AMP-activated Protein Kinase (cid:2) 2 and E2F1 Transcription Factor Mediate Doxorubicin-induced Cytotoxicity by Forming a Positive Signal Loop in Mouse Embryonic Fibroblasts and Non-carcinoma Cells *

Background: Despite the central position in chemotherapy, the clinical application of doxorubicin is compromised by severe adverse effect in different organs. Results: Doxorubicin induces AMPK (cid:2) 2 transcription, and AMPK (cid:2) 2 in turn stabilizes E2F1 in non-carcinoma cells. Conclusion: AMPK (cid:2) 2 and E2F1 mediate cytotoxicity of doxorubicin. Significance: AMPK (cid:2) 2 might serve as a novel target for alleviating the cytotoxicity of doxorubicin. is the drugs, but its clinical application is compromised by severe adverse effects in different organs including cardiotoxicity. In the present study we explored mechanisms of doxorubicin-in-duced cytotoxicity by

Doxorubicin is one of the most widely used anti-cancer drugs, but its clinical application is compromised by severe adverse effects in different organs including cardiotoxicity. In the present study we explored mechanisms of doxorubicin-induced cytotoxicity by revealing a novel role for the AMPactivated protein kinase ␣2 (AMPK␣2) in mouse embryonic fibroblasts (MEFs). Doxorubicin robustly induced the expression of AMPK␣2 in MEFs but slightly reduced AMPK␣1 expression. Our data support the previous notion that AMPK␣1 harbors survival properties under doxorubicin treatment. In contrast, analyses of Ampk␣2 ؊/؊ MEFs, gene knockdown of AMPK␣2 by shRNA, and inhibition of AMPK␣2 activity with an AMPK inhibitor indicated that AMPK␣2 functions as a pro-apoptotic molecule under doxorubicin treatment. Doxorubicin induced AMPK␣2 at the transcription level via E2F1, a transcription factor that regulates apoptosis in response to DNA damage. E2F1 directly transactivated the Ampk␣2 gene promoter. In turn, AMPK␣2 significantly contributed to stabilization and activation of E2F1 by doxorubicin, forming a positive signal amplification loop. AMPK␣2 directly interacted with and phosphorylated E2F1. This signal loop was also detected in H9c2, C2C12, and ECV (human epithelial cells) cells as well as mouse liver under doxorubicin treatment. Resveratrol, which has been suggested to attenuate doxorubicin-induced cytotoxicity, significantly blocked induction of AMPK␣2 and E2F1 by doxorubicin, leading to protection of these cells. This signal loop appears to be non-carcinoma-specific because AMPK␣2 was not induced by doxorubicin in five different tested cancer cell lines. These results suggest that AMPK␣2 may serve as a novel target for alleviating the cytotoxicity of doxorubicin.
Doxorubicin is one of the most effective anti-cancer drugs and is widely used to treat solid tumors and hematological malignancies. However, the clinical use of doxorubicin has been regarded as a double-edged sword (1). Despite its central position in chemotherapy, severe adverse effects in different organs including the heart, brain, kidney, and liver have impeded clinical application of the drug. The most common side effects of doxorubicin are cumulative cytotoxicity, neurological disturbances, liver injury, and bone marrow aplasia (1)(2)(3)(4). Although multiple mechanisms including oxidative stress, DNA damage, mitochondrial impairment, and induction of the apoptotic pathway have been implicated in doxorubicin cytotoxicity (5), the principal mechanisms are not fully understood. Therefore, elucidating the mechanisms underlying doxorubicin cytotoxicity is critical to efforts to increase the utility of doxorubicin for treating cancers.
AMP-activated protein kinase (AMPK) 2 is a heterotrimeric protein kinase that consists of a catalytic subunit (␣) and two regulatory subunits (␤ and ␥), each with multiple isoforms, and plays a central role in cellular energy homeostasis (6). In response to various metabolic stressors, AMPK is activated by the accumulation of AMP as a result of ATP depletion and, in turn, regulates a multitude of metabolic pathways to balance cellular energy (6 -9). In addition to balancing cellular energy, AMPK also induces apoptosis through several tumor suppres-* This work was supported by the National Research Foundation of Korea grant funded by the Korea government (MSIP; 2012009447 and 20120009381). 1 To whom correspondence should be addressed: Dept sors including LKB1, TSC2, and p53. LKB1 is an upstream activating kinase of AMPK (10), and TSC2 and p53 are the direct substrates of AMPK (11)(12)(13). Therefore, AMPK seemingly has dual roles, such as protecting cells from energy depletion and inducing cellular apoptosis, but the underlying mechanisms by which AMPK determines cellular fate remain an enigma. In contrast to the well established role of AMPK in metabolic regulation, there is no consensus on its role or the underlying mechanism under genotoxic stress. Some reports have demonstrated that AMPK is activated in H9c2 (14) and cancer cell lines (15) by doxorubicin. In contrast, AMPK is inhibited by long term treatment of doxorubicin in mouse embryonic fibroblasts (MEFs) (16). Moreover, there is a sharp contrast regarding the role of AMPK in the regulation of cell death under doxorubicin treatment (14,(17)(18)(19). In general, the aforementioned reports have been conducted without distinguishing the potentially differential roles of the AMPK␣ isoforms. Two isoforms, AMPK␣1 and AMPK␣2, are encoded by two distinct genes (20). AMPK␣1 is ubiquitously expressed, whereas AMPK␣2 is highly expressed in metabolically active tissues including heart, skeletal muscle, and liver (21). In fact, AMPK␣ isoforms may have differential roles in the regulation of metabolism and/or blood vessel contraction (6,(22)(23)(24)(25)(26). Therefore, we determined whether each AMPK␣ subunit isoform was differentially associated with cell fate under doxorubicin treatment to reveal novel mechanisms for doxorubicin-induced cytotoxicity.
Here, we report that doxorubicin robustly induced AMPK␣2 expression, whereas it slightly reduced the level of AMPK␣1 in MEFs. In accordance with a previous report (16), our data support that AMPK␣1 harbors survival properties. Moreover, we provide novel evidence that AMPK␣2 and E2F1 mediate doxorubicin cytotoxicity by forming a positive signal loop, revealing the differential role of AMPK␣ isoforms. Doxorubicin induced AMPK␣2 transcription via E2F1, and AMPK␣2 in turn stabilized E2F1 in MEFs. The transcription factor E2F1 is stabilized and activated by DNA damage (27) and transactivates a number of genes including p27 kip1 , Apaf-1, and Bim in the apoptotic pathway (28 -30). This signal loop significantly contributed to cell death in several non-carcinomas including H9c2, C2C12, and ECV cells as well as mouse liver under doxorubicin treatment. Notably, AMPK␣2 was not induced by doxorubicin in cancer cells. Collectively, our results suggest that AMPK␣2 may serve as a novel target for alleviating the cytotoxicity of doxorubicin.
Cell Viability Assay-The Vi-CELL XR Cell Viability analyzer (Beckman Coulter, Fullerton, CA) that performs an automated trypan blue exclusion assay was used to measure cell viability. The assay is based on uptake of trypan blue dye by dead cells due to loss of membrane integrity. A 1-ml aliquot of cell suspension in a plastic cuvette was aspirated and mixed with trypan blue and then pumped into the flow cell for imaging. The instrument collected 50 images of cells to compute viability. Dead cells appear darker than viable cells due to different contrast between live and dead cells.
Reporter Gene Assay-Cell were seeded onto 24-well culture plates at 4 ϫ 10 4 cells/well and incubated for 24 h. Plasmids were transfected into cells using JetPEI (Polyplus Transfection, Illkirch, France) according to the manufacturer's instructions. A 1:1 ratio between reporter gene constructs and expression vector was used for co-transfections. After 24 h of transfection, luciferase activity was determined by mixing 20 g of cell extract with 100 l of luciferase assay reagent, and relative light units were measured for 10 s in a luminometer (TD-20/20 luminometer, Turner Designs, Sunnyvale, CA).
Establishment of Stable Cell Lines-MEFs were transfected with the pLKO.1 vector expressing shRNA for E2F1 or the pGIPZ vector expressing shRNA for AMPK␣1 and -␣2. Stable cells were isolated in media 24 h after the transfection with 3 g/ml puromycin for 2 weeks. Additionally, MEFs were transfected with the pCMV vector expressing E2F1-WT or E2F1⌬TA. After a 24-h transfection, positive-expressing clones were selected with G418 (300 g/ml) after about 2 weeks of selection.
Annexin V-FITC Assay by Flow Cytometry-The percentages of apoptotic cells were determined by flow cytometry using the annexin V-FITC Apoptosis Detection Kit Plus (BioVision, Inc) according to the manufacturer's instructions. Briefly, MEF cells were harvested and resuspended in a binding buffer. Cells (10 6 / ml) were mixed with 5 l of annexin V-FITC. After incubation at room temperature for 15 min in the dark, the stained cells were analyzed using fluorescence-activated cell sorting (BD FACS Canto II, BD Biosciences) reading the emission selected by 530 nm (FITC).
In Situ Proximity Ligation Assay (in Situ PLA)-The proximity ligation assays were performed using the Duolink II detection kit (Olink Bioscience, Uppsala, Sweden). After transfection with the indicated expression vectors or exposure to doxorubicin, the cells were fixed in 4% paraformaldehyde for 5 min at room temperature. After washing with PBS, the cells were treated with 0.1% Triton X-100 in PBS for 3 min at room temperature. To reduce the nonspecific signals, the cells were incubated with a blocking solution for 30 min at 37°C. Primary antibody diluents containing two primary antibodies (1:500 dilutions of anti-FLAG rabbit antibodies and anti-GFP mouse . Under these conditions, the expression level of the indicated proteins and the phosphorylated form of H2A.X was examined via Western blot analyses. Total AMPK␣ protein was also measured (AMPK ␣1/␣2). AMPK␣1 and AMPK␣2 mRNA levels were measured by real-time PCR, and a set of primers that recognize both AMPK␣1 and AMPK␣2 (AMPK ␣1/␣2) was used for PCR to measure the mRNA level of total AMPK␣. The -fold induction is expressed as a ratio of GAPDH mRNA (D). Results are the means Ϯ S.E. n Ն 3.

AMPK␣2 and E2F1 Mediate Doxorubicin Cytotoxicity
antibodies for transfection study or 1:250 dilutions of anti-E2F1 rabbit antibodies and anti-AMPK␣2 goat antibodies for detecting endogenous proteins) were added to the cells and incubated overnight at 4°C. Then the PLA Probes (Detection Reagents Green or Detection Reagents Red) were added and incubated for 2 h at 37°C. The images of the cells were acquired using an LSM 700 Confocal microscope (Carl Zeiss, Jena, Germany).
In Vitro AMPK Kinase Assay-E2F1 was overexpressed and purified as a fusion protein with glutathione S-transferase in Escherichia coli and used as a substrate for in vitro AMPK assay. The AMPK complex was overexpressed in HEK293 cells by cotransfection of HA-AMPK␣2, Myc-AMPK␤1, and Myc-AMPK␥1, immunoprecipitated with anti-HA antibody, eluted with HA peptide, and further purified by desalting with desalting spin column. The recombinant GST-E2F1 and AMPK␣2 complexes were incubated in the kinase assay buffer (62.5 mM HEPES, pH 7.0, 62.5 mM NaCl, 62.5 mM NaF, 6.25 mM sodium pyrophosphate, 1.25 mM EDTA, 1.25 mM EGTA, and 1 mM dithiothreitol containing 200 M AMP) containing 10 M cold ATP and 10 Ci of [␥-32 P]ATP at 37°C for 1 h. The reaction was terminated by adding SDS sample buffer, and the reaction mixtures were subjected to SDS-PAGE and autoradiography. GST-SAMS contains the AMPK phosphorylation peptide sequence of acetyl-CoA carboxylase (HMRSAMSGLHLVKRR) and used as a positive control for AMPK activity.
Statistical Analysis-Results are expressed as the means Ϯ S.E. We used Student's t test and Prism 5 software. Differences were considered significant at a p value of Ͻ 0.01. **, p Ͻ 0.01; ***, p Ͻ 0.001.

AMPK␣2 Expression Is Induced in Response to Doxorubicin
in MEFs-MEFs were exposed to various DNA damaging agents or to glucose-free medium for 12 h, and then the expression levels of the AMPK␣ subunits were examined by Western blot analysis (Fig. 1A). AMPK␣2 was hardly detectable in MEFs under basal conditions, but the DNA damaging agents including H 2 O 2 , doxorubicin, etoposide, and ionizing radiation dramatically induced expression of AMPK␣2 as well as a DNA damage marker, phosphorylation of H2AX. Under these conditions, the level of AMPK␣1 was unaffected or decreased slightly. In contrast to the DNA damaging agents, metabolic stress such as glucose starvation had no effect on AMPK␣2 expression, indicating that AMPK␣2 is specifically induced by DNA damage. Indeed, doxorubicin robustly induced AMPK␣2 expression in a time-and dose-dependent manner, but the level of AMPK␣1 decreased (Fig. 1, B and C). E2F1 was simultaneously induced by doxorubicin (Fig. 1, A-C). The AMPK␣2 mRNA level increased ϳ15-fold by doxorubicin treatment (Fig.  1D). AMPK␣1 is a predominant form in wild type MEF under basal conditions. As a net sum of the decreased AMPK␣1 and the increased AMPK␣2 under doxorubicin treatment, the protein and mRNA amount of total AMPK␣ appears to be same or slightly increased (Fig. B-D). We examined the effect of doxorubicin in subsequent studies.
AMPK␣2 Exerts Pro-apoptotic Properties under Doxorubicin Treatment-To investigate the function of each AMPK␣ subunit, we compared the cellular sensitivity of MEFs from Ampk␣ wild type, Ampk␣1 Ϫ/Ϫ deletion, and Ampk␣2 Ϫ/Ϫ deletion to doxorubicin. When cell viability was measured directly, Ampk␣2 Ϫ/Ϫ MEFs were most resistant to doxorubicin-induced cytotoxicity, whereas Ampk␣1 Ϫ/Ϫ MEFs showed the increased sensitivity to doxorubicin-induced cytotoxicity compared with wild type MEFs ( Fig. 2A). Fluorescence-activated cell sorting analysis of AMPK deletion MEFs with annexin V staining (Fig.  2B) and the comparison of apoptotic markers including the cleaved form of poly(ADP-ribose) polymerase and caspase-3 among AMPK␣ isoform-specifically gene knocked-down MEFs (Fig. 2C) revealed similar results showing that AMPK isoforms exert opposing effects on doxorubicin-induced apoptosis. Our data suggest that AMPK␣1 is anti-apoptotic, whereas AMPK␣2 is pro-apoptotic in response to doxorubicin. In fact, the previous report demonstrated that AMPK␣1 plays the protective role against doxorubicin-induced apoptosis in MEFs and H9c2 cells (16), but the role of AMPK␣2 remains unknown. Therefore, we primarily focused on revealing a novel role of AMPK␣2 and the underlying mechanisms throughout the present study. To unequivocally demonstrate the role of AMPK␣2, we next examined the specific role of AMPK␣2 in Ampk␣1 Ϫ/Ϫ MEFs. AMPK␣2 was still induced by doxorubicin in Ampk␣1 Ϫ/Ϫ MEFs, and knockdown of AMPK␣2 in Ampk␣1 Ϫ/Ϫ MEFs diminished the apoptotic response to doxorubicin (Fig. 2D). Compound C, an AMPK inhibitor, effectively blocked the AMPK␣2 activity in Ampk␣1 Ϫ/Ϫ MEFs after doxorubicin treatment, as judged by the phosphorylation level of acetyl-CoA carboxylase (ACC) Ser 79 (32), the most well characterized AMPK substrate (Fig.  2E), leading to attenuation of the Ampk␣1 Ϫ/Ϫ MEF apoptotic response to doxorubicin (Fig. 2E) and alleviation of the Ampk␣1 Ϫ/Ϫ MEF doxorubicin cytotoxicity (Fig. 2F). In conclusion, when the induction of AMPK␣2 by doxorubicin was hampered by genetic, molecular, and pharmacological approaches, MEFs became resistant to doxorubicin-induced apoptosis and cytotoxicity, indicating that AMPK␣2 exerts pro-apoptotic properties. The pro-apoptotic role of AMPK␣2 was further demonstrated; Ampk␣1 Ϫ/Ϫ ␣2 Ϫ/Ϫ DKO MEFs were transfected with vector expressing the wild type (WT) and dominant negative form (DN) of AMPK␣2 and then treated with doxorubicin (Fig. 2G). The percentage of apoptotic cells under doxorubicin treatment, as  assessed by annexin V staining, significantly increased in the presence of AMPK␣2-WT but not AMPK␣2-DN.
E2F1 Is a Transcriptional Regulator of AMPK␣2 Gene Expression under Doxorubicin Treatment-E2F is a transcription factor that regulates expression of a subset of genes involved in cell proliferation and apoptosis (27)(28)(29)(30)33). A recent report demonstrated that AMPK␣2 is downstream of E2F1 when the phosphatidylinositol 3-kinase signal is blocked; AMPK␣2 is induced in a E2F1-dependent manner under serum-free conditions or with phosphatidylinositol 3-kinase inhibitor treatment (34). E2F1 is stabilized by DNA damage and induces a number of genes involved in apoptosis (28 -30). As E2F1 was induced by doxorubicin in MEFs (Fig. 1), we examined whether AMPK␣2 was induced by doxorubicin in a E2F1dependent manner. First, we constructed wild type functional E2F1 (FLAG-E2F1-WT) and non-functional E2F1 (FLAG-E2F1-⌬TA), in which the transactivation domain was deleted. The activity of E2F1-WT and E2F1-⌬TA was determined by cotransfection with a luciferase reporter vector containing four-repeated E2F1-responsive element ([E2F1 ϫ 4]-Luc) in HEK293 cells; E2F1-WT induced a 2-fold increase in luciferase activity, whereas E2F1-⌬TA showed no activity (Fig. 3A). Then, we established MEFs stably expressing E2F1-WT and E2F1-⌬TA. Overexpression of E2F1-WT, but not E2F1-⌬TA, resulted in a significant induction of AMPK␣2, the protein (Fig.  3B) and mRNA level (Fig. 3C). The known target genes of E2F1 including p27 kip1 , Bim, and Apaf-1 was also induced by E2F1-WT (Fig. 3, B and C). Moreover, AMPK␣2 and Bim expression was profoundly induced by doxorubicin when E2F1-WT, but not E2F1-⌬TA, was overexpressed (Fig. 3D). In contrast to overexpression, knockdown of E2F1 expression by shRNA distinctively blocked doxorubicin-induced expression of AMPK␣2, Bim, and apoptosis biomarkers, indicating that E2F1 acts upstream of AMPK␣2 with doxorubicin treatment (Fig. 3E). We next cloned a human AMPK␣2 promoter (ϳ2.7 kb) into the pGL3 luciferase reporter vector, and cotransfection with E2F1-WT in HEK293 cells resulted in a ϳ14-fold induction of luciferase (Fig. 3F). Cotransfection with E2F1-WT in wild type MEFs induced a ϳ3-fold increase in AMPK␣2 promoter activity, and this induction was further increased by doxorubicin, but E2F1-⌬TA or E2F1-⌬DBD, in which DNAbinding domain was deleted, failed to induce AMPK␣2 promoter activity (Fig. 3G). We also observed the induction of AMPK␣2, E2F1, and its target genes including Bim and ULK1 and the pro-apoptotic markers such as Bax and cleavage of caspase-3 in mouse liver from animals injected with doxorubicin through the tail vein (Fig. 3H). Collectively, these results indicate that AMPK␣2 gene expression is induced by E2F1 in response to doxorubicin and that both E2F1 and AMPK␣2 show pro-apoptotic properties.
AMPK␣2 Positively Regulates E2F1 Protein Stability-We next explored the possibility of feedback regulation between AMPK␣2 and E2F1. Overexpression of wild type AMPK␣2, but not its dominant negative form, in wild type MEFs (Fig. 4A) as well as in Ampk␣1 Ϫ/Ϫ ␣2 Ϫ/Ϫ DKO MEFs (Fig. 4B) induced endogenous E2F1 and its target gene expression. We compared the effect of AICAR, a pharmacological activator of AMPK, on the degree of endogenous E2F1 induction in Ampk␣1 Ϫ/Ϫ MEFs and Ampk␣1 Ϫ/Ϫ ␣2 Ϫ/Ϫ DKO MEFs to investigate the specific role of AMPK␣2. AICAR induced phosphoacetyl-CoA carboxylase and E2F1 in Ampk␣1 Ϫ/Ϫ MEFs, but not in Ampk DKO MEFs, without altering E2F1 mRNA levels (Fig. 4C). Moreover, E2F1 induction and subsequent expression of its target genes by doxorubicin were markedly diminished in Ampk␣ DKO MEFs compared with Ampk␣1 Ϫ/Ϫ MEFs, in which AMPK␣2 was still induced by doxorubicin (Fig. 4D). The AMPK␣2 isoform-specific regulation for E2F1 was further demonstrated by AMPK␣ gene knockdown approaches; AMPK␣2 down-regulation by shRNA, but not AMPK␣1 down-regulation, significantly blocked doxorubicin-induced E2F1 expression in wild type MEFs (Fig. 4E). These data raised the possibility that AMPK␣2 positively regulates E2F1 induction at the post-transcriptional level because the E2F1 mRNA level was not affected. To further confirm this possibility, wild type MEFs were cotransfected with FLAG-tagged E2F1 and AMPK␣2, and overexpression of wild type AMPK␣2 resulted in a significant induction of exogenous E2F1 (Fig. 4F) as well as the transcriptional activity (Fig. 4G). We next transfected FLAG-tagged E2F1 into Ampk␣1 Ϫ/Ϫ MEFs and Ampk␣ DKO MEFs, and the GFP expression vector was cotransfected as a control to assess transfection efficiency. The protein level of exogenously introduced E2F1 was ϳ8-fold higher in Ampk␣1 Ϫ/Ϫ MEFs than that in Ampk␣ DKO MEFs, whereas the GFP level was essentially the same under these conditions (Fig. 4H). We further determined the role of AMPK␣2 in regulation of the half-life of the E2F1 protein. After transfection with FLAG-E2F1, Ampk␣1 Ϫ/Ϫ MEFs were treated with cycloheximide, an inhibitor of new protein synthesis, in the absence or presence of compound C, and the result indicated that E2F1 was rapidly degraded in the presence of the AMPK inhibitor (Fig. 4I). These data collectively suggest that AMPK␣2 positively regulated E2F1 protein stability. As E2F1 induces AMPK␣2 expression at the transcriptional level, these two molecules form a positive signal amplification loop in response to doxorubicin.
AMPK␣2 Directly Interacts with and Phosphorylates E2F1-To examine whether AMPK␣2 directly interacts with E2F, we FIGURE 3. E2F1 regulates gene expression of AMPK␣2 in the presence of doxorubicin. HEK293 cells were cotransfected with FLAG-E2F1-WT or E2F1⌬TA and with [E2F1]x4-Luc, and then luciferase activity was measured (A). Wild type MEFs stably expressing FLAG-E2F1-WT and Fag-E2F1⌬TA were established, and then the indicated protein (B) and mRNA levels (C) were measured. These cells were exposed to doxorubicin (1 M, 6 -12 h), and the expression of AMPK␣2 and Bim were compared (D). Two clones of wild type MEFs stably expressing E2F1 shRNA were isolated, and the indicated protein level was examined in the presence of doxorubicin (0.1 M, 24 h) (E). A human AMPK␣2 promoter (ϳ2.7 kb) was cloned into the pGL3 luciferase reporter vector. This vector was cotransfected into HEK293 cells with FLAG-E2F1-WT or FLAG-E2F1⌬TA, and then the luciferase assay was performed (F). After performing an identical cotransfection in wild type MEFs, the cells were exposed to doxorubicin (DOX; 1 M, 12 h), and the luciferase assay was performed (G). Data represented are the means Ϯ S.E. n Ն 3. Four-week-old male Slc:ICR mice were injected with doxorubicin (5 mg/kg) through the tail vein at 3-day intervals, whereas the control mice were injected with a comparable volume of water. The mice were sacrificed after 15 days, liver lysates were extracted, and Western blot analysis was performed (H). FEBRUARY 21, 2014 • VOLUME 289 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 4845
cotransfected HEK293 cells with E2F1 and Myc-tagged AMPK␣2 or GFP-tagged AMPK␣2 expression vector and then performed a co-immunoprecipitation assay. In a complex precipitated with E2F1 antibody, the presence of Myc-tagged AMPK␣2 or GFP-tagged AMPK␣2 was confirmed (Fig. 5A). Likewise, E2F1 was also detected in a complex precipitated with AMPK␣2 antibody (Fig. 5A). The proximity ligation assay (PLA) also visualized the in situ protein complex between AMPK␣2 and E2F1 (Fig. 5B). PLA is a combination of immunohistochemistry and rolling cycle amplification between adjacent, oligonucleotide-coupled secondary antibodies (35). The protein complex formation between endogenous AMPK␣2 and E2F1 was also detected in AMPK wild type MEFs, and doxorubicin treatment significantly enhanced the protein complex formation (Fig. 5, C and D). Moreover, AMPK␣2 directly phosphorylated GST-E2F1 fusion protein in vitro but not GST alone (Fig. 5E). These data suggest that AMPK␣2 forms a complex with and phosphorylates E2F1 in response to doxorubicin treatment.

Resveratrol Alleviates Doxorubicin Cytotoxicity by Suppressing the AMPK-E2F1 Signal Loop in MEFs and Non-carcinoma
Cells-A number of reports have suggested that resveratrol attenuates doxorubicin-induced cytotoxicity, but precise mechanisms have not been substantiated (36 -39). We next examined the effect of resveratrol on the AMPK-E2F1 signal loop. Wild type MEFs were exposed to doxorubicin in the presence of resveratrol and structurally similar flavonoids including piceatannol and fisetin. These flavonoids significantly blocked doxorubicin-induced expression of E2F1, AMPK␣2, and poly-(ADP-ribose) polymerase cleavage (Fig. 6A). Accordingly, these flavonoids blocked the cytotoxicity induced by doxorubicin (Fig. 6B). A recent report demonstrated that E2F1 activity is down-regulated by Sirt1 deacetylase (40). Because resveratrol can activate Sirt1 (39, 41), we tested whether Sirt1 is involved in the resveratrol action on E2F1 and AMPK␣2. Wild type MEFs and Sirt1 Ϫ/Ϫ MEFs were exposed to doxorubicin in the absence or presence of resveratrol, and the results showed that resveratrol continued to block doxorubicin-induced AMPK␣2 and E2F1 in Sirt1 Ϫ/Ϫ MEFs (Fig. 6C). Sirt1 Ϫ/Ϫ MEFs were more sensitive to doxorubicin-induced apoptosis than that of the wild type, as assessed by the level of cleaved poly(ADP-ribose) polymerase and caspase-3 (Fig. 6C). However, resveratrol decreased the apoptotic markers to a similar degree in both cells, when comparing a long exposure of wild type with short exposure of Sirt1 Ϫ/Ϫ MEFs (Fig. 6C). These data suggest that the AMPK-E2F1 signal loop is a novel target for resveratrol action, but Sirt1 is not likely to be involved in this regulation.
AMPK␣2 Is Not Induced by Doxorubicin in Cancer Cells-We finally examined the effect of doxorubicin on AMPK␣2 and E2F1 in cancer cells. Five different cancer cell types were exposed to the indicated concentrations of doxorubicin for 12 h. Interestingly, E2F1 was still induced by doxorubicin in these cancer cells, but no AMPK␣2 induction was observed (Fig. 8A). Moreover, doxorubicin killed cancer cells regardless of the AMPK␣2 level; AGS cells did not express AMPK␣2, but doxorubicin decreased AGS cell viability as effectively as that in other cancer cells (Fig. 8B). Transfection of wild type AMPK␣2 into AGS cells did not alter the E2F1 expression level or the apoptotic response to doxorubicin (Fig. 8C). Collectively, our data suggest that AMPK␣2 does not form a positive signal loop with E2F1 in cancer cells, further suggesting that a signal loop between AMPK␣2 and E2F1 quite specifically operates in noncarcinoma cells.

DISCUSSION
Multiple mechanisms have been implicated in the development of doxorubicin toxicity including redox damage, but none have been fully substantiated. Administration of compounds harboring anti-oxidant or anti-apoptotic activity has been attempted to reduce doxorubicin cytotoxicity, but some of these attempts failed in clinically relevant animal models and clinical trials (4,42,43). Moreover, the idea of redox damage was challenged by a recent study showing that electron transport chain damage is a major mechanism for doxorubicin cytotoxicity (44). Indeed, despite 40 years of research, there is no effective approach to alleviate doxorubicin cytotoxicity. In the present study we revealed a novel signal pathway that potentially serves as a molecular target for alleviating doxorubicin cytotoxicity in non-carcinoma cells; AMPK␣2 formed a positive signal amplification loop with the E2F1 transcription factor after doxorubicin treatment and mediated doxorubicin cytotoxicity in MEFs.
In the present study we first observed that the AMPK␣2 transcript was dramatically induced by DNA damage including doxorubicin in MEFs (Fig. 1). The major regulatory mechanism for AMPK is short term regulation. AMPK enzyme activity is allosterically regulated by the AMP and ATP cellular ratio (7,8). In addition, AMPK␣ is phosphorylated at Thr 172 and activated by upstream kinases including LKB1, CaMKK (CaM kinase kinase), and TAK (6,9,(45)(46)(47). These two mechanisms have been intensively investigated and largely explain the rapid change in AMPK activity in response to a number of metabolic stressors. However, little information is available on the  FEBRUARY 21, 2014 • VOLUME 289 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 4847
changes in AMPK transcription except that AMPK␣ expression level is tissue-specific; AMPK␣1 is present in every tissue, whereas AMPK␣2 is highly expressed in heart, skeletal muscle, and liver (21). Herein, we report that the AMPK␣2 mRNA level was induced dramatically by doxorubicin in MEFs (Fig. 1), and we further identified E2F1 as a transcription factor for AMPK␣2 during doxorubicin treatment. E2F1 directly increased AMPK␣2 promoter activity and mRNA levels, and gene knockdown of E2F1 under doxorubicin treatment significantly blocked AMPK␣2 induction (Fig. 3). Our data suggest that long term regulation such as transcriptional regulation may be critical for regulating AMPK under genotoxic stress in addition to short term regulation.
Our data also highlight the specific and differential function of AMPK␣ isoforms. Inhibition of AMPK␣2 activity by genetic deletion, gene knockdown, or pharmacologically alleviated doxorubicin cytotoxicity suggests that AMPK␣2 functions as a pro-apoptotic molecule, whereas AMPK␣1 contributes to cell survival under this condition in accordance with a previous report (16) (Fig. 2). The role of each isoform may vary depending on the stress conditions. Indeed, gene knock-out animal studies suggest that the functions of these isoforms are quite different for regulating metabolism as well. When Ampk␣2 Ϫ/Ϫ mice are fed a high fat diet, derangements occur in the regulation of glucose tolerance, insulin sensitivity, and weight gain occurred, whereas these changes were not detected in Ampk␣1 Ϫ/Ϫ mice (23)(24)(25)(26). In contrast, endothelial nitric-oxide synthase activation and blood vessel relaxation are observed in Ampk␣1 Ϫ/Ϫ mice, but not in Ampk␣2 Ϫ/Ϫ mice, in response to the AMPK activator (6,22). At present the precise mechanisms that confer a differential role for each isoform remain almost unknown, although differences in tissue distribution, cellular localization, and substrate specificity of AMPK isoforms have been reported (21,48,49). Therefore, we suggest that future studies should consider the potentially differential role of AMPK␣ isoforms, and such approaches will reveal the more FIGURE 5. AMPK␣2 interacts with E2F1. HEK293 cells were cotransfected with the Myc-tagged or GFP-tagged AMPK␣2 together with pCMV-E2F1, and then immunoprecipitation assay using E2F1 antibody or AMPK␣2 was performed (A). HEK293 cells were seeded on coverslips and cotransfected with GFP-AMPK␣2 and FLAG-E2F1, and then in situ PLA was performed using anti-GFP antibody, anti-FLAG antibody, and PLA probes as described under "Experimental Procedures" (B). After wild type MEFs or Ampk␣1 Ϫ/Ϫ ␣2 Ϫ/Ϫ MEFs were treated with doxorubicin (DOX; 1 M, 12 h or indicated time), and a coimmunoprecipitation assay (C) or in situ PLA assay (D) was performed. In vitro kinase assay was performed; AMPK␣2/␤1/␥1 complex was incubated with GST, GST-SAMS, or GST-E2F1 for 1 h in the presence of [␥-32 P]ATP and then subjected to autoradiography or Western blot (E, left panel). The purified AMPK␣2/␤1/␥1 complex was subjected to Western blot analysis (E, right panel). GST-SAMS was used as a positive control for AMPK activity. IP, immunoprecipitation; WB, Western blot.  FEBRUARY 21, 2014 • VOLUME 289 • NUMBER 8 precise AMPK-involved cellular events under metabolic and genotoxic stressors.

AMPK␣2 and E2F1 Mediate Doxorubicin Cytotoxicity
Our results are summarized in a diagram (Fig. 9). In initial response to DNA damage, it is well known that E2F1 is stabilized by phosphorylation via ATM, ATR, or Chk2 (27,50,51) or by acetylation via p300/CREB-binding protein-associated factor (P/CAF) (52). Our data propose that the stabilized E2F1 after DNA damage acts as an upstream transcription factor for AMPK␣2, promoting the transcription of AMPK␣2 (Fig. 3), which further contributes to E2F1 stability (Fig. 4). The pro-  Five different cancer cell lines, including HCT116, A549, AGS, DU145, and HeLa, were exposed to the indicated concentration of doxorubicin (DOX) for 12 h, and the levels of AMPK␣2 and E2F1 were measured (A). These cells were exposed to doxorubicin for 48 h, and cell viability was measured (B). AGS cells were transfected with AMPK␣2 WT, and then the levels of E2F1 and apoptotic markers were compared. PARP, poly(ADP-ribose) polymerase. apoptotic role of E2F1 in response to DNA damaging agent has been well established. Our data suggest that AMPK␣2 contributes to induction of apoptosis by stabilizing E2F1 by direct phosphorylation. Doxorubicin enhanced the direct protein interaction between AMPK␣2 and E2F1, and furthermore AMPK␣2 directly phosphorylated E2F1 in vitro (Fig. 5). Although we do not have direct evidence at this time, phosphorylation of E2F1 by AMPK␣2 may be critical for E2F1 stability because inhibition of AMPK␣2 activity resulted in rapid protein degradation of E2F1 (Fig. 4I). Characterizing the phosphorylation site(s) of E2F1 by AMPK␣2 will provide us with more detailed information on the underlying mechanisms for E2F1 regulation by AMPK␣2. The tight association between these two molecules was further demonstrated by pretreatment of flavonoids including resveratrol, piceatannol, and fisetin; these flavonoids dramatically blocked the induction of both AMPK␣2 and E2F1 and protected MEFs in the presence of doxorubicin (Fig. 6). This signal loop was also observed in several other cell lines including H9c2, C2C12, and ECV, and resveratrol blocked the signal loop, resulting in reduced doxorubicin cytotoxicity in these cells as well (Fig. 7). We also observed the induction of AMPK␣2, E2F1, and its target genes, apoptotic markers in mouse liver from animal injected with doxorubicin (Fig. 3H).
Notably, doxorubicin did not induced AMPK␣2 in five different cancer cell lines under our experimental conditions despite E2F1 induction (Fig. 8A). Moreover, it was difficult to identify a significant correlation between the expression level of AMPK␣2 and the cellular response to doxorubicin in the tested cancer cells (Fig. 8, B and C). We currently do not understand how AMPK is differentially regulated in normal and cancer cells, but an intensive analysis of the AMPK␣2 gene promoter including epigenetic difference or characterization of E2F1 response element in normal and cancer cells may provide a clue. Collectively, our data suggest that a signal amplification loop between AMPK␣2 and E2F1 in the presence of doxorubicin appears to be non-carcinoma cell-specific. As this signal loop mediates doxorubicin cytotoxicity in non-carcinoma cells and serves as a novel target for resveratrol action, more studies on this signal pathway will provide a better understanding of the nature of doxorubicin cytotoxicity.