Selective Ligand-induced Stabilization of Active and Desensitized Parathyroid Hormone Type 1 Receptor Conformations*

For many G protein-coupled receptors, agonist-in- duced activation is followed by desensitization, internalization, and resensitization. In most cases, these processes are dependent upon interaction of agonist-occupied receptor with cytoplasmic (cid:1) -arrestins. The ligand-induced intramolecular rearrangements of the receptor responsible for the desensitized versus active conformational states, which dictate both the pharma- cological properties of ligands and the biological activity of G protein-coupled receptors, have not been fully elucidated. Here, we identify specific interactions between parathyroid hormone (PTH)-related protein and the human PTH type 1 receptor (PTH1Rc) and the related receptor conformational changes that lead to (cid:1) -ar- restin-2-mediated desensitization. PTH-related protein analogs modified at position 1 induced selective stabili- zation of the active G protein-coupled state of the receptor, resulting in lack of (cid:1) -arrestin-2 recruitment to the cell membrane, sustained cAMP signaling, and absence Mechanisti- the modified at interacting the of VI of a of and that from that the cognate agonist, re- sulting in significantly different conformations of the intracellular loop. These results show how specific interactions between PTH1Rc and its ligands may stabilize distinct conformational states, representing ei- ther the active G protein-coupled or a desensitized (cid:1) -ar-restin-coupled receptor state. In addition, they establish that sustained biological activity of PTH1Rc may be induced by appropriately designed agonist ligands. The simulations utilized a three-layer (water/decane/water) simulation cell, allowing for long simulations while maintaining the overall biphasic hydrophobic/hydro- philic character as well as the molecular motions of the long alkyl chains found in membranes (24). All simulations were performed using the GROMACS package (25). Analysis of motions of the transmembrane (TM) helices was carried out using the microdomain procedure described by Weinstein et al. (26), in which the receptor is defined as a set of domains. The rotation and translation of each domain during the simulation were then mapped upon the torques or forces exerted on the domain, clearly illustrating how interdomain interactions influence the motion of the domain. The structural consequences of the modified ligands, even if involving many sequential interactions, can be unam- biguously identified.

G protein-coupled receptors (GPCRs) 1 represent a major class of membrane-bound proteins that mediate a wide variety of biological functions, including sensitivity to light and odorants, endocrine and cardiovascular control, and neurotransmission. The ligand-induced intracellular signaling of GPCRs is tightly regulated by several mechanisms. For numerous GPCRs, desensitization involves translocation of arrestins from the cytosol to the cell membrane, their direct interaction with agonist-activated GPCRs, and consequent inhibition of G protein coupling (1,2). Additionally, arrestin-mediated internalization of agonist-receptor complexes through clathrincoated vesicles and subsequent receptor recycling to the cell membrane are responsible for the recovery of cellular responsiveness to agonists (resensitization). Receptor desensitization and resensitization accomplish the fundamental physiological role of modulating the cellular responses to both acute and chronic stimulation. An important implication is that not only signal transduction per se, but also the mechanisms regulating signal transduction have a profound influence in determining the pathophysiological processes mediated by GPCRs. Many important questions remain regarding the relationship between receptor occupancy, signaling, and desensitization. In particular, the structural features that differentiate the active and desensitized states and how extracellular interactions with the ligand translate into intracellular events (such as G protein activation and interaction with arrestins) have not been fully characterized.
Parathyroid hormone (PTH) is a major regulator of serum calcium homeostasis and bone metabolism (3). PTH-related protein (PTHrP), first described as the hormone responsible for hypercalcemia of malignancy, is now recognized as an autocrine/paracrine factor with various biological functions in many tissues (4). Both PTH and PTHrP bind to and activate the PTH type 1 receptor (PTH1Rc), a member of the class II subfamily of GPCRs coupled to both G s and G q proteins (5). As for many other GPCRs, interaction of activated PTH1Rc with ␤-arrestin-2 is a primary mechanism for rapid desensitization of cAMP signaling and further internalization of ligand-receptor complexes (6,7). Several lines of evidence suggest that recruitment of ␤-arrestin-2 to the cell membrane and its association with activated PTH1Rc are largely (if not completely) independent of intracellular signaling and receptor phosphorylation (7)(8)(9). Therefore, the increased affinity of agonist-occupied PTH1Rc for ␤-arrestin-2 and the consequent reduction in coupling with G proteins may result from specific conformational states distinct from those of the active G protein-bound receptor. To characterize the specific interactions with agonists that stabilize these distinct functional states, we developed analogs of the biologically active N-terminal fragment of PTHrP (i.e. PTHrP-(1-36)) with selective pharmacological profiles. Analysis of the signaling properties of these analogs combined with fluorescence monitoring of the cellular distribution and extensive molecular modeling of the ligand-receptor complexes allowed us to identify specific interactions between ligand and receptor and the related conformational changes within PTH1Rc that are associated with ␤-arrestin-2-mediated desensitization.  -34)) were carried out as previously described (6). The pure products were characterized by analytical HPLC, electron spray mass spectrometry, and amino acid analysis. Radioiodination and HPLC purification of PTH-(1-34), PTHrP- , and Bpa 1 -PTHrP-(1-36) were carried out as reported (6).
Cell Culture and Transfection Method-Human embryonic kidney cells (HEK-293) and HEK-293 cells stably expressing human PTH1Rc (clone C-21) were cultured in Dulbecco's minimal essential medium supplemented with 10% fetal bovine serum as described (10). Cells were plated at 1.5 ϫ 10 5 cells/coverslip on 35-mm glass coverslips for fluorescence microscopy experiments and at 1.0 ϫ 10 5 cells/well in 24-well plastic dishes (Corning Inc., Corning, NY) for adenylyl cyclase, radioligand binding, and radioligand internalization assays and transfected as previously described (7). All subsequent experiments were performed 24 h after transfection.
Radioreceptor Binding and Internalization Assays-Radioreceptor binding and internalization assays were carried out as reported (7) using HPLC-purified radioiodinated compounds. The binding affinities (K d ) of each PTHrP analog were calculated from Scatchard analysis of radioreceptor binding assays (using each radioiodinated compound as tracer and competition with each unlabeled analog) as described previously (6,7).
Adenylyl Cyclase, Intracellular Calcium, and Inositol Phosphate Assays-cAMP accumulation was determined in subconfluent cell cultures in the presence of the phosphodiesterase inhibitor 3-isobutyl-1methylxanthine as previously described (7). Briefly, ligand-stimulated cAMP accumulation was measured in cells preincubated for 30 min with 3-isobutyl-1-methylxanthine (1 mM) and subsequently exposed to various concentrations of the appropriate ligand (10 Ϫ11 to 10 Ϫ6 M). To determine desensitization of cAMP signaling, cells were treated with the indicated ligands (100 nM) for 30 min, followed by a 2-h washout at 37°C. Residual cAMP accumulation was then measured for 15 min in the presence of 3-isobutyl-1-methylxanthine (1 mM). In all cases, reactions were stopped with 1.2 M trichloroacetic acid, and cAMP was isolated by the two-column chromatographic method (11). Transients of intracellular calcium stimulated by PTHrP-(1-36), PTHrP-(2-36), and Bpa 1 -PTHrP-  were assessed spectroscopically in Fura-2-loaded HEK-293 cells stably transfected with human PTH1Rc (clone C-21) as described (12).
Fluorescence Microscopy-Cellular distribution of fluorescent ligands, receptor-green fluorescent protein (GFP) conjugate (PTH1Rc-GFP), and ␤-arrestin-2-GFP (␤-arr2-GFP) was assessed by real-time fluorescence microscopy as previously described (7). Briefly, transfected cells grown on glass coverslips were rinsed with phosphate-buffered saline and mounted in an open-air, temperature-controlled block on a Nikon Diaphot 300 ® epifluorescence microscope. ␤-arr2-GFP or PTH1Rc-GFP distribution was first analyzed in the absence of ligand by monitoring the cells maintained in phosphate-buffered saline and 0.1% bovine serum albumin at 37°C for up to 30 min. The effects of ligands on ␤-arr2-GFP or PTH1Rc-GFP localization were evaluated by subsequently adding the various ligands (100 -1000 nM) in the same buffer and by monitoring redistribution of the green fluorescence over time. Dual fluorescence microscopy to colocalize ␤-arr2-GFP or PTH1Rc-GFP and rhodamine-labeled ligands was performed by preincubating cells with ice-cold phosphate-buffered saline and 1% bovine serum albumin (blocking of nonspecific binding), followed by incubation with the fluorescent ligand for 10 min on ice or at room temperature. The unbound ligand was then removed by carefully rinsing the coverslips with phosphate-buffered saline, and the cells were warmed to 37°C for microscopy monitoring. Images of the distribution of ␤-arr2-GFP or PTH1Rc-GFP and rhodamine-labeled ligands were acquired sequentially, i.e. within a 10-s interval and in the same cellular plan, using fluorescein and rhodamine filters, respectively. Overlay images were generated using Image-Pro Plus ® software (Media Cybernetics, Silver Spring, MD).
Molecular Modeling-The molecular dynamics simulations of the ligand-receptor complexes were carried out following published procedures (14,15). The starting structure, based on extensive NMR characterization of the ligands and extra-and intracellular portions of PTH1Rc, also includes ligand-receptor contact points derived from photoaffinity labeling experiments (16 -23). The simulations utilized a three-layer (water/decane/water) simulation cell, allowing for long simulations while maintaining the overall biphasic hydrophobic/hydrophilic character as well as the molecular motions of the long alkyl chains found in membranes (24). All simulations were performed using the GROMACS package (25). Analysis of motions of the transmembrane (TM) helices was carried out using the microdomain procedure described by Weinstein et al. (26), in which the receptor is defined as a set of domains. The rotation and translation of each domain during the simulation were then mapped upon the torques or forces exerted on the domain, clearly illustrating how interdomain interactions influence the motion of the domain. The structural consequences of the modified ligands, even if involving many sequential interactions, can be unambiguously identified.
PTH1Rc Desensitization and Resensitization of cAMP Signaling-Because receptor interaction with ␤-arrestin-2 is an essential early step in rapid desensitization of G s /cAMP signaling by PTH1Rc (7), we hypothesized that Bpa 1 -PTHrP-  and PTHrP-(2-36) would induce sustained activation of adenylyl cyclase. Indeed, exposure of HEK cells expressing PTH1Rc either stably (C-21) or transiently to Bpa 1 -PTHrP-(1-36) or PTHrP-(2-36) for 30 min, followed by a 2-h washout, resulted in a 5-7-fold increase in residual levels of cAMP accumulation at the end of this period compared with cells similarly exposed to PTHrP-(1-36) or PTH-(1-34) (Fig. 3B and Table I). Moreover, cells pretreated with PTHrP-(1-36) and PTH-(1-34) were fully responsive to restimulation with these agonists after 2 h (ϩ883 and ϩ796% above residual levels, respectively), indicat-ing that complete receptor resensitization had occurred ( Fig.  3B and Table I). In cells pretreated with Bpa 1 -PTHrP-(1-36) or PTHrP- , restimulation with the same ligands resulted in small increases in cAMP accumulation above the residual levels (13.5 and 36%, respectively) ( Fig. 3B and Table I) (6, 7). b EC 50 was calculated from dose-response curves for cAMP accumulation. Cells were incubated with various concentrations of ligands (10 Ϫ11 to 10 Ϫ6 M) for 15 min at 37°C in the presence of 1 mM 3-isobutyl-1-methylxanthine. cAMP was measured as previously described (6,7). c Maximal acute cAMP accumulation was measured after incubation with 100 nM ligands under the conditions described above. The basal cAMP level was 650 Ϯ 37 cpm/well. d To evaluate residual cAMP accumulation, cells were incubated for 30 min with 100 nM ligands, followed by three washes with phosphatebuffered saline and a 2-h incubation at 37°C in cell culture medium containing [ 3 H]adenine. At the end of this period, 1 mM 3-isobutyl-1methylxanthine was added for 30 min, and cAMP was measured as previously described (6,7). The basal cAMP level was 650 Ϯ 37 cpm/well. e To evaluate maximal cAMP accumulation after re-challenge with the agonist, cells previously incubated with the ligand, washed, and incubated for 2 h, as described above, were restimulated for 15 min with the agonist in the presence of 1 mM 3-isobutyl-1-methylxanthine. The basal cAMP level was 650 Ϯ 37 cpm/well. f Intracellular calcium transients were measured spectroscopically in Fura-2-loaded cells stimulated with 100 nM ligands as reported previously (10). ND, nondetectable generation of calcium transients. differences between PTHrP and its analogs modified at position 1 were dependent on differences in ligand-induced receptor conformation or G q -mediated signaling, we investigated whether regulation of cAMP signaling and cellular distribution of Bpa 1 -PTHrP-(1-36) with ␤-arr2-GFP could be rescued by the PTH1Rc(H223R) mutant. Indeed, this natural mutant is G q signaling-deficient both in COS-7 cells (31) and in HEK-293 cells, where no measurable generation of inositol phosphates in response to PTHrP-(1-34), Bpa 1 -PTHrP-(1-36), or PTHrP-(2-36) was detected (Fig. 3C). Furthermore, in HEK-293 cells, PTH1Rc(H223R) is constitutively associated with ␤-arrestin-2 on the cell membrane and internalizes normally in response to PTH-(1-34) and PTHrP-(1-34) (7). In this system, brief exposure to Bpa 1 -PTHrP-(1-36) and PTHrP-(2-36), followed by a 2-h washout, did not result in cAMP levels higher than the basal activity, which is reflective of the constitutive activity of the receptor mutant (Fig. 3B). Moreover, 2 h after the first agonist stimulation, PTH1Rc(H223R)-expressing cells were fully responsive to a re-challenge with agonist (Fig. 3B), indicating that the elevated constitutive activity in these cells is not a limiting factor for the ligand-induced generation of cAMP. This was accompanied by internalization of 125 I-Bpa 1 -PTHrP-(1-36) and 125 I-PTHrP-(2-36) bound to PTH1Rc(H223R) (68 Ϯ 3 and 54 Ϯ 3%, respectively), which was similar to 78 Ϯ 2% for 125 I-PTH-(1-34). Additionally, rhodamine-labeled Bpa 1 -PTHrP-(1-36) colocalized with PTH1Rc(H223R)-associated ␤-arr2-GFP on the cell membrane and then intracellularly (Fig.  3D), as was previously seen with rhodamine-labeled PTH-(1-34) (7). These experiments show that the absence of activation of the G q /phospholipase C signaling pathway is not responsible for lack of rapid cAMP desensitization of wild-type PTH1Rc in response to PTHrP analogs. Rather, these data indicate that interaction of the receptor with ␤-arrestin-2 is sufficient for the regulation of G s /cAMP signaling.
Molecular Modeling of Ligand-PTH1Rc Complexes-To further explore the structural details underlying these effects, we examined the conformational changes induced by binding of native PTHrP-(1-36) and its position 1-modified analogs to human PTH1Rc using molecular dynamics simulations. The starting ligand-receptor structure, based on extensive NMR characterization of the ligands and extra-and intracellular portions of PTH1Rc, also includes ligand-receptor contact points identified by photoaffinity labeling experiments (16 -23). Among these, of particular significance is the interaction of position 1 of Bpa 1 -PTHrP-  with Met 425 at the extracellular end of TM6 of PTH1Rc (16).
Analysis of simulations with PTHrP-(2-36) and Bpa 1 -PTHrP-(1-36) indicated significant differences from simulations with the native hormone, PTHrP-(1-36) (Fig. 4), in particular with regard to the orientation of TM5 and TM6. The root-mean-square deviation for TM5-TM6 was 3.5 Å between the wild-type ligand and Bpa 1 -PTHrP- . A much smaller value of 1.5 Å was obtained for the remaining helices (TM1-TM4 and TM7). This resulted in major effects on the structure of the third intracellular loop (IC3), particularly the N-terminal helix (Thr 387 -Arg 396 ) of the intracellular loop. With PTHrP- , this helix was found to be parallel to the membrane surface, leading to the exposure of the charged residues, Lys 388 , Arg 390 , and Glu 391 . In contrast, with the G s -selective modified ligands, the helix-turn-helix structure of the third loop, as observed in our NMR studies (23,32), was maintained. In all of the simulations, smaller effects on the orientation of the C-terminal portion of IC3 were observed. These changes, mostly evident with Lys 405 , were identical to those observed for PTH1Rc mutants that constitutively activate adenylyl cyclase (14). The results for PTHrP-(2-36) were very similar to those for Bpa 1 -PTHrP- , suggesting that the first residue of PTHrP is actively involved in the induction of the receptor conformation recognized by ␤-arrestin-2.

DISCUSSION
The objective of this study was to characterize the structural and functional consequences of modifications affecting the first amino acid residue in PTHrP and thereby to provide new insights into the PTH1R conformational states associated with coupling to G proteins and desensitization of intracellular signaling. Our data clearly show that modifications of residue 1 in the biologically active N-terminal fragment of PTHrP result in ligands (i.e. PTHrP-  and Bpa 1 -PTHrP-(1-36)) that, although maintaining high affinity and efficacy in stimulating cAMP production, are unable to stimulate the G q /phospholipase C signaling pathway. Additionally, in contrast to native PTHrP and PTH (6 -9), the modified PTHrP analogs do not stimulate ␤-arrestin-2 translocation to the cell membrane and  (1-34)). B, sustained adenylyl cyclase activity by position 1-modified PTHrP analogs is inhibited by the PTH1Rc(H223R) mutant. HEK-293 cells transiently expressing either wild-type PTH1Rc (white and thinly hatched bars) or the constitutively active PTH1Rc(H223R) mutant (black and thickly hatched bars) (0.4 g of DNA/well in 24-well plates) were treated with the indicated ligands (100 nM) for 30 min, followed by a 2-h washout at 37°C. Residual and re-challenged cAMP accumulation was then measured for 30 min in the presence of 3-isobutyl-1-methylxanthine (1 mM) without (white and black bars) and with re-exposure (hatched bars) to PTHrP or its modified analogs, respectively. In cells expressing wild-type PTH1Rc, residual cAMP accumulation after preincubation with Bpa 1 -PTHrP-(1-36) and PTHrP-(2-36) was significantly higher than after preincubation with PTHrP- . Residual cAMP accumulation after preincubation with Bpa 1 -PTHrP-(1-36) and PTHrP-(2-36) was lower (p Ͻ 0.001) in cells expressing PTH1Rc(H223R) than in cells expressing wild-type PTH1Rc. Also, in cells expressing PTH1Rc(H223R), the response to a re-challenge with the agonist (p Ͻ 0.01 compared with residual cAMP levels) was similar to that in cells not previously treated with the ligand. The maximal cAMP accumulation in these cells, its association with agonist-activated PTH1Rc, functionally resulting in impaired desensitization of cAMP signaling. The observation that regulation of signaling by PTHrP-  and Bpa 1 -PTHrP-(1-36) and ligand-receptor endocytosis were rescued by the mutant receptor PTH1Rc(H223R), which is defective in G q -mediated signaling (31) but is constitutively associated with ␤-arrestin-2 (7), clearly indicates that the signaling selectivity of these analogs per se is not the cause of the impaired desensitization. Rather, when combined with the findings that receptor phosphorylation is not necessary for ␤-arrestin-2 translocation (7,8) and with the previous characterization of the molecular motions of the PTH1Rc(H223R) mutant showing that the conformation of the third intracellular loop of the unbound receptor mutant is very similar to the one of cognate agonist-bound PTH1Rc (14), these experiments suggest that the known interaction between residue 1 of PTHrP and the extracellular end of transmembrane helix VI of PTH1Rc (16) mediates the conformational changes leading to the high affinity state for ␤-arrestin-2. In structural terms, interaction of Bpa 1 -PTHrP-(1-36) and PTHrP-(2-36) with PTH1Rc results in stabilization of a non-desensitized G protein-coupled conformation. In contrast, complexes with native PTHrP-(1-36) allow for a distinct confor-  1-36). A, the ligand is depicted in gray, with the backbone displayed as a ribbon. The receptor is depicted as a ribbon, which is colored dark (N terminus-TM1 and TM4 -second extracellular loop-TM5) and light (TM2-first extracellular loop-TM3 and TM6 -third extracellular loop (EC3)-TM7) green. The molecular dynamics simulations of the ligand-receptor complexes were carried out following published procedures (14,15). B, shown is the interaction of Bpa of Bpa 1 -PTHrP-  with Met 425 (shown in Corey-Pauling-Koltun spheres) of PTH1Rc (16). The color scheme is the same as described for A. C, shown is the structure of PTH1Rc resulting from simulations of the ligand-receptor complex with PTHrP-(1-36) (light green) and Bpa 1 -PTHrP-(1-36) (dark green). mational state in which interactions with ␤-arrestin-2 are favored over G protein coupling (i.e. the desensitized state).
Analysis of the conformational changes within PTH1Rc upon interaction with either PTHrP-  or the non-desensitizing analogs was carried out by molecular modeling of ligand-receptor complexes in a simulation cell that mimics the biphasic amphipathic character of cell membranes (24). The starting ligand-receptor structure, based on experimental characterization of the ligands and relevant portions of PTH1Rc, also includes several ligand-receptor contact points identified by biochemical methodologies (16 -23). Among these, particularly important is the interaction of position 1 of Bpa 1 -PTHrP-  with Met 425 at the extracellular end of TM6 of PTH1Rc (16). The molecular dynamics simulations indicate that the major differences in the receptor structure upon interaction with either PTHrP-  or the non-desensitizing position 1-modified analogs arise from the distinct interaction of the modified residues with TM5 and TM6, resulting in altered conformation of IC3. Interestingly, residues in the N-terminal helix of IC3, particularly Val 384 and Leu 385 , have been previously shown to be involved in G q coupling (33), consistent with the observations that the position 1-modified analogs lose their ability to elicit intracellular calcium transients and phosphatidylinositol hydrolysis. Given the importance of IC3 in coupling to both G proteins and arrestins (33)(34)(35)(36)(37)(38), it is not surprising that major differences in the structures of active and desensitized ligandreceptor complexes are found within this domain. This mechanism of receptor function is in good agreement with mutagenesis studies showing the importance of relative movements between TM3 and TM6 of PTH1Rc for G s and G q activation (9,39). These studies, in which receptor stabilization was achieved by engineered Zn 2ϩ -binding sites, also suggested that G protein activation and PTH1Rc phosphorylation and internalization require different conformational switches (9).
The concept that dynamic conformational changes upon agonist binding result in distinct functional receptor states (i.e. the active and resting states) led to the development of the ternary complex model for G protein-coupled and G proteinuncoupled receptors (40). Recent models propose multiple conformational states for GPCRs (41)(42)(43)(44)(45)(46)(47)(48) that are stabilized by different classes of ligands (i.e. agonists, partial agonists, antagonists, and inverse agonists) and/or receptor oligomerization and that mediate distinct molecular and cellular events. We now provide experimental evidence showing that, in the PTH/PTHrP system, structurally similar agonists can selectively stabilize either the active G protein-coupled or the desensitized ␤-arrestin-coupled receptor conformation. The differences between the active and desensitized receptor conformations appear to be related, at least in part, to the induction of distinct structures of IC3 by the ligand via the interaction of residue 1 with the extracellular end of TM6. Collectively, this study and previous reports (7-9) also suggest that receptor occupancy by agonists, rather than covalent receptor modification (i.e. phosphorylation), may be sufficient for induction of a purely structural desensitized receptor state. Indeed, receptor desensitization is one of the major mechanisms in determining both the physiology and pharmacology of GPCRs (49,50). The possibility of modulating such phenomena through the appropriate design of PTH-and PTHrP-derived agonists may have profound influence on the exploitation of pharmacological interventions on PTH1Rc function.