Research ArticleGPCR SIGNALING

Manifold roles of β-arrestins in GPCR signaling elucidated with siRNA and CRISPR/Cas9

See allHide authors and affiliations

Science Signaling  25 Sep 2018:
Vol. 11, Issue 549, eaat7650
DOI: 10.1126/scisignal.aat7650

The balancing act of β-arrestins

G protein–coupled receptors (GPCRs) are thought to activate the kinases ERK1/2 through G protein– and β-arrestin–dependent pathways. The relative contribution of each is difficult to assess because β-arrestins prevent G protein coupling by GPCRs (see the Focus by Gurevich and Gurevich). Studies based on CRISPR/Cas9-generated cell lines suggested that β-arrestins are dispensable for ERK1/2 activation. Luttrell et al. compared the effects of siRNA-mediated and CRISPR/Cas9-mediated knockdown of β-arrestins on ERK1/2 activation by several GPCRs in independent clones. Their data showed that signaling rewiring in the CRISPR clones rendered GPCR-dependent ERK1/2 activation more G protein–dependent, which was reversed by reconstitution with β-arrestins. Together, these findings suggest that β-arrestins balance signaling through the different pathways from GPCRs to ERK1/2 and suggest that experiments with deletion of β-arrestins or G proteins should be interpreted with caution.

Abstract

G protein–coupled receptors (GPCRs) use diverse mechanisms to regulate the mitogen-activated protein kinases ERK1/2. β-Arrestins (βArr1/2) are ubiquitous inhibitors of G protein signaling, promoting GPCR desensitization and internalization and serving as scaffolds for ERK1/2 activation. Studies using CRISPR/Cas9 to delete βArr1/2 and G proteins have cast doubt on the role of β-arrestins in activating specific pools of ERK1/2. We compared the effects of siRNA-mediated knockdown of βArr1/2 and reconstitution with βArr1/2 in three different parental and CRISPR-derived βArr1/2 knockout HEK293 cell pairs to assess the effect of βArr1/2 deletion on ERK1/2 activation by four Gs-coupled GPCRs. In all parental lines with all receptors, ERK1/2 stimulation was reduced by siRNAs specific for βArr2 or βArr1/2. In contrast, variable effects were observed with CRISPR-derived cell lines both between different lines and with activation of different receptors. For β2 adrenergic receptors (β2ARs) and β1ARs, βArr1/2 deletion increased, decreased, or had no effect on isoproterenol-stimulated ERK1/2 activation in different CRISPR clones. ERK1/2 activation by the vasopressin V2 and follicle-stimulating hormone receptors was reduced in these cells but was enhanced by reconstitution with βArr1/2. Loss of desensitization and receptor internalization in CRISPR βArr1/2 knockout cells caused β2AR-mediated stimulation of ERK1/2 to become more dependent on G proteins, which was reversed by reintroducing βArr1/2. These data suggest that βArr1/2 function as a regulatory hub, determining the balance between mechanistically different pathways that result in activation of ERK1/2, and caution against extrapolating results obtained from βArr1/2- or G protein–deleted cells to GPCR behavior in native systems.

INTRODUCTION

The mitogen-activated protein kinases (MAPKs) extracellular signal–regulated kinase 1 (ERK1) and ERK2 (ERK1/2) play critical roles in cell cycle progression, proliferation, survival, and apoptotic signaling by controlling phosphorylation of nuclear transcription factors, for example, Elk1, as well as diverse regulatory functions mediated through the phosphorylation of cytosolic substrates (1). Given their myriad roles, they are subject to multiple convergent forms of positive and negative control. As key regulators of these processes, G protein–coupled receptors (GPCRs) use multiple pathways to regulate ERK1/2 activity. Depending on the receptor and cell type, Gs-coupled GPCRs, for example, activate ERK1/2 through cAMP (adenosine 3′,5′-monophosphate)–dependent protein kinase (PKA) or direct cAMP-mediated effects on B-Raf and the Epac family of cAMP-regulated Rap1 guanine nucleotide exchange factors (GEFs) (24) or dampen mitogen-stimulated ERK1/2 activation through PKA-mediated inhibition of the upstream kinase cRaf1 (5). Adding to the diversity, Gq/11-coupled receptors often use protein kinase C (PKC)–dependent mechanisms to directly activate cRaf1 (6, 7), whereas Gi/o-coupled receptors use Gβγ subunit–mediated mechanisms to trigger Ras-dependent ERK1/2 activation by stimulating the matrix metalloproteinase–dependent shedding of tethered ligands that “transactivate” epidermal growth factor receptors (EGFRs) (810). Further complicating matters, in some cases, multiple G protein species are involved downstream of the same receptor. For example, β2 adrenergic receptor (β2AR)–mediated ERK1/2 activation can involve “switching” of receptor coupling from Gs to Gi proteins, producing ERK1/2 signaling that is sensitive both to inhibitors of Gs-PKA and to Bordetella pertussis toxin (PTX) (11, 12).

The arrestins are a family of four cytosolic proteins that play a key role in the negative regulation of heterotrimeric G protein signaling pathways by binding to agonist-occupied GPCR kinase (GRK)–phosphorylated receptors and sterically inhibiting G protein coupling at the plasma membrane (13, 14). The two nonvisual arrestins, β-arrestin1 (βArr1) and βArr2, also referred to as arrestin2 and arrestin3, respectively, further control G protein signaling by functioning as adaptor proteins that link activated GPCRs on the plasma membrane to the clathrin-dependent endocytic machinery (1518). These combined actions keep downstream G protein signaling, including G protein–dependent activation of ERK1/2, in check by promoting desensitization and internalization of activated receptors. Arrestins also bind to numerous signaling proteins, notably the three component kinases of the ERK1/2 cascade: cRaf1, MEK1/2 (MAPK kinase 1/2), and ERK1/2 (1921). This binding is sensitive to arrestin conformation such that cRaf1 and ERK1/2 bind efficiently to arrestins in their GPCR-bound conformation yet have almost no affinity for the inactive cytosolic conformation (22, 23). As a result, recruitment of βArr1/2 to activated GPCRs promotes assembly of the cRaf1-MEK1/2-ERK1/2 complex, supporting activation of ERK1/2 and its retention in cytosolic GPCR-arrestin “signalsome” complexes (19, 24). In their role as signaling scaffolds, βArr1/2 affect the kinetics of GPCR-stimulated ERK1/2 activation, favoring prolonged activation, because βArr1/2-bound ERK1/2 is protected from rapid dephosphorylation by nuclear and cytosolic MAPK phosphatases (2528), and determining its function, because the spatial constraints imposed on arrestin-bound ERK1/2 favor phosphorylation of cytosolic substrates but inhibit its nuclear functions. As a result, β-arrestin–bound ERK1/2 is implicated in the regulation of GPCR internalization and trafficking (29, 30), cytoskeletal rearrangement and chemotaxis (31, 32), matrix metalloproteinase–dependent ectodomain shedding (3335), and protein synthesis (36, 37) but dampens Elk1-dependent transcription (38, 39).

Given their functional duality, the absence of βArr1/2 would be expected to impair GPCR desensitization, potentially enhancing G protein–dependent ERK1/2 activation, while at the same time abrogating arrestin-supported signaling. Not surprisingly, germline deletion of ARRB1/2 (the genes that encode βArr1/2) produces complex effects on GPCR regulation of ERK1/2. In murine embryonic fibroblasts (MEFs) derived from βArr1/2 knockout (KO) embryos, stimulation of predominantly Gi/o-coupled lysophosphatidic acid receptors provokes robust and sustained ERK1/2 activation, which is mediated almost exclusively through transactivated EGFRs (40). When βArr2 is restored to the KO cells, the duration of EGFR-dependent ERK1/2 activation is shortened and EGFR-dependent transcription is attenuated, consistent with the dampening of a G protein–dependent pathway by βArr2-mediated desensitization. However, reintroducing βArr2 restores the wild-type phenotype of sustained EGFR-independent ERK1/2 activation, consistent with a βArr2-dependent signaling process that determines the kinetics of ERK1/2 activation and modifies lysophosphatidic acid receptor–driven transcription.

Consistent with previously published work in βArr1/2 KO MEFs, studies that used CRISPR/Cas9 (clustered regularly interspaced short palindromic repeats/CRISPR-associated protein 9) and TALEN (transcription activator-like effector nuclease) genome-editing strategies to delete ARRB1/2 or heterotrimeric Gα subunit proteins in human embryonic kidney (HEK) 293 cells demonstrated that βArr1/2 are “dispensable” for activation of ERK1/2 by β2ARs, that is, that GPCR-dependent stimulation of ERK1/2 can be mediated by G proteins in the absence of β-arrestins. Grundmann et al. (41), using a HEK293 cell model system from which essentially all heterotrimeric G protein activity had been removed by CRISPR/Cas9 gene editing and PTX treatment, reported that several arrestin pathway–selective GPCR agonists fail to activate ERK1/2 in the absence of G proteins. O’Hayre et al. (42) reported that CRISPR/Cas9-mediated deletion of βArr1/2 enhances, rather than diminishes, ERK1/2 activation by the Gs-coupled β2AR and vasopressin V2 receptors (V2Rs) and that β2AR-dependent stimulation of ERK1/2 in a βArr1/2 KO background proceeds through a pathway involving heterotrimeric G proteins and the classical receptor tyrosine kinase mitogenic pathway components, c-Src, SHC, mSOS, and Ras (43). On the basis of the observations that β-arrestin–dependent signaling does not exist in βArr1/2 null or “G-zero” HEK293 cells, the authors of these papers called into question more than 15 years of published work on the role of βArr1/2 in ERK1/2 regulation (4448), suggesting instead that GPCR regulation of ERK1/2 is a G protein–mediated process in which βArr1/2 play a negligible role. The authors further suggest that, because the only ascertainable role of βArr1/2 is to hinder G protein signaling through desensitization and internalization of GPCRs, screening of “biased” GPCR agonists is most appropriately performed in genome-edited ARRB1/2 KO cells, in which the important G protein–mediated signaling is amplified (41, 49).

It is concerning that the conclusions of these first published studies applying CRISPR/Cas9 KO cells to the study of GPCR-mediated ERK1/2 activation contrast so starkly with a large and diverse previous literature using various other validated technologies and physiologically relevant models. Accordingly, we determined the contributions of βArr1/2 to ERK1/2 activation by different Gs-coupled GPCRs in three CRISPR/Cas9 genome–edited HEK293 cell lines derived independently in different laboratories. Each receptor was probed using complementary loss-of-function and gain-of-function approaches wherein endogenous βArr1/2 in parental HEK293 lines were abruptly knocked down using small interfering RNAs (siRNAs), or in which βArr1/2 were reintroduced into CRISPR βArr1/2 KO lines by transient transfection. Together, our data provide insight into the complex roles of arrestins in regulating signal strength in two pathways, one G protein–dependent and the other βArr1/2-dependent, that converge downstream on a common effector.

RESULTS

βArr1/2 gene editing has variable effects on β2AR-stimulated activation of ERK1/2 in three independently derived βArr1/2 CRISPR KO HEK293 cell lines

The ERK1/2 cascade is subject to control by both G protein–dependent and β-arrestin–dependent mechanisms that contribute to pathway activation (47, 48). Because β-arrestins both negatively regulate G protein–dependent signaling to ERK1/2 by promoting GPCR desensitization and support pathway activation by functioning as ligand-regulated scaffolds, deletion of both βArr1/2 isoforms would be expected to augment G protein–dependent ERK1/2 activation while simultaneously abrogating the β-arrestin–dependent pathway. The net effect, either augmentation or reduction of GPCR-mediated ERK1/2 activation, would likely be dependent on the relative contribution of each pathway in a given cell background. For example, data from a study showing enhanced β2AR-stimulated ERK1/2 activation in HEK293 cell lines in which βArr1/2 were deleted using CRISPR/Cas9 or combined βArr1 TALEN KO and βArr2 siRNA approaches suggest that, in those cells, the augmentation of Gs signaling due to loss of βArr1/2-dependent desensitization is greater than the loss of signal arising from β-arrestin–dependent ERK1/2 activation (42).

To test whether deletion of βArr1/2 by genome-editing strategies consistently produced a net enhancement of GPCR-stimulated ERK1/2 activation, we initially compared β2AR-mediated ERK1/2 responses in three independently derived HEK293 cell lines in which both the ARRB1 and ARRB2 genes were targeted using a CRISPR/Cas9 system. The three parental lines differed in the relative ratios of endogenous βArr1 and βArr2 abundance, with βArr1 being more abundant than βArr2 in two lines (AI-HEK293 and SL-HEK293), whereas βArr2 was more abundant than βArr1 in the third cell line (HAR-HEK293; Fig. 1). All three derived CRISPR βArr1/2 KO lines were devoid of detectable βArr1/2. We then compared the time course of ERK1/2 activation in each parental-CRISPR HEK293 cell pair upon stimulation of endogenously expressed β2ARs with a saturating concentration of isoproterenol (Fig. 1, A to C). The maximal isoproterenol-stimulated ratio of phosphorylated ERK1/2 (pERK1/2) to total ERK1/2 in each parental cell line was normalized to 100%, and the response in the corresponding CRISPR cell line was expressed as a percentage of the response in the parental cells. Consistent with the earlier report (42), deletion of βArr1/2 in the AI-CRISPR βArr1/2 KO clone led to a twofold increase in pERK1/2 abundance relative to that in its parental line. However, this increase was not observed in either the SL-CRISPR βArr1/2 KO or HAR-CRISPR βArr1/2 KO cell lines, in which βArr1/2 deletion had no substantial effect on β2AR-mediated ERK1/2 activation. The augmented response seen in the AI-CRISPR βArr1/2 KO line was preserved when the β2AR was overexpressed (Fig. 1D), indicating that, at both high and low β2AR abundance, deletion of βArr1/2 resulted in a net increase in ERK1/2 activation in those cells.

Fig. 1 Variable effects of CRISPR/Cas9 gene editing on β2AR-stimulated activation of ERK1/2 in three independently derived βArr1/2 CRISPR KO HEK293 cell lines.

(A to D) Top: Parental and corresponding CRISPR βArr1/2 KO HEK293 cell lines expressing endogenous β2ARs (A to C) or overexpressed (o/e) FLAG-β2AR (D) were analyzed by Western blotting (IB) with antibody against βArr1/2. β-Actin was used as a loading control. The abundance of endogenous β2AR determined by [125I](−)-iodocyanopindolol binding was 15 ± 5 fmol/mg protein in all parental and CRISPR βArr1/2 KO lines, and the abundance of FLAG-β2AR in the stably transfected AI-parental and CRISPR βArr1/2 KO lines was 2.3 ± 0.01 and 4.7 ± 0.6 pmol/mg protein, respectively. Middle: Serum-deprived cells were stimulated with 100 nM isoproterenol (Iso) for the indicated times, and cell lysates were analyzed by Western blotting sequentially for pERK1/2 (pERK) and total ERK1/2 (ERK). Western blots for each parental-CRISPR pair are representative of four (B to D) or five (A) experiments. Bottom: For each parental-CRISPR βArr1/2 KO pair, isoproterenol-stimulated pERK/ERK ratios were normalized to the maximum signal observed in the parental line, and data are expressed as a percentage of the maximum control response (% max control). Data are means ± SEM of four or five biological replicates. Statistical significance was determined by two-way analysis of variance (ANOVA) and Sidak’s multiple comparison test. *P < 0.05 compared with parental cells; #P < 0.05 compared with parental cells at the corresponding time point. ns, not significant.

Knockdown of βArr1/2 using siRNA in parental HEK293 cells and reconstitution of βArr1/2 in CRISPR βArr1/2 KO cells have reciprocal effects on β2AR-stimulated activation of ERK1/2

A limitation of gene-editing strategies such as CRISPR/Cas9 is that the derived clonal cell lines have, of necessity, been selected for their ability to survive in the absence of the deleted protein(s). In the case of βArr1/2, which are key regulators of virtually all GPCR signaling, surviving CRISPR clones must have undergone some degree of “rewiring” of GPCR signaling pathways to compensate for the absence of these physiologically relevant GPCR regulatory processes. The divergent results we obtained using independently derived CRISPR βArr1/2 KO lines (Fig. 1) suggest that comparison of CRISPR clones with their parental lines is not an optimal strategy for delineating the functions of the missing proteins due to undefined adaptations during clonal selection. To further characterize the complex roles of βArr1/2 in β2AR-mediated ERK1/2 activation, we adopted reciprocal loss-of-function and gain-of-function strategies in which we compared the effects of βArr1/2 knockdown by siRNA in the parental HEK293 lines with those of βArr1/2 reconstitution by transient transfection of the CRISPR βArr1/2 KO lines.

We first determined the effects of knocking down β-arrestins using siRNAs (Fig. 2). Because previous studies of β-arrestin–dependent ERK1/2 activation have suggested that for some GPCRs, for example, the β2AR and the angiotensin AT1A receptor (AT1AR), βArr1 and βArr2 play specialized desensitizing and signaling roles, respectively (5054), these experiments used well-validated siRNA sequences targeting either βArr2 alone or both βArr1/2 (53, 54). We found that siRNA-mediated knockdown of βArr2 or βArr1/2 produced about a 80% reduction in the abundance of the targeted isoform(s) 48 hours after transient transfection. Knockdown of βArr2 alone led to a statistically significant reduction in β2AR-stimulated ERK1/2 activation compared to that in all three HEK293 parental lines and with both endogenously expressed and transiently overexpressed β2AR (Fig. 2, A to D). Similar to previous studies (54), these data indicate that in the continued presence of βArr1, selective knockdown of βArr2 produces a net decrease in β2AR-stimulated ERK1/2 activation that is consistent with βArr2-dependent augmentation of ERK1/2 signaling. To mimic the effects of βArr1/2 deletion in the CRISPR lines, we also used siRNA targeting both β-arrestin isoforms (Fig. 2, E to H). As in the CRISPR cell lines, this would be expected to reduce both βArr2-dependent signaling and any contribution of βArr1 or βArr2 to receptor desensitization, that is, to shift β2AR-stimulated ERK1/2 activation more strongly toward G protein–dependent, and away from β-arrestin–dependent, mechanisms. Knockdown of βArr1/2 in the SL-parental and HAR-parental HEK293 lines reversed the attenuation of ERK1/2 activation that was observed when βArr2 only was targeted, mimicking our findings in the SL-CRISPR and HAR-CRISPR βArr1/2 KO lines (Fig. 2, F and G). The absence of either a net increase or decrease in ERK1/2 activation upon βArr1/2 knockdown suggests that β2AR-mediated ERK1/2 activation in the absence of βArr1/2 reflects the balance between augmented G protein–dependent signaling, which would increase the ERK1/2 signal, and reduced β-arrestin–dependent signaling, which would decrease it. More complex effects were seen using the AI-parental HEK293 line, where siRNA-mediated knockdown of βArr1/2 reduced peak ERK1/2 activation upon stimulation of endogenous, but not overexpressed, β2AR (Fig. 2, E and H). These results are directly opposite those seen in the comparison of the AI-parental and AI-CRISPR βArr1/2 KO lines (Fig. 1), where the gene-edited clone exhibited a G protein–dominant phenotype wherein β2AR-stimulated ERK1/2 activation was consistently enhanced by removal of βArr1/2.

Fig. 2 Consistent effects of siRNA-mediated knockdown of βArr1/2 on β2AR-stimulated ERK1/2 activation in three parental HEK293 cell lines.

(A to H) Top: Parental HEK293 cell lines expressing endogenous β2AR (A to C) and (E to G) or overexpressed FLAG-β2AR (D and H) and transfected with siRNAs targeting no mRNA [control (CTL)], βArr2 (A to D), or both βArr1 and βArr2 (E to H) were analyzed by Western blotting with antibody against βArr1/2. β-Actin was used as a loading control. Middle: Serum-deprived cells were stimulated with 100 nM isoproterenol for the indicated times, and cell lysates were analyzed by Western blotting sequentially for pERK1/2 (pERK) and total ERK1/2 (ERK). Western blots for each pair of CTL siRNA– and β-arrestin siRNA–treated cells are representative of three (H), four (B, C, and E to G), or five (A and D) experiments. Bottom: For each CTL siRNA–treated (open black symbols) and β-arrestin siRNA–treated (red symbols) pair, isoproterenol-stimulated pERK/ERK ratios were normalized to the maximum signal observed in the CTL siRNA–treated cells, and data are expressed as a percentage of the maximum control response (% max control). Data are means ± SEM of three to five biological replicates. Statistical significance was determined by two-way ANOVA and Sidak’s multiple comparison test. *P < 0.05 compared with parental group (control); #P < 0.05 compared with control at the corresponding time point.

We then performed the complementary experiment, in which βArr2 or βArr1/2 were reintroduced into each CRISPR βArr1/2 KO background by transient transfection. Because these cells were devoid of both βArr1 and βArr2, the expectation would be that reintroduction of either or both isoforms would restore β2AR desensitization and dampen G protein–dependent downstream signaling, while at the same time restoring the capacity for β-arrestin–dependent ERK1/2 activation. Because βArr1/2 were reintroduced by transient transfection, these experiments were performed in CRISPR βArr1/2 KO lines cotransfected to express the β2AR. The abundance of βArr1/2 in the transfected CRISPR βArr1/2 KO lines was comparable with that of endogenous βArr1/2 in the respective parental HEK293 lines (Fig. 3, A to C). Not surprisingly, given its complex functions, restoration of βArr2 alone had cell line–specific effects. In the AI-CRISPR βArr1/2 KO line (Fig. 3A), which showed enhanced maximal β2AR–stimulated ERK1/2 activation compared to that of its parental line (Fig. 1), there was no net reduction in ERK1/2 activation, suggesting that any decrease in G protein–dependent signaling that resulted from restored β2AR desensitization was offset by some increase in ERK1/2 activation in the presence of βArr2. Similar results were observed in the HAR-CRISPR βArr1/2 KO line, where reintroduction of βArr2 led to a statistically insignificant reduction in maximal β2AR–stimulated ERK1/2 activation (Fig. 3C). The opposite effect was seen in the SL-CRISPR βArr1/2 KO line, where reintroduction of βArr2 statistically significantly increased the maximal β2AR–stimulated ERK1/2 signal (Fig. 3B). Identical results were obtained when both βArr1/2 were reintroduced into the βArr1/2 KO background to mimic the phenotype of the parental HEK293 cell lines (Fig. 3, D to F). No statistically significant net effect on β2AR-stimulated ERK1/2 activation was seen in the AI-CRISPR βArr1/2 KO line (Fig. 3D), whereas the SL-CRISPR and HAR-CRISPR βArr1/2 KO lines (Fig. 3, E and F) exhibited the enhanced ERK1/2 signal seen when only βArr2 was reintroduced (Fig. 3B), although at different times.

Fig. 3 Variable effects of restoring βArr1/2 on β2AR-stimulated ERK1/2 activation in three CRISPR βArr1/2 KO cell lines.

(A to F) Top: CRISPR βArr1/2 KO cell lines were transiently transfected with plasmid encoding FLAG-β2AR together with either empty vector or plasmids encoding hemagglutinin (HA)–βArr2 (A to C) or both HA-βArr1 and HA-βArr2 (D to F). Parental HEK293 cells (P), vector-transfected CRIPSR cells, and CRISPR cells expressing HA-βArr1, HA-βArr2, or HA-βArr1/2 were analyzed by Western blotting with antibody against βArr1/2. β-Actin was used as a loading control. Middle: Serum-deprived cells were stimulated with 100 nM isoproterenol for the indicated times, and cell lysates were analyzed by Western blotting sequentially for pERK1/2 and total ERK1/2. Western blots for each pair of vector-transfected and HA–β-arrestin–expressing CRISPR cells are representative of three (B, C, E, and F) or four (A and D) experiments. Bottom: For each vector-transfected (black symbols) and HA–β-arrestin–expressing (green symbols) CRISPR cell pair, isoproterenol-stimulated pERK/ERK ratios were normalized to the maximum signal observed in the vector-transfected cells, and data are expressed as a percentage of the maximum control response. Data are means ± SEM of three or four biological replicates. Statistical significance was determined by two-way ANOVA and Sidak’s multiple comparison test. #P < 0.05 compared with control at the corresponding time point.

If the only role of β-arrestins was to inhibit G protein signaling by promoting homologous desensitization and sequestration of GPCRs, then the consequences of βArr1/2 knockdown in parental HEK293 cells by siRNA (Fig. 2) and reconstitution in CRISPR βArr1/2 KO cells by transient transfection (Fig. 3) would be a consistent augmentation and reduction, respectively, of β2AR-stimulated ERK1/2 activation. The most parsimonious interpretation of the seemingly discordant results obtained using different parental-CRISPR pairs is that βArr1/2 perform dual roles, serving both as negative regulators of G protein signaling and as signaling scaffolds that enhance ERK1/2 activation (4448). Hence, their abundance determines the signal strength in two pathways that converge downstream on a common effector, a G protein–dependent pathway that is attenuated by β-arrestin and a β-arrestin–dependent pathway that is conferred by their presence. In such a case, a net increase, decrease, or no effect on ERK1/2 activation might all be observed depending on whether the cell background is poised for G protein–dominant or β-arrestin–dominant activation of ERK1/2. The marked variability in the responses we observed in the three different CRISPR βArr1/2 KO lines supports this interpretation and also highlights a major drawback in using such genetically altered cells for complex signaling experiments. Marked differences in the expression of myriad signaling elements in the different lines due to the rewiring of cellular pathways required to keep these genetically edited cells viable are likely responsible for the different response patterns observed.

βArr1/2 determine both the efficiency of β2AR-Gs coupling and the dominant mechanism of β2AR-mediated ERK1/2 activation

The results of our complementary loss-of-function and gain-of-function analyses of the effect of βArr1/2 on ERK1/2 pathway activation by β2ARs lead to several testable hypotheses. First, deletion of β-arrestins should increase the efficiency of β2AR-Gs coupling by removing the restraint imposed by homologous desensitization. Second, deletion of β-arrestins should reduce or eliminate β2AR internalization, leaving more agonist-responsive receptors on the plasma membrane. Third, deletion of β-arrestins should cause downstream signaling to ERK1/2 to become more dependent on G protein–regulated effectors, such as PKA.

To test the first hypothesis, we measured the efficiency of β2AR-mediated Gs activation using a bioluminescence resonance energy transfer (BRET) approach (55). In this assay, changes in BRET efficiency between a Renilla luciferase-II (Rluc-II)–tagged Gαs subunit and green fluorescent protein 10 (GFP10)–tagged Gγ1 subunit upon isoproterenol stimulation of coexpressed β2ARs report on conformational rearrangements within the Gs heterotrimer that occur upon activation. Isoproterenol stimulation produced a concentration-dependent decrease in BRET (ΔBRET) that was greater in the CRISPR βArr1/2 KO lines than in their respective parental controls for all the clones, indicating a gain in Gs activation efficiency when βArr1/2-mediated desensitization was removed (Fig. 4A). The effects of restoring βArr1 or βArr2 to the βArr1/2 KO background were also examined. Reconstitution with even low amounts of either arrestin isoform was sufficient to fully revert Gs activation to the levels observed in the parental HEK293 lines. The finding that either β-arrestin isoform alone could restore β2AR desensitization is consistent with our ERK1/2 activation data (Fig. 2), where knockdown of βArr2 in the continued presence of endogenous βArr1 consistently led to a net reduction in β2AR-mediated ERK1/2 activation, whereas knockdown of βArr1/2, which would impair desensitization and amplify G protein signaling, produced no net change in the ERK1/2 response.

Fig. 4 Effects of β-arrestin expression on Gαs activation, receptor internalization, and ERK1/2 activation by the β2AR.

(A) Top: Schematic representation of the BRET biosensors used to monitor Gαs activation. Activation of Gαs induces dissociation of Rluc-II–Gαs from GFP10-Gγ1, resulting in a decrease in BRET signal (55). GDP, guanosine diphosphate; GTP, guanosine 5′-triphosphate. Bottom: Isoproterenol concentration-response curves generated in each parental HEK293 cell line (black symbols; solid black line) together with those observed in the corresponding CRISPR βArr1/2 KO line in the absence (open symbols; dashed black line) or presence of exogenous βArr1 (open symbols; dashed blue line) or βArr2 (open symbols; dashed red line). Data are means ± SEM of 4 to 11 biological replicates as follows: SL-parental (n = 11), SL-CRISPR (n = 11), SL-CRISPR + βArr1 (n = 5), SL-CRISPR + βArr2 (n = 5), AI-parental (n = 10), AI-CRISPR (n = 10), AI-CRISPR + βArr1 (n = 4), CRISPR + βArr2 (n = 4), HAR-parental (n = 8), HAR-CRISPR (n = 8), CRISPR + βArr1 (n = 4), and CRISPR + βArr2 (n = 4). (B) Top: Schematic representation of the BRET biosensors used to monitor loss of β2AR from the plasma membrane. Receptor internalization was measured by the decrease in BRET between β2AR–Rluc-II and rGFP-CAAX labeling the plasma membrane (56). Bottom: Isoproterenol concentration-response curves generated in each parental HEK293 cell line (black symbols; solid black line) together with those observed in the corresponding CRISPR βArr1/2 KO line in the absence (open symbols; dashed black line) or presence of exogenous βArr1 (open symbols; dashed blue line) or βArr2 (open symbols; dashed red line). Data are means ± SEM of four to six biological replicates as follows: SL-parental (n = 6), SL-CRISPR (n = 6), SL-CRISPR + βArr1 (n = 5), SL-CRISPR + βArr2 (n = 5), AI-parental (n = 5), AI-CRISPR (n = 5), AI-CRISPR + βArr1 (n = 4), AI-CRISPR + βArr2 (n = 4), HAR-parental (n = 5), HAR-CRISPR (n = 5), HAR-CRISPR + βArr1 (n = 4), and HAR-CRISPR + βArr2 (n = 4). (C) Top: Schematic representation of the FRET-based assay used to monitor ERK1/2 phosphorylation. The AlphaLISA SureFire Ultra system is a sandwich enzyme-linked immunosorbent assay (ELISA) in which bridging of donor and acceptor beads by the activated pThr202/pTyr204 ERK1/2 analyte produces an increase in fluorescence emission (59). Middle: The effect of pretreatment with PTX, the PKA inhibitor 6-22, or both on 1 μM isoproterenol-stimulated ERK1/2 phosphorylation in SL-parental HEK293 cells. Bottom: Results of identical experiments performed using the corresponding SL-CRISPR βArr1/2 KO line. In each graph, responses are expressed as a percentage of the maximal isoproterenol-stimulated response in the absence of inhibitor (control). Data are means ± SEM of three to six biological replicates as follows: SL-parental: control (n = 6), PTX (n = 5), 6-22 (n = 6), and PTX + 6-22 (n = 3); SL-CRISPR: control (n = 6), PTX (n = 5), 6-22 (n = 6), and PTX + 6-22 (n = 3). Statistical significance was assessed by two-way ANOVA. *P < 0.05; **P < 0.01 for the indicated comparisons.

CRISPR KO of βArr1/2 also abolished the isoproterenol-induced loss of cell surface receptor assessed by enhanced bystander BRET (EbBRET), which monitors BRET between a Rluc-tagged β2AR and Renilla reniformis GFP anchored at the plasma membrane by a prenylated CAAX motif (rGFP-CAAX; Fig. 4B) (56). Consistent with the well-established role of β-arrestins in β2AR endocytosis, isoproterenol stimulation produced a concentration-dependent reduction in EbBRET in parental HEK cells with median effective concentration (EC50) values ranging between 208 and 395 nM for the three clones, about 10-fold higher than the EC50 for Gs activation. Reconstitution of either βArr1 or βArr2 in the CRISPR βArr1/2 KO cells was sufficient to restore isoproterenol-stimulated β2AR internalization; however, unlike for the desensitization experiments, the effect of βArr1 was only partial and required greater protein abundance than that of βArr2. This is consistent with published data showing that activated β2ARs have a higher affinity for βArr2 than for βArr1 (53, 57, 58), and suggests that whereas both β-arrestin isoforms readily support β2AR desensitization, βArr2 plays a greater role in β2AR endocytosis.

Finally, to test the hypothesis that ERK1/2 activation in a βArr1/2 KO background would be more dependent upon G protein signaling, we determined the extent to which β2AR-mediated ERK1/2 activation in parental HEK293 and CRISPR βArr1/2 KO cells was sensitive to inhibition of G protein signaling. Previous studies, in both intact HEK293 cells and an in vitro reconstituted system, demonstrated that β2ARs can use multiple G protein–mediated pathways to activate ERK1/2 (24) and that, in some cases, PKA-dependent phosphorylation of the receptor confers the capacity to activate ERK1/2 through PTX-sensitive Gi/o-mediated pathways (11, 12). Depending on the cell background, β2AR-dependent stimulation of ERK1/2 can be sensitive to inhibition of PKA, treatment with PTX, or both. Therefore, we compared the effects of pretreatment with the PKA inhibitor 6-22, PTX, or both on isoproterenol-stimulated ERK1/2 activation assayed using the AlphaLISA SureFire system, a fluorescence resonance energy transfer (FRET)–based assay in which detection of activated ERK1/2 is by immunosandwich capture of endogenous pERK1/2 in cell lysates (Fig. 4C) (59). Isoproterenol stimulation of parental HEK293 cells produced a twofold increase in FRET that was unaffected by 6-22 or PTX alone but was reduced by pretreatment with both inhibitors. In contrast, ERK1/2 activation in the corresponding CRISPR βArr1/2 KO cells was inhibited in the presence of either the PKA inhibitor or PTX alone, confirming a shift toward G protein–dependent ERK activation in the βArr1/2 KO cells.

Together, these data indicate that compared to βArr1/2-replete parental cells, the CRISPR βArr1/2 KO cells have been rewired to shift β2AR-mediated ERK1/2 activation from dual β-arrestin– and G protein–mediated pathways to a predominately Gs-cAMP-PKA pathway, most likely in compensation for the loss of β-arrestin–dependent signaling in the βArr1/2 KO CRISPR clones. The change results from amplified Gs signaling in the absence of βArr1/2-mediated desensitization and abrogated βArr2-dependent β2AR internalization.

β-Arrestin–dependent ERK1/2 activation is conserved among GPCRs that differ in their GPCR-arrestin complex stability

Another facet of GPCR regulation that may affect β-arrestin regulation of ERK1/2 is the lifetime of the GPCR-arrestin complex. Most GPCRs fall into two groups based on their relative avidity for the two β-arrestin isoforms and the stability of the GPCR-arrestin complex (58, 60). Members of the first group, termed “class A,” such as the β2AR, exhibit higher avidity for βArr2 than for βArr1 and form transient GPCR-arrestin complexes that dissociate upon receptor internalization, enabling the receptors to recycle to the plasma membrane. Members of the second group, termed “class B,” such as the V2R, exhibit equivalent avidity for βArr1 and βArr2 and form stable receptor-arrestin complexes that persist as the receptor undergoes endosomal sorting, which slows recycling and promotes lysosomal degradation. The first reports describing β-arrestin–dependent ERK1/2 activation focused on the class B receptors PAR2 (protease-activated receptor 2) and AT1AR, where pERK1/2 is retained in GPCR–β-arrestin signalsome complexes, leading to sustained ERK1/2 activation that is targeted to endosomal vesicle membranes (19, 24, 51). Subsequent studies found evidence for β-arrestin–dependent ERK1/2 activation by class A GPCRs, which also use β-arrestin scaffolds to augment ERK1/2 activation (33, 35, 54).

To investigate the effects of GPCR-arrestin complex stability on ERK1/2 activation using the CRISPR/Cas9 genome-editing approach, we selected an additional panel of receptors that all couple primarily to the Gs-cAMP-PKA pathway but represent different modes of βArr1/2 interaction: the β1AR, V2R, and the follicle-stimulating hormone receptor (FSHR). The β1AR, like the closely related β2AR, uses the class A mode of β-arrestin interaction (34). Previous studies showed that β1AR-mediated ERK1/2 activation is sensitive to knockdown of either β-arrestin isoform individually or together (33, 34) and when stimulated by Carvedilol, it requires concomitant activation of Gi/o proteins (61). We characterized β1AR-mediated ERK1/2 activation both in parental HEK293 lines in which βArr1/2 were knocked down using siRNAs and in CRISPR βArr1/2 KO lines in which βArr1/2 were reconstituted by transient transfection (Fig. 5). The efficiency of siRNA-mediated knockdown of βArr1/2 in the HEK293 parental cell lines cotransfected with plasmid encoding the murine β1AR was comparable to that achieved in experiments using the β2AR (Fig. 5, A to C). Consistent with previous reports (33, 34), siRNA-mediated knockdown of βArr1/2 caused a statistically significant reduction in β1AR-stimulated ERK1/2 activation in all three HEK293 parental lines (Fig. 5, A to C). These results contrast with those from experiments with cells expressing the β2AR, where knockdown of βArr2, but not βArr1/2, consistently impaired ERK1/2 activation across cell lines (Fig. 2), and suggest that, in the case of the β1AR, any amplification of G protein–mediated signaling that might result from knockdown of βArr1/2 is insufficient to compensate for the loss of β-arrestin–mediated signaling. We then compared β1AR-stimulated ERK1/2 activation in CRISPR βArr1/2 KO lines with CRISPR cells in which βArr1/2 was reconstituted by transient transfection (Fig. 5, D to F). We found that restoring βArr1/2 had no net effect on β1AR-stimulated ERK1/2 activation in either the AI-CRISPR or HAR-CRISPR βArr1/2 KO lines and only modestly enhanced it in the SL-CRISPR βArr1/2 KO cells. This finding mirrors our results from experiments with the β2AR (Fig. 3) and supports the interpretation that any attenuation of G protein signaling by βArr1/2-dependent desensitization and internalization of β1ARs obtained by restoring βArr1/2 was offset by the recovery of βArr1/2-dependent ERK1/2 activation.

Fig. 5 Contrast between the consistent effects of siRNA-mediated knockdown of βArr1/2 in parental HEK293 cells and the variable effects of βArr1/2 reconstitution in CRISPR βArr1/2 KO cells on ERK1/2 activation by the β1AR.

(A to C) Top: Parental HEK293 cell lines cotransfected with plasmid encoding FLAG-β1AR and siRNAs targeting no mRNA (CTL) or βArr1/2 were analyzed by Western blotting with antibody against βArr1/2. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a loading control. Middle: Serum-deprived cells were stimulated with 10 μM isoproterenol for the indicated times, and cell lysates were analyzed by Western blotting sequentially for pERK1/2 and total ERK1/2. Western blots for each pair of CTL siRNA– and βArr1/2 siRNA–treated cells are representative of four experiments. Bottom: For each CTL siRNA–treated (black symbols) and βArr1/2 siRNA–treated (red symbols) pair, isoproterenol-stimulated pERK/ERK ratios were normalized to the maximum signal observed in the CTL siRNA–treated cells, and data are expressed as a percentage of the maximum control response. Data are means ± SEM of four biological replicates. (D to F) Top: CRISPR βArr1/2 KO cell lines transiently transfected with plasmid encoding FLAG-β1AR together with either empty vector or plasmids encoding HA-βArr1 and HA-βArr2 were analyzed by Western blotting with antibody against βArr1/2. GAPDH was used as a loading control. Middle: Serum-deprived cells were stimulated with 10 μM isoproterenol for the indicated times, and cell lysates were analyzed by Western blotting sequentially for pERK1/2 and total ERK1/2. Western blots for each pair of vector-transfected and HA-βArr1/2–expressing CRISPR cells are representative of three experiments. Bottom: For each vector-transfected (black symbols) and HA-βArr1/2–expressing (green symbols) CRISPR cell pair, isoproterenol-stimulated pERK/ERK ratios were normalized to the maximum signal observed in the vector-transfected cells, and the data are expressed as a percentage of the maximum control response. Data are means ± SEM of three biological replicates. Statistical significance was determined by two-way ANOVA and Sidak’s multiple comparison test. *P < 0.05 compared with each corresponding control group; #P < 0.05 compared with control at the corresponding time point.

We performed an analogous set of experiments using the class B V2R (Fig. 6). Unlike class A receptors, V2Rs form long-lived GPCR-arrestin complexes that are retained in endosomes after receptor internalization. Because active ERK1/2 have high affinity for βArr1/2 only when the arrestin is in its “active,” GPCR-bound conformation (22), the capacity for βArr1/2 to remain bound to class B GPCRs enables them to regulate both the duration and spatial compartmentalization of ERK1/2 activation (19, 25, 27). Thus, ERK1/2 activation by the V2R is strongly dependent upon βArr1/2 scaffolding (6266). We used siRNA to knock down βArr1/2 in the HEK293 parental cell lines cotransfected with plasmid encoding the human V2R (Fig. 6, A to C). Consistent with previous reports, knockdown of βArr1/2 caused a substantial reduction in V2R-stimulated ERK1/2 activation in all three HEK293 parental lines, indicating that βArr1/2 signaling in arrestin-replete cells cannot be compensated for by the loss of desensitization that occurs when their abundance is abruptly reduced by siRNA-mediated knockdown. We then compared V2R-stimulated ERK1/2 responses in the CRISPR βArr1/2 KO lines with those in which βArr1 or βArr2 was restored by transient transfection (Fig. 6, D to F). Perhaps reflecting the larger contribution of βArr1/2 scaffolds to ERK1/2 activation by class B GPCRs, reintroducing either βArr1 or βArr2 into the βArr1/2-null background was sufficient to rescue ERK1/2 activation, resulting in a substantial increase in V2R-stimulated ERK1/2 activation in all three CRISPR lines.

Fig. 6 Reciprocal effects of βArr1/2 knockdown in parental HEK293 cells and βArr1/2 reconstitution in CRISPR βArr1/2 KO cells on ERK1/2 activation by the V2R.

(A to C) Top: Parental HEK293 cell lines cotransfected with plasmid encoding HA-V2R and siRNAs targeting no mRNA (CTL) or βArr1/2 were analyzed by Western blotting with antibody against βArr1/2. Total ERK1/2 was used as a loading control. Serum-deprived cells were stimulated with 1 μM arginine vasopressin (AVP) for the indicated times, and cell lysates were analyzed by Western blotting sequentially for pERK1/2 and total ERK1/2. Western blots for each pair of CTL siRNA– and βArr1/2 siRNA–treated cells are representative of three experiments. Bottom: For each CTL siRNA–treated (black symbols) and βArr1/2 siRNA–treated (red symbols) pair, AVP-stimulated pERK/ERK ratios were normalized to the maximum signal observed in the CTL siRNA–treated cells, and the data are expressed as a percentage of the maximum control response. Data are means ± SEM of three biological replicates. (D to F) Top: CRISPR βArr1/2 KO cell lines transiently transfected with plasmid encoding HA-V2R together with either empty vector or plasmids encoding FLAG-βArr1 or FLAG-βArr2 were analyzed by Western blotting with antibody against βArr1/2. Total ERK1/2 was used as a loading control. Serum-deprived cells were stimulated with 1 μM AVP for the indicated times, and cell lysates were analyzed by Western blotting sequentially for pERK1/2 and total ERK1/2. Western blots for each pair of vector-transfected and FLAG–β-arrestin–expressing CRISPR cells are representative of three experiments. Bottom: For vector-transfected (black symbols), FLAG-βArr1–expressing (purple symbols), and FLAG-βArr2–expressing (green symbols) cell comparisons, AVP-stimulated pERK/ERK ratios were normalized to the maximum signal observed in the vector-transfected cells, and the data are expressed as a percentage of the maximum control response. Data are means ± SEM of three biological replicates. Statistical significance was determined by two-way ANOVA and Sidak’s multiple comparison test. *P < 0.05 compared with each corresponding control group; &P < 0.05 for βArr1 reconstitution compared with control at the corresponding time point; #P < 0.05 for βArr2 knockdown or βArr2 reconstitution compared with control at the corresponding time point.

We also examined the FSHR, another Gs-coupled GPCR for which β-arrestin–dependent ERK1/2 activation has been well characterized (67, 68). The FSHR exhibits a βArr1/2 binding profile that is distinct from that of either the β1AR or V2R. Whereas it has a class B–like cluster of serine and threonine residues in its C terminus that is lacking in class A GPCRs and recruits both β-arrestin isoforms equivalently, the FSHR releases βArr1/2 soon after its internalization and is efficiently recycled (69, 70). We used siRNAs to target βArr1/2 in the HEK293 parental cell lines cotransfected with plasmid encoding the human FSHR (Fig. 7, A to C). Similar to experiments with the β1AR and V2R, siRNA-mediated knockdown of βArr1/2 caused a statistically significant reduction in FSHR-stimulated ERK1/2 activation in all three HEK293 parental lines. For the complementary reconstitution experiments, we overexpressed βArr1/2 in the CRISPR βArr1/2 KO lines to a similar extent (Fig. 7, D to F). Despite the fact that the FSHR has an atypical mode of arrestin binding, reintroducing βArr1/2 into the βArr1/2 KO background resulted in a marked enhancement in receptor-mediated ERK1/2 activation in all three CRISPR lines, mimicking the results obtained from experiments with the V2R.

Fig. 7 Reciprocal effects of βArr1/2 knockdown in parental HEK293 cells and βArr1/2 reconstitution in CRISPR βArr1/2 KO cells on ERK1/2 activation by the FSHR.

(A to C) Top: Parental HEK293 cell lines cotransfected with plasmid encoding FLAG-FSHR and siRNAs targeting no mRNA (CTL) or βArr1/2 were analyzed by Western blotting with antibody against βArr1/2. Total ERK1/2 was used as a loading control. Middle: Serum-deprived cells were stimulated with recombinant human FSH (100 ng/ml) for the indicated times, and cell lysates were analyzed by Western blotting sequentially for pERK1/2 and total ERK1/2. Western blots for each pair of CTL siRNA– and βArr1/2 siRNA–treated cells are representative of five experiments. Bottom: For each CTL siRNA–treated (black symbols) and βArr1/2 siRNA–treated (red symbols) pair, FSH-stimulated pERK/ERK ratios were normalized to the maximum signal observed in the CTL siRNA–treated cells, and the data are expressed as a percentage of the maximum control response. Data are means ± SEM of five biological replicates. (D to F) Top: CRISPR βArr1/2 KO cell lines transiently transfected with plasmid encoding FLAG-FSHR together with either empty vector or plasmids encoding HA-βArr1 and HA-βArr2 were analyzed by Western blotting with antibody against βArr1/2. Total ERK1/2 was used as a loading control. Middle: Serum-deprived cells were stimulated with FSH (100 ng/ml) for the indicated times, and cell lysates were analyzed by Western blotting sequentially for pERK1/2 and total ERK1/2. Western blots for each pair of vector-transfected and FLAG-βArr1/2–expressing CRISPR cells are representative of four (D and F) or five (E) experiments. Bottom: For each vector-transfected (black symbols) and HA-βArr1/2–expressing (green symbols) CRISPR cell pair, FSH-stimulated pERK/ERK ratios were normalized to the maximum signal observed in the vector-transfected cells, and the data are expressed as a percentage of the maximum control response. Data are means ± SEM of four or five biological replicates. Statistical significance was determined by two-way ANOVA and Sidak’s multiple comparison test. *P < 0.05 compared with each corresponding control group; #P < 0.05 for βArr1/2 knockdown or βArr1/2 reconstitution compared with control at the corresponding time point.

βArr1/2 are required for β1AR- and β2AR-mediated ERK1/2 activation by the biased agonist carvedilol

The complex web of G protein– and βArr1/2-dependent mechanisms regulating ERK1/2 activity downstream of different GPCRs lends itself to divergent outcomes when βArr1/2 are deleted by CRISPR/Cas9 gene editing or knocked down using siRNAs, mostly due to an increase in G protein pathway activity that may or may not offset the loss of β-arrestin–mediated signaling. A more direct way to test the role of β-arrestins in GPCR-stimulated ERK1/2 activation is to use a biased agonist that lacks intrinsic efficacy for heterotrimeric G protein activation but retains the capacity to promote βArr1/2 recruitment to the receptor. With such ligands, the confounding role of βArr1/2 as negative regulators of G protein signaling is eliminated, because the G protein pathways remain inactive. Carvedilol is a clinically useful, nonsubtype selective βAR antagonist (71, 72). In HEK293 cells, carvedilol behaves as an inverse agonist for β2AR stimulation of Gs while retaining the capacity to promote receptor phosphorylation by GRKs and activate ERK1/2 in a β-arrestin–dependent manner (73). In the case of the β1AR, carvedilol behaves as a very low potency partial agonist for Gs (33), while promoting ERK1/2 activation through a βArr1/2-Gi–dependent mechanism (33, 61). As well as recruiting β-arrestins, carvedilol recruits Gi to the β1AR, but not to the β2AR. β-Arrestin–dependent ERK1/2 activation by carvedilol acting on the β1AR is also PTX-sensitive, indicating a requirement for some Gi protein–mediated costimulus (61). Prolonged exposure (24 hours) to high concentrations (>10 μM) of carvedilol has also been reported to directly inhibit ERK1/2 activation in vascular smooth muscle cells (74).

We performed the complementary loss-of-function and gain-of-function experiments using carvedilol to preferentially activate β-arrestin–dependent signaling pathways in cells coexpressing either β2AR or β1AR (Fig. 8). We used siRNAs to knock down βArr1/2 in HEK293 parental cells and monitored the effects of knockdown on the ERK1/2 response to carvedilol in cells expressing the β2AR (Fig. 8A). Knockdown of βArr1/2 caused the nearly complete loss of carvedilol-stimulated ERK1/2 activation in the parental cells. We then tested the β2AR response to carvedilol in the corresponding CRISPR βArr1/2 KO line with and without βArr1/2 reconstitution (Fig. 8B). Carvedilol had no statistically significant effect on ERK1/2 activity in the CRISPR βArr1/2 KO background; however, reintroducing βArr1/2 restored the response. We performed analogous experiments with the β1AR (Fig. 8, C and D). Similar to the experiments with β2AR, carvedilol provoked a modest increase in ERK1/2 activation that was eliminated by βArr1/2 knockdown (Fig. 8C). Conversely, carvedilol had no effect on ERK1/2 activity in the CRISPR βArr1/2 KO background but stimulated a response in cells reconstituted with βArr1/2 (Fig. 8D). Together, these data support the conclusion that the responses of ERK1/2 to carvedilol, which occurred in the absence of measurable Gs signaling, were dependent upon the presence of βArr1/2.

Fig. 8 Reciprocal effects of βArr1/2 knockdown in parental HEK293 cells and βArr1/2 reconstitution in CRISPR βArr1/2 KO cells on ERK1/2 activation by the arrestin pathway–selective biased agonist carvedilol.

(A) Top: SL-parental HEK293 cells cotransfected with plasmid encoding FLAG-β2AR and siRNAs targeting no mRNA (CTL) or βArr1/2 were analyzed by Western blotting with antibody against βArr1/2. β-Actin was used as a loading control. Middle: Serum-deprived cells were stimulated with 10 μM carvedilol (Carv) for the indicated times, and cell lysates were analyzed by Western blotting sequentially for pERK1/2 and total ERK1/2. Western blots of CTL siRNA– and βArr1/2 siRNA–treated cells are representative of three experiments. Bottom: For CTL siRNA–treated (black symbols) and βArr1/2 siRNA–treated (red symbols) cells, carvedilol-stimulated pERK/ERK ratios were normalized to the maximum signal observed in the CTL siRNA–treated cells, and the data are expressed as a percentage of the maximum control response. Data are means ± SEM of three biological replicates. (B) Top: SL-CRISPR βArr1/2 KO cells transiently transfected with plasmid encoding FLAG-β2AR together with either empty vector or plasmids encoding HA-βArr1 and HA-βArr2 were analyzed by Western blotting with antibody against βArr1/2. β-Actin was used as a loading control. Middle: Serum-deprived cells were stimulated with 10 μM carvedilol for the indicated times, and cell lysates were analyzed by Western blotting sequentially for pERK1/2 and total ERK1/2. Western blots of vector-transfected and FLAG-βArr1/2–expressing CRISPR cells are representative of four experiments. Bottom: For vector-transfected (black symbols) and HA-βArr1/2–expressing (green symbols) CRISPR cells, carvedilol-stimulated pERK/ERK ratios were normalized to the maximum signal observed in the vector-transfected cells, and the data are expressed as a percentage of the maximum control response. Data are means ± SEM of four biological replicates. (C and D) Analogous set of experiments to those described in (A) and (B) were performed in HAR-parental HEK293 and HAR-CRISPR βArr1/2 KO cells expressing the β1AR. Western blots and densitometry data represent three (D) or four (C) biological replicates. Statistical significance was determined by two-way ANOVA and Sidak’s multiple comparison test. *P < 0.05 compared with each corresponding control group; #P < 0.05 for βArr1/2 knockdown or reconstitution compared with control at the corresponding time point.

DISCUSSION

Our results underscore the multiple roles played by βArr1/2 in the regulation of ERK1/2, a nodal kinase with essential roles in cell cycle progression and survival that is controlled by convergent signals downstream of multiple GPCRs and growth factor receptors (110). In the physiologic setting (Fig. 9), βArr1/2 play both positive and negative regulatory roles. By promoting the desensitization of activated GPCRs, they maintain a brake on the G protein–mediated signals that activate ERK1/2, while at the same time serving as ligand-regulated scaffolds for the cRaf1-MEK1/2-ERK1/2 cascade (Fig. 9A). Deletion of βArr1/2, by definition, removes both functions. As a result, G protein signaling is amplified, whereas βArr1/2 signaling is eliminated (Fig. 9B). The net effect on GPCR-stimulated ERK1/2 activation may be either an increase or a decrease depending on whether the enhanced signal strength in the unrestrained G protein pathway exceeds the contribution of β-arrestin–mediated signaling that exists in the wild-type environment. Our data, obtained using three independently derived CRISPR/parental HEK293 cell pairs and with four different Gs-coupled GPCRs, illustrate this point. Deletion of βArr1/2 may augment ERK1/2 activation, diminish it, or produce no net effect, as the gain in signal strength in one pathway attempts to offset the loss of another. A more complete understanding of the role of βArr1/2 scaffolds in GPCR control of ERK1/2 requires one to dissociate its two core functions, not to eliminate both. One approach that has been informative is the use of so-called β-arrestin–biased agonists, ligands that have little effect on, or even inhibit, receptor–G protein coupling but that act as agonists or partial agonists for βArr1/2 recruitment (Fig. 9C). Using these probes mitigates the confounding effects of amplified G protein signaling, enabling the effects of βArr1/2 engagement to be observed in isolation. Other tools that either have been, or may potentially be, useful in the future are GPCR mutants deficient in G protein coupling or β-arrestin mutants that retain the ability to activate signaling but lose the ability to desensitize G protein activation or vice versa (51, 54, 75, 76). Pharmacological inhibitors that selectively disrupt only some β-arrestin actions can also be useful tools. As an example, Barbadin, an inhibitor of the interaction between β-arrestins and the adaptor protein AP2 that blocks endocytosis, but not the recruitment of β-arrestin to many GPCRs, inhibits ERK1/2 activation by the V2R (77).

Fig. 9 Model depicting the dual roles of β-arrestins in the GPCR-dependent stimulation of ERK1/2.

(A) In native cells, ERK1/2 activation reflects a balance between G protein–dependent pathways that are attenuated by β-arrestin (βArr)–dependent desensitization and β-arrestin–dependent pathways that are augmented by its scaffolding function. (B) In the CRISPR/Cas9 βArr1/2 KO background, G protein–dependent ERK1/2 activation is unrestrained and β-arrestin–dependent ERK1/2 activation is absent. The resulting increase in signal strength in the G protein pathways may or may not offset the loss of β-arrestin–mediated signaling, leading to a net increase, decrease, or no change in GPCR-stimulated ERK1/2 activity. (C) β-Arrestin–biased ligands that have little or no G protein efficacy rely predominantly on β-arrestin scaffolds to support ERK1/2 activation.

In general, β-arrestins exhibit three properties that enable them to function as central regulators of GPCR signaling (14). First, like their distant cousins, the “arrestin-fold proteins” (78), they have the conformational flexibility to bind to multiple cargos, including, in the case of βArr1/2, several catalytically active proteins, such as ERK1/2, c-Src, and Mdm2 (47, 48, 79). Second, they exist in equilibrium between different intracellular pools, for example, the cytosol, nucleus, microtubules, and the plasma membrane, wherein they adopt different conformations and perform different functions (23, 8082). Finally, they have the unique capacity to recognize the active conformations of hundreds of different GPCRs. Thus, β-arrestins can constrain signaling cargos to one subcellular location, compartmentalizing or dampening pathway activity (83), until an external GPCR-mediated stimulus prompts them to undergo a conformational change that releases some cargos and engages others. Within such a complex network of interactions, differences in the relative abundance of GPCRs and their signaling effectors, as might exist between tissues or in disease states, are likely to cause substantial shifts in signal strength between G protein– and β-arrestin–mediated pathways. Our data from CRISPR βArr1/2 KO cells highlight this. Despite being derived from common HEK293 cell progenitors, independent CRISPR KO clones react differently to the deletion of βArr1/2. In some, for example, AI-CRISPR βArr1/2 KO cells, deletion of βArr1/2 led to an unambiguous amplification of β2AR-dependent stimulation of the ERK1/2 pathway compared to the parental clone, because G protein signaling was unhindered by βArr1/2-dependent desensitization (Fig. 1A) (42). Such findings suggest that the selective pressure applied by CRISPR/Cas9-mediated KO of βArr1/2 can favor the isolation of clones that compensate for the loss of β-arrestin–dependent signaling by rewiring to a predominately Gs-cAMP-PKA pathway. However, this degree of change is not apparent in other clones, for example, SL-CRISPR βArr1/2 KO and HAR-CRISPR βArr1/2 KO cells, likely due to subtle background differences between them (Fig. 1, B and C). With GPCRs that are more tightly βArr1/2-coupled, for example, the V2R, deletion of βArr1/2 produced a consistent reduction in ERK1/2 signaling that was rescued by restoring β-arrestin expression (Fig. 6). Curiously, O’Hayre et al. (42) did not observe a reduction in V2R-stimulated ERK1/2 in the same AI-CRISPR βArr1/2 KO cells that we used in our study, a discrepancy that we currently do not understand. Nonetheless, a range of experimental approaches, including CRISPR βArr1/2 KO cells (this study), siRNA-mediated knockdown, pharmacological inhibition of the β-arrestin–AP2 interaction, and dominant-negative β-arrestin mutants (6266, 77), all converge to indicate that β-arrestins play an important role in the early and late phases of ERK1/2 activation by the V2R.

It should also be stressed that the phrases “arrestin-dependent” and “G protein–independent” are not synonymous. An arrestin-dependent signal is one that is enhanced or conferred by the presence of arrestins and diminished or abolished by their deletion. In considering arrestin-dependent regulation of the ERK1/2 cascade, it is important to recognize that the upstream element, cRaf1, is not a constitutively active kinase. βArr1/2 scaffolding of the c-Jun N-terminal kinase 3 (JNK3) cascade, where simple coexpression of βArr1/2 together with the component kinases ASK1, MKK4 or MKK7, and JNK3 leads to a marked activation of JNK3, is not only G protein–independent but also GPCR-independent (21, 8486). In contrast, the affinity of cRaf1 and ERK1/2 for arrestins is sensitive to βArr1/2 conformation such that only GPCR-bound βArr1/2 supports the efficient assembly of the ERK1/2 activation complex (22). As a result, in the absence of GPCR activation, βArr1/2 act as “silent scaffolds,” sequestering MEK1/2 in a microtubule-bound pool and dampening basal ERK1/2 pathway activity until the ligand-induced formation of high-affinity GPCR docking sites on the plasma membrane recruits βArr1/2, promotes complex assembly, and brings cRaf1 into the proximity of an activating stimulus. In cells in which basal Ras signaling is high, simple βArr1/2-dependent assembly of the cRaf1-MEK1/2-ERK1/2 complex on a membrane surface may be sufficient to exceed the pathway activation threshold. GPCR-independent recruitment of cyclophilin FRB domain–tagged βArr2 to a myristoylated-cyclophilin FKBP domain expressed on the plasma membrane is sufficient to activate ERK1/2, as is expression of a constitutively internalized neurokinin NK1 receptor–βArr1 chimera (28, 87).

On the other hand, formal demonstration that an arrestin-dependent signal is G protein–independent would require either in vitro reconstitution of the complex with pure proteins or examination of GPCR signaling in cells devoid of heterotrimeric G proteins. In more native systems, it is inevitable that β-arrestins and G proteins would act cooperatively to regulate ERK1/2, because activation of G protein effectors can provide a cRaf1-activating stimulus and β-arrestins can recruit and orient the ERK1/2 pathway components on the plasma membrane in proximity to that stimulus. For example, it is known that the β-arrestin–dependent activation of ERK1/2 by many Gi/o-coupled receptors is also PTX-sensitive (61, 88). Similarly, whereas the G protein–coupling efficacy of β-arrestin–biased agonists is disproportionally reduced relative to their ability to stimulate β-arrestin recruitment, most are not devoid of the ability to activate G proteins. For example, Sar1-Ile4-Ile8 acting on the AT1AR (89) and carvedilol acting on the β1AR (61) behave as weak partial agonists for G protein signaling. Thus, it is unsurprising that these ligands can activate ERK1/2 in some βArr1/2 KO backgrounds. β-Arrestin–biased agonists that exhibit true reversal of efficacy, that is, that act as inverse agonists for G protein coupling while still promoting β-arrestin recruitment, for example, [D-Trp12,Tyr34]-bovine parathyroid hormone (PTH) (734) acting on the human type 1 PTH receptor (PTH1R) (90) and carvedilol acting on the β2AR (61, 73), are rare. However, their capacity to cause β-arrestin–dependent ERK1/2 activation while suppressing basal G protein signaling is difficult to reconcile with the proposed universal requirement for heterotrimeric G proteins in the action of β-arrestin–biased agonists (41). Observations from CRISPR/Cas9 genome-edited cell lines virtually devoid of heterotrimeric G protein activity reinforce the notion that G proteins and β-arrestins often act cooperatively, while at the same time demonstrating that ERK1/2 activation by many GPCRs is markedly diminished by CRISPR/Cas9-mediated deletion of βArr1/2 (41). Although our data are not inconsistent with previous studies using CRISPR/Cas9-edited HEK293 cells (41, 42), the focus of those studies on the role of heterotrimeric G proteins fails to recognize that β-arrestin scaffolds not only act as signal amplifiers but also redirect ERK1/2 into functionally discrete pools, targeting ERK1/2 to nonnuclear substrates involved in the regulation of GPCR trafficking, protein translation, and cytoskeletal rearrangement (2939). Assaying total cellular ERK1/2 activity captures only the effects on signal amplitude and duration and can miss β-arrestin functions related to the compartmentalization of functionally distinct ERK1/2 pools (38, 39). Developing a more complete understanding of the roles of β-arrestins would at minimum require the use of reporters that discriminate between the activity of nuclear and cytosolic ERK1/2.

Our data highlight another point—that βArr1/2 do not perform the same functions downstream of all GPCRs even in a common cellular background. Consistent with previous reports (58, 62), we found that βArr2 played a larger role than did βArr1 in supporting β2AR internalization and ERK1/2 activation, whereas both isoforms efficiently support β2AR desensitization (Figs. 2 and 4). Similar evidence of βArr1/2 isoform specialization has been published for the Gq/11-coupled AT1AR, where again βArr2 appears to play the major signaling role, and for the PAR2 receptor, for which βArr1 is the dominant isoform (25, 31, 38, 91). In contrast, our data, together with data from published studies, indicate that for other receptors, for example, the V2R, either βArr1 or βArr2 is sufficient to support ERK1/2 activation (Fig. 6) (6266), whereas for still others, for example, the FSHR and PTH1R, both βArr1 and βArr2 are required (Fig. 7) (27, 67, 68). β-Arrestins also exert GPCR-specific effects on the duration of ERK1/2 activation. In the case of the predominantly Gs-coupled β2AR, β1AR, V2R, and FSHR studied here, depletion of β-arrestins affects the early (2 to 15 min) phase of ERK1/2 activation (33, 54, 66, 68), whereas for receptors with slow rates of ligand dissociation, such as the Gq/11-coupled angiotensin receptor AT1AR, the Gs/Gq/11-coupled PTH1R, and the Gi/Gq-coupled PAR2 receptor, the effects of β-arrestin depletion are most evident at later times (26, 27, 31).

What might be the physical basis for these apparently contradictory results? Certainly, the intracellular domain architecture of different activated GPCRs affects their avidity for βArr1 and βArr2, just as it determines G protein–coupling selectivity (58). Biophysical data obtained using intramolecular FRET or BRET probes suggest even greater nuance, in that different GPCRs appear to induce distinct active β-arrestin conformations, enabling them to determine not only which β-arrestin isoforms are activated but also what functions they perform (92, 93). Comparison of the population average “conformational signatures” of βArr2 induced by different GPCRs in a common cellular background demonstrates that receptors that use β-arrestins to activate ERK1/2 cause characteristic shifts in βArr2 conformation that are not seen with GPCRs that use β-arrestins in a predominantly desensitizing role. Beyond the influences of cell background and GPCR structure, it is likely that the functions of βArr1/2 are determined by multiple factors, including ligand structure (94, 95), GPCR C-tail phosphorylation “bar codes” produced by different GRKs (9698), and posttranslational modifications that stabilize or destabilize the receptor-arrestin complex (99), which collectively influence βArr1/2 conformation.

On the basis of work performed using CRISPR/Cas9 and TALEN KO approaches to demonstrate that βArr1/2 are dispensable for the β2AR-dependent stimulation of ERK1/2 and that β-arrestin–biased agonists are capable of activating ERK1/2 in the absence of β-arrestins, it has been suggested that genome-edited cells are superior to alternatives, such as cells subjected to short-term, siRNA-mediated knockdown, for interrogating the mechanisms underlying biased drug responses (41, 42). Although there are clear advantages to genome-edited cell lines for many applications, CRISPR cell clones are selected on the basis of their capacity to exist in the absence of the deleted protein(s). Hence, whereas such cells are outstanding tools for studying how cells can survive without a missing protein, they are less useful for understanding how cells survive with it or how biased ligands affect cells with intact signaling networks. For example, although germline deletion of either β-arrestin isoform individually is tolerated, simultaneous deletion of βArr1/2 results in embryonic or very early postnatal lethality, which is characterized by severe developmental defects (100). However, it is possible to derive immortalized MEF clones from ARRB1/2 KO embryos (50) and to delete ARRB1/2 using CRISPR/Cas9-dependent gene-editing approaches (101). Thus, although they are apparently indispensable for mammalian development and organismal survival, βArr1/2 are unquestionably dispensable for the survival of cultured cells in vitro, as are apparently the entire set of heterotrimeric Gα proteins (100). βArr1/2 play some vital role(s) in development, but the nature of those roles, or how biased drugs may affect them, cannot be readily elucidated using cells that have adapted to survive without them.

Despite the remarkable contrast in our study between the consistent results we obtained from multiple HEK293 cell clones using siRNA-mediated knockdown of βArr1/2 compared to the variability observed between independently derived CRISPR βArr1/2 clones, it is important to emphasize that no engineered cell line is a perfect model system. Both CRISPR/Cas9 and siRNA approaches are susceptible to off-target effects that can complicate the interpretation of results (102, 103). Complementary gain-of-function and loss-of-function approaches and the use of multiple different siRNA sequences or independent CRISPR/Cas9 clones, as done here, may provide confidence that the observed effects are the consequence of the intended manipulation. Most studies to date of the involvement of β-arrestins in GPCR regulation of ERK1/2 using siRNAs have not included such rescue approaches, just as most studies using CRISPR βArr1/2 KO have been performed using CRISPR/Cas9 clones derived from a single parental background (41, 42). Whereas an abrupt reduction in the abundance of target proteins using siRNAs may reduce the likelihood that compensatory rewiring of intracellular signaling networks will obscure the function of the missing protein, siRNA approaches are limited in that knockdown is invariably incomplete. In this case, functions that do not require large amounts of the targeted protein may be unaffected and important roles may be missed. The β-arrestins, most of whose functions involve stoichiometric interaction with binding partners, may be somewhat less susceptible to this phenomenon than are proteins with catalytic activity, for example, kinases, but the potential for misleading results from experiments with siRNA-mediated knockdown must be highlighted.

Given the many caveats our data raise with respect to the effects of signal rewiring in CRISPR/Cas9 genome-edited cells, it is difficult to justify the claim that they are superior platforms for biased drug screening (41, 49). Studies of ERK1/2 activation in β-arrestin–replete cells using inverse agonists for the β2AR, V2R, and PTH1R (27, 64, 73) are not easily reconciled with CRISPR/Cas9 cell data showing a universal requirement for heterotrimeric G protein signaling. Whether the requirement for G protein signaling observed in CRISPR “G-zero cells” (41) indicates that some background level of constitutive G protein activity or some threshold level of GPCR-dependent G protein activation is required for β-arrestin scaffolds to activate ERK1/2 is unclear. That arrestin-biased agonists can activate ERK1/2 in βArr1/2 KO CRISPR cells or fail to do so in G-zero CRISPR cells may have as much to do with the nature of the genome-edited cells as with the actions of ligands in native backgrounds. GPCR overexpression and βArr1/2 deletion both favor promiscuous coupling to nonphysiologic effectors (104, 105). Deletion of Gα subunits may lead to unknown effects on Gβγ subunit abundance, function, or both and potentially alter the abundance or function of the Gβγ-dependent GRKs (GRK2/3) that promote desensitization versus the Gβγ-independent GRKs (GRK5/6) that favor β-arrestin signaling (65, 96). Ultimately, whether background G protein activity or a low level of G protein activation is required for β-arrestin to scaffold the ERK1/2 module is largely immaterial with respect to the basic paradigms of biased signaling. Hence, strategies such as acute knockdown of proteins by siRNA, inducible short hairpin RNA (shRNA), or CRISPR/Cas9 transient epigenetic modification technology, wherein cells need not survive long term in the absence of the targeted protein, may be more reliable and interpretable. This is particularly relevant to drug development where disease states are more likely due to altered physiology from a modest reduction in the function of a protein rather than its long-term absence. At the end of the day, the choice of model system must be based on the experimental questions being asked, and no data should be overinterpreted or overgeneralized. Engineered HEK293 cells offer many advantages for signal transduction research and as drug discovery platforms, but overexpressed receptors and genetic manipulations inevitably raise concerns about the physiologic relevance of the processes being studied. These systems are invaluable tools for hypothesis generation, but studies conducted in primary cells, organoids, or intact animals are ultimately necessary to fully understand the nuances of GPCR signaling.

MATERIALS AND METHODS

Antibodies and reagents

AVP, (−)-isoproterenol (+)-bitartrate, carvedilol, G418, poly-l-lysine, and o-phenylenediamine dihydrochloride were from Sigma-Aldrich. [125I](−)-Iodocyanopindolol was from PerkinElmer Life Sciences. The PKA inhibitor 6-22 was from Tocris Bioscience. Vybrant DyeCycle orange stain and salmon sperm DNA were from Invitrogen. Coelenterazine 400a was from NanoLight Technology. Recombinant human FSH was a gift from F. Gonal of Merck Serono, Lipofectamine 2000 was from Thermo Fisher Scientific, FuGene6 was from Promega Life Sciences, METAFECTENE was from Biontex Laboratories, polyethylenimine linear 25K was from Polysciences, and GeneSilencer was from Genlantis. Mouse monoclonal anti–phospho-p44/42 ERK1/2 (catalog no. 9106) and rabbit polyclonal anti–phospho-p44/42 ERK1/2 (catalog no. 9101) were from Cell Signaling Technology. Rabbit polyclonal anti-ERK1/2 (catalog no. 9102) was from Cell Signaling Technology or EMD Millipore (catalog no. ABS44). Rabbit polyclonal anti-βArr1/2 antibody (A1CT), which recognizes both βArr1 and βArr2 isoforms, was generated in the laboratory of R. J. Lefkowitz (Duke University, Durham, NC) (106). Anti–β-actin mouse monoclonal antibody (clone AC-15) was from Sigma-Aldrich. Anti-GAPDH mouse monoclonal antibody (clone 6C5) was from Santa Cruz Biotechnology. Horseradish peroxidase (HRP)–conjugated anti-mouse and anti-rabbit immunoglobulin G (IgG) were from GE/Amersham Biosciences, Rockland Immunochemicals, Bio-Rad Laboratories, or Jackson ImmunoResearch Europe. IRDye Alexa Fluor 680 and 800 secondary antibodies were from LI-COR Biosystems. Fluorescein isothiocyanate (FITC)–conjugated anti-FLAG antibody was from Sigma-Aldrich, terbium-conjugated anti-FLAG antibody was from Cisbio, and HRP-conjugated anti-HA IgG was from Roche Diagnostics.

Derivation of CRISPR/Cas9 HEK293 cell lines

The AI-CRISPR βArr1/2 KO and corresponding AI-parental HEK293A cell lines were a gift from A. Inoue (Tohoku University, Sendai, Miyagi, Japan) and derived as previously described in detail (42). This is the CRISPR βArr1/2 KO-parental HEK293 pair used by O’Hayre et al. (42) and Grundmann et al. (41). The SL-CRISPR βArr1/2 KO line was generated in the Laporte laboratory (McGill University, Montreal, Quebec, Canada) by successive deletion of βArr1 and βArr2 from a subclone of HEK293 cells (56) using the CRISPR/Cas9 method. ARRB1 deletion was performed first using the vector pSpCas9(BB)-2A-Puro (Px459v2; Addgene plasmid no. 62988; deposited by F. Zhang) (107) and the two guide RNA (gRNA) sequences 5′-CACCGTGTGGACCACATCGACCTCG-3′ and 5′-CACCGCAACGTACAGTCGTTCCCAC-3′ for the human ARRB1 gene exons 3 and 5, respectively. Each gRNA was cloned into the Px459v2 vector, and cells were transfected with both constructs using Lipofectamine 2000 according to the manufacturer’s instructions. After 24 hours, medium supplemented with puromycin (2 μg/ml) was added for 48 hours and the cells were serially diluted and plated in 96-well plates to generate individual colonies. Deletion of ARRB1 was verified by Western blotting for βArr1 and polymerase chain reaction genotyping using primers flanking both deleted and undeleted regions, and a positive clone was selected. One ARRB1-deleted cell line (clone 13) was subsequently used for ARRB2 deletion using the same method with the two gRNA sequences 5′-CACCGAAGTCGAGCCCTAACTGCA-3′ and 5′-CACCGCCTGTTCATCGCCACCTACC-3′ for the human ARRB2 gene exons 2 and 5, respectively. Selection was performed as described for the ARRB1 deletion, and a positive ARRB1/2-deleted clone was selected (clone 19). This clone is referred to as the SL-CRISPR βArr1/2 KO line used herein. The HAR-CRISPR βArr1/2 KO line was generated in the Rockman laboratory (Duke University, Durham, NC) as described previously (61), using the pSpCas9(BB)-2A-Puro (Px459) vector (Addgene plasmid no. 48139). ARRB1 was targeted using the gRNA sequences 5′-CACCGCATCGACCTCGTGGACCCTG-3′ and 5′-AACCAGGGTCCACGAGGTCGATGC-3′. ARRB2 was targeted using the gRNA sequences 5′-CACCGCGTAGATCACCTGGACAAAG-3′ and 5′-AAACCTTTGTCCAGGTGATCTACGC-3′. The guide sequence oligos were cloned into pSpCas9(BB)-2A-Puro, and plasmids and HEK293 cells were transfected with them using FuGene6 according to the manufacturer’s instructions. Seventy-two hours after transfection, the cells were harvested to check genome insertion/deletion using the Surveyor Nuclease Assay. Puromycin (2.5 μg/ml) was added to the medium of Surveyor-positive cells to select clones containing the puromycin resistance plasmid along with gRNA and Cas9. KO of both β-arrestins was confirmed by Western blotting. One clone referred to as the HAR-CRISPR βArr1/2 KO line was used herein.

Knockdown of βArr1/2 in parental HEK293 cells using siRNA

AI-parental and HAR-parental HEK293 cells were maintained in minimal essential medium (MEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin or gentamicin (20 μg/ml). SL-parental cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) without pyruvate supplemented with 10% FBS and 1% penicillin/streptomycin or gentamicin (20 μg/ml). ERK1/2 activation assays in response to β2AR stimulation were performed in the Shenoy and Rockman laboratories (Duke University, Durham, NC). For experiments comparing the effects of receptor density on β2AR-stimulated ERK1/2 activation, stable lines of AI-parental HEK293 cells and AI-CRISPR βArr1/2 KO cells were generated by transfecting cells with plasmid DNA encoding FLAG epitope–tagged β2AR (pcDNA3.1-FLAG-β2AR) and selecting positive clones against G418 (1 mg/ml). Stable selection was maintained by the inclusion of G418 (400 μg/ml) in the growth medium. Knockdown of β-arrestins in parental HEK293 cells was performed using chemically synthesized, double-stranded siRNAs purchased from GE/Dharmacon. The siRNA sequences used were as follows: human βArr2, 5′-CTAAATCACTAGAAGAGAA-3′ (103); human βArr1/2, 5′-ACCTGCGCCTTCCGCTATG-3′ (27, 108). The nontargeting control siRNA sequence was 5′-AATTCTCCGAACGTGTCACGT-3′. For siRNA experiments, parental HEK293 cells in 10-cm dishes at 40 to 50% confluence were transfected with 20 μg of siRNA using the GeneSilencer transfection reagent according to the manufacturer’s protocols. Forty-eight hours after transfection, cells were split into six-well dishes for ERK1/2 activation assays. Before stimulation, the cells were incubated for 4 hours in serum-free medium containing 10 mM Hepes (pH 7.5) and 0.1% bovine serum albumin (BSA). Western blotting analysis of whole-cell lysates with the A1CT anti-βArr1/2 antibody was performed to assess the extent of βArr2 or βArr1/2 knockdown. Assays assessing β1AR responses were performed in the Rockman laboratory (Duke University, Durham, NC). Before experimentation, the cells were transiently cotransfected with plasmid DNA encoding FLAG epitope–tagged murine β1AR (pcDNA3.1-FLAG-β1AR) and either control siRNA or siRNA targeting βArr1/2 (27, 52, 61). βArr1/2 were targeted by combining three siRNAs: 5′-AAACCTGCGCCTTCCGCTATG-3′, 5′-AAAGCCTTCTGCGCGGAGAAT-3′, and 5′-AAGGACCGCAAAGTGTTTGTG-3′ in a 2:1:1 ratio, respectively. The nonsilencing RNA duplex 5′-AATTCTCCGAACGTGTCACGT-3′ was used as a control. HEK293 cells were seeded into 10-cm dishes and allowed to reach 30 to 40% confluence at the time of transfection. Cotransfection was performed using the GeneSilencer Transfection Reagent according to the manufacturer’s protocols, with 4 μg of pcDNA3.1-FLAG-β1AR plasmid and 20 μg of pooled siRNA per dish. Twenty-four hours after transfection, the cells were split into six-well dishes. Stimulations were performed 72 hours after transfection following overnight incubation in serum-free medium supplemented with 10 mM Hepes (pH 7.5), 0.1% BSA, and 1% penicillin/streptomycin. Cell surface β1AR abundance was assessed by FITC-conjugated anti-FLAG antibody binding and quantified with a BD LSRII flow cytometer (BD Biosciences). Western blotting analysis of whole-cell lysates with the A1CT anti-βArr1/2 antibody was performed to assess the extent of βArr2 or βArr1/2 knockdown. Assays assessing the V2R response were performed in the Laporte laboratory (McGill University, Montreal Quebec, Canada). Before experimentation, the cells were transiently cotransfected with plasmid DNA encoding HA epitope–tagged human V2R (pcDNA3.1-HA-V2R) (104) and either control siRNA or siRNA targeting βArr1/2. The siRNA sequences used were as follows: human βArr1/2, 5′-ACCTGCGCCTTCCGCTATG-3′; nontargeting control siRNA, 5′-TAAGGCTATGAAGAGATAC-3′. HEK293 cells were seeded in poly-l-lysine–coated 12-well plates at a density of 0.7 × 105 cells per well. Forty-eight hours later, the cells were cotransfected with 250 ng of HA-V2R plasmid and 100 nM of either control or βArr1/2-targeted siRNA using the Lipofectamine 2000 transfection reagent according to the manufacturer’s protocols and allowed to grow for an additional 48 hours. Before stimulation, the cells were incubated for 30 min in serum-free medium containing 20 mM Hepes (pH 7.5). Western blotting analysis of whole-cell lysates with the A1CT anti-βArr1/2 antibody was used to assess the extent of βArr2 or βArr1/2 knockdown. Assays to assess FSHR responses were performed in the Reiter laboratory (Institut National de la Recherche Agronomique, CNRS, Université de Tours, Nouzilly, France). Before experimentation, the cells were transiently cotransfected with plasmid DNA encoding FLAG epitope–tagged human FSHR (pcDNA3-FLAG-FSHR) (109) and either control siRNA or siRNA targeting βArr1/2. The siRNA sequences used were as follows: human βArr1/2, 5′-AAACCTGCGCCTTCCGCTATG-3′; nontargeting control siRNA, 5′-TTCTCCGAACGTGTCACGT-3′ (109). For siRNA experiments, parental HEK293 cells in 10-cm dishes at 30% confluence were transfected with 2 μg of pcDNA3-FLAG-FSHR and 20 μg of either control or βArr1/2-targeted siRNA using the GeneSilencer transfection reagent according to the manufacturer’s protocols. Twenty-four hours after transfection, the cells were split into 12-well dishes. Assays were performed 72 hours after transfection following 4 hours of incubation in serum-free medium containing 10 mM Hepes (pH 7.5) and 0.1% BSA. Plasma membrane FSHR abundance was assessed on live cells using Terbium-conjugated anti-FLAG antibody measured with a TriStar2 S LB942 microplate reader (Berthold Technologies). Western blotting analysis of whole-cell lysates with the A1CT anti-βArr1/2 antibody was performed to assess the extent of βArr2 or βArr1/2 knockdown.

Reconstitution of βArr1/2 in CRISPR/Cas9 HEK293 cells by transient transfection

AI-CRISPR and HAR-CRISPR βArr1/2 KO cells were maintained in MEM supplemented with 10% FBS and 1% penicillin/streptomycin or gentamicin (20 μg/ml). SL-CRISPR βArr1/2 KO cells were maintained in DMEM without pyruvate supplemented with 10% FBS and 1% penicillin/streptomycin or gentamicin (20 μg/ml). For ERK1/2 activation assays of cells transiently expressing β2AR, CRISPR βArr1/2 KO cells in 10-cm dishes at 40 to 50% confluence were transiently cotransfected with 2 μg of pcDNA3.1-FLAG-β2AR and 1 μg of plasmid encoding HA epitope–tagged βArr2 (pcDNA3.1-HA-βArr2) or 1 μg each of pcDNA3-HA-βArr2 and a plasmid encoding HA epitope–tagged βArr1 (pcDNA3.1-HA-βArr1) using Lipofectamine 2000 according to the manufacturer’s protocol. Equal amounts of empty vector were used for mock-transfected controls. Twenty-four hours after transfection, cells were plated onto six-well dishes. Stimulations were performed 16 to 24 hours later following 4 hours of serum deprivation. The cell surface abundance of β2AR was determined by [125I](−)-iodocyanopindolol binding. Western blotting analysis of whole-cell lysates with the A1CT anti-βArr1/2 antibody was used to assess the abundance of βArr2 or βArr1/2. For assays of cells transiently expressing β1AR, CRISPR βArr1/2 KO cells in 10-cm dishes were transiently cotransfected with 4 μg of pcDNA3-FLAG-β1AR and 50 ng of pcDNA3.1-HA-βArr1 and 250 ng of pcDNA3.1-HA-βArr2 using FuGene6 according to the manufacturer’s protocol. Equal amounts of empty vector were used for mock-transfected controls. Twenty-four hours after transfection, cells were split into six-well dishes and serum-deprived overnight before stimulation. β1AR cell surface abundance was assessed by FITC-conjugated anti-FLAG labeling, and βArr1/2 abundance was assessed by Western blotting analysis with the A1CT anti-βArr1/2 antibody. For assays of cells transiently expressing V2R, CRISPR βArr1/2 KO cells were seeded in 10-cm dishes at an initial density of 7.5 × 105 cells per dish. Twenty-four hours later, the cells were transfected with 3 μg of pcDNA3.1-HA-V2R with or without 200 ng of pcDNA3.1-FLAG-βArr1 or pcDNA3.1-FLAG-βArr2 (77) using the calcium phosphate DNA precipitation method as previously described (110). Equal amounts of empty vector were used for mock-transfected controls. The next day, cells were detached with phosphate-buffered saline (PBS) and 5 mM EDTA and plated onto poly-l-lysine–coated 12-well plates. Stimulations were performed 24 hours later following 30 min of serum deprivation. Western blotting analysis of whole-cell lysates with the A1CT anti-βArr1/2 antibody was used to assess the abundance of βArr2 or βArr1/2. For assays of cells transiently expressing FSHR, CRISPR βArr1/2 KO cells in 10-cm dishes at 30% confluence were cotransfected with 2 μg of pcDNA3-FLAG-FSHR and 200 ng each of pcDNA3.1-HA-βArr1 and pcDNA3.1-HA-βArr2 or 400 ng of empty vector using METAFECTENE transfection reagent according to the manufacturer’s protocol. Twenty-four hours after transfection, cells were split into 12-well dishes. After a further 24 hours, cells were serum-starved for 4 to 6 hours before stimulation. Plasma membrane FSHR abundance was assessed using Terbium-conjugated anti-FLAG antibody, whereas βArr1/2 abundance was assessed by Western blotting analysis of whole-cell lysates with the A1CT anti-βArr1/2 antibody.

Assessment of ERK1/2 activation by Western blotting analysis

Appropriately transfected and serum-deprived parental and CRISPR βArr1/2 KO cells were stimulated at 37°C with the ligand concentrations and for the times described in the figure legends. After stimulation, the cells were lysed at 4°C in 1% NP-40 lysis buffer [20 mM tris-HCl (pH 7.4), 137 mM NaCl, 20% glycerol, 1% NP-40, 2 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 10 mM sodium fluoride, aprotinin (10 μg/ml), leupeptin (5 μg/ml), and phosphatase inhibitors], protein concentration was determined by Bradford assay (Bio-Rad Laboratories), and equal amounts of cell lysate were mixed with 2× Laemmli sample buffer. Alternatively, stimulation was stopped by directly adding 2× Laemmli buffer [250 mM tris-HCl (pH 6.8), 2% (w/v) SDS, 10% (v/v) glycerol, 0.01% (w/v) bromophenol blue, and 5% (v/v) β-mercaptoethanol] to solubilize the cells. Lysates containing equal amounts of total cell protein were resolved by SDS–polyacrylamide gel electrophoresis on 10% or 4 to 20% gradient tris-glycine polyacrylamide gels (Invitrogen), transferred to nitrocellulose or polyvinylidene difluoride membranes, and incubated with rabbit polyclonal or mouse monoclonal anti–phospho-44/42 ERK1/2 and anti–total ERK1/2, followed by HRP-conjugated anti-rabbit IgG or IRDye conjugates as secondary antibodies. Protein bands were detected with Pierce SuperSignal West Pico Enhanced Chemiluminescence Reagent (Thermo Fisher Scientific) and captured using a ChemiDoc-XRS charge-coupled device camera system (Bio-Rad Laboratories) or either a G:BOX (Syngene) or a LI-COR Odyssey CLX far infrared scanner (LI-COR Biosystems). Bands were quantified by densitometry using Image Lab (Bio-Rad Laboratories) or ImageJ software.

BRET assay of Gαs activation and β2AR internalization

BRET assays of β2AR-mediated Gs activation and receptor internalization were performed in the Bouvier laboratory (University of Montreal, Montreal, Quebec, Canada). Parental HEK293 and CRISPR βArr1/2 KO cells were maintained in DMEM supplemented with 10% FBS, penicillin (100 U/ml), streptomycin (100 μg/ml), and 2 mM l-glutamine. Transient transfections were performed at a density of 0.3 × 106 to 0.7 × 106 cells/ml using a 3:1 ratio of 25-kDa linear polyethylenimine per microgram of DNA, as described previously (111). A maximum of 1.2 μg of plasmid DNA was used to transfect 1.2 ml of cells, and the total amount of DNA was maintained constant with salmon sperm DNA. β2AR-mediated Gs activation was assessed by measuring changes in BRET efficiency between an Rluc-II–tagged Gαs subunit and GFP10-tagged Gγ1 subunit upon stimulation (55). Cells were cotransfected with plasmids encoding HA epitope–tagged β2AR, untagged Gβ1, and Gαs tagged with Rluc-II at amino acid position 117 (Rluc-II–Gαs), and Gγ1 tagged with GFP10 at its N terminus (GFP10-Gγ1). Transfected cells (100 μl per well) were seeded directly into 96-well white wall microplates. For reconstitution experiments, plasmid DNA encoding human βArr1 or βArr2 was included in the transfection. Assays were performed 48 hours after transfection. On the day of the assay, the culture medium was removed, cells were washed with phenol red–free Dulbecco’s PBS (DPBS), and the medium was replaced by Tyrode’s buffer [137 mM NaCl, 0.9 mM KCl, 1 mM MgCl2, 11.9 mM NaHCO3, 3.6 mM NaH2PO4, 5.5 mM glucose, 1 mM CaCl2, and 25 mM Hepes (pH 7.4)]. After the addition of agonist or vehicle, cells were incubated at 37°C for 15 min. The luciferase substrate coelenterazine 400a was added to a final concentration of 2.5 μM 5 min before reading BRET in a Synergy Neo microplate reader (BioTek) equipped with 515 ± 30 nm acceptor and 410 ± 80 nm donor filters. The BRET signal was determined as the ratio of the light emitted by the GFP-tagged biosensor (energy acceptor) and the light emitted by the Rluc-II–tagged biosensor (energy donor). The agonist-promoted BRET signal (ΔBRET) is the difference in BRET recorded from cells treated with agonist and cells treated with vehicle. β2AR internalization was assessed by EbBRET, which monitors BRET between an Rluc-tagged receptor and rGFP anchored at the plasma membrane by a prenylated CAAX motif (56). To measure receptor internalization, cells were cotransfected with plasmid encoding β2AR tagged with Rluc-II at its C terminus (β2AR–Rluc-II) and with a plasmid encoding the prenylated CAAX domain of K-Ras fused to Renilla GFP (rGFP-CAAX) in the absence or presence of human βArr1 or βArr2. Transfected cells (100 μl per well) were seeded directly into 96-well white wall microplates and assayed 48 hours after transfection. On the day of the assay, the culture medium was removed, the cells were washed with DPBS, and the medium was replaced by Tyrode’s buffer. After addition of agonist or vehicle, cells were incubated at 37°C for 15 min and the luciferase substrate coelenterazine 400a was added 5 min before reading BRET in a Synergy Neo microplate reader. The BRET signal was determined as described for the Gαs activation assay, and the results are expressed as ΔBRET.

SureFire ERK1/2 activation assay

In the Bouvier laboratory, β2AR-stimulated ERK1/2 phosphorylation in live cells was assayed using the AlphaLISA SureFire Ultra Kit (PerkinElmer Life Sciences), a FRET assay based on a sandwich ELISA in which the bridging of donor and acceptor beads by the activated pThr202/pTyr204 ERK1/2 analyte produces an increase in fluorescence emission at 520 to 620 nm upon excitation at 680 nm (59). For this assay, cells were transfected with plasmid encoding HA-β2AR, and the transfected cells were seeded (100 μl per well) into 96-well white microplates precoated with poly-d-lysine hydrobromide. Twenty-four hours after transfection, the cells were serum-deprived for 16 hours with or without PTX (0.1 μg/ml) as appropriate. ERK1/2 activation was assayed 48 hours after transfection. The culture medium was first replaced by fresh serum-free medium with or without 100 nM 6-22 (PKA inhibitor) as appropriate, and the cells were incubated at 37°C for 1 hour. After 2 min of stimulation with vehicle or 1 μM isoproterenol, the cells were lysed at 4°C in the lysis buffer provided by the manufacturer and frozen overnight at −20°C. Thereafter, 10 μl of thawed lysates was transferred to a 384-well white AlphaPlate, and ERK1/2 activity was measured according to the manufacturer’s protocol using an EnVision Multilabel plate reader (PerkinElmer Life Sciences). For each experiment, stimulated ERK1/2 values were normalized to isoproterenol-stimulated controls.

Measurement of cell surface β2AR abundance by ELISA

For BRET and SureFire ERK1/2 phosphorylation assays, the equivalence of HA-β2AR expression in parental HEK293 and CRISPR βArr1/2 KO cells was determined by ELISA. For this assay, transfected cells in poly-d-lysine–coated, 96-well plates were fixed for 10 min in DPBS containing 3.7% formaldehyde. Cells were then washed three times with wash buffer (0.5% BSA in DPBS) and incubated for 1 hour with a 1:2000 dilution of HRP-conjugated anti-HA IgG in wash buffer. After three additional washes, the number of cells was measured by incubation for 30 min with 5 μM Vybrant DyeCycle orange stain in wash buffer. Cells were again washed three times with DPBS, after which fluorescence was read at 563 nm with 519-nm excitation using a FlexStation II microplate reader (Molecular Devices). HRP activity, reflecting receptor number, was detected by incubating cells with o-phenylenediamine dihydrochloride. The addition of 0.6 M HCl stopped the reaction, and the absorbance was read at 492 nm using a Tecan Infinite M200 PRO multifunction microplate reader (Tecan Inc.).

Statistical analysis

To foster unbiased data collection and compare reproducibility between laboratories using different approaches to transfection and stimulation, the three independently derived HEK293 parental/CRISPR βArr1/2 KO pairs were shared between the participating laboratories, each of which focused on a single GPCR using the common model systems. Each laboratory performed experiments independently using the protocols, reagents, and tools routinely used by that laboratory to investigate GPCR signaling. All of the raw data were subsequently collated at Duke University Medical Center and analyzed using consistent methods. Statistical analyses were performed using GraphPad Prism 7 software (GraphPad Software). The data were analyzed by two-way ANOVA with Sidak’s multiple comparison test. The mean differences of a minimum of three independent experiments were considered to be statistically significant when P < 0.05.

REFERENCES AND NOTES

Acknowledgments: We thank Q. Lennon and J. Bisson for secretarial help and W. Zou for excellent technical assistance. The AI-CRISPR βArr1/2 KO and corresponding AI-parental HEK293 cell lines used in this study were provided by A. Inoue (Tohoku University, Sendai, Miyagi, Japan). Funding: This work was supported in part by the NIH (grants R01 HL16037 to R.J.L., R01 DK055524 and R35 GM126955 to L.M.L., R01 HL056687 and P01 HL075443 to H.A.R., K08 HL114643 and R01 GM122798 to S.R., and R01 HL118369 to S.K.S.), Canadian Institutes of Health Foundation (grant 14843 to M.B.), and Canadian Institutes of Health Research (grant MOP-74603 to S.A.L.). E.R. is funded by the French National Research Agency Investissements d’avenir program, Grant Agreement LabEx MabImprove: ANR-10-LABX-53, and the Région Centre Val de Loire ARD2020 Biomédicaments program. L.Y. holds a postdoctoral internship from Mitacs Accelerate Canada. B. Plouffe holds a postdoctoral fellowship from Diabetes Canada. S.R. holds a Burroughs Welcome Career Award for medical scientists. M.B. holds the Canada Research Chair in Signal Transduction and Molecular Pharmacology. R.J.L. is an investigator with the Howard Hughes Medical Institute. Author contributions: M.B., S.A.L., R.J.L., L.M.L., E.R., S.R., H.A.R., and S.K.S. conceived and designed all experiments, analyzed the data, wrote the manuscript, and contributed funding for experiments. S.A., C.G., P.-Y.J.-C., M.-H.L., B. Pani, J.K., S.K., E.R., J.W., and L.Y. performed MAPK activation assays, analyzed the data, and wrote the manuscript. B. Pani performed radioligand binding assays. B. Plouffe performed BRET, FRET, and MAPK activation assays; analyzed the data; and wrote the manuscript. J.S.S. performed MAPK activation assays, collated data from all groups, prepared figures, and wrote the manuscript. Competing interests: M.B. is a member of the Scientific Advisory Board of Domain Therapeutics, a company marketing platforms for GPCR-based drug discovery. The BRET-based biosensors used in this study have been licensed to Domain Therapeutics but are available from the Bouvier laboratory free of charge for noncommercial academic use. R.J.L. and H.A.R. are founders of Trevena Inc., a company that discovers and develops novel GPCR-targeted therapeutics. The other authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions of this study are available in the paper.
View Abstract

Navigate This Article