Hypoxia regulates ABCG₂ activity through the activivation of ERK1/2/HIF-1α and contributes to chemoresistance in pancreatic cancer cells

ABSTRACT

Pancreatic cancer is a drug resistant hypovascular tumor. Although there are many studies on the mechanism of chemoresistance in pancreatic cancers, studies on the relationship between ABCG2 and chemoresistance during hypoxia of pancreatic cancer are rare. Hypoxia-inducible factor-1 (HIF-1α) is a master regulator of the transcriptional response to oxygen deprivation in cancer cells. The aim of this study was to examine the role of ABCG₂ and HIF-1α in mediating chemoresistance during hypoxia in pancreatic cancer. In this study, we detected the expression levels of ABCG₂, ERK/phosphorylated-ERK (p-ERK) and HIF-1α by immunohistochemistry in fresh pancreatic cancer and paracarcinoma tissues obtained from 25 patients. The mechanism by which p-ERK1/2 and HIF-1α affect ABCG₂s expression was analyzed in the hypoxic cultured human pancreatic cancer cell line Capan-2. ABCG2-mediatedregulation of gemcitabine response under hypoxic conditions in pancreatic cancer cells was observed. It was found that ABCG₂, ERK/p-ERK and HIF-1α were overexpressed in cancer tissues. ABCG2, HIF-1α and p-ERK levels were demonstrated to be high during hypoxic conditions in pancreatic cancer cells. Hypoxia induced phosphorylation of ERK1/2 to activate HIF-1α and contribute the ABCG₂ expression and mediated gemcitabine chemoresistance in pancreatic cancer cells. Hypoxic conditions induced HIF-1α binding to target gene sequences in the ABCG₂ promoter, resulting in increased transcription in pancreatic cancer cells. We demonstrated that hypoxia-induced chemoresistance is due to the regulation of ABCG₂ through the activation of ERK1/2/HIF-1α. ABCG₂ could serve as a predictor of gemcitabine response and, potentially, as a chemotherapeutic target in pancreatic cancer. Inhibition of ERK1/2 and HIF-1αcould result in increased gemcitabine sensitization in pancreatic cancer with highly expressed ABCG₂ cell member protein.

Introduction

Pancreatic cancer is one of the most malignant forms of cancer. It is the fourth largest cause of cancer-associated deaths across the world, and the 5-year relative survival rate for pancreatic cancer is 5% to 6%.^1,2 Pancreatic cancer is characterized by rapid disease progression, a high metastatic property and difficult early diagnosis.^3,4 It is refractory to treatment, and the recurrence rate is high after pancreatectomy. Chemotherapy is an important therapeutic method, but the sensitivity is low because of growing drug resistance. Therefore, a better understanding of the mechanisms involved in tumor progression is needed to develop a new effective target for the treatment of this fatal disease.

The tumor microenvironment plays a critical role in tumor progression and is associated with therapeutic effects in cancer treatment.⁵ Hypoxia is one of the key features of the tumor microenvironment that contributes to cancer progression and mediates resistance to chemotherapy.

Under hypoxic conditions, tumor cells display a series of adaptive responses that allow for survival and continued proliferation. HIF-1α is a master regulator of the transcriptional response to oxygen deprivation in cancer cells and up-regulates a series of genes that support the tumor cell to compensate for the hypoxic microenvironment.⁶ HIF-1α overexpression has been detected in several solid tumors and is correlated with progression of a variety of cancers, including ovarian carcinoma, cervical carcinoma, oropharyngeal squamous cell carcinoma, non-small-cell lung cancer and pancreatic cancer.^7-11 It has been demonstrated that HIF-1α affects the regulation of cancer cell proliferation, angiogenesis, apoptosis and chemotherapeutic resistance during tumor development.¹² Although HIF-1α has been reported to promote cell survival and may be associated with drug resistance in pancreatic cancer,^11,13 the underlying mechanism still remains to be elucidated. Nuclear accumulation is an important process of HIF-1α activity. Hypoxia has been shown to activate ERK1/2, which subsequently stimulates HIF-1α transcriptional activity in some tumors.¹⁴ ERK1/2 activity was also found to be important for tumor development and drug resistance in pancreatic cancer.¹⁵

Clinical drug resistance remains a significant impediment to the successful treatment of cancer. The ATP-binding cassette subfamily G member 2 (ABCG₂) is the most recently described among the major multidrug-resistance pumps. ABCG₂ mRNA encodes a 655-amino acid, 72 kDa protein, with a single nucleotide binding domain (NBD) and 6 transmembrane domains (TMD). Being a half-transporter, it requires at least 2 NBDs to function as a drug efflux pump. Depending on the system, ABCG₂ expression may be controlled at the promoter level by sex hormones, hypoxia and methylation status.^16,17 Several studies have reported that ABCG₂ expression is related to tumor progression and drug resistance in a variety of malignant tumors including leukemia, myeloma, lung cancer and breast cancer.¹⁸ ABCG₂ expression was also reported to be correlated to drug resistance in a number of pancreatic cancer cell lines.¹⁹

Therefore, ABCG₂ regulates drug response in pancreatic cancer, and further studies are needed to accurately characterize the mechanisms modulating ABCG₂ expression and function. Krishnamurthy et al. were the first to demonstrate that hypoxia regulates ABCG₂ expression.²⁰ Under hypoxic conditions, binding of HIF-1α to a hypoxia response element (HRE) in the ABCG₂ promoter results in increased ABCG₂ transcription.²¹ Therefore, we hypothesized that, in pancreatic cancer, HIF-1α accumulation modulates ABCG₂-induced tumorigenesis and drug resistance. Using histochemical methods, we tested the expression of ABCG₂, ERK1/2 and HIF-1α in human pancreatic cancer tissues and analyzed ERK1/2/HIF-1α/ABCG₂ activity in Capan-2, the hypoxia cultured human pancreatic cancer cell line. Our results demonstrated that ABCG₂ and HIF-1α contribute to the tumorigenesis of human pancreatic cancer and has the potential to serve as a marker for tumor malignancy progression. We also observed that activing ERK1/2/HIF-1α activity increases ABCG₂ expression and function in Capan-2 cells and enhances their drug resistance under hypoxic conditions. These results suggest that the targeting of ERK1/2/HIF-1α and ABCG₂ may be a potentially useful strategy for improving pancreatic cancer therapeutics.

Results

Overexpression of HIF-1α, ABCG₂ and p-ERK/ERK proteins in pancreatic cancer tissues

To determine whether protein expression plays a vital role in the progression of pancreatic cancer, we performed immunohistochemistry analysis to determine HIF-1α, ABCG₂, ERK1/2 and p-ERK1/2 protein levels in pancreatic carcinoma tissues samples. We observed higher levels of HIF-1α, ABCG₂, ERK1/2 and p-ERK1/2 in pancreatic cancer tissues compared with paracarcinoma tissues (Fig. 1A). The high expression of HIF-1α, ABCG₂, ERK1/2 and p-ERK1/2 was confirmed by western blotting (Fig. 1B). In the 25 samples tested, cancer tissue samples were characterized by positive expression of HIF-1α and ABCG₂ (scores>0), while paracarcinoma tissues were characterized by negative expression (score=0 ) (Fig. 1C). The average score for HIF-1α and ABCG₂ was approximately fold4- and fold5-, respectively, in cancer tissue relative to the paracarcinoma tissue (Fig. 1D). These results indicated that HIF-1α, ERK1/2, p-ERK1/2 and ABCG₂ are important for the development of pancreatic cancer.

Figure 1.

The expression of HIF-1α, ERK/p-ERK and ABCG₂ in pancreatic cancer and paracarcinoma tissues. (A) Representative data showing immunohistochemical staining of HIF-1α, ERK/p-ERK and ABCG₂ proteins in human pancreatic carcinoma and paracarcinoma tissues. Brown staining corresponds to proteins. (B) Western blotting analysis of HIF-1α, ERK/p-ERK and ABCG₂ in human pancreatic carcinoma and paracarcinoma tissues. The protein expression levels were normalized to those of β-actin (n = 3 ). (C)The relative rate of positive and negative score for HIF-1α and ABCG₂ expression in pancreatic cancer and paracarcinoma tissues (n = 25 ). (D) Scatter plot to determine the score of HIF-1α and ABCG₂ proteins expressed in pancreatic cancer and paracarcinoma tissues (n=25 ). * p < 0.05.

Hypoxia induced by CoCl₂ regulates the expression of p-ERK, HIF-1α and ABCG2 protein

Exponentially growing Capan-2 cells were exposed to various concentrations of CoCl₂ for 24 hours, and the cytotoxic activity of CoCl₂ was determined by CCK8 analysis to calculate the IC50 value (Fig. 2A). In addition, we performed western blotting to determine the CoCl₂-induced hypoxic effects on p-ERK, HIF-1α and ABCG₂ expression levels. Capan-2 cells were incubated with 400μM CoCl₂ for 0, 6, 12, 24, 48 and 72 h. The expression of HIF-1α started increasing from 6 h and reached peak expression at 24 h after CoCl₂ treatment (Fig. 2B). CoCl₂-induced hypoxia was found to increase p-ERK expression and nuclear translocation (Fig. 2C). The expression of total ABCG₂ increased from 12 h and reached the peak at 48 h after CoCl₂ treatment (Fig. 2D). Moreover, the flow cytometry analysis showed that 400μM CoCl₂-induced hypoxia promoted ABCG₂ accumulation in the cell membrane in a time-dependent manner (Fig. 2E).

Figure 2.

Cellular responses to CoCl₂-induced hypoxia in the pancreatic carcinoma cell line Capan-2. (A) Cytotoxic effects of CoCl₂on Capan-2 cells. The cells were cultured in the presence ofCoCl₂ (0, 100, 300, 500, 1000μM) for 24 h and the survival rate was measured by the CCK8 assay. The IC50 was calculated to be400±57 .5μmol/L. (B) The HIF-1α protein expression was determined by Western blotting analysis. Capan-2 cells were exposed to 400μM CoCl₂ treatment for 6, 12, 24, 48 and 72 h. HIF-1α expression started to increase after 6 h of treatment and reached peak value after 24 h treatment. (C) To analyze the activation and nuclear translocation of ERK1/2, total protein extracts and nuclear extracts from the cells treated with 400μM of CoCl₂ for 30 min, 1 h, 2 h, 4 h and 6 h were subject to Western blotting analysis. Total phospho-ERK 1/2 and ERK1/2 were normalized to β-actin levels, while nuclear p-ERK1/2 was normalized to Histone H3 levels. (D) Western blotting results of ABCG₂ in Capan-2 after the cells were exposed to hypoxic condition (induced by 400μM CoCl₂) for 6, 12, 24, 48 and 72 h. The chart illustrates the variation of ABCG₂ levels normalized to β-actin. (E) Expression of ABCG₂ in the cell membrane after incubation with 400μM CoCl₂ for 12, 24, 48 and 72 h was detected using flow cytometry. The mean value of fluorescence intensity is presented on the right. (n=3, *p < 0.05)

CoCl₂-inducedhypoxia modulatesABCG₂ expression via the modulation of ERK/HIF-1αactivation in Capan-2 cells

To elucidate further mechanistic details, ERK1/2 and HIF-1α were knocked down, and the interference efficiency in Capan-2-siRNA-ERK and Capan-2-siRNA- HIF1α was found to be approximately 75% (Fig. 3A). We tested whether the expression of ABCG₂ induced by hypoxia was related to p-ERK activation. PD98059 (an inhibitor of p-ERK) and si-ERK were used to inhibit the function of ERK, and our result demonstrated that under hypoxic conditions, ERK activation regulates the ABCG₂ expression at mRNA (Fig. 3B, left) and protein levels(Fig. 3B, right). The knockdown or inhibition of ERK1/2 was able to block ABCG₂ cell translocation across the cell membrane (Fig. 3C). To explore the relationship between p-ERK and HIF-1α, we measured total and nuclear HIF-1α protein levels after hypoxia-induced ERK activation. Nuclear accumulation is an important process of HIF-1α activity. We showed that hypoxia-induced ERK phosphorylation resulted in increased expression and nuclear accumulation of HIF-1α  (Fig. 3D). Moreover, p-ERK increased the expression of ABCG₂ via the nuclear accumulation of HIF-1α during hypoxia. The results showed that, under hypoxic conditions, the expression of ABCG₂ reduced after knock-down of HIF-1α or treatment with PD98059. There was no difference in ABCG₂s expression between the PD98059 group and siHIF +PD98059 group, and ABCG₂ was showed decreased after treatment of PD98059 with pc-DNA-HIF-1α in hypoxia, which indicated that hypoxia regulates ABCG₂s expression through ERK / HIF-1α signaling (Fig. 4A). And knockdown of HIF-1α reduced the levels of ABCG₂ in the cell membrane (Fig. 4B). The results suggest that HIF-1α  regulates ABCG₂ expression. Using flow cytometry analysis, we found that overexpression of HIF-1α accelerates the accumulation of ABCG₂ at the cell membrane (Fig. 4D).

Figure 3.

The overexpression of ABCG₂ and HIF-1α and drug resistance under hypoxic condition in Capan-2 through the activation of ERK. (A)ERK1/2 knockdown efficiency by transient transfection with siERK was detected by western blotting (upper) and the knockdown efficiency of HIF-1α by transient transfection with siHIF was detected by Western blotting (lower). (B) RT-PCR and Western blotting were performed to determine mRNA and protein levels of ABCG₂, respectively, after Capan-2 cells were treated with CoCl₂ (400μM, 24 h), CoCl₂ (400μM, 24 h) +siERK, CoCl₂ (400μM, 24 h) +PD98059 (100μM, 2 h). (C) Expression of ABCG₂ in the cell membrane. The results shown are typical in Capan-2 cells after they were treated with CoCl₂ (400μM, 24 h), CoCl₂ (400 μM, 24 h) +siERK, CoCl₂ (400μM, 24 h) +PD98059 (100μM, 2 h). The mean fluorescence intensity is shown. (D) HIF-1α protein expression was determined by Western blotting in 3 treatment groups: CoCl₂ (400μM), CoCl₂ (400μM) +PD98059 (100μM), CoCl₂ (400μM) +siERK. Total HIF-1α was normalized to β-actin and nuclear HIF-1α was normalized to Histone H3. The bar diagram illustrates the western blotting results. (E) The IC50 and cytotoxic effects of gemcitabine on Capan-2 after different treatments. After cells were incubated with CoCl₂ (400μM, 24 h), CoCl₂ (400μM, 24 h) +siERK, CoCl₂ (400μM, 24 h) + PD98059 (100μM, 2 h), 50μM gemcitabine was added. The drug sensitivity was detected using a CCK8 assay. The IC50 was calculated as mean ±SD. (n=3, *p < 0.05).

Figure 4.

The overexpression of ABCG₂ reduced the gemcitabine sensitivity under hypoxic condition by regulating ERK/ HIF-1α molecular functions. (A) The expression of ABCG₂ after the treatments of PD98059, si-HIF, si-HIF+PD98059, pc-DNA-HIF, pc-DNA-HIF+PD98059. 400μM COCl₂ existed in all the groups. (B) The fluorescence intensity of ABCG₂ in Capan-2 treated with CoCl₂ or CoCl₂+si-HIF was analyzed. (C)The IC50 and cytotoxic effects of gemcitabine on Capan-2 treated with CoCl₂ or CoCl₂+si-HIF were evaluated. (D)ABCG₂ fluorescence intensity was analyzed in 3 treatment groups of Capan-2 cells: pcDNA-ctrl and pcDNA-HIF were tested. (E) Relative cytotoxicity of gemcitabine after cells were treated with gemcitabine for 72 h following the exposure to pc-DNA-ctlr and pc-DNA-HIF was analyzed by CCK8 assay. IC50 of gemcitabine on Capan-2 was calculated. (F)Cytotoxic effect of gemcitabine on Capan-2 cells under CoCl₂-hypoxia with si-RNA-ctrl or si-RNA-ABCG_2.(G) Cytotoxic effects of gemcitabine on Capan-2 cells after knocking down ABCG₂. Cells were over-expressed HIF-1α and transftected with si-RNA-ABCG₂ or si-RNA-ctrl were separately cultivated with gemcitabine (0.01, 0.1, 1, 10, 100, 1000μM), and the survival rate was tested by CCK8 assay. Data is presented as mean±SD (n=3, *p < 0.05).

Hypoxia promotes gemcitabine resistance via the regulation of ERK /HIF-1α/ABCG2 function in Capan-2 cells

We performed CCK8 assay to determine the IC50 of gemcitabine in Capan-2 cells. To exclude the possibility that p-ERK affects gemcitabine sensitivity under hypoxia, we performed a CCK8 assay in Capan-2 cells. The results showed that gemcitabine sensitivity in Capan-2 cells was significantly decreased under hypoxia, and that this effect could be inhibited by PD98059 or si-ERK (Fig. 3E). We also analyzed gemcitabine sensitivity and IC50 after HIF-1α knockdown in Capan-2 cells. The results illustrated that siRNA-mediated HIF-1α gene silencing rescues gemcitabine sensitivity and the IC50 compared with siRNA-ctrl-treated Capan-2 cells (Fig. 4C). Overexpression of HIF-1α resulted that gemcitabine sensitivity decreased (Fig. 4E). To detect whether the gemcitabine sensitivity is regulated by ABCG₂ under hypoxia, we knocked down ABCG₂ and tested the CCK8 (Fig. 4F). When the HIF-1α was overexpressed, the gemcitabine sensitivity was reduced with si-RNA-ABCG₂, compared with the treatment of si-RNA-ctrl (Fig. 4G).

Hypoxia increases binding of HIF-1α to the ABCG₂ gene promoter

Having identified that HIF-1α affects the expression of ABCG₂ in a pancreatic cancer cell line, we tested whether HIF-1α binds to the ABCG₂ promoter to regulate its expression during hypoxia. To evaluate whether HIF-1α binds to the ABCG₂ promoter region in Capan-2 (both under normoxia and hypoxia), we performed a CHIP analysis. After precipitation of cell lysates with specific anti-HIF-1α antibody, anti-RNA Polymerase II (positive control) and Normal Human IgG (negative control), the ABCG₂ and GAPDH gene promoters were amplified by PCR using specific primers. The results demonstrated that HIF-1α binds to the ABCG₂ gene promoter in Capan-2 not only during hypoxia but also during normoxic conditions. Moreover, during hypoxia, the affinity of HIF-1α binding to the ABCG₂ gene promoter is higher compared with that during normoxia in Capan-2 (Fig. 5A). To see the relationship of HIF-1α and ABCG₂ intuitively, a part of promoter region of ABCG₂ is shown in Fig. 5B.

Figure 5.

HIF-1α binds to the ABCG₂ gene promoter in Capan-2 cells. (A) Binding of HIF-1αto the ABCG₂gene promoter was detected by CHIP assay. PCR analysis for ABCG₂ gene promoter region was carried out on samplesimmunoprecipitatedwiththeanti-HIF-1α antibody and on purified total input DNA from the Capan-2 cells (hypoxia and normoxia conditions). The anti-RNA Polymerase antibody was used as a positive control, and normal human IgG was used as a negative control. Total input DNA was used as the input control. (B) A schematic plot demonstrating the binding of HIF-1α to the ABCG₂ gene promoter.(n=3, * p < 0.05).

Discussion

Pancreatic cancer is a drug resistant hypovascular tumor. The mechanism underlying hypoxia-induced chemoresistance is important to further improve therapeutic efficacy. Our results demonstrate that, under hypoxic conditions, HIF-1α regulates ABCG2 expression and function to mediate chemoresistance in pancreatic cancer.

In this study, we used immunohistochemistry and Western blotting analysis to demonstrate overexpression of HIF-1α and ABCG₂ proteins in human pancreatic cancer tissues, indicating that the proteins play an important role in the tumorigenesis of pancreatic cancers. Our results suggest that HIF-1α and ABCG₂ serve as a prognostic marker of pancreatic cancer and a marker to distinguish tumor from non-tumor tissues. ABCG₂ has been reported to mediate multidrug resistance, and HIF-1α has been shown to be involved in drug-resistance processes. In our study, further analysis regarding the function of ABCG₂ proteins was performed to elucidate the mechanism underlying mechanism chemoresistance in pancreatic cancer cells under hypoxic conditions.

We used CoCl₂ to induce a hypoxic environment and analyzed the effects on the expression of ABCG₂, HIF-1α and ERK1/2 proteins. We observed that ABCG₂, p-ERK1/2 and HIF-1α protein levels were high under the hypoxic conditions. ABCG₂ was found to be overexpressed in the cell membrane and the cytoplasm. CoCl₂s treatment resulted in increased the activation of p-ERK1/2. Our results demonstrated that all changes in protein expression occurred in a time-dependent manner. It has been established that high levels of ABCG₂ are correlated with poor chemotherapy results for solid tumors.^18,19 Here, we demonstrated that higher levels of ABCG₂ are associated with hypoxia-induced chemoresistance in pancreatic cancer. Because gemcitabine is used as a first-line drug in the treatment of pancreatic cancer, we investigated the mechanism of p-ERK1/2-/HIF-1α-mediated regulation of ABCG₂ activity and the effects of ABCG₂ on gemcitabine response under hypoxic conditions in pancreatic cancer cells.

Our study also revealed that hypoxia-induced phosphorylation of ERK1/2 activates HIF-1α to regulateABCG₂ activity in pancreatic cancer cells. During hypoxia, there is increased cytoplasmic accumulation of HIF-1α, leading to its increased nuclear translocation. In the nucleus, the HIF-1αmolecules heterodimerize with HIF-1βsubunits to bind to specific hypoxia response elements in target genes and regulate transcription.²⁷ We found that hypoxia-induced p-ERK1/2 increases nuclear accumulation of HIF-1α, which in turn promotes ABCG₂ mRNA and protein expression. Activation of HIF-1α leads to increased translocation of ABCG₂ from the cytoplasm to the cell member under conditions of hypoxia. Moreover, gemcitabine is a cytidine analog, and the efflux of its deaminated metabolite, 2′,2′-difluorideoxyuridine (dFdU), by ABC transporters contributes to gemcitabine resistance.²⁸ We further investigated the effects of ABCG₂ on the gemcitabine response during hypoxia. Our results showed that hypoxia regulates the accumulation of ERK1/2/ HIF-1α and has an effect on ABCG₂-mediated modulation of the gemcitabine resistance in pancreatic cancer cells. Our results suggest that ABCG₂ and, in particular, ABCG₂ expressed in the cell membrane might serve as a special predictor of gemcitabine response in pancreatic cancer. Therefore, ABCG₂ could be an effective target for chemotherapeutic resistance in pancreatic cancer. Moreover, we hypothesize that ERK1/2 and HIF-1α inhibitors can increase the gemcitabine sensitization of pancreatic cancer when cell membrane-bound ABCG₂ is expressed at high levels. However, the mechanisms underlying ABCG₂ accumulation and cell membertrans location need further elucidation.

Some studies have reported that the HIF-1 heterodimer binds to a conserved HIF-binding sequence within the hypoxia-responsive element in the promoter or enhancer regions of target genes, thereby eliciting their transactivation and an adaptive hypoxic response.²⁹ We demonstrated that hypoxia enhances HIF-1α binding to a target gene sequence in the hypoxia-responsive elements of the ABCG₂ promoter, increasing its transcript reformation in pancreatic cancer cells. Therefore, we suggest that gene-targeted therapy might be an effective method for improving gemcitabine response in pancreatic cancer.

In conclusion, chemoresistance induced by hypoxia is due to the regulation of ABCG₂ through the actvation of ERK1/2/HIF-1α. ABCG₂ could serve as a predictor of gemcitabine response and potentially as a target for chemotherapy of pancreatic cancer. Inhibition of ERK1/2 and HIF-1α could increase the gemcitabine sensitization in pancreatic cancer with highly expressed ABCG₂ cell member protein.

Materials and methods

Patients and sample preparation

Fresh pancreatic adenocarcinoma and paracarcinoma tissue samples were obtained from patients who received radical pancreatectomy for pancreatic adenocarcinoma without radiation or chemotherapy before the operation. The diagnosis was confirmed according to the standard of American Joint Committee on Cancer (AJCC, 2010).Histological slides were reviewed by 2 experienced pathologists blinded to the clinical data. Each pancreatic cancer surgical specimen was cut into 1 cm³ pieces, quickly fixed in 100 g/L formaldehyde solution, then embedded in paraffin and sliced into 4-mm-thick slices for further studies.

Immunohistochemistry

A 3-step immunoperoxidase technique was used to determine the expression of HIF-1α,p-ERK/ERK and ABCG₂ with 1:100 diluted anti-HIF-1α (rabbit polyclonal antibody, Santa Cruz, USA), 1:100 diluted anti-ERK1/2 (rabbit antibody, Proteintech Group, USA), 1:500 diluted anti-phospho-ERK1/2(p-ERK)( rabbit monoclonal antibody, Cell Signaling Technology, USA) and 1:100 diluted anti-ABCG₂ (rabbit polyclonal antibody, Santa Cruz, USA). Sections were deparaffinized, hydrated, immersed in citrate buffer and autoclaved followed by incubation with rabbit serum for 30 min. Incubations with primary antibodies were carried out for 24 h at room temperature. Slides were washed with PBS and incubated with the corresponding secondary antibodies. Chromogen 3, 3′-diaminobenzidine tetrachloride (DAB) (Serva, Heidelberg, Germany) was used as a substrate. The cell nucleus was stained with Harri'shematoxylin solution. The expression level of HIF-1α and ABCG₂was scored separately according to the percentage of positive and intensity in pancreatic cancer tissues. The scores were defined as follows: <5% was judged negative, >5% was judged positive, 0(negative):<5%, 1(mild): 5% to 33%, 2(moderate): 33% to 67%, 3(intense):>67%. The scoring was determined blindly by 3 independent reviewers without any knowledge of clinical characteristics and pathological grade.²²

Cells and culture condition

The Capan-2 cell line (ATCC) was cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (BI, Australia) and 100 μg/ml of penicillin and 100 μg/ml of streptomycin. The cells were grown at 37°C in 5% CO₂. To achieve a chemical hypoxic condition, the cells were treated with cobalt chloride (CoCl₂). The CoCl₂has been used to substitute cobaltous ions for ferrous ions in heme, resulting in a conformational change in aheme protein O2 sensor.^23,24

RNA interference and cell transfection

The small interfering RNAs (siRNAs) targeting the human HIF-1α, ERK1/2 and ABCG₂genes were designed and synthesized. siRNA-HIF1α (si-HIF1α)(5′-UGU GAG UUC GCA UCU UGA UTT-3′) and its control sequence (5′-AUC AAG AUG CGA ACU CAC A TT-3′); siRNA-ABCG₂ (5′- GAA GAA GAU CAC AGU CUU CTT-3′) and its control sequence (5′-GAA GAC UGU GAU CUU CUU CTT-3′) were synthetized by Jima (Shanghai, China). siRNA targeting ERK1/2(si-ERK,No.6560) was purchased from Cell Signaling Technology (Cell Signaling Technology, USA). Capan-2 cells were transfected with the siRNAs and their controls by using INTERFER in transfection reagent (Polyplus, France). To optimize efficiency, the cells were transfected with siRNA twice at an interval of 48 hours. A total of 2×10⁵ cells were transfected with 110 pmol siRNA. The effects of the knockdown of ERK1/2, HIF-1α and ABCG₂ RNA were determined by western blotting method after 48 h of transfection. As non-silencing control, the silencer negative control siRNA was used.

The HA-HIF1α−wt-pcDNA3 and an empty HA vector plasmid (purchased from Addgene) were co-transformed into competent cells and plated on Ampicillin-LB agar plates overnight at 37°C. The HA- HIF1α−wt- pcDNA3 and the empty HA plasmid were transfected into Capan-2 cells individually using INTERFERin® transfection reagent (Polyplus, France).

Western blotting analysis

To extract proteins from pancreatic cancer and paracarcinoma tissues, the specimens were first cut into small pieces and lysed with 1 mL RIPA buffer (Sangon Biotech, Shanghai, China) per 100 mg tissue. After being homogenized manually with a disperser, the lysates were centrifuged at 16,000 g. Whole cell extracts from cell lines were prepared in RIPA, vortexed and centrifuged. In addition, the nuclear proteins were separately extracted using the Nucleoprotein Extraction Kit (Keygen, China). Proteins from supernatants were quantified by the Bradford method (Axygen, USA), and an equal volume of 1х loading buffer was added before boiling the samples at 100°C for 10 min. Total protein (20μg) mixed with loading buffer were separated by SDS-PAGE and transferred onto PVDF microporous membranes (Bio-Rad). Subsequently, the membranes were blocked with 5% nonfat milk for 1 h at room temperature and incubated overnight at 4°C with the following primary antibodies as required: anti-HIF-1α (1:1000, Santa Cruz, USA), anti-ERK (1:1000, Proteintech, USA), anti-p-ERK (1:1000, Cell Signaling Technology, USA), anti-ABCG₂ (1:1000, Santa Cruz, USA) and anti-β-actin (1:10000).The membranes were then incubated for 2 h with corresponding secondary antibodies at 1:10,000. The immune complexes were detected by chemiluminescence of HRP substrate (Millipore, USA). For quantification, the western blotting bands were quantified by Meta Morph software (MDS Analytical Technologies) after background subtraction.

Real time PCR detection

Total cellular mRNA was extracted with TRIzol (TianGen Biotech, China) according to the manufacturer's instructions and reverse-transcribed into first-strand cDNA using TransScript First-Strand cDNA Synthesis (TransGen, China). Real time PCR was performed with TransStart Green qPCR Super Mix (TransGen, China) and tested by the DA7600 Real-time Nucleic Acid Amplification Fluorescence Detection System (Bio-Rad) in 25μl reaction volumes. The human β−actin was used to normalize mRNA levels. The 2^−ΔΔCt method was used to calculate the relative mRNA fold changes. All experiments were carried out in 3 biological replicates. The primers used in the real-time PCR were synthesized by Sangon (China) are as follows: HIF-1α (F): GTACCCTAACTAGCCGAGGAAGAA, HIF-1α(R): GTGAATGTGGCCTGTGCAGT, ABCG₂ (F): GCTACACCACCTCCTTCTGTC

ABCG₂ (R): GCTGAAACACTGGTTGGTCG, β-actin (F): TGCGTGACAT TAAG GAGAAG, β-actin (R): GCTCGTAGCTCTTCTCCA.

Cell viability measurement

Capan-2 cells were plated in 96-well plates in triplicate at a density of 10³ per well and maintained in complete medium for 12 h before being subject to different treatments. Cells were counted using a Cell Counting Kit-8 (CCK-8, Bevotime, China) following the manufacturer's protocol. Cell viability was measured using ELISA, and optical density (OD) values were read at 450 nm. The cytotoxicity of gemcitabine (Gem, Eli Lilly, USA) was tested using CCK-8. Finally, the cell survival rate was calculated using the following formula: [(As−Ab)/(Ac−Ab)]×100%, where As represents the absorbance of experimental groups, Ab represents the absorbance of the blank sample containing complete medium without cells and Ac stands for the control group containing cells with no treatments. The concentration required for 50% inhibition of cell growth (IC50) of Capan-2 cells was calculated using a non-linear least squares curve fitting method.

Flow cytometry analysis

Capan-2 cells were plated into 6-well plates and cultured at 37°C for 24 h before being subject to different treatments. After the treatments, cells were collected into centrifuge tubes, washed with PBS 3 times and blocked with PBS containing 5% FBS for 1 h. At least 20,000 cells were counted in each test. The cells were then incubated in the dark for 1 h with PE mouse anti-human CD338 (BD, USA). The samples were analyzed by flow cytometry on the FL2-H channel (Flow Jo 7.6).

CHIP assay

A Chromatin Immunoprecipitation Kit (Millipore, USA) was used to detect endogenous binding of HIF-1α to the transporter of ABCG₂. Briefly, Capan-2 cells were cultured in a 150 mm culture dish containing 20 ml RPMI-1640 medium for 24 h before 400μMCoCl₂ was added and incubated for another 24 h. To crosslink the proteins to the DNA, cells were incubated with 1% formaldehyde at 37°C for 10 minutes, and 2 ml 10×Glycine was added for 5 min at room temperature to quench the crosslinking reaction. Cells were washed twice with 20 ml cold PBS, scraped with 2 ml PBS containing 10 ml of 1×Protease Inhibitor Cocktail and resuspended in 1 ml SDS Lysis Buffer containing 1×Protease Inhibitor Cocktail. The cell lysates were sonicated on wet ice with ultrasound sonicator to shear the DNA to a length between 200 bp and 1000 bp. The immunoprecipitation was then performed overnight at 4°C. The supernatants of all groups were supplemented with 8μl 5 M NaCl and incubated at 65°C for 5 h to reverse the DNA-protein crosslinks. After purifying the DNA from the chromatin proteins, PCR was performed using primers for the specific region containing the HREs (5′-GCGTG-3′) of the ABCG₂ promoter. The primers used were as follows: F: 5′-CGAGCAGCGCTTGTGACT-3′R: 5′-ACATCCA GGG GACGAGCTCA-3′.^25,26

Immunofluorescence staining

Small glass coverslips were placed into 24-well plates after they had been immersed first in 1% HCl overnight and, subsequently, absolute alcohol overnight. Capan-2 cells plated on glass coverslips were cultured to 10% confluence. The glass coverslips were then removed and washed with cold PBS 3 times. The cells were fixed with 4% formaldehyde in PBS buffer for 15 min at room temperature, washed with PBST 3 times, permeabilized with Triton X-100(0.25%) and blocked for 1 h with 10% goat serum containing 0.1% Triton X-100. Next, the cells were incubated with rabbit polyclonal anti-ABCG₂ overnight at 4°C in a humidified chamber. The cells were incubated with goat anti-rabbit IgG FITC (Santa cruse, USA) at room temperature in blocking solution for 1 hour. Finally, the cells were treated with DAPI (1:10,000)for 5 minutes, briefly washed with PBST, covered with anti-fade mounting medium (Vectashield, Loerrach, Germany) and placed onto microscope slides. The slides were examined under a laser scanning microscope (TCS-SP2-AOBS-MP, Leica Microsystems CMS).

Statistical analysis

Statistical analyses were performed with SPSS 16.1 software using 2-way analysis as shown in Fig. 1. Other numerical data are reported as the mean ± SEM. Two-tailed Student's t test was used to assess the difference in protein expression, mRNA levels and IC50 values. A p value< 0.05 considered statistically significant.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Funding

This study was funded by the National Natural Fund (31370861) and Tianjin Basic Research Plan Project (13J CZDJC31300). The authors declare that they have no conflict of interest. We thank Tianjin Medical University Cancer Institute and Hospital for providing human pancreatic adenocarcinoma samples and for helpful discussions during the course of our study.