RGD-modified oncolytic adenovirus-harboring shPKM2 exhibits a potent cytotoxic effect in pancreatic cancer via autophagy inhibition and apoptosis promotion

The M2 isoform of pyruvate kinase (PKM2) is a key driver of glycolysis in cancer cells and has critical ‘non-metabolic’ functions in some cancers; however, the role of PKM2 in pancreatic cancer remains unclear. The aim of the current study was to elucidate the role of PKM2 in pancreatic cancer progression and the potential of PKM2 as a therapeutic target. In this study, we observed that PKM2 is highly expressed in patients with pancreatic cancer and is correlated to survival. Elevated PKM2 expression promoted cell proliferation, migration and tumor formation. The inhibition of cell growth by silencing PKM2 is caused by impairment of the autophagy process. To test the potential effects of downregulating PKM2 as a clinical therapy, we constructed an RGD-modified oncolytic adenovirus containing shPKM2 (OAd.R.shPKM2) to knock down PKM2 in pancreatic cancer cells. Cells transduced with OAd.R.shPKM2 exhibited decreased cell viability, and, in a PANC-1 xenograft model, intratumoral injection of OAd.R.shPKM2 resulted in reduced tumor growth. Furthermore, OAd.R.shPKM2 induced apoptosis and impaired autophagy in PANC-1 cells. Our results suggested that targeting PKM2 with an oncolytic adenovirus produced a strong antitumor effect, and that this strategy could broaden the therapeutic options for treating pancreatic cancer.

Pancreatic cancer is projected to become the second mostcommon cause of cancer-related death by 2030. 1 It is nearly undetectable during the early stages, and advanced-stage disease is unresectable and lacks an effective treatment. Despite half a century of research and therapeutic development, the 5-year survival rate is less than 7%, and the median survival rate remains at 6 months. 2 Therefore, it is critical to identify novel therapeutic targets and develop potential therapeutic strategies for pancreatic cancer.
Altered cellular metabolism is a hallmark of cancers. 3 Unlike normal cells, cancer cells can shift their glucose metabolism towards glycolysis, even in an oxygenated environment. This phenomenon is characterized by increased glucose consumption and an elevated rate of lactate production and is known as aerobic glycolysis, or the 'Warburg effect'. 4 Pyruvate kinase is a key enzyme in glycolysis that regulates the final rate-limiting step of catalysing the transfer of a phosphate from phosphoenolpyruvate to adenosine diphosphate to produce pyruvate and energy (ATP). The pyruvate kinase gene comprises four isoenzymes encoded by two distinct genes in mammals, PKM and PKLR. The two splice variants of PKM pre-mRNA produce pyruvate kinase M1 (PKM1) and M2 (PKM2), which include exons 9 or 10, respectively. 5 An increasing number of studies have shown that PKM2 but not PKM1 is crucial for tumorigenic phenotype maintenance, cell cycle progression and tumor growth. 6,7 Recently, PKM2 was identified as a protein kinase and transcription factor coactivator in regulating brain tumorigenesis and colon cancer cell migration, respectively, which are divergent from its canonical role as a pyruvate kinase. 8,9 Modulating PKM2 in tumor angiogenesis was recently reported to be regulated by miR-148a and miR-152 expression. 10 PKM2 prevents apoptosis in hepatocellular carcinoma (HCC), and knockdown of PKM2 inhibited cell proliferation and induced apoptosis in HCC. 11 The outcome of patients with either HCC or pancreatic cancer is inversely correlated with PKM2 expression. [11][12][13] Because of its multiple roles in tumorigenesis, PKM2 should be investigated as a target for pancreatic cancer therapy.
Replication-selective oncolytic adenovirus carrying either therapeutic genes or shRNA has been shown to exert promising antitumor effects on different types of cancers. 14,15 For improved implementation of this vector as a cancer therapy, some modifications have been made. Replacing the original E1A promoter with a tumor-specific promoter can transcriptionally control viral replication to some extent. 16 The vital gene for late viral RNA export is E1B 55K, and viruses with an E1B 55K deletion are incapable of replication in normal cells; however, tumor cells can efficiently export late viral RNA in the absence of E1B 55K. 17 As a binding partner of adenovirus type 5, the coxsackie and adenovirus receptor (CAR), when expressed on tumor cells, restricts the infection efficiency of adenovirus type 5. We previously inserted an Arg-Gly-Asp (RGD) motif into the HI loop of the adenovirus knob, which significantly elevated the infection efficiency of the adenovirus. 18 In the present study, we observed that PKM2 is overexpressed in pancreatic cancer samples and is correlated with patient survival. We showed that PKM2 knockdown inhibited cell proliferation, migration and tumor formation, and that PKM2 supressed autophagy in pancreatic cancer. We constructed an oncolytic adenovirus that expressed an shRNA targeting PKM2 (O Ad .R.shPKM2). Cells transduced with O Ad .R.shPKM2 exhibited increased apoptosis induction and reduced autophagy. This study indicated that PKM2 could be an effective therapeutic target for pancreatic cancer.

Results
PKM2 is highly expressed in pancreatic cancer samples and predicts poor survival. To determine whether PKM2 expression has clinical implications in human pancreatic cancer, we detected PKM2 expression in a tissue microarray containing 282 specimens (198 pancreatic cancer tissues and 84 adjacent noncancerous tissues) using immunohistochemistry (IHC). PKM2 expression in the pancreatic cancer tissues was significantly upregulated compared with that in the adjacent noncancerous tissues (Figures 1a and b). We further assessed the correlation between PKM2 expression and the clinical characteristics of pancreatic cancer patients (Table 1), and found that PKM2 expression was not related to gender or tumor stages but was significantly associated with age (P = 0.0020) and lymph node metastasis (P = 0.0436). High levels of PKM2 expression were correlated with worse overall survival in these pancreatic cancer patients (Figure 1c). In addition, PKM2 expression was higher in several pancreatic cancer cell lines (AsPC-1, PANC-1, BxPC-3 and CFPAC-1) than in the normal pancreatic duct cell line hTERT-HPNE (Figures 1d and e).
PKM2 enhances pancreatic cancer cell proliferation and migration as well as tumor formation. We constructed PKM2 knockdown cells by transducing PANC-1 and BxPC-3 cells with lentivirus containing PKM2-targeted shRNA. The levels of PKM2, but not PKM1, were significantly decreased in stably transduced cells, and shPKM2-1 exhibited better knockdown efficiency (Figures 2a, b, Supplementary Figures S1a and b). Moreover, knockdown of PKM2 decreased cyclin D1 expression, which is a downstream target of PKM2 (Supplementary Figure S1c). 8 PKM2 knockdown significantly inhibited cell viability as measured by the CCK-8 and cell-counting assays (Figure 2c, Supplementary Figures S1d and e). Cell migration in PANC-1 and BxPC-3 cells with PKM2 knockdown was suppressed (Figures 2d and e). We then sought to test the ability of tumor formation in nude mice using PANC-1 shPKM2-1 and shCtrl cells, respectively. Silencing PKM2 significantly inhibited tumor growth ( Figure 2f). The tumor mass and the PKM2 expression levels in tumors derived from cells expressing shPKM2-1 were significantly lower than those in the shCtrl group (Figures 2g and h). In addition, we found that all the mice bearing PANC-1 shCtrl tumors (4/4) but none of the mice bearing PANC-1 shPKM2-1 tumors (0/4) had liver metastases (Supplementary Figures S1f and g). The levels of aspartate aminotransferase (AST) and alanine aminotransferase (ALT) in the PANC-1 shPKM2-1 tumors were outside the normal range observed in the control group (Supplementary Figure S1h).
Lentivirus-driven silencing of PKM2 impaired pancreatic cancer growth by inhibiting autophagy rather than inducing apoptosis. It has been reported that autophagy facilitates the acquisition of additional nutrition for pancreatic cancer cells to increase survival. [19][20][21][22] We detected the basal autophagy levels and observed a high LC3-II/LC3-I ratio as well as elevated Beclin-1 and lower SQSTM1/P62 levels, indicating that basal autophagy levels in pancreatic cancer cells are higher than those in normal cells (Figure 3a). [23][24][25] The autophagy inhibitor 3-methyladenine (3-MA) remarkably declined PANC-1 and BxPC-3 cell proliferation. Conversely, the autophagy activator lithium chloride (LiCl) can accelerate cell proliferation (Figures 3b and c). Knockdown of PKM2 in PANC-1 and BxPC-3 cells caused a significant decrease in the LC3-II/LC3-I ratio and Beclin-1 expression in conjunction with P62 accumulation (Figure 3d). The number of LC3 puncta was reduced in cells with silenced PKM2 (Supplementary Figures S2a-c). A small number of autophagic structures was found in PKM2-knockdown cells (Supplementary Figure S2d), and the mRNA and protein expression levels of the autophagy-related transcription factors HIF-1α and FoxO3a were both reduced in PKM2knockdown cells ( Supplementary Figures S2e-g). LiCl reversed the inhibition of cell proliferation caused by PKM2 knockdown (Figures 3e and f); cell growth was further inhibited by 3-MA in the same cells (Figures 3e and g). These results suggested that PKM2 knockdown inhibited cell growth by suppressing autophagy. We next investigated whether PKM2 knockdown could induce cell apoptosis. Compared to cells with native PKM2 levels, cells with      (Figure 5a), and this autophagy was distinctly inhibited by O Ad .R.shPKM2 treatment (Figure 5b and Supplementary Figures S4a-c). Interestingly, we found that O Ad .R could induce apoptosis in shPKM2-transduced PANC-1 cells (Figures 5c, d and Supplementary Figure S4d). The number of annexin V-positive cells was remarkably elevated after O Ad .R.shPKM2 treatment (Figure 5e and Supplementary Figure S4e); additionally, the protein levels of procaspase-8 and procaspase-3 were reduced, and the cleaved form of PARP protein was increased (Figure 5f), indicating that O Ad .R. shPKM2 could induce cell apoptosis.
We next examined whether O Ad .R.shPKM2 could inhibit autophagy and induce apoptosis in PANC-1 xenograft tumors by using transmission electron microscopy and TUNEL staining, respectively. The number of autophagic structures was decreased in tumors treated with oncolytic adenovirus, and a few autophagic structures were observed in tumors treated with O Ad .R.shPKM2 (Figure 5g). The percentage of TUNEL-positive cells in the O Ad .R.shPKM2 group was significantly higher than that in the O Ad and O Ad .R groups (Figures 5g and h).

Discussion
Here, we showed that PKM2 is highly expressed in pancreatic cancers and is associated with poor patient survival. We also found that PKM2 modulated cell proliferation and migration as well as tumor formation of pancreatic cancer. A mild reduction in autophagy was observed in PANC-1-shPKM2 cells, indicating that impaired autophagy was mediated by PKM2 knockdown in cells with high autophagic flux. Targeting PKM2 using an RGD-modified oncolytic adenovirus (O Ad .R. shPKM2) displayed a potent cell cytotoxic effect. O Ad .R. shPKM2 induced apoptosis and inhibited autophagy both in vitro and in vivo.
We detected the PKM1 and PKM2 expression levels in pancreatic cancer cell lines and in patient samples (Figure 1 and Supplementary Figure S5), and found that PKM2 but not PKM1 was related to patient survival, which was in agreement with the results of other studies (Figure 1c and Supplementary Figure S5c). 12 Our results are also consistent with previous studies reporting that PKM2 levels are increased in other cancer types such as HCC, pulmonary adenocarcinoma, tongue cancer and glioma. 12,[26][27][28][29] Older pancreatic cancer patients expressed lower levels of PKM2 (Table 1). PKM expression is tissue-specific and is controlled by development, diet and hormones, 30,31 which may explain why PKM2 expression decreases with age. Importantly, our data indicated that all patients with lymph node metastasis displayed elevated PKM2 expression. Pancreatic cancer cells lacking PKM2 knockdown also showed enhanced migratory capabilities (Figures 2d and e). Mice bearing tumors with low levels of PKM2 expression had no liver metastases ( Supplementary  Figures S1f and g). The oncolytic adenovirus-treated mice were killed less than 2 months, and no metastases were found. We speculated that this timeframe was too short to form measurable metastases. PKM2-mediated regulation of migration in other cancers has also been reported. PKM2 overexpression was shown to promote colon cancer cell migration and cell adhesion by regulating STAT3-associated signaling. 9 In addition, downregulation of PKM2 by miR-let-7a inhibited migration in gastric cancer. 32 However, elucidating the underlying mechanism of PKM2 in regulating pancreatic cancer metastasis requires further investigation.
Our results showed that PKM2 could promote pancreatic cancer progression (Figures 2c-g), which was similar with previous studies of other cancer types such as non-small cell lung carcinoma, myeloma and gastric cancer. 7,32,33 Our data also revealed that PKM2 knockdown inhibited autophagy in pancreatic cancer cells in vitro and in vivo (Figures 3d, 5g and Supplementary Figures S2a-d). The expression levels of the autophagy-related factors HIF-1α and FoxO3a were decreased in PKM2-knockdown cells ( Supplementary  Figures S2e-g). PKM2 is a target of HIF-1α, and PKM2 expression can influence the transcription and transcriptional activity of HIF-1α as a form of feedback. 34 HIF-1α activates the transcription of the gene encoding BNIP3, a BH3 domain protein that induces selective mitochondrial autophagy by binding to Bcl-2 to release Beclin-1. 35 It was reported that downregulation of HIF-1α by siRNA could decrease autophagy by reducing the levels of free Beclin-1. 36 FoxO3a localizes to the cell nucleus and activates the expression of genes involved in autophagy pathways. 37 Moreover, FoxO3a is able to induce the transcription of target genes involved in cell cycle arrest, cell death, cell metabolism and stress resistance. 38,39 In cells depleted of energy due to blocked p38α activity, FoxO3a was upregulated and autophagy was induced. 40 In pancreatic cancer cells with higher basal levels of autophagy (Figure 3a), PKM2 knockdown reduced the expression of FoxO3a and impaired autophagy ( Supplementary Figures S2f-g). However, how FoxO3a is regulated in pancreatic cancer cells is still unclear, and the mechanism by which PKM2 regulates autophagy still requires validation with additional experiments.
Oncolytic adenovirus is an attractive vector that can selectively replicate in tumor cells and can constantly express either therapeutic proteins or shRNA with virus replication. Adenovirus modifications include (1) an RGD motif containing a peptide sequence to improve viral insertion into the target cell; (2) a survivin promoter driving E1A gene expression; and (3) loss of the regulation of viral E1B 55K RNA export and the apoptosis-restraining domain of E1B 19K. [41][42][43] Owing to the low CAR expression levels in pancreatic cancer cells ( Supplementary Figures S3a and b), the adenovirus infection efficiency was poor. The RGD modification significantly increased the ability of adenovirus to enter cells with low levels of CAR expression ( Supplementary Figures S3f and g). Cells transduced with a survivin promoter-driven oncolytic adenovirus displayed increased clinical safety. Utilizing a pancreatic cancer-specific promoter such as the Mucin1 Promoter (MUC1) to control E1A expression also enhanced the replication selectivity in pancreatic cancer. 44 Systemic administration of oncolytic adenoviruses has favorable toxicity profiles in cancer patients. DNX-2401, which has received fast-track FDA approval, is an orphan drug for the treatment of malignant glioma. 45 In addition, the development of other oncolytic viruses has recently made great strides toward availability for treating cancer patients. Talimogene laherparepvec (T-VEC; IMLYGIC, Amgen, Thousand Oaks, CA, USA) is an oncolytic herpes simplex virus containing granulocyte macrophage colony-stimulating factor (GM-CSF). This virus was tested in patients with unresected melanoma (stages IIIB to IV) in a randomized open-label phase III trial, and was compared to the effects of GM-CSF alone. T-VEC in combination with ipilimumab was administered to previously untreated patients with unresectable stage IIIB-IV melanoma. An intratumoral reovirus infusion for the treatment of recurrent malignant gliomas is currently in a phase I clinical trial. [46][47][48] Calabretta et al. 13 observed that knockdown of PKM2 enhanced the sensitivity of pancreatic cancer cells to chemotherapy-induced apoptosis. In this study, lentiviralmediated knockdown of PKM2 had no effect on apoptosis compared to the control cells (Figures 3h and i). We also found that O Ad .R remarkably enhanced apoptosis in PANC-1 cells infected with lentivirus containing shPKM2 (Figures 5c, d and Supplementary Figure S4d). It has been reported that oncolytic adenovirus can cause apoptosis in various cancer cells. 16,18,49 Therefore, we speculated that oncolytic adenovirus might trigger apoptosis in cells with adenoviral vectormediated PKM2 knockdown. There exists complex crosstalk between autophagy and apoptosis because they share common stimuli and signaling pathways. 50 Induction of O Ad .R.shPKM2 inhibits pancreatic cancer growth Y Xu et al apoptosis is associated with caspase-mediated cleavage of Beclin-1 and PI3K, [51][52][53][54] an event that impairs the autophagic function of Beclin-1. In addition, recent studies have shown that caspase-8 activation can be degraded by autophagy, 24 which further suggests the existence of a feedback mechanism that regulates both autophagy and apoptosis. On the basis of this notion, the apoptosis induced by adenoviral vectors may facilitate the ongoing feedback between autophagy and apoptosis. Hence, targeting the crosstalk between autophagy and apoptosis is a promising new approach for pancreatic cancer therapy. In all, our data showed that PKM2 is critical in pancreatic cancer and is related to cell proliferation, migration and metastasis. PKM2 regulated cell growth by inhibiting autophagy. Expression of O Ad .R.shPKM2 in cells exhibited robust effects in vitro and in vivo by inducing apoptosis and inhibiting autophagy, suggesting a potential antitumor therapy for pancreatic cancer.

Materials and Methods
Patients and tissue microarray. Tissue microarrays consisting of 198 pancreatic cancer and 84 adjacent normal specimens were provided by Bin Xu. The pathological information was retrieved from the Pathology Department of Shanghai 10th People's Hospital. The tumors were categorized as clinical stages I, II, III and IV based on the seventh edition of the American Joint Committee on Cancer criteria. All the human samples were collected after patients provided informed consent, and the approval for specimen usage was obtained from the ethics committee at the 10th People's Hospital. Studies involving these samples were approved and handled in accordance with the Institutional Review Board of the Institute of Biochemistry and Cell Biochemistry, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences.
Immunohistochemistry. The sections of tissue microarrays were subjected to IHC assay according to the established protocols. The stained tissue microarrays were analyzed by a pathologist blinded to the patient status. The levels of PKM2 and PKM1 expression were defined as values ranging from 0 to 3 according to the intensity, and quantification of PKM2 (Signalway Antibody, Washington, MD, USA) and PKM1 (Proteintech, Chicago, IL, USA) expression was calculated as (labeling intensity × percentage of positive cells). Establishing cells with stable PKM2 knockdown. shRNAs against PKM2 (shPKM2-1: 5′-CATCTACCACTTGCAATTA-3′, shPKM2-2: 5′-CCATAATC GTCCTCACCAA-3′) as well as a scramble control shRNA (5′-CTACCGTTG TTATAGGTG-3′) were synthesized and cloned into the pLVX vector according to the standard protocols. Lentivirus was produced in 293T cells by transfecting pLVX-shPKM2 and the corresponding lentiviral helper plasmids, and target cells were infected with the lentiviral vectors and selected using a specific concentration of puromycin.
Quantitative RT-PCR. Total RNA was isolated using TRIzol reagent (CWBIO, Beijing, China), and single-strand cDNA was synthesized using a ReverTra Ace qPCR RT kit (Toyobo, Osaka, Japan). RNA expression was analyzed using SuperReal Premix Plus (TIANGEN, Beijing, China) according to the manufacturer's protocols. The sequences of all the primers are shown in Supplementary Table S1. Each assay was performed in triplicate. Proliferation assays. Cell proliferation was measured using a CCK-8 kit (Dojindo, Kumamoto, Japan). Cells were incubated with 10 μl of CCK-8 at 37°C for 1.5 h, and the absorbance at wavelengths of 450 and 630 nm were applied using a BioTek Eon Microplate Reader (BioTek, Washington, CA, USA).
High-content cell number counting. Cells were fixed with 4% paraformaldehyde and labeled with 10 μM Hoechst (Thermo Fisher Scientific) to stain DNA. Fluorescent images were analyzed using an Operetta high-content imaging system (PerkinElmer, Waltham, MA, USA) at × 20 magnification. Different parameters of the cytoskeletal changes were measured using Harmony High Content Imaging and Analysis Software (PerkinElmer).
Crystal violet nucleus staining. After the cells were exposed to 2% crystal violet in 20% methanol for 15 min, they were washed with distilled water and photographed.
Migration assays. Cell migration assays were performed using transwell inserts (8 μm, BD Biosciences, San Jose, CA, USA) placed in 24-well plates. A total of 8 × 10 4 cells were plated in the upper chambers of the transwell inserts in 200 μl of serum-free medium, whereas the medium in the bottom chamber contained 15% fetal bovine serum. Following a 20 h incubation at 37°C, the inserts were stained with 0.5% crystal violet and gently washed, followed by imaging with an Olympus camera and counting using Image-Pro Plus 7.0 software (Media Cybernetics, Rockville, MD, USA).
Immunofluorescence microscopy. Cells were fixed with 4% (w/v) paraformaldehyde, permeabilized using 0.1% (w/v) Triton X-100 and blocked with 1% bovine serum albumin. The cells were stained overnight with an anti-LC3 antibody (Sigma) and then incubated for 1 h at room temperature with a Cy3conjugated secondary antibody (Abcam) diluted in PBS containing 1% FBS. Images were captured using a laser scanning confocal microscope.
Adenovirus construction and identification. The expression cassette of shPKM2 was inserted into the deleted-E1B region of a shuttle vector (pShuttlesurvivin-E1A-E1B). Different oncolytic adenoviral plasmids were generated using homologous recombination of the shuttle vector and adenoviral backbone plasmid in the Escherichia coli strain BJ5183. Viruses were packaged and amplified in HEK-293 cells followed by centrifugation in a graded CsCl solution for purification. The viral titer was measured using a QuickTiter Adenovirus Titer Immunoassay Kit (Cell Biolabs, San Diego, CA, USA). Viral genomes for identification were extracted using the Blood Genome Extract Kit (Generay, Shanghai, China) according to the manufacturer's instructions. The existence of the RGD peptide-coding sequence and the E1B deletion and wild-type contaminants were shown using PCR and sequencing with the appropriate primers (Supplementary Table S1). PKM2 gene expression was examined using western blotting.
Apoptosis detection. Cell apoptosis analysis was conducted using a fluorescein isothiocyanate-conjugated Annexin V (Annexin V-FITC)/Propidium Iodide Apoptosis Detection kit (BD Biosciences) according to the manufacturer's instructions. Cells were analyzed using a BD LSRII flow cytometer, and all the experiments were repeated three times.
Animal experiments. All animal experiments were performed according to protocols approved by the US Public Health Service Policy on Humane Care and Use of Laboratory Animals. Four-week-old female BALB/c nude mice were purchased from the Animal Core Facility (Shanghai, China).
For the tumorigenicity assay, 5 × 10 6 PANC-1 shCtrl or shPKM2-1 cells were mixed with Matrigel (BD Biosciences) at a ratio of 2 : 1 and subcutaneously injected into the right and left rear back regions of mice, respectively. The tumor volume was measured every 7 days using a Vernier caliper and calculated using the following equation: (length × width 2 )/2.
To determine the in vivo antitumor effect of adenoviruses on PANC-1 xenografts, we premixed 5 × 10 6 PANC-1 cells with Matrigel at a ratio of 1 : 2 and subcutaneously injected the mixture into the right rear back region of each mouse. When the tumors reached 80-120 mm 3 in size, the mice were randomly divided into four groups (six mice per group) using R software (The University of Auckland, Auckland, New Zealand). Either oncolytic adenovirus (2.5 × 10 8 PFU per mouse) or PBS was intratumorally injected every other day (four injections total), and the tumor volume was measured every 3 days. At the end of the experiment, the tumors were resected from the killed mice for histology (hematoxylin and eosin (H&E) staining) and immunofluorescence staining analyses. The serum levels of AST and ALT were also measured.
Histological analysis and TUNEL staining. Formalin-fixed tissues were paraffin-embedded and sliced into 5-μm sections stained using H&E. The TUNEL assay was performed using an in situ cell death detection kit (Roche) according to the manufacturer's instructions. Images were captured using an Olympus BX51 microscope (Olympus Optical, Tokyo, Japan). The number of TUNEL-positive cells was counted in six random fields on each coverslip (two per condition), and the average number of TUNEL-positive cells per random field was determined for each condition tested and quantified using Image-Pro Plus 7.0 software.
Transmission electron microscopy. Tumor tissues were fixed with 2.5% glutaraldehyde overnight, followed by 1% osmium tetroxide for 1.5 h. The tissue samples were then dehydrated using a graded series of ethanol, rinsed with acetone and permeabilized overnight with embedding buffer. Sections (70-nm thickness) were dual-stained with 2% uranyl acetate and lead citrate. Autophagosomes were observed using transmission electron microscopy (FEI, Hillsboro, OR, USA).
Statistical analysis. Data are presented as the mean ± S.D. Comparisons between two groups were performed using the unpaired Student's t-test. The survival rate was calculated using Kaplan-Meier analysis. All tests were conducted using GraphPad Prism 6.0 software (GraphPad Software, LaJolla, CA, USA).