The β-blocker Nebivolol Is a GRK/β-arrestin Biased Agonist

Nebivolol, a third generation β-adrenoceptor (β-AR) antagonist (β-blocker), causes vasodilation by inducing nitric oxide (NO) production. The mechanism via which nebivolol induces NO production remains unknown, resulting in the genesis of much of the controversy regarding the pharmacological action of nebivolol. Carvedilol is another β-blocker that induces NO production. A prominent pharmacological mechanism of carvedilol is biased agonism that is independent of Gαs and involves G protein-coupled receptor kinase (GRK)/β-arrestin signaling with downstream activation of the epidermal growth factor receptor (EGFR) and extracellular signal-regulated kinase (ERK). Due to the pharmacological similarities between nebivolol and carvedilol, we hypothesized that nebivolol is also a GRK/β-arrestin biased agonist. We tested this hypothesis utilizing mouse embryonic fibroblasts (MEFs) that solely express β2-ARs, and HL-1 cardiac myocytes that express β1- and β2-ARs and no detectable β3-ARs. We confirmed previous reports that nebivolol does not significantly alter cAMP levels and thus is not a classical agonist. Moreover, in both cell types, nebivolol induced rapid internalization of β-ARs indicating that nebivolol is also not a classical β-blocker. Furthermore, nebivolol treatment resulted in a time-dependent phosphorylation of ERK that was indistinguishable from carvedilol and similar in duration, but not amplitude, to isoproterenol. Nebivolol-mediated phosphorylation of ERK was sensitive to propranolol (non-selective β-AR-blocker), AG1478 (EGFR inhibitor), indicating that the signaling emanates from β-ARs and involves the EGFR. Furthermore, in MEFs, nebivolol-mediated phosphorylation of ERK was sensitive to pharmacological inhibition of GRK2 as well as siRNA knockdown of β-arrestin 1/2. Additionally, nebivolol induced redistribution of β-arrestin 2 from a diffuse staining pattern into more intense punctate spots. We conclude that nebivolol is a β2-AR, and likely β1-AR, GRK/β-arrestin biased agonist, which suggests that some of the unique clinically beneficial effects of nebivolol may be due to biased agonism at β1- and/or β2-ARs.


Introduction
Nebivolol is classified as a third generation b-adrenoceptor (b-AR) antagonist (b-blocker) that has a higher affinity for b 1adrenoceptors (b 1 -ARs) compared to b 2 -ARs and b 3 -ARs [1,2]. Importantly, nebivolol activates endothelial nitric oxide (NO) synthase (eNOS) leading to vasorelaxation in vivo and ex vivo [3][4][5]. Although the mechanism is unknown, a leading theory is that nebivolol induces vasodilation via b 3 -ARs [5][6][7][8]. However, recent evidence argues against a mechanism involving human b 3 -ARs [9], and nebivolol's reported affinity for b 3 -ARs (K i , 1.5 mM) is much weaker than for b 2 -ARs (K i , 30 nM) and b 1 -ARs (K i , 0.8 nM) [9,10]. Therefore, it seems unlikely that pharmacologically relevant concentrations of nebivolol would act via b 3 -ARs, especially in vivo where 98% of nebivolol is bound to plasma proteins [11]. Because of the ratios of agonist to antagonist utilized the most recent reports regarding nebivolol acting through b 3 -ARs [5] leaves open the possibility that nebivolol could act, at least in part, via b 1 -ARs or b 2 -ARs. The b 3 -AR is not the only theorized target receptor; an alternative theory is that metabolites of nebivolol induce vasodilation through b 2 -ARs [12].
Nebivolol abolishes b 1 -and b 2 -AR-mediated cAMP generation in response to classical agonists and does not generate cAMP through either receptor on its own [9,13,14]. Because of these antagonistic properties at b 1 -and b 2 -ARs, the theories described previously have grown in acceptance due to viewing nebivolol solely as a classic antagonist. However, the recently developed concept of biased agonism indicates that classification based on cAMP production alone is not sufficient to dismiss a ligand from having agonistic properties at b 1 -and b 2 -ARs.
Biased agonism is a relatively new term that defines a subset of functional selectivity [15][16][17]. G protein-coupled receptors (GPCRs) signal through at least two mechanisms: the traditional Ga and Gbc pathways, as well as the more recently appreciated G protein-coupled receptor kinase (GRK)/b-arrestin pathway. Traditional agonists signal through both of these mechanisms simultaneously, and pure antagonists block both pathways simultaneously; however, biased agonists favor one mechanism over the other. This bias allows for a ligand to signal differently than a traditional agonist and, importantly, means that what once were called ''antagonists'' (because they inhibit the Ga and Gbc pathways) may, in fact, activate the GRK/b-arrestin pathway, which would result in unique properties compared to pure antagonists.
Thus far, there is only one known clinically utilized GRK/barrestin biased agonist: carvedilol [18,19]. Carvedilol, which is now available as a generic drug, was originally marketed as a bblocker. In 2007, carvedilol was recognized as a biased agonist that induces b-AR-mediated activation of the GRK/b-arrestin pathway but not the Ga s -coupled/cAMP pathway [19][20][21]. Furthermore, carvedilol along with the other third generation b-blockers, such as nebivolol, are unique among b-blockers because they result in vasodilatation. Like nebivolol, carvedilol-mediated vasodilation is attributed to nitric oxide (NO) production [22]. Moreover, carvedilol-mediated NO production and vasodilation may be due, at least partially, to biased agonism at b-ARs [23].
Given the similarities between carvedilol and nebivolol, it is possible that they signal through similar mechanisms. Although studies have examined most b-blockers for biased-agonist activity [19,20], to date nebivolol has not been examined. Nebivolol is an intriguing candidate for GRK/b-arrestin biased agonism because of its similarities with carvedilol and its debated signaling mechanisms. Since vasodilation can be mediated through an endothelial cell epidermal growth factor receptor (EGFR)-mediated mechanism [24][25][26], b-AR-mediated transactivation of the EGFR through a GRK/b-arrestin biased signaling mechanism could explain nebivolol-mediated NO production without invoking the need for additional ligand-receptor interactions such as via b 3 -ARs. Therefore, we tested the hypothesis that nebivolol is a GRK/b-arrestin biased agonist.
Assessment of GRK/b-arrestin biased agonism can be accomplished through various approaches including phosphorylation of the receptor, receptor internalization, redistribution of b-arrestins, and b-arrestin-mediated signaling. To avoid inducing any overexpression artifacts, we utilized endogenous b-ARs which, due to poor b-AR specific antibodies [27], limited these studies to: b-AR internalization and signaling; b-arrestin-mediated signaling and cellular redistribution; as well as probing for the role of the EGFR and GRKs. Data generated by these studies collectively indicate that nebivolol is indeed a GRK/b-arrestin biased agonist.

Cell Culture
Mouse embryonic fibroblasts lacking all the Gnas exon 2 derived products including Ga s and XLa s (Gnas E22/E22 ) have been described previously [28]. These cells were stably transfected with a plasmid containing human Ga s short driven by an EF-1 promoter and conferring blasticidin resistance. Functional expression of Ga s and b-ARs was confirmed by measuring cAMP production in the presence and absence of 10 mM isoproterenol (data not shown). For simplicity, these cells are referred to as MEFs in this manuscript. The Gnas E22/E22 cells and MEFs were grown in DMEM-F12 supplemented with 10% FBS, 50 units of penicillin-streptomycin (Life Technologies), and 5 mg/mL blasti-  [29]. The myocytes were cultured in complete Claycomb supplemented with 10% FBS (JRH Biosciences, Lenexa, KS), penicillin-streptomycin (100 U/ml), norepinephrine (100 U/ml) and L-glutamine (2 mM) (Life Technologies). The culture flasks were pre-coated with 12.5 mg/ml fibronectin in 0.02% gelatin solution (Sigma). The myocytes were also maintained at 37uC in 5% CO 2 plus 95% air humidified incubator.

Identification of b-ARs and GRKs
Taq-Man qPCR was utilized to identify the b-AR and GRK isoforms expressed in MEFs; HL-1 cells and mouse aorta total RNA (positive control) were also used to identify the b-ARs. The primers were designed utilizing the online primer designer from IDT (Coralville, IA) to amplify mouse b 1 -, b 2 -, and b 3 -ARs, and GRK 2, 3, 4, 5, and 6; additionally a universal 18 s primer was used as an internal control (Table 1). Qiagen (Valencia, CA) RNeasy Plus Mini kit with shredder columns were used as directed to isolate total RNA from confluent plates of cells, and the total RNA was converted to cDNA using Qiagen's Omniscript kit with RNase inhibitor from Promega (Madison, WI). The cDNA, primers, and TaqMan Universal PCR Master Mix (Life Technologies) were placed into 0.2 mL AB MicroAmp tubes from Life Technologies and the samples run on an ABI 7000 with standard cycling (10 minute 95uC followed by 40 cycles of 15 sec 95uC  denaturing step and a 1 minute 60uC anneal/extend step). Each sample was run in triplicate for each of the aforementioned transcripts. An auto baseline was used to determine the Ct for all samples, and DCt was determined by subtracting the 18 s from the paired transcripts. The relative quantification (RQ) method utilizing 2 (2DDCt) was used to compare the transcripts expression levels. For b-ARs in MEFs and HL-1 cells, the pooled DCt of b 1and b 2 -ARs of mouse aorta was used as the control in the DDCt method, and for the GRKs in MEFs the b 2 -AR was used as the control. Nonparametric statistics were used to analyze the data.

Measurement of cAMP, Purines, and Pyrimidines
MEFs were grown in 12-well plates to near confluence and serum starved overnight in 750 mL serum free DMEM/F12. All cells were treated with 500 mM IBMX for 5 minutes, then 1 mM isoproterenol or 10 mM forskolin was added to the cells. After a 30-minute incubation at 37uC, cells were washed with ice-cold phosphate buffered saline (PBS) and 500 mL ice-cold 1-propanol was added to the cells. The cells were incubated for two hours at 4uC, and then the propanol was collected in 1.9 mL tubes and dried in a rotary evaporator. While the samples were drying, the fixed cells were stained with 5 mM of the cell permeant Hoechst 33342 dye in a 3% bovine serum albumin from Jackson ImmunoResearch (West Grove, PA) in PBS for 1 hour. The cells were then washed 3 times in PBS and each plate was imaged 3 times in a BioTek Synergy HT to account for cell number as previously described [30]. The dried samples were reconstituted in 200 mL ultrapure water and analyzed on a Thermo Electron TSQ Quantum-Ultra system with a HESI source utilizing an Agilent Zorbax eclipse XBD-C-18 column as described previously [31].

b-AR Internalization
A modified protocol from Seibold et al. was utilized [32]. Twelve -well plates were coated with poly-D-lysine (Sigma) and cells were immediately seeded on the plates. Once confluent the cells were starved overnight in 750 mL DMEM/F12. For treatment a 4x solution of agonist (nebivolol or isoproterenol) in DMEM/F12 was prepared and 250 mL was added to each well. Two sets of zero time points were used in each experiment. When the incubation time expired, plates were placed on ice and washed 3 times with 1 mL ice cold DMEM/F12. After washing the cells, 1 mL of DMEM/F12 containing 5 nM of the cell impermeant 3 H-CGP [33] and 5 mM Hoechst 33342 was added to all but the second zero time point. The second zero time point acted as a control for nonspecific binding; its media contained 5 nM 3 H-CGP, 5 mM Hoechst 33342, plus 1 mM alprenolol. The dishes were then incubated for 1 hour at 4uC after which the solution was removed and the samples washed twice with ice cold DMEM/F12 and once with ice cold PBS. The PBS was removed and the plate imaged to determine the intensity of the Hoechst stain in a BioTek FLx800 fluorometer 3 times. The cells were removed from the plate by adding 200 mL trypsin/EDTA (Life Technologies) for at least 30 minutes at 37uC. All of the solution was removed from the wells and placed into scintillation vials; 5 mL scintillation fluid was added to the vials and the 3 H-CGP counted on a Beckman LS 6000IC scintillation counter. Cell surface receptors were determined by the ratio of 3 H-CGP/Hoechst; Hoechst 33342 staining intensity is directly related to cell number [30]. The nonspecific binding was then subtracted from each individual point and data were expressed as percent of total surface receptors at time zero. Data were fit to and analyzed with a one-phase decay model in GraphPad Prism 5 for Windows.

Western Blots
MEFs and HL-1 cells were grown in 60 mm and 100 mm tissue culture plates, respectively, serum starved overnight (at least 12 hours) then either not pretreated (for nebivolol signaling experiments) or pretreated for 30 minutes with DMSO (vehicle), 1 mM AG1478, or 30 mM propranolol. Following treatment, cells were stimulated with an agonist (nebivolol, carvedilol, or isoproterenol) for up to 30 minutes, then lysed in lysis buffer (1% Triton X-100, 100 mM NaCl, 20 mM Tris, pH 7.5, 2 mM EDTA, 10 mM MgCl 2 , 10 mM NaF, 40 mM b-glycerol phosphate, supplemented with 1 mM PMSF and 2 mM Na 3 VO 4 ). The samples were clarified by centrifuging the lysate at 14,000 rpm for 5 minutes at 4uC and removing the supernatant, then Laemmli sample buffer was added and the samples were boiled and stored at -20uC before analysis. The samples were separated on a 9% SDS-PAGE gel and transferred to nitrocellulose obtained from BioRad (Hercules, CA). ERK activation was measured via running duplicate gels and blotting with phospho-specific ERK1/2 (E10) antibodies on one blot and for total ERK1/2 on the second blot. When examining barrestins an additional gel (triplicate) was run and probed with barrestin 1/2 (D24H9) antibody. To examine GRK expression level . Nebivolol time-dependently and Ga s -independently signals to ERK. MEFs were treated with 100 nM of isoproterenol, 1 mM of nebivolol, or 100 nM of carvedilol for the indicated time points (A); representative blots are shown above the graph. Data are compared by one-way ANOVA with Kruskal-Wallis Multiple comparison post-hoc test; two asterisk (**) indicates significantly different than control by Bonferroni method (more stringent) and one asterisk (*) indicates significantly different by the regular test (less stringent). Gnas E22/E22 cells were treated with 100 nM of isoproterenol and 1 mM of nebivolol for the indicated time points (B); representative blots are shown above the graph. These data were compared as in panel A. The data obtained from the MEFs and Gnas E22/E22 cells were compared directly for nebivolol-(C) and isoproterenol-mediated (D) phosphorylation of ERK. Data are compared by two-way ANOVA with Sidak's multiple comparisons post-hoc test; the cross ({) indicates that the MEFs and Gnas E22/E22 cells are significantly different from each other. HL-1 cardiomyocytes were treated with 10 mM of nebivolol for the indicated time points (E); representative blots are shown above the graph. These data were compared as in panel A. All data are expressed as mean 6 SEM; n = 3 to 6. doi:10.1371/journal.pone.0071980.g004 the blots were probed with GRK specific antibodies and normalized to actin. The ERK and b-arrestin antibodies were from Cell Signaling Technology (Beverly, MA) and the GRK and actin antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA); the HRP-conjugated secondary antibodies were obtained from Jackson ImmunoResearch Laboratories, and ECL was obtained from Pierce (Rockford, IL). The images were captured and analyzed using a BioRad ChemiDoc XRS system running quantity one software or a Kodak 1500 system running Carestream software.
siRNA Knockdown of b-arrestin 1 and 2 Pooled mouse specific siRNA towards b-arrestin 1 (barr1) (sc-29742) and b-arrestin 2 (barr2) (sc-29743), as well as non-targeting siRNA-A and fluorescein conjugated control siRNA-A were obtained from Santa Cruz Biotechnology. The siRNA was diluted per the manufacturer's specifications and transfected into the MEFs utilizing Life Technologies siPORT Amine following the manufacturer's instructions except that the cells were allowed to remain attached to the tissue culture dish. The final concentration of siRNA transfected into the cells was 50 nM. The transfection media was left in the cells for two days, then the cells were serum starved overnight and the procedures under the western blot section were followed. Transfection efficiency was examined utilizing the fluorescein conjugated control siRNA-A with a Zeiss Axiovert 40 fluorescent microscope.

b-arrestin Redistribution
MEFs were prepared for imaging as described previously [34]. Nebivolol (10 mM) was used to stimulate the cells for up to 30 minutes. The primary antibodies barr1 (sc-9182 K-16) and barr2 (sc-13140 H-9) from Santa Cruz Biotechnology were used at 1:200 dilution, the fluorophore labeled secondary antibodies Alexa Fluor 555 Donkey Anti-Goat IgG for barr1 as well as Alexa Fluor 647 Figure 6. GRK expression and role in nebivolol-mediated phosphorylation of ERK. qPCR was used to identify the GRKs present in MEFs (A); data are expressed as mean 6 SEM, n = 3. Western blots for GRK2, 4, 5, and 6 as well as actin, as a control, from GNAS E22/E22 cells, MEFs, and heterologously expressed GRKs from 293T cells (positive controls) were used to confirm the qPCR data via identifying the protein expression of GRKs (B). MEFs were pretreated with the GRK inhibitor for 30 minutes then stimulated with 10 mM of nebivolol for 7 minutes (C). Representative blots are presented above the histogram of the data. Open bars are controls, and black bars are nebivolol treated. Data are expressed as mean 6 SEM; n = 4. All data were analyzed by a one-way ANOVA with Tukey-Kramer post hoc analysis, and significance is denoted by Greek letters. Any data set with a different Greek letter is statistically different p,0.05. doi:10.1371/journal.pone.0071980.g006 Donkey Anti-Mouse IgG for barr2 were used at a 1:250 dilution when conducting dual staining. barr1 did not generate a strong signal over background; therefore, the replicates utilized single staining with the barr2 antibody and utilized 1:250 Alexa Fluor 488 Donkey Anti-Mouse IgG as a secondary antibody. 5 mM Hoechst 33342 was used to identify the nuclei. The images were acquired with a Leica SPE confocal microscope utilizing a HCX PL APO CS 100x 1.4 NA oil objective and 405, 561, and 635 nm laser lines. All images were acquired at a resolution of 102461024 with identical settings for each image, and each image is an average of 3 repetitive scans. For analysis of barr2 redistribution and accumulation in vesicles, ImageJ was utilized to identify the punctate staining that represents aggregated barr2. Aggregation creates a greater intensity of the pixels than diffuse staining; therefore, the aggregates were identified and counted utilizing a filter to only display pixels that are greater than or equal to 64 (25% of full intensity), 96 (38% of full intensity), 128 (50% of full intensity), and 192 (75% of full intensity) intensity units. The identified punctate staining was then manually counted for each cell, and the following formula was used to integrate the number of vesicles with the threshold where the vesicles were visible: (# only seen at 64 intensity units * 64/255)+(# only seen at 96 intensity units * 96/255)+(# only seen at 128 intensity units * 128/255)+(# only seen at 192 intensity units * 192/255). The original data was obtained from 4 to 5 individual images of a cell; to increase the sample size for analysis by nonparametric statistics, resampling was used to create 8 separate sets of data. The means from each set were used in the analysis. Following analysis the image intensity was enhanced in Photpshop utilizing a sigmoidal green curve to improve visibility of the images; each image was enhanced equally.

Statistics and Data Analysis
All graphs were generated with GraphPad Prism 6, and the Pearson's correlation and two-way ANOVAs were also performed in GraphPad. One-way ANOVAs with Tukey-Kramer post hoc tests for normal data sets and Kruskal-Wallis posthoc test for nonnormal data sets were conducted with NCSS 2007; the post-hoc tests are indicated in each figure legend. The data in the figures are expressed as mean 6 SEM; textual data utilizes confidence intervals and standard deviation as detailed in the text.

Results and Discussion
As there is still debate regarding the role of the b 3 -AR in nebivolol-mediated signaling, the expression of b-ARs in MEFs and HL-1 cardiac myocyte cells were examined via qPCR. Since both of these cells originate from mice, and mouse aorta expresses all three subtypes of b-ARs [35,36], mouse aorta was used as a The efficacy of the siRNA was analyzed via a one-way ANOVA with Tukey-Kramer post hoc analysis; statistical difference is denoted by Greek letters; any data set with a different Greek letter is statistically different p,0.05. Data are expressed as mean 6 SEM; n = 16. The ERK phosphorylation data (B), collected only from samples where b-arrestins were knocked down by 25% or more, are expressed as the percent change induced by nebivolol. These data were analyzed via a one-way ANOVA with a nonparametric Kruskal-Wallis post hoc test. Statistical difference (p,0.05) is denoted by Greek letters; any data set with a different Greek letter is statistically different. Data are expressed as mean 6 SEM; n = 5 to 9. Due to the poor knockdown of b-arrestins, a scatter plot of nebivolol-induced phosphorylation of ERK and b-arrestin level was generated (C) and a Pearson's correlation was run to determine the relationship between barrestin level and nebivolol-mediated phosphorylation of ERK (n = 38); the dotted lines represent the 95% confidence interval of the line. doi:10.1371/journal.pone.0071980.g007 positive control. The primer pairs were able to amplify all b-ARs subtypes in the mouse aorta (Fig. 1A) and no statistical difference was observed between the receptors although the b 3 -AR tended to have lower expression. b 1 -ARs were found only in the HL-1 cells, and b 2 -ARs were found in both MEFs and HL-1 cells (Fig. 1B). b 3 -AR message was below detection levels in both MEFs and HL-1 cells even after 40 cycles of PCR under the conditions used in this study. Expectedly, the cardiac myocyte HL-1 cells contained significantly more b 1 -AR message than b 2 -AR message; in HL-1 cells the b 1 -AR crossed the threshold ,8 cycles before the b 2 -AR, which corresponds to approximately 256-fold greater expression of b 1 -AR mRNA compared to b 2 -AR mRNA. Importantly, these data indicate that none of the following results are likely to be attributed to the b 3 -AR, and the MEFs allow for examining predominately, if not only, b 2 -ARs.
As the GRK/b-arrestin pathway is classically involved in receptor internalization and ligands that activate the GRK/barrestin pathway induce receptor internalization, we examined whether nebivolol stimulation resulted in internalization of b-ARs as a first test of biased agonism. Before examining the effect of nebivolol on b-AR internalization, we validated the internalization assay and determined the rates of isoproterenol-mediated b-AR internalization utilizing 3 H-CGP. CGP is a cell impermeant b 1and b 2 -AR antagonist and a b 3 -AR agonist [33,37]; however, it has much lower affinity for b 3 -ARs compared to the b 1 -and b 2 -ARs [38] and there are no detectable b 3 -ARs in the cells studied. Therefore, the observed effects cannot be attributed to b 3 -ARs. MEFs treated with 100 nM and 1 mM isoproterenol displayed a time and concentration dependent internalization of the b 2 -AR ( Fig. 2A); since MEFs only express b 2 -ARs, we attribute the observed effects in MEFs to b 2 -ARs. There was at least a 2.5 minute time delay in the 100 nM isoproterenol-mediated b 2 -AR internalization; however, once internalization was initiated it progressed with a tK of 6.260 min (95% confidence interval (CI): 3.516 to 28.510 min). The delay in internalization and the rate were concentration dependent; 1 mM isoproterenol induced internalization without any delay and resulted in a statistically shorter tK of 1.315 min (95% CI: 1.046 to 1.769 min; p,0.05 vs. 100 nM) representing a greater rate of internalization. Similarly, 10 mM of isoproterenol (not shown) had no delay and an even more rapid rate of internalization with a tK of 0.6928 min (95% CI: 0.5347 to 0.9836 min, p,0.05, vs. 1 mM). In all cases the plateau of the model, which represents the equilibrium between receptor internalization and recycling to the surface, was approximately 40% of the total surface receptor. Following nebivolol treatment, similar results were observed. 100 nM nebivolol had no effect for at least 2.5 minutes, but then the b 2 -AR internalized with a tK of 2.301 min (95% CI: 1.481 to 5.162 min). 1 mM nebivolol induced internalization of the b 2 -AR with a tK of 1.302 min (95% CI: 0.8900 to 2.422 min) (Fig. 2B). Although the rates were not different in a statistically significant manner, the plateaus were statistically different; 66.46% (95% CI: 60.48 to 72.44%) of receptors remained on the surface versus 39.81% (95% CI: 34.59 to 45.04%) in the 100 nM and 1 mM nebivolol treated samples, respectively (p,0.05). We also examined the effect of nebivolol on the b 1 -and b 2 -ARs in HL-1 cells (Fig. 2C); 1 mM nebivolol induced internalization of the b-ARs with a tK of 0.6555 min (95% CI: 0.4365 to 1.316 min) and a plateau of 40% of the surface receptors. These data demonstrate that nebivolol induces internalization of b-ARs in a manner that is similar to that of a full agonist. Thus, nebivolol is not a pure antagonist because pure antagonists do not induce receptor internalization. Therefore, either nebivolol is a partial agonist at the b-ARs that leads to robust internalization, a GRK/b-arrestin biased agonist at the b 1 -and b 2 -ARs, or acts as an agonist at an unknown receptor that causes heterologous desensitization of the b-ARs.
To begin to determine which of the previous possible explanations describes the observed effects of nebivolol we tested the hypothesis that nebivolol is a partial agonist. If nebivolol is a partial agonist then 39,59-cAMP (cAMP from adenylyl cyclase) should increase in a statistically significant concentration-dependent manner. MEFs treated with 100 nM nebivolol for 30 minutes, a concentration that causes internalization, failed to alter 39,59-cAMP levels (Fig. 3A); control and 100 nM nebivololstimulated levels of 39,59-cAMP were (mean 6 standard deviation) 14.4864.39 and 14.1261.52 (n = 6) pg/100,000 cells, respectively. This result decouples the internalization, which is similar to the full agonist isoproterenol, from cAMP production suggesting that nebivolol is not a partial agonist. On the other hand, 2 and 20 mM nebivolol slightly, and dose-dependently, increased cAMP levels but not to a level that was statistically different from baseline; 2 and 20 mM nebivolol resulted in (mean 6 standard deviation) 17.0065.31 and 21.3366.65 (n = 6), pg/100,000 cells, respectively. This corresponds to a 1.17 and 1.47 fold increase in 39,59-cAMP at 2 and 20 mM nebivolol, respectively. In a separate experiment, MEFs were treated with 10 mM of isoproterenol, which resulted in a greater than 30-fold increase in cAMP levels (data not shown), confirming that we can detect an agonist's effects on 39,59-cAMP, and that the nebivolol-mediated increase in cAMP is exceedingly low. Previous studies indicate that nebivolol has no effect on cAMP [13,14], including studies using individually expressed b-ARs [9]; the discrepancy between those studies and our data are likely due to two factors: 1) LC-MS-MS detection of 39,59-cAMP is the most sensitive technique, and 2) the previous study utilizing cloned b 2 -ARs halted at 1 mM nebivolol and did record a nonsignificant increase in cAMP levels at 1 mM. Collectively the data regarding nebivolol-mediated cAMP production indicates that at concentrations attained in the serum upon clinically used doses of nebivolol, there is no effect on cAMP levels, but at higher concentrations (1 mM and greater) there appears to be marginal cAMP accumulation suggesting that nebivolol may be a weak partial agonist. However, this statistically, and likely biologically, insignificant increase in 3959-cAMP does not correlate with, or explain, the internalization data that mirrors a full agonist. This finding suggests that the other mechanisms to explain nebivolol's effects, biased agonism or a second receptor, are more plausible.
The LC-MS/MS procedure for examining cAMP allows for the quantification of a host of nucleosides and nucleotides, as well as their precursors and metabolites in the same samples [31]. As shown in the rest of Figure 3, nebivolol increased the amount of  (H-9), shown in green, occurred in B through K. All settings remained identical for each image; the scale bar in panel A represents 10 mm. Panel L represents vesicular barr2 in green on the left axis (n = 8) and 10 mM nebivolol-mediated phosphorylation of ERK in black on the right axis (n = 3 to 7). For both data sets a Kruskal-Wallis Multiple comparison test was run and two asterisk (**) indicates significantly different from control by the Bonferroni method (more stringent) and one asterisk (*) indicates significantly different by the regular test (less stringent). doi:10.1371/journal.pone.0071980.g008 inosine, guanosine, uridine, cytidine, 59-AMP, and xanthine in a concentration dependent manner. Nebivolol either had no effect on, or did not statistically increase, the levels of adenosine, 29-AMP, 39-AMP, hypoxanthine, thymidine, and uric acid (only adenosine (Fig. 3H) is shown); additionally, 29,39-cAMP was not detected (Fig. 3I). 29,39-cAMP is generally formed from the breakdown of the poly-A tails on mRNA and is found in abundance in apoptotic cells [39]; therefore, nebivolol did not trigger the degradation of mRNA and is likely not toxic in MEFs. The constellation of nucleosides that were increased suggests that there is increased translation and transcription, which fits with previous studies indicating that nebivolol increases dimethylarginine dimethylaminohydrolase transcript and protein levels [40] and that it alters microRNA expression [41]. Adenosine did not increase, which appears to contradict the translation and transcription hypothesis, yet this may be due to a portion of the increased inosine originating from adenosine degradation. Taken together, the data support previous findings that nebivolol increases translation and transcription. Importantly, all of the statistically different concentration-response changes occurred when there was an absence of a statistically different effect on the cAMP concentration-response. This confirms that nebivolol is not a pure antagonist and demonstrates that nebivolol alters other molecules more potently and efficaciously than 39,59-cAMP, indicating that it is not a classical partial agonist, as it does not affect all downstream targets equally, which is the definition of biased agonism.
A common signaling pathway initiated by GRK/b-arrestin biased agonists is the phosphorylation of ERK through transactivation of the EGFR, and EGFR-mediated signaling as well as ERK-mediated signaling is known to increase transcription and translation. To investigate if nebivolol signals in this manner, its time-, receptor-, and GRK-dependent signaling to ERK was examined. First the time course of 1mM of nebivolol-mediated phosphorylation of ERK was compared to 100 nM of isoproterenol and 100 nM of carvedilol (Fig. 4A). Isoproterenol stimulation resulted in a robust phosphorylation of ERK that peaked at 10 minutes and returned to baseline at 30 minutes; whereas nebivolol and carvedilol stimulation increased ERK phosphorylation only half maximally, compared to isoproterenol, and displayed a shallow peak at 15 minutes. The time profile of nebivolol and carvedilol were nearly identical suggesting that they signal through a common mechanism. It should be noted that nebivolol has an approximately 30-fold lower affinity than carvedilol at the b 2 -AR [10,42], which is why we used greater amounts of nebivolol. Additionally, 10 mM of nebivolol induced a similar time-dependent phosphorylation of ERK as 1 mM of nebivolol (shown with confocal images), except that it appeared to elicit a biphasic response. To further determine if these effects are due to partial agonism at the b 2 -AR, the parental Gnas E22/E22 cells, which lack all Ga s transcripts [28], were treated with 100 nM of isoproterenol and 1 mM of nebivolol (Fig. 4B). Both isoproterenol and nebivolol stimulation resulted in an approximate 2-fold increase in ERK phosphorylation for the duration of the experiment, which was statistically different from control at multiple points for both agonists. The effects of the two drugs were indistinguishable in the Gnas E22/E22 cells suggesting that nebivolol and isoproterenol act similarly when the b 2 -AR cannot signal through Ga s , and that nebivolol signals to ERK independently of Ga s . Nebivolol treatment of MEFs and Gnas E22/E22 cells yielded results that were not different from each other in a statistically significant manner, but the results did not appear identical (Fig. 4C). The two curves match perfectly for the first 5 minutes, and converge again at 20 minutes. This effect may have been due to differences over time in the cells after stable transfection with Ga s or due to the slight accumulation of cAMP; however, the data clearly demonstrated that Ga s is not absolutely required for nebivolol-mediated signaling. Supporting this conclusion, isoproterenol treatment of MEFs and Gnas E22/E22 cells were significantly different from each other at 10 and 15 minutes of stimulation (Fig. 4D). To confirm that nebivolol stimulation leads to ERK phosphorylation in a different cell line, HL-1 cells were stimulated with 10 mM of nebivolol (Fig. 4E). HL-1 cells responded more rapidly and more variably than the MEFs with a peak at 5 minutes. These experiments indicate that nebivolol can signal to ERK and that this signaling does not require Ga s , thus providing more evidence that nebivolol is not a classical antagonist and that nebivolol fits the profile of a biased agonist. However, these experiments leave open the possibility that nebivolol is acting as an agonist at a novel receptor that causes these effects.
To examine directly the role of b-ARs, as well as EGFRs, in nebivolol-mediated phosphorylation of ERK, MEFs and HL-1 cells were pretreated for 30 minutes with vehicle, 30 mM of propranolol, or 1 mM of AG1748 prior to stimulation with 10 mM of nebivolol for 7 minutes. Nebivolol-mediated phosphorylation of ERK in MEFs was fully inhibited by both the b-blocker propranolol and the EGFR kinase inhibitor AG1748 (Fig. 5A). This confirms that nebivolol-mediated phosphorylation of ERK emanates from b 2 -ARs and EGFRs in MEFs, similar to what is known for carvedilol [20]. However, this experiment does not distinguish if b 2 -AR activation precedes EGFR activation or vice versa. Yet, it is unlikely that nebivolol is an EGFR agonist. To confirm that these effects are not exclusive to MEFs, a pair of confirmatory experiments was conducted in HL-1 cells (Fig. 5B). The effects in HL-1 cells are indistinguishable from the MEFs, confirming that nebivolol signals through the b-ARs and EGFR and that this signaling is not cell type dependent. Additionally, identical experiments were conducted in MEFs with 1 mM of nebivolol and similar results were obtained (data not shown). These data confirm that nebivolol signals through b 1 -and b 2 -ARs, as well as EGFRs, to ERK. Collectively, these data strongly suggest that nebivolol is a GRK/b-arrestin biased agonist of b 2and b 1 -ARs and, like carvedilol, is inducing transactivation of the EGFR; however, these data do not directly examine GRKs or barrestin.
To examine the role of GRKs in nebivolol-mediated phosphorylation of ERK we first identified the isoforms of GRKs expressed in MEFs. qPCR indicates that GRK2 is expressed at levels significantly greater than any other GRK (Fig. 6A). As expected, the negative control, the retinal specific GRK1, is not expressed in MEFs isolated from the torso and appendages of the embryo. Surprisingly, GRK3 message is also not expressed. To confirm that the mRNA levels relate to protein we utilized specific GRK antibodies to detect GRKs 2 through 6 in GNAS E22/E22 cells and MEFs (Fig. 6B). Cloned GRKs expressed in 293T cells were used as positive controls, and in this process it was found that the GRK3 antibody also detected GRK2 and therefore GRK3 was not included in the analysis. There appears to be much greater protein levels of GRK2 than the GRK4 family; additionally, the bands for GRK5 are higher than expected as shown by stably expressing full length, sequence verified, GRK5 in MEFs. These data support the conclusion from the qPCR data that GRK2 is the most abundant GRK. Fortunately, there is a small molecule inhibitor of GRK2 available [43], allowing for the examination of the role of GRK2 in nebivolol-mediated phosphorylation of ERK. As shown in Figure 6C, 100 mM of the GRK2 inhibitor reduced nebivolol-mediated phosphorylation of ERK. Although nebivolol significantly increases phosphorylated ERK levels, due to the inhibitor alone slightly increasing phosphorylated ERK levels the samples treated with the GRK2 inhibitor were not statistically distinct from nebivolol treatment or the control. However, nebivolol failed to increase phosphorylated ERK levels in the presence of GRK2 inhibitor, supporting the conclusion that nebivolol-mediated phosphorylation of ERK requires GRK kinase activity.
To determine if there is a role for b-arrestins, siRNA directed against b-arrestins was utilized to knockdown b-arrestin-1 (barr1) and b-arrestin-2 (barr2). If nebivolol-mediated phosphorylation of ERK is initiated through b 2 -ARs and involves b-arrestins, then decreases in barr1 and/or barr2 should reduce nebivolol-mediated phosphorylation of ERK. To determine if this is the case, siRNA was used to decrease barr1 and barr2 separately; however, the siRNA was not specific and minimally reduced the amount of both arrestins (Fig. 7A). Only pooling the barr1 and barr2 siRNA resulted in significant knockdown of both b-arrestins. Therefore, the signaling analysis was conducted by pooling the barr1 and barr2 data and using a cutoff of 25% knockdown for inclusion in the signaling studies (Fig. 7B). These experiments were conducted identically to the inhibitor studies described in Figure 5 and 6, and the data are expressed as the mean change in nebivolol-induced ERK phosphorylation. In control (no siRNA) and MEFs transfected with non-targeting siRNA, nebivolol stimulation led to increased levels of phosphorylated ERK; however, when barrestins were knocked down by 25% or greater, nebivolol, on average, failed to increase phosphorylation of ERK. This finding is consistent with the conclusion that nebivolol is a biased agonist. Since only a fraction of the samples fall within the 25% or more reduction in b-arrestins level, we plotted the level of phosphorylated ERK against the levels of b-arrestin (Fig. 7C). As expected for a GRK/b-arrestin biased agonist, there is a statistically positive Pearson's correlation (r = 0.3926, P = 0.0148) between the level of b-arrestin and the ability for nebivolol to stimulate ERK.
To confirm that b-arrestins are involved in nebivolol-mediated effects, confocal imaging was used to examine the redistribution of b-arrestins following 10 mM of nebivolol treatment. b-arrestin redistribution occurs when arrestins are activated and can be visualized by the accumulation of b-arrestin at the plasma membrane and/or through accumulation of vesicular structures represented by a punctate staining. Therefore, time dependent barrestin redistribution can be used as an assay to identify GRK/barrestin biased agonists [44]. Two independent experiments were conducted to examine the cellular distribution of b-arrestins: dual staining for barr1 and barr2, and staining for just barr2 shown in Figure 8. This sequence demonstrated that the antibody used for barr2 produced a much greater signal than barr1; the barr1 signal was too faint to reliably measure over background. As the levels of barr1 and barr2 are relatively equal in MEFs as shown in Figure 7A, we attributed the observed difference in staining to the quality of the b-arrestin specific antibodies and therefore only analyzed barr2 distribution. Very little signal was present with only the secondary antibody (Fig. 8A), whereas diffuse cytoplasmic staining of barr2 was observed in control cells (Fig. 8B). Treatment with 10 mM of nebivolol for 1 minute had a marginal effect on barr2 distribution, yet some membrane localization was detectable (Fig. 8C, white arrow represents membrane localized barr2). Two to 20 minutes of stimulation with nebivolol resulted in a clear redistribution of b-arrestins into punctate structures (Fig. 8D-J); whereas, 30 minutes of stimulation with nebivolol resulted in a staining pattern similar to control and 1 minute stimulation (Fig. 8K). Quantification of vesicle number and intensity shows that nebivolol-mediated redistribution of b-arrestins follows a biphasic curve with a peak at 3.5 and 15 minutes and then approaches the baseline (Fig. 8L). This suggests that the cellular distribution of barr2 is returning to the basal state 30 minutes after stimulation. Additionally, the biphasic barr2 redistribution pattern is similar to, but precedes, 10 mM nebivolol-mediated phosphorylation of ERK (Fig. 8L). In accordance with the siRNA depletion of b-arrestins (Fig. 7), the barr2 peaks preceding the phosphorylation of ERK peaks suggest that nebivolol-mediated barr2 activation precedes ERK phosphorylation. These data indicate that nebivolol treatment induces redistribution of barr2 further supporting the conclusion that nebivolol is a biased agonist.
The mechanism(s) via which nebivolol induces signal transduction is of interest due to the ability of nebivolol, a highly specific b 1 -AR antagonist [2], to stimulate NO production [3,4]. This fundamental contradiction has been baffling and is at least one of the reasons that numerous theories regarding the mechanism of action of nebivolol have been proposed. Herein, we present data that indicates nebivolol is a b 2 -AR, and likely b 1 -AR, GRK/barrestin biased agonist. This explains how a b-blocker that is highly specific for the b 1 -AR results in activating signaling pathways without invoking the need for a second receptor. However, it remains an open question if this explains the unique in vivo effects attributed to nebivolol [41,45]. Further studies will have to be conducted to answer this question fully; however, these data present a plausible explanation for nebivolol-mediated signaling, specifically in the endothelium. Most transactivation studies indicate that the ligand activating the EGFR is a heparinbound EGF-like growth factor (HB-EGF) [46,47]. Furthermore, HB-EGF stimulates endothelial cell eNOS expression and activity [24] and induces arteriolar vasodilation in a NO-dependent manner [26]. Therefore, it is possible that the biased agonist effect of nebivolol is responsible for generation of NO through transactivation of the EGFR within endothelial cells.
The narrow theory regarding signaling presented above should not be thought of as the only feature of nebivolol. Recent studies utilizing a GRK/b-arrestin biased agonist towards the angiotensin type 1 receptor indicate that GRK/b-arrestin biased agonism leads to phosphorylation of several proteins [48] and that some of the phosphorylation sites are unique to the GRK/b-arrestin biased agonist [49]. Therefore, it is likely that nebivolol shares some signaling pathways with specific b 1 -and/or b 2 -AR agonists, but nebivolol likely induces unique signaling pathways although these pathways still emanate from the b 1 -and/or b 2 -AR. This may have a clinical role; multiple studies indicate that b-arrestins and barrestin-mediated signaling are important in cardiac physiology [50], and GRK/b-arrestin biased agonists promote myocyte survival [23,51]. Therefore, it is likely that the biased agonist property of nebivolol plays a crucial role in its clinical profile.

Conclusion
Nebivolol is a GRK/b-arrestin biased agonist at the b 2 -AR as demonstrated by the data obtained from the MEFs. It is also likely a GRK/b-arrestin biased agonist at the b 1 -AR as demonstrated by the data obtained from HL-1 cells. These conclusions are supported by the following constellation of results. After 40 cycles of qPCR the b 3 -AR was readily observed in the control but not the MEFs or HL-1 cells. Nebivolol induces b-AR internalization; phosphorylation of ERK through b-ARs, GRK, b-arrestin, as well as the EGFR; redistribution of b-arrestin-2; and it increased nucleoside cellular concentrations without significantly altering cAMP. These effects may explain the recently described physiological differences between treatment with nebivolol and classical b-AR antagonists such as atenolol and metoprolol [41,45].