Background
The cancer stem cell (CSC) theory postulates that tumors may arise from a small fraction of cancer cells that are response for tumor initiation and maintenance [
1]. The high metastatic potential and intense resistance to chemotherapy and radiation therapy in several cancers have been linked to CSCs, revealing new opportunities for cancer treatment [
2‐
4]. Pancreatic CSCs have been identified by several putative markers, such as CD44, CD24, EpCAM, CD133, and aldehyde dehydrogenase 1 (ALDH1) [
5‐
7]. The presence of CSCs in pancreatic cancer was associated with poor patient outcomes [
8]. Pancreatic tumors containing higher percentage of CSCs displayed enhanced chemoresistance and metastatic activity [
6,
9,
10]. Accordingly, therapies that selectively target CSCs may offer a greater promise for pancreatic cancer treatment.
Autophagy is a cellular self-degradation process that is required to maintain cell homeostasis. Alterations in autophagy-related signaling pathways frequently occur in many human diseases, including cancer [
11]. Considerable evidence has supported that autophagy maintains the survival of cancer cells through conferring stress tolerance and limiting damages [
12]. In pancreatic cancer, sustained activation of autophagy is associated with poor survival [
13]. Genetic or pharmacologic abrogation of autophagy suppressed the growth of pancreatic cancer cells
in vitro and led to tumor regression
in vivo due to autophagy inhibition-mediated reactive oxygen species production, DNA damages and altered cell metabolism [
14]. Therefore, autophagy is required for pancreatic cancer progression. Because autophagy acts as a survival pathway in cells under stress, much attention has been paid to the role of autophagy in CSC biology. Genetic inhibition of autophagy reduced the proportion of breast cancer cells bearing a CD44
+/CD24
-/low CSC-like phenotype, suggesting the role of autophagy in maintaining the typical breast CSCs [
15]. Blockade of both autophagy flux and lysosomal proteolyic activity by K
+/H
+ ionophore Salinomycin effectively reduced the population of ALDH
+ breast CSCs [
16]. Treatment with the autophagy inhibitor chloroquine (CQ) strongly promoted γIR-induced cell death in highly radioresistant patient-derived stem-like glioma cells [
17]. In pancreatic cancer cells, high levels of autophagy have been observed under basal conditions [
14,
18]; however, the relation between autophagy and pancreatic CSCs remains to be explored.
Osteopontin (OPN), a secreted glycoprotein, has been implicated in a variety of physiological and pathophysiological processes, such as bone remodeling, angiogenesis, immunity, atherosclerosis, and cancer progression [
19,
20]. By interacting with CD44 family of receptors or integrin αvβ3, OPN can activate several downstream signaling pathways, such as PI3K/AKT, NF-κB, and MEK/ERK [
21]. OPN overexpression in many types of cancer has been considered a poor prognostic marker [
22]. Recently, increased OPN expression has been observed in sphere-growing stem-like cells of pancreatic cancer compared with their adherent counterpart [
23]. OPN overexpression significantly increased the formation of spheres derived from the brain tumor cells of p53/PTC double heterozygous mice [
24], suggesting a role of OPN in regulating CSC activity. Given that OPN can induce autophagy directly through integrin/CD44 and p38 MAPK-mediated pathways in vascular smooth muscle cells [
25], we sought to investigate whether OPN can increase pancreatic CSC activity through stimulation of autophagy.
Discussion
CSCs are believed to drive the neoplastic progression, tumor recurrence, and metastasis. A comprehensive understanding of the biology of CSCs allows the development of effective therapies for pancreatic cancer. In this study, we have observed elevated levels of autophagy in pancreatic CSCs. Autophagy blockade profoundly reduced pancreatic CSC activity. OPN could induce autophagy by activating NF-κB and thus upregulate pancreatic CSC activity.
Recently, altered autophagy have been linked to CSCs. Dormant breast cancer stem-like cells induced by farnesyl transferase inhibitors expressed high levels of autophagy markers, such as ATG5, ATG7, and ATG12. Blockade of autophagy by 3-methyladenine decreased the proportion of stem-like cells [
30]. The autophagic flux was upregulated in the sphere-forming breast cancer cells expressing ALDH1 compared with the bulk population. Knockdown of either
BECN1 or
ATG7 dramatically suppressed the sphere formation of breast CSCs [
31]. In pancreatic cancer, cells with stem-like properties displayed stronger autophagic activity than did those with fewer stem cell markers. The enhanced autophagy enabled CSCs to survive under hypoxic stress [
32]. Consistent with these reports, we observed that pancreatic CSCs exhibited elevated autophagy in both clinical specimens and cell lines. Blockade of autophagy by pharmacological or genetic inhibitors reduced CSC populations, sphere-forming ability, drug resistance, and tumor formation, revealing the requirement of autophagy for pancreatic CSC maintenance. Therefore, autophagy may act as a pro-survival regulator for pancreatic CSCs; however, the detailed mechanism is unclear. Similar to normal stem cells, CSCs have the ability to self-renew and differentiate through activation of several embryonic stem cell signaling pathways, including Wnt/β-catenin, Notch, and Hedgehog [
33,
34]. In our study, when autophagy-related genes were knocked down in pancreatic cancer cells, β-catenin and Sonic Hedgehog expression was significantly decreased but Notch1 expression was not (Additional file
6: Figure S6), suggesting that autophagy may regulate pancreatic CSC activity by modulating Wnt/β-catenin and Hedgehog signaling.
Currently, CD44, CD133, and ALDH1 are the most commonly used markers in identification of pancreatic CSCs [
35]. These markers, however, are not universal and perfect to identify CSCs because not all CSCs express the markers, and non-CSCs may also exhibit the markers. Therefore, the markers can be used for identification of CSC-rich subpopulations but not for unambiguous isolation of all of the CSCs. Likewise, autophagy may not be exclusively activated in CSCs but also in non-CSCs of pancreatic cancer. These may be the causes of the weak correlation between autophagy and single CSC marker expression (CD44 or CD133) in pancreatic tumor tissues. Since there is a partial overlap between CD44
+/CD24
+/ESA
+ and CD133
+ pancreatic cancer cells [
6], it is conceivable that a combination of the markers may help to mark more pure CSCs than a single marker. Indeed, our group found that CD44
+/CD133
+ cells isolated from PANC-1 cells were capable of forming tumorspheres
in vitro, exhibited tumor-initiating potentials
in vivo, and profoundly responded to Wnt pathway activation or inhibition [
36]. Accordingly, in our
in vitro experiments, we used CD44 in combination with CD133 to identify a CSC population. We observed that the number of CD44
+CD133
+ cells was increased by the autophagy inducer rapamycin but was decreased by the autophagy inhibitor CQ or knockdown of autophagy-related genes, revealing the requirement of autophagy for the maintenance of CD44
+CD133
+ cells in pancreatic cancer. Unlike CD44 and CD133, ALDH1 showed a high correlation with LC3 expression in pancreatic tumors. This may be because ALDH1 expression and autophagy activation are regulated by some common signaling pathways. ALDH1 reportedly is a target gene of the NF-κB pathway [
37‐
39]. Together with our observation that OPN triggered autophagy by activating NF-κB, the NF-κB pathway may act as an upstream positive regulator of both ALDH1 and autophagy in pancreatic cancer cells. However, further investigation is needed to substantiate our speculation.
The efficacy of current available therapeutic methods against pancreatic cancer remains limited. Gemcitabine is the standard chemotherapy used as first-line treatment for patients with advanced pancreatic cancer; however, the survival extension is only marginal [
40,
41]. The high incidence of tumor relapse following treatment is believed to be caused by the presence of residual CSCs that are resistant to conventional therapies [
42]. Impairment of CSC activity has been reported to reverse gemcitabine resistance [
43,
44]. In this study, gemcitabine alone showed low efficacy in eliminating pancreatic CSCs and preventing xenograft tumor formation. However, blockade of autophagy by genetic knockdown or autophagy inhibitor CQ sensitized pancreatic CSCs to gemcitabine and thus almost eradicated xenograft tumors in mice. Our results are in good agreement with a previous study reporting that autophagy exerts a cytoprotective effect against chemotherapy drugs 5-FU and gemcitabine in pancreatic cancer cells, and inhibition of autophagy by CQ potentiated the growth-inhibitory effects of 5-FU and gemcitabine [
18]. These findings highlight the exciting possibility that gemcitabine in combination with drugs that inhibit autophagy evokes a synergistic antitumor effect for pancreatic cancer treatment. Notably, a recent study demonstrated that CQ can counteract primary pancreatic CSC activity through inhibition of CXCR4 and Hedgehog signaling but independent of autophagy blockade, which is inconsistent with our result that, in primary pancreatic tumor SP-1 cells, CQ caused the accumulation of LC3 puncta, indicative of autophagy inhibition (Additional file
4: Figure S4A). CQ has also been reported to exert its actions not through inhibition of autophagy but through impairment of lysosomal functions [
45]. Therefore, whether CQ-mediated inhibition of pancreatic CSC activity depends on its ability to block autophagy needs further confirmation.
OPN showed a positive association with coexpression of CD44/CD133 or LC3/ALDH1 in pancreatic tumors in this study; however the correlation was not strong. It is known that OPN exerts its biological functions through interaction with the integrin and CD44 families of cell surface receptors [
20]. OPN-mediated functions may be determined by the different receptors. Therefore, analysis of OPN expression in combination with associated receptor expression in pancreatic cancer cells may be more appropriate to evaluate the correlation between OPN signaling and CSC activity than only OPN detection. Previously, OPN has been reported to stimulate autophagy in vascular smooth muscle cells [
25]. We here found that OPN could also trigger autophagy via NF-κB in pancreatic cancer cells, and this finding is supported by several studies showing that NF-κB can act as an inducer of autophagy. For instance, autophagy was induced by NF-κB for cell survival during the heat shock response [
46]. Upon hypoxia, ROS-activated NF-κB can trigger autophagy in breast cancer cells [
47]. Intriguingly, autophagy seemed to be enhanced by the JAK2/STAT3 inhibitor AG490 in PANC-1 cells. Recently, STAT3 has been reported to suppress LC3 expression, thereby inhibiting autophagy and growth of pancreatic cancer cells [
48]. Therefore, whether AG490 induces autophagy by inhibiting STAT3 activity or through its unknown off-target effects remained to be clarified.
Methods
Cell lines and culture conditions
Human pancreatic cancer cell lines PANC-1, MIA PaCa-2, and AsPC-1 were obtained from American Type Culture Collection. All the cell lines and their derived cells were maintained in RPMI 1640 medium (Hyclone, SH30027.02) with 10 % fetal bovine serum (Invitrogen). Cells were incubated at 37 °C in a humidified atmosphere containing 5 % CO2.
Isolation of primary tumor cells
Primary pancreatic tumor SP-1 cells were collected from the centrifugal sedimentation of ascites obtained from a 50-year old male patient. The histopathological examination of the patient’s tumor tissue revealed a well-differentiated ductal adenocarcinoma of the pancreas. Briefly, the bulk of ascites cells were seeded on culture flasks containing RPMI 1640 medium (Hyclone, SH30027.02) with 10 % fetal bovine serum (Invitrogen), 2 mM L-glutamine and 1 % penicillin/streptomycin (Caisson) at 37 °C in a humidified atmosphere with 5 % CO2. After two weeks, fibroblasts and stellate cells were removed by two rounds of serial enzymatic detachment with 0.05 % Trypsin/ EDTA (Life Technologies). The resulting population of cells was confirmed as cancer cells by tumor formation assay in vivo. For the experiments described here, SP-1 cells were used in passages 8 to 12. Immortalization of the isolated tumor cells was not necessary.
Cells were seeded into 6-well ultra-low attachment plates (Corning, 3471) at a density of 4 × 104 cells/well with serum-free medium containing 10 ng/mL basic FGF (PeproTech, 100-18B), 20 ng/mL EGF (PeproTech, AF-100-15), insulin-transferrin-sodium selenite media supplement (Sigma-Aldrich, 3116), and 0.4 % bovine serum albumin (Sigma-Aldrich, A7030). After 14 days of culture, the number of spheres was counted and the spheres were disaggregated into single cells with accutase (Invitrogen, A11105-01). The viable cells were counted by Eosin Y (Sigma-Aldrich, 230251) and cultured for further use.
Measurement of cell viability
The growth or viability of cells was assessed with MTT reagent (Sigma-Aldrich, M2003). Cells were seeded into 96-well plates at a density of 4 × 103/well. After treatments, MTT was added to each well (final concentration of 0.5 mg/mL) followed by incubation at 37 °C for 4 h. The supernatant was removed, and DMSO was added to dissolve the blue-purple crystals of formazan. The optical density of the samples was measured at a wavelength of 540 nm by spectrophotometer (Thermo Scientific, Multiskan EX).
CSC identification and apoptosis detection
For CSC identification, 2 × 10
5 cells were stained for ALDH1, CD44, and CD133 (antibodies are listed in Additional file
7: Table S1). After washing with PBS, cells were analyzed by flow cytometry (FACS Canto II).
For detection of apoptosis, specific binding of Annexin V to phosphatidylserine was performed by incubating the cells in binding buffer containing a saturating concentration of Annexin V-FITC (BD Biosciences, 556547) according to the manufacturer’s protocol. Following incubation, cells were pelleted and analyzed by flow cytometry.
1 × 106 cells were subcutaneously inoculated into the right flank of 8-week old male NOD/SCID mice to form tumors. When the size of individual tumor reached 5 mm in diameter, animals were injected intraperitoneally with gemcitabine (100 mg/kg), CQ (60 mg/kg), and their combination once weekly for 4 weeks. Tumor volume was measured every week at the beginning of treatments using the formula: volume = a
2 × b/2 where a is the major diameter and b is the minor diameter vertical to a. Animals were raised and cared for according to the guidelines set up by the National Science Council, ROC. The animal experiments were approved by the Institutional Animal Care and Use Committee.
Lentiviral transduction and stable cell line generation
OPN, ATG5, ATG7, BECN1, and non-target short hairpin RNA (shRNA) vectors were purchased from the National RNAi Core Facility, Academia Sinica. Both plasmids pMD.2G and psPAX2 for lentiviral packaging were kind gifts from Dr. Mettling Clément (Institut de Génétique Humaine, CNRS UPR1142, Montpellier, France). The lentiviral particles were made by transfecting above-mentioned plasmids into 293 T cells and collected at 48 h post transfection. PANC-1 and MIAPaCa-2 cells were infected in the presence of 4 μg/mL polybrene (Sigma-Aldrich, AL-118). Puromycin (Sigma-Aldrich, P9620) was used for drug selection of infected cells to generate permanent cell lines.
ALDEFLUOR assay
ALDH1 activity was detected by ALDEFLUOR assay kit (Stem Cell Technologies, 01700) according to manufacturer’s protocol. Briefly, cells were incubated in Aldefluor assay buffer containing ALDH1 substrate bodipy-aminoacetaldehyde (1 μmol/L per 1 × 106 cells) at 37 °C for 50 min. As a negative control, a fraction of cells from each sample was incubated under identical condition in the presence of ALDH1 inhibitor diethylaminobenzaldehyde (15 μmol/L per 0.5 × 106 cells). The cells were then subjected to FACS analysis.
Cell lysis and Western blot analysis
The harvested cells were lysed in ice for 30 min with lysis buffer (Cell Signaling Technology, 9803). Lysates were cleared by centrifugation at 14,000 rpm for 10 min at 4 °C. Protein concentrations were measured by the Bradford assay (Bio-Rad Laboratories, 500–0006). For Western blot analysis, cell lysates were boiled for 5 min with sample buffer before being resolved in SDS–polyacrylamide gels. The proteins were transferred to PVDF membrane (Millipore, IPVH00010). The membrane was blocked with 5 % skim milk in TBST buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1 % Tween 20) for one hour and then stained with primary antibodies at 4 °C overnight followed by incubation with secondary antibodies (antibodies are listed in Additional file
7: Table S1). The binding of each antibody was detected using an enhanced chemi-luminescence kit (PerkinElmer, NEL105001EA). The signals were detected by X-ray films (Fuji, 47410 08399) or a UV transilluminator (UVP Ltd., BioSpectrum™ 500 Imaging System) and analyzed by the Gel-Pro Analyzer 4.0 software (Media Cybernetics).
Immunofluorescence staining and analysis of clinical samples
A total of 93 pancreatic cancer patients who underwent resection in National Cheng Kung University Hospital (NCKUH) were included in this study that was approved by Institutional Review Board of NCKUH. Anonymous archived samples of human pancreatic cancer, including both normal and malignant tissues, were obtained from Human Biobank of NCKUH for TMA construction. Paraffin-embedded TMAs were cut into 5 μm-thick sections and stained with primary antibodies (listed in Additional file
7: Table S1) at 4 °C overnight followed by incubation with secondary antibodies (listed in Additional file
7: Table S1) at room temperature for 1 h. Cell nuclei were counterstained with DAPI (blue; Molecular Probes, D3571) or DRAG5 (purple; Abcam, ab108410). Fluorescence imaging was performed using a laser scanning confocal microscope (Olympus, Fluoview 1000), and the signals were quantified with Tissue-Quest software. The percentage of protein staining for each tumor specimen was classified into two staining grades according to the mean value of protein expression (the high grade represents ≥ mean; the low grade represents < mean). Tumor heterogeneity was evaluated using Pearson correlation coefficients (
r) in two different cores from the same tumor blocks, as described previously [
49].
Statistics
Data are expressed as means ± SE from three independent experiments. Statistical analysis in this article was performed by one- or two-way ANOVA using Prism 5.0 software. The median survival was estimated using the Kaplan-Meier method. The association between studied variables was evaluated using Pearson’s correlation coefficient test in SPSS 17.0. Significance was set at P < 0.05.
Supplementary Materials and Methods are available as Additional file
8.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
MCY, HCW, YCH, and YSS conceived the study and designed the experiments. MCY, HCW, YCH, and HLT performed the experiments. HCW, YCH, TJC, and YSS analyzed the data and prepared the figures. HCW and MCY wrote the paper. Most of the clinical patients were treated and followed in the clinic by YSS. YSS supervised the study and revised the manuscript. All authors read and approved the final manuscript.