Evidence for a Tandem Two-site Model of Ligand Binding to Muscarinic Acetylcholine Receptors*

After short preincubations with N -[ 3 H]methylscopol-amine ([ 3 H]NMS) or R ( 2 )-[ 3 H]quinuclidinyl benzilate ([ 3 H]QNB), radioligand dissociation from muscarinic M 1 receptors in Chinese hamster ovary cell membranes was fast, monoexponential, and independent of the concentration of unlabeled NMS or QNB added to reveal dissociation. After long preincubations, the dissociation was slow, not monoexponential, and inversely related to the concentration of the unlabeled ligand. Apparently, the unlabeled ligand becomes able to associate with the receptor simultaneously with the already bound radioligand if the preincubation lasts for a long period, and to hinder radioligand dissociation. When the membranes were preincubated with [ 3 H]NMS and then exposed to benzilylcholine mustard (covalently binding specific li-gand), [ 3 H]NMS dissociation was blocked in wild-type receptors, but not in mutated (D99N) M 1 receptors. Co- valently binding [ 3 H]propylbenzilylcholine mustard detected substantially more binding sites than [ 3 H]NMS. The observations support a model in which the receptor binding domain has two tandemly arranged subsites for classical ligands, a peripheral one and a central one. Ligands bind to the peripheral subsite first (binding with lower affinity) We suggest that subscribers photocopy these corrections and insert the photocopies at the appropriate places where the article to be corrected originally appeared. Authors are urged to introduce these corrections into any reprints they distribute. Secondary (abstract) services are urged to carry notice of these corrections as prominently as they carried the original abstracts.

The agonists and competitive antagonists of muscarinic acetylcholine receptors associate with a binding domain that is located within a cavity formed by the seven transmembrane segments of receptor molecules; the cavity opens into the extracellular space (1,2). Most interactions between muscarinic receptors and their competitive antagonists can be explained in terms of simple bimolecular reactions, whereas the interactions between the receptors and the agonists are complicated by the associated receptor-G protein interactions (3)(4)(5)(6). With regard to some antagonists, however, observations have been described that are difficult to explain in terms of simple singlestep bimolecular reactions. Based on such observations, it has been proposed that the initial binding of the antagonist quinu-clidinyl benzilate (QNB) 1 is followed by a slow change in the conformation of the receptor molecule to a state that binds the ligand more avidly (receptor "isomerization"; Refs. 7-10) and that two "competitive" ligands may bind to the receptor simultaneously (10 -12).
The present study of the complexity of the binding of muscarinic antagonists was started within the context of our investigations of allosteric modulations of muscarinic receptors (13)(14)(15). We wanted to know how the dissociation of muscarinic antagonists N-[ 3

H]methylscopolamine ([ 3 H]NMS) and [ 3 H]QNB
from the receptors is affected by changes in the concentration of the unlabeled ligand added to the system in order to reveal ("induce") the dissociation, by the duration of the preceding radioligand association, and also by the treatment of receptors with a compound known to associate covalently with the classical binding site of muscarinic receptors (benzilylcholine mustard (BCM); see "Experimental Procedures") (16,17). The data we obtained provide strong support for a model in which there are two tandemly arranged binding sites on each muscarinic receptor molecule, a "peripheral" one and a "central" one. In this model, classical ligands bind first to the former and then move to the latter site. The movement of the ligand from the peripheral to the central site may explain some of the phenomena which had been previously interpreted as consequences of receptor isomerization. The results and interpretations have been published as a symposium abstract (18).  19 and 20). Some observations were duplicated on CHO cells stably transfected with the human gene for muscarinic M 2 receptors (hM 2 membranes; Refs. 21 and 22). Cells were grown in plastic dishes in Dulbecco's modified Eagle's medium with 10% fetal calf serum and 0.005% Geneticin (14). They were harvested 7 days after subcul-turing, washed twice through centrifugation (3 min at 300 ϫ g) and resuspension, and homogenized with an Ultra-Turrax homogenizer in a medium consisting of 136 mM NaCl, 5 mM KCl, 1 mM MgSO 4 , 1 mM sodium phosphate buffer (pH 7.4), and 10 mM Na-HEPES buffer (pH 7.4). The homogenate was centrifuged twice for 10 min at 600 ϫ g and combined supernatants were kept frozen at Ϫ20°C. On the day of experiment, membranes were sedimented by 15 min of centrifugation at 60,000 ϫ g, and washed twice by resuspension and recentrifugation (see Ref. 15

for details).
Experiments with Covalently Associating Muscarinic Ligands BCM and [ 3 H]PBCM-When BCM or PBCM are dissolved in water, their amino moieties give rise to aziridinium ions (the agents become "activated"), and this is followed by the formation of a covalent bond between the ligand and the receptor (17). BCM and [ 3 H]PBCM were kept at Ϫ20°C in a stock solution in ethanol and, before being added to the incubation tubes, they were diluted with water and kept at room temperature for 60 -90 min. Their interaction with receptors was stopped by adding 1 mM sodium thiosulfate. In experiments with radioligand association to BCM-pretreated membranes, the products of the reaction between BCM and sodium thiosulfate were removed by washing the membranes with centrifugation, but they remained in the incubation medium in experiments investigating the effect of BCM-pretreatment on radioligand dissociation. Details have been included in the descriptions of individual types of experiments.
Radioligand Binding Experiments-Measurements of radioligand binding were performed essentially as described (15,22,23). Membranes corresponding to 600,000 cells were incubated at 25°C in a final incubation volume of 0.8 ml. The composition of the incubation medium corresponded to that of the homogenization medium (see above), with added ligands as indicated for individual experiments. Atropine ( Treatment of Data-Our observations led us to conclude that there are two binding sites for "classical" antagonists and agonists on each muscarinic receptor: a peripheral one to which the ligands bind first, and a central one to which they move subsequently. Corresponding "two-site" model is proposed under "Discussion," together with a set of differential equations defining its kinetics. With the use of the KINSIM (24) and FITSIM (25) program for chemical kinetics, these equations have been collectively fitted to data from experiments shown in Figs. 1-7, and the best fit kinetic constants obtained have been summarized in Table II. Individual data points in Figs. 1-7 are means of three experiments performed with incubations in triplicate. The curves in these figures have been drawn by computer using the proposed model and the best fit kinetic constants from Table II.  Table I for numerical values at selected data points). A similar relationship between the rate of dissociation of [ 3 H]QNB from rM 1 membranes and the concentration of unlabeled QNB was observed ( Fig. 3A). Sixty min after the start of dissociation, 38% and 16% of originally bound [ 3 H]QNB dissociated in the presence of 1 M and 1 mM QNB, respectively. It is evident that the rate of dissociation of labeled ligands is inversely related to the concentration of the unlabeled ligands applied to reveal dissociation.

Dissociation of [ 3 H]NMS and [ 3 H]QNB from M 1 Muscarinic Receptors in the Presence of Different Concentrations of Unla
Stippled curves in the left parts of Figs. 1 and 3 have been computed as best fits for an assumed monoexponential time course of dissociation. Solid lines correspond to the two-site model described under "Discussion" and to the kinetic constants summarized in Table II. It is apparent that the fit between individual data points and the computed curves is much better for the two-site model. This is confirmed by the plots of residuals shown in the right halves of Figs. 1 and 3. Fig. 2A, 66% of originally bound [ 3 H]NMS dissociated from the rM 1 -D99N membranes 10 min after the start of dissociation in the presence of 1 M NMS and 49% in the presence of 1 mM NMS. This compares to 62% and 35%, respectively, in the case of membranes with wild-type rM 1 receptors (Fig. 1A). In experiments with [ 3 H]QNB, 83% of the originally bound radioligand dissociated from rM 1 -D99N membranes 60 min after the addition of 1 M QNB and 75% after the addition of 1 mM QNB (Fig. 3B); these values contrast with those observed in the case of membranes with wild-type rM 1 receptors (38% and 16%, respectively; Fig. 3A). Apparently, the deceleration of dissociation produced by high concentrations of unlabeled ligands (NMS or QNB) is smaller at mutated D99N receptors than at receptors of the wild type. In addition, the D99N mutation brings about a general increase in the rate of  (Fig. 7A) was clearly different for the one-site and the two-site models, and the actual observations corresponded to what had been predicted by the two-site model. This is best seen from the plots of residuals in the right half of Fig. 7A.

Association of [ 3 H]PBCM with rM 1 and rM 1 -D99N Receptors-[ 3 H]
PBCM is known to covalently associate with the classical binding sites of muscarinic receptors with high specificity and in an atropine-sensitive manner (19,26,27). Fig. 8A Fig. 10, rM 1 membranes (Fig. 10A), rM 1 -D99N membranes (Fig. 10B), and hM 2 membranes (Fig. 10C)   Values of rate constants for radioligand association and dissociation and of equilibrium dissociation constants obtained by fitting the set of differential equations described under "Tandem Two-site Model" of "Discussion" to experimental data See Fig. 11 for explanation of the meaning of individual constants. K  1) After 60 or 90 min of preincubation, the rate of [ 3 H]NMS or [ 3 H]QNB dissociation from receptors was slowed down by increases in the concentration of unlabeled NMS or QNB, applied to reveal dissociation. A similar phenomenon has already been noted in work with atropine and other competitive antagonists (28 -30), without a definite explanation. It seems apparent that unlabeled NMS or QNB bind to receptors simultaneously with [ 3

H]NMS or [ 3 H]QNB (thus creating ternary NMS-receptor-[ 3 H]NMS or QNB-receptor-[ 3 H]QNB complexes and slowing down dissociation), and yet that they prevent [ 3 H]NMS and [ 3 H]QNB reassociation.
2) If the prelabeling of the receptors had been shortened from 60 or 90 min to just 2 or 5 min, the dissociation of [ 3 H]NMS or [ 3 H]QNB was much faster and was not slowed down by higher concentrations of unlabeled NMS or QNB. It seems apparent that, during the short preincubation, the labeled ligand binds to a low affinity site (hence the fast dissociation rate after a short preincubation), while it translocates to a high affinity site during the long preincubation (hence the slow dissociation rate during a long preincubation). The unlabeled NMS or QNB have no effect on the rates of [ 3 H]NMS or [ 3 H]QNB dissociation after short preincubations because they cannot bind to the site that enables them to hinder the dissociation. Most probably, they do so when they sit on the low affinity site and prevent the departure of the labeled ligands from the high affinity site. However, the high affinity sites only become occupied by [  All of these observations are easily explained on the generalizing assumption that there are two tandemly arranged binding sites for classical antagonists on each receptor molecule, a peripheral and a central one, and that the antagonists first bind to the peripheral site, from which they subsequently translocate to the central site. While the receptor-antagonist association is weak as long as the antagonist is attached to the peripheral site (unless a covalent bond is formed, as in the case of BCM and PBCM), it becomes strong after the antagonist moves to the central site. During the procedure of filtration and washing, the radioligand is more easily lost from the peripheral sites (i.e. a proportion of the non-covalent binding to them remains undetected), while the loss from the central sites is smaller. Moreover, if the peripheral sites become occupied by high concentrations of the unlabeled antagonist (added to the medium to induce dissociation of the radioligand), the rate at which the radioligand is able to leave the central sites is further slowed down.
Tandem Two-site Model-A system with the features just outlined has been schematically depicted in Fig. 11. In this Tandem Two-site Arrangement of Muscarinic Binding Domains system, antagonist (A) binding may be described by the following set of differential equations: The total amount of bound antagonist (Y) corresponds to the sum of We used the KINSIM and FITSIM program and fitted these differential equations to data obtained in experiments with [ 3 H]NMS and [ 3 H]QNB association to and dissociation from rM 1 and rM 1 -D99N binding sites. In view of the complexity of the system, it was not possible to obtain separate values for k ϩ3 and k Ϫ3 (describing the association and dissociation of  [ 3 H]NMS or [ 3 H]QNB to and from the peripheral site of a receptor whose central site is already occupied by [ 3 H]NMS or [ 3 H]QNB, respectively). Instead, we could only obtain best fit values of the k Ϫ3 /k ϩ3 ratio, which we call K III in Table I and which describes the affinity for [ 3 H]NMS or [ 3 H]QNB of the peripheral site of a receptor whose central site had already been occupied by the same ligand. If the presence of the ligand at the central site had no effect on the affinity of the peripheral site for the same ligand, K III should be equal to K I . The K III values were consistently higher than the K I values, however, which indicates that the binding of [ 3 H]NMS or [ 3 H]QNB to the central site diminishes the binding of the same ligand to the peripheral site (a negative cooperative effect). The extent of negative cooperativity has been expressed as factor ␣ (31), taking that ␣ ϭ K III /K I , but its nature has not been clarified. We speculate that it depends on direct (steric and coulombic) interaction between the two ligand molecules.
The best-fit values of individual parameters have been listed in Table II It is generally believed that the aspartate residue in position 105 (Asp-105; within the third transmembrane segment) of muscarinic M 1 receptors (and in homologous positions of the other muscarinic receptor subtypes) plays a key role in the binding of positively charged orthosteric muscarinic ligands to receptors (32)(33)(34)(35)(36). In all muscarinic receptor subtypes, there is another Asp present at the extracellular end of the third trans-  in the interactions between the allosteric and orthosteric binding sites (20), but the question of topographical relations between the binding site for allosteric modulators and the peripheral part of the tandem orthosteric site requires special investigation.
Biphasic time courses of [ 3 H]PBCM binding in Fig. 8 (A and  B) raise the question of what determines the rate at which the irreversible ligand associates with the supposed peripheral and central parts of the orthosteric tandem site. We speculate that the association is rapid if it occurs to an empty receptor, or if it occurs at the peripheral site of a receptor the central site of which had already ("beforehand") been occupied. In Fig. 8A Fig. 8B; this may reflect the fact that the affinity of the peripheral sites for NMS is low in receptors in which the central sites had already been occupied by another ligand.
The tandem two-site model assumes that the ligand moves with regard to the receptor after the first association occurred. In a different context, a similar view has already been proposed by Saunders and Freedman (38), who suggested that muscarinic agonists first contact Asp-105 of the M 1 receptors and then translocate to Asp-71 (in the middle of the second transmembrane segment). Acetylcholine is believed to undergo postbinding translocation between the peripheral and the central binding sites of acetylcholinesterase (39). The tandem two-site model provides explanation for most of the phenomena that had been interpreted in terms of receptor "isomerization" (see Introduction); it is indeed likely that ligand translocation is accompanied by some change in receptor conformation.
In an earlier study, Kurtenbach et al. (32) tried to distinguish whether [ 3 H]PBCM associates with Asp-99 or Asp-105 of the M 1 receptors and concluded that the association is mainly with Asp-105, but they (see also Ref. 40 by Curtis et al.) could not exclude that some association with Asp-99 does occur. On the other hand, Fraser et al. (19) observed that the binding of [ 3 H]PBCM to M 1 receptors was strongly diminished after Asp-99 had been mutated to Asn, perhaps reflecting the loss of the peripheral binding site, which we assume to be responsible for the decrease of [ 3 H]PBCM binding observed in the present work (Fig. 8).
The tandem two-site model agrees with earlier kinetic observations indicating that the initial complex formed during [ 3 H]NMS association with the receptor has a low affinity, and that the affinity with which the ligand is bound increases with time (41). An interesting aspect of the tandem two-site arrangement is that it may explain some of the differences in the results of measurements of the numbers of muscarinic binding sites encountered in literature. In particular, higher numbers of muscarinic binding sites have been discovered frequently when [ 3 H]QNB rather than [ 3 H]NMS was applied for receptor labeling (42)(43)(44). Differences in results seem to depend on differences in the speed with which the ligand-receptor association becomes stabilized by ligand translocation from the peripheral to the central binding domain, and in the ease with which double occupancy of receptors by different ligands occurs.

Tandem Two-site Arrangement of Muscarinic Binding Domains
Some of our observations could perhaps be interpreted on an alternative assumption, namely that the receptors work as dimers or oligomers (44,45) with asymmetric properties of their binding sites, and that dimerization is impaired in D99N receptors. It would be difficult to explain, however, why the dissociation of [ 3 H]NMS and [ 3 H]QNB is fast after short preincubations and slow after long preincubations, and why the dissociation of [ 3 H]NMS is blocked by an after-treatment with BCM.
In conclusion, the assumption that the binding of muscarinic ligands to the classical binding site of the M 1 and M 2 muscarinic receptor subtypes is a dynamic process involving ligand translocation between two tandemly arranged binding domains provides excellent explanation of several phenomena that are difficult to interpret otherwise, and deserves further investigation.