Interferon‐alpha modulates the chemosensitivity of CD133‐expressing pancreatic cancer cells to gemcitabine

Abstract

Pancreatic cancer is a lethal disease as current chemotherapies with gemcitabine (GEM) are still insufficient. Accumulating evidence suggests that cancer stem cells (CSC) are responsible for chemoresistance and that CD133 is one of the CSC markers in pancreatic cancer. Interferon‐alpha (IFN‐α), a cytokine with pleiotropic effects, has direct cytotoxic and cytostatic effects on tumor cells. The aim of the present study was to investigate whether IFN‐α can modulate the chemosensitivity of a human pancreatic cancer cell line, Capan‐1, to GEM. Cell cycles were evaluated for response to GEM with and without IFN‐α by BrdU assay. GEM inhibited Capan‐1 cell growth in a dose‐dependent manner. GEM (IC ₅₀; 100 ng/mL) treatment reduced the number of both CD133⁺ and CD133⁻ cells in the S phase, induced apoptosis of CD133⁻ cells more than that of CD133⁺ cells and increased accumulation of CD133⁺ cells into the G0/G1 phase. These results infer that CD133⁺ cells take shelter into the G0/G1 phase from GEM treatment. IFN‐α modulated CD133⁺ cells from the G0/G1 phase to the S phase. Consequently, apoptosis was accelerated in both CD133⁺ and CD133⁻ cells after IFN‐α combined with GEM treatment. Furthermore, GEM combined with IFN‐α treatment showed a significant tumor suppressive effect in the in vivo study. Importantly, CD133⁺ cells showed CSC‐like properties, such as generation of spheres, highly invasive ability and high tumorigenesis. These results suggest that IFN‐α, as a modulator, could contribute to the treatment of CD133⁺ cancer cells and be effective in combined chemotherapies with GEM for pancreatic cancer stem‐like cells. (Cancer Sci 2012; 103: 889–896)

Pancreatic cancer is a lethal disease with a 5‐year survival rate of 5%,1 and recurs despite the use of current chemotherapies. Over the past decade, accumulating evidence has led to the development of the cancer stem cell (CSC) hypothesis for solid tumors.2, 3 This hypothesis might explain the poor prognosis of pancreatic cancer patients because a few CSC can sustain tumor growth and drive relapse after curative treatments. Importantly, CSC contribute to drug resistance.4, 5

CD133 has been implicated as a CSC marker in some cancers.6, 7, 8, 9, 10, 11, 12, 13, 14 In contrast, whether CD133 is a marker of CSC or progenitor cells has been a matter of debate.15, 16 We previously reported that CD133 expression is an unfavorable factor for survival of pancreatic cancer patients.17 Furthermore, CD133‐expressing CSC are essential for the development and perpetuation of pancreatic cancer,18 and may account for resistance to current and standard chemotherapy drugs, such as gemcitabine (GEM).19 However, whether CD133 expression is involved in anti‐cancer drug resistance is unknown.

Interferon‐alpha (IFN‐α), a cytokine with pleiotropic effects, possesses direct cytotoxic and cytostatic effects on tumor cells as well as antiangiogenic effects, and also activates antitumor immunity.20, 21, 22, 23 IFN‐α attracted our interest because of its contribution to the reduction of the side population (SP) enriched in stem cells,24 which was defined by the poor accumulation of Hoechst 33342 in ovarian cancer cells.25 Experimentally, IFN‐α at the optimal biological dose schedule, and in combination with GEM, has been shown to induce apoptosis in tumor‐associated endothelial cells and to decrease the growth of human pancreatic cancer cells.26 However, effects of IFN‐α on CSC in pancreatic cancer have not yet been validated.

Here, we investigate the chemosensitivity of pancreatic cancer cells in the presence of CD133 expression and whether IFN‐α can modulate GEM resistance.

Materials and Methods

Cell culture and reagents

A human pancreatic cancer cell line, Capan‐1, was obtained from the American Type Culture Collection (Manassas, VO, USA). Cells were cultured in DMEM/F12 medium (Gibco, Carlsbad, CA, USA) containing 10% FBS, 100 U/mL penicillin and 100 mg/mL streptomycin at 37°C in 5% CO₂. GEM was supplied by Eli Lilly Japan (Tokyo, Japan), and recombinant human IFN‐α was purchased from Acris Antibodies GmbH (Herford, Germany). GEM and IFN‐α were diluted in culture medium immediately before use.

Flow cytometric analysis and fluorescence‐activated cell sorting

In the present study, 10⁶ single cells were suspended into 100 μL PBS containing 0.5% BSA. A total of 20 μL of FcR blocking reagent and antibodies, carboxyfluorescein conjugated mouse anti‐human IFN‐α/β receptor 1 (R&D Systems, Minneapolis, MN, USA) or R‐phycoerythrin (PE)‐conjugated mouse monoclonal anti‐human IFN‐α/β receptor 2 (PBL InterferonSource, Piscataway, NJ, USA) were added at the appropriate dilutions and incubated on ice for 10–20 min in the dark. Cells were washed and resuspended in a suitable amount of buffer for analysis by flow cytometry. Flow cytometric analysis and FACS were carried out with a FACSAria (Becton Dickinson, Franklin Lakes, NJ, USA) and FACSDiva (BD Biosciences, San Jose, CA, USA). Dead cells were excluded by 7‐amino‐actinomycin‐D (7‐AAD; BD Pharmingen, San Diego, CA, USA) staining. Anti‐CD133‐PE (Miltenyi Biotec, Cologne, Germany) were used for sorting CD133⁺ and CD133⁻ subpopulations of Capan‐1 cells. FACSAria sorting routinely achieved purities exceeding 95% in cell fractions.

Sphere forming assay

Spheres were cultured in DMEM/F12 serum‐free medium supplemented with epidermal growth factor (20 ng/mL; Cell Signaling Tech, Danvers, MA, USA), basic fibroblast growth factor (10 ng/mL; PeproTech, London, UK) and B27 supplements (1:50; Gibco). Capan‐1 cells were cultured for 7 days. Immunofluorescence staining was performed on spheres in culture to identify CD133 expression. Cells were fixed with 10% formalin prior to immunostaining, then incubated with anti‐CD133 mAb for 2 h at 37°C and subsequently washed three times with PBS containing 2% FBS. Next, the cells were incubated for 1 h with PE‐conjugated anti‐mouse secondary antibodies. Nuclear was stained by DAPI. Finally, the stained cells were viewed under a confocal laser scanning microscope (Olympus, Tokyo, Japan). The negative control groups contained cells stained only with the secondary antibody.

Migration and invasion assays

In total, 5 × 10⁴ cells were seeded in serum‐free medium into 24‐well Falcon migration inserts (8 μm pore size). Inserts were placed in Falcon companion plates containing 10% FBS and incubated for 18 h for migration. For invasion, 5 × 10⁴ cells were seeded in serum‐free medium into a Matrigel invasion chamber (Becton Dickinson) for 22 h. Following incubation, the medium and cells were removed from the top chamber using cotton swabs and PBS. The number of migrating or invading cells on the underside of the membrane was determined by staining using Giemsa for 5 min. The number of migrating or invading cells in 10 fields was counted at 20× magnification using light microscopy.

Cytotoxicity assays

Capan‐1 cells were resuspended in fresh medium at a concentration of 5 × 10³ cells/100 μL and seeded in a 96‐well plate, and cells were incubated for 48 h at 37°C. Then, GEM was added to each well at concentrations of 1, 5, 10, 50, 100, 500 and 1000 ng/mL to test the GEM treatment, or IFN‐α was added to each well at concentrations of 500, 1000, 2500, 5000, 10 000, 25 000, 50 000 and 100 000 U/mL to test the IFN‐α treatment. The plate was incubated at 37°C for another 48 h. For the assay, 10 μL of 3‐(4,5‐dimethylthiazol‐2‐yl)‐2,5‐diphenyltetrazolium bromide (MTT; 5 mg/mL) was added to each well, and the plate was incubated for an additional 3 h at 37°C. The medium and MTT solution were then aspirated, and 70 μL of DMSO (Sigma‐Aldrich, St. Louis, MO, USA) was added. The absorbance was measured at 570 nm using a microplate reader.

BrdU assay

Cells (80–90% confluent) were incubated with 1 mM BrdU for 3 h at 37°C and processed using the fluorescein isothiocyanate BrdU Flow Kit (BD Biosciences) according to the manufacturer's instructions. Briefly, 1 × 10⁶ trypsinized cells were fixed, permeabilized and digested with DNase. Cells were then stained with fluorescein isothiocyanate–conjugated anti‐BrdU and 7‐AAD. For cell isolation and characterization, anti‐CD133/1‐APC (Miltenyi Biotec) was used. For each experiment, 10⁴ events were counted by flow cytometry and assays were performed in triplicate. Data were analyzed using FACSDiva.

Western blotting

Cells were lysed on ice in lysis buffer and the lysates were boiled for 5 min, clarified by centrifugation at 15 000 g for 15 min, and then separated by SDS‐PAGE. The proteins were transferred onto nitrocellulose membranes. The membranes were then incubated with a 1:200 dilution of anti‐CD133 mAb (Miltenyi Biotec) followed by a 1:200–1000 dilution of peroxidase‐conjugated anti‐mouse IgG (Jackson ImmunoResearch, West Grove, PA, USA) antibody for the secondary reaction. As an internal control for the amount of protein loaded, β‐actin was detected. The immunocomplex was visualized using the ECL western blot detection system (GE Healthcare Life Science, Amersham, UK).

Mitogen‐activated protein kinase‐integrating kinase 1 inhibition experiments

Capan‐1 cells were pretreated with 20 μM CGP57380 (mitogen‐activated protein kinase‐integrating kinase [Mnk1] inhibitor; Sigma‐Aldrich) for 60 min, and subsequently cultured with 10 μM CGP57380 during the treatment with IFN‐α. Proteins were then detected with antibodies against the phosphorylated Mnk1 (Thr‐197/202, 1:250 dilution; Cell Signaling Tech) and Mnk1 (1:250 dilution); 10% FCS treated cells were used as a positive control for Mnk1 phosphorylation. The signals for pMnk1 were quantified by densitometry using Image J software. To assess the effects of Mnk inhibition on cell viability, Capan‐1 cells were treated with IFN‐α, gemcitabine or Mnk1 inhibitor CGP57380 combination. After 48 h incubation, cell viability was determined by MTT assay.

Animal studies

The animal study was approved by the Committee on the Use of Live Animals for Teaching and Research of Kagoshima University. Non‐obese diabetic (NOD)/SCID and BALB/c nu/nu (nude) mice were purchased from CLEA Japan (Tokyo, Japan).

Tumorigenic assay

For the tumorigenic assay, CD133⁺ or CD133⁻ populations of Capan‐1 cells were collected using FACSAria and PE‐conjugated anti‐CD133 antibody. Freshly isolated CD133⁺ and CD133⁻ cells were subjected to tumorigenic assay. In total, 10, 100 and 1000 cells of each quadrant suspended in 50 μL of DMEM F‐12 medium and 50 μL Matrigel were injected s.c. into 6‐week‐old NOD/SCID mice. Animals were maintained until death resulting from the neoplastic process or the end of the experiment. Xenograft tumors were fixed with 10% buffered formaldehyde and stained with H&E.

In vivo chemotherapies for xenograft tumors

For in vivo treatments, nude mice were randomly assigned to four treatment groups of five mice at week 2 after s.c. injection of Capan‐1 cells (5 × 10⁵). Mice received treatments of vehicle (saline, i.p.), GEM (120 mg/kg/week, i.p.) alone, IFN‐α (20 000 U/mouse/every 2 days, s.c.) alone, or GEM combined with IFN‐α for 4 weeks. Growth curves of xenograft tumors in nude mice were assessed after treatments.

Statistical analysis

Group differences were analyzed statistically using the χ²‐test and Student's t‐test. A P‐value <0.05 was considered statistically significant. All statistical analyses were performed using StatView statistical software version 5.0 (SAS Institute, Cary, NC, USA).

Results

CD133⁺ Capan‐1 cells are resistant to gemcitabine treatment

CD133 expression was examined by flow cytometric analysis in several human pancreatic cancer cell lines. The positive ratio of CD133 in Capan‐1cells is approximately 45%, higher than that in the other cell lines (Table S1), Capan‐1 was chosen to evaluate the sensitivity to GEM. GEM inhibited Capan‐1 cell growth in a dose‐dependent manner and its IC_50, as assessed by MTT growth inhibitory assay, was 100 ng/mL (Fig. 1A). The growth inhibition by GEM treatment (100 ng/mL) showed a significant (P < 0.01) difference between CD133⁺ and CD133⁻ populations of Capan‐1 cells (Fig. 1B). GEM treatment increased the proportion of CD133⁺ Capan‐1 cells (Fig. 1C). Similarly, CD133 protein levels increased in a time‐dependent manner (Fig. 1D). GEM treatment produced a significant increase in the G0/G1 phase and a decrease in the S phase cell populations (Fig. S1). We compared cell cycles between CD133⁺ and CD133⁻ populations of Capan‐1 cells by BrdU assay after GEM treatment (Fig. 2A). The proportion of CD133⁺ cells in the G0/G1 phase increased from 56.4 to 92.5%, and was maintained at 93.2% even after withdrawal of GEM (Fig. 2B). However, there were no significant changes in CD133⁺ cells. Although the proportion of CD133⁺ and CD133⁻ cells in the S phase was remarkably reduced after GEM treatment, the proportion of CD133⁺ cells in the S phase increased compared to that in CD133⁺ cells after withdrawal of GEM (Fig. 2C). The proportion of CD133⁺ cells in the apoptotic phase was significantly lower than that in CD133⁻ cells under control, and apoptotic cells were highly induced in CD133⁺ cells after GEM treatment and withdrawal of GEM (Fig. 2D). These results indicated that CD133⁺ cells were resistant to GEM, compared to CD133⁻ cells.

Figure 1.

CD133⁺ Capan‐1 cells are more resistant to gemcitabine (GEM) treatment than CD133⁻ cells. (A) GEM inhibited Capan‐1 cell growth in a dose‐dependent manner, and its IC ₅₀ was 100 ng/mL. It was measured using the MTT growth inhibitory assay after 24 h of continuous GEM exposure. (B) GEM treatment (100 ng/mL) showed different sensitivities in CD133⁺ and CD133⁻ cells. Error bars indicate SD. (C) GEM treatment for 12 or 24 h increased the proportion of CD133⁺ cells. Error bars indicate SD. *P < 0.01 vs control. (D) Western blot of CD133 expression.

Figure 2.

Comparison of cell cycles between CD133⁺ and CD133⁻ cells. (A) Flow cytometry was used for cell cycle analysis before and after exposure to gemcitabine (GEM). Baseline data are provided in the left panel. Cells were then treated with GEM for 12 h (middle) followed by 24 h of recovery after withdrawal of GEM (right). (B) Comparison of CD133⁺ and CD133⁻ cells in the G0/G1 phase. (C) The S phase and (D) apoptosis by BrdU assay after GEM treatment. Results are based on three independent experiments. Error bars indicate SD. *P < 0.01 vs control.

Interferon‐alpha reduces the CD133⁺ ratio of Capan‐1 cells

All Capan‐1 cells showed expression of IFN‐α/β receptor 2 (Fig. 3A). IFN‐α inhibited Capan‐1 cell growth by up to 30% in a dose‐dependent manner, and concentrations over 5000 U/mL showed similar inhibition rates (Fig. 3B). Importantly, IFN‐α treatment decreased the proportion of CD133⁺ cells in a time‐dependent manner (Fig. S2, Fig. 3C). Similarly, after over 6 h of IFN‐α treatment, CD133 protein levels were decreased (Fig. 3D). To understand the mechanism underlying IFN‐α treatment, Mnk1 expression and inhibition experiment were performed. IFN‐α treatment induced phosphorylation of Mnk1 in a time‐dependent manner (Fig. 3E left and right). Mnk1 inhibitor CGP57380 administration antagonized the IFN‐α effect on cell growth suppression, but not significantly. Mnk1 inhibitor mitigated the antiproliferative response to the co‐administration of IFN‐α and GEM (Fig. 3F).

Figure 3.

Interferon‐alpha (IFN‐α) reduced the proportion of CD133⁺ cells in Capan‐1 cells. (A) Expressions of IFNAR 1 (left) and IFNAR 2 (right). (B) IFN‐α inhibited Capan‐1 cell growth in a dose‐dependent manner for 48 h exposure. (C) IFN‐α (5000 U/mL) treatment for 48 h decreased the proportion of CD133⁺ cells in Capan‐1 cells over time. (D) CD133 protein level analyzed by western blot. (E) IFN‐α‐dependent phosphorylation/activation of Mnk1 in a time‐dependent manner (left and right). (F) Mnk1 mediated the antiproliferative response to the co‐administration of IFN‐α and gemcitabine (GEM). Capan‐1 cells were treated in the combination of IFN‐α, GEM and Mnk1 inhibitor CGP57380. After 48 h incubation, cell viability was determined by MTT assay. These results are the means and SD of values from four wells in one representative experiment.

Interferon‐alpha contributes to combined chemotherapy with gemcitabine

We compared cell cycles between the CD133⁺ and CD133⁻ populations of Capan‐1 cells by BrdU assay after GEM alone, IFN‐α alone or GEM combined with IFN‐α treatment (Fig. S3). GEM treatment increased the ratio of cells in the G0/G1 phase in the CD133⁺ population, while IFN‐α decreased the proportion of cells in the G0/G1 phase (Fig. 4A upper). In the CD133⁻ cells, however, the G0/G1 phases were similar among these treatments (Fig. 4A lower). IFN‐α, but not GEM treatment, remarkably increased the proportion of cells in the S phase in both CD133⁺ and CD133⁻ cells. Furthermore, GEM combined with IFN‐α treatment significantly increased the apoptotic phases in both CD133⁺ and CD133⁻ cells (Fig. 4A). These results suggest that IFN‐α modulates the cell cycle of CD133⁺ Capan‐1 cells (Fig. 4B).

Figure 4.

Interferon‐alpha (IFN‐α) contributes to combined chemotherapy by reducing the proportion of the G0/G1 phase cells and increasing the proportion of the S phase and apoptotic cells. (A) Comparison of G0/G1, S and apoptotic cells by BrdU assay between CD133⁺ and CD133⁻ cells treated with gemcitabine (GEM) alone, IFN‐α alone or GEM + IFN‐α for 24 h. In CD133⁺ (upper) and CD133⁻ (lower) populations, * vs **, * vs # and # vs ## indicate P < 0.01. Error bars indicate SD. (B) Model of IFN‐α modulating CD133⁺ cells from G0/G1 to the S phase and targeting them combined with GEM.

Effect of interferon‐alpha on xenograft tumors of CD133⁺ cells

We attempted to determine the in vivo effect of IFN‐α on xenograft tumors derived from Capan‐1 cells in nude mice. Four weeks' treatment of GEM combined with IFN‐α suppressed tumor growth in nude mice (Fig. 5A) and led to significant differences in tumor growth curves compared to the control, GEM alone or IFN‐α alone (Fig. 5B). However, body weight did not change significantly (Fig. S4A). In the immunohistological study, xenograft tumor cells treated with GEM showed higher CD133⁺ expression than those of the control. In contrast, CD133⁺ expression in xenograft tumors treated with IFN‐α alone was lower than those with saline as control. Interestingly, xenograft tumors treated with GEM combined with IFN‐α showed an intermediate CD133⁺ expression (Fig. 5C). The flow cytometric analysis showed similar results to the immunohistological study (Fig. S4B).

Figure 5.

Effect of interferon‐alpha (IFN‐α) on the growth of xenograft Capan‐1 tumor in nude mice. (A) Xenograft tumors with IFN‐α + gemcitabine (GEM) treatment were smaller than those with controls in nude mice. (B) Tumor growth curves of Capan‐1 xenografts which were treated with GEM (100 ng/mL) alone, IFN‐α (5000 U/mL) alone or GEM plus IFN‐α. IFN‐α + GEM treatment showed a significant effect (P < 0.01). (C) Comparison of histological CD133 expression in Capan‐1 xenografts with treatments at week 5 after inoculation into nude mice.

CD133⁺ Capan‐1 cells identified as a cancer stem‐like population of cells

Flow cytometric analysis was performed on several human pancreatic cancer cell lines. Among these, Capan‐1 showed high expression of CD133 (Table S1, Fig. 6A). Capan‐1 cells showed sphere formations in a stem cell‐permissive medium without serum (Fig. S5A), and CD133 was expressed on the cell surface or in the cytoplasm in these sphere cells (Fig. 6B). These CD133⁺ Capan‐1 cells showed a higher potential of migration and invasion than CD133⁻ cells (Fig. 6C). Furthermore, CD133⁺ cells showed significantly greater tumorigenic potential than CD133⁻ cells (Table 1). In these tumors, the histology was ductal adenocarcinoma and CK expression was observed in all tumor cells, while CD133 expression was shown in a part of the tumor cells by immunohistological staining (Fig. S5B). These results indicate that the CD133⁺ population of Capan‐1 cells exhibits CSC‐like properties.

Figure 6.

Cancer stem‐like characteristics were identified in CD133⁺ Capan‐1 cells. (A) Capan‐1 cells showed monolayer growth in medium with serum (left). In serum‐free culture, a sphere was generated after 10 days of culture (right). Magnification: ×200. (B) These sphere cells expressed CD133 by immunofluorescence staining. Magnification: ×200. (C) Comparison of migration and invasion abilities between CD133⁺ and CD133⁻ Capan‐1 cells.

Table 1.

Comparison of tumorigenesis between CD133⁺ and CD133⁻ population cells in Capan‐1 using non‐obese diabetic/SCID mice

Subset of Capan‐1	Number of implanted cells			Total
	10	102	103
CD133+	1/10 (10%)	7/10** (70%)	8/10** (80%)	16/30*** (53%)
CD133−	0/10	0/10	1/10 (10%)	1/30 (3%)

***P < 0.001; **P < 0.05.

Discussion

Gemcitabine had greater inhibitory effects on the human pancreatic cancer cells, Capan‐1. However, CD133⁺ cells showed more resistance to GEM than CD133⁻ cells, although the growth speed between CD133⁺ and CD133⁻ cells was the same in Capan‐1. Along with GEM treatment, the ratio of CD133⁺ cells in Capan‐1 increased and the resistance against GEM was more drastic.

Interferon‐alpha modulated the cell cycle, resulting in antiproliferative and proapoptotic effects on CD133⁺ cells using combined therapy with GEM. Members of the IFN family are pleiotropic cytokines that have been shown to be important regulators of cell growth. IFN‐α has been recognized to have therapeutic potential for the prevention and treatment of hepatocellular carcinoma.27, 28 Whether pancreatic cancer cells respond to IFN treatment is unknown, although clinical trials including combination therapy with IFN‐α for advanced pancreatic cancer patients have had promising results.29, 30

Type I IFN signaling is mediated by activation of the JAK‐STAT signaling pathway.31 In our study, the increase of the number of cells in the S phase indicates that proliferation of CD133⁺ cells was mediated by IFN‐α treatment. The accumulation of CD133⁺ cells into the G0/G1 phase was remarkably increased after GEM treatment. GEM is a nucleoside analog that can replace one of the building blocks of the nucleic acid during DNA replication, leading to suppression tumor growth. Another target of GEM is to inactivate the enzyme ribonucleotide reductase. GEM shows specificity for proliferation in the S phase of the cell cycle with no effect on progress through early G1, G2 or M phases of the cell cycle.32 However, IFN‐α contributed to the effect on the decrease of CD133⁺ cells in the G0/G1 phase and the increase of them in the S phase. IFN‐α priming provides an efficient way to induce cell cycle entry of dormant cells, such as hematopoietic stem cells.33 IFN‐α makes dormant cells susceptible to elimination by anti‐proliferative chemotherapeutic drugs,34 such as CD133⁺ cells, as shown in Figure 4B. According to a recent report, the Mnk/elF4E kinase pathway is activated in an IFN‐inducible manner and plays important roles in mRNA translation for IFN‐stimulated genes and in the generation of IFN‐inducible antiproliferative responses.35 In our study, IFN‐α treatment induced rapid phosphorylation of Mnk1 that was detectable within 15 min of treatment. Mnk1 inhibitor may mitigate the antiproliferative response to the co‐administration of IFN‐α and GEM. Further clarification of tumor suppression by IFN‐α is necessary.

Numerous studies have identified a “side population” (SP) in various tumor types,36, 37, 38, 39 and SP cells seem to be rich in stem cells.24 These malignant SP cells proliferate in a sustained fashion and readily export many cytotoxic drugs. This high drug efflux capacity correlates with the strong expression of ATP‐binding cassette transporters.36 Interestingly, ovarian cancer containing SP cells have been found to be IFN‐α sensitive in vitro and in vivo due to marked anti‐proliferative and pro‐apoptotic effects.20 In this study, however, the CD133⁺ population of Capan‐1 cells did not coincide with the SP population (data not shown). IFN‐α increased the number of CD133⁺ cells in the S phase compared to that of CD133⁻ cells. Furthermore, IFN‐α combined with GEM induced apoptosis in both CD133⁺ and CD133⁻ cells to a greater extent than GEM or IFN‐α treatment alone. IFN‐α has also been shown to induce differentiation of lung cancer cells 37 and hepatic progenitors.40 In addition, IFN‐α has been shown to regulate the transition from SP into other phenotypes, although this IFN signaling‐related mechanism is unclear.25

In our study, the combination of IFN‐α and GEM significantly inhibited the growth of xenografts of Capan‐1 cells compared to the control, GEM or IFN‐α alone. These results were consistent with the in vitro data. However, using in vivo orthotopic pancreas cancer models, the combination of IFN‐α and GEM has been reported to synergistically induce endothelial cell apoptosis.26 These results suggest that IFN‐α may have multiple biological functions in the modulation of gene expression and regulation of the cell cycle in terms of tumor suppression in vivo.

In contrast, CD133⁺ population of Capan‐1 cells exhibited greater tumorigenesis and the potential to generate spheres and aggressive behavior, such as migration and invasion, compared with CD133⁻ cells. These results suggest that CD133 plays an important role in the cancer stem‐like population of Capan‐1 cells. Hence, the underlying mechanism of the CSC regulation is an important issue. In a recent study, the combined blockade of sonic hedgehog and mTOR signaling together with GEM treatment led to a profound depletion of the CSC compartment and shrinkage of established tumors.41 Our results shed new light on the impact of IFN‐α on the cell cycle of a CSC‐like population in pancreatic cancer cells, although further research into the mechanism of the CSC modulation by IFN‐α is still needed.

In the present study, we demonstrated that GEM could efficiently act on S phase cells in both CD133⁺ and CD133⁻. CD133⁺ cells could escape from GEM treatment by retention in the G0/G1 phase. IFN‐α administration prompted G0/G1 phase CD133⁺ cells to re‐enter the cell cycle. Thus, IFN‐α treatment could increase GEM therapeutic efficacy. Moreover, GEM combined therapy with IFN‐α significantly suppressed xenograft tumor growth. In addition, CD133⁺ cells showed CSC‐like properties, such as generation of spheres in serum‐free culture and tumorigenesis in NOD/SCID mice. Taken together, IFN‐α, as a modulator, could contribute to the treatment of CD133⁺ cancer cells with CSC‐like properties and be effective in combined chemotherapies for pancreatic cancer stem‐like cells.

Disclosure Statement

The authors have no conflict of interest to declare.

Supporting information

Fig. S1. Flow cytometric analysis of cell cycle progression in Capan‐1 cells treated with gemcitabine (100 ng/mL) for 12 or 24 h.

Fig. S2. Flow cytometric analysis of CD133 expression in Capan‐1 cells treated with or without interferon‐alpha (IFN‐α). IFN‐α treatment (5000 IU/mL) for 24 h decreased the ratio of CD133⁺ Capan‐1 cells over time.

Fig. S3. Comparison of BrdU assay between CD133⁺ and CD133⁻ Capan‐1 cells treated with gemcitabine (GEM) (100 ng/mL) alone, interferon‐alpha (IFN‐α) (5000 U/mL) alone or GEM combined with IFN‐α for 24 h.

Fig. S4. (A) Body weight curves of nude mice were not significantly different among the three treatments, which were gemcitabine (GEM) (100 ng/mL) alone, interferon‐alpha (IFN‐α) (5000 U/mL) alone or GEM plus IFN‐α. (B) Comparison of proportions of CD133⁺ cells in Capan‐1 xenografts that were treated with GEM (100 ng/mL) alone, IFN‐α (5000 U/mL) alone or GEM plus IFN‐α at week 2 and 5.

Fig. S5. (A) Comparison of the number of spheres per well (cm³) for pancreatic cancer cell lines, Panc‐1, Capan‐1, MIA PaCa‐2, PK45H and SW1990. The white and black bars indicate the spheres composed of 3–30 and >30 cells, respectively. (B) Immunohistochemical study of a Capan‐1 tumor generated by transplantation into non‐obese diabetic (NOD)/SCID mice. (i) HE staining and (ii, iii, iv) immunostaining performed to identify cytokeratin (CK) and CD133 expression, respectively.

Table S1. Comparison of potential markers related to tumor‐initiating cells in six pancreatic cancer cell lines.