Heterotrimeric Gq proteins act as a switch for GRK5/6 selectivity underlying β-arrestin transducer bias

Signaling-biased ligands acting on G-protein-coupled receptors (GPCRs) differentially activate heterotrimeric G proteins and β-arrestins. Although a wealth of structural knowledge about signaling bias at the GPCR level exists (preferential engagement of a specific transducer), little is known about the bias at the transducer level (different functions mediated by a single transducer), partly due to a poor understanding of GPCR kinase (GRK)-mediated GPCR phosphorylation. Here, we reveal a unique role of the Gq heterotrimer as a determinant for GRK-subtype selectivity that regulates subsequent β-arrestin conformation and function. Using the angiotensin II (Ang II) type-1 receptor (AT1R), we show that β-arrestin recruitment depends on both GRK2/3 and GRK5/6 upon binding of Ang II, but solely on GRK5/6 upon binding of the β-arrestin-biased ligand TRV027. With pharmacological inhibition or genetic loss of Gq, GRK-subtype selectivity and β-arrestin functionality by Ang II is shifted to those of TRV027. Single-molecule imaging identifies relocation of AT1R and GRK5, but not GRK2, to an immobile phase under the Gq-inactive, AT1R-stimulated conditions. These findings uncover a previously unappreciated Gq-regulated mechanism that encodes GRK-subtype selectivity and imparts distinct phosphorylation-barcodes directing downstream β-arrestin functions. GPCR kinases (GRKs) phosphorylate active-form G-protein-coupled receptors (GPCRs). Here, the authors reveal that Gq heterotrimer coupled with the angiotensin II type-1 receptor (AT1R) determines the GRK subtypes recruited to the complex in a microdomain, thus defining subsequent AT1R phosphorylation patterns, β-arrestin conformation and functionality.

G PCRs are the largest family of membrane proteins and are well-established pharmaceutical targets that play a key role in a broad spectrum of physiological events 1 . Agonist stimulation at GPCRs activates heterotrimeric G proteins and their effector proteins, leading to a series of cellular responses. Agonist-bound GPCRs undergo phosphorylation in their cytosolic loops and/or the C-terminal tail by GPCR kinases (GRKs) and subsequently recruit β-arrestin 2 . β-arrestin terminates Gprotein-mediated signaling by spatially competing with G protein and facilitating receptor internalization 3,4 . Besides, β-arrestin induces its own signaling by scaffolding effector proteins 5 .
Biased ligands have attracted much attention owing to its potential to discriminate a therapeutic effect from other on-target adverse effects 6 . Biased agonism, also known as functional selectivity, is divided into "GPCR bias" and "transducer bias" 7 . GPCR bias is defined as different activation states of GPCRs that allow for binding of a specific transducer (e.g., G protein and β-arrestin). Transducer bias arises from variable conformations of a single transducer (typically, β-arrestin) that results in distinct downstream signaling events. Unlike the well-characterized GPCR bias 8,9 , β-arrestin transducer bias remains poorly explored and is proposed to occur through at least two mechanisms: 1) phosphorylation patterns of receptors, 2) β-arrestin-binding modes to receptors.
The former mechanism of the transducer bias is termed a phosphorylation barcode and is produced by a specific pattern of GRK-mediated receptor phosphorylation 10 . In humans, GRKs consist of seven subtypes (GRK1-7) with four subtypes (GRK2, GRK3, GRK5 and GRK6, classified into the GRK2/3 subfamily and the GRK5/6 subfamily) being widely expressed in the body 11 . Biased ligands induce the recruitment of distinct GRK subtypes to the receptor as compared with balanced ligands. It has been reported that individual GRKs recognize different GPCR conformations 12,13 . Further, previous studies identified several mechanisms following G-protein activation that regulate GRKmediated GPCR phosphorylation. For example, in GRK2/3, binding to a free, membrane-anchored Gβγ subunit plays a critical role in receptor phosphorylation by recruiting GRK2/3 to the plasma membrane from the cytosol and enhancing kinase activity 14 . In GRK5/6, calmodulin-binding negatively regulates receptor phosphorylation by dispersing their localization from the plasma membrane 15 . In addition, certain phospholipids (e.g., phosphoinositides) enhance GRK5/6 kinase activity 16 . GRKs, in turn, can regulate a function of their effector proteins. In GRK2/3, binding to an GTP-bound, activated Gαq subunit inhibits GTPase activity 17,18 , while kinase activity of GRK2/3 is not altered 18 . From these findings, it is assumed that G-protein-less-active GPCR agonists rely on GRK5/6, but not GRK2/3, for their βarrestin recruitment, yet without such evidence or existence of a possible positive regulator for GRK5/6. The latter mechanism of the transducer bias is visualized in recent structural studies 19,20 that provide two unique engagement modes of GPCR-β-arrestin complexes: tail hanging and core engagement. The tail-hanging form was initially considered as an intermediate state to the core-engagement form, but recently it has been shown that both β-arrestin forms are capable of inducing GPCR endocytosis and Src activation 21 . Despite the differential engagements of β-arrestin to the receptor, their functional distinction is poorly characterized.
Multicolor single-molecule-tracking (SMT) analysis is a powerful method for analyzing dynamic interaction between GPCRs and their effector proteins. A previous SMT study observed accumulation of ligand-bound GPCR molecules in a compartmentalized region in the plasma membrane called "hotspot", where the GPCR molecules encounter a heterotrimeric G protein 22 . While GPCR-triggered G-protein signal transduction occurs in the hotspot, it is unclear where, i.e. within or outside of the hotspot, subsequent signaling events such as GRK-mediated GPCR phosphorylation and β-arrestin recruitment take place.
AT1R is a widely studied model for biased agonism. TRV027 was developed as a β-arrestin-biased AT1R ligand that induces preferential engagement of Gq over β-arrestin and is tested in a clinical trial for the treatment of heart failure and, more recently, COVID-19-associated acute respiratory distress syndrome [23][24][25] . A series of in vitro structural studies and a cell-based assay have demonstrated a conformational landscape of AT1R that is dependent on its bound ligands [26][27][28][29] , as well as the conformation of β-arrestin as a transducer 30 . However, these analyses have again focused on GPCR bias and the relative impact of transducer bias at β-arrestin remains unclear.
Here, we use a battery of complementary approaches including GRK-deficient HEK293 cell lines, split-luciferase-based proximity assays and high-resolution single-molecule-tracking analysis to demonstrate that, with reference to Ang II, TRV027 induces βarrestin transducer bias by selectively enhancing GRK5/6dependent phosphorylation in AT1R. We surprisingly discover that active GRK5/6 availability for ligand-bound AT1R is negatively regulated by the activity of a G protein, i.e., Gq, which has typically been considered a non-factor in the context of β-arrestin bias at GPCRs.

Results
Selectivity of GRK5/6 at AT1R inversely correlates with Gq activity. Previous studies indicate that phosphorylation of AT1R is mediated by GRKs 31 and members of the protein kinase C (PKC) family 32 . To address whether these kinases are exclusively responsible for β-arrestin recruitment to ligand-bound AT1R, we genetically deleted GRKs and treated the cells with a pan-PKC inhibitor Go6983. Given that the ubiquitously expressed four GRK members (GRK2, 3,5,6) are grouped into GRK2/3 and GRK5/6 33 , we generated HEK293 cell lines lacking GRK2/3, GRK5/6 or combination of the two (referred to as ΔGRK2/3, ΔGRK5/6 or ΔGRK2/3/5/6 cell lines, respectively) by using a CRISPR/Cas9 technology (Fig. 1a). Western blot analysis confirmed that the targeted GRKs were successfully eliminated (Fig. 1b, c). Expressions of untargeted GRKs were increased by two-to-three fold and the altered GRK levels correlated with the variability of the βarrestin-recruitment response (below) between the clones for the same genotypes ( Supplementary Fig. 1, 2a-d). Furthermore, expression and functional analyses of GPCR signal components identified variability among the genotypes and the clones, thus suggesting a need for a cautious interpretation when comparing GPCR signaling responses among the GRK-deficient cell lines ( Supplementary Fig. 2e-l). Nevertheless, we viewed that the GRKdeficient cells would be valuable tools to dissect bias at the level of GRK subtypes and that such clone-associated issues would be minimized when GRK subtypes are re-expressed in ΔGRK2/3/5/6 (below). Using the NanoBiT-based protein-protein interaction assay 34 , we measured the association of β-arrestin1 and β-arrestin2 with AT1R in the presence of Go6983 (hereafter included for pretreatment unless otherwise noted). Ang II-induced β-arrestin1 recruitment to AT1R was abolished in ΔGRK2/3/5/6 cells while the responses were partially reduced in ΔGRK2/3 cells and rather increased in ΔGRK5/6 cells (Fig. 1d, e), the latter of which may be in part attributable to the upregulated GRK2/3 expression. Notably, in the absence of Go6983, we observed essentially the same relative contributions of the GRK subtypes as the Go6983-treated condition ( Supplementary Fig. 3a, b), excluding a direct effect of PKC inhibition on GRK regulation [35][36][37] . β-arrestin2 showed a similar GRK dependency to that of β-arrestin1 although there was a residual recruitment The NanoBiT pair consisting of SmBiT-fused AT1R and LgBiT-fused β-arrestin was co-expressed in the parent or the indicated GRK-deficient cells and AT1R ligand-induced β-arrestin recruitment was monitored. b, c, Validation of successful elimination of the targeted GRK subtypes. Western blot analyses were performed by using lysates derived from wild-type (parent HEK293) and ΔGRK2/3, ΔGRK5/6 (b) and ΔGRK2/3/5/6 (c). Membranes were probed with the indicated antibodies against GRK2, 3, 5, 6 or α-tubulin. α-tubulin was used as a loading control. d-g, Luminescent kinetics (d, f) and quantifications (e, g) of the NanoBiT-β-arrestin-recruitment responses. The indicated cells were stimulated with 1 µM Ang II (d, e) or 1 µM TRV027 (f, g). h, Heatmap representation of the β-arrestin recruitment by Ang II and its analogs. Data are presented as mean from 3-6 independent experiments and β-arrestinrecruitment signal in the parent cells was set as 100% for each ligand. i, G-protein-coupling signatures (derived and re-edited from Namkung et al. 25 ). j, Comparison of the β-arrestin recruitment responses in ΔGRK5/6 cells and G-protein-coupling activity. The scattered plots represent data from the five ligands except for Ang II (see also Supplementary Fig. 3i, j for plots with all of the cell lines and the six ligands). In Fig. 1b, c, representative images of 3-4 independent experiments are shown. In Fig. 1d, f, lines and shaded regions represent mean and SEM, respectively, of 3-6 independent experiments with each performed in duplicate. In Fig. 1e, g, bars and error bars represent mean and SEM, respectively, of 3-6 independent experiments with each performed in duplicate. In Fig. 1j, symbols and error bars represent mean and SEM, respectively, of 3-6 independent experiments with each performed in duplicate. Shaded regions denote 90% confidence of the linear regression analysis. R 2 values are shown at the right. Note that for many data points, vertical error bars are smaller than the size of the symbols, and thus are not visible. In Fig. 1e, g, * and ** represent P < 0.05 and 0.01, respectively, with one-way ANOVA followed by the Dunnett's test for multiple comparison analysis with reference to the Ang II stimulation. ns, not significantly different between the groups. See Supplementary statistics data file for additional statistics and exact P values. response even in ΔGRK2/3/5/6 cells ( Supplementary Fig. 3c, d).
These results indicate that Ang II-stimulated β-arrestin1 and β-arrestin2 recruitment is primarily mediated by GRK2/3 (as in ΔGRK5/6 cells) with a minor contribution from GRK5/6 (as in ΔGRK2/3 cells) and that there are no other kinases involved in β-arrestin1 recruitment except for PKCs ( Supplementary Fig. 3e, f).
Inhibition of Gq activation alters GRK-subtype selectivity and β-arrestin function. Recent studies have shown that Ang II and TRV027 induce unique structural changes in AT1R, suggesting that GPCR bias is a direct consequence of these structural states 29 . Given our finding that the bias exists at the level of GRK-subtype selectivity (Fig. 1h), we questioned whether the lack of Gq activation or the ligand-bound AT1R state is a determinant for the GRK-subtype selectivity. To test this, we pretreated the GRK-deficient cell lines with the chemical Gq inhibitor YM-254890 (YM)) 39 and assessed Ang IIinduced β-arrestin recruitment. YM pretreatment reduced β-arrestinrecruitment response in ΔGRK5/6 cells, but enhanced the response in ΔGRK2/3 cells (Fig. 2a, b and Supplementary Fig. 4a, b). To evaluate the contributions of individual GRKs and to exclude a clonal effect of the knockout cell lines, we restored each GRK subtype in ΔGRK2/3/ 5/6 cells at comparable levels across the GRK subtypes. By quantitative real-time PCR analyses, we observed that mRNA expression levels of the GRK subtypes was highest in GRK2, followed by GRK6, GRK3 and GRK5 ( Supplementary Fig. 2g). Using C-terminally FLAG epitope-tagged GRK recombinant proteins and antibodies for FLAG tag and the individual GRK subtypes, we found that GRK2 was expressed about one order higher than the other GRK subtypes ( Supplementary Fig. 4c, d). In the following add-back experiments, we individually restored the GRK subtypes in ΔGRK2/3/5/6 cells at an expression level comparable to the endogenous expression of GRK2. Restoration of GRK5 or GRK6, but not GRK2 or GRK3, gained β-arrestin-recruitment signal upon YM pretreatment (Fig. 2c, d and Supplementary Fig. 4e, f). Reduction of GRK2/3-mediated βarrestin-recruitment responses in the YM pretreatment condition was likely attributable to reduced level of free Gβγ subunits (Supplementary Fig. 4g,h), which triggers GRK2/3-mediated GPCR phosphorylation 27 . In addition, kinase-inactive mutants failed to restore β-arrestin recruitment response ( Supplementary Fig. 5a-d), indicating that the observed β-arrestin recruitment thus far is dependent on a kinase activity of GRKs. Together, these data demonstrate that inhibiting Gq activation enhances GRK5/6-dependent β-arrestin recruitment to AT1R.
To identify AT1R residue(s) responsible for the GRK5/6dependent β-arrestin recruitment, we individually mutated the seven Ser and Thr residues to Ala within the two potential phosphorylation codes ( Supplementary Fig. 5e) 28 . These single-Ala mutants were similarly expressed to the level of WT AT1R ( Supplementary Fig. 5f), but showed variable levels of ligand- For each experiment, Ang II-induced β-arrestin1-recruitment signal in the parent cells was set as 100%. c, d, Assessment of the individual GRK subtypes for their sensitivity to YM-254890. The C-terminally FLAG epitopetagged GRK2, 3, 5 and 6 constructs were individually expressed at a level equivalent to that of endogenous GRK2 ( Supplementary Fig. 3a, b) along with AT1R-Sm and Lg-β-arrestin1 in ΔGRK2/3/5/6 cells and stimulated with 1 µM Ang II. For each experiment, Ang II-induced β-arrestin1-recruitment signal was set as 100% (d). Cell lines used were ΔGRK2/3, CL2; ΔGRK5/6, CL2; ΔGRK2/3/5/6, CL2 (a-d). In Fig. 2a, c, lines and shaded regions represent mean and SEM, respectively, of 3 independent experiments with each performed in duplicate. In Fig. 2b, d, bars and error bars represent mean and SEM, respectively, of 3 independent experiments with each performed in duplicate. In Fig. 2b, d, * and ** represent P < 0.05 and 0.01, respectively, with two-way ANOVA followed by the Dunnett's test for multiple comparison analysis with reference to the Ang II stimulation. ns, not significantly different between the groups. See Supplementary statistics data file for additional statistics and exact P values.
induced β-arrestin1 recruitment. By normalizing TRV027induced β-arrestin1 recruitment to that of Ang II, we found thatTRV027-induced response was selectively less efficacious in S326A ( Supplementary Fig. 5g, h). These single Ala-mutant results indicate that Ser326 is a key residue for TRV027-induced GRK5/6 phosphorylation, yet multiple or stepwise phosphorylation may be potentially important.
The importance of Gq in switching GRK5/6 utilization led us to investigate whether β-arrestin transducer bias in reference to Ang II occurred upon TRV027 stimulation or YM pretreatment followed by Ang II stimulation (TRV/Ang II + YM treatments). In the following experiments, we examined the functions and conformations of β-arrestin.
Second, we evaluated phosphorylation of extracellular signalregulated kinase (pERK), a prototypical signaling event downstream of β-arrestin activation, by an AlphaLISA assay. To eliminate pERK response by the Gq-PKC pathway 40 , the AT1Rexpressing cells were treated with Go6983. In this condition, we confirmed that the pERK response was dependent on β-arrestin as the pERK signal was completely diminished in the β-arrestindeficient (ΔARRB) cells ( Supplementary Fig. 7a). Ang II stimulation induced a robust pERK signal, pERK response was marginal by TRV027 stimulation 41,42 and nearly undetectable by YM pretreatment (Fig. 3d and Supplementary Fig. 7b-d). We note that the low level of pERK induced by TRV027 was largely diminished upon Go6983 treatment, indicative of a residual Gq-PKC-mediated signaling component ( Supplementary Fig. 7b). GRK-subtype-dependent β-arrestin functionalities. a, Schematic representation of the HiBiT-based AT1R-internalization assay. N-terminally HiBiTfused AT1R was expressed in the parent or the GRK-deficient cell lines and a cell-surface AT1R level upon ligand stimulation was measured by complementation with the recombinant LgBiT protein added to the conditioned media. b, Luminescent kinetics of the AT1R-internalization assay. The indicated cells were stimulated with 10 nM Ang II, 100 nM TRV027 or 10 nM Ang II pretreated with 1 µM YM-254890. At every measurement time point, a ligand-stimulated count was normalized to a vehicle count. c, Quantifications of the AT1R-internalization assay. Bars and error bars represent mean and SEM, respectively, of 5 independent experiments with each performed in duplicate. d, AlphaLISA-based pERK measurement. AT1R-expressing cells were pretreated with or without the PKC inhibitor, Go6983 (3 µM), and stimulated with the indicated ligand for 30 min. e-g, Schematic illustration (e), luminescent kinetics (f) and quantification (g) of the tail-hanging β-arrestin-recruitment assay. AT1R-Sm and Lg-β-arrestin1 with the truncation in the finger-loop region (ΔFLR) were expressed in the parent cells and treated with the indicated conditions. Shaded regions denote SEM (n = 3-5; b, f). Cell lines used were ΔGRK2/3, CL2; ΔGRK5/6, CL2; ΔGRK2/3/5/6, CL2 (a-c). In Fig. 3b, f, lines and shaded regions represent mean and SEM, respectively, of 3 independent experiments with each performed in duplicate. In Fig. 3c, d, g, bars and error bars represent mean and SEM, respectively, of 3 independent experiments with each performed in duplicate. In Fig. 3c, d, g, * and ** represent P < 0.05 and 0.01, respectively, with two-tailed multiple t-test (d) or oneway ANOVA (g) or two-way (c) ANOVA followed by the Dunnett's test for multiple comparison analysis (with reference to the vehicle stimulation (c) or the Ang II stimulation (g)). ns, not significantly different between the groups. See Supplementary statistics data file for additional statistics and exact P values.
These results indicate that internalization and signaling activities of β-arrestin are separable.
Third, we measured a conformational change of β-arrestin using an Ib30 sensor ( Supplementary Fig. 7e), a single-chain antibody that recognizes a specific active form of β-arrestin1 43,44 . The TRV/Ang II + YM treatments apparently decreased Ib30 reactivity to β-arrestin1 as compared with Ang II stimulation ( Supplementary  Fig. 7f, g). Normalization of the Ib30 and β-arrestin-recruitment responses showed that Ib30 recruitment is proportional to the β-arrestin1 recruitment levels ( Supplementary Fig. 7h). Thus we conclude that the Ib30-recognizable active β-arrestin conformation is similarly induced in the TRV/Ang II + YM treatments.
Last, we assessed β-arrestin-binding forms. β-arrestin has at least two distinct modes of engagement with GPCRs: the tailhanging and the core-engagement states 19 . While both are capable of receptor internalization, functional differences, if any, remain obscure 21,45 To distinguish the two modes, we used β-arrestin constructs lacking the finger-loop region (FLR) (Fig. 3e), a critical motif for core engagement 21 . Both ΔFLR β-arrestin1 and 2 were expressed at the level of the WT counterparts ( Supplementary Fig. 8a, b) and showed a reduced recruitment response upon Ang II stimulation (  Fig. 8k-m). These data support a model in which Ang II-induced β-arrestins are in both engagement modes, whereas β-arrestins upon the TRV/Ang II + YM treatments mostly exist in the tail-hanging mode.
Taken together, these results indicate that the GRK-subtype selectivity arises not directly from structural states of the ligandbound receptor, but from the activation level of Gq and, further, that this mechanism governs β-arrestin functionality.

AT1R and GRK5 interaction occurs in a membrane domain.
Using SMT analysis, we evaluated the changes in dynamics of AT1R and GRK molecules. We monitored SF650-labeled AT1R and TMR-labeled GRK2 or GRK5 in ΔGRK2/3/5/6 cells for their diffusion, oligomerization and colocalization ( Fig. 4a and Supplementary Movies 1, 2). In accordance with previous studies 22, 46 , the variational Bayesian-hidden Markov model (VB-HMM) analysis classified diffusion of the molecules into four states (immobile, slow, medium, and fast; Fig. 4b and Supplementary  Fig. 10, 11). Mean square displacement (MSD)-Δt plot analysis showed that AT1R and GRKs in the immobile and the slow states were restricted to membrane domains with a confinement length of 90-250 nm ( Supplementary Fig. 11e, f).
We estimated the distribution of putative oligomer size of the molecules from the intensity histogram of the Gaussian functions ( Supplementary Fig. 12). Assuming that both AT1R and the GRKs are monomeric in the fast states, we inferred that these molecules are dimers or oligomers in the immobile state. Similarly, AT1R exists as a mixture of dimers and monomers in the slow and the medium states, and GRKs a mixture of dimers and tetramers in these states. Ang II stimulation statically significantly decreased the putative oligomer size in GRK5 in the immobile state ( Supplementary Fig. 12c). The TRV/Ang II + YM treatments did not alter it and an increase of the putative oligomer size in the medium and fast states of GRK5. In contrast, no significant change was observed in GKR2 ( Supplementary  Fig. 12c). Next, we assessed the frequency and duration of colocalization of AT1R with GRK molecules (Fig. 4c-e and Supplementary Fig. 13). Compared with vehicle treatment, Ang II stimulation significantly decreased the association rate between AT1R and GRK5 (Fig. 4d). In the meantime, the lifetime of AT1R-GRK5 complexes was slightly increased (Fig. 4e). The YM pretreatment reverted the Ang II-induced decrease of association rate to the similar extent of TRV027 stimulation. The TRV027 stimulation statically significantly increased the lifetime of AT1R-GRK5 complexes ( Fig. 4e and Supplementary Fig. 13e,f.). These results suggested an enhanced availability of ligand-bound AT1R to active GRK5 in the TRV/Ang II + YM treatments.
Finally, we compared the diffusion behaviors of the molecules and the complex (Fig. 4f, g and Supplementary Fig. 11, 13i). Upon Ang II stimulation, AT1R molecules in the GRK2-or the GRK5expressing cells increased the proportion of the slower fractions (immobile, slow). On the other hand, the TRV/Ang II + YM treatments increased the level of the slower fractions in the GRK5-expressing, but not the GRK2-expressing cells (Fig. 4f). Ang II, but not TRV/Ang II + YM, stimulation increased GRK2 molecules in the slower fraction whereas the TRV/Ang II + YM treatments, but not Ang II stimulation, elevated GRK5 molecules in the slower fraction ( Fig. 4g), indicating that AT1R phosphorylation occurs in the slower fraction for both GRKs. Notably, Ang II stimulation decreased the immobile fraction of GRK5 molecules and concurrently accumulated AT1R molecules in the immobile phase. The altered distribution of GRK5 molecules by the Ang II stimulation (i.e., Ang II-induced decrease in the slower fraction) was also observed in the parent cells but not in the Gq family-deficient (ΔGq) cells ( Supplementary Fig. 14)., suggesting that the lack of active Gq, rather than the existence of inactive, is a driver for the altered behaviors in GRK5. Collectively, these data support a model in which activation of Gq promotes dissociation of GRK5 from the AT1R at the immobile domain (Fig. 4h). In the meantime, Gq activation triggers GRK2 recruitment to the immobile domain ( Fig. 4i) presumably via the resulting free Gβγ, and frees up the receptor for phosphorylation by the activated GRK2/3.
Assembly and dissociation of the AT1R-Gq-GRK complex. To gain insight into how Gq activation regulates GRK5/6 selectivity, we assessed the recruitment of GRKs to AT1R (Fig. 5a). β2AR, which was previously shown to be phosphorylated by the four GRK subtypes 11 , exhibited an increase in luminescence, indicative of GRK recruitment to the receptor, for all of the four GRKs upon isoproterenol stimulation (Fig. 5b). In contrast, upon Ang II stimulation, AT1R showed recruitment responses for GRK2 and GRK3, but dissociation responses for GRK5 and GRK6 (Fig. 5b, c and Supplementary Fig. 15a, b). Remarkably, the TRV/Ang II + YM treatments induced GRK5/6 recruitment to AT1R (Fig. 5d, e and Supplementary Fig. 15c, d). GRK2/3 showed recruitment response, but at a lower level as compared with Ang II stimulation. These results are consistent with the selectivity of GRK-mediated β-arrestin-recruitment responses by the NanoBiT-based sensors and the single-molecule imaging (Figs. 1d-g, 2a, b, 4).
Next, we assessed whether the presence of inactive Gq or lack of active Gq, or more broadly, the contribution of the other G proteins, is important for the GRK5/6 behavior. To do so, we conducted the same experiments as above using a panel of the Gprotein-deficient cell lines. The GRK5/6 recruitment pattern was reversed in ΔGq cells, but not in the other G-protein-deficient cell lines (Fig. 5f, g and Supplementary Fig. 15e, f). Inhibitors for the Gq signaling pathway (PLCβ, Ca 2+ and PKC) showed a variable degree of suppression levels, but none of them mimicked the YMlike GRK5/6-association pattern ( Supplementary Fig. 16a, b) further validate a specific effect of Gq, we generated cells lacking the Gq family members in the genetic background of ΔGRK2/3/5/ 6 cells (GRK/Gq-6KO; Supplementary Fig. 16c, d) and assessed GRK5-suppressive activity of the individual Gq members by reexpressing GRK5 along with the NanoBiT-β-arrestin sensor. We found that Ang II-induced β-arrestin-recruitment response in the GRK/Gq-6KO cells was increased to the level of YM-treated ΔGRK2/3/5/6 cells ( Supplementary Fig. 16e, f; note that GRK5 was exogenously expressed in both cell lines), validating the ontarget effects of YM. Among the four Gq family members, the Gαq, the Gα11 and the Gα14 subunits, but not the Gα16 subunit, exhibited an inhibitory effect on GRK5-mediated β-arrestinrecruitment (note that the experiment was performed in the absence of YM) while all of the members showed equivalent Gqsignaling activity ( Supplementary Fig. 16g). Furthermore, the expression of a constitutively active Gαq in the GRK/Gq-6KO cells attenuated the GRK5-mediated β-arrestin recruitment response whereas such effect was not observed for a dominantnegative Gαq ( Supplementary Fig. 16h, i). These data indicate that active Gq, but not downstream Gq signaling, induces dissociation of GRK5/6 from AT1R. To better understand how Gq activation regulates the GRK5/6 behavior, we evaluated whether GRK5/6 would exist proximal to Gq and this colocalization would be affected by Gq activation states (Fig. 6a, d). We detected significant luminescent signals in cells expressing the Gαq-Lg and the GRK5-Sm or the GRK6-Sm sensor ( Supplementary Fig. 17a). Upon Ang II stimulation, both of the Gαq-GRK sensor and the Gβγ-GRK sensor showed decreased luminescence (Fig. 6b, c, e, f and Supplementary Fig. 17b-e). These luminescence changes were suppressed or reversed by the TRV/Ang II + YM treatments, indicating that GRK5/6 dissociates from the Gq heterotrimer in a Gq activation-dependent manner.
Finally, we tested whether the three components (AT1R, Gq and GRK5/6) were in close proximity by using a combined Fig. 4 Single-molecule imaging reveals accumulation of AT1R and GRK5 in the immobile phase under the Gq-inactivated condition. a, Representative total internal reflection fluorescence (TIRF) microscopic image of ΔGRK2/3/5/6 (CL2) cells expressing SF650-labeled AT1R (magenta) and TMR-labeled GRK5 (green). Scale bar: 5 µm. b Representative trajectories of 4 diffusion states of AT1R (left), GRK5 (middle), AT1R-GRK5 colocalization (right). c Representative images of AT1R-GRK5 colocalization. Scale bar: 1 µm. d, e, AT1R-GRK5 association (on-event) and dissociation (colocalization duration) rates estimated from all trajectories. f, g, Fractions of the diffusion states in AT1R (f) and GRK5 (g) molecules by the indicated condition (vehicle, 1 µM Ang II, 1 µM TRV027 or 1 µM Ang II pretreated with 1 µM YM-254890). h, i, Schematic summary of single molecule imaging of AT1R-GRK5 (h) and AT1R-GRK2 (i). In Fig. 4a-c, representative images of 20-29 cells were shown. In Fig. 4d, shaded regions represent a histogram of 20-29 cells. In Fig. 4e, Symbols represent mean of all trajectories in 20-29 cells. In Fig. 4f, g, bars and error bars represent mean and SEM, respectively, of 20-29 cells. In Fig. 4d, e, f, g, * and ** represent P < 0.05 and 0.01, respectively, with one-way ANOVA with followed by the Tukey HSD test among 4 groups (n = 20-29 cells). See Supplementary statistics data file for additional statistics and exact P values. , and quantifications (c) of the NanoBiT-GRK-recruitment assay. AT1R-Sm or β2AR-Sm was expressed together with the indicated GRK-Lg constructs in the parent cells and stimulated with Ang II (1 µM) or isoproterenol (10 µM). Note that owing to the time lag between manual ligand addition and start of kinetics measurement, luminescent signals were changed from the basal count (normalized to the value of 1) at the first reading (set as time = 0 min). d-g, NanoBiT-GRK recruitment in the TRV/Ang II + YM treatments (d, e) and the G-protein-deficient cells (f, g). HEK293A cells deficient for the individual G-protein families (ΔGs, ΔGi, ΔGq, ΔG12; each deficient for the Gα subunits expressed in HEK293A cells; see Methods for detail) were used to monitor Ang II (1 µM)induced AT1R-GRK association (f, g). Note that vehicle plots from all of the cell lines overlapped tightly and only those from the parental cells were shown for clarity (f). In Fig. 5b, d, f, lines and shaded regions represent mean and SEM, respectively, of 3 (f) or 3-4 (b, d) independent experiments with each performed in duplicate. In Fig. 5c, e, g, bars and error bars represent mean and SEM, respectively, of 3 (g) or 3-4 (c, e) independent experiments with each performed in duplicate. In Fig. 5e, g, * and ** represent P < 0.05 and 0.01, respectively, with two-way ANOVA followed by the Dunnett's test for multiple comparison analysis (with reference to the Ang II stimulation (e) or the Parent (g)). See Supplementary statistics data file for additional statistics and exact P values.
NanoBiT and bioluminescence resonance energy transfer (BRET) strategy 47,48 . We used GRK6 because of its higher basal complementation signal than GRK5 (Supplementary Fig. 17a). Prior to ligand stimulation, assembly of the three components was detected as a BRET signal whereas this signal was undetectable in the control conditions ( Fig. 6g and Supplementary Fig. 17f-j). Ang II stimulation attenuated both NanoBiTderived intensity and mVenus-mediated BRET intensity, whereas the BRET ratio was slightly increased (Supplementary Movies 3). The TRV/Ang II + YM treatments showed modest effects in both NanoBiT signal and BRET signal, thus leaving the BRET ratio unchanged (Fig. 6h, i). These results indicate that the three components are in close proximity to each other in the basal state, likely in the immobile domain as well (Fig. 7), and that, upon Ang II stimulation, Gq and GRK6 individually and simultaneously, rather than as the Gq-GRK6 complex, dissociates from AT1R.
Finally, we examined whether the Gq-regulated GRK-subtype selectivity is generalizable to other Gq-coupled receptors. We found that both of the tested Gq-coupled receptors (the serotonin 5-HT 2A receptor and the neurotensin NTS 1 receptor) showed a gain of GRK5/6-mediated β-arrestin recruitment upon YM treatment, whereas the effect was completely absent in the tested non-Gq-coupled receptors (the Gs-coupled β2 adrenergic receptor and the Gi-coupled µOR opioid receptor) (Supplementary Fig. 18).

Discussion
The bidirectional regulatory relationship between G proteins and β-arrestin has received considerable attention in prior studies. The classical role of β-arrestin is to suppress G-protein coupling by competitively engaging in the receptor core 49 . As a signal enhancer, β-arrestin induces sustained G-protein signaling by forming "megaplex," a ternary GPCR-G-protein-β-arrestin complex 50 . Oppositely, G-protein activation triggers the initiation of the β-arrestin system by releasing free Gβγ subunits and activating GRK2/3 14 . A GTP-bound, active Gαq subunit is shown to bind to GRK2/3 17 , yet its functional effect has been only observed in the Gαq side (enhancement of GTPase activity) 18 . To the best of our knowledge, no specific effect on GRKs is reported for the Gα subunits nor is negative feedback regulation of βarrestin by G proteins. In this study, we demonstrate that the Gq heterotrimer acts as a negative modulator of β-arrestin by suppressing the function of GRK5/6, in turn enhancing the access of GRK2/3 to AT1R, which collectively affects phosphorylation code-sensitive β-arrestin function. . In all panels, * and ** represent P < 0.05 and 0.01, respectively, with two-way ANOVA and following the Dunnett's test for multiple comparison analysis (n = 3-5 in dots; with reference to Ang II or Parent). g-i, Schematic representation (g), microscopic imaging (h), and quantifications (i) of the NanoBiT-BRET assay to assess proximity of the three AT1R-Gq-GRK components. Gαq-Lg, Sm-GRK6 and AT1R-mVenus were expressed in the parent cells. The complemented NanoBiT luciferase signal and the BRET signal via mVenus were imaged before and after the indicated stimulation conditions (h). Fold-changes of NanoBiT, mVenus and BRET ratio index were indicated in (i). * and ** represent P < 0.05 and 0.01, respectively, with the Tukey's test among 4 groups (n = 17-21 cells). In Fig. 6b, e, lines and shaded regions represent mean and SEM, respectively, of 3-5 independent experiments with each performed in duplicate. In Fig. 6c, f, Bars and error bars represent mean and SEM, respectively, of 3-5 independent experiments with each performed in duplicate. In Fig. 6h, representative images of 17-21 cells were shown. In Fig. 6i, shaded regions represent a histogram of 20-29 cells. In Fig. 6c, f, i, * and ** represent P < 0.05 and 0.01, respectively, with two-way ANOVA followed by the Dunnett's test (c, f) or the Tukey HSD test (i) for multiple comparison analysis (c, f; with reference to the Ang II stimulation). See Supplementary statistics data file for additional statistics and exact P values. Our results reveal a novel mechanism underlying GRK-subtype selectivity and subsequent transducer bias at β-arrestin (Fig. 7). Previous dual-color SMT analysis for GPCRs and G proteins 22,46 along with single-particle tracking by photo-activation localization microscopy (sptPALM) analysis of β-arrestin 51 demonstrate that, upon ligand stimulation, active GPCRs and G proteins accumulate in immobile domains or "hot spots." Here, we report, for the first time, dual-color SMT analysis for a GPCR and GRKs and observe that the AT1R and GRK2/GRK5 molecules are concentrated in similar domains. We propose that this phenomenon is the molecular basis for efficient receptor phosphorylation. Furthermore, we speculate that the high density of AT1R, Gq and GRK5/6 in the immobile domain allows a complex formation of these molecules and, once Gq is activated, it triggers separation of the complex and dispersal of Gq and GRK5/6 from the domain. Additionally, active Gq facilitates GRK5/6 translocation from the plasma membrane to the cytosol by inducing Ca 2+ -mediated calmodulin activation 15 and in turn recruits and activates GRK2/3 by releasing free Gβγ 14 .
Overall, Gq activity acts as a molecular switch in GRK-subtype selectivity.
The functional diversity of individual GRKs has been wellstudied for the β2AR. In cultured cells, isoproterenol, a reference ligand for β2AR, induces phosphorylation of β2AR by all of the major GRK subtypes, i.e., GRK2, GRK3, GRK5, and GRK6 11 . On the other hand, carvedilol, a unique ligand for β2AR, favors GRK5 and GRK6 for β2AR phosphorylation 11 . A previous study on phospho-site mapping showed that GRK2 and GRK6 phosphorylate the distal and the proximal sites, respectively, in the C-terminal tail of β2AR, demonstrating that isoproterenol and carvedilol provide different phosphorylation codes. In accordance with the β2AR study, our Ala mutant analysis indicated that Ser326, the most proximal Ser/Thr site in AT1R, is selectively phosphorylated by GRK5/6 ( Supplementary Fig. 3k, l). In mice, carvedilol improves cardiac function via β-arrestin2 52 . Isoproterenol, on the other hand, worsens cardiac dysfunction in a β-arrestin2-dependent manner 53 . These reports indicate that pharmacologically distinct ligands elicit different β-arrestindependent responses by preferentially utilizing specific GRKs. However, it remains unclear exactly how, at a molecular level, GRKs influence β-arrestin conformations and functions. Using AT1R as a model, our study is the first to provide clear evidence showing that two ligands, by utilizing different GRKs, induce distinct engagement of β-arrestin, leading to different AT1Rmediated cellular responses.
Pharmacological and genetic studies in mice have demonstrated distinct roles of GRKs in AT1R-involved myocardial contraction. Administration of Ang II or SII, a TRV027-like ligand, causes myocardial contraction via β-arrestin2 54 . Intriguingly, Ang II-induced and SII-induced phenomena were abolished in GRK2-deficient mice and GRK6-deficient mice, respectively, indicating that the biased ligand has GRK preference 54 . Our cell-based study also demonstrates that SII along with TRV027 show preference toward GRK5/6 while Ang II preferentially utilizes GRK2/3. Furthermore, although our study shows that β-arrestin-mediated pERK is different between Ang II and TRV027, among a diverse set of β-arrestin effector molecules 55 , there may be another β-arrestin function that is shared between the two ligands and mediates myocardial contraction.
It has previously been assumed that different structural states of β-arrestin induce distinct functions. Structural studies using negative-stain EM 19 and recent cryo-EM 20,56,57 have shown that β-arrestin has at least two distinct modes with GPCR interactions: the tail hanging (engagement in a phosphorylated receptor via its C-terminus) and the core engagement (C-terminal interaction plus deep binding to the receptor core via the finger-loop region of β-arrestin; here termed "full engagement"). The tail-hanging mode and the full engagement mode are called partially active and fully active, respectively, based on a proposed stepwise activation mechanism of β-arrestin 19 . A discrete mechanism, called catalytic activation, is documented for the β1 adrenergic receptor (β1AR)-induced β-arrestin activation by a study using total internal reflection fluorescence (TIRF) microscopy 51 . In this model, β-arrestin binding to β1AR is a transient event, but its active state sustains even after dissociation from the receptor. Unlike the tail-hanging and full engagement modes, this interaction does not rely on phosphorylation of the β1AR C-tail, but instead requires binding to the β1AR core (here termed "tail-free core engagement"). Thus, at least three modes of a GPCR-βarrestin complex are present.
Although the GPCR-β-arrestin-binding modes have been documented in significant detail as described above, their functional distinctions remain less clear. The tail-hanging and the full engagement modes of β-arrestin are capable of inducing both receptor internalization and β-arrestin-dependent Src signaling 21 . The tail-free core-engagement mode of β-arrestin triggers ERK activation 51 . However, no clear functional difference has previously been reported, in part owing to a lack of knowledge about biased ligands that induce different β-arrestin-engagement modes.
In this study, we described different engagement modes of β-arrestin2 by the three AT1R-stimulated conditions. Ang IIinduced all three modes, whereas the TRV/Ang II + YM treatments caused the occurrence of the two binding modes, except for the core-engagement mode ( Supplementary Figs. 3c, d, 9a-i). These results support the notion that the phosphorylation state of AT1R is important in the transition from the tail-hanging mode to the full engagement mode. The lack of full engagement in the Ang II + YM condition suggests that receptor phosphorylation by GRK5/6 alone is insufficient for binding of β-arrestin to the receptor core and that additional phosphorylation by GRK2/3 is required. Together, our sensor-based analysis provides an insight into how the distinct β-arrestin-engagement modes are regulated.
Importantly, we observed functional differences among these modes. First, we showed that receptor phosphorylation is necessary and sufficient for AT1R internalization. Interestingly, Ang II stimulation caused endosomal translocation of β-arrestin even under the GRK-depleted condition ( Supplementary  Fig. 6f-k), whereas AT1R itself remained on the cell surface (Fig. 3a-c and Supplementary Fig. 6a-e). This suggests that AT1R internalization requires binding of β-arrestin to the phosphorylated receptor tail and that active β-arrestin, via the catalytic activation mechanism, is translocated to the endosome without forming a stable AT1R-β-arrestin complex. Second, we observed differences in β-arrestin-dependent signaling (Fig. 3d). Among the three conditions, Ang II potently induced pERK response, highlighting a signaling role of the full-engaged β-arrestin from the others. Together, our results elucidate the relationship of the β-arrestin-engagement modes with their functional outcomes.
Finally, we observed functional similarities and differences between β-arrestin1 and β-arrestin2. Although numerous signaling molecules have been reported to bind to both β-arrestin1 and β-arrestin2, it is generally assumed that β-arrestin1 plays a role in desensitization and internalization whereas β-arrestin2 does so in signaling. In this regard, many physio-pharmacological effects are attributed specifically to β-arrestin2 58 . In this study, we found that both β-arrestin subtypes showed virtually indistinguishable functionalities including GRK subtype preference, receptor endocytosis and ERK phosphorylation with the exception of β-arrestin2 being capable of the tail-free core engagement ( Supplementary Fig. 3c, d, Supplementary Fig. 9a-c). These observations suggest that β-arrestin1 and β-arrestin2 mostly function redundantly and that GPCRs that are not susceptible to GRK-mediated phosphorylation or that lack the phosphorylation code 59,60 , depend on β-arrestin2 due to their inability to associate with β-arrestin1. Altogether, this study clearly links the mechanisms for activation of multiple β-arrestins and their different functions.
In conclusion, we report a novel regulatory mechanism for GRKs and β-arrestins. In contrast to the previous assumption that G protein and β-arrestin are separable, our finding highlights the close relationship between the two GPCR effectors specifically, through a Gq-regulated GRK switch that leads to transducer bias at β-arrestin. These findings provide a conceptual foundation for biased signaling based drug discovery.
For the HiBiT-based receptor internalization assay, human full-length AT1R was N-terminally fused to a HiBiT cassette (HiBiT-AT1R), which contained the Interleukin 6-derived signal sequence followed by the HiBiT sequence and a linker (MNSFSTSAFGPVAFSLGLLLVLPAAFPAPVSGWRLFKKISGGSGGGGSG; HiBiT tag underlined; gene synthesized with codon optimization). For the NanoBiT-β-arrestin-translocation assay, human full-length β-arrestin1 and β-arrestin2 were N-terminally fused to a small fragment (SmBiT) of the NanoBiT luciferase with a 15-amino acid flexible linker (GGSGGGGSGGSSSGG) and the resulting constructs were referred to as Sm-β-arrestin1 and Sm-β-arrestin2, respectively. For the Ib30 assay, the β-arrestin1 conformation-sensitive intrabody 30 44 was N-terminally fused to SmBiT with the flexible linker and the resulting construct was referred to as Sm-Ib30. A FYVE domain of human Endofin (amino acid regions Gln739-Lys806; gene-synthesized with codon optimization by Genscript) 63 was C-terminally fused to a large fragment (LgBiT) with the flexible linker and the resulting construct was referred to as Endo-Lg.
For the flow cytometry analysis, human full-length AT1R were N-terminally fused to the FLAG epitope tag with the preceding hemagglutinin-derived signal sequence (MKTIIALSYIFCLVFADYKDDDDKGGSGGGGSGGSSSGGG; FLAG epitope tag underlined). The FLR truncation (corresponding to the amino acid regions Gly64-Thr74 and Gly65-Thr75 in β-arrestin1 and β-arrestin2, respectively) was introduced according to a previous study 21 . For Δtail-phos AT1R mutant, all twelve Ser and Thr in the C-terminal region of the AT1R (Supplementary Fig. 9e, f) were replaced with Ala (amino acid regions Pro321-Glu359; PPKAKAHANLAAKMAALAYRPADNVAAAAKKPAPCFEVE; Ala substitutions underlined). For the phospho-code AT1R mutant, the seven Ser and Thr within the potential phosphorylation codes 59 in the C-terminal region ( Supplementary  Fig. 5e) were individually replaced with Ala (PPKAKSHSNLSTKMSTLSYRPSDNVSSSTKKPAPCFEVE; Ala-substituted residues are underlined).
HEK293 cells deficient for the Gq family members and the GRK family members (GRK2, GRK3, GRK5 and GRK6) were generated by mutating the GNAQ and the GNA11 genes (encoding Gαq and Gα11, respectively) from the ΔGRK2/3/5/6 (CL2) cells using previously validated sgRNA sequences for the two genes 66 . We transfected a mixture of GNAQ sgRNA-containing pX330 (a gift from Feng Zhang at the Broad Institute; plasmid No. 42230) and GNA11 sgRNAcontaining PX458 plasmids (sgRNA-targeting sites and the SpCas9 PAM sequences are 5'-AAACAAGAAAGATCTTCTAGAGG-3' for GNAQ [Xba I] and 5'-AGGGTACTCGATGATGCCGGTGG-3' for GNA11 [Hap II]). By following the procedure as described above, we isolated clonal cells and screened for their genotype by the restriction-enzyme method. Successful targeting was confirmed by the immunoblot analysis.
Quantitative real-time PCR. Total RNA of HEK293 cells was isolated using a GenElute Mammalian Total RNA Miniprep Kit (Sigma-Aldrich). The purified RNA was reverse-transcribed using a High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific) according to manufacturer instructions. Real-time quantitative PCRs were performed with SYBR Premix Ex Taq (Takara Bio) and monitored by ABI Prism 7300 (Applied Biosystems). ΔΔCT method was used to calculate relative expression levels of cDNA samples. Primers were as follows: NanoBiT assays. β-arrestin recruitment to AT1R was measured by the NanoBiT β-arrestin-recruitment assay 62 . The parent and the GRK-deficient cell lines were transfected with a mixture of 100 ng Lg-β-arrestin and 500 ng AT1R-Sm plasmids (per well in a 6-well plate unless otherwise noted). After incubation for 1 day, transfected cells were collected with 0.5 mM EDTA-containing Dulbecco's PBS (D-PBS), centrifuged and suspended in 2 mL of HBSS containing 0.01% bovine serum albumin (BSA; fatty-acid-free grade; SERVA) and 5 mM HEPES (pH 7.4) (assay buffer). The cell suspension was dispensed in a white 96-well plate at a volume of 80 μl per well (hereafter 96-well plate) and loaded with 20 μl of 50 μM coelenterazine (CTZ, Carbosynth) diluted in the assay buffer. After 2-h incubation at room temperature, the plate was measured for baseline luminescence (SpectraMax L equipped with 2PMT, Molecular Devices; SoftMax Pro Version 7.03, Molecular Devices) and 20 μl of 6× ligand diluted in the assay buffer or the assay buffer alone (vehicle) were manually added. The plate was read for 15 min with an interval of 20 sec or 40 sec at room temperature. The luminescence counts from 13 min to 15 min (unless otherwise indicated) after ligand addition were averaged and normalized to the initial counts. We note that due to the time lag between the manual ligand addition and start of plate reader measurement (typically, 30-60 s), there were already changes for some of the early responding sensors (e.g., GRK recruitment and GRK-Gβγ association) at the first measurement (labeled as 0 min in the x-axis).
Gq activation was measured by the NanoBiT-G-protein-dissociation assay 62 with minor modifications. Plasmid transfection was performed by using a mixture of 100 ng Gαq with LgBiT insertion at the loop between the αA and the αB helices (the residue 97-98), 500 ng Gβ1, 500 ng SmBiT-fused Gγ2 (C68S), 100 ng RIC8A and 200 ng AT1R plasmids (per well in a 6-well plate). The transfected cells were subjected to the NanoBiT assay as described in above and the luminescence counts from 3 min to 5 min after ligand addition were used for quantification.
An active β-arrestin conformation was detected by the NanoBiT-Ib30 assay 44 with minor modifications. Transfection was performed by using a mixture of 500 ng Sm-Ib30, 500 ng Lg-β-arrestin and 200 ng GPCR plasmids. The transfected cells were subjected to the NanoBiT assay as described above.
β-arrestin endosomal translocation was measured by a bystander approach 63 , but with a NanoBiT modification. Transfection was performed by using a mixture of 500 ng Endo-Lg, 100 ng Sm-β-arrestin and 200 ng GPCR plasmids. The transfected cells were subjected to the NanoBiT assay as described in above.
GRK recruitment to AT1R was measured by using the Lg-GRK and the AT1R-Sm constructs. Transfection was performed by using a mixture of 200 ng Lg-GRK and 200 ng AT1R-Sm plasmids. The transfected cells were subjected to the NanoBiT assay as described in above and the luminescence counts from 3 to 5 min after ligand addition were used for quantification.
For the Gαq-GRK proximity assay, transfection was performed by using a mixture of 500 ng Gαq-Lg, 500 ng Sm-GRK5 or Sm-GRK6, 500 ng Gβ1, 500 ng Gγ2, 100 ng RIC8A and 200 ng AT1R plasmids. For the Gβγ-GRK proximity assay, transfection was performed by using a mixture of 100 ng GRK5-Lg or GRK6-Lg, 500 ng Sm-Gβ1, 500 ng Gγ2, 100 ng RIC8A and 200 ng AT1R plasmids. The transfected cells were subjected to the NanoBiT assay as described in above and the luminescence counts from 3 to 5 min after ligand addition were used for quantification.
For downstream of Gq activation, we used the NanoBiT-IP 3 assay 57 . Transfection was performed by using a mixture of 1 µg Lg-IP3R2-Sm and 200 ng AT1R plasmids. For downstream of G12 activation, we used the NanoBiT-RhoA assay 57 . Transfection was performed by using a mixture of 100 ng Lg-RhoA, 500 ng Sm-PKN1-GBD and 200 ng AT1R plasmids. The transfected cells were subjected to the NanoBiT assay as described in above and the luminescence counts from 3 min to 5 min after ligand addition were used for quantification.
HiBiT-based receptor internalization assay. AT1R internalization was measured by HiBiT-based receptor internalization assay 69 . The parent and the GRK-deficient cell lines were transfected with 100 ng HiBiT-AT1R plasmid (per well in a 6-well plate). For the experiments using the β-arrestin1/2-deficient cells, ΔARRB cells were transfected with 100 ng HiBiT-AT1R plasmid and 100 ng Sm-β-arrestin plasmid. After incubation for 1 day, cells were harvested, suspended in 1 mL of the assay buffer, dispensed in a white 96-well half-area plate at a volume of 25 μL per well, and mixed with 25 μL of 2× substrate buffer consisting of 1:200 of a LgBiT stock solution (Promega) and 20 μM furimazine in the assay buffer. After 40 min incubation at room temperature, the plate was measured for baseline luminescence, and a titrated ligand (10 μL) diluted in the 1× substrate buffer was manually added. The plate was immediately read at room temperature for the following 30 min at a measurement interval of 30 sec. The luminescence counts from 27 min to 30 min after ligand addition were averaged and normalized to the vehicle stimulation count.
AlphaLISA-based ERK1/2 phosphorylation assay. ERK1/2 activation was monitored with the AlphaLISA assay (PerkinElmer). The parent and ΔARRB cells were transfected with 200 ng AT1R plasmid (per well in a 6-well plate) with or without 100 ng Sm-β-arrestin plasmid. After incubation for 1 day, transfected cells were collected, suspended in the complete DMEM at a concentration of 10 5 cells per ml and reseeded in a 96-well plate (10 4 cells per well with 100 µL). One day later, the cells were serum-starved in DMEM overnight. The cells were then stimulated for 30 min with ligands and lysed in 100 µL of a lysis buffer. The lysate was diluted 1:10 and analyzed according to a manufacturer's protocol. Alpha counts were read on the EnSpire microplate reader (PerkinElmer).
Glosensor cAMP assay. Plasmid transfection was performed in a 6-well plate with a mixture of 1 µg of Glo-22F cAMP biosensor (human codon-optimized and genesynthesized)-encoding pCAGGS plasmid and 200 ng of the human vasopressin V2R-encoding plasmid or the human opioid µOR plasmid. After 1-day incubation, the transfected cells were harvested with 0.53 mM EDTA-containing D-PBS, centrifuged at 190 g for 5 min and suspended in the assay buffer described in the NanoBiT assay (vehicle; 1 mL per well). The cells were seeded in a half-area white 96-well plate (30 µL per well) and loaded with D-luciferin potassium solution (10 µL of 8 mM solution per well; FujiFilm Wako Pure Chemical). After 2 h incubation in the dark at room temperature, the plate was read for its initial luminescent count (SpectraMax L, Molecular Devices). For the Gs assay, the V2Rexpressing cells were treated with vehicle, 100 nM arginine vasopressin (AVP, Peptide Institutes) or 10 µM forskolin (FujiFilm Wako Pure Chemical) (10 µL of 5X solution per well). For the Gi assay, the µOR-expressing cells were treated with 10 µM forskolin (vehicle) or forskolin-supplemented [D-Ala2, N-Me-Phe4, Gly5ol]-Enkephalin acetate salt (DAMGO at 1 µM; Sigma-Aldrich). Thereafter, kinetics were measured for 20 min and expressed as fold-change values. Fold-change luminescent signals from 17 min to 20 min after compound addition were averaged and normalized to those in forskolin-treated condition.
TGF-α shedding assay. Gq activation was measured by TGF-α shedding assay 70 . Plasmid transfection was performed in a 6-well plate with a mixture of 500 ng of the alkaline phosphatase-tagged-TGF-α (AP-TGF-α) reporter-encoding pCAGGS plasmid and 200 ng of the human histamine H1 receptor. After 1-day culture, the transfected cells were harvested by trypsinization, pelleted by centrifugation at 190 g for 5 min and washed once with the HBSS containing 5 mM HEPES (pH 7.4). After centrifugation, the cells were resuspended in 6 mL of the HEPES-containing HBSS. The cell suspension was seeded in a clear 96-well culture plate (cell plate) at a volume of 90 µl (per well hereafter) and incubated for 30 min in a 5% CO 2 incubator at 37°C. The cells were treated with 10 µM histamine (diluted in HBSS containing 5 mM HEPES (pH 7.4) and 0.01% (w/v) BSA). After spinning the cell plates, conditioned media (80 µL) was transferred to an empty 96-well plate (conditioned media (CM) plate). AP reaction solution (10 mM p-nitrophenylphosphate (p-NPP), 120 mM Tris-HCl (pH 9.5), 40 mM NaCl, and 10 mM MgCl 2 ) was dispensed into the cell plates and the CM plates (80 µL). Absorbance at 405 nm (Abs 405 ) of the plates was measured, using a microplate reader (SpectraMax 340 PC384, Molecular Devices), before and after 1-h incubation at room temperature. Ligand-induced AP-TGF-α release signal was calculated by subtracting spontaneous AP-TGF-α release signa from total ligand-induced AP-TGF-α release response.
For the Gq addback experiments using the GRK/Gq-6KO cells, 100 ng of the untagged Gα subunit plasmids was individually expressed along with the AP-TGFα and the H1 receptor as above.
In the single-molecule imaging, we visually screened the imaging regions where the density of particles (for each color) was less than~1 particle per µm 2 , an upperthreshold that one can visualize single-particles of fluorescently labeled proteins. In the meantime, a too low density of particles prevents accurate quantification. We typically observed regions contained 0.3~0.6 particles per µm 2 . Once we chose a cell that expressed the optimal particle density, we unbiasedly took 300 frame dualcolor movies per cell with 30.5 ms exposure time between 15 and 30 min after ligand (or vehicle) stimulation (512 × 512 pix images, pixel size: 67 nm/pix). We performed two independent microscopy experiments, for each experimental condition, with recording of 10-15 single cells per coverslip.
In our analysis procedure, after quality checking of recorded movies (e.g., those with out of focus or frame dropping during video recording were omitted), no intentional data exclusion was made. The image processing (subtract background with 25 pixels rolling ball radius, and two frames running average with Running_ZProjector plugin, http://valelab.ucsf.edu/~nstuurman/ijplugins/) were performed by ImageJ with batch processing macro (https://github.com/masatakayanagawa/ImageJ-macro-ImageProcessingSMT). The alignment between the two channels and SMT analysis were performed using AAS, Zido 71 . The VB-HMM analysis was performed by a LabVIEW-based homemade program or AAS 46,71 . The four-state model was selected that showed the highest lower bound in the VB method ( Supplementary Fig. 10g). All subsequent analyses (diffusion dynamics, intensity distribution, colocalization, and statistical analysis) were performed using smDynamicsAnalyzer (https://github.com/masataka-yanagawa/IgorPro8-smDynamicsAnalyzer), an Igor Pro 8.0, WaveMetrix (Igor)-based homemade program. Refer to the reference for the detailed instruction and curve fitting functions 71 . Here, the colocalization of two particles is detected if the two particles are within 100 nm in the same frame, which corresponds to 2~3 SDs of the total position accuracy, and if the two particles are in the same diffusion state. The localization precision of each diffusion state was estimated as shown in Supplementary Fig. 10h based on the equation 6 of Mortensen et al. 72 .
Endocytosis imaging. Plasmid DNAs (2 µg in total per 6-cm dish) encoding FLAG-AT1R and/or SNAP-tag-fused β-arrestin2 in pcDNA3.1 were transfected into the parent and the GRK-deficient cell lines by Lipofectamine 3000 on the day before imaging. SNAP-tag was stained by 100 nM SNAP-Cell TMR-Star ligand and washed as described above. FLAG-tag was labeled by anti-FLAG monoclonal antibody (clone 1E4; 1:100 in 0.01% BSA-HBSS) for 30 min and washed by 0.01% BSA-HBSS on ice. Then, the cells were stained by antimouse IgG-AlexaFluor 488 (1:200 in 0.01% BSA-HBSS) for 30 min and washed with 0.01% BSA-HBSS on ice. Three µM Go6983 with or without 1 µM YM-254890 were added 15 min before ligand stimulation and cells were brought back to room temperature. Fifteen min after ligand stimulation (Vehicle, 1 µM Ang II, or 1 µM TRV027 at final concentration), the cells were fixed by PBS containing 4% paraformaldehyde and 0.2% glutaraldehyde for 30 min at room temperature. Dual-color images (focus: approximately 2 µm above coverslip, exposure time: 30.5 ms, pixel size: 67 nm/pix) were taken by the same microscope system as described in the single-molecule imaging section with oblique illumination settings. The dichroic and emission filters were changed to the following configurations (lasers: Sapphire 488-200, and OBIS 561 nm, Coherent, dichroic: ZT488/561rpc and T560lpxr, Chroma, emission filters in the two-channel imaging system: ET525/50 m, ET605/70 m, Chroma).
The image processing (split channels, merge channels, crop region of interest, and intensity measurement) was performed by ImageJ with batch processing macro (https://github.com/masataka-yanagawa/ImageJ-macro-ImageProcessingSMT). Cell regions were manually cropped, and measured the mean intensity of each channel (I NanoBiT , I Venus ). Calculation of BRET index, data visualization as violin/ box plots and statistical analysis were performed in Igor, where BRET index was defined as follows: BRET index = I Venus /(I NanoBiT + I Venus ). The fold-change was calculated for each cell based on the intensity ratio between before and approximately 14 min after stimulation.
Statistical analyses. Statistical analyses were performed using the GraphPad Prism 8 software (GraphPad) or Igor and methods are described in the legends of the figures. Representation of symbols and error bars is described in the legends. Symbols are either mean values of indicated numbers of independent experiments or numbers of measured cells from a single experiment. Unless otherwise noted, error bars and shaded areas denote SEM. Concentration-response curves were fitted to all data by the Nonlinear Regression: Variable slope (four parameter) in the Prism 8 tool with a constraint of the Hill Slope of absolute value less than 2. For multiple comparison analysis, one-way or two-way ANOVA and following the Dunnett's test or the Tukey's test or the Sidak test, the multiple t-test or two-tailed Welch's t-test was used.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data generated or analyzed during this study are included in this article, its Supplementary Information and Source Data file or from the corresponding authors on reasonable request. The Source data of Figs. 1-6 and Supplementary Fig. 2-18 are provided in the separated Source Data file. Source data are provided with this paper.