Role of AMP-activated protein kinase in cross-talk between apoptosis and autophagy in human colon cancer

Unresectable colorectal liver metastases remain a major unresolved issue and more effective novel regimens are urgently needed. While screening synergistic drug combinations for colon cancer therapy, we identified a novel multidrug treatment for colon cancer: chemotherapeutic agent melphalan in combination with proteasome inhibitor bortezomib and mTOR (mammalian target of rapamycin) inhibitor rapamycin. We investigated the mechanisms of synergistic antitumor efficacy during the multidrug treatment. All experiments were performed with highly metastatic human colon cancer CX-1 and HCT116 cells, and selected critical experiments were repeated with human colon cancer stem Tu-22 cells and mouse embryo fibroblast (MEF) cells. We used immunochemical techniques to investigate a cross-talk between apoptosis and autophagy during the multidrug treatment. We observed that melphalan triggered apoptosis, bortezomib induced apoptosis and autophagy, rapamycin caused autophagy and the combinatorial treatment-induced synergistic apoptosis, which was mediated through an increase in caspase activation. We also observed that mitochondrial dysfunction induced by the combination was linked with altered cellular metabolism, which induced adenosine monophosphate-activated protein kinase (AMPK) activation, resulting in Beclin-1 phosphorylated at Ser 93/96. Interestingly, Beclin-1 phosphorylated at Ser 93/96 is sufficient to induce Beclin-1 cleavage by caspase-8, which switches off autophagy to achieve the synergistic induction of apoptosis. Similar results were observed with the essential autophagy gene, autophagy-related protein 7, -deficient MEF cells. The multidrug treatment-induced Beclin-1 cleavage was abolished in Beclin-1 double-mutant (D133A/D146A) knock-in HCT116 cells, restoring the autophagy-promoting function of Beclin-1 and suppressing the apoptosis induced by the combination therapy. These observations identify a novel mechanism for AMPK-induced apoptosis through interplay between autophagy and apoptosis.

Colorectal cancer ranks third in major causes of cancerrelated mortality worldwide, appearing in~150 000 new cases in the United States annually, and~20-50% of colorectal cancer patients display hepatic metastases. 1,2 Current standard therapies for treating metastatic colon cancer include chemotherapy and biological therapy followed by tumor resection, up front tumor resection followed by systemic therapy, radiofrequency ablation, thermal ablation, selective internal radiation therapy and hyperthermic isolated hepatic perfusion (IHP) therapy. [3][4][5][6][7] Although these therapies are somewhat effective, more effective novel regimens are still needed to improve the survival of patients with liver metastases from colorectal cancer.
As unresectable liver metastases from colorectal cancer are difficult to treat by single modality, we have spent several years developing a multimodality approach for hyperthermic IHP therapy. We previously investigated the mechanism of the synergy between hyperthermia, biological agents (TNFrelated apoptosis-inducing ligand/mapatumumab) and chemotherapeutic agent (oxaliplatin). [8][9][10] However, the clinical grade of those biological agents is no longer available after newly merged companies decided not to produce them. We then investigated potential replacement drugs, which are already Food and Drug Administration (FDA) approved. We screened several FDA-approved drugs including melphalan, chlorquine, bortezomib, carbamazepine, celecoxib, cetuximab and rapamycin using cytoxicity assay. We found that MBR (melphalan+bortezomib+rapamycin) treatment has the best cytotoxic effect on colon cancer cells and also on colon cancer stem cells. Currently, 2807 clinical trials are listed for colon cancer; of these, 225 studies and 195 studies are related to FOLFOX (folinic acid+fluorouracil+oxaliplatin) therapy and FOLFIRI (folinic acid+fluorouracil+irinotecan) therapy, respectively. There are only seven studies for IHP and, specifically, there is no clinical trial with MBR for hyperthermic IHP therapy.
Apoptosis is a major cytotoxic mechanism of chemotherapy; stress-induced apoptosis often proceeds through the intrinsic pathway where permeabilization of the mitochondrial outer membrane releases cytochrome c and activates the caspase cascade. 11,12 Melphalan hydrochloride (trade name Alkeran), which is commonly used in IHP, leads to double-stranded DNA breaks and subsequent cell death through a caspasemediated, apoptotic pathway. 13,14 Bortezomib, the first clinically available proteasome inhibitor, possesses antitumor activity in a variety of human cancers and is often used in the treatment of hematological malignancies. It can induce both proapoptotic effects, including the induction of Bik, Bim and Noxa proteins, and antiapoptotic effects, including the accumulation of Mcl-1 and HSP70, as well as autophagic formation. [15][16][17] mTOR (mammalian target of rapamycin) is known to be well conserved and ubiquitously expressed in endothelial cells and is involved in cell energy metabolism, cell growth, apoptosis and autophagy. Several human cancers display mTOR hyperactivation, thus making mTOR an attractive target in cancer therapy. 18 Sirolimus, known as rapamycin, an mTOR inhibitor, has relatively low cytotoxic activity. Better therapeutic outcomes should be obtained by using rapamycin in combination with other anticancer agents. 19,20 In this study, we observed that melphalan triggered apoptosis, bortezomib induced both apoptosis and autophagy, rapamycin caused autophagy and the combinatorial treatment promoted mitochondrial dysfunction and synergistic apoptosis.
Hallmarks of cancer cells include uncontrolled growth, evasion of apoptosis, immortality, ability to invade other tissues and altered cellular metabolism. 21 In our study, we investigated the effect of multidrug treatment on cellular metabolism. This study revealed that the combinatorial treatment altered cellular metabolism and induced energy sensor AMP-activated protein kinase (AMPK) activation at two stages in the process, resulting in Beclin-1 phosphorylation and autophagy starting with the early stage and Beclin-1 cleavage by caspase-8 and apoptosis concurrent with the late stage. Our observations provided evidence that AMPK has an important role in cross-talk between autophagy and apoptosis.

MBR synergistically induced cytotoxicity and apoptosis.
To investigate the effect on cell viability of the application of MBR, human colorectal carcinoma HCT116 and CX-1 cells were treated with a combination of 10 μg/ml melphalan, along with 50 nM bortezomib and/or 2.5 μg/ml rapamycin for 24 h. Cell viability was determined by 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) assay. As shown in Figures 1a and b, synergistic effect was observed in MBR compared with any single treatment or bitreatment in both cell lines (Po0.01). Our observation was confirmed by combination index (CI) analysis; CI values were o1 (Table 1). In Figure 1c, apoptotic cells detected by Annexin V/PI assay were observed in the upper right quadrant of each plot. Our data clearly show that treatment with MBR enhanced synergistic induction of apoptotic death. These synergistic effects were due to an increased activation of caspases, and thus, the hallmark of apoptosis, poly (ADP-ribose) polymerase (PARP) cleavage (Figures 1d and e). Similar results were observed in human colon cancer stem Tu-22 cells (Figure 1f).
MBR-induced mitochondrial dysfunction and AMPK activation. Stress-induced apoptosis often proceeds through the intrinsic pathway mediated by THE mitochondria. In Figures 2a and b, the upper left quadrant of each plot displays cells with intact mitochondrial membrane potential, whereas the lower right quadrant displays cells with impaired mitochondrial membrane potential. A significant shift occurred to the lower right part of the quadrants in the treatment with MBR. Figure 2c indicates that more cytochrome c was released during multidrug treatment. We then assessed the ATP production as the mitochondria are known to have an important role in providing overall cellular energy supply. ATP production was markedly decreased during the treatment with MBR in HCT116 cells ( Figure 2d) and CX-1 cells (data not shown). Figure 2e shows that MBR induced a large amount of phosphorylation (activation) of AMPK. We then investigated whether Bcl-2-associated X protein (Bax) was involved in the combinatorial treatment-induced apoptosis. Data from Figure 2f clearly demonstrates that MBRinduced apoptosis and caspase activation were effectively suppressed in Bax-deficient cells, indicating that the synergy of MBR-associated apoptosis is partially mediated through Bax. Interestingly, AMPK activation was also decreased in HCT116 Bax − / − cells, implying that Bax contributes to AMPK activation. The hallmark of autophagy, microtubuleassociated protein 1A/1B-light chain 3 (LC3)-II, increased markedly in bortezomib-treated cells and mildly in rapamycintreated cells but no significant increase with melphalan treatment. Notably, there was a significant decrease of LC3-II in the treatment of MBR compared with that of bortezomib alone. Interestingly, LC3-II was increased in the treatment of MBR in HCT116 Bax − / − cells compared with that of HCT116 Bax +/+ cells, indicating that autophagy was increased in the apoptosis-suppressed cells (Figure 2f). These results indicate a cross-talk between autophagy and apoptosis during treatment with MBR. Figure 2g shows that cytotoxicity was significantly decreased in HCT116 Bax − / − compared with HCT116 Bax +/+ (Po0.01).
Interplay between autophagy and apoptosis during treatment with MBR. To further investigate a relationship between apoptosis and autophagy during treatment with MBR, we examined autophagy using as an autophagyspecific marker both processing of LC3-I into LC3-II using immunoblot assay and also green fluorescent protein (GFP)-LC3 puncta formation using confocal microscopy. We observed that LC3-II (Figures 3a and b) and LC3 puncta formation ( Figure 3c) increased significantly in bortezomib alone or in combination with melphalan or rapamycin. However, MBR treatment suppressed LC3-II or LC3 puncta formation compared with bortezomib alone. It is known that the classical autophagy pathway is dependent on Beclin-1, autophagy-related protein 7 (ATG7), and so on. 22 Figure 3d shows that MBR treatment-induced apoptosis was enhanced in mouse embryo fibroblast (MEF) ATG7 − / − cells. Similar results were observed by MTS assay (Figure 3e). These results suggest that inhibition of autophagy enhances multidrug-induced apoptosis.
The role of AMPK in MBR-induced apoptosis. To further investigate a cross-talk between autophagy and apoptosis, we hypothesized that enhancement of MBR-induced apoptosis is mediated through the activation of AMPK. To test this hypothesis, we added an AMPK inhibitor, compound C, in the absence or presence of MBR. Figures 4a and b show that MBR-induced apoptosis was partially suppressed  Figure 4c shows that a combination of AMPK-α1 and AMPK-α2 small interfering RNA (siRNA) downregulated AMPK-α expression level and partially suppressed MBR-induced apoptosis, which was consistent with the effects of compound C. Our results suggest that MBRinduced AMPK activation has an important role in the crosstalk between autophagy and apoptosis. To further investigate the role of AMPK activators metformin and AICAR (aminoimidazole carboxamide ribonucleotide) in MBR-induced cytotoxicity, cells were treated with metformin/AICAR in the absence or presence of MBR. As shown in Figure 4d, metformin alone did not affect CX-1 cell viability. However, MBR-induced cytotoxicity was significantly enhanced in the presence of 20 mM metformin. Similar results were observed in HCT116 cells (Figure 4e). Figures 4f and g show that AICAR alone did not affect cell viability in CX-1 and HCT116 cells but MBR-induced cytotoxicity was enhanced in the presence of AICAR.
AMPK activation-induced Beclin-1 phosphorylation at Ser 93/96 and Beclin-1 cleavage. To examine the role of AMPK in the cross-talk between autophagy and apoptosis, we further investigated AMPK downstream molecules. Recently, AMPK was reported to phosphorylate Beclin-1 at Ser 91/94 for mouse or Ser 93/96 for human upon glucose starvation. 23 C-terminal fragment of Beclin-1 localizes at the mitochondria, inducing cytochrome c release and thus enhancing the apoptosis. 24 We observed Beclin-1 phosphorylation at Ser 93/96 in HCT116 and CX-1 cells (Figures 5a  and b), as well as Beclin-1 cleavage in HCT116 cells (Figure 5c), during the treatment with MBR. Similar results were observed in Tu-22 cells (Figure 5d). We also observed that AMPK inhibitor, compound C, inhibited Beclin-1 phosphorylation at Ser 93/96 and Beclin-1 cleavage in both cell lines (Figures 5e and f). Figure 5g shows that 1 mM AICAR alone, which activated AMPK, did not induce apoptosis in CX-1 and HCT116 cells. However, AICAR enhanced MBRinduced apoptosis. We also observed that MBR-induced PARP cleavage, Beclin-1 phosphorylation at Ser 93/96 and Beclin-1 cleavage were enhanced in the presence of AICAR in both cell lines. AICAR alone increased the level of LC3-II (autophagy). However, a combinatorial treatment of MBR and AICAR reduced MBR-induced LC3-II. These results suggest a strong correlation between AMPK activation, Beclin-1 cleavage and switching the mode of cell death from autophagy to apoptosis.
The kinetics of AMPK-α activation and Beclin-1 phosphorylation and cleavage in the treatment of MBR. Next, we examined the kinetics of MBR-induced apoptosis. As shown in Figures 6a and b, MBR-induced apoptosis was increased as time progressed. In CX-1 cells, AMPK was phosphorylated (activated) very early at 1 h and then decreased. Notably, activation of AMPK was increased again at around 16 h in a sustained manner. Beclin-1 phosphorylation at Ser 93/96 was markedly increased after 3 h. Beclin-1 cleavage increased after 12 h when apoptosis occurred and then apoptosis was significantly increased. We also monitored autophagy in CX-1 cells after MBR treatment, using as an autophagy-specific marker processing of LC3-I into LC3-II. Interestingly, during the first 16 h, MBR treatment co-occurred with increased autophagy indicated by increased LC3-II levels. But then, reduced levels of LC3-II were observed at 24 h after treatment, implying that once caspases were fully activated, autophagy levels were diminished. Similar results were obtained in HCT116 cells, which were more sensitive towards the treatment of MBR compared with CX-1 cells (Figure 6b). To further test the differential role of AMPK activation at different stage, we used compound C to inhibit transiently AMPK at early stage (from 0 to 4 h and washed at 4 h with phosphatebuffered saline) and late stage (from 10 to 24 h) separately. As shown in Figures 6c and d, we observed that inhibiting AMPK at early time point enhanced MBR-induced apoptosis, whereas inhibiting AMPK at late time point abolished MBR-induced apoptosis in CX-1 and HCT116 cells.
We further examined the kinetics of autophagy during treatment with MBR; GFP-LC3 puncta formation was investigated. Figures 6e and f show that LC3 puncta formation was detected at 3 h, and increased until 16 h and then decreased 2-fold at 24 h compared with that of 16 h of MBR treatment. Moreover, Figure 6g shows that PARP cleavage, AMPK activation and Beclin-1 phosphorylation at Ser 93/96, as well as Beclin-1 cleavage, consistently increased even at 36 h. We also observed that LC3-II was markedly decreased at 28 and 36 h. These data suggest that early AMPK activation-induced Beclin-1 phosphorylation and resulted in autophagy, whereas sustained late AMPK activation co-occurring with the activation of caspases induced Beclin-1 cleavage, enhancing apoptotic cell death.
Cleavage of Beclin-1 occurred with caspase activation. As caspase activation occurs during apoptotic signaling, we investigated if Beclin-1 and caspases bind with each other and  Figure 7c, Beclin-1 and its mutant types all bound with caspase-8, whereas binding with caspase-3 was hardly detected. Of note, the binding affinity of caspase-8 with Beclin-1 S91/94A was weaker than that of Beclin-1 WT and Beclin-1 S91/94D, which needs further study. We also observed that the binding affinity of AMPK with Beclin-1 was increased after treatment of MBR in both Beclin-1 WT and its mutant types. These observations were confirmed by immunoprecipitating assay with endogenous Beclin-1 proteins (Figure 7d). Data from Figure 7d reveal that endogenous Beclin-1 indeed associated with AMPK and caspase-8 but not caspase-3. Notably, the binding affinity of Beclin-1 and AMPK was increased after MBR treatment.
Taken together, we summarized our observations in Figure 7g. MBR treatment causes mitochondrial dysfunction and AMPK activation. AMPK-mediated Beclin-1 phosphorylation is sufficient to induce Beclin-1 cleavage and thus contributes to the synergistic induction of apoptosis by decreasing autophagy; specifically, blocking Beclin-1 cleavage promotes autophagy and suppresses the apoptosis induced by MBR.

Discussion
Multidrug treatment can potentially overcome the resistance developed by single agents as well as reduce their side effects. 26,27 Here we have presented a novel combination treatment of melphalan in combination with the proteasome inhibitor bortezomib and the mTOR inhibitor rapamycin for colon cancer cells, as well as colon cancer stem cells, and demonstrated that this multidrug treatment-induced synergistic apoptosis mediated by AMPK activation through facilitation of Beclin-1 cleavage.
In recent research, the relationship between cancer cell growth and cell metabolism has been emphasized, highlighting the role of AMPK as one of the main players. 28 AMPK is a eukaryotic heterotrimeric (α, β and γ) protein kinase activated in response to environmental stresses such as glucose deprivation and hypoxia, which produce changes in cellular ATP levels, resulting in phosphorylation of AMPK at Thr 172. Chemotherapy with certain drugs can activate AMPK to increase cell death and apoptosis. 29,30 Rapamycin and UPS (ubiquitin-proteasome system) inhibitors may upregulate AMPK activity. 31,32 It has been found that AMPK is involved in both cell survival and cell death pathways,  [33][34][35][36][37][38] In our current study, we observed significant AMPK activation in MBR treatment. AMPK inhibitor compound C and AMPK-α siRNA suppressed MBR-induced apoptosis, indicating the proapoptotic effect of AMPK in the combinatorial treatment of MBR. This observation was also supported by experiments using AMPK activator metformin and AICAR. Interestingly, we observed that AMPK was activated at two stages. Early AMPK activation-induced autophagy, whereas late AMPK activation resulted in significant apoptosis. Autophagy, a catabolic degradation process, has recently emerged as an important stress response and cell death regulatory mechanism. The process of conventional macroautophagy is dependent on Atg7 (ubiquitin-activating enzyme (E1)-like) and Beclin-1. 39 We observed that MBR treatmentinduced apoptosis was enhanced in MEF ATG7 − / − cells, indicating that blockage of autophagy promoted apoptosis in MBR treatment. We also observed that autophagy decreased in the combination of MBR compared with bortezomib alone. Thus, these results indicate the role of AMPK in cross-talk between apoptosis and autophagy in the multidrug treatment.
Beclin-1, a mammalian homolog of yeast Atg6, functions in autophagy by initiating autophagosome formation in combination with Vps34 (a class III PI-3 kinase that generates PtdIns3P) and with Vps15 and Atg14; 40 Beclin-1 then regulates autophagosome maturation by binding to UVRAG Figure 4 The role of AMPK in the MBR-induced apoptosis. (a and b) CX-1 (a) and HCT116 (b) cells were pretreated with 10 μM AMPK inhibitor compound C followed by treatment with MBR for 24 h. After treatment, PARP, caspase-8, caspase-9, caspase-3, p-AMPK and AMPK were detected by immunoblotting. (c) HCT116 cells were transfected with nonsense sequence (control) or AMPK-α siRNA targeting AMPK-α mRNA. After 48 h, cells were treated with MBR. The levels of AMPK-α and PARP were detected by immunoblotting. Actin was used as a loading control. (d and e) CX-1 (d) and HCT116 (e) cells were pretreated for 30 min with metformin (1-20 mM) followed by treatment with MBR for 24 h. Cell viability was analyzed by MTS assay. Error bars represent S.D. from triplicate experiments. **Po0.01 represents a statistically significant difference. (f and g) CX-1 (f) and HCT116 (g) cells were pretreated for 30 min with AICAR (0.5-5 mM) followed by treatment with MBR for 24 h. Cell viability was analyzed by MTS assay. Error bars represent S.D. from triplicate experiments. *Po0.05 and **Po0.01 represents a statistically significant difference   (12-36 h). After treatment, cell lysates were immunoblotted with anti-PARP, anti-phospho-AMPK-α, anti-AMPK-α, anti-phospho-Beclin-1, anti-Beclin-1 or anti-LC3 antibody. Actin was used to confirm the equal amount of proteins loaded in each lane and Rubicon. 41 Cross-talk between apoptosis and autophagy was associated with caspase-mediated cleavage of Beclin-1, which both destroys its proautophagic activity 22,25,42,43 and can then amplify mitrochondrion-mediated apoptosis through the cleaved C-terminal fragment. 24 However, how Beclin-1 cleavage is regulated is largely unknown.   46 Recently, AMPK was reported to phosphorylate directly Beclin-1 at Ser 93/96 for activating proautophagy Vps34 complex and subsequently inducing autophagy. 23 The question remains as to how AMPKmediated phosphorylation of Beclin-1 at Ser 93/96 regulates caspase-8-associated cleavage of Beclin-1. Here, we reported that Beclin-1 phosphorylation at Ser 93/96 is a prerequisite for Beclin-1 cleavage and thus contributes to the synergistic induction of apoptosis. AMPK inhibitor compound C inhibited both Beclin-1 phosphorylation at Ser 93/96 and Beclin-1 cleavage. In addition, Myc-tagged murine Beclin-1 S91/94A mutant was resistant to cleavage and suppressed MBR-induced apoptosis. Moreover, the binding affinity of Beclin-1 S91/94A with caspase-8 was decreased, providing one possible mechanism of AMPK-mediated Beclin-1 cleavage through its phosphorylation and formation of an AMPK-Beclin-1-caspase-8 complex. Beclin-1 DM (D133A/D146A) knock-in HCT116 cells partially abolished the apoptosis, confirming the role of Beclin-1 cleavage in the multidrug treatment.
Taken together, we present here that a novel multidrug treatment of chemotherapeutic agent in combination with proteasome inhibitor and mTOR inhibitor induced robust apoptosis in colon cancer cells as well as colon cancer stem cells, an apoptotic process that is linked with altered cellular metabolism and AMPK activation. We believe that these results demonstrate for the first time that the induction of apoptosis by AMPK is associated with Beclin-1 cleavage through Beclin-1 phosphorylation at Ser 93/96. Melphalan, bortezomib and rapamycin are all commonly used FDAapproved drugs and could be considered for colorectal hepatic metastases treatment in clinics.
MTS assays. MTS studies were carried out using the Promega CellTiter 96 AQueous One Solution Cell Proliferation Assay (Promega, Madison, WI, USA). CX-1 cells (1 × 10 5 ) were grown in RPMI-1640 medium containing 10% fetal bovine serum in tissue culture-coated 96-well plates and treated with drugs for 24 h. Cells were then treated with 20 μl MTS/phenazine methosulfate solution for 1-4 h at 37°C. Absorbance at 490 nm was determined using an enzyme-linked immunosorbent assay plate reader.
CI analysis. CIs were calculated using CompuSyn software program (ComboSyn Inc., Paramus, NJ, USA). Base on CI values, extent of synergism/ antagonism is determined. In general, CI value below 1 suggests synergy, whereas CI value above 1 indicates antagonism between the drugs. CI values in the range of 0.9-1.10 would mainly indicate additive effects, those between 0.9 and 0.85 would suggest slight synergy, those in the range of 0.7-0.3 are indicative of moderate synergy and those o0.3 would suggest strong synergy.
Annexin V binding. Cells were treated with drugs, harvested by trypsinization, washed with serum-free medium and suspended in binding buffer (Annexin V-fluorescein isothiocyanate (FITC) Staining Kit; BD PharMingen). This cell suspension was stained with mouse anti-human Annexin V antibody and propidium iodide (PI) and immediately analyzed by flow cytometry.
Immunoprecipitation. Briefly, cells were lysed in CHAPS lysis buffer with protease inhibitor cocktail (Calbiochem). The lysate (0.5-1 mg) was incubated with 1.5 μg of anti-Myc/Beclin-1 antibody or rabbit/mouse IgG (Santa Cruz, Dallas, TX, USA) at 4°C overnight, followed by the addition of protein G PLUS-agarose beads (Santa Cruz) and rotation at room temperature for 2 h followed by immunoblot analysis.
Immunoblot analysis. Cells were lysed with Laemmli lysis buffer and boiled for 10 min. The protein content was measured with BCA Protein Assay Reagent (Pierce, Rockford, IL, USA), separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and electrophoretically transferred to nitrocellulose membrane. The nitrocellulose membrane was blocked with 5% nonfat dry milk in PBS-Tween-20 (0.1%, v/v) for 1 h and incubated with primary antibody at room temperature for 2 h. Horseradish peroxidase-conjugated anti-rabbit or anti-mouse IgG was used as the secondary antibody. Immunoreactive protein was visualized by the chemiluminescence protocol.
JC-1 mitochondrial membrane potential assay. After drug treatment, cells were stained with JC-1 Mitochondrial Membrane Potential Detection Kit (Invitrogen, Carlsbad, CA, USA) for 10 min and analyzed by flow cytometry. Fluorescence intensity was measured with the FACScan flow cytometer (Beckman Coulter, Hialeah, FL, USA). Results were analyzed with CellQuest software (Becton Dickinson Immunocytometry Systems, San Jose, CA, USA).
ATP assay. The ATP content in whole-cell extracts was determined with a luminescent ATP Detection Kit (ATPlite, Perkin-Elmer, Akron, OH, USA) according to the manufacturer's instructions. The luminescence intensity was measured by using a microplate reader (Synergy 2; BioTek Instruments, Winooski, VT, USA). In parallel, the cell numbers in whole-cell samples were counted by Trypan blue exclusion assay. The results were expressed as relative ATP level compared with controls after normalizing for cell numbers.
Measurement of cytochrome c release. To determine the release of cytochrome c from the mitochondria, cells growing in 100 mm dishes were used. After drug treatment, mitochondrial and cytosolic fractions were prepared using Mitochondrial Fractionation Kit (Active Motif, Carlsbad, CA, USA) following company instructions with reagents included in the kit. Cytosolic fractions were subjected to SDS-PAGE gel electrophoresis and analyzed by immunoblotting using anticytochrome c antibody.
Knockdown of AMPK-α with siRNA oligomers. To generate AMPK-α knockdown cells, cells were transfected with 10 nM of siRNA AMPK-α1 and