Key phosphorylation sites in GPCRs orchestrate the contribution of β‐Arrestin 1 in ERK1/2 activation

Abstract

β‐arrestins (βarrs) are key regulators of G protein‐coupled receptor (GPCR) signaling and trafficking, and their knockdown typically leads to a decrease in agonist‐induced ERK1/2 MAP kinase activation. Interestingly, for some GPCRs, knockdown of βarr1 augments agonist‐induced ERK1/2 phosphorylation although a mechanistic basis for this intriguing phenomenon is unclear. Here, we use selected GPCRs to explore a possible correlation between the spatial positioning of receptor phosphorylation sites and the contribution of βarr1 in ERK1/2 activation. We discover that engineering a spatially positioned double‐phosphorylation‐site cluster in the bradykinin receptor (B₂R), analogous to that present in the vasopressin receptor (V₂R), reverses the contribution of βarr1 in ERK1/2 activation from inhibitory to promotive. An intrabody sensor suggests a conformational mechanism for this role reversal of βarr1, and molecular dynamics simulation reveals a bifurcated salt bridge between this double‐phosphorylation site cluster and Lys²⁹⁴ in the lariat loop of βarr1, which directs the orientation of the lariat loop. Our findings provide novel insights into the opposite roles of βarr1 in ERK1/2 activation for different GPCRs with a direct relevance to biased agonism and novel therapeutics.

Distinct spatial positioning of key phosphorylation sites in different GPCRs plays a decisive role in the contribution of β‐arrestin 1 in agonist‐induced ERK1/2 activation. These findings have direct relevance for biased agonism, and for designing GPCR‐targeted novel therapeutics.

Introduction

G protein‐coupled receptors (GPCRs) recognize a diverse array of ligands but exhibit broadly conserved patterns of transducer coupling and regulatory paradigms (Bockaert & Pin, 1999). For example, agonist‐induced receptor phosphorylation promotes coupling of multifunctional proteins called β‐arrestins (βarrs), which are critically involved in the regulation of GPCR signaling and trafficking patterns (Freedman & Lefkowitz, 1996; Lefkowitz & Shenoy, 2005; DeWire et al, 2007). The ability of βarrs to mediate downstream signaling cascades has yielded new paradigms of GPCR signaling and led to the conceptual framework of biased agonism (Azzi et al, 2003; Wei et al, 2003; Shukla et al, 2011; Reiter et al, 2012). G protein and βarr bias has been described for a number of GPCRs, and in many cases, distinct functional profiles of these two pathways in terms of cellular and physiological outcomes have also been established (Luttrell & Gesty‐Palmer, 2010; Appleton & Luttrell, 2013; Gesty‐Palmer et al, 2013; Peterson & Luttrell, 2017). Although βarrs’ contributions are documented in a number of downstream signaling pathways across different GPCRs, agonist‐induced ERK1/2 MAP kinase phosphorylation has been one of the most common readout to probe βarr signaling profile and biased agonism (Azzi et al, 2003; Wei et al, 2003; Lefkowitz & Shenoy, 2005; DeWire et al, 2007).

βarrs are typically observed to contribute positively in ERK1/2 MAP kinase phosphorylation and activation; however, in some cases, the two isoforms, namely βarr1 and 2, play opposite roles (DeWire et al, 2007; Srivastava et al, 2015). For example, depletion of βarr1 results in an increase in ERK1/2 phosphorylation while reducing the levels of βarr2 leads to significant decrease for several GPCRs including the angiotensin receptor (AT_1aR) and bradykinin receptor (B₂R; Ahn et al, 2004; Zimmerman et al, 2011). A mechanistic understanding for this intriguing functional diversity between the two βarr isoforms is currently lacking, and it represents a missing link in our understanding of GPCR‐βarr signaling system. Therefore, understanding the details of βarrs’ contribution in ERK1/2 activation, especially the diversity across different receptor systems, requires additional studies.

Receptor phosphorylation is a key determinant of the interaction between βarrs and GPCRs, and it is well established that differential phosphorylation patterns on the receptor can fine‐tune βarr conformation and ensuing functional outcomes (Gurevich & Gurevich, 2004, 2018b; Reiter & Lefkowitz, 2006; Ranjan et al, 2017; Chen et al, 2018). For example, receptor phosphorylation by different GRKs results in distinct phosphorylation patterns, which in turn guide different conformations in recruited βarrs and functional outcomes, a framework referred to as the “bar code” mechanism (Kim et al, 2005; Ren et al, 2005; Shukla et al, 2008). More recently, a “phosphorylation code”‐based mechanism has been proposed for GPCR‐βarr interaction based on the crystal structure of rhodopsin–arrestin complex (Zhou et al, 2017). Considering the three conserved positively charged pockets on the N‐domain of arrestin, the requirement of at least three phosphorylated residues in GPCRs, which are separated by additional residues forming a spatial pattern, was conceived for high‐affinity interaction between the receptor and arrestin in this study (Zhou et al, 2017). This spatial arrangement of phosphorylation sites on GPCRs was referred to as “phosphorylation code” with two different patterns, i.e., PxPxxP/E/D (short code) and PxxPxxP/E/D (long code), where P refers to a phospho‐Ser or phospho‐Thr and X refers to any other amino acid except proline (Zhou et al, 2017). While the contribution of these phosphorylation patterns in arrestin recruitment was experimentally measured for rhodopsin and β2 adrenergic receptor, a direct correlation between such phosphorylation patterns and βarr‐mediated signaling, if any, was not investigated (Zhou et al, 2017).

Here, we set out to probe the contribution of spatial positioning and pattern of phosphorylation sites in selected GPCRs in determining the role of βarr1 in ERK1/2 phosphorylation. We discover that certain key positions, in the context of phosphorylation site clusters, orchestrate the differential contribution of βarr1 in agonist‐induced ERK1/2 activation for different GPCRs. An intrabody sensor suggests a conformational mechanism for this interesting phenomenon, which is further corroborated by molecular dynamics (MD) simulation. Our data offer mechanistic insights into distinct role of βarr1 in the activation of ERK1/2 MAP kinase downstream of different GPCRs, and it has direct implications for the novel paradigms of GPCR signaling and biased agonism.

Results

Receptor‐specific contribution of βarr1 in ERK1/2 MAP kinase activation

In order to investigate the opposite contribution of βarr1 in agonist‐induced ERK1/2 phosphorylation, and to link it with specific receptor phosphorylation sites, we first chose the V₂R and B₂R as model systems. Both of these receptors were originally categorized as “class B” GPCRs in terms of βarr recruitment and trafficking patterns (Oakley et al, 2000). That is, they interact with βarrs in a stable fashion and prolonged exposure with agonist results in their endocytotic trafficking together with βarrs. A subsequent study however differentiated B₂R from other class B GPCRs including V₂R, by reporting that βarrs rapidly dissociate from this receptor in endosomes followed by receptor recycling to the plasma membrane (Simaan et al, 2005). The most interesting aspect that prompted us to choose these two receptor systems is that previous studies have demonstrated strikingly different contribution of βarr1 in agonist‐induced ERK1/2 phosphorylation for these two receptors (Ren et al, 2005; Charest et al, 2007; Zimmerman et al, 2011; Oligny‐Longpre et al, 2012; Ghosh et al, 2019). While βarr1 knockdown leads to a significant reduction in ERK1/2 phosphorylation downstream of V₂R (Ghosh et al, 2019), it results in an enhanced level of ERK1/2 phosphorylation for the B₂R (Zimmerman et al, 2011). In other words, βarr1 exerts a supportive role in ERK1/2 phosphorylation for the V₂R but an inhibitory role for the B₂R.

We first analyzed the potential phosphorylation site patterns in the carboxyl‐terminus of these two receptors with reference to a recent study that has proposed “phosphorylation codes”‐based mechanism of GPCR‐βarr interaction (Zhou et al, 2017). This study suggested that spatial distribution of potential phosphorylation sites on the receptor is separated by additional residues to constitute PXPXXP and PXXPXXP type patterns, where P is pSer/pThr and X is any other amino acid (Zhou et al, 2017). We observed that both of these receptors (i.e., V₂R and B₂R) harbor two such phosphorylation site patterns in their carboxyl‐terminus, i.e., ³⁵⁷ SCTTAS ³⁶² (i.e., PXPXXP) and ³⁵⁷ SCTTASS ³⁶³ (i.e., PXXPXXP) in V₂R and ³⁶⁶ SMGTLRT ³⁷² (i.e., PXXPXXP) and ³⁶⁹ TLRTSIS ³⁷⁵ (PXXPXXP) in B₂R (Fig 1A). We also confirmed similar recruitment and trafficking patterns of βarr1 for V₂R and B₂R in HEK‐293 cells by confocal microscopy and observed that agonist stimulation leads to surface localization of βarr1‐mCherry first (2–5 min), followed by endosomal trafficking upon prolonged exposure (15–30 min; Fig 1B).

Figure 1. βarr1 has opposite contributions in agonist‐induced ERK1/2 activation for V₂R and B₂R.

• A
  G‐protein‐coupling preference and phospho‐site patterns in the carboxyl‐terminus of V₂R and B₂R, deduced based on a previous study (Zhou et al, 2017). The phospho‐site patterns in the form of PXPXXP and PXXPXXP are underlined and color‐coded. Both V₂R and B₂R recruit βarrs in “class B” pattern (Oakley et al, 2000) as reflected by stable interaction and endosomal trafficking of the receptor‐βarr complexes.
• B
  Confocal microscopy reveals typical “class B” pattern of βarr1 recruitment for V₂R and B₂R as reflected by first the localization at the plasma membrane and subsequently, internalization in endosomal vesicles upon agonist stimulation. HEK‐293 cells expressing V₂R/B₂R and βarr1‐mCherry were stimulated with agonist (AVP; 100 nM and Bradykinin; 100 nM), and the localization of βarr1 was visualized using confocal microscopy. Representative images from three independent experiments are shown here, and the scale bar is 10 μm. Visual scoring of images from three independent experiments revealed agonist‐induced βarr1 recruitment (i.e., membrane and endosomal localization) in approximately 77% of the cells for V₂R (221 cells) and 75% of the cells for B₂R (662 cells).
• C, D
  Agonist‐induced phosphorylation of ERK1/2 in HEK‐293 cells expressing either V₂R or B₂R in the absence (control; CTL) and presence of βarr1 knockdown (βarr1‐KD) are measured using Western blotting. Densitometry‐based quantification of data (mean ± SEM) from four independent experiments is presented as bar graphs in the right panels, normalized with respect to maximal dose under control condition (treated as 100%), and analyzed using two‐way ANOVA with Bonferroni multiple comparisons test (***P < 0.001, **P < 0.01).

Next, we measured agonist‐induced ERK1/2 phosphorylation downstream of V₂R and B₂R in HEK‐293 cells under control and βarr1 knockdown conditions. In agreement with previous studies, we also observed that knockdown of βarr1 yielded a significant reduction in ERK1/2 phosphorylation for V₂R while it augmented the levels of ERK1/2 phosphorylation in the case of B₂R (Fig 1C and D). This striking difference between the V₂R and B₂R suggests that the mechanistic basis of βarr1 contribution in agonist‐induced ERK1/2 activation is determined at levels other than their trafficking patterns and the number of phosphorylation site patterns.

Engineering a double‐threonine cluster in B₂R reverses the contribution of βarr1

A comparison of the carboxyl‐terminus sequences of the V₂R and B₂R revealed key differences in the spatial distribution of the potential phosphorylation sites (Fig 2A). While V₂R has “PXXPXXP” and “PXPXXP” type phospho‐site patterns (where P is Ser/Thr and X is any residues including Ser/Thr), both the phospho‐site patterns present in the B₂R are of “PXXPXXP” type (Fig 2A). This prompted us to generate a set of B₂R mutants that resemble the spatial pattern of phosphorylation sites in V₂R and test the contribution of βarr1 in ERK1/2 activation upon their activation (Fig 2B). These mutants expressed at comparable levels to B₂R^WT in HEK‐293 cells (Fig EV1A), and the surface expression of the individual mutants between the control and βarr1 knockdown conditions were also comparable (Fig EV1B and C). These mutants also displayed typical “class B” pattern of βarr1 trafficking as assessed by confocal microscopy, similar to B₂R^WT (Fig 2C). Moreover, we also observed that these mutants are capable of robustly recruiting βarr1, similar to B₂R^WT, as measured using a NanoBiT assay (Dixon et al, 2016; Shihoya et al, 2018), although B₂R^ΔI374 exhibits slightly lower level of βarr1 interaction (Fig 2D).

Figure 2. Spatial distribution of phospho‐sites and βarr1 recruitment of B₂R mutants.

• A
  Phospho‐site clusters in the carboxyl‐terminus of V₂R and B₂R are underlined and color‐coded to reflect the PXPXXP and PXXPXXP patterns as proposed in a previous study (Zhou et al, 2017).
• B
  A number of B₂R mutants were generated to mimic the spatial distribution of phospho‐site pattern of Ser/Thr as present in the V₂R. The first three mutants were designed to target the proximal part of the phospho‐site pattern (indicated in red dotted box) while the fourth mutant was designed to target the distal part (indicated in blue dotted box).
• C
  Confocal microscopy reveals robust recruitment of βarr1 to B₂R mutant constructs upon agonist stimulation. HEK‐293 cells expressing either B₂R^WT or B₂R mutants along with βarr1‐mCherry were stimulated with agonist (Bradykinin; 100 nM), and the localization of βarr1 was visualized using confocal microscopy. Representative images from three independent experiments are shown here, and the scale bar is 10 μm. Visual scoring from three independent experiments revealed agonist‐induced βarr1 recruitment (i.e., membrane and endosomal localization) in approximately 75% of the cells for B₂R^WT (662 cells), 74% of the cells for B₂R^ΔG368 (169 cells), 75% of the cells for B₂R^L370T (130 cells), 85% of the cells for B₂R^ΔG368/L370T (132 cells), and 77% of the cells for B₂R^ΔI374 (116 cells).
• D
  HEK‐293 cells expressing B₂R constructs with C‐terminal SmBiT and βarr1 with N‐terminal LgBiT were treated with indicated concentrations of bradykinin, and ligand‐induced change in luminescent signal was measured. Concentration‐response curves were plotted using GraphPad Prism, and pEC₅₀ (top) and E_max (bottom) were calculated. Data represent mean ± SEM of five independent experiments (three for B₂R^ΔI374), each performed in duplicate.

Figure EV1. Surface expression of B₂R and AT_1aR mutants.

• A
  HEK‐293 cells were transfected with indicated B₂R constructs, and 48‐h post‐transfection, their surface expression was measured using a whole cell ELISA assay. Data (mean ± SEM) are normalized with respect to B₂R^WT (treated as 100%) and represent four independent experiments.
• B
  For validation of βarr1 knockdown, HEK‐293 cells stably expressing either control or βarr1‐shRNA were lysed, and cellular lysate was subjected to Western blotting. βarrs and β‐actin were detected using the corresponding antibodies. βarr1 knockdown was typically assessed in each experiment measuring ERK1/2 phosphorylation, and representative samples from two different experiments are shown here. Densitometry‐based quantification reflects approximately 50–70% βarr1 knockdown.
• C, D
  Surface expression of B₂R and AT_1aR mutants in control (CTL) and βarr1 knockdown (βarr1‐KD) cells were measured using whole cell surface ELISA and found to be similar under both conditions. Data (mean ± SEM) represent four independent experiments, normalized with respect to the surface expression in control cells for each receptor mutant (treated as 100%).

Among these mutants, the deletion of Gly³⁶⁸, i.e., B₂R^ΔG368, changes the phosphorylation site pattern from PXXPXXP pattern to PXPXXP. Still however, the inhibitory contribution of βarr1 in ERK1/2 activation remains unchanged, i.e., knockdown of βarr1 augments agonist‐induced ERK1/2 phosphorylation, similar to that of B₂R^WT (Fig 3A). Strikingly, either the mutation of Leu³⁷⁰ alone, i.e., B₂R^L370T, or in combination with Gly³⁶⁸ deletion, i.e., B₂R^ΔG368/L370T, reverses the contribution of βarr1 in ERK1/2 activation from inhibitory to supportive, and we observe a significant reduction in ERK1/2 phosphorylation upon βarr1 knockdown (Fig 3B and C). It is interesting to note that the PXXPXXP pattern is not altered in B₂R^L370T and B₂R^ΔG368/L370T mutants. Similar to B₂R^ΔG368, the deletion of Ile³⁷⁴, i.e., B₂R^ΔI374, which disrupts the second PXXPXXP pattern, also does not affect the inhibitory contribution of βarr1 in agonist‐induced ERK1/2 phosphorylation (Fig 3D). These experiments presented in Fig 3 were carried out in HEK‐293 cells where βarr1 was knocked down using stable expression of shRNA against βarr1. We also corroborated the role reversal of βarr1 in ERK1/2 phosphorylation for the selected B₂R mutants (B₂R^L370T and B₂R^L370T/ΔG368) using CRISPR/Cas9 edited HEK‐293 cells expressing either both βarrs or only βarr2 (Fig EV2). We observed a pattern of ERK1/2 phosphorylation (Fig EV2), which is very similar to that measured with shRNA approach (Fig 3).

Figure 3. Contribution of βarr1 in ERK1/2 phosphorylation for B₂R mutants.

(A–D) Knockdown of βarr1 leads to an increase in agonist‐induced ERK1/2 phosphorylation for B₂R^ΔG368 and B₂R^ΔI374 but a decrease for B₂ R^{L 370T} and B₂R^ΔG368/L370T mutants. HEK‐293 cells expressing the indicated B₂R mutants in the presence and absence of βarr1 knockdown were stimulated with indicated doses of bradykinin (Brady) for 10 min followed by detection of phosphorylated ERK1/2 using Western blotting. Densitometry‐based quantification of data (mean ± SEM) from three independent experiments (five for B₂R^ΔG368 and B₂R^ΔI374) is presented as bar graphs in the lower panels. Data are normalized with respect to maximal dose under control condition (treated as 100%) and analyzed using two‐way ANOVA with Bonferroni multiple comparisons test (*P < 0.05; **P < 0.01; ***P < 0.001).

Figure EV2. Contribution of βarr1 in ERK1/2 phosphorylation for B₂R mutants in CRISPR/Cas9 cells.

HEK‐293 cells with βarr1/2 deletion using CRISPR/Cas9 were transfected with βarr2 and indicated B₂R constructs together with either βarr1 or pcDNA vector. Subsequently, cells were stimulated with 100 nM bradykinin (BK) for 10 min and ERK1/2 phosphorylation was measured by Western blotting. Densitometry‐based quantification of data (mean ± SEM) from three independent experiments is normalized with respect to the control condition (i.e., pcDNA transfection; treated as 100%) and analyzed using two‐way ANOVA with Bonferroni multiple comparisons test (*P < 0.05; **P < 0.01).

The carboxyl‐terminus sequence of V₂R contains two clusters of potential phosphorylation sites, which are T³⁵⁹/T³⁶⁰ and S³⁶²/S³⁶³/S³⁶⁴, and the B₂R lacks analogous clusters in its carboxyl‐terminus (Fig 2A). A closer look at the potential phosphorylation sites in B₂R mutants, where βarr1 knockdown augments ERK1/2 activation (i.e., B₂R^L370T and B₂R^L370T/ΔG368), suggests that these mutations construct a double‐Threonine cluster (i.e., Thr³⁶⁹ and Thr³⁷⁰), although they do not alter the PXXPXXP pattern. The B₂R^ΔG368 mutant, which mimics the spatial distribution of Ser³⁶⁶ and Thr³⁶⁹ in B₂R with that of Ser³⁵⁷ and Thr³⁵⁹ in V₂R, and the B₂R^ΔI374 mutant, which constructs a triple phospho‐site cluster in B₂R (Thr³⁷²/Ser³⁷³/Ser³⁷⁵), similar to that in V₂R (Ser³⁶²/Ser³⁶³/Ser³⁶⁴), retain the inhibitory contribution of βarr1 in ERK1/2 phosphorylation. Taken together, these data indicate that the presence of Leu³⁷⁰ in B₂R plays a critical role in the inhibitory contribution of βarr1 in ERK1/2 activation, and spatial positioning of a threonine in its place, which constructs a double‐Thr cluster, reverses the contribution of βarr1.

Intrabody sensor reveals a conformational mechanism for the contribution of βarr1

In order to gain additional mechanistic insight into differential contribution of βarr1 in ERK1/2 activation for V₂R and B₂R, we used a previously described intrabody‐based sensor of βarr1 conformation in cellular context (Ghosh et al, 2019; Baidya et al, 2020). This sensor is based on a synthetic antibody fragment referred to as Fab30 that selectively recognizes V₂R‐bound βarr1 conformation in vitro (Shukla et al, 2013, 2014; Kumari et al, 2016, 2017; Ghosh et al, 2019). Moreover, an intrabody version of Fab30, referred to as Ib30, also recognizes βarr1 upon agonist stimulation of V₂R in cellular context (Ghosh et al, 2019). Recently, a YFP‐tagged version of Ib30 has also been generated and used as a conformational probe for measuring the interaction and trafficking of βarr1 in cells upon GPCR stimulation (Baidya et al, 2020). As previously reported, we observed that Ib30‐YFP efficiently recognizes V₂R‐bound βarr1 and follows the trafficking pattern of βarr1 upon agonist stimulation (Figs 4A and EV3), but fails to recognize βarr1 upon stimulation of B₂R (Figs 4B and EV3). Inferring a potential conformational difference in βarr1 upon interaction with V₂R vs. B₂R, which may be correlated to its opposite contribution in ERK1/2 activation, we measured the ability of Ib30‐YFP to recognize βarr1 in the context of B₂R mutants. Strikingly, we found robust reactivity of Ib30‐YFP to βarr1 in the case of B₂R^L370T and B₂R^L370T/ΔG368 mutants (Fig 4C and D), and it displayed a trafficking profile very similar to that observed of V₂R. On the other hand, we observed only a weak reactivity of Ib30‐YFP for the other two B₂R mutants (B₂R^ΔG368 and B₂R^ΔI374) where βarr1 plays an inhibitory role in ERK1/2 activation, similar to B₂R^WT (Fig EV4).

Figure 4. An intrabody sensor suggests distinct conformations of βarr1 for V₂R and B₂R.

(A–D) HEK‐293 cells expressing V₂R, B₂R, B₂R^ΔG368/L370T, or B₂ R^{L 370T} along with βarr1‐mCherry and YFP‐tagged intrabody30 (Ib30‐YFP) were stimulated with their respective agonists (AVP, 100 nM for V₂R) and (bradykinin, 100 nM for B₂R) and the localization of βarr1 and Ib30 were visualized. Although βarr1 gets recruited for all four constructs, Ib30‐YFP does not colocalize with βarr1 either at the surface or upon internalization in case of B₂ R^WT. Representative data from 3–5 independent experiments are shown, and the scale bar is 10 μm. Visual scoring from 3–5 independent experiments revealed agonist‐induced Ib30 recruitment (i.e., membrane and endosomal localization) in approximately 82% of the cells for V₂R (691 cells), 10% for B₂R^WT (254 cells), 63% for B₂R^L370T (273 cells), and 83% for B₂R^ΔG368/L370T (158 cells). Pearson's correlation coefficients (PCC) were measured to assess the colocalization of βarr1 and Ib30 using JACoP plugin in ImageJ, and the mean ± sem for the basal, surface, and internalized panels, respectively, are presented here. B₂ R^WT — 0.32 ± 0.02 from 16 cells, 0.18 ± 0.01 from 32 cells, and 0.25 ± 0.02 from 31 cells; B₂R^ΔG368/L370T — 0.32 ± 0.03 from 13 cells, 0.92 ± 0.01 from 18 cells, and 0.90 ± 0.01 from 16 cells; B₂ R^{L 370T} — 0.26 ± 0.01 from 10 cells, 0.85 ± 0.01 from 11 cells, and 0.87 ± 0.01 from 25 cells.

Figure EV3. Recognition of βarr1 by intrabody30 sensor for V₂R vs. B₂R.

Confocal microscopy of HEK‐293 cells expressing βarr1‐mCherry and Ib30‐YFP together with the indicated receptors reveals robust recognition of βarr1 by Ib30 upon agonist stimulation of V₂R but not for the B₂R. Line‐scan analysis further corroborates these observations. Representative images from three independent experiments are shown here, and the scale bar is 10 μm.

Figure EV4. Intrabody30 sensor only weakly recognizes βarr1 for B₂R^ΔG368 and B₂R^ΔI374 .

HEK‐293 cells expressing the indicated receptor mutants, βarr1‐mCherry and Ib30‐YFP were stimulated with bradykinin (100 nM) and the localization of βarr1 and Ib30 were visualized using confocal microscopy. Representative images from three independent experiments are shown here, and the scale bar is 10 μm. Visual scoring from 3–5 independent experiments revealed agonist‐induced Ib30 recruitment (i.e., membrane and endosomal localization) in approximately 4% of the cells for B₂R^ΔG368 (202 cells) and 12% for B₂R^ΔI374 (233 cells). Pearson's correlation coefficients (PCC) were measured to assess the colocalization of βarr1 and Ib30 using JACoP plugin in ImageJ, and the mean ± sem for the basal, surface, and internalized panels, respectively, are presented here. B₂R^ΔG368 — 0.30 ± 0.03 from 14 cells, 0.32 ± 0.03 from 18 cells, and 0.31 ± 0.03 from 11 cells; B₂R^ΔI374 — 0.28 ± 0.04 from 10 cells, 0.42 ± 0.02 from 31 cells, and 0.43 ± 0.04 from 13 cells.

To further confirm these findings and measure the reactivity of Ib30 sensor more quantitatively, we designed a NanoBiT construct of Ib30 with N‐terminal fusion of LgBiT. We then measured agonist‐induced luminescence signal in HEK‐293 cells expressing various B₂R constructs together with SmBiT‐βarr1 and LgBiT‐Ib30. We observed a pattern that is reminiscent of confocal microscopy data, i.e., Ib30 robustly recognizes V₂R^WT but does not recognize βarr1 for B₂R^WT (Fig 5A). Moreover, we also observed that Ib30 NanoBiT sensor does not recognize βarr1 for B₂R^ΔG368 and only weakly recognizes βarr1 for B₂R^ΔI374 (Fig 5A). However, Ib30‐NanoBiT sensor exhibits significant interaction with βarr1 for B₂R^L370T and B₂R^ΔG368/L370T (Fig 5A), which is in excellent agreement with the pattern observed by confocal microscopy. Taken together, these findings suggest that spatial signature of receptor phosphorylation sites play a key role in ensuing βarr1 conformation recognizable by Ib30 and, thereby, potentially link βarr1 conformation with its contribution in ERK1/2 activation.

Figure 5. βarr1 conformation and G‐protein coupling for B₂R mutants.

• A
  HEK‐293 cells expressing the indicated receptor constructs (pcDNA vector control for mock‐transfection) together with SmBiT‐βarr1 and LgBiT‐Ib30 were stimulated with the indicated concentrations of respective agonists for 10 min. Subsequently, the luminescence signal generated upon the interaction of SmBiT‐βarr1 and LgBiT‐Ib30 was measured, normalized with basal signal (i.e., no‐agonist stimulation treated as 1), and plotted using GraphPad Prism. pEC₅₀ (top) and E_max (bottom) values are mentioned in the respective graphs, and data represent mean ± SEM of four independent experiments, each carried out in duplicate.
• B
  HEK‐293 cells expressing the NanoBiT‐Gq protein and indicated B₂R constructs were treated different concentrations of bradykinin, and ligand‐induced change in luminescent signal was measured. Concentration‐response curves were plotted using GraphPad Prism to calculate pEC₅₀ (top) and E_max (bottom) values included in the graph. Data represent mean ± SEM of three independent experiments with each performed in duplicate and normalized with respect to the luminescence signal under basal (no‐agonist treatment) condition (treated as 1).

In order to confirm that the mutation of the receptor phosphorylation sites do not alter their G‐protein coupling, we also measured agonist‐induced activation of Gq heterotrimer for B₂R mutants using a previously described NanoBiT‐G‐protein dissociation assay (Inoue et al, 2019). As presented in Fig 5B, we observe that B₂R mutants exhibit G‐protein activation profile similar to the wild‐type receptor.

Structural insights into receptor‐specific βarr1 conformations

In order to gain structural insights into these findings, we employed a molecular dynamics (MD) simulation approach using the previously determined crystal structure of βarr1 in the presence of V₂R phosphopeptide (V₂Rpp; Shukla et al, 2013). We carried out MD simulations with the phosphopeptide sequences derived from the V₂R, B₂R, and B₂R^L370T. All three peptides adopt an overall similar binding mode (Fig 6, middle panel), in which the phosphosensing pockets A, B, and C (Zhou et al, 2017) in βarr1 can accommodate the corresponding phospho‐site patterns (Fig EV5A). The B₂R peptides display a loop bulging between S³⁶⁶ and M³⁶³ compared to the V₂R peptide, due to the presence of G³⁶⁸ (Fig EV5B). Furthermore, we observe that V₂R and B₂R^L370T peptides use two phosphorylated threonines, i.e., T³⁵⁹/T³⁶⁰ and T³⁶⁹/T³⁷⁰ to form a bifurcated electrostatic interaction with K²⁹⁴ in the lariat loop of βarr1 (Fig 6, upper panel). In contrast, the B₂R peptide is able to establish only a single interaction with K²⁹⁴ via the phosphorylated T³⁶⁹, and the second interaction is not possible due to the presence of non‐polar residue, i.e., L³⁷⁰ (Fig 6, upper panel). This is further corroborated by measuring the stability of these two salt bridges for different peptides in complex with βarr1 (Fig 6, lower panel).

Figure 6. Structural insights into phospho‐site interaction and βarr1 conformation.

Structural model of βarr1 in complex with phosphopeptides corresponding to the V₂R, B₂R, and B₂R^L370T were generated based on previously determined structure of V₂Rpp‐βarr1 complex (middle panel). A bifurcated salt bridge links the V₂R and B₂R^L370T peptides to K²⁹⁴ in the lariat loop of βarr1 via T³⁵⁹/T³⁶⁰ and T³⁶⁹/T³⁷⁰, respectively, but not in B₂ R^WT peptide (upper panel). The frequency of salt bridge formation between the phospho‐sites on the receptor and K²⁹⁴ in βarr1 was computed over 4 μs accumulated simulation time per complex using a distance threshold of 3.2 Å between the oxygen of the phosphate group of phosphorylated threonines and the protonated nitrogen of K²⁹⁴. B₂R^L370T mutant establishes a bifurcated interaction with K²⁹⁴ in the lariat loop, analogous to that in V₂R peptide while B₂R peptide forms only a single salt bridge (lower panel). Differential lariat loop stabilization correlates with distinct conformational states. Simulations of 4 μs per complex were clustered with a rmsd cut‐off of 2.2 Å yielding “Cluster 1 and “Cluster 2 while “None reflects conformations that did not fulfill this condition. The phosphopeptides corresponding to V₂R and B₂R^L370T with a conserved bifurcated salt bridge adopt preferential conformational states belonging to “Cluster 1 the B₂R^WT peptide with a single salt bridge favors the conformations of “Cluster 2.

Figure EV5. Structural templates of βarr1 in complex with V₂R^WT, B₂R^WT, and B₂R^L370T peptides.

• A
  The binding mode of indicated peptides to βarr1 are presented here, and the positively charged residues (i.e. Lys and Arg) are shown as blue surface, which were proposed by Zhou et al, to constitute three phosphate‐binding pockets. Both phospho‐site patterns, i.e., the PXPXXP and PXXPXXP can be accommodated in these binding pockets but only the latter type is shown here for clarity. Although Leu³⁷⁰ in B₂R^WT does not fit in pocket B, Thr³⁶⁹, not shown here, can be accommodated in this pocket.
• B
  The presence of G³⁶⁸ in B₂R peptide sequence results in a loop bulging between residues M³⁶³ and S³⁶⁶ compared to the V₂R peptide.

We also observe that the presence or absence of a bifurcated connection between the peptides and K²⁹⁴ impact the conformational states of the lariat loop (Fig 6, lower panel). When present, the lariat loop preferentially samples conformations belonging to “Cluster 1” as seen for the V₂R and B₂R^L370T peptides; however, in its absence, the lariat loop conformations are shifted downwards favoring “Cluster2” as seen in the B₂R^WT peptide (Fig 6, lower panel). Overall, the “Cluster1” resembles the conformation of the lariat loop observed in the V₂Rpp‐βarr1 crystal structure with an average rmsd of 1.9 Å; however, in “Cluster2”, the average rmsd increases to 4.6 Å compared to the crystal structure (Fig 6, lower panel). These observations provide a structural framework to suggest that the relative orientation of the lariat loop in βarr1 is directed by the spatial pattern of phosphorylation sites present in GPCRs. While corroborating experimental evidence is still needed, it is tempting to speculate that the relative orientation of the lariat loop imparts conformational differences in βarr1, which are potentially linked to distinct functional outcomes. It would be interesting to measure a direct contribution of this salt bridge and the relative orientation of the lariat loop in the recruitment of βarr1, its conformational signature, and effect on ERK1/2 phosphorylation in future studies.

Engineering double‐serine cluster reverses the contribution of βarr1 in ERK1/2 activation for AT_1aR

In order to further extend our findings, we next investigated the human angiotensin II type 1a receptor (AT_1aR). AT_1aR consists of two phospho‐site patterns, which can be categorized as PXPXXP and PXXPXXP, similar to that present in the V₂R (Fig 7A). Moreover, it also harbors a double‐phospho‐site cluster i.e. Ser³³⁵/Thr³³⁶, analogous to Thr³⁵⁹/Thr³⁶⁰ in V₂R. Still however, similar to B₂R, siRNA‐mediated βarr1 depletion leads to an augmentation of agonist‐induced ERK1/2 phosphorylation (Ahn et al, 2004), which we also confirmed using shRNA mediated βarr1 knockdown (Fig 7C). Therefore, taking lead from the B₂R data, we focused on Leu³³⁷ in AT_1aR, and we generated two different mutants by either mutating Leu³³⁷ to Thr, i.e., AT_1aR^L337T, or by deleting Leu³³⁷, i.e., AT_1aR^ΔL337 (Fig 7B). These changes either disrupt the PXXPXXP pattern (i.e., in AT_1aR^ΔL337) or leave it unchanged (i.e., in AT_1aR^L337T). Surface expressions of these mutants were comparable under the control vs. βarr1 knockdown conditions (Fig EV1D). In agonist‐induced ERK1/2 phosphorylation assay, we observed that contrary to AT_1aR^WT, depletion of βarr1 resulted in a significant reduction of agonist‐induced ERK1/2 phosphorylation for the AT_1aR^L337T and AT_1aR^ΔL337 mutants (Fig 7D and E). This pattern is reminiscent of B₂R^L370T mutant, where a single point mutation (i.e., Leu³⁷⁰Thr) reverses the contribution of βarr1 in ERK1/2 phosphorylation. Considering the distribution of multiple potential phosphorylation sites in AT_1aR, it should be possible to design additional mutants and probe their ERK1/2 phosphorylation pattern to deduce further insights into βarr1‐mediated ERK1/2 activation. Nonetheless, taken together with B₂R data, our findings underscore the importance of spatially positioned key phosphorylation sites in the receptors in directing the role of βarr1 in ERK1/2 activation and also suggest that PXPXXP/PXXPXXP type patterns may not directly correlate with the contribution of βarr1 in ERK1/2 activation.

Figure 7. Contribution of βarr1 in ERK1/2 phosphorylation for AT _1aR mutants.

• A
  Phospho‐site clusters in the carboxyl‐terminus of V₂R and AT_1aR are underlined and color‐coded to reflect the PXPXXP and PXXPXXP patterns as proposed in a previous study (Zhou et al, 2017).
• B
  Two different AT_1aR mutants were generated by site‐directed mutagenesis to mimic the spatial distribution of Ser/Thr as in V₂R with respect to double‐Thr cluster. The red dotted box highlights the resulting sequence upon either the deletion of Leu³³⁷ or Leu³³⁷Thr mutation.
• C–E
  HEK‐293 cells expressing the indicated AT_1aR constructs, in the absence (CTL; control) or presence of βarr1 knockdown (βarr1‐KD) were stimulated with indicated concentrations of angiotensin II (AngII) for 10 min. Subsequently, ERK1/2 phosphorylation was measured by Western blotting and signal intensities were quantified using densitometry and presented as bar graphs. Data were normalized with respect to the signal at 100 nM agonist concentration in control cells (treated as 100%) and represent mean ± SEM of five (six for AT_1aR^ΔL337) independent experiments. Data were analyzed using two‐way ANOVA with Bonferroni multiple comparisons test (***P < 0.001, *P < 0.05).

Discussion

The interaction of βarrs is mostly conserved across the GPCR family, but there are many instances of distinct functional contribution of βarrs in trafficking and signaling for different receptors (Shenoy & Lefkowitz, 2011; Kang et al, 2014; Srivastava et al, 2015). These functional differences are manifested even for those receptors, which display qualitatively similar βarr recruitment and trafficking patterns (Ahn et al, 2004; Srivastava et al, 2015). Not only the differences in the phosphorylation patterns of different GPCRs can fine tune receptor‐specific conformations and functional capabilities of βarrs (Lee et al, 2016; Nuber et al, 2016; Baidya et al, 2020) but also, differential phosphorylation patterns of a given GPCR can impart distinct conformational changes in recruited βarrs resulting in different functional responses (Charest et al, 2005; Shukla et al, 2008; Nobles et al, 2011; Zimmerman et al, 2012). These examples suggest that different GPCR‐βarr complexes formed in cells, which may grossly look similar, do not necessarily encode identical functional outcomes. Our data presented here reveal that spatial positioning of key receptor phosphorylation sites can play a decisive role in the qualitative contribution of βarr1 in ERK1/2 activation. Our findings also underscore a conformational mechanism for this phenomenon via specific interactions between the key receptor phosphorylation sites and a conserved residue in βarr1.

An interesting point that also emerges from the current study is that βarr1 conformations supporting receptor trafficking vs. ERK1/2 phosphorylation may be different from each other. This is based on the observation that all three receptors tested here induce a similar pattern of agonist‐induced βarr1 trafficking but the contribution of βarr1 in ERK1/2 activation is strikingly different. Going forward, it would be interesting to gather high‐resolution structural information on these different receptor‐βarr1 complexes to further illuminate structural and functional diversity in this signaling system. Recently, cryo‐EM structures of three different GPCRs have been determined in complex with βarr1. These receptors include the human neurotensin receptor (NTS1R; Yin et al, 2019; Huang et al, 2020), the human muscarinic receptor (M2R; Staus et al, 2020) and the turkey β1‐adrenergic receptor (β1AR; Lee et al, 2020). While comparing these structures with previously determined rhodopsin‐visual‐arrestin crystal structure (Zhou et al, 2017), an interesting observation emerges in terms of βarr1 orientation with respect to the membrane bilayer (Chaturvedi et al, 2020). βarr1 in the NTSR1‐βarr1 complexes is rotated by about 85–90° in the plane of the membrane when compared to the rhodopsin‐visual‐arrestin structure, which is not observed in the M₂R‐βarr1 and β₁AR‐βarr1 structures (Chaturvedi et al, 2020). It would be very interesting to probe in future whether distinct phosphorylation patterns may direct specific βarr orientation in GPCR‐βarr complexes, and if they have a direct impact on functional outcomes.

One of the key features driving the functional multiplicity of βarrs, including ERK1/2 activation, is their ability to interact with multiple binding partners and nucleate signalosomes (Lefkowitz & Shenoy, 2005; DeWire et al, 2007; Chen et al, 2018). For example, βarrs can scaffold c‐Raf1, MEK1, and ERK2 to promote agonist‐induced ERK1/2 phosphorylation (Lefkowitz & Shenoy, 2005; DeWire et al, 2007). Considering that engineering a double‐Thr cluster in B₂R reverses the contribution of βarr1 in agonist‐induced ERK1/2 phosphorylation, it would be tantalizing to probe in future studies how the scaffolding properties of βarr1 changes in the context of this modified pattern of receptor phosphorylation sites. Based on the differences observed in the reactivity pattern of intrabody sensor, it is tempting to speculate that distinct βarr1 conformations may encode differential scaffolding abilities for different binding partners and thereby, orchestrating the corresponding functional outcomes. However, this hypothesis remains to be experimentally explored in subsequent studies. We have focused our study on βarr1, which displays opposite contribution in ERK1/2 activation for V₂R vs. B₂R and AT_1aR. βarr2 on the other hand has qualitatively similar contribution for all three receptors, i.e., its knockdown results to a decrease in ERK1/2 activation. This reciprocal regulation of βarr isoforms has been investigated and discussed earlier [14]. In this previous study, authors have systematically evaluated the underlying mechanism for potentiation of ERK1/2 response upon βarr1 knockdown and conclude that it is primarily due to an augmentation of βarr2‐dependent activation of ERK1/2. They also conclude that endogenous βarr1 present at relatively higher levels than βarr2 in HEK‐293 cells essentially exerts a dominant‐negative effect on βarr2 mediated ERK1/2 phosphorylation. Although both βarr1 and 2 are recruited to the receptor, βarr2 is potentially a better scaffold for the components of ERK1/2 cascade and, therefore, βarr1 depletion results in enhanced ERK1/2 phosphorylation.

Although engineering specific phosphorylation sites in the receptors studied here has functional effects, the phosphorylation status of these sites in cellular and physiological context remains to be experimentally established. Moreover, as different sites in GPCRs can be phosphorylated by different GRKs and other kinases, it would be interesting to explore such a possibility for the receptors used here, and probe if that has a role in differential contribution of βarr1 in ERK1/2 activation. It is also worth noting that although the orientation of lariat loop and the propensity of salt bridge formation with Lys²⁹⁴ appear to differ between V₂R and B₂R, future studies are required to correlate it directly with the inhibitory or promotive nature of βarr1 in ERK1/2 phosphorylation. The crystal structure of V₂Rpp‐βarr1 complex suggests that the interaction of the phosphate groups in the double‐Thr cluster (T³⁵⁹/T³⁶⁰) of V₂Rpp with Lys²⁹⁴ is involved in the disruption of the polar core in βarr1, which is an important step in βarr activation. Thus, it is tempting to speculate that different degrees of polar core disruption in βarr1, for example, between V₂R and B₂R, may link the ensuing βarr1 conformation with corresponding ERK1/2 activation. Finally, it is worth mentioning that some recent studies have proposed that the contribution of βarrs in ERK1/2 phosphorylation may not be completely independent of G proteins, and instead an interplay of G proteins and βarrs may be involved (O'Hayre et al, 2017; Grundmann et al, 2018; Gurevich & Gurevich, 2018a; Gutkind & Kostenis, 2018; Luttrell et al, 2018). Although we are not probing this aspect in the current study, our finding provide an additional framework and offer phospho‐site mutants of these receptors as additional tools to delineate the interplay of G proteins and βarrs in agonist‐induced ERK1/2 activation.

In summary, we discover that spatial positioning of the key phosphorylation sites in selected GPCRs triggers the role reversal of βarr1 in agonist‐induced ERK1/2 activation. Our study also identifies the critical involvement of a bifurcated salt bridge interaction between a double‐phospho‐site cluster on the receptor and a conserved lysine residue in βarr1, which directs the orientation of the lariat loop in βarr1. These findings provide a previously missing mechanistic link for the opposite contribution of βarr1 in ERK1/2 activation for different GPCRs, and thereby, improve the current framework of βarr‐dependent GPCR signaling and biased agonism.

Materials and Methods

General reagents, plasmid constructs, and cell culture

HEK‐293 cells (ATCC) were cultured and maintained in DMEM (Thermo Fisher) supplemented with 10% FBS and 100 U/ml penicillin and 100 μg/ml streptomycin. The cells were cultured in sterile 10‐cm plates at 37°C under 5% CO₂ and passaged at 70–80% confluency. For transfection, 50–60% confluent cells were transfected with desired DNA using PolyEthylenImine (PEI) as transfection reagent at the DNA:PEI ratio of 1:3. Expression plasmids for V₂R, B₂R, AT_1aR, Ib30‐YFP, βarr1 shRNA, and βarr1‐mCherry have been described previously (Kumari et al, 2016, 2017; Ghosh et al, 2017, 2019; Pandey et al, 2019a; Baidya et al, 2020). Receptor mutants were generated using site‐directed mutagenesis kit (NEB) and sequenced (Macrogen) before use in the experiments. Commercial antibodies used in various assays are described in the corresponding sections below.

Surface expression of receptor constructs

Surface expression of various GPCRs used in this study was measured using whole cell surface ELISA assay as described previously (Pandey et al, 2019b). In brief, after 24 h of transfection, cells were seeded onto poly‐D‐lysine–coated 24‐well plates at a density of 100,000 cells/well and allowed to adhere for an additional 24 h. Subsequently, the cells were serum starved for 4–6 h and then fixed with 4% formaldehyde for 20 min. Cells were washed three times with 1× TBS, non‐specific sites were blocked for 1 h with 1%BSA in 1× TBS and incubated with HRP‐coupled anti‐FLAG M2 antibody (Sigma, cat. no. A8592, 1:5,000 dilution). After three washes, TMB substrate was added to each well, and once the color was developed, the reaction was stopped by adding 100 μl 1 M H₂SO₄. The absorbance was measured at 450 nm using a multi‐plate reader (Victor X4, Perkin Elmer). Cell density per well was determined by staining with 0.2% Janus Green and measuring absorbance at 595 nm. Normalized surface expression was calculated by the ratio of A₄₅₀ /A_595.

Confocal microscopy

In order to visualize agonist‐induced trafficking of βarr1‐mCherry, and the reactivity of Ib30‐YFP with βarr1 for different receptor constructs, we used confocal microscopy as described earlier (Baidya et al, 2020). Briefly, HEK‐293 cells were co‐transfected with either, the receptor and βarr1‐mCherry, or receptor, βarr1‐mCherry and Ib30‐YFP, and 24 h post‐transfection, they were seeded onto poly‐D‐lysine–coated confocal dishes. After another 24 h, cells were serum starved for 4–6 h and then stimulated with respective agonists at indicated concentrations. Live cell imaging was performed using Zeiss LSM 710 NLO confocal microscope attached to a 32× array GaAsP descanned detector (Zeiss). A Multiline Argon laser for YFP (488 nm) and a Diode Pump Solid State Laser for mCherry (561 nm) were used. Intensity of laser and pinhole settings were maintained in the same range for any parallel set of experiments and filter excitation regions and bandwidths were adjusted to avoid any spectral overlap between the two channels. Images were finally processed in ZEN lite (ZEN‐blue/ZEN‐black) software suite from ZEISS. Agonist‐induced trafficking of βarr1‐mCherry was quantified across multiple cells from independent experiments by manually counting surface (membrane) and endosomal (internalized) localization patterns. Colocalization of βarr1‐mCherry and Ib30‐YFP were measured using either manually, by a line‐scan analysis in ImageJ to measure fluorescence intensities across a drawn line, or by using the JACoP (Just Another Colocalization Plugin) tool in ImageJ (Bolte & Cordelieres, 2006) to calculate Pearson's correlation coefficient (PCC).

NanoBiT‐G‐protein dissociation assay

Ligand‐induced Gq dissociation was measured by a NanoBiT‐G‐protein dissociation assay (Inoue et al, 2019), in which interaction between a Gα subunit and a Gβγ subunit was monitored by a NanoLuc‐based enzyme complementation system called NanoBiT (Promega). Specifically, a NanoBiT‐Gq protein consisting of Gαq subunit fused with a large fragment (LgBiT) in the alpha helical domain and an N‐terminally small fragment (SmBiT)‐fused Gβ1 was expressed along with untagged Gγ2 subunit with a C68S mutation, RIC8A, and a test B₂R construct. HEK‐293A cells (Thermo Fisher Scientific) were seeded in a 6‐well culture plate at a concentration of 2 × 10⁵ cells/ml (2 ml per well in DMEM (Nissui) supplemented with 10% fetal bovine serum (Gibco), glutamine, penicillin, and streptomycin) 1‐day before transfection. Transfection solution was prepared by combining 5 μl (per well in a 6‐well plate hereafter) of polyethylenimine Max solution (Polysciences; 1 mg/ml), 200 μl of Opti‐MEM (Thermo Fisher Scientific) and a plasmid mixture consisting of 200 ng B₂R construct, 100 ng LgBiT‐containing Gαq subunit, 500 ng SmBiT‐Gβ1, 500 ng Gγ2 (C68S), and 100 ng RIC8A. After incubation for 1 day, the transfected cells were harvested with 0.5 mM EDTA‐containing Dulbecco's 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 and loaded with 20 μl of 50 μM coelenterazine (Carbosynth) diluted in the assay buffer. After 2‐h incubation at room temperature, the plate was measured for baseline luminescence (Spectramax L, Molecular Devices) and a titrated Bradykinin (20 μl; 6× of final concentrations) were manually added. The plate was immediately read at room temperature for the following 10 min as a kinetics mode at a measurement interval of 20 s. The luminescence counts over 5–10 min after ligand addition were averaged and normalized to the initial count. The fold‐change values were further normalized to that of vehicle‐treated samples and were used to plot G‐protein dissociation response. Using Prism 8 software (GraphPad Prism), the G‐protein dissociation signals were fitted to a four‐parameter sigmoidal concentration–response curve, from which pEC₅₀ values (negative logarithmic values of EC₅₀ values) and E_max values were used to calculate mean and SEM.

NanoBiT‐β‐arrestin recruitment assay

Ligand‐induced β‐arrestin recruitment to GPCR was measured by a NanoBiT‐β‐arrestin recruitment assay (Dixon et al, 2016; Shihoya et al, 2018), which assesses enzyme complementation of a C‐terminally SmBiT‐fused B₂R construct (B₂R‐Sm) with N‐terminally LgBiT‐fused β‐arrestin (Lg‐βarr). FLAG‐B₂R was C‐terminally fused to SmBiT with a 15‐amino acid flexible linker (GGSGGGGSGGSSSGG) and inserted into a pCAGGS mammalian expression plasmid. Transfection in HEK293A cells was performed as described in the NanoBiT‐G‐protein dissociation assay by using a plasmid mixture consisting of a test B₂R‐Sm construct (500 ng) and Lg‐βarr1 (100 ng). Next day, the transfected cells were subjected to the same assay procedure as described above and increase in luminescent signal was measured.

NanoBiT‐Ib30 assay

Ligand‐induced β‐arrestin conformational change was measured by a NanoBiT‐Ib30 assay, which was developed in this work. Ib30 was N‐terminally fused to LgBiT (codon‐optimized and gene‐synthesized by Genscript) with the 15‐amino acid flexible linker and inserted into the pCAGGS plasmid. Similarly, human β‐arrestin1 was N‐terminally fused to SmBiT with the 15‐amino acid flexible linker and inserted into the pCAGGS plasmid. Transfection in HEK293A cells was performed as described in the NanoBiT‐G‐protein dissociation assay by using a plasmid mixture consisting of a test B₂R construct (200 ng), Sm‐βarr1 (100 ng), and Lg‐Ib30 (500 ng). For a control experiment, a V₂R‐encoding plasmid (200 ng) or an empty plasmid was used in place of a B₂R plasmid. Next day, the transfected cells were subjected to the same assay procedure as described above and increase in luminescent signal was measured.

ERK1/2 phosphorylation assay

For measuring agonist‐induced ERK1/2 phosphorylation, we used HEK‐293 cells with βarr1 knockdown using either shRNA (Ghosh et al, 2019) or CRISPR/Cas9 approach (Luttrell et al, 2018) and followed the protocol described previously (Beautrait et al, 2017). For shRNA approach (Figs 1C and D, 3 and 7B–E), about three million HEK‐293 cells stably expressing either an shRNA against βarr1 or a control shRNA were transfected at approximately 60% confluency with the indicated receptor constructs. We used 0.5, 1, and 2 μg of V₂R, B₂R, and AT_1aR constructs, respectively, and after 24 h, one million cells were seeded into each well of 6‐well plates. Forty‐eight hours post‐transfection, cells were serum starved for 4–6 h in DMEM supplemented with 20 mM HEPES and 0.01% BSA, stimulated with indicated concentrations of the corresponding agonists, and subsequently, lysed using 2× SDS loading buffer followed by boiling at 95°C for 15 min. Phosphorylated and total levels of ERK1/2 were measured by Western blotting using anti‐pERK1/2 and anti‐total ERK1/2 primary antibodies (CST, cat. no. 9101; and 9102, respectively; 1:5,000 dilution) and HRP‐coupled anti‐rabbit secondary antibody (GenScript, cat. no. A00098). Blots were developed using HRP substrate from Promega and Chemi‐Documentation system from Bio‐Rad. Densitometry‐based quantification of bands was carried out using Bio‐Rad software or ImageJ and plotted in GraphPad Prism.

For CRISPR/Cas9‐based βarr1/2‐deleted HEK293 cells (Luttrell et al, 2018) (Fig EV2), about 1.5 million cells were seeded on a 10‐cm cell culture plate a day prior to transfection to achieve ~80% confluency. 1 μg of B₂R, B₂R^ΔG368/L370T or B₂R^L370T constructs were transfected along with 0.3 μg of FLAG‐βarr2, with or without 0.3 μg of FLAG‐βarr1 (Beautrait et al, 2017). After 24 h, one million cells per well were seeded on to a 6‐well plate. Forty‐eight hours post‐transfection, cells were serum starved, stimulated with or without 100 nM bradykinin (BK) for 10 min, followed by sample preparation and Western blotting as mentioned above. Here, HRP‐coupled anti‐rabbit secondary antibody from Bio‐Rad was used (cat. no. 1706515). The expression of transfected βarr1 was detected using A1CT primary antibody (1:2,000 dilution; gifted by the Lefkowitz lab).

Molecular dynamics simulation

In order to generate the complexes of B₂Rpp^WT, B₂Rpp^L370T, and V₂Rpp^WT, we used the previously determined structure of V₂Rpp in complex with βarr1 (PDB code: 4JQI). Missing fragments in the βarr1 and V₂Rpp structures were modeled using the loop modeler module available in the MOE package (https://www.chemcomp.com). In addition, the co‐crystallized Fab30 was also removed from the structural template. For modeling the binding of the B₂R^WT and B₂R^370T peptide sequences to βarr1, we considered two different alignments as indicated below:

Here, the residues numbers are indicated based on the corresponding receptor sequences, and the site for B₂R mutation (i.e., L³⁷⁰T) is colored in red. Both models were obtained by converting the sequence of the V₂Rpp into that of the B₂Rpp and the corresponding B₂Rpp mutant. In alignment A, the studied mutation site 370 is located in the space corresponding to residue 360 in the V₂R/βarr1 complex. In this alignment, the G³⁶⁸ represents an insertion in the loop formed between the first and second β‐sheet. The additional residue results in a loop extension in B₂Rpp and B₂Rpp^L370T compared to V₂Rpp, which can be appreciated in Fig EV5. In such a configuration, the L³⁷⁰T mutation in the B₂R interacts with residue K²⁹⁴ in the lariat loop and mediates key interactions between the N‐ and C‐domain of βarr1. This structurally pivotal role could help explain the dramatic functional shift associated with a L³⁷⁰T mutation, i.e., supportive role of βarr1 in ERK1/2 phosphorylation. In alignment B, L³⁷⁰T is positioned such that it cannot interact with K²⁹⁴ but instead, interacts with K¹⁰. Gaining this additional interaction in comparison with the B₂R^WT is likely result in an increased stabilization of the peptide. However, it is less likely to yield a significant change in the structure of peptide‐βarr1 complex, as it is not located in the critical interface area of the N‐ and C‐domain of βarr1. Based on this reasoning, our experimental observation in terms of ERK1/2 phosphorylation for different B₂R mutants, we proceeded with alignment A in our simulation studies.

The complexes were solvated in TIP3P water, with the ionic strength kept at 0.15 M using NaCl ions. Simulation parameters were obtained from the Charmm36M forcefield (Huang et al, 2017). Systems generated this way were simulated using the ACEMD software package (Harvey et al, 2009). To allow rearrangement of waters and side chains, we carried out a 25 ns equilibration phase in NPT conditions with restraints applied to backbone atoms. The time step was set to 2 fs and the pressure was kept constant, using the Berendsen barostat. After the equilibration, systems were simulated in NVT conditions for 1 μs in four parallel runs employing a 4 fs time step. For all runs, temperature was kept at 300 K using the Langevin thermostat and hydrogen bonds were restrained using the RATTLE algorithm. Non‐bonded interactions were cut off at 9 Å with a smooth switching function applied at 7.5 Å. Before carrying out the structural analysis, simulation frames were aligned using the backbone atoms of the βarr1. To assess the magnitude of salt bridge formation between phosphorylated threonines (pT) residues and K²⁹⁴, we quantified frames in which the protonated nitrogen of K²⁹⁴ and oxygens of the phosphate group of each respective pT were in a distance of <3.2 Å. Conformational variability of the lariat loop was studied with the clustering tool available in VMD (Humphrey et al, 1996). As a clustering parameter, we used RMSD (cut‐off: 2.2) of the backbone atoms of residues 293 to 297 within the lariat loop.

Statistical analysis and data presentation

Quantified data were plotted and analyzed using GraphPad Prism software, and the details of experimental replicates and statistical analysis are mentioned in the corresponding figure legends. Independent experiments mentioned in the figure legends indicate biological replicates.

Conflict of interest

The authors declare that they have no conflict of interest.

Author contributions

Experiment design: AKS, AI, JS, SAL; Experiment and data analysis: AKS, AI, JS, SAL, MB, PK, HD‐A, SP, MC, TMS, KK, YC; Manuscript writing: AKS; Funding: AKS, AI, JS, SAL; Study coordination and supervision: AKS.

Supporting information

Expanded View Figures PDF

Source Data for Expanded View

Review Process File

Source Data for Figure 1

Source Data for Figure 3

Source Data for Figure 7

EMBO Reports (2020) 21: e49886

Data availability

Data included in this manuscript are available from the authors upon reasonable request. No primary datasets have been generated or deposited.