Positive and negative allosteric modulators of the Ca 2+ -sensing receptor interact within overlapping but not identical binding sites in the transmembrane domain

A three-dimensional model of the human extracellular Ca 2+ -sensing receptor (CaSR) has been used to identify specific residues implicated in the recognition of two negative allosteric CaSR modulators of different chemical structure, NPS 2143 and Calhex 231. To demonstrate the involvement of these residues, we have analyzed dose-inhibition response curves for the effect of these calcilytics on Ca 2+ -induced [ 3 H]inositol phosphates accumulation for the selected CaSR mutants transiently expressed in HEK293 cells. These mutants were further used for investigating the binding pocket of two chemically unrelated positive allosteric stimulates (EC 50 = 0.31 µM) increases in [ 3 H]inositol phosphate levels elicited by activating the wild-type CaSR by 2 mM Ca 2+ . Our data validate the involvement of W818 F821 E837 7.39 and I841 7.43 located in TM6 and TM7, in the binding pocket for both calcimimetics and calcilytics, despite important differences observed between each family of compounds. The transmembrane domains (TMs) involved in the recognition of both calcilytics include residues located in TM3 (R680 3.28 , F684 3.32 , F688 3.36 ). However, our study indicates subtle differences between the binding of these two compounds. Importantly, the observation that some mutations which have no effect on calcimimetics recognition but which affect the binding of calcilytics in TM3 and TM5, suggests that the binding pocket of positive and negative allosteric modulators is partially overlapping but not identical. Our CaSR model should facilitate the development of novel drugs of this important therapeutic target and the identification of the molecular


INTRODUCTION
Cloning the extracellular Ca 2+ -sensing receptor (CaSR) from bovine parathyroid has shed light on the molecular mechanisms implicated in the regulation of parathyroid hormone (PTH) secretion (1). The CaSR's role in calcitonin secretion in the thyroid and ion maintenance in the kidney has also been investigated and its presence has been detected in various tissues such as the intestine, the lung as well as in bone (2) where it has been proposed as a molecular target for the actions of strontium ranelate, a candidate drug for the treatment of female osteoporosis (3,4). Its expression in oligodendrocyte cells during development (5,6) and its presence on nerve terminals suggest additional roles for this receptor (7,8).
The CaSR belongs to family 3 of G-protein coupled receptors (GPCRs) characterized by a long bilobed aminoterminal tail proposed to contain the ligand binding sites. This family includes metabotropic glutamate receptors (mGluRs), GABA B , taste and putative pheromone receptors (9,10). Positive allosteric modulators of the CaSR, also called calcimimetics, have been characterized (11)(12)(13)(14). One of these, NPS R-568, a phenylalkylamine ( Fig. 1), has been proposed to selectively activate the parathyroid CaSR resulting in an inhibition of parathyroid hormone (PTH) secretion both in vitro and in vivo (15). The therapeutic potential of NPS R-568 and of its derivatives has been demonstrated in patients with primary hyperparathyroidism as well as in patients exhibiting secondary hyperparathyroidism linked to renal disease. NPS R-568 has been shown to interact with residues located in the extracellular region delimited by the seven transmembrane domains (TMs) (16)(17)(18)(19).
The negative allosteric modulator NPS 2143 (Fig. 1), was recently introduced as the first negative allosteric modulator of the CaSR. This compound, also called a calcilytic, inhibits the biological effects elicited by Ca 2+ or by a calcimimetic acting on the CaSR but does not affect the responses elicited by the activation of several other GPCRs. When applied to bovine parathyroid cells in culture, NPS 2143 stimulated the secretion of PTH (20). After in vivo injection of this molecule, a rapid and sustained increase of plasma PTH was observed in the rat, and long term treatment of ovariectomized rats, an animal model of osteoporosis, was followed by a large increase in bone turnover (20,21). These data suggest that NPS 2143, and calcilytics in general, might be useful for regulating plasma PTH level, thereby representing an interesting pharmacological target for drug development. The identification and characterization of calcimimetics and calcilytics as well as their associated molecular mechanisms of action are therefore an important goals.
In this report, we have identified the binding site(s) of NPS 2143 in the human CaSR using a three dimensional model of this receptor based on the x-ray structure of bovine rhodopsin (22) that we have recently reported (23). Comparison of the ligand binding pocket of both NPS 2143 and Calhex 231 (Fig 1), a structurally different negative allosteric modulator of the CaSR that we have recently characterized (23), led us to demonstrate that these two calcilytics interact with overlapping binding sites in the TMs. Moreover, we report the calcimimetic properties of (R)-2-[1-(1-naphthyl)ethylaminomethyl]-1H-indole (Calindol) ( Fig. 1), which also belongs to a novel structurally different series of calcimimetics (24). We have furthermore examined if the amino acids involved in the recognition of the calcilytics Calhex 231 and NPS 2143 are also implicated in the binding of the calcimimetics Calindol and NPS R-568.
Our data suggest that calcilytics and calcimimetics interact with several identical residues only within the sixth and seventh TMs. These studies further validate our CaSR model based on the crystal structure of bovine rhodopsin and provide a rational framework for the development of more selective and potent allosteric modulators of the CaSR.

Automated docking of NPS 2143
The Surflex docking program (27) was used to automatically dock NPS 2143 to the previously described three dimensional model of the CaSR (23). An idealized active site ligand or protomol (28) was first generated from 33 consensus positions (29) supposed to map the TM cavity of most GPCRs. This protomol consists of the preferred locations of various molecular probes (CH 4 , C=O, N-H) that are then used by the docking engine to search for the best three dimensional morphological similarity between the protomol and the ligand to dock.
A proto_thresh value of 0.5 and a proto_bloat value of 0 were used to generate a compact protomol. A TRIPOS mol2file (TRIPOS Assoc. Inc., St Louis, MO) of the antagonist, obtained from a two dimensional sketch as previously reported (29) was docked into the TM cavity using standard parameters of Surflex used in the "whole" docking approach (27). The best 30 solutions were finally stored in mol2 format.

Hypothesized binding mode of NPS 2143
We have recently developed a model of the human CaSR based on the crystal structure of bovine rhodopsin which has allowed us to delineate the putative TMs of the CaSR ((23) and Fig. 2). We then used this model to dock the novel calcilytic Calhex 231 (Fig. 1) into a hydrophobic cavity centered on Glu837 and formed by two pockets (23). This observation prompted us to use the same model to analyze the potential sites of interaction of NPS 2143, a structurally different calcilytic (Fig. 1). Automated docking of NPS 2143 using the recently-

Functional analysis of CaSR mutants for NPS 2143 inhibition of Ca 2+ -promoted increases of IP response
We first determined the potency of NPS 2143 in inhibiting the increase of IP response induced by 10 mM Ca 2+ in HEK293 cells transiently expressing the human WT CaSR.
Incubation of these cells with NPS 2143 caused a concentration-dependent inhibition of the IP response to 10 mM Ca 2+ (Figs. 4A, B). Analysis of the dose-response curve led to an IC 50 for NPS 2143 of 0.35 ± 0.08 µM (mean ± S.E.M., n = 10) which is comparable to that of Calhex 231 (IC 50 = 0.39 ± 0.08 µM, The dose-response curves of NPS 2143 for the three mutants F684A, F688A located in TM3, and E837A located in TM7, respectively, were profoundly affected as shown in Figs. 4 A and B. Thus, NPS 2143 lost its ability to block the Ca 2+ -induced IP response in CaSR having the point mutation F684A and F688A (<30% inhibition by 10 µM NPS 2143), as well as E837A (<20% inhibition by 10 µM NPS 2143). These data indicate that these residues play a key role in NPS 2143 recognition. Analysis of the dose-response of the R680A mutant revealed a ~12 fold increase of the IC 50 value for NPS 2143 thereby demonstrating that R680, which is located next to the two crucial phenylalanines F684 and F688 in TM3, also participates in the binding of NPS 2143 (Table 1). However, this mutation led to a significant decrease in the IC 50 of Calhex 231 in inhibiting the Ca 2+ -induced IP response (IC 50 = 0.12 ± 0.02 µM) compared to the WT receptor (IC 50 = 0.39 ± 0.08 µM, p < 0.05) ( Table 1). Analysis of the dose response of the I841A mutant indicated a ~12 fold increase of the IC 50 value for NPS 2143 which demonstrates that I841, located next to E837 in TM7, is also implicated in the binding of NPS 2143. The last mutations studied, T764A and H766A, corresponding to two residues located in the extracellular loop 2 which could potentially block access of compounds to the TM region, as well as L776A located in TM5, and W818A, F821A both located in TM6, did not significantly affect the ability of NPS 2143 to block the Ca 2+ -induced IP response in transfected cells ( Table 1).

Potency of Calindol
In a recent preliminary report, we described the synthesis and characterization of a novel series of molecules displaying calcimimetic properties toward the rat CaSR (24). We have now synthesized and investigated the potency of Calindol (Fig. 1), which belongs to this  Table 2). These results show that Calindol and NPS R-568 display similar pharmacological properties toward the human CaSR and that both compounds enhance the affinity of Ca 2+ for its receptor. These data also suggest that these molecules are allosteric modulators, and possibly interact at the level of the CaSR's TM. We therefore examined whether amino acid residues involved in the recognition of the calcilytics are also implicated in the recognition of these two calcimimetics.

Functional analysis of Calindol and NPS R-568 for stimulation of Ca 2+ -promoted increases of IP response in CaSR mutants
CaSRs harboring the F684A, F688A, L776A, W818A, F821A, E837A, I841A mutants were transiently transfected into HEK293 cells and dose-response curves for Ca 2+ ranging from 0.3 to 10 mM alone or in the presence of 1 µM of Calindol or of NPS R-568 were constructed (Fig. 6, Table 2 A substantial reduction in the maximal receptor response to Ca 2+ was observed for the I841A mutant in the presence of both calcimimetics (~ 40%) and a more modest reduction (~ 10-20%) was observed only in the presence of Calindol for the E837A and F684A mutants.
The W818A mutant displayed a marked increase of the maximal response to Ca 2+ (28-35%) in the presence of both calcimimetics ( Fig. 6 and Table 2).

DISCUSSION
Negative allosteric modulation has been reported for many GPCRs including mGluRs that are structurally related to the CaSR (9,10). However, despite the obvious therapeutic interest in identifying such molecules for the CaSR (15), only a limited number of structurally different positive and negative allosteric CaSR modulators has been described. To date, the phenylalkylamine NPS R-568 and its derivatives such as Cinacalcet (30) belong to the first family of calcimimetics to have been evaluated as candidate drugs. In this study, we have characterized a CaSR positive allosteric modulator, Calindol, which belongs to a novel chemically different family of molecules. Moreover, we have identified and compared the sites of interaction of both Calindol and NPS R-568 with the CaSR. As well, prior to this study, the binding sites of NPS 2143, the first and sole calcilytic whose pharmacokinetic properties had been reported in vitro and in vivo (20,21), were not known. We have now characterized several crucial residues involved in its recognition and compared its sites of interaction to those of Calhex 231, a negative allosteric modulator of the CaSR that we have recently described (23). This work has allowed us to identify the presence of a positive and a negative allosteric binding site located at the level of the transmembrane domains of the CaSR and to demonstrate that these sites are overlapping but not identical. Moreover, we show important differences in the binding of the two families of calcimimetics and calcilytics which should allow the development of compounds of higher selectivity and affinity.
We have recently synthesized Calhex 231 and characterized its antagonist properties toward the human CaSR and proposed a model of its allosteric interaction with the seven TM region using a series of CaSR mutants (23). One of the major findings of our present study concerns the identification of five residues Arg680 3.28 , Phe684 3.32 , Phe688 3.36 , Glu837 7.39 and Ile841 7.43 implicated in the recognition of both NPS 2143 and Calhex 231, two structurally different negative allosteric modulators. Moreover, our data allow delineation of ligand binding pockets for both molecules which are largely overlapping, and are located within the bundle formed by the seven TMs (Fig. 3B).
The proposed binding mode of NPS 2143 to the TM cavity of the CaSR exhibits similar features to that previously reported for Calhex 231 and clearly defines the seven TM region as the primary determinant for its sites of interaction with the receptor. Both compounds are primarily anchored through an H-bond assisted salt bridge to a negativelycharged amino acid (E837) located in TM7. Two adjacent hydrophobic pockets are used to locate the two aromatic groups of both compounds, with the bulkier naphthalene moiety anchored to the largest (pocket A) of the two subsites (Fig. 3B). Lastly, the gem-dimethyl moiety of NPS 2143 mainly overlaps with the cyclohexyl ring of Calhex 231 and faces the aromatic ring of F684. However, despite these similarities, significant differences are observed that may explain the herein described different affinity profile of both compounds for selected CaSR mutants (Table 1). NPS 2143 is proposed to directly H-bond to the Arg680 side chain as fully supported by the present study which unambiguously demonstrates that R680A mutation unfavorably affects NPS 2143 binding (Table 1). However, the latter amino acid mutation led to a receptor mutant with significantly enhanced Calhex 231 antagonist activity, which is difficult to predict from our model but that we have already observed for L776A and F821A mutants (23). Furthermore, the naphthalene moiety of Calhex 231 is very close to W818, a residue of TM6 demonstrated to be crucial for binding this compound (23), while the corresponding naphthalene ring of NPS 2143 is more oriented towards the TM3 residues of pocket A (F688, V689). Thus, the W818A mutation is much more detrimental to Calhex 231 binding than to that of NPS 2143. Conversely, NPS 2143 is in closer proximity than Calhex 231 to TM3 amino acids (F688, V689) explaining the completely abolished binding of NPS 2143 to the F688A CaSR mutant (Table 1 and Fig. 4A). Slight differences are also observed in the positioning of the other substituted phenyl ring of both compounds in pocket B (Fig. 3B) that may explain dissimilar loss of affinity resulting from the I841 mutation (Table 1). Interestingly, the negatively-charged E837 side chain presumably interacts with the protonated secondary amine present in both calcilytics which indicates that this residue plays a key role in calcilytic recognition (Fig. 3B). In agreement with our published model and current data (Fig. 3, 4 and Table 1 No attempt was made to model the CaSR in its activated form bound to Calindol or NPS R-568, because of the lack of an adequate 3-D template (33). However, our data emphasizes the crucial role of E837 in anchoring NPS R-568, as previously observed (17,19) and of Calindol, presumably through a salt bridge with the protonated secondary amine of the two compounds as we previously proposed for the calcilytics. Our study also underscores the role of I841 in anchoring the two calcimimetics as indicated by the lack of Ca 2+ potentiation of IP response by both compounds in the mutants. We were also able to demonstrate the noninvolvement of R680, F684, F688 located in TM3 and L776 located in TM5, previously implicated in calcilytics recognition, in anchoring both calcimimetics since we observed a marked left-shift of the dose response curve to Ca 2+ in presence of 1 µM Calindol or NPS R-568 (Table 2 and Fig. 6). A previous report has shown that a F684A mutation exhibits normal activation by Ca 2+ and a profound reduction in the maximal response to Ca 2+ , but attenuated responses to NPS R-568, both with respect to IP formation and to mobilization of intracellular Ca 2+ after transfection of the CaSR mutant in HEK293 cells (19). However, our data indicate that the F684A mutant displays a reduced affinity for Ca 2+ (EC 50 = 5.9 mM) as deduced from dose-response curve analysis of Ca 2+ -induced IP formation as well as a decrease in the maximal response to Ca 2+ (Table 2 and (23)), and shows a left-shift of the IP response in presence of both calcimimetics (EC 50 for Ca 2+ = 5.9 mM, and EC 50 for Ca 2+ + 1 µM Calindol or NPS R-568 = 2.5-2.9 mM) ( Table 2 and Fig. 6). It should be noted that the apparent affinity of NPS R-568 and also of Calindol, varies with the concentration of extracellular Ca 2+ and therefore is directly linked to the EC 50 of Ca 2+ for the receptor mutant. The discrepancies between the two studies of R680A might be attributable to differences in experimental conditions and EC 50 determination.
The current data then suggest that some common features previously reported to participate in the activation of class 1 rhodopsin-like receptors might be conserved in the molecular activation of class 3 GPCRs. Hence, the mutation of two residues (Trp818, Phe821), rather conserved among GPCRs and known to lock class 1 GPCRs in an inactive ground state (32), affects the ability of both calcimimetics to potentiate Ca 2+ binding. Whereas NPS R-568 is proposed to mainly interact with Phe821, Calindol interacts only with Trp818.
Thus, one of these two key residues could be free for inducing, upon binding of one of the latter two compounds, the conformational switch triggering receptor activation.