To measure the activation of the four families of G protein isoforms at the plasma membrane and early endosomes by the V₂R in real-time, we used mG proteins. The mG proteins are homogenously distributed in the cytosol under basal condition but translocate to the subcellular location of GPCRs upon stimulation^17,18,21. In addition, we applied an EbBRET approach instead of a conventional bioluminescence resonance energy transfer (BRET)-based assay to monitor mG protein trafficking. EbBRET displays superior robustness and sensitivity, as well as higher dynamic spectrometric energy transfer signals associated to EbBRET as compared to conventional BRET, which is why this approach was favored¹⁹. We fused mG proteins to the luciferase from Renilla reniformis (Rluc) and anchored green fluorescent protein from the same species (rGFP) to polybasic sequence and prenylation CAAX box of KRas (rGFP-CAAX), which is located at the plasma membrane²², or to the early endosome marker Rab5²³ (Fig. 1a,b, left panels). Four variants of mG proteins (mGs, mGsi, mGsq, and mG12) have been designed and shown to maintain the receptor-Gα protein specificity of the four Gα subunit isoform families¹⁸. In HEK293 cells expressing rGFP-CAAX, V₂R, and similar levels of Rluc-fused mG proteins (Supplementary Fig. 1a), arginine vasopressin (AVP) treatment induced a rapid recruitment of mGs and mGsq but not mGsi nor mG12 to the plasma membrane. Maximal recruitment of the mGs and mGsq were reached ∼10 minutes after initial stimulation (Fig. 1a, right panel). These results suggest that the V₂R activates both Gα_s and Gα_q at the plasma membrane. In addition, AVP stimulation led to the recruitment of the same mG protein isoforms to early endosomes in cells expressing rGFP-Rab5, V₂R, and similar levels of Rluc-fused mG proteins (Fig. 1b, right panel and Supplementary Fig. 1b). In contrast to the plasma membrane response, mGs and mGsq recruitment to early endosomes were slower and reached maximal levels 45-60 minutes after initial stimulation with AVP (Fig. 1b, right panel).

To visualize V₂R-mediated activation of Gα_s and Gα_q at the plasma membrane and early endosomes we used confocal microscopy. For this purpose, we transfected HEK293 cells with mGs or mGsq fused to a HaloTag (Halo-mGs and Halo-mGsq) along with the respective red fluorescent protein (RFP)-fused plasma membrane or early endosomes markers Lck²⁴ or early endosome antigen 1 (EEA1)²⁵. Upon HaloTag labelling with a fluorescent green ligand, mGs and mGsq were visible and homogenously distributed in the cytosol under basal condition (vehicle) (Fig. 1c). In contrast, in cells treated with AVP for 10 minutes, both mGs and mGsq were redistributed along the periphery of the cells where they colocalized with RFP-Lck (Fig. 1c,d). These observations confirm that Gα_s and Gα_q are activated by the V₂R at the plasma membrane. In cells expressing the early endosomal marker RFP-EEA1, robust colocalization between the Halo-mG proteins and RFP-EEA1 were found 45 min after initial AVP-stimulation but not by vehicle treatment (Fig. 1e,f). Together our EbBRET and confocal microscopy imaging data suggest that Gα_s and Gα_q are activated by V₂R first at plasma membrane, and later on, from early endosomes after the V₂R has been internalized.

G protein activation from endosomes by some GPCRs is associated with the ability of the receptor to recruit G protein and βarr simultaneously to form a GPCR-βarr-G protein megaplex. As we already previously demonstrated that the V₂R forms V₂R-βarr-G_s megaplexes upon AVP stimulation⁸, we here explored whether formation of such complexes potentially can be formed with G_q/11 as well. In addition to the V₂R, we also applied a chimeric V₂R harboring the carboxy-terminal tail of the β₂-adrenergic receptor (β₂AR) referred to as V₂β₂AR. We previously showed that the phosphorylated V₂R carboxy-tail forms stable complexes with βarr, a requirement of megaplex formation, whereas the carboxy-tail of the β₂AR does not¹⁰. Therefore, we expected that only the V₂R, but not the V₂β₂AR, recruits G proteins and βarrs simultaneously upon agonist challenge.

Both the V₂R and V₂β₂AR bind to AVP with similar affinities and activate adenylyl cyclase via Gα_s with similar potencies²⁶. We monitored activation of the four Gα protein families at the plasma membrane by the V₂β₂AR upon AVP treatment using the same approach utilized in Fig. 1a. Similarly to the V₂R, the V₂β₂AR activated both Gα_s and Gα_q, but not Gα_i or Gα₁₂, at plasma membrane with a maximal response reached after ∼10 minutes of stimulation (Fig. 2a, and Supplementary Fig. 2). While the V₂R and V₂β₂AR are both reported to internalize via a βarr-dependent mechanism, βarr has been reported to rapidly dissociate from the V₂β₂AR shortly after its recruitment to the plasma membrane due to its low affinity for this receptor chimera²⁶. In contrast, βarr stays associated with the V₂R during its internalization into endosomes owning to its high affinity for the V₂R²⁶. Here, we compared the kinetics of βarr1 and βarr2 recruitment to the V₂R and V₂β₂AR at the plasma membrane and early endosomes by monitoring AVP-promoted EbBRET between Rluc-fused βarrs and rGFP-CAAX or rGFP-Rab5, respectively (Fig. 2b, left panel). At similar levels of receptor and βarr expressions (Supplementary Fig. 3a), both receptors recruited βarr1 and βarr2 at plasma membrane maximally after 10 minutes of stimulation with AVP (Fig. 2b, right upper panel). However, the presence of βarrs at the plasma membrane declined rapidly hereafter 10 minutes in V₂R-expressing cells, while remaining for longer periods of time in V₂β₂AR-expressing cells. These findings are in line with the previous reported observations²⁶. Additionally, the translocation of βarrs to the plasma membrane was more robust for the V₂R as compared to the V₂β₂AR, which is reminiscent from the higher affinity of βarrs for the V₂R as compared to the β₂AR²⁷. In contrast to the rapid translocation of βarrs to the plasma membrane, AVP treatment induced a robust but slower recruitment of βarrs to early endosomes with V₂R reaching a maximal response after approximately 45 minutes of stimulation (Fig. 2b, right bottom panel). Importantly, as opposed to V₂R, AVP-stimulation of V₂β₂AR did not result in βarr translocation to early endosomes (Fig. 2b, right bottom panel, and Supplementary Fig. 3b). Consequently, the V₂β₂AR represents a valuable negative control to investigate the ability to recruit G proteins and βarrs simultaneously at endosomes.

To track the simultaneous coupling of G proteins and βarrs to GPCRs in real-time, a nanoBiT approach was used. Both mGs and mGsq were fused to the large portion of nanoluciferase (large-BiT; LgBiT) and βarr1 to an optimized small peptide BiT (small BiT; SmBiT). Reconstitution of the complete and functional nanoluciferase, which catalyzes the conversion of coelenterazine h, results in emission of a bright luminescence signal. In our setup, close proximity of LgBiT and SmBiT only occurs when LgBiT-mG and SmBiT-βarr1 are recruited simultaneously to the receptor, which is a hall mark of megaplex formation (Fig. 2c, left panel). Using this approach, we detected bright luminescence signals involving mGs/βarr1 (Fig. 2c, upper right panel) and mGsq/βarr1 (Fig. 2c, bottom right panel) upon stimulation of the V₂R but not the V₂β₂AR. Interestingly, the dual coupling of Gα_q/βarr to V₂R appeared to be faster than the co-coupling of Gα_s/βarr. While 20 minutes was required to reach the maximal response of V₂R-stimulated mGs/βarr1 co-coupling, 8 minutes was sufficient to obtain the maximal levels of mGsq/βarr1 recruitment to the V₂R (Fig. 2c, right panels).

To visualize the simultaneous recruitment of G proteins and βarr by confocal microscopy, we transfected HEK293 cells with βarr2 fused to RFP (RFP-βarr2), Halo-mGs or Halo-mGsq, and the V₂R or V₂β₂AR. In vehicle-treated cells, both mGs and βarr2 were homogenously distributed in the cytosol (Fig. 2d). However, after prolonged stimulation of V₂R with AVP, around 75% of βarr2 colocalized with mGs in endocytic vesicles (Fig. 2d,f). In V₂β₂AR-stimulated cells, little to no colocalization was observed between βarr2 and mGs upon prolonged stimulation with AVP (Fig. 2d, f). Surprisingly, however, some clusters of intracellular mGs were clearly visible (Fig. 2d, f). These results suggest simultaneous coupling of Gα_s/βarr2 to the V₂R in endosomes but not to the V₂β₂AR. In cells expressing mGsq, both mGsq and βarr2 were also homogenously distributed in the cytosol when cells were treated with the vehicle in a similar fashion to cells expressing mGs (Fig. 2e). Upon AVP stimulation, approximately 75% of βarr2 colocalized with mGsq in intracellular vesicles in V₂R-expressing cells, whereas poor colocalization was observed in V₂β₂AR-expressing cells (Fig. 2e,f). However, similarly to cells expressing mGs, some clusters of intracellular mGsq were visible in cells expressing V₂β₂AR, suggesting a certain level of endosomal Gα_s/Gα_q signaling despite the absence of βarr2.

Activation of Gα_s and Gα_q by the V₂β₂AR from endosome-like structures in the absence of local βarr raises the possibility that the V₂R can activate these G proteins from endosomes in both βarr-dependent and -independent manners. To test this hypothesis, we compared AVP-induced Gα_s and Gα_q activation at plasma membrane and endosomes in CRISPR/Cas9-engineered βarr1- and βarr2-deficient HEK293 cells (ββarr1/2)¹⁹ as well as their parental cellular counterpart. The surface expression of V₂R was matched in both cellular backgrounds (Supplementary Fig. 4a,b). Using the EbBRET biosensors described in Fig. 1a-b, we performed AVP concentration-response characterization of Gα_s and Gα_q activation at the plasma membrane (Fig. 3a) and early endosomes (Fig. 3b) in parental and ββarr1/2 HEK293 cells. In contrast to Gα_s and Gα_q activation at the plasma membrane, which were not negatively affected by the absence of βarrs (Fig. 3a), we observed a robust decrease in the ability of the V₂R to activate Gα_s and Gα_q at endosomes in ββarr1/2 cells as compared to their parental counterpart (Fig. 3b and Supplementary Table 1). These data demonstrate the important role of βarrs in endosomal Gα_s/Gα_q/11 activation by the V₂R. However, although the ββarr1/2 cells do not express βarrs, we still observed significant residual G protein activation from endosomes. This surprising observation suggests that the V₂R internalizes into endosomes to some extent in a βarr-independent manner from where G proteins are stimulated. To probe this possibility, we compared V₂R internalization in parental and ββarr1/2 HEK293 cells expressing rGFP-CAAX and equivalent amounts of V₂R fused to Rluc at its carboxy-terminal tail (V₂R-Rluc) (Fig. 3c, left panel and Supplementary Fig. 4c). In parental HEK293 cells, AVP-stimulation of the V₂R-Rluc led to a robust decrease of EbBRET values, which indicates strong receptor internalization (Fig. 3c, right panel). Interestingly, we also observed significant internalization of the V₂R in ββarr1/2 HEK293 cells, although less than in the parental cells (Fig. 3c, right panel). These results suggest that a minor population of V₂R internalizes independently of βarrs and contributes to endosomal Gα_s and Gα_q/11 signaling.

Although our data suggest that a minor population of V₂R internalizes in the absence of βarrs and contribute to V₂R-mediated endosomal Gα_s signaling, it has been reported that βarr binding to the V₂R and parathyroid hormone receptor (PTHR) potentiates endosomal Gα_s signaling^7,16. To verify this and determine if this potentiator effect of βarrs also affects endosomal Gα_q/11 signaling, we compared endosomal Gα_s and Gα_q/11 activation in cells expressing similar levels of V₂R or V₂β₂AR (Supplementary Fig. 4d,e). Our rationale for using these two receptors is that if βarrs potentiate endosomal G protein activation, this potentiator effect will be observed to a greater extent for the V₂R since this receptor associates more robustly with βarrs as compared to the V₂β₂AR. Using the same biosensors as in Fig. 1a-b, we performed AVP dose-response curves of mGs and mGsq recruitment to the plasma membrane and early endosomes (Supplementary Fig. 5, and Supplementary Table 2). From the dose-response curves obtained (Supplementary Fig. 5, and Supplementary Table 2), we determined the transduction coefficient log(τ/Ka), a parameter that combines efficiency and potency to determine the overall G protein transduction, for each condition using the operational model of Kenakin and Christopoulos²⁸. In cells expressing mGs, the transduction coefficients of Gα_s activation at the plasma membrane were similar for the V₂R and V₂β₂AR, but higher for the V₂R than V₂β₂AR in early endosomes (Fig. 3d and Supplementary Table 2). Similarly, in cells expressing mGsq, the transduction coefficients of Gα_q/11 activation at the plasma membrane were similar for the V₂R and V₂β₂AR, but higher for the V₂R than V₂β₂AR in early endosomes (Fig. 3e and Supplementary Table 2). Altogether these results indicate that βarrs potentiate activation of G proteins by the V₂R in early endosomes.

HEK293 clonal cell line (HEK293SL cells) and referred as HEK293 cells as well as the HEK293 cells devoid of βarr1 and βarr2 referred as Δβarr1/2 cells were a gift from Stephane Laporte (McGill University, Montreal, Quebec, Canada) and previously described¹⁹. These cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) high glucose (Gibco) supplemented with 10% fetal bovine serum and 100 units per ml penicillin-streptomycin (Gibco), maintained at 37°C and 5% CO₂ and passaged every 3-4 days using trypsin-EDTA 0.05% (Gibco) to detach the cells. DNA to be transfected was combined with salmon sperm DNA (Invitrogen) to obtain a total of 1 μg DNA per condition. Linear polyethyleneimine 25K (PEI; Polysciences) was combined with DNA (3 μg PEI per μg of DNA), vortexed and incubated 20 minutes before adding a cell suspension containing 300,000 cells per ml (1.2 ml of cells per condition). The appropriate volume of cells containing the DNA was seeded and cells were incubated for 48 hours before assay.

All DNA constructs were cloned into pcDNA3.1(+) expression plasmid except if stated otherwise. V₂R and V₂β₂AR were tagged with a HA epitope in amino-terminal of the receptors. HA-V₂R was synthetized by GenScript and HA-V₂β₂AR was generously provided by Dr Robert Lefkowitz (Duke University, USA). C-tRFP-Lck (cloned into PCMV6-AC-RFP expression vector) and TagRFP-T-EEA1 (cloned into pEGFP-C1 vector) were purchased from Addgene (respectively #RC100049 and #42635). Strawberry-tagged βarr2 was a gift from Prof. Marc G. Caron (Duke University, USA). rGFP-CAAX¹⁹, rGFP-Rab5¹⁹, V₂R-Rluc¹⁹, Rluc-βarr1⁴⁷, Rluc-βarr2⁴⁸ were previously described. Rluc-mGs, Rluc-mGsi, Rluc-mGsq, and Rluc-mG12 were synthetized by Twist Bioscience and cloned into pTwistCMV expression vector. The Venus tag in NES-Venus-mGs, NES-Venus-mGsi, NES-Venus-mGsq, and NES-Venus-mG12 previously described¹⁸ was replaced by Rluc. Halo-mGsq was kindly provided by Prof. Nevin A. Lambert (Augusta University, USA). LgBiT-mGsq and SmBiT-βarr1 were synthetized by GenScript. mGsq and βarr1 were tagged in amino-terminal with LgBiT and a linker peptide and SmBiT, respectively.

The cell suspension containing DNA (EbBRET biosensors and receptors) were seeded in white 96-well plates (Greiner) at 30,000 cells/well (100 μl per well). 48 hours after transfection, cells were washed with DPBS (Gibco) and assayed in Tyrode’s buffer containing 137 mM NaCl, 0.9 mM KCl, 1 mM MgCl₂, 11.9 mM NaHCO₃, 3.6 mM NaH₂PO₄, 25 mM Hepes, 5.5 mM glucose, 1 mM CaCl₂ (pH 7.4) at 37°C. AVP or vehicle (water) were added and cells incubated at 37°C for the required time. 5 or 15 minutes before reading, 2.5 μM of the Rluc substrate coelenterazine 400a or 1.33 μM of methoxy e-coelenterazine (NanoLight Technology) was added, respectively. All EbBRET measurements were performed using a FLUOstar Omega microplate reader (BMG Labtech) with an acceptor filter (515 ± 30 nm) and donor filter (410 ± 80 nm). EbBRET values were determined by calculating the ratio of the light intensity emitted by the acceptor over the light intensity emitted by the donor. In kinetics or dose-response curves, ΔEbBRET is defined as the values of EbBRET in presence of AVP minus the value obtained with vehicle. Dose-response curves were fitted using nonlinear regression using a 4-parameter equation and the basal ΔEbBRET was fixed to zero. Statistical significance of parameters of dose-response curves (AVP-induced maximal efficacy or potency) was established by comparing independent fits with a global fit that shares the selected parameter using extra sum-of-squares F test. The transduction coefficients Log(τ/Ka) were determined using the operational from Kenakin and Christopoulos as previously described⁴⁹.

The NanoBiT assay to measure proximity between LgBiT-mG proteins and SmBiT-βarr1 has been reported previously⁵⁰. In short, 2,000,000 cells were seeded per well in 6 well plates. 24 hours later, 125 ng SmBiT-βarr1, 1000 ng V₂R or V₂β₂AR, and 125 ng LgBiT-mGs or 1000 ng LgBiT-mGsq were transfected into the cells using Lipofectamine 3000 transfection reagent. The next day, transfected cells were detached and 100,000 cells/well were plated into a Poly-D-lysine-coated white 96-well Microplate (Falcon) and incubated overnight at 37°C. The cells were equilibrated in Opti-MEM at 37°C for 60 minutes. Coelenterazine-h was added at a final concentration of 10 μM before starting the measurement. After establishing a baseline response for 2 minutes, cells were stimulated with AVP added at a final concentration of 100 nM and the luminescence was measured for additional 20 minutes. The signal was detected at 550 nm using a PHERAstar FSX instrument (BMG LabTech). ΔRLU is defined as the values of relative luminescence in presence of AVP minus the value obtained with vehicle.

Cells containing DNA (fluorescent-tagged localization markers, Halo-mGsq, and receptors) were seeded in 8-well glass chambered slides (Ibidi GMBH) at 30,000 cells per well. The day of the assay, HaloTag® Oregon Green® Ligand (Promega) was added to cells at a final concentration of 1 μM in the culture media and incubated 15 minutes (37°C, 5% CO₂) to label Halo-mGsq. Cells were washed 3 times with the media and incubated 30 minutes (37°C, 5% CO₂) for the last wash. The media was aspirated, replaced by Tyrode’s buffer and cells were stimulated with AVP or vehicle (water) for the required time at 37°C, 5% CO₂. At the end of the incubation, the media was aspirated and cells were fixed by adding 300 μl per well of 4% paraformaldehyde in PBS (Thermo Scientific) and incubated at room temperature for 10 minutes. The paraformaldehyde solution was aspirated, replaced by DPBS and cells were incubated for 10 minutes before being replaced by Tyrode’s buffer (300 μl per well) and visualized on a SP8 confocal microscope (Leica) at 63X magnification. Images were quantified using Imaris cell imaging software version 9.9.1 (Bitplane, Oxford Instruments). For each image, a threshold for the red channel was established by selecting the lower intensity from the region of interest (plasma membrane, endosomes, or βarr2). The percentage of colocalization was determined by the percentage of the material from the red channel above threshold colocalized with the material from the green channel. To quantify the percentage of colocalization for confocal microscopy images, the “surfaces” module was selected isolating cells containing the region of interest for each image. Thresholds for each channel were established by selecting the lower intensity from the region of interest (plasma membrane, endosomes, or βarr2). Data are reported as red volume (red voxels) above the threshold that is co-localized with green volume (green voxels) above the threshold and reported as percentage.

To measure the relative cell surface expression of V₂R and V₂β₂ (both tagged with a HA epitope at their amino-terminal), the same cell suspension containing DNA that was used for EbBRET assays was seeded in white 96-well plates (Greiner) previously coated with Poly-D-lysine at 30,000 cells/well (100 μl per well). Non transfected cells were used to establish the background of the assay. For the coating, Poly-D-Lysine solution (0.1 mg per ml; Cultrex) was added (50 μl per well) and the plates incubated at 37 °C for at least 30 minutes. Following the incubation, the solution was aspirated and wells washed two times with DPBS before adding the cell suspension containing DNA. 48 hours after seeding, cells were washed with DPBS and fixed by adding 50 μl per well of 4% paraformaldehyde in PBS (Thermo Scientific) and incubated at room temperature for 10 minutes. The fixing solution was aspirated and wells washed 3 times with the washing buffer (0.5% BSA in DPBS). The washing buffer was left in the wells for 10 minutes following the last wash. After the 10 minute incubation, the buffer was removed and 50 μl per well of monoclonal 3F10 anti-HA-Peroxidase (Sigma) 12.5 ng/ml in washing buffer was added and the plate incubated 1 hour at room temperature. The antibody was aspirated and wells washed 3 times with the washing buffer. The washing buffer was left in the wells for 10 minutes following the last wash and wells were washed again 3 times with DBPS only. After aspiration of the DPBS, 100 μl per well of SigmaFast™ OPD (Sigma) solution prepared as recommended by the manufacturer was added. Wells were incubated in presence of the OPD solution until the wells containing cells expressing receptors become yellow (typically 10 minutes). The reaction was stopped by addition of 25 μl per well of hydrochloride 3M in water. 100 μl per well were transferred to a transparent clear 96-well flat bottom plate (Corning) and absorbance at 492 nm was measured using a FLUOstar Omega microplate reader (BMG Labtech). The net absorbance represents the absorbance measured in presence of receptor minus the background (i.e. absorbance measured in absence of receptor).

The data and statistical analyses comply with the recommendations on experimental design and analysis in pharmacology. In all experiments at least three independent experiments were performed and for each experiment. n value is provided in the corresponding figure legend. All experiments are performed in quadruplicates. A P value ≤0.05 was considered as statistically significant for all analyses. Normally distributed and normalized data to control for unwanted sources of variation are shown as mean ± standard error on mean (s.e.m). All statistical analyses and nonlinear regressions were performed using GraphPad Prism 9.4.1 software.