β-Arrestin-1 inhibits glucocorticoid receptor turnover and alters glucocorticoid signaling

Glucocorticoids are among the most widely used drugs to treat many autoimmune and inflammatory diseases. Although much research has been focused on investigating glucocorticoid activity, it remains unclear how glucocorticoids regulate distinct processes in different cells. Glucocorticoids exert their effects through the glucocorticoid receptor (GR), which, upon glucocorticoid binding, interacts with regulatory proteins, affecting its activity and function. These protein–protein interactions are necessary for the resolution of glucocorticoid-dependent physiological and pharmacological processes. In this study, we discovered a novel protein interaction between the glucocorticoid receptor and β-arrestin-1, a scaffold protein with a well-established role in G protein–coupled receptor signaling. Using co-immunoprecipitation and in situ proximity ligation assays in A549 cells, we observed that β-arrestin-1 and unliganded GR interact in the cytoplasm and that, following glucocorticoid binding, the protein complex is found in the nucleus. We show that siRNA-mediated β-arrestin-1 knockdown alters GR protein turnover by up-regulating the E3 ubiquitin ligase Pellino-1, which catalyzes GR ubiquitination and thereby marks the receptor for proteasomal degradation. The enhanced GR turnover observed in β-arrestin-1–deficient cells limits the duration of the glucocorticoid response on GR target genes. These results demonstrate that β-arrestin-1 is a crucial player for the stability of the glucocorticoid receptor. The GR/β-arrestin-1 interaction uncovered here may help unravel mechanisms that contribute to the cell type–specific activities of glucocorticoids.

Glucocorticoids (GCs) 2 are lifesaving drugs that are widely prescribed in the treatment of inflammatory diseases and other conditions requiring suppression of the immune system (1). These steroids are physiologically synthesized in the zona fasciculata of the adrenal cortex as end products of the hypothalamic-pituitary-adrenal axis in response to a variety of stressproducing stimuli. Both physiological and pharmacological actions of glucocorticoids occur through the binding to the glucocorticoid receptor (GR; gene ID: NR3C1, nuclear receptor subfamily 3, group C, member 1), a transcription factor that, like other members of the nuclear receptor family, functions as a ligand-activated transcriptional gene expression regulator (2). Unliganded GR is located in the cytoplasm associated in a heterocomplex with chaperone proteins that favor GR maturation and protect the receptor from its degradation (3,4). Upon steroid binding, GR undergoes activation, dissociates from the chaperone complex, and exerts its effects via nongenomic and genomic mechanisms (5). The latter requires the translocation of the receptor to the nucleus, where it can bind either to specific nucleotide sequences on DNA called glucocorticoid-responsive elements (GREs), or it can involve the tethering of GR with other transcription factors (1). Throughout its intracellular journey, liganded GR creates physical contacts with a multitude of regulatory proteins, thus affecting its activity and function (6).
␤-Arrestin-1 (gene ID: ARRB1) and its related family member ␤-arrestin-2 (gene ID: ARRB2) were first described as negative regulators of G protein-coupled receptors (GPCRs) (7). However, ␤-arrestin proteins are now known to exert other actions that go beyond their conventional role of GPCR terminators. Indeed, ␤-arrestins have been reported to take part in a variety of functions, including protein trafficking, protein subcellular redistribution, transcriptional regulation, and protein post-translational modifications, thus affecting cell proliferation, differentiation, and apoptosis signaling (8).
We have previously shown that both ␤-arrestin-1 and ␤-arrestin-2 are glucocorticoid-responsive genes (9). Glucocorticoid treatment induced ␤-arrestin-1 and repressed ␤-arrestin-2 expression, suggesting that a cross-talk between the glucocorticoid receptor and the ␤-arrestin proteins may occur. ␤-Arrestin proteins have been reported to interact with other members of the nuclear receptor superfamily, such as the androgen receptor (AR) (10). Furthermore, ␤-arrestin-1 contributes to internalizing the membrane-bound estrogen recep- cro ARTICLE tor after its activation (11). Herein, we report a novel protein partnership between ␤-arrestin-1 and the glucocorticoid receptor in the A549 lung adenocarcinoma cell line. We found that ␤-arrestin-1 regulates the process of glucocorticoid-induced GR down-regulation by protecting the receptor from enhanced degradation. Surprisingly, the effects of ␤-arrestin-1 on GR are associated with GR transcriptional activity. Mechanistically, glucocorticoid exposure in ␤-arrestin-1 knockdown cells induces the transcription of PELI1, which encodes an E3 ubiquitin ligase responsible for the enhanced turnover of the glucocorticoid receptor.

␤-Arrestin-1 associates in a complex with the glucocorticoid receptor
␤-Arrestin-1 has been shown to be a scaffolding protein that brings numerous proteins together to promote their concerted interactions (12)(13)(14). Moreover, existing evidence suggests that ␤-arrestin-1 acts as a modulator of a variety of cellular processes, such as proliferation, differentiation, and apoptosis, that are also widely known to be regulated by the glucocorticoid receptor (15,16). Therefore, we queried whether ␤-arrestin-1 could associate with GR by performing a co-immunoprecipitation assay in A549 lung adenocarcinoma cells, which were chosen as a classic glucocorticoid-responsive cell line. In the absence of its ligand, the GR co-immunoprecipitated with ␤-arrestin-1, and the association persisted upon 1-and 3-h treatment with a 100 nM concentration of the synthetic glucocorticoid dexamethasone (Dex) (Fig. 1A). To examine the subcellular distribution of GR/␤-arrestin-1 complexes, we next immunoprecipitated ␤-arrestin-1 in both cytoplasmic and nuclear fractions from A549 cells that were treated with or without Dex for 1 and 3 h. The results shown in Fig. 1B demonstrate that, in the absence of glucocorticoids, GR/␤-arrestin-1 complexes were detected in the cytoplasm, whereas in the presence of Dex, the GR/␤-arrestin-1 complexes were detected in the nucleus. To further analyze the existence of GR/␤-arrestin-1 association, we performed an in situ proximity ligation assay (PLA) (17). In the absence of glucocorticoids, the GR/␤arrestin-1 dimerization signals were observed in the cytoplasm, and, upon Dex treatment, we observed a loss of cytoplasmic signals that was accompanied by detection of fluorescent signals in the nuclear compartment (Fig. 1C). These data establish an association between GR and ␤-arrestin-1 that takes place in the cytoplasm and persists even after hormone-induced GR nuclear translocation.

Silencing of endogenous ␤-arrestin-1 alters the glucocorticoid receptor stability
To evaluate the potential impact of ␤-arrestin-1 interaction on the GR function, we transfected A549 cells with ␤-arrestin-1 siRNA to silence ␤-arrestin-1 expression. Forty-eight hours after siRNA transfection, ␤-arrestin-1 mRNA expression was reduced by ϳ80% compared with cells transfected with the nontargeting control (NTC) siRNA ( Fig. 2A). ␤-Arrestin-1 knockdown was also efficient at the protein levels. Indeed, Fig.  2B shows that the well-characterized Dex-induced up-regulation of ␤-arrestin-1 (9) was abolished in ␤-arrestin-1 knockdown (ARRB1-KD) cells up to 48 h after steroid exposure. Dexamethasone treatment led to a decrease of GR levels in a Figure 1. The glucocorticoid receptor associates with ␤-arrestin-1. A, A549 cells were untreated or treated with Dex (100 nM) for 1 and 3 h. Subsequently, whole-cell lysates were immunoprecipitated (IP) with an anti-␤-arrestin-1 antibody, and recovered proteins were immunoblotted for GR. Levels of GR, ␤-arrestin-1 (␤-arr-1), tubulin, and lamin A are shown below the immunoprecipitation blot. Lysates were also immunoprecipitated with goat anti-rabbit IgG as negative control. B, protein lysates from cells untreated or treated with Dex (100 nM) for 1 and 3 h were subjected to subcellular fractionation. Cytosolic and nuclear extracts fractions were then immunoprecipitated with an anti-␤-arrestin-1 antibody, and recovered proteins were immunoblotted for GR. Cytosolic and nuclear lysates were immunoblotted to detect GR and ␤-arrestin-1 subcellular distribution along with tubulin (cytosolic marker) and lamin A (nuclear marker). C, representative confocal microscopic images of the in situ PLA showing the interaction between GR and ␤-arrestin-1 in A549 cells. Each red dot represents the detection of protein-protein interaction complexes in cells treated with Dex (100 nM) for 0, 1, and 3 h. Specificity of the assay was shown by lack of red signal in the negative control, displayed using mouse GR antibody alone with the PLA minus and plus probes. Scale bar, 20 m. Data are for a representative experiment from three independent experiments. ␤-Arrestin-1 modulates the glucocorticoid receptor turnover time-dependent fashion in control cells (NTC) (18). Interestingly, the process of GR protein down-regulation was significantly enhanced when ␤-arrestin-1 was knocked down (Fig. 2, B and C). In NTC cells, Dex repressed the expression of GR protein by ϳ50% after 6 h and remained relatively stable over the 48-h treatment. In ARRB1-KD cells, GR was also down-regulated by dexamethasone in a time-dependent manner but to a greater extent that reached ϳ75% protein reduction over a 48-h period (Fig. 2C). Glucocorticoid-induced GR down-regulation was also enhanced in mouse embryonic fibroblasts (MEFs) with targeted deletion of ␤-arrestin-1 (Fig. 2, D and E) (19), suggesting that GR turnover is affected by ␤-arrestin-1 expression in multiple cell types. Because the difference in GR down-regulation observed in control and ␤-arrestin-1 KD cells occurred in the late phase of Dex exposure, we speculated that ␤-arrestin-1 may affect GR protein stability. To evaluate the GR protein ␤-Arrestin-1 modulates the glucocorticoid receptor turnover turnover over time, cells transfected with NTC and ARRB1 siRNAs were treated with dexamethasone for 2 h prior to the addition of the protein synthesis inhibitor cycloheximide (CHX; 50 M). Compared with control, we observed that GR half-life was reduced from 12.7 to 6.9 h in ␤-arrestin-1 KD cells (Fig. 2, F and G).
Regulation of GR expression is controlled at both transcriptional and post-transcriptional levels, two key mechanisms involved in the regulation of the amount of active GR protein in cells (20,21). The glucocorticoid receptor is primarily degraded via activation of the ubiquitin-proteasome machinery (22). Based on the effects that ␤-arrestin-1 exerts on GR turnover, we investigated whether ␤-arrestin-1 affects GR degradation via engaging ubiquitin and proteasome pathways. A549 cells transfected with NTC and ␤-arrestin-1 siRNAs were then treated with the proteasome inhibitor MG132 (10 M) in the presence or absence of Dex (100 nM) for 6 h, and ubiquitination of GR was evaluated by using the proximity ligation assay. We carried out PLA using an antibody directed to GR and a second antibody directed to ubiquitin, and we examined the PLA signals per cell to determine the amount of endogenous ubiquitinated GR. In untreated cells, knockdown of ␤-arrestin-1 did not affect the basal level of GR ubiquitination. However, silencing of ␤-arrestin-1 increased the content of GR that undergoes ubiquitination following GR activation by Dex (Fig. 2, H and I). It is well-characterized that monoubiquitination regulates protein trafficking, whereas polyubiquitination drives the target protein to proteasomal degradation (23). To assess the type of ubiquitination that marks GR, we performed an in vivo ubiquitination assay using control and ␤-arrestin-1 KD cells. Twenty-four hours after NTC and ARRB1 siRNA transfection, the experimental groups were transfected with HA-ubiquitin for 24 h and then treated for 6 h with or without Dex under condition of proteasome inhibition. Protein lysates were subjected to GR immunoprecipitation followed by immunoblotting with anti-HA antibodies. MG132 treatment alone did not show GR polyubiquitination either in NTC or ARRB1-KD cells. However, GR activation by Dex induced GR polyubiquitination in NTC cells, which was enhanced in the ␤-arrestin-1 KD precipitates (Fig. 2J). These data indicate that ␤-arrestin-1 stabilizes the glucocorticoid receptor by protecting it against an excessive degradation induced by glucocorticoid binding.

␤-Arrestin-1 influences the duration of the GR transcriptional response
GR activation by glucocorticoids results in changes in gene expression levels, with some genes being induced and others repressed (24,25). We hypothesized that the enhanced GR turnover occurring in ␤-arrestin-1deficient cells might affect the glucocorticoid responsiveness by limiting GR-mediated transcriptional activity. To explore this possibility, we evaluated the effects of endogenous ␤-arrestin-1 on GR-mediated transcription. We examined the expression of three known GR target genes TSC22D3 (26), IL-1␤ (27), and CXCL5 (28) in NTC and ␤-arrestin-1 KD A549 cells treated with dexamethasone at different time points over 48 h. As shown in Fig. 3, the duration of Dex-induced activation of TSC22D3 and Dex-induced inhibition of IL-1␤ and CXCL5 genes was attenuated. These data indicate that ␤-arrestin-1, by preventing enhanced GR turnover, plays a regulatory role in the duration of the glucocorticoid gene regulation.

PELI1 as a candidate for ␤-arrestin-1dependent GR turnover
To identify the specific E3 ubiquitin ligase required for ubiquitin-dependent GR degradation in ␤-arrestin-1 KD cells, we screened some of the HECT (homologous to the E6-AP C terminus) and RING-finger E3 ligases that have been implicated in ubiquitination of nuclear steroid receptors (29 -32). We evaluated gene transcription in both control and ␤-arrestin-1 KD cells treated with or without Dex for 3 h. It is noteworthy that knockdown of endogenous ␤-arrestin-1 was efficient at both mRNA and protein levels (Fig. 4A), and we chose the 3-h Dex treatment because at this time point knockdown of endogenous ␤-arrestin-1 did not affect GR mRNA, protein levels ( Fig.  4B), or its subcellular localization (Fig. 4C). Our data showed that PELI1 is the only E3 ubiquitin ligase gene induced by Dex exclusively when ␤-arrestin-1 expression is abolished in A549 cells (Fig. 4D). PELI1 (Entrez Gene: 57162) encodes for the protein E3 ubiquitin-protein ligase Pellino homolog-1 (shorter name: Pellino-1), a 47-kDa protein that, along with its two isoforms Pellino-2 and Pellino-3 (33,34), contains a conserved RING-like domain at the C terminus that confers ubiquitin E3 ligase activity (35). Pellino-1 is emerging as an important component in inflammation, autoimmunity, and tumorigenesis by modulating signaling pathways that elicit inflammatory responses. Indeed, there is a positive correlation between its expression and grade of inflammation (36 -41). For this reason, ␤-Arrestin-1 modulates the glucocorticoid receptor turnover we decided to investigate whether Pellino-1 could play a role in the glucocorticoid receptor degradation enhanced when ␤-arrestin-1 is knocked down. Expression of PELI1 mRNA did not change in NTC cells, either in the presence or absence of glucocorticoids. Interestingly, cells that were suppressed for ␤-arrestin-1 expression had an induction of PELI1 expression ϳ2.3fold after Dex treatment, and PELI1 up-regulation persisted up to 6 h upon Dex exposure, thereafter gradually returning to baseline levels (Fig. 5A). We next evaluated whether the altera-tion in PELI1 mRNA led to corresponding changes in Pellino-1 protein expression. Pellino-1 was expressed at basal levels in NTC cells, and Dex exposure induced its down-regulation in a time-dependent fashion, reaching ϳ50% reduction in 24 h. However, glucocorticoid treatment in ␤-arrestin-1 KD cells sustained Pellino-1 expression (Fig. 5, B and C). These results suggest that the PELI1 gene is a strong candidate for mediating the ␤-arrestin-1 regulation of GR turnover in GCtreated cells. 5A shows that PELI1 mRNA levels were induced within 3 h after Dex treatment in ARRB1-KD cells, suggesting that PELI1 might undergo transcriptional regulation by GR. We analyzed PELI1 nascent RNA following glucocorticoid exposure. Using a primer/probe set spanning an exon-intron boundary, we observed that, compared with NTC cells, Dex promoted an increase in PELI1 nascent RNA transcription (ϳ3-and 4-fold induction after 30 and 60 min, respectively) in ␤-arrestin-1 KD cells, suggesting that only in ␤-arrestin-1 KD cells PELI1 behaves as a primary target gene (Fig. 6A). Moreover, treatment with the glucocorticoid receptor antagonist RU-486 (1 M) prevented the glucocorticoid-mediated induction of PELI1 in ARRB1-KD cells at both mRNA and protein levels (Fig. 6, B and C), indicating that PELI1 is regulated by glucocorticoids. Hormone-bound GR can bind to DNA via palindromic GREs to modulate gene transcription (42)(43)(44)(45)(46). Therefore, we examined the PELI1 gene for putative GRE-binding sites. Sequence analysis disclosed two putative GREs (GRE#1 and GRE#2) that displayed high homology with the consensus GRE sequence (Fig. 6D). Both GRE#1 and GRE#2 are located upstream of the transcription start site of the human PELI1 locus (Ϫ30 and Ϫ8 kb, respectively; Fig. 6D).

PELI1 is a novel GR-responsive gene up-regulated in
Moreover, the GRE#1 sequence appeared to be functional and highly conserved in the mouse genome, and GR recruitment to this regulatory site was associated with induction of Peli1 transcription (47). ChIP assays showed that both GRE putative sites were recognized by GR in a ligand-dependent manner only in A549 cells knocked down for ␤-arrestin-1 (Fig.  6E). To investigate whether there is a differential specificity of GR binding to the GREs in the PELI1 upstream region of the transcription start site, we analyzed the recruitment of activated GR to the GRE located in the proximal promoter of the primary glucocorticoid target gene TSC22D3 (48). We detected a significant GR occupancy to the regulatory sequence of TSC22D3 promoter in both NTC and ␤-arrestin-1 KD cells following Dex treatment (Fig. 6E). These data suggest that PELI1 is a novel glucocorticoid-responsive gene regulated by glucocorticoids extensively following the knockdown of ␤-arrestin-1. These findings demonstrate that the differential regulation of PELI1 depends on ␤-arrestin-1 expression, leading us to hypothesize that ␤-arrestin-1 might regulate the chromosome architecture by allowing, or inhibiting, transcription factors to bind to DNA (49,50). To assess whether ␤-arrestin-1 maintains the GR-responsive region of PELI1 promoter inaccessible to GR binding, we performed a formaldehyde-assisted isolation of regulatory elements (FAIRE) assay. On control and ␤-Arrestin-1 modulates the glucocorticoid receptor turnover ␤-arrestin-1 KD cells, FAIRE-enriched DNA was analyzed by quantitative real-time PCR using the same primer/probe sets designed to detect PELI1 GRE#1 and GRE#2 elements by ChIP. The FAIRE induction seen at the two responsive GR sites in ␤-arrestin-1 KD cells suggests that there is an accessible chromatin environment that may facilitate the recruitment of GR to DNA only in the absence of ␤-arrestin-1 (Fig. 6F). As a control, no difference was found in DNA accessibility at the GRE element in the TSC22D3 locus between NTC and ␤-arrestin-1 KD cells (Fig. 6F).

Knockdown of ␤-arrestin-1 allows Pellino-1 to interact with GR and promotes GR ubiquitination
To evaluate the functional consequence of PELI1 transcriptional regulation in a system lacking ␤-arrestin-1, we performed co-immunoprecipitation experiments to verify whether Pellino-1 associates with GR. A549 cells, transfected with siRNAs for NTC and ␤-arrestin-1 (ARRB1-KD) and co-transfected with siRNAs against ␤-arrestin-1 and PELI1 double knockdown (DKD), were treated with or without Dex for 6 h in the presence of MG132, which inhibits protein proteasomal degradation (Fig. ␤-Arrestin-1 modulates the glucocorticoid receptor turnover 7A, Total lysates). Western blot analysis of endogenous Pellino-1 immunoprecipitation revealed that endogenous GR was readily detected in Pellino-1-associated immunoprecipitates only in cell extracts from ARRB1-KD group treated with dexamethasone (Fig. 7A). Importantly, GR was immunoprecipitated with the anti-Pellino-1 antibody but not with the control IgG antibody (Fig. 7A). These findings raised the possibility that the E3 ubiquitin ligase Pellino-1 can bind to GR only in a context of ␤-arrestin-1 deficiency, suggesting that Pellino-1 could be the "executor" of GR down-regulation. To analyze this question, we evaluated Dex-induced GR down-regulation in control cells (NTC) and cells that lack ␤-arrestin-1 (ARRB1-KD), Pellino-1 (PELI1-KD), or both (DKD) at different time points after glu-cocorticoid treatment. Fig. 7B shows a representative Western blotting, where GR undergoes down-regulation in the four different experimental groups. In NTC cells, Dex-induced GR down-regulation resulted in a GR reduction of ϳ70% after 48-h treatment. Interestingly, PELI1-KD cells revealed a GR degradation kinetics comparable with that in NTC cells (Fig. 7, B  (lanes 13-18 versus lanes 1-6) and C). Reduction in PELI1 expression in cells that are also knocked down for ␤-arrestin-1 (DKD) attenuated the augmented GR down-regulation that occurred in ␤-arrestin-1 KD cells (Fig. 7, B (lanes 19 -24 versus  lanes 7-12) and C). These findings demonstrate that, when ␤-arrestin-1 expression is reduced, up-regulation of Pellino-1 is crucial for GR protein stability.
Pellino-1 has an intrinsic E3 ubiquitin ligase activity that mediates Lys-48 -and Lys-63-linked polyubiquitination, which marks target proteins to proteasomal degradation or stabilization, respectively (34,(52)(53)(54). To add support to the aspect of Pellino-1 in GR ubiquitination, we analyzed its possible involvement in GR degradation by measuring the amount of Lys-48 -linked ubiquitin chains on GR. Using the in situ proximity ligation assay with antibodies recognizing GR and the Lys-48 linkage-specific polyubiquitin chain, we observed an increased ubiquitination signal in Dex-treated ARRB1-KD cells that was ϳ2-fold higher than NTC and PELI1-KD cells, suggesting that the Pellino-1 bound to GR induced the ubiquitination and degradation of the receptor (Fig. 7D). Conversely, the enhanced Lys-48 -linked polyubiquitination of GR decreased in cells where the expression of PELI1 and ␤-arrestin-1 was silenced (Fig. 7D). Collectively, these data indicate that in the absence of ␤-arrestin-1, Pellino-1 expression is sustained; thereby, it binds to activated GR and is responsible for its enhanced polyubiquitination and further proteasomal degradation.

Discussion
The glucocorticoid receptor was described for the very first time in the 1960s (55). Since then, it has been found in almost all cells and tissues and mediates the biological and physiological actions of glucocorticoids either under normal or challenging circumstances. Belonging to the nuclear receptor superfamily, GR is a ligand-dependent transcription factor that acts mainly in the nucleus, where it regulates transcription of a multitude of target genes. Indeed, hormone-bound GR, by regulating cellular mechanisms including cell proliferation, differentiation, and apoptosis, acts as a "bodyguard" ensuring continuous systemic immune surveillance and body homeostasis. However, impairments at any level in the receptor regulation might lead to GR dysfunction. An impaired GR response is linked to polymorphisms of the gene, expression of a specific GR isoform over another, GR reduced expression, altered binding affinity, abnormal nuclear translocation, chromatin reshaping, and GR post-translational modifications. These processes are all key components of the glucocorticoid resistance (56). Moreover, dysfunctions, as well as actions, of the glucocorticoid receptor may be the result of the cross-talk between the receptor and an array of elements, such as co-modulators, co-activators, corepressors, and DNA-remodeling factors (6). For example, some molecules, such as MDFIC (MyoD inhibitor domain- ␤-Arrestin-1 modulates the glucocorticoid receptor turnover containing), can interact with and influence GR conformation by altering the phosphorylation status of the receptor, which in turn reshapes its gene transcription program (57,58). Other co-modulators affect unliganded GR, working to stabilize (59) or reduce (60) the fraction of ligand-bound GR.
Here, we report the identification of the versatile adapter ␤-arrestin-1 protein as a novel GR regulator. We demonstrate for the first time that ␤-arrestin-1 associates in the cytoplasm with unliganded GR, and, upon hormone binding, the complex is detected in the nucleus, where GR can regulate gene tran-scription. Surprisingly, deletion of endogenous ␤-arrestin-1 via siRNA or gene knockout enhances GR turnover, suggesting that the association of ␤-arrestin-1 with GR protects the receptor from degradation.
The glucocorticoid receptor is subjected to numerous posttranslational modifications, which represent a crucial mechanism in the regulation of GR signaling. The best-characterized GR post-translational modification is phosphorylation, mostly because GR has been shown to be a phosphoprotein (57, 61-68). Conversely, less is known about GR ubiquitination. The table shows the position from TSS and the sequence of the two GREs. E, the two identified GR-responsive regions on the PELI1 gene and human TSC22D3 GRE (as control) were tested for GR recruitment in NTC or ARRB1-KD cells by ChIP analysis following 2-h Dex treatment (100 nM). Samples were analyzed by quantitative real-time PCR, and the graphs show fold changes of GR recruitment compared with IgG controls. F, chromatin-remodeling events were measured by FAIRE to evaluate GR accessibility to the GR-responsive regions of PELI1 gene and TSC22D3 gene (as control) in response to knockdown of ␤-arrestin-1. For each GRE, chromatin accessibility was determined by the ⌬⌬Ct method using sonicated input DNA to normalize the FAIRE DNA and expressing this ratio relative to the NTCs that were set equal to 1. Data represent mean Ϯ S.D. (error bars) from three, four, or six independent experiments. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001 for Dex versus control. ##, p Ͻ 0.01; ###, p Ͻ 0.001; ####, p Ͻ 0.0001 to compare ARRB1-KD to NTC.

␤-Arrestin-1 modulates the glucocorticoid receptor turnover
Glucocorticoid-induced GR down-regulation is a classic response occurring in most cells, that is necessary to limit the duration of glucocorticoid action. An excess of this response contributes to glucocorticoid resistance (69). The first observation of ubiquitin-mediated degradation of GR occurred in COS-1 cells when treatment with the proteasome inhibitor MG132 inhibited Dex-induced down-regulation of GR (24).
Seeking the biological meaning of the GR/␤-arrestin-1 protein complex, we found that loss of endogenous ␤-arrestin-1 expression affected GR protein turnover. In A549 cells knocked down for ␤-arrestin-1, GR half-life was shorter along with an ␤-Arrestin-1 modulates the glucocorticoid receptor turnover increase in receptor ubiquitination and degradation via proteasome. The resultant changes in GR turnover limit the duration of GR transcription activity on target genes, suggesting that alteration in ␤-arrestin-1 expression may influence cell typespecific responsiveness to glucocorticoids.
Beyond the very conventional role of G protein-coupled receptor negative modulators, ␤-arrestin-1, as well as ␤-arrestin-2, also participates in the ubiquitination process orchestrating binding specificity and defining timing and dynamics. ␤-Arrestins can act either as adaptors that link a wide array of proteins (GPCRs and non-GPCRs) with their cognate E3 ubiquitin ligases, or they can interfere with the E3 ligase/substrate binding, preventing protein ubiquitination (73). For example, ␤-arrestin-1 brings the tyrosine kinase receptor IGF-1 close to the E3 ubiquitin ligase MDM2 for its ubiquitination and protein degradation (74). A similar process happens to another member of the nuclear receptor superfamily; the AR has ␤-arrestin-2 as one of its co-repressors, which impacts AR-dependent gene expression because it scaffolds MDM2 for androgen receptor ubiquitination and proteasomal degradation (10). Conversely, ␤-arrestin-1 inhibits degradation of ubiquitinated CXCR4 by interacting with signal-transducing adaptor molecule (STAM)-1 (75) but also protects the insulin-dependent IRS-1 from ubiquitination and degradation by inhibiting the formation of IRS-1/MDM2 complexes (76). Our findings suggest that ␤-arrestin-1 protects GR from an enhanced glucocorticoid-induced protein down-regulation, allowing the receptor to remain activated and accomplish its gene regulation task. Our data suggest that the protective role of ␤-arrestin-1 is due to the differential transcriptional regulation of PELI1 in control and ␤-arrestin-1 KD cells. PELI1 is expressed in control cells, and its transcription is independent of GR activation. However, loss of ␤-arrestin-1 expression makes PELI1 become a glucocorticoidresponsive gene. Indeed, we found two functional GR regulatory regions that in ␤-arrestin-1 KD cells are accessible and involved in the binding with activated GR. As a result, there is an induction of PELI1 mRNA levels which sustains the levels of Pellino-1 protein in a context of ␤-arrestin-1 deficiency. An alternative hypothesis could be that, similarly to IRS-1, ␤-arrestin-1, by binding to GR, competes with Pellino-1 and/or other components and impedes GR from becoming a Pellino-1 substrate. Indeed, we provide that in a context of reduced ␤-arrestin-1 expression, Pellino-1 associates with GR causing an increase in ubiquitination of the receptor (Fig. 7, A and D). In conclusion, whether GR degradation is limited due to a prevention of GR-induced PELI1 transcription, or due to a competitive binding between ␤-arrestin-1 and Pellino-1 to GR is currently unknown but is an important question to be addressed in future studies.
Because of the divergent roles exerted by ␤-arrestins, it is complicated to extrapolate how they can contribute to the physiology, and pathophysiology, of the immune system. Loss of ␤-arrestin-1 function favors decreased migration in lymphocytes but promotes the opposite phenotype in neutrophils (77). In a situation involving viral infection, ␤-arrestins hamper the inflammatory axis and virus removal; in the setting of autoimmune disease, ␤-arrestins protect against rheumatoid arthritis exacerbation because they negatively modulate Th1 (T helper 1)-associated cytokine secretion (78).
The GPCR-dependent and -independent roles of ␤-arrestins are essential for multiple aspects of immune cell function. Being regulators of ␤ 2 -adrenergic receptors, ␤-arrestins modulate leukocyte and lymphocyte response by altering the sensitivity and the expression of critical cell surface receptors. In the context of immune function, many studies have shown that ␤-arrestins promote desensitization and internalization of chemokine receptors (CXCR family) and CCR5 (C-C chemokine receptor type 5), hence modulating neutrophil, monocyte, and T lymphocyte response. Despite the fact that ␤-arrestins play a pivotal role in several immune cell processes, it is still early to draw conclusions about the role of ␤-arrestins in inflammation. Our discovery suggests that ␤-arrestin-1 acts as a modulator of the anti-inflammatory effects of glucocorticoids, because it exerts a protective role in stabilizing the glucocorticoid receptor protein. In a recent study by Zhang et al. (79), there is evidence that ␤-arrestin-1 and ␤-arrestin-2 double KO mice die soon after birth due to pulmonary hypoplasia and hepatic impairment. They found that GR protein levels were significantly decreased compared with WT controls, concluding that GR might be a potential downstream effector involved in ␤-arrestin-mediated signaling pathway during development (79).
In conclusion, we define ␤-arrestin-1 as a regulator of the glucocorticoid receptor. ␤-Arrestin-1 is a novel GR-binding partner that stabilizes GR protein expression. Loss of ␤-arrestin-1 compromises GR protein turnover by allowing liganded GR to induce PELI1 transcription, the gene encoding for the E3 ubiquitin ligase Pellino-1 responsible for enhanced GR degradation. These findings reveal a new mechanism by which the actions of glucocorticoids can be regulated in a cell typespecific manner.

Cell culture
Mycoplasma-free human lung adenocarcinoma A549 cells were obtained from ATCC and maintained in Dulbecco's modified Eagle's medium/nutrient mixture F-12 (Gibco) supplemented with 10% fetal bovine serum, 50 units/ml penicillin, 50 g/ml streptomycin, and 2 mM L-glutamine. Mycoplasma-free mouse embryonic fibroblasts, kindly provided by Dr. Robert J. Lefkowitz (Duke University), were maintained in Dulbecco's modified Eagle's medium/high glucose (Gibco) with 10% fetal calf serum, 50 units/ml penicillin, 50 g/ml streptomycin, and 2 mM L-glutamine. To evaluate response to dexamethasone, cell lines were grown overnight in their respective media supplemented with 10% charcoal-stripped fetal bovine serum, 50 units/ml penicillin, 50 g/ml streptomycin, and 2 mM L-glutamine, to deplete steroids, and then treated with vehicle (H 2 O) or 100 nM dexamethasone for the indicated time periods. Cell cultures were kept in an incubator at 37°C in a humidified atmosphere with 5% CO 2 .

siRNA experiments
A549 cells were reseeded in complete growth medium for 24 h prior to transfection (50% confluence). Then cells were cultured in serum-free medium and transfected with 30 nM NTC siRNA or siRNAs targeting ␤-arrestin-1 and/or PELI1, using Dharmafect-1 transfection reagent (Dharmacon-Horizon Discovery, Lafayette, CO) according to the manufacturer's instructions. Twenty-four hours after transfection, transfection medium was replaced, and cells were reseeded with complete medium in tissue culture dishes appropriated for the type of assay performed.
To assess the amount of DNA contamination present in the RNA preparation, a no reverse transcriptase control was included as a negative control. PELI1 nascent RNA levels were quantified relative to the housekeeping gene PPIB using the ⌬⌬Ct method.

Western blotting
For protein expression experiments, the cycloheximide chase assay, and proteasome inhibition experiments, A549 cells were lysed in radioimmune precipitation assay buffer containing protease inhibitor mixture (MilliporeSigma), phosphatase inhibitor mixture (MilliporeSigma), and 1 mM PMSF (Milli-poreSigma). To separate nuclear from cytoplasmic proteins, cells were lysed using the NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher Scientific), following the manufacturer's instructions. The primary antibodies used were anti-GR (D8H2, Cell Signaling Technology), anti-␤-arrestin-1 (D7Z3W, Cell Signaling Technology), anti-Pellino-1 (D2Z4F, Cell Signaling Technology), anti-actin (clone C4, EMD Millipore), anti-lamin A (133A2, Abcam), and anti-tubulin (DM1A, Invitrogen). The secondary antibodies used were IRDye 800CW goat anti-mouse (LI-COR Biosciences, Lincoln, NE) and AF680 goat anti-rabbit (Invitrogen). The immune blots were developed in the linear dynamic range using the LI-COR Odyssey imaging system from LI-COR Biosciences. The specificity of the antibody for ␤-arrestin-1 was confirmed by siRNA-mediated knockdown of ␤-arrestin-1 and by using ARRB1-KO MEF cells. The specificity of the antibody for Pellino-1 was confirmed by siRNA-mediated knockdown of Pellino-1 protein. The specificity of the antibody for GR was already confirmed by siRNA-mediated knockdown of GR in cells (9) and in mouse tissue (80).

Immunoprecipitation assay
For analysis of complex formation in nuclear and cytoplasmic extracts, the above-mentioned NE-PER reagents were used. For analysis of complex formation from whole-cell extracts, a buffer suited for immunoprecipitation was used (81). In both cases, 1000 g of lysates were incubated overnight at 4°C with rotation with anti-␤-arrestin-1 (D7Z3W, Cell Signaling Technology) or anti-Pellino-1 (D2Z4F, Cell Signaling Technology) antibodies. As control of the specificity of antigenantibody binding, lysates were immunoprecipitated with rabbit IgG antibody (MilliporeSigma). The following day, antibodybound proteins were immunoprecipitated with protein A/Gagarose (Thermo Fisher Scientific) for 2 h 30 min at 4°C with rotation. Co-immunoprecipitated proteins were assessed by Western blot analysis using anti-GR antibody.

In vivo ubiquitination assay
The in vivo ubiquitination assay was performed as described by Choo and Zhang (82) with a few modifications. Briefly, cells were transfected with NTC and ARRB1 siRNAs to knock down ␤-arrestin-1 expression, and 24 h after siRNA transfection, cells were transfected with expression plasmid HA-ubiquitin, which was a gift from Edward Yeh (Addgene plasmid number 18712) (83). Twenty-four hours after HA-ubiquitin transfection, cells were treated with or without dexamethasone (100 nM) in the presence of MG132 (10 M). Cells were lysed with complete cell lysis buffer (2% SDS, 150 mM NaCl, 10 mM Tris-HCl, pH 8.0) with 1 mM PMSF, 5 mM N-ethylmaleimide, and protease inhibitor mixture following the paper's instructions. Then cells were diluted using dilution buffer (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 2 mM EDTA, 1% Triton, 1 mM PMSF, and protease inhibitor mixture) and incubated at 4°C for 60 min with rotation. After measuring protein concentration, 1000 g of lysates were incubated overnight at 4°C with rotation with anti-GR antibody. The following day, antibody-bound proteins were immunoprecipitated with protein A/G-agarose (Thermo Fisher Scientific) for 2 h 30 min at 4°C with rotation, and then washed ␤-Arrestin-1 modulates the glucocorticoid receptor turnover with the washing buffer (10 mM Tris-HCl, pH 8.0, 1 M NaCl, 1 mM EDTA, 1% Nonidet P-40, 1 mM PMSF, and protease inhibitor mixture) twice and with PBS once. A/G-agarose resin was spun down at maximum speed to aspirate the residual washing buffer. After adding 25 l of SDS Laemmli buffer and 1 l of 14.3 M ␤-mercaptoethanol, agarose resin was boiled for 5 min, and samples were loaded onto SDS-polyacrylamide gels for immunoblotting. GR ubiquitination was detected by using anti-HA tag antibody (EMD Millipore).

ChIP and FAIRE assays
ChIP experiments were performed in A549 NTC and ARRB1-KD cells that were plated on 150-mm dishes at 90% confluence. After 2-h dexamethasone or vehicle treatment, 1% formaldehyde was added to the dish to cross-link proteins to DNA. After formaldehyde inactivation with glycine, cells were scraped and resuspended in cell lysis buffer provided by the MagnaChIP kit (EMD Millipore) to isolate nuclei. The crude nuclei were collected, resuspended in nuclear lysis buffer (EMD Millipore), subjected to sonication using a Diagenode Bioruptor (15 cycles on high setting, 30 s on, 30 s off, twice). DNA extracts were first run onto 1% agarose gel to evaluate chromatin shear ranging from 200 to 1000 bp and then precleared and immunoprecipitated with anti-GR antibody or IgG control antibody. The GR/DNA complex was pulled down using the MagnaChIP kit (EMD Millipore).
For the FAIRE procedure, we used the same sonicated DNA extracts that were used for the ChIP assay. We followed the protocol described by Simon et al. (84). Quantitative real-time PCR was performed using the primers designed for the two GREs of the PELI1 locus and the GRE Ϫ1500 bp of the TSC22D3 locus.

In situ PLA and immunofluorescence staining
In situ PLA detection was performed using the Duolink PLA kit (Sigma-Aldrich) according to the manufacturer's instructions. Briefly, A549 cells were grown in glass bottom culture dishes (MatTek Corp.) for the time and with the treatment required by the experiment. Then cells were washed with warm PBS, fixed with warm 4% paraformaldehyde for 20 min at room temperature, and blocked in PBS containing 2% BSA and 0.1% Triton X-100 for 30 min at room temperature. Cells were then incubated for 1 h with the Duolink blocking solution at room temperature prior to incubating the specimens at 4°C over-night with fluorescein phalloidin (as cytoplasmic marker; Thermo Fisher Scientific) and a combination of primary antibodies: rabbit anti-␤-arrestin-1 antibody (D7Z3W, Cell Signaling Technology) together with mouse anti-GR antibody (clone E57 (51)); rabbit anti-GR antibody (D8H2, Cell Signaling Technology) with mouse anti-ubiquitin (P4D1; Cell Signaling Technology) antibody; and mouse anti-GR antibody (clone E57) with rabbit polyclonal anti-Lys-48 linkage-specific polyubiquitin antibody (Cell Signaling Technology). The following morning, secondary PLA probes anti-mouse PLA-minus and antirabbit PLA-plus were incubated for 1 h at 37°C in a humid chamber. After washing, ligase and amplification steps were performed as indicated in the protocol. Samples were then washed, air-dried, and mounted with ProLong gold antifade mountant with DAPI (Thermo Scientific). A technical negative control was performed with only anti-GR antibody. A Zeiss laser-scanning confocal microscope (LSM 780 and LSM 880; Carl Zeiss) was used to analyze protein-protein interactions. To quantify the PLA signals (also known as puncta), maximum intensity projections of the raw images were split into singlechannel images by MetaMorph software (Molecular Devices). A minimum fluorescent threshold of PLA puncta was used for all samples within one experiment. The number of puncta measured was divided by the number of cells obtained by thresholding the DAPI channel (used as nuclear stain). For quantification, each bar (mean Ϯ S.D.) represents the mean obtained from three independent experiments. The specificity of the antibodies for ubiquitin and anti-Lys-48-ubiquitin was confirmed, performing ubiquitination assays in the presence or absence of the proteasome inhibitor (data not shown).
For immunofluorescence staining, NTC and ARRB1-KD cells were grown in glass bottom culture dishes (MatTek Corp.) with or without Dex for 3 h. Then, cells were washed twice with warm PBS, fixed with warm 4% paraformaldehyde for 20 min at room temperature, and blocked in 1ϫ PBS containing 2% BSA and 0.1% Triton X-100 for 30 min at room temperature. Cells were then incubated for 1 h with blocking buffer (1ϫ PBS, 5% normal goat serum, and 0.1% Triton X-100) at room temperature prior to incubating the samples at 4°C overnight with rabbit anti-GR antibody (D8H2, Cell Signaling Technology) diluted in 1ϫ PBS containing 2% BSA. The following morning, samples were washed with 1ϫ PBS containing 0.1% Tween and incubated with the secondary antibody goat anti-rabbit AF594 for 1 h at room temperature. Samples were then washed, airdried, and mounted with ProLong gold antifade mountant with DAPI (Thermo Scientific). A Zeiss laser-scanning confocal microscope (LSM 780 and LSM 880; Carl Zeiss) was used to analyze GR cellular distribution. A technical negative control was performed omitting the GR antibody.

Statistical analysis
All data are presented as means Ϯ S.D. Student's t test and one-and two-way analysis of variance with Sidak's and Dunnett's multiple-comparison tests, according to the number of groups, were used to determine whether differences between experimental groups were statistically significant. All analyses were analyzed using the GraphPad Prism 7 software package (GraphPad Software).