Research Article

Noncanonical scaffolding of Gαi and β-arrestin by G protein–coupled receptors

See allHide authors and affiliations

Science  12 Mar 2021:
Vol. 371, Issue 6534, eaay1833
DOI: 10.1126/science.aay1833

Another way for GPCRs to signal

G protein–coupled receptors (GPCRs) normally transmit signals by coupling to heterotrimeric guanine nucleotide–binding proteins (G proteins) or by binding β-arrestin proteins. Smith et al. provide evidence for another mechanism, an approximate combination of the two. They monitored the interaction of vasopressin type 2 receptors (V2Rs) and Gα proteins in cultured cells using bioluminescent resonance energy transfer. Even though V2Rs do not signal canonically through Gαi proteins, they promoted the formation of complexes containing β-arrestin and Gαi, and this led to downstream signaling to extracellular signal-regulated kinase protein kinases.

Science, this issue p. eaay1833

Structured Abstract


G protein–coupled receptors (GPCRs) are a superfamily of seven transmembrane-spanning receptors and are the target of 30% of all U.S. Food and Drug Administration–approved medications. GPCRs are involved in nearly every human physiological process by serving as receptors for a wide array of ligands, including proteins, peptides, fatty acids, small molecules, and ions. The canonical signaling mechanisms of GPCRs involve coupling to specific G protein subtypes (e.g., Gαs, Gαi, Gαq, and Gα12) as well as β-arrestin adaptor proteins. Each G protein subtype regulates specific intracellular second messengers, such as cyclic adenosine 3′,5′-monophosphate (cAMP) responses or calcium mobilization. β-arrestins inhibit, or “arrest,” this canonical G protein–mediated signaling while also promoting their own intracellular signaling events such as extracellular signal–regulated kinase (ERK) activity. Because G proteins and β-arrestins can differentially regulate signaling pathways, often with distinct cellular effects, efforts are under way to design drugs that preferentially target either G protein or β-arrestin signaling to generate more selective GPCR-targeted drugs, i.e., biased ligands. In theory, biased ligands can improve the desired pharmacological properties of drugs while also minimizing their on-target side effects. However, potential coordination between G protein and β-arrestin signaling could have a significant impact on our fundamental understanding of GPCR signaling and the development of biased ligands.


G protein and β-arrestin signaling have broadly been considered separable intracellular pathways with distinct signal transduction mechanisms. However, several lines of evidence have also suggested the potential for coordination between G protein and β-arrestin signaling. For example, some GPCRs can simultaneously bind a G protein and a β-arrestin. Furthermore, when GPCRs were activated in cells lacking functional G proteins, some β-arrestin–mediated effects of GPCR activation were also absent. This could be interpreted as a role for G proteins in what had previously been considered solely β-arrestin–mediated cellular responses. To gain further insight into these experimental findings, we investigated whether there is a GPCR-signaling pathway that involves direct interactions between G proteins and β-arrestins.


We first developed a bioluminescent resonance energy transfer (BRET) approach to monitor tripartite protein interactions among GPCRs, G proteins, and β-arrestin. After confirming that we could monitor such interactions, we found that GPCRs can catalyze a direct interaction between the Gαi protein subtype and β-arrestins at the plasma membrane, and confirmed the interaction using a variety of biochemical and biophysical techniques. This interaction was only observed between β-arrestin and the Gαi protein family, not other Gα protein subtypes. These Gαi:β-arrestin complexes were formed downstream of all receptors tested, even with receptors that do not canonically signal through Gαi, such as the Gαs-coupled β2-adrenergic and vasopressin type 2 receptors. We found that Gαi:β-arrestin complexes formed scaffolds with the signaling kinase ERK downstream of certain receptors. Stimulation of the angiotensin type II receptor with a β-arrestin–biased agonist promoted the formation of the Gαi:β-arrestin complex despite not activating canonical G protein signaling. Cellular migration promoted by this β-arrestin–biased agonist was sensitive to inhibition of both β-arrestins and Gαi, which is consistent with Gαi:β-arrestin complexes regulating cellular migration.


Our results reveal a GPCR-signaling paradigm in which GPCRs promote the formation of Gαi:β-arrestin signaling complexes that are distinct from other canonical forms of GPCR signaling. These Gαi:β-arrestin complexes have the ability to scaffold important cellular effectors such as ERK, and can play a functional role in cellular responses such as cell migration. Demonstration of Noncanonical coordination of Gαi and β-arrestins by GPCRs adds to our understanding of the fundamental mechanisms underlying GPCR signaling, and the potential effects of this complex should be considered when designing and evaluating GPCR ligands.

GPCR signaling by Gαi:β-arrestin complexes.

Complementation and proximity assays identified a complex between Gαi protein family members, but not other Gα protein subtypes, and β-arrestin. This was observed across multiple receptors, including those that do not canonically couple to Gαi. This complex was associated with functional effects, including cell migration. These findings demonstrate GPCR signaling by Gαi:β-arrestin complexes.


Heterotrimeric guanine nucleotide–binding protein (G protein)–coupled receptors (GPCRs) are common drug targets and canonically couple to specific Gα protein subtypes and β-arrestin adaptor proteins. G protein–mediated signaling and β-arrestin–mediated signaling have been considered separable. We show here that GPCRs promote a direct interaction between Gαi protein subtype family members and β-arrestins regardless of their canonical Gα protein subtype coupling. Gαi:β-arrestin complexes bound extracellular signal–regulated kinase (ERK), and their disruption impaired both ERK activation and cell migration, which is consistent with β-arrestins requiring a functional interaction with Gαi for certain signaling events. These results introduce a GPCR signaling mechanism distinct from canonical G protein activation in which GPCRs cause the formation of Gαi:β-arrestin signaling complexes.

Heterotrimeric guanine nucleotide–binding protein (G protein)–coupled receptors (GPCRs) couple to specific Gα protein subtypes and activate G proteins by catalyzing guanine nucleotide exchange (1). These distinct Gα protein subtypes then amplify specific second messenger signaling systems. β-arrestins inhibit, or “arrest,” canonical G protein–mediated signaling and promote other forms of GPCR signaling, such as phosphorylation of extracellular signal–regulated kinase (ERK) (2, 3). G proteins and β-arrestins may coordinate their signaling in part because they can be found in signaling complexes consisting of a GPCR bound to both G protein and β-arrestin (4, 5).

We developed a bioluminescent resonance energy transfer (BRET) method to identify tripartite interactions among GPCRs, Gα proteins, and β-arrestins. Agonist treatment of the Gαs-coupled vasopressin type 2 receptor (V2R) expressed in human embryonic kidney (HEK) 293T cells catalyzed the formation of Gαi:β-arrestin scaffolds that were not observed between β-arrestin and other Gα subtypes. A similar pattern of Gαi:β-arrestin complex formation was also observed after activation of each GPCR that we tested, which included receptors canonically coupled to Gαs, Gαi, and Gαq. The Gαi:β-arrestin scaffolds formed complexes with GPCRs and also bound ERK. Disrupting Gαi and β-arrestin interactions attenuated V2R-mediated activation of ERK phosphorylation. Disrupting Gαi and β-arrestin interactions eliminated GPCR-mediated migration toward a β-arrestin–biased agonist that does not stimulate canonical Gαi signaling. These results reveal the formation of Gαi:β-arrestin signaling scaffolds as a GPCR signaling mechanism.


β-arrestin, Gαi, and receptors form complexes

GPCRs differentially associate with β-arrestins in cells exposed to agonists (6). Activation of V2R results in a long-lived receptor association with β-arrestin (7). This contrasts with other receptors, such as the β2AR, that form transient interactions with β-arrestin resulting in dissociation occurring at or near the plasma membrane (8). Gαs, β-arrestin, and a GPCR can be detected in a complex (4). We used BRET to examine such complexes (Fig. 1A) (9, 10). We used a split luciferase system, nanoluciferase binary technology (NanoBiT), relying on a small 11–amino acid complementing peptide (smBiT) fused to one protein and a nearly full-length nanoluciferase (LgBiT) fused to another protein (fig. S1A) (11). The complementation of the split luciferase components is driven by the interaction of their fusion partners (e.g., β-arrestin and Gα proteins) because smBiT and LgBiT have low affinity for each other (12). When in close proximity, the split luciferase can then transfer energy to a third protein tagged with the fluorescent protein acceptor, monomeric Kusabira orange (mKO), generating a BRET response (fig. S1, B and C) (13). This technology enables real-time quantification of interactions among three proteins in living cells, and we used it to confirm simultaneous interactions among Gαs, β-arrestin, and V2R expressed in HEK 293T cells after agonist treatment (Fig. 1, B and C).

Fig. 1 G protein:β-arrestin:GPCR complex formation.

(A) Arrangement of luciferase fragments and mKO acceptor fluorophore on G protein (LgBiT), β-arrestin-2 (mKO), and V2R (smBiT). HEK 293T cells were transiently transfected and stimulated with agonist or vehicle. (B) BRET ratio of Gαs-LgBiT:β-arrestin-mKO:V2R-smBiT after AVP (500 nM) treatment. After AVP treatment, an increase in the BRET ratio was observed in cells expressing β-arrestin-mKO but not in cells expressing cytosolic mKO. (C) Quantification of Gαs-LgBiT:β-arrestin-mKO:V2R-smBiT complex formation in cells treated with either vehicle or AVP at a single 5-min time point. (D) Similar experiment to panel (B), except testing the ability of Gαi to form a complex. BRET ratio of Gαi-LgBiT:β-arrestin-mKO:V2R-smBiT after treatment with AVP. After AVP treatment, an increase in the BRET ratio was observed in cells expressing β-arrestin-mKO, but not in cells expressing cytosolic mKO [similar to (B)]. (E) Gαi-LgBiT:β-arrestin-mKO:V2R-smBiT formation at a single 5-min time point. (F) Rearrangement of BRET components; Gαi protein (LgBiT), β-arrestin (smBiT), and V2R (mKO). (G) BRET ratio of Gαi-LgBiT:β-arrestin-smBiT:V2R-mKO after AVP treatment. Rearrangement of BRET tags increased the observed signal compared with (D). (H) Five-minute quantification of Gαi-LgBiT:β-arrestin-smBiT:V2R-mKO complexes. (I) Similar experiment to (G), except testing the ability of Gαi-LgBiT:β-arrestin-smBiT to form a complex with the β2AR-mKO as opposed to the V2R-mKO. After isoproterenol (10 μM) treatment, an increase in the BRET ratio was observed in cells expressing β2AR-mKO but not in cells expressing cytosolic mKO. (J) Five-minute quantification of Gαi-LgBiT:β-arrestin-smBiT:β2AR-mKO complexes. For kinetic experiments, *P < 0.05 by two-way ANOVA, Fisher’s post hoc analysis with a significant difference between treatments; for 5-min quantification, *P < 0.05 by Student’s two-tailed t test. For (B) to (E), n = 3; for (G) to (J), n = 4 replicates per condition. Graphs show mean ± SEM. Cyto, cytosolic.

The canonically Gαs-coupled V2R also formed a complex with Gαi and β-arrestin in HEK 293T cells treated with the V2R agonist arginine vasopressin (AVP) (Fig. 1, D and E). We also swapped the location of the dipole donor and acceptor components (Fig. 1F). Rearranging the BRET components increased the observed signal and confirmed that Gαi:β-arrestin complexes can associate with the V2R (Fig. 1, G and H, and fig. S2A). Agonist treatment of the Gαs-coupled β2AR also induced the formation of Gαi:β-arrestin:β2AR complexes (Fig. 1, I and J, and fig. S2B). We validated the specificity of Gαi:β-arrestin:GPCR complex formation by cotransfecting both untagged and mKO-tagged β2AR and V2R. Gαi:β-arrestin:GPCR complexes were only detected in cells treated with the cognate agonist for the mKO-tagged receptor (fig. S2, C to H), indicating a specific interaction of Gαi:β-arrestin with activated receptors.

Gαi:β-arrestin:V2R complexes at the plasma membrane

We visualized localization of Gαi, β-arrestin, and V2R by confocal microscopy (Fig. 2). We validated the imaging parameters using single-color controls to ensure accurate quantification of each component channel (fig. S3). Colocalization of Gαi, β-arrestin, and V2R occurred after agonist treatment and was most prominent at the plasma membrane (Fig. 2, A and B). Line scan analyses demonstrated plasma membrane–localized puncta containing Gαi, β-arrestin, and V2R components after 5 min of agonist treatment (Fig. 2C). Thirty minutes after agonist treatment, clear endosomal localization of β-arrestin and V2R was observed that largely lacked Gαi (Fig. 2D). These observations are consistent with colocalization of Gαi, β-arrestin, and V2R that occurs after agonist treatment and is most prominent at the plasma membrane.

Fig. 2 Confocal microscopy of Gαi:β-arrestin:V2R complexes.

Confocal microscopy analysis of AVP-induced complexes of Gαi:β-arrestin:V2R in HEK 293T cells transfected with mVenus-tagged Gαi, mKO-tagged β-arrestin-2, and Mars1-tagged V2R. (A) Single cells preceding treatment (basal), at 5 min, or at 30 min. Localization of Gαi-mVenus:β-arrestin-mKO:V2R-Mars1 was observed at 5 min, with less at 30 min. Scale bars, 5 mm. (B) Inset of images in (A). Scale bars, 1 mm. (C) Line scan analysis of the 5-min time point demonstrating colocalization of fluorophores after AVP (100 nM) treatment. (D) Line-scan analysis of the 30-min time point. Data are representative of 10 (basal), 20 (5 min), or 15 (30 min) fields of view from three independent experiments.

Distinction of Gαi complexes with β-arrestin from Gαs:β-arrestin:receptor complexes

The difference in magnitude of signal observed in BRET between Gαi and Gαs in G protein:β-arrestin:V2R complexes may reflect distinct interaction orientations. Sustained G protein signaling can exist at receptors after β-arrestin–dependent internalization (14), and β-arrestins can be catalytically activated by an agonist-occupied receptor (15). We therefore investigated whether a direct functional interaction between G proteins and β-arrestins could be catalyzed by agonist treatment of the V2R. We confirmed that the V2R canonically signaled through Gαs (Fig. 3A), increased intracellular concentration of cyclic adenosine monophosphate (cAMP) (Fig. 3B), and recruited β-arrestin (Fig. 3C). V2R did not canonically signal through Gαi in a transforming growth factor–α (TGF-α) shedding assay (16) that works by coupling Gα protein activation to alkaline phosphatase activity to generate an absorbance signal (Fig. 3A).

Fig. 3 Canonically Gαs-coupled V2R forms Gαi:β-arrestin complexes.

(A) Canonical G protein signaling of the V2R. (B) cAMP generation of the V2R. (C) Canonical β-arrestin-2-smBiT recruitment after AVP (500 nM) treatment of the V2R-LgBiT. (D) Arrangement of luciferase fragments on G protein (LgBiT) and β-arrestin (smBiT). (E) Only Gαi-LgBiT formed complexes with β-arrestin-2-smBiT after AVP (500 nM) treatment in cells overexpressing the V2R. (F) Concentration-response on Gαi-LgBiT:β-arrestin-2-smBiT association in cells overexpressing the V2R. (G) Gαi-LgBiT:β-arrestin-2-smBiT complex formation is sensitive to PTX pretreatment. Data are normalized to maximal signal within each replicate. (H) Loss of PTX sensitivity on Gαi-LgBiT:β-arrestin-2-smBiT by mutation of the ADP-ribosylation site (C352). (I) Pretreatment for 30 min with a membrane-permeable (SR121463, 10 μM) or a membrane-impermeable (H3192, 10 μM) V2R antagonist on Gαi-LgBiT:β-arrestin-2-smBiT complex formation. (J) V2R-LgBiT association with β-arrestin-2-smBiT after either membrane-permeable SR121463 or membrane-impermeable H3192 V2R antagonists. (K) Gαi-LgBiT association with β-arrestin-2-smBiT after membrane-permeable (SR121463) or membrane-impermeable (H3192) V2R antagonists. (L) GST-β-arrestin-2 association with purified Gαi. For (A), TGF-α shedding assay was conducted in Δ3G HEK 293T cells. All other experiments were conducted in WT HEK 293T cells overexpressing the indicated assay components. For (A) and (E), *P < 0.05 by two-way ANOVA, with a main effect of Gαs or Gαi, respectively, versus other Gα subunits. For (G) and (H), *P < 0.05 by two-way ANOVA, main effect of PTX treatment. For (I), *P < 0.05 by two-way ANOVA, main effect of vehicle relative to either antagonist. For (J) and (K), *P < 0.05 by two-way ANOVA, Bonferroni post hoc test of vehicle versus either antagonist; #P < 0.05 by two-way ANOVA, Bonferroni post hoc test of SR121463 versus H3192. For (A) to (C), (E), (F), and (H) to (J), n = 3; for (K), n = 6 to 9; for (G), n = 8 replicates per condition. (L) is representative of three pulldown experiments. Graphs show mean ± SEM.

We evaluated the four primary Gα families, Gαs, Gαi, Gαq, and Gα12, in the split luciferase (nanoBiT) system (Fig. 3D). In contrast to canonical V2R-Gαs protein signaling, only Gαi, but not Gαs, Gαq, or Gα12, was detected in complexes with β-arrestin when V2R was treated with AVP (Fig. 3, E and F).

Given that the V2R is not known to signal through Gαi and that there was an absence of canonical Gαi signaling in our assays, we then varied the amounts of Gα subunits transfected by up to 10-fold in our Gα:β-arrestin complex formation assay. We did not observe an interaction between Gαs, Gαq, or Gα12 and β-arrestin with agonist-activated V2R or β2AR. However, we did observe an interaction between Gαi and β-arrestin under these conditions (fig. S4, A to H). Transient overexpression of increased amounts of Gαi directly correlated with increased basal association between Gαi and β-arrestin (fig. S5). Together with the decreased agonist-induced signal, these data are supportive of a specific association between Gαi and β-arrestin that can be promoted by GPCRs. As controls, we verified that the various Gα protein subtypes were expressed in similar amounts (fig. S6). Furthermore, Gαi isoforms 2 and 3, as well as the very similar Gαo, also interacted with β-arrestin after agonist activation of V2R (fig. S7). A similar Gαi-family interaction with β-arrestin-1 was also observed (fig. S8, A and B).

We then tested whether Gαi and β-arrestin could interact in multiple orthogonal assays. We performed nanoBRET between Gα-NanoLuc and β-arrestin-2-mKO to confirm that this complex formed only with Gαi and not Gαs (fig. S9A). Purified Gαi and β-arrestin demonstrated a physical interaction in an immunoprecipitation assay (Fig. 3G). A thermal shift assay (TSA) was used that measures the thermodynamic effects of protein-protein interactions through the binding of a hydrophobic dye (17). The presence of nonhydrolyzable guanosine 5′-triphosphate (GTP) with purified Gαi and β-arrestin-2-394X [which induces an active conformation of β-arrestin (18)] shifted the transition temperatures relative to either Gαi plus nonhydrolyzable GTP or β-arrestin alone, indicating an interaction (fig. S9, B to E). These biochemical experiments are consistent with Gαi and β-arrestin directly interacting without the need for other partners. In a cellular context, overexpressed and tagged β-arrestin coimmunoprecipitated with similarly overexpressed and tagged V2R and Gαi in HEK 293T cells (fig. S10).

The interaction between Gαi and β-arrestin was partially sensitive to pertussis toxin (PTX) (Fig. 3H), which promotes enzymatic ADP ribosylation of cysteine 352 in helix 5 of Gαi (19). Mutation of cysteine 352 to isoleucine relieved sensitivity to PTX (Fig. 3I). PTX pretreatment of HEK 293T cell expressing either the V2R or β2AR did not affect β-arrestin recruitment (fig. S11, A and B). This treatment also did not appear to reduce Gαi:β-arrestin complex formation by interfering with β-arrestin recruitment to the receptor (15, 20).

We investigated whether the V2R could recruit Gαi and if β-arrestin was required. Gαi was recruited to activated V2R in both wild-type (WT) HEK 293T cells and cells depleted of β-arrestin-1 and β-arrestin-2 (fig. S12, A and B). As a control, Gαs was also recruited to the V2R in cells treated with agonist (fig. S12, C and D). These results are consistent with prior findings that canonically Gαs-coupled receptors can also recruit Gαi (21).

Next, we sought to determine whether disrupting β-arrestin recruitment would impair Gαi:β-arrestin complex formation. GPCR kinases (GRKs) phosphorylate intracellular serine and threonine residues of the receptors, often within the C-terminal tail, to promote binding of β-arrestin (22, 23). β-arrestin interacts with both a phosphorylated GPCR tail and the receptor core (24). To determine the contribution of the phosphorylated tail, we used the canonically Gαi-coupled chemokine receptor CXCR3 that has a functional C-terminal truncation mutant (CXCR3 L344X, lacking phosphorylatable tail residues) (25). Both the L344X mutation and a mutation of four C-terminal serine and threonine residues to alanine (CXCR3 4xA) resulted in CXCR3 receptors that were equivalently expressed on the surface of HEK 293T cells as WT CXCR3 (fig. S13, A and B). Canonical Gαi signaling downstream of CXCR3 L344X showed a greater response [left-shifted potency (EC50) and higher maximal efficacy (Emax)] than either WT CXCR3 or CXCR3 4xA (fig. S13C). For β-arrestin-2 recruitment, both CXCR3 L344X and CXCR3 4xA had an attenuated response (lower Emax) relative to that of WT CXCR3 (fig. S13D). Similarly, the mutant CXCR3 receptors displayed reduced Gαi:β-arrestin complex formation (fig. S13E) but had kinetics similar (fig. S13F) to those of WT CXCR3. This suggests that β-arrestin recruitment to the receptor is necessary for Gαi:β-arrestin complex formation (4, 26).

We used V2R antagonists to assess the reversibility of Gαi:β-arrestin complex formation. Treatment of HEK 293T cells overexpressing V2R with either a membrane-permeant V2R antagonist (SR121463) or a membrane-impermeable V2R antagonist (H3192) prevented formation of Gαi:β-arrestin complexes (Fig. 3J). We assessed the ability of these antagonists with different cell permeability properties to disrupt β-arrestin association either with V2R or Gαi. Interactions between the V2R and β-arrestin occur both at the plasma membrane and within intracellular endosomes (6). The membrane-permeant antagonist was more efficient at reversing V2R:β-arrestin association than the membrane-impermeable antagonist (Fig. 3K). By contrast, both antagonists equivalently disrupted Gαi:β-arrestin complexes (Fig. 3L), indicating that Gαi:β-arrestin complexes are present primarily at the plasma membrane, consistent with the confocal findings. The antagonist experiments also demonstrate the reversible nature of this complex and confirm that the underlying interaction between Gαi and β-arrestin generated the signal in the nanoBiT system. After antagonist addition, the half-life of the Gαi:β-arrestin complexes was longer than that of the V2R:β-arrestin complexes (table S1). This Gαi:β-arrestin complex appears to be distinct from the previously described Gαs:β-arrestin:GPCR complex (4).

Multiple GPCRs promote the formation of Gαi:β-arrestin complexes

Given the paradoxical results of the Gαs-coupled V2R catalyzing a Gαi:β-arrestin complex, we investigated whether this phenomenon was generalizable to other GPCRs. We selected five well-studied GPCRs [β2AR, CXCR3, neurotensin 1 receptor (NTS1R), and dopamine receptors D1 (D1R) and D2 (D2R)] to compare the formation of Gαi:β-arrestin complexes with that of β-arrestin and their canonically coupled Gα subtype (or Gαs in the case of canonically Gαi-coupled GPCRs) when overexpressed in HEK 293T cells. We confirmed the canonical G protein coupling of these GPCRs (Fig. 4, A to E). Only CXCR3 and D2R canonically signaled through Gαi, with β2AR and D1R primarily coupling to Gαs and NTS1R primarily to Gαq. All five of these GPCRs formed Gαi:β-arrestin complexes in response to agonist treatment, with no discernable complexes between the canonical Gα subunit for each receptor and β-arrestin (Fig. 4, F to J). Thus, the formation of Gαi:β-arrestin complexes is a common mechanism across GPCRs.

Fig. 4 GPCRs form Gαi:β-arrestin complexes irrespective of canonical G protein coupling.

(A to E) Canonical G protein signaling at (A) β2AR, (B) CXCR3, (C) D1R, (D) D2R, and (E) NTS1R assessed with the TGF-α shedding assay. (F to J) Gαi:β-arrestin complex formation at (F) β2AR (10 μM isoproterenol), (G) CXCR3 (1 μM VUF10661), (H) D1R (500 nM dopamine), (I) D2R (500 nM dopamine), and (J) NTS1R (10 nM neurotensin). *P < 0.05 by two-way ANOVA, main effect of Gαi subtype. For (A) to (E), n = 4. For (F), n = 3 to 6; for (G), n = 3 to 4; for (H), n = 4; for (I), n = 3 to 4; for (J), n = 3 replicates per condition. Graphs show mean + SEM. Iso, isoproterenol.

Gαi:β-arrestin complexes facilitate ERK scaffolding and signaling

We tested the potential functional consequences of Gαi:β-arrestin complexes by assessing whether they could scaffold ERK1 and ERK2, kinases that function in the regulation of cell proliferation, cell survival, and apoptotic signaling (27, 28). GPCRs are thought to regulate ERK activation through phosphorylation by separate G protein and β-arrestin signaling pathways (29). β-arrestins regulate ERK signaling; however, CRISPR-Cas9 genome-editing approaches have demonstrated a lack of ERK signaling in the collective absence of functional G proteins, leading some to conclude that β-arrestin signaling might be dispensable for ERK activation (30, 31). However, other studies using CRISPR-Cas9 demonstrate an essential role for β-arrestin in some pathways regulating ERK activation (32). For Gαi-coupled receptors, β-arrestin–mediated ERK activation is PTX sensitive (3335). However, how the Gαi and β-arrestin transducers cooperate in this pathway is unclear.

To determine whether Gαi:β-arrestin complexes could bind ERK in response to activated V2R, we tagged ERK at either its N or C terminus with the dipole acceptor mKO (Fig. 5A). Data were normalized to an untagged (cytosolic) mKO to account for changes in protein localization. Agonist treatment of V2R catalyzed the formation of a Gαi:β-arrestin:ERK complex (Fig. 5B). The magnitude of the adjusted complex BRET ratio was dependent on the location of the mKO tag on ERK (ERK-mKO compared with mKO-ERK), consistent with the orientation and distance dependence of resonance energy transfer between the luciferase donor and mKO dipole acceptor. To selectively evaluate the contributions of Gαi signaling on ERK phosphorylation, we used triple Gα subfamily knockout (Δ3G) HEK 293T cells depleted by CRISPR-Cas9 technology of three Gα protein subfamilies, Gαs/Gαolf, Gαq/11, and Gα12/13 (only Gαi members present) (fig. S14) (30). Agonist treatment of these Δ3G HEK 293T cells overexpressing V2R increased phosphorylation of ERK (Fig. 5, C and D). Pretreatment with PTX decreased, but did not eliminate, ERK phosphorylation in Δ3G cells or in cells lacking the three Gα proteins (Fig. 5, C and D). In these cells, PTX pretreatment in combination with β-arrestin-1 and β-arrestin-2 small interfering RNA (siRNA) knockdown had a slightly greater effect on ERK phosphorylation (Fig. 5, C and D), consistent with possible functional coordination of Gαi and β-arrestin. Similar effects of PTX pretreatment on ERK phosphorylation were observed in Δ3G cells transiently overexpressing β2AR, D1R, or NTS1R (fig. S15). To support these findings, we used HEK 293T cells lacking all Gα proteins and restored Gαi expression, increased expression of β-arrestin, or both. Increasing expression of both proteins, but not that of either protein alone, significantly increased ERK phosphorylation after AVP treatment in these cells in which all Gα proteins had been suppressed (Fig. 5, E and F). These results are consistent with a role of Gαi:β-arrestin complexes in promoting ERK phosphorylation, although they do not preclude other Gα- or β-arrestin–mediated mechanisms for ERK activation.

Fig. 5 Gαi:β-arrestin scaffolds form functional complexes with ERK.

(A) Arrangement of split luciferase components and mKO on Gα protein (LgBiT), β-arrestin (smBiT), or ERK (mKO). (B) BRET of Gαi-LgBiT, β-arrestin-smBiT, and ERK2-mKO in cells overexpressing untagged V2R treated with AVP (500 nM). (C) Representative immunoblot of phospho and total ERK1/2 in Δ3G HEK 293T cells pretreated with PTX (200 ng/ml) and/or βarr1/2 siRNA stimulated with either vehicle or AVP (500 nM). (D) Quantification of ERK immunoblots in Δ3G HEK 293T cells. “100%” represents the maximal signal of phospho ERK/total ERK in the control, vehicle-treated condition. (E) Representative immunoblot of β-arrestin, Gαi, phospho-ERK, and total ERK1/2 in Δ4G HEK 293T cells (lacking all G proteins) transfected with the indicated amounts of Gαi and/or βarr1/2. “100%” represents the signal of phospho ERK/total ERK in the control siRNA, absent PTX, AVP-stimulated condition. (F) Quantification of ERK immunoblots in Δ4G HEK 293T cells. *P< 0.05, ***P < 0.001, two-way ANOVA with Bonferroni post hoc compared with the no-treatment, control siRNA group. The net BRET ratio of cytosolic mKO control was subtracted from the net BRET ratio of ERK-mKO to yield an adjusted BRET ratio that is the ordinate of (B). Immunoblots are representative of four to five separate experiments. For (B), n = 5; for (D), n = 4; and for (F), n = 5 replicates per condition. NT, no transfection of either Gαi or β-arrestin. Graphs show mean ± SEM.

A β-arrestin–biased angiotensin type 1 receptor agonist promotes formation of the Gαi:β-arrestin complex and promotes PTX-sensitive cell migration

To assess whether Gαi:β-arrestin complexes also occur in the absence of canonical G protein activation, we used the angiotensin type 1 receptor (AT1R) β-arrestin–biased agonist TRV120023, which does not promote canonical G protein signaling but does cause recruitment of β-arrestin (36, 37) (Fig. 6A). The endogenous ligand of AT1R, angiotensin II (AngII), signals through both Gαq and Gαi (Fig. 6B) and recruits β-arrestin (Fig. 6A). In our assays, we verified that TRV120023 is a β-arrestin–biased agonist and that it had no little or no activity at promoting canonical G protein signaling through any of the four Gα-family proteins tested (Fig. 6C) but did stimulate β-arrestin recruitment to the receptor (Fig. 6A). Because TRV120023 does not activate canonical G protein signaling, it would be predicted not to induce cell migration, which is thought to require canonical G protein signaling (38). However, not only did TRV120023 promote migration, this migration was also PTX sensitive, because pretreatment of cells with PTX reduced TRV120023-mediated migration by ~50%. Furthermore, inhibition of both Gαi and β-arrestin through PTX pretreatment and siRNA–mediated depletion of β-arrestin-1 and β-arrestin-2 eliminated migration in HEK 293T cells stably expressing AT1R (Fig. 6, D and E). Migration of primary human pulmonary arterial smooth muscle cells (PASMCs) in response to TRV120023 was also PTX sensitive, indicating that this observation was not unique to immortalized model cell lines, and this mechanism can occur with physiological amounts of receptor and effector expression (fig. S16, A to D). Because treatment of the AT1R with either the endogenous agonist AngII or the β-arrestin–biased agonist TRV120023 induced Gαi:β-arrestin complex formation when overexpressed in HEK 293T cells (Fig. 6F), the migration findings are consistent with the possibility that Gαi:β-arrestin complexes can promote functional responses after GPCR activation.

Fig. 6 Cell migration in response to the β-arrestin–biased ligand TRV120023 requires both Gαi and β-arrestins.

(A) BRET assay quantifying recruitment of β-arrestin-2-YFP to AT1R-RlucII after treatment with either AngII or TRV120023. (B and C) Canonical G protein signaling through the TGF-α shedding assay at the AT1R after treatment with either the endogenous ligand AngII (B) or TRV120023 (TRV023) (C). (D) Representative images of the four TRV120023 migration conditions in HEK 293T cells stably expressing ATIR. (E) Quantification of PTX pretreatment (200 ng/ml) and/or βarr1/2 siRNA on TRV120023-induced migration for the experiment shown in (D). (F) Split luciferase assay for monitoring Gαi-LgBiT:β-arrestin-smBiT association after treatment of AT1R with either AngII or TRV120023. (G) Cartoon schematic of three described GPCR transduction effectors. *P < 0.05, two-way ANOVA with Bonferroni post hoc test compared with the no-treatment, control siRNA group. #P < 0.05, two-way ANOVA with Bonferroni post hoc test compared with control siRNA, PTX-pretreated group. For (A), n = 4; for (B) and (C), n = 4 to 5; for (E) and (F), n = 4 replicates per condition. Graphs show mean ± SEM.


Our results reveal a GPCR signaling mechanism in which GPCRs can promote the formation of Gαi:β-arrestin complexes. These Gαi:β-arrestin complexes were observed downstream of all receptors tested, even receptors that do not canonically signal through Gαi. Various GPCR ligands drove formation of Gαi:β-arrestin complexes, even the β-arrestin–biased ligand TRV120023, which promotes little or no classical G protein–mediated signaling. A major determinant of β-arrestin association with Gαi appears to be GPCR-mediated recruitment of β-arrestin to the plasma membrane. Gαi:β-arrestin scaffolds can include a GPCR, a signaling effector (ERK), or possibly both, and may form functional signaling complexes. Such signaling complexes were associated with V2R-mediated ERK activation, even though the stimulatory GPCR ligand (AVP) does not activate canonical Gαi signaling. It remains to be determined whether Gαs is also a component of these Gαi:β-arrestin complexes formed in response to agonist stimulation of Gαs-coupled receptors. Using HEK 293T cells depleted of the Gαs/q/12 proteins and overexpressing the V2R, we demonstrated that AVP-induced ERK phosphorylation is decreased after Gαi inhibition with PTX and β-arrestin depletion with siRNA. Furthermore, expression of Gαi in cells lacking all functional G proteins combined with overexpression of β-arrestin, but not either β-arrestin overexpression or Gαi rescue alone, significantly increased AVP-induced ERK phosphorylation. PTX impaired migration of both primary human PASMCs and immortalized cells treated with a β-arrestin–biased AT1R ligand, TRV120023. It remains unclear exactly how Gαi:β-arrestin complexes participate in the process of ERK activation and cell migration (Fig. 6G).

β-arrestin signaling is necessary for full efficacy responses in many cellular and physiological activities (22). However, some signaling functions attributed to β-arrestin are absent when functional G proteins are eliminated (30). This study bridges these seemingly contradictory results concerning the interplay of G protein and β-arrestin signaling (31, 32) by delineating a Gαi:β-arrestin scaffolding complex. Furthermore, this study provides further validity to computational simulations and other experimental observations suggesting that the activated state of β-arrestin persists even when it is not bound to a receptor (39). Our results also offer plausible mechanistic insight into our initially paradoxical observations that Gαi can drive ERK phosphorylation downstream of the canonically Gαs-coupled V2R, and that PTX inhibits cell migration in response to a β-arrestin–biased agonist.

Several caveats must be considered when interpreting our results. Many experiments rely on the overexpression of components that have been genetically modified by insertion of various fluorescent reporter probes. Probe architecture can affect the intensity of the signal generated. Because of overexpression, interactions may be detected that would not be seen at physiologically relevant concentrations of these molecules. However, our control experiments, our observation that the lowest concentrations of expression vector provide the highest signal-to-noise ratio for Gαi:β-arrestin complex formation (fig. S4), and our findings in primary human PASMCs support our interpretation that this complex formation is neither an artifact of probe orientation nor dependent on overexpressed conditions. The presence of Gαi:β-arrestin association in orthogonal assays (TSA, immunoprecipitation in both purified and nonpurified conditions, and nanoBRET) provides further support for the existence of Gαi:β-arrestin complexes.

In summary, our findings demonstrate that GPCRs promote the formation of Gαi:β-arrestin complexes, and that these complexes appear necessary for certain aspects of β-arrestin signaling. A better understanding of their physiological roles may provide additional strategies for therapeutic targeting of GPCRs.

Materials and methods

Cell culture and transfection

HEK cells [HEK 293, HEK 293T, Rockman β-arrestin-1/2 HEK 293 knockout, Δ3G HEK 293T cells, and quadruple Gα subfamily knockout (Δ4G) HEK 293T cells] were maintained in minimum essential medium supplemented with 1% antibiotic-antimycotic (Gibco) and 10% fetal bovine serum. Rockman β-arrestin-1/2 HEK 293 knockout cells were supplied by Dr. Howard Rockman and validated as previously described (32). Cells were grown at 37°C with a humidified atmosphere of 5% CO2. For BRET and luminescence studies, HEK 293T cells were transiently transfected with an optimized calcium phosphate protocol as previously described (20). For immunoblot studies using siRNA, Δ3G, or Δ4G HEK 293T cells were transiently transfected with Lipofectamine 3000 (Thermo Fisher) or RNAiMAX (Thermo Fisher) according to the manufacturer’s specifications. For TGF-αalpha shedding assay studies, Δ3G HEK 293T cells were transfected using Fugene 6 (Promega) or Lipofectamine 2000 (Thermo Fisher) according to the manufacturer’s specifications.

Generation of constructs

Construct cloning was performed using conventional techniques such as restriction enzyme and ligation methods. Linkers between the fluorescent proteins or luciferases and the cDNAs for receptors, transducers, kinases, or adaptor proteins were flexible (GGGGS) and ranged between 15 and 18 amino acids. CXCR3 C terminus phosphomutant constructs were generated using a Quikchange mutagenesis kit (Agilent Technologies, Santa Clara, CA). See table S2 for a list of constructs used in this manuscript.


VUF10661, AVP, dopamine, AngII, neurotensin, and isoproterenol were purchased from Sigma-Aldrich (St. Louis, MO). VUF10661 and isoproterenol were dissolved in dimethyl sulfoxide (DMSO) and stored in a desiccator cabinet. Stock solutions of AVP, AngII, and neurotensin (Sigma-Aldrich) were prepared according to the manufacturer’s specifications. TRV120023 was synthesized by Genscript (Piscataway, NJ). Stock solutions of neurotensin were made in 0.1% bovine serum albumin in phosphate-buffered saline (PBS). Dopamine was prepared fresh in BRET medium supplemented with 0.03% ascorbic acid (Sigma-Aldrich). H3192 and SR121463 cells were kindly provided by R. J. Lefkowitz. All drug dilutions were performed with BRET medium or cell culture medium. PTX was obtained from List Biological Laboratories (Campbell, CA). All compound stocks were stored at −20°C until use.

Split luciferase and BRET assays

HEK 293T cells seeded in six-well plates were cotransfected with 500 ng of smBiT-tagged β-arrestin-2, and either 250 ng of LgBiT-tagged receptor or 2000 ng of untagged receptor and varying concentrations of LgBiT Gα protein expression vector (most experiments were conducted with between 50 and 200 ng of Gα plasmid) or 2000 ng of mKO-tagged β-arrestin-2 and 500 ng of smBiT-tagged V2R using a previously described calcium phosphate protocol (40). Twenty-four hours after transfection, cells were plated onto clear-bottomed, white-walled, 96-well plates at 50,000 to 100,000 cells/well in BRET medium [clear minimum essential medium (Gibco) supplemented with 2% fetal bovine serum, 10 mM Hepes, 1× GlutaMax, and 1× Antibiotic-Antimycotic (Gibco)]. Select cells were then treated overnight with PTX at a final concentration of 200 ng/ml. The following day, media were removed and cells were incubated at room temperature with 80 μl of Hanks’ balanced salt solution (Gibco) supplemented with 20 mM Hepes and 3 μM coelenterazine-h (Goldbio or Promega) for ~15 min. For luminescence split luciferase studies, plates were read with a BioTek Synergy Neo2 plate reader set at 37°C with a 485-nm emission filter. Cells were stimulated with either vehicle [Hank’s balanced salt solution (HBSS) with 20 mM Hepes] or the indicated concentration of agonist. For split luciferase luminescence experiments, plates were read both before and after ligand treatment to calculate ∆net change in luminescence and subsequently normalized to vehicle treatment. For BRET experiments, plates were read on a Berthold Mithras LB940 using prewarmed medium and instrument at 37°C using a standard Rluc emissions filter (480 nm) with a custom mKO 542-nm long-pass emission filter (Chroma Technology Co., Bellows Falls, VT). Readings were performed using a kinetic protocol. The BRET ratio was calculated by dividing the mKO signal by the luciferase signal. For some experiments, a net BRET ratio was calculated by subtracting the vehicle BRET ratio from the ligand-stimulated BRET ratio, or an adjusted BRET ratio was calculated by subtracting the ligand-treated cytosolic mKO signal from the ligand-treated effector mKO signal.


Experiments were conducted as previously described (40). Briefly, cells were serum starved for at least 4 hours, stimulated with the indicated ligand, washed once with ice-cold PBS, lysed in ice-cold radioimmunoprecipitation assay buffer containing phosphatase and protease inhibitors [Phos-STOP (Roche) and cOmplete EDTA free (Sigma-Aldrich)], rotated for 45 min, and cleared of insoluble debris by centrifugation at >12,000g (4°C, 15 min), after which the supernatant was collected. Protein was resolved on SDS-10% polyacrylamide gels, transferred to nitrocellulose membranes, and immunoblotted with the indicated primary antibody overnight at 4°C. phospho-ERK (Cell Signaling Technology, #9106) and total ERK (Millipore, #06-182) were used to assess ERK activation. A1-CT antibody that recognizes both isoforms of β-arrestin was used (32), with protein loading assessed by alpha-tubulin (Sigma-Aldrich, #T6074). Galpha i-1 (Santa Cruz Biotechnology, #13533), Galpha q/11/14 (Santa Cruz Biotechnology, #365906), Galpha 12 (Santa Cruz Biotechnology, #515445), Galpha 13 (Santa Cruz Biotechnology, #293424), and Galpha s/olf (Santa Cruz Biotechnology, #55545) antibodies were used to verify Δ3G HEK 293 cells.

Horseradish peroxidase–conjugated polyclonal mouse anti-rabbit immunoglobulin (IgG) IgG or anti-mouse IgG were used as secondary antibodies. Immune complexes on nitrocellulose membrane were imaged by SuperSignal enhanced chemiluminescent substrate (Thermo Fisher). After detection of phospho signal, nitrocellulose membranes were stripped and reblotted for total kinase signal. For quantification, phospho-protein signal was normalized to total protein signal using ImageLab (Bio-Rad) within the same immunoblot.

siRNA knockdown

Δ3G HEK 293T cells were transiently transfected with either Lipofectamine 3000 (Thermo Fisher) or RNAiMAX (Thermo Fisher) per manufacturer specifications in a six-well tissue culture sterile plate with 1 μg of receptor and either 7 μg of either control siRNA or 3.5 μg of siRNA directed to β-arrestin-1 and β-arrestin-2 sequences (Lipofectamine 3000 conditions) or 1000 ng of receptor and 400 ng of β-arrestin-1 and β-arrestin-2 siRNA or 800 ng of control siRNA (RNAiMAX conditions) similarly as previously described (32). Twenty-four hours later, the cells were transiently transfected with 1000 ng of receptor using Lipofectamine 3000 (Thermo Fisher) and left to incubate for an additional 48 hours.

Wound-healing migration assay

Immortalized cells

HEK 293T cells stably expressing the AT1R were used (41). Briefly, 70 μl of cell suspension at a concentration of 5 × 105 cells/ml was applied into each well of silicone inserts (Ibidi, Martinsried, Germany) on a 24-well plate, and after 24 hours of incubation, the inserts were removed to create a wound field. The cells were incubated for an additional 12 hours with 1 μM TRV120023 and visualized with a Zeiss Axio Observer microscope (Carl Zeiss, Thornwood, NY). Wound healing was then analyzed using mageJ (NIH, Bethesda, MD) wound-healing tool macros.

Primary human cells

Primary human PASMCs were isolated from donor lungs as previously described (42) (gift of Suzy Comhair and Serpil Erzurum, Cleveland Clinic Foundation) and seeded into 24-well plates at a density of 1.5 × 104 cells per well in human smooth muscle cell (SMC) basal medium supplemented with human SMC growth supplement (both Cell Application, Inc.). At 30% confluence, cells were transfected with 50 ng of control or β-arrestin-1 and β-arrestin-2 ON-TARGETplus siRNA pools (Dharmacon, Inc.). When the cells reached 100% confluency, they were serum starved for 4 hours with 2 hours of PTX (200 ng/ml) treatment. Scratches were made using small pipette tips. Cells were washed using Dulbecco’s PBS and stimulated with basal SMC medium containing 1 μM AngII (Millipore Sigma) or 10 μM TRV120023 (Genscript USA Inc.). The wound closure was monitored using a live-cell station Zeiss Axio Observer microscope (Carl Zeiss, Thornwood, NY). The images were captured in real time at 0 hours and for every hour up to 16 hours. The initial edges of the scratch at the 0-hour time point were marked, and the migrated distance for 12 to 16 hours afterwards was measured using MetaMorph Premier software (Molecular Devices, San Jose, CA) at the Duke Light Microscopy Core Facility (Durham, NC).

TGF-α shedding assay

GPCR Gα activity was assessed by the TGF-α shedding assay as previously described (16). Briefly, HEK 293 cells lacking Gαq, Gα11, Gαs/olf, and Gα12/13 were transiently transfected with receptor, modified TGF-α–containing alkaline phosphatase (AP-TGF-α), and the indicated Gα subunit. Cells were reseeded 24 hours later in HBSS (Gibco, Gaithersburg, MD) supplemented with 5 mM Hepes in a Costar 96-well plate (Corning Inc., Corning, NY). Cells were then stimulated with the indicated concentration of ligand for 1hour. Conditioned medium (CM) containing the shed AP-TGF-α was transferred to a new 96-well plate. Both the cell and CM plates were treated with para-nitrophenylphosphate (p-NPP, 100 mM; Sigma-Aldrich, St. Louis, MO) substrate for 1 hour; p-NPP is converted to para-nitrophenol (p-NP) by AP-TGF-α. This activity was measured at an optical density at 405 nm (OD405) in a Synergy Neo2 Hybrid Multi-Mode (BioTek, Winooski, VT) plate reader immediately after p-NPP addition and then after a 1-hour incubation. Gα activity was calculated by first determining p-NP amounts by absorbance using the following equation:100*(ΔOD405, CMΔOD405, CM+ΔOD405, cell)where ΔOD405 = OD405, 1hr – OD405, 0 hours and ΔOD405, cell and ΔOD405, CM represent the changes in absorbance after1 hour in the cell and CM plates, respectively. Data were normalized to a single well that produced the maximal signal.


Protein thermal melting shift experiments were performed using the StepOnePlus Real-Time PCR System (Applied Biosystems). Proteins were buffered in 20 mM Hepes, pH 7.5, 100 mM NaCl, and 4 mM MgCl2. β-arrestin-2, Gαiβγ, V2Rpp, Fab30, and GTPγS (a nonhydrolyzable GTP analog) were added at a final concentration of 5, 10, 30, and 120 μM, respectively. All reactions were set up in a 96-well plate at final volumes of 20 μl, and SYPRO Orange (Thermo Fisher Scientific) was added as a probe at a dilution of 1:1000. Excitation and emission filters for the SYPRO-Orange dye were set to 475 and 580 nm, respectively. The temperature was raised at 0.5°C increments every 30 s from 25°C to 99°C, and fluorescence readings were taken at each interval. All measurements were performed three times. Data were analyzed using Applied Biosystems Protein Thermal Shift Software. Expression and purification of heterotrimeric G protein was conducted as previously described (17). In brief, Hive Five insect cells were infected with two viruses made from BestBac baculovirus system, one expressing human Gβ1-His6 and Gγ2 and the other Gαi1. Approximately 48 hours after infection, the cells were harvested, solubilized, and heterotrimeric Gαi was purified using Ni–nitrilotriacetic acid chromatography and HiTrap Q Sepharose anion exchange (GE Healthcare Life Sciences).

Purified component pull-down

Expression and purification of β-arrestin-2 and heterotrimeric Gαi were purified similarly to previously described (43). Briefly, purified glutathione S-transferase–tagged β-arrestin-2 and purified heterotrimeric G protein were added together at a final concentration of 4 and 1 μM, respectively, and placed on a tube revolver at room temperature for 30 min. Once complete, the solution was added to 40 μl of washed glutathione agarose beads and placed on a tube rotator at room temperature for 60 min. After removal of the supernatant, samples were washed three times with 300 μl of 20-mM Hepes, pH 7.5, 100 mM NaCl. The sample was then eluted with 200 μl of 2× SDS-sample buffer (8% SDS, 5% β-mercaptoethanol, 10% glycerol and 25 mM Tris pH 6.8). Protein was resolved on SDS-10% polyacrylamide gels, transferred to nitrocellulose membranes, and immunoblotted with antibodies directed to 400 ng of either β-arrestin-1/2 (A1-CT, from the laboratory of R. J. Lefkowitz) or Gαi1 (Santa Cruz Biotechnology, #13533).


Immunoprecipitation was conducted as previously described (44). Briefly, 4 μg of HA-V2R, 4 μg of Gαi–green fluorescent protein (GFP), and 4 μg of pcDNA-ARRB1-Flag and/or pcDNA were transfected into HEK 293T cells seeded in 6-cm plates. Forty-eight hours after transfection, after ~4 hours of serum starvation, cells were stimulated with AVP for 5 and 10 min at 37°C. Cells were then lysed on ice for 10 min in FLAG lysis buffer (50 mM Tris-HCl, pH 7.4, 1% Triton X-100, 150 mM NaCl, 1 mM EDTA) supplemented with protease inhibitor cocktail tablet (Roche). Cell lysates were incubated with anti-FLAG M2 affinity gel (Sigma-Aldrich, #A2220), overnight and immunoprecipitated ARRB1-FLAG was eluted with Flag peptides (Sigma-Aldrich, #F3290). For primary antibody incubation, GFP polyclonal antibody (Invitrogen, #A6455), HA-Tag (Cell Signaling Technology, #3724S), and ANTI-FLAG M2 antibody (Sigma-Aldrich, #F3165) were used.

Confocal microscopy

HEK293T cells plated in fibronectin-coated, 35-mm, glass-bottomed dishes (MatTek Corp., #P35G-0-14-C) were transiently transfected using the calcium phosphate protocol with DNA encoding Gαi-mVenus (125 ng), βarr2-mKO (125 ng), and/or Mars1-V2R (500 ng). Mars1 binds a membrane-impermeable fluorogen (SCi1) and induces its fluorescence in the near-infrared spectrum (45). Cells were pulse labeled with SCi1 (diluted 1:5000 from 0.5 mg/ml stock) for 15 min before treatment with or without 100 nM AVP. Cells were fixed at basal, 5 min, and 30 min after treatment with 4% paraformaldehyde. Samples were imaged with a Plan-Apochromat 63×/1.4 oil lens on a Zeiss LSM880 confocal laser scanning microscope using the corresponding laser lines to excite mVenus, mKO, or Mars1 (488, 561, and 633 nm, respectively). Spectral gating with a 34–spectral array detector was performed using single-color transfection controls.

Image analysis

Confocal images were arranged and analyzed using ImageJ (NIH, Bethesda, MD). All image adjustments were identical and consistent across all samples and were based on the single-color control samples. The line scan analysis function was used to measure the intensity of each channel.

Graphing and statistical analyses

Data were analyzed in Excel (Microsoft, Redmond, WA) and graphed in Prism 8.0 (GraphPad, San Diego, CA). Dose-response curves were fitted to a log agonist versus stimulus with three parameters (span, baseline, and EC50), with the minimum baseline corrected to zero using Prism 8.0. Statistical analysis was also performed in Prism 8.0. For comparing ligands in concentration-response assays or time-response assays, a two-way analysis of variance (ANOVA) of ligand and concentration or ligand and time, respectively, was conducted. For sample size determination, most experiments were conducted three times, with some additional replicates acting as positive or negative controls for other experiments. Experiments were not randomized, and investigators were not blinded to treatment conditions. Critical plate-based experiments were independently replicated by at least two different investigators when feasible. For some analyses, a post hoc comparison was also made, with details included in the figure legends. Replicates in the figure legends refer to distinct cellular conditions (biological replicates); certain experiments also included technical replicates to show intrareplicate variation. Unless otherwise noted, statistical tests were two-sided, and Bonferroni analyses were corrected for multiple comparisons. Further details of statistical analysis and replicates are included in the figure legends.

Supplementary Materials

Figs. S1 to S16

Tables S1 and S2

MDAR Reproducibility Checklist

References and Notes

Acknowledgments: We thank R. Lefkowitz, L. Luttrell, S. Shenoy, C. Chavkin, and J. Silverman for helpful discussion and thoughtful feedback; C. Ray, M. Orlen, P. Alagesan, J. Gundry, P. Nicholls, and N. Knape for technical assistance; N. Nazo for administrative assistance; R. Lefkowitz, S. Shenoy, and N. Freedman for the use of laboratory equipment; H. Rockman for Rockman β-arrestin-1/2 KO HEK 293T cells; and R. Lefkowitz for constructs outlined in the methods section. Cartoon images were created with Funding: This work was supported by T32GM7171 (J.S.S.), the Duke Medical Scientist Training Program (J.S.S.), F31DA041160 (T.F.P.), PRIME JP17 g5910013 (A.I.), the LEAP JP17 g0010004 from the Japan Agency for Medical Research and Development (A.I.), the JSPS KAKENHI (A.I.), R37MH073853 (M.G.C.), R01GM122798 (S.R.), K08HL114643 (S.R.), and a Burroughs Wellcome Career Award for Medical Scientists (S.R.). J.C.S. is funded by NCI 1K22CA212058. Author contributions: J.S.S. and T.F.P. conceived of the study and designed, generated, and validated receptor and β-arrestin split luciferase and mKO constructs. A.I. designed, generated, and validated all G protein split luciferase constructs and generated the Δ3G and Δ4G cell lines. J.S.S., T.F.P., C.L., K.Z., I.C., D.S.E, X.X., Z.M., G.V., A.W., and I.M.L. performed cell-based and biochemical experiments. X.X. and G.V. performed and analyzed migration assays. A.W.K. performed and analyzed TSA experiments. D.P.S. contributed purified protein for TSA and pull-down experiments. L.K.R. and J.C.S. performed and analyzed confocal experiments. J.S.S., T.F.P., K.Z., I.C., and S.R. analyzed all other data. J.S.S., T.F.P., M.G.C., and S.R. wrote the paper. All authors discussed the results and commented on the manuscript. Competing interests: A patent application for the BRET technique described in this manuscript has been filed on behalf of J.S.S., T.F.P., M.G.C., and S.R. by Duke University. The remaining authors declare no competing interests. Data and materials availability: All data are available in the manuscript or the supplementary materials. Data and materials will be fully available upon reasonable request. CXCR3, AT1R-RLucII, V2R-LgBiT, V2R-smBiT, β2AR-LgBiT, smBiT-βarr2, smBiT-βarr1, V2R-mKO, β2AR-mKO, βarr2-mKO, ERK2-mKO, mKO-ERK2, and Gαi1-mVenus constructs (see table S2) are available from S.R. under a materials transfer agreement with Duke University. The G-protein-KO HEK293 cells and the NanoBiT-G-protein constructs are available from A.I. under a materials transfer agreement with Tohoku University.

Stay Connected to Science

Navigate This Article