and its precursors, fluorescence energy transfer between green fluorescent protein–fused receptors and an Alexa-555–conjugated precursor of methoctramine, and simulation of molecular dynamics of methoctramine association with the receptor. We confirm the hypothesis that methoctramine high-affinity binding to the M2 receptors involves simultaneous interaction with both the orthosteric binding site and the allosteric binding site located between the second and third extracellular loops. Methoctramine can bind solely with low affinity to the allosteric binding site on the extracellular domain of NMS-occupied M2 receptors by interacting primarily with glutamate 175 in the second extracellular loop. In this mode, methoctramine physically prevents dissociation of NMS from the orthosteric binding site. Our results also demonstrate that lysine 523 in the third extracellular loop of the M3 receptors forms a hydrogen bond with glutamate 219 of the second extracellular loop that hinders methoctramine binding to the allosteric site at this receptor subtype. Impaired interaction with the allosteric binding site manifests as low-affinity binding of methoctramine at the M3 receptor.


Introduction
Five muscarinic receptor subtypes have been cloned (Bonner et al., 1987;Bonner, 1989). They all belong to the family of G-protein-coupled receptors with seven transmembrane helices. The binding site of acetylcholine and competitive agonists and antagonists is located deep in a pocket formed by transmembrane helices (Wess, 1996). All subtypes of muscarinic receptors share high homology in the transmembrane domains, making it difficult to discover competitive ligands that are selective for different receptor subtypes. However, a few such selective compounds have been identified. One example is methoctramine (N,N'-bis[6-[[(2-methoxyphenyl)-methyl]hexyl]-1,8-octane] diamine), which binds to the M 2 subtype of muscarinic receptors with high affinity (Melchiorre et al., 1987) and to the M 3 subtype with low affinity (Caulfield, 1993). However, deviations from competitive behavior (slowdown of dissociation of orthosteric antagonists) at high concentrations of methoctramine were revealed soon after the discovery of its selectivity (Giraldo et al., 1988;Lee et al., 1989). Additional investigations have concluded that methoctramine interacts with muscarinic receptors both competitively and allosterically (Waelbroeck, 1994;Boselli and Grana, 1995), and the concept that methoctramine binds simultaneously to the orthosteric and allosteric receptor domains was proposed (Melchiorre et al., 1989). In the present work, we provide experimental support for this concept.
In this study, we investigated whether the extracellular domains of muscarinic receptors are also involved in putative allosteric properties of methoctramine binding, and if they take part in high affinity of the M 2 receptor for methoctramine. To achieve this aim, we availed differences of methoctramine binding between the M 3 and M 2 receptor subtypes. We modified the gene of the M 3 receptor in parts that encode the extracellular domains so that the resulting amino acid sequence resembles that in the M 2 receptor. Using this approach, we demonstrate that methoctramine selectivity for the M 2 receptor arises from its binding to glutamate residues in the second extracellular loop. A similar interaction at the M 3 receptor is prevented by interaction with K523 in the third extracellular loop.

Mutagenesis and Expression
For the sake of brevity, names of mutants consist of receptor subtype followed by a list of mutated amino acids in the third extracellular (o3) loop. Amino acids in either the M 2 or M 3 receptors were always mutated to the corresponding residue in the other receptor subtype (Supplemental Fig. 3). For example, M 2 P means that proline 415 of the M 2 receptor was mutated to the corresponding serine in the M 3 sequence, and M 3 DSKFN means that five amino acids in the o3 loop of the M 3 receptor were mutated to their corresponding residues in the M 2 sequence.
The mammalian expression vector pcDNA3.1 (Invitrogen, Carlsbad, CA) containing the coding sequence of the human variants of M 2 and M 3 subtype of muscarinic acetylcholine receptors was obtained from Missouri S&T cDNA Resource Center (Rolla, MO). Construction of M 3 receptors with the substituted second extracellular (o2) loop and mutations (K523N, KFN, and DSKFN) in the o3 loop were previously described in Krejcí and Tu cek (2001). Additional mutants were generated using the QuikChange II Site-Directed Mutagenesis Kit (Agilent Technologies Company, Santa Clara, CA). To replace serine 519 with proline, a plasmid containing wild-type M 3 receptor as template and 59-TGA ACA CCT TTT GTG ACC CCT GCA TAC CCA AAA CCT TTT GG-39 primer were used. To obtain DKFN mutant, plasmid containing KFN mutant as template and 59-GTG AAC ACC TTT TGT GCC AGC TGC ATA CCC AAT ACC-39 primer were used. To obtain SK mutant plasmid containing K523N mutant as template and 59-TGA ACA CCT TTT GTG ACC CCT GCA TAC CCA ATA CCT TTT GG -39 primer were used. To obtain SKFN mutant, plasmid containing KFN mutant and 59-TGA ACA CCT TTT GTG ACC CCT GCA TAC CCA ATA CCG TTT GG-39 primer were used. To replace proline 415 of the M 2 receptor with serine, a plasmid containing wildtype M 2 receptor as template and 59-AAC ACC TTT TGT GCA TCT TGC ATC CCC AAC ACT GTG-39 primer were used.
Chimeric proteins used in this study consist of three parts: chicken a 7 nicotinic signal peptide, emerald green fluorescent protein (GFP), and human muscarinic M 2 (hM 2 ) or M 3 (hM 3 ) receptor. The DNA sequence of chicken a 7 nicotinic signal peptide (a peptide 31 amino acids long with extracellular protease cleavage site included; Ilien et al., 2003) was fused directly to stop-codon-free GFP (Invitrogen) at its 59 end to ensure extracellular localization of fluorescent protein. cDNA for the hM 2 receptor (cloned into pcDNA3.11 vector) were mutated (QuikChange II Site-Directed Mutagenesis Kit) to obtain AgeI restriction site at the 59 cDNA end. EcoRI and AgeI (New England Biolabs, Ipswitch, MA) double digestion opened the vector and created two cohesive ends for ligation. T4 ligation (New England Biolabs) reaction included opened hM 2 receptor vector, signal peptide-GFP complex, and short linker. hM 2 cDNA was substituted with hM 3 cDNA to obtain GFP-tagged hM 3 chimeric protein. All constructs were sequenced before use. Chimeric protein expression in eukaryotic cell lines was verified under a fluorescent microscope and via [ 3 H]NMS binding (described later). All mutated DNAs were sequenced with the dideoxy method by Dr. J. Felsberg (Academy of Sciences, Institute of Microbiology, Prague, Czech Republic). COS-7 cells were transfected using the DEAE-dextran method. They were grown in 10-cm Petri dishes in Iscove's modified Dulbecco's medium (Sigma-Aldrich) with 10% fetal calf serum. On day 1, 2Â10 6 cells were seeded per dish. On day 3, after washing with phosphatebuffered saline, the transfection mix (2 mg of plasmid DNA and 0.27 mg of DEAE-dextran in 1 ml of phosphate-buffered saline) was applied for 2 hours. Serum-free Iscove's modified Dulbecco's medium with chloroquine (80 mM in 7 ml of medium) was then added for an additional 3 hours. Transfection medium was then removed and fresh medium supplemented with 10% fetal calf serum was applied. Cells were harvested 72 hours after transfection.

Radioligand Binding Experiments
Radioligand binding experiments were performed on membranes of COS-7 cells. Membranes were obtained by dilution of freshly harvested COS-7 cells in a medium composed of 136 mM NaCl, 5 mM KCl, 1 mM MgCl 2 , 10 mM EDTA, and 10 mM Na-HEPES (pH 5 7.4) to a final concentration of 10 7 cells/ml. Cell suspension was homogenized using an Ultra-Turrax homogenizer (Janke & Kunkel IKA-Labortechnik, Staufen, Germany) by two 30-second strokes. Homogenate was centrifuged for 5 minutes at 1000 Â g. The resulting supernatant was centrifuged for 30 minutes at 30,000 Â g and the supernatant was discarded. Pellet was resuspended in 10 times the original volume (before centrifugation), left at 4°C for 30 minutes, and then the latter centrifugation step was repeated. Membranes were kept frozen at 220°C for a maximum of 1 month. Binding experiments to membranes from 1 to 2 million cells per tube were performed as described earlier (Jakubík et al., 1995). The medium used for incubation was the same as described earlier without EDTA, and the incubation volume was 0.8 ml. Incubations performed at 25°C were terminated by filtration through Whatman GF/C glass fiber filters in a Brandel filtration apparatus (Sensat, Herts, UK). Nonspecific binding was determined in the presence of 1 mM NMS.
The affinity of wild-type and mutated muscarinic receptors for [ 3 H]NMS was measured in saturation binding experiments (1-hour incubation with [ 3 H]NMS at concentrations ranging from 32 pM to 1 nM) and expressed as K d(NMS) (equilibrium dissociation constant for the binding of [ 3 H]NMS). The inhibition constants K i of receptors for methoctramine, 4-DAMP, AF-DX 116, and himbacine were determined in competition experiments in which membranes were incubated in the presence of a fixed 200 pM concentration of [ 3 H]NMS and increasing concentrations of the competitor. The incubation lasted 5 hours to achieve full equilibrium. In dissociation experiments, membranes were preincubated for 1 hour with 750 pM [ 3 H]NMS. Dissociation was induced by adding 50 ml of NMS to a final concentration of 1 mM. NMS was added either alone or in mixture with methoctramine or methoctramine precursors at final concentrations ranging from 10 pM to 1 mM.

Fluorescence Resonance Energy Transfer
Fluorescence resonance energy transfer (FRET) between GFP-M 2 or GFP-M 3 receptors and conjugate of long methoctramine precursor (LMP) and Alexa-555 (Alexa-555-LMP) was measured in 96-well plates using a PerkinElmer Victor X4 plate reader (PerkinElmer, Waltham, MA). The following set of filters from Knight Optical (Harrietsham, UK) was used: excitation 470/10 nm, GFP emission 532/10 nm, and Alexa-555 emission 590/10 nm. Acquisition time was 0.1 second. Membranes prepared from COS-7 cells transiently transfected with GFP-M 2 or GFP-M 3 genes (about 50 mg of protein per well) were used. Association was initiated by the addition of Alexa-555-LMP to final concentrations ranging from 30 nM to 1 mM. Dissociation was started by addition of methoctramine to a final concentration of 1 mM. The final sample volume was 0.2 ml. Besides samples containing both fluorescent probes, wells with only GFP-labeled receptors and Alexa-555-LMP (with an added comparable amount of nontransfected COS-7 membranes) were measured to check for fluorescence bleaching. Up to 15% of bleaching was observed for GFP during the 6-hour experiment. No bleaching was observed for Alexa-555-LMP.

Data Treatment
Data were processed and analyzed with open-source software OpenOffice 3.4 (OpenOffice Foundation, www.openoffice.org) and Grace 5.1.12 (Grace Development Team, plasma-gate.weizmann.ac.il/ Grace) on Scientific Linux (www.scientificlinux.org).
The equations for nonlinear regression analysis were as follows: For Saturation Binding Experiments. After subtraction of nonspecific binding, eq. 1 was fitted to the data: where Y is [ 3 H]NMS binding at concentration X of free [ 3 H]NMS, K D is the equilibrium dissociation constant, and B MAX is the number of binding sites.
For Competition Experiments. After subtraction of nonspecific binding and normalization (to express the binding of [ 3 H]NMS in the presence of competitor [methoctramine, 4-DAMP, AF-DX 116, or himbacine) as a percentage of the binding in the absence of competitor), eq. 2 was fitted to the data: where Y is [ 3 H]NMS binding at logarithm of concentration of competitor X, and IC 50 is the concentration of competitor that results in 50% of maximal inhibition. Inhibition constant K i was calculated from the IC 50 value according to Cheng and Prussof (1973):

where [NMS] is the actual concentration of [ 3 H]NMS used in the experiment, and K D is the equilibrium dissociation constant of [ 3 H]NMS.
Dissociation Experiments. After subtraction of nonspecific binding, eq. 4 was fitted to the data: Y 5 100*e ðÀKoff ðobsÞ*XÞ where Y is [ 3 H]NMS binding at time X, and k off(obs) is the observed dissociation rate constant. Apparent equilibrium dissociation constant K D for methoctramine based on dissociation experiments was obtained according to Lazareno and Birdsall (1995): where Y is the observed rate of dissociation k off(obs) at concentration X of the allosteric ligand [methoctramine or (6-aminohexyl)[(2methoxyphenyl)methyl]amine, short methoctramine precursor (SMP)], and k 0 is the observed dissociation rate in the absence of the allosteric ligand.
The error distributions for individual constants were verified according to Christopoulos (1998). IC 50 (eq. 2), and consequently K i (eq. 3), and K D (eq. 5) have log-normal error distribution. Error distributions for NMS K D (eq. 1) and k off(obs) (eq. 4) conform to normal Gaussian distribution.
Parameters of binding kinetics from FRET measurements were obtained by fitting a tandem two-site model (Jakubík et al., 2000) to  the pooled data with subtracted background values using the program COPASI (www.copasi.org) (Hoops et al., 2006). Initial parameter estimates and background values were obtained by fitting two exponential growth and two exponential decay functions, respectively, to the individual data sets using the program Grace.

Molecular Modeling
Preparation of Structures and Systems. The structure of methoctramine (CID 4108) was downloaded from the PubChem database (pubchem.ncbi.nlm.nih.gov) and processed with Schrödinger's (Schrödinger LLC, New Portland, OR) LigPrep. Crystal structures of M 2 (Haga et al., 2012) (PDB code 3UON) and M 3  (PDB code 4DAJ) receptors were downloaded from the RCSB Protein Data Bank (www. rcsb.org), and were preprocessed with Schrödinger's Protein Preparation Wizard to remove nonreceptor parts, fill missing side chains, and energy minimize structures in the OPLSA 2005 force field. A system consisting of receptor, 1,2-dipalmitoylphosphatidylcholine membrane, water, and 0.15 M NaCl with or without methoctramine randomly placed close to the receptor extracellular domain was built with the Desmond System Builder (Bowers et al., 2006).
Simulation of Molecular Dynamics. Molecular dynamics of full membrane systems was simulated using Desmond (Bowers et al., 2006). First, systems were relaxed to 300 K using standard Desmond protocol for membrane systems and subsequently 120 ns of free (without restraints) molecular dynamics (MD) (ensemble class NVE, Coulombic short-range method: cutoff with radius 9 Å, long-range method: smooth particle mesh Ewald) was simulated using Desmond-GPU.
Six systems with methoctramine were built and used for simulation of methoctramine association. After system relaxation prior to free MD simulation, 25 ps of steered MD was run, during which acceleration of 500 pm×ps 22 toward the center of the orthosteric binding site (defined by D103 and N404) was applied to one anisole group of methoctramine. Steered MD and free MD were repeated until the distance between the centers of the anisole group and the orthosteric binding site reached 5 Å. Then 500 ns of free MD was simulated.

Results
Equilibrium Experiments: Methoctramine. In preliminary experiments, we measured the effects of replacement of the entire o2 or o3 loop of the M 3 receptor with the corresponding M 2 sequence on the binding of the muscarinic M 2 -selective ligands AF-DX 116, himbacine, and methoctramine ( Fig. 1) and the M 3 -selective ligand 4-DAMP. Replacement of the o2 loop of the M 3 receptor with that of the M 2 receptor did not change the affinity of the tested ligands except for 4-DAMP, where affinity of the hybrid receptor fell far below the value of the M 2 wild-type receptor ( Table 1). Replacement of the o3 loop of the M 3 receptor with that of the M 2 receptor did not change the affinity of 4-DAMP, AF-DX 116, or himbacine but increased the affinity of methoctramine to a value close to that at the M 2 receptor. (Fig. 2) In subsequent experiments, we measured binding of methoctramine to the seven mutant receptors with mutations in the o3 loop. Mutation of serine 519 (M 3 numbering) to corresponding  proline (M 3 S) only slightly increased the affinity of methoctramine. Mutation of lysine 523 (M 3 numbering) to corresponding asparagine (M 3 K) increased the affinity of methoctramine more than six times. Combination of the M 3 K mutation with the mutation of phenylalanine 525 to valine, asparagine 527 to threonine, and aspartate 518 to alanine, i.e., M 3 KFN and M 3 DKFN, respectively, did not further change the affinity of methoctramine. Double-mutation M 3 SK increased the affinity of methoctramine to a value close to that at the wild-type M 2 receptor. Combination of the double-mutation M 3 SK with mutation of F525, N527, and D518, i.e., M 3 SKFN and M 3 DSKFN, did not further change the affinity of methoctramine (Table 2). Taken together, only the M 3 S and M 3 K mutations increased affinity of methoctramine. Mutation M 3 K contributed most to affinity increase. In accordance with these observations, mutation of proline 415 (M 2 numbering) in o3 loop of the M 2 receptor to serine, i.e., M 2 P, brought about a 10-fold decrease in affinity of methoctramine (Table 2).
Kinetic Experiments: Methoctramine. Kinetic binding experiments were performed to ascertain putative allosteric properties of methoctramine. Membranes were first preincubated with [ 3 H]NMS for 1 hour, then dissociation was started by the addition of unlabeled NMS to a final concentration of 1 mM, either alone or in mixture with various concentrations of methoctramine. In the absence of methoctramine, dissociation of [ 3 H]NMS from the M 2 receptors is about six times faster than that from the M 3 receptors ( Fig. 3; Table 3). In the presence of 100 mM methoctramine, the dissociation from the M 2 receptor was slowed down .50-fold (Fig. 3, top panel; Table 3). In contrast, the observed dissociation rate of [ 3 H]NMS from the M 3 receptor did not significantly differ in the presence of 100 mM methoctramine (Fig. 3, bottom panel; Table 3).
Interestingly, all tested mutations in the o3 loop of the M 3 receptor, except M 3 S, accelerated [ 3 H]NMS dissociation in the absence of methoctramine (second column of Table 3 [ 3 H]NMS dissociation in the presence of 100 mM methoctramine was slower at all mutants (except for M 3 S) than at the M 3 wild type. In fact, the rate of radioligand dissociation at M 3 DKFN, M 3 SK, M 3 SKFN, and M 3 DSKFN mutants was the same as at the M 2 wild type (wt; third column of Table 3). In other words, mutation of only a single residue, K523 to asparagine, elicited allosteric interaction between methoctramine and [ 3 H]NMS on the M 3 receptor, inferred from accelerating the speed of dissociation. In accordance with these observations, mutation of P415 in the o3 loop of the M 2 receptor to serine, i.e., M 2 P, did not change [ 3 H]NMS dissociation significantly, either in the absence or in the presence of methoctramine (Table 3). Only mutations of K523, F525, and N527 (i.e., M 3 KFN, M 3 DKFN, M 3 SKFN, and M 3 DSKFN) significantly altered the ratio of K off(obs) in the absence of methoctramine to K off(obs) in the presence of 100 mM methoctramine (fourth column of Table 3). Again, this parameter was the same for the M 3 SKFN, M 3 DSKFN, and the M 2 wt receptor.
In the next set of experiments, we measured the effects of various concentrations of methoctramine on [ 3 H]NMS dissociation on wild-type receptors and all mutants. The observed dissociation rate constants [K off(obs) ] were calculated according to eq. 4. Calculated K off(obs) were plotted as a function of methoctramine concentration (Fig. 4). Values of apparent equilibrium dissociation constants (K D ) for methoctramine binding to [ 3 H]NMS-occupied receptors were obtained by fitting eq. 5 to the data in Fig. 4 (top panel) and are shown in Table 3. Except for the M 3 S mutant, the calculated K D values are close to the K D of M 2 wt. In other words, mutation of K523 (M 3 K) is crucial for allosteric properties of methoctramine binding.
To gain deeper insight into the mechanisms underlying methoctramine slowing down of [ 3 H]NMS dissociation, we measured [ 3 H]NMS dissociation from M 2 receptors in the presence of concentrations of methoctramine that saturate its  Table 3. binding to [ 3 H]NMS-occupied receptors (Fig. 4, bottom panel). Dependence of K off(obs) on the concentration of methoctramine fits eq. 5 well [negative logarithm of equilibrium dissociation constant (pK D ) 5 5.2 6 0.2 (mean 6 S.E.M., n 5 3)]. A good fit to eq. 5 even at high methoctramine concentrations means that dissociation limits to 0, i.e., methoctramine prevents radioligand dissociation completely.
Equilibrium Experiments: Methoctramine Precursors. To gain insight into substuctures of methoctramine responsible for interaction with the receptor, we synthesized a short and a long methoctramine precursor-namely, SMP and 8-amino-N-(6-{[(2-methoxyphenyl)methyl]amino}hexyl)octanamide (LMP) (Fig. 1). Binding of these precursors was measured at M 2 and M 3 wild-type receptors and M 3 S, M 3 K, M 3 SK, and M 3 DSKFN mutants. Affinity of SMP was lower than the affinity of methoctramine at both M 2 and M 3 wild-type receptors, whereas affinity of LMP was lower than affinity of methoctramine only at the M 2 receptor (Fig. 5). Moreover, at the M 2 receptor, SMP affinity was lower by 100-fold, whereas LMP affinity was only three times lower than affinity of methoctramine (Table 2). Both single mutations in the o3 loop of the M 3 receptor, S519P (M 3 S) and K523N (M 3 K), increased the affinity of both precursors. Similar to methoctramine, the increase in affinity was greater at M 3 K than at M 3 S. Mutants M 3 SK and M 3 DSKFN displayed the same affinity for both precursors as at M 2 wt.
Kinetic Experiments: SMP. To test the allosteric properties of SMP, we tested its effects on [ 3 H]NMS dissociation as described earlier for methoctramine. In the presence of 100 mM SMP, radioligand dissociation from the M 2 wt receptor was slowed down by more than 30-fold ( Fig. 6; Table 3 Table 3. Except for the M 3 S mutant, the calculated K D values are close to the K D at the M 2 wt receptor. Importantly, K D values of SMP were the same as those of methoctramine. As in the case of methoctramine, dependence of K off(obs) of dissociation from the M 2 wt receptor at concentrations of SMP that saturate binding to [ 3 H]NMS-occupied receptors (Fig. 6, bottom panel) fits eq. 5 well [pK D 5 5.5 6 0.2 (mean 6 S.E.M., n 5 3)].
Kinetics Experiments: FRET. For FRET measurements, we constructed emerald GFP attached to the N terminus of either M2 wt or M3 wt receptors as donors. As an acceptor, LMP was conjugated with Alexa-555 (Supplemental Methods). Table 1). The affinity of LMP-Alexa-555 in competition experiments with [ 3 H]NMS at M 2 wt and M 3 wt receptors was the same as that of LMP (Supplemental Table 1). Measurements of FRET allowed direct assessment of binding kinetics of LMP-Alexa-555 (Figs. 7 and 8). LMP-Alexa-555 was added to membranes from COS-7 cells transiently expressing GFP-M 2 receptors (Fig. 7) at final concentrations of 30 (top panel), 100 (middle panel), and 300 nM (bottom panel), and FRET signal was followed for 4 hours (black). Dissociation was initiated by the addition of methoctramine at a final concentration of 1 mM after 5 (red), 60 (green), or 120 minutes (blue) of association. Association was biphasic with a very rapid initial phase followed by a phase with an observed association rate around 0.02 minute -1 (analytical data are in Supplemental Table 2). Nonspecific signal was very strong, being about two-thirds of the total signal under equilibrium even at the lowest concentration. With increasing concentrations, specific signal increased as well as the proportion of fast association that rose from 15% at 30 nM to 53% at 300 nM. The observed association rate of the slower phase remained the same. Dissociation of LMP-Alexa-555 from M 2 receptors was also biphasic. The rate of the fast dissociation phase could not be determined as it was below assay resolution. The rate of slow dissociation was around  Fig. 3 using eq. 4. Values of apparent equilibrium dissociation constant (K D ) for methoctramine and SMP were obtained by fitting eq. 5 to the data in Figs. 4 and 6  Mechanism of Methoctramine Binding 0.007 minute -1 . The proportion of slow dissociation increased with both concentration of LMP and the length of association. The rate of slow dissociation was independent of both the duration of association and concentration used. Fitting a receptor-antagonist complex isomerization model (Järv et al., 1979) to the kinetic data in Fig. 7 using program COPASI provided the following results: k 1 1 5 2 Â 10 8 M -1 ×min -1 ; k -1 5 100 minute -1 ; k 12 5 0.07 minute -1 ; k -2 5 0.008 minute -1 . Higher concentrations of LMP-Alexa-555 were used in FRET measurements at GFP-M 3 receptors because of the lower affinity of methoctramine and its precursors at the wild-type M 3 receptor (Fig. 8). Even still, higher concentrations of LMP-Alexa-555 would be required to achieve comparable occupancy at the M 3 receptors. However, at a 3 mM concentration of LMP-Alexa-555, the background signal was already too strong and background noise obscured a specific signal. Due to slower kinetics of LMP-Alexa-555 at M 3 than M 2 receptors, dissociation was initiated after 60 minutes (green), 2 hours (blue), and 4 hours (yellow) of association. Association was biphasic with a very rapid initial phase followed by a phase with an observed association rate around 0.01 minute -1 (analytical data are in Supplemental Table 3). Dissociation of LMP-Alexa-555 from M 3 receptors was also biphasic. The rate of the fast dissociation cannot be determined as it is below assay resolution. The rate of slow dissociation was around 0.004 minute -1 . Fitting a receptorantagonist complex isomerization model (Järv et al., 1979) to the kinetic data in Fig. 8 using program COPASI gave the following results: k 11 5 2 Â 10 8 M -1 .min -1 ; k -1 5 2000 minute -1 ; k 12 5 0.02 minute -1 ; k -2 5 0.004 minute -1 .

Attachment of GFP to the N termini of the receptors did not change their binding properties as assessed in [ 3 H]NMS saturation and competition experiments (Supplemental
Molecular Modeling. Molecular dynamics of membrane systems with either the M 2 or M 3 receptor were run as described in Materials and Methods and analyzed with Desmond Simulation Interactions Diagram. In analysis of free molecular dynamics without methoctramine, we focused on the o2 and o3 loops. Although in the crystal structure of the M 3 receptor (4DAJ) K523 forms a hydrogen bond with Y127 in the transmembrane a-helix (TM) II (Fig. 9, top-left panel), simulation of molecular dynamics shows that, at 59% of time, K523 forms a hydrogen bond with the conserved glutamate E219 (Fig. 9, topright panel). In addition, K523 binds to E219 by ionic interaction at 18% of time and by water bridge at 21% of time (Supplemental  Figs. 4 and 5). N419 at the M 2 receptor corresponds to K523 of the M 3 receptor. The side chain of N419 is free in the M 2 crystal structure (Fig. 9, bottom-left panel). Simulation of molecular dynamics indicated that N419 is free most of the time. This also applies to E175 (corresponding residue to E219 in the M 3 sequence) (Fig. 9, bottom-right panel; Supplemental Figs. 6 and 7).
In simulations of molecular dynamics of association of methoctramine with the M 2 receptor, methoctramine was placed into the extracellular vestibule by steered molecular dynamics. In continuing molecular dynamics simulations, methoctramine nitrogen in atom position 10 forms a hydrogen bond with E175 at the o2 loop within the first 30 ns (Fig. 10, left panel;Fig. 11,top panel;Supplemental Figs. 8 and 11). After four cycles of alternating free and steered molecular dynamics (480 ns), the anisole group of methoctramine reached the orthosteric binding site defined by D103 and N404. During subsequent free molecular dynamics, methoctramine nitrogen in atom position 10 forms a hydrogen bond with D103, the anisole ring interacts with Y104 in the orthosteric site by p-p stacking interaction, and methoctramine nitrogen in atom position 33 forms a hydrogen bond with E175 in o2 loop (Figs. 10 and 11,middle panel;Supplemental Figs. 9 and 12). Two cycles of free molecular dynamics (total 240 ns) showed the methoctramine anisole group moving toward TM VI to interact with Y403 by p-p stacking interaction, and that the methoctramine oxygen in atom position 7 forms a hydrogen bond with N404 (Fig. 10, right panel; Fig. 11, bottom panel; Supplemental Figs. 10 and 13). Additionally, methoctramine nitrogen in atom position 17 forms a hydrogen bond with D103 in the orthosteric binding site, and nitrogen in atom position 33 forms a hydrogen bond with either E172 or E175 in the o2 loop (E172 59% and E175 35% of the time).   Table 2.

Discussion
In this study, we delineated the molecular mechanism of methoctramine ( Fig. 1) binding and selectivity toward M 2 muscarinic receptors. We show that interaction of methoctramine with both the orthosteric and allosteric binding sites is responsible for its high affinity at the M 2 receptor. Further, we show that K523 in the third extracellular (o3) loop of the M 3 receptor interacts with E219 in the second extracellular (o2) loop and hinders interaction of methoctramine with the allosteric site, which results in low affinity at the M 3 receptor.
Our results confirm earlier findings that, in addition to the high-affinity competitive interaction with [ 3 H]NMS at the M 2 muscarinic receptors, methoctramine also binds to an allosteric domain on the M 2 muscarinic receptor. It has been noted that methoctramine slows down [ 3 H]NMS dissociation from cardiac membranes (Giraldo et al., 1988;Waelbroeck, 1994), and that methoctramine noncompetitively inhibits carbachol-induced phosphoinositide hydrolysis (Lee et al., 1989) and left atria functional response to carbachol (Boselli and Grana, 1995). Our results directly demonstrate an allosteric feature of methoctramine binding at the M 2 receptor and its lack at M 3 receptors (Fig. 3).
To elucidate the amino acid residues involved in high-affinity orthosteric and allosteric binding of methoctramine, we genetically modified the M 3 receptors to resemble the M 2 receptors. We focused on the extracellular domains where allosteric ligands such as alcuronium or gallamine (Krejcí and Tu cek, 2001) have been proposed to bind. We found that replacement of the whole o3 loop of the M 3 receptor with the corresponding sequence of the M 2 receptor increases the affinity of methoctramine and is thus responsible for high-affinity methoctramine binding. On the other hand, finding that this modification of the M 3 receptor does not influence the affinity of two other M 2 -selective ligands, AF-DX 116 and himbacine, and the M 3 -selective ligand 4-DAMP (Table 1) proves that the o3 loop is not involved in M 2 selectivity in a general sense, and means that different ligands gain their selectivity by interacting with different regions on the receptor. Replacement of the o2 loop of the M 3 receptor by the o2 loop of the M 2 receptor does not influence the affinity of methoctramine. This mutation causes a decrease in affinity of the M 3 -selective 4-DAMP, indicating involvement of the o2 loop in its high affinity for the M 3 receptor. These observations demonstrate that at least two additional domains exist which can determine ligand affinity besides the orthosteric site.
Sequence of the o3 loop of the M 2 and M 3 receptors differs in five positions (Supplemental Fig. 3). Mutation of just K523 of the M 3 receptor to asparagine increases the affinity of methoctramine (Fig. 2) and uncovers allosteric properties of methoctramine, similar to those observed at the M 2 receptor (Fig. 4). Mutation K523N is thus sufficient for methoctramine binding to an allosteric site from which it slows down [ 3 H]NMS dissociation. We deduce that methoctramine interacts directly with the extracellular domain of the M 2 receptor, and that this interaction is contingent on the presence of N419 in the o3 loop. This notion is further supported by the observation that the mutation M 2 P decreases affinity of the M 2 receptor for both [ 3 H]NMS and methoctramine (Table 2).
Equilibrium binding experiments on the M 3 receptor mutated at individual positions to corresponding amino acids of the M 2 sequence indicate that the double mutation of K523N and S519P is virtually sufficient to equalize the affinity of methoctramine at the mutated M 3 and the wild-type M 2 receptors (Fig.  2). The mutation S519P at the o3 loop of the M 3 receptor does not change affinity for methoctramine and does not reveal allosteric binding properties of methoctramine. These observations show that N419 of the M 2 receptor is solely required for methoctramine binding to the extracellular domain. However, N419 as well as P415 in the o3 loop of the M 2 receptor are required for high-affinity binding (   Table 3. and approaches a limit of zero at concentrations above those necessary to saturate binding of methoctramine to its secondary allosteric binding site (Fig. 4). This demonstrates that, similar to other allosteric modulators (Pro ska and Tu cek, 1994), methoctramine sterically prevents [ 3 H]NMS dissociation.
To gain insight into substructures involved in methoctramine high-and low-affinity binding, we tested interactions of two methoctramine precursors, the short and the long methoctramine precursors (SMP and LMP) (Fig. 1). Both precursors have lower affinity than methoctramine at the M 2 receptor, Fig. 9. Interaction of K522 at the M 3 receptor and lack of interaction of corresponding N419 in the M 2 receptor. Extracellular view of initial structures (left panel) and average structures (right panels) of simulation of free molecular dynamics of the M 3 (top panels) and M 2 (bottom panels) receptors. Red, TM II; white, o2 loop; blue, o3 loop and TM VII; yellow, hydrogen bond. Elements: cyan, carbon; red, oxygen; blue, nitrogen; white, hydrogen.  whereas only SMP affinity was lower at the M 3 receptor (Table 2). Mutations in the o3 loop of the M 3 receptor affected the affinity of both methoctramine precursors in the same way as it modified the affinity of methoctramine. SMP possesses allosteric properties as evidenced by the slowing down of [ 3 H]NMS dissociation from the M 2 receptor and from the M 3 K, M 3 SK, and M 3 DSKF mutants ( Fig. 6; Table 3). The equilibrium dissociation constant of SMP binding to [ 3 H]NMS-occupied M 2 receptors is the same as that of methoctramine, indicating that this part of the methoctramine molecule is sufficient for binding to the allosteric domain. The observed lower affinity of SMP than that of methoctramine under equilibrium binding to the M 2 receptor indicates that both parts of the methoctramine molecule are involved in (needed for) methoctramine highaffinity binding.
FRET of binding kinetics of LMP at the M 2 receptors (Fig. 7) showed biphasic association as well as biphasic dissociation with the proportion of slow dissociation increasing with time of association. These results are compatible with at least two receptor-ligand interaction models: 1) isomerization of the receptor-antagonist complex (Järv et al., 1979) and 2) the tandem two-site model (Jakubík et al., 2000). The same pattern of binding kinetics of LMP at the M 3 receptors (Fig. 8) can be observed. Much higher concentrations of LMP were required at M 3 receptors for the same effect as at M 2 receptors, indicating dramatically lower affinity of LMP for the allosteric binding site at the M 3 than the M 2 receptors.
Simulations of molecular dynamics of methoctramine association with the M 2 receptor showed that binding is initiated by interaction with E175 at the o2 loop (Fig. 10, left panel) followed by slow translocation to the orthosteric binding site (Fig. 10, middle panel) and methoctramine binding equilibration (Fig. 10, right panel). Thus, initial interaction with E175 (Fig.  11, top panel) represents the fast association step observed in FRET measurements (Fig. 7) and results in methoctramine low-affinity binding that prevents NMS dissociation. Slow translocation then represents the slow association step in FRET measurements (Fig. 7) and manifests itself as slowly dissociating sites whose proportion increases with time of association (proportion of occupied binding sites). Methoctramine interacts with both the orthosteric (D103, Y403, N404) and allosteric (E172, E175) sites (Fig. 11,bottom panel), and this dual interaction results in methoctramine high-affinity binding. As is evident from the molecular model (Fig. 10, middle and right panels; Fig. 11, middle and bottom panels), SMP is too short to interact with both sites and therefore has lower affinity than methoctramine. At the M 3 receptors, methoctramine competes with K523 for interaction with E219, which is, unlike at the M 2 receptor, the only glutamate in the middle of the o2 loop, and thus its binding to the allosteric site is impaired and contributes to low affinity of methoctramine binding at the M 3 receptors under equilibrium.
This work has practical implications. The major difficulty in producing muscarinic subtype-selective ligands is due to the structurally conserved orthosteric binding site. Understanding the molecular mechanisms of methoctramine binding may be useful for designing a novel family of selective compounds that combine elements of known high-affinity orthosteric ligands with known selective allosteric ligands in one molecule. Interaction of such hybrids with the orthosteric binding site would endow them with high affinity, whereas binding to the less conserved extracellular domain would give them subtype selectivity (Mohr et al., 2004;Antony et al., 2009). Modern methods using fluorescent ligands to detect ligand binding have been described (Ilien et. al., 2003;Daval et al., 2012). A common problem with fluorescent labeling of ligands is that fluorescent probe alters ligand affinity. Presented results show that ligand affinity need not be affected by fluorescent probe.
In summary, we demonstrate that the high-affinity methoctramine binding to the M 2 receptors is due to the simultaneous interaction with both the orthosteric and the allosteric binding sites. At the orthosteric binding site, methoctramine forms hydrogen bonds with D103 and N404, and interacts with Y403 via p-p stacking interaction. At the allosteric binding site, methoctramine forms a hydrogen bond alternating between E172 and E175. Methoctramine can bind to the NMS-occupied receptor with low affinity by interaction solely with the allosteric binding site. Although in such cases the interaction between methoctramine and NMS is allosteric (is not mutually exclusive), NMS cannot leave the complex in the presence of methoctramine, which physically prevents its dissociation. Lysine 523 in the o3 loop of the M 3 receptor interacts with E219 in the o2 loop and hinders methoctramine binding to the allosteric site. It results in low affinity of methoctramine binding and lack of allosteric properties at the M 3 receptors.