Main

The two non-visual arrestins, β-arrestin1 and β-arrestin2 (also known as arrestin2 and arrestin3, respectively), bind to and regulate the majority of extraretinal GPCRs1,2. Both static crystallographic structures10,11,12,13,14 and biophysical studies in live cells15,16 indicate that arrestins undergo conformational rearrangement on GPCR binding. To investigate the effect of GPCR activation on the dynamics of β-arrestin2 conformation and function, we prepared a series of FlAsH BRET probes7 by inserting the six-amino-acid motif, CCPGCC, into the β-arrestin2 sequence at sites not predicted to be involved in its interactions with receptors or major binding partners (Fig. 1a). Each probe (rLuc–β-arrestin2–FlAsH1–6) was designed to measure BRET between a Renilla luciferase (rLuc) fluorescence donor at the amino terminus, and a fluorescein arsenical acceptor located at one of six positions along the length of β-arrestin2. We hypothesized that observing changes in BRET efficiency from multiple vantage points would yield an β-arrestin2 conformational signature that would correlate with its molecular functions. We first tested whether insertion of the FlAsH motif compromised β-arrestin2 recruitment by measuring the agonist-induced increase in intermolecular BRET between a C-terminal yellow fluorescent protein (YFP)-tagged GPCR and the N-terminal rLuc moiety of each rLuc–β-arrestin2–FlAsH construct. As shown in Fig. 1b, five of the rLuc–β-arrestin2–FlAsH constructs (F1, F2, F4, F5 and F6) generated BRET signals comparable to unmodified rLuc–β-arrestin2. The sixth construct (F3), which was poorly recruited, was included in subsequent experiments as an internal negative control. We then tested whether GPCR activation would produce an intramolecular rLuc–β-arrestin2–FlAsH BRET signal upon recruitment to an untagged GPCR. Agonist stimulation elicited changes in the β-arrestin2–FlAsH BRET signal (Δnet BRET) that were maintained over at least 10 min (Extended Data Fig. 1a) and proportional to receptor occupancy at less than saturating ligand concentration (Extended Data Fig. 1b). Thus, measuring the Δnet BRET of each construct produced a β-arrestin2–FlAsH BRET signature that was characteristic of the receptor being investigated (Fig. 1c). For the vasopressin type 2 receptor (V2R), ligand stimulation caused significant decreases in the signal from FlAsH sensors in the N-terminal (F1 and F2) and C-terminal (F4 and F5) globular domains, and a significant increase in signal from the sensor located at the C terminus (F6). Predictably, given its poor recruitment, the F3 construct did not significantly change with stimulation.

Figure 1: Design and characterization of rLuc–β-arrestin2–FlAsH BRET reporters.
figure 1

a, Six rLuc–β-arrestin2–FlAsH BRET reporters (F1–F6) were constructed by inserting the amino acid motif CCPGCC after amino acid residues 40, 140, 171, 225, 263 and 410 of β-arrestin2. The location of each FlAsH motif is shown in relation to the globular N and C domains of β-arrestin2, as well as the clathrin and adaptor protein 2 (AP2) binding sites and reported phosphorylation sites (Ser361 and Thr383) in the β-arrestin2 C-terminal regulatory (R2) domain1. b, Intermolecular BRET demonstrating ligand-dependent recruitment of rLuc–β-arrestin2–FlAsH1–6 to human PTH1R. c, rLuc–β-arrestin2–FlAsH1–6 ‘signature’ of β-arrestin2 binding to the V2R. The bar graphs depict mean ± s.e.m. of independent biological replicates (n = 3 (b) and n = 5 (c)). *P < 0.05, #P < 0.005, greater or less than vehicle-stimulated control.

PowerPoint slide

Full size image

To determine whether β-arrestin2–FlAsH signatures were conserved between GPCRs, we selected a panel of six additional receptors with diverse G-protein coupling, arrestin binding, and arrestin-dependent signalling characteristics (Extended Data Table 1). Our test panel included two stable arrestin-binding class ‘B’17 GPCRs: the angiotensin AT1A receptor (AT1AR) and the type 1 parathyroid hormone receptor (PTH1R); three transient arrestin-binding class ‘A’17 GPCRs: the α1B-adrenergic receptor (α1BAR), the β2-adrenergic receptor (β2AR), and the sphingosine-1-phosphate 1 receptor (S1P1R); and the α2A-adrenergic receptor (α2AAR) that does not produce detectable β-arrestin2 translocation. The G-protein-mediated signalling of each receptor was characterized using a FLIPRTETRA fluorescent imaging plate reader to measure ligand-dependent activation or inhibition of adenylyl cyclase and stimulation of transmembrane Ca2+ entry18 (Extended Data Fig. 2). The pattern of arrestin recruitment was confirmed by confocal fluorescence microscopy using GFP-tagged β-arrestin2 (ref. 19) (Fig. 2a). The β-arrestin2–FlAsH BRET signature generated by each receptor is shown in Fig. 2b. As the Δnet BRET observed with each probe reflects the ‘population average’ conformation of the cellular pool of rLuc–β-arrestin2–FlAsH, signatures were generated under conditions of receptor excess and saturating ligand concentration to ensure that the largest possible fraction of the reporter pool was receptor-bound at steady state. Inspection of the rLuc–β-arrestin2–FlAsH BRET signatures revealed that the class B receptors AT1AR, PTH1R and V2R (Fig. 1c), which form stable GPCR–arrestin complexes that transit to endosomes17, produced significant negative Δnet BRET signals at the F4 position and positive Δnet BRET signals at the C terminus (Fig. 2b; black arrows). In contrast, the class A α1BAR, β2AR and S1P1R, which dissociate from arrestin soon after internalization17, produced little to no signal in these positions. Only small N-terminal responses were observed with α2AAR, which interacts weakly with β-arrestin2 (ref. 20).

Figure 2: Relationship between GPCR–β-arrestin2 complex formation, rLuc–β-arrestin2–FlAsH BRET signature, and arrestin-dependent ERK1/2 activation for six different GPCRs.
figure 2

a, Agonist-dependent recruitment of β-arrestin2–GFP. Each panel depicts a representative field of stimulated cells. β-arrestin2–GFP was diffusely cytosolic in the absence of agonist (not shown). Scale bars, 10 μm. b, Receptor-specific rLuc–β-arrestin2–FlAsH1–6 signatures. Each bar graph depicts mean ± s.e.m. of independent biological replicates (n = 5). *P < 0.05, #P < 0.005, † P < 0.001, greater or less than vehicle-stimulated control. Black arrows indicate BRET changes related to GPCR–β-arrestin2 complex stability; grey arrows indicate changes related to β-arrestin2-dependent ERK1/2 activation. c, Effect of downregulating β-arrestin1/2 expression on GPCR-mediated ERK1/2 phosphorylation. A representative phospho-ERK1/2 immunoblot (IB) is shown above a bar graph depicting the mean ± s.e.m. of independent biological replicates (n = 7, AT1AR; n = 9, PTH1R; n = 6, α1BAR; n = 20, β2AR; n = 5, α2AAR and S1P1R). Responses were normalized to the basal level of phospho-ERK1/2 in non-stimulated samples. Dox, doxycycline; iso, isoproterenol; phe, phenylephrine; PTH, parathyroid hormone N-terminal 1–34 fragment; S1P, sphingosine-1-phosphate; UK, UK14303. *P < 0.05, greater or less than stimulated response in non-induced cells. NS, not significant.

PowerPoint slide

Full size image

To relate the β-arrestin2–FlAsH BRET signature to arrestin-dependent signalling, we determined the effect of silencing β-arrestin1/2 expression on ligand-stimulated ERK1/2 activation21,22 using HEK293 FRT/TO β-arrestin1/2 shRNA cells that carry tetracycline-inducible shRNA targeting β-arrestin1/2 (ref. 23). As shown in Fig. 2c, ERK1/2 activation by AT1AR, PTH1R and α1BAR was significantly attenuated by β-arrestin1/2 silencing, indicating a positive signalling role for arrestin scaffolds24. β-arrestin1/2 silencing had no net effect on ERK1/2 activation by β2AR, which reportedly activates ERK1/2 via both Gi/o-dependent and arrestin-dependent pathways in HEK293 cells25, and significantly enhanced ERK1/2 activation by the S1P1R and α2AAR, suggesting that for these receptors the major role of arrestins is to dampen G-protein-dependent ERK1/2 activation by promoting desensitization. Consistent with this, we found that ERK1/2 activation via β2AR, S1P1R and α2AAR was strongly pertussis toxin-sensitive, indicating a predominantly Gi/o-mediated mechanism of activation (Extended Data Fig. 3). Comparison with the rLuc–β-arrestin2–FlAsH BRET signatures revealed a correlation between arrestin-dependent ERK1/2 activation and a significant negative Δnet BRET signal at the F5 position. This was most apparent for the class A α1BAR, which lacked the F4 and F6 signals characteristic of class B receptors, but retained the F5 signal shared by GPCRs mediating arrestin-dependent signals (Fig. 2b; grey arrows). The relationship between α1BAR-induced F5 signal and ERK1/2 activation was present over a range of agonist concentrations (Extended Data Fig. 4a), whereas at saturating ligand concentration the F5 signal readily separated the positive and negative roles of arrestin in ERK1/2 activation by our panel of seven GPCRs (Extended Data Fig. 4b).

We next examined chimaeric GPCRs in which the receptor C-tail was exchanged to reverse the class A and class B patterns of arrestin binding. As shown in Fig. 3a, replacing the C-tail of the class B V2R with that of the class A β2AR (V2β2ctR) is sufficient to reverse the arrestin binding pattern26. Although the C-tail exchange affected the stability of the receptor–arrestin complex, it did not affect arrestin-dependent ERK1/2 activation, which persisted in the V2β2ctR. Comparison of the rLuc–β-arrestin2–FlAsH BRET profiles generated by V2R and V2β2ctR revealed that conversion of class B to class A binding caused the loss of the negative F4 signal characteristic of class B receptors such as AT1AR, PTH1R, and V2R (Fig. 2b). In contrast, the F5 signal was preserved, such that the rLuc–β-arrestin2–FlAsH BRET signature of the chimaeric V2β2ctR resembled that of the α1BAR, the other class A GPCR that retained arrestin signalling. The opposite experiment, involving conversion of a class A receptor to class B, is shown in Fig. 3b. Replacing the C-tail of the class A β2AR with that of the class B V2R (β2V2ctR) reverses the arrestin binding pattern. In this case, β2V2ctR-mediated ERK1/2 activation became more arrestin-dependent, as evidenced by acquired sensitivity to shRNA silencing of β-arrestin1/2 expression. Inspection of the β-arrestin2–FlAsH BRET profiles of β2R and V2β2ctR revealed that conversion of class A to class B produced a significant increase in the F4 signal that was most apparent following 10 min of ligand stimulation. Notably, the F5 signal also increased, consistent with the gain of arrestin-dependent signalling. Thus, reversing the stability of the arrestin–GPCR complex, without altering the other intracellular loops of the receptor, was sufficient to produce loss/gain of FlAsH BRET signal at the F4 position, whereas the magnitude of change in the F5 position correlated with arrestin-dependent signalling.

Figure 3: Effect of GPCR–arrestin trafficking pattern and ligand structure on the rLuc–β-arrestin2–FlAsH BRET conformational signature.
figure 3

a, Effect of converting stable class B β-arrestin2 binding to transient class A binding. Upper, representative confocal fluorescence images showing the pattern of ligand-stimulated GFP–β-arrestin2 recruitment to the V2R or the chimaeric V2β2ctR. Centre, the β-arrestin2–FlAsH1–6 profiles generated by V2R and V2β2ctR. Lower, the arrestin-dependence of ERK1/2 phosphorylation by the V2R and V2β2ctR. b, Analogous experiment demonstrating the effect of converting transient class A β-arrestin2 binding to stable class B binding using the β2AR and the chimaeric β2V2ctAR. In a and b, phospho-ERK1/2 bar graphs depict mean ± s.e.m. of independent biological replicates (n = 12, V2R and V2β2ctR; n = 20, β2AR; n = 12 β2V2ctR). *P < 0.05, less than stimulated response in non-induced cells. Black arrows indicate BRET changes related to GPCR–β-arrestin2 complex stability; grey arrows indicate changes related to β-arrestin2-dependent ERK1/2 activation. c, Effect of ligand structure on the rLuc–β-arrestin2–FlAsH BRET signature. Upper, representative confocal fluorescence images showing the pattern of GFP–β-arrestin2 recruitment to the AT1AR upon stimulation with AngII, SBpA, SVdF, SI or SII. Centre, the F1 and F4–6 profiles generated by each ligand. Bottom, the relationship between the amplitude of the F4 signal and the independently determined avidity of AT1AR and β-arrestin2 measured by FRAP27. SII was not included in the linear fit, as the AT1AR–β-arrestin2 avidity is too low to measure by FRAP. In all panels, the rLuc–β-arrestin2–FlAsH BRET graphs represent mean ± s.e.m. of independent biological replicates (n = 5, V2R and V2β2ctR; n = 6, β2AR and β2V2ctR; n = 6, AT1AR, with each ligand). *P < 0.05, #P < 0.005, † P < 0.001, greater or less than vehicle-stimulated control. Scale bars, 10 μm.

PowerPoint slide

Full size image

We then compared the β-arrestin2–FlAsH BRET signature generated by angiotensin II (AngII) with those of a previously characterized series of arrestin-selective ‘biased’ AngII analogues27 (Fig. 3c). Although all five ligands—AngII, [Sar1,Ile4,Ile8]-AngII (SII), [Sar1,Ile8]-AngII (SI), [Sar1,Val5,d-Phe8]-AngII (SVdF) and [Sar1,Val5,Bpa8]-AngII (SBpA)—promote the assembly of endosomal AT1AR–arrestin complexes, fluorescence recovery after photobleaching (FRAP) has demonstrated that they cause different avidity between the receptor and β-arrestin2, with the rank order of receptor–arrestin complex half-life of AngII > SBpA > SVdF > SI > SII (ref. 27). The efficiency with which these ligands promote arrestin-dependent ERK1/2 activation corresponds to the avidity of the complex, with longer half-life complexes generating proportionally greater arrestin-dependent signalling27. Inspection of the β-arrestin2–FlAsH BRET signatures demonstrated that although different ligands had little effect on the magnitude of the N-terminal F1 shift, the amplitude of the F4 and F5 signals were very sensitive to ligand structure. Plotting the F4 signal versus receptor–arrestin avidity measured by FRAP revealed a strong linear correlation. Thus, the signature presented by β-arrestin2–FlAsH BRET probes in the C-terminal domain reflected the avidity of the AT1AR–β-arrestin2 interaction, even when comparing ligands that all evoke a canonical class B pattern of arrestin recruitment.

The rLuc–β-arrestin2–FlAsH BRET signature reflects both changes in the distance/orientation of the fluorophores due to conformational rearrangement, and steric effects generated by arrestin interaction with its receptor and non-receptor binding partners. Although it is not possible to ascribe the rLuc–β-arrestin2–FlAsH BRET signal at a given position to specific conformational shifts or engagement of binding partners, our data clearly demonstrate that ligand–GPCR complexes confer distinctive β-arrestin2 conformations, and that features of the conformational signature are conserved between receptors with similar arrestin binding/signalling characteristics. Moreover, we find that the Δnet BRET at selected positions correlates with downstream arrestin function, for example, class A versus class B trafficking and arrestin-dependent ERK1/2 activation, suggesting that β-arrestin2–FlAsH BRET probes can predict arrestin function on the basis of the ligand-induced conformational signature. Thus, intramolecular rLuc–β-arrestin2–FlAsH BRET probes may aid in identifying the factors that determine arrestin conformation and function, such as ligand ‘bias’8,9, GPCR C-tail ‘phosphorylation codes’ written by different GRKs28, and post-translational modifications of arrestin that stabilize or destabilize the complex29.

While this work was in progress, we became aware of a complementary study using β-arrestin2–FlAsH fluorescence resonance energy transfer (FRET) sensors30. This study confirms the existence of GPCR-specific β-arrestin2 conformations and, with the superior temporal resolution of FRET, provides key insights into the kinetics of receptor binding and arrestin activation.

Methods

Materials

Cell culture medium and cell culture additives were from Life Technologies. FuGENE HD transfection reagent and Promega GloSensor cAMP reagent were from Fisher Scientific. FLIPR Calcium 5 Assay Kit was from Molecular Devices, Inc. Lipofectamine 2000 and TC-FlAsH II In-Cell Tetracysteine Tag Detection Kits were from Invitrogen. Human PTH(1–34) was obtained from Bachem, Inc. Angiotensin II, [Arg8]-vasopressin, isoproternol, phenylephrine, and UK14303 were from Sigma-Aldrich. S1P was from Avanti Polar Lipids Inc. SII was from MP Biomedicals. SI, SVdF and SBpA were synthesized at the Institut de Pharmacologie de Sherbrooke, Sherbrooke University. Rabbit polyclonal anti-β-arrestin1/2was a gift from R.J. Lefkowitz. Anti-phospho-ERK1/2 IgG (T202/Y204; #9101) and anti-ERK1/2 IgG (#4695) were from Cell Signaling Technology. Horseradish-peroxidase-conjugated donkey anti-rabbit IgG was from Jackson Immuno-Research Laboratories, Inc.

Renilla luciferase–β-arrestin2 FlAsH BRET reporters

The pcDNA3.1 plasmid encoding rat β-arrestin2 tagged at the N terminus with Renilla luciferase (rLuc) was a gift from M. Bouvier. A series of six rLuc–β-arrestin2–FlAsH BRET reporters were constructed by inserting a cDNA sequence encoding the amino acid motif, CCPGCC, immediately following amino acid residues 40, 140, 171, 225, 263 and 410 of β-arrestin2, using a modification of the precise gene fusion PCR method31. For each construct, two PCR steps were performed using the primer sets shown in Extended Data Table 2. The first step was to generate two PCR fragments using the primer pairs: RlucHindF–FlAsHR and FlAsAF–RlucApalR. One PCR product contained a HindIII restriction site at the 5′ end and the CCPGCC FlAsH motif at the 3′ end, and the other contained the complementary FlAsH sequence at the 5′ end and an ApaI restriction site at the 3′ end. A second PCR step was used to fuse the two fragments using three primers: RlucHindF, FlAsHR and RluApalR, and the two PCR fragments as template DNA. The resultant full-length β-arrestin2 PCR product containing the FlAsH motif insert was digested with HindIII and ApaI and cloned into the parent rLuc–β-arrestin2 plasmid to generate the rLuc–β-arrestin2–FlAsH1–6 expression plasmids. All constructs were verified by dideoxynucleotde sequencing.

Cell culture and transfection

HEK293 cells (ATCC CRL1573) were from the American Type Culture Collection. HEK-293 GloSensor cells were from Promega Corporation. HEK293 cells were maintained in minimum essential medium supplemented with 10% fetal bovine serum and 1% antibiotic/antimycotic solution. The HEK293 FRT/TO β-arrestin1/2 shRNA cell line carrying tetracycline-inducible shRNA simultaneously targeting the β-arrestin1 and 2 isoforms (5′-CGTCCACGTCACCAACAAC-3′) was generated as previously described23. These cells were maintained in Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum, 1% antibiotic/antimycotic solution and 50 μg ml−1 zeocin, 50 μg ml−1 blasticidin, and 50 μg ml−1 puromycin to maintain selection. Transient transfections were performed using Lipofectamine 2000 or FuGENE HD according to the manufacturer’s protocols. Before experimentation, cells were serum-deprived overnight in 1% fetal bovine serum growth medium. Cells were not tested for mycoplasma contamination.

FLIPRTETRA assay of calcium influx

HEK293 cells in 6-well plates were transiently transfected with 1 μg of plasmid cDNA encoding the angiotensin AT1A, PTH1R, α1BAR, β2AR, S1P1R or α2AAR, using Lipofectamine 2000. Cells were seeded onto collagen-coated black-wall clear-bottom 96-well plates (BD Biosciences) 24 h after transfection, allowed to grow for a further 24 h, then serum deprived overnight. Fresh FLIPR calcium 5 assay reagent (100 μl per well) was added to 100 μl of serum-deprivation medium and plates were incubated for an additional 1 h before stimulation. Stimulations were carried out on a FLIPRTETRA (Molecular Devices) with 470–495 nm excitation and 515–575 nm emission filters as previously described18. All assays were performed using saturating ligand concentrations: AngII (0.1 μM), hPTH(1–34) (0.1 μM), isoproterenol (1 μM), phenylephrine (10 μM), S1P (1 μM) or UK14303 (10 μM) and run at room temperature. The instrument was programmed to simultaneously dispense 50 μl of vehicle control, 5× ligand, or the calcium ionophore A23187 (10 μM) from the drug plate into the 200 μl of medium in the corresponding wells of the assay plate to achieve the final ligand concentration. Fluorescence was recorded every 1 s for 10 reads to establish baseline fluorescence, then every 1 s for 300 reads. Raw data representing the relationship between time and fluorescence for each well were exported to Microsoft Excel for background subtraction and analysis.

FLIPRTETRA assay of cAMP production

Assays were performed using HEK293 GloSensor cAMP cells that stably express a genetically encoded biosensor composed of a cAMP binding domain fused to a mutated form of Photinus pyralis luciferase32. HEK293 GloSensor cAMP cells were seeded onto poly-D-lysine-coated white-wall clear-bottom 96-well plates (BD Biosciences) 24 h after transient transfection with plasmid cDNA encoding the receptors of interest. cAMP assays were performed 72 h after transfection as previously described18. cAMP reagent medium was prepared by adding 200 μl of freshly thawed GloSensor cAMP reagent to 10 ml of serum free MEM buffered with 10 mM HEPES (pH 7.4). The growth medium was gently aspirated and replaced with 100 μl per well of pre-warmed cAMP reagent medium. Plates were incubated at 37 °C with 5% CO2 for 1.5 h, then removed from the incubator and incubated at room temperature in the dark for an additional 30 min. Stimulations were performed at room temperature in the FLIPRTETRA using saturating ligand concentrations. Luminescence was recorded every 1 s for 10 reads to establish baseline luminescence, then every 1 s for 50 reads. Thereafter, luminescence was recorded every 2 s for 600 reads. Raw data representing the relationship between time and luminescence for each well following ligand addition was exported to Microsoft Excel for background subtraction and analysis. All responses were normalized to the cAMP luminescence generated in response to 10 μM forskolin. To assay Gi-mediated inhibition of cAMP production (α22AR and S1P1R), cells were pre-incubated with or without agonist for 30 min, then stimulated with 10 μM forskolin.

Intermolecular BRET using rLuc–β-arrestin2 and PTH1R–YFP

HEK 293 cells were transiently transfected with 1.5 μg of plasmid DNA encoding the C-terminal yellow fluorescent protein (YFP)-tagged PTH1R33 and 0.15 μg of either rLuc–β-arrestin2 or one of the rLuc–β-arrestin2–FlAsH constructs using Fugene HD. Cells were detached 48 h after transfection, collected by centrifugation, resuspended in BRET buffer (1 mM CaCl2, 140 mM NaCl, 2.7 mM KCl, 900 μM MgCl2, 370 μM NaH2PO4, 5.5 mM D-glucose, 12 mM NaHCO3, 25 mM HEPES (pH 7.4)) and aliquotted into white-wall clear-bottom 96-well plates at a density of 100,000 cells per well. Background and total Venus fluorescence were read on an OptiPlate microplate reader (PerkinElmer) with 485 nm excitation and 525–585 nm emission filters. Cells were stimulated with 0.1 μM PTH(1–34) for 2 min and coelenterazine was then added to a final concentration of 5 μM. Luciferase (440–480 nm) and Venus (525–585 nm) emissions were read to calculate the BRET ratio (emission eYFP/emission Rluc). Net BRET ratio was calculated by background subtracting the BRET ratio measured for vehicle- versus ligand-treated cells in the same experiment.

Intramolecular FlAsH BRET using the rLuc–β-arrestin2–FlAsH constructs

HEK293 cells seeded in 6-well plates were co-transfected with 1.5 μg of plasmid DNA encoding the receptor of interest and 0.1 μg of DNA encoding one rLuc–β-arrestin2–FlAsH construct using Fugene HD. Cells were detached 48 h after transfection, collected by centrifugation, and resuspended in 600 μl of Hank’s balanced salt solution. TC-FlAsH II In-Cell Tetracystein detection reagent was added at 2.5 μM final concentration and the cells incubated at room temperature for 30 min, after which they were washed using 1× BAL buffer from the TC-FlAsH kit, resuspended in BRET buffer and placed in white-wall clear-bottom 96-well plates at a density of 100,000 cells per well. Background and total TC-FlAsH fluorescence were read on an Optiplate microplate reader (Perkin-Elmer) with 485 nm excitation and 525–585 nm emission filters. Except as noted in the figure legends, all stimulations were carried out at saturating ligand concentration: AngII (0.1 μM), [Arg8]–vasopressin (1 μM), hPTH(1–34) (0.1 μM), isoproterenol (1 μM), phenylephrine (10 μM), S1P (1 μM), SBpA (1 μM), SII (1 μM) SVdF (1 μM), or UK14303 (10 μM). Cells were exposed to agonist for 2–10 min, after which coelenterazine was added at a final concentration of 5 μM. Six consecutive readings of luciferase (440–480 nm) and TC-FlAsH (525–585 nm) emissions were taken, and the BRET ratio (emission eYFP/emission Rluc) calculated using Berthold Technologies Tristar 3 LB 941. The Δnet change in intramolecular BRET ratio for each of the six rLuc–β-arrestin2–FlAsH constructs was calculated by background subtracting the BRET ratio measured for cells in the same experiment stimulated with vehicle only.

Confocal microscopy

For determining the pattern of GPCR–arrestin trafficking, HEK293 cells were seeded into collagen-coated 35 mm glass-bottom Petri dishes (MatTek Corporation) and co-transfected with 1.3 μg of plasmid DNA encoding the receptors of interest and 0.7 μg of plasmid encoding green fluorescent protein (GFP)-tagged β-arrestin219 using FuGene HD. Forty-eight hours after transfection, cells were serum derived for 4 h, stimulated with a saturating ligand concentration for 8 min, fixed with 4% paraformaldehyde in phosphate buffered saline for 30 min and washed with 4 °C saline. Arrestin distribution was determined by confocal microscopy performed on a Zeiss LSM510 META laser-scanning microscope with 60× objective using 488 nm excitation and 505–530 nm emission wavelengths. Measurement of AT1AR–β-arrestin2 avidity was performed as previously described27. HEK293 cells stably expressing AT1AR and transfected with β-arrestin2–YFP were stimulated with AngII (1 μM) or analogues (10 μM) for 15 min, after which, endosomes were bleached and fluorescence recovery was monitored every 30 s over a period of 5 min.

Immunoblotting

HEK293 FRT/TO β-arrestin1/2 shRNA cells were used to determine the contribution of arrestins to GPCR-stimulated ERK1/2 activation23,34. Cells in 12-well plates were transiently transfected with 1 μg of plasmid cDNA encoding the receptor of interest using FuGENE HD. Twenty-four hours after transfection, downregulation of β-arrestin1/2 expression was induced by 48 h exposure to 1 μM doxycycline. After overnight serum deprivation, cells were stimulated for 5 min, after which monolayers were lysed in 1× Laemmli sample buffer. Stimulations were performed at saturating ligand concentration, except as noted in the figure legends. Lysates containing 10 μg of whole-cell protein were resolved by sodium dodecyl sulphate polyacrylamide gel electrophoresis and transferred to polyvinylidine difluoride membranes. Immunoblots of phospho-ERK1/2, total ERK1/2, and β-arrestin1/2 were performed using rabbit polyclonal IgG with HRP-conjugated goat anti-rabbit IgG as secondary antibody. Proteins were visualized using enhanced chemiluminescence (PerkinElmer).

Statistical analysis

The sample size (n) reported in each figure legend refers to number of independently performed biological replicates in the data set. All analysable data points were included in the statistical analyses. No statistical methods were used to predetermine sample size. For experimental methods that were highly reproducible, for example, measurement of Δnet BRET, 5 to 6 biological replicates were sufficient to discern effects of ±0.01 with P < 0.05. For experimental methods with greater variability between replicates, for example, fold ERK1/2 activation, 5 to 20 biological replicates were necessary to discern effects of β-arrestin1/2 silencing that were ±10% of the control response with P < 0.05. The investigators were not blinded to allocation during experiments and outcome assessment. All values are expressed as mean ± s.e.m. (n ≥ 5). For comparisons between two groups, statistical significance was assessed with a two-tailed unpaired t-test. Computations were performed and graphs constructed with the GraphPad Prism 4.0 scientific graphing, curve fitting, and statistics program (GraphPad Software). The experiments were not randomized.