Inhibition of AMP-activated Protein Kinase Sensitizes Cancer Cells to Cisplatin-induced Apoptosis via Hyper-induction of p53*

Cisplatin is one of the most effective and widely used chemotherapeutic agents. However, one of the most salient limitations to the clinical application of cisplatin is the acquired or intrinsic drug resistance exhibited by some tumors. In the present study, we have assessed the potential of an intracellular energy balancing system as a target for augmentation of cisplatin sensitivity in tumors. AMP-activated protein kinase (AMPK) regulates the energy balance system by monitoring intracellular energy status. Here we demonstrate that AMPK is rapidly activated by cisplatin in AGS and HCT116 cancer cells. The inhibition of AMPK in those cells and in xenografts of HCT116 resulted in a remarkable increase in cisplatin-induced apoptosis, which was associated with hyper-induction of the tumor suppressor p53. We further showed that ERK, but not ATM (ataxia telangiectasia mutated) and ATR (ATM- and Rad3-related) kinases, was involved in the hyper-induction of p53 by the inhibition of cisplatin-induced AMPK. By way of contrast, cisplatin did not induce AMPK activation in HeLa cells, which appear to have a relatively high sensitivity to cisplatin-induced cytotoxicity, but expression of the constitutive active form of AMPK in HeLa cells resulted in a significant increase of cell viability after cisplatin treatment. Collectively, our data suggest that AMPK performs a pivotal function for protection against the cytotoxic effect of cisplatin, thereby implying that AMPK is one of the cellular factors determining the cellular sensitivity to cisplatin. On the basis of these observations, we propose that a strategy combining cisplatin and AMPK inhibition could be developed into a novel chemotherapeutic modality.

The introduction of cisplatin into clinical trials constitutes a milestone achievement in cancer medicine; cisplatin has since been employed as a first-line chemotherapeutic modality in the treatment of epithelial malignancies, including lung, ovarian, testicular, head-and-neck cancer, and others (1,2). Despite its central position in chemotherapy regimens, however, the clin-ical applications of the drug have been impeded to some degree by the acquired or intrinsic resistance of certain tumor cell variants to the drug (3). Therefore, insight into cellular responses to cisplatin, as well as the mechanism determining drug sensitivity, is critical to efforts to augment the utility of cisplatin in the treatment of cancers.
The principal mechanism of cisplatin-induced damage to tumors involves the induction of apoptosis (3), and its mode of cytotoxicity appears to be initiated as the consequence of DNA damage induced by its interaction with DNA, which activates several signal transduction pathways that culminate in apoptosis. Key molecules for the initiation of the DNA damage response include the ATM 2 (ataxia telangiectasia mutated) and ATR (ATM-and Rad3-related) kinases (4), both of which are members of a family of nuclear kinases known to be essential for the transduction of genotoxic stress signals to the p53 tumor suppressor protein. Upon the infliction of DNA damage, ATM and ATR phosphorylate p53 at Ser 15 (5)(6)(7), which then inhibits the interaction between p53 and Mdm2, leading to p53 stabilization (8). Under normal conditions, the levels of p53 tend to be low, owing primarily to Mdm2-mediated ubiquitin-dependent proteolysis. Depending on the intensity of genotoxic stresses, p53 performs a pivotal function in the action of a cell to undergo either DNA repair or apoptosis (9). When DNA damage exceeds the cellular repair capacity and p53 levels exceed a critical threshold, the cells almost invariably undergo apoptosis (3,10). This is achieved, in part, via the induction of p53 target genes, including Bax, Noxa, and Puma, all of which are known to be pro-apoptotic.
In the present study, as an attempt to develop a novel approach for the augmentation of cisplatin sensitivity in tumors, we have assessed the potential of an intracellular energy-balancing system as a target. To this end, we have determined the effects of cisplatin on AMP-activated protein kinase (AMPK) in cancer cells, as well as its role in cisplatin-induced apoptosis, both in vitro and in vivo. AMPK is a heterotrimer that consists of a catalytic subunit (␣) and two regulatory subunits (␤ and ␥), and operates as an intracellular energy sensor via the monitoring of the cellular AMP/ATP ratio (11). It is activated sensitively in response to a broad spectrum of physiological and pathological ATP-depleting conditions, including exercise, hypoxia, nutrition deprivation, oxidative stress, and exposure to carcinogenic metals. Moreover, recent studies have shown that the AMPK pathway performs a central function in the regulation of glucose and lipid homeostasis, body weight, food intake, insulin signaling, and mitochondria biogenesis (12).
By way of contrast, studies regarding the role of AMPK in cancer are in an embryonic stage at present. Moreover, the effects of genotoxic stress, including the effects of anti-cancer drugs on AMPK or its role in genotoxic stress-induced apoptosis, remain almost completely unknown. In this work, we examined the role of AMPK in cisplatin-induced apoptosis in AGS (human gastric carcinoma), HCT116 (human colon carcinoma), and HeLa (human cervix adenocarcinoma) cells as well as in a human carcinoma xenograft. We also assessed a relevant signaling pathway involving ATM, ATR, and p53. On the basis of the observations given herein, we propose that AMPK is one of the cellular factors determining the cellular sensitivity to cisplatin, further implying that a strategy in which cisplatin and AMPK inhibition were coupled could be developed for a novel chemotherapeutic approach.
Cell Culture and Treatment-AGS (human gastric carcinoma), HCT116 (human colon carcinoma), HeLa S3 (human cervix adenocarcinoma), and GM05849 cells were maintained in RPMI supplemented with 10% heat-inactivated fetal bovine serum and antibiotics at 37°C with 95% air and 5% CO 2 . Cells were exposed to cisplatin in culture media containing 10% fetal bovine serum. The pretreatment of cells with compound C, caffeine, and PD98059 was conducted for 20 min prior to the addition of cisplatin. HCT116 p53ϩ/ϩ and HCT116 p53Ϫ/Ϫ cells were provided by Dr. Bert Vogelstein (Howard Hughes Medical Institute and The Sidney Kimmel Comprehensive Cancer Center, The Johns Hopkins Medical Institutions, Baltimore, MD). GM05849 (ATM Ϫ/Ϫ ) cells were purchased from the Coriell Institute (Camden, NJ).
Cell Viability Assay-The Vi-CELL TM XR Cell viability analyzer (Beckman Coulter) cell counter that performs an auto-mated 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 their membrane integrity. One milliliter of aliquot of the cell suspension in plastic cuvette was aspirated and mixed with trypan blue and then pumped into the flow cell for imaging. The instrument collected 100 images of cells to compute viability. The dead cells appear darker than the viable cells allowing the contrast between live and dead cells to be used in determining cell viability.
Assessment of Cell Apoptosis-Cell apoptosis was assessed via fluorescence-activated cell sorting (FACS) analysis, Hoechst 33342 staining and internucleosomal DNA fragmentation analysis. Total cells were harvested by trypsinization, collected by centrifugation, and washed with PBS. After fixing with 70% ethanol, cells were resuspended in PBS containing 10 g/ml propidium iodide. After sorting out the viable cells, fluorescence intensity was measured by FACSCalibur flow cytometry and CellQuest software (BD Biosciences) using excitation and emission wavelengths of 488 and 525 nm, respectively. Internucleosomal DNA fragmentation was measured as previously described (13). Chromatin was stained with Hoechst 33342 as previously described (14). Cells evidencing condensed chromatin or fragmented nuclei were considered apoptotic.
Measurement of Reactive Oxygen Species-Cells were seeded onto 12-well micro plates on coverglasses. After stimulation, cells were incubated with 10 M 2Ј,7Ј-dichlorofluorescein diacetate (Sigma) for 30 min and washed with PBS, and fluorescence was measured with FACS.
Adenovirus Infection and Plasmid Transfection-c-myctagged AMPK wild-type ␣ subunit (WT), a dominant negative form (DN), and constitutively active form (CA), were generated, prepared, and purified as described previously (15). Infections with adenovirus expressing the AMPK wild type (Ad-AMPK-WT), AMPK dominant negative form (AD-AMPK-DN), or constitutively active form (Ad-AMPK-CA) were conducted in PBS for 30 min at 37°C, after which fresh medium was added. Plasmids were transfected into cells using Cytopure TM (Qbiogene) according to the manufacturer's instructions. AMPK-WT, AMPK-DN, or AMPK-CA in pcDNA was constructed as described previously (15). pG13-luc was provided by Dr. G. Lozano (The University of Texas M. D. Anderson Cancer Center).
ATP Analysis-Cells were washed with ice-cold PBS, and lysis buffer (0.5% Triton X-100, and 2 mM CaCl 2 in PBS) was added to the cells. Intracellular ATP was measured via the luciferin/luciferase method using an ATP Determination Kit (Molecular Probes). The assay buffer (100 l), which contained 0.5 mM luciferin, 1.25 g/ml luciferase, 25 mM Tris, pH 7.8, 5 mM MgSO 4 , 100 M EDTA, and 1 mM dithiothreitol, was mixed with the cell lysates (5 l). Luminescence was analyzed by VICTOR 3TM luminometer (PerkinElmer Life Sciences) and normalized using the cellular proteins.
Animals and in Vivo Anti-tumor Assay-Athymic BALB nu/nu mice were utilized in this study. All animal experiments were approved by the Ethics Committee for Animal Experimentation of the Korea Institute of Radiological & Medical Sciences. Human colon carcinoma cells (HCT116) were injected into the flanks of 5-to 6-week-old nude mice. Five mice were assigned to each of the experimental groups. Intraperitoneal injections of all drugs at 4-day intervals was initiated after the tumor achieved a minimal weight of 200 mg. Tumor weights were evaluated in accordance with the formula (L ϫ l 2 )/2 via the measurement of tumor length (L) and width (l) with a set of calipers.
Statistical Analysis-Results are expressed as the means Ϯ S.E. We used the Student t test. Differences were considered significant at a p value of Ͻ0.05.

Cisplatin Rapidly Induces AMPK Activation in AGS Cells via
Generation of Reactive Oxygen Species-Because gastric carcinoma is one of the most intractable cancers and is known for resistance to cisplatin (16), we used AGS gastric cancer cells in the present study as a working model to investigate a novel mechanism determining the cisplatin sensitivity. To examine the effect of cisplatin on AMPK activity, AGS cells in culture media containing 10% serum were exposed to cisplatin in a time-and dose-dependent manner (Fig. 1A). Cisplatin (60 M) rapidly activated AMPK in AGS cells, and its activation persisted for 6 h, as was evidenced by the phosphorylation level at Thr 172 in the active site of AMPK␣ and Ser 79 of acetyl-CoA carboxylase (ACC), a well characterized AMPK cellular substrate (17) (Fig. 1A, left panel). Because AMPK activation by cisplatin was rapid, we next examined the concentration-dependent effect on AMPK activity after 1-h treatment of cisplatin (Fig. 1A, right panel). Indeed, cisplatin activated AMPK in a dose-dependent manner in AGS cells.
To understand the mechanisms for cisplatin-induced AMPK activation, we measured reactive oxygen species (ROS) levels in AGS cells, because AMPK is also known to be sensitively activated by ROS (18), and cisplatin is known to induce ROS levels (3). Cisplatin (120 M, 1 h) increased ROS generation by ϳ2-fold in AGS cells (Fig. 1B), and anti-oxidants such as N-acetylcysteine (NAC) or catalase significantly blocked cisplatininduced AMPK activation (Fig. 1C), indicating that ROS is a mediator for AMPK activation by cisplatin.
Inhibition of AMPK Activity Sensitizes AGS Cells to Cisplatin-induced Apoptosis-To ascertain whether AMPK activity is essential for cell viability after treatment with cisplatin, we have taken a pharmacological and molecular approach to inhibit the AMPK activity, and then we examined the subsequent effect on cell viability via using several different approaches. Compound C is a potent and selective AMPK inhibitor (19), and it has been widely used to test a role of AMPK. Pretreatment of AGS cells with 20 M compound C for 30 min almost completely prevented the cisplatin-induced AMPK activation, as evidenced by the phosphorylation level of ACC-Ser 79 ( Fig. 2A). We also attempted to confirm our results by using recombinant adenovirus expressing the c-myc-tagged wild-type (Ad-AMPK-WT), dominant negative (Ad-AMPK-DN), or constitutively active (Ad-AMPK-CA) form of AMPK (15). In accordance with the previous report showing that formation of AMPK trimeric complex (␣, ␤, and ␥ subunits) is required for its activity (20,21), overexpression of only the wild-type ␣ subunit did not activate endogenous AMPK activity, as evidenced by the phosphorylation level of ACC-Ser 79 . However, overexpression of AMPK-DN, which replaces the endogenous ␣ subunit in the trimeric complex (20), effectively blocked the cisplatin-induced phosphorylation level at ACC-Ser 79 , whereas AMPK-CA increased its level ( Fig. 2A). We further examined the phosphorylation level of AMPK␣ Thr 172 , but there was no difference in its level between cells overexpressing AMPK␣ WT and DN ( Fig. 2A). Because AMPK-DN contains a mutation in ATP binding site, Thr 172 residue is still available for phosphorylation by the upstream kinase (20).
AGS cells infected by Ad-AMPK-WT, -DN, or -CA were exposed to cisplatin for 24 h, cell apoptosis was assessed via fluorescence-activated cell scanning analysis (sub-G1 fraction), and the results showed that cell apoptosis increased ϳ10% when cisplatin was applied in AGS cells infected with Ad-AMPK-WT. However, the infection with Ad-AMPK-DN in AGS cells did, indeed, result in an augmentation of cisplatinmediated apoptosis (Fig. 2B), suggesting that AMPK plays a protective role in cisplatin-induced cytotoxicity. The increase in cisplatin-induced apoptosis was also verified by Hoechst 33342 staining; distinctive nuclear condensation was observed when AMPK activity was inhibited by infection with Ad-AMPK-DN (Fig. 2C). This result was further supported by DNA fragmentation analysis; a ladder pattern of internucleosomal fragmentation of DNA was apparent when AGS cells infected with Ad-AMPK-DN were treated with cisplatin for 24 h (Fig. 2D). Collectively, these results indicate that AMPK activation generates a survival signal after cisplatin treatment in AGS cells. We next determined the levels of intracellular ATP in AGS cells that were exposed to cisplatin, compound C, or both for different durations (Fig. 2E). Twenty-four hours of cisplatin treatment (60 M) or compound C (20 M) resulted in a ϳ20% reduction in ATP levels. However, combination treatment with cisplatin and compound C on cellular ATP levels decreased dramatically. In accordance with the kinetics inherent to ATP depletion, the cleavage of poly(ADP-ribose) polymerase (PARP), which implies the initiation of apoptosis, was accelerated dramatically as the result of combined treatment with cisplatin and compound C (Fig. 2F). These data indicate, again, that cisplatininduced AMPK activation exerts a protective effect against both ATP depletion and apoptosis.
Inhibition of Cisplatin-induced AMPK Activity Increases p53 Protein Accumulation, Phosphorylation of p53-Ser 15 , and the Expression of the Target Genes-The p53 tumor suppressor protein is a central mediator of a variety of DNA damage responses, and it has been fairly well established that the p53 protein is stabilized and its level thereby increases in response to cisplatininduced DNA damage (9). p53, in turn, triggers cell cycle arrest and apoptosis via the up-regulation of its transcriptional target genes, such as p21 cip1/waf1 and Bax, respectively (22). Therefore, we attempted to determine whether AMPK is involved in the regulation of p53 in AGS cells (Fig. 3). As had been expected, p53 was detected 3-6 h after cisplatin exposure and increased in a time-dependent manner (Fig. 3A). Treatment of cells with a combination of compound C and cisplatin caused a significant induction of p53, and the time course of its induction was also accelerated (Fig. 3A). In accordance with this time course, the phosphorylation level of p53-Ser 15 , which is essential for the stabilization of the p53 protein, as well as p21 cip1/waf1 expression, exhibited similar kinetic patterns (Fig. 3A). Conversely, the pretreatment of cells with the AMPK activator, 5-aminoimidazole-4-carboxamide-1-␤-D-ribofuranoside (AICAR) (11), reduced the cisplatin-induced the protein level of p53 and the phosphorylation of p53-Ser 15 (Fig. 3B). The effect of the modulation of AMPK activity on p53 was further verified via molecular approaches (Fig. 3C). Infection with Ad-AMPK-DN significantly induced the cisplatin-mediated the stabilization of p53, the phosphorylation of p53-Ser 15 , and p21 cip1/waf1 expression in AGS cells. The opposite results were observed when endogenous AMPK activity was stimulated by Ad-AMPK-CA. . These cells were then exposed to cisplatin (60 M) for 1 h, and then Western blot assays were performed using anti-phosphospecific ACC-Ser 79 (P-ACC), anti-ACC (ACC), anti-phosphospecific AMPK␣-Thr 172 (P-AMPK), anti-c-myc (c-myc) antibody. B-D, after infection with Ad-AMPK-WT, Ad-AMPK-DN, or Ad-AMPK-CA, AGS cells were exposed to cisplatin (60 M) for 24 h, and then analyzed for apoptosis via fluorescence-activated cell scanning analysis, and the percentages of sub-G 1 (apoptotic fraction) cells are shown, and the data are expressed as the means Ϯ S.E. for three determinations in triplicate (*, p Ͻ 0.05; compared with the indicated group) (B). Under identical conditions, AGS cells were stained with Hoechst 33342, and the arrows indicate chromosomal DNA fragmentation (C). A ladder pattern of DNA fragmentation was also examined under the identical conditions (D). E, AGS cells were exposed to cisplatin (60 M) for the indicated time periods in the presence or absence of 20 M compound C (Com. C). Then, the intracellular ATP levels were assessed via the luciferin/luciferase method using an ATP determination kit (Molecular Probes). The data are expressed as the means Ϯ S.E. for three determinations in triplicate (*, p Ͻ 0.05; compared with control). F, under identical conditions, the total cell extracts were prepared and subjected to Western blot assays using anti-poly (ADP-ribose) polymerase-1 (PARP) and anti-␣-actinin (␣-actinin) antibody.
Considering the finding that the inhibition of cisplatin-induced AMPK activation resulted in an increase in the levels of the p53 protein, we then assessed the transcriptional activity of p53 and the level of p53 target gene expression. The AGS cells were infected with Ad-AMPK-WT or Ad-AMPK-DN, followed by transfection with pG-13-luc, which harbors the firefly luciferase gene under the control of 13 p53-responsive elements. These cells were then exposed for 12 h to cisplatin. Under these conditions, cisplatin induced p53-dependent luciferase activity by a factor of ϳ4-fold, but infection with Ad-AMPK-DN resulted in a further elevation of cisplatin-induced luciferase activity, resulting in a ϳ14-fold induction (Fig. 4A). Cisplatin unambiguously increased the mRNA level of p53 target genes such as p21 cip1/waf1 and Bax as shown by reverse transcription-PCR, and such inductions were increased significantly by Ad-AMPK-DN (Fig. 4B). Under these conditions, levels of p53 and ␤-actin mRNA remained constant. Consequently, these findings suggest that the inhibition of cisplatin-induced AMPK activity had no effect on the levels of p53 mRNA, but rather resulted in an increase in the level of p53-Ser 15 phosphorylation, which in turn leads to an increase in protein stabilization, transcriptional activity, and target gene expressions.
The Effect of AMPK Inhibition on Cisplatin-induced Apoptosis Is Achieved Principally in a p53-dependent Manner-Thus far, our data have shown that the inhibition of cisplatin-induced AMPK augments apoptosis (Figs. 1 and 2) and leads to an increase in the p53 signal ( Fig. 3 and 4). However, it remained unclear as to whether p53 is a primary mediator of the effects of AMPK inhibition, or whether some route other than p53 might ultimately prove the cause of apoptosis. To assess this possibility, we compared the effects of AMPK inhibition in p53-deficient HCT116 p53Ϫ/Ϫ and its counterpart HCT116 p53ϩ/ϩ (human colon carcinoma). Cisplatin activated AMPK in a dosedependent manner in these cells (Fig. 5A). Treatment of HCT116 p53ϩ/ϩ with a combination of compound C and cisplatin resulted in the cleavage of PARP (23), which implies the initiation of apoptosis, whereas PARP cleavage was not observed in HCT116 p53Ϫ/Ϫ cells under identical conditions (Fig. 5B). Therefore, these results indicate that the apoptosisenhancing effect of AMPK inhibition after cisplatin treatment may be dependent on p53. This idea was further verified when cell viability was assessed under identical conditions (Fig. 5C). Treatment with compound C and cisplatin resulted in a dramatic reduction in the viability of HCT116 p53ϩ/ϩ , but HCT116 p53Ϫ/Ϫ cells proved highly resistant against cisplatin as well as the combined treatment. However, HCT116 p53Ϫ/Ϫ became susceptible to cisplatin, as well as the compound C/cisplatin combination modality, after transfection with an expression vector encoding for wild-type p53. This result clearly indicates that the effects of the inhibition of cisplatininduced AMPK activation are mediated primarily by p53. Furthermore, we also observed that the modulation of AMPK activity exerted identical effects on the phosphorylation level of exogenously introduced p53 in HCT116 p53Ϫ/Ϫ cells (Fig. 5D). HCT116 p53Ϫ/Ϫ cells were cotransfected with p53 wild-type and  (Fig. 5D).
Induction of p53 Phosphorylation at Ser 15 by AMPK Inhibition Occurs Independently of ATM or ATR-Key molecules in the initiation of the DNA damage response include the ATM and ATR kinases, and, upon the infliction of DNA damage, ATM and ATR phosphorylate p53 at Ser 15 . Our data, up to this point in the study, had indicated that the response of p53 to cisplatin, including phosphorylation at Ser 15 , was further potentiated by the inhibition of AMPK. Thus, we attempted to elucidate the relevant signaling pathway by examining whether or not the signal generated by the inhibition of AMPK is associated with ATM and ATR. AGS cells were infected with Ad-AMPK-WT or Ad-AMPK-DN, and then exposed for 6 h to cisplatin in either the absence or presence of the ATM and ATR inhibitor, caffeine (Fig. 6A). Ad-AMPK-DN increased p53-Ser 15 phosphorylation level ϳ1.8-fold in the absence of caffeine. As had been expected, caffeine significantly blocked the effects of cisplatin on the p53-Ser 15 phosphorylation levels. Nevertheless, Ad-AMPK-DN in the presence of caffeine was still able to increase p53-Ser 15 phosphorylation levels ϳ2.2fold, which is approximately the same induction degree as seen in the absence of caffeine. This indicates that the effects of AMPK inhibition are still transmitted to p53, although ATM and ATR activities were blocked (Fig. 6A). Moreover, caffeine evidenced no significant effects on cisplatin-induced AMPK activity, thereby suggesting that AMPK signaling may not be associated directly with that of ATM and ATR. Serine 15 of p53 is phosphorylated by both ATM and ATR, but it is generally thought that genotoxic drugs, including cisplatin, preferentially activate ATR (24,25), whereas ATM responds principally to ionizing radiation-induced DNA damage. This notion was also bolstered by our observation that p53-Ser 15 phosphorylation in the ATM-deficient human fibroblast line, GM05849, was still induced strongly by cisplatin, and the inhibition of cisplatininduced AMPK activation also caused a significant elevation of p53-Ser 15 phosphorylation in GM05849 cells (Fig. 6B). Fig.  6A, caffeine did not completely block the cisplatin-induced phosphorylation of p53-Ser 15 , which suggests that some signaling pathway other than ATR and ATM may also be involved in phosphorylation of p53-Ser 15 . In fact, the involvement of mitogen-activated protein kinases (MAPK) in the regulation of the cisplatin sensitivity of cancer cells remains a subject of substantial interest. MAPKs have been classified into three major subfamilies: extracellular signal-regulated protein kinase (ERK), c-Jun N-terminal protein kinase, and p38 kinase. Among them, ERK has become the focus of our attention due to the following reasons. First, cisplatin activates ERK in many types of cancer cells, and ERK activation has been shown to be critical for cisplatin-induced apoptosis (26). Second, cisplatin-induced ERK activation contributes to an elevated level of p53 phosphorylation at serine 15 and to p53 accumulation (27). Third, we previously showed that AICAR treatment results in the inhibition

(kDa) 85 (kDa)
A FIGURE 5. The effects of the inhibition of cisplatin-induced AMPK activation are mediated primarily by p53. A, HCT116 p53ϩ/ϩ and HCT116 p53Ϫ/Ϫ cells were exposed to cisplatin for 1 h, and then the phosphorylation level at ACC-Ser 79 (P-ACC) and total ACC (ACC) was measured by Western blot analysis. B, HCT116 p53ϩ/ϩ and HCT116 p53Ϫ/Ϫ cells were exposed to cisplatin (60 M) with or without compound C (20 M). Western blot analysis was conducted using anti-poly (ADP-ribose) polymerase-1 (PARP), anti-p53 (p53), and anti-␣-actinin (␣-actinin) antibodies. C, HCT116 p53Ϫ/Ϫ cells were transfected with wild-type p53, and then the cell viability was compared with that of vector-transfected HCT116 p53Ϫ/Ϫ and HCT116 p53ϩ/ϩ after 24 h of cisplatin exposure (60 M) in the presence or absence of compound C (20 M). The data are expressed as the means Ϯ S.E. for three determinations in triplicate (*, p Ͻ 0.05; compared with control). D, HCT116 p53Ϫ/Ϫ cells were cotransfected with p53 wild-type and pcDNA containing AMPK wild-type (WT), dominant-negative form (DN), or constitutively active forms (CA) at a 1:2 ratio. These cells were incubated for 6 h with cisplatin, and then the phosphorylation level of p53 at Ser15 (p53-Ser 15 ) was assessed via Western blotting.
of insulin-like growth factor-1-induced ERK activity (28), although another study showed that AICAR can activate ERK under a certain condition (29). As a result, the interrelationship between AMPK and ERK remains elusive so far. Thus, we attempted to elucidate the AMPK signaling pathway by characterizing the cross-talk with ERK signaling. Cisplatin did, indeed, induce rapid ERK activation in AGS cells (Fig. 7A, left  panel). The inhibition of AMPK activation after cisplatin treatment via either compound C (Fig. 7A) or Ad-AMPK-DN (Fig.  7B) resulted in the hyper-activation of ERK, thereby indicating that AMPK activation in response to cisplatin may result in a suppression of ERK activity. Because the ERK inhibitor, PD98059, almost completely blocked the hyper-activation of ERK induced by the combined cisplatin and AMPK inhibition treatment, the enhanced level of p53-Ser 15 phosphorylation under this condition was simultaneously reduced (Fig. 7, A  (right panel) and B). Consequently, these results indicate the presence of a novel signal cascade involving AMPK, ERK, and p53. Our data appear to initially suggest that ERK activation, in addition to ATM and ATR, is required for the cisplatin-induced p53-Ser 15 phosphorylation. Second, AMPK activation in response to cisplatin seems to transmit a survival signal to the cells via the suppression of the apoptotic property of p53, which occurs due to a suppression of ERK activity. Caffeine treatment did not significantly affect cisplatin-induced AMPK (Fig. 6A) or ERK activation (Fig. 7C), thereby suggesting that this pathway operates independently of ATR or ATM. These signaling pathways are summarized in the diagram shown below in Fig. 13.

Inhibition of AMPK Activation Increases Cellular Response to Cisplatin in Various Cancer
Cells, but Not in HeLa Cells-Next, we determined the sensitivity of a variety of human cancer cells (AGS gastric cancer, HCT116 colon cancer, HepG2 liver cancer, MCF7 breast cancer, and HeLa cervical cancer) to cisplatin. We first examined the level of AMPK activation in these five different cancer cells (Fig. 8A). The results revealed that cisplatin activated AMPK in AGS, HCT116, MCF7, and HepG2 cells, but not in HeLa cells. To ascertain whether AMPK activity is essential for cell viability after treatment with cisplatin, these cells were treated for 24 h with the indicated concentrations of cisplatin in the presence or absence of the AMPK-specific inhibitor, compound C (Fig. 8B). AGS, HCT116, MCF7, and HepG2 cells were found to be relatively insensitive to cisplatininduced cytotoxicity, resulting in ϳ20% decrease in cell viability. In contrast, treatment of HeLa cells with cisplatin (120 M, 24 h) resulted in ϳ60% decrease in cell viability. In fact, among the various cancer cell lines, HeLa cells have been reported to be highly sensitive to anti-cancer drugs such as cisplatin (26). The combined treatment with AMPK inhibitor resulted in a dramatic upshift in cisplatin cytotoxicity in AGS, HCT116, HepG2, and MCF7 cells (Fig. 8B). In the case of the HeLa cells, which are relatively cisplatin-sensitive, AMPK inhibitor evidenced no effects on cisplatin-induced cell death (Fig. 8B). On the basis of our observations so far, we hypothesized that the lack of AMPK activation in HeLa cells contributed to the increased cellular sensitivity to cisplatin, and we further tested this hypothesis.
Overexpression of LKB1 or Constitutive Active Form of AMPK Renders HeLa Cells Relatively Insensitive to Cisplatin-For a more detailed test, HeLa cells were treated with cisplatin under the identical conditions as in Fig. 1A. The results revealed that AMPK was not activated by cisplatin, although the endogenous AMPK was able to be activated by exogenously added hydrogen peroxide (1 mM, 1 h) (Fig. 9A). Moreover, cisplatin did not induce ROS generation in HeLa cells (Fig. 9B), and anti-oxidants such as NAC or catalase showed no effect on AMPK activity (Fig. 9C).
To further demonstrate whether AMPK activation is indeed associated with the sensitivity of cellular response to cisplatin, we attempted to promote AMPK activation in HeLa cells via two different approaches: overexpression of a constitutively active form of AMPK (AMPK CA) or LKB1 which was identified as one of the upstream activating AMPK kinases (30 -32). LKB1 is a tumor suppressor kinase, gene mutations of which induce a dominantly inherited cancer, referred to as Peutz-Jeghers syndrome (33). Moreover, a recent study showed that LKB1 is essential for protection of cells against the apoptosis induced by nutritional deprivation (30). It was also reported that HeLa cells do not express LKB1 (34), which led us to suspect that overexpression of LKB1 would promote AMPK acti- vation in HeLa cells. Indeed, no LKB1 was expressed in the HeLa cells, as demonstrated by reverse transcription-PCR, whereas it was clearly expressed in the AGS cells (Fig. 10A). Transfection of HeLa cells with green fluorescent protein (GFP)-tagged LKB1 expression vector resulted in a significant elevation of AMPK activity, as evidenced by the phosphorylation level of ACC (Fig. 10B, upper panel), and cell viability after cisplatin treatment (60 M, 24 h) was also increased by a significant degree (Fig. 10C). Similar results were also observed when HeLa cells were transfected with c-myc-tagged AMPK CA (Fig. 10B, lower panel  and C). Consequently, our results thus far appear to suggest that AMPK performs a pivotal function for protection against the cytotoxic effects of cisplatin, thereby implying that AMPK is one of the cellular factors determining the sensitivity of cancer cells to the chemotherapeutic agent cisplatin.
We next attempted to examine whether blunt responsiveness of AMPK in HeLa cells is stimuli-specific or not. To this end, HeLa, AGS, and HCT116 cells were exposed to other apoptotic stimuli such as hypoxia (0.1% O 2 ), H 2 O 2 (0.3 mM), as well as cisplatin (60 M) (Fig. 11). Hypoxia and H 2 O 2 induced AMPK activation more significantly than cisplatin in three tested cell lines, whereas cisplatin failed to activate AMPK only in HeLa cells (Fig. 11A), as judged by the phosphorylation level of AMPK-Thr 172 and ACC-Ser 79 , indicating that AMPK in HeLa cells is still responsive to other stimuli. The total form of AMPK and ACC was not altered (data not shown). Therefore, some upstream kinase(s) other than  0 1.4 5.7 1.9 6.3 3.9 12.1 8.9 Cisplatin (60µM, 6h) LKB1 may stimulate AMPK in HeLa cells in response to hypoxia and H 2 O 2 . These three cells were also exposed to hypoxia, H 2 O 2 , and cisplatin for 24 h in the presence or absence of the AMPK-specific inhibitor, compound C, and the cell viability was compared (Fig. 11B). Compound C potentiated cell death induced by these stimuli in all tested cell lines except the case of cisplatin-induced HeLa cell death, where AMPK activation did not occur. These results again support that AMPK plays a critical role for protection against the cytotoxic effects of cisplatin.
The Anti-tumor Efficacy of AMPK Inhibitor in Combination with Cisplatin in HCT116 Human Cancer Xenograft Model-Next, to determine whether the in vitro chemosensitizing effects associated with AMPK inhibition can be reproduced in vivo, we evaluated the anti-tumor effects of cisplatin, compound C, and the combination treatment in nude mice harboring HCT116 p53ϩ/ϩ -tumor xenografts. A total of 2 ϫ 10 6 HCT116 p53ϩ/ϩ cells was implanted on a flank subcutaneously in each of five nude mice in each group. When the tumor masses grew to ϳ2-3 mm in diameter, drugs were intraperitoneally administered once every 4 days (compound C, 2 mg/kg; cisplatin 2 mg/kg; combination of two). Under our experimental conditions, cisplatin alone evidenced no detectable anti-tumor activity (Fig. 12). By way of contrast compound C alone resulted in a relatively effective delay of tumor growth as compared with cisplatin alone. However, when these two agents were administered in combination, dramatic anti-tumor effects were observed (Fig. 12). Collectively, our data show that AMPK inhibition can effect the sensitization of cancer cells to cisplatin, both in vitro and in vivo.

DISCUSSION
Multiple mechanisms have been implicated in the development of cisplatin resistance, including the reduced intracellular accumulation of the drug, the inactivation of cisplatin by sulfurcontaining molecules, an increase in DNA damage repair, increased levels of anti-apoptotic genes, and alterations in the signal transduction pathways associated with apoptosis (35). Therefore, a great deal of effort has been made to augment the therapeutic indices of cisplatin via the modulation of the aforementioned targets. Apart from these mechanisms, we investigated whether inhibition of AMPK, which performs a pivotal function in energy homeostasis via the monitoring of intracellular energy status, could represent a novel approach to the augmentation of tumor cell cisplatin sensitivity (Fig. 13).
Herein, we showed that cisplatin rapidly but transiently activated AMPK in various cancer cells (Figs.  1, 3, and 8), and the inhibition of AMPK activity resulted in a marked augmentation of cisplatin-induced apoptosis (Figs. 2, 5, and 8). Moreover, identical effects were detected in a xenograft model (Fig. 12). In accordance with the previous report (26), the HeLa cells appeared to be relatively cisplatin-sensitive, and moreover addition of AMPK inhib-  itor did not further increase cisplatin-induced cell death (Fig.  8). We determined that this was due, in part, to the blunt responsiveness of AMPK to cisplatin, because HeLa cells did not generate ROS in response to cisplatin (Fig. 9). However, because AMPK activation in HeLa cells was achieved by transfection of AMPK CA form or LKB1, cell viability after cisplatin treatment significantly increased (Fig. 10). Collectively, our data strongly indicate that the activation of AMPK exerts a protective effect against the cytotoxic effects of cisplatin and that the modulation of AMPK activity can affect cellular sensitivity to cisplatin. It seems rational to view AMPK as a survival factor for cancer cells, on the basis of our knowledge of the probable role of AMPK in the augmentation of energy production via the acti-vation of glucose uptake, glycolysis, and fatty acid oxidation in response to ATP-depleting stresses (11). Solid tumors that outgrow the existing vasculature are continuously exposed to a microenvironment in which the supply of both oxygen and nutrition are quite limited. In fact, we and others have previously shown that AMPK is critical for cancer cell adaptation in response to hypoxia or glucose deprivation (15, 36 -38). Moreover, the potential of AMPK as a survival factor has also been well documented in other tissues; AMPK has been clearly shown to protect the heart during ischemia (39). In accordance with the aforementioned reports and the data documented herein, it seems reasonable to conclude that the inhibition of AMPK in cancer cells may prove useful as an approach for the increased induction of apoptosis in tumor cells after cisplatin treatment. However, some have concluded, obversely, that AMPK activation may be employed as a component of an anticancer therapy (40,41). The logic of this approach is predicated on recent observations that AMPK also strongly suppresses cell proliferation. This effect is mediated, in part, by several tumor suppressor proteins associated with the AMPK signaling network, including LKB1 (33) and the tuberous sclerosis complex (TSC2) (42). Moreover, Jones et al. recently reported that the activation of AMPK induces p53-Ser 15 phosphorylation in response to glucose deprivation, resulting in replicative senescence (43). The ability of AMPK to promote senescence or to inhibit cell proliferation in response to energy starvation has been interpreted as a check point that couples glucose availability to the progression of the cell cycle; it was implied that the activation of AMPK might promote the conservation of the remaining energy to support the survival and physiological functions of the cell during cell cycle arrest. In fact, all tumor

FIGURE 12.
In vivo anti-tumor efficacy of AMPK inhibitor alone and in combination with cisplatin. BALB/c nude mice were inoculated subcutaneously into one flank with 2 ϫ 10 6 HCT116 p53ϩ/ϩ in Matrigel. Five mice were assigned to each of the groups. When the tumor masses reached ϳ2-3 mm in diameter, drugs were intraperitoneally administered once every 5 days (compound C, 2 mg/kg; cisplatin 2 mg/kg; combination of two). Tumor growth was measured every 4 days, as described under "Experimental Procedures" (*, p Ͻ 0.01; compared with control). suppressor proteins involved in the AMPK signal network, including LKB1, TSC2, and p53, not only contribute to cell cycle arrest but also protect cells against energy deprivationmediated apoptosis (30,42,43). As a result, active AMPK may facilitate the provision of energy for tumorigenesis, but may also concomitantly diminish the selective advantages of the cancer cells. Therefore, a key question persists as to whether AMPK activation effectively induces apoptosis in malignant cancer cells. Moreover, advanced malignant cancers often retain defective tumor suppressor proteins, including p53, LKB1, and TSC2, and therefore the tumor suppression effect associated with AMPK activation may not be anticipated in the case of certain malignant cancers. One of the most fundamental metabolic alterations occurring during the malignant transformation is the up-regulation of glycolysis, a phenomenon referred to as the Warburg effect, and malignant tumors exploit this pathway for the generation of ATP (44). AMPK also performs a key function in hypoxia-induced glycolysis (45). Therefore, rather than the simple induction of cancer cell arrest, which cannot completely stop the re-emergence of cancer, inhibition of AMPK, which plays a central role in energy homeostasis, is likely to be a more effective approach to cancer chemotherapy, because this approach does, in fact, induce apoptosis in cancer cells. Within a similar conceptual framework, the inhibition of glycolysis can also be considered as a component of an alternative approach to anticancer treatment (46). In this study, we have also shown that AMPK inhibition sensitized cells to cisplatin via hyper-induction of p53. The function of p53 as a signal mediator for AMPK inhibition was corroborated by the observation that AMPK inhibition did not significantly affect the cisplatin resistance of HCT116 cells lacking p53. The introduction of the wild type of p53 into HCT116 p53Ϫ/Ϫ rendered these cells relatively sensitive to the combined AMPK inhibitor and cisplatin treatment (Fig. 5C). Our results are consistent with the general view that p53 levels in excess of a certain threshold almost invariably result in apoptosis or in enhanced drug sensitivity. However, the signaling network that connects AMPK and p53 appears to be a highly complex one, because the particular patterns of biochemical regulation and the ultimate physiological outcome appear to depend substantially upon the types of cells and the stresses induced. Jones et al. demonstrated that AMPK induces p53-Ser 15 phosphorylation in mouse embryonic fibroblasts, and such metabolic activations of p53 were associated with cell survival, whereas the activation of p53 induced by ␥-radiation resulted in a reduction in cell viability (43). However, our results indicate that the inhibition of cisplatin-induced AMPK activation resulted in an upshift in the phosphorylation of p53-Ser 15 in cancer cells, which is associated with reduced cell viability. Consistent with our results, the findings of a recent study indicated that treatment with 5-aminoimidazole-4-carboxamide-1-␤-D-ribofuranoside, a well known AMPK activator, caused a reduction in p53 levels in renal cell carcinoma cells (47). Consequently, to elucidate clearly the AMPK-p53 network, more elaborate and comprehensive studies appear to be required, which take into account the genetic profiles of the cells, the types of stress, its duration, and the degree of persistence of the AMPK activation.