Pro-Death Signaling of GRK2 in Cardiac Myocytes after Ischemic Stress Occurs via ERK-Dependent, Hsp90-Mediated Mitochondrial Targeting

Abstract

Rationale

GRK2 is abundantly expressed in the heart and its expression and activity is increased in injured or stressed myocardium. This up-regulation has been shown to be pathological. GRK2 can promote cell death in ischemic myocytes and its inhibition by a peptide comprised of the last 194 amino acids of GRK2 (known as βARKct) is cardioprotective.

Objective

The aim of this study was to elucidate the signaling mechanism that accounts for the pro-death signaling seen in the presence of elevated GRK2 and the cardioprotection afforded by the βARKct.

Methods and Results

Using in vivo mouse models of ischemic injury and also cultured myocytes we found that GRK2 localizes to mitochondria providing novel insight into GRK2-dependent pathophysiological signaling mechanisms. Mitochondrial localization of GRK2 in cardiomyocytes was enhanced after ischemic and oxidative stress, events that induced pro-death signaling. Localization of GRK2 to mitochondria was dependent upon phosphorylation at residue Ser670 within its extreme carboxyl-terminus by extracellular signal-regulated kinases (ERKs), resulting in enhanced GRK2 binding to heat shock protein 90 (Hsp90), which chaperoned GRK2 to mitochondria. Mechanistic studies invivo and invitro showed that ERK regulation of the C-tail of GRK2 was an absolute requirement for stress-induced, mitochondrial-dependent pro-death signaling, and blocking this led to cardioprotection. Elevated mitochondrial GRK2 also caused increased Ca²⁺-induced opening of the mitochondrial permeability transition pore, a key step in cellular injury.

Conclusions

We identify GRK2 as a pro-death kinase in the heart acting in a novel manner through mitochondrial localization via ERK regulation.

INTRODUCTION

G protein-coupled receptor (GPCR) kinase-2 (GRK2) is the most studied member of the GRK family of serine/threonine kinases, which phosphorylate activated receptors leading to a cessation of signaling, a process known as desensitization^{1, 2}. GRK2 is ubiquitously expressed but is the most abundant GRK isoform in the heart and over the last two decades this kinase has been shown to be a critical regulator of cardiac function, especially in disease, where it is up-regulated resulting in increased levels in the failing human heart^{1, 2}. In addition to classically serving as kinases regulating GPCR coupling, it is becoming increasingly apparent that GRKs play multi-faceted roles in cells since several non-receptor binding partners have been uncovered³. For example, GRK2 has been shown to interact with tubulin^{4, 5}, clathrin⁶, Akt⁷, Hsp90⁸ and IKBα⁹. In addition, there are various non-receptor proteins typically found associated with the plasma membrane that are known to regulate GRK2 localization and activity³. Thus, a fresh look at the emerging evidence suggests that GRK2 is a potential nodal protein intersecting multiple signaling pathways within the cell.

Not only is a large “GRK2 interactome” emerging, but in addition GRK2 can also undergo dynamic regulation via phosphorylation. Protein kinase C (PKC)¹⁰ and c-Src¹¹ have both been shown to phosphorylate GRK2 under conditions of cellular stress. Moreover, GRK2 has been shown to be phosphorylated by ERK MAP kinases at the specific residue Ser670¹². Recently, Ser670 has also was identified as a site for phosphorylation of GRK2 by phosphoinositide-3-kinase (PI3K)¹³. In addition to phosphorylation-dependent regulation of GRK2, an emerging area suggesting novel roles for GRKs beyond GPCR desensitization is novel sub-cellular localization of these kinases. For example, we have recently described that GRK5 is targeted to the nucleus of cardiomyocytes where it acts a novel kinase for Class II histone deactylases (HDACs) and this activity of GRK5 in the heart influences gene transcription in response to hypertrophic stimuli^{14, 15}. Most recently, GRK2 was detected in mitochondria of endothelial cells and HEK cells and had an effect on mitochondrial biogenesis¹⁶.

GRK2 is not targeted to the nucleus of cells but in this study we found that it localizes to heart mitochondria and have elucidated the signal transduction mechanism that targets GRK2 to mitochondria following oxidative and ischemic stress. We previously reported that GRK2 is a pro-death kinase in the heart promoting ischemic injury¹⁷, while its inhibition by the peptide inhibitor, βARKct, comprised of the carboxyl-terminal domain of bovine GRK2, confers cardioprotection^{18, 19}. Specifically, the novel results presented in this study indicate that mitochondrial targeting of GRK2 in myocytes after ischemic injury promotes pro-death signaling as mitochondrial accumulation of GRK2 in myocytes increases after oxidative stress. Preventing this mitochondrial targeting, which can be accomplished with βARKct expression, prevents death signaling and confers cardioprotection. Mechanistically, we found that this novel sub-cellular localization which occurs after oxidative stress is dependent on ERK-mediated phosphorylation of GRK2 at Ser670 and the subsequent movement to mitochondria is dependent on binding of phosphorylated GRK2 to Hsp90, a known mitochondrial chaperone. Thus, the data reported herein reveals that mitochondrial-targeted GRK2 is essential for pro-death signaling occurring after oxidative stress in myocytes and assigns a novel role for this GRK based on specific ERK regulation and its unique cellular localization.

METHODS

Experimental animals

Cardiac-specific transgenic mice (TG) with GRK2 or βARKct overexpression have been described previously¹⁹ and for all in vivo experiments non-transgenic littermate control (NLC) mice were used. All animal procedures were carried out in accordance with the guidelines of the Institutional Animal Care and Use Committee of Temple University School of Medicine.

In vivo Ischemia/Reperfusion (I/R) injury model

Surgical I/R injury was performed on 8–10 weeks old mice as previously described²⁰. All I/R (30 min ischemia/30 min reperfusion) procedures were controlled by a sham procedure without coronary ligation. We did not measure area at risk (AAR) or infarct size in this study but AAR of the LV is predicted to be 40–50% and infarct of 30–40% of AAR would be expected. I/R procedures were done on βARKct TG mice and NLC mice. At the end of 30 min reperfusion the mice were euthanized and the hearts removed, frozen, and saved for biochemical analysis. The number of mice used in a particular experiment is provided in the figure legends. Confirmation of transgene expression was assessed in all mice via Western blotting of cardiac extracts.

Cell culture

Neonatal rat ventricular myocytes (NRVMs) were isolated and cultured in HamsF10 as previously described¹⁷. H9c2 cells, HEK 293 cells, and HeLa cells were cultured in DMEM supplemented with 10% bovine calf serum and penicillin-streptomycin in a humidified chamber with 5% CO₂ at 37°C.

Plasmids and transfection

Flag tagged βARKct (GRK2 amino acids 495–689) was created by PCR amplification using full length bovine GRK2 in the pRK5 vector as a template and the appropriate primer to introduce an N-terminal flag tag. Inserts were subcloned into the pRK5 vector by standard molecular biology techniques.

Flag-S670A and S670D mutations in βARKct and GRK2 were made using the Quick-change kit from Stratagene according to the manufacturer’s instructions. Mutations were confirmed by sequencing. Transient transfections were conducted by using 5ug plasmid dna and 10ug Lipofectamine 2000 (Invitrogen)/100mm dish of either HEK293 or HeLa cells. Experiments were conducted on cells 48 hrs post transfection. Adenoviral-mediated infections of NRVM were performed using Adeno-GFP, Adeno-GRK2, Adeno-GRK2S670A, and Adeno-βARKct at an MOI of 10 on the day after isolation and cells were used for experiments 48 hours later. We thank Dr. S. Kornbluth at Duke University for flag-tagged Hsp90 and the Gst-Hsp90 construct was kindly provided by Dr. A. Chadli at the Medical College of Georgia.

Electron microscopy

Immunogold Electron Microscopy was performed as previously described²¹. Briefly, NLC or GRK2-TG mice at five weeks of age were anesthetized via isoflurane inhalation and heparin was injected at 50U/mouse. Hearts were excised and retrogradely perfused with a solution containing 118mol/L NaCl, 4.8mmol/L KCl, 25mmol/L HEPES, 1.25mmol/L K₂HPO₄, 1.25mmol/L MgSO₄, 11mmol/L Glucose (pH 7.4). Mice were perfusion-fixed with 4% paraformaldehyde in 0.1mol/L phosphate buffer (pH 7.4), and hearts dissected. For immuno-electron microscopy, dissected mice hearts were further fixed in freshly prepared 3% paraformaldehyde in 0.1mol/L phosphate buffer containing 0.1% glutaraldehyde and 4% sucrose (pH 7.4). Tissues were washed, dehydrated, embedded in Lowicryl K4M (Polysciences, Inc., Warrington, PA), and polymerized under UV light (360nm) at −35°C. Ultrathin sections were cut and mounted on Formvar-Carbon coated nickel grids. After incubation with primary antibodies at 4°C overnight, gold conjugated secondary antibodies (15nm Protein A Gold, Cell Microscopy Center, University Medical Center Utrecht, 35584 CX Utrecht, The Netherlands; 18nm Colloidal Gold-AffiniPure Goat Anti-Rabbit IgG (H+L), Jackson ImmunoReasearch Laboratories, Inc., West Grove, PA) were applied and stained with uranyl acetate and lead citrate by standard methods. Control staining was performed using heat-inactivated GRK2 antibody. Stained grids were examined under Philips CM-12 electron microscope (FEI; Eindhoven, The Netherlands) and photographed with a (4k x2.7k) digital camera (Gatan, Inc., Pleasanton, CA).

Immunofluorescence

HEK cells seeded onto glass coverslips were co-transfected with plasmids directing expression of mitochondrial-targeted GFP and GRK2. 48 hr after transfection the cells were fixed in 4% paraformaldehyde for 15 min followed by permeabilization for 5 min with 0.1% triton in DPBS-. Cells were briefly washed with DPBS-several times then blocked 90 min using DPBS-containing 1% BSA, 0.05% Tween 20. Anti-GRK2 antibody (Santa Cruz, sc-562) was added at 1:200 dilution in blocking buffer and applied for 90 min with gentle rocking, which was followed by brief washing and addition of anti-rabbit alexa 594 for 90 min. Cells were washed and then coverslips were mounted on glass slides and images acquired using an Olympus Fluoview FV500 confocal laser scanning microscope.

Mitochondria isolation from mouse hearts and cells

Mitochondria were prepared from mouse hearts or cells as previously described²². Briefly, hearts were minced and resuspended in MSH buffer (210mmol/L mannitol, 70mmol/L sucrose, 5mmol/L Hepes, pH 7.5) supplemented with 1mmol/L EDTA, homogenized with a glass/glass dounce homogenizer and centrifuged at 600 × g for 10 min at 4° C. The supernatant was removed, respun at 600 × g for 10 min then removed and centrifuged again at 5500 × g for 20 min at 4° C. The mitochondrial pellet was resuspended in fresh MSH buffer without EDTA and centrifuged again at 5500 × g for 20 min. The washed mitochondrial pellet was resuspended in RIPA lysis buffer if being used for immunoprecipitation experiments, rotated for 1 hr at 4° C, then centrifuged at 13,000 rpm for 10 min at 4° C. The clarified lysate was taken to be the mitochondrial extract and was used for western blotting or immunoprecipitation. In some cases the mitochondrial pellet was solubilized directly into Laemmli loading buffer. When mitochondria were made from cells a 100mm dish containing about 5 × 10⁶ cells was used, cells were washed once with DPBS, scraped into MSH buffer, dounce homogenized and processed as described above.

Proteinase K assay

For each experiment three male C57 mice at 12–14 weeks old were anesthetized with isofluorane and the hearts excised and homogenized in MSH buffer to prepare mitochondria. Mitochondrial exposure to proteinase k was carried out with minor modification as previously described⁸. Briefly, 40µg of mitochondrial protein was incubated on ice for 30 min in MSH buffer containing 20µg/mL proteinase k (Life Technologies). Digestion was stopped by adding PMSF to a final concentration of 2mM. The mitochondrial pellet was isolated by centrifugation and solubilized directly in Laemmli loading dye and analyzed for GRK2 by western blot. Control antibody used for the outer membrane was VDAC (Neuromab N152B/23) and COX5 (Abcam 110263) was used to examine integrity of the inner membrane.

Immunoprecipitations

GRK2 or Hsp90 was immunoprecipitated from heart RIPA lysates, cells, or mitochondrial extracts using anti-GRK2 or anti-Hsp90 antibody from Santa Cruz (cat # sc-562 and sc-13119 for GRK2 and Hsp90, respectively). Control immunoprecipitations (IPs) were conducted using the same antibodies that were heat inactivated prior to use as previously described²³. Flag-tagged βARKct was immunoprecipitated using Flag M2-agarose conjugate or rabbit polyclonal Flag antibody (Sigma-Aldrich). For IPs 1mg of heart lysate was used and for cells clarified RIPA extracts prepared from 5×10⁶ cells was used. For mitochondrial IPs, detergent extracts of mitochondria obtained from a single heart or from 5×10⁶ cells was used. Lysates were incubated with primary antibody for 2 hrs followed by incubation with protein A/G Plus agarose beads (Santa Cruz) for 90 min. Beads were washed 3 times with RIPA and proteins in the immunocomplex were eluted from the beads by the addition of Laemmli loading buffer then heated for 5min at 95° C before loading onto Tris-Glycine gels (Invitrogen).

Immunoblotting and densitometry

Following SDS-PAGE and transfer to nitrocellulose membranes, primary antibody incubations were performed overnight at 4° C. Fluorescent secondary antibodies were obtained from either Molecular Probes or Li-Cor. Membranes were scanned with the Odyssey infrared imaging system (Li-Cor) and the densities of target bands were quantified using the application software of the Odyssey infrared imaging system. Target band density was normalized to the appropriate loading control as described in Figure legends. Primary antibodies used and sources were as follows: anti-GRK2 from Santa Cruz and Millipore, anti-GAPDH from Millipore, anti-VDAC from BD Biosciences, anti-Hsp90 from Santa Cruz, BD Biosciences and Stressgen, anti-cleaved caspase 3 and anti-MAPK (phospho & total) from Cell Signaling, anti-HIS from Santa Cruz, anti-phosphoSer670–GRK2 from Invitrogen, and anti-Flag from Sigma-Aldrich.

In-vitro binding experiments

Hela cells were transfected with either a control plasmid or with flag-tagged Hsp90 and after 48 hrs the mitochondrial fraction was prepared. Equal amounts of mitochondria were incubated with purified his-tagged GRK2 (Invitrogen) in binding buffer composed of PBS with 0.2mmol/L ATP and rotated for 2hr at 4° C. The mitochondria were pelleted by centrifugation and washed with PBS and resuspended directly in Laemmli loading buffer. Western blots were conducted for flag-Hsp90, His-GRK2, and VDAC. In another set of experiments, wild-type and mutant βARKct were transfected into HEK cells and immunoprecipitated from whole cell lysates using flag antibody. The washed beads containing βARKct were incubated with gst-Hsp90 in binding buffer composed of PBS with 1mmol/L DTT and 0.5% Tween20, rotated for 2 hrs at 4° C then the beads were pelleted and washed in the same buffer. The samples were electrophoresed and transferred to membranes for immunoblotting. Western blotting with anti-gst antibody was conducted to determine the amount of Hsp90 bound to βARKct.

Mitochondrial Ca²⁺ uptake assay

Fluorometric Ca²⁺ uptake experiments were performed as previously described²⁴. Briefly, ten to twelve weeks-old male mice were euthanized via CO₂ inhalation. For each experiment, three mice per group were age-matched and processed together for the isolation of mitochondria, which provided enough protein to run the uptake assay in triplicate each time. Hearts were excised and placed in a beaker containing Buffer A (225mmol/L Mannitol, 70mmol/L Sucrose, 1mmol/L EGTA, 10mmol/L HEPES, pH 7.2). Hearts were washed to remove residual blood and then the atria was removed and frozen for Western blot to confirm transgene overexpression. Ventricles were minced into pieces, put in a beaker with 5mg Protease Type XIV (Sigma) in Buffer A and stirred for 7 min and then BSA was added at 0.02g/mL final concentration. The tissue was homogenized with a Teflon-Potter and spun at 1000g for 3 min. Supernatant was filtered through a 100µm mesh and pellet was re-homogenized in a Teflon-Potter in Buffer A, spun at 1000g for 3 min, and filtered through a 100µ mesh. Pellet was discarded and supernatants spun at 10733g for 10 min. Pellets were combined, resuspended in Buffer B (225mmol/L Mannitol, 70mmol/L Sucrose, 10mmol/L HEPES, pH 7.2) and spun at 10733g for 10min. Pellet was resuspended in Buffer B and protein content determined via BCA protein assay. To begin the assay which was conducted at room temperature, respiration buffer (120mmol/L KCl, 70mmol/L Mannitol, 25mmol/L Sucrose, 20mmol/L HEPES, 5mmol/L KH₂PO₄, 0.5mmol/L EGTA) containing 0.1µmol/L Calcium Green 5N (Invitrogen) was added to a quartz cuvette and a zero baseline established. Mitochondria were added at 0.73mg/mL and then 30µmol/L of free calcium was added every 2 min. Measurements (excitation-emission @ 503–535nm) were obtained using a Cary Eclipse Fluorescent Spectrophotometer.

Statistics

Data are expressed as mean ± SE. Statistical significance was determined by an unpaired t-test or ANOVA and Bonferroni test for multiple comparisons using GraphPad Prism software version 5.9. A P value <0.05 was considered significant.

RESULTS

GRK2 localizes to mitochondria and following myocardial ischemic and oxidative stress there is increased mitochondrial GRK2 translocation

Recent data from our lab has shown GRK2 to be pro-death in the heart following ischemic injury¹⁷ and in pursuit of mechanisms for this pro-death signaling we found GRK2 to be present in the mitochondrial fraction of NRVM and hearts (Figure 1A, 1B). Importantly, we confirmed this localization using anti-GRK2 and immunogold electron microscopy of cardiac sections where indeed GRK2 is present within mitochondria (Figure 1C). This was an unexpected finding as GRK2 has been thought to be primarily cytosolic, especially in the heart where it regulates several GPCRs including β-adrenergic receptors that are crucial regulators of cardiac function^{1, 2}. We also conducted immunofluorescence experiments in HEK cells looking at the possible co-localization of GRK2 with a mitochondrial targeted GFP marker protein. The merged confocal image in Online Figure I indicates co-localization. Finally, to further explore the localization of GRK2 in mitochondria we treated isolated mouse heart mitochondria with proteinase k and following Western blot we found that GRK2 was nearly completely digested, and as an outer membrane positive control we found that VDAC was also decreased after treatment, as would be expected. COX 5, a marker for the inner membrane was not altered indicating that GRK2 was primarily associated at the outer membrane, at least basally under control conditions (Online Figure II). While the functional role of GRK2 in mitochondria is not understood, we found that the mitochondria prepared from the hearts of transgenic mice overexpressing GRK2 specifically in the heart¹⁹ had an increased sensitivity to Ca⁺²-induced opening of the mitochondrial permeability transition pore (MPTP), a key step in oxidative stress-mediated cell injury (Figure 1D). As shown in Figure 1D, the Ca⁺² retention capacity, up to the point at which further addition of a Ca⁺² pulse results in precipitous Ca⁺² release due to opening of the MPTP, was decreased by 16.8% in GRK2 transgenic mice (362.98± 20.6 for control and 288.96 ± 16.29 nmoles Ca²⁺/mg protein, for TG, n=3 separate experiments for both groups). The full pathophysiological role for this negative relationship will be further investigated but as we began to explore whether the unique sub-cellular localization of this GRK2 was related to its pro-death signaling in myocytes, we found, interestingly, more GRK2 within cardiac mitochondrial preparations following ischemia/reperfusion (I/R) injury in mice (Figure 1A). Further, oxidative stress in cultured myocytes following chelerythrine treatment also led to increased mitochondrial GRK2 localization (Figure 1B). Importantly, chelerythrine has been shown to induce apoptosis in myocytes through the generation of reactive oxygen species (ROS), and chelerythrine induces apoptosis despite PKC down-regulation, suggesting that PKC inhibition is not the mechanism of chelerythrine-induced apoptosis²⁵. We confirmed that treatment of cardiomyocytes with 10µmol/L chelerythrine for 30 minutes led to increased cleaved caspase-3 formation (Figure 1E). The chelerythrine-induced mitochondrial localization of GRK2 was inhibited by cellular pretreatment with the anti-oxidant N-acetyl-cysteine (NAC), providing further evidence that apoptosis was dependent upon ROS-generation (Figure 1B). NAC also blocked chelerythrine-mediated apoptotic signaling as measured by cleaved caspase-3 levels (Figure 1E).

Figure 1. Basal and oxidative stress-induced localization of GRK2 to mitochondria.

(A) I/R-induced GRK2 translocation to the mitochondrial fraction in wild-type mouse hearts. Mitochondrial fractions were prepared from hearts of sham and I/R treated mice (30 min I and 30 min R) followed by immunoblotting to determine levels of GRK2. Results were normalized to VDAC, which was used as the loading control. *, P<0.05 I/R vs Sham (n=3). (B) Neonatal rat ventricular myocytes (NRVM) were stimulated for 30 min with 10µmol/L chelerythrine in the absence or presence of NAC (5mmol/L, 40 min pretreatment). Mitochondrial fractions were prepared and immunoblotted for GRK2 and VDAC, which was used as the loading control. *, P<0.05 (n=4). (C) Electron micrographs demonstrating GRK2 immunogold labeling within mouse heart mitochondria. Increased localization is seen in hearts from cardiac-specific GRK2 transgenic (TG) mice compared to normal littermate controls (NLC). (D) Ca⁺² retention capacity of mouse heart mitochondria isolated from GRK2-TG and corresponding NLC mice. Representative tracings from each line show the ability of mitochondria to uptake Ca⁺² from repeated challenges before opening of the MPTP occurs. The bar chart shows the quantification of this result indicating Ca⁺² uptake is decreased in cardiac mitochondria from GRK2-Tg mice. *, P<0.05 GRK2-Tg vs. NLC (n=3 separate experiments). (E) NRVM were treated for 30 min with 10µmol/L chelerythrine in the absence or presence of NAC and whole cell lysates were immunoblotted to detect cleaved caspase 3 (CC3). GAPDH was used as the loading control. *, P<0.05 (n=3).

ROS-mediated mitochondrial localization of GRK2 is dependent on Hsp90 binding

The localization of GRK2 to mitochondria was surprising given that it has no targeting consensus sequence, although it should be noted that approximately fifty percent of all mitochondrial proteins do not have mitochondrial target sequences. Therefore, we hypothesized that an interaction with an intermediate protein could be involved in this mechanism. A potential candidate emerged from known GRK2-interacting proteins. GRK2 has been shown to associate with Hsp90⁸ and this heat shock protein is known to be associated with mitochondria²⁶. Importantly, recent data has shown that mitochondrial association of other kinases such as Akt²⁷ and PKCε^{26, 28} are dependent on Hsp90. Indeed, we found a parallel increase in mitochondrial localized Hsp90 and GRK2 in mouse hearts after ischemic injury (Figure 2A) and in cultured myocytes after chelerythrine treatment (Figure 2B). Previously in HL60 cells, GRK2 and Hsp90 have been shown to co-immunoprecipitate and the interaction was blocked by geldanamycin, a known Hsp90 inhibitor⁸. To explore whether this interaction occurs in myocytes after ischemic or oxidative stress, we treated myocytes with chelerythrine as above without and with pre-treatment with geldanamycin. As shown in Figure 2C, we found decreased chelerythrine-stimulated association of both Hsp90 and GRK2 with mitochondria using geldanamycin. We found that GRK2 and Hsp90 associate in whole cell lysates prepared from myocytes (Figure 3A, lane 1) and co-immunoprecipitation between these two proteins also occurs specifically in mitochondrial preparations (Figure 3A, lane 2). We also found that GRK2 and Hsp90 can interact and bind directly using a cell free system. We transfected cells with either a control plasmid or with flag-tagged Hsp90, prepared the mitochondrial fraction from each and then incubated equal amounts of mitochondria with purified 6×HIS-tagged GRK2. In mitochondria prepared from Hsp90-transfected cells significantly more GRK2 was present compared to control mitochondria (Figure 3B). Finally, arguing in favor of a physiological relevance for this interaction, we found that more GRK2 was associated with Hsp90 (via co-immunoprecipitations) in mouse hearts exposed to I/R (30min I, 30 min R) injury and in myocytes when oxidative stress was increased by treatment with chelerythrine (Figure 3C, D).

Figure 2. GRK2 binds to Hsp90 and this interaction is increased by ischemic and oxidative stress.

(A) I/R increases Hsp90 and GRK2 levels in mitochondrial fractions prepared from hearts of sham and I/R treated NLC mice (30 min I and 30 min R) followed by immunoblotting to determine levels of Hsp90 and GRK2. Results are normalized to VDAC, which was used as the loading control for mitochondrial protein. *, P<0.05 I/R vs Sham (n= 3 each group). (B) Mitochondrial fractions from control or chelerythrine-treated NRVMs were analyzed by immunoblotting to determine levels of Hsp90 and GRK2. *, P<0.05 chele vs basal (n=3). (C) NRVMs were treated with 5µmol/L geldanamycin (GA) for 2 hrs and then stimulated for 30 min with 10µmol/L chelerythrine. Mitochondrial fractions were then prepared and immunobloting conducted for GRK2, Hsp90, and normalized to VDAC levels. *, P<0.05 (n= 3).

Figure 3. GRK2 binds to Hsp90 and this interaction is increased by oxidative stress.

(A) Hsp90 was immunoprecipitated from NRVMs whole cell lysates or from purified mitochondrial fractions as indicated followed by immunoblotting for Hsp90 and GRK2 to demonstrate co-precipitation. A representative immunoblot from 3 separate experiments is shown. (B) HeLa cells were transfected with flag-Hsp90 vector or empty plasmid as a control and mitochondrial fractions were prepared and incubated with purified His-tagged GRK2 protein to assess in-vitro binding of GRK2 to Hsp90. After washing, equal amounts of the mitochondrial pellet were immunoblotted using an anti-histidine antibody to detect exogenous GRK2 and a flag antibody was used to detect the transfected Hsp90 in mitochondria. VDAC was used as the loading control to which the results were normalized. *, P<0.05 GRK2 bound to Hsp90 versus empty vector control (n= 3). (C) Wild-type mice were exposed to I/R (30 min I and 30 min R) and then homogenates were prepared using tissue obtained from the ischemic area of the left ventricle. Hsp90 was immunoprecipitated and blotted to detect the amount of GRK2 bound. *, P<0.05 I/R vs Sham (n=3–4 each group) (D) NRVMs were treated for 30 min with 10µmol/L chelerythrine or not treated and whole cell lysates were used for Hsp90 immunoprecipitation. The amount of GRK2 co-precipitated was determined by western blot. A representative immunoblot and quantitation is shown. *, P<0.05 chele vs basal (n=3–4 each group).

ERK phosphorylation of GRK2 is essential for oxidative stress-induced mitochondrial localization and apoptosis

In addition to ROS being necessary for enhanced localization of GRK2 to mitochondria after stress (in vivo studies) we explored other downstream mediators of ischemic and oxidative stress using chelerythrine-stimulated myocytes. We focused initial attention on activation of ERK (p42/44 MAP kinase) since this has been shown to be activated by chelerythrine²⁹ and because ERK can phosphorylate GRK2 specifically at residue Ser670 within its carboxyl terminus¹². When myocytes were treated for 20 min with 10µmol/L chelerythrine we observed an ~1.5-fold increase in the amount of phosphorylated (i.e. activated) ERK (measured as both p-ERK1 and p-ERK2) (Figure 4A), whereas in cells pretreated for 20 min with the MEK (the upstream ERK-activating kinase) inhibitor PD98059, chelerythrine-activated ERK was significantly inhibited (Figure 4A). Therefore, this pathway downstream of chelerythrine-mediated stress is active in myocytes. Importantly, a site-specific antibody that recognizes GRK2 when phosphorylated at Ser670 is available and we examined whether chelerythrine leads to the phosphorylation of GRK2 at this site in myocytes. These experiments were conducted by immunoprecipitating GRK2 from myocytes and then blotting for pSer670-GRK2. As shown in Figure 4B, the amount of p-Ser670-GRK2 was increased more than 2-fold after a 20 min treatment of cells with chelerythrine and this was significantly decreased in the presence of PD95089. Similarly, we found that the chelerythrine-activation of ERK and concomitant phosphorylation of GRK2 at Ser670 was blocked by pretreatment of cells with the anti-oxidant, NAC (Figure 4C).

Figure 4. Chelerythrine-activated ERK phosphorylates GRK2 at serine 670 regulating its Hsp-90 dependent mitochondrial translocation.

(A) NRVMs were stimulated for 30 min with 10µmol/L chelerythrine in the absence or presence of PD98059 (30 min pretreatment, 50µmol/L) and whole cell lysates were analyzed by immunoblotting to detect phospho- and total-ERK. A representative immunoblot is shown along with quantification of the results which are expressed as fold increase over basal *, P<0.05 (n= 3). (B) The amount of chelerythrine-induced phosphorylation of Ser670 within GRK2 was analyzed by immunoprecipitating GRK2 from NRVMs whole cell lysates followed by immunoblotting for both p-Ser670 and total GRK2. Pretreatment and stimulation conditions were the same as in panel A. The p-Ser670 signal at each condition was normalized to the amount of GRK2 immunoprecipitated *, P<0.05 (n= 3). (C) NRVMs were stimulated for 30 min with 10µmol/L chelerythrine in the absence or presence of NAC (5mmol/L, 40 min pretreatment). Whole cell lysates were immunoblotted for ERK activation (p-ERK) and phosphorylation status of GRK2 Ser670. *, P<0.05 NAC vs. control (n=3). (D) Mitochondrial fractions were prepared from chelerythrine-treated NRVMs in the absence and presence of pretreatment with the ERK inhibitor, PD98059. Immunoblotting was used to determine the amount of p-Ser670 and total GRK2 in mitochondria. VDAC is shown as a loading control *, P<0.05 PD vs DMSO treated control myocytes (n=3). (E) Following treatment of NRVMs with chelerythrine, in the presence or absence of pretreatment with PD98059, Hsp90 was immunoprecipitated from whole cell lysates followed by immunoblotting for Hsp90 and GRK2. A representative immunoblot from 3 separate experiments is shown on the left with quantitation showing PD pretreatment decreased the amount of GRK2 interacting with Hsp90 shown on right. *, P<0.05PD vs control.

Based on these experiments we hypothesized that phosphorylation of GRK2 at Ser670 might regulate stress-induced translocation of GRK2 to mitochondria and cell death. Indeed, when cardiomyocytes were pretreated with PD98059 we found that the amount of GRK2 that was localized to mitochondria after treatment with chelerythrine, which normally is increased, was significantly decreased, and the proportion of GRK2 that was in the phospho-Ser670 state within the mitochondrial fraction was also decreased compared to cells in which ERK was not inhibited (Figure 4D). To strengthen our findings we used a different cellular model to examine the pro-death capability of GRK2. Accordingly, we found that cells overexpressing the S670A GRK2 mutant had less H₂O₂-induced death as read-out by the TUNEL assay. As shown in Online Figure III, myocytes were infected with either wildtype GRK2 or the S670A GRK2 mutant and then treated with 75uM H₂O₂ for 16 hours and processed for TUNEL staining. In cells expressing mutant GRK2 the amount of cell death was significantly reduced by 36.4% (6.38% ± 0.60% for wildtype versus 4.05% ± 0.58% for mutant GRK2). Thus, ERK-dependent phosphorylation of GRK2 is required for its full pro-death capacity.

The GRK2-Hsp90 interaction is dependent on GRK2 phosphorylation at Ser670

Interestingly, we found an ERK-dependence to the Hsp90 binding of GRK2 as PD98059 treatment decreased the amount of wild-type GRK2 bound (Figure 4E), which parallels the PD-induced reduction of GRK2 in mitochondrial fractions after chelerythrine above (Figure 4D). Indeed, the directed phosphorylation of GRK2 at Ser670 appears to mechanistically target binding to Hsp90 prior to mitochondrial localization because a mutant GRK2 in which Ser670 was replaced by the non-phosphorylated Ala (GRK2-S670A) displayed significantly less Hsp90 binding in myocytes after oxidative stress compared to wild-type GRK2 (Figure 5A). Thus, non-phosphorylated GRK2 would not be expected to localize to mitochondria after oxidative stress, which is consistent with less death seen with GRK2-S670A overexpression.

Figure 5. MAPK-stimulated phosphorylation of carboxyl-terminus of GRK2 regulates its binding to Hsp90 and translocation to mitochondria.

(A) NRVMs were infected with adeno-GRK2 or –GRK2 S670A and after 48 hrs, cells were either treated or not treated with 10µmol/L chelerythrine for 20 min followed by immunoprecipitation of Hsp90 from the whole cell lysates. Immunobloting was conducted to determine the amount of GRK2 bound to Hsp90. A representative blot is shown, which was identical in 4 independent experiments. (B) HEK293 cells were transfected with an empty vector (control) or βARKct plasmid and then following immunoprecipitation of Hsp90 from whole cell lysates the immune complex was blotted to detect βARKct. (C) HEK293 cells were transfected with a plasmid containing wild-type βARKct or βARKct-S670A and after chelerythrine treatment whole cell lysates were immunoblotted to determine p-Ser670 (upper panel). The lower panel shows equal expression of the wild-type and mutant βARKct protein. Shown is a representative blot of 3 independent experiments with identical results. (D) Differential binding of Hsp90 to wild-type βARKct and Ser670 mutants. Corresponding wild-type and mutant βARKct’s were immunoprecipitated using flag antibody after cell transfections. The washed beads containing βARKct peptides were incubated with gst-Hsp90 and immunoblotted for gst to determine the amount of Hsp90 bound to βARKct. A representative blot from 3 independent experiments is shown (n=3). (E) HEK293 cells were transfected with plasmids containing wild-type βARKct or βARKct-S670A mutant then stimulated with chelerythrine. Mitochondrial fractions were immunoblotted for flag-βARKct to assess translocation of the wild-type and mutant peptides and the results normalized to VDAC. *, P<0.05 βARKct-S670A vs. wild-type βARKct (n= 3). (F) HEK293 cells were transfected with plasmids containing wild-type βARKct or βARKct-S670 or an empty plasmid control and then stimulated with or without chelerythrine. Mitochondrial fractions were prepared and blotted for endogenous GRK2 and VDAC (left panel) to determine the normalized GRK2 translocated to mitochondria when βARKct peptides were expressed. *, P<0.05 βARKct vs. control (n=6).

Ser670 resides within the βARKct and thus we have a powerful molecular tool to further explore the mechanistic role of ERK-mediated GRK2 recruitment to mitochondria. This is even more intriguing since myocyte expression of βARKct leads to cardioprotection in vivo after ischemic injury with significantly decreased myocardial apoptotic signaling¹⁸. Accordingly, using wild-type βARKct as the standard for pro-survival signaling in stressed myocytes we constructed two βARKct mutants, βARKct-S670A and βARKct-S670D, the latter as a potential phospho-mimetic mutant of this GRK2 peptide and studied these in transiently transfected HEK cells. Importantly, we found that wild-type βARKct (flag-tagged) could be immunoprecipitated with Hsp90 in cell extracts showing that the carboxyl-terminus of GRK2 directs binding (Figure 5B). Further, as with full-length GRK2, chelerythrine treatment resulted in enhanced phosphorylation of the βARKct at Ser670 (Figure 5C). On the other hand, and as would be predicted, the βARKct-S670A mutant showed no signal with the p-Ser670 antibody (Figure 5C), a result that also documents the specificity of this antibody (the βARKct contains nine serine residues).

Next, we used an in vitro binding assay to examine whether the phosphorylation of Ser670 influences the Hsp90 and βARKct interaction. Flag-tagged βARKct, βARKct-S670A or βARKct-S670D was immunoprecipitated from whole cell extracts and then purified GST-Hsp90 was added to these samples. As shown in the Western blot in Figure 5D, βARKct-S670D displayed the highest level of Hsp90 binding while βARKct-S670A was lower compared to wild-type βARKct binding. These results support the hypothesis that phosphorylation at this site places the carboxyl terminus of GRK2 in a configuration that is structurally favorable for binding with Hsp90. Moreover, we found that wild-type βARKct, like GRK2, robustly localizes to mitochondria and this is increased after chelerythrine treatment, however, βARKct-S670A mitochondrial localization after oxidative stress is significantly impaired (Figure 5E). This strongly argues for the importance of ERK phosphorylation of this site on GRK2 to direct Hsp90-dependent mitochondrial translocation. To solidify this mechanism we found that expression of wild-type βARKct inhibits the stress-induced mitochondrial localization of endogenous GRK2 while expression of βARKct-S670A did not have this effect (Figure 5F). The major interpretation of this result is that βARKct competes with GRK2 for Hsp90 binding and since βARKct-S670A binds less favorably to Hsp90 the inhibition of GRK2 translocation is lost, allowing for the mitochondrial-dependent pro-death activity of the kinase.

βARKct-mediated cardioprotection in vivo is mediated via less mitochondrial GRK2 localization in response to ERK phosphorylation

We have shown above that βARKct expression prevents GRK2 mitochondrial localization after oxidative stress in HEK cells and this mechanism also robustly occurs in cardiomyocytes (Figure 6A). Consistent with our findings in isolated myocytes, this is also the case in the ischemic heart in vivo as we have found less GRK2 associated with purified cardiac mitochondria after I/R injury in βARKct transgenic mice compared to non-transgenic mice (Figure 6B). βARKct-mediated cardioprotection in ischemic mouse hearts was associated with lower apoptotic signaling¹⁸ and now we find that mechanistically this is due to inhibition of GRK2 mitochondrial localization, as we find that the lack of increased mitochondrial GRK2 accumulation after βARKct expression in myocytes exposed to chelerythrine results in significantly lower cleaved caspase 3 protein levels (Figure 6C). To bolster the hypothesis that less apoptotic signaling in βARKct-expressing myocytes is due to inhibiting GRK2 movement to mitochondria, we found that the βARKct-S670A mutant does not inhibit chelerythrine-mediated caspase-3 activation (Figure 6C), which is a parallel finding showing that this mutant βARKct does not block GRK2 mitochondrial translocation (Figure 5F).

Figure 6. βARKct inhibits stress-induced translocation of GRK2 to mitochondria and decreases myocyte apoptosis in-vitro and in-vivo.

(A) NRVMs were infected with either Adeno-GFP or Adeno-βARKct and after 48 hrs stimulated with 10µmol/L chelerythrine for 30 min. Mitochondrial fractions were then purified and immunoblotted to detect endogenous GRK2 levels and normalized to VDAC. *, P<0.05 βARKct vs GFP (n=3). (B) βARKct TG and NLC mice were exposed to sham or I/R procedures (30 min I and 30 min R) and then mitochondrial fractions were prepared and immunoblotted for GRK2 and VDAC. *, P<0.05 βARKct-Tg vs NLC (n=6 in each group). (C) NRVMs were infected with Adeno-GFP or Adeno-βARKct then treated with 10µmol/L chelerythrine for 30 min. Whole cell lysates were prepared and immunoblotted to detect cleaved caspase-3 with immunoblotted GAPDH serving as the loading control and normalization. *, P<0.05 βARKct vs GFP (n=3). (D) HEK293 cells were transfected with the indicated plasmids and then treated with 10µmol/L chelerythrine for 30 min. Levels of cleaved caspase-3 and GAPDH were determined by immunoblotting using whole cell lysates and normalized to immunoblotted GAPDH. *, P<0.05 (n=4).

We also have positively linked ERK-mediated phosphorylation of GRK2 to pro-death signaling and pathophysiology after ischemic stress as we found that in cardioprotected βARKct TG mice after I/R injury there is not only less mitochondrial GRK2 (Figure 6B) but there also is significantly less p-Ser670 in the mitochondrial GRK2, despite equal post-I/R cardiac ERK activation (Figure 7A). Interestingly, Ser670 of the βARKct transgenic peptide expressed in βARKct transgenic mice was robustly phosphorylated after I/R (Figure 7A) suggesting that after ERK activation this peptide binds Hsp90 more strongly and prevents GRK2 from inducing mitochondrial-mediated death. Of further importance to our hypothesis, these results were reproduced in cultured myocytes treated with chelerythrine (Figure 7B).

Figure 7. Ischemic and oxidative stress induced ERK phosphorylation of GRK2 is decreased in the presence of βARKct in-vitro and in-vivo.

(A) Following I/R conducted on NLC or βARKct TG mice, hearts were removed and tissue lysates prepared from the ischemic area of the LV, or in the sham group, from the LV area at risk. Immunoblotting was conducted for the indicated proteins and this representative blot indicates a reduction in p-Ser670 phosphorylation within GRK2 and a parallel increase in p-Ser670 within the βARKct. *, P<0.05 GRK2 p-Ser670 levels in βARKct post-I/R mice than in NLC post-I/R mice (n=5 mice per groups). (B) NRVMs were infected with Adeno-GFP or Adeno-βARKct and 48hr later were treated for 20min with 10µmol/L chelerythrine. Whole cell lysates were the prepared and blotted for the indicated proteins shown on the representative blot. Cells expressing βARKct had significantly less pSer670 levels in endogenous GRK2 compared to GFP expressing cells. *, P<0.05 GRK2 p-Ser670 levels in βARKct infected myocytes compared to GFP infected myocytes (n=3).

DISCUSSION

Classically, GRKs phosphorylate agonist-occupied seven-transmembrane spanning receptors, known as GPCRs, and initiate an uncoupling process known as desensitization. The desensitization mechanism includes targeted binding of β-arrestins to GRK-phosphorylated receptors which blocks G protein activation. However, β-arrestins also direct G protein-independent signaling by recruiting signaling molecules themselves³⁰, opening a new area of GPCR signaling that is in addition to the desensitization process. Similarly, evidence is mounting that GRKs can regulate important cellular processes beyond membrane-bound GPCRs, a finding that has gained traction largely as the result of unexpected cellular localization and protein interactions for this family of kinases. Here, we show that GRK2, the most abundant GRK expressed in the myocardium, associates with the outer membrane of mitochondria. The most significant findings from our study are that oxidative stress results in translocation and an increased amount of GRK2 associated mitochondria and we have identified the signaling details that account for this previously unexplored cardiac localization. Further, stressed-induced mitochondrial localization of GRK2 in myocytes promotes pro-death signaling, which adds significant new evidence that GRK2 is a key pathological component in the injured heart. For the last two decades we have used various cellular and animal models to show that the increased GRK2 expression found in failing human myocardium³¹ is a maladaptive molecular alteration that can occur very acutely after stress^{1, 2}. GRK2 is acutely elevated in the myocardium post-MI^32–34 and elevated GRK2 has been shown to promote myocyte death through apoptotic signaling, although other mechanisms of death, such as necrosis or necroptosis should not be ruled out¹⁷. Our current results show that in addition to promoting dysregulation of β-adrenergic receptor signaling in the injured heart, increased GRK2 expression will result in more GRK2 within mitochondria, which appears to be a nodal locale for this kinase in promoting its pro-death activities in the cardiomyocyte following ischemic and oxidative stress.

The discovery of unique cellular localizations and interactomes for a given protein can provide the starting point for the design of studies that could ultimately lead to uncovering new functional roles in cellular signaling. As mentioned earlier, previous studies have found that within the GRK family, certain members have specific cellular localization characteristics. For example, GRK2/GRK3 have been found to be primarily cytosolic enzymes that associate with the sarcolemmal membrane after agonist-occupancy of a GPCR, while GRK5/6 are more avidly associated with the membrane at all times^{2, 35}. However, these previous notions are being challenged, and as an example, GRK5/6 have a functional nuclear localization signal (NLS) within their catalytic domains³⁶ and GRK5 has been shown to reside in the nucleus of myocytes^{14, 37}. This nuclear localization of myocardial GRK5 has uncovered a novel role for this GRK as a Class II HDAC kinase where its enhanced activity in the nucleus can promote maladaptive myocyte hypertrophy and heart failure^{14, 15}.

Although we do not fully understand why GRK2 is associated with mitochondria under basal conditions, we have found that increased mitochondrial GRK2 promotes MPTP opening to a lower threshold of Ca²⁺, which in itself demonstrates GRK2 promoting an altered mitochondrial phenotype. We have identified in this report that GRK2 has crucial partners that direct its stress-induced translocation to mitochondria, including ERK MAP kinase and Hsp90. Of note, without phosphorylation of GRK2 post-ROS formation by ERK, specifically at Ser670, there is no mitochondrial localization leading to cell death as this modification of GRK2 is required for it to bind to Hsp90 which directs its recruitment to the primary pro-death organelle within myocytes. Hsp90 is one of several heat shock proteins that function as molecular chaperones involved in the assurance of correct target protein folding and assembly, and interestingly Hsp90 appears to be more discriminating in terms of client interaction. Evidence suggests a preference for interactions with protein kinases³⁸, and indeed, kinases including Pim-1³⁹, Akt⁴⁰, and PKCε²⁸ bind to Hsp90. Previously, an interaction between Hsp90 and GRK2 was identified with a potential role in the maturation of this kinase⁸. However, our data now uncovers a second functional outcome of this interaction and that is mitochondrial targeting, which is consistent with Hsp90 being identified as an important mitochondrial protein in a proteomic analysis of mouse mitochondria²⁴.

Since most mitochondrial proteins are nuclear encoded and synthesized in the cytosol, the finding that molecular chaperones Hsp90 and Hsp70 deliver immature proteins to the mitochondrial import receptor Tom70⁴¹ was a key discovery that appears to have broader implications in cellular functions, especially in cardiomyocytes after ischemic stress and this previously unappreciated mechanistic role for GRK2. Interestingly, Hsp90 can also play a critical role in cell survival by binding to components of the permeability transition pore and antagonizing its opening, thus preserving organelle integrity and inhibiting the onset of cell death^{42, 43}. Our data suggests that in myocytes there appears to be a more complex relationship between Hsp90 and mitochondrial-mediated death, because the ROS-induced increased localization of GRK2 to mitochondria is associated with increased death signaling. Future studies that are beyond the scope of the present study will examine whether GRK2 specifically antagonizes the pro-survival effects of Hsp90 in mitochondria, as this could occur through a direct inhibition or perhaps GRK2 can antagonize the anti-death effects of other known Hsp90 binding partners in the heart.

The mitochondrial uptake of several cardioprotective proteins, including connexin⁴⁵, Akt and PKCε have been shown to entail Hsp90-mediated import^{27, 28, 44, 45}. Interestingly, a previous study in liver endothelial cells has shown that when GRK2 is elevated it binds to and inhibits Akt in the cytoplasm⁷, suggesting the potential for this interaction to take place at the level of Hsp90 binding or perhaps at mitochondria. Certainly, our mitochondrial-dependent pro-death signaling ascribed to GRK2 in this study is consistent with potential antagonism of the pro-survival effects of Akt, and although we did not specifically find a direct interaction between these two kinases in mitochondrial fractions of myocytes it is possible that the pro-death effects of GRK2 in ischemic myocytes opposes any benefit of Akt within mitochondria. Interestingly, in cardiomyocytes Akt was also shown to phosphorylate mitochondrial hexokinase, a mitochondrial outer membrane protein and protect mitochondria from oxidant or calcium-induced stress⁴⁶. Further studies will be conducted to identify any potential substrates for GRK2 within mitochondria.

In addition to Hsp90 being required for mitochondrial localization of GRK2, we found that an absolute requirement was the ERK-mediated phosphorylation of GRK2 at Ser670 as the phosphorylation status at this site was a key determinant of Hsp90 binding and subsequent mitochondrial accumulation. This ascribes a specific and critical role for the regulation of GRK2 by this MAP kinase within cardiomyocytes and death signaling. Previous studies have shown that phosphorylation of GRK2 by ERK at Ser670 is involved in protein stability and promoted degradation of this kinase^{12, 47}. Similarly, this appears to be the case for the homologous GRK3 after ERK phosphorylation⁴⁸. A more recent study has shown that Ser670 phosphorylation by the cell cycle kinase, CDK2, may be more crucial for GRK2 degradation within its context as a cell cycle regulator⁴⁹. Further, in neuronal cells, phosphorylation of GRK2 at Ser670 by PI3K promoted degradation¹³. Thus, although Ser670 may be the key target to promote GRK2 degradation, differential regulation appears to occur via specific kinases and our data would suggest that ERK-mediated Ser670 phosphorylation targets GRK2 to mitochondria. Importantly, targeting of GRK2 to mitochondria after this phosphorylation event would result in a subsequent loss of cytoplasmic GRK2 consistent with degradation, however without knowledge of the unique cellular localization of GRK2 uncovered in our report this aspect of GRK2 regulation was not studied. A number of studies indicate that activation of the reperfusion injury salvage kinase (RISK) pathway at the time of injury is cardio-protective⁵⁰. ERK1/2 is one of the kinases involved in this pathway and in these studies the beneficial effects of ERK were identified based on the finding that administration of growth factors or other agents at the time of reperfusion resulted in protection through phosphorylation and inhibition of pro-apoptotic proteins. In contrast to the more narrow set of signals which would be triggered by growth factors, ischemic stress activates a more complex set of signaling molecules and the final cellular response derives from the integration of beneficial and pro-death signals. In our study we show ERK activation being involved in the pro-death mechanism of GRK2, however, this finding should not be interpreted as negating other protective aspects of ERK in accordance with the RISK hypothesis.

A recent paper demonstrated mitochondrial localization of GRK2 in HEK and endothelial cells and mitochondria with decreased GRK2 had less ATP production, suggesting that increased GRK2 is beneficial¹⁶. Using GRK2 truncation mutants it was reported that the amino-terminus of GRK2 associated with mitochondria but that the carboxyl-terminus did not. We focused on the carboxyl-terminus because it contains the essential serine residue that, when phosphorylated, regulates its interaction with Hsp90. It is possible that multiple domains within GRK2 could regulate association with mitochondria, such as during basal or oxidative-stress conditions, although given the absence of a mitochondrial targeting sequence within GRK2, the lack of mechanistic detail as to how GRK2 associates with mitochondria through the amino-terminus of the protein is a limitation of their study. Alternatively, cardiac mitochondrial targeting could be different. An important difference between our studies is we show that increased GRK2 localized to cardiac mitochondria is a damaging phenotype and provide mitochondrial Ca²⁺ uptake data to support this. Our data are also consistent with the physiological evidence that we and others have published showing that increased GRK2 levels in the heart are deleterious in the context of ischemic stress.

There are several potential limitations of our study. We measured I/R (30 min I/ 30 min R) – induced translocation of GRK2 to mitochondria and the signaling mechanisms involved but didn’t measure infarct size in a similarly treated cohort, therefore we weren’t able to correlate myocardial injury with translocation of GRK2 to mitochondria. However, using the same I/R injury model, GRK2-Tg mice had significantly greater 24 hr infarcts compared to NLC mice¹⁷. Further, while the mechanism of cell death in mouse models of ischemic injury, such as I/R, remains controversial, it is likely that both apoptosis and necrosis is involved but in this study we focused only on apoptotic signaling. We focused on apoptosis because a prior study¹⁷ reported that mice subjected to 30 minutes of ischemia and 3 hours of reperfusion had increased ventricular caspase 3 activity which is indicative of apoptosis, and the cell death mechanisms in our study would likely be similar to those reported in that study. Accordingly, we treated NRVMs with chelerythrine, which also increased caspase 3 activity, as a model to study the biochemical signaling pathways that would be activated during in vivo injury in the mouse heart. There could be differences in the relative contribution that apoptosis and necrosis play in adult compared to neonatal myocytes exposed to ischemic stress. For example, one study reported that isolated adult myocytes exposed to 6 hrs hypoxia and 6 hrs reperfusion had increased cytochrome c release and caspase activity compared to control cells⁵⁰, which supports apoptotic mechanisms in the adult cell, although a different study showed changes in the level of proteins important in the assembly and formation of the apoptosome in adult compared to neonatal myocytes, supporting the view that adult myocytes are more resistant than neonatal myocytes to mitochondrial driven apoptosis⁵¹. Thus, the relationship between apoptosis and necrosis and the relative importance of each in neonatal and adult cell death during I/R injury continues to evolve and future studies will help resolve these controversies.

Of potential therapeutic importance, our current data assigns a new beneficial mechanism for βARKct’s action in the injured heart as sequences within this peptide direct phosphorylation-dependent binding to Hsp90, and the binding of this expressed and phosphorylated peptide blocks endogenous GRK2 binding to Hsp90 and subsequent mitochondrial localization preventing GRK2-mediated pro-death signaling. We have recently shown that βARKct expression in the infarcted heart significantly reduces cellular apoptosis¹⁷ and in this study we found that Ser670 is critical for βARKct-mediated cardioprotection, because we discovered that pro-death signaling after oxidative stress in myocytes was not blocked by a βARKct mutant that lacks the ERK phosphorylation site. This non-protective βARKct mutant appears to not prevent endogenous GRK2 from binding to Hsp90, thus GRK2 is free to translocate to the mitochondria and promote caspase activation.

In conclusion, our study is the first to report cellular association of GRK2 with mitochondria of cardiomyocytes, and that this localization increases upon ischemic and oxidative stress, events that promote its pro-death signaling. These findings add significant mechanistic information for the maladaptive effect of GRK2 that occurs within the compromised myocardium. Importantly, we ascribe a novel regulatory feature to Ser670 within the carboxyl terminus of GRK2, in that following ROS-dependent ERK phosphorylation there is increased binding to Hsp90 and subsequent translocation to mitochondria. Our results are clear that ERK phosphorylation and Hsp90 dependent mitochondrial localization of GRK2 in ischemic cardiomyocytes promotes non-GPCR mediated pro-death effects of this GRK. This is an apparent key role for GRK2 in the pathogenesis of heart failure and increases the therapeutic importance of GRK2 inhibition as a strategy to combat heart disease.

Novelty and Significance.

What Is Known?

• G-protein coupled receptor kinases (GRKs) phosphorylate activated receptors, which turns off cellular signaling in a process known as desensitization.

• GRK2 is the predominant GRK isoform in the heart and in human heart failure the levels of GRK2 are elevated.

• Recent studies identified non-receptor binding partners for GRK2 suggesting that GRK2 regulates cellular signaling through mechanisms independent of receptor desensitization.

• A peptide comprised of the last 194 amino acids of GRK2 (known as βARKct) is cardioprotective in rodent models of cardiac injury while GRK2 activity promotes cell death after myocardial ischemia.

What New Information Does This Article Contribute?

• GRK2 can be found associated with mitochondria in cardiomyocytes.

• Myocytes in vitro and in vivo exposed to ischemic stress injury cause more GRK2 localization to mitochondria.

• Translocation of GRK2 to mitochondria following ischemic stress is mediated by ERK phosphorylation of GRK2 at residue Ser670 and subsequent binding to Hsp90, which is a known mitochondrial chaperone.

• Mitochondrial localization of GRK2 following ERK phosphorylation appears to promote its pro-death activity as mutation of GRK2 at Ser670 to Ala results in lower apoptosis after oxidative stress.

• βARKct-mediated cardioprotection occurs by competing with and inhibiting endogenous GRK2 binding to Hsp90, thus decreasing the amount of GRK2 at mitochondria.

• Mitochondria with increased GRK2 have decreased calcium uptake capacity, which may mechanistically explain the damaging phenotype associated with increased GRK2 in myocytes.

The classical view of GRK2 as a kinase involved mainly in G protein-coupled receptor signaling is quickly changing. Instead, the idea of a more dynamic GRK2 with multiple roles in regulating cell signaling is emerging. In this paper, we show that GRK2 is not only present at the mitochondria within cardiomyocytes but that its localization to this organelle is altered in response to ischemic injury. Oxidative stress after injury activates multiple signaling molecules including ERK1/2 resulting in the phosphorylation of GRK2 in the cytoplasm, leading to increased binding to Hsp90, which chaperones GRK2 to mitochondria. We also show that mitochondria prepared from mice that overexpress GRK2 in the heart have worsened mitochondrial calcium handling. Further, we show that mice having cardiac-specific expression of a peptide, known as βARKct, fare better after ischemic injury, and this is mechanistically associated with decreased mitochondrial GRK2 translocation/localization. In summary, we elucidate a novel signaling mechanism in the heart in which GRK2 translocates to mitochondria after ischemic stress and acts as a pro-death kinase.

Nonstandard Abbreviations

βARKct

carboxyl-terminus of beta-adrenergic receptor kinase

COX5

cytochrome c oxidase 5

ERK

extracellular signal-regulated kinase

GPCR

g protein-coupled receptor

GRK2

g protein-coupled receptor kinase 2

GST

glutathione S-transferase

GFP

green fluorescent protein

Hsp90

heat shock protein 90

HEK

human embryonic kidney

HDACs

class II histone deactylases

I/R

ischemia-reperfusion

MI

myocardial infarction

MPTP

mitochondrial permeability transition pore

NRVM

neonatal rat ventricular myocyte

NAC

n-acetyl-cysteine

PKC

protein kinase C

PI3K

phosphoinositide-3-kinase

RISK

reperfusion injury salvage kinases

ROS

reactive oxygen species

S670A

serine to alanine mutation at amino acid 670 in grk2

S670D

serine to aspartic acid mutation at amino acid 670 in grk2

VDAC

voltage-dependent anion channel

WT

wild-type