The human pancreatic cancer cell line, Panc-1, was obtained from the American Type Culture Collection (ATCC; Manassas, VA, USA) and frozen in our laboratory in liquid nitrogen and cultured in Dulbecco's modified Eagle's medium (DMEM; Life Technologies; Thermo Fisher Scientific, Inc., Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS; Gibco; Thermo Fisher Scientific, Inc., Paisley, UK), 100 U/ml of penicillin (Gibco; Thermo Fisher Scientific, Inc.), 100 µg/ml of streptomycin (Gibco; Thermo Fisher Scientific, Inc.) in a 5% CO₂ incubator at 37°C. For enriching CSCs from parental Panc-1 cells, a suspension culture supporting proliferation of undifferentiated cells was adopted (26). Briefly, Panc-1 cells were maintained in DMEM/Ham's Nutrient Mixture F-12 (F-12) (1:1) (Life Technologies; Thermo Fisher Scientific, Inc.) with the addition of epidermal growth factor (EGF, 20 ng/ml; Sigma-Aldrich; Merck KGaA, Darmstadt, Germany), human fibroblast growth factor basic (hFGFb, 10 ng/ml; Sigma-Aldrich; Merck KGaA) and 2% B27 (Life Technologies; Thermo Fisher Scientific, Inc.) for 14–21 days. Medium was half-refreshed every 3 days until the spheres were observed.

Cell viability was determined by the Cell Counting Kit-8 (CCK-8; Dojindo Laboratories Co., Ltd., Kumamoto, Japan) assay. Briefly, Panc-1 or Panc-1 cancer stem-like cells (Panc-1 CSCs) (5×10⁴) were cultured in 96-well plates. A total of 10 µl of freshly prepared CCK-8 solution was added to the cell culture for a 2-h co-incubation. The absorbance was measured at 620 nm. The cell viability was expressed as OD (OD620).

The apoptosis of Panc-1 or Panc-1 CSCs was determined by FACS analysis of Annexin V-FITC/propidium iodide (PI; Roche, Basel, Switzerland)-double stained cells. Briefly, cells were grown in a 6-well plate in serum-free medium. After treatment, Panc-1 and Panc-1 CSCs were trypsinized, washed with phosphate-buffered saline (PBS), resuspended in 200 µl PBS with 10 µl RNAase (10 mg/ml) and incubated at 37°C for 30 min. At the end of incubation, 50 µl of Annexin V-FITC/PI labeling solution (BD Biosciences, San Jose, CA, USA) was added and cells were analyzed for apoptosis using a flow cytometry (BD FACSCanto II; BD Biosciences).

For hypoxia induction, cells were cultured under a hypoxic condition (1% O₂, 5% CO₂ and 94% N₂) in a multigas incubator (MCO-19; Sanyo Scientific, Tokyo, Japan) for 6, 12 or 24 h. For normoxia incubation, cells were incubated under the condition of 20% O₂, 5% CO₂ and 75% N₂.

For detecting self-renewal capacity of Panc-1 CSCs, CSCs were grown in 6-well ultra-low attachment plates (Corning Inc., Corning, NY, USA) at a density of 1,000 cells/ml in well-defined serum-free medium at 37°C in a humidified atmosphere of 95% air and 5% CO₂. Fourteen days later, the spheres with a diameter of >40 µm were counted under an Olympus X71 (U-RFL-T) fluorescence microscope (Olympus Corp., Melville, NY, USA). Then, similarly, for secondary spheroids, the same number of CSCs from spheroids were reseeded for another 14 days. All the procedures were repeated 4 times.

Knockdown of the mRNA level of GPX4 was achieved by transient transfection of cells with siRNA duplexes (Thermo Fisher Scientific, Inc., Waltham, MA, USA), specific to the mRNA of GPX4. The relevant siRNA sequences were: GPX4 sense, 5′-UUCGAUAUGUUCAGCAAGAUU-3′ and antisense, 5′-UCUUGCUGAACAUAUCGAAUU-3′; negative control (NC) sense, 5′-GUUCAAUAUUAUCAAGCGGUU-3′ and antisense, 5′-CCGCUUGAUAAUAUUGAACUU-3′. According to the manufacturer's instructions, 5×10⁵ cells were grown in 2 ml of serum-free medium. A siRNA/transfection reagent complex was formed at room temperature by combining siRNA oligomer (50 nM) with 5 µl (2 µg/ml) Lipofectamine™ 2000 transfection reagent (Thermo Fisher Scientific, Inc.) in 0.5 ml Opti-MEM medium (Thermo Fisher Scientific, Inc.) and this was applied to cells for 48 h until they were harvested. Cells transfected with NC siRNA were considered as the control cells. For erastin treatment, 48 h after transfection, the cells were cultured in media supplemented with 2 µM erastin or an equal amount of dimethyl sulfoxide (DMSO) (mock group) for 2, 4, 6 and 8 h, respectively. After 3 washes with PBS, the cells were harvested for further analysis.

For H₂O₂ treatment, 48 h after transfection, cells were cultured in media supplemented with 50, 100, 200 and 500 µM H₂O₂, respectively, for 24 h. After 3 washes with PBS, cells were harvested for further analysis.

The full-length complementary DNA (cDNA) of GPX4 (human GPX4 GenBank accession no. NC_000019) was obtained from RiboBio Co., Ltd. (Guangzhou, China) and ligated into the NheI-XhoI site of the pcDNA3.1 vector (Life Technologies; Thermo Fisher Scientific, Inc.). The NheI and XhoI restriction enzymes and T4 ligase were purchased from Takara Bio (Heidelberg, Germany). A total of 0.8 µg of the plasmid or empty vector was mixed with 4 µl Lipofectamine™ 2000 transfection reagent (Thermo Fisher Scientific, Inc.) in 0.5 ml Opti-MEM medium (Thermo Fisher Scientific, Inc.) and this was applied to cells for 4 h followed by medium-refresh. After 24 h, the culture medium was refreshed containing 500 µg/ml geneticin sulfate 418 (G418; Sigma-Aldrich; Merck KGaA) for antibiotic selection. Two weeks later, survival-transfected cells were collected and maintained in 250 µg/ml G418.

For the apoptosis assay, cells were pretreated with 20 µM caspase inhibitor Z-VAD-FMK (Promega Corporation, Madison, WI, USA), 1 µM Ferrostatin-1 (Ferr-1; Sigma-Aldrich; Merck KGaA), or 0.3 µM Necrostatin-1 (Nec-1; Sigma-Aldrich; Merck KGaA) for 12 h followed by co-incubation of H₂O₂ or erastin, respectively. Twenty-four hours later, the cells were washed 3 times with PBS and harvested for further analysis.

The mRNA expression of GPX4, E-cadherin, vimentin, Slug, Snail and the housekeeping gene β-actin was determined by RT-qPCR using primer oligomers as followed: GPX sense primer, 5′-CGATACGCTGAGTGTGGTTTGC-3′ and antisense primer, 5′-CATTTCCCAGGATGCCCTTG-3′; E-cadherin sense primer, 5′-CGAGAGCTACACGTTCACGG-3′ and antisense primer, 5′-GGGTGTCGAGGGAAAAATAGG-3′; vimentin sense primer, 5′-GACGCCATCAACACCGAGTT-3′ and antisense primer, 5′-CTTTGTCGTTGGTTAGCTGGT-3′; Slug sense primer, 5′-CGAACTGGACACACATACAGTG-3′ and antisense primer, 5′-CTGAGGATCTCTGGTTGTGGT-3′; Snail sense primer, 5′-TCGGAAGCCTAACTACAGCGA-3′ and antisense primer, 5′-AGATGAGCATTGGCAGCGAG-3′; β-actin sense primer, 5′-CATGTACGTTGCTATCCAGGC-3′ and antisense primer, 5′-CTCCTTAATGTCACGCACGAT-3′. Reverse transcription was carried out amplifying 1 µg total RNA using a reverse transcriptase kit (Guangzhou Ribobio Co., Ltd.) according to the manufacturer's instructions. PCR was then carried out using PowerUp SYBR™-Green Master Mix (Thermo Fisher Scientific, Inc.) following the manufacturer's instructions. Briefly, the PCR conditions were started by denaturation for 5 min at 95°C and followed by 35 cycles of 95°C for 10 sec and 60°C for 1 min. mRNA levels were normalized against β-actin mRNA and for relative quantification, 2^−ΔΔCq method was used (27).

Total cellular ROS levels were determined using the Image-iT^™ LIVE Green Reactive Oxygen Species (ROS) Detection kit (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. A total of 1×10⁵ cells were used for analysis. Cells were washed with ice-cold 1X PBS and incubated with 50 µl staining solution containing 25 µM 5-(and-6)-carboxy-2′,7′-dichlorodihydrofluorescein diacetate (carboxy-H₂DCFDA) for 30 min in darkness. Cells were washed twice with ice-cold 1X PBS and ROS levels were measured at 495/529 nm (for carboxy-H₂DCFDA) wavelengths on a Multiskan spectrum microplate reader (Thermo Fisher Scientific, Inc.).

Panc-1 or Panc-1 CSCs were washed twice with ice-cold 1X PBS, and then resuspended in lysis buffer containing 25 mM HEPES buffer (pH 7.6), 3 mM MgCl₂, 40 mM KCl, 2 mM DTT, 5% glycerol and 0.5% NP-40, and 1X protease inhibitor, and were maintained in an ice bath for 30 min. After a 15-min centrifugation at 12,000 × g, at 4°C, the supernatant was collected as total lysate. The protein concentration was determined using the Bradford protein assay procedure (Sigma-Aldrich; Thermo Fisher Scientific, Inc.). Twenty micrograms of protein was subjected to 10% SDS-PAGE, transferred to a polyvinylidene fluoride (PVDF) membrane (Thermo Fisher Scientific, Inc.), and incubated at room temperature overnight in PBS containing 5% dried milk, 0.05% Tween-20 and primary antibodies: Anti-superoxide dismutase 1 antibody (cat. no. ab13498; diluted at 1:2,000), anti-superoxide dismutase 2/MnSOD antibody (diluted at 1:2,000), anti-glutathione peroxidase 1 antibody (cat. no. ab22604; diluted at 1:1,000), anti-glutathione peroxidase 4 antibody (cat. no. ab125066; diluted at 1:2,000) (all from Abcam, Cambridge, MA, USA); anti-Nanog XP^® rabbit monoclonal antibody (mAb) (cat. no. D73G4; diluted at 1:2,500), anti-c-Myc rabbit mAb (cat. no. D3N8F; diluted at 1:1,000), anti-Oct-4 antibody (cat. no. 2750; diluted at 1:1,000), anti-Sox2 antibody (cat. no. 3579; diluted at 1:2,000), anti-E-cadherin rabbit mAb (cat. no. 24E10; diluted at 1:2,000), anti-vimentin XP^® rabbit mAb (cat. no. D21H3; diluted at 1:1,000), anti-Slug rabbit mAb (cat. no. C19G7; diluted at 1:1,000), anti-Snail rabbit mAb (cat. no. C15D3; diluted at 1:2,000), anti-β-actin mouse mAb (cat. no. 8H10D10; diluted at 1:5,000) (all from Cell Signaling Technology, Inc., Danvers, MA, USA). After 4 washes in PBS containing 0.05% Tween-20, the membrane was incubated with secondary antibody [anti-mouse IgG, HRP-linked antibody (cat. no. 7076; diluted at 1:5,000), anti-rabbit IgG, HRP-linked antibody (cat. no. 7074; diluted at 1:5,000) (both from Cell Signaling Technology, Inc,)] for 1 h, washed 4 times and the bound antibody was detected by chemiluminescence, and analyzed with ImageJ software (National Institutes of Health, Bethesda, MD, USA).

All the data are expressed as means ± SD of 3 independent experiments. Unpaired Student's t-tests were used to compare the means of 2 groups. Differences between groups were analyzed by performing the ANOVA Kruskal-Wallis test, with Dunn's multiple group comparison test. P-values <0.05 were considered to indicate a statistically significant result.

After enriched Panc-1 CSCs were cultured for 7 and 21 days, it was obvious that the morphology of the pancreatic CSCs displayed sphere-like appearance (Fig. 1A). The ability of self-renew is the definition of a stem cell that is thought to be functionally mimicked by CSCs (28–30) and serial-replating assay which is widely employed (31), was performed to assess self-renewal of derived cells. As shown in Fig. 1B, no obvious difference in sphere formation ability was found between the cells at the 1st to 4th passage. We then determined whether the spheroid cells have the capacity of high colony formation using Soft agar colony formation assay. Sphere-forming assay was performed to evaluate the stemness property of the Panc-1 cells. The results indicated that the stemness property of the Panc-1 cells was higher than that of their parental counterpart after a period of adaptation to 10% FBS containing medium, instead of serum-free medium (Fig. 1C). And as expected, the capacity of cell proliferation in the spheroid cells was significantly higher than that of their parental counterpart (Fig. 1D). Furthermore, 4 transcription factors which regulate the self-renewal capacity of CSCs (32) were detected by western blot analysis and the results showed that spheroid cells derived from Panc-1 cells displayed obviously higher expression levels of these 4 factors, compared to these levels in the parental counterpart (Fig. 1E).

Hypoxia has been reported to be a promoter of maintaining the stem-like phenotype of CSCs (33). To confirm whether CSCs derived from Panc-1 cells are resistant to hypoxic exposure, Panc-1 or Panc-1 CSCs were exposed to hypoxia for 6, 12 or 24 h and stained using Annexin V-FITC/PI double staining. As shown in Fig. 2A, in normoxia condition, no detectable difference between Panc-1 and Panc-1 CSCs in apoptotic cell death rate (Annexin V-FITC-positive/PI-negative and Annexin V-FITC-positive/PI-positive subpopulation) was noted. After being exposed to hypoxia for 12 and 24 h, the apoptotic cell death rate in the Panc-1 CSCs (4.3±0.5% for 12 h; 0.8±0.6% for 24 h) was obviously lower than that of the Panc-1 cells (9.8±0.9% for 12 h; 11.6±1.7% for 24 h). Then the effects of hypoxia on cell viability were assessed. It was observed that, in Panc-1 CSCs, hypoxia exposure failed to decrease cell viability after 24 h (Fig. 2B). By considering the roles of SODs and GPXs in maintaining low intracellular ROS level in mammalian cells (34), western blot analysis was performed to detect the SOD and GPX protein levels in the Panc-1 and Panc-1 CSCs after hypoxic exposure. Expectedly, SOD1, SOD2 and GPX1 levels were induced by hypoxic exposure (Fig. 2C). However, GPX4 presented a relative high expression level in the Panc-1 CSCs compared with that in the Panc-1 cells, and GPX4 expressing levels were not obviously affected by hypoxic exposure in the Panc-1 CSCs. To confirm whether GPX4 presents a high endogenous level beyond hypoxic exposure, SODs and GPXs were compared in the Panc-1 and Panc-1 CSCs. In Fig. 2D, it was observed that, in the Panc-1 CSCs, GPX4 presented a consistently high endogenous level, which was obviously higher than that noted in the Panc-1 cells and did not respond to hypoxic exposure. This demonstrated that GPX4 potentially play critical roles in regulating physiological processes in panc-1 CSCs, not only after hypoxic exposure.

For identifying the transfecting efficiency of siRNA by Lipofectamine^™ 2000, Alexa Fluor 488-labeled siRNA-NC was introduced into the Panc-1 CSCs and was detected immediately after 3 PBS washes. As shown in Fig. 3A (upper panels), siRNA-NC was successfully introduced into the Panc-1 CSCs. Then, the knockdown efficiency of the siRNA targeted to GPX4 mRNA (siRNA-GPX4) was measured by RT-qPCR and western blot analysis 2 days after transfection. The mRNA level of GPX4 was downregulated by >50%. In addition, a detectable downregulation in the protein level of GPX4 was also confirmed (Fig. 3A, lower panels).

By considering the inactivating ability of erastin to GPX enzymes, especially GPX4 (35), we examined the ROS accumulation of Panc-1 CSCs treated with targeted siRNA 3 days after transfection treated with 2 µM erastin (Fig. 3B). Cells transfected with siRNA-GPX4 presented significant high ROS levels at 2- or 4-h of erastin treatment, while the control cells presented no detectable ROS accumulation. We then examined the effects of GPX4 knockdown on the cell growth of Panc-1 CSCs, which were exposed to hypoxia for 12 h previously, using CCK-8 assay. There was no significant difference in cell viability between cells transfected with GPX4 and control siRNA up to 0 and 2 h after erastin treatment (Fig. 3C). However, at 4 h after erastin treatment, the viability of the siRNA-GPX4 transfected cells was significantly lower than that of the siRNA-NC transfected cells, suggesting that GPX4 exerts protective effect for cell viability against erastin treatment. We further clarify the ROS eliminating effect of GPX4 under oxidative stress condition induced by H₂O₂. As shown in Fig. 3D right panel, NAC pre-treatment scavenged ROS in cells. In Fig. 3D left panel, without NAC treatment (mock groups), GPX4 knockdown promoted the ROS accumulation at 50 and 100 µM (Fig. 3D).

Previously, we showed that the presence of GPX4 exerted protective effects to Panc-1 CSCs under oxidative stresses including H₂O₂ or erastin (Fig. 3). To further investigate whether GPX4 plays certain roles without oxidative stress in Panc-1 CSCs, its effects on cell proliferation was determined. By performing CCK-8 assay and cell counting, it was observed that, without disturbing cell viability, GPX4 knockdown decreased cell proliferation compared to that of the siRNA-NC (Fig. 4A). Flow cytometric analysis after PI staining further showed that GPX4 knockdown arrested the cell cycle at the G₁/G₀ phase, indicating cell cycle arrest at this phase (Fig. 4B). The results of functional analysis showed that knockdown of GPX4 significantly suppressed sphere formation ability (Fig. 4C), migration and invasion capacity (Fig. 4D and E) in Panc-1 CSCs, as compared with siRNA-NC cells. Collectively, our results revealed that knockdown of GPX4 suppressed the stemness phenotype of Panc-1 CSCs and inhibited in vitro cell functions.

We further evaluated the effects of the overexpression of GPX4 on stemness characteristics and chemosensitivity in Panc-1 CSCs. After stable transfection of GPX4, its mRNA and protein levels were obviously overexpressed (Fig. 5A). Expectedly, after oxidative stress exposure, overexpression of GPX4 inhibited the accumulation of ROS after H₂O₂ or erastin treatment (Fig. 5B). By performing serial-replating assay with the existence of 100 µM H₂O₂, it was found that overexpression of GPX4 promoted sphere formation at passage 3 and 4, indicating that the self-renewal capacity of the Panc-1 CSCs was promoted by overexpression of GPX4 (Fig. 5C). Followed by detection of physiological processes under oxidative stress (100 µM H₂O₂), it was revealed that overexpression of GPX4 promoted cell proliferation (Fig. 5D), migration (Fig. 5E) and invasion (Fig. 5F). In order to ascertain whether overexpression of GPX4 inhibits oxidative stress-induced cell death in Panc-1 CSCs, the cells pretreated with apoptotic death inhibitor (Z-VAD-FMK), ferroptotic death inhibitor (Ferr-1) or necrotic death inhibitor (Nec-1) were treated with oxidative stress (100 µM H₂O₂ or 2 µM erastin) and stained using CFSE/PI followed by flow cytometric analysis. As shown in Fig. 5G, overexpression of GPX4 inhibited apoptotic cell death induced by H₂O₂, ferroptotic and necrotic cell death induced by erastin. All these data showed that overexpression of GPX4 exerts protective effects under oxidative stress.

ROS level is reported to be directly and tightly related to EMT-programme activation (35). This promoted us to ascertain whether H₂O₂ or erastin-induced ROS accumulation regulates EMT-programme processes. Firstly, we tested the ROS eliminating effect of N-acetyl cysteine (NAC), a ROS scavenger, after H₂O₂ or erastin induction. As expected, NAC efficiently eliminated ROS after oxidative stress (Fig. 6A), thus, in further experiments, NAC was employed as a ROS scavenger. RT-qPCR results for detection of mRNA levels of EMT hallmark genes, including E-cadherin, vimentin, Slug and Snail showed that both H₂O₂ and erastin treatment upregulated E-cadherin and downregulated vimentin, Slug and Snail mRNA levels (Fig. 6B), while elimination of ROS by a scavenger inhibited these changes, indicating that the generated ROS are attributed to the regulation of EMT. Consistently, protein levels of E-cadherin, vimentin, Slug and Snail were consistent with the tendency of mRNA levels (Fig. 6C).

By considering that the difference between CSCs and non-CSCs is likely to be attributable largely to the cell biological programme termed EMT (37,38), the effects of modified GPX4 on EMT hallmark genes were further detected by RT-qPCR and western blot analysis. Both knockdown and overexpression of GPX4 upregulated E-cadherin, downregulated vimentin, Slug and Snail at the mRNA and protein levels, indicating its inhibitory effect on the EMT programme (Fig. 6D and E), which is consistent with the morphologic changes.