Noncanonical GPCR signaling arising from a PTH receptor–arrestin–Gβγ complex

Abstract

G protein-coupled receptors (GPCRs) participate in ubiquitous transmembrane signal transduction processes by activating heterotrimeric G proteins. In the current “canonical” model of GPCR signaling, arrestins terminate receptor signaling by impairing receptor–G-protein coupling and promoting receptor internalization. However, parathyroid hormone receptor type 1 (PTHR), an essential GPCR involved in bone and mineral metabolism, does not follow this conventional desensitization paradigm. β-Arrestins prolong G protein (G_S)-mediated cAMP generation triggered by PTH, a process that correlates with the persistence of arrestin–PTHR complexes on endosomes and which is thought to be associated with prolonged physiological calcemic and phosphate responses. This presents an inescapable paradox for the current model of arrestin-mediated receptor–G-protein decoupling. Here we show that PTHR forms a ternary complex that includes arrestin and the Gβγ dimer in response to PTH stimulation, which in turn causes an accelerated rate of G_S activation and increases the steady-state levels of activated G_S, leading to prolonged generation of cAMP. This work provides the mechanistic basis for an alternative model of GPCR signaling in which arrestins contribute to sustaining the effect of an agonist hormone on the receptor.

According to the current model for activation and signaling of G protein-coupled receptors (GPCRs), receptor (R) signaling begins when the binding of an agonist ligand (L) stabilizes the active form of the receptor (R*), allowing its coupling to heterotrimeric G proteins (Gαβγ) through a diffusion-controlled process (1–3). The L–R*–G complex catalyzes GDP–GTP exchange on the Gα subunit, promoting the dissociation of the activated G protein into GTP-bound Gα (Gα-GTP) and Gβγ dimers from the receptor. Dissociated Gα-GTP and Gβγ proteins in turn activate specific downstream effectors, such as adenylyl cyclases in the case of Gα_S (Fig. 1A). Activated adenylyl cyclases catalyze a few rounds of cAMP production, the number of which is determined by the rate of GTP hydrolysis of the Gα_S-bound GTP. In order for subsequent rounds of adenylyl cyclase activation to occur, inactive GDP-bound Gα must encounter a Gβγ dimer and then reform an L–R*–G complex (Fig. 1A). In this classical model, arrestin plays a dual role: On one hand, it desensitizes the receptor by preventing further rounds of L–R* and G coupling; on the other, it promotes receptor endocytosis, thus reducing receptor availability on the cell surface. In either case, the classical model predicts that arrestin negatively regulates the levels of the L–R*–G complex, providing the main mechanism by which the signal of a GPCR system is turned off (4, 5). This model, primarily derived from the analysis of the behavior of the β₂-adrenergic receptor (β₂-AR) (4, 5) and many other class 1 GPCRs (6–12), is taken to be universal for GPCR biology (13). Recent findings have challenged this view. We recently showed that parathyroid hormone receptor type 1 (PTHR)–arrestin complexes contribute to prolonging cAMP signaling mediated by PTH or its fully functional N-terminal synthetic analog, PTH(1–34), in cells as diverse as human embryonic kidney (HEK)293 cells expressing recombinant receptor and osteoblastic ROS17/2.8 cells (2, 14) that natively express PTHR. PTH rapidly recruits β-arrestin 1 or β-arrestin 2 to PTHR and stabilizes a persistent ternary PTH–PTHR–arrestin complex that continues signaling via adenylyl cyclases for a considerable time (>30 min) after exposure to a short “pulse” of agonist ligand (14). Because prolonged PTH-induced signaling requires persistent activation of G_S by coupling with PTHR in its L–R* state, this raises the intriguing question of how G_S can bind and be activated by a PTHR that is already associated with arrestin. We reasoned that a long-lived PTH–PTHR–arrestin complex could contribute to sustained cAMP signaling by stabilizing an interaction with the active state of G_S (Fig. 1B, model 1). Alternatively, complexes formed through interactions between β-arrestins and the Gβγ dimer, known to scaffold some signaling complexes (15–17), could promote sustained PTH-induced cAMP signaling by maintaining multiple rounds of Gα_S association with and dissociation from a PTHR–arrestin–Gβγ complex that generates a high level of active Gα_S (Fig. 1B, model 2). Here we tested these hypotheses by using a series of diverse biochemical and biophysical approaches.

Fig. 1.

Signaling models of GPCR. (A) Classical model. A ligand binds the inactive state of a GPCR and stabilizes its active form, which then couples with heterotrimeric G proteins (Gαβγ) through a diffusion-controlled process (step 1). The L–R*–G complex in turn catalyzes GDP–GTP exchange on Gα, leading to dissociation of the GTP-bound Gα (Gα-GTP) along with the Gβγ dimer from the receptor (step 2). In the case of G_S, Gα-GTP activates specific effectors such as adenylyl cyclases (AC), which catalyze the synthesis of cAMP from ATP (step 3). The hydrolysis of GTP to GDP causes the dissociation of Gα_S from adenylyl cyclases, shutting down cAMP production and its reassociation to Gβγ subunits (step 4). In this model, the recruitment of β-arrestin mediates desensitization of G-protein signaling. (B) Noncanonical model. (1) A long-lived PTH–PTHR–arrestin complex could contribute to sustained cAMP signaling by stabilizing an interaction with the active state of G_S. (2) Alternatively, the interaction between the activated PTHR and G_S is stabilized by β-arrestins. After the first round of activation, step 1 is bypassed, such that free Gα-GDP directly reassociates with PTHR–Gβγ complexes to initiate a new cycle of G-protein activation. Arrestin stabilizes the G-protein cycle, resulting in prolonged cAMP production.

Results

We first detected a specific interaction of Gβ₁γ₂ (noted Gβγ) with arrestin induced by PTH(1–34) in transfected HEK293 cells stably expressing PTHR. This interaction was observed by FRET experiments between YFP-tagged β-arrestin 2 (β-arr2^YFP) and the dimer Gβγ labeled by bimolecular fluorescence complementation with the cyan fluorescent protein (Gβγ^CFP) under total internal reflection fluorescence (TIRF) microscopy. Addition of a saturating concentration of PTH(1–34) led to a fast increase in the FRET signal, with a half-life of t_1/2 = 1.4 ± 0.3 min (n = 10) (Fig. 2A, Left). At comparable fluorescence levels, PTH(1–34) did not cause a detectable increase in FRET between β-arr2^YFP and an N-terminally plasma membrane-tagged CFP (CFP^PM) (Fig. 2A, Right). This indicates that the measured FRET between Gβγ and arrestin is not a consequence of random diffusional collisions between arrestin and proteins on the plasma membrane.

Fig. 2.

Formation of a ternary PTHR–arrestin–Gβγ complex by PTH. (A) Time-resolved changes in emission of CFP and YFP fluorescence (F, normalized to the initial value F₀) in single HEK293 cells stably expressing PTHR and transiently expressing β-arr2^YFP and Gβγ^CFP (Left, n = 10). Shown are the changes induced by rapid superfusion with 100 nM PTH(1–34) (arrows). Control experiments were done in cells expressing PTHR, β-arr2^YFP, and CFP^PM (Right, n = 14). Data are the mean ± SEM. (B) Examples of averaged autocorrelation curves from ICCS experiments for Gβγ^CFP and β-arr2^YFP before (Left) and after (Center) PTH(1–34) stimulation was used to calculate fractional binding (Right) for either Gβγ^CFP/β-arr2^YFP or V₂R^CFP/β-arr2^YFP after PTH(1–34) challenge. Data are the mean ± SEM of n = 6 (*P < 0.05, **P < 0.01). (C) HEK293 cells were transiently transfected with PTHR^GFP, Gβγ^CFP, or with (Center) or without (Left) β-arr1[IV-AA]^Tom. (Left) Recovery of CFP fluorescence before (black curve, n = 15) or after (gray curve, n = 4) addition of PTH(1–34) and under control conditions in cells expressing PTHR^GFP and CFP^PM after addition of PTH(1–34) (blue curve, n = 20). (Center) Recovery of CFP in the absence (black curve, n = 7) or presence of PTH(1–34). The gray curve shows the recovery of CFP (n = 7), whereas the red curve indicates the recovery of Tomato fluorescence (n = 5) after addition of PTH(1–34). Note that fluorescence was bleached using a laser focused on the cell-membrane area containing PTHR-GFP and β-arr1[IV-AA]^Tom as visualized by confocal microscope (Left, Upper). Data were used to calculate recovered fluorescence of CFP 60 s after photobleaching (F₆₀, Right) (*P < 0.05, **P < 0.01). Data are the mean ± SEM.

We confirmed the interaction between Gβγ^CFP and β-arr2^YFP using image cross-correlation spectroscopy (ICCS). This technique allows quantitative measurements of the formation of protein complexes in live cells by simultaneously recording fluctuations of fluorescent signals as a function of time in order to calculate a cross-correlation function, a parameter that depends on the relative mobility of each of the two fluorescent proteins (18, 19). Under basal conditions there was no cross-correlation between Gβγ^CFP and β-arr2^YFP (Fig. 2B, Left), indicating that the two proteins diffused independently of one another. Addition of PTH(1–34) significantly increased the cross-correlation of Gβγ^CFP and β-arr2^YFP (Fig. 2B, Center), consistent with an ≈3-fold increase in fractional binding (Fig. 2B, Right). The absence of cross-correlation between β-arr2^YFP and Gβγ^CFP in the presence of isoproterenol (ISO), a β₂-AR agonist, or between β-arr2^YFP and a C-terminally tagged vasopressin V₂ receptor with CFP (V₂R^CFP) after challenge with PTH (Fig. 2B, Right), indicates that the formation of a Gβγ–β-arr complex assembled by the PTH(1–34)–activated PTHR is selective and not due to random proximity of arrestin and proteins on the plasma membrane.

Coimmunoprecipitation experiments further supported the existence of PTHR complexes that include arrestin and/or Gβγ. For these assays, Gβγ^CFP was expressed in HEK293 cells stably expressing HA-tagged PTHR (HA-PTHR), and cell lysates were prepared at different time points after a short exposure to either M-PTH(1–28) or M-PTH(1–14), modified analogs of PTH that induce longer or shorter cAMP generation and biological responses, respectively (20). The difference in duration of cAMP signaling is thought to depend on the capacity of M-PTH(1–28), like PTH(1–34), to form an unusually persistent high-affinity complex with PTHR that is independent of G-protein coupling, whereas M-PTH(1–14) forms a more conventional high-affinity complex that is transient and dependent on G-protein coupling (2, 20, 21). The interactions of HA-PTHR with Gβγ^CFP and β-arrestins, weakly detectable under basal conditions, increased significantly within 5 min after challenge with M-PTH(1–14) or M-PTH(1–28) (Fig. S1). PTHR complexes containing Gβγ and/or β-arrestins were still detectable 30 min after stimulation with M-PTH(1–28) but not M-PTH(1–14) (Fig. S1). Importantly, our previous studies showed that cAMP production induced by M-PTH(1–28) remains elevated for at least 30 min after ligand challenge (14).

It is currently thought that interactions between GPCR and G protein or arrestin are mutually exclusive. However, our FRET/TIRF and coimmunoprecipitation data suggest that G protein and arrestin may bind PTHR simultaneously and thus form PTHR–arrestin–Gβγ complexes. We therefore tested whether stabilizing arrestin association with PTHR would also permit its interaction with Gβγ. To do this, we artificially immobilized an N-terminally GFP-tagged PTHR (PTHR-GFP) (22) at the plasma membrane of HEK293 cells and used a mutant of β-arrestin 1, I³⁸⁶A, V³⁸⁷A (hereafter β-arr1[IV-AA], or β-arr1[IV-AA]^Tom when tagged with tdTomato), that exhibits increased binding affinity for ligand-activated PTHR (14). Cell-surface cross-linking effectively immobilized PTHR without affecting its signaling capacity (Fig. S2). After inducing the formation of stable PTHR–arrestin complexes with PTH(1–34), we photobleached Tomato fluorescence and determined fluorescence recovery as a function of time. We found no fluorescence recovery of β-arr1[IV-AA]^Tom, thus confirming that β-arr1[IV-AA] forms a stable complex with activated PTHR (Fig. 2C, Center). Prior to PTHR activation, the mobile fraction of Gβγ^CFP was similar to that of CFP^PM, which freely diffuses along the plasma membrane. Diffusion of Gβγ^CFP markedly decreased after challenge with PTH(1–34) (Fig. 2C, Left), consistent with a receptor–Gβγ interaction. Furthermore, immobilized PTHR–βarr1[IV-AA] complexes did not prevent the decrease in Gβγ^CFP mobility after challenge with PTH(1–34) (Fig. 2C, Center and Right), indicating that Gβγ can bind activated PTHR even when a stable PTHR–arrestin complex has formed. We observed similar results with β-arr1^Tom (Fig. S3A). When PTHR was activated by M-PTH(1–14), the diffusion of Gβγ remained unchanged (Fig. S3B). These data indicate that interactions between Gβγ and the M-PTH(1–14)–bound receptor did not lead to a stable complex, in accordance with transient kinetics profiles observed when measuring β-arr2–Gβγ interactions (Fig. S4) and earlier kinetic analysis of ligand–PTHR interaction, PTHR activation/deactivation, and G_S activation/deactivation (2, 14). However, the fraction of Gβγ retained by the M-PTH(1–14)–activated PTHR was significantly increased in the presence of β-arr1[IV-AA] (Fig. S3). These data therefore support a model in which the formation of a stable arrestin–receptor complex does not prevent but rather favor interactions of PTHR with G_S.

In a second series of experiments, we sought to identify the specific biochemical steps of PTHR signaling that are influenced by arrestins. We first measured kinetics of association/dissociation between PTHR and Gβγ in the presence of overexpressed Gα_S by measuring intermolecular FRET between C-terminally PTHR tagged with YFP (PTHR^YFP) and G_S^CFP (Gα_S/Gβ₁/Gγ₂^CFP). M-PTH(1–14) induced a fast interaction between PTHR and Gβγ (τ = 1.12 ± 0.05 s), which was rapidly reversed after ligand washout (τ = 1.86 ± 0.38 s) (Fig. 3 A and B, Left). Coexpression of β-arrestin 2, β-arrestin 1, or β-arr1[IV-AA] markedly increased the level of FRET increase after ligand challenge with only a small effect on PTHR–Gβγ association (τ = 3.28 ± 0.12 s) but considerably slowed PTHR–Gβγ dissociation (τ ≈ 20 s for β-arr1/2, τ = 35 s for β-arr1[IV-AA]) (Fig. 3 A and C and Fig. S4). Consistent with these data, when β-arrestin 1 and 2 were depleted using small interfering RNA (siRNA) in HEK293 cells stably expressing PTHR, PTHR–Gβγ dissociation occurred more rapidly than when cells were transfected with control siRNA (τ ≈ 20 s for siRNA-β-arr1/2 vs. 35 s for siRNA-control) (Fig. 3B). These results suggest that β-arrestins prolong the association of Gβγ with ligand-activated PTHR.

Fig. 3.

Kinetics of PTHR–Gβγ interaction modulated by β-arrestins. (A) Time courses of PTHR and Gβγ association/dissociation recorded by changes of the normalized FRET (NFRET) ratio F_YFP:F_CFP in HEK293 cells stably expressing PTHR^YFP and transiently expressing Gβ₁γ₂^CFP without (control) or with β-arrestin 2, β-arrestin 1, or β-arr1[IV-AA] overexpression. Measurements were performed in single cells continuously perfused with control buffer or M-PTH(1–34) (3 μM) for the times indicated by the horizontal bar as previously described (2). (Right) Corresponding averaged time courses for PTHR–Gβγ dissociation. Data are the mean ± SEM of n = 18 (control), n = 11 (β-arr2), n = 16 (β-arr1), and n = 16 (β-arr1[IV-AA]). (B) Similar experiments as in A where cells were transfected with either siRNA targeting human β-arrestin 1 and β-arrestin 2 (si-β-arr1/2) or scrambled siRNA (control). Data represent the mean ± SEM of four independent experiments; n = 6 (control) and n = 12 (si-β-arr1/2) cells. (C and D) Quantification of the change in FRET induced by M-PTH(1–14) in experiments shown in A and B. *P < 0.05, **P < 0.01 when comparing the values obtained for control versus β-arr1, β-arr2, or β-arr1[IV-AA] overexpression, or siRNA-control versus siRNA-β-arr1/2.

We next studied the interaction between PTHR and Gα_S by measuring FRET between PTHR tagged at its C terminus with CFP (PTHR^CFP) and G_S^YFP (Gα_S^YFP/Gβ₁/Gγ₂) to determine whether Gα_S is part of the same PTH–PTHR–Gβγ complex stabilized by arrestin. In accordance with previous bioluminescence resonance energy transfer studies done by the group of Michel Bouvier with β₂-AR (23), measuring PTHR–Gα_S interaction also proved to be difficult due to the small magnitude of change in FRET (≈1.5%) in response to PTH. Nonetheless, we were able to measure kinetics of association and dissociation. As expected, we found that Gα_S interacted with PTHR with similar kinetics as those measured for Gβγ (Figs. S5 and S6). Strikingly, the dissociation kinetics was significantly faster after overexpression of β-arrestin 2 (τ ≈ 2 s for control vs. ≈1 s in the presence of arrestin, P = 0.018) as opposed to the dissociation kinetics of PTHR–Gβγ complexes (τ ≈ 2 s for control vs. ≈20 s in the presence of arrestin). Therefore, these data suggest that at least a fraction of the active state of Gα_S is not in the PTHR–Gβγ–arrestin complex. The dissociation of Gα_S was fast enough to permit its reassociation with Gβγ within the PTH-bound PTHR–arrestin complex, thus favoring an immediate initiation of the next G-protein activation cycle.

Here we used two distinct approaches to identify the specific steps of PTHR signaling that are modulated by arrestin. To this end, we first studied G_S activation by measuring the binding of [³⁵S]GTPγS, a nonhydrolyzable analog of GTP, to Gα_S. Membranes from HEK293 cells stably expressing PTHR were incubated at different time points with either PTH(1–34) or ISO, a β₂-AR agonist, in the absence or presence of purified β-arrestin 2 (Fig. S7). In agreement with previous studies showing that β-arrestins prevent β₂-AR signaling (5, 24), purified β-arrestin 2 reduced GTPγS binding to Gα_S in response to ISO (Fig. 4B). However, the maximal extent of [³⁵S]GTPγS binding on Gα_S increased ≈1.5-fold (P < 0.01) by PTH(1–34) in the presence of purified β-arrestin 2, indicating that arrestin stabilizes the active form of G_S.

Fig. 4.

Consequences of a ternary PTHR–arrestin–Gβγ complex on G_S activation. (A) Time courses of PTH(1–34)–stimulated [³⁵S]GTPγS binding to membrane preparations of HEK293 cells stably expressing PTHR in the absence (gray curve) or presence (black curve) of purified β-arrestin 2 (100 nM). Data are the mean ± SEM of n = 3 experiments. (B) Quantification of [³⁵S]GTPγS binding results shown in A. Control experiments were done with 10 μM ISO. Bars represent the mean ± SEM of n = 3 (GTPγS binding) (**P < 0.01, ***P < 0.001). (C) Examples of the time course of G_S activation measured by FRET in HEK293 cells stably expressing PTHR and transiently expressing G_S^CFP/YFP (Gα_S-YFP/Gβ₁/Gγ₂-CFP) with (Right) or without (Left) overexpression of β-arrestin 2. Measurements were performed in single cells continuously perfused with buffer or briefly perfused with M-PTH(1–14) (3 μM) for the time indicated by the horizontal bar, as previously described (2, 7). G_S activation is associated with a decrease in FRET signal. (D) Quantification of G-protein activation measured by FRET with G_S^CFP/YFP as shown in C. Bars represent the mean ± SEM of n = 9 (β-arr2) and n = 4 (β-arr1) experiments (**P < 0.01). (E) Averaged time courses of G_S activation/deactivation corresponding to experiments shown in C. Data are the mean ± SEM of n = 8 (control) and n = 9 (β-arr2) experiments.

We next used a previously described FRET-based biosensor for G_S activation, Gα_S^YFP/Gβ₁/Gγ₂^CFP (G_S^YFP/CFP), (1, 2) to test whether β-arrestin 2 stabilizes the active state of G_S in live cells. A brief application of M-PTH(1–14) induced a fast decrease of FRET signal in cells stably expressing PTHR, reflecting conformational rearrangements and/or dissociation between Gαβγ subunits that occur during G_S activation (Fig. 4 C and D, Left). Overexpression of β-arrestin 2 accelerated (τ = 0.71 ± 0.08 s) and further increased the magnitude of G_S activation as measured by FRET. Similar results obtained with β-arrestin 1 confirmed that arrestins contribute to increasing the active state of G_S (Fig. 4 C and D, Right), in agreement with GTPγS binding experiments (Fig. 4B). Collectively, these results suggest that Gβγ remains bound to PTHR while in complex with arrestin during the G-protein activation cycle, resulting in accelerated G_S activation and higher levels of active G_S.

If arrestin stabilizes an active PTHR–Gβγ complex, then we would expect that stabilizing the association of PTHR and arrestin would prolong cAMP signaling mediated by even short-acting agonist ligands such as M-PTH(1–14). Indeed, overexpression of β-arr1[IV-AA] prolonged the production of cAMP mediated by M-PTH(1–14) after ligand washout, whereas in control cells cAMP production required the continuous presence of extracellular ligand (Fig. 5A). Conversely, we found a significant decrease in cAMP under a condition with siRNA knocking down β-arrestins upon stimulation with M-PTH(1–14) (Fig. 5A). We also validated this prediction in the osteoblast-like cell line ROS17/2.8, which natively expresses PTHR. Given the undetectable expression of β-arrestin 2 compared with that of β-arrestin 1 in these cells, a finding that we also confirmed in a primary culture of mice calvarial osteoblasts (Fig. 5B), we compared the effect of β-arrestin 1 on cAMP responses to PTH or ISO. Addition of ISO (10 μM) or PTH(1–34) (100 nM) caused a fast increase in cAMP similar in extent to that mediated by forskolin (10 μM) (Fig. 5C). Following this increase, the cAMP response triggered by ISO was rapidly desensitized whereas that mediated by PTH persisted much longer after an initial decline, consistent with previous reports (14). β-Arrestin 1 overexpression markedly reduced cAMP generation in response to ISO, but prolonged cAMP induced by challenge with PTH (Fig. 5C). Arrestin therefore had effects on two distinct receptors that were different both in magnitude and duration: It attenuated signaling by β₂-AR but enhanced cAMP generation by PTHR.

Fig. 5.

Arrestin control of cAMP mediated by PTHR. (A) Averaged cAMP response over 25 min measured by FRET changes from HEK293 cells stably expressing PTHR and transiently expressing a cytoplasmic cAMP FRET sensor, epac-CFP/YFP, in the absence (black curve) or presence (red curve) of β-arr1[IV-AA] (Left, n = 25), and with either siRNA targeting β-arr1/2 or scrambled siRNA (Right, n = 19). Cells were continuously perfused with control buffer or 100 nM M-PTH(1–14) (horizontal bar). Data represent the mean ± SEM. (B) Western blot analysis of β-arrestin expression in HEK293 cells, the osteoblastic-like ROS17/2.8 cell line, and primary calvarial osteoblasts (OB) from mice. (C) Averaged cAMP responses mediated by ISO (10 μM) or PTH(1–34) (100 nM) in osteosarcoma cells. Cyclic AMP induction over a 25-min time course was monitored by FRET changes from ROS17/2.8 osteoblastic-like cells transiently expressing the cAMP FRET-based biosensor epac-CFP/YFP alone (control) or with β-arrestin 1. Fsk, forskolin. (D) Bars represent the average cAMP responses of experiments shown in C determined by measuring the area under the curve from 0 to 25 min for cAMP. Data represent the mean ± SEM of n = 8 (ISO, ctrl), n = 8 (ISO, β-arr1), n = 14 (PTH, ctrl), and n = 8 (PTH, β-arr1) (**P < 0.01).

Discussion

In the classical and widely accepted model of GPCR desensitization, arrestin terminates GPCR signaling by preventing receptor and G-protein association (4, 5, 25, 26) and also by recruiting diverse enzymes such as the phosphodiesterase PDE4 to rapidly degrade cAMP at the plasma membrane (27). Arrestin binding also reduces receptor number at the cell membrane by promoting receptor internalization through a clathrin-coated pit mechanism (28). However, PTHR does not follow this conventional desensitization paradigm. Arrestins interact rapidly with PTHR upon PTH stimulation (29) and induce its internalization, resulting in prolonged cAMP responses in cultured cells expressing recombinant or native PTHR (14). These prolonged signaling responses correlate temporally with the persistence of PTHR–arrestin–G_S complexes on early endosomes (2, 14), a process that is thought to be associated with prolonged physiological calcemic and phosphate responses observed for long-acting PTH analogs in intact animals (20).

Our results set the molecular basis for understanding how long-lived complexes of PTHR–arrestin–Gβγ complexes can sustain cAMP signaling. Our findings strongly support a scaffolding role for arrestin that results in the stabilization of a complex of PTHR and Gβγ that accelerates G-protein activation and increases the steady-state levels of active G_S, thus resulting in prolonged generation of cAMP. As a consequence of stabilizing PTHR–Gβγ complexes, after GTP hydrolysis, Gα_S subunits are likely to encounter preformed L–PTHR–Gβγ complexes. This would bypass the step in which Gα_S-GDP must reassociate with free Gβγ to reform the Gαβγ trimer before binding to the ligand-activated receptor, obligatory for promoting subsequent rounds of Gα_S activation (Fig. 1B, model 2). These observations coupled with the slightly but significantly (P < 0.05) faster kinetics of G_S activation of arrestin (τ = 0.71 min for β-arr2 vs. 1.15 min for control) with no significant change in deactivation kinetics in the presence of arrestin point to model 1 (Fig. 1B) as a possible mechanism by which a PTHR–arrestin–Gβγ complex maintains multiple rounds of Gα_S subunit coupling and activation, thus increasing the pool of active G_S. However, the FRET approach to measuring kinetics of receptor–protein interactions in live cells is limited by ensemble averaging. Therefore, our kinetics analysis of PTHR–Gα_S interaction cannot rule out the possibility that only a fraction of active Gα_S dissociates from PTHR–arrestin–Gβγ complexes, whereas the other fraction remains associated. This would be consistent with model 2 (Fig. 1B), wherein G_S signals without dissociational events. A comprehensive model of our data is then more compatible with the existence of two reversible active forms of Gα_S (i.e., Gα_S-GTP), one of which remains associated with Gβγ subunits in complex with PTHR and arrestin, whereas the other is separated from the PTHR–arrestin–Gβγ complex (Fig. S8). The determination of the free Gα_S-GTP fraction and the relative fraction physically associated with the PTHR–arrestin–Gβγ complex represents an important aim for future studies.

Initially demonstrated for arrestin-dependent MAP kinase pathways (30–33), endosomal signaling via heterotrimeric G proteins has been shown in yeast (34) and recently extended to GPCRs in vertebrates. Sustained cAMP and G_S signaling by receptor internalization in early endosomes was originally reported in 2009 for two distinct GPCRs, the thyroid-stimulating hormone receptor (TSHR) (35) and PTHR (2), and recently expanded to the dopamine D1 receptor (D1R) (36). Endosomal G-protein signaling appears to be an alternative pathway not only for G_S- but also for inhibitory G protein (Gi)-coupled receptors, as demonstrated for the sphingosine-1-phosphate receptor (37). In keeping with the model that we present here, several questions remain to be studied, among them the following. Is the internalization process (i.e., clathrin-coated pit formation) linked to the slow dissociation rate of Gβγ from the PTHR–arrestin complex? What events contribute to slow down the rate of Gβγ association with PTHR besides the dissociation rate? What are the molecular and cellular mechanisms that prevent arrestins from sustaining heterotrimeric Gq/phospholipase C (PLC) signaling by PTHR? Does the “noncanonical” signaling mechanism shown here also apply to other G_S-coupled receptors such as D1R and TSHR? Although the generality of the noncanonical model, in the sense of whether it applies to other receptors, requires further studies, our results provide a molecular understanding linking prolonged cAMP signaling by the PTH/PTHR system and its interaction with arrestin.

Materials and Methods

Image Cross-Correlation Spectroscopy.

These studies were based on the experimental techniques described by Wiseman and coworkers (38–40, as adapted by Wheeler et al. (19). Fluorescence emissions from enhanced cyan fluorescent protein-tagged Gβγ channels and enhanced yellow fluorescent protein-tagged β-arrestin 2 were collected from transiently transfected HEK293 cells. A Nikon A1 confocal microscope focused on the cell plasma membrane adjacent to the coverslip was used. Images (>300) were collected from a small section of the plasma membrane (<50 μm²), which was rapidly scanned (50–60 ms per frame) under low laser power. Data were exported to ImageJ (rsbweb.nih.gov/ij/) and analyzed using a plug-in specifically written to calculate the cross-correlation function of the data (19). Fractional binding was calculated from the ratio of the amplitude of the cross-correlation function to the amplitude of the autocorrelation function for PTHR. These calculations are based on the method described by Kim et al. (41).

Additional Methods.

TIRF, FRET, fluorescence recovery after photobleaching, immunoprecipitation, arrestin purification, GTPγS binding, and immobilization of PTHR are described in SI Materials and Methods.