Research ResourceGPCR SIGNALING

Functional selectivity profiling of the angiotensin II type 1 receptor using pathway-wide BRET signaling sensors

See allHide authors and affiliations

Science Signaling  04 Dec 2018:
Vol. 11, Issue 559, eaat1631
DOI: 10.1126/scisignal.aat1631

Biosensors for GPCR-biased signaling

Upon binding to ligands, G protein–coupled receptors (GPCRs) stimulate heterotrimeric G proteins and recruit β-arrestin, which dampens signaling through G proteins and acts as a scaffold to activate other pathways. Whereas some ligands stimulate both G proteins and β-arrestin, others preferentially stimulate one or the other, a phenomenon known as biased agonism. For receptors engaging more than one G protein subtype, ligands may also bias downstream signaling to a particular G protein. Namkung et al. developed bioluminescence resonance energy transfer (BRET) biosensors, which they used to characterize biased signaling downstream of the angiotensin II (AngII) type 1 receptor (AT1R) in response to various ligands, as well as the bias of naturally occurring AT1R variants. This tool kit can be used to comprehensively dissect the signaling bias of other GPCR-ligand combinations, which may benefit drug development efforts.


G protein–coupled receptors (GPCRs) are important therapeutic targets that exhibit functional selectivity (biased signaling), in which different ligands or receptor variants elicit distinct downstream signaling. Understanding all the signaling events and biases that contribute to both the beneficial and adverse effects of GPCR stimulation by given ligands is important for drug discovery. Here, we report the design, validation, and use of pathway-selective bioluminescence resonance energy transfer (BRET) biosensors that monitor the engagement and activation of signaling effectors downstream of G proteins, including protein kinase C (PKC), phospholipase C (PLC), p63RhoGEF, and Rho. Combined with G protein and β-arrestin BRET biosensors, our sensors enabled real-time monitoring of GPCR signaling at different levels in downstream pathways in both native and engineered cells. Profiling of the responses to 14 angiotensin II (AngII) type 1 receptor (AT1R) ligands enabled the clustering of compounds into different subfamilies of biased ligands and showed that, in addition to the previously reported functional selectivity between Gαq and β-arrestin, there are also biases among G protein subtypes. We also demonstrated that biases observed at the receptor and G protein levels propagated to downstream signaling pathways and that these biases could occur through the engagement of different G proteins to activate a common effector. We also used these tools to determine how naturally occurring AT1R variants affected signaling bias. This suite of BRET biosensors provides a useful resource for fingerprinting biased ligands and mutant receptors and for dissecting functional selectivity at various levels of GPCR signaling.


G protein–coupled receptors (GPCRs) represent the largest class of membrane-bound proteins and are involved in diverse biological processes ranging from hormone action and neurotransmission to cell migration, proliferation, and differentiation (1, 2). They transmit signals from external stimuli conveyed by natural and synthetic ligands, such as hormones and drugs, by engaging different intracellular cascades of signaling effectors and modulators such as the heterotrimeric G proteins, GPCR kinases (GRKs), and β-arrestins, as well as downstream second messenger–generating enzymes and channels. Activation of these receptors by ligands leads to the functional dissociation of Gα and Gβγ subunits within the G protein heterotrimer and the activation of downstream signaling effectors such as adenylyl cyclases (ACs; which can also be inhibited by the action of G proteins), phospholipases such as PLC and PLD, second messenger–dependent protein kinases such as PKA and PKC, and small guanosine triphosphatases (GTPases) such as Rho and Ras. The recruitment of β-arrestin to GRK-phosphorylated, agonist-occupied GPCRs uncouples them from heterotrimeric G proteins at the plasma membrane (PM), leading to a reduction in G protein–dependent signaling, and targets GPCRs for internalization (1, 2). GPCRs have been typically categorized by their ability to couple to and activate a preferred subtype of heterotrimeric G protein, as defined by the identity of the α subunit (Gαs, Gαq, Gαi, etc.); hence, they are often qualified as Gs-, Gq-, or Gi-coupled receptors with the assumption that signaling is restricted to one of these G proteins. However, it has become clear that many GPCRs engage more than one subtype of G protein, albeit with variable efficacies, resulting in the activation of several downstream signaling cascades that specify the responses elicited by the receptor in cells and tissues (24). Similarly, β-arrestins (β-arrestins 1 and 2) have been found to not only play a role in receptor desensitization and act as endocytic adaptors for many GPCRs but also to participate in downstream signaling, affecting cell responsiveness in distinct ways (1, 5). Therefore, detecting how this multidimensionality of signaling repertoire for a given receptor is faithfully translated into intracellular signals remains a challenge.

Observations of distinct ligands acting on the same receptor differentially activating different subsets of signaling effectors in cells have led to the idea that receptors can be conformationally stabilized by ligands in multiple discrete active signaling states (6), a notion that is supported by receptor structural studies (7, 8). The overall signaling profile of a particular GPCR-ligand combination can also be modified by naturally occurring polymorphisms and engineered mutations, which also affect the conformational signaling landscape of the ligand-bound receptor to preferentially engage one pathway over another (912). Stabilization of such distinct conformations by different ligands or mutant forms of ligands or receptors thus enables the preferential activation of specific subsets of effectors and downstream signaling pathways. Conformationally dependent GPCR signaling is now a well-accepted concept that has been encapsulated within the pharmacological notion of “biased agonism,” “functional selectivity,” or “pluridimensional efficacy” (13, 14) and has been described for many GPCRs, including the angiotensin II type 1 receptor (AT1R), which is activated by the octapeptide angiotensin II (AngII) (1518). For example, the peptide SII, the first reported ligand with biased activity on AT1R, has been shown to promote β-arrestin engagement independently of Gαq activation (15, 16). Similarly, AT1R signaling elicited by other AngII analogs such as SI, SVdF, SBpa, DVG, TRV027, and Ang(1–7) have also been shown to preferentially engage β-arrestin over Gαq, albeit with different relative efficiencies (1820), whereas SII was also reported to engage G proteins other than Gαq after its binding to AT1R (17, 21). However, the ability of SII to engage AT1R signaling pathways downstream of these other G proteins or other AngII analogs to promote receptor coupling to distinct G proteins or to other downstream signaling pathways has not been explored carefully.

Exploiting functional selectivity holds great promise for developing more efficient and safer therapeutics (6, 22). Defining the signaling pathways responsible for desired versus undesired effects would facilitate the development of novel drugs with better therapeutic indexes. In principle, this could be achieved by selectively activating the pathways responsible for the desired effects while avoiding the activation of those involved in tolerance or undesirable side effects. This notion has been applied to the AT1R for which TRV027, a biased peptide ligand favoring β-arrestin over Gαq signaling, has been developed to treat heart failure (23). Although TRV027 was efficacious in animal models, in a clinical trial testing the efficacy in human, TRV027 did not improve clinical status compared to placebo in acute heart failure patients (24). Efforts to develop therapeutically advantageous biased drugs can thus be limited by an incomplete understanding of the mechanisms governing functional selectivity, as well as the lack of complete information about the entire repertoire of downstream mediators engaged by biased ligands that could contribute to the desired versus undesired therapeutic effects. Studies on GPCR functional selectivity have often focused on a subset of responses or on a narrow range of ligands, or both, and little information is available about the extent to which the observed biased engagement of G proteins translates into differences in downstream signaling outputs. Moreover, how functional selectivity is generally evaluated (for example, comparing biased effects of one ligand or a receptor mutant for a limited subset of responses, as compared to a reference agonist or the wild-type receptor) may represent a challenge when trying to link biased agonism with therapeutic effects of drugs and of the effects of mutations when those effects are contingent on an ensemble of signaling outputs engaged at various levels and at different times in cells (25). Hence, there is a need for assessing functional selectivity in a more global manner.

Here, we describe the development, validation, and use of a suite of bioluminescence resonance energy transfer (BRET)–based biosensors, including G proteins and their downstream effectors as well as β-arrestin, for studying the functional selectivity of AT1R signaling. Using this platform of pathway-wide BRET-based sensors, we dissected AT1R signaling and demonstrated functional selectivity of various AngII analogs and receptor mutants, revealing much greater signaling profile diversity than previously reported. In addition, the ability to probe signaling activities at different levels downstream of the receptor, in combination with selective pharmacological inhibitors, reveals points of convergence and divergence in the signaling networks that are engaged.


Assessing G protein and β-arrestin signaling elicited by AT1R

The AT1R not only couples to the Gαq/11 family of G proteins but also has been reported to activate other G proteins, such as Gα12/13 and Gαi, as well as β-arrestins (17, 18, 21, 26). Therefore, we first validated the ability of AT1R to activate different G proteins and β-arrestins using BRET-based biosensors (Fig. 1, A and B). These G protein biosensors, which have been previously described, measure separation of the Gα and Gβγ subunits after receptor activation using various Gα subunits tagged with the energy donor RlucII and a Gγ subunit tagged with the energy acceptor GFP10, yielding a reduction in BRET signal upon dissociation of the heterotrimer (2731). We also used the β-arrestin2 double-brilliance (βarr2-DB) BRET sensor that quantifies β-arrestin binding to AT1R (Fig. 1B), wherein the conformational change in β-arrestin after receptor binding results in a reduction of the BRET signal (18). In these experiments, human embryonic kidney (HEK) 293 cells were transiently cotransfected with complementary DNAs (cDNAs) encoding AT1R and either (i) individual Gα-RlucII constructs plus an optimized Gβγ pair consisting of Gβ1 and a GFP10-tagged Gγ1/2 or (ii) βarr2-DB. To compare the efficacy of AT1R coupling to G proteins, we measured the BRET signals at the time of the maximal response after full agonist occupancy. No significant BRET signals over basal were detected with the sensors upon AngII stimulation when AT1R was not expressed in cells (fig. S1A). However, significant changes in maximal BRET signals were observed with the Gαq, Gα12, and Gαi1,2,3 sensors after AngII stimulation (Fig. 1C and fig. S1B, respectively). The Gαs sensor showed no significant BRET change upon AngII stimulation of AT1R (fig. S1B). As expected, the selective Gαq inhibitor UBO-QIC (also known as FR900359) (31) completely blocked the BRET signal from the Gαq sensor but not that from the Gαi2 or Gα12 sensors (Fig. 1C). AngII concentration dependently promoted βarr2 recruitment to AT1R, as revealed by a reduction in BRET of the βarr2-DB sensor (Fig. 1D).

Fig. 1 AT1R downstream signaling pathways and BRET sensors for G protein and β-arrestin activation.

(A) AT1R stimulates signaling through several different classes of heterotrimeric G proteins and activation of βarr2. cAMP, adenosine 3′,5′-cyclic monophosphate. (B) Illustration of BRET-based Gα-Gβγ and βarr2-DB sensors. Upon receptor activation, dissociation of the Gα subunit from Gβγ and conformational changes in β-arrestin cause the BRET signal to decrease. L, ligand; GDP, guanosine diphosphate. (C) AT1R-induced G protein activation. HEK293 cells were transfected with AT1R along with the indicated Gα-RlucII (Gαq, Gα12, or Gαi2) plus GFP10-Gγ and Gβ. Cells were preincubated in the absence or presence of the selective Gαq inhibitor UBO-QIC or vehicle and then stimulated with AngII for 2 min (Gαq and Gαi2) or 10 min (Gα12). Data represent means ± SEM from at least three independent experiments. **P < 0.01, unpaired Student’s t test. (D) Concentration-response curve of the βarr2-DB sensor upon AngII stimulation of AT1R. Cells were transfected with AT1R along with βarr2-DB and stimulated with the indicated concentrations of AngII for 20 min. Data represent means ± SEM from nine independent experiments.

Characterizing downstream G protein–mediated signaling events using new pathway BRET sensors

We next examined Gαq signaling events by engineering new BRET sensors that detect the activity of effectors downstream of this G protein: PLC, PKC, the guanine nucleotide exchange factor (GEF) p63RhoGEF, and the GTPase Rho (Fig. 2, A to D). We assessed these responses in HEK293 cells downstream of endogenous G proteins. Because these cells do not produce endogenous AT1R, we transfected the sensor constructs into HEK293 cells stably producing a small amount of AT1R (around 0.5 pmol/mg; hereafter referred to as HEK293/AT1R). This amount of AT1R is comparable to that in vascular smooth muscle cells (VSMCs) (32).

Fig. 2 Generation and validation of BRET-based sensors for activation of PLC, PKC, p63/Gαq, and Rho.

(A) Schematic diagram of the DAG BRET sensor, which measures the generation of DAG by activated PLC. The recruitment of c1b to the PM by DAG increases the BRET signal. (B) Schematic diagram of the PKC-c1b BRET sensor. Phosphorylation of threonine residues in two PKC consensus sequences (TLKI and TLKD) causes a conformational change in the sensor due to interactions between the phosphothreonines and the phosphothreonine-binding domains (FHA1 and FHA2), producing a BRET signal. (C) Schematic diagram of the p63/Gαq BRET sensor. Upon AT1R stimulation, Gαq-RlucII dissociates from Gβγ and binds to a minimal PH domain of p63RhoGEF (p63BD) fused to GFP10 (p63BD-GFP10). (D) Illustration of the BRET sensor for monitoring Rho activation. The recruitment of the RlucII-tagged Rho-binding domain (RBD) of PKN (PKN-RBD-RlucII) to the PM after Rho activation increases bystander BRET with the membrane-anchored rGFP-CAAX. (E) Pharmacological validation of the DAG sensor. HEK293/AT1R cells expressing the DAG sensor were pretreated in the absence (DMSO) or presence of either the Gαq inhibitor UBO-QIC or the PKC inhibitor Gö6983 and then stimulated with either AngII for 70 s or PMA for 10 min. Data are means ± SEM from at least three independent experiments. DMSO, dimethyl sulfoxide. (F) Pharmacological validation of the PKC-c1b sensor. HEK293/AT1R cells expressing the PKC sensor were treated as in (E). Cells were also stimulated with forskolin (Fsk) for 10 min to activate PKA. Data represent mean ± SEM of at least three independent experiments. (G) Time course of AngII-mediated p63 recruitment to Gαq. HEK293 cells were transfected with AT1R along with p63BD-GFP10 and Gαq-RlucII, preincubated with the vehicle (DMSO) or UBO-QIC before stimulation with AngII (arrow), before BRET measurements. Data represent means ± SEM of triplicate in a representative experiment that was repeated three times with similar results. (H) Rho activation profiles in cells with compromised G protein signaling. Parental HEK293 cells and CRISPR Gq/11 or G12/13 cells (ΔGq/11 and ΔG12/13, respectively) were transfected with PKN-RBD-RlucII and rGFP-CAAX along with AT1R. Cells were incubated with or without UBO-QIC (UBO) and stimulated with the indicated concentrations of AngII before BRET measurements. Data represent means ± SEM of three independent experiments.

To measure the activity of PLC, which cleaves phosphatidylinositol 4,5-bisphosphate (PIP2) and releases diacylglycerol (DAG) as one of the end products, we devised a sensor that detects DAG generation at the PM. This sensor was created by introducing the dual acylation (myristoylation and palmitoylation) signal from the kinase Lyn at the N terminus of GFP10, followed by a 300–amino acid disorganized linker (DIS300), RlucII, and finally the c1b DAG-binding domain of PKC-δ at the C terminus (Fig. 2A). Upon DAG production, the c1b domain is recruited at the PM, allowing RlucII to be in close proximity to GFP10, thus producing an increased intramolecular BRET signal (Fig. 2A). As revealed by microscopy, the DAG BRET sensor localized to the PM, anchored through its fatty-acylated moieties (fig. S2A). Treating cells with phorbol 12-myristate 13-acetate (PMA), a mimic of DAG, was sufficient to promote the recruitment of the c1b domain to the PM, bringing the RlucII and GFP10 moieties into closer proximity and generating a robust BRET signal (Fig. 2, A and E). AngII stimulation of HEK293/AT1R cells expressing the DAG sensor led to a rapid increase in BRET within 30 s of stimulation (Fig. 2E and fig. S2B), consistent with the known time frame of Gαq activation (29). As expected, the BRET signal then gradually decreased toward baseline over 5 min. AT1R-mediated DAG generation was significantly inhibited by the selective Gαq inhibitor UBO-QIC but not by the pan-PKC inhibitor Gö6983, which inhibits both conventional and novel PKCs (cPKC and nPKC, respectively), consistent with a Gαq-dependent activation mechanism (Fig. 2E). To test the ability of the sensor to detect the activity of endogenous receptors, we took advantage of the fact that HEK293 cells produce endogenous muscarinic acetylcholine receptors (mAChRs) (33), including the M3 receptor subtype that couples to Gαq/11. Carbachol stimulation of these cells containing the DAG sensor led to a concentration-dependent increase in the BRET response (fig. S2C), confirming that the sensor is sufficiently sensitive to detect signals promoted by endogenous amounts of receptors.

To assess the activity of PKC without overexpressing the kinase itself, which may affect receptor activity, we generated an intramolecular BRET sensor that consists of an N-terminal GFP10 moiety followed by the FHA1 and FHA2 phosphothreonine binding domains of Rad53, two cassette sequences containing threonine residues in the context of PKC consensus sites (34), RlucII, and the DAG-binding c1b domain at the C terminus (Fig. 2B). A 50–amino acid DIS linker (DIS50) separates the FHA domains and the phosphothreonine cassettes to allow the flexibility necessary for the FHA domains to bind the PKC-phosphorylated sites, which would bring the GFP10 and RlucII moieties into close proximity, yielding an increased BRET signal (Fig. 2B). The c1b domain was added to the PKC BRET sensor in the C terminus to bind DAG at the PM once this second messenger signaling lipid is produced after receptor activation (35). This localized the biosensor in the vicinity of PKC subtypes known to be activated by AT1R (3639). A Strep-tag II sequence was also inserted in the N terminus in front of the GFP10 to purify the sensor. This sensor is hereafter referred to as the PKC-c1b sensor. Under basal conditions, the PKC-c1b sensor was found in both the cytosol and nuclei of HEK293 cells (fig. S3A). AngII stimulation of cells led to a rapid translocation of the PKC-c1b sensor from the cytosol to the PM with no obvious changes in the nuclear signal at the time point studied and an increase in BRET signal (fig. S3, A and B). Activation of the PKC-c1b sensor after AT1R stimulation resulted in a similar kinetic response as the DAG sensor, with the BRET signal peaking at 1 min and then returning to basal after 5 min (figs. S2B and S3B). The phosphorylation of the PKC-c1b biosensor upon either AngII or PMA activation was confirmed by Western blot analysis using an antibody recognizing phosphothreonine after affinity purification of the biosensor (fig. S3C). No increase in the BRET signal was observed with a mutant form of the sensor lacking the threonine residues in the cassette domains (fig. S3D). Direct PMA activation of endogenous PKCs detected by the PKC-c1b BRET sensor was inhibited by Gö6983, but not by UBO-QIC, whereas forskolin-mediated activation of AC and PKA did not promote any activation of the sensor, confirming the selectivity of the sensor for PKCs (Fig. 2F). The AT1R-promoted activation of the PKC-c1b sensor depended on activation of Gαq/11, but not on activation of Gαi, as illustrated by the inhibition of the BRET response by the Gαq/11 inhibitor YM-254890, but not by the Gαi inhibitor pertussis toxin (PTX) (fig. S3E). The PKC-c1b sensor reported on the activity of all PKCs, with PKCβ accounting for half of the response, as revealed by the reduction in the BRET signal in the presence of the selective PKCβ inhibitor LY-333,531 (fig. S3E). Mobilizing intracellular Ca2+ in a Gαq/11- and PLC-independent manner using the ionophore A23187 activated the PKC-c1b sensor, but to a lesser extent than did the agonist-mediated, DAG- and Ca2+-dependent activation of the sensor (fig. S3, E and F), indicating that the sensor was sufficiently sensitive to detect PKC activated solely by Ca2+ and likely relied on the presence of DAG in the membrane under basal conditions. To determine the relative role of the two components of the sensor activation (the translocation to the PM through binding to DAG and the phosphorylation-dependent conformational rearrangement), we designed a PKC sensor constitutively anchored to the PM through the introduction of the dual acylation domain of Lyn in the N terminus of the sensor, hereafter referred to as Lyn-PKC (fig. S3, G to I). The agonist-mediated activation and kinetics of the constitutively (Lyn) anchored and DAG-recruited (c1b) PKC sensors were identical (fig. S3, G and H). However, the Lyn-PKC sensor generated a larger signal, indicative of a greater sensitivity. As was the case for the PKC-c1b, Lyn-PKC was also detected in the nucleus (fig. S3I). As expected, the Lyn-PKC constitutively present at the PM and nucleus did not relocate upon AT1R activation. The causes for the nuclear localization of both PKC-c1b and Lyn-PKC sensors are unknown, but it did not prevent the detection of receptor-promoted PKC-c1b activity even in response to activation of endogenous (not overexpressed) receptors. Similar to the DAG BRET sensor, the PKC-c1b sensor detected the activity of endogenous mAChRs in HEK293 cells (fig. S2C).

The GEF p63RhoGEF, which activates the small G protein RhoA, is also a downstream effector of Gαq (40, 41). We generated a new BRET-based biosensor that detects Gαq activation and engagement of this effector using the minimal PH domain of p63RhoGEF that binds Gαq (40, 42), hereafter referred to as p63 binding domain (p63BD). Pull-down experiments using a glutathione S-transferase (GST)–tagged form of p63BD (GST-p63BD) revealed that p63BD interacted specifically with Gαq but not with Gα12 (fig. S4A). We thus attached the p63BD through its C terminus to GFP10, which when recruited to RlucII-Gαq would generate a BRET signal, and refer to this BRET pair as the p63/Gαq sensor (Fig. 2C). The BRET signal from the p63/Gαq sensor increased over time and reached a maximum after 60 s of agonist stimulation, and the signal lasted over 5 min (Fig. 2G). Its activation was totally blocked by UBO-QIC. The p63/Gαq sensor was not as robust as that from the PKC-c1b sensor and required higher amounts of receptor for activation (compare Fig. 2F to 2G). The greater sensitivity of the PKC-c1b sensor is well illustrated by both its greater maximal BRET response and the steeper increase in signal observed upon stimulation with increasing receptor concentrations (fig. S4B). This is perhaps reflective of the amplified nature of the PKC response versus that of the p63/Gαq sensor, which requires a stoichiometric recruitment of p63BD-GFP10 to RlucII-tagged Gαq, or differences in the sensitivity between these two sensors due to their specific intrinsic design, or both. The p63/Gαq BRET sensor could be activated by receptors known to couple to Gαq/11, such as AT1R and the prostaglandin F2α receptor (FP), but not by receptors known to signal through Gαi [the dopamine D4 receptor (D4R)] or Gαs [the β2-adrenergic receptor (β2AR)] (fig. S4C). In addition, we detected no BRET signal between p63BD-GFP10 and either Gαi2-RlucII or Gαs-RlucII upon activation of the D4R or β2AR, respectively, nor between Gα12-RlucII and p63BD-GFP10 upon activation of AT1R or FP (fig. S4C). Last, consistent with the selectivity shown by the pull-down experiment with GST-p63BD (fig. S4A), YM-254890 blocked the activation of the p63/Gαq BRET sensor by both AT1R and FP, but PTX did not (fig. S4D), also confirming the selectivity of the p63/Gαq BRET sensor for detecting the activation of Gαq-coupled receptors.

To further explore downstream effectors, we generated a sensor to evaluate the activity of the GTPase Rho. We fused the C terminus of the RBD of PKN, which is recruited to the PM upon Rho activation, to RlucII (PKN-RBD-RlucII). We monitored the recruitment of PKN-RBD to the PM by coexpressing PKN-RBD-RlucII with the green fluorescent protein from Renilla reniformis (rGFP) anchored at the PM through prenylation of the CAAX domain of kRas (rGFP-CAAX) (Fig. 2D), which generates an enhanced bystander BRET signal (43) upon PKN-RBD-RlucII translocation to the PM. In HEK293 cells expressing AT1R and the Rho sensor (PKN-RBD-RlucII plus rGFP-CAAX), AngII increased the BRET ratio within 30 s of agonist addition, and the signal persisted for more than 5 min (fig. S5A). Both basal and agonist-mediated BRET responses were significantly reduced by incubating cells with the Rho inhibitor C3 toxin (fig. S5B) (44), confirming the selectivity of the response. Because Rho can be activated by both Gαq/11- and Gα12/13-coupled receptors (45, 46), we evaluated the contributions of these two G protein subfamilies to Rho activation downstream of AT1R using pharmacological and genetic approaches. Either inhibiting Gαq/11 (with UBO-QIC or YM-254890) or using CRISPR-Cas9 Gαq/11 knockout cells (31) reduced the AngII-mediated response by 30 to 40%, without affecting receptor abundance (Fig. 2H and fig. S5C). Consistent with the absence of Gαq/11 in the CRISPR-Cas9–generated cells, UBO-QIC or YM-254890 had no further inhibitory effects on Rho sensor activity in these cells (Fig. 2H). Selective Gα12/13 pharmacological inhibitors are not available, but removing both these G proteins in cells through CRISPR-Cas9 [G12/13 knockout cells (47)] reduced the activation of the Rho sensor by 15 to 20% (Fig. 2H). Adding UBO-QIC to Gα12/13-depleted cells nearly completely blocked the activation of the Rho sensor by AngII (≈95% inhibition; Fig. 2H), demonstrating a role for both Gα12/13 and Gαq/11 in Rho activation. Consistent with our BRET data, we found that both Gα12/13 and Gαq/11 subtypes of heterotrimeric G proteins were involved in Rho activation by AT1R in HEK293 cells (fig. S5, D and E), using a classical GST pull-down assay to monitor Rho activation (48). In addition, to confirm the contribution of the two G proteins to Rho activation in living cells, these data validate the usefulness of the BRET sensor to monitor the relative contributions of each G protein to Rho activation. Together, our findings reveal that these BRET sensors are sensitive surrogates for detecting signaling of heterologously expressed Gαq/11- and Gα12/13-coupled GPCRs in cells. We also tested the PKC-c1b and Rho sensors for endogenous AT1R activity using VSMCs. Similar kinetics and magnitudes of response of PKC-c1b and Rho activation as found in HEK293 cells were also observed in VSMCs (Fig. 3, A and B). Activation of these sensors by AngII in VSMCs increased dose-dependently and was dependent of AT1R and Gαq/11 (Fig. 3, C to F)

Fig. 3 Activation of PKC and Rho sensors in VSMCs.

(A and B) Time course of AngII-mediated activation of virally expressed PKC-c1b (A) and Rho (PKN-RBD-RlucII plus rGFP-CAAX) (B) BRET sensors in VSMCs. Arrow indicates addition of AngII. Data represent means of triplicate in a representative experiment that was repeated three times with similar results. (C and D) Concentration-response curves for activation of the PKC-c1b (C) and Rho (D) BRET sensors in VSMCs by the indicated concentrations of various AngII analogs. BRET signals were normalized to that induced by AngII in the same experiment and expressed as %Emax of AngII and then averaged. Data are means ± SEM of at least three independent experiments. (E and F) Validation of AngII-induced, AT1R-mediated PKC and Rho activation in VSMCs. Cells expressing the PKC-c1b (E) or Rho (F) BRET sensor were preincubated with vehicle, the AngII type 2 receptor (AT2R) antagonist PD 123319, the AT1R antagonist losartan, or the Gαq inhibitor YM-254890 (YM) and then stimulated with or without AngII. Data represent means ± SEM of AngII-mediated changes in the BRET signal (ΔBRET) derived from three to five independent experiments.

Profiling G protein and β-arrestin activation by various AngII analogs

Next, we used the BRET-based sensors to examine the diversity of G protein and β-arrestin activation profiles by 14 AngII analogs including AngII itself (table S1). These analogs were selected on the basis of their partial agonist activities (49) and, for some, their reported bias for inducing activation of β-arrestin versus Gαq [such as SI, SVdF, SBpa, SII, DVG, TRV027, Ang(1–7), and AngIII] (18, 20, 50). By selecting these ligands, we also sought to assess the impact of modifying amino acid positions 1 and 8 of AngII on the overall signaling activity of AT1R. Other analogs, including Sarmesin, SII, and SIII, were selected to also assess the contribution of position 4. We generated concentration-response curves for Gαq, Gαi2, Gαi3, Gα12, and βarr2 activation in HEK293/AT1R cells and compared the potencies and efficacies of each ligand to AngII (Fig. 4, A to E, and Table 1). The AngII analogs did not significantly decrease BRET signals over basal in HEK293 cells lacking AT1R (fig. S1A). In cells overexpressing AT1R, however, AngIII and [Val4]-AngIII showed full agonist activity on Gαq, whereas Sarmesin, [Val5]-Sarmesin, SBpa, SVdF, and SI showed partial activity (Fig. 4A). Saralasin and SIII only marginally activated Gαq, whereas TRV027, DVG, Ang(1–7), and SII were deemed inactive on Gαq, because we detected no reliable signal from concentration-response curves (Fig. 4A). We investigated the activity of AngII analogs on AT1R-mediated Gαi2, Gαi3, and Gα12 responses (Fig. 4, B to D, and Table 1). The ligands that did not activate Gαq [TRV027, DVG, Ang(1–7), and SII] had mild-to-strong partial agonist activities on Gαi2, Gαi3, and Gα12, with Gα12 being the G protein most activated by all four ligands (Fig. 4, B to D). The rank order of potency for activating Gα12, Gαi2, and Gαi3 by Gαq-inactive ligands was identical [TRV027 ≥ DVG > SII > Ang(1–7)]. Saralasin acted as a partial agonist for all three G proteins, with a relatively better efficacy for activating Gα12 over Gαi2 and Gαi3, but had greater potency for engaging Gαi2 and Gαi3 than for engaging Gα12 (Fig. 4, B to D). SIII had similar potencies for activating Gαi2, Gαi3, and Gα12 but better efficacy on Gα12 as compared to Gαi2 and Gαi3 (Fig. 4, B to D). The partial Gαq agonists Sarmesin, [Val5]-Sarmesin, SBpa, and SVdF all act as partial agonists on Gαi2, Gαi3, and Gα12 activation, with similar efficacies and potencies (Fig. 4, A to D). SI, which showed weak Gαq activity, was the least efficacious activator of Gαi2, Gαi3, and Gα12, whereas the full Gαq agonists AngIII and [Val4]-AngIII retained their full agonist properties on these G proteins. Last, despite their reduced efficacies on Gαq activation, Sarmesin, [Val5]-Sarmesin, SBpa, SVdF, and SI all activated βarr2 with efficacies and potencies comparable to that of AngII (Fig. 4E and Table 1). The non–Gαq-activating ligands TRV027, DVG, Ang(1–7), and SII also all promoted βarr2 recruitment to a similar extent as did AngII, albeit with different potencies (Fig. 4E, right panel, and Table 1). The two full Gαq agonists AngIII and [Val4]-AngIII were also full agonists of the βarr2 response. Together, these data suggest that many AngII analogs preferentially stimulate β-arrestin over Gαq signaling downstream of AT1R.

Fig. 4 Concentration-response curves for G protein and β-arrestin activation by AngII analogs.

(A to E) HEK293 cells were transiently transfected with DNA encoding the Gαq (A), Gαi2 (B), Gαi3 (C), Gα12 (D), or βarr2 (E) BRET sensor along with AT1R and stimulated with the indicated concentrations of AngII or various AngII analogs. BRET measurements were recorded and normalized to the response of AngII in the same experiment and expressed as %Emax of AngII. Data are means ± SEM of at least three independent experiments.

Table 1 Potency and relative efficacy (Emax) of AngII and AngII analogs for activating G protein and βarr2 signaling.

HEK293/AT1R cells expressing each indicated BRET sensor were stimulated with various concentrations of AngII and AngII analogs. BRET signals from each sensor were normalized to the maximal response of AngII (%Emax of AngII) and then averaged. pEC50 and Emax were obtained from the nonlinear regression curve of the averaged data. Data represent means ± SEM of three to eight independent experiments. n.d., not determined due to lack of responses.

View this table:

We next evaluated potential pathway-specific effects of the different AngII analogs. We used the operational model (6) to quantitatively determine any bias between two signaling pathways for the tested ligands as compared to the reference ligand, AngII (Table 2 and data file S1). To better visualize ligand rank order, we generated a heat map of the relative activity of the response to each ligand as compared to the response to AngII (Fig. 5A). The relative activity [Δlog(τ/KA), where KA represents the functional affinity of the ligand for the receptor and τ is its efficacy in activating a signaling pathway] is determined from the difference in transduction coefficient of ligands, as compared to that of the reference ligand AngII. Because no reliable Emax or EC50 (median effective concentration) could be determined for TRV027-, DVG-, Ang(1–7)-, and SII-mediated activation of Gαq, we arbitrarily assigned them log(τ/KA) values equal to the weakest Gαq-activating ligand (SIII) for which such coefficient could be determined (Table 2). Similar clustering and rank orders were observed when the effectiveness of responses to the ligands was analyzed according to either the Emax or pEC50 (pEC50, representing the negative log of the concentration of the ligand that gives EC50; fig. S6, A and B). Analysis of Δlog(τ/KA) revealed significant bias responses against Gαq in favor of Gαi2, Gαi3, Gα12, and β-arrestin for SBpa, SVdF, SI, Saralasin, TRV027, DVG, Ang(1–7), SII, and SIII (Fig. 5A and data file S1). As expected from the heat map clustering of the ligands’ relative activity, we observed no strong bias between their abilities to activate Gαi2, Gαi3, Gα12, and β-arrestin, although a small bias, in favor of Gα12, was detected for SBpa, [Val5]-Sarmesin, and SVdF between Gαi3 and Gα12. Sarmesin showed functional selectivity in favor of β-arrestin over Gαq, but not between Gαi2, Gαi3, or Gα12, as compared to Gαq. Ang(1–7) showed a bias against Gαq, and Gα12 while favoring Gαi2, Gαi3, and β-arrestin, but did not distinguish between Gαi2, Gαi3, and β-arrestin. These results reveal a diversity of signaling profiles that had not been previously appreciated and demonstrate that such clustering can be useful to identify pathway-selective ligands with similar biased signaling properties.

Table 2 Transduction ratio and relative effectiveness of AngII and AngII analogs for activating AT1R downstream pathways.

Concentration-response data for each ligand were analyzed by nonlinear regression using the operational model equation in GraphPad Prism with AngII as the reference ligand, as described previously (29). ΔLog(τ/KA)s were calculated by subtracting the log(τ/KA) value of AngII in each pathway (Eq. 1). The SEs of Δlog(τ/KA) were estimated by Eq. 3 as described in Materials and Methods. Data represent means ± SEM of three to eight independent experiments. Relative effectiveness (RE) of the ligand toward each pathway, relative to AngII, was determined using Eq. 2.

View this table:
Fig. 5 Heat map of AT1R signaling signature of various AngII analogs.

(A and B) The transduction coefficients [log(τ/KA)] of each AngII analog were calculated from the concentration-response curves. The relative activity of each ligand [Δlog(τ/KA)] represents the difference between the calculated transduction coefficient [log(τ/KA)] for each ligand and the transduction coefficient of reference ligand (AngII). The relative activity [Δlog(τ/KA)] of AngII analogs in each signaling pathway was expressed as a heat map for (A) G protein signaling and β-arrestin activation and (B) Gαq signaling and activation of its downstream effectors p63RhoGEF (p63/Gαq), PLC (DAG generation), and PKC.

Assessing G protein–mediated effector signaling by different AngII analogs

We assessed the coupling efficiencies of the 14 different AngII analogs on the downstream effectors of Gαq/11 to assess whether their relative activities toward this G protein were also propagated to downstream effectors. We first generated concentration-response curves with the DAG, PKC-c1b, and p63/Gαq sensors for the different ligands in HEK293/AT1R cells (fig. S7). These concentration-response curves for many ligands mimicked those observed for Gαq activation (compare Fig. 4A and fig. S7, A to D). We also observed a similar rank order for the Gαq response to partial agonists in VSMCs, although the amplitude of the signal was lower than in HEK293 cells (Fig. 3C). We performed the corresponding heat map analysis of the relative activities for each ligand in HEK293 cells (Fig. 5B and tables S2 and S3). Again, we arbitrarily assigned the lowest transduction coefficients for TRV027 and Ang(1–7) for the activation of p63/Gαq because we could not extrapolate reliable efficacies or potencies for these ligands. Similarly, we could not determine the transduction coefficient for Saralasin, TRV027, DVG, Ang(1–7), SII, or SIII for the PKC and DAG responses nor for PKC activation by SI (table S3). The heat map analysis of relative activities or the Emax and pEC50 for activating p63/Gαq, DAG, and PKC mirrored that of Gαq activation, but not that of Gαi2, Gαi3, and Gα12, consistent with Gαq/11 playing a primary role in the activation and generation of these effectors and second messengers (Fig. 5B and fig. S8, A and B). Consistent with these observations, we also found a stronger correlation between the transduction coefficients for the different ligands for activating Gαq versus p63/Gαq (fig. S9A), DAG (fig. S9B), and PKC (fig. S9C) than for other G proteins and β-arretin2 with those effectors (fig. S9, A to C, right panels). These findings indicate that the bias of signaling flow through downstream of Gαq was maintained and that the efficiency toward the other pathways had little impact on the coupling of Gαq effectors.

We next generated concentration-response curves of AngII analogs in HEK293 cells transfected with both AT1R and the Rho sensor (Fig. 6, A to D). Because we found that both Gq/11 and G12/13 contributed to the activation of Rho in HEK293 cells, we also assessed the coupling efficiency of the AngII analogs on the Gα12/13-mediated activation of Rho. We stimulated cells expressing the Rho sensor with the different AngII analogs and compared the concentration-response curves and their coupling efficiencies in the presence or absence of the Gαq/11 inhibitor UBO-QIC (Fig. 6, A to D, and tables S4 and S5). UBO-QIC was used to isolate the contribution of Gα12/13 in Rho activation. Rho activation by AngII, AngIII, Sarmesin, SBpa, SVdF, SI, or SII was significantly reduced in cells treated with UBO-QIC (Fig. 6E), suggesting—although to different degrees—the contribution of both Gαq/11 and Gα12/13 in the activation of this effector. Predictably from our analysis, which revealed a stronger bias for the activation of Gα12 over Gαq for Saralasin, TRV027, and DVG among the AngII analogs (data file S1), we observed no significant effects of UBO-QIC on the activation of Rho by these ligands (Fig. 6, A to E). Despite the undetectable efficacy of SII for activating Gαq (Fig. 4A), we nonetheless observed a modest inhibitory effect of UBO-QIC on Rho activation, suggesting the involvement of both Gαq/11 and Gα12/13. These results illustrate that, notwithstanding AngII analogs activating Rho with varied efficacies in HEK293 cells, they do so by differentially engaging Gαq/11 and Gα12/13. Inhibiting Gαq/11 differentially affected the relative effectiveness of activation of Rho for some ligands, whereas for others it had no effect, implying the contribution of only Gα12/13 family members for these biased AngII ligands (fig. S10A). The biased nature of several ligands toward Gα12 was well illustrated by the stronger correlation observed between Gα12 and Rho activation when Gαq/11 was inhibited in these cells (Fig. 6F and fig. S10B, respectively). In VSMCs, however, the activation of Rho by AngII seems to mostly involve the engagement of Gαq/11 (Fig. 3, D and F).

Fig. 6 Rho activation upon AT1R stimulation by various AngII analogs.

(A to D) Concentration-response curves for Rho activation. HEK293 cells expressing the Rho BRET sensor (PKN-RBD-RLucII plus rGFP-CAAX) along with AT1R were pretreated with vehicle or UBO-QIC (UBO) and then stimulated with the indicated concentrations of AngII analogs. BRET signals were normalized to that of AngII in the absence of UBO-QIC and expressed as %Emax of AngII. Data represent means ± SEM from three to four independent experiments. (E) The relative activity of each ligand [Δlog(τ/KA)] represents the difference between the calculated transduction coefficient [log(τ/KA)] for each ligand and pretreatment conditions (vehicle or UBO-QIC) and the transduction coefficient of AngII with vehicle pretreatment. Data represent means ± SEM from three to four independent experiments. *P < 0.05 and **P < 0.01, unpaired Student’s t test. (F) Scatterplot of Δlog(τ/KA) of Rho activation in the presence of UBO-QIC versus Δlog(τ/KA) of Gα12 activation by the AngII analogs. R2 analysis was determined from a linear regression.

Investigating pathway selectivity of AT1R variants

Last, we used the BRET-based platform to study the impact of previously reported naturally occurring AT1R variants ( (5155) on signaling. We focused on nonsynonymous variants in the transmembrane (TM) regions of AT1R (Fig. 7A) because we reasoned that those would be the most likely to affect stabilization of distinct receptor conformations and hence have specific effects on agonist-mediated signaling. We selected the following five variants: I3.27T, A4.60T, A6.39S, T7.33M, and C7.40W [Ballesteros-Weinstein numbering (56); hereafter abbreviated I103T, A163T, A244S, T282M, and C289W, respectively]. We transfected constructs encoding each receptor variant into HEK293 cells along with individual BRET sensors and generated concentration-response curves for AngII stimulation to determine the EC50 and Emax values for each variant as well as for wild-type AT1R for Gαq, Gαi2, Gαi3, Gα12, PKC-c1b, Rho, and β-arrestin translocation to the PM and into endosomes (fig. S11, A to H, and Table 3). Because receptor abundance can affect coupling efficacies, we first examined the cell surface abundance of all constructs by enzyme-linked immunosorbent assay (ELISA; fig. S12A). T282M was as abundant as wild-type AT1R, whereas the abundance of A163T was increased by about 50%, and the three variants C289W, I103T, and A244S were reduced by 50% or less compared to wild-type AT1R. The differences in abundance in this relatively narrow range did not seem to significantly affect the coupling of receptors to Gαq, because all variants exhibited similar efficacy toward this G protein and because the potency was only marginally affected for T282M and C289W (fig. S11A). Consistent with unaffected coupling to Gαq, all variants exhibited similar efficacies and potencies to activate PKC-c1b (fig. S11E). For Gαi coupling, only modest effects of the variants were observed. The potency or efficacy, or both, was slightly reduced for C289W, T282M, A244S, and I103T (in the case of Gαi2) and for C289W and T282M (in the case of Gαi3) (fig. S11, B and C). For C289W, I103T, and A244S, these small differences may result from the reduced receptor abundance having a greater impact on activation of a G protein that is more weakly coupled to the receptor. For Gα12, a reduced efficacy or potency, or both, was also observed for A244S, I103T, and C289W, again consistent with an effect that could be a result of reduced receptor abundance (fig. S11D). However, the largest effect was observed for the T282M variant that showed a reduced potency of more than 100-fold compared to wild-type AT1R, a difference that cannot be attributed to a reduced amount of the protein because this variant exhibited cell surface abundance similar to the wild-type receptor (fig. S12A).

Fig. 7 Signaling profile of AT1R variants relative to the wild-type receptor.

(A) Serpentine structure of human AT1R (obtained from with the variant residues highlighted. (B) Scatterplot of the Δlog(relative activity) of wild-type and mutant AT1R for the indicated signaling outputs. Relative activities (Emax/EC50) were obtained from the AngII concentration-response curves and normalized to that of the wild-type receptor. Data represent means ± SEM from three to five independent experiments. Tukey’s post hoc multiple comparisons tests were used to compare wild-type and mutant receptors across the panel of assays (**P < 0.01), as well as to compare the signaling pathways downstream of each receptor (φP < 0.01).

Table 3 Binding affinity, potency, and relative efficacy of wild-type and variant AT1Rs for activating Gαq, Gαi2, Gαi3, Gα12, PKC, Rho, and βarr2.

HEK293 cells were transfected with wild-type (WT) or variant AT1R or along with the indicated BRET sensor (Gαq, Gαi2, Gαi3, Gα12, PKC-c1b, Rho, βarr2-PM, or βarr2-EE) and stimulated with various concentrations of AngII. BRET signals for each pathway were normalized to the maximal response of wild-type (%Emax of WT) and then averaged. pEC50 and Emax were obtained from the nonlinear regression curve of the averaged data. Data represent means ± SEM of three to five independent experiments. AngII binding affinities were obtained from [125I]-AngII saturation binding assays. Data represent means ± SD of two to three independent experiments performed in duplicate.

View this table:

We also observed reduced efficacies or potencies for A244S, I103T, C289W, and T282M in promoting βarr2 recruitment to the PM. The T282M variant also reduced the translocation of β-arrestin–receptor complexes to the endosome by more than 60% (fig. S11H and Table 3), as assessed using a previously described BRET-based trafficking sensor (43). Together, these data suggest that the relatively modest effects observed for A244S, I103T, and C289W most likely result from the modestly reduced cell surface abundance of the variants, whereas the T282M mutation had a major impact on both Gα12 and β-arrestin pathways (and to a lesser extent on Gαi and Gαq pathways) that could not be attributed to a reduction in the abundance of the receptor at the cell surface. The apparent biased effect of this variant was confirmed by estimating the relative activity, calculated from the ratio of maximal responses over the concentration for half-maximum responses (Emax/EC50) (Table 4). The relative activity of T282M toward Gα12 and βarr2 was substantially more affected than that of Gαq or Gαi. This contrasts with the other variants for which the relative activity for the different pathways was either not affected (A163T, I103T, and A244S) or affected to similar extents (C289W) compared to wild-type AT1R (Fig. 7B and Table 4). Of the panel of variants analyzed, only T282M revealed a biased signaling pattern. Because mutations in the receptor can also have an effect on ligand binding affinity, we measured the affinity of each of the variants for radiolabeled AngII (fig. S12B and Table 3). The ligand-binding affinities of I103T, A163T, and A244S were similar to that observed for wild-type AT1R; however, although we detected ligand binding for C289W and T282M, a loss in binding affinity prevented a reliable estimate of the Kd [dissociation constant (binding affinity)] for these variants. These two variants also showed the greatest impact on downstream signaling. Although the reduction in affinity can explain the reduced potency, it cannot explain the reduction in signaling efficacy observed for these variants. These results also suggest that T282M affected both the affinity for AngII and the transition toward active conformations, which are obviously interrelated.

Table 4 Relative activity of AT1R and mutant receptors for activating each signaling pathway.

Top: Relative activity [log(Emax/EC50)] of each AT1R variant obtained from the concentration-response curves from each individual experiment. Bottom: ΔLog(RA)s were calculated by subtracting the log(RA) value of the wild-type receptor for activating the same pathway (Eq. (5)). Data represent means ± SEM of three to five independent experiments.

View this table:


We report on the development and use of a suite of BRET-based biosensors that monitor G proteins, β-arrestin, and signaling effectors downstream of G proteins to study the functional selectivity of AT1R and naturally occurring receptor mutants. This resource is useful for (i) establishing signaling “fingerprints” for different AngII analogs, allowing a global characterization of their biased signaling properties; (ii) uncovering new biased activities for AngII ligands; (iii) showing the existence of functional selectivity for naturally occurring AT1R mutants; and (iv) demonstrating that the relative contribution of distinct G protein subtypes to a common downstream effector can vary from ligand to ligand, hence revealing a new level of bias.

The new BRET-based sensors monitor receptor-mediated, G protein–dependent engagement of different effectors: PLC (as measured by DAG production), PKC, p63RhoGEF, and Rho. When combined with previously described G protein sensors (2731), they allowed us to demonstrate that the relative biased activity of ligands for the receptor-mediated engagement of G proteins is conservatively propagated to their downstream effectors. This was best observed for the Gαq signaling pathway. We observed that biased ligands activated the PLC, PKC, and p63/Gαq sensors with lower efficacies as their ability to engage Gαq decreased. This suggests that the biased responses observed are intrinsic properties of different ligands on AT1R and did not result from the expression of the sensors. When the DAG, PKC, and Rho sensors were used to record the activities of endogenous G proteins (Gαq/11 and Gα12/13), we detected similar kinetics and relative efficiency rank order of the ligands as those observed with the G protein BRET sensors.

The DAG, PKC, and Rho BRET sensors detected amplified responses to receptor-mediated activation of G proteins. They were sensitive enough to detect responses from endogenous (not overexpressed) receptors, such as AT1R and mAChRs in VSMCs and HEK293 cells, respectively, and to demonstrate the coupling of overexpressed receptors to endogenous G proteins. The use of these BRET biosensors, in combination with pharmacological inhibitors and gene editing tools, also revealed their usefulness for dissecting biases displayed by certain ligands among different G protein subtypes in the engagement of the same effector. We show that some ligands, despite having similar efficiencies for activating Rho, stimulated this effector through the differential engagement of Gαq/11 versus Gα12/13. This was best illustrated by the stronger correlation observed between the relative effectiveness of the biased ligands for engaging Gα12/13 versus Rho when Gαq/11 was inhibited. Because both Gαq/11 and Gα12/13 are known to regulate vascular tone downstream of many vasoactive GPCRs in VSMCs, including AT1R (2, 57, 58), similarly biased ligands may have distinct effects on such cellular responses, as well as other functional consequences in cells producing different relative amounts of these or other G protein subtypes. Although Gαi is itself inhibitory to AC, the Gβγ subunits of Gαi-containing heterotrimeric G proteins can, in some cases, activate specific PLC subtypes, leading to PKC activation (59). However, our analysis of the sensors’ responses reveals a better correlation between the effectiveness of the different biased ligands for engaging Gαq/11 versus the PLC or PKC sensor than between Gαi2 versus these same sensors, again indicating that, in HEK293 cells, AT1R favors signaling through Gαq/11-PLCβ-PKC. These observations are consistent with the lack of effects of PTX, which inhibits Gαi subtypes, on the activation of the PKC sensor. These results obtained with our biosensor set, monitoring the activity at different steps of the signaling cascade for ligands with diverse biased properties, highlight the granular details of signaling flow that exists in these cells.

The activity of ligands can be influenced by the stoichiometry of receptors and G proteins, as well as by the efficacies or kinetics, or both, of G protein activation (25, 6062). Although we did not perform a systematic analysis on the sensor responses in relation to their abundance when expressed in cells, we nonetheless observed the same activation kinetics of Gαq and Gαi by the biased ligands, as well as between Gαq and its downstream effectors, which were also similar to those promoted by AngII. However, differences between the activation kinetics of Gαq and Gα12 were observed, with maximal activation occurring after 2 and 10 min of receptor stimulation, respectively. Such difference in kinetics has been previously observed, although the underlying cause remains unclear (63). This observation, however, had little impact on our analysis of the signaling induced by biased ligands, because they induced the activation of these G proteins similarly to AngII. Nonetheless, the activation of these G proteins may vary among ligands in other systems because of variations in the amount or identities of specific intracellular signaling components that are present in the cell, which may distinctly influence the signaling signature observed. These considerations might be of importance when using the BRET sensors for drug discovery. To evaluate the effect of drugs, they should be used in cell systems containing the appropriate complement of signaling components, and controls should be performed to assess the impact of the stoichiometric variation between receptors and G proteins on the different responses.

Biased agonism has been linked to the affinity or efficacy, or both, of the ligand-bound receptor for the downstream signaling proteins (64). These parameters were well captured in the transduction coefficients determined for each ligand but may have differentially contributed to the biased activity of the ligands and mutant. When comparing the apparent affinities (as determined in competition binding assays) of the different ligands for AT1R, we generally did not find clear relationships with their bias profiles. We previously reported the following rank affinities order: AngII >> SI = SVdF >> SBpa = DVG >>> SII (1 nM, 7 and 8 nM, 17 and 18 nM, and 200 nM, respectively) (18), but found different relative effectiveness for Gαq, Gαi, and Gα12 activation, as outlined in the heat map. For example, the rank order for Gαq is AngII >> SBpa > SVdF = SI >> DVG = SII, whereas for Gα12 it is AngII = SBpa > SVdF > SI > DVG >> SII (Fig. 5). This implies an efficacy-based biased agonism mechanism for many AngII ligands in G protein signaling. However, for β-arrestin, the affinities of the ligands for AT1R seem to contribute more to the bias than do their efficacies.

Mechanisms underlying receptor functional selectivity are not fully understood but have been proposed to involve, at least in part, the stabilization of different receptor conformational states by the ligands (68), which could lead to the binding of different G protein subtypes and β-arrestin with different affinities. In that respect, we previously showed that AngII-biased ligands promoted different conformations of the β-arrestin found in complexes with AT1R (18, 65), suggesting a unique stabilization of the receptor-effector complex conformation by different AngII-biased ligands. Similarly, differences in G protein subtype affinities for the agonist-bound muscarinic M3R that depended on the identity of the ligand have been reported (66).

Our findings also emphasize the existence of a greater than anticipated diversity in the biased signaling profiles among AngII analogs. Substitutions or removal of position 8 in the AngII analogs, such as in SBpa, SVdF, SI, TRV027, DVG, SII, and Ang(1–7), has been shown to bias these ligands toward β-arrestin signaling over Gαq/11 engagement (1820). SII was also shown to bias AT1R against Gαq but in favor of other G proteins, such as Gαi1, Gαi2, Gαi3, and Gα12/13 (17, 21). Our data reveal previously undetected biased activities for known biased ligands [SBpa, SVdF, SI, TRV027, DVG, and Ang(1–7)] as well as for other AngII ligands for which bias had not previously been reported (Sarmesin, Saralasin, and SIII). SBpa, SVdF, SI, TRV027, DVG, and Saralasin have high efficacies, compared to AngII, for activating Gαi2, Gαi3, and Gα12 relative to Gαq. The ability of the ligands to better engage β-arrestin rather than Gαq inversely correlated with the bulkiness of the residue in position 8 of AngII. Our findings with the Sarmesin peptides (where a methyl-tyrosine replaces the tyrosine at position 4 in AngII) are consistent with the importance of the fourth residue of AngII, as recently shown by the biased response to other signaling pathways of the analog [Sar1, Ile4]AngII at the expense of a reduction in its agonist activity on Gα12 as compared to AngII (21). They also demonstrate the importance of this position for the receptor-mediated activation of Gαi2 and Gαi3. Altogether, our data suggest that the presence of Phe at position 8 of AngII is required for stabilizing a conformation in AT1R compatible with its efficient binding and activation of Gαq, whereas the Tyr at position 4 is important for stabilizing a conformation for the activation of Gαq, Gαi, and Gα12 proteins. Notwithstanding that further structure-activity relationship studies on the AngII peptide would be needed to better understand the contribution of each amino acid residue in the biased responses of AT1R, our initial analysis suggests the possibility of developing new classes of AngII ligands with specific biases.

Our findings using BRET sensors and different AngII ligands are compatible with the ability of AT1R to couple to different G proteins in vivo. Although AngII is known to activate Gαq/11 in different tissues, it has been shown to promote AT1R coupling to PTX-sensitive Gαi/o proteins in the rat adrenal glomerulosa, liver, kidney, and pituitary glands and to Gα12/13 proteins in rat portal vein myocytes (26, 57, 58). These observations suggest that some of the AngII-biased ligands described here may have different actions on target tissues, depending on their ability to engage different G proteins and downstream effectors as well as on the relative abundance of these proteins in these tissues.

The octapeptide TRV027, which has been developed for therapeutic use (23) and has been described as a β-arrestin–biased ligand, much like SII (15, 16), also promoted the activation of other G proteins (for example, Gαi2, Gαi3, and Gα12) with somewhat better efficacies than did SII. To what extent the activation of these G proteins contributes to the cardioprotective properties of TRV027 in preclinical models remains an open question. Ang(1–7) is an active circulating metabolite of AngI and AngII that has been proposed to act as a physiological antagonist of AngII, but its main endogenous target(s) still remains open to debate (67). Evidence suggests that Ang(1–7) operates in vivo through the Mas oncogene and Mas-related receptors (Mas-R and MrgD, respectively), but recent findings by us and others demonstrate that it also acts as a β-arrestin–biased agonist for AT1R (20, 50). Consistent with these findings, we show here that Ang(1–7) not only acts as a β-arrestin–biased ligand but also biases AT1R signaling to G proteins other than Gαq with a somewhat similar effectiveness and functional selectivity profile as SII. However, as for SII and TRV027, the physiological implications of such signaling biases by Ang(1–7) are yet to be determined.

Our findings on the functional selectivity of AT1R variants are also compatible with a mechanism whereby biased signaling is conferred by the stabilization of distinct receptor conformations. In that respect, substituting a Thr for a Met residue in AT1R at position 282 generated a biased receptor that most likely was stabilized in a conformation that reduces its efficacy toward Gα12 and β-arrestin activation more than for the other G proteins. Cocrystallization studies using nonpeptidergic antagonists of AT1R have revealed that this amino acid in the N-terminal region of TM domain 7 (TM7) is located close to the binding domain of these antagonists (54). No crystal structure of AT1R with AngII exists, but molecular dynamic and docking studies suggest that position 282 is located near the AngII orthosteric binding site (68, 69). This is consistent with the observed loss of affinity the T282M variant has for AngII. Because TM6 and, to a lesser extent, TM7 reorientation upon agonist binding to GPCRs has been shown to be important for the functional coupling of receptors to G proteins (70, 71), it is tempting to speculate that mutating the Thr at position 282 to Met affects transitional changes in the TM7 conformation. The greater loss of AngII potency to activate Gα12 and β-arrestin than to activate Gαq observed in the T282M variant suggests that this position plays a particular role in driving the active conformation preferred for the engagement or activation of both Gα12 and β-arrestin. These findings are also consistent with the recent suggestion that the binding of Gα12 and β-arrestin to the receptor is functionally linked (63).

This new suite of BRET sensors presented here helped define the signaling pathways downstream of AT1R and should be useful in studying the signaling of other receptors for which coupling to G proteins remains incompletely characterized. These tools have been used to analyze Gαq/11-PLC-PKC signaling downstream of the GPCR Frizzled 5 (FZD5) in response to the ligand WNT-5A (72). The use of these biosensors in combination with the pharmacological analysis of biased agonism for AT1R has also led to the generation of specific “fingerprints” for different ligands and naturally occurring receptor mutants. They allowed us to cluster signaling behaviors of GPCR ligands and mutants, which, when compared to responses in relevant tissues, should also help better understand the relationship between different signaling pathways and the physiological or pathophysiological responses downstream of receptors. Genomic studies have now started to reveal the preponderance of polymorphisms in GPCRs (11, 73) and, in some cases, their impact on drug efficacies. However, such efforts are limited by the lack of resources needed to mechanistically understand the impact of these genetic variations on signaling. Our suite of BRET biosensors should not only aid in reducing this gap but also provide insight on the impact these mutations have on receptor signaling and drug efficacy and hence facilitate the decision-making about which drugs to be used or developed for a more personalized medicine.



Some of the angiotensin ligands used here were described elsewhere (18). The sequences of all ligands used in this study are listed in table S1. AngII (Sigma), DVG (NeoBioLab), and Ang(1–7) (Bachem) were purchased from commercial suppliers. All other AngII analogs were synthesized at the Université de Sherbrooke (Quebec, Canada). Iodine-125 was obtained from PerkinElmer. Dulbecco’s modified Eagle’s medium (DMEM), minimal essential medium (MEM), fetal bovine serum (FBS), and other cell culture reagents were purchased from Gibco, Life Technologies. Coelenterazine 400a was purchased from NanoLight Technology and Gold Biotechnology. UBO-QIC (31) [l-threonine,(3R)-N-acetyl-3-hydroxy-l-leucyl-(aR)-a-hydroxybenzenepropanoyl-2,3- idehydro-N-methylalanyl-l-alanyl-N-methyl-l-alanyl-(3R)-3-[[(2S,3R)-3-hydroxy-4-methyl-1-oxo-2-[(1-oxopropyl)amino]pentyl]oxy]-l-leucyl-N,O-dimethyl-,(7→1)-lactone(9CI)] was purchased from Institute for Pharmaceutical Biology of the University of Bonn (Bonn, Germany). YM-254890 was purchased from FUJIFILM Wako Chemicals U.S.A. Phusion DNA polymerase was from Thermo Scientific. Restriction enzymes, T4 DNA ligase, and Gibson assembly mix were obtained from New England Biolabs. Oligonucleotides were synthesized at Integrated DNA Technologies. Strep-Tactin sepharose and Strep-Tactin conjugated to horseradish peroxidase (Strep-Tactin HRP) were purchased from IBA GmbH. The anti-phosphothreonine (#9381) and anti-RhoA (#2117) antibodies were purchased from Cell Signaling Technology. Anti–Renilla luciferase antibody (MAB4400) was obtained from EMD Millipore. Bradford protein assay was from Bio-Rad. Gö6983 and Gö6976 were from Calbiochem, and LY333531 was from Tocris Bioscience. PD 123319, PTX, and A23187 were purchased from Sigma. Glutathione Sepharose 4B was from GE Healthcare.

BRET biosensor constructs

The polycistronic Gαq BRET sensor (29), Gα12(136)-RlucII, Gαi1(91)-RlucII (30), Gαi2-RlucII (30), Gαi3-RlucII (30), GFP10-Gγ1 (74), GFP10-Gγ2 (74), Flag-Gβ1 (29), GFP10-βarr2-RlucII (29), βarr2-RlucII, rGFP-FYVE and rGFP-CAAX (43), and signal peptide–Flag–tagged human AT1R (sp-Flag-AT1R) were described previously (32). DNA encoding the c1b domain of PKCδ, the FHA1 and FHA2 domains of yeast Rad53, and the DIS300 linker were codon-optimized and synthesized by GenScript. The DIS300 linker and DIS50 are artificially designed linkers 300 and 50 residues in length, respectively. The DIS300 linker was created by generating a random sequence respecting the composition of natural disordered sequences. The random sequence was of a few thousand residues, and the sequence was analyzed for the absence of any nuclear localization sequence motif, for phosphorylation and known protein-protein interaction sites, and for its predicted globularity. Structure prediction was used to identify the most disordered stretches of the sequence, and the best 300- and 50-residue-long sequences were selected. The DIS300 sequence is as follows: KEGEKQKGAMQPSEQQRGKEAQKEKNGKEPNPRPEQPKPAKVEQQEDEPEERPKREPMQLEPAESAKQGRNLPQKVEQGEERPQEADMPGQAQSAMRPQLSNSEEGPARGKPAPEEPDEQLGEPEEAQGEHADEPAPSKPSEKHMVPQMAEPEKGEEAREPQGAEDKPAPVHKPKKEEPQRPNEEKAPKPKGRHVGRQENDDSAGKPEPGRPDRKGKEKEPEEEPAQGHSLPQEPEPMPRPKPEVRKKPHPGASPHQVSDVEDAKGPERKVNPMEGEESAKQAQQEGPAENDEAERPERP. The DIS50 sequence is as follows: EPGRPDRKGKEKEPEEEPAQGHSLPQEPEPMPRPKPEVRKKPHPGASPHQ.

To construct the PKC sensor, Strep-tag II (stII)–GFP10–FHA2–RFRRFQTLKI [a PKC substrate (34)] was synthesized by GenScript and cloned into the N terminus of RlucII in pcDNA3.1 (pPKC2). DNAs encoding FHA1 and c1b were polymerase chain reaction (PCR)–amplified and subcloned into the pPKC2. A DNA oligo encoding the second PKC substrate (RFRRFQTLKD) was inserted by linker ligation. Last, the DIS50 linker was inserted between FHA2 and the substrate cassette. Two threonine residues found in PKC substrates were mutated to alanine residues by linker ligation. Lyn-PKC was generated by inserting the Bsr GI–Xho I fragment from the PKC sensor (FHA1-FHA2-linker-PKC substrates) into a vector containing the Lyn-GFP10 and RlucII-stII by ligation. This generated a PKC sensor lacking the c1b domain and with a Lyn sequence in its N terminus. To construct the DAG sensor, a DIS300 linker and RlucII-c1b were subcloned into the Lyn-GFP10 (43) in pcDNA3.1 by using in-fusion technology. To generate PKN-RBD-RlucII, the coding sequence of the first 93 residues of the human PKN1 (GenBank accession no. AAH94766.1) was synthesized by GeneArt (Thermo Fisher) as a DNA string and subcloned to pcDNA3.1 hygro (+) GFP10-RlucII, using Nhe I and Age I sites, replacing GFP10 by the RBD of PKN1 (PKN-RBD). A linker sequence (IDTGGRAIDIKLPAT) is present between PKN and RlucII. For the pIRES hygro p63-GFP10-stII construct, GFP10 was PCR-amplified from the EPAC db sensor construct (75) using the following primers: 5′-GCTAGCGGATCCGCCGGTACCATGGTGAGCAAGGGCGAGGAG-3′ and 5′-ATCGGATCCTTATTTTTCGAACTGCGGGTGGCTCCACTTGTACAGCTCGTCCATGCC-3′. PCR products were subcloned in pIRES Hygro3 (Clontech) using Nhe I and Eco RV sites. The construct encoding the Gαq binding domain of the human p63RhoGEF (residues 295 to 502, p63BD) tagged with GFP10 was done by PCR amplification from an IMAGE clone (Open Biosystems) encoding the p63RhoGEF and subcloned by Gibson assembly in pIRES Hygro3 GFP10-stII digested with Nhe I. A peptide linker (PASGSAGT) is present between the p63BD and GFP10. For the GST-tagged p63BD, the Gαq binding domain was PCR-amplified from pIRES Hygro p63BD-GFP10-stII using the following primers: 5′-CTGATCGAAGGTCGTGGGATCCCCGAATTCATGATTATGAAGTACCAGTTGCTC-3′ and 5′-GCGGCCGCTCGAGTCGACCCGGGAATTCAACTTCCAACTCCGGGTCCTCTGGG-3′, and subcloned by Gibson assembly in pGEX 5X2 digested with Eco RI.

Site-directed mutagenesis of AT1R

The AT1R variants (A4.60T, T7.33M, and C7.40W; also referred to as A163T, T282M, and C289W) were generated by PCR with overlapping ends and Gibson assembly. Two complementary oligonucleotides containing the desired mutations were used as 5′ and 3′ primers with its matching Xba I reverse primer (5′-GTGACACTATAGAATAGGGCCCTCTAGA-3′) and Eco RI forward primer (5′-CTACTGAAGATGGCATCAAAAGAATTCAAGATGA-3′) from the sp-Flag-AT1R in pcDNA3.1, respectively. Two PCR amplicons, overlapping at the location of the altered bases, were assembled with Eco RI–Xba I–cleaved sp-Flag-AT1R in pcDNA3.1 in one step using the NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs). The 5′ primers were as follows, with the mutated nucleotides underlined: A163T, GCCAGTTTGCCAACTATAATCCATCGA; T282M, AATTGCAGATATTGTGGACATGGCCATGCC; C289W, GCCATGCCTATCACCATTTGGATAGCT. The I3.27T and A6.39S variants (I103T and A244S) were generated by the whole-plasmid PCR technique. Briefly, primers containing the desired mutations were designed complementing opposite strands of the plasmid and extended using DNA polymerase (Phusion Flash High-Fidelity PCR Master Mix, Thermo Scientific). Dpn I endonuclease (Invitrogen, Thermo Fisher Scientific) was used to digest the DNA template and to select the synthesized DNA containing mutations. The 5′ primers were as follows, with the mutated nucleotides underlined: I103T, GGCAATTACCTATGTAAGACTGCTTCAGCCAGCGT (forward) and ACGCTGGCTGAAGCAGTCTTACATAGGTAATTG (reverse); A244S, TTTAAGATAATTATGTCAATTGTGCTTTTCTTT (forward) and AAGCACAATTGACATAATTATCTTAAAAATATCATCATT (reverse). All constructs were validated by sequencing.

Cell culture and transfection

We used a HEK293 clonal cell line (HEK293SL cells) previously described in (43, 76), hereafter referred to as HEK293 cells. HEK293 and HEK293/AT1R cells (HEK293SL cells, which stably express the sp-Flag-AT1R) (32) were cultured in MEM and DMEM, respectively, supplemented with 10% FBS and gentamicin (20 μg/ml). Cells were grown at 37°C in 5% CO2 and 90% humidity. Cells were seeded at a density of 7.5 × 105 cells per 100-mm dish and were transiently transfected the next day with sp-Flag-AT1R (3 μg) along with either the Gαq-polycistronic BRET sensor (4.5 μg), or with the Gαi2-RlucII (60 ng), or the Gαi3-RlucII (0.24 μg) and GFP10-Gγ2 (0.6 μg) and Gβ1 (0.6 μg) sensor, or with the Gα12(136)-RlucII (0.24 μg) and GFP10-Gγ1 (0.6 μg) and Gβ1 (0.6 μg) sensor, or with the GFP10-βarr2-RlucII (0.075 μg) sensor using the calcium phosphate precipitation method, as previously described (77). For the DAG and PKC sensors, 0.15 and 0.18 μg of sensor DNA were transfected in HEK293/AT1R cells, respectively, using the calcium phosphate precipitation methods. For the Rho sensor, 0.67 μg of sp-Flag AT1R, along with 0.12 μg of PKN-RBD-RlucII and 0.48 μg of rGFP-CAAX, was transfected. After 18 hours of transfection, the medium was replaced and cells were divided for subsequent experiments. For the p63/Gαq sensor, cells were seeded onto polyornithine-coated white 96-well plate at a density of 1 × 104 cells per well. The next day, cells were transfected by Lipofectamine 2000 (Invitrogen) with 34 ng of AT1R, 51 ng of p63BD-GFP10, and 3.4 ng of Gαq(118)-RlucII per well. For receptor titration experiments, HEK293 cells were transfected in suspension using polyethyleneimine [PEI; 1 mg/ml in phosphate-buffered saline (PBS), linear 25 kDa, Polysciences]. Briefly, 20 ng of PKC sensor DNA, along with different amounts of AT1R DNA [adjusted total DNA amount to 1 μg by single-stranded salmon sperm DNA (ssDNA, Sigma-Aldrich)] in 100 μl of PBS, was mixed with 100 μl of PBS containing 3 μl of PEI. After 20 min of incubation, the DNA-PEI complexes (200 μl) were mixed with 3.5 × 105 cells in 1 ml of media and then distributed into 12 wells in a poly-d-lysine–coated white 96-well plate (100 μl per well). For the p63/Gαq sensor, 50 ng of Gαq(118)-RlucII and 1 μg of p63-GFP10, along with different amounts of AT1R (0 to 300 ng) and ssDNA (for adjustment of a total of 1.35 μg of DNA), were used. For testing the specificity of p63BD-GFP10 for different receptors and Gα-RlucII, 1 μg of p63-GFP10, along with either Gαq(118)-RlucII (50 ng), Gα12(136)-RlucII (50 ng), Gαs-RlucII (40 ng), or Gαi2-RlucII (50 ng) with different receptor DNAs [β2AR (250 ng), D4R (100 ng), FP (300 ng), or AT1R (300 ng)] and ssDNA (for adjustment of a total of 1.35 μg of DNA), was transfected in cells. All assays were performed 48 hours after transfection. CRISPR Gq/11 and G12/13 knockout cell lines were a gift from A. Inoue (Tohoku University, Sendai, Miyagi, Japan) and derived as previously described (31, 47). Briefly, either GNAQ and GNA11 or GNA12 and GNA13 genes in HEK293 cells were simultaneously targeted using a CRISPR-Cas9 system to generate the Gq/11 and G12/13 knockout cell lines, respectively. For experiments with AT1R mutants, HEK293 cells were transfected with 3 μg of receptor DNA along with the G protein sensors or 1 μg of receptor DNA along with the PKC-c1b or Rho sensor using the calcium phosphate method. For the βarr2 BRET experiments, HEK293 cells were transfected with 120 ng of βarr2-RlucII along with either 480 ng of rGFP-CAAX (PM translocation) or rGFP-FYVE (endosomal translocation) along with 1 μg of the receptor DNA.

BRET measurements

One day after transfection, HEK293 cells were detached and seeded onto polyornithine-coated white 96-well plates at a density of 2.5 × 104 cells per well in media. The next day, cells were washed once with Tyrode’s buffer [140 mM NaCl, 2.7 mM KCl, 1 mM CaCl2, 12 mM NaHCO3, 5.6 mM d-glucose, 0.5 mM MgCl2, 0.37 mM NaH2PO4, 25 mM Hepes (pH 7.4)] and left in Tyrode’s buffer. For kinetic measurements, BRET signals were monitored every 2 s after addition of the cell-permeable coelenterazine 400a at a final concentration of 5 μM, using a Synergy2 (BioTek) microplate reader. The filter set was 410/80 nm and 515/30 nm for detecting the RlucII (Renilla luciferase) (donor) and GFP10 (acceptor) light emissions, respectively. AngII (at a final concentration of 100 nM) was injected after 11 measurements (20 s), and BRET was recorded over a period of 5 min. For concentration-response curves of the Gαq-poly, Gαi2 and Gαi3, p63/Gαq, and Rho sensors, BRET signals were measured after the cells were stimulated with various concentrations of ligand in Tyrode’s buffer for 2 min at room temperature (21°C). For detecting DAG, PKC-c1b, and Lyn-PKC sensors, cells were stimulated with ligands for 1 min before BRET measurement. For the Gα12 sensor, cells were stimulated for 10 min at 37°C. For detecting βarr2 at the PM or in endosomes, cells were stimulated with either various concentrations of AngII for 4 min at room temperature or for 30 min at 37°C, before BRET measurements. Coelenterazine 400a (final concentrations of 5 μM) was added 3 to 5 min before BRET measurements. For the βarr2 sensor, BRET signals were measured 20 min after addition of various concentrations of ligands, and coelenterazine 400a was added 10 min before BRET measurements. The Gαq/11 inhibitors [UBO-QIC (100 nM) and YM-254890 (200 nM)] and PKC inhibitors [Gö6983 (2 μM), Gö6976 (3 μM), and LY333531 (3 μM)] were added for 30 min before ligand stimulation. PMA and forskolin were used at a concentration of 1 and 10 μM, respectively, for 10 min. A23187 was used at a concentration of 1 μM for 1 min. To inhibit Gαi, cells were incubated overnight with PTX (100 ng/ml) before ligand stimulation. PD 123319 and losartan were added at a concentration of 10 μM for 30 min, before adding 1 μM AngII. For single-concentration stimulations of the p63/Gαq sensor, cells were stimulated for 1 min with either AngII (100 nM), PGF2α (100 nM), dopamine (1 μM), or isoproterenol (1 μM). The BRET ratio was determined by calculating the ratio of the light emitted by GFP10 over the light emitted by the RlucII. All BRET experiments were performed in triplicate.

Radioligand binding experiments

Affinities and receptor abundance were assessed by ligand binding assays using [125I]-AngII in either dose-displacement or saturation experiments. [125I]-AngII was prepared using the Iodogen method, as previously described (18). For binding experiments, HEK293 cells from a 10-cm culture dish transiently transfected with 3 μg of wild-type AT1R or mutant receptors or HEK293/AT1R cells were seeded at a density of ~1.5 × 105 cells per well in polyornithine-coated 24-well plates, 1 day before binding experiments. The following day, cells were washed once with cold PBS and then incubated in the absence or presence of increasing concentrations of AngII with a fixed concentration of [125I]-AngII (0.18 nM; ~150,000 cpm at 2200 Ci/mmol) in binding buffer [25 mM tris-HCl (pH 7.4), 100 mM NaCl, 5 mM MgCl2, with 0.2% (w/v) bovine serum albumin (BSA)] at 4°C overnight. For saturation binding experiments, cells were incubated at room temperature for 1 hour in the same binding buffer containing increasing concentrations of [125I]-AngII (at 180 Ci/mmol). Nonspecific binding was determined in the presence of 1 μM AngII. Cells were then washed three times with ice-cold PBS and solubilized in 0.5 ml of 0.2 M NaOH/0.05% SDS, and bound radioactivity was counted using a PerkinElmer Wizard 1470 automatic γ-counter. Protein amounts were determined using a Bradford assay after harvesting cells using 10 mM EDTA. AT1R, in transient transfection experiments, was expressed at levels between 1 and 3 pmol/mg of total proteins, whereas in HEK293/AT1R cells it was expressed at 0.5 pmol/mg. AngII affinity (Kd) for AT1R was determined to be between 0.1 and 0.5 nM for both conditions. Mutant receptors were expressed between 1 and 5 pmol/mg for A163T and between 0.5 and 1.5 pmol/mg for both I103T and A244S with affinities between 0.5 and 0.9 nM. The C289W mutant affinity was between 2 and 5 nM with an expression level between 0.5 and 1.5 pmol/mg. The expression of the T282M mutant could not be determined from the saturation binding isotherm because of the lack of apparent saturation and was rather extrapolated for ELISA data (see below) by comparing it to that of the binding and ELISA data obtained from the wild type expressed at different levels. It was estimated to be between 1 and 3 pmol/mg. The affinity was estimated to be at least greater than 100 nM.

Intact cell ELISA

The cell surface abundance of AT1R mutants was assessed by cell surface ELISA. Cells were transfected with 3 μg of wild-type or mutant receptors. The next day, cells were plated onto polyornithine-coated transparent 96-well plates at a density of 4.5 × 104 cells per well. After 24 hours, cells were washed once with PBS and fixed with 3% paraformaldehyde for 10 min at room temperature. Cells were washed twice with PBS and incubated for 1 hour with 0.05% BSA in PBS for blocking. Then, cells were incubated with M2 anti-FLAG antibody (1:1000 in PBS/BSA) for 1 hour at room temperature, washed twice with PBS/BSA, and reblocked with PBS/BSA for 10 min. Cells were incubated with HRP-conjugated secondary antibody (donkey anti-mouse antibody, 1:1000 in PBS/BSA) for 1 hour and washed four times with PBS, and then, colorimetric HRP substrate (SIGMAFAST OPD) was added. The reaction was stopped after 10 min by adding 3 M HCl, and the plate was read at an absorbance of 492 nm with a microplate reader (Synergy2, Biotek). To obtain specific signal, nonspecific signal from mock DNA (pcDNA)–transfected cells was subtracted. Protein amounts were determined using a Bradford assay after cell lysis using 0.01% SDS.

Strep-Tactin pull-down experiments

Briefly, HEK293 cells were seeded at a density of 1 × 105 cells per well in polyornithine-coated 6-well plates. The next day, the cells were transfected with the PKC-c1b sensor (200 ng per well) along with AT1R (500 ng per well) using the calcium phosphate precipitation method. Forty-eight hours after transfection, the cells were serum-starved with DMEM containing 20 mM Hepes and stimulated with 1 μM of either AngII or PMA for 1 or 10 min, respectively. Cells were then transferred to ice, washed twice with ice-cold PBS, and solubilized for 1 hour at 4°C in 0.4 ml of solubilization buffer [50 mM Hepes, 100 mM NaCl, 1 mM EDTA, 10% glycerol, 1% Triton X-100 (pH 7.4)] supplemented with protease inhibitors [leupeptin (20 μg/ml), aprotinin (10 μg/ml), pepstatin A (2 μg/ml), and 10 mM phenylmethylsulfonyl fluoride (PMSF)] and phosphatase inhibitors (40 mM sodium pyrophosphate and 10 mM NaF). The samples were cleared by centrifugation and then transferred to fresh tubes with 20 μl of Strep-Tactin sepharose (IBA) and incubated for 1 to 2 hours with mixing at 4°C. The samples were then washed three times with solubilization buffer and then incubated in Laemmli buffer for 10 min at 65°C, followed by SDS–polyacrylamide gel electrophoresis (PAGE), transferred to nitrocellulose membranes, and immunoblotted to detect threonine phosphorylation of the PKC sensor. The total amount of pull-down sensor was determined by blotting with Strep-Tactin–HRP after stripping.

GST pull-down assays

GST-tagged Rhotekin-RBD (GST-Rhotekin-RBD) and p63BD (GST-p63BD) were expressed in Escherichia coli BL21 cells by induction with 0.6 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) at 30°C for 3 hours and 0.12 mM IPTG at 15°C for 16 hours, respectively. The GST fusion proteins were purified using Glutathione Sepharose 4B as previously described (78). For the Rho activation assay using GST-Rhotekin-RBD, the parental HEK293 cells and the CRISPR Gαq/11 knockout (ΔGαq/11) and Gα12/13 knockout (ΔGα12/13) cells were transiently transfected in 100-mm dishes with 3 μg of sp-Flag-AT1R by calcium phosphate methods. The next day, the cells were reseeded onto polyornithine-coated six-well plates (~8 × 105 cells per well). Forty-eight hours after transfection, cells were serum-starved for ~3 hours with DMEM containing 20 mM Hepes and incubated in the absence (DMSO, vehicle) or presence of YM-254890 (200 nM) for 30 min before being either left unstimulated or stimulated with 1 μM AngII for 5 min at 37°C. For the GST-p63BD pull-down experiments, HEK293 cells were transfected in 100-mm dishes with AT1R (3 μg) along with either Gαq(118)-RlucII (500 ng) or Gα12(136)-RlucII (500 ng). The next day, cells were reseeded in a six-well plate. Transfected cells were washed once with ice-cold PBS and lysed for 30 min at 4°C in 300 μl of lysis buffer [50 mM tris-HCl (pH 7.4), 137 mM NaCl, 5 mM MgCl2, 10% glycerol, 1% NP-40, 5 mM dithiothreitol, 20 mM NaF] supplemented with protease inhibitors [1 mM PMSF, leupeptin (10 μg/ml), aprotinin (5 μg/ml), and pepstatin A (1 μg/ml)] and phosphatase inhibitors (20 mM NaF and 0.025 mM sodium pervanadate). Samples were cleared by centrifugation, and 30 μl was kept as total cell lysate. The remaining lysate was transferred to fresh tubes; incubated with 30 μg of either GST, GST-Rhotekin-RBD, or GST-p63BD coupled to glutathione resin; and rotated for at least 1 hour at 4°C. Beads were then washed twice with lysis buffer, and proteins were eluted in 25 μl of sample buffer by heating at 65°C for 10 min. Proteins were resolved on SDS-PAGE, transferred to nitrocellulose membranes, and immunoblotted for detecting RhoA or RlucII.

Lentiviral vector production and transduction of rat VSMCs

The coding regions of PKC-c1b sensor, PKN-RBD-RlucII, and rGFP-CAAX were PCR-amplified and subcloned into Asc I/Bam HI, Spe I/Mfe I, and Spe I/Not I sties of pLVXi2H (29), respectively. Sensor coding viruses were generated as described previously (79). Briefly, Lenti-X 293T (Clontech) cells were seeded at a density of 3 × 106 cells onto polyornithine-coated 100-mm dishes. The next day, cells were transfected with either 6 μg of pLVXi2H-PKC or 4 μg of rGFP-CAAX/2 μg of PKN-RBD-RLucII in pLVXi2H along with 4.5 μg of psPAX2 and 1.5 μg of pMD2.G packaging DNA by a PEI method. The medium was changed 16 hours after transfection. Forty-eight to 60 hours after transfection, the virus-containing medium was harvested by centrifugation at 300g for 10 min and filtered through 0.45-μm pore size filters. Virus was stored either at 4°C up to 2 days or at −80°C. Rat VSMCs (passages 13 to 15) were seeded at a density of 20,000 cells per well in polyornithine-coated white 96-well plates 1 day before viral vector transduction. The cells were transduced with the sensor-encoding virus (60 to 100 μl of viral soup for PKC-c1b sensor or 15 to 25 μl for Rho sensor) in the presence of polybrene (8 μg/ml), and the next day, the medium was changed. Forty-eight hours after viral transduction, the cells were washed once with Tyrode’s buffer, and BRET assays were performed as described above.

Confocal microscopy

One day before transfection, cells were seeded in 35-mm glass-bottom dishes at a density of 1 × 105 cells per dish. HEK293 cells were transfected with 67 ng of AT1R along with either 33 ng of DAG sensor or 33 ng of PKC-c1b or Lyn-PKC sensor per dish using the calcium phosphate method. Forty-eight hours after transfection, cells were imaged before and after stimulation with 1 μM AngII with a Zeiss LSM-510 laser scanning microscope. To detect GFP10, an ultraviolet laser was used with 405-nm excitation and 505- to 550-nm bandpass emission filter. Images (1024 × 1024) were collected using a 63× oil immersion lens.

Data analysis

Data were first normalized as percentage of the Emax of AngII (reference ligand) in each experiment. The normalized data were analyzed using “Operational Model” in GraphPad Prism 6 to determine the transduction ratio, log(τ/KA), of each ligand on each signaling pathway, as previously described (29, 80). A user-defined equation AngII, AngIII, and [Val4]-AngIII were treated as full agonists, and others were treated as partial agonists. The relative effectiveness (RE) of the ligands compared to the reference ligand for a specific pathway is then determined from the difference between their log(τ/KA) values using Eqs. 1 and 2Δlog(τKA)= log(τKA)ligand log(τKA)AngII(1)Relative effectiveness(RE)=10Δlog(τKA)(2)

SEs on the Δlog(τ/KA) are calculated using Eq. 3SE(Δlog(τKA))=(SEligand)2+(SEAngII)2(3)

To quantify biased agonism of the mutant AT1R receptors relative to the wild-type receptor, we used a “relative activity (RA)” scale (6, 81). This scale uses a ratio of the maximal response to the EC50 value for an agonist. Data were normalized to the Emax of AngII in wild-type receptor in a specific signal pathway, and EC50 and Emax were estimated from the nonlinear regression curve fitting equations in GraphPad Prism. Relative activity of the mutant receptors compared to the reference receptor (wild-type receptor) for a specific pathway is then determined from the difference between their log(Emax/EC50) values using Eqs. 4 and 5Relative activity(RA)=EmaxEC50EC50Emax=(EmaxEC50)(EmaxEC50)(4)Δlog(EmaxEC50)= log(EmaxEC50)mutant log(EmaxEC50)WT(5)

Statistical analyses were performed using GraphPad Prism 6 software (GraphPad Software Inc.) using Student’s t tests, two-way analyses of variance (ANOVAs), or Tukey’s post hoc multiple comparisons tests, when appropriate. Normality from a typical BRET experiment (AngII-mediated Gαq activation, n = 13) was analyzed using a D’Agostino-Pearson test, and data were found to follow a normal probability distribution. Because all BRET experiments were performed in a similar manner, we infer normal distribution of data for all the other BRET experiments. Curves presented throughout this study represent the best fits and were generated using GraphPad Prism software. P values ≤0.05 were considered significant.


Fig. S1. Responses of G protein and βarr2 BRET sensors in naïve and AT1R-expressing HEK293 cells.

Fig. S2. Characterization of the DAG BRET sensor.

Fig. S3. Characterization of the PKC BRET sensors.

Fig. S4. Characterization of the p63RhoGEF (p63/Gαq) sensor.

Fig. S5. Characterization of the Rho BRET sensor.

Fig. S6. Heat map signature of AT1R signaling induced by AngII analogs.

Fig. S7. Concentration-response curves for the activation of the p63RhoGEF (p63/Gαq), PKC, and PLC (DAG) BRET sensors by AngII analogs.

Fig. S8. Heat map signaling signatures of Gαq and downstream effectors by AngII analogs.

Fig. S9. Correlation plot analysis of Δlog(τ/KA) for G proteins and β-arrestin sensors against downstream signaling effectors.

Fig. S10. Heat map and correlation plot of Δlog(τ/KA) for Rho and Gα12 activation by AngII analogs.

Fig. S11. Concentration-response curves for G proteins, PKC, Rho, and β-arrestin activation by wild-type and mutant AT1Rs.

Fig. S12. Assessment of cell surface abundance and AngII affinities for wild-type and mutant AT1Rs.

Table S1. Sequences of AngII and AngII analogs.

Table S2. Potency and relative efficacy of AngII and AngII analogs for activating the p63RhoGEF (p63/Gαq), PKC, and PLC (DAG) sensors.

Table S3. Transduction ratio and relative effectiveness of AngII and AngII analogs for activating the p63RhoGEF (p63/Gαq), PKC, and PLC (DAG) sensors.

Table S4. Potency and relative efficacy of AngII and AngII analogs for activating the Rho sensor in the absence or presence of Gαq/11 inhibition.

Table S5. Transduction ratio and relative effectiveness of AngII and AngII analogs for activating the Rho sensor.

Data file S1. Statistical analysis of Δlog(τ/KA) of AngII analogs between signaling pathways.


Acknowledgments: We thank the RI-MUHC for the use of their imaging platform and the members of Laporte laboratory for helpful discussion. Funding: This work was supported by grants from the Canadian Institutes of Health Research (CIHR) (MOP-74603 to S.A.L. and FDN-14843 to M.B.), the Sao Paulo Research Foundation (FAPESP 2012/20148-0 to C.M.C.-N.), and the CQDM (Consortium Québécois sur la Découverte du Médicament; to G.P., T.E.H., R.L., S.A.L., and M.B.). M.B. holds a Canada Research Chair in Signal Transduction and Molecular Pharmacology. Author contributions: Y.N., S.K., Y.C., J.G., L.B.T., and S.C.S. performed experiments and produced data for presentation. Y.N., C.L., V.L., J.-M.L., D.D., T.E.H., G.P., R.L., M.B., and S.A.L. contributed to the development and the validation of sensors. Y.N., C.M.C.-N., M.B., and S.A.L. designed experiments and interpreted the data. Y.N., Y.C., R.L., T.E.H., M.B., and S.A.L. contributed to writing the manuscript. Competing interests: The BRET biosensors presented here have been licensed to Domain Therapeutics for commercialization. M.B. is the president of the Scientific Advisory Board of Domain Therapeutics, a company involved in the discovery of drugs targeting GPCRs. Data and materials availability: The BRET sensors can be obtained upon request to S.A.L. or M.B. and used for academic research with a standard academic material transfer agreement (MTA).

Stay Connected to Science Signaling

Navigate This Article